The Cancer Genome Atlas Cases Research Articles

Artificial Intelligence-Based Histopathological Subtyping of High-Grade Serous Ovarian Cancer

Four subtypes of ovarian high-grade serous carcinoma (HGSC) have previously been identified, each with different prognoses and drug sensitivities. However, the accuracy of the classification depended on the assessor’s experience. This study aimed to develop a universal algorithm for HGSC-subtype classification using deep learning techniques. An artificial intelligence (AI)-based classification algorithm, which replicates the consensus diagnosis of pathologists, was formulated to analyze the morphological patterns and tumor-infiltrating lymphocyte counts for each tile extracted from whole slide images of ovarian HGSC available in The Cancer Genome Atlas (TCGA) dataset. The accuracy of the algorithm was determined using the validation set from the Japanese Gynecologic Oncology Group 3022A1 (JGOG3022A1) and Kindai and Kyoto University (Kindai/Kyoto) cohorts. The algorithm classified the four HGSC-subtypes with mean accuracies of 0.933, 0.910, and 0.862 for the TCGA, JGOG3022A1, and Kindai/Kyoto cohorts, respectively. To compare Mesenchymal Transition (MT) with non-MT groups, overall survival analysis was performed in the TCGA dataset. The AI-based prediction of HGSC-subtype classification in TCGA cases showed that the MT group had a worse prognosis than the non-MT group (p = 0.017). Furthermore, Cox proportional hazard regression analysis identified AI-based MT subtype classification prediction as a contributing factor along with residual disease after surgery, stage, and age. In conclusion, a robust AI-based HGSC-subtype classification algorithm was established using virtual slides of ovarian HGSC.

Abstract 3439: Factors affecting the performance of existing epigenetic clocks to predict chronological age in colorectal tumor and normal tissues

Abstract Epigenetic age (EA) may serve as a risk stratification biomarker for early-onset colorectal cancer (EOCRC, <50 years old), which has been rising substantially in recent decades. We utilized publicly available data in a descriptive analysis to compare epigenetic clocks in colorectal tissue and to investigate differences between chronological age (CA) and biological age, as measured by EA, in relation to demographic and clinical features. We analyzed DNA methylation (DNAm) array data on 144 participants (96 CRC cases, 48 healthy controls) from the Colonomics Study (CLX; Barcelona, Spain) and 331 CRC cases (51 EOCRC, 316 average-onset (≥50) colorectal cancer (AOCRC)) from The Cancer Genome Atlas (TCGA) as well as corresponding demographic and clinical variables. Among CLX participants, the 96 CRC cases provided colon tumor and paired normal mucosa samples, while the 48 controls provided healthy colon mucosa samples. TCGA included primary tumor samples from 331 cases with colon or rectal adenocarcinoma (TCGA-COADREAD). The 2013 Horvath clock estimated EA based on CA, while epiTOC, estimated the relative stem cell division rate from tissue biospecimens. We examined the correlations between CA and EA stratified by case/control status, sample type, sex, and age of onset (TCGA only). Tissue from CLX controls demonstrated a strong, positive correlation (r>0.9) between CA and EA for both the 2013 Horvath and epiTOC clocks. Using the Horvath clock, CA and EA correlations in normal mucosa and tumor tissue samples from CLX varied, with r = 0.74 and r = 0.41, respectively, and were weaker using the epiTOC clock, with r = 0.26 (normal) and r = 0.10 (tumor). When CLX cases were stratified by sex, the CA and EA (Horvath) correlations among female cases were r = 0.80 (normal) and r = -0.12 (tumor), while the correlations for males were stronger at r = 0.58 (normal) and r = 0.31 (tumor). This is in comparison to the epiTOC clock with females showing correlations of r = -0.02 (normal) and r = 0.41 (tumor), while males had r = 0.06 (normal) and r = 0.11 (tumor). In TCGA, correlations for EOCRC were r = -0.20 (Horvath) and r = -0.08 (epiTOC) and for AOCRC were r = 0.29 and 0.20, respectively. By sex, correlations for females with EOCRC for Horvath and epiTOC were r = -0.09 and r = -0.07, respectively, while correlations for males with EOCRC were r = -0.02 and r = -0.08, respectively. For TCGA cases with AOCRC, the correlations between CA and EA were r = 0.35 and r = 0.22 for females, and r = 0.26 and r = 0.20 for males, respectively for Horvath and epiTOC. This study revealed substantial variability in the ability of existing epigenetic clocks to accurately predict chronological age, with tissue type and sex strongly influencing the resulting correlations. Our findings highlight the importance of considering these factors when choosing an epigenetic clock. Citation Format: Christopher L. Benson, Ferran Moratalla-Navarro, Anna Diez-Villanueva, Matthew A. Devall, Victor Moreno, Stephanie L. Schmit, Fredrick R. Schumacher. Factors affecting the performance of existing epigenetic clocks to predict chronological age in colorectal tumor and normal tissues [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 3439.

