Tumor Sequencing Data Research Articles

Abstract A024: MatchMiner: An open-source computational platform for genomically-driven matching of cancer patients to precision medicine clinical trials

Abstract To facilitate interpretation of complex tumor sequencing data and clinical trial genomic eligibility criteria, we developed MatchMiner, an open-source platform to computationally match cancer patients to precision medicine clinical trials based on clinical and genomic features. MatchMiner supports two distinct workflows: (1) a patient-centric mode, in which an oncologist can find clinical trial matches for a specific patient, and (2) a trial-centric mode, in which a clinical trial investigator can identify and recruit patients for a specific trial. MatchMiner has been operational at Dana-Farber Cancer Institute since early 2017. There are currently 250+ trials curated in the system and genomic data from 24,000+ patients. Over 82% of living patients match to at least one open clinical trial, with an average of 6 trial matches per patient. At least 85 patients have enrolled on a clinical trial as a result of MatchMiner. The MatchMiner open-source software package is available through GitHub (https://github.com/dfci/matchminer). MatchMiner is a two-tier web application with a Python-based REST application programming interface (API) server and an AngularJS front-end. MatchMiner utilizes Security Assertion Markup Language (SAML)-based authentication and, when hosted behind a secure institutional firewall, is fully HIPAA-compliant. MatchMiner can connect to existing clinical systems, including clinical trial management systems for real-time trial status. To enable computational matching to clinical trials, we developed clinical trial markup language (CTML), a structured format to encode detailed information about a trial. CTML utilizes boolean logic to define clinical (e.g. cancer type), demographic (e.g. age) and genomic (e.g. specific mutations, copy number alterations, structural variants or mutational signatures) eligibility, which can be applied to individual arms of a trial. We recently refactored the core matching algorithm (the matchengine), which improves upon the original matchengine in several ways. While the original matchengine reported the reason for a patient-trial match, the refactored matchengine provides additional, more granular details. In addition, the original matchengine ran at the cohort level, matching all patients to all trials, whereas the refactored matchengine can also run against individual patients or trials, speeding up the matching process. The refactored matchengine is also easily extensible to match based on additional data types. In summary, we have defined a standard for encoding clinical trial information in a structured and computable form, and we have developed an open-source computational trial matching platform to support patient-specific trial identification as well as trial-specific patient recruitment. We are actively collaborating with clinical groups at Dana-Farber Cancer Institute to understand the role of MatchMiner in their clinical workflows, and we are committed to continuing to evolve MatchMiner to meet clinical needs. Citation Format: Tali Mazor, Rachel B Keller, Priti Kumari, James Lindsay, Eric Marriott, Andrea Ovalle, Ethan Siegel, Elizabeth H Williams, Joyce Yu, Michael Hassett, Ethan Cerami. MatchMiner: An open-source computational platform for genomically-driven matching of cancer patients to precision medicine clinical trials [abstract]. In: Proceedings of the AACR-NCI-EORTC International Conference on Molecular Targets and Cancer Therapeutics; 2019 Oct 26-30; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2019;18(12 Suppl):Abstract nr A024. doi:10.1158/1535-7163.TARG-19-A024

Strategy for Identification of an Inherited Leukemia Predisposition in a 299 Patients Cohort with Tumor-Only Sequencing Data

