To explore the spectrum of TNF responses in the B6 and B6.Sst1S BMDMs, we compared gene expression in the BMDM populations either naive, or after 24 hr of stimulation with TNF (10 ng/mL) using single-cell RNA sequencing (scRNA-seq; Figure 1B). The naive macrophage populations of both backgrounds were similar. All BMDMs responded to TNF stimulation, as evidenced by the de novo formation of clusters 1–6, where cluster 3 appeared exclusively in the resistant (B6) and clusters 4 and 5 in the susceptible (B6.Sst1S) macrophage populations (Figure 1C, Figure 1—figure supplement 1A). After TNF stimulation, the expression of the sst1-encoded Sp110 and Sp140 genes coordinately increased only in the B6 macrophages (Figure 1D, Figure 1—figure supplement 1B). The IFN pathway was dramatically upregulated in TNF-stimulated B6.Sst1S cells that lacked the Sp110 and Sp140 expression, but it was upregulated to a lesser degree in cluster 3 exclusive for TNF-stimulated Sp110/140-positive B6 macrophages (Figure 1E, Figure 1—figure supplement 1C).

To determine the relatedness of the diverse macrophage subpopulations, as they emerge during TNF stimulation, we performed trajectory analysis that clearly demonstrated partial overlap and the divergence between the trajectories of TNF-stimulated resistant (RT) and susceptible (ST) subpopulations (Figure 1F). Subpopulations sp1 and sp2 of the B6 and B6.Sst1S macrophages were closely related and were characterized by the upregulation of cell division pathways in both backgrounds (Figure 1G). Cell cycle analysis demonstrated that TNF stimulation increased the fraction of macrophages in S phase (Supplementary file 1).

The ST and RT activation trajectories diverged as the RT transitioned from sp2 to sp3, but the ST transitioned from sp2 to sp4 and 5. The transition from sp2 to sp3 in the resistant B6 BMDMs was characterized by an increase of the G1/S ratio, the downregulation of anabolic pathways involved in cell growth and replication (E2F and Myc), and the upregulation of anti-oxidant genes (Figure 1G and H). In contrast, in the susceptible B6.Sst1S macrophages, the sp2 transitioned to subpopulations sp4 and sp5 that were characterized by the decreased G1/S ratios as compared to sp3, signifying an increased G1 – S transition (Supplementary file 2). The sp4 represented an intermediate state between sp2 and sp5 and was primarily characterized by the upregulation of the IFN pathway (Figure 1G), as evidenced by the upregulation of known IFN-I pathway activation markers Rsad2, IL1rn, Cxcl10; (Figure 1—figure supplement 1C and D).

The ST macrophages in sp5 uniquely demonstrated a coordinated upregulation of the IFN, TGFβ, Myc, E2F, and stress response (p53 and UPR) pathways (Figure 1G). They upregulated the ISR genes (Ddit3/Chop10, Atf3, Ddit4, Trib1, Trib3, Chac1) in parallel with markers of immune suppression and apoptosis (Cd274/PD-L1, Fas, Trail/Tnfsf10, Id2, and a pro-apoptotic ligand – receptor pair Tnfsf12/Tweak and Tnfrsf12a/Tweak receptor; Figure 1—figure supplement 1D). Of note, IFN-I pathway genes were upregulated in all TNF-stimulated B6.Sst1S macrophages in agreement with the paracrine effect of Ifnβ, whose increased production by TNF-stimulated B6.Sst1S macrophages we have described previously (Bhattacharya et al., 2021). Interestingly, the expression of Irf7 and several IFN-inducible genes, such as B2m, Cxcl10, and Ube2l6 was reduced in the sp5 cells, suggesting partial dampening of their IFN-I responsiveness (Figure 1—figure supplement 1C).

Taken together, the single-cell trajectory analysis revealed that sp5 represented a terminal state of the aberrant B6.Sst1S macrophage activation by TNF that was characterized by the coordinated upregulation of stress, pro-apoptotic, and immunosuppression genes. Unexpectedly, the stress escalation coincided with paradoxical activity of Myc and E2F pathways. Transition to this state in the susceptible macrophage population was preceded by IFN-I pathway upregulation (sp2 – sp4). In contrast, the sp2 – sp3 transition in the wild type macrophages was coincident with upregulation of both the Sp110 and Sp140 genes and was accompanied by the termination of cell cycle and the upregulation of antioxidant defense pathways. Therefore, we concluded that in resistant TNF-stimulated macrophages, the sst1 locus-encoded genes promoted the activation of the AOD pathway either directly or by suppressing the IFN-I pathway.

To begin exploring the hierarchy and crosstalk of these pathways, we used an unbiased computational approach to define the Sp110 and Sp140 regulatory networks. First, we inferred a mouse macrophage gene regulatory network using the GENIE3 algorithm (Huynh-Thu and Geurts, 2018) and external gene expression data for mouse macrophages derived from Gene Expression Omnibus (GEO) (Clough and Barrett, 2016). This network represents co-expression dependencies between transcription factors and their potential target genes, calculated based on mutual variation in expression level of gene pairs (Huynh-Thu et al., 2010; Zhernovkov et al., 2019). This analysis revealed that in mouse macrophages, the Sp110 and Sp140 genes co-expressed with targets of Nfe2l1/2 (Nuclear Factor Erythroid 2 Like 1/2) and Mtf (metal-responsive transcription factor) TFs that are involved in regulating macrophage responses to oxidative stress and heavy metals, respectively (Figure 1I). Taken together, our experimental data and the unbiased network analysis suggested that in TNF-stimulated macrophages, the sst1-encoded Sp110 and/or Sp140 gene(s) might be primarily involved in regulating AOD.

Next, we compared the expression of upstream regulators of AOD in B6 and B6.Sst1S macrophages during TNF activation. Our previous studies demonstrated that the earliest differences between the B6 and B6.Sst1S BMDMs occurred between 8 and 12 hr of TNF stimulation, concomitant with the upregulation of the Sp110 protein in the B6 macrophages and heat shock proteins in the mutant cells (Bhattacharya et al., 2021). Comparing the time course of major transcriptional regulators of AOD in TNF-stimulated B6 and B6.Sst1S macrophages during this critical time interval, we observed higher upregulation of Nrf2 protein in TNF-stimulated B6 BMDMs (Figure 2A). The Nrf1 levels were not substantially upregulated after TNF stimulation and were similar in B6 and B6.Sst1S BMDMs (Figure 2—figure supplement 1A and B).

The Nrf2 difference was observed both in cytoplasmic and nuclear fractions (Figure 2B and C, respectively). The levels of Nrf2 negative regulators Keap1 and β-TrCP (Figure 2—figure supplement 1A), and Bach1 (Figure 2B and C) were similar in both backgrounds and did not notably change after TNF stimulation. Quantitative microscopy confirmed that at 12 hr of TNF stimulation, the cytoplasmic and nuclear Nrf2 levels significantly increased in B6 but not in B6.Sst1S BMDMs (Figure 2D and E and Figure 2—figure supplement 1C). In contrast, the levels of Nfe2l1/2 mRNA induced by TNF were higher in the mutant macrophages, suggesting post-transcriptional regulation (Figure 2F). Therefore, we measured the rates of Nrf2 protein degradation 6–8 hr after TNF stimulation but found no difference (Figure 2G and H).

Using EMSA, we demonstrated that binding activity of nuclear Nrf2 to its target DNA at 8 and 12 hr after TNF stimulation was greater in the resistant B6 BMDMs (Figure 2I). To identify core pathways controlled by the sst1 locus during this critical period, we compared global mRNA expression profiles of the B6 and B6.Sst1S macrophages after 12 hr of TNF stimulation using bulk RNA-seq. This analysis confirmed that the Sp110 and Sp140 genes were strongly upregulated by TNF stimulation exclusively in the B6 macrophages (Figure 2J). Gene set enrichment analysis (GSEA) of genes differentially expressed between TNF-stimulated B6.Sst1S and B6 macrophages at this critical junction revealed that the IFN response, Myc, E2F target gene, Hypoxia, UV response, and DNA repair pathways were upregulated in the mutant macrophages, while the detoxification of reactive oxygen species, cholesterol homeostasis, fatty acid metabolism, and oxidative phosphorylation, peroxisome, and lysosome pathways were downregulated (Supplementary file 3). Functional pathway profiling using KEGG and Reactome databases also highlighted the upregulation of genes involved in oxidative stress and cellular senescence in mutant macrophages. In contrast, the wild-type macrophages upregulated genes involved in detoxification of reactive oxygen species, inhibition of ferroptosis, and peroxisome function (Supplementary file 4). Supporting these findings, the total antioxidant capacity of the B6 macrophages after TNF stimulation increased to significantly higher levels, as compared to the B6.Sst1S (Figure 2—figure supplement 1D).

Transcription factor binding site analysis of genes specifically upregulated by TNF in B6, but not B6.Sst1S, macrophages (B6-specific cluster) revealed an enrichment of Nfe2l1/Nfe2l2, Bach1, and Mafk sequence motifs, that is binding sites of transcription factors regulating AOD. In contrast, overrepresentation of E2F, Egr1, and Pbx3 transcription factor binding sites was found for genes in the Sst1S-specific cluster (Supplementary file 5). A master regulator analysis using Virtual Inference of Protein Activity by Enriched Regulon Analysis (VIPER) algorithm also revealed a key role for Nfe2l1/2 (NF-E2-like) transcription factors (TFs) as regulators of genes differentially induced by TNF in B6 and B6.Sst1S BMDMs (Supplementary file 6).

To further investigate this inference, we analyzed the expression of a gene ontology set ‘response to oxidative stress’ (GO0006979, 416 genes) and observed clear separation of these genes in two clusters in a sst1-dependent manner (Figure 2J and Figure 2—figure supplement 1G). This analysis demonstrated that the response to oxidative stress in the sst1 mutant macrophages was dysregulated, but not paralyzed. For example, the upregulation of well-known Nrf2 target genes Heme oxygenase 1 (Hmox1) and (NAD(P)H quinone dehydrogenase 1) (Nqo1) was similar in B6 and B6.Sst1S BMDMs (Figure 2—figure supplement 1E and F, respectively).

A subset of antioxidant defense genes whose expression was concordant with Sp110 and Sp140 in B6 macrophages represented genes that are known to be involved in iron storage (ferritin light and heavy chains, Ftl and Fth), ROS detoxification and maintenance of redox state (Cat, G6pdx, Sod2, Gstm1, Gpx4, Prdx6, Srxn1, Txn2, and Txnrd1; Figure 2J). We hypothesized that their coordinate downregulation in TNF-stimulated B6.Sst1S macrophages sensitized the mutant cells to iron-mediated oxidative damage and played a pivotal role in shaping their divergent activation trajectory.

To test this hypothesis, first we explored the intracellular iron storage. Both the ferritin light (Ftl) and heavy (Fth) chain genes were dysregulated in TNF-stimulated B6.Sst1S BMDMs. While the Fth mRNA was upregulated by TNF in B6 BMDMs, it remained at a basal level in the B6.Sst1S cells (Figure 3A). The Ftl mRNA level was significantly reduced after TNF stimulation in B6.Sst1S macrophages but remained at the basal level in B6 (Figure 3B). Accordingly, the expression of Ftl protein was reduced in B6.Sst1S BMDMs after 12 hr of TNF stimulation, and both Ftl and Fth proteins were reduced at 24 hr (Figure 3C). In parallel, the levels of Gpx1 and Gpx4 proteins were also substantially reduced by 24 hr in B6.Sst1S (Figure 3D). The glutathione peroxidase 4 (Gpx4) protein plays a central role in preventing ferroptosis because of its unique ability to reduce hydroperoxide in complex membrane phospholipids and, thus, limit self-catalytic lipid peroxidation (Stockwell et al., 2017). Thus, TNF-stimulated B6.Sst1S BMDMs had reduced intracellular iron storage capacity accompanied by the decline of the major lipid peroxidation inhibitor Gpx4. Accordingly, we observed increases in an intracellular labile iron pool (LIP, Figure 3E), an intracellular accumulation of oxidized lipids (Figure 3F), a toxic terminal lipid peroxidation (LPO) products malondialdehyde (MDA, Figure 3G) and 4-hydroxynonenal (4-HNE) adducts (Figure 3H, Figure 3—figure supplement 1A). Treatment with the LPO inhibitor Ferrostatin-1 (Fer-1) prevented the 4-HNE adducts accumulation (Figure 3H, Figure 3—figure supplement 1F). The levels of the LIP and LPO remained significantly elevated in B6.Sst1S macrophages after 48 hr of TNF stimulation (Figure 3—figure supplement 1B-F). We also observed an increased ROS production in the B6.Sst1S macrophages during persistent TNF stimulation (Figure 3I and J, and Figure 3—figure supplement 1G). The ASK1 - JNK – cJun stress kinase axis was upregulated in B6.Sst1S macrophages, as compared to B6, 12–36 hr of TNF stimulation, that is during the aberrant activation stage confirming persistent stress (Figure 3K). By 48 hr of TNF stimulation, we noted moderate cell death in B6.Sst1S macrophage cultures (Figure 3L, Figure 3—figure supplement 1H). Treatments with Fer-1, the antioxidant butylated hydroxyanisole (BHA), or IFNAR1 blockade each prevented the cell death (Figure 3M), suggesting that both persistent stress and macrophage death were mediated by ROS, lipid peroxidation, and IFN-I pathway.

Next, we wanted to test whether the IFN-I pathway hyperactivity in TNF-stimulated B6.Sst1S macrophages was responsible for the initial dysregulation of the Ftl, Fth, and AOD gene expression, that is at 8–12 hr of TNF stimulation. The IFNAR1 blockade, however, did not restore the Nrf2 and Ftl protein levels, or the Fth, Ftl, and Gpx1 mRNA levels to the wild type B6 levels (Figure 4A-C). Additionally, we observed that LPO production was increased as early as 6 hr under TNF-stimulated condition in both B6 and B6.Sst1S prior to Ifnb1 superinduction (Figure 4—figure supplement 1A and B). These data suggest that type I IFN signaling does not initiate LPO in our model.

Of note, the IFNAR1 blockade did not prevent the Ifnb1 mRNA superinduction (Figure 4D), thus rejecting a hypothesis that the Ifnb1 superinduction in B6.Sst1S macrophages was driven via an Ifnβ – IFNAR1 positive feedback (Ivashkiv and Donlin, 2014). The mRNA expression of the interferon-inducible gene Rsad2, however, was suppressed, demonstrating the efficiency of the IFNAR1 blockade (Figure 4E).

In contrast, treatment of B6.Sst1S macrophages with Fer-1 or the iron chelator DFO during initial TNF stimulation inhibited both the Ifnb1 and Rsad2 mRNAs upregulation (Figure 4F, Figure 4—figure supplement 1C respectively). Importantly, Fer-1 treatment also reduced the Ifnb1 and Rsad2 levels in B6.Sst1S macrophages when added at 18 hr after TNF stimulation, that is during established aberrant response (Figure 4G, Figure 4—figure supplement 1D). Thus, lipid peroxidation was involved in both the initial Ifnb1 superinduction and in maintenance of the and IFN-I pathway hyperactivity driven by prolonged TNF stimulation.

 Next, we wanted to test whether continuous IFN-I signaling was required for the accumulation of 4-HNE adducts during prolonged macrophage activation. Indeed, the blockade of type I IFN receptor (IFNAR1) after 2, 12, or 24 hr after TNF stimulation prevented the 4-HNE adducts accumulation at 48 hr (Figure 4H and I, and Figure 4—figure supplement 1E-G). Thus, Ifnb1 super-induction and IFN-I pathway hyperactivity in B6.Sst1S macrophages follow the initial LPO production and maintain and amplify it during prolonged TNF stimulation.

