We hypothesized that cells bearing a single inherited “hit” in a tumor suppressor gene express an altered mRNA repertoire that may identify targets for measures that could delay or even prevent progression to carcinoma. We report here on the transcriptomes of primary breast and ovarian epithelial cells cultured from BRCA1 and BRCA2 mutation carriers and controls. Our comparison analyses identified multiple changes in gene expression, in both tissues for both mutations, which were validated independently by real-time reverse transcription-PCR analysis. Several of the differentially expressed genes had been previously proposed as cancer markers, including mammaglobin in breast cancer and serum amyloid in ovarian cancer. These findings show that heterozygosity for a mutant tumor suppressor gene can alter the expression profiles of phenotypically normal epithelial cells in a gene-specific manner; these detectable effects of “one hit” represent early molecular changes in tumorigenesis that may serve as novel biomarkers of cancer risk and as targets for chemoprevention. Cancer Prev Res; 3(1); 48–61

The notion of multistep carcinogenesis posits that rate-limiting mutations accumulate in a single cell and its progeny, marking recognizable histopathologic transitions in the target tissues (1, 2). The time required for this accumulation affords the opportunity to test whether pharmacologic and/or dietary interventions can delay or prevent the transition to malignancy (3). Early targeted intervention would be optimally performed on persons with a very high risk of developing a specific cancer, as with those individuals who carry a germline mutation in a gene known to impose such a risk. Rationale for such an intervention is provided by early studies that showed that cells heterozygous for a cancer-predisposing mutation could show abnormalities in tissue cultures; “one-hit” effects in heterozygous cells were seen in morphologically normal cultured fibroblasts and in epithelial cells derived from familial adenomatous polyposis patients (46). Furthermore, we have recently reported specific changes in protein expression in colonic epithelial cells from familial adenomatous polyposis patients (7). Support for the significance of these early changes comes from observations of similar aberrations in corresponding cancer cells (79). Such changes may play a role in progression to malignancy and therefore constitute targets for strategies to delay or prevent such progression in familial adenomatous polyposis and in other genetic predispositions to cancer.

With this rationale in mind, we have undertaken an investigation of two of the most common predisposing genes, BRCA1 (10, 11) and BRCA2 (refs. 12, 13, and references therein), in two important target tissues, breast and ovary. We note that previous reports on benign cells associated with breast cancer already suggest the possibility of such heterozygous effects.

We have compared the transcriptomes of primary breast and ovarian epithelial cultures from patients predisposed to cancer, bearing monoallelic BRCA1 or BRCA2 mutations, with corresponding cultures from control individuals. We show that the morphologically normal epithelial cells from mutation carriers exhibit abnormalities in a gene- and tissue-specific manner, consistent with detectable single-hit effects. These alterations constitute possible molecular targets for intervention on the path to cancer.

Subject accrual and biopsy specimens

All the subjects were recruited with the approval of the Fox Chase Cancer Center Institutional Review Board, irrespective of gender, race, and age. Individuals with a personal history of cancer and subjects treated previously with either chemotherapy or radiation were ineligible. Eligible cases included unaffected at-risk women in the Fox Chase Family Risk Assessment Program who were shown to be carriers of BRCA1 or BRCA2 mutations. In particular, six BRCA1, six BRCA2 mutation carriers, and six healthy controls were accrued for breast specimens and an equal number for ovary specimens. Normal breast and ovary specimens were obtained by prophylactic oophorectomy or mastectomy or breast reduction surgery.

Cell culture establishment

Surgical breast specimens were placed in transport medium (serum-free Ham's F-12), containing 100 U/mL penicillin, 100 μg/mL of streptomycin, 10 μg/mL ciprofloxacin, 10 μg/mL gentamicin, 2.5 μg/mL of Amphotericin B, and 100 U/mL of Nystatin. The tissue was finely minced using sterile disposable scalpels and was transferred to a tube containing 25 mL of 200 U/mL solution of collagenase (Sigma) prepared in DMEM with 2 g/liter of NaHCO3, supplemented with 160 U/mL of Hyaluronidase, 0.5 μg/mL hydrocortisone, 10 μg/mL insulin, 10 mL of Antibiotic/Antimycotic (Life Technologies), and 10% horse serum. The tissue was digested overnight at 37°C in a rotating water bath and then centrifuged at 2,200 rpm for 10 min. The supernatant was carefully decanted to a sterile tube. The tissue was rinsed four times with transport medium, resuspended in culture medium, and centrifuged one last time. The tissue was then plated in a swine skin gelatin–coated (Sigma) T-25 flask. Cells were cultured for 24 h in high-calcium medium and then refed with low-calcium medium 24 h later. High-calcium medium consists of DMEM/F12 1:1 without calcium (Life Technologies), supplemented with 5% chelated horse serum, 20 ng/mL epidermal growth factor, 100 ng/mL cholera toxin, 10 μg/mL insulin, 0.5 μg/mL hydrocortisone, 1.05 mmol/L calcium chloride, 100 U/mL penicillin, 100 μg/mL streptomycin, 10 μg/mL ciprofloxacin, and 0.25 μg/mL Amphotericin B. Low-calcium medium was the same recipe supplemented with 0.04 mmol/L calcium chloride (14). Cells were cultured 4 to 6 wk until the flask was confluent.

