Recent Progress in Synthetic Strategies to Develop Potent, HDAC8-Selective, Small-Molecule Inhibitors

Abstract

Key players in epigenetic control, the 11 zinc-dependent histone deacetylase (HDAC) enzymes have been validated as therapeutic targets, particularly in the discovery of new chemotherapy drugs: currently there are 3 FDA-approved HDAC inhibitors (HDACi’s), although these are not without side effects. It has been rationalized that achieving HDAC isoform-selective inhibitors will yield safer drugs. HDAC8 has recently been validated as a novel target for the treatment of (pediatric) neuroblastoma, and its overexpression has been implicated in a range of other diseases. Herein, we discuss the topology of the HDAC8 active site in the context of inhibitor design, and we present recent progress in the discovery of potent, HDAC8-selective small-molecule inhibitors, with a discussion on the particular structural criteria required. HDACi’s largely conform to a canonical pharmacophore model of a capping group connected via a linker to a zinc binding group (ZBG), typically a (potentially mutagenic) hydroxamic acid. In particular, for HDAC8-selectivity, the small-molecule should adopt a geometric “L”-shape, which may be accomplished by the linker itself, or the linker and capping group. More recently, isoform-selectivity has been realized with hydroxamic acid surrogates, such as ortho-aminoanilides that yield potent and HDAC1–3-selective inhibitors. Hydrazides have long been utilized as bioisosteres of hydroxamic acids in HDAC inhibitor design, but a recent re-visiting explored the attachment of alkyl groups, and an n-hexyl substituent on the distal, non-acylated nitrogen affords potent and selective HDAC8 inhibitors through targeting the foot pocket of the active site; this discovery was observed across three different HDACi scaffolds, and thus appears quite general. α-Aminoketo ZBGs have also demonstrated HDAC8-selectivity, with associated benzylic moieties likewise engaging the foot pocket. We speculate that a re-imagining of the ortho-aminoanilide through a careful and methodical survey of related moieties, may likewise discover further HDAC8-selective, and safe, pre-clinical inhibitors.

Graphical Abstract

Introduction

Histone deacetylase (HDAC) enzymes play an integral role in epigenetic control, regulating gene expression by catalyzing the deacetylation of acetylated lysine residues on histone proteins (Figure 1), as well as some other non-histone proteins, rendering these side chains cationic leading to compaction of the chromatin.¹ Accordingly, the genes contained within this chromatin are no longer accessible to the transcriptional machinery. In healthy systems, HDACs, or the “erasers”, are present at homeostatic levels, working in concert with histone acetyl transferase (HAT) enzymes, the “writers”, to turn gene transcription off and on as needed. Together with the bromodomain (BRD) “readers”, the HATs and HDACs ensure tight regulation of gene transcription.² Disease states may develop through the overexpression of HDACs (and HATs, and BRDs).³ Of particular note, certain cancers are associated with HDAC upregulation; proteins important for preventing tumorigenesis are no longer expressed.⁴ The tumor suppressor protein p53, aka “the guardian of the genome” was the first nonhistone target of HATs and HDACS to be described. By deacetylating it, HDACs deactivate p53.^5,6 Together, this points to the development of HDAC inhibitors, or HDACi’s, as a promising strategy for drug discovery in oncology (as well as in other diseases).

Figure 1:

A schematic of the homeostatic gene regulation activity of histone deacetylase (HDAC) and histone acetyl transferase (HAT) enzymes. Prepared with Biorender.com.

To date, 18 HDACs have been identified, and these are divided into 4 different classes based on a combination of their architecture, sequence similarity, and their evolutionary relationships to yeast RPD3.⁷ The class I HDACs include HDAC1, 2, 3 and 8. HDAC1/2 share roughly 82% sequence similarity,⁷ and unsurprisingly they share a great deal of similarity in their function and implications in various disease states, and have been extensively studied.⁸ HDAC3 and HDAC8 are more evolutionarily related to one another than to HDAC1/2. However, in contrast to HDAC1/2, they share only 34% overall sequence similarity.⁷ The family of Class II HDACs is divided into two sub-families: class IIa includes HDAC4, 5, 7, and 9, while class IIb comprises HDAC6 and 10. The class IIb HDACs are evolutionarily closer to class I HDACs than to the Class IIa isozymes, and have been more thoroughly investigated.⁷ While class I HDACs are predominantly found in the nucleus, class II HDACs are found in both the nucleus and the cytoplasm⁹ Class IV consists of a solitary member, HDAC11, which may be found in the cytoplasm, nuclei or mitochondria of brain, heart and kidney tissue, among others.¹⁰ The 11 class I, II, and IV enzymes are all zinc-dependent isozymes. On the other hand, class III enzymes, of which there are 7, instead contain a nicotinamide adenine dinucleotide (NAD⁺) ion and are historically referred to as sirtuins rather than HDACs.¹¹

