We have analyzed the chromosome 6q21 breakpoint of a non-constitutional t( 6 ; 15 )(q21;q21) rearrangement in sporadic Wilms' tumor. This identified a novel gene encoding a protein with six N-terminal ankyrin repeats linked to a C-terminal HECT ubiquitin-protein ligase domain. We therefore designated this gene HACE1 (HECT domain and Ankyrin repeat Containing E3 ubiquitin-protein ligase 1). HACE1 is widely expressed in human tissues, including mature and fetal kidney. We show that Hace1 protein possesses intrinsic ubiquitin ligase activity, utilizes UbcH7 as a candidate partner E2 enzyme and localizes predominantly to the endoplasmic reticulum. Although the HACE1 locus was not directly interrupted by the translocation in the index Wilms' case, its expression was markedly lower in tumor tissue compared with adjacent normal kidney. Moreover, HACE1 expression was virtually undetectable in the SK-NEP-1 Wilms' tumor cell line and in four of five additional primary Wilms' tumor cases compared with patient-matched normal kidney. We found no evidence of HACE1 mutations or deletions, but hypermethylation of two upstream CpG islands correlates with low HACE1 expression in tumor samples. Our findings implicate Hace1 as a novel ubiquitin-protein ligase and demonstrate that its expression is very low in primary Wilms' tumors.
Analysis of chromosomal translocations in human malignancies has led to the characterization of numerous genes involved in oncogenesis. For example, numerous translocations have been found to either activate proto-oncogenes or to generate gene fusions encoding dominantly acting chimeric oncoproteins ( 1 ). Less commonly, mapping of translocations has highlighted the positions of novel tumor suppressor genes (TSGs) within deletion hotspots of tumors, including the von Hippel Lindau ( VHL ) gene of renal carcinoma ( 2 ), the NF2 gene in neurofibromatosis type 2 ( 3 ), the hSNF5/INI1 gene in malignant rhabdoid tumors ( 4 , 5 ) and the BCSC-1 locus in breast and other cancers ( 6 ).
We recently identified a balanced non-constitutional t(6;15)(q21;q21) translocation in a sporadic Wilms' tumor occurring in a 5-month-old male ( 7 ). Although Wilms' tumor is the most common renal neoplasm in children, accounting for ∼90% of pediatric kidney tumors and 6% of all childhood cancers, the genetics of sporadic Wilms' tumor remain largely unknown ( 8 ). Rearrangements of the 6q21 region have been previously reported in this tumor, including t(5;6)(q21;q21) and t(2;6)(q35;q21) translocations ( 9 – 11 ). LOH of this region appears to be rare in Wilms' tumor ( 9 ), and there is currently little evidence for a 6q21 Wilms' tumor suppressor locus. However, deletions of 6q21 have been widely reported in human malignancies, including carcinomas of the breast, ovary and prostate, as well as in leukemias and lymphomas ( 12 – 15 ), and this region is hypothesized to harbor one or more TSGs ( 16 ). These observations highlight chromosome 6q21 as a region frequently targeted for genomic alterations in human cancer. We therefore postulated that the t(6;15) translocation of the index case might be contributing to Wilms' tumor oncogenesis, either through oncogene activation or by targeting a 6q21 TSGs. This prompted us to search for the 6q21 gene or genes potentially altered by the t(6;15) rearrangement.
In this study, we have characterized the 6q21 region interrupted by the Wilms' tumor t(6;15) translocation. We report that the 6q21 breakpoint maps to a non-coding region with >200 kb from the nearest known gene, but ∼50 kb upstream of a novel open reading frame (ORF). This ORF encodes a protein with a previously unreported domain architecture consisting of six ankyrin repeats linked to a HECT (homologous to E6-AP C-terminus) domain. HECT domains have thus far only been described in E3 ubiquitin-protein ligases ( 17 , 18 ). E3 ligases are essential components of a highly conserved pathway involving conjugation of one or more ubiquitin (Ub) polypeptides to specific substrate proteins, leading in most cases to substrate proteasomal degradation ( 17 , 19 ). We therefore designated this gene HACE1 (HECT domain and Ankyrin repeat Containing E3 ubiquitin-protein ligase 1). Hace1 protein has in vitro and in vivo ubiquitin ligase activity and represents, to our knowledge, the first documented example of a HECT domain containing ubiquitin ligase possessing N-terminal ankyrin repeats. We found that although HACE1 is ubiquitously expressed in normal tissues, including both adult and fetal kidneys, mRNA and protein levels were almost undetectable in the index case and in four of five institutional Wilms' tumor cases compared with matching adjacent normal kidney, and in the SK-NEP-1 Wilms' tumor cell line. Lower expression was strongly associated with hypermethylation of two CpG islands located upstream of the HACE1 locus. Our findings implicate Hace1 as a novel HECT E3 ubiquitin-protein ligase whose lack of expression is associated with sporadic Wilms' tumor.
Mapping the 6q21 breakpoint in Wilms' tumor
To assess candidate genes potentially affected by the t(6;15)(q21;q21) translocation in the index case, we focused on the 6q21 breakpoint as this region has previously been implicated in Wilms' tumors and other human malignancies. Using 6q21 bacterial artificial chromosomes (BACs) for fluorescence in situ hybridization (FISH), we identified several overlapping BACs that spanned the translocation breakpoint (i.e. that showed three signals as opposed to the normal two; Fig. 1 A). On the basis of sequence information from NCBI ( http://www.ncbi.nlm.nih.gov ) and UCSC ( http://genome.ucsc.edu ) genome databases, we assembled a precise map of the involved 6q21 region and narrowed the breakpoint to an ∼12 kb region between the D6S2097 and STSG6084 polymorphic markers (Fig. 1 C). We next determined the restriction enzyme distribution within this region from publicly available genomic sequences. Southern blotting using genomic DNA from the index case and the SB1 probe (Fig. 1 C) allowed us to confirm the position of the 6q21 breakpoint, as rearrangements were detected in the tumor sample but not in matched normal tissue or control samples (shown for Pst I digests in Fig. 1 B).
