Chromosomal aneuploidy is a common phenomenon in many cancers, and the analysis of copy number variations (CNVs) across multiple samples has helped identify relevant driver genes for human cancers. For example, several oncogenes includingMYC, EGFR, ERBB2 andCCND1 are recurrently amplified through chromosomal or focal gains, while multiple tumor suppressors such asATM, PTEN andCDKN2A are commonly deleted in different cancers¹.

Clear cell renal cell carcinoma (ccRCC), which accounts for 75–80% of all renal cell carcinomas, is characterized by loss of chromosome 3p in about 90% of the sporadic cases². Remarkably, 3p harbors the four most commonly mutated genes in ccRCC whose cancer-driving activities have been established in the disease;VHL³,PBRM1⁴,SETD2⁵, andBAP1⁶, which are mutated in 80%, 40%, 19% and 12% of cases, respectively^7,8,9. Inactivation of VHL leads to constitutive stabilization of the hypoxia inducible transcription factors (HIF), and abnormal activation of their downstream genes, which contribute to cancer development¹⁰. The remaining three genes encode proteins involved in chromatin remodeling and histone modifications, highlighting the important role of epigenome aberration in the disease². While the incidence of ccRCC is increasing worldwide, the male-to-female incidence ratios are typically within the range of 1.5–2:1.0¹¹, arguing for a sex-specific analysis of the genomic abnormalities. Here, we set out to investigate the occurrence and the extent of germline and somatic CNVs in sporadic ccRCC in male and female patients separately, and to further characterize those affecting sex chromosomes.

Loss of chromosome Y is common in ccRCC

Using whole-genome sequencing (WGS) data of ccRCC and matched constitutional DNA sample pairs, which we have reported recently¹², we interrogated CNVs in DNA from 52 male and 41 female patients (discovery set;Supplementary Table 1) by analyzing coverage of sequencing reads mapped to each chromosome (see Methods). In line with previous literature, the most frequent somatic CNV was the loss of 3p detected in 91% of samples, followed by recurrent gains of chromosomes 5q (32%), 7 (23.6%), 12 (13%), and losses of chromosomes 14q (30%), 8p (29%) and 9 (16%). Overall, tumors from male patients exhibited higher prevalence for the recurrent chromosomal aberrations, in particular for gain of 7q (28% in males vs. 17% in females) and deletion of 9p (25% in males vs. 10% in females) (Fig. 1a). In contrast, we observed that loss of chromosome X (LOX) exclusively happens in female patients (17.1% of female cases). Given that several X-linked genes escape X-inactivation, and have therefore two functional copies in females but one in males, this observation suggests that presence of a copy of chromosome X may potentially be essential for the survival of cancer cells. Curiously, whereas no tumors from male patients displayed LOX, loss of chromosome Y (LOY) was the second most frequent somatic chromosome aneuploidy in these tumors (36.5% of male subjects, N = 19;Fig. 1a). The fraction of cells estimated to be affected by somatic LOY in these patients ranged from 11% to 75%, and in 14 patients somatic LOY was detected in at least 20% of the cells (Fig. 1b). Next, we examined the presence of CNVs in constitutional DNA isolated from peripheral samples collected from the same patients. Of significance, LOY was the only recurrent aneuploidy in constitutional DNA of our samples that was detected in 5 male patients (9.6%;Supplementary Figure 1), of which 4 showed the deletion in more than 20% of cells (Fig. 1b). Corroborating previous studies^13,14, the observed LOY was associated with older age in patients (P = 0.04); the average age of patients with LOY in the peripheral blood was 68.9 year in comparison to 58.8 year in those without this abnormality. Notably, we did not observe any association between age of patients and extent of somatic LOY in tumors of the affected patients.

(a) Bar graphs show the frequency of copy number variations across the genome in ccRCC tumors. Frequencies are presented in samples from female and male cases separately. (b) Status of chromosome Y in DNA isolated from tumors (Y-axis) and patient-matched peripheral blood (X-axis) is shown for individual male subjects. In samples affected by LOY, the normalized coverage of chromosome Y, shown on Y and X axes for tumor and normal samples, respectively, is lower than the expected value of 0.5. The color codes define patient groups with different states for LOY.

