Machine learning in onco-pharmacogenomics: a path to precision medicine with many challenges

Over the past two decades, Next-Generation Sequencing (NGS) has revolutionized the approach to cancer research. Applications of NGS include the identification of tumor specific alterations that can influence tumor pathobiology and also impact diagnosis, prognosis and therapeutic options. Pharmacogenomics (PGx) studies the role of inheritance of individual genetic patterns in drug response and has taken advantage of NGS technology as it provides access to high-throughput data that can, however, be difficult to manage. Machine learning (ML) has recently been used in the life sciences to discover hidden patterns from complex NGS data and to solve various PGx problems. In this review, we provide a comprehensive overview of the NGS approaches that can be employed and the different PGx studies implicating the use of NGS data. We also provide an excursus of the ML algorithms that can exert a role as fundamental strategies in the PGx field to improve personalized medicine in cancer.

Machine Learning Application Identifies Germline Markers of Hypertension in Patients With Ovarian Cancer Treated With Carboplatin, Taxane, and Bevacizumab

Pharmacogenomics studies how genes influence a person's response to treatment. When complex phenotypes are influenced by multiple genetic variations with little effect, a single piece of genetic information is often insufficient to explain this variability. The application of machine learning (ML) in pharmacogenomics holds great potential - namely, it can be used to unravel complicated genetic relationships that could explain response to therapy. In this study, ML techniques were used to investigate the relationship between genetic variations affecting more than 60 candidate genes and carboplatin-induced, taxane-induced, and bevacizumab-induced toxicities in 171 patients with ovarian cancer enrolled in the MITO-16A/MaNGO-OV2A trial. Single-nucleotide variation (SNV, formerly SNP) profiles were examined using ML to find and prioritize those associated with drug-induced toxicities, specifically hypertension, hematological toxicity, nonhematological toxicity, and proteinuria. The Boruta algorithm was used in cross-validation to determine the significance of SNVs in predicting toxicities. Important SNVs were then used to train eXtreme gradient boosting models. During cross-validation, the models achieved reliable performance with a Matthews correlation coefficient ranging from 0.375 to 0.410. A total of 43 SNVs critical for predicting toxicity were identified. For each toxicity, key SNVs were used to create a polygenic toxicity risk score that effectively divided individuals into high-risk and low-risk categories. In particular, compared with low-risk individuals, high-risk patients were 28-fold more likely to develop hypertension. The proposed method provided insightful data to improve precision medicine for patients with ovarian cancer, which may be useful for reducing toxicities and improving toxicity management.

Machine learning to improve interpretability of clinical, radiological and panel-based genomic data of glioma grade 4 patients undergoing surgical resection

BACKGROUND

Glioma grade 4 (GG4) tumors, including astrocytoma IDH-mutant grade 4 and the astrocytoma IDH wt are the most common and aggressive primary tumors of the central nervous system. Surgery followed by Stupp protocol still remains the first-line treatment in GG4 tumors. Although Stupp combination can prolong survival, prognosis of treated adult patients with GG4 still remains unfavorable. The introduction of innovative multi-parametric prognostic models may allow refinement of prognosis of these patients. Here, Machine Learning (ML) was applied to investigate the contribution in predicting overall survival (OS) of different available data (e.g. clinical data, radiological data, or panel-based sequencing data such as presence of somatic mutations and amplification) in a mono-institutional GG4 cohort.

METHODS

By next-generation sequencing, using a panel of 523 genes, we performed analysis of copy number variations and of types and distribution of nonsynonymous mutations in 102 cases including 39 carmustine wafer (CW) treated cases. We also calculated tumor mutational burden (TMB). ML was applied using eXtreme Gradient Boosting for survival (XGBoost-Surv) to integrate clinical and radiological information with genomic data.

RESULTS

By ML modeling (concordance (c)- index = 0.682 for the best model), the role of predicting OS of radiological parameters including extent of resection, preoperative volume and residual volume was confirmed. An association between CW application and longer OS was also showed. Regarding gene mutations, a role in predicting OS was defined for mutations of BRAF and of other genes involved in the PI3K-AKT-mTOR signaling pathway. Moreover, an association between high TMB and shorter OS was suggested. Consistently, when a cutoff of 1.7 mutations/megabase was applied, cases with higher TMB showed significantly shorter OS than cases with lower TMB.

CONCLUSIONS

The contribution of tumor volumetric data, somatic gene mutations and TBM in predicting OS of GG4 patients was defined by ML modeling.

Colorectal cancer development is affected by the ECM molecule EMILIN-2 hinging on macrophage polarization via the TLR-4/MyD88 pathway

BACKGROUND

Colorectal cancer is one of the most frequent and deadly tumors. Among the key regulators of CRC growth and progression, the microenvironment has emerged as a crucial player and as a possible route for the development of new therapeutic opportunities. More specifically, the extracellular matrix acts directly on cancer cells and indirectly affecting the behavior of stromal and inflammatory cells, as well as the bioavailability of growth factors. Among the ECM molecules, EMILIN-2 is frequently down-regulated by methylation in CRC and the purpose of this study was to verify the impact of EMILIN-2 loss in CRC development and its possible value as a prognostic biomarker.