Identification of the HACE1 gene
Using public database sequences, we next inspected the region surrounding the 6q21 breakpoint for candidate genes potentially affected by the rearrangement. No previously characterized genes were present in this region. We therefore used GenScan gene prediction software to search for novel ORFs. No coding sequences were identified spanning the breakpoint, indicating that the translocation does not directly disrupt a gene in the 6q21 region. However, a novel 2727 bp ORF located ∼50 kb downstream was identified by this strategy, a portion of which matched to the 5′ end of a non-annotated, IMAGE Consortium expressed sequence tag (EST; GenBank accession no. BC034982 ). We obtained this cDNA clone and sequenced it in its entirety, revealing 100% sequence identity with the predicted full-length ORF from 6q21. Alignment to publicly available 6q21 genomic sequences using the UCSC genome browser BLAT tool predicts that this ORF is organized into 24 exons (Fig. 1 C). The nearest known gene, BVES (blood vessel epicardial substance), lies ∼200 kb upstream of the breakpoint (Fig. 1 C).
Protein domain analysis predicted that the 6q21 ORF encodes a 909 amino acid protein (∼103 kDa) possessing six N-terminal ankyrin repeats and a C-terminal HECT domain (Fig. 1 D). Ankyrin repeats are well-documented to mediate protein–protein interactions ( 20 ), whereas HECT is a catalytic domain possessing ubiquitin-protein ligase activity ( 17 , 18 , 21 ). We therefore designated this gene HACE1 . Comparison of the Hace1 protein sequence to public databases using NCBI BLAST and DART tools ( http://www.ncbi.nlm.nih/BLAST ) highlights Hace1 as the first documented example of an E3 ligase in which a HECT domain is linked to ankyrin repeats. Interestingly, when the HACE1 coding sequence was aligned with those of other genomes using the BLAT tool at the UCSC genome database ( http://genome.ucsc.edu ), orthologues of HACE1 appear to be present only in vertebrate species (data not shown). Hace1 ankyrin repeats show >47% sequence similarity with those of the cyclin-dependent kinase inhibitors p16INK4A and p19INK4D (Fig. 1 D), both of which are known to be inactivated in human malignancies including Wilms' tumor ( 22 ). The Hace1 HECT domain is also highly conserved, with more than 53% sequence similarity to the HECT domains of well-characterized ubiquitin ligases including E6-AP, which destabilizes p53 in HPV infected cervical carcinoma cells ( 23 ), Nedd4, which regulates stability of the epithelial Na+ channel (ENaC) and insulin-like growth factor 1 receptor ( 21 ), and Smurf 1 and 2, which are involved in ubiquitination of TGF-β receptors ( 24 ). HECT family E3 ligases are known to contain an invariant C-terminal cysteine residue necessary for thioester bond formation with ubiquitin ( 18 ); in Hace1 this appears to be represented by Cys-876 (Fig. 1 D).
Normal HACE1 expression patterns
We next examined HACE1 expression profiles in normal human tissues by northern analysis. Using a full-length cDNA probe, we found that human HACE1 is expressed as a single mRNA species of ∼4.6 kb in multiple tissues, including strong expression in heart, brain and kidney (Fig. 2 A). As Wilms' tumor is hypothesized to derive from early metanephrogenic stem cells ( 25 – 28 ), we also assessed HACE1 expression in three (day 54–122) human fetal kidney tissue samples. We found that, although variable, mRNA transcripts are expressed in these cells at levels that are similar to or greater than in mature pediatric kidney (Fig. 2 B). There was no obvious trend in expression patterns with the age of the human fetal kidney samples, although this requires a larger sample size for rigorous evaluation. To examine expression of Hace1 protein, we generated polyclonal α-Hace1 antibodies directed towards either the second ankyrin repeat or to full-length recombinant protein. Both antibodies detect a protein with the expected size of 103 kDa identical to that of recombinant Hace1 in both fetal and pediatric kidney cells, as well as in the HEK293 human embryonic kidney cell line (Fig. 2 C).
Hace1 possesses ubiquitin ligase activity in vitro and in vivo
To determine whether Hace1 possesses a ubiquitin ligase activity, we tested the ability of this protein to form a thioester bond with ubiquitin, a well-documented characteristic of HECT E3 ligases ( 17 ). In vitro [ 35 S]-labeled Hace1 was incubated in the presence of GST–Ub and E1 Ub-activating enzyme, along with a panel of E2 Ub-conjugating enzymes (Fig. 3 A). As HECT domains contain a conserved Cys residue which is crucial for Ub transfer to substrate proteins, we mutated Cys-876 to Ser and tested this Hace1-C876S mutant in thioester bond formation. As seen in Figure 3 A (top panel), an additional higher molecular mass band corresponding to the expected size of a Hace1–GST–Ub conjugate (∼140 kDa) was evident in the presence of the UbcH7 E2 enzyme (lane 8). Thioester bond formation was completely abolished by the C876S substitution in Hace1 (Fig. 3 A, bottom panel) or by disulphide bond reducing agent β-mercaptoethanol (Fig. 3 B; compare lanes 1 and 4), indicating specificity of the reaction. These data confirm Hace1 ubiquitin ligase activity in vitro and identify UbcH7 as a candidate partner E2.
To demonstrate the involvement of Hace1 in the ubiquitination of proteins in vivo , HA-tagged Hace1 was stably expressed in NIH3T3 fibroblasts (Fig. 3 C). Cytoplasmic extracts from these cells were then subjected to immunoprecipitation with α-HA antibodies followed by western blotting with α-Ub antibodies to assess whether Hace1 associates with ubiquitinated proteins. As shown in Figure 3 D, overall levels of protein ubiquitination were similar in HA–Hace1 and vector alone cells. However, high molecular weight ubiquitinated proteins could only be detected in α-HA immunoprecipitates from the HA–Hace1 expressing cells (compare lanes 3 and 4). The level of ubiquitinated proteins in these immunoprecipitates was increased in the presence of proteasome inhibitors such as lactacystin or MG132 (data not shown), indicating that at least some of the proteins targeted for ubiquitination by Hace1 are normally degraded by the proteasome. Moreover, we found that Hace1 directly interacts with the 26S proteasomal complex as Hace1 could be immunoprecipitated using antibodies against the 20S core proteasomal subunits (Fig. 3 E, lane 3). Although the identity of Hace1 target proteins in the earlier mentioned immunoprecipitates remains to be established, our data strongly support the involvement of Hace1 in ubiquitination and degradation of cellular proteins.