Histomolecular classification of urothelial carcinoma of the urinary bladder : From histological phenotype to genotype and back

Of all urothelial carcinomas (UCs), 25% are muscle invasive and associated with a5-year overall survival rate of 50%. Findings regarding the molecular classification of muscle-invasive urothelial carcinomas (MIUCs) have not yet found their way into clinical practice. Prediction of molecular consensus subtypes in MIUCs with artificial intelligence (AI) based on histologic hematoxylin-eosin (HE) sections. Pathologic review and annotation of The Cancer Genome Atlas (TCGA) Bladder Cancer (BLCA) Cohort (N = 412) and the Dr. Senckenberg Institute of Pathology (SIP) BLCA Cohort (N = 181). An AI model for the prediction of molecular subtypes based on annotated histomorphology was trained. For afive-fold cross-validation with TCGA cases (N = 274), an internal TCGA test set (N = 18) and an external SIP test set (N = 27), we reached mean area under the receiver operating characteristic curve (AUROC) scores of 0.73, 0.8 and 0.75 for the classification of the used molecular subtypes "luminal", "basal/squamous" and "stroma-rich". By training on correlations to individual molecular subtypes, rather than training on one subtype assignment per case, the AI prediction of subtypes could be significantly improved. Follow-up studies with RNA extraction from various areas of AI-predicted molecular heterogeneity may improve molecular classifications and thereby AI algorithms trained on these classifications.

TRB CDR3-cancer testis antigen chemical complementarity scoring for identifying productive immune responses in renal cell carcinoma.

Immunogenomics approaches to the characterization of renal cell carcinoma (RCC) have helped to better our understanding of the features of RCC immune dysfunction. However, much is still unknown with regard to specific immune interactions and their impact in the tumor microenvironment. This study applied chemical complementarity scoring for the TRB complementarity determining region-3 (CDR3) amino acid sequences and cancer testis antigens (CTAs) to determine whether such complementarity correlated with survival and the expression of immune marker genes. TRB recombination reads from RCC tumor samples from RNAseq files obtained from two separate databases, Moffitt Cancer Center and The Cancer Genome Atlas (TCGA), were evaluated. Chemical complementarity scores (CSs) were calculated for TRB CDR3-CTA pairs and survival assessments based on those CSs were performed. Moffitt Cancer Center and TCGA cases representing the upper 50th percentile of chemical CSs for TRB CDR3 amino acid sequences and the CTA POTEA were found to be associated with a better overall survival (OS) Also, greater tumor RNA expression of multiple immune signature genes, including granzyme A, granzyme B, and interferon-gamma were correlated with the higher chemical CSs. These results indicate that TRB CDR3-CTA chemical complementarity scoring may be useful in distinguishing RCC cases with a productive, anti-tumor immune response from cases where basic immune parameter assessments are inconsistent with a productive immune response.

Abstract 5433: Representation learning for histological profiling of lung squamous premalignant lesions and tumors

Abstract Lung squamous cell carcinoma (LSCC) is preceded by the development of bronchial premalignant lesions (PMLs). PMLs progress through a series of histologic grades characterized by molecular and morphologic alterations. To enhance our understanding of PML progression to lung cancer, we are developing deep learning methods to quantitate PML histologic features and localize PML severity along the histologic disease spectrum. We leveraged whole slide images (WSIs) of lung tumors (lung adenocarcinoma, LUAD and LSCC) and non-involved adjacent (referred to as ‘normal’) tissue from the Clinical Proteomic Tumor Analysis Consortium (CPTAC), the National Lung Screening Trial (NLST) and the Cancer Genome Atlas (TCGA). The NLST WSIs were used for training a self-supervised contrastive learning (simCLR) model. A graph isomorphism network (GIN) trained on CPTAC WSIs was developed using the features from simCLR and perform WSI-level predictions on tissue subtype (LUAD, LSCC, normal). Model performance was assessed using area under the receiver operating curves (AUC). The model was used to learn WSI-level features of endobronchial biopsies of PMLs from two datasets (PCGA, n=365 samples spanning all histologic grades, and CIS, n=112, only high-grade samples). The features from the final layer of the GIN were used to compute principal components (PCs) and visualize the relationships between samples using t-SNE and UMAP. To interpret how the GIN processes WSI data, we performed gradient-based class activation mapping (Grad-CAMs) on the graphs. High model performance was observed on the CPTAC test dataset (Accuracy = 87%, AUC >= 0.89) but dropped slightly on the TCGA dataset (Accuracy = 72%, AUC >= 0.75). The GIN-CAMs identified WSI regions that were highly associated with the output label based on expert pathologist annotation on TCGA cases. Clustering revealed distinct groupings of normal, LUAD and LSCC subtypes (PC1 & PC2, p<0.01). Most WSIs from the PCGA data clustered with normal tissues (PC1 & PC2: p<0.01 for PCGA v LUAD, PCGA v LSCC). However, a portion of the CIS samples were closely grouped with LSCC tumor cases (74% of CIS WSIs were classified as LSCC tumors. PC1 & PC2: p<0.01 for CIS v LUAD; PC1: p<0.01 for CIS v LSCC). Additionally, the feature distribution of CIS samples that progressed towards LUSC tumors were significantly different than those that regressed (PC1 & PC3: p < 0.05 for Progressive v Regressive). The GIN recognizes and generates robust features that are distinctly and consistently observed in the different histologic groups across multiple cohorts. These features are also able to stratify progressive and regressive high-grade lesions. The stratification of PMLs is an important step in designing novel interception strategies to prevent the development of lung cancer, and our results suggest that pathology data may be efficacious to include in future biomarkers. Citation Format: Rushin Gindra, Yi Zheng, Regan Conrad, Emily Green, Sarah Mazzilli, Ehab Billatos, Mary Reid, Eric Burks, Vijaya Kolachalama, Jennifer E. Beane. Representation learning for histological profiling of lung squamous premalignant lesions and tumors. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 5433.

