Lindau tumor suppressor protein regulates gene expression ... - Nature

Oncogene (2012) 31, 776–786

& 2012 Macmillan Publishers Limited All rights reserved 0950-9232/12 www.nature.com/onc

ORIGINAL ARTICLE

The von Hippel–Lindau tumor suppressor protein regulates gene expression and tumor growth through histone demethylase JARID1C X Niu1, T Zhang1, L Liao1, L Zhou1, DJ Lindner2, M Zhou3, B Rini2, Q Yan4 and H Yang1 1 Department of Cancer Biology, Lerner Research Institute, Cleveland Clinic, Cleveland, OH, USA; 2Taussig Cancer Center, Cleveland Clinic, Cleveland, OH, USA; 3Department of Pathology, Cleveland Clinic, Cleveland, OH, USA and 4Department of Pathology, Yale University School of Medicine, New Haven, CT, USA

In clear-cell renal cell carcinoma (ccRCC), inactivation of the tumor suppressor von Hippel–Lindau (VHL) occurs in the majority of the tumors and is causal for the pathogenesis of ccRCC. Recently, a large-scale genomic sequencing study of ccRCC tumors revealed that enzymes that regulate histone H3 lysine 4 trimethylation (H3K4Me3), such as JARID1C/KDM5C/SMCX and MLL2, were mutated in ccRCC tumors, suggesting that H3K4Me3 might have an important role in regulating gene expression and tumorigenesis. In this study we report that in VHL-deficient ccRCC cells, the overall H3K4Me3 levels were significantly lower than that of VHL þ / þ counterparts. Furthermore, this was hypoxia-inducible factor (HIF) dependent, as depletion of HIF subunits by small hairpin RNA in VHL-deficient ccRCC cells restored H3K4Me3 levels. In addition, we demonstrated that only loss of JARID1C, not JARID1A or JARID1B, abolished the difference of H3K4Me3 levels between VHL/ and VHL þ / þ cells, and JARID1C displayed HIF-dependent expression pattern. JARID1C in VHL/ cells was responsible for the suppression of HIF-responsive genes insulin-like growth factor-binding protein 3 (IGFBP3), DNAJC12, COL6A1, growth and differentiation factor 15 (GDF15) and density-enhanced phosphatase 1. Consistent with these findings, the H3K4Me3 levels at the promoters of IGFBP3, DNAJC12, COL6A1 and GDF15 were lower in VHL/ cells than in VHL þ / þ cells, and the differences disappeared after JARID1C depletion. Although HIF2a is an oncogene in ccRCC, some of its targets might have tumor suppressive activity. Consistent with this, knockdown of JARID1C in 786-O VHL/ ccRCC cells significantly enhanced tumor growth in a xenograft model, suggesting that JARID1C is tumor suppressive and its mutations are tumor promoting in ccRCC. Thus, VHL inactivation decreases H3K4Me3 levels through JARID1C, which alters gene expression and suppresses tumor growth.

Correspondence: Dr H Yang, Department of Cancer Biology, NB-40, Lerner Research Institute, Cleveland Clinic, 9500 Euclid Avenue, Cleveland, OH 44195, USA. E-mail: [email protected] Received 16 December 2010; revised 19 May 2011; accepted 23 May 2011; published online 4 July 2011

Oncogene (2012) 31, 776–786; doi:10.1038/onc.2011.266; published online 4 July 2011 Keywords: von Hippel–Lindau; hypoxia-inducible factor; JARID1C/KDM5C/SMCX; trimethyl H3K4; gene expression

Introduction Inactivation of the von Hippel–Lindau (VHL) tumor suppressor gene has a causal role in the pathogenesis of clear-cell renal cell carcinoma (ccRCC). About 75% of the RCCs are characterized as ccRCC, and among them B70% are sporadic tumors that harbor biallelic inactivation of VHL through mutation, deletion or hypermethylation of the promoter (Kaelin, 2002; Linehan et al., 2004). The protein product of the VHL tumor suppressor gene, pVHL, is best known as the substrate recognition unit of an E3 ubiquitin ligase complex that contains Cul2, Elongin B and C and Rbx1 (Kamura et al., 1999). This complex targets the a subunits of the heterodimeric transcription factor hypoxia-inducible factor (HIF) for ubiquitylation and proteasome-mediated degradation (Ohh et al., 2000). pVHL also regulates other HIF-independent biological processes such as inhibition of NF-kB activity (Yang et al., 2007), maintenance of chromosome stability (Thoma et al., 2009), promoting cilia production (Schraml et al., 2009) and stabilization of RNA polymerase II subunit 1 (Mikhaylova et al., 2008). The HIF transcription factor contains two subunits: the oxygen-sensitive a subunits (HIF1a, HIF2a and HIF3a) and the constitutively expressed HIF1b subunit (also called the arylhydrocarbon nuclear translocator (ARNT)) (Semenza, 2007). The interaction between pVHL and HIFa requires the hydroxylation on either of two prolyl residues by members of the egl nine homolog (EglN) family (also called prolyl hydoxylase domain containing or HIF prolyl hydroxylases) (Epstein et al., 2001; Ivan et al., 2001, 2002; Jaakkola et al., 2001). These enzymes require molecular oxygen, Fe(II) and 2-oxoglutarate for activity. Under normal oxygen conditions (normoxia), the HIFa subunits are prolyl hydroxylated, ubiquitylated and destroyed. When the oxygen level is low (hypoxia), the HIFa subunits fail to

