Oncogene (2010) 29, 5568–5578 & 2010 Macmillan Publishers Limited All rights reserved 0950-9232/10 www.nature.com/onc ORIGINAL ARTICLE PIASy stimulates HIF1a SUMOylation and negatively regulates HIF1a activity in response to hypoxia

X Kang1,2,JLi1, Y Zou1,2,JYi1, H Zhang3, M Cao1, ETH Yeh4,5 and J Cheng1,2

1The Department of Biochemistry and Molecular Cell Biology, Key Laboratory of Cell Differentiation and Apoptosis of Chinese Ministry of Education, E-Institutes of Shanghai Municipal Education Commission, Shanghai Jiao Tong University School of Medicine, Shanghai, China; 2State Key Laboratory of Oncogenes and Related Genes, Shanghai Cancer Institute, Shanghai Jiao Tong University, Shanghai, China; 3The Department of Abdominal Surgery, Cancer Hospital, Chinese Academy of Medical Science and Peking Union Medical College, Beijing, China; 4The Department of Cardiology, The University of Texas MD Anderson Cancer Center, Houston, TX, USA and 5Texas Heart Institute, St Luke’s Episcopal Hospital, Houston, TX, USA

Hypoxia-inducible factor-1a (HIF1a) is a crucial reg- prolyl-4-hydroxylases, and degraded through ubiquiti- ulator of the cellular response to hypoxia through its nation and proteosome-dependent manner (Maxwell regulation of genes that control erythropoiesis, angiogen- et al., 1999; Stebbins et al., 1999; Ohh et al., 2000; esis and anaerobic metabolism. We have previously shown Bruick and McKnight, 2001; Ivan et al., 2001). Because that HIF1a stability is regulated by SUMOylation under prolyl-4-hydroxylase is inactive under low concentration the hypoxic condition. However, how HIF1a became of oxygen, hypoxia limits HIF1a hydroxylation, allow- SUMOylated during hypoxia is still unknown. In this ing HIF1a to escape from normoxia-mediated degrada- study we identify PIASy as a specific E3 ligase for tion. HIF1a then translocates to nucleus to activate hypoxia-induced HIF1a SUMOylation. Hypoxia pro- hypoxia-inducible genes (Jaakkola et al., 2001; Schofield motes translocation of HIF1a to the nucleus to facilitate and Ratcliffe, 2004). Recently, several laboratories its binding to PIASy, enabling the conjugation of HIF1a showed that hypoxia could induce HIF1a SUMOylation by SUMO1. We further show that PIASy negatively (Comerford et al., 2003; Bae et al., 2004; Shao et al., regulates hypoxia-induced HIF1a stability and transacti- 2004; Berta et al., 2007; Carbia-Nagashima et al., 2007; vation. Knocking down PIASy increases the angiogenic Cheng et al., 2007). These studies suggest that SUMOy- activity of endothelial cells. Moreover, we show an inverse lation is important in the regulation of HIF1a under relationship between expression of PIASy and tumor hypoxia condition, although the impact of SUMOyla- angiogenesis in colon cancer. Thus, we define an important tion on HIF1a activity is controversial (Bae et al., 2004; role of PIASy in hypoxia signaling through promoting Berta et al., 2007; Carbia-Nagashima et al., 2007; Cheng HIF1a SUMOylation. et al., 2007). Oncogene (2010) 29, 5568–5578; doi:10.1038/onc.2010.297; SUMOylation is catalyzed by the activating (E1), published online 26 July 2010 conjugating (E2) and ligating (E3) enzymes (Hay, 2005; Geiss-Friedlander and Melchior, 2007; Yeh, 2009). Keywords: HIF1a; hypoxia; PIASy; SUMOylation There are one E1, one E2 and more than eight SUMO E3 ligases reported in mammalian cells (Hay, 2005; Geiss-Friedlander and Melchior, 2007; Yeh, 2009). During SUMOylation reactions, E3 ligase catalyzes Introduction the transfer of SUMO from E2 UBC9 to a substrate (Hay, 2005; Geiss-Friedlander and Melchior, 2007; Yeh, Hypoxia-inducible factor-1a (HIF1a) mediates hypoxia 2009). It is well recognized that E3 ligase is an important response by binding to canonical DNA sequences regulator for protein SUMOylation (Hay, 2005; Geiss- (hypoxia-responsive elements or HREs) in the promo- Friedlander and Melchior, 2007; Yeh, 2009). The largest ters or enhancers of target genes (Wang et al., 1995; group of SUMO E3 ligase is characterized by an SP- Carmeliet et al., 1998; Schofield and Ratcliffe, 2004; Ring motif (Hochstrasser, 2001; Jackson, 2001; Takahashi Gruber and Simon, 2006; Pouyssegur et al., 2006). et al., 2001), which can be subdivided into different During normoxia, HIF1a is hydroxylated at two critical families, including protein inhibitor of activated STAT proline residues within oxygen-dependent degradation (PIAS) family of proteins; MMS21, which is part of an domain (ODD) by a family of oxygen-sensitive enzymes octameric SMC5–SMC6 complex and Zip3, which is the meiosis-specific yeast protein (Hari et al., 2001; Johnson Correspondence: Dr J Cheng, Department of Molecular and Cell and Gupta, 2001; Kahyo et al., 2001; Sachdev et al., Biology, Shanghai Jiao Tong University School of Medicine, Shanghai 2001; Takahashi et al., 2001; Branzei et al., 2006). Other 200025, China. E-mail: [email protected] families of E3 ligases are described as without SP-Ring Received 25 January 2010; revised 16 June 2010; accepted 18 June 2010; motif in its structure and include RanBP2 and Pc2. published online 26 July 2010 RanBP2 has been shown to be essential for topoisomerase PIASy regulates HIF1a SUMOylation X Kang et al 5569 IIa SUMOylation and chromosome segregation in cell different SUMO E3 ligases. Thus, HIF1a SUMOylation mitosis (Dawlaty et al., 2008). Pc2 is a component was analyzed from the lysates of 293 cells transfected of polycomb protein complexes and can enhance with siRNAs plus Flag-HIF1a and HA-SUMO1. PIASy SUMOylation of CtBP, HDAC4, MEF2 and LXRb siRNA is most efficient in abolishing hypoxia-induced (Kagey et al., 2003). HIF1a SUMOylation (Figure 1b). The ability of PIAS proteins were first identified as negative PIASy in downregulating hypoxia-induced HIF1a regulators of STAT signaling (Shuai and Liu, 2005). SUMOylation was further shown by two additional The mammalian PIAS family consists of five PIAS PIASy siRNAs (Figure 1c). We examined SUMOylation proteins, PIAS1, PIAS3, PIASxa, PIASxb and PIASy, of endogenous HIF1a in PIASy siRNA-transfected 293 which have significant sequence identity (more than cells and found a similar result as in cells with co- 40%) and several functional domains and motifs that transfection of HIF1a and SUMO1 (Figure 1d). are conserved among the members (Shuai and Liu, In addition, HIF1a conjugation by SUMO2, another 2005). The Ring-finger domain located in the middle of member of SUMO family, was also significantly PIAS is required for E3 ligase activity of PIAS proteins decreased in PIASy siRNA-transfected cells (Supple- (Kahyo et al., 2001; Schmidt and Muller, 2002). PIAS- mentary Figure S2). mediated SUMOylation of target proteins has been To further determine whether this effect requires shown to involve in a wide range of cellular processes, PIASy E3 ligase activity, we mutated the E3 catalytic including transcriptional control and protein–protein domain of PIASy and compared its activity to interactions (Kahyo et al., 2001; Sachdev et al., 2001; SUMOylate HIF1a with that of wild-type PIASy. As Sapetschnig et al., 2002; Schmidt and Muller, 2002; shown in Figure 1e, overexpression of wild-type PIASy Rogers et al., 2003; Mabb et al., 2006; Carter et al., in 293 cells enhanced endogenous HIF1a SUMOylation. 2007). However, PIASy catalytic-inactivation mutant could In the previous study, we found that SUMOylated not induce HIF1a SUMOylation under the same HIF1a could bind to VHL and be degraded even under experimental condition. We further performed in vitro hypoxia condition (Cheng et al., 2007). SENP1 removes SUMOylation assay and found that addition of PIASy SUMO from SUMOylated HIF1a to stabilize HIF1a, recombinant protein in the reaction mixture markedly allowing for HIF1a to participate in the regulation of increased the SUMOylation of HIF1a ODD, a fragment hypoxia signaling (Cheng et al., 2007). Although, along containing two SUMOylation sites (Figure 1f). with other, we have reported that hypoxia could induce Interestingly, we did not observe any changes in HIF1a SUMOylation (Comerford et al., 2003; Bae overall SUMO1 conjugation in PIASy siRNA-trans- et al., 2004; Berta et al., 2007; Carbia-Nagashima et al., fected cells (Supplementary Figure S3), suggesting that 2007), it is unknown how HIF1a SUMOylation occurs PIASy specifically enhanced HIF1a SUMOylation. in hypoxic condition. In this study, we identify PIASy Because HIF2a was also reported as a SUMOylated as a specific E3 ligase for hypoxia-induced HIF1a protein (van Hagen et al., 2010), we then examined SUMOylation. We find that hypoxia promotes inter- whether PIASy was an E3 ligase for HIF2a SUMOyla- action of PIASy with HIF1a, which is essential for tion. By using siRNA approach, we found that PIASxa, HIF1a conjugation by SUMO1. Furthermore, PIASy is not other PIAS, was a specific E3 ligase for HIF2a shown to regulate negatively hypoxia-HIF1a signaling SUMO conjugation (Supplementary Figure S4). Taken through the induction of HIF1a SUMOylation. together, PIASy is a specific E3 ligase for hypoxia- induced HIF1a SUMOylation.

