Abstract

MATERIALS AND METHODS

RESULTS

Linkage disequilibrium.

In order to assess the clonality of S. maltophilia, the I_A was calculated for different groups of isolates. This measure was significantly different from zero, which indicates a linkage disequilibrium, if all 65 strains or all 48 clinical isolates were considered (Table ​(Table4).4). The I_A differed from zero in regard of all 54 STs, all 41 STs of clinical isolates, or all 17 environmental isolates but did not reach statistical significance in these analyses because the maximum variance in 1,000 random trials exceeded the observed variance.

TABLE 4.

I_A calculated for different groups of S. maltophilia^a

Sample group (no. analyzed)a	Observed variance	Expected variance	IA	Max trial varianceb	Linkage disequilibriumc
Total sample group
Isolates (65)d	0.498	0.142	2.498	0.447	Significant
STs (54)	0.177	0.090	0.971	0.480	Not significant
Clinical group
Isolates (48)d	0.778	0.219	2.562	0.534	Significant
STs (41)	0.304	0.137	1.216	0.589	Not significant
Environmental group
Isolates (17)	0.795	0.203	2.911	1.850	Not significant
STs (15)	0.130	0.132	−0.011	1.708	Not significant

Open in a separate window

^aClinical isolates also included those from the hospital environment; environmental isolates included only those originating strictly outside of hospitals.

^bMax, maximum.

^cThe I_A was considered as significantly different from 0, indicating a disequilibrium, if the observed variance exceeded maximum variance in 1,000 random trials.

^dEvery second isolate from each of the five outbreaks was omitted.

Genogroups in S. maltophilia.

The BURST analysis for clonal complexes revealed three groups of triple locus variants comprised of (i) STs 17 and 30 belonging to genomic group 7, (ii) STs 1, 26, and 27 belonging to genomic group 6, and (iii) STs 2 and 43 also belonging to genomic group 6. All the strains of these STs were of clinical origin.

The NJ tree presentation was chosen (Fig. ​(Fig.1)1) for cluster analysis of all 70 S. maltophilia strains. The 10 different genogroups defined by Hauben et al. (35) could be delineated on the basis of all seven loci. We were able to use the same numbering of genogroups by including strains that had already been assigned in that previous study. Additional groups were given alphabetical designations. None of the other isolates investigated grouped together with strain 888_env, formerly assigned to AFLP genogroup 10. Sequence divergence between different genogroups and calculated k parameters are given in Tables S4 and S5 in the supplemental material. On average, sequence divergence between different genogroups was as high as 0.048 (± 0.002 SE), with the highest values ranking in the order of genogroups 8, A, 9, and 5. All the groups could be separated by a significant ratio of the between-group divergence to the mean of the within-group divergence above 2, except for the comparison of genogroup 2 with groups 3 to 6, B, D, and E. Moreover, five additional groups became apparent, which we designated alphabetically. S. africana belonged to group 4, along with four other S. maltophilia strains. Two isolates, 470_bc and 639_CF could not be assigned to any genogroup.

Open in a separate window

FIG. 1.

NJ tree based on the concatenated data for all seven loci of the 70 S. maltophilia strains. The isolates originated from blood culture, CF patients, tracheal secretions from outbreaks, sputum, pus, conjunctivitis, environmental specimens, hospital environment, patient, and one CSF isolate of S. africana (strain abbreviations are identified in Table ​Table1).1). Previously defined genogroups were numbered (#1 to #9) according to previously investigated strains (35) included here and shown in small rectangular boxes. An uppercase letter indicates groups newly detected in this work. Bootstrap values (200 random seeds) are given as percentages for the main branches. The tree was rooted with the corresponding concatenate of X. campestris (53).

Twenty-two (44%) out of 50 clinical isolates originating from patients in seven German cities and four other countries were clustered in group 6, which consisted of clinical isolates only. Twelve (63%) out of 19 strains isolated from epidemiologically nonassociated CF patients also belonged to this group. Clinical invasive isolates originating from blood cultures or CSF were distributed more evenly among nine genogroups, with just 3 (18.7%) out of 19 strains belonging to genogroup 6. The three hospital environment strains were found in three different groups (2, D, and E), which also included clinical strains. Three genogroups, though containing a limited number of isolates, were exclusively comprised of strains of environmental origin: genogroups 5 (n = 4), 8 (n = 6), and 9 (n = 3).

