Mechanisms of Resistance to Folate Pathway Inhibitors in Burkholderia pseudomallei: Deviation from the Norm

ABSTRACT The trimethoprim and sulfamethoxazole combination, co-trimoxazole, plays a vital role in the treatment of Burkholderia pseudomallei infections. Previous studies demonstrated that the B. pseudomallei BpeEF-OprC efflux pump confers widespread trimethoprim resistance in clinical and environmental isolates, but this is not accompanied by significant resistance to co-trimoxazole. Using the excluded select-agent strain B. pseudomallei Bp82, we now show that in vitro acquired trimethoprim versus co-trimoxazole resistance is mainly mediated by constitutive BpeEF-OprC expression due to bpeT mutations or by BpeEF-OprC overexpression due to bpeS mutations. Mutations in bpeT affect the carboxy-terminal effector-binding domain of the BpeT LysR-type activator protein. Trimethoprim resistance can also be mediated by dihydrofolate reductase (FolA) target mutations, but this occurs rarely unless BpeEF-OprC is absent. BpeS is a transcriptional regulator that is 62% identical to BpeT. Mutations affecting the BpeS DNA-binding or carboxy-terminal effector-binding domains result in constitutive BpeEF-OprC overexpression, leading to trimethoprim and sulfamethoxazole efflux and thus to co-trimoxazole resistance. The majority of laboratory-selected co-trimoxazole-resistant mutants often also contain mutations in folM, encoding a pterin reductase. Genetic analyses of these mutants established that both bpeS mutations and folM mutations contribute to co-trimoxazole resistance, although the exact role of folM remains to be determined. Mutations affecting bpeT, bpeS, and folM are common in co-trimoxazole-resistant clinical isolates, indicating that mutations affecting these genes are clinically significant. Co-trimoxazole resistance in B. pseudomallei is a complex phenomenon, which may explain why resistance to this drug is rare in this bacterium.

IMPORTANCE Burkholderia pseudomallei causes melioidosis, a tropical disease that is difficult to treat. The bacterium's resistance to antibiotics limits therapeutic options. The paucity of orally available drugs further complicates therapy. The oral drug of choice is co-trimoxazole, a combination of trimethoprim and sulfamethoxazole. These antibiotics target two distinct enzymes, FolA (dihydrofolate reductase) and FolP (dihydropteroate synthase), in the bacterial tetrahydrofolate biosynthetic pathway. Although co-trimoxazole resistance is minimized due to two-target inhibition, bacterial resistance due to folA and folP mutations does occur. Co-trimoxazole resistance in B. pseudomallei is rare and has not yet been studied. Co-trimoxazole resistance in this bacterium employs a novel strategy involving differential regulation of BpeEF-OprC efflux pump expression that determines the drug resistance profile.
belongs to the LysR family and is located upstream of mexEF-oprN. Overexpression of MexT activates mexEF-oprN transcription (27). In B. pseudomallei, a similar arrangement exists. The gene encoding the LysR-type regulator BpeT that controls expression of the BpeEF-OprC efflux pump is located upstream of the llpE-bpeE-bpeF-oprC operon ( Fig. 2A) (30,31). However, the role of BpeT in control of BpeEF-OprC is not well understood. LysR family proteins can both activate and repress transcription depending on the interaction(s) with coinducer molecules that bind to their carboxy-terminal binding domain (32). The need for coinducer binding may be alleviated by amino acid substitutions in this carboxy-terminal domain allowing constitutive target gene expression.
Because of the pivotal role that SXT plays in treatment and prophylaxis of B. pseudomallei infections, the focuses of this study were (i) to determine and characterize the mechanisms leading to decreased SXT susceptibility using attenuated excluded (i.e., excluded from the requirements of the Federal select agent regulations; https://www .selectagents.gov/) select-agent strain Bp82 (33), (ii) to determine the molecular basis of SXT r in a collection of clinical isolates that had evolved resistance during infection, and (iii) to provide a possible explanation as to why primary SXT r is uncommon.
(Portions of this research were conducted by N. L. Podnecky and K. A. Rhodes in partial fulfillment of the requirements for a PhD from Colorado State University, Fort Collins, CO, 2013 and 2016, respectively.)

