Previous Article | Next Article ![]()
Journal of Clinical Microbiology, March 2007, p. 865-873, Vol. 45, No. 3
0095-1137/07/$08.00+0 doi:10.1128/JCM.01285-06
Copyright © 2007, American Society for Microbiology. All Rights Reserved.
,
Department of Food Science, Cornell University, Ithaca, New York 14853,1 Wadsworth Center, New York State Department of Health, Albany, New York 12208,2 New York State Department of Agriculture and Markets, Albany, New York 122353
Received 22 June 2006/ Returned for modification 31 August 2006/ Accepted 17 December 2006
|
|
|---|
|
|
|---|
Human listeriosis outbreaks are often difficult to detect, since cases associated with a single outbreak may be geographically dispersed (as illustrated by an outbreak involving patients in as many as 24 U.S. states) (21, 34) and may occur over long periods of time (as illustrated by an outbreak that occurred over more than 5 years) (3, 9). The use of molecular subtyping methods to link human cases that occur over time and space is thus often critical for the initial detection of human listeriosis outbreaks (2, 43, 51, 53). Molecular subtyping methods such as ribotyping (8) and pulsed-field gel electrophoresis (PFGE) (22) have been shown to exhibit much higher discriminatory power than serotyping, which differentiates only 13 L. monocytogenes serotypes (17). PFGE is generally recognized as the most discriminatory subtyping method for L. monocytogenes (22), and widespread and consistent use of PFGE for the routine subtyping of human L. monocytogenes isolates has been shown to facilitate improved detection and control of human listeriosis outbreaks. In particular, the exchange of bacterial PFGE patterns through PulseNet, a national network of public health and food regulatory agency laboratories coordinated by the U.S. CDC (51; http://www.cdc.gov/pulsenet), has shown considerable success (12-14, 20, 21) in identifying and curtailing food-borne listeriosis outbreaks. The importance of PFGE for human food-borne disease surveillance is likely to increase as PulseNet is being expanded internationally (52; http://www.cdc.gov/pulsenet).
In addition to their importance for outbreak detection, molecular subtyping and characterization methods have also provided a better understanding of the population genetics of L. monocytogenes. Specifically, a number of subtyping methods have shown that L. monocytogenes strains can be grouped into at least three distinct genetic lineages (termed genetic lineages I, II, and III) (44, 45, 59). While most subtyping studies of L. monocytogenes have defined two lineages (sometimes also referred to as divisions; see reference 7) with similar serotype groupings, the nomenclature of these lineages is not always consistent. Serotypes 1/2b, 4b, and 3b consistently group into one lineage (37), which has been designated lineage I by us and others (e.g., references 57, 58, and 59) while also being referred to as division II by some investigators (e.g., reference 7). Serotypes 1/2a, 1/2c, and 3a group into another lineage (37), designated lineage II by us and others (e.g., references 57, 58, and 59) while also being referred to as division I (e.g., reference 7). While a third lineage (termed lineage III) has been described as including serotypes 4a and 4c, as well as some serotype 4b strains (37, 57, 59), a recent study indicates the existence of multiple lineages within L. monocytogenes that include isolates with these serotypes (45). Different studies (23, 41) have shown that lineage I strains are generally overrepresented among human isolates, while lineage II strains appear to be overrepresented among food isolates (41). Lineage III strains are rare but have been shown to be overrepresented among isolates from animal listeriosis cases (29). Molecular subtyping studies have been able to identify further specific L. monocytogenes subtypes and clonal groups that appear to be associated with human listeriosis outbreaks and are common among isolates from human listeriosis cases (23, 29, 30). Kathariou (30) specifically designated three L. monocytogenes epidemic clones, (i) epidemic clone I (EC I), linked to listeriosis outbreaks in Nova Scotia (1981), Massachusetts (1983), Los Angeles (1985), Switzerland (1983 to 1987), Denmark (1985 to 1987), and France (1992); (ii) EC II, linked to two outbreaks in the United States in 1998 and 1999 and 2002; and (iii) EC III, a serotype 1/2a strain linked to a single outbreak in the United States in 2000. Increasing evidence indicates that many human disease-associated subtypes, including those representing human epidemic clones, are not only found in foods and food-processing environments but may also be present in urban and natural environments as well as in farm environments (6, 36, 39, 47). Molecular subtyping