CXCL12 derived from endothelial cells promotes atherosclerosis to drive coronary artery disease

Genome-wide association studies (GWAS) have established a link between the genomic locus 10q11, which hosts the CXCL12 gene, and the risk for coronary artery disease (CAD) (1). CAD risk alleles downstream of CXCL12 have been associated with plasma levels of the chemokine CXCL12 (2); however, the nature and directionality of this association remain elusive. Recently, a Mendelian randomization study identifying genetic determinants of biomarkers in the populations of ORIGIN and CARDIoGRAM revealed CXCL12 as a causal mediator of CAD, supported by epidemiological analysis showing a 15% higher risk for cardiovascular events per SD of increased CXCL12 plasma levels in ORIGIN (3). To detail the association between CXCL12 and CAD, we conducted a meta-analysis of GWAS performed in the EPIC-Norfolk and PROMIS cohorts (n=12,657; filters: INFO≥0.5, MAF≥0.01, HWE≥1×10⁻⁶, only SNPs appearing in both cohorts). The study was approved by an institutional review committee and subjects gave informed consent. Applying conditional analysis, we newly identified rs2802492, an intergenic SNP near CXCL12, to be independently associated with CXCL12 plasma levels (β=0.016, P=3.24×10⁻⁸) as determined by ELISA (α-isoform, R&D Systems Quantikine kit) (2) and with increased risk for CAD (OR 1.047, P=1.27×10⁻⁶), corroborating CXCL12 as a driver of CAD. No linkage disequilibrium (r²>0.8) was found between rs2802492 and the CAD-associated SNPs rs1746048 (1) and intergenic rs1482478 (3). Notably, when specifically tested for association with CXCL12 expression using the eQTL calculator in the GTEx dataset, analysis for rs2802492 but not rs1482478 yielded a nominal P<0.05 in the tibial artery, with a consistent direction of effect. However, these data await validation in biological models, namely murine atherosclerosis, to elucidate mechanisms by which CXCL12 mediates CAD. This is particularly relevant, because we found CXCR4, the receptor of CXCL12, to confer cell-specific atheroprotective effects preserving endothelial function and contractile smooth muscle cell (SMC) phenotype in mice, while reduced CXCR4 expression was associated with CAD risk in humans (4), indicating complex effects of this ligand-receptor axis in atherosclerosis and CAD.

