Archived Content

The content on this page is provided for reference purposes only. This content has not been altered or updated since it was archived.

Vaccines, Blood & Biologics

December 11 Transcripts

National Institutes of Health/National Institute of Allergy and Infectious Diseases

Printable Version (PDF - 431 KB)

(Back to Workshop Main Page)

December 11, 2007

Summary of Day 1: Robert Couch and Jacqueline Katz

Session 3: Assays to evaluate vaccine immunogenicity Moderator: Maria Zambon, PhD

Limitations of current serologic assays to detect antibody responses to HA and NA: John Wood, PhD (NIBSC, United Kingdom)

Assays to evaluate cell mediated immunity: Guus Rimmelzwaan, PhD (Erasmus Medical Center, Netherlands)

Novel Assays

Pseudotyped viruses: Gary Nabel, MD, PhD (NIH)

NA inhibition assays: Maryna Eichelberger, PhD (FDA)

Phage display libraries: Hana Golding, PhD (FDA)

M2-specific assays: Walter Gerhard, MD (Wistar Institute)

Assays to evaluate T cell responses: Janet McElhaney, MD, University of British Columbia and University of Connecticut

Session 4. Evaluation of avian influenza vaccine efficacy
Moderator: Kanta Subbarao, MD, MPH - introduction and overview of animal models of influenza (mouse, ferret, monkey)

Evaluation of immunogenicity and efficacy of non- replicating avian influenza vaccines following challenge with homologous and heterologous H5N1 viruses in animal models: Jacqueline Katz, PhD (CDC)

Evaluation of immunogenicity and efficacy of live attenuated avian influenza vaccines efficacy following challenge with homologous and heterologous H5N1 viruses: Kanta Subbarao, MD, MPH (NIH)

Heterosubtypic immunity: Suzanne Epstein, PhD (FDA)

Challenge studies in humans: John Treanor, MD (University of Rochester)

Use of surrogate measures of efficacy: Jerry Weir, PhD (FDA)

Panel Discussion
Led by the moderators (Couch,Katz, Zambon, Subbarao) and panelists Bob Belshe, MD (St Louis University) Ed Kilbourne, MD (NY Medical College) Wendy Keitel, MD (Baylor College of Medicine) and Peter Wright, MD (Vanderbilt)



Agenda item: Summary of Day 1: Robert Couch and Jacqueline Katz

DR. COUCH: Dr. Katz and I are to make some remarks about yesterday's findings. I have a couple of remarks to make before I go ahead and one is that one of the problems with being a moderator is that if you don't take any notes you are in trouble when you sit down and want to make up summaries, and I had no notes from the discussion.

So, I don't know if we will have time here or not but we have a panel session this afternoon so that if I didn't comment on your comments and you want to do it you could get a chance now; otherwise bring it back this afternoon because all of these topics are going to come up here in the discussion period.

So, a lot was not referred to and that is almost self-evident because of the extent of the subject. So, we can't do everything, but what we did yesterday morning, the focus as you know was seasonal influenza and this is the summary that I put together with some help of my loyal secretary who is sitting on the second row back there. I told everybody yesterday I am almost computer illiterate, and this is true. She is not, fortunately.

Okay, seasonal influenza, the first two sessions as you know considered characteristics of immunity and the surrogates and correlates of that immunity.

So, to repeat that homotypic immunity is powerful and long lived. Too bad this thing, this virus has the ability to change. Otherwise we might not be sitting here today.

Heterotypic immunity varies with the extent of antigenic variation. I think that is definitely one of those truisms of influenza as I see it.

Both are highly correlated with the serum anti-HA antibodies to the infecting virus. That is established and has been established for decades and that correlate is highly unlikely to change at any time in the future.

For optimal maximal immunity the mediator must be present at the time of exposure. Dr. Murphy emphasized that and the reason for that, this is a rapidly progressive infection up to its peak for whatever it is going to do with the illness.

Activation from memory can ameliorate, maybe shorten that duration but cannot prevent and unless it is activated very rapidly cannot have a major effect even on the intensity of the initial peak. Those points were made and talked about the dynamics of the infection and that is certainly true for the standard febrile influenza.

Now, with regard to correlates and surrogates there is considerable redundancy in the immune protective modalities. I hope that came across clearly. I think all of us feel that way. Focusing on a single item our high correlate is there. That doesn't change but there are lots of redundancies in the immune system for control of this infection, and all are desirable. I have no hesitation in saying that. Some of them I think are more desirable than others but they are all desirable.

Actually Brian gave me a nice pecking order on which ones he considers the most desirable. Anti-hemagglutinin is the most powerful mediator of immunity to infection.

Anti-hemagglutinin can reduce the severity also. So, it is going to double the edge. It can prevent as well as reduce. So, it is highly desirable from that dual action in itself. Serum and antibody and secretions are necessary and certainly highly desirable for optimal immunity to influenza. That is where we are at this day in time.

Anti-neuraminidase can reduce the severity of the infection and can prevent infection if it is present in secretions and if it is present in high titer. Aramanto jumped me afterwards yesterday and I said, "Absolutely, antibody to neuraminidase can prevent infection." You don't have to have antibody to the hemagglutinin if they have that prevention and when it is present in high quantities, and while I don't have the direct proof, I am sure it does it when it is present in secretions in the right quantities because if you can inhibit that neuraminidase then you can at the time before it reaches that cells you have inhibited the infection entirely, and I think that was verified as true with the anti-neuraminidase drug studies. They prevent infection by inhibiting the neuraminidase, not just modify infection. So, that has a capability that should not be forgotten as well.

Anti-M2 can reduce the severity of infection in mice. That is very clear. Titers in humans are low. Actually I made a correction. Walter Gerhard corrected me that the mouse data, if anybody wrote it down was missing a decimal place in there so that I will make that correction on that slide. The titers in the mice were not as high as they were on my slide.

At any rate anti-M2 can reduce the severity of infection in mice. Titers in humans are low and proof of the value in humans is lacking, what I consider the major deficiency of anti-M2.

CTLs and the cytokines that are reduced, released as a result of antigen-specific stimulation can reduce the severity of infection and data are emerging for an effectiveness in seasonal influenza. So, that is beginning to move and we will cover that a little bit again in terms of the details of the data that Dr. Bennink presented to us as we know it in mice.

Now, that summary as you will recognize is a summary of what I said and what Dr. Murphy said. As I said, I don't want to ignore whoever made comments on the obvious but you are going to have to bring them back if you did on that subject.

So, if we move on to the other subjects that were considered the Cochrane Collaboration, is Tom Jefferson here? All right, Tom, I am going to fuss at you in a minute. I just wanted to be sure you were here so you could defend yourself.

Vaccine efficacy was assessed and for that they identified all field trials possible as far back as they could go and what have you, not volunteer studies, field trials of influenza vaccines and efficacy and they found, I had 338 in my notes. I thought you said, "Three hundred and thirty-one," but I don't think we will quibble over seven.

Only four of these, however, were randomized controlled trials with what the Cochrane Collaboration considered to be a very low bias. That is phenomenal. The examples that were presented were homotypic and heterotypic protection, very high efficacy with homotypic protection. Up to 93 percent in the military trial presented good protection versus heterotypic, about 53 percent. So, both were present and both were well documented and both fit with our concepts of immunity from infection and in this case immunity from vaccination-inducing those same modalities.

The major thing Tom did was make a strong plea for high-quality randomized controlled trials. He said that repeatedly that that is what is missing in order to nail anything down in terms of these concepts.

Now, I said that I was going to fuss at him a little bit and my problem with the Cochrane is that they really are rigorous in requiring high-quality trials. In the process of being rigorous for these high-quality trials they throw out huge amounts of data that can teach us an awful lot, and I wish if you do your Cochrane Collaboration and responsibility then go a step farther and say, "Okay, a little bit of bias. Now what do we see? Maybe a little more bias. What do we see?" and let us have a pecking order of those trials in terms of what is learned from them so that we have more data to look at rather than just a small number of trials that meet these high rigorous criteria which one can't complain about that part of it. It is just that data that we lose in the process of doing that.

At any rate I think it confirms what is available and what they accept as the criteria and the thinking we currently have on immunity.

Now, T cells, this is again pretty basic for me. T cells were presented by Dr. Bennink and he is a superb mouse T cell immunologist. Dissection of the components of a recognition and understanding of the responses to influenza antigens are expanding rapidly. Those of you who were at the Keystone meeting about a year and one-half or maybe 2 years ago, something like that the immunological session was phenomenal in terms of the data that is being developed in relation to influenza.

The role of CTLs and the T cells and mouse model is very clear and the potential significance in humans is clear. I have already indicated early demonstrations are emerging now in humans and the potential is still clear as well, but not documented at the present time. Immunodominance was emphasized by Dr. Bennink as a characteristic of the responses in mice and I am sure in man with what data is available and only a tiny fraction of the T cell epitopes induced responses and that characterizes this virus encounter.

We are going to hear from Dr. Rimmelzwaan. I don't know if you are going to talk about your mutation data but you see frequency significance, re-emergence of alternative, what goes on here and the back up of the redundancy in this system that would prevent us from being worried about the kinds of things that would happen there. That is for humans. That data is not done, not available as yet.

He alluded to the major determinants of responses and they included binding affinity, the T cell repertoire, the processing and some others. The list must be fairly long and my notes didn't have a whole lot more in them other than those that I have listed here, but there are multiple determinants in terms of stimulating the CTL response.

Internal proteins can immunize and can protect, very clearly documented. You see the NP in the mouse is the big one, M1 perhaps in humans, maybe others as well.

The unknowns that he emphasized were features or requirements for optimal benefit have not been demonstrated conclusively and one of those that he identified is one that we have carried in our thinking for some time is that the importance of the precursor cell frequency is not known, and you see actually I just keep giving you my thinking. I can't help that, but sometime ago we tried very hard when we got the NT protein demonstrated in the mice to induce CTL, CHCTLs that protected to get that vaccine, that antigen, that vaccine into humans. We never could get a GMP or never got it done but my thinking has not changed over what it was at that time and that was that if we are going to get a maximum benefit, an improved benefit since we all start with this except maybe those varying quantities in the elderly we have got to start, we have got to change that precursor cell frequency. We have got to start it at a higher level when you are exposed so we can bring it on in quantities to have a greater effect at a center time period. That is my thinking on CTLs with regard to humans and as he said it is not even proven to be of value in mice which did surprise me a little bit, and finally is Dr. Greenberg here? All right, I can't ask him a question unless somebody else knows because I wasn't sure about one of these items. Studies of human cells, Harry emphasized that he was studying human cells here in these responses and these represent human studies. So, they obviously are pretty significant. Live versus inactivated vaccine in infants, children and adults, he didn't give any of the infant data. I don't know much about it but it is only inactivated vaccine. The infants were not given live vaccine.

T cell responses are being studied by analyses for surrogates, interferon gamma certainly and he said porphyrin(?) was the other one and B cell responses for antibody-secreting cells, that is the data that he deals with, and not surprisingly he emphasized that there are a lot of differences by age and vaccine and he only emphasized a few of them. He had a couple of slides there with lines going in every direction that he did too quickly. At any rate this is the one that I had in my notes and I have a little uncertainly about. CD8s and I presume CD8s with markers for CTLs increased in children after live but not in adults. If anybody else knows whether that is correct or not please correct me, but if that is true he emphasized a lot of, several times you know that is sort of what you would expect. This is a greater level of replication in children, more infection, more antigen, more response makes sense, whereas in the adults this thing, this vaccine has trouble doing very much.

Antibody-secreting cells are up at day 9, very clearly demonstrable demonstrated early and then down back to the baseline by 28 in the peripheral blood. He made a note of something I had not heard before or remembered from the reading. CD4 level and I assume this is an antigen specific CD4 level response was the best predictor for an antibody response. Maybe that does make sense, too, that the more helper capability and the more recognition you have got to have the greater the response, and he gave some expected findings as he emphasized and some curious findings and I said, "What are the" -- well, he said that there are fewer changes in adults. It was an expected finding he thought. TIV inactivated vaccine was better at expanding B cell memory cells and certainly in adults which are the data he emphasized and that is perhaps not surprising either in terms of the level of infection in those two.

He said that there was no increase in the second dose after the second dose of TIV in unprimed children, and that I found a little surprising that he couldn't recognize any change in the B cells after that second dose because we think very strongly in terms of the value of that second dose in unprimed individuals and he said that the pre-antibody one year after vaccination in adults was lower in those who had received live than in those who had never had vaccine at all, a curious finding that if it is fortuitous why then that is one thing; if it has an immunological explanation it is important.

That is the summary I had from the notes. I guess we had better move on since we are already getting along and any of your comments and things that I missed they can come up in the panel discussion this afternoon.


DR. KATZ: Okay, I am going to briefly summarize Session 2 which addressed immune responses to avian influenza infection and vaccines in humans and unlike Bob I was fortunate enough to have somebody in the audience taking notes and so I need to thank and acknowledge Catherine Luke from NIAID who really helped me put this together at the last minute.

First of all we heard from Nancy Cox about the extent of genetic and antigenic heterogeneity of H5N1 influenza viruses in birds, and using a unified classification WHO has now classified the viruses into nine different genetic clades that have circulated in the past 3 years and the tenth clade is the viruses that originated in 1996, and this heterogeneity is thought to be both due to the geographic distribution, the heterogeneity due to geographic differences but probably also because the viruses are moving from different bird species from domestic land-based poultry, waterfowl and in and out of the wild bird population and definitely the human isolates that have been isolated from the over 300 cases, human cases reflect the heterogeneity in birds and currently while clade 1 viruses predominated in the 2004 and early 2005, we are now seeing a predominance of clade 2 viruses causing disease in humans and less or very little clade 1 virus activity in humans although it is still out there in places like Vietnam.

So, the heterogeneity that is seen has important implications for preparation of pre-pandemic vaccine candidates and reverse genetics six two PR8 reassortant vaccine candidates have been made from each clade and subclade and as genetic heterogeneity expands the WHO is trying to keep up with potential vaccine candidates.

There is also heterogeneity in the sensitivity to antivirals particularly the amantadines and some of the clade 1 and some clade 2 viruses resistant to these drugs.

Finally, we heard about new methods such as the glycan array technology that can be used to look at variation in receptor binding specificity and that this can be used for a pandemic risk assessment to see if these viruses are acquiring amino acid substitutions in and around the receptor binding site that might promote binding to human-like receptors.

Then I talked about the serological response to avian influenza viruses both in individuals and culture confirmed or PCR-based confirmation of infection as well as in the seroprevalence studies we have conducted in different groups of poultry workers and demonstrated that in the majority of the infected individuals we do indeed if we are able to, to obtain serum at, on and around or after 14 days post symptom onset. We do see serum neutralizing antibody titers to H5N1 viruses and this was demonstrated with the 1997 viruses and we have also done it with the clade 2.1 viruses from Indonesia and if we take this criteria of a titer of 1:80 then in our studies in Hong Kong we found that this level of antibody was also associated significantly associated with the more intense exposure in poultry workers.

Overall, however, the available evidence suggested a seroprevalence for anti-H5 antibody is still relatively low in the populations that have been looked at so far.

However, in some cases we have demonstrated that perhaps mild or asymptomatic infections are eliciting antibody that can be detected by the neutralizing antibody assays but potentially they are of short-lived duration and that has consequences when we are looking for evidence of infection or mild or asymptomatic disease.

Then finally I talked a little bit about the different studies where it had been more difficult in general to detect serum-neutralizing antibody to some of the H7 viruses although we did readily detect them to low pathogenicity North American H7N2 viruses and the questions that arise there that we might come back to as we talk about different assays is is there a limitation in current assays to depict antibody to all H7 viruses

So, three of the talks we had were talking about the immunogenicity of non-replicating vaccines and the clinical trials that have occurred in different parts of the world.

So, I put all of those; so that was Maria Zambon, David Cho and Laszlo. I am having trouble with it again. So, they talked about different clinical studies that occurred in Europe, the US and some that are occurring in China and elsewhere and so in general companies and institutes have, over 10 companies and institutes are producing and evaluating H5N1 vaccines and these fall into inactivated whole virus, split or subunit. Different sorts of adjuvants are being applied including alum MF59 and the GSK adjuvants. People are looking at egg. There are some cell-based studies, the Baxter whole virus vaccine was produced in vero cells and so far most studies have been doing a dose escalation trial where they are looking at low doses and dose sparing with adjuvants.

So far all of the vaccines appear to be safe and well tolerated and in general the immunogenicity is assessed using different functional assays and neutralization assay and the HI assay and as Maria Zambon pointed out we are really only focusing on those functional assays and so we might be missing out on non-neutralizing IgG or other components of the serum antibody response as well as I mean nobody talked about T cell assays although I do understand there are some studies on going with the different vaccines. So, that is something we can perhaps pick up later in the discussion this afternoon

Generally two doses are needed for optimal responses and high doses are needed definitely in the absence of the adjuvant, for example, the US Sanofi-based vaccine which has now been licensed at two times the 19 microgram dose. It was noted from David Cho who described studies where they compared intradermal versus intramuscular immunization that there was no obvious enhanced immunogenicity although this was still at relatively low doses, a 3-and-a-9-microgram dose intradermally and apparently higher dose studies are ongoing. In most of the studies where alum was looked at alum adjuvant had a modest effect which in some studies was age dependent and then in other studies that didn't have a non-adjuvanted arm it was difficult to assess the effects but it was generally felt that alum was a less robust adjuvant compared with some of the more powerful adjuvants such as MF59 or AS and that these adjuvants demonstrated dose sparing at quite low doses.

In general the adjuvanted vaccines induced broader antibody responses to variant clade 2 H5N1 viruses and further boosting in some cases where studies came back and boosted after 16 months or in some cases revaccinated individuals that had been vaccinated with an earlier H5 vaccine 7 or 8 years earlier both of those studies showed a significant boosting effect and enhancement of the magnitude and the breadth of the response and again as Maria pointed out it seems that the higher the immune response to the homologous virus the more likely you are to get some heterologous cross reactive antibody to variant H5N1 viruses.

In some studies age-related pre-existing immunity was detected. In a group, a vaccine study in the elderly it was noted I think it was about 16 percent of the elderly in that vaccine had, in that study had pre-vaccination titers against H5 and in the H9 study conducted in the UK there was a substantial pre-existing immunity to H9 in individuals that were over 35 years of age or so, and clearly the message came through that we need to improve and standardize serologic assays used to evaluate these pandemic vaccine candidates and that we need to have a harmonization and standardization of these assays in different laboratories.

So, Ruth Karron then talked about these live attenuated influenza vaccine studies that she is conducting in collaboration with Kanta Subbarao and three live attenuated vaccines that have been generated on the A/Ann Arbor cold adapted backbone have been developed to H5, H9 and H7 and Ruth presented data on H5 and H9 vaccines that restricted replication in seronegative adults and there was some question there in the discussion as to why that is and Kanta's feeling was that it wasn't just due to receptor binding because H9 and one of the H5N1 vaccine candidates were derived from viruses that had somewhat altered at least in the case of H9 had stronger receptor binding preference for human receptors.

So, there was some discussion about how to really evaluate candidates for their level of attenuation in humans before they go into the clinical trials. So, the H5N1 vaccines were fairly poorly immunogenic with only about 10 percent of subjects seroconverting. However, the H9 vaccine exhibited greater immunogenicity and they saw seroconversion in roughly 80 to 90 percent of subjects depending on the assay used and the proposal was made that additional evaluation of some of these vaccines may be able to proceed in outpatient settings because of the apparent attenuation of the vaccines.

So, finally Fred Hayden gave us an overview of the WHO clinical research initiative which is based on international collaborative efforts that have established the Southeast Asia Influenza Clinical Research Network. The idea is that multiple clinical protocol-based studies will be conducted to improve the understanding of the pathogenesis, immunology, diagnostics and therapeutics for human H5N1 infection and in some cases this will also include severe seasonal influenza infection and so studies that are either ongoing or about to begin include evaluation of the high dose and parenteral administration of the NAI antivirals and he finished up by indicating that new concepts and proposals for studies were welcomed and should be initiated through interaction with the study site PI.

So, I am going to leave it there, and I guess we are going to move on.


Agenda item: Session 3: Assays to evaluate vaccine immunogenicity -Moderator: Maria Zambon, PhD

DR. ZAMBON: Thank you, Jackie and Bob for an excellent summary of yesterday's proceedings. It is my pleasure today to take you through Session 3 where rather than summarizing what is going on and providing if you like overviews of where we are we are starting to look forward to trying to address some of the problems that we have all recognized in terms of specific techniques used for vaccine evaluation and to start us off with that we have Dr. John Wood who is going to talk about the serological assays and hopefully come up with some suggestions for how we might be going forward to achieve that harmonization that we have all been talking about and that will shortly be followed by Guus Rimmelzwaan talking I hope similarly on the cell-mediated immunity side. We then have a number of shorter presentations focusing on specific assays.

Without further ado, John, over to you.


Agenda Item: Limitations of current serologic assays to detect antibody responses to HA and NA

DR. WOOD: Thanks, Maria. Good morning, everyone. I have two words in my title I would just like to pick up. One is limitations. I think that is a very negative part of the talk. I would like to then give a more balanced view rather than just dwell upon the negative aspects of these assays, and secondly, the word "current." I am going to focus on current assays. The novel assays, new developments will follow later this morning.

The assays I am going to talk about are assays to measure antibody to hemagglutinin the hemagglutination inhibition assay, virus neutralization or microneutralization test, single radial hemolysis and then assays to measure neuraminidase antibody.

So, let us start with the HI assay. The advantages I think they are all well known to us, they are technically simple,easy to automate and we have a lot of experience in evaluating antibody responses to infection, vaccination using it for antigenic analysis helping WHO to select vaccine strains, etc.

The correlates of immunity are well documented for season flu and there is on the whole a good correlation with antibody from virus neutralization tests.

The limitations, a summary of those, the HI assay is relatively insensitive to detect antibody to flu B viruses and also as we know for H5s and H7s.

The technical aspects of the test that is the source of the erythrocytes be they avian or mammalian can affect HI titers and also measures to reduce non-specific inhibition; use of RDE can affect HI titers and there is poor reproducibility between the labs.

Now, in the next few slides I am going to focus on some of these aspects, the correlates of immunity, good correlation with virus neutralization, insensitivity and poor reproducibility.

So, correlates of immunity, Yon De Jong did an excellent job in 2003 in reviewing a lot of studies trying to correlate HI antibody titers with protection, studies in healthy adults, in children and involving natural infection and challenge infection with H2N2, with H3N2, with H1N1 and B viruses and overall the median HI titer of 1:28 was associated with 50 percent protection in these individuals, in these populations rather.

If you have more antibody there was greater protection. So, an HI of 1:192 was associated with 90 percent protection. So, De Jong concluded that the dogma of HI 1:40 is justified on these data.

There was also good correlation with virus neutralization titers. In vaccine studies the caveats are that in vaccine studies it is only a good correlation when the strains in the vaccine and the assay are homologous and the virus neutralizing antibody assay tends to be more strain specific than the HI assay. That is information from YonDe Jong, Stevenson and from Maria.

So, the limitations in sensitivity, the insensitivity of B antibodies being partly solved by splitting the antigen but by doing that you decrease the strain specificity which is a slight problem. For H5 and H7 we obviously have the horse erythrocyte HI assay and I will say more about that later.

In terms of poor reproducibility there have been a number of collaborative studies and I have summarized three of them here showing the poor reproducibility when you compare HI titers from lab to lab.

One study I coordinated in 1994, showed a full variation of titer of about 32. A European study done in 2005 showed a greater than 16-fold variation in HI titers between labs.

This actually affected compliance with the European regulatory parameters which is a big issue, a big regulatory issue and then a study coordinated by Ian Stephenson found between 6-and-128-fold variation in HI titers between labs.

So, let us look at the horse HI test. Jackie described this test very well yesterday. It is sensitive for antibody to H5 and H7 hemagglutinins and there is a good correlation with the virus neutralization test. This is data from Jackie with confirmed H5N1 cases and from vaccine trials and the published studies, John Treanor and Bresson et al showing good correlation in general. A big advantage is that you can use inactivated antigen. You don't need high containment for this test to look at H5 antibody.

The limitations, we are not really sure whether we can make the jump from the conventional turkey HI test to the horse HI test and assume a 1:40 titer is going to correlate with 50 percent protection. That needs some more thought, and as Jackie explained yesterday different H5N1 viruses have different affinities for erythrocytes. So, the agglutination can be affected by amino acid changes around the receptor binding site of HA.

So, it is important with different H5 viruses to evaluate specificity and sensitivity. The test may not be as robust as the conventional HI test. The source and the age of the horse erythrocytes do have a bearing on the test and we don't know anything about the reproducibility.

So, let us look at the data that we have to try to correlate HI with virus neutralization for H5 and Jackie showed this slide yesterday showing a very good correlation coefficient of about 0.9 between microneutralizing antibody titers at the bottom and HI titers along the vertical axis.

This is for evaluating H5N1 clinical trials and then another slide from Jackie comparing the specificity of the two tests for detecting antibody for H5, different clades, clade 1 and clade 2 viruses in normal populations. These were US populations looking for the presence of antibody in these normal populations and you see for clade 2 viruses there was 100 percent agreement between the two assays. There were no seropositives detected, but for clade 1 viruses the horse HI assay did detect some positives. So, it indicated a specificity difference between the two assays.

Now, let us look at virus neutralization. The advantage is it is a functional assay. It is suitable for semi-automation as a microneutralization assay. There are equivalent sensitivities to other HI antibody assays for seasonal viruses such as the HI and the SRH test. The virus neutralization assay tends to be more strain specific than HI for seasonal and H5N1 viruses and this is information from De Jong and from Ian Stephnson and from Maria, and there tends to be more or less equivalent sensitivity between neutralizing antibody assays for H5 and assays for HI using horse erythrocytes and the SRH test which I haven't discussed yet.

The limitations, the correlates of immunity are unknown although as Jackie and Maria explained yesterday virus neutralizing antibody titers are between 1:20 and 1:80 depending on which lab is doing the test, can be used to indicate seropositivity for H5N1. Another disadvantage is the need for live virus. So, if you are using highly pathogenic virus you need at least BSL3 plus. If you are using the reverse genetic virus you need 2 plus

Another aspect is that technical features of the test can affect neutralizing antibody titers such as virus growth kinetics. Maria explained yesterday that the kinetics of growth between a highly pathogenic virus and a reverse genetics virus are different and this can affect the neutralizing antibody titers.

Also, differences in protocols can affect the readout, serum treatments, dilution and antivirus added, the time for neutralization, the diluent used and it is really not surprising when you consider all these different technical details that there is poor reproducibility between labs.

I am going to say more about that talking about a study that Ian Stephenson has recently published on a collaborative study which involved Jackie, Ian and myself with the statistician, Rose Das, looking at the comparability, the reproducibility of HI on virus neutralization for measurement of antibody to H3N2 viruses

This is evaluation in 11 laboratories from eight countries and a panel of 19 sera were evaluated in these various labs.

This is a slide showing the reproducibility within a laboratory looking at two of the serum samples R and U and these are replicate serum samples. So, you expect the results of R and U to be the same. Yet in this graph we show the percentage for each laboratory along the bottom the percentage results where they are more than twofold different between R and U, and you see some labs, they get 60 percent of their assays show a greater than twofold difference between R and U and overall for the virus neutralization assay 22 percent of the assays showed such a high degree of variability just lab, within laboratory variation test to test whereas the comparable variability for an HI assay was much less, only 7 percent.

Looking at the absolute titers, the range of titers so we have the labs down the left hand side for the HI test, the variation ranged from a minimum in one lab of less than 10 to 640 and in another lab 10 to 40. So, there was quite a big difference in the range of titers, but it is even more when you look at the virus neutralization assay, from 10 to 160 in one lab, 53 to 81920 in another lab. It really is a huge difference and to summarize the variability of the neutralizing antibody responses if you look at just one serum, serum N the full difference in antibody evaluation for the HI test was 128, for the virus neutralization assay 724-fold difference in neutralizing antibody titers.

If you take the results overall and the median geometric coefficient of variation for the HI it was between 138 to 261 percent, for the neut assay 256 to 323, but what was striking is if you take one of those sera as a standard serum and evaluate all responses in relationship to the standard serum then the variability reduces enormously.

So, in the neut assay the variability goes from this to 85 to 115. It is still quite a lot but a significant reduction by the use of a standard serum.

Single radial hemolysis, this is not as widely used as HI or virus neutralization but it has some advantages. It has equivalent sensitivity to HI for seasonal viruses and greater sensitivity for B viruses. An SRH titer or zone area rather of 25 millimeter squares has been shown to correlate with 50 percent production. So, that is more or less the equivalent of an HI 1:40 and Yon De Jong has summarized this in his review in 2003.

Then for H5 assays SRH has been shown to be more sensitive than the turkey HI which is not too surprising and equivalent sensitivity to the virus neutralization assay and for measurement of seasonal antibody it is actually more reproducible than the HI assay. In the study that I coordinated in 1994 where the HI was 32-fold variable between the labs the SRH was only 3.8-fold variable between labs.

So, it has some potential in reproducibility and this is just a composite of data from my lab looking at antibody to cases of H5N1 from Hong Kong. I measured the SRH antibody. Jackie Katz measured the microneutralizing antibody and you see there was a very good correlation here and this was also echoed in the studies that Maria and I were involved with that Ian Stephenson published looking at antibodies stimulated by the H5N3 vaccine where you have the microneutralizing antibody responses here and the SRH antibody responses here, fairly good agreement between the kind of magnitude of antibody detected by each of the assays.

So, the limitations of SRH are it can detect antibody to virus internal proteins and you have to try and have measures to control that. We aren't sure about correlative immunity to H5N1 and technical details of the test can affect the clarity of the zones and that is the readout. So, that is quite important, the source of complement, the type of erythrocyte, whether the virus is allantoic fluid or cell culture fluid or whether it is more concentrated, that all affects the clarity of the zone. So, it can be quite difficult to read sometimes and the assays we found, I put this last bullet because assays are more demanding for clade 1, H5N1 viruses because in my lab we try to reproduce what we found for the 97 H5 viruses where we had a very good assay for H5 antibody. We couldn't do that for clade 1 viruses. So, the protocol we used just did not work. It was just general lysis throughout the plates. Yet, a colleague of mine, Imanual Monenoly at University of Siena has an SRH protocol that does work very well for this virus. So, there are some lessons to be learned then and this slide just illustrates some of the potential problems with SRH. It is looking at antibody from Hong Kong 97 cases. You see the antibody is shown very well in the SRH test and also a ferret antibody to Hong Kong 97 but also there is antibody in this rabbit serum to chick Germany virus which is an H10N7 virus. Presumably this is antibody to internal proteins because if you adsorb that serum with another influenza A virus that antibody disappears. So, this is something that you have to control.

Now, assays to measure antibody to neuraminidase, the classical one is the neuraminidase enzyme inhibition test developed by Michel Aymard back in 1973. It does allow measurement of antibody to neuraminidase and we all know that antibody to NI antibody has been associated with protection in mice. The vaccine stimulates NI antibody in animals and in humans and it clearly has a role in protection. The NA content of vaccine is not standardized and we rarely measure antibody to neuraminidase stimulated by vaccines and this is probably because of the limitations of the assay. It is laborious. It uses toxic chemicals. It is not suitable for automation.

The enzyme activity of neuraminidase is labile. This affects vaccines. It also could affect the assays, setting up the assay.

It is also not sufficiently sensitive to detect antibody in general stimulated by vaccine. Michel Aymard has said that poor levels of enzyme activity in cell culture viruses does affect suitability of cell culture viruses in the NI assay and finally Ed Kilbourne has published this antibody to hemagglutinin can sterically interfere with detection of antibody to neuraminidase. So, it is advisable to use reassortants with the relevant hemagglutinin.

So, it is really not surprising that alternative tests were developed based on ELISA techniques and two essential approaches have been followed, one to use partially purified neuraminidase and one to use a capture, a monoclonal antibody. The first of the techniques was developed in the USA and the second technique was developed in France.

So, the advantages are that they are technically easier than NI and you can automate the tests. They are more sensitive than NI so that when you look at post-vaccination sera where you have low levels of NI you have equivalent levels of ELISA-based antibody to neuraminidase, equivalent levels to antibody to hemagglutinin.

So, with a bit of effort these assays could be developed to measure antigen content in vaccines which would help standardize vaccines.

The limitations, the French assay does rely on monoclonals. So, it does limit its use for new variants and I think there was only an assay developed for N2 antibody and for the antigen-based assay the specificity of the assay does depend on the presence of relatively pure neuraminidase because internal proteins can affect the assay. We are unsure about reproducibility between laboratories; we are unsure about correlates of immunity.

