Histone Deacetylases in Bone Development and Skeletal Disorders

Abstract

Histone deacetylases (Hdacs) are conserved enzymes that remove acetyl groups from lysine side chains in histones and other proteins. Eleven of the 18 Hdacs encoded by the human and mouse genomes depend on Zn²⁺ for enzymatic activity, while the other 7, the sirtuins (Sirts), require NAD2⁺. Collectively, Hdacs and Sirts regulate numerous cellular and mitochondrial processes including gene transcription, DNA repair, protein stability, cytoskeletal dynamics, and signaling pathways to affect both development and aging. Of clinical relevance, Hdacs inhibitors are United States Food and Drug Administration-approved cancer therapeutics and are candidate therapies for other common diseases including arthritis, diabetes, epilepsy, heart disease, HIV infection, neurodegeneration, and numerous aging-related disorders. Hdacs and Sirts influence skeletal development, maintenance of mineral density and bone strength by affecting intramembranous and endochondral ossification, as well as bone resorption. With few exceptions, inhibition of Hdac or Sirt activity though either loss-of-function mutations or prolonged chemical inhibition has negative and/or toxic effects on skeletal development and bone mineral density. Specifically, Hdac/Sirt suppression causes abnormalities in physiological development such as craniofacial dimorphisms, short stature, and bone fragility that are associated with several human syndromes or diseases. In contrast, activation of Sirts may protect the skeleton from aging and immobilization-related bone loss. This knowledge may prolong healthspan and prevent adverse events caused by epigenetic therapies that are entering the clinical realm at an unprecedented rate. In this review, we summarize the general properties of Hdacs/Sirts and the research that has revealed their essential functions in bone forming cells (e.g., osteoblasts and chondrocytes) and bone resorbing osteoclasts. Finally, we offer predictions on future research in this area and the utility of this knowledge for orthopedic applications and bone tissue engineering.

I. INTRODUCTION

Bone is a highly dynamic tissue with many shapes and functions. Despite a general recognition that most bone fractures heal completely within a few weeks of time and that genomic material recovered from bones at archeological sites offers valuable clues to evolution of vertebrate life on this planet, the astonishing vitality and regeneration capacity of the bony skeleton are vastly underappreciated. Plasticity in the genome, epigenome, and cellular signaling pathways provides cells of mesodermal origin with the capacity to form bones that are flat, long, and/or curved so that they can protect vital organs and support movement. These processes also temporally and spatially orchestrate timely fracture repair and the production and release of molecules that regulate many physiological processes, including hematopoiesis, metabolism, and bone remodeling (13, 97, 109, 126, 137, 147, 207, 213). Many cellular enzymes drive and reverse the intrinsic chemical reactions that produce collagens, growth factors, hormones, and extracellular matrix proteins characteristic of bone (5, 25, 63). This review focuses on one conserved class of enzymes, histone deacetylases, that control both chromatin structure and signaling events, and thus affect the myriad of activities required for bone formation, repair, and regeneration. Hdacs are clinically relevant because numerous molecules that inhibit their activities are used alone or in combination with other drugs to treat a variety of cancers, and are being considered or tested as treatments for other diseases including arthritis, diabetes, epilepsy, heart disease, HIV infection, neurodegeneration, and numerous aging-related disorders (27, 42, 48, 95, 116, 136, 169).

Histone deacetylases (Hdacs; EC 3.5.1.98) catalyze the hydrolysis of an N6-acetyl-lysine residue of histones and other proteins to yield a deacetylated substrate, an acetyl group and one molecule of H₂O. They are thus considered “erasers” of epigenetic marks or posttranslational modifications on chromatin and other intracellular molecular complexes. Hdacs are counteracted by histone/lysine acetyltransferases (HAT/KAT; EC 2.3.1.48), which add acetyl groups to lysines from acetyl coenzyme A (acetyl CoA) and are “writers” of posttranslational modifications. The removal of acetyl groups from lysines by Hdacs restores the positive charge on the side chain and alters the affinity of the modified protein for other molecules, including nucleic acids and proteins with bromodomains or tandem PHD domains (“readers” of acetylated lysines). Histone deacetylation is typically associated with promoter and enhancer inactivity, but deacetylation of non-histone proteins (e.g., transcription factors, DNA repair and replication proteins and chaperones) affects nearly every cellular process by destabilizing proteins or preventing proper cellular trafficking and function (29, 54, 61, 74, 80, 107, 209). A seminal report from 2009 identified 3,600 acetylated residues in over 1,750 proteins by high-resolution mass spectrometry (29). Due to the broad range of substrates, Hdacs are sometimes referred to as lysine deacetylases (Kdacs), but we will refer to them as Hdacs here. In 2011, we reviewed the role of Hdacs in bone development and maintenance and discussed the consequences of Hdac inhibition on bone strength and structure (116). Here we provide an update on research in this area and place the results into the greater context of bone formation, repair, regeneration, aging, and disease. We begin by summarizing general properties of the Hdacs and Sirts, and then discuss their essential functions in bone-forming cells (osteoblasts, chondrocytes, and their mesenchymal progenitors) and bone-resorbing cells (osteoclasts). We will also discuss both the positive and negative clinical implications of pharmaceutically altering Hdac activity by reviewing in vivo studies with Hdac inhibitors and Sirt activators and proposing potential orthopedic applications.

II. HDAC STRUCTURE AND FUNCTION IN BONE

The human and mouse genomes each encode 18 Hdac genes. Hdacs are divided into four families (class I, II, III, and IV) based on their protein structure, function, subcellular localization, and expression (FIGURE 1A). Classes I, II, and IV are zinc-dependent enzymes and will be referred to as Hdacs from here on, while class III Hdacs require NAD⁺ for enzymatic activity and will be referred to as Sirtuins (Sirts). In a small and preliminary RNA-sequencing study, we observe great variability in expression of HDAC and SIRT transcripts in adult human bone and articular cartilage. HDAC1 and HDAC3 are the most abundant HDACs in both tissues (FIGURE 1, B and C). Adult bone also contains many transcripts for HDAC5, HDAC6, and HDAC7, but relatively few for HDAC4, HDAC8, and HDAC9. In normal human articular cartilage, HDAC2, HDAC5, HDAC6, and HDAC7 are expressed at higher levels than HDAC4, HDAC8, HDAC9, and HDAC10. Among the class III Hdacs, SIRT2 is the most highly expressed in adult bone, whereas several SIRTs are equally expressed in normal human articular cartilage (FIGURE 1, D and E). These expression data provide some context for prioritizing studies on Hdacs and Sirts in a mature skeleton, but as highlighted below, do not necessarily reflect the importance of these enzymes (e.g., Hdac4) in the developing skeleton or in different pathophysiological settings.

Figure 1.

HDAC structure and expression in adult human bone and cartilage. A: HDACs are grouped into four classes based on structural considerations. B–E: HDAC gene expression differs between bone and cartilage as illustrated by RNA-sequencing results (van Wijnen, unpublished data). RNA samples were obtained with Institutional Review Board approval from adult human bone (4 specimens) and articular cartilage (3 specimens) as fresh surgical waste. Each colored bar represents a different patient. Values represents reads per kilobase per million (RPKM).

A. Class I Hdacs

Mammalian Hdacs1, 2, 3, and 8 are structurally similar to the S. cerevisiae protein Rpd3 (34, 174). They contain a conserved deacetylase domain flanked by NH₂- and COOH-terminal extensions, are ubiquitously expressed in most tissues, and predominately localize to the nucleus (26, 203) (FIGURE 1A). Class I Hdacs exert high enzymatic activity toward histone substrates and are crucial for gene transcription, as well as for splicing, DNA replication and repair, and cell proliferation and survival (61, 110, 124). Class I Hdacs may be highly expressed in some cancers, but there is no clear evidence that Hdac overexpression is oncogenic. On the other hand, suppression of class I Hdac expression induces cell cycle arrest and other anti-tumor effects (reviewed in Ref. 48). Hdac1 and Hdac2 expression declines during the terminal differentiation of several mesenchymal lineages, including osteoblasts (28, 96).

