An Integrated TCGA Pan-Cancer Clinical Data Resource to Drive High-Quality Survival Outcome Analytics.

For a decade, The Cancer Genome Atlas (TCGA) program collected clinicopathologic annotation data along with multi-platform molecular profiles of more than 11,000 human tumors across 33 different cancer types. TCGA clinical data contain key features representing the democratized nature of the data collection process. To ensure proper use of this large clinical dataset associated with genomic features, we developed a standardized dataset named the TCGA Pan-Cancer Clinical Data Resource (TCGA-CDR), which includes four major clinical outcome endpoints. In addition to detailing major challenges and statistical limitations encountered during the effort of integrating the acquired clinical data, we present a summary that includes endpoint usage recommendations for each cancer type. These TCGA-CDR findings appear to be consistent with cancer genomics studies independent of the TCGA effort and provide opportunities for investigating cancer biology using clinical correlates at an unprecedented scale.

Cell-of-Origin Patterns Dominate the Molecular Classification of 10,000 Tumors from 33 Types of Cancer.

We conducted comprehensive integrative molecular analyses of the complete set of tumors in The Cancer Genome Atlas (TCGA), consisting of approximately 10,000 specimens and representing 33 types of cancer. We performed molecular clustering using data on chromosome-arm-level aneuploidy, DNA hypermethylation, mRNA, and miRNA expression levels and reverse-phase protein arrays, of which all, except for aneuploidy, revealed clustering primarily organized by histology, tissue type, or anatomic origin. The influence of cell type was evident in DNA-methylation-based clustering, even after excluding sites with known preexisting tissue-type-specific methylation. Integrative clustering further emphasized the dominant role of cell-of-origin patterns. Molecular similarities among histologically or anatomically related cancer types provide a basis for focused pan-cancer analyses, such as pan-gastrointestinal, pan-gynecological, pan-kidney, and pan-squamous cancers, and those related by stemness features, which in turn may inform strategies for future therapeutic development.

Machine Learning Identifies Stemness Features Associated with Oncogenic Dedifferentiation.

Cancer progression involves the gradual loss of a differentiated phenotype and acquisition of progenitor and stem-cell-like features. Here, we provide novel stemness indices for assessing the degree of oncogenic dedifferentiation. We used an innovative one-class logistic regression (OCLR) machine-learning algorithm to extract transcriptomic and epigenetic feature sets derived from non-transformed pluripotent stem cells and their differentiated progeny. Using OCLR, we were able to identify previously undiscovered biological mechanisms associated with the dedifferentiated oncogenic state. Analyses of the tumor microenvironment revealed unanticipated correlation of cancer stemness with immune checkpoint expression and infiltrating immune cells. We found that the dedifferentiated oncogenic phenotype was generally most prominent in metastatic tumors. Application of our stemness indices to single-cell data revealed patterns of intra-tumor molecular heterogeneity. Finally, the indices allowed for the identification of novel targets and possible targeted therapies aimed at tumor differentiation.

Perspective on Oncogenic Processes at the End of the Beginning of Cancer Genomics.

The Cancer Genome Atlas (TCGA) has catalyzed systematic characterization of diverse genomic alterations underlying human cancers. At this historic junction marking the completion of genomic characterization of over 11,000 tumors from 33 cancer types, we present our current understanding of the molecular processes governing oncogenesis. We illustrate our insights into cancer through synthesis of the findings of the TCGA PanCancer Atlas project on three facets of oncogenesis: (1) somatic driver mutations, germline pathogenic variants, and their interactions in the tumor; (2) the influence of the tumor genome and epigenome on transcriptome and proteome; and (3) the relationship between tumor and the microenvironment, including implications for drugs targeting driver events and immunotherapies. These results will anchor future characterization of rare and common tumor types, primary and relapsed tumors, and cancers across ancestry groups and will guide the deployment of clinical genomic sequencing.

Comprehensive Molecular Characterization of Muscle-Invasive Bladder Cancer.

