A total of 226 Salmonella isolates were used in this study, and all the non-susceptible strains are listed in Supplementary Table S1, of which 39 were food isolates and 39 were clinical isolates. Among these Salmonella isolates, clinical isolates were collected by Shanghai Municipal Center for Disease Control and Prevention and Wuhan Municipal Center for Disease Control and Prevention, while food isolates were collected from beef, poultry, pork, shrimp, vegetables, fresh juice, and shellfish. PCR-based serotyping was performed as previously described (Ranieri et al., 2013). Two different primer pairs and an additional primer pair were used in this study (Supplementary Table S2). The detailed information of these isolates is listed in Supplementary Table S1.

All 226 Salmonella isolates underwent antimicrobial susceptibility testing using the disk diffusion method against a panel of 18 antibiotics, according to the standards and guidelines recommended by the Clinical and Laboratory Standards Institute (CLSI; CLSI, 2013). A total of 18 antibiotic disks (Oxoid Ltd., Basingstoke, UK) that included ampicillin (AMP, 10 μg), piperacillin/tazobactam (TZP, 100/10 μg), ampicillin/sulbactam (SAM, 10/10 μg), ceftriaxone (CRO, 30 μg), ceftazidime (CAZ, 30 μg), cefepime (FEP, 30 μg), cefotetan (CTT, 30 μg), aztreonam (ATM, 30 μg), cephazolin (CZO, 30 μg), ciprofloxacin (CIP, 5 μg), imipenem (IPM, 10 μg), amikacin (AMK, 30 μg), gentamicin (GEN, 10 μg), tobramycin (TOB, 10 μg), ertapenem (ETP, 10 μg), levofloxacin (LEV, 5 μg), nitrofurantoin (NIT, 300 μg), and sulfamethoxazole/trimethoprim (SXT, 23.75/1.25 μg), were assessed. Escherichia coli ATCC 25922 was used as the control strain. Isolates were defined as MDR if they were resistant to at least three different classes of antibiotics.

Genomic DNA of Salmonella isolates was purified by the cetyltrimethylammonium bromide (CTAB) method as described elsewhere (Wilson, 2001). The antibiotic-resistant isolates identified by antimicrobial susceptibility testing were screened for the presence of PMQR determinants (qnrA, qnrB, qnrS, qnrC, qnrD, aac (6′)-Ib-cr) and β-lactamase genes (bla_TEM-1, bla_OXA-1, bla_PSE-1, and bla_{CMY -2}) by simplex PCR amplification. Primers targeting above antibiotic resistance genes were listed in Supplementary Table S3.

Plasmid incompatibility (Inc) groups were assigned by PBRT using genomic DNA of Salmonella isolates as template. Amplification by PCR was performed with eighteen specific primer pairs designed for FIA, FIB, FIC, HI1, HI2, I1, L/M, N, P, W, T, A/C, K, B/O, X, Y, F, and FIIA basic replicons, and the corresponding protocol was previously described (Carattoli et al., 2005). Resulting amplicons were sequenced by Shanghai Majorbio Bio-pharm Technology Co., Ltd. Comparative analysis of nucleotide sequences was performed using the BLAST program at the National Center for Biotechnology Information (NCBI) site¹.

To determine if the antibiotic resistance observed in this study was plasmid-mediated, the identified antibiotic-resistant Salmonella isolates were subjected to a curing treatment using sodium dodecyl sulphate (SDS), according to Tomoeda et al. (1968) with modifications. Briefly, the overnight culture in Luria–Bertani (LB) broth was inoculated (1:10⁴ dilution) in LB broth containing 5% SDS and incubated for 48 h at 44.5°C with constant agitation. Aliquots of the cultures were then appropriately diluted and spread on LB agar plates without antibiotics and incubated at 37°C for 16–18 h. The resulting single clones were streaked onto two LB agar plates with or without ampicillin. Because ampicillin was the most prevalent among antibiotic-resistant isolates, for the screening convenience of those plasmid-loss mutants, ampicillin was used as the priority selective marker. The LB agar plates were incubated at 37°C overnight, and ampicillin-susceptible clones were considered positive for loss of the incompatible plasmid. After treatment with the curing agent, the putative cured derivatives were examined for the antibiotic susceptibility profile using the same set of antibiotics, and for the presence of related antibiotic resistance genes as well as plasmid replicon type by PCR method as described above.

