Reply to Vargas and Farhat: Mycobacterium tuberculosis glpK mutants in human tuberculosis

We thank Vargas and Farhat (1) for their analyses confirming that reversible glpK frameshifts are common within human Mycobacterium tuberculosis infections. In their thoughtful comments they suggest that additional forces, other than antibiotic pressure, may contribute to the emergence of these mutants. We agree that the mechanisms by which glpK mutants are selected may be complex; however, we question whether their study can resolve if antibiotic pressure in human tuberculosis (TB) increases the fitness differential between wild-type and glpK frameshift M. tuberculosis. Vargas and Farhat (1) describe their experience sequencing the glpK gene in M. tuberculosis cultures isolated from the sputum samples of 200 longitudinally followed patients with TB. They found that glpK frameshifts occurred commonly between longitudinal pairs but that the prevalence of frameshifts was not associated with chronic infection or the duration of antibiotic exposure. However, as we demonstrate (2), small-colony glpK frameshift mutants revert rapidly to wild-type glpK sequences when cultured in standard M. tuberculosis media. It follows that glpK mutants which develop during TB infections may be lost when sputum is cultured in vitro and their disease contribution missed. We suggest that an analysis of glpK variant evolution would be best performed by direct deep sequencing from uncultured sputum samples. Furthermore, we question whether Vargas and Farhat (1) studied the appropriate patients with TB. The authors report that the mean duration between the paired longitudinal sputum cultures was 291 d for pairs without glpK mutations and 167 d, respectively, for pairs with glpK mutations. Thus, each of these groups continued to be culture positive for M. tuberculosis long after TB treatment should normally have converted virtually all of the patient’s sputum to culture negative. This suggests that many of the TB cases in this study either were not being treated or were multidrug resistant. In either case it is unlikely that the study patients were exposed to drugs that had significant effects against the infecting M. tuberculosis strains.

Vargas and Farhat (1) also refer to a mouse study from 2010 (3) which showed that an M. tuberculosis glpK loss-of-function mutant grew in BALB/c mice in a similar manner to the parental wild-type strain. This suggests that glpK, and by implication glycerol catabolism, does not play a role in mouse tuberculosis. However, we show that glpK mutants accumulate in mice infected with wild-type M. tuberculosis. This phenomenon complicates a strict comparison between glpK mutant and wild-type growth kinetics in vivo.

Finally, Vargas and Farhat (1) suggest assessing the potential use of GlpK inhibitors for patient treatment. We disagree that M. tuberculosis GlpK could be a good target for TB drug discovery. Our results suggest that GlpK inhibitors would phenocopy frameshift loss-of-function mutants, inducing M. tuberculosis into a drug-tolerant state. Instead, we propose (2) drug screens to identify compounds that are bactericidal against glpK deletion strains. Drugs that are able to kill drug-tolerant glpK mutants might contribute to rapid eradication of M. tuberculosis, enabling shorter-course TB treatment and improved treatment outcomes.