Ontogeny and intra-lineage diversity are key determinants in macrophage responses to Mtb infection.
To better understand macrophage behavior during Mtb infection, we conducted an extensive multi-modal scRNA-seq analysis of macrophage phenotypes at 2-, 3-, 4-, 6-weeks post-Mtb challenge, which spans the transition from innate to adaptive immune responses2,10,32-36. Our aim was to determine how distinct macrophage populations respond post-Mtb infection, with the goal of identifying the subsets that are responsible for controller responses and those that favor Mtb replication and expansion. Consistent with our previous work9, we infected mice with the hspx’::gfp/smyc’::mCherry Mtb reporter strain. This strain expresses mCherry constitutively and GFP in response to host mediated immune-related stress35. For each infected host cell, we were able to assess the fitness status of intracellular Mtb, quantify the surface marker expression and profile their transcriptome9. After QC, our dataset comprised of 49600 cells, the significant majority of which were identified as macrophages (Figure 1A and Supp. Figure 1A).
Recent studies, including ours, have shown that both resident alveolar macrophages (AMs) and recruited interstitial macrophages (IMs), are ontogenically diverse 6,8,37,38 and comprised of phenotypically-distinct subpopulations9,39-42. These subpopulations exhibit markedly different inflammatory responses to Mtb infection9. In the current study, we extend these findings, further characterizing the discrete AM and IMs subpopulations based on their origin and inflammatory profiles.
AMs are often identified by their expression of surface markers SiglecF and CD11c6,43,44, and in our dataset we could identify AM subpopulations by SiglecF and CD11c staining (Figure 1B). Interestingly, we also noticed a population of cells that cluster with AMs by UMAP analysis, but do not stain for either SiglecF or CD11c (Figure 1A and 1C). To determine the origin of these cells, we focused on the early infection stages (Naïve and 2 weeks post-infection). We performed unbiased weighted gene correlation network analysis (WGCNA) on the resulting 13,125 cells to identify gene co-expression modules that define this population45. The analysis identified two gene expression modules that mirrored the differential staining of the SiglecF and CD11c antibodies (Figure 1D). Module 6 consists of a 28-gene co-expression set and includes genes commonly used to define tissue resident alveolar macrophages—such as Cd9, Mrc1 (CD206), Lpl, Trf, and Chil3 (Ym1) (Figure 1E)8,9,46,47. In contrast, Module 5, represented by a larger 110-gene set and expressed by the SiglecF/CD11c double negative population, is enriched in genes associated with the monocytic lineage (Figure 1E) and monocyte-to-macrophage differentiation (Supp Figure 1B). These AM-like cells express MafB—a driver of monocyte-to-macrophage differentiation—and Ly6c2, a marker for cells of monocyte origins (Figure 1F)48-52. We therefore designated this population as moAMs (monocyte-derived AMs). In the lungs of naïve mice moAMs are absent, implying these monocytes migrate there and differentiate to perform AM-related functions after Mtb infection (Figure 2A). Trajectory and pseudotime analyses corroborate this hypothesis, revealing that moAMs, along with the classical NOS2+ and NOS2- IMs, represent distinct cell fates (grey leaves) that originate from a population of infiltrating monocytes (root 2 - white circle - and branching point 11 – black circle -), characterized by a unique transcriptional signature associated with leukocyte adhesion and extravasation (Figures 2B and 2C)53-64. Uninfected bystander moAMs express genes associated with macrophage functions including phagocytosis, lipid metabolism, and RNA/protein synthesis, indicative of roles beyond mere homeostasis (Figure 2D)65-67. This perspective is supported by the contrasting gene expression in Mtb-infected moAMs, which pivot to a pro-inflammatory state, as evidenced by the upregulation of genes such as Acsl1, Hmox1, Saa3, Ptgs2, Slc7a11, Clec4e among others(Figure 2D)9. This transition is marked by elevated Nos2 expression and association with hspx’::GFP-high bacteria (Figure 2E). Such findings align with prior research which described the role of interferon-g in deactivating enhancer regions bound by the transcription factor Maf. This mechanism is crucial for suppressing M2 genes and increasing activation of monocyte-derived macrophages68. This moAM cohort expresses CD38 (Figure 2F), hence their designation as CD38+ moAMs.