Oophorectomy specimens were collected under aseptic conditions and placed in a transport medium (M199:MCDB105, 1:1) supplemented with 100 U/mL penicillin, 100 μg/mL streptomycin, and 2 mmol/L l-glutamine. The ovaries were processed to establish epithelial cell cultures by gently scraping the ovarian surface with a rubber policeman. Cells were centrifuged and resuspended in fresh medium (M199:MCDB105, 1:1); supplemented with 5% fetal bovine serum, penicillin, streptomycin, glutamine, and 0.3 U/mL insulin; and transferred to tissue culture flasks coated with skin gelatin; they were refed every 4 d and passaged once they reached confluency.

All the breast and ovarian samples were treated with the same tissue-specific culture conditions, including timing for passaging and harvesting. Importantly, all the samples were deidentified, including notation on carrier or control status, and no significant difference in growth or apoptosis among them was noted. At harvest, all the cultures were in log phase.

RNA extraction and amplification

Total RNA was prepared from cultured cells by extraction in guanidinium isothiocyanate–based buffer containing β-mercaptoethanol and acid phenol. RNA integrity was evaluated on the Agilent 2100 Bioanalyzer. All samples showed distinct peaks corresponding to intact 28S and 18S rRNAs and therefore were included in the analysis. Amplification of total RNAs was achieved using the one-cycle Ovation biotin system (NuGEN Technologies, Inc.) as described previously (15).

Hybridization and microarray analysis

For each sample, a total of 2.2 μg of ssDNA, labeled and fragmented with the NuGEN kit, was hybridized to Affymetrix arrays (Human U133 plus 2.0), following the manufacturer's instructions as described previously (15). After washing and staining with biotinylated antibody and streptavidin phycoerythrin, the arrays were scanned with the Affymetrix GeneChip Scanner 3000 for data acquisition.

Real-time reverse transcriptase-PCR validation of microarray data

Validation of microarray findings was conducted by real-time RT-PCR using Taqman low-density arrays (LDA; microfluidic cards from Applied Biosystems). A 48-gene custom-made array (44 candidate biomarkers plus 4 house keeping genes) was designed and prepared by Applied Biosystems. The entire panel of 48 genes was tested across breast and ovarian samples. All samples were tested in quadruplicate to ensure accuracy and reproducibility.

Data were obtained in the form of threshold-cycle number (Ct) for each candidate biomarker identified and the housekeeping gene HPRT1 for each genotype (BRCA1, BRCA2, and WT). For each gene, the Ct values were normalized to the housekeeping gene and the corresponding ΔCt values were obtained for each genotype. Relative quantitation was computed using the Comparative Ct method (Applied Biosystems Reference Manual, User Bulletin no.2) between BRCA1 mutants and WT primary cell RNAs. The relative quantitation is the ratio of the normalized amounts of target for BRCA1 mutant for WT RNAs, and is computed as 2^(−ΔCt) where ΔCt is the difference between the mean ΔCt values for BRCA1 mutant and the mean ΔCt values for WT RNAs. We repeated the relative quantitation analyses for BRCA2 mutant and WT RNA samples.

Statistical analysis

We considered breast and ovarian samples for each of the three genotypes: BRCA1, BRCA2, and mutation negative or WT. There were six biological replicates in each experimental condition resulting in a total of 18 samples for each target organ. For each sample, we obtained probe-level data in the form of raw signal intensities for 54,675 probe sets from Affymetrix.CEL files. Raw data for each target organ were preprocessed separately using the Robust Multichip Average method proposed by Irizarry et al. (16, 17).

We applied the variance-stabilizing and normalizing logarithmic transformation to the data before analysis, and used the Local Pooled Error method (18) for class comparisons. Local Pooled Error is based on pooling errors within genes and between replicate arrays for genes whose expression values are similar. All comparisons were two sided. The Benjamini-Hochberg step-up method (19) was applied to control the false discovery rate (FDR). Genes were defined as differentially expressed, based on statistical significance as well as biological significance. Genes showing a FDR of less than the desired cutoff were considered statistically significant. We accepted a FDR cutoff of 0.20 for breast and 0.10 for ovary. These cutoffs were selected to obtain similar numbers of genes. Biological significance was measured as fold change, i.e., the ratio of the mean expression profiles between two conditions. Genes showing >2-fold change in either direction (upregulated or downregulated) were considered biologically significant. Differentially expressed genes from each of the above filters were combined, and a list of common genes showing statistical and biological significance was identified. These genes were subsequently validated using RT-PCR. The analytic tools available in the R/BioConductor package,10

,11 bioNMF (20), and TMev12 were used in these computations.

