Recent advances with KDM4 inhibitors and potential applications

Abstract

The histone lysine demethylase 4 (KDM4) family plays an important role in regulating gene transcription, DNA repair, and metabolism. The dysregulation of KDM4 functions is associated with many human disorders, including cancer, obesity, and cardiovascular diseases. Selective and potent KDM4 inhibitors may help not only understanding the role of KDM4 in these disorders but also provide potential therapeutic opportunities. Here, we provide an overview of the field and discuss current status, challenges and opportunities laying ahead in the development of KDM4-based anticancer therapeutics.

1. Introduction

Histone methylation plays key roles in regulating chromatin structures, gene transcription and DNA repair. Histone methylation was initially thought to be irreversible until the discovery of lysine-specific demethylase 1 (LSD1).¹ The dynamic methylation of histone lysine residues is regulated by histone methyltransferases (KMTs) (‘writers’) and demethylases (KDMs) (‘erasers’). KDMs are classified into two families based on the catalytic mechanisms. LSD1 (KDM1A) and LSD2 (KDM1B) are flavin adenine dinucleotide (FAD)-dependent amine oxidases,^{1, 2} while JmjC-domain containing demethylases (JmjC-KDMs), including families of KDM2-KDM8, are 2-oxoglutarate (2-OG) and Fe(II)-dependent hydroxylases. The mechanism of demethylation is a dioxygenase reaction.³ Briefly, an electron is transferred from the Fe(II) to the coordinated molecular oxygen, yielding Fe(III) and a superoxide radical that attacks the carbonyl group (C2) in the 2-OG. Decarboxylation of 2-OG ensues, producing succinate and CO₂. During the split of molecular oxygen, a highly unstable Fe(IV)-oxo intermediate is produced, which extracts a proton from the methylated lysine. The resulting Fe(III) hydroxide subsequently hydroxylates the radical on the methyl group, forming a carbonyl group that spontaneously demethylates the lysines such as H3K9me3 (Figure 1A). KDMs demethylate a range of histone lysine residues, including H3K4, H3K9, H3K27, H3K36, H4K20 and H1.4K26. KDM4 specifically catalyzes the removal of di- and tri-methylated lysine 9, lysine 36 of histone 3 and H1.4K26.^{4, 5}

Figure 1.

(A) Demethylation mechanism of KDM4s. 2-OG shows in blue; H3K9me3 shows in red; (B) Protein architecture of KDM4s.

Overexpression of KDM4 proteins is prevalent in various cancers, including breast,⁶ prostate,⁷ colorectal,⁸ lung,⁹ blood,¹⁰ neuroblastoma,¹¹ malignant pleural mesothelioma,¹² and other cancers. Genetic depletion or pharmacologic inhibition of KDM4 has an anticancer effect,^11–25 making KDM4 an attractive target for the development of small molecule inhibitors. Several KDM4 inhibitors with different chemotypes have been reported; however, only one selective and potent KDM4 inhibitor, has entered in phase 1 clinical trial (TACH101), which is currently being evaluated for the treatment of gastrointestinal and high microsatellite instability (MSI-H) metastatic colorectal cancers.²⁶

Here, we provide an overview in the development of KDM4 inhibitors. Firstly, we discuss the structural features and functions of the KDM4 protein family. Then, we review literature reported KDM4 inhibitors, which we classified on the basis of their targeted protein domains (e.g., catalytic domain and reader domain) and chemical structures. We also examine the impact of 2-oxoglutarate competition on the cellular activity of KDM4 inhibitors. We finish with the field outlook, discussing opportunities and challenges in the development of KDM4 inhibitors for clinical applications.

2. Structure of KDM4s

The KDM4 gene family consists of 6 members: KDM4A-F, of which KDM4E and KDM4F have been reported as protein noncoding pseudogenes.²⁷ All KDM4 proteins contain JmjN and JmjC domains that are responsible for the demethylation activity. While the JmjC domain is the catalytic domain that is conserved across all the KDM4 members, the JmjN domain interacts with JmjC domain²⁸ and stabilizes the KDM4 proteins (Figure 1B).^{29, 30} KDM4A-C also have non-catalytic domains including two Tudor domains and two plant homeodomains (PHD) domains, which are absent in KDM4D-F. Both the Tudor and PHD domains are reader domains that recognize specific histone lysine tails.^{31, 32} KDM4A-C catalyze the removal of methyl groups from both H3K9 and H3K36, however, KDM4D-E can only demethylate H3K9.³³ At least some of the selectivity of the KDM4 subfamily residues arises from residues in or close to the catalytic domain.³³Selectivity arising from the catalytic domain may be relatively more important for those members (e.g. KDM4D/E) that do not contain additional domains.³³ Interestingly, when bound to KDM4A, H3K9me3 lysine side chain adopts a ‘W’–shaped conformation, whereas H3K36me3 forms ‘U’–shaped conformation.³⁴ The Tudor domains of KDM4A may also bind to H3K23me3, H3K4me3 and H4K20me3.^{35, 36} While the KDM4B Tudor domains display exclusive binding to H3K23me3, the KDM4C Tudor domains bind specifically to H3K4me3.³⁵ The biological relevance of the selectivity of KDM4 reader domains remains to be elucidated.

