The FA/BRCA Pathway Identified As the Major Predictor of Cisplatin Response in Head and Neck Cancer by Functional Genomics

Published OnlineFirst December 15, 2016; DOI: 10.1158/1535-7163.MCT-16-0457

Companion Diagnostics and Cancer Biomarkers Molecular Cancer Therapeutics The FA/BRCA Pathway Identiﬁed as the Major Predictor of Cisplatin Response in Head and Neck Cancer by Functional Genomics Sanne R. Martens-de Kemp1, Arjen Brink1, Ida H. van der Meulen2, Renee X. de Menezes3, Dennis E. te Beest3, C. Rene Leemans1, Victor W. van Beusechem2, Boudewijn J.M. Braakhuis1, and Ruud H. Brakenhoff1

Abstract

Patients with advanced stage head and neck squamous cell Fanconi anemia/BRCA pathway as the predominant pathway for carcinoma (HNSCC) are often treated with cisplatin-containing cisplatin response in HNSCC cells. We also identified the involve- chemoradiation protocols. Although cisplatin is an effective radi- ment of the SHFM1 gene in the process of DNA cross-link repair. ation sensitizer, it causes severe toxicity and not all patients benefit Furthermore, expression profiles based on these genes predict the from the combination treatment. HNSCCs expectedly not prognosis of radiation- and chemoradiation-treated head and responding to cisplatin may better be treated with surgery and neck cancer patients. This genome-wide functional analysis des- postoperative radiation or cetuximab and radiation, but biomar- ignated the genes that are important in the response of HNSCC to kers to personalize chemoradiotherapy are not available. We cisplatin and may guide further biomarker validation. Cisplatin performed an unbiased genome-wide functional genetic screen imaging as well as biomarkers that indicate the activity of the in vitro to identify genes that influence the response to cisplatin in Fanconi anemia/BRCA pathway in the tumors are the prime HNSCC cells. By siRNA-mediated knockdown, we identified the candidates. Mol Cancer Ther; 16(3); 540–50. 2016 AACR.

Introduction the application of chemoradiation and to treat only those patients whose tumors are likely to respond to the combination therapy. Cancer of the head and neck is the eighth most commonly Major research efforts aimed to find clinical and biomolec- diagnosed cancer worldwide (1). Over 90% of the head and neck ular markers that predict chemoradiation response of HNSCC. cancers are squamous cell carcinomas that arise in the mucosal However, the only clinical factor that shows some predictive linings of the upper aerodigestive tract. Approximately 60% of value for chemoradiotherapy outcome in HNSCC proved to be head and neck squamous cell carcinoma (HNSCC) patients primary tumor volume (4–11), and not even all studies could present with advanced stage disease (stage III and IV), which confirm this (12). relates to a poor prognosis (2). Treatment options for this group Using a candidate gene or protein approach, several biological are surgery followed by radiotherapy or cisplatin-containing and genetic markers have been studied to predict chemoradiation chemoradiation protocols with salvage surgery if needed. outcome in head and neck cancer patients but none found its Although chemoradiation is effective (3), not all tumors way to the clinic (13–20). Other groups have exploited microarray respond well to this combination of cisplatin and radiotherapy. technology to determine expression profiles that might predict Cisplatin is an effective and inexpensive addendum to radiation treatment response in head and neck cancer (12, 21–28), but a protocols, but only between 5% and 10% of the patients benefit predictive profile has not been validated. from the combination at the expense of severe toxicity. Patients There are several explanations why these approaches have not need to be hospitalized during cisplatin infusion and acute and led to breakthroughs. First, tumor biopsies may contain mixed cell long term toxicities are frequent. It would be ideal to personalize populations, and particularly the small subpopulations of treatment-resistant cells (e.g., cancer stem cells) may not be analyzed 1Department of Otolaryngology-Head and Neck Surgery, VU University Medical by expression profiling, while they might be highly relevant for Center, Amsterdam, the Netherlands. 2RNA Interference Functional Oncoge- treatment outcome (29). Second, genes harboring an activating or nomics Laboratory, Department of Medical Oncology, VU University Medical inactivating mutation, but that are expressed at close to normal 3 Center, Amsterdam, the Netherlands. Department of Clinical Epidemiology and levels, will not be identified as predictors of treatment outcome. Biostatistics, VU University Medical Center, Amsterdam, The Netherlands. We therefore aimed to identify all genes that have an important Note: Supplementary data for this article are available at Molecular Cancer role in the response to cisplatin using a functional genomics Therapeutics Online (http://mct.aacrjournals.org/). approach. Unveiling these genes might enable us to find biomar- Corresponding Author: Ruud H. Brakenhoff, VU University Medical Center, PO kers that can be used to predict cisplatin-based chemoradiation Box 7057, Amsterdam 1007 MB, the Netherlands. Phone: 312-0444-0953; Fax: outcome and to personalize HNSCC treatment by only treating 312-0444-3688; E-mail: [email protected] those patients who would benefit from the therapy. doi: 10.1158/1535-7163.MCT-16-0457 The application of loss-of-function high-throughput RNA 2016 American Association for Cancer Research. interference screens has become an important genomic tool in