We report a comprehensive analysis of 412 muscle-invasive bladder cancers characterized by multiple TCGA analytical platforms. Fifty-eight genes were significantly mutated, and the overall mutational load was associated with APOBEC-signature mutagenesis. Clustering by mutation signature identified a high-mutation subset with 75% 5-year survival. mRNA expression clustering refined prior clustering analyses and identified a poor-survival "neuronal" subtype in which the majority of tumors lacked small cell or neuroendocrine histology. Clustering by mRNA, long non-coding RNA (lncRNA), and miRNA expression converged to identify subsets with differential epithelial-mesenchymal transition status, carcinoma in situ scores, histologic features, and survival. Our analyses identified 5 expression subtypes that may stratify response to different treatments.

Comprehensive Molecular Portraits of Invasive Lobular Breast Cancer.

Invasive lobular carcinoma (ILC) is the second most prevalent histologic subtype of invasive breast cancer. Here, we comprehensively profiled 817 breast tumors, including 127 ILC, 490 ductal (IDC), and 88 mixed IDC/ILC. Besides E-cadherin loss, the best known ILC genetic hallmark, we identified mutations targeting PTEN, TBX3, and FOXA1 as ILC enriched features. PTEN loss associated with increased AKT phosphorylation, which was highest in ILC among all breast cancer subtypes. Spatially clustered FOXA1 mutations correlated with increased FOXA1 expression and activity. Conversely, GATA3 mutations and high expression characterized luminal A IDC, suggesting differential modulation of ER activity in ILC and IDC. Proliferation and immune-related signatures determined three ILC transcriptional subtypes associated with survival differences. Mixed IDC/ILC cases were molecularly classified as ILC-like and IDC-like revealing no true hybrid features. This multidimensional molecular atlas sheds new light on the genetic bases of ILC and provides potential clinical options.

Multiplatform analysis of 12 cancer types reveals molecular classification within and across tissues of origin.

Recent genomic analyses of pathologically defined tumor types identify "within-a-tissue" disease subtypes. However, the extent to which genomic signatures are shared across tissues is still unclear. We performed an integrative analysis using five genome-wide platforms and one proteomic platform on 3,527 specimens from 12 cancer types, revealing a unified classification into 11 major subtypes. Five subtypes were nearly identical to their tissue-of-origin counterparts, but several distinct cancer types were found to converge into common subtypes. Lung squamous, head and neck, and a subset of bladder cancers coalesced into one subtype typified by TP53 alterations, TP63 amplifications, and high expression of immune and proliferation pathway genes. Of note, bladder cancers split into three pan-cancer subtypes. The multiplatform classification, while correlated with tissue-of-origin, provides independent information for predicting clinical outcomes. All data sets are available for data-mining from a unified resource to support further biological discoveries and insights into novel therapeutic strategies.

Dynamic reprogramming of the kinome in response to targeted MEK inhibition in triple-negative breast cancer.

Kinase inhibitors have limited success in cancer treatment because tumors circumvent their action. Using a quantitative proteomics approach, we assessed kinome activity in response to MEK inhibition in triple-negative breast cancer (TNBC) cells and genetically engineered mice (GEMMs). MEK inhibition caused acute ERK activity loss, resulting in rapid c-Myc degradation that induced expression and activation of several receptor tyrosine kinases (RTKs). RNAi knockdown of ERK or c-Myc mimicked RTK induction by MEK inhibitors, and prevention of proteasomal c-Myc degradation blocked kinome reprogramming. MEK inhibitor-induced RTK stimulation overcame MEK2 inhibition, but not MEK1 inhibition, reactivating ERK and producing drug resistance. The C3Tag GEMM for TNBC similarly induced RTKs in response to MEK inhibition. The inhibitor-induced RTK profile suggested a kinase inhibitor combination therapy that produced GEMM tumor apoptosis and regression where single agents were ineffective. This approach defines mechanisms of drug resistance, allowing rational design of combination therapies for cancer.

Genomic basis for RNA alterations in cancer.

