|Home | About | Journals | Submit | Contact Us | Français|
Protein phosphatase 2A (PP2A) is a major cellular phosphatase and plays key regulatory roles in growth, differentiation, and apoptosis. Women diagnosed with benign proliferative breast disease are at increased risk for the subsequent development of breast cancer.
We evaluated genetic variation of PP2A holoenzyme subunits for potential contribution to breast cancer risk. We performed a nested case-control investigation of a cohort of women with a history of benign breast disease. Subjects were followed for an average of 18 years; DNA prepared from the original archival benign breast biopsy (1954 – 1995) was available for 450 women diagnosed with breast cancer on follow-up, and for 890 of their 900 controls who were matched on race, age, and year of entry biopsy.
Single allele- and haplotype-based tests of association were conducted, with assessment of significance by permutation testing. We identified significant risk and protective haplotypes of PPP2R1A, giving odds ratios of 1.63 (95% CI 1.3 – 2.1) and 0.55 (95% CI 0.41 – 0.76), respectively. These odds ratios remained significant upon adjustment for multiple comparisons. Women with both the risk PPP2R1A haplotype and a history of proliferative breast disease had an odds ratio of 2.44 (95% CI 1.7 – 3.5) for the subsequent development of breast cancer. The effects of haplotypes for two regulatory subunit genes, PPP2R2A and PPP2R5E on breast cancer risk were nominally significant, but did not remain significant upon adjustment for multiple comparisons.
This evidence supports the previously hypothesized role of PP2A as a tumor suppressor gene in breast cancer.
A large number of genes encode components of the complex mitogen-activated protein kinase (MAPK) signaling network and many are rational candidates that may contribute to breast cancer risk. A breast cancer risk variant of the MAP3K1 gene was recently identified by genome-wide association study.1 However, perturbation of other select nodes of the network are predicted by systems biology modeling to have the greatest influence on signal transduction. Prominent among these candidates is protein phosphatase 2A (PP2A).2 PP2A is a major serine/threonine phosphatase in most tissues. The PP2A holoenzyme is a heterotrimer comprised of a structural, catalytic, and regulatory subunit, each with multiple isoforms that are encoded by at least 16 known genes (Figure 1).3 This combinatorial diversity provides regulatory specificity for a very wide array of cellular proteins, including MAPKs.4 PP2A is in turn inactivated by both receptor and non-receptor protein tyrosine kinases, such as EGFR and Src.5 Given the wide array of cellular functions for PP2A, its activity is tightly controlled to maintain cellular homeostasis. PP2A has been remarkably conserved during evolution and knockouts of subunit genes in model organisms prove embryonic lethal.6–9 Both catalytic subunit genes PPP2CA and PPP2CB, the major structural subunit PPP2R1A gene, and regulatory subunit genes PPP2R2A and PPP2R5E each encode proteins with 100% amino acid identity between mouse and man.
Collectively, significant diverse lines of evidence each support the hypothesis that PP2A can function as a tumor suppressor.10, 11 The tumor promoter effects of okadaic acid (a red tide shellfish toxin produced by marine dinoflagellates) suggest a potential tumor suppressor role of PP2A.4, 12 Inactivating somatic mutations of both structural subunit genes, PPP2R1A and PPP2R1B, have been identified in breast cancer, lung cancer, melanoma, and colon cancer.13–16 Without them, regulatory subunit proteins may also be degraded despite abundant transcripts.17 Loss of PPP2R1A protein expression has also been observed in breast cancer.18 Further, the transforming virus SV40 small T antigen acts by inhibiting the function of PP2A regulatory subunits to spur transformation.19
Here we report our initial investigation of a set of candidate genes of growth factor signal transduction pathways (see Table 1) to address hypotheses of genetic variation influencing breast cancer risk, and to establish feasibility of high-throughput assay of archival DNA. We evaluated each gene by the SNP tagging approach to systematically capture genetic diversity by virtue of patterns of linkage disequilibrium (LD) established by the HapMap.20 We investigated women of the Nashville Breast Cohort, a study population of women who have undergone breast biopsies for benign breast disease, each followed for an average of over 18 years for breast cancer outcome. Subsequent breast cancer risk for these women has ranged from no increased risk for those without proliferative disease, to a five-fold increased risk for those with proliferative disease with atypia.21–25 Benign proliferative breast lesions, characterized by abnormal epithelial proliferation, are generally considered a non-obligate precursor to invasive breast cancer.23 Since women with these lesions may already have taken a first step towards breast cancer, they are a particularly salient group in which to search for genetic factors that participate in this process. Cancer risk in these women could be modified by genetic factors that influence the initiation of proliferative breast disease, or that influence their progression to overt cancer. The formalin-fixed, paraffin-embedded (FFPE) benign tissue blocks of each subject’s study entry biopsy (ranging from greater than one to six decades old) were available for genetic evaluation. Here we report that specific population haplotypes of genes encoding components of PP2A are associated with breast cancer and interact with benign proliferative breast disease to modify the risk.