RESULTS
Characterization of TMP-resistant strains. Following passive selection of Bp82 in the presence of TMP, a collection of isolates with decreased TMP susceptibilities was obtained. Of these, 3 isolates with TMP MICs above the detection limit (Ն32 g/ml), Bp82.102, Bp82.103, and Bp82.104, were selected for further testing and characterization (see Table S1 in the supplemental material). Isolates Bp82.102 and Bp82.103 had over 10-fold increases in SXT MICs, but the observed values were below the 4 g/ml SXT r cutoff for resistance (Table 1). In addition to the folate pathway inhibitors, these two strains also had increased drug MICs for several known substrates of the BpeEF-OprC efflux pump, including acriflavine and chloramphenicol, which exhibited a consistent 2-fold increase and 8-fold increase in MIC, respectively. In contrast, susceptibility to the non-BpeEF-OprC substrates erythromycin and gentamicin remained unaltered. However, the Bp82.104 isolate did not exhibit the same changes in antimicrobial susceptibilities. We confirmed that increased expression of the BpeEF-OprC efflux pump contributes to the increased MICs observed in strains Bp82.102 and Bp82.103 using reverse transcription-quantitative PCR (RT-qPCR). Compared to wild-type Bp82, there was a significant increase in bpeF (Ͼ30-fold) and bpeT (~2-fold) mRNA expression in strains Bp82.102 and Bp82.103 (Fig. 3) but no change in Bp82.104 (data not shown). Genes shown in our studies to be involved in B. pseudomallei TMP, SMX, and SXT resistance are indicated in maroon. The folP gene (indicated in gray) was examined for mutations in this study but was not directly implicated in folate pathway inhibitor resistance. The folE gene is indicated in light green. Sequence coordinates are taken from the GenBank entries for strain 1026b (GenBank accession numbers NC_017831.1 and NC_017832.1) and from the Burkholderia database (18). Old NCBI locus tags are used throughout this paper as they are employed in the Burkholderia database. Where available, current NCBI locus tags for selected genes are provided below. (A) Organization of the llpE-bpeE-bpeF-oprC operon and associated bpeT gene on chromosome II (Chr II). The bpeE, bpeF, and oprC genes encode the membrane fusion protein, RND transporter, and outer membrane channel protein components of the BpeEF-OprC efflux pump, respectively. We confirmed that llpE is cotranscribed with the bpeEF-oprC genes in the llpE-bpeE-bpeF-oprC operon; however, its product LlpE (putative lipase/carboxyl esterase) does not affect efflux pump function, at least not export of known antibiotic substrates (T. Mima and H. P. Schweizer, unpublished data). The bpeT gene encodes a LysR-type regulator, which binds to the 188-bp bpeT-llpE intergenic region. (B) Chromosome I location of the bpeS gene encoding the LysR-type regulator BpeS, which is highly similar to BpeT. The bpeS flanking genes are annotated _I2791 (BP1026B_I2791; current NCBI tag BP1026B_RS13955, peptidase) and _I2789 (BP1026B_I2789; current NCBI tag BP1026B_RS13950, two-component regulator histidine sensor kinase). (C) Location of folM and folE on chromosome II. The folM gene (dihydromonapterin reductase) is in an operon with folE, encoding GTP cyclohydrolase (I) catalyzing the first step in the canonical tetrahydrofolate biosynthetic pathway ( Fig. 1) (54). We have shown that B. pseudomallei folE can complement an E. coli folE mutant (Fig. S4). Neighboring genes are _II0039 (BP1026B_II0039; current NCBI tag BP1026B_ RS18775, serine O-acetyltransferase) and _II0042 (BP1026B_II0042; current NCBI tag BP1026B_RS18770, LysR-type transcriptional regulator). (D) Location of the dihydrofolate reductase gene folA on chromosome I. The folA gene is flanked by pmbA, encoding a protein belonging to the peptidase U62 family, and _I0835 (BP1026B_I0835), encoding a putative 54 -dependent transcriptional regulator. In some isolates, e.g., the originally sequenced strain K96243 (55), a copy of insertion element ISBma2 is located between pmbA and folA (28). (E) Location of folP on chromosome I. The folP gene (dihydropteroate synthase) is located upstream of and probably in the same operon as glmM (phosphoglucosamine mutase). The preceding gene, ftsH, encodes a cell division protein.
These expression patterns suggest that BpeEF-OprC is responsible for resistance to TMP in Bp82.102 and Bp82.103 but not Bp82.104.
