Somatic Mutations and Altered Expression of the Candidate Tumor Suppressors CSNK1⑀, DLG1, and EDD/Hhyd in Mammary Ductal Carcinoma

[CANCER RESEARCH 64, 942–951, February 1, 2004] Somatic Mutations and Altered Expression of the Candidate Tumor Suppressors CSNK1⑀, DLG1, and EDD/hHYD in Mammary Ductal Carcinoma

Tannin J. Fuja,1 Fritz Lin,3 Kathryn E. Osann,2 and Peter J. Bryant1 1Developmental Biology Center and 2Department of Medicine, University of California–Irvine, Irvine, California, and 3Department of Pathology, University of California Irvine Medical Center, Orange, California

ABSTRACT invasive ductal carcinoma, which develops from ductal carcinoma in situ (DCIS) and spreads through the duct walls to invade the breast ⑀ We report somatic mutations in three genes (CSNK1 , encoding the tissue (8). Although the morphological manifestations are well char- Ser/Thr kinase casein kinase I ⑀; DLG1, encoding a membrane-associated acterized, the molecular events involved in this progress have not been putative scaffolding protein; and EDD/hHYD, encoding a progestin induced putative ubiquitin-protein ligase) in mammary ductal carcinoma. identified (9). These genes were suspected of playing a role in cancer because loss-of- Drosophila has been used as a model organism to identify genes function mutations in their Drosophila homologues cause excess tissue involved the regulation of cell proliferation and tumorigenesis (10). growth. Using DNA from 82 laser-microdissected tumor samples, followed Genetic studies have led to the identification of tumor suppressor by microsatellite analysis, denaturing HPLC and direct sequencing, we genes (TSGs) in which recessive mutations lead to neoplastic or found multiple somatic point mutations in all three genes, and these hyperplastic overgrowth of larval imaginal discs (specialized epithe- mutations showed significant association with loss of heterozygosity of lial cells) and other tissues, mainly in the larva (11, 12). Using tissue ⑀ closely linked polymorphic microsatellite markers. For CSNK1 and microarrays, we confirmed that the human homologues of three of DLG1, most of the mutations affected highly conserved residues, some ⑀ were found repetitively in different patients, and no synonymous muta- these TSGs (CSNK1 , DLG1, and EDD/hHYD) are expressed in tions were found, indicating that the observed mutations were selected in normal human breast tissue. Each of them has functional and sequence tumors and may be functionally significant. Immunohistochemical reac- characteristics, suggesting that it may play a role in regulating mam- tivity of each protein was reduced in poorly differentiated tumors, and mary cell growth: there was a positive association between altered protein reactivity, loss of CSNK1⑀. This gene encodes casein kinase 1-⑀ (CKI⑀), a serine- heterozygosity, and somatic mutations. There was a statistically signifi- threonine kinase whose overexpression mimics Wnt signaling by cant association of hDlg staining with p53 and Ki67 reactivity, whereas stabilizing ␤-catenin, thereby increasing expression of ␤-catenin- ⑀ CSK1 and EDD/hHYD staining levels were associated with progesterone dependent genes. CKI⑀ is homologous to the product of the Drosoph- receptor status. The results provide strong indications for a role of all ila TSG dco (discs overgrown) in which some allelic combinations three genes in mammary ductal carcinoma. They also justify additional ⑀ studies of the functional significance of the changes, as well as a search for result in overgrowth phenotypes (13, 14). CKI is found in the additional changes in these and other genes identified from studies on cytoplasm as a monomer and can phosphorylate a variety of proteins, model systems. including itself (15, 16). DLG1. DLG1 encodes hDlg, which is a member of the protein INTRODUCTION family named membrane-associated GUanylate kinase homologues (17, 18). It is a homologue of the Drosophila TSG dlg (discs-large)in Mammary carcinoma is one of the most common neoplasms in which loss-of-function mutations result in overgrowth of the imaginal women (1), and it is the second leading cause of cancer-related deaths discs and larval brain. Membrane-associated GUanylate kinase homo- in women of the Western world (2). Despite significant improvements logues are usually localized at the membrane cytoskeleton and contain in earlier detection and diagnosis of this cancer leading to more several domains, suggesting a role in intracellular signal transduction effective treatment modalities, approximately one-quarter of breast (19). Immunohistochemical localization experiments show that hDlg cancer patients still die of their disease (2). Understanding the mo- is localized along the basolateral surfaces of epithelial cells in the lecular pathology of mammary carcinoma is critical for the develop- intestine and at the plasma membrane at sites of cell-cell contact in ment of novel, molecular-based therapies (3, 4). Progress has been cultured colon and mammary epithelial cells (20). hDlg expression is made in identifying tumor suppressor genes affected in hereditary down-regulated in both cervical (21) and gastric (22) cancer, and loss forms of breast cancer such as BRCA1 and BRCA2 (5). However, of the protein at cell contact sites may be an important step in ϳ95% of breast cancers are sporadic rather than hereditary (6), and development of epithelial cancers (17, 21, 23, 24). The PDZ2 domain very little progress has been made in identifying the molecular lesions of hDlg binds TOPK/PBK, a novel mitotic kinase, and hDlg also responsible for these cases. Here, we show that at least some of the interacts with other tumor suppressors, notably APC (25) and PTEN somatic genetic changes occurring in sporadic breast cancer can be (26), both of which are involved in human cancers. The hDlg protein identified using a candidate gene approach. is highly enriched at the membrane of normal epithelial cells, but it Mammary carcinoma develops through defined clinical and path- shows a higher level of cytoplasmic accumulation in high-grade ological stages, starting with atypical epithelial hyperplasia, progress- cervical neoplasia (21). It is a primary target of the E6 oncoprotein ing to carcinoma in situ, then invasive carcinomas, and culminating in produced by high-risk strains of papillomavirus (27). metastatic disease (7). The most common type of breast cancer is EDD/hHYD/DD5. EDD encodes a progestin-induced protein be- longing to the HECT family (28). The HECT family proteins function Received 7/14/03; revised 11/19/03; accepted 11/21/03. Grant support: National Cancer Institute Research Grant CA91043. T. J. Fuja was as E3 ubiquitin-protein ligases, targeting specific proteins for ubiq- supported by a training grant from the National Institute of Child Health and Human uitin-mediated proteolysis (29, 30). EDD is a homologue of the Development. product of Drosophila hyd (hyperplastic discs) in which loss-of- The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with function mutations cause imaginal disk overgrowth (31). It interacts 18 U.S.C. Section 1734 solely to indicate this fact. with importin ␣-5 through basic nuclear localization signals and binds Note: Supplementary data for this article can be found at Cancer Research Online (http://cancerres.aacrjournals.org). to the progesterone receptor to potentiate progestin-mediated gene Requests for reprints: Peter J. Bryant. E-mail: [email protected]. activation (28). Furthermore, EDD interacts with the DNA topoi- 942

