All 90 strains in this study were routinely isolated from clinical specimens of different patients in China between 2019 and 2023 ( Supplementary Table S1 ). The source from which most isolates were obtained was human sputum, followed by trachea tube, suggesting that most strains isolate from respiratory. All strains were identified as S. maltophilia by MALDI-TOF MS and DNA was extracted using the Genomic DNA Extraction Kit (Tiangen biochemical technology Co. Ltd., Beijing, China).

A total of 758 S. maltophilia genomes were analyzed in this study. Among them, 90 clinical strains were newly sequenced ( Supplementary Table S1 ). Another 668 uncontaminated genomes labeled “Stenotrophomonas maltophilia” with less than 200 contigs were screened and included from NCBI (https://www.ncbi.nlm.nih.gov/datasets/genome/) as of March 2023 ( Supplementary Table S2 ).

Genomes were sequenced using the Illumina HiSeq 2000 platform. The raw sequence data was evaluated by FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/), and the low-quality reads were filtered. The reads data was assembled into contigs using SOAP de novo v2.04 (Li et al., 2019). QUAST v5.0.2 (Gurevich et al., 2013) software was applied to evaluate the genome assembly. The genome size, number of coding sequences (CDSs), and GC content were estimated using Prokka v1.13.3 (Seemann, 2014).

FastANI (Jain et al., 2018) was employed to calculate the pairwise average nucleotide identity (ANI). The similarity matrix was imported into R and plotted as a heatmap of all strains. The strains belonged to the same species when the ANI value was above 95%. The genomes that could not be assigned to a known species were compared against all type strain genomes available in the TYGS (Type (Strain) Genome Server, https://tygs.dsmz.de/) database via the MASH algorithm to determine closely related type strains.

Phylogenetic analysis of the genus Stenotrophomonas follows the previously described method (Huang et al., 2022). Ten representative genomes were selected to cover the overall genetic distance of each species for those species with more than 10 available sequenced genomes, and all the genomes for the species with less than 10 publicly available sequences were adopted. Finally, a total of 190 strains were included, and the non-redundant homologous gene set was computed by CD-HIT (Fu et al., 2012). Next, BLASTp was used to search the homologous proteins in the nonredundant homologous protein set of each strain with a cutoff value of ≥ 90% sequence identity and ≥ 60% length coverage. Each single-copy core protein sequence was aligned separately using ClustalW2 and then merged. Iqtree v1.6.11 (Minh et al., 2020) was used to construct the phylogenetic trees by the maximum-likelihood method based on all these core SNPs (bootstrap replications, 1000). The visualization and beautification of the phylogenetic tree were accomplished by the tvBOT online tool (https://www.chiplot.online/tvbot.html) (Xie et al., 2023).

Nucleotide sequences predicted by Prokka v1.13.3 (Seemann, 2014) and produce *.gff output files were used for the following identification of core genes by Roray v2.3.12 (Page et al., 2015), with default parameters. To identify the virulence-associated genes, the protein sequences of all genomes were aligned using BLASTp with a relatively relaxed cutoff (60% identity, 60% length coverage) and an E value of 1e-5 against the data set from the VFDB (Liu et al., 2022). The antibiotic resistance genes (ARGs) were predicted using Resistance Gene Identifier (RGI) v6.0.2 with cutoffs of 75% identity and 80% length coverage against the data set from the Comprehensive Antibiotic Resistance Database (CARD, https://card.mcmaster.ca/) (Alcock et al., 2023).

Logarithmically grown bacterial culture was adjusted to an optical density at 600 nm (OD₆₀₀) of 0.4–0.5, and an aliquot (2 μL) of the culture was transferred into the LB semisoft (0.15% agar) agar motility plates, followed by incubation at 37°C for 22–24 h. Three parallel controls were set up, and the bacterial expansion zone diameters (millimeters) were measured.

The crystal violet (CV) staining method was used to quantify the ability of each isolate to form biofilm as previously described (Isom et al., 2022). Bacterial cultures of strains were grown overnight in LB medium, and 2 mL was next inoculated into an individual test sterile polystyrene tube (Falcon, Coning; NY, USA) at a 1:50 dilution. Incubation was at 37 °C for 24 h without shaking, under an aerobic atmosphere, and the OD₆₀₀ of each bacterial culture was measured. Then, non-adherent bacteria were removed by washing three times with sterile water carefully, and biofilm samples were fixed for 30 min at 60°C, dyed for 30 min with 0.1% CV, followed by careful rinsing with water to remove excess dyes, and solubilization of the crystal violet using dimethylsulfoxide (DMSO) for 1 h. Finally, the OD₅₇₀ was measured, and the OD₅₇₀/OD₆₀₀ ratio was used to represent the biofilm formation ability.

