Abstract
The mortality caused by tuberculosis (TB) infections is a global concern, and there is a need to improve our understanding of the disease. Current in vitro infection models to study the disease have limitations, such as short investigation durations and divergent transcriptional signatures. This study aims to overcome these limitations by developing a 3D collagen culture system that mimics the biomechanical and extracellular matrix (ECM) of lung microenvironment (collagen fibers, stiffness comparable to in vivo conditions), as the infection primarily manifests in the lungs. The system incorporates Mycobacterium tuberculosis (Mtb) infected human THP-1 or primary monocytes/macrophages. Dual RNA sequencing revealed higher mammalian gene expression similarity with patient samples than 2D macrophage infections. Similarly, bacterial gene expression more accurately recapitulated in vivo gene expression patterns compared to bacteria in 2D infection models. Key phenotypes observed in humans, such as foamy macrophages and mycobacterial cords, were reproduced in the model. This biomaterial system overcomes challenges associated with traditional platforms by modulating immune cells and closely mimicking in vivo infection conditions, including showing efficacy with clinically relevant concentrations of anti-TB drug pyrazinamide, not seen in any other in vitro infection model, making it reliable and readily adoptable for tuberculosis studies and drug screening.
Keywords: Collagen hydrogels, cord formation, foamy macrophages, pyrazinamide
Introduction
In 2022, the World Health Organization (WHO) reported 10.6 million new TB cases and 1.3 million deaths.1 TB is transmitted through bacteria-containing aerosols generated by an infected individual. Upon reaching the alveolus of the new host, the bacteria infect the resident alveolar macrophages, which migrate to the lung interstitium and recruit circulating monocytes and other immune cells, leading to the formation of granuloma.2, 3 Monocytes and macrophages carry a large reservoir of bacteria in vivo and play critical roles in the host response to mycobacterial infection.4–6 These cells are important for studying host-pathogen interactions and hence have been widely employed in the field. 7–9
Conventionally, in vitro Mtb infection experiments to study host-pathogen interaction are performed in two-dimensional (2D) tissue culture systems on polystyrene plates, using primary human monocyte-derived macrophages, mouse bone marrow-derived macrophages (BMDMs), and human THP-1 macrophages.10 These studies have contributed to our understanding of complex processes such as the arrest of phagosomal maturation, activation of innate immune responses, and the underlying mechanisms modulating various cell death pathways.7 However, the existing culture systems have critical drawbacks: First, monolayer (2D) cultures fail to provide the natural microenvironment for the cells, and the cells are usually subjected to very stiff surfaces (~10 GPa), resulting in many dysregulated cellular processes.11 Second, Mtb is a slow-growing pathogen (median doubling time of >100 h) 12, 13 that modulates host responses to help its survival over several days/weeks. However, the length of conventional 2D culture experiments is 4-6 days due to low host cell viability after Mtb infection, failing to recapitulate the intricate host-pathogen dynamics. Third, substantial variation exists in gene expression between infected cells in 2D cultures and human infections.14 Fourth, the current systems seem unreliable as platforms to screen TB drugs due to the disparities they present between drug efficacies with in vivo systems. For instance, pyrazinamide, a first-line anti-TB drug extremely effective in vivo, has shown little to no efficacy in any previously used in vitro culture system at clinically relevant concentrations.15 Moreover, several clinical trials with potential TB drugs and treatment regimens have failed despite showing promise in the initial stage of testing, indicating that we may not be capturing the underlying molecular changes because of the short in vitro infection durations 16. These limitations underscore the need to develop better and more representative culture systems to study host-mycobacteria interactions accurately. An ideal culture system should (a) provide a more extended timeframe (~weeks) for studying host-pathogen dynamics in TB; (b) be able to better recapitulate the genotypic and phenotypic characteristics of human infection; (c) be a more reliable platform for advancing drug susceptibility studies; and (d) be simple with accessible components to enable easy and widespread adoption in the field.
In recent years, three-dimensional (3D) in vitro models have gained significant interest in the field due to their capacity to act as an alternative to animal testing for initiating clinical trials as per the FDA Modernization Act 2.0.17, 18 These 3D models provide a more realistic representation of human tissues than their 2D counterparts, as they mimic the cellular microenvironment, including the extracellular matrix (ECM).19 However, adoption of material-based 3D models for use in infectious diseases like TB is still nascent. A major advancement in using ECM-based matrices to study TB host-pathogen interaction was achieved by Kapoor et al. 20, where the authors used collagen scaffolds to infect peripheral blood mononuclear cells (PBMCs) with Mtb leading to the formation of cellular aggregates resembling TB granulomas.
The authors demonstrated that their model could recapitulate the dormant bacterial phenotype. However, biomechanical aspects (such as Young’s modulus and pore size of the gels), gene expression comparison with primary human samples, and pyrazinamide efficacy were not reported. Another important study used spherical alginate gels doped with collagen as a 3D matrix to infect PBMCs with Mtb to model TB granuloma.20, 21 However, since alginate is not part of the lung ECM, it could result in non-physiological signaling. The synthesis of alginate spheres also requires access to specialized equipment that may not be readily available in BSL-3 settings. A key finding of this study included demonstrating the efficacy of first-line anti-TB drugs in their model, including pyrazinamide. However, a limitation of this study is that the concentrations required to reduce bacterial viability were significantly higher (500 μg/mL) than the clinically relevant concentrations of the drug (50 μg/mL). Our study overcomes the limitations of the current platforms by engineering a readily adoptable collagen-based culture system that mimics human TB infection conditions by recapitulating the in vivo like gene expression and phenotypes of both the bacteria and the host immune cells and drug efficacies using physiologically relevant doses.
We have synthesized hydrogels made of Collagen I, the primary extracellular component in the human lungs, and created a micromechanical environment similar to human lung tissue (Young’s modulus). The collagen gels are easy to form and do not require any specialized equipment. Our study showed that these gels supported the culture of primary human monocytes and THP-1 cells, with the cells sustaining Mtb infection for more than 3 weeks. Dual RNA-sequencing of the host and pathogen transcriptomes from the collagen gels showed a significantly higher similarity of gene expression profiles of the mammalian and bacterial cells with human infection conditions than 2D cultures. Moreover, the monocytes in the infected gels differentiated into macrophages and accumulated lipid bodies, a key phenotype observed in macrophages in human TB granulomas.22 The longevity of this culture system also enabled us to see bacterial growth in the form of cords, a phenotype observed in vivo 5, 23–25 but not in any in vitro macrophage infection model. Finally, our culture system emerged as an excellent drug testing platform, shown by the efficacy of pyrazinamide, at or below the concentrations used clinically (50 μg/mL), unlike previous studies that required a much higher dose. Rapid and widespread adoption of this system is thus likely and would not only aid drug discovery and development but also shed new insights into the complex host-pathogen interactions underlying TB infections.
Results
1. Characterization of collagen gels revealed their micromechanical similarity with human lungs and cytocompatibility
Mtb infection manifests in the lung interstitium 3, primarily comprised of the ECM network of collagen I and III 26, 27 and other proteoglycans. We synthesized and optimized a physically crosslinked collagen I hydrogel that can mimic the biochemical and biophysical cues of the human lung interstitium.28, 29 The overall schematic of the study is represented in Fig. 1A. To characterize the mechanical microenvironment, collagen gels were analyzed using Atomic Force Microscopy (AFM). The Young’s modulus (a measure of the elasticity of the material) obtained was 0.64 kPa ± 0.12 kPa and 2.1 kPa ± 0.3 kPa for 1 mg/mL and 2 mg/mL collagen gels, respectively (Fig. 1B). Since the modulus of healthy human lung tissue is in range of 1.4 to 7 kPa 30, 31, we chose 2 mg/mL collagen gels for all subsequent assays. Next, to characterize whether these gels can mimic the fibrillar structure of ECM in human lungs, the collagen gels were imaged using second harmonic generation (SHG) imaging, a label-free imaging technique based on collagen’s native property to generate the second harmonics when incident with a high intensity laser beam.32 SHG imaging showed distinct fibrils in collagen hydrogels (Fig. 1C) as has been widely reported 28, 33 and morphologically mimicked lung ECM microenvironment. The average pore area in gels with a final collagen I concentration of 2 mg/mL was 8.45 μm2 ± 2.37 μm2, allowing cells to move through the matrix.
Fig. 1. Collagen gels mimic the micromechanical environment of lungs and are cytocompatible.
(A) Schematic of the study (B) Young’s modulus of 1 mg/mL and 2 mg/mL collagen gels. Data in graphs represent the mean ± s.d. (n = 3). (C) Representative SHG image of 2 mg/mL collagen gel. Scale: 5 μm. (D) and (E) Representative images of Calcein-AM (green, live) – PI (red, dead) stained cells images in collagen gels on days 5 and 10, respectively.
The ECM is critical to cell viability as cellular detachment from ECM alters their key biological processes, such as metabolism. The separation from ECM also leads to cell death due to anoikis.34 Collagen plays a critical role in the progress of TB infection as its degradation results in reduced cell viability leading to the caseation of TB granuloma.35 We first tested the cytocompatibility of collagen gels for THP-1 cells by culturing them for 10 days. Minimal cytotoxicity was observed by Calcein -AM – propidium iodide (PI) staining (Fig. 1D and 1E, Day 5 – 92.3 ± 2.76 % viability, Day 10 – 87.1± 4.37 viability).
2. Collagen gels support monocyte and macrophage infection and cell survival for a prolonged duration
We wanted first to determine the multiplicity of infection (MOI) (Mtb: THP-1 cells) that could enable long-term infection studies in our culture system. As a reference, we characterized the infection percentages that could be achieved at a range of MOIs (0.1:1-40:1) in PMA-treated THP-1 macrophages (2D infection) and undifferentiated THP-1 monocytes for our 3D model. To achieve infection in 3D, THP-1 monocytes and a virulent strain of Mtb (H37Rv) were added to the collagen gel before polymerization. We used monocytes to mimic the natural process of TB infection in vivo, during which inflammation leads to the recruitment of circulating monocytes to the site of infection.36 Successful infection was achieved upon incubating THP-1 cells with Mtb in the hydrogels for 12 h, followed by treatment with a mammalian cell impermeable antibiotic (amikacin sulfate) to eliminate extracellular bacteria.37 We performed additional experiments to show that amikacin (at 200 μg/mL dose used in our culture) can eliminate all the extracellular bacteria in the collagen gels through CFU and imaging (Figure S1). As can be seen from the data, there are no detectable extracellular bacteria present two days after the amikacin treatment. Given the large size of these gels (50 μL), we investigated if there were any transport limitations on the diffusion of small molecules (such as antibiotics). We evaluated the diffusion profile of Cy5 dye (MW 653.77 g/mol) whose molecular weight is comparable to Amikacin sulfate (MW 781.76 g/mol), within our collagen gels over time. Briefly, a collagen gel encapsulating THP-1 monocytes was placed on an imaging dish and the edge-to-edge midplane of the gel (in the z-direction) was focused. Live images of the gel were acquired at 20-second intervals for 30 minutes immediately after dye addition (0.1 μg/mL) to the dish. The fluorescence intensity of the dye was plotted against time for fields (in x direction) covering the whole gel (Supplementary Fig. S2). From the intensity profile of the dye, we could conclude that the dye rapidly diffuses throughout the gel indicating that there are no diffusional limitations in the gel. The dye reaches saturation at half the concentration at the center of the gel within 25 minutes (Supplementary Fig. S2). The minimum inhibitory concentration (MIC) of amikacin sulfate against Mtb H37Rv in vitro is 2 μg/mL.38 In our culture system, the amikacin sulfate concentration used is 200 μg/mL, which is 100 times the required minimum concentration. Hence, combined with this data as well as extracellular CFU (Fig. S1), it can be assumed that amikacin sulfate also reaches a saturating concentration of much higher than 2 μg/mL at the center of the gels (in 25 mins).
The infection can be seen by the presence of bacteria (green) inside the cell (nucleus indicated by DAPI staining), and the intracellular presence of the bacteria was confirmed through staining with phalloidin (red) (Fig. 2A, Supplementary Video 1 and 2). An MOI of 40 in the 3D model with monocytes resulted in a similar intracellular infection percentage as an MOI of 1 in 2D infection with macrophages (Supplementary Fig. S3, compare A&B) and was used for further experiments. Harmonizing the initial infection percentages in 2D and 3D was done to allow subsequent comparisons. We also found that infecting undifferentiated monocytes with Mtb (MOI of 40) in the absence of collagen gels resulted in minimal infection (Supplementary Fig. S4), which highlights the importance of collagen in mediating host-Mtb infection.
Fig. 2. Collagen gels support THP-1 cell infection with Mtb over a long duration.
(A) Representative fluorescence microscopy image of infected THP-1 cells in collagen gel (Hoechst-Blue), GFP-Mtb (H37Rv) (Green), and Phalloidin (Red) at 5 days post-infection. (B) The percentage of live THP-1 cells (green), and THP-1 cells infected with H37Rv-GFP (blue) at various time points post-infection (n = 4) as determined using flow-cytometry assays (Gating strategy: Fig. S15). (C) Kinetics of bacterial growth (represented by CFUs) in the 3D collagen culture system (n = 3) over the indicated time points. (D) The percentage of intracellular bacteria containing the replication clock plasmid (pBP10) at various time points in the infected collagen gels, represented by CFUs (mean ± s.d.) on kanamycin vs non-kanamycin plates (n = 3). (E) Plot showing the percentage of live CD68-positive cells in uninfected and infected gels (n = 3) at 10 days post-infection. (F) Mean forward scatter (FSC) and mean side scatter (SSC) plots of CD68-ve and CD68+ve cells from the infected collagen gels (n=3) (G) Representative fluorescence image showing that Mtb (green) forms complex cord-like structures in the infected collagen gels (highlighted using white arrows). Images were acquired from infected collagen 14 days post-infection (Scale: 5 μm). Data in the graphs represent the mean ± s.d., and the p-value was determined by a two-tailed unpaired t-test with Welch’s correction using GraphPad Prism software. p-value < 0.05 was considered significant. *p=0.0282 for E, (**p=0.0056) for FSC and (****p<0.0001) for SSC for F. Experiments related to figure panels B, C, E and G were repeated twice with similar results.
