|Home | About | Journals | Submit | Contact Us | Français|
In 2000 to 2001, 2003 to 2004, and 2005 to 2006, three outbreaks of Salmonella enterica serovar Enteritidis were linked with the consumption of raw almonds. The S. Enteritidis strains from these outbreaks had rare phage types (PT), PT30 and PT9c. Clinical and environmental S. Enteritidis strains were subjected to pulsed-field gel electrophoresis (PFGE), multilocus variable-number tandem repeat analysis (MLVA), and DNA microarray-based comparative genomic indexing (CGI) to evaluate their genetic relatedness. All three methods differentiated these S. Enteritidis strains in a manner that correlated with PT. The CGI analysis confirmed that the majority of the differences between the S. Enteritidis PT9c and PT30 strains corresponded to bacteriophage-related genes present in the sequenced genomes of S. Enteritidis PT4 and S. enterica serovar Typhimurium LT2. However, PFGE, MLVA, and CGI failed to discriminate between S. Enteritidis PT30 strains related to outbreaks from unrelated clinical strains or between strains separated by up to 5 years. However, metabolic fingerprinting demonstrated that S. Enteritidis PT4, PT8, PT13a, and clinical PT30 strains metabolized l-aspartic acid, l-glutamic acid, l-proline, l-alanine, and d-alanine amino acids more efficiently than S. Enteritidis PT30 strains isolated from orchards. These data indicate that S. Enteritidis PT9c and 30 strains are highly related genetically and that PT30 orchard strains differ from clinical PT30 strains metabolically, possibly due to fitness adaptations.
Salmonella enterica is one of the major causes of bacterial food-borne illness worldwide. Many serovars of S. enterica serovar Enteritidis emerged as serious problems in the human food supply during the 1980s, and these cases were associated mostly with undercooked eggs and poultry (26). The phage typing of S. Enteritidis strains associated with egg-associated outbreaks had indicated that phage types 8 (PT8) and PT13a were the most common PTs in the United States (12), and PT4 was the most common in Europe (22). Through education and quality improvements, the incidence of S. Enteritidis due to egg products has decreased in the United States (18). However, several recent outbreaks have identified new sources for S. Enteritidis, specifically mung bean sprouts, tomatoes, and raw whole almonds (3, 13, 31).
At the time of the 2001 outbreak, almonds and other low-moisture foods were considered an unlikely source of food-borne illness. Almonds are California's major tree nut crop and have ranked first in California agricultural exports for many years, accounting for 60% of world production in 2000 (14) and 80% in 2008 (http://www.almondboard.com/AboutTheAlmondBoard/Documents/2008-Almond-Board-Almanac.pdf). However, no outbreaks associated with almonds had been reported before 2001. In the spring of 2001, Canadian health officials identified a link between illnesses caused by S. Enteritidis and the consumption of raw almonds (6). Outbreak-related cases were identified from November 2001 to July 2001 in several provinces across Canada and in several regions in the United States (13). During the traceback investigation, almond retailers, processors, and growers were identified, and S. Enteritidis PT30 was cultured from almond samples, a huller/sheller facility, and environmental samples from the orchards (30). The ability to identify the contaminated food source for this outbreak was aided significantly by the previously rare occurrence of S. Enteritidis PT30. S. Enteritidis PT30 continued to be isolated from one of the outbreak-associated orchards during a 5-year period, suggesting that this organism was highly fit for persistence in this environment (30).
In 2004, another rare S. Enteritidis PT (PT9c) was linked to a second outbreak associated with raw almonds. Similarly to the first outbreak, both phage typing and pulsed-field gel electrophoresis (PFGE) aided the identification of related cases caused by S. Enteritidis PT9c that occurred over a large geographical region of the United States and Canada (3). A third S. Enteritidis PT30 outbreak associated with raw almonds was reported in Sweden in 2005 to 2006 (15).