Abstract 4108: Understanding the role of telomere biology gene variation in cancer etiology

Abstract Telomere biology disorders (TBDs) are cancer-prone syndromes associated with increased risk of acute myeloid leukemia (AML), head and neck squamous cell carcinoma (HNSCC), bone marrow failure, pulmonary fibrosis, and liver disease. Germline variants in at least 15 different telomere biology genes have been implicated (ACD, CTC1, TERT, TERC, STN1, NAF1, NOP10, NHP2, TINF2, RTEL1, PARN, ZCCHC8, DKC1, WRAP53, POT1) with autosomal dominant (AD), autosomal recessive (AR), or X-linked (XLR) inheritance as well as de novo occurrence. We hypothesize that TBDs may be more common than the currently estimated 1:1,000,000. We analyzed the prevalence of germline variants in the 15 TBD-associated genes using 1) The Genome Aggregation Database [gnomAD v2.1.1 non-cancer] and 2) The Cancer Genome Atlas (TCGA). Germline TCGA variants were called using HaplotypeCaller, Freebayes, and UnifiedGenotyper. All variants with minor allele frequency (MAF) <1% in gnomAD non-cancer were considered. They were classified as potentially deleterious based on ClinVar and/or loss of function classification and/or stringent in silico predictions utilizing REVEL, MetaSVM, CADD, BayesDel, Eigen for missense and a combination of spliceAI, spidex, and dbscSNV for splice site variants. For TERC (RNA telomerase template) variants in the pseudoknot/template region were considered deleterious. The known disease-causing inheritance pattern (AD, AR, XLR) for each gene was used to identify TCGA cases with a probable TBD. There were 1215 potentially deleterious variants in TBD associated genes in the gnomAD v2.1.1 non-cancer dataset (n=134,187 samples) for a combined prevalence of 0.9%, without accounting for zygosity. In genes with AD/AR inheritance, variants were most common in: RTEL1 (249), TERT (104), and PARN (104). CTC1 and WRAP53, both associated with solely AR TBDs, also showed high variant frequencies (280 and 102, respectively). In the TCGA dataset, 9089 cancer cases (32 solid tumors and AML) were evaluated. We identified 227 cases (2.5%) with 161 mono- or biallelic rare, potentially deleterious variants in the 15 genes. Ninety-two cases had a probable underlying TBD [1% of total, 41 male, 51 female, median age at cancer diagnosis 46.5 (1-92) years]. Cancer diagnoses in these 92 cases included 24 solid tumors. The frequency of individuals with deleterious TBD-associated variants ranged from 0.4-2.8% in each of the 24 affected cancers. Notably, the heterozygous frequencies in lung squamous cell carcinoma and liver hepatocellular carcinoma were 1.6% and 2.2%, respectively. Our results show that the prevalence of TBDs may be noticeably higher than previously estimated, specifically in individuals with cancer. TBD-associated pathogenic germline variants may be implicated in the etiology of more solid tumor entities than previously recognized, warranting further studies. Citation Format: Marena R. Niewisch, Jung Kim, Judith C. Lunger, Lisa J. McReynolds, Sharon A. Savage. Understanding the role of telomere biology gene variation in cancer etiology [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 4108.

Significance of tumor heterogeneity of p-Smad2 and c-Met in HER2-positive gastric carcinoma with lymph node metastasis

BackgroundTumor heterogeneity has frequently been observed in gastric cancer (GC), but the correlation between patients’ clinico-pathologic features and the tumoral heterogeneity of GC-associated molecules is unclear. We investigated the correlation between lymph node metastasis and the intra-tumoral heterogeneity of driver molecules in GC.Materials and methodsWe retrospectively analyzed the cases of 504 patients who underwent a gastrectomy at the Department of Gastroenterological Surgery, Osaka Metropolitan University and 389 cases drawn from The Cancer Genome Atlas (TCGA) data. We performed a clustering analysis based on eight cancer-associated molecules including HER2, c-Met, and p-Smad2 using the protein expression revealed by our immunohistochemical study of the patients’ and TCGA cases. We determined the correlations between HER2 expression and the other molecules based on the degree of lymph node metastasis.ResultsImmunohistochemical staining data showed that a 43 of the 504 patients with GC (8.5%) were HER2-positive. In the HER2-positive cases, the expressions of c-Met and p-Smad2 were increased in accord with the lymph-node metastatic level. The overall survival of the HER2-positive GC patients with both p-Smad2 and c-Met expression was significantly (p = 0.030) poorer than that of the patients with p-Smad2-negative and/or c-Met-negative expression. The results of the TCGA data analysis revealed that 58 of the 389 GC cases (14.9%) were ERBB2-positive. MET expression was more frequent in the N1 metastasis group than the N0 group. In the high lymph-node metastasis (N2 and N3) group, SMAD2 expression was more frequent, as was ERBB2 and MET expression.Conclusionp-Smad2 and c-Met signaling might play important roles in lymph node metastasis in HER2-positive GC.

