Recent years have witnessed an emergence of a new class of therapeutic agents, termed histone deacetylase 6 (HDAC6) inhibitors. HDAC6 is one isoform of a family of HDAC enzymes that catalyse the removal of functional acetyl groups from proteins. It stands out from its cousins in almost exclusively deacetylating cytoplasmic proteins, in exerting deacetylation-independent effects and in the success that has been achieved in developing relatively isoform-specific inhibitors of its enzymatic action that have reached clinical trial. HDAC6 plays a pivotal role in the removal of misfolded proteins and it is this role that has been most successfully targeted to date. HDAC6 inhibitors are being investigated for use in combination with proteasome inhibitors for the treatment of lymphoid malignancies, whereby HDAC6-dependent protein disposal currently limits the cytotoxic effectiveness of the latter. Similarly, numerous recent studies have linked altered HDAC6 activity to the pathogenesis of neurodegenerative diseases that are characterized by misfolded protein accumulation. It seems likely though that the function of HDAC6 is not limited to malignancy and neurodegeneration, the deacetylase being implicated in a number of other cellular processes and diseases including in cardiovascular disease, inflammation, renal fibrosis and cystogenesis. Here, we review the unique features of HDAC6 that make it so appealing as a drug target and its currently understood role in health and disease. Whether HDAC6 inhibition will ultimately find a clinical niche in the treatment of malignancy or prevalent complex chronic diseases remains to be determined.
Histone deacetylases (HDACs) are a family of 18 enzymes that play diverse roles in mammalian cell homeostasis and in tumour growth. For over half a decade, ‘broad-spectrum’ HDAC inhibitors have been available clinically for the treatment of certain uncommon malignancies. More recently, drug discovery efforts have focused on the development of isoform-specific inhibitors, the HDAC6 isoform standing apart from some of its peers in its relative ‘druggability’ and in its unique functions within the cell. Clinical trials are currently evaluating the effectiveness of the HDAC6 inhibitors ricolinostat (ACY-1215; formerly rocilinostat) and ACY-241 for the treatment of lymphoid malignancies. However, the potential therapeutic scope of HDAC6 inhibitors extends well beyond haematological malignancy alone. Here, we ask the question: what makes the HDAC6 isoform so special and where will HDAC6 inhibitors find their clinical niche?
THE MAGNITUDE AND EXTENT OF PROTEIN LYSINE ACETYLATION
Post-translational modification is a strategy employed by all cells enabling them to rapidly control protein behaviour and adjust and respond to changing shifts in the external and internal microenvironment. The importance of post-translational modification in cellular function is most commonly recognized in the pervasive influence of protein phosphorylation and dephosphorylation, regulated by kinase and phosphatase enzymes respectively. There are approximately 6600 phosphorylation sites on 2200 proteins  and numerous kinase inhibitors are employed in the clinic. However, phosphorylation is not the only chemical modification that is under enzymatic regulation. Protein function, for instance, may also be influenced by the addition or removal of functional acetyl groups.
There are two major types of protein acetylation process. Approximately 85% of proteins are Nα-terminally acetylated . Even though Nα-terminal acetylation is widespread and conserved, its role in biology remains uncertain. The second type of acetylation involves the specific modification of ε-amino groups on lysine residues of proteins. This process is regulated by groups of enzymes called histone acetyltransferases (HATs; which add an acetyl group) and HDACs (which remove an acetyl group). The enzymes are so-named because they were first recognized for their ability to post-translationally modify histones, the proteins that DNA coils around within the cell nucleus to form chromatin [3–5]. However, it is now appreciated that numerous other nuclear and non-nuclear proteins can also be acetylated; a landmark proteomic study identified over 3600 acetylation sites in approximately 1750 proteins , the size of the cellular acetylome thus being comparable to that of the phosphoproteome.
THE 18 MEMBERS OF THE HDAC FAMILY
Most HDACs reside in the nucleus, often cooperatively as part of high molecular mass complexes, where they function to deacetylate lysine residues within the long amino acid tails of histone proteins, contributing to the regulation of gene transcription. Although it is not easy to observe in vivo, histone acetylation may promote chromatin unfolding by neutralizing the basic charge of lysine residues, allowing access to transcriptional activators and thus promoting transcriptional activation . Histone acetylation may also regulate transcription by serving as a recognition point for bromodomains , modules of approximately 110 amino acids in length that are present in many proteins that associate with chromatin . Deacetylation by HDACs opposes these effects. Mammals possess at least 18 HDAC isoforms that are grouped into four classes according to their homology to yeast deacetylases: Class I has four isoforms (HDAC1, 2, 3 and 8); Class II contains six isoforms and is sub-divided into Class IIa (HDAC4, 5, 7 and 9) and IIb (HDAC6 and 10) and Class IV has just one isoform (HDAC11). Each of these 11 isoforms is considered a zinc-dependent HDAC, because it possesses a zinc-binding domain at its active site, which is necessary for enzymatic activity. ‘Broad-spectrum’ HDAC inhibitors target the zinc-dependent HDACs and, among them, primarily Class I HDACs . The remaining HDAC class, Class III, contains seven HDAC isoforms that are referred to as the sirtuins (Sirt1–7). They are distinct in exerting their enzymatic activity through NAD+-dependent mechanisms. The enzymatic activity of HDACs likely extends beyond the removal of acetyl groups, however. Recent evidence indicates that histone lysine residues can also be competitively modified by other acyl groups (i.e. propionyl, butyryl, 2-hydroxyisobutyryl, crotonyl, malonyl, succinyl and glutaryl) [10–15], which may also be substrates for certain HDAC isoforms. For instance, HDAC3 can decrotonylate histone peptides , although the major decrotonylase in living cells appears to be Sirt3 , whereas Sirt2 was shown to remove a propionyl group from histone proteins in one leukaemia cell line .