Class I HDACs are associated with a wide range of pathological disorders, most notably a plethora of human cancers,^12,13 as well as inflammation, infection, cardiac fibrosis, atherosclerosis and neurological diseases.¹⁴ As with the other Class I HDACs, elevated levels of HDAC8 have been reported in T-cell lymphoma, cervical, colon, lung and breast cancer cells;^15,16 uniquely, however, only HDAC8 is overexpressed in neuroblastoma, and its subsequent siRNA knockdown¹⁷ has shown a decrease in neuroblastoma cell proliferation, suggesting that HDAC8 is a potential target for the development of anti-neuroblastoma therapeutics; indeed, this has recently been validated with an HDAC8-selective inhibitor.^16,18 High levels of HDAC8 are also associated with Cornelia de Lange Syndrome,¹⁹ and schistosomiasis, the second most common parasitic infection in the world behind malaria.²⁰ It is caused by the Schistosoma mansoni (Sm) blood fluke, and a large overexpression of SmHDAC8 has been reported in the parasite. Incidentally, SmHDAC8 has significant structural similarity to human HDAC8.²¹ Additionally, HDAC8 inhibition correlated with a relief in hypertension²² a reduction in the viral entry of influenza virus A,²³ and treated acute kidney injury in a zebrafish model.²⁴ The development of selective HDAC8 inhibitors is distinctly of clinical significance. Despite this, only one HDAC8 inhibitor (“NBM-BMX”) has advanced to clinical trials (NCT06012695), and we discuss the reasons why this may be the case below. We hope our review may help address this paucity in HDAC8 clinical candidates.

Histone Deacetylase Inhibitor Pharmacophore

With limited structural data available, initial drug discovery strategies towards the development of HDACi’s yielded pan-HDAC inhibitors, or “pan-HDACi’s”, which inhibited multiple members of the family of enzymes. For class I, II, and IV HDACs, which all contain a catalytic zinc ion within the active site binding pocket, there is an established canonical pharmacophore for the design of HDACi’s (Figure 2A): 1. a capping group (“cap”, blue) that interacts with the HDAC’s solvent-exposed surface surrounding the zinc binding domain, 2. a linker (green) that binds within the hydrophobic channel of the HDAC active site, and also serves to align the capping group most favorably with the protein surface and to optimally deliver the 3. zinc-binding group (“ZBG”, red) that binds/chelates the active site zinc ion. In some cases, a foot pocket-targeting group (“FPG”, yellow) associated with the ZBG is incorporated to garner additional interactions beyond the zinc ion. The most prevalent ZBG is the hydroxamic acid functional group, but occasionally a carboxylic acid, hydrazide, thiol, or a ketone, may be deployed, among others.²⁵ Conventionally, the cap and linker have been exploited to accomplish HDAC isoform selectivity,^26,27 while the ZBG is now drawing attention in this regard; ortho-aminoanilides have emerged as HDAC1/2/3-selective ZBGs,^25,28,29 which has fueled research into harnessing specific nuances within HDAC isoform zinc binding sub-pockets to achieve HDACi selectivity.²⁷ The FDA has approved three pan-HDACi (Figure 2): belinostat (1), for the treatment of peripheral T-cell lymphoma, vorinostat (2), for the treatment of cutaneous T-cell lymphoma, and, until recently, panobinostat (3) for the treatment of multiple myeloma (FDA approval was withdrawn in 2021);³⁰ also, the dual HDAC1/2 inhibitor romidepsin (4) has been approved for cutaneous T-cell lymphoma.³¹ In the recent past, it has been speculated that the toxicity/off-target effects of the FDA-approved, and other pre-clinical, HDACi’s may be associated with either (a) their pan-HDAC inhibitory activities and/or (b) the mutagenicity of the hydroxamic acid that often features in potent HDACi’s,^{25, 32–34} and/or (c) poor pharmacokinetics (PK) in vivo, particularly with those carrying hydroxamic acids.³⁰ Accordingly, research within the realm of HDACi’s has shifted away from the development of broad-spectrum HDACi’s and towards the discovery of inhibitors that are able to inhibit only specific members of the HDAC family to treat the disease state for which that member is commonly dyesregulated.^36–39 It is hoped this strategy may widen the therapeutic windows of HDACi’s. In parallel, many research groups have embarked on quests to discover safer alternatives to the hydroxamic acid ZBG, that ideally also exhibit improved PK profiles.^{25, 33, 34} As already introduced, to accomplish the first goal of HDAC isoform specificity, any one of the 4 main key components of the canonical HDACi pharmacophore may be altered to fit the binding site specificities of a particular HDAC isoform.²⁷ Since the structural diversity between HDAC isoforms is greater on the surface near the rim of the active site,⁴⁰ most efforts towards garnering selectivity for an isoform have focused on developing a unique capping group, although there has been a paradigm shift in recent years, and the ZBG is now receiving increasing attention as a means to achieve selectivity.²⁷ The hydroxamic acid is the “gold standard” ZBG for HDACi’s but research into safer bioisosteres with improved PK properties is gathering pace.^25,34 Below, we discuss synthetic strategies tailoring the capping group, the linker group, the ZBG and the foot pocket-targeting group in the context of the discovery of clinically-viable HDAC8-selective inhibitors.

