PCR-based detection and serovar identification of Salmonella in retail meat collected from wet markets in Metro Manila, Philippines

This study aimed to detect Salmonella from retail meat collected from nine wet markets in Metro Manila, and identify the subtypes of Salmonella isolates using molecular serotyping assays from previously developed primers. Of the 720 collected meat samples, 57.64% were found to be Salmonella-contaminated. The most predominant serogroup was Salmonella O:3, and Salmonella serogroups O:4, O:6,7, O:8, O:9, and undetermined serogroups were also found. Most frequently detected isolates in bovine meat were S. 3:e,h:1,6 (putative identity: S. Anatum) and S: 4:e,h:1,2 (putative identity: S. Saintpaul), in porcine meat was S. 3:e,h:1,6 (putative identity: S. Anatum), and S. 8:i:z6 (putative identity: S. Kentucky) was common in poultry products. This study also demonstrated retail meat samples were contaminated with multiple Salmonella serogroups and serovars. This is the first Philippine study that utilized PCR-based assays to characterize Salmonella isolates down to a serovar level and provides baseline information regarding Salmonella prevalence and serovar distribution in retail meat. Molecular serotyping performed in this study can be used as an alternative approach to traditional serotyping in surveillance of Salmonella in the Philippines since the latter is expensive, time-consuming, and requires skilled technicians.


Introduction
Salmonella is considered one of the most significant pathogens associated with food-borne diseases and outbreaks in the world [1,2]. It is estimated to cause approximately 93.8 million human infections and 155,000 deaths annually worldwide [3] and major economic losses in poultry and livestock industries [4]. The genus Salmonella is classified into serovars, which are based on the presence of somatic (O) and flagellar (H) antigens. Serotyping has been essential in public health monitoring of Salmonella infection for more than 80 years [5]. White-Kauffman-Le Minor (WKL) scheme contains the organized list of 2,659 serovars, which were characterized by serological agglutination of 46 somatic and 119 flagellar antigens [6,7]. However, serological agglutination reaction is limited by antisera production, time-consuming, a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 extremely complex procedure, high cost, and at times inconsistent results due to the presence of capsule phenotype, rough phenotype, or non-expression of flagellar phase [8]. These limitations led to the development of molecular-based serotyping using multiplex PCR, which allows for the simultaneous, rapid, sensitive, and specific detection of various Salmonella genes that express somatic [9] and flagellar antigens [10,11]. Several of these primer pairs were tested for specificity and sensitivity by the National Institute for Public Health and the Environment reference laboratory in The Netherlands (RIVM) as an alternative approach to gold standard [12].
In the Philippines, surveillance data regarding Salmonella serogroup and serotype distribution are limited. Salmonella serotyping is not routinely performed due to previously mentioned disadvantages. Most Salmonella isolates from clinical and environmental settings collected in different parts of the country are submitted to the Antimicrobial Resistance Surveillance Reference Laboratory (ARSRL) in the Research Institute for Tropical Medicine (RITM) for serotyping. Molecular subtyping of Salmonella was previously performed in studies by Soguilon & Rivera, Ng & Rivera, and Calayag et al. [13][14][15] in slaughtered swine and meat samples collected from Metro Manila, Philippines. However, these studies were limited to serogrouping and detection of Sdf-I expressing gene. Hence, a study that determines Salmonella serotype using molecular subtyping is needed.
This study aimed to detect Salmonella spp. from various raw meat and meat products collected from nine selected Metro Manila wet markets using PCR-based detection assays described by Ng & Rivera [14] and Soguilon & Rivera [13], and classify Salmonella based on their O-serogroups and serotypes using PCR-based serotyping assays. This study can contribute to the epidemiology of salmonellosis in the Philippines, particularly on reservoirs of clinically important serotypes. Data regarding the Salmonella prevalence and serotype distribution in retail meat are significant in the identification of important Salmonella reservoirs and initiation of long-term prevention and control measures.

