Immunoregulatory gene GIMAP6 suppresses lethal atherosclerotic vasculopathy and ischemic heart failure

SUMMARY

Controlling hyperlipidemia has reduced but not eliminated atherosclerotic cardiovascular disease (ASCVD) as a major cause of human mortality. Here, we report that loss of the immunoregulatory gene GTPase of immunity-associated protein 6 (Gimap6) in mice causes an inflammatory vasculopathy and accelerated atherosclerosis in the absence of hyperlipidemia. These vascular abnormalities cause progressive cardiac ischemia, myocardial infarction, and heart failure, culminating in early death. GIMAP6 loss precipitates early endothelial barrier failure and a mitochondrial ROS-PPARγ-CD36 lipid-influx cascade that lowers the threshold for foam-cell formation and induces IL-6/MCP-1/TNF-mediated inflammation. Human hypomorphic GIMAP6 variants, including the African ancestry-enriched V65I allele, were associated with altered monocyte redox balance, increased CD36-mediated lipid uptake, and early ASCVD. Thus, GIMAP6 is a crucial safeguard for vascular integrity that restrains endothelial dysfunction, macrophage foam cell generation, and inflammation that can cause lethal ASCVD independently of systemic hyperlipidemia.

Graphical Abstract

In Brief

By destabilizing the endothelium and promoting macrophage lipid loading and inflammation, GIMAP6 deficiency drives atherosclerosis and ischemic heart disease through a mechanism that operates without hyperlipidemia, yet hyperlipidemia significantly exacerbates disease progression

INTRODUCTION

Despite advances in lipid-lowering therapy, atherosclerotic cardiovascular disease (ASCVD) remains a leading cause of death worldwide, with many events occurring in individuals with normal blood lipid or low-density lipoprotein (LDL) levels due to non-lipid vascular risk factors.^1-3 Beyond lipid burden, ASCVD is driven by vessel-wall vulnerability, where endothelial barrier dysfunction, local vascular inflammation, and macrophage foam-cell formation converge to promote atherogenesis.⁴ Thus, intrinsic cellular safeguards must exist within the arterial wall to restrain these non-lipid drivers, but their molecular identity remains incompletely understood.

Monogenic inborn errors of immunity (IEIs) offer a powerful entry point for uncovering such safeguards. Although IEIs are typically recognized in rare individuals with overt infection susceptibility, immune dysregulation, or autoinflammation, milder monoallelic or hypomorphic variation in the same genes can shape common complex diseases.^5-8 Genome-wide association studies (GWAS) have identified several IEI genes with cardiovascular traits,^8,9 but few have been mechanistically connected to vessel-wall pathology, endothelial barrier integrity, or arterial inflammation in ASCVD.

GTPase of immunity-associated protein 6 (GIMAP6) belongs to the GIMAP family of immune-associated GTPases.^10,11 GIMAP proteins are enriched in immune cells and regulate lymphocyte survival, autophagy, redox homeostasis, and host defense.^10-16 Recent work on GIMAP5 deficiency has revealed roles for GIMAP proteins in lipid homeostasis.¹⁷ Human GIMAP6 deficiency causes an IEI characterized by infections, lymphoproliferation, autoimmunity, vasculitis, and cerebral ischemic lesions.^15,16 Gimap6-deficient mice also show immune abnormalities and premature death, previously attributed to microangiopathic kidney disease.¹⁵ However, whether GIMAP6 has any role within the vessel wall to regulate endothelial barrier failure, mitochondrial redox balance, and inflammatory cytokine signaling during atherogenesis remains unknown.

Here, we identify GIMAP6 as a hyperlipidemia-independent safeguard of the vessel wall and describe cellular pathways through which its loss promotes endothelial barrier dysfunction, foam-cell activation, and premature ASCVD in mice and humans.

RESULTS

Gimap6 deficiency in mice drives lethal spontaneous atherosclerosis and heart failure independent of hyperlipidemia

Gimap6^−/− mice on the inbred C57BL/6 background showed fully penetrant premature death, with median survival of 31 weeks in males and 21 weeks in females; heterozygous mice including breeders showed modestly reduced survival (Figures 1A and 1B). GIMAP6 protein expression correlated with gene dosage (Figure S1A). Gimap6^−/− mice appeared initially healthy but lost weight by ~20 weeks, followed by rapid deterioration and death within 24 hours of overt decline (Figure S1B).

Figure 1. Gimap6 deficiency causes premature death due to myocardial infarction and atherosclerosis.

(A) Kaplan-Meier survival curves for standard-fat diet (SFD)-fed Gimap6^−/−, Gimap6^+/−, and Gimap6^+/+ mice (males/females, n = 67/60, 41/53, and 35/47, respectively).

(B) Age at death of Gimap6^−/− mice from (A).

(C) Gross cardiac morphology of 20-week-old male Gimap6^+/+ (WT) and Gimap6^−/− mice. Scale bars, 0.5 cm.

(D) TTC-stained heart sections from WT and moribund Gimap6^−/− mice. Arrows mark infarcted regions. Scale bars, 0.5 cm.

(E, F) Plasma cardiac troponin I over time (E: WT, n = 11; Gimap6^−/−, n = 13) and at moribund/age-matched endpoint (F: WT males/females, n = 6/7; Gimap6^−/− males/females, n = 10/12).

(G) Masson’s trichrome-stained heart sections from WT and Gimap6^−/− mice. Asterisks mark fibrosis; arrows indicate inflammatory infiltrates (green) and vascular abnormalities (yellow). RV, right ventricle; LV, left ventricle. Scale bars, 2.5 mm (black) and 200 μm (white).

(H–J) M-mode echocardiograms (H), longitudinal ejection fraction (I), and QTc/QRS duration measurements (J) in WT and Gimap6^−/− mice (n = 3).

(K) Telemetry ECG recordings from WT and Gimap6^−/− mice (n = 3).

(L) Color Doppler imaging of the aortic arch. Dotted boxes indicate turbulent flow (n = 3).

(M) En face ORO staining and plaque quantification in 20-week-old male WT and Gimap6^−/− mice fed SFD (each point = one mouse, n = 5). Scale bars, 0.5 cm.

Statistics: log-rank test (A); Mann-Whitney test (B, F, M); two-way ANOVA (I, J). See also Figure S1.

Unless otherwise indicated, data are mean ± SD. Statistical significance is denoted as ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Necropsy revealed progressive cardiomegaly and the International Mouse Phenotyping Consortium linked Gimap6 deficiency with increased heart weight (Figures 1C, S1C, and S1D).¹⁸ In moribund mice, triphenyl tetrazolium chloride staining revealed extensive pale, unstained myocardial regions consistent with infarction, and plasma cardiac troponin I rose from 8–12 weeks with marked terminal elevation (Figures 1D-1F).^19,20 Gimap6^−/− hearts showed ventricular dilation, fibrosis, hypertrophic and fibrotic remodeling gene induction, and CD68⁺ macrophage infiltration (Figures 1G, S1E, and S1F). Serial echocardiography, ECG, and telemetry revealed progressive mechanical and electrical deterioration, including a declining ejection fraction that occurred earlier in females, QRS prolongation, ST-segment depression, pathologic Q waves up to 30 days before death, and terminal ventricular arrhythmias (Figures 1H-1K; Video S1; Table S1).²¹

We asked whether the ischemic cardiac disease reflected an underlying vascular disorder. In wild-type mice, Gimap6 was enriched in vascular and immune tissues but low in heart and liver, contrasting with canonical atherosclerosis susceptibility genes Apoe and Ldlr (Figure S1G).²² Doppler echocardiography revealed disturbed aortic flow in Gimap6^−/− mice (Figure 1L).²³ We then tested for atherosclerosis by en face Oil Red O (ORO) staining of the aorta. Whereas WT C57BL/6 mice had no atherosclerosis on a standard-fat diet (SFD),²⁴ 20-week-old SFD-fed Gimap6^−/− mice spontaneously developed atherosclerotic plaques, particularly at turbulence-prone branch points, including the renal arteries (Figure 1M).²⁵ Importantly, SFD-fed Gimap6^−/− mice maintained a normal, HDL-dominant cholesterol profile without elevated total cholesterol and slightly increased triglycerides in males (Figures S1H and S1I). High-fat diet (HFD) feeding increased plaque burden five-fold, to 20% of the aortic surface (Figure S1J). Thus, monogenic Gimap6 deficiency causes spontaneous inflammatory vasculopathy with aortic atherosclerotic lesions and fatal ischemic cardiovascular disease in the absence of hyperlipidemia, although excess lipid exacerbates disease.

Gimap6 deficiency and hyperlipidemia synergize to drive lethal coronary atherosclerosis in Ldlr^−/− mice

To model a more human-like LDL-dominant lipid profile, we crossed Gimap6^−/− mice onto the LDL receptor-deficient (Ldlr^−/−) background, which normally requires aging or HFD to develop atherosclerosis and alone rarely shows coronary involvement, spontaneous infarction, or death.^26,27 Gimap6^−/−Ldlr^−/− mice had markedly reduced survival compared with Gimap6^−/− or Ldlr^−/− mice, while heterozygous Gimap6^+/−Ldlr^−/− mice trended toward reduced survival (Figure 2A).

Figure 2. Gimap6 deficiency synergizes with hyperlipidemia to drive lethal coronary atherosclerosis in Ldlr^−/− mice.

(A) Kaplan-Meier survival curves of SFD-fed Ldlr^−/−, Gimap6^+/−Ldlr^−/−, Gimap6^−/−, and Gimap6^−/−Ldlr^−/− mice (males/females, n = 34/33, 36/33, 67/60, and 21/26). Gimap6^−/− data are replotted from Figure 1A.

(B, C) En face ORO staining (B) and plaque quantification (C) in Ldlr^−/− and Gimap6^−/−Ldlr^−/− mice fed SFD or switched to HFD at 8 weeks and analyzed after 12 or 8 weeks for male and females, respectively. (n = 5 per sex and genotype). Scale bars, 0.5 cm.

(D, E) Masson’s trichrome staining (D) and CD68, α-SMA, and DAPI immunofluorescence staining (E) of aortae from male Ldlr^−/− and Gimap6^−/−Ldlr^−/− mice after 12 weeks of HFD. Red arrows mark foam cells; black arrows mark necrotic cores; dashed yellow lines mark the smooth muscle layer; white arrows indicate macrophage infiltration. Scale bars, 250 μm.

(F, G) Masson’s trichrome staining (F) and quantification (G) of aortic valve plaque and necrotic core areas from Ldlr^−/− and Gimap6^−/−Ldlr^−/− mice after 4 weeks of HFD (males/females, n = 3/3 and 3/4, respectively). Scale bars, 200 μm.

(H) Histological analysis of intramyocardial coronary arteries from HFD-challenged female Ldlr^−/− and Gimap6^−/−Ldlr^−/− mice from (B). Masson’s trichrome (columns 1–3), ORO (column 4), and CD68/α-SMA/DAPI immunofluorescence (column 5). Bottom panels show enlarged views. Scale bars, 100 μm.

Statistics: log-rank (A); two-way ANOVA (C, G). See also Figures S2-S5.

Unless otherwise indicated, data are mean ± SD. Statistical significance is denoted as ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Gimap6^−/−Ldlr^−/− mice developed substantially more severe atherosclerosis than Ldlr^−/− controls on both diets, with aneurysms evident after HFD challenge (Figures 2B, 2C, S2A, and S2B). Lesions in the thoracic and abdominal aorta showed abundant foamy macrophages, CD68⁺ inflammatory infiltrates, smooth-muscle/medial degeneration, and necrotic cores, unlike Ldlr^−/− controls (Figures 2D and 2E). Importantly, female Gimap6^−/−Ldlr^−/− mice developed severe vascular lesions after 8 weeks of HFD challenge, comparable to males after 12 weeks (Figures 2B and S2B).

Critically, Gimap6^−/−Ldlr^−/− mice developed severe coronary artery disease. After only 4 weeks of HFD, Gimap6^−/−Ldlr^−/− mice developed plaques with necrotic cores and thrombi at the aortic valve and coronary ostia, more severe than in Ldlr^−/− controls (Figures 2F, 2G, and S2C). By 8 weeks of HFD, plaques extended into intramyocardial coronary branches, with ORO-positive foam cells, necrotic cores, CD68⁺ macrophage-rich inflammation, and perivascular fibrosis extending into the surrounding myocardium (Figure 2H). These lesions nearly occluded the coronary arteries and were associated with heart-surface pallor, atrial appendage hemorrhage, and elevated cardiac troponin I in moribund mice (Figures 2H, S2D, and S2E). Heterozygous Gimap6^+/−Ldlr^−/− mice had increased aortic plaque burden with no increased heart weight or marked fibrosis (Figures S2F-S2I).

We also investigated whether the cardiac phenotype could reflect a primary myocardial metabolic defect.^28,29 Cardiac tissue had little GIMAP6 expression (Figures S1G and S3A), and 8-week-old Gimap6^−/−Ldlr^−/− hearts, examined before advanced disease, showed no significant changes in fatty acid handling genes nor species, however, the ratio of ATP/AMP and phosphocreatine were decreased, showing a reduced energy charge (Figures S3B-S3E). Together, these data argue against primary cardiomyocyte lipid-metabolic pathology and support ischemia as the disease mechanism.

These findings also explain the renal phenotype previously described as glomerular microangiopathy in Gimap6-deficient mice.¹⁵ In conjunction with renal-artery plaques (Figure 1M), kidneys from Gimap6^−/− mice showed glomerular collapse, tubular epithelial injury, and interstitial fibrosis, while adjacent vessels showed wall thickening and smooth-muscle disorganization (Figures S4A-S4C). Moreover, a 4-week HFD challenge produced severe inflammation, foamy macrophage accumulation, and renal artery vessel-wall disruption in Gimap6^−/−Ldlr^−/− mice (Figure S4D). Thus, the kidney abnormalities suggest vascular insufficiency and ischemia rather than a primary glomerular disease.

Finally, we asked whether accelerated atherosclerosis in Gimap6^−/−Ldlr^−/− mice could be explained by plasma lipid levels. Gimap6^−/− mice maintained a normal, HDL-dominant lipid profile, with only modest changes in triglycerides (Figures S1H and S1I). However, SFD-fed Gimap6^−/−Ldlr^−/− mice exhibited elevated LDL, total cholesterol, and triglyceride levels, which increased further after HFD challenge (Figures S5A and S5B). Notably, while plasma cholesterol and triglyceride levels were comparable, atherosclerosis was significantly more severe in SFD-fed Gimap6^−/−Ldlr^−/− mice than in HFD-fed Ldlr^−/− controls (Figures S5B, S5C, and 2C). Thus, Gimap6 deficiency accelerates atherosclerosis through mechanisms that cannot be explained by plasma lipid levels alone.

Gimap6 deficiency causes early endothelial barrier failure and monocyte adhesion before overt plaque formation

Single-cell transcriptomic analyses localized Gimap6 expression mainly to endothelial cells (ECs) and macrophages in mouse aorta and heart (Figures 3A and 3B), pointing to vascular-wall compartments as candidate drivers of lipid-independent disease.³⁰ Because endothelial dysfunction initiates atherosclerosis,³¹ we tested whether GIMAP6 loss promotes endothelial barrier dysfunction before plaque formation. Evans blue injections in 6–7-week-old mice showed marked extravasation in the aorta and heart of Gimap6^−/− mice compared with WT and age-matched Ldlr^−/− controls (Figure 3C). En face ORO staining showed no detectable plaques in age-matched WT, Gimap6^−/−, and Ldlr^−/− aortae, indicating that widespread vascular leak precedes plaque development (Figure 3D). To test whether this reflected an EC-intrinsic defect, we studied primary Gimap6^−/− aortic ECs. Gimap6^−/− EC monolayers showed increased basal permeability and exaggerated permeability after TNF stimulation (Figure 3E).³² Time-course analysis showed that Gimap6^−/− EC monolayers became progressively more permeable during prolonged inflammatory stress (Figure 3F). Transmission electron microscopy of 6–7-week-old Gimap6^−/−Ldlr^−/− aortic arch endothelium revealed disrupted junctions and widened intercellular gaps as structural defects underlying the EC barrier failure (Figure 3G). The electron micrographs also revealed possible monocyte-like cells adherent to the disrupted endothelium (Figure 3H). We therefore tested adhesion in vitro and observed that Gimap6^−/− EC monolayers bound monocytes more efficiently than WT EC monolayers (Figure 3I).

Figure 3. GIMAP6 deficiency causes early endothelial barrier failure and monocyte adhesion.

(A, B) Single-cell transcriptomic analysis of Gimap6 expression in mouse aorta (SCP289) and heart (SCP283).³⁰ Arrows indicate Gimap6-high populations.

(C) Evans blue vascular permeability in 6–7-week-old WT, Gimap6^−/−, and Ldlr^−/− mice (n = 4 per genotype), with aorta and heart quantification. Scale bars, 0.5 cm.

(D) En face ORO staining of aortae from 6–7-week-old female WT, Gimap6^−/−, and Ldlr^−/− mice before overt plaque formation. Red arrows indicate lesions in the positive-control 32-week Ldlr^−/− aorta. Scale bars, 0.5 cm.

(E, F) FITC-dextran permeability of WT and Gimap6^−/− AEC monolayers under basal conditions or after TNF stimulation (E) and over time (F) (n = 3 independent experiments; one mouse per experiment).

(G, H) Representative transmission electron microscopy images of endothelial ultrastructure in aortae from female 6–7-week-old Ldlr^−/− (n = 1) and Gimap6^−/−Ldlr^−/− (n = 2) mice. Black arrows indicate an intact junction; red arrows indicate widened junctions; orange arrows indicate abnormal vesicles; light blue arrow indicates a disrupted endothelial cell (EC). M, mononuclear cell; L, lumen. Scale bars, 500 nm (G) and 10 μm (H).

(I) Adhesion of CFSE-labeled monocytes to WT and Gimap6^−/− AEC monolayers. Each point represents one independent experiment (n = 4). Scale bars, 200 μm.

Statistics: one-way ANOVA (C); two-way ANOVA (E); mixed-effects analysis (F); Mann-Whitney test (I).

