Antimicrobial residues, non-typhoidal Salmonella, Vibrio spp. and associated microbiological hazards in retail shrimps purchased in Ho Chi Minh city (Vietnam)

We investigated antimicrobial residues, non-typhoidal Salmonella (NTS), Vibrio spp. and their associated antimicrobial resistance (AMR), in shrimps locally purchased in Ho Chi Minh City (Vietnam). In addition, we investigated the relationship between AMR in NTS, Vibrio spp. and antimicrobial residue in the same sample. A total of 40 samples of shrimp heads/shells from different retail sources was cultured using ISO 6579–1:2017 (NTS) and ISO/TS 21872–1:2007 (Vibrio spp.). Phenotypic antimicrobial susceptibility was investigated using Vitek (NTS, 34 antimicrobials) and disk diffusion (Vibrio spp., 12 antimicrobials). A total of 9 (22.5%) samples contained antimicrobial residue, including tetracyclines, fluoroquinolones, sulfonamides and macrolides (in 7.5%, 7.5%, 2.5% and 2.5% of samples, respectively). Shrimp samples from supermarkets had a higher prevalence of antimicrobial residue than those purchased in street markets (50% vs. 13.3%) (p = 0.049). A total of 30 (75%) samples were contaminated with NTS. All samples contained Vibrio spp., with V. parahaemolyticus being most common (87.5% samples). A total of 58.9% NTS isolates were multidrug resistant. With regards to the highest priority, critically important antimicrobials, the highest resistance corresponded to quinolones (14.4–47.8%), followed by 3rd and 4th generation cephalosporins (3.3–7.8%). Vibrio spp. isolates were characterised by their high resistance against ampicillin (82.7%) and 3rd generation cephalosporins (8.3–16.5%). Extended Spectrum Beta-Lactamase (ESBL) activity was detected in 28.1% V. parahaemolyticus isolates. Half of ESBL-positive V. parahaemolyticus strains harboured blaCTX-M1. We found an association between the presence of residues and the number of resistances for NTS (p = 0.075) and Vibrio spp. isolates (p = 0.093) from the same sample. These findings suggest that the presence of residues may contribute to the selection of AMR in foodborne pathogens in shrimps. Authorities should strengthen policies aiming at restricting inappropriate antimicrobial usage in shrimp farming, and step up monitoring of antimicrobial residues and food-borne pathogens at retail in Vietnam.


Introduction
Antimicrobial resistance (AMR) is one of the greatest threats to our society (O'Neill, 2016). Among other sources, humans may acquire AMR bacterial infections or AMR-encoding genes through the consumption of contaminated food, including fish and shellfish (Cabello et al., 2013;Likotrafiti et al., 2018). In recent years shrimp farming has rapidly increased, reaching a global production of 3.2 million metric tons in 2017, much of it taking place in Asia (Anon., 2018). This increase is happening in a context of rapid globalization of markets, as well as the threat of climate change (Lauria et al., 2018). Antimicrobials are widely used in shrimp and aquaculture production, both to treat https://doi.org/10.1016/j.foodcont.2019.106756 Received 16 February 2019; Received in revised form 20 June 2019; Accepted 7 July 2019 and prevent diseases (Henriksson et al., 2018). Contamination of aquaculture food products with antimicrobial residues represents a potential health hazard to the consumer due to food poisoning, the development of allergy problems, changes of the intestinal flora, as well as the emergence and subsequent spread of antimicrobial resistance (Okocha, Olatoye, & Adedeji, 2018).
Non-typhoidal Salmonella (NTS) and certain Vibrio spp. are major microbiological hazards associated with shrimp and seafood consumption (Baker-Austin et al., 2018;Tusevljak et al., 2012). Vibrio parahaemolyticus is the leading cause of seafood-borne bacterial gastroenteritis in the world. Both its thermostable direct hemolysin (tdh) and tdh-related hemolysin (trh) are considered major virulence factors of this micro-organism (Raghunath, 2015). In the late 1990s', V. parahaemolyticus was implicated in a large outbreak of enteric disease in central Vietnam, with 523 cases reported (Chowdhury et al., 2004).
NTS is a major cause of gastroenteritis worldwide (Majowicz et al., 2010). In Vietnam, NTS is recognised as a major cause of pediatric diarrhoea (Thompson et al., 2012). There is also evidence of an increase in the incidence of severe invasive infections in hospitalised patients associated due to this organism (Lan et al., 2016;Nga et al., 2012).
The Vietnamese shrimp industry has experienced a considerably expansion over recent years, with most of its production being aimed the export market (mostly to the USA, Europe and Japan). In 2017, shrimp exports made up about half of the total Vietnam seafood exports, with sales worth 3.8 billion US$ (Hong, Hien, Thu, & Lebailly, 2017).
Shrimp exports are regularly screened for their microbiological safety by the companies themselves. However, little is known about the microbiological safety of shrimps available for domestic consumption. Therefore, the aims of this study were: (1) to investigate major foodborne hazards associated with shrimps from local retail sites in Ho Chi Minh City (HCMC), Vietnam, such as antimicrobial residues, NTS and Vibrio spp.; and (2) to characterise the AMR profile of these organisms, including the presence of Extended Spectrum Beta-Lactamases (ESBL) and colistin resistance. In addition we investigated the relationship between the presence of AMR in the two bacterial species and antimicrobial residues in the same batches, which to our knowledge has not been previously investigated.

