The taxonomy of the genus Stenotrophomonas is complex and requires revision . The individual Stenotrophomonas species recognized by the List of Prokaryotic names with Standing in Nomenclature may be distinguished from each other by using the conservative cut-off for species of 0.95 in the ANIb (Fig. 1, Supplementary Fig. 1) [1, 20]. Using this same criterion, 65 of 111 isolates collected for this study and identified by MALDI-TOF/MS as S. maltophilia would belong to this species. Six isolates were S. pavanii; the other isolates would represent a total of 13 different novel species, which we here indicate as putative species A-M. Isolates that cluster as putative species B-M cluster more closely to S. maltophilia than to the type strains of other Stenotrophomonas species, suggesting that either the divergence of these putative species is more recent or that genetic exchange between these isolates has been more common.
A neighbor-joining tree of the concatenated alleles of the S. maltophila MLST scheme used to compare the ANIb results with those of Ochoa-Sánchez and Vinuesa showed that most groups of isolates or new species grouped comparably (Fig. 2) . Group #1 isolates clustered with the S. pavanii type strain, and also our pairwise clustering suggests that these isolates are S. pavanii. The previously identified group #6 clustered with the S. maltophilia type strain as well as with a number of our isolates. Group #9 was previously proposed as a novel species (Smc4), but, based on their clustering with the type strain in the MLST based phylogeny, the isolates in these group are most likely either S. indicatrix or S. lactitubi. Several previously proposed lineages clustered with the putative species in the ANIb analysis, e.g., lineage #3, #4, #5, #7 with putative species L, E, J, and B, respectively. Some of the putative species from the ANIb clustered with isolates that previously belonged to a different species now considered synonymous with S. maltophilia, such as P. beteli. Our data suggest that these isolates differ significantly from S. maltophilia and may indeed represent different species.
Some groups ((D), (F), Smc1, and Smc2) did not group in our analysis. This may be due to a different algorithm to generate the trees, but another explanation is that concatenated STs are not optimal for defining species [3, 4]. However, it should be noted that the bootstrap values for some branches are low, indicating low confidence for the branching.
To further elucidate the taxonomic position of our isolates, a 16S rDNA sequence analysis was performed. This analysis was basically in agreement with the ANIb analysis. Remarkably, the 16S rDNA sequence of the S. pavanii type strain clustered closely with that of the S. maltophilia type strain and with the S. maltophilia isolates collected for this study, whereas our other S. pavanii isolates clustered elsewhere (Fig. 3). As already suggested by the ANIb, the 16S rDNA data suggest that S. maltophilia, S. pavanii and the putative new species are still closely related. Based on 16S analysis, S. africana, which is currently considered synonymous with S. maltophilia, should be considered as a separate species.
All our isolates harbored a chromosomally encoded L1 and L2 ß-lactamase. However, the (type) strains of species other than S. maltophilia and S. pavanii lacked these ß-lactamases, except for S. indicatrix and S. lactitubi, which encoded the L2 gene. A Neighbor-joining analysis of the amino acid sequences was in agreement with the results of the ANIb and concatenated STs (Fig. 4, Fig. 5): isolates identified as S. maltophilia in the ANIb clustered together with limited sequence variability; the other (putative) species displayed more diversity. The L1 ß-lactamase hydrolyzes carbapenems and other ß-lactam antibiotics (but not monobactams), whereas L2 ß-lactamase is a serine hydrolase that acts as a cephalosporinase. Stenotrophomonas MICs for imipenem and meropenem are generally high; more variation is seen for ceftazidime. The considerable sequence variation for both ß-lactamases, which was already described when the first genes were sequenced, may at least in part explain differences between isolates [6,7,8,9].
All S. pavanii isolates contained a chromosomally encoded aac (6′)-Iak aminoglycoside resistance gene and all isolates that were S. maltophilia based on the ANIb had an aph (3′)-IIc aminoglycoside resistance gene, whereas the other isolates lacked chromosomally encoded aminoglycoside resistance genes. S. maltophilia and S. pavanii tended to have higher MICs for tobramycin than the putative species (Supplementary Table 1). However, this correlation was not perfect, suggesting that other factors play a role in expression and thereby resistance, including regulation of aminoglycoside resistance genes. In addition to aph (3′)-IIc some S. maltophilia had aac (6′)-Iz, a chromosomally encoded aminoglycoside resistance gene (Table 1).
Only a few isolates possessed acquired resistance genes (Table 1), but one isolate encoded 7 different ones, including a sul1 resistance gene. Sul1 is associated with class 1 integrons , but presence of an integron could not be confirmed. Nevertheless, this suggests that S. maltophilia can acquire plasmids or transposons with class 1 integrons, as has been previously described [21, 22]. Differences between the (presence of) chromosomally encoded antibiotic resistance genes further supports the hypothesis that our isolates included different species (S. maltophilia, S. pavanii and a set of putative novel species). Although the chromosomally encoded ß-lactamases and aminoglycoside resistance genes may be clinically relevant, their function is probably to aid in competition with other micro-organisms in their natural niche.