Unless otherwise indicated, data are mean ± SD. Statistical significance is denoted as ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

GIMAP6 restrains macrophage foam-cell formation through a mitochondrial ROS–PPARγ–CD36 axis

We next tested whether Gimap6-deficient macrophages are intrinsically prone to become lipid-laden foam cells, a defining step in atherosclerotic plaque development.⁴ Upon oxidized LDL (oxLDL) exposure, Gimap6^−/− bone marrow-derived macrophages (BMDMs) formed more ORO⁺ foam cells (Figure 4A), and, under basal conditions, accumulated more BODIPY 493/503 neutral lipid (Figure 4B), indicating an intrinsically lowered threshold for lipid loading. Transcriptomic profiling revealed enrichment of immune/pathogen pathways, as previously reported,^15,33 but also of lipid and atherosclerosis programs (Figure S6A). We found that Cd36, a key scavenger receptor for fatty-acid uptake and foam-cell formation, and Pparg, a nuclear receptor that promotes Cd36 transcription, were increased (Figure 4C).^34,35 Cell survival genes were also upregulated, whereas redox-regulatory genes, such as Sod2, were reduced (Figure 4C). Immunoblots confirmed elevated CD36 and PPARγ in Gimap6^−/− BMDMs (Figure 4D). Flow cytometry showed increased surface CD36 and long-chain fatty-acid uptake in Gimap6^−/− BMDMs, with intermediate levels in heterozygous cells (Figures 4E and 4F). This pattern reflected selective CD36 induction because the scavenger receptors Msr1 (SR-A1) and Olr1 (LOX-1) were unchanged; LXR activation by the synthetic agonist GW3965 induced the cholesterol transporters Abca1 and Abcg1 similarly in WT and Gimap6^−/− macrophages and reduced lipid accumulation (Figures S6B-S6D).

Figure 4. GIMAP6 restrains macrophage foam-cell formation through a mitochondrial ROS–PPARγ– CD36 axis.

(A) ORO staining and foam-cell quantification in WT and Gimap6^−/− bone marrow-derived macrophages (BMDMs) after oxLDL treatment. Scale bars, 50 μm. Right, relative foam cell number per field (n = 9–10 fields per experiment).

(B) BODIPY 493/503 lipid-droplet staining and quantification in unstimulated WT and Gimap6^−/− BMDMs. Scale bars, 50 μm. Right, number of lipid droplets per cell (50–100 cells per group).

(C) Heatmap of selected DEGs associated with lipid metabolism, redox regulation, and cell survival in WT (n = 3) and Gimap6^−/− (n = 3) BMDMs.

(D) Immunoblot analysis of CD36, PPARγ, β-actin, and GIMAP6 in BMDMs from Gimap6^+/+, Gimap6^+/−, and Gimap6^−/− mice.

(E, F) Flow-cytometric quantification of CD36 surface expression (E) and BODIPY-C16 uptake (F) in BMDMs of the indicated genotypes (n = 5 per genotype).

(G) Flow-cytometric analysis of mitochondrial ROS by MitoSOX in WT, Gimap6^−/−, and Gimap6^−/−Cd36^−/− BMDMs (n = 3 experiments from 3 mice).

(H) MitoSOX and CD36 expression in WT and Gimap6^−/− BMDMs treated with vehicle or MitoTEMPO.

(I–K) CD36 and PPARγ immunoblots in WT and Gimap6^−/− BMDMs after NAC/MitoTEMPO treatment (I) or oxLDL ± GW9662 treatment (J), and ORO staining with lipid-accumulation quantification after oxLDL ± MitoTEMPO or GW9662 (K). Scale bars, 100 μm; n = 3 independent experiments from 3 mice.

(L) Kaplan-Meier survival curves for Gimap6^−/− and Gimap6^−/−Cd36^−/− mice fed SFD up to 50 weeks (males/females n = 67/60 and 19/33, respectively). Gimap6^−/− data are replotted from Figure 1A and truncated at 50 weeks.

Statistics: t test (A, B); one-way ANOVA (E–H, K); two-way ANOVA (I, J); log-rank (L). See also Figure S6.

Unless otherwise indicated, data are mean ± SD. Statistical significance is denoted as ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

The reduced Sod2 signature, together with prior links between GIMAP6 and mitochondrial homeostasis,^13,15 suggested redox dysregulation. Indeed, Gimap6^−/− macrophages showed increased mitochondrial ROS (mitoROS) that persisted despite Cd36 deletion, placing mitoROS upstream of CD36 (Figure 4G). We next inhibited mitochondrial or total ROS with MitoTEMPO or N-acetylcysteine (NAC), respectively, and found reduced PPARγ and CD36 proteins in Gimap6^−/− macrophages (Figures 4H and 4I). Inhibition of PPARγ with its alkylating inhibitor GW9662 reduced CD36 expression after oxLDL stimulation (Figure 4J). Functionally, MitoTEMPO and GW9662 reduced oxLDL-induced lipid accumulation in Gimap6^−/− macrophages (Figure 4K). Metabolomics and Seahorse profiling of BMDMs revealed energy stress with reduced carnitine and a decreased ATP/AMP ratio, as well as increased levels of the inflammatory lipid LPA (16:0), but no broad impairment of TCA-cycle metabolites, mitochondrial respiration, or glycolytic flux (Figures S6E-S6H). Thus, GIMAP6 loss activates a macrophage-intrinsic mitoROS→PPARγ→CD36→lipid uptake cascade that promotes foam-cell formation.

Gimap6^−/− macrophages contribute to vascular inflammation and endothelial barrier failure

To test whether CD36 contributes to disease in vivo, we generated Gimap6^−/−Cd36^−/− mice. Cd36 deletion substantially but incompletely rescued early lethality (Figure 4L), suggesting other molecular changes contribute to ongoing vascular injury. This partial rescue is consistent with the context-dependent impact of CD36 loss reported across atherosclerosis models.^36,37

Whole-aorta transcriptomics identified an IL-6 transcriptional production program as the most significantly increased inflammation pathway in Gimap6^−/−Ldlr^−/− vessels, including induction of genes linked to endothelial permeability, cytokine signaling, innate immune sensing, and lipid handling (Figures S7A and S7B). Whole-aorta explants showed no genotype-dependent increase in cytokine secretion at 6 weeks, but at 12 weeks, when vascular lesions were established, IL-6, MCP-1, and TNF secretion was elevated with corresponding increases in plasma (Figures 5A and S7C). Moreover, Gimap6^−/− macrophages secreted more IL-6, MCP-1, and TNF after stimulation by oxLDL, and this response was suppressed by MitoTEMPO (Figure 5B). Conditioned medium from these experiments increased WT endothelial monolayer permeability, an effect reduced by MitoTEMPO treatment of the macrophages (Figure 5C). Permeability was further exacerbated using Gimap6^−/− AECs, suggesting that GIMAP6 deficiency creates an intrinsically vulnerable endothelial barrier (Figures 5C and 3E). Thus, redox-activated Gimap6^−/− macrophages release cytokines that worsen endothelial barrier failure.

Figure 5. Redox-activated Gimap6^−/− macrophages amplify endothelial injury.

(A) Aortic explant IL-6, MCP-1, and TNF secretion from Ldlr^−/− and Gimap6^−/−Ldlr^−/− mice at 6 and 12 weeks (n = 5 and 4 per genotype, respectively).

(B) IL-6, MCP-1, and TNF secretion by WT and Gimap6^−/− BMDMs treated with vehicle, oxLDL, or oxLDL plus MitoTEMPO (n = 4 mice from 3 experiments).

(C) Endothelial permeability of WT and Gimap6^−/− (G6^−/−) aortic endothelial cell (AEC) monolayers treated with macrophage-conditioned medium (Mφ) from WT or Gimap6^−/− macrophages exposed to oxLDL ± MitoTEMPO.

(D) CD68, IL-6, and MCP-1 immunohistochemical staining of aortic lesions from Ldlr^−/− and Gimap6^−/−Ldlr^−/− male mice after 12 weeks of HFD. Dashed lines mark the approximate intima-media boundary; arrows indicate positive regions in the vessel wall (black) and foam cells (red). Scale bars, 250 μm.

(E) En face ORO-stained aortae from Ldlr^−/− and Gimap6^−/−Ldlr^−/− male mice after 12 weeks of HFD. The enlarged view highlights severe remodeling and aneurysmal degeneration.

(F) En face ORO staining and plaque quantification in PBS- or MitoTEMPO-treated Ldlr^−/− and Gimap6^−/−Ldlr^−/− female mouse aortae after 4 weeks of HFD (n = 3, 3, 5, and 4, respectively). Scale bars, 0.5 cm.

(G) Gross heart images and heart-weight quantification from mice shown in (F). Scale bars, 0.5 cm.

(H, I) Plasma IL-6, MCP-1, and TNF (H) and cardiac troponin I levels (I) from mice in (F).

(J) Proposed model of GIMAP6-dependent vascular protection. Statistics: two-way ANOVA (A–C, F–I). See also Figure S7.

Unless otherwise indicated, data are mean ± SD. Statistical significance is denoted as ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Histochemical analysis of Gimap6^−/−Ldlr^−/− lesions showed striking CD68⁺ macrophage infiltration with intramural IL-6 and MCP-1 staining (Figure 5D). Lesions progressed to aortic aneurysms and dissection with intramural hemorrhage (Figures 5E and S7D-S7F), consistent with prior evidence that an IL-6/MCP-1 inflammatory loop can promote aortic dissection in mice.³⁸ MitoTEMPO administration to HFD-fed mice reduced aortic plaque burden, attenuated cardiomegaly, decreased circulating IL-6, MCP-1, and TNF, and lowered plasma cardiac troponin I in Gimap6^−/−Ldlr^−/− mice (Figures S7G and 5F-5I). Hence, these findings support a model in which GIMAP6 protects endothelial barrier integrity while suppressing redox-activated foam cells and inflammatory cytokines that amplify vessel-wall injury (Figure 5J).

The GIMAP6-CD36 pathway is conserved in human pro-atherogenic monocytes and macrophages

We next asked whether the GIMAP6–CD36 pathway operates in human monocytes and macrophages during ASCVD.³⁹ Single-cell transcriptomic analyses showed that GIMAP6 was enriched in macrophages, endothelial cells, and lymphocytes and was selectively reduced in aortic tissue from patients with coronary artery disease (CAD) (Figures 6A-6C).^30,40,41 Moreover, GIMAP6 was highest in monocytes from individuals with normal coronary arteries (NCA), lower in those with intermediate coronary lesions (ICL), and lowest in those with acute myocardial infarction (AMI) suggesting that reduced GIMAP6 expression is associated with CAD severity (Figure 6D).⁴² Further, GIMAP6 and PPARG expression were inversely related (Figure 6D). Across human carotid plaque macrophage states, GIMAP6 expression decreased from F13A1⁺ monocyte-like macrophages to SPP1⁺ foam-cell-like macrophages, whereas CD36 and PPARG increased (Figure 6E).⁴³ Thus, reduced GIMAP6 expression correlates with increased PPARG/CD36 expression and symptomatic cardiovascular disease. Silencing GIMAP6 by siRNA in human primary CD14⁺ monocytes strongly upregulated CD36 and PPARγ (Figure 6F), supporting conservation of the GIMAP6→PPARγ→CD36 cascade between humans and mice.

Figure 6. Human GIMAP6 hypofunction activates a conserved macrophage CD36 lipid-uptake pathway.

(A) Single-cell RNA-seq analysis of human aortic cell populations showing GIMAP6 expression (SCP1265).³⁰ Arrows indicate GIMAP6-high populations.

(B) Expression of GIMAP6 across human vascular cell populations from the Human Protein Atlas.⁴¹

(C) Log₂ fold difference in GIMAP6 expression between patients with coronary artery disease (CAD) and healthy controls (HC) from STARNET.⁴⁰

(D) GIMAP6 expression across CAD stages and correlation with PPARG expression in peripheral blood monocytes. NCA, normal coronary arteries; ICL, intermediate coronary lesions; AMI, acute myocardial infarction (n = 8 per group).⁴²

(E) Expression of CD36, PPARG, and GIMAP6 across macrophage subclusters identified by single-cell RNA-seq of human carotid atherosclerotic plaques.⁴³

(F) Immunoblot analysis of GIMAP6, CD36, PPARγ, and β-actin in primary human CD14⁺ monocytes transfected with 3 GIMAP6-targeting (A–C) or negative control (NC) siRNAs.

(G) Immunoblot analysis of GIMAP6, CD36, and HSP90 in negative control (NC), GIMAP6 knockout (G6KO), and GIMAP6/CD36 double-knockout (DKO) THP-1 cells.

(H) Representative flow cytometry histograms of THP-1 cells from (G) stained with BODIPY 493/503.

(I) Flow cytometry quantification of BODIPY-C16 uptake in NC, G6KO, and DKO THP-1 cells with or without sulfo-N-succinimidyl oleate (SSO). Left, uptake kinetics shown as mean fluorescence intensity (MFI); right, relative area under the curve (AUC). Data are mean ± SEM (n = 5 independent experiments).

(J) Pedigrees of two families carrying hypomorphic GIMAP6 variants. The proband in each family is homozygous and marked by an arrow. Asterisk (*) indicates living individuals without available sequencing data. Myocardial infarction, MI; years old, y.o.

(K) Cycloheximide chase analysis of WT and variant GIMAP6 proteins expressed in HEK293T cells (data pooled from 3 independent experiments).

(L) GIMAP6, CD36, and HSP90 immunoblotting in PBMCs from GIMAP6 variant carriers and controls.

(M) CD36 surface expression on CD14⁺CD16⁻ classical monocytes from homozygous, heterozygous carriers, and healthy controls.

Statistics: one-way ANOVA (D, M); two-way ANOVA (I, K); Pearson correlation (D).

Unless otherwise indicated, data are mean ± SD. Statistical significance is denoted as ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

We next investigated lipid accumulation and uptake in human monocytic THP-1 cells with GIMAP6 single knockout (G6KO) or combined GIMAP6/CD36 double knockout (DKO) (Figure 6G). G6KO THP-1 cells showed increased fluorescent BODIPY lipid accumulation, which was largely reversed in DKO cells (Figure 6H). In acute BODIPY-C16 uptake assays, G6KO cells showed increased long-chain fatty-acid uptake, which was reduced by either CD36 deletion (DKO) or the CD36 fatty-acid uptake inhibitor sulfo-N-succinimidyl oleate (SSO) (Figure 6I).⁴⁴ Notably, the lipid-accumulation experiments were performed in RPMI-1640/10% FBS medium, which contains much lower cholesterol and fatty-acid concentrations than human serum.^45,46 Thus, GIMAP6 deficiency enhances CD36-dependent lipid uptake and accumulation even under hypolipidemic conditions.

Hypomorphic GIMAP6 variants link IEI-associated protein instability to CD36 upregulation and human vasculopathy

We previously reported that rare homozygous GIMAP6 stop-gain or missense hypomorphic variants, Trp86Ter (W86*) and Gly153Val (G153V), cause an IEI with immune dysregulation and vasculopathy.^15,16 The W86* variant occurred in two children from Gaza now lost to follow-up. The patient with the G153V variant (Pt. 1) developed inflammatory vasculopathy and acute focal cerebral ischemic lesions in her 20s,¹⁵ later accompanied by mild cardiomegaly and borderline prolonged QTc interval despite low LDL and cholesterol levels (Table S2). We identified two additional patients carrying rare, homozygous missense variants (Figure 6J). Pt. 2, carrying an Ile44Thr variant (I44T; 7-150325555-A>G; CADD 22.8; MAF = 0.0053 in South Asia), had deep venous thrombosis and a family history of midlife stroke, myocardial infarction (MI), and CAD (Figure 6J). Cardiac evaluation at age 38 showed preserved systolic function (ejection fraction 64%) but severe diastolic dysfunction (mitral E/A ratio of 2.68) and ECG nonspecific T-wave abnormalities despite only mildly elevated plasma lipids (Figure 6J; Table S3). Pt. 3, a 10-year-old boy carrying a Val126Met variant (V126M; 7-150628222-C>T; CADD 13.38; MAF = 0.0013 in non-Finnish Europeans) presented at age 9 with an acute-onset stroke despite a normal lipid profile. Neuroimaging revealed diffuse cerebral vasculitis with multifocal arterial narrowing and multiple acute infarcts, accompanied by elevated CRP and TNF. He rapidly progressed to fatal cerebral edema (Figure 6J; Table S4).

We examined whether the I44T and V126M mutations altered protein stability—similar to the published G153V variant—or affected CD36.¹⁵ Cycloheximide (CHX) chase experiments revealed that all three variant GIMAP6 proteins underwent more rapid degradation than wild-type (WT) GIMAP6 (Figure 6K). GIMAP6 protein was reduced in peripheral blood mononuclear cells (PBMCs) from Pts. 1 and 2, indicating that these variants are hypomorphic (Figure 6L). Blots also showed elevated CD36 protein in individuals heterozygous for G153V and in patients homozygous for G153V or I44T (Figure 6L). Flow cytometry confirmed increased CD36 expression in CD14⁺CD16⁻ classical monocytes with evidence of a gene-dosage effect (Figure 6M). Collectively, these families suggest that severe GIMAP6 hypofunction couples immune dysregulation to monocyte CD36 upregulation and early vasculopathy without hyperlipidemia.

GIMAP6 shows female-biased basal expression in immune and vascular systems

GTEx data revealed sexually dimorphic GIMAP6 expression, with higher transcript levels in female than male donors in immune-rich tissues, including spleen and whole blood, as well as in lower-expressing tissues including the aorta and coronary arteries (Figure S8A).⁴⁷ This dimorphism was also observed in mice (Figure S8B).⁴⁸ Other GIMAP family members did not show the same consistent female bias. Moreover, Cd36 and Pparg expression was lower in macrophages from females, consistent with an inverse relationship to Gimap6 expression. Notably, Gimap6 was the most abundantly expressed Gimap family member in mouse macrophages (Figure S8B), as was previously observed in human macrophages.¹⁵ In primary mouse AECs (GSE247364),⁴⁹ Gimap6 expression was higher in females and the only Gimap robustly expressed (Figure S8C). Thus, higher basal GIMAP6 expression in females may be conserved in mammals, potentially contributing to the heightened sensitivity to GIMAP6 deficiency observed in females.

African ancestry-enriched GIMAP6 V65I associates with premature cardiovascular disease independent of hypercholesterolemia

We examined GIMAP6 population-level variation across large-scale biobanks, including the Massachusetts General Brigham Biobank (MGB), UK Biobank (UKBB) and All of Us (AoU) Research Program, using the Broad CVDI portal.⁵⁰ Across data from over 700,000 individuals, the aggregated burden of rare, predicted damaging GIMAP6 variants (LOF + missense with score 0.5) was significantly associated with atherosclerosis but not hypercholesterolemia, unlike LDLR and PCSK9 (Figure 7A). Among individual variants tested across coronary atherosclerosis, ischemic heart disease, heart failure, and heart failure with reduced ejection fraction, Val65Ile (V65I), a variant with increased frequency in African ancestry populations, emerged as the most consistent risk-associated allele, whereas the rare V126M variant showed similar but nonsignificant trends and common variants, such as Q237R and T78A, had negligible effects (Figure 7B). V65I did not correlate with non-cardiovascular diseases, although a modest type 2 diabetes signal was observed (Figure S9A).

