2.1. Expansion of the DfrB FamilyFollowing the recent identification and characterization of two new DfrB family members (DfrB10 and DfrB11) [12], our objective was to investigate whether further new DfrB homologues could be identified in environments that are less likely to be influenced by human activity. To this end, we queried the JGI/IMG metagenomic database and identified over 3000 dfrB gene homologues, from which ten sequences sharing 63 to 92% protein sequence identity were selected to be representative of sequence diversity (Figures S1 and S2). They were defined as DfrB12–DfrB21. High sequence identity of their SH3-like domain suggests that these DfrB12–DfrB21 should fold and tetramerize in a manner analogous to known DfrB enzymes, thus conferring high TMP resistance as a result of their Dfr activity. To investigate this, the minimal inhibitory concentration (MIC) of TMP for E. coli expressing DfrB12–DfrB21 was characterized, followed by determination of Dfr activity in E. coli lysate.Remarkably, all homologues except DfrB12 provided TMP resistance in E. coli comparable to that of DfrB1, up to the highest soluble concentration of TMP (600 µg/mL) (Figure 2A). Furthermore, activity was clearly observed in clarified lysate of E. coli for all DfrB homologues except DfrB12, which conferred resistance up to 150 µg/mL of TMP (Figure 2B). This apparent discrepancy results from little Dfr activity being required to sustain microbial proliferation, such that MIC assays are more sensitive than activity assays in crude lysate [12].The lower TMP resistance and Dfr activity of DfrB12 relative to all other DfrB family members is most likely due to the Q67H substitution in the active-site VQIY tetrad (Figure S1). The Q67H mutation has been previously investigated: the mutation improves binding to both DHF and NADPH by 1–2 orders of magnitude compared to the native enzyme [24]. This favors the formation of the nonproductive DHF⋅DHF substrate or NADPH⋅NADPH cofactor complexes, resulting in an important decrease in activity. We note that this lower activity is sufficient to confer some TMP resistance. On the contrary, the 5- to 10-fold lower activity of DfrB14 and DfrB15 relative to DfrB1 is sufficient to confer the highest level of TMP resistance that we can measure. This work having been performed on crude lysate, we have not determined whether the reduced activity results from sequence variation outside of the conserved VQIY active site or other factors, such as reduced expression or stability.

These results confirm that DfrB12–DfrB21 constitute new DfrB family members. This demonstrates that identification of sequences sharing high sequence identity with dfrB1–dfrB11, including functionally and structurally important residues, is sufficient to identify new DfrB family members. This knowledge will facilitate robust identification of DfrB homologues in the future.

2.2. Genomic Context Analysis of DfrB12–21The DNA sequences containing the newly identified dfrB genes originated from samples isolated from diverse environments not directly associated with the use of antibiotics (Table 1). Consistent with previous studies, dfrB genes were found in Proteobacteria [12]. As the identification of dfrB in diverse environments suggests their widespread presence, we investigated the mobility of the dfrB12–dfrB21 genes by determining whether mobile genetic elements (MGEs, e.g., plasmids, transposons, or integrons) were present in their vicinity. Other antibiotic resistance genes (ARG) were also sought, because a major public health concern is transmission of ARGs associated with MGEs in pathogenic bacteria [25]. To allow comparison to dfrB1–dfrB11, mostly isolated from samples associated with human activity, we characterized the genomic context of dfrB1–dfrB11 according to the same criteria (Figure S3).First, sequences were classified as plasmidic or chromosomal using PlasForest and PlasFlow (Table S1) [26,27]. The resulting predictions obtained were often contradictory, such that it was difficult to conclude on their organization. The poor quality of predictions was expected, since the majority of analyzed contigs in that dataset were too short (26].The association of dfrB genes with integrons and transposon insertion sequences (IS) was investigated using IntegronFinder and ISFinder, respectively (Table 1) [28,29]. These tools rely on frequently updated databases as references and enable robust and precise identification of