Introduction

Mycobacterium tuberculosis (Mtb) infections are associated with approximately 1.5 - 2 million deaths annually worldwide[1]. The current first-line treatments for tuberculosis (TB) disease include a combination of antibiotics (rifampicin, isoniazid, pyrazinamide, and ethambutol) for at least six months[2]. However, the ongoing emergence of multidrug resistant Mtb threatens the effectiveness of the treatment with conventional antibiotics[3]. Host-directed therapy (HDT) strategies targeting the host immune response against Mtb to complement conventional antibiotic treatment strategies have received increasing attention[4–11] to enhance treatment outcomes, shorter treatment durations, and avoidance of resistance development.
HDTs target interactions between the host immune response and the Mtb pathogen. The host immune response to Mtb infection is reliant on the cumulative activities of various defence mechanisms such as macrophage activation, phagocytosis, autophagy, antigen presentation, and cytokine and T-lymphocytes production. Mtb has several mechanisms to modulate the host response enabling evasion of immune system-mediated clearance[1,11]. Pharmacological targeting of specific host-pathogen interaction mechanisms reflects an important approach for HDTs. Understanding the multiscale nature of host-pathogen interactions is essential to identify relevant drug targets for HDTs, and design appropriate combination treatments and dosing schedules.
A major challenge in the discovery and development of HDTs for TB is the prediction of treatment responses associated with specific pharmacological modulation of an immune response-associated target due to complex systems-level host-drug-pathogen interactions[4,6,12]. The translation of systems-level responses to HDT strategies from preclinical models to patients is challenged by inter-species differences in immune responses to Mtb pathogen. Mathematical modelling, and in particular the use of quantitative systems pharmacology (QSP) modelling can serve as a valuable tool to identify relevant HDT targets, and to inform subsequent design of combination drug treatment strategies and dosing schedules[13–18]. The utility of quantitative modelling to design improved treatment strategies for TB have already been demonstrated extensively for conventional antibiotic therapies[17–19]. For design of HDTs, however, QSP approaches remain have not yet been developed.
Here, we discuss the utility of QSP modelling strategies to support discovery and development of HDT strategies. We summarize high-potential host-pathogen interactions of relevance for HDTs. We then provide an overview of several relevant infection models to characterize host-pathogen interactions of Mtb. Based on this, we discuss how QSP models can be developed with a focus on required model components and the integration with experimental and clinical data, for application in target selection, inter-species translation and for clinical study design and treatment optimization.
Host -Pathogen Interactions as basis for Host-Directed Therapy Strategies
Several host-pathogen interactions of Mtb involved in its pathogenesis and immune system evasion offer potential targets for design of HDTs[11] (Figure 1 ), and are of relevance to capture in QSP modelling approaches.

Induction of Autophagy

Autophagy involves the formation of a double-membrane phagophore, elongation of the phagophore, autophagosome maturation, and fusion with lysosomes for degradation of the selected cellular material, and requires a complex interplay between various protein complexes. Autophagy plays an essential role in controlling Mtb infections[20–23] and has been studied extensively as potential HDT strategy for Mtb[4,11,20,24]. Currently, two therapeutic targets, mammalian target of rapamycin (mTOR) and intracellular cholesterol, are being studied to induce autophagy.

mTOR Inhibitors

Mammalian target of rapamycin complex 1 (mTORC1) plays a role in regulation of autophagy by two mechanisms, (1) inhibition of unc-51-like kinase 1 (ULK1) and transcription Factor EB (TFEB) phosphorylation[24] and (2) activation of glycolysis[25]. Mtb activates mTORC1 and thus inhibits autophagy.
Metformin is the most evaluated mTOR inhibitor as potential HDT treatment for Mtb infections. Metformin inhibited the growth of intracellular MDR Mtb strains in vitro[26]. Adjunctive treatment of metformin with isoniazid induced phagosome-lysosome fusion, enhanced the immune response, and reduced intracellular growth of Mtb in mice[26]. Study of transcriptional changes in healthy human volunteers following metformin dosing reported that metformin alters mTOR signalling, induces autophagy, and enhances the host response to Mtb[27]. Multiple reports suggest that metformin adjunctive therapy in diabetic TB patients improved TB therapy success rate and lowered mortality rate[26,28,29].
Everolimus, an mTOR inhibitor, showed significant potential against Mtb as an HDT. In a human granuloma model, everolimus treatment alone or in combination with isoniazid or pyrazinamide showed significant reduction in Mtb load as compared to the controls.[30] Adjunctive everolimus treatment with rifabutin-substituted standard TB therapy improved lung functions as measured by forced expiratory volume (FEV1) when compared to a control group in a randomized clinical trial[31]. A recent study identified that protein kinase inhibitor ibrutinib as a potential HDT against Mtb. Ibrutinib therapy alone significantly promoted auto-lysosome fusion in vitro, inhibited the mTOR pathway in vitro, and reduced Mtb load in mice[32]. Overall, induction of autophagy via mTOR inhibitors, especially in combination with conventional Mtb therapy, holds a potential as an adjunctive HDT strategy for treatment of TB.