Figure 7. GIMAP6 variants associate with premature cardiovascular disease through cholesterol-independent mechanisms.

(A) CVD gene-burden testing for atherosclerosis and hypercholesterolemia using aggregated pLoF and rare damaging missense variants (MAF<0.1%, missense ≥0.5) in GIMAP6, LDLR, and PCSK9. Meta, fixed-effects meta-analysis of UK Biobank, All of Us, and Mass General Brigham cohorts.

(B) Association between GIMAP6 missense variant effect size and CADD score for CAD, IHD, HF, and heart failure with reduced ejection fraction (HFrEF). Bubble size indicates carrier count; color indicates p value.

(C) Cumulative prevalence of CAD, HF, combined CAD+HF diagnosis, and hypercholesterolemia (HChol) in self-reported Black/African American participants in All of Us (AoU) stratified by sex and genotype.

(D) Relative allele frequency of GIMAP6 V65I compared to risk alleles (APOE ε4, TTR V142I), a benign GIMAP6 missense variant (Q237R), and 100 ancestry/MAF-matched controls across age bins. Data shown for the overall cohort of self-reported Black/African American participants in AoU and stratified by sex.

(E) Logistic regression analysis evaluating odds ratios for combined CAD and HF in GIMAP6 V65I carriers vs. non-carriers, adjusted for age, sex, smoking status, BMI category, diabetes mellitus, hypertension, hyperlipidemia status, and 16 ancestry principal components.

(F) Replication of the association between GIMAP6 V65I and cardiovascular phenotypes (CAD, coronary artery disease; CHF, congestive heart failure) in Black/African American participants in the Million Veteran Program.

(G) Immunoblot analysis of GIMAP6 and CD36 in CD14⁺ monocytes from GIMAP6 V65I carriers (n = 5) and healthy controls, including no-race-reported (NR; n = 2) and Black/African American (AA; n = 3) individuals. Right, quantification normalized to HSP90.

(H, I) Flow-cytometric analysis of surface CD36 expression and BODIPY-C16 uptake in CD14⁺ monocytes from GIMAP6 V65I carriers and healthy controls (NR ctrl, n = 3; AA ctrl, n = 3). Representative histograms (H) and quantification (I).

Statistics: Z-score test (D); logistic regression (E); Mann-Whitney test (G) and (I). AoU analyses in (C–E) used participants with both WGS and EHR data. See also Figures S8-S12.

Unless otherwise indicated, data are mean ± SD. Statistical significance is denoted as ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Biochemically, V65I, along with the S105L, V126M, V149F, and R194W substitutions, caused GIMAP6 protein instability (Figures S9B and S9C). Structural modeling placed V65I and V126M near functionally important regions, with V65I located near the conserved GTP-binding domain (Figures S9D). The minor allele frequency (MAF) of the V65I allele is 1.5% in individuals of African ancestry and linkage disequilibrium analyses supported V65I as an independent, ancestry-enriched cardiovascular risk variant (Figures S9E, S9F, and S10A-S10C).⁵¹

Given the strong African-ancestry enrichment of V65I, we next evaluated its clinical impact within African-ancestry participants to reduce population-stratification effects. In AoU, pathogenic LDLR variants served as a lipid-dependent CAD comparator and tracked with CAD and hypercholesterolemia, whereas TTR V142I/V122I served as an African ancestry-enriched late-onset heart-failure comparator and tracked with heart failure primarily in older age groups (Figure 7C).^52-54 By contrast, GIMAP6 V65I showed directionally increased risk across CAD and heart-failure phenotypes, with the strongest and statistically significant association observed for the combined CAD+HF endpoint, without increased hypercholesterolemia (Figures 7C and S11A-S11C). Consistent with a possible survival disadvantage, V65I carrier frequency declined after age 75, particularly among women, exceeding the age-associated depletion observed for ancestry- and MAF-matched control variants, APOE ε4, or TTR V142I (Figure 7D; Table S5).^55,56 PheWAS and adjusted logistic regression analyses showed associations concentrated in cardiovascular disease, and the CAD/HF association was validated in an independent African-ancestry Million Veteran Program cohort (Figures 7E, 7F, S11D, and S11E; Table S6). Age- and sex-stratified analyses showed a more apparent association in females, particularly for CAD+HF, although these analyses do not establish a statistically significant difference between sexes (Figures 7C-E). In ancestry-matched AoU analyses, participants homozygous for V65I showed higher cardiovascular burden, including atherosclerosis, myocardial infarction, stroke, and heart failure, with permutation testing supporting enrichment beyond chance expectation (Figures S11E and S11F). Directionally consistent with this observation, the only two individuals homozygous for V65I identified in UK Biobank both had heart failure, including one with ischemic heart disease.⁵⁷

Finally, we purified primary CD14⁺ monocytes from individuals heterozygous for V65I and found that they exhibited reduced GIMAP6 protein, increased CD36 expression, elevated mitochondrial ROS, and increased fatty-acid uptake recapitulating key features of the GIMAP6-deficient pathogenic pathway (Figures 7G-7I and S12A-S12C). Thus, reduced GIMAP6 function is associated with premature cardiovascular disease in humans through a molecular pathogenic mechanism independent of hypercholesterolemia.

DISCUSSION

Our data show that GIMAP6 ensures a healthy barrier function of ECs and suppresses a mitoROS→PPARγ→CD36→lipid uptake cascade in macrophages. Mechanistically, loss of GIMAP6 triggers elevated mitoROS, possibly by defective mitophagy,^13,15,58 which upregulates PPARγ and increases the scavenger receptor CD36. This lowers the threshold for lipid uptake and subsequent foam-cell formation leading to increased proinflammatory cytokine secretion. While CD36 deletion in mice offers substantial protection, its incomplete rescue identifies it as a major, but nonexclusive, downstream mediator.^36,37 Atherosclerotic vessel wall remodeling likely also involves additional inflammatory, permeability, and extracellular-matrix pathways beyond macrophage lipid-uptake.^59,60

ASCVD is classically framed as a disease of elevated cholesterol and LDL, with inflammation amplifying plaque progression.^4,61,62 However, the GIMAP6 phenotype reveals another potent risk factor that determines vascular disease susceptibility. Gimap6-deficient mice with intact LDL clearance, as well as human GIMAP6 variant carriers, developed vascular disease without hypercholesterolemia. In the Ldlr^−/− background, Gimap6 deficiency boosted LDL and triglycerides, likely reflecting its secondary activated macrophage inflammatory responses,⁶³ but lipid-matched comparisons showed that plasma lipids alone could not explain the accelerated atherosclerosis. Importantly, normolipidemic Gimap6^−/− mice develop systemic vascular leakage, spontaneous aortic atherosclerotic lesions, and lethal ischemic cardiac injury, whereas advanced and near-occlusive coronary atherosclerosis is demonstrated most directly after combining Gimap6 deficiency with the Ldlr^−/− background. Thus, GIMAP6 functions as a distinct cardiovascular disease determinant that protects the endothelial structural integrity and controls the threshold for macrophage response to lipid, rather than systemic lipid levels themselves. Our findings therefore bridge IEIs and cardiovascular biology.^5,64 While severe, biallelic GIMAP6 loss of function causes IEI-associated immune dysregulation and vasculopathy, including cerebral ischemic lesions in affected patients, hypomorphic heterozygous alleles—such as the African ancestry-enriched V65I variant—elevate the risk of premature coronary artery disease (CAD) and heart failure independent of hypercholesterolemia. In mice, Gimap6 deficiency affects multiple vascular compartments, including the aorta, coronary vasculature, brain microvasculature, and kidney, supporting a systemic immune-vascular vulnerability rather than an aortic root-restricted phenotype. Human V65I carrier monocytes exhibit key features of the GIMAP6-deficient pathogenic cascade, supporting the idea that this population-level risk targets an immune-vascular checkpoint conserved in mammals.

The female-biased severity of GIMAP6-associated disease is notable because females are generally relatively protected from ASCVD.⁶⁵ Together with higher basal GIMAP6 expression in female immune and vascular compartments, this suggests that GIMAP6 may act as a protective regulator upon which females are especially dependent. Whether this dependence is driven by cell-intrinsic differences, sex hormones, or other systemic factors remains unresolved. More broadly, GIMAP6 reveals how immune-regulatory failure can couple endothelial dysfunction, foam-cell formation, and inflammation into a self-reinforcing cascade that culminates in lethal ischemic cardiovascular disease. These cardiovascular roles of GIMAP6 suggest that upstream mitochondrial redox reduction or the development of small-molecule chaperones to increase GIMAP6 stability, especially for hypomorphic variants such as V65I, could be therapeutic strategies for atherosclerosis, especially in women and populations with African ancestry, and may be more broadly useful for high-risk ASCVD.

Limitations of the study

While our study strongly supports GIMAP6's role, we acknowledge several limitations. First, our inbred complete knockout mouse model simplifies human genetic complexity and cannot parse the relative contributions of GIMAP6 in macrophages versus endothelial cells. Second, the proximal mechanism linking GIMAP6 loss to increased mitoROS and activation of the mitoROS-PPARγ-CD36 pathway remains a critical open question, including whether defective autophagy contributes to this process. Third, some of our biobank analyses suffer from limited representation of African and non-European ancestries; prospective longitudinal studies are needed to define lifetime risk for presymptomatic V65I carriers. Finally, while we propose that intrinsic cellular abnormalities drive the observed sexual dimorphism, extrinsic modulators—such as sex hormones directly influencing the GIMAP6-PPARγ axis—remain critical areas for future investigation.

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Human participants

Written informed consent was obtained from all participants in accordance with local ethics and Institutional Review Board (IRB) guidelines. Patient 1 was enrolled in NIH protocol 06-I-0015 (NCT00246857), and genetic sequencing was performed as previously reported. Patient 2 was enrolled in NIH protocol 01-I-0202 (NCT00018044), with clinical and laboratory samples collected during inpatient visits. Patient 3 was enrolled under protocol IR.TUMS.CHMC.REC.1398.030, approved by the Tehran University of Medical Sciences Ethics Committee. Individuals carrying the GIMAP6 V65I variant were enrolled under NIH protocol 03-H-0015 (NCT00047996) and 09-I-0032 (NCT00794352), and blood samples were collected during outpatient visits. Control samples and healthy donor blood were obtained through a separate IRB-approved protocol 05-I-0213 (NCT00128973) at the NIH Clinical Center. Detailed clinical summaries and laboratory findings for patients 1–3 are provided in Tables S2-S4.

Mouse strains, housing, and diet

All animal procedures were approved by the NIAID Animal Care and Use Committee (protocol LISB-7E) and followed NIH and Department of Health and Human Services guidelines. Experimental animals were maintained on a congenic C57BL/6 background. B6N(Cg)-Gimap6<tm1b(KOMP)Wtsi>/J mice were generated by the Knockout Mouse Project (KOMP) and maintained as previously described.¹⁵ B6.129S7-Ldlr<tm1Her>/J mice (JAX Stock No. 002207) were crossed with Gimap6^−/− mice to generate Gimap6^−/−Ldlr^−/− double knockout animals. B6.129S1-Cd36tm1Mfe/J (JAX Stock No. 019006) were crossed with Gimap6^−/− mice to generate Gimap6^−/−Cd36^−/− double knockout animals. Genotyping was performed by Transnetyx using real-time PCR. Experimental cohorts were derived from intercrosses of Gimap6^+/−, Gimap6^+/−Cd36^−/− or Gimap6^+/−Ldlr^−/− heterozygous breeders. Littermate knockouts and controls were co-housed to minimize cage and microbiota effects. Mice were housed under specific pathogen-free (SPF) conditions (22 ± 2°C, 12 h light/dark cycle, 40–60% humidity) with up to five per individually ventilated cage, provided autoclaved chow and water ad libitum, and environmental enrichment. For survival studies, mice were maintained on standard chow. For atherosclerosis experiments, mice were switched to a high-fat diet (Research Diets, #D12108C) at 8 weeks of age and maintained on the diet for the durations indicated in the figure legends.

Animals were allowed to age naturally unless removed by veterinary staff due to non-age-related health concerns, in which case they were euthanized by CO₂ inhalation followed by cervical dislocation. Male and female mice were used unless otherwise specified. The age, sex, genotype, diet, and treatment condition for each experiment are provided in the relevant method subsection and figure legends. Age- and sex-matched littermates were used where possible. For treatment experiments, animals were randomly assigned to treatment groups within each genotype.

Cell culture

THP-1 cells (ATCC, TIB-202; RRID: CVCL_0006; human male, acute monocytic leukemia-derived monocytic cell line) were purchased from ATCC and cultured in RPMI-1640 medium supplemented with 10% heat-inactivated fetal bovine serum (FBS), 2 mM L-glutamine, and 1% penicillin-streptomycin. Cells were maintained at 37°C in a humidified 5% CO₂ incubator at 2 × 10⁵ to 6 × 10⁵ cells/mL and passaged every 2–3 days to maintain exponential growth and >90% viability by trypan blue exclusion. CRISPR/Cas9-engineered THP-1 derivatives were generated from the same parental THP-1 stock and maintained under identical culture conditions.

HEK293T cells (ATCC, CRL-3216; RRID: CVCL_0063) were purchased from ATCC and cultured in DMEM supplemented with 10% FBS and 1% penicillin-streptomycin. Cells were maintained at 37°C in a humidified 5% CO₂ incubator and passaged at 70%–90% confluency. Cell-line identity was based on acquisition from ATCC and associated vendor authentication records.

Primary human monocytes

Primary human CD14⁺ monocytes were isolated from fresh peripheral blood mononuclear cells (PBMCs) collected from healthy adult male and female donors. PBMCs were prepared using Ficoll-Paque^™ Plus (Cytiva) density gradient centrifugation, followed by positive selection with Human CD14 MicroBeads (Miltenyi Biotec) and magnetic-activated cell sorting (MACS) per manufacturer’s instructions. The resulting monocyte population was ~95% pure, as confirmed by flow cytometry. Primary human CD14⁺ monocytes were cultured in RPMI-1640 supplemented with 10% heat-inactivated FBS, 2 mM L-glutamine, and 1% penicillin-streptomycin at 37°C with 5% CO₂. Cells were used for downstream assays shortly after isolation unless otherwise indicated.

Primary mouse cultures

Primary mouse aortic endothelial cells (AECs) and bone marrow-derived macrophages (BMDMs) were generated from mice of the indicated genotypes, ages, and sexes. Primary cells were cultured at 37°C in a humidified 5% CO₂ incubator. Passage 2-3 AECs were used for functional assays. AEC cultures were assessed by flow cytometry for CD31 expression and were approximately 90% CD31⁺. Cultures were also monitored for characteristic endothelial morphology and absence of overt fibroblast overgrowth. BMDMs were used immediately after differentiation and were not passaged.

Method details

Sanger sequencing

Genomic DNA was extracted from purified CD14⁺ monocytes from V65I carriers and healthy controls using the DNeasy Blood and Tissue Kit (Qiagen). The region of interest in GIMAP6 was amplified using 2x CloneAmp HiFi PCR Premix (Takara) and primers 5’-CAGTTGTGTCACCAGGAGCA-3’ (forward) and 5’-TTCTATTGGCCAAGTCTCGCT-3’ (reverse) using a VeritiPro Thermal Cycler. PCR products were submitted to Poochon Scientific for Sanger DNA sequencing. Sequence chromatograms were analyzed using SnapGene v8.2.2.

Mouse genotyping

Routine mouse genotyping was performed by Transnetyx using real-time PCR. As an alternative genotyping method, B6N(Cg)-Gimap6<tm1b(KOMP)Wtsi>/J mice were genotyped using allele-specific PCR to distinguish WT and knockout alleles.

The knockout allele was amplified using forward primer 5′-GCTACCATTACCAGTTGGTCTGGTGTC-3′ and reverse primer 5′-ATATAAAGACCCTCCCTTCACCGCC-3′, yielding a 617-bp product.

The WT allele was amplified using forward primer 5′-TTGCTGCCATCTTCCTTACAAGTGC-3′ and reverse primer 5′-AAGTTATTCCACACGTTTGGGGG-3′, yielding a 415-bp product.

PCR was performed in 50 μL reactions containing 25 μL 2× MyTaq Red Mix (Meridian), 0.2 μM of each primer, 1 μL genomic DNA, and nuclease-free water. Thermocycling conditions were 94°C for 5 min; 10 touchdown cycles of 94°C for 15 s, 65°C to 55°C for 30 s with −1°C per cycle, and 72°C for 40 s; followed by 25 cycles of 94°C for 15 s, 55°C for 30 s, and 72°C for 40 s; with a final extension at 72°C for 10 min. PCR products were separated by 1.5% agarose gel electrophoresis. B6.129S7-Ldlr<tm1Her>/J mice were genotyped according to Jackson Laboratory protocol 27075, version 1.2.

Generation of GIMAP6 variant constructs

mCherry-T2A-FLAG-GIMAP6 expression plasmids encoding wild-type or mutant human GIMAP6 variants were synthesized by GenScript Gene Synthesis. All constructs were generated in the same vector backbone and verified by Sanger sequencing. Constructs used for cycloheximide chase experiments encoded WT, E12D, E32G, I44T, V65I, T78A, Q103R, S105L, V126M, V149F, G153V, R194W, or Q237R human GIMAP6, as indicated.

Survival analysis

To assess the impact of Gimap6 deficiency on survival, Kaplan-Meier analysis was performed in mice maintained on standard chow. Gimap6^−/−, Gimap6^+/−, and Gimap6^+/+ mice were generated through intercrosses of Gimap6^+/− heterozygotes. Similarly, Gimap6^−/−Ldlr^−/−, Gimap6^+/−Ldlr^−/−, Gimap6^−/−Cd36^−/− and Ldlr^−/− mice were produced from controlled heterozygous matings. Breeding was continuous, and mice were enrolled in survival cohorts at weaning (postnatal day 21), with heterozygous female Gimap6^+/− breeder mice included in the analyses. Age- and sex-matched mice from different litters were co-housed (up to five per cage) under identical SPF conditions to minimize environmental and microbiota variability. Mice were monitored throughout the study and euthanized at predefined humane endpoints in accordance with NIH Animal Care and Use Committee guidelines. Animals euthanized after reaching predefined humane endpoints were counted as mortality events in survival analyses. Survival data were pooled across multiple litters and generations, and Kaplan-Meier curves were generated and analyzed using GraphPad Prism v10.