HDAC6: A HISTONE DEACETYLASE APART FROM THE REST
Unlike most of the HDAC isoforms, HDAC6 is not localized within the nucleus, but is principally found in the cytoplasm. HDAC6 was first discovered in 1999 for its homology with the Saccharomyces cerevisiae histone deacetylase, HDAC1 [19,20]. Its encoding gene is localized to the sub-band border of chromosome Xp11.22-23 in humans . The structural arrangement of human HDAC6 is shown in Figure 1. The enzyme is made up of 1215 amino acids and has a predicted molecular mass of 131 kDa. HDAC6 possesses two nuclear export signals (NES). The most N-terminally located NES is primarily responsible for the cytoplasmic location of the deacetylase . Human HDAC6 also contains a Ser-Glu-containing tetrapeptide (SE14) motif which helps retain the enzyme within the cytoplasm . At the N-terminus of the enzyme there is also a nuclear localization signal (NLS) that enables the deacetylase to shuttle between the nucleus and the cytoplasm. Acetylation of HDAC6 at the NLS site promotes retention of the enzyme within the cytoplasm [24,25] and can also affect its deacetylase activity . HDAC6 is unique among the HDAC isoforms in possessing two homologous catalytic domains (termed DD1 and DD2; located at the N-terminus and centrally), both of which contribute to its enzymatic activity. Although the enzyme possesses two catalytic domains, only DD2 (the domain with α-tubulin deacetylase activity) has been well-characterized. Indeed, all small molecule HDAC6 inhibitors that have been developed to date have been shown to target this domain. This could result in discrepancies between experimental observations made with pharmacological inhibitors and genetic knockout approaches, the latter targeting both domains. Finally, HDAC6 also possesses a dynein motor-binding domain  and at the C-terminus there is a unique ubiquitin-binding zinc-finger domain [termed the ZnF-UBP (or BUZ) domain]  that enable the protein to exert non-enzymatic effects on cellular function. According to the Human Protein Atlas, the sites of highest HDAC6 protein expression are the kidney (renal tubules) and the testis (cells in the seminiferous ducts) .
Functional domains of the human HDAC6 protein
HDAC6 SUBSTRATES AND INTERACTING PARTNERS
Given the predominantly cytoplasmic distribution of the protein, it is unsurprising that most of the recognized physiological substrates of HDAC6 are cytoplasmic proteins. The first protein that was discovered as being deacetylated by HDAC6 was α-tubulin, which dimerizes with β-tubulin to form microtubules, dynamic components of the cellular cytoskeleton. In 2002/2003, several reports independently demonstrated that HDAC6 interacts with β-tubulin and deacetylates lysine residue 40 on α-tubulin [29–31]. α-Tubulin acetylation is induced by the acetyltransferase MEC-17, also referred to as α-tubulin acetyltransferase 1 (αTAT1) [32,33]. It is noteworthy, however, that HDAC6 is not the only α-tubulin deacetylase, the Class III HDAC, Sirt2 also fulfilling this role , where it could potentially compensate for loss of HDAC6 or a decrease in HDAC6 activity . Other proteins that have subsequently been identified as being substrates for HDAC6 include the cortical actin-binding protein cortactin , the chaperone protein heat shock protein 90 (HSP90) , the redox regulatory protein peroxiredoxin  and the phospho-binding protein 14-3-3ζ . Whereas these are the best known, or the best studied, there are likely to be a number of other HDAC6 protein substrates. For instance, taking a quantitative proteomic approach, investigators recently identified 107 proteins with elevated acetylation levels in the livers of mice genetically deficient in HDAC6, three of which were verified as interacting with HDAC6: myosin heavy chain 9 (MYH9), heat shock cognate protein 70 (Hsc70) and dnaJ homologue subfamily A member 1 (DNAJA1) . Separately, although HDAC6 primarily deacetylates cytoplasmic proteins, it can alter gene transcription by interacting with a number of transcriptional regulators including runt-related transcription factor 2 (RUNX2) , nuclear factor κ-light-chain-enhancer of activated B cells (NF-κB)  and the nuclear receptor corepressor ligand-dependent corepressor (LcoR) [43,44]. These transcriptional regulators are interacting partners of HDAC6 but not necessarily enzymatic substrates.
HDAC6 KNOCKOUT MICE HAVE A GENERALLY BENIGN PHENOTYPE
The genetic knockout of a number of HDAC isoforms has been associated with embryonic or perinatal lethality in mice including the genetic knockout of HDAC1 , HDAC2 , HDAC3 , HDAC4 , HDAC7  and HDAC8 . By contrast, HDAC6 knockout mice are viable and develop normally , indicating that HDAC6 is dispensable for the ordinary functioning of most organs. In a meticulous survey of the phenotypic characteristics of HDAC6 knockout mice, Zhang et al.  observed only minor abnormalities including a sub-optimal immune response and a small increase in cancellous bone mineral density. Importantly, whereas α-tubulin was hyperacetylated in HDAC6 knockout mice, this did not appear to affect either microtubule organization or stability, at least in embryonic fibroblasts . Similarly, more recent direct tests also indicate that α-tubulin acetylation has no effect on microtubule dynamics  or microtubule ultrastructure . Thus, whereas initial descriptions suggested that HDAC6 regulates microtubule stability , its role in this process (if any) is likely to be minor.
NON-ENZYMATIC ACTIONS OF HDAC6
In considering the rationale for the development of HDAC6 inhibitors and their therapeutic applicability, it is noteworthy that HDAC6 itself exerts both enzymatic and non-enzymatic actions on cellular function. In addition to its two catalytic domains, HDAC6 also possesses a ZnF-UBP domain  which enables it to function as a regulator of the ubiquitin and proteasome system (UPS) and thus the cellular response to protein misfolding [26,54]. Accordingly, whereas genetic knockout may be anticipated to affect both enzymatic and non-enzymatic functions of the protein, pharmacological inhibition may only affect the former.
