Phenotype prediction from single-cell RNA-seq data using attention-based neural networks.

MOTIVATION: A patient's disease phenotype can be driven and determined by specific groups of cells whose marker genes are either unknown or can only be detected at late-stage using conventional bulk assays such as RNA-Seq technology. Recent advances in single-cell RNA sequencing (scRNA-seq) enable gene expression profiling in cell-level resolution, and therefore have the potential to identify those cells driving the disease phenotype even while the number of these cells is small. However, most existing methods rely heavily on accurate cell type detection, and the number of available annotated samples is usually too small for training deep learning predictive models. RESULTS: Here, we propose the method ScRAT for phenotype prediction using scRNA-seq data. To train ScRAT with a limited number of samples of different phenotypes, such as coronavirus disease (COVID) and non-COVID, ScRAT first applies a mixup module to increase the number of training samples. A multi-head attention mechanism is employed to learn the most informative cells for each phenotype without relying on a given cell type annotation. Using three public COVID datasets, we show that ScRAT outperforms other phenotype prediction methods. The performance edge of ScRAT over its competitors increases as the number of training samples decreases, indicating the efficacy of our sample mixup. Critical cell types detected based on high-attention cells also support novel findings in the original papers and the recent literature. This suggests that ScRAT overcomes the challenge of missing marker genes and limited sample number with great potential revealing novel molecular mechanisms and/or therapies. AVAILABILITY AND IMPLEMENTATION: The code of our proposed method ScRAT is published at https://github.com/yuzhenmao/ScRAT.

Establishment and validation of preclinical models of SMARCA4-inactivated and ARID1A/ARID1B co-inactivated dedifferentiated endometrial carcinoma.

OBJECTIVE: Dedifferentiated endometrial cancer (DDEC) is an uncommon and clinically highly aggressive subtype of endometrial cancer characterized by genomic inactivation of SWItch/Sucrose Non-Fermentable (SWI/SNF) complex protein. It responds poorly to conventional systemic treatment and its rapidly progressive clinical course limits the therapeutic windows to trial additional lines of therapies. This underscores a pressing need for biologically accurate preclinical tumor models to accelerate therapeutic development. METHODS: DDEC tumor from surgical samples were implanted into immunocompromised mice for patient-derived xenograft (PDX) and cell line development. The histologic, immunophenotypic, genetic and epigenetic features of the patient tumors and the established PDX models were characterized. The SMARCA4-deficienct DDEC model was evaluated for its sensitivity toward a KDM6A/B inhibitor (GSK-J4) that was previously reported to be effective therapy for other SMARCA4-deficient cancer types. RESULTS: All three DDEC models exhibited rapid growth in vitro and in vivo, with two PDX models showing spontaneous development of metastases in vivo. The PDX tumors maintained the same undifferentiated histology and immunophenotype, and exhibited identical genomic and methylation profiles as seen in the respective parental tumors, including a mismatch repair (MMR)-deficient DDEC with genomic inactivation of SMARCA4, and two MMR-deficient DDECs with genomic inactivation of both ARID1A and ARID1B. Although the SMARCA4-deficient cell line showed low micromolecular sensitivity to GSK-J4, no significant tumor growth inhibition was observed in the corresponding PDX model. CONCLUSIONS: These established patient tumor-derived models accurately depict DDEC and represent valuable preclinical tools to gain therapeutic insights into this aggressive tumor type.

Alignment-free clustering of UMI tagged DNA molecules.