Data mining analysis

Pathway and association analyses were conducted to obtain additional insight into the functional relevance of the changes observed. Upregulated and downregulated genes for these exploratory analyses were selected as described above, but using a more relaxed P value cutoff of 0.001. Gene ontology functional categories enriched in differentially expressed genes were identified using conditional hyper-geometric tests in the GOstats package (R/BioConductor). A P value cutoff of 0.01 was used in selecting the Gene ontology terms. Furthermore, gene networks were generated using Ingenuity Pathways Analysis version 6.5 (Ingenuity Systems).13

Gene Set Enrichment Analysis (GSEA; ref. 21) was performed against the lists of differentially expressed genes for BRCA1-WT and BRCA2-WT comparisons. Gene sets from MSigDB (21), including positional, curated, motif, and computational sets, were tested. Default parameters were chosen, except that the maximum intensity of probes was only selected while collapsing probe sets for a single gene.

Next, we compiled and analyzed publicly available microarray gene expression data from the following: (a) a study of mammary gland side population cells (22); (b) a study of two different human breast epithelial cell types, breast primary epithelial cells (BPEC) and human mammary epithelial cells (HMEC), compared with their transformed counterparts (23); (c) a molecular characterization of cancer stem cells in MMTV-Wnt1 murine breast tumors (24); (d) profiles of hereditary breast cancer (25); and (e) a set of annotated genes involved in DNA damage repair. The data from these studies were analyzed (a description of the methods is given in Supplementary Materials). The differentially expressed genes between BRCA1, BRCA2, and WT from our study were compared against the above mentioned gene sets using GSEA.

Genome-wide transcriptome analysis of single-hit BRCA1 and BRCA2 and mutation-negative breast and ovarian epithelial cell cultures

We are interested in the growth behavior of cells that are precursors of cancer; therefore, we chose to study primary cells multiplying in culture. Morphologically normal primary breast and ovarian epithelial cells were established from BRCA1 and BRCA2 mutation carriers, and mutation-negative individuals (Fig. 1). Demographic and mutation data of mutation carriers and control individuals are shown in Table 1. Table 1 shows that our population is mostly Caucasian, and that carriers and controls are well matched for age, race, parity, menopausal status, and body mass index, with the only exception being the group of BRCA1 carriers donating ovarian epithelial cells in which there is a predominance of premenopausal women. This is due to the fact that carriers of highly penetrant BRCA1 mutations were advised to undergo prophylactic oophorectomy. We conducted microarray studies of these primary epithelial cultures to identify differentially expressed genes between BRCA1 mutant and BRCA2 mutant single-hit cells and WT cells, for each target organ.

Class comparison analyses (BRCA1 versus WT, and BRCA2 versus WT) revealed notable changes in gene expression, indicating that the heterozygous mutations in BRCA1 and BRCA2 do indeed affect the expression profiles of cultured primary epithelial cells from the relevant target organs, breast and ovary.

Breast epithelial cells

Table 2 summarizes selected gene expression fold changes on a linear scale in breast cells versus controls. The secretoglobin family of genes (SCGB2A1, SCGB2A2, and SCGB1D2), of unknown function, is highly upregulated in BRCA1 mutant breast cells. The genes have been recently described as novel serum markers of breast cancer with significant prognostic value (26, 27); ∼80% of all breast cancers overexpress this complex (26). In BRCA1 cells, we observed a 3-fold upregulation of mammaglobin (SCGB2A2: FDR, 0.06; P = 4 × 10−10) and a 12-fold increase of lipophilin B and C (SCGB1D2 and SCGB2A: FDR, 0.06 and P = 2 × 10−8; and FDR, 0.16; P = 0.0002, respectively). In BRCA1 cells, we also detected a 3-fold upregulation of the chitinase 3–like 1 gene (FDR, 0.06; P = 1.2 × 10−7), which has proliferative effects on stromal fibroblasts and chemotactic effects on endothelial cells. It can promote angiogenesis, and high serum levels of this protein have been found in patients with glioblastoma (28). IGFBP5 was upregulated 10-fold in BRCA2 breast cells (FDR, 0.04; P = 1.9 × 10−7). It is involved in the stimulation of growth and binding to extracellular matrix, independently of insulin-like growth factor, and is highly overexpressed in breast cancer tissues (29).