Myeloid neoplasms (MN) are usually sporadic late-onset cancers; nevertheless, growing evidence suggests that ~5% of the cases could emerge as a consequence of inherited predisposition. Indeed, the 2016 revision of the World Health Organization (WHO) includes hereditary myeloid malignancies (HMMs) as a separate disease entity. Distinguishing somatic from germline variants is of vital importance, in order to establish an appropriate individualized management and counsel the patients and their relatives. Germline tissues are required to discriminate HMMs alterations, but they are not always available for testing at the time of diagnosis; therefore additional strategies should be put in place in order to identify variants potentially conferring genetic predisposition to MN. The aim of the present work is to develop a tool to identify HMMs cases from tumour-only sequencing data. To this end, we reviewed tumor-only NGS reports from 299 cases of patients with myeloid malignancies sequenced with a custom Pan Myeloid Panel (PMP) that targets 48 genes, of which 21 are described in literature as HMM-associated genes. All patients signed a written informed consent form for genetic testing, research and tissue banking provided by the Biobank of the University of Navarra (UN) and were approved by the Ethical and Scientific Committees of the UN. Variant calling files (VCF) of the 299 cases were run through an algorithm aiming at identifying those variants suggestive of being of germline nature (Figure 1). We considered as indicative of potential inherited origin, variants detected in bone marrow samples at a ~50% VAF classified as pathogenic, likely pathogenic or of unknown significance. This first filter yielded 90 "likely HMMs" patients harboring a total of 104 suspicious of being germline variants affecting at least one of the 21 HMM-associated genes. Our algorithm included three additional filters: mutational patterns suggestive of germline nature of the variants, analysis of variant allele frequency (VAF) on follow up data, and sequencing of non-myeloid tissues (Figure 1). Firstly, we found 10 "likely HMMs" patients presenting double hit mutations in one of three genes: DDX41 (n=7), CEBPA (n=2) and RUNX1 (n=1). Additionally, there was available follow up data for 8 cases included in our cohort, harboring 11 suspicious variants. From them, VAFs in 4 variants affecting ASXL1, CBL, IKZF1 and GATA2 genes, drastically changed over time, indicating that those variants were somatic, whereas the other 7 variants affecting ASXL1 (n=2), DDX41 (n=2), IKZF1 (n=1), and RUNX1 (n=1) genes suspected to be germline maintained VAF ~50% with evident change of the accompanying variants over time. Finally, we collected non-myeloid tissue samples from 8 patients, harboring a total of 9 variants in ASXL1 (n=2), DDX41 (n=1), GATA2 (n=1), IKZF1 (n=1), NF1 (n=3), and SH2B3 (n=1) genes. Sequencing data confirmed germline nature of 6 of the 9 tested variants; 3 of the 6 germline variants have previously been shown to be linked to MN (DDX41 p.Asp140Glyfs*2 in UPN1, ASXL1 p.Gly967del in UPN11, and NF1 p.Met992del in UPN19). Of note, two of these bona-fide pathogenic germline mutations had been previously highlighted as potential germline variants when applying our criteria of mutational patterns and/or follow up data indicative of germline nature, in so showing the effectiveness of our algorithm (DDX41 in UPN1, and ASXL1 in UPN11). Regarding the usefulness of the different tissues, we found that skin fibroblasts, hair follicles and CD3+ T cells helped discriminating the nature of the variants; on the contrary, DNA from hair follicles showed poor DIN values and scarce DNA concentration. Similarly, DNA isolated from buccal swab showed in general poor quality metrics, and high level of contamination with tumour DNA in one case, proving it unsuitable for discrimination of the nature of the variants in some instances. Our data supports the importance of considering variants found upon tumor-only sequencing as potentially of germline origin, and we offer a pipeline to define the nature of the variants. We hope to provide insights for genetic laboratories facing this relatively new challenge of discriminating somatic from germline variants in tumor sequencing data. Disclosures No relevant conflicts of interest to declare.

A DSTYK mutation activates ERK1/2 signaling to promote intraspinal dissemination in a case of solitary fibrous tumor/hemangiopericytoma

Intracranial solitary fibrous tumors/hemangiopericytomas (SFT/HPCs) are vascular tumors that have a high rate of local recurrence and extracranial metastases. Intradural extramedullary spinal dissemination of intracranial SFT/HPC is extremely rare. There is a paucity of data available to elucidate the molecular mechanisms of intraspinal dissemination of intracranial SFT/HPC. Herein, we presented a case of intracranial SFT/HPC with intraspinal metastasis. The resected tumor specimens were enrolled in a clinical sequencing program, including whole-exome and transcriptome sequencing. By comparing genomic sequencing data of the intracranial tumors with intraspinal metastasis, we established the somatic mutational profiles of these tumors. Clonality analysis revealed a distinct subclonal structure in the intracranial tumor and its intraspinal metastasis, which might reflect the possibility of intratumoral clonal selection and evolution during the process of tumor dissemination. Through bioinformatics analysis and Sanger sequencing validation, a DSTYK mutation (Met296Ile) was identified as a candidate driver of intraspinal metastasis in this SFT/HPC case. Further, an intracranial tumor-derived SFT/HPC cell line, HPC3, was established to explore the mechanisms of the DSTYK mutation in promoting SFT/HPC metastasis. Based on the HPC3 cell model, we found that the DSTYK mutation promoted cell migration and invasion of HPC3 cells via activation of ERK1/2 signaling, which was inhibited by the MEK/ERK inhibitor AZD6244. The DSTYK mutation was also shown to upregulate the expression of two metastasis-related molecules: MMP2 and MMP9 in HPC3 cells; however, this effect was attenuated by AZD6244 treatment. Therefore, the DSTYK mutation may activate ERK1/2/MMP2/9 signaling to promote tumor cell metastasis in SFT/HPC. In conclusion, our study revealed the potential role of DSTYK mutation in the regulation of intraspinal metastasis of SFT/HPC, which might provide new biological insights into this rare disease.

Risk factors predisposing to cardia gastric adenocarcinoma: Insights and new perspectives.

