Multigene testing of moderate-risk genes: be mindful of the missense.

BACKGROUND: Moderate-risk genes have not been extensively studied, and missense substitutions in them are generally returned to patients as variants of uncertain significance lacking clearly defined risk estimates. The fraction of early-onset breast cancer cases carrying moderate-risk genotypes and quantitative methods for flagging variants for further analysis have not been established. METHODS: We evaluated rare missense substitutions identified from a mutation screen of ATM, CHEK2, MRE11A, RAD50, NBN, RAD51, RINT1, XRCC2 and BARD1 in 1297 cases of early-onset breast cancer and 1121 controls via scores from Align-Grantham Variation Grantham Deviation (GVGD), combined annotation dependent depletion (CADD), multivariate analysis of protein polymorphism (MAPP) and PolyPhen-2. We also evaluated subjects by polygenotype from 18 breast cancer risk SNPs. From these analyses, we estimated the fraction of cases and controls that reach a breast cancer OR≥2.5 threshold. RESULTS: Analysis of mutation screening data from the nine genes revealed that 7.5% of cases and 2.4% of controls were carriers of at least one rare variant with an average OR≥2.5. 2.1% of cases and 1.2% of controls had a polygenotype with an average OR≥2.5. CONCLUSIONS: Among early-onset breast cancer cases, 9.6% had a genotype associated with an increased risk sufficient to affect clinical management recommendations. Over two-thirds of variants conferring this level of risk were rare missense substitutions in moderate-risk genes. Placement in the estimated OR≥2.5 group by at least two of these missense analysis programs should be used to prioritise variants for further study. Panel testing often creates more heat than light; quantitative approaches to variant prioritisation and classification may facilitate more efficient clinical classification of variants.

Rare mutations in XRCC2 increase the risk of breast cancer.

An exome-sequencing study of families with multiple breast-cancer-affected individuals identified two families with XRCC2 mutations, one with a protein-truncating mutation and one with a probably deleterious missense mutation. We performed a population-based case-control mutation-screening study that identified six probably pathogenic coding variants in 1,308 cases with early-onset breast cancer and no variants in 1,120 controls (the severity grading was p < 0.02). We also performed additional mutation screening in 689 multiple-case families. We identified ten breast-cancer-affected families with protein-truncating or probably deleterious rare missense variants in XRCC2. Our identification of XRCC2 as a breast cancer susceptibility gene thus increases the proportion of breast cancers that are associated with homologous recombination-DNA-repair dysfunction and Fanconi anemia and could therefore benefit from specific targeted treatments such as PARP (poly ADP ribose polymerase) inhibitors. This study demonstrates the power of massively parallel sequencing for discovering susceptibility genes for common, complex diseases.

Differential allelic expression in leukoblast from patients with acute myeloid leukemia suggests genetic regulation of CDA, DCK, NT5C2, NT5C3, and TP53.

mRNA expression levels of certain genes have shown predictive value for the outcome of cytarabine-treated AML-patients. We hypothesized that genetic variants play a role in the regulation of the transcription of these genes. We studied leukoblasts from 82 patients with acute myeloid leukemia and observed various extent and frequency of differential allelic expression in the CDA, DCK, NT5C2, NT5C3, and TP53 genes. Our attempts to identify the causative regulatory single nucleotide polymorphisms by a bioinformatics approach did not succeed. However, our results indicate that genetic variations are at least in part responsible for the differences in overall expression levels of these genes.

A Laboratory Information Management System (LIMS) for a high throughput genetic platform aimed at candidate gene mutation screening.

UNLABELLED: High throughput mutation screening in an automated environment generates large data sets that have to be organized and stored reliably. Complex multistep workflows require strict process management and careful data tracking. We have developed a Laboratory Information Management Systems (LIMS) tailored to high throughput candidate gene mutation scanning and resequencing that respects these requirements. Designed with a client/server architecture, our system is platform independent and based on open-source tools from the database to the web application development strategy. Flexible, expandable and secure, the LIMS, by communicating with most of the laboratory instruments and robots, tracks samples and laboratory information, capturing data at every step of our automated mutation screening workflow. An important feature of our LIMS is that it enables tracking of information through a laboratory workflow where the process at one step is contingent on results from a previous step. AVAILABILITY: Script for MySQL database table creation and source code of the whole JSP application are freely available on our website: http://www-gcs.iarc.fr/lims/. SUPPLEMENTARY INFORMATION: System server configuration, database structure and additional details on the LIMS and the mutation screening workflow are available on our website: http://www-gcs.iarc.fr/lims/

Comprehensive statistical study of 452 BRCA1 missense substitutions with classification of eight recurrent substitutions as neutral.

