Antibiotic-resistant bacteria and gut microbiome communities associated with wild-caught shrimp from the United States versus imported farm-raised retail shrimp

In the United States, farm-raised shrimp accounts for ~ 80% of the market share. Farmed shrimp are cultivated as monoculture and are susceptible to infections. The aquaculture industry is dependent on the application of antibiotics for disease prevention, resulting in the selection of antibiotic-resistant bacteria. We aimed to characterize the prevalence of antibiotic-resistant bacteria and gut microbiome communities in commercially available shrimp. Thirty-one raw and cooked shrimp samples were purchased from supermarkets in Florida and Georgia (U.S.) between March–September 2019. The samples were processed for the isolation of antibiotic-resistant bacteria, and isolates were characterized using an array of molecular and antibiotic susceptibility tests. Aerobic plate counts of the cooked samples (n = 13) varied from < 25 to 6.2 log CFU/g. Isolates obtained (n = 110) were spread across 18 genera, comprised of coliforms and opportunistic pathogens. Interestingly, isolates from cooked shrimp showed higher resistance towards chloramphenicol (18.6%) and tetracycline (20%), while those from raw shrimp exhibited low levels of resistance towards nalidixic acid (10%) and tetracycline (8.2%). Compared to wild-caught shrimp, the imported farm-raised shrimp harbored distinct gut microbiota communities and a higher prevalence of antibiotic-resistance genes in their gut. The presence of antibiotic-resistant strains in cooked shrimps calls for change in processing for their mitigation.

Antibiotic resistance is one of the severe threats faced by the world. In the U.S., more than 2.8 million people are infected with antibiotic-resistant strains each year, resulting in at least 35,000 deaths 1 . A review on global antibiotic resistance reported that by the year 2050, 10 million lives per year are at risk due to the rise of infection by antimicrobial-resistant strains 2 . According to the United States Department of Agriculture (USDA), ~ 97% of fish and shellfish consumed in the U.S. are imported from China, Ecuador, India, Indonesia, Thailand, and Vietnam 3,4 . In the large aquaculture operations, where farmed shrimps are cultivated as monoculture at high density, the prevalence of pathogens is one of the biggest challenges faced by the aquaculture industry 5 . Infection caused by pathogens in aquaculture operations can lead to massive production loss 6 that can completely wipe out the shrimp farms 7 . The post-larvae infection caused by bacterial pathogens in shrimps in 2013 resulted in 1 billion dollars in production loss 6 . To prevent the production loss associated with bacterial infections, the aquaculture industry and shrimp hatcheries are heavily dependent on the application of antibiotics as prophylactic and therapeutic agents [8][9][10][11][12] . The use of multiple classes of antibiotics, e.g., tetracyclines (chlortetracycline, oxytetracycline, and tetracyclines), fluoroquinolone (enrofloxacin and ciprofloxacin), quinolones (oxolinic acid and norfloxacin), sulfonamides, chloramphenicol, and nitrofuran are practiced in shrimp farming 5,13,14 . Antibiotics are administered in aquaculture by mixing them in their diet or by adding them in rearing water. Continuous application of antibiotics by the shrimp hatcheries and farms facilitates the development of antibiotic-resistant bacterial (ARB) strains 15 . Non-biodegradable antibiotics with a long activity period apply selective pressures in an environment 16 . For instance, application of quinolones (oxolinic acid) 17 in a high microbial load environment, forms a perfect combination for the selection of antibiotic-resistant strains in farm-raised shrimp. Once an antibiotic-resistance determinant has been selected, it can easily be transferred to other strains. Popular foods like cooked shrimp or seafood mix are processed using mild heat treatment 18 and can act as perfect vehicles for transferring ARB strains. Currently, FDA cooking recommendations are focused on mitigating Listeria monocytogenes in cooked shrimp 19 , which may not be effective to eliminate antibiotic-resistant bacteria spread over multiple genera. Further, antibiotic-resistant strains can be more tolerant to milder temperature treatments used for shrimp processing 20,21 . Cooked shrimp samples are mostly harvested and processed overseas. Each country and processing facility has its own processing specifications, due to which there is a considerable variation in the microbiological specifications of cooked shrimp available in the retail market. These samples that can be directly consumed after thawing or short heat treatment can be a perfect vehicle for ARB strains and opportunistic pathogens. Past studies have extensively studied the prevalence of antibiotic resistance in fresh produce 22 , beef 23 , poultry 24 , and manure from different livestock 25,26 . However, limited research has been conducted evaluating the prevalence of ARB strains in shrimp. Cross contact of raw shrimp with processed shrimp is another issue of concern 27 . The presence of ARB strains and the opportunistic pathogen can be detrimental to human health and limit treatment options.