data have also shown that L. monocytogenes can persist in processing environments for considerable time periods (up to more than 10 years) (30, 43) and that human listeriosis outbreaks can be traced back to persistent contamination by the outbreak subtype in the source plant. Subtype data for L. monocytogenes isolates from foods and food-processing plants can thus sometimes help detect potential outbreak sources (as shown in a listeriosis outbreak in Finland linked to contaminated butter) (33), even though traditional epidemiological methods are still critical in linking a human outbreak to a specific source (58). Considering the broad distribution of L. monocytogenes, including human disease-associated strains, in different environments, it is thus critical to develop a better understanding of the L. monocytogenes PFGE type diversity across populations from different sources in order to further increase the utility of standard PFGE typing for the identification of human listeriosis outbreaks and their sources.
In this study, we used PFGE to characterize 495 temporally and geographically matched L. monocytogenes isolates in an effort to better understand L. monocytogenes PFGE type diversity and the ecology of different L. monocytogenes PFGE types across populations from different sources, including human clinical cases, farm animals and farm environments, and foods, as well as urban and natural environments. Specifically, the goals of our study were to (i) characterize PFGE type diversity among L. monocytogenes isolates from different sources (i.e., human clinical cases, foods, ruminants and ruminant farm environments, and urban and natural environments); (ii) identify PFGE types associated with specific sources; and (iii) characterize the distribution, ecology, and diversity of human disease-associated L. monocytogenes PFGE types with a specific focus on PFGE types representing previously identified epidemic clones.
|
|
|---|
PFGE. PFGE of all isolates was performed using the standard CDC PulseNet protocol (22). Briefly, isolates were grown on brain heart infusion (Difco/BD, Sparks, MD) agar plates at 37°C for 18 h. Bacterial cultures were embedded in 1% agarose plugs (SeaKem Gold agarose; Cambrex, Rockland, ME), lysed, washed, and digested separately with the restriction enzymes AscI and ApaI for at least 5 h at 37°C and 30°C, respectively. Size separation of restricted DNA fragments was performed for 20 to 22 h in 1% agarose gels by using a contour-clamped homogenous electrical field Mapper XA (Bio-Rad Laboratories, Hercules, CA); voltage was set at 6 V/cm with switch times of 4 s to 40.01 s. XbaI-digested Salmonella enterica serotype Braenderup (CDCH9812) DNA was used as a reference size standard (27). Pattern images were captured with a Bio-Rad Gel Doc and the Multi Analyst software version 1.1 (Bio-Rad Laboratories). PFGE patterns were then analyzed and compared using the BioNumerics version 3.5 software (Applied Maths, Saint-Matins-Latem, Belgium). Similarity clustering analyses were performed with BioNumerics by using the unweighted pair group-matching algorithm and the Dice correlation coefficient with a tolerance of 1.5% (22).
Data analyses. Associations between general source categories (i.e., human, food, farm, and environment) and PFGE types were evaluated using the chi-square test of independence. For these categorical analyses, PFGE types with fewer than five occurrences were pooled into one category termed "rare." Fisher's exact test was used for analyses if the expected frequency in any cell was less than two or if more than 20% of the expected frequencies were less than five. P values for exact tests for large contingency tables (e.g., 14 by 4) were determined using Monte Carlo simulation. All statistical analyses were performed using SAS 9.1 (SAS Institute Inc., Cary, NC). Due to the large number of observations investigated, it is possible that the probability of type I error was inflated; however, as previously described (23, 46), rather than generally lowering the significance threshold by adjusting for multiple comparisons, we have provided observed P values for statistical tests to allow readers to evaluate levels of significance according to their own criteria. Specifically, statistical significance is reported at three levels: P of <0.05, <0.005, and <0.0005. In addition, P values with Bonferroni's correction are reported. Overall, a total of 17 and 43 chi-square tests were conducted using lineage and PFGE distribution data, respectively. The Bonferroni's correction-adjusted P value cutoff should thus be lowered to 0.0029 and 0.0011 for lineage and PFGE distribution data, respectively. Thus, P values of <0.0005 should be considered to indicate significance even after Bonferroni's correction.