So, my final slides are really to pull these things together, the key limitations and the action needed and I think it is clear that we need to standardize the assays to measure antibody to H5 hemagglutinin because of the variability in the HI, the neutralizing antibody assay and SRH to a lesser extent to compare the assays to look at sensitivity and specificity and with this in mind the WHO has now started a collaborative study to evaluate H5N1 serological techniques and to try to establish an international antibody standard for H5N1 antibody and so where we are with this is that we have prepared a freeze-dried candidate standard made from plasma taken from people who have had H5N1 vaccine and we have obtained test sera. They have been filled and coded. We have two sheep antibody, one from NIBSC and one from a US sheep antibody to see whether there is merit in having an animal serum as a standard in the assays and the viruses that we are testing are the clade 1 virus, the Vietnam 1194 virus, reverse genetics and two clade 2 viruses, one from Turkey-Turkey and one from Anhui. This set of reagents was distributed at the end of November and I believe now they have all reached their ultimate destinations. So, we can start the assays fairly soon and we hope to have the results coming back early next year.

There is a team of investigators from the UK, Ian Stephenson, myself and Maria Zambon, from the USA Jackie Katz, Jerry Weir, and Roland Lewondowski and so far we have 17 or possibly 18 participants taking part in this study, and to continue with the limitations in action we need really to, if we are serious about looking for antibody to neuraminidase we need to have some comparative evaluation of the different techniques to assay antibody to NA, those novel and existing assays to evaluate sensitivity, specificity and reproducibility to try to standardize these assays and hopefully to try to standardize the NA content of vaccines as well, and so all of the assays we need to look at correlates of immunity especially for virus neutralizing antibodies and this is really why we are here yesterday and today and we need to prepare for the unexpected.

As Maria and Jackie said yesterday we can detect antibody to H5 and H9 viruses when we least expect it. So,we have to be prepared for that to have adequate controls, back-up assays to try and find out what is happening and to try to investigate the sources of these unexpected reactions

So, to acknowledge the help from my colleagues, Ian Stevenson, Maria Zambon, Jackie Katz and then from NIDSC Rose Das , Diane Major and Bob Newman who have all helped to put together this talk.

Thank you very much.


DR. ZAMBON: Thank you, John. Are there any questions for John?

DR. COUCH: Our laboratory has been doing the hemagglutination inhibition and neutralization test for over 40 years and we are experts on the variability that you can identify with doing these between tests, between antigen, between people and there is no substitute for just a constant overview of quality control within a laboratory. You do high quality control within a laboratory to establish the best precision you can get and then compare your results to another laboratory and they are different and you see we learned this way back years ago in sharing the sera with CDC back in the Alan Kendell and Maury Harman days and we never got the same results.

We have recently relearned it again by doing exactly the same assay, sharing reagents and so forth with Sanofi and we tend to run about four-fold lower than they do on exactly the same sera.

So, it is really difficult to make these the same and the one thing that you can control a lot of this but you can't control the technicians and we have done that internally you see. You have two technicians do the test. We get a different result within the same laboratory on the same day and the two ends of the bench. So you see the number of variables that are built into these things are so difficult.

DR. WOOD: We see this in the HI test every year for the WHO serum selection process. Some labs always read high. Some read low and some are intermediate all the time.

DR. COUCH: I don't mean to say that to discourage but the important thing is that everybody try and do a well-standardized assay and with some knowledge and built in positive and negative controls that you know what you are learning biologically does indeed constitute a biological result. I decided long ago that is all I will ask for, and we have done four different kinds of neutralization tests over the years. So, I can tell you about each one of those and sensitivity and variation and that sort of thing.

In fact, the one we do now is probably the least sensitive and the second thing is you didn't even bring up that working with nasal secretions you see you heard this emphasis that Brian and I both feel about secretion antibody yesterday. It is almost an art rather than a science and I wouldn't recommend anybody start doing this unless they want to devote a lot of time to how you process these things, work with it and the variation is greater than it is in serum. That is just a comment.

You made one comment about neuraminidase. I can only give you one example. When Ed Kilbourne made the purified neuraminidase vaccine we developed an enzyme immunoassay with that purified NA and he ran the standard neuraminidase assay in that vaccine trial and we found the enzyme immunoassay slightly less sensitive than his neuraminidase inhibition assay in his same laboratory.

So, there has got to be some variability in that one as well.

The final thing I want to say and then I will quit and I sound like Ed Kilbourne by saying this one, that we really do need to know what is going on with neuraminidase and the vaccines. You indicated that on your slide. What is there; what should be there to ensure that response because you heard it from both Brian and me and I am sure other people agree that is an important antibody and we are not ensuring that we get it in adequate quantities currently in our vaccines.

DR. WOOD: I agree.

DR. COUCH: I told you no questions. I just wanted to give a lecture.

PARTICIPANT: John, thank you very much for that presentation. It goes to the point of much of the work that I am trying to do at WHO.

One question that I have and I have seen variants of these studies is that even though all these, there is a collaboration between labs usually the sample size that is used to compare these results from laboratory to laboratory is in my opinion fairly small.

So, for example, in some of those studies including 19 sera my question is yes, there is variability but for example if a much larger sera set was to be analyzed, for example, 500 samples I know this may not be practical or whatever but would you expect that some of that variation would be taken care of by increasing the number of samples that are analyzed?

DR. WOOD: Are you saying that the collaborative studies are flawed because there are too few serum samples in there?

PARTICIPANT: Not exactly but if I was to do that study I would include a much larger sample size.

DR. WOOD: I agree. More is better in these but we have to strike a balance and we have to encourage participants to take part in this study. So, we have to be practical.

PARTICIPANT: I agree. I do understand that. For example how was that sample size or this sample size determined statistically? Is that a sample size that is --

DR. WOOD: What we did was to try and get a collection of sera from vaccine trials that had received different types of vaccine that had had post-vaccination levels high, medium and low, so, a kind of cross section of the type of antibody that we will eventually be testing.

PARTICIPANT: The other question is how is the hemolysis measured, visually or by densitromy or how is that done?

DR. WOOD: When I first started doing this test you had a little IC(?) with a micrometer scale on there. That was the easy way of doing it but you can't do that for a long time. Now, you can automate this. You can have scanners.

DR. ZAMBON: Just a comment from the chair here. I have been reminded to ask people when they speak if they could introduce themselves and their institution, please?

DR. MURPHY: Brian Murphy, NIH. I just want to make a couple of comments. In your list of limitations of the HAI assay you should really indicate that it is not useful for measuring antibodies in secretions, okay? It is very difficult to use that assay because of all the inhibitors that are present. Now, that you know 3 or 4 years ago or 5 years ago that might not have been an important comment but now that there are licensed vaccines especially for seasonal influenza in which that so inefficiently stimulates serum antibodies reliance on HAI for correlates of immunity is going to be problematic in the long run and that really needs to be considered.

Another limitation of the HAI assay and the virus neutralizing assay is you can't do subclass things. So, you get only limited amounts of information. It is good because you get the sum of active antibodies but you don't, you can't dissect it and further interpret it.

DR. COX: John, I already know the answer to this question but I would like you to give it to the audience. There were some labs that really did perform quite well in terms if reproducibility and what would you think the factors would be that differentiated those labs which performed well versus those labs that performed less well?

DR. WOOD: Jackie did an evaluation of the type of technique that was used compared with the reproducibility and agreement between the labs to see whether there was some correlation between technique and agreement with neutralizing antibody titers and really there was nothing, no single thing that actually correlated with agreement but there was a trend and that was for the assays that used the microneutralizing antibody assay they were in better agreement than the ones which had more old-fashioned techniques. Is that what you are getting at? Maybe Jackie should --

PARTICIPANT: I think that one of the things that was told to me was that the labs that performed the tests on a routine basis and actually have vast experience in doing the techniques and so they were already doing them routinely there was much less variation than for those labs that don't have as much experience.

DR. WOOD: Yes, and in the H5N1 collaborative study we are circulating protocols for labs that are less experienced so that at least there could be a standardized protocol for all of the participants.

DR. ZAMBON: John, I have one last question. You have indicated the potential value of using standard sera to try to anchor the results and clearly the WHO I think is doing something about this, the intention to make standard serum available.

Now, given that we are looking at vaccines which are varying in their constrained composition and as we see H5 diversifying if you will do you think that there is going to need to be a commitment to making a new international standard on a regular basis to try to capture something that is relevant to circulating strains as time progresses?

DR. WOOD: Yes, that is a worrying question and I think there may well have to be a constant evaluation of the cross reactivity of any antibody standard that we establish to see whether it looks at utility for different H5 variants.

That is really why we have the sheep sera in there because you can obviously update the sheep antibody much quicker than you can a human antibody pool.

So, maybe that is the way to go for the future.

DR. ZAMBON: Thank you very much.

I would like to ask our next speaker, Guus Rimmelzwaan to update us on the assays to evaluate cell-mediated immunity and given what we have heard about assays that we have been using for 40 years I will be interested to hear what Guus has to say about any possible attempt to standardize or harmonize results between laboratories on cell-mediated immunity.


Agenda item: Assays to evaluate cell mediated immunity

DR. RIMMELZWAAN: That is a pretty ambitious statement.

Good morning, everyone. Thank you, Maria. So, this slide shows all the arms of the immune system that are at the site of the virus infection. On the left are the innate immune responses. I am not going to talk about that at all although it is very important. For the induction of immunological memory induced by infection or vaccination we have to focus on responses and there are a number of different cells that contribute to the immune responses.

Yesterday we heard about the B cells. It is very important. They secrete antibodies. They can neutralize the virus. There are also T cells, T helper cells that regulate and control B cell responses and also cytotoxic T cell responses or the CD8 positive T cell responses.

I am mainly going to focus on those cells because they are considered important markers for more broadly protective vaccines since they recognize conserved proteins of the virus.

So, what is the function of these cytotoxic T lymphocytes, CTLs? It is actually quite simple. Their main function is to eliminate virus-infected cells by lysis and the induction of apoptosis through release of foreign granzym and FasL-expression and this way by eliminating effector cells the further spread of virus is limited.

And essential function of T cells is that they can release cytokines like interferon gamma and TS alpha. So, it is important to note that the recognition of viral epitopes or antigenic determinants as Jack Bennink pointed out yesterday is restricted by MHC plus one molecules and I make this point because the human population is the outbred population and there is huge variation in the HLA makeup of all these individuals. HLA is highly polymorphic and it is complicating matters.

So, we have heard that CTL can contribute to protective immunity in the mouse model. That was made clear yesterday and there is evidence that CTL also contributes to protective immunity in man. It is actually quite sparse. You have to go back to a paper from McMichael et al. It was published in 1983 which showed that there an inverse correlation between the lytic activity of PBMC(?) and the final shedding from the nose and this is in the absence of antibodies. So, this is selection of subjects who were experimentally infected with an H1N1 virus and they were seronegative for this virus and so they actually were born before 1967 or after 1967 and never had been exposed with H1N1 virus.

So, the absence of virus antibody CTL can make a difference.

So, how can we detect these cells or specific T cells? There are a number of different principles or properties of these cells that we can use to actually demonstrate them.

First of all we can reactivate these T cells in vitro and measure the subsequent proliferation of cells, and we will come to that. We can also use the functional properties of these cells to detect them, that is we can look for lytic activity, look for cytokine production or we can look for activation markers that are expressed upon excitation of T cells, and more recently other methods are becoming available. We can look for epitope specificity of these subtypes, but more elegantly the study was multimers of MHC plus one peptide complexes to demonstrate virus specific T cells and because the frequency of virus specific T cells may be fairly low we sometimes need to expand these virus specific T cells by a prior round of stimulation and expansion in order to demonstrate the presence of these virus specific T cells.

So, in order to measure proliferation of T cells a number of different methods are available. A very old one is the friction 5 incorporation assay and in this assay the radioactive finding is incorporated into the DNA of defining cells, and after a short incubation for 1 day these cells are harvested. They are lysed and then DNA is captured on the viral filter papers and subsequently be evaluated. We can measure it as a measure for proliferation and the presence of virus specific T cells, and the results of playing with these tests are also expressed as stimulation and this means that the proliferation is defined by controls.

So, we have used this test in the past to demonstrate that five to six specific T cells were induced by vaccination and in this case we had two study groups, a healthy control group and a group that received liver tested patients and we looked for lymphocyte proliferation after infection and you could see that the SI, simulation index, is increased after vaccination.

In another study we saw a similar finding and after vaccination the SI goes up quite dramatically. So the advantages are that these SIs were relatively easy to perform. Of course, it has some disadvantages. It is not very popular anymore. We really don't know the identity of the cells that are proliferating, and it is a black box to find the radioactivity, but we don't know which cells actually incorporate the findings.

We know nothing about the central properties of these cells. We just show proliferation and there is always a possibility that reflects the next patient by non-specific T cells.

When I made the slide I realized that I actually performed this assay more than 25 years ago, and it made me feel old you know, but still this assay has been used for some time.

So, more recently there are alternatives for the thymidine incorporation assay which is based on the incorporation of a fluorescent dye which is known as CFSE. The dye is incorporated and after a subsequent excitation of the T cells the dye is diluted out at each replication cycle of these cells. After each cell division the dye is diluted out. This method was developed by Lyons et al and published in 1994. This is a control cell population and after stimulation these antigen-specific T cells lose their CVC signal upon cell division.

It is nice because proliferation is not affected by flow cytometry which also allows to identify the cells that are actually proliferating and this is an example of that, and I obtained the slide from Cal Van Ballen in our lab who is involved in a huge trial in which HFE candidate vaccines are evaluated and what he does is he is stimulating BUC obtained from HFE affected subjects with dendritic cells that have been transfected with messenger RNA and coding for proteins and you can see that in the controls what you see is a CFSE inactive population, but if you stimulate with dendritic cells expressing the keck(?) protein you can see that this is a population of cells that is losing their CFSE content and since we are able now to identify cells by flow cytometry we can clearly demonstrate that in fact the population, these subjects are in a reactive phase of fonicitra(?) simulation.

So, clearly this assay has some advantages over the classical assays. It is relatively easy to perform, can be detected by flow cytometry. So, no isotopes are used and it also allows the identification of the cells that are actually proliferating in vitro.

So, the potential disadvantages are that only the proliferating cells can be detected. So, we don't know anything about the function of these cells and again there is a potential risk of bystander activation.

So, as we move on we look for another functional property of the cell. This is lytic activity. Then we can use that to demonstrate the presence of virus specific T cells in the population and a very classical task to measure lytic activity of these cells is the chromium release assay and basically target cells are being labeled with a radioactive labeled chromium salt and target cells are usually autologous or MHC plus one matched S1 bulk virus transformer. These are lines for another requirement for an assay that you have to make MHC plus one to make target cells which then can be pulled from peptides or can be affected with virus or a recombinant virus expressing a viral gene of interest.

So most commonly vaccinia viruses are used for this expression for example the nuclear protein of the influenza virus. Then you also have to make these other requirements.

As effector cells in the future expanded PBMC could be used. So, that is really necessary because the frequency of virus specific T cells is too low to measure lytic activity directly in PBMC. So, we need to enrich for virus specific T cells and this assay is also used in limiting dilution assay to determine the CTL precursor frequency of PBMC and it is also used for screening of the clones.

So, after 4 hours of dose incubation of virus cells and effector cells the radioactivity can be measured in a filter supernatant as a measure for lytic activity present in the effector cell.

This is a typical result. As more effector cells are added you see more specific lysis of the target cells and these would be control cells

So, again this is also an assay that has been used for quite a long time and the disadvantages are again use of isotopes. Even now if you use these you see which cells are responsible for the lytic activity that you measure most likely you see it is relatively insensitive and we really need an expansion or enrichment of virus specific T cells in order to measure lytic activity and as I already said there is a need for autologous or MHC plus one matched cells.

So, for this assay there are alternatives, non-radioactive alternatives in which target cells are actually labeled with fluorochromes, fluorescent dyes, and in this way the lysis or the elimination of target cells can be followed, also, by flow cytometry and this has been used predominantly but these are clones and peptides that are used to label target cells.

Another surrogate of lytic activity is the expression of CD biostatin A which is a marker of degranulation and cytotoxicity as was demonstrated by Betts et al in 2003, and you can see that when the CD107A is expressed in the virus specific T cell population that the fluorine content goes down and that is correlated with the really cytotoxic effect on the target cells. So, there actually is CD1O7A expression which is a nice surrogate for the lytic activity of virus specific T cells.

So, we have used that also in our lab and it works actually quite nicely. So, we use the T cell expanded PBC population that we subsequently used to stain with tetramers and after stimulation with virus with peptide pulsed BLCL we can nicely see the expression of CD107a on the surface of the T cells.

More recently in our lab we developed another technique which is called the FATT CTL assay. That is the term for fluorescent antigen transfected target cells and this assay is based on the transfection of target cells with a plasmid that is expressing a gene of interest trying to influence a neutral protein fused to GFP or another reporter gene.

This way we can make target cells that can be easily identified by flow cytometry and we use a nucleofection procedure for that and nucleofection has high possession efficiencies to affect the percent in PBMC and in cell lines. It is important to look higher but since we are able to use PBMC for target cells we no longer have to establish for each individual subject what is called vested autologous B cell lines. So, it saves us a lot of time and we can proceed much quicker if we use PBMC as target cells.

So, these transfected target cells are then incubated with T cell populations and subsequently we can detect the fraction of GFP expression cells that are eliminated from the live gate as a measure for lytic activity with the T cell population.

This is an example of that. In the top panel we see different ET ratios that have been used and the top half of this panel a dicepel(?) was used specific for HFT1 direct proteins and you can see that in the live gate the GFP expressing cells disappear upon incubation with target cells that have been transfected with a rapid GFP expressing plasmid. As control we have the test GFP expressing plasmid and nothing is happening there.

So, we use it for HFE proteins but also for the neutrophil G influenza enzyme and in this case we used the effective PBMC. This was done with T cell lines. Here we transfect with PBMC and you can see that we get a nice specific lysis of PBMC that has been transfected with the plasmid that expressed the leukoprotein of influenza virus.

As I said, it is also more sensitive than classical protein release assay meaning that a lower ET ratio gives the same amount of lysis. So you can go down almost 10-fold with the ET ratio and to demonstrate that this assay was also very specific we used different T cell clones that recognized an epitope of the influenza neutral protein and we know there is variation in this epitope. So, we used three different neutral protein genes containing different versions of this epitope and we already knew what the functional affinity was of these T cell clones for all these variants and we see that the lytic activity that we measure in effector GLSA nicely correlates with functional affinity of these T cell clones for their peptide variants. It is highly specific. Also for HFE1 specific T cells we demonstrated that this is a very specific assay.

Now, in this case the assay was not carried out for 4 hours but for 16 hours. So, this allowed the XC cell detection of lytic activity of BNC directed to case ATC1 proteins. So this is done without a prior expansion of virus specific T cells in vitro. It is direct lytic activity of PBMC ex vivo and we see that the lytic activity that we measure changed to specific HFT1 proteins nicely correlated with the expression of CD69 and interferon gamma after stimulation of the PBMC with peptide pools that were derived from these HFT1 proteins. That is unpublished data generated by Joe Spritzen in my lab. He is a PG student and we used the FATT CTL site to test the level of cross recognition of cytotoxic T cells specific for the seasonal influenza virus, H3N2 with the NP obtained from H5N1 virus and as you can see the level of cross reactivity in most subject is really high, but it is a failure to boost cross recognition of the nuclear coating of H5N1 virus by CTL specific for the NP obtained from a recent H3N2 virus.

So, really this assay has some advantages. Again there is no use of isotopes. There is endogenous antigen processing and presentation like the use of peptides for example and autologous cell lines are not required because we can directly use PBMC as target cells. It is a big advantage. We don't need HLA types of study subjects. That is also a big advantage. We don't have to prepare viral vectors expressing relevant gene products and so we can simply make a plasmid and use that for transfection and it is very sensitive and the prolonged incubation times of the target cells with the specific T cell population allows the detection of lytic activity without prior expansion of these T cells and it is when the frequency of T cells is high enough.

So, another functional property of T cells and CTLs is production of cytokines, for example, interferon gamma and an assay that is very commonly used to detect interferon gamma producing T cells is the ELISPOT technique It is not really standardized but we can use standardized reagents which are commercially available and it is the general principle of this technique that membranes are coated with antibodies that recognize interferon gamma and subsequently T cells are added but have to be stimulated with antigen presenting cells presenting an epitope or proteins and then after some time the cells are washed away and the spots where cells secreted interferon gamma this interferon gamma is then detected by secondary antibody which is subsequently detected by using a substrate that responds to a precipitate and with this precipitate you can then detect quite easily.

In the old days we were doing it by eye and it is really a lot of work. We had always two people counting all the spots so we had an accurate measurement of the number of interferon gamma producing spots but nowadays this is done by use of a digital camera which can produce more or less objective results.

That is the real thing. Here we see the production or the detection of spots here of T cells that produced interferon gamma after stimulation with the corresponding peptide.

So, it is sensitive, can be performed without prior enrichment of virus specific T cells. So, it can be directly performed from population and it is for high throughput testing. It has some disadvantages. The identity of cytokine producing cells is not always known. It also depends on the antigen that is used for stimulation and less cells are sorted prior to testing. So, we see sometimes with virus detected BSELs that we will just try to do some background but if we sort the CD8 positive T cells prior to testing the background is completely gone.

An alternative is intracellular cytokine staining, for example, interferon gamma. So basically you do the same thing and so you detect interferon gamma production by virus specific T cells. It works well for in vitro expanded and T cell clones. The T cells then are stimulated through MHC plus one matched target cells which are effective for example with recombinant virus expressing in the gene of interest where they are loaded with relevant peptides. Then the T cells are treated with blockers of the secretory pathway, monensin, brefeldin A to accumulate interferon gamma for example in the cell. Then the cells are fixed and then stained for cytokines but also for other CD markers like CD3 or CD4C8 to identify cells that are actually producing cytokines and then the cells are analyzed by flow cytometry.

So, we have used this procedure also. So, in this instance we used cells that did not express HLA-A or B alleles themselves but that have been transfected with one allele. So, this way we can actually dissect or detect the HLA usage in the influenza virus specific T cells response. So, the adult response is 50 percent. These are controls and then you see which proportion of the CTL response can be attributed to one of these four HLA-A or B alleles.

So, what we found actually doing that and I just wanted to make this comment to see how complicated the analysis of T cell responses actually can be, but if you look in the groups of HLA matched individuals and these individuals have been matched for both HLA-A and B alleles that in a group that is lacking HLA-A2 but otherwise that is fully matched for the other A and B alleles the response to influenza is much lower, okay? So, this is an HLA-2 negative group of study subjects that lack the capacity to respond to a very immunodominant epitope from the matrix protein.

It is also interesting and it has also been described before by others, so, this group is HLA-B8 positive like all the others but in the presence of HLA-B27 the response to the HLA-B8 is much lower and that has probably something to do with the existence of two overlapping peptides or epitopes that are present in the nuclear protein.

So, maybe there is some level of interference or competition that prevents or inhibits acute HLA-B8 specific responses in the presence of B27.

So, I just showed you the data on HLA usage. It also corresponds with the total influenza virus specific CTL response in these study subjects. These again are the three groups and again we see that in ELISPOT that the response in HLA-A201 negative subjects is significantly lower than in the two other groups.

So, that is a complication that should be taken into account when measuring T cell responses in outbred populations. So, is this really relevant? I mean that is the absence of HLA2 has an impact on the clinical outcome of infection or we really don't know. There is very limited data. This is all we have and we collect the PBMC in the times that the neuraminidase inhibitors were tested.

So, I obtained PBMC from study subjects and the placebo groups of these clinical trials and I looked for expressions of HLA-A2 in these study subjects and I could come up with four subjects that were HLA-A2 negative, and four that were HLA-A2 positive.

As you can see and this is absolutely not statistically significant and much more I think should be done. This is just a trend that we observed that indeed in the HLA-A2 positives that do mount the strongest CTL response the duration of virus shedding was actually shorter than in HLA-A2 negative subjects.

Again, this is not telling. We really need to do a lot more work but it is trying to do it in the right direction.

Quite recently a novel technique was developed by Altman and coworkers in 1996, that was designed which used multimers of MHC plus one peptide complexes known as tetramers and basically what is being done is that tetramers are being made. So, these are MFC heavy chain, five chain peptide complexes which are backed with a biotin that is multimerized by using streptavidin which is carrying a zerogram(?) of HIPE and this way these reagents can be used to detect CTL proper specificity directly either of PBMC or in enriched T cell populations. This is a very powerful technique that I think has revolutionalized the way we can measure T cell responses. It is very easy. Everybody with a flow cytometer can now detect T cells very easily. It identifies all epitopes with a certain specificity without taking into account the functional properties of T cells. It is very simple, can be performed ex vivo and results depends a little bit on the frequency of CTLs that can recognize a given epitope.

There are also some disadvantages. So, the number of available MHC class 1 peptide complexes is still limited but the numbers are growing. It only works for defined epitopes. You really need to know the epitope in order to make these multimers.

Of course you have to know the background, the HLA background of the study subjects because otherwise you cannot interpret the results and there is no information on functionality of the T cells that are detected unless you can stain for example for cytokines and the localization of CD107a for example. You can also use simply the peptides. So you can use peptides to stimulate PBMC to find out what the numbers are of T cell specific response to a peptide stimulus. We have done that for the groups that I indicated earlier These are the HLA epitypes of these groups and as was demonstrated by Jack Bennick yesterday that there is a immunodominance and there is a hierarchy of immunodominance epitopes in the mouse model is also true for human CTL responses. These are individual peptides and you can see that some peptides are recognized better than others.

There is a hierarchy of response, okay, of individual epitopes. It is also clear that there is quite some variation within these HLA identical subjects but there is a huge difference in response to for example the F1 epitope in subjects that are identical and completely matched for the HLA-A alleles.

This is my last slide. There are some general considerations regarding the detection of T cells in study subjects. There are many different technologies, each one with its own advantages and disadvantages, pros and cons. Some require prior restimulation and expansion of specific T cells in vitro. The source of antigen used for the stimulation of the T cells will influence the outcome of the results. There is immunodominance; there is a hierarchy of immunodominant epitopes.

The HLA system is highly polymorphic. There is a huge complication analyzing these responses in the general population and the HLA phenotype dictates which epitopes are recognized and which are not and I gave the example of the M1, of the recognition of the M1 5866 epitope in the matrix protein.

Another complication is that subjects have different history of exposure to infection. That is something that also should be taken into account and I also showed you that the HLA background with non-corresponding HLA alleles can influence response to corresponding HLA alleles but it is also complicated and I don't have the time to go into this and I have to disappoint Bob Couch. I am not going to talk about the variation in T cell epitopes but we have seen in seasonal influenza during 40 years of evolution of H3N2 viruses that these viruses also tend to lose epitopes and can escape from recognition by specific T cells.

So, I will leave it here.


DR. ZAMBON: Thank you for that very comprehensive overview. Just in the interest of time perhaps we have time for one or two quick questions.

Introduce yourself, please?

DR. GREENBERG: Harry Greenberg, Stanford. That was terrific and comprehensive. Do you have any recommendation? There was an awful lot of assays there. We are talking specifically about vaccines and how to evaluate them vis-a-vis their possible ability to stimulate protection.

So, for everybody here if you had to put your dime down on where you would go do you have any favorite way of going? This was a huge discussion.

DR. RIMMELZWAAN: I was already afraid for this question. I mean for none of the assays that I mentioned we know how the outcome correlates with protection; we really don't know. So, I think a lot of work needs to be done in this area.

DR. BRACIALE: Braciale, University of Virginia. Guus, thanks so much. I have a quick question. If we assume that the frequency of CD8 precursors against influenza in an immune individual is .1 percent and we are going to take an unmanipulated PBMC population without in vitro expansion how many mls of blood do we think we have to get in order to determine that frequency?

DR. RIMMELZWAAN: Point 1 percent of CD8 positive T cells.

DR. BRACIALE: Memory cells.

DR. RIMMELZWAAN: Yes, so that is borderline for detecting it by tetramers for example. It is right on the threshold of detection of tetramers. So, in order to measure that and the number of T cells reliably I think you really need to expand it in vitro in order to get a better feel whether T cell memory was present or not. I don't know how many mls of blood that would be.


Agenda Item: Novel assays

DR. ZAMBON: Okay, I think probably in the interest of time we need to move on. I would like to call on our next speaker, Gary Nabel. I think the subject of Gary's talk is given as pseudotype viruses and I am assuming this is retroviral pseudotypes as opposed to influenza pseudotypes.


Agenda item: Pseudotyped viruses

DR. NABEL: My apologies for loading it now. I wasn't able to get here early enough this morning because I had to drop my kids off at school.

So, this will take a second. Thank you for the clarification and the reminder that everybody views their own viruses through the prism that they are used to and it is actually with a little bit of trepidation that I describe these results today because it is a little bit different from what I think most influenza virologists use in terms of evaluating neutralization, but I hope you will see that there are some potential advantages to bringing some newer technology, different technology to the system.

Basically the system that we are using for measuring neutralization involves a lentiviral vector system where essentially we can use a reconstituted lentivirus which contains a reporter gene in this case leuciferase which can be read out quantitatively and quickly using a luminometer(?).

We in a producer cell furnish in trans a gag and a pol open reading frame that is expressed off of a CMB promoter and then in the initial studies we included the influenza hemagglutinin. I will show you later that we have now extended it to include neuraminidase and some other gene products that now allow us essentially to put any hemagglutinin or neuraminidase that we have looked at onto the surface of the lentiviral vector and then to use this vector to infect cells and measure entry.

This is an example from the first description that was in a paper related to the 1918 influenza virus and we were really forced to use this system in that study because at the time there was no replication competent 1918 virus and we were interested in generating a vaccine that could neutralize this virus

What we did essentially was to make the vector, introduce it into cells and initially did the work with H5 derived from the Kam-1 strain from Thailand

You can see that there is quite significant expression compared to controls and we then went on and looked at the ability of these pseudotype viruses to be neutralized by a specific antibody.

What you can see is that if we make pseudotype viruses from the 1918 H1 that basically if we included antisera against H1 we were able to specifically neutralize, not so with the anti-H5 and conversely if we were to make an anti-H5 they neutralized with H5 but not H1 and so this was encouraging and suggested that we could have a way of readily measuring under standard laboratory conditions neutralizing antibodies with a quantitative readout.

Now, that was the good news. What I haven't told you is in order to get the H1s to work in these first studies we actually had to modify the cleavage site for the hemagglutinin. We ended up actually transposing the polybasic cleavage site normally found in H5 into H5 in order to get that to work and so while we were happy that that worked there were some things that we preferred to see in an assay that weren't quite there yet.

So, I will tell you about the improvement that we have made that now make this a little bit more generalizable in terms of different flu strains and before I do that let me just point out what is already in both papers, and I won't spend a lot of time on this. There is excellent correlation between the kind of inhibition that you will see using HAI microneutralization or the lentiviral vector inhibition.

You will notice that with the lentiviral vector inhibition we have a larger dynamic range in the assay because we have such sensitivity from leuciferase and as I will show you later we are actually able to do better than that when we include neuraminidase because we do get more effective gene transfer.

The correlations with protection are actually very similar to what you see with the HAI assay and here is another indication of that where we have done studies in both ferrets and mice in this case with corresponding challenges and you can see again quite nice correlation between all three assays. The only thing we have noticed is that there are instances where the microneutralization assay doesn't read out because of presumably serum factors or other non-specific factors where the lentiviral assay does.

Now, this was the good news. The more guarded news at least when we began these studies was that while we got very good gene transfer of the H5 viruses and we found that when we included the neuraminidase we could actually get even better gene transfer and that actually we subsequently learned was because there was probably more effective and physiologic trimming of the carbohydrates on H5

Now, we then thought that we had the answer to pseudotyping all different influenza viruses and when we went to do this you can see that when we looked at the ability of neuraminidase to increase the gene transfer of H1s or H3s that it really did nothing.

I should point out this is a log scale. So, we are really getting 7 logs of leuciferase, 7 to 8 logs of leuciferase activity and so we were challenged by this lack of gene transfer activity to improve this assay and to see if it could be more generalizable.

The only other point I will make about these HANA pseudotype viruses is on this slide when we fractionated these by purifying them on buoyant density gradients they had the properties of typical lentiviral vectors and I will show you later we can actually measure HAI titers, HA titers on these lentiviruses and infectivity in the co-segregate.

Similarly when we look at the sensitivity to entry to a lysosomal inhibitor of aflamycin(?) like natural flu virus this hemagglutinin is sensitive as opposed to Maloney murine leukemia virus which is a pH independent virus and does not. So, it is behaving essentially like an influenza virus

If we look at the H5s for their sensitivity to neutralizing antibodies we have now looked at a variety of different H5 strains. You can see Kan-1 Vietnam, turkey, the old Hong Kong 21. We have now extended this to more than 20 different H5s. They all mediate transfer. They all are sensitive to neutralization and here is the neutralization of some of the H1s that are essentially insensitive but before I tell you more about that I really need to tell you how we got the H1s and H3s to pseudotype.

It turns out that there was a report in the literature from Matrosovitch and Hans Peter Klenk last year about a serine protease that they had suggested might be involved in the processing of viral hemagglutinins. For a number of years many people had suggested it was furan-like proteases that cleaved the hemagglutinins.