Several cell-to-cell interactions and cell-to-matrix adhesion genes were found to be downregulated in breast cells, including those that code for tensin 4 (TNS4: fold change, 0.26; FDR, 0.06; P = 2.2 × 10−7 in BRCA1; and fold change, 0.5; FDR, 0.03; P = 3.7 × 10−6 in BRCA2), for mucin 16 (MUC16: fold change, 027; FDR, 0.08; P = 3 × 10−5 in BRCA1; and fold change, 024; FDR, 0.03; P = 2 × 10−6 in BRCA2), and for keratin 14 (KRT14: fold change, 0.5; FDR, 0.12; P = 0.0001 in BRCA1; and fold change, 0.5; FDR, 0.03; P = 2.2 × 10−10 in BRCA2). Lack of tensin 4 expression has been reported in prostate and breast cancers (30), suggesting that the downregulation of tensin expression is a functional marker of cell transformation. In addition, loss of keratins, which are necessary for proper structure and function of desmosomes, can cause an increase in cell flexibility and deformability, and may enable a tumor cell to detach from its epithelial layer, and metastasize. Finally, mucin 1 (MUC1 or CA15-3), which is known to be overexpressed in breast cancer (31), is downregulated in “single-hit” BRCA1 cells (fold change, 0.5; FDR, 0.99; P = 0.02). Overrepresentation of even one glycoprotein may affect cell surface protein distribution, with effects on other membrane proteins, such as the downregulation of genes encoding mucin 16 and mucin 1 that we have detected in mutant breast cells. A box plot representation of the top differentially expressed genes for breast epithelial cells is shown in Fig. 2A.

Ovarian epithelial cells

Table 3 summarizes expression changes for selected genes in ovarian epithelial cultures. For example, 5-fold downregulation of the cyclin B1/cdc2 complex (CDC2: FDR, 0.04; P = 7 × 10−5), a key regulator controlling the G2M checkpoint, was observed in BRCA1 mutant ovarian cells. Multiple genes implicated in the mitotic spindle checkpoint, such as nucleolar and spindle-associated protein 1 (NUSAP-1: fold change, 0.08; FDR, 0.02; P = 9 × 10−9 in BRCA1) were downregulated. NUSAP-1 plays a crucial role in spindle microtubule organization, whereas CENP-A, which is downregulated in BRCA2 mutants of breast cells (fold change, 0.5; FDR, 0.03; P = 9 × 10−6), is essential for centromere structure, function, and kinetochore assembly. Because BRCA1 and BRCA2, in addition to their role in DNA repair, are also involved in checkpoint pathways, we suggest that inappropriate expression of these proteins could induce abnormal kinetochore function and chromosome missegregation, a potential cause of aneuploidy and a critical contributor to oncogenesis. Among the upregulated genes in BRCA1 heterozygous ovarian epithelial cells is SAA2, an acute phase component of the innate immune system (fold change, 6.4; FDR, 0.02; P = 2 × 10−7), a candidate marker of epithelial ovarian cancer (refs. 32, 33 and references therein). Among the differentially expressed genes in BRCA2 mutant ovarian epithelial cells, matrix metalloproteinase 3 (MMP3) was upregulated 9- to 12-fold (FDR, 0.1; P = 2 × 10−7); the same tendency has been reported for MMP1 and MMP2 in ovarian cancers (34, 35). Our data also show upregulation of COX1 (PTGS1: fold change, 6.6; FDR, 0.1; P = 8 × 10−9) in BRCA2 ovarian epithelial cells, a finding that is consistent with the reported upregulation of cyclooxygenase 1, but not cyclooxygenase 2, in ovarian cancer (36, 37). A box plot representation of the top differentially expressed genes for ovarian epithelial cells is shown in Fig. 2B.

Real-time reverse transcriptase-PCR validation of microarray data

A validation study on select genes for breast and ovary was done with total RNA using quantitative, real-time RT-PCR on low-density arrays. We selected 44 candidate biomarker genes and 4 housekeeping genes, and the entire panel of 48 genes was tested by the Comparative C1 method across all breast and ovarian samples in quadruplicate using a custom-made array to ensure accuracy and reproducibility. The real-time RT-PCR validation results and the comparisons with the original Affymetrix data are shown in Tables 2 and 3 for breast and ovary, respectively. There was a good correlation (Spearman's ρ) between microarray and LDA data for fold changes of candidate biomarkers in breast and ovarian cultures heterozygous for BRCA1 (0.94 in each case). For candidate biomarkers originally identified in breast and ovarian cultures for one genotype (BRCA1 or BRCA2), there was also a moderate to good correlation between microarray and LDA data for the other genotype (BRCA2 or BRCA1). This is described in more detail in the Supplementary Section and the results are presented in Supplementary Table S1.

