Purpose:

On the basis of strong preclinical rationale, we sought to confirm recommended phase II dose (RP2D) for olaparib, a PARP inhibitor, combined with the AKT inhibitor capivasertib and assess molecular markers of response and resistance.

Patients and Methods:

We performed a safety lead-in followed by expansion in endometrial, triple-negative breast, ovarian, fallopian tube, or peritoneal cancer. Olaparib 300 mg orally twice daily and capivasertib orally twice daily on a 4-day on 3-day off schedule was evaluated. Two dose levels (DL) of capivasertib were planned: 400 mg (DL1) and 320 mg (DL-1). Patients underwent biopsies at baseline and 28 days.

Results:

A total of 38 patients were enrolled. Seven (18%) had germline BRCA1/2 mutations. The first 2 patients on DL1 experienced dose-limiting toxicities (DLT) of diarrhea and vomiting. No DLTs were observed on DL-1 (n = 6); therefore, DL1 was reexplored (n = 6) with no DLTs, confirming DL1 as RP2D. Most common treatment-related grade 3/4 adverse events were anemia (23.7%) and leukopenia (10.5%). Of 32 evaluable subjects, 6 (19%) had partial response (PR); PR rate was 44.4% in endometrial cancer. Seven (22%) additional patients had stable disease greater than 4 months. Tumor analysis demonstrated strong correlations between response and immune activity, cell-cycle alterations, and DNA damage response. Therapy resistance was associated with receptor tyrosine kinase and RAS-MAPK pathway activity, metabolism, and epigenetics.

Conclusions:

The combination of olaparib and capivasertib is associated to no serious adverse events and demonstrates durable activity in ovarian, endometrial, and breast cancers, with promising responses in endometrial cancer. Importantly, tumor samples acquired pre- and on-therapy can help predict patient benefit.

Translational Relevance

This study demonstrates that combination therapy with inhibitors of AKT and PARP has no serious adverse events and has the potential to elicit a prolonged response in ovarian, endometrial, and breast cancers, with promising activity in endometrial cancer. This phase Ib clinical trial defines the recommended phase II dose of capivasertib and olaparib in patients with advanced and/or recurrent endometrial, triple-negative breast, ovarian, fallopian tube, or peritoneal cancer. A translational study was performed on pre- and on-treatment tumor samples, establishing a novel set of biomarkers associated with response to therapy, which may be particularly useful for patient selection in future trials. Furthermore, biomarkers of resistance were identified, which may guide future combination strategies.

PARP inhibitors (PARPi) are now standard of care for breast and ovarian cancer across a number of indications (reviewed in ref. 1). Furthermore, these agents may be relevant in endometrial cancers (2). PARPi demonstrate clear efficacy in tumors harboring aberrations in the homologous recombination (HR) DNA damage repair pathway (1) with limited activity in tumors that demonstrate innate HR proficiency. Unfortunately, PARPi activity can be of limited duration in the setting of innate resistance or the development of acquired resistance through a diverse array of mechanisms (3). Thus, there exists a clear unmet clinical need to expand the depth and durability of PARPi responses.

Conversely, activity of single agents targeting the PI3K pathway has been limited in gynecologic and breast cancers. Ovarian cancer demonstrates amplification of PI3KCA in 17% of tumors (4). However, clinical trials of PI3K pathway targeted agents, alone and in combination with chemotherapy, provide no objective benefit in ovarian cancer (5). In estrogen receptor (ER)-positive breast cancer, the PI3K inhibitor alpelisib demonstrated sufficient activity in combination with fulvestrant to receive an indication in PI3KCA-mutant tumors. In addition, everolimus combined with an aromatase inhibitor improved PFS in patients with hormone receptor–positive advanced breast cancer (6, 7). Despite having the highest reported rate of aberrations in the PI3K pathway among solid tumors (8), endometrial cancer demonstrated only modest benefit from PI3K, AKT, or mTORC inhibition as monotherapy (8).

Combination therapy has arisen as an opportunity to increase the spectrum of patients that benefit from therapy, as well as overcome mechanisms of resistance. Activity of the PI3K pathway has been shown to have a role in tumor cell survival and development of resistance to anti-cancer therapy broadly (9). We previously found treatment with PARPi led to upregulation of PI3K pathway members across a number of tumor models (10, 11). Furthermore, inhibition of the PI3K pathway leads to decreased HR repair through downregulation of BRCA1/2, with subsequent increase in DNA damage and sensitivity to PARPi (12). The AKT serine/threonine protein kinases (AKT1, AKT2, AKT3) are key PI3K pathway downstream mediators and are implicated as oncogenes by constitutive activation through mutation of AKT itself or other constituent of the pathway [loss of PTEN, activating mutation in the catalytic subunit of PI3K (PIK3CA), etc.]. Drugs targeting AKT have demonstrated activity in a number of tumor lineages and are being investigated in phase I/II/III trials (i.e., not approved yet).

Xenograft models of HR-deficient cancers have demonstrated clear synergy from the combination of PARP inhibition and PI3K inhibition (13). Objective activity has been validated in early-phase clinical trials combining these agents in breast and ovarian cancer (14, 15). However, these trials failed to reach the maximum tolerated olaparib dose due to overlapping toxicity (14, 15). AKT inhibitors (AKTi) have been well tolerated in clinical studies and demonstrate a different spectrum of toxicity from PARPi suggesting that drug combinations may be tolerated. We sought to evaluate safety and early antitumor activity of the PARPi, olaparib, in combination with two agents targeting the PI3K pathway, capivasertib (AKTi) and vistusertib (mTORC1/2 inhibitor). We report results from the capivasertib arm herein. We explored quality of life and symptom burden in a subset of patients through longitudinal patient-reported outcomes (PRO). In addition to assessing antitumor activity we also collected pre- and on-treatment tumor biopsies to evaluate predictive markers of response to therapy and explore the effect of combination treatment on relevant molecular pathways.

Study design

This was a phase Ib trial with three noncomparator arms: (i) olaparib + capivasertib given on an intermittent schedule, (ii) olaparib and vistusertib given on a continuous schedule, and (iii) olaparib and vistusertib given on an intermittent schedule. Patients were enrolled on a given arm based on slot availability. This article describes the olaparib and capivasertib arm, which started with a safety lead-in followed by a cohort expansion including ovarian, endometrial, or triple-negative breast cancer (TNBC), to further evaluate safety and efficacy and perform translational analyses. The study was conducted at The University of Texas MD Anderson Cancer Center (MDACC, Houston, TX) under an Institutional Review Board (IRB)-approved protocol. The trial was registered at clinicaltrials.gov as NCT02208375. All procedures involving human participants were carried out in accordance with the Declaration of Helsinki. All subjects provided written informed consent and the study was IRB approved.

Treatment plan

Given that a recommended phase II dose (RP2D) of the combination of 300 mg olaparib and capivasertib 400 mg, twice a day on a 4/3 schedule, had been previously identified in a phase I trial across solid tumors (16), a lead-in phase was planned to obtain additional safety data in ovarian, endometrial, and breast cancers. Cycle length was 28 days. Olaparib tablets were administered orally twice daily and capivasertib was administered orally twice daily on an intermittent schedule 4 days on, 3 days off. Dose level 1 (DL1) was olaparib 300 mg and capivasertib 400 mg. Dose level -1 (DL-1) was olaparib 300 mg and capivasertib 320 mg. Treatment was continued indefinitely until unacceptable toxicity, disease progression, patient withdrawal, use of prohibited medication, or changes in the patient's condition that rendered them ineligible for further treatment. The patient population is described in the Supplementary Materials and Methods.

PROs

A subset of enrolled patients with endometrial and ovarian cancer participated in a companion study of longitudinal PROs (expansion phase) and qualitative interviews (expansion or escalation phase) to explore pattern and severity of symptoms for women who were enrolled in this phase Ib study. The MD Anderson Symptom Inventory (MDASI)–Ovarian Cancer module was administered weekly for the first two cycles of therapy, then once per cycle, and at study exit (17). The EQ5D-5 L and FACT-ovary were administered at baseline, every two cycles, and at study exit (18–20). See Supplementary Table S1 for full schedule of instruments and timepoints and Supplementary Materials and Methods for the description of safety assessment.

Sample collection/molecular testing

Tissue was obtained through biopsy by interventional radiology at baseline and 28 days after beginning treatment for translational testing. Details of molecular analyses can be found in the Supplementary Materials and Methods.