3. Function of KDM4s

KDM4A-C are broadly expressed in normal human tissues, whereas KDM4D protein is predominantly found in the human testes with much lower expression levels in other tissues.³¹ KDM4A-D are involved in a variety of physiological functions, including stem cell self-renewal,^37–41 hematopoietic stem cell maintenance,⁴² cell differentiation,^{43, 44} cell cycle,^{45, 46} DNA damage repair,^{38, 47–51} metabolism,^{52, 53} and spermatogenesis.⁵⁴ One study demonstrated that KDM4A promotes cardiac hypertrophy in response to hypertrophic stimuli in mice.^{55, 56} KDM4A was also found to regulate protein synthesis, an unexpected cytoplasmic and non–chromatin-related role.⁵⁷

Dysregulation of KDM4s has been found in various diseases, including cancers. For example, KDM4A and KDM4B have been noted to cause copy gain at site-specific chromosomal loci such as 1q12 and 1q21,^58–60 which could be problematic for targeted therapies as amplifications (i.e., MCL1 in 1q21) are a mechanism of therapeutic resistance. KDM4A-C overexpression and amplification has been observed in several different subtypes of breast cancer. KDM4A and KDM4D were found overexpressed in 60% of breast cancers,⁶ whereas KDM4B overexpression was observed only in estrogen receptor (ER)-positive subtype breast cancer,⁶¹ as it plays an important role in estrogen-mediated signaling.^62–64 KDM4C is only amplified and overexpressed in basal-like subtypes.⁶⁵ We and others have shown that KDM4A and KDM4B form complexes with ERα and activate ERα-mediated transcription.^{64, 66} Moreover, KDM4B and KDM4C are targets of HIF-1α, driving tumor growth.^{23, 64} Androgen receptor (AR) is a transcriptional factor that plays a crucial role for prostate cancer development. KDM4A-D proteins function as coactivators of AR.^{25, 67, 68} Similar to KDM4C that forms a complex with AR,²⁵ KDM4A and KDM4D interact with AR and promote AR-dependent gene expression.⁶⁷ KDM4B facilitates AR-related gene transcriptional activation by demethylating histone proteins at androgen-regulated chromatin sites and by stabilizing the AR protein.⁶⁸ An additional oncogenic mechanism of KDM4B in prostate cancer is induction of alternative splicing of AR-V7,⁶⁹ an AR isoform that lacks a ligand binding domain and plays an important role in progression of castration-resistant prostate cancer. KDM4A-C are overexpressed in colorectal tumors and are required for tumor cell growth.⁸ KDM4A forms a complex with P53 and depletion of KDM4A causes reduced proliferation in colorectal cancer (CRC) cells.⁷⁰ KDM4A suppresses the p53 pathway and facilitates cellular transformation by cooperating with Ras in lung cancer cells.⁷¹ Depletion of KDM4A causes senescence, restraining the proliferation potential of cancer cells.⁷¹ KDM4B is involved in CRC tumorigenesis and progression by upregulating hypoxia-inducible genes.⁷² Recently, KDM4B has also been found to promote glucose metabolism, thus driving CRC progression.⁷³ KDM4C regulates MALAT1 expression, which leads to the enhanced β-catenin signaling in CRC cell metastasis.⁷⁴ KDM4C is overexpressed in lung cancer and correlates with poor prognosis in patients,⁹ and promotes radioresistance by activating TGF-β2/Smad signaling.⁹ Both KDM4A and KDM4C are required for the survival of acute myeloid leukemia (AML) cells. Knockdown or pharmacologic inhibition of KDM4A and KDM4C induced differentiation and inhibited proliferation of AML cells.^{22, 75} However, KDM4A has a different function from KDM4C in AML and knockdown of KDM4A and KDM4C led to different transcriptional changes.⁷⁵ We have shown that KDM4B protein is highly expressed in MYCN-amplified neuroblastoma, and promotes tumor progression by interacting with N-Myc protein and regulating the Myc pathway.¹¹ KDM4B regulates the proliferation, differentiation, and tumor growth in neuroblastoma models.

The biological consequences of KDM4 inhibition include the induction of cellular differentiation.^{11, 76} inhibition of cancer stem cell proliferation,¹⁶ cell cycle arrest and cell death, probably via repression of oncogenic MYC pathways and hormone-mediated signaling,^{11, 17–19} and other cancer-related pathways. Due to the important functions of KDM4s in various cancers,^{13, 77, 78} developing small molecular inhibitors targeting KDM4s might be a promising strategy for cancer therapy.

4. KDM4 inhibitors

Given the importance of KDM4s in cancers, tremendous efforts have been made in the development of KDM4 inhibitors. As KDM4s are 2-oxoglutarate (2-OG) dependent dioxygenases, most of the small molecule inhibitors are 2-OG analogs and their derivatives that chelate the Fe²⁺ in the catalytic binding pocket, thereby suppressing KDM4 activity. However, most of these inhibitors are lacking selectivity and/or cell permeability.^79–87 Peptide-like inhibitors were also developed by mimicking the KDM4 substrate H3K9me3/K36me3 peptides and showed non-2-OG competitive properties. Some natural products metabolites of bacteria and plants have been identified as KDM4 inhibitors. While most of the KDM4 inhibitors target the JmjC catalytic domain, several groups reported inhibitors targeting the reader domains.

4.1. Metal-Chelating inhibitors

Hydroxamic acid inhibitors

Compounds 1 - 5 (Figure 2A) bearing hydroxamic acid group were found to mimic 2-OG and bind the Fe²⁺ and inhibit KDM activity. HDAC inhibitors that commonly contain the hydroxamic acid group, such as trichostatin A (TSA, 1) and suberoylanilide hydroxamic acid (SAHA, 2), were shown to inhibit KDM4E with IC₅₀ values of 28.4 μM and 14 μM respectively.⁸⁸

Figure 2.

