Background
Despite recent advances in basic and translational research in the field of cutaneous T cell lymphoma (CTCL) [
1,
2], effective treatment options for advanced mycosis fungoides (MF) and Sézary syndrome (SS)—the two most common types of CTLC—are still limited. Usually, conventional therapies result in only short-lived remissions [
3‐
5]. Recent genomic analysis revealed aberrations affecting apoptosis [
6], cytokine, and T cell receptor signaling [
7] as well as epigenetic regulation [
8] as possible drivers in CTCL. In particular, mutations in genes coding for proteins involved in histone modification (acetylation, methylation and ubiquitination) as well as chromatin remodeling were commonly detected [
8]. Moreover, hyper-methylated promotor regions of tumor suppressor genes like p16
INK4A [
9,
10] and hypo-acetylated histones [
11,
12]—both leading to silencing of respective promoters—as well as frequently observed overexpression of epigenetic modifiers such as histone deacetylases (HDACs) [
13] suggested that targeting of epigenetic events in general and especially inhibition of HDACs may be a feasible therapeutic approach in CTCL [
14]. Indeed, two class I HDAC inhibitors, i.e., vorinostat and romidepsin (FK228), achieved good clinical efficacy with objective responses of 25–30% leading to the FDA approval of both drugs for the treatment of CTCL in 2006 and 2009, respectively [
15,
16].
While histone acetylation is generally associated with promoter activation, the role of another epigenetic modification, i.e., histone methylation, in regulation of gene expression is more complex and highly dependent on the specific histone residues modified [
17]. In this regard, tri-methylation of lysine 9 or lysine 27 of histone H3 (H3K9me3 or H3K27me3) facilitates gene repression while di- or tri-methylation of lysine 4 of histone H3 (H3K4me2 or H3K4me3) mediates promoter activation, whereas mono-methylation of H3K4 is a hallmark of active distal enhancer regions [
18].
The histone methylation status is controlled by a balanced action of various methyltransferases and demethylases [
19]. Among these enzymes the first histone demethylase to be discovered was the lysine-specific histone demethylase 1a (LSD1), also known as lysine(K)-specific demethylase 1A (KDM1A). LSD1 demethylates mono- and di-methylated H3K4, thereby repressing transcription [
20]. This gene-repressing function exerts LSD1 in particular as part of the multi-protein CoREST complex, which besides several other proteins, includes also HDAC1 and HDAC2 and mediates promoter silencing in a multi-step process involving consecutive histone-deacetylation and H3K4-demethylation [
21]. In a different context, however, LSD1 may also contribute to transcriptional activation by demethylating the gene-repressive tri-methylated H3K9 [
22]. This requires different binding partners [
22‐
24] and might be restricted to a certain splice variant carrying an additional exon [
25]. Furthermore, non-histone targets of LSD1 have been described and in the context of tumor biology, it is of special interest that LSD1 can demethylate di-methylated lysine 370 on p53 thereby inhibiting transcriptional activity of this tumor suppressor protein [
26].
In various cancers increased expression of LSD1 has been documented and correlates with poor differentiation, higher aggressiveness, epithelial-to-mesenchymal transition, and adverse clinical outcome [
27‐
32]. Therefore, inhibition of LSD1 is considered as a promising antitumor strategy [
33‐
35], and its well-defined active site cavity allowed the design of small molecule inhibitors, which are now evaluated in pre-clinical and clinical studies [
36‐
38]. Notably, due to the functional interplay of LSD1 and HDAC1/2 in the CoREST complex, it has been suggested that combined targeting of HDACs and LSD1 might be superior with respect to cancer-specific cytotoxicity compared to individual inhibition of histone acetylation or methylation [
39]. Indeed, synergistic effects of combined LSD1/HDAC inhibition could be demonstrated, and several single drugs inhibiting HDACs as well as LSD1 have been developed [
39‐
42]. One such compound which has been reported to target class I HDACs as well as LSD1 is domatinostat (4SC-202) [
43], which is currently investigated in a phase I clinical trial in hematological neoplasms with up to now excellent tolerability and good efficacy [
44].
