Cdkn2a is an atherosclerosis modifier locus that regulates monocyte/macrophage proliferation

Abstract

Objective

Common genetic variants in a 58-kilobase region of chr 9p21, near the CDKN2A/CDKN2B tumor suppressor locus, are strongly associated with coronary artery disease. However, the underlying mechanism of action remains unknown.

Methods and Results

We previously reported a congenic mouse model harboring an atherosclerosis susceptibility locus and the region of homology with the human 9p21 locus. Microarray and transcript-specific expression analyses showed markedly decreased Cdkn2a expression, including both p16^INK4a and p19^ARF, but not Cdkn2b (p15^INK4b), in macrophages derived from congenic mice compared to controls. Atherosclerosis studies in subcongenic strains revealed genetic complexity and narrowed one locus to a small interval including Cdkn2a/b. Bone marrow (BM) transplantation studies implicated myeloid lineage cells as the culprit cell type rather than resident vascular cells. To directly test the role of BM-derived Cdkn2a transcripts in atherogenesis and inflammatory cell proliferation, we performed a transplantation study utilizing Cdkn2a^+/− cells in the Ldlr^−/− mouse model. Cdkn2a-deficient BM recipients exhibited accelerated atherosclerosis, increased Ly6C^hi pro-inflammatory monocytes and increased monocyte/macrophage proliferation compared to controls.

Conclusions

These data provide a plausible mechanism for accelerated atherogenesis in susceptible congenic mice, involving decreased expression of Cdkn2a and increased proliferation of monocyte/macrophages, with possible relevance to the 9p21 human locus.

Atherosclerotic vascular disease (ASVD) is the leading cause of death in Western societies. Complications of the disease are usually caused by rupture or erosion of an unstable atherosclerotic plaque, resulting in thrombus formation and arterial occlusion.¹ The complex etiology involves both genetic and environmental factors. While genetic factors underlying traditional risk factors are well known,² new data emerging from genome-wide association studies (GWASs) are revealing novel loci mediating disease susceptibility independent of traditional factors.³ In particular, a risk locus on chromosome (chr) 9p21.3 is strongly associated with ASVD (including myocardial infarction, stroke and aortic aneurysm) but independent of plasma cholesterol levels and hypertension.^4–6 While the association with ASVD is robust and has been widely-replicated in different ethnic groups, the underlying pathogenic mechanism remains unknown.

The association at 9p21 appears to be the most robust common (minor allele frequency > 10%) genetic determinant for ASVD in the human genome. The SNPs most strongly associated with disease risk map to a 58-kb region containing a long, non-coding RNA (ANRIL) and lying ~100 kb centromeric to the CDKN2A/B locus encoding inhibitors of cellular proliferation. Multiple ANRIL splice variants exist, complicating genetic association studies.⁷ Decreased expression of the tumor suppressors p16^INK4a, p19^ARF, and p15^INK4b has been observed among carriers of the ASVD risk allele in some studies^{8, 9} but not others.^10–12 Targeted deletion of the homologous region in mice resulted in markedly decreased expression of Cdkn2a/b transcripts and increased proliferation of vascular smooth muscle cells (vSMCs) in vitro. However, no effect on atherosclerosis was observed in the Western diet-fed wild-type background, a highly athero-resistant model.¹³ Thus, the potential role of these transcripts and the pathogenic mechanism leading to increased atherosclerotic risk at the 9p21 locus remains unclear.

We have used a murine genetic approach to map atherosclerosis susceptibility loci, including a locus on mouse chr 4 that was confirmed in a congenic strain.^{14, 15} The effect on disease susceptibility was independent of plasma cholesterol levels, body weight and plasma glucose levels.¹⁵ We now report refined genetic mapping indicating that the introgressed interval contains at least two loci, one overlapping the Athsq1 QTL and one covering the region of homology with the human 9p21 ASVD locus. Gene expression studies revealed decreased mRNA levels of p16^INK4a and p19^ARF, but not p15^INKb4, in macrophages derived from susceptible congenic mice. Mechanistically, we investigated the hypothesis that increased proliferation of macrophages (or mixed populations of monocyte/macrophages), due to reduced expression of Cdkn2a cell proliferation inhibitor transcripts in myeloid lineage cells, might be responsible, at least in part, for the accelerated atherosclerosis phenotype.