Recent decades have seen an alarming increase in the incidence of cardia gastric adenocarcinoma (CGA) while noncardia gastric adenocarcinoma (NCGA) has decreased. In 2012, 260 000 CGA cases (age‐standardised rate (ASR); 3.3/100 000) and 691 000 NCGA cases (ASR; 8.8/100 000) were reported worldwide. Compared with women, men had greater rates for both the subsites, especially for CGA. Recently, four molecular subtypes of GC have been proposed by the Cancer Genome Atlas (TCGA) and the Asian Cancer Research Group (ACRG); however, these classifications do not take into account predisposing germline variants and their possible interaction with somatic alterations in carcinogenesis. The etiology of adenocarcinoma of the cardia and the gastroesophageal junction (GEJ) is not known. It is thought that CGA is distinct from adenocarcinomas located in the esophagus or distal stomach, both epidemiologically and biologically. Moreover, CGA is often identified in the advanced stage having a poor prognosis. Therefore, understanding the risk and the role of predisposing factors in etiology of CGA can inform clinical practice and counseling for risk reduction. In this paper, we showed that GC family history, lifestyle, demographics, gastroesophageal reflux disease, Helicobacter pylori infection, and multiple genetic and epigenetic risk factors as well as several predisposing conditions may underlie susceptibility to CGA. However, several genome‐wide association studies (GWASs) should be conducted to identify novel high‐penetrance genes and pathways as well as causal germline variants predisposing to CGA. They must include different ethnic groups, especially from high‐incidence countries for CGA, because some risk loci are ancestry‐specific. In parallel, statistical methods can be developed to identify cancer predisposition genes (CPGs) from tumor sequencing data. It is also necessary to find novel long noncoding RNAs related to the risk of CGA. Taken altogether, new cancer risk prediction models, including all genetic and nongenetic factors influencing risk, should be developed to facilitate risk assessment, disease prevention, and early diagnosis and intervention of CGA in the future.

Abstract 4512: Identifying novel mechanisms of p53-mediated tumor suppression

Abstract TP53 is the most frequently mutated gene in human cancer. Classically, p53-mediated tumor suppression is thought to occur via transactivation of canonical targets that induce cell cycle arrest and apoptosis. However, this model has been challenged by studies using mouse p53 mutants. The mouse transactivation domain mutant, p5325,26, is unable to activate most canonical p53 targets and fails to induce cell cycle arrest or apoptosis in response to DNA damage. However, it retains the ability to induce cellular senescence and suppresses tumor development in mouse models. Indeed, some human p53 mutants found in Li-Fraumeni syndrome as well as cancer retain transactivation, while some p53 mutants are not found in cancer despite loss of transactivation. The goal of our study is to identify novel p53 functions that drive tumor suppression by evaluating p53 mutations where transactivation is uncoupled from tumor suppression activity. To identify human TP53 missense mutants with tumor suppressive function dissociated from the ability to activate canonical p53 pathways, we first utilized the results of a 2003 saturation mutagenesis screen performed in yeast which stratified human p53 mutants by their ability to activate 8 canonical p53 targets. We then cross-referenced these results with tumor sequencing data from the IARC, the TCGA PanCancer Atlas, and the MSK-IMPACT Sequencing Cohort. We identified panels of human p53 missense mutants which either 1) retained transactivation ability in yeast, yet had been identified in sequenced human tumors or 2) were unable to transactivate p53 targets in yeast, but had not been identified in sequenced human tumors. To study these p53 mutants compared to wt p53, we cloned their sequences into doxycycline-inducible pLVX-TetOne-Puro plasmids and puromycin-selected for cell lines stably expressing the construct in p53 null H1299 and HCT116 cells. Using qPCR and western blots we validated the expression of mutant or wt p53 and assessed for transactivation of canonical targets p21, mdm2, puma, gadd45α, and 14-3-3. The IncuCyte Live Cell Analysis system and XTT Cell Proliferation assays were used to evaluate p53-mediated growth arrest of human cancer cells following induction of each mutant. From these experiments, we selected three mutants for further study. Missense p53 mutants E224D and R290H are found in human cancers and in the germline of Li-Fraumeni syndrome patients yet retain the capacity to transactivate p53 targets and arrest tumor cell growth in vitro. In contrast, p53 mutant G262A which is not found in cancer lacks the ability to activate p53 targets and fails to arrest tumor cell growth in vitro. We are currently using soft agar assays to assess the transformation potential of the identified p53 mutants in FSF-KRAS ; p53fl/+ MEFs. To dissect tumor suppressive function we are developing transgenic mice expressing the p53 mutants. The results of this ongoing study may uncover novel p53 tumor suppressor mechanisms. Citation Format: Nathan H. Leisenring, Robert W. Floyd, David G. Kirsch. Identifying novel mechanisms of p53-mediated tumor suppression [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 4512.

