|Home | About | Journals | Submit | Contact Us | Français|
Salmonella enterica subsp. enterica is the leading cause of bacterial food-borne disease in the United States. Molecular subtyping methods are powerful tools for tracking the farm-to-fork spread of food-borne pathogens during outbreaks. In order to develop a novel multilocus sequence typing (MLST) scheme for subtyping the major serovars of S. enterica subsp. enterica, the virulence genes sseL and fimH and clustered regularly interspaced short palindromic repeat (CRISPR) loci were sequenced from 171 clinical isolates from nine Salmonella serovars, Salmonella serovars Typhimurium, Enteritidis, Newport, Heidelberg, Javiana, I 4,,12:i:−, Montevideo, Muenchen, and Saintpaul. The MLST scheme using only virulence genes was congruent with serotyping and identified epidemic clones but could not differentiate outbreaks. The addition of CRISPR sequences dramatically improved discriminatory power by differentiating individual outbreak strains/clones. Of particular note, the present MLST scheme provided better discrimination of Salmonella serovar Enteritidis strains than pulsed-field gel electrophoresis (PFGE). This method showed high epidemiologic concordance for all serovars screened except for Salmonella serovar Muenchen. In conclusion, the novel MLST scheme described in the present study accurately differentiated outbreak strains/clones of the major serovars of Salmonella, and therefore, it shows promise for subtyping this important food-borne pathogen during investigations of outbreaks.
Salmonella enterica subsp. enterica is the leading cause of bacterial food-borne disease in the United States, with approximately 1.4 million human cases each year since 1996, resulting in an estimated 17,000 hospitalizations, more than 500 deaths (9, 49), and a cost estimated as 2.6 billion dollars (U.S. Department of Agriculture Economic Research Service Salmonella food-borne illness cost calculator at http://www.ers.usda.gov/Data/FoodborneIllness/salm_Intro.asp). The nine most common human S. enterica serovars, Salmonella serovars Typhimurium, Enteritidis, Newport, Heidelberg, Javiana, I 4,,12:i:−, Montevideo, Muenchen, and Saintpaul, were responsible for more than 60% of human salmonellosis cases based on the Centers for Disease Control and Prevention's (CDC's) annual summary of 2005 (4, 5) and continue to be a major cause of food-borne illness (6, 7, 8, 9, 23). Salmonella has been isolated from a broad range of foods (CDC OutbreakNet Foodborne Outbreak Online Database at http://wwwn.cdc.gov/foodborneoutbreaks/), and widespread distribution of these foods makes tracking the transmission of Salmonella difficult during investigations of outbreaks. In order to define the routes of transmission of Salmonella within the food system, molecular subtyping methods have been employed to distinguish outbreak from non-outbreak-related strains/clones (16).
Serotyping is the most commonly used molecular subtyping method for Salmonella. Serotyping distinguishes Salmonella based on immunological classification of the H and O antigens (19) and is routinely used for surveillance of this organism. However, serotyping cannot distinguish outbreak strains/clones of the same serotype of Salmonella.
Several nucleic acid-based molecular subtyping methods have been used to subtype Salmonella, including amplified fragment length polymorphism (AFLP) (18, 32, 36, 42, 46), multiple-locus variable-number tandem-repeat analysis (MLVA) (2, 30, 31, 37), and pulsed-field gel electrophoresis (PFGE) (35). PFGE is currently considered the “gold standard” method for subtyping food-borne pathogens and is the subtyping method used by PulseNet, the molecular surveillance network in the United States and throughout the world to investigate food-borne illnesses and outbreaks (17). To enhance comparability and interpretation, a standardized PFGE protocol and an extensive quality assurance system have been established in PulseNet (17, 35). The main advantage of PFGE is its high discriminatory power (i.e., ability to separate unrelated strains) for subtyping food-borne pathogens, including most of the major serovars of Salmonella (27). However, PFGE lacks discriminatory power for highly clonal serovars of Salmonella, such as Salmonella serovar Enteritidis (17, 50), or clonal phage types like Salmonella serovar Typhimurium DT104 (17). For example, the multistate Salmonella serovar Enteritidis outbreak associated with shell eggs in 2010 was caused by the most common XbaI PFGE pattern (JEGX01.0004) for Salmonella serovar Enteritidis (7). A similar scenario was also observed recently during the 2010 outbreak associated with Italian-style salami, when the outbreak strain/clone of Salmonella serovar Montevideo had the most common PFGE pattern in the PulseNet database (8).
