• 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

Salmonellosis From Meat by Heather Green, Ph.D., M.S.

DR. GREEN: Thank you Shaohua.

(Slide)

So the rationale behind this project is that commonalities in serovar frequencies have led researchers to believe that retail meats can contribute to a majority of a human non-typhoidal salmonellosis.

Here you see data from the 2007 NARMS survey and you see that Typhimurium, Enteritidis, Heidelberg, and Montevideo which are highlighted in yellow are found in both populations. This data indicates that Salmonella from human and contaminated meat have slightly different but overlapping distributions.

(Slide)

So in the study that we have conducted, we wanted to get a more accurate picture of the relatedness of Salmonella in retail meats and humans at a molecular level. So we did this by looking at Salmonella PFGE subtypes. The idea was maybe that in examining the relatedness of Salmonella patterns found in contaminated meats and infected humans, we might be able to estimate the proportion of salmonellosis that is acquired through consumption of contaminated meat.

(Slide)

So the primary data source for this analysis is PulseNet which most of you are probably already familiar with. It is the national molecular subtyping network for foodborne disease. It is used to detect foodborne disease clusters by PFGE which facilitates early identification of outbreaks.

The members of the PulseNet network include the more than 50-state public health and local public health laboratories, as well as laboratories from the regulatory agencies and this is all coordinated by CDC.

PFGE fingerprints of human isolates are uploaded to PulseNet by the state and local public health laboratories on a daily or weekly basis. Then pattern names are automatically assigned and then confirmed by CDC. The advantage of using the PulseNet network is that the SOP’s used by PulseNet allow us to differentiate most serovars using the primary enzyme or the Xba1 enzyme only.

(Slide)

The study also utilized date collected by NARMS and you are already familiar with the three arms of NARMS. Shaohua went over that yesterday.

(Slide)

You are also probably -– well you are familiar also with the sampling scheme for NARMS as that was also gone over yesterday. But I did want to focus your attention on this last point. All of the isolates are forwarded to the Center for Veterinary Medicine at FDA where they undergo confirmatory serotyping, susceptibility testing, and PFGE subtyping. Those PFGE fingerprints that are generated by the laboratory at CVM are uploaded into PulseNet.

(Slide)

So basically to recap, the retail meat data comes from a subset of PFGE fingerprints that are uploaded to the PulseNet database by CVM for the retail meat study. On average, because we only get about seven percent of meats sampled by the NARMS Retail Meat program as being positive for Salmonella, we decided to combine isolates coming from years 2002 to 2007.

The human data came from the subset of PFGE fingerprints that are uploaded to PulseNet by the more than 50 state and local public health laboratories that are part of the PulseNet network and they came from the same timeframe, 2002 to 2007. Here we were looking only at sporadic clinical isolates because they represent an unbiased sample of the population that is under study.

(Slide)

So, the summary of isolates. In our analysis, we identified over 1500 isolates from the NARMS retail meat study representing 53 serotypes and 479 unique Xba1 patterns from the four retail meat sources. These were coming from the ten states that were part of the NARMS retail meat study during the time period of the data that we were working with.

For the human data, there were over 125,000 human sporadic isolates which is almost 100-fold more than retail meat and that presented some problems later which I will explain a little later. They represented 276 serotypes and over 19,000 unique Xba1 patterns from 50 states.

Among the ten serovars that were most commonly associated with human infections between 2002 and 2007, four were found among the top serovars of retail meats. Those were Typhimurium, Enteritidis, Heidelberg, and Montevideo.

(Slide)

So this table shows the top sixteen retail meat patterns that were identified in the PulseNet database. So this is the pattern name. The first three letters indicate or represent the serotype that this pattern falls under and then the second three letters represent the enzyme that was used to generate this pattern. The last four digits are the isolate identifier.

Then this is the number of isolates that had this particular pattern name in the NARMS retail meat database. This is where this pattern ranked within the database.

Then over here you have the number of human isolates with this pattern name in the human database and this is where they ranked in the human database.

We found that all top 16 retail meat patterns were also found in sporadic case of foodborne illness in humans, however, a majority of the top patterns for retail meat isolates ranked very low within the human subset.

This was not surprising because there are so many more human isolates, as I mentioned earlier. There is almost 100-fold more human isolates than there were retail meat isolates. Also, the human isolates were much more diverse. They were over 19,000 unique Xba1 patterns.

So we tried to figure out how we could look at this data differently in order to assess if there is overlap at he molecular level.

(Slide)

So what we did was we regrouped this table, so this is essentially the same data but now it has been regrouped into serogroups. So when we did this, we found that most of the retail meat patterns were also among the top patterns isolated within the same serotypes in humans.

So for instance, the most frequently isolated pattern was in the Kentucky serogroup. Within the Kentucky serogroup was the third most frequently recovered pattern within the Kentucky serogroup in humans.

(Slide)

So the high prevalence of patterns within serogroups of both human and retail meat subsets might actually indicate that there is an epidemiological linkage between the two sources.

As I mentioned earlier, commonalities and overlap in serovar frequencies have led people to believe that retail meats contribute to the majority of human nontyphoidal salmonellosis and the same can be true of PFGE subtypes.

But there were some exceptions and these are circled here. The most frequently recovered Typhimurium patterns from the NARMS retail meat database ranked very low within the human subset within this Typhimurium serogroup and the same with the Saintpaul pattern which was the second most frequently recovered Saintpaul pattern in the NARMS database. It ranked 95 within the Saintpaul serovar group in the human database.