Mutational landscape of gastric cancer and clinical application of genomic profiling based on target next-generation sequencing

BackgroundGastric cancer (GC) is a leading cause of cancer deaths, and an increased number of GC patients adopt to next-generation sequencing (NGS) to identify tumor genomic alterations for precision medicine.MethodsIn this study, we established a hybridization capture-based NGS panel including 612 cancer-associated genes, and collected sequencing data of tumors and matched bloods from 153 gastric cancer patients. We performed comprehensive analysis of these sequencing and clinical data.Results35 significantly mutated genes were identified such as TP53, AKAP9, DRD2, PTEN, CDH1, LRP2 et al. Among them, 29 genes were novel significantly mutated genes compared with TCGA study. TP53 is the top frequently mutated gene, and tends to mutate in male (p = 0.025) patients and patients whose tumor located in cardia (p = 0.011). High tumor mutation burden (TMB) gathered in TP53 wild-type tumors (p = 0.045). TMB was also significantly associated with DNA damage repair (DDR) genes genotype (p = 0.047), Lauren classification (p = 1.5e−5), differentiation (1.9e−7), and HER2 status (p = 0.023). 38.31% of gastric cancer patients harbored at least one actionable alteration according to OncoKB database.ConclusionsWe drew a comprehensive mutational landscape of 153 gastric tumors and demonstrated utility of target next-generation sequencing to guide clinical management.

A predictive model for survival in non-small cell lung cancer (NSCLC) based on electronic health record (EHR) and tumor sequencing data at the Department of Veterans Affairs (VA).

109 Background: Machine learning tools based on EHR data hold promise to help avoid unnecessary risks associated with lung cancer and its treatment. Additionally, molecular genetic profiling is becoming an integral tool for clinicians to individualize treatment for lung cancer. However, relatively few survival models have been built that integrate this data in individualized predictive models. Here, we combine real-world EHR and tumor sequencing data from the VA Precision Oncology Data Repository (PODR) to build accurate individualized survival predictions in newly-diagnosed NSCLC patients. Methods: We identified a cohort of 356 VA patients newly diagnosed with NSCLC for whom EHR, cancer registry, and targeted tumor sequencing data is available in PODR. We defined 41 features reflecting 15 baseline clinical and demographic characteristics from the EHR and registry, such as age, race, stage, histology, and therapy. We also defined features reflecting 206 clinically actionable somatic variants. We selected 5 important variants for inclusion in the model, as well as the total number of mutations. We trained a random forests algorithm to predict 1-year survival. Precision, recall, and area under the ROC curve (AUC) were assessed using 5-fold cross validation. Results: Mean age at diagnosis was 66 years. The majority of patients had late stage disease (15% stage I, 6% II, 15% III, 44% IV), and 59% of patients received systemic therapy. 45% died within 1 year of diagnosis, and 55% survived past 1 year. Our predictive model for 1-year survival achieves strong results. Cross-validated AUC is 0.79 (SD 0.08), precision is 0.79 (SD 0.07), recall is 0.74 (SD 0.07), suggesting that the trained model combining clinical and genomic features is effective at predicting 1-year survival. Conclusions: By integrating real-world EHR and sequencing data, we built a highly accurate predictive model of 1-year survival in NSCLC patients at the VA. Such a model, after ongoing validation in a larger cohort, offers the ability to make individualized predictions that could inform patient care to improve outcomes.

Integrative analysis identifies lincRNAs up- and downstream of neuroblastoma driver genes

Long intergenic non-coding RNAs (lincRNAs) are emerging as integral components of signaling pathways in various cancer types. In neuroblastoma, only a handful of lincRNAs are known as upstream regulators or downstream effectors of oncogenes. Here, we exploit RNA sequencing data of primary neuroblastoma tumors, neuroblast precursor cells, neuroblastoma cell lines and various cellular perturbation model systems to define the neuroblastoma lincRNome and map lincRNAs up- and downstream of neuroblastoma driver genes MYCN, ALK and PHOX2B. Each of these driver genes controls the expression of a particular subset of lincRNAs, several of which are associated with poor survival and are differentially expressed in neuroblastoma tumors compared to neuroblasts. By integrating RNA sequencing data from both primary tumor tissue and cancer cell lines, we demonstrate that several of these lincRNAs are expressed in stromal cells. Deconvolution of primary tumor gene expression data revealed a strong association between stromal cell composition and driver gene status, resulting in differential expression of these lincRNAs. We also explored lincRNAs that putatively act upstream of neuroblastoma driver genes, either as presumed modulators of driver gene activity, or as modulators of effectors regulating driver gene expression. This analysis revealed strong associations between the neuroblastoma lincRNAs MIAT and MEG3 and MYCN and PHOX2B activity or expression. Together, our results provide a comprehensive catalogue of the neuroblastoma lincRNome, highlighting lincRNAs up- and downstream of key neuroblastoma driver genes. This catalogue forms a solid basis for further functional validation of candidate neuroblastoma lincRNAs.

Landscape of Germline and Somatic Mitochondrial DNA Mutations in Pediatric Malignancies.