Simpson's index of discrimination (D) with 95% confidence intervals was calculated as previously described (25, 26). Two subtyping methods were considered significantly different in their discriminatory powers if 95% confidence intervals for D values for both methods did not overlap.
Spatial analysis. A New York State map was used to visualize the geographical origins of all isolates classified into a given PFGE type that occurred at least five times among the 495 isolates characterized. While information on the county of origin (e.g., the locations of source farms, food establishments sampled, or patient residences) was available for the majority of isolates, geographical origins were plotted using five regions within New York State (western, central, northern, eastern, and metro) (38) to ensure the confidentiality of source locations.
Isolate and data curation. All available isolate data, including PFGE and ribotype images, are publicly available in the PathogenTracker database (http://www.pathogentracker.net). Isolates are stored in brain heart infusion with 15% glycerol at 80°C.
|
|
|---|
Two-enzyme PFGE identified 13 PFGE types occurring at least five times; these 13 PFGE types accounted for 22.8% of the isolates characterized. A total of 235 PFGE types occurred only once and accounted for 47.5% of the isolates. Automated ribotyping identified 25 ribotypes that occurred at least five times, corresponding to 83.5% of the isolates characterized, while the 27 ribotypes that occurred only once corresponded to 5.4% of the isolates. A total of 11 PFGE types included two EcoRI ribotypes. For 47 of the 51 ribotypes that occurred more than once, PFGE was able to further discriminate isolates with the same ribotypes into two or more (up to 41) PFGE types; all ribotypes that occurred
5 times were differentiated into at least two PFGE types (Table 1). In particular, two-enzyme PFGE revealed considerable diversity among the four ribotypes corresponding to L. monocytogenes EC I through EC III. Ribotypes DUP-1038B (40 isolates) and DUP-1042B (48 isolates), both corresponding to EC I, could be differentiated into 18 and 34 PFGE types, respectively. Ribotypes DUP-1044A (EC II; 32 isolates) and DUP-1053A (EC III; 6 isolates) could be discriminated into eight and four PFGE types, respectively.
|
View this table: [in a new window] |
TABLE 1. PFGE subtype discrimination among L. monocytogenes ribotypes
|
|
View this table: [in a new window] |
TABLE 2. Distribution of L. monocytogenes lineages and PFGE types among isolates from human clinical cases, foods, ruminant farms, and urban and pristine environments
|
Spatial distributions of L. monocytogenes isolates from human cases, foods, farms, and natural and urban environments. Maps of the geographical distribution of the isolate sources corresponding to the 13 PFGE types that occurred at least five times (Fig. 1; Fig. S1 [http://www.foodscience.cornell.edu/cals/foodsci/research/labs/wiedmann/links/upload/efuSupplFigure-s11.pdf]) showed that the spatial distribution of these PFGE types included four general patterns: (i) present only in isolates from a single processing facility (PFGE types 50 and 52); (ii) widely distributed and occurring in isolates from human cases as well as other sources, such as farms and natural and urban environments (PFGE types 7, 38, 121, 122, and 240); (iii) widely distributed among isolates from foods and humans (PFGE types 2, 22, and 336); and (iv) widely distributed among isolates from farms (PFGE types 189, 300, and 315).
