Structural stability is a major constraint on the evolution of protein sequences. However, under strong directional selection, mutations that confer novel phenotypes but compromise structural stability of proteins may be permissible. During the evolution of antibiotic resistance, mutations that confer drug resistance often have pleiotropic effects on the structure and function of antibiotic-target proteins, usually essential metabolic enzymes. In the present study, we show that trimethoprim (TMP)-resistant alleles of dihydrofolate reductase from Escherichia coli (EcDHFR) harboring the Trp30Gly, Trp30Arg or Trp30Cys mutations are significantly less stable than the wild-type, making them prone to aggregation and proteolysis. This destabilization is associated with a lower expression level, resulting in a fitness cost and negative epistasis with other TMP-resistant mutations in EcDHFR. Using structure-based mutational analysis, we show that perturbation of critical stabilizing hydrophobic interactions in wild-type EcDHFR enzyme explains the phenotypes of Trp30 mutants. Surprisingly, though crucial for the stability of EcDHFR, significant sequence variation is found at this site among bacterial dihydrofolate reductases (DHFRs). Mutational and computational analyses in EcDHFR and in DHFR enzymes from Staphylococcus aureus and Mycobacterium tuberculosis demonstrate that natural variation at this site and its interacting hydrophobic residues modulates TMP resistance in other bacterial DHFRs as well, and may explain the different susceptibilities of bacterial pathogens to TMP. Our study demonstrates that trade-offs between structural stability and function can influence innate drug resistance as well as the potential for mutationally acquired drug resistance of an enzyme.

Introduction

During the evolution of proteins, structural stability constrains the sequence space that a protein can access and mutations that have large destabilizing effects are usually eliminated by purifying selection. At the same time, there exists a trade-off between the evolution of novel functionalities and structural stability, i.e. novel functions are often accompanied by a loss in protein stability [13]. This trade-off is almost universally operative in proteins, having been demonstrated in diverse systems ranging from metabolic enzymes [4,5] to signaling proteins [6] and even antibodies [7]. It has been invoked to mechanistically explain the evolution of thermostable or cold-adapted enzymes [8], intramolecular epistasis [911] and promiscuous enzyme activity [12]. Stability–function trade-off hence provides an important conceptual framework within which one can understand protein evolution as well as develop tools for protein engineering [13].

The evolution of antimicrobial resistance in the face of antibiotic exposure is a selection process in which an existing protein evolves a novel function, i.e. the ability to resist inhibition by the drug through mutation. Drug resistance is mediated by mutations in antibiotic-metabolizing enzymes or antibiotic-target proteins themselves. In the former, resistance-conferring mutations modify the active site of existing drug-metabolizing enzymes to accommodate novel drugs. The TEM-1 β-lactamase is one of the best-studied drug-metabolizing enzymes in which a mutationally generated change in substrate specificity mediates resistance to second- and third-generation β-lactam antibiotics such as cephalosporins [14,15]. Mutations in antibiotic targets, on the other hand, typically involve loss of binding to the drug by altered active-site architecture. The HIV-I protease remains one of the best-studied systems for these kinds of mutations [16]. In both these cases, it is evident that mutations alter the enzyme active site to generate novel functions, i.e. altered substrate repertoire in one case and altered binding specificity in the other. As predicted by the stability–function trade-off, this novel function is associated with a cost to stability of both these enzymes. Stability–function trade-off in TEM-1 and HIV-1 protease limits the number of evolutionary trajectories that these enzymes can make from drug-sensitive to drug-resistant [911,17]. Furthermore, for the HIV-1 protease, loss of thermodynamic stability has also been causally linked with the fitness costs of drug resistance [17]. Thus, both these systems have provided insights into how stability and functionality trade-off in antibiotic resistance. However, data from proteins other than these ‘model enzymes’ exploring the relationship between function and stability remain scarce. Indeed, computational analyses predict that similar proportions of mutations that confer rifampicin resistance in Escherichia coli [18], metronidazole resistance in Helicobacter pylori [19,20] or trimethoprim (TMP) resistance in E. coli [2123] are predicted to destabilize target proteins as in the case of the HIV-I protease [24] (Figure 1A and Supplementary Figure S1). Thus, stability–function trade-offs may be a common feature during the evolution of drug resistance.

Antibiotic-resistant mutations are associated with reduced protein stability.

Figure 1.
Antibiotic-resistant mutations are associated with reduced protein stability.

(A) Pie charts showing the relative proportions of resistance-conferring mutations that are predicted to be less stable (red) or more stable (green) than wild-type for four different antibiotic-target proteins. Stability prediction was made using reported mutations in RpoB (rifampicin) and DHFR (TMP) from E. coli, and HIV-1 protease (saquinavir) and RdxA (metronidazole) from H. pylori using the iMutant2.0 software. PDB codes of structures used for the analyses are provided. The number in each slice is the number of mutations that fell into that category. (B) Structure of E. coli DHFR (PDB: 7DFR) in cartoon representation with residues that are associated with TMP resistance shown in stick representations (C: magenta; N: blue; S: yellow). Active site-bound dihydrofolate and NADP are shown as lines (C: green; N: blue; O: red; S: yellow). (C) Predicted stabilities of individual TMP resistance-associated mutations. Positive values of ΔΔG indicate stabilization (green), while negative values indicate destabilization (red).

Figure 1.
Antibiotic-resistant mutations are associated with reduced protein stability.

(A) Pie charts showing the relative proportions of resistance-conferring mutations that are predicted to be less stable (red) or more stable (green) than wild-type for four different antibiotic-target proteins. Stability prediction was made using reported mutations in RpoB (rifampicin) and DHFR (TMP) from E. coli, and HIV-1 protease (saquinavir) and RdxA (metronidazole) from H. pylori using the iMutant2.0 software. PDB codes of structures used for the analyses are provided. The number in each slice is the number of mutations that fell into that category. (B) Structure of E. coli DHFR (PDB: 7DFR) in cartoon representation with residues that are associated with TMP resistance shown in stick representations (C: magenta; N: blue; S: yellow). Active site-bound dihydrofolate and NADP are shown as lines (C: green; N: blue; O: red; S: yellow). (C) Predicted stabilities of individual TMP resistance-associated mutations. Positive values of ΔΔG indicate stabilization (green), while negative values indicate destabilization (red).

TMP, an anti-folate drug, is used commonly in the treatment of urinary tract and bladder infections against Gram-negatives such as E. coli. It has also been proposed as a viable alternative treatment against methicillin-resistant Staphylococcus aureus [25] and multidrug-resistant Mycobacterium tuberculosis [26]. Plasmid-borne resistance to TMP is commonly encountered in Gram-negatives like E. coli and Klebsiella pneumoniae. Intrinsic (plasmid-independent) TMP resistance, more common in Gram-positive bacterial species with a few instances reported for Gram-negatives [21,2730], is mediated primarily through mutations in the folA gene that codes for the dihydrofolate reductase (DHFR) protein [21,30,31], an essential enzyme that reduces dihydrofolate to tetrahydrofolate. Inhibition of DHFR by TMP is competitive and leads to bacteriostasis due to perturbation of thymidine and amino acid biosynthesis [32]. TMP resistance has been recapitulated in the laboratory through experimental evolution [22,33] and has proved to be a tractable model to understand the impact of drug-target overexpression [34] and intramolecular epistasis among resistance-conferring mutations [35,36]. Being a relative small globular protein, DHFRs from various different bacteria have been crystallized and biochemically characterized. Additionally, thermodynamic stability of DHFR mutants has been demonstrated to have a direct impact on bacterial fitness [37,38], making this system an attractive tool for structure–function studies. Recent work from the Rodrigues et al. [36] has established that TMP resistance in E. coli is limited by trade-offs with catalytic efficiency. By analysis of three mutations, Pro21Leu, Ala26Thr and Leu28Arg, located in the substrate-binding pocket of E. coli DHFR, it was demonstrated that loss of drug binding was accompanied by reduced catalysis. Like in the case of TEM-1 and HIV-I protease, these mutations also perturbed the structural stability of DHFR and trapped the protein in a molten globule-like state, indicating that stability–function trade-offs are operative during the evolution of TMP resistance in DHFR [36].

Unlike TEM-1, in which mutations are more commonly located in loops [39] or HIV-I protease in which mutations are most common near the active site [24], TMP resistance-conferring mutations are found all over the DHFR enzyme including in core secondary structural elements (Figure 1B). Computational prediction of the stabilities of 12 reported TMP resistance-conferring mutations in EcDHFR showed that some of these mutations can significantly reduce global stability of the protein. Mutations at one such site, Trp30, were predicted to be particularly detrimental to the stability of EcDHFR (Figure 1C). In the present study, we show that TMP resistance-conferring mutations at the Trp30 residue do indeed drastically compromise structural stability of EcDHFR, rendering it susceptible to proteolysis and aggregation. Using structure-guided mutational analyses, we delineate the mechanism of the destabilization of DHFR due to Trp30 mutations and also demonstrate the functional relevance of this destabilization. Finally, using DHFR enzymes from S. aureus and M. tuberculosis, we test whether mutations at Trp30-equivalent positions have similar effects also in other bacterial DHFR enzymes.

Experimental methods

Materials, strains and culture conditions

E. coli K-12 MG155 was used for all protein expression and phenotypic characterization studies. E. coli DH5α was used for DNA manipulation. E. coli ΔfolA strain [40] was a kind gift from Prof. Peter Wright (Scripps Research Institute, U.S.A.). All strains were cultured in Luria Bertani (LB) broth or on LB agar plates (LA). pBKS(+) and pPRO-Ex-HtB plasmids were kind gifts from Prof. Sandhya S. Visweswariah (Indian Institute of Science, India). Proteinase K, IPTG, X-Gal, thymidine, ampicillin and TMP were purchased from Sigma–Aldrich (U.S.A.).

Cloning of wild-type and mutant DHFR enzymes

Wild-type E. coli DHFR (folA gene) was amplified using folA_SfiI_fwd and folA_NotI_rev primers (Table 1) by PCR and pCA24N-folA (Keio Collection) [41] as the template. The PCR amplicon was digested with NotI and cloned into pBKS(+) plasmid digested with SmaI and NotI to yield pBKS-folA plasmid. For expression studies, the folA gene from pBKS-folA was excised by digestion with EcoRV-NotI and was sub-cloned into pPRO-Ex-Htb plasmid digested with StuI-NotI to yield pPRO-folA. Mutations in DHFR were introduced using PCR-based site-directed mutagenesis as described in Shenoy and Visweswariah [42] using the appropriate primer (Table 1) on pBKS-folA template and then sub-cloned into pPRO-Ex-Htb or pPRO-folA directly. S. aureus DHFR was amplified using SafolA_fwd and SafolA_rev primers (Table 1) by PCR using genomic DNA from S. aureus MSSA 29213 strain (kind gift from Dr Harinath Chakrapani, Indian Institute of Science Education and Research, Pune, India). M. tuberculosis DHFR was amplified using MtbfolA_fwd and MtbfolA_rev primers (Table 1) by PCR using genomic DNA from M. bovis BCG strain (kind gift from Prof. Sandhya S. Visweswariah, Indian Institute of Science, Bangalore). Both amplicons were digested with Bsu36I and NotI and used to replace E. coli folA from similarly digested pPRO-folA to yield pPRO-SafolA and pPRO-MtbfolA plasmids. His30Trp mutations were introduced in them by site-directed mutagenesis as described above using the appropriate oligonucleotide primer (Table 1). Inserts in all plasmids were Sanger-sequenced (1st Base, Malaysia) for verification.