We have characterized, by molecular methods, S. Enteritidis strains recovered from clinical, almond, and orchard samples related to these three outbreaks to determine whether they were related genotypically. Additional S. Enteritidis strains representing some common phage types also were examined for comparison. Strains were genotyped by PFGE profiling, multilocus variable-number tandem repeat analysis (MLVA), and comparative genomic indexing (CGI) with a S. enterica serovar Typhimurium LT2/Enteritidis PT4 microarray to determine relatedness and whether an association with the source could be determined.
The S. Enteritidis strains used in this study are listed in Table Table1.1. Strain S. Typhimurium LT2 ATCC 15277 was utilized in DNA microarray analysis. Additional strains of S. Enteritidis PT30, isolated predominantly from the orchard on different sampling trips during a 5-year period (30) but examined only by PFGE, are not listed. Salmonella strains were grown routinely at 37°C on LB agar medium or in LB broth (Becton Dickinson Co., Sparks, MD). Genomic DNA was prepared by growing strains of S. Typhimurium and S. Enteritidis overnight at 37°C and extracting DNA from cells using the Wizard DNA preparation kit (Promega, Madison, WI). These DNA preparations were used for MLVA and CGI analysis.
Salmonella strains were analyzed by PFGE using XbaI and BlnI restriction enzymes essentially as described previously (18). DNA fragments after digestion with enzymes were separated on gels with the CHEF Mapper II system (Bio-Rad, Hercules, CA). The method used for PFGE was not a PulseNet method and was optimized to show differences in the lowest bands. The results are not directly comparable to those within the PulseNet database.
Selected S. Enteritidis strains were analyzed by MLVA using primers described previously (4). Primers for amplifying PCR products for each of eight variable-number tandem repeat (VNTR) loci, designated SE1, SE2, SE3, SE5, SE6, SE7, SE9, and SE10, and their sequencing have been described previously (4). These loci were amplified for each strain in individual PCRs. Each PCR consisted of 1× MasterAmp Taq PCR buffer, 1× MasterAmp Taq enhancer, 2.5 mM MgCl2, 200 μM each deoxynucleoside triphosphate (dNTP), forward and reverse primers (Eurofins MWG Operon, Huntsville, AL) at 0.2 μM each, 0.2 U of MasterAmp Taq DNA polymerase (Epicentre, Madison, WI) or Taq DNA polymerase (NEB, Beverly, MA), and approximately 50 ng of genomic DNA (final reaction volume, 25 μl). The PCR conditions were 30 cycles of 94°C for 30 s, 55°C for 30 s, and 72°C for 60 s, followed by a final elongation of 72°C for 5 min. Thermal cycling was performed with a Tetrad thermal cycler (Bio-Rad, Hercules, CA). All PCR products were analyzed by agarose gel electrophoresis. Positive samples were identified based on the presence of bands of the predicted sizes. The PCR products were purified on a Qiagen 8000 robot using a QIAquick 96-well Biorobot kit (Qiagen, Valencia, CA).
Labeled DNA for sequencing was produced on an MJ Research Tetrad thermocycler using the ABI PRISM BigDye terminator cycle sequencing kit (version 3.0; Applied Biosystems, Foster City, CA) and standard protocols as recommended by the manufacturer. All labeled products were purified on DyeEx spin columns (Qiagen, Valencia, CA). DNA sequencing was performed on an ABI PRISM 3100 Genetic Analyzer (Applied Biosystems) using the POP-6 polymer and ABI PRISM genetic analyzer data collection and ABI PRISM genetic analyzer sequencing analysis software. The DNA primers used for PCR or sequencing were designed using Primer Premier 5.0 (Premier Biosoft International). Sequencing reads were trimmed and assembled using Lasergene Seqman II (version 6.0; DNAstar, Madison, WI). The sequences were verified by BLAST at the NCBI website (http://blast.ncbi.nlm.nih.gov/Blast.cgi). The number of tandem repeats for each locus was measured using Tandem Repeats Finder software (accessible at http://tandem.bu.edu/trf/trf.html).
