Skip to main content
NCBI home page
Antimicrobial Agents and Chemotherapy logoLink to Antimicrobial Agents and Chemotherapy
. 2022 Jul 14;66(8):e00366-22. doi: 10.1128/aac.00366-22

Quantitative Systems Pharmacology Modeling Framework of Autophagy in Tuberculosis: Application to Adjunctive Metformin Host-Directed Therapy

Krina Mehta a, Tingjie Guo a, Robert S Wallis b, Piet H van der Graaf a,c, J G Coen van Hasselt a,✉
PMCID: PMC9380544  PMID: 35862740

ABSTRACT

Quantitative systems pharmacology (QSP) modeling of the host immune response against Mycobacterium tuberculosis can inform the rational design of host-directed therapies (HDTs). We aimed to develop a QSP framework to evaluate the effects of metformin-associated autophagy induction in combination with antibiotics. A QSP framework for autophagy was developed by extending a model for host immune response to include adenosine monophosphate-activated protein kinase (AMPK)-mTOR-autophagy signaling. This model was combined with pharmacokinetic-pharmacodynamic models for metformin and antibiotics against M. tuberculosis. We compared the model predictions to mice infection experiments and derived predictions for the pathogen- and host-associated dynamics in humans treated with metformin in combination with antibiotics. The model adequately captured the observed bacterial load dynamics in mice M. tuberculosis infection models treated with metformin. Simulations for adjunctive metformin therapy in newly diagnosed patients suggested a limited yet dose-dependent effect of metformin on reduction of the intracellular bacterial load when the overall bacterial load is low, late during antibiotic treatment. We present the first QSP framework for HDTs against M. tuberculosis, linking cellular-level autophagy effects to disease progression and adjunctive HDT treatment response. This framework may be extended to guide the design of HDTs against M. tuberculosis.

KEYWORDS: tuberculosis, autophagy, quantitative systems pharmacology, host-directed therapy, mathematical modeling

INTRODUCTION

The increasing burden of Mycobacterium tuberculosis infections is a major global health concern associated with approximately 1.5 to 2 million deaths annually (1). Current first-line treatment for active tuberculosis (TB) disease includes a 2-month intensive phase with rifampicin, isoniazid, pyrazinamide, and ethambutol, followed by a 4-month continuation phase with rifampicin and isoniazid (HRZE). A long treatment duration, common treatment failure, relapse, and the emergence of multidrug-resistant M. tuberculosis strains are key challenges to successful TB treatment (2).

Host-directed therapies (HDTs) aim to exploit the interplay between the pathogen and the host immune response (3, 4). HDTs are increasingly being studied for treatment against M. tuberculosis infections. One of the most studied HDT strategies to date is autophagy induction (5). Autophagy is an intracellular catabolic process involving the delivery of excessive or damaged cellular components, including bacteria, to the lysosome for degradation to maintain homeostasis. The adenosine monophosphate-activated protein kinase (AMPK)-mTOR signaling pathway is an important regulator of autophagy. M. tuberculosis activates phosphorylation of Akt, which stimulates phosphorylation of mTOR complex 1 (mTORC1). Activation of mTORC1 inhibits autophagy by phosphorylation of various autophagy-related proteins (6). Preclinical studies have demonstrated the involvement of the mTOR signaling pathway in the host response to M. tuberculosis, suggesting its relevance as a therapeutic target (6, 7). Therefore, metformin, an antihyperglycemic agent and mTORC1 inhibitor, has been proposed as a potential HDT against M. tuberculosis (6, 7). Metformin adjunctive therapy in diabetic TB patients was found to be associated with an improved therapy success rate and lowered mortality (8, 9).

Quantitative systems pharmacology (QSP) models aim to capture mechanistic details of the interactions between a biological system and the pharmacokinetic-pharmacodynamic (PKPD) properties of a drug (10). QSP-based characterization of drug-host-pathogen interactions may allow evaluation of the expected treatment responses upon the perturbation of specific targets, which may help to identify promising HDT targets and evaluate different potential combination treatment strategies. Within the TB field, mathematical modeling approaches have primarily focused on PKPD modeling and specifically on the design of antibiotics treatment strategies (11, 12). In addition, multiscale systems biology models of the host immune response to M. tuberculosis infection have been developed (13, 14). The prior immune response model (13) has been combined with PKPD models of HRZE to explore the impact of the patient immune response on treatment outcomes (15–17). None of these models included HDT-relevant pathways; however, they did establish a strong basis for further development of a QSP framework to enable the design of HDTs.

To guide the design and development of HDTs, relevant HDT pathways must be added to the QSP framework. The autophagy-regulating AMPK/mTOR pathway represents an important factor for HDTs. There are currently no mathematical models available in the literature describing mTOR signaling-mediated autophagy in TB. The objectives of this work were (i) to develop a QSP framework of the host immune response, including autophagy-mediated interactions, and (ii) to evaluate the effects of metformin-associated autophagy induction in combination with HRZE treatment.

