Identification and characterization of PKF118-310 as a KDM4A inhibitor

ABSTRACT

Epigenetic modifications are functionally involved in gene expression regulation. In particular, histone posttranslational modifications play a crucial role in functional chromatin organization. Several drugs able to inhibit or stimulate some families of proteins involved in epigenetic histone regulation have been found, a number of which are FDA-approved for the treatment of cutaneous T-cell lymphoma or are in phase I/II/III clinical trials for solid tumors. Although some protein families, such as histone deacetylases and their inhibitors, are well characterized, our understanding of histone lysine demethylases is still incomplete. We describe the in silico, in vitro, and cell-based characterization of the compound PKF118-310, an antagonist of transcription factor 4 (TCF4)/β-catenin signaling, as inhibitor of KDM4A. PKF118-310 potential inhibitor activity was discovered via virtual screening on the crystal structure of KDM4A. A peptide-based histone trimethylation assay developed in-house confirmed its potent KDM4A inhibitor activity. Its protein target was identified by cellular thermal shift assay experiments. PKF118-310 anticancer activity was observed in both liquid and solid tumor cells, and shown to have a dose- and time-dependent effect. We demonstrate the previously unreported inhibitory action of PKF118-310 on KDM4A. Our findings open up the possibility of developing the first KDM4A-specific inhibitors and derivatives.

Abbreviations

TCF4

transcription factor 4

PTMs

posttranslational modifications

KDMs

lysine demethylases

JmjC

Jumonji C

JMJs

Jumonji enzymes

HDACs

histone deacetylases

FBS

fetal bovine serum

TEB

triton extraction buffer

FACS

fluorescence-activated cell sorting

DMSO

dimethyl sulfoxide

CETSA

cellular thermal shift assay

Introduction

Epigenetic mechanisms include a plethora of modifications that influence gene expression without altering the DNA sequence.¹ Histone posttranslational modifications (PTMs) play crucial roles in cellular biology, fate, and survival. Histone modifications have been linked to several diseases including cancer, metabolic syndromes, and neurological disorders.² Histones are targets of several PTMs, including methylation, acetylation, and phosphorylation, which can modify chromatin structure and accessibility. Methylation is pivotal in many processes involved in cell differentiation, homeostasis, and disease.^3-6 Methylation occurs mainly on arginine/lysine residues and is mediated by numerous enzymes sharing common domains.^7-9 More than 20 human histone N^ε-methylated lysine residue demethylases (KDMs) have been identified to date. Many exhibit deregulated expression patterns in cancer as well as in neurological and immunological diseases.¹⁰ As with lysine methyltransferases, growing evidence suggests that KDMs are involved in disease initiation and progression.^11,12 KDMs have thus emerged as potential new therapeutic targets.

KDMs are classified into 2 groups. The KDM1 group (or lysine-specific demethylases) shares features with the well-characterized monoamine oxidases, and uses the co-substrate flavin adenine dinucleotide during catalysis of demethylation.^10,13 The second and greater group includes Jumonji C (JmjC) domain-containing KDMs, and currently includes 20 human enzymes divided into 5 subfamilies (KDM2A-B/7A, KDM3A-B, KDM4A-B-C-D-E, KDM5A-B-C-D, and KDM6A-B),¹⁴ known collectively as Jumonji enzymes (JMJs).

JMJs demethylate lysine residues through N-methyl hydroxylation and require Fe²⁺ and α-ketoglutarate as cofactors. Specifically, oxidative decarboxylation converts Fe²⁺ into Fe⁴⁺, which inserts an oxygen atom on the methyl group, making an unstable intermediate product, which leads to the liberation of the demethylated compound and formaldehyde.