Transcript alterations often result from somatic changes in cancer genomes1. Various forms of RNA alterations have been described in cancer, including overexpression2, altered splicing3 and gene fusions4; however, it is difficult to attribute these to underlying genomic changes owing to heterogeneity among patients and tumour types, and the relatively small cohorts of patients for whom samples have been analysed by both transcriptome and whole-genome sequencing. Here we present, to our knowledge, the most comprehensive catalogue of cancer-associated gene alterations to date, obtained by characterizing tumour transcriptomes from 1,188 donors of the Pan-Cancer Analysis of Whole Genomes (PCAWG) Consortium of the International Cancer Genome Consortium (ICGC) and The Cancer Genome Atlas (TCGA)5. Using matched whole-genome sequencing data, we associated several categories of RNA alterations with germline and somatic DNA alterations, and identified probable genetic mechanisms. Somatic copy-number alterations were the major drivers of variations in total gene and allele-specific expression. We identified 649 associations of somatic single-nucleotide variants with gene expression in cis, of which 68.4% involved associations with flanking non-coding regions of the gene. We found 1,900 splicing alterations associated with somatic mutations, including the formation of exons within introns in proximity to Alu elements. In addition, 82% of gene fusions were associated with structural variants, including 75 of a new class, termed 'bridged' fusions, in which a third genomic location bridges two genes. We observed transcriptomic alteration signatures that differ between cancer types and have associations with variations in DNA mutational signatures. This compendium of RNA alterations in the genomic context provides a rich resource for identifying genes and mechanisms that are functionally implicated in cancer.

An approach for normalization and quality control for NanoString RNA expression data.

The NanoString RNA counting assay for formalin-fixed paraffin embedded samples is unique in its sensitivity, technical reproducibility and robustness for analysis of clinical and archival samples. While commercial normalization methods are provided by NanoString, they are not optimal for all settings, particularly when samples exhibit strong technical or biological variation or where housekeeping genes have variable performance across the cohort. Here, we develop and evaluate a more comprehensive normalization procedure for NanoString data with steps for quality control, selection of housekeeping targets, normalization and iterative data visualization and biological validation. The approach was evaluated using a large cohort ($N=\kern0.5em 1649$) from the Carolina Breast Cancer Study, two cohorts of moderate sample size ($N=359$ and$130$) and a small published dataset ($N=12$). The iterative process developed here eliminates technical variation (e.g. from different study phases or sites) more reliably than the three other methods, including NanoString's commercial package, without diminishing biological variation, especially in long-term longitudinal multiphase or multisite cohorts. We also find that probe sets validated for nCounter, such as the PAM50 gene signature, are impervious to batch issues. This work emphasizes that systematic quality control, normalization and visualization of NanoString nCounter data are an imperative component of study design that influences results in downstream analyses.

In silico APC/C substrate discovery reveals cell cycle-dependent degradation of UHRF1 and other chromatin regulators.

The anaphase-promoting complex/cyclosome (APC/C) is an E3 ubiquitin ligase and critical regulator of cell cycle progression. Despite its vital role, it has remained challenging to globally map APC/C substrates. By combining orthogonal features of known substrates, we predicted APC/C substrates in silico. This analysis identified many known substrates and suggested numerous candidates. Unexpectedly, chromatin regulatory proteins are enriched among putative substrates, and we show experimentally that several chromatin proteins bind APC/C, oscillate during the cell cycle, and are degraded following APC/C activation, consistent with being direct APC/C substrates. Additional analysis revealed detailed mechanisms of ubiquitylation for UHRF1, a key chromatin regulator involved in histone ubiquitylation and DNA methylation maintenance. Disrupting UHRF1 degradation at mitotic exit accelerates G1-phase cell cycle progression and perturbs global DNA methylation patterning in the genome. We conclude that APC/C coordinates crosstalk between cell cycle and chromatin regulatory proteins. This has potential consequences in normal cell physiology, where the chromatin environment changes depending on proliferative state, as well as in disease.

A hierarchical spike-and-slab model for pan-cancer survival using pan-omic data.

