Researchers from the Novartis Institutes for BioMedical Research established 1,075 patient-derived tumour xenograft models (PDXs) with a diverse set of driver mutations. With these PDXs, they performed in vivo compound screens to assess the population responses to 62 treatments across six indications. They demonstrated both the reproducibility and the clinical translatability of this approach by identifying associations between a genotype and drug response, and established mechanisms of resistance.
They reported in Nature Medicine that their results suggest that PDX clinical trials may represent a more accurate approach than cell line models for assessing the clinical potential of some therapeutic modalities. This approach represents a new experimental paradigm through which to address the tumour biology of cancer patients, and interrogate targeted therapies in in vivo models that are more relevant to the clinic than traditional oncology models, which could potentially improve the ability of preclinical oncology studies to predict patient response in the clinic.
A need for experimental systems that better replicate the diversity of human tumour biology in a preclinical setting
The Novartis Institutes for Biomedical Research patient-derived tumour xenograft encyclopedia: (a) Distribution of cancer types in the patient-derived tumour xenograft encyclopedia by lineage (n= 1,075). (b) Similarity of patient-derived tumour xenografts between passages and lineages using Affymetrix mRNA expression data. (c) Somatic mutation frequencies in patient-derived tumour xenografts. (d) Genomic landscape analysis of melanoma across patient-derived tumour xenograft encyclopedia, The Cancer Genome Atlas and Cancer Cell Line Encyclopedia data sets.
Roughly 85% of preclinical agents entering oncology clinical trials fail to demonstrate sufficient safety or efficacy to gain regulatory approval. This high failure rate highlights a weak understanding of the complexity of human cancer, the continued limitations of the predictive value of existing preclinical models and the scale at which cancer models are interrogated in the preclinical setting.
Increasing amounts of evidence have suggested that PDXs faithfully recapitulate human tumour biology and predict patient drug response by directly comparing drug responses in patients and their corresponding xenografts. However, these studies have limited value in predicting potential clinical-trial response at the population level, owing to the use of a limited number of PDXs.
Therefore, the researchers from the Novartis Institutes for BioMedical Research have generated an extensive collection of PDXs containing more than 1,000 models, all characterised for their mutations, copy-number alterations and mRNA expression levels.
The study researchers received tumour specimens from the US National Disease Research Interchange, the US National Cancer Institute, the Maine Medical Center, the Tufts Medical Center, the Mt Group Inc. and GenenDesign. They used these PDXs to perform a large-scale in vivo screen to model inter-patient response heterogeneity with a 'one animal per model per treatment' approach (1 × 1 × 1).
By correlating genomic information with observed efficacy, they successfully validated genetic hypotheses and biomarkers derived fromin vitromodel systems, and identified novel therapies that cell line model systems have failed to capture. Furthermore, they obtained notably similar results when comparing the available clinical data with the response of the PDXs.
Finally, deep-sequencing analysis of melanoma-resistant tumours from the PDX clinical trial revealed mechanisms of resistance similar to those reported in patients.
Together, these data demonstrate, retrospectively, the enhanced translatability of this in vivo experimental paradigm, and set the foundation for the use of this population-based approach for the potential prediction of human clinical trial responses.
Effectiveness of experimental paradigm for examining population-basedin vivocompound screens
The authors discussed in their article that their approach enables insight into interpatient response heterogeneity in an efficient manner, and helps to identify responsive subpopulations, thus enabling the discovery of predictive biomarkers. In addition, it can be used to identify clinically relevant mechanisms of resistance. They propose this experimental paradigm for preclinical drug evaluation to enhance the predictability for phase I/II clinical trials.
As with all preclinical models, there are limitations with PDXs, including the lack of an intact immune system, differential influences of mouse stroma versus human stroma and, in addition, the under-representation of very specific genotypes and specific lineage subtypes.
However, conservation of major genetic alterations found in patient tumours in the corresponding PDXs underlies the extensive utility of these models in the preclinical studies. As PDXs closely resemble the genomic landscape of human cancers at the population level, there is a strong rationale for performing preclinical drug screens to investigate the population-based interpatient response heterogeneity.
The researchers demonstrated that PDXs may have an advantage over long-established cell lines. Their genomic analysis revealed that various signaling pathways are under- or overrepresented in cell lines across lineages. For example, they found underrepresentation of alterations in the PI3K pathway in non-small cell lung cancer (NSCLC) and overrepresentation of the TGF-β pathway in pancreatic ductal adenocarcinoma and receptor tyrosine kinase alterations in breast carcinoma. In contrast, these pathways are accurately represented in the PDX collection at a similar mutation frequency as that reported in patient tumours.
The liabilities with cell-based models are borne out in discrepancies with respect to drug response. Anin vitrocombination screen in melanoma failed to identify the combinatorial effect of the CDK4/6 inhibitor with other targeted therapies, whereas the PDX clinical trial subsequently did reveal this effect, as exemplified by the combination of LEE011 and encorafenib.
The study additionally show differential combinatorial effects with IGF1R inhibitors in vitro and in vivo. Indeed, there are several literature reports showing that combinatorial inhibition of IGF1R and MEK1/2 or of PI3K/mTOR are efficacious not only in cell proliferation assays in vitro but also, importantly, in cell line–derived xenografts in vivo. These positive results have led to a number of clinical trials in colorectal cancer and NSCLC, as well as in several other indications, yet the data from the clinical trials are fairly disappointing, concordant with results in the study PDX models.
Analysis of the melanoma-resistant tumours under continuous drug treatment by deep-sequencing analyses has revealed similar mechanisms of resistance to both single-agent (encorafenib) and combination treatments (encorafenib-binimetinib and encorafenib-BKM120) reported in the clinic, supporting the idea that the PDX clinical trial approach can recapitulate the mechanisms of drug resistance found in patients.
Additionally, the researchers have identified a novel mutation (MAP2K2Q218P) that could potentially confer resistance to the binimetinib-LEE011 combination currently being evaluated in the clinic to treat NRAS-mutated melanomas.
Previously reported BRAF amplification, as a mechanism conferring resistance to BRAF inhibitors, was the only mechanism identified. In contrast, the population-based PDX clinical trial studies uncovered diverse mechanisms of resistance to BRAF inhibition, demonstrating a very efficient approach to studying drug resistance in an in vivo setting, and consequently, providing opportunities to develop strategies that are likely to mitigate resistance mechanisms at an early stage of drug development.
This PDX clinical trial concept is the first to evaluate the reproducibility and translatability of the 1 × 1 × 1 concept to drug response using an extensive, well-characterised PDX collection. This approach represents a new experimental paradigm which could potentially improve the ability of preclinical oncology studies to predict patient response in the clinic.
This research was funded by Novartis, Inc. and all authors were employees thereof at the time the study was performed. The authors declare no other competing financial interests.