Predictive Biomarkers of Sensitivity to the Phosphatidylinositol 3′ Kinase Inhibitor GDC-0941 in Breast Cancer Preclinical Models (original) (raw)

Cancer Therapy: Preclinical| July 14 2010

Authors' Affiliations: Departments of 1Development Oncology Diagnostics, 2Cancer Signaling and Translational Oncology, and 3Bioinformatics, Genentech, Inc., South San Francisco, California

Search for other works by this author on:

Jeffrey J. Wallin;

Authors' Affiliations: Departments of 1Development Oncology Diagnostics, 2Cancer Signaling and Translational Oncology, and 3Bioinformatics, Genentech, Inc., South San Francisco, California

Search for other works by this author on:

Deepak Sampath;

Authors' Affiliations: Departments of 1Development Oncology Diagnostics, 2Cancer Signaling and Translational Oncology, and 3Bioinformatics, Genentech, Inc., South San Francisco, California

Search for other works by this author on:

Debraj GuhaThakurta;

Authors' Affiliations: Departments of 1Development Oncology Diagnostics, 2Cancer Signaling and Translational Oncology, and 3Bioinformatics, Genentech, Inc., South San Francisco, California

Search for other works by this author on:

Heidi Savage;

Authors' Affiliations: Departments of 1Development Oncology Diagnostics, 2Cancer Signaling and Translational Oncology, and 3Bioinformatics, Genentech, Inc., South San Francisco, California

Search for other works by this author on:

Elizabeth A. Punnoose;

Authors' Affiliations: Departments of 1Development Oncology Diagnostics, 2Cancer Signaling and Translational Oncology, and 3Bioinformatics, Genentech, Inc., South San Francisco, California

Search for other works by this author on:

Jane Guan;

Authors' Affiliations: Departments of 1Development Oncology Diagnostics, 2Cancer Signaling and Translational Oncology, and 3Bioinformatics, Genentech, Inc., South San Francisco, California

Search for other works by this author on:

Leanne Berry;

Authors' Affiliations: Departments of 1Development Oncology Diagnostics, 2Cancer Signaling and Translational Oncology, and 3Bioinformatics, Genentech, Inc., South San Francisco, California

Search for other works by this author on:

Wei Wei Prior;

Authors' Affiliations: Departments of 1Development Oncology Diagnostics, 2Cancer Signaling and Translational Oncology, and 3Bioinformatics, Genentech, Inc., South San Francisco, California

Search for other works by this author on:

Lukas C. Amler;

Authors' Affiliations: Departments of 1Development Oncology Diagnostics, 2Cancer Signaling and Translational Oncology, and 3Bioinformatics, Genentech, Inc., South San Francisco, California

Search for other works by this author on:

Marcia Belvin;

Authors' Affiliations: Departments of 1Development Oncology Diagnostics, 2Cancer Signaling and Translational Oncology, and 3Bioinformatics, Genentech, Inc., South San Francisco, California

Search for other works by this author on:

Lori S. Friedman;

Authors' Affiliations: Departments of 1Development Oncology Diagnostics, 2Cancer Signaling and Translational Oncology, and 3Bioinformatics, Genentech, Inc., South San Francisco, California

Search for other works by this author on:

Mark R. Lackner

Authors' Affiliations: Departments of 1Development Oncology Diagnostics, 2Cancer Signaling and Translational Oncology, and 3Bioinformatics, Genentech, Inc., South San Francisco, California

Corresponding Author: Mark R. Lackner, Genentech, Inc., 1 DNA Way, South San Francisco, CA 94080. Phone: 650-467-1846; Fax: 650-225-7571; E-mail: mlackner@gene.com.

Search for other works by this author on:

Corresponding Author: Mark R. Lackner, Genentech, Inc., 1 DNA Way, South San Francisco, CA 94080. Phone: 650-467-1846; Fax: 650-225-7571; E-mail: mlackner@gene.com.

Received: October 22 2009

Revision Received: April 29 2010

Accepted: May 04 2010

Online ISSN: 1557-3265

Print ISSN: 1078-0432

2010

Clin Cancer Res (2010) 16 (14): 3670–3683.

Split-Screen
Views Icon Views
Open the PDF for in another window
Tools Icon Tools
Search Site
Article Versions Icon Versions
- Version of Record July 14 2010
- Proof July 6 2010
- Accepted Manuscript May 7 2010

Citation

Carol O'Brien, Jeffrey J. Wallin, Deepak Sampath, Debraj GuhaThakurta, Heidi Savage, Elizabeth A. Punnoose, Jane Guan, Leanne Berry, Wei Wei Prior, Lukas C. Amler, Marcia Belvin, Lori S. Friedman, Mark R. Lackner; Predictive Biomarkers of Sensitivity to the Phosphatidylinositol 3′ Kinase Inhibitor GDC-0941 in Breast Cancer Preclinical Models. _Clin Cancer Res 15 July 2010; 16 (14): 3670–3683. https://doi.org/10.1158/1078-0432.CCR-09-2828

Download citation file:

Abstract

Purpose: The class I phosphatidylinositol 3′ kinase (PI3K) plays a major role in proliferation and survival in a wide variety of human cancers. A key factor in successful development of drugs targeting this pathway is likely to be the identification of responsive patient populations with predictive diagnostic biomarkers. This study sought to identify candidate biomarkers of response to the selective PI3K inhibitor GDC-0941.

Experimental Design: We used a large panel of breast cancer cell lines and in vivo xenograft models to identify candidate predictive biomarkers for a selective inhibitor of class I PI3K that is currently in clinical development. The approach involved pharmacogenomic profiling as well as analysis of gene expression data sets from cells profiled at baseline or after GDC-0941 treatment.

Results: We found that models harboring mutations in PIK3CA, amplification of human epidermal growth factor receptor 2, or dual alterations in two pathway components were exquisitely sensitive to the antitumor effects of GDC-0941. We found that several models that do not harbor these alterations also showed sensitivity, suggesting a need for additional diagnostic markers. Gene expression studies identified a collection of genes whose expression was associated with in vitro sensitivity to GDC-0941, and expression of a subset of these genes was found to be intimately linked to signaling through the pathway.

Conclusion: Pathway focused biomarkers and the gene expression signature described in this study may have utility in the identification of patients likely to benefit from therapy with a selective PI3K inhibitor. Clin Cancer Res; 16(14); 3670–83. ©2010 AACR.

Read the Commentary on this article by Turke and Engelman, p. 3523

This article is featured in Highlights of This Issue, p. 3521

The class I phosphatidylinositol 3′ kinase (PI3K) is activated in a wide variety of human malignancies, and inhibitors targeting the PI3K pathway hold great promise in the treatment and management of cancer. Successful development of such inhibitors will be enhanced by the identification of responsive patients through the use of predictive biomarkers. In this study, we identified a collection of putative biomarkers predictive of response to the selective inhibitor GDC-0941 in breast cancer. These biomarkers include PIK3CA mutations, human epidermal growth factor receptor 2 amplification, and a novel gene signature. We suggest that the evaluation of these biomarkers in clinical trials may enable clinical development of selective inhibitors of PI3K in appropriate patient populations.