Microbiome diversity has a significant impact on host health 28 . The competitive exclusion theory entails that the higher gut microbial diversity, the lower the possibility for pathogenic colonization 29 . The application of antibiotics at hatcheries and shrimp farms can cause dysbiosis. Researchers in the past have characterized and observed major gut microbiome composition differences among diseased and healthy shrimps 30 . Similarly, a comparison of the gut microbiome composition of farmed shrimp raised with antibiotics and wild-caught shrimp raised without antibiotics can be used for developing alternate sustainable methods for shrimp farming. Therefore, the aim of this study was to (a) isolate and characterize ARB strains among commercially available shrimp samples and (b) characterize the gut microbiome diversity of wild-caught shrimp from the U.S. compared to farm-raised shrimp imported from Ecuador.

Results
Distinct arrays of antibiotic resistance spectrum and the prevalence of antibiotic-resistant bacteria in raw versus cooked shrimp. A total of 31 shrimp samples (13 cooked and 18 raw shrimp) ( Table 1) were tested for the presence of antibiotic-resistant bacteria. These shrimp samples originated from eight different countries i.e., Argentina (n = 2), Ecuador (n = 3), India (n = 6), Indonesia (n = 6), Panama (n = 1), Thailand (n = 3), U.S. (n = 7) and Vietnam (n = 3). Aerobic plate count (APC) and coliform count of cooked shrimp samples (n = 13) varied from < 25 CFU/g to 6.2 log CFU/g and < 25 CFU/g to 2.6 log CFU/g, respectively. All shrimp samples tested negative for Salmonella. Presumptive black colonies obtained on Xylose Lysine Deoxycholate agar with tergitol (XLD tn ) and Hektoen Enteric (HE) agar plates tested negative using the Salmonella specific real-time PCR assay.
Distinct signatures of gut microbiome communities and antibiotic-resistant bacteria in wild-caught versus farm-raised shrimp. To further investigate whether and how the overall gut microbiome communities associated with these shrimps differ by geographical origin as well as the rearing method (farm vs. wild-caught), the 16S rRNA metagenomics analysis of the gut microbiome associated with the two groups of shrimps, i.e., farm-raised (from Ecuador) versus wild-caught (from the U.S.) was performed. Shrimp  www.nature.com/scientificreports/ samples collected from countries with a smaller number of samples were removed from the analysis. Interestingly, the analysis revealed remarkably distinct spectrums of microbiome signatures between the two groups of shrimps (Fig. 2). The analysis of β-diversity (in terms of the principal coordinate analysis of the Bray-Curtis index) clearly demonstrated significantly discrete spectrums of microbiome signatures between shrimp from the U.S. versus those from Ecuador (Fig. 2a). Further α-diversity analyses in terms of observed operational taxonomic units (OTU), Chao1 index as well as the Shannon diversity index revealed higher bacterial species richness and diversity in wild-caught shrimps from the U.S. compared with shrimps from Ecuador ( Fig. 2b-d).
