The IRG1–itaconate axis protects from cholesterol-induced inflammation and atherosclerosis

To investigate the expression of IRG1 in human atherosclerosis, we performed single-cell RNA sequencing (scRNA-seq) of human coronary fibroatheromas (FA, n = 2) and paired coronary adaptive intimal thickening (AIT) from the same patients (n = 2, Fig. 1A). IRG1 transcript levels were enriched in FA compared to nondiseased AIT, and its expression was predominantly localized to immune cells (Fig. 1A). IRG1 was most highly expressed by myeloid cells, which amounted to over 20% of all immune cells within the immune cell compartment of coronary plaques (SI Appendix, Fig. S1 A and B), and particularly in monocytes and macrophages (Fig. 1 B–D). This was confirmed by immunostaining of human coronary arteries (Fig. 1E and SI Appendix, Fig. S1C) and carotid artery plaques (SI Appendix, Fig. S1D), which showed colocalization of IRG1 and the monocyte/macrophage marker CD68.

To further investigate how Irg1 expression is modulated during atherogenesis, we fed Ldlr^–/– mice a western diet (WD) for 8, 12, or 16 wk to simulate early, moderate and advanced atherosclerosis. We performed scRNA-seq of CD45⁺ cells isolated from the aortic arch (an area of plaque accumulation) of mice with advanced atherosclerosis (16 wk WD), and detected all major immune cell types (SI Appendix, Fig. S1E). We confirmed that Irg1 was highly expressed in myeloid cells within mouse atherosclerotic tissue, particularly monocytes and neutrophils (Fig. 1F), but not adaptive immune cells. Interestingly, although Irg1 transcript levels were high in monocytes, which are associated with active infiltration of the arterial intima and acute inflammation, it was lower in longer-lived plaque macrophages (Fig. 1F), indicating that its expression becomes limited. To confirm our transcriptomic analyses, we performed IRG1 immunostaining of cross-sections of the aortic root of Ldlr^–/– mice. Within the plaque, IRG1 staining strongly colocalized with the monocyte/macrophage marker CD68, and to a lesser extent with the neutrophil marker Ly6G (Fig. 1G). Notably, we observed that the proportion of IRG1⁺ cells within the plaque was highest in Ldlr^–/– mice with early atherosclerosis (8 wk WD), but declined with continued WD feeding, suggesting that this anti-inflammatory mechanism wanes with plaque progression (Fig. 1H). To investigate the transcriptomic changes associated with expression of Irg1, we built a differential gene expression matrix that compared Irg1-expressing (Irg1⁺) and nonexpressing (Irg1^–) cells within each myeloid subcluster from our scRNA-seq analysis of mouse plaques (Fig. 1I). Using this approach, we observed pronounced transcriptional differences associated with Irg1 expression within the monocyte, macrophage and neutrophil clusters (Fig. 1I). Ingenuity Pathway Analysis of the 526 genes that were up- and 707 genes down-regulated in Irg1 expressing vs. nonexpressing myeloid cells (P-adj < 0.05, −0.5 < Log₂ fold change > 0.5) revealed a transcriptional pattern indicative of increased inflammatory activity and mitochondrial dysfunction in Irg1⁺ cells (Fig. 1J). Notably, Irg1⁺ cells showed down-regulated expression of Apoe, Trem2 and Lpl, which are genes characteristic of the lipid-handling properties of macrophage foam cells (Fig. 1J). Additionally, endogenous ligands predicted to be upstream of these transcriptional changes included proatherosclerotic lipid mediators (24) generated through mitochondrial metabolism (Fig. 1J). These data implicate Irg1 in metabolic reprograming of myeloid cells within the plaque.

To test how atherogenic lipoproteins alter the metabolic state of macrophages, we treated mouse bone marrow-derived macrophages (BMDMs), with aggregated (ag)LDL or oxLDL, and performed targeted liquid chromatography and tandem mass spectrometry (LC–MS) to identify polar intracellular metabolites. We observed significant differences in the accumulation of 29 metabolites within agLDL- or oxLDL-treated cells, including itaconate (Fig. 2A). Using a neat standard curve in LC–MS, we confirmed that BMDM treated with agLDL and oxLDL accumulated itaconate intracellularly (Fig. 2B), and observed twofold and eightfold increases, respectively, in Irg1 mRNA by quantitative RT-PCR (Fig. 2C). Similar findings were observed in human THP1 macrophages and mouse neutrophils (SI Appendix, Fig. S2 A and B).

