All 2,615 genomes of the genus Stenotrophomonas were obtained from the NCBI genome database (https://www.ncbi.nlm.nih.gov/datasets/genome/) as of 1 April 2024. Among them, 402 genomes noted as atypical assemblies (including 22 contaminated genomes, 74 too large and too small genomes, and 306 genomes from unverified sources) were excluded. A total of 2,213 genomes were ultimately included in this analysis. The genome size and GC content of the genomes were ascertained using Seqkit v2.8.2 (https://github.com/shenwei356/seqkit) (Shen et al., 2024).

Currently, there are 31 recognized species of the genus Stenotrophomonas in the NCBI Taxonomy database. Notably, S. indologenes and S. detusculanense do not have available genome sequences. Furthermore, neither S. aracearum nor S. oahuensis has a reference genome assigned in the NCBI database; their only available genomes served as reference genomes. Two novel species of the genus S. pigmentate and S. tuberculopleuritidis (610A2 and 704A1, currently temporarily classified as S. sp.) identified in our previous research (Li et al., 2024) were also incorporated as reference genomes for the study. Consequently, a total of 31 reference genomes were included in this study (Supplementary Table S1).

The 16S rRNA gene sequences were extracted from the genomes using barrnap v0.9 (https://github.com/tseemann/barrnap). Pairwise similarities between 16S rRNA genes were assessed through alignment using BLAST+ v2.15.0 (http://blast.ncbi.nlm.nih.gov/Blast.cgi). The ANI similarity matrix was constructed using skani v0.2.1(https://github.com/bluenote-1577/skani) (Shaw and Yu, 2023) and FastANI v1.33 (https://github.com/ParBLiSS/FastANI) (Jain et al., 2018). dDDH values of the reference genomes were compared using the GGDC 3.0 (https://ggdc.dsmz.de/ggdc.php) (Meier-Kolthoff et al., 2022). The visualization of the results was conducted using Matplotlib v3.8.4 (https://matplotlib.org/) and Seaborn v0.11.0 (https://seaborn.pydata.org/) in Python 3.12.1.

For the phylogenetic analysis of all reference genomes, Pseudoxanthomonas broegbernensis DSM 12573 was selected as the outgroup. We used Prokka v1.14.6 (https://github.com/tseemann/prokka) (Seemann, 2014) to annotate the genomes. The output GFF-formatted files from Prokka were used as input files for Roary v3.13.0 (https://github.com/sanger-pathogens/Roary) (Page et al., 2015) to calculate the genus Stenotrophomonas core and pan-genomes with the minimum blastp percentage identity set at 90%. The phylogenetic tree was constructed using FastTree 2(https://github.com/PavelTorgashov/FastTree) (Price et al., 2010) based on the multi-FASTA alignment of all of the core genes and visualized using Tree Visualization By One Table (tvBOT) (Xie et al., 2023).

The ANI value matrix between the 31 reference genomes and the ANI of all the Stenotrophomonas genomes compared to the 31 reference genomes was generated using both FastANI and skani. For each species, genomes were categorized into three groups based on their ANI compared to the reference genome as follows: ANI ≥ 95, 95% > ANI ≥ 90%, and ANI < 90%. The difference in classification between the two ANI tools was analyzed using the McNemar–Bowker test, and the consistency was evaluated by the kappa coefficient. The difference between the ANI values calculated by the two tools after stratifying was evaluated using a paired t-test. All statistical analyses were conducted using R version 4.4.0 (https://cran.r-project.org/bin/windows/base/old/4.4.0/), with p < 0.05 indicating statistical significance.

The ANI matrix for all genomes of the genus Stenotrophomonas was generated using skani or FastANI. A helper Python script provided using skani was employed to visualize and cluster the lower-triangular similarity matrix that skani output. Building upon this script, the fcluster function from Scipy v1.13.1 (https://docs.scipy.org/doc/scipy/) was integrated to actually obtain clusters by setting an ANI threshold.

We examined the distribution of pairwise ANI values separately between all genomes of the genus Stenotrophomonas (n = 2,213) and between all genomes of S. maltophilia (n = 1,555) using skani. In our evaluation, one genome was taken as the reference genome, and all genomes (including itself) were taken as query genomes. A total of n-squared pairwise comparisons were executed. The density curve of the pairwise ANI values was plotted using ggplot2 (https://ggplot2.tidyverse.org/) in R version 4.4.0.

