Increasing evidence suggests that vascular calcification undergoes a process similar to bone formation, which is mediated by osteoblast-like cells in situ or from the circulation. One osteoblast-like cell that has been implicated in situ is the VSMC (5). Prior to the eventual results of vascular calcification, VSMCs undergo a phenotypic switching from a contractile to a synthetic and osteogenic phenotype through the loss of contractile marker proteins such as smooth muscle α-actin and smooth muscle 22α and the gain of osteogenic proteins like osteocalcin, osteopontin, and alkaline phosphatase (6). These resulting cells can produce a mineralizing matrix that is discrete from normal vascular tissue. In addition to VSMCs in the medial layer, it is noteworthy that osteoblast-like cells can also be derived from endothelial cells in the intima, pericytes in microvessels, calcifying vascular cells, myofibroblasts in the adventitia, and progenitor cells (7).

These osteoblast-like cells can communicate with each other through different signaling mechanisms, thereby contributing to the development of vascular calcification. Thus, understanding of those pathways will provide important insights into the development of a therapeutic platform. The major stimuli for ectopic calcification in blood vessels include oxidative stress, oxidized lipid, and phosphates (8). It is noteworthy that oxidative stress also causes atherosclerosis and other cardiovascular pathologies (9) and is associated with human diseases such as diabetes mellitus and chronic kidney diseases (10). Likewise, oxidative stress can also cause an osteogenic phenotypic switching of VSMCs (11).

Both viable and dead VSMCs contribute to vascular calcification; the contribution of apoptosis has been well documented in experimental models of phosphate-induced VSMC calcification (12). In addition, both viable cells and cells undergoing apoptosis are actively involved in the development of vascular calcification by releasing extracellular vesicles (13). These extracellular vesicles provide the nucleus for the deposition of calcium in the extracellular matrix. It is well known that mitochondrial dysfunction contributes to the development of atherosclerosis. Mitochondrial damage results in a reduction in ATP production and an impairment in overall cellular function. This mitochondrial damage is one of the important initial triggering signals in the early phase of atherosclerosis development (14). Considering that atherosclerosis shares many initial signal cascades with vascular calcification and that both are often found concomitantly in the same pathologic lesions, it is likely that this mitochondrial dysfunction also contributes to the development of vascular calcification. Indeed, mitochondria are involved in the intrinsic apoptosis pathway, which in turn activates caspase-9 signal via release of cytochrome c from the cytosol and thereby leads to nuclear DNA fragmentation and other apoptotic phenotypes (15).

Stress and various stimuli can induce perturbations of the endoplasmic reticulum, which results in the accumulation of misfolded and unfolded proteins and the resultant pathologic consequences of apoptosis (16). Likewise, increasing evidence has elucidated the contribution of endoplasmic reticulum stress in the development of vascular calcification in pathologic conditions with pro-calcifying stimuli, such as high glucose in type 2 diabetes mellitus (17), chronic kidney diseases (18), and atherosclerosis (19). The endoplasmic reticulum stress-induced vascular calcification is known to be mediated by several molecules, such as activating transcription factor 4 (ATF4) and C/EBP homologous protein (CHOP) (2021).

The BMPs, a subgroup of the TGF-β superfamily of growth factors, provide essential signals for normal developmental processes such as embryonic patterning, organogenesis, and bone formation. BMPs also work in pathological conditions including tumor angiogenesis, diabetic vascular complications, and ectopic calcification (22). Of the more than 20 BMP subtypes discovered, BMP2 is closely associated with the development of vascular calcification (23). As discovered in human atherosclerotic calcification lesions, VSMC-specific overexpression by transgenic introduction of BMP2 in apolipoprotein E (apoE)-disrupted mice accelerates vascular calcification (24). Conversely, BMP2 inhibition reduces vascular calcification in atherosclerotic low-density lipoprotein receptor (LDLR)-deficient mice (25). Calcifying phenotypic switching, a trans-differentiation of VSMCs into osteogenic cells, interestingly, is also known to be caused by BMP2 (2325), which further suggests the critical involvement of BMP2 in the development of vascular calcification.

