Mutations in dfrA and its upstream region are associated with para-aminosalicylic acid resistance in Mycobacterium tuberculosis

Abstract

Objectives

para-Aminosalicylic acid is an antifolate prodrug activated in the folate biosynthesis pathway of Mycobacterium tuberculosis (Mtb), ultimately inhibiting dihydrofolate reductase encoded by dfrA. Mutations in dfrA coding and upstream regions have not been thoroughly investigated. This study aimed to assess mutations in dfrA and its upstream regulatory region in clinical Mtb isolates.

Methods

We retrieved and analysed 6547 whole-genome sequences of Mtb clinical isolates. We identified 29 SNPs in the dfrA coding region and eight mutations in the thyA–dfrA upstream region. The most prevalent coding mutations and upstream mutations were selected for functional evaluation using thermal stability analysis, GFP reporter assays and para-aminosalicylic acid drug susceptibility test (DST).

Results

The coding mutations Trp142→Gly, Gln28→Leu and Arg32→Pro were analysed for protein thermal stability and para-aminosalicylic acid susceptibility. Gln28→Leu showed increased thermal stability relative to the WT, whereas Trp142→Gly and Arg32→Pro exhibited reduced stability. Consistently, para-aminosalicylic acid DST demonstrated that Gln28→Leu conferred an 8-fold increase in para-aminosalicylic acid resistance, while Trp142→Gly and Arg32→Pro remained phenotypically susceptible. Moreover, GFP reporter assays revealed that G70T, G59T and C48A increased promoter activity by 13-fold, 4-fold and 3-fold, respectively, whereas the remaining mutations showed no significant effect. Findings were verified by para-aminosalicylic acid DST, where G70T, G59T and C48A showed para-aminosalicylic acid resistance compared with WT.

Conclusions

By integrating genomic analysis with functional assays of thermal stability, promoter activity and drug susceptibility, this study demonstrated that mutations in dfrA coding and upstream regulatory regions contribute to para-aminosalicylic acid resistance in Mtb clinical isolates.

Introduction

TB remains one of the leading causes of death worldwide due to a single infectious agent. It is caused by the Mycobacterium tuberculosis (Mtb) complex and is primarily transmitted via the respiratory route, facilitating rapid person-to-person spread. According to the WHO Global Tuberculosis Report 2024, approximately one-quarter of the global population is infected with TB. Each year, more than 10 million people develop active TB, with incidence rising since 2021. In 2023 alone, TB caused an estimated 1.25 million deaths, including 0.161 million individuals coinfected with HIV.¹

The global TB burden has been further exacerbated by the increasing emergence of drug-resistant Mtb strains. Recent estimates indicate that deaths attributable to drug-resistant TB account for approximately 13% of all antimicrobial resistance (AMR)-related deaths worldwide.² The prevalence of MDR-TB, or rifampicin-resistant TB, is estimated at 3.2% among newly diagnosed patients and 16% among previously treated cases.¹ The rise of MDR-TB severely undermines the effectiveness of existing chemotherapy regimens, highlighting the urgent need for optimized and alternative therapeutic strategies.

Para-Aminosalicylic acid, first introduced for TB treatment in 1946, was widely used for several decades. As early as 1952, para-aminosalicylic acid was combined with streptomycin to prevent the emergence of drug resistance.³ However, following the discovery of more potent and better-tolerated drugs such as isoniazid and rifampicin, para-aminosalicylic acid use was largely discontinued due to its comparatively adverse side-effect profile. The resurgence of MDR-TB in the 1990s necessitated the reintroduction of para-aminosalicylic acid as a second-line anti-TB agent.⁴ Currently, para-aminosalicylic acid is incorporated into longer MDR-TB treatment regimens as a group C drug, when other options to compose a regimen are not possible.⁵ Despite its long history of clinical use, the precise molecular mechanisms underlying para-aminosalicylic acid resistance remain poorly understood.

para-Aminosalicylic acid is an antifolate prodrug that is activated by dihydropteroate synthase and dihydrofolate synthase within the Mtb de novo folate biosynthesis pathway, ultimately inhibiting dihydrofolate reductase (DHFR)⁶ (Figure 1a). Disruption of this activation process, most notably through mutations in folC, impairs para-aminosalicylic acid bioactivation and confers resistance.⁷ Several para-aminosalicylic acid resistance–associated genes, such as folC, thyA, dfrA, ribD and thyX, have been shown to encode enzymes involved in folate metabolism in Mtb.^7–11 Studies using spontaneous resistant mutants of Mtb H37Rv, H37Ra and Mycobacterium bovis BCG have consistently identified folC as a major resistance determinant.^7,9 In contrast, resistance-associated mutations in thyA, dfrA and ribD were initially identified through transposon mutagenesis in M. bovis BCG.⁸

Figure 1.