BACKGROUND: Genetic testing for hereditary cancer syndromes contributes to the medical management of patients who may be at increased risk of one or more cancers. BRCA1 and BRCA2 testing for hereditary breast and ovarian cancer is one such widely used test. However, clinical testing methods with high sensitivity for deleterious mutations in these genes also detect many unclassified variants, primarily missense substitutions. METHODS: We developed an extension of the Grantham difference, called A-GVGD, to score missense substitutions against the range of variation present at their position in a multiple sequence alignment. Combining two methods, co-occurrence of unclassified variants with clearly deleterious mutations and A-GVGD, we analysed most of the missense substitutions observed in BRCA1. RESULTS: A-GVGD was able to resolve known neutral and deleterious missense substitutions into distinct sets. Additionally, eight previously unclassified BRCA1 missense substitutions observed in trans with one or more deleterious mutations, and within the cross-species range of variation observed at their position in the protein, are now classified as neutral. DISCUSSION: The methods combined here can classify as neutral about 50% of missense substitutions that have been observed with two or more clearly deleterious mutations. Furthermore, odds ratios estimated for sets of substitutions grouped by A-GVGD scores are consistent with the hypothesis that most unclassified substitutions that are within the cross-species range of variation at their position in BRCA1 are also neutral. For most of these, clinical reclassification will require integrated application of other methods such as pooled family histories, segregation analysis, or validated functional assay.

Genetic, functional, and histopathological evaluation of two C-terminal BRCA1 missense variants.

BACKGROUND: The vast majority of BRCA1 missense sequence variants remain uncharacterized for their possible effect on protein expression and function, and therefore are unclassified in terms of their pathogenicity. BRCA1 plays diverse cellular roles and it is unlikely that any single functional assay will accurately reflect the total cellular implications of missense mutations in this gene. OBJECTIVE: To elucidate the effect of two BRCA1 variants, 5236G>C (G1706A) and 5242C>A (A1708E) on BRCA1 function, and to survey the relative usefulness of several assays to direct the characterisation of other unclassified variants in BRCA genes. METHODS AND RESULTS: Data from a range of bioinformatic, genetic, and histopathological analyses, and in vitro functional assays indicated that the 1708E variant was associated with the disruption of different cellular functions of BRCA1. In transient transfection experiments in T47D and 293T cells, the 1708E product was mislocalised to the cytoplasm and induced centrosome amplification in 293T cells. The 1708E variant also failed to transactivate transcription of reporter constructs in mammalian transcriptional transactivation assays. In contrast, the 1706A variant displayed a phenotype comparable to wildtype BRCA1 in these assays. Consistent with functional data, tumours from 1708E carriers showed typical BRCA1 pathology, while tumour material from 1706A carriers displayed few histopathological features associated with BRCA1 related tumours. CONCLUSIONS: A comprehensive range of genetic, bioinformatic, and functional analyses have been combined for the characterisation of BRCA1 unclassified sequence variants. Consistent with the functional analyses, the combined odds of causality calculated for the 1706A variant after multifactorial likelihood analysis (1:142) indicates a definitive classification of this variant as "benign". In contrast, functional assays of the 1708E variant indicate that it is pathogenic, possibly through subcellular mislocalisation. However, the combined odds of 262:1 in favour of causality of this variant does not meet the minimal ratio of 1000:1 for classification as pathogenic, and A1708E remains formally designated as unclassified. Our findings highlight the importance of comprehensive genetic information, together with detailed functional analysis for the definitive categorisation of unclassified sequence variants. This combination of analyses may have direct application to the characterisation of other unclassified variants in BRCA1 and BRCA2.

Does nonsense-mediated mRNA decay explain the ovarian cancer cluster region of the BRCA2 gene?

