PRECISE: a domain adaptation approach to transfer predictors of drug response from pre-clinical models to tumors.
ABSTRACT: MOTIVATION:Cell lines and patient-derived xenografts (PDXs) have been used extensively to understand the molecular underpinnings of cancer. While core biological processes are typically conserved, these models also show important differences compared to human tumors, hampering the translation of findings from pre-clinical models to the human setting. In particular, employing drug response predictors generated on data derived from pre-clinical models to predict patient response remains a challenging task. As very large drug response datasets have been collected for pre-clinical models, and patient drug response data are often lacking, there is an urgent need for methods that efficiently transfer drug response predictors from pre-clinical models to the human setting. RESULTS:We show that cell lines and PDXs share common characteristics and processes with human tumors. We quantify this similarity and show that a regression model cannot simply be trained on cell lines or PDXs and then applied on tumors. We developed PRECISE, a novel methodology based on domain adaptation that captures the common information shared amongst pre-clinical models and human tumors in a consensus representation. Employing this representation, we train predictors of drug response on pre-clinical data and apply these predictors to stratify human tumors. We show that the resulting domain-invariant predictors show a small reduction in predictive performance in the pre-clinical domain but, importantly, reliably recover known associations between independent biomarkers and their companion drugs on human tumors. AVAILABILITY AND IMPLEMENTATION:PRECISE and the scripts for running our experiments are available on our GitHub page (https://github.com/NKI-CCB/PRECISE). SUPPLEMENTARY INFORMATION:Supplementary data are available at Bioinformatics online.
Project description:This protocol provides the steps required for the establishment of patient-derived xenograft (PDX) tumors for head and neck squamous cell carcinomas (HNSCCs) and their utility in examining drug responses. PDXs recapitulate the heterogeneity observed in the corresponding human tumors, which makes them an ideal pre-clinical model system. This protocol outlines the detailed steps required for (1) the generation of HNSCC-PDXs, (2) the processing of tumor tissues, and (3) the expansion of PDX models into cohorts for (4) drug testing. For complete details on the use and execution of this protocol please refer to Karamboulas et al. (2018).
Project description:<h4>Background</h4>While patient-derived xenografts (PDXs) offer a powerful modality for translational cancer research, a precise evaluation of how accurately patient responses correlate with matching PDXs in a large, heterogeneous population is needed for assessing the utility of this platform for preclinical drug-testing and personalized patient cancer treatment.<h4>Patients and methods</h4>Tumors obtained from surgical or biopsy procedures from 237 cancer patients with a variety of solid tumors were implanted into immunodeficient mice and whole-exome sequencing was carried out. For 92 patients, responses to anticancer therapies were compared with that of their corresponding PDX models.<h4>Results</h4>We compared whole-exome sequencing of 237 PDX models with equivalent information in The Cancer Genome Atlas database, demonstrating that tumorgrafts faithfully conserve genetic patterns of the primary tumors. We next screened PDXs established for 92 patients with various solid cancers against the same 129 treatments that were administered clinically and correlated patient outcomes with the responses in corresponding models. Our analysis demonstrates that PDXs accurately replicate patients' clinical outcomes, even as patients undergo several additional cycles of therapy over time, indicating the capacity of these models to correctly guide an oncologist to treatments that are most likely to be of clinical benefit.<h4>Conclusions</h4>Integration of PDX models as a preclinical platform for assessment of drug efficacy may allow a higher success-rate in critical end points of clinical benefit.
Project description:The prognosis for children with high-risk neuroblastoma is often poor and survivors can suffer from severe side effects. Predictive preclinical models and novel therapeutic strategies for high-risk disease are therefore a clinical imperative. However, conventional cancer cell line-derived xenografts can deviate substantially from patient tumors in terms of their molecular and phenotypic features. Patient-derived xenografts (PDXs) recapitulate many biologically and clinically relevant features of human cancers. Importantly, PDXs can closely parallel clinical features and outcome and serve as excellent models for biomarker and preclinical drug development. Here, we review progress in and applications of neuroblastoma PDX models. Neuroblastoma orthotopic PDXs share the molecular characteristics, neuroblastoma markers, invasive properties and tumor stroma of aggressive patient tumors and retain spontaneous metastatic capacity to distant organs including bone marrow. The recent identification of genomic changes in relapsed neuroblastomas opens up opportunities to target treatment-resistant tumors in well-characterized neuroblastoma PDXs. We highlight and discuss the features and various sources of neuroblastoma PDXs, methodological considerations when establishing neuroblastoma PDXs, in vitro 3D models, current limitations of PDX models and their application to preclinical drug testing.