To further emphasize the importance of collagen and the 3D culture, we infected THP-1 cells with Mtb on collagen-coated plates and looked at the infection percentage (Supplementary Fig. S5). When compared to infection in 2D, the collagen coating showed an increased infection percentage (0.83 % ± 0.41 % in 2D vs 4.27% ± 0.71% in collagen-coated plates). Since collagen has domains that can bind to monocytes through receptors such as Integrin αVβ339 and Mtb expresses proteins such as Rv0309 which help it bind to collagen and fibronectin40, we hypothesize coating on plates can help in increasing monocyte-bacteria interactions at the 2D surface. However, when these cells are infected in 3D collagen gels, the infection percentage is significantly higher (13.33% ± 0.9%). This suggests that 3D collagen gels support higher infection percentages for monocytes. We also investigated whether successful infection can be achieved if synthetic hydrogels containing collagen are used. To test this hypothesis, we used a polyethylene glycol maleimide (PEG-MAL)-collagen hydrogel system. PEG-MAL is a synthetic polymer and hydrogels made of PEGs, crosslinked with a protease-degradable peptide such as VPM, are commonly used in many biomedical applications.41–43 We added collagen (0.6 mg/mL) to the precursor solution of 4% PEG-MAL and VPM to encapsulate collagen inside the gels along with the cells. THP-1 monocytes were infected Mtb in the (PEG-MAL)-collagen hydrogels using the same methods that were used for our 3D collagen culture system. The viability and infection percentages in PEG-VPM-collagen hydrogels were 91.8±1.99% and 10.8±1.217% respectively. The corresponding viability and infection % in the 3D collagen culture system were 94.5± 0.721% and 12.93±1.24% respectively (Fig. S6A & B). This shows that synthetic gels with collagen can also be used to study THP1-Mtb infection.
A major limitation of the existing 2D infection models is the relatively small time frame for conducting experiments, making it unsuitable as a mimic of the in vivo infection kinetics.44 We verified this observation, where infection with MOI of 1 in 2D resulted in the gradual loss of viable host cells (THP-1) over 7 days, beyond which the cell viability dropped below 10% (Supplementary Fig. S7A). This is a major constraint for long-term mechanistic and drug studies as Mtb is a slow-growing pathogen with a median doubling time >100 h inside mammalian cells 12. Post-infection, the bacteria take 1-2 days to adapt to the mammalian system 13, which further reduces the available time for assays by nearly 25-30%.
In the infected collagen 3D environment, we could successfully culture THP-1 cells for 22 days (Fig. 2B); at various time points, the viability and infection rates were analyzed using flow cytometry where in eight days post-infection, 60±3% of cells were viable, and the corresponding viability for 15- and 22 days post-infection were 63±7% and 40±9%, respectively, The corresponding infection percentages at 8, 15, and 22 days, were 8±0.9%, 10±1.3%, and 30±12.4%, respectively, indicating a prolonged overall survival of cells upon infection in our system. Since THP-1 is an immortal cell line, we performed similar experiments with human primary macrophages in 2D and in the collagen gels and observed that cells in the infected 3D collagen environment (>80% viability after 12 days) were able to withstand Mtb infection significantly better than in 2D (~20% viability after 7 days) at similar starting infection percentages (Supplementary Fig. S7B (2D infection) and S8 (3D collagen culture system). We also performed a colony forming unit (CFU) assay using H37Rv transformed with the pBP10 clock plasmid 45 (replication clock plasmid with kanamycin resistance) to understand the bacterial growth kinetics in THP-1 cells. The plasmid is lost at a quantifiable rate by replicating bacteria in the absence of antibiotic selection and using a mathematical model, bacterial growth rates can be determined.45 We found that during the first 2 days of infection, the CFUs reduced with minimal bacterial replication, as seen from no loss of the replication clock plasmid, which could be due to the time the bacteria need to adapt to the system (Fig 2C). As time progressed, Mtb replicated (indicated by a corresponding loss in clock plasmid) in the THP-1 cells cultured in collagen gels (Fig. 2D), which further corroborated the increased number of infected cells with time (Fig. 2B) while the overall viability of mammalian cells decreased (Fig. 2B). Thus, this platform offers a significantly longer window for testing drugs and studying host-pathogen interaction during infection than 2D culture systems. A similar long time window to study host-pathogen dynamics was shown when PBMCs were infected with Mtb in alginate-collagen matrices 21. Given, the long duration of infection and the large size of the gels, we checked whether there was oxygen diffusion limitation and if the THP-1 cells and Mtb in the infected collagen gels were under hypoxia. To directly stain for the presence of hypoxia, we treated the infected collagen gels (14 days post-infection) with Pimonidazole, a 2-nitroimidazole compound, which forms covalent bonds with cellular macromolecules at oxygen levels below 1.3% 46, and is routinely used to evaluate hypoxia in cancer studies. We found no pimonidazole staining indicating the lack of hypoxia in the culture system (Supplementary Fig. S9).
During lung infection by Mtb, monocytes get recruited to the site of inflammation and differentiate into macrophages.47, 48 In the collagen culture system, to determine whether monocytes were differentiating into macrophages upon Mtb exposure, cells were stained with CD68, a specific intracellular marker for macrophages.21, 49 An earlier study showed that primary human monocytes differentiated to macrophages upon infection with Mtb in alginate-collagen matrices.21 In our 3D collagen culture system, Mtb infection resulted in CD68 expression in nearly 52% of cells 10 days post-infection (Fig. 2E) compared to only 14% in uninfected gels. We then analyzed the Mtb positive vs. bystander cells in the infected conditions using flow cytometry to determine if all cells were uniformly differentiated into macrophages. Nearly all Mtb-infected cells were positive for CD68 (~95%) (Supplementary Fig. S6), implying that infection resulted in the differentiation of monocytes to macrophages in collagen gels, similar to what happens in vivo. We also observed that about 42% of the bystander (uninfected) population showed CD68 expression (Supplementary Fig. S10A), implying that the infected microenvironment promoted the differentiation of THP-1 monocytes to macrophages. We measured the basal CD68 expression levels in THP-1 cells prior to culturing them in our collagen gels and observed no difference in the mean fluorescence intensity (MFI) between the THP1 monocytes (prior to culture) vs the uninfected collagen gels, suggesting that culturing in collagen gels does not increase CD68 expression in monocytes.
However, the infection with Mtb resulted in a significant increase in MFI, indicating the differentiation of THP-1 monocytes to macrophages as a result of Mtb infection (Supplementary Fig. S10B). We also analyzed the cell size (Forward Scatter (FSC) and Side Scatter (SSC)) in the infected collagen gels and found that cells that were gated as CD68+ve had significantly higher mean FSC and mean SSC, indicating a larger size when compared to the non-differentiated cells (Fig. 2F). We also looked at the macrophage activation marker, inducible nitric oxide synthase (iNOS), which has been shown to play an important role in mice, macaques, and human TB.50–53 As expected, we observed a significant upregulation of iNOS upon Mtb infection in differentiated macrophages, further validating the differentiation of monocytes to macrophages upon infection with Mtb in our system (Supplementary Fig. S10C).
Mtb aggregates and forms cord-like structures, known as cording, which is regularly observed in vivo 23, 24, 54, 55 and is associated with the bacteria’s virulence and immune escape mechanism. One of the hypotheses suggests that cording enables the bacteria to prevent phagocytosis once they become extracellular upon killing the host cell 56, 57 and has also been shown to cause macrophage lysis upon contact.58 Cording correlates to the growth of Mtb in the extracellular milieu in vivo 55, 59 and is also observed in human infection 54, 57, although its role in modulating host dynamics during infection is yet to be established. We also observed bacterial cording at later time points in the culture system (post 10 days) (Fig. 2F, green elongated structures indicated with white arrows), which were mostly extracellular (Supplementary Fig. S11). We classified these elongating bacterial structures as cords if the Ferret diameter was above 10 μm and the average in our system was 15.28 ± 2.17 μm, 14 days post-infection. Previous in vitro studies have shown that human lymphatic endothelial cells and alveolar epithelial cell line (A549) are conducive for cord formation by Mtb, 3-4 days post-infection.57 Furthermore, previous literature has shown that macrophages form extracellular traps in response to Mtb cords to restrict the pathogen’s growth; however, in this model, the mammalian cells were infected with cords grown in 7H9 media without the surfactant tween (as opposed to single cell suspension of bacteria which is usually used for infection studies).58, 60 Hence our 3D collagen culture system could recapitulate a key in vivo infection-related phenotypic change of the bacteria.
3. Dual RNA sequencing revealed substantial similarities in gene expression between mammalian cells of the collagen hydrogel system and those in human TB granulomas
We next performed dual RNA sequencing to determine the extent of similarities between the THP-1 cells and the bacteria from the infected collagen gels with in vivo conditions and to dissect the molecular mechanisms governing host-pathogen interaction by studying their transcriptomic differences in the collagen gels 14 days post-infection, allowing sufficient time for the bacteria to adapt to the system. The differentially expressed mammalian genes (|log2fc|>1, FDR<0.05) between uninfected and infected gels, and Mtb genes between planktonic bacteria and from the infected gels are listed in Table S1 and Table S4, respectively. The differentially expressed gene (DEG) distribution for mammalian genes is shown in Fig. 3A. Gene expression patterns from our system were compared with the expression profiles of in vitro infected human cells - 2D macrophage cell line (GSE162729) and primary cells (GSE174566 - This repository had data pertaining to multiple conditions and we used EM 21.1, EM 23.1, and EM 28.1 corresponding to 2D PBMC controls; EM 21.2, EM 23.2, and EM 28.2 corresponding to Mtb infection of PBMCs in 2D). We also compared our data with gene expression from in vivo human infection conditions (caseous lung granuloma (GSE20050) and lymph-node granuloma (GSE17443). The 3D collagen culture system had the highest number of similarly regulated genes with human caseous TB granuloma 22 (656 genes, gene list: Table S2) and with human lymph node granuloma 14 (165 genes, gene list: Table S3) compared with the other 2D infection models (Fig. 3B), which had significantly less similarly regulated genes (65 genes common with both primary granulomas). Human granulomas are also reported to contain both, monocytes and macrophages61 and since our system has both monocytes and macrophages, this could be another reason apart from the 3D culture and ECM mimicking environment that we observe high similarity compared to 2D THP-1 macrophages alone.
Fig. 3. RNA-sequencing of the mammalian cells from the infected collagen gels show maximum similarity with in vivo systems.
(A) Volcano plot depicting differentially expressed genes in THP-1 cells between uninfected and the Mtb infected hydrogels (red: significantly upregulated, green: significantly downregulated, black: non-regulated) (n=3 per sample) (B) Quantification of the number of common genes between the various models and the primary TB granulomas (green: 3D collagen culture system, pink: PBMCs infected with Mtb in 2D, blue: THP-1 macrophages infected with Mtb in 2D). For PBMCs infected in 2D data from GSE174566 was used. This repository had data pertaining to multiple conditions and we used EM 21.1, EM 23.1, and EM 28.1 corresponding to 2D PBMC controls; EM 21.2, EM 23.2, and EM 28.2 corresponding to Mtb infection of PBMCs in 2D.The upregulated genes in our dataset were compared with the upregulated genes from the public datasets and a similar comparison was made for the downregulated genes (C) Comparison of expression in terms of log2fc of the top upregulated genes and downregulated genes common to lymph node granuloma and the 3D collagen culture system.
Moreover, the DEGs common between the lymph node granuloma and the 3D collagen culture system showed similar regulation patterns (Fig. 3C). To better understand the functional relevance of such a gene expression pattern, we performed a REACTOME pathway gene set enrichment analysis for the differentially regulated genes in various models and found a significant overlap in both the upregulated and downregulated pathways between the human TB granulomas and 3D collagen culture system (Supplementary Fig. S12A and S12B). The significantly enriched REACTOME pathways in the collagen model corresponded to ECM organization, signal transduction, etc., while the significantly depleted ones were associated with protein metabolism, organelle biogenesis, and maintenance, among others. The transcriptomic response of the PBMCs to Mtb in a 3D alginate matrix supplemented with collagen also showed many pathways to be similarly regulated with lymph node TB granuloma. However, the ECM remodeling pathways, a significant characteristic of primary human TB granulomas, did not get enriched in the 3D alginate model upon Mtb infection, probably due to the use of alginate in the matrix.14 In our collagen culture system, despite using collagen 1 as the ECM component, the transcriptomic data showed significant upregulation of genes involved in the synthesis of other ECM proteins such as fibronectin (FN1) and laminin (LAMA2, LAMA4), indicating the active remodeling of ECM during infection in our model. The 2D models, on the other hand, showed opposite regulation of many key infection-associated pathways, such as ECM organization, integrin cell-surface interactions, interleukin, and interferon signaling, or those associated with the cell cycle, compared to the human TB granulomas, highlighting significant lacunae of using such models to study Mtb infection dynamics. These results collectively indicated that our system significantly recapitulated the Mtb-mediated in vivo pathway regulations over the 2D systems.