The PIK3CA gene encodes the p110α catalytic subunit of class I phosphatidylinositol 3′ kinase (PI3K) and is mutated or amplified with high frequency in several solid tumor malignancies (1–3). The mutant versions of the protein exhibit increased enzymatic activity independent of upstream signaling, constitutively stimulates signaling through the AKT pathway, and has oncogenic properties such as conferring anchorage-independent growth and increased cell invasion and metastasis (4). Structural analyses have revealed that oncogenic mutations in p110α often lie at the interface between p110 and the p85 regulatory subunit, or between the kinase domain of p110 and other domains within the catalytic subunit, suggesting that they may disrupt the regulation of kinase activity by p85 or increase the catalytic activity of the enzyme, respectively (5, 6). Other key nodes of the PI3K signaling pathway are also frequently deregulated in a wide variety of cancers, including receptor tyrosine kinases that stimulate PI3K activity [e.g., human epidermal growth factor receptor 2 (HER2); ref. 7], the downstream effector kinase AKT1 (8), and the negative regulator PTEN (9). This frequent activation of the PI3K pathway in cancers has led to intensive efforts to identify therapeutics that abrogate PI3K signaling and hence may have utility in patients with cancers addicted to the PI3K pathway (10). Several potent and selective PI3K inhibitors have recently entered early-phase clinical trials (11), among them BEZ-235, a dual PI3K/mammalian target of rapamycin (mTOR) inhibitor (12), and GDC-0941, an inhibitor with low nanomolar potency against all four isoforms of class I PI3K (13, 14).

A major question in the clinical development of these inhibitors is whether key pathway alterations that activate PI3K signaling predict clinical benefit from these inhibitors and hence may have utility in the diagnostic stratification of patients. This question must ultimately be answered clinically by associating diagnostic test results with clinical outcomes, but preclinical studies can play an important role in the elaboration of clinical diagnostic strategies and prioritization of biomarker candidates. Indeed, previous studies have shown that GDC-0941 has preclinical activity in models harboring pathway alterations in PIK3CA, PTEN, or HER2 and have also hinted at broader activity (14), but have not extensively examined molecular determinants of response in the context of a single tissue type. Breast cancer represents an ideal tumor type to evaluate the predictive value of alterations in the PI3K signaling pathway in determining response to PI3K-targeting agents because the pathway is pathologically activated in all three of the major breast cancer subtypes through distinct mechanisms. Luminal subtype tumors are typically hormone receptor positive and have been reported to have a high prevalence (∼30%) of activating mutations in the PIK3CA locus itself (15). HER2-amplified tumors are characterized by dramatic genomic amplification and overexpression of the HER2 gene, an event that potently activates downstream PI3K signaling (16), and also by concomitant activating mutations in PIK3CA in ∼20% of tumors (15). The reason for redundant activation of the pathway at two distinct nodes is unclear but may have to do with overcoming feedback regulation because a well-described negative feedback loop emanates from the downstream pathway effector p70S6K1 and effectively inhibits upstream signaling by inactivating the adaptor protein IRS1 (17). Basal-like breast cancers have been reported to have relatively low prevalence of PIK3CA mutations (15), but frequent loss of expression of the protein product of the tumor suppressor gene PTEN (9). Thus, evaluation of sensitivity to selective PI3K inhibitors across models representing distinct subtypes and genetic alterations could shed light on the need for diagnostics and personalized medicine strategies during clinical development of these agents. To address this question in preclinical models, we carried out pharmacogenomic studies with GDC-0941 in a large panel of well characterized breast cancer cell lines and tumor xenograft models.

Materials and Methods

Breast cancer cell lines

Cell lines used in this study were obtained from the American Type Culture Collection and the Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH (DSMZ). Detailed molecular characterization for the genetic alterations described (PIK3CA, PTEN, KRAS, etc.) as well as molecular subtype assignments have been previously described (18, 19). MCF10A isogenic cells were obtained from Horizon Discovery LTD (http://www.horizondiscovery.com). Cell lines were tested and authenticated using gene expression and single nucleotide polymorphism genotyping arrays, as previously described (18, 20). All cell lines were maintained in RPMI 1640 or DMEM supplemented with 10% fetal bovine serum (Sigma), nonessential amino acids, and 2 mmol/L l-glutamine.

Inhibitors

GDC-0941 is a selective inhibitor of all four isoforms of class I PI3K. Details of its structure, selectivity, and biological properties have been previously described (13). Because selectivity to PI3K over other kinases is crucial to the conclusions of this article, we note that GDC-0941 has been screened against a panel of 288 kinases from Millipore (Upstate) and showed 300-fold selectivity over other assayable kinases (13). Two other kinases (TrkA and FLT1) did exhibit IC50s, but these were in the low micromolar range. PI3KA/Di is an inhibitor selective for the α and δ subunits of class I PI3K (IC50 for p110α, 3 nmol/L; for p110β, 332 nmol/L; for p110δ, 20 nmol/L; for p110γ, 112 nmol/L) whose structure has recently been described (21). PI-103 is a dual inhibitor of both class I PI3K and mTOR (22). All compounds were synthesized by the Genentech Medicinal Chemistry group.

In vitro cell viability studies

For cell viability studies, cells were plated in quadruplicate at a density of 3,000 cells per well in 384-well plates in normal growth medium and allowed to adhere overnight. GDC-0941 dose-response was determined by treating with 10 concentrations based on a 3-fold dilution series. Cell viability was measured 72 hours later using the CellTiter-Glo Luminescent Cell Viability Assay (Promega). The concentration of drug resulting in 50% inhibition of cell viability (IC50) was calculated from a four-parameter curve analysis (XLfit, IDBS software) and was determined from a minimum of three experiments. For cell lines that failed to achieve IC50 and the software was unable to calculate a value, the IC50 was nominally assigned as the highest concentration screened (20 μmol/L), and these cell lines were classified as resistant. Mean IC50 values and SDs from three experiments with each inhibitor are shown in Supplementary Table S1. High-content screening assays were carried out on an Arrayscan VTI (Cellomics, Inc.) using standard bromodeoxyuridine (BrdUrd) incorporation and staining protocols. Briefly, cells were plated at a density of 5,000 to 10,000 cells per well (depending on cell line growth properties) in PackardView 96-well plates and allowed to adhere overnight incubating at 37°C. The following day, the cells were treated with the GDC-0941 and allowed to incubate for 72 hours at 3°C. BrdUrd labeling reagent (Sigma) was then added to the cells at a final concentration of 200 nmol/L for an additional 5 hours, and then the plates were fixed and processed according to the manufacturer's standard protocol. Cells were counterstained with Hoechst-33258 to allow the identification of nuclei, and the percentage of cells positive for BrdUrd immunofluorescence was then quantitated for at least 1,000 cells per well using the Cellomics Target Activation software (http://www.cellomics.com).

Xenograft studies

Xenografted cells were in vivo selected versions of the cell lines used for in vitro analyses, and the identity of each was confirmed by single nucleotide polymorphism genotyping assay. For each experiment, breast cancer cells were implanted subcutaneously near the mammary fat pad of female severe combined immunodeficient mice. After implantation of cells into mice, tumors were monitored until they reached a mean tumor volume of 250 to 350 mm3, and animals were distributed into groups of 10 to 15 animals each before initiating dosing. GDC-0941 was dissolved in 0.5% methylcellulose with 0.2% Tween 80 vehicle and dosed daily for 14 days by oral gavage. Tumor volume was measured in two dimensions (length and width) using Ultra Cal-IV calipers (Model 54-10-111; Fred V. Fowler Company) and was analyzed using Excel version 11.2 (Microsoft Corp.). The tumor volume (mm3) was calculated as (longer measurement × shorter measurement2) × 0.5. Animal body weights were measured using an Adventurer Pro AV812 scale (Ohaus Corp.). Percent weight change was calculated as [1 − (new weight/initial weight)] × 100. Tumor sizes were recorded twice weekly over the course of the study (21 d). Mouse body weights were also recorded twice weekly, and the mice were observed daily. Mice with tumor volumes of ≥2,000 mm3 or with losses in body weight of ≥20% from their initial body weight were promptly euthanized per Institutional Animal Care and Use Committee guidelines. Studies were thus of different duration as animals in some instances had to be sacrificed early. Mean tumor volume and SEM values (n = 10) were calculated using the JMP statistical software, version 5.1.2, at end of treatment. % Tumor inhibition was calculated as 100 × (mean volume of tumors in vehicle-treated animals − mean volume of tumors in test article–treated animals given the test article)/mean volume of tumors in vehicle-treated animals. Data were analyzed, and P values were determined using the Dunnett's t test with the JMP statistical software, version 5.1.2 (SAS Institute). For pharmacodynamic analyses of pAKT (473) in KPL-4 and MDA-MB-231 xenografts, animals harboring 250 to 300 mm3 tumors were dosed with 100 mg/kg GDC-0941 or vehicle for 1 hour. Tumors were then collected and snap frozen.