The sst1-encoded Sp110 and Sp140 genes were described as interferon-induced genes, and Sp140 protein was implicated in maintenance of heterochromatin silencing in activated macrophages (Fraschilla and Jeffrey, 2020a; Mehta et al., 2017). Therefore, we hypothesized that their deficiency in the TNF-stimulated B6.Sst1S mutants may lead to the upregulation of silenced transposons and, thus, trigger the Ifnβ upregulation via intracellular RNA sensors, that is by an autonomous mechanism unrelated to the AOD dysregulation. We examined the transcriptomes of B6 and B6.Sst1S macrophages before and after TNF stimulation for the presence of persistent viruses or transposons using a custom bioinformatics pipeline (Ma et al., 2021). No exogenous mouse viral RNAs were detected. A select set of mouse LTR-containing endogenous retroviruses (ERVs; Jayewickreme et al., 2021), and non-retroviral LINE L1 elements were expressed at a basal level before and after TNF stimulation, but their levels in the B6.Sst1S BMDMs were similar to or lower than those seen in B6 (Figure 4—figure supplement 2A-C). We also tested the accumulation of dsRNA using deep sequencing of macrophage small RNAs and failed to detect evidence of transposon-derived dsRNAs (Figure 4—figure supplement 2D). We concluded from these findings that the majority of the basal transposon RNAs in macrophages exist primarily as single-stranded mRNAs that evade triggering interferon pathway. The above analyses allowed us to exclude the overexpression of persistent viral or transposon RNAs as a primary mechanism of the IFN-I pathway hyperactivity.

Taken together, the above experiments allowed us to reject the hypothesis that IFN-I hyperactivity caused the sst1-dependent AOD dysregulation. In contrast, they established that the hyperactivity of the IFN-I pathway in TNF-stimulated B6.Sst1S macrophages was itself driven by the initial dysregulation of AOD and iron-mediated lipid peroxidation. During prolonged TNF stimulation, however, the IFN-I pathway was upregulated, possibly via ROS/LPO-dependent JNK activation, and acted as a potent amplifier of lipid peroxidation.

We wanted to determine whether the AOD genes were regulated by the sst1-encoded genes directly or indirectly via an intermediary regulator. Previously, we identified two transcription factors whose binding activities were exclusively upregulated in the susceptible macrophages after 12 hr of TNF stimulation: Myc and HSF1 (Bhattacharya et al., 2021). The bulk RNA-seq analysis at this timepoint (12 hr) also demonstrated the upregulation of Myc pathway along with E2F target genes and stress responses specifically in TNF-stimulated B6.Sst1S BMDMs, as compared to B6 (Figure 2). Of note, the scRNA-seq analysis above also demonstrated the association of Myc and stress response pathways in the mutant cells at 24 hr (Figure 1). Therefore, we hypothesized that in susceptible macrophages, Myc might be involved in the dysregulation of AOD and iron storage.

 First, we observed that Myc was regulated in an sst1-dependent manner: in TNF-stimulated B6 wild-type BMDMs, c-Myc mRNA was downregulated, while in the susceptible macrophages, c-Myc mRNA was upregulated (Figure 5A). The c-Myc protein levels were also higher in the B6.Sst1S cells in unstimulated BMDMs and 6–12 hr of TNF stimulation (Figure 5B). Next, we tested whether suppression of Myc activity could ‘normalize’ the susceptible phenotype using Myc-Max dimerization inhibitor 10058-F4 (F4). Indeed, this treatment increased the levels of Fth and Ftl proteins in TNF-stimulated susceptible macrophages (Figure 5C) and decreased the LIP (Figure 5D). Accordingly, the levels of MDA, oxidized lipid,s and 4-HNE adducts also significantly decreased (Figure 5E-G, Figure 5—figure supplement 1A and B), as well as the levels of Ifnb1, Rsad2, and the ISR markers Trib3 and Chac1 (Figure 5H). Possibly, the ISR activation serves as an alternative pathway of Myc inhibition, as it is known to inhibit the oncogene protein translation (Wolfe et al., 2014).

 Next, we wanted to determine whether the upregulation of Myc is driven by TNF alone or in synergy with CSF1, a growth factor that also stimulates Myc. In vitro, we observed the upregulation of Myc shortly after the addition of fresh CSF1-containing media, but no difference in the Myc protein dynamics between B6 and B6.Sst1S BMDMs in the absence of TNF (Figure 5—figure supplement 1C). In addition, we tested whether CSF1R inhibitors could prevent the superinduction of Ifnb1 and Rsad2 mRNAs in TNF-stimulated B6.Sst1S macrophages. The CSF1R inhibitors were used at concentrations that did not cause macrophage death (Figure 5—figure supplement 1D). Neither of these inhibitors prevented the Ifnb1 and Rsad2 upregulation (Figure 5—figure supplement 1E and F). Because Myc induction by CSF1 and TNF is conducted via distinct relays of receptor signaling and transcription factors, we concluded that the sst1 locus specifically controls Myc expression induced by inflammatory signaling. Indeed, Myc promoter has multiple predicted NF-κB and/or AP-1 transcription factors binding sites. To test this hypothesis, we used specific JNK inhibitor D-JNK1, but it only partially reduced Myc protein levels in TNF-stimulated B6.Sst1S macrophages (Figure 5I and J). Because both Sp110 and Sp140 mRNAs and proteins are upregulated during extended TNF stimulation, one of them may participate in feedback regulation of TNF-induced Myc, either directly or via yet unknown intermediates.

Next, we tested whether the described facets of the aberrant macrophage activation conferred by the sst1S allele were relevant to Mtb susceptibility. After macrophage infection with virulent Mtb in vitro, gradual accumulation of LPO product 4-HNE adducts was observed in BMDMs of both B6 and B6Sst1S genetic backgrounds at 3–5 days post infection (dpi). It occurred either in the presence or absence of exogenous TNF (Figure 6A, and Figure 6—figure supplement 1A and B). By day 5 post infection, TNF stimulation significantly increased LPO accumulation only in the B6.Sst1S macrophages (Figure 6A). Both Mtb-infected and bystander non-infected B6.Sst1S macrophages showed 4-HNE adduct accumulation (Figure 6B).

TNF stimulation improved the control of Mtb growth in B6 but not in B6.Sst1S macrophages, as quantified using an Mtb replication reporter strain (Figure 6B, Figure 6—figure supplement 1C and D) and quantitative PCR of Mtb genomes (Figure 6C). The LPO inhibition using Fer-1 improved the survival of the Mtb-infected BMDMs (Figure 6D-E) and prevented the intracellular Mtb growth (Figure 6F) during the five-day in vitro infection. The iron chelator DFO also significantly reduced the Mtb growth, although restricting iron availability may also directly affect the bacterial replication. Importantly, the survival of Mtb-infected BMDMs was also improved, and the bacterial loads were significantly reduced by the Myc inhibitor, F4 (Figure 6G and H). Of note, the effects of the lipid peroxidation inhibitors became prominent between days 3 and 5 post infection (Figure 6—figure supplement 1E), suggesting that these inhibitors do not directly boost the bacterial killing by activated macrophages, but rather prevent macrophage damage.

Next, we tested whether the sst1 susceptible allele compromised responsiveness of Mtb-infected macrophages to mycobacteria-specific T cells. The immune T cells were isolated from the regional lymph nodes of the resistant B6 mice vaccinated with live attenuated BCG vaccine and added either to the B6 or B.Sst1S BMDM monolayers infected with Mtb the day before. The BMDMs were either treated with TNF prior to infection or not. After co-culture with the immune T cells for 1 or 2 days, Mtb loads were significantly reduced in T cell co-cultures with the resistant B6 macrophages. The susceptible B6.Sst1S BMDMs did not respond to the same T cells either in the presence or absence of exogenous TNF (Figure 6I-J and Figure 6—figure supplement 1F). The B6.Sst1S BMDM responsiveness to T cells was improved by inhibitors of lipid peroxidation (Fer1) and Myc (F4): the bacterial loads were significantly reduced 48 hr after co-culture with the immune T cells (Figure 6K, Figure 6—figure supplement 1G). This T-cell-mediated Mtb control was specific for T cells isolated from BCG-immunized mice (Figure 6—figure supplement 1H). The IFNAR1 blockade improved the ability of TNF-stimulated B6.Sst1S macrophages to control Mtb with and without T help (Figure 6L, Figure 6—figure supplement 1I). It also restored their responsiveness to soluble IFN-γ that was inhibited by pre-stimulation with TNF (Figure 6M). These data demonstrate that during prolonged TNF stimulation of B6.Sst1S macrophages, the Myc-driven lipid peroxidation and subsequent IFN-I hyperactivity compromise both the cell autonomous and T-cell-mediated Mtb control.

We wanted to determine whether the aberrantly activated macrophages accumulate within pulmonary TB lesions in vivo. We used a mouse model of pulmonary TB where the lung lesions develop after hematogenous spread from the primary site of infection and progress exclusively in the lungs, despite systemic immunity and control of infection in other organs. Microscopic pulmonary lesions develop in the lungs of both B6 and B6.Sst1S mice, but advanced multibacillary TB lesions develop exclusively in the B6.Sst1S (Yabaji et al., 2025a; Yabaji et al., 2025b).

