Diet-induced Diabetes Activates an Osteogenic Gene Regulatory Program in the Aortas of Low Density Lipoprotein Receptor-deficient Mice*

Vascular calcification is common in people with diabetes and its presence predicts premature mortality. To clarify the underlying mechanisms, we used low density lipoprotein receptor-deficient (LDLR −/−) mice to study vascular calcification in the ascending aorta. LDLR −/− mice on a chow diet did not develop obesity, diabetes, atheroma, or vascular calcification. In contrast, LDLR −/− mice on high fat diets containing cholesterol developed obesity, severe hyperlipidemia, hyperinsulinemic diabetes, and aortic atheroma. A high fat diet without cholesterol also induced obesity and diabetes, but caused only moderate hyperlipidemia and did not result in significant aortic atheroma formation. Regardless of cholesterol content, high fat diets induced mineralization of the proximal aorta (assessed by von Kossa staining) and promoted aortic expression ofMsx2 and Msx1, genes encoding homeodomain transcription factors that regulate mineralization and osseous differentiation programs in the developing skull. Osteopontin(Opn), an osteoblast matrix protein gene also expressed by activated macrophages, was up-regulated in the aorta by these high fat diets. In situ hybridization showed that peri-aortic adventitial cells in high fat-fed mice expressMsx2. Opn was also detected in this adventitial cell population, but in addition was expressed by aortic vascular smooth muscle cells and macrophages of the intimal atheroma. High fat diets associated with hyperinsulinemic diabetes activate an aortic osteoblast transcriptional regulatory program that is independent of intimal atheroma formation. The spatial pattern ofMsx2 and Opn gene expression strongly suggests that vascular calcification, thought to be limited to the media, is an active process that can originate from an osteoprogenitor cell population in the adventitia.

More than a century ago, Virchow (1) recognized two components of vascular disease, atheroma formation and calcification. Cholesterol-containing lipoproteins are probably necessary but not sufficient for atheroma formation. Their presence sets the stage for macrophage activation, foam cell formation, recruitment of smooth muscle cells, and fibrosis (2). In contrast, vascular calcification is largely uncharacterized.
Vascular calcification is particularly relevant to people with diabetes. Atherosclerosis accounts for 80% of diabetes-related deaths, and vascular disease in diabetics carries a far worse prognosis than in nondiabetics (3)(4)(5). Patients with type II diabetes are more likely than nondiabetics to develop macrovascular calcification (6), which is associated with accelerated cardiovascular event rates and death (7). The underlying mechanisms are unknown.
Sometimes considered a passive process of calcium deposition in necrotic tissue, vascular calcification can also be active, associated with expression of osteoblast bone matrix proteins that regulate mineralization. A transcriptional hierarchy controlling osteoblast gene expression and mineralization has emerged (8). The runt domain transcription factor Cbfa1/Osf2 controls neurectodermal and mesodermal tissue mineralization prenatally (9). In contrast, the homeodomain proteins Msx2 and Msx1 control mineralization in the developing skull, programming osteoblasts and odontoblasts that arise from neurectoderm (10,11).
Using LDL 1 receptor-deficient (LDLR Ϫ/Ϫ) mice, we studied mineralization of the ascending aorta, a vascular segment derived from neurectoderm and particularly prone to calcification in diabetic patients. We addressed the question: Does dietinduced diabetes promote vascular heterotopic calcification by recruitment of an osteoblast transcriptional regulatory program? Our results show that high fat diets, regardless of cholesterol content, induce vascular mineralization and a syndrome resembling type II diabetes. A component of this vascular calcification reflects up-regulation of osteoblast transcription factors and associated gene expression programs in perivascular adventitial cells that appear to function as osteoprogenitors.

