ABCC6 deficiency promotes dyslipidemia and atherosclerosis

ABCC6 deficiency promotes ectopic calcification; however, circumstantial evidence suggested that ABCC6 may also influence atherosclerosis. The present study addressed the role of ABCC6 in atherosclerosis using Ldlr−/− mice and pseudoxanthoma elasticum (PXE) patients. Mice lacking the Abcc6 and Ldlr genes were fed an atherogenic diet for 16 weeks before intimal calcification, aortic plaque formation and lipoprotein profile were evaluated. Cholesterol efflux and the expression of several inflammation, atherosclerosis and cholesterol homeostasis-related genes were also determined in murine liver and bone marrow-derived macrophages. Furthermore, we examined plasma lipoproteins, vascular calcification, carotid intima-media thickness and atherosclerosis in a cohort of PXE patients with ABCC6 mutations and compared results to dysmetabolic subjects with increased cardiovascular risk. We found that ABCC6 deficiency causes changes in lipoproteins, with decreased HDL cholesterol in both mice and humans, and induces atherosclerosis. However, we found that the absence of ABCC6 does not influence overall vascular mineralization induced with atherosclerosis. Decreased cholesterol efflux from macrophage cells and other molecular changes such as increased pro-inflammation seen in both humans and mice are likely contributors for the phenotype. However, it is likely that other cellular and/or molecular mechanisms are involved. Our study showed a novel physiological role for ABCC6, influencing plasma lipoproteins and atherosclerosis in a haploinsufficient manner, with significant penetrance.


Material and methods
The data that support the findings of this study are available from the corresponding author upon reasonable request. Because of the sensitive nature of the data from human subjects in this study, these may be subjected to approval by the authors who provided them (LM and GL).
Animals. Abcc6 tm1Aabb mice were generated on a 129/Ola background 25 and backcrossed into a C57BL/6J > 10 times. Ldlr tm1Her mice in the C57BL/6 background were provided by WAB who purchased them from The Jackson Laboratory (Bar Harbor, ME). These mice are hereafter designated as Abcc6 −/− and Ldlr −/− , respectively. By simple crosses, we have generated Abcc6 haploinsufficient (Abcc6 +/− ;Ldlr −/− ) and double knockout (Abcc6 −/− ;Ldlr −/− ) animals to determine if a gene dosage affects atherosclerosis and associated calcification. Ldlr −/− mice with normal Abcc6 expression were used as controls. In this study, both male and female, age-matched mice were used, as gender had no significant impact on results. All mice were housed in pathogen-free rooms in AAALACaccredited animal facilities at the University of Hawaii School of Medicine. Mice were kept under routine laboratory conditions with 12-h light-dark cycle with ad libitum access to water and either a standard chow diet or an atherogenic high fat diet (HFD) that contained 15.8% (wt/wt) fat, 1.25% (wt/wt) cholesterol and no cholate (diet #94059; Harlan Teklad). The University of Hawaii Institutional Animal Care and Use Committee approved these studies. Experiments were conducted according to the NIH Guide for the Care and Use of Laboratory Animals.

PXE patients and healthy controls. Fasting plasma samples and phenotypical data for PXE patients
were obtained from the MAGEC PXE Reference Center of the Angers University Hospital, France. The diagnosis of PXE was based on a combination of established criteria for indisputable PXE, including clinically suggestive skin lesions, angioid streaks and histologically proven fragmented and calcified elastic fibers on skin biopsy 26 . For lipoprotein analyses, PXE and controls plasma samples came from subjects that were not treated with cholesterol lowering medication. Control plasma samples came from a group matched for gender and age without evidence of PXE or other pathology.
For atherosclerotic plaques assessment, results were compared to the NUMEVOX cohort. This cohort of subjects with at least one metabolic syndrome criterion with or without sleep apnea or liver steatosis is registered on clinicaltrials.gov (NCT00997165). The institution review board of the Angers University Hospital, France has approved these studies and the participants have given informed consent when required. All methods were carried out in accordance with relevant guidelines and regulations. www.nature.com/scientificreports/ respond to small dense LDL particles. Similarly, 10 HDL sub-fractions are detected. HDL-1 to -3 represent large particles, sub-fractions − 4 to − 7 are reported as medium particles, and sub-fractions − 8 to − 10 as small particles. This system also calculates the cholesterol concentration for each lipoprotein fraction according to a total cholesterol value obtained for each plasma sample.
