Listeria monocytogenes (Lm) is one of the leading causes of death because of foodborne illness, affecting the elderly, pregnant women, neonates, and people who are immunocompromised. Serologically, Lm can be classified into 13 serotypes, although only 4 are typically linked with food contamination and illness. Since 2000, a shift in serotypes involved in listeriosis outbreaks has been observed, suggesting that tracking of serotypes could help identify emerging trends. A PCR method developed in 2004 allowed detection of the four major serotypes as molecular serogroups, corresponding to broad phylogenetic groups. In this study, a novel quantitative PCR (qPCR) method was developed that uses two multiplex qPCRs, one to confirm the Listeria genus and Lm species and the second for Lm molecular serogrouping. This method was compared with the U.S. Food and Drug Administration Bacteriological Analytical Manual (BAM) method for Lm and the seroagglutination method, using a 208-strain panel. Comparison of the genus and species qPCR assay with the BAM methods found an equal or slightly higher accuracy for the qPCR method (>98%), compared with the BAM protocol (>96%), when evaluated against independent characterization data. Molecular serogrouping using the qPCR method (96.6%) was more accurate than the seroagglutination assay (75.6%). The qPCR method identified Lm 4bV strains, which could not be resolved using seroagglutination. The qPCR could not identify lineage III and IV serotype 4b strains but did correctly identify 16 of 18 lineage III and IV strains. The qPCR method performed genus identification for the Listeria species Lm, L. innocua, L. welshimeri, L. ivanovii, and L. seeligeri. In addition, the method performed species identification for Lm and classified Lm into six molecular serogroups: 2A, 2B, 2C, 4B, NT, and 4bV. This method provided a rapid and accurate confirmation of Lm and serogroup determinations; furthermore, it could help identify otherwise unlinked strains by enabling whole genome sequencing analysis based on broad phylogeny, independent of other information.
The new qPCR method identified major members of Listeria spp. (excluding L. grayii) and members of the Lm species.
The new qPCR-based method identified major serogroups of Lm.
The method was at least equal to the BAM method when identifying Lm species.
Molecular serogroup determination was more accurate than the seroagglutination assay.
Listeria monocytogenes (Lm) is the causative agent of invasive listeriosis, a severe but rare infection that is one of the leading causes of death from foodborne illness. The species can be divided into 13 serotypes, although 3—1/2a, 1/2b, and 4b—are responsible for more than 95% of illnesses (36). A fourth serotype, 1/2c, is commonly isolated in food processing environments, as are 1/2a and 1/2b (36). Identification of these serotypes, as part of a surveillance effort, can provide valuable information about trends in contamination events and illness.
Two options for serotyping Lm are described in the U.S. Food and Drug Administration Bacteriological Analytical Manual (BAM), chapter 10, section I.1, with differing levels of specificity (19). The first option, used more regularly because of reagent availability and simplicity, is a simple serogroup agglutination assay that identifies an isolate as type 1, type 4, or neither. Type 1 strains include serotypes 1/2a, 1/2b, and 1/2c strains, representing distinct phylogenetic groups (28). Conversely, type 4 strains include serotypes 4a, 4b, 4c, 4d, and 4e, which are also found in distinct phylogenetic groups (28). Therefore, subgrouping Lm strains into type 1 and type 4 has little significance for disease association and/or phylogeny. The more complete method involves a three-step seroagglutination process. The first step uses a pair of antisera to determine which downstream, more specific antisera sets should be used in the second step to identify the somatic antigen (O antigen) group. The third step, completed only after several days of culture to enhance flagellar expression, identifies the flagellar antigen (H antigen) group, which allows the final determination of serotype. This method provides specific serotype information that follows phylogenetic grouping, but it is extremely time-consuming and labor-intensive. The availability of the agglutination-based serotyping reagents, single source only, can also make timely acquisition of the reagents challenging. For these two reasons, the second method is rarely performed, resulting in a loss of information. In addition, both methods use a subjective determination of seropositivity that requires a skilled technician, familiar with the method, and can still result in errors. Variable expression of both O and H antigenic determinants can also lead to inaccurate serotype identification of strains.
Doumith et al. (14) developed a multiplex conventional PCR method capable of differentiating most of the 13 serotypes into four molecular serogroups: 2A (1/2a and 3a), 2B (1/2b, 3b, and 7), 2C (1/2c and 3c), and 4B (4b, 4d, and 4e). In addition, this method can detect a subset of 4b strains, termed 4bV or IVb-v1, whose genomes contain a 6.3-kb fragment typically seen only in 1/2a, 1/2c, 3a, and 3c (lineage II) strains (22, 24, 25, 31, 39). Strains lacking a detectable amplicon fall into a fifth category, which appears to be largely composed of serotype 4a and 4c strains, as well as some 4b strains. This latter group has been previously called nontypeable (NT) because of the absence of amplicons. These molecular serogroups also correspond with phylogenetic lineages that divide Lm into four broad molecular lineages: lineage I, which is composed of molecular serogroups 2B and 4B; lineage II, which is composed of molecular serogroups 2A and 2C; and lineages III and IV, which are within molecular serogroup NT (14, 28). However, the method by Doumith et al. (14) relies on the use of gel electrophoresis and visual band identification, which can be subjective for faint bands, particularly for the lmo1118 marker. Amplification of lmo1118 is typically reduced because of a larger amplicon size that can lead to lower amplification efficiency when compared with the other, shorter amplicons. In addition, this method did not include an internal amplification control (IAC), which could result in inaccurate identification of NT strains.