The Salmonella isolate SJTUF 10584 identified in this study was chosen as the plasmid donor for its complete loss of plasmids and MDR after one single trial of plasmid curing. The plasmid was obtained through conjugation experiments by the liquid mating assay using rifampin-resistant E. coli Nk5449 as the recipient strain as previously described (Zurfluh et al., 2014). Transconjugants were selected on LB agar plates containing ampicillin (100 μg/ml) and rifampin (200 μg/ml). Antimicrobial susceptibility tests and antibiotic resistance gene detection were performed to confirm the plasmid transfer, followed by PBRT to test which plasmid and antibiotic resistance markers were transferred.

Plasmid DNA was purified from the transconjugant by the Qiagen Plasmid Midi Kit (Qiagen, Germany) according to the manufacturer’s protocol. Plasmid sequencing was performed on the IIIumina Hiseq sequencing platform at the Shanghai Biotechnology Corporation (Shanghai, China). The complete sequence was annotated using the NCBI Prokaryotic Genome Annotation Pipeline (PGAP), followed by manual inspection. Each predicted protein was further confirmed against the NCBI all-protein database using BLASTP¹. Insertion sequences and repetitive elements were identified using IS finder². DNA sequence comparisons and alignments were performed using BLAST¹ and Geneious software (version 9.1.2; Kearse et al., 2012), and a schematic plasmid map was constructed with WinPlas 2.7 software.

The complete sequence of the plasmid pS10584 has been deposited in Genbank/EMBL/DDBJ with accession number KX058576.

Among 226 Salmonella isolates, 14.2% (32/226) demonstrated resistance to at least one antibiotic and 11.1% (25/226) were resistant to ≥3 antibiotics. In addition, 4.9% (11/226) showed MDR phenotypes. It’s noteworthy that the strain SJTUF 10702 isolated from chicken exhibited strong resistance to 11 antibiotics. In terms of 32 antibiotic-resistant isolates (Figure 1), 25.0% (8/32) presented the quinolone resistance phenotype (LEV or CIP) while 96.9% (31/32) were resistant to the five tested β-lactams (AMP, SAM, CZO, CRO, and CAZ) alone or in combination. Moreover, resistance to individual agents was most frequently observed to AMP (31/32, 96.9%) and SAM (26/32, 81.3%), followed by TOB (14/32, 43.8%) and SXT (11/32, 34.4%). No resistance was detected to CTT, FEP, IPM, and ETP.

Amongst the 32 antibiotic-resistant Salmonella isolates, the PMQR genes aac(6′)-Ib-cr and qnrA were present alone or in combination in 25 (78.1%) and 5 (15.6%) isolates, respectively; three isolates harbored both two determinants (Figure 1). The qnrB, qnrS, qnrC, and qnrD genes were not detected in any of the tested isolates. In terms of β-lactamase genes detected, 75.0% (24/32) were positive for bla_OXA-1, 37.5% (12/32) for bla_TEM-1 and 9.4% (3/32) for bla_{CMY -2}. In addition, seven isolates simultaneously carried bla_OXA-1 and bla_TEM-1, and two carried bla_OXA-1 and bla_{CMY -2}, while one isolate harbored all these three determinants. No isolates contained bla_PSE-1. Of these 27 PMQR-bearing isolates, 23 isolates (85.2%) also carried at least one β-lactamase gene with the prevalent pattern of aac(6′)-Ib-cr-bla_OXA-1 (12/23, 52.2%) and aac(6′)-Ib-cr-bla_OXA-1-bla_TEM-1 (4/23, 17.4%).