The ANI cutoff value was evaluated through the analysis of clustering results and the density distribution. Based on the ANI cutoff values, genomes within the genus Stenotrophomonas were reclassified into the designated species.

A total of 2,213 genomes were ultimately included in this analysis. In this study, the genomes of unclassified species were all included as Stenotrophomonas sp. (including uncultured organisms). In addition to S. sp., the study encompasses 29 defined species composed of 1,841 genomes. The number of each species is shown in Figure 1A. S. maltophilia is the most abundant, with a total of 1,555 genomes, accounting for 70.27% of all genomes. The second most abundant species is S. sp., with a total of 372 genomes. The remaining species all have less than 50 genomes, including 12 species with only one genome. Genomic characterization of different species was shown in Figures 1B,C. The size range of all Stenotrophomonas genomes is from 1.77 to 6.98 Mb, while that of S. maltophilia is 3.80 to 5.54 Mb. The average genome size of each species is similar, except for S. koreensis, whose genomes are significantly smaller than other species, ranging from 2.07 to 3.09 Mb. S. sp. contains various unclassified strains, so its relatively wide genome size range is expected. but the performance of S. nitritireducens is beyond expectation. S. nitritireducens only has 6 genomes, but there is one largest (GCA_037190265.1, 6.98Mb) and one smallest (GCA_017304365.1, 1.92Mb) genome of genus Stenotrophomonas, although the average size of the other 4 genomes (4.22 Mb) is similar to other species.

The GC content of Stenotrophomonas genomes ranges from 62.54 to 70.17%. Apart from S. maltophilia, S. sp., and S. nitritireducens, the GC content distribution of each species is relatively concentrated (fluctuations do not exceed 2%). Although the GC content of over 95% of S. maltophilia genomes concentrates at 65.5–67.0%, it ranges from 63.6 to 68.17%. S. sp. strains are more dispersed, ranging from 62.54 to 69.87%. The GC content of S. nitritireducens genomes exhibited a bimodal distribution, with four genomes ranging from 68.08 to 68.62% and the other two genomes ranging from 63.83 to 63.98%. The GC content of S. acidaminiphila is higher than that of other species, ranging from 68.06 to 69.75%.

One reason for the large fluctuations in genome size and GC content is that 20.83% (461/2213) of the submitted genomes of Stenotrophomonas are metagenome-assembled genomes (Supplementary Table S2), and the majority of these outlier genomes belong to this type, limiting the accuracy of these genomes. The other speculated reason is that many isolates are from diverse sources, leading to genetic variability and diversity. There are nine strains that have been sequenced more than once and have genomes with different levels of completeness in the database. For example, S. maltophilia strain NCTC10498 has two complete genomes (GCA_011386925.1 and GCA_900475685.1) and one scaffold genome (GCA_014235265.1). The ANI values of genomes from different-levels of the same strain were similar to the reference genomes (Supplementary Table S6), indicating that the genome sequencing method and genomic assembly level do not affect the species classification based on the ANI.

The 16S rRNA gene sequences were extracted, and pairwise 16S rRNA gene similarities were calculated (Figure 2A). The 16S rRNA gene similarity between many species is ≥98.65%, and even greater than 99%. Obviously, the 16S rRNA gene sequence has low discriminability for Stenotrophomonas species.

As the gold standard for the delineation of bacterial species, the trends of pairwise ANI and dDDH were essentially consistent across the reference genomes analyzed (Figures 2B–D). The dDDH and ANI values between the defined species are all below the commonly used cutoff values for species determination. The only exceptions were S. pennii Sa5BUN4 and S. lacuserhaii K32. Their 16S rRNA gene sequence similarity, dDDH, and ANI values calculated using skani and FastANI between these two species were 100, 74.5, 97.32, and 97.22%, respectively, all of which exceeded the species cutoffs. Although S. pennii was first reported in 2021 (Gilroy et al., 2021), it was not validly published. S. lacuserhaii was reported in 2022 (Deng et al., 2022) and validly published under the International Code of Nomenclature of Prokaryotes (ICNP). S. pennii and S. lacuserhaii should be merged into one species.

