Genetic variation at the 8q24.21 renal cancer susceptibility locus affects HIF binding to a MYC enhancer

Clear cell renal cell carcinoma (ccRCC) is characterized by loss of function of the von Hippel–Lindau tumour suppressor (VHL) and unrestrained activation of hypoxia-inducible transcription factors (HIFs). Genetic and epigenetic determinants have an impact on HIF pathways. A recent genome-wide association study on renal cancer susceptibility identified single-nucleotide polymorphisms (SNPs) in an intergenic region located between the oncogenes MYC and PVT1. Here using assays of chromatin conformation, allele-specific chromatin immunoprecipitation and genome editing, we show that HIF binding to this regulatory element is necessary to trans-activate MYC and PVT1 expression specifically in cells of renal tubular origins. Moreover, we demonstrate that the risk-associated polymorphisms increase chromatin accessibility and activity as well as HIF binding to the enhancer. These findings provide further evidence that genetic variation at HIF-binding sites modulates the oncogenic transcriptional output of the VHL–HIF axis and provide a functional explanation for the disease-associated effects of SNPs in ccRCC.

I n clear cell renal cell carcinoma (ccRCC), but few other cancers, somatic loss-of-function mutations, chromosomal aberrations or promoter hypermethylation lead to decreased activity of von Hippel-Lindau tumour suppressor protein (pVHL). pVHL is the recognition component of an E3 ubiquitin ligase complex that targets hypoxia-inducible factor (HIF) alpha subunits to the ubiquitin-proteasome pathway. Dysfunctional pVHL therefore disrupts proteasomal degradation of HIF-a subunits (HIF-1a and HIF-2a) and increases expression of HIF target genes 1,2 . VHL mutations are considered to be 'truncal' mutations in ccRCC and HIF stabilization can already be detected in early pre-cancerous lesions in tubular segments bearing biallelic VHL mutations within kidneys of patients with von Hippel-Lindau disease 3 . Though the reasons for the marked tissue restriction of VHL-associated cancer are unclear, genetic and epigenetic factors can influence RCC development 4-7 . In this context, genome-wide association studies have identified single-nucleotide polymorphisms (SNPs) that are specifically associated with renal cancer susceptibility [8][9][10] . So far, two genetic regions with ccRCC-related SNPs may have an impact on the VHL-HIF signalling axis. SNPs on chromosome 2 are located within the first intron of the EPAS1 gene coding for HIF-2a and SNPs on chromosome 11 associate with a HIF-2-binding enhancer, which trans-activates the CCND1 oncogene 11,12 . Recently, a novel variant rs35252396, a two base pair substitution AC4CG, has been detected on chromosome 8q24. 21 (ref. 9). rs35252396 is strongly associated with renal cancer risk in Icelandic and other populations of European descent (odds ratio 1.27, P-value 5.4 Â 10 À 11 , minor allele frequency 0.46 in the combined analysis) 9 . The index polymorphism is located in an intergenic and putative regulatory region interposed between the major oncogenic driver MYC (136 kb upstream) and the oncogenic long noncoding RNA PVT1 (14 kb downstream). MYC orchestrates metabolic and growth-promoting pathways, and dysregulation is a hallmark of tumour initiation 13,14 . With respect to the VHL-HIF axis in ccRCC, MYC interacts differentially with the HIF-1a and HIF-2a subunits, thereby possibly contributing to the isoform-specific effects that are important in ccRCC 15,16 . Across all cancers, the MYC locus displays the highest susceptibility to somatic copy-number gains and both, MYC and PVT1, are co-amplified in most cases (498%) 17,18 . In cancer tissue, PVT1 RNA levels correlate well with MYC levels and appear to be necessary for MYC protein stabilization and tumour growth 18,19 . In addition to chromosomal rearrangements, the 8q24 locus is a hot spot for intergenic SNPs associated with a variety of tumours such as colorectal, ovarian, urinary bladder or prostate cancer, Hodgkin's lymphoma and chronic lymphocytic leukaemia [20][21][22][23][24][25][26][27][28][29][30][31][32][33] . Despite this high density of multiple cancerassociated variants in the 8q24.21 region, functional analyses have mainly been restricted to the SNP rs6983267, which is associated with colorectal and prostate cancer. This SNP resides in a regulatory element 335 kb telomeric of MYC and influences the expression of MYC by affecting activity of an enhancer [34][35][36][37][38] . However, the renal cancer-associated variant rs35252396 observed in the Icelandic population is not in linkage with any other disease-associated SNP in the 8q24.21 region (r 2 o0.02) (ref. 9), suggesting the involvement of renal cancer-specific mechanisms in generating the predisposition.