HMG-CoA Inhibitors

Autophagy is also dependent on intracellular cholesterol. Key proteins, 1A/1B-light chain 3 (LC3) and lysosomal associated membrane protein 3 (LAMP3), and Ca2+ are essential for autophagosome maturation and autophagosome-lysosome fusion. LC3, LAMP3, and Ca2+ are inhibited by intracellular cholesterol[5,33], and thus cholesterol inhibits autophagy and promotes Mtb survival.
The HMG-CoA reductase pathway has been associated with intracellular cholesterol reduction, autophagy induction and improved Mtb clearance. Therapy with HMG-CoA inhibitors, such as simvastatin, pravastatin, and fluvastatin, as adjunctive therapy to conventional anti-TB drugs improved bacterial clearance by the host and improved the efficacy of first-line TB drugs by promoting phagosome maturation and autophagy in macrophage cell cultures and in mice models.[7,34–36] In vitro screening and experiments in mice for eight HMG-CoA inhibitors discovered that pravastatin, simvastatin, and fluvastatin had the most favourable anti-TB activity and pravastatin showed the least toxicity and drug-drug interactions when used as an adjunctive to standard anti-TB treatment[7,33]. On the other hand, a population-based cohort analysis of data from newly diagnosed TB patients recognized no statistically significant difference in hazard ratio between patients who were using statins (as a lipid lowering treatment) in addition to standard TB treatment as compared to patients who did not use statins[37]. Several retrospective clinical studies have identified that chronic use of statins reduced the risk of developing TB; however, to our knowledge, no studies have evaluated statins as a treatment in active TB patients alone or in combination with conventional anti-TB therapy[38]. As such, prospective clinical studies assessing the use of statins, especially pravastatin, at different doses as adjunctive to standard TB therapy may be needed.

Regulation of Host Epigenetics

Infection with Mtb alters some host gene functions important for the ensuring immune response. Two key pathways involved in Mtb-induced host epigenetic alterations are histone deacetylases1 (HDAC1) and TLR3-BMP-miR27a pathway both of which can be pharmacologically exploited for HDTs[39–41].

HDAC Inhibitors

Upregulation of HDAC1 was noted in macrophages containing live Mtb and HDAC1 recruitment suppressed the expression of IL-12B that plays a vital role in initiating type 1 T cell immunity to Mtb. HDAC1 is also known to modulate autophagy associated genes[42,43]. Treatment with a broad-spectrum HDAC inhibitor (Trichostatin A) decreased bacterial growth in both M1 and M2 macrophage cell cultures, while selective HDAC inhibitors (TMP195, and TMP269) reduced bacterial growth in M2 macrophage cell cultures. Vorinostat, an HDAC inhibitor, promoted immune response by human macrophage cell cultures[44]. In zebrafish embryos infected with Mm, HDAC inhibition significantly reduced microbial burden[40]. Additionally, HDAC inhibition significantly inhibited Mtb growth in lungs and showed increased production of key cytokines in mice[45].

Abl Tyrosine Kinase Inhibitors

Abl tyrosine kinase is involved in entry and survival of Mtb within macrophages through TLR3-BMP-miR27a pathway. Abl tyrosine kinase also inhibits expression of vATPase pump-relevant genes, and thus inhibits acidification of autolysosomes. Pharmacological inhibition of Abl tyrosine kinase using imatinib improved containment of Mtb within macrophages, induced expressions of iNOS, increased acidification of phagosomes, and decreased bacterial load in human macrophage cell cultures and in mice[11,41]. A clinical study assessing effects of imatinib alone and in combination with conventional anti-TB drugs in drug-resistant- and HIV co-infected- TB patients[46] is currently ongoing.