To test the role of the IRG1–itaconate axis in limiting atherosclerotic inflammation, we induced hypercholesterolemia in wild-type (WT) and Irg1^–/– mice using an adeno-associated virus (AAV) vector bearing gain-of-function murine PCSK9-D377Y, coupled with WD feeding as we described (25, 26). Immunostaining of cross-sections of the aortic root taken after 12 wk of WD feeding confirmed the absence of IRG1 expression in plaques of Irg1^–/– mice (Fig. 2D), and revealed greater plaque burden throughout the length of the aortic root compared to WT mice (Fig. 2E), in the absence of differences in plasma cholesterol levels or body weight (SI Appendix, Fig. S2C). Furthermore, Irg1^–/– plaques showed higher levels of immunostaining for the macrophage marker CD68 (Fig. 2F) and BODIPY-staining of neutral lipid (Fig. 2G), even after correcting for plaque size. Together, these data suggest increased accumulation of macrophage foam cells and plaque progression in Irg1-deficient mice.

To examine how Irg1-deficiency alters the transcriptional signatures and dynamics of immune cell populations in plaque, we performed scRNA-seq of CD45⁺ cells isolated from the aortic arches of WT or Irg1^–/– atherosclerotic mice. This approach yielded 6,022 single leukocytes belonging to 16 immune cell types as determined by Louvain clustering (Fig. 2H). Of the eight myeloid clusters identified, Lyve1⁺ and Trem2⁺ macrophages constituted the two most abundant subpopulations captured, amounting to ~50% of innate immune cells found within atherosclerotic tissue irrespective of genotype. Lyve1 is recognized as a marker of anti-inflammatory, resident macrophages (27), while Trem2 has been identified as a regulator of foamy macrophage differentiation specifically in the context of increased atherosclerotic burden (28). In the absence of Irg1, Lyve1⁺ macrophages appeared decreased and Trem2⁺ macrophages were significantly more abundant compared to WT plaques (P = 7 × 10^–3; Fig. 2I). Differential expression analysis revealed that the majority of gene expression changes detected in Irg1^–/– plaques were found in Lyve1⁺ macrophages or neutrophils (Fig. 2J). Canonical pathway analyses revealed that compared to their WT counterparts, Irg1^–/– Lyve1⁺ macrophages displayed a proinflammatory transcriptional program, as evidenced by the emergence of terms associated with T cell activation, such as Granzyme A and IL-12 signaling, and the concomitant downregulation of oxidative phosphorylation and NRF2-mediated signaling (Fig. 2J). Using CellChat, a computational tool that infers cell–cell interactions based on gene expression of receptor–ligand pairs from scRNA-seq datasets, we found that WT Lyve1⁺ macrophages engaged in immune dampening IL-10 signaling with multiple lymphoid and myeloid cell clusters, but this signaling was absent in Irg1^–/– Lyve1⁺ macrophages (P = 2.55 × 10^–5; Fig. 2K) (29). These data suggest that the loss of Irg1 enhances the accumulation lipid-loaded macrophages and promotes a proinflammatory transcriptional program in resident macrophages, which may accelerate atherosclerosis.

To confirm that Irg1 expression within myeloid cells protects against atherosclerosis development, we reconstituted lethally irradiated Ldlr^–/– mice with bone marrow from Irg1^–/– (Irg1^–/– → Ldlr^–/–) or WT (WT → Ldlr^–/–) mice, and after 6 wk of recovery, fed them a WD for 12 wk. Similar to mice with global Irg1-deficiency, male and female chimeric Irg1^–/– → Ldlr^–/– mice showed greater plaque burden throughout the length of the aortic root compared to WT → Ldlr^–/– mice of the same sex (Fig. 3A). This was accompanied by higher plaque macrophage and lipid content in Irg1^–/– → Ldlr^–/– mice, even after correction for plaque size (Fig. 3B), despite similar body weights and serum cholesterol levels to WT → Ldlr^–/– mice (SI Appendix, Fig. S3). As previous studies reported that Irg1-deficient mice have increased hepatic lipid accumulation in models of fatty liver disease (30), we investigated whether Irg1 affects macrophage cholesterol handling and foam cell formation in vitro. Irg1^–/– BMDMs treated with oxLDL for 48 h showed a marked increase in intracellular lipid content compared to similarly treated WT BMDMs, as assessed by BODIPY staining of neutral lipids (Fig. 3C). These data suggest an important role for Irg1 in restricting macrophage cholesterol accumulation and foam cell formation.