Results
MYC and PVT1 are direct targets of HIF in ccRCC. In contrast with many other cancers, copy-number gains at the MYC locus are relatively infrequent in ccRCC 18 . As a first step in analysing mechanisms associated with variant rs35252396, we therefore sought to define the frequency of dysregulated MYC and PVT1 expression in renal cancer. Analysis of RNA-seq data from clear cell, papillary and chromophobe RCCs generated by the TCGA consortium [39][40][41] confirmed that MYC and PVT1 RNA are commonly overexpressed in ccRCC (Fig. 1a). In line with the results from RNA analyses, positive MYC protein staining was strongly associated with the clear cell phenotype in a tissue microarray containing 453 unselected renal cancer specimens (Erlangen RCC Cohort, Supplementary Fig. 1). To better understand this contrast between RNA and copy-number changes, we proceeded to investigate direct involvement of the VHL-HIF axis in MYC and PVT1 expression. In pVHL re-expressing RCC4 cells with low HIF, levels of both transcripts were reduced compared with the pVHL-defective parental cells (Fig. 1b). When exposed to the hypoxia mimetic and HIF stabilizer dimethyl oxalylglycine (DMOG), we measured an increase of MYC and PVT1 RNA in pVHL re-expressing RCC4 transfectants, whereas no difference was determined in the respective pVHL-defective cells (Fig. 1b), suggesting that MYC is regulated by the HIF pathway in this context. To examine the specificity of hypoxic MYC and PVT1 regulation, we expanded our analysis to a variety of cells from renal tubular origins (renal cancer cell lines, primary renal tubular cells and immortalized renal tubular cells) and non-tubular origins (immortalized podocytes and non-renal cells) with functional re-expressed or wild-type VHL. Strikingly, significant induction of both MYC and PVT1 RNA by DMOG was specifically observed in renal cancer cell lines and tubular cells ( Fig. 1c; Supplementary Fig. 2). The striking selectivity of MYC messenger RNA regulation by the VHL-hypoxia pathway in renal tubular cells was also concordant with the data from transgenic animals with conditional knockout of VHL in tubular cells in which MYC is strongly induced (Supplementary Fig. 3). Taken together, this suggests that genes encoding MYC and PVT1 are targets of HIF in renal tubule-derived cells.
To test for association between MYC and HIF protein expression, we stained tissue microarray sections from the Erlangen RCC cohort for HIF-1a and HIF-2a. In ccRCC, HIF-1a and HIF-2a correlated significantly with positive MYC staining (Fig. 1d). To directly examine the role of HIF in MYC/PVT1 regulation, we performed short interfering RNA (siRNA)-mediated knockdown of HIF-a subunits in pVHL-competent RCC cells. In pVHL re-expressing RCC4 or 786-O and VHL wild-type RCC L34 cells, induction of MYC and PVT1 by DMOG was significantly reduced after HIF depletion ( Fig. 1e; Supplementary Fig. 4). HIFs are transcription factors that activate gene expression by direct binding to chromatinized DNA 42,43 . Therefore, we interrogated both newly acquired and previously published HIF-1b chromatin immunoprecipitation-DNA sequence (ChIP-seq) data sets at the MYC and PVT1 loci for HIF-DNA binding in a variety of cell types 11,43 . This revealed robust HIF-binding signals across a series of pVHL-defective renal cancer cell lines as well as immortalized proximal tubular and primary tubular cells in which HIF was stabilized by hypoxia or DMOG at intergenic sites located between the MYC and PVT1 genes (Fig. 2a). In line with the lack of hypoxic gene induction, no significant HIF-binding signals were detected at these sites in cells not derived from renal tubules. Very interestingly, consistent HIF-binding signals in the renal tubule-derived cells almost precisely coincided with the renal cancer susceptibility SNP rs35252396, which locates 205 bp downstream of a hypoxiaresponsive element (HRE) centred on the HIF-binding peak.