The role of HDAC6 in the cellular response to protein misfolding
Normal protein function requires the proper spatial organization of proteins in three dimensions. However, up to 30% of all newly synthesized proteins are not folded correctly  and this is even more common in the disease state . These misfolded proteins need to be properly disposed of to prevent their interference with normal cell function. The conventional means by which a cell deals with protein misfolding is by tagging the proteins with a small protein called ubiquitin, a process mediated by ubiquitin ligase enzymes. This ubiquitin tag then acts as a signal for other ligases to attach additional ubiquitin molecules. The polyubiquitin chain is subsequently recognized by the 26S proteasome within the cell, which degrades the protein by proteolysis.
Under certain circumstances, for instance in certain neurodegenerative diseases, protein misfolding exceeds the capacity by which it can be handled by the UPS and this leads to an accumulation of misfolded proteins in aggregates, known as aggresomes, which are too large to be cleared by the proteasome . Within the cell, aggresome formation occurs at the centrosome, a perinuclear structure that also serves as a nidus for the accumulation of UPS components . Centrosomes are one of two microtubule organizing centres within a cell, the other being found at the basal body of cilia. In general, aggresomes serve a protective role by collecting together misfolded or damaged proteins that cannot be processed by the UPS . However, they may impair cell function through their space occupying effects that can affect cellular metabolism and intracellular transport. Aggresomes themselves are disposed of by macroautophagy (more commonly referred to as autophagy), a lysosomal-degradation pathway that takes up aggresomes into autophagosomes that subsequently fuse together with lysosomes promoting the enzymatic degradation and clearance of dysfunctional proteins.
How does HDAC6 regulate the efficient disposal of misfolded proteins? The ZnF-UBP domain at its C-terminus enables HDAC6 to bind to ubiquitinated proteins and the dynein motor-binding domain enables HDAC6 to bind to dynein. Dynein is a motor protein that uses ATP to migrate along microtubules generally towards the ‘minus-end’ (known as retrograde transport). Thus, the binding of HDAC6 to dynein enables the transport of the cellular cargo of misfolded proteins along microtubules into aggresomes . HDAC6 therefore functions as both a metaphorical and a structural bridge between the UPS and aggresome-autophagy pathway (Figure 2). With an apparently superior affinity for ubiquitin binding than other ubiquitin-binding proteins , and by simultaneously binding ubiquitin and dynein motors [26,54], HDAC6 is considered to favour the accumulation of misfolded proteins into aggresomes and decrease their clearance through the UPS . Interestingly, although the physical interactions between ubiquitinated proteins, HDAC6 and dynein motors are required for aggresome formation, so is the deacetylase activity of the enzyme, the reintroduction of deacetylase-deficient HDAC6 to HDAC6 knockout cells being unable to restore aggresome formation .
Role of HDAC6 in the cellular response to protein misfolding
The interaction between HDAC6 and other proteins helps to regulate the intracellular balance between protein degradation through the UPS and protein aggregation mediated by HDAC6 . For example, HDAC6 forms a complex with the ATP chaperone valosin-containing protein (VCP)/p97 , which is a critical regulator of the UPS  and VCP/p97, in turn, prevents the accumulation of polyubiquitinated protein aggregates by HDAC6 by dissociating HDAC6–ubiquitin complexes . Separately, HDAC6 forms a tri-complex with HSP90 and the transcription factor, heat shock transcription factor protein 1 (HSF1) . Upon sensing of ubiquitinated aggregates, HDAC6 dissociates from this complex allowing HSF1 activation and the subsequent induction of molecular chaperone heat shock genes . Further down the pathway of misfolded protein degradation, HDAC6 also functions to recruit and deacetylate cortactin, which is necessary for autophagosome–lysosome fusion  (Figure 2). Misfolded protein accumulation is a common feature of neurodegenerative diseases and a limitation of certain cancer therapies that exert their cytotoxic effects through inhibition of the proteasome and it is in these two major disease classes that the effects of HDAC6 inhibitors have been most extensively investigated to date.
HDAC inhibitors can be considered as broad-spectrum inhibitors, class-specific inhibitors or isoform-specific inhibitors. Four HDAC inhibitors, vorinostat (or suberoylanilide hydroxamic acid, SAHA) , romidepsin , belinostat  and panobinostat  have received regulatory approval for the treatment of haematological malignancies. These HDAC inhibitors demonstrate activity against multiple zinc-dependent HDAC isoforms from different classes including HDAC6. However, they are not necessarily all ‘pan’-HDAC inhibitors, one chemical phylogenetic analysis reporting unexpected isoform selectivity even for compounds perceived to be non-selective . Although these HDAC inhibitors are approved for the treatment of certain cancers, their application in more prevalent chronic diseases is likely to be limited by adverse effects that include haematological toxicity and QT prolongation. This is the rationale behind the development of class- or isoform-specific HDAC inhibitors that may be expected to have a more favourable side-effect profile. The class-specific HDAC inhibitors entinostat (MS-275) and mocetinostat (MGCD0103) exhibit specificity against Class I HDAC isoforms and have entered clinical trial, but have not yet received regulatory authority approval [68,69]. Unlike the other zinc-dependent HDAC isoforms, HDAC6 is unique in having had isoform-specific inhibitors of its enzymatic activity successfully synthesized. One (relatively) isoform-specific HDAC6 inhibitor, ricolinostat (ACY-1215), is currently under phase I/II clinical evaluation for the treatment of patients with relapsed-and-refractory multiple myeloma and lymphoid malignancies (ClinicalTrials.gov identifiers: NCT01323751, NCT01997840, NCT02091063 and NCT02189343) and a separate compound (ACY-241), also selective for HDAC6, is also undergoing phase Ia/b evaluation in multiple myeloma (NCT02400242).
