Abstract Macrophage-to-foam cell transition is an integral part of atherosclerotic plaque progression. Particularly, oxidized low-density lipoprotein (oxLDL) is a driving factor in foam cell formation, altering macrophage function and metabolism. The aim of our research was to understand the impact of oxLDL-induced mitochondrial reactive oxygen species on macrophage-to-foam cell differentiation. We demonstrate that macrophage oxLDL-derived superoxide modulates mitochondrial metabolic reprogramming, facilitating foam cell formation. Mechanistically, mitochondrial superoxide drives signal transducers and activators of transcription 5 (STAT5) activation, leading to reduced tricarboxylic acid cycle activity. In parallel, mitochondrial superoxide enhances chromatin accessibility at STAT5 target genes, establishing a distinct STAT5 signaling signature in foam cells ex vivo and in human and mouse plaques in vivo. Inhibition of STAT5 during atherosclerosis progression prevents the differentiation of macrophages to mature Trem2^hiGpnmb^hi foam cells. Collectively, our data describe an oxLDL-induced, mitochondrial superoxide–dependent STAT5 activation that leads to a self-amplifying feedback loop of reciprocal mitochondrial superoxide production and STAT5 activation, ultimately driving macrophage-to-foam cell transition. __________________________________________________________________ Mitochondrial ROS activate STAT5 to drive foam cell formation, inhibiting STAT5 alters macrophage metabolism and stress response. INTRODUCTION Atherosclerosis is a major contributor to cardiovascular disease development, which remains the leading cause of death worldwide. Initiation and progression of atherosclerosis are closely associated with hyperlipidemia, oxidative stress, and persistent low-grade inflammation ([68]1, [69]2). Among the diverse contributors to atherosclerotic disease pathology, macrophages play a pivotal role in plaque progression and stability through their polarization states and metabolic reprogramming, orchestrating plaque dynamics ([70]3, [71]4). Macrophage-to-foam cell differentiation, driven by persistent lipid uptake, leads to plaque growth but also to necrotic core expansion through necrosis and apoptosis, contributing to plaque instability ([72]5). Foam cell transition is primarily induced by oxidized low-density lipoprotein (oxLDL) that is taken up by macrophages via scavenger receptors including LOX-1 and CD36 ([73]6, [74]7). During macrophage-to-foam cell transition, oxLDL induces a metabolic shift from oxidative phosphorylation (OXPHOS) to superoxide production ([75]8). Reactive oxygen species (ROS) production in general is a major defense mechanism in innate immune cells to neutralize invading pathogens ([76]9). However, ROS can also function as signaling molecules, including a compartment-specific effect of ROS ([77]10, [78]11). Mitochondria are a primary source of ROS in cells, producing them mainly via the electron transport chain (ETC) through OXPHOS ([79]12). Mitochondrial superoxide (O[2]^•−), often simplified as mitochondrial ROS (mitoROS), plays a key role in macrophage biology, particularly in antimicrobial defense and regulation of immune responses ([80]12, [81]13). However, excessive mitoROS production leads to oxidative stress, accelerating atherosclerosis progression but also contributing to myocardial infarction and coronary heart disease ([82]14, [83]15). In atherosclerosis, mitoROS facilitate macrophage-to-foam cell differentiation by driving lipid particle oxidation and enhancing their uptake, thereby contributing to plaque growth ([84]16, [85]17). In our study, we identified that oxLDL uptake reinforces a positive feedback loop that amplifies mitoROS production. MitoROS reshape the macrophage chromatin landscape, specifically enhancing signal transducers and activators of transcription 5 (STAT5) signaling, which drives foam cell transition. In parallel, STAT5 activation promotes further mitoROS production, creating a self-sustaining loop that disrupts OXPHOS by inhibiting the pyruvate dehydrogenase complex (PDC). Inhibiting STAT5 selectively reduces mitoROS, suppresses foam cell formation, and limits disease progression, particularly by targeting the Trem2^hiGpnmb^hi macrophage subset. RESULTS Mitochondrial oxidative stress is predominant in foam cells oxLDL plays a crucial role in the transition of macrophages into foam cells. Within our experimental settings, human monocyte-derived macrophages (hMDMs) readily took up oxLDL particles during a 24-hour stimulation period ([86]Fig. 1A). oxLDL challenge further led to a positive feedback loop by up-regulating both the scavenger receptor CD36 ([87]Fig. 1B) and CD146, an important contributor of foam cell formation ([88]Fig. 1C) ([89]18, [90]19). In addition, oxLDL loading of macrophages significantly increased mitochondrial O[2]^•− production ([91]Fig. 1D). To confirm our experimental results in vivo, we fed apolipoprotein E–deficient (ApoE^−/−) mice with high-fat diet (HFD) for a total of 16 weeks (fig. S1A). To measure mitochondrial O[2]^•− in vivo, we stained aortic, splenic, and bone marrow macrophages using MitoSOX dye. Aortic macrophages derived from atherosclerotic plaque–rich aortas exhibited significantly higher mitochondrial O[2]^•− levels compared to their splenic and bone marrow counterparts ([92]Fig. 1E). Furthermore, circulating blood monocytes from 16-week HFD-fed mice displayed elevated mitochondrial oxidative stress compared to regular chow-fed controls (fig. S1B). To better understand how mitoROS burden develops during atherosclerosis progression, we analyzed mouse atherosclerotic plaques from ApoE^−/− mice fed with an HFD for 4, 10, and 16 weeks using the ROS-derived DNA lesion 8-oxoguanine (8-oxog) as a marker ([93]20). 8-Oxog accumulates in both nuclear and mitochondrial DNA ([94]21–[95]24). To differentiate mitochondrion-specific 8-oxog from cytoplasmic 8-oxog, we measured its colocalization with the mitochondrial outer membrane marker TOM22. A total of 95% of the total cytoplasmic 8-oxog colocalized with mitochondrial TOM22, suggesting that cytoplasmic 8-oxog reflects mitochondrial DNA oxidative damage (fig. S1C). Overall distribution of 8-oxog signal was 60% localized within mitochondria and 40% in the nucleus (fig. S1D). Within atherosclerotic lesions, mitochondrial 8-oxog–positive (8-oxog^+) macrophages progressively increased over time ([96]Fig. 1F), correlating with a gradual increase in foam cell accumulation (fig. S1E). A total of 30% of all mouse macrophages (CD68^+) was positive for mitochondrial 8-oxog staining (fig. S1F). A total of 70% of CD68^+CD80^hi macrophages (fig. S1G) and 20% of CD68^+ CD206^hi macrophages showed mitochondrial oxidative stress (fig. S1H). However, CD146^+ foam cells exhibited the highest mitochondrial oxidative DNA damage burden within atherosclerotic lesions (fig. S1I). Within human atherosclerotic lesions, we observed a similar trend. A total of 30% of all macrophages (CD68^+) in the lesions exhibited mitochondrial oxidative stress ([97]Fig. 1G). This increased to almost 70% in CD68^+CD80^hi macrophages ([98]Fig. 1H) but was lower in CD68^+ CD206^hi macrophages (50%) ([99]Fig. 1I). Similar to mouse atherosclerotic plaques, high ROS burden was observed in CD146^+ foam cells, with more than 70% positivity for 8-oxog ([100]Fig. 1J). Fig. 1. Atherosclerotic macrophages display elevated ROS-induced mitochondrial damage that gradually increases with atherosclerosis progression. [101]Fig. 1. [102]Open in a new tab (A) oxLDL uptake in hMDMs treated with diluted (Dil)-oxLDL for 24 hours, measured by flow cytometry. Data are the MFI of PE-Texas Red signals. (B to D) CD36 (B), CD146 (C), and mitochondrial O[2]^•− (MitoSOX red) (D) staining in hMDMs treated with oxLDL (24 hours), determined by flow cytometry. (D) shows a representative flow cytometry histogram. n ≥ 4. (E) MitoSOX red staining of CD45^+ Ly6g^− Cd11b^+ F4/80^+ macrophages in the aorta, spleen, and bone marrow isolated from ApoE^−/− mice fed with an HFD for 16 weeks. w, weeks. The right panels depict the gating strategy used for the characterization of aortic macrophages. Each dot, one mouse. n = 6. (F) Quantification of mitochondrial 8-oxog staining in mouse aortic plaques after 4, 10, and 16 weeks of HFD and gated from CD68^+ signals (total macrophages). Each dot, one mouse. n = 5 (4-week HFD), n = 4 (10-week HFD), n = 9 (16-week HFD). (G to J) Quantification of mitochondrial 8-oxog staining in human carotid plaques gated from CD68^+ (total macrophages) (G), CD68^+CD80^hi macrophages (H), CD68^+CD206^hi macrophages (I), and CD146^+ foam cells (J). The right panel shows a representative image and a zoomed-in inset. Scale bars, 50 μm; insets, 10 μm. n = 17. All data represent the means ± SD of independent biological replicates. [(A) to (D)] *P < 0.05, **P < 0.01, ****P < 0.0001, unpaired Student’s t test; [(E) and (F)] *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, one-way ANOVA with Tukey’s multiple comparison test; [(G) to (J)] **P < 0.01, paired t test; ns, not significant. These observations suggest that mitochondrial oxidative stress progressively intensifies as atherosclerosis advances, correlating with an increment in foam cell accumulation. The heightened substantial mitochondrial oxidative DNA damage burden in CD146^+ foam cells, compared to the rest of the macrophage subsets, raises the question of whether mitochondrial oxidative DNA damage is merely a byproduct of lipid-induced stress or actively contributes to the shaping of the macrophage phenotype. MitoROS reshape the macrophage chromatin landscape To investigate the role of isolated mitochondrial oxidative stress ex vivo, we treated hMDMs with the redox cycling agent 2,3-dimethoxy-1,4-naphthoquinone (DMNQ) for 7 days during monocyte-to-macrophage differentiation to selectively induce mitoROS production (fig. S2A) ([103]25). We used low-dose DMNQ (0.2 μM) to mimic the chronic oxidative environment of atherosclerotic plaques while avoiding the pleiotropic effects of oxLDL, such as lipid accumulation and receptor-mediated signaling. The selected DMNQ concentration enabled sustained mitochondrial O[2]^•− induction over 7 days ([104]Fig. 2A) without altering the total cellular ROS, as determined by CellROX dye ([105]Fig. 2B), or compromising cell viability (fig. S2B). DMNQ-induced functional changes in macrophages included a reduction in filopodium numbers and a subsequent decrease in migratory capacity ([106]Fig. 2, C and D). Furthermore, DMNQ increased macrophage phagocytic activity ([107]Fig. 2E). MitoROS generation did not affect macrophage polarization, as the polarization markers CD80 and tissue factor (CD142) and secreted interleukin-6 (IL-6) and IL-10 levels were not changed (fig. S2, C and D). Of note, reduced migratory capacity and increased phagocytic activity are characteristic functional changes that occur during foam cell transition within atherosclerotic plaques ([108]26–[109]28). Fig. 2. DMNQ-induced mitochondrial stress enhances chromatin accessibility at STAT5 target genes. [110]Fig. 2. [111]Open in a new tab (A and B) Quantification of mitochondrial (A) and cellular ROS (B) in naïve (M0) and DMNQ-differentiated hMDMs, assessed by flow cytometry. (C) Filopodium quantification via Phalloidin Red staining in hMDMs treated as in (A). The right panel shows a representative image with arrows highlighting filopodia. Scale bar and insert, 10 μm. (D) Migration assay in hMDMs differentiated as in (A). Data represent the percentage of migrated cells normalized to nonmigrated cells, expressed as a fold of M0 macrophages. (E) Phagocytosis assay using pHrodo Zymosan particles in hMDMs differentiated as in (A), shown as the MFI. All data represent the means ± SD of independent biological replicates. n ≥ 4. *P < 0.05, **P < 0.01, unpaired Student’s t test. (F) MA plot of differentially accessible peaks from bulk ATAC-seq comparing normalized counts of naïve (M0) hMDMs and DMNQ-differentiated hMDMs. Purple points indicate significant peaks [−1 < log[2]FC (fold change) > 1; FDR < 0.05]. (G) Summary profile plots and heatmap of chromatin accessibility generated by using normalized read coverages around TSSs (−2 to +2 kb) from merged replicates (n = 4). Colored tracks represent genes with increased accessibility after DMNQ treatment. kbp, kilo–base pairs; bp, base pairs. (H) Dot plot of enriched HALLMARK_PATHWAY terms from genes in (G), ranked by P value. Significant terms (P ≤ 0.05) are highlighted. (I) Summary profile plots and heatmap of chromatin accessibility around the TSS (−2 to +2 kb) from snATAC-seq of human atherosclerotic lesions, comparing lipid-associated macrophages (pink) and resident macrophages (light blue). Colored tracks represent STAT5 target genes with increased accessibility after DMNQ treatment. (J) Heatmap of top DMNQ–positively regulated STAT5 target genes from scRNA-seq of human atherosclerotic lesions, comparing lipid-associated macrophages (pink) and resident macrophages (light blue). Data are tag counts per million. To examine the overall impact of mitoROS on macrophage biology, we evaluated possible changes in human macrophage chromatin accessibility after persistent DMNQ challenge using bulk assay for transposase-accessible chromatin with high-throughput sequencing (ATAC-seq). We found that persistent DMNQ-induced mitochondrial stress resulted in a predominantly positive impact on the macrophage chromatin landscape rather than negative alterations ([112]Fig. 2F). A multitude of gene chromatin sites was opened upon DMNQ stimulation compared to untreated M0 macrophages ([113]Fig. 2G and data S1). To verify that this chromatin alteration is specific to mitochondrial O[2]^•− and not a reaction to general ROS stress, we treated cells with hydrogen peroxide (H[2]O[2]) in parallel to DMNQ. Gene loci affected by DMNQ treatment were not altered by general ROS challenge (fig. S2E). Gene set enrichment analysis (GSEA) of DMNQ–positively regulated gene loci revealed a strong enrichment for IL-2-STAT5 signaling ([114]Fig. 2H and data S1). Notably, lipid-associated gene loci (Lpl, Cd36, and Lgals3) displayed increased chromatin opening upon DMNQ treatment compared to M0 control (fig. S2F). Among the few gene sites negatively affected by DMNQ stimulation, we identified genes enriched for the unfolded protein response and IL-6-STAT3 signaling (fig. S2, G and H). We reanalyzed a previously published single-cell RNA sequencing (scRNA-seq) dataset from human atherosclerotic arteries ([115]29) and a single-nucleus ATAC-seq (snATAC-seq) dataset ([116]30) from human carotid lesions to compare our findings with in vivo human conditions. snATAC-seq analysis showed increased chromatin accessibility of DMNQ–positively regulated STAT5 target genes in the lipid-associated macrophage cluster compared to resident-like macrophages in human atherosclerotic lesions, mirroring our ex vivo results ([117]Fig. 2I). The robust STAT5 target gene signature was also reflected in the scRNA-seq dataset, showing strong up-regulation of these target genes in the lipid-associated macrophage cluster ([118]Fig. 2J). To specify common candidate pathways between ex vivo and in vivo macrophages, we performed pathway enrichment analysis for hallmark gene sets using the Molecular Signatures Database (MSigDB) in resident and lipid-associated scRNA-seq macrophage subsets. Both subsets up-regulated tumor necrosis factor signaling via nuclear factor κB and hypoxia signaling, but only lipid-associated macrophages were enriched for the GSEA hallmark term “IL-2-STAT5 signaling” (fig. S2, I and J). Our observations therefore suggest that mitoROS-induced chromatin alterations, coupled with enhanced STAT5 accessibility and signaling, are hallmarks of foam cells within atherosclerotic lesions. MitoROS activate STAT5 through a self-amplifying feedback loop So far, our data indicate that mitoROS production enhances chromatin accessibility at STAT5 target genes. To determine whether mitoROS also directly alters STAT5 activation and signaling, we measured the tyrosine-phosphorylated STAT5 (pYSTAT5) levels. The antibody used detects the phosphorylation of STAT5a at Tyr^694 and STAT5b at Tyr^699, enabling the precise monitoring of overall STAT5 activation. DMNQ treatment significantly increased pYSTAT5 in macrophages ([119]Fig. 3A). Previous reports already suggested lipopolysaccharide (LPS) and interferon-γ (IFN-γ) as potent inducers of pYSTAT5 in monocytes, which we confirmed in macrophages ([120]31, [121]32). The levels of pYSTAT5 induction were similar with both treatment regimens. A combination of the two stimuli leads to a synergistic increase, suggesting alternate activation routes for pYSTAT5 in inflammation and during mitoROS-induced stress ([122]Fig. 3A). Similar to direct mitoROS induction by DMNQ, also oxLDL stimulation of hMDMs resulted in increased STAT5 activation. Inhibition of mitochondrial O[2]^•− using the mitoROS scavenger MitoTempo (mTEMPO) during oxLDL stimulation abrogated the oxLDL-induced activation of STAT5, supporting the direct role of mitoROS in STAT5 activation ([123]Fig. 3B). Conversely, inhibiting STAT5 activation using a selective STAT5 inhibitor [STAT5-in-1 (S5i)] abrogated oxLDL-dependent up-regulation of mitochondrial O[2]^•− production but had no effect on general cellular ROS ([124]Fig. 3C and fig. S3A). To more accurately assess mitochondrial O[2]^•− production, we normalized its levels to mitochondrial mass. In addition, we included mitoquinone mesylate (MitoQ)—a mitochondria-targeted antioxidant capable of penetrating the mitochondrial membrane and scavenging ROS ([125]33). MitoQ treatment reduced mitochondrial O[2]^•− to a degree comparable with mTEMPO (fig. S3B). To validate and complement these findings, we used MitoNeoD, a more specific and stable probe for mitochondrial O[2]^•−, which allows the quantification of its oxidized product MitoNeoOH and does not intercalate with DNA. This produces a fluorescent signal that is more reflective of superoxide than MitoSOX, whose fluorescence arises from both the specific 2-hydroxyethidium (2-OH-Mito-E^+) and the nonspecific oxidation product ethidium (Mito-E^+) ([126]34–[127]36), making it difficult to distinguish between the two species. When normalized to mitochondrial mass, MitoNeoOH fluorescence, measured by live-cell confocal microscopy, was markedly increased in oxLDL-treated hMDMs, and this increase was abrogated by STAT5 inhibition, mTEMPO, and MitoQ pretreatment, respectively (fig. S3C). We next measured the mitochondrial membrane potential (ΔΨ[m]) via uptake of tetramethylrhodamine methyl ester (TMRM) probe in mitochondria ([128]37), normalized to mitochondrial mass. oxLDL induced a significant decrease in ΔΨ[m], which was not reversed by either S5i, mTEMPO, or MitoQ ([129]Fig. 3D and fig. S3D). Similarly, mitochondrial mass itself was elevated in oxLDL-treated macrophages and remained unchanged following treatment with either inhibitor (fig. S3E). This may suggest an accumulation of depolarized, potentially dysfunctional mitochondria, which appears not to be reversed by STAT5 inhibition or ROS scavenging. This observation is in line with a previous report indicating that oxLDL can influence mitochondrial fission dynamics ([130]8). To connect the increased mitoROS with foam cell formation, we measured the protein levels of the macrophage scavenger receptor CD36, which facilitates oxLDL uptake and drives proatherogenic metabolic reprogramming ([131]8). oxLDL-induced CD36 expression was significantly reduced by both S5i and mTEMPO treatments ([132]Fig. 3E), demonstrating a role of STAT5 signaling in macrophage-to-foam cell transition. The direct requirement of STAT5 activation for foam cell formation is further supported by the observed reduction in oxLDL-induced lipid accumulation upon STAT5 inhibition, as evidenced by Oil Red O staining ([133]Fig. 3F). Together, our findings suggest a positive feedback model in which oxLDL-induced mitochondrial O[2]^•− activates STAT5, which in turn amplifies mitochondrial O[2]^•− production and promotes macrophage-to-foam cell transition through increased CD36 expression and oxLDL uptake. Fig. 3. Mitochondrial superoxide drives pYSTAT5-dependent foam cell formation. [134]Fig. 3. [135]Open in a new tab (A) pYSTAT5 levels in naïve (M0) and LPS + IFN-γ–polarized hMDMs with or without DMNQ, assessed by flow cytometry. n = 7. (B) pYSTAT5 quantification in oxLDL-treated hMDMs (24 hours) ± S5i or mTEMPO pretreatment (both 1 hour). n = 4. (C) Mitochondrial superoxide (O[2]^•−) production in oxLDL-treated hMDMs (24 hours) ± S5i, determined by flow cytometry. n = 5. (D) Mitochondrial membrane potential (ΔΨ[m]) assessment by flow cytometry in oxLDL-treated hMDMs (24 hours) ± S5i, mTEMPO (both 1 hour), or MitoQ (2 hours) pretreatment. Cells stained with TMRM and MTG and data shown as a ratio of TMRM^+ cells to MTG^+ cells (%). n = 5. (E) CD36 expression and representative flow cytometry histogram in hMDMs treated as in (B). n = 4. (F) Oil Red O staining in hMDMs treated as in (C). Lipids in red; nuclei in blue. The right panel shows a representative image and a zoomed-in inset. Arrows indicate lipid-positive cells. Scale bar, 50 μm; inset, 20 μm. n = 5. (G) Migration assay in oxLDL-treated mouse BMDMs from Stat5-deficient mice [Vav1-Cre/+Stat5ab^fl/fl (Stat5^−/−) mice] and their wild-type counterpart (Stat5^fl/fl) ± mTEMPO or both. Data represent the migrated area, normalized to nonmigrated cells (%). n = 5. (H) Phagocytosis of pHrodo-labeled Escherichia coli particles in BMDMs treated as in (F), shown as % positive cells normalized to unstimulated controls, with representative histograms. n = 4. All data represent the means ± SD of independent biological replicates. (A) *P < 0.05, paired t test; [(B) to (F)] *P < 0.05, **P < 0.01, one-way ANOVA with Tukey’s multiple comparison test; [(G) and (H)] *P < 0.05, **P < 0.01, ***P < 0.001, two-way ANOVA with Tukey’s test (intragroup) and Sidak’s test (intergroup). To assess the functional consequences of STAT5 loss in this context, we isolated bone marrow–derived macrophages (BMDMs) from Stat5-deficient mice [Vav1-Cre/+Stat5ab^fl/fl (Stat5^−/−) mice] ([136]38). Stat5 deficiency prevented the oxLDL-induced impairment of migratory capacity, a defect that was rescued by mTEMPO pretreatment in Stat5^fl/fl (wild-type) macrophages. However, mTEMPO treatment had no additional effect on the migratory capacity of Stat5^−/− macrophages during oxLDL exposure ([137]Fig. 3G). In addition, similar to DMNQ challenge, oxLDL stimulation increased phagocytosis in BMDM. This increase was suppressed in Stat5^−/− macrophages or in macrophages treated with mTEMPO ([138]Fig. 3H). Consistent with our findings in human macrophages, STAT5 loss did not restore oxLDL-induced mitochondrial membrane depolarization (fig. S3F). Furthermore, oxLDL-challenged Stat5^−/− BMDMs displayed reduced lipid staining and lower CD36 expression levels compared to their wild-type counterparts (fig. S3, G and H). Together, these findings underscore the role of mitoROS-induced STAT5 activation in driving macrophage dysfunction and foam cell transition in both mice and humans. Mitochondrial STAT5 correlates with high mitochondrial stress burden Activated STAT5 has been reported to locate to both the nucleus and mitochondria ([139]39–[140]41). In macrophages, oxLDL challenge led to a significant increase in pYSTAT5 in both compartments, with the highest levels of pYSTAT5 observed in mitochondria ([141]Fig. 4A). To confirm the localization pattern of STAT5 in vivo, we stained human atherosclerotic plaques and found that most pYSTAT5-positive cells were predominantly associated with high mitochondrial DNA oxidative stress, as determined with 8-oxog staining ([142]Fig. 4B). Among cells coexpressing pYSTAT5 and 8-oxog, 60% showed mitochondrial localization of the signals, while 40% localized to the nucleus, reinforcing the preferential mitochondrial presence of STAT5 under mitochondrial oxidative stress conditions ([143]Fig. 4C). Foam cells, identified as CD146^+ or TREM2^+, exhibited overall the highest pYSTAT5 expression among all atherosclerotic plaque cells ([144]Fig. 4, D and E). Similar findings were observed in mouse plaques from ApoE^−/− fed with an HFD for 16 weeks ([145]Fig. 4, F and G). We further confirmed elevated pYSTAT5 levels in CD146^+ foam cells compared to nonfoamy cells in mouse aortic tissue using flow cytometry ([146]Fig. 4H). To validate the 8-oxog staining in vitro, we quantified 8-hydroxy-2′-deoxyguanosine (8-OHdG) levels in mitochondrial and nuclear DNA using a mass spectrometry–based approach ([147]35) in BMDM from Stat5^−/− mice and their wild-type counterparts. oxLDL treatment substantially increased the proportion of 8-OHdG in mitochondrial DNA, and this effect was attenuated in Stat5^−/− cells, supporting a role for STAT5 in promoting mitochondrial DNA oxidative lesions (fig. S4A). Fig. 4. MitoROS-induced pYSTAT5 is highly expressed in human and mouse foam cells. [148]Fig. 4. [149]Open in a new tab (A) Quantification of nuclear and mitochondrial pYSTAT5 MFI in hMDMs treated with oxLDL for 24 hours. Mitochondria identified by VDAC1/Porin^+ signals; nuclei by DAPI^+ signals. Representative image scale bar, 20 μm. n = 5. (B) Quantification and representative image of mitochondrial pYSTAT5 in human carotid plaques, comparing 8-oxog^+ cells to remaining plaque cells. Scale bar, 50 μm. n = 10. (C) Nuclear and mitochondrial pYSTAT5 quantification in 8-oxog^+ cells within human carotid plaques. Arrows highlight pYSTAT5^+ 8-oxog^+ cells in mitochondria (identified by TOM22^+ signals). Scale bar, 50 μm; inset, 10 μm. n = 10. (D and E) Mitochondrial pYSTAT5 quantification in CD146^+ (D) and TREM2^+ (E) cells versus other plaque cells. Arrows highlight pYSTAT5^+ foam cells. Scale bars, 10 μm. n = 19 (D) and n = 15 (E). (F and G) Relative nuclear and mitochondrial pYSTAT5 localization in mouse aortic plaques from ApoE^−/− mice (HFD, 16 weeks), comparing 8-oxog^+ versus 8-oxog^− cells (F) and CD146^+ foam versus CD146^− cells (G). Each dot, one mouse. Scale bars, 20 μm; insets, 10 μm. n = 11 (G) and n = 9 (H). (H) pYSTAT5 levels in CD45^+ Ly6g^− Cd11b^+ F4/80^+ CD146^− nonfoamy macrophages versus CD45^+ Ly6g^− Cd11b^+ F4/80^+ CD146^+ foam cells from ApoE^−/− mouse aortas (HFD, 16 weeks). Each dot, one mouse. n = 5. All data represent the means ± SD of independent biological replicates. (A) *P < 0.05, two-way ANOVA with Tukey’s test (intragroup) and Sidak’s test (intergroup). [(B) to (G)] *P < 0.05, **P < 0.01, ****P < 0.0001, paired t test. (H) ****P < 0.0001, unpaired Student’s t test. Mitochondrial STAT5 contributes to the oxLDL-induced suppression of oxidative metabolism Mitochondria are highly metabolic organelles involved in several cellular processes such as the tricarboxylic acid (TCA) cycle, adenosine 5′-triphosphate (ATP) production through OXPHOS, and fatty acid oxidation (FAO) ([150]42). To investigate the potential role of mitoROS-induced STAT5 relocalization to mitochondria in cellular metabolism, we first measured the levels of lactate and pyruvate as key readouts of glycolytic flux. In response to oxLDL, lactate production increased in both wild-type and Stat5^−/− BMDMs, although levels in Stat5^−/− BMDMs were significantly lower compared to those in wild-type cells ([151]Fig. 5A). Conversely, pyruvate levels decreased following oxLDL challenge in wild-type macrophages. Stat5^−/− cells already exhibited reduced pyruvate levels at the baseline compared to wild-type macrophages, and these levels remained unchanged upon oxLDL treatment ([152]Fig. 5B). Similar results were observed in hMDMs, where oxLDL also increased lactate production, and this effect was reversed by either STAT5 inhibition or mTEMPO treatment (fig. S4B). Notably, the lactate/pyruvate ratio was elevated in both wild-type and Stat5^−/− BMDMs following oxLDL treatment, indicating a shift toward glycolytic metabolism (fig. S4C) ([153]43). Consistent with this observation, STAT5 inhibition did not alter the oxLDL-induced increase in glucose uptake in human macrophages, which was suppressed by mTEMPO pretreatment, suggesting that oxLDL-induced glycolytic reprogramming is largely STAT5-independent (fig. S4D). Extracellular acidification rate (ECAR) analysis demonstrated that oxLDL treatment promotes glycolysis in wild-type BMDMs, and this effect persisted despite Stat5 loss, confirming that STAT5 is dispensable for the glycolytic shift induced by oxLDL ([154]Fig. 5D and fig. S4E). The oxygen consumption rate (OCR) was