Mycobacterium tuberculosis IDH-PPARγ interaction suppresses GPX4 to drive macrophage ferroptosis and sustain persistent infection

Abstract

Mycobacterium tuberculosis (M.tb) actively reprograms host lipid metabolism during infection; however, the underlying mechanism remains poorly understood. How M.tb manipulates macrophage lipid metabolism to induce lipid peroxidation and ferroptosis for bacterial persistence remains a fundamental question. Here, using single-cell RNA sequencing and proteomics, we show that M.tb infection substantially upregulates peroxisome proliferator-activated receptor gamma (PPARγ) in macrophages. Mechanistically, M.tb isocitrate dehydrogenase (IDH) interacts with PPARγ and impairs its proteasomal degradation. Elevated PPARγ suppresses glutathione peroxidase 4 (Gpx4) expression by recruiting the NCOR/SMRT corepressor complex to the Gpx4 promoter, resulting in increased lipid peroxidation and ferroptosis in infected macrophages. In mice, PPARγ knockout or pharmacological inhibition decreases lung inflammation and M.tb burden, restores GPX4 expression, and enhances macrophage survival. Our findings reveal a mechanism by which M.tb exploits the IDH-PPARγ axis to induce ferroptosis and sustain persistent infection, identifying therapeutic targets for tuberculosis treatment through disruption of this interaction.

Mycobacterium tuberculosis (M.tb) actively intervenes in host cell lipid metabolism during infection. Here the authors show that M.tb modulates the IDH-PPARγ axis to drive ferroptosis and bacterial persistence in macrophages.

Introduction

The intricate relationship between cellular metabolism and immune function plays a critical role in determining the outcome of bacterial infections. Mycobacterium tuberculosis (M.tb), the leading cause of infectious disease mortality worldwide, is a major public health challenge, particularly in many developing countries^1–3. Persistent M.tb infection is closely associated with several metabolic disorders. There is increasing epidemiological evidence that metabolic conditions such as hyperglycemia and dyslipidemia increase susceptibility to tuberculosis (TB) infection⁴. People with type 2 diabetes have a threefold increased risk of TB infection and a twofold increased likelihood of poor treatment outcomes^5,6. Notably, six of the ten countries with the highest diabetes prevalence are also classified as high-burden TB countries, accounting for 80% of global TB cases⁵. Research has shown that TB patients have abnormal metabolism of phospholipids, glycerides, and sphingolipids⁷. After treatment, the metabolic profiles of these patients gradually normalize, suggesting that the interventions targeting lipid metabolism can effectively disrupt energy metabolism and inhibit cell wall synthesis of M.tb⁷.

Lipid metabolism plays a critical role in the growth and virulence of M.tb⁸. Although M.tb shows metabolic flexibility in utilizing different carbon sources during various stages of infection, host-derived lipids are the primary carbon substrate within the host microenvironment⁹. M.tb infection induces significant alterations in cellular lipid homeostasis, leading to the accumulation of cholesteryl ester and triglyceride rich lipid droplets (LDs) in infected macrophages, thereby promoting foam cell formation¹⁰. This transformation facilitates the survival of M.tb and its resistance to host immune responses. Furthermore, sequestration of intracellular triglycerides in lipid inclusion bodies is associated with attenuated growth kinetics, diminished metabolic flux, and enhanced antimicrobial resistance¹¹. Host-derived lipids also contribute to the development of drug-resistant M.tb, such as triacylglycerol and cholesterol¹². These metabolic perturbations manifest key pathological hallmarks of persistent M.tb infection and suggest that precise regulation of intracellular lipid metabolism may influence the dynamics of M.tb infection¹³. While numerous studies indicate that host cellular lipid metabolism may involve both active and passive mechanisms^14–17, the specific mechanisms in which bacteria actively modulate this process remain poorly understood.

Lipid metabolism plays a critical role in regulating cellular lipid toxicity, and aberrant lipid metabolism could trigger ferroptosis, a form of programmed cell death characterized by iron-dependent lipid peroxidation¹⁸. During M.tb infection, there is increasing evidence that elevated iron levels, increased fatty acid oxidation, and glutathione (GSH) depletion contribute to the development of ferroptosis in host cells^19,20. This process is important for the pathogenicity of M.tb and facilitates the spread of bacteria from infected cells in the granulomas. In addition, studies also highlighted the decreased expression of glutathione peroxidase 4 (GPX4) as a critical factor in this context^21,22; however, the underlying mechanisms remain largely unexplored. Some studies suggest that M.tb manipulates host cell iron metabolism and ferroptosis through effector proteins such as PtpA and Rv1324^23,24. However, the interactions between M.tb and macrophage lipid metabolism, particularly in the context of prolonged intracellular survival, remain poorly understood. This issue is particularly relevant in diabetic patients with dyslipidemia infected with M.tb^25–27.

Peroxisome proliferator-activated receptor (PPAR) is a member of the nuclear hormone receptor superfamily. As a master regulator of lipid homeostasis, PPARγ has attracted considerable attention because of its orchestrating roles in metabolic reprogramming and immunological responses during M.tb infection^28–32. PPARγ is critical for regulating the generation of alveolar macrophages, which serve as a niche for M.tb infection³³. Inhibition or knockdown of PPARγ significantly inhibits M.tb growth both in human and mouse macrophages^34,35. Recent studies also demonstrated that M.tb infection increases the abundance of PPARγ in human macrophages, thereby promoting M.tb growth and infectivity³⁶; however, the underlying mechanisms are less understood.

Here, we identify a previously unrecognized mechanism by which M.tb infection increases PPARγ protein levels in macrophages through post-translational modification (PTM) networks mediated by M.tb isocitrate dehydrogenase (IDH). Our mechanistic analyses reveal that this elevated PPARγ accumulation directly suppresses transcription of the critical antioxidant enzyme GPX4, triggering lipid peroxidation-dependent ferroptotic cell death in macrophages. By establishing this molecular link between M.tb-induced metabolic dysregulation, ferroptosis, and bacterial persistence, our findings not only advance the understanding of TB pathogenesis but also identify potential therapeutic targets for host-directed treatment strategies.

Results

M.tb induces excessive accumulation of PPARγ in macrophages

To investigate host cellular responses to M.tb infection, we conducted single-cell RNA sequencing (scRNA-seq) analysis of peripheral blood from TB patients. This analysis revealed significant alterations in macrophage subpopulations, particularly those associated with lipid metabolism (Supplementary Fig. 1a–e). Notably, we observed marked upregulation of the PPAR pathway in these macrophage subpopulations (Fig. 1a, b). This observation was corroborated by RNA-seq (GSE162729) analysis of THP-1 cells infected with virulent M.tb H37Rv and avirulent H37Ra strains, which demonstrated significant enrichment of the PPAR pathway (Supplementary Fig. 1f–h)³⁷.

Fig. 1. Mycobacterial infection induces high expression of PPARγ in macrophages.

a scRNA-seq of clinical PBMC samples. UMAP dimensionality reduction clustering of MPs cell populations from healthy individuals (left, n = 3, biological replicates) and tuberculosis patients (right, n = 3, biological replicates). b Reactome pathway enrichment analysis of MPs cell populations (over-representation analysis). c Transcriptome of THP-1-derived macrophages infected with M.tb strains H37Rv and H37Ra, Venn analysis of DEGs in the NC vs H37Rv and NC vs H37Ra groups. d KEGG pathway analysis of the 282 common DEGs upregulated after H37Rv infection. e Heatmap analysis of PPAR signaling pathway genes, with upregulated (red) and downregulated (blue). f Protein–protein interaction network analysis of upregulated differential genes in the PPAR signaling pathway (red: high confidence score). g Immunofluorescence analysis of BMDMs infected with mCherry-H37Rv. Nuclei were stained with Hoechst, green fluorescence indicates PPARγ expression, and red fluorescence marks H37Rv (independently repeated two times). Scale bars, 20 μm. h Relative fluorescence intensity (RFI) of PPARγ for (g) (n = 6, biological replicates). i Immunofluorescence of BMDMs infected with the mCherry-H37Rv strain, with Hoechst staining of the nucleus and red fluorescence indicating H37Rv (independently repeated two times). Scale bars, 20 μm. j RFI of mCherry-H37Rv for (i) (n = 6, biological replicates). k Intracellular bacterial burden in BMDMs infected with H37Rv and treated with GW9662 was quantified by colony-forming unit (CFU) assays (n = 3, biological replicates, one-way ANOVA). l BODIPY immunofluorescence detection in RAW264.7 macrophages infected with both mCherry-H37Ra and mCherry-H37Rv strains (independently repeated three times). Scale bars, 5 μm. m The corresponding statistical analysis results of BODIPY staining (n = 3, biological replicates). n Western blot showing the levels of PPARγ in RAW264.7 macrophages infected with H37Ra after protein synthesis inhibition with cycloheximide (CHX). Representative blots of three independent experiments. o, p Relative intensity of the PPARγ protein in the NC and H37Ra infection groups for this figure (n) (n = 6, biological replicates). All results are expressed as mean ± s.e.m. Two-tailed t-test for comparing two groups and one-way ANOVA for multiple comparisons. All raw data are provided as a Source Data file.

To dissect the molecular underpinnings of PPAR pathway activation, we performed Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis on differentially expressed genes following M.tb infection in THP-1 cells. This analysis confirmed significant upregulation of the PPAR pathway (Fig. 1c, d and Supplementary Fig. 1i). While numerous downstream PPAR target genes were activated, the PPARG gene itself was not upregulated at the mRNA level, suggesting a potential post-transcriptional regulatory mechanism (Fig. 1e, f and Supplementary Fig. 1j). Intriguingly, we observed a paradoxical relationship between PPARγ mRNA and protein levels. M.tb H37Ra infection significantly decreased Pparg mRNA expression, whereas PPARγ protein levels were markedly elevated (Supplementary Fig. 1k–m). This finding strongly indicates that M.tb infection disrupts the normal post-translational regulation of PPARγ³⁸. Consistent with these findings, immunofluorescence (IF) microscopy revealed a striking accumulation of PPARγ in bone marrow-derived macrophages (BMDMs) upon H37Rv infection (Fig. 1g, h). Similar results were also observed in RAW264.7 macrophages upon H37Ra infection (Supplementary Fig. 2a, b). Additionally, M.tb H37Ra infection induced the accumulation of LDs in macrophages, a hallmark of altered lipid metabolism (Supplementary Fig. 2c–e). To investigate the functional consequences of PPARγ dysregulation, we treated infected macrophages with the PPARγ antagonist GW9662. Although GW9662 did not affect the in vitro H37Rv growth in 7H9 medium (Supplementary Fig. 2f), it significantly reduced the intracellular burden of M.tb (H37Rv and H37Ra) in both BMDMs and RAW264.7 cells (Fig. 1i–k and Supplementary Fig. 2g–i). Suppressing PPARγ also decreased LD accumulation in RAW264.7 cells during infections with H37Rv or H37Ra (Fig. 1l, m). This highlights the importance of PPARγ in M.tb pathogenesis and host lipid metabolism. To further confirm the post-transcriptional regulation of PPARγ, we inhibited cellular protein translation using cycloheximide (CHX) in M.tb-infected macrophages. This resulted in a time-dependent accumulation of PPARγ, even in the absence of new protein synthesis (Fig. 1n–p), confirming that M.tb manipulates PPARγ stability.

Collectively, these results demonstrate that M.tb infection induces significant accumulation of PPARγ in macrophages, likely through post-translational mechanisms, and that this accumulation may play a crucial role in supporting bacterial survival and altering host cell lipid metabolism.

M.tb infection hijacks host cell PPARγ protein through multiple mechanisms

Building upon our initial findings of PPARγ accumulation in M.tb-infected cells, we sought to elucidate the underlying mechanisms. We performed PPARγ immunoprecipitation followed by mass spectrometry in H37Ra-infected RAW264.7 macrophages, which identified three mycobacterial proteins that interact with PPARγ, including IDH (Rv3339c, gene: icd1), oxidoreductase (Rv0547c), and ribosomal protein S19 (Supplementary Fig. 3a, b). Given the pivotal role of M.tb IDH in carbon metabolism and its adaptive capacity to fatty acid environments, we further investigated the potential link between IDH and PPARγ³⁹. The predicted binding model of mouse PPARγ and M.tb protein IDH, as well as the detailed contacts of the complex, are shown in Fig. 2a. Pro110, Thr296, Glu319, and Gln322 of mouse PPARγ interact with residues such as Glu243, Ser364, Lys129, and Gly126 in IDH through hydrogen bonds. Other residues, including Arg316 and Glu319 in mouse PPARγ form salt bridges with residues such as Ser364, Glu265, and Lys129 in IDH. Van der Waals interactions are also observed between mouse PPARγ and IDH. These interactions contribute substantially to the binding energy of mouse PPARγ and IDH. Additionally, based on the predicted binding model, IDH may obstruct the S112 site of PPARγ, which could prevent S112 from binding to other molecules.

