Comparison of several Real-Time PCR Kits versus a Culture-dependent Algorithm to Identify Enteropathogens in Stool Samples

This study aims to validate the current diagnostic method for the clinical detection of gastroenteritis. We analyzed 400 stool samples to detect three of the most common enteropathogens: Salmonella spp., Campylobacter spp., and Yersinia enterocolitica. All specimens were tested with a routine clinical diagnosis algorithm and with five real-time PCR assays. A total of 98 specimens (24.5%) were positive for enteropathogens. We found 24 samples positive for Salmonella enterica, 71 positive for Campylobacter spp., and 4 positive for Yersinia enterocolitica. All evaluated methods exhibited a good performance in identifying Salmonella and Yersinia enterocolitica, being the highest positive percent agreement (PPA) value of 95.8% and 100%, respectively. The clinical algorithm showed the highest PPA value identifying Salmonella, due to the enrichment in selenite broth. However, the evaluated methods showed notable differences in the identification of Campylobacter species, obtaining a wide range of PPA values: 59.2%–100%. The clinical algorithm showed the lowest PPA value since it was only able to detect Campylobacter jejuni and Campylobacter coli species. This study revealed the importance of implementing the real-time PCR technique in a clinical algorithm: it improved the accuracy of the diagnosis and provided results in a shorter time compared to routine clinical methods.

Study design and case definition. Several comparisons were performed simultaneously. All specimens were tested with five qPCR assays. In addition, all specimens were analyzed with the routine clinical algorithm. Discrepant samples among the qPCR assays were resolved with additional qPCR tests. Discrepancies between the qPCR assays and the routine clinical algorithm were resolved with conventional PCR and sequencing.
When an incongruence could not be resolved, the information derived from the other tests was used to resolve it. The growth in culture was considered a positive result, regardless of other results.
Consequently, taking into account the above mentioned factors, we designated the following as "cases": a) Samples that were identified as positive in culture media. b) Discrepant samples, with positive results supported by resolution tests. We consider as resolution tests those that help us to resolve the incongruences. In this study they were the additional qPCR test and the conventional PCR and sequencing. c) Discrepant samples, without successful resolution tests, but a positive result was supported by at least three qPCR assays.

Routine clinical algorithm.
Once samples arrived at the Microbiology laboratory, they were routinely plated onto various selective and differential media, including selenite broth, MacConkey (MCK) agar, Hektoen enteric (HEK) agar, cefsulodin-irgasan-novobiocin (CIN) agar, xylose lysine deoxycholate (XLD) agar, and charcoal cefoperazone deoxycholate agar (CCDA). These cultures were grown to isolate and identify the enteric pathogens. All culture media were obtained from Oxoid (Thermo Fisher Scientific, Massachusetts, US). Stool samples were inoculated into enrichment selenite broth and incubated at 37 °C for 24 h prior to culturing in XLD agar. All media were incubated at 37 °C for 24 h, except for CCDA plates, which were incubated at 42 °C for 48-72 h under microaerophilic conditions, with a GasPak EZ Campy system (BD Diagnostics, New Jersey, US).
Salmonella is expected to grow on MCK, HEK and XLD agar, Campylobacter on CCDA plates and Yersinia enterocolitica on MCK and CIN agar.
Colonies recognized as presumptive pathogens were confirmed with matrix-assisted laser desorption/ ionization-time of flight mass spectrometry (MALDI-TOF MS Biotyper 3.1; Bruker, Massachusetts, US) 11  www.nature.com/scientificreports www.nature.com/scientificreports/ The targets of the three monoplex and one multiplex VIASURE assays are the invA gene (Salmonella), 16S rRNA gene (Campylobacter), and ail gene (Yersinia enterocolitica), whereas RIDA ® GENE Bacterial Stool Panel allows the simultaneous detection of ttr gene (Salmonella), 16S rRNA gene (Campylobacter), and ail gene (Yersinia enterocolitica). All Real-Time PCR assays were performed following the manufacturer's protocol. Positive and negative controls were included in each run. All runs were performed on the AriaMx thermocycler (Agilent Technologies, California, US).

Real-Time PCR resolution test. The Mericon
Campylobacter spp kit (Qiagen, Hilden, Germany) was used for resolving Campylobacter discrepant samples. This Real-Time PCR assay was performed according to the manufacturer's protocol.
Conventional PCR. Conventional PCR was performed with the VIASURE ESSENTIALS DNA Master Mix kit (CerTest Biotec S.L., Zaragoza, Spain), according to the manufacturer's instructions. The kit contained all necessary PCR reagents lyophilized in the wells, except for the primers, which were customizable, depending on the target. The primers used for detecting Salmonella (hilA gene) and Campylobacter (16S rRNA gene) were described previously 12,13 ; we used 1 µM of each primer. We modified Salmonella primers by deleting the first nucleotides; the following forward (5′-ATTTGCGCCATGCTGAGGTAG-3′) and reverse (5′-CCGCCGGCGAGATTGTG-3′) primers were used.
PCR products were separated on a 1.0% agarose gel by electrophoresis.