Results Hypoxia induces the interaction of PIASy and HIF1a During SUMO conjugation, E3 ligase binds to both PIASy is the E3 ligase for hypoxia-induced HIF1a target and UBC9 directly and then facilitates SUMO SUMOylation conjugation by catalyzing the transfer of SUMO from We and others showed that hypoxia could induce global UBC9 to a substrate (Gong et al., 1997; Saitoh et al., SUMO conjugation (Supplementary Figure S1) and 1998; Schwarz et al., 1998; Hay, 2005; Geiss-Friedlander HIF1a SUMOylation (Bae et al., 2004; Shao et al., 2004; and Melchior, 2007). We first examined the binding of Carbia-Nagashima et al., 2007; Cheng et al., 2007). PIASy to HIF1a. 293 cells were transfected with Flag- Because E3 ligase could promote SUMO conjugation, HIF1a and HA-PIASy. Flag-HIF1a was precipitated by we reasoned that an E3 ligase might be involved in anti-Flag antibody. As expected, HA-PIASy was present hypoxia-induced HIF1a SUMOylation. To identify the in HIF1a precipitates (Figure 2a). This interaction was E3 ligase essential for HIF1a SUMOylation, we used an specific because other members of PIAS family could siRNA approach against known SUMO E3 ligases. To not co-precipitate with HIF1a. evaluate the efficiency and specificity of siRNAs, we It is well known that HIF1a translocates into nucleus isolated total RNA from 293 cells transfected with under hypoxia condition (Kallio et al., 1998). This different E3 ligase siRNA and reverse transcription process can be abolished by the mutation of the nuclear (RT)–PCR was performed to determine the expression location signal (NLS) of HIF1a (Kallio et al., 1998). level of endogenous SUMO E3 mRNAs. Figure 1a Interestingly, NLS-mutated HIF1a could not be SU- showed that siRNA could specifically knock down MOylated under hypoxia condition (Figure 2b), indicating

Oncogene PIASy regulates HIF1a SUMOylation X Kang et al 5570 si-RNA

β PIASy-WT /x α PIASy-Mut SUMO- Mock PIAS1 PIASx PIAS3 PIASy RanBP2 Pc2 MMS21 ScramblePIASy-Si1 PIASy-Si2 Control 130 HIF1α PIAS1 SUMO- GAPDH 130 HIF1α α PIASxα/xβ IP:HIF1 GAPDH IP:Flag IB:SUMO1 PIAS3 55 IB:HA 55 GAPDH PIASy 130 130 GAPDH HIF1α IP:HIF1α α RanBP2 IP:Flag IB:HIF1 GAPDH 55 IB:Flag Pc2 55 GAPDH PIASy MMS21 PIASy GAPDH

si-RNA

β /x α Nox Hyx UBC9 250 25 25 (ng)

Mock PIAS1 PIASx PIAS3 PIASy RanBP2Pc2 MMS21 PIASy 25 (ng) Control SUMO- 130 HIF1 α 250 130 IP:Flag ScramblePIASy-Si1 Scramble PIASy-Si1 SUMO- SUMO- 55 IB:HA 130 ODD HIF1 α 100 IP:HIF1α 130 IB:SUMO1 55 GST-ODD WCL 55 Pulldown by GST IB:HA HIF1α IB:HIF1α 55 IP:Flag IB:Flag Figure 1 PIASy is a specific E3 ligase for HIF1a SUMOylation. (a) 293 cells were transfected with control or siRNA against different SUMO E3 ligases. The expression levels of SUMO E3 ligase mRNA in these transfected cells was determined by RT–PCR. GAPDH served as a control. (b) 293 cells were transfected with siRNA as indicated. After 24 h, the cells were transfected with Flag-HIF1a and HA-SUMO1 plasmids (control cells were transfected without HA-SUMO1). After hypoxia treatment, Flag-HIF1a was immunoprecipitated with anti-Flag antibody (IP). Bound proteins were blotted with anti-HA or anti-Flag antibody (IB). WCL was immunoblotted with anti-HA antibody. (c) 293 cells were transfected with Flag-HIF1a and HA-SUMO1 (control cells were transfected without HA-SUMO-1) and scramble shRNAs or two shRNAs (PIASy-Si1 and PIASy-Si2) as indicated. Flag-HIF1a was immunoprecipitated with anti-Flag antibody (IP). Bound proteins were blotted with anti-HA or anti-Flag antibody (IB). WCL was immunoblotted with anti-HA antibody. (d) 293 cells were transfected with scramble or PIASy-Si1 shRNAs as indicated. The cells were treated with normoxia (Nox) or hypoxia (Hpx) for 4 h, and endogenous HIF1a was immunoprecipitated with anti-HIF1a antibody (IP). Bound proteins were blotted with anti-SUMO1 or anti-HIF1a antibody (IB). (e) 293 cells were transfected with 0.5 mg and 1.5 mg PIASy-WT or PIASy-Mut plasmids. Flag-HIF1a was immunoprecipitated with anti-Flag antibody (IP). Bound proteins were blotted with anti-SUMO1 or anti-Flag antibody (IB). WCL was immunoblotted with anti-HA antibody. (f) The mixture of GST-ODDpm, SUMO1 and E1 was incubated for 1 h at 37 1C with UBC9 and GST-PIASy as indicated. The GST-ODDpm was pull-downed from reaction mixture by GST beads and analyzed by western blot with anti-HIF1a antibody.