Table 1
List of oligonucleotide primers used in the present study
Name Sequence (5′ → 3′) 
folA_SfiI_fwd ATCCGGCCCTGAGGGCCATGA 
folA_NotI_rev CCCTTAGCGGCCGCATAGGCC 
folAprom_XbaI_fwd CGGATTCTAGAGAAACGAAACCCTCATCC 
folA_HindIII_rev GGCGAAGCTTCGGCGTCTTAAACACAGCC 
folA p21L NaeI fwd CGCCATGCTGTGGAACCTGCCGGCCGATCTCGCC 
folA W30G NaeI fwd GCCGATCTCGCCGGCTTTAAACGCAACACC 
folA W30R NaeI fwd GGAACCTGCCGGCCGATCTCGCCCGGTTTAAACG 
folA I94L AvaI fwd GAAATCATGGTGCTCGGGGGCGGTCGCG 
folA_K109R_fwd CAAAAGCGCAAAGACTGTATCTGACGC 
folA_A26T_BglII_fwd GAACCTGCCTACAGATCTCGCCTGGTTTAAAC 
folA_H45R_NaeI_fwd CGTGATTATGGGCCGGCGTACCTGGGAATCAATCG 
folA_M20V_NaeI_fwd GGAAAACGCCGTGCCGTGGAACCTGCCGGCCGATCTCG 
folA_V10A_NheI_fwd GATTGCGGCGCTAGCGGCAGATCGCGTTATCG 
folA W30L BglII fwd CCTGCCTGCAGATCTCGCCTTGTTTAAACGCAACACC 
folA_L28R_NaeI_fwd GTGGAACCTGCCGGCCGATCGCGCCTGGTTTAAAC 
folA_P21Q_NaeI_fwd GAAAACGCCATGCAGTGGAACCTGCCGGCCGATCTCG 
folA_W30C_BglII_fwd GAACCTGCCTGCAGATCTCGCCTGTTTTAAACGCAAC 
folA W30Y BglII fwd TGGAACCTGCCTGCAGATCTCGCCTATTTTAAACGCAACACC 
folA W30F PvuII F TGGAACCTGCCAGCTGATCTCGCCTTTTTTAAACGCA 
folA F153A fwd CTCTCACAGCTATTGCGCTGAGATTCTGGAGCGG 
folA F137A fwd ATGACTGGGAATCGGTAGCCAGCGAATTCCACGAT 
folA I155A fwd CAGCTATTGCTTTGAGGCTCTGGAGCGGCGGGGCC 
folA_Saureus_fwd GAATTCCTGCAGCCCCCGGCCCTGAGGGCCATGACTTTATCCATTCTAGTTGCACATGAC 
folA_Saureus_rev GCTCTAGATTCGAAAGCGGCCGCATAGGCCTTTTTTACGAATTAAATGTAGAAAGGTATG 
folA_Mtb_fwd GAATTCCTGCAGCCCCCGGCCCTGAGGGCCATGGTGGGGCTGATCTGGGCTCAAGCGACA 
folA_Mtb_rev GGCTCTAGATTCGAAAGCGGCCGCATAGGCCTGAGCGGTGGTAGCTGTACAACCGGTACCG 
folA_Mtb_H30W_HindIII_F GCCCGAGGACCAAGCTTGGTTCCGGGAGATCACCATGGGGCACAC 
folA_Saureus_H30W_fwd CCAAATGATTTGAAGTGGGTTAAAAAGCTTTCAACAGGTCATACTTTAGTAATGG 
Name Sequence (5′ → 3′) 
folA_SfiI_fwd ATCCGGCCCTGAGGGCCATGA 
folA_NotI_rev CCCTTAGCGGCCGCATAGGCC 
folAprom_XbaI_fwd CGGATTCTAGAGAAACGAAACCCTCATCC 
folA_HindIII_rev GGCGAAGCTTCGGCGTCTTAAACACAGCC 
folA p21L NaeI fwd CGCCATGCTGTGGAACCTGCCGGCCGATCTCGCC 
folA W30G NaeI fwd GCCGATCTCGCCGGCTTTAAACGCAACACC 
folA W30R NaeI fwd GGAACCTGCCGGCCGATCTCGCCCGGTTTAAACG 
folA I94L AvaI fwd GAAATCATGGTGCTCGGGGGCGGTCGCG 
folA_K109R_fwd CAAAAGCGCAAAGACTGTATCTGACGC 
folA_A26T_BglII_fwd GAACCTGCCTACAGATCTCGCCTGGTTTAAAC 
folA_H45R_NaeI_fwd CGTGATTATGGGCCGGCGTACCTGGGAATCAATCG 
folA_M20V_NaeI_fwd GGAAAACGCCGTGCCGTGGAACCTGCCGGCCGATCTCG 
folA_V10A_NheI_fwd GATTGCGGCGCTAGCGGCAGATCGCGTTATCG 
folA W30L BglII fwd CCTGCCTGCAGATCTCGCCTTGTTTAAACGCAACACC 
folA_L28R_NaeI_fwd GTGGAACCTGCCGGCCGATCGCGCCTGGTTTAAAC 
folA_P21Q_NaeI_fwd GAAAACGCCATGCAGTGGAACCTGCCGGCCGATCTCG 
folA_W30C_BglII_fwd GAACCTGCCTGCAGATCTCGCCTGTTTTAAACGCAAC 
folA W30Y BglII fwd TGGAACCTGCCTGCAGATCTCGCCTATTTTAAACGCAACACC 
folA W30F PvuII F TGGAACCTGCCAGCTGATCTCGCCTTTTTTAAACGCA 
folA F153A fwd CTCTCACAGCTATTGCGCTGAGATTCTGGAGCGG 
folA F137A fwd ATGACTGGGAATCGGTAGCCAGCGAATTCCACGAT 
folA I155A fwd CAGCTATTGCTTTGAGGCTCTGGAGCGGCGGGGCC 
folA_Saureus_fwd GAATTCCTGCAGCCCCCGGCCCTGAGGGCCATGACTTTATCCATTCTAGTTGCACATGAC 
folA_Saureus_rev GCTCTAGATTCGAAAGCGGCCGCATAGGCCTTTTTTACGAATTAAATGTAGAAAGGTATG 
folA_Mtb_fwd GAATTCCTGCAGCCCCCGGCCCTGAGGGCCATGGTGGGGCTGATCTGGGCTCAAGCGACA 
folA_Mtb_rev GGCTCTAGATTCGAAAGCGGCCGCATAGGCCTGAGCGGTGGTAGCTGTACAACCGGTACCG 
folA_Mtb_H30W_HindIII_F GCCCGAGGACCAAGCTTGGTTCCGGGAGATCACCATGGGGCACAC 
folA_Saureus_H30W_fwd CCAAATGATTTGAAGTGGGTTAAAAAGCTTTCAACAGGTCATACTTTAGTAATGG 

Solubility characterization

Mid-log phase cultures of E. coli harboring plasmids encoding the appropriate DHFR enzymes were shifted from 37°C to 18°C. After 30 min of acclimatization, protein production was induced by the addition of IPTG (500 µM) for 12 h at 18°C with shaking at 180–200 rpm. For DHFR from S. aureus and M. tuberculosis, induction was performed for 3 h at 37°C. Soluble and insoluble fractions were separated by the following protocol. Cultures were centrifuged and bacterial pellets were lysed in lysis buffer [50 mM Tris–HCl (pH 7.5), 100 mM NaCl, 5 mM β-mercaptoethanol and 10% glycerol] supplemented with PMSF (1 mM) by sonication on ice. Lysates were centrifuged in a refrigerated centrifuge at 3000×g for 10 min to remove unlysed cells followed by 22 000×g for 15 min. Equivalent volumes of total lysate and pellet and supernatant of the high-speed spin were subjected to SDS–PAGE, and protein bands were visualized with Coomassie Brilliant Blue staining.

For solubilization of aggregated DHFR mutants, pellet fractions from the high-speed spin were resuspended in lysis buffer alone (control) or lysis buffer supplemented with NaCl (1 M), SDS (0.2%), NP-40 (0.1%) or Triton X-100 (0.1%) and incubated for 1 h at room temperature on an end-over rocker. Samples were then centrifuged at 22 000 g for 15 min to separate solublized and unsolublized fractions. Equivalent amounts of unsolubilized pellet fraction (‘input’) and solublized fractions were subjected to SDS–PAGE and visualized with Coomassie Brilliant Blue staining.

Limited proteolysis using proteinase K

E. coli harboring DHFR-encoding plasmids were grown to mid-log phase at 37°C without the addition of IPTG. Lysates were prepared from these bacterial cultures as above without the addition of PMSF. Proteinase K (Sigma–Aldrich, U.S.A.) was added to lysates at a concentration of 2 µg/ml, and the reaction was incubated at 37°C for 5 min, unless otherwise mentioned. The reaction was stopped by shifting the tubes to 4°C and the addition of PMSF (4 mM). Control reactions were set up for each DHFR protein that were treated similarly as experimental samples without the addition of proteinase K. An equivalent amount of control and test samples were subjected to SDS–PAGE and electroblotted onto PVDF membranes. For detection of DHFR, the membranes were blocked with 5% skimmed milk for 1 h at room temperature, followed by incubation with anti-DHFR polyclonal IgG raised against purified His-tagged DHFR (BioKlone, India) at a concentration of 100 ng/ml overnight at 4°C in the presence of 0.2% bovine serum albumin. The membrane was washed thrice with TBS-T, incubated with HRP-linked goat anti-rabbit antibody (Thermo Scientific, U.S.A.) at a dilution of 1 : 20 000 for 1 h at room temperature, washed thrice with TBS-T and then visualized using chemiluminescence. Band intensities were analyzed using ImageJ.

Selection of spontaneous TMP-resistant E. coli

E. coli K-12 MG1655 was streaked out onto LA and six single colonies were used to establish replicate evolving populations. Each population was cultured over a period of 8 days in LB supplemented with TMP (250 ng/ml) in wells of a sterile 96-well plate. Each day 10% of the bacteria were sub-cultured into fresh media. At the end of the selection, evolved populations were streaked out onto LA + TMP (1 µg/ml), and TMP-resistant colonies were isolated, regrown and frozen. To map TMP resistance-conferring mutations, the folA gene along with its promoter were amplified by PCR using genomic DNA isolated from the resistant isolates and folAprom_XbaI_fwd and folA_HindIII_rev primers (Table 1). Resistant mutants harboring Trp30Gly and Trp30Cys mutations were used for further characterization.

Endogenous expression levels of DHFR and its mutants

The endogenous expression levels of DHFR and its mutants were analyzed by western blot analysis. Briefly, 5 µg of total lysate protein or the soluble protein fraction from wild-type or TMP-resistant E. coli strains were subjected to SDS–PAGE and immunoblotted as described above. To differentiate between nonspecific bands and DHFR, lysate of an E. coli ΔfolA strain was used as a negative control.