One of the major subfamilies, JMJD, consists of KDM4A-D (also known as JMJD2A-D) proteins, which catalyze demethylation of H3K9me2/me3 and H3K36me3 residues. KDM4A, B, and C display an N-terminal domain Jumonji, a JmjC domain, followed by a C-terminal plant homeodomain and 2 tudor domains. KDM4D lacks the N-terminal region, which includes the plant homeodomain and tudor domain.¹⁵ Studies show that several KDMs are deregulated in cancer. KDM deregulation has been correlated with: i) oncogene expression activation; ii) tumor suppressor silencing; iii) chromosomal stability alteration; and iv) hormone receptor interaction. In line with these findings, overexpression of KDM4 has been reported in various tumors, including breast and prostate cancer, and lymphomas.^16,17

Lysine methylation is not restricted to histone tails, but occurs in many proteins including estrogen, androgen, and NF-kB receptors.¹⁸ Several lysine methylation sites were identified on p53, whose function is activated or suppressed depending on the methylation state.¹⁹ KDM4D is considered a p53 coactivator, decreases H3K9 trimethylation levels, interacts with H3K9 trimethylation, and interacts with the DNA binding domain of p53 thereby activating p21, a cell cycle inhibitor.¹⁹ Silencing of KDM4D in colon cancer cells reveals that KDM4D is a pro-proliferative enzyme.¹⁹ This is, at first glance, counterintuitive since KDM4D is a co-activator of p53, a tumor suppressor that normally reduces cell proliferation, partly by upregulating p21. Increasing evidence indicates that KDM4A may play a key role in tumor initiation, promotion, and progression. KDM4A overexpression has been correlated with Gleason score in human prostate tumors,²⁰ and has been implicated in the development of prostatic intraepithelial tumors.²¹ These results highlight the potential involvement of KDM4A in prostate cancer initiation.²⁰ As previously mentioned, KDM4A is highly expressed in human cancers, where it represses the tumor suppressor ARHI.²² The transcription factor SP1 negatively regulates breast cancer metastasis via DIRAS3 transcriptional activation. Like plicamycin and histone deacetylase (HDAC) inhibitors, KDM4A inhibits SP1 expression in breast cancer, increasing the probability of metastasis.²³

Here, we report the identification and biological characterization of a novel KDM4A inhibitor displaying anticancer action.

Results

Virtual screening and docking identifies PKF118-310 as a novel KDM4A inhibitor

In order to identify new KDM4A inhibitors, we performed a 3D structure-based virtual screening via “mcule.com,” in which predicted 3D structures of small molecules dock into the binding site of the experimentally determined Protein Data Bank (PDB) KDM4A (ID: 2q8d) structure. Over five million compounds were virtually screened, and about one thousand compounds were pre-selected (data not shown). After matching the screened compounds with the available biological data, we focused on PKF118-310 (toxoflavin). This compound has been reported to be a TCF4/β-catenin signaling antagonist. PKF118-310 disrupts the TCF4/β-catenin complex and alters expression of TCF4-responsive genes. In addition, PKF118-310 is able to inhibit expression of survivin²⁴ and induces apoptosis in human colon cancer, human osteosarcoma cells and lymphocytic leukemia cell lines.^24-26 Since no KDM4A action was reported for PKF118-310 so far, we used the chemical structure of PKF118-310 as input and interrogated a library of nearly 10,000 structures (1-CLICK DOCKING) from the sc-PDB, an annotated database of druggable binding sites from the PDB on the same database “mcule.com.”²⁷ The best 4 docking poses were obtained.²⁸ By using this approach, we also found that the KDM4A enzymatic pocket is ‘bona fide’ a good binding candidate site for PKF118-310 (data not shown).