DNA fragments of individual S. Typhimurium LT2 protein-coding sequences (CDS) (19) were amplified using CDS-specific primers from Sigma-Genosys (St. Louis, MO) as described by Clements et al. (8). For the CDS of S. Enteritidis PT4 strain P125109 (29), primers were designed with ArrayDesigner 3.0 (Premier Biosoft, Palo Alto, CA) and synthesized by Illumina (San Diego, CA). Each PCR (total reaction volume, 100 μl) consisted of 1× MasterAmp Taq PCR buffer, 1× MasterAmp Taq enhancer, 2.5 mM MgCl2, 200 μM each deoxynucleoside triphosphate, forward and reverse primers at 0.2 μM each, 0.5 U of MasterAmp Taq DNA polymerase (Epicentre, Madison, WI), and approximately 50 ng of S. enterica genomic DNA (either S. Typhimurium LT2 strain ATCC 15277 or S. Enteritidis PT 4 strain RM3972). Thermal cycling was performed using a Tetrad thermal cycler (Bio-Rad) with the following amplification parameters: 30 cycles of 25 s at 94°C, 25 s at 52°C, and 2 min at 72°C, with a final extension at 72°C for 5 min. PCR products were analyzed by gel electrophoresis in a 1% (wt/vol) agarose gel (containing 0.5 μg/ml of ethidium bromide) in 1× Tris-acetate-EDTA buffer. DNA bands were examined under UV illumination. We amplified a total of 4,422 and 192 PCR products from S. Typhimurium LT2 and S. Enteritidis PT4, respectively. These PCR products were purified on a Qiagen 8000 robot using a QIAquick 96-well Biorobot kit (Qiagen, Valencia, CA), dried, and resuspended to an average concentration of 0.1 to 0.2 μg/μl in 10 μl of 50% dimethyl sulfoxide (DMSO) containing 0.3× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate). All of the PCR probes then were spotted onto Ultra-GAPS glass slides (Corning Inc., Corning, NY) using an OmniGrid Accent (Digilab, Holliston, MA), producing a final array that contained a total of 4,664 features including duplicates. The DNA was UV cross-linked to the microarray slides with a Stratalinker (Agilent Technologies, Redwood City, CA) at 300 mJ.
For each microarray hybridization reaction, 1 μg of genomic DNA from each of the reference strains (S. Typhimurium LT2 strain ATCC 15277 and S. Enteritidis PT4 strain RM3972) and 2 μg of genomic DNA from a test strain were labeled fluorescently with indodicarbocyanine (Cy5) and indocarbocyanine (Cy3), respectively. Briefly, the genomic DNA was mixed with 5 μl 10× NEBlot labeling buffer containing random sequence octamer oligonucleotides (New England Biolabs, Ipswich, MA) and water to a final volume of 41 μl. This mixture was heated to 95°C for 5 min and then cooled for 5 min at 4°C. After this time, the remainder of the labeling reaction mixture components were added: 5 μl of 10× dNTP labeling mix (1.2 mM each dATP, dGTP, and dTTP and 0.5 mM dCTP in 10 mM Tris, pH 8.0, 1 mM EDTA), 3 μl of Cy3 dCTP or Cy5 dCTP (Amersham Biosciences, Piscataway, NJ), and 1 μl of Klenow fragment. The labeling reaction mixtures were incubated overnight at 37°C. Labeled DNA was purified from unincorporated label using QIAquick PCR cleanup kits (Qiagen) and dried by vacuum.
