Lucy F. Stead

Accurately Identifying Low‐Allelic Fraction Variants in Single Samples with Next‐Generation Sequencing: Applications in Tumor Subclone Resolution

Current methods for resolving genetically distinct subclones in tumor samples require somatic mutations to be clustered by allelic frequencies, which are determined by applying a variant calling program to next‐generation sequencing data. Such programs were developed to accurately distinguish true polymorphisms and somatic mutations from the artifactual nonreference alleles introduced during library preparation and sequencing. However, numerous variant callers exist with no clear indication of the best performer for subclonal analysis, in which the accuracy of the assigned variant frequency is as important as correctly indicating whether the variant is present or not. Furthermore, sequencing depth (the number of times that a genomic position is sequenced) affects the ability to detect low‐allelic fraction variants and accurately assign their allele frequencies. We created two synthetic sequencing datasets, and sequenced real KRAS amplicons, with variants spiked in at specific ratios, to assess which caller performs best in terms of both variant detection and assignment of allelic frequencies. We also assessed the sequencing depths required to detect low‐allelic fraction variants. We found that VarScan2 performed best overall with sequencing depths of 100×, 250×, 500×, and 1,000× required to accurately identify variants present at 10%, 5%, 2.5%, and 1%, respectively.

Intravenous delivery of oncolytic reovirus to brain tumor patients immunologically primes for subsequent checkpoint blockade

Intravenous infusion of oncolytic reovirus in patients leads to infection of brain tumors, infiltration by cytotoxic T cells, and up-regulation of PD-L1. Viruses team up with cancer immunotherapy Immune checkpoint inhibitors have shown great promise for cancer therapy, but they do not treat all cancers, and neither breast nor brain tumors are usually treatable with these drugs. However, Bourgeois-Daigneault et al. discovered a way to address this for breast cancer, and Samson et al. discovered a way to address this for brain tumors. In both cases, the authors found that oncolytic virus treatment given early, before surgical resection, alters the antitumor immune response and potentiates the effects of subsequent treatment with immune checkpoint inhibitors. Although these studies differ in the details of their methods and the immune effects induced by the oncolytic viruses, they indicate the potential of such viruses for enhancing the potential of checkpoint therapy and expanding it to new types of cancer. Immune checkpoint inhibitors, including those targeting programmed cell death protein 1 (PD-1), are reshaping cancer therapeutic strategies. Evidence suggests, however, that tumor response and patient survival are determined by tumor programmed death ligand 1 (PD-L1) expression. We hypothesized that preconditioning of the tumor immune microenvironment using targeted, virus-mediated interferon (IFN) stimulation would up-regulate tumor PD-L1 protein expression and increase cytotoxic T cell infiltration, improving the efficacy of subsequent checkpoint blockade. Oncolytic viruses (OVs) represent a promising form of cancer immunotherapy. For brain tumors, almost all studies to date have used direct intralesional injection of OV, because of the largely untested belief that intravenous administration will not deliver virus to this site. We show, in a window-of-opportunity clinical study, that intravenous infusion of oncolytic human Orthoreovirus (referred to herein as reovirus) leads to infection of tumor cells subsequently resected as part of standard clinical care, both in high-grade glioma and in brain metastases, and increases cytotoxic T cell tumor infiltration relative to patients not treated with virus. We further show that reovirus up-regulates IFN-regulated gene expression, as well as the PD-1/PD-L1 axis in tumors, via an IFN-mediated mechanism. Finally, we show that addition of PD-1 blockade to reovirus enhances systemic therapy in a preclinical glioma model. These results support the development of combined systemic immunovirotherapy strategies for the treatment of both primary and secondary tumors in the brain.

Functional and developmental expression of a zebrafish Kir1.1 (ROMK) potassium channel homologue Kcnj1

Non‐technical summary  Due to the conservation of developmental pathways and genetic material over the course of evolution, non‐mammalian ‘model organisms’ such as the zebrafish embryo are emerging as valuable tools to explore causes and potential treatments for human diseases. Ion channels are proteins that form pores and help to establish and control electrical gradients by allowing the flow of ions across biological membranes. A diverse range of key physiological mechanisms in every organ in the body depends on the activity of ion channels. In this paper, we show that a potassium‐selective channel that underlies salt reabsorption and potassium excretion in the human kidney is also expressed in zebrafish in cells that are important regulators of salt balance. Disruption of the channels expression in zebrafish leads to effects on the activity of the heart, consistent with a role for this channel in the control of potassium balance in the embryo.