pVHL controls H3K4Me3 and tumor growth via JARID1C X Niu et al

777

be prolyl hydroxylated, escape recognition by pVHL and the subsequent degradation, and heterodimerize with HIF1b. The heterodimers enter the nucleus, recruit transcriptional co-activator complexes (Arany et al., 1996; Ema et al., 1999) and regulate the expression of target genes by binding to the hypoxia-response element (Semenza, 2003). Activation of HIF leads to a metabolic shift to anerobic glycolysis, increased secretion of proangiogenesis factors, remodeling of the extracellular matrix, and resistance to apoptosis and increased mobility. In VHL-defective ccRCC tumors, enhanced angiogenesis and constitutive activation of the HIF pathway are prominent features. In the preclinical models of ccRCC, HIF was both sufficient (Kondo et al., 2002) and necessary for tumor growth (Kondo et al., 2003; Zimmer et al., 2004). Clinically, drugs that block the activities of the receptors for vascular endothelial growth factor (VEGF), a critical HIF target gene, produced clear and positive clinical results in treating kidney cancer (Rini et al., 2005). Modification on the tails of histones changes the conformations of chromatin and alters the accessibility of transcription factors and co-activators. Large-scale studies have revealed the complex relationship between histone modification and gene expression (Barski et al., 2007). Generally, high levels of histone acetylation and histone H3 lysine 4 (H3K4) methylation at the promoter regions and histone H3 lysine 36 (H3K36) methylation to the transcribed region are associated with active genes, while high levels of histone H3 lysine 27 (H3K27) trimethylation at the promoters and the genic regions correlate with gene repression (Barski et al., 2007). Multiple reports indicate that HIF binds to the promoters and increases the transcription of several jumonji-domain-containing histone demethylases (JMJD1A and JMJD2B) (Beyer et al., 2008; Pollard et al., 2008; Wellmann et al., 2008; Xia et al., 2009; Krieg et al., 2010), enzymes that change epigenetic marks that in turn modulate gene expression. JMJD1A demethylates dimethylated histone H3 lysine 9 (H3K9Me2), while JMJD2B demethylates trimethylated H3K9, and both are upregulated in VHL-defected RCC cells and by hypoxia in many other cell types. Under hypoxia, they increase the expression of adrenomedulin (ADM) and growth and differentiation factor 15 (GDF15) by reducing H3K9 dimethylation on their promoters. Thus JMJD1A and JMJD2B function as signal amplifier of transcriptional response to hypoxia and aid tumor growth. Trimethyl H3K4 (H3K4Me3) at promoters are very tightly linked to active transcription, and 91% of all the RNA polymerase II binding sites on DNA are correlated with H3K4Me3 (Barski et al., 2007). Recently, mutations in genes encoding histone-modifying enzymes such as JARID1C/KDM5C/SMCX, an H3K4Me3 demethylase, were identified in ccRCC (Dalgliesh et al., 2010). Its family members include RBP2/KDM5A, PLU-1/KDM5B and SMCY/KDM5D (Klose et al., 2007). JARID1C is originally identified as an X-linked mental retardation gene that demethylates H3K4Me3 (Iwase et al., 2007), and it binds to Re-1

silencing transcription factor (REST) and represses a subset of REST-target genes (Tahiliani et al., 2007). Furthermore, the region containing JARID1C displays intrinsic escape from X chromosome inactivation (Li and Carrel, 2008). The mutations in JARID1C show selection for truncations in VHL-defective RCC tumors. Gene-expression analysis indicates that JARID1C mutations change expression in 18 genes including the metallothioneins (Dalgliesh et al., 2010). However, it is not known exactly how JARID1C impact on transcription and tumor biology. In this study, we discovered that VHL-defective RCC cells had lower overall H3K4Me3 levels than their VHL þ / þ counterparts. We showed that this was dependent on constitutively active HIF2a and JARID1C. In VHL/ RCC cells, HIF induced JARID1C expression, which in turn changed the expression of hypoxia-responsive genes (HRGs) and reduced the H3K4Me3 levels at the promoters of insulin-like growth factor-binding protein 3 (IGFBP3), COL6A1, DNAJC12 and GDF15. Finally, we found that knockdown of JARID1C greatly enhanced tumor growth, suggesting that JARID1C was tumor suppressive in kidney cancer.

Results VHL-defective kidney cancer cells had lower overall H3K4Me3 than their VHL þ / þ counterparts In VHL/ RCC cells HIF is constitutively active (Kaelin, 2002). JMJD1A and JMJD2B, two histone demethylases for H3K9, are induced by HIF and affect tumor growth (Krieg et al., 2010). JARID1B/PLU-1, a H3K4Me3 demethylase, is also induced by HIF (Xia et al., 2009). However, it is not known how VHL status affects the H3K4Me3 levels. We compared the H3K4Me3 levels in VHL þ / þ and VHL/ RCC cells because of its importance to transcription and mutations in genes encoding H3K4Me3-modifying enzymes JARID1C and MLL2 (Dalgliesh et al., 2010). The overall H3K4Me3 level in 786-O VHL/ cells was significantly lower than that in VHL þ / þ cells (Figure 1a). In contrast, the H3K4Me2 level showed little difference and the H3K4Me1 level was higher in VHL/ cells. To confirm that the overall difference in the H3K4Me3 levels was not limited to just one pair of RCC cells, we examined H3K4Me3 in another isogenic pair of RCC cells: RCC4 VHL þ / þ and VHL/. Again the overall H3K4Me3 level in RCC4 VHL/ cells was significantly lower than that of VHL þ / þ cells (Figure 1b). The lower H3K4Me3 levels in VHL/ RCC cells were HIF dependent As VHL/ RCC cells had constitutively high HIF activity, we addressed whether the decreased H3K4Me3 levels in VHL/ cells were HIF dependent. First we identified five lentiviral small hairpin RNA (shRNA) constructs that could suppress HIF2a (the only HIFa Oncogene


778

Figure 1 VHL-defective kidney cancer cells had reduced overall H3K4Me3 levels than their VHL þ / þ counterparts. (a) Total cell lysates of human renal carcinoma 786-O cells stably transfected to produce wild-type HA-VHL (786-O VHL þ / þ ) or with an empty plasmid (786-O VHL/) were prepared and immunoblotted with the indicated antibodies. (b) The same experiment in a was repeated with human renal carcinoma RCC4 cell lines with or without HA-VHL. The H3K4Me3/H3 ratios were calculated with band intensities measured with NIH imageJ software.

expressed in 786-O cells) (Maxwell et al., 1999) that led to lower expression of the HIF target gene glucose transporter type 1 (GLUT1) (Figure 2a). The 786-O cells were transfected with pCDNA3 vectors expressing wild-type HA-VHL (VHL þ / þ ) and empty vector (VHL/) and selected with G418. 786-O VHL þ / þ and VHL/ cells were then infected with shRNA lentivirus expressing an oligo that did not target any known gene (SCR), HIF2a-566, and HIF2a-1631 then selected with puromycin. As expected, 786-O VHL/ cells had much less H3K4Me3 than VHL þ / þ cells. HIF2a depletion by both constructs did not affect H3K4Me3 level in VHL þ / þ cells, but completely reversed its decrease in VHL/ cells (Figure 2b). Interestingly, very little difference was seen on H3K4Me2 levels, while the H3K4Me1 levels showed a reverse trend of H3K4Me3: higher in VHL/ cells than in VHL þ / þ cells and decreased after HIF2a depletion (Figure 2b). To examine whether the lower H3K4Me3 levels in RCC4 VHL/ cells were also HIF dependent, we suppressed the expression of HIF1b in RCC4 VHL þ / þ and VHL/ cells. As suppression of HIF1b expression would disable the transcriptional activity of HIF (Semenza et al., 1997), we infected both cell lines with SCR and HIF1b-1770 virus. Clearly, HIF1b depletion abolished HIF transcriptional activity as it suppressed the expression of HIF target gene GLUT1 in RCC4 VHL/ cells (Figure 2c). HIF1b suppression also reversed the H3K4Me3 decrease in RCC4 VHL/ cells (Figure 2c). As RCC4 cells express both HIF1a and HIF2a, we examined the contribution of each gene to H3K4me3 level. Depletion of HIF2a in RCC4 VHL/ cells by two different shRNA constructs suppressed the expression of GLUT1 and restored H3K4Me3 levels (Figure 2c). Two shRNA Oncogene