Full size image

LOY is a whole-chromosome event

Given the high prevalence of LOY in tumors and peripheral DNA of male patients, we further analyzed LOY in our sample series, particularly whether the observed LOY spans the whole chromosome or is focal. Analysis of sequencing read coverage along chromosome Y showed that the loss is observed throughout the chromosome in samples affected by LOY (Fig. 2a), suggesting that the deletion affects the whole chromosome. Based on availability of DNA, we subjected samples from seven of the patients affected with somatic LOY to verification by an orthogonal Y-Chromosome deletion detection assay surveying the presence of twenty specific regions of the Y chromosome by polymerase chain reaction (PCR) (see Methods). Somatic LOY at the chromosomal level was confirmed in all examined tumors, evident from an attenuated amplification of Y-chromosome-specific loci in DNA isolated from tumor samples compared to that of the matched constitutional DNA. This pattern was not observed in samples of other male patients who had not been identified as being affected by somatic LOY based on the analysis of their WGS data (Fig. 2b).

(a) Sequencing coverage across chromosome Y is shown in constitutional DNA samples without (top) and with LOY (middle), and in a tumor sample with LOY (bottom). (b) The cartoon on top depicts the location of the loci examined by PCR on Y chromosome. The dot graph on bottom shows average of relative amplification values (Tumor/normal samples of the same patient) for each locus in patients with (red) and without (blue) somatic LOY (SLOY). Error bars show the range across patients of each group.

Full size image

To confirm these findings, we screened tumor and matched control DNA sample pairs of an additional 48 male ccRCC patients (validation set) for LOY using the above PCR-based assay. This analysis revealed somatic LOY in 20 (42.7%) of the validation sample set, demonstrating that this is a common genomic aberration in ccRCC, detected in 39.6% overall (discovery and validation sets; n = 100) of male ccRCC patients (Supplementary Figure 2). Analysis of association between somatic LOY and clinical annotations including tumor stage or grade did not show any significant relationships.

LOY results in downregulation of epigenetic modifier genes

We further examined the possible effect of somatic LOY at the RNA level by interrogating a RNA-Seq dataset on gene expression in normal and tumor samples from male patients within the discovery set¹². We found that 11 genes had significantly different patterns of expression in tumors of the patients with and without somatic LOY (false-discovery rate (FDR) <0.01;Supplementary Table 2). These 11 genes were located on chromosome Y, and while expressed in normal kidney tissue, exhibited lower expression in tumors of patients harboring somatic LOY, indicating that this aberration may have functional consequences through deregulation of the affected genes. Moreover, the level of expression of each gene was found to be inversely correlated to the proportion of cells affected by LOY (Fig. 3). This observation was confirmed using gene expression data generated by microarrays, which was available for 29 tumors of the validation set¹⁵ (Supplementary Figure 3). We surveyed the list of genes affected for potential functionally-relevant candidates. Among these genes,TMSB4Y has recently been identified as a tumor suppressor gene downregulated in male breast cancers¹⁶, but not connected to ccRCC. Likewise, deletion ofKDM5D has been detected in 52% of prostate cancers¹⁷.KDM5D encodes a lysine-specific histone H3 demethylase, which plays an important role in epigenetic regulation¹⁸. Furthermore, it has been shown that knockdown of KDM5D through RNA-interference (RNAi) increases cell proliferation and reduces apoptosis in prostate cancer¹⁹, suggesting a tumor suppressor function for this gene. Intriguingly,KDM5C, the X-linked homologue ofKDM5D is recurrently mutated in ccRCC^8,9,12, and its inactivation leads to genomic instability in ccRCC through deregulation of H3K4 methylation²⁰.KDM5D shows 85% sequence identity toKDM5C, and the products of these two genes possess a similar function in demethylating tri-methyl H3K4^18,20. Given this functional similarity, we surveyed the mutational status ofKDM5C in our discovery set, and investigated possible relationships between mutational status ofKDM5C andKDM5D in tumors of male patients. In female patients,KDM5C was deleted in tumors of 7 cases through somatic LOX, and was affected by focal somatic deletions in two additional patients. Furthermore, somatic mutations ofKDM5C were present in tumors of 3 patients who were also affected with LOX (P = 0.003, Fisher’s exact test,Supplementary Figure 4). Overall,KDM5C was affected with somatic genomic aberrations in 9 out of 41 (22%) female cases. AsKDM5C escapes the X-inactivation²¹, the concomitant mutations ofKDM5C and LOX in the same tumors may suggest that this gene is a classical tumor suppressor affected with bi-allelic inactivation in ccRCC. In male cases, we identifiedKDM5C mutations in tumors of 3 patients (5.8%), of which one was also affected by somatic LOY (Supplementary Figure 4). We did not detect any mutation or a focal CNV affectingKDM5D in tumors of the male patients who did not exhibit LOY.