Structures of Hydroxamic acid inhibitors (A) and Hydrazide inhibitors (B)

Recently, we and others have identified ciclopirox (CPX, 3), an antifungal agent, as a pan-KDM inhibitor.^{76, 89} CPX inhibited KDM4B with an IC₅₀ of 3.8 μM in a TR-FRET assay and showed potent anti-proliferative activity against neuroblastoma cells (IC₅₀: 0.2 – 2.7 μM), but not normal human fibroblast HS68 cells, which suppressed tumor growth in mouse models.⁷⁶

Compound 4a is a KDM4 inhibitor (IC₅₀: 4.3 – 5.9 μM, ESI-TOF MS assay) developed through a bivalent strategy that combined substrate mimic (HDAC inhibitor MS-275) with a 2-OG cofactor mimic (hydroxamate scaffold).⁹⁰ Its methyl ester prodrug 4b (Methylstat) blocked KYSE150 cell growth (GI₅₀ = 5.1 μM) and induced H3K9me3 accumulation in an immunostaining-based assay (EC₅₀ = 8.6 μM, KYSE150; 6.3 μM, MCF7). Methylstat was further modified by introducing a fluorescein isothiocyanate to develop a fluorescent probe (5), which was used in fluorescence polarization (FP) assays for the screening of KDM4A and KDM2A inhibitors.^{91, 92}

Hydrazide inhibitors

Similar to hydroxamic acid containing compounds, hydrazides such as 6 and 7 (Figure 2B) also chelate the catalytic ferrous ion. Tetrazolylhydrazide 6 inhibits KDM4A with an IC₅₀ of 46.64 μM in a formaldehyde dehydrogenase (FDH)-coupled assay and 2.38 μM in the antibody-based LANCE Ultra assay.⁹³ Cocrystal structure of compound 6 in complex with KDM4D (PDB ID: 6ETS) showed that the hydrazide moiety coordinates to the active site Ni²⁺ (a Fe²⁺ replacement for crystallization purpose) with a binding mode similar to the 2-OG cofactor.⁹⁴ Whereas the plant growth regulator daminozide is a selective KDM2/7 inhibitor, Compound 7, which has the two methyl groups removed from daminozide, exhibited selective KDM4E inhibition with IC₅₀ values of 0.2 μM in an AlphaScreen assay.⁸⁹

Pyridine – based inhibitors

2,4-pyridinedicarboxylic acid (8, PDCA) is a non-specific 2OG oxygenase inhibitor showing potent enzymatic inhibition across the JmjC-KDMs family,^{89, 95} as well as collagen and HIF prolyl hydroxylases.⁹⁶ Like the binding mode of 2-OG, PDCA chelates the Fe²⁺ through its pyridine nitrogen and carbonyl oxygen in a bidental way (PDB ID: 2VD7). The pyridine scaffold is a common motif found in KDM4 inhibitors (Figure 3).

Figure 3.

Structures of pyridine - based inhibitors

3-amino-4-pyridinecarboxylic acid 9a lacking the 2-carboxylate group has also been reported as KDM4 inhibitor selective over KDM6B and EGLN3 (pIC₅₀: 6.2, 5.8 for KDM4C and KDM4D, RapidFire mass spectrometry (RFMS)).⁸³ The x-ray crystal structure of 9a (PDB code: 5FP9) showed that the pyridine nitrogen formed a single coordination bond with the metal ion.

Based on this scaffold, GlaxoSmithKline (GSK) developed a cell penetrant KDM4/5 potent dual inhibitor 10 with an IC₅₀ of 79 – 126 nM in a RFMS assay.⁸³ In a follow-up study, the 3-aminopyridine-4-carboxylate scaffold was optimized to a pyrido[3,4-d]pyrimidin-4(3H)-one, which retained key interactions to the proteins, thus eliminating the carboxylate group leading to improvements in physicochemical properties.⁸⁴ Compound 11 also showed cellular activity in a KDM4C/5C cell imaging assay (IC₅₀: 5 μM and 4 μM).

Similar to 11, compound 12, featuring the same pyrido[3,4-d]pyrimidin-4(3H)-one scaffold, was found to be a dual KDM4/5 inhibitor (KDM4B IC₅₀ = 31 nM, KDM5B IC₅₀ = 23 nM, AlphaScreen assay). Compound 12 exhibited cell permeability and increased the methyl marks H3K9me3 and H3K4me3 in Hela cells overexpressing KDM4A and KDM5B demethylases.⁸² Improved affinity for KDM4A was achieved by engagement of residues E169 and V313 by compound 13, a spirocyclic analogue of 12 (KDM4A Ki = 4 nM, KDM5B Ki = 7 nM). The dramatic reduction from biochemical to cell-based activity is likely due to the inhibitor competition with 2-OG.⁸⁶

QC6352 (compound 14) disclosed by a group at Celgene was based on a fragment lead 3-(methylamino)isonicotinic acid 9b (KDM4C IC₅₀ = 1400 nM) developed using a structure-based drug design. QC6352 is a potent and selective KDM4 inhibitor (KDM4A-D IC₅₀ = 35 nM −104 nM, LANCE TR-FRET assay), which displays strong antiproliferative effects (EC₅₀ = 3.5 nM) against the KYSE-150 cell line but has no effect on normal fibroblast cell line IMR-90. Compound 14 upregulated the KDM4 substrates H3K36me3 in KYSE-150^+KDM4C cells with an EC₅₀ of 1.3 nM. In the breast cancer PDX models, QC6352 diminished tumor growth and reduced the tumor initiating cell populations.⁸¹

The imidazopyridine-7-carboxylate, compound 15, has been identified as a selective KDM4A-C inhibitor with >50 fold selectivity over any other 2-OG-dependent dioxygenases, including KDM4D, in a chemical proteomics assay.⁹⁷

EPZ020809 (compound 16) was developed as a KDM4C inhibitor (Ki = 31 nM) by Epizyme, using a high-throughput mass spectrometry assay.⁹⁸ A crystal structure of KDM4A complexed with compound 16 (PDB code: 4GD4) showed that the nitrogen of pyridine and pyrrole ring chelates the metal ion bidentally in a 2-OG-competitive pattern.

