Genome-scale CRISPR screen reveals neddylation to contribute to cisplatin resistance of testicular germ cell tumours

Background Type II testicular germ cell tumours (TGCT) are the most prevalent tumours in young men. Patients suffering from cisplatin-resistant TGCTs are facing very poor prognosis demanding novel therapeutic options. Neddylation is a known posttranslational modification mediating many important biological processes, including tumorigenesis. Overactivation of the neddylation pathway promotes carcinogenesis and tumour progression in various entities by inducing proteasomal degradation of tumour suppressors (e.g., p21, p27). Methods We used a genome-scale CRISPR/Cas9 activation screen to identify cisplatin resistance factors. TGCT cell lines were treated with the neddylation inhibitor (MLN4924)/cisplatin/combination and investigated for changes in viability (XTT assay), apoptosis/cell cycle (flow cytometry) as well as in the transcriptome (3’mRNA sequencing). Results NAE1 overexpression was detected in cisplatin-resistant colonies from the CRISPR screen. Inhibition of neddylation using MLN4924 increased cisplatin cytotoxicity in TGCT cell lines and sensitised cisplatin-resistant cells towards cisplatin. Apoptosis, G2/M-phase cell cycle arrest, γH2A.X/P27 accumulation and mesoderm/endoderm differentiation were observed in TGCT cells, while fibroblast cells were unaffected. Conclusions We identified overactivation of neddylation as a factor for cisplatin resistance in TGCTs and highlighted the additive effect of NAE1 inhibition by MLN4924 in combination with cisplatin as a novel treatment option for TGCTs.


BACKGROUND
In Western countries over the past decades, incidence of testicular germ cell tumours (TGCT) is rising. Type II TGCTs represent the tumour entity with the highest prevalence in young men aged 20-40 years [1,2] arising from primordial germ cells (PGCs), which arrest in early development and forming a germ cell neoplasia in situ (GCNIS). GCNIS eventually progress to seminomas and nonseminomas [3][4][5]. Due to high similarity in global gene expression, epigenetic profile and histology with PGCs, seminomas are viewed as default pathway of TGCT development [6]. Non-seminomas are classified into embryonal carcinomas (EC), choriocarcinomas (CC), yolk-sac tumours (YST) and teratomas (Ter). ECs express markers of totipotency and are therefore considered as type II TGCT stem cell population, which is able to differentiate into CC, YST and Ter [4,7,8]. It has been shown that seminomas are able to reprogramme in a somatic microenvironment towards an EC-like cell fate [9,10]. While 95% of patients with TGCTs are cured by orchiectomy and subsequent cisplatin-based chemotherapy, 20% of metastatic non-seminoma patients develop cisplatin resistance facing poor prognosis due to lack of further treatment options [2,11,12]. Several cisplatin resistance factors have already been described [13][14][15] and grouped in pre-, on-, post-and offtarget effects [15]. Among others, altered cellular drug uptake/ efflux by deregulation of copper transporter 1 and ATP7A/ATP7B expression (pre-target) [16], increased levels of cytoplasmatic scavenger proteins like glutathione (pre-target) [17], upregulated DNA repair systems instead of apoptosis initiation (on-target) [18], overactivated PI3K/AKT signalling supporting p21 induced cell cycle arrest (off-target) as well as upregulated MDM2 expression resulting in p53 degradation thereby avoiding cell death (posttarget) [19,20] were described. All these factors emphasize that cisplatin resistance is multifactorial and cannot be reduced to one key mechanism [15] requiring more research to identify further cisplatin resistance factors in order to reveal potential alternative treatment options for TGCTs.
The process of neddylation (Fig. 1a) describes the conjugation of the ubiquitin-like small molecule NEDD8 (neuronal precursor cell expressed developmentally downregulated protein 8) to a target protein. Neddylation changes the function, stability or subcellular location of the target protein [21,22]. NEDD8 is conjugated to its target protein in a canonical manner by initial binding to the E1-NEDD8 activating enzyme (NAE1/UBA3 complex), further transferred via NEDD8-conjugating enzyme E2 (UBE2M or UBE2F) and substrate-specific NEDD8-E3 ligase (RBX1/2, DCN1, etc.) to the final substrates (cullins and noncullins). Cullins are subunits of the Cullin-RING-ligases (CRL) which are ubiquitin E3 ligases. After neddylation, the CLRs mediate polyubiquitylation of the substrates thereby labelling them for proteasomal degradation. In the cell 20% of all proteasomal degraded proteins are processed via a CLR-based mechanism + lentiSAMv2 plasmid library pMD2.G and psPAX2  [23,24]. Various substrates have been identified including several tumour suppressor proteins like p21, p27, Wee1 and p53 [25].
Overactivation of neddylation leads to increased degradation of these proteins promoting malignant transformation and tumour progression. This has been demonstrated in lung [25], breast [26] and pancreatic cancer [27]. Reduction of neddylation applying the covalent NAE1 inhibitor MLN4924 impaired cell cycle progression and induced apoptosis due to accumulation of p21, p27, Wee1, p53, etc. [25][26][27]. Further, Zheng et al. were able to show increased cisplatin sensitivity in pancreatic cancer triggered by NAE1 inhibition [27]. To date, 40 clinical trials investigating the potential of MLN4924 in terms of cancer treatment have been enrolled (https://www.clinicaltrials.gov/), indicating versatility and potency of the drug [24].
In this study, we applied a genome-scale CRISPR/Cas9 activation screen on parental TGCT cell lines to generate cisplatin-resistant clones. Investigation of candidate genes allowed us to find and describe pathways leading to cisplatin resistance. We identified upregulation of NAE1, a core component of neddylation, to induce cisplatin resistance in 2102EP and JAR cell lines. Application of MLN4924, an inhibitor of neddylation, induced cell cycle arrest and apoptosis in TGCT cell lines. Treatment with both MLN4924 and cisplatin resulted in an additive effect enhancing cell death. Cisplatin-resistant TGCT cell lines were found to revert to cisplatin sensitivity after treatment with MLN4924.