Figure 2:

A. General HDACi pharmacophore: ZBG = zinc-binding group, FPG = foot pocket-targeting group; B. structures of four HDAC inhibitors that have obtained FDA approval (panobinostat has since been withdrawn)1, 2, and 3 are pan-HDAC inhibitors, while romidepsin is a dual HDAC1/2 inhibitor. Unlike the other 3 which utilize a hydroxamic acid ZBG, romidepsin is a disulfide prodrug of an active thiol metabolite, wherein the thiol serves as the ZBG.

The Capping Group: The Unique Side-Pocket

Although HDAC1, HDAC2, HDAC3 and HDAC8 all belong to Class I, there is a large number of compounds that inhibit HDAC1–3 but not HDAC8. HDAC1–3 all share a general hydrophobic region on their surfaces surrounding the active site, while HDAC8 sports a unique and distinct hydrophobic groove.⁴⁰ Results from an inhibitor-bound structural study performed by Marek, Romier and colleagues shows the existence of L1 and L6 loops in HDAC isozymes. All HDAC isozymes aside from HDAC8 present a ridge or an “L1-L6 lock” on the surface of the protein adjacent to the ZBG pocket/active site.⁴⁰ In most HDACs, this ridge blocks access to a hydrophobic groove, which places restrictions on inhibitor design. However, HDAC8 differs in a significant way: this ridge is absent, opening the hydrophobic group for manipulation in HDAC8-selective inhibitor design, capturing interactions with the catalytic residue Y341 (Y306 in hHDAC8).⁴⁰ The Class IIb isozyme HDAC6 demonstrates the greatest similarity in this region to HDAC8 with a less pronounced L1-L6 “lock”, suggesting that leveraging this hydrophobic groove as a strategy to design HDAC8-selective inhibitors may be compromised by unintended HDAC6 inhibitory activity. This may be welcomed if dual HDAC6/8 inhibitory activity is the goal as multiple groups have explored.^41,42 In order to target this uniquely-accessible hydrophobic groove on HDAC8, the capping group should be positioned at a meta- (or analogous) position with respect to the ZBG, affording the inhibitors an overall geometric “L-shape”, a structural feature that is governed by the identity of the linker motif (see next section). Taken together, this survey of biochemical, biophysical and crystallographic data for existing HDAC8-selective inhibitors gave rise to the “L-shape” hypothesis, and that HDAC8–ligand recognition is optimal when the ligand can adopt an “L-shaped” conformation. Free energy and entropy considerations inform this interaction should be favoured further still if the ligand is pre-organized, through a molecular constraint, into that specific geometry.

PCI-34051 (5, Figure 3A) is arguably the most well-studied HDAC8-selective inhibitor: it features a molecular “L-shape” as a result of the relative positionings of the hydroxamic acid ZBG at the 6-position of the indole bicyclic linker and the para-methoxybenzyl capping group at the 1-position, affording a meta-type of relationship. In fact, PCI-34051 served as the prototypical HDAC8-selective inhibitor that led to the development of the “L-shape” hypothesis.⁴⁰ With an IC₅₀ for HDAC8 of 10 nM, PCI-34051 exhibits >200-fold selectivities over HDAC1–3, 6, and 10.³³ A co-crystal structure of PCI-34051 bound to SmHDAC8 (which shares a largely conserved active site with human HDAC8³⁸ (hHDAC8) shows how the “L-shaped” orientation manifests itself with the delivery of the para-methoxybenzyl group in the hydrophobic groove between the L1 and L6 loops, where this group engages with the protein through a “T-shaped” π-π stacking interaction with Y341 (Y306 in hHDAC8) (PDB ID: 6HQY; Figure 3B). Further, the indole linker of PCI-34051 participates in a π-π stacking interaction with H188 (H180 in hHDAC8) in a hydrophobic channel, constructed by the side chains of F151, F216, Y341, G150, H188 and H292 (or F152, F208, Y306, G151, H180 and M274 in hHDAC8, respectively),⁴³ that leads to the catalytic zinc ion, with which the hydroxamic acid of PCI-34051 forms a bidentate interaction. PCI-34051 exhibits a short half-life in vivo, and is not suitable for clinical applications, but it serves as a cornerstone in HDAC8-selective inhibitor design.⁴⁴

Figure 3:

(A) HDAC8 selective inhibitor PCI-34051 and its HDAC inhibitory activity profile in μM. (B) SmHDAC8 co-crystallized with PCI-34051 (PDB ID: 6HQY). Atoms coloured by atom type: protein in grey, ligand in green.