Based on Mtb loads, TB lesions were classified in two categories: the Mtb-controlling paucibacillary lesions and multibacillary lesions, in which the control of Mtb growth was compromised (Figure 7—figure supplement 1A and B). The 4-HNE adduct levels dramatically increased in multibacillary lesions (Figure 7A and Figure 7—figure supplement 2A). The majority of the 4-HNE + cells were CD11b + myeloid cells (Figure 7B and Figure 7—figure supplement 2B).

To characterize macrophages in TB lesions and identify pathways associated with the loss of Mtb control, we performed spatial transcriptomics analysis of intralesional macrophages (Iba1+) using the Nanostring GeoMX Digital Spatial Profiler (DSP) system (Merritt et al., 2020; Figure 7—figure supplement 3A-C).

Comparing the Iba1 + macrophage gene expression profiles in the multibacillary vs paucibacillary lesions, we identified 192 upregulated and 376 downregulated genes at a two and above fold change (Supplementary file 7). Pathway analysis demonstrated a highly significant upregulation of Hypoxia, TNF, and IL6/STAT3 Signaling, Glycolysis, Complement, and Coagulation pathways in the multibacillary lesions consistent with the escalation of hypoxia, inflammation, and macrophage activation in the advanced lesions. Mechanistically, top transcription factors associated with genes upregulated in the multibacillary lesions were NFKB1, JUN, STAT1, STAT3, and SP1 (Figure 7—figure supplement 4A-C).

To specifically interrogate the interferon pathways in paucibacillary vs multibacillary lesions, we compiled a list of 430 interferon type I and type II inducible genes from public databases that were also included in Nanostring Whole Transcriptome Analysis (WTA) probes. Among these, 70 genes were differentially regulated between the Iba1 + macrophages in multi- vs paucibacillary lesions (Figure 7C). Among the upregulated genes were metalloprotease Mmp12, IL6, and Socs3, chemokines Cxcl10, Ccl2, Ccl3, Ccl4, Ccl19, cell stress and senescence marker p21 (Cdkn1a), and many known IFN-I target genes. The most upregulated pathways in Iba1 + cells within multibacillary lesions were interferon, TNF, and IL6/STAT3 signaling, and top transcription factors associated with upregulation of these pathways were NFKB1, STAT3, STAT1, and IRF1 (Figure 7—figure supplement 4D and E).

Among the IFN-inducible genes upregulated in paucibacillary lesions were Ifi44l, a recently described negative regulator of IFN-I that enhances control of Mtb in human macrophages (DeDiego et al., 2019; Jiang et al., 2021) and Ciita, a regulator of MHC class II inducible by Ifnγ, but not IFN-I (Supplementary file 8). Thus, the loss of local Mtb control in advanced pulmonary TB lesions was associated with the accumulation of aberrantly activated macrophages that contained lipid peroxidation products, with upregulation of the IFN-I pathway and downregulation of Ifnγ-inducible genes.

To detect the Ifnβ-expressing cells within TB lesions, we introduced the Ifnb1-YFP reporter described previously (Scheu et al., 2008) in the B6.Sst1S background. The B6.Sst1S,Ifnb1-YFP reporter mice were infected with virulent Mtb (smyc':: mCherry) constitutively expressing the red fluorescent protein reporter (Lavin and Tan, 2022). We validated the Ifnβ reporter in vitro using co-staining of TNF-stimulated B6.Sst1S, Ifnb1-YFP BMDMs with YFP-specific antibodies (Figure 7—figure supplement 5A). We demonstrated the accumulation of Ifnβ-expressing cells in pulmonary TB lesions of B6.Sst1S, Ifnb1-YFP mice using both the YFP reporter and in situ hybridization with the Ifnb1 probe (Figure 7—figure supplement 5B). The majority of the YFP + cells in TB lesions were Iba1 + macrophages a fraction of which were iNOS positive (Figure 7D, Figure 7—figure supplement 6A). The mice with paucibacillary lesions had 62% and multibacillary lesions had 80% activated macrophages expressing YFP. These findings clearly demonstrated the production of Ifnβ by activated M1-like macrophages.

Next, we assessed the expression of stress markers phospho-cJun and Chac1 in total, IFN-I producing, and activated macrophages in TB lesions. These markers were primarily expressed by activated macrophages (Iba1+) expressing iNOS and/or Ifnβ (YFP+)(Figure 7E, Figure 7—figure supplement 6B-F). We also documented the upregulation of PKR in the multibacillary lesions, which is consistent with the upregulation of the upstream IFN-I and downstream Integrated Stress Response (ISR) pathways, as described in our previous studies (Bhattacharya et al., 2021; Figure 7—figure supplement 7A).

 Thus, progression from the Mtb-controlling paucibacillary to non-controlling multibacillary TB lesions in the lungs of TB-susceptible mice was mechanistically linked with a pathological state of macrophage activation characterized by escalating stress (as evidenced by the upregulation of phospho-cJun, PKR, and Chac1), the upregulation of Ifnβ and the IFN-I pathway hyperactivity, with a concurrent reduction of Ifnγ responses. In our in vitro experiments, these stressed macrophages were unresponsive to T cell help (Figure 6J). Consequently, the administration of BCG vaccine to Mtb-infected B6.Sst1S mice did not increase their survival (Figure 7—figure supplement 7B).

 Next, we tested whether the upregulation of Myc pathway is associated with pulmonary TB progression in human TB patients. We used blood samples obtained from 41 individuals recently (<90 days) diagnosed with TB that were enrolled in the Regional Prospective Observational Research for Tuberculosis (RePORT)-India consortium. Patients were infected with drug-susceptible Mtb and treated with rifampicin, isoniazid, ethambutol, and pyrazinamide, per Technical and Operational Guidelines for TB Control by the Ministry of Health and Family Welfare, Government of India, 2016. Blood samples were collected before or within a week of the antibiotic treatment commencement. Patients were monitored for two years during and post-treatment. Individuals who failed the antibiotic treatment (n=21) were identified by positive sputum culture or clinical diagnosis of symptoms at any time after 4 full months of treatment and symptoms determined to not be from another cause. Cured TB patients (controls) remained culture-negative and symptom-negative for the 2-year observation period.

Several oncogene signatures identified previously (Bild et al., 2006) were analyzed with the TBSignatureProfiler (Johnson et al., 2021) to determine their ability to differentiate between treatment failures and controls based on bootstrapped AUC scores. Myc_up ranked within the top 3 signatures with an AUC of 0.74 and p-value of 0.008 (Figure 8A). Boxplots of ssGSEA scores were created to determine whether the myc signatures were differentially enriched between treatment failures and controls. Myc_upregulated genes were enriched in the treatment failure group relative to the treatment control group (Figure 8B, Supplementary file 9). The ROC curve used to generate the AUC scores for myc_up is depicted in Figure 8C and a heatmap of the genes used to generate the ssGSEA scores is depicted in Figure 8D, with ssGSEA scores of each patient sample. These data indicate that the upregulation of Myc pathway in peripheral blood cells of TB patients was associated with poor prognosis and TB persistence even in patients infected with antibiotic-sensitive Mtb. Although pathological evaluation was not included in this study, treatment failures in immunocompetent patients are often associated with massive fibro-necrotic pulmonary TB lesions.

Mouse strains (B6J.C3-Sst1^C3HeB/FejKrmn, B6.Sst1S,Ifnb1-YFP) and all unique/stable reagents generated in this study are available from the Lead Contact with a completed Materials Transfer Agreement. Mouse strain (B6J.C3-Sst1^C3HeB/FejKrmn is available from https://www.mmrrc.org (Stock No: 043908-UNC)).

We support inclusive, diverse, and equitable conduct of research.