HDAC inhibitors have a typical structure made up of three primary domains: a zinc-binding group which chelates the zinc ion at the active site preventing activation of the enzyme (typically a hydroxamic acid, but could also be a thiol, carboxylic acid, ketone or substituted aniline), a linker region and a cap group which binds to the substrate-binding region of the enzyme [70,71] (Figure 3). Variations in the HDAC inhibitor cap region appear to offer the best hope for isoform selectivity because homology-based 3D modelling techniques suggest that each HDAC enzyme has several pockets around the substrate-binding region that are unique and less conserved [72,73].
Typical structure of HDAC inhibitors
The first HDAC6 specific isoform to be developed was Tubacin, so-named because it is a tubulin acetylation inducer . Tubacin was identified through a multidimensional screen of 7392 small molecules [74,75]. It consists of a large cap composed of six hydrophobic rings and a 1,3-dioxane ring and it inhibits α-tubulin acetylation without altering histone acetylation or affecting gene expression or cell cycle progression . However its utility, especially in vivo, is limited by the complexity of the small molecule's synthesis, its lipophilicity and its lack of drug-like structure .
The inhibitor that has been most widely used to study HDAC6 function to date is Tubastatin A, the synthesis of which was originally described by Butler et al. in 2010 . The rational design of Tubastatin A is especially interesting. To select for isoform specificity, the investigators set out to compare HDAC6 with the Class I HDAC isoform, HDAC1. Because crystal structures have not been defined for HDAC6 and HDAC1, the investigators instead elected to use a bioinformatic tool for predicting protein structure based upon amino acid sequence . By comparing the modelled catalytic pockets of HDAC1 and HDAC6, they discovered that although the active site is conserved, the catalytic channel rim differs between the two isoforms being substantially wider in HDAC6 than HDAC1 . The investigators therefore set out to design compounds based upon the canonical HDAC inhibitor structure [i.e. zinc-binding group (hydroxamic acid), linker and cap group] with a cap group that was large enough and inflexible enough to occupy the catalytic channel rim of HDAC6 but not HDAC1 . The cap group that best fulfilled these requirements was the tricyclic structure of a carbazole cap . However, carbazoles are generally too lipophilic to make good drugs offering suboptimal ADMET (absorption, distribution, metabolism, excretion and toxicity) properties . So, the investigators introduced a tertiary amine to disrupt the planarity of the tricyclic ring and reduce lipophilicity . Finally, recognizing that the modelled catalytic channels of HDAC1 and HDAC6 also differ, with the HDAC6 channel being wider and shallower, the investigators sought to adapt the linker region, replacing the typical alkyl chain with bulkier and shorter aromatic moieties . The result was the synthesis of Tubastatin A which, in a cell-free enzyme inhibition assay, had an IC50 for HDAC6 of 0.015 μM, representing >1000-fold selectivity compared with HDAC isoforms 1–11 (except HDAC8, 57-fold selectivity) . Interestingly, however, a more recent study employing HDACs enriched from stable isotope labelling by amino acids in cell culture (SILAC)-encoded MV4-11 cell lysates reported a higher potency of Tubastatin A for HDAC10 (Kd 0.22) than HDAC6 (Kd 0.53) . These observations should be borne in mind before attributing the biological effects of Tubastatin A solely to blockade of the enzymatic actions of HDAC6. Nonetheless, in primary cultured neurons, Tubastatin A increased α-tubulin acetylation without affecting histone acetylation and it dose-dependently protected against oxidative stress-induced neuronal death . Subsequently, the investigators who originally developed Tubastatin A have gone on to synthesize second-generation analogues of the compound by making further modifications to the cap group . These second generation compounds have enhanced HDAC6 selectivity and reduced lipophilicity, some exhibiting subnanomolar activity at HDAC6 and >7000-fold selectivity for HDAC6 compared with HDAC1 .
Whereas Tubastatin A has been relatively widely adopted into pre-clinical mechanistic studies, the only preferentially HDAC6-specific inhibitors to have reached clinical trial are ricolinostat and ACY-241. Ricolinostat is a hydroxamic acid derivative with an IC50 for HDAC6 of 5 nM. However, it also has activity against other HDAC isoforms with IC50s for HDACs 1, 2, 3 and 8 of 58, 48, 51 and 100 nM respectively (IC50 >1 μM for the other HDAC isoforms) . As with other HDAC6 inhibitors, ricolinostat dose-dependently increased α-tubulin acetylation without affecting the acetylation status of histone proteins . It also induced less cytotoxicity in peripheral blood mononuclear cells and T-cells than the pan-HDAC inhibitor, vorinostat . Ricolinostat has mostly been studied for its role in combination with proteasome inhibitors for the treatment of multiple myeloma or lymphoid malignancies [80–83]. Proteasome inhibitors block the actions of proteasomes that break down proteins such as the tumour suppressor, p53. Some proteasome inhibitors, such as bortezomib, have already been approved for the treatment of multiple myeloma and some forms of lymphoma, preferentially exerting their cytotoxic effects on neoplastic cells. However, their effectiveness is limited because blockade of the proteasome causes accumulation of misfolded proteins in aggresomes. Accordingly, it is hypothesized that simultaneously blocking aggresome formation, through HDAC6 inhibition, may confer additive or synergistic benefit alongside proteasome inhibition. In support of this hypothesis ricolinostat, in combination with bortezomib, delayed tumour growth and improved survival in xenograft multiple myeloma mouse models  and in a xenograft model of diffuse large B-cell lymphoma . Likewise, ricolinostat, together with the proteasome inhibitor carfilzomib, demonstrated synergistic interaction in non-Hodgkin lymphoma cells  and inhibited aggresome formation and enhanced apoptosis in multiple myeloma cells . Ricolinostat has also been studied for its use in combination with immunomodulatory drugs (IMiDs) for the treatment of multiple myeloma  and its use in combination with the IMiDs pomalidomide (NCT01997840 and NCT02189343) or lenalidomide (NCT01583283) for this indication is currently undergoing clinical evaluation. Similarly, ACY-241 is currently undergoing phase Ia/b investigation for the treatment of relapsed or relapsed-and-refractory multiple myeloma either as monotherapy or when used in combination with pomalidomide and low-dose dexamethasone (NCT02400242). At least in one in vitro study, however, the cytotoxic effects of ricolinostat when used in combination with lenalidomide were attributed to the modest Class I HDAC inhibitory activity of the agent, rather than HDAC6-specific effects .
