Experimental approaches to inform design of Host-Directed Therapies

Several experimental approaches experimentally modelling (parts of) Mtb infection are of relevance to inform design of HDTs. Such experiments can provide quantitative understanding about components of drug-host-pathogen interactions (Figure 2 ), which can be combined and integrated through the use of QSP modelling. Key aspects include the immune modulatory effect of therapeutic agents on immune cells that in turn lead to changes in Mtb inhibition-dynamics of immune cells, immune system evasion strategies induced by Mtb, and direct pharmacodynamic effects of antimicrobial agents used in combination with HDTs. Parametrization of QSP models requires quantitative data of both the time course of effects (i.e., rates) and the concentration-effect relationships for therapeutic agents studied; as well as the ability to perform time course measurements. Data to be measured during such experiments both include endpoints such as Mtb disease burden, cell counts of lymphocytes, and regulatory proteins and biomarkers that can explain observed treatment response[70]. Here, we discuss key in vitro and in vivo preclinical models that can be of specific relevance for characterization of HDTs using a QSP modelling approach.

In Vitro Macrophage Infection Models and Advanced Cell Culture Systems

Human-derived macrophage and peripheral blood mononuclear cell (PBMC) cultures are extensively used to screen for the activity of antimicrobials but also identify potential compounds with HDT potential[27,71–73]. The in vitro setting allows conducting experiments in high-throughput setting including the use of reporter cell lines to screen for specific effects at the molecular level, e.g., autophagy induction. This approach thus readily allows characterization of the time course and dose response relationship of compounds.
Several advanced cell culture systems have been increasingly used to study Mtb host-pathogen interactions and for screening of compounds including HDTs, such as based on 3D cell cultures and organoids[71,74], and the development of a lung-on-chip system[75]. Similar to simpler cell culture systems, the longitudinal measurement of cytokines, chemokines, and bacterial load is possible, and has been used to study HDT interventions[74]. The use of lung organoids and lung-on-chip and their overall use in drug discovery and development is yet to be advanced further.
The in vitro hollow fibre infection model (HFIM) is commonly used to study the direct effects of antimicrobial agents on Mtb, but it also readily allows to include co-cultures with macrophages to better reflect aforementioned in vitro intracellular infection systems. In the HFIM, Mtb is cultured in a closed chemostat system with continuous flow of medium, while it allows simulation of concentration-time profiles that occur in patients. The corresponding changes in bacterial load over time can be quantified, allowing characterization of underlying PK/PD relationships of antibiotic and/or HDTs in combination with PK/PD modelling[76].
Whilst all in vitro cell culture-based approaches are attractive for purposes of screening and quantitative characterization of key mechanisms and phenotypic response profiles, these systems remain a simplified model system that does not include all aspects related to the host immune response. The use of QSP modelling could facilitate translation of such in vitro responses based on human host cells towards expected in vivo response.

Zebrafish Infection Models

Adult zebrafish Mycobacterium marinum (Mm) infection models have gained increasing attention as a preclinical Mtb infection model[77–79]. Zebrafish embryos and larvae are of interest due to their optical transparency and thus allowing the use of advanced imaging methods. Zebrafish possess an innate immune system that is highly similar to that of mammals; therefore, it has been used in many studies for the analysis of cellular and systemic responses to infection[78,80,81]. Because infection with various mycobacteria, including Mtb leads rapidly to the formation of granuloma structures that are highly similar to those observed in human tuberculosis patients, it has been a successful model to study the progression of tuberculosis and the effects of drug treatment[82,83]. It also enables pharmacological screening of drugs, to treat mycobacterial infection at a high throughput level with an emphasis on the measurement of drug uptake characteristics[84]. Knockdown and overexpression experiments in zebrafish combined with QSP modelling would especially provide insights into contribution of certain component to overall immune response and anti-TB effects. A recent study provided the proof of concept that use of zebrafish larvae combined with translational PK/PD modelling can accurately predict effects of anti-TB drugs in humans[85]. Thus, zebrafish is a promising experimental TB model that can be used to generate the data required for QSP models to evaluate HDT strategies.

Vertebrate Infection Models

Rodent infection models using mice, rabbits, and guinea pigs are commonly used as infection model for Mtb[77]. Mice have been used in TB research from a long time due to the small size, availability of humanized and genetically modified strains, and cost-benefits over other vertebrate in vivo models, such as rabbits, guinea pigs, and NHPs. Even though rodent infection models such as mice incorporate a full immune system, differences between the human immune response remain[86] and lead to translational challenges[71]. QSP models could help address some of those translational challenges. Mice infected with ultra-low dose aerosol Mtb showed heterogeneous disease progression and granuloma formation similar to humans. Analysis of the transcriptomics data obtained from the ultra-low dose Mtb infected mice and the controls enabled predictions of risk of progression to active TB disease following Mtb infection in humans.[87] QSP models can incorporate various doses of Mtb inoculum and findings from the transcriptomics data analysis, and can enable translational predictions of treatment outcome[88–90].
NHPs have been widely used in immunology research and TB vaccine studies. NHP-Mtb infection models are of interest to generate HDT-relevant data due to their similarities to humans in basic physiology, immunology, and disease pathology. The use of these models has been however limited in TB treatment research due to the requirements of scientific and financial resources as well as safety issues due to highly infectious and contagious nature of Mtb.[91] QSP models can fill in the gaps between in vitro, zebrafish, smaller vertebrates such as mice, and humans to overcome the limitations of NHP models. In general, QSP models can link results from various experimental infection models to enable predictions in humans.