Subcellular localization of Hace1
To further characterize Hace1, we analyzed its localization within the cell. Subcellular fractionation followed by western blotting demonstrated that in exponentially growing NIH3T3 fibroblasts both endogenous and ectopically expressed Hace1 are found predominantly in the endoplasmic reticulum (ER) and the cytoplasm, although a small amount of endogenous protein is also present in other fractions (Fig. 4 A). ER localization was confirmed by co-immunostaining of Hace1 with the ER resident, BiP (Grp78). As seen in Figure 4 B, a predominantly perinuclear reticular staining pattern of the ER and of the nuclear envelop was observed for both Hace1 and BiP in NIH3T3 cells using immunofluorescence microscopy. This suggests that a significant portion of Hace1 localizes to the ER.
Altered expression of HACE1 in sporadic Wilms' tumor
As the 6q21 breakpoint in the index case occurred ∼50 kb upstream of the HACE1 locus, we hypothesized that HACE1 expression might be affected by the t(6;15) translocation. Moreover, we wondered whether altered HACE1 expression might represent a recurrent molecular alteration in sporadic Wilms' tumor. To begin to address this possibility, we assessed HACE1 mRNA levels in the index case and in five additional institutional Wilms' tumor cases with patient-matched adjacent normal kidney by quantitative RT–PCR (qRT–PCR). The histology of each tumor sample confirmed the presence of at least 80% tumor cells (data not shown). Of the six paired samples, five tumor specimens showed markedly lower HACE1 mRNA expression compared with their matching normal kidney counterparts, including the index case (Fig. 5 A). Although there was variability in the absolute degree to which HACE1 expression was lower, the average decrease in transcript levels was ∼5-fold. Low expression was confirmed at the protein level by western blotting using α-Hace1 antibodies (Fig. 5 B), as expression of Hace1 protein was very low or undetectable in the index case and in the other four cases with low mRNA levels, compared with matching normal kidney. We next assessed HACE1 expression levels and the presence of variant transcripts in the SK-NEP-1 Wilms' tumor cell line by northern analysis. Transcript levels are virtually undetectable in these cells compared with HEK293 human embryonic kidney cells or Ewing tumor and rhabdomyosarcoma cell lines (Fig. 5 C). Hace1 mRNA expression was also low to non-detectable in neuroblastoma cell lines. In agreement with these findings, Hace1 protein levels are extremely low in SK-NEP-1 and neuroblastoma cell lines compared with HEK293, Ewing tumor and rhabdomyosarcoma cells (Fig. 5 D).
Absence of HACE1 mutations or deletions in sporadic Wilms' tumor
Current views hold that Wilms' tumor most likely derives from embryonic kidney ( 26 – 28 ). As we observed strong Hace1 expression in both mature pediatric and fetal kidneys but not in most of the Wilms' tumors analyzed, one possibility is that Hace1 down regulation or inactivation is a common and potentially etiologic event in this disease. However, it is equally plausible that the specific (as yet unknown) blastemal cell giving rise to Wilms' tumor does not express HACE1 and that the observed lack of expression simply reflects the normal expression profile of the putative neoplastic precursor cell. We therefore assessed the Wilms' tumor cohort for potential genetic mechanisms of HACE1 down regulation as a strategy for distinguishing between these possibilities, as loss of TSGs in malignancies typically involves inactivation of both the alleles ( 29 ). Current literature indicates that chromosomal rearrangements, large scale deletions or LOH of the 6q21 region are rare in Wilms' tumor ( 8 – 11 ). Unfortunately, aside from the index case, karyotypes were not available for the Wilms' tumors analyzed in this study. To search for alternative mechanisms that might underlie altered HACE1 expression in Wilms' tumor, we screened the cohort of cases and matched normal kidney as well as the SK-NEP-1 cell line for HACE1 mutations. All exons and intron–exon boundary regions were sequenced from genomic DNA of each sample. No sequence discrepancies between tumor and normal DNAs were found for any of the six cases. Furthermore, no mutations were found in the SK-NEP-1 cell line. Sequence data were compared with the UCSC human genome database and a total of eight HACE1 polymorphisms were identified in the earlier mentioned cohort (Supplementary Material, Table S1). Two of the polymorphisms were within the coding sequences, but only one represents an amino acid change. However, the latter (Asp to Gly at amino acid 399) does not affect any identifiable functional domains within Hace1 protein, and when a non-disease control cohort of 95 individuals was tested for this polymorphism it was present in 1% of samples. Therefore, it most likely represents a low-frequency polymorphism in the general population. Of the five Wilms' tumor cases with low HACE1 expression, at least one heterozygous polymorphism was found in four of the tumor (and normal) DNAs as well as in the SK-NEP-1 cell line, suggesting that LOH of the HACE1 locus is unlikely in this cohort. However, our sequencing strategy cannot rule out micro-deletions of exons or other genomic sequences located between these polymorphisms. Therefore, our data indicate that a genetic mechanism involving point mutations or deletions is unlikely to explain the low HACE1 expression observed in Wilms' tumor, although larger numbers of cases must be assessed to rigorously test this assumption.