To reduce the contribution of THP-1 monocyte to macrophage differentiation, we compared our DEG list with a previously published work that reported the gene expression changes as a result of THP-1 monocyte to macrophage differentiation.62 We identified the commonly regulated genes (376 upregulated and 48 downregulated) between the two lists and removed these genes from our DEG list for subsequent transcription factor (TF) analysis. This helped us gain better insights into the transcriptional changes in response to Mtb infection and reduce the contribution of THP-1 monocyte to macrophage differentiation. The database of human TF binding profiles, Transcription Factor Target Gene Database (TFBSDB) 63 was used to identify putative TF regulons over-represented (using hypergeometric tests) in the complete set of THP-1 differentially expressed genes (DEGs), upregulated DEGs, and downregulated DEGs identified in this study. To increase the likelihood that a TF was responsible for the over-representation of its putative targets among the DEGs, TFs that were not differentially expressed were removed from downstream analyses. Furthermore, only regulons over-represented (hypergeometric test adjusted p-value < 0.05) in the full list of DEGs and in the upregulated or downregulated DEGs were considered. The same analysis was performed for two datasets of transcriptional profiles of granulomas from human TB infection.14, 22 We dissected the data to identify TFs potentially associated with the observed changes in gene expression in the human cells in the 3D model (Supplementary Fig.S13). Several of the identified TFs seem to also influence expression changes observed using in vivo models (TFs in red font in Supplementary Fig.S13). The link between tuberculosis and some of these TFs await further characterization. Some of the identified TFs have been associated with host response upon M. tuberculosis infection such PO2F12,64 SMAD1,65 RUNX2,65, 66 RUNX1,67 FOXP1,68 We identified functional processes potentially regulated by each TF. Additional experiments and comparisons are necessary to fully understand their implications in the host-pathogen dynamics. Overall, these results show that data collected with our collagen culture offers several functional insights for follow up characterization of the host-pathogen interaction.
The transcriptomic analysis of Mtb from the infected collagen gels identified 460 significantly differentially regulated genes (|log2fc|>1, FDR<0.05, 241 upregulated and 219 downregulated) when compared to the planktonic bacteria cultured in 7H9 medium; DEG distribution for the genes is shown in (Fig. 4A, Table S4). Functional categorization revealed that the upregulated genes are involved in crucial infection-related functions such as virulence and adaptation and cell wall processes, among others (Supplementary Fig. S14). Given the lack of robust Mtb transcriptomic data from human infections, the comparisons were made with observations from infected mice 9, which showed 114 genes to be similarly regulated in our hydrogels (55 genes upregulated and 59 genes downregulated) (Fig. 4B, Table S5). Mtb transcriptome from THP-1 macrophages infected for 24 h in 2D 69 showed only 21 genes (all upregulated) to be similarly regulated compared to mice infection (Fig. 4B, Table S5). This Mtb transcriptomic signature in 2D (from early time points compared to 3D) pointed towards the pathogen’s adaptation to the mammalian cells, wherein genes associated with zinc ion homeostasis, type VII secretion system, and membrane repair were significantly upregulated. Our analyses also revealed a large variation between the Mtb transcriptomes from 2D infected cells 69 and the collagen hydrogels, which could probably be attributed to the time spent by the pathogen in the host (24 h in 2D vs 14 days in collagen gels, Table S5). The upregulated DEGs shared only by the in vivo mice70 and the 3D culture models include virulence factors and other key genes in the adaptation and success of the pathogen in the host environment. We detected upregulation of ripA in our culture system; Mtb uses RipA to influence the host response upon infection to its advantage and long-term survival 71, 72. We also observed the upregulation of Rv1502 together with the transposon-sequencing data reported by Rosas-Magallanes et al.73 supports the identification of key virulence factors of Mtb using our 3D culture model. Furthermore, we detected upregulation of the alternative sigma factor sigD, which has been characterized as a key virulence factor 74. We also detected the upregulation of glfT1, which glfT1 is involved in cell wall remodeling and contributes to bacterial survival in the context of infection 75.
Fig. 4. Dual RNA sequencing reveals the transcriptomic similarity of bacteria from the collagen culture system with Mtb in infected mice.
(A) Volcano plot depicting differentially expressed genes in Mtb between planktonic bacteria and from the infected hydrogels (red: significantly upregulated, green: significantly downregulated, black: non-regulated) (n=3 per sample) (B) Quantification of the number of common genes between the in vitro models (3D collagen culture system, 2D THP-1 macrophages – GSE6209) and in vivo Mtb transcriptome (GSE132354) in mice (red: commonly upregulated, green: commonly deregulated). (C) Reconstruction of M. tuberculosis differentially active transcriptional networks observed in our 3D collagen culture system. The nine transcription factors (TFs) that contribute to the observed transcriptional changes in Mtb from the infected collagen gels were identified using the NetSurgeon algorithm 86 and a precompiled signed transcriptional regulatory network of M. tuberculosis, as explained previously 70. Only DEGs regulated by the shown TFs are displayed. Genes were considered differentially expressed based on |log2fc| ≥1 and FDR of ≤ 0.05.
We then looked at the functional analyses of the Mtb transcriptome in the model and were able to uncover TFs that potentially drive the transcriptional adaptation of M. tuberculosis during growth in our collagen culture model (Fig. 4C). We observed the downregulation of KstR2 regulon (i.e., upregulation of KstR2 repressed genes) which is indicative of the activation of cholesterol metabolism9, 70. This agrees with the expected in vivo use of host-derived cholesterol by the bacterium70, 76. Kstr2 regulon genes are involved in modulating host response to facilitate persistence 77 and fatty acid transport.78, 79 Intriguingly, we found the upregulation of the mce1 operon genes (mec1A, mce1B, mce1C, mce1F, mce1R), which are similarly regulated in vivo, but oppositely regulated in traditional 2D infection cultures. We also found decreased activity of the TF MadR and upregulation of its positively regulated targets (e.g., desA1 and desA2) which fits into a previously described infection adaptation model in which MadR activity is reduced during intermediate stress levels. desA1, which encodes a mycolic acid desaturase and is part of a MadR transcriptionally controlled cell wall remodeling program of M. tuberculosis, was detected as significantly upregulated in the 3D culture but only transiently detected in the BMDM model.70 Taking into account the importance of the MadR regulon for M. tuberculosis fitness (Fig. 4C),70, 80 this finding showcases the ability of the 3D culture model to identify changes in gene expression that contribute to the pathogen survival in the host environment. Furthermore, an increased activity of SigC, an alternative sigma factor was observed wherein the genes upregulated under this TF are important for Mtb virulence 81. The transcriptomic analysis indicated the upregulation of Zur, a zinc-related regulator, due to the downregulation of Zur-repressed genes and a similar regulation of Zur was also observed in an in vivo mouse model of infection.70 However, an opposite trend of Zur regulation was reported in experiments in in vitro BMDM infection model.70 This difference in Zur activity awaits further characterization. Moreover, we also found upregulation of genes involved in lipid degradation (lipL, plcC), TCA cycle, and fatty acid β-oxidation (icd1, Rv0893c, echA20), highlighting the dependence of Mtb on cholesterol and fatty acid utilization in our model, which is also a characteristic feature of the pathogen’s metabolism in the interstitial and alveolar macrophages, respectively, in vivo 9, 36. Moreover, Mtb has the ability to synthesize tryptophan, thus counteracting the attempts by T-cells to starve the pathogen of this essential amino acid 82. We also found upregulation of genes of the tryptophan synthase operon (trpA, trpB, trpC) in Mtb from our model, suggesting its probable adoption of tryptophan synthesis like the pathogen in in vivo infections. These results collectively indicate a robust functional relevance of the gene expression pattern of Mtb from our system with that from established in vivo infection models while highlighting how divergent the transcriptome profile is of conventional 2D systems for functional studies. Overall, the 3D collagen culture system accurately captured multiple mechanisms underlying Mtb infection-associated host-pathogen dynamics significantly better than the current in vitro platforms. To evaluate if our collagen system has hypoxia, mammalian as well as Mtb differentially expressed genes from our RNA-seq data were checked for overlap with hypoxia-associated gene lists derived from studies that looked at the transcriptional adaptation of mammalian and Mtb genes under hypoxic stress 83, 84. Only one mammalian gene (IRAK3) overlapped with genes regulated by HIF (out of 23) and was upregulated in our system. This gene is known to be a modulator of inflammation and can be induced by transcription factors including activator protein-1 (AP1) and glucocorticoid receptor (GR) apart from hypoxia-inducible factor-α (HIF-α)85 (Table S6).
Mtb differentially regulated genes from the culture system were checked for overlap with the genes of the DosR regulon84 and there was an overlap for six genes. However, it was found that these six genes were downregulated in our system while they are known to be upregulated under hypoxia, confirming the lack of any major hypoxic stress on the bacteria in the system (Table S7).
4. THP-1 cells accumulate lipid bodies in response to infection in the 3D collagen culture system
Mtb can persist in the host for long durations and can switch the energy source required for its metabolism 87 88. In the initial stages of infection, the pathogen relies on glucose and fatty acid metabolism as its energy source; however, as the infection progresses, fatty acids and cholesterol become the primary energy source 87. Thus, Mtb induces metabolic reprogramming in the macrophages to accumulate triglycerides, leading to the formation of foamy macrophages, a hallmark of human primary granuloma 88. Hence, we focused on deciphering the Mtb-mediated metabolic alterations to compare how the host-pathogen interactions in our system and in vivo infections are similar in terms of energy utilization. To capture such changes, BODIPY (493/503) was used to visualize and quantify the neutral lipids across systems 89. In a 2D system, owing to the shorter duration in which infection studies can be carried out, the increase in the accumulation of lipids by the infected mammalian cells is not significant (Supplementary Fig. S15), as seen by no difference in the BODIPY signal between the uninfected and infected conditions. However, BODIPY staining of uninfected and infected collagen gels revealed that Mtb infection resulted in a significant accumulation of lipid bodies, as evidenced by more green fluorescence in Fig. 5B (infected) compared to Fig. 5A (uninfected). Infected gels had a 3 to 4 folds higher signal on days 7 and 10 post-infection compared to uninfected gels (Fig. 5C). Additional confirmation by flow cytometry corroborated these results (Fig. 5D). Upon further quantification of BODIPY fluorescence in the infected gels, we observed a higher intensity of the signal in the infected cells with respect to the bystander cells (Fig. 5E). Thus, our collagen culture system resulted in the formation of foamy macrophages similar to in vivo infections. Interestingly, the BODIPY fluorescence in the bystander cells in the infected gels was still significantly higher than cells from uninfected gels, indicating priming of the bystanders to form lipid bodies by a microenvironment-directed mechanism in the Mtb-infected collagen gels (Supplementary Fig. S16). These findings led us to investigate whether our dual-RNA sequencing data reflected the same phenomena. Transcriptomic data from human caseous granuloma (GSE20050) showed upregulation of 55 genes involved in lipid metabolism 22, and nearly one-third of these genes were also upregulated in our culture system (Fig. 5F); this is in contrast to the negligible overlap with the current in vitro 2D models (GSE162729 – THP1 macrophages infected with Mtb in 2D – 1 gene in common; GSE174566-PBMCs infected with Mtb in 2D – zero). This further established transcriptome-guided phenotypic similarities of our system with in vivo infections.
Fig. 5. Cells in the infected collagen gels show accumulation of lipid bodies and differentiate to macrophages:
Representative fluorescence microscopy images of uninfected gels (A) and gels infected with tdTomato expressing Mtb (red) (B) that were stained with the nuclear stain DAPI (blue) and BODIPY (green), 10 days post-infection. (C) Quantification of BODIPY staining per mammalian cell (THP-1 cells) in uninfected vs infected collagen gels (n = 3) (D) Mean fluorescence intensity (MFI) of BODIPY staining of THP-1 cells from uninfected gels (Day 10) and infected gels (Day 10) quantified using flow cytometry (E) MFI of BODIPY staining of bystanders and Mtb infected THP-1 cells from infected gels quantified using flow cytometry. (F) Heat map of the average z-scores, showing the list of genes involved in lipid metabolism (common with a human caseous granuloma). Data in graphs represent the mean ± s.d., and p values were determined by two-tailed unpaired t-tests with Welch’s correction for C and without Welch’s correction for D & E using GraphPad Prism software. *p-value < 0.05 was considered significant. Day 7 (*p=0.012) and Day 10 (*p=0.026) for C, *** p=0.0006 for D, * p=0.0348 for E. Experiments related to figure panels D and E were repeated three times with similar results.
5. Pyrazinamide shows efficacy at or below the clinically relevant concentration in the 3D collagen culture system
Pyrazinamide (PZA) has been a mainstay of antituberculosis treatments since the 1950s and played a major role in shortening the therapy from 12 months to 6 months.90, 91 However, remarkably, it shows no significant activity in any in vitro mammalian cell infection model at clinically relevant concentrations.92, 93 and its anti-tubercular activity was discovered directly on an animal model of infection,94, 95 highlighting how the current in vitro mammalian culture systems are not optimal for drug testing.
