Hypoxia-Induced SUMOylation of E3 Ligase HAF Determines ......SMARTpool (Dharmacon L-017283, siHAF_1), and Silencer #15787 (siHAF_2; Ambion, Life Technologies). Hypoxic expo-sure was

Molecular and Cellular Pathobiology

Hypoxia-Induced SUMOylation of E3 Ligase HAFDetermines Specific Activation of HIF2 inClear-Cell Renal Cell CarcinomaMei Yee Koh1, Vuvi Nguyen2, Robert Lemos Jr1, Bryant G. Darnay2, Galina Kiriakova2,Mena Abdelmelek2, Thai H. Ho3, Jose Karam4, Federico A. Monzon5, Eric Jonasch4, andGarth Powis1

Abstract

Clear-cell renal cell cancer (CRCC) is initiated typically byloss of the tumor-suppressor VHL, driving constitutive acti-vation of hypoxia-inducible factor-1 (HIF1) and HIF2. How-ever, whereas HIF1 has a tumor-suppressor role, HIF2 plays adistinct role in driving CRCC. In this study, we show that theHIF1a E3 ligase hypoxia-associated factor (HAF) complexeswith HIF2a at DNA to promote HIF2-dependent transcrip-tion through a mechanism relying upon HAF SUMOylation.

HAF SUMOylation was induced by hypoxia, whereas HAF-mediated HIF1a degradation was SUMOylation independent.HAF overexpression in mice increased CRCC growth andmetastasis. Clinically, HAF overexpression was associatedwith poor prognosis. Taken together, our results show thatHAF is a specific mediator of HIF2 activation that is critical forCRCC development and morbidity. Cancer Res; 75(2); 316–29.�2014 AACR.

IntroductionThe hypoxia response promotes the adaptation to low oxygen

by shifting cells toward anaerobic metabolism, neovasculariza-tion, and resistance to apoptosis. Hypoxia occurs in all solidtumors and their metastases, and drives responses that contrib-ute to tumor aggressiveness, such as increased invasion, metas-tasis, and a poorly differentiated phenotype. This occurs largelythrough activation of the hypoxia-inducible factors (HIF), HIF1and HIF2, which are central players in the regulation of thephysiologic and pathophysiologic responses to hypoxia (1, 2).HIF1 and HIF2 are nonredundant and play unique and com-plementary roles during bone and vascular development, andalso in regulating cellular transcriptional responses to acute andchronic hypoxia (3, 4). In this regard, HIF1 and HIF2 can drivedistinct downstream target genes and also exhibit antagonism inregulating the hypoxia response (5). The highly divergent out-comes of HIF1 and HIF2 signaling on tumor growth andprogression occur despite their high degree of similarity instructure and mechanisms of regulation, and appear to be

dependent on hypoxic intensity and duration, and other iso-form specific HIF regulators that are beginning to be identified(2). We have established that the hypoxia-associated factor(HAF) can promote the switch from HIF1- to HIF2-dependentsignaling, that it achieves by selectively degrading HIF1a inde-pendently of oxygen or pVHL, and promoting HIF2a transacti-vation without affecting HIF2a levels (4, 6).

Clear-cell renal cell carcinoma (CRCC) is the most commontype of kidney cancer, and is highly refractory to standard che-motherapy and radiation. The etiology of CRCC is uniquelylinked to loss of the von Hippel–Lindau (pVHL) tumor-suppres-sor protein, wherebymore than 90%of cases of both sporadic andhereditary CRCC show pVHL deficiencies (7–9). pVHL is thesubstrate recognition component of the E3 ligase complex thattargets the oxygen labile HIF1a and HIF2a subunits for protea-somal degradation under aerobic conditions. Under hypoxicconditions, or in the presence of pVHL deficiency, HIF1a andHIF2a are stabilized, and enter the nucleus where they hetero-dimerize with HIF1b, forming the HIF1 or HIF2 transcriptionalcomplexes respectively, and activate the transcription of hundredsof genes critical for the adaptation to hypoxia, and for tumorprogression (1, 10). pVHL loss of function is a critical event forCRCC initiation, promoting the constitutive activation of HIF1and HIF2, which play a dominant role in the progression ofCRCC (11). Indeed, CRCC is one of the best-perfused of solidtumors due to the overproduction of HIF-dependent proangio-genic factors, VEGFAandPDGFB.Nevertheless, CRCC tumors stillexperience relatively low oxygen tensions due to the alreadylow physiologic oxygen tensions within the kidney (CRCC isbelieved to originate fromproximal tubular epithelial cells withinthe renal cortex), and to the inherent abnormality of the tumorvasculature (12–14).

Converging lines of evidence support a driving role for HIF2aand not of HIF1a in CRCC. First, although elevated HIF1a isapparent in the earliest preneoplastic lesions in VHL patients, the

1Sanford-Burnham Medical Research Institute, La Jolla, California.2Department of Experimental Therapeutics, The University of TexasMD Anderson Cancer Center, Houston, Texas. 3Department of Hema-tology/Oncology, Mayo Clinic Arizona, Scottsdale, Arizona. 4Depart-ment of GU Medical Oncology, The University of Texas MD AndersonCancer Center, Houston,Texas. 5Department of Pathology and Immu-nology, Baylor College of Medicine, Houston, Texas.

Note: Supplementary data for this article are available at Cancer ResearchOnline (http://cancerres.aacrjournals.org/).

Corresponding Author: Mei Yee Koh, Sanford-Burnham Medical ResearchInstitute, 10901 N Torrey Pines Road, La Jolla, CA 92037. Phone: 858-795-5349; Fax: 858-795-5490; E-mail: [email protected]

doi: 10.1158/0008-5472.CAN-13-2190

�2014 American Association for Cancer Research.

CancerResearch

Cancer Res; 75(2) January 15, 2015316

on May 6, 2021. © 2015 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

Published OnlineFirst November 24, 2014; DOI: 10.1158/0008-5472.CAN-13-2190

http://cancerres.aacrjournals.org/

appearance of HIF2a is associated with increased dysplasia andcellular atypia (15, 16). Hence, CRCC cells and tumors can besubdivided into two subtypes: those that express both HIF1a andHIF2a (pVHL mutant or pVHL wild-type), or those that expressHIF2a exclusively (pVHL mutant only; refs. 17, 18). Second,overexpression of HIF2a promotes, whereas overexpression ofHIF1a inhibits, CRCC growth (19–21). Third, inhibition ofHIF2a by shRNA is sufficient to suppress the growth of pVHL-null CRCC cells (19, 22). Fourth, type IIB pVHL mutants, asso-ciatedwithhigh risk ofCRCC, retain some ability to downregulateHIF1a, but have reduced ability to downregulate HIF2a, com-pared with type IIA pVHL mutants, associated with low risk ofCRCC (23, 24). Taken together, the data suggest that theremay bea temporal nature toHIF activation in CRCC, whereby VHL loss isassociated with early elevation of HIF1a, which then shifts toHIF2a, whichpromotes dysplasia andCRCCprogression. There isalso increasing evidence implicating HIF1A as a kidney cancer–suppressor gene, particularly in advanced CRCC: loss of hetero-zygosity in chromosome 14q in the locus spanning HIF1A hasbeen reported in approximately 40% of human CRCC, whereashomozygous deletion of HIF1A was detected in approximately50% of CRCC cell lines (25, 26). Although the mechanism of theshift to HIF2 is unclear, the specific protumorigenic effect ofHIF2a in CRCCmay be due its increased potency compared withHIF1a in driving protumorigenic factors such as cyclin D1, TGFa,and VEGFA, and in potentiating c-Myc activity, which in contrastis inhibited by HIF1a (18, 21).

The HAF (located at 11q13.1) is a mediator of the switch fromHIF1a to HIF2a, by selectively degrading HIF1a, and promotingHIF2a transactivation (4). HAF overexpression enhances theability of glioblastoma cells (which express both HIF1a andHIF2a) to initiate tumors as intracranial xenografts in mice.However, HAF overexpression decreases xenograft tumor growthof HT29 colon carcinoma cells that only express HIF1a (6).Hence, HAF may either inhibit or promote tumor progression,depending on the dominant HIFa isoform expressed in a givencellular context. However, themechanism bywhichHAF elicits itsselective effects on HIF1a and HIF2a in the context where bothHIFa isoforms are present remains unknown.

Here, we elucidate the mechanism and regulation of HAF-mediated HIF2 transactivation. We show that HAF promotes thetranscription of a subset of HIF2 target genes by binding to a DNAconsensus site located within close proximity to the hypoxia-responsive element (HRE), thus forming a transcriptional complexwith HIF2a to drive transcription. Significantly, we demonstratethat the ability of HAF to bind and transactivate HIF2a is depen-dent upon HAF SUMOylation, which is induced by exposure tohypoxia. In contrast, the ability of HAF to bind and degrade HIF1ais independent of HAF SUMOylation. Thus, in the context ofpVHL loss that results in the constitutive stabilization of HIF1and HIF2, HAF preferentially promotes HIF2-specific activation,and patients with CRCC with high HAF show significantlydecreased progression-free survival (PFS) than thosewith lowHAF.This suggests that HAF may be the determinant of HIF2-specificactivation in CRCC, thus providing a novel avenue for therapy.

Materials and MethodsTissue culture

A498, 786-0, PANC-1, MIAPaCa-2, and ACHN cells were fromATCC andweremaintained in RPMI (A498) andDMEM (ACHN,

786-0, PANC-1, MIAPaCa-2; Invitrogen Corp., Life Technolo-gies), supplemented with 10% fetal bovine serum. The identitiesof all cell lines were confirmed by the Molecular CytogeneticsFacility at MD Anderson Cancer Center (MDACC; Houston, TX)using short tandem repeat (STR)DNA fingerprinting upon receiptfromATCC.Cellswere frozen, thawed, and retested after 3monthsin culture, after which cells were discarded and a fresh vial thawed.Hypoxic incubations (1% O2) were performed using the InVivo2Hypoxia Workstation (Biotrace International Inc.).

