Role of Alanine Racemase Mutations in Mycobacterium tuberculosis d-Cycloserine Resistance

ABSTRACT A screening of more than 1,500 drug-resistant strains of Mycobacterium tuberculosis revealed evolutionary patterns characteristic of positive selection for three alanine racemase (Alr) mutations. We investigated these mutations using molecular modeling, in vitro MIC testing, as well as direct measurements of enzymatic activity, which demonstrated that these mutations likely confer resistance to d-cycloserine.

to each of the 96-wells of the MycoTB plate. All the plates were covered, sealed in plastic bags, and incubated at 37ᵒC during 10 days. After this period, the plates were checked for growth. The MIC was defined as the lowest concentration of antibiotic that inhibited bacterial growth completely. When growth in the growth control-well was insufficient, the plates were re-incubated and re-read at day 21.
Drug susceptibility testing (DST) for DCS was conducted using the BACTEC MGIT 960 system and Epicenter V5.80A/TB eXIST (Becton and Dickinson). DCS (Sigma Aldrich, St. Louis, MO, USA) was tested at 2, 4, 8, 16, 32, and 64 μg/ml. To obtain primary cultures for DST, the strains were grown in MGIT tubes until they reached 100-200 growth units (GU). For DST, the MGIT tubes were inoculated with 0.8 ml of SIRE supplement (Becton Dickinson), 0.1 ml of a solution of DCS to achieve the desired final concentration, and 0.5 ml of the suspension of the strain. For the preparation of the drug-free proportional control, the strain suspension was diluted 1:100 and 0.5 ml inoculated in the tube. To monitor the normal growth of the strain, an absolute control was prepared using 500 μl of the undiluted strain suspension as inoculum. The interpretation of the results was performed as previously described (1,2). Briefly, when growth was observed in the drug-containing tube (GU≥100) before the proportional growth control (1% inoculum) was positive (GU=400), this indicated that more than 1% of the population was growing in the presence of the drug and as per proportion testing, the strain was considered resistant at the corresponding drug concentration. The MIC was considered as the lowest concentration with GU <100 when the drug-free proportional control tube reached the positivity threshold of 400 GU.

WGS and bioinformatics analyses
Strains listed in Table S1 were classified using the nomenclature by Coll et al. and analyzed in the context of a reference collection that encompassed all seven M. tuberculosis and M. africanum lineages, as well as animal strains and M. canettii (3)(4)(5). WGS was performed using Illumina Technology (MiSeq and HiSeq 2500) with Nextera XT library preparation kits as instructed by the manufacturer (Illumina). By contrast, the NEXTflex DNA Sequencing Kit (Bioo Scientific) was used to prepare the Indian strains for sequencing on an Illumina MiSeq.
Fastq files/reads were mapped to the M. tuberculosis H37Rv genome (GenBank ID: NC_000962.3) with BWA (6). Alignments were refined with GATK and Samtools toolkits with regard to base quality re-calibration, alignment corrections for possible PCR and insertions and deletions indel artefact (7,8). Variants were extracted with customized perl scripts employing thresholds of a minimum coverage of 4 reads in both forward and reverse orientation, 4 reads calling the allele with at least a phred score of 20, and 75% allele frequency.
In the combined dataset, we allowed 5% of all samples to fail the aforementioned threshold criteria in individual genome positions to compensate for coverage fluctuations in certain genome regions. Overall, regions annotated as 'repetitive' elements (e.g. PPE and PE-PGRS gene families), resistance associated genes, indels, and consecutive variants in a 12 bp window (putative artefacts flanking indels were excluded. The remaining single nucleotide polymorphisms (SNPs) were considered as valid and used for concatenated sequence alignments.
The maximum likelihood tree was calculated with FastTree using the concatenated sequence alignment, a general time reversible (GTR) nucleotide substitution model, 1,000 resamples and Gamma20 likelihood optimization to account for evolutionary rate heterogeneity among sites. Alr mutations discussed in this manuscript were mapped on the corresponding branches/sub-groups (9).

Site-direct mutagenesis of Alr Mtb
A one-step site-directed mutagenesis protocol was employed to the generate M319T, Y364D and R373L Alr Mtb expression constructs (10). Mutagenesis PCR was performed using pMB2103 (a pET28 vector with a hexa-histidine tag at the N-terminal end and wild-type Alr Mtb (11)) as a template with the primers shown in Table S3. PCR mixes were prepared according to the manufacturer's recommendation (FINNZYMES) using Phusion TM DNA polymerase. The initial denaturation step was for 30 s at 95 °C, followed by 24 cycles of amplification (30 sec at 95 °C, 30 sec at 60 °C and 10 min at 72 °C). A final extension step was carried out at 72 °C for 15 min. 1 μL of the restriction enzyme DpnI was added to the PCR mix and incubated at room temperature overnight. Escherichia coli MC1061 cells were transformed with this mix and transformants were selected on LB agar plates containing kanamycin (50 μg/ml). The expression plasmids were prepared from an overnight 5 ml LB culture of transformants. The fidelity of all constructs was confirmed by Sanger sequencing. The constructs were introduced into E. coli BL21 (DE3) pLysS cells (Novagen) for expression of wild-type Alr Mtb and the three mutants.