The goal of this study was to analyze in vitro how the combined HDAC/LSD1-inhibitor 4SC-202 affects CTCL cells in comparison to the approved therapeutic HDAC-inhibitor FK228. We find that both substances effectively induce cell death in a set of CTCL cell lines with the difference that cell death only in the case of 4SC-202 is preceded by a G2/M arrest. We provide evidence that this difference is not a consequence of targeting LSD1 by 4SC-202. The 4SC-202 anti-cancer activity is, apart from targeting histone-modifying enzymes, rather based on directly affecting tubulin polymerization.
Methods
Cell culture
To study in vitro the impact of FK228 (Selleckchem) and 4SC-202 (4SC AG) on cutaneous lymphoma cells, six different CTCL cell lines were used: CRL-2105 (Synonym: HH, Accesssion: CVCL_1414) [
45], CRL-8294 (MJ, CVCL_1414) [
46], HTB-176 (H9, CVCL_1240) [
47], HuT 78 (NCI-H78, CVCL_0337) [
48], MyLa [
49], and Se-Ax (SeAx, CVCL_5363) [
50]. In addition, HeLa (cervix carcinoma cell line), WaGa (Merkel cell carcinoma), UACC-257 (melanoma), U2OS (osteosarcoma), MCF7 (breast adenocarcinoma), A549 (lung adenocarcinoma), Maver-1 (mantle cell lymphoma), and the human embryonic kidney-derived HEK293T cell line were used in this study. All cell lines were maintained in RPMI supplemented with 10% fetal calf serum (FCS), 100 U/ml penicillin, and 0.1 mg/ml streptomycin. Peripheral blood lymphocytes (PBLs), a dermal fibroblast preparation derived from the skin of an adult donor (fibroblasts A) as well as human foreskin fibroblasts (fibroblasts B), served as untransformed control cells. PBLs were freshly isolated from the blood of a healthy donor by Ficoll-Hypaque density-gradient centrifugation. In vitro doubling times of the different primary cells and cell lines were estimated by cell counting and are summarized in Additional file
1: Table S1.
MTS assay
In order to measure cell viability and cellular metabolic activity, MTS assay was performed according to standard protocols. Cells were seeded as triplicates in a 96-well plate with 5000 or 15,000 cells/well. FK228 (Selleckchem) or 4SC-202 (4SC AG) were added in incremental concentrations. Culture medium served as negative control. After 72 h, 10 μl MTS reagent was added per well and absorbance was measured at 492 nm (with reference wavelength of 630 nm) after 60 min. IC50 doses were calculated by nonlinear fitting of MTS curves using GraphPad Prism Version 7.0.
DNA staining
For DNA staining, cells were fixed using ice-cold ethanol (90%) for at least 1 h followed by treatment with a propidium iodide solution (PI) (phosphate buffered saline supplemented with 1% FCS, 0.1 mg/ml PI, and 0.1 mg/ml RNAse A) at 37 °C for 1 h. Cellular DNA content was then analyzed by flow cytometry.
Annexin V assay
An Annexin V Phycoerythrin conjugate (Annexin V Apoptosis Detection Kit, BD Biosciences) was applied to identify apoptotic cells by flow cytometry according to the manufacturer’s instructions. By double staining with 7-AAD, early apoptotic cells can thereby be identified as 7-AAD−/Annexin V+ cells while the 7-AAD+/Annexin V+ double positive population marks the late apoptotic cells.