Fitness of TMP-resistant E. coli

Relative fitness of TMP-resistant E. coli was calculated as described in Bershtein et al. [38]. Briefly, equal volumes of an overnight-grown culture of the appropriate TMP-resistant strain and the reference strain (E. coli MG1655 ΔlacZ) were mixed and inoculated into 5 ml of fresh LB with or without thymidine (100 µg/ml) or M9 medium supplemented with glucose (0.2%), casein hydrolysate (0.12%) and thiamin (0.5 µg/ml) at a starting density of ∼106 CFU/ml. The strains were allowed to compete for 12 h at 37°C with shaking at 180–200 rpm. The relative proportions of test and reference strains were estimated at 0 and 12 h by plating serial dilutions of the cultures onto LA supplemented with IPTG (1 mM) and X-Gal (40 µg/ml). Since the reference strain had a deletion in the lacZ gene, it formed white colonies, while all other strains formed blue colonies on this medium. Control competition experiments were set up between wild-type E. coli K-12 MG1655 and its ΔlacZ derivative to ensure that deletion of lacZ itself was not costly under these conditions. Fitness was calculated using the following formula [43]:

 
formula

where ‘ft’ and ‘fr’ are the cell densities of the test and reference strains after 12 h of growth, and ‘it’ and ‘ir’ are the initial cell densities of the test and reference strains, respectively.

Characterization of TMP-resistance level

To characterize the level of TMP resistance conferred by various DHFR mutants, saturated cultures of E. coli strains expressing the appropriate DHFR protein were serially diluted and spotted onto LA plates containing TMP (1, 2, 4, 8, 16, 32, 64 and 128 µg/ml). The plates were incubated for 20–24 h at 37°C and resulting growth was photographically recorded. To calculate the IC50, maximum dilution of the cultures allowing growth at each concentration of TMP was plotted against TMP concentration and fitted to a variable slope, dose–response inhibition curve using GraphPad Prism 6 (GraphPad Prism Software, U.S.A.).

Sequence comparison and phylogenetic analysis

To compare the sequences of various bacterial DHFR enzymes, 61 DHFR sequences were obtained from Pfam (pf00186; seed sequence). The sequences of human DHFR and M. tuberculosis DHFR were added to this data set manually. These protein sequences were aligned using ClustalW and a cladogram was constructed using the Neighbor-joining algorithm in MEGA7 [44]. Human DHFR was used as the out-group to root the tree.

Analyses of DHFR structures and mutant stability predictions

Structures of DHFR enzymes from E. coli (EcDHFR, 7DFR) [45], Bacillus anthracis (BaDHFR, 3FL9) [46], M. tuberculosis (MtDHFR, 1DG8) [47], S. aureus (SaDHFR, 2W9T) [28] and Mortiella profunda (MpDHFR, 3IA5) [48] were obtained from the PDB and used for analysis. Structural alignments were done using SuperPose [49] and visualization was done using the PyMOL software (Shrödinger, U.S.A.). Stability of mutant DHFR enzymes was predicted using iMutant2.0 [50] and compared with the hydrophobicity scale calculated for amino acids by Wimley and White [51].

Results

Mutations at Trp30 lead to mis-folding of E. coli DHFR and greater sensitivity to proteolysis

Among the 12 mutations in E. coli DHFR (EcDHFR) reported to confer TMP resistance, mutation of Trp30 to Gly, Arg or Cys were predicted to be the most destabilizing (Figure 1C). To validate these predictions, we cloned wild-type and all 12 TMP-resistant EcDHFR mutants, expressed them as His-tagged fusion proteins in E. coli K-12 MG1655 and analyzed their distribution between soluble and insoluble fractions when overproduced. Large structural perturbation often leads to protein aggregation and, hence, the relative distributions of mutant DHFR enzymes between soluble and insoluble fractions would provide an indication of their stabilities. While wild-type EcDHFR and 9 out of 12 TMP-resistant mutants were primarily localized to the soluble fraction, Trp30Gly, Trp30Arg and Trp30Cys mutant DHFR enzymes were predominantly localized to the insoluble fraction (Figure 2A). These mutant proteins could not be solubilized from the insoluble fraction by treatment with nonionic detergents or high salt buffer, but were readily solubilized by SDS treatment, indicating that the Trp30Gly, Trp30Arg and Trp30Cys mutations lead to mis-folding and aggregation of EcDHFR (Figure 2B). To rule out artifacts of overproduction, we exploited the fact that plasmid-encoded EcDHFR was expressed at a low level (detectable by immunoblotting using DHFR-specific antibodies) even without IPTG induction due to leaky expression. Indeed, at lower expression levels, a significant amount of the Trp30 mutants were found to be soluble (not shown). To check whether the soluble forms of the Trp30 mutants folded as efficiently as the wild-type, we subjected lysates of E. coli-expressing wild-type EcDHFR or its mutants to limited proteolysis by proteinase K. Limited proteolysis is a well-established assay for assessing protein conformation and folding, and proteolytic sensitivity of DHFR has been demonstrated to be dependent on its ability to fold [38,52]. Wild-type EcDHFR clipped naturally to a smaller fragment in lysates of E. coli, and this clipping was exacerbated in the presence of proteinase K. Importantly, this smaller fragment derived from wild-type EcDHFR by the action of proteinase K was significantly more resistant than the full-length protein to further proteolysis and hence most probably represented the compact globular core of the protein (Figure 2C,D). In stark contrast, Trp30Gly DHFR showed a markedly different pattern of clipping and did not accumulate a proteinase K-resistant fragment (Figure 2C,D). Similar results were also observed for Trp30Arg and Trp30Cys mutants, indicating that these mutants were unable to fold as well as the wild-type.

Destabilizing effect of Trp30 mutations on EcDHFR.

Figure 2.
Destabilizing effect of Trp30 mutations on EcDHFR.

(A) Solubility of overproduced His-tagged wild-type and TMP-resistant mutants of EcDHFR using a plasmid-based expression system. Equivalent amounts of L: lysate; S: soluble fraction; P: insoluble pellets were subjected to SDS–PAGE and visualized by staining with Coomassie Brilliant Blue. Data shown are representative of at least three independent experiments for each mutant. (B) Solubilization of Trp30Gly, Trp30Arg and Trp30Cys aggregates using NaCl (1 M), SDS (0.2%), Triton X-100 (TX-100; 0.1%) or NP-40 (0.1%). Equivalent amounts of insoluble aggregate (input) or solubilized fractions with indicated reagent were subjected to SDS–PAGE and visualized by staining with Coomassie Brilliant Blue. Data shown are representative of at least three independent experiments for each mutant. (C) Limited proteolysis of wild-type or Trp30Gly EcDHFR in lysates of E. coli using proteinase K. Lysates were treated with proteinase K for indicated times and EcDHFR, and its fragments were detected using immunoblotting with anti-DHFR polyclonal antibody. The full-length (FL) protein was proteolysed to two shorter bands (I and II) in E. coli lysates, and this was exacerbated by incubation with proteinase K. Wild-type accumulated proteinase K-resistant band II, while Trp30Gly did not. Data shown are representative of three independent experiments. (D) Proteinase K sensitivity of wild-type or Trp30Gly, Trp30Arg and Trp30Cys mutant EcDHFR in lysates of E. coli. The mean intensity of band II normalized to total protein ± SD from three independent experiments is indicated for wild-type and mutant EcDHFR.

Figure 2.
Destabilizing effect of Trp30 mutations on EcDHFR.

(A) Solubility of overproduced His-tagged wild-type and TMP-resistant mutants of EcDHFR using a plasmid-based expression system. Equivalent amounts of L: lysate; S: soluble fraction; P: insoluble pellets were subjected to SDS–PAGE and visualized by staining with Coomassie Brilliant Blue. Data shown are representative of at least three independent experiments for each mutant. (B) Solubilization of Trp30Gly, Trp30Arg and Trp30Cys aggregates using NaCl (1 M), SDS (0.2%), Triton X-100 (TX-100; 0.1%) or NP-40 (0.1%). Equivalent amounts of insoluble aggregate (input) or solubilized fractions with indicated reagent were subjected to SDS–PAGE and visualized by staining with Coomassie Brilliant Blue. Data shown are representative of at least three independent experiments for each mutant. (C) Limited proteolysis of wild-type or Trp30Gly EcDHFR in lysates of E. coli using proteinase K. Lysates were treated with proteinase K for indicated times and EcDHFR, and its fragments were detected using immunoblotting with anti-DHFR polyclonal antibody. The full-length (FL) protein was proteolysed to two shorter bands (I and II) in E. coli lysates, and this was exacerbated by incubation with proteinase K. Wild-type accumulated proteinase K-resistant band II, while Trp30Gly did not. Data shown are representative of three independent experiments. (D) Proteinase K sensitivity of wild-type or Trp30Gly, Trp30Arg and Trp30Cys mutant EcDHFR in lysates of E. coli. The mean intensity of band II normalized to total protein ± SD from three independent experiments is indicated for wild-type and mutant EcDHFR.

Mutations at Trp30 destabilize DHFR due to loss of intramolecular hydrophobic interactions

Trp30 is a buried residue and, unlike several other residues that are mutated in TMP-resistant DHFR (such as Pro21, Ile94 and Met20), is not directly involved in catalysis. Instead, the side chain of Trp30 is part of a hydrophobic tetrad along with three other residues, Phe153, Phe137 and Ile155 (Figure 3A). These three residues are located in a set of antiparallel β-sheets in the globular core of EcDHFR and their interaction with Trp30 may act as a hydrophobic clamp to provide rigidity and stability to the protein. Mutation of Trp30 to Gly, Cys or Arg is likely to perturb this hydrophobic tetrad which may, in turn, alter structural stability and TMP binding. To test this prediction, we analyzed the phenotypes of Phe137Ala, Phe153Ala and Ile155Ala mutant EcDHFR enzymes. All three mutations led to aggregation upon overexpression (Figure 3B) and enhanced proteinase K sensitivity (Figure 3C), similar to the phenotype of Trp30 mutations, demonstrating the importance of this hydrophobic interaction in maintaining the native fold of EcDHFR. Phe153Ala and Ile155Ala mutations also conferred TMP resistance to EcDHFR (Figure 3D) and demonstrated similar IC50 values as the Trp30Gly, Trp30Arg and Trp30Cys mutations (Figure 3E). Phe137Ala did not show enhanced resistance compared with wild-type EcDHFR (Figure 3D,E). Since substitution of Phe153, Ile155 and Phe137 with Ala phenocopied the Trp30 mutations in their reduced stability and Phe153Ala and Ile155Ala mutations also showed resistance to TMP, we concluded that the perturbation of this hydrophobic patch may be the mechanism by which Trp30 mutations confer TMP resistance to EcDHFR.

Intramolecular hydrophobic interactions of Trp30 in EcDHFR.

Figure 3.
Intramolecular hydrophobic interactions of Trp30 in EcDHFR.