540 Mol Cancer Ther; 16(3) March 2017

The FA/BRCA Pathway Predicts Cisplatin Response in HNSCC

research (30–33), although they are rapidly taken over by CRISPR- 37 C/5% CO2. Cell viability was calculated on the basis of Cas9–mediated gene editing approaches. Especially, for the iden- fluorescence measurements after 2 hours incubation with 20 mL tification of genes that modulate drug response, screens using of a 1:1 dilution of CellTiter Blue (Promega) in cell culture siRNAs are increasingly used (34–43). For example, using this medium. Fluorescence was measured using an Infinite 200 micro- technology, Posthuma De Boer and colleagues (44) identified plate reader (Tecan). JIP1 as a gene that modulates the sensitivity of osteosarcoma cells to doxorubicin. We hypothesized that a genome-wide siRNA Genome-wide siRNA screen screen could be the best way to functionally identify all genes VU-SCC-120 cells were subjected to high-throughput forward and signaling pathways involved in cisplatin response. Therefore, transfection as described before (49). In summary, 1,000 cells we employed an arrayed library of 21,121 pools of siRNAs, were seeded and the next day transfected with 25 nmol of each targeting unique human genes in the NCBI RefSeq database siRNA SMARTpool derived from the siARRAY Human Genome (www.ncbi.nlm.nih.gov/RefSeq and ref. 45). This one-gene- library [Catalog items G-003500 (Sept05), G-003600 (Sept05), one-well approach allowed us to comprehensively interrogate G-004600 (Sept05), and G-005000 (Oct05); Dharmacon, the cellular response to cisplatin when expression of each indi- Thermo Fisher Scientific] and 0.03 mL DharmaFECT1 (Thermo vidual gene is knocked down separately. Results may lead to novel Fisher Scientific). The nontargeting siCONTROL#2 and the PLK1 biomarkers that predict which HNSCC patients will benefit from targeting siRNA SMARTpool were used as negative and positive cisplatin-containing protocols. control for transfection efficiency, respectively. After 24 hours of incubation with the transfection reagents, medium volume was m Materials and Methods reduced to 40 L using the Microlab STAR liquid handling station (Hamilton Robotics), and 60 mL fresh medium was added with or Cell lines, clinical specimens, and chemicals without cisplatin to reach final concentrations matching IC0 (no All cancer cell lines were cultured in DMEM (Lonza) containing cisplatin) or IC20 (2 mmol/L) using a Multidrop Combi (Thermo 5% FCS (Lonza) and 2 mmol/L L-glutamine (Lonza). Cells were Fisher Scientific). Plates were incubated for 72 hours at 37C/5% fi grown in a humidi ed atmosphere of 5% CO2 at 37 C. CO2. Cell viability was measured using CellTiter Blue (Promega) HNSCC cell lines UM-SCC-11B (a larynx tumor; T2N2A) and as described. UM-SCC-22B (a hypopharynx tumor; T2N1) were obtained from Deconvolution of siRNA SMARTpools that significantly caused Dr. T. Carey (University of Michigan, Ann Arbor, MI) in the year cellular sensitivity to cisplatin was performed using the same 1984 (46). Cell line VU-SCC-120 (formerly known as 93VU120) procedures, either by hand or using the same automated methods was established in 1993, as described previously (47). All HNSCC as described above. Deconvolution of the same siRNA SMART- cell lines were authenticated to the earliest available passages by pools was investigated in two additional HNSCC cell lines, UM- microsatellite profiling and TP53 mutation analysis (46, 48). SCC-11B and UM-SCC-22B. These were chosen based on their Clinical specimens of a panel of 22 HNSCCs and corresponding sensitivity to cisplatin, having IC20 values that are two times lower macroscopically normal mucosa, adjacent to the tumor, were and four times lower than for VU-SCC-120, respectively. In collected according to the Dutch legislations on research with addition, both cell lines could be reproducibly transfected using human material. This study was approved by the Institutional the robotic platform described previously. Review Board VU University Medical Center (VUmc; Amsterdam, UM-SCC-22B was transfected with 25 nmol siRNA and 0.10 mL the Netherlands). Collection of specimens was done immediately DharmaFECT1. UM-SCC-11B was transfected with 25 nmol siRNA – after surgery, prior to radiotherapy, in the period 2000 2007. and 0.065mL DharmaFECT1.IC20 cisplatin valuesfor UM-SCC-11B Samples were snap frozen and stored in liquid nitrogen. Inclusion and UM-SCC-22B were 1.0 and 0.05 mmol/L, respectively. criteria were squamous cell carcinoma from the oral cavity or oropharynx, and the treatment consisted of surgery with or Data analysis of the genome-wide screen without postoperative radiotherapy. All normal mucosa samples The data from our genome-wide high-throughput screen were were judged free from dysplasia by an experienced pathologist. analyzed with two different statistical methods; Z-score calcula- The panel included 13 males and 9 females with an average age of tion and a Limma-based linear regression model. 60.0 10.1 years (range 33.8–82.5). The majority of the patients For the Z-score analysis, raw fluorescence values were log10 had a history of heavy smoking (30 pack-years; 17/22, 77%) transformed and per plate the trimmed mean was calculated and/or heavy alcohol consumption (100 unit-years, with 1 unit (trimming factor 0.5). Plate normalization was accomplished is equivalent to approximately 15 mL of ethanol; 15/22, 68%). using the overall mean per cisplatin condition (IC0 or IC20) and Specimens were obtained from the oral cavity (17/22, 77%) and per replicate. The normalized fluorescent values were back-trans- the oropharynx (5/22, 23%) with the majority showing a mod- formed. Per cisplatin condition, both replicates were normalized erate degree of differentiation (15/22, 68%), and the remaining toward each other and for each siRNA replicates of each cisplatin seven cases were poorly differentiated. All orophayngeal tumors condition were averaged. IC0/IC20 ratios were calculated and used were tested as human papillomavirus negative. Twenty-three for Z-score calculations. percent of the patients had early disease stage and 77% advanced For the calculations in Limma, the data were read into R stage. Cisplatin was obtained from Teva Pharmachemie in a (version 2.12) and configured using cellHTS2 (50). Subsequently, concentration of 1 mg/mL. the data were log2 transformed and normalized by correcting for linear effects of "day" and "plate" using a regression model. IC50 measurements To estimate the treatment effect, an empirical Bayes linear model Cells were plated in 96-well plates and were treated with 18 accounting for the treatment level (51) was fitted to the normal- different concentrations of cisplatin (ranging from 0.01 to 666 ized data using the BioConductor package Limma (52). The mmol/L) 24 hours later. Plates were incubated for 72 hours at resulting P value list was subsequently corrected for multiple

