Gene Expression In Normal Tissues Research Articles

Notable Histologic Findings in a "Normal" Cohort: The National Institutes of Health Genotype-Tissue Expression (GTEx) Project

The National Institutes of Health (NIH) Genotype-Tissue Expression (GTEx) project was designed to evaluate how genetic variation and epigenetic effects influence gene expression in normal tissue. To ensure that the grossly normal-appearing tissues collected were free from disease, each specimen underwent histologic evaluation. In total, nearly 30 000 tissue aliquots collected from almost 1000 postmortem donors underwent histologic review by project pathologists, and detailed observations of any abnormalities or lesions present were recorded. Despite sampling of normal-appearing tissue, in-depth review revealed incidental findings among GTEx samples that included neoplastic, autoimmune, and genetic conditions; the incidence of some of these conditions among GTEx donors differed from those previously reported for other populations. A number of age-related abnormalities observed during histologic review of tissue specimens are also described. Histologic findings from the GTEx project may serve to improve populational awareness of several conditions and present a unique opportunity for others to explore age- and gender-influenced conditions. Resources from the study, including histologic image and sequencing data, are publicly available for research.

Uniformly shaped harmonization combines human transcriptomic data from different platforms while retaining their biological properties and differential gene expression patterns.

Introduction: Co-normalization of RNA profiles obtained using different experimental platforms and protocols opens avenue for comprehensive comparison of relevant features like differentially expressed genes associated with disease. Currently, most of bioinformatic tools enable normalization in a flexible format that depends on the individual datasets under analysis. Thus, the output data of such normalizations will be poorly compatible with each other. Recently we proposed a new approach to gene expression data normalization termed Shambhala which returns harmonized data in a uniform shape, where every expression profile is transformed into a pre-defined universal format. We previously showed that following shambhalization of human RNA profiles, overall tissue-specific clustering features are strongly retained while platform-specific clustering is dramatically reduced. Methods: Here, we tested Shambhala performance in retention of fold-change gene expression features and other functional characteristics of gene clusters such as pathway activation levels and predicted cancer drug activity scores. Results: Using 6,793 cancer and 11,135 normal tissue gene expression profiles from the literature and experimental datasets, we applied twelve performance criteria for different versions of Shambhala and other methods of transcriptomic harmonization with flexible output data format. Such criteria dealt with the biological type classifiers, hierarchical clustering, correlation/regression properties, stability of drug efficiency scores, and data quality for using machine learning classifiers. Discussion: Shambhala-2 harmonizer demonstrated the best results with the close to 1 correlation and linear regression coefficients for the comparison of training vs validation datasets and more than two times lesser instability for calculation of drug efficiency scores compared to other methods.

348. IDENTIFYING NOVEL MOLECULAR SUBTYPES OF ESOPHAGEAL ADENOCARCINOMA USING LASER CAPTURE MICRODISSECTED RNA-SEQ SAMPLES

Abstract Background Treatment options for esophageal adenocarcinoma (EAC) are limited by a lack of disease stratification methods. In other cancers, gene expression profiling has been successfully used to identify prognostically-relevant and treatment-susceptible molecular subtypes. Previous efforts to subtype EAC have been hindered by low biopsy and resection tumour cellularity. We hypothesize that laser-capture microdissection (LCM) tumour cell enrichment will permit the classification EAC tumours into multiple molecular subtypes that will be important for future therapeutic strategies. Methods Treatment naïve patient samples (N = 52) were collected from primary biopsies (N = 37), resections (N = 10) and metastatic biopsies (N = 5). Samples were laser-capture microdissected to enrich tumour cells followed by total RNA-seq. Gene expression was quantified using salmon and non-negative matrix factorization with 10 components was used to identify gene expression programs. The components identified were validated on a publicly available normal tissue cohort (N = 13 esophagus, gastric, intestinal, Barrett’s samples) and the TCGA EAC cohort (N = 80). We applied NMF with 4 components to the normal tissue (K = cohort) to identify normal tissue gene signatures to validate our RNA-seq data. Results We verified that our RNA-seq samples were depleted for normal tissue using the four NMF signatures from normal tissue cohort and, as expected, our samples were depleted for gene expression from the gastric and esophagus tissues compared to TCGA. Next, we explored the gene expression programs from our tumour samples and identified seven tumour-intrinsic components and three tumour microenvironmental components. The top 50 genes from each tumour intrinsic components were used with consensus cluster plus to identify K = 6 subtypes (Figure 1). Interestingly, two of these subtypes would be difficult to separate from normal tissue contamination without the use of LCM. Conclusion We have successfully used LCM to deplete normal tissue gene expression from our esophageal adenocarcinoma cohort and identified six molecular subtypes. We are currently evaluating the treatment and clinical implications of these subtypes and aiming to accumulate an additional (N = 50) EAC RNA-seq samples.

