ARTICLES

The E3 ligase HACE1 is a critical chromosome 6q21 tumor suppressor involved in multiple cancers

Liyong Zhang1,2,5, Michael S Anglesio3,5, Maureen O’Sullivan3, Fan Zhang3, Ge Yang2, Renu Sarao1,2, Mai P Nghiem1,2, Shane Cronin1, Hiromitsu Hara1, Nataliya Melnyk3, Liheng Li3, Teiji Wada1, Peter P Liu2, Jason Farrar4, Robert J Arceci4, Poul H Sorensen3 & Josef M Penninger1

Transformation and cancer growth are regulated by the coordinate actions of oncogenes and tumor suppressors. Here, we show that the novel E3 ubiquitin ligase HACE1 is frequently downregulated in human tumors and maps to a region of chromosome 6q21 implicated in multiple human cancers. Genetic inactivation of HACE1 in mice results in the development of spontaneous, late-onset cancer. A second hit from either environmental triggers or genetic heterozygosity of another tumor

http://www.nature.com/naturemedicine suppressor, p53, markedly increased tumor incidence in a Hace1-deficient background. Re-expression of HACE1 in human tumor cells directly abrogates in vitro and in vivo tumor growth, whereas downregulation of HACE1 via siRNA allows non-tumorigenic human cells to form tumors in vivo. Mechanistically, the tumor-suppressor function of HACE1 is dependent on its E3 ligase activity and HACE1 controls adhesion-dependent growth and cell cycle progression during cell stress through degradation of cyclin D1. Thus, HACE1 is a candidate chromosome 6q21 tumor-suppressor gene involved in multiple cancers.

Human cancer mapping studies for cataloguing regions of genomic RESULTS loss of heterozygosity (LOH) have defined a large number of LOH HACE1 is epigenetically inactivated in human Wilms’ tumor regions that are common to multiple human cancers1. Deletions or We originally identified the HACE1 gene by cloning the chromosome LOH on human chromosome 6q21 have been widely reported in 6q21 breakpoint of a t(6;15)(q21;q21) translocation in a human human malignancies, including carcinomas of the breast, ovary and Wilms’ tumor12. This rearrangement was associated with marked Nature Publishing Group Group 200 7 Nature Publishing 2–6

© prostate, as well as in leukemias and lymphomas , and it has been downregulation of HACE1 mRNA and protein expression compared hypothesized that this region harbors one or more tumor-suppressor to normal kidney from the same person, even though the transloca- genes7. The key tumor-suppressor gene(s) on chromosome 6q21 have tion did not directly disrupt the HACE1 locus. Fluorescence in situ not been identified to date. hybridization (FISH) analysis confirmed that HACE1 localizes to a We previously identified a t(6;15)(q21;q21) translocation in a region of human chromosome 6q21 that is disrupted in this individual sporadic Wilms’ tumor8. Wilms’ tumors account for over 90% of (Fig. 1a). To extend these findings, we analyzed HACE1 expression in pediatric kidney tumors and 6% of all childhood cancers9. Involve- 26 individuals with Wilms’ tumor. HACE1 levels were markedly ment of the 6q21 region has previously been reported in Wilms’ reduced to virtually undetectable in 20 (77%) of the 26 Wilms’ tumors tumor, including t(5;6)(q21;q21) and t(2;6)(q35;q21) rearrange- compared to patient-matched normal kidney, at both the mRNA ments9–11. We therefore characterized the chromosome 6q21 break- (Supplementary Fig. 1a online) and the protein levels (Fig. 1b). point and found that it mapped directly adjacent to a 6q21 gene12.The Sequencing of all 24 exons of the HACE1 gene plus intron-exon encoded protein has a previously unreported domain architecture, boundaries in tumor and matching normal tissue did not detect consisting of six ankyrin protein-protein interaction motifs with mutations or deletions affecting HACE1 in any of the cases of Wilms’ sequence similarity to those of INK4A, linked to a C-terminal tumor (data not shown). We therefore assessed whether epigenetic homologous to E6-associating protein carboxyl terminus (HECT) inactivation of HACE1 might underlie the observed downregulation. ubiquitin-protein ligase domain13. We designated this gene HECT There are three CpG islands (the most common sites of DNA domain and ankyrin repeat–containing E3 ubiquitin-protein ligase 1, hypermethylation in tumors14) associated with the HACE1 locus: or HACE1.Thein vivo function of this novel HECT E3 ubiquitin CpG-88, containing the transcriptional start site, and CpG-29 and ligase was entirely unknown. CpG-177, located directly upstream of the HACE1 coding sequence12.

1Institute of Molecular Biotechnology of the Austrian Academy of Sciences, Dr. Bohrgasse 3, 1030 Vienna, Austria. 2Heart & Stroke/Richard Lewar Centre of Excellence, University of Toronto and Toronto General Research Institute, University Health Network, Toronto, Ontario M5G 2C4, Canada. 3British Columbia Cancer Research Centre, Vancouver, British Columbia V5Z 1L3, Canada and the Department of Pathology and Laboratory Medicine, University of British Columbia, Vancouver, British Columbia V5Z1L3, Canada. 4Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins University, Baltimore, Maryland 21231, USA. 5These authors contributed equally to this work. Correspondence should be addressed to P.H.S. ([email protected]) or J.M.P. ([email protected]). Received 8 September 2006; accepted 27 June 2007; published online 12 August 2007; doi:10.1038/nm1621

1060 VOLUME 13 [ NUMBER 9 [ SEPTEMBER 2007 NATURE MEDICINE ARTICLES

To evaluate the association of DNA methylation with HACE1 expres- DNA methylation within these three defined CpG islands (Supple- sion in Wilms’ tumor, we used bisulfite genomic sequencing (BGS) mentary Fig. 1b online). We found no evidence for cytosine methyla- and methylation-sensitive PCR (MS-PCR) to assess the pattern of tion at CpG-88. However, both CpG-29 and CpG-177 contained DNA a b 785 015 555 728 586 520 432 N T NTNTNTNTNT NT

HACE1

β-Actin

628 321 395 624 033 277 654 N T NTNTNTNTNT NT

HACE1

β-Actin

c CpG-177 Normal samples Tumor samples Methylation

CpG position

http://www.nature.com/naturemedicine Unmethylated Methylated Not present

Methylation

CpG position Unmethylated Methylated Not present

Methylation

CpG position Unmethylated Methylated Not present

CpG-29 Normal samples Tumor samples Nature Publishing Group Group 200 7 Nature Publishing d

© Methylation

CpG position Unmethylated Methylated Not present

Methylation

CpG position Unmethylated Methylated Not present

Methylation

CpG position Unmethylated Methylated Not present

Unmethylated Methylated Not present In Clone

Figure 1 Loss of HACE1 expression by epigenetic modification in human Wilms’ tumor. (a) FISH analysis. A chromosome 6q21 BAC probe containing the human HACE1 locus, 111D22 (green), was hybridized to metaphase chromosomes from a Wilms’ tumor with a t(6;15)(q21;q21) translocation. Arrows indicate genomic rearrangement of this locus. Chromosomes were stained with 4¢,6-diamidino-2-phenylindole (DAPI). Original magnification 100. (b) HACE1 protein expression is markedly reduced to undetectable in 13 of 14 primary Wilms’ tumors (T) compared to patient-matched adjacent normal kidney (N). Protein expression was determined by western blot. b-actin is shown as a loading control. Numbers refer to individual patients. (c,d) Increased methylation at CpG-177 (c) and relative hypomethylation at CpG-29 (d) in Wilms’ tumor samples compared to paired normal kidney tissue controls. Each box corresponds to a CpG position in the genomic sequence. Colored bars represent the aggregated methylation status as assessed by bisulfite genomic sequencing (BGS) at each position: yellow, methylated; blue, unmethylated; gray, not present or not evaluable in the sequence read. Numbers below each bar indicate the total number of methylated, unmethylated and non-evaluated CpGs at each position. Aggregated normal tissues are shown at left and tumor samples at right.