somerase II-␤-binding protein 1 through its BRCT [BRCA1 COOH ellite primers were used, and primers specific for the 5Ј-untranslated region of terminus domain (32)]. These domains are involved in cell cycle arrest human skeletal ␣-actin were used as internal amplification controls. The in response to activation of the DNA damage checkpoint. conditions for each reaction were optimized using gradient PCR. The optimal To explore the potential role of these three candidate TSGs in Ta for each microsatellite marker was as follows: (d22s1156 and d22s272— mammary ductal carcinoma, we obtained DNA from tumor samples 56.8°C), (d8s1834, d22s423, d8s1789, and d8s521—55.4°C) and (d3s1272 and d3s1265—52.7°C). The optimal annealing temperature of the skeletal ␣-skel- by laser capture microdissection (LCM). We tested for loss of het- etal actin control was 60°C. After 3 min of incubation at 94°C, amplification erozygosity (LOH) of polymorphic markers flanking each gene and was carried out by 35 cycles of: 94°C (30 s), Ta (30 s), and elongation 72°C PCR-amplified genomic fragments encoding critical regions of each (1 min). These cycles were followed by a final elongation step at 72°C for protein. These fragments were used for mutation detection using 7 min. denaturing high-performance liquid chromatography (dHPLC) and Compatibly labeled microsatellite fragments were pooled and scored with direct sequencing. For each gene, we also characterized expression the ABI PRISM 3100 Genetic Analyzer. Using the ABI Genescan and ABI and localization of the protein product in paraffin-embedded sections Genotyper software, cases were scored as informative for a polymorphic of mammary ductal carcinomas using immunoperoxidase and immu- marker if the DNA was effectively amplified and the normal DNA was nofluorescence histochemistry. The results provide strong indications heterozygous for that marker. LOH was characterized according to the equa- ϭ for a role of all three genes in mammary ductal carcinoma. tion Na/Nb (Ca/Cb)v, where Na and Nb were the peak height or peak area of the two predominant alleles within the specific microsatellite size range of the normal DNA, and Ca and Cb were the peak height or peak area of the Յ Ն MATERIALS AND METHODS corresponding alleles in the cancer DNA. When v 0.5 or 2, the case was scored as LOH for that marker (4, 39). PCR and fragment analysis were Tissue Samples. Eighty-two cases of formalin-fixed, paraffin-embedded, repeated three times for each case. When the sample did not show heterozy- archived mammary ductal carcinoma and case/patient-specific paired samples gosity in the normal tissue or did not show conservation of at least one allele of normal tissue were collected from the University of California–Irvine, in the cancer as compared with normal, it was scored as noninformative and Department of Pathology Core Facility, under the direction and approval of the not included in the LOH analysis. University of California Human Subjects Institutional Review Board. In cases dHPLC. Ion-pair reversed-phase dHPLC identifies mutations by detecting showing significant intratumor heterogeneity (e.g., regions of clear DCIS heteroduplex formation between mismatched nucleotides in PCR-amplified isolated from regions of invasive carcinomas), the two disease areas were DNA (40). Using the WAVE Nucleic Acid Fragment Analysis System (Trans- treated independently, so the total case number described in the results is 85. genomics, Inc., Omaha, NE), we analyzed selected regions of each candidate Samples were accompanied by pathology reports that had been stripped of any TSG for the presence of mutations (41, 42). Amplicons encoding the following patient identifiers or links. These reports were held and corroborated by Dr. functional domains were selected for dHPLC analysis: the kinase domain of ⑀ Fritz Lin, a certified pathologist and chief of surgical pathology at University CKI ; the PDZ2 domain of hDlg; and the NH2-terminal region of the EDD of California Irvine Medical Center. HECT domain. A section of the 5Ј-untranslated region of ACT1 was used as Twenty serial 5-␮m thick sections were taken from each tissue block. To a control. After identifying the melt domains of each fragment, we tested the confirm the maintenance of the diagnosis of each specimen throughout the 20 samples for heteroduplex formation at the two best temperatures. sections, standard Meyer’s H&E Y stains were performed at even intervals Direct Sequencing. DNA from LCM samples was sequenced using ABI throughout the entire series. These control H&E stains were reviewed by Dr. Big Dye Terminator Ready v.3.1 cycle sequencing kits (PE Applied Biosys- Lin to ensure consistency of diagnosis. The Scarff-Bloom-Richardson (SBR) tems) after Shrimp Alkaline Phosphatase/Exonuclease I 1-h digestion at 37°C, scoring system (33) was used to classify morphological grades of ductal followed by heat inactivation at 70°C for 15 min. (Promega Corp., Madison, carcinoma. WI). Bi-directional cycle sequencing was performed to confirm each sequence. LCM. Homogeneous regions of cancer or normal tissue from each case The mixture of 60 ng of PCR-amplified SAP/Exo I-digested DNA, 2.3 pmol of ϫ were obtained using the PixCell IIe Laser Capture Microdissection System forward or reverse primer, 1 buffer [1 M Tris (pH 9.0) and 25 mM MgCl2] (Acturus Engineering, Inc., Mountain View, CA; Ref. 34). Each region cap- supplemented with PCRx Enhancer (Invitrogen-Life Technologies, Inc.) was tured was confirmed as cancer or normal by pathologist review. Using recom- held at 95°C for 5 min, followed by 90 cycles of 95°C (30 s), 50°C (10 s), and mended settings (34), 2000 laser hits were taken from each case. To increase 60°C (4 min). The resulting material was precipitated in 75% isopropanol the chance of finding genetic variation, cells were captured from adjacent according to ABI recommendations. Samples were run on a 3730 ABI DNA locations within the same section as well as different regions of the tumor from Analyzer, and sequence data were evaluated using Chromas v1.45 (Technely- each patient, and the samples were mixed. sium). DNA Extraction. DNA was extracted in 50 ␮l of freshly made DNA Immunohistochemistry. To examine protein expression in situ, paraffin- digestion buffer [0.08% proteinase K, 15 mM Tris-HCl (pH 8.0), 1 mM EDTA, embedded sections were analyzed by indirect immunolabeling using both 0.5% Tween 20], which been preheated to 42°C and incubated overnight. The peroxidase and fluorescence-based histochemical methods (43). Two sections contents of the tube were collected by centrifugation and then heated to 95°C from each case were stained using the same immunoperoxidase protocol, and for 8 min. To remove eosin, the DNA was purified using a modified Qiagen these sections were processed by two different technicians to monitor for DNeasy protocol (Qiagen, Valencia, CA) with serial ethanol washes. variability resulting from staining procedure. A third section was stained for Genomic Mapping and PCR Amplification. We analyzed heterozygosity immunofluorescence to ensure staining patterns were not biased by the use of for the following polymorphic microsatellite markers flanking each locus on a single staining method. the Centromeric (C) and telomeric (T) sides: CSNK1⑀: d22s1156 (305 Kb C), Commercially available, affinity-purified goat polyclonal anti-CKI⑀ (C-20) d22s272 (399 Kb T), and d22s423 (1659 Kb T); DLG1: d3s1265 (1664 Kb C) raised against a COOH-terminal peptide of CKI⑀ (Santa Cruz Biotechnology, and d3s1272 (246 Kb T); and EDD/hHYD: d8s1789 (2589 Kb C), d8s521 (958 Santa Cruz, CA) was used at 10 ␮g/ml. Mouse IgG1 monoclonal anti-hDlg Kb C), and d8s1834 (448 Kb T). raised against amino acids 1–229 of hDlg was used at 7 ␮g/ml (Santa Cruz Gene-specific PCR primers were designed to amplify specific regions of Biotechnology). Affinity-purified goat polyclonal anti-EDD raised against a each gene. All genomic PCR amplification reactions were carried out using peptide mapping near the COOH terminus of EDD was used at 5 ␮g/ml (Santa high-fidelity Pfx DNA polymerase (0.625 units/reaction) in a final reaction mix Cruz Biotechnology). Data showing the specificity of these antibodies is ␮ ␮ of 300 M deoxynucleotide triphosphates, 1.5 mM MgSO4, 0.5 M each available from the company’s web site (www.scbt.com). For fluorescently primer, and 1ϫ Pfx amplification buffer (Invitrogen-Life Technologies, Inc., labeled sections, Alexa Fluor 488 and 568 goat antimouse IgG (HϩL) and Carlsbad, CA). PCR yield was imaged and quantified on a NucleoVision 760 Alexa Flour 488 and 568 donkey antigoat (Molecular Probes, Inc., Eugene, gel imager using GelExpert v3.5 software (NucleoTech, San Mateo, CA). OR) were used at 5 ␮g/ml. ToProIII (Molecular Probes, Inc.) was used to LOH. To test for LOH of the polymorphic microsatellite markers, the fluorescently label DNA. purified cancer DNA was compared with case-paired normal DNA extracted Immunofluorescent-labeled sections were viewed using an MRC 1024 by LCM (35–38). 5Ј-Fluorescently labeled (HEX/6-FAM) forward microsat- Bio-Rad/Nikon Diaphot 200 laser-scanning confocal microscope and analyzed 943