Abstract 3136: HYENA detects non-coding genes activated by distal enhancers in cancer

Abstract Genome instability is a hallmark of cancer, resulting in the accumulation of mutation. Among different types of mutations, structural variants (SVs) widely spread in cancer genomes and can be important cancer drivers, demonstrated in multiple cancer types. SVs can induce oncogene activation by redirecting enhancers to promoter regions or disrupting chromatin topology. These events are called “enhancer hijacking” and known to upregulate oncogenes in multiple cancers. Enhancers are among the best understood cis-regulatory elements and are essential regulators for precise gene expression in normal tissue. Since SV associated gene deregulation has been shown in multiple cancer types, it is a feasible way to infer putative oncogenes by detecting enhancer hijacking events. Taking advantage of next generation sequencing, bioinformatic tools were developed to study pan-histology protein-coding gene deregulation caused by SVs. However, algorithms that can unbiasedly detect enhancer hijacking in both coding and non-coding genes are yet to be developed. Here we present a computational algorithm “HYENA” to identify putative oncogenes activated by enhancer hijacking using whole-genome sequencing and RNA sequencing data. It deploys rank-based regression model to test whether an SV breakpoint located close to a gene significantly increases the gene expression. Permutation tests were used to generate a false discovery rate for each gene. By benchmarking with existing tools serving similar purposes, we confirmed that HYENA has superior capabilities to infer enhancer hijacking events. We detected 192 oncogene candidates in 1,148 tumor samples from 25 cancer types. While confirming the known cases of enhancer hijacking, we also identified numerous novel oncogene candidates, of which a substantial proportion was non-coding genes. Specifically, we found that a long non-coding RNA TOB1-AS1 was activated by rearranged enhancers in 10% of pancreatic cancer samples. Furthermore, experimental validation with pancreatic cancer cell lines and immunodeficient mice indicated that high expression of TOB1-AS1 is capable to promote cancer cell invasion and tumor metastasis, suggesting it is a novel cancer driver. As the existing treatments for pancreatic cancer are largely inefficacious and patients still have poor prognosis, new targets are urgently needed to improve pancreatic cancer treatments. Although the molecular mechanisms underlying TOB1-AS1 promoted metastasis remain to be unveiled, this non-coding gene can be a promising target in the patients carrying this enhancer hijacking event. In summary, our results demonstrated that HYENA is a sensitive and reliable tool to infer enhancer hijacking activated oncogenes and spot new potential targets to treat cancer. Citation Format: Anqi Yu, Ali E. Yesilkanal, Ashish Thakur, Fan Wang, Xiaoyang Wu, Alexander Muir, Xin He, Francois Spitz, Lixing Yang. HYENA detects non-coding genes activated by distal enhancers in cancer [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 3136.

Methylation Heterogeneity and Gene Expression of SPG20 in Solid Tumors

Introduction. The downregulation of the Spastic Paraplegia-20 (SPG20) gene is correlated with a rare autosomal recessive disorder called Troyer Syndrome. Only in recent years has SPG20 been studied and partially characterized in cancer. SPG20 has been shown to be hypermethylated in colorectal cancer, gastric cancer, non-Hodgkin’s lymphoma and hepatocellular carcinoma. In this study, we analyze the methylation status and the gene expression of SPG20 in different tumors of various histological origins. Methods. We analyzed the data generated through Infinium Human Methylation 450 BeadChip arrays and RNA-seq approaches extrapolated from The Cancer Genome Atlas (TCGA) database. The statistics were performed with R 4.0.4. Results. We aimed to assess whether the hypermethylation of this target gene was a common characteristic among different tumors and if there was a correlation between the m-values and the gene expression in paired tumor versus solid tissue normal. Overall, our analysis highlighted that SPG20 open sea upstream the TSS is altogether hypermethylated, and the tumor tissues display a higher methylation heterogeneity compared to the solid tissue normal. The gene expression evidences a reproducible, higher gene expression in normal tissues. Conclusion. Our research, based on data mining from TCGA, evidences that colon and liver tumors display a consistent methylation heterogeneity compared to their normal counterparts. This parallels a downregulation of SPG20 gene expression in tumor samples and suggests a role for this multifunctional protein in the control of tumor progression.

3'untranslated regions of tumor suppressor genes evolved specific features to favor cancer resistance.

Cancer-related genes have evolved specific genetic and genomic features to favor tumor suppression. Previously we reported that tumor suppressor genes (TSGs) acquired high promoter CpG dinucleotide frequencies during evolution to maintain high expression in normal tissues and resist cancer-specific downregulation. In this study, we investigated whether 3'untranslated regions (3'UTRs) of TSGs have evolved specific features to carry out similar functions. We found that 3'UTRs of TSGs, especially those involved in multiple histological types and pediatric cancers, are longer than those of non-cancer genes. 3'UTRs of TSGs also exhibit higher density of binding sites for RNA-binding proteins (RBPs), particularly those having high affinities to C-rich motifs. Both longer 3'UTR length and RBP binding sites enrichment are correlated with higher gene expression in normal tissues across tissue types. Moreover, both features together with the correlated N6-methyladenosine modification and the extent of protein-protein interactions are positively associated with the ability of TSGs to resist cancer-specific downregulation. These results were successfully validated with independent datasets. Collectively, these findings indicate that TSGs have evolved longer 3'UTR with increased propensity to RBP binding, N6-methyladenosine modification and protein-protein interactions for optimizing their tumor-suppressing functions.