Molecular docking of a ZINC fragment library followed by fragment linking and hybrid optimization resulted in a pan-JmjC KDM inhibitor (compound 17) with IC₅₀ value of 64 nM against KDM4C (2 μM 2-OG, TR-FRET assay).⁹⁹ Compound 17 formed 1,5-bidentate coordination of iron through pyridine nitrogen and phenol oxygen (PDB code: 5A7W) and exhibited competition with 2-OG.

JIB-04 (compound 18) was identified as a KDM4/5 inhibitor (KDM4A-E IC₅₀ = 290 – 1100 nM; KDM5A IC₅₀ = 230 nM, ELISA assay) through cell-based screening of NCI’s diversity set library.⁸⁰ JIB-04 blocked the proliferation of lung cancer, prostate cancer,⁸⁰ rhabdomyosarcoma,¹⁰⁰ and germ cell tumor cell lines,¹⁰¹ but not the normal cells,⁸⁰and suppressed tumor growth in lung cancer, breast cancer,⁸⁰ Ewing sarcoma,¹⁰² hepatocellular carcinoma¹⁰³ xenograft mouse models. Interestingly, JIB-04 is not a competitive inhibitor of 2-OG, though it bears a metal chelating moiety.⁸⁰ Mechanistic studies indicate that JIB-04 disrupts the binding of O₂ and histone substrates in the KDM4A active site by interacting with Lys 241 and Tyr 177 through hydrogen bonding.⁸⁵ Additionally, the inhibitor may chelate the metal center, thereby altering the binding ability of the co-substrate 2-OG. Interestingly, this study also showed that JIB-04 more effectively inhibits KDM4A in low-O₂ environments, a further indication that the inhibitor is competitive with respect to O₂.⁸⁵

Structure optimization of two hits from virtual screening by Fang at al. led to the discovery of a KDM4D inhibitor (compound 19) with an IC₅₀ of 0.41 μM.¹⁰⁴ The crystal structure of compound 19 in complex with KDM4D (PDB code: 7DYQ) showed that pyridyl nitrogen, instead of the nitrogen of nitrile, coordinates to the metal ion.¹⁰⁵

8-Hydroxyquinoline – based inhibitors

8-hydroxyquinoline (8HQ) derivatives (Figure 4A), which were first identified as 2-OG oxygenase inhibitors of EGLN,¹⁰⁶ were also found to be KDM4 inhibitors. 5-carboxyl-8HQ (compound 20, IOX1) is a KDM4 inhibitor (KDM4A IC₅₀: 1.7 μM, MALDI-TOF MS assay) discovered in a formaldehyde dehydrogenase (FDH)-coupled high-throughput screening assay.¹⁰⁷ Structure-guide modification of IOX led to a cellular permeable KDM4E probe ML324 (compound 21) with IC₅₀ value of 920 nM in an AlphaScreen assay, which displayed potent anti-viral activities.¹⁰⁸ Another 8HQ derivative B3 (compound 22) is a KDM4 inhibitor (KDM4B IC₅₀ = 10 nM, antibody-based fluorometric assay) with weak inhibition of KDM5A and LSD1. Compound 22 inhibited the growth of AR-positive cell lines and breast cancer cell lines with IC₅₀ values of micromolar range in a MTT assay. In addition, compound 22 showed tumor growth inhibition in a PC3 xenograft mouse model.¹⁰⁹ SD70 (compound 23), first discovered in a phenotypic screening of ligand and genotoxic stress-induced translocations in prostate cancer cells, was identified as a KDM4C (IC50 = 30 μM, antibody-based assay) inhibitor.¹¹⁰ SD70 blocks proliferation of prostate cancer cell CWR22Rv1 (concentrations > 5 μM) and inhibits the tumor growth in xenograft mouse model.

Figure 4.

Structures of 8-Hydroxyquinoline - based inhibitors (A), Benzimidazole – based inhibitors (B), and clinically used iron chelating drugs (C)

Benzimidazole – based inhibitors

By means of an FDH-coupled high-throughput screening of the ChemBioNet library, Carter et al. identified the benzimidazole compound 24a (Figure 4B) as a KDM4 inhibitor (KDM4A-E IC₅₀ = 4 – 8 μM).¹¹¹ In a subsequent study, structure optimization resulted in compound 24b with 10-fold increased potency (KDM4E IC₅₀ = 0.9 μM in FDH-based assay). 24b exhibited cytotoxicity against prostate cancer cell lines (GI₅₀ = 8 μM for LnCap and DU145 cell lines) and a non-disease control cell line (GI₅₀ = 26 μM, human prostate epithelial cells).⁸⁷ A kinetics-based study revealed that compound 24b is not competitive with the 2-OG cosubstrate. Furthermore, crystallographic study showed that the benzimidazole pyrazole scaffold binds to a novel surface-exposed binding site formed by residues E118, S207, Y209, T261, K259, F279, and F114. In conclusion, benzimidazole pyrazole compounds inactivate the enzyme by removing metal ion in the active site.

Clinically used iron chelating drugs

Screening a set of clinically used iron chelator drugs resulted in identification of deferoxamine (25), deferiprone (26), and deferasirox (27a) (Figure 4C) as KDM4A inhibitors with IC₅₀ values of 3.22 – 17.4 μM in FDH assay and 3.33 – 4.76 μM in LANCEUltra assay.¹¹² Compounds 25 and 27a also show inhibition of KDM5A and KDM6B, highlighting the general selectivity challenge for iron chelator compounds. 27a increased the histone H3K9me3 level with an EC₅₀ of 40.5 μM in KYSE-150 cells, and inhibited cell proliferation with GI₅₀ of 3.3 μM and 5.5 μM in KYSE-150 cells and HL-60 cells, respectively. Further structure modification resulted in compounds 27b and 27c with improved cell permeability, which showed higher potency in terms of upregulation of histone trimethylation (EC₅₀ 3.3 μM for 27b, and 10 μM for 27c) and more potent inhibition of cancer cell growth (GI₅₀ = 0.45 for KYSE-150 and 1.7 for HL-60 cell).¹¹² Moreover, 27c did not display cytotoxicity against non-transformed lung-derived cell lines IMR-90 and BEAS-2B.