www.aacrjournals.org Mol Cancer Ther; 16(3) March 2017 541

Martens-de Kemp et al.

testing using Benjamini–Hochberg's step-up FDR method (53). determined using a mRNA microarray (4 44 K Whole Human This approach yields more power to find associations, in partic- Genome Arrays G4112F; Agilent). Labeling, hybridization, and ular in studies with a small sample size as the current one (54). scanning was performed according to the manufacturer's proto- col, using a G2505B scanner and Feature Extraction v9.5 (Agilent). Deconvolution of siRNA pools Data analysis was performed as described previously (57) and can Gene-specific knockdown was confirmed by investigation of be downloaded using GEO accession number GSE83519. the phenotype caused by each of the four separate siRNAs that make up the siRNA pool. Triplicate experiments were performed Results as described above and the median of the three IC measurements 0 Genome-wide siRNA screen design as well as the median of the three IC20 measurements were calculated. With these medians, P values against siCONTROL#2 Nineteen HNSCC cell lines were subjected to a selection pro- measurements were obtained with the Student t test. We previ- cedure to identify the cell line most suitable for genome-wide ously analyzed over 1,000 siCONTROL#2 measurements and siRNA application (data not shown). Selection criteria were based observed that the data follow the normal distribution (Supple- on cisplatin sensitivity, growth rate at low density in 96-well fi mentary Fig. S1), allowing t test analysis. plates, limited acidi cation of the medium to allow CellTiter-Blue assays, and reproducibility of siRNA transfection, also on a robotic platform. Cell line VU-SCC-120, a cell line derived from Quantitative real-time PCR a previously untreated moderately differentiated, advanced stage Cells were seeded in 96-well plates and transfected as described tongue cancer (T3N1), matched all criteria. The half-maximal above. RNA was extracted 96 hours post-transfection using the inhibitory concentration (IC50) of cisplatin was determined at 4.0 RNeasy Micro Kit (Qiagen). cDNA was synthesized from 250 ng of mmol/L for this cell line (58). Because we are interested in siRNAs RNA using AMV reverse transcriptase (Promega) and a custom that increase the response to cisplatin, we decided to screen the fi designed reverse primer (Supplementary Table S1). Ampli cation siRNA library in the presence of a cisplatin concentration that of the cDNA was performed on the ABI/Prism 7500 Sequence affected cell survival by 20% (IC ), resulting in a theoretical 20 Detector System (TaqMan-PCR; Applied Biosystems) with 1 screening window of 80%. For VU-SCC-120, this meant that the Power SYBR Green PCR Master Mix (Applied Biosystems) and cells were treated with 2.0 mmol/L of cisplatin (Fig. 1A). custom designed primers for each gene of interest (Supplementary We used these data to design a robust, high-throughput RNA CTR1 Table S1). expression after siRNA transfection was investi- interference screen for the VU-SCC-120 cell line. Cells were trans- gated using TaqMan Gene Expression Assay (Hs00977268_g1; fected in quadruplicate in separate 96-well plates for duplicate Applied Biosystems) according to the manufacturer's instructions. analysis of IC0 (no cisplatin) and IC20 (Fig. 1B). We used the PLK1 For each sample, the cycle number at which the amount of siRNA SMARTpool, which resulted in a decrease in cell viability fi C ampli ed target crossed a predetermined threshold (the t value) (data not shown), as expected (59), as a positive control for was determined. To correct for differences in RNA input and transfection efficiency. The nontargeting siCONTROL#2 was used b GUSB quality, -glucuronidase ( ) was used as a housekeeping as negative control. Cell viability was measured in every plate to gene (55) for each RNA sample. The mRNA expression was evaluate the effect of each siRNA on cell growth (IC0) and the effect calculated relative to that of GUSB using the DDCt method of reduced gene expression on cisplatin sensitivity (IC0 vs. IC20). (56). Quantitative real-time (RT)-PCR reactions without reverse transcriptase were carried out in parallel for each RNA sample to Genome-wide siRNA screen to identify cisplatin susceptibility exclude signal from contaminating genomic DNA. genes Raw cell viability values were inspected for transfection- Cell-cycle analysis induced cell death (siCONTROL#2 vs. untransfected cells; 16% 2 fl Cells were seeded in 25-cm asks and transfected 24 hours cell death) and transfection efficiency (siPLK1 vs. siCONTROL#2; BRCA2 0 later in duplicate with siCONTROL#2, SMARTpool, or 92% cell death), and were found to be acceptable (Z factor ¼ SHFM1 SMARTpool. After another 24 hours, the cultures were 0.570). Comparison of the two independent duplicate screens refreshed with either fresh medium or medium containing 500 showed high reproducibility with Spearman r ¼ 0.823. Subse- nmol/L cisplatin. After 72 hours of incubation, the cells were quently, the data were analyzed using two independent statistical harvested and fixed with 70% EtOH during at least 3 hours. Next, Z methods as described in the Materials and Methods; -score cells were incubated with 0.5 mg/mL RNAse A in PBS at 37 C for analysis and a Limma-based linear regression model. Z-score 30 minutes. The cells were washed and the DNA content was calculations yielded 205 siRNA pools that significantly sensitized stained with propidium iodide (PI). The cell-cycle distribution VU-SCC-120 cells to cisplatin (Z 3; P < 0.001; Fig. 1C), fl was analyzed with a BD FACSCalibur ow cytometer (BD Bios- whereas the linear regression model resulted in 168 hits (raw ciences). Cell-cycle analyses were performed using BD CellQuest P < 0.05). The hit lists obtained with these two methods shared software (BD Biosciences) and calculations in Microsoft Excel. 102 siRNAs (Fig. 1D; Supplementary Table S2). We subsequently selected siRNAs for stratification by decon- Expression microarray volution of the siRNA pools into four separate siRNAs. From the RNA was isolated with TRIzol (Life Technologies) according to 102 siRNA pools that were found to cause a significant sensiti- the manufacturer's instructions. RNA integrity was measured zation to cisplatin by both statistical analyses, we were able to using a RNA nanochip on an Agilent 2100 Bioanalyzer (Agilent order the individual siRNAs of 97 pools. The target sequences of Technologies). Synthesis and labeling of cDNA was performed the remaining five siRNA pools were no longer annotated as according to the recommendations of the manufacturer (Agilent), potential genes. In addition, we randomly selected 28 siRNAs including quality controls. The mRNA expression profiles were that were only identified by Z-score analysis and 23 that were only