![]() View larger version (37K): [in a new window] |
FIG. 1. Geospatial and source distributions of selected PFGE types. Distributions of PFGE type 7 (A) and PFGE type 38 (B) are shown here; distributions of all other PFGE types that occurred 5 times are shown in Fig. S1 (http://www.foodscience.cornell.edu/cals/foodsci/research/labs/wiedmann/links/upload/efuSupplFigure-s11.pdf). Sources of isolates were plotted on New York State maps. Source abbreviations are as follows: H, human; Fd, food; Fm, farm; E, environment. Dates of isolation are given as month/year. While information on the county of origin (e.g., the location of the farm or food establishment sampled or the patient residence) was available for most isolates, geographical origins were plotted using five regions within New York State (western, central, northern, eastern, and metro) (see reference 38) to ensure the confidentiality of source locations. FSL ID, Food Safety Laboratory identification number.
|
Among the widely distributed PFGE types corresponding to isolates from human cases as well as other sources, PFGE type 7 (ribotype DUP-1038B; EC I) represented not only the second most common PFGE type (15 isolates) but also the only PFGE type occurring in isolates from all sources, including seven different ruminant farms throughout New York State (Fig. 1A). This PFGE type also occurred in isolates from all three study years (2001 through 2003) (Fig. 1A). Comparison with PFGE patterns of human listeriosis outbreak-associated strains (19) revealed that PFGE type 7 matched the PFGE patterns of human and food isolates linked to the 1985 listeriosis outbreak in Los Angeles (10, 31) and the 1983 to 1987 listeriosis outbreak in Switzerland (3, 9) (Fig. 2A). PFGE type 121 (ribotype DUP-1042B; EC I) was temporally and spatially widely distributed and corresponded to isolates from a human case and six different farms, as well as environmental samples collected in two cities in New York State from 2001 through 2003 (Fig. S1C at the URL mentioned above). PFGE type 240 (ribotype DUP-1039D) occurred among isolates from a single human case and three farms in eastern and western New York State, as well as five times in samples from a single processing facility (collected over a 3[1/2]-week period) (Fig. S1E at the URL mentioned above). PFGE type 38 (DUP-1038B; EC I) corresponded to one isolate from a human case as well as six isolates from urban environments, including five isolates obtained in Albany, NY, in 2001 and 2002 (Fig. 1B). PFGE type 122 (ribotype DUP-1044B; ECII) occurred among isolates from three human cases as well as isolates from soil samples from two different farms in New York State (Fig. S1D at the URL mentioned above).
![]() View larger version (89K): [in a new window] |
FIG. 2. PFGE patterns for selected commonly occurring PFGE types. PFGE patterns for isolates exhibiting PFGE type 7 (A) and PFGE type 22 (B) are shown. Source abbreviations are as follows: H, human; Fd, food; Fm, farm; E, environment. Dates of isolation are given as month/year.
|
PFGE types widely distributed among isolates from farms included PFGE types 189 (ribotype DUP-1039C), 300 (ribotype DUP-1030B), and 315 (ribotype DUP-1045A). PFGE type 189 corresponded to isolates from samples collected on eight different farms (located in three regions) as well as one sample from an urban environment (Fig. S1I at the URL mentioned above). PFGE type 300 corresponded to isolates from five farms located across three regions of New York State (Fig. S1J at the URL mentioned above), while PFGE type 315 corresponded to isolates from seven farms located across three regions (Fig. S1K at the URL mentioned above).