Heat Shock Proteins HSPA1 and HSP90AA1 Are Upregulated in Colorectal Polyps and Can Be Targeted in Cancer Cells by Anti-Inflammatory Oxicams with Arylpiperazine Pharmacophore and Benzoyl Moiety Substitutions at Thiazine Ring.

Heat shock proteins HSPA1/Hsp70α and HSP90AA1/Hsp90α are crucial for cancer growth but their expression pattern in colorectal polyps or whether they can be modulated by oxicams is unknown. We quantified (RTqPCR) HSPA1 and HSP90AA1 expression in 50 polyp-normal pairs in relation to polyp malignancy potential and examined the effect of piroxicam, meloxicam and five novel analogues on HSPA1 and HSP90AA1 expression (mRNA/protein) in colorectal adenocarcinoma lines. HSPA1 and HSP90AA1 were upregulated in polyps by 3- and 2.9-fold. Expression ratios were higher in polyps with higher dysplasia grade and dominant villous growth pattern, mostly a result of diminished gene expression in normal tissue. Classic oxicams had negligible/non-significant effect on HSP expression. Their most effective analogue inhibited HSPA1 protein and gene by 2.5-fold and 5.7-fold in Caco-2 and by 11.5-fold and 6.8-fold in HCT116 and HSPA1 protein in HT-29 by 1.9-fold. It downregulated HSP90AA1 protein and gene by 1.9-fold and 3.7-fold in Caco-2 and by 2-fold and 5.0-fold in HCT116. HSPA1 and HSP90AA1 are upregulated in colorectal polyps reflecting their potential for malignancy. HSPA1 in cancer cells and, to lesser degree, HSP90AA1 can be reduced by oxicam analogues with thiazine ring substituted via propylene linker by arylpiperazine pharmacophore with fluorine substituents and by benzoyl moiety.

Abstract 1860: Bispecific claudin-6 x CD3 antibodies in a 2 + 1 format demonstrate selectivity and activity on human ovarian cancer cells

Abstract There is a large unmet need for targeted therapies to treat ovarian cancer and other solid tumors. A promising strategy is the use of T cell engaging bispecific antibodies that recruit T cells to kill cancer cells by simultaneously binding to tumor-associated antigens (TAA) on cancer cells and CD3 on T cells. Effective and safe use of these therapeutics depends on selective targeting of cancer cells over normal tissue, feasible when a TAA is more highly expressed on cancer cells versus normal tissues. Selectivity can also be improved by using a mixed valency “2 + 1” bispecific format, coupled with target affinity tuning, to achieve avid binding to the TAA. Claudin-6 (CLDN6) is a tetraspan membrane protein, a member of the claudin family of tight junction proteins, and has been identified as a promising TAA for ovarian cancer due to its high expression on ovarian and other cancer tissues and low expression on normal tissue. However, a complicating factor is that many proteins in the claudin family have high sequence identity. Most similar to CLDN6 is CLDN9, and the two proteins differ at only 3/76 residues in their extracellular loops, creating a challenge for the development of highly selective CLDN6 antibodies. We analyzed CLDN6 and CLDN9 expression using a combination of normal tissue gene expression data (GTEx), cancer cell genomics data (TCGA), and immunohistochemistry (IHC) on a panel of cancer and normal tissues. Data confirmed the tumor-specific expression pattern of CLDN6 but indicated high CLDN9 expression in normal tissues, suggesting that strong antibody selectivity for CLDN6 versus CLDN9 is critical. We created CLDN6-selective T cell engaging bispecific antibodies by starting with a highly selective monoclonal antibody humanized from a mouse hybridoma. Selectivity was further improved by engineering the CDR regions. Variant antibodies were screened for binding to transiently transfected HEK293E cells that highly expressed either CLDN6, CLDN9, or other homologous claudin family members. Selective variants were converted into the XmAb® 2 + 1 bispecific antibody format and tested in T cell-dependent cellular cytotoxicity (TDCC) assays using cancer cell lines expressing CLDN6, or HEK293E cells transfected with claudin homologs. A set of lead CLDN6 x CD3 bispecifics displaying a range of potencies on CLDN6+ cancer cell lines and high selectivity for CLDN6 over CLDN9 and other homologous claudins was identified. Tolerability and pharmacokinetics of these molecules are currently being assessed in non-human primates, and activity will be evaluated in humanized mouse models of ovarian cancer. Citation Format: Matthew S. Faber, Sung-Hyung Lee, Yoon Kyung Kim, Jing Qi, Kendra N. Avery, Duc-Hanh T. Nguyen, Rumana Rashid, Araz Eivazi, Seung Y. Chu, Juan E. Diaz, Connie Ardila, Ruschelle Love, Alex Nisthal, Norman J. Barlow, Christine Bonzon, Umesh S. Muchhal, Matthew J. Bernett, John R. Desjarlais. Bispecific claudin-6 x CD3 antibodies in a 2 + 1 format demonstrate selectivity and activity on human ovarian cancer cells [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 1860.