Class I Hdacs are the enzymatic subunits of several multi-subunit complexes. They contain nuclear localization signals but do not bind chromatin directly. Rather, they are recruited to DNA structures by transcription factors and other proteins. Hdac1 and Hdac2 are highly homologous and can form homo- and heterodimers with each other; thus they are often but not always found together in similar repressive complexes including Sin3, NuRD, CoREST, PRC2, and MiDAC (4, 12, 38, 173, 204). Using a systems biology approach, Joshi et al. (79) described functional interactomes for Hdac1 and other Hdacs. This work demonstrated that the associations between Hdac1 and the above-mentioned chromatin remodeling complexes are relatively stable, as opposed to transient and dynamic interactions with transcription factors (79).

Hdac3 is functionally distinct from Hdac1/2 because it interacts with NCoR-SMRT co-repressor complexes (57, 58, 102, 180, 191). Structural studies on the Hdac3 and SMRT corepressor complex identified an inositol tetraphosphate molecule [d-myo-inositol-(1,4,5,6)-tetrakisphosphate (IP₄)] as a crucial functional component of the complex (189). Further analysis found that IP₄ binding pockets are also present in Hdac1 and Hdac2 and are necessary for engaging Hdac enzymatic activity (72, 121); however, their role in connecting Hdac activity to cell signaling events has not been elucidated yet. The enzymatic activity of Hdac3 is dependent on the phosphorylation of S424 and inversely correlated with the presence of the serine/threonine phosphatase PP4c (214).

Hdac8 was the last member of class I Hdacs to be cloned. Little is known about its association with cofactor complexes. Durst et al. (45) showed that the fusion protein formed by inv(16) rearrangements in acute myeloid leukemias associates with both mSin3A and Hdac8, but direct associations between Hdac8 and mSin3, NuRD, NCor/Smrt, or any of the aforementioned co-repressor complexes have not been reported. Hdac8 was the first zinc-dependent Hdac to have its crystal structure solved with an Hdac inhibitor (HDI) (183). This structure has been used as model for developing more selective HDIs.

B. Class II Hdacs

Class II Hdacs share structural homology to the yeast protein Hda1 (34) and are divided into two subclasses, class IIa and class IIb. Hdac4, Hdac5, Hdac7, and Hdac9 (alternatively called Hdac7b) comprise class IIa, while Hdac6 and Hdac10 are in class IIb because of the presence of two catalytic domains (Figure 1A). Compared with class I Hdacs, class II Hdacs have large NH₂-terminal extensions, which contain conserved binding domains for the myocyte enhancer transcription factor 2 (Mef2) as well as the chaperone protein 14-3-3. Phosphorylation of serine residues in the NH₂-terminal extension promotes the binding of 14-3-3 proteins to class II Hdacs, their translocation from the nucleus to the cytoplasm, and subsequent de-repression of target genes (106, 118, 184). The phosphorylation of class II Hdacs is controlled by many extracellular signals that activate multiple intracellular kinases (reviewed in Ref. 139). Generally, class II Hdacs have more temporally and spatially restricted expression patterns and thus have more specialized functions than class I Hdacs (61). Class II Hdacs appear to be important scaffolds for cellular processes, but have drastically reduced intrinsic enzymatic activity than class I Hdacs due to the presence of a histidine residue instead of tyrosine within the catalytic domain (93). The ability of class II Hdacs to recruit class I Hdacs with their conserved COOH-terminal deacetylase domain accounts for the majority of their enzymatic activity (50, 186, 205). This reduced enzymatic activity may explain why global deletion of most class II Hdacs does not cause embryonic lethality.

C. Class III Hdacs/Sirtuins

Class III Hdacs include the sirtuins (Sirts1-7), which are homologues of the yeast protein, silent information regulator (Sir)2. Sirts are dependent on the cofactor nicotinamide adenine dinucleotide (NAD⁺) and transfer an acetyl group from lysine to NAD⁺, creating O-acetyl ADP-ribose and nicotinamide, which becomes a feedback inhibitor of the enzymatic reaction. Emerging studies indicate that the enzymatic activity of Sirts may extend to different protein modifications. For example, Sirt6 has relatively low deacetylase activity, but efficiently removes myristoyl groups (long-chain fatty-acyl groups) from lysine residues (7, 75), and Sirt4 is primarily an ADP-ribosyltransferase (62). The subcellular localization of Sirts varies: Sirt1-3, 6, and 7 are found in the nucleus; Sirt1 and 2 are predominantly found in the cytoplasm; and Sirt3-5 are strongly expressed in mitochondria (8, 21, 62, 120, 132, 157, 158, 172). Sirts are of scientific interest because they control energy metabolism, inflammation, genome stability, and aging (reviewed in Refs. 27, 136). Sirts are also important for bone health, as Sirt1 and Sirt6 facilitate endochondral ossification and Sirt activators promote bone formation (see sects. IVB and VII).

D. Class IV Hdac

Hdac11 has a deacetylase domain with small COOH- and NH₂-terminal extensions, suggesting structural similarities to both class I and class II Hdacs, but placing it into its own group (class IV) (Figure 1A). Although it is expressed across a variety of tissues, including bone marrow mononuclear cells (152) and adult bone and articular cartilage, little is known about its function other than one report that Hdac11 was unable to suppress Runx2-induced gene transcription (73).

III. HDACS IN BONE DEVELOPMENT AND GENETIC DISEASES

A. Intramembranous Bone Development

Intramembranous ossification occurs when mesenchymal precursor cells condense within a fibrous connective membrane and directly differentiate into matrix-producing osteoblasts. This process generates many of the flat bones in the skull and the medial portion of clavicles and is reactivated during repair of highly stabile fractures and defects. Direct commitment of mesenchymal cells toward the osteoblast lineage requires the transcription factors Runx2 and Osterix (Osx/Sp7) and is largely influenced by canonical Wnt signaling, which promotes osteoblast differentiation and suppresses chondrogenesis of mesenchymal progenitors. Several Hdacs contribute to intramembranous ossification (Figure 2A). Hdac3 is required for proper craniofacial development, as conditional Hdac3 deletion in the skeletal mesenchyme (with Wnt1-Cre or Pax3-Cre) produced microcephaly, nearly undetectable frontal bone formation, and improper formation of the zygomatic arch (168) (Table 1). This phenotype was ascribed to increased expression of Msx1 and Msx2 and increased apoptosis of cranial neural crest cells. Similarly, deletion of Hdac3 in osteochondral progenitor cells (with Osx1-Cre) decreased calvarial bone thickness and density (150). Calvarial cells from these Hdac3-deficient animals expressed high levels of the cyclin-dependent kinase inhibitor Cdkn1a/p21 and a number of Wnt pathway inhibitors.

Figure 2.

Hdacs in bone and cartilage development. A: Hdacs facilitate intramembranous bone ossification. Mesenchymal cells differentiate directly into osteoblasts during intramembranous bone formation. Deletion of Hdac3 or Hdac8 within the developing skeleton alters intramembranous ossification, skull formation, and bone density. B: during endochondral ossification, mesenchymal condensations give rise to a cartilaginous template for bone formation. Several Hdacs regulate chondrogenesis in the sclerotome, the expression of matrix genes associated with chondrocyte proliferation, and hypertrophy. Hdacs control bone modeling by influencing osteoblast commitment, promoting osteoblast proliferation, regulating osteoblast maturation and matrix deposition, and by facilitating osteoclast resorption activity and motility.

Table 1.

