Sequencing of Candidate Chromosome Instability Genes in Endometrial Cancers Reveals Somatic Mutations in ESCO1, CHTF18, and MRE11A

Most endometrial cancers can be classified histologically as endometrioid, serous, or clear cell. Non-endometrioid endometrial cancers (NEECs; serous and clear cell) are the most clinically aggressive of the three major histotypes and are characterized by aneuploidy, a feature of chromosome instability. The genetic alterations that underlie chromosome instability in endometrial cancer are poorly understood. In the present study, we used Sanger sequencing to search for nucleotide variants in the coding exons and splice junctions of 21 candidate chromosome instability genes, including 19 genes implicated in sister chromatid cohesion, from 24 primary, microsatellite-stable NEECs. Somatic mutations were verified by sequencing matched normal DNAs. We subsequently resequenced mutated genes from 41 additional NEECs as well as 42 endometrioid ECs (EECs). We uncovered nonsynonymous somatic mutations in ESCO1, CHTF18, and MRE11A in, respectively, 3.7% (4 of 107), 1.9% (2 of 107), and 1.9% (2 of 107) of endometrial tumors. Overall, 7.7% (5 of 65) of NEECs and 2.4% (1 of 42) of EECs had somatically mutated one or more of the three genes. A subset of mutations are predicted to impact protein function. The co-occurrence of somatic mutations in ESCO1 and CHTF18 was statistically significant (P = 0.0011, two-tailed Fisher's exact test). This is the first report of somatic mutations within ESCO1 and CHTF18 in endometrial tumors and of MRE11A mutations in microsatellite-stable endometrial tumors. Our findings warrant future studies to determine whether these mutations are driver events that contribute to the pathogenesis of endometrial cancer.


Introduction
Uterine cancer is the most commonly diagnosed gynecologic malignancy in the United States and is the eighth leading cause of death from cancer among American women [1]. Endometrial cancers (ECs) account for the vast majority of uterine cancers. Endometrioid, serous, and clear cell carcinomas represent the three major histological subtypes of EC. Each subtype arises from distinct precursor lesions, has distinct clinical behaviors and distinct molecular etiologies [2], [3].
MSI reflects a mutator phenotype resulting from defective mismatch repair (reviewed in [18]). In sporadic endometrial cancers, most instances of MSI are explained by hypermethylation of the MLH1 promoter, loss of MSH2 expression, or somatic mutations in MSH6 (reviewed in [19]). Aneuploidy has recently been suggested to result from a step-wise process resulting from an acquired tolerance for a non-diploid genome, via inactivation of the p53 pathway, as well as aberrant chromosome segregation [20]. Although inactivating mutations in TP53 and p53 protein stabilization are frequent in NEECs, occurring in up to 90% of serous tumors (reviewed in [19]), the genetic basis of chromosome missegregation in NEECs remains poorly understood. In yeast, chromosome missegregation can arise from mutations in genes that regulate sister-chromatid cohesion [21], [22]. Mitotic sister chromatid cohesion refers to the physical linkage of replicated sister chromatids by the cohesin protein complex until anaphase, to ensure the faithful segregation of sister chromatids into daughter cells. In S. cerevisiae, the cohesin complex consists of the Smc1, Smc3, Scc1, and Scc3 subunits and is loaded onto chromatin at the end of G1 by a process that requires the Scc2-Scc4 complex [23], [24], [25]. Subsequent cohesion establishment depends upon the acetylation of Smc3 by the Eco1 acetyltransferase [26], [27], [28], as well as the activities of Chl1 and the alternative replication factor C (Rfc) complex Ctf18-Ctf8-Dcc-Rfc [21], [29]. Cohesion establishment is antagonized by the activities of the Wpl1-Pds5 complex and the Elg1-Rfc complex [30], [31].
In the present study, we sought to determine whether additional sister chromatid cohesion genes are somatically mutated in endometrial tumors. We resequenced the human orthologues of 19 genes implicated in the regulation of sister chromatid cohesion, as well as two additional candidate chromosome instability (CIN) genes, from 24 primary NEECs. Mutated genes were subsequently sequenced from 83 additional endometrial tumors. Our study uncovered nonsynonymous somatic mutations in ESCO1, CHTF18, and MRE11A in a subset of human endometrial tumors.

Ethics statement
The NIH Office of Human Subjects Research determined that this research was not ''human subjects research'' per the Common Rule (45 CFR 46), and therefore that no IRB review was required for sequencing of the anonymized samples in this study. Clinical specimens Anonymized, primary endometrial tumor tissues (45 serous, 20 clear cell, and 42 endometrioid) and matched histologically normal tissues were obtained from the Cooperative Human Tissue Network, or from the Biosample Repository at Fox Chase Cancer Center, Philadelphia PA. Six cases of matched tumor and normal DNAs were procured from Oncomatrix. All tumor tissues were collected before treatment. An hematoxylin and eosin (H&E) stained section of each tumor specimen was reviewed by a pathologist to verify histology and to delineate regions of tissue with high ($70%) tumor cell content.