Methods

An expanded Methods section is available in the Online Supplemental Methods. B6.MOLF-Athsq1 congenic mice were bred as described.¹⁵ Cdkn2a^−/− mice¹⁶ backcrossed into B6¹⁷ were kindly provided by Dr. Sean Morrison (University of Michigan). Gene expression profiling or real-time PCR was carried out using peritoneal or splenic monocyte/macrophages. Raw profiling data are available in the NCBI GEO database (GSE 24342). Atherosclerotic lesions were quantified using morphometric analysis as described.^{14, 15} BM transplantation experiments were performed as described.¹⁸ Monocyte subsets were analyzed by standard flow cytometry analysis.¹⁹ Two-factor ANOVA and t tests were performed using STATVIEW 5.0 (Abacus Concepts, Inc). The threshold for significance was p = 0.05. Data shown are mean ± SEM.

Results

Genetic complexity of an atherosclerosis susceptibility locus containing Athsq1 as well as a 5.4-9 Mb interval containing Cdkn2a/b

A series of subcongenic strains was generated to narrow a previously-reported¹⁵ 54-Mb interval on chr 4 harboring an atherosclerosis susceptibility gene(s). Mice were bred to homozygosity, fed 9-wk WTD and analyzed for atherosclerotic lesion development. Mice carrying a 17- or 11-Mb proximal subinterval, but not the 4-Mb proximal tip, exhibited approximately 2-fold greater mean lesion area than non-congenic controls (p<0.0001) (Figure 1A–B). The effect was observed in both males and females. The negative lesion phenotype of the 4-Mb subcongenic was confirmed in a second cohort (Supplemental Figure I). These data narrow the proximal interval to a 5.4-9 Mb region containing 8–21 known protein coding genes including Cdkn2a/b (Figure 1C). Of note, this locus did not segregate with versican accumulation (data not shown), a phenotype originally observed in the full congenic strain.¹⁵ An additional distal locus was identified in a subcongenic strain carrying a non-overlapping donor interval (Supplemental Figure IB–C). The distal locus overlaps the linkage peak from the original Athsq1 mapping cross, but detailed mapping and phenotypic assessment have not been carried out. Together, the data indicate genetic complexity of the full 54-Mb interval, with refined mapping of the proximal locus leading to identification of a critical region containing 8–21 genes.

Figure 1. Refined mapping of an atherosclerosis susceptibility locus to a 5.4-9-Mb region including the region of homology with a human ASVD risk interval on 9p21.

A, Physical map of mouse chr 4 and MOLF donor intervals (white boxes) carried by congenic strains in the B6-Ldlr^−/− background. Mb, megabase; b/b, homozygosity for B6 alleles; m/m, homozygosity for MOLF alleles. The rectangle indicates the narrowed interval defined by data in (b). B, Mean lesion areas in control (b/b) and congenic strains fed 9-wk WTD. Two-factor ANOVA performed with square root transformation. Horizontal bars indicate group means for males (circles) and females (triangles). NS, not significant. C, Microsatellite markers delineating proximal (D4Mit349/D4Mit27) and distal (D4Mit350/D4Mit154) recombination breakpoints of the refined risk interval. The interval includes the region of homology with a human risk interval on 9p21.

Having previously demonstrated a role for BM-derived cells in determining the pro-atherosclerotic phenotype of Athsq1 congenics,¹⁵ we now performed a reverse BM transplantation to determine whether vSMCs or endothelial cells contributed to the phenotype as well. B6-Ldlr^−/− BM was injected into lethally-irradiated 17-Mb subcongenics and non-congenic controls. Following repopulation of the BM, mice were fed 11-wk WTD and analyzed for atherosclerotic lesion development. No difference in mean lesion area was observed between the groups (Supplemental Figure II). Similar results were obtained for males and females. Importantly, these data indicate that resident cells of the vessel wall are not involved in the accelerated atherosclerosis phenotype associated with the proximal 17-Mb locus. Thus, we focused further mechanistic studies of this strain in BM-derived monocyte/macrophages.

