32 P. mirabilis strains were isolated from different clinical samples of patients at the Hejiang County People’s Hospital, in Luzhou City, Sichuan Province of China, from January to December 2021. Written informed consent from the patients was exempted from this study since the present study only focused on bacteria, and the strains were isolated as a part of the routine hospital laboratory procedures. Isolates were initially identified with both the VITEK 2 system (BioMérieux, Marcy-l’Étoile, France) and 16S rRNA gene sequencing analysis (Lane, 1991).

The minimum inhibitory concentrations (MICs) of 23 antimicrobial agents, including ampicillin, amikacin, ampicillin/sulbactam, azlocillin, aztreonam, cefazolin, cefepime, cefoxitin, cefixime, cefotetan, ceftazidime, ceftriaxone, ciprofloxacin, ertapenem, gentamicin, imipenem, meropenem, levofloxacin, nitrofurantoin, mezlocillin, piperacillin/tazobactam, tobramycin, and trimethoprim-sulfamethoxazole were automatically performed by the VITEK 2 system. MICs of ceftazidime, cefotaxime, and meropenem were manually confirmed using the broth microdilution method, and the susceptibility testing for ceftazidime was also performed by Kirby Bauer disk diffusion method, with Escherichia coli strain ATCC 25922 as the quality control. The results of antibiotic susceptibility testing were interpreted by the breakpoints defined by the Clinical and Laboratory Standards Institute standards for Enterobacterales (CLSI, M100) (CLSI, 2023). MDR strains were defined as non-susceptibility to three or more of the following antibiotic groups: β-lactam-β-lactam inhibitor combinations, cephalosporins, aminoglycosides, fluoroquinolones or trimethoprim-sulfamethoxazole (Korytny et al., 2016).

Fourteen MDR P. mirabilis were selected for genomic DNA extraction using the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany) following the manufacturer’s guidelines. Whole genome sequencing was performed on the HiSeq 2000 (Illumina, San Diego, CA, USA) Sequencer with a 150 bp paired-end library and 200 × coverage by the Beijing Tsingke Bioinformatics Technology Co. Ltd. The raw reads were trimmed using Trimmomatic v0.38 before being assembled into draft genomes using SPAdes v3.12.0 program (Bankevich et al., 2012; Bolger et al., 2014). Annotation was carried out using Prokka (Seemann, 2014). Genome-based species identification was performed by average nucleotide identity (ANI) analysis. The ANI value between genome sequences of these clinical isolates and that of reference strain P. mirabilis HI4320 (Accession no. NC_010554) was calculated with JSpeciesWS (Richter et al., 2016). 96% is the cut-off for defining a bacterial species. Plasmid incompatibility types, antibiotic resistance genes (ARGs), and insertion elements (ISs) were predicted using PlasmidFinder (Carattoli and Hasman, 2020), ResFinder (Bortolaia et al., 2020), and ISfinder (Siguier et al., 2006). The detection of SXT/R391 ICE in the whole genome sequences was performed using the conserved integrase gene (int_SXT). The contigs of SXT/R391 ICE were extracted and assembled against the reference ICE of FZP3105 from our previous study (Li et al., 2022), and gaps between contigs were closed by PCR and Sanger sequencing. The contigs of PmGRI1 were extracted and assembled against the reference PmGRI1-CYPM1 (GenBank accession CP012674). Linear sequence alignment was carried out using BLAST and visualized with Easyfig 2.2.3(Sullivan et al., 2011).

The genome sequences of other representative Proteus isolates in China were retrieved from the GenBank. Genetic relationship of different Proteus isolates was assessed based on single nucleotide polymorphisms (SNPs) in their core genomes, as previously described with minor modification (Li et al., 2022). Briefly, the GFF3 files generated by Prokka were piped into Roary to create a core genome alignment. SNPs were extracted using snp-sites v2.3.2 (Kwok and Hitchins, 2015). A maximum-likelihood phylogenetic tree was constructed based on the SNPs using FastTree version 2.1.10 under the GTRGAMMA model with 1000 bootstrap iterations (Price et al., 2010).

Conjugation experiments were carried out using broth-based method with the E. coli strain J53 (sodium azide-resistant) or EC600 (rifampicin-resistant) as the recipient, as described previously with minor modification (Li et al., 2022). After the donor and recipient strains were grown to exponential stage (the optical density at 600 nm reaches ~0.5), mix them at a ratio of 1:1 before incubation at 37°C for 24 h. Transconjugants were selected on Luria-Bertani (LB) agar plates containing 4 μg/ml cefotaxime plus 150 μg/ml sodium azide (for J53) or 200 μg/ml rifampicin (for EC600). The presence of SXT/R391 ICE was confirmed by PCR using the primers targeting int_SXT (Sato et al., 2020).