bpeT mutations cause BpeEF-OprC overexpression and TMP resistance. To determine if changes to bpeT were the cause of the observed overexpression of the bpeEF-oprC operon and of bpeT in Bp82.102 and Bp82.103, the bpeT gene was PCR amplified and sequenced. We identified two single nucleotide polymorphisms (SNPs) in bpeT resulting in amino acid substitutions in the carboxy-terminal domain of BpeT, namely, C310R in Bp82.102 and L265R in Bp82.103 (Fig. 4). Strain Bp82.104 contained no bpeT mutations. To confirm that these bpeT mutations cause overexpression of bpeF and bpeT mRNA, each mutation was individually introduced into the Bp82 wild-type background by allele replacement. The resulting strains, Bp82.268 (BpeT C310R ) and Bp82.269 (BpeT L265R ), had bpeF mRNA expression profiles similar to those found in the original TMP r isolates (Fig. 3). MIC testing of Bp82.268 and Bp82.269 showed a greater than 5-fold increase in TMP MICs (from 0.75 g/ml in Bp82 to 4 g/ml in Bp82.268 and Bp82.269) ( Table 2). However, this is below the Ն32 g/ml observed in the original Bp82.102 and Bp82.103 strains. There were also a 4-fold increase in SXT MICs and a small 2-fold increase in SMX MICs ( Table 2) that were again below those observed in the originally selected Bp82.102 and Bp82.103 strains. These data suggest that (i) both BpeT amino acid changes (C310R and L265R) result in increased expression of the BpeEF-OprC efflux pump, which in turn results in TMP resistance and reduced susceptibility to SMX and SXT; (ii) the BpeT mutations do promote bpeEF-oprC expression and resistance, not as a result of increased BpeT expression but rather because they likely activate BpeT such that gene expression occurs independently of coeffector binding; and (iii) factors other than mutations in bpeT contribute to the increased TMP, SMX, and SXT resistance levels observed in the selected Tmp r strains. Dihydrofolate reductase mutations contribute to TMP resistance. TMP r strain Bp82.104 did not contain any bpeT mutations and did not overexpress BpeEF-OprC. To identify other potential mechanisms of resistance, the folA gene encoding DHFR, the  TMP drug target ( Fig. 1 and 2D), was examined. DNA sequencing of Bp82.104 folA revealed a single SNP resulting in an I99L amino acid substitution. The putative involvement of this mutation in TMP r in clinical isolate Bp1651 was previously reported, but the involvement was not experimentally verified (34). While Bp82.103 had no changes in the folA sequence, the Bp82.102 folA gene also contained a mutation causing an F158V change. Thus, strain Bp82.102 had mutations in both bpeT and folA. Allelic replacement was used to introduce each folA SNP singly into Bp82, and TMP and SXT MICs were determined for each of the resulting strains, Bp82.183 (folA F158V ) and Bp82.184 (folA I99L ). Both folA mutations caused TMP r in the engineered mutant strains ( Table 2). There was an over 10-fold increase in the SXT MICs (from 0.094 g/ml to 0.38 g/ml), but the data from these isolates did not meet the cutoff for SXT r , suggesting that folA modification alone may not be sufficient for SXT r .
To assess the propensity for selection of folA mutations in the absence of BpeEF-OprC-mediated efflux, we also selected for TMP r in strain Bp82 Δ(bpeEF-oprC) or strain Bp82 ΔbpeT. Twelve resulting TMP r isolates capable of growth on 16 g/ml TMP were sequenced to identify folA mutations. All had the FolA I99L mutation that was identified in Bp82.104, possibly due to clonal expansion (data not shown). Together, these findings suggest that folA target mutations may be a common cause of or contributor to acquired TMP r in the presence or absence of BpeEF-OprC efflux pump expression or in other strongly selective environments.
BpeT is a transcriptional activator of bpeEF-oprC. BpeEF-OprC was previously shown to cause resistance to TMP (28), as was also observed in this study; however, the role of bpeT in transcriptional control of this efflux pump is not well understood. The genetic organization and substrate specificity of the P. aeruginosa MexEF-OprN efflux pump, which is positively regulated by the LysR-type regulator MexT, and those of BpeEF-OprC are similar, which led us to suggest that they may be regulated in analogous fashions (35). We hypothesized that BpeT overexpression would cause activation of bpeEF-oprC transcription and thus would reduce the susceptibility of B. pseudomallei to BpeEF-OprC substrates such as TMP. To test this, bpeT was cloned behind the constitutive P1 promoter on a mini-Tn7 element, which was integrated into the genome of a ΔbpeT Bp82 strain derivative, Bp82.87, to yield Bp82.187. This strain expressed bpeT and bpeF mRNA at levels that were over 20-fold higher than those seen with wild-type Bp82 (Fig. 5). Overexpression of bpeEF-oprC in this strain resulted in TMP r and reduced SMX and SXT susceptibility (Table S2). These data confirm that BpeT is an BpeT is a transcriptional activator of bpeEF-oprC. A mini-Tn7 element containing a bpeT gene whose expression is driven by the constitutive P1 promoter was integrated into the genome of Bp82.87 (Bp82 ΔbpeT) to create Bp82.187. The Bp82.189 control strain was created by introducing the empty mini-Tn7 vector into the same strain background. The bpeF (black bars) and bpeT (gray bars) mRNA levels were determined by RT-qPCR and normalized to expression in Bp82. Error bars represent standard deviations of mean expression levels. Results of statistical analysis performed by one-way ANOVA and Tukey's multiple-comparison test show that bpeF and bpeT expression in response to bpeT overexpression was significantly different from that seen with Bp82, its ΔbpeT derivative (Bp82.87), and empty vector (Bp82.189) controls. ****, P Ͻ 0.0001. activator of bpeEF-oprC transcription and that reduced TMP and SMX susceptibility is a direct result of increased BpeEF-OprC expression. However, the modest increase in SMX resistance bestowed by BpeEF-OprC expression is not sufficient to cause SXT r .