Western blot analyses

Cells were plated in T25 flasks for in vitro analyses. When the cells became 70% to 80% confluent, they were treated with 0, 0.1, or 1 μmol/L GDC-0941 for 24 hours. Cells were washed with cold PBS, and lysate was collected using the T-PER tissue protein extraction reagent (Thermo Scientific). The Pierce BCA Protein Assay (Thermo Scientific) was used to quantitate the lysates. Thirty micrograms of protein were loaded and run on 12 or 20 well NuPAGE Novex 4% to 12% Bis-Tris Gels (Invitrogen) and transferred using the Invitrogen iBlot system. All antibodies were incubated in 5% bovine serum albumin (Sigma). For Western blots on xenograft tumors, samples were placed in 2-mL Eppendorf safe-lock tubes containing a 5-mm stainless steal bead (Qiagen) with 300 μL of lysis buffer and disassociated using the TissueLyser II (Qiagen) as described by the manufacturer. The protein concentration was determined using the Pierce BCA Protein Assay kit (Thermo Scientific). Fifty micrograms of protein per well was loaded onto a NuPAGE Novex 4% to 12% Bis-Tris Midi Gel. Antibodies were obtained from BD BioSciences, Cell Signaling Technology, Millipore (formally Upstate), or Santa Cruz Biotechnology. The antibodies used were pAKT(Ser308), pAKT(Ser473), AKT, pERK(Thr202/Tyr204), CyclinD1, p27, pSGK(T256), p4EBP1(Ser65), pS6(Ser235/6), p70S6K1, cleaved poly ADP ribose polymerase (PARP), Actin, SKP2, CyclinE2, and pRb.

Flow cytometry

For cell cycle analysis, cells were treated with the indicated concentrations of GDC-0941 for 24 hours, then harvested with 0.25% trypsin and washed with PBS. Cells at a density of 1 × 106 were fixed in 100% ice-cold ethanol and stored at −20°C overnight, then washed with PBS and incubated in propidium iodide (PI) solution [0.1% Triton-X, 0.2 mg/mL RNase solution (Sigma), 0.05 mg/mL PI solution (Sigma), in PBS] for 30 minutes at room temperature in the dark. Cells were immediately analyzed with a FACscan flow cytometer (Becton Dickinson). For cell death analyses, 1 × 106 cells were seeded on 10-cm plates for 16 hours. Cells were then treated for 72 hours at the concentrations indicated of GDC-0941. Following treatment, the conditioned medium, 1× PBS wash, and the trypsinized cells were collected together by centrifugation. The cells were then stained with Annexin V-FITC (BD Biosciences) and PI solution (Sigma) according to the manufacturer's instructions and were analyzed by flow cytometry.

Gene expression microarray studies and biomarker analyses

Gene expression analysis of breast cancer cell lines was carried out as previously described (20). Baseline gene expression data for the cell lines described in this study have been deposited in the Gene Expression Omnibus database under accession number GSE12790, and expression data from T47D cells treated with GDC-0941 have been deposited under accession number GSE20719. The Oncomine Concepts map was used to determine association of the GDC-0941 treatment signature with published cancer data sets (23). For categorical analyses, cell lines were binned into sensitive and resistant classes based on an IC50 cutoff of less than or more than 1 μmol/L, respectively, and the Cyber T algorithm was implemented to identify genes differentially expressed between the classes. Cyber T is a variation of the t test that uses a Bayesian estimate of variance to correct for noise and variability commonly seen in microarray data (24). P values were adjusted for multiple testing using the false discovery rate method of Storey and Tibshirani (25). Hierarchical clustering of differentially expressed genes was carried out on median-centered, log-transformed data using the Cluster software (26) and subsequently visualized using the Treeview software (27). We also explored the effects of different sensitivity cutoffs on the genes that emerged from the analysis by using cutoffs of >500 nmol/L or >10 μmol/L to define resistance to GDC-0941. We found that 323 (84%) and 340 (88%) of the 386 genes from the 1 μmol/L Cyber T list were identified as showing statistically significant association with sensitivity when Cyber T analysis was run using the 500 nmol/L and 10 μmol/L cutoffs, respectively, suggesting that the gene list is relatively insensitive to differences in cutoffs (data not shown).

To identify genes that showed expression changes upon GDC-0941 treatment, T47D cells were plated in T-25 flasks and treated with either DMSO or 1 μmol/L GDC-0941 (each in triplicate) for 6 hours, followed by harvesting, RNA preparation, and gene expression profiling on Affymetrix U133P microarrays. Differentially expressed genes were identified by t test, with P values adjusted for multiple testing. Genes from the Cyber T signature that also show altered expression upon GDC-0941 treatment in T47D cells were identified based on a t test cutoff of P < 0.05 and at least a 20% average change in expression for all three replicates. Identical cutoffs were used to identify genes from the Cyber T signature that show altered expression in RNA prepared from MCF10A PIK3CA mutant cells versus isogenic MCF10A wild-type (WT) cells. Genes that showed altered expression in GDC-0941-treated T47D cells and PIK3CA mutant MCF10A cells were selected based on showing reciprocal changes in expression of at least 50% up or down compared with control cells. The Oncomine Concepts map was used to determine the association of the GDC-0941 treatment signature with published cancer data sets by virtue of analyses that assess significance of overlap between the query signature and signatures in the database using a Fisher's exact test (23). Pathway relationships were visualized using the Ingenuity database (http://www.ingenuity.com). Statistical association of pathway biomarkers with sensitivity, along with predictive accuracy of biomarkers, was calculated using 2x2 contingency tables and Fisher's exact test with the GraphPad Prism software.

Results

In vitro sensitivity to GDC-0941

We screened 54 breast cancer cell lines for sensitivity to the selective class I PI3K inhibitor GDC-0941 using an ATP-based cell viability assay and found a wide range of sensitivities, from an IC50 of 0.09 μmol/L in sensitive EFM19 cells to 20 μmol/L in several cell lines in which GDC-0941 did not result in 50% inhibition of viability (Fig. 1A). Overall GDC-0941 showed excellent in vitro activity across the panel of breast cancer cell lines, with 46% (25 of 54) of lines having an IC50 below 500 nmol/L and 72% (39 of 54) of lines having an IC50 of <1 μmol/L. We confirmed these findings in an assay that more directly assesses cell proliferation by determining the effects of 24-hour GDC-0941 treatment on BrdUrd incorporation, a measure of DNA synthesis, in a subset of the cell lines (Supplementary Fig. S1). This analysis confirmed that cell lines that showed sensitivity in the ATP-based viability also showed markedly reduced BrdUrd incorporation, whereas cell lines with lesser sensitivity showed a minimal effect of GDC-0941 treatment on BrdUrd incorporation. We also screened two other inhibitors of PI3K/mTOR signaling across 50 cell lines from the panel and observed a similar profile of sensitivity and resistance. PI3KA/Di is a recently described inhibitor that is selective for the α and δ isoforms of the PI3K catalytic subunit (21). IC50 values for PI3KA/Di across the panel were highly correlated with those for GDC-0941 (Pearson R = 0.80; Fig. 1B). PI-103 is an inhibitor of p110α, mTOR, and DNA-PK (22). IC50 values for PI-103 were also correlated with those for GDC-0941, although to a lesser extent than with PI3KA/Di (Fig. 1B). Together, these results suggest that the sensitivity profile we observe is likely to be due to dependence on PI3K/AKT signaling rather than differences in drug metabolism or transport between cells.