4.2. Nonmetal-chelating inhibitors

In a recent study, Fang et al. reported a KDM4D selective inhibitor 28 (Figure 5A) with a novel scaffold through AlphaLisa-based screening and structural optimization. 28 exhibited high selectivity over other KDM4 family members (KDM4D IC₅₀ = 23 nM, KDM4A > 100 μM) and did not compete with 2-OG. Moreover, 28 displayed anti-proliferation and anti-migration in various CRC cell lines. Molecular docking results showed that 28 occupies the H3K9me3 peptide pocket, but not the 2-OG binding site.¹¹³

Figure 5.

Structures of nonmetal-chelating inhibitors (A), metal cofactor disruptors (B), and peptide-based inhibitors (C)

Screening of a drug-like diversity library resulted in three hits (29 - 31) (Figure 5A) with novel scaffolds without metal-chelating moieties as KDM4A (Ki 0.42 – 3.02 μM) and KDM2A inhibitors.⁹² These compounds displayed cellular activity as evident by the induction of H3K9me3 and H3K36me2 in MiaPaCa2 cells with IC₅₀ values 0.28 μM - 3.96 μM.

4.3. Metal cofactor disruptors

A Cys3-His Zn (II) binding site is present only in KDM4s, but not other 2-OG dioxygenases.³⁴ This Zn ion site is close to the KDM4A catalytic binding pocket, and the Zn ion is important for the KDM4A stability. Zn-ejectors such as disulfiram analogs (32a and 32b) and the selenium-containing compound Ebselen (compound 33) (Figure 5B) can destabilize and inhibit KDM4A at a range of 3 μM −15 μM in a MALDI-TOF MS-based assay.¹¹⁴ Based on the structure of Ebselen, Kim et al. synthesized a benzo[b]tellurophene compound 34 as a KDM4 inhibitor with an IC₅₀ value of 30 μM in a FDH assay.¹¹⁵ Compound 34 specifically increased the trimethylation of H3K9 but not the methylation of other lysine residues in HeLa cells and decreased the HeLa and LoVo cell viability at a concentration of 10 μM without affecting the normal human fibroblast BJ cells.

4.4. Peptide-based inhibitors

Several selective peptide inhibitors (Figure 5C) were developed through a peptidomimetic approach and showed better selectivity compared to the metal chelating inhibitors. Lohse et al. identified that H3(7–14)K9me3 and H3(7–10)K9me3 are the shortest peptides required for the catalytic activity of KDM4A and KDM4C respectively.¹¹⁶ The coupling of uracil, a known iron chelator,¹¹⁶ to the lysine 9 amino group of these peptides resulted in compound 35 with moderate KDM4A and KDM4C inhibition activities. Nielsen et al. synthesized three series of hybrid peptides where uracil-coupled lysine chain was substituted with C-substituted N-triazololysines.¹¹⁷ Compound 36 showed moderate KDM4C enzymatic activity in a TR-FRET assay (IC₅₀ =30 μM).

The same substrate fragment H3K9me3 (7–14) was employed to develop selective inhibitors of the KDM4 subfamilies. Disulfide 37 was developed as a KDM4E binder by MS analysis through cross-linking of N-oxalyl-D-cysteine (NOC) to the peptides.¹¹⁸ The cocrystal structure for KDM4A in complex with 37 (PDB ID 3U4S) revealed 2-OG and histone binding sites were occupied by NOC and the short peptide moieties simultaneously.

Kawamura et al. developed a de novo macrocyclic peptide CP2 (Compound 38a) as a potent non-2OG competitive KDM4A-C inhibitor (33 nM - 42 nM, AlphaScreen assay) with high selectivity over other KDMs by a Peptides Integrated Discovery system.⁷⁹ The crystal structure of KDM4A complexed with 38a (PDB code: 5LY1) reveals that 38a occupies the histone-binding groove of KDM4A, with Arg6 bound in the sub-pocket usually occupied by the H3K9/K36me3 residues. The positively charged residues on the 6-position are necessary for the inhibition activity. In their subsequent work, 38a was optimized to improve the cell permeability by replacing the NNK codons with NKN codons.¹¹⁹ The analogue 38b was more potent than 38a in KDM4A/C inhibition in an AlphaScreen assay (KDM4A IC₅₀: 6 nM, KDM4C IC₅₀ = 2.2 nM). Although these peptide inhibitors achieved selectivity and potency of KDM4s over other KDMs, overcoming the poor cell permeability and instability of these peptide inhibitors to achieve desired cellular activity is a big challenge.

4.5. Natural product inhibitors

A number of natural products such as (Compounds 39 - 46; Figure 6A), showed KDM4 inhibition activity.

Figure 6.

Structures of natural product inhibitors (A) and Tudor domain inhibitors (B)

Tripartin (compound 39) was isolated from the culture broth of the Streptomyces sp. associated with a larva of the dung beetle Copris tripartitus Waterhouse.¹²⁰ Tripartin is the first natural product which showed selective increase of the global H3K9me3 levels in Hela cells.¹²⁰ It was initially defined as a selective KDM4 inhibitor. However, Guillade et al. evaluated the biological activity of synthesized tripartin and found that tripartin analogues did not inhibit isolated KDM4A–E enzymes though manifesting apparent cellular activities.¹²¹ Thus, tripartin may affect histone methylation status via a KDM4-indirect mechanism.