Project description:Introduction and Objectives: Optimizing targeted therapy in patients with metastatic renal cell carcinoma (RCC) would improve clinical outcomes but patient derived xenograft (PDX) models are lacking. We present a novel pre-clinical model that is superior to nude mice for accommodating RCC PDXs. This preclinical model implants RC PDXs into the chorioallantoic membrane (CAM) of avian embryos and is a patient-specific platform that could be advantageous for physicians in the future when deciding what treatment options are best for their patients. This drug panelling platform is rapid, cost-effective, and relies on the highly angiogenic CAM to support RCC PDXs. Methods: Commercial and patient derived RCC cell lines, were grown to full confluence and transduced to generate fluorescently labeled versions of each cell line. Cells were implanted into the CAM. Tumors were treated every two days by applying 10 ?L (10 ?M) of indicated drug onto the tumor onplant. The drugs that were paneled include Sunitinib, Sorafenib, Pazopanib, Axitinib and a vehicle treatment. After 7–8 days of incubation post-implantation, tumor take rate was determined by the presence of tumor growth in the CAM using a fluorescent stereoscope. Results: The highest tumor take rates were observed in the vehicle treatments of the embryos, ranging from 50–86%. Both commercial and primary cell lines saw a reduction in tumor take rate with the application of various anti-angiogenic drugs. Specifically, XP121 tumors were resistant to Sorafenib; 786-0, XP121; XP206 were resistant to Pazopanib; T258, XP121 were resistant to Sunitinib; and lastly, T258 tumors were resistant to Axitinib (Table 1). Conclusions: RCC PDXs onplanted in the CAM of avian embryos offer a robust and cost-effective platform to predict sensitivity/resistance to targeted therapies. When evaluating several patient-derived RCC cell lines, drug paneling revealed other alternative treatment options for these PDXs. More importantly, RCC PDXs that were shown to be Sunitinib-resistant in both the patient and in mouse-based PDXs, also produced the same resistance phenotype in the CAM.
Project description:Patient-derived xenografts (PDXs) are increasingly used in cancer research as a tool to inform cancer biology and drug response. Most available breast cancer PDXs have been generated in the metastatic setting. However, in the setting of operable breast cancer, PDX models both sensitive and resistant to chemotherapy are needed for drug development and prospective data are lacking regarding the clinical and molecular characteristics associated with PDX take rate in this setting.The Breast Cancer Genome Guided Therapy Study (BEAUTY) is a prospective neoadjuvant chemotherapy (NAC) trial of stage I-III breast cancer patients treated with neoadjuvant weekly taxane+/-trastuzumab followed by anthracycline-based chemotherapy. Using percutaneous tumor biopsies (PTB), we established and characterized PDXs from both primary (untreated) and residual (treated) tumors. Tumor take rate was defined as percent of patients with the development of at least one stably transplantable (passed at least for four generations) xenograft that was pathologically confirmed as breast cancer.Baseline PTB samples from 113 women were implanted with an overall take rate of 27.4% (31/113). By clinical subtype, the take rate was 51.3% (20/39) in triple negative (TN) breast cancer, 26.5% (9/34) in HER2+, 5.0% (2/40) in luminal B and 0% (0/3) in luminal A. The take rate for those with pCR did not differ from those with residual disease in TN (p?=?0.999) and HER2+ (p?=?0.2401) tumors. The xenografts from 28 of these 31 patients were such that at least one of the xenografts generated had the same molecular subtype as the patient. Among the 35 patients with residual tumor after NAC adequate for implantation, the take rate was 17.1%. PDX response to paclitaxel mirrored the patients' clinical response in all eight PDX tested.The generation of PDX models both sensitive and resistant to standard NAC is feasible and these models exhibit similar biological and drug response characteristics as the patients' primary tumors. Taken together, these models may be useful for biomarker discovery and future drug development.
Project description:Tumor-sequencing studies have revealed the widespread genetic diversity of melanoma. Sequencing of 108 genes previously implicated in melanomagenesis was performed on 462 patient-derived xenografts (PDXs), cell lines, and tumors to identify mutational and copy number aberrations. Samples came from 371 unique individuals: 263 were naive to treatment, and 108 were previously treated with targeted therapy (34), immunotherapy (54), or both (20). Models of all previously reported major melanoma subtypes (BRAF, NRAS, NF1, KIT, and WT/WT/WT) were identified. Multiple minor melanoma subtypes were also recapitulated, including melanomas with multiple activating mutations in the MAPK-signaling pathway and chromatin-remodeling gene mutations. These well-characterized melanoma PDXs and cell lines can be used not only as reagents for a large array of biological studies but also as pre-clinical models to facilitate drug development.
