Altered PPP2R2A and Cyclin D1 expression defines a subgroup of aggressive luminal-like breast cancer

Background PPP2R2A deletions were recently linked to a subgroup of luminal breast carcinoma (BC) that exhibits poor survival. This subgroup also exhibited amplification of a chromosome region containing the Cyclin D1 coding gene, CCND1. Therefore, we aimed to investigate whether a combination of PPP2R2A (B55α) and Cyclin D1 expression statuses evaluated by immunohistochemistry (IHC) could define a subgroup of luminal BC that exhibits poor survival. Methods First we conducted a retrospective cohort study using sequencing data from The Cancer Genome Atlas initiative to correlate PPP2R2A copy number alteration (CNA) status with its expression level and the corresponding overall survival (OS). Next, also using a retrospective cohort study design, we evaluated the PPP2R2A (B55α) expression levels by IHC in a total of 807 BC patients from two independent cohorts (discovery cohort n = 349 and validation cohort n = 458). Cyclin D1 expression was also evaluated, and the PPP2R2A (B55α)-/low/Cyclin D1high phenotype was evaluated as a predictor of disease-free survival (DFS) and OS in luminal-like BC patients. Results Deletions in the PPP2R2A gene strongly correlate with lower mRNA expression and poorer OS. PPP2R2A (B55α)-/low carcinomas have significantly shorter DFS and OS. Furthermore, in univariate analysis, the PPP2R2A (B55α)-/low/Cyclin D1high phenotype is significantly associated with poorer DFS and OS. In a multivariate analysis, the PPP2R2A (B55α)-/low/Cyclin D1high phenotype is significantly associated with poor DFS, thus defining a group of luminal-like BC with higher risk of relapse. Conclusion We demonstrate that BCs harboring PPP2R2A deletions are associated with worse OS. Moreover, this is the first study to demonstrate that the combination of altered PPP2R2A (B55α) and high Cyclin D1 expression by IHC defines a subgroup of luminal-like BC patients with a high risk of relapse and death. Electronic supplementary material The online version of this article (doi:10.1186/s12885-015-1266-1) contains supplementary material, which is available to authorized users.