Compared to PFGE, multilocus sequence typing (MLST), which targets nucleotide sequence differences of several DNA loci, has the potential to be a less labor-intensive method. Moreover, DNA sequence data are discrete, unambiguous, highly informative, portable, and reproducible. Although MLST is an attractive subtyping approach, a satisfactory MLST scheme for subtyping multiple serovars of Salmonella to the strain level for investigations of outbreaks has yet to be described. MLST schemes targeting housekeeping genes have been developed; however, these schemes usually have much lower discriminatory power than PFGE (14, 24, 29, 46). In order to increase discriminatory power, virulence genes have been included in MLST schemes for subtyping Salmonella (15). Virulence genes are commonly under positive, diversifying selection (13) and therefore tend to have more-variable sequences than housekeeping genes (10, 15). MLST schemes using both housekeeping and virulence genes have been used for subtyping Salmonella to the serovar level (44) or for discriminating Salmonella serovar Typhimurium to the strain level (15). However, with Salmonella serovar Enteritidis, one of the most frequent causes of human salmonellosis, comparative genomic analysis (Salmonella single-nucleotide polymorphism [SNP] database at http://www.ncbi.nlm.nih.gov/genomes/static/Salmonella_SNPS.html) suggested that virulence genes alone are not discriminatory enough for differentiating strains from different outbreaks (Salmonella SNP database). Therefore, additional genome targets with greater sequence diversity than virulence genes are needed in order to create an effective MLST scheme for Salmonella.
One of the fastest evolving genetic elements in bacterial genomes are clustered regularly interspaced short palindromic repeats (CRISPRs) (40). CRISPRs have been identified within the genomes of many archaeal and bacterial species, including Salmonella (26, 40, 47). CRISPRs encode tandem sequences containing 21- to 47-bp direct repeats (DRs) separated by spacers of similar size (see Fig. S1 in the supplemental material). Spacers are derived from foreign nucleic acids, such as those from phage or plasmids and can protect bacteria from subsequent infection by homologous phage and plasmids (1). Many CRISPR loci are flanked by an AT-rich leader sequence and CRISPR-associated (Cas) genes (see Fig. S1 in the supplemental material) (1, 3, 22). As a bacterial immune system against foreign DNA, CRISPRs evolve rapidly in response to changing phage pools (48). Besides the addition of new spacers, deletion of spacers is also frequently observed (11, 34). Because of the high polymorphism of CRISPRs, they have been successfully used to subtype Mycobacterium tuberculosis during investigations of outbreaks (21). CRISPR sequence analysis has also been used to characterize a number of other bacteria, including Yersinia pestis (34), serotype M1 group A Streptococcus strains (25), and Campylobacter jejuni (39).
Two CRISPR loci are found in all Salmonella serovars in the CRISPR database (http://crispr.u-psud.fr/crispr/) (47). Generally, the two CRISPR loci have different numbers of repeats/spacers and different sets of spacers. There have been no reports of CRISPRs being used as markers in an MLST scheme for subtyping Salmonella. Therefore, the purpose of the present study was to investigate whether MLST based on both virulence genes and CRISPRs can accurately differentiate outbreak strains/clones of the major serovars of Salmonella.