MATERIALS AND METHODS Cell culture
The following cell lines were used: 2102EP, 2102EP-R, NCCIT, NCCIT-R, NT2/D1, NT2/D1-R (all ECs), JAR (choriocarcinoma), TCam-2 (seminoma) and HEK293T (human embryonic kidney). As a control MPAF cells (human adult fibroblast) were included. Cell lines with the suffix "-R" represent cisplatin-resistant sublines from the corresponding parental cell line [28,29]. Cells were cultured as described previously [6,30,31]. Short tandem repeat (STR) profiles of all cell lines are checked on a regular basis at the Institute of Forensic Medicine (University Hospital Düsseldorf) and are available upon request (to DN). All cell lines were negative for mycoplasma infection as checked by PCR. Cell line sources are described in Supplementary Methods.

Genome-scale CRISPR/Cas9 transcriptional activation screen
The screen was performed according to Joung et al. [32]. The human CRISPR activation (SAMv2) plasmid library (Supplementary Table S1) was amplified in Endura ElectroCompetent cells (Lucigen, Middleton, WI, USA). To check the sgRNA coverage, next-generation sequencing using NextSeq™550 (Illumina, San Diego, CA, USA) and Python-/Biopython-based analysis were performed [33][34][35]. Lentivirus production was performed in HEK293T cells. For concentration Lenti-X-concentrator was used according to the manufacturer's protocol (Takara BIO INC., Kusatsu, Japan). Multiplicity of Infection (MOI) was determined using transduction via centrifugation and the CellTiter-Glo™ Luminescent Cell Viability Assay Kit (Promega, Madison, WI, USA). This process is described in detail in Supplementary Methods. For the SAM activation approach, 2102EP MPHv2 and JAR MPHv2 helper cell lines stably expressing p65-HSF1 transactivation domains were generated by transduction with MPHv2 lentivirus. The library transduction (1500 × g, 30 min, 32°C) was performed in six-well cell culture plates (10 6 cells/well) at MOI of 0.3 (Supplementary Table S2). After 2 days of incubation, including two PBS wash steps (24 and 48 h), cells were transferred to 15 cm cell culture dishes. One day later, cisplatin treatment was applied until all cells in the wild-type control had died.
After recovery of the surviving cells, 2102EP MPHv2/SAMv2 , JAR MPHv2/SAMv2 and corresponding wild-type cells were re-seeded in six-well cell culture plates (0.25 × 10 4 cells/well). To validate cisplatin resistance, 2102EP, as well as JAR cell lines, were treated with 2.5 µM and 5 µM cisplatin, respectively. Media containing cisplatin was changed every 2nd/3r d day for 2 weeks. The effect on the cells was monitored by bright-field microscopy.
To identify candidate genes, genomic DNA was isolated from 2102EP MPHv2/SAMv2 and JAR MPHv2/SAMv2 cells (Supplementary Table S2), PCR amplification of the region encoding the sgRNAs was performed with NEBNext ® High-Fidelity 2X PCR Master Mix (Ipswich, MA, USA) and obtained PCR products were analysed by next-generation sequencing on the NextSeq™550 device (Illumina). Read counts were analysed using the Python/Biopython-based [33][34][35] count_spacers.py algorithm described by Joung et al. [32]. Further, candidate genes of 2102EP MPHv2/SAMv2 and JAR MPHv2/SAMv2 with read counts >10,000 were evaluated using the STRING 11.5 database [36]. Protein interaction networks were further investigated for overrepresented pathways using the Reactome analysis tool [37]. The CRISPR/Cas9 activation screen was performed n = 1.