DISCUSSION

In the past decades, there have been several changes in the taxonomy of Stenotrophomonas spp. (49). At present, this genus is comprised of nine species, five of which, S. dokdonensis, S. humi, S. koreensis, S. rhizophila, and S. terrae, were not available for the study presented here. The delineation of S. africana from S. maltophilia has been debated (10, 21). This work also confirms that S. africana is a synonym for S. maltophilia.

S. maltophilia can be isolated from a wide variety of environments and geographical regions and may occupy various niches such as soil, rhizosphere, water, and food (17). S. maltophilia has emerged in many hospitals as an important nosocomial pathogen, especially in immunocompromised patients (61). Although this species was previously considered to have limited pathogenicity, reports indicate that infection with the organism is associated with significant morbidity and mortality, particularly in severely compromised patients (46). Yet the clinical importance of this opportunistic pathogen as a mere colonizer or infectious agent often remains unresolved. There is an ongoing debate about the role of this species in later stages of CF. At any rate, isolation of S. maltophilia in CF patients tends to be associated with more advanced disease (15, 30, 32). Treatment of S. maltophilia infection is also complicated by its inherent resistance to many broad-spectrum agents, including carbapenems, which to a considerable extent are mediated by multidrug efflux pumps (1) and broad-spectrum beta-lactamases (4). The emerging resistance to trimethoprim/sulfametoxazole, which is one of the few remaining treatment options, is of major concern (69). Currently, little is known about the virulence factors of S. maltophilia, such as factors for adherence to plastics, an extracellular protease, and a phage-encoded zonula occludens-like toxin (19, 33, 72). More recently, a diffusible signal factor has been described that is assumed to control the expression of virulence and antimicrobial resistance as a cell-cell signaling factor by use of a two-component regulatory system (26).

Genotyping methods have been used successfully in the molecular epidemiology of S. maltophilia. Many typing approaches have revealed that this species is of high genodiversity. Genotyping of isolates from hospitalized CF and non-CF patients by PFGE or randomly amplified polymorphic DNA PCR revealed a high degree of diversity with changes in the strains consecutively isolated from CF patients (18, 38, 70, 71). A multilocus enzyme electrophoresis scheme has been proposed for investigations of hospital epidemiology (60). However, this scheme could not be used here, partly because it is based on markers such as peptidases or elastases, which cannot be assumed to be neutral for selection, and on other markers we could not retrieve unambiguously from the genome sequence databases.

There have been several attempts to delineate the genetic groups of related strains. S. maltophilia could be subdivided into three clusters by use of 16S rRNA gene sequence signatures in the V1 (positions 73 to 97) and V6 (positions 451 to 482) domains (44). A comparative investigation of clinical isolates revealed six 16S rRNA groups based on variable positions in the positions 41 to 109 and four so-called phylogenetic groups based on the smeD-smeT intergenic sequences (31). This region is assumed to be involved in regulation of the smeDEF multidrug efflux pump genes (58). The most extensive work was done by Hauben et al., who used AFLP to assign a highly diverse strain collection of different origins to 10 genomic groups. The results were in part confirmed by 16S rRNA gene sequencing and DNA-DNA hybridization experiments (35). However, 18 out of 107 investigated strains were nongroupable. In a different approach, nine different genomic groups of the genus Stenotrophomonas could be distinguished on the basis of RFLPs in the gyrB gene (11). Surprisingly, 36 out of 40 isolates from CF patients were grouped in just two clusters, group B and group C. In summary, on the basis of RFLP, there is increasing evidence of the existence of particular subgroups of different ecological origin and clinical importance within this bacterial species (11, 35). However, the impact of a few recombination events can be sufficient to obliterate the phylogenetic signal in the gene trees of many species (22). Thus, for future investigation of isolates, this study aimed to develop an MLST scheme to cluster different genotypes in phylogenetically meaningful groups.

The scheme was developed on the basis of a restricted set of carefully selected strains. This diverse strain collection included strains from 10 previously defined genogroups, type strains, and 17 isolates of nonanthropogenic origin. Also, for the purpose of sequence analysis, data were included from two genome sequencing projects of strains K279a and R551-3 even though changes in the sequence data of the latter strain were possible before final publication. However, grouping of the environmental isolate R551-3 together with other environmental isolates in genogroup 5 allowed use of these preliminary data to appear plausible although allelic types were unique in the data set of all seven loci investigated so far.