Macrophage isolation and differentiation. Bone marrow-derived macrophages were isolated and cultured as follows. Briefly, 3-month-old mice were euthanized by carbon dioxide inhalation. Hind legs were exposed and the muscle tissue removed to isolate the femurs and tibiae. The bone marrow was flushed from femurs and tibiae with culture medium (DMEM) containing 10% fetal calf serum and 1% penicillin/ streptomycin/l-glutamine (Invitrogen, Carlsbad, CA) using a syringe and a 25-gauge needle. The cell suspension was drawn through a syringe with an 18-gauge needle to dissociate cell clumps and was passed through a 70-μm pore cell strainer (BD BioSciences, San Jose, CA) to remove tissue debris. The cells were plated and incubated for 5 days in the presence of 25 ng/ml macrophage colony stimulating factor (M-CSF) ( Immunohistochemistry on macrophages. Bone marrow-derived macrophages were seeded in 24-well plates onto small coverslips placed inside the wells. Cells were cultured as described above with or without 50 μg/ mL acetylated LDL (for foam cell differentiation). Cells were then fixed and permeabilized with cold methanol for 5 min at − 20 °C. Rabbit monoclonal anti-ABCA1 (NB400-105SS, Novus Biologicals, Littleton, CO) and anti-ABCG1 (PA5-13462, Thermo Fisher, Waltham, MA) antibodies were used to specifically detect ABCA1 and ABCG1 in BMDMs. The secondary antibody was AlexaFluor 488 (Life Technologies, CA). The coverslips with the cells from the wells were placed onto slides and immunofluorescent images were acquired using an Axioskop 2 fluorescent microscope (Zeiss, Thornwood, NY). Individual images were collected and processed with Photoshop CS6 (Adobe, San Jose, CA).
Western blot. ABCA1 and ABCG1 protein expression in bone marrow-derived macrophages and foam cells were also detected using standard western blot techniques we previously described 18 . Proteins were extracted on ice in RIPA buffer containing 1X protease inhibitor mini complete cocktail (Roche Applied Science, Indianapolis, IN), and 5 mmol/L EDTA. Homogenates were centrifuged at 16,000 g and the protein concentration in the supernatant was determined by absorbance using a Nanodrop spectrophotometer (ThermoScientific, Wilmington, DE). 50 μg was combined with reducing Novex sample buffer (Invitrogen, Carlsbad, CA) and loaded into wells of a 4-12% polyacrylamide gel (Invitrogen, Carlsbad, CA). Proteins were transferred to nitrocellulose membrane and blotted with the anti-ABCA1 and ABCG1 antibodies (NB400-105SS, Novus Biologicals, Littleton, CO and PA5-13462, Thermo Fisher, Waltham, MA). Appropriate secondary antibodies were detected using the Odyssey infrared imaging system (Li-cor, Lincoln, NE). β-actin was used as a loading control (Abcam Ab8229, Cambridge MA) and mouse lung lysates were used as a positive control (not shown).

Data analysis.
Results are presented as the mean +/− standard error of the mean (SEM). Animal numbers used for individual sets of data varied between experiments and are shown on the figures. For multigroup comparisons, one-way ANOVA (F) followed by the Tukey HSD-test was used for detecting the statistical differences.
The ANOVA F values shown in the results are the ratio of two mean square values. If the null hypothesis is true, F is expected to have a value close to 1.0. A large F ratio signifies that the variation among group means is more than you' d expect to see by chance. The Student's t-test was used for detecting differences between two groups by comparing the effect of one variable such as genotype, cell differentiation or diet. Relative gene expression was calculated using the delta-delta threshold method 29 . Difference between relative gene expressions greater than 1.5-fold were considered statistically significant. Results from the comparison between PXE patients and the NUMEVOX cohort (described in the results section and Table 2) were analyzed using Kruskal Wallis and performed with Stata 12.0 software (StataCorp, Texas, USA). The level of statistical significance was set at P < 0.05. Statistical comparison into the groups were performed using Wilcoxon's test. A p value < 0.05 was considered statistically significant. Analyses were performed using GraphPad's PRISM software v.8.4.2 (San Diego, CA).