Little is known about the spectrum of mitochondrial DNA (mtDNA) mutations across pediatric malignancies. In this study, we analyzed matched tumor and normal whole genome sequencing data from 616 pediatric patients with hematopoietic malignancies, solid tumors, and brain tumors. We identified 391 mtDNA mutations in 284 tumors including 45 loss-of-function mutations, which clustered at four statistically significant hotspots in MT-COX3, MT-ND4, and MT-ND5, and at a mutation hotspot in MT-tRNA-MET. A skewed ratio (4.83) of nonsynonymous versus synonymous (dN/dS) mtDNA mutations with high statistical significance was identified on the basis of Monte Carlo simulations in the tumors. In comparison, opposite ratios of 0.44 and 0.93 were observed in 616 matched normal tissues and in 249 blood samples from children without cancer, respectively. mtDNA mutations varied by cancer type and mtDNA haplogroup. Collectively, these results suggest that deleterious mtDNA mutations play a role in the development and progression of pediatric cancers. SIGNIFICANCE: This pan-cancer mtDNA study establishes the landscape of germline and tumor mtDNA mutations and identifies hotspots of tumor mtDNA mutations to pinpoint key mitochondrial functions in pediatric malignancies.

Testis‐specific protein, Y‐linked 1 activates PI3K/AKT and RAS signaling pathways through suppressing IGFBP3 expression during tumor progression

The testis‐specific protein, Y‐linked 1 (TSPY1), a newly recognized cancer/testis antigen, has been suggested to accelerate tumor progression. However, the mechanisms underlying TSPY1 cancer‐related function remain limited. By mining the RNA sequencing data of lung and liver tumors from The Cancer Genome Atlas, we found frequent ectopic expression of TSPY1 in lung adenocarcinoma (LUAD) and liver hepatocellular carcinoma (LIHC), and the male‐specific protein was associated with higher mortality rate and worse overall survival in patients with LUAD and LIHC. Overexpression of TSPY1 promotes cell proliferation, invasiveness, and cycle transition and inhibits apoptosis, whereas TSPY1 knockdown has the opposite effects on these cancer cell phenotypes. Transcriptomic analysis revealed the involvement of TSPY1 in PI3K/AKT and RAS signaling pathways in both LUAD and LIHC cells, which was further confirmed by the increase in the levels of phosphorylated proteins in the PI3K‐AKT and RAS signaling pathways in TSPY1‐overexpressing cancer cells, and by the suppression on the activity of these two pathways in TSPY1‐knockdown cells. Further investigation identified that TSPY1 could directly bind to the promoter of insulin growth factor binding protein 3 (IGFBP3) to inhibit IGFBP3 expression and that downregulation of IGFBP3 increased the activity of PI3K/AKT/mTOR/BCL2 and RAS/RAF/MEK/ERK/JUN signaling in LUAD and LIHC cells. Taken together, the observations reveal a novel mechanism by which TSPY1 could contribute to the progression of LUAD and LIHC. Our finding is of importance for evaluating the potential of TSPY1 in immunotherapy of male tumor patients with TSPY1 expression.

Abstract B079: Evaluation of tools for predicting mutated tumor antigens from exome and RNA sequencing

Abstract Neoantigen vaccination is an emerging modality of cancer immunotherapy with many ongoing trials. One central question of neoantigen vaccination is the method for selecting which mutated tumor-specific antigens to include in a patient’s vaccine. Many in-silico pipelines for neoantigen selection have been published in the past few years, but no comprehensive evaluation has compared them directly on the same tumor/normal sequencing data. We evaluate several publicly available commonly used neoantigen pipelines (pVACtools [1], MuPeXI [2], TIminer [3], OpenVax [4]) on both murine and human cancer samples. Our evaluation highlights the salient differences between these pipelines and shows the divergent results they achieve.

Abstract A005: A phase I study of the safety and immunogenicity of a multipeptide personalized genomic vaccine in the adjuvant treatment of solid cancers