Sample collection
Two hundred seventy-four (274) raw meat (minced beef = 90, chicken wings = 94, ground pork = 90) and 446 processed meat (bacon = 90, burger patties = 70, Filipino-style luncheon meat = 90, sliced ham = 85, Filipino-style sausage = 96, salami = 15) samples were collected from nine wet markets in Metro Manila from 2015-2016. A total of 80 samples were purchased randomly at different stalls in each wet market with varying frequencies depending on market availability of meat products. Raw meat samples were defined as animal muscle and fat that were chopped, minced, or ground, while processed meat samples were defined as meat samples which underwent salting, curing, addition of seasonings and other food materials, and/or heat treatment. Samples were stored in a cooler at approximately 5˚C during transport and immediately processed in the laboratory for the pre-enrichment procedure. All raw meat products were freshly slaughtered from a local city abattoir, while processed meat products were either prepared from wet markets or local meat processing plants. Each market location was given a precursory survey of the conditions, operation facilities, and general practices of the workers at the time of collection. Quality of the market sanitation and practices were ranked on an ordinal scale of five values ranging from poor to very good (S1 Appendix).

Revival and purification
From Salmonella glycerol stocks, 50 μl stock culture was inoculated to 1 ml TSB and incubated at 37˚C for 24 h. A loopful of cultured Salmonella in TSB was streaked onto XLD agar plate. After incubation, a single typical Salmonella colony isolate was selected for inoculation into two microcentrifuge tubes with 1 ml TSB each. Isolation steps using XLD agar plates were repeated until a single colony of presumptive Salmonella was obtained. One of the TSB tubes was subjected to DNA extraction, while the other was stored as a glycerol stock.

DNA extraction
Three different template extraction protocols (TEP) were performed in this study, and DNA templates were extracted in the following Salmonella culture steps: (TEP I) after 18-24 h incubation of pre-enriched samples in RV broth, (TEP II) after 18-24 h incubation of putative Salmonella isolates inoculated in NA slants, and (TEP III) after revival and purification of Salmonella isolates from glycerol stocks. TEP I was intended for the rapid identification of Salmonella within samples. Data regarding Salmonella prevalence were gathered from TEP I and TEP II, while results of TEP III verified that the revived and purified isolates were Salmonella and was used as templates to characterize isolate serotypes. Two milliliters of culture from RV broth (TEP I) and 1 ml of Salmonella culture in TSB (TEP III) were centrifuged at 15,330 ×g for 5 min [13]. The resulting supernatant was discarded while the pooled bacterial pellet was suspended in 1 ml phosphate-buffered saline (PBS). The bacterial suspension was vortexed for 1 min and centrifuged at 15,330 ×g for 5 min. The resulting pellet was suspended in 50 μl and 100 μl sterile Tris EDTA (TE) buffer for RV broth and TSB, respectively. For TEP II, a loopful of Salmonella was collected from the nutrient agar slant and transferred to a microcentrifuge tube containing 100 μl sterile Tris EDTA (TE). Bacterial suspensions were boiled at 100˚C for 10 min using a dry heating block. Resulting lysed cells were centrifuged at 2,656 ×g for 5 min. The supernatants were transferred to a new microcentrifuge tube and stored at -20˚C until use.