Upstream CpG island methylation is associated with low HACE1 expression
Recent studies indicate that promoter hypermethylation may be as frequent as inactivating mutations for reducing TSG expression in human malignancies ( 30 ). Methylation involves the addition of methyl groups to cytosine residues within CpG islands within coding or non-coding sequences of genetic loci, generally resulting in gene silencing ( 31 , 32 ). We therefore investigated the methylation status of CpG islands at the HACE1 locus in the primary Wilms' tumor cohort and in the SK-NEP-1 cell line. There are three CpG islands located within or near the HACE1 locus, one at the promoter extending into exon 1 which contains 88 CG dinucleotides (CpG-88), and two located ∼50 kb upstream of the Hace1 transcriptional start site containing 177 (CpG-177) and 29 (CpG-29) CG dinucleotides repeats, respectively (Fig. 1 C). To assess whether altered methylation might influence HACE1 expression, we analyzed methylation status at each of these CpG islands by digestion with the methylation sensitive restriction enzymes Hae II, Hpa II or Bss HII, followed by semi-quantitative PCR (Fig. 6 A). No evidence of methylation at CpG-88 was found in any of the tumor or normal kidney samples (Fig. 6 A). However, hypermethylation of either CpG-177 and CpG-29, or both, was observed in four out of five Wilms' tumors with low HACE1 mRNA expression (Fig. 6 A) as well as in SK-NEP-1 cells (Fig. 6 B). It was never observed in any of the normal kidney samples (data not shown).
If hypermethylation of CpG-29/CpG-177 is mechanistically related to low HACE1 expression, then pharmacological demethylation of these islands would be expected to restore expression levels. As SK-NEP-1 Wilms' tumor cells express extremely low HACE1 levels and both CpG-29 and CpG-177 are methylated in this line, we treated cells with the methylation inhibitor 5-aza-2-deoxycytidine (5AZ) ( 33 , 34 ) and assessed effects on HACE1 expression. As shown in Fig. 6 B, treatment with this agent reduced methylation at both CpG-29 and CpG-177, particularly at 5 µ M 5AZ. Moreover, this was associated with at least a 4-fold increase in HACE1 mRNA expression in the presence of 5 µ M 5AZ (Fig. 6 C). These findings strongly suggest that methylation status of CpG-29 and CpG-177 influences HACE1 expression. To examine the functional consequences of this observation, we next performed chromatin immunoprecipitation (ChIP) to analyze whether the HACE1 locus exists in an active or inactive chromatin conformation in SK-NEP-1 cells. This was assessed by comparing the relative levels of acetylated-histone H3 versus dimethyl-(Lys79)-histone H3 bound to the HACE1 promoter. Acetylated-histone H3 is known to associate with DNA of active chromatin, whereas dimethyl-(Lys79)-histone H3 correlates with chromatin present in the inactive conformation of heterochromatin or silenced genes ( 35 – 37 ). As shown in Figure 6 D, the promoter region of the HACE1 gene exists predominantly in an inactive chromatin conformation in normally growing SK-NEP-1 cells (i.e. associated with dimethyl-(Lys79)-histone H3 by ChIP). However, treatment with 5 µ M 5AZ reverses this pattern, switching the HACE1 promoter to an active chromatin conformation associated predominantly with the acetylated-histone H3. As a control, we performed identical studies with HEK293 cells (Fig. 6 D, left panel) which abundantly express HACE1 (Fig. 4 ). Consistent with its high expression, the HACE1 locus exists in an active chromatin organization in these cells, and no change in chromatin structure is evident after treatment with 5AZ. Taken together, our data are consistent with a role for CpG-29 and CpG-177 hypermethylations in influencing HACE1 expression, although additional studies are required to confirm this possibility and whether it represents an acquired alteration in tumor cells versus the normal pattern of methylation in the Wilms' tumor neoplastic precursor cell.
In this study, we demonstrate that a previously uncharacterized E3 ubiquitin-protein ligase gene is located ∼50 kb downstream of the 6q21 breakpoint of a t(6;15) translocation in sporadic Wilms' tumor. This gene, which we designated HACE1 , encodes a 103 kDa protein containing six N-terminal ankyrin repeats connected via a linker region to a C-terminal HECT domain. The latter has to date only been described in E3 ligases ( 17 ), implicating the Hace1 protein as a new member of the HECT family of E3 ubiquitin-protein ligases. The protein ubiquitination process involves a highly conserved pathway in which one or more Ub polypeptides become conjugated to specific substrate proteins ( 17 , 19 ). This multistep process involves a cascade of three different classes of enzymes. An ATP-dependent Ub-activating E1 enzyme first forms a thioester bond with Ub, and then transfers the activated Ub moiety to one of a number of different E2 Ub-conjugating (Ubc) enzymes. Finally, E3 ubiquitin ligases catalyze the transfer of Ub from the cognate E2 to a lysine on the substrate protein. Binding of substrate proteins appears to be mediated by the N-terminal protein–protein interaction domains of E3 ligases ( 17 , 18 , 21 ).
HACE1 is widely expressed in human tissues, including both fetal and mature kidneys. The Hace1 protein forms thioester bonds with ubiquitin as expected for a HECT family E3 ligase ( 17 , 18 ), and is associated with high molecular weight ubiquitinated proteins within cells. Among a panel of E2 ubiquitin carrier proteins, thioester bond formation in vitro occurred in the presence the UbcH7 E2 enzyme, and weakly with UbcH6 and UbcH5b, all of which are common partners of other HECT domain containing E3 ubiquitin-protein ligases. We also found that Hace1 associates with ubiquitinated proteins and components of the 26S proteasomal complex, indicating that at least some of the proteins targeted for ubiquitination by Hace1 are likely to be degraded by the proteasome. Using cell fractionation and immunofluorescence microscopy, we found that Hace1 localizes predominantly to the ER indicating its possible involvement in ER-associated protein degradation (ERAD). ERAD is a quality control process that selectively directs degradation of misfolded and aberrant proteins through ubiquitination within the ER followed by transport to the proteasome ( 17 , 38 ). More studies are necessary to demonstrate whether Hace1 contributes to this process.
