Phosphatidylinositol 3-kinase pathway activation in breast cancer brain metastases

Introduction Activation status of the phosphatidylinositol 3-kinase (PI3K) pathway in breast cancer brain metastases (BCBMs) is largely unknown. We examined expression of phospho(p)-AKT, p-S6, and phosphatase and tensin homologue (PTEN) in BCBMs and their implications for overall survival (OS) and survival after BCBMs. Secondary analyses included PI3K pathway activation status and associations with time to distant recurrence (TTDR) and time to BCBMs. Similar analyses were also conducted among the subset of patients with triple-negative BCBMs. Methods p-AKT, p-S6, and PTEN expression was assessed with immunohistochemistry in 52 BCBMs and 12 matched primary BCs. Subtypes were defined as hormone receptor (HR)+/HER2-, HER2+, and triple-negative (TNBC). Survival analyses were performed by using a Cox model, and survival curves were estimated with the Kaplan-Meier method. Results Expression of p-AKT and p-S6 and lack of PTEN (PTEN-) was observed in 75%, 69%, and 25% of BCBMs. Concordance between primary BCs and matched BCBMs was 67% for p-AKT, 58% for p-S6, and 83% for PTEN. PTEN- was more common in TNBC compared with HR+/HER2- and HER2+. Expression of p-AKT, p-S6, and PTEN- was not associated with OS or survival after BCBMs (all, P > 0.06). Interestingly, among all patients, PTEN- correlated with shorter time to distant and brain recurrence. Among patients with TNBC, PTEN- in BCBMs was associated with poorer overall survival. Conclusions The PI3K pathway is active in most BCBMs regardless of subtype. Inhibition of this pathway represents a promising therapeutic strategy for patients with BCBMs, a group of patients with poor prognosis and limited systemic therapeutic options. Although expression of the PI3K pathway did not correlate with OS and survival after BCBM, PTEN- association with time to recurrence and OS (among patients with TNBC) is worthy of further study.


Introduction
The incidence of brain metastases (BMs) is approximately 15% among women newly diagnosed with metastatic breast cancer (BC) [1]. This figure likely underestimates the true incidence, as autopsy studies report a 30% incidence of BMs among women with advanced disease [2]. Current therapeutic interventions include corticosteroids, whole-brain radiotherapy, neurosurgical resection, stereotactic radiosurgery, and systemic chemotherapy [3]. Despite these treatment strategies, prognosis among patients with BCBMs remains poor, with a median overall survival of approximately 6 months [4,5]. Although targeted agents show promise in the treatment of advanced extracranial BC, challenges in delivery of these agents to the central nervous system (CNS) include properties inherent to the blood barrier (that is, efflux mechanisms) and our incomplete understanding the biology underlying BCBMs. Moreover, optimal therapeutic targets within BCBM are largely unknown.
Previous studies indicate that the phosphatidylinositol 3-kinase (PI3K) pathway plays a critical role in the initiation and progression of human BC, and alterations in this pathway have been identified in approximately 50% of these tumors [6,7]. PI3K pathway activation occurs in response to extracellular signals via either growth-factor receptor or integrin pathways. On its recruitment to the cellular membrane via receptor-mediated activation, the p110α catalytic subunit of PI3K phosphorylates phosphatidylinositol-4,5bisphosphate (PIP2) at the 3' position of the inositol ring, generating PIP3 [8]. PIP3 recruits phospholipidbinding domain containing proteins, particularly AKT, to the plasma membrane. Phosphorylated (p-)AKT, the primary downstream effector of PI3K signaling, moves from the cytoplasm to the nucleus to initiate its downstream effects. This cascade, including activation of the mammalian target of rapamycin (mTOR) and its downstream effectors, p70S6 kinase and 4Ebinding protein-1, affects a number of cellular processes, including proliferation and motility, which clinically translate into endocrine and chemotherapy resistance and worse cancer-specific survival [6,[9][10][11][12]. The PI3K/AKT pathway is negatively regulated by PTEN (phosphatase and tensin homologue), a lipid phosphatase that removes the 3-phosphate from PI (3,4)P2 and PI (3,4,5)P3, thus inactivating the signaling cascade [13]. Therefore, loss of PTEN contributes to the activation of the PI3K/AKT signaling cascade through inhibition of degradation of both PI(3,4)P2 and PI (3,4,5)P3.
To date, alterations and activation of the PI3K/AKT pathway are well established in the initiation and progression of extracranial human BC [6,[8][9][10]14,15]. However, the contribution of this important signaling pathway to the pathogenesis of BCBMs has yet to be fully elucidated. This is of clinical importance as smallmolecule inhibitors of the PI3K/AKT/mTOR pathway are in development and show promising activity in the treatment of primary brain tumors, suggesting sufficient blood-brain barrier penetration to elicit therapeutic effects [16,17]. In this study, we quantitated the expression of the PI3K pathway biomarkers p-AKT, p-S6, and PTEN, and evaluated the prognostic implications, primarily overall survival (OS) and survival after BCBMs, of PI3K activation status in BCBMs. As secondary, exploratory end points, we evaluated the associations between PI3K pathway activation and time to distant recurrence (TTDR) and time to BCBM. Finally, similar analyses were also conducted among the subset of patients with triple-negative BCBM.