Statistical methods

This study began as a safety lead-in followed by a cohort expansion of the combination of olaparib and capivasertib to further evaluate safety and efficacy and perform translational analyses. Patients who came off study prior to reaching the first evaluation point due to toxicity or disease progression were considered in the DLT analysis. However, patients who came off early for nonmedical reasons were not considered in the analysis and were replaced. We used descriptive statistics to summarize the demographic and clinical characteristics of patients, and calculated 90% Bayesian estimation and credible interval for the probability of DLT at each dose level. Adverse events (AE) by CTCAE 4.0 were tabulated by grade, dose level, and overall.

Longitudinal symptom burden through the first 8 weeks of treatment was assessed using mixed effect modeling. Covariates included disease site, marital status, education level, body mass index (obese or nonobese), and age (< or ≥65). Changes from baseline through cycle 2 for PROs other than MDASI were compared using paired t test and Cohen D effect size. Wilcoxon signed-rank test was used when the normality assumption was not met. All statistical procedures were performed using Stata v16 or SAS statistical software program for Windows, Version 9.4 (SAS Corporation). A P value of <0.05 was considered statistically significant.

Data availability

Sequence data from the deep targeted sequencing and RNA sequencing in this study have been deposited in the European Genome-phenome Archive (EGA, https://www.ebi.ac.uk/ega/home) under accession numbers EGAD00001006841 and EGAD00001006840, respectively.

Role of the funding source

The funding bodies did not have a role in study design, data collection, data analyses, data interpretation, or writing the article. All authors had full access to all data from the study and had final responsibility for decision to submit the article for publication.

Patient characteristics

Thirty-eight patients were enrolled on the capivasertib arm of the trial between September 9, 2015 and February 22, 2017, 14 in dose escalation and 24 in dose expansion. Patient characteristics are presented in Table 1. Median age was 61 years, patients had a median of three (1–7) lines of prior therapy, and only 1 patient had been previously treated with a PARPi. Sixteen patients (42%) had ovarian, 11 (29%) had endometrial, and 11 (29%) had TNBC. Of the patients with ovarian cancer, 27% had germline BRCA mutations and 87% were platinum resistant or refractory. Among the patients with endometrial cancer, 55% were serous histology and, of those tested, there were no germline BRCA mutations. In the TNBC cohort, 27% had a germline BRCA mutation. Thirty-six patients had biopsies during this trial (30 patients with matched pre- and on-treatment biopsies, 6 patients with pre-treatment biopsy only).

Table 1.

Subject demographics and clinical characteristics (n = 38).

CharacteristicMedian (range)
Age, years 61.0 (30.0–75.0) 
Number of prior lines of therapy (all settings) 3.0 (1.0–7.0) 
 Ovarian 2.0 (1.0–6.0) 
 Endometrial 4.0 (1.0–6.0) 
 Triple-negative breast cancer 3.0 (2.0–7.0) 
 N (%) 
ECOG performance status  
 0 28 (73.7) 
 1 10 (26.32) 
Race 
 Asian 2 (5.3) 
 Black or African American 3 (7.9) 
 Other 4 (10.5) 
 White or Caucasian 29 (76.3) 
Ethnicity 
 Hispanic or Latino 5 (15.6) 
 Not Hispanic or Latino 27 (84.4) 
Primary Cancer 
 Ovarian 16 (42.2) 
 Endometrial 11 (28.9) 
 Triple-negative breast cancer 11 (28.9) 
Ovarian cancer histology 
 High-grade serous 11 (68.7) 
 Endometrioid 1 (6.3) 
 Clear cell 1 (6.3) 
 Mixed 3 (18.7) 
Platinum sensitivity (ovarian patients) 
 Sensitive 2 (13.3) 
 Resistant/refractory 13 (86.7) 
Endometrial cancer histology 
 Endometrioid 3 (27.2) 
 Serous 6 (54.6) 
 Mixed 2 (18.2) 
Breast cancer histology 
 Invasive ductal 11 (100.0) 
Germline BRCA mutation 
BRCA1 mutation 3 (12.0) 
BRCA2 mutation 4 (16.0) 
BRCA VUS 1 (4.0) 
BRCA wildtype 18 (69.0) 
CharacteristicMedian (range)
Age, years 61.0 (30.0–75.0) 
Number of prior lines of therapy (all settings) 3.0 (1.0–7.0) 
 Ovarian 2.0 (1.0–6.0) 
 Endometrial 4.0 (1.0–6.0) 
 Triple-negative breast cancer 3.0 (2.0–7.0) 
 N (%) 
ECOG performance status  
 0 28 (73.7) 
 1 10 (26.32) 
Race 
 Asian 2 (5.3) 
 Black or African American 3 (7.9) 
 Other 4 (10.5) 
 White or Caucasian 29 (76.3) 
Ethnicity 
 Hispanic or Latino 5 (15.6) 
 Not Hispanic or Latino 27 (84.4) 
Primary Cancer 
 Ovarian 16 (42.2) 
 Endometrial 11 (28.9) 
 Triple-negative breast cancer 11 (28.9) 
Ovarian cancer histology 
 High-grade serous 11 (68.7) 
 Endometrioid 1 (6.3) 
 Clear cell 1 (6.3) 
 Mixed 3 (18.7) 
Platinum sensitivity (ovarian patients) 
 Sensitive 2 (13.3) 
 Resistant/refractory 13 (86.7) 
Endometrial cancer histology 
 Endometrioid 3 (27.2) 
 Serous 6 (54.6) 
 Mixed 2 (18.2) 
Breast cancer histology 
 Invasive ductal 11 (100.0) 
Germline BRCA mutation 
BRCA1 mutation 3 (12.0) 
BRCA2 mutation 4 (16.0) 
BRCA VUS 1 (4.0) 
BRCA wildtype 18 (69.0) 

Safety lead-in

Table 2 demonstrates the dose-limiting toxicity (DLT) estimates for the safety lead-in and expansion phases. The first 2 patients on DL1 (olaparib 300 mg; capivasertib 400 mg) experienced DLTs of diarrhea and vomiting. These toxicities occurred in the absence of maximum supportive care as patients did not report any issue until the AEs were already grade 3. Thus, 6 patients were treated on DL-1 (olaparib 300 mg; capivasertib 320 mg). There were no DLTs on DL-1; therefore, 6 additional patients were treated on DL1 with maximun supportive care. There were no DLTs on the reexplored DL1 and this was confirmed as the RP2D. An expansion phase was performed with an additional 24 patients.

Table 2.

DLT in safety lead-in and cohort expansion.

Safety lead-in (n = 14)
Dose levelN# DLTsPosterior probability (90% credible interval)Pr(DLT > 30%)
−1 0.13 (0.01–0.35) 0.082 
1a 0.30 (0.10–0.55) 0.463 
Safety lead-in (n = 14)
Dose levelN# DLTsPosterior probability (90% credible interval)Pr(DLT > 30%)
−1 0.13 (0.01–0.35) 0.082 
1a 0.30 (0.10–0.55) 0.463 
Safety lead-in and expansion phase
Dose levelN# DLTsPosterior probability (90% credible interval)Pr(DLT > 30%)
32 (24 in expansion) 0.12 (0.04–0.22) 0.004 
Safety lead-in and expansion phase
Dose levelN# DLTsPosterior probability (90% credible interval)Pr(DLT > 30%)
32 (24 in expansion) 0.12 (0.04–0.22) 0.004 

aDose used for cohort expansion.

Safety

Thirty-seven (97%) patients experienced at least one treatment-related AE (Table 3). The most frequently observed AEs of any grade were nausea (76%), anemia (63%), diarrhea (61%), elevated creatinine (58%), fatigue (53%), and hyperglycemia (50%). Grade 3/4 AEs occurred in 19 (50%) patients with some patients demonstrating multiple concurrent AEs. Through all dose levels, there were 9 (24%) dose interuptions, 8 (21%) dose reductions, and 3 (8%) patients who discontinued because of toxicity.

Table 3.