that hypoxia-induced nuclear translocation might be a Because hypoxia can promote HIF1a nuclear localiza- prerequisite for HIF1a SUMOylation. Moreover, sta- tion, we hypothesized that it might facilitate binding of bilization of HIF1a by hydroxylase inhibitor DMOG in HIF1a to PIASy. To address this issue, we used a normoxia did not increase HIF1a SUMOylation (Sup- binding assay based on Gal4 interaction reporter plementary Figure S5), suggesting that increasing analysis system to monitor in vivo interaction of PIASy HIF1a stability by hypoxia is not enough for SUMO and HIF1a in cells under hypoxia condition. We conjugation and the nuclear translocation of HIF1a is generated plasmids of HIF1a fused with active domain still need to initiate this process. Thus, hypoxia-driven (AD) and PIASy fused with binding domain (BD). nuclear translocation of HIF1a is proposed to be These plasmids along with Gal4-luciferase were trans- required for its interaction with PIASy, which is located fected into 293 cells. The luciferase activity was in nucleus (Matsuura et al., 2005). Indeed, PIASy co- measured in these transfected cells with or without precipitated with wild-type but not the NLS mutant hypoxia treatment. As shown in Figure 2e, luciferase (K719T) of HIF1a (Figure 2c). To further confirm the activity was significantly increased following the hypox- interaction of PIASy and HIF1a occurred in nucleus, we ia treatment, indicating that hypoxia can induce the extracted cytosolic and nuclear fractions from the 293 interaction between HIF1a and PIASy in vivo. cells following transfection of Flag-HIF1a. Analysis of immunoprecipitates revealed a complex containing endogenous PIASy and Flag-HIF1a in nucleus, but Interaction of PIASy and HIF1a is essential not in cytosol (Figure 2d). for hypoxia-induced HIF1a SUMOylation The above results suggest that HIF1a nuclear To determine the domains responsible for the inter- translocation is critical for its interaction with PIASy. action between HIF1a and PIASy, we generated a series

Oncogene PIASy regulates HIF1a SUMOylation X Kang et al 5571 Flag-HIF1α +

α β Flag-HIF1αWT + −−+ Flag-HIF1αK719T −+− + Vector HA-PIAS1HA-PIASxHA-PIASxHA-PIAS3HA-PIASy −++− 100 HA-SUMO1 IP:Flag PIASy SUMO-HIF1α IB:HA 100 55 IP:Flag IB:HA IP:Flag Flag-HIF1α 55 IB:Flag IP:Flag HIF1αWT/K719T WCL HA-PIAS IB:Flag 250 SUMO- α + −−+ WCL Flag-HIF1 WT 100 Proteins Flag-HIF1αK719T −+− + IB:HA HA-PIASy −++− HIF1αWT/K719T 100 HA-PIASy IP:Flag 210 55 IB:HA Normoxia 180 Hypoxia 150 IP:Flag α IB:Flag Flag-HIF1 120

WCL HA-PIASy 90

60 C N Luciferase Activity Flag-HIF1α −+− + 30 100 IP: Flag PIASy 0 IB:PIASy +++++ 55 Gal4-luc AD −−−++ −−−+ + IP:Flag Flag-HIF1α BD IB:Flag AD-HIF1α −−−++ BD-PIASy −−+ −+ Input Actin

Input Histone 3

Figure 2 Hypoxia promotes the binding of PIASy to HIF1a.(a) 293 cells were co-transfected with Flag-HIF1a and the member of PIAS family as indicated. Flag-HIF1a was immunoprecipitated from the cell lysates with anti-Flag antibody (IP). Bound proteins were detected by immunoblotting with anti-HA or anti-Flag antibody (IB). WCL were immunoblotted with anti-HA antibody. (b) 293 cells were transfected with HA-SUMO1, Flag-HIF1a WT or Flag-HIF1aK719T plasmids as indicated. Flag-HIF1a was immunopreci- pitated from the cell lysates with anti-Flag antibody (IP). Bound proteins were detected by immunoblotting with anti-HA or anti-Flag antibody (IB). WCL were immunoblotted with anti-HA antibody. (c) 293 cells were transfected with HA-PIASy, Flag-HIF1a WT or Flag-HIF1aK719T plasmids as indicated. Flag-HIF1a was immunoprecipitated from the cell lysates with anti-Flag antibody (IP). Bound proteins were detected by immunoblotting with anti-HA or anti-Flag antibody (IB). WCL were immunoblotted with anti-HA antibody. (d) The cytosolic and nuclear fractions were extracted from the 293 cells transfected with Flag- HIF1a. Flag-HIF1a was immunoprecipitated from cell lysates with anti-Flag antibody (IP). Bound proteins were detected by immunoblotting with anti-PIASy or anti-Flag antibody (IB). WCL were immunoblotted with anti-actin or histone 3 antibody. (e) 293 cells were transfected with Gal4 luciferase, active domain (AD), binding domain (BD), AD-HIF1a and BD-PIASy plasmids as indicated. The cells were treated with or without hypoxia (2% O2) for 12 h before luciferase assay. The data are presented in mean±s.d. of three independent experiments. Difference between normoxia and hypoxia was significant (Po0.01, t-test).

of truncated mutants of PIASy or HIF1a. In the association between region of 211–330 of HIF1a and experiments to identify the PIASy binding region on PIASy is unknown. HIF1a, we determined that both regions of 211–330 and Using similar strategy, we mapped the BD of HIF1a 331–698 amino acid of HIF1a could individually to the N-terminal region, most likely between 150 and interact with PIASy (Figure 3a). Interestingly, region 202 amino acid of PIASy (Figure 3b). Because this 331–698 amino acid is the ODD domain of HIF1a fragment can bind to HIF1a, it could serve as a containing SUMO conjugation sites (Ohh et al., 2000; dominant-negative inhibitor for PIASy E3 ligase activity Ivan et al., 2001; Jaakkola et al., 2001). ODD domain through competitively binding with HIF1a. Indeed, has been shown to be a SUMO conjugation target for expression of PIASy (150–202) fragment could decrease PIASy E3 ligase activity in in vitro assay (Figure 1e). the binding of full-length PIASy to HIF1a (Figure 3c). Thus we thought that binding of PIASy to ODD The competitive binding was also confirmed in in vivo domain (331–698) was sufficient for HIF1a SUMOyla- Gal4 interaction reporter assay, which showed that the tion under hypoxia condition. However, the role of the luciferase activity induced by the interaction of PIASy

Oncogene PIASy regulates HIF1a SUMOylation X Kang et al 5572 Flag-HIF1α HA-PIASy 1−826 1-511 PIASy binding region (211-330) (331-698) HIF1α binding region (150-202) Input IP:HA-PIASy Input IP:Flag-HIF1α

Ring Ring Δ Δ 1-84 85-210211-3301-330 331-698699-826331-8261−826 1-84 85-210211-3301-330 331-698699-826331-8261−826

100 1-511 1-100 1-202 1-323 1-511 324-511 1-511 1-100 1-202 1-323 1-511 324-511 70 55 55 27

27 IB: HA

IB: Flag Myc-PIASy

− Myc-PIASy (150-202) 100-202 100-175 125-202 125-175 150-202 100 27 IP:Flag-HIF1α Myc-PIASy IP:Flag-HIF1α IB:Myc IB:Myc 55 7 27 27 Myc-PIASy Input (150-202) IB:Myc Flag-HIF1α 7