Materials and methods
Patients BCBMs (n = 52), including a subset with matched primary BCs (n = 12), from 52 patients treated at the University of North Carolina at Chapel Hill (UNC) (52%) and Duke University (48%) between 1991 and 2008, were studied. Clinical data, including age, race, stage of primary BC at diagnosis (as per the 2003 Sixth Edition of the American Joint Committee on Cancer [AJCC]), treatment history (including local and systemic therapies in the adjuvant and metastatic settings), recurrence, and vital status were available for 50 patients. Given the retrospective nature of clinical data collection, complete information was not available for all 50 patients; thus, denominators may vary throughout the article. This study was approved, and waivers of consent were granted by Institutional Review Boards at both UNC and Duke.
Immunohistochemistry scoring IHC stains were scored by two surgical pathologists (KF and CL). Nuclear ER and PR staining were scored from 0 to 8 by using the Allred system [17], and Allred scores of 0 to 2 and 3 to 8 were defined as negative and positive, respectively. HER2 was scored by using the current American Society of Clinical Oncology ASCO)/College of American Pathologists (CAP) guidelines [18]. Membranous immunoreactivity was scored (0 and 1+ indicates negative; 2+, indeterminate; and 3+, positive for overexpression), and the percentage of tumor cells staining positive was visually estimated.

Identification of breast cancer subtypes
BC subtype was defined by IHC receptor status of the BCBM as follows: Hormone receptor (ER and/or PR) HR+/HER2-, triple-negative (ER-/PR-/HER2-, TN), and HER2+. In five BCBM cases for which ER, PR, and HER2 IHC data were not available, receptor status of the BCBM was obtained per the clinical database and was used to assign subtype. In three of five cases in which HER2 re-staining was indeterminate (IHC = 2+), HER2 classification was based on available clinical data including either IHC and/or FISH (fluorescence in situ hybridization). In the other two cases, HER2 2+ was classified as negative for the purposes of this analysis.

Gene-expression microarray analysis
Normalized gene-expression data from two publicly available datasets [19,20] were evaluated. In Harrell et al. [19] (GSE26338), we analyzed the expression of PTEN in a combined cohort of 855 primary BC patients that were followed, and the first site of distant recurrence was retrospectively annotated, including a subset of 42 patients in whom the first site of relapse (or one of the first sites) was the brain. The intrinsic molecular subtype calls were used as provided in Harrell et al. [19]. In Zhang et al. [20] (GSE14018), we evaluated the expression of PTEN across 36 unpaired BC brain, liver, bone, and lung metastasis samples. In both datasets, all PTEN probes were averaged into a single expression value.