Plasmid replicons detected among 32 antibiotic-resistant Salmonella isolates are listed in Figure 1. Overall, seven different replicons of the 18 classic replicon types associated with Enterobacteriaceae were identified in all the isolates except SJTUF 10023 and SJTUF 10054. The two untypeable isolates may possess divergent or novel replicons (Carattoli, 2009). PBRT demonstrated that IncHI2 (19/32, 59.4%) was the predominant type while the other Inc groups such as P (7/32, 21.9%), A/C (4/32, 12.5%), I1 (3/32, 9.4%), FII_S (2/32, 6.25%), FIA (1/32, 3.1%), and FIB (1/32, 3.1%) were just present in few isolates. Seven isolates were mixed of two replicon types, including the combinations of HI2 and I1 (3/7, 42.9%), HI2 and P (2/7, 28.6%), HI2 and A/C (1/7, 14.3%), and FIA and FIB (1/7, 14.3%). Inc groups HI2, I1, A/C, P, FIA, and FIB were found to be associated with MDR isolates, in agreement with previous studies (Poole et al., 2009; Glenn et al., 2013). In addition, IncHI2 was the dominant replicon type both among MDR isolates (7/11, 63.6%) and Salmonella Typhimurium isolates (15/17, 88.24%). When comparing multiple antibiotic resistance genes with replicon types, 86.7% (19/22) of bla_OXA-1 positive isolates as well as 81.8% (9/11) of bla_OXA-1-aac(6′)-Ib-cr positive isolates were carrying IncHI2 plasmids. And all the three bla_{CMY -2} positive isolates contained IncHI2 and IncI1 plasmids simultaneously. In addition, bla_TEM-1 positive isolates tended to carry the IncHI2 (5/12, 41.7%) and IncP (6/12, 50.0%) plasmids while qnrA positive isolates were likely apt to carry IncA/C plasmids (4/5, 80.0%).

All the 32 antibiotic-resistant Salmonella isolates underwent plasmid curing using 5% SDS as well as elevated temperature (44.5°C) for 48 h to ensure that the antibiotic resistance profile was related to the plasmid. Fifteen phenotypic variants were obtained after up to 10 trials of curing (Figure 2), and the curing efficiency was 46.9% (15/32). The cured plasmids mainly belonged to IncHI2 group (7/8, 87.5%), followed by IncF group (3/3, 100%), and IncI1 group (1/1, 100%); however, no strain showed curing of IncP group plasmids (0/5, 0%). When attributing the loss of resistance phenotypes to the loss of plasmids, we could infer that the antibiotic resistant phenotypes AMP-SAM-TOB of SJTUF 10236, AMP-SAM-GEN-TOB-LEV-SXT of SJTUF 10476, AMP-SAM of SJTUF 10479, AMP-SAM-GEN-TOB-CZO of SJTUF 10565, AMP of SJTUF 10577, and AMP-SAM-TOB-SXT of SJTUF 10580 were all related to IncHI2 plasmids while AMP-SAM-CZO-CAZ-CRO-CIP-SXT of SJTUF 10584 was determined together by IncI1 and IncHI2 plasmids. And the resistant phenotype AMP-SAM-CZO-GEN-CIP-LEV of SJTUF 10713 and AMP-GEN-TOB of SJTUF 10718 were both determined by IncF plasmids. In addition, the elimination of IncHI2 plasmids correlated with the loss of β-lactamase genes (bla_OXA-1 and bla_TEM-1) as well as PMQR genes (qnrA and aac(6′)-Ib-cr), and the loss of IncFIA and FIB plasmids resulted in the absence of the aac(6′)-Ib-cr gene. Furthermore, four cured variants (SJTUF 10211^C, 10700^C, 10702^C and 10705^C) lost antibiotic resistance genes as well as drug resistance phenotypes but still retained the identified plasmids (Figure 2). This may be due to the loss of divergent or novel replicons which were untypeable by the current PBRT scheme. The cured SJTUF 10577^C lost the AMP resistance phenotype, bla genes, and the IncHI2 plasmid simultaneously, even though it retained the IncP plasmid. The function of IncP plasmids in antibiotic resistance needs further study.