The molecular serogroup information can be critical in surveillance efforts, because biases have been observed for these four lineages. Numerous studies have pointed to a higher proportion of lineage I strains among clinical isolates, whereas strains in lineage II, especially molecular serogroup 2A, are more commonly isolated from food and environmental sources; however, geographical differences have also been noted (16, 21, 28). Conversely, the last two lineages, lineages III and IV, are rarely associated with human illness and are more typically associated with animals (3, 20, 39). In addition, there are biases among Lm isolates from clinical cases regarding serotypes, such as a higher association of 1/2b strains with non–pregnancy-associated cases versus 4b strains with a higher association with pregnancy (28). Tracking the prevalence of these subtypes within various sources, e.g., food, food processing environments, natural environments, animals, and humans, can reveal critical information about shifting risk patterns and emergence of new genotypes or clones of Lm.
Association of subtypes with sources and risk of illness can provide valuable information when developing regulatory guidelines, based on risk analyses, because the types of foods associated with listeriosis continue to expand. For example, until this past decade, listeriosis was not associated with fresh produce, except for sprouts, unless the produce was included in a prepared food, such as cabbage in coleslaw or celery in chicken salad. Furthermore, frozen commodities were considered to have minimal risk because of the lack of growth potential in these foods. Recent outbreaks involving cantaloupe, caramel-covered apples, stone fruit, packaged salads, frozen vegetables including corn, and ice cream have changed our understanding of which foods present a risk to public health (9–13, 15). Whether this results from changes in the pathogen, host, or food and its production is unclear, but analyses of subtyping and serotyping determinations of Lm isolates from clinical, food, and environmental sources may provide information to guide more effective risk mitigation.
Separately, the identification of five recent outbreaks caused by 4bV strains (7), with four of these outbreaks linked to a single clade, suggests the need for a method to rapidly identify this recently emerged, or at least recently detected, group to aid regulatory analysis and surveillance. The 4bV strains are members of serotype 4b, which harbor a 6.3-kb DNA island that is found elsewhere only in lineage II strains (e.g., 1/2a, 1/2c, 3a, and 3c) (26) and has been acquired by the 4bV strains through lateral gene transfer. Lineage II strains are typically associated with food and environmental samples, whereas serotype 4b strains have a higher prevalence in clinical cases, especially foodborne outbreaks. Although 4bV strains have been retrospectively identified in older clinical cases, dating back to 1959 (25), there is no information about the frequency of these strains before 2004. Even after the development of the PCR assay that detects this variant, testing for these strains has been sporadic, continuing to leave a gap in our understanding about the historical and current prevalence of 4bV strains. However, the 4bV strains have recently become of increased interest in food safety, because they have been associated with four outbreaks of listeriosis, three of these linked to fresh produce (5, 10, 12, 13). Even more concerning, an examination of whole genome sequencing (WGS) submissions to NCBI suggests the possibility that these strains are expanding as increasing numbers of 4bV strains have been identified; however, this may result from increased sequencing efforts that facilitate screening for these strains (2, 7). Surveillance of isolates obtained from clinical and food samples, as well as the natural environment, will aid in understanding this emerging threat to public health.
Two real-time PCR methods for serogrouping Lm based, at least partly, on the Doumith et al. (14) method have been developed and evaluated (1, 38). However, these studies did not compare the accuracy of the method with traditional standard serological methods, instead assessing their method against the conventional PCR method developed by Doumith et al. (14). Furthermore, these studies only involved strains from lineages I and II, with none from lineages III and IV. Evaluation of a serogrouping method against a diverse panel of strains using the traditional serotyping protocol is critical in establishing the accuracy of the new typing scheme relative to the original standard, as well as identifying deficiencies. The correct identification of these molecular serogroups can be used for various purposes once the limitations are understood.