542 Mol Cancer Ther; 16(3) March 2017 Molecular Cancer Therapeutics

The FA/BRCA Pathway Predicts Cisplatin Response in HNSCC

AB Day 1 Day 2 Day 3 Day 6 Seed cells Transfection Drug treatment Read out

100 4× 4× × 2 Mock (IC0)

IC20

IC50

Surviving population (%) Surviving population × 0 2 Cisplatin (IC20)

10 -2 10 -1 10 0 10 1 10 2 10 3 Sensitivity

Cisplatin (μmol/L)

C 6 D

3 Linear regression 0 Z-score analysis 0 5000 10000 15000 20000 103 102 66 model (Z≥-3) −3 siRNA pool (raw P < 0.05) siCONTROL#2

Z-score −6

−9

−12

−15

Figure 1. Functional genetic screen for cisplatin-sensitizing siRNAs. A, HNSCC cell line VU-SCC-120 was treated with 18 different concentrations of cisplatin to determine the IC50 (4.0 mmol/L) and IC20 (2.0 mmol/L) values. B, Design of high-throughput siRNA screen with VU-SCC-120 cells. White wells contain untransfected control cells, gray wells contain transfection controls (siCONTROL#2 and siPLK1), and black wells contain cells transfected with siRNAs from the genome-wide library. C, Graphic representation of the Z-score distribution. Black dots represent siRNAs from the genome-wide screen. The cut-off for hit list preparation was set at Z 3(P < 0.001). None of the nontargeting siCONTROL#2 siRNAs (gray dots) reached this threshold. D, Schematic representation of the number of siRNAs that were found to signiﬁcantly sensitize VU-SCC-120 cells to cisplatin, either by Z-score calculation or a Limma-based linear regression model. Comparison of the two hit lists showed that 102 siRNAs were identiﬁed by both statistical methods.

found by the linear regression model for further analysis. Decon- (P ¼ 2.13 10 4; Fig. 2A). In addition, according to the STRING volution was done in VU-SCC-120, the cell line used for the database (version 9.05), the only cluster that could be found genome-wide siRNA discovery screen, and two additional among the siRNAs that were confirmed in all three HNSCC cell HNSCC cell lines, UM-SCC-11B and UM-SCC-22B. lines (n ¼ 21; Table 1), harbored these DNA repair genes (Fig. 2B). We marked an siRNA as "confirmed" in case at least two This indicates that an important determinant for cisplatin sensi- separate siRNAs significantly (P < 0.05, Student t test against tivity in HNSCC cells is the proper functioning of particularly the siCONTROL#2 transfections) sensitized the tumor cells to cis- FA/BRCA pathway. platin (Supplementary Table S3). We could confirm 21% of the Indeed, close examination of two important players in the hits identified by only the Z-score analysis and 17% of the hits FA/BRCA pathway, BRCA1 and BRCA2, showed that knockdown identified by the linear regression model only. However, the of these genes enhances cisplatin-induced cell death (Fig. 2C and confirmation rate obtained by siRNAs that were identified by Supplementary Fig. S2A). Also the split hand/foot malformation both methods proved to be superior (42% confirmed) over either type 1 gene (SHFM1 or DSS1), which has been shown to phys- method alone. This observation also holds true for the numbers of ically interact with BRCA2 (60–63), could be confirmed as a siRNAs that were confirmed in at least two cell lines or even in all determinant of cisplatin response with all four separate siRNAs three HNSCC cell lines. In total, 51 of the tested 148 primary hits (Fig. 2C), and this was confirmed in multiple cell lines (Supple- were confirmed in VU-SCC-120 cells, of which 21 were repro- mentary Fig. S2B). Knockdown of mRNA expression by the ducible in all three HNSSC cell lines (Table 1). separate SHFM1 siRNAs was determined by quantitative RT-PCR and showed to correlate with the increase in cisplatin-induced cell The FA/BRCA pathway death (Supplementary Fig. S2A), demonstrating that these are Comprehensive pathway analysis on the complete data set by specific effects caused by knockdown of the SHFM1 gene and not IPA (Ingenuity systems) revealed that particularly genes in the due to off-target effects. Fanconi anemia/BRCA pathway, which is involved in DNA cross- Because SHFM1 is known to interact with BRCA2, we ques- link repair, are implicated in cisplatin response in HNSCC cells tioned whether knockdown of SHFM1 would cause a cell-cycle

www.aacrjournals.org Mol Cancer Ther; 16(3) March 2017 543

Martens-de Kemp et al.