MGEs [28,29]. No contig contained transposon IS, but incomplete integrons (CALIN) were identified in the vicinity of dfrB12, dfrB19, and dfrB20 [30]. Both dfrB12 and dfrB19 were integrated within a CALIN element, indicative of potential mobility of those two genes. The dfrB20 gene was found outside of the CALIN identified in its genomic context, the longest obtained (nearly 10 kbp). Analysis of its content using BLASTP indicated 15 hypothetical or metabolism-associated proteins. Though not indicative of mobility of that dfrB, it demonstrates that genetic mobility occurred in the vicinity of the gene. Overall, this dataset contains evidence of genetic mobility in at most three among the dfrB12–dfrB21 genes, in contrast with our earlier findings based on samples closely linked to human activity [12].The Resistance Gene Identifier (RGI) tool from the Comprehensive Antibiotic Resistance Database (CARD) was used to assess the association of dfrB12–dfrB21 with multidrug resistance (MDR, Table 1) [31]. In contrast to dfrB1–dfrB11, mostly identified in environments associated with the use of antibiotics and generally associated with MGE in a variety of MDR contexts (Table S3) [12], no ARGs were identified in this dfrB12–dfrB21 dataset.A clear limitation of the current dataset is the short length of the contigs (Table 1). Most genetic contexts were of insufficient length to allow identification of additional genetic features with confidence, indicating that analyses on longer contexts should be conducted. 2.3. The Broader DfrB Sequence Space Includes DfrB of ConcernTo gain further information on the genetic context of the dfrB gene family, we identified ten further putative dfrB from a BLASTP search conducted in NCBI (referred to as putative dfrB B1–B10) and ten more from the above-described metagenomic JGI/IMG database search (referred to as putative dfrB C1–C10). We selected sequences with analyzable genomic context (>1 kbp) identified from environments that are not directly associated with the use of antibiotics (e.g., river sediments, soil), although some may be associated with human activity (e.g., polluted river sediment, wastewater). One sequence from a clinical sample (B5) was included as a basis for comparison. Although these new putative DfrB homologues were not functionally characterized, high sequence identity with dfrB1–dfrB21 (63–92%) and conservation of all structurally and functionally important residues are consistent with their being functional DfrB homologues (Figure S3).All sequences were predicted as chromosomal by PlasForest (Table 2), consistent with recent findings [12]. Complete integrons containing a dfrB gene (putative dfrB B1, B2, B5) and proximal transposases (putative dfrB B1, B5) were found only in contigs from samples collected in environments associated with human activity (Table 2). Strikingly, putative dfrB B1, B2, and B5 were also all associated with MDR (Table 2). Notably, previously known dfrB from clinical samples (dfrB1–5, 9–10; Table S3) are all associated with clinical integrons and are in MDR contexts [12]. This is consistent with human-associated settings procuring higher TMP selective pressure, thus inducing mobilization of dfrB and acting as reservoirs for ARGs [32,33,34]. Our findings suggest that these MGEs have contributed to propagating dfrB from diverse sources into clinically relevant settings.Additionally, putative dfrB B3 and B6 from water samples are from environments linked to human activity; they were found in MDR contexts but were not associated with MGEs (Table 2). This suggests vertical transmission or loss of mobility after acquisition of ARGs [35]. This was also the case for putative dfrB C3, which was isolated from soil in the Loxahatchee National Wildlife Refuge. The refuge accommodates a wide variety of recreational activities, although it is in a remote location; the relation of the sample to human activity is plausible but is not clear. Most ARGs found in the vicinity of putative dfrB B1–B3, B5-B6, and C3 are related to aminoglycoside resistance (aadA16, AAC(6′)-IIa, ParS, aadA, baeS), consistent with previous findings for dfrB of clinical origin [12]. Association with resistance to rifampin (arr2), chloramphenicol (catB3, cmlA6), beta-lactam (OXA-21), fosfomycin (fosX), polycationic antibiotics (parS), and macrolides (mtrA) was also noted. This demonstrates association of putative dfrB with MDR in environments linked to human activity beyond