Previously, our group reported the incorporation of the Lm species–specific marker hly, coupled with the BAM type 1 and type 4 agglutination assay (19). This combined approach allowed differentiation of 1/2a, 1/2b, and 1/2c strains from 3a, 3b/7, and 3c strains, respectively (8). Addition of this type 1/type 4 agglutination procedure also verified that strains with an NT PCR result were type 4 positive, supporting their inclusion in serotypes 4a or 4c (8). In this study, we developed a quantitative PCR (qPCR) multiplex method for genus and species confirmation of colony isolates as Listeria spp. or Lm, as well as for molecular serogrouping of confirmed Lm. This novel qPCR method is based on the work of Doumith et al. (14) and can identify molecular serogroups in a format conducive to the workflow associated with high-volume laboratories. The workflow is high throughput, with no need for gel electrophoresis or subjective agglutination determinations. Data generated on a real-time PCR platform can be rapidly analyzed, and the inclusion of an IAC reduces the risk of false-negative results. This method includes both genus- and species-specific markers, iap and hly, respectively. In addition, the method permits screening of Lm isolates to identify 4bV strains, track lineage associations with sources and illnesses, and phylogenetically group strains for WGS analysis during preparation of isolates for sequencing. This method would seamlessly integrate with the current BAM protocols (19) and rapidly provide data to facilitate the monitoring of new trends in listeriosis, which may lead to reduction in future illnesses.
MATERIALS AND METHODS
A total of 208 strains were included in this study. Of these, 120 were previously identified as Lm (Supplemental Table S1), as well as 20 Listeria innocua, 15 Listeria welshimeri, 11 Listeria seeligeri, and 9 Listeria ivanovii (Table S2). This panel also included 33 non-Listeria strains (Table S2). The strains were assigned a blinded identifying (ID) number that was used during subsequent evaluations (Tables S1 and S2).
Preparation of the IAC template
The IAC template, a novel 79-bp DNA sequence from the human adenovirus (35), was amplified using the IAC_F and IAC_R primers (Table 1) and Platinum Blue PCR Supermix (Invitrogen, Carlsbad, CA), with an initial 5 min at 95°C to activate the Taq DNA polymerase enzyme followed by the cycling protocol for 30 cycles and a final extension step of 7 min at 72°C. The cycling protocol was (i) 95°C for 30 s, (ii) 60°C for 30 s, and (iii) 72°C for 30 s.
After amplification, a PCR cleanup was performed with the MinElute PCR cleanup kit (Qiagen, Germantown, MD) to remove residual primers and salts from the initial reaction mixture. Subsequently, the reaction purity was verified via 1% agarose gel electrophoresis. The concentration of the IAC template was determined using a NanoDrop spectrophotometer (Thermo Fisher Scientific, Waltham, MA), and a freezer stock was made at a concentration of 0.05 ng/μL, which was subsequently added to the primer mixes as noted in Tables S3 and S4.
The qPCR was performed using QuantiFast MM without ROX (Qiagen), along with primer mixes designed for the two-part qPCR to verify genus and species (G/S) and serogroup (sero). Each primer mix included primers for the Listeria targets, as well as primers for the IAC template (Table 1). Primer mixes were made separately for the G/S and sero reactions (Tables S3 and S4, respectively). Probe stocks (Table 1) were maintained in separate tubes suspended in 10 mM Tris-EDTA pH 7.4 buffer.
Two separate reaction mixes were prepared: one for the G/S reaction described in Table S5 and one for the sero reaction noted in Table S6. These reaction mixes were used with a template prepared as described previously (8). The template was a lysate prepared by resuspending a single colony measuring approximately 2 mm in diameter, or two to three colonies summing to a similar diameter, in 50 μL of lysis buffer (0.05 N NaOH, 0.25% sodium dodecyl sulfate). The suspension was heated 15 min at 99°C and then cooled at room temperature briefly. Once cooled, 100 μL of sterile distilled water was added to the lysate and the preparation was centrifuged at a force of ≥14 × g for 5 min. Each reaction contained 9.5 μL of the reaction mix and 0.5 μL of the template. The template was collected from the surface of the lysate, near the meniscus, because collection from lower areas of the preparation could result in PCR inhibition.
After the template was added, the plate was sealed with optical adhesive film (Thermo Fisher Scientific, Catalog 43-119-71) and run in an ABI 7500 real-time PCR system (Applied Biosystems, Foster City, CA), as earlier, with fluorescence data collected at the end, using appropriate dye detection selections and disabling of the ROX control. qPCR processing was performed in triplicate, whereas BAM analysis was performed by a separate researcher to prevent results from one assay affecting the assessment of the other.
The cycle threshold was adjusted to determine the cycle quantification (Cq) value near the start of the exponential phase of the sample's sigmoidal curve, indicating proper amplification, and above any observed background signal. Although it would be desirable to have a set threshold for all reactions, the crude nature of the lysate prep required careful consideration of the reaction to make sure extraneous factors from the template material did not result in false negatives or false positives. For example, background fluorescence from proteins or membrane contaminants in the crude template preparation could result in a false-positive result. The threshold was set as low as possible within the exponential phase, while keeping the threshold above nonsigmoidal curves until after cycle 30 to ensure that only specific amplification reactions were considered. Samples qualified as positive if the Cq was less than 30 and the end relative fluorescence unit value surpassed 100. If IAC amplification was negative, the reaction was repeated.