All treatment-related AEs observed in > 10% of subjects.a

Adverse eventGrade 1/2Grade 3Grade 4Any grade
Nausea 28 (73.7) 1 (2.6) 0 (0.0) 29 (76.3) 
Anemia 15 (39.5) 9 (23.7) 0 (0.0) 23 (63.2) 
Hyperglycemia 18 (47.4) 1 (2.6) 0 (0.0) 18 (50.0) 
Fatigue 18 (47.4) 2 (5.3) 0 (0.0) 20 (52.7) 
Elevated creatinine 21 (55.3) 1 (2.6) 0 (0.0) 22 (57.9) 
Leukopenia 10 (26.3) 4 (10.5) 0 (0.0) 14 (36.8) 
Diarrhea 21 (55.2) 2 (5.3) 0 (0.0) 24 (60.5) 
Vomiting 9 (23.7) 2 (5.3) 0 (0.0) 11 (28.9) 
Hypertriglyceridemia 12 (31.6) 0 (0.0) 0 (0.0) 12 (31.6) 
Mucositis 11 (28.9) 0 (0.0) 0 (0.0) 11 (28.9) 
Hypercholesterolemia 8 (21.1) 0 (0.0) 0 (0.0) 8 (21.1) 
Thrombocytopenia 4 (10.5) 1 (2.6) 0 (0.0) 5 (13.1) 
Headache 6 (15.8) 0 (0.0) 0 (0.0) 6 (15.8) 
Neutropenia 6 (15.8) 1 (2.6) 1 (2.6) 8 (21.1) 
Constipation 7 (18.4) 0 (0.0) 0 (0.0) 7 (18.4) 
Hypokalemia 9 (23.7) 0 (0.0) 0 (0.0) 9 (23.7) 
Anorexia 14 (36.8) 0 (0.0) 0 (0.0) 14 (36.8) 
Rash 7 (18.4) 0 (0.0) 0 (0.0) 7 (18.4) 
Hyponatremia 6 (15.8) 1 (2.6) 0 (0.0) 7 (18.4) 
Bladder infection 6 (15.8) 1 (2.6) 0 (0.0) 7 (18.4) 
Abdominal pain 5 (13.2) 0 (0.0) 0 (0.0) 5 (13.2) 
Allergic reaction 0 (0.0) 2 (5.3) 0 (0.0) 2 (5.3) 
Adverse eventGrade 1/2Grade 3Grade 4Any grade
Nausea 28 (73.7) 1 (2.6) 0 (0.0) 29 (76.3) 
Anemia 15 (39.5) 9 (23.7) 0 (0.0) 23 (63.2) 
Hyperglycemia 18 (47.4) 1 (2.6) 0 (0.0) 18 (50.0) 
Fatigue 18 (47.4) 2 (5.3) 0 (0.0) 20 (52.7) 
Elevated creatinine 21 (55.3) 1 (2.6) 0 (0.0) 22 (57.9) 
Leukopenia 10 (26.3) 4 (10.5) 0 (0.0) 14 (36.8) 
Diarrhea 21 (55.2) 2 (5.3) 0 (0.0) 24 (60.5) 
Vomiting 9 (23.7) 2 (5.3) 0 (0.0) 11 (28.9) 
Hypertriglyceridemia 12 (31.6) 0 (0.0) 0 (0.0) 12 (31.6) 
Mucositis 11 (28.9) 0 (0.0) 0 (0.0) 11 (28.9) 
Hypercholesterolemia 8 (21.1) 0 (0.0) 0 (0.0) 8 (21.1) 
Thrombocytopenia 4 (10.5) 1 (2.6) 0 (0.0) 5 (13.1) 
Headache 6 (15.8) 0 (0.0) 0 (0.0) 6 (15.8) 
Neutropenia 6 (15.8) 1 (2.6) 1 (2.6) 8 (21.1) 
Constipation 7 (18.4) 0 (0.0) 0 (0.0) 7 (18.4) 
Hypokalemia 9 (23.7) 0 (0.0) 0 (0.0) 9 (23.7) 
Anorexia 14 (36.8) 0 (0.0) 0 (0.0) 14 (36.8) 
Rash 7 (18.4) 0 (0.0) 0 (0.0) 7 (18.4) 
Hyponatremia 6 (15.8) 1 (2.6) 0 (0.0) 7 (18.4) 
Bladder infection 6 (15.8) 1 (2.6) 0 (0.0) 7 (18.4) 
Abdominal pain 5 (13.2) 0 (0.0) 0 (0.0) 5 (13.2) 
Allergic reaction 0 (0.0) 2 (5.3) 0 (0.0) 2 (5.3) 

aAdverse events less than 10% included if any grade 3/4 toxicity observed. Only causally related adverse events included in this table. A total of 8 patients demonstrated multiple concurrent adverse events.

Efficacy

Median duration of follow-up was 7.4 months (range, 0.7–37.2). Of 32 subjects evaluable for response, objective response rate was 19% (95% confidence interval: 7.2%–36.4%). Seven additional patients (22%) had stable disease (SD) for greater than 4 months for a clinical benefit rate (CBR) of 41%. Figure 1 provides additional detail regarding depth and duration of response. Median duration of response was 169 days. Among patients with ovarian cancer, 1 (7%) had partial response (PR) and 5 had SD > 4 months (CBR 43%). Of note, 5 of the 6 (83%) patients with ovarian cancer with clinical benefit had platinum-resistant disease. Objective response rate among patients with endometrial cancer was 44% (4/9), with an additional patient achieving SD for > 4 months (CBR 57%). Among patients with TNBC, 11% (1/9) achieved objective response and 1 patient (11%) had SD > 4 months (CBR 22%).

Figure 1.

Clinical outcomes. Waterfall plot of best response to olaparib and capivasertib among evaluable patients (n = 32; A), and swimmer plot of duration on study for all evaluable patients (n = 32; B). C, Longitudinal nausea and fatigue during first 8 weeks of therapy, fatigue between responders and non-responders, and interference with physical functioning (walking, activity, work) between responders and non-responders.

Figure 1.

Clinical outcomes. Waterfall plot of best response to olaparib and capivasertib among evaluable patients (n = 32; A), and swimmer plot of duration on study for all evaluable patients (n = 32; B). C, Longitudinal nausea and fatigue during first 8 weeks of therapy, fatigue between responders and non-responders, and interference with physical functioning (walking, activity, work) between responders and non-responders.

Close modal

PROs

A total of 15 patients during the expansion phase participated in the companion PRO study, 7 women with endometrial cancer and 8 with ovarian cancer. Fatigue significantly worsened over time (P = 0.04). Those who had clinical benefit reported significantly increased symptom burden of fatigue over time (P < 0.0001) and increased interference with physical functioning subscore compared with those who did not (P = 0.004). There was no significant change in nausea from baseline, and overall symptom burden from nausea remained in the mild range (Fig. 1).

DNA analysis

To investigate potential biomarkers of sensitivity and resistance to the drug combination, we combined mutations and small insertions/deletions (in/dels) from whole-exome sequencing (WES) data (Fig. 2A; Supplementary Fig. S1). Alterations in several genes were found exclusively in the progressive disease (PD) group, such as KRAS and FGFR2. Alterations in the SWI/SNF complex gene SMARCA2 and cell cycle–related genes CDKN2C and CDC27 were found exclusively in the SD and PR groups, as were alterations in DNA repair related gene ATR. PTEN mutations were found in 2 patients, 1 patient with a PR and 1 patient with a SD. We assessed association between individual gene alterations and response, and found a total of 12 genes with P < 0.10 (KDR, MED12, NTRK2, SPOP, PTEN, FGF5, RARA, ATR, CDC27, PIK3CG, SMARCA2, and TSC2); however, we note that due to small sample size, none of these genes pass multiple correction testing (FDR < 0.05). Interestingly, alterations in other PI3K-AKT pathway members were not associated to outcome.

Figure 2.

