Slc7a7 licenses macrophage glutaminolysis for restorative functions in atherosclerosis

Abstract

Atherosclerosis is a life-threatening condition characterized by chronic inflammation of the arterial wall. Atherosclerotic plaque macrophages are key players at the site of disease, where metabolic reprogramming dictates the progression of pathogenesis. Here we show that reduced macrophage glutaminase activity is related to glutaminase (GLS)-1 and not GLS2 expression. While glutamine synthetase serves as a metabolic rheostat controlling nutrient flux into cells in vitro, macrophage restorative functions in the context of atherosclerosis relies more heavily on glutamine influx. Enhanced glutamine flux is largely mediated by the SLC7A7 exchanger in macrophages: Slc7a7-silenced macrophages have reduced glutamine influx and GLS1-dependent glutaminolysis, impeding downstream signalling involved in macrophage restorative functions. In vivo, macrophage-specific deletion of Slc7a7 accelerates atherosclerosis in mice with more complex necrotic core composition. Finally, cell-intrinsic regulation of glutaminolysis drives macrophage metabolic and transcriptional rewiring in atherosclerosis by diverting exogenous Gln flux to balance remodelling and restorative functions. Thus, we uncover a role of SLC7A7-dependent glutamine uptake upstream of glutaminolysis in atherosclerotic plaque development and stability.

Macrophages are plastic sentinels of tissue health and integrity that are able to adapt their phenotype to the environment and undergo flexible metabolic rewiring^1,2,3,4. This is conceptualized in vitro by specific metabolic features of macrophages classically activated by inflammatory cues, or restorative-type macrophages involving reparative macrophages mediated by interleulin-4 (IL-4)-dependent alternative activation and resolving macrophages following the clearance of apoptotic cells (ACs)^5,6,7. Atherosclerotic cardiovascular diseases are the leading cause of morbidity and mortality, and a defining characteristic of macrophages in atherosclerotic lesions is their perturbed metabolic flexibility, which promotes inflammation and impedes restorative functions^1,2,3,4. However, our understanding of macrophage metabolic reprogramming at the molecular level in vivo remains poor. This represents a clinically relevant knowledge-gap that precludes the development of therapeutics targeting macrophage metabolism.

Glucose (Glc) and glutamine (Gln) are considered ‘fuels for the immune system’ because they support most anaplerosis by supplying mitochondria with the necessary molecules for adenosine triphosphate (ATP) production. We and others have shown that a deficiency in solute carrier (SLC)2A1-mediated glycolysis and glutaminase-1 (GLS1)-mediated glutaminolysis in macrophages accelerates the development of atherosclerosis by limiting the production of energy required for their restorative functions^8,9,10. Yet whether glutaminase-2 (GLS2)-mediated glutaminolysis or Gln uptake and synthesis support the functional metabolic reprogramming of these functions beyond glycolysis and canonical glutaminolysis remains unknown^11,12,13.

Here, we combined the histological and transcriptomic analysis of atherosclerotic plaques from patients with integrated in vitro and in vivo analysis of murine models of atherosclerosis. We dissected the unique roles of GLS1 and GLS2 in macrophages and identified critical roles for Gln synthetase and members of the SLC family of proteins in enabling intrinsic glutaminolysis, and thereby macrophage reparative and resolving functions. Our study provides mechanistic insights into the in vivo Gln-dependent metabolic reprograming in restorative macrophages and evidence that supports its targeting as a new treatment strategy for cardiovascular disease.

Results

GLS2 is dispensable for macrophage restorative functions in atherosclerosis

Given the known relevance of macrophage Gls1 expression in atherosclerosis¹⁰, we first asked whether Gls2 played an additive, complementary or redundant role in the disease. We began by measuring Gls2 expression in samples of human carotid artery plaques: in contrast to Gls1 (ref. 10), there was no significant difference in Gls2 expression between unstable and stable plaques (Fig. 1a, left). Similar findings were observed in atheromatous plaques of western diet (WD)-fed atherogenic mice (Extended Data Fig. 1a) on the basis of publicly available gene expression datasets^14,15. Nevertheless, we found a slight but significant positive correlation between Gls2 expression and that of the reparative M2-specific marker Arg1 (Fig. 1a, right).

Fig. 1 |.

Gls2 is dispensable for macrophage restorative functions in atherosclerosis. a, The Gls2 expression as the mean ± s.e.m. and associations in stable and unstable human carotid artery plaques (16 stable and 27 unstable carotid artery segments from symptomatic patients). The density of microvessel, endothelial cell, lymphocyte (T cell), macrophage (MΦ) and SMC density (n mm–2); plaque size (mm2); collagen content, calcification, macrophage coverage and subsets (%). IPH, intraplaque haemorrhage. b, A schematic representation of transgenic mice (left): MyeΔGls1 mice lack Gls1 in their myeloid cells; Gls2−/− mice lack Gls2 in all tissues; MyeΔGls1 Gls2−/− mice have a total knockout for Gls2 and a myeloid-cell-specific Gls1 knockout (DKO). Quantification of Glu levels in basal and reparative BMDMs from these mice (right). c, The OCR measured by Seahorse in basal or reparative BMDMCtl, BMDMΔGls1, BMDMΔGls2 or BMDMΔDKO (P values displayed for BMDMΔGls1 and BMDMΔDKO versus BMDMCtl).d–f, ATP production measured by bioluminescence assay (d), cell surface expression of CD206 (e) and effferocytic index measured by flow cytometry (f) after 45 min exposure with ACs in these BMDMs. The efferocytic index was calculated as follows: (number of macrophages with AC/total number of macrophages) x 100. g, The efferocytic index measured by flow cytometry after 45 min of AC exposure ex vivo in TIM-4− and TIM-4+ resident peritoneal cavity macrophages (PCMs) of female Ldlr−/− recipient mice that received BM from control, MyeΔGls1, Gls2−/− and DKO mice. All values are means ± s.e.m. and representative of at least one experiment (n = 12 BMDMCtl, n = 10 BMDMΔGls1, n = 10 BMDMΔGls2 and n = 5 BMDMΔDKO (b), n = 3 cultures per genotype (c–f), n = 7 PCMCtl, n = 4 PCMΔGls1, n = 5 PCMΔGls2 and n = 4 PCMΔDKO (g) of biologically independent replicates). h,i, Representative sections (h) and quantification (i) of aortic plaques and necrotic cores (expressed as a percentage of total roots area) from Ldlr−/− mice transplanted with BMCtl (n = 10), BMΔGls1 (n = 12), BMΔGls2 (n = 8) and BMΔDKO (n = 12) at the end of the study period. White dotted lines in representative sections show aortic plaque limits. H&E, hematoxylin and eosin. Scale bar, 200 μm. Data from individual mice are shown and values are presented as the mean ± s.e.m. P values were determined by ordinary two-tailed Student’s t-test (a) or ANOVA with Fisher’s LSD post hoc analysis (b–j). Images in b from Servier Medical Art (https://smart.servier.com/).

To investigate the possibility of complementary roles of GLS1 and GLS2 in reparative macrophage metabolic reprogramming we then crossed germline Gls2 knockout mice (Gls2^−/−) with myeloid-specific Gls1 deficient mice (LysM-Cre × Gls1^fl/fl mice, referred to as Mye^ΔGls1), to generate double knockouts (DKO) (Fig. 1b, left). The efficient deletion of Gls1 and Gls2 was confirmed in bone marrow-derived macrophages (BMDMs) (Extended Data Fig. 1b). We first assessed the cellular Gln and glutamate (Glu) pools. Deficiency of Gls2 did not impact cellular Gln or Glu levels, irrespective of Gls1 status (Fig. 1b and Extended Data Fig. 1c). Consistently, flux measurements did not show any impact of Gls2 deficiency on oxidative phosphorylation (OxPHOS) as measured by the oxygen consumption rate (OCR) (Fig. 1c) or ATP production (Fig. 1d). We next asked whether GLS2 could complement GLS1 activity to regulate reparative and resolving macrophage functions^10,16. Again, we did not observe any difference between Gls1-deficient macrophages in the presence or absence of Gls2 with respect to reduced cell surface expression of the canonical alternatively activated marker CD206 or the efferocytosis marker TIM-4 (Fig. 1e and Extended Data Fig. 1d). Accordingly, GLS1, but not GLS2, was required for efficient efferocytosis of ACs in reparative macrophages (Extended Data Fig. 1e). We also evaluated the efferocytic capacity of TIM-4⁺ and TIM-4⁻ peritoneal macrophage (PCM) subsets of single or DKO mice and confirmed in vivo that, GLS2 did not synergize with GLS1 to affect the clearance of ACs (Fig. 1g and Extended Data Fig. 1f). The overexpression of TIM-4 using lentiviral particles was not sufficient on its own to restore the efferocytic capacity of Gls1-deficient macrophages (Extended Data Fig. 1g). Thus, GLS2 does not participate in GLS1-dependent macrophage clearance functions.

Lastly, to directly assess the in vivo relevance of myeloid GLS2 on atherosclerosis development, the bone marrow (BM) of Ctl, Mye^ΔGls1, Gls2^−/− or DKO mice was transplanted into irradiated atherosclerosis-prone Ldlr^−/− mice, and—after a recovery period—the recipients were placed on a Western-type diet. We did not observe any clear differences in classical risk factors for cardiovascular diseases, including body weight, plasma alanine aminotransferase (ALT), cholesterol and triglyceride levels, and blood leucocyte and myeloid cell counts between the genotypes (Extended Data Fig. 1h). However, while myeloid Gls1 deficiency exacerbated atherosclerosis development and necrotic core composition, myeloid Gls2 deficiency did not (Fig. 1h,i). Thus, the role of GLS2 in macrophages is unrelated to the relevant functions of these cells in the pathogenesis of atherosclerosis.