(A) Interaction of the side chain of Trp30 with side chains of Phe153, Ile155 and Phe137. The crystal structure of EcDHFR protein (PDB: 7DFR) is shown in cartoon representation as an inset. Relevant residues are represented as sticks (C: green; N: blue; O: red). Shortest distances between the amino acid side chains (in Å) are shown and were measured using the PyMol software. (B) Distribution of Ile155Ala, Phe153Ala and Phe137Ala mutant EcDHFR proteins between soluble and insoluble fractions in lysates of E. coli. Equivalent amounts of L: lysate; S: soluble fraction; P: insoluble pellet were subjected to SDS–PAGE and visualized by staining with Coomassie Brilliant Blue. Data shown are representative of at least three independent experiments for each mutant. (C) Limited proteolysis of wild-type or Ile155Ala, Phe153Ala and Phe137Ala mutant EcDHFR proteins in lysates of E. coli using proteinase K. Lysates were treated with proteinase K for indicated times, and EcDHFR and its fragments were detected using immunoblotting with anti-DHFR polyclonal antibody. The mean intensity of band II normalized to total protein ± SD from three independent experiments is indicated for wild-type and mutant EcDHFR. (D and E) TMP sensitivity of E. coli strains expressing wild-type (WT) or indicated mutant DHFR enzymes. Serially diluted cultures (dilution factor indicated) were spotted onto TMP-containing plates and the maximum dilution allowing visible growth was noted at each TMP concentration to estimate IC50. A representative example of plates without TMP (−) or supplemented with 4 µg/ml TMP are shown (D). Mean IC50 ± SD from at least three independent experiments are plotted (E). Statistical significance was tested using an unpaired Student's t-test. ‘*’ indicates P-value <0.05, while ‘n.s.’ indicates P-value ≥0.05.

Figure 3.
Intramolecular hydrophobic interactions of Trp30 in EcDHFR.

(A) Interaction of the side chain of Trp30 with side chains of Phe153, Ile155 and Phe137. The crystal structure of EcDHFR protein (PDB: 7DFR) is shown in cartoon representation as an inset. Relevant residues are represented as sticks (C: green; N: blue; O: red). Shortest distances between the amino acid side chains (in Å) are shown and were measured using the PyMol software. (B) Distribution of Ile155Ala, Phe153Ala and Phe137Ala mutant EcDHFR proteins between soluble and insoluble fractions in lysates of E. coli. Equivalent amounts of L: lysate; S: soluble fraction; P: insoluble pellet were subjected to SDS–PAGE and visualized by staining with Coomassie Brilliant Blue. Data shown are representative of at least three independent experiments for each mutant. (C) Limited proteolysis of wild-type or Ile155Ala, Phe153Ala and Phe137Ala mutant EcDHFR proteins in lysates of E. coli using proteinase K. Lysates were treated with proteinase K for indicated times, and EcDHFR and its fragments were detected using immunoblotting with anti-DHFR polyclonal antibody. The mean intensity of band II normalized to total protein ± SD from three independent experiments is indicated for wild-type and mutant EcDHFR. (D and E) TMP sensitivity of E. coli strains expressing wild-type (WT) or indicated mutant DHFR enzymes. Serially diluted cultures (dilution factor indicated) were spotted onto TMP-containing plates and the maximum dilution allowing visible growth was noted at each TMP concentration to estimate IC50. A representative example of plates without TMP (−) or supplemented with 4 µg/ml TMP are shown (D). Mean IC50 ± SD from at least three independent experiments are plotted (E). Statistical significance was tested using an unpaired Student's t-test. ‘*’ indicates P-value <0.05, while ‘n.s.’ indicates P-value ≥0.05.

Trp30Gly DHFR is costly for E. coli and negatively epistatic to other TMP-resistant DHFR mutants

We next asked whether structural perturbation due to Trp30 mutations in EcDHFR was biologically relevant. Mutations that lower protein stability are expected to have lower steady-state expression levels and are associated with a fitness cost, and this has been shown to be true for many low-stability mutants of DHFR [53]. To test whether Trp30 mutations were costly for E. coli, we isolated spontaneous TMP-resistant E. coli strains by laboratory evolution and sequenced the folA locus of the isolates. Among the resistant isolates, we found two that had evolved the Trp30Gly and Trp30Cys mutations in the genomic copy of DHFR and hence proceeded to characterize the phenotypes of these strains. Both mutations led to a significant increase in IC50 for TMP compared with the wild-type (Figure 4A). As expected, the IC50 for TMP was lower than that seen with plasmid-based expression of these mutants. The Trp30Gly genomic mutation led to a reduction in the expression level of DHFR by ∼50%, consistent with the impact of this mutation on the stability of EcDHFR (Figure 4B,C). The Trp30Cys genomic mutation did not significantly alter the expression level, but reduced the amount of soluble DHFR by ∼40%, suggesting that the Trp30Gly mutation may be more detrimental to the stability of DHFR (Figure 4B,C) in agreement with computational prediction (Figure 1C). In line with the lower expression level, the Trp30Gly mutation also reduced the competitive fitness of E. coli by ∼10% (Figure 4D). The Trp30Cys mutant, which did not affect the levels of total DHFR in the cell, also had no measurable fitness cost (Figure 4D). Similar results were also obtained in defined medium (M9 medium) (Figure 4D). The fitness cost of the Trp30Gly mutant was alleviated upon supplementing the growth media with thymidine, which requires tetrahydrofolate for its biosynthesis, indicating that the fitness cost of Trp30Gly was indeed due to lower DHFR activity in the cell (Figure 4D).

Biological significance of Trp30Gly-mediated destabilization of EcDHFR.

Figure 4.
Biological significance of Trp30Gly-mediated destabilization of EcDHFR.

(A) IC50 of strains harboring the Trp30Cys or Trp30Gly mutations in their genomic copy of DHFR. Mean IC50 ± SD from at least three independent experiments are plotted. Statistical significance was tested using an unpaired Student's t-test. ‘*’ indicates P-value <0.05. (B and C) Endogenous expression level of wild-type or mutant EcDHFR in total lysates (B) or soluble fractions (C) of E. coli K-12 MG1655 or its TMP-resistant derivative strains using immunoblotting with anti-DHFR IgG. E. coli ΔfolA strain was used as a control to distinguish between specific (DHFR) and nonspecific (ns) immunoreactive bands. Mean expression level relative to wild-type ± SD from three independent biological replicates is shown below the immunoblot. (D) Fitness of wild-type or TMP-resistant E. coli strains harboring the indicated mutations relative to an E. coli ΔlacZ reporter strain in LB medium (black bars), LB supplemented with 100 µg/ml thymidine (gray bars) or M9 medium (white bars). Mean ± SD from at least three biological replicates are shown. Statistical significance was tested using an unpaired Student's t-test. ‘*’ indicates a P-value <0.05, while ‘n.s.’ indicates P-value ≥0.05.

Figure 4.
Biological significance of Trp30Gly-mediated destabilization of EcDHFR.

(A) IC50 of strains harboring the Trp30Cys or Trp30Gly mutations in their genomic copy of DHFR. Mean IC50 ± SD from at least three independent experiments are plotted. Statistical significance was tested using an unpaired Student's t-test. ‘*’ indicates P-value <0.05. (B and C) Endogenous expression level of wild-type or mutant EcDHFR in total lysates (B) or soluble fractions (C) of E. coli K-12 MG1655 or its TMP-resistant derivative strains using immunoblotting with anti-DHFR IgG. E. coli ΔfolA strain was used as a control to distinguish between specific (DHFR) and nonspecific (ns) immunoreactive bands. Mean expression level relative to wild-type ± SD from three independent biological replicates is shown below the immunoblot. (D) Fitness of wild-type or TMP-resistant E. coli strains harboring the indicated mutations relative to an E. coli ΔlacZ reporter strain in LB medium (black bars), LB supplemented with 100 µg/ml thymidine (gray bars) or M9 medium (white bars). Mean ± SD from at least three biological replicates are shown. Statistical significance was tested using an unpaired Student's t-test. ‘*’ indicates a P-value <0.05, while ‘n.s.’ indicates P-value ≥0.05.

Mutations in DHFR have been shown to accumulate sequentially during the evolution of TMP resistance and intramolecular epistasis between mutations in EcDHFR is well documented [22,35]. Therefore, we next asked what the impact of the Trp30Gly mutation would be on the phenotype of Pro21Leu and Ile94Leu mutations that independently confer TMP resistance without compromising solubility of DHFR (Figure 2A). Plasmid-borne Pro21Leu-Trp30Gly and Ile94Leu-Trp30Gly double mutants were both insoluble, like the Trp30Gly protein (Figure 5A). Interestingly, the double mutants conferred far lower TMP resistance that either of the single mutants, indicating that the Trp30Gly mutation was negatively epistatic to the Pro21Leu and Ile94Leu mutations (Figure 5B,C). These data indicated that the reduced stability of the Trp30 mutant DHFR enzymes and, in particular, of the Trp30Gly mutant was biologically relevant as not only did E. coli harboring this mutation have lower fitness than the wild-type but also the Trp30Gly mutation altered the resistant phenotypes of other TMP resistance-conferring DHFR mutations.

Biological significance of Trp30Gly-mediated destabilization of EcDHFR.

Figure 5.
Biological significance of Trp30Gly-mediated destabilization of EcDHFR.

(A) Distribution of Pro21Leu-Trp30Gly and Ile94Leu-Trp30Gly double-mutant EcDHFR proteins between soluble and insoluble fractions in lysates of E. coli. Equivalent amounts of L: lysate; S: soluble fraction; P: insoluble pellet were subjected to SDS–PAGE and visualized by staining with Coomassie Brilliant Blue. Data shown are representative of at least three independent experiments for each mutant. (B and C) TMP sensitivity of E. coli strains expressing WT or indicated mutant DHFR enzymes. Serially diluted cultures (dilution factor indicated) were spotted onto TMP-containing plates and the maximum dilution allowing visible growth was noted at each TMP concentration to estimate IC50. A representative example of plates without TMP (−) or supplemented with 4 µg/ml TMP is shown (B). Mean IC50 ± SD from at least three independent experiments are plotted (C). The IC50 values for both double mutants were lower than the lowest concentration of TMP used in the assay.

Figure 5.
Biological significance of Trp30Gly-mediated destabilization of EcDHFR.

(A) Distribution of Pro21Leu-Trp30Gly and Ile94Leu-Trp30Gly double-mutant EcDHFR proteins between soluble and insoluble fractions in lysates of E. coli. Equivalent amounts of L: lysate; S: soluble fraction; P: insoluble pellet were subjected to SDS–PAGE and visualized by staining with Coomassie Brilliant Blue. Data shown are representative of at least three independent experiments for each mutant. (B and C) TMP sensitivity of E. coli strains expressing WT or indicated mutant DHFR enzymes. Serially diluted cultures (dilution factor indicated) were spotted onto TMP-containing plates and the maximum dilution allowing visible growth was noted at each TMP concentration to estimate IC50. A representative example of plates without TMP (−) or supplemented with 4 µg/ml TMP is shown (B). Mean IC50 ± SD from at least three independent experiments are plotted (C). The IC50 values for both double mutants were lower than the lowest concentration of TMP used in the assay.