Decreased SXT susceptibility is dependent on expression of BpeEF-OprC. Mechanisms of SMX resistance in B. pseudomallei have not been described, and several attempts to select for SMX resistance in Bp82 failed for unknown reasons. We instead focused on identifying mechanisms of resistance to the SXT combination to infer potential factors affecting SMX susceptibility. Following serial passage of Bp82 in increasing concentrations of SXT, six strains with decreased SXT susceptibilities were randomly selected for further examination: Bp82.191, Bp82.193, Bp82.199, Bp82.202, Bp82.204, and Bp82.207. These isolates had TMP and SMX MICs above the limits of detection (32 g/ml and 1,024 g/ml, respectively) ( Table 1). The SXT MICs ranged from 2 g/ml to 6 g/ml, and three of the isolates (Bp82.191, Bp82.202, and Bp82.204) were classed as resistant to SXT (MIC, Ն4 g/ml). The drug MICs of SXT mutants were also increased for the known BpeEF-OprC substrates tested, acriflavine and chloramphenicol, although susceptibilities to erythromycin were unchanged and gentamicin susceptibilities were reduced 2-fold to 4-fold ( Table 1). Deletion of the bpeEF-oprC genes resulted in a major drop in the TMP, SMX, and SXT MICs (Table S3), and subsequent single-copy complementation performed with the wild-type bpeEF-oprC operon expressed by the P tac promoter caused increases in the MICs for TMP, SMX, and SXT, though not to the levels observed with the original isolates (Table S3). A possible explanation is that P tac may not give the same high-level BpeEF-OprC expression as the native promoter in the presence of the bpeS mutations contained in these strains. These data show that the BpeEF-OprC efflux pump plays a significant role in SXT resistance.
Overexpression of BpeEF-OprC in SXT-resistant isolates is BpeT independent. RT-qPCR was used to determine the relative expression levels of bpeT and bpeF mRNA in the six isolates (Bp82.191 to Bp82.207) with reduced SXT susceptibility both with and without the BpeT transcriptional regulator. Expression of bpeT mRNA was detected in all six isolates relative to Bp82, but intraisolate bpeT expression level differences were not statistically significant (Fig. 6A). Remarkably, each of the isolates had bpeF mRNA levels that were over 120 times higher than those seen with Bp82 (Fig. 6B). The bpeF mRNA expression levels were also determined in derivatives lacking bpeT. We observed a significant decrease in bpeF mRNA expression in the absence of bpeT, but bpeF mRNA was still overexpressed by at least 30-fold relative to Bp82 (Fig. 6B). This suggests that, while BpeT enhances expression of the bpeEF-oprC operon and BpeT overexpression causes constitutive BpeEF-OprC expression and decreased susceptibilities to several antimicrobials, it is neither essential for expression nor the only transcriptional regulatory component for this efflux pump.
SXT-resistant isolates contain mutations in bpeS encoding a novel LysR-type regulator and folM. Using targeted DNA sequencing of strains Bp82.191 to Bp82.207, we found no mutations in bpeT or in the bpeT-llpE intergenic region, which contains predicted regulatory sequences for both bpeT and the llpE-bpeE-bpeF-oprC operon ( Fig. 2A). Additionally, we did not find any mutations in the folA and folP genes, encoding the targets for TMP and SMX, respectively ( Fig. 2D and E). Whole-genome sequencing of Bp82.202 and Bp82.204 and comparison to the Bp82 parent revealed two SNPs present in both of the mutants. The first mutation was found in BP1026B_I2790 (BPSL0731; strain K96243 annotation) (Fig. 2B), a gene annotated as encoding a LysR-type transcriptional regulator (named BpeS here because of its high similarity to BpeT). BpeS (324 amino acids) showed 61.7% identity overall to the 334-amino-acid BpeT protein (Fig. 4). This identity level increases to 90% over the first 60 amino-terminal amino acids, including the respective predicted helix-turn-helix DNA-binding domains, which are 90% identical and 100% similar. This suggests that BpeT and BpeS likely bind to similar regulatory sequences. The mutation to bpeS results in an amino acid change, K267T, in the putative carboxy-terminal effector-binding domain of the regulator.