Results
ABCC6 deficiency induces dyslipidemia in mice and humans. The premise of this study was that ABCC6 may influence the levels of plasma lipoproteins and susceptibility to atherosclerosis. To explore this possibility, we generated 3 mouse genotypes with normal, reduced and no Abcc6 expression in an atherogenic background (Ldlr −/− ) to determine if gene dosage affects dyslipidemia, atherosclerosis and possibly vascular calcification.
Mice. We first determined the effect of ABCC6 deficiency on lipoprotein profiles using a Lipoprint analyzer in several mouse models fed either normal chow or a high fat diet for 16 weeks. We used Abcc6 +/− ;Ldlr −/− and Abcc6 −/− ;Ldlr −/− mice in comparison with Ldlr −/− controls as well as Abcc6 −/− and wild type (WT) animals. The results, summarized in Table 1, showed that under non-atherogenic conditions (i.e. normal diet and Ldlr +/+ status), ABCC6 deficiency leads to substantially reduced HDL and increased LDL levels (p = 0.03 and p = 0.0035, n = 8 versus n = 3, respectively). Lipoprint analysis allowed the quantitative evaluation of the main HDL and LDL sub-fractions and notably revealed that the large HDL subclass, which shows an inverse relationship to cardiovascular risk, was significantly reduced in Abcc6 −/− mice (Supplemental Table 1). Interestingly, the highly atherogenic sub-fractions LDL 3 through 7 (labelled as "Large LDL" on Supplemental Table 1) were elevated in Abcc6 −/− mice but only in animals fed normal chow. When compared to control mice, total cholesterol and HDL levels were particularly affected in Abcc6 −/− mice fed a lipid-rich diet, though no atherosclerosis was noted in these animals (not shown). Triglyceride levels were significantly higher in all Abcc6-deficient animals as compared to Ldlr −/− or WT controls, except for the double KO on high fat diet (Table 1). Under severe atherogenic conditions (Ldlr −/− and high fat diet), experimental mice displayed a ~ tenfold increase in total cholesterol as compared to Ldlr +/+ mice but there was no significant difference between the three genotypes. However, major alterations of HDL and LDL levels were found in both Abcc6 +/− ;Ldlr −/− and Abcc6 −/− ;Ldlr −/− mice in comparison to Ldlr −/− controls ( Table 1). The gradual decrease to undetectable levels of the small HDL fraction was notable (Supplemental Table 1). Because ABCC6 deficiency did not affect atherosclerosis development in mice with an ApoE −/− background as reported 30 , we specifically examined the plasma levels of this apolipoprotein in Abcc6 +/− ;Ldlr −/− and Abcc6 −/− ;Ldlr −/− mice, as it may have been a contributor to the enhanced atherosclerosis phenotype we observed. However, there was no significant variation as compared to Ldlr −/− controls (Supplemental Fig. 1A). PXE patients. We followed up the murine investigation with a comprehensive analysis of the lipoproteins from PXE patients (Table 1). Total cholesterol levels were not statistically significant from healthy age-and sex-matched controls. However, the patients presented a highly significant reduction in HDL levels (− 24%, p = 0.0004) in the same proportion as Abcc6 −/− vs wild type mice on a normal diet. The small and large HDL sub- www.nature.com/scientificreports/ fractions were also decreased as compared to healthy controls (Supplemental Table 1) but there was no change in LDL, unlike the animal models. We found no difference in plasma triglyceride levels ( Table 1) or plasma ApoE concentration (Supplemental Fig. 1B) overall between PXE patients and healthy controls. However, significant differences in triglyceride levels were found between sub-groups. Triglycerides were elevated in control males (+ 82%, but still within normal range) as compared to control females (p = 0.0013, n = 10 versus n = 24). Average plasma triglyceride levels were also higher (+ 47%) in female patients vs female controls (p = 0.0066, n = 24), but, remarkably, there was no difference between male and female PXE patients (p = 0.88).