Phylogenetic variation at Trp30-equivalent position in bacterial DHFRs modulates innate and mutationally acquired TMP resistance

DHFR enzymes show very high degree of structural conservation among bacteria. Given the importance of Trp30 in the folding of EcDHFR, we assessed the evolutionary conservation of this residue in other bacterial DHFR enzymes. Surprisingly, this position varied greatly among the 61 bacterial DHFR sequences that we compared (Figure 6A,B). On the other hand, at least two other resistance-associated sites in DHFR, namely Pro21 and Ile94 in EcDHFR, are very highly conserved (Figure 6B). Phylogenetic analysis revealed that Trp at position 30 was found exclusively in Enterobacteria, such as Escherichia, Hemophilus and Wigglesworthia. However other bacterial DHFRs have His, Arg, Phe, Tyr, Leu or Asn at this position, His and Arg being the most common (Figure 6A). Unlike Trp30, Phe153 (which forms an aromatic stacking interaction with Trp30 in EcDHFR) was highly conserved. It was conservatively substituted for Tyr is some bacterial enzymes (Figure 6B) and only in 8 of the 61 sequences analyzed were aliphatic residues found at this position. There was too much variation found at the positions equivalent to Ile155 and Phe137 of EcDHFR among bacterial enzymes to be informative (Figure 6B).

Natural variation at Trp30 alters stability and resistance of DHFR enzymes.

Figure 6.
Natural variation at Trp30 alters stability and resistance of DHFR enzymes.

(A) Cladogram of bacterial DHFR enzymes built using the Neighbor-joining method in MEGA7 and rooted using human DHFR. Colored circles next to species names represent the amino acid found at the Trp30-equivalent position in the enzymes. If the amino acid at the Phe153-equivalent position is not aromatic, it is indicated next to the species name in parenthesis. (B) Sequence logos showing the relative conservation of amino acids at the indicated positions (numbered according to EcDHFR) among the 61 bacterial DHFR sequences analyzed. (C) Limited proteolysis of wild-type or Trp30Phe, Trp30Leu, Trp30His and Trp30Tyr mutant EcDHFR proteins in lysates of E. coli using proteinase K. Lysates were treated with proteinase K for indicated times, and EcDHFR and its fragments were detected using immunoblotting with anti-DHFR polyclonal antibody. The mean intensity of band II normalized to total protein ± SD from three independent experiments is indicated for WT and mutant EcDHFR. (D and E) TMP sensitivity of E. coli strains expressing WT or indicated mutant DHFR enzymes. Serially diluted cultures (dilution factor indicated) were spotted onto TMP-containing plates and the maximum dilution allowing visible growth was noted at each TMP concentration to estimate the IC50. A representative example of plates without TMP (∼) or supplemented with 4 µg/ml TMP is shown (D). Mean IC50 ± SD from at least three independent experiments are plotted (E). Statistical significance was tested using an unpaired Student's t-test. ‘*’ indicates P-value <0.05, while ‘n.s.’ indicates P-value ≥0.05.

Figure 6.
Natural variation at Trp30 alters stability and resistance of DHFR enzymes.

(A) Cladogram of bacterial DHFR enzymes built using the Neighbor-joining method in MEGA7 and rooted using human DHFR. Colored circles next to species names represent the amino acid found at the Trp30-equivalent position in the enzymes. If the amino acid at the Phe153-equivalent position is not aromatic, it is indicated next to the species name in parenthesis. (B) Sequence logos showing the relative conservation of amino acids at the indicated positions (numbered according to EcDHFR) among the 61 bacterial DHFR sequences analyzed. (C) Limited proteolysis of wild-type or Trp30Phe, Trp30Leu, Trp30His and Trp30Tyr mutant EcDHFR proteins in lysates of E. coli using proteinase K. Lysates were treated with proteinase K for indicated times, and EcDHFR and its fragments were detected using immunoblotting with anti-DHFR polyclonal antibody. The mean intensity of band II normalized to total protein ± SD from three independent experiments is indicated for WT and mutant EcDHFR. (D and E) TMP sensitivity of E. coli strains expressing WT or indicated mutant DHFR enzymes. Serially diluted cultures (dilution factor indicated) were spotted onto TMP-containing plates and the maximum dilution allowing visible growth was noted at each TMP concentration to estimate the IC50. A representative example of plates without TMP (∼) or supplemented with 4 µg/ml TMP is shown (D). Mean IC50 ± SD from at least three independent experiments are plotted (E). Statistical significance was tested using an unpaired Student's t-test. ‘*’ indicates P-value <0.05, while ‘n.s.’ indicates P-value ≥0.05.

Given that Trp30Arg led to TMP resistance and poor stability in EcDHFR (Figures 2 and 3E), we asked what the consequences of Trp30His, Trp30Phe, Trp30Tyr and Trp30Leu mutations in EcDHFR would be. Trp30Phe and Trp30Tyr were both resistant to proteinase K treatment, indicating that these mutations retained their ability to fold like wild-type EcDHFR (Figure 6C). Trp30His and Trp30Leu mutants failed to accumulate a proteinase K-resistant fragment and hence were mis-folded (Figure 6C). On the other hand, Trp30His and Trp30Tyr substitutions in E. coli DHFR, but not in Trp30Phe and Trp30Leu, conferred TMP resistance with IC50 values very similar to those of the Trp30Arg mutation (Figure 6D,E). These data indicated that the presence of an aromatic residue at the 30th position was crucial for structural stability of bacterial DHFRs, emphasizing the importance of the stacking interaction between Trp30 and Phe153. Furthermore, resistance to TMP was dependent more on the polarity of the residue at the 30th position than on aromaticity, polar/charged amino acids being more resistant to TMP than apolar/uncharged ones. Based on these data, we concluded that natural variation at this site may modulate the stability and TMP-resistance levels of bacterial DHFRs.

Importantly, DHFR from several bacterial pathogens such as M. tuberculosis (MtDHFR) and S. aureus (SaDHFR) naturally contain His instead of Trp30 (Figure 6A and Supplementary Figure S2). Since the Trp30His mutation in EcDHFR conferred TMP resistance, we asked whether His30Trp mutations in these enzymes would alter their TMP susceptibilities. We cloned and expressed MtDHFR and SaDHFR as well as their His30Trp mutant versions in E. coli, and checked the level of TMP resistance that they conferred. Interestingly, wild-type SaDHFR and MtDHFR conferred different levels of TMP resistance when expressed in E. coli, indicating different innate TMP-resistance levels of these two enzymes. SaDHFR expression resulted in a far greater increase in IC50 for TMP compared with EcDHFR, in line with the presence of His at the 30th position in this enzyme (Figure 7A). However, MtDHFR expression resulted in a lower IC50 than EcDHFR, most probably reflective of other changes in the biochemical properties of this enzyme (Figure 7A). Despite these differences, both SaDHFR and MtDHFR were sensitized to TMP by the His30Trp mutation (Figure 7A,B). These data indicated that the sequence variation naturally occurring at the Trp30-equivalent position among bacterial DHFRs modulated the level of TMP resistance also of other bacterial DHFRs.

Analysis of Trp30-equivalent positions in other bacterial DHFRs.

Figure 7.
Analysis of Trp30-equivalent positions in other bacterial DHFRs.

(A and B) TMP sensitivity of E. coli strains expressing wild-type SaDHFR and MtDHFR or His30Trp mutant SaDHFR and MtDHFR enzymes. Serially diluted cultures (dilution factor indicated) were spotted onto TMP-containing plates and the maximum dilution allowing visible growth was noted at each TMP concentration to estimate the IC50. A representative example of plates without TMP (−) or supplemented with 4 µg/ml TMP is shown in (A). Mean IC50 ± SD from at least three independent experiments are plotted in (B). Statistical significance was tested comparing mutant with respective wild-type DHFR enzyme using an unpaired Student's t-test. ‘*’ indicates P-value <0.05. (C) Solubility of overproduced His-tagged wild-type and SaDHFR and MtDHFR and their His30Trp mutant versions using a plasmid-based expression system. Equivalent amounts of L: lysate; S: soluble fraction; P: insoluble pellets were subjected to SDS–PAGE and visualized by staining with Coomassie Brilliant Blue. Data shown are representative of three experiments. (D) Predicted stabilities of mutations at Trp30-equivalent positions in EcDHFR, SaDHFR, MtDHFR, BaDHFR and MpDHFR plotted against ΔG (kcal/mol) of transfer from water to n-octanol (hydrophobicity scale; Wimley and White [51]) of 20 amino acids. The positions of Trp, His and Arg on the hydrophobicity scale are indicated.

Figure 7.
Analysis of Trp30-equivalent positions in other bacterial DHFRs.

(A and B) TMP sensitivity of E. coli strains expressing wild-type SaDHFR and MtDHFR or His30Trp mutant SaDHFR and MtDHFR enzymes. Serially diluted cultures (dilution factor indicated) were spotted onto TMP-containing plates and the maximum dilution allowing visible growth was noted at each TMP concentration to estimate the IC50. A representative example of plates without TMP (−) or supplemented with 4 µg/ml TMP is shown in (A). Mean IC50 ± SD from at least three independent experiments are plotted in (B). Statistical significance was tested comparing mutant with respective wild-type DHFR enzyme using an unpaired Student's t-test. ‘*’ indicates P-value <0.05. (C) Solubility of overproduced His-tagged wild-type and SaDHFR and MtDHFR and their His30Trp mutant versions using a plasmid-based expression system. Equivalent amounts of L: lysate; S: soluble fraction; P: insoluble pellets were subjected to SDS–PAGE and visualized by staining with Coomassie Brilliant Blue. Data shown are representative of three experiments. (D) Predicted stabilities of mutations at Trp30-equivalent positions in EcDHFR, SaDHFR, MtDHFR, BaDHFR and MpDHFR plotted against ΔG (kcal/mol) of transfer from water to n-octanol (hydrophobicity scale; Wimley and White [51]) of 20 amino acids. The positions of Trp, His and Arg on the hydrophobicity scale are indicated.

Was the change in TMP resistance also associated with an alteration in stability of these two bacterial enzymes? Since the EcDHFR polyclonal antibody used in this study reacted poorly with other bacterial DHFR enzymes, we were unable to assess the proteinase K sensitivity of SaDHFR and MtDHFR. Furthermore, when induced at low temperatures (16–18°C), we found that wild-type and His30Trp mutants of MtDHFR and SaDHFR were almost exclusively localized to the soluble fraction (data not shown). Interestingly, when induced at 37°C, His30Trp SaDHFR showed greater solubility than the wild-type, indicating that the mutant may indeed be more stable than the wild-type. Wild-type and His30Trp MtDHFR continued to be soluble even under these conditions. We also used a computational approach to complement these experiments. In line with our experimental data, substitution of Trp30 with hydrophilic residues was predicted to reduce stability of EcDHFR to a far greater extent than hydrophobic and aromatic residues, providing validation for the reliability of these predictions (Figure 7D and Supplementary Figure S3). A similar trend was observed for SaDHFR, and mutation of His30 to hydrophobic/aromatic residues like Phe or Trp had a stabilizing effect in this enzyme as expected (Figure 7D and Supplementary Figure S3). In the case of MtDHFR too, a similar trend was observed, though the differences in stabilities between substitution of His30 with hydrophilic and hydrophobic residues were lower compared with SaDHFR and EcDHFR, indicating that other stabilizing interactions in this protein may reduce the dependence on His30 for global stability (Figure 7D and Supplementary Figure S3). Extending this analysis to two other structurally characterized DHFR enzymes from M. profunda (MpDHFR) and B. anthracis (BaDHFR) also confirmed that the Trp30-equivalent position has a crucial role to play in regulating the stability of bacterial DHFR enzymes in general (Figure 7D and Supplementary Figure S3). Taken together, these observations validated the generality of the importance of Trp30-equivalent residues determining the stability and TMP susceptibility in bacterial DHFR enzymes.