Molecular detection and characterization
Four-step PCR-based assays using T100 Thermal Cyclers (Bio-Rad, USA) were employed in molecular detection and serotyping of Salmonella from meat samples.
Step 1 involved the detection of Salmonella spp. using PCR that targeted the invA gene in DNA extracted from TEP I and II. In step 2, Salmonella positive samples were run as multiplex PCR experiments using DNA extracted from TEP I and TEP III.
Step 3 determined the presence of phase 1 flagellar antigen (H1) encoding genes while step 4 determined the presence of phase 2 flagellar (H2) antigen and Sdf-I encoding genes. The results of the last three PCR assays defined the banding pattern of each Salmonella isolate, which would correspond to a serovar identity. Steps 3 and 4 analyzed DNA extracted from TEP III.
Polymerase Chain Reaction (PCR)-based detection of Salmonella spp. The amplification of the invA gene was performed in a 12.5 μl reaction volume. Each reaction consisted of 6.25 μl 2× Promega Green master mix (1 U i-Taq TM DNA polymerase, 2× PCR buffer, 3 mM MgCl 2 and 0.4 mM dNTPs; Promega, USA), 0.5 μl each of forward and reverse primer (10 uM; Table 1, Reaction group 1), 1 μl DNA template and 4.25 μl of PCR water. PCR was performed using the following conditions: initial denaturation at 95˚C for 3 min; denaturation at 95˚C for 30 s, annealing at 60˚C for 30 sec, extension at 72˚C for 30 sec, repeated for 35 times, and a final extension at 72˚C for 5 min.
O-serogrouping of Salmonella spp. The O-serogrouping assay was performed in a 12.5 μl reaction. Each reaction was composed of 6.25 μl 2× KAPA2G Fast multiplex mix (1 U i-Taq TM DNA polymerase, 1.5× PCR buffer, 3 mM MgCl 2 , and 0.2 mM dNTPs; Kapa Biosystems, USA), 0.25 μl each of 7 forward and reverse primer (10 uM; Table 1, Reaction group 2), 1 μl DNA template and 1.75 μl of PCR water. Multiplex PCR was accomplished using the following conditions: initial denaturation at 95˚C for 3 min; denaturation at 95˚C for 30 sec, annealing at 58˚C for 30 sec and extension at 72˚C for 1 min run in 35 cycles, and a final extension at 72˚C for 5 min.
H-typing I assay. H-typing I assay was performed in a 12.5 μl reaction, wherein each reaction was composed of 6.25 μl 2× KAPA2G Fast multiplex mix (Kapa Biosystems, USA), 3 forward and 7 reverse primers (10 uM; Table 1, Reaction group 3), 1 μl DNA template and 2 μl of PCR water. Multiplex PCR was run using the following conditions: initial denaturation at 95˚C for 3 min; denaturation at 95˚C for 40 sec, annealing at 58˚C for 30 sec and extension at 72˚C for 30 sec for 30 cycles, and a final extension at 72˚C for 7 min.
H-typing II assay. Samples were further characterized in H-typing II assays and primer pairs shown in Table 1 were used for these experiments. H-typing II assay was performed in a 12.5 μl reaction, which was composed of 6.25 μl 2× KAPA2G Fast multiplex mix (Kapa Biosystems, USA), 3 forward and 6 reverse primers (10 uM; Table 1, Reaction group 4), 1 μl DNA template and 0.75 μl of PCR water. Multiplex PCR was run using the following conditions: initial denaturation at 95˚C for 3 min; denaturation at 95˚C for 40 sec, annealing at 58˚C for 20 sec and extension at 72˚C for 20 sec for 30 cycles, and a final extension at 72˚C for 7 min.
Serogroup O:9 was further characterized by a primer pair that detects the Sdf-I gene ( Table 1, Reaction group 5), while samples with a missing band that determined their phase 2-flagellin were further characterized with z6 primers (Table 1, Reaction group 6). PCR reaction mix and conditions for amplification were similar to step 1.

Traditional serotyping
Serological agglutination assays of 16 selected Salmonella isolates with pre-determined antigenic formulae derived from molecular serotyping were outsourced to Antimicrobial Resistance Surveillance Reference Laboratory (ARSRL) at the Research Institute for Tropical Medicine (Metro Manila, Philippines). Table 1. Primers for Salmonella identification and characterization. List of primers for Salmonella detection, serogrouping, and serotyping assays.

Data analyses
Fischer's Exact Test was used to evaluate differences between the number of Salmonella-contaminated samples in raw and processed meat, as well as comparisons of Salmonella detection frequencies for culture versus PCR methods and different TEPs. Statistical tests were performed using IBM SPSS Statistics Version 20 and R 4.0.