METHODS

The AOM/DSS CRC protocol was applied to Emilin-2 null and wild type mice. Tumor development was monitored by endoscopy, the molecular analyses performed by IHC, IF and WB and the immune subpopulations characterized by flow cytometry. Ex vivo cultures of monocyte/macrophages from the murine models were used to verify the molecular pathways. Publicly available datasets were exploited to determine the CRC patients' expression profile; Spearman's correlation analyses and Cox regression were applied to evaluate the association with the inflammatory response; the clinical outcome was predicted by Kaplan-Meier survival curves. Pearson correlation analyses were also applied to a cohort of patients enrolled in our Institute.

RESULTS

In preclinical settings, loss of EMILIN-2 associated with an increased number of tumor lesions upon AOM/DSS treatment. In addition, in the early stages of the disease, the Emilin-2 knockout mice displayed a myeloid-derived suppressor cells-rich infiltrate. Instead, in the late stages, lack of EMILIN-2 associated with a decreased number of M1 macrophages, resulting in a higher percentage of the tumor-promoting M2 macrophages. Mechanistically, EMILIN-2 triggered the activation of the Toll-like Receptor 4/MyD88/NF-κB pathway, instrumental for the polarization of macrophages towards the M1 phenotype. Accordingly, dataset and immunofluorescence analyses indicated that low EMILIN-2 expression levels correlated with an increased M2/M1 ratio and with poor CRC patients' prognosis.

CONCLUSIONS

These novel results indicate that EMILIN-2 is a key regulator of the tumor-associated inflammatory environment and may represent a promising prognostic biomarker for CRC patients.

Machine Learning Application in a Phase I Clinical Trial Allows for the Identification of Clinical-Biomolecular Markers Significantly Associated with Toxicity

Machine learning (ML) algorithms have been used to forecast clinical outcomes or drug adverse effects by analyzing different data sets such as electronic health records, diagnostic data, and molecular data. However, ML implementation in phase I clinical trial is still an unexplored strategy that implies challenges such as the selection of the best development strategy when dealing with limited sample size. In the attempt to better define prechemotherapy baseline clinical and biomolecular predictors of drug toxicity, we trained and compared five ML algorithms starting from clinical, blood biochemistry, and genotype data derived from a previous phase Ib study aimed to define the maximum tolerated dose of irinotecan (FOLFIRI (folinic acid, fluorouracil, and irinotecan) plus bevacizumab regimen) in patients with metastatic colorectal cancer. During cross-validation the Random Forest algorithm achieved the best performance with a mean Matthews correlation coefficient of 0.549 and a mean accuracy of 80.4%; the best predictors of dose-limiting toxicity at baseline were hemoglobin, serum glutamic oxaloacetic transaminase (SGOT), and albumin. The feasibility of a prediction model prototype was in principle assessed using the two distinct dose escalation cohorts, where in the validation cohort the model scored a Matthews correlation coefficient of 0.59 and an accuracy of 82.0%. Moreover, we found a strong relationship between SGOT and irinotecan pharmacokinetics, suggesting its role as surrogates' estimators of the irinotecan metabolism equilibrium. In conclusion, the potential application of ML techniques to phase I study could provide clinicians with early prediction tools useful both to ameliorate the management of clinical trials and to make more adequate treatment decisions.

The benefit of early surgery on overall survival in incidental low grade glioma patients: a multicenter study

BACKGROUND

The role of surgery for incidentally discovered diffuse low-grade gliomas (iLGGs) is debatable and poorly documented in current literature.

OBJECTIVE

The aim was to identify factors that influence survival for patients that underwent surgical resection of iLGGs in a large multicenter population.

METHODS

Clinical, radiological, and surgical data were retrospectively analyzed in 267 patients operated for iLGG from 4 neurosurgical Centers. Univariate and multivariate analyses were performed to identify predictors of overall survival (OS) and tumor recurrence (TR).

RESULTS

The OS rate was 92.41%. The 5- and 10-year estimated OS rates were 98.09% and 93.2% respectively. OS was significantly longer for patients with a lower preoperative tumor volume (p=0.001) and higher extent of resection (EOR) (p=0.037), regardless the WHO defined molecular class (p=0.2). In the final model, OS was influenced only by the preoperative tumor volume (p=0.006), while TR by early surgery (p=0.028). A negative association was found between preoperative tumor volumes and EOR (rs = -0.44, p<0.001).The median preoperative tumor volume was 15 cm 3. The median EOR was 95%. Total or supratotal resection of FLAIR abnormality was achieved in 61.62% of cases.Second surgery was performed in 26.22%. The median time between surgeries was 5.5 years. Histological evolution to high grade glioma was detected in 22.85% of cases (16/70). Permanent mild deficits were observed in 3.08% of cases.

CONCLUSIONS

This multicenter study confirms the results of previous studies investigating surgical management of iLGGs and thereby strengthens the evidence in favour of early surgery for these lesions.

Toward best practice in cancer mutation detection with whole-genome and whole-exome sequencing

Clinical applications of precision oncology require accurate tests that can distinguish true cancer-specific mutations from errors introduced at each step of next-generation sequencing (NGS). To date, no bulk sequencing study has addressed the effects of cross-site reproducibility, nor the biological, technical and computational factors that influence variant identification. Here we report a systematic interrogation of somatic mutations in paired tumor-normal cell lines to identify factors affecting detection reproducibility and accuracy at six different centers. Using whole-genome sequencing (WGS) and whole-exome sequencing (WES), we evaluated the reproducibility of different sample types with varying input amount and tumor purity, and multiple library construction protocols, followed by processing with nine bioinformatics pipelines. We found that read coverage and callers affected both WGS and WES reproducibility, but WES performance was influenced by insert fragment size, genomic copy content and the global imbalance score (GIV; G > T/C > A). Finally, taking into account library preparation protocol, tumor content, read coverage and bioinformatics processes concomitantly, we recommend actionable practices to improve the reproducibility and accuracy of NGS experiments for cancer mutation detection.