Labeled reference and test DNAs were combined in 45 μl Pronto! cDNA hybridization solution (Corning) and heated to 95°C for 10 min. Fifteen microliters of the hybridization mixture then was put onto a microarray slide and sealed with a coverslip. The microarray slide was placed in a hybridization chamber (Corning) and incubated at 42°C for 18 h. This method, known as differential labeling, allows the hybridization of fluorescently labeled reference (Cy5) and test (Cy3) DNA to be measured for each probe on the microarray. As the genetic composition of the control strain is known from the genome sequence, it serves as a control fluorescence signal for each probe and is used for comparison with the test DNA signal during the statistical analysis (see below). Following hybridization, the slides were washed twice in 2× SSC, 0.1% sodium dodecyl sulfate at 42°C for 10 min, followed by twice in 1× SSC at room temperature for 10 min, and finally twice in 0.1× SSC at room temperature for 10 min. The microarray slides were dried by centrifugation at 300 × g for 10 min before scanning. At least two hybridization reactions were performed for each test strain.
Microarrays were scanned and analyzed by methods described previously by Anjum et al. (2) with modifications. Briefly, DNA microarrays were scanned using an Axon GenePix 4000B microarray laser scanner (Axon Instruments Inc., Union City, CA). Features and the local background intensities were detected and quantified with GenePix 4.0 software (Axon Instruments, Inc.). Poor features containing abnormalities or within regions of high fluorescent background were excluded from further analysis. The data were filtered so that spots with a reference signal lower than the background plus two standard deviations of the background were discarded. Signal intensities were corrected by subtracting the local background, and then the Cy5/Cy3 ratios were calculated. To compensate for unequal dye incorporation and slide-to-slide variation, data normalization was performed using an Excel add-in developed by Sacha Lucchini and described previously (2). The CGI analysis of ATCC 15277 and RM3972 compared to that of the reference (ATCC 15277 and RM3972 combined) defined ratio cutoffs for the presence, divergence, and absence status of the genes. Based on these data, we defined the status of a gene as present when the Cy3/Cy5 (test/reference) intensity ratio was >0.6, as divergent or unknown when the Cy3/Cy5 intensity ratio was between 0.6 and 0.3, and absent when the Cy3/Cy5 intensity ratio was <0.3. These values are similar to the intensity ratios utilized in other microarray studies (2, 10, 28). The ATCC 15277 and RM3972 strain-specific spots hybridized to only half of the reference DNA (the Cy5-labeled mixture of ATCC 15277 and RM3972 DNA), increasing the Cy3/Cy5 ratio by 2-fold. Therefore, the ratios for these spots were divided by two before determining the status of the gene. The status for all genes was converted into trinary scores (present, 2; divergent or unknown, 1; absent, 0). The trinary gene scores for each replicate for all strains were analyzed further with GeneSpring microarray analysis software version 7.3 (Agilent) and subjected to average-linkage hierarchical clustering with the standard correlation coefficient and bootstrapping.
Selected S. Enteritidis strains were assayed for respiratory responses to substrates in GN2 microplates (Biolog Inc., Hayward, CA) according to the protocol provided by the vendor. The GN2 microplates were incubated for 24 h at 37°C. Color changes were measured at 600 nm using a Biolog microplate reader and were corrected for the no-substrate control. Three independent replicate assays were performed for each strain. The cutoff point between negative results and positive results was an optical density at 600 nm of 0.2 as described previously (16). Substrates giving values close to the cutoff point were considered positive or negative if there was at least a 4-fold difference in replicate readings between S. Enteritidis PT30 strains from different sources. The Student's t test was used to determine if data sets from orchard isolates and clinical S. Enteritidis PT30 isolates were significantly different as defined by P ≤ 0.01.