Expression of Y chromosome genes downregulated in patients affected by somatic LOY is compared to the proportion of cells estimated to harbor somatic LOY in individual tumor samples.

Full size image

Our list of LOY-associated down-regulated genes (Supplementary Table 2) includes another epigenome modifier with an X-linked homologue that is also recurrently mutated in ccRCC;UTY/KDM6C. KDM6C demethylates H3K27, a function similar to that of KDM6A²². These genes also share over 83% in sequence similarity, resulting in highly conserved active sites in their products. Mutations ofKDM6A leading to its inactivation have been recurrently observed in ccRCC^8,23, highlighting this gene as a potential key tumor suppressor in renal cancer. In addition to being affected by somatic LOX in 7 female patients,KDM6A was also affected by focal deletion in a female patient in our cohort.

KDM5D expression reduces viability of renal cancer cells

Given the reported tumor-suppressive function of KDM5D in prostate cancer¹⁹, and of its X-link homolog KDM5C in renal cancer²⁰, we set out to examine whether KDM5D expression has an anti-tumor activity in renal cancer. We first evaluated KDM5D expression levels in several renal cancer cell lines, which have been derived from tumors resected from male patients. Amongst cell lines examined, ACHN cell line did not show any expression for KDM5D (Fig. 4a). This observation was in line with a previous study reporting the loss of chromosome Y in ACHN cell line²⁴. We therefore selected this cell line for functional analysis of KDM5D expression. Ectopic expression of KDM5D cells reduced cell viability to 65% as compared to control transfection (Fig. 4b,c), suggesting the potential involvement of KDM5D depletion in renal cancer pathology.

(a) Expression levels of KDM5D mRNA in renal cancer cell lines derived from tumors procured from male patients, as measured by qRT-PCR.GAPDH served as a housekeeping gene for measurement of relative gene expression. (b) Over expression of KDM5D in ACHN cell line reduces cell viability. Values are the mean ± SD of six independent experiments. **P < 0.01 when compared to the corresponding results from control (ctrl) (Mann-Whitney U test). (c) Over expression of KDM5D following transfection was confirmed using qRT-PCR.

Full size image

Emerging data emphasizes an association between LOY in peripheral blood and higher risk of cancer²⁵. Likewise focal or chromosome-level somatic LOY occurs recurrently in different malignancies; however, current knowledge of mechanisms by which LOY may contribute to cancer is limited. Recent genomic studies of ccRCC have highlighted the importance of molecular aberrations that impair the function of chromatin remodeling and epigenetic modifiers in ccRCC development^{5,20,26,27,28,29}. Our study expands these findings by highlighting the prevalence of somatic LOY among men affected by ccRCC, and suggesting a functional relevance for this aberration through down-regulation of previously unrecognized epigenetic modifiersKDM5D andKDM6C. Given the functional similarities between these genes and their X-linked homologs, it is plausible that down-regulation ofKDM5D andKDM6C, through somatic LOY, may contribute to ccRCC development or progression. Ourin vitro data shows that over expression of KDM5D in cancer cells that are affected by LOY reduces cell viability. These findings indicate that down-regulation of KDM5D through LOY may contribute to the pathogenesis of renal cancer. However, further detailed analysis through future functional studies is warranted to understand the exact function and pathway context of KDM5D in renal cancer.

