Multiomics analysis of serial PARP inhibitor treated metastatic TNBC inform on rational combination therapies

The study addresses whether longitudinal, real-time multi-omics profiling of serial biopsies from metastatic triple-negative breast cancer (mTNBC) can reveal adaptive resistance mechanisms and inform rational combination therapies with PARP inhibitors. TNBC is molecularly heterogeneous and often unresponsive to uniform therapies. While PARP inhibitors are approved for germline BRCA-mutant disease and immune checkpoint blockade (ICB) has shown benefit in PD-L1–positive mTNBC, resistance frequently develops. Preclinical and early clinical data suggest PARPi can activate STING-mediated immunity and may synergize with ICB, but predictive biomarkers are lacking and serial biopsies have been scarce. The study proposes a precision oncology approach leveraging adaptive changes during PARPi therapy to select patient-specific combinations.

Prior trials (OlympiAD, ABRAZO) established efficacy of PARP inhibitors (olaparib, talazoparib) in germline BRCA-mutant breast cancer. Frontline ICB plus chemotherapy improved outcomes in PD-L1–positive mTNBC (IMpassion130). Preclinical models indicate PARPi induces STING signaling, interferon responses, PD-L1 upregulation, and immune activation, potentially enhancing ICB in BRCA wild-type settings. Clinical combination studies (MEDIOLA, TOPACIO) showed tolerability and activity of PARPi+ICB in subsets but were limited by few serial biopsies, preventing robust biomarker discovery. Additional literature supports adaptive responses to targeted agents underpinning resistance and therapeutic opportunities, with rational combinations (e.g., PARPi with ATR/CHK1/WEE1, PI3K/AKT, MEK/MAPK inhibitors) showing promise across tumor models.

Design: Single-center, open-label, single-arm pilot trial (NCT03544125) at OHSU Knight Cancer Institute evaluating olaparib followed by olaparib plus durvalumab in mTNBC with integrated serial multi-omics profiling. Eligibility: Biopsy-proven mTNBC, ECOG ≤2, ≤2 prior metastatic chemotherapies; no prior PARPi or ICB in metastatic setting. Interventions: Cycle 1 olaparib 300 mg PO BID for 28 days; on-treatment biopsy at end of cycle 1; then up to 12 cycles of olaparib 300 mg PO BID plus durvalumab 1500 mg IV q4 weeks. Biopsies: Image-guided (US/CT) core biopsies (18G), 3–6 passes; processed for FFPE (formalin fixation, paraffin embedding), flash-frozen tissue, and glutaraldehyde fixation. Assays (CLIA and research): - CLIA IHC for ER, PR, AR, HER2, Ki67; PD-L1 IHC 22C3 pharmDx (tumor cell percentage). - Serum tumor markers (CA15-3/CA15-5 noted, CA27-29, CEA). - GeneTrails Comprehensive Solid Tumor Panel (124 genes) for SNVs/indels, CNAs, MSI; RNA fusion panel; performed in CLIA lab. - Whole-exome sequencing (tumor and matched normal): library prep with Agilent SureSelect V5, Illumina NovaSeq (tumor ~500X, normal ~100X), alignment (BWA MEM), calling (GATK4 MuTect2) with specified depth and VAF filters; CNAs via CNVkit. - Multiplex immunohistochemistry (mIHC) on FFPE for immune phenotyping; hierarchical gating; immune cell densities (cells/mm²). - Reverse-phase protein array (RPPA) for proteomics; scaled to TCGA breast RPPA; pathway scores computed from predefined protein/phosphoprotein predictors. - Cyclic immunofluorescence (Cyc-IF) single-cell proteomics (>40 proteins per FFPE slide); iterative staining/quenching; K-means clustering (elbow method) to define tumor cell states. Outcomes: Primary—feasibility of completing CLIA assays within 4 weeks prior to on-treatment biopsy. Secondary—safety/tolerability; preliminary efficacy by RECIST v1.1 and irRECIST, time to progression, OS. Exploratory—identify predictive biomarkers, adaptive resistance mechanisms, PARPi-induced changes, and candidate combination therapies. Analysis: Cross-platform integration of genomics, proteomics, and immune profiling; comparison of pre- vs on-olaparib samples; pathway scoring; cell population frequency changes.