Body weight measurements

Body weight was recorded weekly at the indicated time points using a calibrated digital scale. For the longitudinal body-weight analysis, 15 age- and sex-matched mice per genotype per sex were analyzed as a representative subset of the survival cohort. Gimap6^−/−, Gimap6^+/−, and Gimap6^+/+ mice were housed under SPF conditions with ad libitum access to autoclaved standard chow and water. Body weight was analyzed separately by sex where indicated.

Heart weight measurements

Gross heart morphology and heart weight were assessed in Gimap6^−/−, Gimap6^+/−, Gimap6^+/+, Gimap6^−/−Ldlr^−/−, Gimap6^+/−Ldlr^−/− and Ldlr^−/− mice maintained on standard chow at the indicated time points. After euthanasia, hearts were excised, rinsed in cold PBS, blotted dry, and photographed under standardized lighting and magnification to document gross anatomy. Hearts were weighed using a calibrated analytical balance. Heart weights were analyzed separately by sex. Sample sizes for each genotype, sex, and age group are indicated in the corresponding figure legends.

Cardiac troponin-I quantification

Plasma cardiac troponin-I was measured using a mouse-specific cardiac troponin-I ELISA kit (MyBioSource, #MBS766175; sensitivity, 4.7 pg/mL; dynamic range, 7.8–500 pg/mL) according to the manufacturer’s instructions. For time-course analysis, whole blood was collected at the indicated ages by tail-vein sampling into lithium-heparin tubes. For endpoint analyses, blood was collected by cardiac puncture immediately after euthanasia of moribund mice and age- and sex-matched controls. Moribund animals were defined according to protocol-specific humane endpoint criteria, including severe lethargy, impaired mobility, or >20% body weight loss. For treatment studies, plasma was collected at the experimental endpoint. Plasma was isolated by centrifugation at 3,000 × g for 10 min at 4°C and stored at −80°C until analysis. Plasma samples were assayed at 50 μL per well in ELISA plates, and absorbance was read at 450 nm using a VersaMax Microplate Reader. Cardiac troponin-I concentrations were calculated from a standard curve generated using the supplied standards.

Mouse histopathology

For paraffin sections, tissues were fixed overnight at 4°C in 10% neutral buffered formalin, transferred to 80% ethanol, processed on a Tissue-Tek VIP system (Sakura, GMI Inc.), embedded in paraffin, sectioned at 5 μm (rotary microtome), and mounted on Superfrost Plus slides (Thermo Fisher). Slides were dried overnight at 37°C before staining. For frozen sections, tissues were cryoprotected in sucrose solution (30% sucrose in PBS), embedded in Tissue-Tek^® O.C.T. (Sakura), snap-frozen in isopentane over liquid nitrogen, stored at −80°C, sectioned at 5 μm (cryostat), mounted, air-dried at room temperature, and either stained immediately or stored at −80°C. Histological analyses were performed on heart, aorta, coronary artery, aortic valve, and kidney sections, as indicated in the figure legends. Images were acquired under matched magnification and exposure settings where comparisons were made across groups.

TTC staining

Myocardial viability was assessed by 2,3,5-triphenyltetrazolium chloride (TTC) staining of freshly excised mouse hearts. Gimap6^−/− mice, which typically appear acutely ill and die within ~24 h after symptom onset, were euthanized at moribund stages alongside age- and sex-matched controls. Immediately post-euthanasia, hearts were rinsed in cold PBS (pH 7.4), transversely sliced into 4 to 5 sections, and incubated in 1% TTC solution (Sigma-Aldrich) freshly dissolved in PBS at 37°C for 45 min in the dark. After staining, slices were fixed in 10% neutral buffered formalin at room temperature for at least 24 h. Images were acquired using a Smartphone camera in ambient lighting. Viable myocardium stained red (formazan), while infarcted regions remained pale.

Masson’s trichrome staining

Cardiac fibrosis and extracellular matrix were assessed on formalin-fixed, paraffin-embedded (FFPE) heart sections (5 μm) using the Masson’s trichrome Stain Kit (VitroView). Sections were deparaffinized in xylene (2 × 6 min), rehydrated through graded ethanol (100%, 95%, 70%), and rinsed in distilled water. Slides were mordanted in preheated Bouin’s solution (60°C, 1 h), cooled, and rinsed in tap and distilled water. Nuclear staining was performed with freshly prepared Weigert’s working hematoxylin (10 min), followed by Biebrich Scarlet-Acid Fuchsin cytoplasmic staining (5 min). Sections were treated with phosphomolybdic-phosphotungstic acid (10 min), stained with Aniline Blue (10 min), dipped briefly in 1% acetic acid, and rinsed. Slides were dehydrated through 95% and 100% ethanol, cleared in xylene (3×5 min), and mounted with Permount (Fisher Scientific). Collagen stained blue, muscle fibers red, and nuclei black.

Immunofluorescence staining

Formalin-fixed, paraffin-embedded heart and aorta sections were mounted on Superfrost Plus slides (Thermo Fisher Scientific), deparaffinized in xylene for 2 × 6 min, rehydrated through graded ethanol solutions of 100%, 95%, and 70%, and rinsed in distilled water. Antigen retrieval was performed in citrate buffer consisting of 10 mM sodium citrate, pH 6.0, at 95°C for 15–20 min. Slides were cooled at room temperature for 30 min and washed with PBS. Sections were blocked for 1 h at room temperature in PBS containing 5% normal goat serum and 0.3% Triton X-100. Primary antibodies were diluted in PBS containing 1% BSA and 0.3% Triton X-100 and incubated overnight at 4°C. Primary antibodies included rabbit monoclonal anti-CD68 (clone E3O7V, Cell Signaling Technology), diluted 1:200, and mouse monoclonal anti-alpha-smooth muscle actin (clone 1A4, R&D Systems), used at 10 μg/mL. The next day, sections were washed in PBS containing 0.3% Triton X-100 and incubated for 1 h at room temperature protected from light with goat anti-mouse IgG (H+L) Alexa Fluor Plus 488 (Thermo Fisher Scientific) or goat anti-rabbit IgG (H+L) Alexa Fluor Plus 647 (Thermo Fisher Scientific). Nuclei were counterstained with DAPI. Slides were mounted with Fluoromount-G (SouthernBiotech) and stored at 4°C protected from light. Fluorescence images were acquired using a Leica SP8 confocal microscope under identical acquisition settings across comparison groups. Image analysis was performed using Imaris v10.1.1.

Immunohistochemistry

Immunohistochemistry was performed on formalin-fixed, paraffin-embedded aortic tissue sections cut at 4 μm thickness. Sections were deparaffinized, rehydrated through graded ethanol, and subjected to antigen retrieval in citrate buffer consisting of 10 mM sodium citrate and 0.05% Tween-20, pH 6.0, by microwave heating followed by incubation at 95°C in a steamer. Sections were blocked with normal goat serum for 15 min at room temperature. Sections were incubated overnight at 4°C with primary antibodies against CD68 (clone E3O7V, Cell Signaling Technology), IL-6 (Proteintech), or MCP-1 (Proteintech). Primary antibodies were used at the following dilutions: anti-CD68, 1:800; anti-IL-6, 1:400; and anti-MCP-1, 1:800. Endogenous peroxidase activity was quenched with 1% H₂O₂ in PBS for 15 min. Sections were then incubated with HRP-conjugated polymer secondary antibody (Leica Biosystems) and developed using 3,3-diaminobenzidine substrate (Thermo Fisher Scientific). Nuclei were counterstained with Mayer hematoxylin (Millipore). Brightfield images were acquired using a Leica microscope under matched imaging settings across comparison groups.

Echocardiography

Cardiac structure and function were assessed by high-resolution transthoracic echocardiography in WT and Gimap6^−/− mice of the indicated ages and sexes. Mice were lightly anesthetized with 1–1.5% isoflurane delivered by nose cone and placed supine on a heated rail-mounted imaging platform equipped with ECG leads and a rectal temperature probe. Heart rate, respiratory rate, and core temperature were continuously monitored and maintained within physiological ranges during imaging. Chest fur was removed by shaving followed by depilatory cream, and pre-warmed ultrasound gel was applied for acoustic coupling.

Echocardiographic imaging was performed using a Vevo F2 imaging system with a UHF46X linear transducer. Two-dimensional and M-mode images were acquired from parasternal long-axis and mid-papillary short-axis views. M-mode tracings were used to measure left ventricular internal diameter at end-diastole (LVIDd) and end-systole (LVIDs). Left ventricular ejection fraction and related cardiac functional parameters were calculated using Vevo analysis software according to the manufacturer’s workflow. Fractional shortening was calculated as $\text{FS}(\%)=[(\text{LVIDd}-\text{LVIDs})∕\text{LVIDd}]\times 100$. Pulsed-wave and color Doppler imaging were used to assess ascending aortic outflow, aortic arch flow pattern, and valve function. Additional parameters, including vessel diameter and pulse-wave velocity, were acquired where indicated. Each imaging session lasted approximately 15 min per animal. After imaging, mice recovered on a warming pad or were euthanized according to the experimental protocol.

Telemetry ECG

Long-term cardiac monitoring was performed using surgically implanted ECG telemetry transmitters (Data Sciences International, ETA-F10) under sterile conditions. Mice were anesthetized with 2% isoflurane and positioned in dorsal recumbency. The body temperature was maintained at 37 °C using a water blanket and monitored with rectal temperature probe. Preoperative analgesia (Buprenorphine ER^®, 1 mg/kg subcutaneously) was administered. The chest and abdomen were shaved and disinfected with alternating chlorhexidine and 70% ethanol scrubs. A midline abdominal incision was made caudal to the xiphoid process, and the transmitter body was implanted in the lower abdominal cavity and secured with non-absorbable 6-0. The ECG leads were passed from the abdominal cavity to the exterior and tunneled subcutaneously: the positive lead to the lower chest pectoral muscles, the negative lead to the upper right pectoral muscles. Both leads and the abdomen were anchored with non-absorbable 6-0 sutures. The skin incisions were closed with wound clips; Bupivacaine was applied topically. 1 mL of sterile saline was administered subcutaneously over the dorsal flank. Postoperative care included oxygenated warmed recovery. Mice were housed in oxygen-enriched recovery cages on heated water blankets for 72 h and monitored twice daily by Murine Phenotyping Core or surgical core staff for pain, dehydration, wound healing, and food intake. Analgesia (Buprenorphine ER^®, 1 mg/kg every 2 to 3 days; Meloxicam, 5 mg/kg daily) and subcutaneous fluids (1 to 2 mL as needed) were provided per NHLBI protocols. Staples or wound clips were removed 10 to 14 days post-surgery if incisions were fully healed; mice were euthanized if distress or complications were observed. The ECG data were recorded and analyzed using Ponemah software (Data Sciences International).

PowerLab ECG analysis

ECG signals were acquired using PowerLab ECG equipment (ADInstruments, Octal Bio Amp). Mice were anesthetized with 2% isoflurane and positioned in dorsal recumbency. The body temperature was maintained at 37 °C using a heat lamp and monitored with rectal temperature probe. For surface ECG, 29-gauge needle electrodes were inserted subcutaneously into limbs for 4-lead configurations (Lead I and II). Signals were recorded continuously for 2 min per animal using LabChart 8 software (ADInstruments), and Lead I was used for the analysis.

Plasma lipid analysis

Plasma total cholesterol and triglyceride concentrations were measured using commercial colorimetric assay kits from FUJIFILM Wako Chemicals according to the manufacturer’s instructions. Total cholesterol was measured using the Cholesterol E CHOD-DAOS Method Kit, and triglycerides were measured using the L-Type Triglyceride M GPO-HMMPS Method Kit.

FPLC lipoprotein profiling

For fast protein liquid chromatography (FPLC) lipoprotein profiling, plasma samples were pooled within each experimental group using 150 μL total pooled plasma per group. Each pool contained plasma from 6–8 mice per diet condition. Pooled plasma was applied to a superose 6 10/300 GL column on an ÄKTA Pure FPLC system. Lipoproteins were separated at a flow rate of 0.5 mL/min using 1× PBS containing 0.01% EDTA, and 0.5 mL fractions were collected. Cholesterol concentrations in FPLC fractions were measured using FUJIFILM Wako colorimetric assay kits according to the manufacturer’s instructions. Lipoprotein distribution was assigned according to elution volume, corresponding to very-low-density lipoprotein (VLDL), low-density lipoprotein (LDL), and high-density lipoprotein (HDL) fractions. Where indicated, area under the curve (AUC) was calculated by summing lipid concentrations across the corresponding VLDL, LDL, or HDL elution fractions. For total plasma lipid measurements, each biological data point represents one mouse. For FPLC profiles, each trace represents one pooled plasma sample from 6–8 mice per group and was used to visualize lipoprotein distribution rather than for individual-level statistical testing.

Oil Red O staining of en face aortae

Mice were euthanized per institutional guidelines, and the full aorta (aortic root to iliac bifurcation) was dissected under a stereomicroscope, rinsed in cold PBS (pH 7.4), and fixed in 10% neutral buffered formalin for 5 min at room temperature. After rinsing in PBS, aortae were stored in distilled water in 6-well plates to prevent dehydration. Oil Red O working solution was freshly prepared by diluting 0.5% (w/v) Oil Red O stock in isopropanol (Abcam) at 3:2 with distilled water, then filtered. Aortae were stained for 30 min at room temperature with gentle agitation, washed in 60% isopropanol (30 min), rinsed in distilled water, and stored in PBS at 4°C. Following staining, residual adventitial adipose tissue was carefully removed using 6 cm Cohan-Vannas Spring Scissors (Fine Science Tools) and 11 cm Dumont Tweezers #5 (World Precision Instruments) and aortae were opened longitudinally along the ventral side. For en face imaging, stained aortae were mounted between two glass slides: a drop of distilled water was placed on a Superfrost Plus slide, the aorta (adventitia side down) was flattened with Extra Fine Graefe Forceps (Fine Science Tools), and a second slide was placed on top. Corners were sealed with Clear Silicone Waterproof Sealant (Loctite). After drying (~30 to 60 min), the sandwich was placed vertically in a water-filled dish to promote capillary action. Air bubbles were removed with a 31G insulin needle, and remaining edges were sealed. Slides were air-dried (1 h), stored at 4°C in light-protected boxes, and imaged within one week.

Digital microscope images were captured at matched magnification and exposure settings for the aortic arch and thoracoabdominal regions. Lipid deposition was quantified using ImageJ software (NIH, ImageJ bundled with Java 8). Images were converted to 8-bit grayscale, thresholded to isolate Oil Red O–positive areas and analyzed using the polygon tool to calculate: $\text{Percent lesion area}=(\text{Oil Red O}-\text{positive area}∕\text{total aortic area})\times 100$.

For WT and Gimap6^−/− comparisons, male mice were analyzed at 20 weeks of age on standard-fat diet. For high-fat-diet challenge, male WT and Gimap6^−/− mice were switched to high-fat diet at 8 weeks of age and analyzed at 20 weeks of age after 12 weeks of high-fat diet feeding. For Ldlr^−/−, Gimap6^+/−Ldlr^−/−, and Gimap6^−/−Ldlr^−/− comparisons, mice were maintained on standard-fat diet until 8 weeks of age and then either continued standard-fat diet or switched to high-fat diet. Male mice were analyzed after an additional 12 weeks and female mice after an additional 8 weeks, unless otherwise indicated. For MitoTEMPO rescue experiments, female Ldlr^−/− and Gimap6^−/−Ldlr^−/− mice were analyzed after the treatment schedule described in the in vivo MitoTEMPO section.

Evans blue permeability

Vascular permeability was assessed in 6–7-week-old female WT, Gimap6^−/− and Ldlr^−/− mice before overt aortic plaque formation. Evans blue dye was prepared as a sterile 0.5% (w/v) solution in PBS, and 200 μL of the solution was administered to each mouse by tail-vein injection. After 30 min of circulation, mice were euthanized by CO₂ inhalation. To minimize postmortem vascular permeability artifacts, mice were processed in cohorts of no more than six animals, and all animals within a cohort were euthanized and dissected rapidly under identical timing. Then the circulatory system was perfused via the left ventricle with 20 mL of PBS to wash out intravascular dye. After perfusion, the aorta, heart, and other organs were harvested and photographed under standardized lighting to document areas of dye leakage into the tissue parenchyma.

For quantification, harvested tissues were weighed, minced, and incubated in 2 mL of formamide at 60°C for 48 hours to extract the Evans blue dye as described previously.⁶⁶ The formamide supernatant was then collected after centrifugation (12,000 × g, 10 min), and the absorbance was measured at 620 nm using a VersaMax Microplate Reader. The concentration of the dye was calculated from a standard curve, and the results were expressed as nanograms (ng) of Evans Blue dye per milligram (mg) of tissue.

Transmission electron microscopy

Transmission electron microscopy was used to assess aortic endothelial ultrastructure in 6–7-week-old female Ldlr^−/− and Gimap6^−/−Ldlr^−/− mice before advanced plaque formation. Aortae were dissected immediately after euthanasia, and the aortic arch and abdominal aortic regions were selected for ultrastructural analysis. One Ldlr^−/− control mouse and two Gimap6^−/−Ldlr^−/− mice were analyzed. Aortic tissue was fixed in 2.5% glutaraldehyde and 2% paraformaldehyde in Sorensen’s buffer and processed for transmission electron microscopy. Samples were post-fixed in 1% osmium tetroxide and 1.5% potassium ferrocyanide, stained with uranyl acetate, dehydrated through graded ethanol, and embedded in low-viscosity epoxy resin. Ultrathin sections of 90 nm were prepared using a Leica UC7 ultramicrotome, mounted on copper grids, and post-stained with lead citrate. Images were acquired using a Hitachi HT7800 transmission electron microscope operated at 80 kV. Endothelial junctions, intercellular gaps, endothelial morphology, vesicle accumulation, and luminal mononuclear-cell adhesion were assessed from acquired micrographs. Representative images were selected from multiple fields examined within the analyzed aortic regions.