Despite the fact that the number of isolates investigated is still limited, analysis of the individual loci revealed a high degree of diversity, reflected by a high number of allele types and variable sites, as well as by a Simpson's index of diversity of at least 0.971 in all seven sequences. However, one should bear in mind that the strains had already been selected from different collections on the basis of the maximum diversity available to us. Furthermore, in contrast to MLST or single nucleotide polymorphism (SNP) analysis of core genome genes, it is possible that RFLP-based methods, like AFLP or PFGE, which also cover the entire accessory genome, give greater insight into epidemiological associations or pathoadaptive processes by gain or loss of genomic islands and virulence factors, as has been described for the recombinant species of P. aeruginosa, for instance (45).

Finally, the low dN/dS ratios indicate the absence of a strong positive selection pressure at the chosen loci and their suitability for population genetic studies.

Another distinct feature is the apparent difference in the tree topology of the different loci. Whereas most had a congruent topology, mutM and guaA differed, as demonstrated by the low correlation coefficient of the computed distance matrices. Such variations have been described within the genome of a species, for instance, in Haemophilus influenzae (22).

In a first approach, we attempted to identify clonal complexes with the established BURST algorithm. Despite the restricted and highly diverse strain collection, with a relaxed definition of at least four identical loci, we could already identify three complexes, two of which originated within genogroup 6, which included the largest number of isolates. These findings were also supported by the neighboring phylogenetic tree presentation (Fig. ​(Fig.1)1) of the genogroup 7 strains (isolates 889_conj and 909_bc) as well as group 6 strains (isolates 645_CF, 673_CF, 441_ts_out, 681_CF, and K279a).

There were three indications for clonality in this species. First, we found a high and significant Pearson correlation between the distance matrices calculated in the different loci except for mutM and guaA. One conceivable reason for this might be the large number of variable sites in both loci. However, two other loci with a similarly large number of variable sites (recA and ppsA) revealed high correlation coefficients. This fact constrains the argument of a low correlation due to a high number of SNPs, at least as the sole explanation. Of note, mutM and guaA were the most frequently affected loci in RDP analyses.

Second, for various subsets, calculation of the I_A revealed values different from zero, which tested significant in regard of all 65 copy strain-cleared isolates and all 48 clinical isolates. The significant I_A of all 65 isolates in comparison to a nonsignificant I_A of all 54 STs could be in line with a clonal epidemic population structure. Consideration of all 41 STs of the clinical isolates resulted in an I_A of 1.216. However, this was not significant in contrast to the I_A of all 48 clinical isolates because the observed variance was smaller than the maximal trial variance. This observation could also be in accordance with a clonal epidemic population structure with the successful genomic group 6 or be due to a number of tested strains simply too small to reach statistical significance. The absence of significant evidence for a linkage disequilibrium in the 17 isolates of environmental origin with an I_A of 2.911 can simply be due to the low number of isolates tested.

Finally, the statistical test of congruence using the maximum-likelihood approach, which in contrast to BURST analysis does not require a large MLST data set, pointed toward a clonal population structure, again with the exception of both the mutM and guaA loci.

However, based on today's species definition of S. maltophilia, there are conceivable limitations to the high overall linkage disequilibrium calculated in this study, particularly in consideration of the extensive sequence divergence of different genogroups, e.g., 5, 8, and 9 (see Table S4 in the supplemental material). Future studies could delineate further species within the boundaries of the species designated as belonging to S. maltophilia today. In this case, fixed differences between these new species and a low interspecies recombination rate could prove to be the reason for an apparently high overall linkage disequilibrium of what are considered to be S. maltophilia species today. However, it should be noted that we were also able to show a significant high I_A for the clinical isolates including genogroup 6 with the species-defining type strain and other less distantly related genogroups.

Furthermore, data were scrutinized for putative recombination events by employing different detection algorithms in the RDP suite. Our analyses identified four independent events of HGT involving one-sixth of all analyzed ST profiles. Interestingly, all of these involved isolates of genogroups 5, 8, 9, and 10 with a large DNA sequence distance to the species-defining type strain ATCC 13637^T (isolate 886_pat) in genogroup 6. These findings were corroborated by at least four independent methods for detection of the HGT. Therefore, it seems rather unlikely that the occurrence of true recombinations was misidentified or overlooked. Nevertheless, future investigations of considerably larger numbers of STs might drastically improve the statistical power of detection by the suite of methods applied here (51, 52).