Microarray and transcript-specific gene expression analyses reveal decreased macrophage expression of p16^INK4a and p19^ARF, but not p15^INK4b, in Athsq1 congenic mice

As an independent approach to narrow the list of candidate genes (>600 genes in the full 54-Mb congenic interval and ~160 genes in the 17-Mb interval), we performed gene expression analysis using microarrays. Elicited peritoneal macrophages derived from full (54-Mb) congenics and non-congenic controls were collected after 6-wk WTD feeding. Differentially-expressed genes were defined as exhibiting ≥20% difference in expression level with a significance threshold corrected for multiple testing.²⁰ Three to six of the differentially expressed transcripts reside within the narrowed 5.4-9-Mb proximal interval and could be considered as causal candidate genes (Table 1). Strikingly, the most significant difference was a ~6-fold decrease in Cdkn2a levels in congenic compared to control macrophages. Cdkn2a encodes two transcripts involved in cell proliferation regulation: p16^INK4a and p19^ARF. ²¹ The transcripts utilize different promoters and alternative reading frames resulting in proteins with no amino acid homology.²² Because of the Cdkn2a exonic stucture, the microarray could not discern p16^INK4a vs p19^ARF transcripts. Transcript-specific rtPCR confirmed the results in the 17-Mb subcongenic strain and showed reduced expression of both p16^INK4a and p19^ARF, but not the adjacent p15^INK4b transcript, in macrophages from the atherosclerosis-prone 54- and 17-Mb congenics compared to non-congenic controls (Figure 2A). Similar results were observed for both transcripts in resident peritoneal macrophages, and for p16^INK4a in splenic monocyte/macrophages, from 54- and 17-Mb congenic mice compared to controls (Figure 2B–C). Of note, gene expression in peripheral blood monocytes was low (with Ct values >38) and we could not reliably detect a difference between congenics and controls. Thus, an atherosclerosis-prone murine strain exhibits decreased expression of the p16^INK4a and p19^ARF tumor suppressor genes in macrophages and mixed monocyte/macrophage populations.

Table 1. Cdkn2a is the most differentially expressed macrophage-derived gene residing in the 5.4-9-Mb mapped susceptibility interval.

Fold-change listed for all transcripts encoded within the 5.4–9-Mb interval and exhibiting a ≥20% difference in expression level in peritoneal macrophages derived from Athsq1 54-Mb full congenic or non-congenic (b/b) mice fed 6-wk WTD. N=10 mice/group. Note that Ptplad2 is listed twice (+1.24, −1.30 fold changes), with known splice variants exhibiting different expression patterns.

Probe ID	Map Location (Mb)	Gene symbol	Gene name	Fold change (relative to b/b)	P-value
100093087_TGI_at	88.3	Klhl9	kelch-like 9	1.29	0.0001
100103804_TGI_at	88.0	Ptplad2	protein tyrosine phosphatase-like A domain containing 2	1.24	0.0065
100102382_TGI_at	88.8	Mtap	methylthioadenosine phosphorylase	−1.26	7.77E-05
100116677_TGI_at	88.0	Ptplad2	protein tyrosine phosphatase-like A domain containing 2	−1.30	0.0071
100097525_TGI_at	94.7	Jun	Jun oncogene	−1.39	0.0028
100113480_TGI_at	93.0	Tusc1	tumor suppressor candidate 1	−1.88	5.36E-07
100099086_TGI_at	88.9	Cdkn2a	cyclin-dependent kinase inhibitor 2A	−5.98	1.25E-06

Figure 2. Decreased expression of p16^INK4a and p19^ARF, but not p15^INK4b, cell proliferation inhibitor transcripts in BM-derived cells from chr 4 congenic mice compared to controls.

rtPCR results in cells derived from B6-Ldlr^−/− (b/b) and congenic (54- or 17-Mb) mice. A, concanavalin A-elicited peritoneal macrophages, 6-wk WTD, n=10 mice/group; B, resident peritoneal macrophages, 9-wk WTD, n=8–12 mice/group; C, splenic CD11b⁺ monocyte/macrophages, 15-wk WTD. n=7–9 mice/group. ANOVA (A) or t-test (B, C) performed with log transformation. *p≤0.05, **p≤0.005, †p≤0.0005 compared to b/b controls.