Statistical analysis
The Kaplan-Meier method and log-rank test were used to compare differences among survival curves, and Cox regression analysis was used to evaluate possible predictors in the time-to-event outcomes. Overall survival (OS) was defined as the time from diagnosis of primary BC to death or last contact. CNS-specific survival (CNS survival) was defined as the time from the date of BCBM to the date of death or last follow-up. Time to distant recurrence (TTDR) was defined as the time from primary BC diagnosis to date of distant recurrence. Time to CNS recurrence (TTCNS) was defined as the time from primary BC diagnosis to date of CNS metastases; for those patients whose initial distant recurrence included the CNS, this time was the same as the time to distant recurrence. Differences in PTEN gene expression across the various intrinsic molecular subtypes and between brain metastases and other distant metastastic sites was evaluated by using a Wilcoxon rank sum test. Associations with relapse-free survival were evaluated by using the log-rank test, and Cox regression; PTEN expression was categorized as low/medium or high based on combining the lower two tertiles. Statistical analyses were performed with SAS 9.2 statistical software (SAS Institute, Inc., Cary, NC) and R v.2.8.1 http:// cran.r-project.org.

Patient and tumor characteristics
The clinical characteristics of the study population are presented in Table 1. The median age at diagnosis of primary BC was 48 years (range, 26 to 72 years). Sixtyeight percent of patients were Caucasian, 30%, African-American, and 2%, other ethnicities. Fifty percent of patients were Stage II, and 29% were Stage III at the time of surgery for primary BC.
BC subtype was assigned based on IHC staining of BCBM for 43 patients, and subtype distribution was as follows: 28% HR+/HER2-, 44% TN, and 28% HER2+. Subtype concordance between primary BC and associated BCBM was 57% (four of seven). Of the three cases that were discordant, two HER2+ primary BC lacked HER2 staining in the matched BCBM, whereas one TN primary BC gained HR positivity (HR+/HER2-) in the matched BCBM.
In the metastatic setting, 95% (36 of 38) of patients received some form of systemic therapy; with 32% receiving one line, and 63% receiving two or more lines of therapy. Seventeen percent (six of 35) received systemic therapy both before and after development of CNS metastases; 20% only before and 63% only after diagnosis of BCBM. Therapies in the metastatic setting included the following: chemotherapy (74%), endocrine therapy (70%, HR+ only), and HER2-directed therapy (67%, HER2+ only). Fifty-three percent received cranial radiation (XRT) for BCBM; 9% received radiosurgery. No difference in OS or CNS survival was seen between those who did or did not receive cranial XRT (P = 0.7, 0.5).

Expression of PI3K pathway biomarkers in breast cancer brain metastases
Activation of the PI3K pathway in BCBM was determined by evaluating the expression of p-AKT, p-S6, and PTEN with IHC. Expression of p-AKT and p-S6 was positive (H score, 10 to 300) in 75% and 69% of BCBM, respectively (Table 2a and Figure 1). Twenty-five percent of BCBMs lacked PTEN expression (PTEN-, H score, 0 to 9).
Concordance of PI3K expression between brain metastases and primary breast tumors PI3K pathway biomarkers status in primary BC and their matched BCBM was concordant in 67%, 58%, and 83% of 12 cases for p-AKT, p-S6, and PTEN, respectively (Table 3), and both gains and losses of which were evident for each biomarker evaluated.

Survival outcomes by expression of p-AKT, p-S6, and PTEN
The prognostic implications of p-AKT, p-S6, and PTEN expression in BCBMs were evaluated. Expression of p-AKT, p-S6, and PTEN (H score, < 10 versus ≥10) was not associated with the primary outcome of overall survival or survival after BCBMs (P ≥ 0.06 for all analyses; data not shown). In secondary analyses, neither   Figure 4) and CNS recurrence (2.5 versus 3.4 years; P = 0.06; Figure 5) even when stratified by TNBC (see later section) in exploratory analyses. Moreover, 5-year freedom from distant recurrence were 0 and 22% (95% confidence interval (CI), 10% to 36%) for PTEN-versus PTEN+ BCBMs. Similarly, 5-year freedom from CNS recurrence was 7.6% (95% CI, 0.5% to 29%) and 24% (95% CI, 12% to 39%) for these two groups, respectively. To confirm that these results were not influenced by receipt of systemic therapy, we evaluated the proportion of patients who received treatment by PTEN status. No difference was found in receipt of systemic chemotherapy, either in the curative ( Figure 2 Overall survival, defined as time from primary breast cancer diagnosis to death, by breast cancer subtype.

