• Decrease font size
  • Return font size to normal
  • Increase font size
U.S. Department of Health and Human Services

Animal & Veterinary

  • Print
  • Share
  • E-mail

What Are We Learning From Campylobacter Molecular Epidemiology? by Collette Fitzgerald, Ph.D.

DR. FITZGERALD: Thank you Marie. Good morning everyone. It is a pleasure to be here. I would like to thank the organizing committee for giving me this opportunity to talk to you today.


As Maria mentioned, I am going to be talking about the molecular epidemiology of Campylobacter.


So just a general overview of what I am going to talk about today, just a brief introduction to subtyping, then I am going to talk about some of the Campylobacter outbreak data on PulseNet. I am going to focus a little bit more on an MLST project that we have been embarking on in my lab and then finish up with just two slides on an example of how we are using our subtype data in the context of outbreak investigations and then just some general conclusions.


So many, many subtyping methods have been described since the 1980s to trace sources of human Campylobacter infections and lots of lots of lots of subtyping data have been generated. Before the advent of the high resolution molecular subtyping techniques, I think we could kind of say that no clear conclusions could be made about tracing these sources. So why is this?

So there is a whole number of reasons. I think the first one is really just the scale of the problem. Campylobacter is one of the most common bacterial causes of enteric infections as we all know here in the U.S. and in other developed nations, despite the fact that most infections go undiagnosed and unreported.

As Dr. Tauxe mentioned yesterday, most human cases are sporadic in nature. So subtyping strains to indentify sources of infection therefore demands a substantial effort.

If you then combine that with the other impounding factors, there are many potential reservoirs for human infection and the fact that Campylobacters are genetically diverse organisms, advances in population biology have demonstrated genetic change in these organisms is really driven by recombination events that results in a semiclonal population structure.

So what does that mean to those of us who are typing and tracing? Well, it means that definitive linkage between any two strains is really a difficult task because genotyping results between related organisms can differ.

The upshot of all this is that really the epidemiology of Campylobacter is complex and challenging. Certainly for the last 19 years that I have been working on Campylobacters, they are keeping me on my toes.


So moving on to some of our PulseNet data, before I start talking about the content of the database, I just want to emphasize upfront an important difference about the Campylobacter PulseNet database that is different from the other PulseNet databases, and that is that our participants do not do routine subtyping of their Campylobacters and therefore that means that we are not using this database to detect disease clusters. Rather, we are using it for confirmatory subtyping of strains when outbreaks are detected by other means.

This table shows you just the current figures that we have in PulseNet and these figures are current as of this Monday. We have just over 7,000 entries now in the national database. Sma1 and Kpn1 are our primary secondary enzymes to Campylobacter PFGE. We have a total of 1,697 unique Sma1 patterns, 776 Kpn1 patterns, and 914 unique combined Sma1-Kpn1 patterns.


So on to our outbreak data. We have 682 entries associated with 62 different outbreaks in the database. These represent selected outbreaks for which we have strains available for PFGE. These outbreaks have occurred between 1980 and 2010. The sources of these outbreaks are listed here. Milkborne, foodborne and waterborne outbreaks account for the majority of these outbreaks. A few are linked to food handlers. Two were associated with animal contact and five were unidentified.

So there were a total of 150 different combined PFGE patterns observed among these outbreak strains. A single combined PFGE pattern was seen for each of 38 of the outbreaks and multiple combined PFGE patterns or multiple strains were observed for the other 24 outbreaks.


Now given the large amount of genetic diversity within Campylobacter which I have talked about, it was somewhat surprising to see six strains were responsible for multiple outbreaks. There are examples of three of those here.

So a strain with this combined PFGE pattern was associated two Mexican restaurant outbreaks in two different states over two different time points.

A second strain was responsible for three different raw milk outbreaks, again from different states over different times. Then we have this third strain here which seems to be responsible for multiple raw milk outbreaks over different states and again from 2003 right up to two outbreaks which occurred just this year.

So we are definitely giving a little more thought you know where are we seeing these strains from other places.