Geldanamycin (GA, compound 40a) is an antibiotic discovered in the culture filtrates of Streptomyces hygroscopicus var. geldanus var. nova.¹²² Geldanamycin derivatives have been identified as heat shock protein 90 (Hsp90) inhibitors.¹²³ Using a TR-FRET based high-throughput screening of FDA-approved drugs and bioactive molecules,¹²⁴ our team identified GA and its analogs 17-AAG (compound 40b), 17-DMAG (compound 40c) as pan-JmjC KDMs inhibitors (KDM4B/C IC₅₀: 0.7 μM - 4.8 μM, KDM5A IC₅₀ = 0.13 μM - 0.65 μM, KDM6B IC₅₀ = 0.4 μM - 5.8 μM, Alpha Screen assay).¹²⁵ 17-DMAG upregulated H3K9me3 and H3K36me3 in the RH30 cell line, and delayed tumor growth in alveolar rhabdomyosarcoma xenograft mouse models.

Toxoflavin (compound 41) is a natural product first isolated as a toxin produced by Burkholderia gladioli from Lycoris aurea.¹²⁶ Toxoflavin has shown a variety of activities, and reported as a fungicide,¹²⁷ and antibiotic.¹²⁸ Recently, toxoflavin has been discovered via virtual screening as a KDM4A inhibitor.¹²⁹ A peptide-based histone trimethylation assay confirmed its KDM4A inhibitory activity with an IC₅₀ value of 2.5 μM. Toxoflavin directly bound and stabilized KDM4A protein in an HCT-116 cellular thermal shift assay (CETSA) and inhibited cell proliferation and upregulated H3K9me2/3 levels. Letfus et al. investigated the binding modes using molecular docking of toxoflavin into KDM4A, and predicted the that 8-N and 7-CO oxygen of toxoflavin chelated to the metal ion.¹³⁰

Based on the KDM4A inhibitor purpurogallin (compound 42a), a natural product isolated from nutgalls and oak bark, Souto et al. synthesized a series of purpurogallin analogs with improved enzymatic activity. The most potent compound 42b inhibited KDM4A-D with a modest IC₅₀ value of 24.37 μM - 38.85 μM (AlphaLISA assay), upregulated H3K9/36me3 in MCF7 cells and induced cell death in several solid cancer cell lines.¹³¹ Interestingly, 42b did not inhibit the proliferation of normal mesenchymal cell line Mepr2B.

Flavonoids are the largest family of natural products to be identified as 2-OG dioxygenase inhibitors.¹³² High-throughput screening of a LOPAC compound library using a formaldehyde dehydrogenase-coupled assay resulted in KDM4E inhibitors, of which catechol 43 (epigallocatechin), flavonoids 44 (baicalein), and 45 (myricetin) were the predominant ones.¹³³ The characterized aromatic ene–diol functionality endowed them the ability to chelate to Fe²⁺, which suggest that they may inhibit other 2-OG dioxygenases. Rose et al. also proved that catechols gallic acid (46a) and protocatechuic acid (46b) as pan-KDM inhibitors.⁸⁹

4.6. Reader domain inhibitors

Tudor domains

Recent studies have shown that the Tudor domains of KDM4 family can direct and stimulate localized histone demethylation through binding to specific lysine methylation markers.^35,134 Due to the distinct sequences and histone-binding preferences of Tudor domains between the KDM4 subfamily, targeting Tudor domains provides an alternative approach for developing selective KDM4 inhibitors.

Recently, a novel fragment 47 (Figure 6B) targeting the KDM4A-Tudor domain was identified through a 2D-NMR based fragment screening approach.¹³⁵ Screening a library of rule of three compliant fragments identified fragment 47, which exhibited a modest affinity binding to nearly 80 μM. The co-crystal structure of KDM4A-Tudor domain in complex with 47 (PDB ID: 5VAR) shows that the 47 fits into the aromatic cage formed by the three conserved aromatic residues (F932, W967, Y973). The NanoBRET-based cellular proximity assay demonstrates 47 can inhibit H3K4Me3 binding to the Tudor domain in cells with an EC50 value of 105 μM.

Screening of a NIH Clinical Collection library using the HaloTag assay identified amiodarone as KDM5A-PHD3 inhibitor.¹³⁶ Structure modification resulted in a series of amiodarone analogs with improved potency, of which compound 48a (WAG-003) and 48b (Figure 6B) showed inhibition against both PHD3 domain of KDM5A and Tudor domain of KDM4A with IC₅₀ 26 μM −72 μM in a fluorescence polarization assay.

PHD-finger domains

Like KDM-based Tudor domains, PHD domains are another class of reader domain for histone demethylases. While the function of many JmjC-KDM associated PHD-fingers remain unclear, some of them are partially characterized: KDM5 PHD domains bind to unmodified H3 peptide and activate catalytic domain-mediated removal of methyl marks from H3K4me3 peptide, suggesting that PHD domain inhibitor may allosterically stimulates the activity of the catalytic domain.^{32, 137} While a few papers reported KDM5 inhibitors targeting PHD domain,^{136, 138} no inhibitors targeting PHD domain of KDM4 has been identified.

5. 2-OG challenges

KDM4 is overexpressed in various cancers and plays a critical role in cancer development through transcriptional activation or regression of targeted genes. Being a promising target for cancer therapy, many efforts have been made to develop selective and potent KDM4 inhibitors. In the past decade, different types of inhibitors have been discovered to target different domains of KDM4, including catalytic domain and non-catalytic domain. However, most currently known inhibitors showed poor or no cellular activity, though they possess nanomolar-range inhibition in biochemical assays. While low cellular permeability may account for some of the observed decrease in cellular activity (Table 1), another potential factor could be competition between the cofactor 2-OG and inhibitors,⁸⁶ especially those metal chelating inhibitors. Despite the different chemotypes of this class of inhibitors (hydroxamic acid, hydrazide, pyridine, 8-Hydroxyquinoline), it inevitably competes with the 2-OG when forming the coordination interaction with Fe²⁺ ion in the 2-OG binding pocket. In fact, 2-OG competition is common in all JmjC-KDM inhibitors. KDOAM-25, a selective KDM5A-D inhibitor, is a partial competitor of the 2-OG (IC₅₀ = 27 nM at 4 μM 2-OG, IC₅₀ = 558 nM at 300 μM 2-OG).¹³⁹ A similar pattern was seen in CPI-455 and CPI-4203, which are KDM5 specific inhibitors.¹⁴⁰ The KDM6 inhibitor GSK-J1 is competitive with 2-OG, but not the peptide substrate.¹⁴¹

Table 1.