Recombinant mouse TNF (Cat# 315–01 A) and IL-3 (Cat# 213–13) were procured from Peprotech. The mouse monoclonal antibody to mouse TNF (MAb; clone XT22), isotype control, and mouse monoclonal antibody to mouse Ifnβ (Clone: MAR1-5A3) were obtained from Thermo Fisher Scientific. BHA (Cat# B1253) and Deferoxamine mesylate (Cat#D9533) were sourced from Sigma Aldrich. PLX3397 (Cat# S7818), Myc inhibitor (10058-F4) (Cat# S7153), and ferrostatin-1 (Cat# S7243) were purchased from Selleckchem. D-JNK1 (Cat# HY-P0069), GW2580 (Cat# HY-10917), and BLZ945 (Cat# HY-12768) were acquired from Med Chem Express. Fetal bovine serum (FBS) for cell culture medium was sourced from HyClone. Middlebrook 7H9 and 7H10 mycobacterial growth media were purchased from BD and prepared following the manufacturer’s instructions. A 50 µg/mL hygromycin solution was utilized for reporter strains of Mtb Erdman (SSB-GFP, smyc′::mCherry).

C57BL/6 J inbred mice were obtained from the Jackson Laboratory (Bar Harbor, Maine, USA). The B6J.C3-Sst1^C3HeB/Fej Krmn congenic mice were created by transferring the sst1 susceptible allele from C3HeB/FeJ mouse strain on the B6 (C57BL/6 J) genetic background using twelve backcrosses (referred to as B6.Sst1S in the text). The B6.Sst1S,Ifnb1-YFP mice were produced by breeding B6.Sst1S mice with a reporter mouse containing a Yellow Fluorescent Protein (YFP) reporter gene inserted after the Ifnb1 gene promoter (Scheu et al., 2008). The YFP serves as a substitute for Ifnb1 expression. All animal procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of Boston University (PROTO201800218). Mice were euthanized by CO₂ asphyxiation in accordance with IACUC-approved protocols.

The isolation of mouse bone marrow and the cultivation of bone marrow-derived macrophages (BMDMs) from C57BL/6 J and B6.Sst1S were conducted following the procedures outlined in a prior study (Yabaji et al., 2022). BMDMs were plated in tissue culture plates, followed by treatment with TNF and incubation for 16 hr at 37 °C with 5% CO2. Various inhibitors were applied to the cells until the point of harvest.

For infection experiments, M. tuberculosis H37Rv (Mtb) was cultured in 7H9 liquid media for 3 days and subsequently harvested. The bacteria were then diluted in media containing 10% L929 cell culture medium (LCCM) without antibiotics to achieve the desired Multiplicity of Infection (MOI). Following this, 100 μL of Mtb-containing media with the specified MOI was added to BMDMs cultivated in a 96-well plate format that had been pre-treated with TNF or inhibitors. The plates underwent centrifugation at 500 x g for 5 min, followed by an incubation period of 1 hr at 37 °C. To eliminate any extracellular bacteria, cells were treated with Amikacin at 200 μg/μL for 45 min. Subsequently, cells were washed and maintained with inhibitors, as applicable, in DMEM/F12 containing 10% FBS medium without antibiotics at 37 °C in 5% CO2 for the duration of each experiment. Media changes and inhibitor replacements were carried out every 48 hr. The confirmation of MOIs was achieved by counting colonies on 7H10 agar plates. All procedures involving live M. tuberculosis were conducted within Biosafety Level 3 containment, in accordance with Boston University Institutional Biosafety Committee (IBC) approved protocol #25–875, and adhering to regulations from Environmental Health and Safety at the National Emerging Infectious Disease Laboratories, the Boston Public Health Commission, and the Centers for Disease Control and Prevention.

The cytotoxicity, cell loss, and Mtb growth assays were conducted in accordance with the procedures outlined by Yabaji et al., 2022. In brief, BMDMs were subjected to inhibitor treatment or Mtb infection as previously described for specified time points. Upon harvesting, samples were treated with Live-or-Dye Fixable Viability Stain (Biotium) at a 1:1000 dilution in 1 X PBS/1% FBS for 30 min. Following staining, samples were carefully washed to prevent any loss of dead cells from the plate and subsequently treated with 4% paraformaldehyde (PFA) for 30 min. After rinsing off the fixative, samples were replaced with 1 X PBS. Following decontamination, the sample plates were safely removed from containment. Utilizing the Operetta CLS HCA System (PerkinElmer), both infected and uninfected cells were quantified. The intracellular bacterial load was assessed through quantitative PCR (qPCR) employing a specific set of Mtb and M. bovis-BCG primer/probes, with BCG spikes included as an internal control, as previously described (Yabaji et al., 2022). The percentage of cell numbers was calculated based on the number of cells at Day 0 (immediately after Mtb infection). The fold change of Mtb was calculated after normalization using Mtb load at Day 0 after infection and washes. We quantified the percentage of replicating Mtb using a replication reporter strain, Mtb Erdman (SSB-GFP, smyc′::mCherry). Replicating Mtb was identified by counting the number of SSB-GFP puncta-positive Mtb, while the total number of Mtb per field was determined by counting mCherry-expressing bacteria (red).

The extraction of total RNA was carried out utilizing the RNeasy Plus mini kit (QIAGEN). Subsequently, cDNA synthesis was conducted using the Invitrogen SuperScript III First-Strand Synthesis SuperMix (Cat#18080400). Real-time PCR was executed with the GoTaq qPCR Mastermix (Promega) employing the CFX-96 real-time PCR System (Bio-Rad). Oligonucleotide primers were designed using Integrated DNA Technologies. Either 18 S or β-actin expression served as an internal control, and the quantification of fold induction was computed using the △△Ct method.

Raw sequence reads were mapped to the reference mouse genome build 38 (GRCm38) by STAR (Dobin et al., 2013). The read count per gene for analysis was generated by featureCounts (Liao et al., 2014). Read counts were normalized to the number of fragments per kilobase of transcript per million mapped reads (FPKM) using the DESeq2 Bioconductor R package (Love et al., 2014). Pathway analysis was performed using the GSEA method implemented in the camera function from the limma R package (Wu and Smyth, 2012). Databases KEGG, MSigDB Hallmark, Reactome were used for the GSEA analysis. Transcripts expressed in either 3 conditions with FPKM >1 were included in the pathway analyses. To infer mice macrophage gene regulatory network, we used ARCHS4 collected list of RNE-seq data for mice macrophage cells from Gene Expression Omnibus (GEO; Lachmann et al., 2018). The total number of analyzed experimental datasets was 1960. This gene expression matrix was utilized as input for the GENIE3 algorithm, which calculated the most likely co-expression partners for transcription factors. The list of transcription factors was derived from the Animal TFDB 3.0 database (Hu et al., 2019). The Virtual Inference of Protein Activity by Enriched Regulon Analysis (VIPER) algorithm was used to estimate the prioritized list of transcription factors (Alvarez et al., 2016). The software Cytoscape (version 3.4.0) was used for network visualization (Shannon et al., 2003). The i-cisTarget web service was used to identify transcription factor binding motifs that were over-represented on a gene list (Imrichová et al., 2015). Statistical analysis and data processing were performed with R version 3.6.1 (https://www.r-project.org/) and RStudio version 1.1.383 (https://www.rstudio.com). Raw data were deposited in NCBI’s Gene Expression Omnibus (GEO, accession number-GSE164698).

Single-cell suspensions of mouse bone-marrow-derived macrophages were loaded on a 10 X genomics chip G at concentrations varying between 1,240,000–1,575,000 cells/mL aiming for a targeted cell recovery of 5000 cells per sample and processed using the 10 X Genomics 3’v3.1 Dual index kit according to the manufacturer protocol (10 x Genomics, CA, USA). Following the 10 X Genomics Chromium controller run, the Gel-beads in emulsion (GEMs) were transferred to strip tubes and subsequently put into a thermal cycler using the following parameters: 45 min at 53 °C, 5 min at 85 °C, hold at 4 °C. The samples were then stored in a –20 °C freezer overnight. The next day, samples were thawed at room temperature before adding the recovery agent to break the GEMs. Subsequently, the cDNA was purified and amplified, and the gene expression libraries were generated and purified. The size distribution and molarity of these libraries were assessed using the Bioanalyzer High Sensitivity DNA Assay (Agilent Technologies, Lexington, MA, USA). The libraries were then pooled at 5 nM and sequenced on an Illumina NextSeq 2000 instrument at a 600pM input and 2% PhiX spike-in using a P3 100 cycle flow cell (Illumina, San Diego, CA, USA) resulting in 36,000–58,000 reads per cell.