Survival by PTEN status among patients with triplenegative breast cancer
Recognizing the association between TNBC and PTEN expression, we evaluated the prognostic significance of PTEN expression within the TN BCBM subset as a secondary and exploratory outcome. PTEN-TN BCBMs were associated with inferior overall survival (1.7 years) compared with PTEN+ BCBM (4.7 years; P = 0.03;

Impact of subtype and PTEN status on patient outcome
Consistent with the Kaplan-Meier analyses, the TN IHC subtype was found to be associated with worse overall survival, time to distant and CNS recurrence, and survival after BCBM in univariable Cox regression analyses (all, P ≤ 0.06) ( Table 4). PTEN-was associated with more rapid time to distant recurrence (hazard ratio (HR), 2.2; P = 0.025); a borderline association between PTEN-and shorter time to CNS recurrence was observed (HR, 1.8; P = 0.07).
On multivariable Cox regression analyses, the association between subtype and overall survival, time to distant recurrence, and survival after CNS metastases remained significant when controlling for PTEN status (all, P ≤ 0.04). Similarly, the association between PTENand shorter time to distant recurrence (HR, 2.4; P = 0.02) and time to CNS recurrence (HR, 2; P = 0.06) remained when controlling for subtype among patients with BCBMs.

Evaluation of PTEN gene expression across the intrinsic molecular subtypes and brain metastases
To further explore the association of PTEN-with triplenegative disease and brain metastases, we interrogated two publicly available gene-expression microarray data sets that included: (a) 855 primary breast cancers with annotated intrinsic subtype and relapse-free survival data [19], and (b) 36 unpaired brain, lung, liver, and bone BC metastases [20]. First, we evaluated the expression of the PTEN gene across the intrinsic molecular subtypes in the Harrell et al. [19] dataset. As shown in Figure 7a, basal-like tumors showed an overall lower expression of PTEN compared with the other subtypes of BC. Second, we observed that PTEN expression was expressed at lower levels in BCBMs compared with other distant metastatic sites (Figure 7b). Although we cannot rule out that this observation is due to the fact that these brain metastases were largely of the basal-like subtype, whereas bone and liver metastasis were more of the luminal and HER2-enriched subtypes, these data support the association of lower levels of PTEN, basallike tumors, and the development of brain metastases.

Survival outcomes based on PTEN gene expression
Further to explore the association of PTEN-with poor outcome, we evaluated the Harrell et al. [19] combined microarray data set. In all patients (n = 855), lower levels of PTEN expression were found to be associated with poor prognosis at 5 years (P = 0.002), even when adjusted for ER status (P = 0.012) and ER status plus intrinsic molecular subtype (P = 0.025; Figure 8a). This suggests that PTEN is not just recapitulating the poor prognosis of the basal-like subtype, and supports our IHC-based findings that lack of PTEN expression is also found in the other tumor types. Moreover, in the subset of patients that relapsed to the brain in the first 5 years (n = 42), lower levels of PTEN expression were found to be associated with a shorter time to brain recurrence (1.6 versus 3.4 years; P = 0.003), even when adjusted for ER status (P = 0.001) and ER status plus subtype (P = 0.0003; Figure  8b). Finally, no association of S6K and AKT-1, -2, and -3 genes with outcome was observed (data not shown).