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2004 American Association for Cancer Research. TSG MUTATIONS IN MAMMARY DUCTAL CARCINOMA using LaserSharp image analysis software (Bio-Rad Microscience Division, mutations in CSNK1⑀ (Fig. 1A). The following missense mutations Cambridge, MA). Immunoperoxidase-labeled sections were observed using were identified: I29T (atc3aac); A36G (gcc3ggc); L39Q (ctg3cag; an Olympus BH-2 light microscope equipped with a Sony DXC-960-MD3 detected in 5 of the 9 cases sequenced); L49Q (ctg3cag; detected charge-coupled device (CCD) camera, and images from this system were three times); E52K (gag3aag); G64A (ggg3gcg); G72A (gga3gca); collected and analyzed using Pax-It Image software. A73V (gct3gtt; detected twice); N78T (aac3acc); and S101R Scoring of Protein Reactivity. For each case, the intensity of CKI⑀, hDlg, (agc3agg; detected twice). One mutation to a stop codon was found and EDD staining in mammary ductal tumors was compared with case- (E74*; cag3tag), affecting a residue homologous to the one mutated matched (from the same patient) normal ductal tissue. Each case was inde- 3 pendently scored three times in a single-blind manner to reduce scoring bias. in the dco Drosophila mutant, which was the only dco allele found to In cases where different regions of a section gave scores differing by greater give a robust overgrowth phenotype (13, 14). The A36G and E52K than two scoring units (according to the scale below), each region was reported mutations occur at residues known to participate in ATP binding. Four independently. Otherwise, the mean stain score was used. A comparative scale of the mutations detected resulted in a nonconservative amino acid from 1 to 5 was used to score staining intensity: 1, much less staining in the substitution: I29T; L39Q; L49Q; and S101R. Ten of the 11 residues cancer tissue than in the normal tissue; 2, moderately less staining in the cancer altered in these cases are conserved from Drosophila to human. tissue than in the normal tissue; 3, same intensity of stain in the cancer as in DLG1. We found 13 nonsynonymous mutations and no synony- the normal; 4, moderately more staining in the cancer than in the normal; and mous mutations in DLG1 (Fig. 1B). The following missense mutations 5, much more staining in the cancer tissue than in the normal tissue. were identified: G338R (gga3cga; found twice); I348V (atc3gtc; This relative score reflects the difference between normal tissue and tumor 3 3 tissue. A mean score of 3.0 indicates no difference between normal and tumor found three times); Y349S (tat tct); K352R (aaa aga); K352E 3 3 3 tissue overall, whereas a mean score that exceeds 3.0 by two or more SEs (aaa gaa); G356A (gga gaa); H360Q (cac caa; found twice); and indicates significant overexpression in tumor compared with normal tissue. H360N (cac3aac). Six of the seven residues altered in these cases are Tissue sections were also stained for the following standard clinical indi- conserved from Drosophila to human. There was a single nonsense cators: estrogen receptor (ER), progesterone receptor (PR), p53, Ki67, and mutation Q340* (cag3tag). ErbB2 (HER-2/Neu). These were detected using only the immunoperoxidase EDD/hHYD. This was the only gene tested that showed silent labeling method. These indicators were scored positive or negative using (synonymous) mutations. We found four synonymous and six non- standard clinical pathology guidelines and parameters. synonymous mutations in this gene (Fig. 1C). The following missense Statistical Analysis. Data were analyzed using standard methods appro- mutations were detected: A2411G (gca3gga); M2419V (atg3gtg; priate for continuous and categorical data, including t tests, ANOVA, and ␹2 detected twice); E2422K (gaa3aaa); Q2427E (cag3gag); and tests. The average staining score for each of the three TSGs was treated as a 3 continuous normally distributed variable. Differences between invasive cancer A2433G (gca gaa). All of the missense mutations predicted con- samples and DCIS were evaluated using t tests and ANOVA. For comparisons servative amino acid substitutions. Three of the five residues altered between three or more groups (ANOVA), pairwise t tests were conducted with in these cases are conserved from Drosophila to human. Bonferroni adjustment for multiple comparisons. Associations between level A 185-bp region of the 5Ј-untranslated region of the human skeletal of protein staining for TSGs and standard clinical indicators, including ER, PR, ␣-actin gene was amplified as an internal control for dHPLC and p53, Ki67, and ErbB2 (coded as positive or negative), were tested by t tests. Associations with SBR score were tested using ANOVA. The relationship between standard clinical indicators and TSGs was also investigated using methods appropriate for categorical data. The Cochran ␹2 test for linear trend was used to test for an increasing or decreasing trend in the proportion of subjects for which average staining levels were Ͻ3 (less than normal tissue), equal to 3 (the same as normal tissue), or Ͼ3 (greater than normal tissue). The reliability of the mean of n raters has been estimated as the variance due to true scores divided by the sum of the variance due to true scores and the variance due to the mean of the errors of measurement. Reliability can be estimated as (F Ϫ 1)/F, where F is the ANOVA statistic for testing equality of subject means. Values of Ն0.7 for reliability are considered acceptable agreement between raters (44).

RESULTS Somatic Mutations. dHPLC was used to screen for somatic mutations in regions of each gene that encode protein domains with known functions. Case-specific DNA extracted from normal and cancer tissue was PCR-amplified using gene-specific genomic primers. The DNA encoding the following protein regions was amplified

(primer sequences available in Supplementary Data): the NH2-termi- ⑀ nal portion of the Ser/Thr Kinase domain of CKI ; the NH2-terminal region of the second PDZ domain of hDlg; and the NH2-terminal portion of the HECT domain of EDD/hHYD. After PCR amplification and agarose gel band confirmation, the samples were run on the WAVE Transgenomic dHPLC system. To confirm the mutant sequences, dHPLC was performed three times on each sample. Eight cases showed mutations in CSNK1⑀, 9 showed mutations in DLG1, and 9 showed mutations in EDD/hHYD. We have not yet tested the effects of any of these mutations on protein function. Fig. 1. Summary of mutations. Missense and nonsense mutations found in CSNK1⑀ ⑀ (A), DLG1 (B), and EDD/hHYD (C). Bold: nonconservative mutations; underlined: ATP CSNK1 . Direct sequence analysis of the mutations detected by binding sites in CKI⑀; star: mutation to stop codon. Scoring of nonconservative mutations dHPLC revealed 19 nonsynonymous mutations and no synonymous as in Ref. 58. 944