The 10 x Genomics Cell Ranger pipeline v6.0.1 was used for demultiplexing, alignment, identification of cells, and counting of unique molecular indices (UMIs). The Cell Ranger mkfastq pipeline was used to demultiplex raw base call (BCL) files generated by Illumina sequencers into FASTQ files. The Cell Ranger count pipeline was used to perform alignment and create UMI count matrices using reference genome mm10 (Ensembl 84) and parameters – expect-cells = 5000. Droplets with at least 500 UMIs underwent further quality control with the SCTK-QC pipeline (Hong et al., 2022). The median number of UMIs was 9927, the median number of genes detected was 2637, the median percentage of mitochondrial reads was 5.43%, and the median contamination estimated by decontX was 0.17 across cells from all samples (Yang et al., 2020). Cells with less than 500 counts, less than 500 genes detected, or more than 25% mitochondrial counts were excluded, leaving a total of 14,658 cells for the downstream analysis.

The Seurat package was used to cluster the cells into subpopulations (Stuart et al., 2019). The 2000 most variable genes were selected using the FindVariableFeatures function, after removing features with fewer than three counts in three cells. Cells were embedded in two dimensions with UMAP using the RunUMAP function. The clustering was performed at resolution 0.8, resulting in 15 clusters. Markers for each cluster were identified with the FindAllMarkers function using the Wilcoxon Rank Sum test and default parameters: log fold-change threshold = 0.25 and min.pct=0.1. The trajectory analysis was performed using the slingshot package; the pseudotime trajectory was computed using the slingshot function and selecting the PCA (considering 50 PCA) and unweighted cluster labels as inputs (Street et al., 2018). The pathway analysis was performed using the VAM package in combination with MSigDB’s Hallmark geneset collection. The gene sets were scored using the vamForSeurat function, and the top pathways were found by running the FindAllMarkers analysis on the CDF assay. The following package versions were used: Seurat v4.0.4, VAM v1.0.0, Slingshot v2.3.0, SingleCellTK v2.4.1, Celda v1.10.0.

Equal amounts of protein extracts were separated by SDS-PAGE and transferred to a PVDF membrane (Millipore). Following a 2 hr blocking period in 5% skim milk in TBS-T buffer [20 mM Tris-HCl (pH 7.5), 150 mM NaCl, and 0.1% Tween20], the membranes were incubated overnight with the primary antibody at 4 °C. Subsequently, bands were acquired using the chemiluminescence (ECL) kit (Thermo Fisher Scientific) and GE ImageQuant LAS-4000 Multi-Mode imager. A loading control, either β-actin or β-tubulin, was assessed on the same membrane. Secondary antibodies used included HRP-conjugated goat anti-mouse and anti-rabbit antibodies. Densiometric analysis was performed by quantification of area using Image J.

The B6 and B6.Sst1S BMDM were stimulated with TNF (10 ng/ml) for 6 hr. Cycloheximide (2 uM) was added to the cultures and cells were harvested after 15, 30, 45, 60, 90, and 120 min. Western blotting was carried out for Nrf2 and β-tubulin as described. The densitometry analysis was performed using ImageJ software (NIH, USA). The linear regression curve was plotted using GraphPad Prism software, and the half-life of Nrf2 protein was derived.

The nuclear extracts were prepared using a nuclear extraction kit (Signosis Inc) according to the instructions provided. The EMSA was carried out using the EMSA assay kit (Signosis Inc) following the instructions provided. Briefly, 5 µg of nuclear extract was incubated with biotin-conjugated Nrf2 antioxidant response element (ARE) probe in 1 X binding buffer for 30 min. The reaction mixture was electrophoresed in 6.5% non-denaturing polyacrylamide gel and then blotted onto a nylon membrane. The protein-bound probe and free probe were immobilized on the membrane by UV-light-mediated cross-linking. After blocking the membrane, the probes were detected using streptavidin-HRP mediated chemiluminescence method. The images were captured using ImageQuant LAS 4000.

The method used to measure Labile Intracellular Iron (LIP) involved the Calcein acetoxymethyl (AM) ester (Invitrogen) quenching technique with some modifications, as described in Amaral et al., 2019; Picard et al., 1998; Thomas et al., 1999. BMDMs were plated in 96-well plates and washed with 1×DPBS before being lysed with 0.05% saponin for 10 min. Next, the cell lysates were incubated with 125 nM calcein AM at 37 °C for 30 min. A negative control was established by incubating the lysates with the iron chelator deferoxamine (Sigma-Aldrich) at room temperature for 30 min, while FeSO4-treated cells served as the positive control. The fluorescence of Calcein AM was measured using a fluorescence microplate reader, and the differential MFI of calcein AM between untreated and deferoxamine-treated samples was calculated to determine intracellular labile iron and represented as fold change.

The amount of MDA in the BMDMs was measured using a Lipid Peroxidation (MDA) Assay Kit (Abcam, ab118970) according to the manufacturer’s protocols. Briefly, 2x10⁶ cells were lysed in Lysis Solution containing butylated hydroxytoluene (BHT). The insoluble fraction was removed by centrifugation, and the supernatant was collected, protein concentrations were estimated and used for analysis. The equal amount of supernatants (based on protein concentration) was mixed with thiobarbituric acid (TBA) solution reconstituted in glacial acetic acid and then incubated at 95 °C for 60 min. The supernatants containing MDA-TBA adduct were added into a 96-well microplate for analysis. A microplate reader was used to measure the absorbance at OD 532 nm.

Briefly, the cells were washed three times with 1 X PBS and treated with 1 µM C11-Bodipy 581/591 dye (Invitrogen) suspended in 1 X PBS for 30 min in the dark at 37 °C. The LPOs were measured fluorometrically using Spectramax M5 microplate reader (Molecular Devices).

The total antioxidant capacity was quantified using the Antioxidant assay kit (Cayman chemical) according to the instructions provided. A standard curve was generated by using various concentrations of Trolox (0.068–0.495 mM). 10 µg of the cell lysate was used to determine the Trolox equivalent antioxidant capacity. The percentage of induced antioxidant capacity was calculated using the formula (Trolox equivalent_{TNF stimulated}-Trolox equivalent_unstimulated)/ Trolox equivalent _unstimulated X 100.

BMDMs were cultured on coverslips and subjected to treatment with or without TNF, followed by processing for inhibitor treatment or Mtb infection. Subsequently, cells were fixed with 4% paraformaldehyde (PFA) for 10 min (non-infected) or 30 min (Mtb infected) at room temperature and then blocked for 60 mins with 1% BSA containing 22.52 mg/mL glycine in PBST (PBS + 0.1% Tween 20). After the blocking step, cells were incubated overnight with primary antibodies (4-HNE, Nrf2, or Nrf-1 specific), washed, and then incubated with Alexa Fluor 594-conjugated Goat anti-Rabbit IgG (H+L) secondary Antibody (Invitrogen) in 1% BSA in the dark for 1 hr. The cells were mounted using ProlongTM Gold antifade reagent (Thermo Fisher Scientific), and images were captured using an SP5 confocal microscope. All images were processed using ImageJ software.

To investigate the lipid peroxidation, the Click-iT lipid peroxidation imaging kit (C10446, Thermo Fisher) was used. The BMDMs were cultured on 12 mm coverslips in 24-well plates. Following a 6 hr of TNF treatment, 50 μM Click-iT linoleamide alkyne (LAA) was added to the cell culture and incubated at 37 °C for 1 hr under 5% CO2. The cells were washed, fixed, permeabilized, blocked, and stained with Alexa fluor 488 azide according to the manufacturer’s instructions. Subsequently, the coverslips were mounted onto the microscope slide with ProLongTM Diamond Antifade mountant with DAPI (P36962, Thermo Fisher). The images were captured using the Leica SP5 confocal microscope, and further, the signal quantification was carried out using ImageJ software (Version 1.53 k, NIH).

B6 and B6.Sst1S cells were seeded in a 96-well plate and stimulated with TNF (10 ng/mL) or left untreated for 6, 24, and 36 hr. For ROS inhibition, cells were treated with BHA (100 µM) 2 hr post TNF stimulation. At each time point, CellROX Oxidative Stress Reagent (Cat#C10444, Invitrogen) at 5 µM final concentration was added to the cells and incubated for 30 min at 37 °C, protected from light. After incubation, cells were washed twice with warm PBS, and fluorescence was analyzed using a fluorescence microscope or flow cytometer. ROS levels were quantified by comparing the fluorescence intensity across treatment groups and time points.