Plasmid construction and transfectionsStable cell lines overexpressing HAF and HAF double mutant

(DM; HAF K94R K141R), or shHIF2 were generated by retroviralinfection using pMX-IRES-GFP or pSUPER, respectively (27). ThepCMVFLAG14 construct was used for transient transfections withHAF or HAF DM (6). The P117 Luc reporter was generatedby cloning the P117 sequence (28) upstream of luciferase anda minimal CMV promoter (pGL4.17; Promega Corp.). P117 HREcontaining the HRE and HIF ancillary sequence (29) was gener-ated by cloning the HRE downstream of P117 (SupplementaryMethods). HAF–GAL4 fusion protein was generated by ligatingfull-length FLAG–HAF downstream of the GAL4 DNA-bindingdomain (DBD) using the pBIND plasmid from the CheckmateMammalian Two-Hybrid System (Promega). Transactivationactivity was confirmed using the pG5 Luc vector containing fiveGAL4 binding sites upstream of luciferase, normalized to intronicRenilla luciferase encoded within the pBIND vector. GAL4-VP16(pBIND-id and pACT-MyoD) was used as a positive controlaccording to the manufacturer's protocol. Transient transfectionswere performed using Lipofectamine 2000 (Invitrogen) and cellswere assayed 48 hours after transfection. HIF2a siRNA duplexeswere from Dharmacon (GE Healthcare): siGENOME D-004814-02 (DNA oligos were ligated into pSUPER to generate shH2),D-004814-03 (siH2_2). HAF (SART1) siRNA was ON-TargetplusSMARTpool (Dharmacon L-017283, siHAF_1), and Silencer#15787 (siHAF_2; Ambion, Life Technologies). Hypoxic expo-sure was at 1% O2, performed 24 hours after transfection, andharvested after 16 hours unless otherwise indicated.

TaqMan quantitative RT-PCRTotal RNAwas isolated using the RNEasy Kit withDNAse I step.

TaqMan qRT-PCRwas performed using the ABI 7300 systemwithOne-Step RT-PCRMasterMix Kit and predesigned primer/probes,and normalized to b2-microglobulin as previously described (6).Statistical significance was determined using MS Excel (t test).

Western blotting, immunoprecipitation, andimmunohistochemistry

Western blotting was performed using HIF2a (NB100-122;Novus Biologicals), actin (Santa Cruz Biotechnology Inc.),HIF1a (Cell Signaling Technology), SUMO-1 (Cell SignalingTechnology; #4930S), SUMO-2/3 (ab81371; Abcam), and HAFantibodies (6, 30). Immunoprecipitation to show SUMOylationof HAF was performed using RIPA buffer (with 20 mmol/LNEM) with sonication. Denaturing immunoprecipitation wasperformed by boiling immunoprecipitated proteins conjugatedto protein A beads in 50 mL 1% SDS solution for 5 minutes,resuspending the supernatant in 1 mL of RIPA buffer, andrepeating immunoprecipitation with fresh protein A beadsovernight. Immunocytochemistry was performed using purifiedmouse monoclonal anti-HAF antibody raised against GST-

HAF SUMOylation Drives HIF2 in CRCC

www.aacrjournals.org Cancer Res; 75(2) January 15, 2015 317




HAF432-800 (Biogenes GmbH) and rabbit polyclonal anti-HIF2a(NB 100-122), with Cy3 goat anti-mouse, and Alexa Fluor 488chicken anti-rabbit secondary antibodies, respectively (Invitro-gen). Images were taken using the Nikon A1Rsi confocal micro-scope at a single plane and Z-stack (0.8–1.0 mmol/L). Immu-nohistochemistry (IHC) was performed using HIF2a or HAFmAb antibody, using citrate buffer antigen retrieval (Leicamicrosystems), and the Bond Polymer Intense Detection Kit(Leica Microsystems).

Luciferase assaysThese were performed using the Luciferase Dual-Glo Assay

System (Promega). Activities of reporters were normalized toconstitutive Renilla luciferase (b-actin).

Chromatin immunoprecipitationChromatin immunoprecipitation (ChIP) assays were per-

formed using 786-0 or PANC-1 cells stably expressing HAF,HAF DM, or shHIF2 and the respective empty vector controlspMX-IRES-GFP (Vec) or pSUPER, using the EZ ChIP Kit (EMDMillipore). ChIP antibodies were HIF2a (NB100-122), HAF(6), FLAG (Sigma-Aldrich), and control IgG (Bethyl Biolabs).Semiquantitative PCR was performed using standard protocolswith indicated primer pairs (Supplementary Methods). Quan-titation was performed using the AlphaView Q Imaging soft-ware (Proteinsimple).

In vivo tumor growth studiesNu/nu nude mice (10 per group) were injected subcutaneously

with 107 786-0 cells. Tumor diameters were measured twiceweekly at right angles (dshort and dlong) using electronic calipersand tumor volumes were calculated by the formula volume ¼(dshort)

2 � (dlong)2 (31). Orthotopic renal models were generated

by injection of 5 � 106 786-0 cells (in 50 mL saline) in the rightrenal subcapsule of nude mice (6 per group) as described (32).Tumors were harvested after 2 months. Tumor weight was calcu-lated by subtracting the weight of the normal left kidney from theweight of the implanted right kidney.

CRCC tumor microarray preparation and analysisThe tumormicroarray (TMA) comprised pretreatment nephrec-

tomy specimens of patients with metastatic grade 3/4 CRCCtreated with sorafenib or sorafenib plus interferon (33). Allpatients showed haploinsufficiency at 3p, indicating pVHL defi-ciency, although formal sequencing of VHL was not undertaken.Three cores per patient in replicate arrays were stained for HIF1a(1:100; 2015-1; Epitomics), HIF2a, or HAF. Image capture andanalysis for HIF1a and HIF2a were performed using the Ariolsystem (Applied Imaging). Cores were scanned using TMA Nav-igator software (Applied Imaging). Regions of viable tumor weregated, excluding areas of nonviable tumor and nontumor tissue.Image capture for nuclear HAF was performed using the Vectramultispectral slide analysis system, and analyzed using theinForm image analysis software (both from PerkinElmer, CaliperLife Sciences). The percentage tumor cell involvement of HIF1a,HIF2a, or HAF was determined by a pathologist (F.A. Monzon).

Statistical analysisThese were performed using IBM SPSS (Vs.19) comparing

nuclear HAF staining and PFS with patients stratified to groupsexpressing low and high HAF, respectively. The log-rank test and

Cox regression were used to compute the P value for Kaplan–Meier curves and hazard ratios, respectively.

ResultsHAF promotes the transcription of HIF2 target genes

CRCC cells offer a unique opportunity to study the effects ofHAF on HIF2a, independently of HIF1a, as many CRCC cells donot express HIF1a (34). CRCC also provides a setting in whichHIF2a has a clear driving role in tumor progression. To identifyHIF2 target genes in 786-0 CRCC cells (pVHL deficient, consti-tutive HIF2a, no HIF1a), we stably expressed HIF2a shRNA(shHIF2) or empty vector (pSUPER) in these cells. Knockdownof HIF2a in 786-0 shHIF2 cells decreased the transcription of apanel of HIF2 genes compared with the empty vector control,confirming that these genes were HIF2 dependent (Fig. 1A).Similar results were obtained by transient transfection of parental786-0 cells with a different siRNA duplex targeting HIF2a andnormalized to a nontargeting control siRNA (siH2_2; Fig. 1A). Toinvestigate the role of HAF specifically on HIF2-dependent tran-scription, we overexpressed HAF in the 786-0 cells and found thatit induced the transcription of the aforementioned HIF2 targetgenes without affecting HIF2a levels (Fig. 1B). Among thesegenes, we observed that the greatest induction due to HAF over-expression was in the levels of the pluripotency genes OCT-3/4,SOX2, and NANOG. The induction of these genes was alsoobserved in 786-0 cells stably overexpressing HAF, and thisoccurred in both normoxia and hypoxia (Fig. 1C). Similar resultswere obtained using A498 CRCC cells (pVHL-deficient, constitu-tive HIF2a; Supplementary Fig. S1A and S1B). Consistent withconstitutive HIF2a expression due to loss of pVHL, the 786-0cells did not show any general induction of HIF2 target genesupon exposure to 16 hours of hypoxia. However, we did observesignificant temporal inductions of OCT-3/4, SOX2, and NANOGafter 8 hours of exposure to hypoxia (Fig. 1D), suggestingan alternate, pVHL-independent hypoxia-sensing mechanism forthese genes. To investigate the impact of HAF overexpression incells that have functional pVHL, and therefore hypoxia-inducibleHIF1a and HIF2a, we overexpressed HAF in the ACHN CRCCcells. Here, HAF overexpression decreased the levels of HIF1aprotein without affecting HIF2a levels, and attenuated the hyp-oxic induction of HIF1-specific target genes, BNIP3, BNIP3L(NIX), and PGK1 (Fig. 1E and F). HAF overexpression alsosignificantly induced transcription of the HIF2 target genesOCT-3/4, SOX2, NANOG, and VEGFA to a level above thatinduced by hypoxia (Fig. 1G). Hence, HAF promotes the shifttoward HIF2-dependent transcription in both pVHL wild-typeand pVHL-deficient CRCC cells.