reduced by oxLDL treatment in wild-type BMDMs, as indicated by decreased basal respiration, ATP production, proton leak, and maximal respiration ([155]Fig. 5, E and F, and fig. S4F). Stat5-deficient BMDMs displayed elevated mitochondrial respiratory parameters at the baseline compared to wild-type controls. Following oxLDL treatment, these parameters decreased in both genotypes to a similar degree; however, Stat5-deficient cells maintained significantly higher basal respiration rates and ATP production compared to wild-type cells under the same conditions. The spare respiratory capacity remained largely unchanged across genotypes and conditions, whereas oxLDL exposure modestly enhanced coupling efficiency in both genotypes, likely due to a reduction in proton leak. Notably, mTEMPO partially restored maximal respiration in wild-type cells but had no additive effect in Stat5^−/− BMDMs and did not significantly influence other respiratory parameters (fig. S4, G and H), suggesting that additional oxLDL-induced mitochondrial alterations are likely mediated by mitoROS-independent mechanisms. Indirect measurements of basal OCR in human macrophages showed that STAT5 inhibition mitigated the oxLDL-induced suppression of respiration (fig. S4I). Although not identical to the murine model, these findings suggest a conserved role for STAT5 in regulating mitochondrial metabolism in oxLDL-challenged macrophages. Fig. 5. STAT5 deficiency improves the mitochondrial function. [156]Fig. 5. [157]Open in a new tab (A and B) Lactate and pyruvate metabolite levels in BMDMs from Stat5-deficient mice [Vav1-Cre/+Stat5ab^fl/fl (Stat5^−/−) mice] and their wild-type counterpart (Stat5^fl/fl), following 24-hour oxLDL stimulation, measured by liquid chromatography–mass spectrometry (LC-MS) and normalized to internal amino acid standards. (C) Schematic overview of examined metabolic pathways. Glucose uptake initiates glycolysis, producing pyruvate, which is either converted into lactate or acetyl-CoA, fueling the TCA cycle, leading to ATP production through OXPHOS and FAO. Created in BioRender. Hohensinner, P. (2025). [158]https://biorender.com/ckfllob. (D) Glycolytic activity in oxLDL-treated BMDMs (24 hours) assessed by the ECAR, measured by Seahorse XF Glycolysis stress test. (E) OCR assessment in BMDMs treated as in (D) by the Seahorse XF mito stress test. (F) Quantification of maximal respiration based on the data in (E). (G) Coimmunoprecipitation of STAT5 with PDC-E2, PDHA1, and PDHB in BMDMs differentiated for 7 days with DMNQ and polarized with LPS + IFN-γ ± DMNQ (24 hours). The Western blot shows immunoprecipitated proteins and 20% input, with vinculin as loading control. n = 3. IP, immunoprecipitation; kDa, kilodalton. (H) PDH enzyme activity in BMDMS treated as in (D), presented as the rate of enzymatic activity. (I and J) Scatter plot of pYSTAT5-PDHB colocalization in human carotid plaques (n = 18) (I) and mouse aortic plaques (n = 10; ApoE^−/− mice, HFD, 16 weeks) (J). The red line shows regression with standard error. Representative images scale bars, 10 μm. Pearson correlation assessed by Student’s t test. r, Pearson’s correlation coefficient. All data represent the means ± SD of independent biological replicates. Each dot, one mouse. [(A), (B), (D), (F), (H)] *P < 0.05, **P < 0.01, ***P < 0.001, two-way ANOVA with Tukey’s test (intragroup) and Sidak’s test (intergroup). n = 3. Mitochondrial STAT5 can exert a noncanonical function in mitochondria by interacting with and disrupting the PDC, leading to the reduced conversion of pyruvate into acetyl coenzyme A (acetyl-CoA) ([159]44, [160]45). The PDC is a multisubunit complex composed of E1-E2-E3 dehydrogenase together with the pyruvate dehydrogenase (PDH) E1 subunits alpha and beta (PDHA1 and PDHB) functioning as the E1 subunit ([161]46). hMDMs stimulated with oxLDL showed increased colocalization of the PDC subunit PDC-E2 and pYSTAT5 within mitochondria, as determined by immunofluorescence. This interaction was inhibited upon treatment with S5i or mTEMPO (fig. S5A). Notably, no colocalization was detected within the nucleus (fig. S5B). To further explore the interaction between STAT5 and the PDC in mouse macrophages, we differentiated BMDMs in the presence of DMNQ for mitoROS induction and polarized them with LPS + IFN-γ stimulation to further enhance STAT5 activation. We then prepared mitochondrial and cytosolic extracts and analyzed STAT5 and PDC localization with Western blot analysis. STAT5 was detected in both the cytoplasm and mitochondria, whereas all examined members of the PDC (PDC-E2, PDHA1, and PDHB) were not detectable within the cytoplasm (fig. S5C). Coimmunoprecipitation of STAT5 from total lysates revealed a clear association of the three PDC subunits with STAT5 in all challenged macrophages compared to untreated ones, confirming that STAT5 activation is necessary for its binding to the PDC ([162]Fig. 5G). This interaction was accompanied by a reduction in PDH enzyme activity in wild-type BMDMs upon oxLDL stimulation, whereas Stat5^−/− cells exhibited increased PDH activity under the same conditions ([163]Fig. 5H). However, despite these differences and the observed partial preservation of mitochondrial respiration in Stat5^−/− macrophages, the oxLDL-induced increase in NADH [reduced form of nicotinamide adenine dinucleotide (oxidized form) (NAD^+)]/NAD^+ ratio persisted even in the absence of STAT5 (fig. S5D). pYSTAT5 colocalized with the PDHB subunit in both human (r = 0.46; P = 5.04 × 10^−32) and mouse (r = 0.533; P = 1.60 × 10^−21) atherosclerotic plaque cells, further supporting an in vivo relevance of the STAT5-PDC interaction ([164]Fig. 5, I and J). Collectively, these findings indicate that STAT5 contributes to the oxLDL-induced suppression of mitochondrial respiration, potentially by interacting with and modulating PDC activity. The apparent increase in PDH activity in Stat5^−/− cells did not align with reduced lactate levels or the further decrease in pyruvate and was not reflected in the redox state, suggesting the involvement of additional regulatory mechanisms in controlling metabolic flux under oxLDL stress, thereby underscoring STAT5’s role as an important, although not exclusive, regulator of metabolic reprogramming in foam cells. STAT5 regulates mitochondrial superoxide production via complex III Mitochondrial O[2]^•− production is primarily attributed to complex I via reverse electron transport (RET). This process, previously reported in LPS-treated pro-inflammatory macrophages, occurs under conditions of high ΔΨ[m] and a highly reduced ubiquinone (CoQ) pool ([165]47, [166]48). To determine whether complex I contributes to oxLDL-induced, STAT5-dependent mitochondrial O[2]^•− production, we assessed the impact of succinate dehydrogenase (complex II) inhibition using dimethyl malonate (DMM). DMM limits electron flow from succinate to the CoQ pool, driving RET at complex I ([167]47). DMM pretreatment reduced the OCR in oxLDL-treated wild-type BMDMs ([168]Fig. 6A), consistent with previous findings in LPS-treated macrophages ([169]49). However, DMM also reduced the OCR in Stat5^−/− macrophages, suggesting that complex I–dependent mitochondrial O[2]^•− production is independent of STAT5. This interpretation is further supported by the absence of increased ΔΨ[m] in oxLDL-treated cells, a condition required to support RET at complex I ([170]Fig. 3D and fig. S3F). Fig. 6. oxLDL-induced STAT5 drives mitochondrial superoxide production via complex III. [171]Fig. 6. [172]Open in a new tab (A) OCR, as readouts for OXPHOS, measured by the Seahorse XF mito stress test in BMDMs isolated from Stat5-deficient mice [Vav1-Cre/+Stat5ab^fl/fl (Stat5^−/−) mice] and their wild-type counterpart (Stat5^fl/fl). Cells were treated with oxLDL for 24 ± 3–hour pretreatment using DMM. (B and C) Mitochondrial superoxide (O[2]^•−) production assessed by flow cytometry using either MitoSOX red staining (B) or MitoNeoD staining, followed by quantification of its oxidized product MitoNeoOH (C). BMDMs were treated with oxLDL for 24 hours ± oligomycin in the last 30 min of oxLDL stimulation or for 30 min in M0 macrophage controls. Right panels display representative flow cytometry histograms. All data represent the means ± SD of independent biological replicates. Each dot, one mouse. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, two-way ANOVA with Tukey’s test (intragroup) and Sidak’s test (intergroup). n = 3. Mitochondrial O[2]^•− is also generated at complex III through the Q cycle ([173]50, [174]51). To investigate a possible role of complex III in oxLDL-induced mitochondrial O[2]^•− production, we treated macrophages with antimycin A, a selective complex III inhibitor that binds the Qi site and blocks electron transfer through cytochrome b, thereby promoting superoxide generation ([175]51). Antimycin A treatment was administered for 30 min to evaluate the effect of acute complex III inhibition. Antimycin A significantly increased mitochondrial O[2]^•− production in untreated (M0) BMDMs and further amplified oxLDL-induced superoxide levels in wild-type macrophages, as measured by both MitoSOX Red and MitoNeoD-derived MitoNeoOH ([176]Fig. 6, B and C). Notably, this effect was attenuated in Stat5^−/− macrophages, indicating that oxLDL promotes complex III–dependent mitoROS generation in a STAT5-dependent manner. STAT5 inhibition targets mitoROS during advanced atherosclerosis To investigate the temporal dynamics of STAT5 activation during atherosclerosis progression in foam cells, we measured pYSTAT5 levels in mouse atherosclerotic plaques from ApoE^−/− mice fed with an HFD for 4, 10, and 16 weeks. Our analysis revealed a progressive increase in pYSTAT5-positive CD146^+ foam cells over time ([177]Fig. 7A). To examine the impact of STAT5 inhibition on foam cell composition and phenotype at different stages of disease progression, we used pharmacological inhibition of STAT5 in ApoE^−/− mice at 4, 10, and 16 weeks time points of HFD feeding. The inhibitor was administered for either the whole duration of the experiment or 4 weeks after HFD initiation to allow for the initial formation of atherosclerotic lesions (fig. S6A). Body weight increased throughout the course of HFD regardless of STAT5 inhibition (fig. S6B). pYSTAT5 levels were increased during atherosclerosis development in aortic cells and reduced by STAT5 inhibition during atherosclerosis progression, except in early disease (4 weeks) ([178]Fig. 7B). S5i administration did not alter plasma lipid composition including cholesterol, triglycerides, LDL, or HDL ([179]Fig. 7C and fig. S6C). To assess broader systemic effects, we measured hepatic steatosis and observed a progressive increase in the percentage of liver area infiltrated by fat droplets from 4 to 16 weeks of HFD feeding, with only a mild reduction at 16 weeks following S5i treatment ([180]Fig. 7D). Blood neutrophil levels remained unchanged across all conditions (fig. S6D). Platelet-monocyte aggregates, a marker of vascular inflammation, were elevated at 4 weeks of HFD compared to chow-fed controls and declined at later stages. A similar trend was observed for platelet-neutrophil aggregates, which gradually increased during disease progression and markedly dropped at 16 weeks (fig. S6, E and F). Overall, STAT5 inhibition significantly reduced atherosclerotic plaque size in the aortic arch (fig. S6, G and H) and decreased the plaque necrotic core area (fig. S6I) regardless of S5i treatment start ([181]52). For foam cells specifically, STAT5 inhibition led to a reduction in TREM2^+ foam cells in aortic tissue at 10 and 16 weeks of HFD in both treatment regimens but not at 4 weeks ([182]Fig. 7E). Notably, this reduction occurred independently of changes in total circulating monocyte or subset distribution (fig. S6, J and K). To evaluate mitochondrial oxidative stress, we assessed 8-oxog lesions in plaque foam cells. A significant reduction of 8-oxog lesions in TREM2^+ foam cells was observed at 16 weeks following S5i administration in both treatment regimens. Notably, the proportion of TREM2^+ 8-oxog^+ cells increased progressively over the course of HFD feeding compared to chow-fed controls ([183]Fig. 7F). Moreover, flow cytometric analysis confirmed reduced mitochondrial O[2]^•− in both CD146^+ and TREM2^+ foam cell populations at 16 weeks following STAT5 inhibition (fig. S6L). Together, these findings suggest that STAT5 involvement in atherosclerosis becomes more prominent in advanced disease, coinciding with a substantial increase in mitochondrial oxidative stress and foam cell accumulation. Fig. 7. STAT5 inhibition limits the accumulation of mitoROS-stressed foam cells in advanced atherosclerosis. [184]Fig. 7. [185]Open in a new tab (A) Quantification and representative images of mitochondrial pYSTAT5 localization in CD146^+ foam cells from mouse aortic plaques of ApoE^−/− mice fed with an HFD for 4, 10, and 16 weeks. Scale bar, 50 μm. n = 5 (4 weeks), n = 4 (10 weeks), and n = 11 (16 weeks). (B) pYSTAT5 staining in total aortic cells from ApoE^−/− mice fed with an HFD for 4, 10, or 16 weeks, with an HFD for a total of 16 weeks with or without persistent S5i administration or starting after 4 weeks, or with DMSO control, measured by flow cytometry. n = 5. (C) Plasma cholesterol and LDL levels in ApoE^−/− mice fed and treated as in (B) compared to chow-fed controls (15 and 25 weeks old). n = 6 (D) H&E staining of liver tissue from mice fed and treated as in (B). Data are given as the area covered by fat droplets normalized to the total liver area. Representative images are shown. Scale bars, 20 μm. n = 6. (E) Quantification of TREM2 protein levels in CD45^+ Ly6g^− Cd11b^+ F4/80^+ TREM2^+ aortic cells from ApoE^−/− mice fed and treated as in (B). n = 5. (F) Quantification of mitochondrial 8-oxog in TREM2^+ foam cells from aortic plaques of ApoE^−/− mice fed and treated as in (C). Representative images shown for 16 weeks and S5i 16-week groups. Scale bars, 10 μm. n = 6. All data represent the means ± SD of independent biological replicates. Each dot, one mouse. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, one-way ANOVA with Tukey’s test. STAT5 inhibition selectively targets the Trem2^hiGpnmb^hi foam cell subset in vivo Our in vivo data indicate a robust effect of STAT5 inhibition on foam cell number and mitochondrial oxidative stress during advanced atherosclerosis. To further explore the role of STAT5 in foam cell formation during atherosclerotic plaque development and identify specific subsets of plaque cells affected by STAT5 inhibition, we performed scRNA-seq from single-cell suspensions of whole aortas of HFD-fed animals over a time course of 4, 10, and 16 weeks. We focused on the hematopoietic CD45^+ cells for the analysis ([186]Fig. 8A) and performed subclustering of the aortic mononuclear phagocytes [MPCs; macrophages, antigen-presenting cells (APCs), and monocytes] ([187]52) to identify distinct macrophage subsets present in the aortas of atherosclerotic mice ([188]Fig. 8B). Stat5a and Stat5b isoforms were widely distributed within the main hematopoietic cluster, with Stat5b being predominant over Stat5a in macrophages (30.1 and 16.7%, respectively) (fig. S7A). We identified 11 subclusters including the atherosclerosis-associated foam cells (Trem2, Slamf9, and Gpnmb), a heterogeneous population of both pro- and anti-inflammatory monocytes and macrophages (Il1b, Chil3, and Arg1), an IFN-inducible cell cluster (Ifit3, Ifi44, and Rsad2), and a cluster of resident macrophages, divided into proliferating (Stmn1, Tuba1b, and Dnmt3c), Cd209^hi (Cd209g, Cd209f, and Mmp9) and Lyve1^+ (Lyve1, Cd226, and Retnla) macrophages. Dendritic cells (DCs) were divided into three major subtypes: classical DCs/plasmacytoid DCs (cDC1/pDC; Clec9a, Cd209a, and H2-Ob), cDC2 (Cd5, Ccl5, and Il1r2), and mature DCs (Tmem123, Fscn1, and Socs2) (fig. S7B) ([189]52, [190]53). Overall, we observed increased MPC accumulation in the aorta from 4 to 16 weeks of HFD (fig. S7C). The proportion of the previously defined Trem2^hiGpnmb^hi and Trem2^hiSlamf9^hi macrophage foam cell subsets ([191]52) increased progressively during disease progression, with STAT5 inhibition not significantly affecting their overall abundance (fig. S7D). The Trem2^hiGpnmb^hi subset is associated with lipid metabolism and oxidative stress responses, exhibiting characteristics typical of foamy macrophages. Conversely, the Trem2^hiSlamf9^hi subset is linked to pro-inflammatory responses and innate immune activation, contributing to plaque progression ([192]52, [193]54). Trem2 was found to be mostly expressed in the foam cell clusters but also in resident macrophages while being absent in DCs. Moreover, Slamf9 expression was widely distributed within the macrophage and foam cell subsets as well as in the cDC2 cluster, whereas Gpnmb was most exclusively expressed in the Trem2^hiGpnmb^hi cluster, delineating the specificity of this marker for foam cell signatures ([194]Fig. 8C). We next focused on the comparison between the two foam cell subclusters to understand the role of STAT5 during atherogenesis. Despite expressing similar levels of Trem2, these two subclusters differed in the expression of several atherosclerosis-specific genes. The gene expression of lipid dynamics–associated genes Cd36, Lgals3, Spp1, and Lpl was higher in the Trem2^hiGpnmb^hi cluster compared to the Trem2^hiSlamf9^hi cluster (fig. S7E). Although their functions are known, the roles of these two clusters across different disease stages and the precise dynamics of their prevalence during atherosclerosis progression are not yet fully elucidated. To better understand their distribution during disease progression, we performed trajectory inference analysis with slingshot pseudotime ([195]55) in the 10- and 16-week datasets. We excluded the 4-week dataset because of the low proportion of Trem2^hiGpnmb^hi cell subset (<200 cells). A minor population of Trem2^hiGpnmb^hi cells (orange) appeared early in the pseudotime trajectory before transitioning into the Trem2^hiSlamf9^hi cluster (green), which dominated the early pseudotime regions. At middle stages, the Trem2^hiSlamf9^hi cluster progressively transitioned back into Trem2^hiGpnmb^hi cells, which became the dominant population in the late pseudotime stage. STAT5 inhibition at 16 weeks prolonged the presence of the Trem2^hiSlamf9^hi cluster by reducing the Trem2^hiGpnmb^hi subset at early pseudotime (interval [0,12.1]) (fig. S7F) and impaired the further development of the Trem2^hiGpnmb^hi cluster at late pseudotime (interval [24.2,36.2]; P = 0.0007 S5i 16 weeks versus 16-week HFD). Conversely, only minor alterations were observed at 16-week ([196]Fig. 8D). We therefore suggest that the Trem2^hiGpnmb^hi foam cells represent an end-stage foam cell phenotype that relies on STAT5 signaling, consistent with their observed increased abundance at advanced disease stages ([197]52). To further investigate this dependency, we analyzed the expression of Stat5 and STAT5 target genes within both foam cell clusters. Both Stat5a and Stat5b isoforms were more expressed in Trem2^hiGpnmb^hi (31.3 and 50.6% for Stat5a and Stat5b, respectively) compared to the Trem2^hiSlamf9^hi subset (18.8 and 32.9% for Stat5a and Stat5b, respectively) ([198]Fig. 8E). Moreover, most of the STAT5 target genes obtained from the DMNQ-stimulated hMDM bulk ATAC-seq and expressed mostly in lipid-associated macrophages from the human scRNA-seq and snATAC-seq dataset ([199]Fig. 2, I and J) ([200]29, [201]30) showed higher average expression in Trem2^hiGpnmb^hi cells ([202]Fig. 8F). Notably, STAT5 inhibition specifically down-regulated the expression of these genes in the Trem2^hiGpnmb^hi subset but not in the Trem2^hiSlamf9^hi cluster ([203]Fig. 8G and fig. S7G). To pinpoint the inhibitory effect of the STAT5 inhibitor, we performed a pseudobulk RNA-seq differential expression analysis of the data ([204]https://github.com/cancerbits/DElegate) followed by GSEA to identify the enriched pathway affected ([205]www.broad.mit.edu/gsea/). Persistent pharmacological inhibition of STAT5 for 16 weeks resulted in the down-regulation of 81 genes and up-regulation of 59 genes within the Trem2^hiGpnmb^hi cluster (fig. S7H and data S2). The down-regulated genes were enriched for IL2-STAT5 signaling, together with inflammatory response and hypoxia-related genes, including Nfkbia and Ccl4 ([206]Fig. 8H and fig. S7I). A similar pattern of pathway enrichment was observed for genes down-regulated by STAT5 inhibition at 10 weeks of HFD, although with a lesser transcriptomic impact, resulting in 56 genes down-regulated and 16 genes up-regulated (fig. S7J and data S2). In contrast, STAT5 inhibition had minimal effects on gene expression in the Trem2^hiSlamf9^hi cluster at both 10 weeks (data S2) and 16 weeks of HFD (fig. S7K), with no significant enrichment of hallmark pathways from MSigDB ([207]www.gsea-msigdb.org/gsea/msigdb). Because of the very low proportion of Trem2^hiGpnmb^hi cells at 4 weeks of HFD (fig. S7D), we were unable to perform pseudobulk analysis at this time point. Fig. 8. STAT5 inhibition delays the transition from early Trem2^hiSlamf9^hi to end-stage Trem2^hiGpnmb^hi foam cells. [208]Fig. 8. [209]Open in a new tab (A) DittoBarPlot showing CD45^+ hematopoietic cells (%) from scRNA-seq of aortic cells in ApoE^−/− mice fed with HFD for 4, 10, and 16 weeks ± S5i treatment (see fig. S6A). Six mice pooled per library. (B) t-SNE visualization of aortic MPCs after subsetting and reclustering macrophage and APC/monocyte clusters. (C) t-SNE FeaturePlots showing Trem2, Gpnmb, and Slamf9 expression in MPCs. (D) Slingshot trajectory analysis of Trem2^hiSlamf9^hi and Trem2^hiGpnmb^hi clusters from 10-weeks and 16-week HFD datasets ± S5i. x axis, slingshot pseudotime. (E) Percentage of cells expressing Stat5a and Stat5b in both foam cell clusters from (D). (F) Heatmap of DMNQ–positively regulated STAT5 target genes (from [210]Fig. 2G) in the same clusters as in (D), displayed as a log[2] fold change. (G) Circular bar plot of 15 STAT5 target genes from (F) expressed in Trem2^hiGpnmb^hi cells from ApoE^−/− mice treated with S5i (continuous or from 4 weeks) versus HFD controls. Right: RidgePlot of three representative genes. (H) Dot plot representation of enriched HALLMARK_PATHWAY terms for S5i–down-regulated genes (16 weeks of HFD versus DMSO) in Trem2^hiGpnmb^hi cells. Dot size, transcript numbers. (I) Dot plot showing the expression of representative genes involved in glycolysis, TCA cycle, FAO, and mitochondrial stress response in Trem2^hiGpnmb^hi cells. Dot size, average expression. The color indicates expressing cells (%). (J) RidgePlot of Slc2a1 (glycolysis), Acads (TCA cycle), Aco2 (FAO), and mitochondrial stress response (Pink1) from (I). (K) Quantification of mitochondrial 8-oxog in Trem2^hiGpnmb^hi cells from aortic plaques of ApoE^−/− mice fed with an HFD for 16 weeks ± S5i (continuous or from 4 weeks). Data are the means ± SD of independent biological replicates. *P < 0.05, two-way ANOVA with Tukey’s test (intragroup). n = 6. Each dot, one mouse. Our ex vivo data demonstrated that oxLDL-activated STAT5 increases mitochondrial O[2]^•− production and shifts macrophages toward glycolysis by reducing OXPHOS ([211]Fig. 5). To determine whether STAT5 inhibition induced metabolic changes in vivo in the Trem2^hiGpnmb^hi cluster, we selected target genes involved in glycolysis, TCA cycle, and FAO together with mitochondrial stress response genes. We then analyzed the impact of STAT5 inhibition on their expression from 10 to 16 weeks of HFD ([212]Fig. 8I). Notably, glycolytic genes (e.g., Slc2a1) remained unaffected by S5i treatment, whereas the expression levels of the TCA cycle (Acads)– and FAO (Aco2)–related genes were increased when STAT5 was inhibited either throughout the entire period of HFD feeding or after established disease. In addition, mitochondrial stress response genes (e.g., Pink1) exhibited a marked increase in expression upon STAT5 inhibition ([213]Fig. 8J). Given the prominent effect of STAT5 inhibition observed in later stages of atherosclerosis, and to reconnect mitochondrial O[2]^•− production through fatty acids with the targeted Trem2^hiGpnmb^hi cluster, we stained TREM2- and GPNMB-positive cells from mouse plaques at 16 weeks of HFD with 8-oxog. Approximately 80% of all Trem2^hiGpnmb^hi cells were positive for mitochondrial 8-oxog, representing the highest oxidative burden within macrophage clusters. STAT5 inhibition, both persistent and after established disease, significantly reduced this stress, thereby confirming the link between mitoROS and STAT5 activation ([214]Fig. 8K). Collectively, our data demonstrate that STAT5 affects atherosclerosis by specifically confining the end-stage Trem2^hiGpnmb^hi foam cell cluster. DISCUSSION This study demonstrates the effects of mitochondria-derived ROS during macrophage-to-foam cell differentiation. We identified mitochondrial O[2]^•− as a signaling molecule altering chromatin accessibility in macrophages, particularly within the STAT5 regulatory network. In addition, mitochondrial O[2]^•− directly activated STAT5 within mitochondria, and STAT5-enriched macrophages were predominantly classified as foam cells. By inducing mitochondrial O[2]^•−, oxLDL can therefore induce a feedback loop that promotes macrophage differentiation into foam cells by altering their metabolism. Inhibition of STAT5 activation in an in vivo mouse model of atherosclerosis lowered mitochondrial oxidative stress, limited foam cell accumulation, and altered foam cell differentiation. Using scRNA-seq, we identified the foam cell cluster of Trem2^hiGpnmb^hi to be the main macrophage subset affected by STAT5 inhibition. Mechanistically, we demonstrated that mitochondrial STAT5 rewires macrophage metabolic dynamics, shifting the cells toward glucose by inducing a STAT5-mediated inhibition of the PDC and by inducing mitochondrial O[2]^•− production through complex III of the ETC ([215]Fig. 9). In conclusion, we provide evidence that mitochondrial O[2]^•− induced by oxLDL is a crucial driver of macrophage-to-foam cell differentiation through STAT5 activation, linking oxidative stress, metabolic reprogramming, and epigenetic regulation in atherosclerosis. Fig. 9. Working model of oxLDL-induced metabolic reprogramming in foam cells. [216]Fig. 9. [217]Open in a new tab oxLDL triggers mitochondrial oxidative stress in macrophages through a STAT5-dependent pathway that promotes foam cell formation (lipid-laden foam cell phenotype). Within mitochondria, oxLDL reduces the mitochondrial membrane potential (ΔΨ[m]) and stimulates the production of mitochondrial superoxide (O[2]^•−), which subsequently activates STAT5. Once activated, STAT5 interacts with the PDC, thereby contributing to reduction of PDC activity. This metabolic blockade decreases the OCR, stalling the ETC. Consequently, complex III amplifies mitochondrial O[2]^•− production, creating a self-reinforcing feedback loop. In parallel, oxLDL promotes glycolysis through a mechanism that is largely STAT5-independent. In the absence of STAT5 (macrophage phenotype), mitochondrial respiration is increased already at the baseline, with partially preserved oxidative metabolism upon oxLDL exposure. This is accompanied by reduced mitochondrial O[2]^•− production and increased PDC activity. As a result, mitochondrial respiration and ETC activity are elevated, while mitochondrial O[2]^•− generation is reduced. However, even in the absence of STAT5, oxLDL treatment continues to enhance glycolysis, impair mitochondrial respiration, and disrupt redox homeostasis. Overall, STAT5 mediates oxLDL-induced mitochondrial dysfunction, oxidative stress, and foam cell formation in macrophages. Although STAT5 loss partially restores the mitochondrial function, oxLDL-induced glycolysis and mitochondrial impairment persist, highlighting additional STAT5-independent pathways. Created in BioRender. Hohensinner, P. (2025). [218]https://biorender.com/jx9oiyv. Excessive oxidative stress is considered to be a leading cause of atherosclerosis pathology ([219]56). In the initial phase, ROS oxidize oxLDL particles that infiltrate the subendothelial space, initiating atherosclerotic plaque formation ([220]1). These oxLDL particles in turn can themselves induce mitoROS ([221]57). Using 8-oxog to pinpoint ROS burden sources within human and mouse lesions ([222]58), and measuring mitochondrial O[2]^•− in macrophages and foam cells from aortic tissue via flow cytometry, we found that mitoROS burden was low during early