Results and discussion
In this study, we investigated the Salmonella prevalence in raw and processed meat samples collected from nine wet markets in Metro Manila, Philippines from 2015-2016 using PCRbased assay. We also characterized the serogroups and serotypes of Salmonella isolates using molecular serotyping assays developed by previous published studies [8,10,11,19]. These aims were achieved by performing three template extraction protocols (TEPs). Using TEP I, Salmonella detection in meat samples was achieved within 2 days, while traditional culture techniques requires 3-4 days to obtain negative results and 5-7 days to obtain confirmed positive results [20]. Putative Salmonella colonies from selective growth media were confirmed by extracting the DNA of these colonies grown in NA slants (TEP II). Combined results of PCR targeting Salmonella using DNA extracted using TEP I and II determined the prevalence of the pathogen in collected meat samples. Multiple Salmonella serogroups were detected in DNA extracted using TEPs I and II, thus, further isolation and purification steps were applied to revived Salmonella isolates. Salmonella serogroup frequencies were compared between DNA extracted from TEP I and III. Molecular serogrouping and serotyping assays were performed to DNA of Salmonella isolates extracted using TEP III.

Presence and distribution of Salmonella in retail meat
A wet market is a type of traditional food retailer in the Philippines that offers affordable grains, fresh produce such as meat, fish, fruits, and vegetables [21]. However, this type of market has been linked to major diseases due to poor hygienic conditions [22]. Various studies have detected wet markets as an important source of Salmonella, where prevalence ranged from 6.8-87.5% in various retail meat [13,[22][23][24][25][26][27][28][29][30]. This study found that 415 (57.64%) of 720 purchased raw and processed meat products were contaminated with Salmonella using PCRbased Salmonella detection (Table 2). Extraction protocols TEP I (n = 392) and TEP II (n = 306) yielded different frequencies of Salmonella positive meat samples. Despite utilizing different DNA extraction protocols, there were no statistically significant differences in the detection of Salmonella between TEP I and TEP II (p = 0.162) in both processed and raw meat samples. Also, combining the number of Salmonella positive samples from TEP I and II (overall) had resulted in relatively higher Salmonella positive results (n = 415), but the overall Salmonella prevalence in retail meat was not significantly different when compared to the results of TEP I (p = 0.8314) and TEP II (0.1075). TEP I detected relatively higher Salmonella positives samples than TEP II, however, TEP I also missed 23 positive samples that were confirmed by TEP II. This suggests that utilizing both TEP I and TEP II was important to have a relatively higher sensitivity in the detection of Salmonella. Additionally, comparisons for culture method detection of Salmonella versus TEP I (p = 0.453) and TEP II (p = 0.519) in both raw and processed meat samples also had no significant differences. This Salmonella prevalence in this study was relatively higher compared to previous studies that detected Salmonella in the Philippines, such as Soguilon and Rivera (30.63%, n = 320) [ [26], and Nepal (11.4%, n = 123) [33]. Although this report was relatively lower compared to studies done in Thailand (72%, n = 100) [34], Laos (87.75%, n = 49) [35], and different reports in Malaysia (64.6% n = 82 and 87.5% n = 72) [26,36]. Prevalence of Salmonella contamination among different countries might be affected by differences in sampling procedures, sample types, and Salmonella detection and isolation techniques [37]. Ground pork (87.78%) was found to have the highest prevalence of Salmonella-contamination, followed by minced beef (86.67%), chicken wings (82.98%), bacon (62.22%), Filipinostyle sausage (54.17%), burger (37.14%), Filipino-style luncheon meat (25.88%), ham (25.56%), and salami (6.67%) ( Table 2). Significantly higher number of Salmonella-contaminated samples were observed in raw meat (p = 0.00001) (85.77%, n = 235) compared to processed meat samples (40.36%, n = 180). High prevalence of Salmonella contamination in ground pork can be associated with high frequency (45%, n = 240) of Salmonella found in slaughtered swine from different Metro Manila abattoirs [14,15]. Data regarding Salmonella frequencies in chicken and cows from slaughterhouses and farms are currently lacking. Significantly higher Salmonella contamination rates in raw meat compared to processed meat can be attributed to the absence of interventions which could regulate microbial growth [38]. Also, manual meat handling such as grinding may contribute to an increase of S. enterica counts in raw meat [39]. Cross-contamination of microbes in raw meat is highly probable since they were often presented and sold without packaging. Even in the presence of additives and preservatives, bacon and Filipino style sausage were observed to have a relatively higher percentage of Salmonella contamination (>50%) compared to other processed meat products (<38%). We observed that these meat products were processed directly in wet markets, and without heat treatment. Salmonella contamination in Filipino-style sausage may also come from connective and muscle tissues of the animal's gut which serve as their natural casing [38]. Ham, Filipino-style luncheon meat, and salami had undergone heat and chemical treatments [40][41][42][43][44] and were packaged with either plastic or aluminum foil, hence, a lower rate of Salmonella contamination in these products was expected. But on several occasions, we observed that Table 2. Salmonella-positive retail meat samples from different animal origins and meat types. The number and percentage of samples found positive with Salmonella based on the food matrix using putative culture positives, TEPs I-III and overall positives from TEPs I and II. Short description regarding the animal origins, treatment, and packaging of meat sample weres also included.