Generation of clonal cell lines overexpressing NAE1
Production of lentiviral particles was performed in HEK293T cells using two helper plasmids (pMD2.G, psPAX2) and the pLV-NAE1-GFPSpark plasmid encoding a NAE1/GFP fusion protein (Sino Biological, Beijing, China). For virus concentration, the Lenti-X-concentrator (Takara BIO INC.) was applied according to the manufacturer's protocol. JAR and 2102EP cells were transduced with 50x concentrated lentivirus and expanded for 6 days. For the generation of 2102EP NAE1/GFP and JAR NAE1/GFP clonal cell lines, single cell FACS sorting was performed using the BD FACS ARIA III/BD FACS Melody (BD Biosciences, East Rutherford, NJ, USA). To discriminate dead cells DAPI (AppliChem GmbH, Darmstadt, Germany) staining was applied (5 µg/ml). GFP (NAE1) positive, DAPI-negative (alive) cells were collected in 96-well-cell culture plates and expanded.

Protein extraction, SDS-PAGE and western blot
Protein extraction using RIPA buffer (Cell Signaling, USA), SDSpolyacrylamide gel electrophoresis (SDS-PAGE) and Western Blot analysis were performed as described previously [6,31]. Information on primary and secondary antibodies used are provided in Supplementary Table S3.

Cell viability assay
After treatment of TGCT cells with MLN4924 (solvent dimethylsulfoxid (DMSO)), cisplatin (solvent dimethylformamide (DMF)) or a combination of both drugs, XTT viability assay was performed as described [38]. Means of technical triplicates were calculated and related to solvent control (DMSO, DMF, combination). This procedure was repeated three times resulting in three independent biological replicates (n = 3). Viability assays of NAE1 overexpression cell lines treated with cisplatin or solvent were performed n = 4.

DAPI/AnnexinV (apoptosis) and Hoechst-FACS analysis (cell cycle)
One day after seeding in 6-well cell culture plates (1.5 × 10 5 cells/well), treatment with MLN4924, cisplatin, combination of both drugs or solvent (DMSO, DMF or DMSO/DMF combination) was performed. Subsequent FACS-based apoptosis (DAPI/PE AnnexinV) and cell cycle analysis was applied as described (n = 3) [31]. For apoptosis analysis, combination treatment was compared to MLN4924 or cisplatin-only application while Fig. 1 CRISPR/Cas9 screen for cisplatin resistance factors. a Illustration of the neddylation pathway. NEDD8 is conjugated to an E1 complex and transferred in a canonical manner via E2 and E3 ligases to the substrates which are divided into cullins and non-cullins. MLN4924 inhibits neddylation cascade via NAE1 binding. b Schematic representation of the CRISPR/Cas9 screen using 2102EP and JAR cells. Lentiviral particles were produced in HEK293T cells based on the lentiSAMv2 (activation approach) plasmid library. For the genome-scale activation screen, a stable JAR MPHv2 and 2102EP MPHv2 cell line expressing MS2-p65-HSF1 were transduced with the lentiSAMv2 library encoding dCas9-VP64 and sgRNAs. Gene-edited cells were treated with cisplatin, and surviving cells were recovered in cell culture medium. To identify candidate genes genomic DNA was isolated, sgRNA encoding region was amplified and analysed via next-generation sequencing. c Identification of candidate genes for cisplatin resistance with a read count >10,000 are depicted for JAR MPHv2/SAMv2 and 2102EP MPHv2/SAMv2 cells. d Venn diagram showing unique and common candidate genes for JAR MPHv2/SAMv2 and 2102EP MPHv2/SAMv2 cell lines in the activation screen. e STRING interaction and Reactome pathway analysis on JAR MPHv2/SAMv2 as well as 2102EP MPHv2/SAMv2 candidate genes. The CRISPR/Cas9 screen was performed n = 1.
for cell cycle analysis, the cell cycle phases of the drug-treated groups were compared to the solvent-treated controls (n = 3).