The Nashville Breast Cohort (NBC) is an ongoing retrospective cohort study of 16,946 women who underwent a breast biopsy revealing benign parenchyma or fibroadenoma at Vanderbilt, St. Thomas, and Baptist Hospitals in Nashville, Tennessee since 1954.21, 25 Subjects provided written informed consent under approved institutional review board protocols. To be eligible for inclusion in this cohort a woman could not have had a diagnosis of breast cancer prior to her entry biopsy. Additional details on the NBC are given elsewhere.21, 25 Subjects were followed by telephone interviews or, if deceased, with their next of kin, through medical record reviews, and through searches of the National Death Index and the Tennessee Cancer Registry. Successful follow-up has been obtained on 90% of the women who met the entry criteria for the NBC and who were biopsied before 1990. There were 7038 women among this group whose entry biopsy FFPE blocks were available and who were eligible for this study, 452 of whom had developed breast cancer on follow-up and were Caucasian. We performed a nested case-control study of Caucasian women from this sub-cohort. We selected two controls for each case from the risk set of those who had not been diagnosed with breast cancer by the follow-up time when their case developed this disease. These controls were selected without replacement. Controls were matched to cases by age, race, and year of entry biopsy. Successful DNA extractions from benign archival entry biopsy specimens were performed for 450 out of 452 (99.6%) Caucasian cases and for 890 out of 900 (99%) of their matched controls. The study included a total of 440 trios (1 case: 2 controls) and 10 pairs (1 case: 1 control). Characteristics of these subjects are detailed in Table 2.
DNA samples were extracted from sixteen 5-µm sections of FFPE tissue blocks for each study subject. Our specification for successful extraction was a yield of 50 ng/µl by either of two quantification methods described below. DNA extraction was done by standard paraffin removal, proteinase K digestion, phenol/chloroform extraction, and ethanol precipitation. Where the initial biopsy yielded inadequate DNA, a second replicate extraction was performed and extracted DNA from both rounds was combined. If this too was inadequate, one of the subject’s archival blocks was extracted in total, while ensuring retained blocks and peels of the subject for future use. DNA samples were quantified by PicoGreen assay (Invitrogen, Carlsbad, CA) on a Molecular Devices/LJL Analyst HT (Molecular Devices, Sunnyvale, CA), and by quantitative real-time PCR (qRT-PCR) on an ABI 7900HT (Applied Biosystems, Foster City, CA). The date of the original archival FFPE biopsy block (benign tissue) employed for DNA extraction was generally quite old: 15.0% were from the 1950s; 31.5% were from the 1960s; 36.1% were from the 1970s; 14.6% were from the 1980s; and 2.7% were from the 1990’s (1995 the most recent). An important aspect of our study design was that DNA samples for study also included a total of 67 living subjects for whom both this original archival FFPE benign tissue block (dating from 1963 to 1994) and saliva DNA samples were genotyped. The saliva DNA enabled accuracy estimation for FFPE DNA genotyping. Reference genomic DNA from saliva was prepared using OraGene kits (DNAGenoTek, Ontario, Canada).
We employed the Illumina GoldenGate™ assay (Illumina, San Diego, CA) for genotyping, using five microliters (≥ 250 ng) of each extracted archival DNA. Each 96-well plate of DNA genotyped contained an average of 4.2 (range 1 to 6) reference saliva DNA samples of study subjects for whom DNA from matching blocks was under evaluation. This enabled assessment of genotype accuracy. A total of 242 informative tagging SNPs were successfully genotyped with a completion rate of 99.8%. Each polymorphism was in Hardy-Weinberg equilibrium among controls. The concordance rate for subject saliva DNA – FFPE DNA pairs was 99.8%.