RESULTS

The QSP framework included combined a host-pathogen interaction model, an AMPK-mTORC1 signaling pathway model, including autophagy, and PKPD models of HRZE and metformin (Fig. 1).

FIG 1.

FIG 1

Open in a new tab

Combined TB-autophagy QSP framework. The model captures the dynamics of host immune response in the lungs resulting from M. tuberculosis infection. The model consists of various species of macrophages, lymphocytes, and the key cytokines involved in both the innate and adaptive immune response against M. tuberculosis. The model includes the growth of M. tuberculosis as well as the immune-mediated elimination of M. tuberculosis affecting the overall M. tuberculosis population. The immune-mediated bacterial kill rates include mainly cytokine- and lymphocyte-mediated apoptosis, as well as autophagy. The model also consists of M. tuberculosis evasion mechanisms, such as induction of the AMPK-mTOR pathway and the inhibition of autophagy. Bi, intracellular M. tuberculosis; Be, extracellular M. tuberculosis; Ma, activated macrophage; Mi, infected macrophage; Mr, resident macrophage; T80, precursor-activated CD8+ T cells; T8, subclass (IFN-γ producing) of activated CD8+ T cells; Tc, subclass (cytotoxic lymphocytes) of activated CD8+ T cells; Th0, naive T cells; Th1, type 1 helper T cells; Th2, type 2 helper T cells.

The QSP framework simulations recapitulate the observed in vivo response to metformin.

The model was evaluated by comparing predictions to the observed data. The model predictions for total bacterial load showed good agreement with observed digitized lung bacterial load data from untreated mice infected with M. tuberculosis (6) (Fig. 2A). Simulations using standard TB therapy starting at day 180 postinfection in TB patients predicted the previously reported change in bacterial load from baseline with HRZE treatment reasonably well (Fig. 2B). Overall, these assessments suggested the reliability of the model for the objectives of this analysis.

FIG 2.

FIG 2

Open in a new tab

Time course of predicted and observed lung bacterial loads in (A) M. tuberculosis-infected mice treated or not treated with metformin and (B) tuberculosis patients treated with a standard antibiotic regimen. The model predictions for total bacterial load agree well with the observed data for mice treated with or not treated with metformin. Additionally, model predictions for the effects of standard antibiotic therapy agree reasonably well with the observed change in bacterial load from baseline data from TB patients. Metformin was administered to mice daily from days 7 through 35 using a regimen of 6 days on/1 day off. The dots represent the observed values, and the lines represent the model predictions. Mtb, M. tuberculosis.

Sensitivity analysis provides insights into the mechanistic details of infection.

A global uncertainty and sensitivity analysis suggested that the bacterial load was more sensitive to the parameters of the host-pathogen interaction model than to those of the autophagy model (see section 3 of the supplementary materials at reference 37). In general, the host-pathogen interaction model parameters that correlated most closely with bacterial load were related to macrophage recruitment, macrophage activation or deactivation, phagocytosis, gamma interferon (IFN-γ) production, or interleukin 1b (IL-1b)- or FAS-FAS-mediated apoptosis. Most of these parameters were also identified in the sensitivity analysis in the prior models (18). In the prior models, these parameters were obtained either from literature or were estimated using in vitro or mouse experimental data and therefore are considered relatively reliable. One parameter related to the autophagy model, the AKT dephosphorylation rate, was found to correlate positively with bacterial load and thus negatively with infection control. This highlights the key role of M. tuberculosis evasion and inhibition of autophagy on disease progression. This parameter was unchanged in the current model from the previous AMPK-mTOR signaling model. In the previous work, this parameter was estimated using experimental data from immunoblots and thus deemed reliable. The sensitivity analysis, given the uncertainty in the parameters, provides a thorough picture of the current state of the model (see Fig. 3 of reference 37).

Metformin-associated autophagy induction is predicted to provide dose-dependent reduction in intracellular bacterial load.

The simulations for TB disease progression (i.e., conducted prior to the start of treatment) suggested that M. tuberculosis infection is predicted to reduce autophagy by 55% in a typical subject. We compared the effects of HRZE with or without adjunctive metformin treatment on bacterial load and cytokine levels in a typical virtual TB patient. These simulations considered a typical scenario where the treatment was started upon diagnosis of TB, which was considered to be around day 180 after the initial infection. Adjunctive metformin with HRZE treatment was predicted to show a limited yet apparent dose-dependent increase in autophagy-mediated intracellular and total bacterial elimination when the total bacterial burden was relatively low after the first 2 months of treatment (Fig. 3A). No significant impact was predicted on cytokine levels in the adjunctive metformin with HRZE scenario compared to the HRZE-only scenario (Fig. 3B). The simulations for a scenario where metformin was added 2 months after the start of the intensive phase of treatment with HRZE predicted that adjunctive metformin might reduce the overall treatment duration by 3 to 5 days (see Fig. 4 of reference 37). Overall, we conclude that adjunctive metformin treatment may provide a modest benefit in reducing the M. tuberculosis bacterial load in TB patients during the continuation phase of HRZE treatment.