A total of 90 Smc isolates were tested for antimicrobial susceptibility. Minimum inhibitory concentrations (MICs) of 15 antimicrobial agents (Levofloxacin (LVX), Trimethoprim/Sulfamethoxazole (SXT), Minocycline (MH), Ceftazidime (CAZ), Amikacin (AN), Ampicillin/Sulbactam (SAM), Aztreonam (ATM), Cefepime (FEP), Ceftriaxone (CRO), Ciprofloxacin (CIP), Gentamicin (GM), Imipenem (IPM), Ertapenem (ETP), Piperacillin/Tazobactam (TZP), and Tobramycin (NN)) were detected by the broth microdilution method, and the resistance was defined using the Clinical and Laboratory Standards Institute (CLSI) criteria (Clinical Laboratory Standards Institute, 2023). Pseudomonas aeruginosa ATCC 27853 and Escherichia coli ATCC 25922 were used as control strains.

SPSS 26.0 (IBM Corp., Armonk, NY, USA) was used to conduct the statistical analysis. For categorical variables, the χ² test or Fisher’s exact test was used to determine statistical significance. P < 0.01 was considered to indicate statistical significance.

The genome sequences of the 90 clinical Smc strains sequenced in this study have been deposited at GenBank/DDBJ/ENA under the BioProject ID no. PRJNA1019301.

The 758 genomes were compared with type strains of the Stenotrophomonas genus by ANI calculations, and a total of 32 species-level clades were identified ( Figure 1A ). Among them, in addition to the 637 genomes assigned to 11 well-defined species, 121 genomes remained unassigned to any known species. Subsequently, the 121 strains were compared to the TYGS database, resulting in their identification as Stenotrophomonas spp., and could be further assigned to 21 novel Stenotrophomonas species named Genospecies 1 through Genospecies 21 ( Supplementary Table S2 ). It is worth noting that out of the 734 strains classified as “S. maltophilia”, only 314 (41.4%) were confirmed to be S. maltophilia. The majority of the remaining strains belonged to different species within the genus Stenotrophomonas.

The determination of phylogenetic connections among Stenotrophomonas spp. was based on the analysis of core protein sequences ( Figure 1B ), and the resulting phylogenetic tree revealed 24 species-level clades of Smc. The branches of each species of Smc are characterized by their short and compact arrangement, indicating a close relationship. Consequently, a total of 734 Smc genomes were incorporated into the subsequent analysis, the most prevalent species were S. maltophilia (n = 314, 42.8%), S. sepilia (n = 100, 13.6%), S. geniculata (n = 73, 9.9%), S. muris (n = 68, 9.3%), and Genospecies 1 (n = 31, 4.2%).

The analysis incorporated 734 strains from 39 different countries across 6 continents, with the exclusion of 16 isolates lacking information regarding their geographic origin ( Figure 2A ). These strains were predominantly obtained from China (n = 201, 27.4%), the United States (n = 174, 23.7%), Italy (n = 90, 12.3%), Germany (n = 60, 8.2%), and Spain (n = 28, 3.8%). The majority of strains are geographically spread over Europe (n = 251, 34.2%), Asia (n = 247, 33.7%) and North America (n = 176, 24.0%). Subsequently, smaller proportions of strains are found in Africa (n = 27, 3.6%), South America (n = 10, 1.3%) and Oceania (n = 5, 0.6%) is the least.

Given the disparity in the distribution of strains between continents, our study concentrated specifically on Europe, Asia and North America. These three continents were chosen due to their more abundant number of strains and species composition. In Europe, the prevalence of S. maltophilia was significantly higher (51.8%, P < 0.005) compared to the relatively lower occurrence of S. geniculata (4.0%, P < 0.005). S. muris strains in North America were significantly greater at 14.8% (P < 0.01). However, S. sepilia strains in North America were rather low, accounting for only 2.3% (P < 0.001). In contrast, higher proportions of these strains were found on the other two continents (Asia (19.0%), Europe (14.3%)). Additionally, S. maltophilia exhibits dominance as the prevailing species in South America and Oceania, while S. sepilia strains account for the highest proportion in Africa.