The second SNP was found in the BP1026B_II0040 (BPSS0039; strain K96243 annotation) gene, which is annotated as ptr1 (pteridine reductase 1). This gene likely corresponds to the folM gene found in other bacteria and is referred to as folM from this point on (Fig. 2C). The mutation causes a V15G amino acid change in the aminoterminal domain of FolM. This amino acid change lies within the predicted NADPHbinding site consensus sequence of the protein and thus likely affects its activity (see Fig. S1 in the supplemental material), although this has yet to be shown.
Both the BpeS K267T and FolM V15G mutations were confirmed in strains Bp82.202 and Bp82.204 by targeted Sanger sequencing of PCR-amplified DNA fragments containing the respective mutant genes. Additionally, DNA sequencing of these genes showed that all six SXT-selected isolates-Bp82.191 to Bp82.207-contained the BpeS K267T mutation and that all except Bp82.193 contained the FolM V15G mutation. Bp82.193 instead had a single base deletion at folM nucleotide position 203 causing a frameshift mutation after amino acid 67 and early termination of the protein following residue 92 (Fig. S1). Bp82.193 is phenotypically similar to the other strains in all aspects, supporting the notion that the the FolM V15G mutation may indeed deleteriously affect protein function.
Mutations in bpeS and folM contribute to decreased SXT susceptibility. To confirm the role of the identified mutations in bpeS and folM in decreased SXT  Table 3). Repair of the bpeS gene (Bp82.246 and Bp82.249) caused a greater reduction in MIC, where the susceptibilities of the repaired mutants matched that of the parental strain, Bp82. Repair of the folM gene (Bp82.247 and Bp82.250) reduced the SMX MIC from the detection limit of 1,024 g/ml to 32 g/ml and the SXT MIC from 4 to 6 g/ml to 0.75 g/ml but did not affect the TMP MIC, which remained above the Ն32 g/ml detection limit. For unknown reasons, repair of both SNPs (Bp82.248 and Bp82.251) resulted in MICs below those of the original Bp82 parent strain.
BpeEF-OprC expression was analyzed in strains engineered to contain bpeS K267T or folM V15G or both. Introduction of bpeS K267T into bpeS WT strain Bp82 resulted in Ͼ100fold increased expression of bpeF, and, as expected, there was no change in bpeF expression seen with the introduction of the folM V15G mutation (Fig. 7A). The same bpeF expression patterns were observed in Bp82.202 and Bp82.204 with neither mutation  repaired and in their derivatives that contain only the folM V15G mutation (Bp82.247 and Bp82.250) (Fig. 7B).

BpeS is a conditional transcriptional activator of BpeEF-OprC. Due to the high degree of similarity between BpeT and BpeS, we assessed the role of BpeS in BpeEF-OprC expression specifically via activation of llpE-bpeE-bpeF-oprC operon transcription.
To this end, mini-Tn7 vectors were constructed where bpeS was constitutively expressed from the P1 promoter. The mini-Tn7-P1-bpeS construct was transposed into ΔbpeS strain Bp82.264 to create Bp82.289 or into ΔbpeS ΔbpeT strain Bp82.286 to create Bp82.288. RT-qPCR analyses indicated that bpeS mRNA levels in these strains were increased~12-fold compared to the Bp82 results (data not shown). However, this BpeS overexpression affected neither the expression of bpeF nor the host strain's susceptibility to TMP, SMX, or SXT (Table S2). In contrast, a ΔbpeS strain containing a chromosomally inserted mini-Tn7-P1-bpeS K267T construct (Bp82.320) and a ΔbpeS ΔbpeT strain containing the same construct (Bp82.321) overexpressed BpeEF-OprC with bpeF mRNA levels Ͼ100-fold higher than those seen with Bp82 (Fig. 8A). This overexpression of BpeEF-OprC resulted in TMP, SMX, and SXT resistance (Table S2). These results corroborate the findings described above in that both TMP and SMX are substrates of BpeEF-OprC. They also suggest that altered BpeS function (i.e., activation of bpeEF-oprC transcription independently of the presence of a coinducer), rather than an increase in bpeS transcription, is responsible for differential levels of expression of the llpE-bpeE-bpeF-oprC operon.