To unravel mechanisms underlying the association of plasma CXCL12 and CAD, we employed set of models interrogating the impact of cell-specific CXCL12-deficiency on atherosclerosis in apolipoprotein E-deficient (Apoe^−/−) mice subjected to Western diet (WD, 21% fat, 0.15–0.2% cholesterol) for 12 weeks (Figure), as described (4). Notably, atherosclerotic lesion area, composition and inflammation, as assessed by collagen and macrophage content, in the aorta of mice with CXCL12-deficiency in all somatic cells (Ubi-12-KO) did not differ from controls (Ubi-12-WT), although CXCL12 plasma levels were almost undetectable, suggesting opposing roles of CXCL12 or targeted receptors from different cell types. Hence, we generated chimeric mice with CXCL12-deficiency in bone marrow (BM)-derived hematopoietic (Hem-12-KO) vs non-hematopoietic resident cells (Res-12-KO) by reciprocal BM transplantation. Lesion area or quality did not differ between Hem-12-KO mice and controls (Hem-12-WT), and CXCL12 plasma levels remained unaffected, indicating a minor role of hematopoietic cells for systemic CXCL12 levels. In contrast, lesion area in the thoracoabdominal aorta and the aortic arch was reduced and CXCL12 plasma levels were markedly diminished in Res-12-KO mice compared to Res-12-WT controls, underlining the importance of resident cells as a source of CXCL12 (Figure). To dissect relevant cell populations, we employed mice with SMC-specific (SMC-12-KO) or arterial endothelial cell-specific CXCL12-deficiency (EC-12-KO). Aortic lesion area, composition and CXCL12 plasma levels did not differ between SMC-12-KO and SMC-12-WT mice, excluding a role of SMC-derived CXCL12. Conversely, EC-12-KO mice displayed a marked reduction of lesion area in the thoracoabdominal aorta and the aortic arch, as well as decreased CXCL12 plasma levels as compared to EC-12-WT controls. In EC-12-WT mice, we found a positive correlation of CXCL12 plasma levels and lesion area (Spearman’s Rho=0.349, P=0.04) that was abolished in EC-12-KO (Spearman’s Rho=0.017, P=0.95). In Res-12-KO and EC-12-KO mice, lesional collagen content was increased, whereas macrophage content was not affected. These effects were independent of plasma cholesterol levels, which did not show any relevant alterations in any of the models (Figure). Collectively, these data indicate that effects of CXCL12 on atherosclerosis rely on CXCL12 production in arterial ECs, identifying EC-derived CXCL12 as a crucial driver of atherosclerosis and contributor to CXCL12 levels. As this contribution only amounted to 25% of total plasma levels, other cellular sources, e.g. adipocytes or liver (3), likely give rise to a substantial component of unknown functional relevance. Of note, somatic deficiency in CXCR4 did not reveal equivalent protective effects as deletion in vascular cell types (4), implying pro-atherogenic functions of CXCR4 in a different compartment, as determined by analysis of CXCR4 deficiency in non-vascular, namely BM-derived cells (5). Such pro-atherogenic effects may correspond to the effects of CXCL12 reported herein and may help to tailor therapeutic targeting.

Figure. Effect of cell-specific CXCL12-deficiency on aortic atherosclerosis and CXCL12 plasma levels.

All mice were Apoe^−/− (C57/Bl6) treated with tamoxifen (50 mg/kg body weight for 5 consecutive days) to induce Cre-expression and fed a WD for 12 weeks (A-F). Thoracoabdominal aortas were opened longitudinally, mounted on glass slides and en face-stained with Oil-Red-O to quantify the atherosclerotic lesion area (A). Aortic arches with main branch points (brachiocephalic artery, left subclavian artery and left common carotid artery) was fixed with 4% paraformaldehyde, embedded in paraffin, and lesion area was quantified after HE-staining of 3 transversal sections (B). Atherosclerotic root lesions were stained with Masson’s Trichrome to visualize and quantify collagen within the plaque (C) or reacted with an anti-Mac2 antibody to quantify the percentage of Mac2⁺ macrophages among lesional cells after incubation with a secondary FITC-conjugated antibody (D). An enzymatic assay (Roche Diagnostics) was used to measure cholesterol plasma levels (E) and an ELISA (R&D Systems Quantikine Kit) was used to quantify CXCL12 plasma levels (F). Data represent mean±SEM (n=12–35). *P<0.05; **P<0.01; ****P<0.0001, as analyzed by Mann-Whitney U-test. Procedures were in accordance with institutional guidelines and approved by local authorities. Gloassary: Ubi-12-WT, Cre^ERT2 or Cxcl12^flox/flox mice; Ubi-12-KO, Cre^ERT2Cxcl12^flox/flox mice; Hem-12-WT, Hem-12-KO, mice reconstituted with Cxcl12^flox/flox or Cre^ERT2Cxcl12^flox/flox BM, respectively; Res-12-WT, Res-12-KO, Cxcl12^flox/flox or Cre^ERT2Cxcl12^flox/flox mice reconstituted with Apoe^−/− BM, respectively; SMC-12-WT, SmmhcCre^ERT2 or Cxcl12^flox/flox mice; SMC-12-KO, SmmhcCre^ERT2Cxcl12^flox/flox mice; EC-12-WT, BmxCre^ERT2 or Cxcl12^flox/flox mice; EC-12-KO, BmxCre^ERT2Cxcl12^flox/flox mice.