Patient samples and DNA isolation

Clinical information for patients included in this study is presented inSupplementary Table 1. Patients undergoing nephrectomy for suspected renal cancer during the period December 2008 to March 2011 at St James’s University Hospital in Leeds, UK; University Hospital Motol, Prague, Czech Republic; Masaryk Memorial Cancer Institute, Brno, Czech Republic; Th. Burghele Hospital, Bucharest, Romania; and N. N. Blokhin Cancer Research Centre, Moscow, Russia, were recruited to the study after informed consent was obtained. Recruitment in Central and Eastern Europe was coordinated by the International Agency for Research on Cancer (IARC). All experiments and methods were performed in accordance to the ethics guidelines from the International Cancer Genome Consortium (ICGC) and to the relevant national regulations and with sampling and clinical data collection being undertaken according to predefined standard operating procedures (SOPs) based on guidelines from ICGC. Ethical approvals were obtained from the Leeds (East) Local Research Ethics Committee, the IARC Ethics Committee, as well as from local ethics committee for recruiting centers in Czech Republic, Romania, and Russia. DNA from fresh-frozen tumor tissue samples and buffy coat was isolated using Autopure (Qiagen) as described previously¹², and were quantified by Quant-iT PicoGreen dsDNA Assay Kit (Invitrogen, ON, CAN).

Inference of LOY from WGS data

WGS data of tumor and blood DNA samples studied here were reported previously¹². To detect aneuploidy and LOY from WGS data, we first measured read coverage across the genome in 5 Kbp bins. In each sample, the coverage was normalized by the median coverage across the autosomes. We then estimated, for each sample, the median normalized coverage in each chromosome arm. The only exception was chromosome Y which was considered as a whole. In order to avoid noise due to mappability issues, we used only the top 1000 bins with the lowest median divergence from the expected baseline in the normal samples. We used this normalized median coverage per chromosome arm to test aneuploidy in each sample. For each chromosome arm (or chromosome Y), a mixture of two Gaussian distributions was fitted to the empirical distribution of the median normalized coverage across samples. The main Gaussian was used as the null distribution (Supplementary Figure 5) to derive P-values. A chromosome arm was flagged as aneuploid if the Bonferonni-adjusted P-value was smaller than 0.01 and at least 10% of cells were affected. The proportion of cell with aneuploidy was estimated as the proportion of missing/excess coverage. For LOY, we expect a normalized coverage of 0.5 and the proportion of cells with LOY was (0.5-coverage)/0.5.

We used a logistic regression to test the association of LOY with age. Finally, the CNVs used forKDM5C orKDM6A deletion investigation were detected by PopSV³⁰ using the normal samples as reference and 5 Kbp bins.

PCR-based detection of LOY

To examine the status of LOY in DNA of tumor and blood samples, Y Chromosome Deletion Detection System assay, Version 2 (Promega, WI, USA) was used as instructed by the manufacturer. Briefly, 20 specific regions of the Y chromosome were amplified by PCR using 5 multiplex master mixes, and PCR products were loaded on a QIAxcel instrument (Qiagen, ON, CAN). Densities of PCR products were estimated by BioCalculator software (v.3.2) and a normalization was performed by the control primer pair included in each multiplex master mix to control the amplification efficacy. We also included samples from three male subjects without LOY and one female sample to control the performance of the assay. Similar to the analysis on WGS data, the probes were first normalized by the median probe amplification value across the normal samples. Then the median of the normalized amplification was computed for each sample. It summarized the overall amplification of chromosome Y in each sample. These values were used to produceSupplementary Figure 2 and to identify LOY. Following the same analysis as for the WGS data, the mixture of Gaussian distributions was fitted on the normalized amplification of the normal samples. Samples which deviated significantly (P < 0.01) from the expected amplification and with an estimated proportion of affected of cells >10% were flagged as being affected by LOY.