Sample collection and identification
Batches of shrimps (250-300 g each) were purchased from 40 different retail sites located in 10 districts of HCMC (Vietnam) from March to June 2018. In order to maximize the diversity of sources, from each district three street markets and one supermarket were selected. From each retail site, one batch of live or dead shrimps (chilled, not frozen) was purchased. Shrimps were collected into a clean plastic bag, and were transported to the laboratory within 2 h in an ice-containing box. Five representative specimens per batch were weighted using precision scales. Shrimp species were identified based on their morphological features. Using a pair of sterile scissors, the heads, legs and exoskeleton were separated from the muscle tissue, and were subsequently pooled (shell mix). Muscle tissue samples were investigated for the presence of antimicrobial residues, and the shell mixes were investigated for NTS and Vibrio spp.

Antimicrobial residue analyses
Shrimp muscle tissue samples were investigated for antimicrobial residues using a hierarchical approach. Firstly, they were screened using PremiTest (R-Biopharm AG, Germany), an assay based on the inhibition growth of Bacillus stearothermophilus spores. Positive or inconclusive result samples were then examined for the presence of macrolides, amphenicols, tetracyclines, β-lactams and sulfonamides antimicrobial classes, as well as for the presence of chloramphenicol, streptomycin and gentamicin/neomycin using a Charm II analyzer 7600 (Charm Sciences, USA) (Gaudin, Juhel-Gaugain, Moretain, & Sanders, 2008). Samples that tested positive by Charm II were then confirmed for specific antimicrobials within each class by Ultra-High Performance Liquid Chromatography Tandem Mass Spectrometry (LC-MS/MS). In addition, PremiTest-positive samples were investigated for quinolones by LC-MS/MS. (See Table A with the list of antimicrobials investigated by LC-MS/MS).

Isolation of NTS and vibrio spp.
The shrimp shell mixes were investigated for NTS using a modified ISO 6579-1:2017 method. Briefly, from each sample 25 g of homogenized shell mix was pre-enriched in 225 mL buffered peptone water (BPW, Oxoid, UK) at 37°C for 18 h. A loop of pre-enrichment media was then inoculated on Modified Semi-solid Rappaport-Vassiliadis (MSRV, Oxoid, UK), and incubated at 41.5°C for 24 h. and positive growth was further inoculated on chromogenic Rambach agar (CHROMagar, France) and incubated at 37°C for 24 h (Carrique-Mas, Barnes, McLaren, & Davies, 2009). Matrix-Assisted Laser Desorption Ionization Time-of-Flight Mass Spectrometry (MALDI-TOF MS) (Bruker, Germany) was used to investigate the species identity of three suspected (pink) isolates from each culture. NTS isolates were further classified as either group B, C, D, E or 'others' according to the Kauffmann-White scheme using relevant poly-O antiserum (Grimont P.A. & Weill, 2007). Shrimp shell mixes (25 g) were also investigated for the presence of Vibrio spp. using a modification of the ISO/TS 21872-1:2007 method. Briefly, the steps were: (1) 25 g of the shell mix was suspended in 225 mL of alkaline saline peptone water (ASPW) at 41.5°C for 24 h; (2) a loop of enrichment was cultured on thiosulfate citrate bile and sucrose agar (TCBS, Oxoid, UK) at 37°C for 24 h. Four suspected Vibrio spp. isolates from each sample were confirmed by MALDI-TOF.