FIG 3.

FIG 3

Open in a new tab

Typical patient simulations of adjunctive metformin with standard antibiotic treatment using various dosing regimens starting at day 180 postinfection: (A) bacterial load and (B) cytokines. The simulations suggest dose-dependent effects of metformin on reduction of the intracellular bacterial load. The reduction in intracellular bacterial elimination with adjunctive metformin treatment, however, does not significantly affect the extracellular and thus total bacterial load compared to standard antibiotics, except when the total bacterial burden is high. HRZE refers to a regimen of 2 months of rifampicin, isoniazid, pyrazinamide, and ethambutol, followed by 4 months of rifampicin and isoniazid; the vertical dashed line represents the end of the 2-month regimen.

Metformin may delay disease progression in diabetic TB patients.

We assessed whether metformin would delay TB disease progression if it were administered prior to TB diagnosis, i.e., in scenarios where diabetic patients would be receiving metformin for glycemic control at the time of infection with M. tuberculosis (Fig. 4A). For these simulations, metformin input was added at the same time as the initial bacterial infection. We found that metformin use in diabetic TB patients would delay TB disease progression as assessed by intracellular, extracellular, and total bacterial load. Lower levels of proinflammatory cytokines, IL-1b, tumor necrosis factor alpha (TNF-α), IFN-γ, and IL-12, were also predicted in the typical patient treated with metformin versus not treated with metformin (Fig. 4B). Overall, these simulations suggest protective effects of metformin use toward tissue damage in diabetic TB patients. As these simulations represent scenarios prior to TB diagnosis, they do not provide guidance about metformin treatment for TB patients. However, these simulations provide mechanistic insights into the role of autophagy on the dynamics of TB infection.

FIG 4.

FIG 4

Open in a new tab

Typical patient simulations with or without 500 mg twice daily metformin starting at day 1 postinfection: (A) bacterial load and (B) cytokines. The model predicted some benefits of metformin use in delaying the disease progression in virtual diabetic patients receiving metformin compared to that in nontreated patients.

DISCUSSION

Here, we developed the first QSP framework for the design and evaluation of HDTs focusing on autophagy. Our model was able to recapitulate results from an in vivo study evaluating metformin as an HDT in M. tuberculosis-infected mice. We applied the framework to predict the treatment effects of metformin on autophagy induction in a typical TB patient.

Our analysis identified a modest beneficial effect of adjunctive metformin treatment in a typical TB patient, after the intensive phase of antibiotics treatment, when the total bacillary load is predicted to be relatively low. The predictions suggested that the overall effects of treatment with metformin would depend on the extracellular-to-intracellular bacterium ratio, which may depend on the stage of infection. The model also predicted some benefits of metformin use in delaying the disease progression in virtual diabetic patients receiving metformin. Our results agree with the clinical reports where lowered mortality rates were reported in diabetic patients receiving metformin (8, 9). A key M. tuberculosis survival strategy depends on provoking a nonsterilizing immune response, allowing M. tuberculosis to replicate beyond the reach of most immune mechanisms. As part of the host-pathogen interactions, granuloma formation limits M. tuberculosis growth but also provide a niche for replication by disseminating M. tuberculosis to other areas (19). Metformin, and HDTs in general, may provide beneficiary effects early after the initial infection, e.g., in newly infected TB household contacts, or late during treatment, i.e., after the sputum has been sterilized but when small numbers of persisting bacteria are still present. In these scenarios, small changes in the survival of a rather small bacterial population may have a large effect on the infection outcome, and future studies may consider evaluating this.

Our model provides relevant quantitative insight into the mechanistic details of factors contributing to autophagy-mediated bacterial elimination. The lack of predicted effects of metformin at doses up to 1,000 mg BID (twice a day) on total bacterial load can also be attributed to its potency on AMPK-mTORC1-autophagy signaling and its distribution in the lungs, in addition to the extracellular-to-intracellular bacterial ratio. Previously, a metformin dose-ranging study that evaluated the effects of metformin at doses of 100 to 10,000 μM on M. tuberculosis survival in human monocyte-derived macrophages showed no increased M. tuberculosis survival at doses up to 500 μM. In the same study, an approximately 4% reduction in total bacterial load on day 35 was noted in mice treated with 250 mg/kg or 500 mg/kg metformin daily from days 7 to 35 (6 days on, 1 day off) (6). Our minimal physiologically based pharmacokinetic (mPBPK) model predicted mouse lung Cmax (maximum concentration of drug within lungs) values of 668 μM and 1,336 μM in the 250 mg/kg and 500 mg/kg metformin dose groups, respectively. When these body weight-based doses of metformin that were evaluated in mice are compared to clinically feasible doses, the predicted lung Cmax values in humans are approximately 10- to 15-fold lower than those predicted in mice. As such, it would be no surprise that our predictions showed no significant effects of metformin on the reduction of bacterial load. In fact, a recently completed clinical trial evaluating adjunctive metformin treatment to standard treatment in TB patients reported that metformin treatment did not significantly reduce the time to sputum conversion compared to controls (20).