Statistical analysis
Independent biological replicates are displayed as mean ± SD (standard deviation) given as error bars. For statistical comparison of two groups, two-tailed Student's t test was applied. Datapoints with significant difference (P < 0.05) were indicated with asterisks.

RESULTS
Application of genome-scale CRISPR/Cas9 activation screen yields cisplatin-resistant TGCT cells A saturating CRISPR/Cas9 SAMv2 transcriptional activation screen was applied to generate cisplatin-resistant JAR MPHv2/SAMv2 / 2102EP MPHv2/SAMv2 cells as outlined in Fig. 1b and Supplementary  Fig. S1A. The SAMv2 plasmid library was amplified, purified and analysed by next-generation sequencing (NGS) revealing a coverage of more than 99.9% of sgRNAs and a skew ratio below 6 indicating equal distribution of the sgRNAs, important for maximum screening efficiency ( Supplementary Fig. S2A, B). To reach a sgRNA coverage of 100% >1.1 × 10 8 JAR MPHv2 and 2102EP MPHv2 cells were transduced with the SAMv2 library. Next, cisplatin treatment was started and cytotoxic effects were monitored ( Supplementary Fig. S1B, C). Cell density decreased in the gene-edited cells as well as in the control cells. However, during the recovery phase in medium without cisplatin several gene-edited cells restarted proliferation, indicating that SAMmediated activation of target genes had induced a selective advantage, the control cells died. To confirm cisplatin resistance, surviving cells were re-treated with cisplatin. Bright-field images of the gene-edited cells portended a higher viability compared to the treated control indicated by higher cell number, suggestive of cisplatin resistance ( Supplementary Fig. S3A, B).
To identify genes responsible for cisplatin resistance, the integrated sgRNA regions were analysed using NGS. All candidate genes with a read count >10,000 are depicted in Fig. 1c. A complete list is provided in Supplementary Fig. S4 and Supplementary Data Set S1. Interestingly, when comparing candidate genes identified in 2102EP MPHv2/SAMv2 and JAR MPHv2/SAMv2 , only 3 out of 93 genes overlapped, suggesting cell line/tumour type-specific mechanisms of cisplatin resistance (Fig. 1d). STRING interaction analysis of JAR MPHv2/SAMv2 candidate genes revealed a network including TRAP1, CDC37, DDB1, NAE1, POLE4, NUP133, SGOL2 and TK1. Reactome database analysis performed on this network resulted in significantly enriched terms associated with cell cycle, DNA repair and neddylation (Fig. 1e). For 2102EP MPHv2/SAMv2 , three candidate genes (TK1, DDB1 and AKT3) were found to be part of these pathways. Since disturbed cell cycle and DNA repair are already known in the context of cisplatin resistance in TGCTs [41], we decided to investigate the role of NAE1-mediated neddylation.

Overexpression of NAE1 results in cisplatin resistance
To confirm the results from the genome-scale screen, clonal JAR NAE1/GFP and 2102EP NAE1/GFP cell lines overexpressing NAE1 were generated. A CMV-driven NAE1-GFP fusion was transduced ( Fig. 2a, b), and protein level was analysed via western blot (Fig. 2c, d). The band at 87 kDa was detected with both, NAE1-and GFP-antibody suggesting a NAE1-GFP fusion protein. Wild-type NAE1 was detected at 60 kDa. Further, an additional GFP band at 27 kDa is suggestive of cleavage of the protein.
Neddylation inhibitor MLN4924 in combination with cisplatin effectively reduces cell viability in TGCT cell lines If overexpression leads to cisplatin resistance, then inhibition of neddylation should result in cisplatin sensitivity. We performed NAE1 gene expression meta-analysis of microarray data on different cell lines and tissues previously published by us [42,43]. TGCT cell lines as well as TGCT tissues revealed high levels of NAE1 mRNA ( Fig. 3a and Supplementary Fig. S5A) while expression was lower in MPAF cells and normal testes tissue. Further, Western Blot analysis demonstrated high NAE1 protein levels for 2102EP, 2102EP-R, NCCIT, NCCIT-R, NT2/D1, NT2/D1-R, JAR and TCam-2 cell lines (Fig. 3b). MPAF cells revealed lower levels of the target protein. Further, using Firebrowse database analysis higher NAE1 expression in TGCTs compared to other tumour entities were found ( Supplementary Fig. S5C). UCSC Xena browser-mediated analysis of NAE1 isoforms in TGCT tissue compared to normal testis tissue showed differential expression of several isoforms ( Supplementary Fig. S5B). Next, we performed Western blot analysis revealing increased protein levels of p27 in 2102EP, JAR and TCam-2 cell lines ( Supplementary Fig. S6A-C) treated with MLN4924, pointing towards suppression of neddylation [24,44]. On the same blots, we detected an accumulation of γH2A.X after cisplatin treatment indicating DNA damage induction [45,46]. Consequently, increased p27 and γH2A.X levels were also found in the cells treated with MLN4924/cisplatin combination suggesting inhibition of neddylation and induction of DNA damage, which might trigger a stronger cytotoxic effect compared to monotherapy.