stages of atherosclerosis but progressively increased, ultimately affecting more than 60% of foam cells during advanced disease. Given that ROS play a key signaling role and are essential in macrophage biology ([223]12, [224]13, [225]59), we modeled ROS-induced stress ex vivo during monocyte-to-macrophage differentiation. To specifically induce mitoROS stress, we used DMNQ, as our findings indicated that plaque foam cells are predominantly characterized by high levels of mitochondrial 8-oxog ([226]57, [227]60). MitoROS stimulation induced distinct alterations within macrophages including changes in chromatin accessibility, especially at STAT5 target gene sites. Reanalysis of these target genes in scRNA-seq data and snATAC-seq data from human atherosclerotic lesions ([228]29, [229]30) confirmed STAT5 pathway activation predominantly in foam cells. STAT5 or mitoROS inhibition significantly reduced the protein expression of the scavenger receptor CD36 and decreased lipid-laden foam cell formation in human macrophages. Furthermore, Stat5 deficiency in mouse macrophages restored oxLDL-impaired migratory capacity and reduced their phagocytic activity while also mirroring effects observed in human macrophages. In the context of atherosclerosis, macrophages phagocytose oxidized lipids and lose their migratory ability, causing their retention and accumulation within plaques ([230]26–[231]28). CD36 signaling has been shown inhibit foam cell migration ([232]61), suggesting that the improved migratory capacity observed in STAT5-deficient macrophages may be, at least in part, dependent on the reduction of CD36 expression. This improvement in macrophage migration, combined with reduced phagocytosis, may facilitate the clearance of lipid-laden macrophages and lower the overall macrophage burden within the plaque. Furthermore, we identified ETC complex III as the primary source of mitochondrial O[2]^•− production in response to oxLDL via STAT5-dependent activation. This finding aligns with recent evidence implicating pyruvate kinase M2 in complex III–driven mitoROS generation during oxLDL/CD36 signaling ([233]62); thus, STAT5 could be required for establishing the redox environment that facilitates the interaction of pyruvate kinase M2 with the mitochondrial network. Previous data suggest that STAT5 can modulate cellular metabolism by interacting with the mitochondrial PDC ([234]44, [235]45). The PDC catalyzes the irreversible decarboxylation of pyruvate to acetyl-CoA, thereby providing the substrate for the TCA cycle ([236]46). Activated STAT5 can disrupt this shuttle by disrupting PDC integrity, thereby diverting metabolic flux toward glycolysis and promoting a glycolytic phenotype ([237]44, [238]45). Our ex vivo results demonstrate that STAT5 interacts with the PDC within mitochondria, specifically binding the PDC-E2, PDHA1, and PDHB subunits. In addition, we observed a colocalization of mitochondrial—but not nuclear—PDC-E2 and pYSTAT5 in oxLDL-stimulated human macrophages that was abrogated by STAT5 inhibition as well as mTEMPO treatment. STAT5-deficient macrophages exhibited improved PDC activity upon oxLDL exposure, indicating that mitoROS-dependent STAT5 activation impairs the PDH function. As a consequence of these metabolic alterations, oxLDL stimulation increased lactate production coupled with impaired OXPHOS in wild-type macrophages, consistent with a metabolic shift toward glycolysis. Notably, although STAT5 deficiency enhanced basal mitochondrial respiration and prevented the oxLDL-induced inhibition of PDC activity, it did not fully normalize the mitochondrial function under oxLDL stress. In Stat5^−/− macrophages, oxLDL stimulation still led to a decline in basal respiration and ATP production, indicating that STAT5 contributes to the metabolic impairments induced by oxLDL but is not solely responsible for them. Similarly, lactate levels and NADH/NAD^+ ratio remained elevated in oxLDL-treated Stat5^−/− cells, suggesting that ongoing glycolytic NADH production and redox imbalance persist independently of STAT5. Together, these findings indicate that STAT5 plays an important role in the regulation of basal mitochondrial function and is involved in oxLDL-induced mitochondrial suppression likely through its interaction with the PDC. However, oxLDL also activates additional STAT5-independent mechanisms that contribute to the broader metabolic remodeling, including the sustained glycolytic shift and partial mitochondrial dysfunction. Our findings are consistent with previous data from Chen et al. ([239]8), which suggest that STAT5 may act downstream of glycolysis by repressing the conversion of pyruvate into acetyl-CoA, thereby necessitating increased glucose uptake to compensate for mitochondrial dysfunction. Overall, our findings suggest that mitochondrial STAT5 and mitochondrial O[2]^•− form a feedback loop in macrophages, decoupling metabolism from OXPHOS and favoring glucose as the primary energy substrate. This metabolic shift, together with chromatin remodeling at lipid uptake receptors and lipid-associated proteins, indicates that mitoROS reprogram macrophages toward a lipid storing phenotype that drives foam cell expansion. Building on previous data demonstrating the beneficial effects of ex vivo and in vivo STAT5 inhibition at the onset of the disease ([240]63, [241]64), we showed that inhibiting STAT5 after initial plaque formation led to reduced atherosclerotic lesions. Regarding foam cells, we would anticipate an overall reduction in their number, driven by decreased macrophage-to-foam cell differentiation coupled with altered oxidative stress. This effect is likely due to the disruption of the oxLDL-induced, STAT5-dependent mitochondrial O[2]^•− feedback loop. Within our mouse model, STAT5 activation was markedly induced over time, with more than 90% of foam cells exhibiting activated STAT5 at 16 weeks of HFD. Inhibition of STAT5 activation markedly reduced mitochondrial 8-oxog lesions at 16 weeks of HFD, restoring levels to those observed in macrophages derived from animals fed with regular chow diet, and attenuated mitochondrial oxidative DNA damage levels in aortic foam cells. This reduction was independent of circulating monocyte levels as well as cholesterol and LDL serum levels, which remained elevated and similar to vehicle-treated HFD-fed animals. To address specific alterations within atherosclerotic plaque lesions in detail, we performed scRNA-seq at three different time points in HFD-fed mice with and without STAT5 inhibition, focusing on alterations within the macrophage population and specifically on foam cell changes. Plaque macrophages are categorized into five major subsets including foamy lipid–associated, inflammatory, and resident macrophages ([242]29, [243]30, [244]52, [245]53, [246]65, [247]66). The foamy macrophage cluster in mice consists of Trem2^hiGpnmb^hi and Trem2^hiSlamf9^hi subpopulations that have distinct characteristics. TREM2 is not only a foam cell marker but exerts a dual role in atherosclerosis, acting as protective in early atherosclerosis, where its deletion leads to increased necrotic core formation, but becoming detrimental, possibly at later stages, by promoting lipid uptake, thereby contributing to plaque growth ([248]67, [249]68). Within Trem2-expressing macrophages, the Trem2^hiSlamf9^hi macrophages are enriched for inflammatory markers and have low lipid metabolism, while Trem2^hiGpnmb^hi macrophages express a higher level of genes involved in lipid metabolism and scavenger receptor signaling ([250]52, [251]54). We showed that both foam cell clusters and their associated marker genes increased from 10 to 16 weeks of HFD, with only minimal foam cells present at 4 weeks. Through trajectory analysis, we showed that a small proportion of the Trem2^hiGpnmb^hi cluster emerges earlier in pseudotime and is suppressed by STAT5 inhibition. However, the main trajectory observed for foam cells in our dataset is a progression from a Trem2^hiSlamf9^hi macrophage to a Trem2^hiGpnmb^hi macrophage. This transition is altered under STAT5 inhibition, leading to a prolonged presence of the early Trem2^hiSlamf9^hi foam cell phenotype and a loss or delay of the late Trem2^hiGpnmb^hi foam cell phenotype. We confirmed the reduced presence of the late Trem2^hiGpnmb^hi foam cells in atherosclerotic lesions using immunohistochemistry. On a transcriptional level, the Trem2^hiGpnmb^hi was also defined by a STAT5 target gene expression signature, further confirming our data from human scRNA-seq and snATAC-seq datasets with inhibition of STAT5 activation having a large impact on gene regulation within this cluster. Specifically, STAT5 inhibition suppressed inflammatory and hypoxia-related genes while enhancing the expression of genes involved in OXPHOS, FAO, and mitochondrial stress responses. In contrast, the Trem2^hiSlamf9^hi cluster was only minimally affected by STAT5 inhibition. Notably, promoting TREM2 signaling in macrophages reduced apoptosis and increased OXPHOS in plaque macrophages, together with reducing STAT5 signaling, indirectly confirming our results of STAT5 inhibition on metabolism and macrophage function ([252]69). Together, we propose that oxLDL-induced changes in macrophages involve the production of mitochondrial O[2]^•−, leading to subsequent STAT5 activation. Once activated, STAT5 reinforces mitochondrial O[2]^•− generation via complex III, establishing a self-sustaining positive feedback loop that amplifies oxidative stress. In parallel, STAT5-dependent inhibition of PDC appears to contribute to reduced oxygen consumption, consistent with attenuated mitochondrial activity. This metabolic adaptation supports lipid accumulation and foam cell formation but likely acts in coordination with additional STAT5-independent mechanisms. Moreover, mitoROS reshape the macrophage chromatin landscape, increasing accessibility at STAT5 target gene sites, thus establishing a STAT5 pathway signature in foam cells. By inhibiting STAT5 activation, we disrupted this mitoROS-STAT5 feedback loop, reducing oxidative stress and delaying the transition from early Trem2^hiSlamf9^hi foam cells to mature Trem2^hiGpnmb^hi foam cells, ultimately decreasing atherosclerotic plaque formation. Limitations of the study While our study establishes a strong link between mitoROS, STAT5 activation, and foam cell formation, additional research is needed to dissect stage-specific effects and explore the STAT5-independent mechanism of oxLDL-induced metabolic reprogramming. Moreover, while this study focuses on macrophages, STAT5 expression in endothelial and smooth muscle cells, which are also targeted by lipids, may influence plaque stability and atherosclerosis progression, highlighting the need for further investigation. MATERIALS AND METHODS Experimental model and study participant details Human atherosclerotic plaques samples Human plaque tissue was derived from explanted carotid arteries from 21 patients undergoing carotid endarterectomy (mean age, 70.5 ± 8 years; 71% females) ([253]3, [254]70). Demographic and clinical data were collected for each patient, including sex, age, plaque status (symptomatic or asymptomatic), degree and location of carotid stenosis, comorbidities, and body mass index, and are reported in table S1. Patients with high-grade carotid artery stenosis undergoing carotid endarterectomy were prospectively included in the study. All surgical procedures were performed at the Department of Vascular Surgery, Medical University of Vienna. This single-center study was approved by the local institutional review board (approval number 2449/2020) and was performed in accordance with the principles of the Declaration of Helsinki. All patients gave written informed consent. The indication for surgery included symptomatic carotid artery stenosis (≧50%) or high-grade asymptomatic carotid artery stenosis (≧70%). All surgeries were performed by trained vascular surgeons. The degree of luminal narrowing was determined by carotid Duplex sonography and/or computed tomography angiography using the criteria of the North American Symptomatic Carotid Endarterectomy Trial ([255]71). Duplex sonography examinations were performed by independent, trained medical technical assistants. The morphological evaluation of carotid artery stenosis before surgery was assessed using carotid Duplex ultrasound and/or computed tomography angiography. Plaques were classified according to ultrasound echogenicity into echolucent (noncalcified, unstable, or vulnerable plaque), echogenic (calcified, stable, or nonvulnerable plaque), and mixed plaques (partly calcified) by the expert vascular laboratory at the Medical University of Vienna. For histological classification, endarterectomy specimens were formalin fixed and embedded in paraffin at the Department of Pathology. Transverse sections were cut each 3 mm along the plaque by experienced technicians. Specimens were stained with hematoxylin and eosin (H&E) and elastic van Giessen staining and classified according to modified American Heart Association classification based on morphological description. Human blood samples Blood was obtained from healthy volunteers according to the recommendations of the ethical board of the Medical University of Vienna including informed consent (approval number 1575/2014). Mice Vav1-Cre/+Stat5ab^fl/fl (Stat5^−/−) mice ([256]38) were housed and bred at the University of Veterinary Medicine of Vienna, Vienna, Austria. Mice were used at an age of 8 to 12 weeks for the isolation of bone marrow. For in vivo experiments, we used female and male littermate ApoE^−/− mice (B6.129P2-Apoetm1Unc/J, Charles River Laboratories) between 8 to 12 weeks of age. Animals were obtained from the Institute of Biomedical Research, Medical University of Vienna (Himberg, Austria), and bred in-house. They were housed in groups of three to five animals grouped by sex in a controlled environment [temperature of 21°C (±2°C) with 55% (±10%) humidity, 12-hour dark/12-hour light cycle] and had access to conventional chow diet and water ad libitum. For induction of atherosclerosis, mice were fed an HFD, with 21% fat and 0.20% cholesterol (TD88137, ssniff). No exclusion criteria were necessary. All mouse experiments were approved by the Animal Care and Use Committee of the Medical