MOTIVATION: Next-Generation Sequencing has led to the availability of massive genomic datasets whose processing raises many challenges, including the handling of sequencing errors. This is especially pertinent in cancer genomics, e.g. for detecting low allele frequency variations from circulating tumor DNA. Barcode tagging of DNA molecules with unique molecular identifiers (UMI) attempts to mitigate sequencing errors; UMI tagged molecules are polymerase chain reaction (PCR) amplified, and the PCR copies of UMI tagged molecules are sequenced independently. However, the PCR and sequencing steps can generate errors in the sequenced reads that can be located in the barcode and/or the DNA sequence. Analyzing UMI tagged sequencing data requires an initial clustering step, with the aim of grouping reads sequenced from PCR duplicates of the same UMI tagged molecule into a single cluster, and the size of the current datasets requires this clustering process to be resource-efficient. RESULTS: We introduce Calib, a computational tool that clusters paired-end reads from UMI tagged sequencing experiments generated by substitution-error-dominant sequencing platforms such as Illumina. Calib clusters are defined as connected components of a graph whose edges are defined in terms of both barcode similarity and read sequence similarity. The graph is constructed efficiently using locality sensitive hashing and MinHashing techniques. Calib's default clustering parameters are optimized empirically, for different UMI and read lengths, using a simulation module that is packaged with Calib. Compared to other tools, Calib has the best accuracy on simulated data, while maintaining reasonable runtime and memory footprint. On a real dataset, Calib runs with far less resources than alignment-based methods, and its clusters reduce the number of tentative false positive in downstream variation calling. AVAILABILITY AND IMPLEMENTATION: Calib is implemented in C++ and its simulation module is implemented in Python. Calib is available at https://github.com/vpc-ccg/calib. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Metagenomic and metatranscriptomic analysis of human prostate microbiota from patients with prostate cancer.

BACKGROUND: Prostate cancer (PCa) is the most common malignant neoplasm among men in many countries. Since most precancerous and cancerous tissues show signs of inflammation, chronic bacterial prostatitis has been hypothesized to be a possible etiology. However, establishing a causal relationship between microbial inflammation and PCa requires a comprehensive analysis of the prostate microbiome. The aim of this study was to characterize the microbiome in prostate tissue of PCa patients and investigate its association with tumour clinical characteristics as well as host expression profiles. RESULTS: The metagenome and metatranscriptome of tumour and the adjacent benign tissues were assessed in 65 Chinese radical prostatectomy specimens. Escherichia, Propionibacterium, Acinetobacter and Pseudomonas were abundant in both metagenome and metatranscriptome, thus constituting the core of the prostate microbiome. The biodiversity of the microbiomes could not be differentiated between the matched tumour/benign specimens or between the tumour specimens of low and high Gleason Scores. The expression profile of ten Pseudomonas genes was strongly correlated with that of eight host small RNA genes; three of the RNA genes may negatively associate with metastasis. Few viruses could be identified from the prostate microbiomes. CONCLUSIONS: This is the first study of the human prostate microbiome employing an integrated metagenomics and metatranscriptomics approach. In this Chinese cohort, both metagenome and metatranscriptome analyses showed a non-sterile microenvironment in the prostate of PCa patients, but we did not find links between the microbiome and local progression of PCa. However, the correlated expression of Pseudomonas genes and human small RNA genes may provide tantalizing preliminary evidence that Pseudomonas infection may impede metastasis.

Markers of MEK inhibitor resistance in low-grade serous ovarian cancer: EGFR is a potential therapeutic target.

BACKGROUND: Although low-grade serous ovarian cancer (LGSC) is rare, case-fatality rates are high as most patients present with advanced disease and current cytotoxic therapies are not overly effective. Recognizing that these cancers may be driven by MAPK pathway activation, MEK inhibitors (MEKi) are being tested in clinical trials. LGSC respond to MEKi only in a subgroup of patients, so predictive biomarkers and better therapies will be needed. METHODS: We evaluated a number of patient-derived LGSC cell lines, previously classified according to their MEKi sensitivity. Two cell lines were genomically compared against their matching tumors samples. MEKi-sensitive and MEKi-resistant lines were compared using whole exome sequencing and reverse phase protein array. Two treatment combinations targeting MEKi resistance markers were also evaluated using cell proliferation, cell viability, cell signaling, and drug synergism assays. RESULTS: Low-grade serous ovarian cancer cell lines recapitulated the genomic aberrations from their matching tumor samples. We identified three potential predictive biomarkers that distinguish MEKi sensitive and resistant lines: KRAS mutation status, and EGFR and PKC-alpha protein expression. The biomarkers were validated in three newly developed LGSC cell lines. Sub-lethal combination of MEK and EGFR inhibition showed drug synergy and caused complete cell death in two of four MEKi-resistant cell lines tested. CONCLUSIONS: KRAS mutations and the protein expression of EGFR and PKC-alpha should be evaluated as predictive biomarkers in patients with LGSC treated with MEKi. Combination therapy using a MEKi with EGFR inhibition may represent a promising new therapy for patients with MEKi-resistant LGSC.