Table 1. List of siRNAs that were confirmed to sensitize VU-SCC-120 (and additional HNSCC cell lines) to cisplatin with at least two separate siRNAs Gene symbol Origin Confirmed in Accession number Gene name ALKBH2 Overlap 3 HNSCC cell lines NM_001145374 AlkB, alkylation repair homolog 2 (E. coli) APP Overlap VU-SCC-120 NM_000484 Amyloid beta (A4) precursor protein (Alzheimer disease) ATR Overlap 3 HNSCC cell lines NM_001184 Ataxia telangiectasia and Rad3 related BCL11A Overlap VU-SCC-120 NM_138559 B-cell CLL/lymphoma 11A (zinc finger protein) BRCA1 Overlap 3 HNSCC cell lines NM_007294 Breast cancer 1, early onset BRCA2 Overlap 3 HNSCC cell lines NM_000059 Breast cancer 2, early onset BRUNOL6 Overlap VU-SCC-120 NM_052840 Bruno-like 6, RNA binding protein (Drosophila) C17ORF35 Overlap VU-SCC-120 NM_003876 Chromosome 17 open reading frame 35 CAMK2B Overlap 3 HNSCC cell lines NM_001220 Calcium/calmodulin-dependent protein kinase II beta CDK19 Overlap 3 HNSCC cell lines NM_015076 Cell division cycle 2-like 6 (CDK8-like) CENPT Lin. reg. model VU-SCC-120 NM_025082 Centromere Protein T CHRNA6 Overlap VU-SCC-120 NM_004198 Cholinergic receptor, nicotinic, alpha polypeptide 6 CHUK Overlap VU-SCC-120 NM_001278 Conserved helix-loop-helix ubiquitous kinase CLK3 Overlap 3 HNSCC cell lines NM_001292 CDC-like kinase 3 COASY Overlap VU-SCC-120 NM_025233 Coenzyme A synthase DNAH3 Overlap VU-SCC-120 NM_017539 Dynein, axonemal, heavy polypeptide 3 DSCR1 Z-score analysis VU-SCC-120 NM_004414 Down syndrome critical region gene 1 ETNK2 Overlap VU-SCC-120 NM_018208 Ethanolamine kinase 2 ETV3L Z-score analysis VU-SCC-120 NM_001004341 Ets Variant 3-Like FANCM Z-score analysis 3 HNSCC cell lines NM_020937 Fanconi anemia, complementation group M FOXL1 Overlap VU-SCC-120 NM_005250 Forkhead box L1 ING1L Overlap 3 HNSCC cell lines NM_001564 Inhibitor of growth family, member 2 KCNJ2 Overlap 3 HNSCC cell lines NM_000891 Potassium inwardly-rectifying channel, subfamily J, member 2 LCE1B Overlap 3 HNSCC cell lines NM_178349 Late cornified envelope 1B LOC115704 Lin. reg. model VU-SCC-120 NM_145245 MARK3 Overlap 3 HNSCC cell lines NM_002376 MAP/microtubule affinity-regulating kinase 3 MTNR1B Overlap VU-SCC-120 NM_005959 Melatonin receptor 1B MTX1 Overlap 3 HNSCC cell lines NM_002455 Metaxin 1 NR1H2 Overlap VU-SCC-120 NM_007121 Nuclear receptor subfamily 1, group H, member 2 NUP153 Z-score analysis VU-SCC-120 NM_005124 Nucleoporin 153 kDa OR1D2 Overlap VU-SCC-120 NM_002548 Olfactory receptor, family 1, subfamily D, member 2 PALB2 Overlap VU-SCC-120 NM_024675 Partner and localizer of BRCA2 PRDM13 Overlap VU-SCC-120 NM_021620 PR domain containing 13 PTAFR Overlap 3 HNSCC cell lines NM_000952 Platelet-activating factor receptor PTMA Overlap 3 HNSCC cell lines NM_002823 Prothymosin, alpha (gene sequence 28) RBMS3 Overlap VU-SCC-120 NM_014483 RNA binding motif, single stranded interacting protein REV3L Overlap 3 HNSCC cell lines NM_002912 REV3-like, catalytic subunit of DNA polymerase zeta (yeast) RHAG Overlap VU-SCC-120 NM_000324 Rhesus blood group-associated glycoprotein RNF26 Overlap VU-SCC-120 NM_032015 Ring finger protein 26 SHFM1 Z-score analysis VU-SCC-120 NM_006304 Split hand/foot malformation (ectrodactyly) type 1 SLC25A27 Overlap 3 HNSCC cell lines NM_004277 Solute carrier family 25, member 27 SLC44A1 Overlap VU-SCC-120 NM_080546 Solute carrier family 44, member 1 SSUH2 Lin. reg. model 3 HNSCC cell lines NM_001256748 Ssu-2 Homolog (C. elegans) TADA3L Overlap 3 HNSCC cell lines NM_006354 Transcriptional adaptor 3 (NGG1 homolog, yeast)-like TF Overlap 3 HNSCC cell lines NM_001063 Transferrin TFF1 Z-score analysis 3 HNSCC cell lines NM_003225 Trefoil factor 1 (breast cancer estrogen-inducible protein) TNFRSF13B Overlap VU-SCC-120 NM_012452 Tumor necrosis factor receptor superfamily, member 13B TNK1 Overlap VU-SCC-120 NM_003985 Tyrosine kinase, nonreceptor, 1 TOP2A Lin. reg. model VU-SCC-120 NM_001067 Topoisomerase (DNA) II Alpha 170 kDa TREM2 Overlap VU-SCC-120 NM_018965 Triggering receptor expressed on myeloid cells 2 TTBK2 Overlap VU-SCC-120 NM_173500 Tau tubulin kinase 2

arrest in G2-M after cisplatin treatment, as is the case after BRCA2 in each phase of the cell cycle was significantly different after knockdown. We compared cell-cycle distributions among cell cisplatin incubation (P < 0.02). This indicates that SHFM1 is a cultures treated with or without cisplatin (Fig. 3A). We applied critical component in the FA/BRCA pathway, and reduced expres- 500 nmol/L cisplatin, the highest concentration that does not sion causes increased cisplatin sensitivity and G2–M arrest. show an effect on cell survival (Fig. 1A). Indeed, this concentra- To investigate whether BRCA1, BRCA2, and SHFM1, identified tion did not result in cell-cycle arrest. BRCA2 knockdown (no as determinants of cisplatin sensitivity by our genome-wide cisplatin added) caused a significant shift toward a G2–M arrest in siRNA screen, could be suitable candidates to serve as markers the cell-cycle distribution (P < 0.05, two-sided Fisher exact test). that predict whether cisplatin-containing chemoradiation is suc- This further increased after cisplatin treatment (P < 0.001). For the cessful in HNSCC patients, we mined microarray mRNA expres- SHFM1 knockdown cells, we found a small, but not significant sion profiles made from 22 normal mucosa/tumor pairs for these shift in cell-cycle distribution in the cultures that were not treated genes. We found that the expression of BRCA1, BRCA2, and with cisplatin (P ¼ 0.13, two-sided Fisher exact test in comparison SHFM1 was significantly upregulated (P < 0.001) in HNSCC cells with siCONTROL#2-transfected cells), but the proportion of cells compared with the paired normal mucosa (1.94x 0.60, 1.27x

544 Mol Cancer Ther; 16(3) March 2017 Molecular Cancer Therapeutics

The FA/BRCA Pathway Predicts Cisplatin Response in HNSCC

Figure 2. Functional relationship of the cisplatin-sensitizing siRNA targets. A, Comprehensive pathway analysis of all genes present in our genome-wide siRNA library and the associated Z-scores by IPA (Ingenuity Pathway Analysis) showed that many siRNAs that target members of the FA/BRCA pathway significantly enhance the cisplatin response in VU-SCC-120. A darker gray color is associated with a lower Z-score. B, String analysis of siRNAs that were confirmed to increase cisplatin sensitivity in all three HNSCC cell lines (n ¼ 21) showed that the only cluster that could be identified, harbors DNA repair genes. C, The siRNA SMARTpools targeting BRCA1 (left), BRCA2 (middle), and SHFM1 (right) were deconvoluted in VU-SCC-120 cells. Dark gray bars represent IC0-treated cells (no cisplatin), lighter bars correspond to IC20-treated cells. Bars represent cell viability as compared with the IC0 value of each transfection; error bars, SD. , P < 0.05 compared with siCONTROL#2 IC20 value of the same gene (Student t test).

0.40, and 1.62x 0.69 fold, respectively; Fig. 3B). In addition, the lation with the overall survival of the patients (P ¼ 0.039). range of gene expression varies tremendously among the different Next, we fitted a multivariable cox model with ridge regular- tumor samples, leaving the possibility open that variation in gene ization using package glmnet (66) and R version 3.1.3. The expression might determine the variation in clinical response to accuracy of this model was assessed with a leave one out cross cisplatin. validation. Across the first 3 years after treatment, this gave an To investigate this possibility, we analyzed the publicly integrated area under the curve (iAUC; ref. 67) of 0.60. Indi- available gene expression dataset of Wichmann and colleagues vidual HRs resulting from the cross-validation where divided in (64), which contains informationontheexpressionprofiles of three equal groups (Supplementary Table S5) and the survival 290 HNSCC tumors and progression-free (PFS) and overall of each of these groups was visualized by Kaplan–Meier anal- survival (OS). All tumors were treated with a range of multi- ysis (Fig. 4). Together, these results indicate that the 84-gene set modal therapies with curative intent. We first identified all identified in our cisplatin-sensitivity screen is correlated with primary tumors that did not contain HPV DNA and/or RNA overall survival and confirmed the importance of our in vitro (n ¼ 121), as this is a major prognostic factor. Next, we selected findings in an independent study in a patient cohort. all probes targeting the 102 genes that sensitize HNSCC cells for cisplatin as identified by the two different statistical methods used for analysis of the genome-wide siRNA screen (Fig. 1D). Discussion This left us with 127 probes targeting 84 different genes Resistance to platinum-containing chemoradiation may be a (Supplementary Table S4), and we analyzed whether the limiting factor in the successful treatment of advanced stage expression of these genes (that sensitize for cisplatin treatment) HNSCC. It would be of significant clinical benefittounder- is significantly associated with patient survival. A (cox) global stand the underlying factors that cause response, because this test (65) indicated that our gene set shows a significant corre- may reveal biomarkers for response prediction and pave the

www.aacrjournals.org Mol Cancer Ther; 16(3) March 2017 545

Martens-de Kemp et al.