Demographic disparities in genomic data and clinical trials for multiple myeloma.

8043 Background: There are considerable outcome disparities among patients with multiple myeloma (MM). African-Americans have higher risk of developing MM, earlier age at diagnosis, and higher mortality compared to Whites. While outcomes have multifactorial socioeconomic etiologies, race and ethnicity (R&E) correlate with genomic ancestry, and any role of genetics should be explored. This requires equitable representation in biological databases like The Cancer Genome Atlas (TCGA). Here, we characterize disparities in MM cases of TCGA and clinical trials (CT). Methods: MM incidence was obtained from North American Association of Central Cancer Registries (NAACCR) for R&E, sex and age ( < 50, 50-64, 65+ years). Race includes White, Black, and Asian, along with Hispanic ethnicity, as TCGA MM is limited to these groups. Genes with oncogenic potential and > 5% mutation were stratified by R&E and age. TCGA MM cases are from USA, Canada, Italy, and Spain. On ClinicalTrials.gov, completed MM CT limited to these countries were identified. Student’s t-test and chi-square test were used to analyze disparities and gene mutations. Kaplan-Meier curve was generated to evaluate survival. Results: TCGA MM representation is in Table, calculated as (demographic TCGA cases divided by total TCGA cases) divided by (demographic incidence divided by total incidence). Race was not reported in 20% of TCGA, 4% of CT, and 3% of NAACCR. Of cases reporting R&E, Hispanics are underrepresented by 26% in TCGA and 47% in CT relative to incidence, Blacks by 22% in TCGA and 41% in CT, and Asians by 40% in TCGA and 10% in CT. Whites are overrepresented, by 8% in TCGA and 7% in CT. More men are in TCGA than women, in all R&E. Less than 25% of patients age < 50 relative to incidence are in TCGA, in all R&E. BCL7A is mutated more in Blacks age < 50 (n = 20, 50%) than Whites (n = 39, 10%) (p < 0.01). FAT4 is mutated in Whites age 65+ (n = 31, 10%) but not Blacks. In all ages, KRAS is mutated more in Blacks than Whites (31-35% vs. 20-25%, p < 0.05). Black patients have lower survival than Whites (p < 0.02). In patients age 65+, KRAS mutation is associated with 10% lower 4.5-year survival. Conclusions: Substantial disparities in Black, Hispanic, women, and age representation exist for MM cases in TCGA and CT. This stratification by R&E and age offers new insight on BCL7A, FAT4, and KRAS mutation in MM. Mutation status is associated with survival in older patients. Equitable demographic representation should be pursued to improve quality of available data and access to medical resources for all populations. [Table: see text]

Do publicly available OncoGenomic databases represent the population? A comparative analysis.

6588 Background: Large publicly available databases are important repositories for analyses of clinicogenomic research used for identifying clinically relevant biomarkers. Diversity among individuals in these repositories is key for ensuring applicability of findings to patient populations. Methods: We compared two publicly available pancancer databases from academic institutions: The Cancer Genome Atlas (TCGA) and United States (US) institutions from the American Association for Cancer Research (AACR) Project GENIE version 11.0 (APG) with cancer incidence statistics from The US Cancer Statistics (USCS) in 2018, the most recently available data. We compared demographic data from key individual cancer types (Lung, Colorectal, Prostate, Breast, Gliomas, and Leukemias) for gender, race, and ethnicity. Frequencies are displayed as percentages and compared by Chi-Squared method. Results: The USCS includes 1,708,921 new cases in 2018 while the TCGA includes 12,958 cases and APG includes 109,041 cases. Women account for 49.5% of all cancer diagnosis and similarly 50.4% of all cases in AACR and 51.2% of TCGA cases. Table summarizes key demographic differences. Amongst all cancer types, 78% of all US cancers occur in White patients however 84% and 83% of patients were White in AACR and TCGA respectively. 16% of prostate cancer cases occur in Black patients, but only 9% (n = 328/3993) of AACR and 11% (n = 51/484) of TCGA cases were Black (p < 0.01). However, while Black patients are only 2% of all breast cancer diagnoses, they accounted for 9% (n = 841/9871) of AACR and 16% (n = 203/1343) of TCGA cases p < 0.01). Patients of Hispanic ethnicity were underrepresented amongst the population and all single tumor types with Hispanics accounting for 8% of cases in USCS but only 5% in the AACR and TCGA (p < 0.01). Pancreatic and lung cancers, which have historically short survivals, both had lower median ages at sequencing compared to the median age of diagnosis. Median age of sequencing for pancreatic cancers was 65 in both TCGA and AACR, while median age at diagnosis in US is 70. Median age at diagnosis of lung cancers is 71 years in US, however median age was 67 in both AACR and TCGA. Conclusions: Patients with advanced age and minority races are underrepresented in publicly available American databases. To make informed analyses from genomic databases, the diversity of the population must be reflected in these databases and efforts must be made to increase representation of underrepresented groups. [Table: see text]

Diverse Macrophages Constituted the Glioma Microenvironment and Influenced by PTEN Status.