We found that PZA consistently reduced the bacterial burden in our culture system, evidenced by the decrease in the intracellular bacteria (green, mammalian cell boundary stained with phalloidin (red)) in the treated gels compared to the untreated gels (Fig. 6A, compare the left panel vs right) and also by the reduction in CFUs at 50 μg/mL (Fig. 6B), which is in the range of average peak plasma concentration (Cmax) in individuals undergoing TB treatment with PZA. 96 We further validated this finding using primary human monocytes isolated from human PBMCs, which also showed a reduction in CFU upon PZA treatment (Fig. 6C). However, at this concentration, PZA failed to act in 2D infection models (Fig. S17A) and was also ineffective on planktonic bacteria and bacteria embedded in gels without mammalian cells (Supplementary Fig. S17B-C). A previous study using collagen-alginate matrix showed PZA efficacy, but the drug concentration was 500 μg/mL, ten times higher than Cmax.97 Furthermore, we carried out a dose-response-based study to check the concentration at which PZA becomes effective and observed a significant CFU reduction even at 10 μg/mL of the drug concentration (Fig. 6D). The efficacy of PZA at a physiologically relevant concentration in our 3D collagen model reiterates the similarity of our infection model with in vivo conditions, further establishing its potential as a more reliable drug testing platform.
Fig. 6. Pyrazinamide showed efficacy at a clinically relevant concentration in infected collagen gels:
(A) Representative fluorescence images of sections from Mtb infected THP-1 in collagen gels treated with PZA at 50 μg/mL (right) with respect to the vehicle-treated control (left). In these gels, the infection was carried out for 4 days, post which the experimental groups were treated with PZA at 50 μg/mL for 6 days. Intracellular CFU plots obtained after treating (B) Mtb-infected THP-1 in collagen gels (n = 4), the treatment with PZA was started 7-days post-infection and was carried out for 3 days; hence the CFUs shown are 10 days post-infection (C) Mtb infected primary human monocytes (n = 4), the treatment with PZA was started 7-days post-infection and was carried out for 3 days; hence the CFUs shown are 10 days post-infection. (D) Intracellular CFU plots were obtained after treating Mtb infected THP-1 in collagen gels (n = 4 for untreated and n = 3 for other groups), the treatment with PZA was started 4-days post-infection and was carried out for 6 days; hence the CFUs shown are 10 days post-infection. (E) Mtb (POA resistant strain) infected THP-1 cells with PZA at 50 μg/mL (ns: non-significant) (n = 3), the treatment with PZA was started 4-days post-infection and was carried out for 6 days; hence the CFUs shown are 10 days post-infection. The mutations are listed in Table S6. Data in graphs represent the mean ± s.d., and p values were determined by two-tailed unpaired t-tests with Welch’s correction for B&C and Tukey’s multiple comparison tests for D using GraphPad Prism Software. p-value < 0.05 was considered significant. **p=0.0089 for B and *p=0.0171 for C; *p=0.0068 for untreated vs 2 μg/mL, *p=0.0002 for untreated vs 10 μg/mL, *p<0.0001 for untreated vs 50 μg/mL, *p<0.0001 for untreated vs 150 μg/mL for D, ns = non-significant for E. Experiments were repeated four times for B and twice for C with similar results.
Given the exclusive effectivity of pyrazinamide in vivo, it has been difficult to understand its mechanism of action fully, and it is only in the recent past that some insights have emerged. 15, 98–102 Pyrazinamide is a prodrug catalyzed to its active form pyrazinoic acid (POA), by the bacterial gene pncA.103 One of the hypotheses of PZA action points to its ability to kill the bacteria by perturbing the intrabacterial pH.104 Briefly, PZA gets metabolized to POA inside the bacteria and is transported out through the efflux pumps. However, in an acidic pH microenvironment (such as in a phagosome), POA gets protonated to HPOA, which then re-enters the bacteria. This cycle eventually leads to the bacteria being unable to maintain its cytoplasmic pH and membrane potential, leading to death. We tested this hypothesis by checking for phagosomal maturation (indicating low pH) in the infected 2D and 3D conditions. Lysosomal staining using LAMP-1 to quantify the colocalization with Mtb did not reveal any significant differences in the 2D infection scenario compared to 3D (Supplementary Fig. S18). We also compared the Mtb genes from our 3D collagen culture system and THP-1 macrophages (2D infection - GSE6209) to those induced by low pH as an in vitro stimulus in planktonic culture.105 This analysis showed 16 Mtb genes in our 3D collagen culture system to be commonly upregulated with those induced by low pH in planktonic culture compared to 19 from the 2D infection, further pointing to the relatively similar pH stress in our system and 2D infection (Table S8). These results ruled out the possibility of the pH-dependent effectivity of PZA in our model, prompting further mechanistic investigations.
Although pncA mutations are responsible for PZA resistance in clinical samples, other mechanisms have also been shown.106–108 A few studies have highlighted the critical role of Aspartate Decarboxylase (panD) and the Unfoldase (clpC1) genes in mediating resistance to PZA. 100, 101 PanD is involved in the synthesis of pantothenic acid (vitamin B5), critical for biosynthesis of coenzyme A and acyl carrier protein (ACP), which regulates various steps of lipid metabolism, and hence, bacterial deletion mutants of these genes were found to be highly attenuated.109 The active form of the drug, POA, binds to PanD and the C-terminal of the PanD-POA complex is recognized by the caseinolytic protease, ClpC1-ClpP, leading to its degradation.98 To determine whether the PZA-PanD axis was the mechanism of action of the drug in our system, we generated four spontaneous POA-resistant mutants of H37Rv (labeled mutant A to D) as previously described 100 and corroborated that mutations were present in the panD gene corresponding to the C-terminal domain of PanD (Table S9); pncA and clpC1 genes did not have any mutations in these bacteria. We infected THP-1 cells with these POA mutants in the collagen gels and treated the experimental groups with PZA (50 μg/mL) (Fig. 6E, Supplementary Fig. S19A-C), which was ineffective in reducing the bacterial burden for all 4 mutants. This clearly emphasized the critical role of PanD in the efficacy of PZA in our 3D model, strengthening the case for adopting this culture system in drug discovery and mechanistic studies.
Several other Mtb proteins like Fad2, SigE, Mas, etc., are hypothesized to synergize/antagonize PZA activity,110, 111 however, the associated genes were not differentially regulated in our collagen culture system compared to planktonic cultures (Table S10). Thus, the transcriptomic data suggest that these proteins may have a minor role in PZA efficacy in our culture system, although there may be differences at the translational level. Further validation of the system was done using the other first-line anti-TB drugs (rifampicin, ethambutol, and isoniazid), which were also effective in reducing bacterial counts when used at their respective minimum inhibitory concentrations (MIC) of broth culture (Supplementary Fig. S20A-C). These results collectively demonstrated the suitability of our biomaterial-based system in mimicking the in vivo drug efficacies.
Discussion
Since tuberculosis causes a significant social and economic burden on both patients and society, there is a need for active research to understand the bacteria-host interaction and find new drug candidates. Although experiments on planktonic cultures of Mtb and in vitro 2D infection experiments with mammalian cells continue to be the primary step for many studies, their inherent limitations, such as short infection durations and non-physiological microenvironments, often lead to findings that do not translate well to animal models and humans. Such efforts consume precious time and resources or could also curtail the development of effective therapies, warranting the need to develop better drug-testing platforms. In this study, we propose a 3D collagen-hydrogel culture system as a next-generation in vitro platform to delineate the complexities of Mtb infection by mimicking an in vivo system and be a reliable drug-testing platform. 3D culture platforms such as static microphysiological systems, organ-on-a-chip systems, and organoid cultures are being used to study many diseases and are gaining wide acceptance.19 The current culture system uses hydrogels made of collagen 1, the main ECM component of human lungs. ECM plays a fundamental role in TB pathogenesis during human infection, as Mtb is known to induce the production of matrix metalloproteases (MMPs), which digest collagen, leading to ECM degradation and disease progression.112 Hence, inhibitors that act on these proteases are being studied as potential supplements to the existing treatment regimens.113 The drawbacks of the current model systems include a lack of ECM in 2D infection studies and human orthologs of some MMPs in mice.114 A previous study cultured PBMCs with Mtb in a 3D matrix to form micro aggregate structures of cells of nearly 100 μm surrounding an infected cell or bacteria 115 and revealed the critical role of collagen in reducing the mammalian cell death and Mtb growth.21, 116 Primary granulomas from zebrafish have also been cultured ex vivo.117 More advanced 3D systems include engineering lung organoids using stem cells, though its potential is currently unexplored in TB research.10 A recent study has used human airway organoids to study mycobacteria interaction with airway epithelial cells.118 Furthermore, an in vitro hollow fiber system for TB (HFS-TB) was used to recapitulate the human pharmacokinetics and pharmacodynamics of a few TB drugs, resulting in optimal dosing regimens for moxifloxacin and linezolid.119 Although all these platforms add significant value to the field, their use is still limited due to difficulties in procuring primary lung granulomas, lung biopsies, and specialized equipment in BSL-3 containment. Even though these systems are primarily used to test drug efficacies, none has shown the efficacy of pyrazinamide, a critical first-line anti-TB drug, at clinically relevant concentrations.
The collagen gels used in this study are easy to synthesize, are widely used to mimic in vivo cellular microenvironments,120 and are easy to work with even in BSL-3 facilities. These collagen gels are modular, for example, the matrix characteristics, such as pore size, stiffness, etc., depend on polymerization temperature and can be tuned in accordance with the human lung ECM and have been extensively characterized.121
THP-1 monocytes and BMDMs are commonly used for studying host-pathogen interaction in vitro, but the short duration of these studies continues to be a major challenge. BMDM infection offers a slightly longer duration to study host-pathogen interaction elucidated by the transcriptional and physiological profile of Mtb in a BMDM infection model over 14 days, which showed the adaptation of the pathogen in the face of host stress.13 However, the kinetics of the BMDMs response in terms of viability varies greatly with MOI, with approximately 75% at day 6, with an MOI of 1, which falls to less than 25% with an MOI of 5.122 In contrast to these systems, our culture system overcomes the shortcomings of the mentioned studies, offering the flexibility of looking at host-pathogen interaction at multiple levels such as time (1 day – 3 weeks) and scale: single cell to multiple cells behavior for mammalian and Mtb; and is suitable for analysis through tools such as genetic mutations in bacteria and host, histology, fluorescence microscopy, CFU assays, live-cell imaging, and flow cytometry.
We also adopt unbiased high throughput analysis to establish the genotypic and phenotypic similarity of mammalian cells in the infected collagen gels to human infection. For example, Mtb is known to induce the formation of lipid bodies in macrophages in vivo. 22, 123 Previously, it has been shown that 2D infection of human macrophages can lead to a high accumulation of lipid bodies compared to uninfected macrophages,124 however, more recent reports indicate necrosis of other cells in the milieu to be critical for this phenomenon.125 An MOI of 1 or 5 in 2D cultures did not result in foamy macrophage formation but instead was seen with a high MOI of 50, which led to more than 60 % mammalian cell death.125 The necrotic cells served as a source of triglycerides for the live infected cells. Other commonly used methods include supplementation of oleate in the growth media to generate foamy macrophages, followed by infecting these cells with Mtb to study lipid retention/turnover.126, 127 Another model used PBMCs exposed to sepharose beads coated with Mtb purified protein derivative (PPD) to show aggregate formation accompanied by foamy macrophages.128 Mtb infection of murine adipocytes 129 or supplementation with lipoproteins 130 have also been tried to induce the formation of foamy macrophages. Furthermore, human macrophages differentiated in 2D have a propensity to accumulate lipids even in the absence of Mtb infection.22 In our 2D infection studies, a low MOI of 1 (leading to nearly 10% infection) did not induce significant necrosis, which in turn did not lead to an increase in the lipid content in the infected group compared to the uninfected group. In contrast, in our collagen culture system, even a low burden of Mtb infection led to the increase in lipid content in the mammalian cells without any additional supplementation of fatty acids, suggesting Mtb-mediated metabolomic reprogramming similar to in vivo infections. We also show cord-like Mtb replication in the infected collagen gels, another prominent phenotype found in vivo. 23, 24, 54, 55 Furthermore, in our Mtb transcriptomic data, upregulation of the gene fbpC, a member of the antigen 85 family of proteins that confer fibronectin binding affinity and maintain the cell wall integrity of Mtb by synthesizing alpha trehalose-dimycolate (TDM, cord factor),131 elucidated the molecular basis of cording in our system. Host cell autophagy is an important defense mechanism against intracellular pathogens, and the upregulation of autophagic genes and pathways in our dual RNA sequencing data points towards such a phenomenon also occurring in our system, which the bacteria successfully evade by cording.57 Transcriptomic analysis of the mammalian cells in the infected collagen gels revealed a significantly greater overlap with human TB granulomas than in the current 2D culture systems. Such shortcomings in how mammalian cells in 2D respond to Mtb infection render them somewhat unreliable systems for testing potential drug candidates. For example, none of the current in vitro systems could demonstrate significant PZA efficacy at clinically relevant concentrations (≤ 50 μg/mL), which could be observed using our 3D collagen culture system. PZA efficacy in vitro has been shown at an MIC of 200 μg/mL on planktonic bacteria at a pH of 5.95 104. Further reduction in the pH of the medium in which these planktonic bacteria are cultured resulted in growth inhibition of several strains 104. Similarly, in human monocyte-derived macrophages, treatment with concentrations as high as 100-400 μg/mL of PZA only resulted in about 70% reduction in bacteria as determined by fluorescence imaging 102, 132. Thus, our platform is potentially an excellent substitute for current in vitro models and a starting point for more accurate drug testing and to give deeper insights into host-pathogen interaction before large-scale animal studies.