Mouse AEC isolation

Primary mouse aortic endothelial cells (AECs) were isolated from 7–8-week-old female WT and Gimap6^−/− mice using a ring explant-based sprouting method. Mice were euthanized, disinfected with 70% ethanol, and the thoracic cavity was opened under sterile conditions. Aortae were excised into PBS containing penicillin-streptomycin and surrounding adipose and connective tissues were removed under a stereomicroscope. Cleaned aortae were cut into 1–2 mm rings and placed in 35 mm culture dishes precoated with 50 μL Matrigel (Corning Life Sciences) diluted 1:1 in cold PBS. After Matrigel polymerization for 20 min at room temperature, explants were overlaid with 1–2 additional drops of Matrigel and allowed to polymerize for another 20 min. Aortic explants were cultured at 37°C in a humidified 5% CO₂ incubator in endothelial cell growth medium consisting of Ham F12/DMEM 1:1, 10% heat-inactivated FBS, 50 μg/mL endothelial cell growth supplement (Sigma-Aldrich), 10 U/mL heparin, and 1% penicillin-streptomycin. Medium was changed every 2–3 days. Endothelial sprouting was typically observed within 5–7 days and aortic rings were removed after observing sufficient outgrowth. 3–4 days following aortic ring removal, endothelial sprouts were rinsed with PBS, dissociated with Dispase I (Sigma-Aldrich, 2 U/mL in PBS, 0.1 mL/cm²) at 37°C for 20 min, diluted with 10 mL cold endothelial cell growth medium, centrifuged at 500 x g for 10 min at 4°C, and resuspended in fresh endothelial cell growth medium. For expansion, AECs were seeded onto fibronectin-coated dishes. Dishes were coated overnight at 4°C with fibronectin (Sigma-Aldrich) diluted to 10 μg/mL in PBS at 0.2 mL/cm² and rinsed with PBS before use. AECs were maintained at 37°C in 5% CO₂ and passaged at approximately 90% confluency. Passage 2–3 AECs were used for experiments. AEC cultures were approximately 90% CD31⁺ by flow cytometry. Cultures were additionally monitored for characteristic endothelial morphology and absence of overt fibroblast overgrowth.

AEC permeability assay

Endothelial monolayer permeability was measured using the In Vitro Vascular Permeability Assay, 96-well format (Millipore/Merck), according to the manufacturer’s protocol with the following modifications. Passage 2–3 primary mouse AECs were used after formation of a confluent monolayer and confirmation of typical endothelial cobblestone morphology. Cells were detached, counted by hemocytometer, and viability was assessed by trypan blue exclusion. Collagen-coated 96-well permeability inserts were hydrated with growth medium before cell seeding. AECs were seeded at 50,000 cells per insert in 100 μL endothelial cell growth medium. Each receiver well contained 250 μL endothelial cell growth medium. Cells were cultured for 72 h at 37°C in 5% CO₂ to allow monolayer formation. No-monolayer wells containing growth medium, but no cells were included as permeability controls. Monolayer formation was assessed by visual inspection and monolayer staining according to the assay protocol. After monolayer formation, inserts were transferred to a fresh receiver tray. AEC monolayers were left untreated in growth medium or treated with recombinant TNF (BioLegend) in growth medium. For dose-response experiments, cells were treated with 100 ng/mL TNF. For time-course experiments, cells were treated with 100 ng/mL TNF for the indicated durations. After treatment, inserts were transferred to a fresh receiver tray for FITC-dextran flux measurement. FITC-dextran working solution was prepared by diluting the kit-provided FITC-dextran reagent 1:40 in medium and was protected from light. The receiver wells were filled with 250 μL medium, and 75 μL FITC-dextran working solution was added to the upper chamber of each insert. Plates were incubated for 20 min at room temperature protected from light. Permeation was stopped by removing the inserts, and the receiver-well medium was mixed thoroughly. A 100 μL aliquot from each receiver well was transferred to a black opaque 96-well plate. Fluorescence intensity was measured using a VICTOR2 1420 Multilabel Counter with Wallac software using FITC-compatible excitation/emission settings of 485/535 nm and a 1 s counting time. Receiver-well FITC-dextran fluorescence was used as the measure of relative endothelial permeability. Experiments were performed using AECs from three independent biological replicates, with two technical inserts per condition. Wells were excluded only if monolayer formation failed before treatment or if inserts were physically damaged before measurement; no wells were excluded after fluorescence measurement.

In vitro monocyte adhesion assay

Monocyte adhesion to primary mouse AECs was measured using a CFSE-based fluorescence adhesion assay. Passage 2–3 primary AECs isolated from WT and Gimap6^−/− mice were seeded at 2 × 10₅ cells per well in fibronectin-coated 24-well plates and cultured at 37°C in 5% CO₂ until confluent monolayers formed. Mouse monocyte-enriched adherent cells were prepared from pooled PBMCs isolated from 7–10-week-old WT mice by plastic adherence. Briefly, PBMCs were plated in tissue-culture-treated dishes and incubated for 2 h at 37°C. Non-adherent cells were removed by gentle washing, and adherent cells were collected and labeled with 5 μM CFSE in serum-free medium for 30 min at 37°C. Cells were washed thoroughly to remove excess dye and resuspended in endothelial cell growth medium. A total of 4 × 10₅ CFSE-labeled monocyte-enriched cells were added to each well containing confluent WT or Gimap6^−/− AEC monolayers and incubated for 2.5 h at 37°C in 5% CO₂. After incubation, non-adherent cells were removed by gently washing each well three times with pre-warmed PBS. Adherent CFSE-positive cells were imaged by fluorescence microscopy using a 20× objective under identical acquisition settings across groups. Adherent CFSE-positive cells were quantified using ImageJ. Monocyte adhesion was normalized within each independent experiment to the WT AEC condition, which was set to 1, and reported as relative monocyte adhesion.

BMDM isolation and culture

Mice were euthanized, and femurs were harvested. Bone marrow was flushed using a 25G needle attached to a 10 mL syringe filled with cold sterile PBS, collected in 50 mL conical tubes on ice, and centrifuged at 500 × g for 10 min at room temperature. Pellets were resuspended in macrophage complete medium containing DMEM supplemented with 10% FBS, 1% penicillin-streptomycin, and 20 ng/mL recombinant murine M-CSF (GenScript), filtered through a 70 μm cell strainer to remove debris, and counted by hemocytometer. Cells were seeded at 4×10⁶ per 10 cm non-tissue culture-treated Petri dish in 10 mL medium and maintained at 37°C, 5% CO₂. On day 3, 5 mL of fresh medium was added. On day 7, nonadherent cells were removed by PBS wash, and adherent macrophages were detached with 0.25% trypsin-EDTA with phenol red (1 mL/dish) for 3 min at room temperature. BMDM cultures showed typical macrophage morphology, and macrophage identity was confirmed by F4/80 expression in representative preparations.

Foam-cell formation assay

Foam-cell formation was assessed by Oil Red O staining after oxLDL loading. Differentiated BMDMs were seeded onto poly-L-lysine–pretreated glass coverslips placed in 24-well tissue-culture plates at 4 × 10₅ cells per well and allowed to adhere overnight. Cells were left untreated or treated with oxidized LDL (Thermo Fisher Scientific, 50 μg/mL) for 24 h at 37°C in 5% CO₂. After treatment, cells were washed with PBS and fixed with 10% neutral buffered formalin for 10 min at room temperature. Fixed cells were stained with freshly prepared Oil Red O working solution, generated by diluting 3 parts 0.5% (w/v) Oil Red O stock solution in isopropanol with 2 parts distilled water, for 30 min at room temperature. Cells were then washed with 60% isopropanol for 10 min, followed by PBS washes to remove residual dye. Stained cells were imaged by brightfield microscopy under identical acquisition settings across groups. Foam-cell formation was quantified using ImageJ as Oil Red O-positive foam-cell number, Oil Red O-positive area, or relative foam-cell number, as indicated in the figure legends.

BODIPY 493/503 lipid-droplet staining

Neutral lipid accumulation was assessed by BODIPY 493/503 staining. A 10 mM BODIPY 493/503 stock solution was prepared by dissolving 1 mg BODIPY 493/503 (GLPbio Technology Inc.) in 382 μL DMSO. BODIPY 493/503 working solution was prepared fresh by diluting the stock in PBS to a final concentration of 5 μM. Differentiated BMDMs from 8–10-week-old male or female WT and Gimap6^−/− mice were seeded onto sterile coverslips in 24-well plates at 4 × 10⁵ cells per well and allowed to adhere overnight. For basal lipid-droplet analysis, BMDMs were maintained in standard macrophage culture medium without additional stimulation. Culture medium was removed, and cells were fixed with 10% neutral buffered formalin for 10 min at room temperature. Cells were washed with PBS and incubated with 300 μL BODIPY 493/503 working solution per coverslip for 30 min at room temperature protected from light. Cells were then washed with PBS and counterstained with DAPI. Coverslips were mounted and imaged by fluorescence microscopy under identical acquisition settings across groups. Lipid droplets were quantified using ImageJ as the number of BODIPY-positive lipid droplets per cell. DAPI-positive nuclei were used to define cell number.

Pharmacologic inhibition experiments

For pharmacologic inhibition experiments, differentiated BMDMs were generated as described above from 8–10-week-old male or female mice of the indicated genotypes. Cells were seeded at the density appropriate for each downstream assay and allowed to adhere overnight before treatment. Unless otherwise indicated, oxLDL (Thermo Fisher Scientific) stimulation was performed at 50 μg/mL for 24 h before analysis. For mitochondrial ROS inhibition, BMDMs were pretreated with MitoTEMPO (MCE) at 5 μM for 2 h before downstream stimulation or staining. For antioxidant treatment, BMDMs were pretreated with N-acetyl-L-cysteine (NAC, Millipore Sigma) at 5 mM for 2 h. For PPARγ inhibition, BMDMs were pretreated with GW9662 (MCE) at 5 μM for 2 h. MitoTEMPO and GW9662 were prepared as DMSO stocks, and vehicle-control wells received the same final DMSO concentration as the corresponding drug-treated wells. NAC was dissolved in cell culture grade water, and vehicle-control wells received the corresponding solvent. For oxLDL stimulation, control wells received an equal volume of the buffer used to deliver oxLDL. After treatment, cells were processed for MitoSOX staining, CD36 surface staining, immunoblotting, Oil Red O staining, BODIPY-based lipid uptake assays, or cytokine measurements as described in the corresponding method sections.

In vivo MitoTEMPO treatment

Female Ldlr^−/− and Gimap6^−/−Ldlr^−/− mice were used for in vivo mitochondrial ROS inhibition experiments. At 7 weeks of age, mice were randomly assigned within genotype to receive MitoTEMPO or vehicle. MitoTEMPO (MCE) was dissolved in DMSO as a concentrated stock and diluted in sterile PBS before administration. Mice received MitoTEMPO at 0.7 mg/kg body weight by intraperitoneal injection once daily from 7 to 12 weeks of age. Vehicle-treated control mice received the same volume of PBS containing the same final DMSO concentration. Mice were maintained on standard-fat diet from 7 to 8 weeks of age and were then switched to high-fat diet from 8 to 12 weeks of age while daily treatment continued. At 12 weeks of age, mice were euthanized, and aortae, hearts, and plasma were collected for Oil Red O plaque analysis, heart-weight measurement, plasma cytokine quantification, and cardiac troponin-I measurement.

Lipid uptake assay

Long-chain fatty-acid uptake was measured using BODIPY FL C16 (Thermo Fisher Scientific). For THP-1 cell-line experiments, normal control, GIMAP6 knockout, CD36 knockout, and GIMAP6/CD36 double-knockout THP-1 cells were used as indicated. For CD36 inhibition experiments, THP-1 cells were incubated with or without 25 μM sulfosuccinimidyl oleate (SSO, Cayman Chemical) for 15 min at 37°C in HEPES-buffered saline containing 20 mM HEPES, 138 mM NaCl, 1 mM MgCl2, and 6 mM KCl, pH 7.4. After SSO treatment, cells were washed once with PBS and incubated with 2 μM BODIPY FL C16 for 30 min at 37°C. Cells were then washed with PBS, resuspended in FACS buffer consisting of PBS supplemented with 2% FBS, and analyzed by flow cytometry using a BD FACSymphony A3 instrument.

For BMDM lipid-uptake experiments, differentiated BMDMs from 8–10-week-old male or female mice of the indicated genotypes were seeded in non-coated 24-well plates at 4 × 10⁵ cells per well. Non-coated plates were used to minimize firm adherence and facilitate recovery of cells for flow-cytometric analysis. BMDMs were stained directly in the plate with 2 μM BODIPY FL C16 for 30 min at 37°C. After staining, cells were washed with PBS to remove excess probe and harvested using PBS-EDTA. Cells were then collected, resuspended in FACS buffer consisting of PBS supplemented with 2% FBS, and analyzed by flow cytometry using a BD FACSymphony A3 instrument. Mean fluorescence intensity (MFI) of BODIPY FL C16 was used as the measure of fatty-acid uptake.

MitoSOX and CD36 flow cytometry

Mitochondrial ROS and CD36 surface expression were measured by flow cytometry in mouse BMDMs, THP-1 cells, and human PBMC-derived monocyte populations, as indicated. Mouse BMDMs were generated from 6–10-week-old male or female Gimap6^+/+, Gimap6^+/−, and Gimap6^−/− mice, with additional genotypes included where indicated. Human samples included healthy controls and GIMAP6 variant carriers or patients, as indicated. For each staining reaction, 4–5 × 10⁵ cells were used.

For mitochondrial ROS measurement, live cells were stained with MitoSOX Red mitochondrial superoxide indicator (Thermo Fisher Scientific) at a final concentration of 5 μM in phenol red-free HBSS for 20 min at 37°C protected from light. Cells were then washed with HBSS. Mouse BMDMs and THP-1 cells were analyzed live without live/dead dye. For human PBMC samples, a live/dead viability dye (Thermo Fisher Scientific) was included to exclude dead cells during analysis. Cells were acquired live on a flow cytometer immediately after staining. For CD36 surface staining, cells were incubated with Fc block and stained in FACS buffer with anti-CD36 antibody diluted 1:200 for 40 min at 4°C protected from light. Human cells were stained with anti-human CD36 (BioLegend), and mouse cells were stained with anti-mouse CD36 (BioLegend). Mouse BMDMs and THP-1 cells were analyzed without live/dead dye, whereas human PBMC samples included a live/dead viability dye. CD36 staining and MitoSOX staining were performed as separate assays. Where indicated, cells were pretreated with MitoTEMPO (MCE) at 5 μM for 2 h before MitoSOX or CD36 staining. Vehicle-treated cells were processed in parallel. Flow cytometry was performed using a BD FACSymphony A3 instrument, and data were analyzed using FlowJo v10.10.0. Cells were gated on singlets and live cells where a viability dye was used. Human classical monocytes were analyzed as live singlet CD14⁺CD16⁻ cells where indicated. MitoSOX and CD36 signals were quantified as mean fluorescence intensity (MFI) using identical gating strategies across samples.

Aortic explant and cytokine analysis.

Aortic explant cytokine secretion was measured using Ldlr^−/− and Gimap6^−/−Ldlr^−/− female mice maintained on standard-fat diet. Mice were analyzed at 6 weeks of age, before advanced inflammatory disease, and at 12 weeks of age. Mice were anesthetized with isoflurane and perfused through the left ventricle with sterile PBS before tissue collection. Hearts and aortae were excised en bloc into sterile PBS, and periadventitial fat was removed under a stereomicroscope. The entire aorta was dissected and immediately placed into 0.5 mL DMEM containing 1× insulin-transferrin-selenium (ITS) supplement (Sigma-Aldrich) and 0.1% BSA. Aortae were incubated for 4 h at 37°C in 5% CO₂. Explant culture medium was then collected, clarified by brief centrifugation, and stored at −80°C until cytokine measurement. IL-6, MCP-1, and TNF were measured using the GeniePlex Mouse Inflammation 5-Plex Panel 1 kit according to the manufacturer’s instructions. Cytokine concentrations were reported as pg/mL from the explant-conditioned medium, with the entire aorta cultured under matched volume and incubation conditions for each sample.

BMDM cytokine secretion assay

BMDM cytokine secretion was measured using differentiated BMDMs from 8–10-week-old male WT and Gimap6^−/− mice. BMDMs were seeded in 24-well plates at 5 × 10₅ cells per well in DMEM supplemented with 10% FBS and 1% penicillin-streptomycin and allowed to adhere overnight before treatment. Where indicated, BMDMs were pretreated with MitoTEMPO (MCE; 5 μM) for 2 h before oxLDL stimulation. OxLDL (Thermo Fisher Scientific) was then added at 50 μg/mL, and cells were cultured for 24 h at 37°C in 5% CO₂. Vehicle-treated and unstimulated control wells were processed in parallel. Supernatants were collected after 24 h and stored at −80°C until cytokine measurement. IL-6, MCP-1, and TNF were measured using the GeniePlex Mouse Inflammation 5-Plex Panel 1 kit according to the manufacturer’s instructions. Cytokine concentrations were reported as pg/mL.

For conditioned-medium experiments, BMDMs were first treated with oxLDL as described above. After 24 h, cells were washed with PBS to remove residual oxLDL and then incubated in fresh medium for an additional 24 h to generate macrophage-conditioned medium. Conditioned medium was collected and applied to primary mouse AEC monolayers for permeability assays, as described in the AEC permeability assay section. Conditioned medium from WT or Gimap6^−/− BMDMs, with or without MitoTEMPO treatment, was applied to WT or Gimap6^−/− AEC monolayers as indicated. FITC-dextran permeability was measured using the 96-well vascular permeability assay described above.

Immunoblot

Mouse spleens, hearts, BMDMs, THP-1 cells, HEK293T cells, and human PBMCs or purified CD14⁺ monocytes were lysed in RIPA Lysis and Extraction Buffer (Thermo Fisher Scientific) supplemented with Halt Protease and Phosphatase Inhibitor Cocktails (Thermo Fisher Scientific). Lysates were incubated on ice for 30 min and centrifuged at 16,000 × g for 10 min at 4°C. Protein concentrations were determined using the BCA Protein Assay Kit (Thermo Fisher Scientific). Equal amounts of total protein (25 μg per lane) were mixed with NuPAGE LDS Sample Buffer (4×) (Thermo Fisher Scientific) and Reducing Agent (10×) (Thermo Fisher Scientific), heated at 80°C for 10 min, and resolved by electrophoresis on NuPAGE 4–12% Bis-Tris Protein Gels (Thermo Fisher Scientific). Proteins were transferred onto nitrocellulose membranes (Bio-Rad) using standard wet transfer methods. Membranes were blocked with EveryBlot Blocking Buffer (Bio-Rad) for 10 min at room temperature, then incubated overnight at 4°C with primary antibodies diluted in blocking buffer. Primary antibodies included anti-human GIMAP6 (MAC445),⁶⁷ anti-mouse GIMAP6 (MAC431, MAC432, MAC434, or a mixture of all three),¹³ anti-mouse CD36 (E8B7S, Cell Signaling Technology), anti-PPAR-γ (Proteintech), anti-CD68 (Cell Signaling Technology), anti-human CD36 (D8L9T, Cell Signaling Technology), anti-OLR1 (Proteintech), anti-MSR1 (Proteintech), anti-β-actin (BioLegend), and anti-HSP90 (BioLegend). After washing in TBST (0.05% Tween-20), membranes were incubated for 1 h at room temperature with HRP-conjugated secondary antibodies (SouthernBiotech, 1:2,000 dilution). Protein bands were visualized using chemiluminescent substrates, including Immobilon Classico or Forte Western HRP Substrate (Millipore), SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Fisher Scientific), or SuperSignal West Dura Extended Duration Substrate (Thermo Fisher Scientific) with an Azure 500 system (Azure Biosystems). For membrane stripping and reprobing, membranes were treated with Restore Western Blot Stripping Buffer (Thermo Fisher Scientific) for 1 h at room temperature, followed by three 5-min washes in TBST.