The glioma immune microenvironment (GIM), consisting of glioma cells, stromal cells, and immune cells, accelerates the initiation, development, immune evasion, chemoresistance, and radioresistance of glioblastoma (GBM), whereas the immunosuppressive mechanisms of GBM have not been thoroughly elucidated to date. The glioma data downloaded from The Cancer Genome Atlas (TCGA) and the Chinese Glioma Genome Atlas (CGGA) databases were used to evaluate the composition of tumor-infiltrating immune cells (TICs) by the CIBERSORT algorithm. RNA-seq datasets from the TCGA and CGGA were used to analyze the relationship between immune scores with patients’ characteristics and TICs, which showed higher ratios of tumor-inhibiting/tumor-promoting signatures (M2/M1 macrophages) along with higher immune scores. The distribution of TICs among different glioma patients and the correlation with hazard ratio (HR) analysis suggested that M2 macrophages were abundant in malignant gliomas and indicated an unfavorable prognosis. We further analyzed TCGA cases with available mutation and copy-number alteration information, which showed that the status of PTEN could influence the immune microenvironment of glioma patients. Tissue microarrays of 39 GBM patients were carried out to confirm the clinical significance of PTEN and macrophage markers. We found that the high expression of PTEN was associated with a more extended survival period of glioma patients, positively correlated with M2 macrophages and negatively with M1 macrophages. Transwell and flow cytometry analyses demonstrated that PTEN status could prevent M1 to M2 polarization and M2 macrophage recruitment of gliomas in vitro. The newly discovered immunoregulatory activity of PTEN opens innovative avenues for investigations relevant to counteracting cancer development and progression.

Association of differential expression of immunoregulatory molecules and presence of targetable mutations may inform rational design of clinical trials

BackgroundImmune checkpoint inhibitors (ICIs) and genomic biomarker-driven targeted therapies have revolutionized the modern oncologic treatment arsenal. The next step has been to combine targeted agents and ICIs. In doing so, some combination regimens may be more logical than others.Patients and methodsWhole-exome and whole-transcriptome sequencing were performed on 2739 unselected later-stage clinical cases from 24 solid tumor subtypes in the NantHealth database, and data were also curated from 5746 similarly sequenced patients across 28 solid tumor subtypes in The Cancer Genome Atlas (TCGA). Significant differential expression of 10 immunoregulatory molecules [IRMs (genes)] was analyzed for association with mutant versus wild-type genes.ResultsTwenty-three significant associations between currently actionable variants and RNA-expressed checkpoint genes were identified in the TCGA cases; 10 were validated in the external cohort of 2739 clinical cases from NantHealth (P values were adjusted using Benjamini–Hochberg multiple hypothesis correction to reduce false-discovery rate). Within the same 5746 TCGA profiles, 2740 TCGA patients were identified as having one or more potentially oncogenic single-nucleotide variant (SNV) mutation within an established 50-gene hotspot panel. Of the 50 genes, SNVs within 15 were found to be significantly associated with differential expression of at least one IRM after adjusting for tissue enrichment; six were confirmed significant associations in an independent set of 2739 clinical cases from NantHealth.ConclusionsLogically combining ICIs with targeted therapies may offer unique treatment strategies for patients with cancer. The presence of specific mutations impacts the expression of IRMs, an observation of potential importance for selecting combinations of gene- and immune-targeted therapeutics.

Autocrine signaling by receptor tyrosine kinases in urothelial carcinoma of the bladder.

Comprehensive characterizations of bladder cancer (BCa) have established molecular phenotype classes with distinct alterations and survival trends. Extending these studies within the tyrosine kinase (TK) family to identify disease drivers could improve our use of TK inhibitors to treat specific patient groups or individuals. We examined the expression distribution of TKs as a class (n = 89) in The Cancer Genome Atlas (TCGA) muscle invasive BCa data set (n >400). Patient profiles of potentially oncogenic alterations (overexpression and/or amplification) clustered TKs into 3 groups; alterations of group 1 and 3 TKs were associated with significantly worse patient survival relative to those without alterations. Many TK pathways induce epithelial-to-mesenchymal transition (EMT), which promotes tumor invasiveness and metastasis. Overexpression and/or amplification among 9 EMT transcriptional activators occurred in 43% of TCGA cases. Co-occurring alterations of TKs and EMT transcriptional activators involved most group 1 TKs; 24% of these events were associated with significantly worse patient survival. Co-occurring alterations of receptor TKs and their cognate ligands occurred in 16% of TCGA cases and several BCa-derived cell lines. Suppression of GAS6, MST1 or CSF1, or their respective receptors (AXL, MST1R and CSF1R), in BCa cell lines was associated with decreased receptor activation, cell migration, cell proliferation and anchorage independent cell growth. These studies reveal the patterns and prevalence of potentially oncogenic TK pathway-related alterations in BCa and identify specific alterations associated with reduced BCa patient survival. Detection of these features in BCa patients could better inform TK inhibitor use and improve clinical outcomes.