Clearly that is possible maybe for a subset of viral hemagglutinins but not all and in this very nice study in the Journal of Virology Bottcher et al had shown that there was a type 2 transmembrane serine protease called TMPRSS 2 which is basically a protease that is found in the small intestine, in the respiratory tract, in the prostate, colon, stomach, salivary glands and what they showed in their paper was that it seemed to mediate pretty efficient cleavage of a variety of viral hemagglutinins.

So, we then went on and basically co-expressed the hemagglutinin, the neuraminidase and the TMPRSS 2 expression vectors in the producer lines and you can now see that for H1s and H3s and even H9s that where they weren't processed and where they weren't active previously they now are functional and giving quite measurable readouts above background.

Not all of them are 7 to 8 logs of magnitude but there clearly are at least 3 to 4 orders of magnitude above the controls.

Interestingly the TMPRSS 2 is not needed for certain strains notably the H5s and the H7s which have this polybasic cleavage site and in response to I think John Wood's comment in the first talk this does work for H7 as well.

We essentially now haven't found, I should say that we don't find TMPRSS 2 interfering with any cleavage that would happen and we have found that it seems to work for any combination for which a virus has been found.

I mentioned the fact that these pseudotype lentiviruses actually will give you a hemagglutination signal and here is an example.

You can see that if you just express the H5 without the N1 it will mediate gene transfer. It is down by a couple of logs in terms of that but if you now express the neuraminidase you can see that the hemagglutination is readily observed. The same is true for the H7 and as you can see with the H1N1s without the TMPRSS 2 you don't see hemagglutination. It essentially looks like this but when you have the functional viral spike it does mediate hemagglutination as well.

We have used a variety of different reference sera from the CDC and these simply just show that taking the standardized reference sera against H1s they work against H1s but the H3s don't work against them. The reference sera against H3s work against H3s but not against H1s and again you will notice the quite high titers to which these can be found.

I mentioned earlier that there had been a variety of different H5s that we have looked at. The beauty of the system if that simply by making some mutations very quickly in different hemagglutinins we can generate new strains and have assays ready to test quickly and what is shown on this slide you don't really have to focus on all of the details but I show immunization. These are mice that have been immunized with either homologous vaccines compared to the reference and that is shown in the upper bar or a heterologous in this case the Vietnam strain and you can see in all instances that we are able to pick up good neutralization against a variety of different H5s including the clade 2s.

We have also been able to do multivalent immunization. So, we have actually in mice been able to take mixtures of 10 different H5s, vaccinate and then we can read out the specificities of those neutralizing antibodies on the different strains and what we then do is we can actually stratify the different hemagglutinins and ask if we mix certain combinations do we get better neutralization or if we mix other combinations do we get worse and you can actually see here is a mixture of five that did less well than the 10 but another mixture of five does better and so it allows us to stratify and to prioritize mixtures that give us the broadest coverage.

Finally I want to emphasize that this has been enormously valuable to us as a discovery tool. This is work that has been previously published. Nancy Cox referred to some of the structural studies yesterday from James Stevens and Ian Wilson and what we have been doing is based on many of the analyses that they have done and previous genetic information from the literature we have been able to make site-directed mutations in the receptor binding domain to use the system both to do glycan arrays to show that we can modify the specificity or 2,3-linked sialic acid to generate monoclonal antibodies and that is another feature of this system.

We can actually do high throughout screening because it is all done on the bench. We can literally do thousands of screenings of samples and we have now generated monoclonal antibodies not only against the H5s which has been done by many labs but now also against our mutant viruses that in this case I am actually showing you three different types of monoclonal antibodies, one that reacts with the wild-type virus but not with the mutant, one monoclonal that reacts with the wild type and the mutant and a third monoclonal that reacts with the mutant and not the wild type.

So, very quickly again it gives us tools to look at specificity, to map determinants of neutralization and equally importantly if one wants to quantitate the affinity of neutralization because we have this very quantitative readout with leuciferase various different antisera or monoclonals can be readily compared as you can see in the bottom panel here and this if you would like to look at in more detail is in the Science paper that we published.

We, also, are able then to use the same lentiviruses to look at differences in specificity and confirm them by hemagglutination and I will just draw your attention to this interesting virus that we defined in the paper that has not only lost 2.3 specificity relative to the wild type but it has begun to gain 2,6 specificity relative to the wild type as well and so we are able to begin to prospectively develop vaccine candidates that may be more 2,6 preferring in terms of their receptor specificity, and finally based on this kind of work we have also been able to take prototype DNA vaccines that we have run through animal systems and bring them forward into human studies where we are also able to measure neutralizing antibody responses with these assays.

So, to summarize it is now possible to generate functional pseudotype lentiviral vectors with a large proportion of HAs and NAs found in nature. The assay shows high sensitivity, specificity and the ability to be performed under GLP with standardization between labs. There is actually a commercial company that does such testing for HIV using lentiviral reporter genes, a company called Monogram in California and they for example could seamlessly incorporate the kinds of HAs and NAs that I have told you about here into their assay systems and it could be done by others. It could be done by laboratories just as a research tool as well.

Neutralization titers parallel those determined be hemagglutination inhibition as well as microneutralization, eliminates the need for high containment and can be used to screen for vaccine efficacy monoclonal antibodies and HA specificity.

I have also briefly mentioned its use as a discovery tool to understand the structural basis of influenza substrate recognition and to new vaccine strategies and monoclonal antibody preparation and then the fact that we can use this to move various new platforms particularly the DNA platform into human study.

With that I would just like to thank a number of people who contributed to this work, Zhi-Yong Yang, C. J. Wei, Wing Kong and Ling Xu with the Vaccine Research Center, David Smith who has done the glycan array assays, Terry Tumpy at the CDC who has been very helpful in looking at the HAI titers for both H5 and H1 and Adolfo Garcia-Sastre for the H1 reconstituted virus.



DR. ZAMBON: There is just time for one question.

DR. MURPHY: Brian Murphy, NIH. Gary, why can't you just use trypsin in your producer cell lines rather than having to use the other protease? Maybe you said it and I didn't hear.

DR. NABEL: We tried it, Brian. It just doesn't work. I think it has to do with where the cleavage occurs and the requirement, I suspect with TRMPSS 2 that there is a certain confirmation of the cleavage site that has to be exposed and it is probably cleaved intracellularly and once you actually have it on the surface of the lentivirus it doesn't work, for whatever reason, and we have also tried to express trypsin and it didn't work.

So, it may have something to do with specificity, access of the protease but it for whatever reason doesn't seem to help.

DR. ZAMBON: Thank you very much. I would like to introduce our next speaker, Maryna Eichelberger who is going to be talking to us about her recent work on NA inhibition assays.


Agenda item: NA inhibition assays

DR. EICHELBERGER: Good morning. I think what you are going to hear from me is a reiteration of what John Wood has told you earlier. I am going to be talking about assays to measure neuraminidase inhibition antibody titers and give you an update on an assay that we used. It uses neuraminidase as an endpoint to measure virus neutralization titer.

So, I am going to go fairly quickly over the background which was very nicely introduced yesterday when we heard about HI responses correlating with protection but that indeed NI titers, also, contribute to protection and that when we think about correlates of immune protection perhaps we should be thinking about the sum of the responses and so indeed we do need to consider antibody responses to neuraminidase as well, and this has been recognized for a very long time and Dr. Kilbourne a very long time ago demonstrated that in vitro antibodies that neutralize neuraminidase activity reduce plaque size and the total virus output in a culture, and it was demonstrated by Compans first and then by several others that antibodies to neuraminidase will block virus shedding in vitro.

I think there is still some room to understand the mechanisms by which antibodies to neuraminidase contribute to protection in vivo.

It has been demonstrated in numerous human clinical trials that neuraminidase specific responses reduce viral replication and prevent disease and that this has some relevance when we think about pandemics and the 1968 shift from H2N2 to H3N2 viruses that the anti-N2 response did contribute to lessening the impact of that pandemic and for our current sera of an H5N1 pandemic I think it is worthwhile considering the animal studies that have been done to date that demonstrate antibody to neuraminidase contribute to lethal doses of avian influenza viruses in chickens as well as when we simply in mice transfer studies of serum that contained only anti-neuraminidase responses, provided partial protection to an H5N1 lethal dose and these studies were done by Matt Sandbulte when he was a postdoc with Richard Webbing.

So, why don't we all measure anti-NI titers? It seems like this should be at the forefront of what we do on a routine basis and yet it is not, and I think John emphasized that this assay, the traditional NI assay is rather cumbersome. There are quite a few limitations in taking large numbers of samples and running them through the assay. The assay, however, is quite rigorous and is reproducible within a lot and if we had an assay of similar rigor perhaps we could adapt it to become more sensitive and use it on a more routine basis.

So, Matt actually has joined us at the FDA and has been working on miniaturizing this thiobarbiturate assay, the traditional assay and we are not at the point to say that this can be used as an alternative.

There is a lot of validation studies that need to be performed to show that this indeed will give us the endpoint that we want.

Perhaps just to outline the objective of this miniaturization assay, obviously to analyze large numbers of samples we reduced the volumes that are used in the traditional assays and are performing the assay in either PCR tube strips or in 96-well plates. This allows you to do a heating set that is required with absorbent readings in a plate reader instead of individual cuvettes.

So, we hope that this would be an assay that perhaps we could share and validate by having it performed in several other laboratories as well.

John has also mentioned several limitations to the assay and I believe he pointed out that the choice of virus is important and in Dr. Kilborn's original studies he was very careful to use antigenic hybrid viruses. When we are going to be looking at antibody-responses to H5N1 it is going to be necessary to create antigenic hybrid viruses so that we don't confuse NI titers with antibodies to hemagglutinin that might also block and interfere with the NI assay.

We, also need reference antisera so that we can compare results between labs and to point out that the choice of substrates is really important. Fetuin is currently used as a large substrate because antibodies to neuraminidase often induce and primarily work by simply blocking the substrate from accessing the enzyme access site which is within a pocket within the neuraminidase head.

The smaller substrates like NANL or chemiluminescent(?) or fluorochrome-mediated small substrates of neuraminic acid are so small that they act as the enzyme active pockets very easily and are poorly inhibited by some antibodies and so we do need to think about continuing to use a large substrate molecule and for this reason the assay really hasn't come to the point where we can do it very easily with a conjugated product, fetuin or other large substrates and from what I know have not yet gotten to that point of chemistry.

I just wanted to point out that there are vast numbers of structural studies of neuraminidase that have been performed over the years primarily through the Webster lab where Rob's lab created a lot of monoclonal antibodies that was used to generate escape mutants. The escape mutants have been sequenced and Julian Ayres' lab in particular was responsible for identifying specific epitopes on neuraminidase that are recognized by antibodies and then these complexes have been crystallized and Peter Coleman's lab has generated numerous structures that help us to understand the interaction between antibody and neuraminidase and the take-home message is that there is a vast number of loops on the antibody side that recognize several loops on the top surface of the neuraminidase molecules and as I said before the enzyme-active target is within the surface and you can take two different monoclonals. NC 41 will inhibit only the fetuin products from cleavage whereas NC 10 when it interacts with N9 will inhibit both the small and large substrate molecules So, there are monoclonal antibodies that will inhibit small enzymes from the NCs but obviously we need to take into consideration both of these types of antibody responses that will be present in polyclonal antisera.

So, what alternative assays are available? There are some very old assays, for example hemagglutination inhibition that was done to show that neuraminidase-specific antibodies can inhibit hemagglutination and so if you use mismatched hemagglutinin you can tease apart the HI response from an NI response.

Lectins have been used on a more high-throughput screening kind of method. These haven't taken off quite as well and it might be worthwhile considering some of these approaches. The basis for the approaches in the traditional assays we measure the product of the enzyme reaction of sialic acid or the product that is left bound to the glycoconjugate is going to be galactose or in a lactose group that is recognized by lectins.

So, you can through knowing the intensity of the color that is produced by the HRPO or if you use biphenylated RCA lectin identify how much of the product is left within the well that was coated with fetuin as a large substrate.

So, it is possible to do an ELISA-based technique. Ideally we would like to synthesize large substrates that are labeled with a fluorochrome that is easy to measure or with a chemiluminescent output.

So, these types of things are under investigation at the moment and hopefully as time goes on we will be able to compare these types of assays with the more traditional assay so that we can come to some agreement about which assays are best to use.

We will need to determine assay sensitivity, the robustness of the assays and of course validate these assays.

So, in thinking about including an anti-neuraminidase response in our evaluation of the serologic responses to either vaccination or wild-type infection I like the idea of a virus-neutralization assay. I think we really need to understand the antibody specificities that are playing a role in that assay and perhaps we will find that anti-NA antibodies, also contribute to that neutralization, but in thinking about this we have tried to come up with a microneutralization assay that perhaps lends itself to a higher throughput type of screening method in addition to having an assay that has less within-lab variation.

So, this slide basically introduces some ideas about virus replication assays and of course the traditional assay that would include looking at anti-NI or looking at NI titers is in plaque-reduction neutralization assay where the number of plaques and the size of the plaques really can reflect anti-HA and anti-NA responses. The more traditional microneut assay that is currently being used by many people is an ELISA-based detection with antibodies to NP detecting how much virus has replicated in that well.

We have taken this as the basis of this assay and decided to look at virus in a different way, quantitating it based on neuraminidase as the readout and just to explain to you that of course neuraminidase activity is a hallmark of all influenza A&B viruses and this type of identity tests has been the basis of quite a few rapid diagnostic tests and the basis for these tests really used the small substrates that I was referring to previously, n-acetyl neuraminic acid that is conjugated to a chemiluminescent compound and in our case we use a fluorescent compound, methyllumbelliferyl and so when neuraminidase comes along and sees this product the n-acetyl neuraminic group is released as well as the lumbelliferyl group which now becomes fluorescent and you can measure fluorescence then as your measure of enzyme activity.

You can use this assay on whole virus preparations and see that the neuraminidase activity is proportional to your input virus. You can use it as your measure of virus that has replicated in a dish by simply either adding the substrates to your culture dish or taking supernates and measuring how much virus is in that supernatant and this is a very sensitive assay and you can use multiplicities of infection that are quite low to get very large relative fluorescence units.

So, our adaptive microneut assay we call the accelerated viral inhibition assay with NA as the readout and it simply is methodologically very, very simple. You simply make serum dilutions. You add your virus of course with some added trypsin in BSA and you add your washed cells, incubate overnight and we like to remove supernatant for our neuraminidase assay which is a simple 1-hour assay in the presence of the small substrates and then we read fluorescence on a microtiter plate.

So, we really developed this assay to have a high throughput method for screening antivirals and this work is being done by a student in the lab of Arosh Hasstoufighi in collaboration with a high-throughput screening laboratory at Johns Hopkins University and we use it to identify antiviral activities but as I said we have been adapting it to measure neutralizing antibody responses and so if you look at the titers for our assay compared to the traditional microneuts you certainly see that the assay is as specific and when we look at several pre- and post-human sera samples we believe that the assay is possibly a little bit more sensitive but is on the same, within the same ballpark as the microneut assay.

So, simply in conclusion I think NI titers should be measured routinely in vaccine trials and that having a practical method to enable this group would really be very helpful.

I have also shown you that neuraminidase provides a marker that you can use to follow virus replication and perhaps we can exploit this to increase sensitivity and reproducibility of our virus assay.

Any questions?


DR. ZAMBON: We have time for just one question.

PARTICIPANT: Just a few quick things. First of all we all saw trying to do microneut to minimize assay, as a standard assay; however, human H1N1 viruses have such low activity and they also do not grow well and so both assays have limitations because of human influenza viruses H1N1 and another thing is we also tried this galactose exposure detection assay. In our hands, and it is commercially available, it didn't work well, but we will try to do more if you have good results with this.

DR. EICHELBERGER: We actually haven't tried the assays. So, that is good to know.

PARTICIPANT: And I totally agree with we need large substrates but have it labeled differently like more advanced, should be more advanced kind of technology and I have to say that in equine vaccine production we are already using neuraminidase activity in the vaccine making process.

DR. ZAMBON: I am going to ask you to save your questions until the break because we are running about half an hour behind time. So, if I could call on our next speaker, please to give her presentation on phage display libraries, Dr. Hana Golding?

Thank you.


Agenda item: Phage display libraries

DR. GOLDING: Thank you very much.

So, I really would like to focus on the target of ours and this is really we found that there was a major research gap in influenza vaccines and the main gap is the limited availability of analytical assays for qualitative and quantitative evaluation of the antibody responses elicited by the new or improved influenza vaccine candidates, and very much on everybody's mind is to try a method recognized by the heterotypic neutralizing antibody that we all hope to elicit by the improved vaccine and we feel a little bit like Alice in Wonderland looking through the keyhole, and the keyhole is really the available assays and we decided to go into the hole and see what else we can find that will tell us what are the differences qualitatively between the type of antibodies that ultimately would provide protection.

In order to do that Surender Khurana in the lab took the approach of constructing phage display libraries from both seasonal H1N1, H3N2 and most important for this talk the H5N1 and H7N1, the avian influenza, and what does it mean to make phage display libraries?

We are basically taking the cDNA of the entire genome,all the H fragments and you then subject them to limited DNA treatment. You then take the size of the fragment that you are interested in. It can be either 50 to 200 or as in our case we actually took 201 KT and you basically express it as a protein in the G3 code protein of phages and you create libraries that express every possible open reading frame for all eight genes of the strain that is indicated.

The key question then is how to use them and we use them by panning against sera from either animals or humans and finally those that are bound with high avidity are basically sequenced to identify the insert.

So, how can we use it? The first set of slides will demonstrate our collaboration with Antonia Nazelvechio, Kanta Subbarao and Cameron Simmons and as you all know they were able to go to the survivors of H5N1 Vietnam exposed individuals and generate not just human monoclonal antibodies but several of them turned out to be broadly neutralizing and even more importantly broadly protecting in animal studies. These were conducted by Kanta Subbarao and what I am going to show you how our phage display libraries even at that stage allow us to map the context residues and the recognition side for these human monoclonal antibodies and especially will focus on the FLA 5.10 which as you can see is clade 1 specific, does not give much protection against clade 2 and two other monoclonal antibodies at 21.14 that give you broader cross protection even though it doesn't always correlate with in vitro mutant titers and the FLA 3.14 will give you broader cross protection and also microneut.

So, let us look at the 5.10. We found out that using this phage display library we identified a segment that was recognized by these antibodies. It included part of the receptor binding domain with a little bit an N terminal During this period Kanta's lab had looked for escaped mutants from this antibody in vivo and identified two amino acids that are shown here, this lysine and serine. However, those could not quite explain the differences in cross protection because there were amino acids that were actually conserved between Vietnam and Indonesia. We followed these libraries by random peptide libraries that allow us to identify what we consider contact residues for these particular antibodies that are shown here and you can see at least one amino acid, this lysine is clearly different between the Vietnam and Indonesia and in subsequent studies Surender demonstrated by removing or making alanine substitution here. It is really does not bind anymore to FLA 5.10.

Now, let us see where is it in the structure if you map them actually to the structure, this is the lysine and this is in the system and this is the second strip and you can see that in the structure this is a trimer. It is in the monomer. They are very close together and what is very nice is that if we now add to amino acid the escaped mutants that were generated by Kanta's group; so this is again the lysine in yellow, and those are the two amino acids that they identified by escape and in the structural strip, trimer, they are clearly all together.

So, you can envision how escape or changes in this amino acid may indeed change the binding of these antibodies.

So, this was the one that was very clade specific. The second one was the 21.14 that had a broader cross protection. We found out that we needed to go to a fairly large segment expressed by the library in order to find a fragment that is bound by these particular antibodies and when we mapped it further we found two stretches of amino acid that were a little bit further away in the linear structure. However, again, when you put it on the three-dimensional structure you see that they are very close together.

We, also, took then the fragments, the protein fragment that we identified for the FLA 5.10 and we synthetically expressed it in bacteria in E. coli and then generated ELISA to see the binding and indeed not only in the context of phage but also in ELISA you have very nice binding of this monoclonal antibody and as might have expected this other antibody did give definitely binding to this region although we lost the dose response which shifted to the left.

The third monoclonal antibodies that have actually brought cross protection did not bind to this particular fragment.

So, what does this monoclonal recognize? We haven't finished the contact residue it is making but we definitely require a fairly large segment of the HA to be expressed. It does include the receptor binding but also some N and C terminal and we hope to actually identify the contact residue in the near future.

So, it is very nice to use this assay for mapping of monoclonal antibody. You have one antibody to deal with. What do you do when you have multiple antibodies and the most relevant of course next one is the sera from the patients from which these antibodies were generated. So, through our collaboration with Cameron Simmons we were able to get pooled sera from those exposed individuals that survived the Vietnam H5N1 and what I am actually showing you is what the data really look like.

Each of these little sort of rectangles represent one of the phages that bound well to these antibodies and now we looking of course in pool antibodies and you can see that you have, this is first I am going to show you just the HA and NA and then the rest of the gene.

As far as the HA you see a very strong actually recognition of epitopes in the HA2 part of it. In the HA1 you see again multiple different epitopes. Some of them are in the receptor binding site. Some are more N terminal.

When we tried to look are these neutralizing antibodies represented in the whole sera we definitely found an FLA 5.10-like epitope recognized by the pool sera but we haven't found this very large fragment that is required for recognition of the more broadly cross-protecting monoclonal antibodies that I showed you.

What does this mean? It could be that simply these kind of antibodies represent a very low frequency among the overall antibodies that is well generated in these individuals or that at this point we were not able to identify them in our library.

I, also, wanted to point out that this patient did indeed generate antibodies against NA that actually included at least some that could be mapped to the NA catalytic site.

So, to summarize this part I wanted to again show you that using sera from survivors of the H5N1 we find very broad recognition of both the receptor binding domain as well as some that were more restricted to the N terminal of the HA1, very large epitopes in the HA 2 domain and the one in the NA that is shown here in green is actually very close to the catalytic site and this is the predicted binding site for also oseltamivir which I think indeed corresponds to the same site.

So, yes, infected people do generate the NCMA antibodies and some of them may have therefore the ability to block NA activity.

So, now we come to the big question. Oh, sorry before I go to the next step what do we see in the rest of the genomes and this is also something that was a bit of a surprise to us because we expected that the rest of the genome will be recognized by many, many antibodies in these patients but this was not the case. We found a fairly focused response. It was focused of course on the proteins that were known to be dominant in terms of the frequency and clearly most of the epitopes were in the MP and in the M1. We did find a recognition of MS1 and we did find recognition of M2 in these sera but unfortunately so far all of them are not mapping to the XO domain but rather to the transmembrane in cytoplasmic. So, even in this group of patients we haven't found evidence for antibodies against the XO domain of the M2.

This we also thought was very interesting because as you know several investigators found a new and important role for PB1F2 as a potential pathogenicity factor both in the 1918 and possibly the more pathogenic H5N1 avian influenza and to our surprise there was a very robust response against these proteins suggesting that this particular isoform of PB1 is indeed expressed in vivo and is generating antibody responses.

When we did a similar study with people who were exposed to seasonal strains through our collaboration with the monitor we found much lower frequency of PB1 F2 specificity, but now I would like then to shift to our attempt to understand vaccine quality and of course we hear a lot about attempts to improve the quality of vaccines. For a start we managed to collaborate with John Beigel and Julie Metcalf at the NIAID who have initiated the study with the purpose of producing FLU-IGIV and what we used for a start was the currently licensed vaccine which is the H5N1 Vietnam produced by Sanofi Pasteur and their thinking was maybe we should start with the highest doses that have been used in the Phase III clinical trials and go up from there. Maybe it will improve somehow either the percent of seroconverters or the breadth of the response.

So, I am going to show you only a single slide. We analyzed sera from these vaccine recipients that received the 90-microgram dose which is the highest dose that is used in the current trial and what do we see? In the top part we are looking at those that have actually very low HAI titer. The black ones are basically epitopes that are common to both of them and you see they are mainly mapping to the HA2. When we look at the HA1 those that have high HAI titers have this sort of unique set of epitopes that actually map to the N terminal a little bit N terminal to the, upstream I would say of the receptor binding.

As a matter of fact we found very few epitopes in the receptor binding domain. Maybe this one isn't the terminal which is quite different from what I showed you before in the exposed individuals.

So, it seems that the current vaccines that have been licensed generate at least in our assays fairly limited immune responses, and maybe this antibody is involved in the HAI but we don't see even what you see in exposed individuals.

We did find at least two of the three epitopes in the NA overlapped the catalytic site. So, this vaccine does elicit probably some NA inhibiting antibodies as well.

So, we are actually at this point really ready to move on from this vaccine to the next generation of vaccine that was shown by different companies and other sponsors to have both increasing titer but it also improved quality in terms of breadth of neutralization and we hope that through collaboration with such sponsors we will be able to show whether or not we are able to see increased breadth of repertoire including large fragments in the receptor binding domain. So, I just want to summarize this GFPDL was successfully used to identify the epitope recognized by antibodies in avian influenza-exposed individuals and H5N1 vaccine recipients and in H5 human monoclonal antibodies and thought that this tool will be used to understand antibody-mediated correlated protection ultimately as an add-on assay and eventually this can help in the design of improved vaccines, and I really want to thank all of our collaborators. Surender Khurana was the lead scientist. Yonaira Rivera helped. Kanta Subbarao was very important in our collaboration, John Beigel and Julie Metcalf, Antonio Lanzavecchia for the monoclonal antibody. This is an ongoing collaboration with Arnold Monto and Suzanne Ohmit to help us to compare seasonal flu vaccine including the LAIV Bill Cerebri and our collaborator in University of Oxford, Cameron Simmon and we have started a collaboration on H7N7 with Adam Meijer.


DR. ZAMBON: I am going to suggest that we have one or two quick questions and then actually take a short break because we are rather far behind in this session.

PARTICIPANT: That was wonderful. In terms of looking at children's sera who responded to live attenuated vaccine we observed a very broad immune response in those children versus children given inactivated vaccine who have a very narrow immune response. I wonder if you have had an opportunity to look at those sera?

DR. GOLDING: No, but we will be very --

PARTICIPANT: It is a terrific tool to do that.

DR. GOLDING: Absolutely, yes.

PARTICIPANT: Do the expressed epitopes assume the native conformation of the marker?

DR. GOLDING: We think that they do for the very simple reason that we really found that, first of all we found out that if you use very short peptides or small fragments we very often don't get any binding and we really need to expand the size of the fragments and the fact that we are able to map three monoclonal antibodies, at least two of them requiring very large segments meant to us that yes they do acquire the conformation of a structure. Now, whether it will be true for every single antibody of course we will find out what the limitation of the system is. Then of course we will follow up by expressing these proteins either in bacteria or in mammalian cells and see that as a mini protein whether they can indeed be recognized by the same sera. So, there is of course a process but this is the first start that gives us at least sort of the window to know where to go.

DR. ZAMBON: May I suggest that we reconvene after a brief 15-minute break, so, if we could be back at five past eleven?

Thank you.

(Brief recess.)

DR. ZAMBON: Thank you very much for taking a reduced coffee break. I am sorry to move everyone along but just in the interests of trying to keep us on track this morning I think we need to make a start now.

I would like to introduce our next speaker, Walter Gerhard who is going to talk to us about M2-specific assays.

Thank you, Walter.


Agenda item: M2-specific assays

DR. GERHARD: On the first slide I would like to give you some general information about M2 influenza A virus. So, this is an non-binding glycosylated transmembrane protein of 97 amino acids that has a relatively small 24 amino acid long ectodomain and has been shown to form in its mature state homotetramers. It is pH-inducible proton transfer activity. It is known to be expressed at very high density in the plasma membrane of injected cells at the time of virus maturation but it gets incorporated at low density into the envelope of the mature virion and what is most important in the present context is that the antibodies that are directed to the ectodomain which I will refer to as M2e, restrict virus replication in vitro and also in vivo and this information together is information that can evolve from sequence analysis that M2e has remained surprisingly well conserved in human influenza virus strains but this leads to the possibility or the hope that perhaps M2e may be a vaccine or a target that can provide broadly protective activity and against influenza A virus. So, the question then becomes how should one measure this immunity. Unfortunately in this case there is no useful functional assay because the virus reduction or the virus yield reduction of applied size reduction is not specific for M2e. It can also be displayed by HA and NA specific antibodies.

So, that leads straightforward with us with binding assays and the most frequently used binding assay has been simply the peptides which correspond to the ectodomain and that peptide can be used on its own or coupled to a large carrier protein and this assay is very easy to perform but has the potential problem that this does not detect antibodies that are specific for determinants of the tetrameric, of the mature tetrameric M2e and so to, and so to develop also an assay that could detect that we thought that the best possibility would be to have a cell line that is stably transfected with M2e and many initial attempts to achieve such a stable transfectant with the standard adherent cell lines and using a strong promoter always failed. We got very good transient expression but we never could select a stable line.

We thought that this probably was because of toxicity of M2e expression to identify and so we turned to the HeLa Tet-on system. In this case M2e could be put under the control of a minimal CMV promoter. One has very low constitutive expression but its expression can then be up regulated by adding to the cells the tetracycline or doxycycline and here all lines be made what I will refer to as the M2 expressing HeLa-M2 and the control cell line which simply was transfected with MT expression vector.

So, in terms of the performance of the assay with these transfected here and these are M2 cells it was very simple. We simply seeded HeLa M2 or HeLa C10 cells into 96 well microtiter plates and we used half of the plate for the HeLa N2, half for the HeLa C10 and the cells were grown for 2 days in the presence of doxycycline to induce M2 expression and monolayers were then fixed with glutaraldehyde. They were blocked and the plates could be stored for several weeks in the cold.

After that at the time of use we used these plates in a totally standard ELISA assay, by simply titrating serum samples in parallel on HeLa M2 and HeLa C10 and then using the difference of optical density between the two cell lines to quantitate the concentration of M2e specific antibodies relative to purified and specific antibodies that were titrated in parallel on the same units. These were always our standards and an example of such an assay is shown here with mouse serum which was titrated on the HeLa M2 and on the C10 and the red one shows the delta ODs, the difference in ODs and the ODs at the trailing edge of this titration curve was then used to quantitate the adjuvanted concentration relative to purified M3 specific monoclonal antibodies.

We used the same assay for human sera except that in this case the standard antibody was chimeric antibody which had variable regions of a mouse MP specific monoclonal antibody and human concentrations and the assay was developed in this case here with a mixture of goat anti-human C kappa, C lambda which should detect all antibodies.

The mouse assay was developed with anti-C kappa which detects approximately 95 percent of all the antibodies.

Here I only want to use this slide to show that we used exactly the same assay for detection of peptide specific antibodies and so this was, in our case this was our specific antigen and this was the non-specific antigen.

So, with these two assays we could differentiate three major distinct and three specific antibody populations, antibodies that bound only to L2 peptide, antibodies that cross reacted with M3 peptides and HeLa 2 cells and then antibodies that only bound to the HeLa M2 cells. We called the latter ones M2e not for native. These M3 pep and these pep-not because they cross reacted and we could show that in mice that were immunized with M2 vaccine which we are using which is a synthetic M2 vaccine approximately 85 percent of the antibodies on average were M2-pep specific and approximately 15 percent were M2 pep-not specific, but the ratio here varies greatly amongst individuals.

In mice that were immunized by repetitive heterosubtypic infections and in this case actually three consecutive infections we found that those antibodies consisted of approximately 40 percent M2e pep-not specific and approximately 60 percent M2e not specific.

So, now in the following I want to show you some actual experimental data and in the first few slides I will show you that in mice that are immunized with this construct it is only these antibodies here which provide protection and so the experimental protocol was as follows. Vaccination with M2e by peripheral route subcu, im or ip two or three times 3 weeks after the last vaccination they were blastomized(?) to determine the antibody titer and a week later, they were challenged with influenza virus and in this challenge we used a very small volume of influenza virus so that we initiated a nasally-confined infection and we let this infection progress into the lower respiratory tract and 5 days later we determined how much virus is at the very site of the respiratory tract and what we found is this is one sector of the samples and I should say that each single indicates the virus titer on or off bases in the nose or in the lungs. The trait is not shown here, and the bars represent the mean, the degenerative mean within each group. So, we use now the difference in the mean, the reduction in mean virus titer between the control group and the immune group as a measure of protection in nose and in lungs. These are indicated by these red arrows and we wanted to see how does the M2e specific serum antibody titer, does it correlate with this protection, and what we found is there is absolutely no correlation between M2e pep specific antibodies in serum and the level of protection.

However, when we only looked at the M3 pep-not specific antibody titer then we saw that at each site of the respiratory tract we had a highly significant correlation and the relationship between antibody titer in serum and protection followed a sigmoidal curve in both sites and these are kind of strange looking sigmoidal curves but these are a few samples really and 50 percent effective dose in both cases was approximately around 20 micrograms per milliliter of serum antibody and now that was very promising. However, in these experiments we also used not only peripheral or parenteral immunization, we, also, immunized other groups of mice by the intranasal route and when we now looked at how these correlated, serum antibody titers with the protection in intranasally immunized mice we found there is no correlation and this is, all right, I must have lost something here. Okay, so, in that case there was no correlation.

So, we also used this assay to look at M2e specific antibody titers in human sera and this was done in collaboration with Bob Couch who provided the sera from 24 human subjects that were obtained during the acute and convalescent phase of a natural influenza virus infection and I am told that all these subjects made a significant HI specific antibody response.