NATURE MEDICINE VOLUME 13 [ NUMBER 9 [ SEPTEMBER 2007 1061 ARTICLES

Figure 2 Reduced HACE1 expression in multiple NT N T N T N T Melanoma Hepatocellular carcinoma human tumors and spontaneous tumor formation a c in Hace1 mutant mice. (a) HACE1 mRNA expres- sion levels in paired human tumor (T) and patient- matched normal tissue (N) arrays probed with a

full-length HACE1 cDNA. (b)Chromosome6 Vulva ideogram and expanded 6q21 region illustrating previously described LOH regions in multiple human tumors. Vertical bars indicate tumor subtypes with LOH between the Breast Renal Thyroid Liver Lung adenocarcinoma Angiosarcoma polymorphic markers D6S1845 and D6S1424.

I, ovarian carcinoma and benign ovarian 25 32 33, b 24 I tumors , non-Hodgkin’s lymphoma (NHL) 23 II 34 22 childhood acute lymphoblastic lymphoma , III gastric carcinoma35, T and NK cell lymphomas36, p 21 IV invasive ductal breast carcinoma37–41; V 42 12 II, pancreatic carcinoma ; III, prostate q16.3 D6S1845 carcinoma2,43; IV, central nervous system (CNS) 12 D6S1021 lymphoma44; V, sporadic endocrine pancreatic 13 HACE 1 14 D6S268 45 15 q21 Mammary carcinoma Lymphoma tumors .(c) Photomicrographs of H&E-stained HACE 1 sections showing tumors that developed 16 D6S261 q21.1 spontaneously in aged (41 year) Hace1–/– mice. 21 Melanoma: the black arrow indicates intact q 22 q22.31 epidermis and the white arrow an ulceration 23 D6S1702 overlying the dermal tumor (original magnification 24 100). Arrowhead in inset points to melanin q22.33 25 http://www.nature.com/naturemedicine pigment production by pleomorphic cells 26 27 q23.1 infiltrating the dermis (400). Hepatocellular D6S1424 Chr 6 carcinoma: effacement of normal porto-lobular architecture (40). Inset (400) shows d 100 *** P < 0.0001 carcinoma cells forming pseudoglandular 90 *** P < 0.0001 P < 0.01 structures (white arrow). Lung adenocarcinoma: 80 ** arrows show interface between adenocarcinoma 70 (right) and lung parenchyma (left; 40). Inset 60 Hace1+/– Tp53 +/+ shows haphazard arrangement of lung 50 *** Hace1–/– Tp53 +/+ ****** adenocarcinoma cells with mitotic activity 40 Hace1–/– Tp53 +/– *** (arrows) ( 400). Angiosarcoma: formation of 30 Hace1+/– Tp53 –/– **

tumor free Percent –/– –/– blood vessels (black arrow). Inset shows 20 Hace1 Tp53 ** endothelial cell pleomorphism. Mammary 10 carcinoma: arrow shows border between 0

Nature Publishing Group Group 200 7 Nature Publishing carcinoma (bottom) and normal breast 0255075 100 Age (weeks) © tissue (top). Inset shows focal keratinization. Lymphoma: arrow shows starry sky appearance of lymphoma cells (right) expanding into normal pancreatic parenchyma (left; 40). In the inset (400), a mitotic figure is circled. (d) Spontaneous tumor incidence in mice carrying the indicated genotypes is shown by Kaplan-Meier plots as the tumor-free proportion of each cohort of mice. Tumor incidencein mice over the indicated follow-up period is shown as the ratio of the number of mice with tumors / total mice analyzed for each genotype. This represents 68% (22/69) for Hace1–/–Tp53+/– mice, 79% (11/14) for Hace1+/–Tp53–/– mice, 100% (24/24) for Hace1–/–Tp53–/– mice, 12% (29/252) for Hace1–/–Tp53+/+ mice and 1.3% (4/295) for Hace1+/–Tp53+/+ mice.

methylation and were further evaluated by detailed cloning and Fig. 1a) and western blotting (Fig. 1b). In contrast, for the six Wilms’ sequencing of the initial 30-40 CpG sites. We observed increased tumor cases in which HACE1 was not reduced compared to matching methylation at CpG-177 in the tumor samples (Fig. 1c and Supple- normal kidney, there was either no difference in CpG-177 methylation mentary Fig. 2 online) when compared to normal controls. In in tumor versus normal tissue (n ¼ 3) or no methylation in either contrast, we found relative hypomethylation at CpG-29 (Fig. 1 and tumor or normal tissue (n ¼ 3). This association of decreased HACE1 Supplementary Fig. 3 online) in Wilms’ tumor samples. expression with hypermethylation of CpG-177 in Wilms’ tumor but Because CpG island hypermethylation, rather than hypomethyla- not normal kidney was statistically significant (Supplementary Fig. 4c tion, is more broadly implicated in epigenetic gene silencing14,wenext online). Consistent with these observations, we previously found that used MS-PCR to screen Wilms’ tumor samples for methylation at the DNA methylation inhibitor 5-aza-2-deoxycytidine induces HACE1 CpG-177 (Supplementary Fig. 4a online). In agreement with results re-expression in a human tumor cell line (SK-NEP1) with HACE1 from the BGS analysis (Fig. 1c), we found that 73% of tumor samples inactivation12. Our data indicate that expression of the HACE1 gene is (19/26) but only 35% of normal samples (9/26) were methylated at epigenetically downregulated in primary human Wilms’ tumor. CpG-177 (Supplementary Fig. 4b online). Moreover, in more than half of the CpG-177–methylated tumors (11/19), there was complete HACE1 is downregulated in multiple human cancers absence of corresponding methylation in paired normal tissue as HACE1 mRNA transcripts are widely expressed in normal human detected by MS-PCR. All 11 of these individuals showed markedly tissues (Supplementary Fig. 5a online). We therefore assessed mRNA reduced HACE1 expression in the tumor sample compared to expression in human tumors and organ-matched normal tissue paired normal kidney by real-time quantitative PCR (Supplementary samples from the same individuals. HACE1 expression was decreased

1062 VOLUME 13 [ NUMBER 9 [ SEPTEMBER 2007 NATURE MEDICINE ARTICLES

in B50% of all primary human tumors studied (Fig. 2a). HACE1 Table 1 Spontaneous tumor development in Hace1-deficient mice transcripts were markedly downregulated in almost all lines from the NCI-60 panel of human cancer cell lines compared to control a Hace1–/– (29 mice with tumors from total 252 mice analyzed, tumor incidence HEK293 cells (Supplementary Fig. 5b online). HACE1 transcripts are 12%) also markedly reduced in human tumors propagated in vivo as xenografts in athymic nude mice (Supplementary Fig 5a). Mapping Mouse Sex Age (weeks) Type of tumor studies confirmed that the HACE1 locus is located within previously B196 M 36 Mammary carcinoma described regions of deletions or LOH in multiple human tumors B390 M 70 Lymphoma including T- and NK-cell lymphomas, B-cell non-Hodgkin’s lympho- B395 M 56 Melanoma mas, central nervous system lymphomas, childhood acute lympho- B491 M 96 Hepatocellular carcinoma blastic leukemia, prostate cancers, pancreatic tumors, gastric F6 F 105 Lymphoma carcinomas, ovarian carcinomas, cervical adenocarcinomas and F26 F 105 Melanoma aggressive breast cancer (Fig. 2b). Thus, HACE1 is downregulated in F156 F 90 Hepatocellular carcinoma multiple human cancers and maps to a prominent tumor-suppressor F191 F 90 Adenocarcinoma of lung, region of human chromosome 6q21. angiosarcoma, splenic lymphoma F209 M 90 Lymphoma Generation of Hace1 mutant mice H6 M 78 Lung adenocarcinoma To study the in vivo function of HACE1, we generated Hace1 mutant H11 F 90 Splenic lymphoma mice. The mouse and human HACE1 proteins share 96% identity. H20 M 82 Hepatocellular carcinoma Expression of mouse Hace1 mRNA transcripts was detected in H41 M 90 Hepatocellular carcinoma essentially all adult tissues and throughout embryonic development H49 F 81 Lymphoma H120 F 80 Lymphoma (Supplementary Fig. 6a online). We disrupted the Hace1 gene in H128 F 90 Angiosarcoma http://www.nature.com/naturemedicine mouse embryonic stem (ES) cells using a targeting vector in H143 F 95 Histiocytic sarcoma which nucleotides encompassing exons 3 and 4, which encode the H57 M 80 Lung carcinoma first two ankyrin repeats, were deleted (Supplementary Fig. 6b H60 M 80 Sarcoma, thymic lymphoma +/– online). The intercross of Hace1 mice produced homozygous H76 M 90 Hepatocellular carcinoma –/– Hace1 mice, as confirmed by Southern blot analysis using a 5¢ H190 F 20 Fibrohistiocytic sarcoma flanking probe (Supplementary Fig. 6c online). Hace1 mRNA tran- H200 M 26 Myxosarcoma –/– –/– scripts were undetectable in kidneys of Hace1 mice and Hace1 B527 F 66 Soft tissue tumor embryonic fibroblasts by RT-PCR and northern blot analyses using a F69 M 102 Soft tissue tumor full-length probe (Supplementary Fig. 6d online and data not F126 M 96 Soft tissue tumor shown). Moreover, Hace1–/– mice do not express any detectable G60 M 95 Soft tissue tumor HACE1 protein (Supplementary Fig. 6e online). Notably, all of Thymic lymphoma the Hace1 mice used in this study were from a mixed 129/Ola and G76 M 100 Soft tissue tumor C57BL/6 (F2) background and only littermate mice were included in H2 M 90 Lung tumor Nature Publishing Group Group 200 7 Nature Publishing the cohorts. H5 M 80 Lung tumor ©