Drs. Page and Sanders conducted the entry biopsy slide review using criteria of the Cancer Committee of the College of American Pathologists, without knowledge of subsequent cancer outcome.26–29 Independent agreement on histologic diagnoses was achieved in 95% of biopsies. Consensus was reached for discrepant diagnoses. Entry biopsy slides were classified into one of three broad categories: proliferative disease with atypia (atypical hyperplasia), proliferative disease without atypia, and no proliferative disease. Dr Page reviewed the histologic slides and surgical pathology records of subsequent cancer for all cases to confirm the diagnosis.
Tagging SNPs were selected from genotype data of HapMap CEU subjects 20 using LDSelect 30 with a minor allele frequency threshold of 0.05 and an r2 threshold of 0.8. Calculation of Hardy-Weinberg equilibrium and pairwise LD was accomplished using Haploview v3.2.31 Individual study subject diplotypes were estimated by PHASE v2.1 32–34, run for all subjects as a group. For a given window of N consecutive SNPs at a gene, the diplotype may be assigned with evaluation of the specified window or with evaluation of larger windows encompassing the same markers. The window diplotype of highest probability ≥ 0.9 among them was employed for subsequent tests of association.
A sliding window approach tested a haplotype window of N markers, sliding the window along the map in single marker increments.35–37 Each N-marker haplotype of at least 5% frequency was compared to the remaining haplotypes of the window as a group among cases and controls. The resulting 2 × 2 contingency table was evaluated by a χ2 test statistic. The most significant risk and protective haplotypes identified in contingency tests were subsequently modeled by conditional logistic regression employing the matched study design.
Conditional logistic regression analyses were used to estimate breast cancer odds ratios (OR) and 95% confidence intervals (CI) under additive, dominant, and recessive models (Stata Release 10, Stata Corporation, College Station, TX). For additive models indicator covariates were used to separately model the risk of women who were heterozygous or homozygous for the haplotype under consideration. Models that assessed the joint effects of proliferative disease and haplotype contained cross-product terms of the indicator covariates that allowed breast cancer risk to be affected by the interactions between histologic and genetic variables. These interaction terms avoided an assumption that the effects of haplotype and proliferative disease on breast cancer risk are multiplicative. Regression models included age at biopsy and year of biopsy as covariates to adjust for residual confounding. Models additionally adjusting for body mass index, menopausal status, parity, and first-degree family history of breast cancer yielded similar results (not presented). In this study, 86% of cases had an invasive malignant lesion and 14% had ductal carcinoma in situ. Analysis of data restricted to cases with invasive cancer (not presented) yielded results that were similar to those with all cancers combined.
Our search for haplotypes that affect breast cancer risk involved the evaluation of many haplotypes in numerous haplotype windows, as well as the evaluation of single alleles. We explored 29 genes by the tagging SNP approach with sliding window haplotype-based analysis. We used study-wide permutation testing to assess the extent to which results could be explained as multiple comparisons artifacts. We generated 500 copies of the study-wide data set in which pseudo case status was permuted among matched case-control trios and pairs. In order to derive the haplotype or allele at each gene most strongly associated with the pseudo case status, we used the same search algorithm that was used for the real data in each of these simulated data sets. A z value for this haplotype or allele was calculated in the same way as for the real data. Since the null hypothesis is true for each randomized subject set, the proportion of simulated z values whose absolute value is greater than the absolute value of the real z value was used as a P value for the breast cancer association, adjusted study-wide for multiple comparisons.
A set of candidate genes of growth factor signal transduction pathways with hypothesized roles in breast cancer risk were evaluated in this study (Table 1). For each of these genes, we evaluated tagging SNPs, 96% of which (242 of 251) were informative within the study population.
