A total of 89 P. mirabilis strains (62 from chicken meat, 20 from pork meat and 7 from aquatic products) were isolated from 347 samples, and the overall isolation rate of P. mirabilis was 25.65% (89/347). The chicken meat samples had the highest isolation rate, which reached 54.39% (62/114). The isolation rates of P. mirabilis in pork and aquatic products were low, which were 14.18% (20/141) and 7.61% (7/92), respectively. In 14 food markets, the isolation rates of P. mirabilis varied from 4.76% (1/21) to 53.33% (8/15) (Table 1). The detailed information of the 89 P. mirabilis strains reported in this study is listed in Supplementary Table S1.

The mean of the draft genomes of 89 P. mirabilis strains was 4.00 Mb (from 3.79 to 4.32 Mb) in size with average GC content of 38.8%. The draft genomes consisted of an average of 95 (from 49 to 260) contigs above 200 bp (Supplementary Table S1). A total of 83,751 SNPs were identified in the genomes of the 89 P. mirabilis strains, and the phylogenetic tree based on the SNPs was conducted (Figure 1). All 89 strains were divided into 81 different clones (SNPs >5), indicating high genetic diversity of P. mirabilis in food markets. However, four clonal transmission events were also found: (i) four clonal strains JZ5, JZ35, JZ47 and JZ66 (0–1 SNPs) were isolated from pork meat samples in food market F14; (ii) two strains JZ6 and JZ37 (4 SNPs) that harboured bla_DHA-1 came from chicken meat samples in F11; (iii) four strains (JZ4, JZ7, JZ53 and JZ70) were clonally related (0–1 SNPs), in which three from aquatic products in F7, but JZ70 from chicken meat in F5; and (iv) strains JZ26 and JZ117 (5 SNPs) respectively came from chicken meat in F14 and aquatic product in F1, both of which harboured tigecycline resistance gene cluster tmexCD3-toprJ1.

Antimicrobial susceptibility testing showed that the P. mirabilis strains exhibited high resistance rate to SXT (94.38%), STR (93.26%), CHL (86.52%), FFC (83.15%) and AMP (83.15%), and low resistance rate to CAZ (16.85%), IPM (13.48%) and AMK (5.62%). The resistance percentage to other agents was as follows: NAL (79.78%), NOR (69.66%), GEN (49.44%), CRO (49.44%), CIP (42.70%), FOS (30.34%), FOX (26.97%) and AMC (22.47%). It’s worth noting that 81 (91.01%) strains displayed MDR profiles.

A total of 73 different resistance genes (or variants) were found in the 89 P. mirabilis strains (Supplementary Table S1). The prevalence of β-Lactam, quinolone, fosfomycin, aminoglycoside, chloramphenicol, sulfonamide, trimethoprim, tetracycline and lincosamide resistance genes were shown in Figure 2. Intrinsical resistance genes tet(J) (tetracycline) and cat (chloramphenicol) were detected in 89 (100%) and 87 (97.75%) strains, respectively. bla_OXA-1 was the most prevalent β-lactamase genes, and found in 55 (61.80%) strains. Thirty-five strains carried ESBLs gene bla_CTX-M, including 30 bla_CTX-M-65, 3 bla_CTX-M-14, 1 bla_CTX-M-3 and 1 bla_CTX-M-55. AmpC β-lactamases genes bla_DHA-1 and bla_CMY-2 were found in 16 and 3 strains, respectively. More importantly, one strain (JZ109) carried carbapenemase gene bla_NDM-1. Quinolone resistance gene aac(6′)-Ib-cr was found in 69 (77.53%) strains. Nineteen strains carried plasmid-mediated quinolone resistance gene qnrD, including 16 qnrD1, 1 qnrD2 and 2 qnrD3. Fosfomycin resistance gene fosA3 was detected in 27 (30.34%) strains. The prevalence of other resistance genes above 60% was as follows: aminoglycoside resistance genes aph(4)-Ia (75.28%), aac(3)-IVa (74.16%), aadA1 (74.16%), aadA2 (73.03%) and aph(3′)-Ia (70.79%); chloramphenicol resistance genes floR (85.39%) and catB3 (62.92%); sulfonamide resistance genes sul1 (88.76%) and sul2 (88.76%). It’s worth noting that multiresistance gene cfr and tigecycline resistance gene cluster tmexCD3-toprJ1 were found in 4 and 4 strains, respectively.