SiNVICT: ultra-sensitive detection of single nucleotide variants and indels in circulating tumour DNA.

MOTIVATION: Successful development and application of precision oncology approaches require robust elucidation of the genomic landscape of a patient's cancer and, ideally, the ability to monitor therapy-induced genomic changes in the tumour in an inexpensive and minimally invasive manner. Thanks to recent advances in sequencing technologies, 'liquid biopsy', the sampling of patient's bodily fluids such as blood and urine, is considered as one of the most promising approaches to achieve this goal. In many cancer patients, and especially those with advanced metastatic disease, deep sequencing of circulating cell free DNA (cfDNA) obtained from patient's blood yields a mixture of reads originating from the normal DNA and from multiple tumour subclones-called circulating tumour DNA or ctDNA. The ctDNA/cfDNA ratio as well as the proportion of ctDNA originating from specific tumour subclones depend on multiple factors, making comprehensive detection of mutations difficult, especially at early stages of cancer. Furthermore, sensitive and accurate detection of single nucleotide variants (SNVs) and indels from cfDNA is constrained by several factors such as the sequencing errors and PCR artifacts, and mapping errors related to repeat regions within the genome. In this article, we introduce SiNVICT, a computational method that increases the sensitivity and specificity of SNV and indel detection at very low variant allele frequencies. SiNVICT has the capability to handle multiple sequencing platforms with different error properties; it minimizes false positives resulting from mapping errors and other technology specific artifacts including strand bias and low base quality at read ends. SiNVICT also has the capability to perform time-series analysis, where samples from a patient sequenced at multiple time points are jointly examined to report locations of interest where there is a possibility that certain clones were wiped out by some treatment while some subclones gained selective advantage. RESULTS: We tested SiNVICT on simulated data as well as prostate cancer cell lines and cfDNA obtained from castration-resistant prostate cancer patients. On both simulated and biological data, SiNVICT was able to detect SNVs and indels with variant allele percentages as low as 0.5%. The lowest amounts of total DNA used for the biological data where SNVs and indels could be detected with very high sensitivity were 2.5 ng on the Ion Torrent platform and 10 ng on Illumina. With increased sequencing and mapping accuracy, SiNVICT might be utilized in clinical settings, making it possible to track the progress of point mutations and indels that are associated with resistance to cancer therapies and provide patients personalized treatment. We also compared SiNVICT with other popular SNV callers such as MuTect, VarScan2 and Freebayes. Our results show that SiNVICT performs better than these tools in most cases and allows further data exploration such as time-series analysis on cfDNA sequencing data. AVAILABILITY AND IMPLEMENTATION: SiNVICT is available at: https://sfu-compbio.github.io/sinvictSupplementary information: Supplementary data are available at Bioinformatics online. CONTACT: cenk@sfu.ca.

Integrated genome and transcriptome sequencing identifies a novel form of hybrid and aggressive prostate cancer.

Next-generation sequencing is making sequence-based molecular pathology and personalized oncology viable. We selected an individual initially diagnosed with conventional but aggressive prostate adenocarcinoma and sequenced the genome and transcriptome from primary and metastatic tissues collected prior to hormone therapy. The histology-pathology and copy number profiles were remarkably homogeneous, yet it was possible to propose the quadrant of the prostate tumour that likely seeded the metastatic diaspora. Despite a homogeneous cell type, our transcriptome analysis revealed signatures of both luminal and neuroendocrine cell types. Remarkably, the repertoire of expressed but apparently private gene fusions, including C15orf21:MYC, recapitulated this biology. We hypothesize that the amplification and over-expression of the stem cell gene MSI2 may have contributed to the stable hybrid cellular identity. This hybrid luminal-neuroendocrine tumour appears to represent a novel and highly aggressive case of prostate cancer with unique biological features and, conceivably, a propensity for rapid progression to castrate-resistance. Overall, this work highlights the importance of integrated analyses of genome, exome and transcriptome sequences for basic tumour biology, sequence-based molecular pathology and personalized oncology.