Accordingly, further analysis at the level of major bacterial phyla and genera showed clearly distinct microbiome composition between the two groups of shrimps ( Fig. 2e-f). Compared to farm-raised shrimps from Ecuador, the wild-caught shrimps from the U.S. harbored a higher abundance of Proteobacteria and Planctomyces and had a lower abundance of Tenericutes and Bacteroidetes (Fig. 2e). Interestingly, the analysis of organism-level phenotypes showed that the shrimps from the U.S. harbor a significantly higher overall ratio of Gram-negative to Gram-positive bacterial taxa as compared to shrimps from Ecuador (Fig. 2g). In addition, the proportion of aerobic taxa was higher while that of facultative anaerobic taxa was lower in shrimps from the U.S. versus those from Ecuador (Fig. 2h). Linear discriminatory analysis (LDA) effect size (LEfSe) analysis was performed to infer the relative abundance of unique taxonomic clades and Kyoto Encyclopedia of Genes and Genomes (KEGG) metagenome orthologs that were significantly (p < 0.01) over-or under-represented between shrimp from Ecuador and the U.S.. The LEfSe analysis identified several unique bacterial taxa that drive differences in the gut communities between the two groups of shrimps (Fig. 2i). Figure 2j further simplifies these data and presents the LDA scores of the taxa that were significantly (p < 0.01) different between the two groups of shrimps. The shrimps from the U.S. were uniquely distinguished by a higher proportion of Proteobacteria, Alteromonadales, Rhizobiales, Synechococcaceae, Myxococcales, and Planctomyces, whereas shrimps from Ecuador were characterized by a higher abundance of Tenericutes, Flavobacteriales, Desulfovibrio, Oscillatoriales, Mycobacterium, and Bacteroidales ( Fig. 2i-j). In addition, data from the study showed several bacterial species that were significantly different between the two groups of shrimps ( Supplementary Fig. 1a). These differentially abundant markers could be considered as potential metagenomic biomarkers for the identification of shrimp from Ecuador (farm-raised) and the U.S. (wild-caught). Further, the PICRUSt (Phylogenetic Investigation of Communities by Reconstruction of Unobserved States) analysis of the KEGG orthologs associated with these bacterial signatures also revealed distinct clusters among the two groups ( Supplementary Fig. 1b). The shrimps from the U.S. were characterized mainly by a higher (p < 0.01) abundance of bacterial taxa associated with the metabolism of several amino acids, lipids, cofactors, and vitamins in addition to the xenobiotic degradation ( Supplementary Fig. 1b).
In contrast, the shrimps from Ecuador presented a higher proportion of taxa associated with the metabolism of amino acid-related enzymes and gluconeogenesis in addition to the taxa related to primary immunodeficiency ( Supplementary Fig. 1b).
In addition to gut microbiome analysis, the bacterial genomic DNA from raw shrimp gut from the U.S. (wild-caught) and Ecuador (farm-raised) was subjected to real-time PCR assays for examining the prevalence of antibiotic-resistance genes. Interestingly, compared to raw shrimp sample of U.S. origin (wild-caught), the shrimp from Ecuador (farm-raised) showed a significantly higher prevalence of bla sul-1 , bla sul-2 , bla qnrS , bla aadA , bla mec-A , bla CMY , bla floR , bla cat , and bla tet-A (p < 0.01) while demonstrating a lower prevalence of bla CTX-M and bla van-A (p < 0.01) (Fig. 3).

Discussion
According to the Food and Agricultural Organization (FAO), the microbiological limits for frozen cooked shrimp samples with aerobic plate count 10 7 CFU/g is the boundary between marginally acceptable counts and unacceptable counts 32 . In our study, the APC of cooked shrimp samples ranged from < 25 CFU/g to 6.2 logs CFU/g. Cooked shrimp samples from India (sample No. = 30) and Vietnam (sample No. = 31) had a high APC. Variations in hazard analysis critical control point and processing plans in different countries and temperature fluctuation during transportation may be possible reasons for the observed variation. Some samples showed as low as < 25 CFU/g APC, while other samples had high APC and even tested positive for the presence of antibioticresistant bacteria. We found that 22% (n = 24/107) of the bacterial isolates were resistant to one or more antibiotics. Interestingly, compared to isolates from raw shrimp, the resistance among isolates from cooked shrimp was significantly higher towards chloramphenicol (19%), aztreonam (7.7%), cefotaxime (7.3%), and meropenem (7.3%) (p < 0.01) (Fig. 1a). In a similar study, Nawaz et al. 33 reported resistance to nalidixic acid, ampicillin, tetracycline, and chloramphenicol in 52% (55/105) of E. coli isolates from imported black Asian shrimp (n = 330).