Background PP2A (Protein Phosphatase 2) is a major heterotrimeric serine/threonine phosphatase that consists of one structural (PP2A/A), one catalytic (PP2A/C) and one regulatory (PP2A/B) subunit [1]. The PP2A/A and PP2A/C subunits each have two known isoforms (α, β) and comprise the core dimer (reviewed in [2]). Conversely, the PP2A/B subunit has multiple isoforms, and it is the major determinant of substrate specificity [3]. As a Ser/Thr phosphatase, PP2A counteracts the actions of Ser/Thr kinases, which are often defective or deregulated in cancer. Examples include components of the MAP kinase and AKT pathways, the tumor suppressors pRB and p53, and CDK1 substrates (reviewed in [2]).
Substantial evidence points to a role for PP2A as a tumor suppressor. Early studies showed that toxinmediated inhibition of PP2A was a potent tumorpromoter stimulus and that PP2A had a pivotal role in modulating the transforming ability of oncogenic viruses [4,5]. More recent reports have demonstrated that PP2A is inhibited in colorectal carcinoma [6]. Such inhibition is due, at least partially, to the downregulation of some of PP2A regulatory subunits encoded by the PPP2R5E and PPP2R2A genes [6]. Alterations in PPP2R5E have also recently been shown to be a driver of clonal evolution in breast carcinoma [7]. Specifically, PPP2R2A is located on chromosome 8p21.2 and encodes B55α, a B regulatory subunit of PP2A. Loss of PPP2R2A is a common event in non-small cell lung cancer, and it results in the impairment of the homologous recombination repair pathway [8]. Additionally, shRNA-mediated silencing of PPP2R2A has been shown to be a potent oncogenic signal in experimental models of colon carcinoma [8]. Furthermore, in patients with acute myeloid leukemia (AML), low levels of B55α expression in blast cells have been associated with shorter complete remission duration, demonstrating that in AML, the decreased expression of PPP2R2A (B55α) is an adverse prognostic factor [9]. Taken together, these findings lead to the conclusion that PP2A, as well as some of its regulatory subunits, such as PPP2R2A, have definitive roles as tumor suppressors in a large variety of cancers.
In breast carcinoma (BC), the expression status and genomic alterations of PP2A and it subunits, such as PPP2R2A, are only now being elucidated. Despite recent reports of the possible therapeutic targeting of PP2A with inhibitors in BC [10], knowledge is still limited. Different haplotypes of PPP2R2A modify BC risk [11], but mutations in PPP2R2A are rarely detected in BC (The Cancer Genome Atlas Research Network BRCA dataset, provisional). More interestingly, in a recent study, copy number aberrations (CNAs) centered on PPP2R2A, along with heterozygous and homozygous deletions were commonly found in BC, and they were associated with a mitotically active estrogen receptor (ER)-positive BC subgroup [12]. The same study also identified an ER-positive BC subgroup, comprising 11q13/14 cis-acting luminal tumors that exhibited a steep mortality trajectory with elevated hazard ratios (Integrative Custer 2) [12]. The CNAs within this subgroup also contain the amplification of a region at 11q13 that harbors the Cyclin D1-encoding gene CCND1 [12]. Nevertheless, rapid translation of these findings to a clinical setting has been limited because of the need to use genomic and transcriptomic profiling, technologies that are not routinely available in most pathology laboratories.
Here, we test the hypothesis that PPP2R2A protein expression, detected using immunohistochemistry (IHC), will uncover a group of ER-positive BCs that exhibit a differential outcome. Additionally, we hypothesized that the combined assessment of PPP2R2A and Cyclin D1 protein expression would further refine the subgroups' outcome associations and be potentially useful in identifying Integrative Cluster 2-like BC.

Methods
The Cancer Genome Atlas (TCGA) data Data from the Invasive Breast Cancer dataset (TCGA, Provisional) of the TCGA database on PPP2R2A mRNA expression levels as determined by RNA sequencing (RNA Seq V2 RSEM -Illumina HiSeq 2000 platform), copy number alteration (CNA) as determined by GISTIC 2.0 analysis, recurrence status and overall survival (OS) were downloaded from the cBIO Cancer Genomics Portal on March 24, 2014 [13,14]. Survival analysis was performed by comparing patients with no CNAs with those with heterozygous or homozygous deletions.

Patient samples
Additionally we used the two following independent cohorts of clinical samples: (a) a discovery cohort comprised of 349 non-consecutive, randomly selected patients with BC who had undergone surgery at Centro Hospitalar de São João, EPE, Porto, Portugal, between January 2001 and March 2006; (b) a validation cohort comprised of 458 non-consecutive, randomly selected patients with BC who had undergone surgery between 1978 and 1992 at the Hospital Xeral-Cíes, Vigo, Spain. Only patients with histologically confirmed BC from which adequate formalin-fixed, paraffin-embedded tissue was available were included. The tumor grade was determined according to the modified Bloom and Richardson classification [15]. The exclusion criteria were similar for both cohorts and included the following requirements: previous history of BC, hereditary BC, bilateral BC at diagnosis, refusal of surgical treatment, synchronous tumors and cases with neoadjuvant therapy as the initial treatment. All estrogen receptor (ER)-negative and lymph node (LN)-positive patients in the discovery cohort had received adjuvant chemotherapy. None of the HER2+ patients had received trastuzumab as the initial treatment. The clinical definitions used for OS and Disease-Free Survival (DFS) are detailed in the Additional file 1. Women in whom the envisaged endpoint was not reached were censored as of the last follow-up date. Subjects with missing values for survival and PPP2R2A (B55α) expression in both cohorts were excluded from the analyses. A diagram of the complete analytical strategy and the flow of patients through the study, including the numbers of patients included in each stage of the analysis, are shown in Figure 1.
This study was conducted according to the Portuguese National Regulative Law for the handling of biological specimens from tumor banks (waiving the need of specify approval for samples from a tumor archive and used in retrospective analysis) and complied with The Helsinki Declaration statement.