The presence of genes related to bacterial virulence factors including fimbriae (mrpA, ucaA, pmfA, pmpA), hemolysin (hpmAB), Urease (ureC), biofilm formation (pst, rcsD), autotransporters (pta, aipA), proteases (zapA), and siderophore (nrpR) were identified using the Virulence Factors of Pathogenic Bacteria Database (VFDB) (Chen et al., 2005). The identification of flagella genes (flhA, fliF, fliG, fliP, fliL, flgN, flaD) was performed in a pairwise BLASTn alignment.

Among the 32 P. mirabilis strains, 14 (43.8%) of them were identified as MDR, which were recovered from sputum (n=7, 50%), urine (n=4, 28.6%), stool (n=1, 7.1%), ascites (n=1, 7.1%), and wound secretion (n=1, 7.1%). Results of the antimicrobial susceptibility test indicated that all the MDR strains were resistant to ampicillin, azlocillin, cefazolin, cefixime, mezlocillin, ampicillin/sulbactam, and trimethoprim/sulfamethoxazole, in addition to their intrinsic resistance profiles (Table 1). They also showed high resistance rates (>70.0%) to ceftriaxone (n =12, 85.7%), and a lesser extent to cefotetan (n =8, 57.1%), ciprofloxacin (n=8, 57.1%), and levofloxacin (n=8, 57.1%). Four isolates (28.6%) were resistant to piperacillin/tazobactam. Resistance to antimicrobial agents, including cefepime, cefoxitin, aztreonam, gentamicin, and tobramycin was rarely (n<3) detected. All the MDR strains were sensitive to meropenem (MIC ≤ 1), ertapenem (MIC ≤ 0.5), ceftazidime (MIC ≤ 1), and amikacin (MIC ≤ 2). The Kirby Bauer test confirmed the sensitivity of these strains to ceftazidime. Consistent with the observation that P. mirabilis possesses intrinsic decreased susceptibility to imipenem (but not meropenem and ertapenem), 92.9% of MDR strains (n=13) were resistant to imipenem in this study, which was thought to be caused by weak affinity with penicillin-binding proteins or porin loss (Girlich et al., 2020).

We sequenced the genomes of all 14 MDR isolates on the Illumina platform. The genomes vary from 3,964,688 bp to 4,212,070 bp, with an average GC content of 38.80% to 39.09% (Table S1). The genomic analysis identified the presence of 35 different ARGs in the 14 MDR P. mirabilis strains, and 11 (78.6%) of them carried at least 15 ARGs (Table 1). Three different ESBLs genes, bla_CTX-M-65, bla_CTX-M-63, and bla_CMY-2, were detected. Among them, bla_CTX-M-65, which has been identified in many P. mirabilis strains from humans and animals in China (Li et al., 2022; Qu et al., 2022), also represents the most prevalent one in this study. All the strains except for HJP18031 carried abundant aminoglycosides resistance genes, with aac(3)-IV (n=11), aph(4)-Ia (n=11), aph(6)-Id (n=8), and aph(3’’)-Ib (n=8) being the most prevalent. Of the detected quinolone resistance genes, aac(6’)-Ib-cr (n=10, 71.4%) is the most common, followed by qnrD1 (n=8, 57.1%), and qnrS1 was only detected in one isolate. Furthermore, we found that genotypic and phenotypic resistance in these strains often does not match. Almost all the bla_CTX-M-65-harboring isolates (except for HJP22016) remain sensitive to aztreonam. Also, the presence of aac(6’)-Ib-cr did not lead to quinolone resistance in several strains, such as HJP21048, HJP31010, and HJP31030. These findings suggest that the functions of these ARGs may be regulated by some unknown mechanisms in P. mirabilis strains (Li et al., 2022).