Transfection and cycloheximide chase assay

HEK293T cells were seeded in DMEM with 10% heat-inactivated FBS and 1% penicillin-streptomycin in 12-well plates and grown to ~70% confluency. Plasmids were diluted in Opti-MEM^™ Reduced Serum Medium (Thermo Fisher) and transfected using Lipofectamine^™ 3000 (Thermo Fisher) per manufacturer’s instructions. Cells were incubated at 37°C in a humidified CO₂ incubator. After 24 h, transfected cells were treated with cycloheximide (50 μg/mL; GLPbio) and lysed at the time points indicated in each figure panel. FLAG-GIMAP6 expression was quantified using ImageJ, normalized to HSP90, and, for chase experiments with a 0 h sample, expressed relative to the 0 h time point.

CRISPR/Cas9-mediated gene knockout in THP-1 cells

Each crRNA and ATTO550-labeled tracrRNA (Integrated DNA Technologies) was reconstituted to 100 μM in Nuclease-Free Duplex Buffer (IDT) according to the manufacturer’s instructions. Equal molar amounts of crRNA and tracrRNA were combined in sterile PCR tubes to form gRNA duplexes, annealed at 95°C for 5 min, and gradually cooled to 23°C. For ribonucleoprotein assembly, 3 μL of each annealed crRNA:tracrRNA duplex, corresponding to 150 pmol of each gRNA, was combined with 4.5 μL Cas9 protein (30 μg, 180 pmol; MacroLab, UC Berkeley) and 1.5 μL Nuclease-Free Duplex Buffer and incubated at 23°C for 10 min. For genome editing, 1 × 10⁶ THP-1 cells were nucleofected with RNPs using the SG Cell Line 4D-Nucleofector X Kit L (Lonza) and 4D-Nucleofector X Unit, program FF-100, according to the manufacturer’s protocol. After 24 h, the top 50% of ATTO550-positive cells were isolated by flow cytometry. Both single-cell clones and pooled populations were expanded in culture for downstream analyses. Loss of target protein expression was confirmed by immunoblotting before downstream experiments. crRNAs targeting GIMAP6 (Hs.Cas9.GIMAP6.1.AA, Hs.Cas9.GIMAP6.1.AB, Hs.Cas9.GIMAP6.1.AC) and CD36 (Hs.Cas9.CD36.1.AA, Hs.Cas9.CD36.1.AB, Hs.Cas9.CD36.1.AC) were used.

siRNA transfection of human CD14+ monocytes

Primary human CD14⁺ monocytes were isolated from fresh PBMCs from four healthy adult donors, including three female donors and one male donor, aged 30–50 years. PBMCs were isolated by Ficoll-Paque^™ Plus (Cytiva) density-gradient centrifugation followed by positive selection using CD14 MicroBeads (Miltenyi Biotec) and magnetic-activated cell sorting, as described above. Monocyte purity was ≥95%, as confirmed by flow cytometry. Purified CD14⁺ monocytes were cultured in RPMI-1640 supplemented with 10% heat-inactivated FBS, 2 mM L-glutamine, 100 IU/mL penicillin, and 100 μg/mL streptomycin and were used within 48 h of isolation. For siRNA knockdown, primary CD14+ monocytes were transfected using the P3 Primary Cell 4D-Nucleofector X Kit S and the Lonza 4D-Nucleofector System with program EA-100, according to the manufacturer’s protocol. For each condition, 3 × 10⁶ cells were electroporated with 30 pmol total of either non-targeting control siRNA or one of three GIMAP6-targeting siRNAs (OriGene). siRNAs were reconstituted in RNase-free siRNA duplex resuspension buffer (OriGene) before use. At 24 h after transfection, medium was replaced with fresh monocyte culture medium, and cells were incubated for an additional 24 h before downstream analysis. At 48 h after transfection, cells were harvested for immunoblot analysis of GIMAP6, CD36, PPARγ, and β-actin as described in the immunoblot section. Donors were analyzed independently and were not pooled. GIMAP6-targeting siRNA duplex sequences are provided in Table S7.

mRNA expression analysis

Total RNA was extracted from flash-frozen heart tissue and BMDMs using the RNeasy Plus Mini Kit according to the manufacturer’s instructions. First-strand cDNA synthesis was performed using random hexamer primers and the SuperScript III First-Strand Synthesis System (Invitrogen). Quantitative real-time PCR was performed using Power SYBR Green PCR Master Mix (Thermo Fisher Scientific) on a QuantStudio 7 Flex Real-Time PCR System. For cardiac remodeling analysis, target genes included hypertrophy-associated genes Mybpc3, Nppa, Tnnt2, and Myh7 and fibrosis-associated genes Acta2 and Col1a1. For cardiac lipid-metabolism analysis, target genes included Cd36, Fabp3, Cpt1b, Acadm, Acadvl, Ppara, and Ppargc1a. For macrophage lipid-efflux analysis, target genes included Abca1 and Abcg1. Gene expression was normalized to Actb unless otherwise indicated. Relative expression was calculated using the ΔΔCt method. Each biological data point represents one mouse or one independently prepared cell culture. Primer sequences were obtained from PrimerBank and are provided in Table S7.⁶⁸

mRNA-seq analysis

Raw RNA-seq reads were trimmed for quality and adapter sequences, then aligned to the GRCm39 (mm39) reference genome using STAR ⁶⁹ with Ensembl annotations. Gene counts were obtained using HTSeq⁷⁰ using union mode and reverse-stranded counting with Ensembl GTF annotation files. One aortic RNA-seq sample was excluded from downstream analysis because of RNA degradation and insufficient RNA quality before library preparation. Differential expression was analyzed using DESeq2;⁷¹ genes with low read counts (< 10) were excluded, and log₂ fold changes were shrinkage-adjusted. Principal component analysis (PCA) was performed on rlog-transformed data. For non-replicate or small-sample datasets, DEGseq⁷² or edgeR ⁷³ was used. Differentially expressed genes (DEGs), defined as Benjamini-Hochberg adjusted p value < 0.05, were subjected to Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis⁷⁴ using hypergeometric tests. For gene set enrichment analysis, ranked gene lists were analyzed using the WEB-based GEne SeT AnaLysis Toolkit (WebGestalt), and normalized enrichment scores (NES) and false discovery rate (FDR) were reported where applicable.⁷⁵ Results were visualized as clustered heatmaps using the R package pheatmap. Selected DEGs or pathway-associated DEGs were displayed as row-scaled Z scores.

Public single-cell and transcriptomic analyses

Publicly available single-cell RNA sequencing (scRNA-seq) datasets from mouse heart and aorta were obtained from the Broad Institute’s Single Cell Portal.³⁰ Datasets were selected based on tissue relevance and the availability of processed gene expression matrices and cell-type annotations. Normalized gene expression values and corresponding cell-type labels were downloaded directly from the portal. All visualizations, including t-distributed stochastic neighbor embedding (t-SNE) plots and violin plots, were generated using the portal’s built-in tools. Expression of Gimap6 (mouse) or GIMAP6 (human) was queried across annotated cell populations. No additional filtering, normalization, or reclustering was performed; all analyses reflect the original processing pipelines provided with each dataset. The datasets analyzed included human thoracic aorta (SCP1265),⁷⁶ mouse aorta (SCP289),⁷⁷ and mouse heart (SCP283).⁷⁸

Human vascular cell-type expression of GIMAP6 was additionally queried from the Human Protein Atlas vascular single-cell resource. For coronary artery disease-associated tissue expression, GIMAP6 expression in CAD and healthy-control tissues was obtained from the STARNET database (http://starnet.mssm.edu/). For peripheral blood monocyte expression across CAD stages, GIMAP6 and PPARG expression values were obtained from GEO dataset GSE166780, which includes individuals with normal coronary arteries, intermediate coronary lesions, or acute myocardial infarction. Correlation between GIMAP6 and PPARG expression was assessed using Pearson’s correlation coefficient. For human carotid atherosclerotic plaque macrophage states, processed single-cell RNA-seq data were obtained from the SICOF carotid plaque portal. Expression of CD36, PPARG, and GIMAP6 was examined across annotated macrophage subclusters, including F13A1-positive monocyte-like macrophage clusters and SPP1-positive foam-cell-associated clusters. For sex-biased expression analyses, human tissue gene-expression values, reported as transcripts per million, were obtained from the GTEx Portal, Analysis Release V10. Mouse immune-cell expression data were obtained from the Immunological Genome Project portal. Mouse aortic endothelial cell expression data were obtained from GEO dataset GSE247364.⁴⁹ Female and male samples were analyzed separately where sex information was available. Expression values were visualized as violin plots or bar graphs, as indicated in the corresponding figures.

BMDM polar metabolomics

For BMDM metabolomics, differentiated bone marrow-derived macrophages from WT and Gimap6^−/− mice were seeded at 7 × 10⁵ cells per well in 6-well plates. After 24 h, cells were washed once with ice-cold PBS and extracted in 1 mL ice-cold 80% methanol/20% water. Extracts were transferred to 1.5 mL microcentrifuge tubes and centrifuged at 16,000 × g for 10 min at 4°C. Supernatants were collected and dried under nitrogen at 4°C. Dried extracts were resuspended in 200 μL 40% acetonitrile/40% methanol/20% water containing ¹³C- and ²H-labeled internal standards listed in Table S8. Polar metabolites were analyzed in positive and negative electrospray ionization modes using hydrophilic interaction chromatography. BMDM samples were analyzed using a Vanquish UPLC system coupled to an Exploris 480 mass spectrometer. Chromatographic separation was performed using a SeQuant ZIC-pHILIC column, 5 μm, 200 Å, 150 × 2.1 mm. Mobile phase A consisted of 20 mM ammonium carbonate in water, pH 9.2, containing 1 μM medronic acid. Mobile phase B consisted of 95% acetonitrile and 5% mobile phase A. The flow rate was 150 μL/min, and the injection volume was 2 μL. Data were acquired in data-dependent acquisition mode. MS1 acquisition on the Exploris 480 was performed at 90,000 resolution with an AGC target of 100%, maximum injection time of 100 ms, and scan range of m/z 70–1,050. MS2 acquisition was performed at 15,000 resolution with an AGC target of 100%, maximum injection time of 75 ms, 1 s cycle time, 1.2 m/z isolation window, and dynamic exclusion of 8 s. Normalized collision energies were 18, 32, and 59 in positive ion mode and 18, 38, and 73 in negative ion mode.

Heart metabolomics and lipidomics

For heart metabolomics and lipidomics, mice were terminally anesthetized with isoflurane, hearts were rapidly harvested, and euthanasia was completed by decapitation. Hearts were dissected, snap-frozen in liquid nitrogen, and stored at −80°C. Frozen tissue sections of 20 mg were pulverized using a cryogenic ball mill at 25 Hz for 15 s. Tissue powder was extracted in 500 μL ice-cold 80% methanol/20% water containing internal standards. Samples were vortexed twice for 15 s during a 10 min incubation on ice. A total of 300 μL ice-cold water was added, and samples were extracted twice with 800 μL methyl tert-butyl ether. Organic and aqueous fractions were collected separately and dried under nitrogen at 4°C.

Polar metabolites were resuspended in 200 μL 40% methanol/40% acetonitrile/20% water containing internal standards. Nonpolar metabolites and lipids were resuspended in 200 μL 60% butanol/20% methanol/20% water containing internal standards. Heart polar metabolites were analyzed using the same hydrophilic interaction chromatography method described above, with a 5 μL injection volume, on a Q Exactive Plus mass spectrometer operated in data-dependent acquisition mode. MS1 acquisition was performed at 70,000 resolution with a maximum injection time of 100 ms and a scan range of m/z 70–1,050. MS2 acquisition was performed at 17,500 resolution with an AGC target of 1 × 10⁵, maximum injection time of 50 ms, loop count of 6, isolation window of 1 m/z, underfill ratio of 1.00%, apex trigger of 3–10 s, and dynamic exclusion of 25 s. Normalized collision energies were 20, 40, and 80 in positive ion mode and 20, 50, and 100 in negative ion mode. Heart nonpolar metabolites and lipids were analyzed in negative ionization mode using an Accucore C30 column, 250 × 2.1 mm, 2.6 μm, coupled to an Orbitrap ID-X mass spectrometer. Mobile phase A consisted of 20 mM ammonium formate in 60:40 acetonitrile:water containing 0.25 μM medronic acid. Mobile phase B consisted of 20 mM ammonium formate in 90:10 isopropanol:acetonitrile containing 0.25 μM medronic acid. The flow rate was 200 μL/min, the injection volume was 5 μL, and the column temperature was 35°C. MS1 data were acquired at 120,000 resolution with an AGC target of 4 × 10⁵, maximum injection time of 50 ms, and scan range of m/z 200–2,000. MS2 acquisition was performed at 15,000 resolution with loop count of 10, AGC target of 5 × 10⁴, maximum injection time of 150 ms, isolation window of 1 m/z, dynamic exclusion of 10 s, and normalized collision energies of 25, 35, and 45.

LC-MS data processing

Raw LC-MS files were converted to mzML format using ProteoWizard v3. Compound identification and peak grouping were performed using the R package OpenCLaM. Peaks were matched within 10 ppm for precursor masses and 20 ppm for fragment masses. For metabolomics, fragmentation spectra and retention times were compared with an in-house library generated from authentic standards. For lipidomics, fragmentation spectra were compared with an in-house in silico lipid library. Automated annotations were manually inspected, validated, and corrected where necessary using MAVEN2 v2.11.2. Peak intensity was quantified using the sum of the maximum peak intensity and the immediately preceding and following scans, referred to as “peak area top.” Peak area top values were log₂ transformed. Peaks below the limit of detection were assigned a lower-bound value of 12 after log₂ transformation. For each identified metabolite or lipid, genotype effects were tested using linear models in R. For heart analyses, the model was log₂ abundance = β0 + β1 genotype, where genotype was coded as 0 for Ldlr^−/− and 1 for Gimap6^−/−Ldlr^−/−. For BMDM analyses, the same model structure was used with genotype coded as 0 for WT and 1 for Gimap6^−/−. False discovery rates were calculated using the qvalue R package, which implements Storey’s method. Features with q < 0.05 were considered significant. Heatmaps were visualized as row-scaled Z scores, and selected pathway summaries were displayed as log₂ fold change with dot size reflecting statistical significance.

Seahorse metabolic analysis

Oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were measured using a Seahorse XFe96 Analyzer (Agilent) and the Seahorse XF Cell Mito Stress Test Kit (Agilent) according to the manufacturer’s instructions. Differentiated BMDMs were generated from 7–10-week-old female WT and Gimap6^−/− mice. BMDMs were detached using Accutase (BD Biosciences) and seeded at 1 × 10⁵ cells per well onto poly-D-lysine-coated Seahorse XFe96/XF Pro microplates (Agilent). Cells were allowed to adhere overnight at 37°C in a humidified 5% CO₂ incubator. Before analysis, culture medium was replaced with phenol red-free Seahorse XF DMEM assay medium (Agilent) supplemented with 1 mM sodium pyruvate, 2 mM L-glutamine, and 10 mM D-glucose. Cells were equilibrated for 1 h at 37°C in a non-CO₂ incubator before the assay. Basal OCR and ECAR were recorded, followed by sequential injections of oligomycin at 2 μM, FCCP at 2 μM, and rotenone plus antimycin A at 0.5 μM each. OCR and ECAR values were normalized to cell number. Basal respiration, ATP-linked respiration, maximal respiration, spare respiratory capacity, and peak ECAR were calculated using Seahorse XF Analytics software. Experiments were performed across three independent BMDM preparations. Technical replicate wells were included for each genotype in each experiment. Data were analyzed using identical assay settings across groups. No wells or samples were excluded from analysis unless there was technical injection failure or visible loss of the cell monolayer.

GW3965 LXR activation assay

LXR activation experiments were performed using differentiated BMDMs from 7–10-week-old female WT and Gimap6^−/− mice. BMDMs were seeded in 24-well plates at 5 × 10⁵ cells per well in DMEM supplemented with 10% FBS and 1% penicillin-streptomycin and allowed to adhere overnight. For lipid-efflux gene induction assays, BMDMs were treated with the LXR agonist GW3965 (MCE) at 5 μM for 24 h. Vehicle-treated control wells received the same final concentration of DMSO. Target genes included Abca1, Abcg1, and Nr1h3. Gene expression was normalized to Actb, and relative expression was calculated using the ΔΔCt method. Primer sequences are provided in Table S7. For lipid-accumulation assays, BMDMs were pretreated with GW3965 at 5 μM for 2 h before oxLDL loading. GW3965 was maintained in the medium during oxLDL stimulation. OxLDL (Thermo Fisher Scientific) was added at 50 μg/mL, and cells were incubated for 24 h at 37°C in 5% CO₂. Vehicle-treated control wells received matched DMSO. Oil Red O (Abcam) staining was then performed as described in the foam-cell formation assay section. Lipid accumulation was quantified using ImageJ as Oil Red O-positive lipid area per field normalized to WT.

Rare variant association data from the Broad CVDI portal

Gene-level rare variant association data were obtained from the Broad Cardiovascular Disease Initiative (CVDI) Human Disease Portal.⁵⁰ This public resource provides pre-computed results from meta-analyses combining whole-genome sequencing data from the All of Us cohort (n = 242,902) with whole-exome sequencing data from the UK Biobank (n = 454,162) and Mass General Brigham (MGB) biobank (n = 51,815). All data were derived from the portal's combined UKB_450k_AoU_250k_MGB_53k cohort analysis.

From the portal, we selected gene-level burden test results generated using a previously defined variant aggregation strategy. This strategy included high-confidence loss-of-function (LOF) variants (frameshift indels, stop-gain, and splice site-disrupting variants), excluding those flagged as low-confidence by the Loss-Of-Function Transcript Effect Estimator (LOFTEE). It also incorporated damaging missense variants scored using the fraction of 30 prediction algorithms in the dbNSFP database that classified a variant as deleterious. For our analysis, we downloaded results derived from a combined variant set of high-confidence LOF and missense variants with a damaging prediction score ≥ 0.5 (designated by the portal as LOF+missense0.5).