Tissue Microarrays (TMAs) and IHC
Representative areas of invasive BCs were selected on hematoxylin and eosin (H&E)-stained sections, and two tissue cores (2 mm in diameter) were obtained from each specimen for TMA construction as previously described [16]. An H&E-stained section from each TMA block was reviewed to confirm the presence of morphologically representative areas of the original lesions. Appropriate controls with normal breast tissue were also included. After pretreatment and antigen retrieval, the slides were incubated with the primary antibodies for PPP2R (sc-81606; Santa Cruz Biotechnology, diluted 1:50) and Cyclin D1 (SP4; Neomarkers, diluted 1:50). For the discovery cohort, clinicopathological data were obtained from original pathology reports and Ki-67 IHC performed in cases missing this information. ER, PgR, Her2 and Ki-67 IHC were performed on the validation cohort, and used as surrogates for molecular subtyping, based on 13th St Gallen International Breast Cancer Conference (2013) Expert Panel recommendations [17]. The cut-off level for ER and PgR receptor positivity was nuclear staining in more than 1% of tumor cells, and the cut-off level for Ki-67 was based on work described in Cheang et al. [18]. In both cohorts, HER2 evaluation was carried out in compliance with the most recent American Society of Clinical Oncology/College of American Pathologists (ASCO/CAP) guidelines [19].
The details regarding TMA construction, antibodies used and the detailed conditions for each antibody, including those targeting PPP2R2A (B55α) and Cyclin D1, are included in the supplementary data (Additional file 1: Table S1). All morphological and IHC assessments were made by a pathologist (FS). Cyclin D1 expression was determined as previously described [20].

PPP2R2A (B55α) scoring
Because to the best of our knowledge there is no published data regarding a PPP2R2A (B55α) scoring method in BC, we first used the discovery cohort to establish the ideal assessment criteria for PPP2R2A (B55α) expression by IHC and applied the same methodology to the validation cohort. Briefly, PPP2R2A (B55α) IHC expression was semi-quantitatively evaluated in the cytoplasm of invasive carcinoma according to previous methodology [21]. Afterwards, using the minimum p-value method on the log-rank test for OS in the discovery cohort, we determined the ideal cut-off for dichotomization of PPP2R2A (B55α) scoring (further details are provided in the Additional file 1) [22]. After establishing the IHC assessment criteria in the discovery cohort, expression in the validation cohort was determined using the same methodology. Briefly, tumors were considered PPP2R2A (B55α) -/low when there was no detectable expression or if it was weak in less than 5% of tumor cells ( Figure 2).

Statistical analysis
Where appropriate, Pearson's χ2, Fisher's exact and Student's t tests were used. Cumulative survival probabilities were estimated using the Kaplan-Meier method, and differences between survival rates were tested for significance using the log-rank test. Multivariate analysis for survival was performed using the Cox proportional hazards model. Variables included in the models were defined a priori and are indicated under each table.
Departures from the proportional hazards assumption were assessed based on the Schoenfeld residuals. Hazard ratios (HR) and 95% confidence intervals (95% CI) were estimated for each variable. All tests were two-sided with a 95% CI and a p-value of < 0.05 considered significant. Correction for multiple comparisons was performed where indicated using the Benjamini-Hochberg procedure. TCGA data were analyzed using R Version 3.0.2 for Mac OS X, and all other statistical analysis were carried out using Stata® software, version 13.
All analyses and data reporting were performed according to the "REporting recommendations for tumour MARKer prognostic studies" (REMARK) guidelines and STrengthening the Reporting of OBservational studies in Epidemiology (Additional file 2) statement [23,24].

