Targeted Reduction of Vascular Msx1 and Msx2 Mitigates Arteriosclerotic Calcification and Aortic Stiffness in LDLR-Deficient Mice Fed Diabetogenic Diets

Vascular calcification increasingly afflicts the aging population, driven by the dysmetabolic milieus of diabetes, dyslipidemia, and uremia (1–4). In type 2 diabetes, arterial vascular calcification primarily occurs within the tunica media, with contributions from atherosclerotic plaques as they accrue. Both medial and atherosclerotic calcification increase vascular stiffness, thus impairing Windkessel physiology, the elasticity of conduit vessels that enables smooth distal tissue perfusion throughout the cardiac cycle (3,5). During systole, compliant conduit vessels capture kinetic energy as potential energy then release this energy during diastole as a mechanism that sustains perfusion pressure with cardiac relaxation (5). When conduit vessels lose elasticity, cardiac afterload and myocardial oxygen consumption are increased, tissue perfusion becomes increasingly pulsatile in distal vascular beds, and risks for end-organ barotrauma and ischemia are increased during systole and diastole, respectively (5). The net consequence is increased cardiovascular morbidity and mortality through stroke, myocardial infarction, congestive heart failure, and lower-extremity amputation (1,2).

When fed high-fat diets, male LDLR^−/− mice develop obesity, hyperlipidemia, insulin-resistant diabetes, and arterial calcification (6–9). These responses phenocopy the arteriosclerotic pathobiology observed in patients with diabetes (1). Msx1 and Msx2 are homeodomain transcription factors indispensible for craniofacial bone formation (10) and cardiac valve morphogenesis (11,12), and the osteogenic Msx gene regulatory program is concomitantly upregulated in calcifying arteries of diabetic mice and humans with diabetes, dyslipidemia, and/or uremia-induced vascular disease (1,2,13). Thus, mineralization programs regulated by Msx genes are activated during vascular calcium accrual in the setting of type 2 diabetes.

In previous studies, we demonstrated that augmenting aortic Msx2 gene expression through either tumor necrosis factor–dependent proinflammatory signals or direct Msx2 transgenic strategies worsens arterial calcification (8,9,14). We addressed whether reducing aortic Msx gene expression mitigates arteriosclerotic calcification to provide additional proof for the role of this gene regulatory pathway in arterial disease biology. Implementing Cre-lox technology and the SM22-Cre transgenic mouse (15), we show that targeted reductions in vascular smooth muscle Msx2 and Msx1 (10,16,17) reduce arterial calcification and improve arterial compliance in diabetic LDLR^−/− mice, with concomitant reductions in the mineralizing potential of vascular osteoprogenitors.

Molecular, biochemical, genotyping, and histological methods have been previously detailed (7–9,18). The indicated, inventoried TaqMan Gene Expression assays were purchased from Life Technologies for quantifying mRNA accumulation by real-time fluorescence quantitative RT-PCR (qRT-PCR). Genotyping primers were ordered from Life Technologies. Amplimer pairs were as follows: SM22-Cre, 5′-CAG ACA CCG AAG CTA CTC TCC TTC C-3′ and 5′-CGC ATA ACC AGT GAA ACA GCA TTG C-3′ (500 bp); Msx1, 5′-ACA CTA TGC TTG ATG TGG TCC CAG GCG-3′ and 5′-GGG CTC GGC CAA TCA AAT TAG AGA G-3′ (wild type [WT] = 165 bp; flox = 233 bp); Msx2, 5′-GTT TCA TGA CCT CAT TAC TCA CGC TG-3′ and 5′-GGT ACC TTT GTC AAA TCT GTG AG-3′ (WT = 158 bp; flox = 226 bp); and LDLR^−/−, 5′-ACC CCA AGA CGT GCT CCC AGG ATG-3′ and 5′-CGC AGT GCT CCT CAT CTG ACT TGT C-3′ for the genomic site of insertion (intact WT = 383 bp) and 5′-AGG ATC TCG TCG TGA CCC ATG GCG A-3′ and 5′-GAG CGG CGA TAC CGT AAA GCA CGA GG-3′ for neomycin (200 bp). ELISA kits quantifying matrix metabolism markers desmosine (CSB-E14196M; American Research Products) and type I collagen propeptide P1NP (AC-33F1; IDS Inc.) were purchased from commercial sources. Lipofectamine RNAiMAX, ON-TARGETplus control and SMARTpool small interfering RNAs (siRNAs) targeting Msx1 and Msx2 were purchased from Life Technologies. GSK3β/phospho-GSK3β antibodies were from Cell Signaling Technologies (8,9,14).