For the G/S reaction, samples with a negative result (Cq > 30) for iap and hly were considered non-Listeria spp., whereas samples positive (Cq < 30) for iap but not hly were Listeria spp. Samples positive for both iap and hly were confirmatory for Lm.
For samples determined to be Lm, the sero reaction was performed. Listeria isolates other than Lm were not serotyped. The amplification pattern was used to identify the serogroup. This pattern is based on the original patterns from the initial conventional PCR method, with the added profile for the 4bV group (8, 14, 25).
Analysis based on the BAM protocol
In parallel, isolates were inoculated onto Rapid'L.mono (RLM) agar (Bio-Rad, Hercules, CA) to identify Listeria spp. and Lm in concordance with the BAM (19). Isolates showing a typical Lm colony phenotype were further tested with Analytab Products (API) Listeria strips (bioMérieux, Inc., Durham, NC). Additional API Listeria tests were run to evaluate ambiguous results or to verify a subset of strains as Listeria spp. Seroagglutination assays were performed for those strains determined to be Lm using the kit available from Denka Seiken (Tokyo, Japan) or Mast Group (Bootle, UK), according the manufacturer's protocol, both distributed by Hardy Diagnostics (Santa Maria, CA). Blinded ID strains 157 to 208 were added to expand the serotype analyses. These strains were evaluated to verify that they were Lm using only RLM and then assessed for serotype using the Mast Group kit. The first 156 strains were subjected to the complete BAM analysis, with serotyping performed using the Denka Seiken kit.
In the current study, a process was needed to determine result accuracy for those strains that lacked well-characterized serotype information and revealed results that were deemed nonequivalent. Agglutination-based serotyping has a greater degree of resolution than the qPCR method, because it can distinguish 13 serotypes versus 5 serogroups. Therefore, if the qPCR molecular serogrouping assay results identified a strain as a member of a serogroup that would include the serotype identified via agglutination, as well as prior characterization as noted in Table S12, the results were deemed equivalent. Strains not meeting this result were considered to have conflicting results, or to be nonequivalent. To resolve these latter results, WGS comparisons were performed using a data set that included well-characterized strains and strains requiring independent verification. The well-characterized strains allowed clear identification of phylogenetic groups that could be used to identify the WGS group of the strains with nonequivalent results, based on the groups in which these strains were clustered. The WGS subgroups were named in a manner corresponding to the molecular serogroups identified using qPCR. This would allow determination of the similarity of results.
WGS was obtained through sequence read archive (SRA) files already available at NCBI or, for those strains lacking WGS, genome sequencing was performed on a MiSeq sequencing platform (Illumina, San Diego, CA). Tagmentation and library indexing were performed according to the GenomeTrakr protocols as described previously (6, 37). The SRA files were uploaded to NCBI and biosample numbers are provided in Table S1, along with the accession numbers used for those strains previously sequenced. The FASTQ files were analyzed using the Center for Food Safety and Applied Nutrition (CFSAN) single-nucleotide polymorphism (SNP) pipeline (30), with the Lm strain F2365 genome as the reference genome. The SNP alignment fasta file was used to generate phylogenetic trees in Molecular Evolutionary Genetics Analysis 7.0.26 (MEGA7) (18) using the maximum likelihood algorithms, with bootstrap testing of the phylogeny result (n = 1,000) (32), using the Tamura-Nei substitution model. Phylogenetic groups were identified and used to assess the accuracy of the BAM and the novel qPCR methods.
A second WGS analysis was employed, because the SNP pipeline was unable to resolve the serotype designation of strains within lineages III and IV. For a serotype-based assessment, this was a critical need, because serotype 4b strains have been identified in these lineages. Characterization of Listeria genome assemblies was performed using a local BLAST search with target loci designed to classify strains into species, lineage, and serotype. Species were determined using the prs gene, with species organized according to the classification system of Orsi and Wiedmann (29). Strains classified as L. innocua, L. ivanovii, L. marthii, Lm, L. seeligeri, or L. welshimeri by prs were further characterized; internal fragments of three housekeeping genes commonly applied in multilocus sequence typing analysis (cat, dat, and lhkA) were used, resulting in lineage determination for Lm strains and confirmation of the prs assignment for the other Listeria species. Strains of Lm were then molecularly serotyped using three genes found within the cell wall teichoic acid gene cluster: the tagH gene of the ATP-binding protein encoding the teichoic acid ATP-binding cassette transporter, an unnamed gene encoding a putative membrane protein (abbreviated as memP here), and an unnamed glycosyltransferase gene (abbreviated as glyT here). Genes used for the typing tool, termed the Listeria typing tool (LTT), are described in Table S7, and the interpretation of results is outlined in Table S8.
Accuracy was assessed as the percentage correctly detected. The 95% Wald confidence intervals (CIs) were calculated around the accuracy, with nonoverlapping CIs signifying significant differences. Agreement was assessed using Cohen's kappa (κ) with 95% CIs.