• Enrollment/feasibility: Three mTNBC patients enrolled (median age 51.3; range 39–66). All completed pre- and on-treatment biopsies. Predetermined CLIA analytics were completed within the 4-week window; median completion time 13 days, meeting the primary feasibility endpoint.
• Safety: No biopsy-related adverse events. One grade 3–4 hip fracture (33%; 1/3), unrelated to treatment. Common grade 1–2 AEs: fatigue 66% (2/3), nausea 66% (2/3), constipation 66% (2/3), musculoskeletal pain 66% (2/3); no immune-related AEs; no AE-led discontinuations.
• Clinical responses: Case 1 (AR-positive LAR phenotype): progressive disease (PD), deceased ~2 months after start (survival 1.77 months). Case 2: stable disease (SD) for 7 months; survival 12.53 months; overall response PD. Case 3: partial response (near complete) lasting 15.1 months; progression later at 19 months; alive at 21-month data cutoff.
• Tumor biomarkers (selected): Case 1: AR 100% pre/post; ER negative pre, 5% post; PR negative; HER2 negative; Ki67 50%→30%; PD-L1 low positive (1–10%→10–20%); CNA: CCND1 (84→51 copies), FGF3 (44→45); SNVs: TP53 (63%→86% VAF), FGFR4 VUS (~64–65% VAF); MSI stable. Case 2: ER/PR/AR negative; Ki67 50%→80%; PD-L1 negative (post); mutations in TP53, RB1, ALK VUS, MDC1 VUS; MSI stable. Case 3: ER/PR/AR negative; Ki67 60–80% pre, >50% post; PD-L1 negative pre; WES pre: TP53 mutation (70% VAF), MYC amplification (8.5 copies), RB1 copy loss (0.96); germline BRCA1 exons 13–15 deletion (VUS for function); MSI stable.
• Immune contexture (mIHC): Case 1: modest increase in CD8 T-cell density on-treatment but predominantly memory/naïve; effector CD8+ cells ~16 cells/mm²; overall low immune infiltration, consistent with PD. Case 2 (post only): diverse immune populations with predominance of CD4+ T cells and CD163− myelomonocytic cells; CD8+ T-cell density low (~12 cells/mm²) with few effectors. Case 3: marked increase in immune infiltration; CD8+ T-cell population increased to ~150 cells/mm² with rises in early and late effectors; tumor content dropped from ~90% to ~5% on-treatment—consistent with robust response and PARPi-induced immune activation.
• Proteomics (RPPA): Confirmed target engagement with drastic decrease in protein PARylation on-treatment. Case 1: activation of DNA damage response (↑p-H2AX, ↑p-RPA32), but minimal global protein network changes; baseline high pathway activity (RTK, RAS-MAPK, PI3K-AKT, TSC-mTOR) and strong DNA damage checkpoint and cell cycle progression scores (replication stress), with low immune scores; suggests indifference to PARP inhibition and potential vulnerability to DNA damage checkpoint inhibitors rather than ICB combination. Case 3: extensive protein network rewiring post-olaparib with increased immune infiltration signatures, DNA damage checkpoint activation, apoptosis, and increased RTK/RAS-MAPK signaling; decreased tumor content and cell cycle progression consistent with sensitivity.
• Single-cell proteomics (Cyc-IF): K-means clustering defined tumor cell states. Case 1: high cyclin D1 and GATA3 (consistent with CCND1 amplification and AR biology); small shifts toward proliferative cells with replication stress post-treatment; resting population unchanged—indicative of olaparib indifference. Case 3: major shift from ~75% proliferative cells pre-treatment to >90% DDR/stress phenotype on-treatment, indicating strong PARPi response.
• Overall: Cross-platform concordance demonstrated that early, PARPi-induced ecosystem changes can be detected after one cycle and may predict benefit; findings nominate rational, patient-specific combinations (e.g., PARPi with ATR/CHK1/WEE1, or with RTK/PI3K-AKT/MAPK inhibitors; PARPi+ICB in tumors with PARPi-induced immune activation).

This pilot demonstrates that serial, real-time multi-omics of mTNBC under PARPi stress is feasible and yields actionable biological insights beyond baseline profiling. Early adaptive changes in tumor and immune states within one cycle of olaparib correlated with clinical outcomes: an exceptional responder (basal, BRCA1-altered) exhibited marked immune infiltration and proteomic rewiring, supporting synergy of PARPi with ICB, whereas an AR-positive LAR subtype patient showed minimal ecosystem change and rapid progression, suggesting poor suitability for PARPi+ICB but potential vulnerability to combinations targeting replication stress and DNA damage checkpoints. Integrated analysis across CLIA genomics, WES, RPPA, mIHC, and Cyc-IF provided a coherent view of pathway activation, cell states, and immune contexture, enabling hypothesis generation for individualized combinations. The results support a precision approach that uses induced adaptive responses to guide next-step therapies rather than relying solely on static pre-treatment biomarkers.

Serial multi-omics profiling of mTNBC during initial PARP inhibitor exposure is feasible within clinical timelines and reveals dynamic tumor and immune adaptations that can inform rational combination therapies. Early changes in protein networks, cell states, and immune infiltration may predict benefit and identify patient-specific partners for PARPi, including ICB, AR-targeted strategies in LAR tumors, DNA damage checkpoint inhibitors (ATR/CHK1/WEE1), and RTK/PI3K-AKT/MAPK pathway inhibitors. Larger prospective trials are warranted to validate dynamic biomarkers, refine patient selection for PARPi-based combinations, and to explore less invasive modalities (e.g., liquid biopsies) to monitor tumor restructuring and guide therapy.

This was a small pilot with only three patients, limiting generalizability and statistical inference. Tissue quantity and quality constrained some planned analyses (e.g., insufficient material from Case 2 for RPPA, Cyc-IF, and pre-treatment mIHC), potentially biasing comparisons. Biopsies were obtained from specific metastatic sites, and site-specific microenvironmental differences may influence adaptive responses; the study did not systematically assess inter-site variability. Functional impact of certain genomic alterations (e.g., BRCA1 exons 13–15 deletion) was uncertain. The design did not isolate contributions of olaparib versus durvalumab to clinical benefit. Longer-term outcomes and validation in larger cohorts are needed.