Procedures for handling mice were approved by the Washington University and Sanford-Burnham institutional animal care and use committees. LDLR^−/−B6.129S7-Ldlrtm1Her/J (19) and SM22-CreTg(Tagln-cre)1Her/J (15) mice were obtained from The Jackson Laboratory. Msx1(fl/fl);Msx2(fl/fl) mice have been described (20) and were bred onto the LDLR^−/− background. Experimental SM22-Cre;Msx1(fl/fl);Msx2(fl/fl);LDLR^−/−, SM22-Cre;Msx2(fl/fl);LDLR^−/−, and Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− control animals were obtained through the breeding scheme outlined in Fig. 1. At 5–10 weeks of age, animals were weighed. Male sibling cohorts (n = 4–13 per genotype as indicated) of Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− and SM22-Cre;Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− mice with equivalent starting weights were challenged with high-fat western diet (HFD) (TD88137; Teklad Lab Animal Diets; Harlan) for 2 months. At the end of the dietary challenge, thoracic aortas were harvested, weighed, and extracted for calcium, collagen, or total RNA, implementing methods previously detailed (9,21).

Cohorts of male Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− and SM22-Cre;Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− mice were challenged with HFD for 3 months. Aortic arch pulse wave velocity (PWV) was determined as described (21) using a modification of the transit time method (22), implementing a Vevo 770 ultrasound system with a 30-MHz transducer (VisualSonics Inc., Toronto, ON, Canada). Under isofluorane anesthesia, the ascending aorta, aortic arch, and proximal portion of the descending aorta were imaged in one two-dimensional imaging plane from the right-side superior parasternal view. The pulse wave Doppler sample volume was placed first near the aortic valve to record blood flow velocity in the proximal aorta then promptly moved to the visualized portion of the descending aorta without changing the imaging plane to record blood flow velocity in the descending aorta. The curvilinear distance between the proximal and distal points of the aortic velocity interrogation (D2–D1 in millimeters) was measured using the exact coordinates of the Doppler sample volumes. The time delay between the onset of flow velocity in the distal and proximal portions of the aorta (T2–T1 in milliseconds) was measured relative to the simultaneously recorded electrocardiogram signal. PWV was calculated as the ratio of D2–D1 to T2–T1 expressed as meters per second.

Aortic adventitial myofibroblasts were prepared using a modification of published methods (14,23). Aortas were isolated from Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− and SM22-Cre;Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− mice by dissection from diaphragm to aortic outflow and processed, and cell isolates were pooled (8–10 male mice per genotype). After rinsing twice in PBS supplemented with penicillin-streptomycin-fungizone (PSF) 200 Units/mL, 200 μg/mL, 2.5 μg/mL with intervening blotting on Kimwipes to remove blood, each aorta was rinsed three times in fresh DMEM supplemented with PSF. Individual aortas were then minced into four fragments. Two aortas per genotype were placed in a 50-mL conical screw cap tube and then digested at 37°C in a 5% CO₂ incubator in 4 mL of type I collagenase 1 mg/mL (catalog number LS004149; Worthington Biochemical Corporation), DNase I 60 Units/mL (D5025; Sigma-Aldrich), and hyaluronidase 0.5 mg/mL (H3506; Sigma-Aldrich) in DMEM with PSF using a sterile flea stir bar to provide gentle agitation. Two sequential 1-h digestions were performed and combined, and cells were pelleted at 1,500 rpm for 5 min. The cell pellet was resuspended in DMEM with 10% FBS supplemented with PSF and then plated onto 10-cm tissue culture plates coated with rat tail type I collagen 6 μg/cm². After 3 days, one-half of the medium was removed from each culture and replaced with fresh growth medium. Two days later, cells were rinsed once with growth medium lacking PSF and then maintained in growth media with penicillin-streptomycin 100 International Units/mL, 100 μg/mL and fed every 3 days until confluence. Only passage 1 and passage 2 cultures were used for experimentation. The methods for preparation of SFG-LacZ and SFG-Wnt7b retroviruses and myofibroblast transduction have been described (18). RNA interference (RNAi) in primary adventitial cultures was performed in 12-well cluster dishes (100,000 cells/well) in triplicate with 100 nmol/L nontargeting control siRNA (NTC), 50 nmol/L Msx1 siRNA + 50 nmol/L NTC, 50 nmol/L Msx2 siRNA + 50 nmol/L NTC, or 50 nmol/L Msx1 + 50 nmol/L Msx2 siRNA using RNAiMAX, following the manufacturer’s protocol. Two days later, RNA was isolated for qRT-PCR using previously detailed methods (18).