Fig. 2. Inhibition of PPARγ degradation via hijacking mechanisms in RAW264.7 cells upon infection with H37Rv and H37Ra strains.

a The predicted binding model of mouse PPARγ and M.tb IDH was generated using AlphaFold and calculated via the HDOCK server platform. The backbone of mouse PPARγ is colored in orange, while IDH protein in blue. The interacting residues in mouse PPARγ and IDH protein are depicted as orange, marine sticks, respectively. The hydrogen bonds, salt bridge interactions between mouse PPARγ and IDH protein are depicted as green and red dashed lines, respectively. Residue Ser112 in mouse PPARγ is shown as green balls and sticks. b Co-IP was performed to detect the interaction between Flag-PPARγ and ZsGreen-IDH in HEK293T cells (independently repeated three times). c Western blot analysis of M.tb H37Rv IDH expression in bacterial lysate, culture supernatants, and in infected BMDMs (n = 2, biological replicates, independently repeated two times). d Endogenous Co-IP array reveals interaction between PPARγ and IDH in BMDMs following H37Rv infection (independently repeated three times). e Co-localization of Flag-PPARγ with ZsGreen-IDH was analyzed (independently repeated two times). Scale bars, 5 μm. f Western blot analysis showing the levels of PPARγ and p-PPARγ (S112) after overexpression of ZsGreen-NC or ZsGreen-IDH in RAW264.7 cells (independently repeated three times). g Proteomic sequencing following H37Ra strain infection in RAW264.7 cells, with KEGG analysis of downregulated proteins (n = 4, biological replicates, over-representation analysis). h Heatmap analysis of the expression levels of 20S proteasome catalytic β subunits in the proteome (n = 4, biological replicates). i Analysis of 20S proteasome catalytic following H37Ra strain infection in RAW264.7 cells (n = 12, biological replicates). j Western blot detection of ubiquitinated protein levels at different time points after infection of RAW264.7 cells with the H37Ra strain (n = 2, biological replicates, independently repeated three times). k Western blot analysis showing the levels of 20S proteasome catalytic β subunits following H37Ra strain infection in RAW264.7 cells (n = 2, biological replicates, independently repeated three times). l Schematic diagram of M.tb IDH hijacking PPARγ and inhibiting its protein degradation. The illustration was created in BioRender. Zhao, C. (2026) https://BioRender.com/vf6b7cn. All results are expressed as mean ± s.e.m. One-way ANOVA for multiple comparisons. All raw data are provided as a Source Data file.

To validate this interaction, we overexpressed both M.tb ZsGreen-IDH and Flag-PPARγ in HEK293T cells. Co-immunoprecipitation (Co-IP) experiments confirmed the interaction between IDH and PPARγ (Fig. 2b). To determine whether IDH is actively secreted by M.tb, we analyzed culture supernatants from H37Rv grown under axenic conditions. IDH was undetectable in the culture supernatant, while it was robustly detected in H37Rv lysates (GROEL1 as a loading control), indicating that IDH is not a classically secreted protein (Fig. 2c). In contrast, IDH was detected in BMDMs upon infection with H37Rv (Fig. 2c). These findings suggest that IDH release occurs specifically within the host intracellular environment. Given that no secretion was detected in the M.tb culture supernatant, we speculate that IDH release may result from bacterial damage or lysis during infection. Additionally, we detected an interaction between endogenous PPARγ and IDH in H37Rv-infected BMDMs (Fig. 2d), further indicating that the infection stabilizes PPARγ expression.

Next, we ectopically expressed M.tb IDH in HEK293T cells to study its interaction with PPARγ. Immunofluorescence (IF) staining further demonstrated co-localization of PPARγ and M.tb IDH in the cytoplasm (Fig. 2e). Notably, overexpression of M.tb IDH alone was sufficient to induce PPARγ accumulation within HEK293T cells (Fig. 2e), suggesting a causal relationship. Importantly, ectopic overexpression of M.tb IDH in RAW264.7 cells inhibited the phosphorylation of PPARγ at serine 112 (S112), a key residue that targets PPARγ for degradation^40,41(Fig. 2f).

Given that PPARγ degradation is primarily mediated by the ubiquitin-proteasome system (UPS)⁴², we conducted global proteomic analysis of H37Ra-infected cells (Supplementary Fig. 3c, d). This analysis revealed a significant downregulation of the UPS following M.tb infection (Fig. 2g). Specifically, several subunits of the 26S proteasome were downregulated (Supplementary Fig. 3e), including the 20S proteasome catalytic β subunits PSMB8, PSMB9, and PSMB10 (Fig. 2h). Functional assays confirmed a decrease in cellular proteasome 20S activity after H37Ra infection (Fig. 2i). Since the expression of 20S proteasome catalytic β subunits is regulated by interferon-related pathways⁴³, we investigated the impact of M.tb infection on signal transducer and activator of transcription 1 (STAT1) signaling. We observed a significant inhibition of STAT1 phosphorylation at S727, which was accompanied by the decreased expression of PSMB8, PSMB9, and PSMB10 (Fig. 2k). Notably, as a consequence of impaired proteasome function, we detected an accumulation of ubiquitinated proteins in H37Ra-infected RAW264.7 cells in a time-dependent manner, including the ubiquitinated PPARγ (Fig. 2j and Supplementary Fig. 3f).

Collectively, these results reveal a multifaceted strategy employed by M.tb to manipulate the stability of PPARγ in host-cell. The bacterium achieves this through direct protein-protein interactions, modulation of PPARγ phosphorylation, and global dysregulation of the host-cell protein degradation machinery (Fig. 2l).

Accumulation of PPARγ following M.tb infection induces ferroptosis in macrophages

Having established the mechanisms by which M.tb infection leads to PPARγ accumulation, we next investigated the downstream consequences of this phenomenon. Proteomic analysis of infected cells revealed significant upregulation of the ferroptosis pathway (Fig. 3a)^24,44. Further examination of intracellular protein dynamics during H37Ra infection showed substantial accumulation of PPARγ in both the cytoplasm and nucleus, accompanied by a significant downregulation of GPX4⁴⁵, a key inhibitor of ferroptosis (Supplementary Fig. 4a). We then examined the temporal dynamics of PPARγ-mediated ferroptosis in RAW264.7 cells, where GPX4 expression progressively decreased while PPARγ increased during H37Ra infection (Supplementary Fig. 4b, c). In H37Rv-infected BMDMs, GPX4 was reduced by infection but restored by PPARγ inhibition, while solute carrier family 7 member 11 (SLC7A11) remained unchanged (Fig. 3b, c). Notably, the lipid peroxidation product 4-hydroxynonenal (4HNE) progressively accumulated, indicating ongoing ferroptotic damage (Supplementary Fig. 4b, c). In contrast, the expression of the ferroptosis-related membrane protein SLC7A11 remained relatively stable (Fig. 3b, c and Supplementary Fig. 4b, c). To further explore the role of PPARγ in M.tb-induced ferroptosis, we employed IF microscopy to assess intracellular iron deposition, a crucial feature of ferroptosis. M.tb infection significantly increased iron accumulation, which was attenuated by PPARγ inhibition in both BMDM (H37Rv) and RAW264.7 (H37Ra) cells (Fig. 3d, f and Supplementary Fig. 4d, e). DCFH-DA staining indicated that PPARγ inhibition significantly attenuated H37Rv-induced reactive oxygen species (ROS) accumulation in BMDMs (Fig. 3e, g). Similarly, lipid peroxidation, another hallmark of ferroptosis, was markedly increased following M.tb infection. PPARγ inhibition significantly reduced lipid peroxidation products in macrophages infected with H37Rv or H37Ra (Fig. 3h, i and Supplementary Fig. 4f–h). TUNEL staining revealed a significant increase in apoptosis in H37Rv-infected BMDMs, which was associated with ferroptosis-related processes (Fig. 3j, k and Supplementary Fig. 4i, j). PPARγ inhibition significantly reduced apoptosis in these cells (Fig. 3j, k and Supplementary Fig. 4i, j). This finding provides a mechanistic link between PPARγ accumulation and GPX4 inhibition, potentially explaining how M.tb infection promotes ferroptosis.

Fig. 3. PPARγ promotes ferroptosis in macrophages.

a Proteomic sequencing following H37Ra strain infection in RAW264.7 cells, with KEGG analysis of upregulated proteins (n = 4, biological replicates, over-representation analysis). b Western blot analysis of PPARγ, GPX4, SLC7A11, and p-PPARγ in BMDMs infected with H37Rv and treated with GW9662 (n = 2, biological replicates, independently repeated two times). c Relative intensity of protein in the Control and H37Rv infection groups from (b) (n = 6, biological replicates). d FerroOrange immunofluorescence detection in BMDM cells infected with the H37Rv and treated with GW9662 (independently repeated two times). Scale bars, 25 μm. e DCFH-DA staining in H37Rv-infected BMDMs treated with GW9662. Fluorescence indicates intracellular ROS levels (independently repeated two times). Scale bars, 25 μm. f RFI of FerroOrange for (d) (n = 6, biological replicates). g RFI of DCFH-DA staining for (e) (n = 6, biological replicates). h Immunofluorescence analysis of intracellular lipid peroxidation in BMDMs infected with H37Rv strain (independently repeated two times). Blue indicates nuclei, green represents lipid peroxidation, and red indicates non-peroxidized lipids. Scale bars, 25 μm. i Quantification of lipid peroxidation and non-peroxidized lipids for this figure (h) (n = 6, biological replicates). j TUNEL staining revealed apoptosis in GW9662-treated and H37Rv-infected BMDMs (independently repeated three times). Scale bars, 20 μm. k The corresponding statistical analysis results of TUNEL staining (n = 6, biological replicates). All results are expressed as mean ± s.e.m. One-way ANOVA for multiple comparisons. All raw data are provided as a Source Data file.

Collectively, these results demonstrate that M.tb-induced accumulation of PPARγ plays a central role in promoting ferroptosis in infected macrophages. This process involves multiple hallmarks of ferroptosis, including iron accumulation and lipid peroxidation. The regulation of GPX4 by PPARγ provides a molecular mechanism for this phenomenon, offering insights into how M.tb manipulates host cell death pathways to potentially benefit its own survival and persistence.

PPARγ modulates macrophage mitochondrial metabolism during M.tb infection

Our results indicate that M.tb modulates lipid metabolism and promotes ferroptosis via the PPARγ-GPX4 axis. However, mitochondrial dynamics are also increasingly recognized as key regulators of ferroptosis and remain to be elucidated in the context of M.tb infection^46,47. Transmission electron microscopy (TEM) results demonstrated that H37Ra infection disrupts mitochondrial morphology, a key morphological feature associated with ferroptosis, whereas PPARγ inhibition partially restored mitochondrial morphology (Fig. 4a, b). These alterations were accompanied by increased expression of mitofusin 1 (MFN1) and dynamin-related protein 1 (DRP1) in H37Rv-infected BMDMs, suggesting modulation of mitochondrial fusion-fission dynamics following PPARγ inhibition (Fig. 4c, d)⁴⁸. Interestingly, PGC1α, a key regulator of mitochondrial metabolism, was also significantly upregulated upon PPARγ inhibition (Fig. 4c, d). This finding suggests a potential link between PPARγ-associated lipid metabolism and M.tb-induced metabolic reprogramming. In BMDMs, H37Rv infection markedly decreased mitochondrial membrane potential, as indicated by decreased MitoTracker Red staining (Fig. 4e, g). However, PPARγ inhibition largely restored the potential (Fig. 4e, g), supporting an association between mitochondrial dysfunction and PPARγ-GPX4-regulated ferroptosis. To further investigate the role of mitochondrial metabolism in PPARγ-mediated cellular responses in macrophages, we treated BMDMs with Mito-TEMPO, a mitochondria-targeted antioxidant⁴⁹. Treatment with Mito-TEMPO reduced the accumulation of mitochondrial ROS in H37Rv-infected macrophages and significantly reduced iron accumulation (Supplementary Fig. 4k–n). In fact, depletion of mitochondrial ROS by Mito-TEMPO also restored mitochondrial membrane potential and reduced apoptosis in H37Rv-infected cells (Fig. 4f, h and Supplementary Fig. 4o, p). This was further accompanied by the reduction of cellular lipid peroxidation levels (Fig. 4i–k). However, unlike PPARγ inhibition (Fig. 1i, k), which decreased intracellular M.tb viability in macrophages, antioxidant treatment did not alter the clearance of M.tb in BMDMs (Fig. 4l, m). This finding was further confirmed by H37Rv culture results (Fig. 4n). These results suggest that mitochondrial metabolism and PPARγ-driven ferroptosis play distinct roles in macrophage-mediated anti-M.tb immunity.