GS-dependent Gln synthesis balances Gln flux and macrophage restorative functions

Classically, the balance between metabolite absorption and biosynthesis maintains metabolic homoeostasis, and a role of the Gln synthetase (GS encoded by the Glul gene) has been proposed in tumour-associated macrophages to limit inflammation¹⁷. We found that—relative to genes encoding proteins involved in Gln and Glu metabolism in macrophages—Glul mRNA expression was downregulated by IL-4 stimulation in a validated gene expression dataset (Extended Data Fig. 2a)¹⁸. This was also confirmed in our previous datasets, partially dependent on the expression of Gls1 (Fig. 2a, top)¹⁰. Thus, the downregulation of Glul opposed the upregulation of Gls1 after IL-4 treatment (Fig. 2a, bottom). We next assessed the influence of GS on resting and reparative BMDM from control or Mye^ΔGls1 mice using the GS inhibitor methionine sulfoximine (MSO) (Fig. 2b). MSO was not toxic to the cells (Extended Data Fig. 2b), but—in contrast to Gls1 deficiency— the inhibition of GS had no impact on cellular Glu levels (Fig. 2c). Despite this, GS inhibition did reduce intracellular Gln levels in both resting and reparative BMDM^Ctl and BMDM^ΔGls1 (Fig. 2d). A PCA from RNA sequencing (RNAseq) data revealed that MSO-treated resting BMDMs clustered aside from their respective controls, but in contrast to Gls1 deficiency, this effect was attenuated upon IL-4 stimulation (Fig. 2e). Metabolic pathway-enrichment analysis confirmed a similar pattern (Extended Data Fig. 2c). This suggests that MSO treatment has a differential impact on macrophage metabolism in resting and reparative conditions, probably because Glul expression is already repressed under reparative conditions. Alongside these effects, MSO treatment of resting macrophages improved maximal respiration (Extended Data Fig. 2d), primed cells towards aerobic metabolism (Fig. 2f and Extended Data Fig. 2e) and enhanced ATP production (Fig. 2g). Consistently, we saw an increase in CD206 expression under resting conditions almost to the level induced by IL-4 stimulation (Fig. 2h). A similar increase was observed when AC uptake was quantified as a readout of efferocytosis (Fig. 2i and Extended Data Fig. 2f); however, most of these effects were lost upon Gls1 deficiency (Fig. 2f–i and Extended Data Fig. 2d–f), suggesting that GLS1-dependent glutaminolysis dictates Glu availability for GS metabolic reaction. Thus, GS inhibition through pharmacological treatment or IL-4 stimulation leads to a shared metabolic reprogramming that sustains ATP production to support macrophage reparative functions. Upon reparative conditions, in which Glul expression is repressed, MSO-treated macrophages reduced OCR and became slightly more glycolytic (Fig. 2f) but failed to impact ATP production, M2 marker expression or efferocytosis (Fig. 2g–i). Genetic inhibition of Glul using short-interferring RNA (siRNA) in BMDMs confirmed the modulation of efferocytosis (Fig. 2j). These findings indicate that GS opposes GLS1 as a cellular metabolic rheostat that dictates macrophage restorative functions depending on environmental cues.

Fig. 2 |.

Loss of GS-dependent metabolic rheostat for macrophage restorative functions in atherosclerosis. a, RNAseq analysis (top) of Glul expression in basal (M0) and reparative (M2) control (PCMCtl) and Gls1-deficient PCMs (PCMΔGls1) (GSE183176), and schematic representation of the impact of the regulation of Glul and Gls1 by IL-4 on the Gln–Glu cycle (bottom). b, The experimental outline. BMDMs from control and MyeΔGls1 mice were exposed for 16 h to the GS inhibitor MSO at a concentration of 1 mM. c,d, The quantification of Gln (c) and Glu (d) levels in these cells under basal or reparative conditions (n = 12 BMDMCtl saline, n = 11 BMDMCtl MSO, n = 10 BMDMΔGls1 saline and MSO of biologically independent replicates). e, PCA plot of RNAseq from basal (M0) and reparative (M2) BMDMCtl and BMDMΔGls1 in the presence or absence of MSO. f,g, Energy map generated from Seahorse metabolic flux measurements (f) and ATP production (g) for these cells. h,i, The cell surface expression of CD206 (h) and efferocytic index (i) in these BMDMs as measured by flow cytometry after 45 min exposure to ACs. Efferocytic index was calculated as follows: (number of macrophages with AC/total number of macrophages) × 100. j, The efferocytic index in basal and reparative BMDMs transfected with scramble or Glul siRNA in comparison with overnight MSO stimulation. Data from individual mice are shown and values are means ± s.e.m. (n = 3 biologically independent cultures per genotype (f–j)). k, Glul expression as the mean ± s.e.m. and the association with pathological features in stable (n = 16) and unstable (n = 27) human carotid artery plaques. The density of microvessel, endothelial cell, lymphocyte (T cell) and macrophage (MΦ) density (n mm–2); plaque size (mm2); collagen content, calcification, macrophage coverage and subsets (%). IPH, intraplaque haemorrhage. l, The cellular Gln and Glu levels, cell surface expression of CD206 and efferocytic index in PCMs isolated from atherogenic diet-fed Apoe−/− mice with i.p injection of saline or 20 mg kg–1 BW of MSO every other day for 12 weeks. Data from individual mice are shown and results are presented as means ± s.e.m. of cultures from four independent mice. Two-tailed Student’s t-test (k–l) or ANOVA with Fisher’s LSD post hoc analysis (c, d and g–j) was used. Images in a from Servier Medical Art (https://smart.servier.com/).

Although we recently showed that the inhibition of GS by MSO had a limited impact on the development of atherosclerosis¹⁹, its relevance to macrophage behaviour in this context remains unknown. We found variable Glul expression in unstable human carotid artery plaques, which positively correlated with the extent of microvessel and endothelial cell density, and, to some degree, macrophage density (Fig. 2k). It was also inversely correlated with the expression of the M2-specific marker Arg1 (Fig. 2k). These findings suggest that GS could be involved in macrophage expansion and functions in atherosclerosis as it was previously described as part of a core macrophage identity signature²⁰. Taking advantage of publicly available gene expression datasets of WD-fed atherogenic mice^14,15, we confirmed the upregulation of Glul expression over time in atheromatous plaques, which paralleled the expansion of macrophages (Extended Data Fig. 2g). Therefore, we assessed the functional characterization of PCMs isolated from control and MSO-treated, atherogenic diet-fed mice. As anticipated, MSO treatment reduced Gln levels in atherogenic PMCs, but this was not accompanied by significant changes in Glu levels (Fig. 2l). We also did not observe any significant effects on CD206 cell surface expression or efferocytosis in these cells (Fig. 2l). Thus, Gln synthesis is not a prerequisite for reparative macrophage function in the setting of atherosclerosis.

Fine tuning of Gln uptake controls macrophage restorative functions

We then asked whether GLS1-mediated macrophage glutaminolysis relies on Gln influx rather than Gln synthesis to support ATP production for macrophage reparative functions. Thus, we quantified the uptake of the radiolabelled [¹⁴C] Gln in BMDM^Ctl and BMDM^ΔGls1 treated with or without MSO. As expected, IL-4 stimulation enhanced [¹⁴C] Gln uptake^16,21,22 and this occurred independently of Gls1 expression (Extended Data Fig. 3a). Unexpectedly, the inhibition of GS with MSO also raised [¹⁴C] Gln uptake in reparative BMDM^Ctl and BMDM^ΔGls1 (Extended Data Fig. 3a). This led us to investigate the impact of removing Gln from the culture medium on macrophage metabolism, alternative activation and alternate resolving functions. As expected, Gln deprivation significantly reduced cellular Gln levels to the same extent across all conditions, even after GS inhibition (Extended Data Fig. 3b). We next observed that Gln deprivation reduced maximal respiration and ATP production, similar to Gls1 deficiency, even after GS inhibition (Extended Data Fig. 3c,d). In agreement with previous work^16,23, we confirmed that Gln depletion reduced the cell surface expression of the canonical alternatively activated macrophage markers CD206, similar to Gls1 deficiency, even after GS inhibition (Extended Data Fig. 3e). Similar findings were also observed for efferocytosis as Gln depletion reduced the efferocytic index after AC exposure in control and GS-inhibited reparative macrophages (Extended Data Fig. 3f). These data suggest that macrophage restorative functions rely on Gln influx, which is used as a rheostat when Gln synthesis is disrupted.