To assess whether the increased accumulation of lipid-laden macrophages in Irg1^–/– → Ldlr^–/– plaques was associated with markers of plaque vulnerability, we quantified smooth muscle cell (SMC) abundance via immunostaining of ACTA2 and measured necrotic core area. We observed a trend toward reduced plaque SMC content and a significant increase in the ratio of macrophage-to-SMC in Irg1^–/– → Ldlr^–/– plaques (Fig. 3 D and E), a feature found in inflammatory plaques. However, we observed no difference in necrotic core area in Irg1^–/– → Ldlr^–/– mice compared to WT → Ldlr^–/– mice (Fig. 3F). Notably, we also observed a marked increase in immunostaining for IL-1β in plaques of Irg1^–/– → Ldlr^–/– mice compared to WT → Ldlr^–/– mice (Fig. 3G). Irg1-deficiency has been reported to increase NLRP3-inflammasome activation and maturation of pro-IL-1β during infection (17), but whether it regulates inflammasome activation in atherosclerosis remains unclear. To test this, we first primed WT and Irg^–/– BMDMs with LPS, and then exposed them to CC, an atherogenic activator of inflammasome assembly and caspase-1 activation (5, 6). We found that CC-induced IL-1β secretion was potentiated in Irg^–/– BMDM compared to WT BMDM (Fig. 3H), consistent with increased inflammasome activation.

Our cell–cell communication analysis (Fig. 2K) indicated important anti-inflammatory cross-talk between Lyve1⁺ macrophages and neutrophils that was absent in Irg1-deficient mice. Neutrophils contribute to inflammation in the plaque through the production of NETs (7), ROS, and proteases, which are triggered by CC (31, 32). Consistent with a role for Irg1 and itaconate in restricting neutrophil proatherogenic responses, we measured a significant increase in NET accumulation in plaques of Irg1^–/– mice compared to WT mice, as assessed by immunostaining for myeloperoxidase (MPO) and citrullinated histone H3 (CitH3, Fig. 4A). To test whether this effect was cell autonomous, we isolated neutrophils from the bone marrow of WT and Irg1^–/– mice, stimulated with PMA (100 nM) or vehicle (DMSO), and quantified NET formation via user-trained deep-learning image analysis in the IncuCyte platform. In PMA-stimulated neutrophils, NETosis was potentiated in absence of Irg1 (Fig. 4B). Further, compared to WT neutrophils, Irg1^–/– neutrophils elaborated more NET DNA in response to CC (Fig. 4C). Notably, NETs can activate surrounding macrophages by priming the NLRP3-inflammasome, leading to increased expression of pro-IL1b and inflammasome components (e.g., Pycard, Casp1 and Nlrp3). To test whether Irg1 regulates this neutrophil-macrophage cross talk, we collected NET DNA from neutrophils exposed to CC and applied it on WT and Irg1^–/– BMDMs (Fig. 4D; steps 1 and 2). We observed higher levels of ll1b and Pycard mRNA in Irg1^–/– compared to WT BMDM, indicating increased NLRP3 inflammasome priming (Fig. 4E), and exposure of NET-primed BMDMs to CC (step 3) resulted in increased IL-1β secretion from Irg1^–/– BMDM compared to WT (Fig. 4F). Collectively, these data highlight the role of Irg1 in tempering the inflammatory cross talk by neutrophils and macrophages that regulates inflammasome priming and activation in atherosclerosis.