University of Vienna and the Austrian Federal Ministry of Education, Science and Research (authorization numbers BMBWF 2020-0.719.600, 2021-0.588.496, and 2023–0.794.836) and by the Animal Welfare Committee at the University of Veterinary Medicine of Vienna (authorization number BMBWF 2022-0.404.452). The three R’s of ethics in animal testing were followed and taken into account while planning all experiments. All mice were housed under identical conditions in a specific pathogen–free facility according to the Federation of European Laboratory Animal Science Association guidelines and additionally monitored for being norovirus-negative. For aortic plaque and flow cytometry in vivo staining ([257]Figs. 1 to [258]5 and figs. S1 to S3), 4 ApoE^−/− mice were fed with an HFD for 4 weeks, 5 mice for 10 weeks, and 12 mice for 16 weeks (fig. S1A). For the in vivo experiments in [259]Figs. 7 and [260]8 and figs. S6 and S7, 60 ApoE^−/− mice were used. Six mice (8 weeks old) were fed with normal chow for 7 weeks and were used as “young” controls; six mice (8 weeks old) were fed with normal chow for 17 weeks and were used as “old” controls; 12 mice were fed for 4 weeks with an HFD, together with daily administrations of dimethyl sulfoxide (DMSO) or S5i (cat. no. S6784; Selleckchem); 12 mice were fed for 10 weeks with an HFD, together with daily administrations of DMSO or S5i administered for 10 weeks or 4 weeks after initiation of HFD for the remaining 6 weeks; and 12 mice were fed for 10 weeks with an HFD, together with daily administrations of DMSO or S5i administered for 16 weeks or 4 weeks after initiation of HFD for the remaining 12 weeks. The S5i inhibitor was dissolved in 5% DMSO (Sigma-Aldrich) and administered at 10 mg/kg in a mixture of Nutella and soy milk. DMSO (5%) + milk and Nutella were administered in HFD control mice. Primary macrophages Primary cells were grown in a humidified incubator (5% CO[2] and 37°C). BMDMs were differentiated from bone marrow isolated from femurs and tibias of 8- to 12-week-old Vav1-Cre/+Stat5ab^fl/fl (Stat5^−/−) mice (both genders) and their wild-type counterpart (Stat5^fl/fl) ([261]38). Generation of hMDMs is described in the “Method details” section. Method details Isolation and culture of monocyte-derived macrophages hMDMs were generated as previously described ([262]70). Briefly, peripheral blood mononuclear cells were isolated from whole blood via density gradient centrifugation using lymphocyte separation medium (cat. no. C-44010; PromoCell). Monocytes were isolated via adherence to plastic-treated tissue culture plates. Monocytes were differentiated into macrophages for 7 days in differentiation medium [RPMI 1640 medium (Sigma-Aldrich) supplemented with human macrophage colony-stimulating factor (M-CSF; 100 ng/ml; cat. no. 574808; BioLegend), 10% fetal bovine serum (FBS; Biochrome Millipore), penicillin (100 U/ml), streptomycin (100 U/ml), fungizone (0.25 μg/ml), and 2 mM l-glutamine (all Gibco)] with two medium changes in between. BMDM generation BMDMs were differentiated for 7 days in differentiation medium [RPMI 1640 medium (Sigma-Aldrich) supplemented with mouse M-CSF (100 ng/ml; cat. no. 576404; BioLegend), 10% FBS (Biochrome Millipore), penicillin (100 U/ml), streptomycin (100 U/ml), fungizone (0.25 μg/ml), and 2 mM l-glutamine (all Gibco)] and cultured into nontreated tissue culture plates. Polarization and treatment of human and mouse macrophages Both human and mouse macrophages were seeded into treated tissue culture plates for ex vivo experiments in complete medium [RPMI 1640 medium (Sigma-Aldrich) supplemented with 10% FBS (Biochrome Millipore), penicillin (100 U/ml), streptomycin (100 U/ml), fungizone (0.25 μg/ml), and 2 mM l-glutamine (all Gibco)]. hMDMs were stimulated for 24 hours with LPS (100 ng/ml; cat. no. L2880; Sigma-Aldrich) and IFN-γ [100 ng/ml; cat. no. 570206 (human)]. Naïve (M0) macrophages were left unstimulated. Both macrophage types were differentiated either with or without addition to the differentiation medium of 0.2 μM DMNQ (cat. no. ALX-420-02; Enzo Life Sciences). The compound was used at a very low concentration that did not affect cell viability (fig. S2B). For the ATAC-seq dataset, macrophages were differentiated in the presence of 100 μM H[2]O[2] (Sigma-Aldrich) as the control. For inhibition of STAT5 activation and signaling in hMDMs, S5i (cat. no. S6784; Selleckchem) was used at the concentration of 300 μM with 1-hour pretreatment. For inhibition of mitoROS, the ROS scavenger mTEMPO (cat. no. sc-221945; Santa Cruz Biotechnology) was used with 1-hour pretreatment at the concentrations of 500 μM in hMDMs and 300 μM in BMDMs, and the mitochondrion-targeted antioxidant MitoQ (cat. no. S8978; Selleckchem) was used with 2-hour pretreatment at the concentration of 500 nM. oxLDL (cat. no. [263]L34357; Thermo Fisher Scientific) was used at the concentration of 50 μg/ml in both cell types and DiI-oxLDL (cat. no. L34358; Thermo Fisher Scientific) at 50 μg/ml in hMDMs, both for 24 hours of stimulation. oxLDL and DiI-oxLDL stimulation was carried out in complete medium depleted of 10% FBS with the addition of 0.2% bovine serum albumin (BSA; Sigma-Aldrich). Aorta, spleen, and bone marrow isolation At the end of each time point of atherosclerosis induction, blood was collected from the vena cava inferior of the euthanized mice, and vascular perfusion was performed with phosphate-buffered saline (PBS) via cardiac puncture. The aorta was dissected, cleaned from perivascular adipose tissue, and either fixed in 4% paraformaldehyde (PFA) overnight for subsequent histology or minced into fine pieces and digested with 2.5 ml of digestion solution [Hanks’ balanced salt solution (Thermo Fisher Scientific) and collagenase IV (2 mg/ml; Sigma-Aldrich) + deoxyribonuclease I (50 U/ml; Sigma-Aldrich)] at 37°C for 1 hour by gentle shaking at 200 rpm. The resulting suspension was pipetted until it was homogenized and filtered through a 70-μm cell strainer. The cell suspension was washed with PBS + 2% FBS before spinning down. Red blood cells were lysed with Erylysis Buffer [150 mM NH[4]Cl, 10 mM KHCO[3], and 0.1 mM Na[2]EDTA] for 5 min and then washed twice with PBS + 2% FBS before usage. The spleen was isolated as previously described ([264]72). Flow cytometry For flow cytometry analysis, hMDMs and BMDMs were detached from the plates with Accutase solution (cat. no. 423201; BioLegend), pelleted, and washed once with PBS before staining. Human antibodies used for staining were 1:20 CD80 [FITC (fluorescein isothiocyanate); cat. no. 11-0809-42; eBioscience], 1:20 CD142 (APCs; cat. no. 365206; BioLegend), 1:20 CD146 (APC-Cy7; cat. no. 361041; BioLegend), and 1:20 CD36 (Pe-Cy7; cat. no. 336221; BioLegend). For in vivo staining of total murine aortic macrophages and foam cells and blood monocytes, neutrophils, and platelets, we used CD45 (BV650; cat. no. 103151; 1:100), CD11b (Pe-Cy5; cat. no. 101210; 1:100), F4/80 (Pe-Cy7; cat. no. 123114; 1:25), Ly6G (AF700; cat. no. 127622; 1:100), CD146 (AF594; cat. no. 134720; 1:100), TREM2 (PE; cat. no. 824806; 1:50), CD41 (FITC; cat. no. 133904; 1:100), and CD115 (APC/Fire 750; cat. no. 135536; 1:25) all from BioLegend. Macrophages were defined as CD45^+ Ly6g^− Cd11b^+ F4/80^+ and foam cells as either CD45^+ Ly6g^− Cd11b^+ F4/80^+ Trem2^+ or CD45^+ Ly6g^− Cd11b^+ F4/80^+ CD146^+ subsets. For circulating monocyte and neutrophil identification, mouse blood was collected from vena cava and immediately supplemented with 10% acid citrate-dextrose (Sigma-Aldrich) as an anticoagulant. For each staining, 10 μl of blood was used. Red blood cells were lysed with Erylysis Buffer [150 mM NH[4]Cl, 10 mM KHCO[3], and 0.1 mM Na[2]EDTA] for 5 min and then stained with the appropriate fluorochrome-conjugated antibodies for 20 min. Monocytes were identified as CD45^+ Ly6g^− Cd11b^+ CD115^+ Ly6c^+ cells. These were further subdivided into Ly6C^hi and Ly6C^lo subsets on the basis of Ly6C expression levels. Neutrophils were defined as CD45^+ Ly6g^+ Cd11b^+ cells. Platelet-monocyte aggregates were identified by the coexpression of CD41 (platelet marker) with monocyte markers. Specifically, CD41^+ events coexpressing Ly6C^hi and Ly6C^lo monocyte markers were gated. Similarly, platelet-neutrophil aggregates were identified by gating CD41^+ events coexpressing Ly6G^+ neutrophil markers. For the phagocytosis of pHrodo BioParticles, we used pHrodo Red Zymosan BioParticles (cat. no. [265]P35364) for hMDMs and pHrodo Green E. coli BioParticles (cat. no. [266]P35366; both from Thermo Fisher Scientific) for BMDMs according to the manufacturer’s protocol. Briefly, particles were suspended in PBS and sonicated for 10 min in a sonicator bath. Thirty micrograms of Zymosan BioParticles or 100 μg of E. coli BioParticles were added directly onto cells and incubated for 1 hour at 37°C. Phagocytosis was stopped by ice incubation of cells. After PBS washing, cells were detached from the plate, resuspended in PBS, and acquired. For intracellular staining, cell pellets were fixed in 2% PFA for 10 min and then permeabilized with ice-cold 90% methanol in PBS + 2% FBS on ice for 30 min. Cells were then washed twice with PBS + 2% FBS and stained for 1 hour with anti–pYSTAT5-APC (1:20; cat. no. 17-9010-42; Thermo Fisher Scientific). Cells were then washed twice with PBS + 2% FBS and immediately acquired. Fluorescence-activated cell sorting (FACS) analysis was carried out with the CytoFLEX LX Flow Cytometer (Beckman Coulter) at the Flow Cytometry Core Facility of the Medical University of Vienna ([267]https://corefacilities.meduniwien.ac.at/unsere-units/flow-cytomet ry/). Data were analyzed with CytExpert Software version 2.4. Representative FACS plots have been created with CytExpert Software version 2.4 and FlowJo Software (version 10.8.1; BD Life Sciences). Immunohistochemistry and immunocytochemistry Preparation of the mouse aortic root was performed according to an established protocol and described in the “Aorta, spleen, and bone marrow isolation” section (hearts were fixed in 4% PFA and embedded in paraffin). After the appearance of all three aortic valve leaflets, the aortic root was cut into 5-μm-thick sections. H&E staining was performed as previously described ([268]60). Specimens were then deparaffinized and rehydrated. Slides were then stained for 4 min with hematoxylin (cat. no. 11895; Morphisto), rinsed for 4 min with tap water, stained for 30 s with eosin Y (cat. no. HT110116-500ML; Sigma-Aldrich), and finished by dehydration with ethanol and xylene. Hepatic steatosis was evaluated in liver tissue by H&E staining and data shown as the percentage of area covered by fat droplets normalized on the total liver area. Immunohistochemistry was performed as previously described ([269]61). Briefly, human plaque tissue was fixed in 4% PFA and embedded in paraffin. Specimens were then deparaffinized and rehydrated; target retrieval was performed using a citrate buffer solution. For 8-oxog staining, target retrieval was performed with pepsin reagent (cat. no. 10690501; Sigma-Aldrich). Sections were blocked with blocking buffer (2% BSA, 0.5% fish gelatine, and 0.3% Tween 20 in PBS) for 90 min, followed by primary antibody incubation overnight in PBG buffer (0.5% BSA and 0.2% fish gelatine in PBS). The next day, the specimens were stained with secondary antibodies when appropriate. Nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI; 1:2000), and samples were mounted in Fluoromount-G Mounting Medium (cat. no. 00-4958-02; Thermo Fisher Scientific). Primary antibodies were CD68 [cat. no. sc7083 (human); 1:200; Santa Cruz Biotechnology; cat. no. ab283654 (mouse); 1:500; Abcam], CD80 (cat. no. PA5-85913; 1:100; Thermo Fisher Scientific), CD206 (cat. no. A5-16871; 1:100; Thermo Fisher Scientific), CD146 [cat. no. AF932 (human); 1:100; Bio-Techne; cat. no. ab228540 (mouse); 1:500; Abcam], TREM2 (cat. no. MAB17291; 1:500; Bio-Techne), pYSTAT5 (cat. no. 9359; 1:200; Cell Signaling), and 8-oxog (cat. no. ab206461; 1:200; Abcam). Cy2-, Cy3-, and Cy5-labeled secondary antibodies (1:250; all from Thermo Fisher Scientific) were used for visualization when appropriate. Mitochondria were stained with TOM22 (cat. no. 11278-1-AP; 1:100; Thermo Fisher Scientific) and nuclei with DAPI (1:2000; Thermo Fisher Scientific). Images were scanned on a Tissue FAXS automated microscopy stage (TissueGnostics) mounted on an Observer Z1 microscope at 20× magnification [numerical aperture (NA), 0.5]. Whole aortic sections were systematically scanned in an automated manner to ensure uniform coverage and eliminate user bias. Images from the same experiment were acquired under identical exposure settings, magnifications, and Z-stack parameters to maintain consistency. Secondary antibody staining alone served as the negative control for tissue autofluorescence. Regions with excessive autofluorescence or tissue artifacts were excluded from the analysis. No significant differences in staining were observed between symptomatic and asymptomatic human plaques. For colocalization analysis in atherosclerotic plaques, pYSTAT5 (cat. no. 9359; 1:200; Cell Signaling), PDHB (cat. no. 860401; 1:100; BioLegend), and CD68 [cat. no. sc7083 (human); 1:200; Santa Cruz Biotechnology; cat. no. ab283654 (mouse); 1:500; Abcam] antibodies were used and stained as described above. Colocalization analysis was performed by measuring the area occupied by pYSTAT5 and PDHB signals within plaque cells. To quantify colocalization, Pearson’s correlation coefficient was calculated, providing an objective measure of the spatial correlation between the two protein signals. Immunocytochemistry was performed as previously described ([270]62). Briefly, hMDMs (2.5 × 10^5) were differentiated for 7 days on eight-well glass chamber slides (cat. no. 80841; Ibidi) precoated with gelatin for 30 min. On the day of the experiments, cells were stimulated accordingly to the experiment requirement and fixed with 4% PFA for 15 min at room temperature. Blocking was carried out in blocking buffer for 90 min. The primary antibodies [pYSTAT5 (cat. no. 9359; 1:200; Cell Signaling), PDC-E2 (cat. no. B614251; 1:100; BioLegend), and VDAC (cat. no. 600-101-HB2; 1:200; Thermo Fisher Scientific)] were diluted in PBG buffer and incubated overnight at 4°C. Cy2 (cat. no. SA5-10038), Cy3 (cat. no. [271]A31570), and Cy5 (cat. no. SA5-10089) antibodies (1:200; all from Thermo Fisher Scientific) were used for visualization and incubated for 30 min in PBS, together with DAPI (1:2000). Secondary antibody staining alone served as the negative control for tissue autofluorescence. Samples were mounted in Fluoromount-G Mounting Medium (cat. no. 00-4958-02; Thermo Fisher Scientific) and left overnight at 4°C before image acquisition. Images were acquired using a Zeiss LSM 980 Confocal Microscope with a 20× air objective and ×0.8 magnification at the Imaging Facility of the Medical University of Vienna ([272]https://corefacilities.meduniwien.ac.at/en/our-units/imaging/). To analyze pYSTAT5 levels in the nucleus (DAPI^+ signals) and mitochondria (VDAC^+ signals), the mean fluorescence intensity (MFI) was quantified. For colocalization between pYSTAT5 and PDC-E2, Pearson’s correlation coefficient was calculated. All analyses were performed using CellProfiler version 4.2.5. pipelines ([273]63). H&E-stained aortic root sections were analyzed at 50-μm intervals, with the plaque size and necrotic core area quantified with QuPath version 0.5.1 ([274]64). For image analysis, a minimum of 1000 cells was evaluated under each condition. For in vivo analysis, the entire plaque section was included in the analysis. For representative panels in all figures, all images were processed using ImageJ software ([275]65). Background intensity levels were standardized across all images with the following settings: DAPI (616-38318), Cy2 (932-8657), Cy3 (34-35263), and Cy5 (649-21840). All adjustments were applied to the entire image. The color depth was changed to 16 bit before generating composite overlays. DAPI was set to cyan when displayed together with Cy5 (red) and Cy3 (yellow) and to blue when combined with Cy2 (green), Cy3 (yellow), and Cy5 (set to magenta). The color space of the picture was converted to red, green, and blue, and a scale bar was inserted for each picture. MitoROS assessment To assess cellular ROS levels, CellROX Green (cat. no. [276]C10444; Thermo Fisher Scientific) was used at the concentration of 5 μM, and cells were stained in complete medium for 30 min at 37°C according to the manufacturer’s protocol. For mitochondrion-specific O[2]^•− quantification, cells were stained with either MitoSOX red (cat. no. [277]M36008; Thermo Fisher Scientific) or MitoNeoD (cat. no. [278]AOB37866, AOBIOUS), both at the concentration of 5 μM. Staining was performed for 20 min at 37°C in Hanks’ balanced salt solution or PBS, as recommended by the manufacturers. MitoROS signals were subsequently analyzed by flow cytometry. For high-resolution imaging of mitochondrial O[2]^•− in hMDMs, cells were co-incubated with 5 μM MitoNeoD (cat. no. [279]AOB37866, AOBIOUS), 100 nM MitoTracker Green (MTG) FM (cat. no. M7514; Thermo Fisher Scientific), and 1 μM Hoechst (no. H1399; Thermo Fisher Scientific) in Dulbecco’s modified Eagle’s medium (10% FBS and 1% penicillin-streptomycin) for 20 min. The cells were then washed with PBS, and the medium was replaced with phenol red–free Dulbecco’s modified Eagle’s medium (Gibco). All images were acquired using a Zeiss LSM 980 confocal microscope with a 20× air objective (NA, 0.8) at the Imaging Facility of the Medical University of Vienna ([280]https://corefacilities.meduniwien.ac.at/en/our-units/imaging/). The MFI was quantified in 100 to 200 cells per biological replicate to assess O[2]^•− levels. Mitochondrial membrane potential assessment To evaluate the mitochondrial membrane potential (ΔΨ[m]), macrophages were gently detached using a cell scraper, washed once with 1× PBS, and incubated with 20 nM TMRM (cat. no. [281]M20036, Thermo Fisher Scientific) and 100 nM MTG FM (cat. no. M7514, Thermo Fisher Scientific) in PBS for 20 min at 37°C, following the manufacturer’s protocol. After staining, cells were washed with 1× PBS and resuspended in PBS supplemented with 1% BSA. Samples were acquired using a CytoFLEX LX flow cytometer (Beckman Coulter) at the Flow Cytometry Core Facility of the Medical University of Vienna ([282]https://corefacilities.meduniwien.ac.at/unsere-units/flow-cytomet ry/). As a positive control for mitochondrial depolarization, cells were treated with 50 μM carbonyl cyanide 3-chlorophenylhydrazone for 5 min at 37°C and 5% CO[2] before staining, as described above. Flow cytometry data were analyzed using CytExpert Software (version 2.4). The TMRM fluorescence intensity was normalized to the percentage of MTG-positive cells to quantify ΔΨ[m], allowing for correction based on mitochondrial mass. Transwell migration assay hMDMs and BMDMs (1 × 10^5) were seeded in transwell 24-well plates with a 5.0-μm pore size (cat. no. CLS3421-48EA; Corning) for 3 hours in complete serum-free medium. Then, the medium was changed into complete medium (to induce migration) and into fresh complete serum-free medium (for the negative control). After 4 hours of migration, cells were washed once with PBS, and then the membrane of the transwell was carefully passed through a cotton swap to remove nonmigrated cells. Cells were stained with DAPI (1:2000) for 30 min, and nuclei were visualized from images scanned on an automated Tissue FAXS microscopy stage on an Observer Z1 microscope at 40× magnification (NA, 0.5). The area of the DAPI-positive signal was calculated with ImageJ ([283]73), and migrated cells were expressed as the percentage of nonmigrated cells or further normalized to M0 unstimulated cells. Isolation of mitochondria Mitochondria were isolated using the Mitochondria Isolation Kit for Cultured Cells (cat. no. 89801; Thermo Fisher Scientific) according to the manufacturer’s instructions. Briefly, 2 × 10^7 BMDMs were seeded in 15-cm nontreated tissue culture plates, treated accordingly (see fig. S5C), and washed once with PBS before isolation. Mitochondria were isolated through multiple centrifugation steps and ultimately separated from the cytoplasm. Whole-cell lysates were obtained from the same sample through the lysis of cells in CHAPS buffer [1× tris-buffered saline (TBS) + 2% CHAPS (Sigma-Aldrich)]. Mitochondria were lysed in 100 μl of CHAPS buffer. All lysates were stored at −20°C before immunoblot analysis. For immunoblotting, whole-cell lysates, mitochondria, and cytoplasmatic lysates were boiled at 95°C for 5 min in 1× SDS sample buffer (62.5 mM tris-HCl, pH 8, 2% SDS, 10% glycerol, 10% β-mercaptoethanol, and 0.002% bromophenol blue). Coimmunoprecipitation and immunoblotting BMDMs were lysed in lysis buffer [50 mM tris-HCl (pH 7.5), 150 mM NaCl, 1 mM Na[3]VO[4], 50 mM NaF (sodium fluoride), 2 mM EDTA, 1% Triton X-100, 0.1 mM phenylmethylsulfonyl fluoride, and 1× protease inhibitor], incubated for 5 min on ice, and then centrifuged for 5 min at 4°C and 13,400g. The supernatant was transferred to a new tube. The protein concentration was determined with the Pierce BCA Protein Assay Kit (cat. no. 23227; Thermo Fisher Scientific), and 100 μg of protein in 200 μl of lysis buffer was used for further immunoprecipitation. Thirty microliters of magnetic beads (Dynabeads protein G, cat. no. 10003D; Thermo Fisher Scientific) was added to the lysates to preclear for unspecific binding and was rotated for 30 min at room temperature. The precleared lysate was transferred into a new tube. Forty microliters (20% of the lysate used for the immunoprecipitation) was used as an input control. The STAT5 (no. 94205; 1 μl; Cell Signaling) antibody or IgG (immunoglobulin G) control (no. 3900S; 1 μl; Cell Signaling) was added to 200 μl of lysate and incubated overnight at 4°C while rotating. Ten microliters of magnetic beads was added to each sample and incubated for 3 hours at 4°C while rotating. Afterward, the beads were washed five times with 1 ml of lysis buffer, and proteins were eluted in 50 μl of 2× SDS sample buffer at 95°C for 10 min. Proteins were blotted on a nitrocellulose membrane at 4°C for 16 hours at 200 mA and then for 2 hours at 400 mA in carbonate transfer buffer (3 mM Na[2]CO[3], 10 mM NaHCO[3], and 20% ethanol). The membrane was blocked in 5% milk powder in TBS with Tween 20 for 1 hour at room temperature and then washed three times with TBS with Tween 20. The primary antibody to Vinculin (cat. no. V9131, 1:5000) was purchased from Sigma-Aldrich. The primary antibodies to PDC-E2 (cat. no. sc-271534, 1:500), ATP5A (cat. no. sc-136178, 1:500), PDHA1 (cat. no. sc-377092, 1:1000), Enolase (cat. no. sc-271384; 1:1000), and STAT5 for immunoblot detection (cat. no. sc-835-G; 1:1000) were purchased from Santa Cruz Biotechnology. PDHB (cat. no. 860401, 1:1000) was purchased from BioLegend. STAT5 for coimmunoprecipitation (no. 94205; 1:200) was purchased from Cell Signaling. The horseradish peroxidase–coupled secondary antibodies used were purchased from Cell Signaling (cat. nos. 7076S and 7074S, each used at 1:2000). For the development of protein signals, Amersham ECL Prime Western Blotting Detection Reagent (cat. no. 12994780; Cytiva Lifescience) was used. For signal detection, a Fusion FX (Vilber) imaging system was used. Serum lipid analysis Blood was collected using 23-gauge needles postmortem after more than 4 hours of fasting via the vena cava into collection tubes with citrate-dextrose solution (cat. no. C3821; Sigma-Aldrich). Plasma was obtained by centrifugation at 3000g for 5 min, then at 13,000g for 1 min, and at 13,000g for an additional 1 min to obtain platelet-free plasma. Plasma cholesterol, triglyceride, LDL, and HDL levels were measured with a Hitachi CobasC311 Bioanalyzer (Roche) according to the manufacturer’s protocol. Cell viability measurement ATP levels were determined using the CellTiter-Glo Luminescent Cell Viability assay (cat. no. G7570; Promega) according to the manufacturer’s instructions. Glucose uptake assay The effects on glucose uptake were measured using a nonradioactive, homogeneous bioluminescent 2DG Uptake-Glo kit (cat. no. J1341; Promega). hMDMs were differentiated in 96-well plates and treated as indicated (see fig. S4D). The complete medium was removed, and cells were washed twice with PBS and resuspended in PBS, and the assay was performed according to the manufacturer’s instructions. The luminescent signal in terms of relative light units was measured after 3 hours using the GloMax Plate Reader (Promega). l-Lactate measurement For intracellular l-lactate assessment, hMDMs were differentiated in 96-well plates and stimulated as indicated (see fig. S4B). Lactate levels were determined using the Lactate-Glo Luminescent assay (cat. no. J5021; Promega) according to the manufacturer’s instructions. Lactate was measured after 1 hour from the addition of the lactate detection reagent according to the manufacturer’s protocol. PDH enzyme activity PDH activity was measured using the Pyruvate Dehydrogenase Enzyme Activity Microplate Assay Kit (cat. no. ab109902; Abcam) according to the manufacturer’s protocol. Briefly, 5 × 10^6 BMDMs were plated in 6-cm plates and treated for 24 hours with oxLDL. The complete medium was removed, and cells were washed twice with PBS and lysed with a detergent solution. All buffers used for lysate preparation were supplemented with 1× protease inhibitor cocktail (cat. no. 11836170001; Roche) and 10 mM NaF (Sigma-Aldrich) to preserve the phosphorylated status of endogenous PDH. One hundred micrograms of lysates was then loaded onto the 96-well plate coated with an anti-PDH monoclonal antibody and incubated for 3 hours at room temperature. Two hundred microliters of assay solution was then added to each well, and the optical density (OD[450 nm]) was measured in kinetic mode at room temperature for 30 min with 20-s intervals. The relative activity rate was measured as Rate (mOD/min) = Absorbance 2 − Absorbance 1/Time (minutes). Absorbance 2 and Absorbance 1 represent the two time points for all the samples where the increase in absorbance is the most linear. Metabolic flux measurements The OCR and ECAR were measured on a Seahorse XFe96 platform (Agilent) using the Seahorse XF Cell Mito Stress test kit (Agilent) and the Seahorse XF Glycolysis Stress Test Kit, respectively, according to the manufacturer’s instructions. Briefly, 100,000 wild-type and Stat5^−/− BMDMs were seeded overnight in Seahorse XFe96 Cell Culture Microplates and underwent the respective treatment before proceeding with metabolic flux measurements. Cells were either treated with oxLDL for 24 hours or pretreated with 10 mM DMM (cat. no. [284]D97754; Sigma-Aldrich) for 3 hours or with 300 μM mTEMPO (cat. no. sc-221945; Santa Cruz Biotechnology) for 1 hour before oxLDL stimulation. Eight technical replicates and three biological replicates were used for each condition. Before OCR measurement, the medium was changed to XF Base Medium (Agilent) containing glucose (10 mM), sodium pyruvate (1 mM), and l-glutamine (2 mM); and cells were incubated for 1 hour in a CO[2]-free incubator. The basal OCR was measured over time followed by sequential injections with the mitochondrial inhibitors oligomycin (1.5 μM), carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP; 2 μM), and rotenone/antimycin A (500 nM). Before ECAR measurement, the medium was changed to XF Base Medium (Agilent) containing l-glutamine (2 mM), and cells were incubated for 1 hour in a CO[2]-free incubator. The basal ECAR was measured over time followed by sequential injections with glucose (10 mM), oligomycin (1.5 μM), and 2-deoxy-d-glucose (100 nM). Cells were lysed in 10 μl of 10 mM tris + 0.1% Triton X-100 (pH 7.4), and lysates were quantified with BCA (Thermo Fisher Scientific). Raw data were analyzed using Wave Desktop Software (Agilent; version 2.0). Data were normalized on protein content and analyzed with GraphPad Prism software (GraphPad; version 8.0.2). Metabolomics For the measurement of pyruvate, lactate, NAD^+, and NADH, Stat5^fl/fl and Stat5^−/− BMDMs were treated with oxLDL for 24 hours, then the medium was removed, and cells were quickly washed with ice-cold 1× PBS. Metabolites were extracted with 1 ml of mass spectrometry (MS) extraction buffer [50% (v/v) methanol, 30% (v/v) acetonitrile, and 20% (v/v) high-performance liquid chromatography (HPLC)–grade water] and incubated for 1 hour at 4°C. Supernatants were transferred into prechilled Protein LoBind tubes (Eppendorf), snap-frozen