Gene expression analysis

Transcriptome profiles of the tumor samples included in this study (previously reported in our earlier publication¹²), were used to examine differential gene expression between male subjects affected with somatic LOY and those without this abnormality. RNA-seq data was available for tumors of 34 patients, of which 21 had RNA-seq for matched normal kidney samples. Differentially expressed genes between tumors affected with somatic LOY and those without this abnormality were identified using Student’s T-Test on log2-transformed RPKM data, and the Benjamini-Hochberg method was used to correct multiple testing. Genes with a FDR< 0.01 were considered differentially expressed. A linear regression was used to test the association between the proportion of cells with somatic LOY and gene expression (RPKM).

Gene expression microarray data for 29 tumors of validation samples had previously been reported¹⁵, and were used to confirm the anti-correlation between the proportion of cells with somatic LOY and gene expression levels (log2 intensity).

Cell viability assay

Renal cancer cell lines 786-O, A704, Caki-2, ACHN were obtained from ATCC (Rockville, USA) and cultured in RPMI, EMEM and McCoy medium supplemented with 10% (v/v) fetal bovine serum (FBS), 100 U/ml penicillin and 100 lg/ml streptomycin. Cells were incubated at 37 °C and 5% (v/v) CO2. For viability assays, 5000 cells were transfected with 100 ng of either KDM5D cDNA-expressing (courtesy of Dr. Stephane Richard) or control empty vector (Sigma, Oakville, Canada) in 96-well plates using Lipofectamine 3000 (Invitrogen) according to the manufacturer’s instructions. CellTiter-Glo assay (Promega, WI, USA) was used to assess cell viability after 72 hours post-transfection.

Quantitative real-time PCR (qRT-PCR)

Total RNA was extracted from cells using miRNeasy kit (Qiagen, Toronto, Canada) according to the supplier protocols. 1 μg RNA was reverse transcribed into complementary DNA (cDNA) using Transcriptor First Strand cDNA Synthesis Kit (Roche, Laval, Canada) following instructions provided by the manufacturer. Real-time PCR reactions were prepared using LightCycler 480 SYBR green I master kit (Roche), and were run on a LightCycler 480 instrument (Roche) according to the manufacturer’s recommendations. Triplicate PCR reactions were performed for each sample to ensure reliability. Expression ofKDM5D mRNA was normalized to the expression of the housekeeping geneGAPDH, and was reported as 2^−ΔCt. All the primers were purchased from IDT (Coralville, IA, US). The sequences of primers were CGTGGAAGGACTCATGACCA (GAPDH forward), GCCATCACGCCACAGTTTC (GAPDH reverse), CGCAGCTTTGAAGAGCTAAG (KDM5D forward) and CAGCTGTGGAGTGTCCATCC (KDM5D reverse).

How to cite this article: Arseneault, M.et al. Loss of chromosome Y leads to down regulation of KDM5D and KDM6C epigenetic modifiers in clear cell renal cell carcinoma.Sci. Rep.7, 44876; doi: 10.1038/srep44876 (2017).

Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

This study was supported by funding from EU FP7 under grant agreement number 241669 (the CAGEKID project,http://www.cng.fr/cagekid), and funding from Génome Québec as well as Ministère de l’Enseignement supérieur, de la Recherche, de la Science et de la Technologie (MESRST). We also acknowledge the support of Cancer Research UK Centre and ECMC infrastructure funding in Leeds for contributions to sample collection. The Czech-Brno group was supported by MH CZ - DRO (MMCI, 00209805). JM is supported by funds from National Sciences and Engineering Research Council (NSERC-448167-2013).