Sensitivity (percentage of isolates with the indicated subtype correctly detected) and specificity (percentage of isolates not belonging to the indicated subtype that were correctly not detected) were calculated for each subtype. The 95% Wald CIs were included.
Genus and species evaluation
A total of 208 strains were included in this study, although only 156 were analyzed fully by the BAM protocol for both genus and species and serotyping characterization (Tables S1 and S2). The strains evaluated in this study were analyzed as described in Figure 1, which shows a generalized outline of the workflow. Many Lm strains were also analyzed by WGS to verify results (Table S1). The BAM protocol provides four options for secondary confirmation of Lm. The API Listeria strips were used in this study as the secondary confirmation when analyzing the first 156 strains described in the panel. This system has a robust database for Lm identification, partially provided in the product insert, is commonly used in regulatory laboratories, and requires no special equipment. Critically, because the qPCR method evaluated in this study shares targets with the BAM qPCR method, the genus portions of both qPCRs were essentially identical. Use of the BAM qPCR assay would therefore result in duplicate results, rather than a true independent assessment of the new qPCR method.
The G/S and sero qPCR data obtained are included in Tables S9 and S10, respectively. Results from using the BAM protocol are described in Table S11. A detailed comparison of the results is presented in Table S12, and a summary of the results is found in Table 2.
Agreement across the qPCR G/S reactions was strong; only one strain (ID 136) had differing results between qPCR replicates. Separately, for most strains evaluated, the BAM serotyping and qPCR serogrouping methods agreed with each other and with prior historical strain characterizations. However, in three instances (ID 6, 67, and 112), the BAM method failed to identify Lm strains because of lack of the typical Lm colony phenotype displayed on RLM agar. Subsequent testing using the API Listeria test strips verified that these three strains were Lm, but in a typical laboratory evaluation, these strains would have been excluded before this step. A fourth strain (ID 24) could not be verified as Lm by the API Listeria test strip because of a nonstandard profile. Two strains (ID 3 and 136), previously identified as Lm, were identified as Listeria spp. by both the BAM and the qPCR methods in this evaluation. WGS analyses of these strains suggested that strain LS168 (ID 3) was inaccurately identified as Lm in its original characterization (Fig. 2). The other strain, LS95 (ID 136), was classified as Lm in the WGS analysis (Fig. 2). This strain was deliberately included in the panel, because it has a history of being challenging to identify and serotype. LS242 (ID 24) was identified as Lm on RLM agar but gave atypical results in the API Listeria test, as noted earlier. However, qPCR and WGS analysis both agree that this strain is Lm.
In addition, RLM agar failed to properly identify blinded ID 30, 63, 113, 126, and 143. These were L. ivanovii strains that should typically show phospholipase C activity and xylose fermentation phenotypes on RLM agar, allowing them to be differentiated from Lm. However, because of weak xylose fermentation on RLM agar, these strains were initially identified as Lm. Subsequent testing using API Listeria test strips was required to correctly exclude these strains. The qPCR assay correctly excluded all of these strains as non-Lm. Evaluation of blinded ID 157 to 208 strains via the BAM protocol identified all of these strains as Lm when they were grown for the full 48 h on RLM agar. However, a shorter incubation would have been problematic, because three of these additional strains (ID 167, 181, and 193) had atypical Lm colony morphologies at 24 h. At 24 h, these colonies were not blue, indicating reduced phospholipase C activity, and the agar around the colonies was slightly yellow, indicating possible weak xylose fermentation. Although these phenotypes resolved properly after the full 48 h, these observations highlight the critical need for a full 48-h incubation before interpreting colony morphology. Finally, although not critical to the evaluation of this method for Lm identification, both the BAM and the qPCR G/S methods identified LS537 (ID 23) as non-Listeria even though prior characterization suggested it was L. welshimeri. WGS analysis was not performed on this strain, because it was still properly excluded as not Lm, but it may be that this strain, like LS168, was misidentified originally. Alternatively, it may represent an atypical Listeria spp.
Overall accuracies for the BAM and qPCR methods to identify strains to the genus and species level as Lm were 96.8 and 98.7%, respectively (Table 2). Agreement was found to be strong, with no significant difference between the methods (Fig. 3). These results suggest that the novel qPCR method was comparable to the BAM Listeria method for genus and species identification.
Strains that were misidentified as not being Lm during the blinded BAM protocol G/S testing were included in the subsequent seroagglutination analysis to ensure a complete dataset for serotype assessment was obtained. This was accomplished by partially unblinding the strains to identify missing Lm strains. Difficulties occurred when assessing the accuracy of the qPCR method versus the agglutination method, because prior characterization by external labs was typically limited to a type 1/type 4 seroagglutination assay, as noted in the BAM protocol (19). Previous conventional PCR results were used to supplement this when selecting the panel. However, using prior PCR-derived data would reasonably bias the assessment of a qPCR assay derived from that initial method. Instead, WGS analysis was used to separately serogroup strains, because serotypes and molecular serogroups correspond with phylogenetic clusters. Strains with incomplete prior serotyping data or disagreement within current results were assessed via WGS to resolve disagreements between the agglutination assay (Table S9) and the qPCR serotype determinations (Table S8) (23). Comparisons of the BAM and qPCR methods were made between each other, with prior characterization data, and with WGS data where applicable (Table S10). A summary of the results of the serogrouping analysis can be found in Table 2.