To address the importance of this locus in renal oncogenesis, we analysed the function of the SNP-associated HIF-binding site in detail. Inspection of epigenetic data from our laboratories and the ENCODE consortium revealed enriched signals for open (FAIRE, formaldehyde-assisted isolation of regulatory elements) and active chromatin (H3K4me1 and H3K27ac) at this site in pVHL-defective 786-O cells that are homozygous for the risk allele at rs35252396 (Fig. 2b). Confirming the cell-type specificity, levels of these markers of active chromatin were low in MCF-7 breast cancer cells that lack both HIF-a binding at the SNP-associated site and regulation of MYC or PVT1 RNA by hypoxia (Fig. 2b). As with studies of gene expression, we expanded the analysis of chromatin accessibility at the enhancer to panels of cells from tubular and non-tubular origins. In FAIRE experiments, enrichment of open chromatin at the HIF-binding locus was significantly greater in renal tubulederived cells compared with non-tubular cells ( Supplementary  Fig. 5). Taken together, the data from both expression and ChIP-seq studies suggest that HIF binds at an enhancer located between the MYC and PVT1 locus, and likely drives transcription of these genes. Open chromatin, HIF-binding and HIF-dependent regulation of MYC and PVT1 appear to be highly cell-type specific and restricted to cells of renal tubular origins.
The SNP-associated enhancer is necessary for MYC regulation. To further validate a functional relation between this HIF-binding enhancer and the promoters of nearby genes, we conducted chromatin conformation (Capture-C) and genome editing experiments in pVHL-defective cells. The Capture-C assay examines physical interaction of selected anchor sites with any distant chromatin region 44 . Using this technique, we observed chromatin interactions of the HIF-binding region with both the MYC and PVT1 promoters in 786-O cells ( Fig. 3a; Supplementary  Fig. 6). However, since we had identified more than one HIF-binding signal in the B20 kb stretch of DNA surrounding the putative enhancer ( Fig. 2a), we wished to specifically analyse the transactivation potential of the HIF binding that was most clearly associated with the predisposing SNPs. We therefore designed a guide RNA targeting the HRE in the centre of the SNP-associated HIF-binding signal and disrupted this HRE in 786-O renal cancer cells using CRISPR/Cas9 technology 45 . We screened 36 clones of cells for indel mutations at this site by PCR amplification followed by polyacrylamide gel analysis and identified 7 clones of cells with mutations that affected the a c e d b 25  pVHL-competent cells exposed to vehicle (ctrl, black) or 1 mM DMOG (grey) for 16 h. Data present mean ± s.d. from six independent experiments. *One-sample t-test, Po0.05). Fig. 7). We confirmed reduced binding of HIF and decreased marks of activity at the 8q24.21 enhancer in a selection of these cells with a defective HRE using HIF, RNApol2 and H3K27ac ChIP-quantitative PCR (qPCR; Fig. 3b,c; Supplementary Fig. 8). Further confirming the transactivation potential of this site, MYC and PVT1 RNA expression was significantly reduced by 40% and 32%, respectively, in comparison with the control HPRT gene, in cells with mutations that affected the HRE when compared with non-mutant clones of cells (Fig. 3d). MYC overexpression acts as a global amplifier of gene expression in tumour cells 46,47 . To test for the effects of reduced MYC levels on total RNA content in our cells, we measured RNA levels of HRE-defective and control cells. In line with the hypothesis of MYC-induced global gene expression, HRE-defective cells with lower MYC levels had lower RNA content than control cells (Fig. 3e). Thus, results from  epigenetic analyses, Capture-C and genome editing indicate that this enhancer site interacts with MYC and PVT1 promoters, is necessary for HIF-mediated transactivation of both genes and influences downstream effects of MYC.