To dissect the mechanism by which Gln uptake is enhanced in this setting we first investigated the role of the canonical Gln exchangers SLC1A5 and SLC7A5²⁴. We excluded a role of SLC1A5 in cellular Gln homoeostasis using the inhibitor V9302, while the inhibition of SLC7A5 with JPH203 inhibitors raised the Gln content in reparative BMDM^Ctl and BMDM^ΔGls1 (Extended Data Fig. 4a). We did not observe any significant effect on CD206 cell surface expression or efferocytosis following the inhibition of SLC1A5 or SLC7A5 (Extended Data Fig. 4b,c). We next scrutinized our RNAseq data for the expression of genes encoding Gln exchangers²⁴ in BMDM. MSO treatment upregulated the expression of several plasma membrane solute carriers (Slc) capable of importing Gln in basal or reparative macrophages, including Slc7a6, Slc7a7 and Slc38a1 (Extended Data Fig. 4d). We therefore tested whether silencing these transporters using siRNA would perturb macrophage Gln uptake and reparative or resolving functions. At steady state, macrophages with silenced Slc7a6 or Slc7a7, but not Slc38a1, exhibited reduced Gln content in BMDM^Ctl and BMDM^ΔGls1 transfected with scramble (Extended Data Fig. 4e). The silencing of Slc7a7 also reduced cellular Gln content after Gln synthesis inhibition by MSO (Extended Data Fig. 4f), indicating that this transporter dominates Gln flux both at steady state and in response to cellular needs. Consistently, reduced Gln content was only evident in Slc7a7-silenced reparative macrophages (BMDM^ΔSlc7a7) (Fig. 3a).

Fig. 3 |.

Myeloid-Slc7a7 deletion impairs macrophage restorative functions in the pathological process of atherosclerosis. a, The quantification of Gln levels in reparative BMDMCtl and BMDMΔGls1 transfected with scramble (Scbl) or Sl7a6, Slc7a7 and Slc38a1 siRNA. b,c, The cell surface expression of CD206 (b) and the effferocytic index (c) measured by flow cytometry after 45 min exposure to ACs in these BMDMs. d, The efferocytic index in BMDMs transfected with scramble or Slc7a7 siRNA after one (45 min) or two (45 min + 1 h rest + 45 min) incubations with ACs. The results are presented as means ± s.e.m. (n = 3 biologically independent cultures per genotype (a–d)). e, A schematic representation of transgenic mice (left). MyeΔSlc7a7 represents mice lacking Slc7a7 in their myeloid cells compared with control (LyzMCre). Reduced CD206 expression and effferocytic index were confirmed in reparative BMDMs generated from these animals (n = 4 per group). The efferocytic index was calculated as follows: (number of macrophages with AC/total number of macrophages) × 100. f, Slc7a7 expression as the mean ± s.e.m. and impact in stable (n = 16) and unstable (n = 27) human carotid artery plaque. The density of microvessel, endothelial cell, lymphocyte (T cell), macrophage (MΦ) and SMC density (n mm–2); plaque size (mm2); collagen content, calcification, macrophage and fibroblast subsets (%). IPH, intraplaque haemorrhage. g, A dot plot showing the average scaled expression (colour intensity) and percent of expressing cells (dot size) for Slc7a7 and macrophage subset markers (Trem2, Cd163 and Ccl3) across four macrophage clusters (GSE260656 and GSE155513, top). UMAP representation of major macrophage subsets from an integrated single-cell RNAseq analysis of mouse atherosclerotic aortas (22,852 cells), with Slc7a7 expression projected onto the UMAP plot (bottom). h, The experimental outline. BM from control and MyeΔSlc7a7 mice were transplanted into lethally irradiated atherosclerotic Ldlr−/− recipient mice. After a 5-week recovery period, the mice were fed an atherogenic diet for 11 weeks. i, Representative sections (left) and quantification (right) of aortic plaques and necrotic cores from female Ldlr−/− recipient mice that received BM from control and MyeΔSlc7a7 mice (n = 12 mice per genotype). Dotted lines in representative sections show aortic plaque limits. H&E, hematoxylin and eosin. Scale bar, 200 μm. j, Enzymatically digested aortas from these animals were analysed by spectral flow cytometry (n = 8 mice per genotype). t-SNE plots depicting macrophage subsets within CD45+Lin− fraction (left) and quantification of the proportion of these populations for each group of animals (right). Data from individual mice are shown and values are presented as the mean ± s.e.m. Significance was determined by two-tailed Student’s t-test (e, f, i and j) or ANOVA with Fisher’s LSD post hoc analysis (a–d). Images in e, g, h and j from Servier Medical Art (https://smart.servier.com/).

In contrast to SLC38A1, SLC7A6 and SLC7A7 are obligatory exchangers, meaning that the influx of Gln is coupled to the efflux of other substrates such as arginine²⁴. Thus, we next investigated how SLC-dependent Gln uptake and GLS1-dependent glutaminolysis interact to modulate downstream metabolic pathways, such as hexosamine biosynthesis, which relies on phosphoglucomutase 3 enzymatic activity or the activation of mechanistic target of rapamycin (mTOR) and DBL signalling (Extended Data Fig. 4g)^25,26. We first observed that silencing of Slc7a7, but not Slc7a6 or Slc38a1, reduced Glu content in reparative BMDM^Ctl, recapitulating the defect observed in BMDM^ΔGls1 (Extended Data Fig. 4h). Using sambucus nigra agglutinin lectin (SNL) staining of 2,6-linked sialylglycoprotein modification as a readout of hexosamine biosynthesis, we observed that BMDM^ΔSlc7a7 did not exhibit any major effect on hexosamine flux, except that they limited Gln shunt induced by Gls1 deficiency (Extended Data Fig. 4i). Silencing of Slc7a7 was also associated with reduced levels of phosphor-mTOR and a trend towards increased Dbl expression in reparative macrophages (Extended Data Fig. 4j,k); this was independent of GLS1, suggesting that reduced cellular Gln levels after the silencing of Slc7a7 may have a broader impact on macrophage functions. Nevertheless, Slc7a7 but not Slc7a6 and Slc38a1-silenced macrophages showed reduced CD206 cell surface expression (Fig. 3b) and efferocytosis (Fig. 3c) in reparative control macrophages recapitulating the phenotype of BMDM^ΔGls1. These findings align with the reduced ATP levels in Slc7a7-silenced cells (Extended Data Fig. 4l). After successive rounds of AC clearance, Slc7a7-silenced macrophages still showed reduced efferocytosis in both BMDMs and PCMs (Fig. 3d and Extended Data Fig. 4m). This was further illustrated by reduced cell surface expression of CD206 and efferocytosis in macrophages from mice bearing a conditional allele for SLC7A7 crossed with lysozyme M-Cre transgenic mice (Mye^ΔSlc7a7 mice) (Fig. 3e). Altogether, these results identify SLC7A7 as the main Gln importer responsible for macrophage restorative functions.

Impaired SLC7A7-dependent Gln uptake by macrophages accelerates atherosclerosis

We and others previously observed enhanced Gln uptake in macrophage-rich atherosclerotic lesions^10,27,28. To evaluate the potential clinical significance of our findings we first asked whether Slc7a7 expression correlated with human atherosclerotic plaque complexity. We found higher Slc7a7 expression in unstable versus stable human carotid artery plaques (Fig. 3f, left). Moreover, we uncovered a positive correlation between Slc7a7 expression and plaque size (Fig. 3f, right). Similar findings were observed over time in murine atheromatous plaques^14,15 (Extended Data Fig. 5a). Notably, high Slc7a7 expression was detected in macrophage subsets, including the TREM2⁺ population, as shown in available single-cell RNAseq datasets from human and murine atherogenic aortas²⁹ (Fig. 3g and Extended Data Fig. 5b). However, as we also saw a positive correlation between Slc7a7 expression and macrophage density (Fig. 3f), it was possible that the former correlations might primarily reflect the extent of macrophage infiltration. Thus, we next tested the in vivo relevance of SLC7A7 for atherosclerosis development in Ldlr^−/−-recipient mice that received BM from control or Mye^ΔSlc7a7 mice fed with an atherogenic diet for 11 weeks (Fig. 3h). We did not observe any significant difference in plasma cholesterol levels between mice whose macrophages lacked or expressed SLC7A7 (Extended Data Fig. 5c). Nevertheless, we observed an approximately 1.5-fold increase in atherosclerosis plaque area and more complex necrotic core composition in the aortic sinus of Ldlr^−/− recipient mice transplanted with BM from Mye^ΔSlc7a7 mice compared with controls, independent of changes in fibrous cap thickness (Fig. 3i and Extended Data Fig. 5d). This was associated with a trend towards decreased Gln levels and significantly reduced Glu levels in aortas isolated at the end of the study period in a subgroup of animals (Extended Data Fig. 5e). Building on the identification of macrophage subset markers in murine single-cell RNAseq data (Fig. 3g), we developed a spectral flow cytometry strategy to investigate how Slc7a7 deficiency affects macrophage diversity (Extended Data Fig. 5f). To validate our experimental model, Ldlr^−/− recipient mice were transplanted with BM from adult myelopoiesis fate mapper (Ms4a3^tdTomato) mice³⁰ and fed an atherogenic diet for 11 weeks (Extended Data Fig. 5g). tdTomato labelling in CD45⁺ myeloid cells isolated from aortas at the end of the study period ranged from 50 to 100%, depending on the ontogenetically distinct macrophage subsets, with no labelling observed in lymphoid cells (Extended Data Fig. 5h). Thus, we applied spectral flow cytometry to the aortas of control or Mye^ΔSlc7a7 BM transplanted Ldlr^−/−-recipient mice (Fig. 3j). Mice with myeloid-specific Slc7a7 deficiency exhibited increased plaque cellularity, primarily driven by inflammatory macrophage subsets (CD11b⁺, MHCII⁺ and Ly6C⁺), along with the CD9⁺CD72⁺ subset, which represents the recently described foamy TREM2 population²⁹ (Fig. 3j). These findings demonstrate that SLC7A7 contributes to macrophage restorative functions and plasticity in the pathological process of human and murine atherosclerosis.