Abstract 2226: Prioritizing tumor types for clinical study of novel Sec61 inhibitors by searching for expression profiles of sensitive cell lines in tumor sample databases

Abstract Blockade of secreted and transmembrane protein translocation into the endoplasmic reticulum represents a potentially multifactorial approach to inhibiting tumor growth. We have shown that small molecule inhibitors of the Sec61 translocon can inhibit expression of autocrine growth factors, receptor oncogenes, and immune checkpoints. We nominated KZR-261 for clinical development and describe here bioinformatic analysis of in vitro cancer cell line screens and primary tumor samples that prioritizes potentially sensitive tumor types for clinical evaluation. Anti-tumor activity of KZR-834, a structurally related analog of KZR-261, was assessed in two cell line cytotoxicity screens totaling over 400 cell lines. Broad anti-cancer activity across tumor types was observed in both screens, with median IC50 of 0.39 µM and 0.15 µM. We next demonstrated that KZR-261 has equivalent activity to KZR-834 by testing it on a subset of 48 cell lines (R2=0.91). For cell lines with RNA-Seq gene expression profiles available in CCLE (N=232), we correlated sensitivity to KZR-834 with each gene's expression. This analysis revealed a modest correlation between expression and cell line sensitivity for many genes. However, even the gene with the strongest correlation between expression and sensitivity provided only a partial explanation for KZR-834 sensitivity (CCND2, Spearman ρ = 0.31, P-val = 10−4). Levels of Sec61 were not associated with cell line sensitivity (SEC61A1, ρ = -0.05, P-val = 0.57). We therefore surveyed the gene-level results for enrichment of over 7,000 modules by performing fGSEA analysis. Enriched modules were hierarchically clustered revealing 6 primary clusters associated with sensitivity. The 6 clusters highlighted processes such as transcription, translation, metabolism, golgi vesicle transport, as well as genes encoding proteins that are secreted or plasma membrane localized. While fGSEA did not identify basal levels of ER stress as a process correlated with sensitivity, cell lines treated with KZR-834 have consistently shown activation of ER stress pathways at the mRNA and protein level, which led to its inclusion in subsequent analyses. Modules representing these 7 processes were then utilized to search tumor and normal tissue gene expression data from TCGA, TARGET and GTEx. Tumor types such as ALL, DLBCL, uveal and cutaneous melanoma, and mesothelioma were more enriched for expression of these modules than other tumor types and the vast majority of normal tissues. Our novel small molecule inhibitors of Sec61 dependent protein translocation exhibit broad anti-tumor activity in vitro. Large cell line screens and bioinformatic analyses identifying tumor intrinsic sensitivities led to a prioritized list of potentially sensitive tumor types that has guided tumor type selection for our initial Phase 1 clinical trial of KZR-261 in solid tumors. Citation Format: Eric Lowe, R. Andrea Fan, Henry W.B. Johnson, Christopher J. Kirk, Dustin McMinn, Yu Qian, Brian Tuch. Prioritizing tumor types for clinical study of novel Sec61 inhibitors by searching for expression profiles of sensitive cell lines in tumor sample databases [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 2226.

Genome-Wide Transcriptomic Analysis of Non-Tumorigenic Tissues Reveals Aging-Related Prognostic Markers and Drug Targets in Renal Cell Carcinoma.

Simple SummaryFew studies have employed expression data of normal tissues to explore the survival of cancer patients. Hence, we identified aging-related genes and microRNAs in normal tissues from patients with various types of cancer and investigated their effects on survival. We built statistical models to predict the survival of cancer patients using the expression levels of aging-related genes and microRNAs. We found that five aging-related genes (DUSP22, MAPK14, MAPKAPK3, STAT1, and VCP) from kidney cancer patients were significantly related to patient survival. We experimentally showed that there is an increase in the five aging-related genes in the macrophages of aged mice and, among these five genes, the invasion of cancer cells is inhibited by downregulating DUSP22. Using the zebrafish model of tumor cell dissemination, it was also confirmed that in vivo treatment with a small molecule inhibitor of DUSP22 blocked the metastasis of cancer cells in the early stage.The relationship between expression of aging-related genes in normal tissues and cancer patient survival has not been assessed. We developed a genome-wide transcriptomic analysis approach for normal tissues adjacent to the tumor to identify aging-related transcripts associated with survival outcome, and applied it to 12 cancer types. As a result, five aging-related genes (DUSP22, MAPK14, MAPKAPK3, STAT1, and VCP) in normal tissues were found to be significantly associated with a worse survival outcome in patients with renal cell carcinoma (RCC). This computational approach was investigated using nontumorigenic immune cells purified from young and aged mice. Aged immune cells showed upregulated expression of all five aging-related genes and promoted RCC invasion compared to young immune cells. Further studies revealed DUSP22 as a regulator and druggable target of metastasis. DUSP22 gene knockdown reduced RCC invasion and the small molecule inhibitor BML-260 prevented RCC dissemination in a tumor/immune cell xenograft model. Overall, these results demonstrate that deciphering the relationship between aging-related gene expression in normal tissues and cancer patient survival can provide new prognostic markers, regulators of tumorigenesis and novel targets for drug development.