The Efficacy of Anti-PD-L1 Treatment in Melanoma Is Associated with the Expression of the ECM Molecule EMILIN2

The use of immune checkpoint inhibitors has revolutionized the treatment of melanoma patients, leading to remarkable improvements in the cure. However, to ensure a safe and effective treatment, there is the need to develop markers to identify the patients that would most likely respond to the therapies. The microenvironment is gaining attention in this context, since it can regulate both the immunotherapy efficacyand angiogenesis, which is known to be affected by treatment. Here, we investigated the putative role of the ECM molecule EMILIN-2, a tumor suppressive and pro-angiogenic molecule. We verified that the EMILIN2 expression is variable among melanoma patients and is associated with the response to PD-L1 inhibitors. Consistently, in preclinical settings, the absence of EMILIN-2 is associated with higher PD-L1 expression and increased immunotherapy efficacy. We verified that EMILIN-2 modulates PD-L1 expression in melanoma cells through indirect immune-dependent mechanisms. Notably, upon PD-L1 blockage, Emilin2^−/− mice displayed improved intra-tumoral vessel normalization and decreased tumor hypoxia. Finally, we provide evidence indicating that the inclusion of EMILIN2 in a number of gene expression signatures improves their predictive potential, a further indication that the analysis of this molecule may be key for the development of new markers to predict immunotherapy efficacy.

A New Epigenetic Model to Stratify Glioma Patients According to Their Immunosuppressive State

Gliomas are the most common primary neoplasm of the central nervous system. A promising frontier in the definition of glioma prognosis and treatment is represented by epigenetics. Furthermore, in this study, we developed a machine learning classification model based on epigenetic data (CpG probes) to separate patients according to their state of immunosuppression. We considered 573 cases of low-grade glioma (LGG) and glioblastoma (GBM) from The Cancer Genome Atlas (TCGA). First, from gene expression data, we derived a novel binary indicator to flag patients with a favorable immune state. Then, based on previous studies, we selected the genes related to the immune state of tumor microenvironment. After, we improved the selection with a data-driven procedure, based on Boruta. Finally, we tuned, trained, and evaluated both random forest and neural network classifiers on the resulting dataset. We found that a multi-layer perceptron network fed by the 338 probes selected by applying both expert choice and Boruta results in the best performance, achieving an out-of-sample accuracy of 82.8%, a Matthews correlation coefficient of 0.657, and an area under the ROC curve of 0.9. Based on the proposed model, we provided a method to stratify glioma patients according to their epigenomic state.

Long Noncoding RNA NEAT1 Acts as a Molecular Switch for BRD4 Transcriptional Activity and Mediates Repression of BRD4/WDR5 Target Genes

BRD4 is an epigenome reader known to exert key roles at the interface between chromatin remodeling and transcriptional regulation, and is primarily known for its role in promoting gene expression. In selective contexts, however, BRD4 may work as negative regulator of transcription. Here, we reported that BRD4 binds several long noncoding RNAs (lncRNA). Among these, the lncRNA NEAT1 was found to interfere with BRD4 transcriptional activity. Mechanistically, lncNEAT1 forms a complex with BRD4 and WDR5 and maintains them in a low-activity state. Treatment with Bromodomains and Extraterminal (BET) inhibitor caused the lncRNA NEAT1 to dissociate from the BRD4/WDR5 complex, restored the acetyl-transferase capacity of BRD4, and restored the availability of WDR5 to promote histone trimethylation, thereby promoting BRD4/WDR5 transcriptional activity and activation of target gene expression. In addition, the lncRNA NEAT1 then became available to bind and to inhibit EZH2, cooperatively increasing transcriptional activation. IMPLICATIONS: Our results revealed an epigenetic program that involves the interaction between the lncRNA NEAT1 and BRD4, functioning as a molecular switch between BRD4's activator and repressor chromatin complexes.

A Novel Epigenetic Machine Learning Model to Define Risk of Progression for Hepatocellular Carcinoma Patients

Although extensive advancements have been made in treatment against hepatocellular carcinoma (HCC), the prognosis of HCC patients remains unsatisfied. It is now clearly established that extensive epigenetic changes act as a driver in human tumors. This study exploits HCC epigenetic deregulation to define a novel prognostic model for monitoring the progression of HCC. We analyzed the genome-wide DNA methylation profile of 374 primary tumor specimens using the Illumina 450 K array data from The Cancer Genome Atlas. We initially used a novel combination of Machine Learning algorithms (Recursive Features Selection, Boruta) to capture early tumor progression features. The subsets of probes obtained were used to train and validate Random Forest models to predict a Progression Free Survival greater or less than 6 months. The model based on 34 epigenetic probes showed the best performance, scoring 0.80 accuracy and 0.51 Matthews Correlation Coefficient on testset. Then, we generated and validated a progression signature based on 4 methylation probes capable of stratifying HCC patients at high and low risk of progression. Survival analysis showed that high risk patients are characterized by a poorer progression free survival compared to low risk patients. Moreover, decision curve analysis confirmed the strength of this predictive tool over conventional clinical parameters. Functional enrichment analysis highlighted that high risk patients differentiated themselves by the upregulation of proliferative pathways. Ultimately, we propose the oncogenic MCM2 gene as a methylation-driven gene of which the representative epigenetic markers could serve both as predictive and prognostic markers. Briefly, our work provides several potential HCC progression epigenetic biomarkers as well as a new signature that may enhance patients surveillance and advances in personalized treatment.