S. Enteritidis strains from almond-associated outbreaks in 2000 to 2001 (OB1), 2003 to 2004 (OB2), and 2005 to 2006 (OB3) were typed by PFGE using the restriction enzymes XbaI and BlnI for digestion. It should be noted that this PFGE method was not a PulseNet method, and the results are not directly comparable to PulseNet data. The PFGE patterns for 98 strains of S. Enteritidis PT30 and 10 strains of S. Enteritidis PT9c were compared to several epidemiologically unrelated S. Enteritidis strains (Table (Table2).2). The S. Enteritidis PT30 strains had four closely related PFGE patterns with XbaI (types X1, X2, X3, and X7) and two PFGE patterns with BlnI (B1 and B2) (Table (Table2;2; Fig. Fig.1).1). The predominant PFGE pattern combination was X2-B1; 72 strains, including clinical strains from OB1, OB3, and numerous environmental strains, were X2-B1 (Table (Table2).2). The PFGE pattern combination X1-B1 was detected for clinical strains from Canada associated with OB1, almonds linked to OB1, and some environmental strains. Three PT30 strains isolated in Malaysia from almonds shipped from California had a profile designated X7-B1. This pattern was highly similar to the PT 30 X1-B1 profile except for subtle differences in the lower-molecular-weight bands (Fig. (Fig.1;1; Table Table2).2). Similarly, three human PT30 strains suspected of being linked to a small outbreak associated with tomatoes prepared in multiple taqueria restaurants in California (California Department of Health, personal communication) had the X2-B1 profile. These clinical strains were included for the comparison of genotypes to those of almond- and orchard-related strains. For all S. Enteritidis PT9c strains from OB2, there was only one XbaI (X6) pattern and one BlnI (B3) pattern. The PFGE patterns for both the S. Enteritidis PT30 and S. Enteritidis PT9c strains using either XbaI or BlnI were distinguishable from each other and from those of other S. Enteritidis strains analyzed in this study (Fig. (Fig.1).1). However, PT13a and PT8 strains had the XbaI (X4) pattern and BlnI B5 and BlnI B6 patterns, respectively. The PFGE patterns were almost identical, with the exception of a single band that is in the middle of BlnI gel lane 10 but missing in lane 4 in Fig. Fig.1.1. The high similarity in the PFGE patterns for both of these PTs augments the difficulty in determining the phylogeny of strains that are suspected of being related epidemiologically but that are unrelated spatially and temporally. Therefore, we analyzed a subset of strains by additional genotyping methods to assess their relatedness.
MLVA distinguished the S. Enteritidis PT30 and PT9c strains from the other S. Enteritidis strains in this study (Table (Table3).3). Of the S. Enteritidis PT30 strains analyzed, 20 of the 21 strains had identical alleles at each of the nine VNTR loci that were examined. This included all of the strains from the almond orchards that had been collected during a 5-year period. Of the S. Enteritidis PT30 clinical strains from OB1, OB3, and PT30 strains unrelated to the almond outbreaks, only strain RM4292 from OB1 had a distinct MLVA type; this difference was observed only at VNTR locus SE5 (Table (Table3).3). All five of the S. Enteritidis PT9c strains analyzed by MLVA in this study had identical alleles at each of the nine VNTR loci and were different from the major MLVA type of the PT30 strains at three VNTR loci (SE2, SE5, and SE7). All of the S. Enteritidis PT30 and the S. Enteritidis PT9c strains analyzed failed to amplify the SE3 locus located in the S. Enteritidis PT4 ROD21 genomic island. Thus, by MLVA, the PT30 and PT9c strains appeared to be similar within the PT. However, MLVA was able to discriminate the PT30 and PT9c strains from strains of the other five PTs tested at multiple loci and sometimes multiple tandem repeats (Table (Table33).
The genetic content of 15 S. Enteritidis PT30 strains, 4 S. Enteritidis PT9c strains, 8 S. Enteritidis strains with differing PTs, and an S. Typhimurium strain described in Table Table11 were examined by microarray. S. Enteritidis PT30 strains analyzed by CGI were the following: (i) a geographically diverse set of clinical strains associated with OB1 and OB3; (ii) sporadic clinical strains not associated with almonds and strains from a non-almond California outbreak (OB4); (iii) environmental strains from almonds related to OB1; (iv) strains from almond-processing equipment related to OB1; and (v) strains obtained during a 5-year period from an almond orchard that were related to OB1. Three S. Enteritidis PT9c clinical strains (OB2) and one S. Enteritidis PT9c isolate from almonds during a survey (9) also were analyzed for comparison.