The lack of ABCC6 enhances atherosclerosis. Mice As the highest level of Abcc6 expression is normally found in the liver 12 , we verified whether the lack of hepatic ABCC6 function could affect the expression of selected genes involved in reverse cholesterol transport (Abca1, Abcg1, Sr-b1, ApoA1) or cholesterol metabolism (3-hydroxy-3-methylglutaryl CoA reductase or Hmgcr and Cyp7α1). There was no significant change in Abcc6 −/− ;Ldlr −/− mice as compared to control Ldlr −/− mice fed normal or high fat diets except in Abcg1 expression which was significantly downregulated (p = 0.02) under high fat dietary conditions in an Abcc6-dependent manner (Supplemental Fig. 2B).

PXE patients.
To determine if PXE patients are susceptible to vascular remodeling, we performed a retrospective analysis of data collected from 142 patients routinely followed at the University Hospital of Angers, France (by L.M. and G.L.). The effect of ABCC6 deficiency on the vasculature has already been extensively documented in many comparisons with healthy controls 22,[31][32][33][34][35] . Therefore, to provide a novel disease perspective to which the effects of the PXE pathophysiology could be gauged against, we compared data from PXE patients with a cohort already at risk for significant cardiovascular anomaly i.e. 668 dysmetabolic individuals screened from the NUMEVOX cohort 36 .
In both cohorts, carotid intima-media thickness (IMT) was determined using ultrasound imaging coupled with echotracking according to standardized methods while plaques were scored (i.e. presence/absence) bilaterally at the carotid, abdominal aorta and femoral bifurcations. Note that in the absence of a direct histopathology analysis, the nature of the plaque seen in PXE patients and control individuals is herein assumed to be of atheromatous nature. PXE patients showed no significant difference in intima-media thickness (IMT) as compared to the cohort at risk (617 ± 164 µm versus 650 ± 121 µm, p = 0.720). Adjustment for age and number of sites with atherosclerotic plaques did not affect the lack of statistical significance (β = − 0.0011, p = 0.312). However, PXE patients who had different (better) lipid profiles with modest dyslipidemia (average cholesterol ratio of 3.7) presented a higher number of atherosclerotic sites as compared to the NUMEVOX cohort (1 ± 3 versus 2 ± 4, Enhanced systemic inflammation is associated with ABCC6 dysfunction. Abcc6-deficient mice. It is now well accepted that atherosclerosis is a chronic inflammatory disease mediated by the concerted action of many circulating pro-inflammatory cytokines. Therefore, we measured selected cytokines in the plas-  (Fig. 2). This result suggested that the www.nature.com/scientificreports/ increased calcification was only a consequence of atherosclerotic plaque (intimal) and not linked to the Abcc6 genotype. This conclusion is supported by the increased mineralization in Abcc6 +/− ;Ldlr −/− and Abcc6 −/− ;Ldlr −/− mice fed regular chow when compared to control Ldlr −/− and Abcc6 −/− mice, as the first two mouse models presented atherosclerotic lesions whereas the two controls did not (Fig. 2). For control purposes, we evaluated whether the absence of the Ldlr gene and a lipid-rich diet would alter the typical PXE calcification phenotype found in Abcc6 −/− mice. The lack of Ldlr alone or in combination with high fat diet had no significant influence on the mineralization of the vibrissae capsule (Supplemental Fig. 3A). High fat diet can also induce cardiac calcification in mice lacking functional ABCC6 20 . Here, there was no evidence of high fat diet-related DCC in the relatively short timeframe (16 weeks) of this study (Supplemental Fig. 3B).

PXE patients.
As skin calcification cannot be readily quantified in PXE patients, we determined arterial calcium score as previously described 34 . Not surprisingly, our cohort of patients displayed significantly elevated arterial calcium score as compared to healthy controls (TEP scan units 1356 ± 430.3, n = 16 vs 137.9 ± 68.3, n = 17 respectively, p = 0.007.), however, these measures did not readily distinguish medial from intimal calcification.