HIF-binding site (Supplementary
rs35252396 affects chromatin accessibility and HIF binding. The overlap of a HIF-binding enhancer and renal cancer predisposing SNPs led us to consider whether the SNPs could affect HIF-DNA interactions and enhancer activity. In reporter assays using a DNA sequence spanning the HIF-binding site and SNPs, we measured a significant hypoxic induction of the reporter gene but no effect of the SNPs on reporter gene activity or hypoxic induction ( Supplementary Fig. 9). Reporter assays use non-chromatinized DNA and therefore do not capture epigenetic effects caused by the action of SNPs in chromatinized DNA.
To test for such an action, we proceeded to assays that examine SNP-associated effects on native chromatin. rs35252396 is located 205 bp downstream of the HIF-binding HRE and thus is likely to be present in most of the DNA fragments that are immunoprecipitated with HIF antibodies from native chromatin or isolated by FAIRE. We reasoned that this would allow testing of captured DNA for allelic imbalance of rs35252396 with respect to HIF binding or open chromatin. For these experiments, we identified primary tubular cells as well as pVHL-defective RCC4 and RCC L13 renal cancer cells that are heterozygous for rs35252396. In genotype-specific qPCR assays, the risk allele at rs35252396 was significantly enriched in chromatin fragments that bound HIF and had marks of active chromatin (Fig. 4a,   Similar results were obtained in both RCC cell lines and primary renal tubular cells (Fig. 4c,d; Supplementary Fig. 13). These data indicate that differential HIF binding at the renal cancerassociated 8q24. 21   Allelic imbalance in MYC and PVT1 expression. Allele-specific HIF binding should result in an allelic imbalance in MYC and PVT1 expression. To examine this, we identified primary renal tubular cells from 12 individuals who are heterozygous for rs11604, a SNP in the coding region of PVT1. This SNP is in weak LD with rs35252396 (r 2 ¼ 0.049; D 0 ¼ 0.270) and therefore the phase in the primary renal tubular cells is not known. We measured the allelic balance of rs11604 in genomic DNA and cDNA from control (untreated) or DMOG-treated cells using genotype-specific qPCR assays. Allelic imbalance in control samples was comparable between cells from individuals with a heterozygous (n ¼ 8) and homozygous (n ¼ 4) genotype at rs35252396 (Supplementary Fig. 14). However, the change in allelic expression of PVT1 induced by HIF stabilization compared with control was significantly greater in cells from individuals with a heterozygous genotype at rs35252396 (Fig. 5a). This is consistent with the hypothesis that unequal HIF binding at the enhancer drives differential expression of PVT1 from the two alleles. We were unable to establish a similar assay for the MYC-coding region and therefore resorted to genotype and tumour gene expression data from the TCGA consortium. rs35252396 was not genotyped in this cohort, but we identified SNPs in the TCGA cohort (rs10111989, rs4733579 and rs17775239) that were genotyped and are in LD with rs35252396 (ref. 9). Analysis of genotype expression correlations revealed that the risk allele of SNP rs10111989 (pairwise LD with rs35252396: r 2 ¼ 0.33, D 0 ¼ 0.98 (ref. 9)) exhibited a significant association with higher MYC expression (w 2 -test, P ¼ 0.0296, Fig. 5b). The other SNPs showed a correlation with MYC expression, but did not reach statistical significance (Supplementary Fig. 15). Though this analysis itself cannot implicate any specific polymorphism in the regulation of MYC in renal cancer, it is consistent with the data above implicating an rs35252396-associated phenotype.
Taken together, our findings identify a HIF-binding enhancer of oncogenic MYC and PVT1 expression (Fig. 5c). HIF binding, activity and accessibility of the enhancer as well as MYC/PVT1 induction are restricted to cells from renal tubular origin and dependent on the genotype of rs35252396, a polymorphism associated with renal cancer susceptibility.