MLN4924 and cisplatin induce apoptosis and cell cycle arrest
For further investigation of cellular effects, we focused on 2102EP, JAR, TCam-2 and MPAF cells. Apoptosis was analysed by FACS after 2 and 5 days of MLN4924, cisplatin or combination treatment (Fig. 4a). In 2102EP cells induction of apoptosis could be observed already after 2 days of treatment. Apoptosis induction at day 2 was 3.2 times stronger upon co-treatment, increasing to 3.6 times  100 μm). c, d Western Blot analysis detecting GFP, NAE1 and β-actin (loading control) in JAR and 2102EP cells. e JAR NAE1/GFP 2, 3, 5 and JAR WT as well as (f) 2102EP NAE1/GFP 1, 2, 3 and 2102EP WT cell lines were treated with 5 and 2.5 μM cisplatin, respectively. Next, viability was measured via XTT assay at days indicated and was normalised to DMSO control which was set to 100% and to the wild-type cell line at the respective day. Asterisks indicate a significant difference between NAE1/ GFP cell lines and WT line at each time point (*P < 0.05). n = 4. Data represent independent biological replicates mean ± SD.

Transcriptome analysis revealed apoptosis induction, cell cycle arrest and cell differentiation after MLN4924/cisplatin combination treatment
To further investigate the molecular mechanisms of MLN4924 and cisplatin treatment in TGCTs, the transcriptome of 2102EP as well as JAR cells was analysed by 3'mRNA sequencing after 2 days of treatment (Supplementary Data Set S1).
STRING interaction analysis of differentially upregulated genes in 2102EP and JAR cells after MLN4924, cisplatin and combination treatment revealed interaction networks which were associated with the GO terms apoptosis and cell death (Fig. 5a, b and Supplementary Fig. S7A-F). The data suggested apoptosis induction after single drug and combination therapy, confirming the results from the FACS analysis. Further, based on the fold change of the top 10 upregulated genes in the network of 2102EP cells, apoptosis initiation appeared stronger upon combination treatment. Comparing the transcript levels of KRT17, an important player in apoptosis, showed a log2 fold change of 5.8/6.8/9.9 after MLN4924/cisplatin/combination treatment, respectively. In the JAR cell line, the effect of combination treatment was even more obvious. Here, the interaction network for MLN4924/cisplatin includes 48 proteins, while for MLN4924 only and cisplatin only there were just 10 proteins found for each condition associated with apoptosis and cell death. The increased apoptotic effect due to combination treatment can be explained when comparing the top ten upregulated genes in the interaction network, which showed strongly increased transcript abundancy compared to MLN4924 or cisplatin-only application.
Further, transcriptome analysis revealed cell cycle arrest, indicated by comprehensive STRING networks for 2102EP cells treated with MLN4924 and combination treatment (Supplementary Fig. S8A-F). MLN4924 treated 2102EP cells revealed a complex network with differentially downregulated genes like CCNA1, TUBB4B, KIFC1, BUB3 suggesting downregulation of cell cycle progression. In contrast, cisplatin application in 2102EP cells resulted only in a small network of upregulated genes associated with cell cycle arrest, including the cell cycle regulator p21 (CDKN1A). Combination treatment revealed a network of differentially upregulated genes involved in negative regulation of cell cycle, like SFN, GPER, etc. In JAR cells, many differentially downregulated genes associated with cell cycle progression were found ( Supplementary Fig. S9A-F). Especially after combination treatment, the results tended towards decreased cell cycle processes (FES, HSF1, PIN1), while after MLN4924 or cisplatin application only few cell cycle-related transcripts were differentially expressed.
Analysis of 3'mRNA sequencing data after two days of treatment with MLN4924, cisplatin or combination revealed initiation of cellular differentiation in 2102EP and JAR cells (Fig. 6a, b and Supplementary Fig. S10A-F). We found highly increased transcript abundance of cell differentiation markers, like HAND1, CDX2, CLDN1 and SOX15 after all three treatment conditions in 2102EP cells. Cotreatment of JAR cells also revealed strong upregulation of CDX2 and SOX15. Especially, the combination treatment seemed to drive the 2102EP and JAR cells into mesoderm and endoderm fate, which was supported by differential upregulation of marker genes like MIXL1, COL7A1, SMAD3, TXNRD1, DUSP5 and TBX3. In parallel to the induction of differentiation, pluripotency markers, such as SOX2, SOX21, KLF15, TCFL1 and HESX1 (only deregulated in MLN4924/ cisplatin sample) were strongly downregulated in 2102EP cells treated with MLN4924 only or MLN4924 in combination with cisplatin ( Supplementary Fig. S11A-D).