Folate pathway in Mtb and associated genes. (a) Partial folate pathway in Mtb. (b) Genomic organization of the dfrA–thyA locus showing two transcription start sites (TSSs). TSS-1 is located at +9 relative to the dfrA start codon, while TSS-2 is located −47 bp upstream of the dfrA start codon within the dfrA–thyA intergenic region (71 bp). CDS, coding sequence; DHF, dihydrofolate; DHFR, dihydrofolate reductase; DHFS, dihydrofolate synthase; DHP, dihydropteroate; DHPP, 6-hydroxymethyl-7,8-dihydropterin pyrophosphate; DHPS, dihydropteroate synthase; dTMP, deoxythymidine monophosphate; dUMP, deoxyuridine monophosphate; FDTS, flavin-dependent thymidylate synthase; mTHF, methylenetetrahydrofolate; pABA, para-aminobenzoic acid; SHMT, serine hydroxyl methyltransferase; THF, tetrahydrofolate; TS, thymidylate synthase.

Overexpression of dfrA or ribD has been shown to confer resistance to para-aminosalicylic acid. At high expression levels, RibD can function as an alternative DHFR, thereby compensating for the loss of dfrA. Although a mutation in the upstream region of dfrA has been reported in two clinical isolates exhibiting high-level para-aminosalicylic acid resistance (MIC > 32 mg/L),¹² experimental evidence confirming a causal role for this upstream mutation in para-aminosalicylic acid resistance is lacking. However, the contribution to para-aminosalicylic acid resistance of mutations within the coding region of dfrA is incompletely characterized. The genomic organization of dfrA is illustrated in Figure 1b. The gene is located adjacent to thyA, separated by a 71 bp intergenic spacer, and is predicted to have two transcriptional start sites (TSS-1 and TSS-2) relative to the dfrA start codon (ATG), located at +9 and −47, respectively.¹³ In this study, we identified mutations in the coding and upstream regions of dfrA, and demonstrated that some but not all of those mutations contribute to para-aminosalicylic acid resistance in Mtb clinical isolates.

Materials and methods

Bacterial strains, plasmids and growth conditions

Mtb strains H37Ra (WT) and H37Ra (mutants) were grown in Middlebrook 7H9 broth with 0.05% Tween 80, 0.5% glycerol, and 10% (v/v) oleic acid–albumin–dextrose–catalase (OADC) at 37°C for 2 weeks, or on 7H10 agar supplemented with 0.5% glycerol and 10% OADC at 37°C for 4 weeks. Mycobacterium smegmatis mc²155 (MS) was cultured in Middlebrook 7H9 broth supplemented with 0.05% (v/v) Tween 80% and 0.5% (v/v) glycerol at 37°C for 72 h, or on 7H10 agar supplemented with 0.5% glycerol for 5 days at 37°C. Escherichia coli DH5α was cultured in LB medium at 37°C with shaking at 220 rpm overnight, or on LB agar for 24 h. Plasmids included p519ngfp (for gfp amplification), and pMV261 (shuttle vector for mycobacterial gene expression). Antibiotics were added as kanamycin (100 mg/L for E. coli, 25 mg/L for MS and H37Ra).

Identification and analysis of SNPs in dfrA and its upstream region in clinical isolates

We retrieved a total of 6547 whole-genome sequences of Mtb clinical isolates from the NCBI database and imported them into bioinformatics software Geneious Prime (version 2022.2.1) for analysis. The dfrA gene (480 bp) and its upstream thyA–dfrA intergenic region (71 bp) from the H37Rv reference genome (GenBank accession no. AL123456.3) were used as reference sequences. All genomes were aligned using the default Geneious Prime alignment algorithm to identify shared nucleotide sequences, after which SNPs were systematically detected.