In addition, we observed that the ANI values between S. panacihumi GSS15 and other genomes were low, with values calculated using skani being below 82% and values calculated using FastANI ranging between 81 and 84%. S. panacihumi is a non-validly described and poorly characterized species (Yi et al., 2010). It only has two assembled genomes. The study by Vinuesa et al. also demonstrated the inconsistency of S. panacihumi in the phylogeny of the genus Stenotrophomonas (Vinuesa et al., 2018). This could be the reason why the GTDB classified S. panacihumi into the new genus Stenotrophomonas_A. We also discovered the anomaly of S. koreensis and S. ginsengisoli in ANI comparisons, although both were validly described species (Yang et al., 2006; Kim et al., 2010). In addition, the paired ANI between S. koreensis DSM 17805 and S. ginsengisoli DSM 24757 was 86.76% (skani) and 86.92% (FastANI). However, their ANI with the other 29 reference genomes is very low, with values below 82% (skani) or ranging between 81 and 82% (FastANI).

Furthermore, we compared the concordance between the ANI clustering heatmap (Figure 3A) and the phylogenetic tree constructed based on the core genes (Figure 3B). Considering that many pairwise ANI values between reference genomes are below 82% and output as “0” in skani as the default parameter setting (Shaw and Yu, 2023), we utilized FastANI to calculate accurate ANI values for clustering. The results showed that the classification of the four clades of the genus Stenotrophomonas based on these two methods and the composition of the species in each clade were similar. The relatively simple ANI calculation can replace the complex core gene analysis for identifying closely related species and classifying them on the phylogenetic tree. In the phylogenetic tree, we also observed inconsistencies between S. panacihumi GSS15 and other genomes of the genus Stenotrophomonas. While S. koreensis DSM 17805 and S. ginsengisoli DSM 24757 had lower ANI values than other genomes, they still belonged to the genus Stenotrophomonas on the phylogenetic tree.

We compared the ANI values between all 1,841 genomes of 29 defined species and their reference genomes using two ANI tools to assess their consistency. The results indicated that when using 90 and 95 as thresholds, the ANI of each genome calculated by these two ANI tools was completely consistent in classification (Table 1 and Supplementary Table S3). The McNemar–Bowker test showed no significant difference (kappa coefficient = 1).

We conducted paired comparisons for ANI values across each range (Figure 4) and observed that for ANI ≥ 95%, ANI values calculated using skani were higher than those calculated using FastANI, exhibiting a notable statistical difference (p < 0.05). For ANI values between 90 and 95%, the ANI values computed by both tools were essentially consistent (p > 0.05). For ANI < 90%, the ANI values calculated using skani were slightly lower than those using FastANI. However, the difference was not statistically significant (p > 0.05). Despite a few differences in the ANI values calculated by the two ANI tools, they demonstrated identical discriminatory capabilities at thresholds of 90% or 95%. Considering that skani’s computations were several to tens of times faster than FastANI, we primarily utilized skani for large-scale paired ANI calculations in this study.

We conducted paired ANI comparisons of all 2,213 Stenotrophomonas genomes and performed clustering. Out of a total of 4,897,369 comparisons of the genus Stenotrophomonas genomes, 4,196,423 yielded ANI values greater than 82%. For S. maltophilia, out of a total of 2,418,025 comparisons, 2,403,945 yielded ANI values in the 82–100% range. The distribution of paired ANI values of the genus Stenotrophomonas and S. maltophilia species is quite similar to the multimodal distribution (Figure 5A), which may be partially due to the fact that the genomes annotated as S. maltophilia account for the majority of the genus Stenotrophomonas.

We found that the peak of the ANI distribution is most abundant between 90 and 94.7%, accounting for approximately half of all ANI values. This indicated that numerous genomes have been mistakenly annotated as S. maltophilia, which actually belonged to other species closely related to S. maltophilia in the ANI clustering. The ANI values between the species classified in Smc, S. sepilia, P. hibiscicola, and S. maltophilia also fell in this range. Therefore, we recommended that a 90% ANI value could potentially serve as the threshold for distinguishing complex/group species within the genus Stenotrophomonas.

In our study, despite a minor peak between 95.5 and 96.5%, the curve representing ANI values between 94.7 and 96.5% remained flat and very low, not exceeding 0.45% of all pairs, especially around 95%. Therefore, we suggested that the interspecies ANI threshold for the genus Stenotrophomonas might fall within this range. We examined the number of ANI clusters formed at different ANI cutoffs from 94 to 96% with every 0.1% increment and did not observe a stable plateau in the number of clusters (Figure 5B). Considering that 95% is the nadir of the kernel density distribution curve, as the cutoff for species delineation in prokaryotes, ANI ≥95% can correctly classify the genus Stenotrophomonas at the species level.