DISCUSSION
Here, we used a genome-scale CRISPR/Cas9 activation screen to uncover pathways contributing to cisplatin resistance in TGCTs. Among known genes involved in DNA repair and cell cycle control, we identified components of the neddylation pathway to mediate resistance. Overexpression of NAE1 in 2102EP and JAR cells resulted in increased cisplatin resistance, validating our finding. In turn, inhibition of neddylation increased cisplatin sensitivity of TGCT cell lines. The combination treatment of the covalent NAE1 inhibitor MLN4924 and cisplatin revealed a highly therapeutic efficacy, not only in the tested TGCT parental cell lines, but also in the cisplatin-resistant lines. Further, we found that the fibroblast control cells were unaffected by the combination therapy, suggesting no severe side effects. Next, we could show that MLN4924 in combination with cisplatin resulted in stronger G2/M-phase cell cycle arrest and increased apoptosis induction compared to monotherapy. In 2102EP, JAR and TCam-2 cells on protein level strong accumulation of p27 after MLN4924 only and MLN4924/cisplatin treatment was found. Further, accumulation of γH2A.X after cisplatin and MLN4924/cisplatin treatment was observed validating known targets of MLN4924 and cisplatin. Transcriptome analysis in 2102EP and JAR cells revealed strong induction of meso-and endoderm fate, which was most prominent after combination treatment. In 2102EP cells, downregulation of pluripotency markers was observable after MLN4924 only and MLN4924/cisplatin combination treatment.
The neddylation core component NAE1 was the most interesting factor we found in the genome-scale CRISPR/Cas9 activation screen. Surprisingly, there is nothing known about the role of neddylation in TGCTs. Overactivation of neddylation in different Drug application was repeated after 2 days. Treated samples were normalised to the solvent control referred to as fold change 1. Asterisks indicate significant difference in apoptosis between MLN4924 only or cisplatin only compared to the combination treatment (n.s. not significant, *P < 0.05, **P < 0.01). (n = 3). b For Hoechst-FACS-based cell cycle analysis cells were measured after 24 and 48 h of treatment (same concentrations as indicated above). Solvent-treated cells served as control sample. Asterisks indicate significant difference of cell portion in the respective cell cycle phase after treatment compared to control sample (*P < 0.05, **P < 0.01). (n = 3) Data represent independent biological replicates mean ± SD.
tumour entities like liver, lung and colon cancer suggest an important role in carcinogenesis. In pancreatic cancer overexpression of NAE1 has even been shown to be responsible for cisplatin resistance [27]. Further, we found overactivated DDB1 in cisplatin-resistant TGCT cells which is aside from DNA repair also involved in the neddylation cascade. In detail, DDB1 is part of the CRL4 complex which in cisplatin-resistant ovarian cancer seems to be considerably overactivated. There cisplatin resistance could be overcome by knockdown of CLR4 [47]. Here, we suggest overactivated NAE1/CLR4 axis to represent a putative pathway for cisplatin resistance. In this study, we could already proof that overexpression of NAE1 in 2102EP NAE1/GFP and JAR NAE1/GFP cells induced cisplatin resistance which makes neddylation a very interesting and promising therapeutic target in treatment of TGCTs.
The genome-scale CRISPR/Cas9 activation screen we applied in our study also revealed already known cisplatin resistance factors, like altered cell cycle and disturbed DNA repair mechanisms in TGCTs LGALS7B LGALS7 LGALS7 LGALS7B [ 14,15]. In cisplatin-resistant JAR MPHv2/SAMv2 and 2102EP MPHv2/SAMv2 cells, we found POLE4 as well as AKT3 to be upregulated, respectively. Both proteins have already been described as factors contributing to cisplatin resistance in TGCTs [14,15,48,49]. POLE4 is a subunit of DNA polymerase ε playing a role in nucleotide excision repair, base excision repair, DNA double-strand break repair and proofreading, supporting genome stability [48,49]. High levels of AKT3 induce overactivation of the PI3K/AKT signalling pathway preventing cell cycle arrest and apoptosis [14,15]. These results clearly