Salmonella clinical isolates were isolated from patients’ specimens such as blood, feces, synovial fluid and pus from abdominal and skin and soft tissue infections in the First People’s Hospital of Hangzhou, Zhejiang Province, China, between 2009 and 2018. Bacterial species were identified by the automated Vitek 2 system (BioMérieux, Marcy-l’Étoile, France) and matrix-assisted laser desorption ionization time of flight mass spectrometry (MALDI-TOF MS) (Bruker, Bremen, Germany). Salmonella serotyping was identified by slide agglutination with specific antisera (Tianrun Bio-Pharmaceutical Co., Ltd., Ningbo, China) according to the White-Kauffmann-Le Minor scheme (9th edition).

All Salmonella isolates were screened for mcr-2 to mcr-10 by using PCR with corresponding pairs of primers (Table 1). The amplification products were subsequently sequenced by Sanger sequencing for confirmation.

Antimicrobial susceptibility testing was performed by broth microdilution, including colistin, ampicillin, amoxicillin, piperacillin-tazobactam, cefazolin, cefoxitin, ceftriaxone, cefepime, ceftazidime, aztreonam, ertapenem, imipenem, meropenem, amikacin, gentamicin, kanamycin, ciprofloxacin, levofloxacin, tigecycline, tetracycline, trimethoprim-sulfamethoxazole, and fosfomycin. The minimum inhibitory concentration (MIC) of nitrofurantoin was performed using E-test method. Antimicrobial susceptibility testing of the transconjugants was performed by broth microdilution, including colistin, ampicillin, amoxicillin, piperacillin-tazobactam, amikacin, gentamicin, kanamycin, and tetracycline.

The results of antimicrobial susceptibility testing were interpreted by Clinical and Laboratory Standards Institute guidelines (CLSI) (M100, 30th ed.; CLSI, 2020), except that colistin and tigecycline were used the European Committee on Antimicrobial Susceptibility Testing (EUCAST) breakpoints v8.1. The quality control strain was E. coli ATCC 25922.

Genomic DNA of two mcr-9-positive strains S15 and S639 were sequenced by HiSeq (Illumina, San Diego, CA, United States) and MinION sequencer (Oxford Nanopore Technologies, Oxford, United Kingdom). The short read and long read sequence data were hybrid de novo assembled by Unicycler v0.4.8 (Wick et al., 2017). The gene sequences were annotated by Prokka (Seemann, 2014) and NCBI Blast (Camacho et al., 2009). Resistance genes and insertion sequence (IS) were identified by BacAnt (Hua et al., 2021). Multi-locus sequence typing (MLST) was identified by using mlst.¹ The gene sequences were compared and visualized by Easyfig 2.2.5 (Sullivan et al., 2011) and BRIG-0.95 (Alikhan et al., 2011).

Conjugation assays were conducted by using rifampicin-resistant Salmonella strain XH1984 and E. coli strain EC600 as the recipient strain and strain S15 as the donor. The Mueller-Hinton agar plates containing rifampicin (100 μg/ml) and ampicillin (4 μg/ml for S15 and XH1984; 32 μg/ml for S15 and EC600) were used for selection. The successful transconjugants of S15 and XH1984 were verified by PCR using two pairs of primers: XH104-F and XH104-R; S15-mcr9.1-F and S15-mcr9.1-R (Table 1). The former pair of primers was used to identify XH1984 and the latter was used to identify pS15. The transconjugants of S15 and EC600 were verified by S15-mcr9.1 primers and MALDI-TOF MS. The conjugation frequency of pS15 was determined.

The assembled mcr-9 carrying S. Thompson and S. Typhimurium genomes were downloaded from NCBI and annotated using prokka 1.13 (Seemann, 2014). The maximum likelihood phylogenetic tree was constructed with IQTree 2.1.2 (Nguyen et al., 2015) from a multiple alignment of the core genomes generated by Roary 3.7.0 (Page et al., 2015). The trees were visualized with ggtree (Yu, 2020) and ggtreeExtra (Xu et al., 2021) in R.

We previously screened 689 clinical Salmonella isolates from hospital patient specimens for the presence of the colistin resistance gene mcr-1 (Fan et al., 2020). This follow-up study, we have screened all isolates for other types of mcr genes, including mcr-2 to mcr-10. While none of the strains contained any mcr-2 to mcr-8 or mcr-10 genes, we found two (0.29%) mcr-9-positive Salmonella spp. strains, S15 and S639. The strain S15 was isolated from the stool of a 53-year-old female patient in 2011, while S639 was a stool sample isolate from a 24-year-old woman obtained in 2018. Both patients came to the outpatient service with symptoms of diarrhea.

The antimicrobial susceptibility results are displayed in Table 2. Two mcr-9-positive strains were both resistant to ampicillin, amoxicillin, and tetracycline. S639 was additionally resistant to cefazolin, cefoxitin, ceftriaxone, ceftazidime, aztreonam, amikacin, kanamycin, ciprofloxacin, levofloxacin, and trimethoprim-sulfamethoxazole. Both strains were sensitive to colistin, cefepime, ertapenem, imipenem, meropenem, gentamicin, tigecycline, and nitrofurantoin.

The serotype of S15 was Salmonella Typhimurium (O4:Hi), belonging to ST 34. The strain contained a single plasmid only which we called pS15. The plasmid encoded the mcr-9 gene which showed a 100% identity and coverage to the previously reported mcr-9 in Enterobacterales (WP_044704969.1). The genome size of IncHI2 plasmid pS15 was 266,098 bp and the GC content was 46%. This strain contains 12 resistance genes on the plasmid and 20 resistance genes on the chromosome (Table 3).