Alkaline phosphatase–positive cells in aortic adventitial cell cultures were visualized as previously described, but using the Vector Red fluorescent substrate (14,18). Briefly, aortic mesenchymal cells were cultured on a type I collagen–precoated four-chamber slide (60,000 cells/chamber) for 8 days. During the final 6 days, cells were treated with β-glycerophosphate 5 mmol/L and ascorbic acid 50 μg/mL. Cells were then washed three times with Tris-buffered saline (TBS) (20 mmol/L Tris HCl, 1.5 mol/L NaCl, pH 7.5), fixed with 4% paraformaldehyde in TBS for 4 min, and rinsed three times with TBS. Staining with Vector Red reagent (Vector Red Alkaline Phosphatase Substrate Kit I, catalog number SK-5100; Vector Laboratories) was performed for 1 h in the dark per the manufacturer’s protocol. After washing three times with TBS and twice with distilled water, nuclei were stained with DAPI (Prolong Gold Antifade Reagent, catalog number P36931; Molecular Probes) and the slides coverslipped. Photomicrographs of alkaline phosphatase–positive fluorescent cells (TX2 filter cube), nuclei (DAPI filter cube), and phase contrast images were captured by a Leica DM4000 B fluorescence microscope at ×200 magnification.

Aortic adventitial cells were seeded in type I collagen–coated 24-well cluster plates (15,000 cells/well, three wells per genotype). The next day and every other day thereafter, cells were fed with DMEM containing 10% FBS, ascorbic acid 50 μg/mL, and β-glycerophosphate 5 mmol/L for a total of 12 days. Mineralized matrix was detected by Alizarin Red S staining as described (14,18), and images (12–16 images/well) were captured by a Nikon Eclipse Ti microscope at ×100 magnification. Alizarin red–stained areas were quantified using National Institutes of Health ImageJ software as previously detailed (14).

All experiments were performed with 3–12 independent replicates per group. Statistical analyses were performed using GraphPad InStat version 3.06 software, implementing standard parametric or nonparametric methods (two-tailed testing) when indicated. Graphic data are presented as mean ± SEM.