SXT-resistant clinical isolates contain folM and BpeEF-OprC regulatory mutations. SXT resistance in clinical isolates is rare but can occasionally occur (23,24,36). To assess whether any of the genes mutated in laboratory-selected SXT r strain Bp82 derivatives also contained mutations in SXT r clinical isolates, we analyzed genes of the few SXT r strains identified to date by targeted sequencing or genome sequencing or both. Strain pair 354b (SXT s ) and 354e (SXT r ) represents two sputum isolates obtained from the same patient 75 months apart (30,31). Phenotypically, these strains differ in that the 354e relapse isolate exhibits a multidrug resistance phenotype consistent with BpeEF-OprC expression, likely due to the observed truncation of BpeT (30,31) (Fig. 4). Strain 354e also contains a previously noted 7-bp deletion in folM which frameshifts FolM after amino acid 20 (30) (Fig. S1). Strains MSHR664 (SXT s ) and MSHR663 (SXT r ) are sequential isolates obtained from a relapse melioidosis patient in Darwin, Australia (37). In addition to an SNP that causes an H278Y amino acid substitution in BpeT (Fig. 4), MSHR663 contains the same 7-bp deletion in folM that is present in SXT r Thai isolate 354e (Fig. S1). Australian strains MSHR8441 and MSHR8442 are SXT r isolates from a cystic fibrosis patient (36,38). Although not noted at the time, both MSHR8441 and MSHR8442 contain an SNP causing an R163L amino acid substitution in BpeS in comparison to 1026b (Fig. 4) and a 9-bp in-frame insertion in folM (Fig. S1). The latter mutation duplicates FolM amino acids 20 to 22. Lastly, strain 5041a is an SXT r sputum isolate from a Thai melioidosis patient. The folM gene of 5041a contains the same 7-bp deletion present in strains MSHR663 and 354e. Compared to several SXT s strains, a single unique SNP in bpeS was identified in 5041a that caused a P29S amino acid substitution in the putative helix-turn-helix DNA-binding domain of BpeS (Fig. 4). Of note is that two SXT r strains, 1374a from Thailand (23,24) and Bp1651 from Australia (34), did not contain any bpeT or folM mutations but rather only the bpeS P29S allele. Several lines of evidence indicate that the BpeS P29S substitution plays a role in BpeEF-OprC expression and thus in SXT resistance. First, this mutation is present in multiple SXT r strains, including 5041a and 1374a from Thailand and Bp1651 from Australia. Second, we observed by RT-qPCR that bpeF is highly expressed in 5041a and 1374a, i.e., at levels that are 78-fold and 118-fold higher, respectively, than that seen with 1026b; a similar high level of expression was observed in a Bp82 derivative, Bp82.284, which contains chromosomal bpeS P29S (Fig. S2). Third, overexpression of bpeS P29S in a Bp82 ΔbpeS strain (Bp82.320) or a Bp82 ΔbpeT ΔbpeS strain (Bp82.321) led to high-level bpeF mRNA expression (Fig. 8B) and SXT resistance (Table S2). Fourth, a bpeEF-oprC deletion abolishes TMP r , SMX r , and SXT r in both strain 1374a and strain 5041a (Table S3). These results further support the notion that altered BpeS function and not an increase in protein levels is responsible for the observed changes in BpeEF-OprC expression levels.

DISCUSSION
B. pseudomallei is intrinsically resistant to numerous antimicrobials; however, it is generally susceptible to drugs used in the treatment of melioidosis, including the tetrahydrofolate biosynthetic pathway inhibitor combination SXT (9,10). SXT is the drug of choice for eradication-phase melioidosis treatment and postexposure prophylaxis of accidental laboratory exposures (9,11). SXT-resistant isolates are rare in nature but are of concern as SXT resistance significantly reduces available treatment options (23). The main purposes of this study were to explore and characterize the molecular had no effect on bpeF transcription; however, bpeT levels were slightly increased in Bp82.323, but this expression was not statistically significantly different from that seen with Bp82. mechanisms responsible for antimicrobial resistance to the clinically relevant tetrahydrofolate pathway inhibitors TMP and SMX and to better understand why SXT r resistance is uncommon in B. pseudomallei.
Our findings illustrate that the nature of resistance to SXT in B. pseudomallei is complex and does not fit the pattern typically seen in other SXT r bacteria. Typically, bacterial resistance to TMP and SMX, the two components of SXT, is the result of mutations to the DHFR (FolA) and DHPS (FolP) drug targets (Fig. 1) (12, 20). While we have shown that mutation of FolA alone can cause high-level TMP r in laboratoryselected mutants, SMX resistance due to folP mutations could not be demonstrated. This finding is paralleled by observations made with clinical isolates where the presence of FolA mutations, e.g., the FolA I99L mutation in SXT r Australian strain Bp1651 (34), had previously been implicated in TMP r but where folP mutations have yet to be identified. In this study, we showed that SXT r in B. pseudomallei employs a novel strategy of regulation of BpeEF-OprC efflux pump expression that employs two closely related LysR-type transcriptional regulators, BpeT and BpeS. Further contributing to SXT r is FolM, a folate pathway-associated protein whose function is not yet well understood and which has not been previously been implicated in folate inhibitor resistance.