Nucleic acid isolation and identity testing
Genomic DNA was isolated from macrodissected tissue using the Puregene kit (Qiagen). Paired, tumor-normal DNAs were genotyped using the Coriell Identity Mapping kit (Coriell) according to the manufacturer's instructions. Genotyping fragments were size separated on an ABI-3730xl DNA analyzer (Applied Biosystems) and alleles were scored using GeneMapper (Applied Biosystems).

Identification of orthologous genes
A consolidated list of known and candidate human orthologues of yeast chromosome stability genes (with demonstrated roles in sister chromatid cohesion) was identified through standard crossspecies approaches. Briefly, InParanoid 7 and HomoloGene databases were queried to identify known orthologues, while BLASTp was employed to identify the top-hit candidates (based on E-value) from the non-redundant protein sequences within the Homo sapiens database.

Reverse transcriptase PCR (RT-PCR)
Total RNA was extracted from 5 endometrioid and 2 serous endometrial cancer cell lines using Trizol Reagent (Ambion). A commercially available human total RNA control mix (Applied Biosystems) was used as a positive control. cDNA synthesis was performed on 1mg of total RNA with the high-capacity cDNA archive kit using random hexamers (Applied Biosystems). cDNAs (0.2ml) were amplified by PCR using the primer pairs provided in Table S1. Amplification consisted of 40 cycles using the following parameters: 94uC for 30 s, 58uC for 30 s and 72uC for 30 s, with a final extension step at 72uC for 10 min. PCR products were separated on a 1% agarose gel stained with ethidium bromide in 0.56 TAE buffer and visualized under ultraviolet illumination.

Primer design and PCR amplification
Primer pairs were designed, using published methods [47], to target 97.4% (458 of 470) of all exons of the 21 genes in the mutation discovery screen (Table S2), and all exons of the three genes in the mutation prevalence screen (Table S3). PCR conditions are available on request.
Nucleotide sequencing PCR products were subjected to bidirectional Sanger sequencing using M13 primers and the BigDye Terminator Version 3.1 Cycle Sequencing Kit (Applied Biosystems). Sequencing reactions were run on ABI 3730xl DNA Analyzers (Applied Biosystems). Sequence trace quality was assessed with the base-calling program, Phred [48], [49]. All traces were included in the subsequent analysis, since deletion-insertion polymorphisms can mimic poor quality data from a Phred-quality measure, but may contain valid sequence data. All sequences for a given primer pair were assembled using Consed [50]; overlapping amplimers were assembled separately to allow independent cross-validation of calls in overlapping regions. Sequence variants, including singlenucleotide differences and short (,100 base pair) insertions and deletions, were identified using PolyPhred v6.11 [51] and an inhouse algorithm (DIPDetector) optimized for improved sensitivity in finding insertions and deletions from aligned trace data. DIPDetector analyzes Sanger sequencing traces and predicts insertions and deletions by first examining read alignments for homozygous variants. It then searches for signatures of heterozygous insertions and deletions within the output of the basecaller phred run with the -poly option [49]. After forming two vectors containing the bases with highest peak areas at each position of the read (or assigning the highest area peak to both vectors when the second largest peak has an area less than 10% the size of the largest peak), DIPDetector attempts to phase these vectors by inserting potential shifts of all possible sizes into all possible positions of the read, and scores these shifts according to how well the resulting shifted vectors match the observed bases within the trace. Human genome assembly hg18 (NCBI Build 36.1) was used as the reference sequence. Variant positions were cross-referenced to dbSNP (Build 129) entries to identify known polymorphisms. To determine whether novel variants were somatic mutations or germline polymorphisms, the appropriate tumor DNA and matched normal DNA were re-amplified in an independent PCR followed by sequence analysis of the variant position. The predicted impact of somatic mutations on protein function was evaluated in silico using Mutation Assessor release 2 (http:// mutationassessor.org/), SIFT (http://sift.jcvi.org/), and Polyphen-2 (http://genetics.bwh.harvard.edu/pph2/index.shtml).

Calculation of discovery screen power
The estimated power to detect one gene mutation in a set of 24 tumors is 1-(1-X)ˆ24, where X is the actual fraction of tumors with a mutation in that gene.