PKF118-310 is a KDM4A inhibitor in vitro and in living cells

To corroborate the virtual screening, we performed an in vitro enzymatic assay for KDM4A. Specifically, KDM4A acts on substrate demethylation with formaldehyde production (Fig. S1). The combination of formaldehyde, ammonia, and acetoacetanilide produces a fluorescent compound that reacts at an excitation wavelength of 370 nm and emission wavelength of 470 nm. When a putative compound is an inhibitor, KDM4A activity is blocked, and the final fluorescent compound is decreased compared to the control. Based on reports in the literature that H3K9me3, H3K9me2, and H3K36me3 are targets for KDM4A activity,^29,30 we validated the virtual screening on the histone targets of KDM4A. H3K9me3 and H3K36me3 peptides were used as substrates for the enzymatic activity of KDM4A. PKF118-310 was incubated with recombinant GST-KDM4A enzyme, cofactors, and substrate, and enzyme activity was recorded. Compared to control (DMSO), PKF118-310 strongly inhibited KDM4A activity on both substrates (Fig. 1A). A PKF118-310 concentration curve was generated for KDM4A activity. We confirmed that PKF118-310 inhibition activity occurs in a dose- and time-dependent manner (Fig. 1B). Fifty percent of KDM4A activity was assessed with PKF118-310 at 10 μM in vitro.

Figure 1.

In vitro PKF118-310 characterization. (A) Fluorescent acquisition of reaction with 2 different peptide substrates in presence of PKF118-310 at 10 μM as final concentration. (B) PKF118-310 IC₅₀ evaluation based on a dose-dependent enzymatic activity acquired as described in Fig. S1. (C) Relative quantization of Western blot signals based on CETSA. Cells were treated with PKF118-310 (10 μM) and an equal amount of DMSO for 1 h. The respective samples were divided into aliquots (100 μl) and heated at 25°C, 37°C, 44°C and 47°C for 3 min. Aliquots of treated cells were heated at the indicated temperature. Total protein extracts were obtained in RIPA buffer (50 mM Tris-HCl pH 7.4; 1% NP40; 0.5% Na-deoxycholate; 0.1% SDS; 150 mM NaCl; 2 mM EDTA; 50 mM NaF; one tablet of protease/phosphatase inhibitors) and quantified by Bradford assay. KDM4A primary antibody was used for protein revelation. Results were normalized and integrated. The relative abundance was achieved using Fuji software. P-value < 0.005.

PKF118-310 directly binds KDM4A

To test the direct binding between PKF118-310 and KDM4A, we set up a CETSA assay.³¹ This assay is able to prevent or retard thermal degradation of the specific target protein as a result of its stabilization by drug binding. Our findings show that at an increased temperature compared to room temperature, KDM4A is stabilized in response to PKF118-310 stimuli (Fig. 1C). Specifically, we observed KDM4A stabilization at 44°C. These results confirm the direct interaction between the protein target KDM4A and PKF118-310.

PKF118-310 inhibits cancer cell proliferation

To investigate the effects of PKF118-310 on cell proliferation, we performed a proliferation assay. Specifically, the graph in Fig. 2A shows the CI (a value of cellular density) acquired for over 100 h. The CI is a specific parameter allowing evaluation of cell proliferation. At the higher concentration, PKF118-310 follows a similar pattern to cells without FBS, displaying a very strong and rapid inhibition of cellular duplication, readily detectable already within 30 h of treatment (Fig. 2B). At the lower concentration, PKF118-310 is still able to modulate proliferation, unlike the controls. Fig. S2A records the doubling time demonstrating how the higher concentration strongly inhibits cellular life, while Fig. S2B shows the single curve in order to demonstrate the reproducibility of experiments.

Figure 2.

Real-time anticancer doubling effect of PKF118-310. (A) Average of biological duplicate proliferation curve relative to HCT-116 cell line recorded for more than 100 h. The graph shows the wild type cell line (dark blue), cell treated with PKF118-310 at 10 μM (light blue) and 1 μM (purple), background (red) and cells without FBS (green). (B) Detail of first 30 h post treatment combination.