Upon binding of BMP2, both type I and type II BMP2 receptors form a heterodimeric complex, which then phosphorylates the signal proteins named small mothers against decapentaplegic (SMAD) 1/5/8 to relay the signal inside the cells. SMAD1/5/8 again makes a heteromeric complex with SMAD4 and shuttles into the nucleus to initiate transcription of the downstream genes (26). The main transcription factors in association with the BMP2/SMAD signal are runt-related transcription factor 2 (RUNX2) and muscle segment homeobox 2 (MSX2) (523). RUNX2 plays an important role in the differentiation of osteoblasts and bone formation (2728). RUNX2 is not expressed in quiescent normal vessels. Its expression is, however, highly upregulated in calcified vessels as BMP2 (2930).

Like RUNX2, MSX2 is an important transcription factor in osteoblast differentiation and bone formation (3132). In blood vessels, it was reported that BMP2 also activates MSX2 for pro-calcific effects (3334). Vascular MSX2 expression is increased in the vascular calcification in LDLR-knockout mice fed a high-fat diet (35), whereas VSMC-specific knockout of both MSX1 and MSX2 attenuates the vascular calcification in those mice (36).

Wnt/β-catenin signals are critical for mediating cell differentiation, proliferation, and death; Wnt binds to its membrane-bound receptors of Frizzled and LRP-5/6, which causes translocation of nuclear β-catenin and downstream target gene transcription (37). Wnt also targets many vascular calcification-related proteins, such as RUNX2 and osteocalcin. Like MSX-2, in response to inflammation and reactive oxygen species, the Wnt signal in the adventitia is activated by BMP2/4 released from pericytes and the endothelium (38). In addition, the Wnt/β-catenin signal induces deposition of calcium in pericytes via BMP2 (39), which suggests crosstalk between BMP and Wnt signals in the development of vascular calcification. These osteogenic proteins, BMP-2, MSX2, and Wnt, are also known to induce endoplasmic reticulum stress, which might provide an alternative pathway for their pro-calcifying effects (3440).

As observed in patients with chronic kidney disease, loss of anti-calcific molecules also provides an important mechanism for the development of vascular calcification. Matrix G1a protein (MGP), fetuin-A, and osteoprotegerin (OPG) have been implicated as anti-calcific molecules (6). In addition to its direct binding to calcium to prevent vascular calcification, MGP reduces BMP2 activity by direct binding to BMP2 (41), which was further supported by the results of an in vivo animal study (42). Fetuin-A, a glycoprotein that is released from adipose tissue and liver, binds to calcium and hydroxyapatite in the blood to prevent vascular calcification (643).

The recent discovery of OPG, a phosphoprotein, and its associated proteins, receptor activator of nuclear factor-κB (RANK)/RANK ligand (RANKL), provides evidence of a direct association of vascular calcification with bone formation and loss (44). As expected, OPG knockout mice show severe osteoporosis, whereas OPG overexpression attenuates osteoclast differentiation and then causes increased bone mass (45). As expected, a decrease in OPG activity results in vascular calcification (46). Mechanistically, RANKL binds to its physiologic receptor, RANK, to induce osteoclast differentiation (47). OPG, however, interrupts the RANK/RANKL signal. This RANKL/RANK/OPG signal complex has an emerging role in the cellular communications in vascular calcification development (44). RANKL, interestingly, directly induces the release of BMP2 from human aortic endothelial cells, which suggests direct crosstalk between the two signal cascades (48).

Proteins are not stable but are subject to active enzymatic modification, such as phosphorylation, acetylation, methylation, ubiquitinylation, carboxylation, nitrosylation, and sumoylation. These modifications, named posttranslational modifications, are finely balanced by two opposite enzymes, such as kinases vs. phosphatases, acetyltransferases vs. deacetylases, methyltransferases (methylases) vs. demethylases, etc. These paired proteins are often referred to as writers (kinases, acetyltransferases, methyltransferases, etc.) and erasers (phosphatases, deacetylases, demethylases, etc.). These modifications relay biological signals that are recognized by a series of molecules, called readers, containing particular structures such as a bromodomain extra-terminal domain. The posttranslational modifications of histone molecules are especially emphasized as epigenetic regulatory mechanisms due to their properties of inheritance of the gene expression patterns. It is noteworthy that these histone posttranslational modifications work in combinatorial ways, called the histone code, which is critical in the fine regulation of diverse cellular events (49). Interestingly, not only histones but multiple other proteins undergo posttranslational modifications, which results in the fine regulation of diverse biological and pathological functions. For example, protein acetylation takes place either at lysine residues inside the proteins or at the N-terminal of the target proteins. This lysine acetylation is associated with the regulation of enzyme activity, DNA recruitment, and transcription. Typically, histone acetylation opens the nucleosome structure, thereby enabling access of the transcriptional machinery to induce the transcription of downstream genes. In contrast, histone deacetylation represses transcription (49).