Msx1 and Msx2 are highly related members of the NK-like homeodomain transcription factor family (11,12). Msx1 and Msx2 play quantitatively distinct, but functionally redundant roles in neural crest biology, craniofacial and heart valve morphogenesis, and skeletal and ectodermal organ growth at sites where dynamic epithelial-mesenchymal interactions occur (11,12). Our previous studies demonstrated that an HFD upregulates Msx1 and Msx2 in the aortic myofibroblasts and smooth muscle cells in male LDLR^−/− mice (6,23), presaging subsequent cardiovascular calcification and the expression of several osteogenic Wnt ligands, including Wnt3, Wnt5, and Wnt7 family members (7). Moreover, transgenic augmentation of aortic Msx2 expression promoted calcification in the tunica media in mice fed HFD, with mineralization proceeding through activation of an osteogenic Wnt cascade (7). To better understand the role of endogenous Msx1 and Msx2 in vascular disease processes, including arteriosclerotic calcification, we implemented Cre-Lox technology to reduce aortic Msx1 and Msx2 tone in genetically engineered mice possessing floxed (fl) Msx alleles (20). The SM22-Cre transgene was used to deliver Cre recombinase to vascular smooth muscle cells (VSMCs) and aortic myofibroblasts (15). Using the breeding scheme outlined in Fig. 1, SM22-Cre;Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− and Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− mice were generated and male sibling cohorts studied. Msx1 and Msx2 mRNAs were significantly reduced in total aortic RNA from SM22-Cre;Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− mice versus Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− controls, with concomitant reductions in osterix (Osx), a zinc finger transcription factor upregulated by Msx2 (23,24) and necessary for osteogenic mineralization (25) (Fig. 2A). Although the abundance of Runx1, Runx2, and Sox9 messages were unaltered in total aortic RNA, expression of Sonic hedgehog (Shh), a marker of the vascular multipotent mesenchymal progenitor (26), was concomitantly reduced with vascular Msx gene deficiency. We next challenged sibling male cohorts with the diabetogenic HFD for 2 months and assessed the impact on aortic calcium accumulation. Compared with Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− controls, SM22-Cre;Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− mice exhibited a significant 31% reduction in total aortic calcium accrual (Fig. 2B). Alizarin red staining for aortic calcium confirmed that changes in mineral deposition largely localized within the tunica media (Fig. 2C and D). By contrast, total aortic collagen accumulation (Fig. 2E) and ascending aorta wall thickness (Fig. 2F) were unaltered. SM22-Cre;Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− mice were significantly more obese than Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− controls (Fig. 3A), but bone mineral density and bone mineral content by DEXA were also unaltered (Fig. 3B). No significant differences were observed with respect to diet-induced changes in fasting plasma glucose, triglyceride, or cholesterol levels (Fig. 3C). Furthermore, plasma haptoglobin and 8-F-isoprostanes, markers of systemic inflammation and oxidative stress upregulated by diabetogenic HFD (8), did not differ with vascular Msx gene deletion (Fig. 3D), indicating that genetic manipulation of vascular Msx genes had not mitigated the proinflammatory metabolic milieu. Thus, aortic Msx1 and Msx2 expression in aortic VSMCs and myofibroblasts supports arteriosclerotic calcification in LDLR^−/− mice fed a diabetogenic HFD.

Arterial calcification is a harbinger of vascular stiffness, a composite of material properties and geometric properties that alter the distensibility of conduit vessels during the cardiac cycle (3). From the Moens-Korteweg relationship (27), the product of the elastic modulus and the arterial wall thickness-to-lumen diameter ratio is proportional to the square of the conduit vessel PWV; thus, PWV is an index of vessel stiffness arising from contributions of vessel material properties (e.g., calcification, fibrosis) and geometric properties (wall thickness, diameter) (3). We therefore assessed aortic arch PWV in SM22-Cre;Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− mice and Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− controls after 3 months of HFD, implementing Doppler echocardiography as a noninvasive and clinically useful method to assess aortic stiffness (22). As shown in Fig. 4A, aortic arch PWV was significantly reduced by 30% in SM22-Cre;Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− mice (P < 0.05 vs. controls), indicating reduced vascular stiffness. Importantly, aortic diameters did not differ between Cre-minus and Cre-plus cohorts at the ascending aorta (1.3 ± 0.05 vs. 1.4 ± 0.05 mm), aortic arch (1.1 ± 0.05 vs. 1.2 ± 0.03 mm), or descending aorta (0.9 ± 0.02 vs. 1.0 ± 0.04 mm; all P not significant) (Fig. 4B). Plasma markers of matrix remodeling and synthesis, namely the elastin degradation product desmosine (28) and type I procollagen N-telopeptide (9), did not significantly differ between Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− and SM22-Cre;Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− animals (Supplementary Fig. 1). Thus, reducing arterial Msx1 and Msx2 expression mitigates vascular stiffness in a murine model of diet-induced diabetic arteriosclerosis. Reductions in vascular calcification, but not changes in vascular diameter, wall thickness, or fibrosis, accompany reductions in vascular stiffness.