Next, we focused on TR-AMs and identified two subpopulations based on their CD38 expression. The CD38+ population exhibits increased Nos2 expression and is associated with stressed (hspx’::GFP high) bacteria (Figure 2E and 2F). Notably, these TR-AMs are dual positive for SiglecF and CD11c (Figure 1B). They also express the TR-AM gene signature from Module 6 (Figure 1D), leading us to annotate them as CD38+ TR-AMs. Importantly, trajectory and pseudotime analyses establish the CD38+ TR-AMs as a distinct lineage from moAMs, as evidenced by their origin from root 1 (white circle) (Figure 2B). Additionally, we also identified CD38- subpopulations of TR-AMs, associated with hspx’::GFP-low Mtb, indicating bacteria experiencing minimal host-derived stress (Figures 2E and 2F).
In examining the IM subpopulations, we observed a similar level of heterogeneity. While both IM populations appear to originate from infiltrating monocytes (Figure 2B), the NOS2+ IMs are CD38+, whereas the NOS2- IM lack CD38 expression (Figure 2F). Looking at the transcriptional profiles of these populations revealed that CD38+ IMs, akin to CD38+ moAMs and TR-AMs, display a gene expression profile consistent with classic pro-inflammatory responses, which were previously linked to effective tuberculosis control6,9 (Figure 2G) (Suppl. Table 1).
Conversely, the IM subpopulation that is CD38- not only lack Nos2 expression, but is also associated with hspx’::GFP-low Mtb (Figure 2E and 2F). Relative to NOS2+ IMs, these cells show amplified expression of genes linked to anti-inflammatory responses. This includes transcripts for complement proteins C1q, Ccl8, Ms4a7 among others, which have been previously characterized (Figure 2G)9,69-73. Furthermore, this subset still shows a partial expression of extravasation and adhesion markers shared with infiltrating monocytes (Figure 2C). This suggests that they represent recently arrived interstitial monocytes that have just been infected and have not yet undergone immune activation (Suppl Table 1). In summary, our analysis identified distinct subpopulations of AMs and IMs with varying ontogeny and inflammatory profiles. Importantly, we discovered a population of monocyte-derived AMs capable of transitioning to a pro-inflammatory state upon Mtb infection, and two subpopulations of TR-AMs based on CD38 expression, which are associated with different bacterial phenotypes following Mtb infection.
Pro-inflammatory immune responses are associated with rising CD38 expression in macrophages over time.
To understand the temporal evolution of macrophage immune responses following Mtb infection, we examined the expression patterns of pro-inflammatory gene signatures and associated markers over time. From our past studies, there was a discernible correlation between Nos2 expression in host macrophages and bacterial stress (Figure 2E)9. Building on this, we tracked the temporal trends of Nos2 expression within the infected macrophage populations. As the infection progresses, we observed an increase in nitric oxide production, reflected by both an increase in the level of Nos2 expression per cell (intensity) and the number of cells expressing Nos2 (prevalence), as shown in Figure 3A. Flow cytometry data confirmed this trend; the median fluorescence intensity (MFI) of the GFP signal from hspx’::GFP-infected cells increased over time, indicating heightened bacterial stress as the infection proceeded (Figure 3B). Importantly, infecting Nos2-KO mice with the hspx’::GFP Mtb reporter strain resulted in minimal GFP induction by Mtb at both 2 and 4 weeks post-infection (Figure 3C). The limited induction of GFP expression by Mtb in Nos2-KO mice was also accompanied by an increase in bacterial load (Figure 3D). Finally, flow cytometry analysis at 2 wpi of Rag1 and IFN-γ KO mice revealed no induction of hspx’::GFP (Supplementary Figure 2A), confirming that production of nitric oxide by pro-inflammatory macrophages is critically dependent on the adaptive immune response.