Fig. 4. Mitochondrial metabolism changes in BMDMs upon M.tb infection.

a Transmission electron microscopy (TEM) observation of mitochondrial morphological changes in RAW264.7 cells infected with the H37Ra strain. Scale bars, 500 nm. b The mitochondrial mean length analysis of TEM results (n = 6, biological replicates). c Western blot analysis of MFN1 and DRP1 (mitochondrial dynamics), TOM20, and PGC1α in H37Rv-infected BMDMs treated with GW9662 (independently repeated two times). d The corresponding statistical quantification results of mitochondrial proteins of this figure (c) (n = 6, biological replicates). e MitoTracker staining was used to assess mitochondrial integrity in H37Rv-infected BMDMs treated with GW9662 (independently repeated two times). Scale bars, 10 μm. f MitoTracker staining was used to assess mitochondrial integrity in H37Rv-infected BMDMs treated with Mito-TEMPO (independently repeated two times). Scale bars, 10 μm. g The corresponding statistical analysis results of MitoTracker staining for this figure (e) (n = 6, biological replicates). h The corresponding statistical analysis results of mitoTracker staining for this figure (f) (n = 6, biological replicates). i Immunofluorescence detection of intracellular lipid peroxidation in BMDMs infected with H37Rv and antioxidant treatment (independently repeated three times). Scale bars, 25 μm. j, k RFI of lipid peroxidation (j) and non-peroxidized lipids (k) for (i) (n = 6, biological replicates). l Immunofluorescence detection of M.tb intracellular survival in BMDMs infected with H37Rv and antioxidant treatment (independently repeated two times). Scale bars, 20 μm. m The corresponding quantification of H37Rv signals for this figure (l) (n = 6, biological replicates). n Intracellular bacterial burden in BMDMs infected with H37Rv and treated with antioxidant was quantified by CFU assays (n = 3, biological replicates). All results are expressed as mean ± s.e.m. Two-tailed t-test for comparing two groups and one-way ANOVA for multiple comparisons. All raw data are provided as a Source Data file.

PPARγ recruits the NCOR/SMRT complex to repress Gpx4 transcription and promote ferroptosis

Our previous studies systematically investigated the regulatory mechanisms of transcriptional repression in macrophages and provided mechanistic insights into nuclear receptor-mediated transcriptional repression^50,51. To elucidate the mechanism by which PPARγ represses Gpx4 transcription, we performed chromatin immunoprecipitation and sequencing (ChIP-seq) analysis of PPARγ and histone H3 lysine 27 acetylation (H3K27ac) in H37Ra-infected macrophages. This analysis revealed that after infection, PPARγ chromatin binding increased globally across the macrophage genome (Fig. 5a). Based on these observations, we hypothesized that PPARγ might directly regulate GPX4-mediated ferroptosis^52,53. To test this hypothesis, we then analyzed ChIP-seq data, which revealed that PPARγ binds to the promoter of Gpx4 and affects its transcription in macrophages (Fig. 5b). Notably, the PPARγ binding to the Gpx4 promoter increased by 53% while the H3K27ac decreased by 13% (Fig. 5b). These findings were further validated by ChIP-qPCR, which confirmed increased PPARγ binding and decreased H3K27ac at both the Gpx4 promoter and enhancer regions in infected macrophages (Fig. 5c). These results were independently corroborated by ChIP-PCR, providing further evidence that H37Ra infection enhances PPARγ-mediated repression of Gpx4 (Fig. 5d).

Fig. 5. PPARγ recruits NCOR/SMRT to inhibit Gpx4 transcription and chromatin accessibility.

a Heatmap analysis of ChIP-seq peaks for PPARγ and H3K27ac in RAW264.7 cells infected with the H37Ra strain. b IGV genome browser tracks representing the PPARγ and H3K27ac ChIP-seq peaks at the Gpx4 locus in control and infected with H37Ra strain in RAW264.7 cells. c ChIP-qPCR analysis of PPARγ and H3K27ac at the Gpx4 promoter and enhancer regions (n = 3, biological replicates). d ChIP-PCR analysis of PPARγ and H3K27ac at the Gpx4 promoter region (n = 2, biological replicates). e MA plot showing log2-fold changes of ATAC-seq peaks between control and infected with H37Ra strain in RAW264.7 cells. The significantly upregulated (red) or downregulated peaks (blue). Peaks at marker gene loci are highlighted in green (n = 3, biological replicates). f IGV genome browser tracks of ATAC-seq data at the Gpx4 locus in control and infected with H37Ra strain in RAW264.7 cells (n = 3, biological replicates). g Gpx4-qPCR analysis of RAW264.7 cells after NCOR and SMRT knockdown and infection with H37Ra strain (n = 3, biological replicates). h PPARγ and GPX4 western blot analysis of RAW264.7 cells after NCOR and SMRT knockdown and infection with H37Ra strain (independently repeated three times). i, k IF detection of TUNEL and intracellular lipid peroxidation in RAW264.7 cells after NCOR and SMRT knockdown followed by H37Ra infection (independently repeated two times). Scale bars, 50 μm. j, l The corresponding quantification of immunofluorescence signals for this figure (i, k) (n = 6, biological replicates). m Schematic diagram of PPARγ recruiting NCOR/SMRT to suppress Gpx4 gene transcription in the nucleus. The illustration was created in BioRender. Zhao, C. (2026) https://BioRender.com/2n2ys5k. All results are expressed as mean ± s.e.m. One-way ANOVA for multiple comparisons. All raw data are provided as a Source Data file.

To gain a broader understanding of the changes in chromatin accessibility induced by M.tb infection, which may reflect the consequences of nuclear-enriched PPARγ, we performed assay for transposase-accessible chromatin with sequencing (ATAC-seq) in RAW264.7 cells (Supplementary Fig. 5a). The results indicated that the chromatin regions becoming more accessible after H37Ra infection were associated with the AP1 family proteins, including ELF4, CEBP, TGIF1, and IRF8 (Supplementary Fig. 5b, c). In contrast, the less accessible chromatin regions were predominantly associated with the PU.1 family proteins, such as SP1, TFE3, NRF, and RXR (Supplementary Fig. 5d, e). Importantly, H37Ra infection reduced the chromatin accessibility at the Gpx4 promoter by 53% (Fig. 5e, f).

Given the involvement of the NCOR/SMRT corepressor complex in gene regulation⁵⁴, we hypothesized that NCOR/SMRT may participate in PPARγ-mediated transcriptional regulation (Supplementary Fig. 5f, g). Remarkably, depletion of either NCOR or SMRT significantly restored GPX4 expression in H37Ra-infected RAW264.7 cells (Fig. 5g, h). To investigate the direct regulatory role of the NCOR/SMRT complex on Gpx4 at the transcriptional level, we performed knockdown experiments in BMDMs. Notably, NCOR knockdown induced significant transcriptional activation of Gpx4 and other ferroptosis-associated genes (Supplementary Fig. 5h, i). Consistent with these findings, SMRT depletion similarly resulted in moderate but discernible upregulation of Gpx4 expression at the transcriptional level (Supplementary Fig. 5j, k). To further elucidate the regulatory relationship, we examined the effects of NCOR and SMRT overexpression on GPX4 expression. Individual overexpression of NCOR or SMRT significantly suppressed GPX4 protein expression (Supplementary Fig. 5n, o). Notably, this suppressive effect was markedly enhanced by pharmacological activation of PPARγ with rosiglitazone, suggesting a synergistic regulatory mechanism through PPARγ signaling pathways (Supplementary Fig. 5n, o). IF analysis further confirmed that the depletion of NCOR or SMRT attenuated infection-induced lipid peroxidation and ferroptosis (Fig. 5i–l). These findings collectively demonstrate that M.tb infection facilitates the recruitment of the NCOR/SMRT complex via PPARγ to repress Gpx4 transcription, thereby driving lipid peroxidation and ferroptosis (Fig. 5m).

To relate our findings to a broader range of infectious stimuli, we performed epigenomic profiling targeting PPARγ in macrophages using CUT&Tag and compared these results with relevant data from our previous studies (Supplementary Fig. 5)⁴⁹. Activation of PPARγ and TLR4 signaling both significantly decreased the transcriptional activity of Gpx4, whereas activation of the IL-4 anti-inflammatory pathway and retinoid X receptors (RXR) did not have this effect (Supplementary Fig. 5l). Lipopolysaccharide (LPS)-mediated pathway activation, known to be a critical mechanism of cellular immunity mediated by various pathogens, significantly affected the transcriptional repression of Gpx4 (Supplementary Fig. 5m). To determine whether M.tb-induced ferroptosis shares conserved mechanisms with other infectious pathogens, we employed Klebsiella pneumoniae-derived LPS as a comparative control. Both H37Ra infection and LPS treatment comparably downregulated GPX4 expression (Supplementary Fig. 5p, q). Intriguingly, pharmacological blockade of PPARγ specifically reversed GPX4 suppression in H37Ra-infected cells, suggesting a potential regulatory axis involving PPARγ signaling that is distinct from LPS-mediated pathways. This analysis provides insight into the specificity of the M.tb-induced Gpx4 transcriptional repression and its potential relevance to other infections.

Targeting PPARγ attenuates lung inflammation and reduces bacterial burden during M.tb infection

To further assess the role of PPARγ targeting during virulent H37Rv infection, H37Rv-infected mice were treated with the PPARγ antagonist GW9662 in a BSL-3 facility (Fig. 6a). PPARγ expression was significantly upregulated in H37Rv-infected lung tissues (Fig. 6b, c). At 4 weeks post-H37Rv infection, GW9662 treatment significantly improved disease phenotype (Supplementary Fig. 6a). By 8 weeks post-infection, lungs exhibited extensive inflammatory infiltration, which was markedly alleviated by PPARγ inhibition (Fig. 6d).

Fig. 6. Targeting PPARγ alleviates pulmonary inflammation and reduces the bacterial load in mice.

a Establishment of H37Rv aerosol infection in C57BL/6 mice treated with GW9662. The illustration was created in BioRender. Zhao, C. (2026) https://BioRender.com/smlelax. b Immunofluorescence staining of PPARγ in the lungs of H37Rv-infected mice (repeated two times). Scale bars, 200 μm. c The corresponding quantification of immunofluorescence signals for this figure (b) (n = 6, biological replicates, two-tailed t-test). d H&E staining shows lung pathology in H37Rv-infected mice at eight weeks post-infection. Scale bars, 500 μm. e H&E staining reveals lung histopathology in H37Ra-infected PPARγ KO mice. Scale bars, 2 mm. f Microscopic imaging of stained inflammatory cells in BALF from PPARγ knockout mice. Scale bars, 100 μm. g The count of the inflammatory cells in the BALF from PPARγ KO mice (n = 6, biological replicates). h The protein level in the BALF (n = 6, biological replicates; one-way ANOVA). i CFU of lung bacterial load in mice 8 weeks after infection with H37Rv strain (n = 5, biological replicates; two-tailed t-test). j CFU of lung bacterial load in PPARγ KO mice after infection with H37Ra strain (n = 10, biological replicates; two-tailed t-test). k Acid-fast staining of lung bacteria in mice 8 weeks after infection with H37Rv strain. Scale bars, 20 μm. l Flow chart of the flow cytometry workflow for lungs from H37Rv-infected mice. m Macrophage and T cell alterations in the lungs of H37Rv-infected mice (n = 5, biological replicates; one-way ANOVA). n Macrophage and T cell alterations in the lungs of H37Ra-infected PPARγ KO mice (n = 6, biological replicates; one-way ANOVA). All results are expressed as mean ± s.e.m. One-way ANOVA for multiple comparisons. All raw data are provided as a Source Data file.