Figure 2 HIF2a was required for the reduced H3K4Me3 levels in 786-O VHL-defective cells. (a) Immunoblot analysis of 786-O cells stably infected with the indicated lentiviral shRNA constructs and selected with puromycin. PLKO-SCR: control. The rest were constructs targeting HIF2a. (b) 786-O VHL þ / þ and VHL/ cells expressing indicated shRNA constructs were analyzed with the indicated antibodies. (c) RCC4 VHL þ / þ and VHL/ cells expressing indicated shRNA constructs were analyzed with the indicated antibodies.

constructs against HIF1a failed to do so, although they efficiently suppressed HIF1a expression (Figure 2c). We will study the effect of H3K4Me3 levels on RCC tumor growth in a xenograft model. As 786-O cells form tumors much more efficiently than RCC4 cells in nude mice, we decided to focus our further analysis on 786-O cells. In VHL/ RCC cells HIF-dependent increase of JARID1C caused the decrease of overall H3K4Me3 levels The lower H3K4Me3 levels in VHL/ cells could be due to reduced methyltransferase activity or increased demethylase activity or both. As this reduction is HIF dependent and HIF is known to increase the transcription of histone demethylases, we examined the protein expression of JARID1A/RBP2/KDM5A, JARID1B/PLU-1/KDM5B, JARID1C/SMCX/KDM5C, and JARID1D/SMCY/KDM5D. The JARID1D expression was not detectable (data not shown). No difference in protein expression on RBP2 or PLU-1 was found between 786-O VHL þ / þ and VHL/ cells (Figures 2b, 3a and b), while a reproducible increase of JARID1C protein in VHL/ cells was observed (Figures 2b, 3a and b). Moreover, this increase was reversed after HIF2a depletion in VHL/ cells (Figure 2b).


779

Figure 3 JARID1C was responsible for the different levels of H3K4Me3 in 786-O VHL þ / þ and VHL/ cells. (a) 786-O VHL þ / þ and VHL/ cells expressing indicated shRNA constructs were analyzed with the indicated antibodies. (b) 786-O VHL þ / þ and VHL/ cells with or without shRNA against JARID1C were fractionated into cytoplasmic and nuclear fractions. The lysates were blotted with indicated antibodies. *Indicates background bands.

To study the contribution of each demethylase to the different H3K4Me3 levels in 786-O VHL þ / þ and VHL/ cells, we individually suppressed RBP2, PLU-1 or JARID1C expression by shRNA. These two cell lines were also infected with a lentivral virus expressing SCR as controls. The shRNA constructs against RBP2 or PLU-1 successfully depleted the corresponding protein without restoring the overall H3K4Me3 levels in 786-O VHL/ cells (Figure 3a). In contrast, the depletion of JARID1C by three different shRNA constructs all abolished the difference of H3K4Me3 levels between 786-O VHL þ / þ and VHL/ cells (Figure 3a). To more accurately compare the protein abundance of JARID1C, we isolated cytoplasmic and nuclear fractions from VHL þ / þ and VHL/ cells and cells with JARID1C depleted. The successful partitioning of these cellular compartments was

confirmed by immunoblots of the IKKa (mostly cytoplasmic (Huang et al., 2003)) and TopoIIb (a nuclear protein (Berrios et al., 1985)). RBP2 and PLU1 were strictly nuclear and showed no difference. In contrast, JARID1C was both cytoplasmic and nuclear, and VHL/ cells had more JARID1C protein than VHL þ / þ cells in both compartments (Figure 3b). To examine whether there is HIF-dependent induction of H3K4Me3 demethylases mRNA, we compared their expressions in 786-O VHL þ / þ , VHL/ and VHL/ cells with endogenous HIF2a depleted (HIF2a-1631 or HIF2a-566). Quantitative real-time PCR (qRT–PCR) result suggested that the mRNA expression of JARID1C, but not RBP2 or PLU-1, showed a HIF-dependent increase in VHL/ cells (Figures 4a and 5a). The increase (about twofold) was consistent with the increased protein levels (Figures 2b, 3a and b). To further investigate whether JARID1C protein is a HIF target gene, we used Ren-01 RCC cell line that was newly derived at Cleveland Clinic (Negrotto et al., 2011). Ren-01 cells were infected with lentiviral vectors expressing SCR or HIF1b-1770, which very effectively depleted HIF1b (Figure 4b). Both cells were either untreated or treated with hypoxia mimetic DFO or cobalt chloride (CoCl2), two chemicals that abolished the degradation of HIFa subunits by inhibiting the activity of EGLNs (Yan et al., 2007; Pollard et al., 2008). Both hypoxia mimetics induced the expression of HIF1a protein and HIF target gene GLUT1 as expected in Ren-01 SCR cells, while in Ren-01 HIF1b-1770 cells the GLUT1 protein had much lower basal expression and failed to be efficiently induced by hypoxia mimetics. The expression of JARID1C protein behaved like that of GLUT1, suggesting that HIF controlled JARID1C expression. Less HIF1a proteins was induced by hypoxia mimetics in Ren-01 HIF1b-1770 cells, probably due to lower stability in the absence of its binding partner HIF1b (Isaacs et al., 2004). As we found that HIF2a was the major regulator of JARID1C expression in some RCC cell lines (Figures 2b and c), we examined whether this was also true in Ren01 cells. Suppression of HIF1a expression by two shRNA constructs significantly reduced the induction of both GLUT1 and JARID1C by DFO in Ren-01 cells, while HIF2a suppression had a much smaller effect (Figure 4c). This suggests that HIF1a could also contribute significantly to the induction of JARID1C in some RCC cell lines. As VHL loss or treatment by hypoxia mimetics affects the functions of proteins other than HIFa, we investigated whether HIF activation alone in VHL þ / þ RCC cells was capable of inducing JARID1C expression and reducing the overall H3K4Me3 level. 786-O VHL þ / þ cells were infected with empty retrovirus, retrovirus expressing either a stable and transcriptionally active HIF2a mutant (HIF2a-dPA) or a stable but transcriptionally inactive HIF2a mutant (HIF2a-PA-dTA) (described in Yan et al., 2007). Polyclonal pools of cells were selected. As expected, HIF2a-dPA induced the expression of HIF target GLUT1, while HIF2a-PA-dTA Oncogene


780

Figure 4 JARID1C mRNA and protein expressions are HIF dependent. (a) 786-O VHL þ / þ , 786-O VHL/ and 786-O VHL/ cells expressing HIF2a-1631 or HIF2a-566 were cultured at standard condition, and total RNA was extracted from these cells. The relative abundances of RBP2/JARID1A, PLU-1/JARID1B and JARID1C were measured with qRT–PCR. The expression levels of these genes were normalized against actin. (b) Ren-01 cells stably infected with SCR or HIF1b-1770 were untreated or treated with hypoxia mimetics DFO or CoCl2. The lysates were immunobloted with the indicated antibodies. (c) Ren-01 cells stably infected with indicated shRNA constructs were untreated or treated with hypoxia mimetic DFO. The lysates were immunobloted with the indicated antibodies. (d) 786-O VHL þ / þ cells stably expressing an empty vector, a stable and functional HIF2a mutant (HIF2a-dPA), and a stable but non-functional HIF2a mutant (HIF2a-dPA-dTA) were analyzed with indicated antibodies.