Fig. 1.

In vitro response to GDC-0941 and relationship to PI3K pathway alterations and breast cancer subtype. A, the half-maximal inhibitory concentration (IC50) of GDC-0941 for 54 breast cancer cell lines, determined from an ATP-based cell viability assay and ordered from lowest to highest. Boxes below the chart indicate molecular subtype (L, luminal; H, HER2 amplified; B, basal-like), activating hotspot mutations in PIK3CA (blue), HER2 amplification (green), or PTEN protein loss determined by Western blotting (red). Molecular subtype of each cell line was determined by gene expression profiling and assessing HER2 amplification status, as previously described (18, 19). B, sensitivity profile of three different PI3K inhibitors across the breast cancer cell line panel. Cell lines in the heat map are clustered by z score–transformed IC50 values. GDC-0941 is a pan-inhibitor of all four class I subunits of PI3K; PI3KA/Di is selective for the α and δ subunits; and PI-103 is a dual inhibitor of class I PI3K and mTOR. C, scatter plot showing relationship between GDC-0941 IC50 (_Y_-axis) and genetic alterations in key signaling pathway components in the cell lines (_X_-axis). D, scatter plot showing relationship between GDC-0941 IC50 (_Y_-axis) and molecular subtype of the cell lines (_X_-axis).

Fig. 1.

Close modal

In an effort to understand the molecular predictors of response to GDC-0941, we determined whether key alterations in the PI3K pathway such as HER2 amplification, PIK3CA mutations, or PTEN loss (protein null by Western blotting) were associated with greater sensitivity to GDC-0941. We found that cell lines harboring oncogenic mutations in PIK3CA, or HER2 amplification, were significantly more sensitive than cell lines without these alterations (P = 0.005 PIK3CA versus WT, P = 0.048 HER2 versus WT; Fig. 1C). In addition, we found that cell lines harboring double alterations in PIK3CA and HER2, PIK3CA and PTEN, or HER2 and PTEN were common and were all significantly more sensitive than cell lines with no detectable alterations in the signaling pathway (P = 0.009, double mutants versus none). In contrast, we found no significant association between PTEN status and response to GDC-0941 (P = 0.71). We also sought to determine whether GDC-0941 response was different across cell line models representing different molecular subtypes of breast cancer. This analysis suggested that both HER2-amplified and luminal cell lines are on average significantly more sensitive to GDC-0941 than basal-like cell lines (P = 0.002 and P = 0.02, respectively, Mann-Whitney U test; Fig. 1D).

To evaluate the prediction accuracy of these pathway-focused biomarkers, we calculated sensitivity, specificity, and positive and negative predictive value for each marker as an individual test or all three markers taken together (Supplementary Fig. S2). PIK3CA mutations and HER2 amplification showed excellent specificity (100% and 95%, respectively) and high positive predictive value, but relatively low sensitivity (∼30%) and poor negative predictive value as single markers in predicting drug responsiveness in the cell line panel. Thus, a positive test for HER2 or PIK3CA would be predicted to accurately identify responsive patients (low false-positive rate), but to miss many responsive patients and have a high rate of false negatives. When PIK3CA, HER2, and PTEN were considered altogether as a composite diagnostic test, sensitivity was 69% and specificity was 67%. This suggests that although these known pathway alterations are likely to have value in identifying responsive patients, other biomarkers may still be required to identify all responsive patients.

We next sought to confirm our in vitro observation that molecular subtype and genetic activation of the PI3K signaling pathway play a role in determining responsiveness to GDC-0941 through in vivo studies in xenograft tumors from a core set of cell lines. We found that the in vivo studies recapitulated our in vitro findings (Fig. 2). Specifically, GDC-0941 showed excellent antitumor activity in xenograft models with HER2 amplification, PIK3CA mutations, or concomitant alterations in two pathway components (e.g., PTEN loss and PIK3CA mutation), but little or no effect in xenografts of basal-like KRAS mutant MDA-MB-231 cells (Fig. 2). We note that in all of these studies, the doses of GDC-0941 tested were well tolerated with no observable gross toxicities and minimal difference in body weights between control and GDC-0941–treated animals (data not shown). In addition, we found that treatment with GDC-0941 at a dose of 100 mg/kg substantially downregulated levels of pAKT(S473) in xenograft tumors of both sensitive KPL-4 cells and resistant MDA-MB-231 cells after 1 hour (Fig. 2), suggesting effective pharmacodynamic inhibition of PI3K signaling at this dose in both models.

Fig. 2.

In vivo antitumor activity of GDC-0941 in breast tumor xenograft models is related to PI3K pathway alterations. Daily dosing of GDC-0941 at 100 or 150 mg/kg was carried out in mice harboring s.c. implanted tumors for each of the five indicated models; points, mean tumor volume from treated animals and vehicle controls; bars, SEM. The legends for each graph indicate molecular subtype as well as key genetic alterations in PIK3CA, HER2, PTEN, or KRAS. Models are WT at each locus unless indicated. Percent tumor growth inhibition at the end of the study is also shown. Western blots at the lower right showing that 100 mg/kg GDC-0941 treatment effectively inhibited pAKT(S473) in sensitive KPL4 and resistant MDA-MB-231 xenograft tumors compared with vehicle control treatment. Tumors were collected after 1 h of treatment, and each lane represents an independent tumor sample.

Fig. 2.

Close modal

Flow cytometric analysis of GDC-0941 effects on cell cycle and apoptosis

Because the PI3K/AKT axis has been implicated in cell proliferation and antiapoptotic signaling, we conducted fluorescence-activated cell sorting (FACS)–based cell cycle and apoptosis analyses before and after GDC-0941 treatment in a subset of the cell lines to assay for effects on these processes. We found that 24-hour GDC-0941 treatment caused accumulation of the G1 fraction and concomitant reduction in the S-phase fraction in sensitive EVSA-T PTEN null and MDA-MB-361 HER2 PIK3CA mutant cells, although the effects were very modest in MDA-MB-361 cells. In contrast, we observed no effect on G1 arrest in resistant MDA-MB-231 KRAS mutant cells (Fig. 3A). This analysis also revealed the presence of a sub-G1 fraction in EVSA-T cells and MDA-MB-361 cells that suggested GDC-0941 induces apoptosis in these cells. To investigate this more specifically, we performed a FACS-based assay for Annexin V and PI staining, and a 72-hour time course of treatment. We found that 1 μmol/L GDC-0941 treatment induced apoptosis by 2- to 12-fold in sensitive lines that harbor pathway alterations in PIK3CA (T47D and MDA-MB-361) or PTEN (EVSA-T and HCC-70) but caused no induction of apoptosis in resistant cell lines (Fig. 3B and C). Effects were maximal at 72 hours of treatment (data not shown). These findings suggest that GDC-0941 has cytotoxic as well as antiproliferative effects in tumor models that are addicted to PI3K signaling through activating pathway alterations.