EXPERIMENTAL PROCEDURES
Animals and Diets-LDLR Ϫ/Ϫ mice in the C57BL/6J background were purchased from the Jackson Laboratory (Bar Harbor, ME). The genotype was confirmed in animals upon arrival and periodically in offspring using multiplex PCR as described previously (12). After weaning, pups were fed a chow diet (1:1 mixture of PicoLab rodent chow 20 and mouse chow 20 with a total fat content of 6.75%) until the age of 8 -10 weeks. At that time, mice were either maintained on the chow diet or started on a high fat diet. A high fat, high carbohydrate diet without cholesterol (referred to as "high fat") providing 58% calories as fat was purchased from BioServ (Frenchtown, NJ, product F1850). A high fat diet containing 0.15% cholesterol (referred to as "Western") providing 42% calories as fat was purchased from Harlan Teklad (Madison, WI, product TD 88137). Most experiments used the high fat and Western diets. One type of experiment also used a common but more harsh dietary intervention in an attempt to elicit Cbfa1 expression (see Fig.  4A). For this experiment, a high fat diet containing 1.25% cholesterol and 0.5% cholate (referred to as "Chol. ϩ Fat") was purchased from Harlan Teklad (product TD 95046).
Serum Analyses-Mice were fasted for 4 h then anesthetized with metofane. Blood was collected from the retro-orbital venous plexus with a capillary tube, and serum concentrations of glucose, triglycerides, and cholesterol were determined enzymatically using kits purchased from Sigma. Lipoprotein profiles were analyzed by fast protein liquid chromatography after pooling serum from several animals in each cohort as described previously (13). Insulin concentrations were determined by radioimmunoassay using an antibody to rat insulin (Linco, St. Charles, MO).
Intimal Lesion Quantitation-After 17 weeks of dietary intervention, mice were exsanguinated under anesthesia. Vessels were perfused with phosphate-buffered saline then 4% paraformaldehyde, and the entire aorta was incised and pinned en face as described previously (12). Aortic images were captured with a digital camera and edited to ensure that artifacts were not scored as plaques. Data are reported as the percent involvement of the intimal surface area for the arch, thoracic aorta, and the abdominal aorta.
Reverse Transcription-Polymerase Chain Reaction Analyses of mRNA Accumulation-Cohorts of 8 -10 week old LDLR Ϫ/Ϫ mice were maintained on chow (control), high fat, and high fat diets supplemented with cholesterol as described above for 2, 5, or 17 weeks. At each time point, Ն5 animals were sacrificed in each dietary group. Vessels were perfused with cold lactated Ringer's solution, the ascending aorta (including the aortic valve) with adjacent adventitia dissected, and total RNA isolated as described previously (14). cDNA synthesis and RT-PCR analyses were carried out as described previously (15)  Histologic Specimen Preparation and Staining-LDLR Ϫ/Ϫ mice maintained on the indicated diets were euthanized and perfused as described above. The ascending aorta was dissected, rinsed in phosphate-buffered saline, then fixed in 4% paraformaldehyde in phosphatebuffered saline overnight at 4°C. After fixation, specimens were prepared for in situ hybridization analysis as described previously (16). The von Kossa silver stain for mineral deposition was carried out on paraffin sections using nuclear fast red to counterstain tissue components.
Plasmids and cDNA Templates for Riboprobe Generation and in Situ Hybridization-The plasmids for generating Msx2 (nucleotides 748 -1173, GenBank TM X59292) and osteocalcin (nucleotides 9 -376, Gen-Bank TM X04142) antisense riboprobes were described previously (16). The cDNA fragment for making antisense riboprobe specific for mouse vascular smooth muscle cell ␣-actin (nucleotides 1051-1292, Gen-Bank TM X13297) was generated by RT-PCR from mouse vascular smooth muscle cell RNA using the amplimers 5Ј-GAG CGG ATC CAG AAC GCA AGT ACT CTG TCT GG-3Ј and 5Ј-GCA CGG TAC CAC GAG TAA CAA ATC AAA GCT TTG G-3Ј. The amplified product was subcloned into the KpnI/EcoRI sites of pcDNA3 in antisense orientation to the T7 promoter. The cDNA fragment for making antisense riboprobe for mouse Opn (nucleotides 367-951, GenBank TM J04806) was generated by RT-PCR from mouse MC3T3E1 calvarial osteoblasts using the amplimers 5Ј-AGC GAG GAT TCT GTG GAC TC-3Ј and 5Ј-GTT GAC CTC AGA AGA TGA AC-3Ј. The amplified product was subcloned into the T-Easy cloning vector (Promega, Madison, WI), and sequenced to identify insert orientation antisense to the T7 promoter. Radiolabeled [ 35 S] antisense riboprobe synthesis and in situ hybridization were carried out as described previously (16).