Only one of the 29 genes remained significant after correction for multiple testing, detailed further below. This gene was PPP2R1A, one of several genes encoding subunits of the PP2A holoenzyme that were evaluated. Eight tagging SNPs within PPP2R1A, PPP2R2A, and PPP2R5E that were nominally associated with breast cancer risk in single allele analyses are given in Table 3. However, the principal analytic strategy for tagged genes was haplotype-based. The haplotype combines information of genotyped SNPs to provide a more powerful surrogate for untested variation than that afforded by the individual alleles.32, 35, 38–41 Each of our tagged candidate genes may be evaluated at windows of adjacent SNPs; these include windows of sizes 2, 3, 4,… up to the total number of SNPs evaluated at a given gene. The sliding window analytic approach systematically evaluates windows of adjacent SNPs, moving the window along the gene map in single SNP increments. This analytic approach is computationally demanding and requires correction for statistical over-fitting given the large number of correlated multiple comparisons.
Figure 2 illustrates the results of sliding window analyses for the tagged PPP2R1A, PPP2R2A, and PPP2R5E genes. As our initial screen, we performed simple contingency tests of a given haplotype at a window of consecutive SNPs, relative to the remaining haplotypes as a group. A total of 1,843 windows were evaluated at the 29 tagged genes, including 105 at PPP2R1A, 78 at PPP2R2A, and 210 at PPP2R5E. Haplotypes nominally significantly associated with breast cancer were observed at these genes, often redundantly within numerous windows. At PPP2R1A, haplotypes at 6 windows were consistent with risk haplotype A of Figure 2 and were significant with a P ≤ 0.05. PPP2R1A protective haplotype B was identified by 46 windows with P ≤ 0.05. At PPP2R2A the only nominally significant test was observed at rs7838409. PPP2R5E risk haplotype E was identified by 13 windows and protective haplotype F was observed at a single window, each window with P ≤ 0.05.
We evaluated additive, dominant, and recessive genetic models for each of the haplotypes designated A – F (red font) in Figure 2. Haplotypes A – D proved nominally significantly associated with breast cancer in these conditional logistic regression models, the most significant of which is detailed for each haplotype in Table 4. Particularly noteworthy were the risk and protective haplotypes at PPP2R1A, which had significant estimated odds ratios of 1.63 and 0.555, respectively (both with P = 0.0002).
In total, 5,529 haplotype tests were done across the candidate genes. At a gene evaluated by the sliding window approach many of the tests were highly correlated. Our estimation of significance for each of the haplotypes of interest was therefore adjusted through permutation testing. We randomized case/control status within the matched sets of subjects (trios or pairs), conducted the sliding window analysis to identify the most significant test at the gene among the trait-randomized subjects, fitted conditional logistic regression models, and derived a z value from the model most associated with the pseudo-case status. This resulted in 29×3 = 87 z values for each permuted data set. After 500 such permutations of the entire study-wide data set, the proportion of the 43,500 z values more extreme than that observed in the actual data provided an estimate of significance that is adjusted for multiple comparisons. Under this approach, only PPP2R1A haplotypes A and B remained significant, permuted P = 0.003 for both haplotypes.
The women of the Nashville Breast Cohort have each undergone biopsy for benign breast disease. On histologic evaluation, a subset of these women had epithelial hyperplasia without, or with, atypia. Each of these forms of proliferative breast disease carries an increased risk for the subsequent development of breast cancer (two- and five-fold, respectively). We investigated the hypothesis that each of the salient haplotypes of PPP2R1A, PPP2R2A, and PPP2R5E might affect the potential of pre-malignancy to influence breast cancer risk. Results of conditional logistic regression models evaluating interaction with proliferative breast disease are shown in Table 5. The overall breast cancer odds ratio associated with proliferative breast disease was 1.61 (95% CI 1.3 – 2.1, P = 0.0002). Noteworthy interactions were observed between proliferative disease and haplotypes for PPP2R1A and PPP2R2A. Risk haplotypes for these two genes augmented the risk of breast cancer for women with proliferative breast disease, while protective haplotypes nullified the risk of proliferative breast disease. Similar, although less significant, effects were seen for PPP2R5E. These results suggest that the effect of these haplotypes on breast cancer risk is not solely due to their influence on the risk of developing proliferative disease.