Fig. 3.

Flow cytometric analysis of cell cycle and apoptotic effects of GDC-0941 in sensitive and resistant cell lines. A, FACS cell analysis of GDC-0941 effects at 24 h. _X_-axis, DNA content assessed by PI staining; _Y_-axis, relative cell number. The legend to the right of each plot shows the percentage of cells in sub-G1, G1, S, and G2-M phase of the cell cycle. Cell line identity is shown to the left of each plot. B, FACS Annexin V assay for induction of apoptosis by GDC-0941 at 72 h. Density plots, cell populations stained with PI (_Y_-axis) or Annexin V (_X_-axis). Lower-right quandrant (Annexin V+, PI−), the early apoptotic cell population. C, quantitation of Annexin V/PI FACS assays similar to those shown in B for four GDC-0941–sensitive (green text) and three GDC-0941–resistant cell lines (red text). _Y_-axis, the fold increase in Annexin V+/PI− cells (lower-right quadrant) in cells treated with 0.1 or 1.0 μmol/L GDC-0941 relative to control-treated cells.

Fig. 3.

Close modal

GDC-0941 treatment downregulates PI3K/AKT and mTOR signaling

To investigate the mechanism whereby GDC-0941 inhibits cell viability in sensitive breast cancer cell lines, we conducted experiments on a core set of cell lines representing both sensitive and resistant models. Clear differential sensitivity between selected models is shown by in vitro viability dose-response curves shown in Supplementary Fig. S3. Immunoblotting of lysates prepared from treated cells showed that GDC-0941 caused marked dose-dependent decreases in pAKT(SER473) in sensitive and resistant cell lines (Fig. 4A). This effect was modest and only observed with 5 to 10 μmol/L GDC-0941 in MDA-MB-468 cells, and we note that this cell line has also been described to show only partial reductions in pAKT(S473) in response to a selective inhibitor of AKT1/2 (28). MCF-7 cells did not show detectable baseline pAKT(S473) so it was not possible to determine whether levels of this epitope were decreased upon treatment with GDC-0941, and a recent report has suggested that PI3K signaling is mediated via SGK3 rather than AKT1/2 activity in this and other cell lines with helical domain mutations in PIK3CA (29). We also determined phosphorylation levels of the key AKT substrates GSK-3β and FOXO1/3a in a subset of cell lines and found that in general and as expected, phosphorylation of these proteins tracked with those of pAKT (S473), with the exception that decreases in GSK3β phosphorylation were not observed in treated T47D cells (Supplementary Fig. S4). The AKT pathway regulates apoptosis through multiple mechanisms, and the FACS studies described above suggested apoptotic effects in sensitive cell lines, so we also looked at induction of cleaved PARP, a key substrate of activated caspases and an early indicator of apoptosis. Notably, we observed substantial increases in cleaved PARP accumulation upon GDC-0941 treatment in all three sensitive cell lines but no change in the three resistant cell lines (Supplementary Fig. S4; Fig. 4B).

Fig. 4.

Effects of GDC-0941 treatment on signaling pathway activation in sensitive and resistant cells. A, Western blot analysis of AKT(S473) and total AKT from cell lysates prepared after 6 h of treatment with a range of concentrations of GDC-0941. Text color indicates whether a cell line is sensitive (green) or resistant (red) to the effects of pharmacologic inhibition with GDC-0941. B, Western blots showing effects of 24-h 1 μmol/L GDC-0941 treatment on levels of phospho-proteins downstream of mTORC1 (pp70S6K1, pS6, and p4EB-P1) in lysates from four sensitive and four resistant cell lines. C, Western blots showing effects of 24-h 1 μmol/L GDC-0941 treatment on key regulators of the G1-S cell cycle transition in sensitive T47D and resistant MDA-MB-231 cells. D, pathway diagram showing relationship of analytes tested in this experiment for treatment response to PI3K/mTOR signaling, as well as whether levels of each analyte decreased (green fill) or increased (red fill) in sensitive cell lines.

Fig. 4.

Close modal

PI3K/AKT signaling also plays a key role in regulating the cellular translational machinery by releasing the mTORC1 complex from inhibition by TSC1 and TSC2 (30), so we next looked at whether pathway inhibition with GDC-0941 had effects on the signaling of mTORC1-regulated proteins in sensitive and resistant cell lines (Fig. 4B). We found that 24-hour treatment with GDC-0941 resulted in substantial decreases in phosphorylation of p70S6K1, concomitant decreases in the p70S6K1 substrate S6 ribosomal protein (pS6), as well as decreases in phosphorylation of the key translation regulator 4EB-P1 in all four sensitive cell lines (Fig. 4B). In contrast, we noticed more variable effects on these analytes in a similar experiment in the four insensitive cell lines. CAL-120 and MDA-MB-436 did show some decreases in pp70S6K1 and pS6 levels, and some decrease in p4EB-P1 levels in response to GDC-0941, although the effects were generally weaker and more variable than in the sensitive lines. We also observed that these two cell lines did not show substantial increases in cleaved PARP, suggesting that factors required for GDC-0941–induced apoptosis may be lacking in these cell lines. GDC-0941 had virtually no effect on pp70S6K1, pS6, 4EB-P1, or cleaved PARP in MDA-MB-231 and HCC-1428 cells. Overall, these results suggest that pS6K1, pS6, and p4EB-P1 may represent useful pharmacodynamic biomarkers of GDC-0941 activity because they are most substantially decreased in cell lines that show decreased viability in response to GDC-0941.

The PI3K/PTEN pathway has also been shown to play a key role in regulating the G1-S cell cycle transition (31), and we observed G1 arrest in the FACS studies described above, so we examined the expression of key regulators of this transition in GDC-0941–treated sensitive T47D cells and resistant MDA-MB-231 cells (Fig. 4C). We observed a robust increase in the expression of the negative regulator p27 upon 1 μmol/L treatment in T47D cells but no effect in MDA-MB-231 cells. In addition, expression of the ubiquitin ligase SKP2, Cyclin E2, and Cyclin D1, as well as phosphorylated Rb, all showed modest decreases in sensitive T47D cells but no change in resistant MDA-MB-231 cells (Fig. 4C and D).

Gene expression predictive markers of response to GDC-0941

Although our studies suggest that PIK3CA mutations and HER2 amplification may have value as predictive biomarkers of response to GDC-0941, there were also cell lines lacking either of these alterations that showed substantial in vitro sensitivity. To better understand additional response factors, we set out to identify gene expression predictors of response and resistance to GDC-0941 in an unbiased manner. We used the Cyber T algorithm and identified 386 microarray probe sets (294 named genes) that showed a statistically significant difference in expression between sensitive and resistant cells (P < 0.002; false discovery rate, <0.25). Figure 5 shows hierarchical clustering of all cell lines with the 35 genes whose expression was most significant in this analysis (false discovery rate, <0.1), and the complete list of genes is shown in Supplementary Table S2. It is notable and helps validate the approach that HER2 was identified in this unbiased analysis, given our earlier analyses that linked HER2 status with GDC-0941 sensitivity.

Fig. 5.

A gene expression signature predictive of pharmacologic sensitivity to GDC-0941. Heat map showing breast cancer cell lines hierarchically clustered by the top 35 genes (false discovery rate, <10%) whose baseline expression was associated with in vitro sensitivity to GDC-0941. Boxes at the top of the heat map indicate whether a cell was sensitive (S) or resistant (R) to GDC-0941. _X_-axis, cell lines; Y axis, genes; data were derived from log transformation and median centering for each gene. Red, high expression; green, low expression according to z scores.