Our study has a few limitations, such as the tuberculosis bacilli in a host are taken up by monocytes, macrophages, dendritic cells, and neutrophils 7 and also have a niche in mesenchymal stem cells and lymphatic endothelial cells 57, 133. An MOI of 40:1 was used to establish successful infection in our culture system and we hypothesize that a higher MOI was required as monocytes take up bacteria with lesser efficiency compared to macrophages. Monocytes in 2D culture do not take up bacteria as much as they do in 3D, as can be seen in Supplementary Fig. S5, which supports that monocytes are less phagocytic compared to macrophages. However other effects of using a collagen matrix such as change in cell phenotype, higher bactericidal activity, etc can’t be ruled out. Furthermore, the THP-1 monocytes may proliferate in the culture system and can result in a low number of infected cells.
We found that for both THP-1 monocytes and primary human monocytes derived macrophages, the cell viability was much higher in our 3D gels as compared to 2D cultures. PZA was also efficacious when used on Mtb-infected primary human monocytes. However, other observations regarding mammalian and bacterial phenotypes need to be validated with primary cells. Another limitation is that the culture system does not incorporate other immune and non-immune cells that are relevant in vivo 134–136. During infection, Mtb primarily resides in alveolar macrophages (AM) and interstitial macrophages (IM) in vivo and has distinct interactions with each of these macrophage populations 9; AMs offer a more permissive niche for bacterial growth compared to IMs 36, and hence, it would be interesting to see how AMs respond to infection in our collagen hydrogels. However, there are no established human AM cell lines, and primary AMs from humans are difficult to obtain as they require access to bronchoalveolar lavage (BAL). Another limitation of the study is the use of publicly available datasets instead of generating them in-house. However, it is difficult to access the primary tuberculosis samples from hospital settings given the infectious nature of the samples. Furthermore, the bacteria encounter multiple stresses in vivo, such as adaptive immune response and hypoxia in a necrotic granulomatous core, which was not present in our platform. The absence of other components of the adaptive immune system also means that Mtb can continue to proliferate in the macrophages as the duration of the infection progresses. It is possible to increase the density of cells cultured in collagen gels, which may lead to low oxygen concentration near the center of the gel, as shown in spheroid cultures 137 and will be explored in future studies.
Conclusions
We show that our 3D collagen cell culture system offers a significantly longer window for studying host immune cell-pathogen interaction. Immune cells in the infected microenvironment have properties similar to cells in vivo, in terms of lipid accumulation, cord formation, and gene expression, among other characteristics. The effectivity of pyrazinamide, a first in any in vitro model, in reducing the bacterial burden further emphasizes the robustness of the platform, and we hypothesize that potentially undiscovered mechanistic phenomena can be unraveled using our culture system. The system offers utility with ease of usage and modularity while incorporating resources that are easily accessible to the research community and can, therefore be readily adopted.
Methods
1. Culture of mammalian cells and isolation of primary human monocytes
Human THP-1 monocytes were seeded at a concentration of 105 cells/mL and passaged every 3 days. The cells were cultured in RPMI 1640 (Invitrogen) media containing 10% Fetal bovine serum (FBS) and antibiotics (Penicillin/Streptomycin).
For human primary monocytes, 10 mL of whole blood was collected from volunteers (Institutional Human Ethics Committee Approval Number: 20-14012020), and peripheral blood mononuclear cells (PBMCs) were isolated by density gradient centrifugation over Histopaque®-1077 (Sigma Aldrich). From these PBMCs, the primary monocytes were isolated using a monocyte isolation kit (Mojosort™ Human CD14+ Monocytes Isolation Kit, BioLegend®), per the manufacturer’s instructions. These CD14+ monocytes were cultured in RPMI 1640 media supplemented with 10% human serum (H4522, Sigma Aldrich) in non-adherent dishes for 6 days to allow their differentiation to macrophages, post-which the macrophages were detached from the plate using 5 mM EDTA in ice-cold PBS, and these were further used for experiments in 2D by plating in 48-well plates at 105 cells/well or in 3D.
All mammalian cells were cultured in a humidified incubator supplemented with CO2 and maintained at 37 °C.
2. Culture of H37Rv
Mycobacterium tuberculosis strain H37Rv was cultured in media composed of 7H9, tween 80 (0.05%), glycerol (0.1%) with albumin-dextrose-catalase (ADC) (10%) enrichment at 37 °C. GFP-H37Rv carrying the plasmid pmyc1tetO-gfp, tetR(B), and tdTomato H37Rv carrying the plasmid pTEC27 (pMSP12: tdTomato), with hygromycin (50 μg/mL) as the selection marker, were used for imaging and flow cytometry experiments. Mtb H37Rv-pBP10 was cultured similar to above with 30 μg/mL kanamycin.
3. Collagen gel preparation
Rat tail Collagen I (3 mg/mL, Gibco™) was used. Collagen gels with final concentrations of 1 mg/mL to 2 mg/mL were made as per the manufacturer’s recommendations. The pH of these gels ranged from 7-7.4. The polymerization was carried out in a humidified incubator at 37 °C for 20 minutes, following which they were transferred into a 24-well plate containing 1X phosphate-buffered saline (PBS). In experiments involving mammalian cells (THP-1 or primary macrophages) and bacteria, the cell suspension (4x106 cells/mL) was added in place of 1X PBS, and the gels were transferred to wells containing RPMI 1640 containing 10% FBS. Each collagen gel had 2x105 cells and was 50 μL in volume with a diameter of approximately 2.5 mm.
4. Infection experiments
Infected collagen gels were made as follows: single-cell suspension of H37Rv (OD of 0.5-1) was prepared and added to THP-1 monocytes at a multiplicity of infection (MOI) of 40:1 (bacteria: mammalian cells) in the required volume for making the gels. Post gelation, the infection was carried out for 12 h, after which amikacin sulfate (200 μg/mL) was added to the media for 6 h to remove extracellular bacteria. The media was changed every 48 h during the experiment. The gels were treated with collagenase, type 1 (0.4 mg/mL, GIBCO™) for 1 h at 37 °C to degrade the collagen for flow cytometry experiments. The cells were stained with LIVE/DEAD™ (L34955 LIVE/DEAD™ Fixable Violet Dead Cell Stain Kit, for 405 nm excitation), followed by fixation with 4% paraformaldehyde (PFA), and used for further analysis.
For 2D infection experiments, THP-1 cells were differentiated into adherent macrophages using 20 nM Phorbol 12-Myristate 13-Acetate (PMA). Briefly, cells were incubated in media containing PMA for 24 h, post which the PMA was removed, and the cells were allowed to rest for 2 days with fresh media. Cells were then infected with Mtb at an MOI of 1:1 (bacteria: mammalian cells) for 4 h and treated with amikacin sulfate (200 μg/mL) for 2 h.
For experiments using primary human monocyte-derived macrophages, an MOI of 1 was used, and infection in 2D and 3D was carried out similarly to THP-1 cells as described above.
We also carried out 2D infection experiments on collagen coated plates for which we coated well-plates with collagen. Briefly, collagen I (3 mg/mL) was diluted to a final concentration of 50 μg/mL in 0.02M acetic acid on ice and added to the well plates such that the surface was sufficiently covered. The plates were kept at 4° C overnight and excess collagen was removed. The plates were then air dried and washed with sterile distilled water twice and air dried again. Collagen coated plates were stored at 4° C. Following this, THP-1 cells were co cultured with Mtb H37Rv (MOI of 40:1) on 2D collagen coated plates for 12 h, followed by amikacin sulfate (200 μg/mL) treatment for 6 hours. The samples were processed stained with LIVE/DEAD™ (L34955 LIVE/DEAD™ Fixable Violet Dead Cell Stain Kit, for 405 nm excitation), followed by fixation with 4% paraformaldehyde (PFA), and analyzed using flow cytometry.
10kDa PEG-MAL (4-arm PEG-MAL, Creative PEG works) was used to make hydrogels with a final PEG concentration of 4%. VPM (GCRDVPMSMRGGDRCG) (Genscript), which is a matrix metalloproteinase (MMP) degradable peptide, was used as a crosslinker at a 2:1 ratio of PEG-MAL: VPM. The solution was mixed with collagen type I such that the final concentration of collagen in the hydrogel was 0.6 mg/mL. 10X PBS was used as a buffer to make up the final volume. A mixture of THP-1 cells and Mtb at a multiplicity of infection of 40:1 (bacteria: cells) was suspended in the buffer and added to the mixture of PEG-VPM-collagen and polymerized. The infection was carried out for 12 h, followed by the amikacin sulfate treatment for 6 h to kill the extracellular bacteria. The cells were then immediately released from the hydrogel using 0.5 mg/mL collagenase, stained with LIVE/DEAD™ (L34955 LIVE/DEAD™ Fixable Violet Dead Cell Stain Kit, for 405 nm excitation), followed by fixation with 4% paraformaldehyde (PFA), and analyzed using flow cytometry.
5. Pimonidazole staining
To check if the gel develops hypoxia overtime, THP-1 monocytes and Mtb H37Rv were co-cultured in the collagen hydrogels for 14 days. As a positive control for hypoxia staining, HEK cells were grown into spheroids on agar-coated plates for 2 days. The spheroids and infected collagen gels were treated with 150 μM pimonidazole hydrochloride (Hydroxyprobe™) at a concentration of 150 mM for 12 hours, followed by fixation with 4% paraformaldehyde. The spheroids and infected collagen hydrogels were then cryosectioned and stained with MAb1 (mouse monoclonal antibody, 1.2 μg/mL), followed by a suitable secondary antibody, and visualized under a confocal microscope.
6. Mtb growth kinetics
To study the kinetics of Mtb replication during infection in the hydrogel system, we infected THP-1 cells with H37Rv carrying the replication clock plasmid, pBP10 45 (a kind gift from Dr. David Sherman). Intracellular CFUs were carried out by treating the gels with collagenase, lysing the mammalian cells with 0.1% Triton X-100, followed by spotting serial dilutions on Middlebrook 7H11 agar plates supplemented with 10% OADC, with and without 30 μg/mL kanamycin and incubated at 37 ºC for 20-24 days. The replication and death rates of Mtb were quantified using the mathematical model proposed by Gill et al.45.
7. Second-harmonic imaging of collagen gels and lung tissues from mice
Second-harmonic generation (SHG) images of the samples were acquired using a mode-locked fiber laser (Coherent Fidelity HP) with an operational wavelength of 1040 nm, a pulse width of 140 fs, and a repetition rate of 80 MHz. The incident beam was scanned using a galvo-scanner (GVS001, Thorlabs), with the beam focused on the sample using a 60x/1.2 NA water immersion objective (UPLSAPO60XW, Olympus). A photomultiplier tube (R3896, Hamamatsu) was used to collect the SHG emission at 520 nm from collagen. The incident power at the focus for imaging collagen was 1.7 mW. The average pore size of the collagen gels was calculated using the particle analyzer in the BoneJ plugin of ImageJ (NIH, USA). The resolution of the microscope was around 400 nm, and hence the pores below the size of 0.16 μm2 were excluded from the analysis.
8. Atomic Force Microscopy (AFM) measurement
Collagen gels were loaded into custom polydimethylsiloxane (PDMS)-based AFM holder. The sample was kept hydrated during measurement, and indentation (with a depth of 500 microns) was done at different points using Park Systems NX-10 AFM in a liquid medium. Force distance spectroscopy was carried out using the HYDRA6V-200NG-TL cantilever with a 5 μm diameter sphere. The force constant was 0.045 N/m, and the force limit was 0.91 nN. The up and down speed was 0.3 μm/s. The data generated were further analyzed through a custom-written MATLAB code to determine Young’s modulus 31.
9. Ferret diameter calculation of Mtb cords
THP-1 cells were infected with GFP-H37Rv in collagen gels for 10-14 days, post which the gels were fixed with 4% paraformaldehyde (PFA). The gels were stained with Hoechst 3342 (Thermo Scientific™) and Phalloidin (A12380, Thermo Scientific™) and imaged using a confocal microscope (SP8, Lecia Microsystems). The GFP signal from H37Rv from 5 images was thresholded, and only aggregates above 5 μm2 were used for further analysis to filter out the single bacteria/noise. The ferret diameter was then calculated using the Analyze option in ImageJ.
10. TB antibiotic treatments
THP-1 cells were infected with Mtb in collagen gels for 7 days, post which the gels in the treatment group were treated with the first line of antituberculosis drugs – rifampicin (Himedia®) (0.3 μg/mL), isoniazid (Himedia®) (0.25 μg/mL), and ethambutol (Himedia®) (5 μg/mL). At 3- and 7-days post-treatment, gels were degraded with collagenase, and cells were lysed with 0.1% Triton X-100. Cell lysates were plated on 7H11 plates supplemented with 10% OADC and incubated at 37° C. The CFUs were counted 3-4 weeks after plating. For studying pyrazinamide (Himedia®) (50 μg /mL) efficacy, the infection was carried out for either 4 or 7 days, post which the cells were treated for 6 or 3 days, respectively.
11. Generation of pyrazinoic acid (POA) mutants
The H37Rv POA mutants were generated as described by Pooja et.al. 98. Briefly, Mtb was grown on 7H11 plates supplemented with 2 mM POA and 10% OADC, and a few colonies were picked and restreaked on fresh 2 mM POA plates. Colonies from 2nd plate were further grown in 7H9 media supplemented with 2 mM POA and 10% ADC. Genomic DNA was isolated from these mutants and sequenced with gene-specific primer pairs (Table S11) for pncA, panD, and clpC1 genes using Phusion™ High-Fidelity DNA polymerase (Thermo Scientific™) as per the manufacturer’s instructions. The products were sequenced using Sanger sequencing and the results were analyzed using the NCBI BLAST tool.