LMNA functions as an oncogene in hepatocellular carcinoma by regulating the proliferation and migration ability.

The role of the LMNA gene in the development and progression of hepatocellular carcinoma (HCC) and the associated molecular mechanism is not yet clear. Therefore, the purpose of this study was to evaluate the relationship between LMNA and HCC. LMNA gene expression in normal tissues and corresponding tumours was evaluated and the Kaplan–Meier survival analysis was performed. Next, the LMNA gene was knocked out in the 293T and HepG2 cell lines using the CRISPR/Cas9 technique. Subsequently, the proliferation, migration and colony formation rate of the two LMNA knockout cell lines were analysed. Finally, the molecular mechanism affecting the tumorigenesis due to the loss of the LMNA gene was evaluated. The results showed that the LMNA gene was abnormally expressed in many tumours, and the survival rate of the HCC patients with a high expression of the LMNA gene was significantly reduced compared with the rate in patients with a low LMNA expression. The knockout of the LMNA gene in the HCC cell line HepG2 resulted in a decreased tumorigenicity, up‐regulation of the P16 expression and down‐regulation of the CDK1 expression. These findings suggested that LMNA might function as an oncogene in HCC and provided a potential new target for the diagnosis and treatment of HCC.

Identification and prediction of novel non-coding and coding RNA-associated competing endogenous RNA networks in colorectal cancer.

AIMTo identify and predict the competing endogenous RNA (ceRNA) networks in colorectal cancer (CRC) by bioinformatics analysis.METHODSIn the present study, we obtained CRC tissue and normal tissue gene expression profiles from The Cancer Genome Atlas project. Differentially expressed (DE) genes (DEGs) were identified. Then, upregulated and downregulated miRNA-centered ceRNA networks were constructed by analyzing the DEGs using multiple bioinformatics approaches. DEmRNAs in the ceRNA networks were identified in Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways using KEGG Orthology Based Annotation System 3.0. The interactions between proteins were analyzed using the STRING database. Kaplan-Meier survival analysis was conducted for DEGs and real time quantitative polymerase chain reaction (RT-qPCR) was also performed to validate the prognosis-associated lncRNAs in CRC cell lines.RESULTSEighty-one DElncRNAs, 20 DEmiRNAs, and 54 DEmRNAs were identified to construct the ceRNA networks of CRC. The KEGG pathway analysis indicated that nine out of top ten pathways were related with cancer and the most significant pathway was “colorectal cancer”. Kaplan-Meier survival analysis showed that the overall survival was positively associated with five DEGs (IGF2-AS, POU6F2-AS2, hsa-miR-32, hsa-miR-141, and SERPINE1) and it was negatively related to three DEGs (LINC00488, hsa-miR-375, and PHLPP2). Based on the STRING protein database, it was found that SERPINE1 and PHLPP2 interact with AKT1. Besides, SERPINE1 can interact with VEGFA, VTN, TGFB1, PLAU, PLAUR, PLG, and PLAT. PHLPP2 can interact with AKT2 and AKT3. RT-qPCR revealed that the expression of IGF2-AS, POU6F2-AS2, and LINC00488 in CRC cell lines was consistent with the in silico results.CONCLUSIONCeRNA networks play an important role in CRC. Multiple DEGs are related with clinical prognosis, suggesting that they may be potential targets in tumor diagnosis and treatment.

Cancer-specific expression quantitative loci are affected by expression dysregulation.

Expression quantitative trait loci (eQTLs) have been touted as the missing piece that can bridge the gap between genetic variants and phenotypes. Over the past decade, we have witnessed a sharp rise of effort in the identification and application of eQTLs. The successful application of eQTLs relies heavily on their reproducibility. The current eQTL databases such as Genotype-Tissue Expression (GTEx) were populated primarily with eQTLs deriving from germline single nucleotide polymorphisms and normal tissue gene expression. The novel scenarios that employ eQTL models for prediction purposes often involve disease phenotypes characterized by altered gene expressions. To evaluate eQTL reproducibility across diverse data sources and the effect of disease-specific gene expression alteration on eQTL identification, we conducted an eQTL study using 5178 samples from The Cancer Genome Atlas (TCGA). We found that the reproducibility of eQTLs between normal and tumor tissues was low in terms of the number of shared eQTLs. However, among the shared eQTLs, the effect directions were generally concordant. This suggests that the source of the gene expression (normal or tumor tissue) has a strong effect on the detectable eQTLs and the effect direction of the eQTLs. Additional analyses demonstrated good directional concordance of eQTLs between GTEx and TCGA. Furthermore, we found that multi-tissue eQTLs may exert opposite effects across multiple tissue types. In summary, our results suggest that eQTL prediction models need to carefully address tissue and disease dependency of eQTLs. Tissue-disease-specific eQTL databases can afford more accurate prediction models for future studies.