The extensive resistance pattern of S. maltophilia severely limits the antibiotic treatment options for infections, and only co-trimoxazole is considered a reliable treatment option. Described alternatives include fluoroquinolones and tetracyclines , but the MICs may vary considerably. Currently little is known about the genetic factors determining resistance to these antibiotics with the exception of fluoroquinolone resistance, which is associated with overexpression of efflux pumps. These are either SmQnr, SmeDEF or SmeVWX [14,15,16]. Overexpression of SmeVWX is associated with amino acid changes in its repressor, SmeRv [15, 16]. Five isolates had previously described amino acid substitutions; however, their ciprofloxacin MICs ranged from 2 to > 32 mg/L. For the newly described variants (a C310W in 534,828 and truncated sequences) the range was 2 to 16 mg/L. No mutations in the quinolone resistance determining regions of gyrAB and parCE were observed .
The presence of the genes for the proteases StmPr1, StmPr2, and StmP3, as well as the DNase, phospholipase C and D, esterase, fimbriae, TadE-like protein, and the RpfC regulator genes in all or nearly all isolates (Table 2, Supplementary Table 2). However, somewhat lower percentages have been reported in literature. StmPr1, StmPr2, esterase, and fimbriae were present in more than 90% of the isolates in a collection from CF patients attending a pediatric hospital in Rome, Italy . Possibly, the collections differed in the distribution of different (putative) species, and that these genes do not belong to the core genome of all (putative) species.
Differences in the type IV pilus adhesion precursor should likely be sought in the different adherence properties that may lead to advantages in different natural niches.
StmPr1, − 2, and − 3 degrade a wide variety of extracellular matrix components. StmPr1 degrades collagen type I and IL-8, and can kill A549 lung epithelial cells; StmP3 degrades fibronectin, fibrinogen, and IL-8, and contributes to cell rounding and detachment in vitro; the activities of StmPr2 have not exactly been defined, but this protease contributes to degradation of extracellular matrix proteins and cell rounding. The three proteases are secreted in Xps-dependent manner [24, 25]. However, the genes encoding Xps type II secretion system could only be identified in 90% of the S. maltophilia isolates and in putative species B and J andabsent in the other (putative) species. Since phenotypic protease activity did not correlate with species or STs, proteases may likely also be secreted by an alternative mechanism. All isolates contained protease genes and an esterase gene, but phenotypic expression of protease and esterase activity was found in only 71.6 and 40.4% of the isolates respectively. This suggests that the regulation of expression of these activities is complex. Alternatively, additional proteases may be present which are secreted by a different mechanism.
Polysaccharide lyase degrades alginate, poly-ß-D-glucuronic acid and hyaluronic acid . These first two compounds are found widely in nature and hyaluronic acid is an important constituent of human skin, but it is also found elsewhere in the human body. The polysaccharide lyase was present in more than three quarters of the isolates, and its absence/presence did not follow the (putative) species boundaries. The absence of the gene in many isolates obtained from CF and lung infections indicate that it is not essential for colonization or infection of the human lung.
The secreted ankyrin-repeat protein, which interacts with actin, was present in only approximately a quarter of the isolates . This interaction is thought to alter the cytoskeletal structure. However, StmPr1 and StmPr2 also contribute to this process .
The RPF quorum sensing system showed considerable sequence variability, in particular for the sensor protein (Supplementary Fig. 2). This quorum sensing system has been reported to regulate motility, biofilm development, antibiotic resistance and virulence in S. maltophilia, but the implications for (CF) lung infections are not immediately clear [18, 19, 28]. It has been reported that the signal molecule DSF generated by the expression of the RPF system influences Pseudomonas aeruginosa biofilm formation and antibiotic resistance [29,30,31]. Some isolates lack a 190 amino acid sequence in the sensor domain region in the middle of the protein. Huedo et al showed sequence variation in the RPF cluster of S. maltophilia and designated two variants: rpf-1 and rpf-2 . The first is more similar to the system found in Xanthomonas, whereas the latter is more similar to the system in Pseudoxanthomonas, Arenimonas, and Lysobacter. In a rpf-2 system, with a shorter sensor domain, it appears that some quorum sensing signal DSF needs to be present for activation of DSF synthesis. It has been speculated that under some conditions the rpf-2 system saves energy, and that its activation is dependent on the presence of other bacterial isolates or species .
The majority of S. maltophilia isolates are capable of biofilm production, but marked differences have been observed. On average more biofilm formation on polystyrene was observed by non-CF isolates than by CF isolates, which in turn displayed more biofilm formation than environmental isolates, but large variation among isolates within the groups was present . Variable biofilm formation was also observed on IB3–1 bronchial cells in vitro .
The spgM, and rmlA genes, encoding aphosphoglucomutase/phosphomannose bifunctional protein and glucose-1-phosphate thymylyltransferase, respectively, have been implicated in biofilm formation and were reported in 83.3 and 87.5% of 37 isolates tested . However, all our isolates contained these genes; therefore, either there were differences in the populations tested, or the PCR used in prior studies did not detect all genes due to sequence variation in the primer region(s). Although the RPF system has been reported to influence biofilm formation, no correlation with the detection of rpfF by PCR was found . The nitrate reductase has also been associated with biofilm by enabling growth in micro-oxic conditions . Nearly half of our isolates harbored the narG gene encoding the nitrate reductase; this was in agreement with a previous study that reported 37/63 positive isolates . The Ax21 outer membrane protein is also involved in biofilm formation, as well as motility, reduced tolerance to tobramycin, and virulence in an insect model, but its regulation and mode action of action have not been resolved .
Although these data help to identify putative virulence their role should ultimately be proven in vivo.