The Molecular and Microenvironmental Landscape of Glioblastomas: Implications for the Novel Treatment Choices

Glioblastoma (GBM) is the most frequent and aggressive primary central nervous system tumor. Surgery followed by radiotherapy and chemotherapy with alkylating agents constitutes standard first-line treatment of GBM. Complete resection of the GBM tumors is generally not possible given its high invasive features. Although this combination therapy can prolong survival, the prognosis is still poor due to several factors including chemoresistance. In recent years, a comprehensive characterization of the GBM-associated molecular signature has been performed. This has allowed the possibility to introduce a more personalized therapeutic approach for GBM, in which novel targeted therapies, including those employing tyrosine kinase inhibitors (TKIs), could be employed. The GBM tumor microenvironment (TME) exerts a key role in GBM tumor progression, in particular by providing an immunosuppressive state with low numbers of tumor-infiltrating lymphocytes (TILs) and other immune effector cell types that contributes to tumor proliferation and growth. The use of immune checkpoint inhibitors (ICIs) has been successfully introduced in numerous advanced cancers as well as promising results have been shown for the use of these antibodies in untreated brain metastases from melanoma and from non-small cell lung carcinoma (NSCLC). Consequently, the use of PD-1/PD-L1 inhibitors has also been proposed in several clinical trials for the treatment of GBM. In the present review, we will outline the main GBM molecular and TME aspects providing also the grounds for novel targeted therapies and immunotherapies using ICIs for GBM.

Prognostic role of macrophage infiltration and monocyte-to-lymphocyte ratio in stage III colon cancer: The MIRROR study

e16118

Background: Changes in peripheral blood cells composition may reflect immune microenvironment and its role in cancer growth. High monocyte-to-lymphocyte ratio (MLR) could be a marker of tumor’s recruitment of suppressive cells, showing a prognostic role. This study aimed to assess the prognostic impact of macrophage infiltration and MLR in stage III colon cancer (CC) patients (pts). Methods: This multicentric study retrospectively analyzed a consecutive cohort of 423 CC pts treated between 2008-2019 at the Cancer Centre of Aviano (Italy) (n = 300) and at the European Georges Pompidou Hospital of Paris (France) (n = 123). The association of MLR with disease-free survival (DFS) and overall survival (OS) was evaluated with Cox regression analyses. Random Forrest was implemented on python using h2oai. Performance was assessed in terms of accuracy (ACC) and Matthews Coefficient (MCC). Analyses was adjusted on classical prognostic factors of stage III CC such as pT, pN, grade, location, ECOG PS. Results: Overall, 77% had pT1-3, 30% pN2 and 73% G1-2 tumors. Interestingly, 25% had a lymphatic and vascular invasion, 42/230 (18%) had MSI status, 69/152(45%) and 19/114 (13%) were KRAS and BRAF mutant. 56% had CEA > 5. Pts were treated with fluoropyrimidine and oxaliplatin as adjuvant therapy. Notably, 130 cases were analyzed according to lymphocytic and macrophage infiltration (CD163, CD68, CD3, CD8). Of them, 78% had a CD163/CD8 ratio ≤3 and 74% a CD8/CD3 ratio ≤1.5. At median follow-up of 57 months, median DFS and OS were not reached, 31% of pts relapsed and 23% dead. By multivariate analysis, including statistically significant prognostic variables, CD163/CD8 ratio (HR 1.15, p = 0.039, 95%C.I. 1.1-1.32) and MLR > 0.45 (HR 2.98, p = 0.008, 95%C.I. 1.33-6.67) were associated with worse DFS. By multivariate analysis for OS, including statistically significant confounding variables, MLR > 0.45 (HR4.32, P = 0.012, 95%C.I. 1.37-9) and BRAF mutation predicted worse OS. According random survival forest for OS, CD68/CD3 were the first variable of importance (0.06), followed by MLR (0.009) and CD8 (0.007). Interestingly, high MLR followed by CEA, MSI, KRAS were the features linked with organotropism on liver (ACC 0.6 ±0.3), while high MLR, KRAS, pN, pT were mainly linked with lung colonization (ACC 0.6 ±0.2). Conclusions: High pre-treatment levels of MLR and CD163/CD8 ratio in stage III CC are independently associated with worse prognosis. The present study paves the way to a prospective validation of these promising cost-effective biomarkers.