BACKGROUND: Pan-omics, pan-cancer analysis has advanced our understanding of the molecular heterogeneity of cancer. However, such analyses have been limited in their ability to use information from multiple sources of data (e.g., omics platforms) and multiple sample sets (e.g., cancer types) to predict clinical outcomes. We address the issue of prediction across multiple high-dimensional sources of data and sample sets by using molecular patterns identified by BIDIFAC+, a method for integrative dimension reduction of bidimensionally-linked matrices, in a Bayesian hierarchical model. Our model performs variable selection through spike-and-slab priors that borrow information across clustered data. We use this model to predict overall patient survival from the Cancer Genome Atlas with data from 29 cancer types and 4 omics sources and use simulations to characterize the performance of the hierarchical spike-and-slab prior. RESULTS: We found that molecular patterns shared across all or most cancers were largely not predictive of survival. However, our model selected patterns unique to subsets of cancers that differentiate clinical tumor subtypes with markedly different survival outcomes. Some of these subtypes were previously established, such as subtypes of uterine corpus endometrial carcinoma, while others may be novel, such as subtypes within a set of kidney carcinomas. Through simulations, we found that the hierarchical spike-and-slab prior performs best in terms of variable selection accuracy and predictive power when borrowing information is advantageous, but also offers competitive performance when it is not. CONCLUSIONS: We address the issue of prediction across multiple sources of data by using results from BIDIFAC+ in a Bayesian hierarchical model for overall patient survival. By incorporating spike-and-slab priors that borrow information across cancers, we identified molecular patterns that distinguish clinical tumor subtypes within a single cancer and within a group of cancers. We also corroborate the flexibility and performance of using spike-and-slab priors as a Bayesian variable selection approach.

Racial differences in breast cancer outcomes by hepatocyte growth factor pathway expression.

PURPOSE: Black women have a 40% increased risk of breast cancer-related mortality. These outcome disparities may reflect differences in tumor pathways and a lack of targetable therapies for specific subtypes that are more common in Black women. Hepatocyte growth factor (HGF) is a targetable pathway that promotes breast cancer tumorigenesis, is associated with basal-like breast cancer, and is differentially expressed by race. This study assessed whether a 38-gene HGF expression signature is associated with recurrence and survival in Black and non-Black women. METHODS: Study participants included 1957 invasive breast cancer cases from the Carolina Breast Cancer Study. The HGF signature was evaluated in association with recurrence (n = 1251, 171 recurrences), overall, and breast cancer-specific mortality (n = 706, 190/328 breast cancer/overall deaths) using Cox proportional hazard models. RESULTS: Women with HGF-positive tumors had higher recurrence rates [HR 1.88, 95% CI (1.19, 2.98)], breast cancer-specific mortality [HR 1.90, 95% CI (1.26, 2.85)], and overall mortality [HR 1.69; 95% CI (1.17, 2.43)]. Among Black women, HGF positivity was significantly associated with higher 5-year rate of recurrence [HR 1.73; 95% CI (1.01, 2.99)], but this association was not significant in non-Black women [HR 1.68; 95% CI (0.72, 3.90)]. Among Black women, HGF-positive tumors had elevated breast cancer-specific mortality [HR 1.80, 95% CI (1.05, 3.09)], which was not significant in non-Black women [HR 1.52; 95% CI (0.78, 2.99)]. CONCLUSION: This multi-gene HGF signature is a poor-prognosis feature for breast cancer and may identify patients who could benefit from HGF-targeted treatments, an unmet need for Black and triple-negative patients.

Hepatocyte growth factor pathway expression in breast cancer by race and subtype.