A surface rendering of the co-crystal structure of PCI-34051-SmHDAC8 is shown in Figure 4A. By aligning and overlaying analogous surface rendered crystal structures of HDAC1 and HDAC6 to the PCI-34051-bound HDAC8 co-crystal structure, we demonstrate how the geometric “L-shape” is not accommodated by the HDAC1 (Figure 4B) and, to a lesser extent, the HDAC6 (Figure 4C) active sites. These renderings clearly show how the L1-L6 locked ridge in HDAC1 and HDAC6 sterically prevents the accommodation of a hydrophobic capping group of an HDAC8i if the inhibitor is geometrically forced into an “L-shape”.

Figure 4:

PyMOL surface renderings of crystal structures of: (A) smHDAC8 co-crystalized with PCI-34051 (PDB ID: 6HQY), (B) HDAC1 in maroon (PDB: 4BKX) overlaid on HDAC8 with PCI-34051, and (C) HDAC6 in orange (PDB: 5EEN) overlaid on HDAC8 with PCI-34051. These surface renderings highlight the ridge between the L1 and L6 loops in HDAC1 and HDAC6, which is not receptive to the para-methoxybenzyl moiety of PCI-34051.

Many other HDAC8-selective inhibitors have also been developed, and the capping group variation amongst HDAC8 inhibitors suggests that the relative position of the capping group is far more important than the nature of the capping group.⁴⁰ The HDAC8 selective inhibitors in Figure 5 highlight the widespread diversity of structures and sizes of the capping groups accommodated by the hydrophobic groove on the HDAC8 surface.^39,45–50 Despite this high degree of variability, all these inhibitors conform to specific criteria yielding highly potent and selective HDAC8 inhibitors: 1. The linker presents a meta (or analogous) relationship between the hydroxamic acid ZBG and the capping group; 2. the capping group is generally hydrophobic to elicit Van der Waals interactions with the hydrophobic groove on the HDAC8 surface, which may include residues of the H6 loop, such as P291 and H292 (hHDAC8 P273 and M274), and potentially a hydrogen bond with the terminal amine of K20 (hHDAC8 K33).⁴⁰ MD simulations have identified additional residues that may also play a role in binding the capping group;⁴⁴ 3. the capping group should preferably contain an aromatic species capable of engaging in favorable π−π stacking interactions with Y341 (hDAC8 Y306).^40,44,51

Figure 5:

Structures and Inhibition Profiles from HDAC inhibition assays (IC₅₀ values presented in μM) of “L-Shaped” HDAC8 inhibitors highlighting the diversity of capping groups and rigidity of the linking groups.^{38,42,45–50}

Other prominent HDAC8-selective inhibitors besides PCI-34051 (5) include NCC-149 (6) and its derivative 7, which exhibit HDAC8 IC₅₀s of 70 and 53 nM, respectively, and are >100-fold selective over HDAC1 and HDAC6. These compounds utilize a longer capping group that includes a triazole and sulfide.³⁹ Compounds 8 and 9 from Stuart Schreiber’s laboratory are two instructive examples of the degree of capping group variability.⁴⁵ These two compounds both utilize a benzene ring as a linking group with a hydrazide-functionalized hydrophobic “arm” in a meta-position with respect to the ZBG, but the rest of the capping group differs greatly: 9 is a much larger 12-membered ring structure comprised of an amide and an ester while 8 incorporates a simple phenyl ring. Despite these differences, both compounds exhibit similar HDAC8 inhibitory activities, and share >100-fold selectivities for HDAC8 against HDAC2 and HDAC3.⁴⁵ Although these compounds were not tested against the structurally similar HDAC1 or HDAC6 isozymes, the data nonetheless exemplify the significant tolerability in the capping group. Similarly, compounds 10–13 showcase that variability is accepted in the connection between the linking and capping groups with the respective authors citing the importance of positioning their capping groups in the hydrophobic groove to elicit van der Waals and π-π stacking interactions with the catalytic Y341^39,40 Generally speaking, these compounds feature the capping group at the aforementioned meta position of the linker group relative to the ZBG hydroxamic acid, realizing the L-shaped geometry for HDAC8 selectivity. Hassan et al. designed and synthesized compounds 14 and 15 which appear to contradict this molecular shape mandate as these compounds contain capping groups at the para position of the linkers. However, the authors obtained X-ray crystal structures and performed computational docking studies that determined these benzanilide compounds still adopt an L-shaped configuration likely induced sterically by the isopropyl substituent.⁴⁸ In summary, along with the linker-governed positioning of the capping group resulting in an overall molecular L-shape, inclusion of an aromatic species and overall hydrophobicity in the capping group are beneficial to garnering HDAC8 affinity and selectivity.