Meat Samples
Meat type Animalorigin

PLOS ONE
some vendors opened packaged meat products and sold them separately. The presence of Salmonella in ham, Filipino-style luncheon meat, and salami might also be associated with failure of treatment or cross-contamination during the manufacturing and retailing stages [45][46][47], however, a more elaborate study must be performed to prove this hypothesis. We found considerable variation (5-76.25%) of Salmonella-contamination rate among different wet markets (Table 3), even though the handling procedures of retail meat products and environmental conditions were almost similar in all visited wet markets (S1 Appendix). Specifically, practices of meat being displayed by either hanging or lying on the counter without any covering for several hours, exposure to ambient temperature, excess moisture, and handled using bare hands made the retail meat highly vulnerable to microbial population growth and further microbial cross-contamination. Remarkably, we only found four (5%) Salmonella contaminated meat products (Filipino style sausages) at Quezon City District IV wet market, even in the presence of poultry evisceration activities near the meat retail area. We also noticed that the meat display counters in Makati and Paranaque wet markets were relatively drier and cleaner compared to other wet markets, and their meat counters were made up of metal or smooth concrete finishing, still, Salmonella-contamination rates in these markets were high (71.25% and 66.25%, respectively). Further investigation of other hygienic parameters in the wet market, such as water, equipment, vendors, and potentially other animal vectors must be performed to determine the important factors that affect Salmonella prevalence in wet markets [48]. Initial Salmonella contamination of animals from farms, slaughterhouses, and distribution may also have affected the different rates of Salmonella-contamination in wet markets. In previous Salmonella prevalence studies in abattoirs, varying Salmonella frequencies in slaughtered swine were observed in different slaughterhouses in Metro Manila [14,15]. We strongly suggest implementing strict hazard analysis and critical control point (HACCP) practices in slaughterhouses and transport. Hygienic practices and proper product segregation must be ensured in wet markets to prevent contamination of Salmonella to other fresh produce such as fruits and vegetables. Meat consumers must cook their meat thoroughly before consumption to kill microbial pathogens. Also, consumers must be cautious in handling and preparing meat to prevent cross-contamination in the kitchen and spread the pathogens to ready-to-eat food.

Salmonella serogroup and serovar distribution in retail meat
Salmonella-positive DNA extracted using TEP I (n = 392 meat samples) and TEP III (n = 641 Salmonella isolates from 274 meat samples) were characterized using the molecular serogrouping assay. Most meat samples were contaminated with more than one Salmonella serogroups Table 3. Salmonella-positive retail meat samples from nine wet markets in Metro Manila. The number and percentage of samples found positive with Salmonella based on the wet market using putative culture positives, results of TEPs I-III and overall positives using results from TEP I and II.