Consistent with these data, our surface marker analysis mirrored the trends observed in Nos2 RNA levels. Specifically, CD38 protein levels – a marker associated with inflammation20,22 – increase in infected macrophages during the latter stages (Figure 3E). Parallel trends were also seen with the marker CD11b (Suppl Figure 2B). To gain deeper insights into the molecular mechanisms and biological processes driving these phenotypes, we performed a time-dependent pathway analysis74. This analysis identified 68 pathways that exhibited variations in expression as the infection progressed (p.adj < 0.05)(Table 1). Pathways showing increased representation over time were predominantly associated with inflammatory and anti-bacterial responses and were overly represented in CD38+ macrophages (Figure 3F and Supp figure 2C). Conversely, pathways that dominated early stages of infection were mostly associated with CD38- macrophages, and aligned with processes linked to M2 polarization, tissue homeostasis and bacterial survival, such as negative regulation of protein kinase c signaling75, suppression of apoptosis and cell proliferation among others (Figure 3F and Supp figure 2D). Confirming these observations, our analysis revealed an increased recovery of Mtb mRNA reads from CD38+ macrophage populations over time, indicative of active bacterial degradation within these cells. Specifically, we noted low levels of Mtb reads at 2 weeks post-infection, which significantly increased from 3 to 6 weeks, as illustrated in Supplementary Figure 2E.
Intriguingly, when focusing on bystander macrophages, our analysis revealed no change in their CD38 staining levels, irrespective of the timepoint examined, contrasting starkly with their infected counterparts (Figure 3E). To validate this, we leveraged a force-directed layout to conduct unbiased graph-based clustering of both cell types76. While infected cells clustered according to duration of infection, bystander cells grouped based on their ontogeny without temporal distinctions (Figure 3G). This pronounced difference emphasizes that the inflammatory responses we observe aren't merely a result of a broad immune activation of host macrophages over time; rather, they are finely tuned and specifically driven by active Mtb infection.
In conclusion, our findings highlight a dynamic shift in the phenotype and relative proportion of infected macrophage populations over the course of infection. This transition emphasizes the crucial role of CD38+ macrophages, which appear to be closely associated with the control of infection.
The relative proportion of Mtb-infected cells shifts from AMs to IMs over time.
To quantify the changes in the proportion of infected macrophage populations over time, we employed a recently developed statistical framework designed for differential abundance testing (DAB)77. We used this approach to assess whether changes in abundance of both infected and bystander cells occur over the course of infection. Our analysis identified variations in the abundance of 400 neighbors across different timepoints in infected AMs and IMs subpopulations (with FDR < 0.1). This contrasted with bystander populations that exhibited minimal variation (Figure 4A).
Mtb-infected neighbors belonging to the CD38- TR-AMs subpopulations showed a marked decline in numbers late in infection. Conversely, neighbors within the NOS2+ IMs showed a significant increase in abundance during the same later timepoints (Figure 4B). Observations at the broader population level further highlighted this trend: at 2 weeks post-infection (wpi), AMs constituted 80% of all the infected macrophages, but by 3wpi this percentage was reduced to approximately 39%, and by the 4th and 6th wpi, it dropped to around 15% (Figure 4C). Overall, our data highlight a substantial shift in abundance from TR-AMs to IMs among the infected macrophages as the infection advances, as also confirmed by flow cytometry analysis (Supp. Figure 3). This trend sharply contrast the patterns seen in bystander AMs and IMs, whose proportions remain consistent across the analyzed timepoints (Figure 4A). The stability in the proportion of bystander macrophages in the Mtb-infected lung from the 2 weeks post-infection (wpi) suggests that, within our infection model, monocyte-derived cells have already been recruited to the lung by this time but remain largely uninfected (Figure 4D), similar to a previous report78.
This premise is further supported by examining the distinct timelines of infection rates between CD38+ IMs and moAMs. Despite pseudotime analysis indicating a shared origin for both populations from the early monocyte cluster (as illustrated in Figure 2B), CD38+ IMs only became the more abundant infected group by the 3rd week post-infection (Figures 4A and 4C). In contrast, a substantial proportion of CD38+ moAMs were infected at a much earlier stage, by 2wpi (Figures 4C and 2A). Recent studies provide context to these observations. The initial delay in the infection of monocyte-derived IM may be attributed to the early-stage preference of the infection for the lung's alveolar space during the onset of tuberculosis. As the disease progresses, leading to the translocation of infected AMs from the alveolar to the interstitial lung space2,6,8,10, the infection rate of CD38+ Nos2+ IMs increases, leading to the overall shift in the abundance of infected macrophages from AMs to IMs as described above.