Statistical analysis.
We compared the tests under study in terms of the positive percent agreement (PPA) and negative percent agreement (NPA), instead of sensitivity and specificity. This approach conforms to FDA recommendations 14 , which suggest that sensitivity and specificity should only be used when the comparator method is a reference standard. In this study, we did not use the culture as the reference method. Instead, we used the "case" definition for the comparator method. Samples were designated "cases" when they were positive in the culture media, in the discrepant sample analysis, or at least in three qPCR assays. The quantification cycle (C q ) value determines the initial copy number of the target. Samples with low target concentrations display high C q values. A C q value greater than 40 was considered negative.
All statistical analyses were performed with SPSS software (version 22; IBM Corp., Madrid, Spain), except for the 95% confidence intervals (CI), which were calculated with the modified Wald method, with GraphPad QuickCalcs. The Chi-square test was performed to determine correlations between enteropathogen infections and patient groups. A P-value <0.05 was considered statistically significant.

Results
A total of 400 stool samples, acquired from 385 patients, were included in this study. Samples were grouped by stool consistency, and they were also classified according to the patient's age group, gender, and hospitalization status. These data are shown in Table 1.
Among 400 stool samples, 98 (24.5%) met the case definition. Of these, 69 were found positive in culture media, 25 were discrepant samples that were resolved, and 5 were discrepant samples that were found positive at least in three qPCR assays. Although there were 98 cases, there were 99 positive results, because one sample corresponded to a co-infection by Campylobacter and Yersinia enterocolitica. A more detailed analysis is given in Table 2.
Nine different species of Campylobacter were found. All of these were detected with the VIASURE monoplex assay (PPA = 100%, CI = 93.9-100; NPA = 100%, CI = 98.6-100); the next most accurate assays, in descending order were: the VIASURE multiplex (PPA = 95.8%, CI = 87. No false positive results were found in this evaluation; consequently, all NPA values were 100%. Moreover, no inhibition occurred in any qPCR reaction; therefore, it was not necessary to re-test any sample. Samples were grouped according to the C q value in three categories: C q <25 (high copy samples), C q = 25-35 (medium copy samples) and C q >35 (low copy samples, near the detection limit). They were also classified into two groups: positive by culture and negative by culture. The minimum and maximum C q values were calculated for each group. These data are shown in Table 4. We excluded Yersinia enterocolitica samples that were negative by culture as none of them were positive by qPCR assays. (2020) 10:4301 | https://doi.org/10.1038/s41598-020-61202-z www.nature.com/scientificreports www.nature.com/scientificreports/ We also evaluated whether some factors (age, hospitalization status, and stool sample consistency) might be positively associated with infections by these three enteropathogens. We performed a Chi-square test, but we did not find any statistically significant link between enteropathogen infections and the age group (P = 0.199), hospitalization status (P = 0.556), or stool sample consistency (P = 0.073). Hence, we can only report trends. When analyzed according to age, most patients affected seem to be the youngest; infection rates were 32.2% in pre-school children; 30.6% in infants; 23.1% in school-age children and adolescents; and 20.9% in adults. When analyzed according to hospitalization status, infection rates were higher among patients that required hospitalization (28.0%), compared to those attended in outpatient clinics (25.1%) and in the emergency department (18.2%). Finally, when analyzed according to sample consistency, infections were found most frequently in stools with mucous (43.8%), liquid (38.2%), and crumbly (35.7%) consistencies and in stools with traces of blood (28.6%).