For the comparative genomic analysis, an index of the gene content (present/divergent or unknown/absent) relative to those from the DNA microarray (4,398 genes from S. Typhimurium LT2 and 178 genes from S. Enteritidis PT4) was generated for each strain (Fig. (Fig.2).2). The genomic relationship among the S. Enteritidis and S. Typhimurium strains was determined by hierarchical clustering analysis using a standard correlation function and bootstrapping. The cluster analysis demonstrated that strains grouped together into clusters according to their PT (Fig. (Fig.2).2). Interestingly, the S. Enteritidis PT30 and PT9c strains formed a distinct cluster away from the other S. Enteritidis PTs, and as predicted, all of the S. Enteritidis strains examined in the study were clearly distinct from the S. Typhimurium strain. However, clustering analysis did not further differentiate the PT30 strains according to outbreak (OB1, OB3, or OB4), source, or time of collection.
The genetic content as determined by CGI analyses of S. Enteritidis strains indicated that the majority of the differences between the S. Enteritidis strains with distinct PTs corresponded to genes of the bacteriophage Fels-2 of S. Typhimurium LT2 and the bacteriophage-related genes of S. Enteritidis PT4 (Table (Table4).4). Specifically, the majority of genes of the bacteriophage Fels-2 were absent from S. Enteritidis PT30, PT9c, PT6a, and PT4 strains, whereas they were present frequently in S. Enteritidis PT13a, PT8, and PT33. Among the genes in bacteriophage Fels-2, the STM2705 gene was present in S. Enteritidis PT30 and PT33 strains and absent from S. Enteritidis PT4, PT6a, PT13a, PT8, and PT9c strains. The S. Enteritidis PT4 SE14 and SE20 prophages had regions that were present variably in the other PTs. The S. Enteritidis PT4 ROD21 genomic island was absent from PT30 and PT9c but present in PT4, PT6a, PT13a, PT8, and PT33. A number of S. Typhimurium LT2 bacteriophage regions were mostly absent from all S. Enteritidis strains (Table (Table4).4). These regions included most genes of the Fels-1 and Gifsy-1 bacteriophages as well as the putative phage-related genes STM2230 to STM2234 and STM4195 to STM4218 (data not shown). These data from the CGI analyses are consistent with our previous genotyping findings, indicating that the 30 and PT9c strains are highly related compared to the lack of relatedness of S. Enteritidis strains of different PTs and sources.
Furthermore, the CGI results revealed that S. Enteritidis PT30 strains possessed several genes that were absent from PT9c strains. In particular, these genes included three prophage regions from S. Enteritidis PT4: SEN1137 to SEN1139 of the Gifsy-2-like prophage, SEN1389 to SEN1391 of the SE14 prophage, and SEN1958 and SEN1959 of the SE20 prophage. Also included were genes STM1546 and STM1547, which are not associated with any prophage (Table (Table4).4). We confirmed these CGI results by PCR (data not shown).
Within the CGI data set, there were a few genes that were present or absent variably among PT30 strains, suggesting some genomic variation. Two of these variable genes, SEN1958 and SEN1964, then were examined by PCR in the PT30 strains. The results of the PCR demonstrated that in all PT30 strains SEN1958 was present, while SEN1964 was absent (data not shown). These PCR results suggest that differences in CGI-measured genetic content for some genes in PT30 strains are due partly to experimental variation in the microarray analysis rather than genomic content variation.