Our data suggest that levels of Irg1, and thus itaconate, decrease with plaque progression in mice (Fig. 1H), indicating that itaconate supplementation could have therapeutic potential in atherosclerosis. To investigate this, we tested the ability of the cell-permeable itaconate derivative 4-OI to restrict neutrophil NETosis and macrophage NLRP3-inflammasome activation. Treatment with 4-OI decreased CC-induced NET formation in vitro by ~50% (Fig. 5A). Furthermore, in LPS-primed BMDMs, 4-OI abrogated inflammasome activation and IL-1β secretion in response to CC or ATP (Fig. 5B). Given these beneficial effects, we conducted a 4-wk trial of 4-OI or vehicle administration (intraperitoneal injection, 3x/week) in Ldlr^–/– mice with advanced, complex atherosclerotic plaques induced by 30-wk WD feeding (Fig. 5C). Administration of 4-OI induced favorable changes in plaque morphology, characterized by marked reductions in plaque macrophage content (Fig. 5D) and necrotic core area (Fig. 5E). Moreover, 4-OI reduced the proportion of complex, advanced plaques (Fig. 5F; stage V, Stary classification), in the absence of changes in plaque size (Fig. 5D), consistent with beneficial plaque remodeling in treated mice. Notably, 4-OI treatment reduced IL-1β immunostaining in plaques (Fig. 5G), consistent with our in vitro findings that 4-OI reduces NLRP3-inflammasome activation by CC.

To assess the therapeutic potential of 4-OI in humans, we leveraged a validated systems-immunology approach to investigate CVD-related inflammation ex vivo (33). We isolated PBMCs from healthy donors (n = 6) and incubated them with plasma from patients with carotid atherosclerosis (CVD plasma) with or without pretreatment with 4-OI (250 µM; n = 4 for all conditions, Fig. 6A). To assess the inflammatory signaling events that occur in response to treatment with CVD plasma, we used phospho-cytometry by time-of-flight (phospho-CyTOF) to resolve phospho-signaling at the single-cell level. Using viSNE, we identified 11 major immune cell populations based on canonical marker expression patterns [CD1c⁺ dendritic cells (DC), CD14⁺ monocytes, CD16⁺ monocytes, CD14/16⁺ monocytes, plasma DCs, B cells, CD4⁺ and CD8⁺ T cells, exhausted T cells, natural killer (NK) cells, and basophils] (Fig. 6B), and then quantified the phosphorylation of seven intracellular proteins–AKT serine/threonine kinase 1 (AKT), nuclear factor of κ light chain polypeptide gene enhancer in B cells inhibitor-α (IκBα), mitogen-activated protein (MAP) kinase-activated protein kinase 2 (MAPKAPK2), ribosomal protein S6 (pS6), phosphorylated cAMP-response element binding protein (pCREB), and extracellular signal-regulated kinase 1 and 2 (pERK1/2) across immune populations and treatments (Fig. 6 C and D). Because of the restricted expression of Irg1 in myeloid cells, we focused our analysis on monocyte populations, where we observed a relative increase in AKT, MAPKAPK2, CREB, ERK1/2 and S6 kinase signaling upon treatment with CVD plasma, which was in turn, attenuated by 4-OI pretreatment (Fig. 6D). The significant increase in phospho-signaling following CVD plasma treatment could be visualized by t-statistics, with phosphorylated proteins showing consistent upregulation in all three monocyte subtypes (Fig. 6 E, Top). 4-OI supplementation reverted this pattern, with significant decreases in pCREB in CD14/16⁺ and CD14⁺ monocytes, and pS6 in CD16⁺ monocytes, and a trending decrease for all phosphoproteins across these cell populations (Fig. 6 E, Bottom).