When we looked at M2 specific and this is now an assay done only with HeLa M2 cells; so it detects antibodies that could also potentially cross with native M2. We saw that 11 made significant increase of antibody against M2 and 13 showed no increase whatsoever, but interestingly at the onset in the acute phase all patients, all 24 started very low. They had very low antibody titers indicating that also in this increase we saw here these antibody responses obviously are very transient and an antibody response which didn't lead really to a good memory phase and this is stated there and this is actually in marked contrast to the antibody response to another very conserved viral protein like MP where it is shown here titration of sera from healthy human subjects or 5 sera these sera had hardly detectible M2 specific antibody titers but as you can see they had very high nuclear protein specific antibody titers obviously from previous infections.

So, because we have to save time I was told I will go only to the last slide, the acknowledgement and I would like to acknowledge the people in my lab, Krystyna Mozdzanowska and three postdocs, Darya Zharikova, Jinqui Feng and Manxin Zhang who contributed to this study, also, Bill Wunner and Henry Hoff from the protein expression core facility of the Wistar Institute who made contributions and then Laszlo Otvos and two of his postdocs who did the peptide synthesis and then obviously I mentioned I would like to acknowledge Bob Couch for the collaboration in the human studies.

So, with this, thank you.


DR. ZAMBON: Thank you very much. We have time for one question.

DR. WEN: Juan Wen(?) International Vaccine Institute, South Korea. I just have one very quick comment that recently we used the recombinant protein consisting of the multiple M2e and the results obtained from ELISA using Z cell recombinant protein are very consistent with the assay using your sera expressing M2e.

DR. ZAMBON: With that I would like to introduce our next speaker, Dr. Janet McElhaney from both British Columbia and University of Connecticut to give the final talk of this morning's Session 3.


Agenda item: Assays to evaluate T cell responses

DR. MC ELHANEY: So, my whole goal of this talk is to Brian to turn his half dumbbell into a whole dumbbell on T cell responses and I just want to acknowledge to start off with the support that I have had through the development of this project and assay.

So, I just want to start out with to put this into context that right now we have an unprecedented opportunity to develop better vaccines and we clearly need these in older people.

This is my most dreaded complication of influenza. It is looking at this looking at risk for catastrophic disability. This means that you lose over the course of an illness three or more of your basic activities of daily living. This isn't about going to the grocery store. This is about bathing, grooming and toileting and quite frankly I would rather die of the illness than experience this. So, it is really, and most of the other things that are on this list along with pneumonia and influenza have actually been linked to influenza as well, and other studies by Bill Barker and Ann Falsi have also shown that in the context of flu illness people lose independence in activities of daily living.

This is a very worthwhile endpoint to be looking at and the reason for looking at outcomes that we are going to make a difference with new vaccines and I get concerned with the data I am going to show you about using antibodies for responses for accelerated programs for vaccine development for this population.

So, this is just a cartoon of what happens when we vaccinate. So, you will recognize this from other talks but the killed virus vaccines really stimulate T helper cells fairly well and B cells to produce antibodies but they don't stimulate cytotoxic T lymphocytes very well so that in the context of an influenza infection this is meant to illustrate an airway that what we might see in an aging person is that we tend to produce more of these TH2 cytokines and what we are trying to do with vaccines is stimulate more of a TH1 response so that we can help to turn on these cytotoxic T lymphocytes, and this is important because it doesn't matter which T helper cell response you get you still get antibodies but you don't get CTL response with a TH2 response to this virus.

So, those are the molecules we looked at and then just importantly at the time of activation of the CTL when it recognizes the virus infected cell these little granules migrate to the point of contact at the immunologic synapse and at the time of activation of the CTL one of these granzymes that is going to cross over is activated by terminal dipeptidylase(?) cleavage and through a process facilitated by perferin(?) it crosses over into the cell and generates an enzyme cascade that leads to apoptotic cell death of that virus infected cell, and so this is just data just to kind of prove the point around looking at antibody responses to influenza vaccines.

This started off with Doug Powers' study where he was looking in a cohort manner at the effect age in repeated vaccination and we found no effect of either of these two things on the response to vaccination in terms of HI antibody titers.

So, I went to Derek Smith and Walter Beyer who added a whole bunch of more subjects into this database so that we could ask this in a more robust way about the effect of age and repeated annual vaccinations. So, we had Doug's data from 1996 to 1999, and some of Walter's data that he has previously included in a publication and Derek has adjusted the baseline antibody titers in all of these subjects. So, we are starting with the same baseline and then calculating the number of twofold increases in antibody titer with each vaccination and what you can see for the H1N1 strain, the top are the young and the pink are the older and by the time you get to fourth vaccination the response to vaccination is identical in young and older people whereas if you looked at H3N2 and we had actually a hard time finding; we could only find older people who hadn't been vaccinated in the previous year where most of these young people had never been vaccinated with flu vaccine before but you can see by the second vaccination titers are, that the antibody response to vaccination is identical in this group and similarly for the B strains by the third vaccine identical antibody responses in terms of the number of twofold rises that we were seeing.

This is data, the graph on the left shows the antibody response to vaccination and then with infection in a group of young and older people most of whom were older adults and the yellow this is now a geometric mean titer pre-vaccination 4 and 10 weeks post-vaccination and then this year we had flu right after and right around the time of the 4-week post-vaccination point and you can see that antibody titers were not different in people who got flu, the yellow versus the turquoise and then this is their seroconversion to the infection.

The following year we had eight of the nine people who had flu in this study came back for the study as well as 68 of the people who never got flu along with an additional about another 70 new people into the study, and these are the people who had flu the previous year. Antibody titers start out high but in general their response to vaccination parallels that of the group who did not get flu in the study and importantly again we showed that just prior to the flu season these people who went on to get flu had virtually identical titers to the people who never got influenza in the study.

So, I am going to get you to follow the green, the turquoise and the yellow dots as I go through some of these other slides.

This is just the assay that we have developed so it is simply putting PBMCs into culture with live virus and harvesting them after 20 hours in culture. Our assay does not work. It doesn't correlate with protection if we take these out to 5 days and culture as many of the other cell-mediated immune responses as are measured at that time and I am going to show you some data that comes from ex vivo harvested lymphocytes in this assay, cytokines and supernatants for enzyme B in the lysates and a little bit of flow cytometry.

So, this is looking at, we looked at both interferon gamma and IL10 levels and I can tell you that the best thing is to look at the ratio because this is really about a regulation of these cytokine levels and remember on that cartoon slide we want to go to the left to increase the amount of interferon gamma relative to IL10 to protect against flu; go the other direction and we should have increased susceptibility and so we find that generally in people who don't get flu they don't really change this ratio much and this really an effective adjustment that I will talk about in a minute but the people who went on to get flu had low levels of this ratio prior to infection.

The concern about using cytokines is that these levels interact a lot with the medications that these people are on and we had people with heart failure in this study and that is measured by their performance on a 6-minute walk test. So, all this data has been adjusted for that and so that is not something that is very practical in a large trial in terms of trying to use this as a correlate of protection.

This is now 2 years' worth of data. So,this is the granzyme B assay. We simply make a lysate of those stimulated PBMCs.

We have a commercially available purified granzyme B standard that we can now run on every plate and that helps to standardize the units in the assay when this thing cleaves this substrate which has an aspartate residue here that is unique to this enzyme. The paranitroanilide(?) is cleaved and turned yellow and you simply read it in a plate reader.

Over two influenza seasons this is now looking at 209 people who never got flu in either influenza season compared to 12 subjects who got flu in the first year after their 4-week vaccination time point and eight more subjects. So, it was 12 in total who got flu after a 4-week time point in the study so we could evaluate their response to flu before they had infection and these folks have significantly lower levels of granzyme B prior to influenza infection.

So, the important point about this as well is that in case people are getting nihilistic about what the opportunity is to develop better vaccines through this population these are actually now the eight people who had flu in the previous year. So, they presumably had this flat response to vaccination in the previous year but you can see that by the next year you can really restimulate their CTL memory and this tells me I think that there is some capacity of this system to respond to differently formulated vaccines that might do a better job of stimulating CTL.

Oh dear, we have got a map problem here. So, some of the criticisms of this assay have been how do I know these aren't NK cells that are producing the granzyme. This is flow data to show that the granzyme B activity, this is using a single chain antibody that we got from Kevin Kane that only the T cells are actually up regulating granzyme B in their cytoplasm and not the NK cells and in fact there are very few NK cells that are expressing granzyme B in this assay and so this is actually the other assay that we did with the CD107B marker that Guus talked about this morning and we actually show that without virus stimulation low levels of cells expressing granzyme B and virtually none of them expressing the degranulation marker CD107B and as that is dramatically up regulated with a stimulation with the virus so, too, is that degranulation marker.

So, we think that these cells we are measuring something that has an effector function in the assay and for anybody that is interested I would be happy to show you on my computer what those flow cytometry data look like.

So, in summary I hope that I have presented some data that would raise concerns about using antibody titers as a sole measure of vaccine efficacy in this population. I think that we can use some of these T cell responses and make complement antibody titers in the evaluation of vaccine efficacy and I hope that I have convinced you that the aging immune system may senesce but memory is not lost, and would just finally a summary slide.

This really takes an enormous effort from a team of people that are not just the investigators but the on-the-ground folks who helped me out with this study at the University of Connecticut, some long-time collaborators at the University of Alberta who have done all the stuff with the granzyme B and Derek Smith and his assistants for the analysis of data and Walter Baer and most of all Doug for getting me going down this track.

Thank you.


DR. ZAMBON: We have time for one question, I think.

DR. COUCH: I first of all want to thank Janet for telling me that age is not necessarily a basis for immunodeficiency but I would just quibble with you a little bit. I think you are a little too strong saying that inactivated vaccine don't stimulate this response because you --

DR. MC ELHANEY: No, I am sorry. Let me correct myself.

DR. COUCH: There is a gradient there. There is no question about that.

DR. MC ELHANEY: Yes, they do stimulate some of the CTL response, certainly restimulate memory. So, thank you for that clarification, Bob because that is a battle I have been fighting for a long time is to convince people we can actually measure this response in the CTLs.

DR. ZAMBON: Thank you very much. I would like to thank all the speakers in this last session for some very interesting and comprehensive talks. I am sorry to hand over so far behind time but hope that we can catch up now and Kanta over to you.


Agenda item: Session 4. Evaluation of avian influenza vaccine efficacy -Moderator: Kanta Subbarao, MD, MPH

DR. SUBBARAO: Thank you. What we will try to do is try to get you off to lunch about 15 minutes late, hopefully not much longer than that and the session that I am going to talk about is the Session 4 which is on evaluation of avian influenza vaccine efficacy.

So, what we are talking about here is how efficacy of candidate vaccines has been or can be evaluated and what we are dependent on to link the immunogenicity data to protection and efficacy data is to depend on animal models. Clearly you have heard from the Session 2 about the evaluation of immuongenicity of these vaccines in clinical trials.

So, I am going to start with a very brief summary of the available animal models and then Jackie Katz is going to talk about the immunogenicity of efficacy data from the inactivated vaccine. I will follow with the talk on the immunogenicity and efficacy data from live attenuated vaccines that we have developed and then we will break for lunch and have three talks afterwards.

So, what I am going to start with is just a very brief overview of the available animal models that I assume Jackie and I will both be referring to and so I just wanted to summarize just to make sure everybody is on the same page some of the features of the mouse model.

As you heard yesterday we have learned an awful lot from studying influenza in mice. The human influenza viruses replicate in the respiratory tract of mice without prior adaptation but they rarely cause disease without prior adaptation.

Some avian influenza viruses including the highly pathogenic H5N1 viruses replicate efficiently in the respiratory tract and many of them cause disease without prior adaptation though the extent of disease can vary even among closely related viruses.

So, for instance of the 16 isolates from 1997, about nine are lethal for mice and the others are not lethal for mice. So, they all replicate well in the respiratory tract.

The receptor distribution in the respiratory tract of mice appears to be it is a little hard to find the exact citations for these but it appears to be alpha 2,3 specific in the ciliated epithelial cells and in type 2 alveolar epithelial cells and in studies that were published in the last 2 years H5N1 viruses appear to bind to trachea of mice.

Clinical signs of illness in mice include ruffled fur, hunching, labored breathing, hyperthermia, weight loss and mortality.

Intranasal infection under anesthesia results in viral pneumonia and you heard from Walter that you can infect with a very small volume and have an infection that is primarily restricted to the upper respiratory tract or you could use inhalation anesthetic or injectable anesthetic and get a complete respiratory tract infection.

Some influenza isolates replicate in extrapulmonary sites. The significance of this is unknown but in H5N1 infections extrapulmonary spread in mice correlates with lethality.

Doses of inactivated virus vaccines that have been used in mice range greatly from less than 1 microgram in some studies to 10 micrograms in other studies. They are usually administered with adjuvants such as complete or incomplete Freund's adjuvant, Ribi or AI salts and the dose of live virus vaccines used in mice range from 10 to the 5th to 10 to the 6th TCID50 of infectious virus.

The features of the ferret model are human and avian influenza viruses replicate efficiently in the respiratory tract of ferrets without prior adaptation. The receptor distribution in terms of alpha 2, 3 and alpha 2,6 appears to be similar to that seen in humans with alpha 2,6 primarily in the upper respiratory tract and alpha 2,3 in the alveoli. However, the avian influenza viruses administered intranasally will replicate in the upper respiratory tract as you can recover virus from the nasal wash in very high titer.

Clinical signs of illness include fever, sneezing, rhinorrhea and weight loss. Neurologic and gastrointestinal symptoms have been seen following infection with some highly pathogenic avian influenza viruses.

You see mild inflammation in the respiratory tract following infection with human influenza viruses and some influenza viruses replicate in extrapulmonary tissue at sites including the brain. The significance of this finding is not known.

The doses of inactivated virus vaccines that have been used in ferrets typically range between 7 and 15 micrograms. They are generally administered with adjuvant or in two doses.

In some much earlier studies the human influenza vaccines had to be administered with adjuvant as a single dose of an unadjuvanted inactivated vaccine typically did not elicit an antibody response and you needed two doses or vaccine with adjuvant.

The live virus vaccine dose is usually 10 to the 7th TCID50 and we found in recent studies that have just been published that the volume of inoculum should be .2 mls, not more. If you administer .5 or 1 ml you do get some virus going down into the lower respiratory tract.

There are studies that have been done in hamsters and guinea pigs and in cotton rat models and there are fewer of these studies than the mouse and ferret and I kind of smashed them all onto one slide.

Human and avian influenza viruses replicate efficiently in the respiratory tract of hamsters without prior adaptation. The clinical signs of infection are not observed with human influenza viruses. Some avian influenza viruses do cause lethal infections in hamsters and extrapulmonary spread of some avian influenza viruses has also been reported.

In guinea pigs human influenza viruses replicate efficiently in the respiratory tract of guinea pigs without prior adaptation. Avian influenza viruses have not yet been evaluated in this model. The Hartley strain guinea pigs are highly susceptible to the non-adapted A/Panama '99 strain. These are data from Peter Palese’s lab that were published. And infection transmitted to contact animals, clinical signs of infection are not observed. He does not report coughing or sneezing or rhinorrhea in these guinea pigs but pneumonia has been reported in the past with the Hong Kong '68 H3N2 infection.

Cotton rats. Non-adapted human influenza isolates replicate in the upper and lower respiratory tract following intranasal inoculation under light anesthesia.

There is an increase in respiratory rate which is a clinical sign of infection and histopathological changes have been observed in the lungs following infection. So, that is a short summary of those models.

Cats and non-human primates, and Osterhaus is not here to remind us. He wasn't able to join us at this meeting; so, this slide is for him on his behalf to remind you that they have done some studies in cats where intranasal infection with human influenza virus does not result in clinical signs of influenza but virus can be recovered from pharyngeal secretions and the virus can be transmitted to contact animals.

The H7 avian influenza viruses replicate in the upper respiratory tract but do not cause illness, but the highly pathogenic H5N1 viruses cause severely lethal infection after intratracheal inoculation or after the cats are fed on infected bird carcasses.

The receptor binding in the lower respiratory tract appears to be similar to that in human tissue and the highly pathogenic H5N1 viruses replicate in extrapulmonary sites including the GI tract and these viruses are transmitted to contact cats.

Non-human primates, there is more interest in studies in non-human primates recently although non-human primates -- human influenza viruses were evaluated in non-human primates for many years. Virginia Henshaw and Brian Murphy did many of these studies.

Inoculation by a variety of routes, intranasal and intratracheal appears to be necessary to establish an infection in non-human primates with human and avian influenza viruses. So, intranasal infection alone is much less reproducibly successful. You have to administer the virus intratracheally and with or without additional routes.

Some non-human primate species exhibit clinical signs of infection following intratracheal inoculation but the findings of clinical illness are not uniform.

Highly pathogenic H5N1 viruses cause disease in some cynomolgus macaques. So, I had actually two or three more slides that just outlined some of the advantages and disadvantages of these different models. Primarily the advantages of the mouse model are the wide range of strains and of mice and the wide range of immunological reagents that can be used to study them.

However, as we were reminded yesterday most of the laboratory strains of mice do not have a functional MX gene and many influenza viruses require adaptation in mice and the advantage of using ferrets is that they are naturally susceptible to influenza viruses and the clinical signs of infection are similar to human disease.

The problems are they are difficult to obtain or can be difficult to obtain seronegative ferrets especially in human influenza season. There are a limited number of suppliers and a limited number of immunological reagents although I believe that is changing and they have specialized housing requirements.

They may not be the ideal model to study temperature-sensitive viruses because they have a higher core body temperature and in terms of the non-human primates the advantage is that there is a closer evolutionary relationship to humans but the disadvantage is they seem to require high doses of inoculum and multiple routes of inoculation and they are very expensive and require specialized housing.

So, with that very brief summary of the animal models we will have Jackie Katz come and talk to us about the evaluation of immunogenicity and efficacy in inactivated vaccines.


Agenda item: Evaluation of immunogenicity and efficacy of non-replicating avian influenza vaccines following challenges with homologous and heterologous H5N1 viruses in animal models

DR. KATZ: With that introduction I can probably cut out several of my slides. So, thank you for that. So, I am just going to use this slide to orient you to the two models mainly that I am going to focus on in my talk and also to mention that obviously there is a lot of data in the literature about the mouse model in particular. There is less with the ferret with respect to pandemic vaccine evaluation.

So, I have really limited my talk to some of the work we have done and some of the work that Elena Gorbokova has done with the ferret model using fairly recent H5N1 vaccine candidates and in some cases H9N2.

So, Kanta has already gone over the pros and cons of these models and I am going to focus primarily on the mouse and ferret because these have been used as the challenge models and we do have good systems set up for looking at H5, H9 and H7 subtype challenges in these animals So, for the mouse model and the other thing I am trying to do today is to orient you to the different parameters that can be used to evaluate protection in either the mouse or the ferret model and I hope I demonstrate that some correlate better than others in the broad sense with immune responses.

So, typically we use inbred strains of mice. They can be Balb/C, B6 or B6-129. We have found that they all generally are equally susceptible although Richard Webby has demonstrated that there is a gradient of sensitivity with different inbred strains of mice, but the three key parameters that we look at are the body weight, so a highly virulent H5N1 virus will cause a decline in body weight and then a fairly rapid death of the animals out to day 7 or 9 whereas as Kanta mentioned some less virulent H5N1 strains in the mouse model may only cause transient weight loss and the animals survive, and then typically the third marker in addition to survival is viral replication, which can be addressed in the upper and lower respiratory tract, typically the lungs and other systemic organs since the highly pathogenic H5 viruses are generally spread systemically if they are highly virulent in this model and so the brain or other lymphoid organs such as the spleen or thymus can be used as a marker for spread of the virus.

So, back in 1997, when H5N1 first emerged in humans we realized that a surrogate H5N3, a low pathogenic virus was at that time the only suitable vaccine candidate and we heard yesterday that Cal Nicholson and Maria Zambon and Stevenson and others performed several studies in humans using a duck Singapore based vaccine.

So, we were also interested in how this vaccine could induce protection in the mouse model and we generated a whole virus formalin inactivated vaccine using purified duck Singapore virus and gave an equivalent of 3 micrograms of HA and this was determined through densitometric analysis of the total viral protein of the purified vaccine and we found that in fact when we didn't administer any adjuvant that we got fairly poor and this is traditional serum HI antibody titers; we got fairly low titers with one dose. They were boosted somewhat with two doses but if we added alum adjuvant we got a substantial boost to response and although alum adjuvant wasn't used in the H5N3 clinical trial MF59 was the adjuvant, these results reflect in general the need for two doses or an adjuvant in particular to boost responses.

When we looked at challenging giving these animals a lethal challenge with one of the 1997 viruses and duck Singapore was fairly well antigenically matched with the 1997 viruses and we found that after two doses of the vaccine alone we got lower serum antibody, HI antibody titer and neutralizing antibody titer and when we added alum we got much higher titers as you saw in the previous slide but in either case although we saw fairly different levels of antibody here all the animals were equally well protected from death and it is only when we went to look at the weight loss which isn't shown here and I will show in another slide but the actual viral titers that we could see a difference. Although this wasn't statistically significant in this case there was a trend whereby the lower antibody was also associated with a higher viral titer. So, whereas higher antibody was associated with total clearance and this is really just looking at one time point but total removal of viral load in the lung at that time.

We were also interested to better understand what appeared to be the correlate of serum HI antibody protection and this was before the horse hemagglutination inhibition assay. So, we were doing this with turkey erythrocytes.

So, this is probably under represented and would be higher now if we used horse HI assay, but what we did here was we had animals that were infected with the live either H5N1 virus that showed lower virulence in the mouse model or the surrogate duck Singapore virus which also infected and replicated well but didn't cause lethal disease in this model, and you can see that when we had higher titers, titers of 40 or greater we got essentially complete protection from lethal challenge with again the 48397 strain whereas titers of 20 or lower were generally associated with a lack of survival.

Moving on we tried in subsequent assays to look at the comparison, and this is now moving forward several years where we now have heterologous viruses, Vietnam 1203 clade 1 vaccine, but we again challenged animals with either in this case one dose of the duck Singapore vaccine or the vaccine plus alum and also in another group of animals had a whole virus inactivated vaccine produced with wild-type Hong Kong 213 2003, and what we are looking at here is the antigenic distance between Hong Kong 213 and Vietnam 1203 that is obviously closer than the duck Singapore vaccine and whether we could reflect that in a challenge model, whether our correlates of what were our, what was the parameters whereby we could detect a difference in this antigenic difference. So, again here this is showing, this is microneutralization on our horse HI antibody to the, not to the homologous but to the heterologous Vietnam 1203 virus and you can see that the duck Singapore vaccine didn't really stimulate a titer. With alum it was fairly weak as you might expect with the antigenic distance but when we got to the Hong Kong 213 with alum we got quite a robust HI and microneutralizing antibody response.

So, then we challenged these animals with a high dose of Vietnam 1203, and we found again that even though there was a substantial difference in antibody response in the different groups all the animals were equally well protected from death, and we had to look at other parameter such as weight loss to actually show a significant difference and here you can see that control animals had a mean maximum weight loss of 19 percent whereas the group that had a high antibody had no weight loss and I realize here that there is some inequivalence here with the antibody but there is a general trend whereby the animals that really had no antibody response were showing more weight loss than those that had some level of antibody.

So, one other thing we looked at because it seemed again that we weren't completely picking up the total spectrum of antibody responses possibly in the serum, we did ELISA using either recombinant H5HA or recombinant N1NA, and again you can see that where we have a vaccine with a matched N1 neuraminidase with alum we were detecting some anti-neuraminidase response although I would guess that since this is an H5N3 this represents a level of cross reactivity in this indirect ELISA, but more importantly we were also seeing IgG directed against the HA and we routinely see this and I will show you another example where we see IgG responses when we don't see HI or neutralizing antibody responses.

So, moving to another study where we looked at the H9N2 vaccines we produced the vaccines in the same way and as you will recall from some of the work I mentioned yesterday these two H9N2 groups were circulating in the late nineties in chickens in Hong Kong and elsewhere and these two groups referred to as the G1 group; this was actually the virus isolated from a human case and the G9 or Y280-like viruses were genetically and antigenically distinct.

So, in this study we were interested to see what level of cross protection we would see with the two vaccines.

So, we again generated a whole virus vaccine formalin inactivated and delivered it either without alum or with alum and similarly for the G9 vaccine candidates and here you can see using the HI again we saw the good homologous titers of antibody induced by the Hong Kong 1073, some level of cross-reactive antibody but when we challenged these animals the animals were completely protected from the homologous strain and showed some limited level of virus replication but were also essentially protected from the heterologous strain.

H9N2 viruses don't cause lethal disease in mice. So, this was really what we, we were really only able to address viral titers in this case and when we looked at the G9 vaccines they also elicited very substantial homologous antibody titers but the reciprocal cross reactivity was very, very low and this is also what was seen in reference ferret sera that it appeared to be more of a one-way cross reactivity between these two genetically distinct groups, and so these animals when challenged were all infected and showed some level of reduced, but nevertheless showed some level of reduced viral replication indicating that again there was some degree of protection although clearly not as good as homotypic vaccine.

So, the point I wanted to make here was when we also looked at the IgG anti-HA IgG response here we did see a substantial response, so indicating that there is antibody being made but it is not being picked up in this case by the HI titer nor was it picked up by the neutralizing antibody assay in this case.

So, to summarize the mouse studies quickly it looks like in general we do see the dose response enhanced immunogenicity with adjuvants that generally reflect the observations of immunogenicity seen in humans.

However, protection against homologous and heterologous challenge strains are best assessed by determining reduction of viral replication or a level of morbidity such as weight loss and that protection from death is really not an optimal parameter in this mouse model.

We can discriminate relative strengths of heterotypic immunity in this case and the level of protection we see is not always apparently correlated with the serum antibody detected by the functional HI or neutralization assays because we do see protection reduced viral titers in the absence of detectible serum anti-HA antibody by these assays and so we need to look further and use other antibody assays and other cell-based assays and look further and have better methods for looking and evaluating the effects of say the antibody to neuraminidase or other anti-HA but non-neutralizing IgG.

So, I am going to move to the ferret model and again we have multiple parameters where we can assess the vaccine efficacy. For the highly virulent H5N1 viruses, they can cause lethal disease and morbidity that again can be measured by weight loss, sometimes fever although we don't use it because it can be very variable and lethargy which can be measured by a documented index score and then again virus replication in the upper nasal passages.

This is an example of the kinetics of weight loss that we see in animals infected with highly pathogenic viruses. I cut and pasted here. So, we are really only looking at H5N1 infected ferrets, not H3N2 and you can see that the viral replication again varies depending on the virulence of the strain with the more recent viruses giving very high titers that don't drop down at all.

So, I am moving now to some work because there really hasn't been that much work in the literature done with the ferret and in evaluating current vaccines and that is going to be my message at the end of the talk but last year Elena Govorkova and her colleagues at St. Jude used the ferret model to evaluate their RG Hong Kong 213 PR8 reassortant vaccine and they also generated a whole virus vaccine that was beta-propriolactone inactivated but the one caveat that you have to understand in looking at this work is that they used animals that were antibody positive for H3 and in some cases that might confound the cross reactivity that we see in these results. So, they used either two doses of seven or a 15-microgram dose with alum or two doses without alum and you can see that they were robust responses, neutralizing antibody and HI titers achieved after just a single low dose with alum and the titers with two doses without alum were even higher. I made a typo there. I made the talk in a great hurry last night. Sorry. So, this is actually, this value is around 100 and this is around 200.

So, there is a very substantial boost with the second dose in the absence of alum and again these titers appear very high and I wonder about whether there is some heterotypic protection in this situation, but nevertheless they challenged the ferrets with 10 to the 6th egg infectious doses intranasally and then looked at the level of nasal wash viral titers achieved and you can see that even in the presence of this substantial pre-challenge antibody the titers have only been reduced about 100-fold compared to the controls by day 3 and really what you see in the ferret is earlier enhanced clearance by day 5, but you do see that the viral titers again generally with the higher antibody titers there is far less virus at the early time point in the nasal washes.

So, they then looked at what would happen with a heterotypic challenge with Vietnam 1203 2004, and first of all they looked at the antibody profile and the upper level is the HI antibody response. The lower is the neutralizing response and interestingly here the neut is quite a bit more strain or is recognizing earlier Hong Kong 156 better than it recognizes the 213 and the bottom line here is that for either the HI or the neut only the group of animals that got two doses of the 7 micrograms in the absence of alum did they see a heterologous antibody response to Vietnam 1203, and so then they went on to challenge these animals and even though these animals did have pre-existing or had been infected with H3 they nevertheless all succumbed to infection. All of the animals, three out of three animals died and showed substantial weight loss and as you can see had high viral titers over the 3 to 7 days where it was examined whereas regardless of the vaccine protocol all of the animals that received any vaccine were completely protected from death and however there was somewhat of a gradient with the weight loss and with the ability of the animals to clear the virus with the animals that had the highest antibody titers being able to, having less virus at the later time points of day 5 and day 7.

I think I am going to skip over that. Am I running short of time? Okay, well, let me just quickly show you another study that Terry Tompy did in our group. This is again with an H9N2 vaccine. This was done with the Novivax virus-like particles here. He inoculated ferrets intramuscularly, boosted them at 3 weeks and then challenged them again with the 1073 '99 H9N2 virus.

Again, this isn't lethal for ferrets and so what we are looking at is the viral shedding and also the antibody response which is shown in this slide, and you can see that in the animals receiving either 1.5 or 15 micrograms you can get a dose-dependent increase in HI antibody response but when you look at the ability of these animals to be protected from viral replication you can see even though they had very robust responses that at day 3 there is really not much clearance but where significant efficacy comes in or protection from viral replication is really by day 5 and day 7 where groups that received either the 1.5 or the 15 micrograms were protected equally well.

So, in ferrets again there seems to be a dose response to vaccination. You can enhance with adjuvant but it is difficult to reduce the early virus replication in the upper respiratory tract for ferrets even in the presence of strong serum antibody responses, and I believe this has also been seen with seasonal influenza virus challenge of ferrets, and the effect you primarily see is in enhanced clearance and I wonder whether this is really because we are dealing with the upper respiratory tract, the nasal turbinate and perhaps IgA is really playing more of a role there and the inactivated vaccines that we are testing aren't doing a very good job of inducing IgA.

So, we do see protection against homologous and heterologous challenge and again protection from death is not the optimal parameter because you can see protection from death over a wide range of antibody titers, and so looking at reduced viral titers and changes in weight loss might be a better way of assessing the protective effect of pandemic vaccines and we do see it in a pet on occasion in the absence of detectible serum anti-HA antibody just as we do in the mouse.

In this the caveat here is that the animals were H3N2 primed. So, it is not clear to what extent there is some perhaps heterosubtypic protection going on but clearly it also demonstrates that there are other protective parameters or correlates that we really need to be assessing in this model.

I will just finish up by saying that, well, I think we have heard all that from Kanta; so, I will just move on to the conclusions, and just want to emphasize that obviously animal models are necessary to assess immunogenicity and protective efficacy of pre-pandemic avian influenza vaccines and they are tremendously useful for proof of concept novel vaccine strategies and here the mouse really is an optimal model for screening for vaccine strategies and optimizing strategies before you get to perhaps the ferret as the next tier of evaluation, but in both models protection from death is really not an optimal parameter for protection and we need to do better than that and really look at extensively as much as possible with extended kinetics to really understand how the vaccines are able to reduce the viral load or prevent infection in these animal models.

So, ferrets can be used to establish a protective efficacy of vaccines in humans but as far as I know this has rarely been done. There was some discussion with a study ongoing with HHS that this will be done in parallel with a Phase II portion of a human vaccine trial and I think that is a very good approach and one that perhaps we or other people and partners should be thinking about that once you have optimized your vaccine and you are testing it in humans the only way we are really going to know what the level of protection is is if we can use this animal model to better understand and link to the immune correlates of protection. I think more work still needs to be done and hopefully there will be improved reagent availability for the ferrets and that will improve this model. We will be able to perhaps look at the simple T cell-based functions but nevertheless any vaccine that is going into clinical evaluation we should consider using the ferret as a preclinical model to at least link vaccine efficacy to what is seen in the clinical trials and I will just acknowledge particularly Xiuhua Lu who did all of the mouse work and Terry Tumpy and Jessica, Neal and Taronna who were involved in the ferret particle work and our colleagues at Novivax and other colleagues who provided the viruses for us.

Thank you.


DR. SUBBARAO: We have time for a few questions.

DR. KILBOURNE: I am going to utter a squeak in defense of the mouse. I think we have got to be careful about our terminology here and before we dismiss the fact that pulmonary replication occurs with almost influenza A strains we have ever looked at, what is usually is not sought because we are looking for easy parameters like death which is a definitive endpoint but as you pointed out over and over again is just not useful, I think there is a problem in confusing the ability of a mouse to acquire lobar pneumonia and demonstrably pneumonic consolidation with whether the virus is replicating there and if one really looks and takes out the lungs, grinds them up and measures virus titer then you will find that almost all influenza A strains will replicate in the lung of a mouse particularly if ether anesthesia or perisal(?) transmission is the method of infection. Intranasal infection I would call -- PARTICIPANT: Please identify yourself?

