Deciphering the molecular effects of romidepsin on germ cell tumours: DHRS2 is involved in cell cycle arrest but not apoptosis or induction of romidepsin effectors

Abstract Testicular germ cell tumours (GCTs) mostly affect young men at age 17‐40. Although high cure rates can be achieved by orchiectomy and chemotherapy, GCTs can still be a lethal threat to young patients with metastases or therapy resistance. Thus, alternative treatment options are needed. Based on studies utilising GCT cell lines, the histone deacetylase inhibitor romidepsin is a promising therapeutic option, showing high toxicity at very low doses towards cisplatin‐resistant GCT cells, but not fibroblasts or Sertoli cells. In this study, we extended our analysis of the molecular effects of romidepsin to deepen our understanding of the underlying mechanisms. Patients will benefit from these analyses, since detailed knowledge of the romidepsin effects allows for a better risk and side‐effect assessment. We screened for changes in histone acetylation of specific lysine residues and analysed changes in the DNA methylation landscape after romidepsin treatment of the GCT cell lines TCam‐2, 2102EP, NCCIT and JAR, while human fibroblasts were used as controls. In addition, we focused on the role of the dehydrogenase/reductase DHRS2, which was strongly up‐regulated in romidepsin treated cells, by generating DHRS2‐deficient TCam‐2 cells using CRISPR/Cas9 gene editing. We show that DHRS2 is dispensable for up‐regulation of romidepsin effectors (GADD45B,DUSP1,ZFP36,ATF3,FOS,CDKN1A,ID2) but contributes to induction of cell cycle arrest. Finally, we show that a combinatory treatment of romidepsin plus the gluccocorticoid dexamethasone further boosts expression of the romidepsin effectors and reduces viability of GCT cells more strongly than under single agent treatment. Thus, romidepsin and dexamethasone might represent a new combinatorial approach for treatment of GCT.


| INTRODUCTION
Testicular type II germ cell tumours (GCTs), which are sub-divided into seminomas and non-seminomas, arise from the precursor lesion germ cell neoplasia in situ (GCNIS). 1,2 GCNIS and seminoma cells are highly similar to primordial germ cells with regard to histology, gene expression and epigenetics, while the stem cell population of the non-seminomas, the embryonal carcinoma (EC), shows features of pluri-to totipotency. 1 Thus, ECs are able to differentiate into cells of all germ layers (teratoma) and into cells resembling extra-embryonic tissues (yolk-sac tumours, choriocarcinomas). 1 Generally, GCTs are treated by orchiectomy followed by additional chemotherapy. By this, high cure rates of >90% are achieved, however, patients with metastatic disease or resistance towards standard chemotherapy require alternative therapeutic options. 3 In previous studies, we demonstrated that treatment of (cisplatin-resistant) GCT cell lines with the histone deacetylase inhibitor (HDACi) romidepsin (ISTODAX, FR228, FR901228) efficiently induced apoptosis and blocked the cell cycle at very low doses, but did not affect survival of fibroblasts or Sertoli cells. [4][5][6] We showed that romidepsin treatment of GCT cell lines resulted in heterochromatin formation within the promotor of ARID1A causing down-regulation of ARID1A, 5 which is a subunit of the chromatin remodelling SWI/SNF-complex. As a result, stress and apoptosis sensors as well as cell cycle regulators GADD45B, ATF3, ZFP36, DUSP1, FOS, ID2 and CDKN1A were up-regulated. 5 In addition, we identified four genes (DHRS2, RHOB, CRISPLD2, BAIAP2), which were up-regulated in all GCT cell lines tested as well as the controls (fibroblasts and the Sertoli cell line FS1), suggesting that these genes represent a common effect of romidepsin on gene expression. 5 Among all samples analysed, DHRS2 was the most prominently up-regulated gene. 5 In this study, we extended our analysis of the molecular mode of action of romidepsin and also focused on the role of DHRS2, a NADPHdependent dehydrogenase/reductase, in the romidepsin-response cascade, to gain a better understanding of the effects of romidepsin on GCTs and normal healthy cells. In general, understanding the molecular effects of a therapeutic drug is required for the assessment of risks and side effects before administering it to a patient.

| Generation of DHRS2-deficient TCam-2 cells
TCam-2 cells heterozygous or homozygous deficient for DHRS2 were generated as published. 5,8 Deletions within the coding sequence of DHRS2 in each clone were detected by PCR ( Figure S1C,D). See Table 1 for guideRNA sequences and genotyping primers.

| DNA, RNA and protein isolation
Genomic DNA, total RNA and proteins were isolated as described previously. 5 Briefly, DNA was isolated by phenol/chloroform/isoamylalcohol precipitation, RNA by the RNAeasy mini kit (Qiagen, Hilden, Germany) and proteins by RIPA buffer.

| Western blot
Western blots were performed as described previously. 5 Beta-ACTIN was used as housekeeper and loading control. See Table 2 for antibody details.

| Quantitative RT-PCR
Quantitative RT-PCR was performed as published previously. 5 500 ng of total RNA was used for first strand synthesis. GAPDH was used as housekeeping gene and for data normalisation. In general, all samples were analysed in technical triplicates and biological triplicates/quadruplicates (see individual figure legend for more detailed information). 2.8 | FACS-based propidium iodide and AnnexinV/ 7AAD measurement FACS-based measurement of cell cycle distribution and apoptosis levels were performed as described previously. 5,6 All samples were analysed in technical and biological triplicates.