Skeletal phenotypes resulting from altered Hdac and Sirt expression in mice

Hdac	Genetic Alteration	Cells Targeted	Bone Phenotype	Reference Nos.
Hdac1	KO	All	Embryonic lethal	92, 122
Hdac2	KO	All	Reduced body size, smaller vertebrae and pelvis	178, 218
Hdac3	KO	All	Embryonic lethal, E9.5	6, 123
	CKO	Osteochondral progenitors (Osx1-Cre)	Cortical and trabecular bone loss, decreased bone formation rate, fewer osteoblasts, increased marrow adiposity, abnormal chondrocyte hypertrophy	11, 150
	CKO	Osteoblasts (OCN-Cre)	Age related cortical and trabecular bone loss, decreased bone formation rate, reduced osteoblast numbers	114
	CKO	Neural crest cells (Wnt1-Cre)	Craniofacial abnormalities	168
	CKO	Neural crest cells (Pax3-Cre)	Craniofacial abnormalities	168
	CKO	Chondrocyte (Col2-Cre)	Embryonic lethal	Carpio et al., unpublished data
	CKO	Chondrocyte (postnatal Col2-CreERT)	Reduced trabecular bone density, increased cortical bone thickness, short stature, delayed secondary ossification center development, residual hypertrophic cartilage in growth plate, increased osteoclast activity in primary spongiosa	Carpio et al., unpublished data
Hdac4	KO	All	Ectopic chondrocyte hypertrophy and enhanced endochondral ossification, increased MMP13 expression	165, 185
	CKO	Osteoblast (Runx2-Cre)	Reduced bone density, increased Rankl expression and bone resorption, reduced osteocalcin expression	108, 133
	Tg	Chondrocytes (Col2a1 promoter)	No endochondral ossification	185
Hdac5	KO	All	No gross skeletal phenotypes at birth	19
	KO	All	Reduced bone density, increased Rankl expression and bone resorption, normal osteocalcin expression	108, 133
	KO	All	Reduced bone trabecular bone density, reduced bone formation, increased Sost expression	190
Hdac6	KO	All	Increased trabecular bone mineral density	215
Hdac7	KO	All	Embryonic lethality	20
	CKO	Chondrocytes (Col2a1-Cre)	Embryonic lethal	10
	CKO	Chondrocytes (postnatal Col2a1-CreERT)	Reduced trabecular bone volume and trabecular number, increased chondrocyte proliferation	10
	CKO	Osteoclasts (LysM-Cre)	Reduced bone mass, increased bone resorption, normal bone formation	76, 170
	CKO	Early osteoclasts, hematopoietic cells (Tie2-Cre)	Embyronic lethality	76
	CKO	Osteoclasts (PT-Cre)	Reduced bone mass, increased bone resorption, normal bone formation	76
Hdac8	KO	All	Decreased bone mineral density	IMPC*
	KO	All	Ossification defects in frontal and interparietal bones	60
	CKO	Neural crest cells (Wnt1-Cre)	Ossification defects in frontal and interparietal bones
	CKO	Osteoblasts (Col1a1-Cre)	None
	CKO	Chondrocytes (Col2a1-Cre)	None
	CKO	Mesenchymal precursors (Twist1-Cre)	None
Hdac9	KO	All	Reduced bone mass, increased bone resorption and reduced bone formation	77
Hdac10	KO	All	Increased circulating alkaline phosphatase levels in females, no gross phenotype	IMPC*
Hdac11	KO	All	No gross skeletal phenotypes	152
Sirt1	KO	All	Shorter stature, craniofacial abnormalities, reduced endochondral ossification, reduced cortical thickness	51, 119
	Het	All	Trabecular bone loss in females, increased osteoclast number, decreased osteoblast numbers, increased marrow adiposity, smaller in size, increased osteoarthritis and articular chondrocyte apoptosis	30, 51
	CKO	Osteoblasts (2.3Col1a1-Cre)	Decreased bone mass in females, reduced osteoblast differentiation and increased NFκB activity	47
	CKO	Osteoclasts (LysM-Cre)	Decreased bone mass in females, increased osteoclastogenesis and NFκB acetylation	47
	CKO	Mesenchymal progenitors (Prx1-Cre)	Cortical bone loss with aging, reduced bone formation	166
	CKO	Osteochondral progenitors (Osx1-Cre)	Decreased cortical thickness in females, reduced bone formation	71
	Inducible HDAC	Postnatal-all cells (CreERT2)	Decreased cortical bone thickness	119
	CKO	Chondrocytes (Col2a1-Cre)	Normal size and long bones growth, increased susceptibility to osteoarthritis	112
Sirt2	KO	All	No gross skeletal phenotype, genomic instability, prone to tumors	160
	KO	All	Increased bone mineral density in males	IMPC*
Sirt3	KO	All	No gross skeletal phenotype, normal bone mineral density and content	105
Sirt4	KO	All	No gross skeletal phenotype	62
Sirt5	KO	All	No gross skeletal phenotype	105
Sirt6	KO	All	Osteoporosis, shorter long bones, fewer proliferating and hypertrophic cells in growth plates, failed to thrive after 3 wk of age	125, 143
Sirt7	KO	All	No reported skeletal phenotype	182

CKO, conditional knockout mice; Het, heterozygous mice; KO, knockout mice; Tg, transgenic mice.

^*IMPC: International Mouse Phenotyping Consortium (http://www.mousephenotype.org/data/genes/).

Hdac8 is also required for proper skull development. Conditional deletion of Hdac8 within neural crest cells using Wnt1-Cre impaired calvarial development through decreased expression of the transcription factors Otx2 and Lhx1 (60) (Table 1). Loss-of-function mutations in HDAC8 were found in a subset of patients with Cornelia de Lange syndrome (CdLS) and with a condition resembling Wilson-Turner linked mental retardation syndrome (35, 36, 49, 64, 81) (Table 2). In the latter syndrome, mutations alter splicing of HDAC8 and cause premature stops in translation. The CdLS mutations inactivate the deacetylase activity of HDAC8, thus stabilizing the acetylation of the cohesion subunit, structural maintenance of chromosomes 3 (SMC3). Patients with HDAC8 mutations have typical CdLS phenotypes, but delays in anterior fontanel closure are notable because they are highly reminiscent of the Hdac8-deficient mice. Distinct craniofacial dimorphisms and shorter bones characterize both syndromes.

Table 2.

HDAC mutations in human diseases with bone phenotypes

HDAC (MIM No.)	Disease (OMIM No.)	Mutation/Variant	Bone Phenotype	Reference Nos.
HDAC2 (605164)	Interstitial 6q21q22.1 Deletion Syndrome-medial	Deletion of 3–10 genes, including HDAC2	Hypertelorism, thoracic scoliosis, vertebral rotation	173a
HDAC4 (605314)	Brachydactyly-Mental Retardation Syndrome; aka Albright Hereditary Osteo-Dystrophy-Like Syndrome or 2q37 Microdeletion Syndrome (600430)	Loss of function; deletions result in haploinsufficiency	Distinctive craniofacial features, and skeletal abnormalities, including brachydactyly type E, and shortened metatarsals and metacarpels	187, 195
HDAC5 (605315)	Bone mineral density quantitative trait locus 15 (613418)	Mutation in miR2861 prevents binding to HDAC5 UTR and increases HDAC5 expression	Low bone mineral density	101
	Osteoporosis (166710)	SNP rs228769 (8 kb upstream of HDAC5)	Low bone mineral density in lumbar spine and femoral neck	151
HDAC6 (300272)	Chondrodysplasia with platyspondyly, distinctive brachydactyly, hydrocephaly, and microphthalmia (300863)	Gain of function; mutations in miR binding site in HDAC6 UTR	Distinctive craniofacial features, generalized reductions in bone mineralization, small growth plates	167
HDAC8 (300269)	Cornelia de Lange Syndrome 5 (300882)	Loss of function; mutations increase acetylation of cohesion subunit SMC3	Distinctive craniofacial features, delayed fontanel closure, small hands and feet	35, 81
	Wilson-Turner X-linked Mental Retardation Syndrome (309585)	Loss of function; mutations alter splicing and introduce a premature stop codon	Distinctive craniofacial features, short stature	64

MIM, gene locus number for the Mendelian Inheritance in Man catalog; OMIM, Online Mendelian Inheritance in Man; SNP, single nucleotide polymorphism; UTR, untranslated region.

Intramembranous bone defects have not been described in mouse models of altered class II Hdac expression, but other model systems suggest that these Hdacs may also contribute to intramembranous bone formation. Zebrafish embryos injected with hdac4 morpholinos showed improper palate formation and shortened faces (37). Interestingly, deletions or mutations in HDAC2, HDAC4, and HDAC6 are associated with distinctive craniofacial phenotypes or skeletal abnormalities in rare human syndromes; however, a causative role for altered deacetylase function in these phenotypes has not been proven and requires further investigation (Table 2) (167, 173a, 187, 195).