Fig. 5.

Close modal

Gene discovery efforts from large data sets suffer from a substantial false discovery problem (25, 32) and an independent test set of cell lines was not available to confirm the predictive value of these gene sets in determining GDC-0941 response, so we also sought to prioritize genes by identifying those in which expression is under control of the PI3K pathway. We hypothesized that a collection of genes whose expression was associated with GDC-0941 sensitivity and was also regulated by PI3K signaling might be more likely to constitute a robust diagnostic signature of response. To identify such genes, we treated sensitive T47D cells with 1 μmol/L GDC-0941 (see Materials and Methods) and identified a large number of genes that showed a statistically significant change in gene expression after 6 hours of treatment (498 genes, adjusted P < 0.0001; Supplementary Table S3). This collection of genes that changed upon GDC-0941 treatment yielded interesting insights into PI3K signaling. For instance, pathway analysis showed that components of this signature were involved in the cell cycle, apoptosis, and biology of the proteasome (Supplementary Fig. S5). In addition, analysis of components of this treatment signature in the Oncomine gene database revealed that the GDC-0941 treatment signature is closely related to two gene signatures derived from perturbations of the insulin-like growth factor I-R axis, and also showed statistically significant overlap with gene sets overexpressed in high-grade breast tumors compared with low-grade tumors (33–38), as well as gene sets overexpressed in estrogen receptor–positive breast cancer (Supplementary Fig. S6; refs. 33, 35, 39). Our analysis revealed that 91 of the probe sets (83 genes) that showed significant changes in expression upon GDC-0941 treatment were also among the genes differentially expressed between sensitive and resistant cell lines from the Cyber T analysis (Supplementary Table S4). To further prioritize the gene list, we asked which of these genes were differentially expressed in MCF10A cells with an activated H1047R-encoding allele of PIK3CA compared with isogenic cells WT for PIK3CA, based on the hypothesis that true pathway-regulated genes should also show altered expression in the presence of an oncogenic mutation in PIK3CA. A total of 122 probe sets (103 genes) from the Cyber T signature were differentially expressed in PIK3CA mutant cells compared with WT isogenic cells (Supplementary Table S4). Seventeen genes were common to all three data sets, showing not only a significant difference in expression at baseline between sensitive and resistant cell lines, but also reciprocal changes in expression of at least 1.5-fold when the pathway was inhibited by GDC-0941 or activated by an oncogenic PIK3CA mutation (Fig. 6A and B).

Fig. 6.

Identification of genes from the pharmacologic sensitivity signature that are under control of the PI3K pathway. A composite gene expression signature composed of genes that are differentially expressed between sensitive and resistant cell lines and, in addition, are modulated by both GDC-0941 treatment and an activating mutation in PIK3CA. A, Venn diagram showing conceptual overlap of the three gene expression data sets that were used to identify the composite gene expression signature. The final 17 genes are differentially expressed between PIK3CA mutant and WT cells (Cyber T signature), show altered expression upon GDC-0941 treatment (treatment signature), and are differentially expressed between PIK3CA mutant and WT cells (isogenic cells). B, heat map showing fold change in expression of each of the 17 genes upon GDC-0941 treatment in T47D cells compared with vehicle control, as well as fold change in expression in MCF10A cells harboring a mutant H1047R-encoding PIK3CA allele compared with WT isogenic control cells. C, a model for negative feedback regulation of expression of key upstream genes that explains reciprocal changes in expression based on pathway activation or inhibition. Active signaling through the pathway engages negative feedback loops and inhibits expression of key regulators such as ERBB3 and IRS2, whereas inhibition of PI3K relieves this negative feedback and allows increased expression of upstream genes.

Fig. 6.

Close modal

A key theme that emerged from this analysis is that genes directly involved in upstream signaling through the PI3K pathway showed significantly increased mRNA expression upon treatment of sensitive cells with GDC-0941. For example, mRNA levels of ERBB3 and the adaptor protein IRS2 are both significantly induced upon GDC-0941 treatment (Fig. 6B). The gene products encoded by these genes have been associated with the regulation of the activity of the PI3K pathway (7, 17), and the observed increase in expression is consistent with the existence of negative feedback loops that maintain low expression levels of these upstream pathway regulators when the PI3K pathway is actively signaling (Fig. 6C). Consistent with this hypothesis, we found that ERBB3 and IRS2 were in contrast expressed at significantly lower levels at baseline in cells with PIK3CA-activating mutations compared with isogenic WT cells. Paradoxically, low expression of key upstream activators such as ERBB3 and IRS2 may serve as a sign that the PI3K signaling pathway is turned on in a particular tumor (Fig. 6C). We suggest that genes that show differential expression between sensitive and resistant cell lines as well as reciprocal expression changes in response to pathway inhibition and activation might be more likely to comprise bona fide predictive biomarkers that will translate from the in vitro setting to clinical practice. Such a signature could readily be formatted into an exploratory quantitative real-time PCR–based diagnostic test that could be used to test this hypothesis in patients enrolled in clinical trials for selective inhibitors of class I PI3K.

Discussion

The PI3K/AKT signaling axis is generally regarded as the most frequently deregulated pathway in solid tumors because nearly every major node can be activated and lead to constitutive signaling (40). In breast cancer, amplification of HER2, loss of the tumor suppressor PTEN, and activating mutations in PIK3CA and AKT1 have all been reported to occur with reasonable frequency and are thought to be key drivers of enhanced proliferation and survival (9, 15, 40). Tumor genetics thus clearly implicates this pathway as an important focal point for the development of novel targeted therapeutics, and indeed intensive efforts are under way to develop pan- and isoform-selective inhibitors of PI3K, dual PI3K and mTOR inhibitors, and selective inhibitors of AKT and mTOR (10). We have focused in this report on identifying molecular predictors of response to a pan-inhibitor of class I PI3K in breast cancer, starting with key alterations implicated by tumor genetics but also casting a wider net to identify more generic signatures of drug sensitivity.

Our studies showed that although breast cancer models exhibited strong overall sensitivity to the PI3K inhibitor GDC-0941, oncogenic alterations in PIK3CA and HER2 were predictive of sensitivity to this agent both in vitro and in vivo. Similar findings have been reported for the PI3K/mTOR inhibitor BEZ-235 (41). In addition, we found models harboring concomitant activating alterations in core pathway components (e.g., PIK3CA mutations and HER2 amplification) were exquisitely sensitive to GDC-0941 both in vitro and in vivo, suggesting that perhaps selection for multiple oncogenic events renders a tumor model particularly dependent on continued signaling through the pathway and vulnerable to inhibition with a selective inhibitor. Notably, we observed tumor regressions in xenograft models with HER2 amplification and/or PIK3CA mutations, suggesting the possibility that GDC-0941 could have single-agent tumor killing activity in this setting. Some but not all models harboring PTEN were also sensitive to GDC-0941, suggesting that PTEN may have lower value as a predictive biomarker of single-agent GDC-0941 response and perhaps that other factors may influence response in PTEN null models. It is notable that the PTEN null lines that are insensitive to GDC-0941 are all basal like in phenotype because we have previously shown that basal-like breast cancer cell lines exhibit an activated RAS-like transcriptional program and dependence on RAS/mitogen-activated protein/extracellular signal-regulated kinase kinase (MEK) signaling and, furthermore, that such PTEN null basal-like lines require concomitant inhibition of MEK and PI3K signaling for full suppression of tumor growth (20).