Abstract P3-01-05: Liquid biopsy methods and machine learning modeling to understand organ tropism and metastatization behavior of metastatic breast cancer

Background: Liquid biopsy provides a growing amount of real-time data about prognosis and the genomic landscape of metastatic breast cancer (MBC) and its comprehensive analysis is an emerging clinical need. Machine Learning (ML) data-driven models are able to “learn” information about a system and to adaptively improve their performance by directly observing its data. This enables the discovery of hidden patterns in complex heterogeneous and high dimensional data. The aim of this study was to explore the combination of clinical characteristics, circulating tumor DNA-detected aberrations (ctDNA) and CTC enumeration in estimating target organs more susceptible to MBC involvement using a ML modeling approach. Methods: The study retrospectively analyzed 88 MBC patients (pts) treated and characterized for CTCs and circulating tumor DNA (ctDNA) at Northwestern University (Chicago, IL) independently from treatment line. Blood samples were collected at baseline, concomitantly with imaging. CTCs were isolated through the CellSearch™ kit (Menarini Silicon Biosystems, PA), while ctDNA was analyzed using the Guardant360™ NGS-based assay (Guardant Health, CA). All features were normalized and included in a random forest algorithm implemented in Python (Scikit learn, BSD license), node splitting criterion for the decision tree classifiers was varied using Gini index and entropy. Hyperparameters of the random forest were then optimized including number of trees and the minimum leaf size by implementing hyperparameter grid search using 10-fold cross validation. Results: The median number of lines at baseline collection was 1 (interquartile range: 1-3), while the median number of metastatic sites was 3 (inter quartile range: 1-3) with the most observed sites being bone (37%), lymph nodes (29%), lung (27%) and liver (25%). The cohort consisted of 43% hormone receptor positive (HRpos), 32% TNBC, and 25% HER2-positive MBC. In the overall population, continuous CTC number (n_CTC), inflammatory breast cancer diagnosis (IBC), and aberrations in ESR1, KITand CDK4were the main features linked to liver metastases (AUC: 0.842), n_CTC, ESR1, PIK3CA, CCNE1and CDK6were the features linked to bone involvement (AUC: 0.770), while PIK3CA, METand MYC, were linked to lung organotropism (AUC: 0.701). Factors linked to the metastatization net combination pattern were then explored within each MBC subtype. Intriguingly, AR, n_CTC, TP53and ESR1were the main drivers in HRpos MBC (Mean per class error0.46), while EGFR, KITand NOTCH1were the main features in TNBC (Mean per class error 0.605). Consistently, n_CTC, ERBB2, PIK3CAwere the driving features among HER2 positive MBC pts (Mean per class error 0.87). Conclusions: This novel analysis demonstrates that liquid biopsy integrating both CTCs enumeration and genomic characterization by ctDNA could prove useful in a detailed description of the metastatic process, allowing a more tailored monitoring and therapeutic approach. Intriguingly, features linked to Epithelial to Mesenchymal transition were found to be a potential driver of the metastatization behavior, underlining the need to further elucidate the clinical impact of this process.

Citation Format: Lorenzo Gerratana, Andrew A Davis, Maurizio Polano, Qiang Zhang, Ami N Shah, Chenyu Lin, Debora Basile, Giuseppe Toffoli, Firas Wehbe, Fabio Puglisi, Amir Behdad, Leonidas C Platanias, William J Gradishar, Massimo Cristofanilli. Liquid biopsy methods and machine learning modeling to understand organ tropism and metastatization behavior of metastatic breast cancer [abstract]. In: Proceedings of the 2019 San Antonio Breast Cancer Symposium; 2019 Dec 10-14; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2020;80(4 Suppl):Abstract nr P3-01-05.

A Novel Comprehensive Clinical Stratification Model to Refine Prognosis of Glioblastoma Patients Undergoing Surgical Resection

Despite recent discoveries in genetics and molecular fields, glioblastoma (GBM) prognosis still remains unfavorable with less than 10% of patients alive 5 years after diagnosis. Numerous studies have focused on the research of biological biomarkers to stratify GBM patients. We addressed this issue in our study by using clinical/molecular and image data, which is generally available to Neurosurgical Departments in order to create a prognostic score that can be useful to stratify GBM patients undergoing surgical resection. By using the random forest approach [CART analysis (classification and regression tree)] on Survival time data of 465 cases, we developed a new prediction score resulting in 10 groups based on extent of resection (EOR), age, tumor volumetric features, intraoperative protocols and tumor molecular classes. The resulting tree was trimmed according to similarities in the relative hazard ratios amongst groups, giving rise to a 5-group classification tree. These 5 groups were different in terms of overall survival (OS) (p < 0.000). The score performance in predicting death was defined by a Harrell’s c-index of 0.79 (95% confidence interval [0.76–0.81]). The proposed score could be useful in a clinical setting to refine the prognosis of GBM patients after surgery and prior to postoperative treatment.

Clonal Selection of a Novel Deleterious TP53 Somatic Mutation Discovered in ctDNA of a KIT/PDGFRA Wild-Type Gastrointestinal Stromal Tumor Resistant to Imatinib