Immunoblot
Immunoblotting was performed as previously described [
51]. Mouse-derived primary antibodies, anti-HDAC1 (10E2, Cell Signaling, 1:1000), anti-H3K9ac (1B10, Active Motif, 1:1000), anti-β-tubulin (TUB 2.1, Santa Cruz Biotechnology, 1:1000), anti-β-actin (AC-15, Sigma Aldrich, 1:3000), and anti-Caspase-3 (3G2, Cell Signaling, 1:1000), were used in this study while anti-LSD1 (C69G12, Cell Signaling, 1:1000), anti-H3K4me2 (Y47, Abcam, 1:1000), and anti-Cleaved Caspase-3 (D175, Cell Signaling, 1:1000) were rabbit antibodies.
Real time PCR
Expression levels of
LSD1 and
HDAC 1-
3 genes were determined by qPCR with SYBR Green technology. RNA was isolated as described in the instruction manual of the peqGOLD Total RNA Kit® (Peqlab), transcribed into cDNA by SuperScript II, and amplified by the primers given in Additional file
1: Table S2. Expression of the target genes was depicted as ∆Ct (target-RPLP0).
NanoString nCounter® analysis
Alterations of gene expression under treatment with 4SC-202 or FK228 were assessed by NanoString nCounter® analysis (NanoString technologies). One hundred nanograms total RNA were subjected to hybridization with the NanoString kinase Kit (Kinase_V2_Panel-48rxn Kit, NanoString technologies) containing probes for 519 kinase and six housekeeping genes. Following nCounter digital reading the values were globally normalized according to the manufacturer’s protocol.
Time-lapse microscopy
Since live cell imaging turned out to be not feasible with suspension cells such as CTCL cell lines, adherent histone H2B-GFP and additionally RFP-tubulin expressing HeLa cells were used as a representative model for time-lapse microscopy. Cells were seeded into 4-well slides (ibidi®) in phenol red-free medium, and placed in a live cell imaging chamber that assured standard culture conditions (37 °C, 95% humidity, 5% CO2). Images were taken every 10 to 20 min using Eclipse Ti (Nikon).
Lentiviral LSD1 knockdown and knockout
To knockdown LSD1, we first generated a selectable lentiviral one-vector system which allows Golden Gate cloning of an shRNA coding sequence under the control of a Doxycyclin (Dox)-inducible promoter (induc shRNA EYFP-P2A-Puro; Genbank: MH749464). As shRNA target sequence for
LSD1, we used AGGCCTAGACATTAAACTGAA. Lentiviral supernatants were produced as previously described [
52]. MyLa cells were infected and following puromycin selection, shRNA expression was induced by addition of 1 μg/ml Doxycyclin.
To achieve an LSD1 knockout, we first cloned an oligonucleotide sequence targeting CGCGGAGGCTCTTTCTTGCG in exon 1 of the
LSD1 gene into the LentiGuide-BSD vector, which had been derived from LentiGuide-Puro [
53] (kind gift from Feng Zhang; Addgene plasmid #52963) by replacing the puromycin with a blasticidin resistance. Virus generated with this LentiGuide-BSD-LSD1 construct was used to infect HeLa cells stably expressing Cas9, due to prior transduction with the lentiviral pcdh puro Cas9 followed by Puromycin selection. Blasticidin-resistant single-cell clones were established, and clones lacking LSD1 expression were identified by immunoblot. Knockout was confirmed by sequencing of the genomic region targeted by the
LSD1 guide RNA.
siRNA transfection
siRNAs targeting HDAC1 and HDAC3 were purchased from Sigma (Mission® esiRNA) and transfection was performed with Lipofectamine® RNAimax (Thermofisher) according to the manufacturer’s instructions.
In vitro tubulin polymerization assay
An in vitro tubulin polymerization assay kit (Merck) containing > 99% pure bovine tubulin was applied according to the manufacturer’s instructions. Spontaneous formation of microtubules in the presence of GTP at 37 °C was monitored by measuring the OD at 340 nm in an absorbance microplate reader over time.