HAF promotes the binding of HIF2a to the HRETo determine the mechanism by which HAF promotes HIF2

transcription, we investigated the impact of HAF overexpres-sion on HIF2a DNA binding activity within the promoter ofVEGFA. HAF has been shown to bind the VEGFA promoter ata site located approximately 600 bp downstream of the HRE(Fig. 2A, i; ref. 35). Using specific primers flanking the VEGFAHRE in 786-0 cells (Fig. 2A), we show that HAF overexpres-sion significantly increased the binding of HIF2a to the HRE(Fig. 2A, ii, left, 2C). In addition, HAF itself bound the HREbut binding was reduced when HIF2a was knocked down,suggesting that HAF binding to the HRE is HIF2a-dependent

Koh et al.

Cancer Res; 75(2) January 15, 2015 Cancer Research318




F G

A B i) ii)

C D E

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

NANOGSOX2OCT-3/4VEGFMMP9HIF2A

No

rmal

ized

fo

ld c

han

ge

Vector

HAF

*

*

**

0

10

20

30

40

50

60

70

80

90

100

NANOGSOX2OCT-3/4

No

rmal

ized

fo

ld c

han

ge

Nor vecNor HAFHyp vecHyp HAF

HIF2 targets

*

*

*

**

*

0

1

2

3

4

5

6

7

PGK1NIXBNIP3

No

rmal

ized

fo

ld c

han

ge

Nor vec

Nor HAF

Hyp vec

Hyp HAF

HIF1 targets

*

*

*

*

*

Vec HAF Vec HAFNor Hyp

HIF1a

HIF2a

Actin

HAF

0

5

10

15

20

25

OCT-3/4 SOX2 NANOG

No

rmal

ized

fold

ch

ang

e

Nor vec

Nor HAF

Hyp vec

Hyp HAF

* *

*

* * *

FLAG-HAF

HAFVec

HAF

HIF2a

Actin

0

0.2

0.4

0.6

0.8

1

1.2

VEGFANANOGSOX2OCT-3/4MMP9HIF2A

No

rmal

ized

fo

ld c

han

ge

Control shH2 siH2_2

** * * *

*

*

0

1

2

3

4

5

6

HAF HIF2A MMP9 OCT-3/4 SOX2 NANOG VEGFA

No

rmal

ized

fold

ch

ang

e

Normoxia

Hyp 8 h

Hyp 24 h

*

0

0.5

1

1.5

2

2.5

3

VEGFA

No

rmal

ized

fold

ch

ang

e

*

*

Figure 1.A, the effect of HIF2a knockdown using stable expression of HIF2 shRNA or transient transfection with an alternative HIF2a siRNA duplex on levels of HIF2atranscript and a panel of HIF2 target genes as determined by qRT-PCR normalized to stably or transiently expressed nontargeting controls, respectively. B, the effectof transient HAF overexpression in 786-0 cells on HIF2a protein levels determined by Western blotting (i) and transcription of HIF2a and a panel of HIF2target genes as determined by qRT-PCR normalized to vector control (ii). C, the effect of stable HAF overexpression in 786-0 cells on transcription of a subset of HIF2target genes in both normoxia and hypoxia as determined by qRT-PCR normalized to vector cells in normoxia. D, qRT-PCR showing effect of duration ofhypoxic exposure on the transcription of the panel of HIF2 target genes. Data shown are themean of at least two independent experiments� SE; � , P <0.05. E and F,the effect of HAF overexpression in ACHN cells on HIF1/2a protein levels by Western blotting (E) and HIF1 (F), or HIF2-dependent target genes (G) in normoxiaand hypoxia. Data are representative of two independent experiments � SE; �, P < 0.05.






(Fig. 2A, ii, right; 2E). Using specific primers flanking the HAFsite in the VEGFA promoter, we confirmed that both endoge-nous and overexpressed HAF bound to the HAF site (Fig. 2A, iii,

left; 2D and F). HAF binding to this site was unchanged whenHIF2a was knocked down, suggesting that HAF binding to theHAF site is HIF2a independent (Fig. 2A, iii, right; 2G). In

Vec HAF Vec HAF Vec HAF Vec HAF

Input IgG IP HAF IP HIF2α

pS shHIF2 pS shHIF2 pS shHIF2

Input IgG IP HAF

Vec HAF Vec HAF Vec HAF Vec HAF

IP HAF IP HIF2αInput IgG

pS shHIF2 pS shHIF2 pS shHIF2

Input IgG IP HAF

A VEGFA promoter

HRE HAF site

–974 –507

..catacgtgggctcc.. cccccagcccca+1

ATG(i)

(ii) HRE

(iii) HAF site


Vec HAF Vec HAF Vec HAF Vec HAF pS shHIF2 pS shHIF2 pS shHIF2

Input IgG IP HAF


Vec HAF Vec HAF Vec HAF Vec HAF pS shHIF2 pS shHIF2 pS shHIF2

Input IgG IP HAF

B OCT-3/4 promoter

–2557 –1956 –1510 –130 +1

(i)

ATGHRE (putative)

–2547 gcccctccccct …aggtgccgtgatg..

HAF site

–2488 CR4 CR3 CR2 CR1

HRE

(ii) CR3 (HRE; 100 bp)

(iii) CR4 (HAF site and putative HRE; 100 bp)

0

1

2

3

4

5

6

7

8

Vec HAF Vec HAF

VEGFHRE

OCT4HRE

Nor

mal

ized

inte

nsity

IP HIF2A

0

0.5

1

1.5

2

2.5

3

3.5

4

Vec HAF Vec HAF

VEGFHRE

OCT4HRE

Nor

mal

ized

inte

nsity

IP HAF

0

0.2

0.4

0.6

0.8

1

1.2

pS

shH

2

pS

shH

2

VEGFHRE

OCT4HRE

Nor

mal

ized

inte

nsity

IP HAF

0

0.5

1

1.5

2

pS

shH

2 pS

shH

2

VEGFHAF

OCT4HAF

Nor

mal

ized

inte

nsity

IP HAFGC D E F

00.5

11.5

22.5

33.5

44.5

5

Vec HAF Vec HAF

VEGFHAF

OCT4HAF

Nor

mal

ized

inte

nsity

IP HAF

*

*

*

*

*

* *

Figure 2.A, i, schematic of regulatory regions in VEGFA promoter showing the HRE and the HAF DNA binding site. Numbers indicate nucleotides with the transcription start siteindicated by þ1. Arrows, binding location of primers used for semiquantitative PCR. A, ii and iii, the representative results of ChIP assay followed by semiquantitativePCR (30 cycles) showing the effects of HAF overexpression (left) and stable HIF2a knockdown (right) on binding of HAF or HIF2a to the HRE (ii) and HAF site(iii) in the VEGFA promoter. B, i, schematic of theOCT-3/4 promoter showing CRs 1–4, HREs, and the putative HAF binding site. ChIP assays were performed as in A toshow the effects ofHAF overexpression (left) andHIF2a knockdown (right) onbinding ofHAForHIF2a to theHREwithin CR3 (ii) andputativeHAFbinding site andHREwithin in CR4 (iii). C–G, quantitation of ChIP results obtained from three independent experiments. Results are the mean � SE; �P < 0.05. IP, immunoprecipitation.

Koh et al.





contrast, we did not detect any binding of HIF2a to the HAF site(Fig. 2B, iii, left).

To investigate another HIF2-dependent target gene, we exam-ined the promoter sequence of the stem cell factor,OCT-3/4. Fourconserved regions (CR) between the promoters of the human,mouse, and bovine orthologs of OCT-3/4 have been reported(36). Although six putative HREs have been identified withinthese CRs, only two (CR3 and CR4) have been described to bindHIF2a (Fig. 2B, i; ref. 37). We found that endogenous HIF2abound the OCT-3/4 promoter within the HRE on CR3, but notwithin CR4, and the binding of HIF2a to CR3 was significantlyincreased when HAF was overexpressed (Fig. 2B, ii, iii; 2C).Similar to the VEGFA promoter, HAF overexpression increasedits recruitment to the CR3 HRE, but its binding was decreasedwhen HIF2a was knocked down, suggesting, again, that thebinding of HAF to the HRE is HIF2a dependent (Fig. 2B, ii,right; 2E). HIF2a binding to the HRE in CR4 was undetectable,and hence the impact ofHAFoverexpression onHIF2abinding to

this site could not be assessed (Fig. 2B, iii, left). Hereafter, theHREwithin CR3 will be referred to as the OCT-3/4 HRE. Because HAFbinds to similar sites (CCCCCACCCCC and CCCCCAGCCCC)within the EPO and VEGFA promoters, respectively (35, 38), wesearched the CRs within the OCT-3/4 promoter for putative HAFbinding sites based on the CCCCRRCCCC motif deduced fromthe two HAF binding sites. We identified a putative HAF site(CCCCTCCCCC) within a highly CR of CR4, known as the OCT-3/4 distal enhancer (39). Using ChIP primers flanking this region,we observe binding of HAF to CR4, which was not affected byHIF2a knockdown, suggesting thatHAF binds to CR4 in aHIF2a-independent manner (Fig. 2B, iii, right, 2G).