| XTT assay
The XTT assay was performed as described previously. 5 2.10 | Chromatin immunoprecipitation followed by sequencing Data of the chromatin immunoprecipitation followed by sequencing experiment are publically available via GEO (GSE78262) and were re-analysed in context of this study. 5 2.11 | Illumina HT-12v4 expression and Infinium 450k DNA methylation array The Illumina expression and DNA methylation array analyses were performed exactly as published. 5,9 The microarray data sets are available via GEO (ncbi.nlm.nih.gov/geo/) (GSE76709; GSE71239; Data S1E).

| Affymetrix expression microarray analysis of GCT tissues
The whole procedure has already been published. 10 The array was re-analysed in context of this study.

| Statistics
We checked for significance of measured values by performing twotailed Student's t-tests. Significance was assumed at P ≤0.05. For all measurements, standard deviations were calculated and given above the bars.

| RESULTS
Previously, we demonstrated that romidepsin causes global hyperacetylation of histones 3 and 4. 5 Now, we addressed the question, whether romidepsin treatment elicits an alteration at specific lysine residues and acts in a cell-type specific manner. We used western blotting to screen for changes in lysine acetylation on histones H3 and H4 16 hours after romidepsin application ( Figure 1A). General efficacy of the romidepsin treatment was validated by detection of pan-H3 and -H4 acetylation. GCT cell lines (TCam-2, 2012EP, JAR) showed considerably higher levels of acetylation compared to human fibroblasts (MPAF). Within the group of GCT samples, non-seminomatous cell lines (2102EP, JAR) showed highest levels of acetylation at all analysed H3-and H4-lysine residues. Four lysine residues (H3K4, H3K14, H3K79, H4K16) showed an increase in acetylation in non-seminomatous cell lines only. Although, the overall increase in acetylation at these lysines was low compared to the other lysine residues analysed. H4K8 acetylation was low before and remained low after romidepsin treatment in all tested cell lines. No lysine residue could be identified that showed a specific increase in acetylation in TCam-2 or MPAF cells. Fibroblasts did not respond as strongly as GCT cells to romidepsin, which is in line with the strongly reduced induction of apoptosis in fibroblasts compared to GCT cells. 5 We analysed if the increase in histone acetylation might also affect the DNA methylation (5-methylcytosine; 5mC) landscape. It has been proposed that acetylated histones are associated with unmethylated DNA and methyated DNA is able to recruit HDACs to repress transcription. 11 In addition, in a previous study we demonstrated that GCT cell lines are able to actively demethylate their DNA via the oxidative pathway involving the TET enzymes, allowing for a rapid change in the 5mC pattern. 12 We performed 5mC microarray analysis (Illumina 450k)  50% change in 5mC to control) (Data S1A).
In our previous study, we observed that DHRS2 was the most prominently up-regulated gene in response to romidepsin in GCT cell lines, prompting us to analyse the role of DHRS2 in more detail. 5 DHRS2 is a NADPH-dependent dehydrogenase/reductase with 3,4hexanedione, 2,3-heptanedione and 1-phenyl-1,2-propanedione as substrates. DHRS2 has been shown to attenuate MDM2-mediated P53 degradation, leading to P53 stabilisation and MDM2/P21 accumulation. 13 First, we demonstrated that DHRS2 is not expressed in different GCT tissues and normal testis tissues ( Figure S1). Furthermore, we demonstrated previously that DHRS2 is also absent in GCT cell lines and strongly induced upon romidepsin treatment of (cisplatin-resistant) GCT cell lines, fibroblasts (MPAF and ARZ) and Sertoli cells (FS1). 5 Thus, up-regulation of DHRS2 is a common effect provoked by romidepsin.
Of note, we screened for changes in expression of other DHRS genes in response to romidepsin in GCT cell lines, including the important DHRS2 paralogue DHRS4 ( Figure S1B). We found that, besides DHRS2, expression levels of all other DHRS molecules analysed remained either unchanged or were down-regulated upon romidepsin treatment, suggesting that the other DHRS genes are not involved in the romidepsin response ( Figure S1B).
We asked how DHRS2 expression might be regulated in GCTs.
As shown by chromatin-immunoprecipitation followed by sequencing in TCam-2 cells 16 hours after romidepsin treatment, the induction of DHRS2 was accompanied by an increase in histone H3 acetylation

| DISCUSSION
In this study, we further characterised the molecular and epigenetic effects of the HDACi romidepsin on GCT cells.
We demonstrated that romidepsin causes hyperacetylation of the majority of H3/H4 lysine residues across GCT cell lines. Furthermore, non-seminomas are more sensitive to romidepsin-provoked changes in histone acetylation than seminomas, indicated by higher levels of acetylation at single lysine residue resolution. In response to romidepsin, we detected acetylation on four lysine residues specifically in EC cell lines (H3K4, H3K14, H3K79, H4K16  Hep27 (encoded by DHSR2) also has non-enzymatic activity, that is, a proteolytically processed form of Hep27 can bind the P53-inhibiting protein MDM2 in the nucleus, leading to accumulation of P53, thereby controlling onset of cell cycle arrest and apoptosis. 13 Although this mechanism seems plausible in explaining the role of DHRS2 in induction of cell cycle arrest, we found that neither expression nor activity (phosphorylation) of P53 is up-regulated/induced upon romidepsin stimulus 5 or DHRS2 knock out (this study). So, this mode of action of Hep27 can be excluded for romidepsin treated GCTs. Further studies will address the question how

DISCLOSURE OF POTEN TI AL CONFLICTS OF INTEREST
The authors confirm that there are no conflicts of interest.