Assessment of isochromosome 12p and 12p abnormalities in germ cell tumors using fluorescence in situ hybridization, single-nucleotide polymorphism arrays, and next-generation sequencing/mate-pair sequencing

The identification of isochromosome 12p [i(12p)] and 12p gains have significant clinical utility in the diagnosis of germ cell tumors (GCTs). We have summarized the results of fluorescence in situ hybridization (FISH) assays to identify i(12p), performed in a Clinical Laboratory Improvement Amendments (CLIA)-validated setting for 536 specimens. In addition, the American Association for Cancer Research (AACR) Project GENIE registry and The Cancer Genome Atlas (TCGA) data sets were evaluated for chromosome 12p gains, and a limited number of cases were concurrently evaluated using FISH, single-nucleotide polymorphism (SNP) arrays and next-generation sequencing (NGS; including mate-pair sequencing). Specimens submitted for FISH testing were frequently from potential sites of metastases (male: 70.9% and female: 69.3%), and polysomy of chromosome 12 with or without concurrent i(12p) was a frequent finding, seen in 3% (16/536) and 35% (186/536) of cases, respectively. Our analysis suggests that 12p gains are likely to be present in approximately 73% of male GCTand in 32% of female GCT (AACR GENIE, n=555). When comparing TCGA cases of testicular GCT (n=149) to combined cases of sarcoma, colorectal, prostate, and urothelial carcinoma (n=1754), 12p gains had a sensitivity of 77.2% and specificity of 97.3% for GCT. Some advantages of FISH over SNP arrays/NGS include relatively lower cost, rapid turnaround time, the ability to analyze biopsy material with a limited number of tumor cells (50cells), and the ability to distinguish i(12p) from polysomy. The ability to spatially restrict the analysis to cells of interest is critical, as specimens submitted for testing often have low tumor purity. Disadvantages include false negative results due to an inability to detect segmental gains due to FISH probe design. With the availability of numerous testing modalities, including FISH, SNP arrays, and NGS-based assays, a nuanced understanding of the advantages and disadvantages of each methodology, as has been presented in this study, may inform appropriate testing strategies.

Molecular characterization of high-grade serous ovarian cancers occurring in younger and older women

ObjectivesTo determine if the mutational landscapes and genomic features of homologous recombination DNA repair defects (HRD) vary between younger and older patients with high-grade serous ovarian cancer (HGSOC). MethodsYounger and older women were defined as bottom and top age quartiles, respectively. HGSOCs from 15 younger (median 49 years, range 35–53) and 15 older women (median 72 years, range 70–87) were subjected to whole-exome sequencing (WES). For validation, HGSOC WES data were obtained from The Cancer Genome Atlas (TCGA), including 38 younger (median 45 years, range 34–50) and 30 older women (median 74 years, range 68–84). Mutational profiles, BRCA1/2 status, genomic HRD features, and for TCGA cases RNA-sequencing-based HRD transcriptomic signatures were assessed. ResultsIn the institutional cohort, pathogenic germline BRCA1/2 mutations were more frequent in younger (5/15) than older women (0/15, p = 0.042). No somatic BRCA1/2 mutations were identified. HGSOCs from older patients preferentially displayed aging-related mutational signatures and, in contrast to younger patients, harbored CCNE1 amplifications (3/15, 20%). In the TCGA cohort, pathogenic germline BRCA1 (younger 8/38, older 0/30, p = 0.007) but not BRCA2 mutations (young 3/38, older 4/30, p = 0.691) were more frequent in younger patients. Again, no somatic BRCA1/2 mutations were identified. HGSOCs from younger women more frequently displayed genomic features of HRD (all, p < 0.05), a significant HRD gene-signature enrichment, but less frequently CCNE1 amplification (p = 0.05). Immunoreactive CLOVAR subtypes were more common in HGSOCs from younger women, and proliferative subtypes in HGSOCs from older women (p = 0.041). ConclusionsHGSOC patients diagnosed at an older age less frequently harbor pathogenic BRCA1 germline mutations and genomic features of HRD than younger women. Individualized treatment options, particularly pertaining to use of PARP inhibitors, in older women may be warranted.

CILP2 overexpression correlates with tumor progression and poor prognosis in patients with colorectal cancer in The Cancer Genome Atlas (TCGA) study

BackgroundGenetic alterations play an important role in the progression of colorectal cancer (CRC). Identifying new biomarkers to assess the prognosis of patients with CRC is critical. Cartilage intermediate layer protein 2 (CILP2) gene, screened from TCGA database by bioinformatics, may be closely related to the progression of CRC. CILP2 was barely reported with clinical features of tumors.Materials and methodsClinical information and RNA-seq data were derived from TCGA colorectal carcinoma cohort. CILP2 expression at mRNA level was estimated by bioinformatical analysis of TCGA cases. Tissue microarray (TMA) was constructed containing paraffin-embedded 64 pairs of CRC and matched adjacent normal tissues. The expression at the protein level was detected in 64 pairs of CRC and matched adjacent normal tissues by immunohistochemical analysis. CILP2 expression level and its clinical value were estimated by bioinformatical analysis with linear and logistic regression. Survival analysis was performed between high and low groups of CILP2 expression by Cox regression analysis, and the P value was calculated by the log-rank test. The Kaplan-Meier curves were tested by the log-rank test.ResultsCILP2 was statistically significantly higher expressed in the CRC tissues when compared with paired adjacent normal tissues in TCGA cohort (P < 0.001) and in the TMA cohort (P = 0.001). Also, CILP2 high expression was strongly correlated with T3/4 stage (P = 0.001), N1/2/3 stage (P = 0.005), M1 stage (P = 0.048), and higher clinical stage (UICC 2010 stage) (P < 0.001) in TCGA cohort, and also positively associated with T3/4 stage (P = 0.022) and higher clinical stage (UICC 2010 stage) (P = 0.03) in TMA cohort. Furthermore, CILP2 overexpression predicted poor prognosis and could be an independent prognostic factor (P = 0.003).ConclusionWe revealed that CILP2 is associated with advanced stages and could play a role as an independent predictor of poor survival in CRC.