Molecular mechanisms driving the response to therapy. DNA alterations. A, Cohort distribution for DNA alterations by response category. Mutations were aggregated from CLIA panel and WES; insertion and deletions were called from WES. Alterations were aggregated across pre- and on-treatment samples for each patient. B, Pathway enrichment for DNA alterations by response group. For each response group (PR, SD, and PD), pathways that had an overrepresentation of altered genes (using both mutation and insertion/deletions) were assessed. Pathway categories are shown on the plot according to number of genes hit in the pathway as well as number of patients with gene alterations, colored according to response group. FDR shown on plot is the minimum of the set of pathways that were collapsed. Full pathway table of top enriched pathways can be found in Supplementary Table S2. Gene expression. C, Heatmap representing the hierarchical unsupervised clustering of the RNA expression in pre- and on-treatment sample of all patients. Only the most significant genes between PR and PD patients were included. D, GSEA was performed to compare the expression of genes between PR and PD patients. The most representative pathways enriched in either PR and PD patient's samples were included. Protein expression. E, RPPA analysis was used to determine protein alteration in the on-treatment samples compared to pre-treatment. Level of total PARylation and phosphorylated GSK3a-b (S21/S9) was measured in the different outcome groups. A paired t test was used to assess the significant differences between the different groups. P ≤ 0.05 is considered significant. F, Heatmap representing the hierarchical unsupervised clustering of the on- to pre-treatment ratio of all proteins and patient samples. Most relevant proteins involved in the cell cycle and DNA damage response, signaling pathways, immune system, and epigenetics are identified. G, Pathway analysis comparing pre, on, and on to pre change in all samples. The heatmaps were constructed using hierarchical unsupervised clustering of both the pathway scores and the samples. A full list of predictors used to calculate the pathways can be found in Supplementary Table S3. Kaplan–Meier curves were built using the days on-treatment for patients with high and low RTK and RAS-MAPK pathway activity pre-treatment (H), replication stress and DNA damage response on-therapy (I), and immune/IFN gene expression on-therapy (J). Patients with RAS mutation were included in the high RTK-RAS-MAPK group. Only patients with RNA sequencing data were included in the immune/IFN Kaplan–Meier curve. K, Kaplan–Meier curves showing PI3K-AKT-mTOR inhibition status with and without DNA damage response. Each patient group was defined on the basis of the pathway scores showed in G.

Figure 2.

Molecular mechanisms driving the response to therapy. DNA alterations. A, Cohort distribution for DNA alterations by response category. Mutations were aggregated from CLIA panel and WES; insertion and deletions were called from WES. Alterations were aggregated across pre- and on-treatment samples for each patient. B, Pathway enrichment for DNA alterations by response group. For each response group (PR, SD, and PD), pathways that had an overrepresentation of altered genes (using both mutation and insertion/deletions) were assessed. Pathway categories are shown on the plot according to number of genes hit in the pathway as well as number of patients with gene alterations, colored according to response group. FDR shown on plot is the minimum of the set of pathways that were collapsed. Full pathway table of top enriched pathways can be found in Supplementary Table S2. Gene expression. C, Heatmap representing the hierarchical unsupervised clustering of the RNA expression in pre- and on-treatment sample of all patients. Only the most significant genes between PR and PD patients were included. D, GSEA was performed to compare the expression of genes between PR and PD patients. The most representative pathways enriched in either PR and PD patient's samples were included. Protein expression. E, RPPA analysis was used to determine protein alteration in the on-treatment samples compared to pre-treatment. Level of total PARylation and phosphorylated GSK3a-b (S21/S9) was measured in the different outcome groups. A paired t test was used to assess the significant differences between the different groups. P ≤ 0.05 is considered significant. F, Heatmap representing the hierarchical unsupervised clustering of the on- to pre-treatment ratio of all proteins and patient samples. Most relevant proteins involved in the cell cycle and DNA damage response, signaling pathways, immune system, and epigenetics are identified. G, Pathway analysis comparing pre, on, and on to pre change in all samples. The heatmaps were constructed using hierarchical unsupervised clustering of both the pathway scores and the samples. A full list of predictors used to calculate the pathways can be found in Supplementary Table S3. Kaplan–Meier curves were built using the days on-treatment for patients with high and low RTK and RAS-MAPK pathway activity pre-treatment (H), replication stress and DNA damage response on-therapy (I), and immune/IFN gene expression on-therapy (J). Patients with RAS mutation were included in the high RTK-RAS-MAPK group. Only patients with RNA sequencing data were included in the immune/IFN Kaplan–Meier curve. K, Kaplan–Meier curves showing PI3K-AKT-mTOR inhibition status with and without DNA damage response. Each patient group was defined on the basis of the pathway scores showed in G.

Close modal

Given the small number observed for both somatic mutations and in/dels, we aggregated gene level alterations to higher functional pathways. Reactome was used to determine whether the presence of single-nucleotide variants and in/dels of genes in shared pathways was associated with patient response group (Fig. 2B; Supplementary Table S2). Aberrant cell-cycle checkpoint pathways and signaling NTRK pathways were highly enriched in the PR group, while “signaling by nuclear receptor pathway” was enriched in the SD group. Strong enrichment of cell cycle and mitosis alterations was found for both PR and SD outcomes. Conversely, several pathways were enriched in the PD group such as the signaling by MAPK family members, NOTCH, WNT, cytokine signaling, and ERBB4 receptor tyrosine kinase (RTK). Transcriptional regulation by TP53 and the AP-2 transcription factor family was also enriched in the PD group. Interestingly, although FGFR gene alterations were only found in the PD group, aberration in the FGFR signaling pathway was associated with the PR group. This contradictory result could be attributed to the small number of patients in the PR group.

RNA analysis

To further investigate mechanisms of sensitivity and resistance to the drug combination, tumor samples were analyzed by RNA sequencing. A principal component analysis (PCA; Supplementary Fig. S2) did not demonstrate clustering based on tissue sites, suggesting no tissue-dependent response and thus all tumors were analyzed as a set. Furthermore, RNA-based clustering was not dependent on response to therapy or treatment status (pre- or on-therapy). The comparison of gene expression between PR and PD tumors showed enrichment of immune-related genes in the PR group both pre- and on-treatment (Fig. 2C). Indeed, most samples from the PD group had markedly lower HLA expression, suggesting reduced antigen presentation by tumor cells. This finding was supported by decreases of other genes involved in immune responses, as shown in Fig. 2C. Gene set enrichment analysis (GSEA; ref. 21) demonstrated that PR pre-treatment samples were enriched in genes involved in IFN signaling, complement cascade, and FCGR activation. Conversely, genes involved in DNA methylation, epigenetic regulation of gene expression, complex I biogenesis, and respiratory electron transport were enriched in the PD group. In on-treatment samples, immune signatures were strongly associated to PR (IFN signaling, adaptive immune response, antigen presentation, and cytokine signaling) and increased metabolism was associated to PD (biological oxidations, fatty acid and steroid metabolism, peroxisomal protein import; Fig. 2D). To further our findings, a gene set variation analysis (GSVA) was used to investigate pathways altered by treatment. As shown in Supplementary Fig. S3, G2–M DNA damage checkpoint, Myc and E2F targets as well as WNT beta-catenin signaling were decreased on therapy. Because AKTi block cells in G1–S-phase, these alterations strongly suggest reduced proliferation, and thus reduced replication stress. Four samples (two PD and two SD) did not have any major alteration of these pathways, suggesting the AKT pathway was not fully inhibited at the doses delivered or at the time of biopsy.

Protein analysis

Pre- and on-treatment tumors were analyzed by reverse phase protein array (RPPA) and a PCA (Supplementary Fig. S2) revealed clustering based primarily on treatment status. Importantly, no clustering by tissue site was observed. A few samples from the PD group (PT-8, PT-17, PT-21, PT-26) clustered separately from the rest and all of those samples, except PT-17 also showed a close relationship between pre- and on-treatment, indicating that the treatment did not induce protein network rewiring that overrode the patient-specific protein network. In Volcano plots (Supplementary Fig. S4), PARylation (PAR) was the most downregulated protein, followed by the oxidative stress sensor DJ1, the AKT downstream target p-GSK3a-b (S21/S9), and the replication stress protein RPA32. Proteins upregulated included stress response and apoptotic proteins HSP27, cleaved-caspase-3, and c-IAP2, as well as hormone-related p-ERα (S118) and GATA3. Dimethylated lysine 9 Histone-H3 was also increased, supporting activation of the DNA damage response. PARylation and p-GSK3a-b (S21/S9) were used as predictors of target engagement by olaparib and capivasertib. As shown in Fig. 2E, PARylation was low in all on-treatment samples, although only SD and PD samples had a significant decrease when comparing matched on- and pre-treatment samples. This could be explained by a lower PARylation level in the PR samples prior to treatment, although this was not significant. AKT inhibition by capivasertib was more pronounced in the PR groups than in the SD and PD groups, as shown by a significant decrease in p-GSK3a-b (S21/S9).