The Linker Group

Compared to pan HDACi’s 1–3 (Figure 2), selective HDAC8I’s (Figure 4) generally have much less chemical diversity within their linkers, both in length and rigidity. To date, the majority of selective HDAC8 inhibitors utilize either a single aromatic ring, or a bicylic aromatic, linker group as shown with compounds 5–15, wherein the capping group is tethered through a meta (or equivalent) position, relative to the ZBG, affording the inhibitor with an overall geometric “L-shape”. As already discussed, this is due to the need to probe the unique hydrophobic pocket between the L1 and L6 loops whilst also facilitating the simultaneous binding/chelation of the zinc ion with the ZBG.^38,44 With these constraints in mind, only a few selective HDAC8 inhibitors with non-aromatic linking groups have been reported, although they often feature some rigidity within the capping group that likely compensates for the lack of the requisite linker shape. Taha et al. demonstrate this hypothesis via compounds 16 and 17.⁴⁹ Structurally, these compounds are quite similar: each contains the hydroxamic acid ZBG, and uses a short ethyl linker with a tertiary amine leading into the capping group. While the capping group for 17 is arranged in a more rigid tetrahydroisoquinoline moiety with a para-methoxyphenyl substituent, the capping group for 16 contains two flexible benzyl groups. 17 demonstrated >100-fold selectivity for HDAC8 over HDAC1/2/3 while 16 exhibited minimal HDAC8 selectivity, which may be ascribed to the greater rigidity of 17’s capping group.⁴⁹ Similarly, Huang, et al. have also developed highly HDAC8-selective inhibitors without an aromatic linker instead employing rigid alkenes, as illustrated by compounds 18 and 19.⁵⁰ In particular, the trans-alkene in 18 facilitates an analogous “meta” projection of the distal phenyl ring that takes on the role of the capping group, and in the case of 19, while the phenyl ring is ortho,para-substituted, the more acute L-shaped geometry afforded by this smaller linker is accommodated by HDAC8 as the alkene introduces a spacer that allows for the ZBG and capping groups to align optimally within the HDAC8 active site.⁵⁰ In summary, the unique need to establish the geometric alignment manifested through a molecular “L-shape” decreases the variability of HDAC8 linkers, and has resulted in greater focus on variations of the capping group and, more recently, the ZBG.

The Zinc-Binding Group, ZBG (and Foot Pocket-Targeting Group, FPG)

Perhaps the most intriguing and clinically relevant component of the HDAC pharmacophore is the ZBG, the most prominent of which is the hydroxamic acid: without a ZBG, an HDACi is inert.^21,27 As a potent zinc chelator,^27,46 the hydroxamic acid rapidly became the ZBG of choice in HDACi development, and it features in 2 of the 3 HDAC inhibitors that are currently approved by the FDA (Figure 2B). While the hydroxamic acid has proven to be a prolific ZBG in the design of selective HDAC8 inhibitors in vitro, this functional group is not optimal for clinical use, presenting a drug design quandary, and significant re-design may be required for translation into a pre-clinical candidate. As mentioned previously, hydroxamic acids exhibit poor PK properties.³⁵ Moreover, they are potentially mutagenic: chelation to metal ions, especially Zn²⁺ ions, may initiate a variant of the Lossen rearrangement affording a highly electrophilic isocyanate.⁵³ Surrogates of the hydroxamic acid group (examples shown in Figure 6) such as carboxylic acid and ketone functional groups typically elicit weaker biological activities, and although thiols, in general, are equipotent⁵⁴ with hydroxamic acids, they suffer from off-target effects through their high reactivity, as well as facile oxidation.^25,55 To the best of our knowledge, there are no reports of thiols, ketones, or carboxylic acids eliciting selective HDAC8 inhibition. Likewise, although the ortho-aminoanilide ZBG has emerged as a HDAC1–3 selectivity determinant, it only weakly inhibits Class I co-family member HDAC8.^7,56

Figure 6:

Alternative ZBGs to the hydroxamic acid. R¹ = aryl; R² = H or alkyl.

The most comprehensive study exploring the replacement of the hydroxamic acid for alternative ZBGs for HDAC8 inhibition was published recently by Huryn et al., wherein the hydroxamic acid of PCI-34051 (5) was methodically replaced with almost 30 potential alternative ZBGs.³⁴ In the HDAC8 inhibitory assay with these congeners, the best, as showcased in Table 1, were those with a propanoyl group (20), a 1,2,4-Oxadiazol-5(2H)-one (21), an ortho-aminobenzamide (22), and an ortho-aminoanilide (23) as putative hydroxamic acid bioisosteres; that said, each of these alternatives showed 1600-fold less potency compared to the hydroxamic acid.³⁴ In the case of the first 2 surrogates (20, 21), we surmise this is because these functional groups cannot chelate the zinc ion, unlike the hydroxamic acid. For 22 and 23, which are capable of chelating zinc,⁵⁷ the lack of activity for HDAC8 is presumably due to a steric constraint, such that these functional groups cannot be physically accommodated in the foot pocket. Huryn’s study is consistent with previous research that the ortho-aminoanilide ZBG is largely inactive against HDAC8, and so it is not a pan-Class I ZBG.^25,34,58 Included in Huryn’s study were various other putative ZBGs with 5 or 6-membered rings such as an ortho-hydroxylbenzamide, ortho-methylthiobenzamide, isoxazole, oxadiazole, and a few others, although all were essentially inactive, suggesting that the foot pocket extending beyond the zinc binding domain of HDAC8 cannot accommodate bulky, planar, ring structures.^25,33,58