Comprehensive Genomic Analysis of Translocation Renal Cell Carcinoma Reveals Copy-Number Variations as Drivers of Disease Progression.

Translocation renal cell carcinoma (tRCC) is a rare, aggressive renal cell carcinoma (RCC) subtype. There is currently limited understanding on the role of molecular alterations in the pathogenesis and progression of these tumors. We investigated the association between somatic alterations and clinical outcomes in two independent cohorts profiled using DNA sequencing. Twenty-two tRCCs underwent targeted sequencing [Memorial Sloan Kettering Cancer Center (MSK)-IMPACT]; a subset was profiled using exome-sequencing and combined with exome data from The Cancer Genome Atlas (TCGA) for analysis. The prognostic value of specific somatic aberrations, tumor mutation burden (TMB), and fraction of copy-number-altered genome (FCNAg) was explored. In TCGA cases, neoantigen prediction and immune cell deconvolution were performed using RNA-sequencing and exome data. Overall survival estimates were computed using the Kaplan-Meier method; time-on-treatment was calculated for 14 MSK-IMPACT patients who underwent systemic therapy. Associations between molecular features and outcomes were evaluated using nonparametric testing. Copy-number aberrant tRCCs were associated with poor overall survival (P = 0.03). Pediatric patients had tumors with lower FCNAg (P = 0.01). In one adult case with two chronologically distinct tumor samples sequenced, we confirmed that copy-number events occurred early during evolution. TERT promoter mutations were found exclusively in high-stage tumors. We found that tRCCs displayed distinct angiogenesis and PD-L1 gene expression profiles compared with other RCC subtypes. Tumors molecularly defined by increased copy-number variations were associated with aggressive disease in tRCC. A higher burden of genomic events in adults compared with pediatric cases likely reflects a more aggressive clinical course. The unique immunophenotypic characteristics of tRCC merit further exploration.

Multigene mutation signatures in colorectal cancer patients: Predict for the diagnosis, pathological classification, staging and prognosis.

e16113 Background: Like most cancers, early diagnosis of colorectal cancer (CRC) contributes to a good prognosis. Recently, some studies indicated that mutation gene signatures have good performances in predicting the treatment and prognosis in CRC patients. However, incomplete understanding of the genotype is not conducive to proper treatment. The objective was to give further genetic insights into CRC and provide more comprehensive references for clinical practice. Methods: In the training group, the whole genome sequencing (WGS), clinical, and demographic data of 531 patients were downloaded from The Cancer Genome Atlas (TCGA). The validation group contained 53 patients, which were collected in Tianjin Medical University Cancer Institute and Hospital (Tianjin, China) from April 2014 to November 2018. The fresh tissues collected within 24 hours after operation were used to extract genomic DNA, and then sequenced with targeted next-generation sequencing (NGS) technology to examine somatic mutations and analyze relevant gene concerning clinical indicators. The correlation between gene variation distribution and various indexes such as cancer species, stage, total survival period, sex, age and race were evaluated. Results: A total of 44 mutant genes with mutation frequencies above 5% were found in both TCGA cases and validated cases. Mutations of TP53, APC, KRAS, BRAF and ATM covered 97.55% of TCGA population and 83.02% validation patients, which were proved to be associated with the development of CRC and could be used as a diagnostic signature. Importantly, mutations of TP53, PIK3CA, FAT4, FMN2 and TRRAP had a remarkable difference between early (I/II) and advanced (III/IV) stage patients (P &lt; 0.0001). Besides, we also confirmed that PIK3CA, LRP1B, FAT4 and ROS1 were a mutated gene signature for the prognosis, LRP1B portended to a higher of recurrence and shorter progression-free survival (PFS); mutation of FAT4 portended to a lower of recurrence and longer PFS, which could predict the recurrence and survival of CRC patients. Conclusions: We have indicated gene mutation signatures related to the diagnosis, pathological classification, staging and prognosis in CRC, which provides further insights into the study of CRC genotype. It is helpful to further improve the personalized diagnosis and treatment.