Construction of reporter vectors

We used the ClonExpress Ultra One Step Cloning Kit V2 (Vazyme, China) for constructing reporter vectors as per the manufacturer’s manual. First the plasmid pMV261 was linearized with restriction enzymes such as AatII and HpaI and the hsp60 promoter removed. Primers were designed using SnapGene (Version6.0.2) software using target genes as templates and the terminal sequences (15–20 bp) of the linearized vector at the 5′ end of each forward and reverse primer were introduced. The mutant vectors, WT vector and negative control vector were constructed using different combinations of primers synthesized by TSINGKE. For the negative control reporter vector, we used the primers gfp-FP and gfp-RP, with the gfp sequence from the plasmid p519ngfp serving as template (Figure 2a).

Figure 2.

Vector construction. (a) gfp-negative control reporter vector without promoter. (b) gfp-positive control reporter vector with WT dfrA promoter. (c) gfp reporter vector with mutated dfrA promoter amplified by inverse PCR. (d) dfrA-negative control vector with only dfrA promoter. (e) Vector with WT dfrA promoter and CDS. (f) Vector with dfrA CDS and mutated dfrA promoter amplified by inverse PCR. A pMV261 plasmid digested with AatII and HpaI restriction enzymes was utilized.

To construct a WT reporter vector, the 500 bp upstream region of the dfrA gene was fused to gfp. The upstream region and gfp were amplified with two pairs of primers: wt pro(dfrA)-FP/wt-RP and wt- df.gfp.FP/RP, using Mtb H37Ra genomic DNA and p519ngfp as templates. To facilitate successful homologous recombination, 15–20 bp homologous sequences matching either the linearized vector or the fusion junctions were added to the 5′ ends of the primers (Figure 2b). For generating mutant reporter vectors, we adapted an inverse PC-based strategy and amplified the whole plasmid by using primers with mutations incorporated and wt plasmid as a template (see Supplementary methods; available as Supplementary data at JAC-AMR Online).

We recombined the amplified fragments and pre-digested plasmid in a total of 10 µL reaction volume consisting of 20 ng of amplified product, 80 ng of digested plasmid and 5 μL of Exnase (2× CE mix), with nuclease-free water added to reach the final volume. The mixture was incubated at 50°C for 15 min to allow recombination. Inverse PCR reactions were first treated with DpnI enzyme at 37°C for 1–2 h to eliminate the template plasmids and then amplicons were gel purified.

The resultant products were transformed into E. coli DH5α, plated on LB agar containing kanamycin, and incubated at 37°C for 24 h. Colonies were selected for sequencing and correct vectors were electroporated into M. smegmatis for phenotypic analysis. The electroporated suspension was plated on Middlebrook 7H10 agar with kanamycin and incubated at 37°C for 5 days. After the incubation period, single colonies were selected in 7H9 broth, and subjected to colony PCR (see Table S1 for details of primers, plasmids and strains used in this study).¹¹

GFP fluorescence detection

GFP fluorescence was quantified using a previously reported method with slight modifications, and relative fluorescence units (RFUs) were normalized to OD at 600 nm (OD₆₀₀).¹¹ MS cultures were grown to mid-log phase, and 1% of the culture was inoculated into 5 mL of Middlebrook 7H9 broth supplemented with the appropriate antibiotics. Cultures were incubated at 37°C until an OD₆₀₀ of approximately 0.6 was reached. Cells (2 mL) were then harvested by centrifugation at 2000 × g for 10 min at room temperature, washed twice with PBS, and resuspended in 2 mL of PBS. An aliquot of 200 μL was transferred to a black 96-well plate with a clear bottom. Absorbance at 600 nm was measured, followed by GFP fluorescence detection using an excitation wavelength of 488 nm and an emission wavelength of 525 nm. To quantify GFP expression, raw RFU were first divided by the OD₆₀₀ to normalize for cell density (RFU/ODU). (For Figures 4b and 4c, these values were further normalized to the WT strain to calculate the relative percentage, where the mean RFU/ODU of the WT strain was set to 100%.) To calculate RFU (%), the following formula was used: RFU (%) = [(RFU/ODU)_sample/(RFU/ODU)_wildtype] × 100.

Figure 4.

Growth curves and gfp expression in WT, negative control and mutant strains. (a) Growth curves of MS strains, including WT, negative control and mutants. (b, c) gfp expression in MS WT, negative control and mutant strains. Values shown here are for three biological repeats. **P < 0.01, *P < 0.05. ns, not significant. Error bars represent the standard deviation (SD).