The standard of care for the first-line treatment of advanced gastrointestinal stromal tumor (GIST) is represented by imatinib, which is given daily at a standard dosage until tumor progression. Resistance to imatinib commonly occurs through the clonal selection of genetic mutations in the tumor DNA, and an increase in imatinib dosage was demonstrated to be efficacious to overcome imatinib resistance. Wild-type GISTs, which do not display KIT or platelet-derived growth factor receptor alpha (PDGFRA) mutations, are usually primarily insensitive to imatinib and tend to rapidly relapse in course of treatment. Here we report the case of a 53-year-old male patient with gastric GIST who primarily did not respond to imatinib and that, despite the administration of an increased imatinib dose, led to patient death. By using a deep next-generation sequencing barcode-aware approach, we analyzed a panel of actionable cancer-related genes in the patient cfDNA to investigate somatic changes responsible for imatinib resistance. We identified, in two serial circulating tumor DNA (ctDNA) samples, a sharp increase in the allele frequency of a never described TP53 mutation (c.560-7_560-2delCTCTTAinsT) located in a splice acceptor site and responsible for a protein loss of function. The same TP53 mutation was retrospectively identified in the primary tumor by digital droplet PCR at a subclonal frequency (0.1%). The mutation was detected at a very high allelic frequency (99%) in the metastatic hepatic lesion, suggesting a rapid clonal selection of the mutation during tumor progression. Imatinib plasma concentration at steady state was above the threshold of 760 ng/ml reported in the literature for the minimum efficacious concentration. The de novo TP53 (c.560-7_560-2delCTCTTAinsT) mutation was in silico predicted to be associated with an aberrant RNA splicing and with an aggressive phenotype which might have contributed to a rapid disease spread despite the administration of an increased imatinib dosage. This result underlies the need of a better investigation upon the role of TP53 in the pathogenesis of GISTs and sustains the use of next-generation sequencing (NGS) in cfDNA for the identification of novel genetic markers in wild-type GISTs.

NR4A3 fusion proteins trigger an axon guidance switch that marks the difference between EWSR1 and TAF15 translocated extraskeletal myxoid chondrosarcomas

Extraskeletal myxoid chondrosarcoma (EMC) is a rare sarcoma histotype with uncertain differentiation. EMC is hallmarked by the rearrangement of the NR4A3 gene, which in most cases fuses with EWSR1 or TAF15. TAF15‐translocated EMC seem to feature a more aggressive course compared to EWSR1‐positive EMCs, but whether the type of NR4A3 chimera impinges upon EMC biology is still largely undefined. To gain insights on this issue, a series of EMC samples (7 EWSR1‐NR4A3 and 5 TAF15‐NR4A3) were transcriptionally profiled. Our study unveiled that the two EMC variants display a distinct transcriptional profile and that the axon guidance pathway is a major discriminant. In particular, class 4–6 semaphorins and axonal guidance cues endowed with pro‐tumorigenic activity were more expressed in TAF15‐NR4A3 tumors; vice versa, class 3 semaphorins, considered to convey growth inhibitory signals, were more abundant in EWSR1‐NR4A3 EMC. Intriguingly, the dichotomy in axon guidance signaling observed in the two tumor variants was recapitulated in in vitro cell models engineered to ectopically express EWSR1‐NR4A3 or TAF15‐NR4A3. Moreover, TAF15‐NR4A3 cells displayed a more pronounced tumorigenic potential, as assessed by anchorage‐independent growth. Overall, our results indicate that the type of NR4A3 chimera dictates an axon guidance switch and impacts on tumor cell biology. These findings may provide a framework for interpretation of the different clinical–pathological features of the two EMC variants and lay down the bases for the development of novel patient stratification criteria and therapeutic approaches. © 2019 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of Pathological Society of Great Britain and Ireland.

HSPA8 as a novel fusion partner of NR4A3 in extraskeletal myxoid chondrosarcoma

Extraskeletal myxoid chondrosarcoma (EMC) is a very rare sarcoma most often arising in the soft tissue. Rare EMC of the bone have been reported. EMC exhibits distinctive clinico-pathological and genetic features; however, despite the name, it lacks any feature of cartilaginous differentiation. EMC is characterized by the rearrangement of the NR4A3, which, in most cases (about 62-75%), is fused with EWSR1 and less frequently with other partners, including TAF15 (27%), TCF12 (4%), TFG, and FUS. We herein report the identification by whole-transcriptome sequencing of HSPA8 as a novel fusion partner of NR4A3 in a case of EMC. FISH analysis confirmed the presence of a genomic HSPA8-NR4A3 translocation in the vast majority of tumor cells. Our findings expand the spectrum of NR4A3 fusion partners involved in EMC pathobiology.

The co-existence of transcriptional activator and transcriptional repressor MEF2 complexes influences tumor aggressiveness

The contribution of MEF2 TFs to the tumorigenic process is still mysterious. Here we clarify that MEF2 can support both pro-oncogenic or tumor suppressive activities depending on the interaction with co-activators or co-repressors partners. Through these interactions MEF2 supervise histone modifications associated with gene activation/repression, such as H3K4 methylation and H3K27 acetylation. Critical switches for the generation of a MEF2 repressive environment are class IIa HDACs. In leiomyosarcomas (LMS), this two-faced trait of MEF2 is relevant for tumor aggressiveness. Class IIa HDACs are overexpressed in 22% of LMS, where high levels of MEF2, HDAC4 and HDAC9 inversely correlate with overall survival. The knock out of HDAC9 suppresses the transformed phenotype of LMS cells, by restoring the transcriptional proficiency of some MEF2-target loci. HDAC9 coordinates also the demethylation of H3K4me3 at the promoters of MEF2-target genes. Moreover, we show that class IIa HDACs do not bind all the regulative elements bound by MEF2. Hence, in a cell MEF2-target genes actively transcribed and strongly repressed can coexist. However, these repressed MEF2-targets are poised in terms of chromatin signature. Overall our results candidate class IIa HDACs and HDAC9 in particular, as druggable targets for a therapeutic intervention in LMS.