Most HECT E3 ligases characterized to date contain N-terminal WW domains for substrate interaction and targeting ( 17 , 18 , 21 ). In fact, the presence of ankyrin repeats within an E3 ligase as in Hace1 has not been previously described. Given this modular structure, Hace1 may represent a novel sub-family of HECT E3 ubiquitin-protein ligases. Ankyrin motifs are well known to mediate protein–protein interactions ( 20 ), and therefore likely serve the same function in Hace1. Alternatively, the ankyrin repeats may be necessary for Hace1 to link with non-substrate proteins in higher order signaling complexes, a function that has been described previously for ankyrin motifs ( 20 , 39 ). Interestingly, the Hace1 ankyrin repeats show high sequence similarity to those of the cyclin-dependent kinase inhibitors p16INK4A (CDKN2A) and p19INK4D (CDKN2D), which inhibit the cell cycle by binding to and inhibiting CDK4 and CDK6.
HACE1 expression was very low both at the mRNA and protein levels in five of the six sporadic Wilms' tumors compared with patient-matched normal kidney, including the index case. Moreover, expression of this gene is virtually undetectable in SK-NEP-1 Wilms' tumor cells. This raises the possibility that loss of HACE1 expression may be a recurrent alteration in Wilms' tumor, and that HACE1 inactivation plays a role in the pathogenesis of Wilms' tumor. On the other hand, this may reflect the normal expression profile of HACE1 in the Wilms' tumor neoplastic precursor cell; i.e. HACE1 may not normally be expressed in the putative blastemal cell of origin of this disease. Typically, inactivation of TSGs in human malignancies involves targeting of both alleles by genetic or epigenetic mechanisms ( 29 ). Therefore, we searched for mechanistic evidence of HACE1 inactivation in Wilms' tumor. As chromosomal translocations or other cytogenetic abnormalities involving 6q21 have only infrequently been reported in Wilms' tumors ( 7 , 9 – 11 ), this is unlikely as a mechanism of HACE1 loss. Moreover, in the index case of the present study the 6q21 breakpoint did not directly disrupt the HACE1 gene, but occurred outside of the coding region. We also failed to detect HACE1 mutations in any of the cases with low HACE1 expression, and LOH of this gene was not evident by analysis of several informative markers identified within the HACE1 locus. The latter is consistent with other studies indicating that LOH of 6q21 is rare in Wilms' tumor ( 9 ). Although further studies using larger case cohorts to definitively rule out that mutations or deletions of HACE1 occur in sporadic Wilms' tumor, our results to date do not support a role for genetic mechanisms in HACE1 loss in Wilms' tumor.
Gene inactivation by methylation appears to be at least as frequent as inactivating mutations for disrupting TSG expression in sporadic tumors ( 30 ). For example, in sporadic breast and ovarian carcinomas, BRCA1 mutations are highly infrequent and epigenetic silencing by promoter methylation is the only apparent mechanism for loss of function in most cases ( 40 ). We therefore examined whether the CpG islands associated with the HACE1 locus were differentially methylated in tumor versus normal kidney samples. Although the HACE1 promoter CpG island was not methylated, we detected hypermethylation of CpG-29 and CpG-177, located ∼50 kb upstream of HACE1 exon 1, in four of the five Wilms' tumors with reduced mRNA expression, and also in the SK-NEP-1 cell line. Moreover, treatment with the methyltransferase inhibitor 5AZ not only blocked methylation of these islands but also restored HACE1 mRNA expression in SK-NEP-1 cells. One obvious caveat of these experiments is that 5AZ as a pharmacological inhibitor has a global effect on CpG methylation. We therefore performed ChIP analysis of the HACE1 promoter region using antibodies to acetylated-histone H3 or dimethyl-(Lys79)-histone H3, which associate with actively transcribed versus inactive, silenced chromatin, respectively ( 36 , 37 ). These experiments indicated that 5AZ treatment of SK-NEP-1 cells shifts the HACE1 promoter from an inactive chromatin conformation to that of active chromatin, in keeping with the increased expression of this gene in response to 5AZ. In contrast, the HACE1 promoter of HEK293 human embryonic kidney cells, which express high levels of HACE1 , exists in an active conformation under normal growth conditions and there is no change in chromatin structure upon treatment with 5AZ. These findings provide compelling evidence that HACE1 expression is influenced by upstream CpG island methylation. It is important to note, however, that our findings do not establish whether the correlation between CpG island hypermethylation and low HACE1 expression is specific to transformed cells. That is, they do not distinguish between acquired hypermethylation leading to HACE1 inactivation as an etiologic event in Wilms' tumor oncogenesis, versus hypermethylation leading to low HACE1 expression as part of the normal transcriptional regulation of this gene in the Wilms' tumor precursor cell. Functional studies are necessary to determine whether the Hace1 protein has suppressive effects on cell growth, and whether its loss contributes to Wilms' tumor oncogenesis.
Although methylation of CpG dinucleotides outside of promoters has been documented to influence gene transcription ( 41 ), most studies which describe epigenetic silencing of genes have focused on CpG islands located within promoters or coding regions of genes ( 30 ). Recently, distance effects on insulator methylation were reported for the H19 gene, whereby imprinting by CpG methylation at more distant sites regulated mono-allelic expression of both H19 and Igf2 ( 42 , 43 ). In addition, reduced expression of Sonic hedgehog ( SHH ) has been associated with chromosomal rearrangements occurring 15–250 kb from the SHH locus in holoprosencephaly ( 44 , 45 ). It is possible that long distance effects of CpG island methylation on transcription may be a more general mechanism for gene silencing in human cancers. This might explain how chromosomal translocation breakpoints located at long distances from promoters of potential TSGs could influence gene silencing. It is well-documented that heterochromatin has a silencing effect on adjacent euchromatin, likely through spreading of histone methylation ( 46 – 48 ). It is possible that by juxtaposing heterochromatin in the vicinity of potential TSGs, a chromosomal translocation can exert long distance transcriptional suppression on that locus by effecting local methylation changes. For the index case in this study, the 6q21 breakpoint mapped very closely to if not within CpG-29 or CpG-177. Moreover, both of these islands were methylated in tumor tissue but not in adjacent normal kidney tissue. Therefore, the direct effects of the t(6;15) in this case might have been to trigger methylation of these CpG islands, resulting in HACE1 silencing. Alternatively, effects on methylation due to translocation breakpoints occurring in this region may not be limited to HACE1. It is possible that long distance effects on expression may also affect other genes in the region, and that targeting of these genes for activation or inactivation may actually be contributing to Wilms' tumor oncogenesis rather than HACE1. However, the only known genes mapping to within ∼500 kb of the breakpoint include the telomeric genes BVES , Popeye domain containing protein 3 ( POPDC3 ) and prolyl endopeptidase ( PREP ) ( http://genome.ucsc.edu/ ). None of these genes have been implicated as oncogenes or TSGs, although this remains to be rigorously determined. No known genes are located within at least 1 Mb centromeric to the breakpoint. Interestingly, a recent reported case of Wilms' tumor with a 6q21 breakpoint maps to a 1.3 Mb region we now know includes the HACE1 locus ( 9 ), suggesting that other rearrangements of this region may similarly occur near HACE1 and adjacent loci in Wilms' tumor.