1. Madeleine Arseneault and Jean Monlong: These authors contributed equally to this work.

Authors and Affiliations

1. Department of Human Genetics, McGill University, 1205 Dr Penfield Avenue, Montreal, H3A 1B1, QC, Canada

   Madeleine Arseneault, Jean Monlong, Maryam Safisamghabadi, Nazanin Nourbehesht, Pudchalaluck Panichnantakul, Pouria Jandaghi, Mark Lathrop, Guillaume Bourque & Yasser Riazalhosseini

2. McGill University and Genome Quebec Innovation Centre, 740 Doctor Penfield Avenue, Montreal, H3A 0G1, QC, Canada

   Madeleine Arseneault, Jean Monlong, Maryam Safisamghabadi, Nazanin Nourbehesht, Pudchalaluck Panichnantakul, Pouria Jandaghi, Mark Lathrop, Guillaume Bourque & Yasser Riazalhosseini

3. Leeds Institute of Cancer and Pathology, University of Leeds, Cancer Research Building, St James’s University Hospital, Leeds, LS9 7TF, UK

   Naveen S. Vasudev, Patricia Harnden & Rosamonde E. Banks

4. International Agency for Research on Cancer (IARC), 150 cours Albert Thomas, 69008, Lyon, France

   Ruhina S. Laskar, Paul Brennan & Ghislaine Scelo

5. Department of Pathology, Karolinska Institutet, Stockholm, SE-171 77, Sweden

   Lars Egevad

6. First Faculty of Medicine, Institute of Hygiene and Epidemiology, Charles University in Prague, Studničkova 7, Praha 2, 128 00, Prague, Czech Republic

   Ivana Holcatova

7. University Hospital Motol, V Úvalu 84, 150 06, Prague, Czech Republic

   Antonin Brisuda

8. Department of Preventive Medicine, Faculty of Medicine, Palacky University, Hnevotinska 3, 775 15, Olomouc, Czech Republic

   Vladimir Janout & Helena Kollarova

9. Department of Cancer Epidemiology and Genetics, Masaryk Memorial Cancer Institute and MF MU, Zluty Kopec 7, 656 53, Brno, Czech Republic

   Lenka Foretova & Marie Navratilova

10. National Institute of Public Health, Dr Leonte Anastasievici 1–3, sector 5, Bucuresti, 050463, Romania

    Dana Mates

11. Carol Davila University of Medicine and Pharmacy, Th. Burghele Hospital, 20 Panduri Street, 050659, Bucharest, Romania

    Viorel Jinga

12. Russian N.N. Blokhin Cancer Research Centre, Kashirskoye shosse 24, Moscow, 115478, Russian Federation

    David Zaridze & Anush Mukeria

13. European Molecular Biology Laboratory, European Bioinformatics Institute, EMBL-EBI, Wellcome Trust Genome Campus, Hinxton, CB10 1SD, UK

    Alvis Brazma

14. Laboratory for Epigenetics & Environment, Centre National de Génotypage, CEA-Institut de Génomique, 2 rue Gaston Crémieux, Evry, 91000, France

    Jorg Tost

Contributions

Y.R. conceived the study and designed the experiments with contribution from G.B. N.V., P.H., L.E., I.H., A.B., V.J., H.K., L.F., M.N., D.M., Vi.J., D.Z., A.M., P.B., G.S. and R.E.B. were responsible for patient selection, sample collection, sample preparation and pathological reviews. M.A. and N.N. prepared DNA, performed experiments to detect LOY and analyzed the data with contribution from P.J. J.M. developed the computational methods for WGS analysis, analyzed WGS data and generated the figures. P.P., R.S.L., P.J. and Al.B. performed the gene expression analysis. M.L. provided critical advice on data analysis and statistical approaches. M.S. and P.J. performed functional analysis with cell line models. M.A. and Y.R. wrote the manuscript, with assistance from J.T., G.S. and R.E.B.

Corresponding author

Correspondence toYasser Riazalhosseini.

Competing interests

The authors declare no competing financial interests.

This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visithttp://creativecommons.org/licenses/by/4.0/

Reprints and permissions