B. Endochondral Bone Development

The majority of bones (including vertebrae and long bones) form and heal by endochondral ossification, where a temporary cartilaginous anlage serves as a template for longitudinal bone growth, repair, and mineralization. During endochondral bone development (Figure 2B), mesenchymal condensations form a growth plate consisting of three zones of resting, proliferative, and hypertrophic chondrocytes. The resting cells are quiescent, but after chondrocytes begin proliferating, they produce a matrix rich in proteoglycan, type IIa collagen, and aggrecan. Chondrocyte maturation through the proliferative zone, into the hypertrophic zone, and subsequently out of hypertrophy is heavily regulated through a feedback mechanism involving parathyroid hormone-related protein (PTHrP) and Indian hedgehog (Ihh) (reviewed in Refs. 83, 91). Hypertrophic chondrocytes downregulate expression of type IIa collagen and begin to synthesize a matrix consisting of type X collagen. In addition, hypertrophic chondrocytes mineralize their surrounding extracellular matrix (131) and promote vascularization of the growth plate through the production of vascular endothelial growth factor (VEGF) and other angiogenic factors. This facilitates the recruitment of myeloid progenitor cells, which differentiate into osteoclasts and chondroclasts that digest the calcified cartilage matrix. Matrix metalloproteinases (MMPs) produced by hypertrophic chondrocytes degrade the matrix within the calcified cartilage matrix. The majority of hypertrophic chondrocytes undergo apoptosis and leave behind a calcified matrix that serves as a scaffold for bone formation, but recent lineage tracing studies provide new support for the hypothesis that some hypertrophic chondrocytes give rise to bone marrow osteoblasts and matrix-residing osteocytes (202, 217). As discussed next, many Hdacs contribute to various steps of the endochondral ossification process.

1. Class I Hdacs and endochondral ossification

a) hdac1/hdac2. Genetic deletion of several class I Hdacs adversely affects endochondral ossification. Loss of hdac1 in zebrafish impaired cartilage formation in zebrafish pectoral fins and produced defective craniofacial cartilage by promoting premature cell apoptosis in some locations (posterior branchial arch) and impairing terminal chondrocyte differentiation in others (anterior mandibular and hyoid arches) (70, 144). Thus far, the role of Hdac1 in postnatal murine bone development has not been determined as Hdac1 KO mice die during early embryogenesis (92, 122). However, in the initial derivation of cartilage from the sclerotome during axial skeleton formation, Hdac1 associates with a crucial transcription factor, Nkx3.2, to confer the repressive activity required for chondrogenesis (85, 128, 211). Wuelling et al. (199) found that the zinc finger nuclear regulator Trps1 promoted chondrocyte mitosis by increasing the activity of Hdac1, as well as Hdac4. Hong et al. (66) found that HDAC1 and HDAC2 levels were elevated in articular cartilage from osteoarthritis patients and that they suppress expression of many matrix proteins through a common domain. They and others documented that Hdac1 repressed expression of cartilage extracellular matrix genes including collagen type 2, aggrecan, and cartilage oligomeric protein (Comp) (66, 68, 103), while Hdac2 suppressed collagen type 2, aggrecan, and collagen type 11 (66). Germline deletion of Hdac2 results in partially penetrant embryonic lethality and decreased survival of postnatal null animals (178, 218). Of the surviving Hdac2 null animals, significant decreases in body size and long bone length were observed with reductions in insulin-like growth factor I (IGF-I) signaling. These data point towards impairment of growth plate development, but this hypothesis has also not yet been tested in depth.

b) hdac3. Compared with Hdac1 and Hdac2, the role of Hdac3 in endochondral ossification has been extensively probed (Table 1). Germline deletion of Hdac3 caused early embryonic lethality (6, 123); however, conditional deletion from Osx1-Cre expressing osteochondroprogenitor cells reduced long bone length, delayed exit from the hypertrophic zone, and produced weak skeletons that were prone to fracture (11, 150). Despite a larger hypertrophic zone, these Hdac3-depleted mice produced less cartilage matrix and had lower expression levels of bone specific genes, aggrecan, Mepe, and osteopontin as well as VEGF (11). The reduced matrix production was at least partially explained by the blunted activation of Akt and its downstream substrates, including the protein translation factor p70 S6 kinase in Hdac3-depleted cells (Figure 3). Akt phosphorylation and activity were reduced in Hdac3-deficient chondrocytes due to increased expression of the phosphatase Phlpp1 (Figure 3) (11). Hdac3 bound to the Phlpp1 promoter in a region containing a Smad3 binding site, indicating that Hdac3 may also repress transforming growth factor (TGF)-β responses (11). More recent studies show that depletion of Hdac3 even earlier in chondrogenesis with Col2a1-Cre produces more severe phenotypes. Hdac3-CKO_Col2 animals rarely survived to birth. Those that did survive had severe endochondral bone defects and were hypomorphic for Hdac3 expression (16). Specifically, these animals had decreased femur length, poor trabecular bone architecture, and thin cortical bones. Due to the difficulty in obtaining viable Hdac3-CKO_Col2 mice, Hdac3 was also deleted postnatally with a tamoxifen inducible Col2a1-Cre transgene to generate Hdac3-CKO_Col2ERT pups. At 8 wk of age (7 wk post induction), Hdac3-CKO_Col2ERT animals had decreased cancellous bone and increased cortical bone. The latter phenotype is likely a compensatory adaptation to mechanical loading forces exerted on the long bones. In other words, to compensate for decreased chondrocyte maturation and cancellous bone formation, the cortical bone adapted by increasing in thickness. Mechanistically, Hdac3 deficiency in the Col2-Cre expressing chondrocytes increased production of inflammatory proteins, elevated Stat3 and Erk phosphorylation, and produced subsequent increases in Mmp13 expression in vitro and in vivo. Increased osteoclastogenesis is also apparent in the primary spongiosa of these mice (Carpio and Westendorf, unpublished data). In contrast, conditional loss of Hdac3 in mature osteoblasts and osteocytes expressing OCN-Cre is not associated with growth plate or long bone length defects, but these mice do show an osteopenic phenotype and defective matrix mineralization that worsens with age, leading to increased skeletal fragility (114). Surprisingly, Hdac3 CKO mice made with 2.3Col1a1-Cre transgenic mice were not viable (Westendorf, unpublished data), possibly because early transient expression of this promoter in the embryo caused unintended Hdac3 deletion in vital tissues/organs (104).

Figure 3.

Hdacs in bone and cartilage signaling pathways. Hdacs integrate multiple cytokine signaling cascades to transcription factors and gene transcription during bone and cartilage development.

c) hdac8. As discussed earlier, Hdac8 deficiency primarily altered cranial facial features, but its conditional deletion in either Col2-Cre expressing chondrocytes or Col1a1-Cre expressing osteoblasts produced no obvious long bone phenotypes (60). Further examination of these animals would be warranted in light of discoveries that patients with HDAC8 mutations have shortened long bones (Table 2; Refs. 35, 49, 64, 81).

d) summary of class i hdacs in bone formation. Class I Hdacs (Hdacs 1, 2, 3, and 8) are essential regulators of endochondral ossification. Germline deletion of many of the class I Hdacs results in early embryonic lethality, as does skeletal specific deletion of several class I Hdacs. The requirement of these Hdacs during endochondral bone development may stem from their ability to control the activity of essential cartilage and bone transcription and chromosome maintenance factors.

2. Class II Hdacs and endochondral ossification

a) hdac4. The roles of class II Hdacs in endochondral ossification are under intensive investigation because they integrate extracellular signals required for bone formation with transcriptional regulators (Figure 3). The first evidence that these enzymes contribute to endochondral ossification appeared with generation of mice globally deficient in Hdac4. These mice are viable, but exhibit premature endochondral ossification leading to many skeletal defects including vertebral body fusion, decreased long bone length, and premature ossification of the skull (185). Conversely, transgenic expression of Hdac4 from the Col2a1 promoter inhibited chondrocyte maturation by blocking Runx2 activity (185).