PLOS ONE
Molecular detection of Salmonella in Metro Manila, Philippines using DNA from TEP I (Fig 1A; 73.47%). Since multiple putative Salmonella colonies (both from XLD and BG agars, if present) were isolated per samples, contamination with more than one serogroup per sample was also observed but to a lesser quantity (Fig 1B; 31.75%). These results were expected since DNA from TEP I was extracted from Salmonella selective enrichment broth and will likely contain mixed groups of Salmonella, while DNA from TEP III was isolated and purified from a single colony. Multiple Salmonella serogroup contaminations were also observed in slaughtered pigs from Metro Manila [14]. Detection of the same Salmonella serogroups from multiple isolates was also observed in several samples, hence, frequencies of Salmonella serogroup from TEP III (isolates) were 1.5-1.8 times higher than TEP III (samples) (Fig 1C). The Salmonella serogroup data produced from TEP I provide a more comprehensive overview of serogroup distribution in retail meat and the presence of mixed Salmonella serotype contamination. A possible underestimation of samples containing multiple serogroup contamination may have occurred since the primer pairs used in this study was limited to the detection of six Salmonella serogroups. Salmonella serogroup O:3 was the most frequently detected serogroup in meat samples, followed by serogroup O:4 using both TEP I and TEP III. Frequencies of serogroup O:6,7 were relatively higher than O:8 using TEP I, but their frequencies were almost the same using TEP III (per sample and per isolate). The number of Salmonella serogroup O:9 and undetermined serogroups was almost similar in TEP I, but relatively higher frequencies of undetermined serogroup compared to serogroup O:9 were found in TEP III (per samples and per isolate). Salmonella serogroup O:2 was not detected in any meat sample. There was a portion of samples that did not exhibit amplification band on the internal amplification control in O-serogrouping assay likely due to either poor DNA recovery or PCR inhibitors. Previous Salmonella

PLOS ONE
studies in Metro Manila demonstrated that Salmonella serogroup O:3 was the most dominant serogroup detected in raw and processed meat (78.57%) from wet markets [13], and slaughtered swine (75%) from slaughterhouses [15]. However, Ng & Rivera [14] found that serogroup O:4 (73%) was the most frequently isolated serogroup in slaughtered swine in abattoirs followed by serogroup O:3 (24%). Reports of Calayag et al. and Soguilon et al. accounted for the distribution of Salmonella serogroups per sample, while Ng and Rivera reported frequencies of Salmonella serogroups per isolate.
We detected 29 Salmonella band patterns that correspond to putative Salmonella serotypes (Table 4) Anatum, S. Newport, and S. Meleagridis, respectively. The serovar identities of the remaining six band patterns need further confirmation since the primers that target the missing somatic or flagellar antigens were not included in this study. The putative serovar identities of 6,7:g:-, 3:g:-, -:b:enx, 4:g:1,2, and 3:-:1,6, were designated as S. Rissen, S. Senftenberg, S. Hvitingfoss, S. Agona, and S. London, respectively, since they had been previously reported in the Philippines [49][50][51]. Further analyses of these band patterns together with S. 4:g:-must be performed to confirm serovar identity. Isolates with multiple band patterns and/or unknown serovar identities were observed in more than 100 meat samples. Further purification of the samples must be done and more primers that target different somatic and flagellar antigens must be tested to unravel the identities of these Salmonella isolates. This is the first study to report Salmonella  (Table 4), only one Salmonella isolate with band pattern S. 6,7:r:1,2 (S. Virchow) did not match with the serotype identity from serological testing (S. Gabon; band pattern: 6,7:l,w:1,2). This disagreement may have arisen from the primer pair targeting fliC-r gene. It was reported that this primer pair had a 99.5% specificity [12]. Hence, it is suggested to design a more specific primer pair that targets fliC-r gene. It is also possible that the gene that expresses the flagellar phase I (l,w) antigen was present in the isolate, but not detected due to the limited number of primers pairs used in the study. As observed in the serogroup data, contamination with multiple Salmonella serovars per sample was also observed. In one chicken sample, four Salmonella isolates were identified as S. Schwarzengrund, S. Livingstone, S. Rissen, and S. Anatum, while three Salmonella isolates from a burger sample were identified as S. Augustenborg, S. Kentucky and S. Typhimurium. More evidence of multiple Salmonella serovar contamination can be discovered when more primers that targeted other somatic and flagellar antigens are used in the future.
The most frequently detected serovar in chicken was S. Kentucky, while S. Anatum was the most frequently detected in meat originating from pigs and cows. In addition to S. Anatum, S. Saintpaul had the same detection frequency in minced beef and burger. In the Philippines, Salmonella Kentucky was isolated from cloacal swabs of healthy layer chickens in San Jose, Batangas in 2011 [51]. In other countries, S. Kentucky was the most commonly reported serovar for poultry in Canada and the US, ground chicken and broilers in the US, and broiler meat in 11 Table 4. Frequencies of Salmonella serovars in retail meat. Frequencies of Salmonella band patterns and their putative Salmonella serovar identities in retail meat that originated from cow, chicken, and pork. Results of serological serotyping of 16 Salmonella isolates from ARSRL were also included. Some band patterns have more than one possible identity.

