Year: 2016
Type: editorial
Abstract: While the disconnect between the reality of the human breast cancer experience and the preclinical models used to study the disease has been of concern for many years, two technical breakthroughs are the principle drivers of the contemporary debate on this critical issue. Next-generation sequencing (NGS) of DNA and RNA has revealed the stunning complexity of breast cancer genomes, with tumors often harboring hundreds, even thousands, of somatic changes and marked clonal heterogeneity (1). Clearly cell line collections do not reflect the clinical and genomic complexity of breast cancer. The second breakthrough is the patient-derived xenograft (PDX) field where enhanced engraftment rates for breast cancer in new strains of immunocompromised mice and new engraftment techniques are creating a rapidly enlarging repertoire of new breast cancer models with detailed clinical annotation (2). The potential of merging these fields was first illustrated in 2010 when a trio of samples from a patient with basal-like breast cancer was examined by whole-genome sequencing (3). The comparison between the breast primary, a brain metastasis, and a PDX from the primary revealed that the PDX has a mutational repertoire more similar to the brain metastasis, even though it was grown from the primary. This suggested that the PDX captured the most metastatic subclones, as if the mouse was functioning as a metastatic site. Subsequently PDX methodologies have been used to reveal endocrine therapy resistance mechanisms (4) and to model new treatment approaches (5). In general PDX models are reasonably genomically and phenotypically stable across successive generations; however, further clonal evolution upon passaging is inevitable (6) and repeated genomic analysis is likely necessary to ensure the models maintain genomic features under study (4). The study by Goetz et al., published in this issue of the Journal, nicely illustrates the synergies that can be achieved by combining NGS and PDX research but also the complexities and uncertainties of this approach (7). The Breast Cancer Genome Guided Therapy Study (BEAUTY) study is a prospective study of women with invasive early-stage breast cancer who qualify for neoadjuvant chemotherapy (NAC). Women who enrolled in this study underwent biopsy of the primary tumor for exome and RNA sequencing and for PDX development. After completion of NAC, if the surgical tumor specimen contained viable tumor, it was used to generate post-treatment PDXs. This effort generated a pretreatment PDX from 31 patients and six PDX from residual disease. The expected bias in PDX take rate was observed (4,8), highest with basal-like and lowest with luminal estrogen receptor–positive (ER + ) tumors. In the future, intraductal injection techniques may resolve the engraftment imbalance between subtypes (9). Surprisingly Goetz et al. do not clearly report the relationship between PDX success rate and pCR—a critical question if we are to use PDX models to dissect responses to multi-agent chemotherapy. In terms of the NGS data, disease subtypes, and clinical outcomes, TP53 mutation was enriched in both TNBC and HER2+ tumors in accordance with published data. No patient in the luminal A group had pCR, and only 10% of the luminal B achieved pCR, reflecting the current understanding of the breast cancer–intrinsic subtypes and response to neoadjuvant chemotherapy (10). From the perspective of interactions between somatic mutation data and clinical outcomes, their data illustrates the complexity of correlating specific mutations to a drug response phenotype such as pathological response because this outcome is the result of a very complex set of not only genomic variables but also cell lineage effects (basal vs luminal) and host effects including the immune response (11). Nonetheless the chemotherapy-resistant nature of luminal androgen receptor–positive (LAR+) breast cancer is a useful point of emphasis, and it would be of value to investigate whether the counterpart LAR+ PDXs were responsive to drugs targeting AR. The potential study of drug effectiveness in PDX from chemotherapy-resistant tumors is clearly illustrated with their “N of one” study of olaparib in a PDX from a tumor with frame-shift mutations in both BRCA1 and PALB2 that must have generated a marked defect in DNA repair based on homologous recombination. A more extensive analysis of PARP inhibitor sensitivity in a larger number of PDXs is needed to determine if this class of agents could have wider applicability in chemotherapy-resistant disease. Somatic mutations in BRCA1/2, as well as genes serving other homologous recombination (HR) pathway components, are reasonably common but are not routinely identified. Restricting PARP inhibitors to only germline HR defect diagnoses likely represents a missed therapeutic opportunity. Next steps will need to focus on improving the rigor of PDX studies. Credentialing every PDX against the genome of the progenitor tumor is an important step taken by other investigators (4) but is a pending analysis in the BEAUTY study. Matching drug treatment sensitivities between human and each counterpart PDX (the defining feature of a co-clinical trial) is another (8), and proteogenomic analysis to bring in data on signal transduction would be a third (12). Proteomic approaches could address many unanswered questions concerning PDX, including whether perturbed signaling networks driven by somatic mutation are preserved after transplantation. Because the stromal and immune environments are quite different in the PDX vs the human breast, it would be unwise to assume that PDX modeling will reproduce all therapeutically targetable biological events in breast cancer. Nonetheless it does seem likely that in the near future the acquisition of PDX for mechanistic studies will be considered an important aspect of drug development as we continue to probe new molecular classifications of breast cancer with targeted agents. Finally, researchers involved in the PDX field should consider entering data on each PDX model into a database so that common targetable pathways present in the repertoire can be identified, paving the way for researchers to collaborate more effectively and move beyond the N of one approach. An example is The Breast Cancer Research Foundation PDX cBioPortal, which houses clinico-pathologic and genomics data on PDX models and tools for data visualization and reanalysis, accessible through http://www.cbioportal.org/pdx. The authors have no conflicts of interest to disclose.
DOI: 10.1093/jnci/djw307
Institution: Baylor College of Medicine
Language: English
Subfield: Cancer Research
Domain: Life Sciences
Open Access status: bronze
Funder: National Cancer Institute
Award: G4023538194
OpenAlex ID: W2688529415