Development of spontaneous tumors in Hace1–/– mice b Hace1+/– (4 mice with tumors from total 295 mice analyzed, tumor incidence Hace1–/– mice were born at the expected mendelian frequency and 1.3%) were fertile. Hace1–/– mice were indistinguishable from wild-type controls and showed apparently normal morphologies of all organs Mouse Sex Age (weeks) Type of tumor –/– analyzed. As Hace1 mice aged, these mutant mice spontaneously B84 F 90 Angiosarcoma developed a remarkable spectrum of tumors such as melanoma, G36 F 90 Hepatocellular carcinoma, lymphoma hepatocellular carcinoma, spontaneous lung adenocarcinoma, angio- G71 M 90 Lymphoma sarcoma, mammary carcinomas and lymphomas (Fig. 2c). The B58 F 78 Soft tissue tumor first spontaneous tumor in Hace1–/– mice was observed at 20 weeks B All tumors listed in plain face were detected macroscopically and analyzed by of age and the incidence of spontaneous tumors reached 12% in microscopy; tumors listed in boldface were analyzed either by weight and flow cytometry 2-year-old mice (Fig. 2d and Table 1). The incidence of spontaneous for detection of thymic lymphomas or by gross examination (for macroscopically visible tumors in the Hace1+/– and Hace1+/+ littermate cohorts was B1–2% tumors) in the remaining cases. Age indicates age of mice when tumors where detected. at a comparable age. Thus, genetic inactivation of Hace1 in mice results in the formation of a wide array of spontaneous tumors in g-irradiation resulted in B30% incidence of thymic lymphomas diverse tissues. (Supplementary Fig. 7a–c online). Flow cytometric analysis of such thymic lymphomas revealed a CD4negCD8negTCRablow phenotype Hace1 loss renders mice susceptible to other cancer triggers (Supplementary Fig. 7d online). The incidence of cancer has been observed to increase markedly in To extend these findings to a second model of environmentally tumor-suppressor mutant mice upon exposure to a second genetic or triggered cancer and a different class of cancer-promoting agents, we environmental insult15,16. We speculated that inactivation of Hace1 treated mice with urethane, a DNA alkylating agent that causes lung might also render mice susceptible to a second cancer trigger. To test carcinomas after repeated administration15,17. Consistent with pre- this hypothesis, we challenged Hace1–/– and control littermates with vious reports18, a single challenge of urethane did not trigger cancer in g-irradiation at a dose that results in a very low incidence of thymic wild-type littermates (Fig. 3a,b). Only a low percentage of Hace1+/– lymphomas on a wild-type background. In Hace1–/– mice, low-dose mice developed lung tumors (Fig. 3b). Notably, almost all Hace1–/–

NATURE MEDICINE VOLUME 13 [ NUMBER 9 [ SEPTEMBER 2007 1063 ARTICLES

a bc

Hace1+/+ 100 90 80 70 µ 60 0.5 mm 50 m 50 40 30 20 10

Percentage of mice with lung tumor Percentage 0 –/– +/– +/+ Hace1 Hace1 Hace1 0.5 mm 50 µm Hace1–/–

Figure 3 Hace1 mutant mice are susceptible to urethane-induced lung cancer. (a)Lungtumors in Hace1–/– mice. Three typical cases of lung tumors in Hace1–/– mice are shown 6 months after a single injection of urethane (1,000 mg/kg of body weight; i.p.) at 4 weeks of age. All Hace1+/+ littermates treated with the same procedure remained tumor free. (b) Lung cancer 0.5 mm 50 µm incidence. Data are summarized from 30 Hace1–/–,10Hace1+/+ and 20 Hace1+/– mice at 6 months after the single urethane injection. P o 0.001 between groups, w2 test. (c) Histology of lung tumors in Hace1–/–mice. Top, normal lung from a urethane-treated Hace1+/+ mouse. Middle and bottom, lung adenocarcinomas in Hace1–/– mice. All pictures are from H&E-stained sections. Scale bars, 0.5 mm (left panels) and 50 mm (right panels). http://www.nature.com/naturemedicine mice developed large and rapidly growing lung tumors in response in the lung over the time period of assessment (Fig. 4h). In addition to to urethane (Fig. 3a,b). Within the same mice, multifocal lung the high prevalence of sarcomas and carcinomas in Hace1–/–Tp53+/– tumor development was observed, suggesting multiple transformation and Hace1–/–Tp53–/– mice, high-grade lymphomas were also observed events. Tumors were histologically assessed as early as 3 months (Supplementary Fig. 8 online). In one tumor that developed in after urethane treatment and classified as adenocarcinomas Hace1–/–Tp53+/– mice, we observed loss of p53 expression (Fig. 4i). (Fig. 3c). During the observation period, we did not observe Additional samples need to be analyzed to allow for solid conclusions metastatic spread of these primary lung tumors. Our results show on p53 LOH in such tumors. These data show that loss of HACE1 that loss of HACE1 renders mice susceptible to different environ- renders mice susceptible to environmental and genetic second hits for mental triggers of cancer. the development of multiple cancers and that HACE1 cooperates with p53 in cancer formation. Cooperativity between p53 and HACE1 in tumor suppression To assess whether loss of HACE1 shows cooperativity in tumorigenesis HACE1 controls tumorigenicity of human cancer cells Nature Publishing Group Group 200 7 Nature Publishing with other tumor suppressors, we crossed the Hace1 mutation onto Because cancer is a multi-step process and HACE1 inactivation © Tp53 or Cdkn1a mutant backgrounds. Tp53 encodes the tumor cooperates with other cancer-promoting agents, we tested whether suppressor p53. The Cdkn1a gene encodes the tumor suppressor HACE1 per se has tumor-suppressor activity in transformed cells. We p21/CIP1. We first generated Hace1/Cdkn1a double-mutant mice. first examined the growth properties of NIH3T3 mouse fibroblasts Compared with Hace1 or Cdkn1a single-knockout mice, no significant stably expressing hemagglutinin (HA)-tagged HACE1 alone or in statistical differences in tumor development were detected among combination with an activated form of Ki-Ras (encoded by KiRas, those mice (data not shown). We next assessed potential cooperativity also known as Kras). Although HACE1 overexpression in either non- between HACE1 and p53. Heterozygosity at the Tp53 locus did not transformed or KiRas-transformed NIH3T3 cells had no effect on cell + + affect tumor development on a Hace1 heterozygous background, but morphology (data not shown), Ki-Ras HACE1 cells reached a –/– +/+ older Hace1 Tp53 mice showed an incidence of cancer similar to growth plateau earlier than Ki-Ras cells transduced with empty vector that described above (Fig. 2d and Fig. 4a). As expected19,20,lossof (MSCV) (Fig. 5a). Cell-cycle analysis showed that at confluence, a + + both Tp53 alleles resulted in high incidence of tumor formation. greater percentage of the Ki-Ras HACE1 cells accumulate in G0/G1 Notably, loss of a single Tp53 allele on a Hace1–/– background greatly compared with control MSCV cells (Fig. 5b). Moreover, the propor- + + increased tumor incidence (Fig. 2d, Fig. 4a and Supplementary tion of mitotic Ki-Ras HACE1 cells was decreased in comparison to Table 1 online). In some instances, these tumors grew to enormous their Ki-Ras counterparts (Fig. 5b). sizes within 2 weeks of first detection (Fig. 4b). Inactivation of both We next tested whether HACE1 can mediate growth suppression in p53 and Hace1 alleles (Hace1–/–Tp53–/–) also markedly increased human kidney tumor or neuroblastoma cell lines that express low tumor incidence in younger mice (Fig. 2d and Fig. 4a). Whereas levels of HACE1. To test whether the growth suppressive effects of Tp53–/– mice primarily developed thymic lymphomas19,20,lossof HACE1 are dependent on its E3 ligase activity, we used a ligase-dead HACE1 markedly expanded the profile of cancer types (Supplemen- HACE1 mutant with a cysteine-876-to-serine (C876S) substitution12. tary Table 1) to include osteosarcoma (Fig. 4c), angiosarcoma When SK-NEP-1 human kidney tumor cells were stably transduced (Fig. 4d), leiomyosarcoma (Fig. 4e), mammary carcinoma (Fig. 4f), with HA-tagged HACE1, growth rates were markedly suppressed keratinizing squamous cell carcinoma (Fig. 4g)andothers.The compared to those of cells transduced with vector alone cells histogenesis of tumors was confirmed by electron microscopy and (Fig. 5c and Supplementary Fig. 9a online). Moreover, whereas immunohistochemistry (data not shown). Notably, two of the sarco- SK-NEP-1 cells show anchorage-independent growth, expression of mas were associated with the development of micrometastatic disease wild-type HACE1 significantly reduced the ability of SK-NEP-1 cells