Genetic environments of clinically important antimicrobial resistance genes cfr, tmexCD3-toprJ1 and bla_NDM-1 were characterized, which showed that those genes were located on diverse genetic elements including Tn7 transposon, plasmid, SXT/R391 ICEs and SGI1 (Figure 3).

cfr gene was located on Tn7 in strains JZ29 and JZ94, pJZ27-cfr in strain JZ27 and ICEPmiChnJZ57 in strain JZ57. Strains JZ29 and JZ94 showed high genetic similarity (18 SNPs) and shared the same genetic environment for cfr. The cfr gene in four strains was flanked by insertion sequence IS26. The Tn7 region in JZ29 chromosome harboured various resistance genes including cfr and bla_DHA-1, and was divided into two parts separated by a chromosomal DNA fragment of 112.54 kb (PMI3068-PMI3171 and truncated PMI3172; Figure 3A). We hypothesized that an initial single IS26 transposition event occurred in JZ29 chromosome in locus PMI3172 and has subsequently generated an inversion between two IS26 elements in opposite orientation, resulting in the right end of Tn7 and this 112.54 kb chromosomal DNA fragment being in inverse orientation. The cfr gene in strain JZ27 was also located on a Tn7 in plasmid pJZ27-cfr. pJZ27-cfr was 132,678 bp in size, and showed 100% nucleotide identity to plasmid pHBRJC7 (MK630213) recovered from P. mirabilis strain of broiler origin in China (Chen et al., 2022). Compared with pHBRJC7, pJZ27-cfr supplemented IS26-cfr-repUS18 region (Figure 3A). The ICEPmiChnJZ57 harbouring cfr was 110,965 bp is size, and showed 99.99% nucleotide identity to ICEPmiChnSTP3 (MT449450) in P. mirabilis strain from swine in China (He et al., 2021). ICEPmiChnJZ57 lost floR in variable region VRIII and aphA1 gene in hot spot HS4 compared with ICEPmiChnSTP3 (Figure 3B). Conjugation experiments indicated that pJZ27-cfr and ICEPmiChnJZ57 that harboured cfr gene could be conjugated into E. coli EC600, and ICEPmiChnJZ57 was chromosomally integrated into the 5′ end of the prfC gene.

tmexCD3-toprJ1 was located on a novel SXT/R391 ICE designated as ICEPmiChnJZ26 according to the SXT/R391 nomenclature (Burrus et al., 2006). ICEPmiChnJZ26 was 105,682 bp in size, and had 61% coverage to tmexCD3-toprJ1-harbouring ICEPciChn180 (CP073356) in Proteus terrae subsp. cibarius strain SDQ8C180-2 T from chicken in China (Wang et al., 2021a). tmexCD3-toprJ1 gene cluster was located in VRIII in both ICEPmiChnJZ26 and ICEPciChn180 associated with ISCfr1 (Figure 3B). Conjugative transfer of ICEPmiChnJZ26 failed despite three independent attempts to recover transconjugants carrying ICEPmiChnJZ26. However, the tmexCD3-toprJ1-harbouring ICE was detected in four strains, three of which were not clonally related, indicating the mobility of tmexCD3-toprJ1-harbouring ICE.

The bla_NDM-1 in JZ109 was located on the chromosomal SGI1 (Figure 3C), which was 40,275 bp in size and showed 100% nucleotide identity (only one base change) to SGI1-1NDM that was firstly reported in clinical P. mirabilis strain PmBJ020-1 (CP065146) from China (Yang et al., 2021). Mobilization assays confirmed that SGI1-1NDM in JZ109 could be successfully transferred to E. coli and incorporated into the 3′ end of trmE with the help of IncC plasmid pR55.