To determine how 4-OI affects the transcriptional signature of human immune cells exposed to CVD plasma, we next performed scRNA-seq of treated PBMCs. CVD plasma induced significant increases in abundance of both CD163⁺ and PLIN2⁺ macrophage subpopulations (P = 3.48 × 10^–6 and 1.15 × 10^–5, respectively) at the expense of the CD14⁺ monocyte cluster (P = 3.48 × 10^–6), evidence of a transition from a naïve to a mature state following stimulation (Fig. 6F and SI Appendix, Fig. S4A). IRG1 expression was detected across all myeloid cell populations, and was highest in monocytes and macrophages (Fig. 6F). While CVD plasma stimulation decreased IRG1 expression, this was prevented by 4-OI pretreatment, particularly in CD14⁺ monocytes (Fig. 6G and SI Appendix, Fig. S4B). No changes were detected in the abundance of T cell subpopulations, with the exception of double-negative T cells marked by expression of the chemokine gene CXCL3 (34), that amounted to less than 1% of that group (SI Appendix, Fig. S4C), indicating that CVD plasma stimulation primarily affects innate immune cell populations. Of those, CD163⁺ and PLIN2⁺ macrophages displayed the highest number of differentially expressed genes (DEG) between the CVD plasma and CVD plasma + 4-OI conditions (Fig. 6H). PLIN2⁺ macrophages, which highly express the PLIN2 gene encoding a protein that coats intracellular lipid droplets, have been implicated with both lipid loading and inflammation within human atherosclerotic plaques (35), whereas CD163⁺ highly express CD163, a scavenger receptor associated with downstream anti-inflammatory responses through IL-10 signaling (36). Canonical pathway analyses of genes differentially expressed in these macrophage populations in the presence of 4-OI showed an upregulation of NRF2 and IL-10 signaling, and a downregulation of genes that contribute to neutrophil degranulation (Fig. 6 H and I). Consistent with this, in silico modeling of cell–cell interactions predicted that the bulk of IL-10 signaling took place through CD163⁺ macrophages, and this was increased by 4-OI pretreatment (Fig. 6J). These data show that IRG1 expression by myeloid cells declines in the CVD environment and that 4-OI supplementation can up-regulate antioxidant and anti-inflammatory transcriptional programs that protect from atherosclerosis.

Two patients with end-stage heart failure undergoing heart transplantation were enrolled in a clinical study approved by the Institutional Review Board (IRB) (IRB no. 21-00429). Patient characteristics are detailed in SI Appendix, Table S1. Inclusion criteria were patients over 18 y old undergoing heart transplantation at NYU Langone Health for either ischemic or nonischemic heart failure. Exclusion criteria were active infection, autoimmune diseases, peripheral arterial occlusive disease with rest pain, and renal dialysis. Sections of the same coronary arteries were stained with hematoxylin and eosin (H&E) and classified as AIT and FA according to established criteria, as described (47, 48).

Single-cell suspensions were prepared from the left anterior descending and right coronary artery immediately after heart removal. Briefly, plaques were washed extensively, digested, and sequentially filtered before removal of dead cells and automated counting. scRNA-seq libraries were prepared using the Chromium Next GEM Single Cell 3’ GEM, Library & Gel Bead Kit v3.1, Chip G Single Cell Kit and Dual Index Kit TT Set A (10× Genomics; PN-1000268, PN-1000127, PN-100215). DNA library quality was measured by Qubit dsDNA, and sequenced on an Illumina NovaSeq X+ sequencer as described (49).

Eight patients undergoing carotid endarterectomy (CEA) were enrolled in a clinical study approved by the IRB of NYU Langone Health (IRB no. 21-00429) and the Icahn School of Medicine at Mount Sinai (IRB no. 11-01427). Patient characteristics and inclusion and exclusion criteria are described elsewhere (47, 49). Paraffin-embedded tissues were processed as described (47). Samples were costained for IRG1 and CD68 (SI Appendix, Table S2) and counterstained for nuclei (DAPI, Invitrogen D1306). Autofluorescence was quenched with TrueBlack Lipofuscin Autofluorescence Quencher (Biotium, Cat. #23007). Images were acquired on a Phenoimage HT (Akoya Biosciences), and manually outlined before spectral unmixing and quantitative analysis, as described (47).

All experimental procedures were approved by the NYU Grossman School of Medicine’s Institutional Animal Care and Use Committee and conducted in accordance with the US Department of Agriculture Animal Welfare Act and the US Public Health Service Policy on Humane Care and Use of Laboratory Animals. Atherosclerosis was induced in male or female Ldlr^−/− mice (Jackson Lab: #002077), by WD (21% [wt/wt] fat, 40% fat kcal, 0.3% cholesterol; Dyets 101977GI) feeding for 12 to 16 wk. Atherosclerosis was induced in 8-wk-old Irg1^–/–or control C57BL/6NJ (Jackson Lab: #029340, #005304) male mice by delivery of D377Y-mPCSK9 by AAV (Penn Vector core) and WD feeding for 12 to 16 wk as described (25). To evaluate the effects of 4-OI on advanced atherosclerotic plaques, 8-wk-old Ldlr^−/− male mice were fed WD for 30 wk to induce atherosclerosis and treated 4-OI (25 mg/kg i.p. in corn oil; MedChem Express, HY-112675) or vehicle (7.5% DMSO in corn oil) 3x/week for 4 wk (n = 5 mice/group). For transplant studies, bone marrow cells from male or female Irg1^–/– or C57BL/6NJ were retro-orbitally transferred into lethally irradiated (2 × 4.5 Gy, administered 3 h apart) male or female Ldlr^−/− mice (n = 3 mice/genotype/sex donor, n = 13 to 15 mice/group/sex recipient), and mice were kept on antibiotic-containing water for 5 wk. After one additional week of recovery, mice were placed on WD for 12 wk. Bone marrow chimerism was assessed by PCR (primers used are listed in SI Appendix, Table S3).