GLS1-mediated intrinsic glutaminolysis moderates Gln flux in macrophages to balance remodelling and restorative functions

Despite enhanced Gln uptake, lower GLS-dependent metabolism has been observed in murine macrophage-rich atherosclerotic lesions¹⁰. We now show that reduced Gls1 expression also correlated with lower aortic Glu levels in human plaques despite the enhanced Slc7a7 expression (Extended Data Fig. 6a). We then asked whether macrophage glutaminolysis might participate in hyperglutaminaemia-induced atherosclerosis. Indeed, we recently reported that modulating the plasma Gln-to-Glu ratio (GGR) through GLS2-dependent hepatic glutaminolysis enhances aortic Gln uptake and accelerates atherosclerosis¹⁹. However, the effect of this modulation on macrophage functions—particularly macrophage glutaminolysis¹⁰—is unknown. We first assessed the potential of serum from Ldlr⁻/⁻ and Ldlr⁻/⁻Gls2⁻/⁻ mice to reprogramme Gln metabolic pathways in Gls1- and Slc7a7-silenced macrophages (Fig. 4a). While Gls1 and Slc7a7 deficiency had minimal impact on hexosamine pathway activity under basal conditions, they exerted opposing effects on this flux in the presence of elevated Gln levels from Ldlr⁻/⁻Gls2⁻/⁻ serum, as evidenced by SNL staining (Fig. 4b). Despite this, we confirmed that both deficiencies reduced cellular Glu and ATP levels relative to controls, regardless of the increased serum Gln concentrations in Ldlr⁻/⁻Gls2⁻/⁻ mice (Fig. 4b). These results were further supported by a reduction in maximal respiration (Extended Data Fig. 6b), prompting us to investigate their relevance in vivo. We tested this in Ldlr^−/− and Ldlr^−/−Gls2^−/− recipient mice that received control or Mye^ΔGls1 BM then were fed with an atherogenic WD (Fig. 4c). We first confirmed increased plasma levels of Gln and GGR in Ldlr^−/−Gls2^−/− recipient mice, and that this effect was independent of macrophage GLS1-dependent glutaminolysis (Extended Data Fig. 6c). This was associated with enhanced levels of Gln and GGR in aortas isolated at the end of the study period in a subgroup of animals (Fig. 4d). Conversely, a reduction in Glu levels occurred in the aortas of Mye^ΔGls1 BM recipient mice, and this effect was independent of GLS2-dependent hepatic glutaminolysis (Fig. 4d). A similar trend in Gln and Glu levels was observed in PCMs isolated from these mice at the end of the study period (Fig. 4e). Using SNL staining, we confirmed that Gln flux could be diverted when macrophages were deficient in glutaminolysis, particularly when confronted to high Gln levels in Ldlr^−/−Gls2^−/− mice (Fig. 4e). However, succinate, aspartate, proline and ATP levels—downstream of Glu availability—were reduced in Gls1-deficient cells exposed to high Gln (Fig. 4e and Extended Data Fig. 6d). These data indicate that macrophages can adapt their metabolic flux in response to the environment.

Fig. 4 |.

Cell-intrinsic glutaminolysis governs Gln flux in atherosclerosis. a, A schematic representation of potential Gln flux in macrophages from Ldlr−/− and Ldlr−/−Gls2−/− animals. b, Quantification of SNL (a readout of hexosamine biosynthesis), Glu and ATP levels in reparative BMDMs transfected with scramble (Scbl) or Gls1 and Slc7a7 siRNA in the presence of 5% MS from atherogenic Ldlr−/− or Ldlr−/−Gls2−/− mice. The results are means ± s.e.m. (n = 3 biologically independent cultures per genotype). c, Experimental outline. BM cells from control and MyeΔGls1 mice were transplanted into lethally irradiated atherosclerosis-prone Ldlr−/− or Ldlr−/−Gls2−/− recipient mice. After a 5-week recovery period, the mice were fed an atherogenic diet for 11 weeks. d, Quantification of aortic Gln (left) and Glu levels (Glu, middle) and GGR (Gln/Glu, right) in a subgroup of these animals. Both sexes were analysed (female, empty circle; male, filled circle). Data from individual mice are shown and values are presented as the mean ± s.e.m. (n = 7 BMDMCtl and BMDMΔGls1 transplanted into Ldlr−/− mice and n = 6 BMDMCtl and BMDMΔGls1 transplanted into Ldlr−/−Gls2−/− mice). e, Quantification of Gln, Glu, SNL and ATP levels in PCMs isolated from Ldlr−/− and Ldlr−/−Gls2−/− recipient mice that received BM from control and MyeΔGls1 mice at the end of the study period (n = 5 per group). Data from individual mice are shown, and values are presented as the mean ± s.e.m. ANOVA with Fisher’s LSD post hoc analysis was used. Images in a and c from Servier Medical Art (https://smart.servier.com/).

To understand the transcriptional effects of the different genotypes upon PCMs, we carried out whole-transcriptome profiling at the end of the study. The Venn diagram of DEGs revealed that each genotype had its own signature, with little overlap (Extended Data Fig. 7a). As expected¹⁰, GSEA of DEGs between conditions showed an increase in genes associated with inflammation and a decrease in genes associated with ‘response to stress’ and ‘microtubule cytoskeleton’ in Gls1-deficient PCMs (Fig. 5a). Consistently, defective glutaminolysis dominantly decreased cell surface expression of CD206 (Extended Data Fig. 7b) and efferocytosis, particularly in reparative TIM-4⁺ PCMs (Fig. 5b). These in vivo findings aligned with our in vitro observations, where Gls1- and Slc7a7-silenced macrophages displayed similarly impaired efferocytosis, independent of the elevated serum Gln levels from Ldlr⁻/⁻Gls2⁻/⁻ mice (Extended Data Fig. 7c). By contrast, PCMs isolated from Ldlr^−/−Gls2^−/− mice showed an increase in genes involved in ‘response to stress’ along with ‘membrane plasticity’, opposing a decrease in genes involved in ‘organelle organization’ (Extended Data Fig. 7d). Gls1-deficient PCMs exposed to high levels of Gln from Ldlr^−/−Gls2^−/− mice showed a more drastic phenotype with an upregulation of genes involved in ‘DNA damage response’ and ‘cellular morphogenesis’ along with a downregulation of genes involved in ‘organelle organization’ and ‘immune response’. These results suggest that divergent cellular fluxes of Gln control the transcriptional landscape of macrophages; nevertheless, filtering the matrisome transcriptomic signature³¹ revealed that the enhanced hexosamine pathway in GLS2-deficient mice was associated with shared matrisome reorganization (Fig. 5c,d). The impact of this metabolic reprogramming on efferocytosis and matrix remodelling (Fig. 5e) was first demonstrated by the accelerated atherosclerosis and increased necrotic core density in Ldlr^−/− and Ldlr^−/−Gls2^−/− recipient mice transplanted with Mye^ΔGls1 BM relative to their respective counterparts receiving control BM (Fig. 5f and Extended Data Fig. 7e). Despite reduced plasma cholesterol levels in Ldlr^−/−Gls2^−/− recipient mice (Extended Data Fig. 7f), a similar development of atherosclerosis occurred at 80% of the cholesterol level in these mice with a more prominent necrotic core and enhanced fibrous cap thickness (Fig. 5f and Extended Data Fig. 7e). Notably, Ldlr^−/−Gls2^−/− recipient mice exhibited exacerbated fibrous cap thickness compared with Ldlr^−/− recipients, regardless of Mye^ΔGls1 BM transplantation (Fig. 5f). Altogether, these findings indicate that impaired macrophage glutaminolysis in a pathological setting redirects Gln flux, thereby impeding macrophage restorative functions, disrupting matrix remodelling and driving atherogenesis (Fig. 6).

Fig. 5 |.

Cell-intrinsic glutaminolysis dominates macrophage restorative functions in atherosclerosis. a, To identify the regulation of specific transcriptional pathways between PCMΔGls1 and PCMCtl from Ldlr−/− recipient mice transplanted with BM cells from MyeΔGls1 and control mice, differentially expressed genes quantified by RNAseq were subjected to GSEA analysis. The top four upregulated and downregulated pathways are displayed. b, The gating strategy (left) and efferocytic index (right) in TIM-4− and TIM-4+ PCMs isolated from atherosclerosis-prone Ldlr−/− or Ldlr−/−Gls2−/− recipient mice transplanted with BM cells from control and MyeΔGls1 mice at the end of the study period after 45 min of ex vivo AC exposure. Data from individual mice are shown and values are presented as the mean ± s.e.m. (n = 5 mice transplanted with BMDMCtl, n = 6 BMDMΔGls1 into Ldlr−/− mice and n = 7 BMDMΔGls1 into Ldlr−/−Gls2−/− mice). c, A Venn diagram highlighting the common and specifically regulated matrisome transcripts quantified by RNAseq across the different PCMs, as indicated in the figures. d, A radar chart representation of matrisome module enrichment across these different PCMs. Eight radii of the radar chart are devoted to matrisome and KEGG specific modules indicated in the figures. e, A schematic representation of the potential impact of Gln flux on macrophagemediated matrix remodelling and efferocytosis in Ldlr−/− and Ldlr−/−Gls2−/− animals. f, Representative sections (left) and quantification (right) of necrotic cores and fibrous cap thickness (mircometre) from Ldlr−/− and Ldlr−/−Gls2−/− recipient mice that received BM from control and MyeΔGls1mice (n = 13 mice transplanted with BMDMCtl and n = 14 mice transplanted with BMDMΔGls1). The dotted lines in representative sections show aortic plaque limits. Scale bar, 200 μm. Data from individual mice are shown and values are presented as the mean ± s.e.m. Significance was determined ANOVA with Fisher’s LSD post hoc analysis. Images in e from Servier Medical Art (https://smart.servier.com/).