This strain with PFGE pattern Sma1 pattern 8, Kpn1 pattern 28 was associated with a tetracycline resistant C. jejuni clone which was described Dr. Sahin and colleagues from Iowa State University in 2008. This particular clone was important because it is commonly associated with outbreaks of ovine abortion here in the U.S.

In their study, they showed that 66 of 71 strains from 33 different farms in Iowa between 2003 and 2007 also have this same PFGE pattern as our outbreak strains. All had the same MLST type. It was ST8.

Representative strains from that study were sent to our lab and we did Penner serotyping on them. They are all Penner 1,8. Dr. Sahin described this as their “SA” sheep clone.


So where else have we seen this strain? Well, an additional 46 strains in the national database also have this pattern. 43 of them were sporadic isolates. Human isolates from five different states isolated between 2004 and 2009. We have one ovine and one bovine isolate also with the same pattern.

We went back to the five states and asked them did they still have the isolates in the freezer. Thankfully, 28 of the isolates were still available for further characterization. These were also all ST8 and Penner serotype 1,8.

So the value here of the subtyping data is that we are really able to link that this clone responsible for the majority of Campylobacter-associated sheep abortions in the U.S. is also associated with milkborne outbreaks and sporadic cases of human infections.


So I am going to move on now to just a few slides on out MLST data. This project is being coordinated by Dr. Patrick Kwan from my group. He is here at the meeting. He is actually sitting here at the back. If anyone has questions for him related to the details of the study, please feel free to talk to him in the breaks.


So just some background for the MLST. There has been an explosion in MLST data for Campylobacter jejuni since this method was first described in 2001. These studies are showing, you know, a high degree of overlap between genotypes recovered from human and food animal isolates. There is an association of particular genotypes with certain hosts.

Because of this association, molecular data can be used to attribute cases of human infection to particular host sources.
Several countries have been actively engaging in this area. Source attribution models based on molecular genotyping has proved valuable for identifying the most likely sources of human campylobacteriosis in Europe and New Zealand.

Really the take home from all of these studies, I think there are two clear messages that come through. Chicken has been identified as the most important reservoir for human infection but the relative contribution of the other sources to the burden of disease certainly does require further clarity.


So what about the situation here in the U.S? Well, this shows you the relative contribution by country to the PubMLST database which is the online resource for submission of your MLST data. You can see the majority of data has been provided by the U.K. where the method was developed and where they have been actively engaging in this area. There is only limited date from the U.S.

So we have little information on the background of circulating C. jejuni genotypes causing sporadic infection here in the United States. There are only four published studies. These are all great studies but there is only less than 50 isolates in each of these studies, so we have limited data from human infection and food-producing animals and no data from wild animals or the environment.


So the aim of our study was really just to take stock, get a sense of the baseline for what diversity are we seeing of the genotypes circulating here in the U.S., so to determine the major genotypes causing human infections here in the U.S.

We are focusing on two types of both sporadic and outbreak isolates. We initially started with characterization of isolates from the 1999 FoodNet Campylobacter Case Control Study. This is a valuable resource of isolates because these isolates come with detailed clinical, epidemiologic and antimicrobial resistance data. We are also looking at isolates from 2008. These are the most recently linked isolates between FoodNet and NARMS so they can give us a good ten-year comparison.


So where are we today? Well this project is still a work-in-progress so what I am showing you for the next couple of slides are still preliminary data. Some of the numbers are small, but we are expanding on that.

We have looked at 471 isolates to date, 296 from the 1998 collection. Seven FoodNet sites participated in that case control study and those are shown here, Minnesota, New York, Connecticut, Maryland, Georgia, Oregon, and California. Antimicrobial resistance data, susceptibility data is available on all of these isolates through NARMS. 156 have been characterized so far.

On the 2008 dataset, we requested isolates from all of the FoodNet sites and we have one non-FoodNet site participating right now. Wyoming has an interest for a special project going with Campylobacter so we are working with Wyoming as well. We do hope to include many more FoodNet sites as we continue on with this project.

So far we have analyzed data from the number of states that are listed here. Again, we are comparing the MSLT data to the susceptibility data.