A B BRCA1 siCONTROL#2 BRCA2 SHFM1 0.8 * G : 79.2% G : 66.1% G : 66.3% 1 1 1 S: 10.1% S: 12.8% S: 15.9% G −M: 9.9% G −M: 19.9% G −M: 16.3% 2 2 2 0.6 0 nmol/L

0.4 Relative expression

0.2

G : 75.1% G : 47.8% G : 56.4% Mucosa Tumor 1 1 1 S: 11.0% S: 20.9% S: 16.2% BRCA2 G −M: 12.9% G −M: 28.9% G −M: 25.9% 2 2 2 1.4 500 nmol/L * 1.2

1.0

0.8

* 0.6 ** Relative expression * 0.4 100% Mucosa Tumor SHFM1 4.0 75% * G −M 2 3.0 S 50% G1 2.0

25% 1.0 Relative expression

0% 0.0 siCON 0 nmol/L siCON 500 nmol/L BRCA2 0 nmol/L BRCA2 500 nmol/L SHFM1 0 nmol/L SHFM1 500 nmol/L Mucosa Tumor

Figure 3. Functional relationship of the cisplatin-sensitizing siRNA targets. A, Cell-cycle distribution of VU-SCC-120 cells transfected with siCONTROL#2, BRCA2,or SHFM1-targeting siRNA SMARTpools. Cultures were not treated with cisplatin (0 nmol/L) or treated with 500 nmol/L cisplatin during 72 hours. Percentages of cells in designated cell-cycle phases are given and are schematically represented in the bottom panel (, P < 0.02; , P < 0.001). Error bars, SD. B, Microarray gene expression data from 22 patient-derived pairs of tumor tissue and healthy mucosa were examined for BRCA1 (top), BRCA2 (middle), and SHFM1 (bottom) expression. Thick lines within the boxplots represent the median values. The expression of all three genes was signiﬁcantly higher in the tumor samples than in the normal mucosa ( raw P < 0.001, moderated t test implemented in Limma).

way to a more personalized treatment. In this study, we iden- therapy would succeed, for example by screening for gene tified determinants of response to cisplatin in an unbiased loss- expression or mutations. However, mutations in these particu- of-function screen. lar genes are uncommon in HNSCC (COSMIC database; cancer. In a previous publication, we described that DNA-bound sanger.ac.uk/cancergenome/projects/cosmic). This highlights platinum is the most important determinant of cisplatin sen- the necessity for the identification of biomarkers that might sitivity in HNSCC cell lines (58), and not intracellular platinum disclose whether the FA/BRCA pathway and/or the involved levels. These data already pointed in the direction that the genes are at high activity or not. Such biomarkers could be the activity of the DNA cross-link repair machinery might be the expression of the genes themselves as the first analysis of the critical factor in HNSCC that determines cisplatin response. In expression data suggests. A functional DNA crosslinking assay or this study, we identified the genes involved in repair of DNA imaging of radioactive cisplatin in tumor DNA might be the crosslinks, such as BRCA1, BRCA2,andATR, as most important biomarker of choice to personalize treatment of patients that genes that modulate cisplatin response. Patients suffering from will not benefit from cisplatin-containing chemoradiation pro- FA/BRCA pathway deficiency are predisposed for squamous cell tocols. Alternatively, the combination of chemoradiation and cancer most particularly in the head and neck likely by the toxic an agent that inhibits DNA crosslink repair might enhance effect of aldehydes as natural cross-linkers (68). The same efficacy of cisplatin and lead to higher chemoradiation response pathway apparently determines response to DNA crosslinkers rates, although also toxicity will likely increase. used as cytotoxic drugs. Examination of the activity level of the We identifiedSHFM1asanimportantplayerintheresponse FA/BRCA pathway prior to cisplatin-containing treatment is to cisplatin treatment. SHFM1 is known to bind to BRCA2 an option to predict upfront whether addition of cisplatin to (61, 62) and to stabilize this protein (60), although this could

546 Mol Cancer Ther; 16(3) March 2017 Molecular Cancer Therapeutics

The FA/BRCA Pathway Predicts Cisplatin Response in HNSCC

carcinoma of the kidney (75). Furthermore, it might have occurred that compensatory mechanisms abolishe the effect of knockdown expression of a particular gene (76, 77). For example, the functions of AKT1, AKT2, and AKT3 seem partly redundant, implying that knockdown of either one of these genes might not result in a phenotype as the other two genes are able to take over some of the functions of the silenced gene. Finally, we identified novel genes that may influence cisplatin response in HNSCC cells, and that have not been described before. An example is KCNJ2, a potassium inwardly-rectifying channel, that has a small but significant effect on cisplatin sensitivity and we were able confirm this in all three cell lines used (Supplementary Fig. S4A). Also the blocking of the channeling function of KCNJ2, by barium chloride (78–80), showed that cell cultures are significantly sensitized to cisplatin (Supplementary Fig. S4B). However, the observed effect is small and not compa- rable with that of knockdown of, for example, BRCA2. We investigated the importance of the genes we identified as cisplatin response determinants in an independent, publicly available expression dataset based on a cohort of HNSCC patients Figure 4. treated with multimodality treatment (64). We found that the Patient survival is correlated with expression of cisplatin-sensitizing genes. expression of genes in our gene set correlates with patient survival The Kaplan–Meier curves are based on the cross-validated HRs calculated with (integrated AUC ¼ 0.60) and that higher expression of the genes the multivariable cox model using the expression of 84 genes that were fi result in worse patient survival. This indicates that the gene set we identi ed in our genome-wide siRNA screen. The patient cohort was divided in fi in vitro fl three equal groups based on these HRs. Survival curves indicate that patients identi ed to in uence cisplatin sensitivity also contributes with higher (cross-validated) HR have a worse overall survival. in vivo to HNSCC treatment outcome. Larger studies of more homogenous patient series will be required to validate these findings, and to establish that some of these genes might be not be confirmed in every study (63). SHFM1 has been previ- exploited as biomarker. ously found to sensitize HeLa cells to cisplatin (35). In that Taken together, our data show that the application of an study, the authors also focused on members of the FA/BRCA unbiased genome-wide functional genetic screen leads to the pathway, which indicates that this pathway is not only impor- identification of unexpected and novel determinants of cisplatin tant in the cisplatin response in SCCs from the head and neck, response, which might be exploited as biomarkers for clinical but also in other squamous cell carcinomas. In addition, a prediction of chemoradiation response. Our data will redirect recent study in ovarian cancer showed that SHFM1 has a role in clinical research towards reliable response predictors for plati- cisplatin resistance (69) and that SHFM1 levels are elevated in num-containing compounds applied in chemoradiation and tumor cells as compared with the nontumorigenic neighboring systemic therapy. cells. This observation serves as a platform to use SHFM1 expression as a biomarker for ovarian cancer. Also in breast Disclosure of Potential Conflicts of Interest SHFM1 cancer, high expression was found to correlate to poor No potential conflicts of interest were disclosed. prognosis and poor relapse-free survival (70). We found that the SHFM1 expression levels are overall elevated in HNSCC as Authors' Contributions compared with healthy mucosa (Fig. 3B), and that the level of Conception and design: S.R. Martens-de Kemp, B.J.M. Braakhuis, expression varies tremendously. R.H. Brakenhoff We noticed that several genes that are known to influence the Development of methodology: S.R. Martens-de Kemp, I.H. van der Meulen, cisplatin response were not found in our genome-wide siRNA V.W. van Beusechem, R.H. Brakenhoff screen. For some of these genes, it might have occurred that the Acquisition of data (provided animals, acquired and managed patients, siRNA-induced gene knockdown was insufficient, as we deter- provided facilities, etc.): S.R. Martens-de Kemp, A. Brink, I.H. van der Meulen, V.W. van Beusechem mined for ATM (Supplementary Fig. S3A). However, the knock- CTR1 fi Analysis and interpretation of data (e.g., statistical analysis, biostatistics, down of was signi cant (Supplementary Fig. S3B), but we computational analysis): S.R. Martens-de Kemp, A. Brink, R.X. de Menezes, could not find clues for the involvement of this gene in cisplatin D.E. te Beest, C.R. Leemans, B.J.M. Braakhuis, R.H. Brakenhoff response (Supplementary Fig. S3C), as we could not before (58). Writing, review, and/or revision of the manuscript: S.R. Martens-de Kemp, This is surprising as many studies implicated an important role for I.H. van der Meulen, C.R. Leemans, V.W. van Beusechem, B.J.M. Braakhuis, CTR1 in cisplatin uptake (71–73). Because these studies were not R.H. Brakenhoff Administrative, technical, or material support (i.e., reporting or organizing performed in HNSCC, one might argue that other genes than CTR1 data, constructing databases): S.R. Martens-de Kemp, I.H. van der Meulen, are of greater importance in cisplatin response in HNSCC. R.H. Brakenhoff This may also be the reason why we identified some of the SLC Study supervision: B.J.M. Braakhuis, R.H. Brakenhoff family members that are previously implicated in cisplatin (SLC25A27 and SLC44A1; Table 1; Supplementary Table S2), Acknowledgments but not all, such as SLC7A11, which proved to be important in The authors would like to thank Dirk J. Kuik for assistance during the cisplatin resistance in bladder cancer (74) and collecting duct statistical analysis of the genome-wide siRNA screen. This study was performed

