Mind the gap: Understanding discordance between culture- and a non-culture-based measure of bacterial burden in murine tuberculosis treatment models

The standard pharmacodynamic marker in murine tuberculosis drug studies is colony-forming units (CFU). A faster PCR-based marker of bacterial burden is 16S rRNA. For unclear reasons, treatment reduces CFU more than 16S rRNA. We evaluated this CFU-16S gap and estimated the fraction potentially attributable to slow decay of 16S rRNA from dead Mycobacterium tuberculosis (Mtb) versus transition to a viable but not culturable on solid agar (VBNCSA) population. We quantified the CFU-16S gap during and following treatment in six BALB/c mouse studies and one in vitro study. Applying a two-population ordinary differential equation-based model of Mtb death and 16S rRNA decay, we estimated the fraction of the gap potentially attributable to dead Mtb. Using meta-regression, we estimated the association between CFU or 16S rRNA with relapse. For all regimens, CFU fell more than 16S rRNA, ranging from isoniazid-rifampin-pyrazinamide-ethambutol (CFU decreased 39-times more than 16S rRNA at week 4) to bedaquiline-pretomanid-moxifloxacin-pyrazinamide (CFU decreased >500,000-times more). The two-population model suggested that the fraction of the CFU-16S gap attributable to residual 16S rRNA from dead Mtb is modest and decreases over time. After treatment, 16S rRNA often fell while CFU rose. Four-week CFU change explained most variation in relapse (R{superscript 2}=0.90) while four-week 16S rRNA change did not (R{superscript 2}=0.24). CFU-16S gap is only partially explained by slow decay of residual 16S, suggesting development of a VBNCSA population. However, continued decrease in 16S rRNA after treatment cessation and its limited association with relapse suggests VBNCSA may be a transient rather than persistent state.