Mice with cholesterol <525 mg/dL were excluded from the study. Heart, aorta, and blood were collected and processed as described (25). Morphometric analysis of plaque burden, necrotic area, and plaque complexity was performed using ImageJ software (https://fiji.sc/) as described (25). To assess plaque composition, immunohistochemical and histochemical staining was performed to evaluate neutral lipid content (BODIPY), NETosis (Citrullinated H3, MPO), IRG1, CD68, Ly6G, and ACTA2 as described (25). Aortic arches were digested, filtered through a 70-µm strainer, stained for viability and CD45 expression, and hash-tagged. Live CD45⁺ cells were purified by fluorescence-activated cell sorting using a BD FACSAria II instrument equipped with a 100 µm nozzle, and specimens were pooled by the experimental group (n = 5/group), and scRNA-seq performed as described above for human coronary plaques and in SI Appendix (25).

scRNA sequencing data alignment, preprocessing, filtering, normalization, dimensionality reduction, and visualization followed protocols established in previous human (49) and mouse (25) studies. Briefly, in human coronary plaque, Louvain clustering revealed 23 distinct cell populations. Subsequent subclustering of myeloid cells identified nine distinct groups, with a total of 3,556 myeloid cells sequenced and annotated. For CD45+ cells from aortic arch plaques in Ldlr^–/– mice, 21 distinct populations were identified using Louvain clustering. Subclustering on monocytes, macrophages, DCs, and neutrophils (myeloid cells) identified 8,669 myeloid cells. For immune cells from aortic arch plaques of WT and Irg1^–/– mice, 20 distinct populations were identified by Louvain clustering, with 6,022 cells sequenced. For human PBMCs, initial clustering with the Louvain algorithm identified 20 distinct major populations, with a total of 41,407 cells sequenced. Further subclustering identified 23 distinct subpopulations of myeloid cells and T cells. Immune cells within the atherosclerotic plaque were annotated based on expression of canonical gene markers identified for human (49) and mouse (50) atherosclerosis. Human PBMCs were annotated using established gene markers (33). DEG between cell types were identified using the presto package (v1.0.0) and pathway enrichment analysis was performed using Ingenuity Pathway Analysis software (Qiagen). Cell–cell communication was examined using the CellChat package (v1.6.1) for the entire dataset (29).

Mouse BMDMs were prepared as described (25) and treated with agLDL (50 µg/mL), oxLDL (50 µg/mL; Thermo Fisher L34357) or vehicle for 8 h, or primed with NET-DNA (500 ng/mL, 4 h). To assess the effect of 4-OI on NLRP3-activation, WT BMDMs were primed with LPS (10 ng/mL, 3 h), treated with 4-OI (250 µM) or vehicle for 2 h, and then activated with CC (500 µg/mL, 8 h) or ATP (5 mM, 2 h). Supernatant was collected and IL-1β was measured using BD CBA Mouse IL-1β Enhanced Sensitivity Flex Set (BD, 562278). To induce foam cell formation, BMDMs were treated with oxLDL (50 µg/mL, 48 h), stained with BODIPY, and intracellular neutral lipid was quantified from eight fields of view in triplicate and normalized to cell number. All experiments performed with n = 3 mice/group.

Mouse neutrophils were isolated from bone marrow of WT and Irg1^–/– mice using the Neutrophil Isolation Kit (Mitenyi Biotec, 130-097-658). NETosis was assessed using the Incucyte Live-Cell Analysis System (Sartorius) in neutrophils treated with PMA (100 nM), CC (500 µg/mL), or vehicle, and stained with Cytotox Green reagent DNA dye (250 nM, Sartorius). To assess the effect of 4-OI, neutrophils were pretreated with 4-OI (250 µM) for 1 h prior to stimulation. NET-DNA extrusion was measured from neutrophils treated with CC (500 µg/mL, 4 h) as described (7). To detect NETosis in situ, plaques were stained for Histone H3 citrulline R2+R8+R17, MPO, and counterstained with DAPI (SI Appendix, Table S2).