To evaluate the robustness of the data, we quantitated andcombined ChIP results from three independent experiments.Hence, we confirm that HAF overexpression promotes a signifi-cant increase in the binding of HIF2a to the HREs of both VEGFAand OCT-3/4 (CR3), and that HAF itself is also recruited to theHRE in a HIF2a-dependent manner (Fig. 2C–E). In addition, we

0

0.5

1

1.5

2

2.5

3

3.5

4

0 Vec HAF Vec HAF

P117 P117D

RL

U

786-0

CMVmPHAF site

Luc

P117

A

B C D

CMVmPHAF site HRE

P117 HRE

LucE

F G H

HAF siteHRE

HIF2α HIF2 targets

*

I

0

0.5

1

1.5

2

2.5

3

3.5

4

Vec HAF Vec HAF

Nor Hyp

P11

7 H

RE

RL

U

*

*

*

0

0.5

1

1.5

2

2.5

Vec HAF Vec HAF

Nor Hyp

P11

7del

HR

E R

LU

*

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

RL

U

*

*

*

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

P117 +HRE

P117 +HRE del

P117del+ HRE

P117del+ HREdel

RL

U

Reporter construct

Nor

Hyp

HAF

28

26

24

22

6

4

2

0

RL

U

GAL4alone

pG5

GA

L4

GA

L4-H

AF

GA

L4-V

P16

Figure 3.A, schematic showing the P117 reporterconstruct containing the HAF siteinserted upstream of a CMV minimalpromoter and luciferase. B, effects ofHAF overexpression on luciferaseactivity of cotransfected P117 Luc or P117containing a deletion of the HAF site(P117D) in 786-0 cells in normoxia.Results are given as relative light units(RLU) of firefly luciferase normalized to acotransfected constitutive Renillaluciferase construct. � , P <0.05. C, effectsof HAF or HIF2a (siH2) knockdown usingtwo different siRNA duplexes on activityof P117 in 786-0 cells in normoxia. D,transcriptional activity of the pG5luciferase reporter when cotransfectedwith the GAL4 DBD alone, GAL4 DBD–HAF fusion, or positive controlGAL4DBD–VP16. Results given as RLU offirefly luciferase normalized to Renillaencoded in the GAL4 DBD (pBIND)vector. E, schematic showing the P117HRE reporter construct containing theP117 sequence in tandem with the HREfused to luciferase. F and G, the effect ofHAF overexpression in ACHN CRCC cellswhen cotransfected with P117 HRE Luc(F) or P117del HRE (G) in normoxia andhypoxia. Results are given as RLU ofluciferase normalized to a cotransfectedconstitutive Renilla luciferase construct.� , P < 0.05. H, effect of deletion of theconsensus HAF site or HRE site (HAFdelor HREdel, respectively) on the hypoxiainducibility of P117 HREwhen transfectedinto ACHN cells and exposed tonormoxia or hypoxia. Data given as RLUas in F and G, and normalized to P117HREin normoxia. � , P <0.05. I, cartoondepicting the formation of HAF/HIF2transcriptional complex that promotesHIF2-dependent transcription.






HIF1α

HIF2α

HAF

Actin

Vec HAF DM Vec HAF DM Vec HAF DM Vec HAF DMNor Nor +MG132 Hyp Hyp +MG132

0

0.5

1

1.5

2

2.5

IP HIF2 IP FLAG

No

rmal

ized

inte

nsi

ty

VecHAFDM

OCT-3/4 HRE

* *

A i ii B

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

IP FLAG

No

rmal

ized

inte

nsi

ty

VecHAFDM

OCT-3/4 HAF*

IP FLAG anti-SUMO1

SUMO1-HAF

FLAG-HAF100

150

Vec HAF DM 100

FLAG-HAF

HIF2α

HAF

Actin

Vec HAF DM Vec HAF DM Vec HAF DM Nor Hyp (4 h) Hyp (24 h)

*

*

0

0.5

1

1.5

2

2.5

OCT-3/4 SOX2

No

rmliz

ed fo

ld c

han

ge

Vec

HAF

DM

C D E

K*

P11

7 H

RE

(R

LU

)

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

Vec HAF DM

0

0.5

1

1.5

2

2.5

Vec HAF DM

P11

7 H

RE

RL

U

*

L M N

%O2 20 3 1 20 3 1 20 3 1

FLAG-HAF100

150

Vec HAF DM

IP FLAG anti-SUMO1

100 FLAG-HAF

SUMOylation (Hypoxia)

HAF site

Baseline transcription

HAF

HAF siteHRE

HIF2αHIF2 targets

S

Maximal induction

HAF

100

150

100

50

FLAG-HAF(SUMO1)-HAF(SUMO1)2-HAF

FLAG-HAF

igG heavy chain

Vec HAF DM

IP FLAG, boil in 1% SDSIP FLAG anti-SUMO1

SUMO1-HAF

150

100

(h) 1%O2 0 4 24 0 4 24 0 4 24Vec HAF DM

FLAG-HAF

IP FLAG anti-SUMO1

100 FLAG-HAF

SUMO1-HAF

HAF

SUMO1

100

150

100

IP HAF anti-SUMO1

IP IgG IP HAFNor Nor Hyp

NS

HIF2αNS

Vec HAF DM

100

IP HIF2α anti-FLAG

FLAG-HAF

100

IP HIF2α anti-HAF

HAF

100 HIF2α

NS

50 Heavy chain

100

IP IgG IP HAFNor Nor Hyp

Nor Hyp Nor Hyp Nor HypVec HAF DM

FLAG-HAF

HIF2αNS

IP HIF2α anti-FLAG

100

100

*

0

1

2

3

4

5

6

GA

L4

GA

L4-H

AF

GA

L4-D

M

GAL4 pG5

RL

U

S

F G H i ii

I J

Koh et al.





confirm the binding of both endogenous and overexpressed HAFto the HAF site within VEGFA, and to its putative binding sitewithin OCT-3/4 (CR4), both of which occur independently ofHIF2a (Fig. 2F and G).

HAF binding to DNA cooperates with the HRE for maximalhypoxic induction

To determine whether HAF binding to its DNA binding siteconfers transcriptional activation, we used theminimal promoterof EPO, P117 (40), which contained a HAF binding site (Sup-plementary Materials) fused to a CMV minimal promoterupstream of luciferase (P117 Luc). HAF overexpression signifi-cantly increased P117 Luc activity in 786-0, but did not affect theactivity of P117del, which contained the flanking sequences ofP117, and a deletion of the HAF binding sequence(CCCCCACCCCC; Fig. 3B). Similar results were obtained usingPANC-1 cells (Supplementary Fig. S2A). Furthermore, P117 Lucactivity was significantly decreased in the presence of HAF siRNA,butwas not affected by transfectionwithHIF2a siRNA, suggestingthat HAF binding to its site confers transcriptional activity inde-pendently ofHIF2a (Fig. 3C). Efficiencyof siRNAknockdownwasverified using Western blotting (Supplementary Fig. S2B). Todetermine whether HAF on its own is sufficient to confer tran-scriptional activation, we used the yeast GAL4/UAS system. TheGAL4 DBD binds a consensus UAS site on DNA, but is unable toactivate transcription unless fused to a transactivator, such as theherpes simplex virus protein, VP16. When we generated aGAL4DBD–HAF fusion, we observed a significant induction ofluciferase from a cotransfected 5xUAS domain containing lucif-erase reporter construct, compared with transfection with theGAL4DBD alone, suggesting that HAF is sufficient to confertranscriptional activation when tethered to DNA (Fig. 3D). Itshould be noted that the HAF-induced increase in luciferase wassubstantially lower than that with GAL4-VP16, which has beendescribed as an unusually potent transcriptional activator, pos-sibly due to its viral origin (41).

To assess the contribution of HAF and HIF binding to hypoxia-induced gene transcription, we inserted theHRE containedwithinthe EPO enhancer in tandem with P117 Luc, to generate P117HRE Luc (Fig. 3E). Using ACHN CRCC cells expressing wild-type

pVHL for hypoxia-inducibility, we found that cotransfection ofP117 HRE with HAF in normoxia, or transfection with P117 HREalone with exposure to hypoxia, resulted in significant inductionsof luciferase activity (1.5-fold each), compared with P117 HRE innormoxia (Fig. 3F). When we cotransfected P117 HRE with HAFand also exposed cells to hypoxia, we obtained a 3-fold increasein luciferasewhen comparedwith P117HRE in normoxia, suggest-ing that the HAF site and the HRE both contribute to hypoxia-induced gene transcription (Fig. 3F). In contrast, cotransfectionof HAF with P117del HRE (P117 HRE containing a deletion ofthe HAF binding sequence), did not significantly affect luciferaseactivity, suggesting that the binding of HAF to the HAF site isrequired for its transcriptional activity (Fig. 3G). To address theeffects of potential nonspecific activity conferred by flankingsequences, we also evaluated the transcriptional activities of con-structs containing various combinations of the deletion of the HAFsite (P117del), or of the HRE (HREdel). We found that deletion oftheHRE but not of theHAF site abrogated the hypoxia-inducibilityofP117HRE, suggesting that theHAFsitedoesnot contribute to thehypoxia-inducibility of transcription, the latter of which is entirelydependent on the presence of the HRE (Fig. 3H).

Taken together, the results suggest that HAF, by binding to aconserved HAF DNA binding sequence within close proximity tothe HRE, cooperates specifically with HIF2a to promote HIF2-dependent transcription (Fig. 3I).