Parathyroid hormone (PTH) and PTHrP are crucial regulators of Hdac4 activity during bone and cartilage development (Figure 2). Shimizu and co-workers (164, 165) showed that PTH induced PKA to phosphorylate Hdac4 at serine 740, which released it from Runx2, induced expression of the Mmp13 promoter, and promoted Hdac4 degradation by the lysosome. PTH also stimulated Hdac4 expression to suppress Mmp13 in a negative feedback loop (165). In another study, PTHrP promoted Hdac4 association with Mef2c, thereby inhibiting the ability of Mef2c to promote chondrocyte hypertrophy (2, 90) (Figure 3). The repressive effects of the Hdac4/Mef2c complex on transcription and chondrocyte hypertrophy were prevented by salt-inducible kinase (SIK)3 and Ca²⁺/calmodulin-dependent kinase IV, which induced cytosolic localization of Hdac4 (55, 155). PTHrP also stimulated expression of the transcriptional co-repressor Zfp521, which recruited Hdac4 and inhibited Runx2 activity and chondrocyte hypertrophy (31) (Figure 3). MiR365, a microRNA induced in parallel with Ihh signaling in vitro and within the developing growth plate, may promote chondrocyte hypertrophy by decreasing Hdac4 expression (56). A more recent study demonstrated that conditional deletion of Hdac4 in osteoblasts expressing a Runx2-Cre transgene reduced vertebral trabecular bone density in young mice by increasing Rankl expression in osteoblasts, which in turn promoted osteoclastogenesis and bone resorption (133). In this study, PTH promoted the ubiquitination and degradation of Hdac4 in osteoblasts via Smurf2, thereby allowing Mef2c to activate Rankl transcription. Meanwhile, induction of sympathetic signaling in osteoblasts induced Hdac4 nuclear localization and association with the transcription factor Atf4 (133). Osteocalcin expression is enhanced by Atf4, but was reduced in Hdac4-deficient mice, leading the authors to suggest that Hdac4 may support osteocalcin expression through Atf4 (108).

Hdac4 also controls TGF-β and bone morphogenetic protein (BMP) signaling during bone development (Figure 3). Pei et al. (141) found that Hdac4 overexpression enhanced TGF-β-induced chondrogenic differentiation of synovial-derived cells. Hdac4 and Hdac5 bound to Smad3 in response to TGF-β and inhibited Runx2 activity (82). The association of Hdac4 and Hdac5 to TGF-β-activated phospho-Smad2 and 3 complexes may require Ski, which can inhibit the signaling pathway and accelerate chondrocyte differentiation (88). Hdac4 can also deacetylate multiple lysines in Runx2 in response to Bmp2 and promote its degradation via Smurf1 (74).

Loss-of-function, haploinsufficiency-inducing deletions or mutations in HDAC4 were detected in patients with Brachydactyly-Mental Retardation Syndrome (also called 2q37 Microdeletion Syndrome or Albright Hereditary Osteodystrophy-Like Syndrome; OMIM no. 600430) (187, 195) (Table 2). The afflicted individuals have distinctive craniofacial features and shortened metatarsals and metacarpals, consistent with a crucial role for Hdac4 in murine endochondral ossification.

b) hdac5. Hdac5 knockout (KO) mice showed no gross structural patterning or growth defects at birth (19), but two recent and independent studies demonstrated that Hdac5 is required for optimal bone density. Obri et al. (133) reported that Hdac5 KO mice had reduced trabecular bone density at 2-3 mo of age despite modest increases in bone formation rates because Rankl expression was increased in osteoblasts, which stimulated greater bone resorption by osteoclasts. Osteocalcin expression was not changed in this Hdac5 KO colony (108). Using mice from the same strain, Wein et al. (190) found that osteocytes from 8-wk-old Hdac5 KO mice produce more sclerostin, a secreted Wnt/Lrp antagonist, because Mef2c is more active; they did detect increases in osteoclast activity. The discrepancies between these studies might be resolved with tissue-directed depletion studies, which are needed to fully understand Hdac5's role in bone. Hdac5 KO mice are of normal size, but interestingly, crossing them with Hdac9 KO mice, which are also of normal size, decreased long bone length via unknown mechanisms. These results suggest some level of redundancy amongst these Hdacs (19).

HDAC5 was identified as a gene locus affecting bone mineral density (151), and another study showed that mutation in miR2861 prevented its binding to the HDAC5 UTR and increased HDAC5 expression in juveniles with low bone mass (101). HDAC5 can deacetylate multiple lysines in RUNX2 and promote its degradation via Smurf1 (74). RUNX2 levels were suppressed in the juvenile osteoporosis patients, supporting a role for HDAC5 in bone mineral density (101).

c) hdac6. The class IIb Hdac, Hdac6, is intriguing because existing evidence suggests that it has unique roles in endochondral ossification. Hdac6 is best known for its cytoplasmic functions as a tubulin deacetylase and a component of ubiquitin and autophagic complexes (113), but experiments in osteoblasts indicate that it rapidly transits through the nucleus where it can interact with Runx2 (193). Germline deletion of murine Hdac6 modestly increased cancellous bone density (215). The mechanisms contributing to this phenotype are not well understood but are likely to include cytoskeletal changes in osteoclasts and bone resorption (see sect. V). Recent discoveries linking HDAC6 to chondrodysplasia (OMIM no. 300683) should revive interest in this molecule (167) (Table 2).

d) hdac7. Hdac7 is another class IIa Hdac that plays crucial roles in endochondral ossification by suppressing chondrocyte proliferation. Deletion of Hdac7 in all tissues causes embryonic lethality due to vascular dilation and blood vessel ruptures (20). Deletion of Hdac7 in Col2-Cre expressing chondrocytes produced few viable animals, suggesting that Hdac7 is also essential for proper endochondral ossification in utero (10). Postnatal deletion of Hdac7 in the same populations of chondrocytes led to an expansion of the proliferative zone, narrowing of the hypertrophic zone, reduced femur lengths, and decreased trabecular bone density (10). Canonical β-catenin signaling was enhanced in the absence of Hdac7. Cytokines that promote chondrocyte proliferation (insulin and IGF-I) also induced the translocation of Hdac7 from the nucleus to the cytoplasm in wild-type chondrocytes. These results demonstrate that reducing Hdac7 levels may stimulate chondrocyte expansion and cartilage regeneration, but that Hdac7 is necessary for endochondral ossification.

e) summary of class ii hdacs in bone formation. Class II Hdacs are critical regulators of endochondral ossification due to their ability to integrate gene transcription with responses to growth factors and cytokines promoting cartilage and bone formation (Figure 3). Most class II Hdacs bind and regulate the activity of the Mef2 proteins, although it is likely that they control other transcription factors as well. Unlike the class I Hdacs, germline deletion of most class II Hdacs does not cause embryonic lethality, indicating some level of functional redundancy. Class II Hdacs have low intrinsic deacetylase activity and may work through functional complexes involving a class I Hdac. As class II Hdacs shuttle between the nucleus and the cytoplasm, it will be interesting to determine if nontranscriptional functions contribute to endochondral ossification. It will also be important to specifically study the skeletons of class II Hdac KO mice by imaging and histomorphometry because phenotypes may not be readily apparent. Indeed, a recent study identified osteopenia in Hdac9 KO mice (77). This phenotype is discussed in section V because the bone mass alterations were most influenced by increased osteoclastogenesis, although bone formation was also reduced.

3. Class III Hdacs and endochondral ossification

a) sirt1. Several studies have demonstrated that Sirt1 is essential for endochondral ossification in female mice. Germline Sirt1 deficiency increased p53-mediated apoptosis and produced severe developmental defects (24), including shorter stature, craniofacial abnormalities, increased cartilage apoptosis, and reduced endochondral ossification and cortical thickness (51, 119). Trabecular bone loss in Sirt1^+/− mice corresponded with fewer osteoblasts, more osteoclasts, increased Sost expression, and greater marrow adiposity (30). Conditional deletion of Sirt1 in committed osteoblasts with the 2.3Col1a1-Cre transgene increased NF-κB activity, which suppressed osteoblast differentiation and decreased trabecular bone mass in female mice (47, 161). Paradoxically, deletion of Sirt1 at earlier stages of osteoblastogenesis with Osx1-Cre and Prx1-Cre, or an inducible CreERT2, did not affect cancellous bone density, but decreased cortical bone thickness in female mice (71, 166). These studies found that Sirt1 promotes cortical bone formation by promoting β-catenin accumulation in the nucleus through deacetylation of K49 and K345 (166) and preventing sequestration by FoxO transcription factors (71). Sirt1 also regulated production of sclerostin by altering histone 3 acetylation of the Sost promoter (1, 30). Sirt1 deletion in chondrocyte lineages did not cause any developmental delays, but made joints more susceptible to osteoarthritis (112). The reason(s) why early deletion of Sirt1 in mesenchymal progenitor cells (and their progeny) causes cortical but not trabecular bone loss, while deletion in committed osteoblasts causes cancellous bone loss is unclear. One possible explanation is that the 2.3Col1a1-Cre transgene is transiently expressed in the early embryo (104) and could prevent skeletal development through other physiological mechanisms.