1064 VOLUME 13 [ NUMBER 9 [ SEPTEMBER 2007 NATURE MEDICINE ARTICLES

Figure 4 Cooperativity between p53 and HACE1 in tumor suppression. (a) Incidence of ab +/– +/– spontaneous, macroscopically visible tumors in 100 Hace1 tp53 Hace1–/– tp53 +/+ mice carrying the indicated Hace1 and Tp53 90 Hace1–/– tp53 +/– genotypes at 4 months and 10 months of age. 80 Hace1+/– tp53 –/– +/– +/– Numbers per group were Hace1 Tp53 Hace1–/– tp53 –/– 70 (n ¼ 85); Hace1–/–Tp53+/+ (n ¼ 29); Hace1+/–Tp53–/– (n ¼ 11); Hace1–/–Tp53+/– 60 (n ¼ 69); Hace1–/–Tp53–/– (n ¼ 24). (P o 0.05 50 between groups, w2 test.) (b)Tumorsina 40 i Tumor Normal 12-week-old Hace1–/–Tp53+/– mouse. Note that 30 in this mouse, there was only a 2-week period p53 20

between first gross detection of the tumor and of mice with tumor Percentage β-Actin the development of the tumor as shown (arrows). 10

– – (c–h) Photomicrographs of H&E sections from 0 +/– +/ +/– +/ +/+ Up to 4 months old Up to 10 months old representative tumors: osteosarcoma in a Ip53 – Ip53 Ip53 Ip53 – Ip53 Hace1–/–Tp53+/– mouse, with arrow showing Genotypes and ages of mice –/– +/ +/– –/– +/

‘lace-like’ osteoid formed by malignant, Hace1 Hace1 Hace1 Hace1 Hace1 pleomorphic cells (200; c); angiosarcoma in a Hace1–/–Tp53–/– mouse, with arrow showing c de highly atypical endothelial lining cell (400; d); leiomyosarcoma in a Hace1–/–Tp53+/– mouse, with arrows demonstrating the interdigitating fascicles of pleomorphic, mitotically active spindle cells (200; e); mammary carcinoma in a Hace1–/–Tp53+/– mouse, with black arrow http://www.nature.com/naturemedicine showing malignant glands with focal keratinization in left upper quadrant, adjacent to a normal duct (white arrow) (40; f); keratinizing squamous cell carcinoma arising fgh in skin of a Hace1–/–Tp53–/– mouse, with arrow showing focus of keratinization (100; g); metastatic osteosarcoma in lung of a Hace1–/–Tp53+/– mouse, with arrow showing a focus of osteosarcoma with calcification within lung parenchyma (40; h). (i)Lossof p53 expression in a tumor isolated from a Hace1–/–Tp53+/– mouse. p53 expression in a tumor from a Hace1+/–Tp53–/– mouse and in non-transformed kidney tissue from Hace1+/–Tp53+/–,Hace1–/–Tp53+/– and Hace1+/–Tp53+/+ mice is shown. p53 and control b-actin levels were determined by western blotting. Nature Publishing Group Group 200 7 Nature Publishing © to form colonies in soft agar (Fig. 5d). By contrast, transfection with To demonstrate the effects of downregulating endogenous HACE1, the E3 ligase–dead HACE1 C876S mutant had no effect on cell growth we stably expressed small hairpin RNAs (shRNAs) directed against (data not shown) and did not change soft agar colony formation of HACE1 mRNA in human HEK293 cells. HEK293 cells express HACE1, SK-NEP1 cells (Supplementary Fig. 9b online). In addition, transfec- show anchorage-dependent growth and form colonies only poorly in tion with wild-type HACE1 significantly reduced the ability of human soft agar. As shown for the representative HACE1-specific N1 shRNA, IMR32 neuroblastoma cells to form anchorage-independent colonies these molecules were effective in reducing HACE1 protein expression, compared to transfection with vector alone or with the HACE1 C876S whereas the scrambled S1 control shRNA was not (Fig. 6a). Reduced mutant (Supplementary Fig. 9c online). Thus, HACE1, through its E3 HACE1 expression resulted in increased soft agar colony formation of ligase activity, can suppress cell growth and anchorage independence HEK293 cells expressing N1 shRNA (Fig. 6a,b). Down-modulation of of human tumor cells. HACE1 expression in HEK293 cells was also achieved using three To assess the growth-suppressive effects of HACE1 in vivo, alternative siRNAs, and in all instances, decreased HACE1 expression SK-NEP-1 cells stably expressing wild-type HACE1 or the HACE1 levels correlated with the ability of HEK293 cells to form colonies in C876S mutant were compared to cells expressing vector alone for their soft agar (Supplementary Fig. 10a,b online). HEK293 cells expressing ability to form tumors after injection into nude mice. Consistent with N1 versus scrambled shRNA were then injected into nude mice. Where- previous reports21,22, SK-NEP-1 tumor cells stably expressing vector as cells expressing scrambled shRNA were non-tumorigenic, those alone formed visible tumors 15–20 d after injection, which continued expressing HACE1 shRNA showed a marked increase in tumorigenicity to grow over time (Fig. 5e,f). Tumorigenicity was profoundly inhib- in vivo (Fig. 6c).ThesedataindicatethatHACE1canmodulatesoft ited for SK-NEP-1 cells overexpressing wild-type HACE1 (Fig. 5e,f), agar colony formation and in vivo tumorigenicity of different cell types. although very small nodules were occasionally observed in mice injected with these cells. The HACE1 C876S mutant–expressing HACE1 modulates cell proliferation pathways via cyclin D1 SK-NEP1 cells showed tumor formation rates that were comparable To examine the potential effects of HACE1 on cell proliferation, we to control cells (Fig. 5e). A similar trend was observed for mice analyzed a panel of cell cycle–related proteins in HEK293T cells, injected with Ki-Ras–transformed NIH3T3 cells in the presence or which, in contrast to the parental HEK293 cells, express low HACE1 absence of HACE1 cotransfection, although the differences in tumori- levels. These cells were engineered to stably overexpress HACE1, genicity were less marked (data not shown). ligase-dead C876S HACE1 or vector alone (Supplementary