12. Antibody & BODIPY staining
Differentiation of monocytes to macrophages was quantified by CD68 expression. Briefly, cells in the infected and uninfected hydrogels were released using collagenase (0.4 mg/mL). The cells were stained with LIVE/DEAD™ (L34955 LIVE/DEAD™ Fixable Violet Dead Cell Stain Kit, for 405 nm excitation), post which they were fixed using 4% PFA. The cells were then permeabilized using 0.1% Triton X-100 and stained with anti-CD68 antibody (PE-Mouse Anti-Human CD68 set, BD Biosciences) (1:50 dilution) overnight in a buffer made of 1X PBS supplemented with 1% bovine serum albumin (BSA).
To study the activation of the THP-1 cells post-infection, cells were stained with an anti-iNOS antibody (NB300-605 – Novus Biologics). Briefly, at day 10, cells in the infected and uninfected hydrogels were released using collagenase (0.4 mg/mL). The cells were stained with LIVE/DEAD™ (L34955 LIVE/DEAD™ Fixable Violet Dead Cell Stain Kit, for 405 nm excitation), post which they were fixed using 4% PFA. The cells were incubated in a blocking buffer made of 1X PBS supplemented with 1 % BSA, 0.1% Triton X-100 and 5% donkey serum for two hours at room temperature. This was followed by washing with 1X PBS and staining with the primary antibody (iNOS - 7 μg/mL) diluted in FACS buffer and incubation at 4 °C overnight. This was followed by staining with the secondary antibody (Donkey anti-Rabbit IgG (H+L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor™ Plus 647) diluted in the blocking buffer (0.5 μg/mL) for 3 hours, followed by running the samples through a flow cytometer.
To study the localization of Mtb phagosomes with lysosomes in the infected 2D and 3D scenarios, cells were stained with anti-(LAMP-1) antibody (ab24170, Abcam). Briefly, the cells in 2D were incubated in a blocking buffer made of 1X PBS supplemented with 1 % BSA, 0.1% Triton X-100 and 5% donkey serum for two hours at room temperature. The cells were then washed with PBS and stained with the primary antibody (0.5 μg/mL) diluted in the blocking buffer and incubated at 4 °C overnight. This was followed by staining with the secondary antibody (Donkey anti-Rabbit IgG (H+L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor™ Plus 647) diluted in the blocking buffer (1 μg/mL), for 3 hours and imaged at 60X using a fluorescence microscope (IN Cell Analyzer 6000). For LAMP-1 staining in hydrogels, the gels were fixed with 4% PFA, post which cryo-sectioning 138 was carried out. The sections were stained with primary and secondary antibodies similar to 2D cultures. For the 2D infection, analysis was done on days 3 and 5 post-infection, while for 3D, it was done on days 5 and 7 post-infection. The images were acquired with a z-stack of 0.5 slice thickness over 20 μm. The co-localization of Mtb and LAMP-1 in the images was analyzed using the Just Another Co-localization Plug-in (JACoP) in ImageJ. Briefly, the intensities of the fluorescence signals (green – Mtb and red-LAMP1-1) were thresholded, and the Pearson’s coefficient from the plug-in was used to calculate the extent of co-localization.
For staining the mammalian cells in 2D and in the hydrogels with BODIPY™ 493/503 (Thermo Scientific™), the cells were fixed with 4 % PFA for 1 hr, followed by incubation with the dye diluted in 1X PBS (1 μg/mL) for 15 mins at 37 °C. The dye was washed, followed by staining with Hoechst 3342 (Thermo Scientific™) and imaged using a fluorescence microscope (IN Cell Analyzer 6000) with a 40X objective. The intensity of the fluorescence signal (green – BODIPY) was thresholded, and the area was calculated using the Analyze Particles option in ImageJ. The obtained BODIPY area per field was then divided by the number of mammalian cells in the field to arrive at the BODIPY area per cell in the field. Analysis was carried out for 10 fields per gel. For flow cytometric quantification of BODIPY, the gels were degraded with collagenase, type 1 (0.4 mg/mL, GIBCO™) for 1 h at 37 °C. These cells were stained with LIVE/DEAD™ (L34955, Thermo Fisher Scientific), followed by fixation with 4% PFA. The cells were then stained with BODIPY, as described above.
For all the flow cytometry experiments, the cells were suspended in 300 μL of FACS buffer, and data were acquired using BD FACSDiva™ software, followed by analysis using FlowJo (v9). The gating strategy to identify the suitable cell populations for CD68, iNOS and BODIPY staining is shown in Supplementary Fig. S21, and the viability assay is in Supplementary Fig. S22.
13. Preparation of samples for dual RNA-sequencing
The gels were degraded using collagenase (0.4 mg/mL) at day 14. The cells were treated with Guanidinium thiocyanate buffer (GTC buffer) containing beta-mercaptoethanol. The homogenate was centrifuged at 5000 g resulting in pellets of Mtb. The supernatant was collected and processed for THP-1 RNA extraction. Mtb RNA was isolated from the pellet using FastRNA® Pro Blue kit (MP Biomedicals™) per the manufacturer’s recommendations. Briefly, 1 mL of RNApro™ solution was added to the bacterial pellet and was transferred to a lysing matrix tube. Bead beating was carried using a FastPrep® instrument (2 cycles of 20 seconds each at a speed setting of 4.5 m/s). The isolated RNA was purified using a NucleoSpin RNA kit (MACHEREY-NAGEL) per the manufacturer’s recommendations and quantified using a NanoDrop™ One/OneC Microvolume UV-Vis Spectrophotometer (Thermo Scientific™).
14. Bioinformatic pipeline and analysis
The libraries were paired-end sequenced on Illumina HiSeq X Ten sequencer for 150 cycles (Illumina, San Diego, USA) following the manufacturer’s instructions. The data obtained from the sequencing run was de-multiplexed using Bcl2fastq software v2.20, and FastQ files were generated based on the unique dual barcode sequences. The sequencing quality was assessed using FastQC v0.11.8 software. The adapter sequences were trimmed, bases above Q30 were considered, and low-quality bases were filtered off during read pre-processing and used for downstream analysis. The pre-processed high-quality data were aligned to the human reference genome using HISAT with the default parameters to identify the alignment percentage. Reads were classified into aligned (which align to the reference genome) and unaligned reads. The tool featureCount was used to estimate and calculate transcript abundance. Absolute counts for transcripts were identified and used in differential expression calculations. DESeq was used to calculate the differentially expressed transcripts. Transcripts were categorized into up, down, and neutrally regulated based on the log2fold change (log2fc) cutoff of ±1 value, and with a false-discovery rate (FDR) < 0.05 were selected for further analysis (GEO accession code GSE216503).
The following publicly available data sets were used for analysis: GSE162729 – for THP-1 macrophages; GSE20050 – for human caseous granuloma, GSE17443 – for lymph node granulomas and GSE174566 – PBMCs infected with Mtb in 2D. GSE174566 14 has the data for 36 samples from PBMCs derived from 3 donors for – 2D cell culture, 3D alginate, and 3D collagen models (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE174566). Of these 36 samples, EM 21.1, EM 23.1, and EM 28.1 correspond to 2D controls, whereas EM 21.2, EM 23.2, and EM 28.2 correspond to 2D Mtb infection. Mtb datasets include GSE6209 – for the transcriptional response of Mtb from THP-1 macrophages (2D infection) and GSE132354 – for the transcriptional response of Mtb from mice. From these data sets, genes with a |log2fc|>1 having a false-discovery rate (FDR)<0.05 were selected for further analysis. The genes in the five datasets were compared, and the similarity of the in vitro models (3D collagen culture system, THP-1 macrophages, 2D PBMCs) with the primary granulomas (lymph node and caseous) was determined by the number of similarly regulated genes (upregulated and downregulated). The ReactomePA package in R was used for gene ontology enrichment analysis of the differentially regulated genes.139 From the human caseous and lymph node granulomas datasets, the significantly enriched (p<0.05) similarly regulated (upregulated and downregulated) REACTOME pathways in both the data sets were selected for representation, and the regulation of these pathways was compared with the in vitro model’s data sets. Heatmaps were made using the ComplexHeatmap function in R.
The transcriptional changes in Mtb from the infected collagen gels were identified using the NetSurgeon algorithm 86 and a precompiled signed transcriptional regulatory network of M. tuberculosis, as explained previously 70.
15. Statistical analysis
Data were presented as mean ± standard deviation and plotted using GraphPad PRISM 9 software. Measurements for each data point were taken from distinct samples. Unpaired two-tailed t-tests with or without Welch’s correction were used to detect statistical differences between groups depending on the number of variables, with p <0.05 considered significant.
Supplementary Material
Acknowledgements
We thank Dr. David Russell (Cornell University) for providing valuable external feedback on this manuscript. We thank Dr. Deepak Saini (IISc) for providing us with GFP and tdTomato conjugated stains of H37Rv and Dr. Amit Singh (IISc) and Dr. Narendra Dixit (IISc) for their invaluable inputs on this project. We thank the Centre for Infectious Diseases Research (CIDR) at IISc Bangalore for allowing us to carry out the studies in the Biosafety Level-3 facility. We thank Dr. Siddharth Jhunjhunwala (IISc) for access to the flow cytometer. We also acknowledge Dr. Thomas Dick’s (Center for Discovery and Innovation, Hackensack Meridian Health) advice for generating spontaneous mutants of PanD. The fluorescence image acquisition was carried out at the Microbiology and Cell Biology (MCB) imaging facility at the Indian Institute of Science.
Funding
This work was supported by the Wellcome Trust–DBT India Alliance Intermediate Fellowship to Rachit Agarwal (IA/I/20/1/504906), Mr. Lakshmi Narayanan, and the Indian Institute of Science start-up fund. We also thank Dr. Vijaya and Rajagopal Rao for funding Biomedical Engineering research at the Department of Bioengineering.
Footnotes
Declaration of interests:
A patent application (IDR-BSSE-2022-080) has been filed for this work. The authors declare no other conflict of interest.
Author contributions:
RA and VG wrote the manuscript. RA and VG designed experiments. APS helped with the initial optimization of collagen gels. VG performed experiments and analyzed the results. MLA and NB performed network analysis. APS and JKM performed the second-harmonic generation and 2-photon imaging. SJ helped with characterizing the POA mutants. VV helped in sample preparation for imaging and validation of PZA data.