Integrated transcriptomic analysis of distance-related field cancerization in rectal cancer patients

Field cancerization (FC) occurs in various epithelial carcinomas, including colorectal cancer, which indicates that the molecular events in carcinogenesis might occur in normal tissues extending from tumors. However, the transcriptomic characteristics of FC in colorectal cancer (CRC) remain largely unexplored. To investigate the changes in gene expression associated with proximity to the tumor, we analyzed the global gene expression profiles of cancer tissues and histologically normal tissues taken at various distances from the tumor (1 cm, 5 cm and the proximal end of the resected sample) from 32 rectal cancer patients. Significantly differentially expressed genes related to the distance from the tumor were screened by linear mixed effects analysis using the lme4 package in R. The distance-related differentially expressed genes that were gradually up-regulated (n=302) or gradually down-regulated (n=568) from normal tissues to the tumor were used to construct protein-protein interaction (PPI) networks. Three subnetworks among the gradually up-regulated genes and four subnetworks among the gradually down-regulated genes were identified using the MCODE plugin in the Cytoscape software program. The most significantly enriched Gene Ontology (GO) biological process terms were “ribosome biogenesis”, “mRNA splicing via spliceosome”, and “positive regulation of leukocyte migration” for the gradually up-regulated subnetworks and “cellular calcium ion homeostasis”, “cell separation after cytokinesis”, “cell junction assembly”, and “fatty acid metabolic process” for the gradually down-regulated subnetworks. Combined with the previously constructed multistep carcinogenesis model used for the analysis, 50.59% of the genes in the subnetworks (43/85) displayed identical changes in expression from normal colon tissues to adenoma and colon cancer. We focused on the 7 genes associated with fatty acid metabolic processes in the distance-related down-regulated subnetwork. Survival analysis of patients in the CRC dataset from The Cancer Genome Atlas (TCGA) revealed that higher expression of these 7 genes, especially CPT2, ACAA2 and ACADM, was associated with better prognosis (p = 0.034, p = 0.00058, p = 0.039, p = 0.04). Cox proportional hazards regression analysis revealed that CPT2 was an independent prognostic factor (p = 0.004131). Our results demonstrate that field cancerization occurs in CRC and affects gene expression in normal tissues extending from the tumor, which may provide new insights into CRC oncogenesis and patient progression.

Identification of breast cancer recurrence risk factors based on functional pathways in tumor and normal tissues.

The recurrence of breast cancer (BC) is a serious therapeutic problem, and the risk factors for recurrence urgently need to be identified. In this study, we examined the functional pathways in tumor and normal tissues to more comprehensively identify biomarkers for the risk of BC recurrence. We collected tumor and normal tissue gene expression profiles of recurrent BC patients and non-recurrent BC patients from the TCGA database.We derived an expression interval (mean ± 1.96SD) based on non-recurrent patients rather than a single value, such as a mean or median. If the expression of a gene was significantly different from its normal expression interval, it was considered a differentially expressed gene. Eight pathways that significantly distinguished recurrent and non-recurrent BC patients were obtained based on 65% accuracy, and these pathways were all associated with the immune response and sensitivity to drugs. The genes in these eight pathways were also used to analyze survival, and the significance level reached 0.003 in an independent dataset (p = 0.02 in tumor and p = 0.03 in normal tissue). Our results reveal that the integration of tumor and normal tissue functional analyses can comprehensively enhance the understanding of BC prognosis.

Abstract 4442: Very low methylation is a novel epigenetic regulatory mechanism in normal tissues and predisposes to hypermethylation in cancer

Abstract High level DNA methylation of promoter CpG Island (CGI) represses gene expression. However, low-level CGI methylation is usually not distinguished from complete absence of methylation. Here we show that Very Low Methylated Regions (VLMRs 1-20%) on promoter CGIs are present in up to 20% of human and mouse genes and negatively correlate with gene expression in all tissues examined. In vitro, low-level methylation directly represses reporter gene expression; an effect mediated by methyl-CpG-binding proteins as transient knockdown of MBD4 reverses this repression. In vivo VLMR genes are enriched for polycomb occupancy but can also occur in H3K4me3 occupied promoters, where it correlates with gene repression independent of polycomb complex binding. VLMRs in promoter CGIs in healthy WBCs are 19 fold and 65 fold more likely to gain hypermethylation in Acute Myelogenous Leukemia and Myelodysplastic Syndrome patients respectively compared to unmethylated CGIs (<1%). Our study reveals a novel epigenetic control mechanism that has strong effects on gene expression in normal tissues and accounts for most of the gene specificity of hypermethylation in cancer. Citation Format: priyanka madireddi, Jaroslav Jelinek, Justin Lee, Takahiro Sato, Jean-Pierre Issa. Very low methylation is a novel epigenetic regulatory mechanism in normal tissues and predisposes to hypermethylation in cancer. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 4442.