The 6q21 region is commonly deleted in several other human neoplasms. Examples include malignancies of the prostate, breast and ovaries, as well as leukemias and lymphomas ( 12 – 15 ). For example, a recent study found that ∼50% of prostate cancers had deletions of the 6q21 region ( 15 ). A number of candidate metastasis suppressor genes have also been mapped to chromosome 6q21 ( 49 ). There is emerging interest in the role of protein degradation in neoplasia, and E3 ligases have been implicated in both tumor formation and suppression ( 50 ). In fact, the E6-AP HECT E3 ligase was originally defined by its ability to ubiquitinate and promote p53 degradation after recruitment by the E6 oncoprotein in HPV infected cervical carcinoma cells ( 23 , 51 , 52 ). It will be important to further assess the possibility that the Hace1 E3 ligase plays a role in human cancer.
MATERIALS AND METHODS
FISH and Southern blotting
Single color FISH was performed on metaphase and interphase nuclei from the index Wilms' tumor case. BAC probes used for mapping were obtained from the RPCI-11 human BAC library, labeled with Spectrum Green or Spectrum Red kits (Vysis) and hybridized to denatured slides as described previously ( 53 ). For Southern blotting, genomic DNA (∼15 µg) extracted from frozen tissue sections was first digested with respective restriction enzymes, and then analyzed by hybridization with [α- 32 P]dATP labeled genomic probe generated by PCR from a human DNA template SB1 (primers: forward, GGAAA CAAAA GCAAA GCGAC CCAAC TAT; reverse, GGCGG CCGAG ACCTG AGACC).
The HACE1 coding region from the IMAGE cDNA clone no. 4838835 (ATCC; GenBank accession no. BC034982 ) was excised using Asc I and Bsa AI, blunted with T4 DNA polymerase and ligated in frame into blunted Xho I site of the mammalian expression vector pcDNA3–HA (Invitrogen) or pET-15b (Novagen). The HA-tagged (hemagglutinin peptide tag, YPYDVPDYA) construct was excised from pcDNA3–HA with Sac II, blunted with T4 DNA polymerase, and subcloned into the Hpa I site of pMSCV hygro (Clontech). To construct the plasmid encoding the Cys to Ser mutant (C876S), the 3′ Xba I fragment of HACE1 was subcloned from pcDNA3-HA-Hace1 into the Xba I site of pBluescript-II. Site directed mutagenesis was done using the QuickChange kit (Stratagene) (primers: forward, TCAAG CACAT CCATC AACAT G; reverse, CATGT TGATG GATGT GCTTGA). The mutation was sequence verified and the Xba I fragment was than replaced into the original pcDNA3–HA–Hace1 construct. The N-terminal Hace1 fragment, p75, was generated by removal of a 3′ Xba I fragment of HACE1 from the pET-15b–Hace1.
NIH3T3 cells (ATCC) were grown in Dulbecco's modified Eagle's medium (DMEM) (Invitrogen) containing 10% calf serum (CS) (Invitrogen). SK-NEP-1, SK-N-SH, SK-N-MC, HEK293 and Bosc23 cells were grown in DMEM supplemented with 10% fetal bovine serum (FBS) (Invitrogen). KCNR, SAN-2, IMR-32, Birch and CT-10 were maintained in RPMI 1640 (Invitrogen) with 15% FBS. NIH3T3 cell lines stably expressing HA–Hace1 were generated using a retroviral system as described previously ( 54 ). Briefly, ecotropic virus packaging Bosc23 cells were transfected with empty vector (pMSCV hygro ) or pMSCV hygro –HA–Hace1 plasmid using calcium phosphate. Retrovirus-containing supernatants were collected 48 h after transfection, filtered and incubated with NIH3T3 cells for 24 h. Infected cells were selected for 4–8 days with hygromycin B (600–800 µg/ml) (Invitrogen).
Methylation sensitive restriction enzyme digests followed by semi-quantitative PCR have been performed as described ( 55 ). Briefly, 100 ng of DNA, extracted from tumor and normal kidneys frozen tissue sections, was digested to completion (overnight) with Eco RI ± Hae II, Eco RI ± Bss HII or Eco RI ± Hpa II (NEB) as indicated in Figure 6 . Ten nanograms of digested DNA was used for semi-quantitative PCR with amplicons surrounding CpG islands near the HACE1 locus (Fig. 1 ). The 5′ region of the MIC2 gene, previously shown to be unmethylated ( 56 ), was used as a control for complete digestion. All reactions were performed using Platinum Taq (Invitrogen) with the included buffer supplemented with 10% DMSO and carried out with the following conditions: CpG-29 (794 bp forward: GGAAA CAAAA GCAAA GCGAC CCAAC TAT, reverse: GGCGG CCGAG ACCTG AGACC, 100 n M each), 50 µ M dNTPs, 1.0 m M MgCl 2 , 94°C 30 s, 56°C 10 s, 72°C 60 s, 30 cycles. CpG-177 (804 bp, forward: TGTGC TGTTC GGAAT GATGT, reverse: CTAGC CTGGG TGTGA GAGGG, 100 n M each), 100 µ M dNTPs, 1.5 m M MgCl 2 , 94°C 45 s, 58°C 20 s, 72°C 120 s, five cycles, 94°C 30 s, 56°C 10 s, 72°C 90 s, 30 cycles. CpG-88 (939 bp, forward: CCCCG ATGCA GCTTA AAGTA, reverse: GAGGG TAGGA GGAGC AGGG, 100 n M each), 50 µ M dNTPs, 1.0 m M MgCl 2 , 94°C 45 s, 58°C 10 s, 72°C 90 s, 30 cycles. MIC2 (512 bp, forward: AGTAT CTGTC CTGCC GCC, reverse: TTTGC AACTC CGACA ACAAA CGC, 100 n M each), 50 µ M dNTPs, 1.0 m M MgCl 2 , 94°C 45 s, 55°C 15 s, 72°C 45 s, 35 cycles. The optimal number of cycles for each reaction was determined empirically by testing a range of 25–40 cycles. Above conditions are those in which the most reproducible differences in amplification within the exponential phase of amplification were achieved. All reactions were performed in triplicate.