In summary, while the total bystander IM and AM populations in the lung remain constant from 2wpi onwards, our data indicates a decline in the number of newly infected TR-AMs as the infection shifts from the alveolar to the interstitial lung space. In the established infection, our data demonstrate that monocyte-derived IMs are the dominant host cell population accounting for the majority of new infection events.
The early infection timepoints are dominated by CD38- TR-AMs that are relatively nonresponsive to tuberculosis infection.
Work by Rothchild et al. highlighted that murine AMs exhibit a robust anti-inflammatory response during the early stages of tuberculosis infection, up to 10 dpi. This early-phase response is postulated to be pivotal to their role in the initial infection process2,6,8,10. Building on this previous work, our focus on the AM populations revealed that the CD38- AM subset dominates this early infection landscape. Specifically, they account for approximately 70% of the Mtb-infected host cells at 14 dpi (Figure 5A). To gain a more nuanced understanding of the CD38- AMs and the impact of their changing numbers, we employed differential abundance testing (DAB) to group previously identified neighbors based on their log-fold change (LFC) depletion rates, as an alternative approach to the traditional cluster-based categorization (Figure 5B)77. Through this approach, we identified two neighbor groups (group 2 and 5) exhibiting significant depletion rates in late infection (> 5 LFC). Examining the cell types associated with these high LFC depletion rate groups (FDR <0.1), we found that 98% of cells from group 2 belonged to the CD38- TR-AMs subsets (spanning CD38- TR-AM_1, CD38- TR-AM_2, CD38- TR-AM_3). For group 5, 57% were associated with the CD38- TR-AMs and 33% with CD38+ moAMs (Figure 5C). The over-representation of CD38- TR-AMs in groups marked by pronounced depletion rates, identified through our DAB approach and independently confirmed by compositional analysis through scCODA79 (Supp. File 1) , suggests an innate susceptibility of these cells to Mtb infection. This vulnerability becomes more obvious when compared with the unchanging relative abundance of CD38- TR-AMs in bystander populations across different infection timepoints (Figure 4A and 4D). Coupled with the observed decline in the overall count of infected AMs as the infection unfolds (Figure 4C), it becomes evident that very few CD38- TR-AMs become infected in the later stages. This underscores the idea that the CD38- TR-AM clusters, dominant at 14 dpi, are poorly equipped to survive Mtb infection. This hypothesis is supported by the absence of Nos2 mapping reads across all infection timepoints in CD38- TR-AMs (Figure 5D), excepting cells bordering the MoAM in the CD38- TR-AM_1 cluster (likely signifying a common scRNA-seq occurrence of cells being misattributed among neighboring clusters).
DGE analysis of the CD38- AMs further supports this hypothesis. We found CD38- AMs to be associated with known markers of AM populations (Chil3, Lpl, Marco, Mrc1, Trf)8,9 (Figure 5E). Examining the transcriptional profile of these populations we observed an upregulation of genes involved with lipid metabolism such as Fabp4, Mgll, and Lpl8,80. Additionally, we noted increased expression in genes linked to the electron transport chain (like mt-Nd1, mt-Nd2, mt-Nd4, mt-Cytb, and mt-Atp6) which play pivotal roles in cellular energy metabolism. This metabolic transcriptional signature was further complemented by the upregulation of genes connected to fatty acid uptake and transport (e.g., Dbi)81, lipid droplet formation (e.g., Cidec, Plin2)82,83, and protection against oxidative stress (e.g., Gpx1, Gpx4)84 (Supp Table 2). Collectively, this data indicates a shift towards fatty acid oxidation (FAO) metabolism. Importantly, the upregulation of genes involved in metabolic and homeostatic functions stands in contrast to a diminished expression of pro-inflammatory genes, critical in combating tuberculosis infection (Figure 5E) (Supp. Table 2). Moreover, Gene Ontology (GO) functional enrichment analysis of the transcriptional profile of CD38- AMs also supports these observations. The functional profile of the CD38- TR-AMs aligns with alveolar macrophage characteristics that have been shown to promote Mtb growth6. These include increased oxidative phosphorylation and fatty acid metabolism (Figure 5F). In summary, our data suggests that the broad range of anti-inflammatory responses often associated with AM population are mediated through CD38- TR-AMs, which are highly susceptible and sensitive to Mtb infection during the early phases of TB.