Discussion

In the present study, we have focused our attention on TMP resistance acquired by mutation of Trp30 in EcDHFR. This residue is set apart from other resistance-associated sites in DHFR that have been investigated in earlier studies [36] due to the following reasons. Firstly, Trp30, though proximate to the active site, contributes to substrate binding only indirectly by co-ordinating a substrate-interacting water molecule along with two other neighboring residues. The fact that a mutation-like Trp30Gly, or replacement of Trp with residues like Phe in other bacterial DHFRs, is tolerated indicates that while this residue may contribute to substrate binding, its role in this regard may be an accessory one. Indeed, the greater degree of sequence variability at this site among bacterial DHFRs compared with other resistance-associated sites that directly participate in substrate binding and catalysis is in line with this fact (Figure 6B). Secondly, Trp30 has been suggested to participate in the folding and stabilization of DHFR [54]. Mutations at such structurally crucial sites are generally prohibited due to large structural perturbations that they are expected to produce. Yet, mutations at Trp30 are commonly isolated in screens for TMP resistance [21]. Thirdly, Trp30 is part of the structural core of DHFR, rather than being present in a loop and, hence, mutations at this site would affect the global stability of the protein, as observed from our experiments, rather than more local effects as have been previously observed with active-site residues. Finally, as we have shown in this study, natural variation at this site among bacterial DHFRs can potentially contribute to the intrinsic TMP resistance of the DHFR enzyme family, providing novelty to this study.

The interactions between Trp30 and other residues in EcDHFR provide a likely structural explanation for the stability–function trade-off associated with mutations at this position. As demonstrated experimentally in this study, Trp30 forms hydrophobic interactions with Ile155, Phe137 and Phe153 (Figure 3). The side chains of Ile5 and Tyr111 are also within interaction distance of Trp30 and hence may also interact with it. In addition, Trp30 forms main chain interactions with Ala26, Asp27 and Leu28 that are part of a structurally conserved alpha helix involved in substrate binding, as well as with Arg33 and Asn34 (Figure 8A). Asp27 is directly involved in catalysis [55], and Leu28 is known to be involved in binding folate as well as TMP [56,57] and is itself a TMP resistance-associated site [22,36]. We propose that the anchoring of Trp30 to Ile155, Phe137 and Phe153 through side chain–side chain hydrophobic interactions may provide necessary conformational stability to substrate-binding and catalytic residues (Figure 8A,B). Mutation of Trp30 to residues like Gly, Arg and His would perturb this hydrophobic anchoring, leading to greater flexibility of substrate-binding residues, reduced binding to TMP and consequently drug resistance, but at the cost of global stability. This model is substantiated by structural characterization of the Trp30Arg mutant by Cammarata et al. [58] using UVPD that showed greater structural flexibility in the substrate-binding pocket in the mutant relative to the wild-type and folding studies on the Trp30Leu mutant by Ohmae et al. [54] that reported greater structural perturbation in the mutant compared with the wild-type. We propose that Trp30 in EcDHFR acts as a ‘toggle switch’ that can be ‘flipped’ to allow greater flexibility and TMP resistance by mutation to less hydrophobic/aliphatic residues.

Intramolecular interactions between Trp30-equivalent positions in bacterial DHFRs.

Figure 8.
Intramolecular interactions between Trp30-equivalent positions in bacterial DHFRs.

(A) Structural superposition of EcDHFR (7DFR; cyan cartoon) with SaDHFR (2W9T, pink cartoon), MtDHFR (1DG8, salmon cartoon) and BaDHFR (3FL9, yellow cartoon). The global RMSD between structures is shown. Trp30, Phe153, Phe137 and Ile155 from EcDHFR are shown in line representations (C: green; N: blue; O: red) and corresponding residues from other enzymes are shown as sticks (C: magenta; N: blue; O: red). (B) Schematic representation of the intramolecular interactions between Trp30 and other residues in EcDHFR. A role for Trp30 as a ‘toggle switch’ between maintenance of active-site conformation and global stability is suggested. Equivalent residues and interactions in BaDHFR, SaDHFR and MtDHFR are shown. Key for the interactions is provided.

Figure 8.
Intramolecular interactions between Trp30-equivalent positions in bacterial DHFRs.

(A) Structural superposition of EcDHFR (7DFR; cyan cartoon) with SaDHFR (2W9T, pink cartoon), MtDHFR (1DG8, salmon cartoon) and BaDHFR (3FL9, yellow cartoon). The global RMSD between structures is shown. Trp30, Phe153, Phe137 and Ile155 from EcDHFR are shown in line representations (C: green; N: blue; O: red) and corresponding residues from other enzymes are shown as sticks (C: magenta; N: blue; O: red). (B) Schematic representation of the intramolecular interactions between Trp30 and other residues in EcDHFR. A role for Trp30 as a ‘toggle switch’ between maintenance of active-site conformation and global stability is suggested. Equivalent residues and interactions in BaDHFR, SaDHFR and MtDHFR are shown. Key for the interactions is provided.

Among 30 bacterial DHFR enzymes for which thermostability has been experimentally determined [59], 8 of the 10 most thermostable enzymes harbor either a Trp, Tyr or Phe residue at the Trp30-equivalent position. The other two harbor His at this position, while enzymes with Arg or amino acids with aliphatic side chains tend to have lower thermostabilities (Supplementary Figure S4). Though correlative, these analyses, coupled with the experimental results of this study, make a compelling case for a ‘toggle-switch’ like role of Trp30-equivalent residues in other bacterial DHFR enzymes as well. This may not be surprising, given the conserved structural context of Trp30-equivalent residues in bacterial DHFRs. For instance, in BaDHFR, Tyr31 forms a side chain–side chain aromatic interaction with Tyr155 (Trp30–Phe153 in EcDHFR) and main chain–main chain interactions with catalytic residues, Glu28 and Leu29 reminiscent of the interactions seen in EcDHFR. Furthermore, the TMP-resistance/stability effects of mutations at Trp30-equivalent residues in SaDHFR and MtDHFR are also readily explained by intramolecular interactions (Figure 8A,B). In SaDHFR for instance, His30, unable to interact through its side chain with aromatic residues in its neighborhood (Figure 8A,B), results in phenocopying the EcDHFR Trp30His mutant in its TMP resistance as well as stability (Figure 7). Interestingly, MtDHFR may have evolved to compensate for lower stability due to the presence of His30 (instead of Trp30) by establishing a novel side chain ionic interaction with Asp111 (Figure 8A,B). Indeed, negatively charged residues equivalent to Asp111 are seen also in other bacterial DHFRs such as Lactobacillus delbruckeii and Comamonas testeroni that have a His instead of Trp30. This novel ionic interaction may also explain the replacement of a bulky aromatic Phe153 (EcDHFR) with the smaller aliphatic Leu153 in MtDHFR (Figure 8A,B), the markedly different stability profile predictions for mutation of His30 in MtDHFR and SaDHFR as well as different levels of TMP resistance of these enzymes despite the presence of His30 in both of them (Figure 7). It is important to note, however, that MtDHFR represents the exception rather than the rule and is 1 of only 8 DHFR enzymes among the 61 sequences analyzed by us that may have evolved to rewire interactions at this site. The applicability of these findings to other bacterial DHFRs is particularly relevant since clinical TMP resistance is more common in Gram-positives like S. aureus than in bacteria such as E. coli in which plasmid-borne resistance is the dominant form [21,2730].

Loss in stability due to the mutation of Trp30 had two important biologically relevant repercussions on E. coli (Figures 4 and 5). The first of these was a loss in fitness of E. coli harboring the Trp30Gly mutation. Fitness costs of antimicrobial resistance are a widespread phenomenon in bacteria, though in the case of TMP resistance costs of resistance have not been reported [60]. The lack of reversal of TMP resistance in the clinical setting has been causally linked to the lack of fitness costs in vitro [61]. Our finding that Trp30Gly is costly due to its poor stability and lower expression level therefore needs further investigation and demonstrates that TMP resistance can occasionally be costly in E. coli. Interestingly, the Trp30Cys mutation, despite lower stability, did not have a fitness cost, indicating that DHFR may be able to tolerate reasonably large structural perturbations. A possible reason for this could be the low endogenous expression level of EcDHFR [62], which may preclude aggregation. The second important repercussion of the Trp30Gly mutation was its negative epistasis with other TMP resistance-associated mutations. This observation is reminiscent of results in TEM-1 β-lactamase and HIV-I protease, where loss in stability due to a single mutation inevitably affected the phenotypes of subsequent mutations [911]. Furthermore, in the context of TMP-resistant mutations, Palmer et al. [35] demonstrated that they are largely positively epistatic. However, in their study too, some instances of negative epistasis were observed and several of these involved Trp30Gly or Trp30Arg [35]. Our study may provide a mechanistic explanation for their observations. We propose that since interactions between Trp30 and its neighboring residues in EcDHFR provide global stability to the protein, these interactions may serve to buffer local destabilization due to mutations at active-site/catalytic residues. In the absence of this stabilizing interaction, the thermodynamic costs of other resistance-conferring mutations may be too great to permit their survival.

Surprisingly, the Pro21Leu mutation has been demonstrated to destabilize EcDHFR [36], even though in our study we did not detect a significant change in the distribution of the Pro21Leu mutant between soluble and insoluble fractions when overproduced using a plasmid-based system. This discrepancy may be explained by the fact that small/local destabilization may be insufficient to lead to a change in solubility upon overproduction. In this regard, it is also important to note that the Pro21Leu mutation was computationally predicted to be mildly stabilizing in contrast with published results [36]. Similarly, while the Leu28Arg mutation was predicted to be mildly destabilizing, it has been demonstrated to enhance the stability of EcDHFR [36]. In light of these discrepancies between experimental data and computation predictions of mutant stabilities, we would advise caution in relying on computational predictions alone in assessing the impact of mutations on the stability of proteins. This is particularly true in the case of changes with small effect sizes, as seen for mutants like Pro21Leu or Leu28Arg. On the other hand, for mutants like Trp30Gly that were predicted to have large effects on the stability of DHFR, computational prediction may accurately assess mutational impact on stability as demonstrated by this study.