The mechanism of action of GDC-0941 seems similar to that described for other PI3K and AKT inhibitors (41). We found that levels of phospho-AKT (S473) were generally decreased by GDC-0941 treatment in the majority of cell lines in a dose-dependent manner, independent of whether a cell line was sensitive or resistant to GDC-0941. This finding suggests some utility in monitoring this epitope as a pharmacodynamic biomarker of target modulation, although PI3K can clearly signal through other effectors (29) and pAKT may not be a universal response marker. In contrast, although we found that the sensitive cell lines we examined showed substantial GDC-0941–mediated downregulation of mTORC1-regulated proteins such as pp70S6K1, pS6, and p4EB-P1, this response was either weaker or absent in resistant cell lines. This suggests that although PI3K may signal through multiple upstream mechanisms in sensitive lines, the ability to couple these mechanisms to mTORC1 signaling may be a hallmark of sensitive models. Similarly, we saw evidence of G1 arrest and downregulation of key components of the cell cycle machinery such as Cyclin D1 and Cyclin E2 at both the mRNA and protein level in sensitive cell lines. Finally, we observed substantial induction of apoptotic markers and apoptosis itself specifically in sensitive cell lines harboring alterations in HER2, PIK3CA, and PTEN, indicating that in some contexts, single-agent GDC-0941 may exert cytotoxic effects in addition to cytostatic effects on tumor cells.

Our studies also have implications for the development of diagnostic tests aimed at predicting response to GDC-0941 from analysis of archival tissue. First and foremost, such tissue should be used for the evaluation of PIK3CA mutation status, confirmation of HER2 status, and evaluation of PTEN status. It is tempting to speculate that early selection of patients based on the presence of HER2 amplification or PIK3CA mutation, perhaps at the expansion cohort stage (42), could be used to rapidly show clinical benefit and validate PI3K as a drug target. Subsequent clinical development could then focus on evaluating activity in a broader population and the utility of other candidate biomarkers in patient stratification. Toward this end, we identified a relatively large set of genes that are differentially expressed between sensitive and resistant cell lines and could potentially have utility in the identification of responsive patients.

We also identified a small collection of genes from our predictive response signature, including IRS2 and ERBB3, whose expression was feedback inhibited by PI3K signaling. These findings are reminiscent of a recent report showing ERBB3 levels are increased by treatment with the HER2/ERBB3 dimerization inhibitor pertuzumab and decreased by growth factor stimulation (43). Intriguingly from a diagnostic perspective, a phase II clinical study recently showed that low ERRB3 levels were associated with longer progression-free and overall survival in patients with platinum-resistant epithelial ovarian cancer treated with pertuzumab and gemcitabine (44), implying that low ERRB3 mRNA expression levels may be a potential prognostic and predictive diagnostic biomarker in this setting. Release of negative feedback loops has also been shown in the presence of both mTOR and RAS/mitogen-activated protein kinase pathway inhibition, with consequences such as increased levels of upstream signaling or activation of alternate pathways (20, 45–49). These feedback loops have also been shown to affect response to targeted agents. In particular, mTOR inhibition with rapamycin has been shown to induce IRS-1 expression and abrogate feedback inhibition of the pathway, resulting in AKT activation both in cancer cell lines and in patient tumors treated with the rapamycin derivative, RAD001, potentially limiting efficacy in response to rapalogs (47). Similarly, inhibition of MEK has been shown to induce compensatory upregulation of pAKT and blunt the effects of single-agent MEK inhibition (20, 46). With these parallels in mind, a clinical implication of the feedback loop we describe is that activation of upstream signaling by ERBB3 could theoretically lead to escape from the inhibitory effects of GDC-0941 through activation of compensatory pathways, potentially suggesting a need for combined targeting of PI3K and key nodes in other signaling pathways (50).

Tumor genetics clearly implicates PI3K as an attractive and important target in the treatment of breast cancer, and early focus on the identification of appropriate patient populations may enhance the speed and probability of successful development of agents that selectively target PI3K signaling. Our findings suggest that tumors with activating mutations in PIK3CA or amplification of HER2 are selectively addicted to PI3K signaling and that patients with tumors harboring these alterations are particularly good candidates for PI3K-targeting therapies. Our results also suggest the potential for broader activity outside of this patient population, and it is our hope that additional biomarkers such as the gene expression signatures we describe and other efforts in the field will help deliver on the promise of personalized cancer therapy with PI3K inhibitors.

Disclosure of Potential Conflicts of Interest

All of the authors are employees of Genentech, Inc.

Acknowledgments

We thank David Shames, Elaine Storm, Richard Neve, and Mark Merchant for comments on the manuscript; the Genentech Medicinal Chemistry group for providing GDC-0941; Jill Spoerke for the assistance with Western blots; and Zora Modrusan and the Genentech microarray core facility for the gene expression profiling efforts.

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.

References

Bachman

Argani

Samuels

, et al.

The PIK3CA gene is mutated with high frequency in human breast cancers

Cancer Biol Ther

2004

;

772

–

Samuels

Wang

Bardelli

, et al.

High frequency of mutations of the PIK3CA gene in human cancers

Science

2004

;

304

554

Shayesteh

Kuo

, et al.

PIK3CA is implicated as an oncogene in ovarian cancer

Nat Genet

1999

;

–

102

Kang

Bader

Vogt

Phosphatidylinositol 3-kinase mutations identified in human cancer are oncogenic

Proc Natl Acad Sci U S A

2005

;

102

802

–

Huang

Mandelker

Gabelli

Amzel

Insights into the oncogenic effects of PIK3CA mutations from the structure of p110α/p85α

Cell cycle

2008

;

1151

–

Miled

Yan

Hon

, et al.

Mechanism of two classes of cancer mutations in the phosphoinositide 3-kinase catalytic subunit

Science

2007

;

317

239

–

Way

Lin

Role of HER2/HER3 co-receptor in breast carcinogenesis

Future Oncol

2005

;

841

–

Carpten

Faber

Horn

, et al.

A transforming mutation in the pleckstrin homology domain of AKT1 in cancer

Nature

2007

;

448

439

–

Saal

Gruvberger-Saal

Persson

, et al.

Recurrent gross mutations of the PTEN tumor suppressor gene in breast cancers with deficient DSB repair

Nat Genet

2008

;

102

–

Workman

Clarke

Guillard

Raynaud

Drugging the PI3 kinome

Nat Biotechnol

2006

;

794

–

Lackner

Prospects for personalized medicine with inhibitors targeting the RAS and PI3K pathways

Expert Rev Mol Diagn

2010

;

–

Maira

Stauffer

Brueggen

, et al.

Identification and characterization of NVP-BEZ235, a new orally available dual phosphatidylinositol 3-kinase/mammalian target of rapamycin inhibitor with potent in vivo antitumor activity

Mol Cancer Ther

2008

;

1851

–

Folkes

Ahmadi

Alderton

, et al.

The identification of 2-(1H-indazol-4-yl)-6-(4-methanesulfonyl-piperazin-1-ylmethyl)-4-morpholin-4-yl-thieno[3,2-d]pyrimidine (GDC-0941) as a potent, selective, orally bioavailable inhibitor of class I PI3 kinase for the treatment of cancer

J Med Chem

2008

;

5522

–

Raynaud

Eccles

Patel

, et al.

Biological properties of potent inhibitors of class I phosphatidylinositide 3-kinases: from PI-103 through PI-540, PI-620 to the oral agent GDC-0941

Mol Cancer Ther

2009

;

1725

–

Stemke-Hale

Gonzalez-Angulo

Lluch

, et al.