Discussion
We demonstrate here that unrestrained activation of HIF in pVHL-defective renal cancer enhances expression of MYC and PVT1 via long distance interactions with a HIF-binding intergenic enhancer. This regulatory potential already exists in non-transformed tubular cells and in our analysis appears to be restricted to renal tubular cells. The pathway might therefore be important for cell homeostasis even in normal tubular epithelium and the earliest tubular neoplastic lesions that arise following inactivation of VHL. Modulation of HIF activity and MYC/PVT1 expression by polymorphisms located in this intergenic region could then promote or retard renal tumorigenesis. That many of the currently known renal cancer-associated SNPs (EPAS1, CCND1 and MYC/PVT1) can be linked to modulation of a single pathway (that is, the HIF pathway) is striking and to our knowledge unique in tumour biology. Interestingly, analysis of two of these loci, CCND1 and MYC/PVT1, indicates that the susceptibility determinants appear to operate on the chromatin structure at tissue-specific HIF-binding loci 11 . We conclude that the predisposing or protective effects of renal cancer-associated polymorphisms are explained at least in part by their ability to promote or inhibit, respectively, HIF expression or HIF-mediated transactivation of key oncogenic pathways. Cell lines were grown in DMEM, 100 U ml À 1 penicillin, 100 mg ml À 1 streptomycin and 10% fetal bovine serum (Sigma Aldrich). HKC-8 cells were cultured in DMEM/Ham's F-12 supplemented with 10% fetal calf serum, 2 mM Lglutamine, 100 U ml À 1 penicillin and 100 mg ml À 1 streptomycin, 5 mg ml À 1 insulin, 5 mg ml À 1 transferrin, and 5 ng ml À 1 selenium (Sigma Aldrich). Healthy human kidney cortical tissue from patients undergoing tumour nephrectomy was used for tubular cell isolation. Informed consent was given by each patient and the use of the tissue was approved by the local ethical committee at the University of Erlangen-Nürnberg. After mincing and DNAse I (Roche Diagnostics) and collagenase II (Gibco) digest, cells were sieved through a 100 mm and a 70 mm filter. Primary human tubular cells were cultured in DMEM/Ham's F-12 supplemented with 10% fetal calf serum, 2 mM L-glutamine, 100 U ml À 1 penicillin and 100 mg ml À 1 streptomycin, 5 mg ml À 1 insulin, 5 mg ml À 1 transferrin, 5 ng ml À 1 selenium (Sigma), tri-jodothyronin (T3) 10 ng ml À 1 , hydrocortisone 1 mg ml À 1 , and epidermal growth factor 100 mg ml À 1 (Peprotech). Epithelial origin was confirmed by immunocytochemistry for N-and E-Cadherin. HT1080 cells were cultured in minimal essential medium 10% fetal calf serum, 2 mM L-glutamine, 100 U ml À 1 penicillin and 100 mg ml À 1 streptomycin. As indicated sub-confluent cell cultures were exposed to 1 mM DMOG (Cayman) before collecting.

Methods
Chromatin immunoprecipitation. ChIP experiments were performed using the Upstate protocol (Millipore). Two 15 cm dishes of sub-confluent cells were used for crosslinking (1% formaldehyde for 12 min on ice). After 5 min incubation with glycine 125 mM on ice, cells were lysed in 1 ml lysis buffer and sonificated using a Bioruptor Plus sonicator (Diagenode) using 28 cycles with 15 s on and 15 s off. For immunoprecipitations, 6-10 ml of antibodies against HIF-1a (rabbit polyclonal, PM14 or Cayman Chemicals, Cay10006421), HIF-2a (rabbit polyclonal, PM9 or PM8), HIF-1b (rabbit polyclonal, Novus Biologicals, NB100-110), RNA polymerase II (rabbit polyclonal, Santa Cruz, SC-899), H3K27ac (rabbit polyclonal, Abcam, ab4729), H3K27ac (rabbit polyclonal, Diagenode, pAb-174-050), H3K4me1 (rabbit polyclonal, Abcam, ab8895) and H3 (rabbit polyclonal, Abcam, ab1791) 11,43,48,49 were used per 150 ml cell lysate. Rabbit serum or IgG (Millipore, 12-370) were used as negative controls as appropriate. Antibody-chromatin complexes were pulled down by proteinase A agarose beads (Millipore). ARTICLE After reversal of the crosslinking by heat, DNA was isolated by phenol-chloroform extraction. For ChIP-qPCR experiments, primers spanning the HIF-binding site within the 8q24.21 enhancer, a positive control at an EGLN3 intronic enhancer and a negative control at the CCND1 locus were used. qPCRs were performed using SYBRgreen chemistry (Thermo Scientifc) on a Step-one plus real-time PCR cycler (Applied Biosystems). Primers are listed in Supplementary Table 1. siRNA transfection and RNA isolation. siRNA against HIF-1a, HIF-2a and dHIF (drosophila HIF, control) are listed in Supplementary Table 2 and have been described earlier 50 . siRNA was transfected using Saint red (Synvolux) transfection reagent at a final concentration of 40 nM. Transfection was repeated after 24 h and cells were collected 48 h after the first transfection. Total RNA from cells or tissue was isolated using Tri Reagent (Sigma Aldrich) or peqGold total RNA kit (PeqLab) according to the manufacturer's protocol and transcribed into cDNA using the high capacity cDNA reverse transcription kit (Life Technologies). qPCRs were performed as described above and primers are listed in Supplementary Table 3.