Properties of cellular active metal-chelating inhibitors

Cmpd	KDM4 IC50	Cell permeability	Cellular activity IC50	Cell viability IC50	Other targets
3	KDM4B: 3.8 μM			NB line: 0.2–2.7 μM	KDM4C, PHF8, KDM6A
4a-b	KDM4A/C/E: 3.4 – 5.9 μM		H3K9me3: MCF7: 6.3 μM KYSE150: 8.6 μM	KYSE150: 5.1 μM	KDM6B: 43 μM PHF8: 10 μM
10	KDM4A/C/D/E: 40 – 100 nM	ChromLogDpH7.4 : 0.1	Cell imaging: KDM4C: 7.9 μM		KDM5C: 100 nM KDM6B: 12 μM
11	KDM4A/C/D/E: 398 – 630 nM	ChromLogDpH7.4 : 1.8	Cell imaging: KDM4C: 5 μM		KDM5C: 63 nM
12	KDM4A: 102 nM KDM4B: 31 nM	Caco-2: 11.8×10−6 cm/s	Cell imaging: confirmed		KDM5B: 23 nM KDM5C: 65 nM
13	KDM4A: 100 nM KDM4B: 43 nM	Caco-2: 11.6×10−6 cm/s	IF KDM4A: 4.7 μM InCELL Hunter: KDM4B: 1.3 μM		KDM5B: 38 nM KDM5C: 123 nM
14	KDM4A-D: 35 – 104 nM	PAMPA: 51.2 nm/s	H3K36me3: 1.3 nM	KYSE150: 3.5 nM	KDM5B: 750 nM
18	KDM4A-E: 290 – 1100 nM			H358: 100 nM A549: 250 nM	KDM5A: 230 nM KDM6B: 855 nM
21	KDM4E: 920 nM	Caco-2: 12.5×10−6 cm/s	Viral IE infection inhibition: 10 μM
22	KDM4B: 10 nM KDM4A/C/D			PC3: 40 nM LNCaP, VcaP: submicromolar
23	KDM4C: 30 μM
24b	Pan-KDM4 KDM4E: 900 nM			LnCap, DU145: 8 μM
27c	Pan-KDM KDM4A: 4.1 μM		H3K9me3: 10 μM	KYSE150: 0.46 μM HL60: 1.7 μM

To overcome the 2-OG competition, developing covalent KDM4 inhibitors could be a potential strategy to enhance cellular activity, which has been proven to be feasible in KDM5 covalent inhibitors.¹⁴² Another way to overcome the 2-OG competition is to degrade KDM4 using Proteolysis Targeting Chimeras (PROTAC) technology. Unlike the traditional small molecule inhibitors, which require high levels of sustained occupancy, the event-driven catalytic mechanism of action makes PROTACs effective at very low, catalytic concentrations.¹⁴³ Recently, a KDM5C PROTAC has been developed,¹⁴⁴ suggesting that demethylases are tractable targets for proteasomal degradation.

However, the disconnect between potent biochemical activity and weak cellular activity for Fe chelating inhibitors may not be necessarily attributed to 2-OG competition. Compound 14 is a Fe chelator and possesses potent cellular activity in vitro and in vivo. Presumably in these settings, the concentrations of endogenous 2-OG are high. Furthermore, non-chelating inhibitors do not appear to show much cellular activity as well. These findings suggest this is a general challenge in targeting the KDMs family.

6. Perspectives of KDM4 inhibitors in clinical applications

Oncogenic transcription factors such as MYC, AR, ER, and PAX3-FOXO1 are drivers of multiple cancers including neuroblastoma, prostate cancer, breast cancer, and alveolar rhabdomyosarcoma. However, developing small molecular inhibitors targeting transcriptional factors is challenging due to the flat protein-protein interaction surface and absence of deep pockets presented in enzyme active sites.^{145, 146} KDM4 acts as a coactivator of those transcription factor and drives tumor progression, thus targeting KDM4 offers an alternative approach for the treatment of those cancers.

Since KDM4 is involved in DNA damage repair, pharmacologically inhibition of KDM4 by small molecules may enhance sensitivity to radiotherapy or chemotherapy that induces DNA damage. Indeed, JIB-04 was able to sensitize resistant triple-negative inflammatory breast cancer cells and their parental cell line SUM149 to the chemotherapeutic drugs doxorubicin and paclitaxel,¹⁴⁷ restore leukemia cell sensitivity to cytarabine,¹⁴⁸ inhibit growth of temozolomide-resistant glioblastoma cells,¹⁴⁹ and overcome radioresistance of lung cancer cells.¹⁵⁰ KDM4A plays an important role in innate immunity. Depletion of KDM4A has been shown to trigger tumor cell intrinsic immune responses and enhance anti-PD-1 therapy.¹⁵ JIB-04 also increases the IFN-γ+ NK cell ratio in the presence of IL-12.¹⁵¹ Thus, combining a KDM4 inhibitor with immunotherapy might be a promising strategy for cancer therapy.