PIASy(150-202) Flag-HIF1α + + + HA-SUMO1 − + + 250 Myc-PIASy(150-202) −−+ 200 SUMO- * 130 HIF1α 150 ** IP:Flag 100 IB:HA 50 55 IgG Luciferase Activity 0 IP:Flag − HIF1α PIASy (150-202) IB:Flag Gal4-luc + BD-PIASy PIASy IB:Myc + AD-HIF1α (150-202) Figure 3 The binding of HIF1a to PIASy is critical for hypoxia-induced HIF1a SUMOylation. (a) 293 cells were co-transfected with HA-PIASy- and Flag-HIF1a-truncated plasmids as indicated. HA-PIASy was immunoprecipitated from cell lysates with anti-HA antibody (IP). Bound proteins were detected by immunoblotting with anti-Flag (IB). Input was immunoblotted with anti-Flag antibody. (b) In top panel, 293 cells were transfected with Flag-HIF1a- and HA-PIASy-truncated plasmids as indicated. Flag-HIF1a was immunoprecipitated from cell lysates with anti-Flag antibody (IP). Bound proteins were detected by immunoblotting with anti-HA (IB). Inputs were immunoblotted with anti-HA antibody. In the bottom panel, 293 cells were transfected with Flag-HIF1a- and Myc- PIASy-truncated plasmids as indicated. Flag-HIF1a was immunoprecipitated from cell lysates with anti-Flag antibody (IP). Bound proteins were detected by immunoblotting with anti-Myc (IB). Input was immunoblotted with anti-Myc antibody. (c) 293 cells were transfected with Flag-HIF1a, Myc-PIASy-WT or titrated Myc-PIASy (150–202) plasmids. Flag-HIF1a was immunoprecipitated from cell lysates with anti-Flag antibody (IP). Bound proteins were detected by immunoblotting with anti-Myc or anti-Flag (IB). (d) 293 cells were transfected with Gal4 luciferase, BD-PIASy, AD-HIF1a and titrated Myc-PIASy (150–202) as indicated. The cells were treated with hypoxia (2% O2) for 12 h before luciferase assay. The data are presented in mean±s.d. of three independent experiments (*Po0.02,**Po0.005, t-test). (e) 293 cells were co-transfected with Flag-HIF1a, HA-SUMO1 and Myc-PIASy (150–202) as indicated. The cells were treated with MG132 (10 mM) and hypoxia (2% O2) for 4 h. Flag-HIF1a was then immunoprecipitated from cell lysates with anti-HA antibody (IP). Bound proteins were detected by immunoblotting with anti-HA or anti-Flag (IB). WCL was immunoblotted with anti-Myc.

and HIF1a was decreased by expression of PIASy et al., 2007; Carbia-Nagashima et al., 2007; Cheng et al., (150–202) fragment (Figure 3d). More importantly, 2007). Bae et al. (2004) showed that SUMO-1 increased overexpression of PIASy (150–202) decreased HIF1a HIF1a protein level and its activity in SUMO1-over- SUMOylation promoted by PIASy (Figure 3e), indicating expressed cells. Carbia-Nagashima et al. (2007) reported that the disruption of the interaction could abrogate the E3 that RSUME could enhance HIF1a SUMOylation and ligase activity of PIASy for HIF1a SUMOylation. These its activity through its interaction with UBC9. However, results also suggest that the binding of PIASy to HIF1a is Berta et al. (2007) reported that HIF-dependent crucial for hypoxia-induced HIF1a SUMOylation. transcriptional activity is increased in SUMO-deficient HIF1a by co-transfection of siRNA targeted to endogenous HIF1a together with HIF1a siRNA-resis- PIASy negatively regulates HIF1a stability and activity tant expression vectors carrying mutations for SUMO We and others reported a controversial consequence of modification. In our previous study, we used SENP1À/À SUMOylation on HIF1a activity (Bae et al., 2004; Berta MEF cells to show the importance of SENP1 in the

Oncogene PIASy regulates HIF1a SUMOylation X Kang et al 5573 regulation of HIF1a stability through de-SUMOylation HIF1a activity by performing an HIF1a-dependent of HIF1a (Cheng et al., 2007). To show directly the luciferase assay in SENP1À/À MEF cells. Overexpression relationship between HIF1a SUMOylation and its of PIASy significantly decreased hypoxia-induced activity, we performed an HRE reporter gene assay in HIF1a transactivation (PIASy WT vs control, SENP1À/À MEF cells and showed that the mutation of Figure 4d). However, mutation of PIASy E3 ligase HIF1a SUMOylation sites could increase its transcrip- activity (PIASy Mut) abrogated the effect of PIASy on tional activity in response to hypoxia (Figure 4a), HIF1a transactivation (Figure 4d). Similarly, PIASy indicating that SUMOylation could affect HIF1a (150–202), a fragment that can specifically disrupt the transactivation. Thus, we hypothesized that PIASy PIASy interaction with HIF1a, markedly increased could inhibit HIF1a-mediated hypoxia signaling hypoxia-induced HRE-luciferase in dose-dependent through HIF1a SUMOylation. To test it, we first manner (Figure 4e). These results suggest PIASy exerted examined the effect of PIASy on the stability of HIF1a a negative effect on HIF1a stability and activity. because we have previously shown that SUMOylation could cause HIF1a degradation under hypoxia condi- tion. Because SENP1 can regulate HIF1a degradation PIASy negatively regulates angiogenesis by deconjugation of SUMOylated HIF1a (Cheng et al., Above results suggest that PIASy might regulate HIF1a 2007), we performed this assay in SENP1À/À MEF cells biological function through HIF1a SUMOylation. to exclude action of SENP1. We generated PIASy À/À 20 siRNA stably-transfected SENP1 MEF cells and Normoxia treated them with or without hypoxia for 12 h. PIASy SENP1-/- 15 Hypoxia was markedly downregulated by 80% in PIASy siRNA Hypoxia −+− + À/À stably-transfected SENP1 MEF cells (Supplementary mPIASy-Si −++ − Figure S6). Importantly, HIF1a protein level signifi- 10 cantly increased in PIASy-silenced SENP1À/À MEF cells, HIF1α 5 especially under hypoxia condition (Figure 4b), indicat- Fold: 1.02.4 2.8 6.8 ing that endogenous PIASy could decrease HIF1a Luciferase Activity 0 PIASy stability. To determine the functional consequence of α Mut Actin action of PIASy on HIF1a stabilization, we analyzed HIF1 α− the expression of vascular endothelial growth factor HIF1 (VEGF) and Glut-1, which are HIF1a target genes, in À/À VEGF Glut-1 PIASy-silenced SENP1 MEF cells. As shown in 8 10 Figure 4c, hypoxia-induced expression of both target Normoxia Normoxia À/À Hypoxia genes markedly increased in PIASy-silenced SENP1 Hypoxia 8 MEF cells in comparison to that of scramble siRNA 6 cells. We further confirmed the effect of PIASy on 6 4 4