In vitro exposure to either TMP or SXT selects for regulatory mutants that constitutively express BpeEF-OprC. However, the natures of the regulatory mutations and the levels of efflux pump expression selected by the two drugs are different and this differential regulation of expression determines the drug resistance profile. Exposure to TMP alone selected for mutations affecting the putative carboxy-terminal effectorbinding domain of BpeT. These are likely activating mutations that relieve the protein's dependence on interaction with an unknown coinducer, possibly a pump substrate. Constitutive BpeEF-OprC expression in the TMP r isolates was accompanied by differences in the SMX MICs, but this level of expression is insufficient to cause SMX r and thus to cause clinically significant SXT r . Exposure to increasing concentrations of SXT led to SXT resistance. This resistance is largely due to constitutive overexpression of BpeEF-OprC, likely by activated mutant BpeS. The underlying mutations affect either the carboxy-terminal domain or the amino-terminal helix-turn-helix DNA-binding domain of BpeS. The constitutive overexpression of the BpeEF-OprC efflux pump mainly confers TMP resistance but also confers resistance to SMX and thus to SXT.
Despite that these results clearly implicate BpeT and BpeS in regulation of bpeEF-oprC gene expression, at present we do not fully understand the involvement of two LysR-type regulators in regulation of BpeEF-OprC expression. What we do know is that neither regulator is essential for control of bpeEF-oprC gene expression. If either bpeT or bpeS were an essential activator of pump gene expression, loss of one or both regulatory genes would result in abrogated pump expression, which is not what we observed in induction studies. These studies showed that bpeF expression was induced to similar levels in the wild type and in a bpeS bpeT double mutant by the BpeEF-OprC substrates chloramphenicol and doxycycline (see Fig. S3 in the supplemental material). Wild-type-like induction in the absence of both BpeT and BpeS suggests that neither regulator is solely responsible for expression of bpeEF-oprC. These data support the notion that an additional regulatory factor is responsible for substrate-mediated pump induction.
We have shown that purified BpeT and BpeS, as well as representative mutant derivatives, e.g., BpeT with carboxy-terminal mutations, BpeS K267T , and BpeS P29S , bind to the same sequence corresponding to the bpeT-llpE intergenic region (K. A. Rhodes and H. P. Schweizer, unpublished data). Thus, both regulators and their mutant versions affect transcription of the llpE-bpeE-bpeF-oprC operon directly, but likely independently, which suggests that they might activate BpeEF-OprC expression in response to different stimuli. This idea is especially interesting when combined with the fact that RND efflux systems such as BpeEF-OprC are adaptation mechanisms needed for response to environmental stressors other than antimicrobials (39)(40)(41)(42). The expression of multiple, highly contextual regulatory proteins may allow flexible employment of BpeEF-OprC to manage multiple environmental conditions. The finding that bpeS mutations seem to drive greater efflux gene expression than bpeT mutations perhaps reflects greater affinity of the respective proteins for the common binding site.
Involvement of several transcriptional regulators in efflux pump expression is well documented. For instance, expression of the MexAB-OprM efflux pump in P. aeruginosa is controlled by at least eight regulators, five of which bind in the regulatory region of the mexAB-oprM operon (reviewed in reference 43). One of these regulatory proteins, MexT, the LysR-type transcriptional activator of MexEF-OprN, the P. aeruginosa homolog of BpeEF-OprC, exerts a negative effect on MexAB-OprM expression, but the underlying mechanism remains largely unknown (44). Such "cross talk" between efflux systems is not uncommon and has been proposed for AmrAB-OprA, BpeAB-OprC, and BpeEF-OprC in B. thailandensis although the underlying mechanisms were not investigated (45). Although not explored in this study, the unexpected lowering of gentamicin resistance in the SXT r mutants can possibly be attributed to a negative effect that BpeEF-OprC may exert on AmrAB-OprA expression. However, gentamicin MICs remain well above the clinical breakpoints for susceptibility that would enable potential clinical use of gentamicin and or other aminoglycosides for treating SXT r B. pseudomallei.