The tumorigenic process is characterized by profound alterations of the transcriptional landscape, aimed to sustain uncontrolled cell growth, resistance to apoptosis and metastasis. The contribution of MEF2, a pleiotropic family of transcription factors, to these changes is controversial, since both pro-oncogenic and tumor-suppressive activities have been reported. To clarify this paradox, we studied the role of MEF2 in an aggressive type of soft-tissue sarcomas, the leiomyosarcomas (LMS). We found that in LMS cells MEF2 become oncogenes when in complex with class IIa HDACs. We have identified different sub-classes of MEF2-target genes and observed that HDAC9 converts MEF2 into transcriptional repressors on some, but not all, MEF2-regulated loci. This conversion correlates with the acquisition by MEF2 of oncogenic properties. We have also elucidated some epigenetic re-arrangements supervised by MEF2. In summary, our studies suggest that the paradoxical actions of MEF2 in cancer can be explained by their dual role as activators/repressors of transcription and open new possibilities for therapeutic interventions.

Abstract LB-329: Enhancing the resolution and accelerating the pace of translational fusion characterization in oncology by RNA sequencing

Chromosomal rearrangements are common markers of cancer progression across a wide range of cancer types, and therefore, identification of fusion transcripts in cancer biopsies may have potential to provide tumor-specific insight toward diagnosis, prognosis and precision treatment. Currently, routine methods for fusion detection using fluorescent in-situ hybridization (FISH) provide a low-resolution view of the aberrant fusion transcript. We describe an RNA-Seq approach designed to survey cancer fusions in a single assay by selectively enriching the cancer transcriptome using probes that target the coding regions of over 1385 cancer-associated genes.

We tested the performance of the 1385 gene, RNA-Seq Pan-Cancer panel on RNA extracted from 47 patient-derived samples from brain, sarcoma and leukemia, including blood, bone marrow, and formalin-fixed paraffin-embedded (FFPE) samples. Each sample harbored at least one orthogonally verified gene fusion transcript, previously confirmed by FISH or Reverse Transcriptase PCR (RT-PCR). RNA-Seq libraries were prepared from 10-100 ng of total RNA from blood or bone marrow and 20-200 ng total RNA from FFPE tissue and subsequently enriched by hybridization to the Pan-Cancer panel. All samples yielded sufficient library and were sequenced with 76 base-pair paired-end reads on an Illumina MiSeq at 8 samples per flow cell (∼3 million reads per sample). Sequencing data was analyzed using RNA-Seq with STAR aligner and Manta fusion caller. Using this capture-based single-assay approach, we successfully detected fusions commonly associated with leukemia (BCR-ABL1, MLL-MLLT3, MLL-AFF1, RUNX1-ETV6, EBF1-PDGFRB, TCF3-PBX1, IKZF1-PAX5), sarcoma (EWSR1-ATF1, EWSR1-FLI1, JAZF1-SUZ12, SS18-SSX, FUS-DDIT3, FUS-KLF17, YWHAE-FAM22B) and brain cancer (KIAA1459-BRAF) consistent with previously confirmed RT-PCR or FISH results. Several examples of previously unknown fusion partners or additional structural information that were not identified from the FISH or RT-PCR testing were also uncovered in this study. These cases are described in detail.

In summary, we show that selective enrichment of RNA-Seq libraries with cancer-specific probes enables detection of known and novel fusions across a broad range of cancer pathologies in a single reaction, creating new opportunities for discovery and translational cancer studies.

Citation Format: Lisa C. Watson, Stephen M. Gross, Felix Schlesinger, Anthony Mai, Mariko Kellogg, Steve Lee, Claire Attwooll, Monica Brenca, David Swanson, Andrew Wong, Angelo P. Dei Tos, Claudia Haferlach, Torsten Haferlach, Wolfgang Kern, Roberta Maestro, Manja Meggendorfer, Niroshan Nadarajah, Maurizio Polano, Sabrina Rossi, Marta Sbaraglia, George S. Charames, Gary P. Schroth, Grace DeSantis. Enhancing the resolution and accelerating the pace of translational fusion characterization in oncology by RNA sequencing. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr LB-329.

Quadruple-Negative GIST Is a Sentinel for Unrecognized Neurofibromatosis Type 1 Syndrome

PURPOSE

The majority of gastrointestinal stromal tumors (GIST) are driven by KIT, PDGFRA, or, less commonly, BRAF mutations, and SDH gene inactivation is involved in a limited fraction of gastric lesions. However, about 10% of GISTs are devoid of any of such alterations and are poorly responsive to standard treatments. This study aims to shed light on the molecular drivers of quadruple-negative GISTs.

EXPERIMENTAL DESIGN

Twenty-two sporadic quadruple-negative GISTs with no prior association with Neurofibromatosis Type 1 syndrome were molecularly profiled for a panel of genes belonging to tyrosine kinase pathways or previously implicated in GISTs. For comparison purposes, 24 GISTs carrying KIT, PDGFRA, or SDH gene mutations were also analyzed. Molecular findings were correlated to clinicopathologic features.

RESULTS

Most quadruple-negative GISTs featured intestinal localization, with a female predilection. About 60% (13/22) of quadruple-negative tumors carried NF1 pathogenic mutations, often associated with biallelic inactivation. The analysis of normal tissues, available in 11 cases, indicated the constitutional nature of the NF1 mutation in 7 of 11 cases, unveiling an unrecognized Neurofibromatosis Type 1 syndromic condition. Multifocality and a multinodular pattern of growth were common findings in NF1-mutated quadruple-negative GISTs.