PKF118-310 impacts on cell cycle and induces apoptosis

Several reports describe the effect of PKF118-310 stimulation on cell cycle.^24,32,33 We therefore investigated the effects mediated by the compound on wild type HCT-116, HCT-116 p53^−/−, and U937 (a leukemia cell line with p53 deletion mutation) cells,³⁴ and an immortalized cell line, MePR2B,³⁵ used as control. We stimulated the cells with PKF118-310 at 2 concentrations (1 and 10 μM). Cells were collected and cell cycle was analyzed via FACS (Fig. 3A–D). Pre-G1 analysis confirmed that PKF118-310 is a potent cytotoxic anticancer agent (Fig. 3E). We obtained similar results for the HCT-116 p53^−/− (Fig. 3F and Fig. S3A) and U937 (Fig. 3G) cell lines. Furthermore, in MePR2B cells, the amount of cell death induced was lower (Fig. 3H), indicating that PKF118-310 acts on cancer cells in a preferential manner. Despite some toxicity that seems to be exerted by PKF118-310, our findings suggest that PKF118-310 may be used at a lower concentration with potentially reduced side effects.

Figure 3.

Effect of PKF118-310 on cell cycle. Cell cycle analysis in HCT-116 (A, E), HCT-116 p53^−/− (B, F), U937 (C, G) and MePR2B (D, H) cell lines. PKF118-310 was tested at 1 μM and 10 μM in all cell lines at 12 h, 24 h, 48 h and 48 h*(compound was re-added after 24 h). Cell cycle and pre-G1 phases were determined and analyzed by ModFit software.

Polycombinatorial effect of PKF118-310 in cancer models

Cancer treatment guidelines currently support a polycombinatorial approach to drug delivery. We therefore performed a cell cycle and histone modification analysis on PKF118-310, the well-characterized HDAC inhibitor SAHA, and the methyltransferase inhibitor GSK-J4.^36,37 In colon cancer HCT-116 cells, the combination of PKF118-310 and GSK-J4 was found to have a synergistic effect on cell cycle modulation with a burst in pre-G1 phase (Fig. 4A). The increased number of cells showing fragmented DNA was not directly correlated with H3K9me2 modification (Fig. 4B).

Figure 4.

Combinatorial effect with HDAC and KDM inhibitors. (A) Cell cycle analysis of PKF118-310 at 0.1 μM and 1 μM for 24 h in HCT-116 cells, alone and in combination with SAHA at 0.5 μM and 5 μM, and 10 μM GSK-J4. (B) H3K9me2 Western blot and its quantization by Fuji software. H4 was used as loading control.

PKF118-310 target evaluation

To corroborate our data, PKF118-310 was assessed in HCT-116 and U937 cancer cells. After stimulating HCT-116 cells with PKF118-310 or respective control (DMSO), levels of H3K9me3, H3K9me2, and H3K9me1 were evaluated by Western blot on histone extracts (Fig. 5A-C). Compared to DMSO and H3K9me1, PKF118-310 incubation generated an increased signal of H3K9me3 and H3K9me2. Moreover, H3K9me3 and H3K9me2 upregulation was time- and PKF118-310 concentration-dependent. Similar results were obtained with the U937 cell line (Fig. S3B). The analyses revealed a 14-fold signal increase for H3K9me3 challenged with PKF118-310 for ‘24 h + 24 h’ (compound added again 24 h after initial stimulation). Notably, the specificity of anti-H3K9me3 antibody was validated in Western blot by detecting the phosphorylation in Ser10: H3S10ph signal was equally modulated at 48 h and 24 h + 24 h points, differently from H3K9me3, for which a modulation between these 2 points was clearly detectable (Fig. S4). These results strongly indicate that PKF118-310 is a KDM4A inhibitor in vitro and in living cells.

Figure 5.

Specific histone targets. (A, B, C) Western blot of HCT-116 histone extract incubated with H3K9me1, H3K9me2, and H3K9me3, respectively. ‘24h+24h’ indicates that PKF118-310 has been added a second time after 24 h. Densitometry analysis was performed using Fuji software. Results are the average of independent experiments.

Discussion

‘Readers’, ‘writers’, and ‘erasers’ are the mediators of epigenetic mechanisms in physiological and disease conditions. Fine-tuning their activity is the goal of epigenetic drug discovery, and huge advances are continually being made. While HDAC inhibitors are already in clinical use, our understanding of methylation regulators has lagged somewhat behind. The scientific community shares the view that histone methylation is one of the major crossroads in gene expression and regulation. However, the finer features of players involved in molecular machinery, known collectively as demethylase enzymes, are yet to be clarified. Investigators are looking for small molecules able to modulate these enzymatic families. KDM4A, one of the demethylase enzyme family members, is currently one of the main targets used in drug discovery.