Abstract Introduction: Mutation-derived tumor antigens (MTAs) arise as a direct result of somatic variations, including nucleotide substitutions, insertions, and deletions that occur during carcinogenesis. These somatic variations can be characterized via genetic sequencing and used to identify MTAs. We developed a platform for a fully-personalized MTA-based vaccine in the adjuvant treatment of solid and hematologic malignanicies. Methods: This is a single-arm, open label, proof-of-concept phase I study designed to test the safety and immunogenicity of Personalized Genomic Vaccine 001 (PGV001) that targets up to 10 predicted personal tumor neoantigens. The single-center study will enroll 20 eligible subjects with histologic diagnosis of solid and hematologic malignancies. Subjects must have no measurable disease at time of first vaccine administration, and 5-year disease recurrence risk of > 30%. Toxicity will be defined by CTCAE v5.0. Blood samples will be collected at various time points for immune response monitoring. Each patient’s vaccine peptides are selected by identifying somatic mutations from comparison of tumor and normal exome sequencing data, phasing somatic variants with co-occurring germline variants using tumor RNA sequencing data, and ranking mutated peptide sequences ”Openvax pipeline.” The process for determining somatic variants hews closely to the Broad Institute’s “Best Practices” for cancer SNVs and indels. The phasing of somatic and germline variants is implemented in a custom bioinformatics tool called Isovar. Mutated protein sequences containing phased variants are ranked according to two criteria: expression of the mutant allele in tumor RNA and aggregated predicted affinity to the patient’s Class I MHCs. Both quantities are normalized and multiplied together to create single ranked ordering of the candidate mutant sequences. Results: PGV001_002 (head and neck squamous cell cancer), who has completed vaccination, received 10 doses of vaccine comprising 10 long peptides (25 amino acid length) combined with poly-ICLC (toll-like receptor-3 agonist) intradermally. Vaccine-induced blood T-cell responses were determined, at weeks 0 (before-treatment) and 27 (after-treatment), ex vivo by interferon (IFN)-g enzyme-linked immunospot (ELISPOT) assay and after in vitro expansion by intracellular cytokine staining (ICS). Overlapping 15-16-mer assays peptides (OLPs) spanning the entirety of each ILP and 9-10-mer peptides corresponding to each predicted class I epitope (Min) were pooled and used to monitor immunogenicity. Ex vivo responses to these peptide pools were undetectable at week 0 but were evident at week 27 against 2 OLPs out of 10 (20%) and in 5 Min out of 10 (50%). After in vitro expansion, neoantigen-specific CD4+ and CD8+ T-cell responses were found in 5 out of 10 pooled peptides (50%). 7 out of 10 (70%) epitopes elicited polyfunctional T-cell responses (secretion of INF-α, TNF-α, and/or IL-2) from either CD4+ or CD8+ T-cells. Conclusion: To identify which predicted epitopes within the peptides pools stimulated the T-cell responses, we deconvoluted all the pools by either ex vivo and in vitro expansion. Ex vivo IFN-α production was detected in 1 (15-mer) peptide out of 15 (6.7%) and in 4 (9-10-mer) peptides out of 22 (18.2%). After expansion with single peptides, of 22 (9-10-mer) peptides tested, CD8+ T-cells were reactive against 13 peptides (59%), while CD4+ responses were seen in response to 11 of 15 (15-16-mer) peptides tested. Both CD4+ and CD8+ T-cell responses were polyfunctional. The PGV001 vaccine in our first patient showed both safety and immunogenicity, eliciting both CD4+ and CD8+ responses to the vaccine peptides. As we are enrolling additional patients, the information learned from this clinical trial will instruct the next generation of MTA-based vaccines, future development of immunotherapeutic approaches and rational combinations. Citation Format: Ana B. Blazquez, Alex Rubinsteyn, Julia Kodysh, John P. Finnigan, Thomas Marron, Rachel L. Sabado, Marcia Meseck, Timothy J. O'Donnell, Jeffrey Hammerbacher, Michael Donovan, John Holt, Milind Mahajan, John Mandeli, Krysztof Misiukiewicz, Eric M. Genden, Brett A. Milles, Hooman Khorasani, Peter R. Dottino, Hanna Irie, Amy B. Tiersten, Elisa R. Port, Andrea S. Wolf, Hern J. Cho, Ashutosh Tewari, Samir S. Parekh, Sujit Nair, Matthew D. Galsky, William K. Oh, Sacha Gnjatic, Eric E. Schadt, Phillip A. Friedlander, Nina Bhardwaj. A phase I study of the safety and immunogenicity of a multipeptide personalized genomic vaccine in the adjuvant treatment of solid cancers [abstract]. In: Proceedings of the Fourth CRI-CIMT-EATI-AACR International Cancer Immunotherapy Conference: Translating Science into Survival; Sept 30-Oct 3, 2018; New York, NY. Philadelphia (PA): AACR; Cancer Immunol Res 2019;7(2 Suppl):Abstract nr A005.

SMuRF: portable and accurate ensemble prediction of somatic mutations.

SummarySomatic Mutation calling method using a Random Forest (SMuRF) integrates predictions and auxiliary features from multiple somatic mutation callers using a supervised machine learning approach. SMuRF is trained on community-curated matched tumor and normal whole genome sequencing data. SMuRF predicts both SNVs and indels with high accuracy in genome or exome-level sequencing data. Furthermore, the method is robust across multiple tested cancer types and predicts low allele frequency variants with high accuracy. In contrast to existing ensemble-based somatic mutation calling approaches, SMuRF works out-of-the-box and is orders of magnitudes faster.Availability and implementationThe method is implemented in R and available at https://github.com/skandlab/SMuRF. SMuRF operates as an add-on to the community-developed bcbio-nextgen somatic variant calling pipeline.Supplementary information Supplementary data are available at Bioinformatics online.