We employed tamoxifen-inducible PPARγ knockout mice to further validate the physiological relevance of PPARγ targeting in the context of H37Ra infection (Supplementary Fig. 6b–d). Immunoblot analysis confirmed that tamoxifen induction led to a sustained decrease in PPARγ expression in lung tissues and BMDMs (Supplementary Fig. 6c). Consistent with our in vitro findings (Fig. 1), H37Ra infection also led to a significant increase in PPARγ expression in lung tissues (Supplementary Fig. 6e–h). Consistent with the GW9662 inhibition experiments, PPARγ knockout significantly reduced H37Ra-induced inflammatory infiltration in the lungs (Fig. 6e). This effect appeared to be lung-specific, as we observed no significant changes in spleen weight (Supplementary Fig. 6i), suggesting a localized rather than systemic impact of PPARγ deletion. Analysis of bronchoalveolar lavage fluid (BALF) revealed a marked decrease in inflammatory cells in PPARγ knockout mice following H37Ra infection (Fig. 6f, g). This reduction in cellular infiltration was accompanied by a significant decrease in BALF protein levels (Fig. 6h), indicating reduced lung inflammation and vascular permeability.

Notably, both pharmacological inhibition and genetic ablation of PPARγ significantly reduced both H37Rv and H37Ra growth (Fig. 6i–k and Supplementary Fig. 6j–l). Consistently, acid-fast staining showed a significant reduction in M.tb burden in GW9662-treated mice infected with H37Rv and in PPARγ knockout mice infected with H37Ra (Fig. 6k and Supplementary Fig. 6j). We next performed flow cytometric analysis of lung tissues collected at 4 and 8 weeks post-infection (Fig. 6l). PPARγ inhibition significantly rescued the H37Rv-associated loss of macrophages, whereas a reduction in CD3⁺ T cell accumulation was observed only at 8 weeks (Fig. 6m and Supplementary Fig. 6m, n). These immune cell compositions were further corroborated in the H37Ra-infected PPARγ knockout mouse model (Fig. 6n and Supplementary Fig. 6o). These findings suggest that PPARγ plays a key role in regulating macrophage survival or recruitment during infection⁵⁵.

Collectively, our results in PPARγ-deficient mice indicate that PPARγ contributes to infection-associated pathology and modulates pulmonary bacterial burden. Moreover, pharmacological inhibition of PPARγ with GW9662 in the H37Rv model highlights PPARγ as a promising target for host-directed therapy in TB.

PPARγ modulation restores GPX4 expression and mitigates ferroptosis in lung macrophages during M.tb infection

Based on in vitro findings and the observed effects of PPARγ knockout on lung inflammation and bacterial burden, we further investigated the molecular mechanisms underlying these phenomena in vivo. Immunoblot analysis of lung tissues revealed a clear association with the ferroptosis pathway. GPX4 expression, which was significantly inhibited after H37Rv or H37Ra infection, markedly increased following PPARγ knockout or GW9662 treatment (Fig. 7a–d). Concurrently, 4HNE levels were significantly increased after infection but were decreased upon PPARγ knockout, whereas SLC7A11 remained relatively unchanged (Fig. 7c, d). These results were consistent with in vitro findings and support the hypothesis that PPARγ regulates GPX4 expression and ferroptosis during H37Ra infection in vivo. Immunohistochemical (IHC) staining confirmed increased PPARγ expression in response to both H37Ra and H37Rv infection (Fig. 7e, f and Supplementary Fig. 7a, b). In both infection models, PPARγ knockout or inhibition significantly attenuated M.tb-induced 4HNE accumulation. This reduction was accompanied by increased GPX4 expression in lung tissues upon PPARγ deletion or inhibition (Fig. 7g, h and Supplementary Fig. 7c–e). IF co-localization further confirmed these GPX4 expression patterns in both infection models (Fig. 7i, j and Supplementary Fig. 7f, g).

Fig. 7. PPARγ inhibition ameliorates M.tb-induced macrophage ferroptosis in mouse lungs.

a Western blot analysis of ferroptosis-related proteins in lung tissues after infection with H37Rv and GW9662 treatment (n = 2, biological replicates, independently repeated three times). b The corresponding quantification of proteins expression for this figure (a) (n = 6). c Western blot analysis of ferroptosis-related proteins in lung tissue after infection with H37Ra in PPARγ KO mice (n = 2, biological replicates, independently repeated three times). d The corresponding quantification of proteins expression for this figure (c) (n = 6). e IHC of PPARγ expression in mouse lung after infection with H37Rv. Scale bars, 100 μm (repeated two times). f The corresponding quantification of PPARγ expression for this figure (e) (n = 6). g IHC detection of GPX4 and 4HNE in mouse lung tissue after H37Rv infection and GW9662 treatment. Scale bars, 100 μm (independently repeated two times). h Relative quantification of GPX4 and 4HNE IHC for this figure (g) (n = 6). i Fluorescence colocalization staining of F4/80 (red) and GPX4 (yellow) in mouse lung tissue after infection with the H37Rv strain. Scale bars, 100 μm (repeated two times). j The corresponding quantification of proteins expression for this figure (i) (n = 6). k Immunofluorescence staining of mitochondria-related proteins in lung tissues from H37Rv-infected mice with GW9662 treatment (independently repeated two times). Scale bars, 200 μm. l The corresponding quantification of proteins expression for this figure (k) (n = 6). All results are expressed as mean ± s.e.m. Two-tailed t-test for comparing two groups and one-way ANOVA for multiple comparisons. All raw data are provided as a Source Data file.

Analysis of serum lipid profiles revealed that high-density lipoprotein (HDL) levels, which were significantly decreased after H37Ra infection, were markedly restored following PPARγ knockout (Supplementary Fig. 7h). This finding suggests a potential role for PPARγ in modulating systemic lipid metabolism during infection^56–58. Additionally, serum levels of malondialdehyde (MDA), a marker of lipid peroxidation, increased after infection and showed a trend toward reduction following PPARγ knockout, although this change was not statistically significant (Supplementary Fig. 7h). To directly assess the impact of PPARγ on macrophage survival during M.tb infection, we infected BMDMs with H37Ra. We observed significant cell death after H37Ra infection, which was markedly reduced following PPARγ knockout (Supplementary Fig. 7i, j). This result provides evidence that PPARγ-mediated ferroptosis contributes to macrophage death during M.tb infection. Inhibition of PPARγ reduced F4/80⁺ macrophage infiltration in the lungs of H37Rv-infected mice (Fig. 7k, l). Consistent with our findings in BMDMs, PPARγ inhibition in H37Rv-infected mouse lungs significantly improved mitochondrial membrane morphology, as reflected by restored translocase of the outer mitochondrial membrane 20 (TOM20) expression (Fig. 7k, l). In addition, GW9662 treatment significantly reduced H37Rv-induced mitochondrial stress-associated mitochondrial transcription factor A (TFAM) expression in lung tissue (Fig. 7k, l).

In summary, using two M.tb strains with distinct virulence (H37Ra and H37Rv), we demonstrate that M.tb infection induces PPARγ upregulation in lung tissue in vivo, suppressing GPX4 and promoting lipid peroxidation and ferroptosis. PPARγ deletion or pharmacological inhibition restores GPX4 expression, reduces lipid peroxidation, and enhances macrophage survival during M.tb infection. These results collectively establish a crucial role for PPARγ in mediating ferroptosis during M.tb infection. These findings highlight the potential of targeting the PPARγ-GPX4 axis as a host-directed therapeutic strategy for TB, aimed at preserving macrophage function and enhancing bacterial clearance (Fig. 8).

Fig. 8. Mechanisms of M.tb-mediated regulation of ferroptosis in macrophages via PPARγ.

During M.tb infection, the bacterial isocitrate dehydrogenase (IDH), which hijacks the PPARγ protein and inhibits its phosphorylation at Ser112. This inhibition leads to the accumulation of PPARγ in cells, which in turn causes macrophages to accumulate large amounts of lipid droplets, providing nutrients for bacterial survival. The excessive accumulation of PPARγ not only supports nutrient storage in macrophages but also results in its translocation to the nucleus. PPARγ recruits the NCOR/SMRT complex, directly inhibiting the transcription and chromatin accessibility of the Gpx4 gene. This inhibition leads to ferroptosis in macrophages, thereby facilitating the dissemination of M.tb. The illustration was created in BioRender. Zhao, C. (2026) https://BioRender.com/uw9jxg4.

Discussion

Pathogen infection of host cells often alters host lipid metabolism through multiple mechanisms, with inflammatory signals and hypoxic microenvironments contributing to this process and creating conditions favorable for pathogen survival⁵⁹. Macrophages are highly phagocytic innate immune cells that play a crucial role in combating pathogens⁶⁰. We found that after M.tb infects macrophages, the bacteria release the metabolic enzyme IDH. A key unresolved question concerns how bacterial IDH, a non-classically secreted enzyme, gains access to the host intracellular environment to interact with PPARγ. Our data demonstrate that IDH is not detectable in culture supernatants of M.tb, arguing against a canonical active secretion mechanism. Instead, IDH becomes detectable during macrophage infection, indicating that its release is infection-dependent. One plausible explanation, in line with previous studies, is that IDH is released following the lysis of bacteria within the harsh environment of the host macrophage. Following lysis, the extensive phagosomal permeabilization driven by the ESX-1 secretion system of surviving bacilli allows exposed bacterial cytosolic proteins to translocate into the host cytosol⁶¹. This mechanism mirrors the delivery of other leaderless mycobacterial effectors, such as PtpA²⁴. Furthermore, it aligns with the multifunctional properties of other released metabolic enzymes, such as GAPDH and Ndk, which exert secondary immunomodulatory functions⁶². This strategy of utilizing membrane disruption to release effectors is also conserved in other pathogens, such as Listeria monocytogenes, which relies on listeriolysin O for vacuolar rupture⁶³. Alternatively, non-canonical release mechanisms, such as extracellular vesicle-mediated export, cannot be excluded and warrant further investigation^64,65. Released IDH disrupts host metabolism by hijacking PPARγ and inhibiting its phosphorylation, leading to changes in PPARγ activity and nuclear localization. This mechanism differs from previously described active defense strategies by M.tb (e.g., PtpA), highlights the potential role of infection-dependent IDH release in modulating host cell signaling²⁴. IDH-mediated abnormal activation of PPARγ promotes LDs accumulation in macrophages, which provides potential nutrients for M.tb and also serves as a key carrier for host immune responses⁶⁶. The accumulation of PPARγ facilitates the recruitment of NCOR/SMRT corepressor complexes, resulting in the repression of Gpx4 and a consequent decrease in GPX4 protein, thereby triggering ferroptosis. This ferroptosis is a significant driver of M.tb spread. Consequently, M.tb uses intracellular parasitism and reprograms host lipid metabolism to establish a unique escape and survival strategy. The cell wall of M.tb is composed of a large number of diverse lipid components. Mycobacteria can hijack host cells and promote LDs accumulation to establish a cellular environment critical for their intracellular survival^35,58. Therefore, lipids are considered important for the survival, invasion, parasitism, and proliferation of mycobacteria within host cells. Studies have shown that mycobacteria regulate PPAR signaling and utilize host-derived triacylglycerol (TAG) and cholesterol as nutrient sources and to evade the host immune system^67–70. Our study also extends these findings. Through single-cell transcriptomics analysis of human PBMCs, we discovered the activation of lipid metabolism-related pathways in macrophages following M.tb infection. Additionally, activation of the PPAR pathway was observed in RNA-seq analysis of M.tb-infected macrophages, and we also observed LDs accumulation in these cells. In future work, we will continue to investigate the biological functions of PPAR in different cell subsets during M.tb infection.