BRCA2 (BReast CAncer susceptibility gene 2) germline mutation carriers are at increased risk for breast and ovarian cancers. Mutations occurring in the ovarian cancer cluster region (OCCR) are linked to higher ovarian cancer and/or lower breast cancer risk(s) than mutations occurring elsewhere in BRCA2. Most BRCA2 germline mutations introduce premature termination codons (PTCs), making their mRNAs likely targets of nonsense-mediated mRNA decay (NMD), a mechanism that eliminates PTC-bearing transcripts to prevent expression of truncated proteins. Contradictory evidence exists regarding whether NMD can be triggered by PTCs located far upstream of the nearest exon-exon junction (EEJ). Since the OCCR comprises a major portion of the 4.9 kb exon 11 of BRCA2, we investigated if transcripts bearing PTCs in this large exon are unable to trigger NMD, and if this might contribute to the phenotypic difference associated with the OCCR. We examined cDNA from 18 carriers of PTC-introducing germline mutations located throughout BRCA2, and found that PTC-bearing transcripts were 1.4-3.3-fold less prevalent than their nonmutated counterparts irregardless of PTC position. We conclude that NMD can recognize PTCs up to 4.5 kb upstream of the nearest EEJ, demonstrating that a general inability of NMD to recognize PTCs in exon 11 is unlikely to explain the genotype-phenotype correlation associated with the OCCR.

Classification of BRCA1 missense variants of unknown clinical significance.

BACKGROUND: BRCA1 is a tumour suppressor with pleiotropic actions. Germline mutations in BRCA1 are responsible for a large proportion of breast-ovarian cancer families. Several missense variants have been identified throughout the gene but because of lack of information about their impact on the function of BRCA1, predictive testing is not always informative. Classification of missense variants into deleterious/high risk or neutral/low clinical significance is essential to identify individuals at risk. OBJECTIVE: To investigate a panel of missense variants. METHODS AND RESULTS: The panel was investigated in a comprehensive framework that included (1) a functional assay based on transcription activation; (2) segregation analysis and a method of using incomplete pedigree data to calculate the odds of causality; (3) a method based on interspecific sequence variation. It was shown that the transcriptional activation assay could be used as a test to characterise mutations in the carboxy-terminus region of BRCA1 encompassing residues 1396-1863. Thirteen missense variants (H1402Y, L1407P, H1421Y, S1512I, M1628T, M1628V, T1685I, G1706A, T1720A, A1752P, G1788V, V1809F, and W1837R) were specifically investigated. CONCLUSIONS: While individual classification schemes for BRCA1 alleles still present limitations, a combination of several methods provides a more powerful way of identifying variants that are causally linked to a high risk of breast and ovarian cancer. The framework presented here brings these variants nearer to clinical applicability.

Analysis of missense variation in human BRCA1 in the context of interspecific sequence variation.

INTRODUCTION: Interpretation of results from mutation screening of tumour suppressor genes known to harbour high risk susceptibility mutations, such as APC, BRCA1, BRCA2, MLH1, MSH2, TP53, and PTEN, is becoming an increasingly important part of clinical practice. Interpretation of truncating mutations, gene rearrangements, and obvious splice junction mutations, is generally straightforward. However, classification of missense variants often presents a difficult problem. From a series of 20,000 full sequence tests of BRCA1 carried out at Myriad Genetic Laboratories, a total of 314 different missense changes and eight in-frame deletions were observed. Before this study, only 21 of these missense changes were classified as deleterious or suspected deleterious and 14 as neutral or of little clinical significance. METHODS: We have used a combination of a multiple sequence alignment of orthologous BRCA1 sequences and a measure of the chemical difference between the amino acids present at individual residues in the sequence alignment to classify missense variants and in-frame deletions detected during mutation screening of BRCA1. RESULTS: In the present analysis we were able to classify an additional 50 missense variants and two in-frame deletions as probably deleterious and 92 missense variants as probably neutral. Thus we have tentatively classified about 50% of the unclassified missense variants observed during clinical testing of BRCA1. DISCUSSION: An internal test of the analysis is consistent with our classification of the variants designated probably deleterious; however, we must stress that this classification is tentative and does not have sufficient independent confirmation to serve as a clinically applicable stand alone method.

Prostate cancer susceptibility genes: lessons learned and challenges posed.