Functional mining of microarray data

To define the biological underpinnings of the observed gene expression differences, we conducted additional mining of the microarray data using gene ontology, pathway, and association analyses. Gene ontology analysis revealed overrepresentation of several biological processes in BRCA1 and BRCA2 mutant cells (see Supplementary Information; Fig. 3). Next, using differentially expressed genes as input to the Ingenuity Pathway Analysis software, we generated networks and overlaid pathways onto genes to understand their interactions and functional importance. In the case of breast BRCA1 cells, two gene networks had many downregulated genes involved in G2-M DNA damage checkpoint regulation, DNA damage response, p38 mitogen-activated protein kinase signaling, and tight junction signaling. Genes such as ATR, PMS1, and MCM6 that are involved in BRCA1-related DNA damage response were upregulated in BRCA1 heterozygous cells (Fig. 4A). Similarly, in breast BRCA2 cells, one significant network that contained genes involved in G1-S checkpoint regulation and ephrin receptor signaling was identified (Fig. 4B). In BRCA1 ovarian cells, two significant networks, involved in cell cycle control (G2-M DNA damage checkpoint regulation) and pyrimidine metabolism, and in glucocorticoid receptor signaling, contained downregulated genes (Fig. 4C). We did not find any significant networks for genes differentially expressed in ovarian BRCA2 heterozygous cells.

To identify unifying biological themes central to mutant breast and ovarian cells compared with WT cells, we used the GSEA for the detection of complex relationships among coregulated genes. GSEA is an analytic method that allows the evaluation of lists of differentially expressed genes of interest against known biological modules, such as gene sets specific to pathways, processes, and profiles, of previous profiling experiments (21). To determine whether any specific pathway or profile is enriched in the four different gene lists, we tested 1,892 curated gene sets obtained through MsigDB, which is a constituent database of gene sets available through GSEA. Supplementary Table S3 shows the gene sets from MsigDb that are enriched in breast and ovarian samples. Figure 5A is a heat map of differentially expressed genes in both breast and ovary showing various gene sets that were identified to be enriched. We did not find any significant enrichment for canonical pathways. However, we found a variety of gene sets that are listed in Supplementary Table S3. Among these, we found two stem cell–related gene sets.

Because the gene expression profiles in BRCA1 and BRCA2 mutant cells have similarities to those of stem and progenitor cells, we tested the four gene lists from this study against a cohort of gene sets obtained from various studies including breast stem cells from both human and mouse and DNA repair genes (2224). Breast BRCA1 and BRCA2 cells show significant enrichment with the differentially expressed gene sets of transformed HMECs versus control HMEC cells and BPECs versus control BPEC population (Fig. 5B). We observed that downregulated genes from breast BRCA1 heterozygous cells show significant association with transformed HMEC and BPEC cells, suggesting that BRCA1 single-hit cells and transformed breast primary cells share a common fingerprint.

Finally, we compared our breast BRCA1 data set to the three data sets of Hedenfalk et al. (25) on sporadic breast cancers as well as breast cancers in families with BRCA1 and BRCA2 mutations. The GSEA comparison revealed that our breast epithelial BRCA1 data set is most similar to the BRCA1 tumors with matching hits corresponding to the following genes: KRT8, TGFB1, S100A2, S100P, EPHA2, TRIM29, OSBP2, FLNB, and MUC1. Figure 5B shows that genes from BRCA1 and BRCA2 mutant and sporadic tumors associated with breast BRCA1 heterozygous cells. These genes have already been implicated in breast cancer. For instance, decreased levels of KRT8 in cytoplasm are detected in breast cancer cells (38). TGFB1 is highly expressed in sporadic breast tumors or tumors from BRCA1 and BRCA2 kindreds, whereas it is downregulated in heterozygous BRCA1 breast cells. Calcium binding proteins A2 and P are downregulated across all tissues. Aberrant expression of S100A2 has been implicated in breast cancer. In all genes that are common to three tumor types (BRCA1, BRCA2, and sporadic) and heterozygous BRCA1 cells, expression of TGFB1 is observed to be different. This finding confirms our hypothesis of the earliest significant molecular changes in one-hit cells and their relationship with transformed breast cells.

Our study shows that mRNA expression profiles are altered in morphologically normal breast and ovarian epithelial cells heterozygous for mutation in BRCA1 or BRCA2, and include functionally critical genes. Remarkably, these single-hit cells bear significant transcriptomic changes that share features of the profiles of the corresponding cancer cells. It is well known that BRCA1 and BRCA2 mutant breast cancers exhibit distinct expression profiles (25), and the same is true for ovarian cancer (39). In the case of our single-hit breast and ovarian epithelial cell cultures, gene expression differences related to a given genotype clearly emerge when supervised methods are used (Tables 2 and 3), and they are reflected in separate clusters (Fig. 3). On the other hand, genome-wide unsupervised analyses using hierarchical clustering and nonnegative matrix factorization revealed clusters that differentiate tissue of origin but not genotype (see Supplementary Material and Supplementary Fig. S2).

Although these specific molecular changes are yet to be placed in the context of cancer initiation and progression, it should be noted that both BRCA proteins have clear functional links to transcription. Indeed, both are mediators of the cellular response to DNA damage that includes a transcriptional component (40, 41). Of course, damage does occur in normal cells as a consequence of physiologic DNA replication processes, although it is repaired with high efficiency (42), and BRCA1 and BRCA2 are part of a protein complex with RNA polymerase II and the CREB-binding protein and p300 histone acetyltransferases that is involved in chromatin remodeling and transcription (43). Because of these links to transcription (44), alterations in the levels of BRCA1 and BRCA2 proteins in single-hit cells might be expected to lead to multiple gene expression differences.