NATURE MEDICINE VOLUME 13 [ NUMBER 9 [ SEPTEMBER 2007 1065 ARTICLES

abcdMSCV 7 HA-HACE1 KiRas KiRas 6 1,200 KiRas-MSCV 1,500 MSCV Hace1 3 5 1,000 * 10 KiRas-Hace1 × 4 MSCV 2 mm 800 1,000 2.35 H]thymidine 70.5 3 1.96 3 10.1 82.8 * 600 12 9.09 4.75 incorporation 2 400 500 *

Relative [ Relative 1

200 of cells number Relative Number of cells 0 0 123456 0 200 400 0 200 400 1234 HA-HACE1 Days in culture Days in culture Fluorescence intensity

1,400 MSCV HA-Hace1 Figure 5 HACE1 controls growth and tumorigenesis of transformed e MSCV f 1,200 HA-HACE1 human cancer cells. (a,b) HACE1 expression attenuates K -Ras–induced )

i 3 HA-C876S proliferation and cell cycle progression in KiRas-transformed NIH3T3 1,000 cells. Growth curves (a) and cell cycle profiles (b; PI staining) of KiRas- 800 transformed NIH3T3 cells that co-express HACE1, compared with KiRas-transformed NIH3T3 cells co-transfected with an empty vector 600 (MSCV). Percentages of cells in sub-G1 (apoptotic cells), G1, G2/M 400 Tumor volume (mm volume Tumor and S phases are indicated. Data are representative of three separate 200 experiments. (c) Proliferation of human SK-NEP-1 kidney tumor cells 0 expressing HA-tagged wild-type human HACE1 (HA-HACE1) or empty 0 5 10 15 20 25 30 control vector (MSCV). Proliferation was determined in triplicate Days after injection (±s.d.) by H3-thymidine uptake and compared to proliferation of http://www.nature.com/naturemedicine non-transfected cells. *P o 0.02 between groups, Student’s t-test. (d) Soft agar colony formation of control SK-NEP-1 cells transfected with the empty control vector MSCV and SK-NEP-1 cells transfected with HA-HACE1. The images show typical colony sizes in control MSCV- and HA-HACE1–transfected SK-NEP1 tumor cells. Note that expression of HACE1 reduced not only colony numbers but also the sizes of individual colonies. Representative colonies on day 21 are shown. Data are representative of at least three separate experiments. (e) In vivo tumor formation. SK-NEP-1 cells engineered to stably overexpress HA-tagged HACE1, empty vector (MSCV) or the C876S HACE1 mutant (HA-C876S) were injected subcutaneously into nude mice and tumor volumes determined. Cells expressing mutant HA-C876S HACE1 or carrying empty vector reached maximal tumor burden in less than 30 days, whereas HA-HACE1–expressing cells did not achieve a maximal tumor burden (P o 0.05, Mann-Whitney rank test). (f) Typical tumor growth of HACE1-expressing and control vector–transfected SK-NEP-1 tumor cells injected into nude mice. Arrows indicate tumors.

Fig. 11a online). Cells were synchronized by adding thymidine and To further probe the effects of HACE1 on cyclin D1 expression, aphidicolin in fresh medium to block cells at the G1-S phase, released HEK293T cells treated with the translation inhibitor cycloheximide by medium change, and then assayed for expression of cell cycle pro- showed markedly reduced stability of cyclin D1 when HACE1 was teins. There was no difference in levels of cyclin A, B, E or G1 among overexpressed (Supplementary Fig. 12 online). Moreover, cyclin D1 Nature Publishing Group Group 200 7 Nature Publishing the different cell lines over the time period analyzed (Supplementary levels were stabilized in HACE1-overexpressing cells in the presence of © Fig. 11b). However, we observed a marked reduction in cyclin D1 in the MG132 proteasome inhibitor (Fig. 6f). Thus, loss of cyclin D1 in cells overexpressing HACE1 compared with C876S HACE1 or vector these cells occurs through degradation by the 26S proteasome and not alone cells (Supplementary Fig. 11b). Cyclins D2 and D3 were either through alterations in cyclin D1 translation. Cyclin D1 is known to be unaffected or not detectable in these cells (data not shown). After targeted for proteasomal degradation through glycogen synthase overnight serum starvation, there was again no difference among cell kinase-3b (GSK3b)–mediated phosphorylation of threonine-28624. lines in levels of cyclins A, B, E and G1 with 10% serum treatment over We therefore assessed the effects on HACE1 growth suppression of a an 8-h period (Fig. 6d). In contrast, starved HACE1-overexpressing cyclin D1 mutant with a threonine-286-to-alanine substitution, which cells, but not cells treated with vector alone, failed to upregulate cyclin stabilizes the cyclin D1 protein24.Notably,thiscyclinD1T286A D1 after re-stimulation with serum (Fig. 6d). Cells expressing the mutant was able to reverse HACE1-mediated reduction of HACE1 C876S mutant were identical to those expressing vector alone HEK293T soft agar colony formation, whereas overexpression of (data not shown). In keeping with previous data indicating that wild-type cyclin D1 in the same cells did not block HACE1 effects cyclin-dependent kinase 4 (CDK4) is stabilized through the formation (Fig. 7a). Moreover, HACE1 overexpression did not induce degrada- of CDK4/cyclin D1 complexes23, we also observed a reproducible tion of the cyclin D1 T286A mutant, whereas the wild-type cyclin D1 reduction in CDK4 levels in cells overexpressing HACE1 compared to protein was rapidly degraded over the same time period (Fig. 7b). vector alone or ligase-dead HACE1 after serum stimulation (data not These data suggest that HACE1 is involved in phosphorylation- shown). We next attempted to rescue cyclin D1 expression in HACE1- dependent degradation of cyclin D1. We have not found any evidence overexpressing HEK293T cells by reducing HACE1 levels with siRNA. that cyclin D1 is directly targeted by the HACE1 E3 ligase. Although Downregulation of HACE1 expression using two independent siRNAs, not precluding the involvement of other signaling pathways, our but not scrambled siRNA, increased cyclin D1 expression (Fig. 6e). findings point to a role for HACE1 in inhibiting cell cycle progression Moreover, we observed increased and more rapid cyclin D1 induction during cell stress through regulation of cyclin D1 degradation. in Hace1–/– mouse embryonic fibroblasts (MEFs) after mitogenic stimulation of serum-starved cells compared to wild-type MEFs DISCUSSION (Supplementary Figure 11c). Finally, treatment with g-irradiation Genomic losses involving human chromosome 6q21 have been failed to abolish cyclin D1 expression in Hace1–/– MEFs (Supplemen- described for a wide spectrum of tumor types including prostate, tary Figure 11d). breast or ovarian cancers, as well as leukemias and lymphomas,

1066 VOLUME 13 [ NUMBER 9 [ SEPTEMBER 2007 NATURE MEDICINE ARTICLES

HA-Hace1 ad80 bc1 mm MSCV 600

) HEK293-N1 3 HEK293-S1 Serum, h: Ctrl– 2 4 6 8 Ctrl– 2 4 6 8 60 HEK293-S1HEK293-N1 100 HA-Hace1 400 Cyclin A HACE1 54.4 40 Cyclin B HEK293-N1 54.4 1 mm Cyclin D1 β-Actin 200 38.7 * Cyclin E 20 54.4

Mean tumor volume (mm Mean tumor volume 38.7 Cyclin G1 0 25 30 35 40 45 Colonies formed (% of cells plated) Colonies formed 0 HEK293-S1 Days after injection