CD38+ AMs are key contributors in controlling tuberculosis infection.
We re-clustered the AM subsets based on their surface expression of CD38 protein. Consistent with the previous observations, CD38+ cells only constituted 12% of the total infected AMs at 2 weeks, but this proportion increased to 89% and 93% at the 4 and 6-week timepoints, respectively (Figure 5G). Intriguingly, DAB testing revealed unchanging abundance of infected CD38+ AMs across timepoints (Figure 4A), and mo_AMs and CD38+ TR-AMs constitute the majority of the infected alveolar macrophages in the later stages of infection (Figure 4B-4C). This stability in numbers, even amidst a pronounced decline in the overall infected AMs (Figure 4C), implies that these cellular subsets might be inherently more resilient to Mtb infection. This interpretation aligns with their observed Nos2 expression patterns. Both CD38+ TR-AMs and mo_AMs exhibit a bimodal Nos2 expression, with increasing cell counts and Nos2 expression intensity, suggesting their antimicrobial capacity increases over time (Figure 5D). The elevated CD38 expression and pro-inflammatory pathways in AMs in latter stages, as described earlier, are predominantly represented by these two CD38+ populations.
Trajectory and pseudotime analyses provide additional insights into these populations. While mo_AMs, CD38- TR-AM_1, and CD38- TR_AM_2 appear as mature endpoints (grey circles 12,18,17,7,5 and root 3, white circle), CD38+ TR-AMs seem to arise from enhanced polarization of the CD38- TR-AM_3 subset (root 1, white circle and branching point 12,17, black circles, Figure 2B). Thus, we hypothesize that the decreased presence of the CD38- TR-AM_3 subset in subsequent infection stages is likely a consequence of these cells transitioning to CD38+ TR-AMs, as highlighted in the subsequent paragraphs. Finally, enrichment analysis of the genes upregulated by CD38+ AMs reveals up-regulation of the same pathways that are associated with control of Mtb infection (Figure 5H). These findings suggest that the divergent behavior of CD38+ and CD38- AMs in response to Mtb infection is programmed into the cell subsets prior to infection, with the CD38+ AMs playing a potentially crucial role in managing Mtb growth and spread.
Intrapulmonary BCG vaccination amplifies the CD38+ macrophage population resulting in enhanced control of Mtb infection.
Recent investigations, including those by Mata et al. 202114, have shown that intrapulmonary BCG vaccination prior to Mtb infection induces protective responses in lung resident macrophages. To extend this observation, we assessed how pulmonary BCG administration modulates the responses of CD38+/- macrophages to Mtb infection. Mice were vaccinated intranasally with live BCG and after a two-month period, were infected with the reporter smyc’::mCherry/hspx’::GFP Mtb Erdman for a duration of two weeks. Subsequently, we integrated scRNA-seq datasets from these vaccinated mice with our existing timepoint analysis data.
Analysis of infected macrophage populations from both BCG-vaccinated (n=3931 cells) and unvaccinated (n=2278 cells) mouse lungs, two weeks post-Mtb infection, revealed marked differences. In the BCG-vaccinated cohort, the monocyte-derived macrophages, associated with hspx’::GFP-high Mtb (Figures 1F and 2E), constituted the vast majority of infected macrophages, at 74.43%. This is in stark contrast to the unvaccinated cohort, where such cells accounted for just 43.19% (Figure 6A). Flow cytometry data from the sorted populations reinforced these findings. Macrophages from the BCG-treated group mounted a more effective response, limiting Mtb replication more efficiently. This is underscored by a marked reduction of the mCherry signal in infected cells of BCG-vaccinated mice, which was approximately ~2 log10 times lower than that in infected cells from unvaccinated mice. Moreover, there was a substantial difference in the proportion of hspx’::GFPstressed bacteria between the two cohorts: GFP-high cells represented 22.6% of the total infected cells in the unvaccinated group compared to 77.78% in their BCG-vaccinated counterparts (Figure 6B). The distribution of hspx’::GFP+ cells between the two groups, as revealed in our scRNA-seq dataset (Supp. Figure 4B), aligned with our flow cytometry data. Complementing these findings, we observed a marked decrease in the overall percentage of infected cells in BCG treated mice (0.45%) compared to unvaccinated mice (2.08%) (Figure 6B), as also confirmed by independent flow cytometry analysis (Supp Figure 4A). This was also supported by an overall reduction in bacterial burden in BCG-treated mice, as assessed by CFU counts (Figure 6C). These differences in cellular responses between BCG-treated and unvaccinated mice highlight significant alterations in their macrophage populations. BCG-treated mouse lungs were dominated by highly pro-inflammatory monocytes, and there was a marked reduction in infected CD38- TR-AMs compared to unvaccinated controls (Figure 6D and 6E). The reduced proportion of infected CD38- TR-AMs in BCG-treated mice is likely the result of increased recruitment of pro-inflammatory monocytes, in addition to the transition of the CD38- TR-AMs_3 into their CD38+ counterparts (Figure 6E).