DR. KILBOURNE: What is my name? Edward Kilbourne, and I am from home.


DR.KATZ: I would just like to respond to those nice comments, Ed. I don't want to give the impression that we shouldn't use the mouse. I think it is a good model and obviously there are many things as Kanta pointed out that you can really dissect the immune response and it is really our only tool for really addressing the contribution of different effectors and I think that is very important.

DR. WOOD: David Wood from the World Health Organization. I think one of the challenges that we are facing at the moment as we go forward is as we have vaccines being donated or put away in stockpiles and trying to assess the continued relevance of the vaccines that have been put in stockpiles against drifted strains how do you see the animal models or do you see the animal models being useful in that context and if so what would be the best models and how best to use them?

DR. KATZ: That is actually what Shualu was trying to do with the mouse model and I think it potentially could be used and the idea was to have mice that were vaccinated with the current pre-pandemic vaccine or you could vaccinate them with a spectrum of the stockpiled vaccines and then assess as new variants come up what the level of protection was and I think that is definitely something that should be considered and whether it should be done in the mouse which would be a simpler approach and use larger numbers is possible. The ferret is more difficult because of the numbers and the scope of the work but I think that looking at the potential for the different candidates that we have to protect with contemporary H5 is one thing the animal models can do.

DR. REED: Steve Reed from the Infectious Disease Research Institute. One thing about the mouse is it allows us to evaluate state-of-the-art vaccine formulations and whether or not the pathology is perfect, I mean obviously that is a situation to debate but as far as the ferret goes what do we know about the expression of TLR receptors on cells from these animals because some of the new formulations with include compounds that stimulate through TLR, and I imagine even with the limited immune reagents that are available one could address that in vitro. I just want to know if there is any work being done on that.

DR. KATZ: Yes, that is a great question and as far as I know there isn't any work being done right now although there is a big push I think and NIH has recently funded a consortium to sequence the entire ferret genome and so the hope would be that in the very near future there would be the genetic information readily available to make suitable primers and that is already happening for a limited number of cytokines.

DR. REED: In the leprosy field we had a similar situation with armadillos, but we didn't have to go through sequencing the whole genome. We just made degenerate primers and some of the cytokines can be found quite readily.

DR. KATZ: Exactly and that is the approach that has been taken with some of the cytokines to date. So, it is a good point.

DR. SUBBARAO: I have one comment and question for you, Jackie. With the data where you find that despite the fact that you detect neutralizing antibody or HI antibody in ferrets but you still see a fair amount of virus in the upper respiratory tract, that actually fits with what we heard yesterday about the gradient of being able to protect the lower respiratory tract better. How much of an effort has been made to look for virus in the lower respiratory tract as well? Should that be a better outcome measure?

DR. KATZ: It could be. Again, it is a hard thing to do and it uses more animals of course but because you can get the titers in the nose without euthanizing the animals but I think there should be more of an effort especially with the highly virulent H5 challenges to do that. Good point.


Agenda item: Evaluation of immunogenicity and efficacy of live attenuated avian influenza vaccines efficacy following challenge with homologous and heterologous H5N1 viruses

DR. SUBBARAO: I am going to talk now about essentially a very topic but looking at the live attenuated vaccine and hopefully we won't have any Mac to PC problem.

I am going to start with two slides from a study that we did that essentially supports what Bob Couch and Brian Murphy told us yesterday about how anti-hemagglutinin antibodies are protective and so this is a study in which we took human monoclonal antibodies that were generated from convalescent, from PBMC from convalescent individuals from Vietnam that Hana Golding told you about that Antonio Lanzavechia immortalized and in the first study, these were studies done in mice we tested the ability of these monoclonal antibodies to protect mice from infection.

So, this was a prophylaxis study. The top row, D2.2 and A146 are irrelevant antibodies. They are generated by the same technology against other pathogens and then we tested the 5.1 and 3.14 which are two of the antibodies that Hana talked about and administered the antibodies intraperitoneally to groups of mice prior to infection. So, we administered the antibodies IP and then the next day challenged the animals with the virulent virus and you are looking at Kaplan Meier survival curves where on the Y axis you are looking at proportion of mice surviving over time and you can see that the mice that received the irrelevant antibodies succumb to infection between day 5 and 8.

The sheep antiserum is actually a hyperimmune anti-H5 antiserum made by hyperimmunizing the sheep with baculovirus expressed H5HA and so, all the mice that received the hyperimmune serum survived and with one of the two antibodies we don't even see a dose response; even at the lowest dose all the mice survived and with one of the antibodies we see a step-wise dose response, and one of the things we looked at was what the correlate of the survival was with virus replication in the lungs and with virus titers and I am just showing you one selection. This is the paper that was already published and just looking at the histopathology, so looking at H&E or immunohistochemistry in the lungs of mice on the left panel that received the irrelevant antibody and on the right panel the mice that received the H5 specific, one of the H5 specific monoclonals and what you can see with the H&E staining is that there is more pneumonitis in the mice that received the irrelevant antibody and the brown staining is the antigen positive cells and in the mice that received the H5 monoclonal there are antigen-staining cells but they are largely lining the larger airways and there is much less antigen in the parenchyma of the lungs compared to the mice that received the irrelevant antibody.

So, I would just like to add to what Jackie said in addition to looking at virus titers in the lungs because when we ground up these lungs we really didn't know what the distribution of the antigen was and so the virus titers were about a log and one-half apart, statistically significant but there was still a lot of virus there but by looking at the immunohistochemistry we could see where the distribution of the virus was and we did pulse ox on some of these and if you do pulse oximetry on some of these animals you may actually detect a difference that correlates with where the virus is.

So, that is just something new that we are thinking about adding to our studies. We also looked at the efficacy of these monoclonal antibodies following the onset of infection. So, this was in a treatment strategy and so the mice received the antibodies 24, 48 or 72 hours after the onset of infection. So, the mice were infected intranasally and then 24, 48 or 72 hours later they were given the irrelevant antibody on the top row and the H5 specific antibodies in these others and you can see that the irrelevant antibody did not protect the mice at all but surprisingly even when the monoclonal antibodies were administered 72 hours after the onset of infection we saw survival.

So, this was a study where we only looked at survival and we really didn't look at virus titers. I suspect there was virus present but they did survive infection.

So, this was just to reiterate the part of the message yesterday that the HA specific antibodies are protective in these models

So, now, I will switch to talking about what we do for the most part which is the development and evaluation of vaccines and this is the collaborative research effort with Medimmune vaccines.

So, we are using the internal protein gene of the Ann Arbor cold-adapted virus and we put in the hemagglutinin and neuraminidase genes of different viruses.

So, one of the first things we do when we make these reverse genetics-derived viruses is to evaluate their attenuation phenotype and if the HA is derived from a highly pathogenic avian influenza virus like an H5 or H7 virus then we evaluate the attenuation in chickens. Otherwise we look in mice and in ferrets. In mice if the parent virus is lethal we look at two measures, lethality and virus replication in the respiratory tract and in ferrets we use the convention that Medimmune has already established with the FluMist and that is to look at virus replication in the respiratory tract at day 3 post-infection.

So, what we found when we made three different H5N1 cold adapted virus candidates, and you heard about the clinical data on two of them yesterday, the 2003 and 2004 but we had also made a 1997, and we gave these viruses to David Swain at the Southeast Poultry Research Laboratory and when these viruses, the wild type or the cold-adapted viruses are administered to chickens intravenously eight out of eight birds die if they get the wild-type virus within a day or two but none of the cold-adapted viruses killed the chickens.

It turns out of course that these are temperature sensitive viruses and they didn't actually even replicate in the chickens. The body temperature of chickens is about 41 degrees, but it is reassuring to our agricultural colleagues that these viruses would not pose a threat to poultry.

So, we then evaluated the level of attenuation in mice. So, these are viruses that are administered intranasally to lightly anesthetized mice and then at day 2, 4 or 6 through 12 and I have collapsed the data from day 6 through 12 because we didn't detect virus and the mice, you are looking at the mean virus titer in log to the base 10 TCID50 per gram of tissue in the lungs, nasal turbinates or brain and the mice that are represented in the blue bars received the wild-type virus and the mice that are represented in the yellow bars got the cold-adapted Vietnam 2004 virus.

So, you can see that there is a significant reduction in the level of replication of the cold-adapted virus in the lungs and in the nasal turbinates and there is no systemic spread. So, there is no virus detected in the brain.

We, also, looked at how these viruses replicated in ferrets and I apologize this is a glitch in the Office 2004 software. These two yellows should be just slightly a different color. So, the blue bar represents the 2004 wild-type virus. Next to it is the 2004 cold-adapted virus and this is the Ann Arbor parent. This is the 1997. The red bar is the 1997 wild-type virus. It is cold adapted reassortant and the Ann Arbor cold-adapted parent and these are ferrets that received 10 to the 7th TCID50 or virus intranasally and then on day 3 they were sacrificed and we looked at virus in the nasal turbinates, brain and lungs.

So, the cold-adapted viruses do not replicate in the lungs or in the brain and they are significantly reduced in replication in the nasal turbinates.

So, these viruses met the criteria of attenuation in ferrets, mice and in chickens. So, based on those data we moved forward and evaluated these viruses in mice and in ferrets. In immunogenicity studies we looked at one dose or two doses and we have tested the sera against homologous and heterologous viruses.

So, this is just a sample of some of the data that we have. So, we gave the mice when we immunized mice with the 2004 cold-adapted virus or the 2003 cold-adapted virus in a single dose, the 2004 cold-adapted virus did not elicit neutralizing antibodies even tested against the homologous antigen whereas the 2003 cold-adapted virus actually did elicit some neutralizing activity against the homologous 2003 strain.

This is consistent with what Rob Webster has published and has been saying that the 2003 virus is more immunogenic and possibly more sensitive.

As Jackie pointed out when you actually, if you don't look at neutralizing antibody and look just for binding antibodies, so if you look at ELISA antibody against the recombinant H5 hemoagglutinin in the absence of neutralizing antibody you can still detect quite a significant ELISA response but the ratio of the titers stays pretty much similar to the boost that you see within the neutralizing antibody titer with the second dose.

So, we then wondered what would happen if we gave these two doses in terms of the breadth of protection and so these are, you are looking at the geometric mean titers against the wild-type 1997, 2003, 2004 or Indonesia virus if you administered one or two doses of each of our cold-adapted vaccines and there is a lot of information here but what is highlighted in yellow are the titers against the homologous challenge virus and what we found was that two doses of vaccine in mice elicited a very significant boost in neutralizing activity and in two of the cases there was a fairly broad cross reactivity. So, the 1997 virus actually did elicit some cross reactivity and the 2004 cold-adapted virus certainly did. The 2003 cold-adapted virus actually elicited much more of a strain-specific response than the other two.

So, we looked at whether you really needed two doses or whether you just could wait, give one dose and wait 8 weeks. Maybe it was the kinetics of the response.

So, in this experiment we gave mice a single dose of vaccine and waited 8 weeks or we gave them a dose of vaccine and tested them 4 weeks later or gave them two doses of vaccine and you can see that even if you just waited 8 weeks without the second dose of vaccine there is a gradual increase in the neutralizing antibody response and we see this with HI as well although there is a boost with the second dose and we are seeing this pattern with H7, with H6 and so on but the kinetics of the response seem to be different than we are used to seeing with the H1 and H3 viruses.

So, we are seeing peak antibody responses at day 42 rather than at day 21 or day 28. Now, when we looked at the serum-neutralizing antibody responses in ferrets we gave again in the same structure, we gave the 1997, 2003 or 2004 cold-adapted virus in a single dose or in two doses and here you see that there is a neutralizing antibody response that is detected with the second dose of vaccine but 2003 was by far much more immunogenic than the other two and the 2004 even with two doses just barely elicited a detectible neutralizing antibody response.

So, when we look at efficacy we evaluate the efficacy of these vaccines in mice and in ferrets. We evaluate the efficacy following a single dose or two doses and then tried to use homologous and heterologous challenges. If a lethal challenge is relevant we do that and if not we look at pulmonary virus replication and systemic spread of the challenge virus.

So, you have heard before a single dose of any of the vaccines, the cold-adapted 1997, 2003 or 2004 cold-adapted viruses when mice were challenged with 50, 500 or 5000 LD50 of the homologous or heterologous virus all the mice survived and this is following a single dose in the absence of detectible neutralizing antibodies.

However, these mice do have ELISA antibody. It is just not detectible neutralizing antibody. So, when we looked at these data we thought that suggested that it was a cellular immune response that was protecting these mice from lethality but we depleted CD4s and CD8s in both CD4 and CD8 positive T cells and they still survived.

So, then we went back and transferred serum from these mice to naive mice and although the serum did not have any detectible neutralizing activity it does confer protection.

So, we are looking at the mechanism by which this is occurring. It appears to be antibody. It appears to be something that either is below the level of sensitivity of our neutralizing assay or it is non-neutralizing antibody or complement may play a role. So, there are a number of possibilities that we are currently evaluating.

As Jackie pointed out in the studies that she summarized, instead of looking at lethality as the endpoint if you look at the level of virus replication and here we are looking at level of virus replication in the lungs of mice following a single dose the blue bars represent mice that were mock immunized. The yellow bars are the mice that received our vaccine and if you challenge mice a month after a single dose of vaccine in the absence of neutralizing antibody there is a statistically significant reduction if you just focus on the part that is boxed in red that is the homologous challenge.

There is a statistically significant reduction in virus replication but it is still 6 logs of virus. So, although that translates to protection from lethality there are still 6 logs of virus in the lungs, but after 2 doses when you do detect neutralizing antibody there is a very good correlation with the complete protection from pulmonary virus replication and this was true with homologous or heterologous virus challenge. So, the blue bars again are mock-immunized mice challenged with the 1997, 2003 or 2004 or Indonesia 2005 strain and all of the vaccines following two doses provide complete protection from pulmonary virus replication of homologous virus or heterologous virus.

We looked at the efficacy following two doses of cold-adapted vaccine in ferrets and again the blue bars represent the mock immunized animals. The red bars are ferrets that received the 1997 vaccine. The yellow bars got the 2004 cold-adapted vaccine and two doses of vaccine provided complete protection from pulmonary virus replication with homologous or heterologous challenge viruses.

However, there is still virus replication in the upper respiratory tract and I should point out that this was a study where detected virus in the brain but this was a time when we weren't separating the olfactory bulb from the rest of the brain and in subsequent studies when we separate them the virus that we see is in the olfactory bulb and presumably is just extended up from the nose. There is no virus in the rest of the brain.

So, you saw these data yesterday from Ruth where despite all of this evidence from preclinical studies that we have got viruses that are infectious, immunogenic and protective in the animal models when we went to clinical trials and these are just data from the Vietnam 2004 cold-adapted virus we really saw poor infectivity and poor immunogenicity.

So, what have we learned? The kinetics of the neutralizing antibody response to the H5 viruses in mice are slow. The magnitude of the neutralizing antibody response in mice and in ferrets was poor. When the vaccines elicit a robust neutralizing antibody response in mice and ferrets then the antibodies tend to be cross reactive and the vaccines are cross protective. So, I think there the magnitude of the antibody is generally important. There is a caveat in that even though a single dose of the 2003 vaccine was more immunogenic than the other two it tended to give you much more of a strain-specific antibody response and those animals were not protected from cross challenge.

The changes in the H5 hemagglutinin therefore from our data on the preclinical data on immunogenicity and efficacy would suggest that the changes that are occurring in the H5HA over these 10 years may not be driven by positive selection by antibody pressure but they may be changes that are accumulated as a result of the replication of the virus in a large number of hosts and it argues, at least our preclinical data argues against the need to have each clade-specific vaccine generated. At least in the preclinical studies if you get a good enough antibody response you will get cross protection.

A vaccine that elicits a sufficiently robust neutralizing antibody I just said appears to provide cross protection.

The poor infectivity of the cold-adapted H5N1 viruses was not predicted from the preclinical data and in data that I haven't shown you these viruses replicate very well in embryonated egg, in MDCK cells, in ferrets and in mice.

So, the cold-adapted virus replicates well in all of these systems. We are now evaluating human airway cells and other tissues to try to find the model in which we could predict the outcome that we saw in the clinical trials, but the poor infectivity of these live attenuated vaccines may in fact correlate with the lack of efficient person-to-person transmission of the wild-type H5N1 viruses in the world right now.

So, this is just a snapshot of the H5 data. We have a program in which as Ruth pointed out we have made vaccines against H7, H9 and we have also made some H6 vaccines and we have evaluated each of these in a very similar strategy in in vitro as well as in immunogenicity and efficacy studies in mice and in ferrets and to just give you a short overview of how some of these other subtypes are behaving the H7N3 vaccine which was based on the British Columbia strain from 2004 when we gave this virus to mice in a single dose or two-dose schedule we found that actually at 4 weeks following a single dose unlike the H5N1 the H7N3 viruses did elicit a neutralizing antibody response. So, they are not quite as poor in eliciting the neutralizing antibody response as we saw with the H5 viruses and then we see a very significant boost with the second dose and over time.

In ferrets if we give a single dose of the H7N3 vaccine and look at the response by HI or by neutralization at day 28 we detect a much more robust HI or neutralizing antibody response than we did with the H5 viruses and we get a boost with the two doses and the British Columbia virus is the North American lineage virus and so we tested the sera from ferrets that were infected with our cold-adapted virus against the Netherlands 2003 wild-type virus to see how much cross reactivity there would be and there is some cross reactivity by HI that we are not detecting by neutralization.

Now, when we evaluate the protective efficacy of the British Columbia H7N3 virus in ferrets the immunizing dose is 10 to the 7th. So, the bars in blue again represent the mock immunized animals just as they were before and you are looking at the virus titer in the nasal turbinates, lung, brain or olfactory bulb. Here we are separating the two and the yellow bars without the shading are ferrets that received a single dose of the H7N3 vaccine and the shaded yellow bars got two doses of the H7N3 vaccine.

So, you can see that even a single dose of the H7N3 vaccine provides complete protection from pulmonary virus replication with the homologous virus and provides some protection from heterologous virus challenge with the Netherlands virus.

There is a less complete protection from virus replication in the nasal turbinates but there is complete protection from replication in the brain, and we have done similar studies with H6 viruses. We have developed three candidate vaccines with the H6's. We have got an H6N9 duck virus, and H6N1 teal virus and an H6N2 mallard virus and in cross challenge studies in ferrets the red bars represent the unimmunized or mock immunized ferrets. The yellow bars are animals that receive the duck cold adapted vaccine. The blue color got the teal vaccine and the green color got the mallard vaccine and you can see that the only vaccine that provides complete cross protection in the lungs is the teal virus. So, that is our sort of lead candidate for the development.

So, just to summarize all of these data candidate live virus vaccines have been generated against four avian influenza virus subtypes. They are attenuated in chickens, mice and ferrets. In terms of immunogenicity two doses are required to observe consistent serum antibody responses in mice and ferrets against the H5N1 viruses but one dose was immunogenic for H7N3 and H9N2 viruses in mice and ferrets.

In protection or challenge studies we have used mice and ferret models to test against homologous challenge as well as heterologous challenge and in mice a single dose protects against lethal challenge but as Jackie pointed out that really doesn't fit with the level of virus replication. There can still a lot of virus in the lungs. So, we also do not recommend lethal challenges as an outcome to follow although it can be certainly one of the parameters you look at, but when one dose of a vaccine and this is the neutralizing antibody response that protects well from pulmonary virus replication and in other cases two doses may be required and in ferrets we found the single dose of all of the vaccines that we have tested provides protection from pulmonary virus replication but even after two doses you do not completely abolish replication of the wild-type H5N1 virus in the upper respiratory tract.

So, all of this is part of a very ambitious program to try to develop a library of vaccines against viruses of each subtype.

We develop the vaccines at the NIH in collaboration with MedImmune Vaccines and then proceed to clinical trials to evaluate the safety, infectivity and immunogenicity of these vaccines in healthy adults.

The intent is to actually bank sera from the vaccinated volunteers to have on hand and I think this will be a very valuable part if we have viruses that are immunogenic to test against avian viruses as they emerge in the world. The program as I said was a CRADA with Medimmune Vaccines and the clinical trials have been done in collaboration with the Center for Immunization Research at Johns Hopkins University, and I will stop by acknowledging all the wonderful people in my lab who do all this work and Jack is dressed up the way we do to work on highly pathogenic viruses. He helps us with all of our animal studies. The work is done in LID and I have acknowledged Medimmune and the Southeast Poultry and Johns Hopkins University.

The monoclonal antibody work was done in collaboration with Antonio Lanzavechia, Cameron Simmons and Jerry Ward.

Thank you.


DR. KEITEL: Thank you, Kanta. I have two questions, Wendy Keitel from Baylor College of Medicine. First is have the vets requested that you challenge chickens by another route in view of the fact that this is a temperature-sensitive virus, in other words oral or --

DR. SUBBARAO: Right. They have not asked us to do it but David Swain did that anyway. He gave the virus intranasally as well and the virus does not replicate even when delivered intranasally.

So, there is no lethality in the chickens when the virus is delivered intranasally but the chickens obviously survive and then when they are bled and we look for serologic evidence of infection there is none.

DR. KEITEL: The second question I will preface with a comment because it didn't come up yesterday, but it relates to a presentation that was given at WHO by Larisa Redenko using an LAIV based on a Russian construct and she used, and now this virus differs in many ways from the viruses that you are working with. It was an apathogenic H5. It was an H3 neuraminidase and it was the Russian backbone but despite all the differences when she inoculated human volunteers she had good levels of virus replication and good immunogenicity. So, I just want you to comment or maybe Ruth would like to comment on the differences between the US LAIV and the Russian LAIV and the potential contribution of neuraminidase antibody.

DR. SUBBARAO: I will start. Why don't you come to the microphone, too and see if you have anything to add. Wendy, as you pointed out there are a number of differences in the backbone of these viruses. The Leningrad strain and the Ann Arbor strains are quite different. They used a non-pathogenic H5N2, a Potsdam as I recall, N2 I thought.




DR. SUBBARAO: So, that is one less difference. So, I really would like to know. We would certainly like to have a virus that replicates well but I don't know, I can't really comment on the specifics of why that might be.

DR. KARRON: I guess the only other thing I was going to say is that my understanding is that there are two viruses that are currently being used in Russia as live attenuated vaccine. I believe this backbone is from the less attenuated of the two, is that correct, the one that is usually given to adults? Yes.

DR. SUBBARAO: Laszlo, do you have something to add?

DR. PALKONYAY: That is correct. There are two viruses in the different passage line but actually today the vaccine which is approved for children and adults is the less, lower passage level one. So, the higher passage level one is not used anymore.

DR. WOOD: David Wood, WHO. Kanta, both yourself and Jackie have stressed that reduction in virus titer is sort of the outcome measure that you would like to see in the animal models but how much reduction in virus are you actually looking for? Are you in agreement as to how much reduction there should be? Is it complete abolishment or what are you actually looking for?

DR. SUBBARAO: The greater the reduction the better. So, I am delighted if I see no pulmonary virus replication but I think that is a gradation. I mean I think historically we used to look for a 2-log reduction and I am not sure if Brian wants to comment on where the 2-log reduction came from but with viruses that elicit a good neutralizing antibody response we have been seeing complete lack of pulmonary virus replication.

Brian, do you want to comment?

DR. MURPHY: Brian Murphy, NIH. In humans these viruses go to 6, 7 logs. If you reduce it by 10-fold you will be changing 105 fever to 104 fever which is not really tremendous but a 10-fold reduction in the mouse will completely prevent the, will allow them to survive. Okay, so that is really not a useful amount of reduction. You need to go down around 10 to the 3 to 10 to the 4 in order to get a successful level of protection. You want to reduce a human infection from 10 to the 6 or 10 to the 7 or 10 to the 5 down to that level below which symptoms are seen which is 10 to the 2 to 10 to the 3.

DR. SUBBARAO: So, Brian, you are suggesting at least a 3-to-4-log reduction?

DR.MURPHY: Yes, that is what I would say, 3-to-4-log reduction would be something in the ballpark of what you are aiming for.

DR. TOPHAM: Dave Topham, Rochester. Have you done any studies to look for durability of the protection that you get in the animals? So, when you do your challenge studies after vaccination if you wait 3 months or 6 months is it maintained?

DR. SUBBARAO: Yes, we have not done those studies. We have done essentially studies where we have challenged a month after one dose or a month after two doses, but that is something that we could do. These studies are difficult because they are in by containment and so at some point we should do those to just see how long the protection lasts. We don't know that.

I want to thank you very much for your attention. We will stop now and have you come back at two to restart the session. We have got three speakers and then a panel discussion.

(Thereupon, at 12:50 p.m., a recess was taken until 2 p.m., the same day.)


DR. SUBBARAO: Thanks for returning on time. What I would like to do is start the second half of Session 4 and we are going to have the first talk by Sue Epstein from the FDA who is going to talk to us about heterosubtypic immunity to influenza A.



Agenda item: Heterosubtypic immunity

DR. EPSTEIN: I want to thank the organizers and in particular Kanta, for inviting me to speak here and as she said I am going to be speaking about heterosubtypic immunity to influenza A and I would like to start by saying that that the background on heterosubtypic immunity and its characteristics has already begun due to the help of Robert Couch, Brian Murphy and Jack Bennink.

I want to thank them for covering some of this and I also want to point out that let us take the limitations they were alluding to as givens, that this form of immunity is not going to meet the expectations for an HA vaccine. It is not going to be as powerful and complete a form of immunity but I will be discussing what is the potential of it; how could it make a contribution and how can we improve this form of immunity compared to the induction referred to.

So, the issues of potency and duration definitely come up. Then I will be discussing induction by natural infection and by various vaccine types and the implications for immune correlates or surrogates that might be relevant to this type of protection.

So, this immunity will involve not just the outer proteins but a variety of components of the virus. I will be focusing on NP and M2. M2 extends through the viral membrane and a variety of components can induce heterosubtypic immunity. It is defined as immunity induced by virus of one influenza A subtype or its antigens that can protect against virus of another subtype and this has long been studied in animals starting with the work of Schulman and Kilbourne in the 1960s.

This has already been discussed, that it is not sterilizing, permits some replication in most cases although it can be close to sterilizing in some mouse strains. It accelerates clearance and reduces morbidity and mortality. A variety of conserved antigens can be recognized and protection is seen against various influenza A subtypes. It does not require the presence of cross-reactive serum antibody.

I want to highlight one experiment I feel is quite elegant. This is that this immunity is specific to influenza A at the effector stage. There is no control of bystander influenza B virus during a mixed challenge infection. This was done by Walter Gerhard's lab. So, it is not just an innate response or local release of interferon, cytokines or something of that nature. It is a receptor-specific response.

First just to give a little background on natural infection in humans and what we do know and don't know this is data from the 1957 pandemic which was a shift of both the hemagglutinin and neuraminidase from H1N1 to H2N2.

Some participants in the Cleveland family study were monitored through epidemics of influenza A in Cleveland in 1950, 1951 and 1953, and all of this is culture confirmed influenza A in laboratory tests. They were all seronegative to H2N2 prior to the pandemic by serological testing. What you see is in blue those with no prior episodes during those years and in yellow those who did have an episode.

You can see in children there was no alteration in susceptibility during the pandemic by in the case of children the single exposure they had ever had.

In the adults with a longer history of potentially more exposure there is first of all a substantial difference between children and adults which could have this immunity as a basis and in addition those with a recent prior exposure have a reduced susceptibility despite being exposed to active cases of pandemic influenza in their own households but the numbers are not sufficient for statistical significance.

Another study at the time by Slepushkin looked at a much larger population. This is 15,000 people but without the laboratory virus testing of the Cleveland study. So, this is based on symptoms only and an episode in the spring could have been H1N1 or could have been other respiratory viruses.

What you see is that there was a highly significant reduction in susceptibility during the summer and the fall waves of the pandemic due to an illness that was described as flu-like in the spring, so, a recent exposure. People often ask if there is immunity of this type why are there even pandemics and not everybody has had a recent exposure, a recent influenza infection even in the past few years.

So, this is highly significant but again is just descriptive. Historical data cannot answer this question by itself.

To move on to animal models in mice which are the models studied most often heterosubtypic immunity is a multifactorial form of protection. There can be a role of antibodies, of IgG or IgA subtypes or T cells of various subclasses depending on the antigen used and the form of immunization.

The anatomical site is very important and as has been known since the 1950s and a variety of labs have contributed to study mucosal immunization and the differences in the upper and lower respiratory tract and to save time I am just acknowledging the many contributions of people who are here and others briefly this way.

The mediators have been studied by adoptive transfer of T cells, passive transfer of antibodies for in vitro treatment of virus before use and challenge by antibodies, foster nursing, depletion of T cells or NK cells and the use of knockout mice that lack a variety of components, IgA, all immunoglobulins, cytokines, receptors, beta-2 microglobulin and so on and clearly there is a role of T cells in some situations especially for NP. Clearly there is a role of antibodies in other situations including mucosal immunization and ABCC has been reported in one instance.

This is not just a phenomenon in mice. Other animal models studied have included chickens, ferrets, pigs and cotton rats. Depending on the system the parameters monitored include viral replication which is reduced, viral shedding, morbidity and mortality and histopathology and just to give a couple of examples that are particularly powerful in a study of ferrets earlier clearance was shown by viral shedding in animals that were heterosubtypically immune and this protection persisted at 18 months. Then in a study of chickens from the Webster lab H9N2 protected against H5N1 and it was shown by elegant adoptive T cell transfers in inbred chickens that this was T cell mediated.

You have already heard today about a variety of materials that can induce some form of broad immunity. This can include inactivated virus given intranasally especially with adjuvants, cold-adapted viruses and then a variety of experimental vaccines which can include proteins, peptides, conjugates and vectored vaccines.

In the case of NP Jack Bennink already mentioned that antibodies do not transfer protection but T cells do especially CD8 cells and in the case of M2 a variety of constructs can induce protective immunity, fusion proteins, multiple antigenic peptide constructs, peptide conjugates and in this case serum antibodies do transfer protection.

Here is an example using cold-adapted virus from my lab and the acknowledgements I will not go through to save time but they are indicated in green. Cold-adapted H1N1 protected against H3N2 challenge and vice versa. Cold-adapted influenza B virus, however, did not and I think this is an important type of control to indicate the specificity of the response. It is not just that anything protects against anything. This is two doses of vaccine given intranasally.

Moving now to an experimental form of vaccination DNA prime and adeno boost in a study with Gary Nabel's lab if we used control constructs induced by influenza B nuclear protein and compared to influenza A nuclear protein we can observe the following T cell responses. These are cytokine producing T cells after immunization with DNA plus adeno or just DNA or the irrelevant constructs.

You can see there is no response to the irrelevant constructs. There is a response to DNA alone in both the CD4 and CD8 compartments but it is greatly enhanced by adeno boosting and this is the response to a variety of influenza NP peptides.

We then challenged animals vaccinated in that way in collaboration with Terry Tumpy using H5N1 viruses. The animals received either prime boost to B&P or to A&P and you can see that the animals receiving the flu B control constructs lost weight very rapidly, a measure of morbidity and went on to die.

The A&P construct the animals lost weight only transiently and recovered, went on to survive in the case of Hong Kong 156 and a subset of them went on to survive in the case of the even more virulent Hong Kong 483.

This is challenge 5 months after adeno boosting and I will mention as an aside that in other examples of heterosubtypic immunity we have gone out as far as 8 months to a year and seen protection.

Going on to another conserved antigen, M2 there is a consensus sequence shared by most viruses of the circulating human subtypes, H1, 2 and 3. Other viruses can differ at a single position or multiple positions including there is an H1N1 virus with multiple differences. So, we used that as a model for divergence as well as looking at H5 sequences which diverged as well.

Shown here are immunizations with KLH conjugates of the M2 ectodomain peptides and the antibodies by ELISA did cross react in some instances with the M2E sequence from an H5 virus. So, that was encouraging about the potential breadth of the protective immunity. Based on that encouragement we went on to use the DNA prime adeno vector approach because that would use the entire M2 gene and provide potential helper sites, other sites that might be active in different MHC types.

Here is challenge of mice receiving DNA plus adeno boost to the M2 consensus sequence. They are then being challenged with FM, a virus that differs by three amino acids. It is almost as divergent, well, as divergent as some H5 isolates and a bit less than some others and it is quite a virulent challenge.

What you see is that the control animals receiving irrelevant constructs succumb, but the M2 immunization was protective.

We further found that this type of protection in a PRA challenge example was passively transferred by serum. On the other hand T cell depletion during the challenge period abrogated much of the protection. So, it is a combination of antibody and T-cell mediated protection. It is not solely mediated by antibodies to the peptides.

We then went on to challenge mice that had been immunized again to M2 with an H5N1 virus in collaboration with Terrence Tumpy. The mice are receiving B&P controls that are irrelevant, A&P constructs as a positive control or the M2 consensus sequence construct as a test and they were monitored for morbidity, mortality and lung viral titers. The flu B controls have high levels of virus in the lungs and they go on to succumb, and P reduced the titers in the lungs significantly and the animals all survived and M2 reduced the titers in the lung a bit less but significantly and again the animals went on to survive.