HEK293-S1HEK293-N1 efMSCV HA-Hace1 MG132, h: – 2 4 6 8 10 – 2 4 6 8 10 Figure 6 HACE1 controls anchorage-dependent growth and cyclin D1 HA-Hace1 siRNA-S1siRNA-N1siRNA-N2 100 expression. (a) Decreased HACE1 expression in HEK293 cells after treatment Cyclin A with HACE1-specific shRNA (N1) results in increased colony formation in soft HACE1 54.4 Cyclin B agar. Scrambled shRNA (S1) is shown as a control. The western blot inset Cyclin D1 54.4 Cyclin D1 indicates HACE1 and control -actin expression in HEK293 cells treated with 38.7 b Grb2 HACE1 siRNA (N1) and scrambled siRNA (S1). * P o 0.01 between groups, 54.4 Cyclin E Student’s t-test. (b) Typical soft agar colonies in HEK293-S1 and HEK293-N1 cells. Scale bars, 1 mm. (c) In vivo tumor formation of control HEK293-S1 and HEK293-N1 cells. Cells were injected into immunodeficient nude mice and tumor volumes determined on the indicated days after injection. (d) Expression of cell cycle molecules in HEK293T cells stably overexpressing HACE1 or transfected with vector alone (MSCV). Cells were stimulated for 0 (–), 2, 4, 6 or 8 h with 10% serum after overnight serum starvation. Ctrl, control cells under exponential growth conditions in 10% serum. Expression of the indicated cell cycle molecules was determined by western blotting. Molecular weight

http://www.nature.com/naturemedicine markers are shown at left. (e) Downregulation of HACE1 expression in HACE1-overexpressing HEK293T cells using two independent HACE1-specific siRNA oligos corresponding to N1 and N2 resulted in elevated cyclin D1 expression compared to the scrambled siRNA (S1) control. HACE1 levels were detected using anti-HACE1 antibodies. Grb2 protein levels are shown as controls. (f) Protein expression of the indicated cyclins in MSCV- and HA-HACE1–transfected HEK293T cells after treatment with 10 mM of the proteasome inhibitor MG132. Data are representative of at least three separate experiments. Molecular weight markers are shown on the left side.

suggesting that this region encompasses one or more major tumor- angiosarcoma, breast cancer, hepatocellular carcinoma, lymphoma suppressor genes2–6. For example, a recent study found that B50% of and lung carcinoma. Importantly, loss of HACE1 renders mice prostate cancers have deletions in the 6q21 region2.Ourresultsnow susceptible to environmental second hits for the development of show that expression of the 6q21 HACE1 gene is downregulated in multiple cancers. For instance, almost all Hace1 mutant mice devel- multiple human tumor types. Loss of HACE1 expression correlates oped large, multifocal and rapidly growing lung adenocarcinomas in with hypermethylation of a CpG island, CpG-177, located directly response to urethane challenge. Many of these tumor types have upstream of the HACE1 locus. This suggests that at least in some previously-documented genomic alterations of chromosome 6q21, Nature Publishing Group Group 200 7 Nature Publishing human malignancies, HACE1 may be epigenetically regulated. Silen- most of which have not been characterized at the genetic level. © cing of tumor-suppressor genes by DNA or histone methylation Moreover, our studies showed that HACE1 cooperates with p53 in appears to be at least as frequent as inactivating mutations in sporadic preventing cancer formation, and mutation of both genes dramatically tumors24, and aberrant gene silencing may be an early event in increased tumor incidence and significantly expanded the profile of neoplastic progression14. Further studies are necessary to determine cancer types from that normally observed in p53 single mutant mice. the role of epigenetic inactivation of HACE1 in other malignancies. At the molecular level, HACE1 appears to regulate cell cycle Gene inactivation in mice results in spontaneous, late onset cancers progression during cell stress by influencing the degradation of cyclin and renders mice exceedingly sensitive to second environmental D1. Importantly, expression of a cyclin D1 mutant (cyclin D1 T286A) and genetic cancer triggers. Intriguingly, Hace1 mutant mice exhibit that cannot be degraded via GSK3b-mediated phosphorylation25 a remarkable spectrum of tumor types, including osteosarcoma, abolished the growth-inhibitory effects of HACE1 overexpression.

Figure 7 HACE1 modulates cell proliferation ab pathways via cyclin D1 degradation. (a)Co- 80 * * MSCV Hace1 expression of the T286A cyclin D1 mutant blocks Serum –2 4 6 8 – 2 4 6 8 h HACE1-mediated reduction of HEK293T soft 60 Ctrl Ctrl agar colony formation. Soft agar colony formation HA-HACE1 HA–cyclin1 of HEK293T cells transfected with the empty 40 Cyclin D1 control vector MSCV, HA-HACE1 as well as co- Cyclin D1 β

transfected with wild-type cyclin D1 or cyclin D1- Colonies formed 20 -Actin T286A mutant. (*P o 0.01 between groups, HA-HACE1 Student’s t test). (b) Expression of cell cycle (percentage of cells plated) 0 molecules in HEK293T cells stably HA–cyclin1 Cyclin D1-T286A overexpressing HACE1 or transfected with vector MSCVHace1 Cyclin D1 MSCV-D1Hace1-D1 alone (MSCV) as well as cotransfected with wild- β-Actin type cyclin D1 or the T286A cyclin D1 mutant. MSCVHace1 D1-T286A D1-T286A Cells were stimulated for 0 (–), 2, 4, 6 or 8 h with 10% serum after overnight serum starvation. Ctrl, control cells under exponential growth conditions in 10% serum. Expression of the indicated cell cycle molecules was determined by western blotting. b-actin levels are shown as controls. Data are representative of at least three separate experiments.