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2004 American Association for Cancer Research. TSG MUTATIONS IN MAMMARY DUCTAL CARCINOMA direct sequencing. This analysis was done on all 13 cases that showed was significantly associated with EDD/hHYD mutations (P ϭ 0.033, mutations in one or more of the three candidate genes analyzed. No ␹2 test). heteroduplex patterns resulted from dHPLC analysis of this fragment. Immunohistochemical Reactivity of CKI⑀. CKI⑀ gave a fairly These DNA fragments were sequenced and found to have no detect- uniform cytoplasmic staining pattern (Figs. 2 and 3, A–H), although in able mutations, indicating that the mutations we detected in the three some cases the staining was slightly more intense near the cell TSGs were not the result of widespread genomic instability. membrane. In the 85 cases evaluated for CKI⑀ reactivity, this protein LOH. LOH was observed at the following loci at the stated fre- showed increased reactivity (mean 3.98; SE ϭ 0.22; n ϭ 17) in DCIS, quency: d22s1156 (15 of 39; 38.5%); d22s272 (9 of 42; 21.4%); whereas invasive carcinoma (mean 3.04; SE ϭ 0.13; n ϭ 68) showed d22s423 (5 of 38; 13.2%); d8s1789 (7 of 35; 20.0%); d8s521 (4 of 28; little difference from normal mammary tissue. The difference between 14.3%); d8s1834 (3 of 39; 7.6%); d3s1265 (5 of 41; 12.2%); and CKI⑀ staining in DCIS and invasive carcinoma was statistically sig- d3s1272 (9 of 32; 28.1%). nificant (P ϭ 0.00007, t test). Therefore, CKI⑀ appears to be overex- Association between Mutations and LOH. We compared the pressed in DCIS. CKI⑀ staining was also found to vary according to mutation detection by dHPLC with LOH detected by fluorescent SBR score in a statistically significant manner (P ϭ 0.006, f test). microsatellite fragment analysis. We found that those microsatellite Well-differentiated ductal carcinoma gave a mean staining score of markers that mapped more closely to the given candidate TSG locus 3.93 (SE ϭ 0.29), with moderately differentiated ductal carcinoma showed statistically significant association between LOH and muta- giving a mean staining score of 2.95 (SE ϭ 0.16), and poorly differ- tion in the gene. LOH of the CSNK1⑀ microsatellite marker d22s1156 entiated mammary carcinoma giving a score of 2.76 (SE ϭ 0.22). was significantly associated with CSNK1⑀ mutations (P ϭ 0.0062, ␹2 Thus, CKI⑀ reactivity decreased in a statistically significant manner test); LOH of the telomeric DLG1 microsatellite marker d3s1272 was with increasing SBR Score. The reliability of the mean score from the significantly associated with DLG1 mutations (P ϭ 0.0027, ␹2 test); three observations was defined as the variance due to true scores and LOH of the centromeric microsatellite marker of EDD, d8s521, divided by the sum of the variance due to true scores and the variance

Fig. 2. Immunoreactive CKI⑀ expression. A and B, H&E control of normal breast and invasive mammary carcinoma, respectively; C and D, immunoperoxidase reactivity of CKI⑀ in normal and invasive cancer; and E and F, confocal immunofluorescent images using anti-CKI⑀ detected by Alexa Fluor 488 (green)-conjugated secondary antibodies and TOPROIII/DNA (blue).

945

Fig. 3. Immunoperoxidase reactivity of hDlg, CKI⑀, and EDD/hHYD. A and B, H&E control of case-paired normal breast and DCIS; C and D, immunoperoxidase reactivity of CKI⑀ in case- paired normal breast and ductal carcinoma in situ (DCIS); E and F, H&E control of case-paired normal breast and invasive mammary carcinoma; G and H, immunoperoxidase reactivity of CKI⑀ in case-paired normal breast and invasive cancer; I and J, H&E control of case-paired normal breast and DCIS; K and L, immunoperoxidase reactivity of hDlg in case-paired normal breast and DCIS; M and N, H&E control of case-paired normal breast and invasive mammary carcinoma; O and P, immunoperoxidase reactivity of hDlg in case- paired normal breast and invasive cancer; Q and R, H&E control of case paired normal breast and progesterone receptor [PR(Ϫ)] invasive cancer; S and T, immunoperoxidase reactivity of EDD/ hHYD in case-paired normal breast and PR(Ϫ) invasive cancer; U and V, H&E control of case paired normal breast and PR(ϩ) invasive cancer; and W and X, immunoperoxidase reactivity of EDD/hHYD in case-paired normal breast and PR(ϩ) invasive cancer. (Note the H&E stains are included to confirm pathology and may not be of the same regions of the case shown in the immunoperoxidase stain. They are from the same case).

due to the mean of the errors of the three scores. The three staining 2.27 (SE ϭ 0.26; P ϭ 0.136, f test). This final score was the lowest scores showed good agreement with a reliability estimate of 0.889. observed of the three candidate TSGs, suggesting that EDD/hHYD Immunohistochemical Reactivity of hDlg. In DCIS cases, hDlg reactivity was reduced the most of the three TSGs in mammary ductal was up-regulated with a mean staining score of 3.91 (SE ϭ 0.14, carcinoma. The reliability estimate for EDD/hHYD was the highest, n ϭ 17), whereas in invasive cases, it was down-regulated, giving a scoring 0.897, also showing good statistical agreement in the scoring score of 2.87 (SE ϭ 0.11, n ϭ 66). The difference in staining of EDD/hHYD reactivity. intensities of hDlg in DCIS versus invasive carcinoma was statisti- Mutations and Protein Expression. We compared the immuno- cally significant (P Ͻ 0.000005). Additionally, hDlg expression was histochemical staining scores for those cases that were evaluated reduced with increasing SBR score. Well-differentiated mammary using dHPLC to detect mutations. In all cases, there was a statistically ductal carcinoma gave a mean score of 3.26 (SE ϭ 0.25), whereas significant association between these two measures; however, the moderately differentiated ductal carcinoma gave a mean score of 2.91 trend of staining score was not the same for each gene (Fig. 6). Those (SE ϭ 0.15), and poorly differentiated mammary ductal carcinoma cases with mutations in CSNK1⑀ had a mean staining score of 2.50 gave a mean staining score of 2.58 (SE ϭ 0.21; P ϭ 0.119, f test). (SE ϭ 0.19) compared with 3.66 (SE ϭ 0.14) in cases with no Also, the membrane localization of hDlg was evaluated semiquanti- mutations detected (P Ͻ 0.0005). Cases with DLG1 mutations de- tatively (Figs. 3, I–P and 4). It was localized to the membrane in tected had a mean staining score of 2.50 (SE ϭ 0.16) compared with normal mammary epithelial ducts, but membrane localization was lost 3.62 (SE ϭ 0.13) observed in those without detectable mutations with increasing SBR score, and expression was intensely cytoplasmic (P Ͻ 0.0005). The mean EDD/hHYD staining score for cases with in DCIS cases. The reliability estimate for hDlg scoring was 0.842, mutations detected in that gene was 3.69 (SE ϭ 0.30) compared with showing good statistical agreement (44). 2.74 (SE ϭ 0.22) in cases without mutations (P ϭ 0.021, t test). Immunohistochemical Reactivity of EDD/hHYD. EDD/hHYD LOH and Protein Expression. We compared the staining score showed nuclear localization in normal mammary tissue as previously for each protein with the LOH frequency. The mean staining score of described in T47 mammary cell lines; however, in cancer sections, CKI⑀ was reduced in those cases showing LOH for two of the three staining was also reproducibly observed in the perinuclear cytoplasm microsatellite markers flanking the CSNK1⑀ gene. In all cases, there (Figs. 3, Q–T and 5). EDD/hHYD showed slightly increased staining was a reduced mean staining score for hDlg when LOH was detected intensity in DCIS (mean 3.41; SE ϭ 0.19, n ϭ 16), whereas a reduced in the markers flanking DLG1. EDD/hHYD gave the opposite result. intensity was seen in invasive carcinoma (mean 2.70; SE ϭ 0.14; LOH of three microsatellite markers was found to be significantly n ϭ 67). The difference between these two scores is statistically linked to reduced staining score. They were d22s272 (P ϭ 0.0014, t significant (P ϭ 0.004, t test). EDD/hHYD reactivity decreased with test), d3s1265 (P ϭ 0.014, t test), and d3s1272 (P ϭ 0.00033, t test). increasing SBR score. Well-differentiated tumors gave a mean stain Immunohistochemical Reactivity and Clinical Indicators. To score of 2.97 (SE ϭ 0.35), indicating little decrease compared with further evaluate the role of DLG1, CSNK1⑀, and EDD/hHYD in reactivity observed in normal mammary ducts. Moderately differen- mammary ductal carcinoma, we looked for association of protein stain tiated tumors gave a reactivity score of 2.85, and poorly differentiated of each of these TSGs with known clinical indicators (Table 1). We mammary ductal carcinoma demonstrated a mean staining score of examined lymph node, ER, PR, p53, Ki67, and ErbB2 status of each 946