Enzyme assay and determination of D-cycloserine (DCS) IC 50 for Alr Mtb
The Alr Mtb activity was measured using the D-alanine → L-alanine coupled reaction as described previously, except for the assay temperature that was changed from 23°C to 30°C. 6 Rates were measured on an Ultrospec 3100 pro with SWIFTII software and analyzed using GraphPad Prism version -3-6.00 for Windows (GraphPad Software, La Jolla California USA). The DCS inhibition assay was measured at a fixed substrate concentration of 2.5 mM D-alanine. The activity of wild-type Alr Mtb and single mutants were measured at various DCS concentrations ranging from 1.4 µM to 1.0 mM. The enzyme concentration used to determine IC 50 was varied due to the large reduction in the turnover rates of some of Alr mutants. 47 nM enzyme was used for the wild-type and Y364D mutant, and 470 and 235 nM were used for M319T and R373L mutants, respectively. Each reaction mix was pre-incubated with DCS for 30 minutes and the absorbance at 340 nm was measured for 2 minutes to set a baseline. The reaction was started by adding D-alanine to the assay mix and, following equilibration, the rates were measured for 10 minutes. The activity was normalised against a control with no DCS present in the assay mix. Activity assay at each concentration was performed in triplicate.

Structural analysis of Alr Mtb mutations
For the structure-based analysis of the Alr mutations, an Alr Mtb -DCS complex model was created by superposing the crystal structures of M. tuberculosis Alr Mtb (PDB: 1XFC) and the Geobacillus stearothermophilus Alr adduct with DCS (PDB: 1EPV) (12,13). PyMOL was used for inspecting the structure and preparing Figure S1 (14). The overall structure of Alr was highly conserved across bacterial species, including the core residues that form the active site (13,(15)(16)(17)(18)(19). Indeed 12 out of 13 active site residues were identical between two Alrs used in this model generation. Given a Cα root-mean-square deviation of 0.4 Å for these 13 residues, the two structures were superposed to place the PLP-DCS molecule into the Alr Mtb structure. Although W88 (Alr Mtb numbering) was replaced with leucine in the G. stearothermophilus Alr, this single mutation did not influence positioning of the co-factor PLP in both structures.

Ethics approvals
Estonia

Table S1
Overview of the alr mutants included in this study with the associated MIC data, where applicable. Details of the study population from Swaziland can be found elsewhere (20,21). H37Rv was tested as a negative control in all cases. Moreover, four wild-type Portuguese strains were used as controls for the Portuguese alr M319T mutants ( Figure  1). We also included the baseline isolate PBm0, which developed DCS resistance during treatment (PBm14). Isolates in bold were deemed resistant using the following criteria: MIC>CC=30 μg/ml for the 1% critical proportion method on LJ, MIC>tentative ECOFF=20 μg/ml for the 10% critical proportion method on LJ, and at least a two-fold MIC increase compared with the wild-type controls for MGIT and Sensititre.

Table S2
Overview of mutations in other known or potential DCS resistance genes in 22 isolates from Table S1, for which WGS data were available. The ald T-32C, cycA R93L and ddlA T365A mutations are most likely polymorphisms that do not confer resistance as these occurred in isolates with low as well as high DCS MICs (Table S1).   Figure S1. Alr Mtb -DCS complex model (A) Overall schematic view of the Alr Mtb -DCS complex structure model, including the PLP co-factor. The dimeric Alr Mtb is shown as a ribbon diagram with each monomer of the homodimer colored in light yellow and cyan, respectively. A truncated molecular surface was used to highlight one of the two active sites in the dimer. Some residues were removed for clarity. The left panel shows the entryway of the active site, whereas the view from the opposite side is shown in the right panel. (B) Zoomed-in view of the boxed areas from Panel A. The M319, Y364 and R373 side chains are shown in green. The PLP-DCS molecule is shown in lavender, blue and red. The active site surface model is also shown. Some residues were removed for clarity. (C) Same view as Figure S1B but showing mutated side chains for M319T, Y364D and R373L as purple balls and sticks.