Whole cell analysis of tubulin polymerization
To measure the degree of intracellular tubulin polymerization by flow cytometry we performed tubulin staining under conditions where only polymerized tubulin is retained in the cells while tubulin monomers tend to get lost during the fixation procedure [
54]. To this end, cells were fixed in 1 ml of 0.5% glutaraldehyde in microtubule-stabilizing buffer (80 mM Pipes [pH 6.8], 1 mM MgCl
2, 5 mM EDTA, and 0.5% Triton X-100 [TX-100]). After 10 min, 0.7 ml NaBH
4 (1 mg/ml in PBS) was added, and the cells were pelleted by centrifugation. Cells were then re-suspended in 20 μl of PBS containing 50 μg/ml RNase A, 0.2% TX-100, 2% bovine serum albumin [BSA], and 0.1% NaN
3, and incubated overnight at 4 °C. After additional 3 h of incubation with an Alexa Fluor® 488 anti-Tubulin-α Antibody (clone 10D8; BioLegend), 200 μl of 50 μg/ml propidium iodide in PBS were added followed by flow cytometry. Median tubulin fluorescence was determined for the 4 N fraction.
Discussion
Recent pre-clinical studies on 4SC-202 have demonstrated activity against hepatocellular carcinoma [
68], colorectal cancer [
69], urothelial carcinoma [
57], medulloblastoma [
63,
70], and pancreatic cancer cells [
71]. Our study adds cutaneous T cell lymphoma to this list. Moreover, in line with previous publications [
69,
70], we observed that cytotoxicity induced by 4SC-202 was elevated in cancer cells compared to untransformed control cells.
Interestingly, different mechanisms have been proposed for how 4SC-202 would inhibit cellular growth and viability. In this respect, activation of the ASK1-dependent mitochondrial apoptosis pathway in hepatocellular carcinoma cells [
68], inhibition of hedgehog/GLI signaling in medulloblastoma cells [
63], or increased expression of BRD4- and MYC-dependent epithelial genes in pancreatic cancer cells [
71] have been suggested as critical anti-tumorigenic outcome following inhibition of its primary targets by 4SC-202. Indeed, all previous studies on 4SC-202 have in common that the authors either assume or provide evidence that these primary targets are class I HDACs and/or LSD1. In contrast, our results suggest that 4SC-202 has the potential to inhibit cancer cell growth also independent of targeting these epigenetic modifiers, and independent of altering cellular transcription.
Our study on CTCL cells was largely based on the comparison of 4SC-202 with FK228, a class I HDAC inhibitor approved in the US for the treatment of CTCL [
72,
73]. One major difference regarding the biological effects induced by the two inhibitors was that cell death induction by 4SC-202 was preceded by a G2/M arrest, a feature of 4SC-202 reported by others as well [
57,
69]. Since HDACs1–3 are also inhibited by FK228 [
55,
56], the only other described target of 4SC-202, i.e., LSD-1, [
43] seemed to be the best candidate to mediate the observed G2/M arrest. However, neither knockdown nor knockout of LSD-1 affected cell cycle distribution of untreated or 4SC-202-treated cells. Furthermore, and in contrast to FK228, 4SC-202-induced cytotoxicity could not be attenuated by HDAC1 overexpression, and was observed already at concentrations associated with only minor changes in histone modifications and gene transcription. Finally, even under conditions of completely abolished cellular transcription, 4SC-202 could maintain a G2/M arrest. Together, these results suggest that 4SC-202 can induce mitotic arrest and cell death by targeting molecules which differ from its targets described so far. Our data, however, do not suggest that 4SC-202 cannot inhibit HDACs and LSD-1 in CTCL or other cells. Nevertheless, it is of interest that a recent publication raised doubts whether 4SC-202 would target LSD1 at all [
40].