Fig. 4. Immunoreactive hDlg expression. A and B, H&E control of normal breast and invasive mammary carcinoma, respectively; C and D, immunoperoxidase reactivity of hDlg in normal and invasive cancer; and E and F, confocal immunofluorescent images using anti-hDlg detected by Alexa Fluor 488 (green)-conjugated secondary antibodies and TOPROIII/DNA (blue).

case studied. We also associated protein stain of the three TSGs and of the poorly differentiated cancers were ER(Ϫ). This trend was each of the clinical indicators listed above with SBR score. In this significant (P Ͻ 0.0005). Similarly, 91.7% (11 of 12) of the well- evaluation, reduced stain corresponds to cancer cases wherein protein differentiated cancers were PR(ϩ), whereas 83.3% (15 of 18) of the staining intensity was less than that observed in case-specific normal poorly differentiated cancers were PR(Ϫ). This trend was significant tissue. (P Ͻ 0.0005). Of the tested indicators, CKI⑀ staining showed a significant asso- CKI⑀ and hDlg staining were both reduced in less differentiated ciation only with PR and ER status. The mean CKI⑀ staining score of cancers. A total of 9.1% (1 of 11) of well-differentiated, 41.7% (15 of PR(Ϫ) cancers was 2.63 (SE ϭ 0.22) compared with 3.26 (SE ϭ 0.15) 36) of moderately differentiated, and 55% (11 of 20) of poorly in PR(ϩ) cancers (P ϭ 0.02, ␹2 test). CKI⑀ reactivity was decreased differentiated cancers showed reduced CKI⑀ reactivity (P ϭ 0.017, ␹2 in 65% (13 of 20) of ER(Ϫ) cancers (P ϭ 0.035, t test), and a test for trend). Likewise, reduced hDlg reactivity was found in 25% (3 statistically significant trend of reduced CKI⑀ stain in ER(Ϫ) cancers of 12) of well-differentiated, 40% (14 of 35) of moderately differen- and increased or normal expression in ER(ϩ) cancers was found tiated, and 66.7% (12 of 18) of poorly differentiated ductal mammary (cltP ϭ 0.023). hDlg staining showed a statistically significant asso- carcinoma (P ϭ 0.019). A similar trend for EDD/hHYD was not ciation only with Ki67 and p53 status. The mean hDlg staining score observed (P ϭ 0.170); however, the mean staining score decreased as was 2.28 (SE ϭ 0.18) in Ki67(Ϫ) cancers and 3.15 (SE ϭ 0.12) in the mammary carcinomas became more poorly differentiated. Ki67(ϩ) cancers (P ϭ 0.00045, t test). The mean hDlg staining score ϭ Ϫ ϭ was 2.70 (SE 0.15) in p53( ) cancers and 3.19 (SE 0.16) in DISCUSSION p53(ϩ) cancers (P ϭ 0.03, t test). EDD/hHYD staining was reduced in both ER(Ϫ) and PR(Ϫ) cancers (2.44 and 2.53, respectively) The high frequency of somatic mutations we have discovered in compared with ER(ϩ) and PR(ϩ) cancers (2.87 and 2.86, respec- CSNK1⑀, together with mounting evidence that CKI⑀ activates the tively). Of the indicators assayed, SBR score was significantly asso- Wnt pathway, suggests that Wnt pathway activity changes may con- ciated only with ER and PR status. All of the well-differentiated tribute to at least some cases of breast cancer. The role of Wnt ductal mammary carcinoma cases were ER(ϩ), and 77.7% (14 of 18) signaling in breast cancer has been uncertain, although Wnt-1 was 947

Fig. 5. Immunoreactive EDD/hHYD expression. A and B, H&E control of normal breast and invasive mammary carcinoma, respectively; C and D, immunoperoxidase reactivity of EDD/hHYD in normal and invasive cancer; and E and F, confocal immunofluorescent images using anti-EDD/hHYD detected by Alexa Fluor 488 (green)-conjugated secondary antibodies and TOPROIII/DNA (blue).

originally identified as a mammary oncogene in mice infected with 48), and high ␤-catenin activity (indicated by high levels of the mouse mammary tumor virus (45). The Wnt pathway components protein in the cytoplasm and/or nucleus) is associated with poor APC, ␤-catenin, and Axin do not show high mutation rates in breast prognosis in breast cancer (47). It has therefore been suggested that cancer in contrast to the situation with colorectal cancer (46). How- Wnt pathway activation in breast cancer may result from somatic ever, several components of the Wnt signaling pathway, including mutations in other regulators of the pathway that have yet to be Wnt-1, ␤-catenin, and the target genes c-myc and cyclin D1, have discovered or analyzed (46). Our results suggest that CKI⑀ may be been reported as showing elevated expression in breast cancer (47, such a regulator of the pathway. In vertebrates, CKI⑀ has been shown to be a positive regulator of Wnt signaling and to be required to transduce Wnt signals. Further- more, inhibition of endogenous CKI⑀ with a kinase-defective version of the enzyme or with CKI⑀ antisense oligonucleotides attenuates Wnt signaling (49). In mammalian cells, CKI⑀ activates expression of a Lef-1 reporter gene ϳ10-fold, and this can be inhibited by coexpres- sion of axin (49). These results led to the suggestion that CKI⑀ activates the Wnt pathway by interacting with the ␤-catenin-Lef-1/ TCF complex. It is now known that CKI⑀ regulates the Wnt pathway Fig. 6. Association between staining reactivity and mutations. Staining score in cases both positively and negatively, by interacting directly with most of the with (right) and without (left) mutations detectable by denaturing high-performance liquid chromatography using the conditions described. The y axis is staining score, and the x axis known components of the pathway (50, 51). The reported apparently is case number. A, CKI⑀; B, hDlg; and C, EDD/hHYD. Mean staining score is indicated opposite effects of CKI⑀ on the Wnt pathway have led to the sugges- by sidebars. There was a statistically significant difference in staining score between cases ⑀ with and without mutations. Note that staining decreases in cases with mutations in DLG1 tion (52) that CKI could have opposite effects in the presence or and CSNK1⑀ but increases in cases with mutations in EDD/hHYD. absence of Wnt signaling. When the pathway is activated by Wnt 948

Table 1 Statistical association of reactivity, clinical indicators, and tumor phenotype

Data summary of mean staining scores of CKI␧, hDlg, and EDD/hHYD and with SD or SE-indicated by asterisk. Data are shown for select clinical indicators, ductal carcinoma in situ (DCIS) versus invasive cancer and by Scarff-Bloom-Richardson score.

