Fitness Effect of the Isoniazid Resistance Mutation S315T of the Catalase-Peroxidase Enzyme KatG of Mycobacterium tuberculosis

Abstract

The mutation S315T of the catalase-peroxidase (CP) protein KatG of Mycobacterium tuberculosis is the most common mutation that confers resistance to the prodrug isoniazid. Here, we reconstruct its evolutionary history in 145 whole-genome sequences of M. tuberculosis from Russian hospitals, inferring 11 independent appearances of this mutation and 5 reversion events, with an estimated reversion rate 1,500 times higher than the rate of preserved nonsynonymous or intragenic mutations. This suggests that, contrary to the commonly held view, the mutation KatG(S315T) results in a fitness cost, possibly because of reduced tolerance to oxidative stress. Consistent with this interpretation, the mutant enzyme presents reduced CP activities. Applying the torsional network model (TNM), we found that the mutant protein shows more restricted thermal dynamics, although its functional site moves quite similarly to the wild type. Of the four internal clones where KatG(S315T) arose, two present high reproductive rates and secondary mutations at the 5'-UTR region of the gene encoding superoxide dismutase A (sodA), while the other two present significantly lower reproductive rates and lack mutations at genes related with tolerance to oxidative stress. Our results suggest that the resistance mutation KatG(S315T) incurs a fitness cost, which may be alleviated through compensatory mutations at the gene sodA or other genes that respond to oxidative stress, such as the previously known gene ahpC. This suggests that isoniazid treatment could be complemented with drugs that produce oxidative stress in order to hinder the propagation of resistant strains devoid of compensatory mutations.

Keywords: Mycobacterium tuberculosis; S315T; antimicrobial resistance; isoniazid; oxidative stress; regularized maximum likelihood and minimum evolution (REGMLAME).

Fig. 1.

Relationship between normalized log-likelihood score and mean branch length for the sixteen inferred three topologies and branch lengths fitted with different substitution models. First column: amino-acid models JTT and STMTREV. They provide higher likelihood and shorter branch lengths to all considered topologies than the nucleotide-based models GTR and TVM (fourth column). Combined models (amino-acids and DNA) are shown in the second and third column and they provide intermediate branch lengths and log-likelihood. Note that all models show strong negative correlation between the log-likelihood and branch lengths.

Fig. 2.

For each of the 16 tested tree topologies, we show the mean across all models of the REGMLAME score (a) and the BIC score (b). The error bars represent the SEM. We rank the trees according to the ratio between the mean and SEM of the REGMLAME score.

Fig. 3.

Inferred phylogenetic tree and reconstructed ancestral sequences. For each ancestral sequence at internal nodes, we indicate whether the residue at site KatG(315) is S or T and the number of secondary mutations of genes related to oxidative stress. We denote the events that produced the first occurrence of the resistance mutations or its reversion to S315 as ST (red) and TS (green), respectively. For each extant sequence (leaf), we indicate on the right the identity of the residue KatG(315), the number of secondary mutations, the lineage (either two or four for all of the sequenced strains) and the resistance (r) or susceptibility (s) to the four drugs Streptomycin, INH, Rifampicin, and Ethionamide. The red color indicates multi-resistant strains, green means susceptible, and the other colors indicate conditionally susceptible strains. The internal node labels indicate the amino acid at position 315 (either S or T) and the number of secondary mutations at genes involved with oxidative stress (0, 1, or 2). Since the label is difficult to read, we differentiate them based on color: dark green (S_sec0), blue (S_sec1), light green (T_sec0), red (T_sec1), and purple (T_sec2).

Fig. 4.

Relative rates of various types of preserved mutations inferred by adopting six different topologies of the phylogenetic tree. From left to right, the shown mutations are: (i) the resistance mutation KatG(S315T); (ii) its reversion KatG(T315S); (iii) other amino acid mutations known to confer antimicrobial resistance; (iv) nonsynonymous mutations at other genes; (v) synonymous mutations and mutations at the 5′-UTR regulatory region of genes known to be related with resistance to the oxidative stress, which we denote as secondary mutations; (vi) synonymous mutations at other genes; and (vii) mutations at integenic regions of other genes, which we adopt for normalizing the rates.

Fig. 5.

Predicted atomic fluctuations (a) and torsion angle fluctuations b) of the wild-type protein KatG (black) and the S315T mutant (soft lines).

Fig. 6.

Predicted dynamical couplings within (a, c, e) and between (b, d, f) functional sites of the wild-type protein KatG (black) and the S315T mutant (soft lines).

Fig. 7.

Mean and standard error of the mean of the reproductive rate of ancestral sequences of the mutant KatG(S315T) that did not present (left) and did present (right) mutations at the 5′-UTR region of the superoxide dismutase gene sodA.