Particular plasmids belonging to the Inc families usually carry multiple physically linked genetic determinants, playing a major role in the diffusion of specific antibiotic resistance genes among different bacteria (Carattoli, 2013; Wang et al., 2013). In this study, we characterized IncHI2 plasmids as the main vehicle in horizontal transfer of β-lactamase genes (bla_OXA-1 and bla_TEM-1) as well as PMQR genes (qnrA and acc(6′)-ib-cr). The co-existence of aac(6′)-Ib-cr with bla_OXA-1, with or without bla_TEM-1 was also identified. IncHI2 plasmids are known to play a significant role in the acquisition of antibiotic resistance, and they have recently been implicated in the dissemination of genes encoding extended-spectrum β-lactamase (Cain and Hall, 2012). In contrast, IncA/C plasmids are prevalent in several MDR Salmonella serovars and closely linked to the expansion of MDR in the United States (Folster et al., 2010, 2015). We also discovered the high prevalence of IncHI2 plasmids in S. Typhimurium, similar to the report from Li et al. (2013). The identical genetic context of antibiotic resistance genes on IncHI2 plasmids was also detected in S. Indiana in China (Lai et al., 2013) as well as S. Typhimurium in Europe (Campos et al., 2015), suggesting a similar evolutionary origin and highlighting the potentially global spread of IncHI2 plasmids among Salmonella. Furthermore, a similar genetic context of aac(6′)-Ib-cr with bla_OXA-1 was also identified on an IncR plasmid of a K. oxytoca strain in Spain (Ruiz et al., 2011), and an IncN plasmid of an E. coli isolate in Hong Kong (Ho et al., 2013). The similar resistance gene module infers that the transfer of resistance-associated module was driven by mobile genetic elements like integrons, insertion sequences or transposons between different plasmid replicons, and was probably mediated by IS26 in terms of the above cases (Miriagou et al., 2005; Li et al., 2013; He et al., 2015). Thus, further active surveillance is needed to minimize the spread of particular plasmids such as IncHI2 group to control the dissemination of antibiotic resistance. As mobile genetic elements play a major role in the acquisition and dissemination of antibiotic resistance genes, further study may be required to determine the distinct genetic environment of antibiotic resistance genes on plasmids.

The MDR Salmonella enterica serovar Indiana isolate SJTUF 10584 identified in this study was chosen as the plasmid donor. This isolate contains IncHI2 and IncI1 plasmids simultaneously, and these two types of plasmids were easily eliminated after one trial of curing. The plasmid designated as pS10584 was acquired through conjugation experiments recovered from ampicillin and rifampin resistance selection. Replicon typing showed that the only transferred plasmid pS10584 belonged to the IncI1 group and PCR analysis confirmed that it was positive for bla_{CMY -2} gene but not for bla_OXA-1 or aac(6′)-Ib-cr genes. All of the transconjugants shared the same antibiotic resistance pattern of AMP-SAM-CZO-CAZ-CRO.

Complete DNA sequence of pS10584 revealed that it was a circular molecule of 94,697-bp with an average G+C content of 50.0% and 110 putative CDSs. In general, pS10584 possessed a plasmid scaffold typical for IncI1 plasmids, with important IncI1-associated genetic modules including traABCD regulatory cluster, traLMNOPQRSTUVWXY transfer region, nikB-trbABC region, ssb-psiAB region, and type IV thin pilus formation region pilIJKLMNOPQRSTUV (Figure 3). Besides the approximate 89,866-bp “core genome” region encoding characteristic IncI1 plasmid replication, transfer, maintenance and stability functions, a 4,831-bp accessory module encoding the antibiotic resistance gene as well as flanking insertion sequences was also discovered. BLASTN comparison revealed that the overall genetic organization of pS10584 was very similar to that of another fully sequenced IncI1 plasmid pJIE512b (GenBank accession no. HG970648; Tagg et al., 2014) with high sequence identity (>99% at nucleotide level) except an inversion within the pilV shufflon and a 2,307-bp deletion (nucleotide 6,685^th to 8,991^nd of pS10584, containing a truncated colicin gene cia and an intact putative DNA-binding transcriptional regulator gene yagA) in pJIE5112b.