Due to the reported persistence of S. Enteritidis PT30 in the orchard environment (30), we reasoned that S. Enteritidis PT30 exhibited a distinct metabolic fingerprint that may confer an environmental fitness compared to that of other S. Enteritidis strains. To examine this possibility, a representative group of S. Enteritidis PT30 orchard strains (>20), clinical strains (10 strains from both Canada and different locations in the United States), and strains representing four other PTs (PT4, PT8, PT13a, and PT33) were tested in the Biolog assay with GN2 microplates (Table (Table5).5). The S. Enteritidis PT30 clinical strains and four strains of other S. Enteritidis PTs exhibited similar metabolic fingerprints and generally were distinct from the metabolic fingerprint of the S. Enteritidis PT30 orchard strains (Table (Table5).5). Of 95 different substrates tested with the Biolog assay, the largest differences were with l-aspartic acid, l-glutamic acid, l-proline, l-alanine, and d-alanine. These substrates induced higher respiratory responses in S. Enteritidis PT30 clinical strains and the five other S. Enteritidis PTs compared to the responses by S. Enteritidis PT30 orchard strains. One exception was RM3966, an orchard isolate that had a metabolic profile comparable to that of the clinical PT30 strains.
Three outbreaks of salmonellosis due to the consumption of raw almonds have been caused by two rare PTs of S. Enteritidis: PT30 and PT9c. S. Enteritidis PT30 was associated with two temporally distinct outbreaks occurring at the end of 2000 and beginning of 2001 (6, 13) and the end of 2005 and beginning of 2006 (15); S. Enteritidis PT9c was associated with an outbreak in 2003 to 2004 (3). The involvement of S. Enteritidis PT30 in two outbreaks more than 3 years apart may reflect fitness characteristics facilitating persistence in the orchard environment. The isolation of S. Enteritidis PT30 strains during a 5-year period from an orchard linked to OB1 and indistinguishable by PFGE from outbreak strains is consistent with a persistence hypothesis (30). In addition, the thought that the PT30 and PT9c strains are related genetically, even though they are distinguishable by PFGE, stimulated us to analyze whether MLVA and CGI could measure any genetic differences among the almond outbreak strains (S. Enteritidis PT30 and PT9c) compared to results with PFGE. Our study demonstrated that the CGI facilitated the comparison of the gene content of S. Enteritidis PT30 and PT9c relative to each other and to S. Enteritidis of different PTs.
It has been reported previously that S. Enteritidis strains are quite homogeneous within a particular PT (21, 23, 27). By using all three genotyping methods, our study also indicated that S. Enteritidis PT30 and PT9c strains were related genetically within their PT. It is noteworthy that particular S. Enteritidis PT30 strains with distinct PFGE and MLVA results were not distinguished as a separate cluster in the microarray-based CGI analysis. These differences in the discriminatory power of the genotyping methods may be consistent with single-nucleotide polymorphisms (SNPs) and small insertions/deletions, sometimes at tandem repeats, resulting in restriction site and fragment size differences or priming differences measurable by PFGE and MLVA, respectively, without affecting hybridization ratios measured on microarray-based CDS analysis (23). Moreover, neither PFGE nor MLVA could distinguish all of the S. Enteritidis PT30 from various sources. In particular, these two genotyping methods did not discriminate S. Enteritidis PT30 strains relating to OB1 from the ones relating to OB3, nor could S. Enteritidis PT30 almond outbreak-related strains be distinguished from unrelated S. Enteritidis PT30 clinical strains. Specifically, the subtle differences detected in the low-molecular-weight region of PFGE profiles were not informative of any epidemiological and/or source information available (e.g., spatial, temporal, outbreak). Furthermore, the results of the present study showing similar MLVA types for S. Enteritidis PT30 regardless of source contrast with results observed previously in two separate studies for other S. Enteritidis PTs (4, 7). In those previous studies, encompassing over 100 strains each, MLVA was determined to have sufficient discriminatory power as calculated by Simpson's index of diversity and demonstrated that S. Enteritidis strains, which had the same PT but were temporally and geographically distinct, could differ in their MLVA types (4). Thus, our results suggest that our S. Enteritidis PT30 strains represent a common geographical source (California) and have VNTR loci that are too stable temporally within S. Enteritidis PT30, and perhaps other PTs, for the fine resolution of epidemiological relationships. One possibility is that the MLVA method is more adequate for differentiating strains of wider temporal and geographic distribution.