For the 119 confirmed Lm strains, results between the agglutination and the molecular serogrouping qPCR assays were found to be equivalent, as defined in “Materials and Methods,” for 89 strains. However, 4 of these were only able to be serotyped because of a modification of the serotyping protocol. These strains showed no agglutination with either of the sera in the first step of the O-antigen typing, which is required to identify the antisera used in the next part of the analysis. These four strains (ID 65, 74, 99, and 181) were tested with all six sera in the second portion of the assay because of the initial negative result. In addition, the BAM G/S results would have excluded three other strains from serotype assessment, while a fourth was excluded by both the qPCR and the BAM G/S assays. However, we elected to assess the accuracy of the agglutination-based serotyping assay independent of G/S identification in this study, so these strains were all analyzed. In an overall analysis, accuracy was even lower when combined with the G/S identification errors, especially for the agglutination assay results, because the qPCR assessment included no partial unblinding to ensure inclusion of all Lm strains in the sero analysis.
Results were also compared with externally provided serotype information and were found to largely agree. However, for some strains, there was disagreement when assessed by the agglutination assay as to whether these strains were type 1 or type 3 when compared with external results. In some instances, the seroagglutination assay was unable to resolve a type 4 strain to its correct serotype. However, because the qPCR results would not resolve all of these details either, these errors were not considered when assessing the qPCR serogrouping accuracy unless the molecular serogroup could differentiate them.
In this study, 10 strains belonged to molecular serogroup 4bV. Nine of these strains were correctly identified as 4bV, whereas the 10th strain (ID 169) appears to have been swapped during sample processing for the qPCR analysis. Subsequent reanalysis by qPCR from the same culture used for the agglutination assays identified this strain as 4bV (data not shown), supporting this conclusion. The inability of agglutination assays to identify 4bV strains is a well-known limitation of the agglutination assay, so none of these strains could be identified via the BAM protocol.
For the remaining strains, an independent method was required to determine whether the qPCR or BAM serogroup determinations were accurate. This approach used the CFSAN SNP pipeline and LTT to serogroup via WGS data, as described in the supplemental text. Based on this analysis, it was determined that the CFSAN SNP pipeline analysis, partnered with the LTT analysis for lineages III and IV, was suitable for determine the accuracy of the sero analyses.
The results of the LTT revealed weaknesses in both the qPCR and the traditional agglutination methods when evaluating lineage III and IV isolates. They showed that 5 of the 18 lineage III and IV strains were unable to be serotyped beyond a determination as type 4 using the seroagglutination assay. Conversely, the qPCR method correctly identified all lineage III strains as NT but could not differentiate serotype 4b strains from serotype 4a or 4c strains within this lineage. Evaluation of the lineage IV isolates was more challenging, given that there are only three such isolates in this study. However, the qPCR incorrectly assessed two of these isolates as serogroup 2B, because they were positive for ORF2819, indicating a 2B serogroup identity. Evaluation of the sequence file for these two strains verified the presence of ORF2819 within these genomes (data not shown).
Analysis of the accuracy for the sero subtyping could be considered two ways. One was by serotype correlation. However, because there are serotype 4b isolates outside of lineage I and serotyping cannot identify the 4bV subgroup, an alternate analysis, based on phylogenetic resolution, could be considered. In a serotyping-based comparison, a lineage III or IV strain that serotyped as 4b but serogrouped via qPCR as NT would be considered accurate for the BAM method. If phylogenetic relevance was deemed paramount, then the qPCR would be deemed correct. Cohen's kappa was calculated to assess the agreement using both definitions.
In a serotype-based analysis of methods, the BAM method was 87.4% accurate and the qPCR method was 91.6% accurate; these results were not significantly different because of overlapping CIs. If a phylogenetic-based analysis was used, the BAM method fell to 75.6% accuracy, whereas the qPCR method rose to 96.6% accuracy. In this analysis, because the CIs did not overlap, the qPCR sero method was determined to be significantly more accurate than the BAM method when using a phylogenetic-based assessment. The two methods also had the lowest percentage of agreement (74%, κ = 0.68) for the phylogenetic-based analysis. In addition, when evaluating specificity and sensitivity, substantially lower sensitivity was seen for the BAM method for either assessment when evaluating the 2C and NT groups (Table 3). Conversely, the qPCR method was deemed less sensitive for identifying the 4B group when using a serotype-based assessment because of lineage III or IV 4b strains. These data, plus the inclusion of the IAC, allowed the determination that the NT group could be considered a defined molecular serogroup, as opposed to being nontypeable, capable of identifying most lineage III and IV strains with a high degree of sensitivity and accuracy (89.5%), with all errors attributed to two lineage IV strains.