Shrimp processing facilities rely on the application of a shorter and milder thermal treatment, which may not be sufficient for mitigation of ARB strains spread over multiple genera. Several studies have demonstrated the interplay between stress associated with food processing, antibiotic resistance, and vice-versa 34 . In fact, exposure to mild temperatures has been shown to enhance the selection of ARB strains. Ebinesh et al. observed a reduction in the inhibition zone for amikacin, imipenem, meropenem, norfloxacin, piperacillin, and tazobactam among six A. baumannii strains following heat treatment at 45 °C 21 . In another study, the ARB strains of E. coli O157:H7 were found to be less susceptible to 55 °C treatment compared to the wild strain 20    The presence of ARB strains in cooked food may promote the transfer of resistance genes to bacterial strains in the human gut microbiome or pathogens. Oxytetracycline is widely used by hatcheries across the world 36,37 . Along with the widespread use of oxytetracycline for aquaculture, resistance to this drug is commonly observed in bacterial isolates from shrimp 38 . In the present study, tetracycline resistance was associated with Enterobacter hormaechei, Enterococcus faecalis, Morganella morganii, Proteus mirabilis, and Vibrio parahaemolyticus isolates from the shrimp imported from India, Vietnam, and Indonesia. Chloramphenicol, a broad-spectrum antibiotic, is banned for use in food-producing animals by the U.S., European Union, and many other countries due to its toxic side-effects in humans 39 . However, the presence of chloramphenicol residue is a common reason for the rejection of imported shrimp by the FDA 40 . The presence of chloramphenicol residues reflects the indiscriminate use of banned antibiotics in shrimp-exporting counties. In the present study, chloramphenicol resistance was associated with Enterobacter hormaechei, Enterobacter lignolyticus, Proteus mirabilis, Morganella morganii, and Serratia marcescens isolates. In cooked samples, chloramphenicol resistance was associated with Enterococcus faecalis, M. morganii, P. mirabilis, and S. maltophilia, all of which are known to be opportunistic nosocomial pathogens and cause infection in the immunosuppressed population 41 . Not many reports are available on the isolation of carbapenem-resistant bacteria from food-producing animals, including shrimp and their environment 42 . A pilot study conducted by the National Antimicrobial Resistance Monitoring System (NARMS) reported carbapenem-resistant isolates with bla NDM-1 resistance genes in Acinetobacter baumannii and Aeromonas sobria from seafood samples collected from the U.S. 43 ), and V. parahaemolyticus (bla OXA48 ). The carbapenem-resistance genes are plasmid-borne, which can facilitate their spread in the food-producing environment and can promote the development of co-resistance 42 .
In the present study, low levels of resistance were observed towards ESBL antibiotics. All isolates were susceptible to ceftazidime, 7.7% of isolates showed resistance to cefotaxime and aztreonam, and 16.6% of isolates showed intermediate resistance to cefotaxime. Intermediate resistance to cefotaxime was primarily associated with strains belonging to the Acinetobacter pittii. A previous study also reported resistance to third-generation (cefotaxime, ceftriaxone, and ceftazidime) and fourth-generation (cefepime) cephalosporins in two E. cloacae isolate from fresh seafood 44 . A similar study reported the isolation of ESBL-producing Enterobacteriaceae from fresh seafood samples collected from the India market, wherein 80% (n = 53/66) of E. coli isolates were found to be ESBL producers 45 . MDR phenotype was reported among 169 isolates with resistance to third-generation cephalosporins, aztreonam, ertapenem, and meropenem 45 . Further, the isolates from the study were found to be positive for ESBL-encoding genes bla CTX , bla SHV, and bla TEM and the metallo-β-lactamase gene bla NDM-1 . The extensive use of unapproved antibiotics for aquaculture operations in India can be a possible reason for this higher prevalence of antibiotic-resistant bacteria in shrimp harvested from India. In 2019, a total of 61 shrimp entry lines were refused by the FDA due to the presence of banned antibiotic residues from different countries. Among all refused products, shrimp products exported from India (n = 33/61) have consistently tested positive for banned antibiotics reflecting extensive use of banned antibiotics for aquaculture in India [46][47][48] .
The strains of S. maltophilia were isolated from farm-raised shrimp from Vietnam and following the disc diffusion assay antimicrobial susceptibility testing (AST) analysis of the strain confirmed its intrinsic resistance (i.e., towards chloramphenicol and ceftazidime) 49 . Vietnam allows the use of 32 antibiotics for aquaculture 50  Based on the observed antibiotic resistance profile by disc diffusion assay, we selected six isolates and further characterized their antibiotic-susceptibility array using Sensititer Gram-Negative GN2F plates. The data revealed resistance to beta-lactams, cephalosporins, quinolones, dihydrofolate reductase inhibitor/sulfonamide, and fluoroquinolones class of antibiotics and intermediate resistance to carbapenem antibiotic (i.e., imipenem) (Supplementary Table 1). Based on the updated definition of MDR viz. "the isolate is non-susceptible to at least 1 agent in ≥ 3 antimicrobial categories" 53 , all of the isolates tested herein using Sensititer Gram-Negative GN2F plates were characterized as MDR isolates. The widespread trade of shrimp can act as a vehicle for the dissemination of antibiotic-resistant genes. Currently, the FDA relies on only testing antibiotic residue 54 in imported seafood. Hence, the presence of MDR strains in cooked shrimp samples points toward a gap in monitoring measures.