HAF SUMOylation is required for HIF2 activationHAF has been reported to be a target of conjugation with

small ubiquitin-related modifiers (SUMO), and two SUMOyla-tion sites (K94 and K141) have been confirmed using massspectrometric approaches (42, 43). Protein SUMOylation anddeSUMOylation can regulate a variety of diverse outcomes,including changes in protein stability, localization, transcrip-tional activation, and protein–protein interactions (44). Todetermine whether the reported sites are indeed the sites ofHAF SUMOylation, we overexpressed FLAG-tagged wild-typeHAF (referred to as HAF), or FLAG–HAF containing lysine toarginine mutations of the residues known to be SUMOylated(K94R and K141R), referred to as HAF DM. We observedconstitutive SUMO-1 conjugation of overexpressed wild-type

Figure 4.A, i, Western blotting showing immunoprecipitation (IP) of FLAG from 786-0 cell lysates overexpressing FLAG–HAF or a SUMOylation-deficient mutant of HAF(FLAG–HAF DM), then probed for SUMO-1. A, ii, Western blotting showing effect of denaturing immunoprecipitation on HAF SUMOylation. Lysates wereimmunoprecipitated with anti-FLAG, boiled in 1% SDS, and supernatant was subjected to a second IP with anti-FLAG, then probed for SUMO-1. Molecular weightsconsistent with unconjugated HAF or with HAF potentially conjugated with one or two SUMO-1 molecules are indicated. B, Western blotting showing SUMO-1modification of immunoprecipitated FLAG–HAF in 786-0 cells after 4 hours of exposure to indicated oxygen tensions. C, time course of HAF SUMOylation inresponse to hypoxia. Stable FLAG–HAF expressing 786-0 cells were exposed to the indicated durations of 1% O2 and analyzed as in B. D, Western blottingshowing levels of FLAG–HAF, FLAG–HAF DM, and HIF2a in normoxia, and after exposure to indicated durations of hypoxia in 786-0 cells. E, Western blottingshowing SUMO-1 modification of endogenous HAF. Endogenous HAF was immunoprecipitated from lysates of ACHN cells after 8 hours of hypoxia andprobedwith SUMO-1. F andG, impact of transient HAForHAFDMoverexpressiononP117HRE luciferase activity (F) or transcription ofHIF2 target genesOCT-3/4 andSOX2 in 786-0 cells (G) in normoxia determined by qRT-PCR normalized to vector control. H, i, top, Western blotting showing immunoprecipitation of HIF2afrom lysates of 786-0 cells (exposed to 4 hours of hypoxia) stably expressing FLAG–HAF or FLAG–HAF DM, then probed with FLAG. Bottom, Western blottingshowing immunoprecipitation using 786-0 cell lysates, as in top, of cells grown in normoxia or exposed to 4 hours of hypoxia. ii, Western blotting showingimmunoprecipitation of HIF2a from parental 786-0 cells in normoxia or exposed to 4 hours of hypoxia, then probed for HAF (to detect endogenous HAF). I, Westernblotting showing levels of HIF1a and HIF2a in normoxia and hypoxia� proteasome inhibitor MG132 (5 mmol/L) in PANC-1 cells stably overexpressing FLAG–HAF orFLAG–HAF DM. J, luciferase activity in FLAG–HAF or FLAG–HAF DM overexpressing in PANC-1 cells transiently transfected with P117 HRE and normalized toRenilla luciferase. K and L, quantitation of ChIP assays for CR4 (HAF site; K) and CR3 (HRE site; L) on the OCT-3/4 promoter. Antibodies used for ChIP are indicatedbelow axes. � , P < 0.05. M, transcriptional activity of the pG5 luciferase reporter when cotransfected with the GAL4 DBD alone, GAL4–DBD–HAF fusion, orGAL4–DBD–HAFDMmutant. Results given as relative light units (RLU) offirefly luciferase normalized toRenillaencoded in theGAL4DBD (pBIND) vector. � ,P<0.05.N, model for the regulation of HAF-mediated HIF2 activation. HAF, by binding to its consensus site, regulates baseline transcription of target genes. SUMOylationof HAF (which can be induced by hypoxia) promotes the binding of HAF to HIF2a, hence driving maximal activation of HIF2 target genes.






HAF but not of cells overexpressing HAF DM, confirming thatK94R and K141R are indeed sites of HAF SUMOylation(Fig. 4A, i). To confirm that the SUMOylation observed wasspecific for HAF, we immunoprecipitated FLAG–HAF, thenboiled the samples in 1% SDS to dissociate all associatedproteins, and performed a second immunoprecipitation usinganti-FLAG. Indeed we found that the SUMOylated bands wereretained even after denaturation, confirming that it was indeedHAF that was SUMOylated (Fig. 4A, ii). Two bands wereobserved above the band for unmodified FLAG–HAF (120 kD)with molecular weight shifts consistent with conjugation to oneor two SUMO-1 molecules (each SUMO �10 kD). To deter-mine whether HAF SUMOylation is dependent on oxygentension, we immunoprecipitated FLAG–HAF (wt and DM)from 786-0 CRCC cells grown in normoxia (20% O2) orexposed to 3% O2 or 1% O2 for 4 hours. Oxygen tensions of3% to 6% O2 and 1% to 3% (30–50 mmHg; 8–15 mmHg) arephysiologic in the kidney cortex and medulla, respectively,whereas tensions of �1% O2 are typical of solid tumors(45). Intriguingly, we observed that exposure to 1%O2 resultedin a marked increase in SUMOylation (by SUMO-1) comparedwith exposure to 3% or 20% O2 (Fig. 4B). The increase in HAFSUMOylation was sustained, although slightly decreased after24 hours at 1% O2, whereas levels of HAF and HIF2a wereunchanged (Fig. 4C and D). In contrast, modification bySUMO-2/3 was weak and barely detectable (SupplementaryFig. S3A), and further studies hence focused exclusively onSUMO-1 modification. To determine whether SUMOylationof endogenous HAF was also hypoxia dependent, we immu-noprecipitated endogenous HAF from cells grown in normoxia,or from cells exposed to 4 hours of hypoxia (1% O2). Indeed,we found that SUMOylation of endogenous HAF was alsoincreased by hypoxia, thus supporting our findings with over-expressed HAF (Fig. 4E). To assess the impact of HAF SUMOy-lation on its regulation of HIF2 activity, wild-type HAF or HAFDM was transiently transfected into 786-0 cells, together withthe P117 HRE reporter. Indeed, we found that HAF DM, unlikeHAF wt, was unable to induce the activity of cotransfected P117HRE (Fig. 4F). Using qRT-PCR, we confirmed that HAF DMoverexpression did not promote the transcription of the HIF2target genes OCT-3/4 and SOX2, supporting our observationswith P117 HRE (Fig. 4G). Because we have previously shownthat HAF-mediated HIF2 transactivation requires HAF bindingto HIF2a (4), we examined the ability of HAF DM to bindHIF2a. Here, we found a clear reduction in the amount ofFLAG–HAF DM that could be immunoprecipitated with HIF2acompared with that with FLAG-tagged wild-type HAF (Fig. 4H,i, top). In addition, we observed a marked increase in thecoimmunoprecipitation of FLAG–HAF (but not of HAF DM)with HIF2a after exposure of 786-0 cells to 4 hours of hypoxia,although equal amounts of HIF2a were immunoprecipitated,suggesting that the binding of HAF to HIF2a is enhanced byhypoxia-dependent SUMOylation (Fig. 4H, i, bottom). We alsoobserved increased coimmunoprecipitation of endogenousHAF with HIF2a in hypoxia, thus supporting our findings withoverexpressed HAF (Fig. 4H, ii). To investigate the impact ofHAF SUMOylation in the context of functional pVHL in cellsexpressing both HIF1a and HIF2a, we also overexpressed HAFand HAF DM in PANC-1 cells. Similar to VHL-deficient cells, weobserved that HAF SUMOylation in pVHL-competent PANC-1cells was also increased in hypoxia (Supplementary Fig. S3B).

In addition, similar to wild-type HAF, HAF DM mediated theproteasomal-dependent degradation of HIF1a in both normoxiaand hypoxia, while not affecting levels of HIF2a (Fig. 4I). Inaddition, HAF DM expression did not induce the activity ofcotransfected P117 HRE, and showed reduced binding to HIF2a,thus confirming our findings in the 786-0 cells (Fig. 4J andSupplementary Fig. S3C). Furthermore, using ChIP assays, wefound that although HAF DM retained the ability to bind theHAF site within the OCT-3/4 promoter in a similar manner towild-type HAF (Fig. 4K), HAF DM neither bound the OCT-3/4HRE, nor promoted the binding of HIF2a to theOCT-3/4HRE, asobservedwithwild-typeHAF (Fig. 4L).However,when tethered toDNA using the GAL4/UAS system, we found that GAL4–HAF DMmutant demonstrated comparable transcriptional activity withGAL4–HAF, suggesting that HAFmay also possess SUMOylation-independent transcriptional activity (Fig. 4M).

Hence, we show that HAF SUMOylation is required for thebinding of HAF to HIF2a, and the formation of a HAF/HIF2atranscriptional complex that drives maximal induction of HIF2downstream target genes (Fig. 4N).

HAF promotes tumor metastasis in CRCC cells in vivoBecause HIF2a has been shown to play a unique role in

driving CRCC progression, we investigated the role of HAF ontumor growth using the 786-0 CRCC cells. Using Z-stackingmultispectral confocal microscopy, we confirmed that endoge-nous HAF (red) and HIF2a (green) were both localized withinclose proximity in the nucleus of 786-0 cells (Fig. 5A). Confir-mation of the specificities of the antibodies used is shown inSupplementary Fig. S4. When injected subcutaneously into theflanks of nude mice, we found that xenografts from cells over-expressing wild-type HAF showed a trend of increased tumorgrowth (P < 0.1), whereas cells overexpressing HAF DM grew ata slower rate, similar to the vector control cells, consistent withour findings that SUMOylation of HAF is required for HIF2transactivation (Fig. 5B). The growth-promoting effect mediatedby HAF overexpression was only evident in vivo, as we didnot observe any effects of HAF overexpression on cell prolifer-ation in vitro (Supplementary Fig. S5A). Because metastasis isfrequently observed in advanced CRCC, we also investigatedthe impact of HAF overexpression in a 786-0 CRCC mouserenal orthotopic tumor model (46). Here, we found that HAFoverexpression increased the incidence and number of meta-static lesions in the gastrointestinal (GI) tract of the mice (5 of 6in HAF-injected mice) versus 1 of 6 in vector-injected mice (Fig.5C and D), although there were no significant differencesbetween the sizes of the HAF versus vector-derived primarytumors. The lesions were confirmed by the presence of GFP,encoded by the pMX vector backbone, which was not detectedin the GI tract of non-tumor-bearing mice, or in mice bearingvector-derived tumors (Supplementary Fig. S5B). The spleen ofmice bearing HAF-overexpressing orthotopic tumors were alsoenlarged although metastatic lesions were not grossly apparent(Fig. 5E). Hence, HAF overexpression in vivo promotes thegrowth of subcutaneous tumor xenografts, and in a kidneyorthotopic model, results in increased metastasis.