Table 1:

HDAC8 inhibitory activity IC₅₀ values for a selection of the best performing ZBGs from Huryn’s study.³⁴

	HDAC8 IC50 (μM)
	0.072
	120
	140
	130
	150

Looking more closely at the ortho-aminoanilide ZBG, the Class I selective HDAC inhibitor Tucinodinostat (chidamide (24)), also known as CS055, which has received approval from the Chinese FDA to treat T-cell lymphoma,⁵⁹ demonstrates IC₅₀s of 95, 160 and 67 nM IC₅₀s to HDACs 1,2 and 3, respectively but an IC₅₀ of only 733 nM to HDAC8.⁵⁸ In addition to chidamide, mocetinostat (25)²⁸ and MS-275 (26)²⁹ are two additional HDAC inhibitors that have advanced to clinical trials in the United States as selective HDAC1/2/3 inhibitors. While chidamide is only 10-fold selective for HDAC1/2/3 over HDAC8, Mocetinostat and MS-275 are >100-fold selective, presumably ascribed to their different capping groups.^28,29 Given that these compounds do not adopt an “L-shape”, the lack of HDAC8 activity may be due to the para-positioning of the capping group, the inability of HDAC8’s foot-pocket to accommodate a bulky, rigid structure, such as the ortho-aminoanilide, or a combination of these factors. In sharp contrast, Whitehead and colleagues discovered the two HDAC8-selective compounds 27 and 28 (Figure 8) that incorporate a chelating α-aminoketo ZBG and associated chlorobenzyl groups which were found to occupy the foot pocket, making van der Waals contacts with the side chains of I34, W141, G303, G304 and Y306 (hHDAC8 numbering); in particular π-π stacking was observed with W141 (Figure 8B).^60,61 Indeed, compound 28 was included in Huryn’s study as a control compound, and the data acquired was in reasonable agreement with Whitehead’s data.³⁴ Taken together, these findings suggest the presence of a highly discriminatory foot pocket in the HDAC8 active site that is accommodating of benzylic moieties but not more rigid phenyl moieties. We speculate that increased flexibility, and smaller aromatics, such as furans, and possibly also small, cyclic aliphatic groups may be accommodated by the HDAC8 active site in place of the phenyl ring, particularly in light of the hydrazide discussion below, although achieving selectivity presents another challenge and will require extensive SAR campaigns.

Figure 8:

(A) Structures and inhibitory activities of HDAC8-selective inhibitors sporting the α-aminoketo ZBG; (B) a co-crystal structure of 28 bound to hHDAC8 (PDB ID: 3SFH), highlighting the foot pocket residues in close proximity to the o,p-dichlorobenzyl group.

Acyl hydrazides (RCONHNH₂) have been deployed as the ZBG in HDACi’s for over two decades, and have been extensively reviewed elsewhere.¹³ In the context of drug molecules, there are some general advantages of the hydrazide versus the hydroxamic acid motif, including better oral bioavailability and greatly-reduced toxicity, as supported by negative results in the Ames mutagenicity test.¹³ The most widely evaluated HDAC inhibitor utilizing a hydrazide as a ZBG is UF010 (29, Figure 9), which inhibits HDAC1, 2, 3 and 8 with IC₅₀s of 0.45, 0.133, 0.190, and 2.83 μM, respectively. Likewise, derivative SR-3208 (30) inhibits HDAC1, 2 and 3 with IC₅₀s of 0.23, 0.88, 0.12 μM, respectively).¹³ Besides the stated advantages of the hydrazide as a ZBG, functionalization of the distal (non-acylated) nitrogen permits the probing of foot pockets that may then yield isoform specificity. Since the ortho-aminoanilides achieve HDAC1/2/3 selectivity over HDAC8, it was presumed HDAC8’s foot pocket is not as pronounced compared to other class I isozymes.^{27–29,33,58} Based on these findings, one might conclude the introduction of increasingly lengthier alkyl groups at this terminal nitrogen may translate into higher selectivities for HDAC1, 2, and 3 over HDAC8. Sippl’s group delved more deeply into this speculation, and expanded on the earlier finding that propyl or butyl hydrocarbon chains attached to the terminal nitrogen was a source of HDAC3 selectivity.⁶² Surprisingly, however, they discovered that replacing these groups with a hexyl hydrocarbon chain afforded HDAC8 selectivity, as in 31 (Figure 10).³³ Crucially, they also demonstrated that this HDAC8-selective ZBG/FPG could be transposed onto different HDACi scaffolds, as in 32 and 33 (Figure 9), providing some evidence of the generality of this ZBG/FPG. Shorter alkyl chains (n=1–4) were less selective for HDAC8.³³ For example, 32d with an n-hexyl hydrazide potently inhibited HDAC8 with an IC₅₀ of 36 nM, and selectivities of 50-fold over HDAC1, and >300-fold over HDAC2 and >500-fold over HDAC3. Reducing the n-hexyl group to a shorter n-butyl (32c) or n-propyl (32b) yielded compounds that were similar to 32d in inhibitory activity of HDAC1–3 but crucially, the IC₅₀’s for HDAC8 were reduced by an order of magnitude, which resulted in significant erosions of isozyme selectivity.