PLOS ONE
European countries in 2007 [3]. Salmonella Kentucky was also the most commonly isolated serovar in retail chicken meat in Vietnam [29], and Trinidad [52]. Salmonella Anatum was previously reported in mesenteric lymph nodes of asymptomatic cattle in Laguna and Batangas [53], and bile specimens of healthy cattle [54] in the Philippines. In other countries, S. Anatum was included in ten most frequently isolated serovars in cattle in the United Kingdom, US and Tunisia, in bovine meat in Canada, ground beef in the US, included in the seven most common serovars isolated in pigs and livestock in the United Kingdom in 2007, and most frequently isolated serovar in pork within Thailand for 2003 [55]. This serovar was also the most prevalent in cattle and swine from Vietnam in 2004 [56]. Salmonella Saintpaul was previously detected in asymptomatic cattle from Laguna and Batangas, Philippines [53]. And frequently isolated in apparently healthy cattle in Australasia (5.4%) and Africa (2.1%) [57]. Other Salmonella serovars such as S. Newport, S. Stanley, S. Derby, S. Weltevreden, and S. Enteritidis were previously reported in animal and retail meat [15,31,53,54], while S. Senftenberg and S. Hvittingfoss were reported in peanut butter products in the Philippines. These detected serovars can pose risks to retail meat vendors and consumers through unhygienic handling, cross-contamination with ready-to-eat food such as fresh produce, and consumption of poorly cooked meat. All previously reported serovars in the Philippines had been isolated in human clinical cases in the country. Although S. Anatum, S. Kentucky, and S. Saintpaul were the most frequently detected serovars in retail meat, reports regarding their incidence rates among submitted clinical isolates from 2004-2018 were very low. The incidence rate of S. Anatum was 0.54% (n = 13) in clinical isolates in the Philippines from 2004-2018 [58], while the incidence rates of S. Kentucky and S. Saintpaul were both 0.26% (n = 2) from 2014-2018 [59,[60][61][62][63]. These serovars may possess low virulence and induce moderate symptoms in humans, wherein patients no longer seek medical care. Hence, clinical cases caused by these serovars may be underreported. It was estimated that the true incidence of salmonellosis is at least 29-fold greater than the number of reported cases [64]. The most commonly reported non-typhoidal Salmonella serovars in human clinical cases in the Philippines are S. Enteritidis and S. Typhimurium from 2003-2018 [49,50,59,65], however, they ranked 11 th and 10 th frequently isolated Salmonella serovars in this study. Low prevalence rates of these serovars in meat samples suggest that their mode of transmission may come from other significant reservoirs, which have not yet been fully studied. It is recommended to perform more screening and serotype distribution studies in live and slaughtered animals, meat and other food sources (eggs, fresh produce, milk, etc.), and environmental samples in different parts of the country to obtain a more comprehensive picture of the extent of atypical Salmonella serovars associated with food contamination. Another hypothesis is that these serovars are more virulent to humans compared to other predominant Salmonella serovars, which resulted in infections that required medical attention to patients and were followed up by strain isolation and incidence reports. In contrast to this findings, most commonly reported serovars in human clinical cases in the United States (S. Enteritidis, S. Typhimurium, S. Newport, S. Javiana, S. I 4, [5],12:I:-, and S. Muenchen) were also reported as commonly isolated Salmonella serotypes (except for S. Javiana and S. Muenchen) in meat and meat-producing animals in the US [64].
Published reports of Salmonella serogroup and serovar distribution in animal, food, and environmental samples are limited and outdated in the Philippines. Studies that characterized Salmonella up to the serovar level were performed mostly by researchers from different academic institutions more than a decade ago [31,54,55]. Moreover, the determination of Salmonella serovar distribution in the food production chain is often neglected. Reports from the Philippine Department of Agriculture (DA)-mandated laboratories regarding Salmonella were focused on faster detection of the pathogen in animal, food, and environmental samples, and determination of their antimicrobial susceptibility profiles (personal communication with the National Meat Inspection Service, Philippines). Thus, this study offers a more recent overview of prevailing Salmonella serogroups and serovars present in retail meat that originates from cattle, poultry, and swine. Surveillance of Salmonella and its serovars in different reservoirs is important in source tracking and attribution, mitigation, and control of Salmonella-related outbreaks. In Denmark, Salmonella surveillance data were used in their targeted national control program incorporating HACCP, a systemic preventive approach to food safety, wherein key steps in the food to fork chain were identified for interventions that have an impact on reduction (or elimination) of food safety hazards. This control program caused a significant reduction in human salmonellosis in Denmark. The prevalence of salmonellosis in the broiler, pork, and eggs was also reduced by 95%, 85%, and 75%, respectively [66].