Despite emerging interest and research on marine and seafood-related microbiome, the microbial communities associated with shrimp gut remain relatively underexplored. In our study, the gut microbiota of shrimp samples collected from the U.S. had a higher abundance of Proteobacteria and Planctomyces and had a lower abundance of Tenericutes and Bacteroidetes. A study characterizing the gut microbiota of adult Litopenaeus vannamei raised in farms located in China reported that the adult shrimp gut from farm-A was dominated by Proteobacteria, Tenericutes, Firmicutes, Planctomyces, and Verrucomicrobia 55 . Parallel findings were reported wherein L. vannamei samples were collected from Mexico, showing a higher abundance of Proteobacteria, Cyanobacteria, Actinobacteria, Gemmatimonadetes, Bacteroidetes, and Firmicutes 56 .
Further, we observed higher bacterial species richness and diversity in wild-caught shrimps from the U.S. when compared to farm-raised shrimps from Ecuador. Rungrassamee et al. reported the presence of higher bacterial diversity in wild-caught Penaeus monodon (Tiger shrimp) compared to farmed-raised samples 57 . The major factor affecting the composition of aquatic organism gut microbiota are egg microbiota 58 , larval rearing water 59 , shrimp diet [60][61][62] , and environmental factors 63 . Fish on an herbivorous diet has been shown to have higher diversity compared to fish on a carnivorous diet [60][61][62]64 . Farm-raised shrimp are commonly fed an artificially formulated diet consisting of various animal proteins. A similar significant difference (p < 0.05) in the diversity indices were observed among the wild-caught and farm-raised L. vannamei samples collected from Mexico 56 , wherein Actinobacteria and Nitrospirae were found to be significantly abundant in wild-caught L. vannamei samples compared to Bacteroidetes, Gemmatimonadetes, Fusobacteria and Spirochaetes phyla, which were abundant in farm-raised samples 56 .
The two-dimensional PCoA plot for the beta-diversity for the shrimp samples from the U.S. and Ecuador showed clear clustering of representative samples into two distinct groups. Shrimp samples collected in our study were comprised of Pandalus borealis (Pink shrimp) and Litopenaeus setiferus (White shrimp). These two shrimp species clustered together in the PCoA plot, showing that the region of shrimp origin is a more crucial factor affecting gut microbiota than host genetics. Tzeng et al. 65 characterized the gut microbiome composition of two closely related shrimp species, i.e., Macrobrachium nipponense (oriental river prawn) and M. asperulum, harvested from river and lake water located in Taiwan. Penaeus monodon samples were used as control. Data from the study showed that the host genetics and habitat both play important roles in determining the gut microbial composition of shrimp. Further, the PCoA plot from the study revealed that the host habitat has a much greater www.nature.com/scientificreports/ impact on the shrimp gut microbial composition. The effect of habitat on the gut microbial composition can be attributed to the similar diet and exposure to similar aquatic microbiota among different species of shrimp from a geographical region. The PICRUSt data from a previous study showed an abundance of pathways dedicated to amino acid metabolism, lipid metabolism, and xenobiotics biodegradation in wild-caught L. vannamei 56 , which is similar to the PICRUSt data for wild-caught shrimp collected from the U.S. in our study. These distinct microbiome signatures observed among the U.S. wild-caught shrimp, when confirmed by further studies, could be used for designing region-specific probiotic therapies for improving shrimp health and can support the aquaculture of shrimp in the U.S. Pure culture strains of E. hormaechei were isolated in this study, which has been used in the past as a shrimp probiotic 66 .