All 171 Salmonella enterica isolates used in this study (Table (Table1)1) were from the PulseNet culture collection maintained by the Centers for Disease Control and Prevention (CDC) in Atlanta, GA. This set of isolates represents the 9 serovars most commonly associated with human disease and includes isolates involved in multiple outbreaks, with 2 or 3 isolates per outbreak. In some cases, isolates with different PFGE patterns that were obtained from the same outbreak (had poor epidemiologic concordance by PFGE) were deliberately included. All isolates were previously analyzed by serotyping, and most isolates were analyzed by PFGE by the CDC. Bacterial isolates were stored at −80°C in 20% glycerol. When needed, isolates were grown overnight in tryptic soy broth (TSB) (Difco Laboratories, Becton Dickinson, Sparks, MD) at 37°C. For all isolates, DNA was extracted using the UltraClean microbial DNA extraction kit (Mo Bio Laboratories, Solana Beach, CA) and stored at −20°C before use.
In silico analysis of 9 publically available whole-genome sequences of S. enterica (serovar Agona SL483, GenBank accession no. CP001138; serovar Choleraesuis SC-B67, GenBank accession no. AE017220; serovar Dublin CT_02021853, GenBank accession no. CP001144; serovar Enteritidis P125109, GenBank accession no. AM933172; serovar Gallinarum strain 287191, GenBank accession no. AM933173; serovar Heidelberg SL476, GenBank accession no. CP001120; serovar Newport SL254, GenBank accession no. CP001113; serovar Schwarzengrun CVM19633, GenBank accession no. CP001127; and serovar Typhimurium LT2, GenBank accession no. AE006468) was used to identify 14 virulence genes (hilA, fimH1, fimH2, pipB, sopE, sseF, sseL, sseJ, siiA, sifB, stdA, fimA, bcfC, and phoQ) (Tables (Tables22 and and3;3; see Table S1 in the supplemental material) that were present in all genomes but displayed differences in their DNA sequences. Primers for amplifying these genes were designed using Primer 3.0 (http://frodo.wi.mit.edu/primer3/) and were based on the published Salmonella serovar Typhimurium LT2 (GenBank accession no. AE006468) genome (Table (Table3;3; see Table S1 in the supplemental material). Primers for amplifying CRISPR1 were designed based upon consensus alignments of the published Salmonella serovar Typhimurium LT2 (GenBank accession no. AE006468) and serovar Newport strain SL254 genomes (GenBank accession no. CP001113), and the Salmonella serovar Javiana strain GA_MM04042433 (GenBank accession no. ABEH00000000) whole-genome shotgun sequence (Table (Table3).3). Primers for amplifying CRISPR2 were designed based on the Salmonella serovar Typhimurium LT2 genome. All primers annealed to conserved regions located 5′ or 3′ of the CRISPR loci. PCR amplifications were performed using a Taq PCR master mix kit (Qiagen Inc., Valencia, CA) and a Mastercycler PCR thermocycler (Eppendorf Scientific, Hamburg, Germany). A 25-μl PCR system contained 12.5 μl of Taq PCR 2× master mix, 9.5 μl of PCR-grade water, 1.0 μl of DNA template, 1.0 μl of forward primer (final concentration, 0.4 μM), and 1.0 μl of reverse primer (final concentration, 0.4 μM). A single PCR cycling condition was used for separately amplifying all four markers. PCR was performed as follows: initial denaturation step of 10 min at 94°C; 28 cycles, with 1 cycle consisting of 1 min at 94°C, 1 min at 55°C, and 1 min at 72°C; final extension step of 10 min at 72°C.
After PCR, products for sequencing were treated with 1/20 volume of shrimp alkaline phosphatase (1 U/μl) (USB Corp., Cleveland, OH) and 1/20 volume of exonuclease I (10 U/μl) (USB Corp). The mixture was then incubated at 37°C for 45 min to degrade remaining primers and unincorporated deoxynucleoside triphosphates (dNTPs). After that, the mixture was incubated at 80°C for 15 min to inactivate the added enzymes. PCR products were sent to the Genomics Core Facility at The Pennsylvania State University for sequencing using the ABI data 3730XL DNA analyzer. In order to obtain complete DNA sequences of fimH and sseL, two more primers targeting the internal regions of these two genes were used together with the forward and reverse primers (Table (Table3).3). Both DNA strands of the amplicons were sequenced.
