Search

Tumors are characterized by somatic mutations that drive biological processes ultimately reflected in tumor phenotype. With regard to radiographic phenotypes, generally unconnected through present understanding to the presence of specific mutations, artificial intelligence methods can automatically quantify phenotypic characters by using predefined, engineered algorithms or automatic deep-learning methods, a process also known as radiomics. Here we demonstrate how imaging phenotypes can be connected to somatic mutations through an integrated analysis of independent datasets of 763 lung adenocarcinoma patients with somatic mutation testing and engineered CT image analytics. We developed radiomic signatures capable of distinguishing between tumor genotypes in a discovery cohort (n = 353) and verified them in an independent validation cohort (n = 352). All radiomic signatures significantly outperformed conventional radiographic predictors (tumor volume and maximum diameter). We found a radiomic signature related to radiographic heterogeneity that successfully discriminated between EGFR+ and EGFR− cases (AUC = 0.69). Combining this signature with a clinical model of EGFR status (AUC = 0.70) significantly improved prediction accuracy (AUC = 0.75). The highest performing signature was capable of distinguishing between EGFR+ and KRAS+ tumors (AUC = 0.80) and, when combined with a clinical model (AUC = 0.81), substantially improved its performance (AUC = 0.86). A KRAS+/KRAS− radiomic signature also showed significant albeit lower performance (AUC = 0.63) and did not improve the accuracy of a clinical predictor of KRAS status. Our results argue that somatic mutations drive distinct radiographic phenotypes that can be predicted by radiomics. This work has implications for the use of imaging-based biomarkers in the clinic, as applied noninvasively, repeatedly, and at low cost. Cancer Res; 77(14); 3922–30. ©2017 AACR.

Supplemental Figures and Tables to augment (1) the mathematical modeling (providing details on specific therapeutics); (2) patient characteristics (treatment history, clinical and predicted response), (3) synergistic activities identified by EMMA; and (4) data processing guides

Multiple myeloma remains treatable but incurable. Despite a growing armamentarium of effective agents, choice of therapy, especially in relapse, still relies almost exclusively on clinical acumen. We have developed a system, Ex vivo Mathematical Myeloma Advisor (EMMA), consisting of patient-specific mathematical models parameterized by an ex vivo assay that reverse engineers the intensity and heterogeneity of chemosensitivity of primary cells from multiple myeloma patients, allowing us to predict clinical response to up to 31 drugs within 5 days after bone marrow biopsy. From a cohort of 52 multiple myeloma patients, EMMA correctly classified 96% as responders/nonresponders and correctly classified 79% according to International Myeloma Working Group stratification of level of response. We also observed a significant correlation between predicted and actual tumor burden measurements (Pearson r = 0.5658, P < 0.0001). Preliminary estimates indicate that, among the patients enrolled in this study, 60% were treated with at least one ineffective agent from their therapy combination regimen, whereas 30% would have responded better if treated with another available drug or combination. Two in silico clinical trials with experimental agents ricolinostat and venetoclax, in a cohort of 19 multiple myeloma patient samples, yielded consistent results with recent phase I/II trials, suggesting that EMMA is a feasible platform for estimating clinical efficacy of drugs and inclusion criteria screening. This unique platform, specifically designed to predict therapeutic response in multiple myeloma patients within a clinically actionable time frame, has shown high predictive accuracy in patients treated with combinations of different classes of drugs. The accuracy, reproducibility, short turnaround time, and high-throughput potential of this platform demonstrate EMMA's promise as a decision support system for therapeutic management of multiple myeloma. Cancer Res; 77(12); 3336–51. ©2017 AACR.

Supplements with additional descriptions of datasets, feature selection, and results.

Individual patient EMMA predicitons of actual responses to therapy (companion to Figure 3i).

PDF file - 69K, Table S1. Tumor grade and histology of samples used for global metabolomic profiling. Table S2. Metabolites making up the identified metabolomic profiles that demonstrated statistical significance in malignant glioma

PDF file - 224K, Transcriptional pathways associated with metabolic pathways

PDF file - 609K, Kruskal-Wallis test for transciptional differences in metabolic subtypes

PDF file - 237K, Figure S1. RF analysis was performed to determine the accuracy of metabolomic profiles in predicting glioma grade. Figure S2. Intra-sample concordance of glycolytic intermediates PEP and 3-PG in glioma. Figure S3. Glioma subtype classification based on transcriptional profiles was performed using unsupervised hierarchical clustering from a total of 35 probe sets, as described as Phillips et al. (15). Figure S4. PCA of each metabolic profile by plotting the first two components, indicating no discernable separation between transcriptional subtypes or grade.

Malignant tumors exhibit altered metabolism and consume higher levels of glucose compared to surrounding normal tissue, resulting in acidic extracellular microenvironment. Acidity in the microenvironment is a critical stress factor and selection force for the evolution of aggressive tumor types. There are several acid sensing cell surface receptors and ion channels (ASICs) that can sense acidity in the microenvironment; among them proton sensing G protein-coupled receptors (GPCRs) such as OGR1 (GPR68), TDAG8 (GPR65), GPR4 and GPR132 form a major class of acid receptors. Our previous studies demonstrated that acid adaptation is associated with survival mechanisms like chronic activation of autophagy and redistribution of the lysosomal proteins to the plasma membrane. When grown under acidic pH, breast cancer cells accumulate lipids as revealed by staining with Nile Red and perilipin 2, a protein that coats adiposomes or lipid droplets. These are dynamic organelles that store neutral lipids surrounded by a shell of proteins and a phospholipid monolayer. The lipids stored in adiposomes are produced de novo, as acid-induced lipogenic phenotype is maintained, even if cells are grown with de-lipidated serum. Inhibition of fatty acid synthesis was selectively toxic under acidic conditions and attenuated adiposome accumulation. Among the acid sensing receptors, TDAG8 and OGR1 are highly expressed in a panel of breast cancer cell lines compared to non-malignant breast epithelial cells. Highly invasive and acid tolerant MDA-MB-231 cells express significantly higher levels of TDAG8 while MCF7 and T47D cells have high levels of OGR1. We investigated the role of these receptors in transducing the acid signal that results in the accumulation of lipid droplets. CRISPR/Cas9 mediated depletion of the major acid sensors in breast cancer cell lines MCF7 and T47D showed that depletion of OGR1, not TDAG8, specifically inhibited acid induced adiposome accumulation. Additionally, inducing the cells with ogerin, an allosteric activator of OGR1 resulted in adiposomogenesis. Further, OGR1 knock out cells showed inhibition in cell growth under acidic growth conditions compared to OGR1 expressing cells. OGR1 knockout cells were defective in acid induced autophagy. OGR1 is coupled with Gq/11 and, upon ligand (H+) binding, triggers activation of phospholipase C stimulating the formation of the second messenger IP3. Adiposome formation was inhibited in presence of PLC inhibitors, Edelfosine or U73122. These results indicate that OGR1 is the major acid sensing GPCR in these cells that mediate adiposomogenesis. Accumulation of adiposomes is a highly regulated process related to storing autophagic products, and appears to be important in cell survival in acid stress. Taken together, increased dependence on lipid metabolism by cancer cells under acidic conditions reveals novel therapeutic vulnerabilities. Citation Format: Smitha Pillai, Michael Langsen, Jonathan Nguyen, Jonathan Wojtkowiak, Marilyn Bui, Robert Gatenby, Robert Gillies. Acid sensing G protein-coupled receptor OGR1 is required for acid induced adiposomogenesis in breast cancer cells [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 1846.