Results and Discussion
In a mutation discovery screen, we analyzed 24 primary NEECs for the presence of nucleotide variants within the coding exons and splice junctions of 21 candidate chromosome instability genes, which are expressed, at variable levels, in endometrial cancer cell lines ( Figure S1). Nineteen of these genes are implicated in the regulation of sister-chromatid cohesion, based on their sequence homology to cohesion genes in S. cerevisiae ( Table 1). The 24 NEECs consisted of 17 serous ECs and 7 clear cell ECs; five of the serous tumors (T33, T45, T65, T69, T70) were recently subjected to whole exome sequencing [52]. We included only MSI-stable tumors in the discovery screen; the MSI data have been reported elsewhere [52].
We obtained high quality sequence data for 87.6% (5.64 Mb) of bases (6.44 Mb) targeted. After excluding variants that were annotated as single nucleotide polymorphisms (SNPs) within dbSNP (Build 129), there were 109 unique nucleotide variants that represented potential somatic mutations. To determine whether these variants were somatic mutations or germline variants, we reamplified and sequenced the variant positions from the appropriate tumor DNA and matched normal DNA. Three variants were bone fide somatic mutations, present in the tumor DNA but absent from the matched normal DNA. The somatically mutated genes were ESCO1 (establishment of cohesion 1 homolog 1 (S. cerevisiae)), CHTF18 (chromosome transmission fidelity factor 18 homolog (S. cerevisiae)), and MRE11A (meiotic recombination 11 homolog A (S. cerevisiae)); each gene was mutated in 4% (1 of 24) of NEECs in the discovery screen. Although we found no evidence for somatic mutations in the remaining 18 candidate CIN genes, it is important to acknowledge that our discovery screen has insufficient power to detect all somatic mutations present in NEECs. We estimate that in a screen of 24 NEECs, the power to detect genes that are somatically mutated in 5%, 10% or 15% of all NEECs is 71%, 92%, and 98% respectively.
We next sought to more precisely determine the frequency and spectrum of somatic mutations in ESCO1, CHTF18, and MRE11A in endometrial cancer. To do this, we performed a prevalence screen in which we resequenced the coding exons and splice sites of the three genes from an additional 28 serous tumors, 13 clear cell tumors, and 42 endometrioid tumors, unselected for MSI status.
In the combined discovery and prevalence screens, we uncovered nonsynonymous somatic mutations within ESCO1, CHTF18, and MRE11A in, respectively, 3.7% (4 of 107), 1.9% (2 of 107), and 1.9% (2 of 107) of endometrial tumors ( Table 2 and Figure S2). Overall, 7.7% (5 of 65) of NEECs and 2.4% (1 of 42) of EECs had somatic mutations in one or more of the three genes. Compared to known consensus cancer genes with established roles in endometrial cancer, and to significantly mutated cancer genes, ESCO1, CHTF18, and MRE11A were infrequently mutated ( Figure S3, Figure S4, Figure S5) [44], [52], [53], [54], suggesting that these three genes are either rare pathogenic driver genes for endometrial cancer or that they are non-pathogenic genes that have acquired passenger mutations. Immunoblotting confirmed the expression of MRE11A and CHTF18 in panel of endometrial cancer cell lines ( Figure S6); ESCO1 was variably expressed among these same cell lines.
ESCO1, which encodes a lysine acetyltransferase that is essential for the establishment of sister chromatid cohesion in mammalian cells, was somatically mutated in 2.2% (1 of 45) of serous ECs, 10% (2 of 20) of clear cell ECs, and 2.4% (1 of 42) of endometrioid ECs. Two of the ESCO1 mutations are predicted to impact protein function. The ESCO1 R786C missense mutant, within the acetyltransferase domain, is predicted to impact protein function by both the SIFT and Polyphen algorithms ( Table 2). We speculate that the ESCO1 E338X nonsense mutant, which we uncovered in a serous-EC, may be a loss-of function mutant since a protein produced by this allele would be prematurely truncated and fail to include the acetyltransferase domain. Alternatively, nonsensemediated decay of the ESCO1 E338X transcript might lead to haploinsufficiency.
CHTF18 was somatically mutated in 2.2% (1 of 45) of serous ECs and 2.4% (1 of 42) of endometrioid ECs. In human cells, the CHTF18-RFC complex regulates the acetylation of the SMC3 cohesion-subunit by ESCO1 and ESCO2 acetyltransferases [34], thereby contributing to the establishment of sister chromatid cohesion. The CHTF18-RFC complex has also been implicated in the stimulation of DNA polymerase g activity, and in the recruitment of DNA polymerase e to sites of gap-filling repair synthesis [55], [56]. Both of the CHTF18 mutants we uncovered in endometrial cancer localize to the carboxy-terminus of the protein (Figure 1), within a region (residues 576-876) that mediates binding to RFC2-5 [57]. The CHTF18 R854W mutant is predicted to possibly affect protein function by the Mutation Assessor and SIFT algorithms ( Table 2). Interestingly, the majority of CHTF18 mutations observed in other cancers also localize to the C-terminus of the encoded protein [58]. These observations raise the possibility that somatic missense mutations in the C-terminus of CHTF18, found here and in other cancers, might disrupt the CHTF18-RFC interaction.