Virulence gene sequences were aligned, and single-nucleotide polymorphisms (SNPs) were identified using MEGA 4.0 (43). For CRISPR1 and CRISPR2, analyses of the spacer arrangements were performed using CRISPRcompar (20), and spacers were visualized by the method of Deveau et al. (11). Different allelic types (ATs) (sequences with at least one-nucleotide difference or one-spacer difference in the case of CRISPRs) were assigned arbitrary numbers. The combination of 4 alleles (fimH, sseL, CRISPR1, and CRISPR2) determined its allelic profile, and each unique allelic profile was designated a unique sequence type (ST). The epidemiological relationships of the strains were kept from the investigators until the data had been analyzed and sequence types assigned.
Epidemiologic concordance (E) was calculated using the equation developed by the European Study Group on Epidemiologic Markers (41).
Cluster analyses were performed based on allelic profile data, and results were visualized using the tree drawing tool on PubMLST (www.pubmlst.org). CRISPR1 and CRISPR2 might be genetically linked due to their proximity in the genome (47); therefore, they were combined into one allele to reduce their weight in the cluster analysis (Fig. (Fig.1c1c).
The standard deviations of the average numbers of spacers in CRISPR1 and CRISPR2 were calculated using Microsoft Excel.
DNA sequences of the four genetic MLST markers were deposited in GenBank under accession numbers HQ329797 to HQ329931.
We began this study by sequencing 14 virulence genes (fimH, sseL, hilA, fimH2, pipB, sopE, sseF, sseJ, siiA, sifB, stdA, fimA, bcfC, and phoQ) from 20 Salmonella serovar Typhimurium, 15 Salmonella serovar Newport, and 15 Salmonella serovar Enteritidis isolates. Two virulence genes, fimH and sseL, were found to provide discrimination equal to the combined discrimination of all 14 virulence genes (data not shown); therefore, the other 12 virulence genes were excluded from the rest of the study. Virulence genes fimH and sseL were sequenced from the remaining isolates, and the total number of allelic types was 17 for fimH and 16 for sseL (Table (Table4).4). Only epidemiologically unrelated strains were included in the calculation of polymorphic sites. The total number (percentage of polymorphic sites) for fimH was 48 (4.76%), and for sseL, it was 69 (7.23%) (Table (Table5).5). Within each serovar, the percentage of polymorphic sites in fimH ranged from 0% to 1.79%; for sseL, the percentage of polymorphic sites ranged from 0% to 3.88%. For both fimH and sseL, less polymorphism was observed for Salmonella serovars Typhimurium, Enteritidis, Heidelberg, Javiana, and I 4,,12:i:− than for Salmonella serovars Newport, Montevideo, Muenchen, and Saintpaul (Table (Table5).5). Sequences of sseL were especially conserved in Salmonella serovars Typhimurium, Heidelberg, Javiana, and I 4,,12:i:−, with no SNPs observed within each serovar. For all serovars, a total of 39 polymorphic sites in sseL were nonsynonymous, and 13 polymorphic sites in fimH were nonsynonymous (Table (Table55).
Since the discrimination provided by virulence genes was limited (separation to outbreak level was not achieved), the addition of CRISPR1 and CRISPR2 to the MLST scheme was investigated. The total numbers of unique spacers in CRISPR1 and CRISPR2 for all 171 isolates analyzed were 166 and 182, respectively (Table (Table6;6; see Fig. S2 in the supplemental material). Repeat sequences of the two CRISPRs were generally conserved as shown by the typical repeat in Table S2 in the supplemental material, however, SNPs were sometimes observed and we define these as “repeat variants” (see Table S2 in the supplemental material). The number of spacers in CRISPR1 ranged from 3 to 24, while the number of spacers in CRISPR2 ranged from 2 to 25 (Table (Table6;6; see Fig. S2 in the supplemental material). CRISPR2 had more spacers than CRISPR1 for all serovars except serovar Muenchen (Table (Table66 and Fig. S2).