Frequent KRAS mutations in oncocytic papillary renal neoplasm with inverted nuclei

Papillary renal neoplasm with reverse polarity (PRNRP) is a newly documented rare tumour type. Its molecular pathological features have thus far been very little studied. There were 13 PRNRP cases including 3 The Cancer Genome Atlas (TCGA) cases and our 10 cases in this study. The 3 TCGA cases were found by a combined analysis of GATA3 mRNA expression levels and digital slides from the TCGA papillary renal cell carcinoma project. KRAS codon 12 mutations were identified in the three PRNRPs from TCGA. Of our 10 PRNRP cases, the mutations were also discovered using Sanger sequencing in seven (77.8%) of nine cases with available DNA, where KRAS p.G12V (n=3), p.G12D (n=2), p.G12R (n=1) and p.G12C (n=1) alterations were found. PRNRP shared similar gene expression profiles with renal distal tubules via an interprofile correlation analysis. Gene set enrichment analysis revealed that genes involved in 'KEGG aldosterone regulated sodium reabsorption' or 'hallmark apical surface' were enriched in PRNRP. Moreover, polarised immunostaining patterns for L1CAM and EMA in the distal tubule were maintained in PRNRP. These results imply that the tumour potentially originates from the distal tubule, especially from the cortical collecting duct, and probably retains its cell polarity, except for nuclear inversion. We therefore propose that oncocytic papillary renal neoplasm with inverted nuclei (OPRNIN) is a better name for this tumour type. OPRNIN is a kidney site-specific KRAS mutation neoplasm different from conventional papillary renal cell carcinoma.

A Genomic Classifier for Predicting Clinically Aggressive Luminal Bladder Tumors with Higher Rates of Pathological Up Staging.

Urothelial carcinoma of the luminal molecular subtype is associated with lower rates of pathological up staging from clinical stage T1-T2 to nonorgan confined (pT3 or greater and/or pN+) disease at radical cystectomy. However, approximately a third of luminal urothelial carcinoma cases were up staged to nonorgan confined disease, and these may be under treated if neoadjuvant chemotherapy is withheld. In this study we trained a genomic classifier to predict luminal nonorgan confined disease in patients diagnosed with clinically organ confined (cT1/T2) disease. Specimens from transurethral resected high grade cT1-T2N0M0 urothelial carcinoma of the bladder that belonged to the luminal subtype (Seiler 2017) were randomly split into training (75) and testing (25) sets for the development of a single sample luminal up staging classifier using lasso/ridge-penalized logistic regression. All patients underwent radical cystectomy without neoadjuvant chemotherapy and the primary end point was up staging to nonorgan confined disease. A radical cystectomy cohort and a platinum treated neoadjuvant chemotherapy cohort were used to evaluate the luminal up staging classifier. Up staging to nonorgan confined disease occurred in 34% of luminal cases. The luminal up staging classifier predicted up staging in 32 of 34 cases, with 6 false-positives (AUC 0.96). The sensitivity for detection of luminal pN+ disease was 95% (20 of 21). Patients with predicted nonorgan confined luminal tumors had worse survival than those with organ confined luminal tumors (p=0.001). On multivariable analysis the luminal up staging classifier was a significant predictor of overall survival after adjusting for clinical variables available at transurethral resection. The luminal up staging classifier also predicted overall survival for aggressive luminal TCGA (The Cancer Genome Atlas) cases (n=83, p=0.043). In the neoadjuvant chemotherapy cohort the luminal up staging classifier predicted 9 up staging cases, all of which had excellent prognosis. A luminal up staging classifier was developed that distinguishes a subset of cT1-T2N0M0 luminal urothelial carcinoma cases at high risk for up staging to nonorgan confined disease at radical cystectomy and of death. Validation of this model in an independent, large patient cohort is necessary to determine how molecular stratification of luminal tumors could be used to guide treatment of these patients.

PATH-02. A COMBINATION OF MGMT METHYLATION AND NFKBIA COPY NUMBER ALTERATION REFINES PROGNOSTICATION OF IDH-WT GLIOBLASTOMAS

Abstract INTRODUCTION Recent studies have reported that NFKBIA deletion (dNFKBIA) was potentially associated with worse prognosis in glioblastoma (GBM) patients. However, no consensus has been reached to its universal prognostic value. Here, we investigated the survival impact of dNFKBIA using two primary IDH wild-type GBM cohorts: an original Japanese cohort and a dataset from The Cancer Genome Atlas (TCGA). Additionally, prognostic impact of a combination of NFKBIA copy number and MGMT methylation status was evaluated. METHOD The Japanese cohort was collected from cases registered in Kansai Molecular Diagnosis Network for CNS tumors (KNBTG). The survival impact of dNFKBIA and/or unmethylated MGMT (uMGMT) were analyzed for 212 KNBTG cases and 265 TCGA cases. The hazard ratio (HR) and p-value were computed using Cox regression analysis. RESULTS dNFKBIA was less frequently observed in KNBTG (47 cases, 22.2%) than in TCGA (84 cases, 31.7%). dNFKBIA was associated with unfavorable prognosis in KNBTG (HR 1.52, p = 0.031), while this was not validated in TCGA (HR 1.14, p=0.406). uMGMT was a common adverse prognostic factor in KNBGT (HR 1.72, p = 0.001) and TCGA (HR 1.50, p = 0.008) cohort. When stratified by NFKBIA status, uMGMT was also associated with shorter survival in NFKBIA deleted cases both in KNBTG (HR 1.87, p = 0.002) and TCGA (HR 1.59, p = 0.014). On the other hand, MGMT status was not significantly associated with prognosis in NFKBIA intact cases in either KNBTG (HR 1.45, p = 0.279) or TCGA (HR 1.55, p = 0.131). DISCUSSION Although the prognostic value of dNFKBIA in IDH wild-type GBM patients was not validated in TCGA cohort, our results indicated that the prognostication based on MGMT methylation was potentially interacted by NFKBIA status.