A recent investigation highlighted the critical role of PPAR-mediated lipid metabolism in the formation of foamy macrophages following infection with M.tb. Among the PPARs, PPARγ has been particularly implicated in mycobacterial infections. For instance, M.tb infection disrupts host lipid homeostasis and promotes the development of foamy macrophages, which are essential for the intracellular survival and proliferation of the pathogen⁷¹. The virulent strain H37Rv of M.tb has been shown to upregulate PPARγ expression³⁵, while the attenuated strain Bacillus Calmette-Guerin (BCG) demonstrates a lesser degree of PPARγ activation⁷². Experimental disruption of PPARγ signaling in macrophages infected with M.tb results in reduced intracellular lipid accumulation and enhanced bactericidal activity³⁴. Furthermore, treatment with a PPARγ antagonist significantly inhibits the accumulation of intracellular LDs induced by both BCG and M.tb⁷³. Additionally, the growth of M.tb is diminished in human lung macrophages following PPARγ knockout or in PPARγ-deficient mice. Collectively, these findings underscore the necessity of PPARγ for foam macrophages formation within tuberculous granulomas, which is closely associated with the survival of M.tb. Our study also supports these findings; however, the specific molecular mechanisms by which M.tb infection leads to PPARγ accumulation remain unclear.

Interestingly, our study found that following M.tb infection, PPARγ mRNA levels were suppressed, while protein levels significantly increased. This suggests that M.tb infection regulates PPARγ at the PTM level, leading to its accumulation. We further found that M.tb IDH is involved in the post-translational regulation of PPARγ. The genome of M.tb contains two isoforms of isocitrate dehydrogenase, namely Rv3339c (ICD-1) and Rv0066c (ICD-2)^74,75. We found that the M.tb gene Rv3339c (ICD-1) is involved in the regulation of PPARγ. Research indicates that the ability of M.tb ICD-1 to tolerate a broader range of pH and temperature compared to ICD-2, highlighting its robustness⁷⁵. This suggests that Rv3339c can better adapt to the host cell environment while interacting with host proteins. Additionally, studies have shown that both ICD-1 and ICD-2 are NADP-dependent members of the ICD family, with ICD-1 exhibiting greater homology to eukaryotic ICDs, while ICD-2 is more closely related to prokaryotic counterparts⁷⁶. These studies also provided a foundation for our findings. Furthermore, we report a previously uncharacterized role of M.tb ICD-1 in host macrophages. Our study found that M.tb IDH hijacks host PPARγ, inhibiting phosphorylation at the Ser112 site, which results in its inability to be degraded. However, a limitation of our study is that it remains unclear whether IDH directly blocks the phosphorylation sites of PPARγ, which requires further investigation.

During M.tb infection, PPARγ-mediated macrophage dysfunction not only leads to foamy macrophages formation but also influences inflammatory responses, autophagy, and fibrosis^77–79. Research has reported that PPARγ inhibits IL-1β, IL-6, and TNF-$\mathrm{\alpha }$ in human monocytes stimulated with phorbol 12-myristate 13-acetate⁸⁰. Human macrophages activate PPARγ through the mannose receptor CD206 when phagocytosing M.tb, thereby reducing pro-inflammatory responses³⁵. In mouse models, the absence of PPARγ in lung macrophages enhances pro-inflammatory cytokine production and reduces M.tb growth³⁶. Mouse macrophages treated with mannose-capped lipoarabinomannan, either alone or alongside the PPARγ antagonist GW9662, showed that PPARγ inhibition downregulates Prkag2 expression, which is essential for AMPK activation^78,81,82. Our study reveals a function mediated by PPARγ during M.tb infection. We have substantial evidence that, during M.tb infection, PPARγ directly inhibits the transcription of the Gpx4 gene by recruiting the NCOR/SMRT complex, leading to ferroptosis in macrophages. Iron overload plays a key role in ferroptosis, as it promotes lipid peroxidation when GPX4 activity is inhibited, leading to plasma membrane damage^83–85. Ferroptosis is induced by hydrogen peroxide generated from the Fenton reaction, which interacts with membrane lipids to produce toxic lipid peroxides. Under homeostatic conditions, these lipid peroxides are rapidly reduced by GPX4 through GSH oxidation. However, when the expression or activity of GPX4 is inhibited under conditions of iron overload, the peroxidized lipids accumulate uncontrollably, leading to cell death. Therefore, our study provides a mechanistic insight into this process. Recent studies have found that M.tb PtpA interacts with host RanGDP at its Cys11 site to enter the host cell nucleus. Once inside the nucleus, PtpA enhances the asymmetric dimethylation of histone H3 arginine 2 (H3R2me2a) by targeting arginine methyltransferase 6 (PRMT6), thereby inhibiting the expression of GPX4 and ultimately inducing ferroptosis²⁴. Our research also identified a molecular mechanism by which M.tb regulates GPX4 to induce ferroptosis, suggesting a potential target for anti-tuberculosis therapy.

Ferroptosis is recognized as a key contributor to M.tb pathogenesis and can be triggered by multiple distinct mechanistic pathways during infection^21,24,86. Unlike the M.tb PtpA-driven ferroptosis, the IDH-PPARγ axis appears to be an important passive survival mechanism used by M.tb²⁴. Consistent with M.tb’s reliance on host lipids, bacterial co-option of PPARγ increases macrophage lipid uptake and de novo lipogenesis, promotes foam cell formation, and exacerbates iron dyshomeostasis^87,88. We identify PPARγ as a central integrator of lipid and iron metabolic programs that also directly represses GPX4, thereby serving as a key trigger for ferroptosis. Although restoring mitochondrial ROS homeostasis attenuates M.tb-induced ferroptosis, it does not improve intracellular bacterial clearance, suggesting that improving mitochondrial function does not ensure the concurrent reversal of cell membrane damage caused by lipid peroxidation^89,90. These findings support a model in which M.tb evades macrophage cellular immunity by rewiring PPARγ-dependent lipid metabolism and promoting ferroptosis through GPX4 repression.

Here, we reveal that M.tb regulates PPARγ to trigger ferroptosis by directly targeting GPX4, thereby disrupting the core ferroptosis axis. This suggests that M.tb IDH is a key pro-ferroptotic effector in host-pathogen interactions. Additionally, our study shows that M.tb infection inhibits the activity of intracellular immunoproteasome. Therefore, we speculate that in addition to the M.tb IDH-PPARγ-GPX4 axis-mediated activation of ferroptosis, M.tb may also regulate host ferroptosis through additional pathways. Overall, our findings provide pathogen-induced mechanisms of ferroptosis and suggest that blocking host PPARγ to prevent GPX4 suppression and ferroptosis may represent a potential therapeutic approach for tuberculosis (Fig. 8). Our results also imply that diabetic patients with M.tb infection taking the thiazolidinediones class of hypoglycemic drugs may experience enhanced bacterial replication, adversely affecting their treatment.

Methods

Ethics statement

This study was approved by the Ethics Committee of the Second Affiliated Hospital of Nanjing University of Chinese Medicine (Nanjing Public Health Center, Approval No. 2022-LS-KY043), and written informed consent was obtained from all participants in accordance with the principles of the Declaration of Helsinki. All animal experiments in this study were approved by the Nanjing University Animal Care and Use Committee (IACUC-2003124). All animal experiments complied with Nanjing University’s animal ethics guidelines. The H37Ra M.tb animal infections were performed in a biosafety level 2 (BSL-2) laboratory. All animal infection and cell-based experiments were authorised by the BSL-3 laboratory at Shenzhen University and conducted under appropriate biosafety containment measures (Approval No. SZUABSL3-S20251003). The BSL-3 facility is accredited by the China National Accreditation Service for Conformity Assessment (CNAS; Registration No. CNASBM0134).

Clinical patient samples

All patients provided written informed consent. Clinical samples were collected from three healthy individuals and three tuberculosis patients. Inclusion criteria were as follows: male sex, non-smokers, non-drinkers; BMI < 24 kg/m²; normal blood glucose levels (fasting blood glucose 3.9–6.1 mmol/L, postprandial blood glucose < 7.8 mmol/L); patients who had not received medication after admission; blood was collected in EDTA anticoagulant tubes; 5 mL of blood was collected and stored at 4 °C. Detailed information is provided in Supplementary Table 1.

scRNA-seq of clinical patient samples

Single-cell sequencing was performed by Singleron Biotechnologies Co., Ltd. (Nanjing, China). Single-cell suspensions were prepared from whole blood samples using peripheral blood mononuclear cells (PBMCs) separation liquid and diluted to 2–2.5 × 10⁵ cells/mL. Single cells were isolated and labeled using SCOPE-chip microfluidic chips, based on Poisson distribution principles. Magnetic beads with unique cell barcodes and molecular barcodes (UMI) were added to the chip microwells to capture and label mRNA. Reverse transcription and amplification were performed on the captured mRNA, followed by the construction of single-cell sequencing libraries for sequencing on the Illumina platform.

scRNA-seq analysis

Raw data were quality controlled using CeleScope software to obtain an expression matrix. The standard analysis pipeline removed low-quality cells based on the number of genes, UMI counts, and mitochondrial gene content distribution. Scanpy 1.8.2 was used to cluster cells based on gene expression levels. Low-quality cells, cells with abnormally high gene counts, or high mitochondrial content were filtered out according to specific thresholds. Data normalization and dimensionality reduction were performed using Principal Component Analysis (PCA). The top 50 principal components were selected for cell clustering using the Louvain algorithm. UMAP was used for dimensionality reduction and visualization. CellID 0.1.0 was employed using MCA and HGT algorithms to automate cell type annotation based on the SynEcoSys single-cell database. Differentially expressed genes among cell types were subjected to enrichment analysis. Gene set enrichment analyses, including Gene Ontology (GO) functional enrichment, Reactome pathway enrichment, and KEGG pathway enrichment, were performed using the clusterProfiler 3.8.1^91,92.

RNA-seq analysis

In this study, we reanalyzed RNA sequencing (RNA-seq) data from our published study (GSE162729) to explore conserved regulatory mechanisms³⁷. The study examined four groups (control, H37Rv, H37Ra, and BCG infected), with three biological replicates in each group. We used bioinformatics approaches to identify overlapping differentially expressed genes (DEGs) in THP-1-derived macrophages infected with M.tb. The criteria for defining genes as DEGs were |log2 fold change| >1.0 and P < 0.05. Standard bioinformatics analysis, including PCA, gene expression, heat map, KEGG pathway, volcano plot, and Venn diagram, were performed by BGI. RNA-seq data from NCOR- and SMRT-knockdown BMDMs were obtained from our previous work (n = 3). We reanalyzed the GSE291538 dataset and generated volcano plots and heatmaps for selected genes⁹³. GO and KEGG plots were generated based on over-representation analysis using the clusterProfiler R package.

Bacterial culture

mCherry-H37Ra, H37Ra, mCherry-H37Rv, and H37Rv M.tb strains (Shanghai Gene-Optimal Science & Technology Co., Ltd.) were used in this study. Suspensions of the standard M.tb strains were prepared in BSL-2 and BSL-3 laboratories and plated on 7H10 agar (BD Middlebrook, Difco, Sparks). Cultures were incubated at 37 °C for 6–8 weeks. Single colonies were picked and inoculated into 7H9 liquid medium (BD Middlebrook, Difco, Sparks) containing 10% oleic acid-albumin-dextrose-catalase (OADC, Becton Dickinson Microbiology Systems) and 0.5% glycerol (Sigma). Liquid cultures were incubated without shaking at 37 °C for 4–6 weeks.

Cell culture and infection

RAW264.7 (ATCC, TIB-71) cells, BMDMs, and HEK293T (ATCC, CRL-3216) cells were cultured in DMEM (Gibco, 11965092) supplemented with 10% fetal bovine serum (Biochannel Biological Technology Co., Ltd., China) and 100 U/mL penicillin and streptomycin (SenBeiJia Biological Technology Co., Ltd., China). Cells were maintained in a humidified incubator at 37 °C with 5% CO₂. When cells reached 90% confluence, cells were passaged at a ratio of 1:10 every 2–3 days. For in vitro macrophage infection assays, RAW264.7 and BMDMs were seeded at 1 × 10⁶ cells per well in 6-well plates, and 5 × 10⁵ cells per well in 12-well plates (n = 3). To achieve the indicated MOI of 10 for the H37Ra strain, cells were infected with an absolute inoculum of 1 × 10⁷ bacilli per well (in 6-well plates) or 5 × 10⁶ bacilli per well (in 12-well plates). For infections with the H37Rv strain at an MOI of 5, the absolute inoculum was 5 × 10⁶ bacilli per well (in 6-well plates) or 2.5 × 10⁶ bacilli per well (in 12-well plates) (n = 3).