b) sirt6. Sirt6 facilitates endochondral ossification by controlling chondrocyte proliferation and differentiation. Sirt6 KO mice were born at Mendelian frequencies, but displayed growth retardation shortly after birth, failed to thrive, and died around 3.5 wk of age due to degeneration of multiple organs and genomic instability (125). The Sirt6 KO mice had a 30% reduction in bone mineral density and severe spine curvature. Growth plate chondrocyte proliferation and differentiation were impaired in the Sirt6-deficient animals and linked to reduced Ihh expression and increased cellular senescence (143).

c) summary of class iii hdacs in bone. Thus far, Sirt1 and Sirt6 are the only class III Hdacs that have been specifically shown to have necessary roles in bone. Sirt3 KO mice were found to have normal bone mineral density (105). Interestingly, current information on the International Mouse Phenotyping Consortium's (IMPC) website (http://www.mousephenotype.org/data/genes/) indicates that Sirt2 KO mice may have increased bone mass even though they appear phenotypically normal (160). While the bone cell autonomous roles of Sirt1 have been extensively analyzed genetically (Table 2), much work remains on the other Sirts. Fortunately, mice carrying mutated alleles are available for all of the Sirt genes (Table 2) (62, 105, 160, 182); thus focused studies may reveal bone phenotypes that are not readily apparent.

IV. HDACS IN OSTEOCLASTS AND BONE RESORPTION

Osteoclasts are large, multinucleated cells derived from hematopoietic stem cells and monocyte precursors. These specialized bone-resorbing cells play an essential role in endochondral ossification as they remove calcified cartilage at the growth plate during the initial formation of the mineralized bone (bone modeling), facilitate bone turnover in response to damage or physiological needs (bone remodeling), and clear fracture calluses to aid healing. Osteoclasts remove bone by secreting acids and proteolytic enzymes, such as cathepsin K, into the area of resorption. Osteoclasts also secrete factors that promote the recruitment of preosteoblasts to the resorption site (140, 171). Normally, the amount of bone resorbed by osteoclasts is precisely offset by new bone formation. However, at menopause and in conditions of increased bone resorption (e.g., osteoporosis, Paget's disease, rheumatoid arthritis, and tumor-induced osteolysis), osteoclast-mediated resorption outpaces new bone deposition by osteoblasts and causes net bone loss. The roles of Hdacs in osteoclasts have not been as widely studied as they have been in osteoblasts; however, this is an emerging area of scientific interest.

A. Hdac7 in Osteoclasts

Suppression of the class II Hdac, Hdac7, facilitated osteoclastogenesis and increased osteoclast size (142, 170). This observation was attributed to the ability of Hdac7 to inhibit the transcriptional activity of Mitf, a transcription factor required for Rankl-induced gene expression. The NH₂ terminus of Hdac7 was sufficient to bind and inhibit Mitf activity, indicating that the deacetylase domain of Hdac7 was not necessary for this repression (170). In a separate report, conditional deletion of Hdac7 using a LysM-Cre promoted the proliferation of osteoclast precursor cells, and increased Rankl-induced osteoclast differentiation (76). Hdac7 suppressed β-catenin activity and cyclin D1 expression in marrow progenitor populations and differentiated osteoclasts (76).

B. Hdac9

Hdac9 KO mice were originally reported to have no gross phenotypic changes in bone structure up to 3 mo of age (212); however, more focused analyses revealed a large (∼45%) reduction in bone volume in both cancellous and cortical bone (77). Hdac9 levels naturally declined during osteoclastogenesis, and accordingly, osteoclast numbers and bone resorption indexes were elevated in Hdac9 KO mice. Bone formation indexes were also reduced in the Hdac9 KO mice; however, the major driver of the osteopenic phenotype was increased osteoclastogenesis. Bone marrow transplantation experiments showed that Hdac9 depletion in osteoclasts affected bone resorption in a cell autonomous fashion. Thus transplantation of Hdac9 KO bone marrow into wild-type mice was sufficient to cause osteopenia. Conversely, wild-type bone marrow transplantation rescued the osteopenic phenotype of the Hdac9 KO mice. Hdac9 was found to participate in a negative regulatory circuit with Pparγ and Rankl, as both suppressed Hdac9 mRNA levels, and Hdac9 partnered with NCoR and Smrt to suppress Pparγ-dependent transcription. Overall, these studies show that Hdac9 expression suppresses osteoclastogenesis.

C. Other Hdacs in Osteoclasts

Several other Hdacs have been studied in osteoclastic models. Pham et al. (142) demonstrated that shRNA-mediated knockdown of Hdac3 in bone marrow-derived osteoclasts or RAW246.7 cells inhibited osteoclast formation and decreased osteoclast size in response to RANKL. Expression of osteoclast-specific genes, including NFATc1, Ctsk, and DC-Stamp, was also suppressed by Hdac3 knockdown. Meanwhile, Hdac6, a class IIb Hdac and tubulin deacetylase, destabilizes the osteoclast cytoskeleton and inhibits osteoclast migration and podosome formation (40, 146). The class III Hdac Sirt1 also inhibits osteoclast differentiation as conditional deletion of Sirt1 in osteoclasts increased NF-κB activity, osteoclastogenesis, and bone loss in female mice (47, 161).

V. HDAC INHIBITORS

The expanding appreciation that epigenetic events contribute to varied phenotypes and disease etiologies has produced great interest in natural and synthetic compounds that modify the activity of Hdacs and other epigenetic regulators. HDIs are generally classified as short-chain fatty acids, hydroxamic acids, cyclic peptides, or benzamides. Although they have diverse structures and variable specificities, several HDIs with broad HDAC specificity have received Food and Drug Administration approval for clinical use in the United States. The oldest is valproic acid (divalproex sodium; Depakote; valproate), which has been used as an anticonvulsive therapy in epilepsy and a mood stabilizing therapy for bipolar disorder for decades and long before the cloning of Hdacs. Vorinostat (SAHA, Zolinza) and romidepsin (depsipeptide, Istodax) were first approved for treatment of cutaneous T-cell lymphoma, but are now routinely used in combination with other chemotherapeutic agents to treat a variety of hematopoietic and solid tumors. Romidepsin and belinostat (PXD101) are approved for the treatment of peripheral T-cell lymphomas. These and other HDIs, including panobinostat (LBH589), entinostat (MS-275), and givinostat (ITI2357), are in various clinical trials as potential treatments for many diseases including cancer, HIV infection, inflammatory conditions, neurodegeneration, and diabetes (27, 42, 48, 95, 116, 136, 169). The effectiveness of HDIs against cancer is attributed to many mechanisms, including increased acetylation of histones and other proteins, and consequent induction of tumor suppressor genes, stimulation of cell cycle arrest and apoptosis, and inhibition of DNA repair. HDI-induced chromatin relaxation may also make cancerous cells more susceptible to DNA damage from radiation or chemotherapies.

The ubiquitous expression patterns of Hdacs raise important concerns about the safety of HDI-based therapies especially for chronic diseases. Surprisingly, HDIs are fairly well tolerated by cancer patients, possibly because HDIs have short half-lives in serum and their effects can be reversed by acetyltransferases in normal cells that, contrary to tumor cells, are not addicted to oncogenes and have stable genomes (48). Nearly all patients receiving HDIs experience nausea, thrombocytopenia, fatigue, diarrhea, and dehydration, but all of these symptoms are temporary and cease after therapy ends (reviewed in Ref. 95). Some patients experience confusion, vertigo, and other neuropsychological problems or rare pulmonary side effects indicated by coughing and dyspnea. More severe problems, including gastrointestinal toxicity and cardiac conduction abnormalities, have been reported but are rare. Thus far, the effects of HDI therapy on bone density have not been assessed in a prospective randomized trial. Three independent retrospective studies on long-term valproic acid therapy produced varied results, with bone mineral density either decreased or unchanged, and markers of bone resorption increased, decreased, or unchanged (89, 156, 179). However, epidemiological studies indicate that long-term use of HDIs increases fracture risk (a widely accepted surrogate for reduced bone quality) in children and adults and is teratogenic to fetuses (9, 53, 59, 134, 156, 159, 163). Children exposed to the Hdac inhibitor valproic acid in utero developed craniosynostosis, axial skeletal defects, and limb malformations (94, 135, 162, 181). These phenotypes demonstrate the crucial role of Hdac activity in orchestrating human skeletal development. Animal studies have begun to shed light into how HDIs affect the developing and established skeleton.