To ensure adequate statistical power for our genes of interest, we utilized the results generated with a minor allele frequency (MAF) filter of < 0.1% for the LOF+missense0.5 variant set. This filter was chosen because alternative, more stringent filters (e.g., MAF < 0.001%) resulted in a combined allele frequency too low for robust analysis. Using these criteria, we obtained association data for the GIMAP6, LDLR, and PCSK9.

GIMAP6 missense variant association analysis in All by All browser (All of Us)

Missense variant filtering and phenotype association analyses were conducted using the All by All browser populated with data from the All of Us Research Program (browser version 0.1.0-20fd1751-202501311616; data freeze July 1, 2022). Variants in GIMAP6 were filtered to include only missense variants with a minor allele frequency (MAF) < 10% and a minimum of 100 carriers, which yielded 11 variants for analysis. Precomputed association statistics were extracted for a range of phenotypes. Primary cardiovascular disease (CVD) phenotypes of interest included coronary atherosclerosis (PheCode: 411.4), ischemic heart disease (PheCode: 411), and heart failure (PheCode: 424, including HFrEF, PheCode: 428.3). We also examined associations with key CVD risk factors, such as hypertension (PheCode: 401) and type 2 diabetes (PheCode: 250.2), and a selection of non-cardiovascular conditions, including chronic kidney disease (PheCode: 582.2), cancer (PheCode: CA_130), asthma (PheCode: 495), autoimmune disease (PheCode: BI_181), infections (PheCode: ID_089), and varicose veins (PheCode: 454). For each variant-phenotype pair, we obtained the effect size (β), p value, and allele count (AC) as reported by the All by All browser. The predicted functional impact of each variant was assessed using Combined Annotation Dependent Depletion (CADD) scores (v1.7).⁷⁹ The results were visualized by plotting the effect size (β) against the CADD score for each variant, with the bubble size proportional to the carrier count and color-coded by statistical significance (p value).

Genetic variant and linkage disequilibrium analysis

Analysis of the genetic relationship between variants was conducted using the LDlink suite of web-based tools and public genomic databases.⁸⁰ The regional linkage disequilibrium (LD) structure around the GIMAP6 V65I variant (7-150628405-C-T) was assessed using the LDproxy module. The analysis was performed using the 1000 Genomes African (AFR) population as the reference panel. Direct pairwise LD between the GIMAP6 V65I variant (rs17173519) and the NOS3 variant (rs3918226) was calculated using the LDpair module. The analysis utilized the same AFR reference population to determine the r² and p values, establishing the genetic independence of the two signals. Allele frequencies for rs3918226 were obtained from the Genome Aggregation Database (gnomAD v4.1.0) to compare ancestry-specific frequency distributions with GIMAP6 V65I.

Age-stratified disease prevalence analysis

Study Population and Data Source:

A cross-sectional analysis was conducted using genomic and electronic health record (EHR) data from the All of Us Research Program Researcher Workbench (Controlled Tier Dataset, version C2024Q3R5). The study cohort was restricted to participants who self-reported as Black or African American (race concept ID: 8516) and had both whole-genome sequencing (WGS) and EHR data available. The age for each participant was calculated as the interval between their date of birth and the date of their last recorded visit, and the final cohort for analysis included individuals aged 18 to 100 years.

Genetic Variant Status:

Carrier status for specific genetic variants was determined from WGS data using the cb_variant_to_person table. The primary variants of interest were GIMAP6 p.Val65Ile (V65I), TTR p.Val142Ile (V142I), and a control variant, GIMAP6 p.Gln237Arg (Q237R). Additionally, carriers of pathogenic or likely pathogenic variants in the LDLR gene were identified by querying the cb_variant_attribute table for variants with a ClinVar clinical significance containing “pathogenic” or “likely pathogenic”. Individuals not identified as carriers for any of these specified variants constituted the non-carrier reference group for all comparative analyses.

Phenotype Definition:

Disease phenotypes were ascertained using curated sets of Systematized Nomenclature of Medicine (SNOMED) and International Classification of Diseases (ICD) concept IDs obtained from the cb_search_all_events table. The presence of a phenotype was defined by the occurrence of one or more of its associated concept IDs in a participant's record. The primary phenotypes analyzed for this figure were coronary atherosclerosis (CAD), heart failure (HF), the composite of both “CAD and HF,” and hypercholesterolemia (HChol).

Statistical Analysis:

The primary analysis focused on evaluating the age-dependent prevalence of diseases among genetic variant carriers and non-carriers. Age-stratified cumulative prevalence was calculated for each phenotype. For each 5-year age increment from 35 to 85, prevalence was computed as the proportion of individuals with the disease among all individuals younger than that age cutoff. This calculation was performed for the non-carrier reference group and for each specified genetic variant carrier group. All cumulative prevalence analyses were performed on the overall cohort and then stratified by sex (male, female). All analyses were performed using custom Python scripts utilizing the Pandas, NumPy, and SciPy libraries.

Age-stratified carrier frequency analysis

Study Population and Data Source:

This cross-sectional analysis utilized genomic and EHR data from the All of Us Research Program Researcher Workbench (Controlled Tier Dataset: vC2024Q3R5). The study cohort was restricted to participants who self-reported as Black or African American (race concept ID: 8516) aged 18 to 100 years, with available whole-genome sequencing (WGS) and EHR data. The cohort was stratified into four age bins, 18–64, 65–74, 75–84, and ≥85 years, to balance the effect of aging and statistical power.

Genetic Variant Definition and APOE Allele Determination:

Carrier status for variants of interest (GIMAP6 p.Val65Ile, TTR p.Val142Ile, GIMAP6 p.Gln237Arg), two APOE SNPs, and a large panel of randomly selected control variants was determined from WGS data using the cb_variant_to_person table. APOE allele carrier status was not directly queried but was inferred from the genotypes of two SNPs: rs429358 (vid: '19-44908684-T-C') and rs7412 (vid: '19-44908822-C-T'). APOE ε4 carrier was defined by the presence of the rs429358 risk allele ('C') and the absence of the rs7412 risk allele ('T').

To generate a control distribution for age-associated allele-frequency changes, 5,000 missense variants were randomly selected from the All of Us cohort with global MAF between 0.1% and 1%, approximating the global frequency of GIMAP6 V65I. This variant pool was then filtered within Black or African American participants to identify variants with carrier frequency between 2% and 4%, approximating the population-specific frequency of V65I. From the 696 variants passing this filter, 100 variants were randomly selected as the final control variant panel.

For each variant, overall carrier frequency was calculated across the full cohort and compared with carrier frequency in each age bin. Normalized frequency change was calculated as: $\text{normalized change}(\%)=[(\text{frequency in age bin}-\text{frequency overall})∕\text{frequency overall}]\times 100$. For each age bin, the mean and 99.99% confidence interval of normalized frequency change among the 100 control variants were calculated to define the expected neutral range. Normalized frequency changes for GIMAP6 V65I, APOE ε4, TTR p.Val142Ile, and GIMAP6 Q237R were plotted relative to this control range. Differences between GIMAP6 V65I and the control variant distribution were quantified by Z-score testing. Analyses were performed for the full cohort and in male and female subgroups using custom Python scripts.

Adjusted logistic regression

Adjusted association analyses were performed using genomic and electronic health record data from the All of Us Researcher Workbench Controlled Tier Dataset C2024Q3R5. The analytic cohort was restricted to adult participants who self-reported as Black or African American and had available whole-genome sequencing and electronic health record data. Age at last observation was calculated from date of birth to the date of the last recorded clinical event.

The primary endpoint was the presence of both coronary artery disease (CAD) and heart failure (HF). Cases were defined as participants with at least one coronary artery disease-associated diagnosis code and at least one heart failure-associated diagnosis code, based on curated SNOMED and ICD concept sets. Controls were participants who did not meet the combined CAD+HF case definition. Sensitivity analyses compared CAD+HF cases with participants with neither CAD nor HF and compared CAD-or-HF cases with participants with neither diagnosis.

GIMAP6 V65I carrier status, chr7:150628405 C>T, was determined from WGS data. Multivariable logistic regression was used to test the association between V65I carrier status and disease status. Models were adjusted for age at last observation, sex or gender where applicable, smoking status, BMI category, diabetes, hypertension, hyperlipidemia, and 16 WGS-derived genetic ancestry principal components. Models including sex or gender were restricted to participants recorded as male or female to avoid sparse-category non-convergence. Male-only and female-only models were analyzed separately without sex or gender as a covariate. Additional sensitivity analyses adjusted for EHR follow-up duration and log-transformed EHR event count. Results are reported as odds ratios with 95% confidence intervals. P values from stratified analyses were adjusted using the Benjamini-Hochberg false discovery rate. Analyses were performed using custom Python scripts with statsmodels, pandas, NumPy, and SciPy.

GIMAP6 V65I homozygous analysis

GIMAP6 V65I homozygotes were identified from All of Us Controlled Tier Dataset v8 whole-genome sequencing data aligned to GRCh38. Genotypes were extracted using Hail v0.2.130. Variants were filtered by genomic locus chr7:150628405, reference allele C, and alternate allele T. Homozygous alternate individuals were defined as participants with genotype GT = 1/1. This analysis identified 24 V65I homozygous individuals among 2,868 V65I carriers. Electronic health records of homozygous individuals were reviewed, and participants were required to meet a minimum threshold of EHR activity for phenotype analysis, yielding a final clinical-analysis cohort of fewer than 20 individuals, consistent with All of Us participant disclosure policies.

For the matched case-control analysis, cases were V65I homozygous individuals with at least 5 years of longitudinal EHR coverage. Controls were selected from Black or African American participants with WGS and EHR data, at least 5 years of EHR coverage, and no known GIMAP6 V65I allele. Controls were matched to cases at a 100:1 ratio without replacement by exact sex at birth, age at first EHR visit within ±5 years, and total EHR duration within ±5 years. Controls were randomly selected using random_state = 1.

Clinical phenotypes were defined using curated SNOMED and ICD concept sets from condition_occurrence. Cardiovascular phenotypes included atherosclerosis and infarction, stroke and cerebrovascular disease, heart failure and enlarged heart, arrhythmias and ECG abnormalities, and angina pectoris. Major cardiovascular disease was defined as a composite of infarction, stroke, and heart failure. Background comorbidities included hypercholesterolemia, diabetes, hypertension, and chronic kidney disease.

Baseline age was compared using an independent t test, and sex distribution was compared using Fisher’s exact test. Phenotype associations were tested using Fisher’s exact test, and odds ratios and 95% confidence intervals were calculated. Continuity correction was applied for confidence intervals when groups contained zero events. A p value < 0.05 was considered significant.

Permutation testing

Permutation testing was used to assess whether cardiovascular disease prevalence among GIMAP6 V65I homozygotes exceeded that expected by chance. A matched master pool was generated by identifying all potential controls that matched at least one case by sex, age at first EHR visit within ±5 years, and duration of EHR coverage within ±5 years. This matched control pool was combined with the case cohort. The observed test statistic was the number of individuals in the true case cohort with the phenotype of interest.

A null distribution was generated using 10,000 permutations. In each permutation, a random group of individuals equal in size to the case cohort was sampled without replacement from the matched master pool, and the number of individuals with the phenotype was counted. A one-sided empirical p value was calculated as the proportion of permutations in which the number of affected individuals was greater than or equal to the observed number in the true case cohort. A p value < 0.05 was considered significant.

Phenome-wide association study (PheWAS)

Phenome-wide association analysis was performed using genomic and electronic health record data from the All of Us Research Program Researcher Workbench Controlled Tier Dataset C2024Q3R5. The cohort was restricted to adult participants aged 18–100 years who self-reported as Black or African American, race concept ID 8516, and had available whole-genome sequencing and EHR data. For the primary PheWAS, the cohort was further restricted to participants aged ≤70 years at the last recorded clinical visit to enrich for earlier-onset disease.

GIMAP6 V65I carrier status was determined from the cb_variant_to_person table using variant ID 7-150628405-C-T. Clinical phenotypes were defined using curated SNOMED and ICD concept sets and grouped into primary cardiovascular diseases, cardiometabolic conditions, negative-control diseases, and composite phenotypes. Primary cardiovascular phenotypes included coronary atherosclerosis, heart failure, and major infarction events. Cardiometabolic phenotypes included hypercholesterolemia, diabetes, hypertension, and chronic kidney disease. Negative-control phenotypes included rheumatoid arthritis, Alzheimer’s disease, and Parkinson’s disease.

For each phenotype, cases were defined as participants with at least one corresponding diagnosis code, and controls were all other participants without that phenotype. Associations between V65I carrier status and each phenotype were tested using Fisher’s exact test. Odds ratios, 95% confidence intervals, and p values were calculated for each phenotype. Benjamini-Hochberg correction was applied across tested phenotypes to control false discovery rate. FDR-adjusted p value < 0.05 was considered significant. Results were visualized as forest plots.

Public cohort validation of GIMAP6 V65I associations

Validation of GIMAP6 V65I associations in populations with African ancestry was performed using publicly available summary statistics from independent cohorts. Single-variant association results for GIMAP6 V65I in individuals with genetically inferred African ancestry (GIA) were obtained from the Million Veteran Program (MVP) Genome-wide×Phenome-wide Association Study (gwPheWAS) via the Phenomics VA portal.⁸¹ Single-variant association results for GIMAP6 V65I were additionally queried for UK Biobank participants with African ancestry using the AZ PheWAS portal.⁵⁷ Only summary statistics provided by the public portals were analyzed. No individual-level data from these external cohorts were accessed or reanalyzed.

QUANTIFICATION AND STATISTICAL ANALYSIS

All statistical analyses were performed using GraphPad Prism v10.2.3 unless otherwise specified. Data are presented as mean ± SD or mean ± SEM, as indicated in the figures. The value of n represents independent biological replicates, individual animals, or independent experiments.

For comparisons between two groups, two-tailed unpaired Student’s t tests or Mann–Whitney U tests were used as appropriate. Comparisons among multiple groups were performed using one-way or two-way ANOVA with appropriate post hoc multiple-comparison correction. Survival analyses were performed using the log-rank (Mantel–Cox) test. Correlations were assessed using Pearson’s correlation coefficient. Statistical analyses used for individual experiments are indicated in the corresponding figures.

For human genetic analyses, multivariable logistic regression was used to calculate odds ratios (ORs) while adjusting for relevant covariates. Fisher’s exact test was used for case-control comparisons. Pearson correlation was used for association analyses. Z-score testing was used to compare observed allele frequencies with matched control variant distributions. Multiple testing was controlled using the Benjamini– Hochberg false discovery rate (FDR) procedure where indicated. Disease enrichment among V65I homozygotes was assessed using a non-parametric permutation test (10,000 iterations).

A two-sided p value < 0.05 was considered statistically significant. Significance is indicated as ns, not significant; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

For animal studies, age- and sex-matched littermate controls were co-housed and randomly assigned to experimental groups where applicable. Sample sizes were based on previous studies and pilot experiments. Image acquisition and quantification were performed by investigators blinded to genotype, sex, diet, or treatment condition where applicable. Samples were excluded only for pre-specified technical reasons, such as failed monolayer formation, physical damage to assay inserts, insufficient sample quality, or RNA degradation before analysis. Image analysis was performed using ImageJ v1.54p, flow cytometry data were analyzed using FlowJo v10.10.0, and RNA-seq analyses were performed as described in the mRNA-seq analysis section.

Supplementary Material

1

Figures S1–S12. Supplemental figures, related to Figures 1-7.

2

Table S1. Longitudinal echocardiographic measurements in male and female Gimap6 knockout and control mice, related to Figure 1.

3

Table S2. Cardiovascular evaluation and longitudinal lipid profile of the GIMAP6 G153V homozygous patient (Pt. 1), related to Figure 6.

4

Table S3. Summary of GIMAP6 I44T homozygous patient (Pt. 2) findings, related to Figure 6.

5

Table S4. Summary of GIMAP6 V126M homozygous patient (Pt. 3) findings, related to Figure 6.

6

Table S5. Raw counts and normalized age-stratified frequency changes for variants in All of Us participants self-identified as Black or African American, related to Figure 7.

7

Table S6. Association of GIMAP6 V65I with cardiovascular, metabolic, and chronic disease phenotypes in All of Us participants self-identified as Black or African American, related to Figure 7.

8

Table S7. Oligonucleotides used in this paper, related to STAR Methods.

9

Table S8. Metabolomics appendix, related to STAR Methods.