BM-specific Cdkn2a-deficiency results in accelerated atherosclerosis, increased Ly6C^hi monocytes, and increased BrdU incorporation into monocytes and macrophages

To directly test the hypothesis that decreased Cdkn2a expression in BM-derived cells is pro-atherogenic, we performed a BM transplantation study using a previously described Cdkn2a-deficient mouse. The targeted mutation replaces exons 2/3 with a neo cassette, knocking out both p16^INK4a and p19^ARF expression¹⁶, and has been crossed into a uniform B6 background.¹⁷ In an effort to simulate the modest reduction of p16^INK4a and p19^ARF expression observed in congenics compared to controls, we used donor B6-Ldlr^+/−mice heterozygous for the Cdkn2a null allele. B6-Ldlr^−/− recipients were injected with B6-Ldlr^+/− or B6-Ldlr^+/−, Cdkn2a^+/− BM and fed 10-wk WTD following repopulation of the BM. No differences were observed in body weight, plasma total cholesterol, HDL or triglycerides (Supplemental Table I). Resident peritoneal macrophages exhibited a 2.5-fold decrease in expression of p16^INK4a but not p19^ARF or p18^INK4c (another INK4-class gene encoded by a distant region of mouse chr 4) in B6-Ldlr^+/−, Cdkn2a^+/− recipients compared to controls (Figure 3A). Similarly, splenic CD11b⁺ monocyte/macrophages exhibited ~4-fold decreases in both p16^INK4a and p19^ARF but not p18^INK4c (Figure 3B). Importantly, mean atherosclerotic lesion area was increased in B6-Ldlr^+/−, Cdkn2a^+/− recipients compared to controls (p=0.04, two-factor ANOVA) (Figure 3C), with non-significant single-sex increases of 34% and 17% for males and females, respectively. Thus, heterozygous BM-specific deficiency of Cdkn2a is sufficient to confer a modestly accelerated atherosclerosis phenotype in mice.

Figure 3. BM-specific Cdkn2a deficiency is sufficient to promote atherosclerosis in B6-Ldlr^−/− mice.

A–B, Transcript-specific rtPCR results for Cdkn2a (p16^INK4a and p19^ARF) and p18^INK4c (another INK4-class gene encoded by a distant region of mouse chr 4). N=7 mice/group. Unpaired t-test performed with log transformation. C, Mean lesion areas from B6-Ldlr^−/− mice transplanted with B6-Ldr^+/− or B6-Ldr^+/−, Cdkn2a^+/− BM and fed 10-wk WTD. Two-factor ANOVA performed with square root transformation. Horizontal bars indicate group means for males (circles) and females (triangles). *p≤0.05.

To test potential atherogenic mechanisms consistent with decreased p16^INK4a and/or p19^ARF expression in monocyte/macrophages, we assessed apoptosis and measures of monocytosis in the BM transplanted mice. TUNEL staining of atherosclerotic lesions revealed few positive cells and no difference between B6-Ldlr^+/−, Cdkn2a^+/− recipients compared to controls (data not shown). No differences were observed in circulating white blood cell counts or total monocytes in blood derived from B6-Ldlr^+/−, Cdkn2a^+/− recipients compared to controls (data not shown). However, flow cytometry analysis of blood monocytes revealed increased Ly6C^hi monocytes after chow (0-wk timepoint) and WTD feeding (Figure 4A–B). Increased percentages of Ly6C^hi and decreased percentages of Ly6C^lo subsets (Figure 4B) resulted in a significant increase in the ratio of Ly6C^hi:Ly6C^lo monocytes in the circulation of B6-Ldlr^+/−, Cdkn2a^+/− recipients compared to controls (Supplemental Figure 3A). Note that the 0- and 2-wk timepoints are pre-lesional indicating that increased Ly6C^hi monocytes are not secondary to accelerated atherosclerosis.¹⁵ Similar results were observed for the 17-Mb congenic strain compared to non-congenic controls (Supplemental Figure IIIB). Moreover, an increased ratio of Ly6C^hi:Ly6C^lo monocytes was observed in splenic monocytes derived from B6-Ldlr^−/−, Cdkn2a^+/− mice compared to B6-Ldlr^−/− controls (p=0.003) (Supplemental Figure IVA–B).

Figure 4. Increased inflammatory Ly6C^hi monocytes, mediated by increased cell proliferation, in the circulation of B6-Ldlr^−/− mice transplanted with Ldlr^+/−, Cdkn2a^+/− BM compared to controls.