Antimicrobial susceptibility testing
All confirmed NTS isolates were tested for their antimicrobial susceptibility against a panel of 34 antimicrobials belonging to 11 classes by Vitek (bioMérieux, Marcy l'Etoile, France) (Livermorea et al., 2002) (33 antimicrobials), as well as by Etest (BioMérieux, France) (colistin). All Vibrio spp. isolates were tested using the disk diffusion method for 12 antimicrobials representative of eight classes (Oxoid, UK). The full list of antimicrobials investigated is displayed in Tables C4 and C5. NTS and Vibrio spp. isolates were classified as susceptible, intermediate or resistant according to CLSI guidelines (M100-S27 for NTS, M45-A2 for Vibrio spp.) (Anon., 2010;2017a). A strain was defined as 'multidrug resistant' (MDR) if it was fully resistant to antimicrobials belonging to at least three different classes. The potential production of ESBLs was investigated by the 'comparative disk diffusion method', using cefotaxime and ceftazidime disks alone, as well as in combination with clavulanate (Anon., 2017a). Antimicrobial susceptibility results were sorted according to the WHO list of antimicrobials of human health importance (Anon., 2017b).

Determination of serovar identity of NTS
NTS isolates were further classified as belonging to either group B, C, D, E or 'others' according to the Kauffmann-White scheme using poly-O antiserum (Grimont P.A. & Weill, 2007). From each sample, one isolate representative of each a serogroup-antimicrobial susceptibility testing result pattern was investigated by Multi-Locus Sequence Typing (MLST). The MLST scheme followed is based on seven loci aroC, dnaN, hemD, hisD, purE, sucA and thrA (Yun et al., 2015).

Investigation of tdh and trh genes of vibrio spp. by PCR
The presence of genes encoding a thermostable direct hemolysin (tdh) and a tdh-related hemolysin gene (trh) was investigated by PCR in all Vibrio spp. isolates (Tada et al., 1992). Positive and negative control isolates were used. The positive control isolates originated from confirmed human cases.

Statistical analyses
We investigated 40 shrimp batches with the aim of determining the prevalence of residues, based on an expected prevalence of 8%, a 95% level of confidence and an 8% relative precision. We expected to obtain 50 NTS isolates from these batches. This sample size (50) allowed determining a prevalence of MDR of ∼27% (based on published data) with a 95% level of confidence and a 9% relative precision. The prevalence of contamination and resistance across variables was compared using chi-square tests. The level of agreement between presence of residues and presence of intermediate (or resistant) isolates in each shrimp batch sample was investigated using the kappa statistic.

Antimicrobial residues
Antimicrobial residues were detected in 9/40 (22.5%) samples by PremiTest. Shrimp samples from supermarkets had a higher prevalence of antimicrobial residues than those purchased in street markets (50% vs. 13.3%) (χ 2 = 3.871, p = 0.049). Four of the nine PremiTest-positive samples were positive by Charm II, whereas the remaining five tested negative (Table 2). Tetracyclines, sulfonamides, and macrolides were detected by Charm II in 7.5%, 2.5%, and 2.5% of samples, respectively. Antimicrobials identified by LC-MS/MS included two tetracycline antimicrobials (tetracycline and oxytetracycline) and two fluoroquinolone antimicrobials (ciprofloxacin and flumequine). The Charm II macrolidepositive sample tested negative for both tylosin and erythromycin by LC-MS/MS. Two samples contained antimicrobial residues above the maximum residue limits (MRL) according to Vietnamese regulations (one contained tetracycline and sulfamethoxazole; another oxytetracycline, and ciprofloxacin). Oxytetracycline and flumequine were also found in two other samples although at concentrations below the MRL (43.7 μg/kg and 41.2 μg/kg, respectively).