Although Tubacin, Tubastatin A and ricolinostat have been the most extensively studied agents to date, other HDAC6 inhibitors have also been synthesized. For example, arylalanine containing hydroxamic acids have been reported as another class of HDAC6 selective inhibitors, potent in low micromolar concentrations [85,86]. Because most HDAC inhibitors share a common structure, to enhance the HDAC inhibitor pool, Inks et al.  elected to screen the Library of Pharmacologically Active Compounds for agents that exhibit HDAC inhibitory properties in a search for novel compounds with a novel structure. Out of the library of 1280 compounds, they identified five with HDAC inhibitory properties, one of which (a dual-specificity phosphatase inhibitor, NSC-95397) was selective for HDAC6 . A number of analogues of the parent compound were synthesized and one, NQN-1, demonstrated an IC50 for HDAC6 of 5.5 μM, with minimal inhibitory activity against other HDAC isoforms . Molecules with a cyclic peptide scaffold or chiral structure derivatives [88,89] and sulfamide- , thiolate- , trithiocarbonate-  and mercaptoacetamide-  based compounds have also been explored as potential selective HDAC6 inhibitors.
ROLE OF HDAC6 IN DISEASE
HDAC6 and neurodegenerative diseases
Neurodegeneration is a catch-all term used to describe clinical conditions caused by the functional loss of neurons. These clinical conditions, which include Parkinson's disease, Alzheimer's disease, Huntington's disease and Charcot–Marie–Tooth disease, share a common pathological feature in that they are associated with the abnormal accumulation of misfolded proteins . Since HDAC6 is a master regulator of the misfolded protein response, significant research efforts have been made in understanding the role of the protein in neurodegenerative disease models. For example, in a fly model of the neurodegenerative disease spinobulbar muscular atrophy, HDAC6 was found to facilitate compensatory autophagy in the context of UPS impairment .
Parkinson's disease is caused by loss of dopamine-producing cells within the substantia nigra and is characterized by the accumulation of cytoplasmic inclusions, termed Lewy bodies, which are principally composed of the protein, α-synuclein. In Drosophila, HDAC6 promotes inclusion formation and protects dopaminergic neurons from the injurious cellular effects of α-synuclein  and, in brain sections from people with Parkinson's disease, Lewy bodies are enriched for HDAC6 . Together, these observations imply that HDAC6 up-regulation in Parkinson's disease brains may be a protective response suggesting that therapeutic augmentation of HDAC6 may slow the progression of the disease .
Familial early-onset forms of Parkinson's disease have also been linked to HDAC6. For instance, one autosomal recessive form of the condition that is devoid of Lewy bodies is caused by mutations in PARK2 the gene encoding the protein Parkin, a component of the E3 ubiquitin ligase that plays a pivotal role in the UPS. Although the normal role of Parkin has not been fully unravelled, the protein has been linked to the autophagic clearance of dysfunctional mitochondria (termed mitophagy) . It does this by catalysing mitochondrial ubiquitination, which facilitates the recruitment of HDAC6 and p62. Disease causing mutations in Parkin impair HDAC6-dependent PARK2 mitophagy . Through its interaction with HDAC6, Parkin also facilitates the incorporation into aggresomes of misfolded forms of the peptidase C56 family member DJ-1 , which is also associated with an autosomal recessive form of Parkinson's disease .
Alzheimer's disease is one of the clinical conditions in which the role of HDAC6 has been most intensively investigated . One of the characteristic histopathological features of Alzheimer's disease is the formation of neurofibrillary tangles that are made up of aggregates of the hyperphosphorylated form of the microtubule-associated protein, tau . Neurofibrillary tangles are not, however, unique to Alzheimer's disease, with more than 20 clinical conditions included under the umbrella term, tauopathy (e.g. Pick's disease, progressive supranuclear palsy, post-encephalitic Parkinsonism and corticobasal degeneration) . The phosphorylation and accumulation of tau into aggresomes is dependent on the microtubule-binding domain of tau and the SE14 of HDAC6 .
Tau is a client protein for HSP90  and HDAC6 levels correlate with tau burden, with a decrease in HDAC6 expression or activity favouring clearance of tau, potentially through the promotion of HSP90 acetylation and consequent attenuation of its tau-chaperoning actions . In Drosophila, an HDAC6 null-mutation rescued tau-induced microtubule defects . Conversely, HDAC6 levels were observed to be elevated in postmortem brain tissue from people with Alzheimer's disease . Even though HDAC6 has been associated with Alzheimer's disease in a number of studies, its precise role has not yet been fully established. For instance, tau itself has been reported to be an inhibitor of HDAC6  and, like in Parkinson's disease, HDAC6 up-regulation in Alzheimer's disease may be protective, the deacetylase regulating the appropriate degradation of misfolded proteins. Alternatively, the role of HDAC6 may be dependent on the stage of disease. Early up-regulation of HDAC6 may confer protective benefits, but over time this may lead to accelerated neuronal damage . Nonetheless, two separate groups have each recently reported an improvement in cognition with HDAC6 inhibition in mouse models of Alzheimer's disease [109,110].