Mouse strains

Male wild-type C57BL/6J and Pparg^{fl/fl/Cre-ERT2} inducible knockout mice (C57BL/6J), aged 6–8 weeks, were purchased from the Model Animal Research Center of Nanjing University. All mice were randomly assigned to either the control group or the infection group. The mice were acclimatized for one week in an SPF-grade animal facility under controlled conditions: 45 ± 15% humidity, 12-h light/dark cycle, and a constant temperature of 24 ± 2 °C, with ad libitum access to food and water.

RNA extraction

Twenty-four hours post-infection, the culture medium was removed, and cells were washed twice with pre-chilled PBS. 2 mL of TRIzol (Vazyme Biotech Co., Ltd., China) reagent was added to the culture flask (n = 3), and the mixture was gently pipetted and incubated on ice for 3–5 min. Cells were collected, and chloroform (0.2 mL per milliliter of TRIzol) was added. The mixture was shaken for 15 s, incubated at room temperature for 2 min, and centrifuged at 13,000 × g at 4 °C for 15 min. The upper aqueous phase was carefully aspirated, and isopropanol (0.5 mL per milliliter of TRIzol) was added. The mixture was thoroughly mixed, incubated at room temperature for 10 min, and centrifuged at 13,000 × g at 4 °C for 10 min. The supernatant was discarded, and the pellet was washed with 70% ethanol (equal volume to TRIzol) and centrifuged at 5000 × g at 4 °C for 10 min. The pellet was air-dried for 5–10 min, resuspended in 30 μL of ddH₂O, and stored at −80 °C. Then, 1 μg of total RNA was reverse transcribed into cDNA using the HiScript III RT SuperMix Kit (Vazyme, R323-01). RT-qPCR analysis was performed using SYBR Green (ABclonal, RK21219) on a real-time fluorescence qPCR instrument (n = 3). Oligonucleotides for RT-qPCR were synthesized and provided by Sangon Biotech Co., Ltd. Relative mRNA expression was calculated using the 2^−ΔΔCt method and β-actin as the reference to assess the expression of the genes Gpx4, Pparg, Ncor, and Smrt. RT-qPCR primers were provided in Supplementary Table 2.

Protein extraction and quantification

Following the completion of the cell infection model, RAW264.7 cells in 6-well plates were washed three times with PBS. RIPA buffer (Beyotime Biotechnology, China) containing protease and phosphatase inhibitors (Vazyme Biotech Co., Ltd., China) was added, and cells were scraped off and collected on ice. The lysate was incubated on ice for 30 min and centrifuged at 13,000 × g for 15 min. The supernatant was transferred to a new EP tube. We quantified PPARγ and GPX4 in RAW264.7 cells using a nuclear-cytoplasmic fractionation approach. The Nuclear-Cytoplasmic Protein Extraction Kit (Beyotime, P0028) was used to isolate nuclear and cytoplasmic proteins according to the manufacturer’s standard protocol. Protein concentration was quantified using a BCA (Beyotime Biotechnology, China) protein assay kit. A standard curve was prepared by diluting the standard protein from 2 mg/mL to various concentrations. Cell proteins were diluted 1:4. BCA assay reagent A and B were mixed at a 50:1 ratio to prepare the working solution. In a 96-well plate, 200 μL of the working solution was added to each well, followed by 10 μL of diluted standards and protein samples. The plate was incubated at 37 °C for 30 min, and the OD value was measured at 570 nm using a microplate reader. A standard curve was fitted to calculate the protein concentration. For protein electrophoresis, reducing SDS-PAGE sample buffer was mixed with protein samples at a 1:4 ratio, heated at 95 °C for 10 min, and cooled to room temperature. Each experiment was replicated 2–3 times, and representative samples (n = 2–3) were chosen to present the results, ensuring they reflected the overall observed trends.

Protein electrophoresis and western blot

Gels were prepared according to Bio-Rad instructions. The separating gel was prepared by mixing 30% Acrylamide-Bis (29:1), separating gel buffer (1.5 M Tris-HCl, pH 8.8), distilled water, 10% APS, and TEMED. The stacking gel was prepared similarly using stacking gel buffer (1 M Tris-HCl, pH 6.8). Protein samples (30 μg) were loaded into the gel wells. Electrophoresis was performed at 80 V until samples reached the boundary between the stacking and separating gels, then increased to 120 V until the bromophenol blue reached the bottom of the separating gel. Post-electrophoresis, the gel was equilibrated in transfer buffer for 15 min. PVDF membranes and filter paper were cut and equilibrated in transfer buffer. The transfer stack was assembled with the membrane and gel in a transfer cassette, and proteins were transferred at 350 mA for 70 min. After transfer, the membrane was blocked in 5% BSA (Invitrogen, EN0521) in TBST at room temperature for 1 h. The membrane was incubated with diluted primary antibody in blocking solution at 4 °C overnight, then rinsed with TBST. The membrane was incubated with diluted secondary antibody at room temperature for 1 h, rinsed with TBST, and detected using enhanced chemiluminescence detection reagent. Chemiluminescence signals were scanned with a gel imaging system, and the results were recorded.

The primary antibodies used in this study were obtained from the following sources: GAPDH (60004-1-lg), PPARα (66826-1-Ig), and PPARγ (66936-1-Ig) from Proteintech (US); CD36 (A19016), Ubiquitin (A19686), p-STAT1 (Ser727, AP1000), p-STAT1 (Tyr701, AP0054), PSMB8 (A23378), PSMB9 (A9549), PSMB10 (A21123), SLC7A11 (A13685), β-tubulin (A12289), Lamin B (A5001), Flag (AE092), GPX4 (A1933), and β-actin (AC026) from ABclonal (China); 4-Hydroxynonenal (4-HNE, ab46545) from Abcam (Cambridge, UK); NCOR (AF0270) and SMRT (DF8896) from Affinity Biosciences (China); p-PPARγ (bs-3737R) from Bioss (China); and ZsGreen (TA180002S) from Origene (US). Secondary antibodies used were HRP AffiniPure Goat Anti-Mouse IgG (H + L) (FDM007) and HRP AffiniPure Goat Anti-Rabbit IgG (H + L) (FDR007), both from Fudebio-tech (China).

ChIP-seq

Chromatin immunoprecipitation followed by sequencing (ChIP-seq) was performed according to established protocols⁵⁰. Briefly, control and M.tb-infected RAW264.7 cells from 15 cm² culture dishes were cross-linked with 2 mM disuccinimidyl glutarate (ThermoFisher, 20593) for 30 min, followed by 1% formaldehyde for 10 min. The cross-linking reaction was quenched by adding 2.5 M glycine to a final concentration of 0.125 M and incubating for 5 min. Nuclei from lysed RAW264.7 cells were sonicated for 30 min (30 s ON/30 s OFF) using a Bioruptor Plus (Diagenode, B01020001). Protein A Dynabeads (Invitrogen, 10002D) were incubated with antibodies against H3K27ac (Abcam, ab4729) and PPARγ (Proteintech, 66936-1-Ig) (n = 2). Immunoprecipitated DNA was purified using the Clean & Concentrator Capped Zymo-Spin I kit (Zymo Research, D4013). Sequencing was performed on the NovaSeq 6000 platform (Illumina) using 150 bp paired-end reads. FASTQ files were aligned to the mouse reference genome NCBI37/mm9 using Bowtie2 within the Galaxy-Europe platform⁹⁴. Sequencing tags were processed and imported into HOMER⁹⁵. Peak calling was performed using HOMER with default settings and minor adjustments to parameters for histone marks and transcription factors/coregulators. Overlapping peaks were determined by combining individual peak files. Heatmaps were generated using deepTools⁹⁶. ChIP-PCR experiments were conducted using primers listed in Supplementary Table 2.

ATAC-seq

Chromatin accessibility was assessed using the Hyperactive ATAC-Seq Library Prep Kit for Illumina (Vazyme, HUF302900) following the manufacturer’s protocol. Briefly, nuclei were isolated from cell samples and subjected to Tn5 transposase-mediated fragmentation and adapter ligation in a single reaction. The resulting libraries were amplified and purified according to the kit instructions. The prepared libraries were then sequenced on an Illumina platform at Personal Biotechnology Co., Ltd. (NovaSeq 6000, PE150) to generate paired-end reads (n = 3). The sequencing data were further processed for downstream analysis. FASTQ files were aligned to the mm9 genome using Bowtie2, and peak calling was performed using MACS2⁹⁷. The resulting BAM files were then analyzed using HOMER⁹⁵. Differential accessibility analysis between the control and M.tb-infected groups was performed using DESeq2, with peaks considered differential if the adjusted P < 0.05. Data visualization was performed using the IGV genome browser and RStudio.

CUT&Tag

CUT&Tag experiments were performed according to a published protocol using the H3K27ac antibody⁹⁸. RAW264.7 cells were treated with rosiglitazone (Rosi, 5 µM, a PPARγ ligand) and retinoic acid (RA, 3 µM, an RXR ligand) for 1 h prior to CUT&Tag assays (n = 2). Library samples were sequenced on the NextSeq 2000 platform (PE100) with paired-end output at the BEA Core Facility, Karolinska Institutet, Sweden. The results of the CUT&Tag experiments were then compared with our dataset, including LPS and IL-4 treatments in RAW264.7 cells (GSE130383, GSE184884, and GSE235408)^50,51,93. The analysis of the CUT&Tag results was performed as described for ATAC-seq above.

Immunofluorescence (IF)

RAW264.7 cells and BMDMs were seeded on coverslips in 12-well plates at 1 × 10⁵ cells per well and infected when 60% confluent. After 24 h, cells were washed with PBS, fixed with paraformaldehyde for 30 min, permeabilized with 0.3% Triton X-100 for 20 min, and blocked with 3% BSA for 30 min. Primary antibodies were diluted (1:200) and incubated overnight at 4 °C. Fluorescent secondary antibodies (1:1000) were added for 1 h at room temperature, followed by DAPI staining (1:1000). Coverslips were mounted using antifade medium, and images were captured using an Olympus FV3000 laser confocal microscope. High-content imaging and analysis were performed on a Megarobo system (CellVue® T3000) equipped with 20, 40, and 60× objectives, with image acquisition controlled by the CellVue® T3000 software. Fluorescence intensity was calculated using Fiji software by multiplying the average relative fluorescence intensity (RFI) by the stained area.

For tissue section immunofluorescence staining, paraffin sections were dewaxed, rehydrated, and subjected to antigen retrieval in sodium citrate buffer. Sections were permeabilized with 0.5% Triton X-100, blocked with 3% BSA, and incubated with primary antibodies overnight at 4 °C. Fluorescent secondary antibodies (1:2000) were applied for 2 h at room temperature, followed by DAPI staining (1:2000). Sections were mounted using antifade mounting medium and imaged using a confocal microscope. The primary antibodies used in this study were obtained from the following sources: PPARγ (66936-1-Ig) from Proteintech (US), Flag (AE092) from ABclonal (China), ZsGreen (TA180002S) from Origene (US). Secondary antibodies used were Alexa Fluor 488-labeled Goat Anti-Rabbit IgG (H + L) (A0423) and Cy3-labeled Goat Anti-Rabbit IgG (H + L) (A0516), both from Beyotime (China). For assessment of MitoTracker, LDs, ROS, FerroOrange, and C11 BODIPY 581/591, staining was performed exclusively on live cells without fixation. To ensure biosafety compliance, all unfixed samples infected with H37Rv were imaged using a confocal microscope equipped with a live-cell imaging workstation located within the BSL-3 facility. Each experiment was replicated 2–3 times, and representative samples were chosen to present the results, ensuring they reflected the overall observed trends.

Image analysis and fluorescence quantification

All quantitative analyses of the microscopy-based readouts (including PPAR$\mathrm{\gamma }$, H37Rv, H37Ra, TUNEL, F4/80, GPX4, TFAM, TOM20, MitoTracker, LDs, ROS, FerroOrange, and C11 BODIPY 581/591) were performed using ImageJ/Fiji software (NIH). To ensure rigorous and unbiased quantification, identical thresholds and background subtraction parameters were applied across all images within a given experiment.