The major structural elements of the mitotic spindle are microtubule polymers consisting of α/β tubulin heterodimers. [
74] Spindle assembly as well as chromosome segregation during mitosis are highly complex processes involving multiple accessory proteins [
74,
75]. Nevertheless, spindle formation and function are based on the fundamental capability of tubulin to undergo spontaneous microtubule polymerization in the presence of GTP [
76]. Disturbing microtubule polymerization has proven efficacy in the clinic against a broad range of malignancies [
77], and respective antimitotic drugs are usually classified as either microtubule-destabilizing agents, which inhibit microtubule polymerization (e.g., nocodazole) or microtubule-stabilizing agents (e.g., paclitaxel) [
74]. In a very simple in vitro system consisting of highly purified α- and β-tubulin and GTP, we demonstrated here that 4SC-202 is able to inhibit tubulin polymerization just like nocodazole suggesting that it might act as a microtubule-destabilizing spindle poison. Indeed, using time-lapse microscopy, we could confirm that 4SC-202 is able to inhibit formation of the mitotic spindle in HeLa cells. Since 4SC-202-treated CTCL cells display a reduction of polymerized tubulin in 4-N cells as well as activation of the spindle assembly checkpoint, it is very likely that also in CTCL cells 4SC-202 is affecting formation of the mitotic spindle and thereby induces the observed G2/M arrest. Furthermore, since a 4SC-202-induced mitotic arrest has been described for colorectal cancer [
69] and urothelial carcinoma cells [
57], and we observed the same in six out of six cell lines from further cancer entities, targeting of tubulin polymerization might be a general mechanism contributing to impairment of tumor cells by 4SC-202. This probably includes also cytotoxic effects induced by 4SC-202 since it is well known that inhibiting mitosis finally leads to cell death [
77]. Because it is indicating a potentially large therapeutic window, it is of special interest that arrest and cell death induced by 4SC-202 occurred only in cancer cells. Indeed, even fast-proliferating control cells were barely affected by 4SC-202 treatment. Since it has been shown that modifying microtubule stability can change the sensitivity towards nocodazole [
78], a possible explanation for the differential sensitivity towards 4SC-202 might be diverse microtubule stability in cancer and control cells.
Emerging therapies for leukemia and lymphoma are frequently based on the application of epigenetic modifiers and especially HDAC inhibitors [
79‐
81]. Our data suggest that in contrast to most other HDAC inhibitors, 4SC-202 is additionally directly targeting microtubule formation. Polypharmacology, a concept encompassing both, multiple drugs binding different targets as well as one drug binding multiple targets within a pathological network, is considered as a very useful strategy for the treatment of complex and refractory diseases, in particular of cancer [
82]. In this respect, attempts have been made to design new drugs targeting HDACs as well as tubulin polymerization, and recently such compounds demonstrating excellent anti-proliferative activity have been described [
83]. Furthermore, mocetinostat, an inhibitor of class I/IV HDACs which is currently under clinical evaluation [
84], has been demonstrated to also bear microtubule inhibitory activity although direct interaction with tubulin was not formally proven [
85]. Notably, HDAC inhibitors can disrupt mitosis by affecting a number of components of the mitotic machinery [
86]. Our results, however, argue against such an indirect effect. Indeed, here we describe that 4SC-202, which so far was considered as a drug targeting class I HDACs as well as the histone demethylase LSD1, is also a potent microtubule-destabilizing agent. The latter function was dominant for the impairment of CTCL cell lines in our in vitro experiments. This, however, does not exclude a contribution of the other inhibitory features to the anti-tumoral activity of the substance. This holds true for direct effects on cancer cells but even more for indirect effects, which cannot be detected in vitro. In this respect, it is intriguing that very recently LSD1 was demonstrated to be critical for inhibiting tumor cell immunogenicity [
87]. Furthermore, also HDAC inhibitors are discussed as potential immunomodulating agents to treat cancer [
88]. Therefore, it will be interesting to evaluate whether the pleiotropic molecular features of 4SC-202 may turn into benefits for treated cancer patients. In this respect, a first-in-man phase I clinical trial with 4SC-202 (
www.clinicaltrials.gov, NCT01344707) for patients with advanced hematological malignancies has been launched [
44] and a report of the results is awaited soon.