Abstract P3-07-01: Towards a therapeutically relevant subtyping scheme for triple-negative breast cancer (TNBC), profiling results from A Randomized, TNBC Enrolling trial to confirm Molecular profiling Improves Survival (ARTEMIS)

Triple-negative breast cancer is a highly diverse group of cancers, with poor prognosis, and currently, there are no targeted drugs available in the clinic. In TNBC around 50% percent of the patients respond to chemotherapy, while, the other 50% percent relapse with poor prognosis. There is a need to understand better the targetable mechanisms driving TNBC via integrative analysis of gene-expression, copy-number, and mutational data.

Samples from 220 triple-negative breast cancer (TNBC) pts treated with NACT were prioritized for transcriptomic and genomic profiling. Non-negative matrix factorization was used on array-based profiling to identify six robust (ARTEMIS) subtypes. Comparing ARTEMIS subtypes with Vanderbilt subtypes, revealed significant overlap with 4/6 clusters while identifying two new clusters. Logistic regression on ssGSEA scores vs. subtypes revealed several pathways, selectively enriched specific subtypes. CL1/IM (Immune subtype), was enriched in INFg and INFa, while CL2 (MYC/mTOR), showed enrichment of several proliferation-related pathways. In addition, LAR and M (Mesenchymal) pts formed overlapping clusters, using either method.

Two new subtypes did not associate significantly with any of the previous subtypes. The majority of the tumors from the Vanderbilt BL2 and MSL were reclassified into a CL5 (ANGIO) cluster, which was enriched in angiogenesis geneset, including targetable genes like VEGF and FGFR. Also, an MYO (CL3) subtype was identified, with myogenesis-related genes. Of note, TIL (tumor infiltrating lymphocytes) and LAR quantification using IHC were associated with respective ARTEMIS subtypes. Finally, the IM subtype was significantly associated with higher rates of RCB 0-I and the M (CL4) subtype was associated with higher rates of RCB II-III, irrespective of the neoadjuvant treatment regimen.

ARTEMIS subtypes are a novel classification system for TNBC that is focused on therapeutic translation. Further, we show a possibility to classify previously un-classified (UNS) tumors, which will be validated using additional cohorts (TCGA/METABRIC).

Citation Format: Seth S, Huo L, Rauch G, Lau R, Gilcrease M, Adrada B, Piwnica-Worms H, Symmans WF, Draetta G, Futreal AP, Moulder S, Chang JT. Towards a therapeutically relevant subtyping scheme for triple-negative breast cancer (TNBC), profiling results from A Randomized, TNBC Enrolling trial to confirm Molecular profiling Improves Survival (ARTEMIS) [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr P3-07-01.

Abstract P5-05-01: A molecularly annotated collection of breast cancer patient-derived xenograft models aligned with ongoing clinical trials built from fine needle aspiration samples throughout neoadjuvant treatment

BACKGROUND: Patient-derived xenograft (PDX) models of breast cancer replicate the diverse histologic and molecular features of patient tumors and provide a renewable source of human tumor tissue. However, collection of tissue by core needle biopsy is problematic due to patient discomfort, bleeding risk and the limited number of passes a patient can tolerate. Several studies have catalogued the maintenance of molecular features of patient tumors in PDX models of breast cancer.

METHODS: To support the neoadjuvant molecular diagnostic and drug development program in triple negative breast cancer (TNBC), a pilot study was conducted to determine if fine needle aspiration (FNA) could be used for building PDX models. Subsequently, PDX models are being established in alignment with ongoing clinical trials at MDACC. The molecular evolution of patient's tumors, matched with PDXs engrafted from their tumors, is under study throughout the neoadjuvant treatment of TNBC using RNA sequencing, whole-exome sequencing, deep sequencing of cancer genes, and histologic analyses.

RESULTS: To date, 20 established PDX models have been developed and stable PDX models continue to be generated at a rate of 2-3 per month. Several of these models are derived from serial FNAs derived from patients throughout neoadjuvant treatment. These models retain histologic and molecular features of the original patient tumors. Serial patient biopsies, matched with PDX models, have enabled measurement of the mutational and transcriptomic evolution in vivo of TNBC undergoing neoadjuvant treatment.