CSNK1␧ hDlg EDD/hHYD

n Mean SE n Mean SE n Mean SE Select clinical indicators ERϩ 45 3.204 0.144 44 2.974 0.171 44 2.870 0.137 ERϪ 20 2.633 0.255 19 2.821 0.263 20 2.438 0.178 PRϩ 41 3.263 0.153 40 2.989 0.190 40 2.856 0.148 PRϪ 24 2.627 0.216 23 2.822 0.217 24 2.533 0.156 p53Ϫ 34 2.891 0.147 33 2.700 0.191 33 2.498 0.147 p53ϩ 29 3.181 0.234 28 3.191 0.222 29 3.038 0.164 Ki67Ϫ 18 2.844 0.222 16 2.281 0.211 17 2.576 0.184 Ki67ϩ 45 3.097 0.165 45 3.154 0.189 45 2.817 0.122 ErbB2Ϫ 27 2.960 0.216 27 3.056 0.237 27 2.687 0.196 ErbB2ϩ 34 3.019 0.177 32 2.741 0.191 33 2.706 0.124 LymphϪ 21 2.671 0.192 20 2.795 0.214 21 2.581 0.193 Lymphϩ 32 3.161 0.205 31 2.860 0.210 31 2.618 0.183 DCIS versus invasive score Invasive 68 3.038 0.126 66 2.870 0.11 67 2.698 0.141 DCIS 17 3.976 0.166 17 3.912 0.136 16 3.406 0.185 Scarff-Bloom-Richardson score Well 11 3.927 0.294 12 3.258 0.253 11 2.973 0.345 Moderately 36 2.949 0.163 35 2.913 0.148 35 2.850 0.193 Poorly 20 2.760 0.218 18 2.578 0.207 20 2.265 0.265 ligand, CKI⑀ could contribute positively to activation by phosphoryl- most of the residues involved in the active site. In DLG1, the mutation ating and activating Dsh, leading to glycogen synthase kinase-3␤ is in the PDZ2 domain, which could produce a protein that allowed (GSK-3␤) inactivation, but CKI⑀ would be unable to exert its inhib- faulty organization and possibly constitutive activation of the pathway itory effect on APC and ␤-catenin because GSK-3␤ is also required controlled by this scaffolding protein. for this effect. In the absence of Wnt ligand, GSK-3␤ would remain We found numerous examples of microsatellite LOH and clear active and CKI⑀ could contribute to inactivation of APC and ␤-cate- evidence of linkage between LOH and the somatic gene mutations, nin by creating priming sites for GSK-3␤ (52). This dependence of especially in those cases where the marker was physically close to the CKI⑀ action on the presence and activity of GSK-3␤ has been termed gene. For CSNK1⑀, four of six tumors showing mutations in the gene a molecular switch for the Wnt pathway (15). also showed LOH for both centromeric and telomeric microsatellite Mutations in hDlg could also contribute to breast and other cancers markers. For DLG1, two of six tumors showing mutations in the gene by affecting the interactions of this protein with the Wnt signaling also showed LOH for the centromeric marker and two showed LOH pathway. hDlg binds directly to APC, which is an important Wnt for the telomeric marker (see supplementary data published online for pathway component and known colorectal tumor suppressor protein details of individual cases). LOH could be selected in tumors either (53). This binding occurs both in vitro and in vivo, and the hDlg/APC because it removes a dominant tumor-suppressing allele or because it complex can block cell cycle progression (25). Another potential increases the dose of an oncogenic allele. We cannot distinguish interaction of hDlg with the Wnt pathway is suggested by a recent between these possibilities in our studies, but the statistically signif- report (54) that hDlg and other members of the PSD-95 subfamily of icant association between mutations and LOH does support our con- MAGUKs can bind directly to the Frizzled-1, -2, -4, and -7 receptors, clusion that the mutations are selected during growth of the tumor and each of which carries a type I PDZ-binding consensus sequence at the are functionally significant. COOH terminus, and that PSD-95 can form a ternary complex with The presence of somatic mutations in CSNK1⑀ and DLG1 was Frizzled-2 and the APC protein. Loss or alteration of hDlg could associated in a statistically significant manner with decreased anti- therefore affect the Wnt pathway and possibly contribute to oncogen- body reactivity for each gene product. Complete correlations are not esis by interfering with the localization of these receptors and their expected in these studies because all of the mutations we found are interactions with APC. point mutations, which may affect protein activity rather than amount To understand the effects of mutations in CSNK1⑀ or DLG1 on the of protein or its reaction with the antibody. The two tumors in which Wnt pathway and tumorigenesis, it will be important to determine stop codons were detected showed, as expected, very low levels of whether the mutations cause gain or loss of function. All of the 19 corresponding protein reactivity, although the antibody against CKI⑀ mutations we found in CSNK1⑀ and the 13 we found in DLG1 were recognized a COOH-terminal peptide, whereas the antibody against nonsynonymous and some predict nonconservative changes in the hDlg recognized an NH2-terminal peptide. protein. As previously reported, in the absence of cell selection after Our results with EDD/hHYD are different from those with CSNK1⑀ somatic mutation, the predicted ratio of nonsynonymous to synony- and DLG1 in several respects. First, 40% (4 of 10) of the mutations mous mutations (N:S) in human DNA is 2:1 (55). The ratio for detected in EDD/hHYD were synonymous or silent mutations, and the CSNK1⑀ is statistically significant from the prediction for unselected ratio between nonsynonymous and synonymous was not significantly mutations at the level of P ϭ 0.0005, and for DLG1, it is significant different from that predicted for unselected mutations. Second, all of at the level of P ϭ 0.0055. Therefore, the mutations we have de- the nonsynonymous mutations found in EDD/hHYD were missense scribed are probably selected during tumor growth. changes predicting conservative changes in the protein. Third, muta- In both CSNK1⑀ and DLG1, we found one example of a nonsense tions detected in EDD/hHYD were associated with an increase rather mutation producing a stop codon, predicting a COOH-terminally than a decrease in reactivity for the gene product. We interpret these truncated protein. However, in both cases, the position of the mutation differences as showing that the mutations did not lead to functionally allows for the possibility that the truncated protein has a gain rather significant protein alterations. However, of the three gene products, than a loss of function. In CSNK1⑀, the mutation is downstream of EDD/hHYD showed the greatest mean decrease in reactivity in inva- 949