|
|
|---|
PFGE is more discriminatory for the subtyping of L. monocytogenes than ribotyping. Rapid, discriminatory, and standardized molecular subtyping methods are critical for effective food-borne disease surveillance and outbreak investigations. While automated ribotyping provides superior standardization and speed compared to many other molecular subtyping methods for bacterial isolates (58), not surprisingly and consistent with the findings of a number of previous studies (1, 5, 24, 32, 56), two-enzyme PFGE using the standard PulseNet protocol provides significantly higher subtype discrimination than EcoRI ribotyping. While automated ribotyping is thus suitable for population-based studies or subtype characterization of large numbers of isolates (1), the discriminatory power of PFGE clearly is critical for human disease outbreak investigations. Interestingly, while the previous observation (7) that ApaI PFGE may be more discriminatory than AscI PFGE was confirmed by the fact that more ApaI than AscI PFGE types among the isolates characterized here were differentiated, discriminatory powers as determined by Simpson's index of discrimination did not differ significantly between ApaI and AscI PFGE methods. Combined ApaI and AscI PFGE showed significantly higher discriminatory power than single-enzyme PFGE, supporting the importance of PFGE typing with at least two enzymes for L. monocytogenes disease outbreak surveillance and investigations.
Some L. monocytogenes PFGE types are associated with specific sources. While L. monocytogenes PFGE types present in various sources in New York State show considerable overall diversity, we have identified specific PFGE types that are associated with specific sources. Overall, a total of three specific PFGE types occurred in multiple samples collected at different times from the same processing facilities, consistent with previous observations that L. monocytogenes can persist in processing plants over considerable time periods, from months to more than a decade (4, 19, 43, 50). Interestingly, two PFGE types (50 and 52) were found only in samples from a given processing plant. While it is possible that these PFGE types may be found in other sources if more isolates are characterized, these results illustrate the potential power of large PFGE databases containing a comprehensive selection of food-associated isolates for linking listeriosis cases to potential food sources. For example, matching the PFGE patterns of a number of L. monocytogenes isolates from human listeriosis cases in Finland in 1998 and 1999 and those of an isolate obtained from butter produced in a specific plant in 1997 provided important source tracking information for this outbreak investigation (33). Similarly, matching PFGE patterns of an isolate from a hot dog produced in a specific processing plant in 1989 and those of isolates from a number of patients in a 2000 listeriosis outbreak ultimately helped link this outbreak to RTE deli products produced in the same physical facility that produced the hot dog that yielded the 1989 isolate (11, 13, 43). While PFGE databases may yield apparently food-processing plant-specific L. monocytogenes PFGE patterns, which can help detect outbreak sources, it is critical to understand that PFGE matches alone and in the absence of strong epidemiological linkages cannot be used to implicate a specific food facility as an outbreak source.
In addition to the possibly processing facility-specific PFGE types discussed above, we identified other PFGE types that, although prevalent in foods, were infrequently associated with human cases. In particular, PFGE type 336 (DUP-1062A) corresponded to seven isolates from different widely distributed RTE foods and only two human cases. Interestingly, isolates with this PFGE type represent ribotype DUP-1062A, a clonal group that was previously identified (40) as being characterized by attenuated invasiveness for human intestinal epithelial cells due to a premature stop codon in inlA, an L. monocytogenes gene critical for the invasion of human intestinal epithelial cells. This association between ribotype DUP-1062A and premature stop codons in inlA had been confirmed using 62 isolates with this ribotype (40). PFGE, as well as ribotyping and other subtyping methods (16, 23), thus may also sometimes help to define L. monocytogenes subtypes that are associated with specific sources, including some subtypes for which phenotypic and genetic data provide explanations as to the cause of source associations observed for these specific subtypes.
Some L. monocytogenes PFGE types are widely distributed, including PFGE types corresponding to select epidemic clones that appear to be stable and pandemic. In addition to PFGE types that appear to be associated with specific sources, we have found a number of PFGE types that appear to be widely distributed and can be detected among isolates from a variety of different sources. Some PFGE types (189, 300, and 315) were widespread among the isolates from the environment and farms but were not found in isolates from humans. Other PFGE types (38, 121, 240, and 336) were widespread and found in multiple sources but were rarely associated with human cases. For example, PFGE type 38, a PFGE type corresponding to isolates with ribotype DUP-1038B (which is grouped into EC I), was assigned to a single human isolate as well as to six isolates from urban environments, including five isolates obtained at two different times over 1 year in a single city, further supporting the persistence of this specific subtype in this city and consistent with initial ribotyping data (47). Similarly, PFGE type 121 (ribotype DUP-1042B; EC I) was widely distributed among isolates from farms and the environment and corresponded to only a single human isolate. Similar to our findings, results from other groups have also previously shown that identical L. monocytogenes PFGE types can sometimes be found among isolates from different foods, food animals, or environments, as well as from humans, without an apparent linkage between nonhuman sources and human cases of disease that yielded identical L. monocytogenes subtypes (6, 24, 42).