was ineffective (Figure 4d). We found that HIF2a-dPA also induced the expression of JARID1C and suppressed the H3K4Me3 level, while HIF2a-PA-dTA failed to do so (Figure 4d), suggesting that activated HIF2a alone was sufficient to reduce H3K4Me3 level through JARID1C in VHL þ / þ RCC cells. Similar result was observed in ACHN RCC cells with a wildtype VHL (Supplementary Figure S1). It was reported that the yeast histone H3K4Me3 demethylase Jhd2, a JARID1C homolog, was polyubiquitylated by the E3 ubiquitin ligase Not4 and degraded by the proteasome (Mersman et al., 2009). As pVHL is also a part of an E3 ubiquitin ligase complex, it was possible that JARID1C was a target of pVHL and this contributed to the higher abundance of JARID1C in VHL/ cells. To test this, we blocked the protein synthesis with cycloheximide (CHX) in 786-O VHL þ / þ Oncogene

and VHL/ cells before examining how fast the JARID1C protein disappeared. JARID1C was a moderately unstable protein but it did not exhibit higher stability in VHL/ cells than that in VHL þ / þ cells (Supplementary Figure S2a). In addition, disruption of proteasome function by MG132 induced accumulation of poly-ubiquitylated proteins in the whole-cell extracts without significantly increasing the protein levels of JARID1C in VHL þ / þ cells or VHL/ cells (Supplementary Figure S2b). These results suggested that the stability of JARID1C was not affected by pVHL in RCC cells. HIF2a alone was capable of increasing JARID1C expression and reducing H3K4Me3 level (Figure 4d and Supplementary Figure S1). As transcription regulators often physically interact, we examined whether they could bind to each other and whether the enzymatic


781

activity of JARID1C was affected by VHL status. We found that JARID1C and HIF2a associated with each other (Supplementary Figure S3). This association might increase the enzymatic activity of JARID1C in VHL/ cells to contribute to the lower overall H3K4Me3 levels. However, the semiquantitative in vitro essay revealed that JARID1C purified from 786-O VHL þ / þ cells was at least as active as JARID1C purified from VHL/ cells in demethylating H3K4Me3 (Supplementary Figure S4), suggesting that HIF2a did not significantly change the enzymatic activity of JARID1C. Lower H3K4Me3 levels in VHL/ RCC cells affected the transcription of hypoxia-responsive genes (HRG) As VHL þ / þ and VHL/ RCC cells had different overall levels of H3K4Me3, some of the HRGs’ expression might be affected by JARID1C. We used qRT–PCR to compare the expression of some well-

characterized HRGs in VHL þ / þ and VHL/ cells with or without JARID1C expression silenced. JARID1C expressions were efficiently suppressed by two shRNA constructs (Figure 5a). The expression of IGFBP3, a HIF target in RCC cells (Nakamura et al., 2006), was elevated in VHL/ cells after JARID1C loss, while the expressions of cyclin D1 (CCND1), VEGF, or GLUT1 did not show significant difference. Western blots were used to corroborate the mRNA results. Higher levels of HIF2a were present in VHL/ cells and were not changed by JARID1C depletion (Figure 5b), so it was unlikely the cause of the increased expression of IGFBP3. Consistent with the mRNA results, the protein levels of IGFBP3, but not GLUT1 or Cyclin D1, was elevated in VHL/ cells after JARID1C suppression. Chromatin immunoprecipitation (ChIP) assay further confirmed that the H3K4Me3 level at the promoter of IGFBP3 was lower in VHL/ cells than that in VHL þ / þ cells, and the difference disappeared after JARID1C depletion

Figure 5 JARID1C suppressed the expression of HIF-responsive gene IGFBP3 in RCC cells. (a) The relative abundances of JARID1C, IGFBP3, CCND1, VEGF and GLUT1 were measured with qRT–PCR in 786-O VHL þ / þ and VHL/ cells with or without shRNA constructs against JARID1C. The expression levels of these genes were normalized against actin. (b) The whole-cell lysates of 786-O VHL þ / þ and VHL/ cells with or without shRNA constructs against JARID1C (JARID1C-1056 and JARID1C2767) were extracted with 1% SDS and immunoblotted with indicated antibodies. (c) ChIP was used to examine the H3K4Me3 levels at the promoter regions of different genes. 786-O VHL þ / þ and VHL/ cells with or without shRNA constructs against JARID1C (JARID1C-2767 (top), JARID1C-1056 (bottom)) were used for this assay. The amount of DNA bound to H3K4Me3 was visualized with semiquantitative PCR. One percent of the initial materials used for ChIP were amplified by PCR as the input. (d) The relative abundances of DEP-1, GDF15, DNAJC12, COL6A1, CASP1 and SGK2 were measured with qRT–PCR in 786-O VHL þ / þ and VHL/ cells with or without shRNA constructs against JARID1C. The expression levels of these genes were normalized against actin. (e) ChIP assay as described in c was used to examine the H3K4Me3 levels at the promoter regions of COL6A1, DNAJC12, GDF15 and GAPDH. 786-O VHL þ / þ and VHL/ cells with or without shRNA constructs against JARID1C (JARID1C-2767 (top) and JARID1C-1056 (bottom)) were used for this assay. Oncogene


782

(Figure 5c). Consistent with the mRNA and protein results, the H3K4Me3 levels on Cyclin D1, GLUT1 and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) promoters showed no difference between VHL þ / þ and VHL/ cells. As VHL/ RCC cells had lower overall levels of H3K4Me3, JARID1C might be repressing some of the HIF-repressed genes. We examined the effect of JARID1C suppression on some of these genes identified previously (Yan, unpublished data). We found that JARID1C was responsible for suppressing the expression of density-enhanced phosphatase 1 (DEP-1), GDF15, DnaJ (Hsp40) homolog subfamily c member 12 (DNAJC12) and collagen type VI a 1 (COL6A1), but not for the lower expression of caspase 1 (CASP1) or serum/glucocorticoid-regulated kinase 2 (SGK2) in VHL/ cells (Figure 5d). As their expressions were all suppressed by a stable and active HIF2a under normoxia in 786-O VHL þ / þ cells (Yan, unpublished data), they were HIF-repressed genes. We further tested the H3K4Me3 levels at the promoters of DNAJC12, COL6A1 or GDF15 with ChIP assay. The result showed that they were lower in VHL/ cells than these in VHL þ / þ cells, and the difference disappeared after JARID1C depletion (Figure 5e).

