Milena M. Boczar 1, Raymonde Bamboukou Bekale 2, Paolo Denti 3, Samantha Sampson 2, Elin M. Svensson 1,4
1 Department of Pharmacy, Uppsala University (Uppsala, Sweden), 2 Division of Molecular Biology & Human Genetics, Stellenbosch University (Stellenbosch, South Africa), 3 Division of Clinical Pharmacology, Department of Medicine, University of Cape Town (Cape Town, South Africa), 4 Department of Pharmacy, Pharmacology and Toxicology, Radboud University Medical Center (Nijmegen, The Netherlands)
Objectives
Mycobacterium tuberculosis (Mtb) can exist in different metabolic states, ranging from fast-growing populations to viable but non- or slowly replicating (VBNR) bacteria. These subpopulations have varying sensitivity to antimicrobial drugs, and resting bacteria could contribute to relapse after completion of treatment. Current methods for differentiating between subpopulations are expensive, labour-intensive, and require specialized infrastructure. If similar information could be obtained through interpretation and analysis of routinely used metrics like colony-forming units (CFU) and time-to-positivity (TTP) from the BD BACTEC mycobacteria growth indicator tubes (MGIT) system, where continuously recorded growth units (GU) are currently underutilized, that would be of great benefit. While CFU represents a static measurement at one time point, MGIT provides temporal readout of bacterial growth. GU inspection allows distinction between initial time-to-growth (TTG) phase, during which no measurable increase is observed, and subsequent growth phase.
The aim of this pilot study was to test whether the effect of acid stress, which is expected to induce VBNR Mtb [1,2], can be detected by analysing GU data from MGIT together with CFU.
Methods
Mtb H37Rv::pTiGc [3,4] was used for the experiments. Because the stain was auxotrophic [4], all media were supplemented with leucine and pantothenate. Starting cultures were grown in 7H9 broth supplemented with kanamycin, incubated at 37°C with 5% CO₂ until optical density 600 reached 1. Cultures were subcultured for 7 days. Bacteria were transferred to media adjusted to pH 4.5 or maintained at pH 6.5, and incubated for 2 days. Bacterial suspensions were prepared as seven 10-fold dilutions (10¹–10⁷), inoculated into MGIT tubes (500 µL/tube) and solid agar plates (100 µL/plate) for CFU enumeration. MGIT tubes with PANTA enrichment were monitored in the BD BACTEC MGIT system to record TTP. Using Epicenter software, continuous GU were recorded from inoculation onward. CFU were counted after 21 days of incubation. Media-only controls were included throughout the experiment. Experiments were performed in duplicate on two occasions.
Two previously developed models [5] were used as the starting point for analysis. The TTG-CFU relationship was analysed using linear mixed-effects modelling in NONMEM (7.5.1) and Perl-speaks-NONMEM (5.3.1). TTG (natural log (ln) scale) was modelled as dependent variable, CFU was log10-transformed as independent variable. The linear model estimated intercept and slope. Between-biological replicate variability on both parameters and effect of pH on intercept were evaluated. Residual variability was described using additive error model (ln scale).
The relationship between GU and incubation time in MGIT was analysed using nonlinear mixed-effects modelling (same software). TTG was subtracted from incubation time to isolate bacterial growth phase of the experiment. GU (log10-transformed) was described using a logistic function characterized by carrying capacity (asymptote), growth rate (slope), and time to half of carrying capacity (T50). Between-MGIT variability on all three parameters and the covariates pH and dilution on slope and asymptote were tested. Residual variability was described using additive error model (log10 scale).
Results
The linear model described the TTG-CFU relationship well and included 25 experiments. The estimated intercept was 578 hours [relative standard error (RSE): 0.8%], slope –0.3 hours (ln) per log10 CFU [3.7%], and additive residual error of 8.3% [12.4%]. pH had non-significant effect.
The relationship between GU and incubation time was described well with the logistic function. In total, 6,188 observations from 27 experiments were included. The parameter estimate for asymptote was 22,500 GU [15.2%], slope 0.1 GU/hour [4%], 75 hours [3.8%] for T50 and additive residual error of 12.3% [6.5%]. Dilutions had significant effect (ΔOFV=-60.9) on the asymptote. Compared with dilution 10¹, the asymptote increased across dilutions 10²–10⁵ and decreased at 10⁷. Between-MGIT variability was 16.5% [22.7%], 20.3% [22.1%], 19.5% [25.9%] on asymptote, slope and T50, respectively.
Conclusions
The relationship between GU and incubation time in MGIT was not affected by pH, which is expected. Although dilution effects on the asymptote were anticipated, the observed pattern was not as expected. One possible explanation is that the lowest dilutions did not reach the plateau phase within the incubation period and require further investigation. Acidic stress did not significantly alter the TTG–CFU relationship. Given prior evidence that low pH favours formation of VBNR Mtb, the absence of an observed effect likely reflects experimental constraints rather than lack of biological impact, and additional study is needed.
References:
[1] Early JV, Mullen S, Parish T. A rapid, low pH, nutrient stress, assay to determine the bactericidal activity of compounds against non-replicating Mycobacterium tuberculosis. PLoS ONE. 2019 Oct 7;14(10):e0222970. doi:10.1371/journal.pone.0222970 PubMed PMID: 31589621; PubMed Central PMCID: PMC6779252.
[2] Gold B, Warrier T, Nathan C. A Multi-stress Model for High Throughput Screening Against Non-replicating Mycobacterium tuberculosis. In: Parish T, Roberts DM, editors. Mycobacteria Protocols [Internet]. New York, NY: Springer; 2015 [cited 2026 Feb 25]. p. 293–315. Available from: https://doi.org/10.1007/978-1-4939-2450-9_18 doi:10.1007/978-1-4939-2450-9_18
[3] Mouton JM, Helaine S, Holden DW, Sampson SL. Elucidating population-wide mycobacterial replication dynamics at the single-cell level. Microbiology. 2016 Jun;162(6):966–78. doi:10.1099/mic.0.000288 PubMed PMID: 27027532; PubMed Central PMCID: PMC5042079.
[4] Mouton JM, Heunis T, Dippenaar A, Gallant JL, Kleynhans L, Sampson SL. Comprehensive Characterization of the Attenuated Double Auxotroph Mycobacterium tuberculosisΔleuDΔpanCD as an Alternative to H37Rv. Front Microbiol. 2019 Aug 20;10:1922. doi:10.3389/fmicb.2019.01922 PubMed PMID: 31481950; PubMed Central PMCID: PMC6710366.
[5] Maitra A, Wijk M, Margaryan H, Denti P, McHugh TD, Kloprogge F. The impact of physiological state and environmental stress on bacterial load estimation methodologies for Mycobacterium tuberculosis. Sci Rep. 2024 Oct 30;14(1):26108. doi:10.1038/s41598-024-74318-3
Funding
The project was funded by the Swedish Heart-Lung Foundation (Hjärt-Lungfonden; E.M.S., project no. 20230736).
Reference: PAGE 34 (2026) Abstr 12163 [www.page-meeting.org/?abstract=12163]
Poster: Drug/Disease Modelling - Infection