There were a number of reasons that we figured this might be occurring. That could include lower pathogenicity of these strains or higher susceptibility, the cooking, or preparatory methods. We have also found some preliminary evidence indicating that these strains have limited geographic distribution which might also explain their low prevalence in the human subset.

That is basically what I have just said but it is also printed in your packets in your folders so you get the take home message.

(Slide)

So we also did a crosswise analysis and we looked at the top patterns in the human subset and looked for their respective rankings in retail meats. Here in this chart the patterns are already in serogroups and we found that six of the top patterns are not found in retail meats.

Patterns that are more prevalent in humans than in retail meats could have been missed because the sampling for retail meats is relatively small compared to the amount of retail meats that are sold nationwide.

The reason that these six patterns are not in the NARMS database could also be because they represent transmission routes other than retail meat.

That point is exemplified by this pattern which not only is this pattern not found in the NARMS retail meat database, but this entire serovar is not found in the NARMS retail meat database, which strengthens the common idea that Javiana is transmitted through other nonmeat sources.

(Slide)

So we also looked at the percent of human patterns that are shared with retail meats. We did this because we wanted to be able to make some statements about the proportion of salmonellosis that may be acquired through consumption of contaminated meat.

So looking at the first row, of the Heidelberg isolates, there were 52 unique NARMS patterns. Five were not found in the human subset and 47 of these patterns were also found in the human subset. Those 47 patterns represented 4,941 isolated in the human subset. When you take that and divide it by the total number of Heidelberg isolates in the human subset, you get 75.6 percent and that is the percent of human isolates that share patterns with retail meats.

So looking at this table, we see that retail meats do share a significant number of Salmonella patterns with humans and these numbers seem to vary by serovar. The top four have been highlighted because there are at least 50 percent of unique Xba1 patterns found in humans that are shared with retail meats for serovars Heidelberg, Enteritidis, Hadar, and Dublin.

The percentage of patterns shared between humans and meats can actually give us a rough estimate of the maximum number of cases that may be attributed to consumption of retail meats. But in order to do that extrapolation, we would have to assume that these patterns are only found in meats and humans which likely is not the case.

We would also need additional data on the relative consumption of each of the meat commodities sampled in order to estimate the true risk of infection. But this is a starting point for us and we are hoping that we can include some of this data.

(Slide)

This is actually the chart that is continued for the other top serovars found in retail meats and humans. We are hoping we can include some of this data in a mathematical model in the very near future.

(Slide)

And so then finally we found that Salmonella strains as defined by their PFGE patterns appear to have different affinities for different sources. So here we looked for Salmonella patterns that were found only or exclusively in one meat source and for this we instituted an arbitrary nine isolate cutoff limit.

(Slide)

This table lists the exclusive patterns and the number of isolates that were found in a particular meat type. You only see ground turkey and chicken breast here because there were the only sources in which isolates met that nine-isolate cutoff limit. There were too few pork chop or ground beef isolates in our database.

So was also looked for these patterns in the human subset. From most of the PFGE patterns, there were very low numbers of corresponding isolates in humans. In some cases, less isolates in humans than there were in retail meats.

We thought that maybe this is because retail meat patterns found exclusively in one NARMS retail meat source with very low numbers of corresponding isolates in humans could suggest causality.

Then finally we conducted a national PulseNet database search for these patters in other foods that were not chicken, turkey, beef or pork. We found that this particular Mbandaka pattern is also found in animal feed and that is really interesting because it might actually represent a farm-to-fork link.

(Slide)

So the limitations of this study, there were actually some large ones, but one of them being that the sample sizes of the databases were vastly different. So again, the human database was about 100-fold -– contained about 100-fold more isolates that the NARMS retail meat subset of the PulseNet database. That prevented us from doing some sophisticated statistical analysis.

We also could not assess the source of contamination so the data that we are collecting from the NARMS retail meat study we are not sure if these retail meats are contaminated at the retail level, at the farm level, or at the distribution site level.

Furthermore, the analysis was limited to one enzyme which affected the discriminatory power. Again, we only used Xba1 enzyme. If we had used Bln1, the second enzyme, or even if we had for perhaps enteritidis gone out to using six enzymes, we would have gotten some more precise estimates, but then the drawback of that is that we might have also had a lot of human isolates that were unaccounted for because it is too discriminatory.

Furthermore, something that I did not list here, but which is obvious is that we are only looking at retail meats. We are not looking at any other food sources so that when trying to estimate the contribution of retail meats to total salmonellosis based on all different foods that can be consumed, we are only looking at a small portion of that percentage.

(Slide)

Okay, so the conclusion from the work that we have done is that even though within the recent past there is increasing evidence that other food sources such as fresh produce may be attributed to Salmonella infection, the similarities that we found between human and retail meat patterns support the idea that contaminated meats may continue to play a major role in the transmission of at least some Salmonella serovars to people.

(Slide)

These are the people that have helped me generate this data. If you have any questions, I would be happy to answer them during the break or at any other time during this talk.

(Applause)

DR. ZHAO: Thank you Heather. Okay. Our last speaker is Dr. Rebecca Lindsey. She is a microbiologist from USDA and today she is going to talk about the Inc A/C Plasmids in Salmonella enterica.