A total of 347 fresh samples (114 chicken meat, 141 pork meat and 92 aquatic products) were purchased from 14 food markets at different locations in Chengdu city, Sichuan province of China from December 2021 to January 2022. The numbers of market stalls selling retail meat and aquatic products were varied from 0 to 20. We collected only one sample from each market stall. The meat samples were stored in sterile bags at 4°C between sampling and further processing. The processing method of P. mirabilis identification was modified from the protocol reported previously (Wong et al., 2013). All culture mediums were purchased from Beijing Land Bridge Technology Co., Beijing, China. Briefly, 25 g of meat or aquatic product was placed in 100 ml of buffered peptone water (BPW) and incubated at 37°C for 24 h. Then, the 1 ml BPW-incubated culture was mixed with 9 ml of Rappaport-Vassiliadis Salmonella enrichment broth and incubated at 42°C for 24 h. One loopful of broth was streaked onto Mueller Hinton Agar and incubated at 37°C for 16 h to check for the motility of the isolates. The colonies with swarming growth phenotype were streaked on Shigella and Salmonella agar to obtain single clone. P. mirabilis isolates were confirmed by using the BD Phoenix™ 100 Automated Microbiology System (Becton Dickinson, United States). Only one isolate from each sample was selected for further study.

Genomic DNA was extracted using a bacterial genomic DNA extraction kit (Tiangen, China). Whole genomes were sequenced using the Illumina HiSeq platform (150 bp paired-end reads with about 200-fold average coverage), and the clear data were assembled to draft genomes using the software SPAdes v3.15.4. Antimicrobial resistance genes were found using ResFinder 4.1.¹ SNPs from genomes of the P. mirabilis strains were called and a phylogeny based on the concatenated alignment of the high-quality SNPs was inferred by using CSI Phylogeny 1.4² with parameters as defaults. A threshold of 5 SNPs among isolates was considered likely to have an clonal relationship (Dallman et al., 2015; Kanagarajah et al., 2018). The phylogenetic tree was modified in Itol.³

Antimicrobial susceptibility testing was performed using the standard Kirby–Bauer disk diffusion method according to Clinical and Laboratory Standards Institutes (CLSI) guidelines (CLSI, 2020). Susceptibilities to 16 antimicrobial agents were tested including ampicillin (AMP), amoxicillin-clavulanate acid (AMC), cefotaxime (CTX), ceftazidime (CAZ), cefoxitin (FOX), imipenem (IPM), nalidixic acid (NAL), norfloxacin (NOR), ciprofloxacin (CIP), streptomycin (STR), gentamicin (GEN), amikacin (AMK), chloramphenicol (CHL), florfenicol (FFC), fosfomycin (FOS), and trimethoprim-sulfamethoxazole (SXT). Escherichia coli ATCC25922 was used as quality control. MDR was defined as resistance to at least one agent in three or more chemical classes of antimicrobials.

The genomes of the P. mirabilis strains harbouring cfr, tmexCD3-toprJ1 and bla_NDM-1 were further sequenced using Nanopore MinION using Rapid Sequencing Kit. The complete genome sequences were assembled by Unicycler v0.5.0 using Nanopore sequencing data combined with Illumina sequencing data. Genetic environments of cfr, tmexCD3-toprJ1 and bla_NDM-1 were analyzed using the BLAST program⁴ and Easyfig v2.2.2.⁵

Conjugation experiments were performed by using P. mirabilis JZ27 (harbouring plasmid pJZ27-cfr), JZ26 (harbouring ICEPmiChnJZ26), and JZ57 (harbouring ICEPmiChnJZ57) as the donor strains and rifampin-resistant E. coli EC600 as the recipient strain with selection on nutrient agar plates containing 300 mg/l rifampin and 8 mg/l florfenicol (for cfr) or 2 mg/l tigecycline (for tmexCD3-toprJ1). The positive transconjugants were further determined by detection of antimicrobial resistance profile and the location of ICEPmiChnJZ26 and ICEPmiChnJZ57 in E. coli with primers LE1/LE4 and RE4/RE1 (McGrath et al., 2006). Mobilization assay for SGI1 harbouring bla_NDM-1 was carried out as previously described (Siebor et al., 2016), using E. coli EC600 harboring an IncC plasmid pR55 as the recipient strain. The transconjugants (EC600 carrying SGI1) were further examined for the presence of the bla_NDM-1 and the location of SGI1 in E. coli (Girlich et al., 2015; Siebor et al., 2016).

The genomes of 89 P. mirabilis strains reported in this study have been deposited in National Center for Biotechnology Information and registered BioProject number PRJNA847328. The nucleotide sequences of ICEPmiChnJZ57, Tn7 in JZ29, pJZ27-cfr, SGI1 in JZ109 and ICEPmiChnJZ26 have been deposited in GenBank under accession numbers ON553412, ON553413, ON553414, ON553415 and ON630400, respectively.