From sequence to molecular pathology, and a mechanism driving the neuroendocrine phenotype in prostate cancer.

The current paradigm of cancer care relies on predictive nomograms which integrate detailed histopathology with clinical data. However, when predictions fail, the consequences for patients are often catastrophic, especially in prostate cancer where nomograms influence the decision to therapeutically intervene. We hypothesized that the high dimensional data afforded by massively parallel sequencing (MPS) is not only capable of providing biological insights, but may aid molecular pathology of prostate tumours. We assembled a cohort of six patients with high-risk disease, and performed deep RNA and shallow DNA sequencing in primary tumours and matched metastases where available. Our analysis identified copy number abnormalities, accurately profiled gene expression levels, and detected both differential splicing and expressed fusion genes. We revealed occult and potentially dormant metastases, unambiguously supporting the patients' clinical history, and implicated the REST transcriptional complex in the development of neuroendocrine prostate cancer, validating this finding in a large independent cohort. We massively expand on the number of novel fusion genes described in prostate cancer; provide fresh evidence for the growing link between fusion gene aetiology and gene expression profiles; and show the utility of fusion genes for molecular pathology. Finally, we identified chromothripsis in a patient with chronic prostatitis. Our results provide a strong foundation for further development of MPS-based molecular pathology.

Integrated genomic and epigenomic analyses pinpoint biallelic gene inactivation in tumors.

Aberrant methylation of CpG islands and genomic deletion are two predominant mechanisms of gene inactivation in tumorigenesis, but the extent to which they interact is largely unknown. The lack of an integrated approach to study these mechanisms has limited the understanding of tumor genomes and cancer genes. Restriction landmark genomic scanning (RLGS; ref. 1) is useful for global analysis of aberrant methylation of CpG islands, but has not been amenable to alignment with deletion maps because the identity of most RLGS fragments is unknown. Here, we determined the nucleotide sequence and exact chromosomal position of RLGS fragments throughout the genome using the whole chromosome of origin of the fragments and in silico restriction digestion of the human genome sequence. To study the interaction of these gene-inactivation mechanisms in primary brain tumors, we integrated RLGS-based methylation analysis with high-resolution deletion maps from microarray-based comparative genomic hybridization (array CGH; ref. 3). Certain subsets of gene-associated CpG islands were preferentially affected by convergent methylation and deletion, including genes that exhibit tumor-suppressor activity, such as CISH1 (encoding SOCS1; ref. 4), as well as genes such as COE3 that have been missed by traditional non-integrated approaches. Our results show that most aberrant methylation events are focal and independent of deletions, and the rare convergence of these mechanisms can pinpoint biallelic gene inactivation without the use of positional cloning.

Poly-gene fusion transcripts and chromothripsis in prostate cancer.

Complex genome rearrangements are frequently observed in cancer but their impact on tumor molecular biology is largely unknown. Recent studies have identified a new phenomenon involving the simultaneous generation of tens to hundreds of genomic rearrangements, called chromothripsis. To understand the molecular consequences of these events, we sequenced the genomes and transcriptomes of two prostate tumors exhibiting evidence of chromothripsis. We identified several complex fusion transcripts, each containing sequence from three different genes, originating from different parts of the genome. One such poly-gene fusion transcript appeared to be expressed from a chain of small genomic fragments. Furthermore, we detected poly-gene fusion transcripts in the prostate cancer cell line LNCaP, suggesting they may represent a common phenomenon. Finally in one tumor with chromothripsis, we identified multiple mutations in the p53 signaling pathway, expanding on recent work associating aberrant DNA damage response mechanisms with chromothripsis. Overall, our data show that chromothripsis can manifest as massively rearranged transcriptomes. The implication that multigenic changes can give rise to poly-gene fusion transcripts is potentially of great significance to cancer genetics.

Influence of ADT on B7-H3 expression during CRPC progression from hormone-naïve prostate cancer.