Critical errors found for the qPCR molecular serogrouping assay the identification of one strain, LS95, that confounded both the BAM and the qPCR methods at the genus and species level and the identification of lineage IV isolates. These errors can be directly attributed to the method being consistently unable to resolve these strains correctly. Most other errors in the qPCR assay appeared to be linked to human error.
As a separate result of this work, we were able to determine that two strains received in our laboratory did not match the strains that were requested. Seroagglutination and qPCR results for LS573 and LS574 disagreed. During analysis of the SRA files associated with strains LS573 (SRR3372411) and LS574 (SRR3372412), we noticed a discrepancy between the WGS analysis, which indicated WGS group 1, 2B, and the qPCR analysis, which grouped them in WGS group 2, 2C. Because these could have represented strain variants that were presenting atypically, an effort was made to determine whether this resulted from human error or was strain intrinsic. To resolve this, we resequenced LS573 and LS574, and the WGS analysis confirmed that the strains belong to WGS group 2, 2C. The discrepancy could have resulted from mislabeling of the strains. Thus, the qPCR method can also provide a verification step when transferring strains from laboratory to laboratory to reduce the risk of misidentified strain attribution.
Genus and species identification
The Lall probe and primers, based on the iap gene of Listeria, were derived from a prior method present in the BAM protocol (19). This method was able to detect Lm, L. innocua, L. welshimeri, L. ivanovii, and L. seeligeri but was not evaluated for L. grayi. Evaluation of this primer set using an alignment of the iap coding regions from several representative strains found degeneracy in some non-Lm sequences that could result in a failure to detect more distantly related Listeria spp. Detection of Listeria spp. can serve as indicators that the conditions from which the sample was obtained are supportive for the potential of future Lm contamination and should be addressed to limit this risk. Because detection of a broad diversity of Listeria spp. is valuable, the primer set was modified from the sequences described in the BAM protocol, section H-2d (19), to allow degeneracy in four locations to improve detection of L. seeligeri isolates. However, L. grayi iap sequences were found to have diverged significantly from the other members of the Listeria genus. Therefore, it is unlikely that an iap-based genus detection system will detect L. grayi isolates without incorporating distinct and specific oligomers. A detailed discussion of Listeria spp. and their relationships with one another was reported by Orsi and Wiedmann (29) and suggested that L. grayi should belong to a separate genus, proposed as Murraya. No modifications, either to the sequence or to the TaqMan chemistry, were made to the probe used in the initial real-time PCR method that is described in the BAM protocol, section H-2d (19).
The hly3 probe or primers were designed to detect the hly gene, encoding listeriolysin of Lm. This gene is found in both Lm and L. ivanovii. However, the primers used in this study did not amplify the L. ivanovii hly because of sequence divergence within the target regions. None of the nine L. ivanovii strains screened yielded a positive hly qPCR result in any of the three trials, supporting the specificity of this method for only Lm. This is in direct contrast to RLM agar, which could yield erroneous results because of weak xylose fermentation of some L. ivanovii strains. Further testing resolved this issue but required another 48 h to complete. Conversely, the qPCR assay was also able to resolve this within a single working day after the initial overnight incubation for growth on suitable medium. In addition, preliminary results suggest that the qPCR method can be performed on colonies from the initial selective agar plates (no need for subculture), resulting in even faster identification.
In addition to rapidly discriminating Lm from L. ivanovii, the qPCR method was able to detect the presence of the virulence gene hly (whose product is listeriolysin O). Prior reports have noted the presence of naturally occurring prfA mutants (4, 27), whose product regulates several virulence genes, including hly and plcA. Loss of expression of these genes would result in phenotypes indistinguishable from L. innocua on blood agar, ALOA (Agar Listeria Ottavani & Agosti) agar, and RLM agar (4). Although these prfA mutants may not be pathogenic in animal and tissue culture models, the proper identification of these isolates is important, because a few Lm with prfA mutations have been isolated from humans (27). Therefore, determination of a strain as Lm by genetic methods rather than phenotype is of great importance for food safety.