Interestingly, phylogenetic analysis undertaken in this study revealed that Trp30 of EcDHFR is typically replaced by a His/Arg in most bacteria (Figure 6A,B). Both these substitutions in EcDHFR destabilized the enzyme. Furthermore, since Trp30 was limited to the Enterobacteriaceae, this may be a clade-specific adaptation (Figure 6A). Why then have other bacterial enzymes not evolved to be more stable? While there can be no definitive answer to this based on our study, we propose that the presence of His/Arg at the Trp30-equivalent position may confer other desirable properties to bacterial DHFRs that may be selected for, making these the most common amino acids at the Trp30-equivalent position. For instance, the presence of His at this position in Streptococcus pneumoniae DHFR (StDHFR) has been shown to alter the pH range of the enzyme [63]. StDHFR was also found to have greater substrate turnover than EcDHFR [63], despite higher substrate Km and this higher turnover may be under selection. Alternatively, the cellular environment of Enterobacteria like E. coli may have more stringent protein-folding checks than other bacteria or may face extracellular environments that demand greater protein stability. Indeed, Behrshtein et al. [59] showed that DHFRs from other bacteria are less stable in E. coli cells, leading to their degradation by the Lon protease. Finally, it would be imprudent to negate the possibility that the reason Enterobacteria harbor a Trp at this position may be due to traits other than stability, and the higher stability due to Trp30 may be a spandrel [64] rather than having been directly selected. Further biochemical characterization of Trp30 mutant EcDHFR enzymes will no doubt shed light on this. Regardless of the mechanistic details, however, sequence variability at Trp30-equivalent sites may alter the potential for acquisition of TMP resistance by mutations at this site. Indeed, TMP-resistant SaDHFR rarely harbors mutations at His30, already a destabilizing amino acid at this position, and the only known mutation at this site among clinical strains is His30Asn [27].

In summary, we have shown that resistance to TMP mediated by mutation at a structurally conserved, but non-catalytic site in bacterial DHFRs comes at the expense of global protein stability. We have also shown that analogous to mutationally acquired TMP resistance, natural variation at this site among bacterial DHFR enzymes alters their innate TMP resistance and hence may contribute to the variable susceptibility of bacterial species to TMP. Our results add to the growing evidence supporting the relevance of trade-offs in molecular evolution. Moreover, we show that compromise in global stability may be permissible during direction selection and that bacterial DHFRs may provide an opportunity to understand the mechanistic details of how these trade-offs are mediated and negotiated. In the future, it will also be instructive to explore such trade-offs in the evolution of resistance to other antibiotics to test the generality of our conclusions.

Abbreviations

     
  • DHFR

    dihydrofolate reductase

  •  
  • EcDHFR

    dihydrofolate reductase from Escherichia coli

  •  
  • LA

    Luria Bertani agar plates

  •  
  • LB

    Luria Bertani

  •  
  • MtDHFR

    DHFR from Mycobacterium tuberculosis

  •  
  • PCR

    polymerase chain reaction

  •  
  • SaDHFR

    DHFR from Staphylococcus aureus

  •  
  • StDHFR

    Streptococcus pneumoniae DHFR

  •  
  • TMP

    trimethoprim

Author Contribution

N.M. conceived the present study and wrote the manuscript. N.M., S.B., M.P. and P.S. performed the experiments and analyzed data.

Funding

This project was funded by IISER (Pune) and the Department of Science and Technology (DST), Govt. of India. N.M. is a recipient of the Innovation in Science Pursuit for Inspired Research (INSPIRE) fellowship from the DST. S.B. is a recipient of the INSPIRE scholarship from the DST.

Acknowledgments

We acknowledge Prof. Sandhya S. Visweswariah for providing plasmids and M. bovis BCG genomic DNA. We acknowledge Prof. Peter Wright for providing the E. coli ΔfolA strain. We acknowledge Dr Harinath Chakrapani for providing S. aureus genomic DNA.

Competing Interests

The Authors declare that there are no competing interests associated with the manuscript.