NATURE MEDICINE VOLUME 13 [ NUMBER 9 [ SEPTEMBER 2007 1067 ARTICLES

The roles of D-type cyclins in oncogenesis are well established26,and Cruz), GRB2 (growth factor receptor–bound protein 2) (BD Biochemical), the overexpression of cyclin D1 has been described in a wide range of HA peptide tag YPYDVPDYA (Bio-Can Scientific), p21 (Upstate) and p53 human malignancies27. These findings suggest a possible mechanism CM5 (Novocastra, UK). To normalize signals, filters were re-probed with whereby HACE1 can regulate exit from the cell cycle in response to antibodies against b-actin or GAPDH. various forms of cell stress by reducing cyclin D1 protein levels in an Cell proliferation and modulation of HACE1 expression. To analyze DNA E3 ligase– and proteasome–dependent manner. In accordance with content, synchronized cells were stained with propidium iodide and analyzed this notion, recent studies have demonstrated that MEFs lacking by flow cytometry. To measure cyclin D1 stability, HEK293 cells were treated D-type cyclins show reduced susceptibility to oncogenic transforma- with cycloheximide and analyzed for cell cycle proteins by western blotting. To tion by various oncogenes, and that these cells have an attenuated block proteasomal degradation, HEK293 cells were treated with MG132. For ability to respond to extracellular mitogenic stimulation28. cyclin D1 overexpression studies, pcDNA3 plasmids containing HA-tagged The data presented here demonstrate that loss of HACE1 expression wild-type cyclin D1 or the cyclin D1 T286A mutant were cloned into MSCV 12,31 leads to the development of diverse cancers and renders mice vectors for retroviral infection of HEK293T cells . For HACE1 down- modulation, short hairpin constructs were transferred to the lentiviral packa- susceptible to environmental and second genetic cancer triggers ging construct pLenti6/Block-iT-DEST and then used for cell infection or in vivo. Our data therefore identify HACE1 as a candidate tumor- transient RNA interference using synthetic siRNA. Quantitative soft agar assays suppressor gene that maps to a region of chromosome 6q21 impli- were performed by plating SK-NEP-1 and IMR32 cell lines in 6-well plates. In cated in multiple human tumor types. addition, SK-NEP1 cells expressing HA-HACE1, C876S HACE1 or empty vector (MSCVhygro), or HEK293-N1 and HEK293-S1 cells, were injected METHODS subcutaneously in the flanks of 6–8 week-old male athymic nude mice. Fluorescence in situ hybridization (FISH). FISH was performed using standard protocols on metaphase nuclei. In vivo tumor analysis. Spontaneous primary tumor formation was followed in Hace1-deficient mice and wild-type littermates for a period of 2 years. For Detection of DNA methylation. Bisulfite modification was performed on induction of lung cancer, mice were injected intraperitoneally at 4 weeks of age genomic DNA from Wilms’ tumor samples and patient-matched normal with urethane (1,000 mg per kg body weight). For g-irradiation–induced http://www.nature.com/naturemedicine kidney using methylSEQr Bisulfite Conversion. For BGS, PCR products were tumors, 50 Hace1–/– and 50 Hace1+/+ littermates at 4 weeks of age were cloned, isolated using TOPO TA Cloning, sequenced and exported for final exposed to a single dose of 5-Gy whole-body g-irradiation from a cobalt 60 bisulfite sequence analysis using BiQ Analyzer29. MS-PCR products were source at a dose rate of 1.095 Gy/min. Mice were monitored once per week and electrophoretically separated on 2% agarose gels and visualized by ethidium at the first sign of morbidity animals were killed according to institutional bromide staining. guidelines. Autopsy was performed on all moribund mice or mice with visible tumors. At the time of sacrifice, necropsies were performed as follows: lungs, Generation of Hace1–/– mice. To disrupt the mouse Hace1 gene in embryonic thymus, liver, kidneys, spleen, heart, stomach, intestine and colon were stem cells, a portion of the Hace1 genomic DNA containing exons 3 and 4, examined macroscopically for the presence of tumors. Tumor types were which encode the first two ankyrin repeats, was replaced with a neomycin (neo) diagnosed using standard pathologic criteria by two of the authors (M.O. resistance cassette. The targeting construct was electroporated into embryonic and P.H.S.) who are board-certified pathologists. Lymphoma phenotypes as day 14 (E14) ES cells (129/Ola). Two targeted ES cell lines yielded germline well as thymocyte and peripheral T cell populations were determined by flow transmission. Data reported in this manuscript were consistent between the cytometry using antibodies directed against CD4, CD8 and CD3/TCR surface two mutant mouse lines. Tp53+/– and Cdkn1a+/– mice were intercrossed with antigens (BD Biosciences). Hace1–/– mice. All mice used in this study are on a mixed 129/Ola and C57BL/6 Nature Publishing Group Group 200 7 Nature Publishing background. Only littermate mice from the same breeding were used. All mice Statistical analysis. Statistical analysis was performed using either Student’s © 2 were maintained at the animal facilities of the Ontario Cancer Institute and the t-test or w tests. P o 0.05 was considered significant. Institute of Molecular Biology at the Austrian Academy of Sciences in Additional methods. Detailed methodology is described in Supplementary accordance with institutional guidelines. All experimental procedures per- Methods. formed on mice were in accordance with institutional guidelines. Note: Supplementary information is available on the Nature Medicine website. Cell culture and expression constructs. NIH3T3 cells were propagated in DMEM with 10% calf serum. Human SK-NEP-1 tumor cells, IMR32 human ACKNOWLEDGMENTS neuroblastoma cells, and HEK293, HEK293T and Phoenix A cells were We thank P. Grundy and the Children’s Oncology Group as well as the propagated as described12. MEFs were prepared from E13.5 mouse embryos Cooperative Human Tissue Network (CHTN) for providing clinical Wilms’ generated from Hace1 heterozygous crosses. A plasmid containing green tumor samples for this study; A. Brooks-Wilson and M. Marra for HACE1 30 sequencing assistance; R. Kandel and D. Holmyard for help with electron fluorescence protein (GFP)–tagged activated Ki-Ras was a gift (see Acknow- ledgments). Ectopic expression of wild-type or C876S mutant HACE1, and/or microscopy of the sarcoma samples; D. Bouchard and N.-J. Chen for assistance with FACS, M. Bowden and M. Gleave for assistance with nude mouse studies; GFP-tagged activated Ki-Ras in NIH3T3 and SK-NEP-1 cells, was established by 31 and V. Evdokimova, T. Zoranovic, K. Kuba, N. Joza, A. Oliveira-dos Santos, retroviral infection . M. Crackower, I. Kozieradski, T. Nakashima, H. Jones Taggart, M. Cheung, M. Rangachari, Y. Liu, M. Sun and M. Pollard for discussions, technical help and Protein and mRNA analyses. mRNA expression was detected by real-time reagents. The GFP-tagged activated Ki-Ras plasmid was provided by R. Kay, Terry quantitative PCR and northern blot analyses using standard procedures. For Fox Laboratory, British Columbia Cancer Research Centre. This work was analysis of HACE1 expression in human tumors, mRNA expression was supported by the National Cancer Institute of Canada (NCIC) (to P.H.S.), a quantified in the NCI-60 panel of human cancer cell lines. In addition, HACE1 Canadian Institutes of Health Research (CIHR) post-doctoral fellowship (to F.Z.), expression was assessed in human normal and tumor tissues using cDNA the Institute for Molecular Biotechnology of the Austrian Academy of Sciences templates from the Clontech Human Multiple Tissue cDNA (MTC) Panel I (IMBA), the Jubilaeumsfonds of the Austrian National Bank, GEN-AU and an EU (cat. no. K1420-1) and Panel II (cat. no. K1421-1), as well as the Human Tumor Excellence grant to J.M.P. Multiple Tissue cDNA (MTC) Panel (cat. no. K1422-1). For detection of AUTHOR CONTRIBUTIONS HACE1 protein, we raised polyclonal and monoclonal antibodies against a L.Z. made Hace1 mutant mice in the laboratory of J.M.P. and later in the mouse N-terminal HACE1 peptide and the full-length HACE1 protein. In laboratory of P.P.L. and analyzed the tumors in vivo with significant help from addition, we used antibodies reactive to cyclin A (Upstate), cyclin B (Upstate), G.Y., R.S., M.P.N., S.C., H.H., T.W. and P.P.L. M.S.A. performed most of the cyclin D1 (Cell Signaling), cyclin D1/D2 (Upstate), cyclin D3 (Santa Cruz), in vitro cell line experiments, with significant assistance from F.Z., N.M. and L.L. cyclin E (Upstate), cyclin G1 (Santa Cruz), Cdk4 (Neomarkers), Cdk6 (Santa M.O. and P.H.S. are trained pathologists and assessed the tumors histologically.

