Difference in persistent tuberculosis bacteria between in vitro and sputum from patients: implications for translational predictions

Abstract

This study aimed to investigate the number of persistent bacteria in sputum from tuberculosis patients compared to in vitro and to suggest a model-based approach for accounting for the potential difference. Sputum smear positive patients (n = 25) provided sputum samples prior to onset of chemotherapy. The number of cells detected by conventional agar colony forming unit (CFU) and most probable number (MPN) with Rpf supplementation were quantified. Persistent bacteria was assumed to be the difference between MPNrpf and CFU. The difference in persistent bacteria between in vitro and human sputum prior to chemotherapy was quantified using different model-based approaches. The persistent bacteria in sputum was 17% of the in vitro levels, suggesting a difference in phenotypic resistance, whereas no difference was found for multiplying bacterial subpopulations. Clinical trial simulations showed that the predicted time to 2 log fall in MPNrpf in a Phase 2a setting using in vitro pre-clinical efficacy information, would be almost 3 days longer if drug response was predicted ignoring the difference in phenotypic resistance. The discovered phenotypic differences between in vitro and humans prior to chemotherapy could have implications on translational efforts but can be accounted for using a model-based approach for translating in vitro to human drug response.

Combined quantitative tuberculosis biomarker model for time-to-positivity and colony forming unit to support tuberculosis drug development

“…In addition, the use of a correction factor ( ) that scaled down the ratio of the non-multiplying bacteria to the total bacteria that is transferred from the MTP model to the tube bacterial compartment was also evaluated ( . The parameter was both estimated and fixed to 17% ( Faraj et al, 2020a ). This is because it has been previously shown that the ratio of the non-multiplying sub-state to the total bacteria had to be scaled down to correctly predict the clinical bacterial number using an MTP model developed using in vitro data ( Faraj et al, 2020a ).…”

“…The parameter was both estimated and fixed to 17% ( Faraj et al, 2020a ). This is because it has been previously shown that the ratio of the non-multiplying sub-state to the total bacteria had to be scaled down to correctly predict the clinical bacterial number using an MTP model developed using in vitro data ( Faraj et al, 2020a ).…”

“…The MTP model was developed using CFU counts from natural growth data of Mtb in an in vitro hypoxia system together with the decline in CFU counts in response to treatment in log and stationary phase cultures. The MTP model was since successfully used to describe other in vitro systems ( Chen et al, 2018 ; Susanto et al, 2020 ), and was validated and used in in vivo settings ( Chen et al, 2017 ; Clewe et al, 2020 ), and in clinical settings ( Svensson and Simonsson, 2016 ; Faraj et al, 2020a ; Faraj et al, 2020b ) to predict the changes in the numbers of bacteria in the different sub-states and subsequently CFU with and without treatment. Likewise, time-to-event approaches for describing drug efficacy using TTP biomarker data in clinical trials have been developed ( Chigutsa et al, 2013 ; Svensson and Karlsson, 2017 ; Svensson et al, 2018b ).…”

Pharmacometrics in tuberculosis: progress and opportunities

“…It can be employed to account for immunity, and integrate the effects of multiple drugs [72] . The exploration of phenotypic differences in persistent TB bacteria between in-vitro systems and patient sputum [73] ; computational models of granuloma formation and function [74] ; models describing the spatio-temporal distribution of drugs in granulomas and cavitary TB lesions [75] , the multi-state TB pharmacometric (MTP) model, which provides predictions of the change in bacterial counts for fast-, slow-and nonmultiplying bacterial sub-populations, with and without drug effects [ 76 , 77 ]; and models linking the MTP with the General Pharmacodynamic Interaction (GPDI) model to account for the PD interactions between concomitantly administered drugs [78][79][80] , are examples of recently developed 'sub-systems' platforms contributing to ever-more complex models informing optimal drug combinations and doses [81][82][83][84][85][86] . Models describing drug distribution to cerebrospinal fluid (CSF) have proven useful in informing the treatment of TB meningitis, with recent work linking low rifampicin [87] and isoniazid [88] concentrations in CSF to negative outcomes.…”

Clinically encountered growth phenotypes of tuberculosis-causing bacilli and their in vitro study: A review

“…See Table 1 for a summary of in vitro DD Mtb work; Figure 3 highlights the different mechanisms influencing growth heterogeneity in Mtb. While promising, much work remains to determine how the in vitro studies relate to in vivo findings; discrepancies in, for example, antibiotic response kinetics between one in vitro model of persistence and clinical sputum have been noted ( Faraj et al., 2020 ).…”