The number of allelic types for CRISPR1 (44 allelic types) and CRISPR2 (51 allelic types) were significantly greater than those for virulence genes (Table (Table4).4). In total, there were 61 sequence types based on both virulence genes and CRISPRs for all 158 isolates that were epidemiologically unrelated (Table (Table4).4). An equal number of allelic types was observed in both CRISPR1 and CRISPR2 for Salmonella serovars Javiana and Montevideo (Table (Table4).4). However, for Salmonella serovars Typhimurium, Enteritidis, Newport, Heidelberg, and Saintpaul, CRISPR2 yielded more allelic types than CRISPR1. In contrast, for Salmonella serovar Muenchen, CRISPR1 yielded more allelic types than CRISPR2 (Table (Table44).
Cluster diagrams based on allelic profiles were constructed using only the two virulence genes (Fig. (Fig.1a),1a), only CRISPR1 and CRISPR2 (Fig. (Fig.1b),1b), and using virulence genes combined with CRISPR (Fig. (Fig.1c).1c). Virulence genes alone were effective at separating isolates of different serovars, while the addition of CRISPR1 and CRISPR2 provided additional discrimination between isolates within the same serovar (compare Fig. Fig.1a1a to Fig. Fig.1c).1c). CRISPR sequencing alone provided the same level of discrimination as the combination of CRISPRs and virulence genes for all serovars except Salmonella serovar Enteriditis and Salmonella serovar Heidelberg (compare Fig. Fig.1b1b to Fig. Fig.1c).1c). MLST results showed high congruence with serotypes of Salmonella, as isolates of the same serovars typically occupied the same branch of the cluster diagram (Fig. (Fig.1c).1c). The three exceptions to this were strains SST4, McnST12, and MvoST3, which occupied unique branches. MLST also did not separate all isolates of the related Salmonella serovars Typhimurium and I 4,,12:i:−.
Inclusion of CRISPR in the present MLST scheme added to the discrimination provided by PFGE for outbreak isolates of Salmonella serovar Enteritidis (Tables (Tables11 and and4).4). Most isolates of Salmonella serovar Enteritidis (25 out of 34) had either the XbaI and BlnI PFGE profile JEGX01.0005 and JEGA26.0004 or JEGX01.0004 and JEGA26.0002 (Table (Table1).1). Isolates SE1, SE2, SE23, SE18, SE17, SE20, SE32, and SE33 (CDC code for isolates explained in Table Table1,1, footnote a) had the same PFGE profile (JEGX01.0005 and JEGA26.0004) but had two MLST sequence types (E ST1 and E ST 9; MLST sequence types explained in Table Table1,1, footnote c) (Table (Table1).1). Also, the PFGE profile (JEGX01.0004 and JEGA26.0002), which included isolates SE6, SE7, SE8, SE9, SE15, SE16, SE19, SE30, SE12, SE13, SE14, SE26, SE31, SE28, SE29, SE24, and SE34, were further separated into five sequence types (E ST3, E ST4, E ST6, E ST7, and E ST8) by MLST (Table (Table1).1). We did not calculate the discriminatory power (27) of PFGE and MLST, because the isolates used were not randomly selected but were biased toward outbreak strains that showed poor epidemiologic concordance by PFGE.