Histone deacetylation is mediated by a family of HDACs, which consists of 18 different HDAC molecules. Based on their structural similarities and substrates, the 18 HDACs are divided into 4 classes. HDAC1/2/3/8 belong to class I, whereas HDAC4/5/6/7/9/10 are class II. The requirement for NAD⁺ for their deacetylase activity results in HDAC12–18 being classified as class III HDACs. The class IV HDAC group has a single member: HDAC11. In addition, class II HDACs can be divided into class IIa (HDAC4/5/7/9) and IIb (HDAC6/10). Among these HDACs, the deacetylase domain is well-conserved. Fig. 1 shows the molecular structures and tissue distributions of the HDACs.

It is somewhat in debate whether these HDACs have eventual deacetylase activity (50). Rather, as shown in the case of class IIa HDACs, some HDACs provide a scaffolding function by binding to other transcription factors including myocyte enhancer factor 2 (MEF2), thereby regulating the transcriptional activity of their binding partners (51) (Fig. 1). Some reports, including ours, however, have elucidated the existence of intrinsic deacetylase activity of the class II HDACs (5253). In the case of the class I HDACs, although the molecular sizes are relatively small and the deacetylase domain covers almost the entire structure (Fig. 1), their intrinsic activities are greater than those of the class IIa HDACs. Recently, instead of HDAC, a new term of lysine deacetylase (KDAC) is being used, because many proteins other than histones have been discovered as substrates of HDACs. This modification of nonhistone proteins provides diversity in their functions, and the acetylation status is finely regulated by protein acetyltransferases and KDACs (54).

Interestingly, the HDACs themselves undergo diverse posttranslational modifications. These modifications regulate HDAC activity by altering their intrinsic enzymatic activity, their intracellular localization, their binding partners, or their protein stability. For example, regardless of their intrinsic activity, the intracellular localization of the class II HDACs provides their repressive regulatory mechanism (51). Phosphorylation causes cytosolic shuttling of the class IIa HDACs from the nucleus, which causes de-repression of prohypertrophic genes and thereby induces cardiac hypertrophy (5556). In contrast, a recent report demonstrated the opposite action of nuclear accumulation of HDAC5, a class II HDAC, in response to β-adrenergic simulation in cardiomyocytes (57). Our group also reported that either phosphorylation (58) or acetylation (52) induces activation of HDAC2, a class I HDAC, and thereby induces cardiac hypertrophy.

It is interesting that each posttranslational modification can affect other modifications and that multiple modifications in one molecule at different locations or even at a single residue (e.g., lysine) can take place. These multiple modifications may act synergistically or antagonistically. As an example of multiple modifications at a single residue, consider lysine. Lysine can be a target of acetylation, methylation, or ubiquitination. One of the well-known mechanisms of multiple modifications at a single residue is histone 3 lysine-9 (H3K9), which serves as a substrate for both acetylation and methylation. Trimethylated histone 3 lysine-9 (simply, H3K9me3) is associated with repression of transcription, whereas acetylation of H3K9 indicates an increase in transcription. The posttranslational modifications of the HDACs and the regulatory mechanisms have been reviewed previously by our group (54).

We first observed that treatment with either trichostatin A (a non-class-specific HDAC inhibitor) or apicidin (a class I-specific HDAC inhibitor) did not affect calcium content in primarily cultured VSMCs (78). Interestingly, however, trichostatin A and apicidin potentiated vascular calcification induced by either inorganic phosphate in VSMCs or vitamin D3 in mice. In the following series of experiments, we found that loss of activity of HDAC1, one of class I HDACs, results in the exaggeration of vascular calcification. Specifically, we found that 1) HDAC1 siRNA imitated the effects of HDAC inhibitors, and 2) VSMC-specific deletion of HDAC1 in the mice exhibited enhancement of vascular calcification phenotypes. We further examined the expression levels of class I HDACs (HDAC1, HDAC2, HDAC3, and HDAC8) and found that vascular calcification reduces the protein expression level of HDAC1. Indeed, in vascular calcification samples obtained from diverse animal models or from human calcified coronary artery samples, we found that the expression of HDAC1 was significantly reduced (78).