I do want to thank Dr. Jean Whichard and Kevin Joyce from the Human NARMS lab at CDC for all of their help and support in getting our hands on these isolates for the characterization. We really appreciate their assistance.


So on to some of the results. This shows you the distribution of the clonal complexes among the 471 sporadic isolates. As expected, we are seeing a lot of genetic diversity within the dataset. 71 sequence types and 24 clonal complexes were identified. I supposed if you are not familiar with MLST data, the clonal complexes are really just groups of related sequence types of strains. That is sort of our epidemiologic unit for looking at this type of data.

So you can see here from the dataset that we have four major clonal complexes which accounted for 62 percent of the isolates. That was ST-21, 45, 353, and 48. Across the board the complexes that we saw in our study were found in both animal and human isolates from past studies outside the U.S. However, there are differences in the prevalence of the STs. For example, 353 which is common in our sporadic dataset is fairly uncommon among human isolates in the U.K.


Again, this is the same dataset but just broken out by year. What we see here is variation in the prevalence of the complexes between the two different years. That really shows us that the prevalence of these complexes is not static. It is dynamic over time as we might expect.

So in 1998, ST-48 was the most prevalent ST and then in descending order for these STs, but by 2008, this had dropped significantly and clonal complex 21 had taken over as the most predominant ST.


So if we break this down by the distribution of those four major complexes by state, this is the 1998 data. This is the 2008 data. These is a lot going on here but really what we are saying is that we are seeing state to state variation in the prevalence of these complexes.


We have the most data so far from Minnesota and New York, about 150 isolates from each. So focus a little bit more here on those two. Again, we are seeing these site-specific changes within Minnesota. Prevalence for the top three complexes in 1993 dropped while ST-21 complex increased in 2008. For New York, the prevalence of ST-48 complex dropped sharply while the other three complexes rose.

So our take home message here is that prevalence of these complexes not only differs between states but also within the states over time, as we might expect.


So looking at some correlation here between the complexes and ciprofloxacin susceptibility, a key question is, are there specific lineages of C. jejuni that are more resistant to ciprofloxacin? If we look at the four major complexes here, we have compared 317 isolates with the susceptibility data to the MLST data and you see that isolates from the four major complexes were resistant to ciprofloxacin, but we have two STs, 353 and 21, that appear to be more frequently associated with cipro resistance.


If we break that down again by the years, what you see here for the different four complexes, you see here that isolates from all of the major complexes showed an increase in resistance over the ten-year time period, but there was a notable sharp increase in the cipro resistance among the 353 complex isolates.


So moving on to look at some of the outbreak data, we have characterized now strains from 18 outbreaks to date. What you see here is the clonal complex, the prevalence in sporadic data, the number of outbreaks that were associated with the different complexes, and just a little description of the outbreaks.

The take home from here is that ST-21 is the complex most commonly now associated with sporadic infection, also caused most of the outbreaks. The majority of these outbreaks were raw milk-associated outbreaks.

ST-48 and 353 were much less -– were less –- were more uncommonly associated with the outbreaks. 48 only caused one of these outbreaks, even though they are circulating with high prevalence in the sporadic infections.

So we only have 18 outbreaks characterized so far. We are going back to look at all of the outbreaks in PulseNet plus we are prospectively asking our state lab partners to send us isolates. So the picture here may change over time as we increase the number of outbreaks, but let’s just wait and see how that turns out.


So the next steps for our project, we will continue to characterize isolates from the 2008 FoodNet sites. We would also like to add in additional isolates from non-FoodNet sites. We are going to continue to characterize outbreak isolates in collaboration with PulseNet and OutbreakNet.

Really, then once we have a sense of what is going on on the human side, we would like to move into phase II which is collaboration with partners. We need to have a discussion of inclusion of strain subtype data from farm animals, retail meat, and environmental samples.
We are showing here MLST data. We are very mindful of the limitation of looking at seven housekeeping genes when we have a whole genome worth of data. The microarray data, comparative genomic data, whole genome data, all of that hopefully we just take stock of that data and see if we can include that in our source attribution models.

Also collaborating with academia on the source attribution and analytical aspects of the project so that the models that we are using we can keep fine tuning those models.