To assess the effect of CpG island methylation on HACE1 expression efficiency, SK-NEP-1 cells were treated with methylation inhibitor 5-AZ (Sigma). Treatment was done twice for 24 h over a 7-day period (on days 2 and 5) as previously described ( 57 ). DNA and RNA were isolated on day 6 and HACE1 expression and methylation of the upstream CpG islands were assayed by qRT–PCR and semi-quantitative PCR, respectively. Each experiment was repeated four times.
Total RNA was extracted using Trizol reagent (Invitrogen) and converted to cDNA using 2 µg RNA in a random primed synthesis with the Superscript II Reverse transcriptase kit (Invitrogen). qRT–PCR was done using the TaqMan 5′ exonuclease assay. All primer/probe sets were designed to span exon boundaries to eliminate the risk of template contamination by genomic DNA and the need for DNase1 pre-treatment of RNA. HACE1 gene specific primer/probe set (forward: TCTTA CAGTT TGTTA CGGGC AGTT, probe: [6FAM]CAAAC CCACC ATGTG GGACC CTG[TAMRA], reverse: CAATC CACTT CCACC CATGAT) was multiplexed with VIC-MGB labeled β-actin endogenous control primer/probe kit (Applied Biosystems) using TaqMan universal PCR master mix (Applied Biosystems) and standard conditions. An ABI 7000 sequence detection system (Applied Biosystems) was used to run the PCR reactions and measure fluorescence at each cycle. Generated data were further analyzed using Microsoft Excel. Each PCR reaction was performed in quadruplet and each sample was analyzed independently at least twice.
SK-NEP-1 or HEK293 cells were treated with or without 5 µ M 5AZ as described earlier. Immunoprecipitation of DNA representing active versus inactive chromatin was performed using a ChIP assay kit (Upstate) according to the manufacturer's protocol. Briefly, DNA and protein were cross-linked in culture with 1% formaldehyde. For each reaction 1×10 6 cells were collected in 200 µl SDS lysis buffer (1% SDS, 10 m M EDTA, 50 m M Tris–HCl, pH 8.1, 1 m M PMSF, 1 µg/ml aprotinin, 1 µg/ml pepstatin A) and sonicated to shear genomic DNA to a size range of ∼300–1000 bp. Lysates were diluted 1 : 10 with ChIP buffer (0.01% SDS, 1.1% Triton X-100, 1.2 m M EDTA, 16.7 m M Tris–HCl, pH 8.1, 167 m M NaCl) and immunoprecipitated with the appropriate antibody overnight at 4°C with agitation. α-Acetyl-histone H3 (Lys9/Lys14) antibodies (Upstate) were used to immunoprecipitate active chromatin fragments ( 58 ), whereas α-dimethyl-histone H3 (Lys79) antibodies (Upstate) were used to immunoprecipitate inactive chromatin ( 35 ) with protein A beads. Protein A agarose/Salmon sperm DNA slurry (Upstate) added to lysates without antibody was used as a negative control (data not shown). DNA was recovered by proteinase K digestion and phenol/chloroform extraction after washing and reversal of cross-links. Equal volumes were utilized to maintain the quantitative nature of the assay. The recovered DNA pellet was air dried and resuspended in 25 µl of TE (10 m M Tris–HCl, 1 m M EDTA, pH 8.0). Immunoprecipitated HACE1 levels were determined by using PCR amplifying the 5′ end of the gene, overlapping the CpG-88 island. PCR was carried out with Platinum Taq (Invitrogen) with the included buffer, supplemented with 10% DMSO and the following conditions: 2.5 µl recovered ChIP DNA, primers (297 bp product, forward: CGGCT CACCC TCGGG CAACT CC, reverse: CGGCG GCGGG TGTAC TGTAG GTGGT C, 200 n M each), 100 µ M dNTPs, 2 m M MgCl 2 , 94°C 30 s, 55°C 30 s, 72°C 45 s, 32 cycles. Input DNA amounts were verified by PCR using the same conditions as mentioned using DNA extracted from a 100 µl aliquot of ‘ChIP buffer’ diluted cell lysate taken prior to addition of antibodies, as recommended by the ChIP assay kit (Upstate) (shown as ‘input fraction’ in Fig. 6 D). Input was also verified with 2.5 µl of recovered immunoprecipitated DNA using the a PCR assay for 5′ end of the unmethylated MIC2 gene ( 55 , 56 ), as described earlier (data not shown). The optimal number of cycles for each reaction was determined empirically by testing a range of 25–40 cycles. The conditions mentioned earlier are those in which the most reproducible differences in amplification within the exponential phase of amplification were achieved. The entire ChIP experiment was repeated four times for each cell line.
Genomic DNA from Wilms' tumors, patient-matched normal kidney (same cohort used for expression analysis) and the SK-NEP-1 cell line were used as templates for sequencing. Reference sequence for primer design, exon layout and assembly was obtained from the UCSC human genome database ( http://genome.ucsc.edu/ ). Forward and reverse primers for each of the 24 HACE1 exons contained the 21M13F (TGTAAAACGACGGCCAGT) or M13R (CAGGAAACAGCTATGAC) sequences at their 5′ ends, respectively. After PCR of each exon from genomic DNA samples automated sequencing was performed by standard methods. Sequence reads were base-called using Phred software and subsequently assembled with reference sequences using Phrap software ( 59 , 60 ). Contigs of sequence traces corresponding to each exon were examined using PolyPhred software ( 61 ) for detection of heterozygotes and visualized in Consed software ( 62 ) to facilitate verification of sequence variants by examination of individual traces.