The integrated QSP framework connects a complex intracellular process, autophagy, to the disease outcome at the organism level. Our model can be easily adapted in the future to perform evaluations of other mTORC1 inhibitors and mTORC1-independent autophagy inducers. Some candidate drugs include everolimus, statins, PI3K (phosphatidylinositol 3-kinase) inhibitors, and tyrosine kinase inhibitors. The model can also facilitate in silico evaluations of the perturbations of various proteins involved in autophagy and predict their effects on the outcome, as such enabling target identification for optimal autophagy induction. For example, the model may be used in combination with screening assays to prioritize further development of potential HDTs.

One of the limitations of our model was that it built upon a prior relatively simple TB host immune response model (18). In our model, we added an empirical transition from fast to slow M. tuberculosis growth phases to resemble the initial log-phase increase in bacterial load. However, growth and treatment effects on different subpopulations of M. tuberculosis, i.e., nonpersisters and persisters, at a given time are not included in the model. Future work may integrate a mechanistic model with various subpopulations of both intra- and extracellular M. tuberculosis into the QSP framework. For this, measurements of various M. tuberculosis subpopulations in the sputum or bronchoalveolar lavage fluid of TB patients are required. The proposed integration may also require applying antibiotic bacterial kill rates specific to the subpopulation. In general, the current construct of the TB host immune response model is relevant for our primary objective, i.e., evaluating different treatment scenarios with and without metformin in TB patients when the bacterial load is already relatively high.

Our model includes the relative activity of the key proteins involved in AMPK-mTOR-autophagy signaling; however, it does not consider total concentrations of these proteins. The original data-driven AMPK-mTOR model that was adapted in this work was developed using immunoblot data from HeLa cells and therefore considered the relative activity of the proteins (21). Direct measurements of total protein concentrations are not available to date. As such, uncertainty exists in the parameters, impacting the effects of AMPK-mTOR signaling on autophagy and treatment-effect predictions. However, the current approach of using the relative activities of AMPK-mTOR signaling proteins to evaluate their downstream effects on autophagy provides a useful alternative in the absence of absolute protein data.

Conclusion.

To summarize, we developed a QSP framework for autophagy-inducing HDT by integrating previously developed models for AMPK-mTOR signaling, host-pathogen interactions, and PKPD. We extended the framework to include autophagy to enable in silico evaluations of adjunctive metformin to antibiotics in TB patients. Our predictions suggest that metformin may provide some beneficiary effects when the overall bacterial load, or extracellular-to-intracellular bacterial ratio, is low. Overall, this is the first QSP framework that links cellular-level events affecting autophagy to disease progression, and it may further be developed to guide HDT design and development for the treatment of TB.

MATERIALS AND METHODS

The QSP framework developed (Fig. 1) included (i) PK models for standard antibiotics and metformin, (ii) a TB host immune response model including the PD effects of HRZE, and (iii) an autophagy model including the PD effects of metformin. The QSP model development was facilitated by the adaptation of various models presented in literature (17, 18, 21). A set of ordinary differential equations (ODEs) describing the dynamics of intra- and extracellular bacteria in host lungs as functions of time, and the dynamics of immune response components, such as macrophages, cytokines, and lymphocytes, as functions of time and bacterial load, form the core of our QSP framework (18). The core model was then linked to a model describing the dynamics of AMPK-mTOR signaling proteins leading to autophagy (21). The interactions between M. tuberculosis and autophagy connect these two models. Moreover, the combined TB host immune response-autophagy model was linked with models capturing the PKPD relationships of HRZE and metformin.

Model development.

The details of the model development process are provided (see section 1 of reference 37), and the key steps are presented below.

Pharmacokinetics.

PK models of four antibiotics, HRZE, were reproduced from literature-based population PK models (22, 23). Plasma concentrations of HRZE following standard-of-care dosing were simulated using the PK models. HRZE intra- and extracellular lung concentrations were predicted by applying plasma-to-lung alveolar cell and plasma-to-lung epithelial lining fluid ratios obtained from literature (24). To predict lung concentrations of metformin, we developed a minimal physiologically based PK model for metformin including a lung compartment (25) (see section 1.1 of reference 37).

TB host immune response and pharmacodynamics of standard antibiotics.

A published model that captured the dynamics of host-pathogen interactions following M. tuberculosis infection was implemented (18). This host immune response model contained host-pathogen interactions in lungs and included three populations of macrophages (resting, activated, and infected), various cytokines (IFN-γ, TNF-α, IL-10, IL-4, and IL-12) and lymphocytes, as well as intra- and extracellular M. tuberculosis populations. An update was made to this model to add the turnover of IL-1b and IL-1b-mediated bacterial elimination (26) (see section 1.2 of reference 37).