Fig. 6 |.

Gln metabolism remodels macrophage functions and atherosclerotic lesions. Top left: healthy macrophages utilize intrinsic Gln fluxes in vitro as a metabolic rheostat to regulate restorative functions, such as matrisome reorganization and efferocytosis. This involves enhanced SLC7A7-dependent Gln influx, GLS1-dependent glutaminolysis (but not GLS2) and reduced GSdependent Gln synthesis. Top right: elevated systemic Gln levels in mice lacking hepatic GLS2 raise the pathogenic threshold of plasma GGR, promoting distal macrophage Gln influx. However, in the context of atherosclerosis—where GS is upregulated and GLS1 downregulated—the hexosamine (HBP) pathway impacts macrophage-driven extracellular matrix (ECM) remodelling. Bottom left: blocking SLC7A7-dependent Gln influx to limit GLS1-dependent glutaminolyis raises the pathogenic threshold of aortic GGR, impairs ATP production and promotes a dysferocytic phenotype, ultimately contributing to necrotic core formation. Bottom right: combining elevated systemic Gln levels in mice lacking hepatic GLS2 with impaired macrophage intrinsic glutaminolysis exacerbates the hexosamine pathway, while limiting ATP production, which ultimately impact the metabolic fingerprint of atherosclerosis. Parts of the figure were created using images from Servier Medical Art (https://smart.servier.com/).

Discussion

A role of Gln for reparative-type macrophages induced by IL-4-dependent alternative activation was proposed almost a decade ago¹⁶. However, it is only recently that the physiopathological relevance to atherosclerosis has been revealed in a mouse model with specific deficiency of GLS1-mediated glutaminolysis in macrophages¹⁰. Despite impaired glutaminolysis in the inflamed aortas of atherosclerotic mice, we and others reported enhanced Gln uptake, which correlated with inflamed atherosclerotic lesions and plaque macrophage content better than Glc utilization^10,27,28. Thus, there is a need to better understand the Gln routes that probably depend on intrinsic macrophage properties and the tissue environment to enable atherosclerotic plaque development^32,33. The metabolic fingerprint of mouse atherosclerosis and its relevance to the human condition, as shown in this study, illuminates the finely tuned regulation of Gln flux in macrophages. At least, GS-dependent Gln synthesis serves as a metabolic rheostat driving Gln uptake, which can influence macrophage reparative and resolving functions depending on the availability of exogenous Gln and the efficiency of glutaminolysis. The latter is mediated by GLS1 but not GLS2 in macrophages and Gln uptake is mediated in part by SLC7A7. Consistently, mice with macrophage-specific deletion of Slc7a7 exhibited an altered macrophage landscape in the aortas of Ldlr^−/− atherosclerotic mice, and accelerated atherosclerosis. These results emphasize that plaque macrophages have elevated Gln requirements, probably because they cannot efficiently support anaplerosis in the context of defective glutaminolysis.

The modulation of Gln metabolism can lead to strikingly different phenotypes in macrophages. Classically activated macrophages rely on aerobic glycolysis and the conversion of Gln through the GABA shunt to support pro-inflammatory responses^34,35. By contrast, reparative macrophages rely on non-canonical transamination of Gln to sustain OxPHOS and efferocytosis¹⁰. Thus, there is growing interest in identifying mechanistic differences between the metabolic reprogramming of macrophage polarization states. In this regard, Schilperoort et al.³⁶ showed that, in contrast to the common feature of classically activated macrophages boosting inflammation by prolonged aerobic glycolysis with Glc-to-lactate conversion (Warburg effect), macrophages ingesting ACs exhibit a transient glycolysis producing a burst of lactate for pro-resolving processes. Labelling experiments previously suggested that Gln rather than Glc is the major nutrient fuelling the TCA cycle after IL-4 treatment favouring OxPHOS¹⁶, an effect that could even be exacerbated when cells are cultured without Glc³⁷. Here, we report that GS is a metabolic rheostat that controls Gln influx in a negative feedback loop. Indeed, while it has been shown that Gln deprivation increases Glul expression¹⁷, we now found that the inhibition of GS led to an increase in Gln uptake promoting macrophage metabolic rewiring. A similar scenario is well known with cholesterol synthesis³⁸. It is therefore tempting to assume that this mechanism is part of a ‘fight-or-flight’ process for cell survival in response to their changing environment that could therapeutically be harnessed in atherosclerotic lesions³⁹. Nevertheless, this metabolic flexibility is severely impaired in the context of atherosclerosis, confirming the dependence on glutaminolysis rather than Gln synthesis in this setting to sustain the Krebs cycle for mitochondrial anapleurosis. These findings highlight the limitation of simply imaging Gln flux in atherosclerotic plaques and the need to better assess the in vivo activity of mitochondrial electron transport chain that supports the balance between OxPHOS and ROS production^40,41,42.

Plaque macrophages can be metabolically reprogrammed by systemic and local factors depending on the stage of the disease, their origin and zonation or the presence of different metabolic stressors^1,2,3,4. Enhanced Gln flux in atherosclerotic plaque macrophages is illustrated by the uptake of ¹⁸F-FGln by specific SLC7A7-positive atherosclerotic plaque macrophages^27,28. Testing the relevance of macrophages SLC7A7 in murine atherosclerosis, we found an unexpected acceleration of atherosclerosis with more complex necrotic core composition. This phenotype resembles the increased development of atherosclerosis in mice with macrophage-specific deletion of the Glc transporter Slc2a1^8,9, despite the positive correlation between Slc2a1expression and plaque size. In both cases, the absence of these transporters led to defective efferocytosis. Consistently, it has recently been shown that Slc7a7 expression increases within hours after efferocytosis in zebrafish tissue macrophages most likely to handle high metabolic demand⁴³, and perturbed efferocytosis is observed in macrophages from patients with lysinuric protein intolerance with defective SLC7A7⁴⁴. Of note, we now report that SLC7A7 dominates the inflammatory macrophage landscape of atherosclerotic plaques, probably owing to impaired resolution of inflammation, including the TREM2⁺CD9⁺ macrophages—also known as scar-associated macrophages—in which TREM2 signalling contributes to plaque stability^45,46,47. In the present study, we observed that SLC7A7 regulated several Gln-dependent downstream pathways and its role in promoting macrophage resolution and repair functions was mainly attributed to feeding GLS1-dependent glutaminolysis. Consistently, high Slc7a7 expression negatively correlated with low plaque Glu levels alongside the decrease in Gls1 expression in unstable human plaques and murine atherosclerotic plaques¹⁰. Furthermore, we observed that plaque remodelling was the result of combined systemic and local Gln flux perturbations. Indeed, beyond the role of glutaminolysis in regulating efferocytosis, elevated plasma Gln levels in a mouse model with hepatic Gls2 deficiency promoted macrophage metabolic rewiring towards the hexosamine pathway—especially in the absence of macrophage glutaminolysis—leading to impaired remodelling functions and exacerbated fibrous cap thickening.

Overall, our data support the requirement of Gln metabolism as a potential metabolic liability of macrophages in atherosclerotic plaques that might be exploited for cardiovascular therapy.

Methods

Human atherosclerosis

The tissues used were part of the Maastricht Pathology Tissue Collection; the further storage and use of the tissue was in line with the Dutch Code for Proper Secondary use of Human Tissue, with approval from the local Medical Ethical Committee (protocol number 16–4-181). Carotid arteries were collected from 22 symptomatic male patients undergoing carotid endarterectomy as part of the Maastricht Human Plaque Study (72.9 ± 6.3 years old)^10,33. Formalin-fixed, paraffin embedded segments of 5 mm in size were alternated with frozen segments for RNA isolation. Two independent pathologists then classified the segments in a blind fashion as fibrous cap atheroma with or without intraplaque haemorrhage (16 stable segments and 27 unstable segments, respectively) according to hematoxylin and eosin (HE) staining. Stable and unstable snap-frozen segments were used for further microarray analysis. Snap-frozen segments were pulverized and 5–20 mg of material was subjected to transcriptomics. RNA isolation was performed by guanidium thiocyanate extraction and RNA was further purified with the Nucleospin RNA II kit (Macherey–Nage). RNA quality and integrity were determined using the Agilent 2100 Bioanalyzer. Biotinylated complementary RNA was prepared with the Illumina TotalPrep RNA Amplification kit (Ambion), and 750 ng of complementary RNA per sample was used for hybridization (Illumina Human Sentrix-8 v.2.0, Beadchip). Scanning was performed on the Illumina Beadstation 500 (Illumina) and image analysis was conducted using the Illumina Beadstudio v.3 gene expression software. A total of 22,184 human transcripts were analysed in the R Bioconductor lumi package. First, a variance stabilizing transformation was performed. Second, the Robust Spline Normalization algorithm was applied to normalize the data. Differential gene expression analysis was performed using the function lmFit provided in Limma R package on preprocessed transcriptomics data. The 88 adjacent tissue sections underwent assessment for plaque size, necrosis, inflammation (CD68, CD3, arginase and iNOS), smooth muscle cells (SMCs) (α-smooth muscle actin) collagen (Sirius red) and angiogenesis (CD31⁺ microvessel density, newly formed CD105⁺ microvessels, α-smooth muscle actin⁺ mature microvessels and Lyve⁺ lymphatic density). Gln and Glu levels were quantified by metabolomics analysis³³. Pearson correlation analysis was performed to assess the association between gene expression and plaque phenotypical traits.