To control for potential variations in cell density, the total number of cells per field of view (FOV) was consistently counted based on nuclear staining (DAPI or Hoechst). Statistical analysis confirmed no significant differences in cell numbers across the compared experimental groups. For normalization, the total integrated density of the target signal in each FOV was divided by the respective cell count within that FOV, yielding the average fluorescence signal per cell. The relative fluorescence intensity was then calculated by normalizing these per-cell values to the average value of the uninfected or untreated control group, which was arbitrarily set to 1.0. For each condition, a minimum of 2–3 randomly selected FOVs from at least three independent biological replicates were analyzed.

MitoTracker deep red dyes for mitochondria labeling

Mitochondria were stained using a commercial kit (Beyotime, C1032). Briefly, fully differentiated BMDMs were seeded at 5 × 10⁵ cells per well in 12-well plates and infected with H37Rv at an MOI of 5 (2.5 × 10⁶ bacilli per well) in the presence of the mitochondrial ROS scavenger Mito-TEMPO (MCE, HY-112879) or GW9662 (10 μM; MCE, HY-16578). At 24 h post-infection, cells were incubated with MitoTracker Deep Red diluted to a final concentration of 50 nM at 37 °C for 30 min (n = 3). The dye-containing medium was then removed, cells were washed and replenished with pre-warmed fresh culture medium, and fluorescence images were acquired using a fluorescence microscope. Each experiment used 3 biological replicates, with representative samples illustrating the results.

Lipid droplet detection

Quantification of intracellular LDs was performed using the BODIPY 493/503-based LDs Green Fluorescence Assay Kit (Beyotime Biotechnology, C2053S) (n = 3). Cellular staining was conducted according to the manufacturer protocols with a fluorophore-conjugated probe, enabling specific visualization of neutral LDs through 488 nm excitation. The experiment was performed three times and the results presented using representative samples.

Mass spectrometry analysis

The mass spectrometry sequencing in this study was performed by Bioprofile Technology Company Ltd., China. Protein digestion and LC-MS/MS analysis were performed for large-scale protein identification (Shotgun). Total proteins extracted from M.tb-infected cells were digested using trypsin after reduction and alkylation. The resulting peptides were desalted, vacuum-dried, and reconstituted in 0.1% formic acid for LC-MS/MS analysis (n = 4). LC-MS/MS analysis was conducted using an Easy nLC1200 chromatography system (Thermo Scientific) coupled to a Q-Exactive HF mass spectrometer (Thermo Scientific). Peptides were separated on a C18 column (75 µm × 150 mm, 3 µm, Dr. Maisch GmbH) with a 120-min gradient of increasing acetonitrile concentration. DDA mass spectrometry was performed in positive ion mode with a scan range of 350–1800 m/z. MS1 resolution was 120,000 @ m/z 200, and the top 20 precursor ions were selected for MS2 analysis (resolution: 15,000 @ m/z 200; activation type: HCD). Data were processed using MaxQuant 2.0.1.0 and searched against the UniProt Mus musculus and M.tb protein databases (ATCC 25177/H37Ra)⁹⁹.

Structure modeling

The three-dimensional structure of full-length mouse PPARγ (UniProt ID: P37238) was predicted by the I-TASSER server¹⁰⁰, which is a hierarchical approach to protein structure prediction. It first identifies structural templates from the PDB by the multiple-threading approach LOMETS (Local Meta-Threading Server, version 3), with full-length atomic models constructed by iterative template-based fragment assembly simulations. All models are ranked by C-score, which is a confidence score for estimating the quality of predicted models by I-TASSER. It is calculated based on the significance of threading template alignments and the convergence parameters of the structure assembly simulations. C-score is typically in the range of −5 to 2, where a higher C-score signifies a model with high confidence and vice versa.

Molecular docking

HDOCK server¹⁰¹ was used for docking simulation between mouse PPARγ and IDH. The predicted structure of full-length mouse PPARγ was used in the docking simulations. The structure of IDH predicted by AlphaFold¹⁰² was downloaded from AlphaFold protein structure database (https://alphafold.ebi.ac.uk/) under the UniProt accession A5U813. During the docking process, mouse PPARγ was set as the receptor while IDH was set as the ligand. The HDOCK server automatically predicts their interaction through a hybrid algorithm of template-based and template-free docking. By analyzing the binding conformations between mouse PPARγ and IDH protein, we chose one conformation that was considered the most probable complex conformation of mouse PPARγ and IDH protein. The interactions between mouse PPARγ and IDH protein were analyzed by the Complex Interface Analysis module in WeMol (Wecomput Technology Co., Ltd., China) and illustrated by PyMOL (Schrödinger, LLC, Version 2.5.0).

Proteasome 20S activity assay

Proteolytic 20S activity was quantified using the Amplite 20S Proteasome Assay System (AAT Bioquest). Substrate cleavage reactions were carried out under strictly controlled conditions (37 °C, 60 min) in assay-specific buffer, followed by fluorometric detection employing 488 nm excitation with synchronous emission capture (n = 12).

Flow cytometry

Approximately 1 × 10⁵ cells from mouse bronchoalveolar lavage fluid were resuspended in 100 µL PBS and stained with the following antibodies: 1 µL Anti-Mouse CD3 FITC, 1.5 µL Anti-Mouse CD45 PE/Cy7, and 5 µL Anti-Mouse F4/80 PE (Biolegend, US) (n = 3). After incubation (4 °C, 30 min), cells were centrifuged (500 × g, 10 min), resuspended in 500 µL PBS, filtered through a 300-mesh strainer, and analyzed using a flow cytometer. Single-cell suspensions were prepared from digested lung tissue. Cells were stained with a panel of fluorophore-conjugated antibodies, targeting surface markers including CD45 APC/CY7, CD3 FITC, and F4/80 PE, following incubation with a live/dead viability dye (BV510). Apoptosis in BMDMs was analyzed by flow cytometry using Apotracker Green (BioLegend, 427402) and a live/dead viability dye (BV510). Data were acquired on a BD FACSAria II flow cytometer and analyzed with FlowJo software (version 10.8).

Hematoxylin and eosin (H&E) staining

Lung tissue samples were embedded in paraffin and sectioned into 5 μm thick transverse sections using a microtome (n = 3). Sections were deparaffinized in xylene (2 × 5–10 min) and rehydrated through a graded ethanol series (100% for 5 min, 90% for 2 min, 80% for 2 min, and 70% for 2 min), followed by rinsing in distilled water (2 × 2 min). Sections were stained with hematoxylin for 5 min, with staining time adjusted based on intensity, and rinsed in double-distilled water for 5 min. For immediate observation, sections were washed in 70% ethanol (2 × 10 s). Sections were dehydrated through a graded ethanol series (70, 80, 90, and 100% for 10 s each), cleared in xylene (2 × 5 min), and mounted using neutral resin or another suitable mounting medium. Nuclei appeared blue, and cytoplasm appeared pink or red under the microscope. Quantitative analysis was performed using ImageJ 2.15.1 software.

Oil Red O staining

Following M.tb infection, the culture medium was removed from the 6-well plate, and wells were washed with 1× PBS (3 times). Cells were fixed with 2 mL of 4% neutral paraformaldehyde solution per well for 30 min. After fixation, the paraformaldehyde solution was removed, and wells were rinsed with 1 × PBS (3 times). For lipid staining, 1 mL of Oil Red O working solution (prepared by mixing Oil Red O stock solution with distilled water in a 3:2 ratio and filtering through neutral filter paper) was added to each well and incubated for 30 min. The Oil Red O solution was removed, and wells were rinsed with 1 × PBS (3 times). Cells were observed under a microscope to assess lipid staining (n = 3). The experiment was performed three times and the results presented using representative samples.

Acid-fast staining

RAW264.7 cells were cultured on tissue culture-treated cell coverslips (n = 3). Following M.tb infection, coverslips were rinsed with PBS (3 times). Cells were stained with Kinyoun carbol fuchsin staining solution for 10 min, thoroughly rinsed with PBS, and decolorized with Kinyoun decolorizing solution until no red color remained. Coverslips were washed with PBS (3 times), counterstained with methylene blue staining solution for 30–60 s, rinsed with PBS, and mounted and observed under an oil immersion microscope. The experiment was performed three times and the results presented using representative samples.

Bacterial load

In vitro experiments

RAW264.7 cells were infected with H37Ra for 24 h and washed with PBS (3 times). Cells were scraped and collected into EP tubes using 1 mL of sterile water. The cell suspension was serially diluted (1:10) five times, and 100 μL from each dilution was plated onto 7H10 agar plates specific for H37Ra. Plates were incubated at 37 °C in a humidified incubator for 25–30 days to allow colony formation, and colonies were counted (n = 3). BMDMs were differentiated from mouse bone marrow cells using DMEM containing 10% FBS and 20 ng/mL M-CSF (PeproTech) over 7 days. Differentiated BMDMs were seeded at 5 × 10⁵ cells per well in 12-well plates, then infected with H37Rv at a multiplicity of infection MOI of 5 (2.5 × 10⁶ bacilli per well) for 4 h in the BSL-3 facility under appropriate biosafety conditions (n = 3). Following infection, cells were washed and further incubated for 72 h. To determine bacterial load, macrophages were lysed with 0.1% Triton X-100, and the lysates were plated on 7H10 agar supplemented with 10% OADC. Bacterial colonies were enumerated after 3–4 weeks of incubation at 37 °C.

In vivo experiments

Six to eight-week-old male C57BL/6 mice were randomly assigned to either the control group or the M.tb infection group, with 5–6 mice per group. Following animal experiments, the mouse lung tissues were aseptically removed and washed with PBS (3 times). Tissues were homogenized in 1 mL of sterile water in EP tubes. The homogenized tissue suspension was serially diluted (1:10) five times, and 100 μL from each dilution was spread onto 7H10 agar plates specific for M.tb. Plates were incubated at 37 °C in a humidified incubator for 25–30 days to allow colony formation, and colonies were counted (n = 5–6).

Transmission electron microscopy (TEM)

Cultured cells were fixed in pre-cooled 2.5% glutaraldehyde for 1 h without rinsing. Using a cell scraper at a 45° angle, cells were quickly scraped off (without trypsin digestion) and transferred into a 15 mL centrifuge tube. Cells were centrifuged at 60 × g to form a pellet (approximately half to one mung bean size). Most of the supernatant was removed, and the cell pellet was gently dispersed before transferring to a 1.5 mL EP tube. Cells were allowed to settle naturally for 1 h in an upright position, and the supernatant was gently removed without losing cells. Fresh pre-cooled 2.5% glutaraldehyde (1 mL) was slowly added along the tube wall, and the sample was stored at 4 °C for fixation. Following standard TEM sample preparation, cells were washed, fixed with osmium tetroxide, washed again, dehydrated, infiltrated, and embedded in Epon812 (n = 3). Semi-thin sections were prepared and oriented, followed by ultra-thin sectioning (70–100 nm) using an LKB-V ultra-microtome. Sections were stained with lead citrate and uranyl acetate, and observed using a JEOL-1200EX transmission electron microscope. Images were recorded with a MORADA-G2 camera.

Immunohistochemistry (IHC)

Mouse lung tissue samples, clinical samples, and tissue microarray samples were fixed with 4% paraformaldehyde and embedded in paraffin. Paraffin sections (5 μm) were deparaffinized, rehydrated, and subjected to antigen retrieval using Tris-EDTA buffer (pH 9.0) in a microwave oven. Blocking and color development were performed using the Thermo Immunohistochemistry Kit according to the manufacturer’s instructions (n = 3). Endogenous peroxidase activity was blocked, followed by the blocking of non-specific protein adsorption. Desired antibodies (1:100) were incubated overnight at 4 °C. Sections were then incubated with Primary Antibody Amplifier Quanto and HRP Polymer Quanto. DAB Quanto Chromogen and DAB Quanto Substrate were used for color development. Samples were counterstained with hematoxylin, differentiated, blued, dehydrated, cleared, and mounted with neutral resin. IHC staining was performed by Wuhan Sevier Biosciences Company.