The role of macrophages in dyslipidemia and atherosclerosis associated with ABCC6 deficiency in mice. Macrophages play a prominent role in the development of atherosclerotic lesions, and this has been extensively documented 37,38 . These immune cells normally efflux cholesterol to ApoA1, nascent and p-values were determined by one-way ANOVA and Tukey's post hoc test (*) or Student's t-test (#).*, # p < 0.05, **p < 0.01, ***, ### p < 0.001. Table 2. Logistic regression between disease status and age, number of atherosclerotic plaque sites and intima media thickness (IMT). β = slope coefficient, SEM = standard error of the mean. Constant = predicted value when all the X variables = 0. p-values were determined using Wilcoxon's test. *p < 0.05, ***p < 0.0001. www.nature.com/scientificreports/ mature HDL via the ATP-binding cassette transporter A1 (ABCA1) and ABCG1, which are key steps in reverse cholesterol transport. The dysregulation of this process in vascular macrophages promotes atherogenesis 39 . To determine the role of macrophages in the dyslipidemia and atherosclerosis phenotypes we observed in ABCC6deficient mice (and PXE patients), we first used bone marrow-derived macrophages isolated from our experimental mice to study cholesterol efflux. We found no difference of baseline efflux (without acceptor) or efflux towards ApoA1 between the three genotypes. However, when HDL particle acceptors were present, there was a small but significant decrease in efflux in equivalent proportions in Abcc6 +/− ;Ldlr −/− and Abcc6 −/− ;Ldlr −/− macrophages (Fig. 3).

Model
To investigate a possible cause of this decrease in efflux, we quantified the expression of Abca1, Abcg1, ApoE, Hmgcr, Cd36 and Sr-b1 genes in both macrophages and foam cells from Ldlr −/− and Abcc6 −/− ;Ldlr −/− mice. The differentiation of macrophages into foam cells enhanced the expression of Abca1 and Abcg1, but there was no Abcc6specific effect (Fig. 4A,B). These results were confirmed at the protein level as well (Fig. 4C,D,E). Similarly, the expression of Hmgcr and Sr-b1 dropped in foam cells but no Abcc6-dependent effect was observed (Fig. 4G,H,I). This result is somewhat similar to previous findings 40 . ApoE expression (Fig. 4F) showed no difference, which was consistent with plasma levels (Supplemental Fig. 1A), though Cd36 showed a modest (but not significant) expression decrease in Abcc6 −/− ;Ldlr −/− macrophages (Fig. 4H). Because the changes in gene expression were either small and/or inconsistent between macrophages and foam cells, they seemed unlikely to be physiologically relevant. The expression of Enpp1 and Nt5e was also analyzed as these genes encode proteins immediately downstream of the pathway initiated by ABCC6 5,9,10 . There was no significant change for Enpp1, but the expression of Nt5e was upregulated in both macrophages and foam cells from Abcc6 −/− ;Ldlr −/− mice (Fig. 5A,B). Furthermore, we analyzed the expression of selected cytokine genes that play a role in atherosclerosis. We only found a significant increase in the expression of Ccl-2  in Ldlr −/− ;Abcc6 −/− macrophages vs Ldlr −/− controls, Fig. 5C) whereas Ccr-2 and TNF-α showed not significant change (Fig. 5D,E).
Remarkably, we found no significant Abcc6 expression in bone marrow-derived macrophages M1 or M2 activated or foam cells from Ldlr −/− mice (not shown).
Changes in plasma bile acids in mice lacking ABCC6. Cholesterol (as HDL) is brought to the liver via reverse transport for recycling or elimination. Because bile acids are quickly absorbed in the intestines and recy- In both dietary conditions, we found a profound decrease in the circulation of the following bile acid concentrations: cholic, α-and β-muricholic acids, chenodeoxycholic acid, deoxycholic acid, ω-muricholic acid, hyodeoxycholic acid and ursodeoxycholic acid. However, this decrease was more pronounced in mice fed high fat diet than normal chow (Fig. 6). Lithocholic acid was not detectable in our samples (not shown) but the other 8 primary, secondary and tertiary bile acids showed drastic reductions ranging from − 55 to − 99% (p < 0.0001) in Abcc6 −/− ;Ldlr −/− mice with cholic, β-muricholic and ω-muricholic acids being the most affected, decreasing to undetectable levels.