BACKGROUND: African American women have the highest risk of breast cancer mortality compared to other racial groups. Differences in tumor characteristics have been implicated as a possible cause; however, the tumor microenvironment may also contribute to this disparity in mortality. Hepatocyte growth factor (HGF) is a stroma-derived marker of the tumor microenvironment that may affect tumor progression differentially by race. OBJECTIVE: To examine whether an HGF gene expression signature is differentially expressed by race and tumor characteristics. METHODS: Invasive breast tumors from 1957 patients were assessed for a 38-gene RNA-based HGF gene expression signature. Participants were black (n = 1033) and non-black (n = 924) women from the population-based Carolina Breast Cancer Study (1993-2013). Generalized linear models were used to estimate the relative frequency differences (RFD) in HGF status by race, clinical, and demographic factors. RESULTS: Thirty-two percent of tumors were positive for the HGF signature. Black women were more likely [42% vs. 21%; RFD = + 19.93% (95% CI 16.00, 23.87)] to have HGF-positive tumors compared to non-black women. Triple-negative patients had a higher frequency of HGF positivity [82% vs. 13% in non-triple-negative; RFD = + 65.85% (95% CI 61.71, 69.98)], and HGF positivity was a defining feature of basal-like subtype [92% vs. 8% in non-basal; RFD = + 81.84% (95% CI 78.84, 84.83)]. HGF positivity was associated with younger age, stage, higher grade, and high genomic risk of recurrence (ROR-PT) score. CONCLUSION: HGF expression is a defining feature of basal-like tumors, and its association with black race and young women suggests it may be a candidate pathway for understanding breast cancer disparities.

Protein-based immune profiles of basal-like vs. luminal breast cancers.

Tumor-infiltrating lymphocytes play an important, but incompletely understood role in chemotherapy response and prognosis. In breast cancer, there appear to be distinct immune responses by subtype, but most studies have used limited numbers of protein markers or bulk sequencing of RNA to characterize immune response, in which spatial organization cannot be assessed. To identify immune phenotypes of Basal-like vs. Luminal breast cancer we used the GeoMx® (NanoString) platform to perform digital spatial profiling of immune-related proteins in tumor whole sections and tissue microarrays (TMA). Visualization of CD45, CD68, or pan-Cytokeratin by immunofluorescence was used to select regions of interest in formalin-fixed paraffin embedded tissue sections. Forty-four antibodies representing stromal markers and multiple immune cell types were applied to quantify the tumor microenvironment. In whole tumor slides, immune hot spots (CD45+) had increased expression of many immune markers, suggesting a diverse and robust immune response. In epithelium-enriched areas, immune signals were also detectable and varied by subtype, with regulatory T-cell (Treg) markers (CD4, CD25, and FOXP3) being higher in Basal-like vs. Luminal breast cancer. Extending these findings to TMAs with more patients (n = 75), we confirmed subtype-specific immune profiles, including enrichment of Treg markers in Basal-likes. This work demonstrated that immune responses can be detected in epithelium-rich tissue, and that TMAs are a viable approach for obtaining important immunoprofiling data. In addition, we found that immune marker expression is associated with breast cancer subtype, suggesting possible prognostic, or targetable differences.

Virus expression detection reveals RNA-sequencing contamination in TCGA.

BACKGROUND: Contamination of reagents and cross contamination across samples is a long-recognized issue in molecular biology laboratories. While often innocuous, contamination can lead to inaccurate results. Cantalupo et al., for example, found HeLa-derived human papillomavirus 18 (H-HPV18) in several of The Cancer Genome Atlas (TCGA) RNA-sequencing samples. This work motivated us to assess a greater number of samples and determine the origin of possible contaminations using viral sequences. To detect viruses with high specificity, we developed the publicly available workflow, VirDetect, that detects virus and laboratory vector sequences in RNA-seq samples. We applied VirDetect to 9143 RNA-seq samples sequenced at one TCGA sequencing center (28/33 cancer types) over 5 years. RESULTS: We confirmed that H-HPV18 was present in many samples and determined that viral transcripts from H-HPV18 significantly co-occurred with those from xenotropic mouse leukemia virus-related virus (XMRV). Using laboratory metadata and viral transcription, we determined that the likely contaminant was a pool of cell lines known as the "common reference", which was sequenced alongside TCGA RNA-seq samples as a control to monitor quality across technology transitions (i.e. microarray to GAII to HiSeq), and to link RNA-seq to previous generation microarrays that standardly used the "common reference". One of the cell lines in the pool was a laboratory isolate of MCF-7, which we discovered was infected with XMRV; another constituent of the pool was likely HeLa cells. CONCLUSIONS: Altogether, this indicates a multi-step contamination process. First, MCF-7 was infected with an XMRV. Second, this infected cell line was added to a pool of cell lines, which contained HeLa. Finally, RNA from this pool of cell lines contaminated several TCGA tumor samples most-likely during library construction. Thus, these human tumors with H-HPV or XMRV reads were likely not infected with H-HPV 18 or XMRV.