Project description:Colorectal cancer (CRC) is the second most common cancer in Europe and a leading cause of death worldwide. Patient-derived xenograft (PDX) models maintain complex intratumoral biology and heterogeneity and therefore remain the platform of choice for translational drug discovery. In this study, we implanted 37 primary CRC tumors and five CRC cell lines into NU/J mice to develop xenograft models. Primary tumors and established xenografts were histologically assessed and surveyed for genetic variants and gene expression using a panel of 409 cancer-related genes and RNA-seq, respectively. More than half of CRC tumors (20 out of 37, 54%) developed into a PDX. Histological assessment confirmed that PDX grading, stromal components, inflammation, and budding were consistent with those of the primary tumors. DNA sequencing identified an average of 0.14 variants per gene per sample. The percentage of mutated variants in PDXs increased with successive passages, indicating a decrease in clonal heterogeneity. Gene Ontology analyses of 4180 differentially expressed transcripts (adj. p value < 0.05) revealed overrepresentation of genes involved in cell division and catabolic processes among the transcripts upregulated in PDXs; downregulated transcripts were associated with GO terms related to extracellular matrix organization, immune responses, and angiogenesis. Neither a transcriptome-based consensus molecular subtype (CMS) classifier nor three other predictors reliably matched PDX molecular subtypes with those of the primary tumors. In sum, both genetic and transcriptomic profiles differed between donor tumors and PDXs, likely as a consequence of subclonal evolution at the early phase of xenograft development, making molecular stratification of PDXs challenging.
Project description:INTRODUCTION:Improvement in the ability to target underlying drivers and vulnerabilities of high-grade serous ovarian cancer (HG-SOC) requires the development of molecularly annotated pre-clinical models reflective of clinical responses. METHODS:We generated patient-derived xenografts (PDXs) from consecutive, chemotherapy-naïve, human HG-SOC by transplanting fresh human HG-SOC fragments into subcutaneous and intra-ovarian bursal sites of NOD/SCID IL2R?(null) recipient mice, completed molecular annotation and assessed platinum sensitivity. RESULTS:The success rate of xenografting was 83%. Of ten HG-SOC PDXs, all contained mutations in TP53, two were mutated for BRCA1, three for BRCA2, and in two, BRCA1 was methylated. In vivo cisplatin response, determined as platinum sensitive (progression-free interval ? 100 d, n = 4), resistant (progression-free interval <100 d, n = 3) or refractory (n = 3), was largely consistent with patient outcome. Three of four platinum sensitive HG-SOC PDXs contained DNA repair gene mutations, and the fourth was methylated for BRCA1. In contrast, all three platinum refractory PDXs overexpressed dominant oncogenes (CCNE1, LIN28B and/or BCL2). CONCLUSIONS:Because PDX platinum response reflected clinical outcome, these annotated PDXs will provide a unique model system for preclinical testing of novel therapies for HG-SOC.
Project description:Although clinical management of melanoma has changed considerably in recent years, intrinsic treatment resistance remains a severe problem and strategies to design personal treatment regimens are highly warranted. We have applied a three-dimensional (3D) ex vivo drug efficacy assay, exposing disaggregated cells from 38 freshly harvested melanoma lymph node metastases and 21 patient derived xenografts (PDXs) to clinical relevant drugs for 7 days, and examined its potential to evaluate therapy response. A strong association between Vemurafenib response and BRAF mutation status was achieved (P?<?.0001), while enhanced viability was seen in some NRAS mutated tumors. BRAF and NRAS mutated tumors responded comparably to the MEK inhibitor Cobimetinib. Based on the ex vivo results, two tumors diagnosed as BRAF wild-type by routine pathology examinations had to be re-evaluated; one was subsequently found to have a complex V600E mutation, the other a double BRAF mutation (V600E/K601 N). No BRAF inhibitor resistance mechanisms were identified, but PIK3CA and NF1 mutations were identified in two highly responsive tumors. Concordance between ex vivo drug responses using tissue from PDXs and corresponding patient tumors demonstrate that PDX models represent an indefinite source of tumor material that may allow ex vivo evaluation of numerous drugs and combinations, as well as studies of underlying molecular mechanisms. In conclusion, we have established a rapid and low cost ex vivo drug efficacy assay applicable on tumor tissue from patient biopsies. The 3D/spheroid format, limiting the influence from normal adjacent cells and allowing assessment of drug sensitivity to numerous drugs in one week, confirms its potential as a supplement to guide clinical decision, in particular in identifying non-responding patients.