HAF is associated with decreased PFS in CRCCTo investigate the clinical application of our findings, we

examined HAF and HIF2a distribution in patient CRCC tissue.When we stained serial sections of tumor tissue obtained from

Koh et al.





human stage III CRCC, we observed that regions of the tumorthat expressed high levels of HIF2a also expressed high levelsof HAF (Fig. 6A). This suggests that the HAF/HIF2a complexmay also be present in vivo to promote HIF2-dependent tran-scription. To investigate the relationship between HAF andCRCC patient prognosis, we stained a TMA from patients withstage III/IV advanced metastatic CRCC (33). Staining intensi-ties for HAF, HIF1a, and HIF2a together with other clinicalparameters are shown in Supplementary Fig. S6A. Represen-tative sections of high and low HIF1/2a-expressing cores withnuclear and cytoplasmic HAF expression are depicted in Fig.6B. Overall, we found that HIF1a staining was almost exclu-sively nuclear, whereas HIF2a staining was both nuclearand cytoplasmic. HAF staining was also both nuclear andcytoplasmic, and there was a significant correlation betweencytoplasmic and nuclear HAF expression (Supplementary Fig.S6B). When we divided the patients into two groups according

to the levels of nuclear HAF, we observed a significant corre-lation between patients with high HAF and decreased PFS, asurrogate endpoint for survival (P ¼ 0.045; Fig. 6C). Thehazard ratio for high HAF and time to progression was2.263 [95% confidence interval (CI), 1.001–5.116]. The medi-an PFS for patients with high versus low HAF was 5.94 � 0.62versus 8.97 � 3.1 months and approached significance (P ¼0.08). The only other univariate parameter that also showedsignificant association with PFS was Motzer criteria (47), witha hazard ratio of 2.34 (95% CI, 0.9985–5.492) for interme-diate versus good criteria, and time to progression (Supple-mentary Fig. S7). Other parameters such as age, sex, andnumber of disease sites were not significantly associated withPFS (not shown). In addition, we did not detect any significantcorrelations between levels of HIF1a or HIF2a and PFS (Sup-plementary Fig. S8). Hence, nuclear HAF may be a novelpredictor of poor patient response in high-grade CRCC.

Cell line Primary tumor weight (g)

Vec 0.164 ± 0.056

HAF 0.159 ± 0.088

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0.2

Vec HAF

Sp

leen

wei

gh

t (g

)

P = 0.06

Vec

HAF

HAF1 cm1 cm

Vec

1 cm

GI lesions (HAF)

Hoechst (Nucleus)HAF (Cy3) HIF2α (Alexa 488) Merge (Z-stack)

0

100

200

300

400

500

600

700

800

900

15 35 55 75 95

Mea

n t

um

or

bu

rden

(m

m3 )

Days since cell injection

Vec

HAF

DM

0

20

40

60

80

100

120

Vec HAF

% In

cid

ence

Mets (n)

3–6

≤2

0

B C

A

D E

Figure 5.A, confocal immunocytochemistryshowing localization of endogenousHAF andHIF2a to the nucleusof 786-0CRCC cells (magnification, �40).Merge image shows a representativeslice of the Z-stack, whereas green andpurple boxes (on the x and y axes)indicate axial views of the 0.8- to1.0-mmslice. B, in vivo tumor growth of786-0 HAF and 786-0 HAF DM cellscompared with vector control whengrown as subcutaneous xenograftsin nude mice (10 per group). C,comparison of primary tumor weightsobtained with orthotopic implantationof HAF or vector 786-0 cells into therenal subcapsule of nude mice (�SE).Representative photographs ofinjected and control kidneys areshown. D, percentage incidence andnumber of metastases observed in GItracts of mice injected with HAF orvector 786-0 cells. Representativephotograph of metastases is shown inthe inset. E, average spleen weightsand representative photographs ofspleens in the mice injected with theHAF- or vector-overexpressing cells.






Discussion

We have previously shown that HAF promotes the switch fromHIF1- to HIF2-dependent transcription in variety of cancer celllines (4). HAF is an E3 ubiquitin ligase that promotes theubiquitination and degradation of HIF1a (6). HAF also bindsto HIF2a but does not promote degradation, but rather increasesHIF2 transactivation. Here, we show that HAF-mediated HIF2transactivation requires both the binding of HAF to its DNAconsensus site, and the hypoxia-dependent SUMOylation ofHAF,which promotes its interaction with HIF2a. Hence, SUMOylatedHAF contributes to the maximal induction of HIF2 target genes,

and this is particularly clear in the pluripotency genes OCT-3/4,SOX2, and NANOG. The fact that this subset of HIF2-dependentgenes retains hypoxia inducibility despite loss of VHL (Fig. 1D)supports the existence of an alternative oxygen-sensing mecha-nism, such as HAF SUMOylation, which may play a role in theprogression of VHL-deficient CRCC.

Wehave observed that theHAFmutant that cannot be SUMOy-lated appears to be more efficient at degrading HIF1a than wild-type HAF (Fig. 4I). This suggests that HAF SUMOylation is adeterminant of its HIFa isoform selectivity. In this regard, HAFbinds andubiquitinatesHIF1a through theHAFE3 ligase domainlocated at the HAF C-terminus, and this occurs independently of

HIF2α

HAF

C

P = 0.045

PFS (months)0 5 10 15 20 25

Su

rviv

al p

rob

abili

ty (

K-M

)

Low HAF High HAF

A

#9

#1

HIF1α HIF2α HAF HAF (enlarged)

B

NuclearHAF

HiLow

0.0

0.2

0.4

0.6

0.8

1.0

Figure 6.A, localization of HAF and HIF2a insequential sections of human grade 3CRCC (magnification, �10). B,representative photomicrographsshowing expression of HIF1a, HIF2a,and HAF in CRCC tissue cores fromtwo different patients. Enlargedsections show differences in nuclearlocalization of HAF. C, the Kaplan–Meier plot representing theprobability of PFS in patients withCRCC, stratified according to high orlow nuclear HAF expression levels.Representative images of low andhigh HAF-expressing samples areshown in the inset.

Koh et al.





HAF SUMOylation. In contrast, HAF SUMOylation, which occursat theHAFN-terminal region, close to theHIF2a-binding domain(HAF 300–500), facilitates the binding of HAF to HIF2a, thuspromoting HIF2-dependent transcription. In this regard,decreased HAF SUMOylation may reduce binding to HIF2a, thusincreasing the amount of available HAF to bind and degradeHIF1a, resulting in the increased HIF1a degradation. Alternative-ly, it is also possible that HAF SUMOylationmay interfere with itsE3 ubiquitin ligase activity. Whichever the case may be, modu-lators of HAF SUMOylation may dictate whether it acts primarilyas an activator of HIF2a, or as an inhibitor of HIF1a.

On the basis of our findings, we propose that under ambientconditions, when HIF1/2a are largely absent, HAF drives base-line transcription by binding the HAF site within the promotersof target genes. As the HAF site within the OCT-3/4 promoterresides within the distal enhancer, which is required forOCT-3/4 transcription in embryonic stem cells, HAF may alsobe an important regulator of OCT-3/4 transcription in thesecells (39). During VHL loss (CRCC tumor initiation), bothHIF1a and HIF2a are stabilized and activate downstream tran-scription. Under these conditions, HAF functions as an E3ligase for HIF1a, hence attenuating the increase in HIF1acaused by pVHL loss. As CRCC tumors develop and begin toexperience regions of hypoxia, HAF becomes SUMOylated andis able to bind and transactivate HIF2, driving maximal induc-tion of HIF2 target genes (particularly of the HIF2-dependentpluripotency genes OCT-3/4, SOX2, and NANOG), while con-tinuing to degrade HIF1a. Thus, the combination of the non-hypoxic constitutive activation of the HIFs (pseudohypoxia)manifest during pVHL loss, and the presence of SUMOylatedHAF in hypoxic CRCC tumors, promote preferential activationof HIF2a, thus mediating the shift to HIF2-dependent tran-scription frequently observed in pVHL-deficient CRCC (Fig. 7).In contrast, differences in hypoxia-dependent spatiotemporalstabilization of HIF1a versus HIF2a may limit the extent ofHAF-mediated HIF2 specific activation due to fluctuations inHIF2a availability in other solid tumor types not associatedwith VHL-loss (2).