Construction of expression vectors for dfrA expression in H37Ra

We constructed the negative control vector pMV260-P_dfrA and WT vector pMV260-P_dfrA-dfrA by amplifying 500 bp upstream of dfrA and P_dfrA-dfrA (−500 to +480) regions, respectively, using different primers from H37Ra genomic DNA. Each amplified fragment was ligated to pMV261, pre-digested with AatII and HpaI. For the mutant vectors, harbouring different mutations in the upstream region of dfrA, we employed an inverse PCR-based approach and amplified the mutant vectors from WT vector pMV260-P_dfrA-dfrA, using primers that contained the mutated sequences. The amplified plasmids were digested with DpnI as described above and transformed into E. coli. Well-grown colonies were selected and sequence-verified. Following verification, the correct plasmids were electrotransformed into Mtb H37Ra for phenotypic analysis.

Assessment of thermal stability of protein point mutations

The thermal stability of protein structures was evaluated using Gibbs free energy (ΔG), which is naturally negative; lower ΔG values indicate greater thermal stability. To assess the effect of point mutations on protein stability, changes in free energy (ΔΔG) were calculated. In this study, the PremPS web server (https://lilab.jysw.suda.edu.cn/research/PremPS/) was employed to predict the thermal stability alterations caused by single point mutations. PremPS, developed by the Systems Biology Research Center at Soochow University, provides reliable predictions across different structural models of the same protein.¹⁴ The PDB file of the target protein, either downloaded from the Protein Data Bank or generated through structural prediction software, along with the specified point mutation information, was submitted to PremPS. The server outputs ΔΔG values (ΔG_mut – ΔG_wt) for each mutation.

Drug susceptibility testing

The MIC, defined as the lowest concentration of an antibiotic that prevents visible growth in vitro, was determined for para-aminosalicylic acid (Sigma-Aldrich) in Mtb H37Ra. Drug was dissolved in double-distilled water, filter sterilized and tested at final concentrations of 0, 0.005, 0.01, 0.02, 0.04, 0.08, 0.16, 0.32, 0.64, 1.28, 2.56 and 5.12 mg/L. Bacterial cultures were grown to a mid-log phase in Middlebrook 7H9 medium with OADC, adjusted to approximately 10⁷cfu/mL, and further diluted 100-fold in fresh medium. A volume of 10 μL (∼10³cfu) was spotted onto Middlebrook 7H10 agar plates containing the indicated concentrations of para-aminosalicylic acid. Plates were incubated at 37°C for 4 weeks. MIC values were determined based on growth inhibition and confirmed using three independent technical replicates.

Results

Analysis of dfrA mutations in coding and upstream region

A total of 6547 whole genomes of Mtb clinical isolates were downloaded from NCBI and mapped to the reference dfrA and upstream thyA–dfrA sequences of H37Rv reference sequence using Geneious Prime software, and a total of 37 SNPs were found. Among these SNPs, 29 belonged to the dfrA coding region, including 20 missense mutations, 5 synonymous mutations, 3 frameshift mutations and 1 termination mutation. The remaining eight SNPs were observed in the upstream thyA–dfrA region. There were 33 isolates with missense mutations, of which Trp142→Gly (27.27%, 9/33) was the most abundant; the other mutations were not concentrated and had a low proportion, of which 2 strains were found in Gln28→Leu. SNPs in the upstream region of thyA–dfrA were found in 73 isolates, with G70T being the most abundant (87.67%, 64/73). We selected the most frequent mutations in the coding region and all upstream mutations for phenotypic analysis (Table 1, Figure 3). All mutations are being reported for the first time except for G70T, which has been previously reported.

Table 1.