KvSNP: accurately predicting the effect of genetic variants in voltage-gated potassium channels

MOTIVATION Non-synonymous single nucleotide polymorphisms (nsSNPs) in voltage-gated potassium (Kv) channels cause diseases with potentially fatal consequences in seemingly healthy individuals. Identifying disease-causing genetic variation will aid presymptomatic diagnosis and treatment of such disorders. NsSNP-effect predictors are hypothesized to perform best when developed for specific gene families. We, thus, created KvSNP: a method that assigns a disease-causing probability to Kv-channel nsSNPs. RESULTS KvSNP outperforms popular non gene-family-specific methods (SNPs&GO, SIFT and Polyphen) in predicting the disease potential of Kv-channel variants, according to all tested metrics (accuracy, Matthews correlation coefficient and area under receiver operator characteristic curve). Most significantly, it increases the separation of the median predicted disease probabilities between benign and disease-causing SNPs by 26% on the next-best competitor. KvSNP has ranked 172 uncharacterized Kv-channel nsSNPs by disease-causing probability. AVAILABILITY AND IMPLEMENTATION KvSNP, a WEKA implementation is available at www.bioinformatics.leeds.ac.uk/KvDB/KvSNP.html. CONTACT d.r.westhead@leeds.ac.uk SUPPLEMENTARY INFORMATION Supplementary data are available at Bioinformatics online.

Genomic Subtypes of Non-invasive Bladder Cancer with Distinct Metabolic Profile and Female Gender Bias in KDM6A Mutation Frequency.

Bladder cancer incurs a higher lifetime treatment cost than other cancers due to frequent recurrence of non-invasive disease. Improved prognostic biomarkers and localized therapy are needed for this large patient group. We defined two major genomic subtypes of primary stage Ta tumors. One of these was characterized by loss of 9q including TSC1, increased KI67 labeling index, upregulated glycolysis, DNA repair, mTORC1 signaling, features of the unfolded protein response, and altered cholesterol homeostasis. Comparison with muscle-invasive bladder cancer mutation profiles revealed lower overall mutation rates and more frequent mutations in RHOB and chromatin modifier genes. More mutations in the histone lysine demethylase KDM6A were present in non-invasive tumors from females than males.

The clonal relationships between pre‐cancer and cancer revealed by ultra‐deep sequencing

The study of the relationships between pre‐cancer and cancer and identification of early driver mutations is becoming increasingly important as the value of molecular markers of early disease and personalised drug targets is recognized, especially now the extent of clonal heterogeneity in fully invasive disease is being realized. It has been assumed that pre‐cancerous lesions exhibit a fairly passive progression to invasive disease; the degree to which they, too, are heterogeneous is unknown. We performed ultra‐deep sequencing of thousands of selected mutations, together with copy number analysis, from multiple, matched pre‐invasive lesions, primary tumours and metastases from five patients with oral cancer, some with multiple primary tumours presenting either synchronously or metachronously, totalling 75 samples. This allowed the clonal relationships between the samples to be observed for each patient. We expose for the first time the unexpected variety and complexity of the relationships between this group of oral dysplasias and their associated carcinomas and, ultimately, the diversity of processes by which tumours are initiated, spread and metastasize. Instead of a series of genomic precursors of their adjacent invasive disease, we have shown dysplasia to be a distinct dynamic entity, refuting the belief that pre‐cancer and invasive tumours with a close spatial relationship always have linearly related genomes. We show that oral pre‐cancer exhibits considerable subclonal heterogeneity in its own right, that mutational changes in pre‐cancer do not predict the onset of invasion, and that the genomic pathway to invasion is neither unified nor predictable. Sequence data from this study have been deposited in the European Nucleotide Archive, Accession No. PRJEB6588. Copyright

Down-Regulation of miR-92 in Breast Epithelial Cells and in Normal but Not Tumour Fibroblasts Contributes to Breast Carcinogenesis.