In terms of sample sources of Smc, a majority of the strains (n = 471, 64.2%) were obtained from clinical sources. The isolation source was categorized into five distinct groups ( Figure 2B ). The predominant isolates from the human respiratory tract were strains of Genospecies 19 (100%), Genospecies 2 (57.1%) and S. maltophilia (52.5%). Genospecies 4 (50.0%), Genospecies 5 (50.0%) and Genospecies 3 (40.0%) contained significantly more human-invasive strains compared with all other isolation sources. The prevalence of human-non-invasive origin strains was found to be significantly higher in Genospecies 14 (100%), Genospecies 7 (61.5%) and S. muris (20.6%). Additionally, environmental strains were more common in Genospecies 18 (100%), Genospecies 20 (100%), Genospecies 21 (100%) and Genospecies 9 (66.7%).

According to the annotations and predictions made by the VFDB database, 17 virulence factor classes, involving 154 virulence-associated genes, were obtained from 734 Smc genomes ( Table 1 ; Figure 3 ), each genome contains an average of 34 virulence genes. Belonging to 7 virulence factor classes (flagella, pili, LPS, capsule, protease, Hsp60 and EF-Tu) including 25 genes were present in almost all of the Smc genomes. In addition, there are variations in the distribution of certain genes among the species of the Smc. Specifically, the cheA and manC genes were exclusively found in S. maltophilia, where they play a crucial role in flagella formation and adhesion, respectively. Moreover, we found genes associated with type VI secretion system (T6SS) in numerous species of the complex, including Genospecies 7 (100%), Genospecies 12 (100%), Genospecies 3 (90%), P. hibiscicola (35.7%), Genospecies 1 (35.5%) and S. sepilia (33.0%). The presence of the gtrB gene, which plays a role in the modification of lipopolysaccharide (LPS) O-antigen, was merely seen in S. muris. The tppA gene, encoding Type IV pili, was found solely in S. sepilia. Similarly, the xcpS gene, encoding the general secretion pathway protein F, was only found in Genospecies 11. It is worth noting that a higher percentage of S. sepilia strains (22%) lack the katA gene, which is a catalase enzyme that plays a role in increasing persistence levels in response to hydrogen peroxide disinfectants and is found in nearly all other genomes.

A total of 46 ARGs, referring to 7 antimicrobial classes of drugs, were identified against the CARD ( Figure 3 ). Almost all genomes carry genes encoding resistance-nodulation-cell division (RND) antibiotic efflux pump, beta-lactamase and aminoglycoside phosphotransferase (APH), indicating the principal resistance mechanism of Smc. The prevalence of genes producing aminoglycoside acetyltransferase (AAC) was found to be approximately 37.1%. These genes were predominantly observed in strains of S. maltophilia, S. pavanii, and S. speilia. Another aminoglycoside-modifying enzyme, aminoglycoside nucleotidyltransferase (ANT), was exclusively encoded in S. maltophilia, S. muris, Genospecies 1 and P. hibiscicola. In addition, all strains of P. hibiscicola, Genospecies 1, Genospecies 4, and Genospecies 7 did not encode smeABC (a multidrug resistance RND-type efflux pump that confers resistance to aminoglycosides, beta-lactams, and fluoroquinolones) or smeRS (a two-component regulatory system for smeABC). The sulfonamide-resistance-conferring sul1 or sul2 were found in 26 strains (3.6%), mostly occurring in clinical strains, which suggests a relatively low prevalence of trimethoprim/sulfamethoxazole-resistant strains in our collection, which is the recommended first-line agent for the treatment of S. maltophilia infection.

We conducted an assessment of the swimming motility and biofilm forming capabilities of 90 clinical strains obtained from Smc, encompassing 13 distinct species. The findings indicated significant variations in swimming motility across the different species. S. sepilia exhibits the highest level of swimming proficiency, followed by S. pavanii, whereas Genospecies 2 displays the lowest swimming motility ( Figure 4A ). Meanwhile, these species have been identified with varying degrees of biofilm-forming capability. S. maltophilia and S. sepilia in particular have shown stronger biofilm formation activity, followed by Genospecies 2 and S. africana, whereas Genospecies 3 was classified as the weakest biofilm producer ( Figure 4B ).

A resistance phenotypic test was performed on 15 antibiotics against Smc ( Supplementary Table S3 ). Our analyses focused on the resistance of four therapeutic drugs against S. maltophilia infections ( Table 2 ). From all of the isolates, 44 were susceptible to the four antibiotics tested, whereas three trains, SMYL 1 (S. geniculata), SMYL 8 (Genospecies 1), SMYL 55 (S. maltophilia), were resistant or intermediary to three antibiotics. Surprisingly, all 14 strains of S. geniculata revealed resistance to Ceftazidime and exhibited significantly higher than other species (P < 0.001, Fisher’s exact test for n < 5). No statistically significant variation was observed between species in terms of resistance to the remaining three antibiotics.