Androgen deprivation therapy (ADT) is the standard care for advanced prostate cancer (PCa) patients. Unfortunately, although tumors respond well initially, they enter dormancy and eventually progress to fatal/incurable castration-resistant prostate cancer (CRPC). B7-H3 is a promising new target for PCa immunotherapy. CD276 (B7-H3) gene has a presumptive androgen receptor (AR) binding site, suggesting potential AR regulation. However, the relationship between B7-H3 and AR is controversial. Meanwhile, the expression pattern of B7-H3 following ADT and during CRPC progression is largely unknown, but critically important for identifying patients and determining the optimal timing of B7-H3 targeting immunotherapy. In this study, we performed a longitudinal study using our unique PCa patient-derived xenograft (PDX) models and assessed B7-H3 expression during post-ADT disease progression. We further validated our findings at the clinical level in PCa patient samples. We found that B7-H3 expression was negatively regulated by AR during the early phase of ADT treatment, but positively associated with PCa proliferation during the remainder of disease progression. Our findings suggest its use as a biomarker for diagnosis, prognosis, and ADT treatment response, and the potential of combining ADT and B7-H3 targeting immunotherapy for hormone-naïve PCa treatment to prevent fatal CRPC relapse.

Comrad: detection of expressed rearrangements by integrated analysis of RNA-Seq and low coverage genome sequence data.

MOTIVATION: Comrad is a novel algorithmic framework for the integrated analysis of RNA-Seq and whole genome shotgun sequencing (WGSS) data for the purposes of discovering genomic rearrangements and aberrant transcripts. The Comrad framework leverages the advantages of both RNA-Seq and WGSS data, providing accurate classification of rearrangements as expressed or not expressed and accurate classification of the genomic or non-genomic origin of aberrant transcripts. A major benefit of Comrad is its ability to accurately identify aberrant transcripts and associated rearrangements using low coverage genome data. As a result, a Comrad analysis can be performed at a cost comparable to that of two RNA-Seq experiments, significantly lower than an analysis requiring high coverage genome data. RESULTS: We have applied Comrad to the discovery of gene fusions and read-throughs in prostate cancer cell line C4-2, a derivative of the LNCaP cell line with androgen-independent characteristics. As a proof of concept, we have rediscovered in the C4-2 data 4 of the 6 fusions previously identified in LNCaP. We also identified six novel fusion transcripts and associated genomic breakpoints, and verified their existence in LNCaP, suggesting that Comrad may be more sensitive than previous methods that have been applied to fusion discovery in LNCaP. We show that many of the gene fusions discovered using Comrad would be difficult to identify using currently available techniques. AVAILABILITY: A C++ and Perl implementation of the method demonstrated in this article is available at http://compbio.cs.sfu.ca/.

Multiomics Characterization of Low-Grade Serous Ovarian Carcinoma Identifies Potential Biomarkers of MEK Inhibitor Sensitivity and Therapeutic Vulnerability.

Low-grade serous ovarian carcinoma (LGSOC) is a rare tumor subtype with high case fatality rates in patients with metastatic disease. There is a pressing need to develop effective treatments using newly available preclinical models for therapeutic discovery and drug evaluation. Here, we use multiomics integration of whole-exome sequencing, RNA sequencing, and mass spectrometry-based proteomics on 14 LGSOC cell lines to elucidate novel biomarkers and therapeutic vulnerabilities. Comparison of LGSOC cell line data with LGSOC tumor data enabled predictive biomarker identification of MEK inhibitor (MEKi) efficacy, with KRAS mutations found exclusively in MEKi-sensitive cell lines and NRAS mutations found mostly in MEKi-resistant cell lines. Distinct patterns of Catalogue of Somatic Mutations in Cancer mutational signatures were identified in MEKi-sensitive and MEKi-resistant cell lines. Deletions of CDKN2A/B and MTAP genes were more frequent in cell lines than tumor samples and possibly represent key driver events in the absence of KRAS/NRAS/BRAF mutations. These LGSOC cell lines were representative models of the molecular aberrations found in LGSOC tumors. For prediction of in vitro MEKi efficacy, proteomic data provided better discrimination than gene expression data. Condensin, minichromosome maintenance, and replication factor C protein complexes were identified as potential treatment targets in MEKi-resistant cell lines. This study suggests that CDKN2A/B or MTAP deficiency may be exploited using synthetically lethal treatment strategies, highlighting the importance of using proteomic data as a tool for molecular drug prediction. Multiomics approaches are crucial to improving our understanding of the molecular underpinnings of LGSOC and applying this information to develop new therapies. SIGNIFICANCE: These findings highlight the utility of global multiomics to characterize LGSOC cell lines as research models, to determine biomarkers of MEKi resistance, and to identify potential novel therapeutic targets.