Mice

Gls1^fl/fl mice (Glstm2.1Sray/J, The Jackson Laboratory) were crossed with Lyz2^Cre mice (B6.129P2-Lyz2tm1(cre)Ifo/J, The Jackson Laboratory). Gls2^−/− mice were generated by Taconic Biosciences (Cryopreserved-Tyr^c-Brd) and have been crossed to atherogenic Ldlr^−/− (B6.129S7-Ldlr^tm1Her/J) animals¹⁹. Slc7a7^fl/fl mice were crossed with Lyz2^Cre mice (B6.129P2-Lyz2tm1(cre)Ifo/J, The Jackson Laboratory) and BM from these mice were kindly provided by S. Bodoy. BM from Ms4a3^tdTomato mice were kindly provided by F. Ginhoux. For each experiment, co-housed C57BL/6 littermate male and female controls were used between 8 and 14 weeks of age. Animal protocols were approved by the Institutional Animal Care and Use Committee of the French Ministry of Higher Education and Research and the Mediterranean Centre of Molecular Medicine (Inserm U1065) and were undertaken in accordance with the European Guidelines for Care and Use of Experimental Animals. Animals had free access to food (chow diet A04, Safe) and water and were housed in a controlled environment with a 12-h light–dark cycle, at constant temperature (21.7–22.8 °C) and relative humidity (50–60%). Water and cages were autoclaved. Cages were changed once weekly, and the health status of the mice was monitored using a dirty bedding sentinel programme. Hyperlipidaemia was induced by feeding the mice with a WD (TD88137, Ssniff) for 11 weeks as indicated in the figure legends.

Treatment

GS inhibition was induced by intraperitoneal injection of 20 mg kg⁻¹ MSO (Sigma, M5379) dissolved in sterile phosphate-buffered saline (PBS), administered every 2 days during the 12 weeks.

BM transplantation

The 12–16-week–old female recipient mice (Ldlr^−/− or Ldlr^−/−Gls2^−/−) were irradiated 16 h before BM transplantation¹⁰. They then underwent intravenous injection with 4 × 106 BM cells from donor mice (Mye^ΔGls1, Mye^ΔSlc7a7 or Ms4a3^tdTomato mice). Mice were then allowed to recover for 5 weeks before feeding them an atherogenic diet (WD, TD88137, Ssniff) for 11 weeks.

Blood leucocytes

Leucocytes were quantified from 50 μl of whole blood collected into EDTA tubes using a haematology cell counter (Hemavet, Beckman Coulter).

Plasma biochemical parameters

Plasma multi-analyte profiling was performed using a clinical chemistry analyser (Mindray BS-240 Pro, BioSentec) with the following colorimetric kits: ALT (ALT-0102 from Mindray). Plasma cholesterol and triglyceride content were measured with LabAssayTM Cholesterol (Sobioda) and Triglycerides Reagent (Diasys) according to the manufacturer’s protocol.

Gln/Glu measurements

Plasma, tissue or cells were collected and a commercially available Gln/Glu-GloTM Assay kit (Promega- J8022) was used to measure Gln and Glu levels in accordance with the manufacturer’s instructions.

Analysis of atherosclerotic plaque

Mice were killed by dislocation and slowly perfused with 10 ml of ice-cold PBS. The hearts and aortas were carefully excised. In some experiments, aorta samples were used fresh for spectral flow cytometry analysis (see below). The hearts were fixed in 4% paraformaldehyde containing 30% sucrose and embedded in paraffin to analyse tissue architecture by HE staining¹⁰. Then, 7-μm-thick paraffin sections of the aortic sinus were prepared using a HM340E microtome (Microm Microtech). Histological slides were examined using a Nikon upright microscope coupled with a DS-Ri 1 colour camera (Nikon Eclipse Ci). Images were acquired with a Ds-L3 software (Nikon), and plaque areas were quantified with Fiji software. Fibrous cap thickness covering the lipid core was measured and expressed in micrometres. Necrotic core size (quantified by measuring the area of HE-negative, acellular white areas in the intima) was expressed as a percentage of the total root area¹⁹.

Tissue leucocyte analysis

Aortas were flushed, cut into small pieces and digested for 30 min with HBSS medium containing 1.5 mg ml⁻¹ of collagenase D (Sigma,11088882001) and 0.5 U ml⁻¹ of Liberase (Sigma, 5401054001) at 37 °C. Single-cell suspensions were subjected to red blood cell lysis, filtration and centrifugation for 5 min at 400g, before incubation for 30 min on ice protected from light, with a cocktail of fluorochrome-conjugated antibodies for extracellular antigens. The following antibodies are used at 1:200 dilution for flow cytometric analysis: CD45 BUV805 (BDBiosciences Clone I3/2.3), CD115 BV711 (BioLegend Clone AFS98), CD8a BUV735 (BDBiosciences Clone 53–6.7), CD19 BUV563 (BDBiosciences Clone 1D3), TCR β chain Pacific Blue (BioLegend Clone H57–597), Ly6G BUV395 (BDBiosciences Clone 1A8), CD64 (FcγRI) BV786 (BDBiosciences Clone X54–5/7.1), CD31 (PECAM-1) PercP/Cy5.5 (BioLegend Clone 390), CD206 AFR647 (BioLegend Clone C068C2), IA/IE BUV615 (BDBiosciences Clone M5/114.15.2), CD209 BV421 (BDBiosciences Clone 5H10), CD9 BV480 (BDBiosciences Clone KMC8), TIM-4 BV605 (BDBiosciences Clone RMT4–54)CD72 BV650 (BDBiosciences Clone K10.6), CD74 BV750 (BDBiosciences Clone In-1), LYVE1 AFR488 (Invitrogen Clone ALY7), CD11c PE/Cy7 (BioLegend CloneN418), Ly6C AFR700 (BioLegend CloneHK1.4) and FR-b APC (clone10/FR2, cat. no. 153304, BioLegend). Cells were then washed and centrifuged, and data were acquired on Cytek Aurora flow cytometer using SpectroFlow with a five-laser configuration. Forward scatter/side scatter gating was used to exclude dead cells and debris followed by FSC-A/FSC-H gating to select singlets. All analyses, including unsupervised t-distributed stochastic neighbour embedding analysis, were performed using FlowJo software (Tree Star).

Cell culture, PCMs and BMDMs

BM cells were collected from mouse femur and tibia and differentiated in the presence of recombinant mouse M-CSF (20 ng ml⁻¹; Miltenyi) in complete Roswell Park Memorial Institute (RPMI) 1640 medium (Corning) containing 10 mM of Glc, 2 mM of L-Gln, 100 U ml⁻¹ of penicillin/streptomycin and 10% foetal bovine serum (FBS) for 7 days at 37 °C and 5% CO₂. On day 7, BMDMs were incubated overnight with IL-4 (20 ng ml⁻¹, Peprotech). PCMs were obtained by peritoneal lavage with 5 ml of PBS. The resultant single-cell suspension was subjected to red blood cell lysis, filtration and centrifugation for 5 min at 400g.

Cell treatments

Macrophages were cultured in complete RPMI 1640 medium (Corning) containing 10 mM of Glc, 2 mM of l-Gln, 100 U ml⁻¹ of penicillin–streptomycin and 10% FBS at 37 °C and 5% CO₂. Cells were incubated overnight with the following treatments: IL-4 (20 ng ml⁻¹, Peprotech), l-methionone sulfoximine (1 mM MSO, Sigma), SLC1A5 inhibitor (V-9302, 10 μM, Sigma) and SLC7A5 inhibitor (JPH203, 5 μM, Tocris). Cells were starved of Gln overnight using the same culture medium without Gln. In some experiments, cells were cultured overnight without Gln and FBS but supplemented with 5% mouse serum (MS) from atherogenic Ldlr^−/− and Ldlr^−/−Gls2^−/− mice (Supplementary Table 1).

siRNA transfection

Cells were transfected with Glul, Slc7a6, Slc7a7 and Slc38a1 siRNA (ON-TARGETplus-SMARTpool, Dharmacon) or control siRNA (D-001810–01-05, Dharmacon) (referred to as Scbl) at 30 nM using Lipofectamine RNAiMAX (Life Technologies), according to the manufacturer’s instructions.

Lentivirus overexpression

Cells were spin-transfected for 90 min at 1,300g with Timd4 lentivirus (LVM(VB250129–1156mar)-K1, Vectorbuilder) or control lentivirus (LVM(VB010000–9492agg)-b, Vectorbuilder) (referred to as ‘empty’) and used at a multiplicity of infection of 10. After 6 h, cells were washed and the medium was replaced with fresh medium for 48 h before treatment

Seahorse extracellular flux analysis

For extracellular flux assay, 1 × 10⁵ BMDMs were plated in a Seahorse Bioscience culture plate. Cells were then incubated overnight with different drugs and metabolites. The OCR was measured by an XF96 Seahorse Extracellular Flux Analyser following the manufacturer’s instructions. In the Seahorse assay, cells were treated with oligomycin (1 μM), FCCP (1.5 μM), rotenone (1 μM) and antimycin A (0.1 μM). ATP production was measured after the injection of oligomycin and calculated using the Seahorse XF report generator. Each condition was performed with three replicates.