Discussion
This study aimed to validate the current routine clinical diagnosis method for identifying the three main bacterial enteropathogens: Salmonella, Campylobacter, and Yersinia enterocolitica. We compared the standard culture-based techniques to three types of qPCR assays (VIASURE monoplex, VIASURE multiplex, and R-Biopharm).
In recent years, a large number of studies have been conducted to validate novel gastrointestinal molecular panels as BD Max enteric bacterial panel 4 , ProGastro SSCS assay 7 , Allplex GI-Bacteria assay 8 or FilmArray GI Panel 5,15 . These previous studies emphasized the importance of not considering the culture technique as the reference method, due to its lack of sensitivity. Consequently, we created a reference method based on different techniques. According to this reference, samples were classified as "cases" or "non-cases" (see Materials and Methods section).
All methods showed acceptable PPA values (87.5%-95.8%) for identifying Salmonella spp. The highest PPA value was obtained with the routine clinical algorithm. This could be explained by the fact that stool samples were inoculated into a selenite enrichment broth, as reported in previous studies 16 . This broth contributed to bacterial replication, which enhanced the culture method. The culture method was able to detect samples that showed C q values extremely near the detection limit (C q > 35).The maximum C q values obtained in Salmonella identification were 39.7, 35.8, and 39.6 with VIASURE monoplex, VIASURE multiplex, and R-Biopharm assays, respectively (Table 4). This observation revealed the high sensitivity that could be achieved with the culture method, due to the enrichment in selenite broth. We performed qPCR assays to identify Salmonella in two selenite broths whose stool sample results were negative with qPCR. As we suspected, the Salmonella DNA could be detected in the selenite broth with qPCR assays but could not be detected in the stool sample. The benefits of the Real-Time PCR in combination with the enrichment in selenite broth have been recently reported. Boer et al. 17 found that the number of Salmonella positive samples increased by 2.2% when qPCR was performed after enrichment in selenite broth. Tang et al. 18 have also validated the high sensitivity and specificity of a national standard protocol based on Real-Time PCR combined with guided culture. www.nature.com/scientificreports www.nature.com/scientificreports/ Relative to Campylobacter, the qPCR assays exhibited notable differences in their abilities to detect many Campylobacter species. The highlight was the high capacities of the VIASURE monoplex and multiplex assays (PPA = 100% and PPA = 95.8%, respectively) in diagnosing the majority of Campylobacter species that are considered pathogenic in humans 19 .
The R-Biopharm test displayed less diagnostic capacity (PPA = 81.7%). It seems to be due to a failure in the recognition of less common Campylobacter species (other than C.jejuni and C.coli) rather than in the sensitivity of the test, since most of the samples were in high concentration (C q < 25). The R-Biopharm assay could not detect any positive sample for C. upsaliensis, C. hyointestinalis, C. helveticus, or C. rectus (Table 2).  Table 2. Comparison of different molecular assays and the routine clinical method for enteropathogen identification. Abbreviations: +, positive detection; −, negative detection; S., Salmonella; C., Campylobacter; N/A, not applicable. b Conventional diagnostic assays included the culture method, MALDI-TOF mass spectrometry, agglutination assays, and biochemical tests. c A discrepancy analysis was not applicable, because these samples were positive in culture, which immediately defined them as a "case". d Conventional PCR was unsuccessful; therefore, the Salmonella serotype was not resolved in this sample; however, it was considered a "case", based on three diagnostic methods, which supported a positive result. e Sequencing was not successful in 4/16 samples, but the Mericon assay (the qPCR resolution test) performed with these four samples provided a positive result; thus, all were considered "cases". f All of these samples were positive in the Mericon assay, and the positive result was confirmed with sequencing in 9/10 samples. g This sample was positive in the Mericon assay. h These two samples were positive in the Mericon assay, and the positive result was confirmed by sequencing in both samples.  Table 3. Campylobacter species identification by 16S rRNA sequencing.
The routine clinical algorithm was only capable of detecting C. jejuni and C. coli species (PPA = 59.2%) ( Table 2). This finding was consistent with the literature, where different studies [20][21][22] indicated that some antimicrobial agents incorporated into the commonly used selective media (like cefoperazone in CCDA media) could inhibit the growth of several species, including C. hyointestinalis, C. upsaliensis, and C. fetus. Moreover, not all Campylobacter species are thermophilic; consequently, the incubation at 42 °C was not suitable for all species. Additionally, some species, like C. concisus, C. rectus, C. curvus, C. gracilis, and C. showae require a special atmosphere enriched with hydrogen.
The culture technique failed to detect 3 C. jejuni and 3 C. coli samples ( Table 2). None of them were low copy samples since 4/6 showed values of C q < 25 and 2/6 showed a C q value = 25-35. This observation revealed that sometimes even the growth of C. jejuni and C. coli in selective media could be complicated.
For the above reasons, the qPCR technique appears to be a reliable solution for Campylobacter identification since it could ensure the identification of most Campylobacter species without excessive complications in the diagnostic algorithm.
Regarding Yersinia enterocolitica, all the methods studied could detect all samples positive for Yersinia enterocolitica O:3 (PPA = 100%).
An important aspect of this study was that no false positive result was found in any qPCR assay. This high specificity (100%) could represent a key difference between molecular-based and antigen-based CIDTs. Couturier 10 stated that the problem with stool antigen-based CIDTs was that they exhibited highly variable sensitivities and specificities. That problem has not occurred with molecular-based CIDTs, which, to date, have shown excellent sensitivity and specificity compared to cultures. Our findings correlate with those previous results; therefore, molecular CIDTs could be considered a reliable diagnostic tool, although more studies are needed. Additionally, some molecular CIDTs, like VIASURE assays, are very easy to use. All the components are provided lyophilized inside the wells, which implies that the user is not required to mix any reagents. Also, the lyophilized format facilitates the storage and shipping conditions as lyophilized reagents are stable at room temperature.
We did not encounter inhibition in any of the qPCR assays, although we tested a vast range of stool consistencies. This finding suggested that the qPCR assays used in this study are robust tests with a stable matrix formulation, which neutralized PCR inhibitor effects.
The main limitation of this study was that the number of positive samples depended on the prevalence of each pathogen in the study area. Consequently, we recommend that a multicenter study should be performed, and that the sample collection time should be extended.
We conclude that the three types of qPCR assays (VIASURE monoplex, VIASURE multiplex and R-Biopharm) tested could improve the routine clinical diagnosis algorithm, because they had a noticeable impact on the Campylobacter diagnosis. Additionally, they exhibited good performance in Salmonella and Yersinia enterocolitica identification too, and they substantially reduced the turnaround time (from days to hours). The ease of using molecular methods and the time they save will allow laboratories to increase the number of samples processed. Moreover, saving time is critical for avoiding the spread of infections and for providing effective treatment plans.