So, this is as before an imperfect form of immunity but has a large impact. Given that multiple antigens can induce heterosubtypic immunity one can imagine that a multi-antigen cocktail might be a desirable vaccine. You could induce a wider variety of immune effector mechanisms, T cell, B cell, whatever. It would reduce the probability of escape mutations and it would reduce the chance that people of a variety of HLA haplotypes might be non-responders. There would be probably something there for everyone.

A possible disadvantage would be antigenic competition or interference. We are now making comparisons in the lab of a variety of vaccine types and combinations. The most potent heterosubtypic immunity is not necessary achieved by just giving everything, giving all antigens. So, the best candidates can be chosen by comparison.

Immune correlates are a bit more complicated in this situation than if we know that HA is the only active moiety. So, for the general case of vaccines of different types serum HAI or neutralizing antibodies can be involved, respiratory tract IgA, cytokines, T cells, ABCC combinations and so on. This is just a conceptual diagram of the idea that a response can be mediated by one major type of immunity, by multiple types that are additive and there can be a threshold of protection that is exceeded by a single response or it may require multiple responses to achieve that and rather than just add up one, two, three and four in all cases clinically for practical reasons we could measure certain activities in priority order. This is a little bit analogous to in HIV testing not doing a Western blot on everybody.

So, suppose the vaccine warranted looking at levels of HA specific antibody. That might be the first test and if those levels were protective and confirmed if necessary the testing might be done. If not one could go on to the next mediator suggested by the content of the vaccine, for example, M2 antibody, is that protective alone or in combination with the above? If not maybe certain T cell responses; so the algorithm would be a way of testing in priority order for relative responses.

To summarize a variety of vaccine types can induce heterosubtypic immunity in animals that can be long lived, reduces morbidity and mortality and reduces viral load, can involve T cells and/or antibodies. Protection can work against challenge with strains of various subtypes including some H5N1 strains.

Historical data suggest this immunity in humans may alter susceptibility during a pandemic though those historical results are not definitive and so these findings and the gaps in them call for comparison of vaccines in preclinical models selecting the best candidates to be examined in clinical trials and for surveillance of at risk populations.

The public health implications are described here. This is not intended to replace an HA-based vaccine or an HA-containing vaccine. This would be a first line of defense to reduce morbidity and mortality despite the imperfect immunity.

The idea would be to prime in advance and that combination vaccine could contain some hemagglutinin but suppose we have guessed completely wrong; suppose as someone mentioned earlier in the meeting H5 is not what ends up spreading widely. We could prime in advance. In the event of an outbreak of an unexpected strain or a pandemic we could boost in high-risk areas and this immunity is intended to be augmented by antigenically matched vaccines whether inactivated or live attenuated when sufficient supplies become available.

I just want to acknowledge the past and present members of my lab and the past and present collaborators in particular Terrence Tumpy did all the H5N1 challenge studies. Julia Misplon and Chia-Yun Lou participated in all these studies, Mark Tompkins especially in the M2 study and thank you for your attention.


DR. SUBBARAO: We have time for a few questions.

DR. MURPHY: Brian Murphy, NIH. In those studies where you looked at reduction of replication in the lower respiratory tract you have information on reduction of replication in the upper respiratory tract rather than just the lower respiratory tract and is there a difference; is this immune mechanism operative equally in both sites in the respiratory tract?

DR. EPSTEIN: In the H5N1 study we have not done that. We have done it previously in other mice and Walter Gerhard's lab has done that.

We did see a reduction in nasal virus if we looked at general heterosubtypic immunity induced by previous infection.

Walter Gerhard had shown the mediators in the upper and lower respiratory tract can differ and change somewhat with time. So, the dominance of CD8 cells versus CD4 cells differs but there is control of viral replication in both areas.

DR. SUBBARAO: Since this mechanism is not specific to a single subtype how would you see implementing this into seasonal vaccine so that it really isn't even a pre-pandemic strategy; it is just routinely part of immunization?

DR. EPSTEIN: If we knew enough to optimize the vaccine formulations that give potent heterosubtypic immunity the annual routine vaccination could include whatever strains were selected and then with a heightened and focused response to say NP and M2. The difference between what I was showing you, the data I was showing you and simply cold-adapted vaccine is that if you narrow the response and focus it on certain antigens by the DNA or adeno vectors you seem to enhance the potency of the cross protection, well of the heterosubtypic immunity. The term "cross protection" has been diverted to use for two different uses.

So, it is premature. This will be a speculation but suppose you prepared your annual vaccine of whatever type you wanted but you spiked it with something that would greatly enhance the response to conserved epitopes that might appear in any subtype.

DR. SUBBARAO: Any other questions?

Thank you.

So, we will go on to the next talk and that is by John Treanor who is going to talk to us about challenge studies in humans.


Agenda item: Challenge studies in humans

DR. TREANOR: Thanks for the invitation. I am going to very briefly review sort of the potential utility of a challenge model in humans for understanding correlates of protective immunity.

So, when I talk about a challenge model the concept basically is that we are going to administer something to human subjects in order to stimulate a response which we hope mimics what would happen if those subjects had actually been exposed to the real influenza virus and so the utility of this model is very dependent on how well whatever stimulus we are using really mimics the natural infection and this is one of the limitations of the model that we can discuss but having said that challenge models in humans have been used in many different ways specifically for influenza including establishing the etiology of the syndrome of flu as being due to influenza virus many, many years ago when it was noted that transmission from infected ferrets could induce the illness in laboratory workers. It has been used extensively particularly by Brian Murphy in assessing the relative attenuation of candidate live vaccines by comparing their level of replication and symptom induction in human volunteers with that of a corresponding wild-type virus, for proof of concept studies looking at both antiviral agents and as well as vaccines and immunomodulators like interferon and their ability either prevent or treat influenza infection in this sort of challenge model and to make measurements of the kinetics of the immune response or other types of responses that could be useful in developing computer models of the immune response or other things and generally in the context of vaccine studies in a limited way to try to look at what would be correlates of immune protection at least in the model

Now, what I put together here and it may be difficult to see from the back of the room, it is just sort of a conceptual schematic of what is involved in a human challenge model and really sort of the schematic that could be applicable to any sort of study of this type.

We begin with subjects who are in an immune state which we might call immune state zero. Oftentimes this involves selection of subjects from the general population because they possess some characteristic of the immune system typically susceptibility to the challenge virus that we desire and then we are looking at that measuring any number of outputs which would be measurements of the immune system. These individuals may be then subjected to an input or an intervention of some type which might be the administration of a vaccine, passive administration of antibodies or immunomodulators or antivirals and then of course in this scenario we are evaluating sort of the safety of that intervention and its effect; what would be optimal dose for achieving what we are looking for and pharmacovigilance and after this intervention is applied people are moving into a second immune state which you might call immune state one, again characterized by measurements of nasal or serum antibody, measurements of cellular immunity or innate immunity and then subsequently after they achieve this state we input some sort of challenge material which is designed to tease the immune system and see how well it is working and that challenge material might be a wild-type virus which had been produced under the right kind of conditions to be able to be administered to humans or it might be an attenuated virus or some other kind of system that we are using to challenge the system perhaps even inactivated antigen and individuals respond to this challenge with an output of various things including evidence of infection in the case of live viral challenges, the development of illness, production of virus and perhaps other responses like cytokine responses and then following this challenge they move into some new immune state which you could call immune state two where there has been a response of the immune system to that challenge and we can set a model up depending on what we are looking for to really use any kind of input and look at what this does to the various immune states before and after challenge.

Now, a couple of things to bear in mind specifically with the use of this model in humans. One is if we are talking about seasonal influenza and we are doing studies in adults we recognize that all of the subjects have some level of pre-existing immunity to influenza. This is because obviously they are being exposed to influenza on a yearly basis and so in order to achieve infection if we are using an infection model it may be necessary to screen the subjects and select them based on their apparent susceptibility to that particular virus and this is typically done by measuring serum HAI antibody and probably is the most direct demonstration of the role of HAI antibody in protection at least in this model in that it is quite difficult to experimentally infect individuals if they already have high levels of HAI antibody to that particular antigen.

In some instances it may be better to use neutralizing antibody because particularly for example for influenza B there is good evidence that neutralizing antibody is more sensitive and is a better predictor of susceptibility to influenza B and in theory you might use other types of immune measurements to select susceptible subjects.

This would clearly not be as important if we were talking about a pandemic virus although we have learned that there may be unexpected cross reactivity in individuals who in theory should never have been previously. exposed to a pandemic virus.

Then having selected the subjects input to that is the challenge intervention and most of the time this is a wild-type influenza virus or it might be a live attenuated vaccine, but I think we have to bear in mind whether or not the responses that we are going to get are going to be valid for the intended use of the model so that if the model is intended to look at whether or not an intervention reduces symptomatology it would be necessary to use a challenge material that induced symptoms and things like that.

This relationship between susceptibility and dose I think is important. One can push the system to make people have higher infection rates by using a . higher dose or make antibody levels more important by using a lower dose and things like that and then the third thing about this model to bear in mind is that in many circumstances the model is intended to induce illness in the subjects. I think this is a very important thing to bear in mind because we are talking about studies that are non-therapeutic by their nature in a scenario where we are exposing subjects to risk and so that risk/benefit ratio has to be considered very carefully when using a challenge model for any purpose of proof of concept.

The clinical outputs if the model is intended to induce clinical output could include both objective measurements like viral titers and physiologic measurements and a large component of subjective measurements of symptoms which can be difficult to validate just to keep that in mind and likewise the immune outputs may or may not be well validated. This is an issue that has been discussed already.

Now, here is a typical result of challenge of susceptible volunteers with a wild-type virus. This is a study done by Fred Hayden and Steve Strauss a number of years ago in which individuals were selected based on having low levels of HAI antibody to the A/Texas '91 virus. It was the H1N1 virus. So, subjects were then inoculated with a specific strain A/Texas 3691 at a dose of 10 to the 5th TCID50 intranasally and this is a composite of data that was presented in this paper and in order to put everything on the same graph I have eliminated the actual Y values but you can see here that challenge is associated with the development of virus shedding with the peak viral titers achieved on day 2. This would be very typical and declining viral titers over the next 7 days so that by day 8 after challenge essentially none of the subjects are still shedding virus.

This is accompanied in a very direct way by increases in symptoms scores and a subjective measurement of illness, symptoms typically consisting of upper respiratory tract signs and symptoms, a minority of subjects experiencing fever or systemic signs and here is one example of an objective measurement of illness induction which in this case is the total nasal mucus weight by day and you can see how these things tend to track together. Fred and Steve also did a very thorough analysis of both systemic as well as nasal secretion cytokines and you can see particularly an association with the development of symptoms you see increases in levels of intranasal interleukin 6 and interferon alpha and these correlated well with the levels of viral replication as well as the level of clinical symptoms.

Now, the model has been used in many ways, one of which is to predict the efficacy of vaccines. The models have been predictive for the efficacy of both vaccines as well as antivirals.

This is a study that we did a number of years ago. It was an incredibly complicated study probably overly complex study design with many faults but the concept was to immunize volunteers with a trivalent cold-adapted influenza vaccine and then challenge them with monovalent wild-type pools representing H1N1, H3N2 or B viruses that were present in the vaccine in order to get a preliminary idea of the vaccine efficacy.

Now, there were a lot of technical problems with the study. It didn't work out very well but what we did see is that both the cold-adapted vaccine as well as trivalent inactivated vaccine were protective in this model with slightly better protection seen with the inactivated vaccine although the differences were not statistically significant and I will just point out that very, very similar results were recently reported by Sue Ohmit and Arnold Monto in a field trial with essentially similar design where inactivated vaccine and live vaccine were tested in adults against placebo and you can see again actually quite similar estimates of the protective efficacy of the vaccine and not to say that this proves that the model is a good one but it is suggestive that the model can be predictive of results in the field.

Now, in this study that we did we had an opportunity to look then at the relationships between pre-challenge antibody levels that is post-vaccine and pre-challenge and you can get a feeling from this about the kind of measurements that can be made.

So, this graph looks at the pre-challenge level of nasal secretory IgA which is measured in an ELISA. It is not a very easy to standardize ELISA. In this particular case these are units and the pre-challenge serum HAI and in placebo recipients you can see that the infected subjects tend to be those who begin the study with relatively low levels of serum antibody and among those with low levels of serum antibody there is a suggestion that perhaps the level of nasal antibody is slightly lower in those who are protected although that was not significant.

In individuals who receive the cold-adapted vaccine you can see a little bit more of this effect of nasal antibody and a protective effect of serum antibody when comparing those who ended up being affected with those who were not and individuals who received the inactivated vaccine have higher levels of antibody prior to challenge and you can see that in general those who ended up with infection were those who had the lowest levels of serum antibody prior to challenge, but the other thing I would point out is that at least in this particular study and I think it is true in other studies as well these are not black and white differentiations so that there is not a level where you could say that 100 percent of the subjects below this level are infected; 100 percent of the subjects above this level are not infected, but if you look generally at the rates of infectivity in those who had low levels of both serum antibody and nasal antibody and compared that to either those who just had nasal antibody you can see a significant reduction or compared to those who just had serum antibody you can see a reduction and those who had a positive test prior to challenge for both serum and nasal antibody had the lowest rates of infectivity, sort of confirming that either one of those arms of the immune system could be useful in preventing infection and this is very, very similar to data that was generated by Mary Lou Clemments and Brian Murphy many years ago looking at challenge this time with one of two viruses, A/Washington, an H3N2 virus or A/California, an H1N1 and what they found interestingly was the correlates of immunity to infection with the challenge virus differed slightly depending on what the source of that immunity was. So, you would get slightly different answers if you looked at individuals who had acquired their immunity from natural infection than you did in individuals who had been vaccinated with inactivated vaccine or who had received a live vaccine and in addition to looking at nasal antibody and serum HAI antibody they also showed that serum neuraminidase-inhibiting antibody could be protective against infection and illness in the challenge model.

Now, the other thing that has been looked at I think Guus Rimmelzwaan already showed this slide is that the model has also been used to show the potential protective effect of cellular immunity.

So, this is a study that was done with the emergence of H1N1 viruses back in the late 1970's and these individuals were challenged with an A/Munich '79 H1N1 virus in I think 1979 or 1980, when the H1N1 viruses had first emerged and this looks at the results in 22 subjects who did not have detectible serum HAI or neuraminidase inhibiting antibody at the beginning of the study and you can see here that if you look at the pre-challenge level of cytotoxic immunity as evaluated by a chromium release lysis assay you can see a correlation between having higher levels of cellular immunity and relatively lower levels of virus shedding with the highest levels of virus shedding seen in those with the lowest levels of immunity.

So, the models also demonstrated that. Now, clearly there are circumstances where it is not feasible to challenge individuals with wild-type viruses. It would be particularly true of children and so in that circumstance there has been some information that has been generated by simply using live attenuated vaccines as a challenge model.

So, this is a study that was done by Bob Belshe and colleagues that looks at children who had either received the live attenuated cold-adapted vaccine or who had received placebo who were then challenged with a monovalent cold-adapted vaccine with the idea of seeing whether or not the trivalent vaccine could prevent infection with the cold-adapted vaccine and you can see here this shows the percentage of subjects who were shedding on each day after challenge, and you can see a significant reduction in the proportion of individuals who shed virus in the vaccine recipients compared to placebo recipients.

Now, Brian had asked earlier about the proportion of individuals who shed virus in children with the live vaccine sort of in comparison to what was seen with the H5.

In this particular study 24 percent of children who are on average about 4 or 5 years old, 24 percent of these children shed the cold-adapted monovalent virus on at least one day following administration.

Now, again, Bob used this model to look at what the correlates of immunity were in individuals who had received previous placebo whose immunity presumably exists because of prior natural infection and whom we feel probably had not been exposed to H1N1 viruses for at least two previous years because they had not circulated.

You can see a very strong effect of serum antibody as well as a somewhat lesser effect of nasal antibody in that almost all of the infections that were seen were in those who did not have detectible serum or nasal antibody.

In a much larger number of previous vaccine recipients it is a little bit more complicated but I think you can see that having relatively high amounts of nasal IgA provided protection against challenge.

Now, the interesting thing about this was that if you simply looked at those individuals who had neither serum nor nasal antibody there was still a protective effect, about 50 percent decreased shedding of the challenge virus in those doubly negative individuals comparing the vaccine recipients to placebo.

So, there were additional protective effects induced by vaccination which were not measured well in this study but perhaps were reflected in neutralizing antibody because it looked as though those individuals also had more neutralizing antibody.

So, how would this be used for pandemic vaccines? I think that is a very big question which I don't have a good answer for. I think if Ruth had presented a study that showed that the cold-adapted H5N1 vaccine were shed in reasonable amounts or by a relatively large proportion of the subjects you could contemplate using that virus to tease out correlates of immunity with all of the caveats that it is only a poor mimic probably of what really happens with natural infection, but if you had a suitable live attenuated vaccine that had enough infectivity in replication and still maintained an acceptable safety profile and perhaps the A/Leningrad-based reassortants would meet this, then you could use the model potentially to validate some of the concepts that we are working on right now about the relationship between antibody and protection, whether we are overestimating how much antibody we need or potentially underestimating and in addition a model like that could be a convenient way to explore novel immune mechanisms but again I think I would say first of all that such a model really isn't possible right now because there is no suitable live attenuated vaccine that could be used in this way and even if we had one I think you do have some significant limitations when you start using an attenuated virus in trying to guess particularly if you are looking at specific cutoff numbers what would be required for protection against an actual wild-type challenge but with those caveats I think you could learn something about what you need to induce with a vaccine if this model could be developed sufficiently.

So, I will end with that.


DR. SUBBARAO: Any questions?

DR. COUCH: Just a couple of quick remarks, John. As you know I have cut a lot of my research history on volunteer challenge studies including influenza but probably more rhinovirus than influenza.

I think it is very useful. You just have to understand and accept the limitations of the model, and I think you have done very well on that.

There is a couple of things that I think do need to be pointed out. The challenge model as we use it and as it was developed is an intranasal challenge. So, unlike a lot of the mouse and ferret data where you are using volumes and you are challenging both the upper and the lower respiratory tract and we are looking at lung virus and that is sort of an out but here you are challenging the nasopharynx.

So, if you are looking at an immune mechanism you are asking for that immune mechanism to be able to function efficiently at that location.

So, you have to keep that in mind with what you have learned from the challenge model.

Second is that you can, I think, do this safely you see and the reason for saying that is we did it in 1968, when H3N2 came in as the new pandemic virus and we did it within individuals that I now know in retrospect were because of the nature of our neutralization test, we always selected for neut negative not HI negative but let us not go into why that is a bit better for selection. We were selecting with an assay at that time that in addition to selecting for no anti-hemagglutinin antibodies selected for no neuraminidase antibodies.

You give the volunteers an HID50 of 3 to 10, 80 percent of them have a classic febrile influenza illness except for lacking the lower respiratory tract. As soon as we got into the era which is where we are with anything we do now we began to have less illness, less fever and a compromise from that cross-reacting heterotypic immunity was our interpretation of it but a lot can be learned.

I think we learned a lot. I think you learned a lot. Too bad we don't have a model readily available now but it has its limitations. We have to keep those in mind.

DR. TREANOR: That observation that the virulence of the wild-type challenge tends to decline as you get further into the era even though you are still screening with antibody tests I do think reflects some underlying heterotypic immunity that is developing which is not easy to measure, but there is something that happens to people as they are repeatedly exposed that doesn't get reflected well in the HAI assay that does dampen down the response to the challenge, and I think your point about the upper respiratory is important. This is actually the reason why I have always been a little bit surprised at how the McMichael study turned out because if you asked me whether or not having detectible cytolytic T cells in your peripheral blood would protect you against the kind of upper respiratory tract replication of virus I wouldn't have thought that it would, but apparently it did.

DR. WRIGHT: Peter Wright. John, do you want to just comment on the difficulty of making pools and logistically of carrying out studies like this?

DR. TREANOR: From the point of view of pandemic viruses I don't think there even is any thought that someone would be interested in making a wild-type pool of virus to inoculate people with but for the other seasonal flu viruses it is actually not technically difficult to make wild-type pools but they are expensive because they require as would any live viral vaccine very extensive testing for adventitious agents which is expensive and so it requires basically that someone would really want to do that and produce the viruses so that they could be used, but technically I think it is a fairly straightforward kind o thing to do.

To my knowledge there is not such a virus available but there could be at some point in the future.

DR. MURPHY: Brian Murphy, NIH. When you give the cold-adapted virus to very young infants who have never seen influenza before you obviously can't give wild-type virus to those individuals. You do find a higher level of virus shedding and a greater number of individuals who replicate the virus and aren't I right, Peter? You get about 70, 80 percent of kids who are given --

DR. WRIGHT: Right.

DR. MURPHY: It is a very different pattern. So, I think it is a more useful model for trying to get information out of the young seronegative child. I mean it is the only way to get information and we found this very, useful for the development of RSV and parainfluenza viruses as well. That really is the target population. So, if you are thinking about vaccines for children who are less than 1 and influenza causes a lot of disease in that age group using the cold-adapted virus as a challenge can provide very useful information and your frequency uptakes and your magnitude of replication is higher in that population than in the older population where you have some heterotypic immunity.

DR. TREANOR: Yes, I should point out the data that I showed I think those kids are on average about 5 years old, something like that. So, they are not infants by any means.

DR. SUBBARAO: All right, thank you very much.

The last talk in this session is going to be given by Jerry Weir from the FDA on the use of surrogate measures of efficacy.


Agenda item: Use of surrogate measures of efficacy

DR. WEIR: I actually have two announcements before we get going. One is we have been asked by a lot of participants in this workshop about the availability of speaker slides and so I just thought I would update you that we were trying to contact all of the speakers now to see if they are agreeable to this, but what we would hope to do is put the slide sets on the web site that was used to announce the meeting and I can't tell you right now when they will be available but hopefully within a few days and the speakers will get a chance to make edits and corrections if they want to.

So, that is our plan. We will be talking to them and hopefully within a couple of days or a few days we will have those so everybody can find them. I think that will be useful because a lot of data has been presented here.

The second announcement is that of course we are going to have a panel discussion after I talk and after a break and we will be flashing up slides of a sort of collated group of questions and comments and I actually have handouts for all of the panelists here that you might want to grab before the break so that you can look at them, at least for 15 minutes.

Anyway that is the announcement. Oh, a third announcement. I think we have agreed, Kanta and Maria that we will most likely at this point skip the wrap-up of the Sessions 3 and 4. We are hoping everybody since it is today can remember what went on anyway and we are also hopeful that the panel discussion which most of us think could be the most interesting part of this workshop anyway will be entertaining enough.

So, with that I will go ahead and talk. What I was asked to do was to talk about the use of surrogate markers for efficacy. So, one of the disadvantages of giving the last talk in a 2-day workshop like this is that there is always a chance that somebody else has said the same thing that you were going to say, and in my case since I wasn't going to present any data that is doubly true and so you are going to see a lot of repetition of what you have heard already.

Now, the advantage is that I can sort of gloss over most of this and we can move on to the panel discussion in a few minutes, but nevertheless I will go through this. Like I said you will find it is pretty repetitious if you have been here for the 2 days but hopefully it will serve somewhat as a summary, sort of a springboard into the panel discussion anyway.

Okay, so, what I was going to do was give a couple of slides on the current state of efficacy evaluation of pandemic influenza vaccines. So, I think it is obvious to say that evaluation of pandemic influenza vaccines is going to depend on surrogate, to some extent on surrogate measures of efficacy. Currently the principle surrogate measure of efficacy is an immunogenicity response in clinical trials and by and large most trials look at hemagglutination inhibition antibody titer and this, the use of this endpoint of course as you have heard is based on data primarily from seasonal influenza vaccines, primarily inactivated vaccines.

Okay, so we have been through this as well. HI antibody titer as an endpoint. You have seen several times the different criteria that are used. So, I don't think it is worth me spending any time on it. Depending on the area of the world seroprotection, seroconversion or GMT increases following vaccination.

Okay, so there are several challenges though in the efficacy evaluation of pandemic influenza vaccines. The first one sort of relates to the HI assay and the use of this to evaluate vaccines and the way I phrased this is simply is it appropriate to extrapolate what we know from seasonal influenza vaccination to pandemic influenza vaccines particularly the use of HA as an endpoint and the endpoint criteria previously defined and as an example of this I quote from the EMEA guideline on dossier, structure and content for pandemic influenza vaccine marketing authorization application. They state in this document that the criterion of an HI titer of at least 40 units is based on the assumption of a correlation with a reduction in influenza-like illness when most of the vaccinated population has some degree of pre-existing immunity against interpandemic strain. This may or may not be valid for pandemic influenza vaccines.

Other challenges, again maybe obvious is an HI antibody surrogate or any antibody surrogate appropriate for all types of pandemic influenza vaccines. Other examples as we have discussed quite a lot, live attenuated influenza vaccines, but there are other vaccines on the horizon, recombinant subunit vaccines, vector vaccines and so what is the appropriate surrogate that one would use to evaluate these in trials?

How do we define the appropriate immunogenicity endpoints that would serve as a surrogate for evaluating pandemic influenza vaccines and finally at least finally in the ones that I put down how do we establish the protective levels associated with newly defined surrogate endpoints and accurately quantify the responses following vaccination?

Now, Bob Couch mentioned this on day one and pointed out again that there is a spectrum in responses and how difficult it is to assign a value but on the other hand if you are going to have a surrogate you have to draw a line someplace with a number that is meaningful.

Okay, so briefly I am going to walk through some of the immunogenicity endpoints and again this is where it is going to get pretty repetitious.

HI antibody response obviously is the one that we have looked at in the greatest and the most examples. There are several advantages as an endpoint as I think John Wood pointed out earlier today. The assays are relatively simple and high throughput. For the most part they are validated and reasonably standardized and as I said already a titer of 1 to 40 has been correlated with reduction in influenza-like illness. There are still some disadvantages and hence some concerns.

One question is whether a 1 to 40 titer is a protective level for all strains of influenza including a pandemic strain. Would a 1 to 40 titer provide a similar protection in a naive population and does a 1 to 40 titer correlate with protection if assay changes are necessary to achieve sensitivity for pandemic vaccine response, i.e., the example of having switched to horse red blood cells to get a sensitive assay and is a 1 to 40 HI antibody titer necessary if other potentially even more appropriate surrogates could be identified?

We have had a lot of discussion about neutralizing antibody response as a clinical endpoint for pandemic influenza evaluation. There are clearly some advantages to such an endpoint. Neutralization is thought to be a measure of function that would be easily thought of as important for protection.

Neutralization assays are at least in many people's hands, some are sensitive compared to HI assays and as you also heard this morning there are and there is an international effort under way to try to standardize neutralization assays and reduce some of the variability in these assays from lab to lab.

Of course there are some disadvantages and concerns as an endpoint. One, there is no specific neutralizing antibody titer at this point that has been correlated with a reduction in influenza-like illness.

There are difficulties in standardization and variability between the labs or among the labs and then of course how does one bridge neutralizing titers to either HI titers and/or protection?

Mucosal antibodies, we have heard that mucosal antibody is a potential surrogate. There are some advantages to using this as an endpoint. It is thought to be important for protection. Some studies for example seasonal live attenuated influenza vaccines have correlated mucosal antibody, IgA in nasal washes with protection.

The disadvantages of course, no specific IgA antibody titer has been correlated with a reduction in influenza-like illness. There is a question of how protective mucosal antibody titers were determined and then technical difficulties in standardization. I heard someone describe this as more of an art than a science earlier today.

I am actually skipping over other things like N2 antibody, antibody to neuraminidase but some of the same questions would apply to many of the issues of surrogate endpoints and in this slide I am lumping all of the cell-mediated immune responses in one slide. Obviously they are not easy to lump together but some of the advantages are the same. CMI is a likely contributor to protection and as you heard just a few minutes ago may actually provide some degree of cross protection.

There are disadvantages and concerns about using any number of these as endpoints. One is that no specific measure of cell-mediated immunity has been correlated with a reduction in influenza-like illness. How would an appropriate CMI endpoint for a pandemic influenza vaccine evaluation actually be defined? Then of course technical difficulties in developing validated standardized assays not to mention high throughput assays and then there is always with some of these the relevance of the assay, questions about the relevance of the assays and the clinical samples themselves to functional mechanisms involved in protection.

Okay, in this slide I listed some of the potential attributes of a useful surrogate immunogenicity endpoint for pandemic influenza vaccine evaluation. I don't know that this is all inclusive or not but some of the things that I think one would want to see in a useful endpoint would be a correlation of the immunogenicity endpoint with a protection in relevant animal models.

One might want to see a correlation of an analogous surrogate immunogenicity endpoint with protection against seasonal influenza strains in actual clinical trials.

The third one I listed was and again this might depend on your viewpoint a correlation of the immunogenicity endpoint in vaccine trials with protective levels measured in subjects following natural infection. I think the question here of course is what is a protective level following natural infection and then last but not least the potential attribute or one of the attributes that would be key would be the availability of a validated assay that is practical, quantifiable and can be standardized among various laboratories.

I will explore these in just a little bit of detail. The role of animal models in defining relevant surrogate immunogenicity endpoints, there are some obvious strengths here. We have heard a lot about this today. Animal model studies can provide an important proof of concept for identifying the potential surrogate and vaccination strategies that might be useful.

Some of the strengths include the ability to correlate immunogenicity with protection. Of course the protection is defined by the model or in the model. Animal models can be used to determine the protective levels of an identified marker and finally they actually can be useful to facilitate the development of an appropriate assay to quantify that protective level.

Now, the weaknesses of animal models and we have seen many slides on this today. There are clear differences between animals and humans, for example, inbred populations, and I am referring to the animals, the physiology of influenza infection and then there is always the question of relevance of animal models to clinical studies. Are the protective levels in a model the same as for humans and are you actually measuring the same thing in the model as what you would measure in the clinic?

Extrapolation of surrogate endpoints as defined in seasonal influenza vaccine studies to the pandemic situation, the strengths of this sort of extrapolation are that the current surrogate marker HI antibody seems to be appropriate and applicable for all currently circulating strains and these influenza subtypes were novel at some point in time.

Seasonal vaccines or vaccine studies are useful for establishing new surrogate markers for protection and these may only be feasible or the ability to identify and correlate a new marker may only be feasible using a seasonal influenza vaccine and finally seasonal vaccines might provide some support for new vaccination strategies, for example, adjuvants that might provide support to determine cross protection and disease protection in special populations, for example, the elderly.

Weaknesses of this extrapolation are that seasonal studies are conducted in a non-naive population unlike those exposed in a pandemic strain and of course the nagging question is how certain are we that what is an identified level of protective immunity for a given marker is the same for all influenza subtypes including pandemic strains.

Studies of natural infection as a guide to define relevant surrogate immunogenicity endpoints, the strengths of this possibly that the human immune response to infection with novel strains of influenza may shed light on protective mechanisms and levels of specific surrogate endpoints needed for a successful vaccine.

Weaknesses, recovery from infection likely involves multiple mechanisms and distinguishing relative protective effects is probably difficult. This of course, even applies to vaccination studies that I think is doubly hard when you are looking at recovery from natural infection.

There is the assumption that recovery means protection. We did see data I think yesterday that suggested that this was true for seasonal influenza strains but can you extrapolate that to a pandemic strain and finally of course in using natural infection there is always the problem of timing of analysis may limit the usefulness of the data collected.

Development and evaluation of assays for surrogate immunogenicity endpoints, these are not actually strengths and weaknesses; these are just attributes. The assays for proposed surrogate immunogenicity marker must be appropriate to that endpoint. In other words are you really measuring what you think you are measuring.

The assays must be robust enough for evaluation of large-scale clinical trials, high throughput, quantitative and ideally the same assay could be applied to the animal model as well as in a clinical setting and the assays should be standardized with a low interlaboratory variability so that the results from multiple clinical trials can be compared.

There are several advantages to trying to define new surrogate endpoints now rather than later. One is that this can guide the development of new and improved vaccines for pandemic influenza. We could clearly increase our confidence in the effectiveness of a pandemic influenza vaccine before that pandemic actually occurred and it can actually guide the development of vaccination policy. All of these I think are important.

So, to summarize and conclude as I said this is pretty much what others have concluded and summarized earlier, new surrogate immunogenicity endpoints are needed for evaluating pandemic influenza vaccines. This is to increase our confidence in the effectiveness of vaccines against novel strains of influenza and to facilitate the development and evaluation of new generation influenza vaccines. There are of course unique challenges to identifying new surrogate immunogenicity endpoints and for establishing the levels necessary for protective immunity with a pandemic influenza vaccine.

You could imagine situations in which there are multiple accepted surrogate immunogenicity endpoints tailored for the evaluation of a particular specific type of vaccine and new and improved assays capable of quantifying such endpoints will likely be needed to be developed and standardized and as I said our next session will be the panel discussion and so since I was the last speaker I got to put up my own panel exercise in advance of the rest of them, but all of us probably have different things we would like to see come out of this workshop but this is something that I would like for both the panelists and the participants to think about over the next few minutes as part of this panel discussion and here I gave you an assumption. One is I don't think we can explore every potential surrogate endpoint in detail over the next couple of hours. I think that is not realistic.