ATP production assay

Cells were collected and a commercially available ATP determination kit (Invitrogen-A22066) was used to measure ATP levels in accordance with the manufacturer’s instructions.

Metabolite measurements

Cellular succinate, aspartate and proline content were measured using colorimetric assay kits (Sigma and Abbexa, respectively) according to the manufacturer’s protocol.

2-Deoxy-[¹⁴C] Glc and [¹⁴C] Gln uptake

Cells were incubated with 0.1 μCi of 2-deoxy-[¹⁴C] Glc or [¹⁴C] Gln in 2% serum albumin Krebs–Ringer bicarbonate buffer, pH 7.4 for 1 h at 37 °C, washed three times with PBS and homogenized with 5% HClO₄ solution. The radioactivity incorporated was measured and expressed as total radioactivity per 10⁶ cells.

In vitro flow cytometry analysis

Cell suspensions were incubated with the appropriate antibodies for 30 min on ice, protected from light. The following antibodies and dyes were used at 1:200 dilution for macrophage flow cytometric analysis: CD45 APC-Cy7 (BDBiosciences Clone 30-F11), CD206 PercP Cy5.5 (BioLegend Clone C068C2), TIM-4 Pe/Cy7 (BioLegend Clone RMT4–54), CD11b BV510 (BioLegend Clone M1/70), CD301 PeCy7 (BioLegend Clone LOM-14) and Sambucus Nigra Lectin-FITC (ThermoFisher). Antibody validations were performed by suppliers and antibodies were used according to the manufacturers’ instructions. In some experiments, apoptosis was measured with the APC Annexin V apoptosis detection kit with 7-AAD according to the manufacturer’s instructions (BioLegend). Cells were then washed and centrifuged, and data were acquired on BD FACSCanto flow cytometer. Forward scatter/side scatter gating was used to exclude dead cells and debris followed by FSC-A/FSC-H to select singlets. Gating strategies are described in the manuscript. Data collection was performed using the FACS DIVA software and analysis was performed using FlowJo software (Tree Star).

Intracellular flow cytometry

Before labelling, cells were removed from medium then incubated with antibodies recognizing surface molecules. Next, cells were fixed and permeabilized using a transcription factor staining buffer set (eBioscience Foxp3–130-093–142) according to the manufacturer’s instructions. Lastly, intracellular antigens were detected by incubation with phospho-mTor (Ser2448) PE conjugated (clone O21–404, BDphosphoflow) and MCF2 APC conjugated (clone AA804–833, AntibodiesOnline) for 30 min on ice, protected from light.

In vitro efferocytosis analysis

BMDMs and PCMs were generated and stimulated as described above. To generate ACs, spleens from C57BL/6J mice were collected and mechanistically dissociated, filtered on 100-μm-pore-diameter nylons (Falcon), pelleted and resuspended in RPMI medium supplemented with 10% FBS. Apoptosis was induced by ultraviolet exposure at 312 nm for 10 min and cells were maintained in culture for an additional 2 h. ACs were labelled with CellTrace Violet Cell Proliferation kit (Thermo Fisher) according to the manufacturer’s instructions. Fluorescent ACs were washed twice with PBS before use. For efferocytosis: stained ACs were added at a 5:1 ratio onto plated macrophages for 45 min. For two rounds of efferocytosis: ACs were added at a 5:1 ratio on plated macrophages for 45 min. Cells were then washed three times and macrophages were incubated for 1 h. Stained ACs were then added at a 5:1 ratio on macrophages for 45 min. Cells were washed three times and macrophages were stained for AC content and activation markers and analysed by flow cytometry.

Real-time qPCR and RNAseq

Tissues were collected and stored at −80 °C in RNAprotect tissue reagent (Qiagen). BMDMs were obtained as described above. Total RNA was extracted with RNeasy Mini Kit (Qiagen) according to the manufacturer’s protocol, and quality was assessed by Nanodrop (Ozyme). For real-time quantitative polymerase chain reaction (qPCR), cDNA was prepared using 10 ng μl⁻¹ total RNA by polymerase chain reaction with reverse transcription using a high-capacity cDNA reverse transcription kit according to the manufacturer’s instructions (Applied Biosystems). Real-time qPCR was performed on cDNA using SYBR Green. qPCRs were performed on the StepOne device (Applied Biosystems). The results were normalized on m36B4 gene expression to account for variability in initial messenger RNA levels. Gls1 (F: GCACATTATTCACCCGGTAACC; R: CTGCCCACCCACCATCC) Gls2 (F: ACAAGATGGCTGGGAACGAAT; R: TGACACTGCCTGACTCACAGG) were used. For RNAseq, library constructions were conducted and sequenced at BGI Hong Kong Company Limited using a DNBseq^™ platform.

Network analyses

The identification of up- and downregulated genes was based on P values and false discovery rates calculated with Phantasus¹⁹.

Statistics

Data are expressed as mean ± s.e.m. Outliers were tested with the Grubb’s test, and statistical analysis was performed using a two-tailed t-test or one-way analysis of variance (ANOVA) with Fisher’s least significant difference (LSD) post hoc analysis with GraphPad Prism software. P values <0.05 were considered statistically significant.

Extended Data

Extended Data Fig. 1 |.

Myeloid Gls1 deficiency, but not Gls2, impairs macrophage restorative functions. (a) RNAseq analysis of Gls2 expression in relation to canonical macrophage markers in the aortas of atherogenic Western diet (WD)-fed Apoe−/− mice (GSE10000). When indicated, *P < 0.05 vs controls. Violin plot representing the mean expression of Gls2 in aortic macrophages from control and WD-fed Ldlr−/− mice (GSE97310). (b) mRNA expression of Gls1 and Gls2 in control (BMDMCtl), Gls1-deificient BMDMs (BMDMΔGls1), Gls2-deificient BMDMs (BMDMΔGls2) and BMDMΔDKO (values were normalized to m36B4 and expressed as arbitrary unit)(n = 4 mice per group). (c) Gln levels in these BMDMs at steady state (basal) or after overnight IL-4 stimulation. (d) Cell surface expression of TIM-4 measured with flow cytometry in basal and reparative (IL-4 stimulated) BMDMCtl, BMDMΔGls1, BMDMΔGls2 and BMDMΔDKO. The results are means ± s.e.m. (n = 12 BMDMCtl, n = 10 BMDMΔGls1 and BMDMΔGls2 and n = 5 BMDMΔDKO (c), n = 3 cultures per genotype (d) of biologically independent replicates). (e) Efferocytic index gating strategy for TIM-4+ cells after 45 min incubation with ACs in reparative BMDMCtl, BMDMΔGls1, BMDMΔGls2 and BMDMΔDKO. (f) Efferocytic index gating strategy after 45 min of ex vivo AC exposure in TIM-4− and TIM-4+ resident peritoneal cavity macrophages (PCMs) of Ldlr−/− recipient mice that received BM from control, MyeΔGls1, Gls2−/− and DKO mice. (g) Cell surface expression of TIM-4 measured with flow cytometry and efferocytic index in PCMCtl and PCMΔGls1 stimulated overnight with IL-4 after TIM-4 lentivirus overexpression (left) and efferocytic index in basal and reparative BMDMCtl and BMDMΔGls1 after TIM-4 lentivirus overexpression (right)(n = 3 mice per group). (h) Body weight and plasma alanine aminotransferase (ALT) levels (left), plasma cholesterol and triglyceride levels (middle) and blood leukocyte count and blood myeloid cell percentage (right) in these mice at the end of the study period (n = 10 BMDMCtl, n = 11 BMDMΔGls1, n = 8 BMDMΔGls2, and n = 12 BMDMΔDKO transplanted into Ldlr−/− mice). Data from individual mice are shown and values are presented as the mean ± s.e.m. Significance was determined by two-tailed Student’s t-test or one-way analysis of variance (ANOVA) with Fisher’s least significant difference (LSD) post hoc analysis. Source data are provided as a Source Data file (b,c,d,g,h).

Extended Data Fig. 2 |.

Glul is inhibited by alternative polarization, blunting its steady-state effect on macrophage priming. (a) RNAseq analysis of Gln metabolism KEGG pathways in basal and reparative control and BMDMΔGls1 (GSE53053). When indicated, *P < 0.05 vs M0. (b) Gating strategy and quantification of apoptosis measured by flow cytometry in basal and reparative control and BMDMΔGls1 in the presence or absence of MSO. The results are means ± s.e.m. (n = 3 biologically independent cultures per genotype). (c) Metabolic pathway-enrichment analysis of RNA-seq profiling in these BMDMs. Pathways highlighted in red or blue indicate significant up- or down- regulation, respectively. (d) OCR measured by Seahorse in basal (left) or reparative (right) control and BMDMΔGls1 in presence or absence of MSO. The results are means ± s.e.m. (n = 3 biologically independent cultures per genotype. Pvalues displayed for M0MSO vs. M0Ctl and M2ΔGls1 and M2ΔGls1-MSO vs. M2Ctl). (e) Uptake of 2-deoxy-[14C] glucose in basal and reparative BMDMCtl and BMDMΔGls1 in the presence or absence of MSO. The results are means ± s.e.m. (n = 3 biologically independent cultures per genotype). P values were determined by two-tailed Student’s t-test or one-way analysis of variance (ANOVA) with Fisher’s least significant difference (LSD) post hoc analysis. (f) Efferocytic index gating strategy for these BMDMs after 45 min incubation with ACs. (g) RNAseq analysis of Glul expression in relation to canonical macrophage markers in the aortas of atherogenic-diet-fed Apoe−/− mice (GSE10000). When indicated, *P < 0.05 vs controls. Violin plot representing the mean expression of Glul in aortic macrophages from control and atherogenic-diet-fed Ldlr−/− mice (GSE97310). Source data are provided as a Source Data file (b,d,e).