While the functions of KDM4 in virus infection are unknown, a recent study has shown that JIB-04 has broad-spectrum antiviral activity and inhibits SARS-CoV-2 replication and coronavirus pathogenesis,¹⁵² indicating that KDM4 inhibition may has the potential to suppress infectious diseases. In a transgenic dilated cardiomyopathy mouse model, JIB-04 reduces heart size and prolongs mouse survival.¹⁵³ Importantly, JIB-04 has no negative effect on healthy hearts.¹⁵⁴ Thus, KDM4 inhibitors might be suitable to treat cardiovascular diseases.

TACH101, a selective, first-in-class, potent small-molecule inhibitor of KDM4, was shown to block the proliferation of multiple cancer cell lines and patient-derived organoid models, reduce tumor growth by up to 100% in animal xenograft models representing various cancer types (https://tachyontx.com/our-pipeline/tach101/). TACH101 has recently entered clinical trial in 2022, and its outcome is eagerly awaited by the researchers and clinicians in the field.

However, obstacles remain to translate KDM4 inhibitors to medical applications. First, the clinical benefit of KDM4 inhibition is still unclear and we have no answer yet to what patients would maximally benefit from KDM4 inhibitor treatment. While KDM4 isoforms have been demonstrated to play a role in multiple cancer types, their pathophysiology is poorly understood. Whether KDM4s are important in other cancer types needs to be studied. As KDM4s are epigenetic modulators whose activity is affected by oxygen levels, iron and 2-OG concentrations, phenotypic in vitro screening may not faithfully recapitulate the biological functions of KDM4s in vivo. Therefore, appropriate disease models are required to validate KDM4s as bona fide therapeutic targets in relevant cancers and other diseases. Patient stratification based on biomarkers are key to maximize the efficacy of targeted therapies. However, reports of suitable biomarkers for KDM4 inhibitors are scarce. Tachyon reported that colorectal cancer cells with deficiency of mismatch repair is more sensitive to TACH101. This is the first report of the clinical biomarker for KDM4 inhibition. A recent clinical investigation using PD-1 immunotherapy has achieved remarkable success in treating locally advanced rectal cancer with mismatch repair deficiency.¹⁵⁵ Considering that KDM4 inhibition induces type I interferon response, it would be interesting to see if combination of TACH101 or other KDM4 inhibitors with immune checkpoint inhibitors could be a promising strategy to treat cancers with mismatch repair deficiency. In addition, cancers driven by dysregulated chimeric transcriptional factors or hormone receptors have shown dependency of KDM4 activity, suggesting potential therapeutic application of KDM4 inhibitors. More indications may be waiting to be discovered through further in-depth research of KDM4 physiology and pathophysiology, and the development of novel chemical probes and KDM4-based therapies is likely to continue attracting attention in both industry and academia.

ABBREVIATIONS USED

2-OG

2-oxoglutarate

8HQ

8-hydroxyquinoline

AML

acute myeloid leukemia

AR

androgen receptor

CETSA

cellular thermal shift assay

Chem-seq

chemical affinity capture and massively parallel DNA sequencing

CPX

ciclopirox

CRC

colorectal cancer

ER

estrogen receptor

FAD

flavin adenine dinucleotide

FDH

formaldehyde dehydrogenase

FP

fluorescence polarization

GA

geldanamycin

JmjC-KDM

JmjC-domain containing demethylase

KDM

histone lysine demethylase

KMT

histone methyltransferases

LSD1

lysine-specific demethylase 1

MSI-H

high microsatellite instability

NOC

N-oxalyl-D-cysteine

PDCA

pyridinedicarboxylic acid

PHD

plant homeodomains

PROTAC

proteolysis targeting chimeras

RFMS

RapidFire mass spectrometry

SAHA

suberoylanilide hydroxamic acid

TSA

trichostatin A

Biographies

Qiong Wu completed his Ph.D. in medicinal chemistry from Peking University (PKU) in 2018, after which he joined St Jude children’s research hospital as a postdoctoral fellow. Currently, he is working on the demethylase project under the supervision of Dr. Rankovic and Dr. Yang in the department of chemical biology & therapeutics and surgery department, respectively.

Brandon Young obtained his Ph.D. in organic synthesis from The Johns Hopkins University in 2001. After five years in the pharmaceutical industry, he joined St. Jude Children’s Research Hospital as a member of the High Throughput Chemistry Center (HTC). Currently, he is leader of the Medicinal Chemistry Center at SJCRH.

Yan Wang obtained her MD from Qingdao Medical College. She is an Associate Chief Physician in Department of Geriatrics and Occupational Disease at Qingdao Central Hospital. Her focus is clinical investigation and treatment of a variety of cardiovascular diseases and other critical illness.

Andrew M Davidoff obtained his MD from University of Pennsylvania. He is a board-certified pediatric surgeon, having completed residency training in general surgery at Duke University Medical Center and general and thoracic pediatric surgery at the Children’s Hospital of Philadelphia. He is the Chair of Department of Surgery at St Jude Children’s Research Hospital. His academic interests are focused on clinical and translational investigation and treatment of pediatric solid tumors and gene therapy.

Zoran Rankovic received his Ph.D. in organic chemistry from the University of Leeds, UK. That same year he joined Organon UK, and then continued his drug discovery career as medicinal chemistry director and research fellow at Schering-Plough, Merck, Eli Lilly, and most recently St. Jude Children’s Research Hospital, where he directs Chemistry Centers at the department of Chemical Biology and Therapeutics. His current research interests focus on developing small molecule protein degraders and epigenetic modulators for the treatment of pediatric cancers.

Jun Yang obtained his MBBS from Qingdao Medical College, MMed from National Institute of Pharmaceutical and Biological Products, China, PhD from the Institute of Cancer Research, UK. He is an Assistant Member of Department of Surgery at St Jude Children’s Research Hospital, USA. His research focuses on epigenetics, therapeutics, and drug discovery.