Using WGCNA, we identified a unique co-expression module of 463 genes exclusive to the macrophage populations of BCG-vaccinated mice (Supp. Figure 4C) (Supp. Table 3). This module was highly enriched in pro-inflammatory genes, including Nos2 and Cd38 (Supp Figure 4D). Further analysis revealed significant presence of genes involved in small GTPase signal transduction, encompassing Rab GTPases, guanine nucleotide exchange factors (GEFs) and GTPase activators (GAPs) (Supp. Figure 4E). Rab GTPases are known for their roles in endosomal trafficking, while GAPs and GEFs are vital for membrane transport, phagocytosis, and controlling of the actin cytoskeleton85-87. Their increased expression suggests modifications in intracellular trafficking within the BCG-treated macrophages. Altered Mtb intracellular trafficking and increased lysosomal fusion can limit bacterial growth 88-90 and amplify killing mechanisms due to immune activation and increased autophagy91-95. Our data confirms the previous findings as we observe that increased intracellular trafficking is tightly linked to increased expression of genes involved in lysosomal and autophagy functions in BCG-treated macrophages (Figure 6F). Additionally, genes facilitating macrophage migration and tissue invasion saw heightened expression in macrophages from BCG-treated lungs (Supp.Table 3). This observation aligns with the gene-set enrichment analysis of the 463 co-expressed genes from the scWGCNA module. Here, cell migration, motility, programmed cell death and autophagy emerged as the pathways enriched in BCG-treated macrophages, in both GO and KEGG analysis (Supp Figure 5A) (Supp Table 4).
The increased expression of pro-inflammatory gene signatures involved in control of Mtb infection in BCG-treated macrophages aligns with their observed phenotype. A hallmark feature across BCG-treated mice was the increased expression of CD38 (Supp. Figure 5B). This immune control phenotype was still evident at 4 weeks post-infection, corroborating Mata et al.'s findings (Supp Figure 5C). Intriguingly, we observed discrepancies between the two studies regarding the origin of infected macrophages after BCG vaccination. In our study, the majority of infected macrophages in BCG-treated mice were monocyte-derived, while previous data indicated Mtb was mostly confined to AMs after BCG vaccination14. We believe the disparity was due to weak CD64 staining on pro-inflammatory monocyte-derived cells (Supp Figure 5D). To test this hypothesis, we re-infected BCG-treated and control mice with a high infection dose (5x103 CFU), similarly to Mata et al.14 The published CD64 flow cytometry gating approach will identify most infected macrophages as AMs in BCG-treated mice14 (Supp Figure 5E). However, when gating only on SiglecF, expressed uniquely by AMs (Figure 1B), the results indicate the majority of infected cells are monocyte-derived (Supp Figure 5F), as in our scRNA-seq dataset. Regardless of the gating method, we found a consistent increase in the percentage of infected AMs in BCG-vaccinated mice compared to the unvaccinated counterparts, underscoring their improved ability to restrict Mtb growth, both at 2wpi (Supp Figure 5F) and at 4wpi (Supp figure 5G)14. This aligns with our scRNA-seq results and highlights the fundamental role of the increased polarization of the CD38- TR-AMs_3 transitioning to CD38+ TR-AMs for enhanced Mtb control in BCG-treated AMs (Figure 6E). In conclusion, our data suggests that the heightened defense against Mtb reinfection observed after intranasal BCG vaccination is due to both the increased presence of CD38+ monocyte-derived macrophages and the activation of resident CD38+ TR-AMs.