PPP2R2A CNA status is associated with mRNA expression levels and OS in the TCGA dataset
To investigate whether there was a correlation between CNA and PPP2R2A expression, namely if CNA with any deletion was associated with lower PPPR2A expression levels, data for 852 BC samples were retrieved from TCGA. We selected the cases displaying either normal copy number, heterozygous PPP2R2A deletion or homozygous deletion (n = 736) for further analysis. Cases displaying either heterozygous or homozygous deletions at the PPP2R2A gene locus constituted 53.8% (n = 458) of all cases. The PPP2R2A mRNA expression level shows a significant association with CNA, with cases with heterozygous or homozygous deletion of the PPP2R2A locus displaying a significantly lower mRNA expression Z-Score ( Figure 3). The p-values of the t-test performed for every pairwise comparison were <0.001, after multiplicity correction using the Benjamini-Hochberg procedure.
Using the same dataset (Invasive Breast Cancer dataset, TCGA, provisional), we performed survival analysis on all cases available with CNA data (either HomDel or HetDel) and survival information (n = 582 patients). We compared the OS of patients with CNA (n = 374, 33 reported events) with those without CNA at the PPP2R2A locus (n = 208, 9 reported events) and show that patients with CNA at PPP2R2A have significantly worse survival  (Wilcoxon test for equality of survival functions p = 0.047) (Figure 4). Due to the extensive follow-up time and unavailability of the time to progression in the currently publically available version of the dataset, we used the Wilcoxon test because we considered the earlier events to have occurred due to disease recurrence. We performed a sensitivity analysis that supports the assumption that disease-specific events occurred earlier (more details in Additional file 1: Figure S1).

PPP2R2A (B55α) -/low expression is associated with higher grade and ER positive carcinomas
In this study, we included 280 tumors from the discovery cohort and 378 from the validation cohort that had PPP2R2A (B55α) expression and outcome data. A summary of the patients' characteristics is provided in Additional file 1: Table S3 for the discovery cohort and in Additional file 1: Table S4 for the validation cohort. The prevalence of carcinomas with PPP2R2A (B55α) -/low expression was similar in both cohorts, with 51.8% (n = 145) and 54.0% (n = 204) of tumors expressing PPP2R2A (B55α) in the discovery and validation cohorts, respectively. In the discovery cohort, tumors displaying PPP2R2A (B55α) -/low expression are associated with ER expression (p < 0.001), higher grade (p = 0.004) and Her2 overexpression (p < 0.001). Tumors displaying PPP2R2A (B55α) -/low expression are also associated with shortened DFS and OS (log-rank p < 0.001 and p = 0.002, respectively) (Additional file 1: Table S3). Associations between PPP2R2A (B55α) -/low expression and ER expression and short DFS and OS were further confirmed in the validation cohort (p = 0.001 for ER; log-rank p = 0.006 for DFS and p =0.033 for OS) (Additional file 1: Table S4).
Due to the hypothesis set a priori regarding the association between PPP2R2A (B55α) expression and ER status, which was subsequently tested on the two independent cohorts used, we conducted further analysis restricted to the luminal-like BC subgroup.