Reductions in arteriosclerotic calcification and vascular stiffness with reduced Shh expression suggested that key components of vascular mesenchymal progenitor differentiation were perturbed in aortic myofibroblasts from SM22-Cre;Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− mice. Therefore, we studied expression of transcriptional markers of early osteogenic, adipogenic, chondrogenic, and myogenic differentiation in aortic adventitial myofibroblasts stimulated with a dexamethasone cocktail used to enhance mesenchymal cell differentiation from progenitors (29). As shown in Fig. 5A, primary aortic myofibroblasts generated from male SM22-Cre;Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− mice exhibit deficiencies in Dlx5, PPARG, Sox9, and myocardin (Myocd) gene expression with concomitant Cre-dependent reductions in Msx1, and Msx2. An impact at an early stage of osteogenic differentiation is likely because levels of BSP (bone sialoprotein) and OCN (osteocalcin), the latter a gene of mature osteoblasts directly inhibited by Msx2 (30), were both reduced in SM22-Cre;Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− cultures (Fig. 5B). Consistent with this notion, Shh and Sca1, markers of adventitial osteoprogenitors (26), were concomitantly diminished with combined Msx1+Msx2 deficiency (Fig. 5B). Expression of leptin and OPN (osteopontin) were unaffected (Fig. 5B).

Increasing Msx2 tone in adventitial myofibroblasts by either viral transduction or transgene expression upregulates a cohort of genes encoding Wnt ligands (7). Therefore, we examined the impact of Msx1+Msx2 deficiency on the expression of all 19 Wnt ligands. As shown in Fig. 5C and D, the relative expression of Wnt1, Wnt2, Wnt2b, Wnt3, Wnt4, Wnt5a, Wnt5b, Wnt7a, Wnt7b, Wnt8b, Wnt9a, and Wnt11 was reduced by ≥50% in SM22-Cre;Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− cultures. Wnt10a and Wnt10b were unaffected. Axin2, a genomic target entrained to canonical Wnt/β-catenin signaling in multiple cell types (31), was also downregulated (Fig. 5D). GSK3β phosphorylation, another index of activated Wnt signaling (32), was concomitantly reduced in SM22-Cre;Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− cultures as well (Fig. 5E and F). Thus, combined Msx1 and Msx2 deficiency in SM22-Cre;Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− aortic myofibroblasts reduces elaboration of differentiated mesenchymal cell transcripts, including the expression of osteogenic and Wnt gene regulatory programs.

We and others previously demonstrated that aortic adventitial myofibroblasts and calcifying vascular cells residing within the tunica media exhibit osteogenic, myogenic, and limited adipogenic potential when cultured (23,26,33). Moreover, transduction with a retrovirus driving expression of Msx2 promotes osteogenesis in aortic myofibroblasts (23). However, the consequences of depleting endogenous Msx signaling upon vascular mesenchymal osteogenic differentiation are unknown. Therefore, we prepared primary aortic myofibroblasts from male SM22-Cre;Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− and Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− sibling controls, cultured them under mineralizing conditions, and quantified alkaline phosphatase and mineralizing Alizarin red osteogenic colonies as previously described (14,23). As shown in Fig. 6A and B, the numbers of alkaline phosphatase–positive adventitial myofibroblasts decreased by 78% in aortas of Msx-deficient mice compared with controls. Moreover, through Alizarin red staining quantifying calcified colonies, mineralization was shown to be concomitantly decreased by 43% (Fig. 6C and D). Reduced calcification was partly due to diminished Wnt gene bioactivity because transduction of SM22-Cre;Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− cells with a retrovirus driving Wnt7b expression restored Alizarin red staining (Supplementary Fig. 2). Thus, the combined Msx1 and Msx2 deficiency in SM22-Cre;Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− aortas reduces the osteogenic mineralization potential of vascular myofibroblasts.