Interestingly, we also identified two PFGE types that were previously linked to human listeriosis outbreaks as being widely distributed. PFGE type 7, which is identical to the PFGE type linked to human listeriosis outbreaks in Los Angeles (1985) (10) and Switzerland (1983 to 1987) (9) and thus represents EC I (30), was found among isolates from humans, foods, farms, and farm animals, as well as isolates from environmental sources in the study reported here. We thus hypothesize that this PFGE type represents a stable pandemic clone, which appears to be able to survive successfully in different environments as well as grow in foods and cause human disease. Similarly, PFGE type 22, which is identical to the two-enzyme PFGE type of the strain linked to a multistate listeriosis outbreak in the northeastern United States in 2002 (14), was found among human and food isolates that spanned a three year period and included isolates from foods not associated with the outbreak (i.e., soft cheese), possibly indicating that this outbreak-associated PFGE type is also more widely distributed than previously assumed (14). Overall, our data indicate that some PFGE types are stable and widely distributed. These findings are important for the application of PFGE in food-borne disease surveillance and outbreak detection as they clearly support the fact that alone, PFGE subtype matches between food or environmental isolates and human clinical isolates do not necessarily imply a causal relationship. The more common a PFGE pattern linked to an outbreak is, the more critical strong epidemiological evidence is for linking an outbreak to a specific food source; sometimes the use of additional subtyping methods to differentiate subtypes within highly stable two-enzyme PFGE types may also be needed to further characterize isolates linked to a given outbreak with the necessary level of confidence.
Conclusion. While the characterization of L. monocytogenes by PFGE typing or other subtyping methods can aid in the identification of listeriosis outbreaks and their food sources (12-14, 20, 21), the data reported here show that large subtype databases representing isolates from different sources may help in the interpretation of subtype data in epidemiological investigations and may facilitate the identification of common, as well as source-specific, subtypes. Due to the ever-increasing complexity and broadening distribution patterns of the food system (43, 51), databases that include subtype patterns from sources around the globe will be particularly useful. Efforts to expand the current U.S. PulseNet database internationally (52), as well as efforts to include PFGE patterns for isolates obtained from foods (e.g., food isolates obtained by the U.S. Department of Agriculture's Food Safety Inspection Service and the Food and Drug Administration) (30, 51), are thus critical to further improve the value of PulseNet and PFGE typing. While the addition of PFGE patterns for animal isolates (e.g., through the proposed VetNet system) (18, 28) is likely to enhance the value of the PulseNet databases even further, even more comprehensive databases that include isolates from different environmental sources may be needed to fully understand food-borne pathogen and L. monocytogenes diversity and to provide data that can be useful in providing context to epidemiological investigations. The large number of PFGE types found among L. monocytogenes isolates specifically indicates that PFGE databases may need to contain information on thousands of isolates from broad geographical and source ranges.
We thank P. McGann, M. Garner, and K. Nightingale for helpful discussions and comments on manuscript drafts.
Published ahead of print on 3 January 2007. ![]()
Supplemental material for this article may be found at http://jcm.asm.org/. ![]()
|
|
|---|
dms/lmr2-toc.html.This article has been cited by other articles:
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Copyright © 2009 by the American Society for Microbiology. For an alternate route to Journals.ASM.org, visit: http://intl-journals.asm.org | More Info»