Formaldehyde-assisted isolation of regulatory elements (FAIRE). Following the protocols from Giresi et al. 51 with some modifications, two 15 cm dishes of sub-confluent cells were used for crosslinking (1% formaldehyde for 5 min at room temperature) and for the preparation of input DNA. The isolation of DNA was performed using three rounds of phenol-chloroform extraction. SYBRgreen qPCR was performed on FAIRE DNA and input DNA. Values were normalized to input Twelve primary renal tubular cell cultures (PTC) were identified that are heterozygous for SNP rs11604 that resides in the coding region of PVT1. Four individuals had an AC/AC genotype and eight individuals had an AC/CG genotype at rs35252396. For rs11604, the ratio of allele-specific signals was measured by qPCR (FAM/VIC) from input genomic DNA and cDNA derived from control (untreated) or DMOG-treated cells. No significant allelic imbalance at rs11604 was detected between the two groups (AC/AC or AC/CG at rs35252396) in input genomic DNA or cDNA from untreated cells ( Supplementary Fig. 14). We then calculated the change of allelic ratios of cDNA from DMOG-treated cells compared with the respective control untreated cDNA. We detected a significant greater change in allelic PVT1 expression induced by DMOG in cells from individuals with an AC/CG genotype at rs35252396. Values are mean ± s.d. from four (AC/AC) or eight (AC/CG) individuals. qPCRs for DNA and cDNA were performed in triplicates per individual. *Student's t-test, Po0.05. (b) Genotype expression correlation for rs10111989 and MYC in the KIRC TCGA data. rs10111989 is in LD with rs35252396 (r 2 ¼ 0.33, D 0 ¼ 0.98) and allele C is associated with RCC development (odds ratio 1.16, Po0.05) in data from a meta-analysis of UK and NCI cohorts 54 . *w 2 -test; Po0.05 for higher expression in CC individuals compared with TT individuals. Whiskers extend to 1.5 times of the inter quartile range. (c) Schematic representation of the 8q24.21 RCC enhancer of MYC and PVT1 expression. In cells from renal tubular origin (non-transformed tubular cells or RCC cells), HIFs can bind to the enhancer and regulate MYC and PVT1 expression, but binding is dependent on the rs35252396 genotype that affects accessibility of the site.
DNA and compared with a region just outside of the putative regulatory region. Primer sequences are listed in Supplementary Table 4.  44,52 . Capture efficiency was determined with qPCR relative to a standard curve of genomic DNA before sequencing.
DNA extraction. DNA was isolated using DNA cell lysis buffer (NaCl 100 mM, Tris pH 8.0 10 mM, EDTA 25 mM, SDS 0.5%, Proteinase K 0.1 mg ml À 1 ) for 1 h at 45°C. After proteinase K digest, the isolation of DNA was performed with phenol-chloroform and salt precipitation. DNA content was measured by NanoDrop (Peqlab).