Small RNA libraries were prepared from size fractionation of 10 µg of total macrophage RNA on a denaturing polyacrylamide gel to purify 18-35nt small RNAs, and converted into Illumina sequencing libraries with the NEBNext Small RNA Library kit (NEB). Libraries were sequenced on the Illumina NextSeq-550 in the Boston University Microarray and Sequencing Core. We applied long RNAs and small RNAs RNAseq fastq files to a transposon and small RNA genomics analysis pipeline previously developed for mosquitoes (Dayama et al., 2022; Ma et al., 2021) with the mouse transposon consensus sequences loaded instead. Mouse transposon consensus sequences were downloaded from RepBase (Bao et al., 2015), and RNA expression levels were normalized internally to each library’s depth by the Reads Per Million counting method.

The subcutaneous infection of mice was conducted following the procedure outlined in Yabaji et al., 2025b. In brief, mice were anesthetized using a ketamine-xylazine solution administered intraperitoneally. Each mouse was subcutaneously injected in the hock, specifically in the lateral tarsal region just above the ankle (by placing the animal in the restrainer), with 50 µl of 1 X PBS containing 10^6 CFU of Mtb H37Rv or Mtb Erdman (SSB-GFP, smyc′::mCherry). At the designated time points, the mice were anesthetized, lung perfusion was performed using 1 X PBS/heparin (10 U/ml), and organs were collected.

Female mice were used unless otherwise specified, and mice were randomly assigned to experimental groups. No animals were excluded after enrollment. Investigators were blinded to group assignments during outcome assessment. Sample size was determined based on prior studies; no formal power analysis was conducted.

Immunofluorescence of thick lung sections (50–100 µm) was conducted following the detailed procedures outlined earlier (Yabaji et al., 2025b). Briefly, lung slices were prepared by embedding formalin-fixed lung lobes in 4% agarose and cutting 50 μm sections using a Leica VT1200S vibratome. Sections were permeabilized in 2% Triton X-100 for 1 day at room temperature, followed by washing and blocking with 3% BSA in PBS and 0.1% Triton X-100 for 1 hr. Primary antibodies were applied overnight at room temperature, followed by washing and incubation with secondary antibodies for 2 hr. Samples were stained with Hoechst solution for nuclei detection, cleared using RapiClear 1.47 solution for 2 days, and mounted with ProLong Gold Antifade Mountant. Imaging was performed using a Leica SP5 spectral confocal microscope. Primary rabbit anti-4-HNE, anti-Iba1, and anti-iNOS polyclonal antibodies were detected using goat anti-rabbit antibodies labeled with Alexa Fluor 647. Custom-designed HCR RNA-FISH probe sets targeting mouse Ifnb1 mRNA, along with amplifiers and buffers (v3 chemistry), were obtained from Molecular Instruments (www.molecularinstruments.com). Hybridization and amplification were performed according to the manufacturer’s instructions, and samples were imaged using a confocal microscope.

Paraffin-embedded 5 µm sections were stained using New Fuchsin method (Poly Scientific R and D Corp., cat no. K093) and counterstained with methylene blue following the manufacturer’s instructions.

Tissue samples were submersion fixed for 48 hr in 10% neutral buffered formalin, processed in a Tissue-Tek VIP-5 automated vacuum infiltration processor (Sakura Finetek, Torrance, CA, USA), followed by paraffin embedding with a HistoCore Arcadia paraffin embedding machine (Leica, Wetzlar, Germany) to generate formalin-fixed, paraffin-embedded (FFPE) blocks, which were sectioned to 5 µm, transferred to positively charged slides, deparaffinized in xylene, and dehydrated in graded ethanol. A subset of slides from each sample was stained with hematoxylin and eosin (H&E), and consensus qualitative histopathology analysis was conducted by a board-certified veterinary pathologist (N.A.C.) to characterize the overall heterogeneity and severity of lesions.

A rabbit-specific HRP/DAB detection kit was employed (Abcam catalog #ab64261, Cambridge, United Kingdom). In brief, slides were deparaffinized and rehydrated, endogenous peroxidases were blocked with hydrogen peroxide, antigen retrieval was performed with a citrate buffer for 40 min at 90 °C using a NxGen Decloaking chamber (Biocare Medical, Pacheco, California), non-specific binding was blocked using a kit protein block, the primary antibody was applied at a 1:200 dilution, which cross-reacts with mycobacterium species (Biocare Medical catalog#CP140A,C) and was incubated for 1 hr at room temperature, followed by an anti-rabbit antibody, DAB chromogen, and hematoxylin counterstain. Uninfected mouse lung was examined in parallel under identical conditions with no immunolabeling observed serving as a negative control.

A Ventana Discovery Ultra (Roche, Basel, Switzerland) tissue autostainer was used for brightfield and multiplex fluorescent immunohistochemistry (fmIHC). In brief, tyramide signaling amplification (TSA) was used in an iterative approach to covalently bind Opal fluorophores (Akoya Bioscience, Marlborough, MA) to tyrosine residues in tissue sections, with subsequent heat stripping of primary-secondary antibody complexes until all antibodies were developed. Before multiplex-IHC was performed, each antibody was individually optimized using a single-plex-IHC assay using an appropriate positive control tissue. Optimizations were performed to determine ideal primary antibody dilution, sequential order of antibody development, assignment of each primary antibody to an Opal fluorophore, and fluorophore dilution. Once an optimal protocol was established, 5 µm tissue sections were cut from FFPE lung arrays. All Opal TSA-conjugated fluorophore reactions took place for 20 min. Fluorescent slides were counterstained with spectral DAPI (Akoya Biosciences) for 16 min before being mounted with ProLong gold antifade (Thermo Fisher, Waltham, MA). Antibodies utilized in 4-plex 5 color (DAPI counterstained) analysis included: Iba1, iNOS, YFP, Chac1, and P-c-Jun. All rabbit antibodies were developed with a secondary goat anti-rabbit HRP-polymer antibody (Vector Laboratories, Burlingame, CA) for 20 min at 37° C and all mouse-derived antibodies were developed with a secondary goat anti-mouse HRP-polymer antibody (Vector Laboratories).

Antigen retrieval was conducted using a Tris-based buffer-Cell Conditioning 1 (CC1)-Catalog # 950–124 (Roche). The MHCII primary was of mouse origin, so a goat anti-mouse HRP-polymer antibody (Vector Laboratories) was utilized. Brightfield slides utilized A ChromoMap DAB (3,3′-Diaminobenzidine) Kit-Catalog #760–159 (Roche) to form a brown precipitate at the site of primary-secondary antibody complexes containing HRP. Slides were counterstained with hematoxylin and mounted.

Fluorescent and chromogen-labeled slides were imaged at 40 X using a Vectra Polaris Quantitative Pathology Imaging System (Akoya Biosciences). For fluorescently labeled slides, exposures for all Opal dyes were set based upon regions of interest with strong signal intensities to minimize exposure times and maximize the specificity of signal detected. A brightfield acquisition protocol at 40 X was used for chromogenically labeled slides.

Whole slide fluorescent images were segmented into QPTIFFs, uploaded into Inform software version 2.4.9 (Akoya Biosciences), unmixed using spectral libraries affiliated with each respective opal fluorophore including removal of autofluorescence, then fused together as a single whole slide image in HALO (Indica Labs, Inc, Corrales, NM).

View settings were adjusted to allow for optimal visibility of immunomarkers and to reduce background signal by setting threshold gates to minimum signal intensities. After optimizing view settings, annotations around the entire tissue were created to define analysis area using the flood tool, and artifacts were excluded using the exclusion pen tool. To analyze lesions independently, a tissue classifier that uses a machine-learning approach was utilized to identify lesions. The outputs from this classifier were annotations for each lesion, which were then isolated into independent layers, which were analyzed separately from other layers. For quantifying myeloid markers in multiplexes, an algorithm called the HALO (v3.4.2986.151, Indica Labs, Albuquerque, NM, USA) Area Quantification (AQ) module was created and finetuned to quantify the immunoreactivity for all targets. Thresholds were set to define positive staining based on a real-time tuning feature. AQ outputted the percentage of total area displaying immunoreactivity across the annotated whole slide scan in micrometers squared (μm²). AQ output the total percentage positive for each phenotype. AQ and HP algorithms were run across all layers for each individual lesion and exported as a.csv file.