RESULTS
Feeding high fat diets to LDLR Ϫ/Ϫ mice caused hyperlipidemia (Table I). This effect was apparent after only 2 weeks of dietary intervention, reflected a greater elevation of cholesterol than triglycerides, and was more pronounced with the Western diet (containing 0.15% cholesterol). Compared with chow-fed mice, serum cholesterol was 62% higher in high fat-fed mice (p ϭ 0.0011) and 286% higher in Western-fed mice (p Ͻ 0.0001) after 17 weeks.
Diet-induced hyperlipidemia in these mice was characterized by an elevation of atherogenic lipoproteins ( Fig. 1). Consistent with relative effects on total serum cholesterol (Table I), increases in VLDL/LDL cholesterol fractions were greater in Western-fed than high fat-fed animals.
Higher levels of serum cholesterol were associated with more aortic atheroma formation (Fig. 2). After 17 weeks, Western diet-fed mice showed considerable involvement of the intima with atheromatous lesions that mostly affected the arch. Only 2 of 10 mice on the high fat diet had lesions, both detected in the arch. For these data, the Kruskal-Wallis p value was 0.0025. Lesion area was significantly greater in Western dietfed mice compared with chow-fed (p Ͻ 0.05) and high fat diet-fed mice (p Ͻ 0.01 by Dunn's multiple comparisons test).
Although the effects of high fat diets on hyperlipidemia and atheroma formation were disparate and cholesterol content-dependent, both high fat diets produced similar degrees of hyperglycemia, hyperinsulinemia, and obesity (Table II). High fat feeding for only 2 weeks, before the development of obesity, produces insulin resistance in rats (17). High fat diets have the same effect on LDLR Ϫ/Ϫ mice in the C57BL/6J background (Table II). At 2 weeks, body weights were not significantly different but fasting glucose and insulin levels were higher in animals eating the experimental diets. By 5 weeks, weights were significantly higher in animals on both high fat diets. By 17 weeks, animals on both the high fat and Western diets were hyperinsulinemic, hyperglycemic, and obese compared with chow-fed mice. C57BL/6J mice are obesity and diabetes prone (18).
Calcification of human vasculature can occur independent of intimal atheroma formation (19). We used von Kossa staining to determine whether aortic calcification occurs in diabetic LDLR Ϫ/Ϫ mice and if such mineralization is atheroma-dependent. Mineralization was observed in calcifying tracheal cartilage (Fig. 3A, arrow to right of panel) from mice fed the high fat (no added cholesterol) diet for 5 weeks. Surprisingly, mineral deposition was also visualized in the adventitia immediately surrounding the coronary arteries near the ascending aorta (arrow to left of panel A with enlarged view of the same region shown in panel B) in animals fed the high fat diet. This diet also promoted mineral deposition on aortic valve leaflets (Fig. 3C). Aortic mineralization was consistently observed in animals on the high fat (no added cholesterol, minimal atheroma formation) and Western (0.15% cholesterol, extensive ath- eroma formation) diet (data not shown), but not in animals eating chow (Fig. 3, D and E) even though calcifying tracheal cartilage was visualized within the same specimen (not shown). Thus, diabetogenic diets promote mineral deposition in periaortic adventitia and valves of LDLR Ϫ/Ϫ mice independent of the extent of atheroma formation.
Mineral deposition can be passive or active. To determine whether diabetogenic diets cause active deposition, we assayed expression patterns of genes encoding three known transcriptional regulators of orthotopic tissue mineralization and osse-ous differentiation, the homeodomain proteins, Msx2 and Msx1, and the runt domain protein, Cbfa1/Osf2. High fat feeding with or without cholesterol gave rise to higher levels of aortic Msx2 and Msx1 mRNA as compared with chow (Fig. 4A). Expression of Cbfa1 mRNA was not observed in any aortic specimen, even though the positive control (from MC3T3E1 calvarial osteoblasts) was readily detected. Opn message levels were also higher in aortas of animals fed high fat diets with or without cholesterol supplementation (Fig. 4A).