Thioester bond formation assay
L -[ 35 S]-Methionine labeled Hace1 proteins were synthesized in vitro in a coupled transcription/translation system (Promega) using pcDNA3-HA–Hace1 plasmids coding for the wild-type and mutant Hace1 proteins. Reaction mixtures contained 2 µl of the translation reaction, GST–Ub (1 µg), 50 m M KCl, 5 m M MgCl 2 , 2 m M ATP, 10 m M creatine phosphate, 0.5 U of creatine phosphokinase, 1 m M DTT, and 20 m M Tris–HCl (pH 7.8). Reactions were also supplemented with E1 (∼0.5 µg) and the respective E2s (∼0.4 µg) (BostonBiochem). After incubation for 30 min at 30°C, reactions were stopped with 2×SDS–PAGE sample buffer at the absence or presence of 300 m M β-mercaptoethanol. Reactions containing β-mercaptoethanol were then boiled for 3 min, and those lacking β-mercaptoethanol were incubated at room temperature for 20 min before loading. Reaction products were resolved by SDS–7% PAGE and visualized by autoradiography.
Purification of the recombinant proteins from bacteria and antibody preparation
Hace1 or p75, lacking the HECT domain, were expressed as His-tagged fusion proteins in Escherichia coli BL21(DE3). After 4 h of IPTG induction, bacteria were lysed by sonication in buffer containing 2 M NaCl, 10 m M Tris–HCl (pH 7.6) and 0.5 m M PMSF. Cell debris was removed by centrifugation at 10 000 g for 20 min at 4°C. The supernatant was diluted 4-fold with 10 m M Tris–HCl (pH 7.6), 0.5 m M PMSF and passed through a Ni + –Sepharose column (Qiagen). After washing the column with loading buffer [500 m M NaCl, 10 m M Tris–HCl (pH 7.6)], bound proteins were eluted with the same buffer containing 300 m M imidazole and dialyzed against 200 m M KCl, 10 m M Tris–HCl (pH 7.6). α-Hace1 antibodies were generated by subcutaneous immunization of rabbits with the full-length Hace1 recombinant protein or with the N-terminal Hace1 peptide (CLVLL LKKGA NPNYQ DISG) and used for western blotting at 1 : 2000 dilution.
Immunoprecipitation and western blotting
Immunoprecipitation and western blotting were done essentially as described previously ( 53 , 63 ). Briefly, immunoprecipitations were performed using 500 µl of cytoplasmic cell extracts (∼500 µg–1 mg of total protein) and 3 µl (∼3 µg) of the appropriate antibodies immobilized on 20 µl of protein (A+G)–Sepharose beads (Qiagen) for 2–3 h at 4°C with rotation. For immunoprecipitation of 26S proteasome, the mixture of antibodies directed against α and β subunits [Calbiochem; 5 µl of each antibody per 30 µl of protein(A)–Sepharose] has been used. Monoclonal mouse α-HA (Babco), α-BiP, α-Grb2 and α-UbcH7 (Transduction Laboratories), α-p97 (VCP; Research Diagnostics Inc.), goat α-actin, rabbit α-Ub (Santa Cruz) and α-acetyl-histone H3 (Upstate) antibodies were used at 1 : 1000 dilution.
To prepare cell extracts, NIH3T3 cells expressing HA-Hace1 were trypsinized, washed three times with PBS and lysed with buffer containing 50 m M KCl, 2 m M MgCl 2 , 2 m M DTT, 0.25% NP-40, 20 m M Hepes–KOH (pH 7.8) and the proteasome inhibitors MG132 (20 µ M ) or lactacystin (10 µ M ). Nuclear, mitochondrial and post-mitochondrial fractions were separated by sequential centrifugation at 1000 g for 5 min followed by centrifugation at 10 000 g for 15 min. Nuclei were then additionally purified from the re-suspended 1000 g pellet fraction by spinning down through a 50% glycerol cushion. The post-ribosomal supernatant was obtained by a further 20 min centrifugation of the post-mitochondrial fraction at 100 000 rpm in a TLA-100 centrifuge (Beckman). Cell fractions were normalized for protein concentration using Protein Assay kit (BioRad).
NIH3T3 cells exponentially growing on coverslips were rinsed with PBS and fixed with −20°C cold methanol for 10 min. To detect Hace1 and BiP, coverslips were incubated overnight with a mixture of rabbit α-Hace1 (1 : 2000 dilution) and mouse α-BiP antibodies (1 : 1000 dilution) followed by the secondary antibodies Rhodamine Red-X-conjugated goat α-rabbit and Oregon Green 514-conjugated goat α-mouse (Molecular Probes). Slides were counterstained with DAPI and analyzed using a Zeiss Axioplan epifluorescent microscope equipped with a COHU-CCD camera. Control staining with pre-immune antibodies or no primary antibodies showed no signal (data not shown).
Supplementary Material is available at HMG Online.
We would like to thank Seong-Jin Kim for reagents, Joan Mathers and Heather Wildgrove for technical assistance, Francis Ouellette, Yaron Butterfield, Jacqueline Schein, Stefanie Butland and Sohrab Shah for bioinformatics support, as well as Fred Barr, Catherine Anderson, Lola Maksumova, Gregor Reid and Fan Zhang for helpful discussions. Normal fetal kidney samples were provided by Alan G. Fantel, Birth Defects Research Laboratory, University of Washington. This study was supported by the National Cancer Institute of Canada and by a Translational Research Grant from the Children's Oncology Group (to P.H.B.S.). This work was also funded by the Johal Program in Pediatric Oncology Basic and Translational Research at the BC Research Institute for Children's and Women's Health.