To perform spatial transcriptomics analysis, we used the Nanostring GeoMX Digital Spatial Profiler (DSP) system (Nanostring, Seattle, WA) (Danaher et al., 2022; Merritt et al., 2020). To identify pathways dominating early vs. late (advanced) lesions, we selected lungs from 2 mice with paucibacillary early lesions and 2 mice with advanced multibacillary lesions with necrotic areas. Slides were stained with fluorescent CD45-, pan-keratin-specific antibodies, and Iba1-specific antibodies (sc-32725, Santa Cruz Biotechnology, CA, USA) and DAPI. Diseased regions of interest (ROI) for expression profiling of Iba1 + cells were selected to focus on myeloid-rich areas avoiding areas of micronecrosis and tertiary lymphoid tissue (Figure 7—figure supplement 3). Eight ROI each of normal lung, early and late lesions (respectively) were studied. The profiling used the Mouse Whole Transcriptome Atlas (WTA) panel which targets ~21,000 + transcripts for mouse protein coding genes plus ERCC-negative controls to profile the whole transcriptome, excluding uninformative high expressing targets such as ribosomal subunits. A subsequent study focused on gene expression within macrophages identified by Iba1 labeling in lung sections showing uncontrolled (advanced multibacillary) lesions in comparison to controlled (paucibacillary stage). Samples from each ROI were packaged into a library for sequencing (NextSeq550, Illumina) following the procedure recommended by Nanostring. After sequencing, the data analysis workflow began with QC evaluation of each ROI based on thresholds for number of raw and aligned reads, sequencing saturation, negative control probe means, and number of nuclei and surface area. Background correction is performed using subtraction of the mean of negative probe counts. Q3 normalization (recommended by Nanostring) results in scaling of all ROIs to the same value for their 3rd quartile value. The count matrix was imported into the Qlucore Genomics Explorer software package (Qlucore, Stockholm, Sweden) and log2 transformed for further analysis. Statistical analysis was then performed to identify differentially expressed genes (typically unpaired t-test with Benjamini-Hochberg control for FDR rate at q<.05). Lists of differentially expressed genes (DEGs) from comparisons of distinct stages were further analyzed for enrichment of pathways or predicted transcription factors using the Enrichr online tool (Chen et al., 2013; Liberzon et al., 2015; Xie et al., 2021). To specifically explore interferon-related gene expression in Iba1 + macrophages, we used: (1) a list of genes unique to type I genes derived by pooling four common type I and type II interferon gene list respectively and identifying the non-overlapping (unique) genes for each type (see Supplementary file 8, Figure 7—figure supplement 4); (2) using the GSEA analysis tool of the Qlucore software to evaluate enrichment of the MSigDB Hallmark collection of pathways (which include Interferon-alpha and Interferon-gamma, and other pathways of interest [Liberzon et al., 2015]). Raw data were deposited in NCBI’s Gene Expression Omnibus (GEO, accession number- GSE292392).

Participant demographics including age, sex, body mass index (BMI), diabetes status, alcohol use, and tobacco use were recorded and matched for Blood collection. Blood samples were collected from 41 individuals and recently (<90 days) diagnosed with TB, within one week of treatment commencement from five clinical sites within the Regional Prospective Observational Research for Tuberculosis (RePORT)-India consortium. Individuals for this study were newly diagnosed with TB (within 90 days; sputum smear-positive or Xpert MTB/RIF assay positive (Cepheid, Sunnyvale, CA, USA)), at least 15 years of age, multi-drug-resistant negative, and had received less than one week of treatment. Sociodemographic data such as age, sex, behavioral characteristics (smoking, alcohol use (using the Alcohol Use Disorders Identification Test [AUDIT-C])), tobacco use, and body mass index (BMI) were obtained through a questionnaire and the medical history of the participants. There were no significant differences in risk factors for TB or treatment failure, including sex, risky alcohol use, tobacco use, BMI, or diabetes. Patients were monitored for two years during and post-treatment with Rifampicin, isoniazid, ethambutol, and pyrazinamide, per India national guidelines Division CT. Technical and Operational Guidelines for TB Control in India 2016. Central TB Division, Directorate General of Health Service, Ministry of Health and Family Welfare; Government of India, New Delhi. 2016 (Bush et al., 1998). Individuals who failed treatment (n=21) were identified by positive sputum culture or clinical diagnosis of symptoms at any time after 4 full months of treatment and symptoms determined to not be from another cause. Treatment controls were determined by those who remained culture-negative and symptom-negative for the two-year observation period.

Blood samples from each person were collected and stored in PAXgene tubes (Cat #762165, BD Biosciences, San Jose, CA, USA) at –80 °C until processing using the PAXgene Blood RNA kit (Cat #762164, QIAGEN, Hilden, Germany). PAXgene tubes were sent to MedGenome (Bangalore, India) for processing. Library preparation and sequencing were performed at the GenoTypic Technology Pvt. Ltd. Genomics facility in Bangalore, India, with the SureSelect Strand-Specific mRNA Library Prep kit (Cat #5190–6411, Agilent, Santa Clara, USA). One μg of RNA was used for mRNA enrichment by poly(A) selection and fragmented using RNAseq Fragmentation Mix containing divalent cations at 94°C for 4 min. Enriched mRNA underwent fragmentation through exposure to RNASeq Fragmentation Mix with divalent cations at 94 °C for 4 min. Subsequently, single-strand cDNA was synthesized in the presence of Actinomycin D (Gibco, Life Technologies, Carlsbad, CA, USA) and purified utilizing HighPrep magnetic beads (Magbio Genomics Inc, USA). Following this step, double-stranded cDNA was synthesized, and the ends were repaired before undergoing further purification. Adenylation of the 3′-ends of cDNA preceded the ligation of Illumina Universal sequencing adaptors, which were then purified and amplified with 10 PCR cycles. The final cDNA sequencing libraries were purified and quantified using Qubit, while the fragment size distribution was assessed using Agilent TapeStation. Subsequently, the libraries were combined in equimolar proportions, and the resulting multiplexed library pools were sequenced utilizing the Illumina NextSeq 500 platform for 75 bp single-end reads.

The RNA samples were processed sequentially in two batches. Batch effects created by combining the two batches were removed using ComBat-Seq (Zhang et al., 2020). The ComBat-Seq adjusted counts were normalized using a log₂-counts per million (logCPM) adjustment, and the logCPM values were used for downstream analysis.

Differential expression was performed on batch-corrected counts/logcpm using limma (Ritchie et al., 2015) and signature genes were selected from the limma results to obtain logFC, p-values, and p-adjusted values (using a Benjamini-Hochberg false-discovery rate).

Normal Human mammary epithelial cells (HMEC) were transfected with adenoviruses containing either the gene of interest or GFP as described in Bild et al., 2006; McQuerry et al., 2019. Differentially expressed genes between transfected cells and controls with a Benjamini-Hochberg cutoff of < 0.05 were used in for the signatures of each transfected oncogene.

Heatmaps, boxplots, receiver-operating characteristic (ROC) curve, and area under the ROC curve (AUC) were calculated and depicted using the functions within the TBSignatureProfiler (Johnson et al., 2021). Bootstrapping was used to iteratively calculate AUC values using leave-one-out cross-validation to obtain mean AUCs and 95% confidence intervals (CI) for 100 repeats for each signature. We generated mean AUC values for the predictive performance of collected oncogene-specific signatures (Bild et al., 2006; McQuerry et al., 2019) in terms of their ability to distinguish between treatment-failure individuals and control samples in our cohort. The single sample Gene Set Enrichment Analysis (ssGSEA; Barbie et al., 2009) was used to generate enrichment scores for each signature.

Statistical analyses were performed using GraphPad Prism 9 software (RRID:SCR_002798). Differences among groups involving two or more variables were assessed using two-way analysis of variance (ANOVA) with adjustments for multiple post hoc comparisons. For comparisons across multiple groups based on a single variable, one-way ANOVA with post hoc testing was applied. Two-tailed paired or unpaired t-tests were used for comparisons between two groups after verifying data normality. For non-parametric datasets, the Wilcoxon Rank Sum test (Mann-Whitney U test equivalent) was employed. Sample sizes were chosen based on prior studies using the same infection model; no formal power calculation was performed. Animals were randomly assigned to experimental groups. Statistical significance was defined as p<0.05. Significance levels are indicated as follows: *, p<0.05; **, p < 0.01; ***, p < 0.001; and ****, p < 0.0001.