JARID1C was tumor suppressing in VHL/ RCC cells Inactivating mutations of JARID1C were identified in VHL-defective RCC tumors (Dalgliesh et al., 2010). To investigate whether this would promote or suppress tumor growth, we compared the growth rates of 786-O RCC cells with or without shRNA against JARID1C in vitro and in a xenograft study. As reported, expression of pVHL did not affect the 786-O cells growth under standard culturing condition with serum (Iliopoulos et al., 1995), and JARID1C suppression did not cause any obvious change in growth rates either (Figure 6a). As the growth conditions of in vitro cell culture differs significantly from those of the in vivo environment, the contribution of many genes of interest to tumorigenesis, such as VHL and HIF2a, can only be truthfully evaluated by xenograft models (Iliopoulos et al., 1995; Kondo et al., 2002, 2003). Following subcutaneous injection into nude mice, 786-O VHL/ cells with JARID1C suppressed formed much larger tumors than the cells expressing SCR (Figures 6b and c). The tumors with or without JARID1C suppression were very similar histologically (data not shown). Therefore, JARID1C was likely a HIF target gene, decreased the overall H3K4Me3 and suppressed gene transcription, and retarded tumor growth (Figure 6d).

Figure 6 Loss of JARID1C enhanced tumor growth of VHL-defective RCC cells. (a) 786-O VHL þ / þ and VHL/ cells expressing PLKO-SCR or JARID1C-2767 were grown in a 96-well plates in the presence of serum. Viable cell number at the indicated time points after seeding was determined by XTT assay. (b) 786-O VHL/ cells infected to produce JARID1C-2767 or with the PLKO-SCR were injected subcutaneously in the flanks of nude mice. Approximately 9 weeks later, tumors were excised and weighed. Eight tumors per line were analyzed. Error bars ¼ s.e.m. Pp0.001 according to a Mann–Whitney U statistic analysis. (c) Representative photographs of nude mice and tumors analyzed in b. Left: tumor from cells expressing PLKO-SCR; right: tumor from cells expressing JARID1C-2767. (d) Model of how pVHL regulates JARID1C through HIF to modulate overall H3K4Me3, gene expression and tumor growth. Oncogene


783

Discussion Inactivation of VHL is the dominant functional mutation in ccRCC tumors. Recently, mutations in enzymes modifying histones were also identified in ccRCC— SetD2, a histone H3 lysine 36 methyltransferase; JARID1C, an H3K4Me3 demethylase; MLL2, an H3K4Me3 methyltransferase; and UTX, a histone H3 lysine 27 demethylase (Dalgliesh et al., 2010). Little was known on how the mutations of these histone modifiers would impact on gene transcriptions and tumorigenesis. In this study, we found that the overall H3K4Me3 levels in VHL/ RCC cells were significantly lower than that of their VHL þ / þ counterparts. This difference was both HIF and JARID1C dependent, and there was a HIF-dependent increase of JARID1C mRNA and protein levels. The increased JARID1C activity in VHL/ RCC cells was suppressive for the expression of IGFBP3 and a few other HIF-repressed genes, which correlated with the decreased H3K4Me3 levels on their promoters. Downregulation of JARID1C in 786-O VHL/ RCC cells promoted tumor growth, providing the first in vivo evidence that JARID1C is tumor suppressive, and its mutations in ccRCC might lead to higher tumor burden. It was widely reported that multiple histone demethylases, including JMJD1A and JMJD2B, were direct HIF target genes and induced by hypoxia (Beyer et al., 2008; Pollard et al., 2008; Wellmann et al., 2008; Xia et al., 2009; Krieg et al., 2010). Hypoxia induces HIF1 activity, which in turn inhibits mitochondrial biogenesis (Zhang et al., 2007). This would lead to reduced production of a-ketoglutarate, an essential cofactor for histone demethylases. It was hypothesized that the hypoxic induction of these enzymes would at least partially compensate for the inhibitory effect of hypoxia to maintain histone methylation homeostasis (Xia et al., 2009). Without this increase, hypoxia would inhibit the enzymatic activity of JARID1A protein in Beas-2B lung epithelial cells to significantly increase overall H3K4Me3 (Zhou et al., 2010). In RCC and colon cancer cells, the induction of HRGs, such as ADM and GDF15, was dependent on JMJD1A (Krieg et al., 2010). JARID1B was induced by hypoxia in HepG2, and suppression of hypoxic induction of JARID1B caused the increase of overall H3K4Me3 levels to greater extents (Xia et al., 2009). In this study, we have identified that JARID1C was the dominant H3K4Me3 demethylase in VHL/ ccRCC cells. As the promoter regions of JARID1C contained HIF-binding motifs (data not shown), it is possible that JARID1C is a direct HIF target gene. Thus hypoxia, through HIF, induces histone-modifying enzymes. The combination of transcriptional activity of HIF, the amplifying activity of JMJD1A and the suppressive activity of JARID1 proteins, and potentially other transcription-modifying factors, shaped the landscape of the hypoxia-induced transcription. HIF2a was essential for the JARID1C induction and H3K4Me3 reduction in 786-O and RCC4 VHL/ cells (Figures 2b and c). A stable and functional mutant of

HIF2a alone was sufficient for these events in 786-O VHL þ / þ and ACHN RCC cells (Figure 4d and Supplementary Figure S1). HIF1a did not seem to have a role in JARID1C induction in RCC4 VHL/ cells (Figure 2c). However, it was required for JARID1C induction in Ren-01 cells by DFO. We noticed that HIF1a was not required for GLUT1 expression in RCC4 cells (Figure 2c) but was necessary for GLUT1 expression in Ren-01 cells (Figure 4c), suggesting that HIF1a was very potent in inducing HRGs in Ren-01 cells. Thus both HIF1a and HIF2a were capable of inducing the expression of JARID1C. Which gene has a more important role in JARID1C induction might be cell-context dependent. The yeast homolog of JARID1C, Jhd2, was polyubiquitylated by the E3 ubiquitin ligase Not4 and degraded by the proteasome (Mersman et al., 2009). As pVHL is also a part of an E3 ubiquitin ligase complex, pVHL might promote the turnover of JARID1C like HIFa subunits. However, our results showed that pVHL did not have any obvious effect on JARID1C stability (Supplementary Figure S2). On the other hand, we found that JARID1C and HIF2a bound to each other, at least when overexpressed (Supplementary Figure S3). Although this interaction might not lead to enhanced enzymatic activity of JARID1C purified from VHL/ cells (Supplementary Figure S4), it might still enhance the access of JARID1C to its substrates as HIF has high affinity toward chromatins. Through physical interaction, HIF might recruit JARID1C to decrease the H3K4Me3 levels on the promoters of HIF-repressed genes. Through testing previously identified HRGs, we identified that the expressions of IGFBP3, DNAJC12, COL6A1, GDF15 and DEP-1 were repressed by JARID1C in VHL/ RCC cells. For IGFBP3, HIF was also increasing its expression (Nakamura et al., 2006), so the end result was a combination of these two forces. The H3K4Me3 levels associated with their promoters correlated well with their mRNA expressions. The overall decrease of H3K4Me3 in VHL/ cells was likely to affect more genes than what we identified through candidate approach in this study. A comprehensive genome-wide comparison of gene expressions among VHL þ / þ and VHL/ RCC cells with or without JARID1C suppressed, is necessary to thoroughly survey the impact of JARID1C and H3K4Me3 change. It will also be necessary to perform a genome-wide comparison of H3K4Me3 marks. The global comparisons of H3K4Me3 marks and gene expression would provide valuable insights into how JARID1C and H3K4Me3 regulate gene expression and other cellular phenotypes. Recently, it was reported that increased expression of JARID1A protein was responsible for the drugresistance to EGFR inhibitors in non-small-cell lung cancer cell lines harboring EGFR mutation (Sharma et al., 2010). JARID1A upregulation in drug-resistant non-small-cell lung cancer cells significantly reduced overall H3K4Me3 levels and changed the chromatin state, and knockdown of JARID1A prevented the appearance of drug-resistant cancer cells during conOncogene


784

tinuous drug treatment. In another report, JARID1Bpositive melanoma cells were slow –cycling but JARID1B was required for continued tumor growth, as knockdown of JARID1B led to an initial acceleration of tumor growth followed by the exhaustion of long-term tumor growth (Roesch et al., 2010). Although loss of JARID1C in the 786-O VHL/ RCC cells led to bigger tumors, it remains to be seen how JARID1C would impact on the long-term growth of the tumors and the drug-responsiveness to different therapeutic agents.