Background MicroRNA (miR) expression is commonly dysregulated in many cancers, including breast. MiR–92 is one of six miRs encoded by the miR-17-92 cluster, one of the best-characterised oncogenic miR clusters. We examined expression of miR–92 in the breast epithelium and stroma during breast cancer progression. We also investigated the role of miR–92 in fibroblasts in vitro and showed that down-regulation in normal fibroblasts enhances the invasion of breast cancer epithelial cells. Methodology/Principal Findings We used laser microdissection (LMD) to isolate epithelial cells from matched normal, DCIS and invasive tissue from 9 breast cancer patients and analysed miR–92 expression by qRT-PCR. Expression of ERβ1, a direct miR–92 target, was concurrently analysed for each case by immunohistochemistry. LMD was also used to isolate matched normal (NFs) and cancer-associated fibroblasts (CAFs) from 14 further cases. Effects of miR–92 inhibition in fibroblasts on epithelial cell invasion in vitro was examined using a Matrigel™ assay. miR–92 levels decreased in microdissected epithelial cells during breast cancer progression with highest levels in normal breast epithelium, decreasing in DCIS (p<0.01) and being lowest in invasive breast tissue (p<0.01). This was accompanied by a shift in cell localisation of ERβ1 from nuclear expression in normal breast epithelium to increased cytoplasmic expression during progression to DCIS (p = 0.0078) and invasive breast cancer (p = 0.031). ERβ1 immunoreactivity was also seen in stromal fibroblasts in tissues. Where miR–92 expression was low in microdissected NFs this increased in matched CAFs; a trend also seen in cultured primary fibroblasts. Down-regulation of miR–92 levels in NFs but not CAFs enhanced invasion of both MCF–7 and MDA-MB–231 breast cancer epithelial cells. Conclusions miR–92 is gradually lost in breast epithelial cells during cancer progression correlating with a shift in ERβ1 immunoreactivity from nuclei to the cytoplasm. Our data support a functional role in fibroblasts where modification of miR–92 expression can influence the invasive capacity of breast cancer epithelial cells. However in silico analysis suggests that ERβ1 may not be the most important miR–92 target in breast cancer.

Elucidating drivers of oral epithelial dysplasia formation and malignant transformation to cancer using RNAseq

Oral squamous cell carcinoma (OSCC) is a prevalent cancer with poor prognosis. Most OSCC progresses via a non-malignant stage called dysplasia. Effective treatment of dysplasia prior to potential malignant transformation is an unmet clinical need. To identify markers of early disease, we performed RNA sequencing of 19 matched HPV negative patient trios: normal oral mucosa, dysplasia and associated OSCC. We performed differential gene expression, principal component and correlated gene network analysis using these data. We found differences in the immune cell signatures present at different disease stages and were able to distinguish early events in pathogenesis, such as upregulation of many HOX genes, from later events, such as down-regulation of adherens junctions. We herein highlight novel coding and non-coding candidates for involvement in oral dysplasia development and malignant transformation, and speculate on how our findings may guide further translational research into the treatment of oral dysplasia.

Exploring the surfaceome of Ewing sarcoma identifies a new and unique therapeutic target.

Significance By investigating cell surface proteins of Ewing sarcoma we have identified an antigen that is uniquely expressed on these tumor cells compared with mesenchymal stem cells. This protein acts as a target for antibody drug conjugates that are internalized and can kill these tumor cells, presaging translating to clinical use in treating Ewing sarcoma, especially metastatic disease. The cell surface proteome of tumors mediates the interface between the transformed cells and the general microenvironment, including interactions with stromal cells in the tumor niche and immune cells such as T cells. In addition, the cell surface proteome of individual cancers defines biomarkers for that tumor type and potential proteins that can be the target of antibody-mediated therapy. We have used next-generation deep RNA sequencing (RNA-seq) coupled to an in-house database of genes encoding cell surface proteins (herein referred to as the surfaceome) as a tool to define a cell surface proteome of Ewing sarcoma compared with progenitor mesenchymal stem cells. This subtractive RNA-seq analysis revealed a specific surfaceome of Ewing and showed unexpectedly that the leucine-rich repeat and Ig domain protein 1 (LINGO1) is expressed in over 90% of Ewing sarcoma tumors, but not expressed in any other somatic tissue apart from the brain. We found that the LINGO1 protein acts as a gateway protein internalizing into the tumor cells when engaged by antibody and can carry antibody conjugated with drugs to kill Ewing sarcoma cells. Therefore, LINGO1 is a new, unique, and specific biomarker and drug target for the treatment of Ewing sarcoma.

Mutation detection by clonal sequencing of PCR amplicons and grouped read typing is applicable to clinical diagnostics.

We describe a sensitive technique for mutation detection using clonal sequencing. We analyzed DNA extracted from 13 cancer cell lines and 35 tumor samples and applied a novel approach to identify disease‐associated somatic mutations. By matching reads against an index of known variants, noise can be dramatically reduced, enabling the detection and quantification of those variants, even when they are present at less than 1% of the total sequenced population; this is comparable to, or better than, current diagnostic methods. Following the identification or exclusion of known variants, unmatched reads are grouped for BLAST searching to identify novel variants or contaminants. Known variants, novel variants, and contaminants were readily identified in tumor tissue using this approach. Our approach also enables an estimation of the per‐base sequencing error rate, providing a confidence threshold for interpretation of the results in the clinic. This novel approach has immediate applicability to clinical testing for disease‐associated genetic variants.