Prevalence of contamination with NTS serovars and vibrio spp.
A total of 90 NTS isolates were recovered from 30/40 (75%) shrimp batches ( Table 1). The prevalence of NTS among samples purchased in street markets was higher than that from supermarket samples, although this difference was not significant (80% vs. 60%; χ 2 = 0.711; p = 0.399). There was lower probability of recovering NTS from Pre-miTest-positive samples than from PremiTest-negative samples (5/9 vs. 25/31), although this difference was not significant (Fisher's exact test, p = 0.190). MLST was performed on 62 isolates with a unique serogroup-antimicrobial susceptibility pattern. The remaining 28 isolates were assigned to serovar based on MLST results of isolates recovered from the same sample and with the same serogroup-AST pattern. A total of 28 MLST sequence types (ST) corresponding to 25 NTS serovars were identified. One NTS strain (Group B) could not be assigned to ST, and therefore its serovar identity was not determined (Table B). The most prevalent serovars identified were Braenderup (present in 20% samples), Anatum (16.7% samples), Saintpaul (13.3% samples), Rissen and Litchfield (10% samples each). All (100%) samples were positive for

Detection of toxin-encoding genes
None of the 133 Vibro spp. isolates tested positive for either the tdh or trh genes.

Detection of ESBL genes in NTS and vibrio spp. isolates
The phenotypically ESBL-positive serovar Branderup isolate tested negative for all ESBL genes investigated. All three serovar Infantis isolates (from the same sample) were positive for bla CTX-M9 . The three serovar Give isolates tested positive for both bla CTX-M1 and bla TEM . Interestingly, all isolates were fully susceptible to cefepime and cefoxitin. A total of 9/18 (50%) ESBL-positive V. parahaemolyticus strains were positive for bla CTX-M1 . In addition, one of these isolates tested positive to the bla TEM gene.

Relationship between residues, and AMR in NTS and vibrio spp. isolates
NTS isolates from shrimp samples that tested positive to PremiTest (n = 15) were resistant to a median of 10 antimicrobials [IQR 3-13], whereas NTS isolates from samples testing negative (n = 75) were resistant to 5 antimicrobials [IQR 0-9] (Wilcoxon test W = 724.5, p = 0.075). Vibrio spp. isolates from PremiTest-positive samples (n = 29) were resistant to a median of 1 [IQR 1-4] antimicrobial, compared with 1 [IQR 1-2] among Vibrio spp. from PremiTest-negative samples (n = 104) (Wilcoxon = 1792; p = 0.093) (Fig. 3). We found fair agreement between presence of residue (PremiTest) and presence of co-trimoxazole and ciprofloxacin resistant NTS from the same sample (kappa values 0.265 and 0.365; p ≤ 0.016). We also found a fair agreement between presence of a quinolone residue and ciprofloxacin resistance in Vibrio spp. isolates from the same sample (kappa = 0.383; p = 0.005). In addition, there was a moderate agreement between samples that contained ESBL-positive Vibrio spp. and ESBL-positive NTS isolates (kappa = 0.515, p < 0.001) (Table D). However, in neither of the two samples that contained both phenotypically ESBL-positive NTS and Vibrio spp. we could demonstrate the presence of the same genes: in one sample V. parahaemolyticus was positive for both bla TEM and bla CTX-M1 , whereas no ESBL genes were detected in the NTS isolate; in the other, NTS harboured the bla CTX-M9 , wheras V. parahaemolyticus tested negative for all ESBL genes.