1068 VOLUME 13 [ NUMBER 9 [ SEPTEMBER 2007 NATURE MEDICINE ARTICLES

J.F. and R.J.A. did all methylation studies reported in the paper. P.H.S. and J.M.P. 20. Harvey, M. et al. Spontaneous and carcinogen-induced tumorigenesis in p53-deficient supervised and initiated the project from the outset and together wrote the paper. mice. Nat. Genet. 5, 225–229 (1993). 21. Fogh, J., Fogh, J.M. & Orfeo, T. One hundred and twenty-seven cultured human COMPETING INTERESTS STATEMENT tumor cell lines producing tumors in nude mice. J. Natl. Cancer Inst. 59, 221–226 (1977). The authors declare competing financial interests: details accompany the full-text 22. Lovvorn, H.N., III, Savani, R.C., Ruchelli, E., Cass, D.L. & Adzick, N.S. Serum HTML version of the paper at http://www.nature.com/naturemedicine/. hyaluronan and its association with unfavorable histology and aggressiveness of heterotransplanted Wilms’ tumor. J. Pediatr. Surg. 35, 1070–1078 (2000). Published online at http://www.nature.com/naturemedicine 23. Bagui, T.K., Mohapatra, S., Haura, E. & Pledger, W.J. P27Kip1 and p21Cip1 are not Reprints and permissions information is available online at http://npg.nature.com/ required for the formation of active D cyclin-cdk4 complexes. Mol. Cell. Biol. 23, reprintsandpermissions 7285–7290 (2003). 24. Diehl, J.A., Cheng, M., Roussel, M.F. & Sherr, C.J. Glycogen synthase kinase- 3beta regulates cyclin D1 proteolysis and subcellular localization. Genes Dev. 12, 1. Thiagalingam, S. et al. Loss of heterozygosity as a predictor to map tumor suppressor 3499–3511 (1998). genes in cancer: molecular basis of its occurrence. Curr. Opin. Oncol. 14,65–72 25. Malumbres, M. & Barbacid, M. To cycle or not to cycle: a critical decision in cancer. (2002). Nat. Rev. Cancer 1, 222–231 (2001). 2. Hyytinen, E.R. et al. Defining the region(s) of deletion at 6q16–q22 in human prostate 26. Jones, P.A. & Baylin, S.B. The fundamental role of epigenetic events in cancer. Nat. cancer. Genes Chromosom. Cancer 34, 306–312 (2002). Rev. Genet. 3, 415–428 (2002). 3. Inoue, M. et al. Chromosome 6 suffers frequent and multiple aberrations in thymoma. 27. Diehl, J.A. Cycling to cancer with cyclin D1. Cancer Biol. Ther. 1, 226–231 Am. J. Pathol. 161, 1507–1513 (2002). (2002). 4. Orphanos, V. et al. Allelic imbalance of chromosome 6q in ovarian tumors. Br. J. Cancer 28. Kozar, K. et al. Mouse development and cell proliferation in the absence of D-cyclins. 71, 666–669 (1995). Cell 118, 477–491 (2004). 5. Utada, Y. et al. Mapping of target regions of allelic loss in primary breast cancers to 29. Bock, C. et al. BiQ Analyzer: visualization and quality control for DNA methylation data 1-cM intervals on genomic contigs at 6q21 and 6q25.3. Jpn. J. Cancer Res. 91, from bisulfite sequencing. Bioinformatics 21, 4067–4068 (2005). 293–300 (2000). 30. Tognon, C.E. et al. Regulation of RasGRP via a phorbol ester-responsive C1 domain. 6. Zhang, Y. et al. A 3-cM commonly deleted region in 6q21 in leukemias and lymphomas Mol. Cell. Biol. 18, 6995–7008 (1998). delineated by fluorescence in situ hybridization. Genes Chromosom. Cancer 27,52–58 31. Tognon, C. et al. Expression of the ETV6-NTRK3 gene fusion as a primary event in (2000). human secretory breast carcinoma. Cancer Cell 2, 367–376 (2002). 7. Morelli, C. et al. Cloning and characterization of the common fragile site FRA6F 32. Thomas, N.A., Neville, P.J., Baxter, S.W. & Campbell, I.G. Genetic analysis of benign harboring a replicative senescence gene and frequently deleted in human tumors. ovarian tumors. Int. J. Cancer 105, 499–505 (2003). Oncogene 21, 7266–7276 (2002). 33. Gaidano, G. et al. Deletions involving two distinct regions of 6q in B-cell non-Hodgkin http://www.nature.com/naturemedicine 8. Fernandez, C.V., Lestou, V.S., Wildish, J., Lee, C.L. & Sorensen, P.H. Detection of a lymphoma. Blood 80, 1781–1787 (1992). novel t(6;15)(q21;q21) in a pediatric Wilms’ tumor. Cancer Genet. Cytogenet. 129, 34. Takeuchi, S. et al. Allelotype analysis in relapsed childhood acute lymphoblastic 165–167 (2001). leukemia. Oncogene 22, 6970–6976 (2003). 9. Solis, V., Pritchard, J. & Cowell, J.K. Cytogenetic changes in Wilms’ tumors. Cancer 35. Li, B.C. et al. Allelic loss of chromosome 6q in gastric carcinoma. Diagn. Mol. Pathol. Genet. Cytogenet. 34, 223–234 (1988). 12, 193–200 (2003). 10. Hoban, P.R. et al. Physical localisation of the breakpoints of a constitutional transloca- 36. Yoon, J. & Ko, Y.H. Deletion mapping of the long arm of chromosome 6 in peripheral T tion t(5;6)(q21;q21) in a child with bilateral Wilms’ tumor. J. Med. Genet. 34, and NK cell lymphomas. Leuk. Lymphoma 44, 2077–2082 (2003). 343–345 (1997). 37. Fujii, H., Zhou, W. & Gabrielson, E. Detection of frequent allelic loss of 6q23-q25.2 in 11. Bruce, C.K., Howard, P., Nowak, N.J. & Hoban, P.R. Molecular analysis of region microdissected human breast cancer tissues. Genes Chromosom. Cancer 16,35–39 t(5;6)(q21;q21) in Wilms tumor. Cancer Genet. Cytogenet. 141, 106–113 (1996). (2003). 38. Fujii, S., Takeshima, Y., Arihiro, K., Kaneko, M. & Inai, K. Microsatellite instability in 12. Anglesio, M.S. et al. Differential expression of a novel ankyrin containing E3 ubiquitin- breast cancers with special reference to patients’ age and bilaterality. Hiroshima protein ligase, Hace1, in sporadic Wilms’ tumor versus normal kidney. Hum. Mol. J. Med. Sci. 47, 89–97 (1998). Genet. 13, 2061–2074 (2004). 39. Fujii, H. et al. Mucinous cancers have fewer genomic alterations than more common 13. Huibregtse, J.M., Scheffner, M., Beaudenon, S. & Howley, P.M. A family of proteins classes of breast cancer. Breast Cancer Res. Treat. 76, 255–260 (2002). structurally and functionally related to the E6-AP ubiquitin-protein ligase. Proc. Natl. 40. Trink, B. et al. Absence of TSG101 transcript abnormalities in human cancers. Acad. Sci. USA 92, 2563–2567 (1995). Oncogene 16, 2815–2818 (1998). Nature Publishing Group Group 200 7 Nature Publishing 14. Jones, P.A. & Baylin, S.B. The epigenomics of cancer. Cell 128, 683–692 (2007). 41. Smeds, J. et al. Ductal carcinoma in situ of the breast with different histopathological © 15. Sharpless, N.E. et al. Loss of p16Ink4a with retention of p19Arf predisposes mice to grades and corresponding new breast tumor events: analysis of loss of heterozygosity. tumorigenesis. Nature 413, 86–91 (2001). Acta Oncol. 44, 41–49 (2005). 16. Sharpless, N.E. et al. p16INK4a and p53 deficiency cooperate in tumorigenesis. Cancer 42. Yatsuoka, T. et al. Association of poor prognosis with loss of 12q, 17p, and 18q, and Res. 62, 2761–2765 (2002). concordant loss of 6q/17p and 12q/18q in human pancreatic ductal adenocarcinoma. 17. Herzog, C.R., Noh, S., Lantry, L.E., Guan, K.L. & You, M. Cdkn2a encodes functional Am. J. Gastroenterol. 95, 2080–2085 (2000). variation of p16INK4a but not p14ARF, which confers selection in mouse lung tumori- 43. Konishi, N. et al. Genetic mapping of allelic loss on chromosome 6q within hetero- genesis. Mol. Carcinog. 25, 92–98 (1999). geneous prostate carcinoma. Cancer Sci. 94, 764–768 (2003). 18. Jackson, R.J. et al. Loss of the cell cycle inhibitors p21(Cip1) and p27(Kip1) 44. Nakamura, M. et al. Novel tumor suppressor loci on 6q22–23 in primary central enhances tumorigenesis in knockout mouse models. Oncogene 21, 8486–8497 nervous system lymphomas. Cancer Res. 63, 737–741 (2003). (2002). 45. Barghorn, A. et al. Putative tumor suppressor loci at 6q22 and 6q23-q24 are involved 19. Donehower, L.A. et al. Mice deficient for p53 are developmentally normal but in the malignant progression of sporadic endocrine pancreatic tumors. Am. J. Pathol. susceptible to spontaneous tumors. Nature 356, 215–221 (1992). 158, 1903–1911 (2001).

NATURE MEDICINE VOLUME 13 [ NUMBER 9 [ SEPTEMBER 2007 1069