Starting from an in silico screening, we selected a number of potential candidate inhibitors. Of these, PKF118-310 was not previously described as a KDM4A inhibitor. In addition, since PKF118-310 is reported to be a TCF4/β-catenin modulator, we analyzed its histone demethylase modulation on H3K9me3 and H3K9me2 but not on H3K9me1, corroborating the hypothesis of KDM4A specific activity. Both in vitro and ex vivo experiments identified PKF118-310 inhibition of KDM4A. Interestingly, we observed a greater impact on cell cycle in the U937 leukemia cell line. We also focused more closely on the direct binding of PKF118-310 with KDM4A via CETSA, identifying direct binding and inhibitory activities in vitro and in cell-based settings. Taken together, our findings provide an insight into the biological effects of PKF118-310, and demonstrate its ability to bind KDM4A, probably in the enzymatic pocket.

Our results propose KDM4A-related targeting of PKF118-310 as a novel therapeutic strategy in cancer therapy, highlighting the crucial role of this enzyme in tumorigenesis. Further studies will be necessary to better understand the molecular mechanism(s) of action of the drug and define the specific role of the target KDM4A in the described anti-cancer activity. To this aim, comparative genetic (mutation in KDM4A gene) and epigenetic (methylation levels of H3K9) analyses between normal and cancer patients samples could provide a useful tool to further establish the critical function of the enzyme in cancerogenesis.

In conclusion, we report for the first time a KDM4A-specific inhibitor. Our findings suggest the possibility of inducing cancer (solid and liquid) cell-specific death by modulating demethylases and in particular KDM4A. The advantage of using PKF118-310 for cancer treatment may lie in its ability to achieve a synergistic effect in combination with other methylation regulators, such as GSK-J4. This evidence highlights the growing interest to develop combination therapies as viable approaches to increase the effectiveness of a histone methylation modulator. Since several aberrant histone modifications are involved in cancer, the effect of combination therapy is synergistic, targeting simultaneously 2 or more epi-modifications. In addition, combo-therapies are useful to overcome drug resistance and minimize toxicities and side effects.

Materials and methods

Cell lines

Human colon cancer HCT-116 (ATCC, VA, USA) and HCT-116 p53^−/− (DSMZ, Braunschweig, Germany) cells were propagated in McCoy's 5A medium (Euroclone, Milan, Italy) with 10% fetal bovine serum (FBS) (Euroclone), 2 mM L-glutamine (Euroclone) and antibiotics (100 U/ml penicillin, 100 lg/ml streptomycin) (Euroclone).³⁸ Human leukemic monocyte lymphoma U937 cells³⁹ (ATCC) and mesenchymal progenitor (MePR2B) cells⁴⁰ were propagated in RPMI 1640 medium containing 4.5 g/L glucose (Euroclone) supplemented with 10% FBS (Euroclone), 100 U/mL penicillin-streptomycin (Euroclone) and 2 mM L-glutamine (Euroclone). Cells were stimulated with drugs for 12 h, 24 h, 48 h, and 24 h + 24 h (drugs were added again 24 h post-24 h stimulation).

Protein histone extraction

To quantify H3K9 modification, the cells were harvested, washed with PBS (Euroclone), and lysed in triton extraction buffer (TEB; PBS containing 0.5% Triton X 100 (v/v), 2 mM PMSF, 0.02% (w/v) NaN3) at a cell density of 10⁷ cells/mL for 10 min on ice, with gentle stirring. After centrifugation (2,000 rpm at 4°C for 10 min), the supernatant was removed, and the pellet was washed in half the volume of TEB and centrifuged as before. The pellet was suspended in 0.2 N HCl at a cell density of 4 × 10⁷ cells/mL overnight at 4°C on a rolling table. The samples were then centrifuged at 2,000 rpm for 10 min at 4°C, the supernatant was removed, and the protein content was determined using a Bradford assay (Bio-Rad, CA, USA).