Extended Data Fig. 3 |.

Gln is the main energy source for macrophage restorative functions. (a) Uptake of [14C] Gln in basal and reparative BMDMCtl and BMDMΔGls1 in the presence or absence of MSO. (b) Quantification of Gln levels in reparative BMDMCtl and BMDMΔGls1 in the presence or absence of MSO after overnight Gln deprivation. (c) ATP production measurements in these BMDMs. (d) OCR measured by Seahorse in reparative BMDMCtl and BMDMΔGls1 in the presence or absence of MSO at steady state (left) or after Gln deprivation (right). Pvalues displayed for basal M2MSO, M2ΔGls1 and M2ΔGls1-MSO vs. M2Ctl and for Gln-deprived M2MSO vs. M2Ctl. (e) Cell surface expression of CD206 measured by flow cytometry in these BMDMs. (f) Gating strategy (left) and quantification (right) of effferocytic index measured by flow cytometry after 45 min exposure to apoptotic cells (ACs) in these BMDMs. Efferocytic index was calculated as follows: (number of macrophages with AC/total number of macrophages) × 100. Data from individual mice are shown and values are presented as the mean ± s.e.m. (n = 3 biologically independent cultures per genotype (a-f)). Two-tailed Student’s t-test or one-way analysis of variance (ANOVA) with Fisher’s least significant difference (LSD) post hoc analysis was used. Source data are provided as a Source Data file (a,b,c,d,e,f).

Extended Data Fig. 4 |.

Slc7a7 is the main Gln transporter for macrophage reparative and resolving functions. (a) Gln levels in reparative BMDMCtl and BMDMΔGls1 exposed for 16 hours to the SLC1A5 inhibitor (V9302) or SLC7A5 inhibitor (JPH203). (b) Cell surface expression of CD206 and (c) effferocytic index measured by flow cytometry after 45 minutes’ exposure to apoptotic cells (ACs) in these BMDMs. (d) Heatmap illustrating RNA-seq analysis of SLC Gln exchangers in basal (M0) and reparative (M2) BMDMCtl and BMDMΔGls1 in the presence or absence of MSO. When indicated, *P < 0.05 Gls1-deficient BMDMs vs controls§;P < 0.05 MSO treatment vs controls#;P < 0.05 MSO-treated Gls1-deficient BMDMs vs controls. Gln levels in resting BMDMCtl and BMDMΔGls1 transfected with scramble (Scbl) or Sl7a6, Slc7a7 and Slc38a1 siRNA in the absence (e) or presence (f) of MSO. (g) Schematic representation of potential signaling pathways downstream of SLC7A7. Quantification of Glu levels (h), sambucus nigra agglutinin lectin (SNL, a readout of hexosamine biosynthesis) levels (i), intracellular phosphor-mTorc1 ( j), and intracellular Dbl (k) measured by flow cytometry in these BMDMs. (l) Efferocytic index in PCMs transfected with scramble or Slc7a7 siRNA after one (45 min) or two (45 min + 1 h rest + 45 min) incubations with ACs. The results are means ± s.e.m. (n = 3 biologically independent cultures per genotype (a-c, e-m)). Two-tailed Student’s t-test or one-way analysis of variance (ANOVA) with Fisher’s least significant difference (LSD) post hoc analysis was used. Images in g from Servier Medical Art (https://smart.servier.com/). Source data are provided as a Source Data file (a,b,c,e,f,h,I,j,k,l,m).

Extended Data Fig. 5 |.

SLC7A7-dependent plasticity of atherosclerotic macrophages assessed by spectral flow cytometry. (a) RNAseq analysis of Gln exchanger Slc expression in in the aortas of atherogenic-diet-fed Apoe−/− mice (GSE10000) #.P < 0.05 vs baseline. Violin plot representing the mean expression of Slc7a7 in aortic macrophages from control and atherogenic-diet-fed Ldlr−/− mice (GSE97310). (b) UMAP representation of major myeloid cell subsets from an integrated scRNA-seq analysis of human coronary artery samples containing atherosclerotic lesions, with Slc7a7 expression projected onto the UMAP plot. (c) Plasma cholesterol levels and (d) quantification of fibrous cap thickness (μm) in Ldlr−/− recipient mice transplanted with BM cells from control and MyeΔSlc7a7 mice (n = 12 mice per genotype). (e) Aortic Gln, and Glu levels in a subgroup of these mice (n = 4 mice per genotype). Data from individual mice are shown and values are presented as the mean ± s.e.m. (c,e). Two-tailed Student’s t-tests were used. (f) t-SNE plots depicting the macrophage population gating strategy determined by spectral cytometry in enzymatically digested atherogenic aortas. (g) Experimental outline. BM from Ms4a3tdTomato mice were transplanted into lethally irradiated atherosclerotic Ldlr−/− recipient mice. After a five-week recovery period, the mice were fed an atherogenic diet for 11 weeks. (h) Gating strategy (left) and quantification (right) of tdTomato labelling in macrophage subsets and lymphoid cells isolated from the aortas of these mice at the end of the study period. The results are means ± s.e.m. (n = 7 mice). Images in f and g from Servier Medical Art (https://smart.servier.com/). Source data are provided as a Source Data file (c,d,e,h).

Extended Data Fig. 6 |.

Cell-intrinsic glutaminolysis dominates macrophage metabolic rewiring in atherosclerosis. (a) Pearson correlation between Gln-related gene expression and Gln or Glu levels in stable and unstable human carotid artery plaques (n = 32). (b) OCR measured by Seahorse in reparative BMDMs transfected with scramble (Scbl) or Gls1 and Slc7a7 siRNA in presence of mouse serum (MS) from atherogenic Ldlr−/− mice (left) or Ldlr−/−Gls2−/− mice (right). (n = 3 biologically independent cultures per genotype. Pvalues displayed for BMDMΔGls1 and BMDMΔSlc7a7 vs. BMDMCtl). (c) Quantification of plasma Gln (left), Glu (middle) and Gln/Glu (right) levels in Ldlr−/− and Ldlr−/−Gls2−/− recipient mice transplanted with BM cells from control and MyeΔGls1 mice. Both sexes were analysed (female, empty circle; male, filled circle) (n = 13 mice per genotype). (d) Schematic representation of potential Glu flux in macrophages from Ldlr−/− and Ldlr−/−Gls2−/− animals (left) and quantification of succinate, aspartate and proline levels in PCMs isolated from these mice at the end of the study period (n = 3 per group). Data from individual mice are shown and values are presented as the mean ± s.e.m. (b,c,d). Significance was determined by two-tailed Student’s t-test or one-way analysis of variance (ANOVA) with Fisher’s least significant difference (LSD) post hoc analysis. Images in d from Servier Medical Art (https://smart.servier.com/). Source data are provided as a Source Data file (a,b,c,d).

Extended Data Fig. 7 |.

Gln homeostasis influences the transcriptional landscape of macrophages and atherosclerosis development. (a) Venn diagram highlighting the common and specifically regulated genes quantified by RNAseq across the different PCMs isolated from Ldlr−/− and Ldlr−/−Gls2−/− recipient mice transplanted with BM cells from control and MyeΔGls1 mice, as indicated in the figure. (b) Representative histogram and quantification of cell surface expression of CD206 in these cells. Data from individual mice are shown and values are presented as the mean ± s.e.m. (n = 10 mice transplanted with BMDMCtl, n = 9 BMDMΔGls1 into Ldlr−/− mice, and n = 10 BMDMΔGls1 into Ldlr−/−Gls2−/− mice). (c) Efferocytic index measured by flow cytometry after 45 min exposure to ACs in reparative BMDMs transfected with scramble (Scbl) or Gls1 and Slc7a7 siRNA in presence of 5% mouse serum from atherogenic Ldlr−/− or Ldlr−/−Gls2−/− mice. The results are means ± s.e.m. (n = 3 biologically independent cultures per genotype) (d) To identify the regulation of specific pathways between PCMs exposed to high Gln levels from Ldlr−/−Gls2−/− recipient mice that received BM from control and MyeΔGls1mice, differentially expressed genes were subjected to GSEA analysis. The top four upregulated and downregulated pathways are displayed. (e) Quantification of aortic plaques (mm2) and (f) plasma cholesterol levels in Ldlr−/− and Ldlr−/−Gls2−/− recipient mice that received BM from control and MyeΔGls1mice (n = 13 mice transplanted with BMDMCtl and n = 14 mice transplanted with BMDMΔGls1). Data from individual mice are shown and values are presented as the mean ± s.e.m. One-way analysis of variance (ANOVA) with Fisher’s least significant difference (LSD) post hoc analysis was used. Source data are provided as a Source Data file (b,c,e,f).

Data availability

Other information is available from the corresponding author upon reasonable request. RNAseq data of BMDMs and PCMs have been deposited at the Gene Expression Omnibus under accession number GSE302738. Source data are provided with this paper.