Discussion
This study evidenced a high prevalence of contamination of shrimp samples with antimicrobial residues (22.5%), NTS (75%), and Vibrio spp. (100%). This result is in line with a previous survey of shrimps from local markets in the Red River and Mekong Delta regions of Vietnam (13.0% and 33.3%, respectively) (Pham et al., 2015), but higher than previous results reported from the Vietnamese provinces of HCMC, Thai Binh and Nha Trang (8.8, 1.8 and 3.2%, respectively) (Uchida et al., 2016). However, we were only able to establish the identity of the antimicrobial residue in 4 of 9 samples that tested positive by a bacterial inhibition test. This may be the result of a falsepositive result in our screening test, or (more likely) due to the presence of antimicrobial residues not investigated in this study. All antimicrobials confirmed in our samples (tetracyclines, ciprofloxacin,  flumequine) are known to be commonly used in Vietnamese aquaculture (Pham et al., 2015), although levels of flumequine (two samples) and oxytetracycline (one sample) where within the legal limits. We found considerable agreement between the screening results by Charm II and the residues detected by LC-MS/MS. The exception was a macrolide-positive sample by Charm that subsequently tested negative for both erythromycin and tylosin. We hypothesize that this may reflect the use of spiramycin, another macrolide antimicrobial known to be used in aquaculture (Sekkim & Kum, 2011) that was not investigated. It is likely that the observed prevalence of residues in shrimp samples reflects the use of antimicrobials late in the production cycle, thought to be common practice in Vietnamese aquaculture (Pham et al., 2015). A surprising finding was the higher prevalence of residues in shrimps procured from supermarkets than from (more informal) street markets. This suggests that intensive production systems (normally associated with the supermarket supply chains) may be associated with a higher levels of antimicrobial usage. This is consistent with a previous study that reported a higher prevalence of residues in shrimp products procured in supermarkets compared with wholesale markets (Uchida et al., 2016). The observed prevalence of contamination with NTS in this study (75%), is higher than in previous studies from Vietnam, China, Thailand and Bangladesh (prevalence levels ranging between 13% and 55%) (Minami et al., 2010;Phan et al., 2005;Pinu, Yeasmin, Bar, & Rahman, 2007;Uddin, Larsen, Barco, Phu, & Dalsgaard, 2015). A recent study investigating NTS in 25 g meat samples purchased across markets in Vietnam resulted in a comparatively lower prevalence (71.8% in chicken, 70.7% in pork and 62.2% in beef samples) (Nhung et al., 2018). In contrast with the antimicrobial residue results, shrimp samples purchased in street markets had higher prevalence of NTS contamination than shrimps purchased in supermarkets, although this difference was not statistically significant. We hypothesize that this may reflect deficiencies in the cold chain associated with street markets. A study in Thailand reported absence of NTS in shrimps purchased in supermarkets, compared with a 50.2% prevalence among shrimps purchased in street markets (Minami et al., 2010).
We found a considerable serovar diversity (25) of NTS in our samples, with serovars Braenderup, Anatum, Saintpaul and Rissen accounting for 46.7% of all NTS isolates. Most serovars detected in shrimps have been isolated in outbreaks of human disease in different countries. A study from 2004 on 56 human clinical (diarrhoea, fever) in Vietnam identified 12 different serovars (Vo et al., 2006), of which four (Anatum, Albany, Typhimurium and Enteritidis) were detected in our shrimp samples. However, in that study, serovars Typhimurium and Enteritidis accounted for 50% of human cases, whereas in our study they only accounted for 6.6% of all isolates. A more recent study on invasive (bloodstream infection) isolates identified 19 serovars, of which 8 were identified in our study (Lan et al., 2016). Surprisingly, serovar Weltevreden, which was predominant in a previous study on shrimp farms in the Mekong Delta of Vietnam (Uddin et al., 2015) was found only in one sample, and was a pansusceptible strain. A possible explanation for this discrepancy is that NTS contamination of shrimps mainly occurs after harvesting. It has been shown experimentally that seafood is an excellent nutrient media supporting vigorous post-harvest NTS multiplication at ambient temperatures (Kumar, Datta, & Lalitha, 2015). Since no routine diagnostic or surveillance data exists for human salmonellosis, it is not possible to know to what extent shrimps are a significant source of NTS to the community in Vietnam. A total of 58.9% NTS isolates were MDR, a figure higher compared with overall MDR prevalence in isolates from meat in a previous study using the same antimicrobial panel (52.2%) (Nhung et al., 2018). Overall resistance to ciprofloxacin (33.3%) is of great concern, since this antimicrobial is often used to treat enteric infections. We found that 7.8% isolates were ESBL-producers and were fully resistant to at least one of the four 3rd-4th generation cephalosporins investigated. ESBL-producing NTS organisms were identified as belonging to serovars Braenderup (1), Infantis (3), and Give (3). These ESBL-producing isolates were MDR (including quinolone resistance), although were susceptible to carbapenems. The latter is the first choice drug in the treatment of ESBL-producing microorganisms (Zhanel et al., 2007).