The results of our study implicate two genes encoding subunits of PP2A as modifiers of breast cancer risk. This enzyme is the principal phosphatase counterbalancing cellular kinases of signaling pathways regulating cellular growth, differentiation, and apoptosis. The results of our investigation are consistent with multiple independent lines of investigation that suggest a key role for the enzyme in the development of cancer. These include viral- and toxin-induced transformation of cell culture systems, and somatic mutation of genes encoding subunits of PP2A in multiple tumor types.10 Kinetic modeling of the MAPK signaling network predicts that perturbation of the PP2A node is among those that would most alter signaling.2 Components of the MAPK signaling network are believed to play a role in the initiation of breast cancer. The MAP2K4 and FGFR2 genes were among the most significant genes detected by two recently published genome-wide scans of breast cancer.1, 42 There is substantial a priori reason to consider population variation of genes encoding subunits of PP2A as potentially important modifiers of cancer risk.
We observe that genetic variation of PPP2R1A, PPP2R2A, and possibly PPP2R5E modifies risk of breast cancer among women of the Nashville Breast Cohort. Cohort members have a history of benign breast disease; those with a history of benign proliferative breast disease have an elevated breast cancer risk.21–25, 43 These lesions are characterized by abnormal epithelial proliferation, potentially an initial step in the progression to breast cancer. The observation that the breast cancer risk associated with proliferative breast disease appears to be exacerbated in women with risk haplotypes and negated in women with protective haplotypes may have clinical importance in terms of breast cancer surveillance screening.
Our permutation tests revealed anticipated evidence of statistical over-fitting in the sliding window analytic approach, and appropriately adjusted for multiple comparisons. Among nominally significant haplotypes at these genes, only haplotypes A and B of PPP2R1A were significant after adjustment for multiple testing. Other nominally significant haplotypes remain credible candidates for further evaluation as modifiers of breast cancer risk, particularly in light of their interaction with the established breast cancer risk factor, proliferative breast disease (Table 5). These haplotypes were detected based upon a tagging strategy. Potential underlying functional variants remain unknown, though two candidate missense SNPs at PPP2R1A are presently annotated within dbSNP. Studies to investigate additional PP2A holoenzyme subunit genes by as modifiers of breast cancer risk appear warranted and may reveal important epistatic effects. This study is the first of its type to be conducted within a cohort of women with a history of benign breast disease; replication within similar cohorts is planned. The CGEMS GWAS, which was not conducted in such a high-risk study population42, did not observe nominally significant single allele analyses at PPP2R1A, PPP2R2A, or PPP2R5E at alternative sets of tagging SNPs. We were unable, however, attribute our observations at PPP2R1A to an artifact of multiple comparisons.
To our knowledge, this study is the first to employ a high-throughput genotyping technology for a genetic association study of truly archival tissue specimens. The level of data completeness (99.8%) and accuracy (observed 100%) achieved in this study should encourage other investigators similarly considering the use of FFPE tissue as a DNA source for investigation, even when specimens are up to six decades old. The median age of archival specimens in this study was 35 years.
The role of PP2A in maintenance and progression of breast cancer and potential promise of therapies targeting PP2A within tumors is also an intriguing subject. Somatic mutation of the structural subunit genes has been observed in several tumor types, including breast cancer.13, 14 Partial insufficiency of PPP2R1A can transform immortalized cells and spur tumorigenicity. Without heterotrimer complex formation, other PP2A subunits (including those encoded by PPP2R2A and PPP2R5E) are degraded despite their persistent expression.17 Pharmacologic enhancement of specific PP2A holoenzyme activities may thus be a therapeutic strategy relevant in breast cancer. We note that the small molecule tyrosine kinase inhibitors dasatinib and imatinib both inhibit the abl kinase, which precludes activation of the PP2A inhibitory protein SET, in turn restoring PP2A activity to check growth.44 Dasatinib has been demonstrated to selectively inhibit growth of basal type/“triple-negative” breast cancer cell lines.45 The numerous independent lines of evidence implicating PP2A in breast cancer suggest that PP2A may be a rational target for breast cancer therapeutics.
In summary, we observe evidence that haplotypes of genes encoding PP2A subunits affect breast cancer risk. Permutation tests revealed a significant effect after multiple comparisons adjustment for haplotypes of the PPP2R1A gene. As with any exploratory study, these findings need to be confirmed in subsequent independent studies. Further study of PP2A genetic variation has the potential to lead to important advances in our understanding of breast cancer and its treatment.
Grant support: National Cancer Institute grants 1P50 CA098131-01, R01 CA050468 and P30 CA068485, an award from the V Foundation, by a MERIT grant from the US Department of Veterans Affairs.
There are no financial disclosures.