In the present study, the prevalence of sulfonamide (bla sul1 and bla sul2 ), quinolone (bla qnrS ), aminoglycoside (bla aadA ), beta-lactam (bla mecA and bla CMY ), phenicol (bla floR and bla cat ), and tetracycline (bla tetA ) resistance genes was significantly higher in the gut of shrimp imported from Ecuador (Fig. 3). The higher abundance of these antibiotic-resistance genes in farm-raised shrimp can be attributed to the application of antibiotics for aquaculture operations facilitating the selection of strains with resistance genes. The presence of these genetic elements can facilitate the transfer of selected genes to genetically related and unrelated strains via the food chain 67 . A previous study has also shown a strong positive correlation of the presence of bla sul1 and bla qnrD genes with the abundance of Proteobacteria; bla cmlA and bla floR genes with the abundance of Bacteroidetes; bla floR and bla sul2 with Firmicutes; and bla cmlA , bla tetG , bla qnrB, and bla qnrA with the abundance of Acidobacteria and Cyanobacteria 55 . Notably, these five phyla (i.e., Proteobacteria, Bacteroidetes, Firmicutes, Acidobacteria, and Cyanobacteria) were dominant phyla among farm-raised shrimp from Ecuador 55 .
One of the limitations of this study was our dependence on Sanger sequencing of 16S rDNA region for the identification of the obtained isolates. 16S rDNA sequencing is not very accurate for the differentiation of closely related species. To our knowledge, this is the first report demonstrating the isolation and prevalence of MDR strains of S. maltophilia from imported cooked shrimp. Selected cooked shrimp showed high APC as well as coliform counts. The data demonstrate that the presence of MDR bacterial strains and opportunistic pathogens in imported cooked shrimp products may pose a threat to human health and calls for the need for standardization of effective cooking time-temperature combination for potential mitigation of antibiotic-resistant strains in cooked shrimp samples. Further, data from our study can be used as baseline data for the U.S. wild-caught shrimp microbiome composition, and with larger validation studies, the microbiome markers identified in our study can be used for the development of novel PCR-based detection assays.

Methods
Shrimp samples. Raw (n = 18) and cooked (n = 13) shrimp samples were purchased from March to September 2019 from stores, supermarkets, and retailers located in the states of Florida and Georgia. The collected shrimp samples were either frozen pre-packaged or thawed shrimp placed on ice in bulk. The pre-packaged samples had a clear country of origin label. Similarly, thawed shrimp placed on ice in bulk collected from the supermarket had a country-of-origin label. Cooked shrimp samples were already peeled, deveined, and precooked, and could be directly consumed after thawing. Sample details are listed in Table 1.
Aerobic plate count and coliform count. Twenty-five grams of cooked shrimp samples from each collected sample were weighed in Whirl-Pak filtered stomacher bags (Nasco, MI, U.S.), and samples were diluted using 225 mL of sterile 0.1% peptone water (w/v). Samples were stomached at 230 rpm for 2 min and were serially diluted using 9 mL of 0.1% (w/v) peptone water tubes. Diluted samples were plated on plate count agar (Hardy Diagnostics, CA, U.S.) and violet red bile agar (Hardy Diagnostics, CA, U.S.). Plates were incubated at 37 °C for 48 h, and colonies were enumerated after the incubation period, and obtained counts were expressed in log10 CFU/mL. Detection of antibiotic-resistance genes using real-time PCR assays. Pure culture bacterial DNA isolated from broth culture was used for real-time PCR assays. DNA samples were tested for the presence of antibiotic resistance genes (bla KPC , bla OXA-48 , bla NDM , bla VIM , bla CTX-M , bla IMP , bla CMY , bla ACC , bla SHV , and bla TEM ) by real-time PCR assay as previously described 55,68,74,75 . Microbiome analysis. Shrimp gut microbiome was analyzed according to our previously described methods [76][77][78] . In brief, the Earth Microbiome Project (EMP) benchmarked protocol 79 was used for the characterization of shrimp gut microbiome. Microbial community analysis, protocols, and standards 80 were employed by following a barcoded high-throughput sequencing approach, as described by Caporaso et al. 81 . Bacterial genomic DNA from the gastrointestinal tract of raw shrimp samples (Table 2) from the U.S. and Ecuador was extracted using the PowerFecal DNA kit (Qiagen, CA, U.S.). The V4 hypervariable region of the 16S rDNA gene was PCRamplified using the universal primer pair 515F/806R 81 ; the resulting uniquely barcoded amplicons were purified using Agencourt AMPure XP magnetic purification beads (Beckman Coulter, CA, U.S.) and quantified with Qubit fluorimeter (Qubit3; Invitrogen, CA, U.S.) and