Intriguingly, some of the gene expression profiles enriched in breast BRCA1 one-hit cells are similar to those detected in stem and progenitor cells (Fig. 5A). This does not seem to be true for one-hit BRCA2 breast epithelial cells. Indeed, recent findings from the Wicha laboratory indicate that the BRCA1 gene is involved in regulating stemness and differentiation of breast progenitor cells (45). In addition, more recently, HMECs from BRCA1 mutation carriers were found to form progenitor cell colonies on semisolid medium with higher plating efficiency compared with mammary epithelial cells from reduction mammoplasty controls (46).

In general, we found more expression changes in BRCA1 versus WT cells than with BRCA2 versus WT cells (Table 2), which may reflect the fact that BRCA2 is primarily involved in double-strand break repair, whereas BRCA1 may also bridge double-strand break repair and signal transduction pathways. Thus, BRCA1 may act both as a sensor of DNA damage and as a repair factor, whereas BRCA2 is thought to be involved primarily in actual repair. Even small alterations in levels of the sensor (BRCA1) may have phenotypic consequences in terms of differentially expressed genes. This is reminiscent of animal models with hypomorphic alleles of the mismatch repair protein MSH2 (part of the damage sensor) that are proficient for mismatch repair per se but defective for activation of the cellular DNA damage response (47).

An additional important finding from this study is that some of the molecular changes detected correspond to candidate biomarkers described previously for breast and ovarian cancer, such as mammaglobin and serum amyloid protein for breast and ovarian cancer, respectively (Tables 2 and 3). Furthermore, GSEA analysis reveals that the data set of breast epithelial BRCA1 one-hit cells shows similarities to that of hereditary breast carcinomas associated with BRCA1 mutations (Fig. 5B).

In conclusion, the findings from this study are largely consistent with what is known about the pathophysiology of BRCA1 and BRCA2 and sporadic cancers. However, there are genes with abnormal patterns of expression that seem unrelated. Both their number and their unrelatedness are unexplainable based on what is known about BRCA1 and BRCA2 cancers, but they may be early changes associated indirectly with cancer initiation (7, 8). For example, heterozygosity may trigger a phenomenon such as induction of expression of small interfering RNAs, each of which might affect the expression of multiple genes.

In principle, therefore, the genetic approach used in this study may serve as a model for the identification of biomarkers for epithelial malignancies in general, and for the use of such markers as targets for chemoprevention measures that would decrease the probability of a second “hit” or greatly reduce its occurrence.

No potential conflicts of interest were disclosed.

We thank Dr. Jeff Boyd for the critical reading of the manuscript, R. Sonlin for the expert secretarial assistance, Fannie E. Rippel Biotechnology Facility, the Biomarker and Genotyping Facility, the Cell Culture Facility, and the Biosample Repository at the Fox Chase Cancer Center for the technical support.

Grant Support: NIH contracts N01 CN-95037, N01-CN-43309, the Ovarian Cancer Specialized Programs of Research Excellence at Fox Chase Cancer Center (P50 CA083638), NIH grant P30 CA06927, an appropriation from the Commonwealth of Pennsylvania to the Fox Chase Cancer Center, and the Eileen Stein Jacoby Fund.

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 18 U.S.C. Section 1734 solely to indicate this fact.