PPP2R2A (B55α) expression defines luminal-like BC groups with distinctive outcomes
When restricting the analysis to the luminal-like BC subgroup, in the discovery cohort, 41.4% (n = 96) of the cases display PPP2R2A (B55α) -/low expression, while in the validation cohort, these tumors correspond to 42.3% (n = 116) of luminal BC (Additional file 1: Table S5 and Additional file 1: Table S6, respectively). Again, significant associations with grade and PPP2R2A (B55α) expression status are found, with PPP2R2A (B55α) -/low tumors displaying higher grade (p = 0.004). Luminal BCs with PPP2R2A (B55α) -/low are also associated with a higher prevalence of Her-2 overexpression (p < 0.001), although no association with the Ki-67 proliferation index was found (p = n.s.) (Additional file 1: Table S5). PPP2R2A (B55α) -/ low expression status also defined subgroups of luminal BCs with differential outcomes, with PPP2R2A (B55α) -/ low -expressing BC associating with worse DFS and OS (log-rank: p = 0.002 and p = 0.001, respectively ( Figure 5). Despite considerable differences in the baseline characteristics between luminallike cases in the discovery and validation cohorts, Figure 4 Survival probability according to CNA on PPP2R2A loci. In the TCGA database, carcinomas harboring CNA (either HetDel or HomDel) at the PPP2R2A locus (n = 374) display significantly shorter OS (Wilcoxon p = 0.047).
namely regarding age, tumor size and grade (Additional file 1: Table S7), the association between PPP2R2A (B55α) expression status and worse DFS and OS was further confirmed in this subset of patients in the validation cohort with PPP2R2A (B55α) -/ low BC (p = 0.031 for DFS and p = 0.027 for OS). In the luminal-like BC subgroup, in both cohorts, PPP2R2A (B55α) -/ low is significantly associated with steeper mortality curves, thus further validating the scoring methodology used and the association between PPP2R2A (B55α) expression status and survival in this subset of patients ( Figure 5).

PPP2R2A (B55α) -/low /Cyclin D1 high expression defines a group of luminal BC with worse outcomes
To address the secondary hypothesis that the combination of altered PPP2R2A (B55α) protein expression and Cyclin D1 expression would define a group of luminal BC with worse outcomes, we first evaluated Cyclin D1 expression in the luminal BC cases of the discovery cohort. In this cohort, the overexpression of Cyclin D1 is associated with higher grade carcinomas (p = 0.029), but there was no significant association with any of the survival outcomes (Additional file 1: Figure S2). Next, we selected the cases with Cyclin D1 overexpression from those with negative/low expression of PPP2R2A (B55α), thus defining the phenotype PPP2R2A -/low /Cyclin D1 high . The prevalence of this phenotype is 11.9% in the complete discovery cohort and 14.5% when restricting luminal-like BC (Table 1). Interestingly, despite the association with higher grade (p = 0.014) and Her-2 OE carcinomas (p = 0.020), 51.5% of the cases displaying the phenotype PPP2R2A -/low /Cyclin D1 high would be otherwise classifiable as luminal A-like (Table 1) and the presence of the phenotype was also associated with both worse OS (HR 2.71, 95% CI 1.17-6.28, p = 0.020) and DFS (HR 2.63, 95% CI 1. 26-5.11, p = 0.010) (Table 1 & Figure 6). Since this effect on survival could be largely dependent on HER2 status, disease stage and proliferation index, we performed multivariate Cox proportional hazards survival analysis adjusting for HER2 status, tumor size and lymph node status, after performing multiple imputation of missing values for lymph node status (more information on modeling is provided in the Additional file 1). The proliferation index variable was dropped due to a lack of significance in the univariate analysis using the set cut-off of 14% (HR 1.88, 95% CI 0.84-4.19, p = 0.120 for OS; Additional file 1: Figure S3). In this multivariate model for PPP2R2A -/low /Cyclin D1 high cases, the HR for OS is 2.09 (95% CI 0.89-4.96, p = 0.091), and the HR for DFS is 2.18 (95% CI 1.02-4.66, p = 0.043) ( Table 2). Thus, in luminal-like BC, the PPP2R2A -/low /Cyclin D1 high phenotype is significantly associated with worse DFS, independent of HER2 status, tumor size and lymph node status ( Table 2).