Distribution of SNPs in dfrA coding and upstream thyA–dfrA regions

S. No.	Base mutation	Type of mutation	Amino acid change	No. of isolates
1	83A → T	Missense mutation	Q28L	2
2	95G → C	Missense mutation	R32P	2
3	424T → G	Missense mutation	W142G	9
4	53G → A	Missense mutation	G18D	1
5	64T → C	Missense mutation	W22R	1
6	85G → A	Missense mutation	A29T	1
7	147G → C + 424T → G + 427C → G	Missense mutation	S49S + W142G + R143G	1
8	170T → C	Missense mutation	L57P	1
9	220A → G	Missense mutation	S74G	2
10	231G → T	Missense mutation	E77D	1
11	305T → G	Missense mutation	L102R	1
12	313C → T	Missense mutation	P105S	1
13	330T → G	Missense mutation	C110W	1
14	337A → G + 341A → T	Missense mutation	T113A + Q114V	1
15	345C → T + 424T → G + 427C → G	Missense mutation	V115V + W142G + R143G	1
16	353G → C	Missense mutation	G118A	1
17	370G → C	Missense mutation	G124R	2
18	422A → G	Missense mutation	E141G	1
19	422A → G + 424T → G	Missense mutation	E141G + W142G	2
20	440C → G	Missense mutation	S147C	1
21	129C → T	Synonymous mutation	G43	1
22	156T → G	Synonymous mutation	A52	1
23	279G → C	Synonymous mutation	V93	1
24	411C → G	Synonymous mutation	G137	1
25	96G → A	Synonymous mutation	R32	1
26	242C → G	Termination mutation	S81 Stop	1
27	163C del	Frameshift mutation	Frameshift	1
28	387C del	Frameshift mutation	Frameshift	1
29	411C del	Frameshift mutation	Frameshift	1
30	−11C → G	Regulatory mutation	NA	1
31	−14T → G	Regulatory mutation	NA	2
32	−48C → A	Regulatory mutation	NA	1
33	−50G → C	Regulatory mutation	NA	1
34	−59G → T	Regulatory mutation	NA	1
35	−62G → C	Regulatory mutation	NA	2
36	−70G → A	Regulatory mutation	NA	1
37	−70G → Ta	Regulatory mutation	NA	64

NA, not applicable; S.No., Serial Number.

^aMutation previously reported.

Figure 3.

Distribution of tested mutations in the upstream thyA–dfrA region and the position of most frequent amino acid substitutions within the dfrA coding sequence.

Impact of mutations on the thermal stability of DHFR

The PDB file with ID 6nnd and the corresponding mutation sites and mutation information were entered into the PremPS website to obtain the ΔΔG values for each mutation, as shown in Table 2. ΔΔG is defined as the ΔG value of the mutant minus that of the WT. A negative value indicates that the mutant protein is more stable than the WT protein, while a positive value indicates that the mutant protein is less stable. The PremPS analysis showed that the ΔΔG value of Gln28→Leu is negative, indicating increased stability relative to the WT, whereas Arg32→Pro and Trp142→Gly have positive values, indicating possible destabilizing effects on protein structure, suggesting that these mutations have a significant impact on the thermal stability of the enzyme.

Table 2.

Assessment of thermal stability of protein point mutations

S. No.	Amino acid mutation	PremPS ΔΔGa (kcal/mol)
1	Q28L	−0.57
2	R32P	+0.31
3	W142G	+1.37
4	G18D	+0.92
5	W22R	+2.06
6	A29T	+0.57
7	L57P	+2.70
8	S74G	+0.46
9	E77D	+0.39
10	L102R	+0.70
11	P105S	+0.48
12	C110W	+1.78
13	G118A	−0.23
14	G124R	+0.85
15	E141G	+0.17
16	S147C	+0.33

^aΔΔG=ΔG_mut−ΔG_wt.

Mutations in the thyA–dfrA region affected promoter activity of dfrA

We generated recombinant MS reporter strains (MS mc²155) carrying the plasmid pMV260, in which the upstream regulatory region of dfrA was fused to the GFP reporter gene. Briefly, a 500 bp fragment upstream of dfrA, encompassing the transcription start sites and regulatory elements, was amplified and cloned upstream of gfp in pMV260 after removal of the native hsp60 promoter, in order to assess dfrA promoter activity in M. smegmatis. Growth curve analysis was performed for all reporter strains, including the negative control (pMV260-gfp), the WT promoter construct (pMV260-PdfrA-gfp), and strains harbouring mutant dfrA promoter variants. All strains exhibited comparable growth kinetics, with no statistically significant differences observed. These results indicate that neither the promoter mutations nor the shuttle vector imposed detectable growth defects, thereby excluding growth-related effects on reporter gene expression (Figure 4a). Comparison of GFP expression levels revealed that three strains with mutations in the dfrA promoter significantly enhanced promoter activity relative to the WT, while the other mutations had either reduced or no significant effect. Among the increased-activity group, G70T, G59T and C48A increased promoter activity by approximately 13-fold, 4-fold and 3-fold, respectively (Figure 4b and 4c).