We included two M. tuberculosis growth phases, fast and slow, as a simple implementation of the initial rapid progression of active disease (13, 18). The switch from fast to slow growth phases was empirically set to 21 days postinfection based on mouse infection experimental results (27, 28). We used the slow bacterial growth rate estimates from the reproduced TB host immune response model (18). The growth rates for the initial fast growth phase were optimized using digitized data from a mouse M. tuberculosis infection experiment (27) (see also section 1.2.1 of reference 37). Bactericidal effects on the intra- and extracellular bacterial population and bacteriostatic effects on the growth rates of bacteria driven by intra- and extracellular lung concentrations of HRZE were reproduced from literature (17).

Autophagy and the pharmacodynamics of metformin.

The AMPK-mTOR cell signaling network model from reference 21 was reproduced. This model captures the dynamics of key proteins involved in the AMPK-mTOR signaling pathway and includes relative interactions between proteins involved in the AMPK-mTOR signaling pathway, such as insulin receptor substrate, class I phosphatidylinositol 3-kinases, AMPK, mTORC1, and mTOR complex 2 (mTORC2). This model was updated to include various M. tuberculosis- and autophagy-related components. The updates can be categorized into (i) the effect of M. tuberculosis infection on autophagy inhibition due to the activation of AMPK-mTOR signaling and (ii) the effect of autophagy on M. tuberculosis elimination. The gene AKT3, a key upstream regulator of the AMPK-mTOR signaling pathway, was found to be induced 1.38-fold in M. tuberculosis-infected versus -uninfected mice based on differential expression in the lungs (6). This ratio was added as a proportional scaling factor in the model of production of AKT to simulate the presence of M. tuberculosis and its impact on key downstream proteins involved in AMPK-mTOR signaling, including mTORC1 (see section 1.3 of reference 37). Due to limited data availability, time course effects of the progression of M. tuberculosis infection on autophagy are not included in the current model.

The effects of AMPK-mTOR signaling on autophagy were modeled using a direct-effect saturable Emax (maximum effect) model. Autophagy at the time of M. tuberculosis infection was set to 100% to represent a healthy state prior to infection. Then, the percent inhibition of autophagy due to M. tuberculosis infection and the subsequent AMPK-mTOR signaling activation were modeled. Next, the autophagy model was combined with the TB host immune response model by introducing autophagy-mediated intracellular bacterial kill rates and autophagy-mediated extracellular to intracellular bacterial uptake. These processes were incorporated as first-order processes, and the parameters were informed by M. tuberculosis survival data from in vitro infection experiments with and without metformin treatment (6) (see also section 1.3 of reference reference 37). The inhibitory effect of metformin on mTORC1 phosphorylation was incorporated using an indirect effect saturable Emax model, and the parameters were obtained from the literature (29, 30).

Model evaluations.

The combined TB-autophagy QSP framework predictions were first compared to observed digitized lung bacterial load data from untreated and metformin-treated mice infected with M. tuberculosis (6). To this end, the QSP model was scaled from humans to mice by applying volume differences between the species. To evaluate HRZE PKPD components of the combined QSP model, the predicted change in bacterial load over time after start of HRZE treatment was compared to reported values for TB patients (31–33).

Sensitivity analysis.

High uncertainty existed in some parameters, especially those related to the autophagy model, due to limited data availability. To further understand the impact of uncertainty in the parameters on the model predictions, a global uncertainty and sensitivity analysis using Latin hypercube sampling (LHS) and the partial rank correlation coefficient method was performed using 500 samples (34, 35). The outcome used in this analysis was the predicted total bacterial load. All parameters, except those related to PK and PD, were evaluated in the global uncertainty and sensitivity analysis. The parameter ranges used for the LHS were the same as those used in the previous model for TB host immune response model components and were varied by 20% for autophagy-related components (18).

Simulations of metformin-associated autophagy induction in humans.

Typical TB patient simulations were conducted using the QSP framework to predict the effects of autophagy induction with metformin on the overall treatment outcome. Typical TB patient simulations were performed using the parameter values presented in section 2 of reference 37. A typical virtual TB patient was defined as a 70-kg human. No random effects or uncertainty components were included in the simulations. An initial extracellular M. tuberculosis inoculum of 100 bacteria was introduced at day 0 in all simulations.