In the final steps of the reverse cholesterol transport, the cholesterol half-transporters ABCG5 (ATP-binding cassette transporter G5) and ABCG8 (ATP-binding cassette transporter G8) play a critical role. Both hemitransporters are present on the apical membrane of hepatocytes facilitating excretion of cholesterol and plant sterols into the bile and are also expressed in the intestinal epithelial cells 41,42 . Because elevated expression of both ABCG5/G8 in Ldlr −/− mice contributes to the attenuation of diet-induced atherosclerosis 43 , we examined the expression of Abcg5/8 in the liver of WT, Abcc6 −/− , Ldlr −/− and Abcc6 −/− ;Ldlr −/− mice. We found that the deletion There was a trend towards lower expression of Abcg5/8 in Ldlr −/− mice lacking ABCC6 but the difference was not statistically significant with the limited number of samples available for this experiment (Fig. 7A). However, with mice fed high fat diet, the hepatic expression of Abcg5 and Abcg8 decreased significantly (p = 0.020, p = 0.016 respectively, n = 4) in Abcc6 −/− ;Ldlr −/− mice as compared to Ldlr −/− controls (Fig. 7B), which is consistent with the reduction in plasma bile acids.

Discussion
Several publications describing PXE patients and Abcc6-deficient mouse models have previously reported sporadic and circumstantial evidence of alteration of lipoprotein levels and arteriosclerosis 16,25,[44][45][46][47] . However, the relationship between ABCC6, lipoproteins and calcification has never been specifically studied. This study examined this association using mouse models under normal or atherogenic conditions and compared results with human PXE patients. We found that ABCC6 deficiency caused significant changes in lipoprotein levels and enhanced atherosclerosis, which were associated with decreased cholesterol efflux from macrophages and the dysregulation of genes involved in purinergic signaling including Nt5e 5,6,48,49 , and in reverse cholesterol transport (ABCG5/8). The presence of pro-inflammatory conditions was likely a compounding factor. Remarkably, dyslipidemia and the development of atherosclerotic lesions occurred in a haploinsufficient manner, unlike the calcification phenotype caused by ABCC6 mutations.  47 . The authors of this study reported an association between an ABCC6 variant (now considered non-pathogenic 35,50,51 ) and elevated HDL levels and low triglycerides. However, these results are inconsistent with what others have since described 22,31,33 . A few years later, Gorgel et al. obtained more solid but limited lipoprotein data using mice 25 . The results presented here have now largely expanded upon these initial observations with a more detailed analysis of both mice and humans lacking functional ABCC6. Remarkably, we observed in our mouse model with an atherogenic background (Abcc6 −/− ;Ldlr −/− ) the formation of atherosclerotic plaque in the aorta even under normal dietary conditions and in the absence of dyslipidemia. These results suggest that plaque development occurs in these mice independently of LDL/HDL levels, and imply that other molecular, cellular and/or physiological mechanism(s) are involved. The elevated triglycerides under these normal dietary conditions were probably one of the contributing factors 52 . When these mice were challenged with high fat diets, ABCC6-specific differences in lipoprotein levels and plaque development were dramatically enhanced indicating that ABCC6 influences lipoprotein levels and atherosclerosis through independent but compounding mechanisms. animal model used to study human diseases, the biology of rodents can be noticeably divergent 53 . Abcc6 −/− mice, despite the known differences with the human phenotype 54 , have been extraordinarily useful for the study of PXE 55,56 and the development of therapeutic approaches 49,[57][58][59][60] . However, mice present singular differences in lipoproteins and atherosclerosis development 61 . For this reason, the overlapping characteristics between animal and PXE patient data weighed heavily in our interpretations. Even though we could only speculate as to what the link between ABCC6 molecular function and HDL metabolism could be, we distinguished in our data and the literature several arguments suggesting that ABCC6 modulates atherosclerosis by influencing HDL quantity and/or quality 62 , as it is the most common denominator between our mouse models and human subjects.