A, Flow cytometry analysis of blood monocytes from Ldlr^+/− or Ldlr^+/−, Cdkn2a^+/− BM recipients fed chow or WTD. Monocytes were gated as CD45⁺CD115⁺ Ly6C^hi or CD45⁺CD115⁺Ly6C^lo. B, Quantification of Ly6C^hi/ Ly6C^lo cells among total CD45⁺CD115⁺ cells. C, Analysis of proliferating CD45⁺CD115+ monocytes at the 10-wk timepoint. Cells were gated as Ly6C^hi BrdU⁺. D, Quantification of BrdU⁺ cells among total CD45⁺CD115⁺ monocytes. N=11 mice/group. *p≤0.05, **p≤0.005, †p≤0.0005.

To test for a direct effect of heterozygous Cdkn2a deficiency on monocyte proliferation, we injected mice with BrdU before sacrifice at the 10-wk timepoint. B6-Ldlr^+/−, Cdkn2a^+/− recipients exhibited an increased percentage of BrdU⁺ monocytes compared to controls, indicating increased monocyte proliferation (Figure 4C–D). Moreover, the increase in BrdU incorporation occurs mainly in the Ly6C^hi subset (Figure 4C, left panel).

To test for an effect of BM-derived Cdkn2a deficiency on tissue macrophage proliferation, we assayed elicited peritoneal macrophages by flow cytometry following intraperitoneal BrdU injection. While there was no difference in the total number of CD45⁺CD115⁺F4/80⁺ macrophages (Figure 5A), there was a significant increase in the percentage of BrdU⁺ macrophages from B6-Ldlr^−/−, Cdkn2a^+/− mice compared to B6-Ldlr^−/− controls (p<0.04) (Figure 5B–C). We also tested for increased BrdU incorporation within lesions of a small cohort of B6-Ldlr^−/−, Cdkn2a^+/− mice compared to controls using immunohistochemistry. While ample numbers of cells (mean=42–47 cells) stained positively following a 7-day pulse, the within-group variation was high and power calculations indicated that we would need to study 50 mice/group to have an 80% chance of detecting a significant difference (p<0.05) comparable in magnitude to the difference in lesion area (i.e. ~30%).

Figure 5. Increased proliferation of tissue macrophages derived from B6-Ldlr^−/− mice transplanted with Ldlr^+/−, Cdkn2a^+/− BM compared to controls.

A, Flow cytometry analysis of concanavalin A-elicited peritoneal macrophages from B6-Ldlr^−/−or B6-Ldlr^−/−, Cdkn2a^+/− mice fed 4–5 wks WTD. Macrophages were gated as CD45⁺CD115⁺F4/80⁺. B, Analysis of proliferating CD45⁺CD115⁺F4/80⁺ macrophages at the 4–5 wk timepoint. Cells were gated as BrdU⁺. C, Quantification of BrdU⁺ cells among total CD45⁺CD115⁺F4/80⁺ macrophages. N=6–7 mice/group. *p≤0.05.

Together, these data provide direct evidence for a suppressive effect of BM-derived p16^INK4a and/or p19^ARF expression on inflammatory monocyte/macrophage proliferation.

Discussion

Prior to the identification of 9p21 as a risk locus for human ASVD, we used a murine genetic approach to identify the homologous region of mouse chr 4 as a modifier of atherosclerosis susceptibility. In the current study, we provide refined genetic mapping and transcriptional evidence that Cdkn2a mediates at least some of the altered atherosclerosis susceptibility of this region through altered expression in macrophages. In support of this model, heterozygous deficiency of Cdkn2a transcripts in BM-derived cells was found to be sufficient to confer accelerated atherogenesis in the B6-Ldlr^−/− background. The moderate (~2–4-fold) reduction in p16^INK4a and p19^ARF expression associated with increased atherosclerosis in these murine models is consistent with the reduction of p16^INK4a and p19^ARF expression in T cells of carriers of the 9p21 risk allele.⁸ Moreover, our study shows for the first time that the underlying pathogenic mechanism may involve increased proliferation/expansion of the Ly6C^hi inflammatory monocyte population in the circulation as well as increased proliferation of tissue macrophages. This provides a plausible mechanism to account, in part, for accelerated atherogenesis in the chr 4 congenic mouse, with possible relevance to the mechanism of increased atherogenesis in humans bearing the 9p21 risk allele.