Mutations in dfrA coding and thyA–dfrA upstream regions affect para-aminosalicylic acid resistance

To further investigate the effect of dfrA mutations on para-aminosalicylic acid susceptibility, we constructed dfrA expression vectors carrying mutations either in the coding region or in the thyA–dfrA upstream regulatory region and transformed them into Mtb H37Ra. MICs were determined to evaluate the impact of coding-region mutations on drug susceptibility and to assess the association between dfrA promoter activity and para-aminosalicylic acid resistance. Among the coding-region mutants, the Gln28→Leu substitution resulted in an 8-fold increase in MIC compared with the WT, whereas the Trp142→Gly and Arg32→Pro mutations did not cause any significant change in para-aminosalicylic acid susceptibility. In contrast, mutations within the dfrA promoter region that enhanced promoter activity were consistently associated with increased para-aminosalicylic acid resistance (Table 3). Specifically, expression constructs harbouring the thyA–dfrA upstream mutations G70T and G59T conferred the highest level of resistance (≥64-fold increase in MIC) relative to the WT strain. The C48A mutation resulted in a 32-fold increase, while the T14G mutation caused a 2-fold increase in para-aminosalicylic acid resistance. Other promoter mutations did not produce a significant change in MIC; notably, the G50C mutation was associated with a lower MIC than that of the WT strain. We observed a very strong positive correlation between GFP expression and para-aminosalicylic acid MIC (r_s = 0.88, n = 9, P = 0.0036), indicating that higher levels of GFP expression/promoter activity are significantly associated with increased antibiotic resistance in these strains, except for T14G, which had no significant promoter activity yet showed slightly raised MIC compared with the WT.

Table 3.

MICs of PAS against dfrA expression strains

S#	Strain	PAS MIC (mg/L)	Fold change
1	H37Ra pMV260-PdfrA-dfrA	0.08	1
2	H37Ra pMV260-PdfrA-dfrA Q28L	0.64	8
3	H37Ra pMV260-PdfrA-dfrA R32P	0.04	0.5
4	H37RapMV260-PdfrA-dfrA W142G	0.04	0.5
5	H37Ra pMV260-(C-11G)-dfrA	0.04	0.5
6	H37Ra pMV260-(T-14G)-dfrA	0.16	2
7	H37Ra pMV260-(C-48A)-dfrA	2.56	32
8	H37Ra pMV260-(G-50C)-dfrA	0.02	0.25
9	H37Ra pMV260-(G-59T)-dfrA	5.12	64
10	H37Ra pMV260-(G-62C)-dfrA	0.04	0.5
11	H37Ra pMV260-(G-70A)-dfrA	0.04	0.5
12	H37Ra pMV260-(G-70T)-dfrA	>5.12	>64

PAS, para-aminosalicylic acid.

Discussion

The emergence of resistance to first-line anti-TB drugs has increased reliance on second-line agents such as para-aminosalicylic acid, particularly for the treatment of MDR- and XDR-TB. para-Aminosalicylic acid may be incorporated into individualized treatment regimens for patients with highly drug-resistant TB, including those with XDR-TB or treatment failure when resistance or intolerance limits the use of preferred agents.^15,16 With the growing adoption of WGS for predicting drug resistance, identifying reliable genetic markers for para-aminosalicylic acid resistance has become increasingly important.^17–19 However, resistance prediction based solely on WGS remains challenging because para-aminosalicylic acid resistance mechanisms are incompletely characterized. Previous studies have identified mutations in thyA, folC and ribD, and also indicated that mutations in the upstream region of dfrA may contribute to para-aminosalicylic acid resistance.¹² These observations motivated us to systematically investigate mutations in dfrA and its upstream regulatory region in clinical Mtb isolates.

In this study, analysis of 6547 clinical genomes revealed 29 SNPs in the dfrA coding region, including 20 missense, 5 synonymous, 3 frameshift and 1 termination mutation. Among the missense variants, Trp142→Gly was the most frequent (27.27%, 9/33), whereas all other mutations were observed in no more than two strains. In parallel, we identified eight SNPs in the thyA–dfrA upstream region, with G70T being highly prevalent (87.67%, 64/73), followed by T14G and G62C, each detected in two isolates.