Materials and methods Cell culture and treatments 786-O and RCC4 cells with or without retroviral or pCDNA3based wild-type HA-VHL were described previously (Kondo et al., 2003). Ren-01 was developed by Daniel Lindner’s lab. In brief, a 2-mm diameter biopsy sample from a patient with sunitinib- and bevacizumab-resistant metastatic RCC was implanted subcutaneously into the flank of an athymic nu/nu mouse. After two more round of passage in mice, tumor cells were dissociated in vitro and Ren-01 was developed. The VHL status of the cell lines was confirmed by immunoblots. Cell lines were maintained in DMEM medium with glutamine supplemented with 10% fetal bovine serum plus 1% penicillin and streptomycin. For hypoxia mimetic treatment, 200 mM deferoxamine (DFO, an iron chelator) or 20 mM cobalt chloride (CoCl2, which replaces iron at the active site of the prolyl hydroxylases) was added to the cell culture media for 12 h before the cells were harvested for analysis. For cycloheximide (CHX) treatment, 70–80% confluent cells were treated with CHX at 10 mg/ml for the indicated time. Plasmids pBaba-Puro-HA-tagged HIF2a-dPA had proline 405 and 531 mutated to alanines. pBaba-Puro-HA-tagged HIF2a-PA-dTA had proline 405 mutated to alanine, amino-acid residues 24–29 RCRRSK mutated to ACAASA in the basic helix-loop-helix domain that disrupted DNA binding, and deletions of the N-terminal and C-terminal transactivation domains: aminoacid residues 450–572 and 820–870). In the Co-IP experiment, pCDNA3.0-HA-HIF2a-dPA was used. Flag-tagged wild-type JARID1C was constructed by the following procedure: complementary DNA was amplified from a plasmid with primers (forward, 50 -GGAAGCTTATGGAGCCGGGGTCC GACGATT-30 , reverse, 50 -GCGAATTCCAACTGTTGCTG AGGCGGCTGCTG-30 ). The PCR products were digested by the restriction enzymes HindIII and EcoRI, and ligated into the vector of p3xFlag-CMV-10. Western blot analysis Total cellular lysates were prepared with 1% sodium dodecyl sulfate (SDS) and sonicated to break up DNA before being boiled with 5 sample buffer. The relative intensity of protein bands was measured with NIH imageJ software (Wayne Rasband, National Institutes of Health, USA). NE-PER Nuclear and Cytoplasmic Extraction Kit (Thermo Scientific, Rockford, IL, USA, Cat# 78833) was used to fractionate the cells according to the manufacturer’s instruction. In brief, the cells were trypsinized and precipitated before addition of reagents CERI and CERII to extract the cytoplasmic proteins. The remaining pellets were further added NER reagent and Oncogene

vortexed to extract the nuclear proteins. The same volume of lysates was loaded and resolved by SDS–PAGE and analyzed with standard western blot techniques. The blots were developed with Super Signal Pico substrate (Pierce Biotechnology, Rockford, IL,USA) or Immobilon Western substrate (Millipore, Billerica, MA, USA). Antibodies against H3K4Me3 (17-614) and H3k4Me2 (07-030) are from Millipore, antibodies against H3K4Me1 (ab8895) and histone H3 were from Abcam (Cambridge, MA, USA). Antibody against RBP2/JARID1A was described previously (Klose et al., 2007). Antibodies against PLU-1/JARID1B (A301-813A), JARID1C (A301-035A) and EGLN3 (A300-327A) were from Bethyl Laboratories (Montgomery, TX, USA). Anti-GLUT1 antibody (NB300-666) and anti-HIF2a (NB100-132) were purchased from Novus Biologicals (Littleton, CO, USA). AntiIGFBP3 (AF-675) was from R&D Systems (Minneapolis, MN, USA). Antibodies against HIF1a (610959) and HIF1b (611079) were both purchased from BD transduction laboratories (Sparks, MD, USA). Antibodies against IKKa (sc7218), TopoIIb (sc-25330), PARP1 (sc-7150), anti-HA-epitope (sc-7392), anti-Ub (P4D1) (sc-8017) and vinculin (sc-73614) were from Santa Cruz Biotechnology (Santa Cruz, CA, USA), and anti-cyclin D1 antibody (#2926) was from Cell Signaling Technology (Boston, MA, USA). All the antibodies were used at 0.5 to 1 mg/ml for western blots. Short hairpin RNAs (shRNAs) shRNA constructs were obtained from Sigma (St Louis, MO, USA). The sequences (50 –30 ) were as follows: SCR: GCGCG CUUUGUAGGAUUCGTT; HIF1a-1492: CCGCTGGAGA CACAATCATAT; HIF1a-1048: GTGATGAAAGAATTAC CGAAT; HIF2a-1631: CGACCTGAAGATTGAAGTGAT; HIF2a-566: CCATGAGGAGATTCGTGAGAA; RBP23130: CCTTGAAAGAAGCCTTACAAA; PLU-1-5366: CG AGATGGAATTAACAGTCTT; JARID1C-2767: AGTACC TGCGGTATCGGTATA; JARID1C-1056: CAGTGTAACA CACGTCCATTT; JARID1C-643: TCGCAGAGAAATCGG GCATTT; HIF1b-1770: GAGAAGTCAGATGGTTTATTT. Chromatin immunoprecipitation (ChIP) The method was a modified version from Cheung et al. (2010). In brief, cells were crosslinked using 1% formaldehyde for 10 min. The samples was precleared with 50 ml protein A/G agarose slurry (Roche, Indianapolis, IN, USA) in 1 ml ChIP dilution buffer (0.01% SDS, 1.1%, Triton X-100, 1.2 mM EDTA, 16.7 mM Tris–HCl, pH 8.1 and 167 mM NaCl). One percent of the precleared chromatin sample was set aside as input DNA. Antibodies against H3K4Me3 and normal IgG (Upstate, Billerica, MA, USA) were added and rotated at 4 1C for 1 h. A volume of 100 ml of protein A/G agarose slurry was added to each sample and rotated at 4 1C for 2 h. Each sample was washed once with 1 ml low-salt buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris–HCl, pH 8.1 and 150 mM NaCl), followed by 15 min in 1 ml high-salt buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris–HCl, pH 8.1 and 500 mM NaCl), LiCl (0.25 M LiCl, 1% IgepalCA630, 1% deoxycholic acid (sodium salt), 1 mM EDTA and 10 mM Tris, pH 8.1) and twice with TE buffer (10 mM Tris– HCl, 1 mM EDTA, pH 8.0). The protein–DNA complexes were then eluted with 200 ml elution buffer (0.1 M NaHCO3 and 1% SDS), and the protein–DNA complexes were decoupled by incubation at 65 1C overnight in the presence of RNase A and proteinase K. Finally, the DNA was purified with spin columns (QIAquick PCR purification kit, Qiagen, Valencia, CA, USA). A volume of 100 ml per sample was used to elute DNA, and 4 ml per reaction was used for quantification by