Well-Differentiated Papillary Mesothelioma of the Peritoneum Is Genetically Distinct from Malignant Mesothelioma.

Well-differentiated papillary mesothelioma (WDPM) is an uncommon mesothelial proliferation that is most commonly encountered as an incidental finding in the peritoneal cavity. There is controversy in the literature about whether WDPM is a neoplasm or a reactive process and, if neoplastic, whether it is a variant or precursor of epithelial malignant mesothelioma or is a different entity. Using whole exome sequencing of five WDPMs of the peritoneum, we have identified distinct mutations in EHD1, ATM, FBXO10, SH2D2A, CDH5, MAGED1, and TP73 shared by WDPM cases but not reported in malignant mesotheliomas. Furthermore, we show that WDPM is strongly enriched with C > A transversion substitution mutations, a pattern that is also not found in malignant mesotheliomas. The WDPMs lacked the alterations involving BAP1, SETD2, NF2, CDKN2A/B, LASTS1/2, PBRM1, and SMARCC1 that are frequently found in malignant mesotheliomas. We conclude that WDPMs are neoplasms that are genetically distinct from malignant mesotheliomas and, based on observed mutations, do not appear to be precursors of malignant mesotheliomas.

Evaluation of paired-end sequencing strategies for detection of genome rearrangements in cancer.

Paired-end sequencing is emerging as a key technique for assessing genome rearrangements and structural variation on a genome-wide scale. This technique is particularly useful for detecting copy-neutral rearrangements, such as inversions and translocations, which are common in cancer and can produce novel fusion genes. We address the question of how much sequencing is required to detect rearrangement breakpoints and to localize them precisely using both theoretical models and simulation. We derive a formula for the probability that a fusion gene exists in a cancer genome given a collection of paired-end sequences from this genome. We use this formula to compute fusion gene probabilities in several breast cancer samples, and we find that we are able to accurately predict fusion genes in these samples with a relatively small number of fragments of large size. We further demonstrate how the ability to detect fusion genes depends on the distribution of gene lengths, and we evaluate how different parameters of a sequencing strategy impact breakpoint detection, breakpoint localization, and fusion gene detection, even in the presence of errors that suggest false rearrangements. These results will be useful in calibrating future cancer sequencing efforts, particularly large-scale studies of many cancer genomes that are enabled by next-generation sequencing technologies.

BAP1 haploinsufficiency predicts a distinct immunogenic class of malignant peritoneal mesothelioma.

BACKGROUND: Malignant peritoneal mesothelioma (PeM) is a rare and fatal cancer that originates from the peritoneal lining of the abdomen. Standard treatment of PeM is limited to cytoreductive surgery and/or chemotherapy, and no effective targeted therapies for PeM exist. Some immune checkpoint inhibitor studies of mesothelioma have found positivity to be associated with a worse prognosis. METHODS: To search for novel therapeutic targets for PeM, we performed a comprehensive integrative multi-omics analysis of the genome, transcriptome, and proteome of 19 treatment-naïve PeM, and in particular, we examined BAP1 mutation and copy number status and its relationship to immune checkpoint inhibitor activation. RESULTS: We found that PeM could be divided into tumors with an inflammatory tumor microenvironment and those without and that this distinction correlated with haploinsufficiency of BAP1. To further investigate the role of BAP1, we used our recently developed cancer driver gene prioritization algorithm, HIT'nDRIVE, and observed that PeM with BAP1 haploinsufficiency form a distinct molecular subtype characterized by distinct gene expression patterns of chromatin remodeling, DNA repair pathways, and immune checkpoint receptor activation. We demonstrate that this subtype is correlated with an inflammatory tumor microenvironment and thus is a candidate for immune checkpoint blockade therapies. CONCLUSIONS: Our findings reveal BAP1 to be a potential, easily trackable prognostic and predictive biomarker for PeM immunotherapy that refines PeM disease classification. BAP1 stratification may improve drug response rates in ongoing phases I and II clinical trials exploring the use of immune checkpoint blockade therapies in PeM in which BAP1 status is not considered. This integrated molecular characterization provides a comprehensive foundation for improved management of a subset of PeM patients.