We next questioned how vascular calcification causes the reduction in HDAC1 expression. Through a series of experiments, we found that HDAC1 protein degradation was enhanced in vascular calcification conditions. These results raised the possibility of involvement of the protein-degradation pathway in the reduction of the HDAC1 expression level. Indeed, we found that the polyubiquitination-dependent protein degradation pathway is involved in the degradation of HDAC1 in vascular calcification events. Together with the autophagic degradation system, proteasomal degradation is the most important system of the protein quality control pathways to remove abnormal or unnecessary proteins and recycle them. In the proteasomal degradation pathway, polyubiquitination of proteins is the key initial signal to recognize the target to be degraded (79). As exemplified in the catabolic process during muscle atrophy, these ubiquitin-proteasome and autophage-lysosome pathways are very important in both pathophysiology and normal homeostasis (80).

Polyubiquitination is caused by the covalent bonding of target proteins with a serial conjugation of small proteins named ubiquitin (81). Of note, polyubiquitination by multiple conjugations of ubiquitin K48 initiates 26S proteasome-dependent degradation, while K65 polyubiquitination as well as monoubiquitination (attachment of a single ubiquitin molecule to a target) is not associated with protein degradation but is involved in endocytosis, intracellular shuttling, enzymatic activation, or transcriptional regulation (82).

To induce polyubiquitination-dependent protein degradation, the target protein should undergo sequential conjugations catalyzed by serial enzymes of E1, E2, and E3 ligases (83). E1 and E2 ligases are involved in a common pathway; they work to make the multiple conjugations of ubiquitin proteins themselves. Then the E3 ligases determine the target protein and make the final conjugation of the multiple ubiquitin molecules generated by the E1/E2 ligases to the target (83). Several hundreds to thousands of E3 ligases are expected to be involved in the final process. Because target protein specificity is determined by the E3 ligase, it is important to find which E3 ligase is involved in the proteasomal degradation pathway of a protein of interest. Likewise, we believed that finding an HDAC1-specific E3 ligase was critical for further investigation of the HDAC1-associated mechanism in the development of vascular calcification.

Thus, we next tried to find the HDAC1 E3 ligase involved in vascular calcification. First, by literature searching, we found four HDAC1-specific E3 ligases: CHRF (84), MDM2 (85), PIRH2 (86), and Cullin3-REN (87). In addition, we performed a cDNA microarray and searched for E3 ligases that were upregulated by treatment with inorganic phosphate. In both the literature search and the cDNA microarray, the expression of MDM2 was dramatically increased (78). MDM2 has multiple targets to induce polyubiquitination and was first elucidated as a p53-specific E3 ligase (88). Because of its property to antagonize p53 activity (88) and the observation of an increase in its expression in cancer (89), MDM2 is considered an oncogene that serves as a clinical biomarker in certain cancers. MDM2 has multiple targets, of which HDAC1 is one (85).

Next, we further confirmed that MDM2 is involved in an HDAC1-associated vascular calcification mechanism as an E3 ligase by obtaining the results of 1) MDM2-upregulation by inorganic phosphate, 2) MDM2-induced increase in calcium contents in VSMCs, 3) direct binding of MDM2 to HDAC1, 4) MDM2-induced degradation of HDAC1, and 5) MDM2 RING domain-dependent ubiquitination of HDAC1. In addition, we observed that MDM2 is upregulated in diverse animal models and in human coronary samples with either medial calcification or intimal calcification (78). The most striking observation in our work is the attenuation of vascular calcification by RG 7112, an MDM2 inhibitor (78). The working hypothesis of vascular calcification regarding MDM2/HDAC1 is shown in Fig. 2.

Our findings are of major significance for several reasons. First, we have demonstrated that polyubiquitination-dependent protein degradation induces vascular calcification and that MDM2/HDAC1 signal cascades are involved in the cardiovascular pathophysiology. To our knowledge, only two E3 ligases demonstrated by two research groups including ours are known to be involved in the polyubiquitination-associated vascular calcification: MDM2 (78) and NEDD4 (90). Second, we have shown that HDAC inhibition potentiates vascular calcification. Considering that many HDAC inhibitors are being developed and are under extensive investigation as anti-cancer drugs, the likeliness of an increase in vascular calcification by HDAC inhibitors should be noted as a potential adverse effect. Third and most importantly, we have provided new concepts for drug development against vascular calcification by demonstration of the beneficial effect of blockade of either HDAC1 degradation or MDM2 activity. Additionally, RG 7112 and its associated drugs would be excellent models.