Unsupervised hierarchical clustering was used to compare expression of proteins in pre- and on-treatment samples as well as changes induced by treatment (Fig. 2F; Supplementary Fig. S5). Pre-treatment samples showed two major clusters with one enriched in samples from the PD and SD groups (CL2). This cluster was associated with high PI3K-AKT-mTOR activity, as indicated by the strong phosphorylation of AKT, 4E-BP1, GSK3a-b, mTOR, and S6. In CL1, a subset of samples from the PD group were enriched in high RTK and RAS-MAPK signaling, as indicated by an elevated phosphorylation of EGFR, MAPK, and MEK1. One sample with KRAS mutation (p6) and all FGFR2-mutated (p14, p22, p30) samples were associated with high MAPK pathway activity and PD. Furthermore, two samples from the PR group had strong immune markers (CD4, CD45, and HLAs) consistent with the findings from RNA expression data. The on-treatment samples showed three clusters with two enriched in samples from the PD group (CL2, CL3). CL2 had a high expression of proteins from the mTOR pathway, as well as increased expression and demethylation of histone-H3. CL3 showed high activity of the PI3K-AKT-mTOR pathway with high phosphorylation levels of AKT, GSK3a-b, mTOR, and S6. Interestingly, sample 28 from the PR group was part of this cluster. Patient 28 displayed BRCA2 gene mutation, which could explain sensitivity to treatment in spite of having a high PI3K-AKT-mTORC activity on-treatment. Finally, CL1 was enriched in PR and SD samples and showed a strong RTK and RAS-MAPK activity, which is consistent with a compensatory response to effective AKT inhibition (22, 23). This is markedly different from pre-treatment samples, where high RTK and RAS-MAPK activity was associated with a poor response.

In contrast to the clustering of post-treatment samples, clustering of protein changes induced by treatment resulted in four clusters that separated mostly by outcome (Fig. 2F). CL1, enriched in PD and SD, showed increased histone alterations (Histone-H3, DM-Histone-H3, Ub-Histone-H2, p-RPA32), consistent with replication stress and DNA damage response. Although PI3K-AKT pathway activity was reduced, these samples had increased mTOR pathway activity (p-4E-BP1 and p-S6) and a moderate RTK-MAPK pathway activation (p-MEK1, p-Met, p-EGFR, p-IGF1R) potentially due to release of feedback inhibition in response to AKT inhibition. These events may be linked as mTORC pathway activity is downstream of RTK-MAPK pathway activity in many epithelial cells (24). CL2, which contains 2 PR patients and 1 PD patient showed inhibition of the PI3K-AKT-mTORC pathway and an induction of DNA damage response as shown by increased p-H2AX and G2–M DNA damage checkpoint proteins p-ATR, p-WEE1, and p-CDC2. CL3 was enriched for both PR and SD patients. This cluster showed the overall strongest PI3K-AKT-mTOR pathway inhibition, as well as increased RAS-MAPK pathway activity potentially due to strong PI3K-AKT pathway inhibition. This cluster also demonstrated activation of several proteins involved in the DNA damage response. Finally, PD-enriched CL4 showed increased PI3K-AKT-mTOR pathway activity (p-AKT, p-GSK3a-b, p-mTOR, p-4EBP1, and p-S6) consistent with bypass of the AKTi.

To investigate the roles of cancer-associated pathways, we used calculated pathway scores from RPPA data (Fig. 2G). Unsupervised clustering of these scores demonstrated that although pre-treatment samples did not cluster by outcome, high RTK or RAS-MAPK activity prior to treatment is, in most cases, associated with a poor outcome. In on-treatment samples, the clustering showed that histone alterations and mTOR pathway activity are associated with the PD group. Conversely, in on-therapy biopsies responders showed low AKT activity (p-GSK3b and mTORC) and a high DDR and G0–G1 cell-cycle arrest, suggesting a mechanistic response to the drugs. These samples also displayed high RTK, RAS-MAPK, and PI3K-AKT (AKT and upstream) signaling activity, suggesting activation of compensatory mechanisms due to inhibition of negative feedback loops. Interestingly, the samples also displayed a high immune checkpoint activity, which again support the involvement of the immune system in the response. Finally, by comparing pre-treatment score changes, four types of responses to therapy were observed: (i) inhibition of AKT signaling with an increased mTOR/S6 pathway and histone alteration was associated with bad outcome; (ii) inhibition of AKT signaling with almost no change in other pathways, suggesting an indifference to therapy was associated with a poor outcome; (iii) increased AKT and mTOR signaling as well as increased G1–S cell-cycle phase was associated with therapy resistance; and (iv) inhibition of AKT and mTORC/S6 activity as well as a reduced cell-cycle progression was associated to good outcome. In responders, DNA damage and immune checkpoints as well as compensatory mechanisms (RTK and MAPK) were activated.

Biomarkers of response to therapy

Kaplan–Meier curves were constructed with the different categories of potential biomarkers (Fig. 2H–K; Supplementary Fig. S6). Integration of the DNA and RNA analysis revealed an association between high RTK and RAS-MAPK pathway activity in pre-treatment samples and a bad outcome [median 83.5 compared with 151.5 days in control patients (P = 0.0054)]. Patients with high replication stress and DNA damage response in on-treatment samples had reduced time on therapy [median 86 compared with 152 days in control patients (P = 0.0041)]. Strikingly, patients with a high IFN pathway and immune signature in on-treatment sample were treated for a longer period of time (275 days) compared with control patients (85 days; P = 0.038). Combined AKT and S6 pathway activation with concurrent induction of the DNA damage was associated with a good outcome (Fig. 2K). Taken together, these analyses reveal a potential set of biomarkers that could improve patient selection for this particular combination and help predict outcome.

PDX model

Our clinical trial was of olaparib and capivasertib. Thus, the trial did not include translational analysis of tumors on PARPi and AKTi monotherapies, which would have been useful in understanding how PARP and AKT co-inhibition alters the gene and protein expression in the tumors. Specifically, the trial cannot provide “contribution of components” which is important for future studies. To investigate independent contributions of olaparib and capivasertib in patient response and identify and validate potential biomarkers involved in the synergistic activity of the combination of these drugs, we assessed the effects of monotherapy and combination therapy in a PARP-resistant TNBC patient-derived xenograft (PDX) model (Fig. 3; Supplementary Fig. S6). NGS immune-deficient mice were treated for five cycles with either vehicle (n = 6), olaparib (n = 7), capivasertib (n = 7), or their combination (n = 6). Tumor growth monitoring confirmed resistance to olaparib. Capivasertib alone slightly decreased tumor growth, while the combination had a synergistic activity (P < 0.001; Fig. 3A). Unsupervised clustering of the protein expression data after five cycles of treatment demonstrated highly conserved protein changes across tumors based on treatment (Fig. 3B). The PARP and AKTis both displayed target engagement, as indicated by decreased PARylation and p-GSK3a-b. Although the olaparib monotherapy group was similar to the vehicle-treated group, a slight decrease in G2–M-phase proteins (CCNB1 and CDK1) was observed, suggesting a decrease in cell-cycle progression. Capivasertib monotherapy triggered several changes with a subset being conserved in the combination-therapy group, such as decreased phosphorylation levels of GSK3a-b, S6, CCNB1, and RPA32 and increased expression of p16INK4a, CCND1, p-AKT, and p-WEE1. Conversely, some changes induced by capivasertib were reversed by the addition of olaparib. For example, several structural proteins (p-cadherin, d-a-tubulin, b-actin, and fibronectin) were decreased following capivasertib monotherapy but not following combination therapy. Others such as PTEN, PDGFR-β, and PARylation were increased exclusively in the capivasertib group. The addition of olaparib reversed this phenotype returning levels to that of vehicle-treated tumors. Interestingly, the drug combination altered some proteins independently from olaparib or capivasertib monotherapy. Among these changes, we observed an increase in several stress-related proteins (HSP70, p-ATR, and BCL-xL) and a decrease in the epigenetic mediators BRD4 and ARID1A; the DNA damage checkpoint proteins p-chk2 and p-cdc2; and signaling molecules p-MET, p-srcpErk5, and p-MEK1. Pathway score analysis revealed that PARP itself did not alter major pathway activity, except for a slight G2–M-phase decrease. AKTi showed a strong decrease of PI3K-AKT-mTOR signaling. The PARP and AKTi combination blocked both PARylation and PI3K-AKT-mTOR pathway activity. In addition, G2–M-phase, total and dimethylated histone-H3, and p-S6 expression were reduced, while the G0–G1-phase increased in combination-treated tumors.

Figure 3.