To understand the genetic relationship of 14 MDR P. mirabilis strains in this study with other isolates from different geographic locations in China (Table S2), a core genome-based phylogenetic tree was constructed, which revealed five distinct groups (Figure 1). The tested strains are diffusely distributed in the tree, suggesting the genetic diversity of them. HJP21048 and HJP31006 are present in the subclade II that is distant from other tested strains, suggesting a different origin of them. In subclade IV, four strains HJP26021, HJP31010, HJP31030, and HJP01027 are tightly clustered, and they share identical core genomes (SNP=0) as well as resistance phenotypes, showing the clonal nature of them. Notably, HJP26021, HJP31010, and HJP01027 were subsequent isolates from the sputum of a single patient a few days apart, whereas HJP31030 was from another patient in the same inpatient ward but a different room, indicating the clone spread of this MDR strain. In subclade V, strains from this study cluster with several isolates from animals and humans from different locations in China. Especially, HJP17012 and HJP05004 are closely related (105 SNPs), and tightly cluster together with several clinical isolates and animal-derived strains. These findings suggest the possibility of circulation and clonal spread of these MDR P. mirabilis strains across different regions of China.

Only two types of plasmids, IncQ1 and Col3M, were identified in these P. mirabilis strains and eight of them harbor the 2,683-bp Col3M plasmid, which encodes a quinolone resistance protein QnrD1 and a hypothetical protein. BLASTn analysis showed that the Col3M plasmid is very common in P. mirabilis strains of different origins. Two IncQ1-type replication genes were detected in HJP05004 and HJP16012, respectively. In HJP05004, the IncQ1 plasmid pCTX-M_HJP05004 is 40,486 bp in size and has no other known ARGs except bla_CTX-M-63 (Figure 2A). It consists of a 38-kb backbone comprising genes for replication (repA), maintenance (topB), and conjugation (vir genes). The sole acquired resistance gene region was identified downstream of dnaA (encoding the chromosomal replication initiation ATPase DnaA), wherein bla_CTX-M-63 was present in an ISEcp1-bla_CTX-M-63 transposition unit. The insertion sequence ISEcp1 is known to mobilize adjacent sequences, including bla_CTX-M genes, by using its own left inverted repeat (IRL) in conjunction with alternative sequences resembling its right inverted repeat (IRR) (Zong et al., 2010). BLASTn revealed that pCTX-M_HJP05004 is partly similar (71% coverage, >97.1% nucleotide identity) to p1_FZP3115 (CP098451, P. mirabilis, patient) from China, and pPM64421b (MF150117, P. mirabilis, unknown) from Brazil, indicating global spread of this type of plasmid.

On the chromosome of HJP16012, thirteen ARGs conferring resistance to aminoglycosides (aadA5, aph(6)-Id, aph(3’’)-Ib, aph(3’)-Ia, aac(3)-IIa, aph(6)-Id, aph(3’’)-Ib), β-lactam (bla_TEM-1B), macrolide (mph(A)), amphenicol (catA1), sulphonamides (sul1, sul2), and trimethoprim (dfrA17) are clustered in a 39, 086-bp MDR region, wherein an IncQ1-type replication gene repA is present. The MDR region showed high similarity (98% coverage, 99.56% identity) to the one found in a P. mirabilis strain CY32 (CP118227, chicken, China), and was partly similar (85% coverage, 99.87% identity) to the plasmid p24362-1 (CP051379) recovering from a Salmonella enterica strain from swine in the USA. The result reveals that the MDR regions on the chromosomes of P. mirabilis HJP16012 and CY32 seem to be derived from a p24362-1-like plasmid. Further sequence analysis showed that two copies of IS1 flanked the MDR region in CY32 and HJP16012 (Figure 2B), suggesting a possible role of IS1-mediated composite transposon in the integration process (Partridge et al., 2018).

In recent years, five Salmonella genomic island 1 (SGI1)-relative elements, including SGI1(Ahmed et al., 2006), Proteus genomic island 1 (PGI1)(Siebor and Neuwirth, 2014), Proteus genomic island 2 (PGI2)(Lei et al., 2018a), GIPmi1(Siebor et al., 2018), and PmGRI1(Lei et al., 2020) have been identified in P. mirabilis strains. The presence of SGI-relative elements in our P. mirabilis isolates was screened by searching the integrase gene of SGI1 (KM234279), PGI1(KJ411925), PGI2 (MG201402), GIPmi1 (MF490433), and PmGRI1(MW699445) from the whole sequenced genomes. As a result, five (35.71%, 5/14) strains contained the PmGRI1 integrase gene (Table 1). No other integrase genes of SGI1-relative elements were detected in these strains.