In the ANI classification, 758 (41.17% of 1,841) genomes had ANI values below 95%, with 158 genomes (8.58%) even below 90% (Table 1). Table 2 provides a detailed list of species with ANI values below 95% compared to the reference genomes. These results indicated that a large proportion of genomes had been incorrectly identified and annotated. Meanwhile, only 135 out of 372 S. sp. strains had ANI > 95% compared to the reference genomes and could be classified into the defined species.

Gröschel et al. classified Smc into 23 monophyletic lineages based on MALDI-TOF MS and 98.8% 16S rRNA gene sequence identity as the strain inclusion criteria (Gröschel et al., 2020). For consistency, we used and amended the naming convention of lineages from previous reports. In our study, 850 genomes, including the S. maltophilia reference genome NCTC10258, belonged to the lineage Sm6, which is strictly defined as S. maltophilia species by 95% ANI. Through ANI comparison, we discovered that the ANI values of several Smc lineages described by Gröschel et al. with species within or outside Smc were ≥ 95%, for example, Sm9 and Sm16, Sm3 and S. sepilia, and Sm5 and P. hibiscicola within Smc; Sm2 and S. pavanii, Sm4a and S. muris, Sm10 and S. geniculata, Sm15, and S. riyadhensis, and Sgin3 and S. indicatrix outside Smc. Moreover, it is worth mentioning that when the ANI threshold for the complex/group is set to 90%, only lineages Sm1–Sm18 should be kept in Smc, while lineages Sgn1-Sgn4 with more distant relationships should be removed.

The clustering heatmap of paired ANI comparisons of all 2,213 Stenotrophomonas genomes indicated that there are a considerable number of clusters within the genus Stenotrophomonas (Supplementary Figure S1). Using 95 and 90% ANI as species and complex/group thresholds, 2,213 Stenotrophomonas genomes in the NCBI database were re-clustered into 126 ANI genomic species and 51 ANI genomic complex groups (Supplementary Tables S4, S5). The clustering result in our study suggested that Smc should be redefined. It may consist of 1,668 genomes from 33 species, including 18 previously described species and lineages (P. hibiscicola/Sm5, S. geniculate/Sm10, S. maltophilia/Sm6, S. muris/Sm4a, S. pavanii/Sm2, S. riyadhensis/Sm15, S. sepilia/Sm3, Sm1, Sm4b, Sm7-8, Sm9(Sm16), Sm11-14, and Sm17-18) and 15 novel ANI species.

The 10 genomes closely related to S. panacihumi were clustered into 4 complexes composed of S. panacihumi and other 7 ANI species (ANI species 117, 119–124). The ANI values between these genomes and other species in the genus Stenotrophomonas were below 82%. Whether plotted on the ANI clustering tree or the core gene phylogenetic tree, S. panacihumi is a clade that is far apart from other Stenotrophomonas species. Moreover, Stenotrophomonas sp. MMGLT7 (ANI species 125) should also be removed from the genus Stenotrophomonas as its genome (GCA_021168555.1) showed the highest identity with Xanthomonas massiliensis SN8 (GCA_900018785.1), with an ANI of 86.1%. The classification of S. koreensis and S. ginsengisoli was challenging. Their ANI values for other Stenotrophomonas species were even lower than those of S. panacihumi. However, on the core gene phylogenetic tree (Figure 3B), they were related to the clade including S. terrae, S. nitritireducens, and S. acidaminiphila. If S. koreensis and S. ginsengisoli are retained, there would be 116 ANI genomic species and 47 ANI complexes in the genus Stenotrophomonas. In addition to the aforementioned species and their reference genomes for other ANI species or lineages, high-quality genomes were selected to serve as their respective reference genomes. A hierarchical clustering dendrogram was constructed using the ANI matrix of 125 reference genomes and annotated with both ANI genomic species and currently defined species (or lineages) information (Figure 6).

It should be pointed out that ANI clustering is based on all genomes within a cluster, which may result in some ‘transboundary’ occurrences. For example, S. lactitubi Bi35 (GCA_918698175.1) exhibited ANI values above 95% when compared to the reference genome of S. lactitubi (GCA_900188015.1) and 10 other genomes within this species. However, its ANI values to the remaining 13 genomes of the species were < 95%. According to the genomic similarity to the reference genome, we kept it in S. lactitubi. In the 54 paired comparisons between lineage sm4b (18 genomes) and ANI species 88 (3 genomes), 12 values were > 95%, and 42 values were < 95%; therefore, they were divided into separate species.