Figure 9:

The structures of class I HDACi’s containing a hydrazide ZBG-FPG.

Figure 10:

Structures and HDAC activity inhibition profiles (IC₅₀ values in μM) of three selective HDAC8i’s incorporating an alkyl hydrazide ZBG.²⁸

The authors performed some molecular modeling studies towards rationalizing these isoform selectivities. In HDAC8, the hydrazide of 31 chelates the zinc ion, whilst its associated n-hexyl group forms hydrophobic interactions with I34 and W141 in the foot pocket (hHDAC8 numbering). However, 31 appears unable to chelate the zinc ion in HDAC1–3, accounting for the reduced activities. It should also be noted that 32d exhibits much reduced selectivities for HDAC8 over the other HDACs than does 31. While it is acknowledged that 31 and 32d are not isomers of each other, the most significant difference is that with its meta-substituted linker, 31 adopts an “L-shape” that is optimal in HDAC8 inhibitor design, as previously discussed; molecular modeling indicates the meta-aminobenzyl group may form interactions with F152 and Y306 in the side pocket of HDAC8 (hHDAC8 numbering). On the other hand 32d has a para-substituted linker, so it would be predicted (see Linker section) its affinity to HDAC8 and selectivities would be reduced, which was observed experimentally. In summary, these data demonstrate that flexible aliphatic chains – optimally an n-hexyl group – can occupy the foot pocket of HDAC8 despite it not being able to accommodate bulkier rigid groups like anilides.^{25,28,29,33,58} Furthermore, since 32d is a potent inhibitor of HDAC8 with good selectivities over HDAC1–3, these data hint that the “L-shape” hypothesis may apply only when the ZBG is a hydroxamic acid. Given the potential benefits of the hydrazide ZBG, it is important to note that compounds 31-33, are comparably selective and equipotent with the hydroxamic acid-containing, and “L-shaped” PCI-34051 (Figure 3A). While no co-crystal structures of a hydrazide-based HDAC8-selective inhibitor have been solved, the hypothesis that the HDAC8 foot pocket can accommodate flexible aliphatic chains is evidenced by the co-crystal structure in Figure 11 of an N-alkylated hydroxamic acid 34 in the SmHDAC8 active site (PDB: 6FU1).⁶³ As reported by Simoben et al.,^27,63 the hydroxamate portion of 34 interacts with the zinc ion as expected, and the associated alkyl chain interacts with hydrophobic residues F21, W140, and C152 in the foot pocket (SmHDAC8 numbering). While the hydroxamate motif itself is undesirable, this crystal structure provides crucial insights into the ability of the HDAC8 foot pocket to accommodate flexible alkyl chains: further enhancing the excitement surrounding the hexyl hydrazide and its potential as a safer, alternative ZBG that also elicits HDAC8-selectivity.

Figure 11:

Crystal structure of the N-alkylated hydroxamic acid 34 in the smHDAC8 active site (PDB ID: 6FU1).⁵¹

Recently, N-thiomethylated azetidinones (β-lactams) have featured as HDAC8-selective ZBG.⁶⁴ For example, the generic β-lactam compound 35 had no measurable activity against HDAC8, but compound 36 (Figure 12) demonstrated an IC₅₀ for HDAC8 of 4.5 μM, with a >222-fold selectivity over HDAC6. This demonstrates that the N-thiomethyl group is crucial for biological activity, however N-S bonds are weak covalent bonds,⁶⁵ so it remains to be seen how effective compounds with this ZBG will be in cells, and also in vivo. Last, Meyer-Almes et al. described a pyrimido[1,2-c][1,3]benzothiazine-6-imine scaffold, for example compound 37 which demonstrated potent inhibition of HDAC8 (IC₅₀ = of 0.0029 μM) and >500-fold selectivity for HDAC8 over all other isozymes with IC₅₀s of 1.7 μM, >50 μM, 6.7 μM, and 2.8 μM for HDAC1, 2, 3, and 6 respectively.⁶⁶ In a follow-up publication, the same group discovered that 37 undergoes a rearrangement to reveal a thiophenol that forms a disulfide with C153 in the HDAC8 active site. For this reason, the authors conceded that 37 will likely have multiple off-target effects,⁶⁷ and that, although the pyrimido[1,2-c][1,3]benzothiazine-6-imine scaffold is an intriguing model, it is not therapeutically relevant.

Figure 12:

Structures of the inactive generic β-lactam (35), an azetidinone-incorporating compound (36) and a pyrimido[1,2-c][1,3]benzothiazine-6-imine compound (37).