Serogrouping analyses described within the BAM protocol were generally limited to the results obtained through a simple type 1/type 4 agglutination assay, because proper serotyping would require a time-consuming protocol and reagents that have limitations in accessibility. The kit for full serotyping of Lm is produced by only one company. This kit is expensive, can take several months for shipment from the company, and can be delayed because of international shipping logistics. This can be problematic for labs that need ready access to these reagents. In addition, the protocol is time-consuming, because detection of the flagellar-based H-antigen requires a week of passaging in semisolid broth cultures to maximize flagellar gene expression. The type 1/type 4 agglutination assay, which is readily available within the United States, only evaluates O-antigen profiles, resulting in a simplistic serogrouping of isolates that does not correspond to phylogenetic clustering results. Both methods rely on subjective assessment of a positive or negative agglutination reaction. In addition, the sensitivity of these types of reactions may be problematic, leading to erroneous results (34). Expression of the antigenic markers may be variable, especially for flagellar markers, resulting in inaccurate assessment. More than half of the errors occurring in this evaluation of the agglutination method were linked to incorrect flagellar determinations. This was consistent with a prior evaluation of the agglutination method in 1996, which found errors linked frequently to the H-antigen (33). Separately, type 1 O-antigen determination appeared to be somewhat challenging, because the agglutination reaction was not as consistently robust as observed for the type 4 reaction. Schönberg et al. (33) also observed difficulties in differentiating type 1/2 from type 3 strains and suggested this resulted from a commonly shared somatic antigen that required good reactivity from one of the other sera to resolve. The accuracy levels observed in this current work were in line with the level reported by Schönberg et al. (33), which identified an overall reproducibility of 83%.
A qPCR-based molecular serogrouping method would remove the ambiguity of determining a positive agglutination reaction, as well as variability in antigen expression levels and antiserum reactivity. Based on comparison with WGS phylogenetic analyses, the qPCR molecular serogrouping method was highly reliable. As noted earlier, strain LS95 also was unable to be correctly identified as Lm using the BAM protocol. This was not the only strain that gave potentially confounding results using both methods, with LS537 yielding results that suggested either mischaracterization or another atypical profile. Further evaluation of such atypical strains should be considered critical, because they may represent strains that could yield inaccurate results when evaluating samples. Determination of the basis of these differences would enable the development of better methods to identify them.
Separate from the potential variation in antigen expression and antisera reactivity, the requirement of technical expertise was also less stringent with the qPCR method, resulting in less human error. All three analyses for this portion of the evaluation were performed by undergraduate interns, unlike the BAM method that required evaluation by an experienced staff member to determine serum reactivity. Although results demonstrated the need for care when processing samples, especially in large numbers, the overall training needed was less rigorous than that needed for seroagglutination.
Beyond the accurate determination of serogroup, the qPCR method accurately identified 9 of the 10 4bV strains in the study, with one missed because of technician error. The seroagglutination assay could only identify them as serotype 4b. This was critical, because evaluation of this subtype as an emerging threat to food safety requires the availability of fast, accurate tests to facilitate the identification of these strains. The utilization of this method would enable labs to rapidly identify 4bV strains to allow analysis of the implications of these strains separate from individual outbreak events. This type of evaluation led to the identification of two large clades that appear to be expanding in food and clinical isolates (7). Determination of whether this is true for either or both clades is critical for addressing emerging threats to mitigate their impact.
Use of WGS SNP analyses to validate the qPCR method has a reciprocal implication, suggesting that the molecular serogrouping qPCR assay could be used to phylogenetically group strains into broad groups, thus streamlining subsequent WGS-based phylogenetic analyses independent of pulsed-field gel electrophoresis. Comparison against an entire database would by default require more computational resources than would be required if the analysis could be focused to a defined subset. This could potentially speed analysis downstream and is performed as a routine method elsewhere using the conventional PCR serogrouping method (17). Furthermore, the availability of this information would allow surveillance for trends in Lm contamination and disease patterns.
Although this method was not compared with the Doumith et al. (14) method, the shift from a gel electrophoresis–based analysis to a qPCR-based format is expected to reduce time and improve throughput compared with the original method. Critically, this method reduces the PCR cycling times; removes gel preparation, loading, and running; and eliminates visualization of the PCR bands. In addition to the removal of the time necessary to process these samples, it is anticipated that the lab waste generated would be reduced. Finally, inclusion of an IAC reduces the risk of PCR inhibition leading to inaccurate characterization.
In summary, we have developed a rapid, high-throughput qPCR method for verification of the Listeria genus and Lm with a high degree of accuracy (96.6%); this method can be combined with parallel analyses of serogroups, which can aid in the identification of the emerging 4bV subset, as well as help streamline subsequent WGS analyses. In addition to being rapid, with analyses capable of being completed within a few hours, the described method was highly accurate, with slightly higher accuracy (98.7%) for the G/S verification, compared with 96.8% for the BAM protocol, and significantly higher accuracy 96.6% for phylogenetic-based serogrouping, greatly exceeding the 75.6% accuracy using seroagglutination, including the ability to detect 4bV strains. Critically, reduced sensitivity could largely be attributed to the lineage III and IV strains (Table 3), and these strains are not typically associated with clinical, food, or food processing sources.
We thank Yiping He, of the U.S. Department of Agriculture's Agricultural Research Service (ARS), for the cloned IAC DNA. We also thank Travis Adkins at the ARS Culture Collection for the NRRL isolates used to expand serotype assessment. This work was funded in part by the University of Maryland Joint Institute for Food Safety and Applied Nutrition through a cooperative agreement with the U.S. Food and Drug Administration, FDU001418.
Present address: Office of Food Safety, Center for Food Safety and Applied Nutrition, College Park, MD 20740, USA.