All (100%) retail shrimp samples were contaminated with Vibrio spp., being V. parahaemolyticus the most prevalent species (87.5% samples). Other Vibrio species were isolated in 10.0-60.0% samples. These levels of contamination are comparable with studies in northern Vietnam (99.5% prevalence) (Tra et al., 2016) and Malaysia (100%) (Letchumanan, Yin, Lee, & Chan, 2015), although higher than results from Turkey (67%) (Mus, Cetinkaya, & Celik, 2014), confirming that Vibrio spp. organisms are omnipresent in the shrimp farm aquatic environment (Gopal et al., 2005). In addition to V. parahaemolyticus, both V. vulnificus and V. cholerae non-O1 are also known to cause severe human disease (Deshayes et al., 2015). We did not, however, find evidence of any of the two major virulence genes investigated (tdh and trh) in any of the 133 Vibrio spp. isolates. Previous research in Malaysia has shown a low prevalence of tdh (4%) and trh (12%) genes in nonclinical V. parahaemolyticus isolates (Paydar, Teh, & Thong, 2013). In a study on 47 environmental isolates from India, only 4.2% and 2.1% harboured the tdh and trh genes, respectively (Koralage et al., 2012). However there was no evidence of these genes in isolates investigated in northern Vietnam (Tra et al., 2016), Hong Kong (Wong, Liu, Wan, & Chen, 2012), or Sri Lanka (Koralage et al., 2012). In the late 1990s', V. parahaemolyticus was implicated in a large outbreak of enteric disease in central Vietnam, with 548 cases reported (Tuyet et al., 2002). It was determined in further analyses that the prevailing serovar changed over time (O3:K6 in 1997, O4:K68 in 1998, O1:K25 in 1998-1999), and that 85% clinical isolates harboured either the tdh or trh genes (Chowdhury et al., 2004).
We found that 82.7% of Vibrio spp. isolates were resistant to ampicillin (100% for V. parahaemolyticus and V. alginolyticus). This prevalence was comparable to published resistance levels among V. parahaemolyticus isolates from shrimps in northern Vietnam (87%) and Malaysia (82%) (Letchumanan et al., 2015;Tra et al., 2016). A total of 31.3% V. parahaemolyticus isolates were resistant to third generation cephalosporins and 28.1% were ESBL producers. In about half of those strains, the gene carried was bla CTX-M1 . However, in another half the molecular basis for ESBL activity could not be established. Given that these genes are highly mobile, and are often inserted in plasmids and transposons, we hypothesize that V. parahaemolyticus may act as a reservoir of ESBL genes (Canton, Jose, & Galan, 2012). Interestingly we found an association between the presence of ESBL in NTS and V. parahaemolyticus isolated from the same sample.
The study confirmed the association between presence of antimicrobial residues and phenotypic resistance in NTS and Vibrio spp. in the same samples. This may be a reflection of AMR selection during the farming process, or may alternatively reflect post-harvesting contamination. The latter would be likely if the antimicrobial residue in the sample contributed to preferentially select for contamination with more resistant strains.
This study provides evidence of high levels of contamination with antimicrobial residues, NTS and Vibrio spp. among shrimps purchased in retail sites in HCMC. We found a high prevalence of MDR among NTS, with worryingly high levels of quinolone resistance. Although most Vibrio spp. isolates are unlikely to be pathogenic, the high carriage levels of ESBL in V. parahaemolyticus is of concern. We recommend authorities to enforce existing policies aiming at restricting inappropriate antimicrobial usage on shrimp farms, stepping up hygiene conditions during harvesting, transporting and retailing of shrimps, and to establish monitoring of antimicrobial residues, NTS and Vibrio spp. (focused on ESBL and virulence factors) in Vietnam. These findings should encourage the establishment of microbiological surveillance systems focused on health hazards in aquaculture food products, as well as strengthening laboratory capacity to enable comparisons between NTS and Vibrio spp. isolates from shrimps and human cases of disease.

Conflicts of interest
The authors declare no conflict of interest.