The serotype of S639 was Salmonella Thompson (O7:Hk:H1,5), belonging to ST 26. Again, this strain contained only a single plasmid which we called pS639. Here, the mcr-9 exhibited 100% identity and coverage to another colistin resistance gene in the Enterobacterales, WP_001572373, mcr-9.1. This sequence was missing a single codon (for tryptophan) right before the STOP codon compared to mcr-9 in S15. The genome size of IncHI2 plasmid pS639 was 308,491 bp and the GC content was 48%. This strain contains 27 resistance genes on the plasmid while only four resistance genes are found on the chromosome (Table 3).

Most of the chromosome-encoded resistance genes in S15 and S639 are related to metal resistance aside from the multidrug efflux RND transporter gene mdsB. The resistance genes in two plasmids differ greatly from each other: for example, pS15 contains genes associated with resistance to tellurium (terW and terZ), mercury (mer gene cluster), β-lactams (bla_TEM-1), tetracycline [tet(A)], trimethoprim (dfrA16), streptomycin (aadA2), colistin (mcr-9), and copper (pcoS and pcoE). Plasmid pS639 encodes genes mediating resistance to all substances which are facilitating resistance in pS15, some of which were different alleles like dfrA19 and tet(D). Additionally, pS639 encodes genes associated with resistance to sulfonamide (sul1 and sul2), macrolide [mph(A)], β-lactams (bla_OXA-10), aminoglycoside (aacA34), rifamycin (arr-3), streptomycin [aph(6)-ld and aph(3″)-lb], chloramphenicol (catA2), cephalosporin (bla_DHA-1), and quinolone (qnrB4). This abundance of additional ARGs might explain why S639 exhibited higher MIC values and a wider resistance to more antibiotics than S15.

The two plasmids we found are similar to other IncHI2 plasmids. Table 4 lists plasmids for comparison, some of which share high query coverage and identity with pS15 and pS639 from different species. The backbone structures of pS15, pEcl10-1 (CP048704), sLN794248 (LN794248), and pC45-VIM4 (LT991958) are closely related (Figure 1). The main differences between them are the resistance gene cluster regions where most insertion elements (ISs) were located. Although pEcl10-1 and sLN794248 were more similar to pS15 in their sequence, they do not contain mcr-9.1 and dfrA16. Plasmid pC45-VIM4 encodes a mcr-9.1 gene but shows more differences in the 90–140 kbp region of pS15 compared to other plasmids. When comparing pS15 to three other plasmids, pS15 contains a gene encoding group II intron reverse transcriptase/maturase (around 210 kb). Two resistance gene clusters are present in pS639, found in two sections, from ~100 to 180 kb and from ~240 to 270 kb, respectively (Figure 2). The first section shares some similarities but also substantial differences with two other plasmid sequences, p48212_MCR (CP059413) and pMOL665_IncHI2 (OU015720). pS639 additionally encodes mph(A), bla_OXA-10, aacA34, aar-3, and catA2, genes which are able to facilitate the resistance to macrolide, β-lactam, aminoglycoside, rifamycin, and chloramphenicol. In the other resistance gene cluster which includes mcr-9.1, the main difference is an insertion of the two resistance genes bla_DHA-1 and qnrB4 while a gene cluster encoding phage shock protein is also present.

Our genetic analyses regarding the sequences surrounding the mcr-9 genes revealed two types (Figure 3). The mcr-9 surrounding structure of pS639 was pcoE-pcoS-IS903B-mcr-9-WbuC-IS26 similar to p48212_MCR and pMOL665_IncHI2. However, in our case, the genetic context that embedded the mcr-9 gene was pcoE-pcoS-IS903B-mcr-9-IS1 which is present in both, the Salmonella plasmid pS15 and pC45-VIM4, a plasmid found in a bacterium of the Enterobacter cloacae complex. There was an insertion of IS1 and catA in the position of mcr-9 and IS903B.

We also tested the ability of the plasmids to be transferred to other strains and their ability to convey antibiotic resistance. We first tested the plasmid pS15 which was successfully transferred to the rifamycin-resistant Salmonella strain XH1984 and E. coli strain EC600. When determining the MICs for ampicillin, amoxicillin, piperacillin-tazobactam, and tetracycline, we found increased resistance likely conferred by the presence of the plasmid-encoded bla_TEM-1 and tet(A) genes (Table 2). However, mcr-9 in pS15 could not confer colistin resistance in neither Salmonella nor E. coli strains. The conjugation efficiency was calculated in pS15 plasmid conjugation assays, which was 1.9 × 10⁻⁶ transconjugants per donor when pS15 was transferred to XH1984 and 2.1 × 10⁻⁸ transferred to EC600 (Table 5).

Our attempt in transferring the plasmid pS639 was unsuccessful. The reason for this is that we did not have a suitable recipient strain available which would allow the use of an antibiotic selection marker, as our strains exhibited resistance to the antibiotics encoded on the plasmid. Also, the possibility that the recipient strains were genetically not suitable to receive this specific plasmid cannot be excluded.

A total of 175 S. Typhimurium strains and 21 S. Thompson strains carrying mcr-9 have been deposited in NCBI till today (February 2022). The phylogenetic trees of two serotypes were displayed in Figures 4, 5, respectively. S. Typhimurium strains were mostly isolated from clinical samples. Australia and the United Kingdom were the countries where the most assembled sequences were uploaded from, which does not necessarily reflect the prevalence of the strains in these countries. S. Typhimurium S15 was most similar to FSIS32003798 isolated from pork in the United States. As for S. Thompson, only three of 21 strains were clinical origin including S639, which had the closest relationship with 813,389 isolated in the United Kingdom. Interestingly, both S15 and S639 were more closely related to strains isolated outside China, despite being isolated within the country.