So by using the molecular data in combination with epidemiologic and phylogenetic analysis really means that we need a coordinated multidisciplinary approach if we are going to attempt to do source attribution of Campylobacter here in the United States.


I am just going to share with you two slides highlighting how we are currently using our outbreak data. We assisted out colleagues in Alaska back in 2008 with an outbreak that was associated with consumption of raw peas that were contaminated by Sandhill cranes.

There were 99 human cases, 54 that were lab confirmed during this outbreak. We did an extensive environmental investigation as part of this outbreak. We found that 16 of 42 samples were positive for Campylobacter. We did five colony picks from each sample, used four different isolation methods and we found that we came up with a total of 125 C. jejuni isolates, and eight C. canadenesis isolates which only came through with filtration only.

On the birds what we found was there was up to seven different subtypes per bird sample. So these birds are certainly carrying multiple Campylobacter strains in them.

We also found, which is an aside to what I am talking about here that the molecular types that we recovered from these birds really varied depending on the isolation methods that we used. That certainly has implications for our source attribution work. This may be discussion for another day.

But PFGE was done in real time and as part of this outbreak. 15 human isolates with four different PFGE patterns were indistinguishable from isolates from four environmental samples, so two pea samples and two crane feces samples by PFGE.

So we had a link between the human isolates and the birds and the peas, but the question we asked ourselves was, well these are 15 of the human cases, what about the other 39 cases? Can was attribute the other human cases that didn’t match by PFGE also to the Sandhill cranes?


So we did MSLT on all of the human and environmental isolates associated with the outbreak. What you see here is a radial neighbor-joining tree showing majestic genetic relatedness of the outbreak isolates comparing it to the PubMLST data.

So what you see here which I have shown you red -- the red dots are actually the crane isolates. The orange dots are our human isolates. The other dots on the tree, the blue are wild birds, the green are human isolates, and the pink are chicken isolates, all from the PubMLST database.

What you can see is that the crane and the human isolates all cluster together with each other, but they also clustered with the wild bird isolates from the PubMLST database which showed and supported our hypothesis that the Campylobacter jejuni isolates from these human cases were of avian origin.

What we would like to do is actually compare our data to additional data from isolates from here in the U.S. We currently do not have that data available at the moment, so we have got some data gaps. So that is why we compared it to the PubMLST data.

I think being able to compare it to national data here from the U.S. will take the uncertainty and buy us out of the results from attribution work. So I definitely think that is an area we should all work on together.


So just some general conclusions are the strength of our PFGE data of doing PFGE analysis if Campylobacter strains in PulseNet lies in having one-day rapid protocol that we use in combination with epidemiologic information to facilitate the timely investigation of outbreaks.

C. jejuni isolates from human infections exhibit a wide range of variation. This is apparent whatever subtyping strategies are adopted. Stable clones do exist.

Our date is showing it on the human side, Shaohua showed it yesterday on the retail meat side. So we are looking forward now to comparing our data on the retail meat, and even with the VetNet data.

We are certainly starting to see a clearer picture which is starting to emerge on what genotypes are causing outbreaks and sporadic human infection here in the U.S.

Source attribution models require broad reference populations from a full range of possible sources and reservoirs in order to minimize uncertainty and bias in attribution results that we generate.

So there really is a clear need for us to take stock of our currently available molecular data and our strain collections. Which sources are they from? When were they isolated? Where were they isolated? How were they isolated? What are our holes? Let’s take stock of the data and really let’s collaborate is really sort of the take-home message from this.


I just want to acknowledge a few people, Patrick Kwan for all his hard work with the MLST data, Monica and Yueren who are also here in the audience. They are our summer interns working on their MLST projects. So thank you for all their help. Then to our state public health lab partners, a huge thank you because without them we would not have the isolates to do our work. Thank you.


DR. KARLSSON: Thank you Collette. If it is okay, we would like to save the questions for the end of the session.
Our next speaker is Dr. Lucie Dutil from the Public Health Agency of Canada and CIPARS. She will be presenting on Ceftiofur Resistance in Salmonella enterica Serovar Heidelberg from Chicken Meat and Humans.