PmGRI1 is a multidrug-resistant GI that has been found in several P. mirabilis isolates of animal and human origins from different locations in China (Ma et al., 2021; Li et al., 2022). Based on the genome data, the complete sequences of PmGRI1 in strain HJP17012 (PmGRI1-17012) was successfully assembled. PmGRI1-17012 is 40,678 bp in size, and catA1 is the only known and complete antimicrobial resistance gene (Figure 3). BLASTn analysis showed that the configuration of PmGRI1-17012 is identical to that of PmGRI1-CYPM1 (CP012674), which was recovered from a clinical P. mirabilis strain from Taiwan, China, in 2012. Due to possible complex structures or high numbers of transposases and ISs, the complete sequences of PmGRI1 in HJP05004, HJP22021, and HJP20004 were fragmented in two or more contigs. PmGRI1-05004 and PmGRI1-20004 are also possible variants of PmGRI1-CYPM1 as revealed by the existing assembly. PmGRI1-22021 seems to have a more complex genetic structure, with a number of genes, including several ARGs, being inserted upstream and downstream of tnpA of Tn21, respectively (Figure 3). Altogether, the presence of PmGRI1 in the tested strains reconfirmed its prevalence in P. mirabilis in China, highlighting that PmGRI1 may serve as an important vehicle in capturing and spreading ARGs.

In addition, we detected the presence of SXT/R391 ICEs by targeting the conserved integrase gene (int_SXT) and found that ten out of the fourteen (71.4%, 10/14) MDR P. mirabilis strains were positive for the int_SXT gene (> 96% identity to that of SXT, AY055428, Table 1). The complete sequences of seven SXT/R391 ICEs were successfully assembled, ranging in size from 91,092 bp to 131,933 bp. Among them, ICEPmiChn26021, ICEPmiChn31010, ICEPmiChn31030, and ICEPmiChn01027 are almost identical, mainly differing by an insertion or deletion of the aph(6)-Id gene (an aminoglycosides resistance gene). BLASTn analysis showed that ICEPmiChn21049 is similar (94% coverage, 99.96% nucleotide identity) to ICEPmiChn-HERJC4 (MZ221994) recovering from P. mirabilis strain HERJC4 from chicken in China. ICEPmiChn20004 is closely related (99% coverage, >99.9% identity) to several ICEs of human and animal origins from different locations in China, such as those in P. mirabilis strain PM8762 (CP092652, patient, 2021), and P. mirabilis strain HBNNC12 (MZ277865, cow, 2018). ICEPmiChn31006 is nearly identical (99% query coverage, 99.95% identity) to the element ICEPmiJpn1 initially described in Japan (Harada et al., 2010) and was later detected in other parts of the world (Sato et al., 2020), confirming the global spread of this element. The four clonally related ICEs, ICEPmiChn26021, ICEPmiChn31010, ICEPmiChn31030, and ICEPmiChn01027, showed the highest similarity (94% coverage, 99.95% nucleotide identity) to ICEPmiChn3105 (Accession no. CP098444), while HJP31010 (or HJP31030, or HJP01027, or HJP26021) is genetically distant from FZP3105 according to the species tree (Figure 1). The results suggest the cryptic dissemination and independent acquisition of MDR ICEs in P. mirabilis strains.

Genetic analysis revealed that these ICEs shared a common backbone structure with most SXT/R391 ICEs (Figure 4). Additionally, all ICEs except for ICEPmiChn31006 contained four hotspots (HS1, HS2, HS4, HS5) and one variable region (VR) III. Their HS4 (traN-s063) regions contain 15 to 19 ARGs (encoding for β-lactam, aminoglycoside, fluoroquinolone, fosfomycin, tetracycline, macrolide, sulphonamide, trimethoprim, and rifamycin resistance) that are clustered together in an ISPpu12-mediated composite transposon (Figure 4). The MDR HS4 regions identified in this study are highly similar to each other, with IS26-mediated excision/replacement upstream of tet(A) and ISCR2-mediated insertion of ARGs downstream of floR representing two major modular differences of them (Figure S1). These HS4 regions are also similar to that in ICEPmiChnHBSZC16 (MZ277866, P. mirabilis, animal, China), ICEEcoChnXH1815 (CP069386, E. coli, patient, China), and ICEKpnChnQD23 (CP042858, Klebsiella pneumoniae, patient, China), suggesting that ICEs may sever as an important vehicle in mediating the accumulation and dissemination of ARGs across bacterial species from different sources. ICEPmiChn31006 carries no VRIII and MDR HS4, but an HS3 region, with inserted genes encoding a pair of methylcytosine-specific restriction enzyme mcrBC. The sole ARG bla_CMY-2 (encoding β-lactam resistance) in ICEPmiChn31006 is present in the HS5 region and locates downstream of the truncated ISEcp1, the element assumed to be involved in mobilization and expression of bla_CMY-2 (Figure 4) (Harada et al., 2010). To determine the transfer ability of these ICEs, we selected four ICE-carrying strains (HJP21049, HJP17012, HJP31006, and HJP26021) for the conjugation experiments. However, no transconjugants were obtained after repeated attempts, despite that conjugative genes were complete. The unsuccessful conjugation might result from the exceptionally low transferability that is below detectable limits.