References
- 1.Bagcchi S. Lancet Microbe. 2023;4:e20. doi: 10.1016/S2666-5247(22)00359-7. [DOI] [PubMed] [Google Scholar]
- 2.Ernst JD. Nature reviews Immunology. 2012;12:581–591. doi: 10.1038/nri3259. [DOI] [PubMed] [Google Scholar]
- 3.Cohen SB, Gern BH, Delahaye JL, Adams KN, Plumlee CR, Winkler JK, Sherman DR, Gerner MY, Urdahl KB. Cell Host Microbe. 2018;24:439–446.:e434. doi: 10.1016/j.chom.2018.08.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Samstein M, Schreiber HA, Leiner IM, Susac B, Glickman MS, Pamer EG. Elife. 2013;2:e01086. doi: 10.7554/eLife.01086. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Cambier CJ, O’Leary SM, O’Sullivan MP, Keane J, Ramakrishnan L. Immunity. 2017;47:552–565.:e554. doi: 10.1016/j.immuni.2017.08.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Rovetta AI, Pena D, Hernandez Del Pino RE, Recalde GM, Pellegrini J, Bigi F, Musella RM, Palmero DJ, Gutierrez M, Colombo MI, Garcia VE. Autophagy. 2014;10:2109–2121. doi: 10.4161/15548627.2014.981791. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Bussi C, Gutierrez MG. FEMS Microbiol Rev. 2019;43:341–361. doi: 10.1093/femsre/fuz006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Bai X, Kim SH, Azam T, McGibney MT, Huang H, Dinarello CA, Chan ED. J Immunol. 2010;184:3830–3840. doi: 10.4049/jimmunol.0901913. [DOI] [PubMed] [Google Scholar]
- 9.Pisu D, Huang L, Grenier JK, Russell DG. Cell Rep. 2020;30:335–350.:e334. doi: 10.1016/j.celrep.2019.12.033. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Fonseca KL, Rodrigues PNS, Olsson IAS, Saraiva M. PLoS Pathog. 2017;13:e1006421. doi: 10.1371/journal.ppat.1006421. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Kapalczynska M, Kolenda T, Przybyla W, Zajaczkowska M, Teresiak A, Filas V, Ibbs M, Blizniak R, Luczewski L, Lamperska K. Archives of medical science : AMS. 2018;14:910–919. doi: 10.5114/aoms.2016.63743. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Mahamed D, Boulle M, Ganga Y, Mc Arthur C, Skroch S, Oom L, Catinas O, Pillay K, Naicker M, Rampersad S, Mathonsi C, et al. Elife. 2017;6 doi: 10.7554/eLife.28205. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Rohde KH, Veiga DF, Caldwell S, Balazsi G, Russell DG. PLoS Pathog. 2012;8:e1002769. doi: 10.1371/journal.ppat.1002769. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Reichmann MT, Tezera LB, Vallejo AF, Vukmirovic M, Xiao R, Reynolds J, Jogai S, Wilson S, Marshall B, Jones MG, Leslie A, et al. J Clin Invest. 2021;131 doi: 10.1172/JCI148136. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Sun Q, Li X, Perez LM, Shi W, Zhang Y, Sacchettini JC. Nat Commun. 2020;11:339. doi: 10.1038/s41467-019-14238-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Gumbo T, Lenaerts AJ, Hanna D, Romero K, Nuermberger E. The Journal of infectious diseases. 2015;211(Suppl 3):S83–95. doi: 10.1093/infdis/jiv183. [DOI] [PubMed] [Google Scholar]
- 17.Adashi EY, O’Mahony DP, Cohen IG. Am J Med. 2023 doi: 10.1016/j.amjmed.2023.03.033. [DOI] [Google Scholar]
- 18.Han JJ. Artif Organs. 2023;47:449–450. doi: 10.1111/aor.14503. [DOI] [PubMed] [Google Scholar]
- 19.Schwartz MA, Chen CS. Science. 2013;339:402–404. doi: 10.1126/science.1233814. [DOI] [PubMed] [Google Scholar]
- 20.Kapoor N, Pawar S, Sirakova TD, Deb C, Warren WL, Kolattukudy PE. PLoS One. 2013;8:e53657. doi: 10.1371/journal.pone.0053657. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Tezera LB, Bielecka MK, Chancellor A, Reichmann MT, Shammari BA, Brace P, Batty A, Tocheva A, Jogai S, Marshall BG, Tebruegge M, et al. Elife. 2017;6 doi: 10.7554/eLife.21283. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Kim MJ, Wainwright HC, Locketz M, Bekker LG, Walther GB, Dittrich C, Visser A, Wang W, Hsu FF, Wiehart U, Tsenova L, et al. EMBO Mol Med. 2010;2:258–274. doi: 10.1002/emmm.201000079. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Pagan AJ, Yang CT, Cameron J, Swaim LE, Ellett F, Lieschke GJ, Ramakrishnan L. Cell Host Microbe. 2015;18:15–26. doi: 10.1016/j.chom.2015.06.008. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Berg RD, Levitte S, O’Sullivan MP, O’Leary SM, Cambier CJ, Cameron J, Takaki KK, Moens CB, Tobin DM, Keane J, Ramakrishnan L. Cell. 2016;165:139–152. doi: 10.1016/j.cell.2016.02.034. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Pagan AJ, Lee LJ, Edwards-Hicks J, Moens CB, Tobin DM, Busch-Nentwich EM, Pearce EL, Ramakrishnan L. Cell. 2022;185:3720–3738.:e3713. doi: 10.1016/j.cell.2022.08.018. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Madri JA, Furthmayr H. Hum Pathol. 1980;11:353–366. doi: 10.1016/s0046-8177(80)80031-1. [DOI] [PubMed] [Google Scholar]
- 27.Bateman ED, Turner-Warwick M, Adelmann-Grill BC. Thorax. 1981;36:645–653. doi: 10.1136/thx.36.9.645. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Raub CB, Suresh V, Krasieva T, Lyubovitsky J, Mih JD, Putnam AJ, Tromberg BJ, George SC. Biophys J. 2007;92:2212–2222. doi: 10.1529/biophysj.106.097998. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Hameed P, Manivasagam G. Biophys Rev. 2021;13:387–403. doi: 10.1007/s12551-021-00804-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Sicard D, Haak AJ, Choi KM, Craig AR, Fredenburgh LE, Tschumperlin DJ. Am J Physiol Lung Cell Mol Physiol. 2018;314:L946–L955. doi: 10.1152/ajplung.00415.2017. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Polio SR, Kundu AN, Dougan CE, Birch NP, Aurian-Blajeni DE, Schiffman JD, Crosby AJ, Peyton SR. PLoS One. 2018;13:e0204765. doi: 10.1371/journal.pone.0204765. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Chen X, Nadiarynkh O, Plotnikov S, Campagnola PJ. Nature protocols. 2012;7:654–669. doi: 10.1038/nprot.2012.009. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Abraham T, Hirota JA, Wadsworth S, Knight DA. Pulm Pharmacol Ther. 2011;24:487–496. doi: 10.1016/j.pupt.2011.03.008. [DOI] [PubMed] [Google Scholar]
- 34.Buchheit CL, Rayavarapu RR, Schafer ZT. Semin Cell Dev Biol. 2012;23:402–411. doi: 10.1016/j.semcdb.2012.04.007. [DOI] [PubMed] [Google Scholar]
- 35.Al Shammari B, Shiomi T, Tezera L, Bielecka MK, Workman V, Sathyamoorthy T, Mauri F, Jayasinghe SN, Robertson BD, D’Armiento J, Friedland JS, et al. J Infect Dis. 2015;212:463–473. doi: 10.1093/infdis/jiv076. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Huang L, Nazarova EV, Tan S, Liu Y, Russell DG. J Exp Med. 2018;215:1135–1152. doi: 10.1084/jem.20172020. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Haas H, Michel J, Sacks TG. Chemotherapy. 1982;28:1–5. doi: 10.1159/000238053. [DOI] [PubMed] [Google Scholar]
- 38.Dijkstra JA, van der Laan T, Akkerman OW, Bolhuis MS, de Lange WCM, Kosterink JGW, van der Werf TS, Alffenaar JWC, van Soolingen D. Antimicrob Agents Chemother. 2018;62 doi: 10.1128/AAC.01724-17. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Brilha S, Wysoczanski R, Whittington AM, Friedland JS, Porter JC. J Immunol. 2017;199:982–991. doi: 10.4049/jimmunol.1700128. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Kumar S, Puniya BL, Parween S, Nahar P, Ramachandran S. PLoS One. 2013;8:e69790. doi: 10.1371/journal.pone.0069790. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Foster GA, Headen DM, Gonzalez-Garcia C, Salmeron-Sanchez M, Shirwan H, Garcia AJ. Biomaterials. 2017;113:170–175. doi: 10.1016/j.biomaterials.2016.10.044. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Phelps EA, Enemchukwu NO, Fiore VF, Sy JC, Murthy N, Sulchek TA, Barker TH, Garcia AJ. Adv Mater. 2012;24:64–70. doi: 10.1002/adma.201103574. 62. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Clark AY, Martin KE, Garcia JR, Johnson CT, Theriault HS, Han WM, Zhou DW, Botchwey EA, Garcia AJ. Nat Commun. 2020;11:114. doi: 10.1038/s41467-019-14000-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Parish T. Expert Opin Drug Discov. 2020;15:349–358. doi: 10.1080/17460441.2020.1707801. [DOI] [PubMed] [Google Scholar]
- 45.Gill WP, Harik NS, Whiddon MR, Liao RP, Mittler JE, Sherman DR. Nat Med. 2009;15:211–214. doi: 10.1038/nm.1915. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Gross MW, Karbach U, Groebe K, Franko AJ, Mueller-Klieser W. Int J Cancer. 1995;61:567–573. doi: 10.1002/ijc.2910610422. [DOI] [PubMed] [Google Scholar]
- 47.Srivastava S, Ernst JD, Desvignes L. Immunol Rev. 2014;262:179–192. doi: 10.1111/imr.12217. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Geissmann F, Manz MG, Jung S, Sieweke MH, Merad M, Ley K. Science. 2010;327:656–661. doi: 10.1126/science.1178331. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Chistiakov DA, Killingsworth MC, Myasoedova VA, Orekhov AN, Bobryshev YV. Laboratory investigation; a journal of technical methods and pathology. 2017;97:4–13. doi: 10.1038/labinvest.2016.116. [DOI] [PubMed] [Google Scholar]
- 50.Rutschmann O, Toniolo C, McKinney JD. mBio. 2022;13:e0225122. doi: 10.1128/mbio.02251-22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51.Choi HS, Rai PR, Chu HW, Cool C, Chan ED. Am J Respir Crit Care Med. 2002;166:178–186. doi: 10.1164/rccm.2201023. [DOI] [PubMed] [Google Scholar]
- 52.Cunningham-Bussel A, Zhang T, Nathan CF. Proc Natl Acad Sci U S A. 2013;110:E4256–4265. doi: 10.1073/pnas.1316894110. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 53.Mattila JT, Ojo OO, Kepka-Lenhart D, Marino S, Kim JH, Eum SY, Via LE, Barry CE, 3rd, Klein E, Kirschner DE, Morris SM, et al. J Immunol. 2013;191:773–784. doi: 10.4049/jimmunol.1300113. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 54.Ufimtseva EG, Eremeeva NI, Petrunina EM, Umpeleva TV, Bayborodin SI, Vakhrusheva DV, Skornyakov SN. Tuberculosis (Edinb) 2018;112:1–10. doi: 10.1016/j.tube.2018.07.001. [DOI] [PubMed] [Google Scholar]
- 55.Clay H, Volkman HE, Ramakrishnan L. Immunity. 2008;29:283–294. doi: 10.1016/j.immuni.2008.06.011. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 56.Bernut A, Herrmann JL, Kissa K, Dubremetz JF, Gaillard JL, Lutfalla G, Kremer L. Proc Natl Acad Sci U S A. 2014;111:E943–952. doi: 10.1073/pnas.1321390111. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 57.Lerner TR, Queval CJ, Lai RP, Russell MR, Fearns A, Greenwood DJ, Collinson L, Wilkinson RJ, Gutierrez MG. JCI Insight. 2020;5 doi: 10.1172/jci.insight.136937. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 58.Toniolo C, Dhar N, McKinney JD. EMBO J. 2023;42:e113490. doi: 10.15252/embj.2023113490. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 59.Roca FJ, Whitworth LJ, Prag HA, Murphy MP, Ramakrishnan L. Science. 2022;376:eabh2841. doi: 10.1126/science.abh2841. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 60.Kalsum S, Braian C, Koeken V, Raffetseder J, Lindroth M, van Crevel R, Lerm M. Front Cell Infect Microbiol. 2017;7:278. doi: 10.3389/fcimb.2017.00278. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 61.McCaffrey EF, Donato M, Keren L, Chen Z, Delmastro A, Fitzpatrick MB, Gupta S, Greenwald NF, Baranski A, Graf W, Kumar R, et al. Nat Immunol. 2022;23:814. doi: 10.1038/s41590-022-01178-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 62.Liu T, Huang T, Li J, Li A, Li C, Huang X, Li D, Wang S, Liang M. PLoS One. 2023;18:e0286056. doi: 10.1371/journal.pone.0286056. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 63.Plaisier CL, Bernard B, Reynolds S, Simon Z, Toledo CM, Ding Y, Reiss DJ, Paddison PJ, Baliga NS. Cell Syst. 2016;3:172–186. doi: 10.1016/j.cels.2016.06.006. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 64.Gawad S, Schild L, Renaud PH. Lab Chip. 2001;1:76–82. doi: 10.1039/b103933b. [DOI] [PubMed] [Google Scholar]
- 65.Chai Q, Lu Z, Liu Z, Zhong Y, Zhang F, Qiu C, Li B, Wang J, Zhang L, Pang Y, Liu CH. Commun Biol. 2020;3:604. doi: 10.1038/s42003-020-01318-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 66.Cliff JM, Cho JE, Lee JS, Ronacher K, King EC, van Helden P, Walzl G, Dockrell HM. J Infect Dis. 2016;213:485–495. doi: 10.1093/infdis/jiv447. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 67.Koh HJ, Kim YR, Kim JS, Yun JS, Kim S, Kim SY, Jang K, Yang CS. Exp Mol Med. 2018;50:1–15. doi: 10.1038/s12276-018-0091-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 68.Grant AV, Sabri A, Abid A, Abderrahmani Rhorfi I, Benkirane M, Souhi H, Naji Amrani H, Alaoui-Tahiri K, Gharbaoui Y, Lazrak F, Sentissi I, et al. Hum Genet. 2016;135:299–307. doi: 10.1007/s00439-016-1633-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 69.Fontan P, Aris V, Ghanny S, Soteropoulos P, Smith I. Infect Immun. 2008;76:717–725. doi: 10.1128/IAI.00974-07. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 70.Peterson EJ, Bailo R, Rothchild AC, Arrieta-Ortiz ML, Kaur A, Pan M, Mai D, Abidi AA, Cooper C, Aderem A, Bhatt A, et al. Mol Syst Biol. 2019;15:e8584. doi: 10.15252/msb.20188584. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 71.Shariq M, Quadir N, Sharma N, Singh J, Sheikh JA, Khubaib M, Hasnain SE, Ehtesham NZ. Front Immunol. 2021;12:636644. doi: 10.3389/fimmu.2021.636644. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 72.Healy C, Gouzy A, Ehrt S. mBio. 2020;11 doi: 10.1128/mBio.03315-19. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 73.Rosas-Magallanes V, Stadthagen-Gomez G, Rauzier J, Barreiro LB, Tailleux L, Boudou F, Griffin R, Nigou J, Jackson M, Gicquel B, Neyrolles O. Infect Immun. 2007;75:504–507. doi: 10.1128/IAI.00058-06. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 74.Calamita H, Ko C, Tyagi S, Yoshimatsu T, Morrison NE, Bishai WR. Cell Microbiol. 2005;7:233–244. doi: 10.1111/j.1462-5822.2004.00454.x. [DOI] [PubMed] [Google Scholar]