Reactive oxygen species (ROS) assay

Measurement of macrophage ROS levels was performed according to the instructions of a commercial kit (Dojindo, R252). Differentiated BMDMs were seeded at 5 × 10⁵ cells per well in 12-well plates and cultured at 37 °C with 5% CO₂ overnight (n = 3). The cells were then infected with M.tb H37Rv at an MOI of 5 (2.5 × 10⁶ bacilli per well) and treated with the mitochondrial scavenger Mito-TEMPO (MCE, HY-112879) or 10 μM GW9662 (MCE, HY-16578). After 24 h, the media was removed, and the cells were washed twice with HBSS. A diluted high-sensitivity DCFH-DA working solution was then added, and the cells were incubated for 30 min at 37 °C/5% CO₂. After washing, an ROS-inducing buffer was added, and the cells were washed again with HBSS. ROS levels were then assessed by fluorescence microscopy.

Detection of cell apoptosis by TUNEL assay in vitro

RAW264.7 cells and BMDMs were cultured on coverslips, infected with M.tb, and subjected to the TUNEL assay using the Novus TUNEL FITC Apoptosis Detection Kit. Cells were fixed with 4% paraformaldehyde, permeabilized with 0.2% Triton X-100, and equilibrated with 1× equilibration buffer (n = 3 per group). TdT enzyme was added, and samples were incubated at 37 °C for 60 min in a humidified chamber. Samples were washed, counterstained with DAPI, and mounted with glycerol. Green fluorescence was observed under a fluorescence microscope using a standard fluorescence filter set at 520 ± 20 nm.

Lentivirus transduction and knockdown/overexpression

Lentivirus packaging was performed using plasmids psPAX2 (addgene, 12260), pMD2.G (addgene, 12259), pLKO.1-Puro-shNCOR, pLKO.1-Puro-shSMRT, pLV4ltr-PGK-ZsGreen(2 A)-Puro-CMV-NC, and pLV4ltr-Puro-CMV-coH37Ra (IDH)-ZsGreen. All plasmids were amplified in LB medium containing ampicillin (100 μg/mL) and extracted using the Endo-free Plasmid Mini Kit II according to the manufacturer’s instructions. For lentivirus production, HEK293T cells (7 × 10⁵) were seeded in a 6 cm dish and transfected with a mixture of 1 μg of each plasmid, 750 ng of psPAX2 packaging plasmid, and 250 ng of pMD2.G envelope plasmid using a transfection reagent. After incubation and medium replacement, the medium containing lentivirus particles was collected, centrifuged to remove cell debris, and stored at −80 °C. RAW264.7 and HEK293T cells were transduced with the lentivirus when approximately 50–60% confluent. After incubation and medium replacement, puromycin selection was initiated two days after transduction. Stable transfected cell lines were obtained after one month of continuous passaging.

Protein extraction and quantification from lung tissue

Lung tissue (40 mg) was homogenized in RIPA lysis buffer containing protease and phosphatase inhibitors using a homogenizer (60 Hz, 120 sec) (n = 5–6 per group). The homogenate was incubated on ice for 30 min and centrifuged at 12,000 × g for 1 min. The supernatant was mixed with 1:4 of 5× SDS-PAGE sample buffer, heated at 95 °C for 10 min, and stored at −20 °C until further analysis. Protein concentration was measured using the BCA assay according to the kit instructions. Absorbance at 562 nm was measured using a microplate reader, and a standard curve was fitted to calculate the protein concentration. The experiment was performed three times and the results presented using representative samples.

Co-Immunoprecipitation (Co-IP)

The Flag-PPAR$\mathrm{\gamma }$ and ZsGreen-IDH tagged plasmids were synthesized by our laboratory. HEK293T cells were transiently transfected with mammalian expression plasmids using Lipofectamine 3000, following a 48-h incubation period for protein overexpression. Treated cells were washed twice with cold PBS on ice, and 200 μL of WB/IP buffer was added to each well. Cells were scraped off the plate and incubated on ice for 30 min. The lysate was centrifuged at 12,000 × g for 15 min, and the supernatant was transferred to pre-chilled EP tubes. For each 100 μL of cell lysate, 1 μg of the corresponding antibody was added and incubated at 4 °C with gentle mixing overnight. Protein A/G agarose beads were resuspended to create a uniform suspension. 50 μL of the resuspended beads were added to each 100 μL of the lysate-antibody mixture and incubated at room temperature with gentle mixing for 4 h. The mixture was centrifuged at 12,000 × g at 4 °C for 3 min, and the beads were washed twice with WB/IP buffer. 50 μL of 2 × SDS sample buffer was added to the pellet and heated at 95 °C for 5 min. The supernatant was collected after centrifugation at 12,000 × g at 4 °C for 3 min and used for Western blot analysis (n = 3 per group). The HA-tagged expression constructs pLV-3HA-mNCOR and pLV-3HA-mSMRT were generously provided by Prof. Eckardt Treuter’s laboratory (Karolinska Institutet). Overexpression of NCOR and SMRT was used to confirm the repressive role of the repressor complex on the Gpx4 gene. The experiment was performed three times and the results presented using representative samples.

Establishment of a M.tb respiratory infection animal model

Six to eight-week-old male C57BL/6 mice were randomly assigned to either the control or M.tb infection group, with 6 mice per group. Mice were anesthetized using an inhalation anesthesia system with isoflurane and intranasally administered 1 × 10⁶ CFU of H37Ra. Fourteen days post-infection, mice were euthanized, and lung tissue samples were collected. Tissues were either rapidly frozen in liquid nitrogen or stored at −80 °C for Western blot and qPCR analysis or fixed in 4% paraformaldehyde for pathological analysis, including IF and IHC. The Western blot experiment was performed three times and the results presented using representative samples.

H37Rv-infected mice were maintained in a specific pathogen-free facility at Shenzhen University. Aerosol infection was performed to deliver ~200 CFU of H37Rv per mouse; the bacterial load in lungs was confirmed to be ~200 CFU on day 1 after the first two infections, establishing a standardized protocol for all later experiments. Lung samples were collected at 4 and 8 weeks post-infection for time-course studies (n = 5 per group). All procedures received prior approval from the university’s IACUC.

Analysis of mouse bronchoalveolar lavage fluid (BALF)

BALF was obtained by injecting 2 mL of PBS into the mouse lungs via the trachea and aspirating (n = 6 per group). A portion of the BALF was used for cell counting with Countstar, while another portion was stained with Diff-Quik Stain for observation. For flow cytometry analysis, BALF was centrifuged at 500 × g for 10 min, resuspended in PBS, and analyzed. The remaining BALF was centrifuged at 12,000 × g for 5 min, and the supernatant was collected for protein content determination.

Serology analysis

Serum biochemical parameters, including total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, MDA, and non-esterified fatty acids, were quantified using standardized commercial assay kits (Jiancheng Bioengineering Institute, Nanjing) following enzymatic spectrophotometric methods (n = 4–9 per group). Absorbance measurements were acquired with a 96-well plate reader under manufacturer-specified optical conditions.

Preparation and validation of polyclonal antibodies against Rv3339c (M.tb IDH)

Polyclonal antibodies against the M.tb Rv3339c protein were generated commercially by Shanghai Gene-Optimal Science & Technology Co., Ltd. Briefly, five BALB/c mice were immunized subcutaneously with 100 μg of purified Rv3339c protein emulsified in Freund’s adjuvant. Booster immunizations were administered at 2–3 week intervals. Serum titers against Rv3339c were monitored by indirect ELISA. Mice with satisfactory titers were exsanguinated via ocular bleeding 10 days after the final immunization. The obtained polyclonal antiserum was stored at −80 °C. The purified antibodies were subsequently characterized: the titer against Rv3339c was determined to be approximately 409 K by ELISA, the protein concentration was measured using a BCA assay kit, and specificity was confirmed by Western blot analysis.

Intracellular labile iron and lipid peroxidation staining

FerroOrange staining

After treatment, cells were washed three times with Hank’s Balanced Salt Solution (HBSS). Cells were then incubated with 1 μM FerroOrange (Dojindo, F374) in HBSS at 37 °C under 5% CO₂ for 30 min. Following incubation, the staining solution was removed, cells were washed once with HBSS, and immediately imaged under a confocal laser scanning microscope.

C11 BODIPY 581/591 staining

For the detection of lipid peroxidation, treated cells were washed twice with DMEM. The cells were then incubated with the BDP 581/591 C11 working solution (Dojindo, L267) at 37 °C with 5% CO₂ for 30 min. After incubation, the solution was aspirated, and cells were washed twice with HBSS before imaging in HBSS using a confocal microscope. The experiments were performed three times and the results presented using representative samples.

Statistical analysis

All experiments were performed using biological replicates and repeated at least twice. Statistical analyses were performed using GraphPad Prism 9.0, and data are presented as mean ± standard error of the mean (s.e.m.). Group comparisons were evaluated using Student’s t-test for comparing two groups and one-way ANOVA followed by Tukey’s or Dunnett’s post hoc test for multiple comparisons, as appropriate. All statistical tests were two-tailed, with significance set at P < 0.05. Sample sizes were not predetermined by statistical methods, and all samples or animals included in the analyses were considered without exclusion.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

W.P., W.C. and Z.H. designed this experiment and wrote this manuscript. W.P., X.M.Z., Z.Y.H., J.Y., C.G. and J.W. performed the experiments related to the M.tb-infected model of mice. W.P., C.W., Y.S., Y.Z. and K.L. participated in specimen collection. W.P. and S.S. analyzed the data from sequencing and visualized the results. L.L., P.L., H.C., E.P., R.L., A.Y., N.L. and R.W. assisted in the experiments. M.T., J.Z., X.Z., R.F. and E.T. assisted in the linguistic modification of this manuscript. Z.H., H.W. and X.C. supervised this research.

Peer review

Peer review information

Nature Communications thanks Vinicius Soares, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was supported by the National Key Research and Development Program of China (Grant 2022YFF0710801 to H.W.); the National Natural Science Foundation of China (Grants 82571040 to Z.H., 82370899 to H.W., 82070912 to H.W., 825B2024 to R.W.); the Fundamental Research Funds for the Central Universities (Grants 14380538 to Z.H. and 14380550 to Z.H.); State Key Laboratory of Analytical Chemistry for Life Science (Grants 5431ZZXM2615 and 5431ZZXM2404 to H.W.); Visiting Researcher Fund Program of State Key Laboratory of Metabolism and Regulation in Complex Organisms (Grant KF20250005 to Z.H.); the Natural Science Foundation of Jiangsu Province (Grant BK20251988 to Z.H., BK20251803 to H.W.); the NUS-NJU Research Collaboration Fund (Grants 14915200 to H.W., 2025-NJUNUS-0001 to E.P.); the Fundamental Research Funds for the Central Universities and Nanjing University International Collaboration Initiative (Grant 14380549 to H.W.); the Young Talent Development Promotion Association of Nanjing University (2026 to Z.H.); the Team Building and Start-up Funds of Nanjing University (Grant 14912217 to Z.H.); and the Nanjing University Laboratory Safety Research Project (Grant LSK202402 to Z.H.). This work was also supported by the Swedish Cancer Society (Grants 232891Pj to R.F., 211582Pj to E.T., and 243547Pj to E.T.); the Swedish Research Council (Grants 2023-02311 to R.F. and 2022-00545 to E.T.); the EFSD Novo Nordisk Future Leaders Award (to R.F.); the Novo Nordisk Foundation (Grant NNF23OC0084552 to E.T.); the Karolinska Institute Strategic Research Programme in Diabetes (SRP) Rolf Luft Grants (to R.F.); the Jiangsu Funding Program for Excellent Postdoctoral Talent (Grant 2025ZB871 to W.P.) and the National Medical Research Council of Singapore (Grant MOH-OFIRG24jan-0010 to E.P.).

Data availability

The ATAC-seq, ChIP-seq, GRO-seq, and CUT&Tag data generated in this study have been deposited in the NCBI Gene Expression Omnibus (GEO) under the accession number GSE275146. The relevant CUT&Tag and RNA-seq data associated with this study are available under accession numbers GSE235408 (H3K27ac CUT&Tag) and GSE162729 (THP-1 RNA-seq). The other ChIP-seq results were from GSM1555714, GSM4848500, and GSM5701262 datasets. The global run-on sequencing (GRO-seq) results are from our previous datasets (GSE130383) [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE130383], including samples GSM4848613, GSM4848614, GSM4848615, and GSM4848616. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Consent for publication

All authors have read and approved the publication of this manuscript.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-74032-w.