In most developed countries, prostate cancer is the most frequently diagnosed malignancy in men. The extent to which the marked racial/ethnic difference in its incidence rate is attributable to screening methods, environmental, hormonal and/or genetic factors remains unknown. A positive family history is among the strongest epidemiological risk factors for prostate cancer. It is now well recognized that the role of candidate genetic markers to this multifactorial malignancy is more difficult to identify than the identification of other cancer susceptibility genes. Indeed, despite the localization of several susceptibility loci, there has been limited success in identifying high-risk susceptibility genes analogous to BRCA1 or BRCA2 for breast and ovarian cancer. Nonetheless, three strong candidate susceptibility genes have been described, namely ELAC2 (chromosome 17p11/HPC2 region), 2'-5'-oligoadenylate-dependent ribonuclease L (RNASEL), a gene in the HPC1 region, and Macrophage Scavenger Receptor 1 (MSR1), a gene within a region of linkage on chromosome 8p. Additional studies using larger cohorts are needed to fully evaluate the role of these susceptibility genes in prostate cancer risk. It is also of interest to mention that a significant percentage of men with early-onset prostate cancer harbor germline mutation in the BRCA2 gene thus confirming its role as a high-risk prostate cancer susceptibility gene. Although initial segregation analyses supported the hypothesis that a number of rare highly penetrant loci contribute to the Mendelian inheritance of prostate cancer, current experimental evidence better supports the hypothesis that some of the familial risks may be due to inheritance of multiple moderate-risk genetic variants. In this regard, it is not surprising that analyses of genes encoding key proteins involved in androgen biosynthesis and action led to the observation of a significant association between a susceptibility to prostate cancer and common genetic variants in some of those genes.

Interpreting epidemiological research: blinded comparison of methods used to estimate the prevalence of inherited mutations in BRCA1.

While sequence analysis is considered by many to be the most sensitive method of detecting unknown mutations in large genes such as BRCA1, most published estimates of the prevalence of mutations in this gene have been derived from studies that have used other methods of gene analysis. In order to determine the relative sensitivity of techniques that are widely used in research on BRCA1, a set of blinded samples containing 58 distinct mutations were analysed by four separate laboratories. Each used one of the following methods: single strand conformational polymorphism analysis (SSCP), conformation sensitive gel electrophoresis (CSGE), two dimensional gene scanning (TDGS), and denaturing high performance liquid chromatography (DHPLC). Only the laboratory using DHPLC correctly identified each of the mutations. The laboratory using TDGS correctly identified 91% of the mutations but produced three apparent false positive results. The laboratories using SSCP and CSGE detected abnormal migration for 72% and 76% of the mutations, respectively, but subsequently confirmed and reported only 65% and 60% of mutations, respectively. False negatives therefore resulted not only from failure of the techniques to distinguish wild type from mutant, but also from failure to confirm the mutation by sequence analysis as well as from human errors leading to misreporting of results. These findings characterise sources of error in commonly used methods of mutation detection that should be addressed by laboratories using these methods. Based upon sources of error identified in this comparison, it is likely that mutations in BRCA1 and BRCA2 are more prevalent than some studies have previously reported. The findings of this comparison provide a basis for interpreting studies of mutations in susceptibility genes across many inherited cancer syndromes.

Mutation analyses of 268 candidate genes in human tumor cell lines.

We have performed a homozygous deletion screen on 268 candidate genes in 90 human tumor cell lines derived from multiple types of cancers. Most of the candidate genes investigated have been proposed to be involved in cellular processes that are germane to cancer progression, such as cell cycle control, genome maintenance, chromatin remodeling, cell adhesion, and apoptosis. We have detected novel homozygous deletions affecting four independent loci: Brahma-related gene (SMARCA4) on chromosome 19p in the TSU-Pr1 prostate and A427 lung carcinoma lines, Map Kinase Kinase 3 (MAP2K3) on 17q in the NCI-H774 lung tumor cell line, TMPRSS2 on 21q in the Bx PC-3 pancreatic carcinoma line, and Cadherin 6 (CDH6) on 5p in the SK-LU-1 lung carcinoma line. Subsequent analyses of the coding sequences of these four genes using cDNAs from a panel of tumor cell lines revealed multiple sequence variants. The results of this mutation study serve to demonstrate the feasibility of performing high-throughput screens of candidate genes in tumor cell lines to identify genes that may be targeted for mutation during the development of cancer.

HPC2 variants and screen-detected prostate cancer.