PDX model. A TNBC PDX model was implanted in NGS mice and treated with vehicle, olaparib, capivasertib, or their combination. Tumor growth was assessed (A) and proteins were analyzed by RPPA (B). The heatmap representing the hierarchical unsupervised clustering of all proteins with different level of expression across treatment groups. Groups of proteins that were altered by AKT only (green), AKT and the combination (blue), PARP and the combination (purple), and the combination alone (red) were identified. A Student t test was used to compare the tumor growth between control mice and mice treated with olaparib, capivasertib, or their combination. **, P < 0.01; ***, P < 0.001. § indicates a significant difference between the tumor size of capivasertib-treated mice and mice treated with the combination.

Figure 3.

PDX model. A TNBC PDX model was implanted in NGS mice and treated with vehicle, olaparib, capivasertib, or their combination. Tumor growth was assessed (A) and proteins were analyzed by RPPA (B). The heatmap representing the hierarchical unsupervised clustering of all proteins with different level of expression across treatment groups. Groups of proteins that were altered by AKT only (green), AKT and the combination (blue), PARP and the combination (purple), and the combination alone (red) were identified. A Student t test was used to compare the tumor growth between control mice and mice treated with olaparib, capivasertib, or their combination. **, P < 0.01; ***, P < 0.001. § indicates a significant difference between the tumor size of capivasertib-treated mice and mice treated with the combination.

Close modal

This phase Ib study with planned expansions in recurrent ovarian, endometrial, and TNBC confirmed the RP2D for olaparib and capivasertib. In future trials, olaparib should be given at a dose of 300 mg twice daily continuously and capivasertib at a dose of 400 mg twice daily on a 4-day on, 3-day off schedule. Importantly, there were no unexpected safety signals in this cohort of breast and gynecologic cancers. Furthermore, there was encouraging clinical activity in a highly pretreated cohort of patients, including impressive durable responses in women with recurrent endometrial cancer, which is typically resistant to therapy. The extensive translational analyses including analysis of change in protein expression and pathway activity on therapy provide critical molecular insights regarding predictors of response and resistance that could be applied in future trials.

Although AEs were common on this trial, the majority were grade 1 or 2 and could be mitigated with protocol-directed supportive care. AEs were as expected, including known class effects from PARP inhibition and AKT inhibition such as fatigue, gastrointestinal, and hematologic toxicity. Our experience with the first 2 patients on study highlights the need for early intervention for gastrointestinal effects including nausea and diarrhea. After these two early DLTs occurred, our team increased pre-treatment counseling and encouraged patients to report side effects early so that severe toxicity could be avoided.

Overall, the combination of olaparib and capivasertib appeared to have no serious AEs from a PRO perspective. Despite frequency of side effects noted such as nausea, anemia, and fatigue, the mean symptom burden was in the mild range. Interestingly, patients who received clinical benefit from the combination therapy appeared to have more fatigue and greater interference with physical functioning compared with those who did not receive clinical benefit and experienced disease progression. Given the small numbers, this observation is hypothesis generating and suggestive of more effective target inhibition but warrants further investigation.

Antitumor activity was seen across all tumor types, regardless of presence of BRCA mutation or aberrations in the PI3K/AKT pathway. The overall objective response rate and clinical benefit rate are impressive given the high proportion of platinum resistance (87% of patients with ovarian cancer) and the low rates of BRCA mutations across the study population (0% endometrial cancer, 27% TNBC, and 27% patients with ovarian cancer). Previous studies of PARP and PI3K inhibition have demonstrated encouraging activity in similar populations; however, prior studies included higher proportions of patients with germline and somatic BRCA mutations and did not include endometrial cancers, making it difficult to perform direct comparisons (14, 15). For example, in a phase I trial, Yap and colleagues tested the combination of olaparib and capivasertib in solid tumors. They obtained a similar clinical benefit of 44.6%, but the PR proportion was higher (34%), which can be attributed to the high proportion of patients with BRCA alterations (16). Thus, although BRCA alteration is a predictor of response, our study demonstrates that the BRCA wildtype population demonstrates significant benefit from this combination therapy.

Previously, activity of single-agent PARP inhibition has been most impressive and durable in BRCA aberrant disease. Thus, it is important to demonstrate the activity of PARP combinations in patients with BRCA wildtype tumors. Indeed, the combination of PARP and AKT had unexpected activity in BRCA wildtype tumors. Furthermore, as more patients are treated with PARPi in the upfront setting in breast and ovarian cancer, there is emergent need to overcome PARP resistance through combination therapy. We have previously shown that the PI3K/AKT pathway is upregulated in response to PARP inhibition, even after only a short window of treatment (10). The combination of PARP and AKT inhibition holds promise to reverse or prevent the emergence of PARP resistance, and further studies are necessary to explore this potential. Importantly, this is the first study to show significant activity of PARP inhibition, albeit as part of a combination therapy, in endometrial cancer. It is particularly intriguing that there was clinical activity in this study regardless of endometrial cancer histology or presence of PI3K/AKT pathway aberrations. Moving forward, it will be essential to determine the relative contribution of the PARPi versus the AKTi in this population. Of note, in prior studies, activity of capivasertib monotherapy was limited to tumors with known PI3K pathway aberrations (25, 26). The TNBC PDX model that we explored in this study helps elicit the contribution of each drug and the synergistic effect of the combination. Indeed, this PARP-resistant model had almost no protein changes with PARPi monotherapy, while AKT inhibition increased DNA damage checkpoint proteins. Importantly, the combination of both drugs drastically reduced tumor growth likely through the induction of major stress responses and decreased cell-cycle progression that was not apparent with either agent alone.

By collecting samples pre- and on-therapy from each patient and performing extensive DNA, RNA, and protein analysis, we demonstrated that longitudinal analyses can help identify patients likely to benefit from the olaparib and capivasertib combination early during the course of treatment, as well as identify potentially targetable mechanisms of resistance. This could be used to determine whether to continue or change therapy as well as to determine whether additional drugs should be added to the therapy regimen. We demonstrated, through a pathway-oriented analysis, that patients with cell-cycle and DNA damage repair pathway alterations, as well as patients with high immune/IFN activity are more likely to respond to therapy. Because of the alteration in the immune activity of responders and because immunotherapy in combination with PARP or PI3K pathway inhibitors has shown clinical activity in several models as well as patients, it would be interesting to test whether the addition of immunotherapy in combination with olaparib and capivasertib could increase the number of patients with clinical benefit. Conversely, patients with KRAS mutations or with elevated RTK and RAS-MAPK pathway activity in pre-therapy biopsies were resistant to the drug combination. These patients will likely require a different therapy approach. We previously demonstrated that RAS-mutant tumors are sensitive to the combination of PARP and MEK inhibitors, which suggest these patients could benefit from such therapy (11). There was also a strong association between changes in epigenetic mediators and metabolism with therapy resistance, which suggest a proficient DNA repair mechanism and a possible bypass of the AKT pathway inhibition through mTOR activation. Indeed, previous studies demonstrated that epigenetic events are involved in the resistance to PARPi through the protection of the replication fork and reduced replication stress (27–29). Furthermore, resistance to AKTi has been previously associated with increased mTOR signaling, mitochondrial biogenesis, and oxidative phosphorylation (30–32). When comparing on- with pre-treatment samples, we observed major protein network rewiring in responders, consistent with the inhibition of negative feedback loops (increased RTK and RAS-MAPK activity; ref. 33) as well as the induction of a stress response (reduced cell-cycle progression and induced DNA damage response). Conversely, several progressing tumors showed signs of incomplete AKT pathway inhibition, potentially as the result of activation of bypass mechanisms. Interestingly, protein data indicate that inhibition of the mTOR axis along with AKTi could improve responses and that the mTOR axis might be a significant contributor to DNA damage repair in response to PARPi (34, 35). Unfortunately, due to the small number of patients, we were unable to analyze the translational data by cancer type, which would have been informative on cancer-specific biomarkers. In a larger trial, it will be important to investigate each cancer type separately and determine whether cancer-specific biomarkers can improve patient selection. We are currently designing a trial in which we utilize specific markers of benefit to direct patients to the combination of olaparib and capivasertib based on CLIA-compliant results in pre- and on-treatment biopsies. Longitudinal analyses represent a clinical challenge, because serial biopsies might be impossible to collect in some patients and there is an added cost and potential for morbidity. It will be important in future studies to determine whether noninvasive approaches, such as liquid biopsies, could be used to inform on the adaptive responses of the tumor to therapy and help identify patients likely to benefit PARPi and AKTi combination. Furthermore, liquid biopsies could identify patients likely to benefit from addition of other drugs such as immune checkpoint inhibitors.