It has been known that SXT/R391 ICEs have the potential to be transferred and thus are important agents in the dissemination of antimicrobial resistance. SXT/R391 ICEs are commonly detected in Proteus, as the vehicle of various resistance genes, including several clinically important ARGs, such as bla_CTX-M-65 (Li et al., 2022), bla_NDM-1 (He et al., 2021), and tet(X6) (Peng et al., 2020). The findings in this study demonstrates highly genetic plasticity in the MDR HS4 region of ICEs carried by Proteus strains and further highlights them as a public health threat that requires continuous monitoring.

All the tested strains produce strong crystalline biofilm in the urine environment and all of them, except for HJP16012, are urease producers (Figures S2, S3). The hemolysis was observed in seven strains, including HJP21049, HJP05004, HJP17012, HJP20004, HJP26021, HJP16012, and HJP31030 (Figure S4). In addition, different degrees of swarming migration on agar surfaces, with diameter lengths of 42 to 80 mm, were observed in these strains (Figure S5). Whether motility is positively correlated with bacterial virulence needs further verification.

The fimbriae of P. mirabilis are essential virulence factors in UTIs (Schaffer et al., 2015). In this study, all the MDR P. mirabilis strains contained mrpA encoding the MR/P fimbriae, which is known to contribute to biofilm formation and virulence. 85.71% (12/14) strains contained the pmpA (encoding the P. mirabilis P-like fimbria) and 71.42% (10/14) of them contained the uroepithelial cell adhesin (UCA) fimbriae gene ucaA, while only two strains (14.29%) had the Proteus mirabilis fimbria (PMF) gene pmfA (Figure 5). CAUTI is generally initiated by biofilm formation on the urinary catheter (Schaffer et al., 2015). Several factors contribute to P. mirabilis biofilm formation in vitro, including but not limited to urease, MR/P fimbriae, Pst (phosphate transporter), and RcsD (flagellar regulation) (Schaffer et al., 2015). In addition to mrpA as mentioned above, all of the tested strains also contained ureC (encoding the urease), pst, and rcsD (Figure 5). This result provides a possible explanation for the strains’ strong biofilms. It is known that flagella are essential for swimming and swarming motility, which is an important characteristic feature of P. mirabilis uropathogenesis (Belas and Suvanasuthi, 2005). Some flagella genes, including flhA (encoding the flagellar assembly protein), fliF (encoding the flagellar MS-ring protein), fliG (encoding the flagellar motor switch protein), fliP (encoding the flagellar biosynthetic protein), fliL (encoding the flagellar basal body-associated protein), and flgN (encoding the flagella filament assembly) were identified in all the tested strains. flaD (encoding the flagellar capping protein) was absent in HJP05004, HJP17012, and HJP18031, whose swarming abilities, however, were not affected at all, suggesting that flaD maybe not very necessary for swarming motility. As previously reported (Cestari et al., 2013), hpmAB (encoding the hemolysin HpmA and its secretion protein HpmB) is highly conserved across the tested P. mirabilis isolates, whereas only seven of them exhibited hemolytic activity, suggesting an unknown regulatory mechanism is involved. Besides, 78.57% (11/14) of these P. mirabilis strains contained the siderophore synthase gene nrpR, which is important for colonizing the urinary tract (Schaffer et al., 2015). It was determined that non-fimbrial adhesins AipA (adhesion and invasion autotransporter) and Pta (Proteus toxic agglutinin) are required for invading and colonizing of P. mirabilis in the bladder (Armbruster et al., 2018). All of the tested strains carried pta, and four (28.57%) of them had aipA (Figure 5). Taken together, these results demonstrate a number of virulence genes associated with bacterial pathogenicity in these MDR P. mirabilis strains, revealing the virulence potential of them.