Two studies have reported significant associations between susceptibility to prostate cancer and two common missense variants of the HPC2/ELAC2 gene, with estimated relative risks in the range of two- to threefold. We investigated whether these polymorphisms could be informative in the prediction of the presence of prostate cancer in men undergoing prostatic biopsy for the evaluation of an elevated serum-PSA level (> or = 4.0 ng/ml). We genotyped 944 men who underwent a prostate biopsy at our institution, as well as a control population of 922 healthy, unselected women from the same population. The prevalence of the HPC2 Ala541Thr allele was similar in men with prostate cancer (6.3%), men with other prostatic conditions (6.8%), and healthy women (6.3%) (P = .83). We conclude that HPC2 genotyping is unlikely to be a useful adjunct to PSA in the prediction of the presence of biopsy-detected prostate cancer in asymptomatic men.

A candidate prostate cancer susceptibility gene at chromosome 17p.

It is difficult to identify genes that predispose to prostate cancer due to late age at diagnosis, presence of phenocopies within high-risk pedigrees and genetic complexity. A genome-wide scan of large, high-risk pedigrees from Utah has provided evidence for linkage to a locus on chromosome 17p. We carried out positional cloning and mutation screening within the refined interval, identifying a gene, ELAC2, harboring mutations (including a frameshift and a nonconservative missense change) that segregate with prostate cancer in two pedigrees. In addition, two common missense variants in the gene are associated with the occurrence of prostate cancer. ELAC2 is a member of an uncharacterized gene family predicted to encode a metal-dependent hydrolase domain that is conserved among eukaryotes, archaebacteria and eubacteria. The gene product bears amino acid sequence similarity to two better understood protein families, namely the PSO2 (SNM1) DNA interstrand crosslink repair proteins and the 73-kD subunit of mRNA 3' end cleavage and polyadenylation specificity factor (CPSF73).

Microdissection, DOP-PCR, and comparative genomic hybridization of paraffin-embedded familial prostate cancers.

There is a clear genetic component to prostate cancer susceptibility. Regions reported to be linked to prostate cancer include 1q24-25 (HPC-1), 1q42.2-43, and Xq27-28. There is limited genetic information on familial prostate tumors. We used the Utah Population Database to identify familial prostate cancer cases and selected 35 cases from high-risk families. Tissue blocks containing discernable tumor were available from 19 cases; 13 of these yielded adequate specimens for analysis. Six cases came from families with linkage to HPC-1, 3 were known to have linkage to Xq27-28, and 4 had no linkage to a known locus; 7 cases were analyzed from patients who showed no known linkage (sporadic tumors) as controls. These paraffin-embedded tumors were laser microdissected, degenerate oligonucleotide (DOP)-amplified, and labeled for fluorescence detection by comparative genomic hybridization (CGH). Loss of 7q, 10q, and 16q and gain of 8q were common abnormalities present in both familial and sporadic tumors. Distinctive abnormalities included loss of 3p12-3p22 in 3 of 6 HPC-7-linked cases and in 2 of 3 X-linked cases and gain of 6q11-6q21 in 2 each of HPC-1 and X-linked tumors. In conclusion, laser microdissection, DOP-PCR, and CGH is a feasible method for analysis of paraffin-embedded prostate tumors. This study provides preliminary data suggesting that familial prostate cancer harbors some unique genetic changes when compared with sporadic prostate tumors.

BRG1, a component of the SWI-SNF complex, is mutated in multiple human tumor cell lines.

Human BRG1 is a component of the evolutionarily conserved SWI-SNF chromatin remodeling complex. BRG1 has been implicated in growth control through its interaction with the tumor suppressor pRb and may consequently serve as a negative regulator of proliferation. Postulating that BRG1 may itself be a tumor suppressor gene, we screened a panel of tumor cell lines to determine whether the gene is targeted for mutation. We report that the COOH-terminal region of BRG1 is homozygously deleted in two carcinoma cell lines, prostate TSU-Pr1 and lung A-427. In addition, biallelic inactivations of BRG1 were observed in four other cell lines derived from carcinomas of the breast, lung, pancreas, and prostate; their mutations in BRG1 included three frameshift lesions and one nonsense lesion. Point mutations were also discovered in a number of other cell lines, however in most cases any effect of these mutations on BRG1 function remains to be established. A variety of different mutations within BRG1, in several cell lines, suggest that BRG1 may be targeted for disruption in human tumors. Significantly, reintroduction of BRG1 into cells lacking BRG1 expression was sufficient to reverse their transformed phenotype inducing growth arrest and a flattened morphology. These data strongly support the model that BRG1 may function as a tumor suppressor and strengthen the hypothesis that the regulation of gene expression through chromatin remodeling is critical for cancer progression. It will be important to confirm these observations in primary tumors.