Western blotting

To quantify histone H3K9 modification, 5 mg of total histone extracts were separated on a 15% polyacrylamide gel. Western blots revealed H3 modification with H3K9me1, H3K9me2 and H3K9me3 antibodies (Diagenode, Liège, Belgium; cat n° pAb-065-050, C15410060, C15410193, respectively) and with H3S10ph (Abcam Cambridge, UK; cat n° ab5176). Histone H4 (Abcam Cambridge, UK; cat n° ab10158) antibodies were used to normalize for equal loading. Semi-quantitative analysis was performed using Fiji software.⁴¹

Cell proliferation assay

Cancer cell proliferation was tested with the xCELLigence system (Roche, Mannheim, Germany). HCT-116 cells were dispensed in 96-well plates (E-Plate, Roche) in duplicate at 2 × 10⁴ cells/mL to evaluate cellular impedance. Impedance is dependent on confluence level. An arbitrary unit parameter, ‘Cell Index' (CI), was assigned to express confluence.⁴² PKF118-310 (10 μM and 1 μM) (Sigma Aldrich, MO, USA; cat n° K4394) and DMSO (Sigma) were added 6 h post-seeding, in duplicate. Dynamic CI values were monitored at 30-min intervals from the time of plating until the end of the experiment. CI values were calculated and plotted on a line graph and histogram.

Cell cycle analysis

Cells were plated (2 × 10⁵ cells/mL) and collected after stimulation. The cells were then centrifuged (1,200 rpm for 5 min) and suspended in a solution containing 1 × PBS, 0.1% sodium citrate, 0.1% NP40 and 50 mg/mL propidium iodide. After 30 min of incubation at room temperature in the dark, cell cycle was evaluated by fluorescence-activated cell sorting (FACS) flow cytometer (FACSCalibur BD Bioscience, CA, USA) and analyzed with ModFit v3 software (Verity Software House).

FACS analysis of apoptosis

Apoptosis was measured as pre-G1 analyzed by FACS with Cell Quest software (BD Biosciences), as previously reported.^43,44

Cellular thermal shift assay (CETSA)

HCT-116 cells were harvested and washed with PBS after treatment with PKF118-310 (10 μM) and an equal amount of DMSO, as control, for 1 h. The respective samples were suspended in PBS (1.5 mL), divided into aliquots (100 μl), and heated at different temperatures for 3 min by Thermo Mixer (Eppendorf, Milan, Italy), followed by cooling for 3 min at 4°C. After incubation, lysis buffer (100 μl) was added to the samples and incubated for 15 min. The samples were then centrifuged at 13,000 rpm for 30 min at 4°C, the supernatant was removed, and the protein content was determined using a Bradford assay (Bio-Rad). Of the total protein extract, 20 μg was loaded on 10% SDS-PAGE. Nitrocellulose filters were stained with Ponceau red (Sigma) as an additional control for equal loading. KDM4 was from Abcam (cat n° ab1059) and ERK1 antibody was from Santa Cruz (CA, USA; cat n° A1916). Semiquantitative analysis (Fig. 2C) was performed using Fiji software.

Availability of data and materials

The datasets generated during the current study are not publicly available but are deposited on www.mcule.com server and are available from the corresponding author on reasonable request.

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Acknowledgment

We thank C. Fisher for linguistic editing.

Funding

This work was supported by: Blueprint (282510); EPIGEN (MIUR-CNR); MIUR (PRIN-2012ZHN9YH); AIRC (17217).

Author contributions

GF and FS made substantial contributions to conception and design, acquisition of data, analysis and interpretation of data, manuscript drafting and revising it critically for important intellectual content; AN contributed to experimental conception and design, manuscript drafting and revising it critically for important intellectual content; LA conceived experiments and wrote the manuscript.