A similar pattern of gene expression was observed after 17 , or Western (42% fat, 0.15% cholesterol) diets, mice were sacrificed. Vessels were fixed with paraformaldehyde, and the entire aorta was pinned en face. The percentage of the aortic intimal surface area affected by atheroma was determined by image processing. Aortas from animals eating chow (n ϭ 3), high fat (n ϭ 10), and Western (n ϭ 9) diets, were assayed at each of three sites, the aortic arch (extending from the aortic valve to a point 4-mm distal to the left subclavian artery), thoracic aorta (from the end of the arch to the final intercostal artery), and abdominal (Abd) aorta (from the final intercostal artery to the ileal bifurcation). weeks (Fig. 4B). Msx1 expression was not affected by fat feeding at this time point, but both Msx2 and Opn message levels were still up-regulated by high fat diets. Induction of Msx2 mRNA accumulation by high fat diets is detectable as early as 2 weeks (Fig. 4C) when animals already manifest hyperglycemia and hyperinsulinemia (Table II). Thus, at least a component of diet-induced aortic calcification reflects active regulation of transcription factors that control skeletal mineralization programs. High fat diabetogenic diets consistently induce aor-tic expression of Msx2, a skeletal transcription factor that controls neurectodermal mineralization and gene expression in the developing skull.
To identify the responsible cell types, Msx2 expression in aortic cross-sections of fat-fed LDLR Ϫ/Ϫ mice was analyzed by in situ hybridization. As shown in Fig. 5, A, C, and E, Msx2 expression was detected in a subset of peri-aortic adventitial cells and cells at the medial-adventitial junction, and most readily observed adjacent to the coronary arteries and the  epicardium. Analysis of adjacent sections revealed that signals arose from cells within adventitial adipose tissue (Fig. 5, B and  D). The von Kossa stain of an adjacent section demonstrated mineral accumulation (Fig. 5D). Smooth muscle cell ␣-actin expression (Fig. 5F) partially overlapped Msx2 (Fig. 5E) and Opn (Fig. 5G) signals, suggesting that adventitial vascular smooth muscle cells such as the pericyte participate in this response. Little if any Msx2 expression was detected in myocardium, valvular tissue, tunica media (Fig. 5, A and E) or atheroma (not shown). Osteocalcin mRNA was not detected by in situ hybridization in these aortic regions (Fig. 5H) as predicted by elevated levels of Msx2 expression (Fig. 4; Refs. 16 and 20). Data in Fig. 5 were obtained from animals fed the high fat (no added cholesterol) diet for 5 weeks; the same expression patterns were obtained using animals fed the Western (0.15% cholesterol) diet (not shown). Thus, Msx2 is expressed in periaortic adventitial cells in the aortas of LDLR Ϫ/Ϫ mice fed high fat diabetogenic diets.
Opn is a multifunctional matrix protein that binds calcium and mediates ␣ v ␤ 3 integrin-dependent adhesion (21,22). It is expressed by osteoblasts and odontoblasts during normal development, but also by cells comprising the atheroma, including activated macrophages and vascular smooth muscle cells (23). Because Opn induction by the high fat (no added cholesterol) diet (Fig. 4) occurs independent of atheroma formation (Fig. 2), expression reflects in part activation of the osteogenic program. To determine whether Opn expression in response to the Western (0.15% cholesterol) diet reflects both activation of the osteogenic program and foam cell activity within the atheroma, aortic Opn expression was compared by in situ hybridization for mice fed the high fat (Fig. 6, A and B) and Western (Fig. 6, C and D) diets. Light field photomicrography was used for these comparisons, because the Opn signal was very intense. As shown in Fig. 6, A and B, a subset of adventitial cells expressed Opn, similar to analyses of Msx2 expression (Fig. 5). Unlike Msx2, Opn was expressed by smooth muscle cells of the vessel wall and foam cells of the atheroma (Fig. 6, C and D). Thus, diet induction of Opn reflects contributions from both peri-aortic adventitial cells and cells that form the intimal atheroma.

DISCUSSION
Stiff blood vessels are found in otherwise healthy humans with well defined metabolic disorders such as familial hypercholesterolemia (caused by LDLR deficiency) and diabetes (24 -26). Hyperlipidemia and hyperglycemia impair endothelial vasodilation, but calcified vessels are also unlikely to dilate in response to stress. So it is not surprising that arterial calcification predisposes type II diabetics to premature death (7).