For some Salmonella serovars (Salmonella serovars Newport, Typhimurium, I 4,,12:i:−, Montevideo, and Saintpaul), PFGE provided greater separation than MLST among strains associated with different outbreaks. For example, PFGE separated Salmonella serovar I 4,,12:i:− isolates (ST1, ST2, and ST3) of an outbreak linked to consumption of turkey pot pies (cluster 0706PAJPX-1c) from isolates (ST14 and ST15) of cluster 0607INjpx-1c, while these isolates could not be distinguished by MLST (Table (Table1).1). PFGE also distinguished Salmonella serovar Typhimurium isolates from an outbreak linked to raw milk consumption (designated “outbreak a” in Table Table1)1) and outbreak cluster 0309ORJPX-1c (Table (Table1).1). Also, in contrast to MLST, PFGE was able to discriminate the outbreak linked to raw chicken (cluster 0807AZJIX-1c) from the outbreak linked to salami/pepper (cluster 0908ORJIX-1) of Salmonella serovar Montevideo (Table (Table1).1). Multiple PFGE patterns were seen among Salmonella serovar Newport isolates from MLST sequence types N ST4 and N ST5. For Salmonella serovar Saintpaul, both methods allowed accurate separation and identification of all outbreaks due to this serovar, although PFGE provided better separation of sporadic isolates SS18, SS19, and SS15 from outbreak isolates (Table (Table11).
For Salmonella serovar Heidelberg, the most accurate outbreak identification was achieved by combining MLST and PFGE. MLST provided separation for the isolates from an outbreak on a cruise ship (cluster 0607NYJF6-1c) and the isolates from an outbreak in a religious camp (cluster 0607PAJF6-1c), which could not be distinguished by PFGE (Table (Table1).1). Similarly, MLST distinguished between isolates from an outbreak linked to hummus (cluster JF6X01.0032) and isolates from an outbreak (cluster 0702TNJF6-1c), which had the same pulsotypes in Table Table1.1. However, PFGE separated the isolates from the outbreak on a cruise ship from the isolates from the outbreak linked to hummus, which were indistinguishable by MLST (Table (Table11).
Values of epidemiologic concordance of MLST and PFGE for each serovar were calculated (Table (Table7),7), except for the Salmonella serovar Javiana which did not contain any isolates with a defined PulseNet cluster code. Epidemiologic concordance values were calculated based on isolates from well-defined outbreaks (isolates with the same cluster codes), so sporadic isolates and isolates without cluster code designations were excluded. It is important to note that many outbreak isolates included in this study were deliberately chosen because they displayed poor epidemiologic concordance by PFGE. For instance, isolates ST6, ST7, and ST8 were all associated with a 2008 outbreak linked to peanut butter, but each of these isolates had a distinct PFGE pattern (Table (Table1).1). MLST showed high epidemiologic concordance (epidemiologic concordance between 0.88 and 1.0) for subtyping all serovars included in this study, except for Salmonella serovar Muenchen (epidemiologic concordance of 0.39) (Table (Table7).7). On the basis of the limited number of strains analyzed in the present study, MLST showed higher epidemiologic concordance than PFGE for Salmonella serovars Enteritidis, Typhimurium, Newport, and Montevideo, equal epidemiologic concordance for Salmonella serovar Saintpaul, but lower epidemiologic concordance for Salmonella serovars Heidelberg and Muenchen (Table (Table77).
There are several important criteria to follow when selecting genetic markers to use in an MLST scheme. First, the selected genetic markers should exhibit adequate sequence variations to provide separation of unrelated strains (33). Second, genetic markers that provide epidemiologically meaningful information should be selected so that the MLST scheme can exhibit high epidemiologic concordance. Last but not least, genetic markers should be present in the genome within all strains of the species of interest. Previous studies demonstrated that MLST schemes based on Salmonella enterica housekeeping genes showed poor discriminatory power compared to PFGE (14, 24, 46). Inclusion of virulence genes into one published MLST scheme for subtyping Salmonella serovar Typhimurium increased discriminatory power to 0.98, which was comparable to that of PFGE (0.96) (15). Similarly, virulence genes provided epidemiologically meaningful separation and clustering of strains of Listeria monocytogenes (10). Besides virulence genes, CRISPRs were selected as markers in the current MLST scheme because they were found to be one of the fastest evolving genetic elements in bacterial genomes (40).