Huntington's disease is an autosomal dominant condition caused by an expanded poly-glutamine sequence in the huntingtin gene, IT15, that also causes misfolding and aggregation of the protein . It has been suggested that the mutant huntingtin protein may cause neuronal toxicity through microtubule destabilization  and that HDAC6 inhibition may improve neuronal transport by increasing α-tubulin acetylation [113,114].
Charcot–Marie–Tooth disease is characterized by progressive muscle wasting, typically in the lower legs, and loss of sensation to light touch. Approximately 20–40% of individuals with Charcot–Marie–Tooth have the axonal type of the condition (CMT2), which has a heterogeneous genetic basis. One of the subgroups of CMT2 (CMT2F) is associated with missense mutations in the gene encoding heat shock protein 27 (HSPB1) . This causes decreased α-tubulin acetylation and defects in axonal transport affecting the peripheral nerves . In mice carrying mutations in the gene encoding HSP27 in neurons, treatment with Tubastatin A reversed the axonal defect .
HDAC6 and cancer
Whereas HDAC6 undoubtedly plays a role (albeit complex) in the pathogenesis of or protection against neurodegenerative disease, to date clinical trials of HDAC6 inhibitors have been restricted to the treatment of certain malignancies. The link between HDAC6 and aggresome formation represents probably the most clearly defined and (at present) clinically significant relationship between modulation of HDAC6 activity and altered cancer outcomes. Transformed cells accumulate misfolded proteins at a faster rate than non-transformed cells and, for cancer cell survival, these misfolded proteins must be appropriately disposed of through either the UPS or the aggresome-autophagy pathway . Proteasome inhibitors prevent disposal of misfolded proteins by the UPS and their use in combination with HDAC6 inhibitors may promote cytotoxicity by inhibiting both the UPS and the aggresome-autophagy pathway . The relationship between HDAC6 and cancer, however, is not limited solely to the role of the deacetylase in the misfolded protein response.
Whereas HDAC6 expression has been reported to be increased in a range of cancer types [119–123], the association between HDAC6 expression and prognosis is likely to be dependent on cancer type. For example, in breast cancer increased HDAC6 expression correlates with improved survival , which may relate to the fact that HDAC6 is an oestrogen-regulated gene . Up-regulated HDAC6 may function as a surrogate for oestrogen receptor α (ERα) status, where ERα-positive tumours are responsive to oestrogen-blocking therapy and associated with lower mortality than ERα-negative breast cancers [120,125]. Conversely, there are several points in the malignant process where HDAC6 may contribute to cancer development and progression [126,127].
Lee et al.  reported that HDAC6 is induced in Ras-oncogene transformed cells and that HDAC6-deficient cells and mice are resistant to Ras-induced oncogenesis, implying that HDAC6 likely facilitates the activation of Ras and the downstream phosphoinositide 3-kinase (PI3K) and mitogen-activated protein kinase (MAPK) pathways. The authors also found that HDAC6 is necessary for tumour cells to acquire the ability to resist anoikis (a type of programmed cell death that is normally induced when anchorage dependent cells detach from the surrounding matrix) thereby facilitating tumour invasion and metastasis [127,128].
A separate mechanism through which HDAC6 may contribute to cancer development and progression involves the interaction of HDAC6 with the chaperone protein, HSP90. For instance, HSP90 hyperacetylation through HDAC6 knockdown prevents the nuclear translocation of other client proteins, including the glucocorticoid receptor [37,129] and the androgen receptor , which in the latter case impaired xenograft establishment in a model of prostate cancer . These observations serve to illustrate that, even though HDAC6 resides in the cytoplasm, changes in its activity can eventually affect gene transcription, in this case through the altered nuclear activity of steroid hormone receptors. HDAC6 inhibition may also block tumour growth by preventing the nuclear localization of β-catenin . Ordinarily a transducer of Wnt signalling, translocation of β-catenin to the nucleus can also be induced by epidermal growth factor (EGF) where it activates target genes, including the transcription factor c-myc. A number of cancers are associated with overexpression of or mutations in β-catenin. Upon EGF stimulation, HDAC6 associates with β-catenin at the caveolae membrane deacetylating β-catenin at lysine residue 49 and inhibiting β-catenin serine 45 phosphorylation . These effects ultimately act to prevent the nuclear translocation of β-catenin, decrease c-myc expression and attenuate tumour proliferation . Separately, HDAC6 has been shown to form a ternary complex with β-catenin and the membrane glycoprotein CD133, which is a marker of cancer progenitor cells . This association stabilizes β-catenin thus facilitating β-catenin signalling .
Aside from tumour growth, HDAC6 may also adversely affect cancer outcomes by promoting tumour invasion and metastasis. Cortactin is an established HDAC6 substrate  that is up-regulated in a range of cancers and that plays a pivotal role in tumour invasiveness by promoting cell motility . It does this, at least in part, through its role in the formation of invadopodia, actin-rich protrusions that extend from cancer cells and that are associated with extracellular matrix degradation [133–135]. HDAC6 has been linked to cortactin-dependent invadopodia formation and cancer cell invasion in several studies [134–136].