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2004 American Association for Cancer Research. TSG MUTATIONS IN MAMMARY DUCTAL CARCINOMA sive breast cancer, suggesting that changes in its expression may be 12. Potter, C. J., Turenchalk, G. S., and Xu, T. Drosophila in cancer research. An functionally significant even if genetic changes do not contribute to expanding role. Trends. Genet., 16: 33–39, 2000. 13. Jursnich, V. A., Fraser, S. E., Held, L. I. J., Ryerse, J., and Bryant, P. J. Defective tumor formation and underscores the possibility that mutations in gap-junctional communication associated with imaginal disc overgrowth and degen- other regions of EDD/hHYD may be affecting EDD/hHYD expression. eration caused by mutations of the dco gene in Drosophila. Dev. Biol., 140: 413–429, 1990. We found that reactivity of each gene product decreased as tumor 14. Zilian, O., Frei, E., Burke, R., Brentrup, D., Gutjahr, T., Bryant, P. J., and Noll, M. differentiation decreased and SBR score increased. As the ductal Double-time is identical to discs overgrown, which is required for cell survival, carcinoma became more poorly differentiated, the staining for each proliferation and growth arrest in Drosophila imaginal discs. Development (Camb.), 126: 5409–5420, 1999. protein decreased or was lost completely suggesting that these pro- 15. Amit, S., Hatzubai, A., Birman, Y., Andersen, J. S., Ben Shushan, E., Mann, M., teins may be involved in tumor suppression as suggested from the Ben Neriah, Y., and Alkalay, I. Axin-mediated CKI phosphorylation of ␤-catenin at mutant phenotypes in model organisms. We also found that protein Ser 45: a molecular switch for the Wnt pathway. Genes Dev., 16: 1066–1076, 2002. 16. Yanagawa, S., Matsuda, Y., Lee, J. S., Matsubayashi, H., Sese, S., Kadowaki, T., and reactivity for each gene product increased in DCIS. This phenomenon Ishimoto, A. Casein kinase I phosphorylates the Armadillo protein and induces its was most pronounced for CKI⑀ and hDlg. degradation in Drosophila. EMBO J., 21: 1733–1742, 2002. Associations were found between hDlg stain reactivity and p53 17. Caruana, G. Genetic studies define MAGUK proteins as regulators of epithelial cell polarity. Int. J. Dev. Biol., 46: 511–518, 2002. status as well as Ki67 status. We also found that 62.5 and 75% of 18. Marfatia, S. M., Byron, O., Campbell, G. Liu, S. C., and Chishti, A. H. Human PR(Ϫ) cases showed a decrease in CKI⑀ and EDD/hHYD staining. homologue of the Drosophila discs large tumor suppressor protein forms an oligomer Similarly, we found that 65 and 75% of ER(Ϫ) cases showed a in solution. Identification of the self-association site. J. Biol. Chem., 275: 13759– ⑀ 13770, 2000. decrease in CKI and EDD/hHYD staining. As EDD/hHYD was first 19. Purmonen, S., Ahola, T. M., Pennanen, P., Aksenov, N., Zhuang, Y. H., Tuohimaa, characterized as a progestin-regulated gene, it follows that EDD/ P., and Ylikomi, T. HDLG5/KIAA0583, encoding a MAGUK-family protein, is a hHYD reactivity would associate with PR status, but it is not clear why primary progesterone target gene in breast cancer cells. Int. J. Cancer, 102: 1–6, 2002. 20. Mu¨ller, B. M., Kistner, U., Veh, R. W., Cases-Langhoff, C., Becker, B., Gundelfinger, CKI⑀ should show a similar trend. We found no association between E. D., and Garner, C. C. Molecular characterization and spatial distribution of SAP97, hDlg staining and ER/PR status, and no significant association be- a novel presynaptic protein homologous to SAP90 and the Drosophila discs-large tween reactivity of any of the gene products and ErbB2 (Her2/Neu) or tumor suppressor protein. J. Neurosci., 15: 2354–2366, 1995. 21. Watson, R. A., Rollason, T. P., Reynolds, G. M., Murray, P. G., Banks, L., and lymph node status. Roberts, S. Changes in expression of the human homologue of the Drosophila discs The somatic mutations we discovered in CSNK1⑀, DLG1, and large tumour suppressor protein in high-grade premalignant cervical neoplasias. Carcinogenesis (Lond.), 23: 1791–1796, 2002. EDD/hHYD show that this candidate gene approach can be successful 22. Liu, L. X., Liu, Z. H., Jiang, H. C., Qu, X., Zhang, W. H., Wu, L. F., Zhu, A. L., in identifying genes that may be functionally significant in human Wang, X. Q., and Wu, M. Profiling of differentially expressed genes in human gastric cancer. At least 50 additional TSGs have been identified in Drosophila carcinoma by cDNA expression array. World J. Gastroenterol., 8: 580–585, 2002. 23. Caruana, G., and Bernstein, A. Craniofacial dysmorphogenesis including cleft palate through the finding of excess growth caused by loss-of-function in mice with an insertional mutation in the discs large gene. Mol. Cell. Biol., 21: mutations, and Ͼ20 of these have been cloned and studied at the 1475–1483, 2001. molecular level (56, 57). Their homologues provide a rich source of 24. Nix, S. L., Chishti, A. H., Anderson, J. M., and Walther, Z. hCASK and hDlg associate in epithelia, and their src homology 3 and guanylate kinase domains candidate genes that can be tested using techniques similar to those participate in both intramolecular and intermolecular interactions. J. Biol Chem., 275: described here and that may also be shown to play significant roles in 41192–41200, 2000. many kinds of cancer. 25. Ishidate, T., Matsumine, A., Toyoshima, K., and Akiyama, T. The APC-hDLG complex negatively regulates cell cycle progression from the G0-G1 to S phase. Oncogene, 19: 365–372, 2000. 26. Adey, N. B., Huang, L., Ormonde, P. A., Baumgard, M. L., Pero, R., Byreddy, D. V., ACKNOWLEDGMENTS Tavtigian, S. V., and Bartel, P. L. Threonine phosphorylation of the MMAC1/PTEN PDZ binding domain both inhibits and stimulates PDZ binding. Cancer Res., 60: We thank Dr. Knut Ketterer for LCM training. We also thank Dr. Kevin 35–37, 2000. Kiesler of the Fragment Analysis Facility at the University of Arizona and 27. Kiyono, T., Hiraiwa, A., Fujita, M., Hayashi, Y., Akiyama, T., and Ishibashi, M. Binding of high-risk human papillomavirus E6 oncoproteins to the human homologue Karen Sutton for their assistance in DNA fragment analysis. We thank Manxu of the Drosophila discs large tumor suppressor protein. Proc. Natl. Acad. Sci. USA, Zhao and Guangli Cheng for their help in tissue preparation and DNA extrac- 94: 11612–11616, 1997. tion. 28. Henderson, M. J., Russell, A. J., Hird, S., Munoz, M., Clancy, J. L., Lehrbach, G. M., Calanni, S. T., Jans, D. A., Sutherland, R. L., and Watts, C. K. EDD, the human hyperplastic discs protein, has a role in progesterone receptor coactivation and potential involvement in DNA damage response. J. Biol. Chem., 277: 26468–26478, REFERENCES 2002. 1. Li, C. I., Anderson, B. O., Daling, J. R., and Moe, R. E. Trends in incidence rates of 29. Callaghan, M. J., Russell, A. J., Woollatt, E., Sutherland, G. R., Sutherland, R. L., and invasive lobular and ductal breast carcinoma. J. Am. Med. Assoc., 289: 1421–1424, Watts, C. K. Identification of a human HECT family protein with homology to the 2003. Drosophila tumor suppressor gene hyperplastic discs. Oncogene, 17: 3479–3491, 2. Weiss, N. S. Breast cancer mortality in relation to clinical breast examination and 1998. breast self-examination. Breast J., 9 (Suppl. 2): S86–S89, 2003. 30. Lee, J. D., Amanai, K., Shearn, A., and Treisman, J. E. The ubiquitin ligase 3. Gruvberger, S. K., Ringner, M., Eden, P., Borg, A., Ferno, M., Peterson, C., and Hyperplastic discs negatively regulates hedgehog and decapentaplegic expression by Meltzer, P. S. Expression profiling to predict outcome in breast cancer: the influence independent mechanisms. Development (Camb.), 129: 5697–5706, 2002. of sample selection. Breast Cancer Res., 5: 23–26, 2002. 31. Mansfield, E., Hersperger, E., Biggs, J., and Shearn, A. Genetic and molecular 4. Balmain, A., Gray, J., and Ponder, B. The genetics and genomics of cancer. Nat. analysis of hyperplastic discs, a gene whose product is required for regulation of cell Genet., 33 (Suppl.): 238–244, 2003. proliferation in Drosophila melanogaster imaginal discs and germ cells. Dev. Biol., 5. Lou, Z., and Chen, J. Brca proteins and DNA damage checkpoints. Front Biosci., 8: 165: 507–526, 1994. S718–S721, 2003. 32. Honda, Y., Tojo, M., Matsuzaki, K., Anan, T., Matsumoto, M., Ando, M., Saya, H., 6. Vasen, H. F. Screening in breast cancer families: is it useful? Ann. Med., 26: and Nakao, M. Cooperation of HECT-domain ubiquitin ligase hHYD and DNA 185–190, 1994. topoisomerase II-binding protein for DNA damage response. J. Biol. Chem., 277: 7. Keen, J. C., and Davidson, N. E. The biology of breast carcinoma. Cancer (Phila.), 97: 3599–3605, 2002. 825–833, 2003. 33. Amat, S., Penault-Llorca, F., Cure, H., Le Bouedec, G., Achard, J. L., Van, P., I, 8. Porter, D., Lahti-Domenici, J., Keshaviah, A., Bae, Y. K., Argani, P., Marks, J., Feillel, V., Mouret-Reynier, M. A., Dauplat, J., and Chollet, P. Scarff-Bloom- Richardson, A., Cooper, A., Strausberg, R., Riggins, G. J., Schnitt, S., Gabrielson, E., Richardson (SBR) grading: a pleiotropic marker of chemosensitivity in invasive Gelman, R., and Polyak, K. Molecular markers in ductal carcinoma in situ of the ductal breast carcinomas treated by neoadjuvant chemotherapy. Int. J. Oncol., 20: breast. Mol. Cancer Res., 1: 362–375, 2003. 791–796, 2002. 9. Balmain, A. Cancer as a complex genetic trait: tumor susceptibility in humans and 34. Blumenstein, R., Dias, M., Russo, I. H., Tahin, Q., and Russo, J. DNA content and mouse models. Cell, 108: 145–152, 2002. cell number determination in microdissected samples of breast carcinoma in situ. Int. 10. Rubin, G. M., and Lewis, E. B. A brief history of Drosophila’s contributions to J. Oncol., 21: 447–450, 2002. genome research. Science (Wash. DC), 287: 2216–2218, 2000. 35. Zhu, W., Zou, H., Beck, A., Chervinsky, D., Malik, D., Brooks, J., and Tan, D. Loss 11. Bryant, P. J., and Schmidt, O. The genetic control of cell proliferation in Drosophila of heterozygosity in primary lung cancer using laser capture microdissection and imaginal discs. J. Cell Sci. Suppl., 13: 169–189, 1990. WAVE DNA fragment analysis techniques. Med. Sci. Monit., 8: BR95–BR99, 2002. 950