www.aacrjournals.org Mol Cancer Ther; 16(3) March 2017 547

Martens-de Kemp et al.

within the framework of CTMM, the Centre for Translational Molecular Med- ing the RNA Interference Functional Oncogenomics Laboratory (RIFOL) core icine AIRFORCE project. High-throughput screens were conducted at the RNA facility. Interference Functional Oncogenomics Laboratory (RIFOL) core facility at the The costs of publication of this article were defrayed in part by the VUmc Cancer Center Amsterdam. payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate Grant Support this fact. R.H. Brakenhoff received a grant from CTMM, the Centre for Translational Molecular Medicine (AIRFORCE project, grant 03O-103). V.W. van Beuse- Received July 18, 2016; revised November 18, 2016; accepted December 6, chem received ﬁnancial support from the Stichting VUmc CCA for establish- 2016; published OnlineFirst December 15, 2016.

References 1. Ferlay J, Shin HR, Bray F, Forman D, Mathers C, Parkin DM. Estimates of resistance to chemoradiotherapy and poor outcome. Int J Radiat Oncol worldwide burden of cancer in 2008: GLOBOCAN 2008. Int J Cancer Biol Phys 2001;50:27–36. 2010;127:2893–917. 17. Lavertu P, Adelstein DJ, Myles J, Secic M. P53 and Ki-67 as outcome 2. Leemans CR, Braakhuis BJ, Brakenhoff RH. The molecular biology of head predictors for advanced squamous cell cancers of the head and neck treated and neck cancer. Nat Rev Cancer 2011;11:9–22. with chemoradiotherapy. Laryngoscope 2001;111:1878–92. 3. Pignon JP, le MA, Maillard E, Bourhis J. Meta-analysis of chemotherapy in 18. Osman I, Sherman E, Singh B, Venkatraman E, Zelefsky M, Bosl G, et al. head and neck cancer (MACH-NC): an update on 93 randomised trials and Alteration of p53 pathway in squamous cell carcinoma of the head and 17,346 patients. Radiother Oncol 2009;92:4–14. neck: impact on treatment outcome in patients treated with larynx pres- 4. Chen SW, Yang SN, Liang JA, Lin FJ, Tsai MH. Prognostic impact of tumor ervation intent. J Clin Oncol 2002;20:2980–7. volume in patients with stage III-IVA hypopharyngeal cancer without bulky 19. van den Broek GB, Wildeman M, Rasch CR, Armstrong N, Schuuring E, lymph nodes treated with definitive concurrent chemoradiotherapy. Head Begg AC, et al. Molecular markers predict outcome in squamous cell Neck 2009;31:709–16. carcinoma of the head and neck after concomitant cisplatin-based che- 5. Doweck I, Denys D, Robbins KT. Tumor volume predicts outcome for moradiation. Int J Cancer 2009;124:2643–50. advanced head and neck cancer treated with targeted chemoradiotherapy. 20. Warnakulasuriya S, Jia C, Johnson N, Houghton J. p53 and P-glyco- Laryngoscope 2002;112:1742–9. protein expression are significant prognostic markers in advanced head 6. Hoebers FJ, Pameijer FA, de BJ, Heemsbergen W, Balm AJ, Schornagel JH, and neck cancer treated with chemo/radiotherapy. J Pathol 2000; et al. Prognostic value of primary tumor volume after concurrent chemor- 191:33–8. adiation with daily low-dose cisplatin for advanced-stage head and neck 21. Akervall J, Guo X, Qian CN, Schoumans J, Leeser B, Kort E, et al. Genetic and carcinoma. Head Neck 2008;30:1216–23. expression profiles of squamous cell carcinoma of the head and neck 7. Knegjens JL, Hauptmann M, Pameijer FA, Balm AJ, Hoebers FJ, de Bois JA, correlate with cisplatin sensitivity and resistance in cell lines and patients. et al. Tumor volume as prognostic factor in chemoradiation for advanced Clin Cancer Res 2004;10:8204–13. head and neck cancer. Head Neck 2010;33:375–82. 22. Dumur CI, Ladd AC, Wright HV, Penberthy LT, Wilkinson DS, Powers CN, 8. Kurek R, Kalogera-Fountzila A, Muskalla K, Dafni U, Schnabel T, Kober B, et al. Genes involved in radiation therapy response in head and neck et al. Usefulness of tumor volumetry as a prognostic factor of survival in cancers. Laryngoscope 2009;119:91–101. head and neck cancer. Strahlenther Onkol 2003;179:292–7. 23. Duong C, Greenawalt DM, Kowalczyk A, Ciavarella ML, Raskutti G, Murray 9. Studer G, Lutolf UM, El-Bassiouni M, Rousson V, Glanzmann C. Volu- WK, et al. Pretreatment gene expression profiles can be used to predict metric staging (VS) is superior to TNM and AJCC staging in predicting response to neoadjuvant chemoradiotherapy in esophageal cancer. Ann outcome of head and neck cancer treated with IMRT. Acta Oncol Surg Oncol 2007;14:3602–9. 2007;46:386–94. 24. Ganly I, Talbot S, Carlson D, Viale A, Maghami E, Osman I, et al. Iden- 10. Tsou YA, Hua JH, Lin MH, Tsai MH. Analysis of prognostic factors of tification of angiogenesis/metastases genes predicting chemoradiotherapy chemoradiation therapy for advanced hypopharyngeal cancer–does tumor response in patients with laryngopharyngeal carcinoma. J Clin Oncol volume correlate with central necrosis and tumor pathology? ORL 2007;25:1369–76. J Otorhinolaryngol Relat Spec 2006;68:206–12. 25. Hanna E, Shrieve DC, Ratanatharathorn V, Xia X, Breau R, Suen J, et al. A 11. van den Broek GB, Rasch CR, Pameijer FA, Peter E, van den Brekel MW, Tan novel alternative approach for prediction of radiation response of squa- IB, et al. Pretreatment probability model for predicting outcome after mous cell carcinoma of head and neck. Cancer Res 2001;61:2376–80. intraarterial chemoradiation for advanced head and neck carcinoma. 26. Maher SG, Gillham CM, Duggan SP, Smyth PC, Miller N, Muldoon C, et al. Cancer 2004;101:1809–17. Gene expression analysis of diagnostic biopsies predicts pathological 12. de Jong MC, Pramana J, Knegjens JL, Balm AJ, van den Brekel MW, response to neoadjuvant chemoradiotherapy of esophageal cancer. Ann Hauptmann M, et al. HPV and high-risk gene expression profiles predict Surg 2009;250:729–37. response to chemoradiotherapy in head and neck cancer, independent of 27. Pramana J, van den Brekel MW, van Velthuysen ML, Wessels LF, Nuyten DS, clinical factors. Radiother Oncol 2010;95:365–70. Hofland I, et al. Gene expression profiling to predict outcome after 13. Handra-Luca A, Hernandez J, Mountzios G, Taranchon E, Lacau-St-Guily J, chemoradiation in head and neck cancer. Int J Radiat Oncol Biol Phys Soria JC, et al. Excision repair cross complementation group 1 immuno- 2007;69:1544–52. histochemical expression predicts objective response and cancer-specific 28. Schauer M, Janssen KP, Rimkus C, Raggi M, Feith M, Friess H, et al. survival in patients treated by cisplatin-based induction chemotherapy for Microarray-based response prediction in esophageal adenocarcinoma. locally advanced head and neck squamous cell carcinoma. Clin Cancer Res Clin Cancer Res 2010;16:330–7. 2007;13:3855–9. 29. Bao S, Wu Q, McLendon RE, Hao Y, Shi Q, Hjelmeland AB, et al. Glioma 14. Homma A, Furuta Y, Oridate N, Nakano Y, Kohashi G, Yagi K, et al. stem cells promote radioresistance by preferential activation of the DNA Prognostic significance of clinical parameters and biological markers in damage response. Nature 2006;444:756–60. patients with squamous cell carcinoma of the head and neck treated with 30. Berns K, Hijmans EM, Mullenders J, Brummelkamp TR, Velds A, Heimerikx concurrent chemoradiotherapy. Clin Cancer Res 1999;5:801–6. M, et al. A large-scale RNAi screen in human cells identifies new compo- 15. Jun HJ, Ahn MJ, Kim HS, Yi SY, Han J, Lee SK, et al. ERCC1 expression as a nents of the p53 pathway. Nature 2004;428:431–7. predictive marker of squamous cell carcinoma of the head and neck treated 31. Gobeil S, Zhu X, Doillon CJ, Green MR. A genome-wide shRNA screen with cisplatin-based concurrent chemoradiation. Br J Cancer 2008; identifies GAS1 as a novel melanoma metastasis suppressor gene. Genes 99:167–72. Dev 2008;22:2932–40. 16. Koukourakis MI, Giatromanolaki A, Kakolyris S, Sivridis E, Georgoulias V, 32. Jian R, Cheng X, Jiang J, Deng S, Hu F, Zhang J. A cDNA-based random RNA Funtzilas G, et al. Nuclear expression of human apurinic/apyrimidinic interference library for functional genetic screens in embryonic stem cells. endonuclease (HAP1/Ref-1) in head-and-neck cancer is associated with Stem Cells 2007;25:1904–12.