in liquid nitrogen, and stored at −80°C until analysis. Metabolomic profiling was performed by the Metabolomics Facility at the Vienna BioCenter Core Facilities (VBCF; [285]www.viennabiocenter.org/vbcf/metabolomics/). Intracellular metabolites were detected and quantified with hydrophilic interaction liquid chromatography (HILIC) liquid chromatography–tandem MS (LC-MS/MS). Retention times, selected reaction monitoring (SRM) transitions, and optimal collisional energies were determined by authentic standards. All data interpretation was performed using Xcalibur (Thermo Fisher Scientific). In HILIC-MS/MS, an Ultimate 3000 HPLC system coupled to a TSQ Quantiva mass spectrometer (Thermo Fisher Scientific) was used. The HPLC system has been directly coupled to the mass spectrometer via electrospray ionization, operated in SRM mode. One microliter of each sample was directly injected onto a polymeric iHILIC-(P) Classic HPLC column (HILICON, 100 by 2.1 mm; 5 μm) connected with a guard column. A flow rate of 100 μl/min was used, and a linear gradient [A: acetonitrile; B: 10 mM aqueous ammonium bicarbonate, supplemented with medronic acid (0.1 μg/ml)] was applied. The linear 8-min gradient started with 25% B, increasing to 70% B. The following SRM transitions were used for quantitation in the negative ion mode: mass/charge ratio (m/z) 87 to 43 (pyruvate); m/z 89 to 43 (lactate); m/z 124 to 80 (taurine); m/z 662 to 273, m/z 662 to 328, and m/z 662 to 540 (NAD); and m/z 664 to 346, m/z 664 to 397, and m/z 664 to 408 (NADH). The following metabolites were measured for data validation in the positive ion mode: m/z 118 to 72 (valine), m/z 132 to 86 (leucine and isoleucine), m/z 150 to 133 (methionine), and m/z 205 to 188 (tryptophan). Relative concentrations of the metabolites were determined using the area ratio of the spectra normalized by leucine, isoleucine, valine, methionine, and tryptophan. Measurement of 2-OHdG/2-dG ratio in nuclear and mitochondrial DNA Mitochondrial DNA was isolated using the mitochondrial DNA isolation kit (cat. no.ab65321; Abcam) according to the manufacturer’s protocol. Briefly, 5 × 10^6 BMDMs were resuspended in cytosol extraction buffer and homogenized with a Dounce tissue grinder for 100 passes. Nucleus suspension was verified by viewing under a microscope. After the first centrifugation step, the supernatant was transferred into a new tube for mitochondrial DNA isolation, while the pellet (containing nuclei) was subjected to DNA isolation using the QIAamp DNA Mini Kit (cat. no.51304; Qiagen) according to the manufacturer’s instructions. Both nuclear DNA and mitochondrial DNA were quantified using the nanophotometer N120 (IMPLEN). DNA purity was assessed by quantitative polymerase chain reaction using the Luna Universal qPCR Master Mix (cat. no. M3003L; NEB) according to the manufacturer’s protocol. Primers used for the mitochondrial gene MT-ND1 were as follows: forward, 5′-CTA ATC GCC ATA GCC TTC CTA A-3′; reverse, 5′-GTT GTT AAA GGG CGT ATT GGT T-3′. For 18S ribosomal RNA, primers were as follows: forward, 5′-CAC GGA CAG GAT TGA CAG ATT-3′; reverse, 5′-GCC AGA GTC TCG TTC GTT ATC-3′. Quantification of deoxyguanosine (dG) oxidation in DNA was performed after digestion of the DNA into nucleosides with DNA Degradase Plus (cat. no. E2021; Zymo Research). Nucleosides from digested DNA were measured with reversed-phase LC-MS/MS. Analysis was performed by the Metabolomics Facility at the VBCF ([286]www.viennabiocenter.org/vbcf/metabolomics/). Retention times, SRM transitions, and optimal collisional energies were determined by authentic standards. All data interpretation was performed using Xcalibur (Thermo Fisher Scientific). Reversed-phase LC-MS/MS was performed by using an Ultimate 3000 HPLC system (Dionex, Thermo Fisher Scientific). The HPLC system was coupled to a TSQ Altis mass spectrometer (Thermo Fisher Scientific). Authentic standards used for the acquisition of calibration curves and biological samples were acidified to a final concentration of 1% formic acid. One microliter of each sample was injected onto a Kinetex (Phenomenex) C18 column (100 Å, 150 by 2.1 mm) connected with a guard column and using a 4-min-long linear gradient from 98% A (1% acetonitrile and 0.1% formic acid in water) to 60% B (0.1% formic acid in acetonitrile) at a flow rate of 80 μl/min. Using SRM in the positive ion mode, nucleosides were quantified with the SRM transitions m/z 284 to 168 (8-OHdG) and m/z 268 to 152 (dG). Concentrations of dG and 8-OHdG were determined using an external calibration with a dilution series of defined standards of both nucleosides. Oxidation of dG in the samples was calculated by dividing the measured molar concentration of 8-OHdG by the sum of the measured molar concentration of dG and 8-OHdG. Bulk ATAC-seq hMDMs (0.5 × 10^6) were seeded onto 12-cm nontreated tissue culture plates on day 7 of differentiation with or without DMNQ stimulation. H[2]O[2] stimulation was used as the control for general ROS induction. Cells were processed with the ATAC-Seq Kit (cat. no. 53150; Active Motif) according to the manufacturer’s protocol. The quality of the libraries was confirmed on a bioanalyzer to further determine the size distribution. Libraries were sequenced on a NextSeq 2000 PE100. ATAC-seq analysis ATAC-seq data were processed as previously described ([287]74). Reads were aligned against the Illumina iGenome human GRCh38 reference genome. For the generation of summary profile plots and heatmaps, density information (bigwig) for gene regions and surrounding regions [2 kb upstream of the transcription start site (TSS) and 2 kb downstream of the transcription end site (TES)] was plotted using deeptools (version 3.4.3) ([288]https://zenodo.org/records/3965985#.Yh-DPejMJPZ). For differential peak analysis, the package DiffBind (version 3.16.0.) was used ([289]https://bioconductor.org/packages/DiffBind/), and data were normalized using loess fit normalization. The log fold changes between two groups measured for identifying these sites as differentially bound and the multiple testing–corrected false discovery rate (FDR) were calculated by the DESeq2 analysis. MA plots and box plots were generated with dba.plotMA and dba.plotBox functions, respectively. GSEA was performed using the ChIP-Enrich (version 2.30.0; Bioconductor) ([290]https://code.bioconductor.org/browse/chipenrich/) R tool with the hybridenrich method. All enriched pathways were considered significant with an FDR P value of ≤0.05. The built-in gene set used in this analysis is the hallmark gene sets ([291]www.gsea-msigdb.org/gsea/msigdb/human/collections.jsp). The dot plot was generated using ggplot2 (version 3.5.1). Manipulation of the published dataset Previously analyzed and published scRNA-seq ([292]29) and snATAC-seq ([293]30) datasets were used. For the generation of summary profile plots and heatmaps, density information (bigwig) for gene regions and surrounding regions (2 kb upstream of the TSS and 2 kb downstream of the TES) was plotted using deeptools (version 3.4.3) ([294]https://zenodo.org/records/3965985#.Yh-DPejMJPZ). Heatmaps were generated using pheatmap tool (version 1.0.12) with the normalized tag counts per million from the scRNA-seq dataset and clustered by row. The list of genes used for summary profile plots and heatmaps was obtained according to the bulk ATAC-seq data. GSEA was performed using the ChIP-Enrich (version 2.30.0; Bioconductor; [295]https://code.bioconductor.org/browse/chipenrich/) R tool with the hybridenrich method (see the “ATAC-seq analysis” section). Whole-aorta scRNA-seq processing ApoE^−/− mice were fed as described in fig. S6A. Freshly isolated whole aortic cells from a pool of n = 6 mice were fixed with the Chromium Fixed RNA Profiling kit (cat. no. [296]CG000478; 10x Genomics) according to the manufacturer’s instructions and cryoconserved at −80°C until the last sample was processed. The viability of cells was measured with LIVE/DEAD Fixable Scarlet stain (cat. no. [297]L34986; Thermo Fisher Scientific) and measured by flow cytometry. All samples had viability ≥70%, indicating no cytotoxic effect of the STAT5 inhibitor. For quantification, nuclei were stained with a fluorescent nucleic acid staining dye and counted on a Countess 3 FL automated cell counter (Invitrogen). Libraries were prepared by the Next-Generation Sequencing (NGS) Facility of the Medical University of Vienna ([298]https://corefacilities.meduniwien.ac.at/en/our-units/genomics-rna /). Briefly, libraries were prepared using 10x Chromium Fixed RNA kit Mouse Transcriptome (cat. no. BC PN-1000496; 10x Genomics), according to the manufacturer’s instructions. We used the Chromium Next GEM Chip Q Single Cell Kit (cat. no. PN-1000422; 10x Genomics) for the generation of gel beads in emulsion. All libraries were quantified by a Qubit 3.0 Fluorometer using the Qubit dsDNA HS Assay kit (cat. no. [299]Q32854; Thermo Fisher Scientific), and quality was checked using a Bioanalyzer 2100 with High Sensitivity DNA Kit (cat. no. NC1738319; Agilent). Sequencing was performed using an S1 flow cell with a NextSeq 2000 P3 platform (100 cycles; Illumina) into two different runs and two different lanes (four samples per chip) targeting 26,837 and 22,378 reads per cell. Sequencing data were demultiplexed and mapped with Cell Ranger software version 8.0.1 (10x Genomics). The mouse mm10 (Ensembl 98) reference was used for the alignment. scRNA-seq data analysis Cell Ranger outputs (sample_filtered_feature_bc_matrix) were loaded in R (version 4.4.1), preprocessed, and analyzed using Seurat (version 5.1.0) [300]https://satijalab.org/seurat/. Samples were merged into a single Seurat object and was prefiltered as follows: Nuclei with more than 5% mitochondrial transcripts, nFeature_RNA > 200, and nFeature_RNA < 6000 were excluded. The filtered data were normalized and log transformed, and the log-transformed matrix was used for all downstream analyses. Data were integrated using RPCAIntegration batch correction ([301]https://satijalab.org/seurat/articles/seurat5_integration). Initial clustering analysis was performed in Seurat using 20 principal components and a 0.1 resolution. Only clusters expressing CD45 were considered for further analysis. MPCs were identified by subclustering of macrophage and APC/monocyte clusters. Subclusters were identified using 30 principal components, 0.2 resolution, and harmony reduction. Cluster markers were identified using “FindAllMarkers” from the Seurat package. A dot plot of top markers expressed in each MPC cluster was generated using the DotPlot function from Seurat. For the pseudobulk differential expression analysis of foam cell subsets, the package DElegate was used ([302]https://github.com/cancerbits/DElegate). The function findDE() was used with the compare argument “16 weeks vs s5i16w” and DE method “deseq” for each foam cell cluster. The 4-week HFD dataset was excluded for this analysis as it had low numbers (<200 nuclei) within the foam cell clusters. An adjusted P value cutoff of 0.05 and 1 ≤ log[2] fold change ≥ 1 were applied for the identification of differentially expressed genes. t-Distributed stochastic neighbor embedding (t-SNE) was used for data visualization in two dimensions. DittoBarPlot from the Dittoseq package ([303]www.bioconductor.org/packages/release/bioc/html/dittoSeq.html) was generated to show the percent composition of cells from each cluster. GSEA analysis was performed using the fgsea package ([304]https://bioconductor.org/packages/release/bioc/html/fgsea.html) with the hallmark gene sets from the differentially expressed genes obtained from pseudobulk analysis. Violin plot and heatmaps were generated using the VlnPlot and DoHeatmap function of the Seurat package. The circular bar plot and dot plot of GSEA analysis were generated using ggplot2 (version 3.5.1). The volcano plot was generated using EnhancedVolcano (version 1.24.0; [305]https://bioconductor.org/packages/release/bioc/html/EnhancedVolcan o.html). Trajectory analysis was performed using the slingshot trajectory inference method, implemented in R through the Bioconductor package ([306]https://bioconductor.org/books/3.14/OSCA.advanced/trajectory-anal ysis.html) ([307]55). Before trajectory inference, principal components analysis was performed on the normalized single-cell dataset. Pseudotime values were assigned to each cell, indicating its relative position along the inferred trajectory. Comparisons between categorical data (Trem2^hiGpnmb^hi cells within different pseudotime intervals and treatments) were calculated with Fisher’s exact test to account for small sample sizes. Quantification and statistical analysis Data are shown as the means ± SD. Each dot represents one independent biological replicate. At least n = 4 biological samples were used in each experiment. For statistical analysis between two groups, unpaired Student’s t test was used. For the histological analysis of plaques, a paired t test was used to compare measurements within the same samples, accounting for intrasample variability. One-way analysis of variance (ANOVA) with Tukey’s multiple comparison test was used for multiple comparisons. For the comparison of data between Stat5^−/− BMDMs and their wild-type counterpart and for comparisons of different cell subsets from in vivo flow cytometry mouse data, we used two-way ANOVA with Tukey’s multiple comparison test for intragroup analysis and two-way ANOVA with Sidak’s multiple comparison test for intergroup analysis. For the colocalization analysis of aortic plaques, Student’s t test for Pearson correlation was used. For trajectory analysis, Fisher’s exact test was used. All statistical analyses were performed using GraphPad Prism software (GraphPad; version 8.0.2). In all tests, a P value <0.05 was considered significant. Data acquisition and analysis were performed in a blinded manner. For immunoblots, immunofluorescence, and FACS histogram, the reported images are representative of n ≥ 3 independent experiments. n samples from hMDMs were derived from blood donated by different individuals, with each experiment conducted using cells from distinct blood donors. For in vivo mouse data, exclusion of n samples was performed, if necessary, because of poor sample quality or loss of material. Acknowledgments