2 2

Figure 4 PIASy negatively regulates HIF1a stability and activity. Relative mRNA levels Relative mRNA levels (a) SENP1À/À MEF cells were transfected with HRE reporter gene, plus HIF1a wide-type (WT) or HIF1a SUMOylation site mutant 0 0 (Mut) plasmids as indicated, and treated with or without hypoxia (2% O2) for 12 h before luciferase assay. The data are presented in mean±s.d. of three independent experiments. Difference in Scramble Scramble hypoxia-induced HRE activity between HIF1a and HIF1a-Mut mPIASy-Si mPIASy-Si was significant (Po0.01, t-test). (b) mPIASy-Si or scramble stably- À/À transfected SENP1 MEF cells were treated with or without HRE-luc HRE-luc hypoxia. The cell lysates were analyzed by immunoblotting with 5 25 anti-HIF1a, anti-PIASy or anti-actin antibody. (c) The mRNA Normoxia Normoxia levels of VEGF and Glut-1 in mPIASy-Si or scramble SENP1À/À Hypoxia Hypoxia 4 20 MEF cells were measured by using real-time PCR. The data are ** presented in mean±s.d. of three independent experiments. * Difference in hypoxia-induced VEGF and Glut-1expression 3 15 between Scramble and mPIASy siRNA was significant (Po0.01, t-test). (d) 293 cells were transfected with HRE luciferase, plus PIASy wide-type (WT) or PIASy mutant (Mut) plasmids as 2 10 indicated, and treated with or without hypoxia (2% O2) for 12 h Luciferase Activity

± Luciferase Activity before luciferase assay. The data are presented in mean s.d. of 1 5 three independent experiments. Difference in hypoxia-induced HRE activity between control and PIASyWT was significant 0 (Po0.01, t-test). (e) 293 cells were transfected with HRE luciferase 0 and titrated PIASy (150–202) plasmids as indicated, and treated PIASy with or without hypoxia (2% O ) for 12 h before luciferase assay. − 2 Control (150-202) The data are presented in mean±s.d. of three independent PIASy WTPIASy Mut experiments (*Po0.01, **Po0.005, t-test).

Oncogene PIASy regulates HIF1a SUMOylation X Kang et al 5574 2.5 Normal Case 1 Case 2 2

1.5 αPIASy

1

0.5 α

Relative mRNA level of VEGF CD31 0

Scramble PIASy-Si1 PIASy-Si2

Scramble PIASy-Si1 PIASy-Si2 80 Normoxia Hypoxia 60 Normoxia 40

20 Number of tubes

0 Hypoxia

Scramble PIASy-Si1PIASy-Si2

Hypoxia

SU SENP1

PIASy SU

Proteasomal Degradation VEGF

Nucleus Cytoplasm

Figure 5 PIASy regulates angiogenic activity. (a) The mRNA levels of VEGF in scramble-, PIASy-Si1- or PIASy-Si2-HUVEC cells were measured by real-time PCR. The data are presented in mean±s.d. of three independent experiments. Difference in VEGF expression between Scramble and PIASy-Si1 or -Si2 cells was significant (Po0.01, t-test). (b) A total of 4 Â 104 scramble-, PIASy-Si1- or PIASy-Si2-HUVEC cells were cultured on Matrigel-coated plates under normoxia or hypoxia condition. The tubule formation was analyzed at 16 h after treatment (left panel). Right panel shows the number of tubes per field and present as mean±s.d. of three independent experiments. Difference in hypoxia-induced tube formation between Scramble and PIASy-Si1 or -Si2 cells was significant (Po0.05, t-test). (c) Representative human normal and colon cancer tissue samples with PIASy and CD31 staining. Case 1: high expression of PIASy and low density of angiogenesis. Case 2: low expression of PIASy and high density of angiogenesis. (d) Model depicting the role of PIASy in regulating hypoxia-induced HIF1a SUMOylation and hypoxia signaling.

Because silencing of PIASy can increase HIF1a target of endothelial cells. Expression of PIASy in PIASy- VEGF gene expression, we reasoned that PIASy could specific siRNAs-transfected HUVEC cells was silenced regulate VEGF-mediated angiogenesis. Therefore we by 75% (Supplementary Figure S7). Real-time PCR performed an angiogenic tube assay in HUVEC cells to assay confirmed the increase of VEGF mRNA expres- test whether PIASy has an effect on angiogenic activity sion in the PIASy siRNA cells (Figure 5a). Importantly,

Oncogene PIASy regulates HIF1a SUMOylation X Kang et al 5575 Table 1 Summary of PIASy and CD31 immunohistochemical tional activity by precluding the binding of p300/CBP staining in colon cancer samples (Lando et al., 2002). Phosphorylation and acetylation of Angiogenesis (CD31) Total HIF1a have also been reported as regulation mechan- isms for its activation (Brahimi-Horn et al., 2005). Low High We and others have shown that HIF1a is modified by SUMO under hypoxia condition. Carbia-Nagashima PIASy High 10 (58.8%) 7 (41.2%) 17 et al. (2007) have reported that RSUME can enhance Low 3 (20%) 12 (80%) 15 global SUMO conjugation by interacting with UBC9. Total 13 19 32 They further found that the expression of RSUME is induced by hypoxia, and consequently HIF1a SUMOy- The 32 surgical specimens of colon cancer were stained with anti- lation increased. In this study, we identify a specific E3 PIASy and anti-CD31 antibodies as shown in Figure 5c. Expression patterns in all samples are summarized here. The Pearson’s w2-test was ligase for HIF1a SUMOylation, which is promoted by used to analyze the significance of the relationship between PIASy hypoxia (Figure 5d). The conclusion is made based on expression and intensity of angiogenesis expression (Po0.025). the following evidences. First, PIASy siRNA, not other E3 siRNAs, decreases SUMO1 modification of HIF1a and increases the activity of HIF1a in response to PIASy silencing increased the tube formation of hypoxia. Second, overexpression of PIASy promotes the HUVEC cells during hypoxia treatment (Figure 5b), HIF1a SUMOylation even in the cells without hypoxia indicating that PIASy negatively regulates angiogenic treatment. In contrast, a catalytic mutant of PIASy activity of endothelial cells. blocks HIF1a SUMOylation. Third, recombinant To further investigate the significance of PIASy action PIASy protein significantly increases HIF1a SUMO on angiogenesis, we determined the association between conjugation in vitro. Fourth, hypoxia induces the PIASy and angiogenesis in colon cancer samples by binding of PIASy to HIF1a. immunohistochemistry. A series of sections from 32 Bae et al. (2004) previously reported that SUMO-1 colon cancer samples and 4 normal colon tissue samples increased HIF1a protein level and its activity in were immunostained with anti-PIASy (aPIASy) and SUMO1-overexpressed cells. Carbia-Nagashima et al. CD31 (aCD31) antibodies. As PIASy-positive cells were (2007) reported that RSUME could enhance HIF1a detected to distribute mainly in the stroma of normal SUMOylation and its activity. These studies were and colon cancer tissues (Figure 5c), we focused the carried out in the presence of SENP1. In contrast, our analysis on the cells in stroma. Immunohistochemical previous report (Cheng et al., 2007) and the current scoring was determined by the intensity of the distribu- studies have used SENP1 1À/À MEF cells to show the tion of positive cells (for PIASy staining) or angiogen- importance of SENP1 in the regulation of HIF1a esis (CD31 staining) in the stroma. We set the average stability and activity through de-SUMOylation of score of normal tissues as baseline. The tumor tissues HIF1a. In support of our results, Berta et al. (2007) with scores more than 50% the baseline were designated reported that HIF-dependent transcriptional activity is as PIASy or angiogenesis high, and others were increased in SUMO-deficient HIF1a by co-transfection designated as PIASy or angiogenesis low. We found a of siRNA targeted to endogenous HIF1a together with significantly negative correlation between PIASy expres- HIF1a siRNA-resistant expression vectors carrying sion and intensity of angiogenesis (Table 1). Among 17 mutations for SUMO modification. Another possibility cases with PIASy high, 10 cases were with angiogenesis for the difference observed by Bae et al. (2004) and low (Table 1; case 1 in Figure 5c). However, the 12 of 15 Carbia-Nagashima et al. (2007) may be due to an cases with PIASy low showed high density of blood indirect effect on HIF1a through SUMOylation of other vessels indicated by anti-CD31 staining within the proteins caused by overexpression of SUMO1 or stroma of these samples (Table 1; case 2 in Figure 5c). RSUME in their system. These results suggest that PIASy is a negative regulator Our study shows that hypoxia-driven HIF1a nuclear for tumor angiogenesis in colon cancer. translocation is a crucial process for HIF1a SUMOyla- tion promoted by PIASy. It has been shown that HIF1a nuclear accumulation is dependent on an NLS within Discussion the C-terminal end of HIF1a (Kallio et al., 1998). A single amino-acid substitution within the NLS HIF1a is pivotal in the regulation of gene transcription sequence motif (K719T) can inhibit HIF1a nuclear in response to hypoxia. The stability and subsequent translocation (Kallio et al., 1998) and results in transactivational function of HIF1a has been shown inhibition of HIF1a SUMOylation (Figure 2b). More- to be regulated by posttranslational modification over, the interaction between HIF1a and PIASy occurs (Brahimi-Horn et al., 2005). Hydroxylation of HIF1a mainly in nucleus, which is essential for PIASy-mediated on two proline residues by prolyl-4-hydroxylase is the HIF1a SUMOylation. Thus, these results suggest that signal for interaction with the E3 ubiquitin ligase VHL, the HIF1a nuclear accumulation is a prerequisite for and subsequent polyubiquitination and degradation of hypoxia-induced HIF1a SUMOylation. HIF1a under normoxia (Ivan et al., 2001; Jaakkola We mapped a PIASy BD on HIF1a locating in et al., 2001). Hydroxylation of N803 in the C terminus residues 331–698 of HIF1a. This domain is in the ODD, of HIF1a by FIH leads to inactivation of its transcrip- which contains two prolyl residues (P402 and 564) for