The mechanisms underlying arterial calcification are not well understood. Most data come from surgical specimens representing end-stage disease. Our work establishes a murine model of the initiation of vascular calcification induced by diabetogenic diets. High fat feeding of LDLR Ϫ/Ϫ mice induces expression of Msx2 and Msx1 (transcription factors controlling neurectodermal osteoblast differentiation), Opn (a secreted protein characteristic of the mineralizing osteoblast), and causes aortic mineralization. This pattern is independent of accumulation in aortas from mice fed chow, high fat (no added cholesterol), and Western (0.15% cholesterol) diets for 2 weeks. Again, the mRNA for Msx2 accumulates to relatively higher levels in the aortas of mice fed the high fat diets; even at this early time point both diabetes (Table I)  Adult LDLR Ϫ/Ϫ mice were fed experimental diets for 2 to 17 weeks. Total RNA from the proximal aorta was prepared from at least five animals in each dietary group, combined, and mRNA accumulation for specific genes assessed by RT-PCR as described under "Experimental Procedures." Panel A shows relative mRNA accumulation in aortas from mice fed chow, a high fat diet containing 1.25% cholesterol and 0.5% cholate (Chol. ϩ Fat), and the high fat (58% fat, no added cholesterol) diet for 5 weeks. Note that the mRNAs for Msx2, Msx1, and osteopontin (Opn) accumulate to relatively higher levels in the aortas of mice maintained on high fat diets compared with mice maintained on mouse chow. Further, note there is no apparent difference in the accumulation of mRNA for glyceraldehyde phosphate dehydrogenase (GAPD) in these same specimens, and that mRNA for Cbfa1/ Osf2 is not detected, even though a positive control (the single lane just to the left of the mouse dietary intervention samples) is readily visualized. The positive control was derived from MC3T3E1 calvarial osteoblasts. Panel B shows relative mRNA accumulation in aortas from mice fed chow, the high fat (58% fat, no added cholesterol), and the Western (42% fat, 0.15% cholesterol) diet for 17 weeks. Note that the mRNAs for Msx2 and osteopontin (Opn) accumulate to relatively higher levels in the aortas of mice maintained on the high fat diets, whereas no apparent differences exist for Msx1 and glyceraldehyde phosphate dehydrogenase accumulation at this time point. Panel C shows relative mRNA FIG. 5. Msx2 is expressed in peri-aortic adventitial cells in LDLR ؊/؊ mice fed high fat diabetogenic diets. LDLR Ϫ/Ϫ mice were fed mouse chow or high fat diabetogenic diets for 5 weeks. Aortic tissues were dissected, fixed, embedded, sectioned, and analyzed for Msx2, osteopontin (Opn), osteocalcin (Osc) and vascular smooth muscle cell ␣-actin (␣-Act) expression by radioactive in situ hybridization as described under "Experimental Procedures." Specimens were viewed under dark field illumination for photomicrographic visualization of hybridization signals. Little or no signal for these mRNAs is observed in tissues from chow-fed animals (data not shown). However, gene expression was readily detected in aortic tissues from mice fed the high fat diabetogenic diet (Panels A-H). Panel A shows the Msx2 expression pattern. Note the intense expression of Msx2 in peri-aortic adventitia immediately adjacent to a coronary artery (arrows). Panel B shows light field hematoxylin/eosin histology of the same region. Panel C shows a higher power magnification of signal in Panel A between the two arrows. Note again the signal arising from peri-aortic adventitial cells, and the absence of signal from vascular smooth muscle of the tunica media. Panel D shows light field histology of an adjacent 5 section, stained with the von Kossa technique and nuclear fast red. Note the significant adipose content of this adventitial region. A small amount of regional adventitial mineral deposition is seen on the right side of the panel. Panels E, F, and G show relative expression patterns for Msx2, vascular smooth muscle cell ␣-actin, and Opn, respectively. Note that the adventitial pattern of vascular smooth muscle cell ␣-actin (panel F) overlaps that of Msx2 (panel E), and Opn (panel G), suggesting that adventitial vascular smooth muscle cells are recruited to express the cholesterol supplementation and atheroma formation. These data suggest that diet-induced diabetes causes vascular calcification through an active process, the initiation of an osteoblast transcriptional regulatory program.
Demer and colleagues (27) provided molecular evidence that vascular calcification is in part an active process. They demonstrated that cells within human aortic plaques express BMP2, a transforming growth factor-␤ superfamily member that can direct heterotopic ossification in susceptible cell types. Fitzpatrick and colleagues (28) detected osteopontin and osteocalcin expression in end-stage calcified native and bioprosthetic valves, again suggesting an active process.