Gene expression in prostate tissue according to frequency of ejaculation.

25 Background: Long-term prospective data from within the Health Professionals Follow-up Study (HPFS) suggest more frequent ejaculation throughout adult life may be protective against future development of prostate cancer (PCa). Compared to men with an average monthly ejaculation frequency (EF) of 4-7 times/month during adulthood, men who ejaculated at least 21 times per month were 30% less likely to be diagnosed with PCa after adjustment for a number of potential confounders and PCa screening. We sought to explore mechanisms through which EF may influence PCa development utilizing gene expression data assessed on tumor and adjacent normal tissue. Methods: In 1992, 31,925 HPFS participants were questioned on their average monthly EF at ages 20-29, 40-49, and in the previous year; 3,839 of these men were later diagnosed with PCa. Expression of 20,254 genes was available from archival tumor tissue from 156 cases; 84 of these cases also had data from adjacent normal tissue. We regressed each gene expression value as a continuous outcome on EF, age at diagnosis, and year at diagnosis. We calculated the p value associated with the coefficient for the EF variable as a measure of the strength of the trend test. We then used t values for the coefficients to rank the genes, and ran a gene set enrichment analysis to identify pathways with genes associated with EF. Results: We observed that while there was not much signal in the tumor tissue, there was enrichment of smaller p values in the adjacent normal tissue. EF in the time period most proximal to diagnosis is most strongly associated with gene expression in normal tissue. Focusing on the association between gene expression in normal tissue and EF in the year prior to the questionnaire (1991), the top six differentially expressed genes were OR4K1, CCDC138, SERPINA9, TRMT61B, ZNF302, and ASB4. Gene set enrichment analysis revealed a substantial number of pathways differentially expressed at the FDR <0.05 threshold. Conclusions: Identifying changes in gene expression associated with EF in normal prostate tissue provides further support for a biological role of ejaculation frequency in the etiology of PCa. Notably, pathways downregulated with increasing ejaculation frequency were primarily immune related.

MERAV: a tool for comparing gene expression across human tissues and cell types.

The oncogenic transformation of normal cells into malignant, rapidly proliferating cells requires major alterations in cell physiology. For example, the transformed cells remodel their metabolic processes to supply the additional demand for cellular building blocks. We have recently demonstrated essential metabolic processes in tumor progression through the development of a methodological analysis of gene expression. Here, we present the Metabolic gEne RApid Visualizer (MERAV, http://merav.wi.mit.edu), a web-based tool that can query a database comprising ∼4300 microarrays, representing human gene expression in normal tissues, cancer cell lines and primary tumors. MERAV has been designed as a powerful tool for whole genome analysis which offers multiple advantages: one can search many genes in parallel; compare gene expression among different tissue types as well as between normal and cancer cells; download raw data; and generate heatmaps; and finally, use its internal statistical tool. Most importantly, MERAV has been designed as a unique tool for analyzing metabolic processes as it includes matrixes specifically focused on metabolic genes and is linked to the Kyoto Encyclopedia of Genes and Genomes pathway search.

Abstract 1592: CXCR4 pathways in CTCs: from bioinformatics to immunophenotype

Abstract Introduction: Bioinformatics’ analysis regarding gene expression in Normal tissue vs cancer tissue, Normal blood vs cancer patients’ blood and Normal blood vs cancer tissue revealed a subgroup of 24 genes which potentially could be overexpressed in CTCs. Among these genes are: CXCR4, a chemokine receptor which is involved in tumor metastasis and, JUNB a transcription factor participating in CXCR4 pathway. The objective of this study was to confirm the results of bioinformatics’ analysis in samples from breast cancer patients and therefore to evaluate the expression of CXCR4 and JUNB in CTCs. Methods: Triple staining immunofluoresence with panytokeratin/CXCR4/JUNB antibodies were performed in SKBR3, MDA-MB231, MCF7 and Hela cell lines. The same experiments were performed in PBMCs from normal (n = 10.) subjects and in PBMCs (n = 55) from untreated metastatic breast cancer patients. Results: Since CXCR4, as a chemokine receptor is also expressed on hematopoietic cells, its expression on CTCs was quantified using the ARIOL system. We used cell lines and both normal vs patients’ PBMCs in order to establish an expression pattern of both molecules. Quantification experiments among cell lines revealed that CXCR4 was overexpressed in SKBR3 following the subsequent hierarchy SKBR3>MCF7>MDA-MB231>Hela. Accordingly, the expression pattern of JUNB in cell lines was SKBR3>Hela>MCF7>MDA-MB231 with higher expression in SKBR3 and lower in MDA-MB 231. Statistical analysis revealed significant differences in the expression of both molecules between healthy donors’ PBMCs and patient's PBMCs. Subsequently, CTCs were detected in 17 out of 55 screened patients (31%). Mean intensity of CTCs for CXCR4 was 34.38 and it was higher than Hela and lower than SKBR3. Patients with CXCR4-positive CTCs (with mean expression higher than 95% of normal PBMCs) were 53%. In addition 84% of the examined CTCs had expression profile higher than normal PBMCs. JUNB expression in CTCs (above 95% of normal PBMCs) was identified in 76.92% of patients. JUNB mean expression in CTCs was 26.1 and it was lower than SKBR3 and higher than MCF7. Furthermore 64.3% of the total CTCs have expression higher than Normal PBMCs. Conclusion: CXCR4 and JUNB are highly expressed in CTCs derived from breast cancer patients, in agreement to bioinformatics’ analysis. Quantification of immunofluoresence potentially delineates a subgroup of patients with high expression of CXCR4 and JUNB that could benefit from target therapies. Citation Format: G. Kallergi, V. Tsintari, S. Sfakianakis, M. Zervakis, D. Mavroudis, V. Georgoulias. CXCR4 pathways in CTCs: from bioinformatics to immunophenotype. [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr 1592. doi:10.1158/1538-7445.AM2015-1592