Consistent with our BM transplantation study, whole-body deletion of p19^ARF was recently shown to be pro-atherosclerotic in the B6-Apoe^−/− background.²³ The mechanism of action suggested in the Apoe^−/− model was decreased apoptosis, although the culprit cell type was not identified. We did not observe differences in plaque apoptosis in either the Athsq1 congenic model or mice carrying BM-deficiency of Cdkn2a compared to respective controls. The discrepancy between the two studies could be due to different experimental designs (BM vs whole body deficiency), different targeted alleles (p16^INK4a/p19^ARF vs p19^ARF alone), different mouse models (Ldlr^−/− vs Apoe^−/−), or a non-target effect of a carrier gene in one of the mutant models. However, while decreased apoptosis may accelerate early lesion formation, increased apoptosis likely contributes to lesion progression to advanced plaques with clinically significant consequences²⁴ such as MI associated with 9p21.

Chr 4 mouse deletion mutants with decreased expression of all three Cdkn2a/b transcripts exhibited increased proliferation of vSMCs ²⁵. However, vSMC proliferation does not readily explain the human athero-thrombotic phenotype since this generally leads to cap thickening and plaque stabilization.²⁶ The results of a reverse BM transplantation experiment reported herein were inconsistent with a role for vSMCs in the atherosclerosis phenotype of Athsq1 17-Mb subcongenic mice (Supplementary Figure 2). In line with this, Folkersen et al²⁷ failed to demonstrate an association between vascular gene expression of CDKN2A/B and 9p21 risk genotype.

Evidence for an association between elevated WBC counts and ASVD risk has been documented in more than a dozen prospective epidemiological studies.²⁸ There is also substantial evidence in animal models for a causal relationship between monocytosis and atherogenesis.^29–32 In mouse models of diet-induced atherosclerosis, the subset of Ly6C^hi expressing monocytes expands with hypercholesterolemia and selectively enters sites of inflammation including atherosclerotic lesions, compared to the Ly6C^lo subset.^{33, 34} In addition, single gene mutations resulting in reduced blood monocytes also reduced atherosclerosis independent of plasma cholesterol levels.^35–38 While the relationship between monocytosis and atherosclerosis has been well-documented, detailed mechanisms responsible for monocytosis are lacking. Consistent with a recent finding,¹⁹ our study suggests that the control of myeloid cell proliferation during hypercholesterolemia is an important factor determining magnitude of leukocytosis and the atherogenic response.

A variety of mechanisms for CDKN2A/B and/or ANRIL involvement in the 9p21 ASVD locus have been proposed. Regulation of CDKN2A/B gene expression has been suggested to occur via cis or trans mechansim(s) involving either the structurally overlapping ANRIL,^{9, 39} or other regulatory motifs residing within the 58-kb risk locus.^{11, 40} It is also possible that ANRIL may mediate effects at 9p21 without involvement of CDKN2A/B, through regulation of gene expression at an unlinked locus.⁴¹ The mouse genome does not contain a contiguous sequence with homology to ANRIL. However, markedly decreased expression of Cdkn2a/b transcripts in the 70-kb deletion mutant suggests the existence of a cis-acting enhancer.¹³ In our Athsq1 congenic mouse, there is a loss of coordinate regulation with decreased expression of Cdkn2a (p16^INK4a and p19^ARF) transcripts but not Cdkn2b (p15^INK4b). Thus, in our model, the causal variant is likely to be cis-acting and specific to the Cdkn2a locus.

The 9p21 locus has been associated with multiple vascular phenotypes. These include myocardial infarction,⁵ abdominal aortic and intracranial aneurysms,⁴² ischemic stroke,^{43, 44} and peripheral artery disease.⁴⁵ Interestingly, monocyte/macrophage recruitment is an important process in cerebral aneurysm formation and Ccl2 deficiency has been shown to inhibit macrophage accumulation in aneurysmal walls and significantly decrease aneurysm formation in an experimentally induced mouse model.⁴⁶ In another study, increased immunostaining of CD68 antigen was observed in intracranial aneurysms compared to control tissue.⁴⁷ In ischemic stroke, macrophage accumulation has been associated with severity of brain injury.⁴⁸ Thus, alterations in monocyte/macrophage proliferation could potentially provide a common underlying mechanism for vascular phenotypes associated with the 9p21 locus.