References

References
1
Soskine
,
M.
and
Tawfik
,
D.S.
(
2010
)
Mutational effects and the evolution of new protein functions
.
Nat. Rev. Genet.
11
,
572
582
2
Studer
,
R.A.
,
Dessailly
,
B.H.
and
Orengo
,
C.A.
(
2013
)
Residue mutations and their impact on protein structure and function: detecting beneficial and pathogenic changes
.
Biochem. J.
449
,
581
594
3
Tokuriki
,
N.
,
Stricher
,
F.
,
Serrano
,
L.
and
Tawfik
,
D.S.
(
2008
)
How protein stability and new functions trade off
.
PLoS Comput. Biol.
4
,
e1000002
4
Nagatani
,
R.A.
,
Gonzalez
,
A.
,
Shoichet
,
B.K.
,
Brinen
,
L.S.
and
Babbitt
,
P.C.
(
2007
)
Stability for function trade-offs in the enolase superfamily ‘catalytic module’
.
Biochemistry
46
,
6688
6695
5
Studer
,
R.A.
,
Christin
,
P.-A.
,
Williams
,
M.A.
and
Orengo
,
C.A.
(
2014
)
Stability-activity tradeoffs constrain the adaptive evolution of RubisCO
.
Proc. Natl Acad. Sci. U.S.A.
111
,
2223
2228
6
Li
,
M.
,
Kales
,
S.C.
,
Ma
,
K.
,
Shoemaker
,
B.A.
,
Crespo-Barreto
,
J.
,
Cangelosi
,
A.L.
et al. 
(
2016
)
Balancing protein stability and activity in cancer: a new approach for identifying driver mutations affecting CBL ubiquitin ligase activation
.
Cancer Res.
76
,
561
571
7
Julian
,
M.C.
,
Li
,
L.
,
Garde
,
S.
,
Wilen
,
R.
and
Tessier
,
P.M.
(
2017
)
Efficient affinity maturation of antibody variable domains requires co-selection of compensatory mutations to maintain thermodynamic stability
.
Sci. Rep.
7
,
45259
8
Xu
,
B.-L.
,
Dai
,
M.
,
Chen
,
Y.
,
Meng
,
D.
,
Wang
,
Y.
,
Fang
,
N.
et al. 
(
2015
)
Improving the thermostability and activity of a thermophilic subtilase by incorporating structural elements of its psychrophilic counterpart
.
Appl. Environ. Microbiol.
81
,
6302
6313
9
Figliuzzi
,
M.
,
Jacquier
,
H.
,
Schug
,
A.
,
Tenaillon
,
O.
and
Weigt
,
M.
(
2016
)
Coevolutionary landscape inference and the context-dependence of mutations in beta-lactamase TEM-1
.
Mol. Biol. Evol.
33
,
268
280
10
Jacquier
,
H.
,
Birgy
,
A.
,
Le Nagard
,
H.
,
Mechulam
,
Y.
,
Schmitt
,
E.
,
Glodt
,
J.
et al. 
(
2013
)
Capturing the mutational landscape of the beta-lactamase TEM-1
.
Proc. Natl Acad. Sci. U.S.A.
110
,
13067
13072
11
Olabode
,
A.S.
,
Kandathil
,
S.M.
,
Lovell
,
S.C.
and
Robertson
,
D.L.
(
2017
)
Adaptive HIV-1 evolutionary trajectories are constrained by protein stability
.
Virus Evol.
3
,
vex019
12
Kaltenbach
,
M.
,
Emond
,
S.
,
Hollfelder
,
F.
and
Tokuriki
,
N.
(
2016
)
Functional trade-offs in promiscuous enzymes cannot be explained by intrinsic mutational robustness of the native activity
.
PLoS Genet.
12
,
e1006305
13
Siddiqui
,
K.S.
(
2017
)
Defying the activity–stability trade-off in enzymes: taking advantage of entropy to enhance activity and thermostability
.
Crit. Rev. Biotechnol.
37
,
309
322
14
Bradford
,
P.A.
(
2001
)
Extended-spectrum β-lactamases in the 21st century: characterization, epidemiology, and detection of this important resistance threat
.
Clin. Microbiol. Rev.
14
,
933
951
15
Weinreich
,
D.M.
,
Delaney
,
N.F.
,
Depristo
,
M.A.
and
Hartl
,
D.L.
(
2006
)
Darwinian evolution can follow only very few mutational paths to fitter proteins
.
Science
312
,
111
114
16
Shafer
,
R.W.
and
Schapiro
,
J.M.
(
2008
)
HIV-1 drug resistance mutations: an updated framework for the second decade of HAART
.
AIDS Rev.
10
,
67
84
PMID:
[PubMed]
17
Chang
,
M.W.
and
Torbett
,
B.E.
(
2011
)
Accessory mutations maintain stability in drug-resistant HIV-1 protease
.
J. Mol. Biol.
410
,
756
760
18
Campbell
,
E.A.
,
Korzheva
,
N.
,
Mustaev
,
A.
,
Murakami
,
K.
,
Nair
,
S.
,
Goldfarb
,
A.
et al. 
(
2001
)
Structural mechanism for rifampicin inhibition of bacterial RNA polymerase
.
Cell
104
,
901
912
19
Bereswill
,
S.
,
Krainick
,
C.
,
Stähler
,
F.
,
Herrmann
,
L.
and
Kist
,
M.
(
2003
)
Analysis of the rdxA gene in high-level metronidazole-resistant clinical isolates confirms a limited use of rdxA mutations as a marker for prediction of metronidazole resistance in Helicobacter pylori
.
FEMS Immunol. Med. Microbiol.
36
,
193
198
20
Mirzaei
,
N.
,
Poursina
,
F.
,
Moghim
,
S.
,
Rahimi
,
E.
and
Safaei
,
H.G.
(
2014
)
The mutation of the rdxA gene in metronidazole-resistant Helicobacter pylori clinical isolates
.
Adv. Biomed. Res.
3
,
90
21
Flensburg
,
J.
and
Skold
,
O.
(
1987
)
Massive overproduction of dihydrofolate reductase in bacteria as a response to the use of trimethoprim
.
Eur. J. Biochem.
162
,
473
476
22
Toprak
,
E.
,
Veres
,
A.
,
Michel
,
J.-B.
,
Chait
,
R.
,
Hartl
,
D.L.
and
Kishony
,
R.
(
2012
)
Evolutionary paths to antibiotic resistance under dynamically sustained drug selection
.
Nat. Genet.
44
,
101
105
23
Watson
,
M.
,
Liu
,
J.-W.
and
Ollis
,
D.
(
2007
)
Directed evolution of trimethoprim resistance in Escherichia coli
.
FEBS J.
274
,
2661
2671
24
Weber
,
I.T.
and
Agniswamy
,
J.
(
2009
)
HIV-1 protease: structural perspectives on drug resistance
.
Viruses
1
,
1110
1136
25
Grim
,
S.A.
,
Rapp
,
R.P.
,
Martin
,
C.A.
and
Evans
,
M.E.
(
2005
)
Trimethoprim-sulfamethoxazole as a viable treatment option for infections caused by methicillin-resistant Staphylococcus aureus
.
Pharmacotherapy
25
,
253
264
26
Palomino
,
J.C.
and
Martin
,
A.
(
2016
)
The potential role of trimethoprim-sulfamethoxazole in the treatment of drug-resistant tuberculosis
.
Future Microbiol.
11
,
539
547
27
Dale
,
G.E.
,
Broger
,
C.
,
D'Arcy
,
A.
,
Hartman
,
P.G.
,
DeHoogt
,
R.
,
Jolidon
,
S.
et al. 
(
1997
)
A single amino acid substitution in Staphylococcus aureus dihydrofolate reductase determines trimethoprim resistance
.
J. Mol. Biol.
266
,
23
30
28
Heaslet
,
H.
,
Harris
,
M.
,
Fahnoe
,
K.
,
Sarver
,
R.
,
Putz
,
H.
,
Chang
,
J.
et al. 
(
2009
)
Structural comparison of chromosomal and exogenous dihydrofolate reductase from Staphylococcus aureus in complex with the potent inhibitor trimethoprim
.
Proteins: Struct., Funct., Bioinf.
76
,
706
717
29
Then
,
R.L.
and
Riggenbach
,
H.
(
1978
)
Dihydrofolate reductases in some folate-requiring bacteria with low trimethoprim susceptibility
.
Antimicrob. Agents Chemother.
14
,
112
117
30
Vickers
,
A.A.
,
Potter
,
N.J.
,
Fishwick
,
C.W.
,
Chopra
,
I.
and
O'Neill
,
A.J.
(
2009
)
Analysis of mutational resistance to trimethoprim in Staphylococcus aureus by genetic and structural modelling techniques
.
J. Antimicrob. Chemother.
63
,
1112
1117
31
Adrian
,
P.V.
and
Klugman
,
K.P.
(
1997
)
Mutations in the dihydrofolate reductase gene of trimethoprim-resistant isolates of Streptococcus pneumoniae
.
Antimicrob. Agents Chemother.
41
,
2406
2413
PMID:
[PubMed]
32
Stokes
,
A.
and
Lacey
,
R.W.
(
1978
)
Effect of thymidine on activity of trimethoprim and sulphamethoxazole
.
J. Clin. Pathol.
31
,
165
171
33
Baym
,
M.
,
Lieberman
,
T.D.
,
Kelsic
,
E.D.
,
Chait
,
R.
,
Gross
,
R.
,
Yelin
,
I.
et al. 
(
2016
)
Spatiotemporal microbial evolution on antibiotic landscapes
.
Science
353
,
1147
1151
34
Palmer
,
A.C.
and
Kishony
,
R.
(
2014
)
Opposing effects of target overexpression reveal drug mechanisms
.
Nat. Commun.
5
,
4296
35
Palmer
,
A.C.
,
Toprak
,
E.
,
Baym
,
M.
,
Kim
,
S.
,
Veres
,
A.
,
Bershtein
,
S.
et al. 
(
2015
)
Delayed commitment to evolutionary fate in antibiotic resistance fitness landscapes
.
Nat. Commun.
6
,
7385
36
Rodrigues
,
J.V.
,
Bershtein
,
S.
,
Li
,
A.
,
Lozovsky
,
E.R.
,
Hartl
,
D.L.
and
Shakhnovich
,
E.I.
(
2016
)
Biophysical principles predict fitness landscapes of drug resistance
.
Proc. Natl Acad. Sci. U.S.A.
113
,
E1470
E1478
37
Bershtein
,
S.
,
Choi
,
J.-M.
,
Bhattacharyya
,
S.
,
Budnik
,
B.
and
Shakhnovich
,
E.
(
2015
)
Systems-level response to point mutations in a core metabolic enzyme modulates genotype-phenotype relationship
.
Cell Rep.
11
,
645
656
38
Bershtein
,
S.
,
Mu
,
W.
and
Shakhnovich
,
E.I.
(
2012
)
Soluble oligomerization provides a beneficial fitness effect on destabilizing mutations
.
Proc. Natl Acad. Sci. U.S.A.
109
,
4857
4862
39
Abriata
,
L.A.
,
Salverda
,
M.L.
and
Tomatis
,
P.E.
(
2012
)
Sequence-function-stability relationships in proteins from datasets of functionally annotated variants: the case of TEM β-lactamases
.
FEBS Lett.
586
,
3330
3335
40
Bhabha
,
G.
,
Ekiert
,
D.C.
,
Jennewein
,
M.
,
Zmasek
,
C.M.
,
Tuttle
,
L.M.
,
Kroon
,
G.
et al. 
(
2013
)
Divergent evolution of protein conformational dynamics in dihydrofolate reductase
.
Nat. Struct. Mol. Biol.
20
,
1243
1249
41
Kitagawa
,
M.
,
Ara
,
T.
,
Arifuzzaman
,
M.
,
Ioka-Nakamichi
,
T.
,
Inamoto
,
E.
,
Toyonaga
,
H.
et al. 
(
2006
)
Complete set of ORF clones of Escherichia coli ASKA library (a complete set of E. coli K-12 ORF archive): unique resources for biological research
.
DNA Res.
12
,
291
299
42
Shenoy
,
A.R.
and
Visweswariah
,
S.S.
(
2003
)
Site-directed mutagenesis using a single mutagenic oligonucleotide and DpnI digestion of template DNA
.
Anal. Biochem.
319
,
335
336
43
Wiser
,
M.J.
and
Lenski
,
R.E.
(
2015
)
A comparison of methods to measure fitness in Escherichia coli
.
PLoS ONE
10
,
e0126210
44
Kumar
,
S.
,
Stecher
,
G.
and
Tamura
,
K.
(
2016
)
MEGA7: molecular evolutionary genetics analysis version 7.0 for bigger datasets
.
Mol. Biol. Evol.
33
,
1870
1874
45
Bystroff
,
C.
,
Oatley
,
S.J.
and
Kraut
,
J.
(
1990
)
Crystal structures of Escherichia coli dihydrofolate reductase: the NADP+ holoenzyme and the folate. NADP+ ternary complex. Substrate binding and a model for the transition state
.
Biochemistry
29
,
3263
3277
46
Bourne
,
C.R.
,
Bunce
,
R.A.
,
Bourne
,
P.C.
,
Berlin
,
K.D.
,
Barrow
,
E.W.
and
Barrow
,
W.W.
(
2009
)
Crystal structure of Bacillus anthracis dihydrofolate reductase with the dihydrophthalazine-based trimethoprim derivative RAB1 provides a structural explanation of potency and selectivity
.
Antimicrob. Agents Chemother.
53
,
3065
3073
47
Li
,
R.
,
Sirawaraporn
,
R.
,
Chitnumsub
,
P.
,
Sirawaraporn
,
W.
,
Wooden
,
J.
,
Athappilly
,
F.
et al. 
(
2000
)
Three-dimensional structure of M. tuberculosis dihydrofolate reductase reveals opportunities for the design of novel tuberculosis drugs
.
J. Mol. Biol.
295
,
307
323
48
Hay
,
S.
,
Evans
,
R.M.
,
Levy
,
C.
,
Loveridge
,
E.J.
,
Wang
,
X.
,
Leys
,
D.
et al. 
(
2009
)
Are the catalytic properties of enzymes from piezophilic organisms pressure adapted?
ChemBioChem
10
,
2348
2353
49
Maiti
,
R.
,
Van Domselaar
,
G.H.
,
Zhang
,
H.
and
Wishart
,
D.S.
(
2004
)
Superpose: a simple server for sophisticated structural superposition
.
Nucleic Acids Res.
32
,
W590
W594
50
Capriotti
,
E.
,
Fariselli
,
P.
and
Casadio
,
R.
(
2005
)
I-Mutant2.0: predicting stability changes upon mutation from the protein sequence or structure
.
Nucleic Acids Res.
33
,
W306
W310
51
Wimley
,
W.C.
and
White
,
S.H.
(
1996
)
Experimentally determined hydrophobicity scale for proteins at membrane interfaces
.
Nat. Struct. Biol.
3
,
842
848
52
Kasper
,
J.R.
,
Liu
,
P.-F.
and
Park
,
C.
(
2014
)
Structure of a partially unfolded form of Escherichia coli dihydrofolate reductase provides insight into its folding pathway
.
Protein Sci.
23
,
1728
1737
53
Bershtein
,
S.
,
Mu
,
W.
,
Serohijos
,
A.W.R.
,
Zhou
,
J.
and
Shakhnovich
,
E.I.
(
2013
)
Protein quality control acts on folding intermediates to shape the effects of mutations on organismal fitness
.
Mol. Cell
49
,
133
144
54
Ohmae
,
E.
,
Sasaki
,
Y.
and
Gekko
,
K.
(
2001
)
Effects of five-tryptophan mutations on structure, stability and function of Escherichia coli dihydrofolate reductase
.
J. Biochem.
130
,
439
447
55
Casarotto
,
M.G.
,
Basran
,
J.
,
Badii
,
R.
,
Sze
,
K.-H.
and
Roberts
,
G.C.
(
1999
)
Direct measurement of the pKa of aspartic acid 26 in Lactobacillus casei dihydrofolate reductase: implications for the catalytic mechanism
.
Biochemistry
38
,
8038
8044
56
Baker
,
D.J.
,
Beddell
,
C.R.
,
Champness
,
J.N.
,
Goodford
,
P.J.
,
Norrington
,
F.E.A.
,
Smith
,
D.R.
et al. 
(
1981
)
The binding of trimethoprim to bacterial dihydrofolate reductase
.
FEBS Lett.
126
,
49
52
57
Birdsall
,
B.
,
Roberts
,
G.C.K.
,
Feeney
,
J.
,
Dann
,
J.G.
and
Burgen
,
A.S.V.
(
1983
)
Trimethoprim binding to bacterial and mammalian dihydrofolate reductase: a comparison by proton and carbon-13 nuclear magnetic resonance
.
Biochemistry
22
,
5597
5604
58
Cammarata
,
M.
,
Thyer
,
R.
,
Lombardo
,
M.
,
Anderson
,
A.
,
Wright
,
D.
,
Ellington
,
A.
et al. 
(
2017
)
Characterization of trimethoprim resistant E. coli dihydrofolate reductase mutants by mass spectrometry and inhibition by propargyl-linked antifolates
.
Chem. Sci.
8
,
4062
4072
59
Bershtein
,
S.
,
Serohijos
,
A.W.R.
,
Bhattacharyya
,
S.
,
Manhart
,
M.
,
Choi
,
J.-M.
,
Mu
,
W.
et al. 
(
2015
)
Protein homeostasis imposes a barrier on functional integration of horizontally transferred genes in bacteria
.
PLoS Genet.
11
,
e1005612
60
Melnyk
,
A.H.
,
Wong
,
A.
and
Kassen
,
R.
(
2015
)
The fitness costs of antibiotic resistance mutations
.
Evol. Appl.
8
,
273
283
61
Sundqvist
,
M.
,
Geli
,
P.
,
Andersson
,
D.I.
,
Sjolund-Karlsson
,
M.
,
Runehagen
,
A.
,
Cars
,
H.
et al. 
(
2010
)
Little evidence for reversibility of trimethoprim resistance after a drastic reduction in trimethoprim use
.
J. Antimicrob. Chemother.
65
,
350
360
62
Taniguchi
,
Y.
,
Choi
,
P.J.
,
Li
,
G.-W.
,
Chen
,
H.
,
Babu
,
M.
,
Hearn
,
J.
et al. 
(
2010
)
Quantifying E. coli proteome and transcriptome with single-molecule sensitivity in single cells
.
Science
329
,
533
538
63
Lee
,
J.
,
Yennawar
,
N.H.
,
Gam
,
J.
and
Benkovic
,
S.J.
(
2010
)
Kinetic and structural characterization of dihydrofolate reductase from Streptococcus pneumoniae
.
Biochemistry
49
,
195
206
64
Gould
,
S.J.
and
Lewontin
,
R.C.
(
1979
)
The spandrels of San Marco and the Panglossian paradigm: a critique of the adaptationist programme
.
Proc. R. Soc. Lond. B Biol. Sci.
205
,
581
598

Supplementary data