785 PCR. A volume of 20 ml of input DNA skipped the immunoprecipitation and was processed the same way as the ChIP samples. The DNAs were amplified for 32 cycles with the exception of IGFBP3 (29 cycles), and were visualized on agarose gels. The PCR primers used were listed in Supplementary Table 1. Real-time reverse-transcription PCR Total RNA was extracted from cells using Trizol reagent (Invitrogen, Carlsbad, CA, USA) following the instructions from the manufacturer. RNA concentration was determined by absorbance at 260 nm. First-strand complementary DNA was synthesized using First-strand cDNA Synthesis kit (Origene, Rockville, MD, USA). QRT–PCR was performed using the 7500HT Fast Real-time PCR System (Applied Biosystems, Carlsbad, CA, USA) with RT2 Real-Time ROX PCR Master Mix from SABiosciences (Frederick, MD, USA). Genes were amplified using the primers described in Supplementary Table 1. All quantifications were normalized to b-actin. In vitro proliferation assays In vitro cell proliferation assays were carried out using a Cell Proliferation Kit II (XTT) from Roche Diagnostics (Pleasanton, CA, USA) following the manufacturer’s instructions.

Nude-mice xenograft assays All animal experiments were conducted in accordance with a Cleveland Clinic Institutional Animal Care and Use Committeeapproved protocol. Subcutaneous nude-mice xenograft assays were performed as previously described (Kondo et al., 2002). For each cell line 107 viable cells were injected subcutaneously into the flanks of nude mice. The mice were killed at 8 to 10 weeks after injection, and tumors were excised and weighed. Results are presented as mean±s.e. of the mean. Eight pairs of tumors were compared. Results were evaluated statistically using Mann–Whitney U statistic analysis from SigmaPlot. Conflict of interest The authors declare no conflict of interest. Acknowledgements The 786-O and RCC4 cell lines were kind gifts from Dr William Kaelin Jr. This work was supported by American Cancer Society Pilot Award (to HY) and the seed fund from Cleveland Clinic to HY. QY is a Breast Cancer Alliance Young Investigator who was also supported by V scholar Award. We thank Dr Donal Luse for his critical reading of the manuscript.

References Arany Z, Huang LE, Eckner R, Bhattacharya S, Jiang C, Goldberg MA et al. (1996). An essential role for p300/CBP in the cellular response to hypoxia. Proc Natl Acad Sci USA 93: 12969–12973. Barski A, Cuddapah S, Cui K, Roh TY, Schones DE, Wang Z et al. (2007). High-resolution profiling of histone methylations in the human genome. Cell 129: 823–837. Berrios M, Osheroff N, Fisher PA. (1985). In situ localization of DNA topoisomerase II, a major polypeptide component of the Drosophila nuclear matrix fraction. Proc Natl Acad Sci USA 82: 4142–4146. Beyer S, Kristensen MM, Jensen KS, Johansen JV, Staller P. (2008). The histone demethylases JMJD1A and JMJD2B are transcriptional targets of hypoxia-inducible factor HIF. J Biol Chem 283: 36542–36552. Cheung I, Shulha HP, Jiang Y, Matevossian A, Wang J, Weng ZP et al. (2010). Developmental regulation and individual differences of neuronal H3K4me3 epigenomes in the prefrontal cortex. Proc Natl Acad Sci USA 107: 8824–8829. Dalgliesh GL, Furge K, Greenman C, Chen LN, Bignell G, Butler A et al. (2010). Systematic sequencing of renal carcinoma reveals inactivation of histone modifying genes. Nature 463: 360–363. Ema M, Hirota K, Mimura J, Abe H, Yodoi J, Sogawa K et al. (1999). Molecular mechanisms of transcription activation by HLF and HIF1alpha in response to hypoxia: their stabilization and redox signal-induced interaction with CBP/p300. EMBO J 18: 1905–1914. Epstein AC, Gleadle JM, McNeill LA, Hewitson KS, O’Rourke J, Mole DR et al. (2001). C. elegans EGL-9 and mammalian homologs define a family of dioxygenases that regulate HIF by prolyl hydroxylation. Cell 107: 43–54. Huang TT, Wuerzberger-Davis SM, Wu ZH, Miyamoto S. (2003). Sequential modification of NEMO/IKKgamma by SUMO-1 and ubiquitin mediates NF-kappaB activation by genotoxic stress. Cell 115: 565–576.

Iliopoulos O, Kibel A, Gray S, Kaelin Jr WG. (1995). Tumour suppression by the human von Hippel-Lindau gene product. Nat Med 1: 822–826. Isaacs JS, Jung YJ, Neckers L. (2004). Aryl hydrocarbon nuclear translocator (ARNT) promotes oxygen-independent stabilization of hypoxia-inducible factor-1 alpha by modulating an Hsp90-dependent regulatory pathway. J Biol Chem 279: 16128–16135. Ivan M, Haberberger T, Gervasi DC, Michelson KS, Gunzler V, Kondo K et al. (2002). Biochemical purification and pharmacological inhibition of a mammalian prolyl hydroxylase acting on hypoxia-inducible factor. Proc Natl Acad Sci USA 99: 13459–13464. Ivan M, Kondo K, Yang H, Kim W, Valiando J, Ohh M et al. (2001). HIFalpha targeted for VHL-mediated destruction by proline hydroxylation: implications for O2 sensing. Science 292: 464–468. Iwase S, Lan F, Bayliss P, de la Torre-Ubieta L, Huarte M, Qi HH et al. (2007). The X-linked mental retardation gene SMCX/ JARID1C defines a family of histone H3 lysine 4 demethylases. Cell 128: 1077–1088. Jaakkola P, Mole DR, Tian YM, Wilson MI, Gielbert J, Gaskell SJ et al. (2001). Targeting of HIF-alpha to the von Hippel-Lindau ubiquitylation complex by O2-regulated prolyl hydroxylation. Science 292: 468–472. Kaelin Jr WG. (2002). Molecular basis of the VHL hereditary cancer syndrome. Nat Rev Cancer 2: 673–682. Kamura T, Koepp DM, Conrad MN, Skowyra D, Moreland RJ, Iliopoulos O et al. (1999). Rbx1, a component of the VHL tumor suppressor complex and SCF ubiquitin ligase. Science 284: 657–661. Klose RJ, Yan Q, Tothova Z, Yamane K, Erdjument-Bromage H, Tempst P et al. (2007). The retinoblastoma binding protein RBP2 is an H3K4 demethylase. Cell 128: 889–900. Kondo K, Kim WY, Lechpammer M, Kaelin Jr WG. (2003). Inhibition of HIF2alpha is sufficient to suppress pVHL-defective tumor growth. PLoS Biol 1: E83. Oncogene