Bioinformatic analysis predicted that most dfrA coding mutations reduced protein thermal stability, whereas Gln28→Leu was predicted to increase stability relative to the WT enzyme. Structural modelling suggested that Trp142→Gly and Arg32→Pro induced varying degrees of conformational alteration. Reduced thermal stability or structural disruption is often associated with diminished enzyme activity, whereas increased stability may enhance its function. We selected Gln28→Leu, Trp142→Gly and Arg32→Pro for experimental validation.

Using expression constructs driven by the native dfrA promoter, we demonstrated that Gln28→Leu conferred an 8-fold increase in drug MIC, while Trp142→Gly and Arg32→Pro retained a phenotype susceptible to para-aminosalicylic acid, although Trp142→Gly seems to be more prevalent in Mtb clinical isolates. Ribeiro et al.²⁰ reported that dihydrofolate interacts with residues including Trp22, Asp27, Leu24 and Gln28 in DHFR, highlighting Gln28 as a key binding site. We therefore speculate that the Gln28→Leu substitution may enhance DHFR–substrate affinity or enzymatic efficiency, thereby contributing to drug resistance. Previous studies have shown that dfrA mutations in resistant strains are often accompanied by thyA mutations;^21,22 here we demonstrate that Gln28→Leu alone causes para-aminosalicylic acid resistance.

To our knowledge, this is the first study to systematically characterize mutations in the dfrA upstream region and experimentally validate their role in drug resistance. GFP reporter assays revealed that G70T, G59T and C48A significantly increased promoter activity, with G70T showing the strongest effect, whereas the remaining mutations had minimal impact. These findings were verified by para-aminosalicylic acid susceptibility testing, which exhibited >64-fold, 64-fold and 32-fold increases in para-aminosalicylic acid MIC for G70T, G59T and C48A, respectively. A strong concordance between GFP expression and MIC values was observed, indicating that enhanced dfrA expression directly contributes to para-aminosalicylic acid resistance. Similar regulatory mechanisms have been reported for mutations in the rv2172c coding region and the thyX upstream region, both implicated in para-aminosalicylic acid resistance in clinical Mtb isolates.^23,24 Together, these data highlight the importance of promoter mutations as an underappreciated mechanism of antifolate resistance.

It is important to note that among the SNPs identified in dfrA and its upstream region, only a subset actually confer drug resistance. This necessitates the combined application of genomic analysis and functional characterization to validate reliable molecular markers for para-aminosalicylic acid resistance. The newly verified SNPs in dfrA and its upstream region could be incorporated into the WHO catalogue of mutations in Mtb complex.²⁵

This study has some limitations. para-Aminosalicylic acid MIC could not be determined for all dfrA coding mutations due to workload constraints. Functional effects of promoter mutations on gene expression were evaluated using a GFP reporter system in M. smegmatis rather than directly in Mtb. Although M. smegmatis is a well-established surrogate host for mycobacterial promoter studies, differences in transcriptional regulation between the two species may influence promoter activity. Resistance to para-aminosalicylic acid was assessed using in vitro MIC measurements, and the observed effects may not fully reflect the clinical impact of these mutations in vivo. Further research is needed to correlate aforementioned mutations and para-aminosalicylic acid resistance.

Funding

This study was supported by Shenzhen Medical Research Fund, Grant No. C2401026; Prevention and Control of Emerging and Major Infectious Diseases-National Science and Technology Major Project, Grant No.2025ZD01999804; Major Project of Guangzhou National Laboratory, Grant No. GZNL2024A01009.

Transparency declarations

The authors have no conflicts of interest related to this work to disclose.

Author contributions

Rong Zeng (Conceptualization, Methodology, Investigation, Formal analysis, Resources, Writing—review & editing), Najeeb Ullah (Investigation, Formal analysis, Data curation, Visualization, Writing—original draft), Jing Gu (Lab management) and Jiao-Yu Deng (Conceptualization, Formal analysis, Supervision, Funding acquisition, Writing—review & editing). All authors have read and agree to the final version of this manuscript.

Disclaimer

AI assistance was employed only for grammar check and language refinement; no AI tools were used in data analysis, interpretation or scientific content generation.

Supplementary data

Supplementary methods and Table S1 are available as Supplementary data at JAC-AMR Online.