- 75.Chauhan A, Singh N, Kumar R, Kushwaha NK, Prajapati VM, Singh SK. Tuberculosis (Edinb) 2023;141:102352. doi: 10.1016/j.tube.2023.102352. [DOI] [PubMed] [Google Scholar]
- 76.Casabon I, Zhu SH, Otani H, Liu J, Mohn WW, Eltis LD. Mol Microbiol. 2013;89:1201–1212. doi: 10.1111/mmi.12340. [DOI] [PubMed] [Google Scholar]
- 77.Shimono N, Morici L, Casali N, Cantrell S, Sidders B, Ehrt S, Riley LW. Proc Natl Acad Sci U S A. 2003;100:15918–15923. doi: 10.1073/pnas.2433882100. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 78.Nazarova EV, Montague CR, La T, Wilburn KM, Sukumar N, Lee W, Caldwell S, Russell DG, VanderVen BC. Elife. 2017;6 doi: 10.7554/eLife.26969. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 79.Nazarova EV, Montague CR, Huang L, La T, Russell D, VanderVen BC. Elife. 2019;8 doi: 10.7554/eLife.43621. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 80.Cooper C, Peterson EJR, Bailo R, Pan M, Singh A, Moynihan P, Nakaya M, Fujiwara N, Baliga N, Bhatt A. Proc Natl Acad Sci U S A. 2022;119 doi: 10.1073/pnas.2111059119. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 81.Sun R, Converse PJ, Ko C, Tyagi S, Morrison NE, Bishai WR. Mol Microbiol. 2004;52:25–38. doi: 10.1111/j.1365-2958.2003.03958.x. [DOI] [PubMed] [Google Scholar]
- 82.Zhang YJ, Reddy MC, Ioerger TR, Rothchild AC, Dartois V, Schuster BM, Trauner A, Wallis D, Galaviz S, Huttenhower C, Sacchettini JC, et al. Cell. 2013;155:1296–1308. doi: 10.1016/j.cell.2013.10.045. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 83.Hudock TA, Foreman TW, Bandyopadhyay N, Gautam US, Veatch AV, LoBato DN, Gentry KM, Golden NA, Cavigli A, Mueller M, Hwang SA, et al. Am J Respir Cell Mol Biol. 2017;56:637–647. doi: 10.1165/rcmb.2016-0239OC. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 84.Park HD, Guinn KM, Harrell MI, Liao R, Voskuil MI, Tompa M, Schoolnik GK, Sherman DR. Mol Microbiol. 2003;48:833–843. doi: 10.1046/j.1365-2958.2003.03474.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 85.Nguyen TH, Turek I, Meehan-Andrews T, Zacharias A, Irving HR. PLoS One. 2022;17:e0263968. doi: 10.1371/journal.pone.0263968. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 86.Michael DG, Maier EJ, Brown H, Gish SR, Fiore C, Brown RH, Brent MR. Proc Natl Acad Sci U S A. 2016;113:E7428–E7437. doi: 10.1073/pnas.1603577113. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 87.Warner DF. Cold Spring Harb Perspect Med. 2014;5 doi: 10.1101/cshperspect.a021121. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 88.Ehrt S, Schnappinger D, Rhee KY. Nature reviews Microbiology. 2018;16:496–507. doi: 10.1038/s41579-018-0013-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 89.Li G, Li J, Otsuka Y, Zhang S, Takahashi M, Yamada K. Materials. 2020;13 doi: 10.3390/ma13030677. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 90.Lancet. 1974;2:237–240. [Google Scholar]
- 91.Am Rev Respir Dis. 1978;118:219–228. doi: 10.1164/arrd.1978.118.2.219. [DOI] [PubMed] [Google Scholar]
- 92.Tarshis MS, Weed WA., Jr Am Rev Tuberc. 1953;67:391–395. doi: 10.1164/art.1953.67.3.391. [DOI] [PubMed] [Google Scholar]
- 93.Zhang Y, Shi W, Zhang W, Mitchison D. Microbiol Spectr. 2014;2:MGM2-0023-2013. doi: 10.1128/microbiolspec.MGM2-0023-2013. [DOI] [PubMed] [Google Scholar]
- 94.McCune RM, Jr, McDermott W, Tompsett R. J Exp Med. 1956;104:763–802. doi: 10.1084/jem.104.5.763. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 95.Malone L, Schurr A, Lindh H, Mc KD, Kiser JS, Williams JH. Am Rev Tuberc. 1952;65:511–518. [PubMed] [Google Scholar]
- 96.Nuermberger E, Grosset J. European journal of clinical microbiology & infectious diseases : official publication of the European Society of Clinical Microbiology. 2004;23:243–255. doi: 10.1007/s10096-004-1109-5. [DOI] [PubMed] [Google Scholar]
- 97.Bielecka MK, Tezera LB, Zmijan R, Drobniewski F, Zhang X, Jayasinghe S, Elkington P. mBio. 2017;8 doi: 10.1128/mBio.02073-16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 98.Gopal P, Tasneen R, Yee M, Lanoix JP, Sarathy J, Rasic G, Li L, Dartois V, Nuermberger E, Dick T. ACS Infect Dis. 2017;3:492–501. doi: 10.1021/acsinfecdis.7b00017. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 99.Yee M, Gopal P, Dick T. Antimicrob Agents Chemother. 2017;61 doi: 10.1128/AAC.02342-16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 100.Gopal P, Nartey W, Ragunathan P, Sarathy J, Kaya F, Yee M, Setzer C, Manimekalai MSS, Dartois V, Gruber G, Dick T. ACS Infect Dis. 2017;3:807–819. doi: 10.1021/acsinfecdis.7b00079. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 101.Gopal P, Sarathy JP, Yee M, Ragunathan P, Shin J, Bhushan S, Zhu J, Akopian T, Kandror O, Lim TK, Gengenbacher M, et al. Nat Commun. 2020;11:1661. doi: 10.1038/s41467-020-15516-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 102.Santucci P, Greenwood DJ, Fearns A, Chen K, Jiang H, Gutierrez MG. Nat Commun. 2021;12:3816. doi: 10.1038/s41467-021-24127-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 103.Jureen P, Werngren J, Toro JC, Hoffner S. Antimicrob Agents Chemother. 2008;52:1852–1854. doi: 10.1128/AAC.00110-08. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 104.Salfinger M, Heifets LB. Antimicrob Agents Chemother. 1988;32:1002–1004. doi: 10.1128/aac.32.7.1002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 105.Rohde KH, Abramovitch RB, Russell DG. Cell Host Microbe. 2007;2:352–364. doi: 10.1016/j.chom.2007.09.006. [DOI] [PubMed] [Google Scholar]
- 106.Casali N, Nikolayevskyy V, Balabanova Y, Harris SR, Ignatyeva O, Kontsevaya I, Corander J, Bryant J, Parkhill J, Nejentsev S, Horstmann RD, et al. Nat Genet. 2014;46:279–286. doi: 10.1038/ng.2878. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 107.Stoffels K, Mathys V, Fauville-Dufaux M, Wintjens R, Bifani P. Antimicrob Agents Chemother. 2012;56:5186–5193. doi: 10.1128/AAC.05385-11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 108.Zhang S, Chen J, Shi W, Liu W, Zhang W, Zhang Y. Emerg Microbes Infect. 2013;2:e34. doi: 10.1038/emi.2013.38. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 109.Sambandamurthy VK, Wang X, Chen B, Russell RG, Derrick S, Collins FM, Morris SL, Jacobs WR., Jr Nat Med. 2002;8:1171–1174. doi: 10.1038/nm765. [DOI] [PubMed] [Google Scholar]
- 110.Gopal P, Gruber G, Dartois V, Dick T. Trends Pharmacol Sci. 2019;40:930–940. doi: 10.1016/j.tips.2019.10.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 111.Thiede JM, Dillon NA, Howe MD, Aflakpui R, Modlin SJ, Hoffner SE, Valafar F, Minato Y, Baughn AD. mBio. 2022;13:e00439-00421. doi: 10.1128/mbio.00439-21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 112.Elkington PT, D’Armiento JM, Friedland JS. Sci Transl Med. 2011;3:71ps76. doi: 10.1126/scitranslmed.3001847. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 113.Xu Y, Wang L, Zimmerman MD, Chen KY, Huang L, Fu DJ, Kaya F, Rakhilin N, Nazarova EV, Bu P, Dartois V, et al. PLoS Pathog. 2018;14:e1006974. doi: 10.1371/journal.ppat.1006974. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 114.Jackson BC, Nebert DW, Vasiliou V. Hum Genomics. 2010;4:194–201. doi: 10.1186/1479-7364-4-3-194. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 115.Fitzgerald LE, Abendano N, Juste RA, Alonso-Hearn M. BioMed research international. 2014;2014:623856. doi: 10.1155/2014/623856. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 116.Bryson BD, Rosebrock TR, Tafesse FG, Itoh CY, Nibasumba A, Babunovic GH, Corleis B, Martin C, Keegan C, Andrade P, Realegeno S, et al. Nat Commun. 2019;10:2329. doi: 10.1038/s41467-019-10065-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 117.Cronan MR, Matty MA, Rosenberg AF, Blanc L, Pyle CJ, Espenschied ST, Rawls JF, Dartois V, Tobin DM. Nat Methods. 2018;15:1098–1107. doi: 10.1038/s41592-018-0215-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 118.Iakobachvili N, Leon-Icaza SA, Knoops K, Sachs N, Mazeres S, Simeone R, Peixoto A, Bernard C, Murris-Espin M, Mazieres J, Cam K, et al. Mol Microbiol. 2022;117:682–692. doi: 10.1111/mmi.14824. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 119.Pasipanodya JG, Nuermberger E, Romero K, Hanna D, Gumbo T. Clin Infect Dis. 2015;61(Suppl 1):S10–17. doi: 10.1093/cid/civ425. [DOI] [PubMed] [Google Scholar]
- 120.Caliari SR, Burdick JA. Nat Methods. 2016;13:405–414. doi: 10.1038/nmeth.3839. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 121.Doyle AD. Current protocols in cell biology. 2016;72:10 20 11-10 20 16. doi: 10.1002/cpcb.9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 122.Zhang L, Jiang X, Pfau D, Ling Y, Nathan CF. J Exp Med. 2021;218 doi: 10.1084/jem.20200887. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 123.Russell DG, Cardona PJ, Kim MJ, Allain S, Altare F. Nat Immunol. 2009;10:943–948. doi: 10.1038/ni.1781. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 124.Singh V, Jamwal S, Jain R, Verma P, Gokhale R, Rao KV. Cell Host Microbe. 2012;12:669–681. doi: 10.1016/j.chom.2012.09.012. [DOI] [PubMed] [Google Scholar]
- 125.Jaisinghani N, Dawa S, Singh K, Nandy A, Menon D, Bhandari PD, Khare G, Tyagi A, Gandotra S. Front Immunol. 2018;9:1490. doi: 10.3389/fimmu.2018.01490. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 126.Lee W, VanderVen BC, Fahey RJ, Russell DG. J Biol Chem. 2013;288:6788–6800. doi: 10.1074/jbc.M112.445056. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 127.Podinovskaia M, Lee W, Caldwell S, Russell DG. Cell Microbiol. 2013;15:843–859. doi: 10.1111/cmi.12092. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 128.Puissegur MP, Botanch C, Duteyrat JL, Delsol G, Caratero C, Altare F. Cell Microbiol. 2004;6:423–433. doi: 10.1111/j.1462-5822.2004.00371.x. [DOI] [PubMed] [Google Scholar]
- 129.Agarwal P, Khan SR, Verma SC, Beg M, Singh K, Mitra K, Gaikwad AN, Akhtar MS, Krishnan MY. Microbes Infect. 2014;16:571–580. doi: 10.1016/j.micinf.2014.04.006. [DOI] [PubMed] [Google Scholar]
- 130.Caire-Brandli I, Papadopoulos A, Malaga W, Marais D, Canaan S, Thilo L, de Chastellier C. Infect Immun. 2014;82:476–490. doi: 10.1128/IAI.01196-13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 131.Belisle JT, Vissa VD, Sievert T, Takayama K, Brennan PJ, Besra GS. Science. 1997;276:1420–1422. doi: 10.1126/science.276.5317.1420. [DOI] [PubMed] [Google Scholar]
- 132.Santucci P, Aylan B, Botella L, Bernard EM, Bussi C, Pellegrino E, Athanasiadi N, Gutierrez MG. mBio. 2022;13:e0011722. doi: 10.1128/mbio.00117-22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 133.Jain N, Kalam H, Singh L, Sharma V, Kedia S, Das P, Ahuja V, Kumar D. Nat Commun. 2020;11:3062. doi: 10.1038/s41467-020-16877-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 134.Cooper AM. Annu Rev Immunol. 2009;27:393–422. doi: 10.1146/annurev.immunol.021908.132703. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 135.de Martino M, Lodi L, Galli L, Chiappini E. Front Pediatr. 2019;7:350. doi: 10.3389/fped.2019.00350. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 136.Ravesloot-Chavez MM, Van Dis E, Stanley SA. Annu Rev Immunol. 2021;39:611–637. doi: 10.1146/annurev-immunol-093019-010426. [DOI] [PubMed] [Google Scholar]
- 137.Richards DJ, Li Y, Kerr CM, Yao J, Beeson GC, Coyle RC, Chen X, Jia J, Damon B, Wilson R, Hazard E Starr, et al. Nat Biomed Eng. 2020;4:446–462. doi: 10.1038/s41551-020-0539-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 138.Yang CC, Jenkins L, Burg KJL. Journal of Histotechnology. 2007;30:185–191. [Google Scholar]
- 139.Yu G, He QY. Mol Biosyst. 2016;12:477–479. doi: 10.1039/c5mb00663e. [DOI] [PubMed] [Google Scholar]
- 140.Liberzon A, Birger C, Thorvaldsdottir H, Ghandi M, Mesirov JP, Tamayo P. Cell Syst. 2015;1:417–425. doi: 10.1016/j.cels.2015.12.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.