Discussion
BCBMs represent one of the most challenging aspects in the clinical care of patients with advanced BC. Not only does intracranial recurrence limit survival, but associated symptoms also decrease functional status, limit independence, and negatively affect quality of life. No approved systemic therapies are available to treat patients with BCBMs, and it is unclear whether therapeutic targets, such as PI3K, differ between primary BC and BCBMs. In the present study, we explored the expression and prognostic implications of a panel of PI3K-pathway biomarkers, p-AKT, p-S6, and PTEN, in 52 BCBMs and 12 matched primary BCs. Our central goal was to improve our current understanding of the complex biology underlying BCBMs in hopes of guiding the future use of targeted agents to treat this aggressive disease. Our results show that the PI3K pathway is active in most BCBMs, regardless of IHC subtype; however, activation status does not appear to affect overall survival or survival after BCBMs in this cohort of patients. Interestingly, our secondary analyses indicate Although alterations of the PI3K pathway are frequently observed in primary BC [25], different mechanisms of pathway activation exist [26]. Among them, activating PIK3CA mutations have been identified in about 15% to 30% of breast tumors and are more commonly associated with ER+ disease [27][28][29][30][31]. Conversely, alternative mechanisms of PI3K pathway activation, such as loss of PTEN and loss of the tumor-suppressor inositol polyphosphate 4-    [27,[32][33][34][35][36][37]. Although our data indicate that PI3K pathway activation in BCBM is not entirely subtype specific, lack of PTEN expression was more commonly observed in the TN and basallike subtypes when compared with the other tumor types. Given that brain metastases across subtypes were included in this study, multiple mechanisms of PI3K activation (that is, PIK3CA mutations, PTEN loss, and/or INPP4B loss) may be responsible for the high levels of PI3K pathway activation observed in this cohort. Future studies aimed at identifying subtypespecific mechanisms of PI3K activation are certainly warranted, both in primary BC and BCBM, to refine our current understanding of the biologic processes driving this disease process. The role of PI3K-pathway activation as a prognostic and/or predictive biomarker is under investigation. Although our primary analyses did not reveal associations between PI3K-pathway activation and overall survival or survival after BCBM, several secondary analyses are worthy of discussion. Our exploratory analysis indicates PTEN may be prognostic, with lack of PTEN expression being associated with more rapid time to disease recurrence (across subtypes) and worse overall survival in the TN subset of patients [38]. Interestingly, of the three biomarkers evaluated in this study, PTEN showed the highest concordance (83%) between matched primary BC and BCBM. In the clinical setting, biologic specimens from brain metastases are not commonly available, as resection is generally reserved for solitary lesions, and biopsies are reserved for cases with equivocal radiographic findings. Given the high concordance of PTEN status between primary BCs and their BCBMs, PTEN status in primary breast tumors may also be prognostic, and potentially predictive of distant and CNS recurrence. Confirming these findings in a large, unselected cohort of patients with primary breast tumor tissue available for PTEN testing would certainly be of value.
We recognize that the data presented in this study have several limitations. First, all patients included in this study underwent a neurosurgical procedure, so the population studied here might not be representative of all patients with BCBMs. Second, the sample size in this study is small, but comparable to previously reported studies evaluating BCBM tissues [39]. Although this is the largest study evaluating PI3K activation in BCBM tissues to date, subset analyses should be interpreted with caution because of the small sample size and multiple comparisons. A laudable future goal would be to validate these findings in a larger cohort; however, the inherent difficulty of obtaining brain metastasis tissues remains an obstacle. Thus, the development of clinically annotated brain metastases and primary BC tissue repositories housing both paraffin-embedded and fresh, frozen tissues should be a priority among the scientific community.
Finally, we used an IHC definition to identify the intrinsic molecular subtypes based on ER, PR, and HER2 status. We recognize that considerable discordance may exist between subtype assignment by IHC biomarkers and molecular profiling [40]. However, similar results were observed when we evaluated publicly available gene-expression data in more than 800 tumors in which molecular profiling had been performed [19,20]. Overall, this genomic analysis supports our IHC findings, in which PTEN expression was associated with time to distant and brain recurrence, basal-like tumors, and the development of BCBM.

Conclusions
In summary, results of this study indicate that the PI3K pathway is active in the majority of BCBMs across the spectrum of IHC subtypes. Although expression of the PI3K pathway did not correlate with OS and survival after BCBM, loss of PTEN may hold prognostic and/or predictive value among this group of very high risk patients. Presently, small-molecule inhibitors of the PI3K pathway are in clinical development to treat multiple malignancies, including BC [41,42], and several cross the blood-brain barrier. Thus, inhibition of the PI3K pathway represents a promising therapeutic strategy for patients with BCBMs, with the ultimate goal of improving outcome and quality of life for patients diagnosed with this devastating disease.