In a separate experimental cohort, conditional deletion of the Msx2 gene alone was insufficient to reduce aortic calcium accumulation in response to an HFD challenge (1.2 ± 0.2 vs. 1.1 ± 0.2 μg calcium/g aorta in Msx2(fl/fl);LDLR^−/− and SM22-Cre;Msx2(fl/fl);LDLR^−/− mice, respectively; n = 9–11 per genotype; P not significant), indicating some functional redundancy. To better understand the individual contributions of Msx1 and Msx2 to adventitial myofibroblast biology, we implemented RNAi to reduce Msx1, Msx2, or Msx1+Msx2 expression in primary cell cultures, quantifying the impact on osteogenic lineage programming, TNAP (34), and cardiogenic (32) Wnt (Wnt2, Wnt5a, Wnt7b, Wnt11) gene expression. As shown in Fig. 7A, RNAi selectively and specifically reduced Msx gene expression in transfected aortic adventitial myofibroblast cultures. TNAP and Wnt7b messages were downregulated by siRNA targeting either Msx1 or Msx2. By contrast, Wnt5a and Wnt2 expression were downregulated primarily in response to Msx2 depletion. Unlike cultures from SM22-Cre;Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− mice, Wnt11 expression was not acutely altered by RNAi targeting Msx1+Msx2. Passman et al. (26) previously identified an Shh-, Sca1-, and Msx1-expressing multipotent adventitial cell population capable of osteogenic differentiation. We therefore assessed the impact of Msx1 and Msx2 RNAi on expression of these specific adventitial progenitor markers, further encompassing CD90+CD105 as general markers of mesenchymal stem cell populations (26), and Sox9 as a marker of the osteochondroprogenitor (35). As shown in Fig. 7B, siRNA targeting Msx1 selectively reduced Shh expression. By contrast, siRNA targeting Msx2 preferentially downregulated Sca1 and Sox9, whereas CD90 and CD105 were unaffected. Thus, Msx1 and Msx2 program aortic adventitial osteoprogenitors through overlapping, yet distinct mechanisms that control Wnt and Shh gene expression.

Msx1 and Msx2 play fundamental roles in signaling cascades controlling craniofacial morphogenesis and epithelial-mesenchymal interactions driving odontogenesis, cardiac cushion formation, and breast development (10–12). Msx2 is also upregulated in aortic valve and vascular myofibroblasts during the progression of diet-induced diabetes in preclinical models of vascular calcification (6,7). Atherogenic stimuli such as tumor necrosis factor, interleukin-1β, and lipid products derived from oxidized LDL all upregulate Msx2 in vascular mesenchymal cells (8,14,36). Furthermore, transgenic overexpression of Msx2 in the vessel wall promotes vascular calcification and activation of Wnt-dependent osteogenic gene regulatory programs (7). Importantly, Msx2 is elaborated in sclerotic vessel (1,2) and valve (13) segments in humans with arteriosclerosis. However, the contributions of endogenous Msx genes to cardiovascular disease during postnatal life, including vascular calcification, have not been previously examined. Because of the functionally redundant roles of Msx1 and Msx2 in vital developmental processes (10–12,17), we addressed the impact of combined Msx1 and Msx2 vascular deficiency on arteriosclerotic disease in LDLR^−/− mice.

Consistent with the current results demonstrating the activation of aortic osteogenic mineralization in transgenic mice overexpressing Msx2 (7), reductions in vascular Msx tone elicit reciprocal changes; aortic calcium accrual is reduced alongside reductions in vascular stiffness in SM22-Cre;Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− mice versus Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− controls. Reductions in aortic calcification and stiffness occurred in the absence of significant improvements in metabolic status. Aortic collagen accumulation and vascular wall thickness, other contributors to vascular stiffness (3), were unaltered, suggesting a fundamental contribution of endogenous Msx gene tone to vascular calcium and osteoprogenitor homeostasis. Consistent with this notion, ex vivo analysis of vascular adventitial myofibroblasts from SM22-Cre;Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− mice revealed reductions in Alizarin red calcium staining and expression of bone-specific markers. Because Msx1 and Msx2 are cell-autonomous inhibitors of OCN gene transcription (6,30), but OCN mRNA levels were reduced with Msx deficiency, we considered the possibility that reductions in Msx tone might reduce osteoprogenitor numbers. This was confirmed by enumerating alkaline phosphatase–positive cells arising in the adventitial myofibroblast population. Passman et al. (26) first identified that Shh demarcates the renewable mesenchymal progenitors with osteogenic potential within the aorta. We therefore examined the impact of Msx1+Msx2 deficiency on Shh expression and identified significant reductions in Shh in aortic myofibroblast cultures. Concomitant reductions in multiple Wnts not only confirm our previous studies indicating that Wnt genes are entrained to Msx tone but also suggest that the Wnt ligand milieu may be critical to development and maintenance of osteoprogenitors. Canonical Wnt signals can also upregulate both Msx1 and Msx2 gene expression (37), introducing the notion of feed-forward regulation. Thus, endogenous Msx signaling contributes to vascular calcification, directing prosclerotic gene regulatory programs that drive osteoprogenitor development and differentiation (7,23,24).