In the present study, cluster analyses based on the two virulence genes and two CRISPRs accurately grouped isolates according to their specific serovars, except for Salmonella serovar Typhimurium and Salmonella serovar I 4,,12:i:−, which were clustered together. As Salmonella serovar I 4,,12:i:− is a monophasic variant of Salmonella serovar Typhimurium (12), this was not unexpected. Virulence genes were previously found to provide accurate identification of different serovars of Salmonella in other studies as well (38, 44, 45).
Addition of CRISPRs significantly increased discriminatory power (Fig. (Fig.1)1) compared to previously published MLST schemes, and the identification of individual outbreak strains/clones was achieved. For example, one MLST scheme based on three housekeeping genes (manB, pduF, and glnA) and one virulence gene (spaM) identified one sequence type among 85 Salmonella serovar Typhimurium isolates and discriminatory power for the MLST scheme was 0 (14). Another MLST scheme targeted seven housekeeping genes, aroC, dnaN, hemD, hisD, purE, sucA, and thrA, and identified 12 sequence types among a total of 81 Salmonella serovar Newport isolates, which also resulted in poor discriminatory power (0.61) (24). One MLST study based on virulence genes (hilA, pefB, and fimH), 16S rRNA gene, and housekeeping genes showed high discriminatory power (0.98) for subtyping Salmonella serovar Typhimurium (15); however, its capacity to discriminate strains from more-clonal serovars, such as Salmonella serovar Enteritidis, was not tested. In conclusion, the MLST scheme described in the present study has superior discriminatory power compared to previously published MLST schemes for subtyping the major serovars of Salmonella, especially for the highly clonal Salmonella serovar Enteritidis.
As mentioned previously, the isolates selected for this study were biased toward those that had poor epidemiologic concordance by PFGE; therefore, future studies comparing MLST and PFGE need to be performed using a nonbiased strain collection. Generally speaking, the current MLST scheme showed high epidemiologic concordance for subtyping the major serovars of Salmonella, except for Salmonella serovar Muenchen (E = 0.39) (Table (Table7).7). All Salmonella serovar Muenchen isolates had different sequence types, except isolates SMcn13 and SMcn15 from the outbreak linked to orange juice (Table (Table1).1). Interestingly, the allelic types of fimH and sseL were the same for all the Salmonella serovar Muenchen isolates, except for isolate SMcn12 (Fig. (Fig.1a),1a), which means that CRISPR1 and CRISPR2 provided almost all of the discriminatory power in the case of Salmonella serovar Muenchen isolates (Fig. 1b and c). One possible explanation for this unexpected diversity may be that CRISPRs are evolving too fast for investigations of Salmonella serovar Muenchen outbreaks, either because the specific niche where Salmonella serovar Muenchen resides harbors a large number of different phages and/or because phage pools of Salmonella serovar Muenchen are very dynamic. Dramatic differences have been observed in the rate of spacer acquisition between different eubacteria. In Streptococcus thermophilus, CRISPRs are very active, and new spacer acquisition appears to be the primary mechanism by which this species evolves phage resistance (11); however, the rate of new spacer acquisition in other bacteria, such as Escherichia coli, appears to be much slower (47). Alternatively, CRISPRs may have detected epidemiologically relevant differences in Salmonella serovar Muenchen isolates that were not detected by PFGE or discovered in the investigations of outbreaks.