Oncogene PIASy regulates HIF1a SUMOylation X Kang et al 5576 hydroxylation that lead to degradation under normoxia banks maintained by the Cancer Hospital of Chinese Academy (Ivan et al., 2001; Jaakkola et al., 2001). ODD domain of Medical Science and Peking Union Medical College. The also contains two SUMOylation sites (K391 and 477), tissue specimens had been obtained during colectomy in 32 and SUMOylation can cause HIF1a degradation under colon cancer patients. The histopathology of the specimens hypoxia (Cheng et al., 2007). We showed this domain is was determined by pathologists in the same hospital. SUMOylated in PIASy-dependent manner (Figure 1f). RT–PCR In conclusion, in this study we found that PIASy RT–PCR was performed on 20 ng of total RNA with specific is an E3 ligase, which negatively regulates HIF1a primers and a Qiagen one-step RT–PCR kit (Qiagen, Valencia, stability and transactivation, and downregulates VEGF- CA, USA) according to the manufacturer’s protocol. The mediated angiogenesis. Thus, we defined an important primers used for the expression analysis of genes were the role of PIASy in hypoxia-HIF1a signaling. These results following: Pc2 (forward: 50-GAGAACATCCTGGACCCCA may provide a molecular basis to use PIASy as a new GGC-30,reverse:50-CTGGGCGCCTCCTTGTGGCCGC-30), target for antiangiogenic therapies. PIAS1 (forward: 50-GAACTCCAAGTACTGTTGGGCT-30, reverse: 50-CGGATGGACTGGGTGAAGAGCT-30), PIASx (forward: 50-CCAGCCAACCGTGTACAAAAATAG-30,re- 0 0 Materials and methods verse: 5 -TTCTTTGTTCTCCTGGCAAATC-3 ), PIASx (for- ward: 50-CCAGCCAACCGTGTACAAAAATAG-30,reverse: 0 0 Antibodies, plasmids, siRNA and cell lines 5 -CTGGTGGTGGTGACAGACGTAC-3 ), PIAS3 (forward: 50-GGTGCTTCTTGGCTTTGCTGGC-30, reverse: 50-GCTGG Antibodies against HIF1a (catalog no. NB-105) were pur- 0 0 chased from Novus (Littleton, CO, USA). PIASy (catalog no. CTAGAAGTGGATGCAAG-3 ), PIASy (forward: 5 -CTTT AATATGCTGGATGAGCTG-30, reverse: 50- CTCCTTGACC sc-50347) and Myc (catalog no. sc-40) antibodies were 0 0 purchased from Santa Cruz Corp. (Santa Cruz, CA, USA). AGTGCCTTGCAC-3 ), Ranbp2 (forward: 5 -ACTTCAGAGA CAAGCAAGGCTCCA-30, reverse: 50- TGCAATCCCACTGT Anti-Flag antibody (M2, catalog no. F3165) was from Sigma 0 0 (St Louis, MO, USA). Anti-HA antibody (16B12, catalog no. CCTTCCTTCT-3 ), MMS21 (forward: 5 -AAGCTGACAGA GAAGCTGACGGAA-30,reverse:50-GGCTACAGCCAATTT MMS-101P) was purchased from Covance (Princeton, NJ, 0 0 USA). Anti-SUMO1 antibody (catalog no. 18-2306) was GAGGGCAAT-3 ), GAPDH (forward: 5 -CATGTTCGTCA TGGGTGTGAACCA-30, reverse: 50-AGTGATGGCATGGA purchased from Zymed (Mason, OH, USA). 0 Plasmid pcDNA3-Flag-PIASy truncates (1-100, 1-202, CTGTGGTCAT-3 ). 1-323, del RF, 324-511, 1-511) were provided by Guangchao Sui (Wake Forest University, USA); HA-PIAS1, HA-PIASxa, Real-time quantitative PCR HA-PIASxb, HA-PIAS3, HA-PIASy and GST-PIASy were Total RNA was isolated by Trizol kit (Invitrogen). RNA was provided by Xinhua Feng (Baylor College of Medicine, USA); treated with DNase (Promega, Madison, WI, USA). Com- pcDNA3-myc PIASy wt, pCMV-Flag-PIASy WT and pCMV- plementary DNA was synthesized using the cDNA synthesis Flag-PIASy Mut, a PIASy catalytic mutant, were provided by kit (Takara, Shiga, Japan) according to the manufacturer’s Shigeki Miyamoto (University of Wisconsin–Madison, USA). instructions. Fluorescence real-time RT-PCR was performed Myc-HIF2a was provided by Longsheng Lu (Texas Heart with the double-stranded DNA dye SYBRGreen PCR Core Institute, USA). HA-SUMO-1, GST-ODDpm and HRE-luc Reagents (PE Biosystems, Warrington, UK) using the ABI were previously described. Flag-HIF1a,Flag-HIF1a Mut, PRISM 7300 System (PerkinElmer, Torrance, CA, USA). HA-HIF1a,Flag-HIF1aK719T, Flag-HIF1a truncates (1-84, PCR was carried out in triplicate and standard deviations 85-210, 211-330, 1-330, 331-698, 699-826, 331-826), 3xmyc- representing experimental errors were calculated. All data were PIASy (100-202, 100-175, 125-202, 125-175, 150-202) were analyzed using ABI PRISM SDS 2.0 software (PerkinElmer). generated using standard cloning procedures and PCR-based Pairs of PCR primers used to amplify the target genes were as follows: hVEGF (forward: 50-TTTCTGCTGTCTTGGGTG mutagenesis. AD-HIF1a and BD-PIASy were generated based 0 0 on Mammalian Matchmaker two-hybrid system purchased from CATTGG-3 , reverse: 5 -ACCACTTCGTGATGATTCTGC CCT-30), mVEGF (forward: 50-TCACCAAAGCCAGCACA Clontech (Mountain View, CA, USA). 0 0 siRNA against PIAS1 (catalog no. M-008167-01), PIASxa/ TAGGAGA-3 , reverse: 5 -TTTCTCCGCTCTGAACAAGG CTCA-30), mGlut-1 (forward: 50-TCAACGAGCATCTTCG xb (catalog no. J-006033-05), PIAS3 (catalog no. J-006034-05), 0 0 PIASy (catalog no. J-005946-05), RanBp2 (catalog no. J- AGAAGGCA-3 , reverse: 5 -TCGTCCAGCTCGCTCTACA ACAAA-30), 18S (forward: 50-AGGCCCTGTAATTGGAAT 006044-05), Pc2 (catalog no. J-006035-05), MMS21 (catalog 0 0 0 no. L-018070-00) and nonspecific siRNA (catalog no. D- GAGTC-3 , reverse: 5 -GCTCCCAAGATCCAACTACGAG-3 ). 001810-01) were purchased from Dharmacon RNA Technol- Immunoprecipitation ogies (Lafayette, CO, USA). PIASy-Si1 (against human 6 PIASy1: 50-GTGCTCTACGGAAAGTACTT-30), PIASy-Si2 293 cells (5 Â 10 ) pellets were lysed in 1 ml of radioimmune (against human PIASy: 50-CAAGACAGGTGGAGTTGAT-30) precipitation assay buffer (50 mM Tris–HCl (pH 7.4), 1% and mPIASy-Si (against mouse PIASy: 50-GTGCTGTACGG Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS and GAAGTACTT-30) were generated based on Knockout RNAi- 350 mM NaCl) containing 1 mM phenylmethylsulfonyl fluoride, Ready pSIREN-RetroQ Vector system purchased from 1 mg/ml leupeptin, 10 mg/ml aprotinin, 1.5 mM pepstatin, 1 mM M N Clontech. Na3VO4 and 5 m -ethylmaleimide. For hypoxia treatment, HUVEC cell was purchased from ATCC. 293 cell was the cells were be treated by addition of MG132 (10 mM) for 4 h purchased from Invitrogen (Carlsbad, CA, USA). SENP1 before collection. Cell lysis was performed for 30 min on ice, MEF cells were previously described (Cheng et al., 2007). and DNA in the sample was sheared with a 22-gauge needle. After centrifugation at 20 000 g for 10 min at 4 1C, the supernatants were added to appropriate antibody coupled to Human colon cancer tissue specimens 20 ml of protein G-Sepharose beads (Amersham Pharmacia All the formalin-fixed and paraffin-embedded colon cancer Biotech, Piscataway, NJ, USA). The bead suspensions were tissue specimens were obtained from the colon cancer tissue rotated for 3 h at 4 1C. Beads were then washed five times with