Comprehensive Molecular Characterization of Papillary Renal-Cell Carcinoma.

BACKGROUND: Papillary renal-cell carcinoma, which accounts for 15 to 20% of renal-cell carcinomas, is a heterogeneous disease that consists of various types of renal cancer, including tumors with indolent, multifocal presentation and solitary tumors with an aggressive, highly lethal phenotype. Little is known about the genetic basis of sporadic papillary renal-cell carcinoma, and no effective forms of therapy for advanced disease exist. METHODS: We performed comprehensive molecular characterization of 161 primary papillary renal-cell carcinomas, using whole-exome sequencing, copy-number analysis, messenger RNA and microRNA sequencing, DNA-methylation analysis, and proteomic analysis. RESULTS: Type 1 and type 2 papillary renal-cell carcinomas were shown to be different types of renal cancer characterized by specific genetic alterations, with type 2 further classified into three individual subgroups on the basis of molecular differences associated with patient survival. Type 1 tumors were associated with MET alterations, whereas type 2 tumors were characterized by CDKN2A silencing, SETD2 mutations, TFE3 fusions, and increased expression of the NRF2-antioxidant response element (ARE) pathway. A CpG island methylator phenotype (CIMP) was observed in a distinct subgroup of type 2 papillary renal-cell carcinomas that was characterized by poor survival and mutation of the gene encoding fumarate hydratase (FH). CONCLUSIONS: Type 1 and type 2 papillary renal-cell carcinomas were shown to be clinically and biologically distinct. Alterations in the MET pathway were associated with type 1, and activation of the NRF2-ARE pathway was associated with type 2; CDKN2A loss and CIMP in type 2 conveyed a poor prognosis. Furthermore, type 2 papillary renal-cell carcinoma consisted of at least three subtypes based on molecular and phenotypic features. (Funded by the National Institutes of Health.).

Differences in race, molecular and tumor characteristics among women diagnosed with invasive ductal and lobular breast carcinomas.

BACKGROUND: The dominant invasive breast cancer histologic subtype, ductal carcinoma, shows intrinsic subtype diversity. However, lobular breast cancers are predominantly Luminal A. Both histologic subtypes show distinct relationships with patient and tumor characteristics, but it is unclear if these associations remain after accounting for intrinsic subtype. METHODS: Generalized linear models were used to estimate relative frequency differences (RFDs) and 95% confidence intervals (95% CIs) for the associations between age, race, tumor characteristics, immunohistochemistry (IHC) and RNA-based intrinsic subtype, TP53 status, and histologic subtype in the Carolina Breast Cancer Study (CBCS, n = 3,182) and The Cancer Genome Atlas (TCGA, n = 808). RESULTS: Relative to ductal tumors, lobular tumors were significantly more likely to be Luminal A [CBCS RNA RFD: 44.9%, 95% CI (39.6, 50.1); TCGA: RFD: 50.5%, 95% CI (43.9, 57.1)], were less frequent among young (≤ 50 years) and black women, were larger in size, low grade, less frequently had TP53 pathway defects, and were diagnosed at later stages. These associations persisted among Luminal A tumors (n = 242). CONCLUSIONS: While histology is strongly associated with molecular characteristics, histologic associations with age, race, size, grade, and stage persisted after restricting to Luminal A subtype. Histology may continue to be clinically relevant among Luminal A breast cancers.