Conclusions

HDAC8 has a few distinct active site features that are targetable for its selective inhibition. Most importantly is the major hydrophobic groove caused by a separation of the L1 and L6 loops that renders a hydrophobic pocket accessible for ligand binding. Specifically, this domain accommodates the capping group of the canonical HDACi pharmacophore: optimal binding is realized when the capping group is hydrophobic to elicit Van Der Waals interactions, preferably contains an aromatic moiety to π-π stack with Y306 (SmHDAC8 Y341), and crucially should be positioned in the meta- (or analogous) position on the linker with respect to the ZBG. Given the requisite ZBG serves as an anchor point, this meta-positioning will geometrically oblige the hydrophobic capping group to be delivered towards the hydrophobic groove while simultaneously ensuring it is excluded from the analogous domains on the surfaces of other isozymes due to their L1-L6 locked ridges, thereby affording selectivity. Beyond these criteria, there is good tolerance of structural diversity in the capping group (Figure 5), permitting other aspects such as bioavailability and solubility to be given due consideration. On the other hand, while there exists room for chemical creativity in the capping group, which may also feature bioavailability considerations, diversity in the linking group is quite limited, again due to the necessity of the rigid “L-shaped” overall structural geometry. With the ZBG bound to the zinc ion, the depth of the zinc binding pocket and position of the hydrophobic groove is mostly targeted with an aromatic mono- or bicycle scaffold with a handful of exceptions. Those few selective HDAC8i’s that have incorporated more flexible alkyl linking groups require rigid, cyclic capping groups as a compensatory mechanism of facilitating formation of the optimal L-shaped molecular geometry. Figure 13 provides an overview of some of the chemical moieties that have been used in promoting HDAC8 affinity and selectivity, along with the more prominent residues with which the various constituents interact (hHDAC8 numbering shown).

Figure 13:

Most potent and selective HDAC8 inhibitors adopt, or are constrained in, a molecular “L-shape” for optimal interactions with HDAC8.³⁵ Included here is an example of some of the chemical constituents – and the more prominent residues with which they interact – of the canonical HDACi pharmacophore that have featured in HDAC8 inhibitors: C = capping group; L = linker; ZBG = zinc-binding group; FPG = foot pocket-targeting group. For clarity, residue numbering shown for only hHDAC8.

Hydroxamic acids have gained notoriety for their lack of selectivity, toxicity and poor PK properties,^25,32–35 therefore the ZBG, which is most commonly represented by a hydroxamic acid, is perhaps the most important pharmacophore component towards the development of clinically relevant and selective inhibitors. All other aspects being equal, unsubstituted acyl hydrazides as ZBG are a little less potent than their hydroxamic acid counterparts but crucially also less toxic.¹¹ Very recent research has revealed that alkylation of the primary amino group of acyl hydrazides may contribute to both enhanced affinities and selectivities of HDAC inhibitions. Seminal work in this area was recently reported by Sippl’s group,²⁸ who expanded research into the hydrazide functionality as a surrogate for the traditional hydroxamic acid, which yielded further insights into the nature of the HDAC8 foot pocket suggesting that the foot pocket is able to accommodate alkyl chains, opening the door towards the development of clinically relevant and selective HDAC8i. Indeed, they discovered that n-hexyl-substituted hydrazides (RCONHNH(n-hex)) are accommodated in the HDAC8 foot pocket. Moreover, this functionality also elicited selectivity for HDAC8 over the other class I HDACs and HDAC6. As a result, incorporation of a hexyl hydrazide in an inhibitor with a rigid linker and hydrophobic arm capping group geometrically positioned in an “L-shaped” conformation with respect to the hydrazide should be the blueprint towards the development of future HDAC8-selective inhibitors and clinical relevancy. We speculate that, although ortho-aminoanilides are inactive towards HDAC8, given the success with Sippl’s hydrazide research in acquiring HDAC8 affinity and selectivity, careful re-exploration of the ortho-aminoanilide ZBG may yield novel HDAC8-selective ZBGs. Further, although α-aminoketo groups as HDAC8-selective ZBGs have not gained traction, the published co-crystal structures of 27 and 28 may provide helpful insight into future HDAC8 inhibitor design, particularly in terms of strategies to engage the foot pocket.

Last, targeted protein degradation, such as with bivalent PROTACs, has emerged as a new therapeutic modality wherein the pathogenic protein may be completely eliminated, rather than merely inhibited.⁶⁸ Meanwhile, (non-degrading) bivalent inhibitors that inhibit both HDAC6 and additional targets through the incorporation of a secondary ligand have recently been reported.⁶⁹ With the advent of synthetic strategies to acquire HDAC8-selectivity, the stage is set for the development of HDAC8 PROTACs⁷⁰ and also HDAC8 polypharmacological agents.^71–73

Figure 7:

The structure of the ortho-aminoanilide-containing HDAC1/2/3-selective inhibitors.

Abbreviations

FPG

foot-pocket-targeting group

HDAC

histone deacetylase

HDACi

histone deacetylase inhibitor

HDAC8

histone deacetylase 8

ZBG

zinc-binding group

Sm

Schisotosomsa mansoni

h

human