Modulation of Cytokine Response

The kinetics of the key cytokines, such as interferon gamma (IFN-γ) , tumour necrosis alpha (TNF-α), interleukin (IL)-1β, IL-10, IL-4, IL-12, and IL-2, during the course of Mtb infections have been well studied in vitro and in vivo[47–51]. IFN-γ is one of the most important players to the host immune response and its main role is activation of macrophages. IFN-γ also induces infected macrophage apoptosis via induction of more than 200 pro-apoptotic genes (i.e. Fas/Fas ligand, cathepsin, protein kinase R, etc.)[52,53]. Activated macrophage produce reactive nitrogen intermediates (RNIs) and pro-inflammatory cytokines, TNF-α and IL-1β, that possess microbicidal properties against Mtb. Resident macrophages also produce RNIs, TNF-α, and, IL-1β; however activated macrophage-mediated production is much more efficient[54–56]. Excessive production of pro-inflammatory cytokines, however, can lead to tissue damage[57]. Anti-inflammatory cytokines, IL-10 and IL-4, are also induced upon macrophage phagocytosis and balance pro-inflammatory cytokine levels by macrophage deactivation[57]. However, excessive production of anti-inflammatory cytokines may result in limiting the host immune systems’ microbicidal activities[58]. Thus, the fine balance between the pro- and anti-inflammatory cytokines may determine the overall outcome of the Mtb infection.
Adjunctive treatment with IFN-y have been evaluated in various clinical studies; however, different patient conditions, routes of administration (intravenous vs. subcutaneous) and dosing regimen resulted in varying outcomes[59]. Adjunctive treatment with aerosolized IFN-y showed benefits in reducing cavitary lesions and induced negative sputum conversion in TB patients in clinical studies[60,61].
Anti-inflammatory agents such as cyclooxygenase (COX) 1/2 inhibitors, corticosteroids, 5-lipoxygenase inhibitor (Zileuton), phosphodiesterase (PDE) inhibitors, and matrix metalloproteinases (MMP) inhibitors have been shown to reduce Mtb burden in vitro or in preclinical species[4]. However, treatments with corticosteroids and celecoxib (COX1 inhibitor) in combination with conventional anti-TB drugs did not show significant benefits of these additional HDT in human subjects[4,62]. Adjunctive treatments with PDE inhibitors and MMP inhibitors have not been evaluated yet in human subjects to our knowledge. Retrospective analysis of existing data where TB patients took approved anti-inflammatory drugs, especially PDE inhibitors, as concomitant medications for other conditions and their impact on TB outcome can be a valuable approach.

Enhancing T-cell Mediated Host Response

The overall innate immune reaction play an important role in the initiation of adaptive immune response by antigen presentation, cytokines, and costimulatory signals[55]. Two to three weeks after the initial infection, antigen-presenting cells (APCs) that drain into regional lymph nodes initiate adaptive T-lymphocytes mediated immune response. Upon antigen presentation, the APCs, through antigen presentation via major histocompatibility molecules (MHC)-I and II , prime CD8+ T cells (cytotoxic T cells) and CD4+ T cells, respectively to initiate adaptive immune response.[63–65] Both activated CD4+ and CD8+ T cells secrete IFN‑γ, IL-2, IL-17A, and IL-10. The production of CD4+ mediated IFN-γ is further stimulated by activated macrophages, whereas the production of CD8+ mediated IFN-γ is driven by concentrations of IL-12 and correlates with bacterial load[50]. Mature dendritic cells secrete IL-12p70 which helps increasing recruitment of additional CD4+ T cells[66,67]. IL-2 play a role in further proliferation of T cells[57]. CD8+ cells have direct microbicidal capabilities through perforin, granzymes, and granulysin or induce apoptosis through Fas/Fas ligand interaction[68].
Adjunctive cytokine supplementation with IL-12 and IL-2 have been evaluated in clinical studies, but did not result in significant benefits[4,59]. However, recombinant human IL-2 supplementation showed significant improvements in negative sputum culture conversion rates and in enhanced X-ray resolution in MDR TB patients[69]. Therefore, the use of recombinant IL-2 supplementation as HDT strategy for TB should be further evaluated.