Stromal Gene Expression is Predictive for Metastatic Primary Prostate Cancer.

BACKGROUND: Clinical grading systems using clinical features alongside nomograms lack precision in guiding treatment decisions in prostate cancer (PCa). There is a critical need for identification of biomarkers that can more accurately stratify patients with primary PCa. OBJECTIVE: To identify a robust prognostic signature to better distinguish indolent from aggressive prostate cancer (PCa). DESIGN, SETTING, AND PARTICIPANTS: To develop the signature, whole-genome and whole-transcriptome sequencing was conducted on five PCa patient-derived xenograft (PDX) models collected from independent foci of a single primary tumor and exhibiting variable metastatic phenotypes. Multiple independent clinical cohorts including an intermediate-risk cohort were used to validate the biomarkers. OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS: The outcome measurement defining aggressive PCa was metastasis following radical prostatectomy. A generalized linear model with lasso regularization was used to build a 93-gene stroma-derived metastasis signature (SDMS). The SDMS association with metastasis was assessed using a Wilcoxon rank-sum test. Performance was evaluated using the area under the curve (AUC) for the receiver operating characteristic, and Kaplan-Meier curves. Univariable and multivariable regression models were used to compare the SDMS alongside clinicopathological variables and reported signatures. AUC was assessed to determine if SDMS is additive or synergistic to previously reported signatures. RESULTS AND LIMITATIONS: A close association between stromal gene expression and metastatic phenotype was observed. Accordingly, the SDMS was modeled and validated in multiple independent clinical cohorts. Patients with higher SDMS scores were found to have worse prognosis. Furthermore, SDMS was an independent prognostic factor, can stratify risk in intermediate-risk PCa, and can improve the performance of other previously reported signatures. CONCLUSIONS: Profiling of stromal gene expression led to development of an SDMS that was validated as independently prognostic for the metastatic potential of prostate tumors. PATIENT SUMMARY: Our stroma-derived metastasis signature can predict the metastatic potential of early stage disease and will strengthen decisions regarding selection of active surveillance versus surgery and/or radiation therapy for prostate cancer patients. Furthermore, profiling of stroma cells should be more consistent than profiling of diverse cellular populations of heterogeneous tumors.

The long noncoding RNA landscape of neuroendocrine prostate cancer and its clinical implications.