A. Hdac Inhibitors and the Fetal Rodent Skeleton

Preclinical studies showed that HDIs cause fetal toxicity, growth retardation, and skeletal defects (127, 138, 181, 196). The negative effects of Hdac inhibitors on axial skeletal formation and limb development may in part be due to altered commitment of progenitors to the chondrocyte lineage, as valproate repressed chondrogenesis in ex vivo cultured murine limbs and reduced expression of key endochondral ossification genes, including Sox5, Sox6, Sox9, and type 2 collagen (138). Valproate also reduced expression of type 1 collagen and osteonectin in human fetal osteoblastic precursor hFOB1.19 cells (69). The precise target of the broad-acting HDIs in vivo is unclear; however, the phenotypes of genetically altered mice (reviewed above and in Table 1) indicate that inhibition of any single Hdac or multiple Hdacs may contribute to improper skeletal formation and long bone development in utero.

B. Hdac Inhibitors and the Mature Rodent Skeleton

A number of studies have measured the effects of HDIs on mature rodent skeletons. Depending on dosage and treatment regimen, in vivo treatment with valproate (159) or vorinostat (115, 145) caused bone loss in mouse and rat models. In a study designed to study the effectiveness of vorinostat at reducing tumor burden and associated osteolysis in immunocompromised mice, daily HDI treatment (100 mg/kg ip) for 4 wk prevented bone resorption associated with early stages of tumor growth, but it also caused significant cancellous bone loss in the contralateral non-tumor injected limbs and increased overall osteolysis despite decreasing tumor burden (145). Furthermore, vorinostat increased osteoclast differentiation in immunocompromised mice lacking tumors (145). In an identically designed study, C57Bl/6 mice treated with vorinostat had reduced cancellous bone mineral density, but normal cortical bone density and osteoclast activity (115). Paradoxically, while osteoblast numbers were reduced in HDI-treated C57Bl/6 bones, osteoblast activity (measured by mineral apposition rate) was increased. In vitro studies showed that immature osteoblasts that were not yet making a bone matrix were highly susceptible to DNA damage and cell cycle arrest when cultured in the presence of vorinostat, whereas mature and committed osteoblasts were resistant to the negative effects of HDIs on the cell cycle and possibly produced more extracellular matrix (115, 200).

Several groups found that HDIs like vorinostat, MS-275, and trichostatin (TSA) promote cell cycle arrest, senescence, and apoptosis in stromal cells derived from bone marrow or adipose tissues (41, 43, 98, 111, 115). For example, treatment of human bone marrow-derived mesenchymal stem cells with vorinostat activated caspase 9 and apoptotic pathways and downregulated genes associated with cell stemness (41). Valproate caused G₂/M phase arrest of human mesenchymal stem cells (98). Although low concentrations were tolerated, high amounts of vorinostat caused apoptosis of human mesenchymal stromal cells (200). These effects can functionally impair the cells upon further differentiation. For example, pretreatment of human adipose-derived stromal cells with HDI prior to osteogenic differentiation decreased the amount of mineralized matrix formed (111), and murine bone marrow stromal cells exposed to vorinostat at early stages of osteogenic differentiation produce less matrix compared with vehicle-treated cells (115). Histone H4 acetylation patterns throughout the genome were correlated with changes in gene expression in vorinostat-treated MC3T3 osteoblasts, revealing a common mechanism of decreased signaling through the insulin/Akt/FoxO1 pathway in committed osteoblast precursors (44).

The effects of HDIs on bone resorption and osteoclast differentiation warrant further study because interest in their use as treatments for more cancers as well as chronic inflammatory and epigenetic diseases is growing. A body of evidence indicates that HDIs inhibit osteoclasts and joint destruction in animal models of chronic inflammatory diseases, including rheumatoid arthritis and periodontal disease (14). In three preclinical models of inflammatory arthritis, the Hdac inhibitor ITF2357 reduced joint inflammation and cartilage destruction (78). TSA also reduced inflammation and cartilage destruction in inflammatory arthritis (130). In vitro, HDIs reduced inflammatory responses by repressing expression of prostaglandin E₂ synthase-1 (mPGES-1) and prostaglandin E₂ production in response to interleukin-1 (18, 210).

A number of in vitro studies demonstrated that HDIs directly suppress early osteoclastogenesis. Romidepsin (FR901228, Isodax) induced expression of interferon-β, which decreased expression of two genes, c-Fos and Socs-3, required for osteoclastogenesis (129). MS-275 (Entinostat, SNDX-275) also suppressed osteoclast differentiation by repressing c-Fos expression (87). TSA and sodium butyrate repressed osteoclast formation by blocking the induction of NFATc1 and c-Fos in response to Rankl signaling and inducing the expression of two anti-osteoclastogenic factors, CCAAT enhancer binding protein (C/EBP)-β, and mitogen-activated protein kinase phosphatase (MKP)-1 (148, 194). TSA promoted apoptosis of mature multinucleated osteoclasts by increasing the expression of Cdkn1a/p21^WAF (206). Studies on human osteoclasts generated from peripheral blood mononuclear cells indicate that simultaneous inhibition of both class I and II Hdacs may be necessary for osteoclastic suppression by Hdac inhibitors (14). Moreover, Hdac3 and Hdac7 may have distinct if not opposing effects on osteoclastogenesis in vitro (76, 142).

VI. SIRTUIN ACTIVATORS

Because Sirt1 expression lengthens the lifespan of many lower organisms and improves the healthspan of vertebrates, there is great interest in identifying Sirt activating compounds. A number of polyphenols, including resveratrol, are natural Sirt1 activators (39), and numerous studies have documented the benefits of resveratrol on preserving and enhancing bone mass in mice (reviewed in Ref. 177). Resveratrol's antioxidant, proestrogenic, and anti-inflammatory activities both promote bone formation and prevent bone resorption. Several synthetic Sirtuin activating compounds have been tested for their ability to protect bone mass in animal models. SRT2104 increased cancellous bone volume, but not cortical bone mass in male mice subjected to hindlimb suspension (119). In vitro, SRT2104 increased osteoblastic mineralization and inhibited osteoclast formation. Another SIRT activator, SRT3025, increased vertebral bone mass, femoral biomechanical properties, and cortical periosteal mineralizing surface in ovariectomized mice, while suppressing sclerostin expression and increasing β-catenin signaling in osteocytes (3).

VII. HDACS AND ORTHOPEDIC APPLICATIONS

Early in vitro studies with HDIs created a great deal of excitement in the bone field because they pointed to anabolic properties. Indeed, contrary to the studies described above, multiple laboratories throughout the world have reported that HDIs promote maturation of osteoblastic cell lines and calvarial osteoblast cultures in part by enhancing the activity of Runx2 and other transcription factors, as well as by activating BMP signaling pathways and preventing adipogenesis (reviewed in Ref. 116). But as summarized above, in vivo inactivation of Hdacs is generally unfavorable for bone development and maintenance because mesenchymal progenitors are particularly sensitive to Hdac loss of function. While sustained systemic administration of current broad acting HDIs may have adverse effects on the skeleton, these drugs could have some utility for bone tissue engineering (33) and orthopedic applications, such as heterotopic ossification, fracture repair, osteoarthritis, and treatment of bone tumors.

A. Hdacs and Bone Regeneration

Bone fracture healing and repair occur when progenitor cells from the surrounding tissues (e.g., periosteum, marrow, or vasculature) infiltrate the wound site and regenerate new tissue. Rigidly stabilized fractures (e.g., externally fixed breaks, calvarial or single cortex defects) that permit little to no motion of the healing site will heal by intramembranous ossification. Localized HDI treatment can augment these healing processes. For example, addition of the cyclic depsipeptide largazole to a macroporous biphasic calcium phosphate scaffold promoted greater bone formation than scaffold alone in rabbit calvarial bone defects. This was attributed to the induction of Runx2 and BMPs (99). Similarly, treatment of murine critical sized calvarial defects with MS-275 promoted bone formation that exceeded controls and was comparable to the enhanced healing caused by BMPs (86). Thus it has been suggested that HDIs could be developed into adjuvant therapies to promote faster skeletal healing (67). However, given the negative effects of HDI administration on bone mass in vivo and on osteoprogenitor survival in vitro, this concept requires further investigation.