1
Fearon
ER
,
Vogelstein
B
. 
A genetic model for colorectal tumorigenesis
.
Cell
1990
;
61
:
759
67
.
2
Armitage
P
,
Doll
R
. 
The age distribution of cancer and a multistage theory of carcinogenesis
.
Br J Cancer
1954
;
8
:
1
12
.
3
Erickson
LC
. 
The role of O-6 methylguanine DNA methyltransferase (MGMT) in drug resistance and strategies for its inhibition
.
Semin Cancer Biol
1991
;
2
:
257
65
.
4
Danes
BS
. 
Increased in vitro tetraploidy: tissue specific within the heritable colorectal cancer syndromes with polyposis coli
.
Cancer
1978
;
41
:
2330
4
.
5
Kopelovich
L
. 
Phenotypic markers in human skin fibroblasts as possible diagnostic indices of hereditary adenomatosis of the colon and rectum
.
Cancer
1977
;
40
:
2534
41
.
6
Kopelovich
L
. 
Heritable colorectal cancer and cancer genes: systemic expressions
.
Mol Carcinog
1993
;
8
:
3
6
.
7
Yeung
AT
,
Patel
BB
,
Li
XM
, et al
. 
One-hit effects in cancer: altered proteome of morphologically normal colon crypts in familial adenomatous polyposis
.
Cancer Res
2008
;
68
:
7579
86
.
8
Kopelovich
L
. 
Genetic predisposition to cancer in man: in vitro studies
.
Int Rev Cytol
1982
;
77
:
63
88
.
9
Stoyanova
R
,
Clapper
ML
,
Bellacosa
A
, et al
. 
Altered gene expression in phenotypically normal renal cells from carriers of tumor suppressor gene mutations
.
Cancer Biol Ther
2004
;
3
:
1313
21
.
10
Futreal
PA
,
Liu
Q
,
Shattuck-Eidens
D
, et al
. 
BRCA1 mutations in primary breast and ovarian carcinomas
.
Science
1994
;
266
:
120
2
.
11
Miki
Y
,
Swensen
J
,
Shattuck-Eidens
D
, et al
. 
A strong candidate for the breast and ovarian cancer susceptibility gene BRCA1
.
Science
1994
;
266
:
66
71
.
12
Tavtigian
SV
,
Simard
J
,
Rommens
J
, et al
. 
The complete BRCA2 gene and mutations in chromosome 13q-linked kindreds
.
Nat Genet
1996
;
12
:
333
7
.
13
Wooster
R
,
Bignell
G
,
Lancaster
J
, et al
. 
Identification of the breast cancer susceptibility gene BRCA2
.
Nature
1995
;
378
:
789
92
.
14
Soule
HD
,
Maloney
TM
,
Wolman
SR
, et al
. 
Isolation and characterization of a spontaneously immortalized human breast epithelial cell line, MCF-10
.
Cancer Res
1990
;
50
:
6075
86
.
15
Caretti
E
,
Devarajan
K
,
Coudry
R
, et al
. 
Comparison of RNA amplification methods and chip platforms for microarray analysis of samples processed by laser capture microdissection
.
J Cell Biochem
2008
;
103
:
556
63
.
16
Irizarry
RA
,
Bolstad
BM
,
Collin
F
,
Cope
LM
,
Hobbs
B
,
Speed
TP
. 
Summaries of Affymetrix GeneChip probe level data
.
Nucleic Acids Res
2003
;
31
:
e15
.
17
Irizarry
RA
,
Hobbs
B
,
Collin
F
, et al
. 
Exploration, normalization, and summaries of high density oligonucleotide array probe level data
.
Biostatistics
2003
;
4
:
249
64
.
18
Jain
N
,
Thatte
J
,
Braciale
T
,
Ley
K
,
O'Connell
M
,
Lee
JK
. 
Local-pooled-error test for identifying differentially expressed genes with a small number of replicated microarrays
.
Bioinformatics
2003
;
19
:
1945
51
.
19
Benjamini
Y
,
Hochberg
Y
. 
Controlling the false discovery rate: a practical and powerful approach to multiple testing
.
J Royal Stat Soc B
1995
;
57
:
289
300
.
20
Pascual-Montano
A
,
Carmona-Saez
P
,
Chagoyen
M
,
Tirado
F
,
Carazo
JM
,
Pascual-Marqui
RD
. 
bioNMF: a versatile tool for non-negative matrix factorization in biology
.
BMC Bioinformatics
2006
;
7
:
366
74
.
21
Subramanian
A
,
Tamayo
P
,
Mootha
VK
, et al
. 
Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles
.
Proc Natl Acad Sci U S A
2005
;
102
:
15545
50
.
22
Behbod
F
,
Xian
W
,
Shaw
CA
,
Hilsenbeck
SG
,
Tsimelzon
A
,
Rosen
JM
. 
Transcriptional profiling of mammary gland side population cells
.
Stem Cells
2006
;
24
:
1065
74
.
23
Ince
TA
,
Richardson
AL
,
Bell
GW
, et al
. 
Transformation of different human breast epithelial cell types leads to distinct tumor phenotypes
.
Cancer Cell
2007
;
12
:
160
70
.
24
Cho
RW
,
Wang
X
,
Diehn
M
, et al
. 
Isolation and molecular characterization of cancer stem cells in MMTV-Wnt-1 murine breast tumors
.
Stem Cells
2008
;
26
:
364
71