Oncogene PIASy regulates HIF1a SUMOylation X Kang et al 5577 radioimmune precipitation assay buffer. The immunoprecipi- 0.5 ml medium were seeded onto the solidified gel. After tates were treated with 30 ml of 2% SDS treating solution incubation for 16 h, the endothelial tubes were assessed with a containing 5% b-mercaptoethanol and analyzed by western photomicroscope. blotting. Immunohistochemistry In vitro SUMOylation assay Tissue samples were fixed in buffered formalin and were In vitro SUMOylation kit was purchased from LAE Biotech embedded in paraffin. Antibodies were used for specific tissues International (Rockville, MD, USA). The reaction mixture immunostaining including mouse anti-human CD31 antibody contained ATP (2 mM), E1 (150 ng), GST-ODDpm (300 ng), (Dako Corp., Carpinteria, CA, USA), and goat anti-human GST-PIASy (25 ng) and E2 (amount as indicated) and reaction PIASy (Santa Cruz Corp.). Vectastain ABC system (Vector was carried out at 37 1C for 1 h. Laboratories, Burlingame, CA, USA) was used for immuno- histochemistry. FITC- or rhodamine-conjugated secondary antibodies were used for immunofluorescence detection. siRNA stably-transfected HUVEC cells and MEF cells The retrovirus containing PIASy siRNA or nonspecific siRNA infected HUVECs to generate PIASy1-Si1-HUVEC, PIASy-Si2-HUVEC or scramble HUVEC cells. All HUVECs Conflict of interest normally grow in a complete endothelial cell medium (ScienCell, Carlsbad, CA, USA). The retrovirus containing The authors declare no conflict of interest. mPIASy siRNA or scramble siRNA infected SENP1À/À MEF cells to generate mPIASy-Si-SENP1À/À MEF cell, or scramble SENP1À/À MEF cell after puromycin selection. Hypoxic Acknowledgements treatment (2% O2) was performed in a specially designed hypoxia incubator (Thermo Electron, Forma, MA, USA) and This work was supported in part by National Natural Science generated by flushing a mixture of air and N2 in combination Foundation of China (30772462, to JC; 30800579 to XK), with 5% CO2. National Key Scientific Program in China (2009CB918403, to JC), State Key Laboratory of Oncogenes and Related Genes In vitro tube formation of HUVEC (91-08-06, to JC), E-Institutes of Shanghai Municipal Educa- Each well of prechilled 24-well plates was coated with 100 ml tion Commission (E09013, to JC) and NIH grants (CA239520, per well of Matrigel (BD Biosciences, San Jose, CA, USA) and to ETHY). ETHY is the McNair Scholar of the Texas Heart incubated at 37 1C for 1 h. A total of 4 Â 104 HUVEC cells in Institute, St Luke Episcopal Hospital.