In pS10584, the 4,831-bp antibiotic resistance-associated module comprised 161-bp of the right end of ISEcp1 truncated by IS1294 in the reverse orientation and a 2,823-bp region originating from the Citrobacter freundii chromosome including a bla_{CMY -2} gene, with blc (an outer membrane lipoprotein, lipocalin), sugE (a small MDR transporter), and ΔecnR (LuxR family transcriptional regulator) genes immediately downstream, followed by a 159-bp fragment of IncA/C ending in GTTC which matched the last 4-bp of terIS of IS1294 (Figure 4). In addition, the 4831-bp bla_{CMY -2} segment inserted within the cia gene at the 698-bp from the 5′ end without causing adjacent deletion, followed by the remaining 1,184-bp segment of the cia gene.

A similar genetic context surrounding the bla_{CMY -2} gene was identified in an IncI1 plasmid p2735 (KP017243; Yassine et al., 2014) isolated from Klebsiella pneumoniae as well as six IncI1 plasmids isolated from E. coli including two completely sequenced plasmids pJIE512b (HG970648; Tagg et al., 2014), and pR7AC (KF434766; Minoia et al., 2012), and four partially sequenced plasmids pTN38148 (FM246883), pTN10, pTN13, and pTN386 (Verdet et al., 2009) (Figure 4). Compared to pS10584, the bla_{CMY -2} insertion was within the yagA gene 16-bp from the 5′ end, removing the rest of the yagA gene, as well as a large part of the cia gene, leaving the same part of the cia gene in the 3′ end followed by the same 159-bp IncA/C backbone ending in GTTC in pJIE512b. Whereas, the flanking fragments of bla_{CMY -2} gene in other six IncI1 plasmids (pTN10, pTN13, pTN386, pTN38148, p2735, and pR7AC) were much more different from that in pS10584, including an IS1294-like element designated as IS1294b (Yassine et al., 2014), followed by a 372-bp truncated ISEcp1, the identical 2,823-bp region of the C. freundii chromosome that contained bla_{CMY -2} and the additional 1395-bp IncA/C backbone also ending in GTTC (Figure 4). This bla_{CMY -2} insertion was downstream of the repZ gene, removing a larger adjacent part and truncating the 3′ end of yafA followed by yafB.

Although S. Indiana frequently exhibited MDR phenotypes, very little research has been done to clarify the molecular mechanisms conferring its MDR or the transferability of the resistance properties (Lai et al., 2013). Among ampC-like genes, the bla_{CMY -2} gene which probably originated from C. freundii, is the most prevalent gene worldwide conferring resistance to clinically important β-lactam-based compounds particularly the third-generation cephalosporins. (Verdet et al., 2009; Call et al., 2010; Karczmarczyk et al., 2014; Tagg et al., 2014). In this study, two bla_{CMY -2}-positive isolates both showed resistance phenotypes against the third-generation cephalosporins (CRO, CAZ, and CZO) while the other one was susceptible to them, probably due to the low-level expression or silence of this gene. In previous research, it was observed that resistance to a second generation cephalosporin, cefoxitin in Salmonella isolates harboring bla_{CMY -2} (Giles et al., 2004; Martin et al., 2012; Noda et al., 2015). However, none of the three bla_{CMY -2}-positive isolates was resistant to cefotetan, which is a second-generation cephalosporin like cefoxitin. This was also found in a previous study (Noda et al., 2015). This may also be attributed to the low-level expression or silence of bla_{CMY -2} gene.