An integrative genomic and proteomic analysis of PIK3CA, PTEN, AKT mutations in breast cancer

Cancer Res

2008

;

6084

–

Yakes

Chinratanalab

Ritter

King

Seelig

Arteaga

Herceptin-induced inhibition of phosphatidylinositol-3 kinase and Akt Is required for antibody-mediated effects on p27, cyclin D1, and antitumor action

Cancer Res

2002

;

4132

–

Shah

Hunter

Turnover of the active fraction of IRS1 involves raptor-mTOR- and S6K1-dependent serine phosphorylation in cell culture models of tuberous sclerosis

Mol Cell Biol

2006

;

6425

–

Stern

, et al.

Genetic alterations and oncogenic pathways associated with breast cancer subtypes

Mol Cancer Res

2009

;

511

–

O'Brien

Cavet

Pandita

, et al.

Functional genomics identifies ABCC3 as a mediator of taxane resistance in HER2-amplified breast cancer

Cancer Res

2008

;

5380

–

Hoeflich

O'Brien

Boyd

, et al.

In vivo antitumor activity of MEK and phosphatidylinositol 3-kinase inhibitors in basal-like breast cancer models

Clin Cancer Res

2009

;

4649

–

Jaiswal

Janakiraman

Kljavin

, et al.

Somatic mutations in p85α promote tumorigenesis through class IA PI3K activation

Cancer Cell

2009

;

463

–

Fan

Knight

Goldenberg

, et al.

A dual PI3 kinase/mTOR inhibitor reveals emergent efficacy in glioma

Cancer Cell

2006

;

341

–

Rhodes

Kalyana-Sundaram

Mahavisno

, et al.

Oncomine 3.0: genes, pathways, and networks in a collection of 18,000 cancer gene expression profiles

Neoplasia

2007

;

166

–

Baldi

Long

A Bayesian framework for the analysis of microarray expression data: regularized t -test and statistical inferences of gene changes

Bioinformatics

2001

;

509

–

Storey

Tibshirani

Statistical significance for genomewide studies

Proc Natl Acad Sci U S A

2003

;

100

9440

–

Eisen

Spellman

Brown

Botstein

Cluster analysis and display of genome-wide expression patterns

Proc Natl Acad Sci U S A

1998

;

14863

–

Page

TreeView: an application to display phylogenetic trees on personal computers

Comput Appl Biosci

1996

;

357

–

She

Chandarlapaty

, et al.

Breast tumor cells with PI3K mutation or HER2 amplification are selectively addicted to Akt signaling

PloS One

2008

;

e3065

Vasudevan

Barbie

Davies

, et al.

AKT-independent signaling downstream of oncogenic PIK3CA mutations in human cancer

Cancer Cell

2009

;

–

Huang

Manning

A complex interplay between Akt, TSC2 and the two mTOR complexes

Biochem Soc Trans

2009

;

217

–

Gottschalk

Basila

Wong

, et al.

p27Kip1 is required for PTEN-induced G1 growth arrest

Cancer Res

2001

;

2105

–

Benjamini

Hochberg

Controlling the false discovery rate: a practical and powerful approach to multiple testing

JRSSB

1995

;

289

–

300

Ivshina

George

Senko

, et al.

Genetic reclassification of histologic grade delineates new clinical subtypes of breast cancer

Cancer Res

2006

;

10292

–

301

Wang

Iglehart

Zhang

Richardson

Predicting features of breast cancer with gene expression patterns

Breast Cancer Res Treat

2008

;

108

191

–

201

Miller

Smeds

George

, et al.

An expression signature for p53 status in human breast cancer predicts mutation status, transcriptional effects, and patient survival

Proc Natl Acad Sci U S A

2005

;

102

13550

–

Sotiriou

Wirapati

Loi

, et al.

Gene expression profiling in breast cancer: understanding the molecular basis of histologic grade to improve prognosis

J Natl Cancer Inst

2006

;

262

–

van 't Veer

Dai

van de Vijver

, et al.

Gene expression profiling predicts clinical outcome of breast cancer

Nature

2002

;

415

530

–

Ganesan

Miller

Tan

A modular analysis of breast cancer reveals a novel low-grade molecular signature in estrogen receptor-positive tumors

Clin Cancer Res

2006

;

3288

–

Chin

DeVries

Fridlyand

, et al.

Genomic and transcriptional aberrations linked to breast cancer pathophysiologies

Cancer Cell

2006

;

529

–

Yuan

Cantley

PI3K pathway alterations in cancer: variations on a theme

Oncogene

2008

;

5497

–

510

Serra

Markman

Scaltriti

, et al.

NVP-BEZ235, a dual PI3K/mTOR inhibitor, prevents PI3K signaling and inhibits the growth of cancer cells with activating PI3K mutations

Cancer Res

2008

;

8022

–

Carden

Sarker

Postel-Vinay

, et al.

Can molecular biomarker-based patient selection in phase I trials accelerate anticancer drug development?

Drug Discov Today

2010

;

–

Nagumo

Faratian

Mullen

Harrison

Hasmann

Langdon

Modulation of HER3 is a marker of dynamic cell signaling in ovarian cancer: implications for pertuzumab sensitivity

Mol Cancer Res

2009

;

1563

–

Amler

Makhija

Januario

, et al.

HER pathway gene expression analysis in a phase II study of pertuzumab + gemcitabine vs. gemcitabine + placebo in patients with platinum-resistant epithelial ovarian cancer

J Clin Oncol

2008

;

5552

Carracedo

Teruya-Feldstein

, et al.

Inhibition of mTORC1 leads to MAPK pathway activation through a PI3K-dependent feedback loop in human cancer

J Clin Invest

2008

;

118

3065

–

Mirzoeva

Das

Heiser

, et al.

Basal subtype and MAPK/ERK kinase (MEK)-phosphoinositide 3-kinase feedback signaling determine susceptibility of breast cancer cells to MEK inhibition

Cancer Res

2009

;

565

–

O'Reilly

Rojo

She

, et al.

mTOR inhibition induces upstream receptor tyrosine kinase signaling and activates Akt

Cancer Res

2006

;

1500

–

Pratilas

Taylor

, et al.

(V600E)BRAF is associated with disabled feedback inhibition of RAF-MEK signaling and elevated transcriptional output of the pathway

Proc Natl Acad Sci U S A

2009

;

106

4519

–

Sos

Fischer

Ullrich

, et al.

Identifying genotype-dependent efficacy of single and combined PI3K- and MAPK-pathway inhibition in cancer

Proc Natl Acad Sci U S A

2009

;

106

18351

–

Rexer

Ghosh

Arteaga

Inhibition of PI3K and MEK: it is all about combinations and biomarkers

Clin Cancer Res

2009

;

4518

–

2010

Predictive Biomarkers of Sensitivity to the Phosphatidylinositol 3′ Kinase Inhibitor GDC-0941 in Breast Cancer Preclinical Models (original) (raw)

Abstract

Materials and Methods

Breast cancer cell lines

Inhibitors

In vitro cell viability studies

Xenograft studies

Western blot analyses

Flow cytometry

Gene expression microarray studies and biomarker analyses

Results

In vitro sensitivity to GDC-0941

Flow cytometric analysis of GDC-0941 effects on cell cycle and apoptosis

GDC-0941 treatment downregulates PI3K/AKT and mTOR signaling

Gene expression predictive markers of response to GDC-0941

Discussion

Disclosure of Potential Conflicts of Interest

Acknowledgments

References

Supplementary data