One novel feature of our work is the induction of Msx2 expression in the proximal aorta by insulin-resistant diabetes. Transcripts for Msx2, a homeobox gene important for normal embryogenesis, have been detected during development in structures that give rise to the aorta (29). Msx2 is expressed at embryonic day 14.5 in mesenchymal cells of the newly formed aorta but decreases to undetectable levels by parturition. 2 Proximal aorta mesenchymal cells are of neural crest origin. Neural crest-derived populations are plastic, retaining the potential for differentiation along different pathways depending on cues from their environment (30). We speculate that a dietinduced metabolic signal acts on mesenchymal cell progenitors in the proximal aorta to activate an Msx2-driven tissue mineralization pathway. Candidate signals include hyperlipidemia, hyperinsulinemia, and hyperglycemia. Hyperglycemia can induce expression of transforming growth factor-␤, a growth factor known to commit neural crest cells to certain lineages (31).  6. Opn is expressed in perivascular adventitial cells, vascular smooth muscle cells, and atheroma foam cells. LDLR Ϫ/Ϫ mice were fed the high fat (no added cholesterol) or Western (0.15% cholesterol) diet for 5 weeks. Aortic tissues were dissected, fixed, embedded, sectioned, and analyzed for Opn expression by radioactive in situ hybridization as described under "Experimental Procedures." Because the intense Opn signal permits light field visualization of both silver grains and subjacent cell type, light field photomicrography is presented. Panel A shows Opn expression in the aorta of a mouse fed the high fat (no added cholesterol) diet for 5 weeks. Opn expression was detected in peri-aortic adventitia (arrowhead) and was particularly intense in the tunica media at valve leaflet insertions (arrows). Note the absence of atheroma in animals fed the high fat (no added cholesterol) diet, as shown in Fig. 2. Panel B shows a higher magnification of the peri-aortic adventitia of Panel A. Panel C shows Opn expression in the aorta of a mouse fed the Western (0.15% cholesterol) diet for 5 weeks. Note Opn expression in the cholesterol-induced atheroma (lower arrow). The upper arrow shows signal at a valve leaflet insertion. Panel D is a higher magnification of panel C. See text for details.
The pericyte appears to be one of the mesenchymal cell progenitors responsible for the mineralization pathway. Pericytes, vascular smooth muscle cells supporting the microvasculature, can differentiate into multiple mesenchymal cell types including osteoblasts (32). Bovine retinal pericytes can form mineralizing bone nodules and express osteoblast genes when grown in culture (33). Retinal pericytes will form bone, cartilage, and fat when seeded into subcutaneous diffusion chambers (33). We recently identified Msx2 and Opn gene expression in an aortic pericyte cell line. 2 Another novel feature of our work is the adventitia-based rather than tunica media-based pattern of Msx2 and Opn expression. Prevailing concepts of the pathophysiology of vascular calcification emphasize the role of medial vascular smooth muscle cells. This idea was mostly derived from studies using (i) vitamin D to promote hypercalcemia and calciphylaxis (34), (ii) mechanical injury (35), or (iii) genetic lesions that remove natural arterial mineralization inhibitors (36,37). The latter models produce diffuse vascular calcification in immature mice and do not require dietary manipulation.
In our model, high fat diets promote vascular calcification and induce metabolic insults characteristic of type II diabetes and mixed dyslipidemia; these insults activate an osteogenic regulatory program in adventitial cells near the adventitialmedial junction. Of note, surgical removal of the adventitia in rodents decreases diet-induced vascular calcification of the media (38), suggesting that adventitial pericytes migrate into the aortic vessel wall and participate in disease progression. Taken together with our data demonstrating molecular fingerprints of early osteoblast gene transcription within the adventitia, these findings strongly support a model in which metabolic insults such as diabetes and hyperlipidemia promote vascular calcification in part by recruiting adventitial cells with osteogenic potential. The small amount of calcium deposition detected at an early stage of vascular disease (5 weeks; see Figs. 3 and 5) probably does not have physiological consequences, but the effects of chronic, diffuse calcium deposition on murine vascular function are likely to be severe. Future studies will test these notions directly and pursue the specific metabolic signals that mediate osteoprogenitor activation.