High-throughput sequencing. ChIP-seq library preparations were carried out using Illumina protocols and libraries were sequenced on a HiSeq 2000 platform (Illumina). Sequences were mapped to NCBI build 37 (hg19) using BWA and peaks were called with MACS as previously described 53 . Capture-C libraries were sequenced on the HiSeq 4000 (Illumina). Capture-C data were analysed as previously described 52 . In brief, reads were trimmed, in silico digested for DpnII and aligned to the GRCh37 (hg19) with Bowtie 1.0 and interaction frequencies were determined using CCanalyser2.pl (ref. 44). Significant interactions were called using a background model of distance-dependent decay from the capture site, interaction frequencies above the background level were analysed for significance 52 45 . The guide has a quality score of 93. The CRISPR-Cas9 plasmid was cloned following the manufacturer's protocol. A total of 2 Â 10 7 cells were transfected by electroporation with 3 mg vector. Clones of cells were generated by dilution. For mutation screens, genomic DNA of single-cell clones was isolated and the CRISPR/Cas9 target region was amplified by PCR. Products were resolved by polyacrylamide gel electrophoresis. Genomic DNA of clones of cells with putative indel mutations was PCR amplified and cloned into pGL3 vector (Promega) and subjected to Sanger sequencing. The top five potential off-targets in DNA regions were tested for the off-target effects by PCR amplification and polyacrylamide gel electrophoresis analysis. Expression of all genes with potential off-target sites in the coding regions was tested by qPCR in the clones. No off-target effects were detected. From 37 clones of cells, we identified 7 clones of cells with indel mutations at the HRE and chose 10 non-altered clones as controls.
Allele-specific assays. To identify samples heterozygous for the common and rare alleles at rs35252396 and rs11604, DNA from cell lines and primary tubular cells was genotyped using customized Taqman assays (Life Technologies). All primers, probes and conditions used are available on request. For allele-specific assays DNA from FAIRE and ChIP experiments as well as cDNA from primary renal tubular cell cultures was used. Genomic DNA from untreated samples from the same experiments was used in serial dilutions as an input control. Homozygous cell lines for both alleles were used as positive controls (rs35252396: AC/AC-Caki1 cells, CG/CG-786-O cells; rs11604: T/T-RCC L15, C/C-RCC L13) in all allele-specific assays. Data were analysed using the TaqMan Genotyper Software V1.3 (Life Technologies). For both assays, the mean ratio of minor allele/major allele (FAM/VIC) for the input DNA was arbitrarily set to 1 and the ratios of DNA from assays (FAIRE; ChIP or cDNA) were normalized to input DNA ratios. For the intragenic SNP, rs11604 allelic ratios of cDNA from DMOG-treated primary tubular cells were compared with the allelic ratio of the corresponding untreated control cells.
Tumour samples. A total of 453 renal tumours and corresponding normal renal tissue were collected from the archives of the Department of Pathology, University of Erlangen-Nürnberg, for tissue microarray construction. The tumour collection consists in part of older samples (before 2008) that were collected anonymously. The local Ethical Committee has approved the use of these samples for this study without informed consent. For more recent samples, informed consent was provided by the patients. The entire study has been approved by the local Ethical Committee at the University of Erlangen-Nürnberg and specimens were collected in accordance with the World Medical Association Declaration of Helsinki. Details on tissue microarray (TMA) composition and tumour characteristics have been published previously 55 . In short, archival FFPE tissues were reclassified according to the 2004 World Health Organization classification of renal tumours and the 2002 tumour node metastasis (TNM) classification by two uropathologists. One representative punch from each tumour and from corresponding normal tissue was transferred to a new block for TMA construction.
Immunohistochemistry. Immunostainings were conducted on paraffin-embedded tissue arrays as described earlier 3 . Antibodies were anti-MYC (1:200, rabbit monoclonal (Y69), Abcam, ab32072), anti-HIF-1a (1:10,000, rabbit polyclonal, Cayman Chemicals, Cay10006421) and anti-HIF-2a (1:10,000, rabbit polyclonal, PM8), and were applied after an antigen retrieval procedure (Dako). Stainings were analysed by two researchers blinded for tumour phenotype and results of other stainings. Stainings were scored according to intensity of nuclear staining (0-4) and per cent of positive cells (0-100%) using an immunoreactive score (IRS) according to Remmele and Stegner 56 . HIF and MYC stainings were then divided into four categories: no (IRS ¼ 0), low (IRS 1-3), medium (IRS 4-8) or high (9-12) levels of immunoreactivity. A small proportion of samples from the TMA (o5%) could not be analysed due to the lack of tissue or due to the presence of normal kidney tissue at the respective position on the TMA. Software Inc). IRSs and eQTL analyses were evaluated using the w 2 -test (IBM SPSS Statistics 21).