validate the CRISPR screen since we found not only novel but also already known factors of cisplatin resistance in TGCTs. Further, in cisplatin-resistant cells from our screen upregulated genes like TRAP1, TK1 and SGOL2 were identified. Elevated expression level of the mitochondrial heat shock protein TRAP1 is associated with cisplatin resistance in osteosarcoma, colorectal, prostate and lung cancer due to reduction of reactive oxygen species production and prevention of apoptosis. Knockdown of TRAP1 increased sensitivity toward cisplatin [50,51]. Thus, TRAP1 could also be an interesting prognostic indicator and therapeutic target for cisplatin-resistant TGCTs. TK1 and SGOL2 have been identified as tumour progression factors in other tumour entities [52,53]. Increased TK1 protein levels were found to serve as a biomarker for poor prognosis after initial chemotherapy of lung, ovarian and breast cancer [52]. SGOL2 is defined as an oncogene in hepatocellular carcinoma cells [53]. We hypothesize that upregulation of TK1 and SGOL2 promotes cell proliferation in TGCTs thereby initiating cisplatin resistance which makes these genes interesting candidates for further investigations.
The genome-scale CRISPR/Cas9 screen applied here revealed already known and novel factors responsible for cisplatin resistance in TGCTs. Since the screen was performed n = 1 we do not claim to display a comprehensive list of all possible cisplatin resistance factors in TGCTs. Instead, this study revealed several interesting candidate genes which might contribute to cisplatin resistance in TGCTs. The screen can only deliver initial suggestions which have to be confirmed as shown for NAE1.
Next, treating TGCT cells with both MLN4924 and cisplatin revealed an additive cytotoxic effect displayed by decreased viability. Of note, MLN4924/cisplatin combination treatment in a xenograft mouse model of pancreatic cancer resulted in significantly decreased tumour volume [27]. Further, cotreatment of the TGCT cell lines led to G2/M-phase cell cycle arrest and apoptosis induction. Application of both drugs also resulted in protein accumulation of p27 and γH2A.X indicating the additional effect of combination treatment. In other tumour entities like lung cancer and oesophageal squamous cell carcinoma MLN4924 has been shown to induce G2/M-phase cell cycle arrest [25,54]. Both publications describe the accumulation of p21, p27 and Wee1 after MLN4924 treatment as a cause for cell cycle arrest and observed subsequent apoptosis [25,54]. Cisplatinbased apoptosis follows a similar pattern, including DNA damage indicated by increased γH2A.X levels, subsequent G2/M-phase cell cycle arrest and Caspase as well as PUMA and NOXA-mediated apoptosis [41,48]. Thus, our findings are in accordance with the literature revealing MLN4924 and cisplatin combination treatment leads to enhanced G2/M-phase cell cycle arrest and subsequent apoptosis induction in TGCTs.
Interestingly, in 2102EP as well as in JAR cells, after MLN4924, cisplatin and combination treatment we found mRNA levels of KRT17 to be highly upregulated. According to the gene ontology analysis KRT17 is involved in apoptosis induction. In pancreatic cancer overexpression of KRT17 was shown to decrease CyclinD1 levels and increase cleaved caspase 3 amounts causing cell cycle arrest and apoptosis [55]. Thus, upregulation of KRT17 in JAR and 2102EP cells might contribute to induction of apoptosis. We observed reduced transcript levels of SOX21 and of the core EC marker SOX2 after MLN4924 only and cisplatin/MLN4924 combination treatment of 2102EP cells, suggesting downregulation of pluripotency. At the same time in 2102EP as well as in JAR cells treated with both drugs strong induction of cell differentiation was found. Especially mRNA levels of CDX2 and HAND1 were highly upregulated in 2102EP cells after treatment whereas in JAR cells a moderate upregulation has been detected. Concomitant downregulation of SOX2 and SOX21 while maintaining CDX2 levels induces cell differentiation [56]. In 2102EP mesoderm fate is induced indicated by upregulation of 13 mesoderm-associated genes including HAND1, LEF1, etc. JQ1 mediated inhibition of BRD2/ 4 results in downregulation of pluripotency and induction