So, I think we should choose one, in this case neutralization because I think most people would agree that neutralization is a little further along than some of the other potential endpoints not counting HI. Assume that that would be a useful surrogate or likely surrogate for effectiveness of an inactivated pandemic influenza vaccine containing HA. Like it or not most of the vaccines that are being developed are inactivated vaccines at this point and so what I would like to hear from everyone is what sort of combination of preclinical studies, clinical trials, assay development and validation would be needed? What do you think would be useful to establish neutralization as an acceptable endpoint for evaluating trials for inactivated pandemic influenza vaccines and two, and this is just as important what would be needed to render it practical and useful for doing so?

So, with that I will stop and like I said, hopefully it will be a good lead in for our panel discussion.


DR. SUBBARAO: Thank you. Does anyone have a question for Jerry?

DR. ALFONSO: Is it possible for us to have copies of your presentation available for this afternoon or I don't know how you want to distribute this? It is a pretty good summary of what happened more or less in these 2 days of discussion. So, it would be good if you could print it out for us this afternoon.

DR. WEIR: I will see what we can do at the break. Just come out to the table.

DR. SUBBARAO: Thanks and try to be back here by three-twenty.

(Brief recess.)


Agenda item: Panel Discussion

DR. ZAMBON: The four of us were moderating different sessions. A number of questions have been submitted and we have got some of the questions up there. We are going to first direct the questions to a panelist and then back to the audience, okay?

So, Jackie, I think you are going to take the first set of questions.

DR. KATZ: The first set of questions are about the antigen-specific immune responses in humans and what else do we need to do and how should we focus our efforts.

So, the first question I am going to direct to Wendy Keitel and that is is there agreement on the immune response priority to evaluate vaccine immunogenicity and efficacy.

DR. KEITEL: I drew the short stick this afternoon but I would say in general I think we would all agree that yes in terms of looking at HAI antibody responses I think it is generally agreed based on the types of constructs that we are currently studying HA-based vaccines, both inactivated as well as live attenuated we would agree that HAI should be measured as well as neutralizing antibody responses.

I think it has been reiterated over the course of the past 2 days that perhaps we have neglected the neuraminidase and I will leave a discussion of anti-NA responses to Dr. Kilbourne which will follow and then after HAI and neutralizing antibody responses I think we have heard a lot about newer assays that can look at cell-mediated immune responses but these are quite developmental and not well standardized from laboratory to laboratory.

So, the base assays for HA-based vaccines will be HAI and neut with consideration to neuraminidase and then as we look at newer constructs we are going to have to select assays based on the type of immune response which we expect to be elicited by the vaccine.

DR. KATZ: So, is there any comment or any further discussion from the audience on that point?

DR. KEITEL: I have another question which will come later which will relate live attenuated and trivalent and activated. So, maybe I will just address that at this point. It is very interesting to me over the past couple of days to hear about all of the potential assays that we could use to correlate with immunity to flu focusing both on surface glycoproteins as well as other components of the virus, the MTE, other internal proteins and obviously the live attenuated influenza virus is carrying a full complement of these proteins and in addition they are administered directly to the mucosal site and are more efficient at eliciting mucosal antibody responses and I think studies that have been alluded to yesterday and today it is clear that the nasal wash antibody is an additional correlate of immunity. The issue and the problem there is that assaying nasal wash antibody responses, standardizing and validating these assays is extremely difficult. So, it would be hard to imagine unless we have a really successful development process how we are going to measure the nasal wash antibody responses.

So, comparing live attenuated and TIV clearly the serum antibody response is not the whole or even necessarily, it is a major contributor but nasal wash antibody responses are going to be very important to assays comparing live and inactivated vaccines as well as the cell-mediated immune response against the internal proteins.

So, it makes me want to think about this whole issue of surrogates of immunity and correlates of immune responses and perhaps what we need to be doing is thinking about building a model for building in these various immunologic parameters to be able to predict for any given vaccine what contribution a particular type of immune response might have. We might need some help from some of the mathematical modelers to start working these in for any type of vaccine so that You could predict how much or what percentage of protection would be contributed by immune response to DHA, in the serum and in the nasal wash to internal proteins and so forth.

DR. KATZ: Any comments from the audience?

DR. COUCH: My comment would not be a very specific one from there but I did want to differ with Marie a little bit earlier. I think the gold standard is HI. It may not be the most sensitive and our best one and it may not be the most desirable one but nevertheless that is a pretty established gold standard that we can relate everything to. I think you ought to measure everything you can in one of these assays and that is how we develop data that can correlate with those responses then with the HI that forward some of these new surrogates that Jerry was talking about trying to get.

DR. ZAMBON: Could I just respond to that? I am not aware that I actually disagreed with you.

DR. EPSTEIN: Sue Epstein, FDA. This is a comment on the suggestion of mathematical modeling. I think we are nowhere near there and the prediction of the percent of the response that each component might be responsible for it would be profoundly difficult and I think it has to be empirical. I really think there is no alternative to empirical and when you administer different combinations they influence each other.

DR. KEITEL: I think, however, we could start building these models for interpandemic influenza based on the types of immune responses that are elicited. You start putting things into the model and seeing what independently could predict infection and what contribution is being made.

DR. EPSTEIN: If you measured the component responses and predict the outcome, okay. What you can't predict is if the vaccine consists of a mixture of the following components what immune response profile you will get and that is where we are doing empirical comparisons of vaccines.

DR. WOOD: David Wood, WHO. Wendy, maybe you said this and maybe I missed it but just to go back to the first question on the priority of the immune responses do I assume that you mean that the responses that you are looking for are in serum or mucosal immune responses and how would you prioritize serum versus mucosal?

DR. KEITEL: I think for the inactivated vaccines administered by the parenteral route based on the best information at this point we would prioritize serum antibody responses. I think we are missing an awful lot of information if we focus exclusively on serum antibody responses for vaccines such as live attenuated viruses that are administered by the intranasal route and I think we urgently need some way to measure the immune responses in the respiratory tract for the live attenuated vaccine development program.

PARTICIPANT: Sue, everybody is in love with antibodies and I think the most exciting, and I am not a B cell guy but the most exciting thing I heard today or yesterday and today is that 7-day assay for the cells that are, the blasting cells that are on their way to wherever they are going. The assumption is that they are going to the marrow to secrete antibodies that are going to be found in the entire periphery but they may be on their way to the nasal mucosa to secrete IgM.

So, it seems to me that rather than worrying about the mucosal part separate from the circulating antibodies that an effort should be really made to sort of take advantage of the fact that you can find these blasting cells after immunization. On days 7, 8 and 9 it starts falling off and then go in and actually see is it IgM; is it IgG?

I have heard that there is some data that is going to be published soon that in one study Don Caper actually sequenced the V genes, made some monoclonals and actually showed that it was a case of original sin, that the blasting cells were actually for a previous immunization and not the one that was given just 8 days prior to that time.

So, by looking at those blasting cells I think one, if you are interested in the antibody part of it that is the part that at that point you can get a new novel insight into what your vaccine is doing without all those downstream assumptions about what is happening later in terms of which tissue is it going to localize after it leaves the bloodstream, etc., and the other thing I want to put in a pitch for is that perhaps helper T cells should not be neglected when one thinks about antibody production.

DR. KEITEL: I agree with that and I think there will be further discussion on the development of some of these assays. They are not ready for prime time in everybody's laboratory and I think it is very important. Brian Murphy may want to make some comment about serum IgA as a marker for respiratory IgA levels.

DR. MURPHY: I do think although you can detect these IgA cells in the blood it is very important to measure where they go and once they get there how much antibody they are making, how much of that transudates into the mucosal surfaces of the respiratory tract where it is active. It is not active submucosally. If most of those are going to your lower thigh or something they are not going to help you very much against influenza.

So, you really want to measure, ultimately you want to measure for IgA antibody in your respiratory secretions and I think you need to optimize those assays. This is good, but the best thing is what actually winds up and how long it stays there.

DR. COUCH: I agree with Brian. That’s where it needs to be and that’s where the assays ought to be and if we could get those cells out, the way we can get these out, we’d be doing those assays on the cells that are up in the submucosal area. I wanted to tell Harry and I kept forgetting about it, at least put into your spot forming cells an antisecretory piece antibody and see if any of those cells that are recirculating around there are producing the dimeric and polymeric and could give you any guidance. My guess would be that there are probably some there but it is such a low percentage that you wouldn’t do any good with the assay, but you may as well try.

Hana Golding, CBER: I think a lot of the comments that were made today up to now and even the way it is posted can fairly much apply to both seasonal vaccine and pre-pandemic vaccines and we are raising a lot of issues related to the new form of vaccines and what do we know about how the cold-adapted work and actually we have already licensed products that really the quickest way to do and was suggested to go to the vaccinated people and see what are the best immune correlates of their protection if it is not simply HI, but what I would like to hear from the panel in the upcoming discussion is really trying to ask ourselves how are the pre-pandemic vaccines different from the seasonal; what it is that we would like to get from a pre-pandemic vaccine. Are we only worried about homotypic protection or do we really want to get something that would give us heterotypic protection against other clades? If so, what are the assays that we discussed today that are more likely to give us this kind of readout and will be sensitive to the different correlates of immunity?

So, instead of just in general talking about is it HI or microneut or CTL, well, if we need to look at CTL do we have the assays that will give us the answer?

I think we have to identify what it is about the avian influenza infection that is unique and different from the seasonal one that we want to address in our moving forward with pre-pandemic vaccines and try to be a little bit maybe more rigorous about what we would like to ask from this type of vaccine and this type of assays.

DR. KATZ: Does anybody want to take that?

DR. COUCH: In a more general sense I can't answer H5 questions. I am not that close to it and you have people on this panel that are and that was a fairly H5 specific one, but coming to this meeting did remind me of something I did 3 or 4 years ago for a meeting and that was to try to go back over some of that pandemic data on swine USR 1957 and what I could get 68 so far and when I started trying to sort out from that information what were the markers of immunity in '57 in those vaccines and in '68 and in '77 with the three unique ones that came in that were our guidance for immunity, like a 1 to 40 or some kind -- and I threw up my hands. I could not identify one and so most people who were at that meeting know that I came away from that review with a conclusion that we don't know what it takes to protect humans adequately to a novel, brand new novel virus which was the case in each one of those and in the absence of that knowledge we know that immune mechanisms are there and immune mechanisms are working and so the only thing that I could recommend was let us get as much as we can of what we can measure from the vaccine and I would be confident that would give us protection and hopefully would be a large amount. That was the best I could do. Now, if somebody could take on H5, I can't.

DR. KATZ: I would say that based on the knowledge of the diversity of H5 out there at the moment and the sense that we need to be ready for any clade or subclade that might emerge I think that is a fairly novel situation that we really haven't found ourselves in before.

So, I think there is a lot to be said for coming up with ideas of how we can make the broadest response possible and whether that is to as Bob said to make the highest antibody responses possible so that we will see better cross protection, that is one approach or we design, attempt to design vaccines that are intrinsically more broadly cross reactive and I think that addresses the, I mean that is a question for seasonal vaccines as well but I think the situation with H5 is that we are not, if an H5 virus does emerge as a pandemic strain the sense now is that we really want to have the best antigenic match possible and we are not in a situation to predict what that might be.

So, having the broadest response possible is the best we can do at the moment.

DR. KILBOURNE: I would like to make a comment. It seems to me that we are milking mice here that we are concentrating on what is not even a pandemic virus and the title of this whole thing relates to pandemic. I think when we start to focus on H5 and its clades and everything it is entirely premature. I think we don't know the next pandemic virus. We know a lot about an epizootic that occasionally infects humans and I hope that somehow that is going to open up the course of discussion.

DR. WOOD: While I agree with that very response I think there are some practical questions or practical reasons why we want to know some of these answers. For example, vaccines are being stockpiled. Current H5 vaccines are being stockpiled and the question I was going to ask Jackie is you are saying that we need to have a response that is as broad as possible. How would you assess that response when we have 10 clades? Do you need to test all 10 or what would you do?

DR. KATZ: I guess right now what we are trying to do is test the ones that have proven to infect humans at least and probably some labs could go beyond that and test broader, the 10 clades, but I think the relevant ones right now are, you can only keep up with what is relevant and current at the time, and I think certainly testing three or four, clade 1 and multiple clades 2 subclades is appropriate and that is ongoing and that is a major question for any of the vaccine clinical trials as far as I am aware.

DR. ZAMBON: Could I add to that in the sense that there has been quite a bit of data produced in various different sorts of studies over the last few days which indicates that the height, the sort of magnitude of what we currently measure however good or poor that might be is at least some sort of surrogate for kind of breadth of immune response and so when one thinks along those lines the actual magnitude of response of whatever it is that you are measuring I think is good as a way of putting it and the quality that we will be looking for in whatever we measure would be I think magnitude, duration and breadth of response. Breadth is probably linked to magnitude and I think that sort of data has come from human clinical trials and from at least semi-predictive animal work. So, I think that is the consensus which we could probably agree on.

DR. TOPHAM: Dave Topham from Rochester. I have heard a lot of talk about measuring immune responses to various vaccine candidates and formulations. My question is what do we still have to learn from what is an optimum immune response to these seasonal natural flu infections. As an immunologist I can't answer that question for human subjects.

So besides having a well-matched neutralizing antibody in the absence of that even with seasonal flu what sorts of immune responses do you need to resist disease, not necessarily infection but disease? We have seen some evidence that is starting to come out but it seems like we don't really have good assays in place yet and we are not sure what those correlates are, what the level should be and can we learn from that as opposed to focusing on the pandemic?

DR. BELSHE: Let me tackle that. One of our formal questions is on here. Do we define the correlates of protection for different ages and it really gets to the question of why do we want to do correlates of protection. I think it is so that we can improve the vaccines. If you have a correlate of protection you don't have to go through an efficacy field trial every time you tinker with the product and as Tom Jefferson pointed out yesterday we really haven't done much in the way of correlates of protection studies in parallel to efficacy field trials. It is fairly straightforward to collect an extra serum or two when you are doing field trials.

It is not easy to collect cells and it is not easy to collect nasal washes. So, we don't generally do that. It is enormously complex but getting an extra serum the day before your volunteers become sick is pretty easy and obviously it wouldn't be day before but when do you do that? Simply building in an extra serum on January 1 or January 2, if you want to take the day off will get you a serum pretty close to infection that you could then use to at least measure some correlates of protection to injected vaccines. It is much more difficult to try to do that for mucosally applied vaccines or for a vaccine designed to do cell-mediated immunity. Those are much more complex studies to think about and those studies you might want to think about going to John Treanor as the challenge model.

DR. TOPHAM: My point is are we going to get the best information from doing field trials of vaccine candidates? Are we going to get the best information from looking at natural flu infections naturally acquired? I am not sure what the answer to that is. I think it might be actually the latter.

DR. BELSHE: The other point I wanted to make was we don't have to look very far to get a good model of pandemic flu. You look at children who are naive and they behave like the rest of us are going to behave when we get a novel flu come through. The elderly it is another question. If you are worrying about vaccines for the elderly that is a whole other set of correlates that we need to look at but looking at young children I think is the way to go in characterizing what are the immune responses to natural influenza and what are the immune responses to various vaccines and which of those responses reduce their likelihood of getting infected with influenza?

DR. WEIR: Jerry Weir, CBER. I am going to put Bob and Jackie and to some extent Maria on the spot based on your last answers.

So, the three of you to some extent said that you didn't know what you would be looking for except a better immune response, a broader response but again I get back to my question. Did that imply that you are satisfied with continuing to use HAI titers or do you need something else and if so what is that something else and how are we going to get there?

DR. COUCH: We didn't say that, Jerry, but in trying to look at the data which I did can I find a number; you see that was part of the question. I could not. I gave up, but there is no question about immune responses protecting. So, what I would say is that you would like that standard but I would have to say that my best judgment of vaccines and the best way to do that is have a comparison, not live. Live versus inactivated will always be compared where possible but we haven't really dug into much on the immune mechanisms there, but if you are comparing two inactivated vaccines I want the highest and broadest response I can get and that is the most I can offer to the public and I can't pick a surrogate that says, "Now, I know I have got a good one." I just don't know how to pick it. So, therefore I have got to get the most I can and then if we do the studies during the pandemic I will tell you that we had a meeting saying, "Okay, let us get all ready to collect the data we need." In 1968 we did the same thing and again in 1976, and we still got the same questions, but we still should continue to ask those questions and try to gather that data to see if that will be the last time we have to.

DR. GERHARD: Walter Gerhard, Wistar Institute. I would like to make a remark that one should achieve obviously a strong protective response against homotypic virus but it also should be a broad response and in that regard I think it is now clear that humans lack immune activity which in animals has been proven to provide very broad protection although not as strong for example as HA antibody, but I think what we really should not forget is that there is this component of the immune response which at this point humans lack and one should try to improve that to see how much we can improve the protection in that way.

DR. COUCH: That is a point of information that we need to understand.

DR. BENNETT(?): Hello, I am Jillian Bennett. I am from CSL in Australia. I wanted to just provide you some insight that we had in terms of developing our pandemic vaccine. One of the panelists earlier in the session had alluded to the fact that maybe we should look at the responses that we see in children because they are a naive population and I thought that you would be interested to know what we did in terms of developing our pandemic vaccine. We did do a clinical study initially with our trivalent inactivated vaccine in a pediatric population age from 6 months to 8 years and we gave them the licensed trivalent vaccine for Australia and those studies indicated that probably if we developed an H5N1 vaccine that included 15 micrograms or 7.5 micrograms of hemagglutinin we would expect to get good immune responses. So, what we did is we made vaccines with and without aluminum adjuvant and we were really disappointed because the H5N1 vaccine did not give us the same type of responses that we had seen in our naive pediatric population, and then one of the other questions that we asked from those studies and Professor Zambon presented the data yesterday, after 2 doses of this vaccine what happens if you give a third dose. We hadn't actually achieved the serological threshold that is specified in Europe but by actually giving the third dose of vaccine we approached that threshold. Learning from all of that though we further applied and we have actually, and by the way we started this study before the data from the US Sanofi vaccine became available and then it was also the same time as GSK was doing their work and ultimately what we are proposing to the Australian Government right now is a 30-microgram HA vaccine adjuvanted with alum. Although we wanted to learn from the pediatric population it wasn't as insightful as we would have expected.

DR.BELSHE: Just to clarify what you can learn from children is not how they respond to vaccines but how they respond to natural infection after vaccine and so I think what you have recapitulated is what we have observed over and over again. H5 is not very immunogenic and there is good reason to do these trials in advance of any pandemic figure out what the dose is and the number of doses.

The alum adjuvant is kind of an odd thing. I mean we have seen in our NIH trials that alum had a detrimental effect and so I am puzzled by some of these other trials that are showing a positive effect and so that I think is something that needs to be worked out and so what is it about some alum adjuvant that may enhance the immune response whereas other alum adjuvants are detrimental is somewhat of a puzzle.

DR. ZAMBON: I would like to speak up on that because when I commented yesterday on the effects of alum I think the one thing that we take from our experiences in Europe is that the effect of alum is somewhat unpredictable both from vaccine types or formulation and you can't always predict even within a particular manufacturer's portfolio what is going to happen and one of the things that we think might contribute to improving predictability is understanding the relationship between the amount of adjuvant and the amount of antigen, that is the ratios that are used and quite often when you try to find out this sort of information it is not available and so an adjuvant is by its very nature enhancing the immune response to antigen. So, therefore it seems like a fundamental principle that we should understand what is the ratio; what is the optimum amount of antigen to a particular antigen type, and until we have a better understanding of that I don't think we are going to be able to predict exactly what is going to be the outcome of adding something like alum to vaccines.

DR. KEITEL: I will just add that the element of salts also differs.

DR. TREANOR: I will echo that it is amazing how little we know about the mechanism of action of aluminum given the fact that they have been used as adjuvants for such a long time, but I wonder if you would find it useful to divide your thinking into two different categories. There have been many comments about the kinds of immune responses that would be desirable to generate because they might improve the vaccine but the other issue that I think is also relevant is simple assays that we can use to distinguish between vaccines that are likely to work and vaccines that are not likely to work which don’t necessarily encompass all immune responses that might be useful but are simple markers that would be used to judge vaccine efficacy and the reason this is important is because I think the philosophy that generating more antibody is always better is probably true but you are always going to be faced with tradeoffs between what it takes to drive the immune response to its absolute maximum and what you really want to use is a vaccine. So, if you are confident that your vaccine has driven immune responses to the point that you need to get to you might stop there even though it might be possible to even make them go higher because you would have achieved what you needed and I think that is where the cut off issue becomes more important. It is to try to somehow judge how much you have to push things to get where you need to be and it is difficult I think fundamentally because of the fact that the antigen changes every year and I don't see a way around that problem. Every assays is going to be different because the antigen itself is a critical component of the assay and I don't know how you get to the point as you said. I think that is the reason why it is impossible to come up with a specific level.

DR. KILBOURNE: I think it is important to know what adjuvant affects and probably it is worthwhile looking at the alum but just recognize you are never going to use it for general immunization without lawsuits out the ears.

The Influenza Commission of the United States Army, the military branch found out that this is totally unacceptable definitively because of abscesses in the arm and lumps in the arm and so forth which in our delicate society today would not be tolerated. I think that it is swell and gives you your alum if you wish to get some kind of indication of whether you can stimulate with any kind of adjuvant but don't think it is going to be a definitive adjuvant.

DR. COUCH: I agree with John. I just want to clarify a little bit. The best way to sort out and he would agree with that, the best way to sort out new surrogates is going to be in efficacy trials where they are measured and none of us want to stick to antibodies but that is where we are right now but there are a lot of new candidates out there including N2 antibody that Walter identified but they need to be measured in relation to efficacy and what happens with infection and exposure.

PARTICIPANT: I just wanted to go back to the pediatric issue and I am wondering if anybody had actually compared the intensity and duration of virus shedding in children after they have received inactivated vaccines and what that might tell us in a pandemic where all of us are naive.

DR. BELSHE: There are 3 comparative trials comparing live vaccines to inactivated vaccines, two of them in children, relatively young, either 6 months to 5 years of age or 1 year to 6 years of age. Both of those comparative trials showed that the live vaccine was significantly better, about 50 percent significantly better.

Older children with asthma the live vaccine was about 35 percent better but in terms of looking at the parameters associated with infection other than culture positivity and symptoms associated with that culture positivity we don't have duration of shedding. One might presume for example that the live vaccine was better at shutting down viral shedding in the breakthrough infections compared to the inactivated vaccine but we don't actually have that data..

PARTICIPANT: I am just wondering if we might see more virus shedding following inactivated vaccine in the general population during the pandemic when we are all basically naive to that virus and if vaccination may contribute to greater circulation or exposure to others.

DR. BELSHE: In the absence of secretory IgA I think that is a reasonable hypothesis but I don't know of any data for exposure.

DR. KEITEL: I think some of the vaccines that are being evaluated now contain novel adjuvants and the novel adjuvants have been shown to augment serum antibody responses and it is possible but not known whether they will also improve the antibody responses in the respiratory tract.

Of course, one of the issues about taking these novel adjuvants down into the pediatric is particularly in very young children; so there are safety issues and so forth that have to be taken into consideration, but it is very clear that if you even in a non-adjuvanted vaccine albeit in prime populations if you increase the dosage of a primal vaccine you can augment the appearance of nasal wash antibodies and so I think either increasing a dosage to the higher serum antibody level or adding a novel adjuvant which also augments the antibody you are going to be able to get respiratory tract antibody and then the question would be how does the inactivated compare with live when you optimize the responses both in the serum and in the respiratory tract.

DR. WRIGHT: I was recently at a meeting on polio vaccine and there are so many analogies between the Salk and Sabin vaccines and what we are thinking about with influenza and there was a recent study done of giving the Salk vaccine and then giving Sabin polio vaccine and looking at shedding and the influence on shedding is there. It is a little bit in duration, a little bit in the amount of virus being shed but it is fairly minimal and in contrast to Sabin. Sabin, there may be some things you can learn or think about in relation to other viruses that are relevant.

I guess the only other comment I would make in response to Hana is I don't think of although I haven't been intimately involved in it, the H5 has not been an influenza virus and not having the same correlates of protection although we may not know that and my position would be that serum antibody might be very valuable because I think of this as so much of a lower risk and I am very convinced of the relative importance of IgG in the lower respiratory tract and IgA in the upper respiratory tract.

DR. KATZ: (Inaudible.)

PARTICIPANT: We couldn't hear the question.

DR. BELSHE: The question was do we need to define correlates of immune protection for different age groups and certainly I would agree that we need to do that. In particular if you look at the elderly here is an age group that we have relatively modest efficacy with our inactivated vaccine, perhaps 30 or 40 percent protection against culture confirmed influenza, maybe a little better against hospitalization and death, maybe 60 percent with those endpoints.

If we knew what the correlates of the immune protection were in the elderly we could then tinker with a vaccine that might be something simple like doing a high-dose vaccine trial with more hemagglutinin. Maybe it will require combining vaccines so they get both parenteral and intranasal vaccines. Maybe we need to add an adjuvant. So, if we had a better handle on the correlate we could move forward with improving that vaccine.

At the other end of the age spectrum it is more difficult to collect samples in infants but again we have relatively modest efficacy with inactivated vaccine in children under 2 and having a good correlate of protection might give us the handle. Maybe it as simple as giving a third dose of vaccine to that age range but those would be the reasons to look at correlates at different ages so that we can adjust our vaccines without going through enormously expensive and large-scale trials to measure efficacy, and secondly we have the ethical problem I have already talked about. You can't really do placebo-controlled trials in those populations. So, you are left with doing vaccine A versus vaccine B where vaccine A is your known standard vaccine and vaccine B is presumably a better product and to show an improved product in that sort of situation drives your sample size to enormously large numbers of subjects.

DR. KATZ: Okay, thanks, Bob.

So, one final question that I am going to direct to Ed Kilbourne pertains to the neuraminidase. We have heard a lot about attempts to improve neuraminidase antibody assays but can you discuss what the criteria for the NA content in current vaccines should be and your thoughts about neuraminidase responses, Ed?

DR. KILBOURNE: Thanks a lot. I think this was covered very well by Dr. Eichelberger this morning but I think if I can try to give an explanation as to why this is not done and we don't have a standard it is because NI tests are notoriously difficult to do. So, we are hung up with a technical problem which makes it understandable why this is not used and -- throughout the world for example in small laboratories. I think another hurdle that may not be sufficiently well appreciated is the fact that from the same isolate you can get mutants which are either very highly active enzymatically or almost no activity at all so that again it is mutation, mutation, mutation.

I am ducking the question obviously. I am sidestepping it, but I do want to point out that there is a test which is a functional assay which is even more difficult perhaps than the NI and that is plaque size reduction. In other words like neuraminidase and like a neutralization test in a monolayer of cells but in this case the antibody is incorporated in the arteries. It is continuously present and when they are infected the plaque develops more slowly and to a lesser degree and with classical neutralization the plaque doesn't appear at all.

So, if we want to come to an ultimate standard I think in terms of functional assays there you have it right in the Petri dish and of course it is not easy to do either but it measures what you really want to measure that is the functional action of antibody against the neuraminidase.

The other thing I would like to point out emphasizing the difficulty here of even deciding which vaccine to use is in the notorious nonpandemic of 1976. One of the drug houses either accidentally or I wouldn't dare say deliberately used a high yielding vaccine strain of swine virus S15 going back to (?) before and it turned out that this was easily the thing which probably the actual New Jersey 1176 virus which was used in the vaccine and was causing the epidemic by the fact that the neuraminidase was highly stable in this fake vaccine and therefore the vaccine that 53 million people got in the United States lacked neuraminidase at all. Now, I am a neuraminidase advocate but I want to say two things, to point out the primary efficacy and importance of the HA, but having said that I would like to say that I, personally would like to see every child in the United States given a cocktail of neuraminidase as the original and sole vaccine and rely thereafter on the repeated rounds of replication from natural infection to end up with definitive immunization with an occasional boost. So, those are my prejudices and absolve them if you will.

DR. COUCH: I want to agree with Ed's prejudices and ask him a question just to see if I am still thinking correctly with what I say because I have sort of picked up your mantle at the FDA meetings. They have heard me promote neuraminidase annually now since you no longer attend them annually but at one time you said that the best indicator that you were aware of for an in vitro measure of immunogenicity was enzyme activity. Would you still say that currently is the best marker we have for immunogenicity in that vaccine?

DR. KILBOURNE: It is true in a sense and it is a handy marker for protein denaturation or lack of it. (Inaudible.)

DR. COUCH: If that enzyme is not stable you probably don't have stable immunogenicity. Would you agree with that?

DR. KILBOURNE: I think you could account for the immunogenicity if it is active but there is a gradient of the denaturation which occurs. So, it is a nice marker for full antigenic activity. That correlates almost one to one.

DR. COUCH: But in vitro standards still need to be developed but that is a good starting point.

DR.KILBOURNE: Absolutely.

DR. COUCH: Okay.

PARTICIPANT: How would you incorporate Gary Nabel's pseudotype assay into that range of assays? Do you think that added something that might be easier?

DR. KILBOURNE: Yes, I do. I was very interested in that paper. I think that is worth exploring.

DR. SUBBARAO: We are going to move on to the next set of questions. These were questions that were submitted or that were mentioned that we sort of grouped together and so on this set of questions I am going to go in the order that they are written here but I was going to start with the second one and direct it to Bob Belshe. What information do you think is outstanding that could be gathered from animal studies?

DR. BELSHE: One of our questions has been a titer of 1 to 40 using horse cells, what does that mean and I think one might be able to quantitate antibodies in terms of milligrams or nanograms of antibody to nitrogen and figure that out using the animal sera and relate that to protection.

I think the other thing that could be done with animal studies is instead of using an overwhelmingly large challenge to perhaps do larger numbers of animals with an aerosol challenge that might more closely mimic the human experience. I think that infectivity of humans is by fairly low quantities of virus and probably mostly by aerosol or perhaps a little bit occasionally by (inaudible) but that is an area that could be explored. If we could get ferrets for example to get flu with low-dose aerosol I think that might more closely mimic the human experience and I would feel more comfortable about relating ferret vaccine studies to humans in that context.

DR. SUBBARAO: I went to an aerobiology workshop a year or two ago and one of the take-home messages that I came away with is that the aerosol-generating systems are very complex and they are very technically demanding in terms of who operates them and how they are operated.

So, although I agree with you that perhaps an aerosol challenge is more real than the intranasally administered materials with the pipette or drop I am also uncomfortable about aerosolizing wild-type H5 despite the Pappers that we wear

DR.BELSHE: You might start with H3.


DR. COUCH: I am going to have to differ a little bit with Kanta here. I think what you did was to go to a meeting where all the purists were talking about their subjects because you can do small particle aerosols. The method is very well understood and it can be worked very well, and we have had two groups who have done that in mice, but one person only does all of his that way, and it is very simple and easy to put those mice in a box to quantitate the respiratory rate and you can get a reasonable estimate of the small particle aerosol dose without having to have fancy equipment.

Now, that is not ferrets. That is mice now, but it would seem to me that it ought to be possible to make the adaptation to be able to do ferrets. Okay, but the last one is I don't think anybody would propose at the present time that because of the safety considerations that you aerosolize H5 even for ferrets or mice.

PARTICIPANT: Bob, I have to war with you on that one because very tight controls have to be used to have an effective aerosol infect mice, low humidity, low temperature.

DR. ZAMBON: Could I just comment aerosol vaccines are used for measles and they are used in developing world situations, but I think we ought to get away from the idea that we can't do aerosol delivery in experimental animal situations because if we can do it in the middle of Africa we can surely do it in experimental animals.

DR. WEIR: Since Bob has had experience with this I would like to ask what does this add; does it add something?

DR. COUCH: This is humans,not mice, but go ahead.

DR. WEIR: Does the aerosol challenge add something to the intranasal challenge? What do you get in addition by doing the aerosol challenge?

DR. COUCH: Actually Ed Kilbourne needs to comment on that, too, but when you are talking about animal challenges what do you gain, when you put it into the lower respiratory tract and everything we know, not direct but everything we know indirectly says that at least most of the disease, I am sure it is not all of it, most of the infections that are acquired by humans are acquired by the airborne route, not all of them for sure. So, it is important that that lower respiratory tract component be a part of what you do and that would be a closer inoculation than a big volume of liquid is in the nose which also puts it into the lower respiratory tract. That is all. It refines the system. By and large it is not going to give you a whole lot of new data. It just refines the system.

Now, we don't do that in humans. A little bit was done with wild virus way back and enough to sort of tantalizingly touch it you see and I have been asked on a couple of occasions if I would in the early days of cold adapted put it in by aerosol into humans but I just said,"No, I don't want to take that chance. I don't want that risk," and most everything that we are learning I think we can learn intranasally through the volunteer challenge model.