36. Reifenberger, J., Knobbe, C. B., Wolter, M., Blaschke, B., Schulte, K. W., Pietsch, T., 47. Lin, S. Y., Xia, W., Wang, J. C., Kwong, K. Y., Spohn, B., Wen, Y., Pestell, R. G., Ruzicka, T., and Reifenberger, G. Molecular genetic analysis of malignant melano- and Hung, M. C. ␤-Catenin, a novel prognostic marker for breast cancer: its roles in mas for aberrations of the WNT signaling pathway genes CTNNB1, APC, ICAT, and cyclin D1 expression and cancer progression. Proc. Natl. Acad. Sci. USA, 97: BTRC. Int. J. Cancer, 100: 549–556, 2002. 4262–4266, 2000. 37. Schildkraut, J. M., Collins, N. K., Dent, G. A., Tucker, J. A., Barrett, J. C., Berchuck, 48. Wong, S. C., Lo, S. F., Lee, K. C., Yam, J. W., Chan, J. K., and Wendy Hsiao, W. L. A., and Boyd, J. Loss of heterozygosity on chromosome 17q11-21 in cancers of Expression of frizzled-related protein and Wnt-signalling molecules in invasive women who have both breast and ovarian cancer. Am. J. Obstet. Gynecol., 172: human breast tumours. J. Pathol., 196: 145–153, 2002. 908–913, 1995. 49. Sakanaka, C., Leong, P., Xu, L., Harrison, S. D., and Williams, L. T. Casein kinase 38. Callen, D. F., Crawford, J., Derwas, C., Cleton-Jansen, A. M., Cornelisse, C. J., and I epsilon in the wnt pathway: regulation of ␤-catenin function. Proc. Natl. Acad. Sci. Baker, E. Defining regions of loss of heterozygosity of 16q in breast cancer cell lines. USA, 96: 12548–12552, 1999. Cancer Genet. Cytogenet., 133: 76–82, 2002. 50. Peters, J. M., McKay, R. M., McKay, J. P., and Graff, J. M. Casein kinase I transduces 39. Kleymenova, E., and Walker, C. L. Determination of loss of heterozygosity in frozen Wnt signals. Nature (Lond.), 401: 345–350, 1999. and paraffin embedded tumors by denaturing high-performance liquid chromatogra- 51. Lee, E., Salic, A., and Kirschner, M. W. Physiological regulation of ␤-catenin phy (DHPLC). J. Biochem. Biophys. Methods, 47: 83–90, 2001. stability by Tcf3 and CK1⑀. J. Cell Biol., 154: 983–993, 2001. 40. Liu, W., Smith, D. I., Rechtzigel, K. J., Thibodeau, S. N., and James, C. D. 52. Rubinfeld, B., Tice, D. A., and Polakis, P. Axin-dependent phosphorylation of the Denaturing high performance liquid chromatography (DHPLC) used in the detection adenomatous polyposis coli protein mediated by casein kinase 1epsilon. J. Biol. of germline and somatic mutations. Nucleic Acids Res., 26: 1396–1400, 1998. Chem., 276: 39037–39045, 2001. 41. Cobb, C. J., Scott, G., Swingler, R. J., Wilson, S., Ellis, J., MacEwen, C. J., and 53. Matsumine, A., Ogai, A., Senda, T., Okumura, N., Satoh, K., Baeg, G. H., Kawahara, McLean, W. H. Rapid mutation detection by the transgenomic wave analyser DHPLC T., Kobayashi, S., Okada, M., Toyoshima, K., and Akiyama, T. Binding of APC to the identifies MYOC mutations in patients with ocular hypertension and/or open angle human homolog of the Drosophila discs large tumor suppressor protein. Science glaucoma. Br. J. Ophthalmol., 86: 191–195, 2002. (Wash. DC), 272: 1020–1023, 1996. 42. Rossetti, S., Chauveau, D., Walker, D., Saggar-Malik, A., Winearls, C. G., Torres, 54. Hering, H., and Sheng, M. Direct interaction of Frizzled-1, -2, -4, and -7 with PDZ V. E., and Harris, P. C. A complete mutation screen of the ADPKD genes by DHPLC. domains of PSD-95. FEBS Lett., 521: 185–189, 2002. Kidney Int., 61: 1588–1599, 2002. 55. Bardelli, A., Parsons, D. W., Silliman, N., Ptak, J., Szabo, S., Saha, S., Markowitz, S., 43. Frost, A. R., Sparks, D., and Grizzle, W. E. Methods of antigen recovery vary in their Willson, J. K., Parmigiani, G., Kinzler, K. W., Vogelstein, B., and Velculescu, V. E. usefulness in unmasking specific antigens in immunohistochemistry. Appl. Immuno- Mutational analysis of the tyrosine kinome in colorectal cancers. Science (Wash. DC), histochem. Mol. Morphol., 8: 236–243, 2000. 300: 949, 2003. 44. Winer, B. J. Statistical Principles in Experimental Design, Ed. 2. New York City: 56. Watson, K. L., Justice, R. W., and Bryant, P. J. Drosophila in cancer research: the first McGraw-Hill, 1971. fifty tumor suppressor genes. J. Cell Sci. Suppl., 18: 19–33, 1994. 45. Nusse, R. The int genes in mouse mammary tumorigenesis and in normal develop- 57. Hay, B. A., and Guo, M. Coupling cell growth, proliferation, and death. Hippo weighs ment. CIBA Found. Symp., 150: 212–222; discussion 222–226, 1990. in. Dev. Cell, 5: 361–363, 2003. 46. Brown, A. M. Wnt signaling in breast cancer: have we come full circle? Breast 58. Pearson, W. R., Wood, T., Zhang, Z., and Miller, W. Comparison of DNA sequences Cancer Res., 3: 351–355, 2001. with protein sequences. Genomics, 46: 24–36, 1997.

951

Downloaded from cancerres.aacrjournals.org on October 2, 2021. © 2004 American Association for Cancer Research. Somatic Mutations and Altered Expression of the Candidate Tumor Suppressors CSNK1ε, DLG1, and EDD/hHYD in Mammary Ductal Carcinoma

Tannin J. Fuja, Fritz Lin, Kathryn E. Osann, et al.

Cancer Res 2004;64:942-951.

Updated version Access the most recent version of this article at: http://cancerres.aacrjournals.org/content/64/3/942

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2004/04/23/64.3.942.DC1

Cited articles This article cites 57 articles, 20 of which you can access for free at: http://cancerres.aacrjournals.org/content/64/3/942.full#ref-list-1

Citing articles This article has been cited by 23 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/64/3/942.full#related-urls

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

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/64/3/942. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.