Hypoxia-dependent SUMOylation has increasingly beenshown to regulate the hypoxia response. PIASy, a SUMOE3 ligase

upregulated in hypoxia, SUMOylates pVHL, resulting in pVHLoligomerization, abolishing its ability to degrade HIF1a (48).Hypoxia-dependent SUMOylation of HIF1/2a has also beenreported, and this promotes HIF1/2a degradation by pVHL,although this is typically reversed by the SUMO-specific protease1 (SENP1; refs. 49, 50). Conversely, SUMOylation of HIF1a canalso promote HIF1a stabilization and transcriptional activity(51). Hence, SUMOylation and deSUMOylation have clear con-sequences on the HIF pathway, and their physiologic (and path-ophysiologic) significance remain to be elucidated. We have notyet identified the SUMO E3 ligase responsible for the hypoxia-dependent SUMOylationofHAF.However, to our knowledge, thehypoxia-dependent SUMOylation of HAF is the first report ofSUMO-mediated HIFa isoform-specific regulation.

In human patients, increased HAF levels correlate to decreasedPFS in patients with metastatic grade 3/4 CRCC. Indeed, the onlyother univariate parameter that also showed significant associa-tion with PFS was Motzer criteria, a classification system foradvanced RCC that has shown excellent predictive value, andthat has been used to select and stratify patients for treatment(47, 52). We did not detect any significant correlations betweenlevels of HIF1a or HIF2a and PFS (Supplementary Fig. S8). Thismay be due to the advanced stage of the tumors used in our study,compared with previously reported relationships between theHIFs and patient survival, which were identified in samplescomprising multiple tumor stages (15, 26, 53).

In summary, we have identified a novel mechanism determin-ing the HIFa isoform specificity of HAF. Hypoxia-dependentSUMOylation of HAF enables its binding to HIF2a to promotethe transcription of HIF2 target genes without affecting HAFsability to bind and degrade HIF1a. In the context of the pVHLdeficiency frequently observed in CRCC, this may result in pre-ferential HIF2-specific activation that drives tumor progression,resulting in poor patient prognosis. Hence, the targeting of theHAF–HIF2 axis offers a promising therapeutic strategy for thetreatment of CRCC.

Disclosure of Potential Conflicts of InterestG. Powis has ownership interest (including patents) in Oncothyreon. No

potential conflicts of interest were disclosed by the other authors.

Ambient (normal kidney) VHL loss (tumor initiation) VHL loss + hypoxia (advanced tumors)

HAF

HIF1αDegradationHAF

SS

HAF siteHRE

SHIF2α

HIF2 targets

S

Maximal induction

O2

HAF site

Baseline transcription HIF1α

HRE

HIF2αHRE

3%–6% O2 ≤1% O2

HAF

HIF1αDegradationHAF

Basal activity HIF1 and HIF2 active Selective HIF2 activation Poor prognosis

Figure 7.Proposed mechanism for HAF mediated HIFa isoform-specific regulation. Under ambient conditions, HAF is deSUMOylated and drives baseline transcription oftarget genes, whereas both HIF1a and HIF2a are absent. During tumor initiation, VHL loss results in the constitutive activation of both HIF1a and HIF2a, andHAF functions as an E3 ligase for HIF1a. Oxygen becomes limiting as tumors become established, resulting in HAF SUMOylation. This promotes HAF bindingto HIF2a, drivingmaximal activation of HIF2 target genes. At the same time, HAF still binds and degrades HIF1a independently of SUMOylation, effectively inhibitingHIF1a, and promoting activation of HIF2 target genes in CRCC.






Authors' ContributionsConception and design: M.Y. Koh, G. PowisDevelopment of methodology: M.Y. Koh, B.G. Darnay, G. PowisAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): V. Nguyen, R. Lemos Jr, T.H. Ho, J. Karam, E. JonaschAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): M.Y. Koh, M. Abdelmelek, T.H. Ho, J. Karam, F.A.Monzon, G. PowisWriting, review, and/or revision of the manuscript: M.Y. Koh, T.H. Ho,J. Karam, F.A. Monzon, E. Jonasch, G. PowisAdministrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): B.G. Darnay, G. Kiriakova, G. PowisStudy supervision: G. Powis

Grant SupportThis work was supported by the NIH grant CA181106 (M.Y. Koh), ASCO

Young Investigator Award-Kidney Cancer Association (T.H. Ho), and NIHgrant CA098920 (G. Powis), and the Cancer Center Support grant NIHCA030199.

The costs of publication of this article were defrayed in part by thepayment of page charges. This article must therefore be hereby markedadvertisement in accordance with 18 U.S.C. Section 1734 solely to indicatethis fact.

Received July 31, 2013; revised October 15, 2014; accepted November 5,2014; published OnlineFirst November 24, 2014.

References1. Wilson WR, Hay MP. Targeting hypoxia in cancer therapy. Nat Rev Cancer

2011;11:393–410.2. Koh MY, Powis G. Passing the baton: the HIF switch. Trends Biochem Sci

2012;37:364–72.3. Holmquist-Mengelbier L, Fredlund E, Lofstedt T, Noguera R, Navarro S,

Nilsson H, et al. Recruitment of HIF-1alpha and HIF-2alpha to commontarget genes is differentially regulated in neuroblastoma: HIF-2alphapromotes an aggressive phenotype. Cancer Cell 2006;10:413–23.

4. KohMY, Lemos R Jr, Liu X, Powis G. The hypoxia-associated factor switchescells from HIF-1{alpha}- to HIF-2{alpha}-dependent signaling promot-ing stem cell characteristics, aggressive tumor growth and invasion. CancerRes 2011;71:4015–27.

5. Keith B, Johnson RS, Simon MC. HIF1alpha and HIF2alpha: siblingrivalry in hypoxic tumour growth and progression. Nat Rev Cancer2012;12:9–22.

6. Koh MY, Darnay BG, Powis G. Hypoxia-associated factor, a novel E3-ubiquitin ligase, binds and ubiquitinates hypoxia-inducible factor 1alpha,leading to its oxygen-independent degradation. Mol Cell Biol 2008;28:7081–95.

7. Gnarra JR, Tory K, Weng Y, Schmidt L, Wei MH, Li H, et al. Mutations ofthe VHL tumour suppressor gene in renal carcinoma. Nat Genet 1994;7:85–90.

8. Moore LE, Nickerson ML, Brennan P, Toro JR, Jaeger E, Rinsky J, et al. VonHippel-Lindau (VHL) inactivation in sporadic clear cell renal cancer:associations with germline VHL polymorphisms and etiologic risk factors.PLoS Genet 2011;7:e1002312.

9. Banks RE, Tirukonda P, Taylor C, Hornigold N, Astuti D, Cohen D, et al.Genetic and epigenetic analysis of von Hippel-Lindau (VHL) gene altera-tions and relationship with clinical variables in sporadic renal cancer.Cancer Res 2006;66:2000–11.

10. Wang GL, Jiang BH, Rue EA, Semenza GL. Hypoxia-inducible factor 1 is abasic-helix-loop-helix-PAS heterodimer regulated by cellular O2 tension.Proc Natl Acad Sci U S A 1995;92:5510–4.

11. Baldewijns MM, van Vlodrop IJ, Vermeulen PB, Soetekouw PM, vanEngeland M, de Bruine AP. VHL and HIF signalling in renal cell carcino-genesis. J Pathol 2010;221:125–38.

12. Welford SM, Dorie MJ, Li X, Haase VH, Giaccia AJ. Renal oxygenationsuppresses VHL loss-induced senescence that is caused by increased sen-sitivity to oxidative stress. Mol Cell Biol 2010;30:4595–603.

13. Bratslavsky G, Sudarshan S, Neckers L, Linehan WM. Pseudohypoxicpathways in renal cell Carcinoma. Clin Cancer Res 2007;13:4667–71.

14. Chen Y, Zhang J, Dai J, Feng X, Lu H, Zhou C. Angiogenesis of renal cellcarcinoma: perfusion CT findings. Abdom Imaging 2010;35:622–8.

15. Mandriota SJ, Turner KJ, Davies DR, Murray PG, Morgan NV, Sowter HM,et al. HIF activation identifies early lesions in VHL kidneys: evidence forsite-specific tumor suppressor function in the nephron. Cancer Cell 2002;1:459–68.

16. Schietke RE, Hackenbeck T, Tran M, Gunther R, Klanke B, Warnecke CL,et al. Renal tubular HIF-2alpha expression requires VHL inactivation andcauses fibrosis and cysts. PLoS ONE 2012;7:e31034.

17. Turner KJ,Moore JW, Jones A, Taylor CF, Cuthbert-HeavensD,HanC, et al.Expression of hypoxia-inducible factors in human renal cancer: relation-ship to angiogenesis and to the von Hippel-Lindau gene mutation. CancerRes 2002;62:2957–61.

18. Gordan JD, Lal P, Dondeti VR, Letrero R, Parekh KN, Oquendo CE,et al. HIF-alpha effects on c-Myc distinguish two subtypes of sporadicVHL-deficient clear cell renal carcinoma. Cancer Cell 2008;14:435–46.

19. Kondo K, KimWY, LechpammerM, KaelinWG Jr. Inhibition of HIF2alphais sufficient to suppress pVHL-defective tumor growth. PLoS Biol 2003;1:E83.

20. Maranchie JK, Vasselli JR, Riss J, Bonifacino JS, LinehanWM, Klausner RD.The contribution of VHL substrate binding andHIF1-a to the phenotype ofVHL loss in renal cell carcinoma. Cancer Cell 2002;1:247–55.

21. Raval RR, Lau KW, Tran MG, Sowter HM, Mandriota SJ, Li JL, et al.Contrasting properties of hypoxia-inducible factor 1 (HIF-1) and HIF-2in von Hippel-Lindau-associated renal cell carcinoma. Mol Cell Biol2005;25:5675–86.