The current MLST scheme also separated Salmonella serovar Enteritidis isolates with common PFGE patterns (Table (Table1).1). The predominant XbaI PFGE patterns for Salmonella serovar Enteritidis in the PulseNet database are JEGX01.0004 and JEGX01.0005 making up 45% and 13% of the database, respectively (CDC, unpublished data). This lack of PFGE pattern diversity makes it difficult to differentiate potential outbreak-related isolates from sporadic isolates (17). The discriminatory power of PFGE has been increased by the combination of multiple restriction enzymes (50). However, whether the increased discrimination provided by additional restriction enzymes caused potential loss of epidemiologic concordance was not addressed in that study. The present MLST scheme allowed separation of the two predominant PFGE patterns of Salmonella serovar Enteritidis isolates (Table (Table1)1) and resulted in high epidemiologic concordance (Table (Table7).7). CRISPRs provided most of the discrimination among outbreak strains/clones (Fig. 1b and c). CRISPRs in Salmonella serovar Enteritidis are evolving due to plasmids and/or phages present in the environment (48). Fortunately, the rate of spacer insertion and deletion in CRISPRs is slow enough such that they do not appear to change during the time course of an outbreak (Table (Table1).1). CRISPRs may also reflect the specific phage and plasmid pool in the environment and hence contain ecologically and geographically meaningful information for bacteria (28, 48) that may be useful for tracking strains of Salmonella to their specific source (farm or food processing plant). In conclusion, the current MLST scheme effectively subtyped the two most common PFGE patterns of Salmonella serovar Enteritidis and thus could enhance cluster definition and outbreak investigation capabilities of public surveillance laboratories.
It has been previously suggested that CRISPRs are poor epidemiological markers in enterobacteria due to the low rate of spacer acquisition (47). However, that study analyzed only 16 complete Salmonella genomes for CRISPRs, and only four of them were from the same serovar as strains analyzed in the current study. Additionally, the authors included in their study only one isolate from Salmonella serovars Typhimurium, Enteritidis, Newport, and Heidelberg, so the true value of CRISPRs for epidemiologic investigations could not be adequately evaluated given their sampling limitations. Our study analyzed 26, 34, 15, and 20 isolates from Salmonella serovars Typhimurium, Enteritidis, Newport, and Heidelberg, respectively, and demonstrated that CRISPR sequences may provide the discriminatory power and epidemiologic concordance needed for epidemiologic investigations. We are currently testing this hypothesis further using larger numbers of isolates obtained from current and past Salmonella outbreaks. The previously observed discrepancy between CRISPR sequences and strain phylogeny (47) suggests that the MLST method reported here would not be useful for determining the long-term phylogeny of Salmonella isolates.
This MLST scheme has several other advantages that make it a potential subtyping method for routine surveillance of Salmonella. First, the primers in this MLST scheme were designed to have the same annealing temperature for all four markers so that it can be conveniently performed in large-scale epidemiologic investigations. Second, the number of the markers targeted was minimized to two virulence genes and two CRISPRs so that time and expense can be saved during routine typing of Salmonella strains (33). Third, all four markers, fimH, sseL, CRISPR1, and CRISPR2, are present in the major serovars of Salmonella and also in all published genomes of Salmonella serovars, so the current MSLT scheme is widely applicable. Although this MLST scheme shows great promise, future research is needed to further validate it for epidemiologic purposes and to compare it to more commonly used molecular subtyping tools for Salmonella, including PFGE and MLVA. One advantage our method has over MLVA though is that it appears to be universally applicable to the most clinically relevant Salmonella serovars, where MLVA protocols for only a limited number of serovars have been described so far.
In conclusion, this study suggests that an MLST protocol that includes CRISPR1 and CRISPR2 may be a useful subtyping method for tracking the farm-to-fork spread of the most prevalent serovars of Salmonella during investigations of outbreaks.
We thank Bindhu Verghese for technical guidance throughout the study, especially for the idea of combining CRISPRs into one allele for the cluster analysis. We also acknowledge the Penn State Genomics Core Facility-University Park, PA, for DNA sequencing and Eija Hyytia-Trees at the CDC for assistance in selecting the strain collection.
This study was supported by a U.S. Department of Agriculture Special Milk Safety grant to the Pennsylvania State University (contract 2009-34163-20132).
Published ahead of print on 28 January 2011.
†Supplemental material for this article may be found at http://aem.asm.org/.