786 Kondo K, Klco J, Nakamura E, Lechpammer M, Kaelin Jr WG. (2002). Inhibition of HIF is necessary for tumor suppression by the von Hippel-Lindau protein. Cancer Cell 1: 237–246. Krieg AJ, Rankin EB, Chan D, Razorenova O, Fernandez S, Giaccia AJ. (2010). Regulation of the histone demethylase JMJD1A by hypoxia-inducible factor 1 alpha enhances hypoxic gene expression and tumor growth. Mol Cell Biol 30: 344–353. Li N, Carrel L. (2008). Escape from X chromosome inactivation is an intrinsic property of the Jarid1c locus. Proc Natl Acad Sci USA 105: 17055–17060. Linehan WM, Vasselli J, Srinivasan R, Walther MM, Merino M, Choyke P et al. (2004). Genetic basis of cancer of the kidney: disease-specific approaches to therapy. Clin Cancer Res 10: 6282S–6289S. Maxwell PH, Wiesener MS, Chang GW, Clifford SC, Vaux EC, Cockman ME et al. (1999). The tumour suppressor protein VHL targets hypoxia-inducible factors for oxygen-dependent proteolysis. Nature 399: 271–275. Mersman DP, Du HN, Fingerman IM, South PF, Briggs SD. (2009). Polyubiquitination of the demethylase Jhd2 controls histone methylation and gene expression. Genes Dev 23: 951–962. Mikhaylova O, Ignacak ML, Barankiewicz TJ, Harbaugh SV, Yi Y, Maxwell PH et al. (2008). The von Hippel-Lindau tumor suppressor protein and Egl-9-Type proline hydroxylases regulate the large subunit of RNA polymerase II in response to oxidative stress. Mol Cell Biol 28: 2701–2717. Nakamura E, Abreu-e-Lima P, Awakura Y, Inoue T, Kamoto T, Ogawa O et al. (2006). Clusterin is a secreted marker for a hypoxiainducible factor-independent function of the von Hippel-Lindau tumor suppressor protein. Am J Pathol 168: 574–584. Negrotto S, Hu ZB, Alcazar O, Ng KP, Triozzi P, Lindner D et al. (2011). Noncytotoxic differentiation treatment of renal cell cancer. Cancer Res 71: 1431–1441. Ohh M, Park CW, Ivan M, Hoffman MA, Kim TY, Huang LE et al. (2000). Ubiquitination of hypoxia-inducible factor requires direct binding to the beta-domain of the von Hippel-Lindau protein. Nat Cell Biol 2: 423–427. Pollard PJ, Loenarz C, Mole DR, McDonough MA, Gleadle JM, Schofield CJ et al. (2008). Regulation of Jumonji-domain-containing histone demethylases by hypoxia-inducible factor (HIF)-1alpha. Biochem J 416: 387–394. Rini BI, Sosman JA, Motzer RJ. (2005). Therapy targeted at vascular endothelial growth factor in metastatic renal cell carcinoma: biology, clinical results and future development. BJU Int 96: 286–290. Roesch A, Fukunaga-Kalabis M, Schmidt EC, Zabierowski SE, Brafford PA, Vultur A et al. (2010). A temporarily distinct subpopulation of slow-cycling melanoma cells is required for continuous tumor growth. Cell 141: 583–594.

Schraml P, Frew IJ, Thoma CR, Boysen G, Struckmann K, Krek W et al. (2009). Sporadic clear cell renal cell carcinoma but not the papillary type is characterized by severely reduced frequency of primary cilia. Mod Pathol 22: 31–36. Semenza GL. (2003). Targeting HIF-1 for cancer therapy. Nat Rev Cancer 3: 721–732. Semenza GL. (2007). Hypoxia-inducible factor 1 (HIF-1) pathway. Sci STKE 2007: cm8. Semenza GL, Agani F, Booth G, Forsythe J, Iyer N, Jiang BH et al. (1997). Structural and functional analysis of hypoxia-inducible factor 1. Kidney Int 51: 553–555. Sharma SV, Lee DY, Li BH, Quinlan MP, Takahashi F, Maheswaran S et al. (2010). A chromatin-mediated reversible drug-tolerant state in cancer cell subpopulations. Cell 141: 69–80. Tahiliani M, Mei P, Fang R, Leonor T, Rutenberg M, Shimizu F et al. (2007). The histone H3K4 demethylase SMCX links REST target genes to X-linked mental retardation. Nature 447: 601–605. Thoma CR, Toso A, Gutbrodt KL, Reggi SP, Frew IJ, Schraml P et al. (2009). VHL loss causes spindle misorientation and chromosome instability. Nat Cell Biol 11: 994–1001. Wellmann S, Bettkober M, Zelmer A, Seeger K, Faigle M, Eltzschig HK et al. (2008). Hypoxia upregulates the histone demethylase JMJD1A via HIF-1. Biochem Biophys Res Commun 372: 892–897. Xia X, Lemieux ME, Li W, Carroll JS, Brown M, Liu XS et al. (2009). Integrative analysis of HIF binding and transactivation reveals its role in maintaining histone methylation homeostasis. Proc Natl Acad Sci USA 106: 4260–4265. Yan Q, Bartz S, Mao M, Li L, Kaelin Jr WG. (2007). The hypoxiainducible factor 2alpha N-terminal and C-terminal transactivation domains cooperate to promote renal tumorigenesis in vivo. Mol Cell Biol 27: 2092–2102. Yang H, Minamishima YA, Yan Q, Schlisio S, Ebert BL, Zhang X et al. (2007). pVHL acts as an adaptor to promote the inhibitory phosphorylation of the NF-kappaB agonist Card9 by CK2. Mol Cell 28: 15–27. Zhang HF, Gao P, Fukuda R, Kumar G, Krishnamachary B, Zeller KI et al. (2007). HIF-1 inhibits mitochondrial biogenesis and cellular respiration in VHL-deficient renal cell carcinoma by repression of C-MYC activity. Cancer Cell 11: 407–420. Zhou X, Sun H, Chen HB, Zavadil J, Kluz T, Arita A et al. (2010). Hypoxia induces trimethylated H3 lysine 4 by inhibition of JARID1A demethylase. Cancer Res 70: 4214–4221. Zimmer M, Doucette D, Siddiqui N, Iliopoulos O. (2004). Inhibition of hypoxia-inducible factor is sufficient for growth suppression of VHL/ tumors. Mol Cancer Res 2: 89–95.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

Oncogene