clinical contexts.Conversely, indications of genetic mobility were found in the genomic context of putative dfrB B4 and B10 isolated from soil and putative dfrB C1 isolated from freshwater sediment, without association with MDR (Table 2). Strikingly, whereas analyses using CARD reveal no ARGs associated with those putative dfrB, BLASTP analyses of the integron content in the vicinity of putative dfrB C1 indicate the presence of ten proteins associated with metabolism and detoxification. This suggests the coevolution of putative dfrB C1 with metabolism- and defense-associated genes, rather than with antibiotic resistance genes. The remaining putative dfrB isolated from soil (C5, C6, C8–C10) and from water samples (B7, B8, C2, and C7) were not associated with MGEs or MDR (Table 2). These findings suggest that DfrB may confer an evolutionary advantage in environmental contexts that are not directly associated with the use of antibiotics.All putative dfrB genes were isolated from Proteobacteria as for all known dfrB, most of which have been reported in clinical settings, often linked to ARGs and mobility (Table S3) [12]. Our findings highlight a new pattern: with few exceptions, the 12 putative dfrB genes identified in settings that are not associated with human activity were not associated with ARGs or mobility. Exceptions include the putative dfrB C4 linked with β-lactam resistance (OmpA) and an incomplete transposase (Table 2) and observation of dfrB7 in a clinical integron (Table S3), despite both having been isolated from environmental sources. These examples could be the result of environmental contamination with clinically relevant pathogens [36]. Inversely, putative dfrB B9 was found in a wastewater sample but is not associated with MDR nor MGEs.

As the isolation sources and genomic contexts of these putative dfrB are heterogeneous, more information is needed to conclude on the influence of environment on mobility and prevalence of dfrB. Nonetheless, the association of dfrB from environmental sources with MGEs and/or ARGs demonstrates that the broader DfrB sequence space includes DfrB of concern and justifies the need to closely monitor them.

2.4. DfrB Genes with Similar Level of Mobility Share Closer Evolutionary RelationshipsThe identification of dfrB genes in various settings led us to investigate whether closer phylogenetic relationships exist between dfrB isolated from similar environments because of a higher likelihood of horizontal gene transfer [37]. To this end, the phylogeny of dfrB1–dfrB21 and the 20 putative dfrB (B1–B10, C1–C10) was reconstructed using IQ-Tree [38].These results highlight evolutionary proximity between sequences that have similar levels of mobility (Figure 3). For instance, most dfrB contained in integrons and associated with MDR (dfrB1–dfrB5, dfrB9, dfrB10, putative dfrB B1, dfrB B2, dfrB B5) share their closest ancestor with another integron-associated sequence.These results also indicate evolutionary proximity between sequences from similar environments in the absence of indicators of mobility. For instance, pairs of dfrB from terrestrial samples (dfrB16 and dfrB19; dfrB12 and dfrB17; putative dfrB C8 and B10) share their most proximal common ancestor, although none hold clear markers of genetic mobility (Figure 3). This could suggest high conservation of the dfrB sequence owing to similar selection pressures from a similar environment and/or loss of mobility of an ancestral dfrB. In contrast, dfrB from samples isolated in aquatic or wastewater that are not associated with mobility are evenly distributed throughout the tree, suggesting that various evolutionary paths define their relationships.Because all dfrB analyzed were found in Proteobacteria (Tables S2 and S3), it is difficult to distinguish events that are due to taxonomy from those due to horizontal gene transfer in our reconstructed phylogeny. Interestingly, dfrB from the same genus (e.g., Rhodoferax sp., putative dfrB B3 and B9) are not associated with mobility and do not share a close common ancestor. This could reflect different evolutionary pressures from different environments, as putative dfrB B3 was isolated from groundwater, whereas putative dfrB B9 was isolated from activated sludge. This also indicates that dfrB genes can exist in bacterial strains that are not typically associated with clinical settings [39], suggesting that DfrB enzymes could confer an evolutionary advantage in environmental contexts.