Finally, an important component of the cellular response to stress is the formation of stress granules, membraneless cytoplasmic RNA–protein complexes 100–200 nm in size  that can affect mRNA translation and stability . These complexes can sequester apoptotic regulatory proteins, thereby facilitating cell survival . HDAC6 associates with stress granules through its interaction with the stress granule protein, Ras-GTPase-activating protein SH3 domain-binding protein-1 (G3BP1) and either HDAC6 inhibition or knockdown prevents stress granule formation . Since G3BP1 down-regulation blocks tumour metastasis , prevention of stress granule formation may represent a further mechanism for the anti-cancer effects of HDAC6 inhibition.
HDAC6 and cardiovascular (patho)physiology
A collective body of evidence indicates that HDAC6 also plays a role in cardiovascular (patho)physiology. Cardiac HDAC6 activity is increased in multiple models of left ventricular and right ventricular pressure overload, including in deoxycorticosterone acetate (DOCA)-salt hypertensive rats, spontaneously hypertensive heart failure rats and rats with right ventricular hypertrophy due to chronic hypoxia, but not in rats with physiological (exercise-induced) cardiac hypertrophy [141,142]. When mice were subjected to angiotensin II infusion, both wild-type and HDAC6 knockout animals developed cardiac hypertrophy and fibrosis but only wild-type animals developed systolic dysfunction . Similarly, treatment of angiotensin II-infused mice with Tubastatin A prevented systolic dysfunction and, when transverse aortic constriction was induced in HDAC6 knockout mice, the mice developed cardiac hypertrophy but were protected from the development of left ventricular systolic dysfunction . These findings suggest that HDAC6 may have some function in the regulation of cardiac contractility. Indeed, HDAC6 co-localizes with myofibrils and isolated cardiac myofibrils from HDAC6 knockout mice exhibit an increase in calcium-activated force generation in response to angiotensin II, suggesting that the protein may function to regulate contractile strength by deacetylating sarcomeric proteins . Nonetheless, the precise role of HDAC6 in regulating cardiac contractility (if any) remains unclear. Using the Langendorff isolated perfused heart model, for example, investigators found that HDAC6 inhibition with Tubastatin A did not improve cardiac contractility following ischaemia reperfusion injury and may actually have been detrimental by decreasing levels of the antioxidant, catalase .
Separately, the activity of HDAC6 has been linked to atrial fibrillation . Tachypacing of atrial cardiomyocytes increased HDAC6 expression and activity that disrupted the microtubule network through α-tubulin deacetylation, depolymerization and calpain-dependent degradation, ultimately impairing cardiac function . HDAC6 expression and activity were similarly increased in the atrial appendages of patients with atrial fibrillation, and HDAC6 inhibition, with Tubastatin A, prevented tachypacing-induced electrical remodelling in dogs .
Cardiomyocytes are essentially post-mitotic and therefore unable to regenerate. As a result, they are vulnerable to the deleterious effects of the accumulation of misfolded proteins, which can cause heart failure. McLendon et al.  observed that hyperacetylation of α-tubulin occurred in a mouse model of proteinopathy-induced heart failure. Reasoning that this is an adaptive response, the investigators observed that knockdown or inhibition of HDAC6 increased autophagy and reduced aggresome accumulation in cultured cardiomyocytes and that pan-HDAC inhibition in vivo prevented aggresome formation and improved cardiac function . Because the aging heart has a reduced capacity to remove protein aggregates by autophagy , this has led the investigators to postulate that HDAC6 inhibition may improve cardiac function in the elderly .
HDAC6 and other diseases
HDAC6 and inflammation
In the original report describing the phenotype of HDAC6 knockout mice, the authors noted a moderately impaired immune response . Since then, several pre-clinical studies have described an anti-inflammatory effect of HDAC6 inhibition [149–151]. Naturally occurring regulatory T-cells promote tolerance to autoantigens and prevent against the development of autoimmunity. They can be identified by their expression of the transcription factor and marker, forkhead box P3 (Foxp3). In models of inflammation or autoimmunity (including models of colitis and cardiac allograft rejection) either HDAC6 knockdown or inhibition promoted the suppressive activity of Foxp3+ T regulatory cells .
Interestingly, four recent reports specifically described an effect of HDAC6 on the production of the anti-inflammatory cytokine, interleukin-10 (IL-10) with strikingly contrasting findings [152–155]. In 2014, Wang and colleagues reported that deficiency of HDAC6 resulted in augmented production of IL-10 when a cultured macrophage cell line was exposed to bacterial lipopolysaccharide (LPS) . The investigators went on to demonstrate that this effect was mediated by microtubule hyperacetylation, which selectively amplified p38 signalling and subsequent SP-1 dependent IL-10 transcription . The same year, in two separate articles, Cheng et al. [153,154] reported that knockdown of HDAC6 in the same macrophage cell line (RAW264.7) inhibited IL-10 expression. They attributed this effect to the formation of complexes between HDAC6 and either HDAC11  or the transcription factor, signal transducer and activator of transcription 3 (STAT3) . The reasons for this discordance between the studies are not immediately apparent and may relate to differences in the experimental conditions including the passage number of the cells or the extent to which HDAC6 knockdown was achieved. However, a more recent study also reported an up-regulation of IL-10 with HDAC6 inhibition in a mouse model of silicone-induced inflammation .
HDAC6 and mood disorders
Whereas the focus of many studies of the role of HDAC6 in the central nervous system has been on the function of the deacetylase in neurodegenerative diseases, a growing body of evidence separately suggests that HDAC6 may also be important in the development of mood disorders. Loss of HDAC6 function, for example, has an antidepressant like effect in rodents [156–158]. One possible mechanism for this effect is the role of HDAC6 in modulating the interaction between HSP90 and the glucocorticoid receptor, with HDAC6 knockdown or inhibition preventing translocation of the receptor to the nucleus in neuronal cells , as similarly described in other cell-types [37,129]. Indeed, two novel, brain-penetrant HDAC6-selective inhibitors, ACY-738 and ACY-775 (with low nanomolar potency and 60- to 1500-fold selectivity compared with Class I HDACs), were recently shown to have exploration-enhancing effects in mice , a common feature of antidepressant therapies .