CONCLUSIONS

NF1 gene mutations are frequent in quadruple-negative GISTs and are often constitutional, indicating that a significant fraction of patients with apparently sporadic quadruple-negative GISTs are affected by unrecognized Neurofibromatosis Type 1 syndrome. Hence, a diagnosis of quadruple-negative GIST, especially if multifocal or with a multinodular growth pattern and a nongastric location, should alert the clinician to a possible Neurofibromatosis Type 1 syndromic condition. Clin Cancer Res; 23(1); 273-82. ©2016 AACR.

Transcriptome sequencing identifies ETV6-NTRK3 as a gene fusion involved in GIST

Gastrointestinal stromal tumours (GISTs) are the most common mesenchymal neoplasms of the gastrointestinal tract. The vast majority of GISTs are driven by oncogenic activation of KIT, PDGFRA or, less commonly, BRAF. Loss of succinate dehydrogenase complex activity has been identified in subsets of KIT/PDGFRA/BRAF-mutation negative tumours, yet a significant fraction of GISTs are devoid of any of such alterations. To address the pathobiology of these 'quadruple-negative' GISTs, we sought to explore the possible involvement of fusion genes. To this end we performed transcriptome sequencing on five KIT/PDGFRA/BRAF-mutation negative, SDH-proficient tumours. Intriguingly, the analysis unveiled the presence of an ETV6-NTRK3 gene fusion. The screening by FISH of 26 additional cases, including KIT/PDGFRA-mutated GISTs, failed to detect other ETV6 rearrangements beside the index case. This was a 'quadruple-negative' GIST located in the rectum, an uncommon primary site for GIST development (∼4% of all GISTs). The fusion transcript identified encompasses exon 4 of ETV6 and exon 14 of NTRK3 and therefore differs from the canonical ETV6-NTRK3 chimera of infantile fibrosarcomas. However, it retains the ability to induce IRS1 phosphorylation, activate the IGF1R downstream signalling pathway and to be targeted by IGF1R and ALK inhibitors. Thus, the ETV6-NTRK3 fusion might identify a subset of GISTs with peculiar clinicopathological characteristics which could be eligible for such therapies. Copyright © 2015 Pathological Society of Great Britain and Ireland. Published by John Wiley & Sons, Ltd.

Monitoring the interaction between β2-microglobulin and the molecular chaperone αB-crystallin by NMR and mass spectrometry: αB-crystallin dissociates β2-microglobulin oligomers

The interaction at neutral pH between wild-type and a variant form (R3A) of the amyloid fibril-forming protein β2-microglobulin (β2m) and the molecular chaperone αB-crystallin was investigated by thioflavin T fluorescence, NMR spectroscopy, and mass spectrometry. Fibril formation of R3Aβ2m was potently prevented by αB-crystallin. αB-crystallin also prevented the unfolding and nonfibrillar aggregation of R3Aβ2m. From analysis of the NMR spectra collected at various R3Aβ2m to αB-crystallin molar subunit ratios, it is concluded that the structured β-sheet core and the apical loops of R3Aβ2m interact in a nonspecific manner with the αB-crystallin. Complementary information was derived from NMR diffusion coefficient measurements of wild-type β2m at a 100-fold concentration excess with respect to αB-crystallin. Mass spectrometry acquired in the native state showed that the onset of wild-type β2m oligomerization was effectively reduced by αB-crystallin. Furthermore, and most importantly, αB-crystallin reversibly dissociated β2m oligomers formed spontaneously in aged samples. These results, coupled with our previous studies, highlight the potent effectiveness of αB-crystallin in preventing β2m aggregation at the various stages of its aggregation pathway. Our findings are highly relevant to the emerging view that molecular chaperone action is intimately involved in the prevention of in vivo amyloid fibril formation.

The interaction of humic substances with the human prion protein fragment 90-231 affects its protease K resistance and cell internalization

In this paper we analyzed the determinants and the structural effects of the interaction of human prion protein fragment 90-231 (HuPrP) with humic substances, (HS) including humic (HA) and fulvic (FA) acids, natural refractory organic polyanions widely diffused in soils and waters. We show that this interaction is mainly driven by non-specific electrostatic attraction involving regions situated within alpha-helix A and beta-sheet S1 of human PrP. FA binding to HuPrP altered its ability to acquire some PrPSc-like characteristics induced by the mild thermal denaturation of the peptide (1 h at 53 degrees C). In particular, in the presence of FA, HuPrP shows a reduced amount of beta-sheet content (as demonstrated by the reduced binding of thioflavin T), an increased sensitivity to protease K and an inhibition of the entering in the fibrillogenic pathway. FA/HuPrP interaction caused the aggregation of the peptide in unstructured macrocomplexes, as demonstrated by the altered electrophoretic migration in semi-denaturing detergent-agarose gel assay. Importantly, in the presence of FA the rate of internalization of HuPrP in human neuroblastoma cells was significantly reduced as compared to that of the beta-structured peptide. Therefore, HS inhibited the acquisition of PrP(Sc)-like structural properties that, in turn, are responsible for HuPrP intracellular accumulation and lead to neuronal death. Important implications of these data are that HuPrP-HS complexes, being unable to be internalized in living cells may represent a molecular mechanism for the reduced transmission of prion transmission from HS-rich soil also in the presence of contamination from infected animals.