First, the simulations were performed to evaluate the effects on bacterial load and on cytokine levels following HRZE therapy with and without adjunctive metformin treatment at three different dosing regimens starting at day 180 postinfection, i.e., upon diagnosis. Day 180 postinfection was selected as the approximate time to diagnosis and as such, the starting point for treatment, based on a prior model (17). In this first set of simulations, metformin was added at the start of HRZE treatment. Additional simulations were performed to predict the effects on total bacterial load if metformin were added at the end of 2 months of intensive HRZE treatment. The metformin dosing regimen used in the simulations included 250 mg, 500 mg, and 1,000 mg, all BID. The HRZE regimen in the simulations included 300 mg isoniazid, 600 mg rifampin, 1,500 mg pyrazinamide, and 1,100 mg ethambutol, all QD (once a day) for 2 months, followed by the same dose of isoniazid and rifampin for 4 months. Next, to understand the effects of metformin on TB disease progression in scenarios where diabetic patients would be receiving metformin for their glycemic control at the time of infection with TB (8, 9), simulations were performed to predict the effects of 500 mg BID metformin treatment starting at day 1 postinfection.

Software.

All parameter optimization and model simulations were conducted using R and RStudio with the nlmixr and RxODE packages (36). The literature model for autophagy was converted from an SBML file to ODEs in R using the IQRsbml package (https://iqrsbml.intiquan.com/main.html).

ACKNOWLEDGMENTS

We thank Sud et al., Sonntag et al., and Fors et al. for making their model codes available in the literature.

K.M. and J.G.C.V.H. were involved in designing the analysis; K.M. performed the hands-on analysis and preparation of the first draft of the manuscript; T.G. performed quality checks of the model codes. All authors were involved in reviews and revisions of the manuscript.

No funding was received for this work.

R.S.W. is a coinvestigator in a clinical study of metformin use in TB supported by the NIH and a principal investigator of metformin use in TB supported by the EC Horizon 2020 program. All other authors declare no conflicts of interest.

All data used in the analysis were obtained from literature, and source literature is cited where applicable. The supplementary materials and final model code used in the simulations is available at https://github.com/krinaj/TB_Autophagy_Metformin_Model.