In addition to BpeS and BpeT, all strains examined in this study (except one) with increased SXT resistance also contained mutations in the annotated ptr1 gene in B. pseudomallei, which is likely folM and whose true role in the biology of B. pseudomallei and contribution to SXT resistance remain to be established. Although FolM was originally postulated to function as an alternative DHFR (13), subsequent studies suggested that this is not its physiological function. A more likely scenario is that in bacteria expressing FolA, FolM does not function as a DHFR, as dihydrofolate is a poor substrate for purified FolM and its intracellular levels are kept extremely low by FolA (14). The physiological function now attributed to FolM is reduction of dihydromonapterin to tetrahydromonapterin (14) (Fig. 1). In bacteria expressing phenylalanine hydroxylase (PhhA), e.g., Burkholderia species, and thus catalyzing the conversion of phenylalanine to tyrosine, terahydromonapterin is a required PhhA cofactor, and FolM is required for its synthesis (14). Tetrahydromonapterin levels often outrank folate levels as an end product of pterin biosynthesis, and tetrahydromonapterin synthesis would establish competition between tetrahydropterin and folate synthesis (14). In this scenario, inactivation of FolM by mutation would increase the level of substrates flowing through the folate biosynthetic pathway, possibly resulting in decreased susceptibilities to folate pathway inhibitors. This notion is supported by the well-established findings showing that hyperproduction of p-aminobenzoic acid and increased substrate flux through the folate biosynthetic pathway represent a documented mechanism of sulfonamide resistance (46). Our data are consistent with the notion that mutations that compromise FolM function contribute to but are not alone sufficient to confer TMP, SMX, and SXT resistance.
Our findings determined with laboratory-selected SXT r strains are corroborated by data obtained with clinical B. pseudomallei isolates. First, it has been postulated that the previously introduced B. pseudomallei SXT r clinical isolate 354e likely expresses BpeEF-OprC due to a truncation of the BpeT carboxy terminus, and this strain has been shown to exhibit an SXT r phenotype (30). Second, SNPs in bpeS affecting protein integrity are also found in SXT r clinical isolates. Third, all but one of SXT r clinical B. pseudomallei isolates studied to date also possess mutations in folM.
In conclusion, our data show that acquired TMP and SXT resistance in B. pseudomallei is multifactorial and reflects complex regulation of BpeEF-OprC efflux pump expression and interplay with known and novel folate and pterin pathway constituents. A more complete understanding of these mechanisms will require further studies of the central and yet likely differential roles of BpeS and BpeT in bpeEF-oprC gene expression, as well as those of folM and its gene product. Knowledge of the factors involved in the SXT resistance seen in B. pseudomallei is vital for its rapid identification in clinical settings and for forensic applications. The finding that achieving even modest levels of SXT r requires multiple mutations provides a likely explanation for the rarity of clinically occurring SXT-resistant mutants despite lengthy eradication-phase therapy (23,24). followed by DNA sequencing to confirm insertion or repair of SNP mutations. Details of plasmid and strain construction are provided in Text S1.
Deletion mutant complementation. Deletion strains were complemented with a bpeT or bpeEF-oprC gene(s) originating from strain 1026b or Bp82 utilizing the mini-Tn7 system, which allows stable and site-specific single-copy insertions into the B. pseudomallei genome at three possible glmS-associated attTn7 sites (50). The respective mini-Tn7 elements, along with an empty-mini-Tn7 element used as a control, were transferred to the target B. pseudomallei strains either via conjugation from E. coli or by electroporation, and glmS-associated insertions were verified as previously described (50,51). Mini-Tn7 insertions at the glmS2-associated attTn7 site were routinely retained for further studies, unless noted otherwise. The inducible E. coli trp/lac operon hybrid P tac promoter was used for regulated expression of the bpeEF-oprC genes. BpeEF-OprC expression was induced by addition of 1 mM isopropyl-␤-Dthiogalactopyranoside (IPTG; Gold Biotechnology, St. Louis, MO) (28).
RNA extraction and reverse transcription-quantitative PCR (RT-qPCR). Expression levels of mRNA of target genes were analyzed in bacteria grown to mid-log phase (OD 600 ϭ 0.6 to 0.8) in LB. RNA was isolated using an RNeasy Protect Bacteria minikit (Qiagen, Valencia, CA) in biological triplicate, as previously described (28). Relative expression levels were determined using gene-specific primer sets (Table S4) in technical triplicate. 23S rRNA was used for normalization, and relative fold expression compared to that seen with the parental strains was determined using iCycler iQ Optical System software version 2.0 (Bio-Rad, Hercules, CA) with experimentally defined amplification efficiencies for each primer set. Expression values were pooled between biological replicates, and relative expression data were analyzed by two-way analysis of variance (ANOVA) followed by either Tukey's multiple-comparison test or Dunnet's posttest using GraphPad Prism (GraphPad Software, Inc., La Jolla, CA). P values of Ͻ0.05 were considered significant.