In the end, we were able to confirm previously described genogroups and identify new ones. It is important to note that genogroups defined on the basis of DNA similarities can comprise strains with entirely different allelic profiles, which are already changed by introduction of one SNP in each locus and have a remaining sequence similarity of >99%. For instance, both genogroup 6 strains 325 and 396 differed in 31 SNPs out of 3,591 nucleotides, which varied from one SNP in gapA to 11 in the guaA locus. Despite entirely different allelic profiles, the strains have a sequence similarity of 99.14%. This MLST scheme provides some further evidence that the distribution of isolates from the nonanthropogenic environment, like the rhizosphere or water, apparently differs from that of humans. Like another analysis of previously described genogroups (35) containing isolates of predominantly (>75%) or entirely clinical origin (2, 6, and 7) or environmental origin (5, 8, and 9), the study presented here also found the same genogroups comprising exclusively isolates of the corresponding different ecological origins. This is reminiscent of niche separation of different ecotypes. Yet, to some extent, the small number of isolates (<10) assigned to the particular genogroups is a clear limitation of both studies. In addition, this MLST scheme confirmed a genetic relatedness among the majority of genotypes found in isolates from CF patients (11). In the previous study based on a gyrB RFLP, 90% of all the 40 CF isolates were grouped in just two clusters. However, the delineation in the study appeared to be rather fuzzy because group B corresponded to the AFLP groups 1 and 3, and group C included the AFLP groups 2, 4, 6, and 7 as well as the type strains of S. maltophilia, S. africana, and S. rhizophilia. More precisely, our study assigned 12 out of 19 isolates from epidemiological nonassociated CF patients to genogroup 6. The remaining CF isolates belonged to five other genogroups, and one could not be assigned to any of the groups. Of note, genogroup 6 included additional respiratory tract isolates from three different hospital outbreaks and the S. maltophilia type strain ATCC 13637^T. In contrast, this typing scheme could not provide evidence for particularly virulent, i.e., invasive, genogroups. Sixteen isolates from blood cultures or CSF isolates were found among eight genogroups. One might speculate that these strains gave raise to invasive infections due mainly to the impaired immune response of the host, and thus there was no need for particular virulence factors in the pathogen. The association of genogroup 6 isolates with respiratory tract specimens and CF or ICU patients may lead to the hypothesis that these strains are particularly adapted to colonization of this environment or to interacting as a specialist with particular microbial consortia prevailing in this ecological niche. Similarly, 4 out of 10 respiratory tract isolates and 2 out of 3 CF isolates have previously been described as belonging to AFLP genogroup 6 (35).

We noted a high average sequence divergence of 0.048 (± 0.016 standard deviation) between different genogroups, with a range from 0.026 to 0.088, in the calculation of k parameters to assess the significance of genogroups identified in the tree presentation. Based on published 16S rRNA gene sequence data, the percentage of sequence similarities ranged from 91.6% to 99% (mean, 96.5%) (35), which is somewhat below the species definition, which usually proposes a 16S rRNA gene sequence similarity of >97% (65). However, in the absence of characteristic phenotypes and relatively high intergroup DNA-DNA hybridization values, the authors of the previous study refrained from terming the genogroups as separate species.

In conclusion, this MLST scheme for S. maltophilia presents a discriminatory typing method with stable markers appropriate for studying the population structure. Based on DNA similarities, S. africana belongs to the species S. maltophilia. MLST data confirmed the existence of previously defined genogroups and identified an additional five new genogroups. Thus, this work provides a sequence database and a method for assigning further isolates to already defined or new genogroups and for refining population structure analyses. Initial data analyses for inferring population structure provide additional evidence for a clonal rather than a recombinant structure. However, to corroborate these findings, a greater number of isolates must be investigated with this newly established scheme, especially isolates belonging to the environmental genogroups, which are apparently just distantly related to the S. maltophilia type strain. The predominance of clinical isolates, particularly in genogroup 6, requires further elucidation, as does the strict environmental origin of isolates in genogroups 5 and 8 to 10. Further taxonomic studies are required to assess whether S. maltophilia must be separated into several distinct species.

Supplementary Material

Acknowledgments

Footnotes

REFERENCES