Abstract A82: Virtual microdissection reveals tumor specific heterogeneity in pancreatic cancer

Abstract Using matrix factorization, we have removed confounding normal tissue gene expression from profiles of primary and metastatic tumors, facilitating the study of underlying tumor biology without the need for laser capture microdissection. Understanding molecular mechanisms of metastasis in pancreatic cancer has the potential to yield rationally designed therapies. Previous work has used gene expression arrays and next generation sequencing to look for potential targets, focusing on identifying differences between metastatic lesions and primary tumors. However, such analyses are hampered by the low cellularity of malignant epithelium in patient samples. Since profiles of primary and metastatic tumors often contain contributions from adjacent normal tissues from the tumor site, comparison of these samples are confounded by the differences between normal pancreas and normal distant organs. We seek to avoid this pitfall by leveraging the blind source separation technique, nonnegative matrix factorization (NMF)[1], to the analysis of gene expression data. Microarray data were obtained from matched primary (n=8) and metastatic tumors (n=31) as well as adjacent normal tissues (n=85) for both local and distant sites of patients who died of metastatic pancreatic cancer[2]. Pathological analysis showed that tumor cellularity varied with a mean of 48% and a standard deviation of 31% across all sites. We applied NMF to our cohort, thus computationally microdissecting tumors into the source tissues composing our samples. 200 iterations of 5-fold resampling were performed to achieve stable NMF results. Genes with expression ranked in the top 50 of any factor together were recorded in a consensus matrix. This consensus matrix was then used as the basis of a hierarchical clustering as to yield k gene clusters. These k gene-clusters were used to seed a supervised NMF using the MATLAB NMF multiplicative update solver until completion. Using an unbiased approach, we identified distinct molecular signatures associated with adjacent normal and tumor tissue in primary and metastatic pancreatic cancer samples. Specifically, we identified confounding liver, lung, lymph, and muscle tissue gene expression from our metastatic samples as well as both endocrine and exocrine pancreas gene expression from our primary tumors. Furthermore, we estimated the relative composition of each of our samples in terms of a weighted sum of these tissue-specific signatures. Both primary tumors and metastases appeared to be admixtures of pancreatic tumor epithelial expression and adjacent normal tissue. For example, in liver tissues histological assessment of tumor content correlated well with NMF predictions of tumor content, (n=39, R2=0.72, p<0.01). Prior to NMF, unsupervised clustering of all data caused samples to cluster by site of harvest. After digital subtraction of confounding tissue factors identified by NMF, samples grouped by patient of origin, suggesting intrinsic similarity among tumor sites within a patient. We have demonstrated a method to virtually microdissect tissues, thereby identifying tumor-specific gene expression data in pancreatic cancer. By applying a fresh computational approach to a large cohort of data, we can generate new insight into the complex nature of low-cellularity tumors such as pancreatic cancer, and facilitate the study of inter- and intra-patient heterogeneity. This approach may be further leveraged to study the role of tumor and stroma signatures in pancreatic cancer. 1. Brunet, J.P., et al., Metagenes and molecular pattern discovery using matrix factorization. Proceedings of the National Academy of Sciences, 2004. 101(12): p. 4164-4169. 2. Stratford, J.K., et al., A six-gene signature predicts survival of patients with localized pancreatic ductal adenocarcinoma. PLoS medicine, 2010. 7(7): p. e1000307. Citation Format: Richard A. Moffitt, Keith A. Volmar, Judy M. Anderson, Michael A. Hollingsworth, Jen Jen Yeh. Virtual microdissection reveals tumor specific heterogeneity in pancreatic cancer. [abstract]. In: Proceedings of the AACR Special Conference on Pancreatic Cancer: Innovations in Research and Treatment; May 18-21, 2014; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2015;75(13 Suppl):Abstract nr A82.