RNAi confirmed the role of Msx genes in TNAP and Wnt expression. Paracrine Wnt signals are important for the elaboration of osteogenic TNAP expression downstream of BMP2 and Shh (38). Of note, although commonalities exist in Msx1 and Msx2 regulation of TNAP and Wnt7b, Msx1 appears to play a uniquely important role in support of Shh expression. By contrast, Msx2 is more important in the elaboration of Wnt5a and Wnt2. Taken together, these data suggest that Msx1 and Msx2 differentially support Shh and Wnt signaling pathways to promote the osteogenic programming of adventitial progenitors. Future studies will identify genomic complexes targeted by Msx1 versus Msx2.

Of note, global Msx1+Msx2 deficiency results in embryonic lethality and abnormal cardiac morphogenesis due to cardiac neural crest hyperproliferation (12). Additionally, hypoplastic myocardial atrioventricular cushions in Msx1^−/−;Msx2^−/− embryos arise from impaired endothelial-mesenchymal transitioning necessary for atrioventricular valve formation (11). Thus, the cardiac defects of Msx1^−/−;Msx2^−/− embryos relate to aberrant tissue remodeling and hemodynamics arising from perturbed endothelial-mesenchymal transitioning and cell type–specific proliferative control (11,12). We speculate that the cardiovascular developmental consequences of Msx deficiency may be more significant in the endothelial lineage than in the VSMC lineage.

There are limitations to our studies. The current data newly establish the role for endogenous vascular Msx genes in diabetic arteriosclerosis, vessel stiffness, and the osteogenic potential of vascular myofibroblasts and confirm the impact of Msx1+Msx2 on Wnt gene expression. However, the specific Wnt ligands programmed by Msx1+Msx2 as necessary for osteoprogenitor development and differentiation have yet to be characterized. Multiple noncanonical as well as canonical Wnt ligands are regulated. Wnt11, Wnt5a, and Wnt7b are capable of provoking both noncanonical and canonical signals, depending on context (39), and these ligands participate in both osteogenesis and fibrogenesis (9,40,41). Other important Wnts are also expressed by aortic myofibroblasts (42,43), and the increased body fat observed in SM22-Cre;Msx1(fl/fl);Msx2(fl/fl);LDLR^−/− mice following HFD challenge likely arises in part due to reduced Wnt tone (44). Finally, Msx1+Msx2 deficiency did not completely eliminate vascular calcium accrual. As Speer et al. (45) and Byon et al. (46) have shown, a Runx2-positive cell population is responsible for ∼80% of the vascular osteogenic cells within the vessel wall, driving osteogenic mineralization in response to oxidative stress. Msx1+Msx2 deficiency did not abrogate vascular Runx2 expression, and we speculate that Runx2 osteogenic programming contributes to residual disease. Nevertheless, the current data demonstrate that significant reductions in vascular calcification, with concomitant reductions in arterial stiffness, can be achieved through reductions in arterial Msx signaling. Future studies will identify the specific Wnt ligands required for programming osteogenic differentiation and vascular calcification in conjunction with Msx1+Msx2. Targeted inhibition of these ligands and their cognate receptors may phenocopy, in part, the beneficial actions of myofibroblast Msx deficiency in arteriosclerotic disease and thus offer a therapeutic strategy.

Funding. This work was supported by National Institutes of Health grants HL-69229 and HL-81138 to D.A.T. and DE-016320 to R.M., the Barnes-Jewish Hospital Foundation, and the Sanford-Burnham Medical Research Institute.

Duality of Interest. D.A.T. serves as a consultant for Eli Lilly, Daiichi-Sankyo, and Merck & Co. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. S.-L.C., A.B., J.-S.S., and A.K. researched data and reviewed and edited the manuscript. B.R., K.K., Y.B.A., and M.M. researched data. R.M. reviewed and edited the manuscript. D.A.T. researched data and wrote the manuscript. D.A.T. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.