B. Hdacs and Osteoarthritis

Hdacs may also regulate osteoarthritic phenotypes. Osteoarthritis (OA) is the most common form of arthritis and a leading cause of disability that poses enormous social and economic burdens in developed countries. Whereas articular chondrocytes are normally maintained in a quiescent or low metabolic state, joint injuries and repetitive use can induce damage and initiate the OA disease process. OA is characterized by thinning of the articular cartilage, induction of chondrocyte proliferation (cloning), aberrant chondrocyte hypertrophy, enzymatic degradation of the cartilaginous matrix, increased matrix calcification, thickening of subchondral bone, and/or inflammation of the joint. Since Hdacs regulate many of the processes involved in OA progression (e.g., inflammation, proliferation, hypertrophy, and matrix synthesis and degradation), their role in controlling OA progression is of interest.

Multiple HDACs are expressed in human articular cartilage (FIGURE 1, D and E), and several were found expressed at higher levels in diseased cartilage from OA joints. HDAC1 and HDAC2 were enriched in chondrocytes obtained from OA cartilage and ectopic expression of these enzymes decreased matrix gene expression (66). Articular cartilage from osteoarthritis patients also has elevated levels of HDAC7 mRNA compared with cartilage obtained from healthy controls (65). HDAC7 was more prevalent in the middle and deep zones of the articular cartilage of OA patients compared with healthy controls. SIRT1 was also elevated and associated with an increased Sox9-driven type 2 collagen expression in human chondrocytes derived from osteoarthritic patients (46); however, in osteoblasts from osteoarthritic joints, SIRT1 expression was reduced and correlated with increased Sost expression (1). Subsequent studies demonstrated that SIRT1 suppression reduced articular chondrocyte matrix gene expression and increased MMP expression and chondrocyte apoptosis (52).

HDIs alter the integrity of the cartilaginous extracellular matrix and OA disease progression. Using a surgical model of OA in the mouse, Culley et al. (32) found that systemic administration of TSA diminished cartilage damage caused by cytokine-induced MMP expression in human articular chondrocytes. Similar results were obtained using a rabbit model of OA where TSA protected cartilage and reduced MMP and cathepsin expression (22, 23, 68, 188); however, TSA also suppressed expression of two major cartilage extracellular components, aggrecan and collagen type 2 (68). Young et al. (208) found that TSA and sodium butyrate inhibited expression of MMPs and aggrecan-degrading enzymes such as ADAMTS5 in human chondrocytes, supporting a potential role for the protective effects of HDIs (208). The effects of Hdac inhibition on MMP production may be mediated through decreased activation of the MAPK pathway as SAHA, TSA, and MS-275 decreased ERK activity in vitro (153, 216) (Figure 3). Thus HDIs have the potential to protect articular cartilage by preventing matrix degradation; however, a role in promoting regeneration in unlikely as HDIs reduced MSC proliferation as well as differentiation into defined lineages, including chondrocytes (98).

C. Hdacs and Bone Tumors

The antiproliferative and chemo/radio-sensitizing properties of HDIs have made them attractive epigenetic therapies for cancers. The relatively modest side effects of these drugs have led to their widespread testing in many clinical trials. Xenograft models of prostate and breast cancer bone metastases indicate that HDIs can access bone marrow and reduce tumor burden (145). The in vivo effectiveness of HDIs against other bone tumors, such as osteosarcoma and myeloma, transplanted into mice is also an active area of investigation (84, 197). The ability of HDIs to kill cells from human and canine osteosarcomas, chondrosarcomas, and other primary bone tumors has been well documented by a number of in vitro studies (15, 175, 176, 197, 198, 201). Sarcoma patients have been included in a number of ongoing clinical trials with HDIs, and time will tell if these cancers are susceptible to HDIs (17, 154). Local application of HDIs may be an effective therapeutic approach for some primary tumors. A recent study demonstrated that HDIs loaded into bone cement scaffolds were released over time in culture and retained their activity (176). But because primary bone tumors are most lethal when they metastasize to the other organs, the effects of HDIs on bone cancer metastases are of interest. One study showed that oral administration of MS-275 caused regression of osteosarcoma lung metastases in a mouse model (149). Thus a number of lines of evidence indicate that HDIs have potential to treat at least some forms of primary and secondary bone tumors.

VIII. SUMMARY AND FUTURE DIRECTIONS

Hdacs have emerged as crucial regulators of both intramembranous and endochondral bone formation. The Zn-dependent Hdacs, especially those in classes I and IIa, appear to be essential for proper skeletal development as their in vivo inhibition generates multiple skeletal defects. This reality is sometimes in conflict with and confused by results from in vitro studies where Hdac inhibition promoted terminal differentiation of committed osteoblasts, but it aligns with data showing that HDIs suppress osteoblastic progenitors. Together, these results suggest that current HDIs may be effective for bone tissue engineering where cells can be coaxed past the proliferative phase before exposure to HDIs, and for some orthopedic applications, such as heterotopic ossification and primary bone tumor inhibition, where local suppression of progenitor cells may be advantageous. However, HDIs may not be effective therapies for systemic and/or aging-related diseases, such as osteoporosis, where stem/progenitor cell pools are essential for new bone formation and where osteoclastogenesis could be enhanced by HDIs. Existing evidence suggests that specific inhibitors of HDAC6 and activators of SIRTs may promote bone formation and/or prevent bone loss, but more studies are needed to understand the mechanisms behind these phenotypes. A challenge for drug studies is deciphering the specificity of HDIs in vivo. Although some HDIs are described or marketed as being specific for certain Hdacs, these conclusions are often based on studies with immunoprecipitated complexes, which may contain multiple Hdacs in mega-Dalton, multiprotein co-repressor complexes. Some evidence indicates the co-repressors can influence HDI specificity for an Hdac (4, 192). It is also possible that the absence of activity towards a particular purified protein in vitro may not represent its activity in vivo or in situ. The rapid emergence of epigenetic inhibitors as potential therapies for a variety of diseases and the advancement of bone imaging technologies should provide opportunities for prospective studies on how Hdac inhibitors, Sirt activators, and modulators of other epigenetic enzymes such as HATs and methyltransferases affect bone quantity and quality in human subjects.

Despite tremendous progress, there is still much to be learned about how Hdacs and Sirts control bone formation and regeneration through mouse genetics. Within the next decade, it is likely that all Hdacs (especially the most highly expressed Hdacs, Hdac1 and Sirt2) will have been individually deleted in bone cells of both the osteoblastic and osteoclastic lineages and that both their unique as well as redundant roles in bone development will be defined. The consequences of altering Hdac activity on chromatin structure and epigenetic memory in osteocytes (the longest living bone cells) and articular chondrocytes will also be of interest. As the crucial roles of Hdacs become more broadly recognized, it is likely that more studies will examine how various stimuli (e.g., Rankl and PTH), mechanical forces, aging, and altered physiological states affect Hdac expression and activity in osteoblasts, osteocytes, and osteoclasts (77, 100). More challenging studies will involve determining their role in adult skeletons and aging. For instance, one interesting question is whether or not deletion of Hdac4 will be as destructive to adult skeletons as it is to developing skeletons given its relatively low expression in adult bone and articular cartilage. Future work will also undoubtedly examine how Hdacs and Sirts expressed in bone regulate (and are regulated by) physiological processes, such as energy metabolism and sympathetic tone. Two recent studies on Hdac3 and Hdac4 CKO offer provocative evidence that Hdacs can control the expression of osteoblastic hormones that regulate systemic metabolism (108, 117). Finally, the mechanisms of Hdac action still need to be defined. A fundamental question is what are the important substrates of Hdacs besides histones. Some studies indicate that deacetylase activity and domains are dispensable for Hdac-mediated suppression and phenotypes, but these mechanisms are not fully understood. Addressing these and other questions are crucial for fully realizing the functions of Hdacs in the skeleton and understanding the advantages and disadvantages of HDI therapies on the skeleton.

GRANTS

The authors receive support from the Mayo Foundation for Education and Research and National Institutes of Health (NIH) Grants AR048069, AR056950, AR060140, AR068103, AR065397, DE20194, and DK050456. The contents of the article are the sole responsibility of the authors and do not represent official views of the Mayo Clinic or NIH.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.