Platelet activation must be finely balanced to ensure haemostasis without excessive clot formation and the regulation of microtubule acetylation by HDAC6 has also been implicated in the control of this process . The marginal band is a structure that is essential to the maintenance of the normal discoid shape of resting platelets  that is enriched in hyperacetylated tubules . During platelet activation, HDAC6-mediated microtubule deacetylation enables platelet shape change to take place, which is followed by reacetylation of microtubules in spread platelets [161,163].
HDAC6 and kidney disease
A requirement for HDAC6 in transforming growth factor-β (TGF-β) induced epithelial–mesenchymal transition [164,165] and a reduction in TGF-β expression in the kidneys of angiotensin II-infused mice treated with Tubastatin A both suggest that HDAC6 may play some role in renal fibrosis . Separately, HDAC6 has also been implicated in cystic diseases of both the liver  and the kidney . This association likely relates to the importance of HDAC6 in the formation of the primary cilium. Almost all mammalian cells possess a single primary cilium. Far from being vestigial organelles, primary cilia play important roles in intercellular signalling and in the regulation of cell division, their dysfunction contributing to a range of diseases including Bardet–Biedl syndrome, neural tube defects, retinal degeneration and polycystic kidney disease . Primary cilia contain a microtubule-based internal structure called an axoneme that arises from the centrosome, a microtubule-organizing centre, comprised of two centrioles surrounded by a pericentriolar matrix . The membrane of the primary cilium is attached to the distal end of one centriole and, thus, in order for cell division to occur, the primary cilium must be disassembled . This disassembly process involves an interaction between the scaffolding protein, neural precursor cell expressed developmentally down-regulated protein 9 (NEDD9) and the oncogenic kinase, Aurora A that results in the phosphorylation and activation of HDAC6 . Because cyst growth occurs as a result of persistent proliferation of de-differentiated epithelial cells , dysregulation of HDAC6-mediated ciliary disassembly may contribute to the development of renal cysts .
The regulation of primary cilium disassembly is not the sole mechanism through which HDAC6 may contribute to the development of renal cysts. Through its α-tubulin deacetylating actions, HDAC6 also regulates the intracellular transport of the epidermal growth factor receptor (EGFR) , whose increased activity promotes cyst formation . In kidney epithelial cells with a mutation in the PKD1 gene, that encodes the protein polycystin-1 and that is associated with autosomal dominant polycystic kidney disease, HDAC6 expression was observed to be increased, whereas HDAC6 inhibition promoted EGFR degradation and normalized EGFR localization . Autosomal dominant polycystic kidney disease can be caused by mutations in either the PKD1 gene or in the PKD2 gene, the latter encoding the protein polycystin-2. Polycystin-1 and -2 interact with each other . Separate to its role in EGFR trafficking, HDAC6 also binds polycystin-2 and expression of full-length polycystin-1 accelerates transport of the polycystin-2/HDAC6 complex towards aggresomes, facilitating the degradation of polycystin-2 by autophagy and thus negatively regulating its expression . The balance between increased and decreased activity of polycystin-1 and -2 therefore appears to be tightly regulated in renal epithelial cells and either up-regulation or down-regulation of either protein may result in cyst formation . It is possible that inhibiting HDAC6 can redress an imbalance in polycystin-1/2 activity attenuating the development of renal cysts.
In summary, despite the moniker of being a histone deacetylase, HDAC6 is an enzyme that is unique among HDACs in its cytoplasmic functionality. It deacetylates non-histone proteins and, independent of its catalytic activity, it acts as a bridge linking the UPS and the aggresome-autophagy pathway, regulating the disposal of misfolded proteins. HDAC6 expression or activity is altered in cancer, neurodegenerative diseases, cardiovascular disease, inflammation and other diseases, where it may contribute to the pathogenesis of the condition or where it may be a compensatory response (Figure 4). Based upon its mechanism of action, however, it is not a prerequisite for HDAC6 activity to be altered in a particular disease for pharmacological inhibitors of its activity to find their clinical niche there. Most advanced in clinical development is the use of HDAC6 inhibition in combination with proteasome inhibitors for additive cytotoxic effects in haematological malignancies. Given the likely benign effects of HDAC6 inhibition in normal cells, whether these therapies will be applied in more prevalent complex chronic diseases awaits to be seen.
Conditions associated with altered HDAC6 activity or in which HDAC6 inhibition may confer therapeutic benefit
We thank Kryski Biomedia for the elegant artwork.
This work was supported by the Heart and Stroke Foundation of Canada [grant number G-14-0005877 (to A.A.)]; a Keenan Family Foundation KRESCENT Post-doctoral Fellowship (to S.N.B.); a Heart and Stroke/Richard Lewar Center of Excellence Fellowship Award (to S.N.B.); a Queen Elizabeth II/Dr Arnie Aberman Graduate Scholarship in Science and Technology (to A.S.B.); and a Yow Kam-Yuen Graduate Scholarship in Diabetes Research from the Banting and Best Diabetes Centre (to A.S.B.).
epidermal growth factor
epidermal growth factor receptor
oestrogen receptor α
forkhead box P3
Ras-GTPase-activating protein SH3 domain-binding protein-1
heat shock transcription factor protein 1
heat shock protein 27
heat shock protein 90
nuclear export signal
nuclear localization signal
transforming growth factor-β
ubiquitin and proteasome system
ubiquitin-binding zinc-finger domain
These authors contributed equally to this work.