DR. BELSHE: My concern is that using 10 to the 6th challenge is very much like doing a neutralization in a test tube. I would like to see something mimic more closely what we think happens in nature which is a relatively low inoculum to cause infection. I think it would add something if we can get it to work.

DR. SUBBARAO: But I would add that we have in some of our challenge studies in animals, we have tried a lower challenge dose and what we find with the difference of the challenge doses being four logs apart we get a similar level of replication in the lower respiratory tract. So, I think we haven't tried less than 10 to 2. We tried 10 to the 2 and 10 to the 6 and got very similar data. So, I think there may be a threshold and maybe if we challenged with 10 TCID50 or less we might get a different result, but when we give 100 TCID50 or a million we get very similar results.

DR. BELSHE: That makes me a lot more comfortable actually hearing that you are only using 100 particles and getting infection and that makes me believe in that animal model a lot more.

DR. COUCH: In the animal model the volume makes a big difference because if you have done a big volume you put in the upper and the lower. If you put 10 to the 6 you put it into the upper and lower still just a bigger dosage. If you really want to separate the two the best way to cleanly do that is by aerosol. You can't purely inoculate the lower respiratory tract by aerosol. You inoculate both, upper and lower, but you can set the system up so that you only inoculate the nose.

DR. KILBOURNE: Bob, I will send you a paper because the volume is not that important in terms of what part of the respiratory tract it lands in.

We got tired of doing aerosol chamber experiments. So, we just tried simple intranasal instillation of relatively small doses.

. DR. COUCH: You got no lung growth?


DR. COUCH: It grows in the lungs.

DR. KILBOURNE: Oh, yes, absolutely.

DR. COUCH: Oh, no, you can inoculate the lung with lot to the nose. I thought that was what I was saying, but you can restrict it to the nose if you will restrict the volume.

DR. KILBOURNE: I thought you were talking about a greater volume was necessary with intranasal instillation.

PARTICIPANT: I have just been listening and realizing it is better to be thought a fool than speak and everyone know you are a fool, but having just listened to this I work in a company that specializes in aerosol challenge of etiological agents for the testing of vaccines particularly vaccines that are at BL3 levels and we chose not to pursue aerosolizing influenza although we could do it safely and our aerosol system has all the fancy what’s its and the high containment class 3 cabinets to ensure safety, but we did not pursue that because it was believed number one that this community believes that intranasal inoculation was sufficient for an animal model and secondly we believed that influenza transmission was mainly by droplet infection and not by aerosol that being it would be an infection from person to person that initiates in the upper respiratory tract and then translates to the lower respiratory tract. Now, if I am wrong I am wrong, and I would like you to tell me that because if this community believes that such a model is warranted then I need to go back and tell my bosses that we need some more money to do some development.

DR. COUCH: It depends on what you want to do with it you see as to how you do it. That is what we said a minute ago and there are people as he said who feel very strongly that this is an infection that is initiated the majority of the time with drops. You are talking about the large drops into the nasal pharynx and that anything that goes into the lower respiratory tract is a later aspect of the infection and there are those of us who are on the other side of that and I know Dr. Kilbourne is on the other side. I don't know about everybody else, but I think the majority of the data --

DR. KILBOURNE: I want to qualify something very importantly and that is --

DR. COUCH: Let me first of all before you qualify, I think the data, it is circumstantial data, but it is very strong for airborne transmission for a significant proportion of the influenza infections and the majority of them in acute outbreaks.

DR. KILBOURNE: Amen. I am totally with you but the thing I didn't say is that intranasal instillation is as good as aerosol if a mouse is anesthetized. So, I don't think you want to anesthetize.

DR. COUCH: I don't do these things. I just read them, but if you take the mouse and either don't anesthetize him or just lightly anesthetize him and get your volume not 40 or 50 microliters but down to 4 or 2 microliters and drop it into the nose you get an upper respiratory tract infection.

DR. KATZ: But you can also get an upper respiratory tract infection if you deliver a larger volume. Then you will get an upper and a lower. You will get both and that is ideal in terms of really for vaccine evaluation because you can evaluate protection in both the upper and lower and the point I just wanted to make with respect to Bob's desire for an aerosol is that certainly with the avian influenza viruses that we are working with you can, as Kanta said, you can drop the dose a lot lower. High doses are used so that we have a consistent and reproducible event. So we can use a smaller number of animals and have --

DR. COUCH: It is based on the question and the response you want to look at. If you want to say what is the best modality of protection you can do that. There is no question that there are attack rates in society. It is what is out there circulating. It has got to be a relatively low dose that people are being infected with and something close to that HID50 so that if you want to see, get the maximum of test effect of your vaccine I think you ought to challenge that vaccine with as low a dose as possible with that question.

Now, that is degree of effectiveness. You give it a big dose, it can tell you whether it works, yes or no. You can get black and white answers for sure.

DR. GERHARD: Walter Gerhard, Wistar Institute. So, we initially started with the aerosol and we stopped using aerosol because it simply was not reproducible. However, I agree with Bob that I would feel much more comfortable with using a good aerosol and the reason is that if you give a bolus injection you totally disturb all the lining fluid of the respiratory tract and you dilute it 100,000-fold and so you don't see the real protection which actually would act on the undisturbed respiratory tract by bolus injection and you would see by an aerosol. So, if I would get a good reproducible aerosol machine I certainly would use that.

DR. SUBBARAO: Any other comments on this topic? Brian, you look like you are half deciding.

The question is how should animal studies be designed to provide meaningful data about the vaccine's potential effectiveness against a pandemic virus.

DR. MURPHY: I was very impressed by the information that you got from your passive transfer of your human monoclonal antibodies in mice in that by doing that particular study where you looked at the ability of antibody in this case of purified monoclonal but you could use the polyclonal system induced by a vaccine in the same test. Its ability to do three things, protect against disease, okay and that you can measure by symptoms as well as by histopathology, you are able to look at its ability to protect against replication of virus which you confirm by doing immunohistochemistry. It could also be done by looking at PCR, amount of virus that is present by PCR, but there is always the possibility as you grind your lungs up, and we have done some of those studies you can get in vitro neutralization where your passage of transfer serum can neutralize the virus that is present in your lung homogenate. That can be obviated or checked by looking at PCR.

I would suggest that if I was in your position, I am happy actually where I am, Jerry. If I was in the position of trying to make rules about how would I look at vaccine effectiveness, okay, what I would do for pandemic vaccine is I would get a model with one or two or three of these pandemic viruses that differ in terms of clade X, Y or Z. I would have a pneumonia model in a mouse and then I would say then that I want to get the following sets of information in my process of evaluating these vaccines. I would like to get the HA titer. I would like to get the neut titer. I am very happy with Jackie's microneut, but also I would like to get its ability to provide passive protection, okay, in a mouse model against several variant viruses. I don't call these guys antigenic variants.

You see some weirdness following a single infection of a ferret and when you get a good final immunological response you get a nice broadly reactive response with these H5 viruses. That is what I would do and then you would have a nice biological assay. You would be able to validate. FDA likes to use the word "validate." I have learned that.

The other thing I would do, Jerry, as an aside here, I’d stop asking everyone to do these foolish toxicology studies and have them give you meaningful data on a study such as this where you get very good sort of efficacy trials but if you did that I think you would have a high comfort level, okay, in your ability to say that this level of immunogenicity of your vaccine has a clear-cut biological effect and that biological effect is not just preventing an infection in a tissue culture system but actually in an in vivo model of organized pulmonary epithelium where the virus has to transudate across the epithelium and actually be functional in the respiratory tract.

DR. WOOD: David Wood from WHO. Just following up from that in terms of what is meaningful data to get from animal studies I thought we heard this morning that the single most important endpoint was the reduction in virus titer in the challenge model which I didn't see highlighted on the slides here and I didn't hear Brian mention because I mean that seemed to me one of the clear things that came out of this morning that the single most important thing was the reduction in the amount of virus in the challenge system in an animal model. Are we in agreement with that?

DR.MURPHY: The point I was making is that you verify that, you verify the reduction in the virus titer by looking for a concomitant effect on immunohistopathology or quantitative PCR both of which would give you independent information on the level of reduction that you are seeing in the target organ.

DR. COUCH: I would like to add, too, that we don't do these human volunteer challenges anymore, but I think John would agree with me. I have always felt for a long time after I got started on these the biggest value of the volunteer challenge model was the relation, your ability to quantitate the effect on infection and it was the peak titer, the number of days of duration and the illness was secondary because of your ability to be precise with that quantitative virology and illness as we have said before, that is an artificial circumstance. Illness is an important measurement but the primary measurement there has always been the degree of the antiviral effect.

PARTICIPANT: (Inaudible.)

DR. COUCH: Nasal washes. There are no aerosol challenges done of volunteers. All are nasal challenges but quantitative virology has been done on nasal secretions, oral secretions, sputum and so forth. The nasal washes are the best and more reproducible for quantity than human volunteers.

DR. SUBBARAO: I think she is asking about in animals where would you like to see the reduction.

DR. COUCH: I don't do animals. You tell her that.

DR. MURPHY: I agree with Bob. I mean I think the human trials you can get quantitative virology. Almost all the points I was making in my talk related to quantitative virology in which you took not just the titer on one day and the peak titer which works perfectly well but you can actually get sums of titers which give you an area under your virus curve and it really gives you a good idea of the magnitude of the response.

That is one number and you can correlate that with your levels of antibody that you get. Now, that is in humans. Now, humans, the way you do the studies in humans differs considerably from animals because you sacrifice animals, but the other immunological thing the way it differs is that the virus that you have in the wash material you don't mix it. It is not a ground, it is not a homogenate in humans. So, you don't get the possibility of mixing. It is a direct readout of how much virus is present. In animals you have that compounding effect.

Now,what would I measure in an animal? Lungs and nasal turbinates. That is all you need to do. You do both, okay. You don't just do lungs. You do lungs and nasal turbinates. There are two major sites of disease with these viruses and Bob would agree that otitis media is a major effect of upper respiratory tract infection in these which you can prevent, right, Bob, with vaccine? So, you look at the nasal turbinates and you look at the lungs and you get your readout.

DR. SUBBARAO: Any comments?

DR. WRIGHT: I the rules particularly for regulatory but even for pathogenesis are that you have got to develop I think at this point a fairly uniform model. You have got a limited number of sites in which the studies can be done. It probably should be done in a similar way. You should take all of the vaccine candidates that are out there now and not just inactivated vaccines and in a uniform way put them through a model. I think that histology is important. Obviously the viral titer is important. I think we have all agreed that mortality is not a very good endpoint. You see an awful lot of murine flu papers that simply report mortality with nothing else reported and then you study as much as you can the immunologic correlates but recognize and I think we have seen some examples of that, you are not going to in any absolute way predict how these vaccines behave in man. You probably can establish some sense of priority and optimization of the vaccines and get an idea of whether some of the newer adjuvants work better in animal models and if they do and that correlates with an increased antibody response however you measure it that is obviously something to go for, but I think a very systematic uniform assessment of all of the vaccines in the animal model and I hate to put this on Terry and Kanta and so forth but it is going to be very important to do. It is not each person setting up his own model in a situation where there is some biohazard in the challenge virus and so it has got to be done in one --

DR. SUBBARAO: I agree. I think that one of the and I think Jackie said this in her slides, too, one of the problems with trying to develop a surrogate measure of efficacy from animal models is that the vaccines, the exact formulations that have gone into people have not been evaluated systematically in animal models. So, we don't have the data to bridge the two at all. So, I couldn't agree with you more.

I think we are going to move on to the next category of questions just in fairness so that we don't neglect the last two sets completely and Maria is going to take this.

DR. ZAMBON: Okay, we have a series of questions up there that have actually been posed from various people and some of those we have actually touched on and in regard to the amount of time and rather lively discussion that we have had I am proposing to start off in reverse order and ask Wendy what opportunity do you think has been missed in the development of an effective H5N1 vaccine over the last few years. There has been a lot of work with H5N1 vaccines. What do you think has been missed?

DR. KEITEL: If you will permit I think I will try to spin this as what opportunities do we have now which will imply maybe opportunities we missed earlier on and let us tie in the question above it, what lessons were learned from past pandemics that can guide development of an effective H5N1 vaccine. We have learned a lot of lessons from previous pandemics. We learned a lot about priming and so and so forth.

Dr. Couch has just pointed out another lesson we learned was being right on our speed. There was a very expeditious response to vaccine development in prior pandemics so that between 1997 and the present we now have a number of different entities around the world developing vaccines and we have some that are emerging as appearing more promising than others. Yet we have no real frame of reference because of the difference in formulations, the difference in dosages, presence or not of adjuvants, the diversity of laboratory assays, not only neutralizing which is extremely variable as we have heard about but even HAI which has a variability. We are not even using the same substrates from study to study.

So, moving from here I would like to see a more programmatic approach. I understand companies take a particular product and evaluate it and then move it on if it looks promising but certain non-pharmaceutical governmental agencies have an opportunity I think to approach it very programmatically. We know that the hemagglutinin for whatever reason appears poorly immunogenic relative to other influenza hemagglutinins, that the dosages that have been evaluated we know that it is likely going to require inclusion of an adjuvant unless we can as they say fix the hemagglutinin or figure out something about an assay methodology. So, what I would like to see now is if we are going to look at a clade 2 vaccine I would like to look at the various options that we have available to us simultaneously using uniform standardized laboratory assays so that we have a frame of reference for what one vaccine is doing relative to another.

I remind myself of the wonderful and rapid progress that was made in the development of the acellular pertussis vaccines and in the Hemophilus influenzae conjugate vaccines where we were able to acquire products from different manufacturers and put them into large field trials, rapidly generate information and you could make some sense from one study to the other, have at least a frame of reference for knowing how one vaccine was behaving relative to another. So, where we are right now is we have a handful of promising candidates that are actually being licensed or registered and are going into stockpiles. Can we get these and start looking at them in a single trial using validated and standardized serologic assays to see how they compare, one relative to the other, and then finally I guess I think we often tend to say that there is going to be the best vaccine or what is it going to be. I don't think that is true.

We are probably going to be if we ever succeed in getting an effective live attenuated vaccine, there are probably going to be large segments of the population for which that vaccine will be contraindicated. If these whole virus vaccines appear so promising actually cause severe febrile reactions then we would probably not tolerate severe but if they cause a lot of severe reactions maybe the subunit vaccine would be more appropriate and safer for use in infants and young children.

So, I would like to see rather than a linear development at this point taking what we have learned and moving on with the new clades and trying to compare within single trials using standardized serology compare the different formulations and approaches to vaccination.

DR. ZAMBON: Does anybody have a view on that? Is that the next logical step?

Brian, you look as though you are going to say something.

DR.MURPHY: Just something underlying in that which represents a real problem, not with the concept but right now there is no place that is making a purified HA reagent or the reagents that you need to develop your SOPs for such an analysis and I would think that one of the things that I would really suggest, I mean I really do suggest that the FDA because your job is to regulate these products, okay, it would be to support the development of assays, okay,which would be purified hemagglutinins, a microneut test, get the whole types of assays that you would like to have somebody actually run on new preparations and then I would do this for seasonal and for pandemic flu and with the idea in mind of using those assays and study sites such as Wendy's and Bob's and the whole VTU system to try and get some of the information that you are asking for which are the correlates of immunity but the problem is that one person is making the antigen over here. Another guy is developing a test over here. There is no uniformity in this and I think that there is probably a need for it. I think the fact that it doesn't exist indicates a deficiency. I think if you did both, made the reagents, developed standardized assays and then work out ways which would be really probably somebody else, the governmental sort of functions to get all the different products and do the comparative test, then at the end of that you will have a much better idea of what is going on.

DR. KEITEL: I think we heard from John Wood a really promising approach to look at, first of all to understand how much variability there is around the world in terms of the assays even when it maybe looks like they are similar in terms of their assays. So, you can do that. You can establish an international standard of the assays and they will need to be done in many different laboratories because no one single laboratory as the process moves forward can be responsible for doing them.

So, we do have a certain number of products and we have new clades that are coming forward and one way to standardize and get a frame of reference is to use a single laboratory and to test a number of products in the same clinical trial and we have probably five now that would be interesting to put into that type of a comparison.

DR. SUBBARAO: I was going to say that you probably know better than I do but I have heard that certainly with the HIV vaccine trials there are core laboratories that perform all the assays of a given type. So, there is more uniformity.

DR. WRIGHT: That is the analogy that is occurring to me as you speak and I see sort of you know people looking as though this isn't possible to do but it certainly is absolutely possible to do. A very standardized way is required for certain cells. There is a component to the extent that you want to look at cells drawing up to 300 ml of blood if that is what is necessary to accurately determine a cellular response, all of these things are possible and all the work is done in a very standardized way and virtually one or two laboratories do all of the assays and are very committed to that and so I think that that is where you want to go.

There is a danger if you don't know enough in making a decision that you are going to go in a very kind of a regimented way but I think I agree with Wendy that we are at a point where promising vaccines have emerged and they need to be looked at in a very uniform way.

DR. WOOD: David Wood, WHO. Can I just push Wendy and Peter on that point as to why you feel that doing a single study with the various different vaccines would be or is that the best way to get this data because as we heard from John Wood that the development of an international standard should that be effective will give us a way of calibrating the responses from the clinical trials in a uniform way? We would define an international unit as an example of an antibody response and we will be able to compare the responses from different trials and would that give us, I mean that may be a quicker way perhaps of getting some of this data and can I just sort of push as to do you really think that doing a single trial with all the different vaccines is going to be the best way to get this data?

DR. WRIGHT; I agree that an international standard and I think John showed that that you markedly lowered your variability when you adopted a single standard and corrected all of the or some of the interlaboratory variability and I don't know that you can design such a big trial. I don't know that it all has to be a single trial but it has to be a trial performed in a very uniform way with the analysis I would say in a very limited number of laboratories that cross validate themselves and that becomes almost so much work that it is almost better to do it in a single laboratory. I think both need to proceed simultaneously. The standards are very important.

DR. COUCH: This is repeating a little bit about what we said, but one of the strongest, I was around during these things. One of the strongest components, aspects of the 1976 swine episode if you will was that in the preparation for that there was an extreme degree of collaboration which you can call it critical if you want. We have not seen it in relation to the H5 problem but all of the investigators and the government agencies and the pharmaceutical companies gathered and coordinated an approach to that one and when we did the vaccine trials we did all of them compared side by side and the best way to avoid variation in assays is to do them all in one spot. Nobody would question that and for that particular assay sera from every volunteer in that whole trial went to CDC and they did all of the HAIs on everything. In addition to that in our laboratory I told you we have been doing neutralization tests for 40 years. We did our own HI and we also did neutralization tests and a number of people did other things. So, you are not prevented from doing things in your own laboratory as well but there was a central coordinated site that had the same data on every single vaccine from every location, and it was extremely well coordinated and comparison side by side is an awfully powerful way to do things.

DR. ALFONSO; I just wonder that rather than setting a whole new clinical study or something to test these I believe that people in this, the collection of samples that we have or the collections that you all have I am sure that there could be a sample size representative, large enough with the representative standards, the reference standards from WHO or any reference lab in which we can compare the intralaboratory variation and interlab variation. So, I think that among all of you there has got to be a good enough sample size that that standardization could at least be, we can take a look at that because I think we are, what we have been discussing in these 2 days, we are all in the book. We know it is happening but I think I see a lot of noise and I think we need to clean up what we have because I think we do have differences.

DR. KEITEL: I think that that is a third option which is you have an international standard. You do a trial with everything run at the same time in the same laboratory or we could actually make use of existing samples and perhaps select the dosage levels that look promising in terms of regimens or the ones that are actually going for registration on these products and get those serum samples collected together and run them all in a single assay as opposed to different labs with an international standard. I think an international standard solves certain issues across laboratories but I think the best assay is one that is done in the same laboratory and I would just like to make one other comment.

In the past results from the clinical trials have really moved out quickly and become a matter of public information and the US FDA has imposed upon clinical researchers a follow-up visit 6 months after receipt of the last dose of an experimental product.

Now, it doesn't sound too complicated and the actual information that is collected at that time at a minimum consists of and it can even be a telephone call, consists of collecting information about the occurrence of serious adverse events during that 6-month period after receiving the last dose of vaccine. So, it doesn't sound too ominous and complicated but what it does is it prevents us from releasing any information for many, many, many months after all the information is already available and all of the serologic assays have been run because we can't break the code and the investigator can't know and so this is in a massive way impeding the delivery of information to the international community about the performance of these vaccines and the primary endpoint of these studies in addition to safety and you usually whether you know what group they are in or not have a good idea about whether a vaccine was safe or not by the end of 2 months. The primary information is the antibody response after the receipt of the second dose of the vaccine but we can't do anything until usually 8 or more months after that time point. We can't release the information and so I would make a plea that we, if we need to keep the investigators blinded and collect this SAE information that we develop approaches that will allow the timely delivery of information which is already available about the immunogenicity of a product.

DR. COUCH: Let me second that, plea and it goes on to the next session which we are getting started but Suzanne Epstein said -- stand up. Are you sure? Okay, all right, then we have moved on to the next session and what constitutes an ideal clinical trial you talk to the people, the scientists and they will tell you what constitutes. Now, you talk to the biometrician and what constitutes an ideal clinical trial and that data is locked. You don't touch anything that until all of it is completed and in place and you see we sit there for 6 months waiting for that phone call to see if anybody got put in the hospital before that study is completed and that data can then be unlocked to look at. See what kind of time we are losing on something that we want to move on? So, that is an appeal that I am supporting Wendy on that. We would like for you to give some consideration to modification or that requirement which is hurting progression of developing data. Now, this one, the last session you see, I looked at the four questions and let me do the last one first and make a comment and then we have got people who do clinical trials down here that are going to answer the first two for me and that is overall sense that a program is emerging to systematically evaluate new vaccine candidates and the more you talk the more it will emerge and this is part of that. There has probably not been enough talking of what needs to be done and what could be coordinated. Can we coordinate candidates between Europe and US and I would say that is partly scientific maybe a little bit money and then maybe even a little political, but I think what we would all say is it is highly desirable to do as much of that as possible.

Now, to answer that in terms of can we, certainly we can do it to some extent and we ought to try to do it as much as possible. I don't think any of us would differ with that.

The data doesn't have to be developed here. It can be developed in Europe. It can be developed here and the questions can be coordinated and the more coordination that goes on the better it will be.

Now, what are the characteristics of an ideal trial and I need to make a comment, too. Tom Jefferson was unhappy with my remark this morning and I need to apologize because, is Tom here? Okay, I apologize, Tom. But I had said that we come down with just four trials. We are throwing away too much data, and he said that he didn't throw away anything and okay, we are not using the data. Then how did we get to the four trials, and he explained to me that we got to four trials because he was applying RCA criteria?


DR. COUCH: NRA criteria and I am ignorant of NRA criteria. Maybe there are some others and applying the NRA criteria to the analysis of trials restricted the quality trials to four and you see without knowing the NRA criteria I just have to say, "Gee, that is" -- Bob says, "Why didn't I ask him." I don't know if you want to give us a quickie or anything like that or if anybody else does but it restricts us to four trials. We need to look at it and think about changing it.

DR. JEFFERSON: It is WHO document, a very well written WHO document that summarizes the requirements across I think six NRAs, six national registration authorities, isn't that right, and I forget which ones they are. Well, I don't forget but I am being diplomatic and the characteristics of these, common characteristics are serology and mainly HAI like you said this morning.

So, what we did is we simply restricted the 330 studies in our database which are comparative studies; they are not all trials; they are comparative studies, things like cohorts and case controls with prospective goals, prospective cohorts, semi-randomized trials and randomized controlled trials. We restricted them all the way down applying inclusion criteria simply to try to answer the question that was set to us which is what are the evidence-based criteria and we come to the business of these four trials which I wasn't very good at explaining this morning. So, I apologize in turn.

DR. COUCH: And I didn't ask you, like you told me I should have

DR. JEFFERSON: The problem with the trial quality the study quality is that study quality impacts on conclusions. They are not independent. They are strictly linked. They are P 0.000001 linked, okay? The chi, I forget is something like 35. So, they are the same thing. The conclusions impact on the data and the data impacts or the quality impacts on the funder and the funder is tied to the impact factor of the journal that the study is published on and it is tied to the crude citation factor of that same article, the number of citations that an article gets in the 12 months after it has been published, the crude citation factor, okay? Do I need to go on?

DR. COUCH: No, you can stop anytime you want unless people have a question for you.

DR. JEFFERSON: In other words it is a study which drops half of its follow on without telling you why. You can't interpret it. So, you can't really say very much if you are trying to answer the question of what relationship there is between serological outcomes and field protection. How are you going to say? What are you going to say? Are you going to say that maybe it is 50; maybe it is 70 and maybe it 80 and maybe it is 90?

DR. COUCH: And part of the problem as you pointed out is your plea was for good clinical trials. It is for also a good description in your write-up of those clinical trials because all of us have contacted people and did said, "Did you?" and they said, "Oh, yes. I just didn't put it in the paper."

DR. JEFFERSON: There are two elements to this. There is the design of the trial which is important, vital but equally important is the reporting of the trial. That is because a third party who was not party to the trial and not party to any study could understand what is going on. You, yourself, said this morning that you actually were wondering about some of these studies, how they reached the conclusions. Somebody today said,"How do they reach those conclusions?" We don't know. If I don't know how they reached this conclusion how can I answer the question or how can we answer the question that WHO set us. The answer is we don't know. Judging those four studies there is notable uncertainty. Now, let us move forward, okay? One suggestion we heard today from Dr. Belshe is very simple. Instead of doing a single sera you do a paired sera. Now, paired doesn't seem to be a big problem so I understand and I have never done a big clinical trial. So, the expert tells us it is not a big deal to actually put in an extra. So, let us have a look. Let us have six or seven of these big trials that are going on at the moment or just about to start and let us have a look by inserting another look through the keyhole, again another expression that we use today. We can make some kind of sense of correlation between surrogates and real outcomes, okay? That is all I said.

DR. COUCH: That is fine, Tom. In relation to that first question, what are the characteristics I am just going to make one general, well, nobody questions that randomized placebo-controlled trials are the best when you are evaluating any clinical question not just a vaccine question just to be able to draw the conclusion and the generality, Wendy has heard this from me, look, when we get ready to do one of these we design the trial that we want to do that would be clearly the best trial we can do an then we start compromising and it is rare for us not to have to compromise you see, like you are not going to necessarily get two extra bloods on a young child and you are not going to necessarily be able to do all the lymphocyte studies on everybody but we start compromising and try to get as close to that random controlled complete analysis trial as possible.

Now, I have got experts on my list.

Wendy had some comments about the characteristics of an ideal trial and you can go on to the second one that includes outcome measures.

DR. KEITEL: Actually I am going to let Bob address this one.

DR. COUCH: And, Peter, you, too, and Ed. It has been a while but Ed has done these, too.

DR. KEITEL: We need standardized case definitions for whatever our endpoints are and I think the answer to the question on the previous page is yes, of course we could develop standardized case definitions. The Brighton collaboration has been moving towards this. They would have to be tailored for specific age groups but I think one component of a good clinical trial is that you have really well-defined case definitions.

DR. BELSHE: Let me just second that. I think what we are referring to here are what the characteristics of the ideal clinical trial designed to measure vaccine efficacy. That is an important addition and so we have already heard we want it to be a placebo-controlled trial. What that means is in the US that limits you to children over 5 and adults under 64 who don't have a high-risk condition if you are going to do a placebo-controlled trial and obviously you have to power your sample size large enough to measure the effect. Typically we use an attack rate of 4 percent in a placebo group and 1 percent in a vaccine to power our trial. We would want to measure effects around the vaccine with serum before and after vaccine and then some later time point to try and get proximate to the natural infection. Depending on the characteristics of the vaccine you might want to measure nasal washes or cell-mediated immunity but again that is totally dependent on the product you are looking at and then having some sort of well-defined laboratory endpoint, I, personally like culture positive influenza-like illness as the definition of the endpoint. I personally don't like PCR. The reason is you get a few false positives in PCR and it is false positives that drive your efficacy estimates down because they tend to occur in both groups and they dilute out your vaccine effect.

So get the best endpoint I like culture positive flu rather than PCR positive flu but those are some of the characteristics that I would argue for.

DR. COUCH: Peter and then Tom and then we are going to quit this question.


DR. WRIGHT: I agree with Bob. I think there is another type of trial and maybe we’re more at the other type of trial at least with pandemic vaccines a little and that is the small immunologically intense virologically intense if it happens to be a live vaccine trial that really informs and begins to rank your vaccine candidates and so it is a Phase I or even some people refer to a Phase 0 trial and I think a series of these can be done, can and should be done quite quickly and lead up to an efficacy trial. I think you have to be lucky. Three years with a cold-adapted vaccine initially yielded very good data on two of the components of the vaccine, a third had to be approached as was indicated by a live attenuated challenge. You obviously do need a lot of biostatistical input but I say that you have to be lucky because influenza is so unpredictable and you can be all set up to do a trial. You have to have product available well before the influenza season and that is a setting that we find ourselves in consistently of products appearing in September or October with the idea that we are going to give two doses of vaccine and measure response to the second dose of vaccine before influenza season.

We probably need to think about setting up trials in settings that employ both the Northern and Southern Hemisphere because otherwise you are using only 6 or 8 months of the year in which to do trials.

So, there are lots of things that go into it. Most importantly is a lot of collective thought and a very carefully drawn out protocol.

DR. JEFFERSON: Just another plea. I agree with the standardization of outcomes. We need to know what we are measuring and we need to somehow diminish the interrater variability.

The example of the Brighton collaboration is an excellent example that Wendy mentioned. The Brighton collaboration is a voluntary collaboration, international not-for-profit collaboration set up to standardize mainly safety outcomes with definitions for both the clinical trials and surveillance, and I would make the same plea for the efficacy and effectiveness outcomes.

I would like personally to see fewer death for all causes as outcomes and a few more relevant outcomes like hospitalizations and possibly hospitalization for PNI but even better something which is culture positive.

DR. COUCH: An objective outcome of illness like visit to a doctor, hospitalization, medication, things that you can quantitate and are not necessarily subjective.

DR. JEFFERSON: But are relevant.

DR. COUCH: And relevant.

DR.JEFFERSON: Like culture positive.

DR. COUCH: Any illness is relevant.

DR. BELSHE: Let me just expand on that just a little bit. I also like intent-to-treat analysis rather than according-to-protocol analysis particularly for flu trials because that is what is important where you are looking at the vaccine effect on everybody who enrolled regardless of whether or not they in fact got the vaccine. I like intent-to-treat analysis for this and typically the intent-to-treat analysis and according-to-protocol analysis are virtually the same in flu trials not for other vaccines but for flu trials they are and the other thing I like is to look at all culture-confirmed flu regardless of antigenic match because again that is the impact on the public is all flu not whether or not we guessed right and put it in the vaccine and so that is a couple of the nuances I would like to see in the trials.

DR. MURPHY: For the pandemic type of trials where you might not get a challenge you are still left with the question is this antibody or whatever is induced in the vaccinee, is it efficacious. It certainly can be built into these trials. You need to remove enough serum from some of the adult volunteers to do passive transfer studies so you can compare, you can directly compare how does the human antibody respond to this vaccine and how does that compare with your models. So, then you could start building your database on some of these parameters.

DR. SUBBARAO: So we put up the question that Jerry had left us with just before the coffee break and I think we have got parts of the answer.

Are you satisfied, Jerry or do you want some more discussion?

DR. WEIR: Do you have any more ideas?

DR. WRIGHT: I can just say and I think that other people who have done the neutralization test would agree that this is the ultimate biologic assay and has a tremendous amount of inherent, it is not variability and it does have variability, it is variables that are absolutely critical to make the test work and so this is a ten-fold more difficult test to do than the HAI, I really believe and we need to be very careful about making that, I think it is very important to look at but making that a surrogate that replaces the HAI at this point. I mean this is a small lab trying to do an assay but it is not easy.

DR. WEIR: (Off microphone.)

DR. WRIGHT: I think that exemplifies, it is a very nice first effort. It exemplifies the problem. It was all sort of voluntary. It came out of I think more than anything Jeffrey Shill's initiative to do it. There was no funding for it. Everybody did the test as they could do it. It took a long time to get the results back and they look terrible. So, I don't know.


DR. WEIR: But I was referring to the one that we are starting now.

DR. COUCH: The assay just has too many additional variables that the HI doesn't have and so you add more variables into an assay and you are going to get more variation from different people doing the assays and in the first place different neutralization assays are being done. If you want to pick one and say, "That is the one everybody has got to do," then that would narrow it a little bit but I don't want to do a new assay. I am in the biologic business not in the assay business.

DR. WEIR: With that if no one else has any comments on behalf of all of the organizers I want to thank the panelists, speakers, everybody that ponied up money. Anyway I think I speak for all of us. I think it has been very useful and I guess maybe we should all think about when to get together and do it again, but anyway I would like to thank all of you. I think we have done a really good job and I appreciate everyone's efforts.

(Thereupon, at 5:20 p.m., the meeting was adjourned.)

December 10, 2007 Transcripts

Page Last Updated: 01/26/2015
Note: If you need help accessing information in different file formats, see Instructions for Downloading Viewers and Players.