MRE11A was somatically mutated in 4.4% (2 of 45) of serous ECs. No MRE11A mutations were observed among clear cell or endometrioid tumors. MRE11A possesses both endonuclease activity and 39-59 exonuclease activity and, as a component the MRE11A-RAD50-NBS1 (MRN) complex, it plays an essential role in the cellular response to double strand breaks (reviewed in [59]). In mammalian cells, the MRN complex is also required for ATR-mediated phosphorylation of the SMC1 subunit of cohesin [60], and siRNA depletion of MRE11A in human cells results in cohesion defects [37]. The MRE11A D131N somatic mutant, which we uncovered in a serous EC, occurs at a highly evolutionarily conserved residue in the third phosphoesterase motif within the nuclease domain [61] and is predicted to impact protein function ( Figure 1, and Table 2). The MRE11A D692Y mutant, in the DNA binding domain, is also predicted to be functionally significant ( Table 2). Although intronic somatic mutations in MRE11A have been reported in microsatellite unstable endometrial cancers [62], [63], [64], to our knowledge, the present study is the first report of somatic mutations of MRE11A in microsatellite stable endometrial tumors ( Table 2). Of note, the MRE11A D131N variant, which was somatic in our study, has also been observed as a rare population variant (TMP_ESP_11_94212851) in the NHLBI Exome Sequencing Project (URL: http://evs.gs. washington.edu/EVS/), with a minor allele frequency of 0.0233% in the EuropeanAmerican population.
The mutual exclusivity or co-occurrence of somatic mutations in two or more genes can indicate functional redundancy or functional synergy, respectively. To determine the pattern of somatic mutations within cohesion genes in endometrial cancer, we combined the results of the present study with our previous analysis of the ATAD5 (hELG1) gene in this same cohort of ECs [44]. Although the number of mutated cases is small, we observed that somatic mutations in ESCO1 and ATAD5 tended to co-occur in endometrial cancer (P = 0.0102, two-tailed Fisher's exact test), as did somatic mutations in ESCO1 and CHTF18 (P = 0.0011) ( Figure 2, and Table 3). These observations raise the possibility that there might be functional synergy between ESCO1 and ATAD5 mutants, and between ESCO1 and CHTF18 mutants, in endometrial cancer. In this regard, it is noteworthy that somatic mutations in ESCO1 and ATAD5 tend to also co-occur in colorectal tumors (P = 0.000001) ( Figure S7), based on an analysis of the publically available mutation data generated by The Cancer Genome Atlas [http://cbio.mskcc.org/ cancergenomics/]. An alternative, but not mutually exclusive, possibility is that the co-occurring mutations of cohesion genes in endometrial cancer may reflect an underlying hypermutable phenotype. We previously evaluated the cohort of 107 tumors in this study for microsatellite instability and MSH6 mutations [44], [52], both of which can give rise to hypermutability due to defective mismatch repair (MMR). Although three of the tumors with cohesion gene mutations in this study were either MSIunstable or MSH6-mutated (Figure 2), we observed no statistically significant association between mutations in sister chromatid cohesion genes and defects in mismatch repair (Table S4 and  Table S5).
In summary, we have identified rare, nonsynonymous, somatic mutations within ESCO1, CHTF18, and MRE11A in a subset of primary endometrial tumors. Future studies will be required to determine whether these mutations are driver events that contribute to the pathogenesis of endometrial cancer.  Figure S3 Oncoprints displaying the distribution of somatic mutations in serous endometrial tumors as reported in this study (*) and elsewhere [44], [52], [53], [54]. Each blue bar represents an individual tumor (T). Nonsynonymous somatic mutations and MSI+ are indicated by the red bars. For MSH6, germline variants of unknown functional significance are displayed by orange bars. The observed frequency (%) of mutated cases, for each gene, is shown on the right. (TIF) Figure S4 Oncoprints displaying the distribution of somatic mutations in clear cell endometrial tumors as reported in this study (*) and elsewhere [44], [52], [53], [54]. Each blue bar represents an individual tumor (T). Nonsynonymous somatic mutations and MSI+ are indicated by the red bars. For MSH6, a germline variant of unknown functional significance is displayed by the orange bar. The observed frequency (%) of mutated cases, for each gene, is shown on the right. (TIF) Figure S5 Oncoprints displaying the distribution of somatic mutations in endometrioid endometrial tumors as reported in this study (*) and elsewhere [44], [52], [53], [54]. Each blue bar represents an individual tumor (T). Nonsynonymous somatic mutations and MSI+ are indicated by the red bars. For MSH6, germline variants of unknown functional significance are displayed by orange bars. The observed frequency (%) of mutated cases, for each gene, is shown on the right.