REFERENCES

  • 1.Nahid P, Dorman SE, Alipanah N, Barry PM, Brozek JL, Cattamanchi A, Chaisson LH, Chaisson RE, Daley CL, Grzemska M, Higashi JM, Ho CS, Hopewell PC, Keshavjee SA, Lienhardt C, Menzies R, Merrifield C, Narita M, O'Brien R, Peloquin CA, Raftery A, Saukkonen J, Schaaf HS, Sotgiu G, Starke JR, Migliori GB, Vernon A. 2016. Executive summary: official American Thoracic Society/Centers for Disease Control and Prevention/Infectious Diseases Society of America clinical practice guidelines: treatment of drug-susceptible tuberculosis. Clin Infect Dis 63:853–867. 10.1093/cid/ciw566. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Ginsberg AM, Spigelman M. 2007. Challenges in tuberculosis drug research and development. Nat Med 13:290–294. 10.1038/nm0307-290. [DOI] [PubMed] [Google Scholar]
  • 3.Mehta K, Spaink HP, Ottenhoff THM, van der Graaf PH, van Hasselt JGC. 2022. Host-directed therapies for tuberculosis: quantitative systems pharmacology approaches. Trends Pharmacol Sci 43:293–304. 10.1016/j.tips.2021.11.016. [DOI] [PubMed] [Google Scholar]
  • 4.Kilinç G, Saris A, Ottenhoff THM, Haks MC. 2021. Host-directed therapy to combat mycobacterial infections. Immunol Rev 301:62–83. 10.1111/imr.12951. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Young C, Walzl G, Du Plessis N. 2020. Therapeutic host-directed strategies to improve outcome in tuberculosis. Mucosal Immunol 13:190–204. 10.1038/s41385-019-0226-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Singhal A, Jie L, Kumar P, Hong GS, Leow MK-S, Paleja B, Tsenova L, Kurepina N, Chen J, Zolezzi F, Kreiswirth B, Poidinger M, Chee C, Kaplan G, Wang YT, De Libero G. 2014. Metformin as adjunct antituberculosis therapy. Sci Transl Med 6:263ra159. 10.1126/scitranslmed.3009885. [DOI] [PubMed] [Google Scholar]
  • 7.Lachmandas E, Eckold C, Böhme J, Koeken VACM, Marzuki MB, Blok B, Arts RJW, Chen J, Teng KWW, Ratter J, Smolders EJ, Van den Heuvel C, Stienstra R, Dockrell HM, Newell E, Netea MG, Singhal A, Cliff JM, Van Crevel R. 2019. Metformin alters human host responses to Mycobacterium tuberculosis in healthy subjects. J Infect Dis 220:139–150. 10.1093/infdis/jiz064. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Degner NR, Wang J-Y, Golub JE, Karakousis PC. 2018. Metformin use reverses the increased mortality associated with diabetes mellitus during tuberculosis treatment. Clin Infect Dis 66:198–205. 10.1093/cid/cix819. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Yu X, Li L, Xia L, Feng X, Chen F, Cao S, Wei X. 2019. Impact of metformin on the risk and treatment outcomes of tuberculosis in diabetics: a systematic review. BMC Infect Dis 19:859. 10.1186/s12879-019-4548-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.van der Graaf PH, Benson N. 2018. The role of quantitative systems pharmacology in the design of first-in-human trials. Clin Pharmacol Ther 104:797. 10.1002/cpt.1145. [DOI] [PubMed] [Google Scholar]
  • 11.Gumbo T, Angulo-Barturen I, Ferrer-Bazaga S. 2015. Pharmacokinetic-pharmacodynamic and dose-response relationships of antituberculosis drugs: recommendations and standards for industry and academia. J Infect Dis 211:S96–S106. 10.1093/infdis/jiu610. [DOI] [PubMed] [Google Scholar]
  • 12.Gaohua L, Wedagedera J, Small BG, Almond L, Romero K, Hermann D, Hanna D, Jamei M, Gardner I. 2015. Development of a multicompartment permeability-limited lung PBPK model and its application in predicting pulmonary pharmacokinetics of antituberculosis drugs. CPT Pharmacometrics Syst Pharmacol 4:605–613. 10.1002/psp4.12034. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Marino S, Kirschner DE. 2004. The human immune response to Mycobacterium tuberculosis in lung and lymph node. J Theor Biol 227:463–486. 10.1016/j.jtbi.2003.11.023. [DOI] [PubMed] [Google Scholar]
  • 14.Kirschner D, Pienaar E, Marino S, Linderman JJ. 2017. A review of computational and mathematical modeling contributions to our understanding of Mycobacterium tuberculosis within-host infection and treatment. Curr Opin Syst Biol 3:170–185. 10.1016/j.coisb.2017.05.014. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Goutelle S, Bourguignon L, Jelliffe RW, Conte JE, Maire P. 2011. Mathematical modeling of pulmonary tuberculosis therapy: insights from a prototype model with rifampin. J Theor Biol 282:80–92. 10.1016/j.jtbi.2011.05.013. [DOI] [PubMed] [Google Scholar]
  • 16.Lyons MA, Lenaerts AJ. 2015. Computational pharmacokinetics/pharmacodynamics of rifampin in a mouse tuberculosis infection model. J Pharmacokinet Pharmacodyn 42:375–389. 10.1007/s10928-015-9419-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Fors J, Strydom N, Fox WS, Keizer RJ, Savic RM. 2020. Mathematical model and tool to explore shorter multi-drug therapy options for active pulmonary tuberculosis. PLoS Comput Biol 16:e1008107. 10.1371/journal.pcbi.1008107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Sud D, Bigbee C, Flynn JL, Kirschner DE. 2006. Contribution of CD8+ T cells to control of Mycobacterium tuberculosis infection. J Immunol 176:4296–4314. 10.4049/jimmunol.176.7.4296. [DOI] [PubMed] [Google Scholar]
  • 19.Ehlers S, Schaible UE. 2012. The granuloma in tuberculosis: dynamics of a host-pathogen collusion. Front Immunol 3:411. 10.3389/fimmu.2012.00411. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Padmapriydarsini C, Mamulwar M, Mohan A, Shanmugam P, Gomathy NS, Mane A, Singh UB, Pavankumar N, Kadam A, Kumar H, Suresh C, Reddy D, Devi P, Ramesh PM, Sekar L, Jawahar S, Shandil RK, Singh M, Menon J, Guleria R, METRIF Team . 2021. Randomized trial of metformin with anti-tuberculosis drugs for early sputum conversion in adults with pulmonary tuberculosis. Clin Infect Dis:ciab964. 10.1093/cid/ciab964. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Sonntag AG, Dalle Pezze P, Shanley DP, Thedieck K. 2012. A modelling-experimental approach reveals insulin receptor substrate (IRS)-dependent regulation of adenosine monosphosphate-dependent kinase (AMPK) by insulin. FEBS J 279:3314–3328. 10.1111/j.1742-4658.2012.08582.x. [DOI] [PubMed] [Google Scholar]