Among various plasmid types associated with the carriage of the bla_{CMY -2} gene, IncI1 and IncA/C plasmids are often reported as the predominant carriers (Mata et al., 2012; Bortolaia et al., 2014). Additional research discovered that bla_{CMY -2} carrying IncA/C plasmids always conferred additional antimicrobial resistance phenotypes such as chloramphenicol and sulfisoxazole, whereas in this study, bla_{CMY -2} carrying IncI1 plasmids only conferred a bla_{CMY -2}-associated phenotype, which was also observed in Folster et al. (2011). The bla_{CMY -2} genetic context on the IncI1 plasmid identified in this study was IS1294-ΔISEcp1-bla_{CMY -2}-blc-sugE-ΔecnR followed by 159-bp IncA/C backbone, which was identical to that of pJIE512b from an E. coli isolate (Tagg et al., 2014). The genetic arrangement ISEcp1-bla_{CMY -2}-blc-sugE-ecnR was conserved on IncA/C plasmids (Verdet et al., 2009; Ye et al., 2016). ISEcp1 was presumed to be responsible for the mobilization of bla_{CMY -2} (Verdet et al., 2009) or bla_CTX-M-2 (Lartigue et al., 2006) as well as its adjacent region from the bacterial chromosome to the IncA/C plasmids by a one-ended mobilization mechanism. On the basis of sequence alignments, it was suggested that mobilization of bla_{CMY -2} from the IncA/C plasmid to the IncI1 replicon occurred through IS1294-mediated transposition activity (Tagg et al., 2014). The bla_{CMY -2} insertion region of pS10584 started in CTTG and ended in GTTC, both of which matched the last 4-bp of oriIS and terIS of IS1294 (Tavakoli et al., 2000; Yassine et al., 2014), implying that IS1294 inserted into ISEcp1 in an IncA/C plasmid and subsequently mobilized the adjacent region, including the bla_{CMY -2} segment plus a 159-bp IncA/C backbone (Tagg et al., 2014). A similar bla_{CMY -2} context was also observed in six reported IncI1 plasmids p2735 (Yassine et al., 2014), pR7AC (Minoia et al., 2012), pTN38148, TN10, TN13, and TN386 (Verdet et al., 2009); however, in these plasmids the bla_{CMY -2} insertion segment varied in the length of IncA/C backbone and the position that ISEcp1 interrupted by an IS1294-like element designated as IS1294b. And the integration of the bla_{CMY -2} gene by IS1294b between IncA/C and IncI1 replicons was experimentally confirmed by Yassine et al. (2014). By comparing the integration sites of bla_{CMY -2} insertion segments on these reported IncI1 plasmids, we inferred that the downstream of repZ gene could represent a “hot spot” for the integration of IS1294b element in the IncI1 plasmid scaffold but determination of the “hot spot” for IS1294 may require further study.

Besides the resistance determinants IncI1 plasmids encode, additional factors may be responsible for their prevalent diffusion. The shufflon multiple inversion system on IncI1 plasmids involves four segments separated by seven inverted repeats and a shufflon-specific recombinase (Rci) which acts to catalyze recombination between these repeats (Sampei et al., 2010). This shufflon system is responsible for generating variation in the C terminus of the pilV tip adhesins of the pili, leading to the specificity of recipient cells in liquid mating (Tagg et al., 2014). The resulting type IV pilus is also a known virulence factor, contributing to adhesion and invasion of Shiga toxin-producing E. coli (Kim and Komano, 1997), which accounts for the prevalent existence of IncI1 plasmids in pathogenic E. coli (Johnson et al., 2007). Thus, the association of epidemic ability as well as resistance determinants may favor the wide dissemination of plasmids belonging to the IncI1 group (Carattoli, 2009; Folster et al., 2010).