Targeted metabolomics was performed on BMDMs treated with agLDL (50 µg/mL), oxLDL (50 µg/mL), or vehicle for 24 h by the NYU Metabolomics Core Resource Laboratory, as described (51). Briefly, metabolites were extracted using 80% methanol and a metabolomics amino acid mix standard (Cambridge Isotope Laboratories Inc.), homogenized, centrifuged, dried by speedvac (Thermo Fisher Scientific), and reconstituted in Optima LC/MS grade water (Thermo Fisher Scientific). Samples were sonicated and subjected to LC/MS analysis.

Total RNA was isolated, reverse transcribed, and quantitative PCR performed as described (5). Fold change in mRNA expression was calculated using the comparative cycle method (2^−ΔΔCt) normalized to housekeeping genes. Primers used are listed in SI Appendix, Table S3.

PBMCs were isolated from whole blood of six healthy donors (patient characteristics are detailed in SI Appendix, Table S4) from the New York Blood Center, within 2 h of receipt, as described (49) and stored in liquid N₂ until experiment date. To isolate CVD plasma, fasting peripheral venous blood was preoperatively collected from 4 CVD patients undergoing CEA and enrolled in an IRB-approved clinical study (NYU Langone Health-IRB no. 21-00429 and the Icahn School of Medicine at Mount Sinai-IRB no. 11-01427), as described in ref. 49. Patient characteristics are detailed in SI Appendix, Table S5. PBMCs (RPMI, 1% FBS, 1% P/S) were pretreated with 4-OI (250 µM, 1 h), and stimulated with CVD plasma from 4 CVD patients at a final concentration of 20% as described (33) for 15 min for phospho-CyTOF analysis, or for 4 h for scRNA-seq.

PBMCs were labeled with Rh103 viability marker, fixed with formaldehyde, barcoded using the Cell-ID 20-Plex Pd Barcoding Kit (Fluidigm) and labeled with surface antibodies to identify immune subsets, followed by methanol permeabilization and staining with intracellular phosphoprotein antibodies (SI Appendix, Table S6), as described (33). Samples were stored in formaldehyde with an iridium intercalator until acquisition using a CyTOF2 Helios mass cytometer (Standard BioTools–FKA Fluidigm). Data normalization and sample processing were performed using Helios software, with barcodes deconvoluted and doublets filtered as described in ref. 49.

Statistical significance between two groups of independent biological replicates was evaluated with Student’s t test; or visually represented using t-statistics. One-way ANOVA was performed when comparing three groups or more for univariate comparisons. Tukey’s post hoc multiple comparison test (MCT) was used to compare all groups if the ANOVA revealed significant group differences. Two-way ANOVA was used when comparing two groups or more for bivariate analyses or to assess interaction between two variables, and Sidak’s post hoc MCT was used for comparisons between groups selected a priori. Statistical analyses were performed using GraphPad Prism software or R studio. Threshold for statistical significance was P ≤ 0.05. All quantitative data are presented as mean ± SEM.

Y.C., F.K.B., C.G., and K.J.M. designed research; Y.C., F.K.B., L.A., P.S., M. Gourvest, D.D., T.Z., K.M.W., R.V.I., P.M.S., S.D.D., B.F.S., and C.v.S. performed research; A.M.S. contributed new reagents/analytic tools; Y.C., F.K.B., J.G.B.D., A.A.C.N., L.A., P.S., M. Gourvest, D.D., M. Gildea, R.K., T.Z., K.M.W., R.V.I., P.M.S., E.A.F., C.v.S., and C.G. analyzed data; and Y.C., F.K.B., and K.J.M. wrote the paper.

K.J.M. is on the scientific advisory Board of Beren Therapeutics and Bitterroot Bio. K.J.M. and A.M.S. have patents and patent applications through NYU Grossman School of Medicine that have been submitted/published and that are not related to the work detailed in this manuscript. The other authors declare no conflict of interest.