References

Bae SH, Jeong JW, Park JA, Kim SH, Bae MK, Choi SJ et al. (2004). Dawlaty MM, Malureanu L, Jeganathan KB, Kao E, Sustmann C, Sumoylation increases HIF-1alpha stability and its transcriptional Tahk S et al. (2008). Resolution of sister centromeres requires activity. Biochem Biophys Res Commun 324: 394–400. RanBP2-mediated SUMOylation of topoisomerase IIalpha. Cell Berta MA, Mazure N, Hattab M, Pouyssegur J, Brahimi-Horn MC. 133: 103–115. (2007). SUMOylation of hypoxia-inducible factor-1alpha reduces its Geiss-Friedlander R, Melchior F. (2007). Concepts in sumoylation: a transcriptional activity. Biochem Biophys Res Commun 360: 646–652. decade on. Nat Rev Mol Cell Biol 8: 947–956. Brahimi-Horn C, Mazure N, Pouyssegur J. (2005). Signalling via the Gong L, Kamitani T, Fujise K, Caskey LS, Yeh ET. (1997). hypoxia-inducible factor-1alpha requires multiple posttranslational Preferential interaction of sentrin with a ubiquitin-conjugating modifications. Cell Signal 17: 1–9. enzyme, Ubc9. J Biol Chem 272: 28198–28201. Branzei D, Sollier J, Liberi G, Zhao X, Maeda D, Seki M et al. (2006). Gruber M, Simon MC. (2006). Hypoxia-inducible factors, hypoxia, Ubc9- and mms21-mediated sumoylation counteracts recombino- and tumor angiogenesis. Curr Opin Hematol 13: 169–174. genic events at damaged replication forks. Cell 127: 509–522. Hari KL, Cook KR, Karpen GH. (2001). The Drosophila Su(var)2-10 Bruick RK, McKnight SL. (2001). A conserved family of prolyl-4- locus regulates chromosome structure and function and encodes a hydroxylases that modify HIF. Science 294: 1337–1340. member of the PIAS protein family. Genes Dev 15: 1334–1348. Carbia-Nagashima A, Gerez J, Perez-Castro C, Paez-Pereda M, Hay RT. (2005). SUMO: a history of modification. Mol Cell Silberstein S, Stalla GK et al. (2007). RSUME, a small RWD- 18: 1–12. containing protein, enhances SUMO conjugation and stabilizes Hochstrasser M. (2001). SP-RING for SUMO: new functions bloom HIF-1alpha during hypoxia. Cell 131: 309–323. for a ubiquitin-like protein. Cell 107: 5–8. Carmeliet P, Dor Y, Herbert JM, Fukumura D, Brusselmans K, Ivan M, Kondo K, Yang H, Kim W, Valiando J, Ohh M et al. (2001). Dewerchin M et al. (1998). Role of HIF-1alpha in hypoxia-mediated HIFalpha targeted for VHL-mediated destruction by proline apoptosis, cell proliferation and tumour angiogenesis. Nature 394: hydroxylation: implications for O2 sensing. Science 292: 464–468. 485–490. Jaakkola P, Mole DR, Tian YM, Wilson MI, Gielbert J, Gaskell SJ Carter S, Bischof O, Dejean A, Vousden KH. (2007). C-terminal et al. (2001). Targeting of HIF-alpha to the von Hippel–Lindau modifications regulate MDM2 dissociation and nuclear export of ubiquitylation complex by O2-regulated prolyl hydroxylation. p53. Nat Cell Biol 9: 428–435. Science 292: 468–472. Cheng J, Kang X, Zhang S, Yeh ET. (2007). SUMO-specific protease 1 Jackson PK. (2001). A new RING for SUMO: wrestling transcrip- is essential for stabilization of HIF1alpha during hypoxia. Cell 131: tional responses into nuclear bodies with PIAS family E3 SUMO 584–595. ligases. Genes Dev 15: 3053–3058. Comerford KM, Leonard MO, Karhausen J, Carey R, Colgan SP, Johnson ES, Gupta AA. (2001). An E3-like factor that promotes Taylor CT. (2003). Small ubiquitin-related modifier-1 modification SUMO conjugation to the yeast septins. Cell 106: 735–744. mediates resolution of CREB-dependent responses to hypoxia. Proc Kagey MH, Melhuish TA, Wotton D. (2003). The polycomb protein Natl Acad Sci USA 100: 986–991. Pc2 is a SUMO E3. Cell 113: 127–137.

Oncogene PIASy regulates HIF1a SUMOylation X Kang et al 5578 Kahyo T, Nishida T, Yasuda H. (2001). Involvement of PIAS1 in the Sapetschnig A, Rischitor G, Braun H, Doll A, Schergaut M, sumoylation of tumor suppressor p53. Mol Cell 8: 713–718. Melchior F et al. (2002). Transcription factor Sp3 is Kallio PJ, Okamoto K, O’Brien S, Carrero P, Makino Y, Tanaka H silenced through SUMO modification by PIAS1. EMBO J 21: et al. (1998). Signal transduction in hypoxic cells: inducible nuclear 5206–5215. translocation and recruitment of the CBP/p300 coactivator by the Schmidt D, Muller S. (2002). Members of the PIAS family act as hypoxia-inducible factor-1alpha. EMBO J 17: 6573–6586. SUMO ligases for c-Jun and p53 and repress p53 activity. Proc Natl Lando D, Peet DJ, Gorman JJ, Whelan DA, Whitelaw ML, Bruick Acad Sci USA 99: 2872–2877. RK. (2002). FIH-1 is an asparaginyl hydroxylase enzyme that Schofield CJ, Ratcliffe PJ. (2004). Oxygen sensing by HIF hydro- regulates the transcriptional activity of hypoxia-inducible factor. xylases. Nat Rev Mol Cell Biol 5: 343–354. Genes Dev 16: 1466–1471. Schwarz SE, Matuschewski K, Liakopoulos D, Scheffner M, Jentsch Mabb AM, Wuerzberger-Davis SM, Miyamoto S. (2006). PIASy S. (1998). The ubiquitin-like proteins SMT3 and SUMO-1 are mediates NEMO sumoylation and NF-kappaB activation in conjugated by the UBC9 E2 enzyme. Proc Natl Acad Sci USA 95: response to genotoxic stress. Nat Cell Biol 8: 986–993. 560–564. Matsuura T, Shimono Y, Kawai K, Murakami H, Urano T, Niwa Y Shao R, Zhang FP, Tian F, Anders Friberg P, Wang X, Sjoland H et al. (2005). PIAS proteins are involved in the SUMO-1 et al. (2004). Increase of SUMO-1 expression in response to modification, intracellular translocation and transcriptional repres- hypoxia: direct interaction with HIF-1alpha in adult mouse brain sive activity of RET finger protein. Exp Cell Res 308: 65–77. and heart in vivo. FEBS Lett 569: 293–300. Maxwell PH, Wiesener MS, Chang GW, Clifford SC, Vaux EC, Shuai K, Liu B. (2005). Regulation of gene-activation pathways Cockman ME et al. (1999). The tumour suppressor protein VHL by PIAS proteins in the immune system. Nat Rev Immunol 5: targets hypoxia-inducible factors for oxygen-dependent proteolysis. 593–605. Nature 399: 271–275. Stebbins CE, Kaelin Jr WG, Pavletich NP. (1999). Structure of the Ohh M, Park CW, Ivan M, Hoffman MA, Kim TY, Huang LE et al. VHL–ElonginC–ElonginB complex: implications for VHL tumor (2000). Ubiquitination of hypoxia-inducible factor requires direct suppressor function. Science 284: 455–461. binding to the beta-domain of the von Hippel–Lindau protein. Nat Takahashi Y, Kahyo T, Toh EA, Yasuda H, Kikuchi Y. (2001). Yeast Cell Biol 2: 423–427. Ull1/Siz1 is a novel SUMO1/Smt3 ligase for septin components and Pouyssegur J, Dayan F, Mazure NM. (2006). Hypoxia signalling in cancer functions as an adaptor between conjugating enzyme and substrates. and approaches to enforce tumour regression. Nature 441: 437–443. J Biol Chem 276: 48973–48977. Rogers RS, Horvath CM, Matunis MJ. (2003). SUMO modification of van Hagen M, Overmeer RM, Abolvardi SS, Vertegaal AC. (2010). STAT1 and its role in PIAS-mediated inhibition of gene activation. RNF4 and VHL regulate the proteasomal degradation of SUMO- J Biol Chem 278: 30091–30097. conjugated hypoxia-inducible factor-2alpha. Nucleic Acids Res 38: Sachdev S, Bruhn L, Sieber H, Pichler A, Melchior F, Grosschedl R. 1922–1931. (2001). PIASy, a nuclear matrix-associated SUMO E3 ligase, Wang GL, Jiang BH, Rue EA, Semenza GL. (1995). Hypoxia- represses LEF1 activity by sequestration into nuclear bodies. Genes inducible factor 1 is a basic-helix-loop-helix-PAS heterodimer Dev 15: 3088–3103. regulated by cellular O2 tension. Proc Natl Acad Sci USA 92: Saitoh H, Sparrow DB, Shiomi T, Pu RT, Nishimoto T, Mohun TJ 5510–5514. et al. (1998). Ubc9p and the conjugation of SUMO-1 to RanGAP1 Yeh ET. (2009). SUMOylation and De-SUMOylation: wrestling with and RanBP2. Curr Biol 8: 121–124. life’s processes. J Biol Chem 284: 8223–8227.

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

Oncogene