SeqClone: sequential Monte Carlo based inference of tumor subclones

BackgroundTumor samples are heterogeneous. They consist of varying cell populations or subclones and each subclone is characterized with a distinct single nucleotide variant (SNV) profile. This explains the source of genetic heterogeneity observed in tumor sequencing data. To make precise prognosis and design effective therapy for cancer, ascertaining the subclonal composition of a tumor is of great importance.ResultsIn this paper, we propose a state-space formulation of the feature allocation model. This model is interpreted as the blind deconvolution of the expected variant allele fractions (VAFs). VAFs are deconvolved into a binary matrix of genotypes and a matrix of genotype proportions in the samples. Specifically, we consider a sequential construction of the genotype matrix which we model by Indian buffet process (IBP). We describe an efficient sequential Monte Carlo (SMC) algorithm, SeqClone, that jointly estimates the genotypes of subclones and their proportions in the samples. When compared to other methods for resolving tumor heterogeneity, SeqClone provides comparable and sometimes, better estimates of model parameters. By design, SeqClone conveniently handles any number of probed SNVs in the samples. In particular, we can analyze VAFs from newly probed SNVs to improve existing estimates, an attribute not present in existing solutions.ConclusionsWe show that the SMC algorithm for deconvolving VAFs from tumor sequencing data is a robust and promising alternative for explaining the observed genetic heterogeneity in tumor samples.

Molecular characterization of an MLL1 fusion and its role in chromosomal instability.

Chromosomal rearrangements involving the mixed‐lineage leukemia (MLL1) gene are common in a unique group of acute leukemias, with more than 100 fusion partners in this malignancy alone. However, do these fusions occur or have a role in solid tumors? We performed extensive network analyses of MLL1‐fusion partners in patient datasets, revealing that multiple MLL1‐fusion partners exhibited significant interactions with the androgen‐receptor signaling pathway. Further exploration of tumor sequence data from TCGA predicts the presence of MLL1 fusions with truncated SET domain in prostate tumors. To investigate the physiological relevance of MLL1 fusions in solid tumors, we engineered a truncated version of MLL1 by fusing it with one of its known fusion partners, ZC3H13, to use as a model system. Functional characterization with cell‐based assays revealed that MLL1‐ZC3H13 fusion induced chromosomal instability, affected mitotic progression, and enhanced tumorsphere formation. The MLL1‐ZC3H13 chimera consistently increased the expression of a cancer stem cell marker (CD44); in addition, we detected potential collateral lethality between DOT1L and MLL1 fusions. Our work reveals that MLL1 fusions are likely prevalent in solid tumors and exhibit a potential pro‐tumorigenic role.

Development and Characterization of a Spontaneously Metastatic Patient-Derived Xenograft Model of Human Prostate Cancer

Here we describe the establishment and characterization of an AR+, PSMA+, ERG+, PTEN−/−, CHD1+/− patient-derived xenograft (PDX) model termed ‘C5’, which has been developed from a 60 years old patient suffering from castration-resistant prostate cancer (CRPC). The patient underwent radical prostatectomy, showed early tumor marker PSA recurrence and, one year after surgery, abiraterone resistance. Subcutaneous C5 tumors can be serially transplanted between mice and grow within ~90 days to 1.5–2 cm³ tumors in SCID Balb/c mice (take rate 100%), NOD-scid IL2Rgnull (NSG) mice (100%) and C57BL/6 pfp−/−/rag2−/− mice (66%). In contrast, no tumor growth is observed in female mice. C5 tumors can be cryopreserved and show the same growth characteristics in vivo afterwards. C5 tumor cells do not grow stably in vitro, neither under two- nor three-dimensional cell culture conditions. Upon serial transplantation, some C5 tumors spontaneously disseminated to distant sites with an observable trend towards higher metastatic cell loads in scid compared to NSG mice. Lung metastases could be verified by histology by means of anti-PSMA immunohistochemistry, exclusively demonstrating single disseminated tumor cells (DTCs) and micro-metastases. Upon surgical resection of the primary tumors, such pulmonary foci rarely grew out to multi-cellular metastatic colonies despite doubled overall survival span. In the brain and bone marrow, the metastatic cell load present at surgery even disappeared during the post-surgical period. We provide shallow whole genome sequencing and whole exome sequencing data of C5 tumors demonstrating the copy number aberration/ mutation status of this PCa model and proving genomic stability over several passages. Moreover, we analyzed genomic and transcriptomic alterations during metastatic progression achieved by serial transplantation. This study describes a novel PCa PDX model that enables future research on several aspects of metastatic PCa, particularly for the AR+ , ERG+ , PTEN−/− PCa subtype.