We have standardized the use of FNAs to generate PDX models both pre- and post-neoadjuvant therapy in the following ongoing neoadjuvant clinical trials:

1. MDACC 2014-0185 (PI Stacy Moulder, 360 patients), 'ARTEMIS: A Randomized TNBC-Enrolling trial to confirm Molecular profiling Improves Survival'

2. MDACC 2014-0045 (PI Jennifer Litton, 20+ patients), 'A pilot study of BMN673 as a neoadjuvant study in patients with a diagnosis of invasive breast cancer and a deleterious BRCA mutation'

CONCLUSION: We demonstrated that PDX models from tissue collected by FNA recapitulate the biology and clinical course of the patient's tumor. Sequencing analyses revealed that neoadjuvant chemotherapy and PDX engraftment enrich for cancer gene mutations. We observe association of the rate of successful PDX engraftment with clinical parameters such as the patient's residual cancer burden (RCB) status at the time of surgery (upon completion of neoadjuvant treatment). In addition, we observe that PDX models derived from serial patient biopsies throughout treatment are more resistant to chemotherapy treatment. These models recapitulate the variety of chemotherapy responses observed in patients with TNBC and serve as powerful tools for preclinical biomarker and discovery studies.

Citation Format: Echeverria GV, Cai S, Tu Y, McCoy A, Lau R, Redwood A, Rauch G, Adrada B, Candelaria R, Santiago L, Thompson A, Litton J, Moulder S, Symmans F, Chang JT, Piwnica-Worms H. A molecularly annotated collection of breast cancer patient-derived xenograft models aligned with ongoing clinical trials built from fine needle aspiration samples throughout neoadjuvant treatment [abstract]. In: Proceedings of the 2017 San Antonio Breast Cancer Symposium; 2017 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2018;78(4 Suppl):Abstract nr P5-05-01.

Abstract P1-07-06: Effect of biospecimen variables on proteomic biomarker assessment in breast cancer

Background: PI3K/Akt/mTOR signaling is being actively pursued as a therapeutic target. We sought to determine how tumor heterogeneity and biospecimen variables affect assessment of PI3K/Akt/mTOR pathway activation.

Methods: Intraoperative image-guided core-needle biopsies (CNB) of primary breast tumors were prospectively collected in 53 patients with invasive breast cancer. After surgery, specimens were collected from the center and periphery of the excised tumor. CNB, central and peripheral surgical specimens were assessed with reverse phase proteomic arrays (RPPA), H&E and immunohistochemistry (IHC).

Results: The expression of standard of care markers ER, PR, and HER2 by RPPA correlated well between biospecimen types. Overall, there was a significant correlation between the expression of 132 (86%) of 154 different markers in the center and periphery; the correlation was significantly higher for smaller tumors, and with shorter cold ischemia time. Expression of many investigational prognostic markers and druggable targets on CNB correlated with expression in the surgical specimen (average of center and periphery), while others, such as EGFR and c-MET, had a weak correlation. Of 154 RPPA markers, 132 (86%) were not statistically different between the center and periphery, and 97 (67%) were not different between the CNB and the surgical specimen. On analysis of the PI3K/AKT/mTOR pathway, pAkt S473 and PTEN had a significant correlation between central and peripheral specimens, and between CNB and surgical specimens. However, pAkt S473, pS6 S235/236 and pS6 240/244 levels were higher in CNB than the central specimens both by RPPA and by IHC. When patients were classified by RPPA PI3K pathway activation score, there was a moderate agreement between classification on the CNB and central specimens (Cohen's Kappa 0.539). However 9 of 20 tumors classified as having PI3K activation on CNB were classified as not having pathway activation on central specimens.

Conclusions: There is remarkable homogeneity in expression of biomarkers within a tumor. However, proteomic markers are differentially expressed by biospecimen type and other preanalytic variables. PI3K pathway activation is greater in CNB compared to surgical samples.

Citation Information: Cancer Res 2012;72(24 Suppl):Abstract nr P1-07-06.