.
25
Hedenfalk
I
,
Duggan
D
,
Chen
Y
, et al
. 
Gene-expression profiles in hereditary breast cancer
.
N Engl J Med
2001
;
344
:
539
48
.
26
Bernstein
JL
,
Godbold
JH
,
Raptis
G
, et al
. 
Identification of mammaglobin as a novel serum marker for breast cancer
.
Clin Cancer Res
2005
;
11
:
6528
35
.
27
Ntoulia
M
,
Stathopoulou
A
,
Ignatiadis
M
, et al
. 
Detection of Mammaglobin A-mRNA-positive circulating tumor cells in peripheral blood of patients with operable breast cancer with nested RT-PCR
.
Clin Biochem
2006
;
39
:
879
87
.
28
Junker
N
,
Johansen
JS
,
Hansen
LT
,
Lund
EL
,
Kristjansen
PE
. 
Regulation of YKL-40 expression during genotoxic or microenvironmental stress in human glioblastoma cells
.
Cancer Sci
2005
;
96
:
183
90
.
29
Li
X
,
Cao
X
,
Zhang
W
,
Feng
Y
. 
Expression level of insulin-like growth factor binding protein 5 mRNA is a prognostic factor for breast cancer
.
Cancer Sci
2007
;
98
:
1592
6
.
30
Lo
SH
,
Lo
TB
. 
Cten, a COOH-terminal tensin-like protein with prostate restricted expression, is down-regulated in prostate cancer
.
Cancer Res
2002
;
62
:
4217
21
.
31
von Mensdorff-Pouilly
S
,
Snijdewint
FG
,
Verstraeten
AA
,
Verheijen
RH
,
Kenemans
P
. 
Human MUC1 mucin: a multifaceted glycoprotein
.
Int J Biol Markers
2000
;
15
:
343
56
.
32
Khan
N
,
Cromer
CJ
,
Campa
M
,
Patz
EF
 Jr
. 
Clinical utility of serum amyloid A and macrophage migration inhibitory factor as serum biomarkers for the detection of nonsmall cell lung carcinoma
.
Cancer
2004
;
101
:
379
84
.
33
Moshkovskii
SA
,
Serebryakova
MV
,
Kuteykin-Teplyakov
KB
, et al
. 
Ovarian cancer marker of 11.7 kDa detected by proteomics is a serum amyloid A1
.
Proteomics
2005
;
5
:
3790
7
.
34
Wenham
RM
,
Calingaert
B
,
Ali
S
, et al
. 
Matrix metalloproteinase-1 gene promoter polymorphism and risk of ovarian cancer
.
J Soc Gynecol Investig
2003
;
10
:
381
7
.
35
Kenny
HA
,
Kaur
S
,
Coussens
LM
,
Lengyel
E
. 
The initial steps of ovarian cancer cell metastasis are mediated by MMP-2 cleavage of vitronectin and fibronectin
.
J Clin Invest
2008
;
118
:
1367
79
.
36
Daikoku
T
,
Wang
D
,
Tranguch
S
, et al
. 
Cyclooxygenase-1 is a potential target for prevention and treatment of ovarian epithelial cancer
.
Cancer Res
2005
;
65
:
3735
44
.
37
Gupta
RA
,
Tejada
LV
,
Tong
BJ
, et al
. 
Cyclooxygenase-1 is overexpressed and promotes angiogenic growth factor production in ovarian cancer
.
Cancer Res
2003
;
63
:
906
11
.
38
Iwaya
K
,
Ogawa
H
,
Mukai
Y
,
Iwamatsu
A
,
Mukai
K
. 
Ubiquitin-immunoreactive degradation products of cytokeratin 8/18 correlate with aggressive breast cancer
.
Cancer Sci
2003
;
94
:
864
70
.
39
Jazaeri
AA
,
Yee
CJ
,
Sotiriou
C
,
Brantley
KR
,
Boyd
J
,
Liu
ET
. 
Gene expression profiles of BRCA1-linked, BRCA2-linked, and sporadic ovarian cancers
.
J Natl Cancer Inst
2002
;
94
:
990
1000
.
40
Harper
JW
,
Elledge
SJ
. 
The DNA damage response: ten years after
.
Mol Cell
2007
;
28
:
739
45
.
41
Zhou
BB
,
Elledge
SJ
. 
The DNA damage response: putting checkpoints in perspective
.
Nature
2000
;
408
:
433
9
.
42
Vilenchik
MM
,
Knudson
AG
. 
Endogenous DNA double-strand breaks: production, fidelity of repair, and induction of cancer
.
Proc Natl Acad Sci U S A
2003
;
100
:
12871
6
.
43
Welcsh
PL
,
Owens
KN
,
King
MC
. 
Insights into the functions of BRCA1 and BRCA2
.
Trends Genet
2000
;
16
:
69
74
.
44
Yoshida
K
,
Miki
Y
. 
Role of BRCA1 and BRCA2 as regulators of DNA repair, transcription, and cell cycle in response to DNA damage
.
Cancer Sci
2004
;
95
:
866
71
.
45
Liu
S
,
Ginestier
C
,
Charafe-Jauffret
E
, et al
. 
BRCA1 regulates human mammary stem/progenitor cell fate
.
Proc Natl Acad Sci U S A
2008
;
105
:
1680
5
.
46
Burga
LN
,
Tung
NM
,
Troyan
SL
, et al
. 
Altered proliferation and differentiation properties of primary mammary epithelial cells from BRCA1 mutation carriers
.
Cancer Res
2009
;
69
:
1273
8
.
47
Claij
N
,
Te Riele
H
. 
Methylation tolerance in mismatch repair proficient cells with low MSH2 protein level
.
Oncogene
2002
;
21
:
2873
9
.