CGI data identified several of the precise genetic differences between S. Enteritidis strains of different PTs. The cluster analysis of the CGI data indicated that the almond-related PTs (30 and 9c) were more closely related to each other than to the other PTs assessed (Fig. (Fig.2).2). In contrast to other S. Enteritidis PTs analyzed in this study, neither of the almond-related PTs (30 and 9c) possessed the ROD21 genomic island that is predicted to encode a putative type IV pilin system, and PT9c strains also lacked STM1546 and STM1547, which encode a hypothetical protein and a putative marR family transcriptional regulator. It is unknown whether the absence of these genes affects the physiology of the PT30 and PT9c strains and their fitness in the almond environment. Other differences we observed between the PTs were predominantly in prophage regions, which is in agreement with previous reports (21, 27), and these prophages may be partly responsible for different PTs. However, we did not observe any significant differences in the content of prophage genes between PT8 and PT13a strains, suggesting that mechanisms other than prophage differences can alter the PT. Indeed, a curious relationship between PT8 and PT13a was demonstrated by the phage type conversion from PT8 and PT13a that occurred following the introduction of a plasmid (5). Moreover, it is possible that the different phage types of S. Enteritidis possess distinct genes specific for each phage type but that are not present on the microarray for analysis by CGI.
Although metabolic analysis by Biolog assay indicated that S. Enteritidis PT30 clinical strains were similar metabolically to the other S. Enteritidis PTs examined, the analysis revealed differences in S. Enteritidis PT30 strains that were from different sources but highly related genotypically. In particular, clinical strains responded to (and possibly utilized) amino acids l-aspartic acid, l-glutamic acid, l-proline, l-alanine, and d-alanine more efficiently than the S. Enteritidis PT30 orchard strains. Similar metabolic differences were reported recently for common S. Enteritidis PTs and S. Enteritidis PT11 (24); however, the metabolic differences appeared to be related directly to the absence of putative amino acid transport genes in the PT11 strains. In our investigation, the metabolic differences occur between genotypically identical S. Enteritidis strains of the same PT and are indistinguishable by PFGE, MLVA, and CGI. These findings suggest that differences in metabolism result from selected adaptations for increasing fitness in an orchard environment and maintained subsequently to isolation from the orchard. A significant association between environmental fitness and pathogenesis for humans is relevant for understanding the mechanisms of food-borne illness, the increasing number of outbreaks, and the increasing virulence of some pathogens (17). Thus, the fact that phenotypic differences in PT30 strains could be measured in the absence of correlation with genotype illustrates the need for higher-resolution genotyping methods, the measurement of functional activities related to fitness in the environment (e.g., animal, soil, water, plants, and biofilms), and other methods for studying food-borne pathogens and outbreaks. SNPs within the cyaA gene previously have been shown to discriminate between S. Enteritidis strains with the same phage type (11, 20); however, both the PT30 and PT9c strains possessed a single cyaA allele that was identical to the S. Enteritidis PT4 strain P125109 allele (data not shown). Currently, we are performing whole-genome sequencing of representative S. Enteritidis PT30 and PT9c strains to facilitate the identification of novel genes and SNPs that may be related to observed phenotypic differences and for the development of improved methods for genotyping S. Enteritidis strains.
We are indebted to Jay Hinton, Sacha Lucchini, and Arthur Thompson at the Institute for Food Research, Norwich, United Kingdom, for providing PCR products for S. Typhimurium LT2. We thank Sharon Horn, Felicidad Bautista, and Anna Bates for technical assistance in this study. We are grateful for the cooperation and help of the Almond Board of California and the California almond industry.
This work was funded by the Almond Board of California and U.S. Department of Agriculture Cooperative State Research Education and Extension Service project 2002-03886. This work also was supported by the U.S. Department of Agriculture, Agricultural Research Service, CRIS project 5325-42000-045.
Published ahead of print on 2 April 2010.