the dsDNA H.S. assay kit (Life Technologies, CA, U.S.); and the amplicon library was generated according to methods of Caporaso et al. 81 . The purified PCR products were pooled in equal molar concentrations and sequenced on one 2 × 300-bp Illumina MiSeq run (Miseq reagent kit v3; Illumina Inc., CA, U.S.) for paired-end sequencing. The sequencing quality control was carried out with on-board Miseq Control Software and Miseq Reporter (Illumina Inc.). The resultant sequences were demultiplexed, quality-filtered, clustered and taxonomically-assigned (at 97% similarity level against GreenGenes database, May 2013 version) with Ribosomal Database Project (RDP)-classifier workflow as described by Wang et al. 82 using QIIME software suite 83 as described previously 76,78 . Quality filtering was performed as described previously 84 . Subsequent data processing included filtering by clustering similar sequences with less than 3% dissimilarity by using the USEARCH algorithm version 5.2.32 85 with an open-reference OTU clustering method using the Greengenes database and detecting and removing chimeras by using the UCHIME algorithm 86 . The most abundant sequence in each OTU was selected as the representative sequence and the resulting OTUs were assigned to taxa by using the Ribosomal Database Project classifier 82 trained on the Greengenes reference database. A total of 2,183,494 sequences (mean ± SEM: 77,981.929 ± 7707.829) were obtained after quality-filtering. To avoid the bias of sequencing depth, the sequences were rarefied to the minimum sequence reads per sample for downstream analyses. To avoid the bias of sequencing errors or low-level contaminations, the OTUs with very low read count (less than 4) in very few samples (less than 10% prevalence) were filtered out from the subsequent analyses. The data of taxon abundance were subjected to the total sum scaling, and the taxa with less than 1% mean relative abundance were further excluded from the subsequent downstream analyses. To avoid the bias of nucleic acid extraction, PCR reaction conditions, and primers on community composition obtained by amplicon sequencing, all the samples were batch-processed identically and simultaneously. Bacterial community composition was measured at taxonomic levels of phylum, class, order, family, and genus. Alpha-diversity measures were calculated within QIIME. Beta-diversity was computed using principal coordinate analysis (PCoA) of the Bray-Curtis dissimilarity index, as described previously 87  Prevalence of antibiotic resistance genes in raw shrimp gut. Prevalence of antibiotic resistance genes in shrimp gut was determined using DNA isolated from shrimp gut and real-time PCR assays. Presence of antibiotic resistance gene (i.e., bla sul-1 , bla sul-2 , bla qnrS , bla qnrD , bla aadA , bla TEM , bla SHV , bla CTX-M , bla CMY , bla KPC , bla NDM , bla mecA , bla vanA , bla vanB , bla vanC , bla cat , bla cmlA , bla floR , and bla tetA ) were tested as using primer-pair and PCR conditions as previously described 55,68,75,90 . Data analysis. Alpha-diversity indices and bacterial abundance data between the two groups of shrimps were compared by two-tailed unpaired t-test. The data of the prevalence of antibiotic resistance and the associated genes were compared with Fisher's exact test. LEfSe (Linear discriminatory analysis [LDA] Effect Size) was applied to identify the bacterial taxa driving the differences between the different groups of shrimps 91 , wherein the alpha parameter significance threshold for the Kruskal-Wallis as well as the Wilcoxon test implemented among classes was set to 0.01, and the logarithmic LDA score cut-off was set to 3.0. Bray-Curtis similarity scores inferred from the taxonomic data were reduced to a two-dimensional space using principal coordinate analysis (PCoA) for the computation of structural similarity (beta-diversity) of bacteriomes from different groups of shrimps. Differences in beta-diversity were computed by permutational multivariate analysis of variance (PER-MANOVA) using the algorithm of web-based tool Microbiome Analyst 92 . Hierarchical clustering heatmaps were constructed within 'R' statistical software package (version 3.6.1; https ://www.r-proje ct.org/) using the 'heatmap.2' and "ggplots" packages. Unless otherwise stated, the data are presented as means ± SEM. P < 0.05 was considered statistically significant unless specified.

Data availability
All data generated or analyzed during this study are included in this published article (and its Supplementary  Information files). The raw sequences have been deposited at the Sequence Read Archive (SRA) under the SRA number (SUB7840299) and BioProject number PRJNA648917 for unrestricted, public access. Other information/ data related to the current study are available from the corresponding author on reasonable request.