10

Video S1. Serial echocardiographic measurements of WT and Gimap6^−/− mice, related to Figure 1.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit monoclonal anti-CD68 (E3O7V)	Cell Signaling Technology	Cat# 97778; RRID: AB_2928056
Mouse monoclonal anti-alpha-smooth muscle actin (1A4)	R&D Systems	Cat# MAB1420; RRID: AB_262054
Goat anti-mouse IgG (H+L), Alexa Fluor Plus 488	Thermo Fisher Scientific	Cat# A32723; RRID: AB_2633275
Goat anti-rabbit IgG (H+L), Alexa Fluor Plus 647	Thermo Fisher Scientific	Cat# A32733; RRID: AB_2633282
Rat anti-human GIMAP6 (MAC445)	Lab of Geoffrey W. Butcher, Babraham Institute	Pascall et al.67
Rat anti-mouse GIMAP6 (MAC431)	Lab of Geoffrey W. Butcher, Babraham Institute	Pascall et al.13
Rat anti-mouse GIMAP6 (MAC432)	Lab of Geoffrey W. Butcher, Babraham Institute	Pascall et al.13
Rat anti-mouse GIMAP6 (MAC434)	Lab of Geoffrey W. Butcher, Babraham Institute	Pascall et al.13
Mouse monoclonal anti-FLAG M2	Sigma-Aldrich	Cat# F3165; RRID: AB_259529
Rabbit polyclonal anti-PPARgamma	Proteintech	Cat# 16643-1-AP; RRID: AB_10596794
Rabbit monoclonal anti-CD36 (D8L9T)	Cell Signaling Technology	Cat# 14347; RRID: AB_2798458
Rabbit monoclonal anti-CD36 (E8B7S)	Cell Signaling Technology	Cat# 28109; RRID: AB_3675401
Rat monoclonal anti-beta-actin (W16197A)	BioLegend	Cat# 664802; RRID: AB_2721349
Mouse monoclonal anti-HSP90alpha/beta (3H3C27)	BioLegend	Cat# 675402; RRID: AB_2565512
Rabbit anti-IL-6	Proteintech	Cat# 26404-1-AP; RRID: AB_3085866
Rabbit anti-MCP-1	Proteintech	Cat# 26161-1-AP; RRID: AB_2918100
FITC anti-human CD36	BioLegend	Cat# 336204; RRID: AB_1575025
PE anti-mouse CD36	BioLegend	Cat# 102605; RRID: AB_389348
PE anti-human CD36	BioLegend	Cat# 336206; RRID: AB_2072513
FITC anti-human CD14	BioLegend	Cat# 325604; RRID: AB_830677
Brilliant Violet 510(TM) anti-human CD16	BioLegend	Cat# 302048; RRID: AB_2562085
Brilliant Violet 421(TM) anti-mouse CD31	BioLegend	Cat# 102424; RRID: AB_2650892
OLR1 antibody	Proteintech	Cat# 11837-1-AP; RRID: AB_2299037
MSR1 antibody	Proteintech	Cat# 17858-1-AP; RRID: AB_2878455
Goat polyclonal anti-rat Ig-HRP	SouthernBiotech	Cat# 3010-05; RRID: AB_2795801
Goat polyclonal anti-rabbit Ig-HRP	SouthernBiotech	Cat# 4050-05; RRID: AB_2795955
Goat polyclonal anti-mouse Ig-HRP	SouthernBiotech	Cat# 1010-05; RRID: AB_2728714
PowerVision Poly-HRP anti-Rabbit IHC Detection Systems	Leica biosystems	Cat# PV6119; RRID: AB_3751740
Bacterial and virus strains
Stellar Competent Cells	Takara	Cat# 636766
Biological samples
Healthy human PBMCs	This paper	NIH Blood Bank / NIH protocol 05-I-0213 / NCT00128973
Patient and family PBMCs (Pt 1)	This paper	NIH protocol 06-I-0015/ NCT00246857
Patient PBMCs (Pt 2)	This paper	NIH protocol 01-I-0202 / NCT00018044
Patient sample (Pt 3)	This paper	Protocol IR.TUMS.CHMC.REC.1398.030
V65I carrier samples	This paper	NIH protocol 03-H-0015 / NCT00047996 and 09-I-0032 / NCT00794352
Chemicals, peptides, and recombinant proteins
MyTaq Red Mix	Meridian	Cat# BIO-25043
2xCloneAmp HiFi PCR Premix	Takara	Cat# 639298
DNeasy Blood and Tissue Kit	Qiagen	Cat# 69506
RPMI-1640	Gibco	Cat# 11875
DMEM	Gibco	Cat# 11965
DMEM/F-12	Gibco	Cat# 11320033
Fetal bovine serum	Gibco	Cat# A5670
Penicillin-streptomycin	Gibco	Cat# 15140122
L-glutamine	Gibco	Cat# 25030
Trypsin-EDTA	Gibco	Cat# 25300054
Accutase	BD Biosciences	Cat# 561527
Ficoll-Paque PLUS density gradient medium	Cytiva	Cat# 17144002
CD14 MicroBeads, human	Miltenyi Biotec	Cat# 130-050-201
Formalin solution, neutral buffered, 10%	Sigma-Aldrich	Cat# HT5012
2,3,5-triphenyltetrazolium chloride	Sigma-Aldrich	Cat# T8877
Permount	Fisher Scientific	Cat# SP15
Oil Red O	Abcam	Cat# ab146295
Evans Blue	Sigma-Aldrich	Cat# E2129
Formamide	Sigma-Aldrich	Cat# F9037
Tissue-Tek O.C.T. compound	Sakura	Cat# 4583
DAPI	GlpBio	Cat# GC43378
Fluoromount-G	SouthernBiotech	Cat# 0100-01
Citrate buffer, pH 6.0	BioLegend	Cat# 828502
Triton X-100	Sigma Aldrich	Cat# T8532-500ML
Bovine serum albumin	Sigma-Aldrich	Cat# A7030-10G
Pierce™ DAB Substrate Kit	Thermo Fisher Scientific	Cat# 34002
Mayer′s Hematoxylin Solution	Millipore	Cat# MHS32
Power SYBR Green PCR Master Mix	Thermo Fisher Scientific	Cat# 4368702
SuperScript III First-Strand Synthesis System	Invitrogen	Cat# 18080-051
Opti-MEM I Reduced Serum Medium	Thermo Fisher Scientific	Cat# 31985062
Lipofectamine 3000 transfection reagent	Thermo Fisher Scientific	Cat# L3000075
RIPA Lysis and Extraction Buffer	Thermo Fisher Scientific	Cat# 89901
Halt Protease Inhibitor Cocktail	Thermo Fisher Scientific	Cat# 78438
Halt Phosphatase Inhibitor Cocktail	Thermo Fisher Scientific	Cat# 78427
NuPAGE LDS Sample Buffer, 4x	Thermo Fisher Scientific	Cat# NP0008
NuPAGE Reducing Agent, 10x	Thermo Fisher Scientific	Cat# NP0004
NuPAGE 4%-12% Bis-Tris Protein Gels	Thermo Fisher Scientific	Cat# NP0336BOX
Nitrocellulose membrane	Bio-Rad	Cat# 1620112
EveryBlot Blocking Buffer	Bio-Rad	Cat# 12010020
Immobilon Classico Western HRP substrate	Millipore	Cat# WBLUC0500
Immobilon Forte Western HRP substrate	Millipore	Cat# WBLUF0500
SuperSignal West Femto Maximum Sensitivity Substrate	Thermo Fisher Scientific	Cat# 34096
SuperSignal West Dura Extended Duration Substrate	Thermo Fisher Scientific	Cat# 34076
Restore Western Blot Stripping Buffer	Thermo Fisher Scientific	Cat# 21063
Cycloheximide	GlpBio	Cat# GC17198
Recombinant mouse M-CSF	GenScript	Cat# Z02930
Recombinant Mouse TNF-α (carrier-free)	BioLegend	Cat# 575202
Matrigel Growth Factor Reduced Basement Membrane Matrix	Corning Life Sciences	Cat# 356231
Endothelial cell growth supplement	Sigma-Aldrich	Cat# E2759
Heparin sodium salt	Sigma-Aldrich	Cat# H3393
Dispase I	Sigma-Aldrich	Cat# D4818
Fibronectin	Sigma-Aldrich	Cat# F1141
Sulfosuccinimidyl oleate	Cayman Chemical	Cat# 11211
HEPES sodium salt	Sigma-Aldrich	Cat# H7006
ITS Liquid Media Supplement (100×)	Sigma-Aldrich	Cat# I3146
BODIPY FL C16	Thermo Fisher Scientific	Cat# D3821
BODIPY 493/503	GLPbio Technology Inc	Cat# GC42959
MitoSOX Red mitochondrial superoxide indicator	Thermo Fisher Scientific	Cat# M36008
MitoTEMPO	MCE	Cat# HY-112879
N-acetyl-L-cysteine	Millipore Sigma	Cat# A9165-5G
GW9662	MCE	Cat# HY-16578
GW3965	MCE	Cat# HY-10627
Oxidized Low-density Lipoprotein (OxLDL)	Thermo Fisher Scientific	Cat# L34357
Pierce™ Dimethylsulfoxide (DMSO), Sequencing Grade	Thermo Fisher Scientific	Cat# 20688
eBioscience™ Fixable Viability Dye eFluor™ 450	Thermo Fisher Scientific	Cat# 65-0863-14
Human TruStain FcX	BioLegend	Cat# 422302
Nuclease-Free Duplex Buffer	Integrated DNA Technologies	Cat# 11-01-03-01
ATTO550-labeled tracrRNA	Integrated DNA Technologies	Cat# 1075927
Cas9 protein	MacroLab, UC Berkeley
RNase-free siRNA duplex resuspension buffer	OriGene	Cat# SR30005
Non-targeting control siRNA	OriGene	Cat# SR30004
GIMAP6 siRNA pool	OriGene	Cat# SR318628
SPLASH Lipidomix Mass Spec Standard	Avanti Polar Lipids	Cat# 330707
Methanol, LC-MS grade	Fisher Scientific	Cat# A456
Isopropanol, LC-MS grade	Fisher Scientific	Cat# A461
Acetonitrile, LC-MS grade	Fisher Scientific	Cat# A955
Water, LC-MS grade	Fisher Scientific	Cat# W6212
Chloroform, LC-MS grade	Sigma-Aldrich	Cat# 650471-1L
Methyl tert-butyl ether	Honeywell	Cat# 34875-1L
Butylated hydroxytoluene	Sigma-Aldrich	Cat# B1253-500G
Ammonium formate	Thermo Fisher Scientific	Cat# 014517.30
Ammonium carbonate	Thermo Fisher Scientific	Cat# 196655000
Formic acid, LC-MS grade	Thermo Fisher Scientific	Cat# 28905
Medronic acid	MCE	Cat# HY-108309
Sodium Pyruvate Solution	MCE	Cat# HY-K3024
D-glucose	Sigma-Aldrich	Cat# G8270
Poly-L-lysine solution	Sigma-Aldrich	Cat# P8920
FORANE (isoflurane, USP)	Baxter	Cat# 1001936060
High-fat diet, 40% kcal fat, 1.25% cholesterol	Research Diets	Cat# D12108C
Critical commercial assays
Mouse Cardiac Troponin-I ELISA Kit	MyBioSource	Cat# MBS766175
Masson’s Trichrome Stain Kit	VitroView	Cat# VB-3016
RNeasy Plus Mini Kit	Qiagen	Cat# 74136
SG Cell Line 4D-Nucleofector X Kit L	Lonza	Cat# V4XC-3012
P3 Primary Cell 4D-Nucleofector X Kit S	Lonza	Cat# V4XP-3032
Cholesterol E CHOD-DAOS Method Kit	FUJIFILM Wako	Cat# 999-02601
L-Type Triglyceride M GPO-HMMPS Method Kit	FUJIFILM Wako	Cat# 994-02891
In Vitro Vascular Permeability Assay, 96-well	Millipore/Merck	Cat# ECM642
GeniePlex Mouse Inflammation 5-Plex Panel 1	Assay Genie	Cat# MOAMPM016
Seahorse XF Cell Mito Stress Test Kit	Agilent	Cat# 103015-100
Deposited data
International Mouse Phenotyping Consortium	IMPC	MP:0000274
CZ CELLxGENE Discover	Chan Zuckerberg Initiative	https://cellxgene.cziscience.com/gene-expression
scRNA-seq dataset of human thoracic aorta	Pirruccello et al.76	SCP1265
scRNA-seq dataset of mouse aorta	Kalluri et al. 77	SCP289
scRNA-seq dataset of mouse heart	Hu et al.78	SCP283
scRNA-seq dataset of carotid plaques	SICOF portal	Mocci et al.43; https://sicof.medh.ki.se/athero/carotid/index.html
Multi-tissue transcriptomic, eQTL, and GRN dataset from CAD patients	STARNET portal	Koplev et al.40 ; http://starnet.mssm.edu/
RNA-seq dataset of Gimap6-deficient BMDMs	This paper	GEO: GSE307778
RNA-seq dataset of Gimap6-deficient Aortae	This paper	GEO: GSE335627
CAD-stage-specific monocyte RNA-seq dataset	Wang et al.42	GEO: GSE166780
Mouse AEC sex-comparison RNA-seq dataset	Shin et al.49	GSE247364
GTEx Analysis Release v10	GTEx Portal	dbGaP phs000424.v10.p2
Pt 1 and Pt 1’s family Whole-exome sequencing (WES)	Yao et al.15	dbGaP phs002816.v1.p1
ImmGen expression dataset	Immunological Genome Project	https://www.immgen.org/
CADD v1.7	Schubach et al.79	https://cadd.gs.washington.edu/
gnomAD v4.1.0	Broad Institute	https://gnomad.broadinstitute.org/
GIMAP6 ASCVD All of Us analysis code	This paper	https://github.com/xiangchennih-25/GIMAP6_CVD_Analysis-; Zenodo DOI https://doi.org/10.5281/zenodo.17100995
All of Us Controlled Tier Dataset v8	All of Us Research Program	Researcher Workbench
All of Us Controlled Tier Dataset C2024Q3R5	All of Us Research Program	Researcher Workbench
All by All browser	All of Us Research Program	Version 0.1.0-20fd1751-202501311616; data freeze July 1, 2022
UK Biobank 500K WGS v2 Public Release	AstraZeneca PheWAS Portal / UK Biobank	https://azphewas.com/
Million Veteran Program gwPheWAS	Million Veteran Program	https://phenomics.va.ornl.gov/web/cipher/pheweb
Human Protein Atlas vascular single-cell resource	Human Protein Atlas	https://www.proteinatlas.org/ENSG00000133561-GIMAP6/single+cell
Broad CVDI Human Disease Portal	Broad Institute	https://hugeampkpn.org/research.html?pageid=600_diseases_home
Experimental models: Cell lines
THP-1	ATCC	Cat# TIB-202; RRID: CVCL_0006; human male
HEK293T	ATCC	Cat# CRL-3216; RRID: CVCL_0063; human female fetal kidney-derived
THP-1 GIMAP6 knockout cells	This paper	Derived from ATCC TIB-202; confirmed by immunoblot; Pool
THP-1 CD36 knockout cells	This paper	Derived from ATCC TIB-202; confirmed by immunoblot; Pool
THP-1 GIMAP6/CD36 double-knockout cells	This paper	Derived from ATCC TIB-202; confirmed by immunoblot; Pool
Experimental models: Organisms/strains
Mouse: C57BL/6N-Gimap6tm1b(KOMP)Wtsi/JMmucd	KOMP/MMRRC	RRID:MMRRC_047846-UCD
Mouse: B6.129S7-Ldlrtm1Her/J	The Jackson Laboratory / Taconic Biosciences	RRID: IMSR_JAX:002207
Mouse: B6.129S1-Cd36tm1Mfe/J	The Jackson Laboratory	RRID: IMSR_JAX:019006
Mouse: Gimap6−/− Ldlr−/−	This paper	Generated by crossing Gimap6-deficient and Ldlr-deficient mice
Mouse: Gimap6−/− Cd36−/−	This paper	Generated by crossing Gimap6-deficient and Cd36-deficient mice
Oligonucleotides
qPCR primers	PrimerBank / this paper	Sequences in Table S7
GIMAP6 crRNAs	Integrated DNA Technologies	Hs.Cas9.GIMAP6.1.AA; Hs.Cas9.GIMAP6.1.AB; Hs.Cas9.GIMAP6.1.AC; sequences in Table S7
CD36 crRNAs	Integrated DNA Technologies	Hs.Cas9.CD36.1.AA; Hs.Cas9.CD36.1.AB; Hs.Cas9.CD36.1.AC; sequences in Table S7
GIMAP6 Sanger primers	This paper	Forward 5-CAGTTGTGTCACCAGGAGCA-3; reverse 5-TTCTATTGGCCAAGTCTCGCT-3
Alternative Gimap6 mouse genotyping primers	This paper	Sequences listed in Method text
Recombinant DNA
pCI-mCherry-T2A-3xF-GIMAP6-Iso1	GenScript	Custom design
FLAG-tagged human GIMAP6 variant constructs	GenScript	WT, I44T, G153V, V126M, V65I, T78A, Q237R, E12D, Q103R, E32G, S105L, V149F, R194W; Plasmid sequence files: available upon request.
Software and algorithms
LabChart	ADInstruments	Version 8.1.30
Ponemah	Data Sciences International	Version 6.x
MetaboScape	Bruker Daltonics	Version 2023
DataAnalysis	Bruker Daltonics	Version 6.1
FlowJo	FlowJo, LLC	Version 10.10.0
ImageJ	NIH	Version 1.54p
GraphPad Prism	GraphPad Software	Version 10.2.3
Adobe Illustrator	Adobe Inc.	Version 28.7.7
Imaris	Oxford Instruments	Version 10.1.1
SnapGene	SnapGene Software	Version 8.2.2
Seahorse XF Analytics	Agilent	Version 2.6.4.39
R	R Foundation	Version 4.3.3
ProteoWizard	ProteoWizard	Version 3
LDpair	LDlink / NIH	https://ldlink.nih.gov/?tab=ldpair
LDproxy	LDlink / NIH	https://ldlink.nih.gov/?tab=ldproxy
Other
Elute UHPLC system	Bruker Daltonics	Model# Elute UHPLC
timsTOF pro2 mass spectrometer	Bruker Daltonics	Model# timsTOF pro2
Vanquish UPLC system	Thermo Fisher Scientific	Model# Vanquish
Exploris 480 mass spectrometer	Thermo Fisher Scientific	Model# Exploris 480
Q Exactive Plus mass spectrometer	Thermo Fisher Scientific	Model# Q Exactive Plus
Orbitrap ID-X mass spectrometer	Thermo Fisher Scientific	Model# Orbitrap ID-X
ACQUITY UHPLC CSH C18 column	Waters	Cat# 186005297
AKTA Pure FPLC system	GE Healthcare/Cytiva	Model# AKTA Pure
Leica SP8 confocal microscope	Leica Microsystems	Model# SP8
Leica UC7 ultramicrotome	Leica Microsystems	Model# UC7
Hitachi HT7800 TEM	Hitachi High-Tech	Model# HT7800
BD FACSymphony A3	BD Biosciences	Model# FACSymphony A3
VICTOR2 1420 Multilabel Counter	PerkinElmer Life Sciences	Model# VICTOR2 1420
VersaMax Microplate Reader	Molecular Devices	Model# VersaMax
Azure 500 imaging system	Azure Biosystems	Model# Azure 500
QuantStudio 7 Flex Real-Time PCR System	Thermo Fisher Scientific	Model# QuantStudio 7 Flex
Vevo F2 imaging system	FUJIFILM VisualSonics	Model# Vevo F2
UHF46X linear transducer	FUJIFILM VisualSonics	Model# UHF46X
DSI PhysioTel ETA-F10 telemetry transmitter	Data Sciences International	Model# ETA-F10
Octal Bio Amp	ADInstruments	Model# Octal Bio Amp
4D-Nucleofector X Unit	Lonza	Model# 4D-Nucleofector X Unit
Seahorse XFe96 Analyzer	Agilent	Model# XFe96
Seahorse XFe96/XF Pro microplates	Agilent	Cat# 103794-100
Seahorse XF DMEM assay medium, pH 7.4	Agilent	Cat# 103680-100
Steriflip filter unit	Millipore	Cat# SE1M003M00
Superfrost Plus microscope slides	Thermo Fisher Scientific	Cat# 12-550-15
Extra Fine Graefe Forceps	Fine Science Tools	Cat# 11150-10
Loctite Waterproof Sealant	Loctite	Cat# 908570

Highlights.

• GIMAP6 deficiency causes lethal atherosclerotic heart attack without hyperlipidemia

• GIMAP6 protects endothelial barrier integrity and limits monocyte adhesion

• GIMAP6 loss drives mitoROS–PPARγ–CD36-dependent foam-cell formation and inflammation

• Hypomorphic GIMAP6 is linked to human ASCVD risk