Conclusion
The molecular identification and characterization of Salmonella provide baseline information regarding its prevalence and serovar distribution in food samples originated from chicken, pigs, and cows. The PCR-based assay that targeted invA gene revealed that 415 (57.64%) retail meat samples were contaminated with Salmonella, while molecular serotyping showed that most retail meats were contaminated with multiple Salmonella serovars. We have demonstrated a faster means of Salmonella detection in food using the TEP I method, which confirmed the pathogen within 2 days compared to the traditional culture technique that requires 5-7 days, and higher sensitivity of Salmonella detection when the protocols TEP I and II were combined. Classical serotyping of Salmonella in animal and environmental samples are not routinely performed in the Philippines since the technique is expensive, time-consuming, and requires skilled technicians. Thus, there is a need for the development or adaptation of rapid, specific, sensitive, less costly, and high throughput techniques. This study adapted previously developed primers to characterize the serogroups and serovars of Salmonella isolates from retail meat using multiplex PCR. We detected five serogroups and at least 29 band patterns that lead to putative Salmonella serovar identities. We also found evidence of multiple Salmonella serogroup and serovar contaminations per sample using TEP I and TEP III. It is highly recommended to adapt or design more primers that target somatic and flagellar genes to determine the identities of unknown isolates. This molecular Salmonella serotyping assay may be adapted by laboratories from health and agricultural departments, particularly for developing countries, as an alternative approach to classical serotyping techniques to increase the Salmonella surveillance data in these regions.
This surveillance data is important in the development of Salmonella control strategy, intervention, and prevention schemes. We determined the most predominant Salmonella serovars in retail meat originating from were S. 3:e,h:1,6 (putative identity: S. Anatum) from pig, S. 3:e, h:1,6 (putative identity: S. Anatum), and S. 4:e,h:1,2 (putative identity: S. Saintpaul) from cow, and S. 8:i:z6 (putative identity: S. Kentucky) from chicken. On the other hand, we observed the low prevalence of S. Typhimurium and S. Enteritidis in retail meat even when they are the most commonly reported serovars in human clinical samples. This is also the first study to report Salmonella serovars S. Agama, S. Livingstone, S. Meleagridis, S. Schwarzengrund, monophasic S. Typhimurium, S. Augustenbirg, S. Rechovot, S. Newlands, S. Podiensis, and S. Gabon in the Philippines. It is proposed to include non-human and non-clinical isolates in routine Salmonella serotyping, particularly in the food production chain, to track the sources of Salmonella contamination in food and future outbreaks in humans and/or animals. Investigation of the antimicrobial susceptibility of Salmonella isolates acquired from this study shall be performed in future studies to identify emerging resistant strains circulating in foods of animal origin, in particular from cows, swine, and poultry.
Supporting information S1 Appendix. Wet market assessment sheets. Guide questions used to assess the hygienic conditions of each wet markets in Metro Manila. The name of wet markets was concealed. "Tocino", "Embutido" and "Longganisa" are Filipino terms for bacon, Filipino style meatloaf, and Filipino style sausage. (DOCX)