of cell differentiation (mesoderm development) in NCCIT cells [6]. Differentiation of tumour cells can be advantageous by limiting their progression and metastatic potential. Especially ECs show reduced tumorigenicity due to treatment with differentiation agents like thioridazine and salinomycin [57]. Surprisingly, MLN4924 treatment of JAR cells resulted in upregulation of four mesoderm-(TXNRD1, BMP7, etc.) and four endoderm-associated (DUSP4, MIXL1, etc.) genes. Of note, JAR is a cell line derived from choriocarcinoma and displays extraembryonic characteristics. Due to the absence of upregulation of EC markers we speculate that MLN4924 induces a direct reprogramming from extraembryonic to meso-and endoderm fate. Since we previously demonstrated reprogramming from seminoma to EC-like fate and from seminoma to mixed-non-seminoma fate [9,10,58], conversion of JAR cells reported here further underscores the inherent plasticity of TGCTs.
CDX2 overexpression can be indicative for differentiation into YST and Ter. Further, we found upregulation of GATA3 and KRT17 in 2102EP cells after combination treatment. Increased GATA3 levels were detected in several YST patients but seem to be more common in CC [59]. Elevated KRT17 expression rather points towards Ter differentiation [60]. However, since the key markers α-Fetoprotein, Glypican-3, β-HCG are not expressed differentiation into YST, CC or Ter seems rather unlikely.
In the context of current pre-, on-, post-or off-target effect classification of cisplatin resistance, we consider global upregulation of neddylation as a pre-or a post-target effect. It remains elusive whether patients with resistant TGCTs display already high levels of NAE1 or whether the treatment itself upregulates NAE1 levels. Increased neddylation of substrates such as cullins results in subsequent degradation of many cellular proteins, like tumour suppressors (p21, p27) and failure of apoptosis induction. In turn, inhibition of neddylation by MLN4924 results in accumulation of cell cycle inhibitors p21, p27, Wee1, etc. and factors for induction of apoptosis (e.g., NOXA accumulation) [24].
In other tumour entities an additive effect of MLN4924 in combination with cisplatin was observed at MLN4924 concentrations which were in the same range (0.2 to 1 µM) [27,54] as in our study for the TGCTs. Here, MLN4924 treatment resulted in resensitisation of cisplatin-resistant cells. Interestingly, MLN4924 inhibitor in combination with cisplatin did not show any effect on the fibroblast control cells. NAE1 expression is highest in TGCTs compared to other tumour entities and normal tissues. According to The Human Protein Atlas and the Firebrowse database, in TCGT tissues neddylation pathway members are expressed at moderate to high levels. Thus, MLN4924 in combination with cisplatin appears as an attractive treatment option in TGCTs. At present, 40 clinical trials investigate the potential of MLN4924 in various tumour entities (https://www.clinicaltrials.gov/). This will allow an off-label use for patients suffering from TGCTs as soon as MLN4924 is approved.
To conclude, we identified upregulation of neddylation to contribute to cisplatin resistance in TGCTs. Combination treatment of NAE1 inhibitor MLN4924 and cisplatin elicits a strong additive cytotoxic effect in TGCTs while leaving fibroblast control cells unaffected. Further, inhibition of neddylation by MLN4924 effectively re-sensitised TGCT-R cells towards cisplatin. The fact that TGCTs display the highest levels of NAE1 compared to other human tumour entities or regular tissue suggest neddylation to be potential novel therapeutic option for TGCTs.

DATA AVAILABILITY
Raw and processed 3'mRNA sequencing data generated in this study have been deposited in the Gene Expression Omnibus database under accession code GSE216043. The raw and processed DNA sequencing data (CRISPR/Cas9 screen) are available in the Gene Expression Omnibus database under accession code GSE225852. The Cancer Genome Atlas' (TCGA) datasets can be accessed via the UCSC Xena browser (https://xena.ucsc.edu) and Firebrowse (https://www.http:// firebrowse.org/). Protein expression datasets are available via The Human Protein Atlas (https://www.proteinatlas.org/).