In summary, this study found that the combination of olaparib with capivasertib, demonstrated encouraging activity with acceptable toxicity in endometrial, ovarian, and TNBC cancers. The striking response rates, particularly in endometrial cancer that often is resistant to therapy, support further exploration of the combination. Furthermore, we have identified a series of biomarkers associated with response (cell-cycle gene aberration, high IFN signaling, intact antigen presentation, inhibition of AKT pathway activity in on-therapy biopsies), and resistance (KRAS and FGFR2 mutations, RTK and RAS-MAPK pathway activity, high epigenetic regulation and oxidative phosphorylation) that, following confirmation in additional studies, could be used to identify and select patients most likely to benefit. On the basis of our data, treatment genomic evaluation may inform therapeutic decision making by enabling strategic therapeutic pivoting to agents targeting adaptive responses. Furthermore, although we demonstrated that serial on-therapy biopsies provided the most predictive information for patient outcomes, predictive biomarkers in the pre-treatment sample had considerable ability to identify patients likely to benefit.

S.N. Westin reports grants from AstraZeneca during the conduct of the study. S.N. Westin also reports personal fees from AstraZeneca, Merck, Pfizer, Eisai, and Zentalis; grants and personal fees from Clovis Oncology, GSK/Tesaro, Roche/Genentech, and Novartis; grants from Bayer, Mereo, and Cotinga Pharmaceuticals outside the submitted work. J.K. Litton reports other support from Pfizer/Medivation, Genentech, Zenith, and AstraZeneca during the conduct of the study, as well as other support from EMD Serono, Merck, and GSK outside the submitted work. J.K. Litton also reports membership on advisory committees or review panels, board membership, etc. in the past include for AstraZeneca, Ayala, and Pfizer (all uncompensated), as well as participation on review panels for NCCN, ASCO, NIH PDQ, and SITC. B. Fellman reports grants from NIH during the conduct of the study. Y. Yuan reports personal fees from AbbVie, Amgen, BeyondSpring Pharmaceuticals, Boehringer Ingelheim Pharmaceuticals, Bristol Myers Squibb, Servier Pharmaceuticals, Starpax Pharmaceuticals, and Vertex Pharmaceuticals outside the submitted work. N. Kabil is an employee of AstraZeneca LP. P.T. Soliman reports grants from Novartis and Incyte, as well as other support from Amgen outside the submitted work. M. Frumovitz reports personal fees from Stryker and Seagen, as well as grants from AstraZeneca and GlaxoSmithKline outside the submitted work. K.M. Schmeler reports grants, non-financial support, and other support from AstraZeneca during the conduct of the study. A. Jazaeri reports other support from Iovance, BMS, AstraZeneca, Aravive, Merck, and Eli Lilly, as well as personal fees from Nuprobe, AvengeBio, Genentech-Roche, EMD-Serono, Agenus, Macrogenics, TwoXAR, and Instil Bio outside the submitted work. R. Murthy reports grants from EMD Serono and Pfizer; grants and personal fees from Seattle Genetics, Genentech, and AstraZeneca; and personal fees from Puma, Novartis, and Sanofi outside the submitted work. L.A. Meyer reports other support from AstraZeneca and grants from NIH/NCI during the conduct of the study, as well as other support from GSK outside the submitted work. C.C. Sun reports other support from AstraZeneca during the conduct of the study, as well as other support from Inform Genomics outside the submitted work. A.K. Sood reports grants from NCI during the conduct of the study. A.K. Sood also reports personal fees from Kiyatec, Merck, and AstraZeneca, as well as other support from BioPath outside the submitted work; in addition, A.K. Sood has a patent for SiRNA delivery issued. R.L. Coleman reports grants and personal fees from AstraZeneca, Clovis, Genentech/Roche, Agenus, Seagen, Genelux, and Immunogen; R.L. Coleman also reports personal fees from GSK, Janssen, Karyopharm, and Novocure, as well as grants from Merck outside the submitted work. G.B. Mills reports grants, personal fees, non-financial, and other support from Amphista, AstraZeneca, Chrysallis Biotechnology, GSK, ImmunoMET, Ionis, Lilly, PDX Pharmaceuticals, Signalchem Lifesciences, Symphogen, Tarveda, Turbine, Zentalis Pharmaceuticals, Adelson Medical Research Foundation, Breast Cancer Research Foundation, Komen Research Foundation, Ovarian Cancer Research Foundation, Prospect Creek Foundation, Nanostring Center of Excellence, HRD assay to Myriad Genetics DSP to Nanostring, Genentech, and Catena during the conduct of the study. No disclosures were reported by the other authors.

S.N. Westin: Conceptualization, resources, formal analysis, supervision, funding acquisition, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. M. Labrie: Data curation, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. J.K. Litton: Investigation, writing–review and editing. A. Blucher: Data curation, formal analysis, writing–review and editing. Y. Fang: Formal analysis. C.P. Vellano: Investigation. J.R. Marszalek: Investigation, writing–review and editing. N. Feng: Investigation. X. Ma: Investigation. A. Creason: Formal analysis. B. Fellman: Investigation, writing–review and editing. Y. Yuan: Investigation, writing–review and editing. S. Lee: Formal analysis, investigation, methodology, writing–review and editing. T.-B. Kim: Data curation. J. Liu: Investigation. A. Chelariu-Raicu: Data curation. T.H. Chen: Investigation. N. Kabil: Resources, investigation. P.T. Soliman: Investigation, writing–review and editing. M. Frumovitz: Investigation, writing–review and editing. K.M. Schmeler: Investigation, writing–review and editing. A. Jazaeri: Investigation, writing–review and editing. K.H. Lu: Investigation, writing–review and editing. R. Murthy: Investigation, writing–review and editing. L.A. Meyer: Investigation, writing–review and editing. C.C. Sun: Investigation, writing–review and editing. A.K. Sood: Investigation, writing–review and editing. R.L. Coleman: Resources, investigation, writing–review and editing. G.B. Mills: Conceptualization, resources, supervision, funding acquisition, investigation, visualization, writing–original draft, writing–review and editing.