22. Zimmer M, Doucette D, Siddiqui N, Iliopoulos O. Inhibition of hypoxia-inducible factor is sufficient for growth suppression of VHL�/� tumors1 1NIH grant R29CA78358-06 (O. I.), Bertucci Fund for Urologic Malignan-cies (O. I.),DavidP. Foss Fund (O. I.), andVHLFamily Alliance 2003 award(M. Z.). Mol Cancer Res 2004;2:89–95.

23. Li L, Zhang L, Zhang X, YanQ,Minamishima YA,Olumi AF, et al. Hypoxia-inducible factor linked to differential kidney cancer risk seen with type 2Aand type 2B VHL mutations. Mol Cell Biol 2007;27:5381–92.

24. Hacker KE, Lee CM, Rathmell WK. VHL type 2B mutations retain VBCcomplex form and function. PLoS ONE 2008;3:e3801.

25. Shen C, Beroukhim R, Schumacher S, Zhou J, Chang M, Signoretti S, et al.

Genetic and functional studies implicate HIF1I� as a 14q kidney cancersuppressor gene. Cancer Discov 2011;1:222–35.

26. Monzon FA, Alvarez K, Peterson L, Truong L, Amato RJ, Hernandez-McClain J, et al. Chromosome 14q loss defines a molecular subtype ofclear-cell renal cell carcinoma associated with poor prognosis. Mod Pathol2011;24:1470–9.

27. Lamothe B, Besse A, Campos AD, Webster WK, Wu H, Darnay BG. Site-specific Lys-63–linked tumor necrosis factor receptor-associated Factor 6auto-ubiquitination is a critical determinant of IkappaB kinase activation. JBiol Chem 2007;282:4102–12.

28. Blanchard KL, Acquaviva AM, Galson DL, Bunn HF. Hypoxic induction ofthe human erythropoietin gene: cooperation between the promoter andenhancer, each of which contains steroid receptor response elements. MolCell Biol 1992;12:5373–85.

29. Kimura H, Weisz A, Ogura T, Hitomi Y, Kurashima Y, Hashimoto K, et al.Identification of hypoxia-inducible factor 1 ancillary sequence and itsfunction in vascular endothelial growth factor gene induction by hypoxiaand nitric oxide. J Biol Chem 2001;276:2292–8.

30. Polek TC, Talpaz M, Darnay BG, Spivak-Kroizman T. TWEAK mediatessignal transduction and differentiation of RAW264.7 cells in the absence ofFn14/TweakR. Evidence for a second TWEAK receptor. J Biol Chem 2003;278:32317–23.

31. Paine-Murrieta GD, Taylor CW, Curtis RA, Lopez MH, Dorr RT, JohnsonCS, et al. Human tumor models in the severe combined immune deficient(scid) mouse. Cancer Chemother Pharmacol 1997;40:209–14.

32. Zhang L, Zhang X, Barrisford GW, Olumi AF. Lexatumumab (TRAIL-receptor 2 mAb) induces expression of DR5 and promotes apoptosisin primary and metastatic renal cell carcinoma in a mouse orthotopicmodel. Cancer Lett 2007;251:146–57.

Koh et al.





33. Jonasch E, Corn P, Pagliaro LC,WarnekeCL, JohnsonMM, Tamboli P, et al.Upfront, randomized, phase 2 trial of sorafenib versus sorafenib and low-dose interferon alfa in patients with advanced renal cell carcinoma: clinicaland biomarker analysis. Cancer 2010;116:57–65.

34. Shen C, Beroukhim R, Schumacher SE, Zhou J, ChangM, Signoretti S, et al.Genetic and functional studies implicateHIF1alpha as a 14q kidney cancersuppressor gene. Cancer Discov 2011;1:222–35.

35. Gupta M, Mungai PT, Goldwasser E. A new transacting factor that mod-ulates hypoxia-induced expression of the erythropoietin gene. Blood2000;96:491–7.

36. Nordhoff V, Hubner K, Bauer A, Orlova I, Malapetsa A, Scholer HR.Comparative analysis of human, bovine, and murine Oct-4 upstreampromoter sequences. Mamm Genome 2001;12:309–17.

37. Covello KL, Kehler J, Yu H, Gordan JD, Arsham AM, Hu CJ, et al. HIF-2alpha regulatesOct-4: effects of hypoxia on stem cell function, embryonicdevelopment, and tumor growth. Genes Dev 2006;20:557–70.

38. Warnecke C, Zaborowska Z, Kurreck J, Erdmann VA, Frei U, Wiesener M,et al. Differentiating the functional role of hypoxia-inducible factor (HIF)-1alpha and HIF-2alpha (EPAS-1) by the use of RNA interference: eryth-ropoietin is a HIF-2alpha target gene in Hep3B and Kelly cells. FASEB J2004;18:1462–4.

39. Yeom YI, Fuhrmann G, Ovitt CE, Brehm A, Ohbo K, Gross M, et al.Germline regulatory element of Oct-4 specific for the totipotent cycle ofembryonal cells. Development 1996;122:881–94.

40. Blanchard KL, Acquaviva AM, Galson DL, Bunn HF. Hypoxic induction ofthe human erythropoietin gene: cooperation between the promoter andenhancer, each of which contains steroid receptor response elements. MolCell Biol 1992;12:5373–85.

41. Sadowski I, Ma J, Triezenberg S, Ptashne M. GAL4-VP16 is an unusuallypotent transcriptional activator. Nature 1988;335:563–4.

42. Schimmel J, Balog CI, Deelder AM, Drijfhout JW, Hensbergen PJ, VertegaalAC. Positively charged amino acids flanking a sumoylation consensus

tetramer on the 110kDa tri-snRNP component SART1 enhance sumoyla-tion efficiency. J Proteomics 2010;73:1523–34.

43. Vertegaal AC, Ogg SC, Jaffray E, Rodriguez MS, Hay RT, Andersen JS,et al. A proteomic study of SUMO-2 target proteins. J Biol Chem 2004;279:33791–8.

44. Geiss-Friedlander R, Melchior F. Concepts in sumoylation: a decade on.Nat Rev Mol Cell Biol 2007;8:947–56.

45. Nangaku M, Eckardt KU. Hypoxia and the HIF system in kidney disease. JMol Med 2007;85:1325–30.

46. Naito S, von Eschenbach AC, Fidler IJ. Different growth pattern andbiologic behavior of human renal cell carcinoma implanted into differentorgans of nude mice. J Natl Cancer Inst 1987;78:377–85.

47. Motzer RJ, Bacik J, Schwartz LH, Reuter V, Russo P, Marion S, et al.Prognostic factors for survival in previously treated patients withmetastaticrenal cell carcinoma. J Clin Oncol 2004;22:454–63.

48. Cai Q, Verma SC, Kumar P, Ma M, Robertson ES. Hypoxia inactivates theVHL tumor suppressor through PIASy-mediated SUMO modification.PLoS ONE 2010;5:e9720.

49. Cheng J, Kang X, Zhang S, Yeh ET. SUMO-specific protease 1 is essential forstabilization of HIF1alpha during hypoxia. Cell 2007;131:584–95.

50. van Hagen M, Overmeer RM, Abolvardi SS, Vertegaal AC. RNF4 and VHLregulate the proteasomal degradation of SUMO-conjugated Hypoxia-Inducible Factor-2alpha. Nucleic Acids Res 2010;38:1922–31.

51. Carbia-Nagashima A, Gerez J, Perez-Castro C, Paez-Pereda M, Silber-stein S, Stalla GK, et al. RSUME, a small RWD-containing protein,enhances SUMO conjugation and stabilizes HIF-1a during hypoxia.Cell 2007;131:309–23.

52. Dorff TB, Goldkorn A, Quinn DI. Review: targeted therapy in renal cancer.Ther Adv Med Oncol 2009;1:183–205.

53. Sandlund J, Ljungberg B, Wikstrom P, Grankvist K, Lindh G, Rasmuson T.Hypoxia-inducible factor-2alpha mRNA expression in human renal cellcarcinoma. Acta Oncol 2009;48:909–14.






2015;75:316-329. Published OnlineFirst November 24, 2014.Cancer Res Mei Yee Koh, Vuvi Nguyen, Robert Lemos, Jr, et al. Specific Activation of HIF2 in Clear-Cell Renal Cell CarcinomaHypoxia-Induced SUMOylation of E3 Ligase HAF Determines

Updated version

10.1158/0008-5472.CAN-13-2190doi:

Access the most recent version of this article at:

Material

Supplementary

http://cancerres.aacrjournals.org/content/suppl/2014/11/22/0008-5472.CAN-13-2190.DC1

Access the most recent supplemental material at:

Cited articles

http://cancerres.aacrjournals.org/content/75/2/316.full#ref-list-1

This article cites 53 articles, 22 of which you can access for free at:

Citing articles

http://cancerres.aacrjournals.org/content/75/2/316.full#related-urls

This article has been cited by 3 HighWire-hosted articles. Access the articles at:

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

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://cancerres.aacrjournals.org/content/75/2/316To request permission to re-use all or part of this article, use this link



http://cancerres.aacrjournals.org/lookup/doi/10.1158/0008-5472.CAN-13-2190

http://cancerres.aacrjournals.org/content/suppl/2014/11/22/0008-5472.CAN-13-2190.DC1

http://cancerres.aacrjournals.org/content/75/2/316.full#ref-list-1

http://cancerres.aacrjournals.org/content/75/2/316.full#related-urls

http://cancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://cancerres.aacrjournals.org/content/75/2/316


Documents

Hypoxia-Induced SUMOylation of E3 Ligase HAF Determines ......SMARTpool (Dharmacon L-017283, siHAF_1), and Silencer #15787 (siHAF_2; Ambion, Life Technologies). Hypoxic expo-sure was