  • 22.Svensson RJ, Aarnoutse RE, Diacon AH, Dawson R, Gillespie SH, Boeree MJ, Simonsson USH. 2018. A population pharmacokinetic model incorporating saturable pharmacokinetics and autoinduction for high rifampicin doses. Clin Pharmacol Ther 103:674–683. 10.1002/cpt.778. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Denti P, Jeremiah K, Chigutsa E, Faurholt-Jepsen D, PrayGod G, Range N, Castel S, Wiesner L, Hagen CM, Christiansen M, Changalucha J, McIlleron H, Friis H, Andersen AB. 2015. Pharmacokinetics of isoniazid, pyrazinamide, and ethambutol in newly diagnosed pulmonary TB patients in Tanzania. PLoS One 10:e0141002. 10.1371/journal.pone.0141002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.McCallum AD, Pertinez HE, Else LJ, Dilly-Penchala S, Chirambo AP, Sheha I, Chasweka M, Chitani A, Malamba RD, Meghji JZ, Gordon SB, Davies GR, Khoo SH, Sloan DJ, Mwandumba HC. 2021. Intrapulmonary pharmacokinetics of first-line anti-tuberculosis drugs in Malawian patients with tuberculosis. Clin Infect Dis 73:e3365–e3373. 10.1093/cid/ciaa1265. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Jermain B, Hanafin PO, Cao Y, Lifschitz A, Lanusse C, Rao GG. 2020. Development of a minimal physiologically-based pharmacokinetic model to simulate lung exposure in humans following oral administration of ivermectin for COVID-19 drug repurposing. J Pharm Sci 109:3574–3578. 10.1016/j.xphs.2020.08.024. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Jayaraman P, Sada-Ovalle I, Nishimura T, Anderson AC, Kuchroo VK, Remold HG, Behar SM. 2013. IL-1β promotes antimicrobial immunity in macrophages by regulating TNFR signaling and caspase-3 activation. J Immunol 190:4196–4204. 10.4049/jimmunol.1202688. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Reiling N, Homolka S, Walter K, Brandenburg J, Niwinski L, Ernst M, Herzmann C, Lange C, Diel R, Ehlers S, Niemann S. 2013. Clade-specific virulence patterns of Mycobacterium tuberculosis complex strains in human primary macrophages and aerogenically infected mice. mBio 4:e00250-13. 10.1128/mBio.00250-13. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Musuka S, Srivastava S, Siyambalapitiyage Dona CW, Meek C, Leff R, Pasipanodya J, Gumbo T. 2013. Thioridazine pharmacokinetic-pharmacodynamic parameters “wobble” during treatment of tuberculosis: a theoretical basis for shorter-duration curative monotherapy with congeners. Antimicrob Agents Chemother 57:5870–5877. 10.1128/AAC.00829-13. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Howell JJ, Hellberg K, Turner M, Talbott G, Kolar MJ, Ross DS, Hoxhaj G, Saghatelian A, Shaw RJ, Manning BD. 2017. Metformin inhibits hepatic mTORC1 signaling via dose-dependent mechanisms involving AMPK and the TSC complex. Cell Metab 25:463–471. 10.1016/j.cmet.2016.12.009. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Hawley SA, Gadalla AE, Olsen GS, Hardie DG. 2002. The antidiabetic drug metformin activates the AMP-activated protein kinase cascade via an adenine nucleotide-independent mechanism. Diabetes 51:2420–2425. 10.2337/diabetes.51.8.2420. [DOI] [PubMed] [Google Scholar]
  • 31.Wallis RS, Vinhas SA, Johnson JL, Ribeiro FC, Palaci M, Peres RL, Sá RT, Dietze R, Chiunda A, Eisenach K, Ellner JJ. 2003. Whole blood bactericidal activity during treatment of pulmonary tuberculosis. J Infect Dis 187:270–278. 10.1086/346053. [DOI] [PubMed] [Google Scholar]
  • 32.Sabiiti WS, Bowness RB, Gillespie SHG, Consortium P. 2016. Rapid monitoring of tuberculosis treatment response: a multi-centre study. Eur Respir J 48:OA1513. 10.1183/13993003.congress-2016.OA1513. [DOI] [Google Scholar]
  • 33.Hai HT, Vinh DN, Thu DDA, Hanh NT, Phu NH, Srinivasan V, Thwaites GE, Thuong N. 2019. Comparison of the Mycobacterium tuberculosis molecular bacterial load assay, microscopy and GeneXpert versus liquid culture for viable bacterial load quantification before and after starting pulmonary tuberculosis treatment. Tuberculosis (Edinb) 119:101864. 10.1016/j.tube.2019.101864. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Alden K, Read M, Timmis J, Andrews PS, Veiga-Fernandes H, Coles M. 2013. Spartan: a comprehensive tool for understanding uncertainty in simulations of biological systems. PLoS Comput Biol 9:e1002916. 10.1371/annotation/cfcaffb6-8515-4595-b6cc-85b84703d7d5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Marino S, Hogue IB, Ray CJ, Kirschner DE. 2008. A methodology for performing global uncertainty and sensitivity analysis in systems biology. J Theor Biol 254:178–196. 10.1016/j.jtbi.2008.04.011. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Fidler M, Wilkins JJ, Hooijmaijers R, Post TM, Schoemaker R, Trame MN, Xiong Y, Wang W. 2019. Nonlinear mixed-effects model development and simulation using nlmixr and related R open-source packages. CPT Pharmacometrics Syst Pharmacol 8:621–633. 10.1002/psp4.12445. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Mehta K, Guo T, Wallis R, van der Graaf PH, Coen van Hasselt JG. 2022. krinaj/TB_Autophagy_Metformin_Model. https://github.com/krinaj/TB_Autophagy_Metformin_Model. Retrieved 7 July 2022. [Google Scholar]

Articles from Antimicrobial Agents and Chemotherapy are provided here courtesy of American Society for Microbiology (ASM)

RESOURCES