ALPHLARD: a Bayesian method for analyzing HLA genes from whole genome sequence data

BackgroundAlthough human leukocyte antigen (HLA) genotyping based on amplicon, whole exome sequence (WES), and RNA sequence data has been achieved in recent years, accurate genotyping from whole genome sequence (WGS) data remains a challenge due to the low depth. Furthermore, there is no method to identify the sequences of unknown HLA types not registered in HLA databases.ResultsWe developed a Bayesian model, called ALPHLARD, that collects reads potentially generated from HLA genes and accurately determines a pair of HLA types for each of HLA-A, -B, -C, -DPA1, -DPB1, -DQA1, -DQB1, and -DRB1 genes at 3rd field resolution. Furthermore, ALPHLARD can detect rare germline variants not stored in HLA databases and call somatic mutations from paired normal and tumor sequence data. We illustrate the capability of ALPHLARD using 253 WES data and 25 WGS data from Illumina platforms. By comparing the results of HLA genotyping from SBT and amplicon sequencing methods, ALPHLARD achieved 98.8% for WES data and 98.5% for WGS data at 2nd field resolution. We also detected three somatic point mutations and one case of loss of heterozygosity in the HLA genes from the WGS data.ConclusionsALPHLARD showed good performance for HLA genotyping even from low-coverage data. It also has a potential to detect rare germline variants and somatic mutations in HLA genes. It would help to fill in the current gaps in HLA reference databases and unveil the immunological significance of somatic mutations identified in HLA genes.

Daxx Functions Are p53-Independent In Vivo.

Mutations in the death domain-associated protein (DAXX) have been recently identified in a substantial proportion of human pancreatic neuroendocrine tumors (PanNETs). Remarkably, however, little is known about the physiologic role(s) of DAXX despite in vitro studies suggesting potential functions. Most prominently, and supported by tumor sequencing data, DAXX functions in concert with alpha thalassemia/mental retardation X-linked (ATRX) as a histone chaperone complex for the H3.3 variant. Studies have also identified potential roles in apoptosis, transcription, and negative regulation of the p53 tumor suppressor pathway. Herein, a mouse modeling approach was used to specifically address the latter and no significant genetic interaction between Daxx and the p53 pathway was determined. The embryonic lethal phenotype of Daxx loss is not p53-dependent. In addition, Daxx heterozygosity does not sensitize mice to a sublethal dose of ionizing radiation or alter the survival or tumor phenotype of Mdm2 transgenic mice. However, the data support a tumor suppressor role for DAXX as low-dose ionizing radiation produced a higher proportion of carcinomas in Daxx heterozygous mice than wild-type controls.Implications: While DAXX has important in vivo functions, they are independent of an inhibitory role on the p53 tumor suppressor pathway. Mol Cancer Res; 16(10); 1523-9. ©2018 AACR.

Distribution of EGFR amplification, combined chromosome 7 gain and chromosome 10 loss, and TERT promoter mutation in brain tumors and their potential for the reclassification of IDHwt astrocytoma to glioblastoma.

EGFR amplification (EGFRamp), the combination of gain of chromosome 7 and loss of chromosome 10 (7+/10-), and TERT promoter mutation (pTERTmut) are alterations frequently observed in adult IDH-wild-type (IDHwt) glioblastoma (GBM). In the absence of endothelial proliferation and/or necrosis, these alterations currently are considered to serve as a surrogate for upgrading IDHwt diffuse or anaplastic astrocytoma to GBM. Here, we set out to determine the distribution of EGFRamp, 7+/10-, and pTERTmut by analyzing high-resolution copy-number profiles and next-generation sequencing data of primary brain tumors. In addition, we addressed the question whether combinations of partial gains on chromosome 7 and partial losses on chromosome 10 exhibited a diagnostic and prognostic value similar to that of complete 7+/10-. Several such combinations proved relevant and were combined as the 7/10 signature. Our results demonstrate that EGFRamp and the 7/10 signature are closely associated with IDHwt GBM. In contrast, pTERTmut is less specific for IDHwt GBM. We conclude that, in the absence of endothelial proliferation and/or necrosis, the detection of EGFRamp is a very strong surrogate marker for the diagnosis of GBM in IDHwt diffuse astrocytic tumors. The 7/10 signature is also a strong surrogate marker. However, care should be taken to exclude pleomorphic xanthoastrocytoma. pTERTmut is less restricted to this entity and needs companion analysis by other molecular markers to serve as a surrogate for diagnosing IDHwt GBM. A combination of any two of EGFRamp, the 7/10 signature and pTERTmut, is highly specific for IDHwt GBM and the combination of all three alterations is frequent and exclusively seen in IDHwt GBM.

Detecting repeated cancer evolution from multi-region tumor sequencing data.

Recurrent successions of genomic changes, both within and between patients, reflect repeated evolutionary processes that are valuable for the anticipation of cancer progression. Multi-region sequencing allows the temporal order of some genomic changes in a tumor to be inferred, but the robust identification of repeated evolution across patients remains a challenge. We developed a machine-learning method based on transfer learning that allowed us to overcome the stochastic effects of cancer evolution and noise in data and identified hidden evolutionary patterns in cancer cohorts. When applied to multi-region sequencing datasets from lung, breast, renal, and colorectal cancer (768 samples from 178 patients), our method detected repeated evolutionary trajectories in subgroups of patients, which were reproduced in single-sample cohorts (n = 2,935). Our method provides a means of classifying patients on the basis of how their tumor evolved, with implications for the anticipation of disease progression.