Discussion
Currently, the PP2A holoenzyme complex and certain subunits including PPP2R2A (B55α) are thought to be tumor suppressors in many cancer types [6,9,25,26]. However, the expression status of the PP2A holoenzyme complex subunits, specifically PPP2R2A, as a biomarker has hardly been explored in any cancer type, including BC. Here, we present the first study to evaluate PPP2R2A (B55α) expression in BC and show that CNA, namely deletions centered on PPP2R2A in the most recent TCGA Breast Cancer dataset, is associated with worse OS. We also provide pioneering evidence that absence of PPP2R2A (B55α) expression in BC is associated with an aggressive phenotype, even in strikingly different cohorts of patients.
In contrast to the lack of data on PPP2R2A (B55α), there are numerous publications on the prognostic and predictive roles of Cyclin D1 overexpression and CCND1 amplification, which are usually associated with early relapse and poor prognosis in ER-positive BC [20,[27][28][29][30][31]. Taking in account some of the reported characteristics of the Integrative Cluster 2 BC [9,12], we hypothesized that using a combined analysis of Cyclin D1 and PPP2R2A (B55α) expression assessed by IHC, we would be able to identify an immunophenotype of luminal BC that is associated with poor OS and DFS. Indeed, we show that in luminal BC, the PPP2R2A -/low / Cyclin D1 high phenotype is associated with higher histological grade and concomitant HER-2 overexpression, as well as shorter DFS and OS. Moreover, approximately half of the cases with the PPP2R2A -/low /Cyclin D1 high phenotype correspond to luminal B-like BC, which is similar to the prevalence of luminal B BC in Integrative Cluster 2 BC [12]. However, we also show that the association with worse DFS is independent of HER2-status, as well as tumor size and lymph node status. Thus, we provide additional clinicopathological support for the role of PPP2R2A (B55α) as a tumor suppressor in BC and present pilot evidence that the PPP2R2A -/low /Cyclin D1 high phenotype may be clinically relevant in BC and should be further investigated.
We consider the PPP2R2A (B55α) scoring methodology proposed in our study to be fairly robust, as both cohorts had similar proportions of cases assigned to the various groups. Moreover, given the clinicopathological differences between the cohorts, the PPP2R2A (B55α) scoring methodology was able to define groups displaying comparable magnitudes of HR, thus further reinforcing the reproducibility of the method. Regarding Cyclin D1

Conclusion
In conclusion, the results of this study show that CNAs in the PPP2R2A gene and its expression are associated with worse OS and that the PPP2R2A -/low /Cyclin D1 high phenotype is associated with aggressiveness in luminal-like BC. Importantly, not only does the PPP2R2A -/low /Cyclin D1 high phenotype add prognostic information to otherwise HER2expressing luminal BC, it also allows the identification of a subgroup of luminal BC with reduced DFS that would be otherwise classified as luminal A-like. Consequently, these results provide one of the first indications for the clinical relevance of this phenotype in human BC. Based on the functional characteristics and cancer-specific expression pattern of PPP2R2A (B55α) and Cyclin D1, it is evident that both PPP2R2A and the PPP2R2A -/low /Cyclin D1 high phenotype deserve further attention to better uncover their possible clinical and biological importance in BC.  Table 1 -Antibodies used in the immunohistochemistry study, PPP2R2A (B55α) and Cyclin D1 IHC, PPP2R2A (B55α) expression evaluation and cut-off determination, SD Table 2 -Cut-off determination using minimum p-value; Supplementary Results: SD Table 3 -Discovery cohort: Clinico-pathological data of the patients, SD Table 4 -Validation cohort: Clinico-pathological data of the patients, SD Table 5 -Discovery cohort: Clinico-pathological data of Luminal BC patients, SD Table 6 -Validation cohort: Clinico-pathological data of Luminal BC patients, SD Table 7 -Comparison of the clinico-pathological characteristic of the cohorts, SD Figure 1 -Survival probability according to copy number status on PPP2R2A and sensitivity analysis, SD Figure 2 -Survival probability according to Cyclin D1 (CCND1) expression status, SD Figure 3-Forrest plot with odds-ratio for predictors of OS, Survival Model of PPP2R2A -/low /Cyclin D1 high phenotype; Supplementary Data References.