Background: Treatment-induced neuroendocrine prostate cancer (tNEPC) is an aggressive variant of late-stage metastatic castrate-resistant prostate cancer that commonly arises through neuroendocrine transdifferentiation (NEtD). Treatment options are limited, ineffective, and, for most patients, result in death in less than a year. We previously developed a first-in-field patient-derived xenograft (PDX) model of NEtD. Longitudinal deep transcriptome profiling of this model enabled monitoring of dynamic transcriptional changes during NEtD and in the context of androgen deprivation. Long non-coding RNA (lncRNA) are implicated in cancer where they can control gene regulation. Until now, the expression of lncRNAs during NEtD and their clinical associations were unexplored. Results: We implemented a next-generation sequence analysis pipeline that can detect transcripts at low expression levels and built a genome-wide catalogue (n = 37,749) of lncRNAs. We applied this pipeline to 927 clinical samples and our high-fidelity NEtD model LTL331 and identified 821 lncRNAs in NEPC. Among these are 122 lncRNAs that robustly distinguish NEPC from prostate adenocarcinoma (AD) patient tumours. The highest expressed lncRNAs within this signature are H19, LINC00617, and SSTR5-AS1. Another 742 are associated with the NEtD process and fall into four distinct patterns of expression (NEtD lncRNA Class I, II, III, and IV) in our PDX model and clinical samples. Each class has significant (z-scores >2) and unique enrichment for transcription factor binding site (TFBS) motifs in their sequences. Enriched TFBS include (1) TP53 and BRN1 in Class I, (2) ELF5, SPIC, and HOXD1 in Class II, (3) SPDEF in Class III, (4) HSF1 and FOXA1 in Class IV, and (5) TWIST1 when merging Class III with IV. Common TFBS in all NEtD lncRNA were also identified and include E2F, REST, PAX5, PAX9, and STAF. Interrogation of the top deregulated candidates (n = 100) in radical prostatectomy adenocarcinoma samples with long-term follow-up (median 18 years) revealed significant clinicopathological associations. Specifically, we identified 25 that are associated with rapid metastasis following androgen deprivation therapy (ADT). Two of these lncRNAs (SSTR5-AS1 and LINC00514) stratified patients undergoing ADT based on patient outcome. Discussion: To date, a comprehensive characterization of the dynamic landscape of lncRNAs during the NEtD process has not been performed. A temporal analysis of the PDX-based NEtD model has for the first time provided this dynamic landscape. TFBS analysis identified NEPC-related TF motifs present within the NEtD lncRNA sequences, suggesting functional roles for these lncRNAs in NEPC pathogenesis. Furthermore, select NEtD lncRNAs appear to be associated with metastasis and patients receiving ADT. Treatment-related metastasis is a clinical consequence of NEPC tumours. Top candidate lncRNAs FENDRR, H19, LINC00514, LINC00617, and SSTR5-AS1 identified in this study are implicated in the development of NEPC. We present here for the first time a genome-wide catalogue of NEtD lncRNAs that characterize the transdifferentiation process and a robust NEPC lncRNA patient expression signature. To accomplish this, we carried out the largest integrative study that applied a PDX NEtD model to clinical samples. These NEtD and NEPC lncRNAs are strong candidates for clinical biomarkers and therapeutic targets and warrant further investigation.

Whole-genome and Transcriptome Sequencing of Prostate Cancer Identify New Genetic Alterations Driving Disease Progression.

BACKGROUND: Global disparities in prostate cancer (PCa) incidence highlight the urgent need to identify genomic abnormalities in prostate tumors in different ethnic populations including Asian men. OBJECTIVE: To systematically explore the genomic complexity and define disease-driven genetic alterations in PCa. DESIGN, SETTING, AND PARTICIPANTS: The study sequenced whole-genome and transcriptome of tumor-benign paired tissues from 65 treatment-naive Chinese PCa patients. Subsequent targeted deep sequencing of 293 PCa-relevant genes was performed in another cohort of 145 prostate tumors. OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS: The genomic alteration landscape in PCa was analyzed using an integrated computational pipeline. Relationships with PCa progression and survival were analyzed using nonparametric test, log-rank, and multivariable Cox regression analyses. RESULTS AND LIMITATIONS: We demonstrated an association of high frequency of CHD1 deletion with a low rate of TMPRSS2-ERG fusion and relatively high percentage of mutations in androgen receptor upstream activator genes in Chinese patients. We identified five putative clustered deleted tumor suppressor genes and provided experimental and clinical evidence that PCDH9, deleted/loss in approximately 23% of tumors, functions as a novel tumor suppressor gene with prognostic potential in PCa. Furthermore, axon guidance pathway genes were frequently deregulated, including gain/amplification of PLXNA1 gene in approximately 17% of tumors. Functional and clinical data analyses showed that increased expression of PLXNA1 promoted prostate tumor growth and independently predicted prostate tumor biochemical recurrence, metastasis, and poor survival in multi-institutional cohorts of patients with PCa. A limitation of this study is that other genetic alterations were not experimentally investigated. CONCLUSIONS: There are shared and salient genetic characteristics of PCa in Chinese and Caucasian men. Novel genetic alterations in PCDH9 and PLXNA1 were associated with disease progression. PATIENT SUMMARY: We reported the first large-scale and comprehensive genomic data of prostate cancer from Asian population. Identification of these genetic alterations may help advance prostate cancer diagnosis, prognosis, and treatment.