1. We explored possible cellular mechanisms that could enhance atherosclerosis. We first examined cholesterol efflux from bone marrow-derived macrophages after lipid loading. Reduced cholesterol efflux from macrophages within the plaque is probably a significant contributing factor to the enhanced atherosclerotic plaque development we observed in the animal models and PXE patients especially in the absence of changes to LDL in humans. However, the reason behind the strictly ABCC6-dependent decrease in cholesterol efflux is unclear. Indeed, there was no significant change in Abca1 or Abcg1 expression and protein levels or in other related genes between control and experimental macrophages and foam cells. Furthermore, the notable absence of Abcc6 mRNA in macrophages likely reflected an acquired phenotype that persisted ex-vivo after bone marrow isolation and macrophage differentiation. We and others have previously described similar situations with primary skin fibroblasts, which express little to no ABCC6, from PXE patients that retain altered characteristics ex-vivo 40,63,64 . An alternative explanation could be that even a minimal ABCC6 presence may exert a meaningful impact on cellular functions as Hendig and co-workers suggested 65,66 . 2. A recent study of ABCC6-deficient mice in ApoE −/− background showed no significant changes in atherosclerosis (or arterial calcification) as compared to ApoE −/− controls 30 . The results of this study are not necessarily discordant with ours and in fact provided some support to our overall conclusion regarding HDL. ApoE −/− mice are a very useful and popular animal model. However, atherogenesis in mice is quite distinct from humans as HDL cholesterol is virtually absent in these mice 67 and atherosclerosis is not a distinguishing feature of humans lacking ApoE 68 . In that sense, Ldlr −/− mice provide a pathophysiology somewhat closer to that of humans as plaque formation is primarily driven by higher LDL and lower HDL 69 and this is what we observed in the present study (Table 1, and Fig. 1). Since Van der Veken et al. reported no difference in atherosclerosis or lipoproteins between Abcc6 −/− ;ApoE −/− mice and ApoE −/− controls, we first suspected that ApoE could play a role in our model. Indeed, the absence of ApoE in macrophage promotes atherosclerosis without changing plasma cholesterol and we found that atherosclerosis occurred independently of dyslipidemia 70 . However, we found no significant variation in the plasma of mice or in PXE patients. Another likely possibility was that ApoE −/− mice have very little or no circulating HDL. If ABCC6 modulates atherogenesis by its differential effect on HDL quantity (Table 1), and/or quality 62    www.nature.com/scientificreports/ 3. Reverse cholesterol transport, which is primarily mediated by HDL, is a mechanism by which excess cholesterol is removed from peripheral tissues and delivered to the liver for either redistribution or excretion in the form of bile acids. About 95% of bile acids are reabsorbed during the digestive process. Thus, lower plasma bile acid level signals either a reduction of the enterohepatic recirculation (intestinal absorption and liver reabsorption) and/or decreased reverse cholesterol transport. Lower HDL plasma concentration, reduced cholesterol efflux from foam cells and the dramatic changes in plasma bile acids are all consistent with altered reverse cholesterol transport. The lack of change in Sr-b1, ApoA1 and Cyp7α expression indicated that the canonical FXR pathway was unaffected in liver and also agrees with this interpretation. 4. As circumstantial evidence, a recent comparative study of polar and brown bear genomes identified ABCC6 as one of 20 genes under positive selection that allowed polar bears to adapt to a lipid-rich diet and elevated plasma LDL 71 . 5. Furthermore, we found clear evidence for an ABCC6-dependent pro-inflammatory status both at the cellular levels in murine macrophages and foam cells with overproduction of CCL-2 as well at the systemic levels with elevated plasma levels of IL-6 both in PXE patients and our mouse models. The role of inflammation in atherosclerosis development is well-documented and thus these cytokines and chemokines are likely factors aggravating atherosclerosis associated with ABCC6 deficiency.
In conclusion, the present study reveals a novel physiological role for ABCC6, influencing plasma lipoproteins and atherosclerosis in a haploinsufficient manner, with significant penetrance. Figure 8. The proteins/enzymes in this pathway regulate both calcification and atherosclerosis. ABCC6 activity facilitates the cellular efflux of ATP from liver and other tissues/cells, which is quickly converted to pyrophosphate (PPi), a potent inhibitor of mineralization. Decreased plasma PPi levels cause calcification in PXE (OMIM #264800) and GACI (OMIM #614473 and #208000). NT5E activity leads to adenosine production which has numerous biological activities towards inflammation, atherosclerosis and inhibition of TNAP synthesis. TNAP degrades PPi into inorganic phosphate (Pi), an activator of calcification which leads to vascular calcification in Calcification of Joints and Arteries patients (CALJA, OMIM #211800). In addition to ABCC6, both ENPP1 and NT5E functions have been linked to atherosclerosis development in ApoE −/− mice. ENTPD1 function overlaps with that of ENPP1 and also plays a role in atherosclerosis in mice lacking ApoE, though this ectoenzyme does not seem to regulate ectopic calcification. Symbols: ∅ : no known effect; ➚: increase; ➘: decrease.