AstraZeneca, MDACC Moonshots Program, MDACC Support Grant CA016672 NCI SPOREs in Ovarian (CA217685) and Uterine (CA098258) Cancer, and a kind gift from the Miriam and Sheldon Medical Research Foundation. AZD5363 was discovered by AstraZeneca subsequent to a collaboration with Astex Therapeutics (and its collaboration with the Institute of Cancer Research and Cancer Research Technology Limited). Ovarian Cancer Research Alliance and Ruth and Steve Anderson, in honor of Shae Anderson Gerlinger.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Curtin
NJ
,
Szabo
C
. 
Poly(ADP-ribose) polymerase inhibition: past, present and future
.
Nat Rev Drug Discov
2020
;
19
:
711
36
.
2.
Musacchio
L
,
Caruso
G
,
Pisano
C
,
Cecere
SC
,
Di Napoli
M
,
Attademo
L
, et al
PARP inhibitors in endometrial cancer: current status and perspectives
.
Cancer Manag Res
2020
;
12
:
6123
35
.
3.
Li
H
,
Liu
ZY
,
Wu
N
,
Chen
YC
,
Cheng
Q
,
Wang
J
. 
PARP inhibitor resistance: the underlying mechanisms and clinical implications
.
Mol Cancer
2020
;
19
:
107
.
4.
Salvesen
HB
,
Werner
HM
,
Krakstad
C
. 
PI3K pathway in gynecologic malignancies
.
Am Soc Clin Oncol Educ Book
2013
.
doi: 10.1200/EdBook_AM.2013.33.e218
.
5.
Yang
J
,
Nie
J
,
Ma
X
,
Wei
Y
,
Peng
Y
,
Wei
X
. 
Targeting PI3K in cancer: mechanisms and advances in clinical trials
.
Mol Cancer
2019
;
18
:
26
.
6.
Markham
A
. 
Alpelisib: first global approval
.
Drugs
2019
;
79
:
1249
53
.
7.
Baselga
J
,
Campone
M
,
Piccart
M
,
Burris
HA
,
Rugo
HS
,
Sahmoud
T
, et al
Everolimus in postmenopausal hormone-receptor-positive advanced breast cancer
.
N Engl J Med
2012
;
366
:
520
9
.
8.
Roncolato
F
,
Lindemann
K
,
Willson
ML
,
Martyn
J
,
Mileshkin
L
. 
PI3K/AKT/mTOR inhibitors for advanced or recurrent endometrial cancer
.
Cochrane Database Syst Rev
2019
;
10
:
CD012160
.
9.
Jiang
N
,
Dai
Q
,
Su
X
,
Fu
J
,
Feng
X
,
Peng
J
. 
Role of PI3K/AKT pathway in cancer: the framework of malignant behavior
.
Mol Biol Rep
2020
;
47
:
4587
629
.
10.
Labrie
M
,
Kim
T-B
,
Ju
Z
,
Lee
S
,
Zhao
W
,
Fang
Y
, et al
Adaptive responses in a PARP inhibitor window of opportunity trial illustrate limited functional interlesional heterogeneity and potential combination therapy options
.
Oncotarget
2019
;
10
:
3533
46
.
11.
Sun
C
,
Fang
Y
,
Yin
J
,
Chen
J
,
Ju
Z
,
Zhang
D
, et al
Rational combination therapy with PARP and MEK inhibitors capitalizes on therapeutic liabilities in RAS mutant cancers
.
Sci Transl Med
2017
;
9
:
eaal5148
.
12.
Ibrahim
YH
,
García-García
C
,
Serra
V
,
He
L
,
Torres-Lockhart
K
,
Prat
A
, et al
PI3K inhibition impairs BRCA1/2 expression and sensitizes BRCA-proficient triple-negative breast cancer to PARP inhibition
.
Cancer Discov
2012
;
2
:
1036
47
.
13.
Juvekar
A
,
Burga
LN
,
Hu
H
,
Lunsford
EP
,
Ibrahim
YH
,
Balmañà
J
, et al
Combining a PI3K inhibitor with a PARP inhibitor provides an effective therapy for BRCA1-related breast cancer
.
Cancer Discov
2012
;
2
:
1048
63
.
14.
Konstantinopoulos
PA
,
Barry
WT
,
Birrer
M
,
Westin
SN
,
Cadoo
KA
,
Shapiro
GI
, et al
Olaparib and alpha-specific PI3K inhibitor alpelisib for patients with epithelial ovarian cancer: a dose-escalation and dose-expansion phase 1b trial
.
Lancet Oncol
2019
;
20
:
570
80
.
15.
Matulonis
UA
,
Wulf
GM
,
Barry
WT
,
Birrer
M
,
Westin
SN
,
Farooq
S
, et al
Phase I dose escalation study of the PI3kinase pathway inhibitor BKM120 and the oral poly (ADP ribose) polymerase (PARP) inhibitor olaparib for the treatment of high-grade serous ovarian and breast cancer
.
Ann Oncol
2017
;
28
:
512
8
.
16.
Yap
TA
,
Kristeleit
R
,
Michalarea
V
,
Pettitt
SJ
,
Lim
JSJ
,
Carreira
S
, et al
Phase I trial of the PARP inhibitor olaparib and AKT inhibitor capivasertib in patients with BRCA1/2- and non-BRCA1/2-mutant cancers
.
Cancer Discov
2020
;
10
:
1528
43
.
17.
Sailors
MH
,
Bodurka
DC
,
Gning
I
,
Ramondetta
LM
,
Williams
LA
,
Mendoza
TR
, et al
Validating the M. D. Anderson Symptom Inventory (MDASI) for use in patients with ovarian cancer
.
Gynecol Oncol
2013
;
130
:
323
8
.
18.
Rabin
R
,
de Charro
F
. 
EQ-5D: a measure of health status from the EuroQol group
.
Ann Med
2001
;
33
:
337
43
.
19.
Herdman
M
,
Gudex
C
,
Lloyd
A
,
Janssen
M
,
Kind
P
,
Parkin
D
, et al
Development and preliminary testing of the new five-level version of EQ-5D (EQ-5D-5L)
.
Qual Life Res
2011
;
20
:
1727
36
.
20.
Basen-Engquist
K
,
Bodurka-Bevers
D
,
Fitzgerald
MA
,
Webster
K
,
Cella
D
,
Hu
S
, et al
Reliability and validity of the functional assessment of cancer therapy-ovarian
.
J Clin Oncol
2001
;
19
:
1809
17
.
21.
Subramanian
A
,
Tamayo
P
,
Mootha
VK
,
Mukherjee
S
,
Ebert
BL
,
Gillette
MA
, et al
Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles
.
Proc Natl Acad Sci U S A
2005
;
102
:
15545
50
.
22.
Mendoza
MC
,
Er
EE
,
Blenis
J
. 
The Ras-ERK and PI3K-mTOR pathways: cross-talk and compensation
.
Trends Biochem Sci
2011
;
36
:
320
8
.
23.
Jahangiri
A
,
Weiss
WA
. 
It takes two to tango: Dual inhibition of PI3K and MAPK in rhabdomyosarcoma
.
Clin Cancer Res
2013
;
19
:
5811
3
.
24.
Akbani
R
,
Ng
PKS
,
Werner
HMJ
,
Shahmoradgoli
M
,
Zhang
F
,
Ju
Z
, et al
A pan-cancer proteomic perspective on The Cancer Genome Atlas
.
Nat Commun
2014
;
5
:
3887
.
25.
Capivasertib active against AKT1-mutated cancers
.
Cancer Discov
2019
;
9
:
OF7
.
26.
Li
J
,
Davies
BR
,
Han
S
,
Zhou
M
,
Bai
Y
,
Zhang
J
, et al
The AKT inhibitor AZD5363 is selectively active in PI3KCA mutant gastric cancer, and sensitizes a patient-derived gastric cancer xenograft model with PTEN loss to taxotere
.
J Transl Med
2013
;
11
:
241
.
27.
Taglialatela
A
,
Alvarez
S
,
Leuzzi
G
,
Sannino
V
,
Ranjha
L
,
Huang
J-W
, et al
Restoration of replication fork stability in BRCA1- and BRCA2-deficient cells by inactivation of SNF2-family fork remodelers
.
Mol Cell
2017
;
68
:
414
30
.
28.
Karakashev
S
,
Fukumoto
T
,
Zhao
B
,
Lin
J
,
Wu
S
,
Fatkhutdinov
N
, et al
EZH2 inhibition sensitizes CARM1-high, homologous recombination proficient ovarian cancers to PARP inhibition
.
Cancer Cell
2020
;
37
:
157
67
.
29.
Rondinelli
B
,
Gogola
E
,
Yücel
H
,
Duarte
AA
,
van de Ven
M
,
van der Sluijs
R
, et al
EZH2 promotes degradation of stalled replication forks by recruiting MUS81 through histone H3 trimethylation
.
Nat Cell Biol
2017
;
19
:
1371
8
.
30.
Arasanz
H
,
Hernández
C
,
Bocanegra
A
,
Chocarro
L
,
Zuazo
M
,
Gato
M
, et al
Profound reprogramming towards stemness in pancreatic cancer cells as adaptation to AKT inhibition
.
Cancers
2020
;
12
:
2181
.
31.
Morita
M
,
Gravel
S-P
,
Hulea
L
,
Larsson
O
,
Pollak
M
,
St-Pierre
J
, et al
mTOR coordinates protein synthesis, mitochondrial activity and proliferation
.
Cell Cycle
2015
;
14
:
473
80
.
32.
Woo
SU
,
Sangai
T
,
Akcakanat
A
,
Chen
H
,
Wei
C
,
Meric-Bernstam
F
. 
Vertical inhibition of the PI3K/Akt/mTOR pathway is synergistic in breast cancer
.
Oncogenesis
2017
;
6
:
e385
.
33.
O'Reilly
KE
,
Rojo
F
,
She
Q-B
,
Solit
D
,
Mills
GB
,
Smith
D
, et al
mTOR inhibition induces upstream receptor tyrosine kinase signaling and activates Akt
.
Cancer Res
2006
;
66
:
1500
8
.
34.
Hsieh
H-J
,
Zhang
W
,
Lin
S-H
,
Yang
W-H
,
Wang
J-Z
,
Shen
J
, et al
Systems biology approach reveals a link between mTORC1 and G2–M DNA damage checkpoint recovery
.
Nat Commun
2018
;
9
:
3982
.
35.
Mo
W
,
Liu
Q
,
Lin
CC-J
,
Dai
H
,
Peng
Y
,
Liang
Y
, et al
mTOR inhibitors suppress homologous recombination repair and synergize with PARP inhibitors via regulating SUV39H1 in BRCA-proficient triple-negative breast cancer
.
Clin Cancer Res
2016
;
22
:
1699
712
.