Pre-existing differences in the chromatin organization of CD38+ vs CD38- AMs are linked to differential responses to Mtb infection.
Our scRNA-seq analysis revealed distinct differences in AM populations between naïve and Mtb-infected mice. Specifically, we observed that CD38+ moAMs are recruited to the lung in response to Mtb infection (Figure 2A). In contrast, CD38+ TR-AMs were already present prior to infection, predominantly exhibiting a less active CD38- phenotype, which we previously defined as CD38- TR-AM_3 (Figure 2A and 2B).
To investigate the chromatin landscape and potential epigenetic regulation of AM subsets prior to infection, we performed scATAC-seq on CD45+ cells isolated from the lungs of naïve mice. Unbiased clustering based on differential chromatin accessibility identified 10 distinct clusters (Figure 7A). Integrating this scATAC-seq dataset with our timepoint-specific scRNA-seq data, which includes naïve, bystander, and infected cells, revealed that variations in chromatin organization before infection align closely with the diverse transcriptional phenotypes observed during tuberculosis infection. Using gene scores, we inferred the potential gene expression profiles for each cell in the scATAC-seq sample, based on the accessibility of regulatory elements adjacent to each gene. We then performed data integration with the scRNA-seq dataset, as previously described96 (Figure 7B).
Given that our scATAC-seq sample comprised only cells from naïve mice, we anticipated that the inferred gene expression profiles from the scATAC-seq dataset would predominantly align with those of naïve cells from our scRNA-seq dataset. Surprisingly, we found that the inferred gene expression of cells from cluster C7 mirrored that of the pro-inflammatory CD38+ subsets from our scRNA-seq data (Figure 7B and 7C), which by transcriptional profiling are not present in naïve mice (Figure 2A). In contrast, clusters C6 and C8 aligned with CD38- TR-AMs, while cluster C5 correlated with Mki67+ AMs (Figures 7B and 7C). To validate that cluster C7 represented CD38+ TR-AMs and not monocyte-derived AMs, we probed for open chromatin within the promoter regions of monocyte markers Mafb and Ly6a, whose expression is restricted to monocyte-derived macrophages in our scRNA-seq dataset, as noted earlier (Figure 1F). We found high levels of open chromatin for these markers only in monocyte-derived cells, but not in cluster C7 (Supp. Figure 6B).
To further understand why the inferred gene expression of cluster C7 aligns with pro-inflammatory populations in scRNA-seq, we first assessed transcription factor dynamics, performing marker peak and motif enrichment analysis (FDR < 0.1, log2FC > 0.5) to identify transcription factor binding sites (TFBS) that are enriched across the different scATAC-seq clusters. Cluster C7 exhibited highly significant enrichment for binding sites of transcription factors such as Smarcc1, Bach1, and Rela, known to drive pro-inflammatory gene activation in macrophages97-99 (Figure 7D). We further validated the expression of these TFs in our scRNA-seq dataset and found them to be uniquely expressed by CD38+ pro-inflammatory macrophages following Mtb infection (Figure 7E).
Finally, our analysis of open chromatin peaks within the regulatory regions (± 5k from TSS) of pro-inflammatory genes that define CD38+ AMs, such as Nos2, Cd38, Slc7a11, Ccl5, Ptgs2, Il1b, and Cxcl2, revealed higher open chromatin in cells from cluster C7 compared to clusters C6 and C8, further validating that CD38+ TR-AMs are pre-primed for a pro-inflammatory response (Supp Figure 6C).
Overall, our analysis demonstrates that CD38+ TR-AMs are present in naïve mouse lungs before infection, but are transcriptionally quiescent and have an epigenetic profile markedly distinct from their CD38- TR-AM counterparts. These CD38+ TR-AMs have increased chromatin accessibility in promoter regions of pro-inflammatory genes, with significant enrichment of transcription factor binding sites that have the potential to drive pro-inflammatory macrophage activity and control Mtb growth. Our findings support the hypothesis that the response of different TR-AM subsets to Mtb infection is largely predetermined by their intrinsic chromatin organization prior to infection.