Threonine phosphorylation of the MMAC1/PTEN PDZ binding domain both inhibits and stimulates PDZ binding.

Two-hybrid searches with the tumor suppressor MMAC1/PTEN isolated the proteins hDLG and hMAST205. Further two-hybrid analysis and microtiter plate binding assays localized the sites of interaction to PDZ domains from hDLG and hMAST205 and the PDZ binding domain at the COOH terminus of MMAC1/PTEN. A synthetic peptide derived from the MMAC1/PTEN PDZ binding domain (MMAC1/PTEN-PDZBD) was used to coprecipitate proteins from A431 human cell lysate. The recovered proteins were resolved by SDS-PAGE and immobilized on a nitrocellulose membrane. Treatment of this membrane with an anti-hDLG antibody identified a Mr 140,000 band, consistent with the size of hDLG. Treatment of this membrane with the MMAC1/PTEN-PDZBD peptide identified a single prominent band of slightly larger than Mr 200,000 (Mr 200,000 kDa). Threonine phosphorylation of the MMAC1/ PTEN-PDZBD peptide inhibited both microtiter plate binding to the hDLG and hMAST205 PDZ domains and coprecipitation of the Mr 140,000 and > 200,000 proteins, but promoted coprecipitation of proteins of approximately Mr 90,000 and Mr 120,000 from A431 cell lysate. This result suggests phosphorylation of the MMAC1/PTEN PDZ binding domain can both inhibit and promote PDZ interactions.

Prostate cancer susceptibility locus HPC1 in Utah high-risk pedigrees.

A prostate cancer susceptibility locus ( HPC1 ) at 1q24-25 has been identified. Subsequent analysis showed that the majority of the evidence for localization was provided by families with relatively young (<65 years) average age at diagnosis. We examined evidence for linkage to this region in a set of 41 extended multi-case prostate cancer pedigrees containing 440 prostate cancer cases. Genotyping of five short tandem repeat markers in the region was performed on DNA from 1724 individuals, including 284 prostate cancer cases. In comparison with the families reported in the initial localization, the Utah pedigrees are generally much larger (average of 10.7 versus 5.1 cases) and have an older average age at diagnosis (69 versus 65 years). Two- and three-point linkage analyses were conducted using a previously reported model and provided replication for HPC1 (two-point: LOD = 1.73, P = 0.005 at D1S196; three-point: LOD = 2.06, P = 0.002 for the interval D1S196-D1S416 ). The youngest quartile (by median age at diagnosis) yielded a maximum LOD of 2.82, P = 0. 0003 (at D1S215-D1S222 ), compared with a maximum LOD of 0.73, P = 0. 07 for the oldest quartile pedigrees at the same locus. Further analysis with an age-dependent model, specifying higher sporadic rates for older cases, suggests that the linkage evidence may be lower than expected given the power of the resource due to a high sporadic rate in the large Utah pedigrees.

Genomic structure, chromosomal location, and mutation analysis of the human CDC14A gene.

Human CDC14A is a dual-specificity phosphatase that shares sequence similarity with the recently identified tumor suppressor, MMAC1/PTEN/TEP1. By radiation hybrid mapping, we localized CDC14A to chromosome band 1p21, a region that has been shown to exhibit loss of heterozygosity in highly differentiated breast carcinoma and malignant mesothelioma. We have mapped the exon-intron structure of CDC14A gene and found an in-frame ATG at 14 codons upstream of the previously reported start site (GenBank Accession No. AF000367). In screening a panel of 136 cDNAs from tumor cell lines for coding mutations, we have identified a 48-bp in-frame deletion in the cDNA of the breast carcinoma cell line, MDA-MB-436. This deletion is the result of an acceptor splice site mutation (AG to AT) in intron 12 that causes the skipping of exon 13 in the gene. Loss of expression of the wildtype allele in the same breast cell line supports the possibility that CDC14A may be a tumor suppressor gene that is targeted for inactivation during tumorigenesis.