548 Mol Cancer Ther; 16(3) March 2017 Molecular Cancer Therapeutics

The FA/BRCA Pathway Predicts Cisplatin Response in HNSCC

33. Krishnan MN, Ng A, Sukumaran B, Gilfoy FD, Uchil PD, Sultana H, et al. 55. Aerts JL, Gonzales MI, Topalian SL. Selection of appropriate control genes RNA interference screen for human genes associated with West Nile virus to assess expression of tumor antigens using real-time RT-PCR. Biotechni- infection. Nature 2008;455:242–5. ques 2004;36:84–1. 34. Arora S, Bisanz KM, Peralta LA, Basu GD, Choudhary A, Tibes R, et al. RNAi 56. Livak KJ, Schmittgen TD. Analysis of relative gene expression data using screening of the kinome identifies modulators of cisplatin response in real-time quantitative PCR and the 2(-Delta Delta C(T)) Method. Methods ovarian cancer cells. Gynecol Oncol 2010;118:220–7. 2001;25:402–8. 35. Bartz SR, Zhang Z, Burchard J, Imakura M, Martin M, Palmieri A, et al. Small 57. Martens-de Kemp SR, Brink A, Stigter-van Walsum M, Damen JM, Rusten- interfering RNA screens reveal enhanced cisplatin cytotoxicity in tumor burg F, Wu T, et al. CD98 marks a subpopulation of head and neck cells having both BRCA network and TP53 disruptions. Mol Cell Biol squamous cell carcinoma cells with stem cell properties. Stem Cell Res 2006;26:9377–86. 2013;10:477–88. 36. Bauer JA, Ye F, Marshall CB, Lehmann BD, Pendleton CS, Shyr Y, et al. 58. Martens-de Kemp SR, Dalm SU, Wijnolts FM,Brink A, Honeywell RJ, Peters GJ, RNA interference (RNAi) screening approach identifies agents that et al. DNA-bound platinum is the major determinant of cisplatin sensitivity in enhance paclitaxel activity in breast cancer cells. Breast Cancer Res head and neck squamous carcinoma cells. PLoS One 2013;8:e61555. 2010;12:R41–56. 59. Spankuch-Schmitt B, Bereiter-Hahn J, Kaufmann M, Strebhardt K. Effect of 37. Chen S, Blank JL, Peters T, Liu XJ, Rappoli DM, Pickard MD, et al. Genome- RNA silencing of polo-like kinase-1 (PLK1) on apoptosis and spindle wide siRNA screen for modulators of cell death induced by proteasome formation in human cancer cells. J Natl Cancer Inst 2002;94:1863–77. inhibitor bortezomib. Cancer Res 2010;70:4318–26. 60. Li J, Zou C, Bai Y, Wazer DE, Band V, Gao Q. DSS1 is required for the 38. Harradine KA, Kassner M, Chow D, Aziz M, Von Hoff DD, Baker JB, et al. stability of BRCA2. Oncogene 2006;25:1186–94. Functional genomics reveals diverse cellular processes that modulate 61. Marston NJ, Richards WJ, Hughes D, Bertwistle D, Marshall CJ, Ashworth A. tumor cell response to oxaliplatin. Mol Cancer Res 2011;9:173–82. Interaction between the product of the breast cancer susceptibility gene 39. Iorns E, Turner NC, Elliott R, Syed N, Garrone O, Gasco M, et al. Identi- BRCA2 and DSS1, a protein functionally conserved from yeast to mam- fication of CDK10 as an important determinant of resistance to endocrine mals. Mol Cell Biol 1999;19:4633–42. therapy for breast cancer. Cancer Cell 2008;13:91–104. 62. Yang H, Jeffrey PD, Miller J, Kinnucan E, Sun Y, Thoma NH, et al. BRCA2 40. Mackeigan JP, Murphy LO, Blenis J. Sensitized RNAi screen of human function in DNA binding and recombination from a BRCA2-DSS1-ssDNA kinases and phosphatases identifies new regulators of apoptosis and structure. Science 2002;297:1837–48. chemoresistance. Nat Cell Biol 2005;7:591–600. 63. Gudmundsdottir K, Lord CJ, Witt E, Tutt AN, Ashworth A. DSS1 is required 41. Swanton C, Marani M, Pardo O, Warne PH, Kelly G, Sahai E, et al. for RAD51 focus formation and genomic stability in mammalian cells. Regulators of mitotic arrest and ceramide metabolism are determinants EMBO Rep 2004;5:989–93. of sensitivity to paclitaxel and other chemotherapeutic drugs. Cancer Cell 64. Wichmann G, Rosolowski M, Krohn K, Kreuz M, Boehm A, Reiche A, et al. 2007;11:498–512. The role of HPV RNA transcription, immune response-related gene expres- 42. Turner NC, Lord CJ, Iorns E, Brough R, Swift S, Elliott R, et al. A synthetic sion and disruptive TP53 mutations in diagnostic and prognostic profiling lethal siRNA screen identifying genes mediating sensitivity to a PARP of head and neck cancer. Int J Cancer 2015;137:2846–57. inhibitor. EMBO J 2008;27:1368–77. 65. Goeman JJ, Oosting J, Cleton-Jansen AM, Anninga JK, van Houwelingen 43. Whitehurst AW, Bodemann BO, Cardenas J, Ferguson D, Girard L, Peyton HC. Testing association of a pathway with survival using gene expression M, et al. Synthetic lethal screen identification of chemosensitizer loci in data. Bioinformatics 2005;21:1950–7. cancer cells. Nature 2007;446:815–9. 66. Simon N, Friedman J, Hastie T, Tibshirani R. Regularization paths for Cox's 44. Posthuma De Boer J, van Egmond PW, Helder MN, de Menezes RX, Cleton- proportional hazards model via coordinate descent. J Stat Softw Jansen AM, Belien JA, et al. Targeting JNK-interacting-protein-1 (JIP1) 2011;39:1–13. sensitises osteosarcoma to doxorubicin. Oncotarget 2012;3:1169–81. 67. Heagerty PJ, Zheng Y. Survival model predictive accuracy and ROC curves. 45. Pruitt KD, Tatusova T, Maglott DR. NCBI Reference Sequence (RefSeq): a Biometrics 2005;61:92–105. curated non-redundant sequence database of genomes, transcripts and 68. Langevin F, Crossan GP, Rosado IV, Arends MJ, Patel KJ. Fancd2 counteracts proteins. Nucleic Acids Res 2005;33:D501–4. the toxic effects of naturally produced aldehydes in mice. Nature 2011; 46. Lin CJ, Grandis JR, Carey TE, Gollin SM, Whiteside TL, Koch WM, et al. 475:53–8. Head and neck squamous cell carcinoma cell lines: established models and 69. Ma YY, Lin H, Chang FM, Chang TC, Trieu T, Pridgen HI, et al. Identification rationale for selection. Head Neck 2007;29:163–88. of the deleted in split hand/split foot 1 protein as a novel biomarker for 47. Hermsen MA, Joenje H, Arwert F, Welters MJ, Braakhuis BJ, Bagnay M, et al. human cervical cancer. Carcinogenesis 2013;34:68–78. Centromeric breakage as a major cause of cytogenetic abnormalities in oral 70. Rezano A, Kuwahara K, Yamamoto-Ibusuki M, Kitabatake M, Moolthiya P, squamous cell carcinoma. Genes Chromosomes Cancer 1996;15:1–9. Phimsen S, et al. Breast cancers with high DSS1 expression that potentially 48. Brenner JC, Graham MP, Kumar B, Saunders LM, Kupfer R, Lyons RH, et al. maintains BRCA2 stability have poor prognosis in the relapse-free survival. Genotyping of 73 UM-SCC head and neck squamous cell carcinoma cell BMC Cancer 2013;13:562. lines. Head Neck 2010;32:417–26. 71. Holzer AK, Samimi G, Katano K, Naerdemann W, Lin X, Safaei R, et al. The 49. Martens-de Kemp SR, Nagel R, Stigter-van Walsum M, van der Meulen I, copper influx transporter human copper transport protein 1 regulates the van Beusechem V, Braakhuis BJ, et al. Functional genetic screens identify uptake of cisplatin in human ovarian carcinoma cells. Mol Pharmacol genes essential for tumor cell survival in head and neck and lung cancer. 2004;66:817–23. Clin Cancer Res 2013;19:1994–2003. 72. Ishida S, Lee J, Thiele DJ, Herskowitz I. Uptake of the anticancer drug 50. Boutros M, Bras LP, Huber W. Analysis of cell-based RNAi screens. Genome cisplatin mediated by the copper transporter Ctr1 in yeast and mammals. Biol 2006;7:R66. Proc Natl Acad Sci U S A 2002;99:14298–302. 51. Smyth GK. Linear models and empirical Bayes methods for assessing 73. Lin X, Okuda T, Holzer A, Howell SB. The copper transporter CTR1 differential expression in microarray experiments. Stat Appl Genet Mol regulates cisplatin uptake in Saccharomycescerevisiae. Mol Pharmacol Biol 2004;3:Article3. 2002;62:1154–9. 52. Smyth GK. Limma: linear models for microarray data. In: Gentleman R, 74. Drayton RM, Dudziec E, Peter S, Bertz S, Hartmann A, Bryant HE, et al. Carey VJ, Dudoit S, Irizarry RA, Huber W, editors. Bioinformatics and Reduced expression of miRNA27a modulates cisplatin resistance in blad- computational biology solutions using r and bioconductor. New York, NY: der cancer by targeting the cysteine/glutamate exchanger SLC7A11. Clin Springer; 2005. p.397–420. Cancer Res 2014;20:1990–2000. 53. Benjamini Y, Hochberg Y. Controlling the false discovery rate - a practical 75. Wang J, Papanicolau-Sengos A, Chintala S, Wei L, Liu B, Hu Q, et al. and powerful approach to multiple testing. J Roy Stat Soc B Met Collecting duct carcinoma of the kidney is associated with CDKN2A deletion 1995;57:289–300. and SLC family gene up-regulation. Oncotarget 2016;7:29901–15. 54. de Menezes RX, Boer JM, van Houwelingen HC. Microarray data analysis: a 76. Degtyarev M, De MA, Orr C, Lin J, Lee BB, Tien JY, et al. Akt inhibition hierarchical T-test to handle heteroscedasticity. Appl Bioinformatics 2004; promotes autophagy and sensitizes PTEN-null tumors to lysosomotropic 3:229–35. agents. J Cell Biol 2008;183:101–16.

www.aacrjournals.org Mol Cancer Ther; 16(3) March 2017 549

Martens-de Kemp et al.

77. Gill C, Dowling C, O'Neill AJ, Watson RW. Effects of cIAP-1, cIAP-2 and 79. Chun TU, Epstein MR, Dick M, Andelﬁnger G, Ballester L, Vanoye CG, et al. XIAP triple knockdown on prostate cancer cell susceptibility to apoptosis, Polymorphic ventricular tachycardia and KCNJ2 mutations. Heart Rhythm cell survival and proliferation. Mol Cancer 2009;8:39. 2004;1:235–41. 78. Alagem N, Dvir M, Reuveny E. Mechanism of Ba(2þ) block of a mouse 80. Tsuboi M, Antzelevitch C. Cellular basis for electrocardiographic and inwardly rectifying Kþ channel: differential contribution by two discrete arrhythmic manifestations of Andersen-Tawil syndrome (LQT7). Heart residues. J Physiol 2001;534:381–93. Rhythm 2006;3:328–35.

550 Mol Cancer Ther; 16(3) March 2017 Molecular Cancer Therapeutics

The FA/BRCA Pathway Identified as the Major Predictor of Cisplatin Response in Head and Neck Cancer by Functional Genomics

Sanne R. Martens-de Kemp, Arjen Brink, Ida H. van der Meulen, et al.

Mol Cancer Ther 2017;16:540-550. Published OnlineFirst December 15, 2016.

Updated version Access the most recent version of this article at: doi:10.1158/1535-7163.MCT-16-0457

Supplementary Access the most recent supplemental material at: Material http://mct.aacrjournals.org/content/suppl/2016/12/14/1535-7163.MCT-16-0457.DC1

Cited articles This article cites 79 articles, 21 of which you can access for free at: http://mct.aacrjournals.org/content/16/3/540.full#ref-list-1

Citing articles This article has been cited by 1 HighWire-hosted articles. Access the articles at: http://mct.aacrjournals.org/content/16/3/540.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at Subscriptions [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://mct.aacrjournals.org/content/16/3/540. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.