Prediction of cancer treatment response from histopathology images through imputed transcriptomics.

Advances in artificial intelligence have paved the way for leveraging hematoxylin and eosin (H&E)-stained tumor slides for precision oncology. We present ENLIGHT-DeepPT, an approach for predicting response to multiple targeted and immunotherapies from H&E-slides. In difference from existing approaches that aim to predict treatment response directly from the slides, ENLIGHT-DeepPT is an indirect two-step approach consisting of (1) DeepPT, a new deep-learning framework that predicts genome-wide tumor mRNA expression from slides, and (2) ENLIGHT, which predicts response based on the DeepPT inferred expression values. DeepPT successfully predicts transcriptomics in all 16 TCGA cohorts tested and generalizes well to two independent datasets. Our key contribution is showing that ENLIGHT-DeepPT successfully predicts true responders in five independent patients' cohorts involving four different treatments spanning six cancer types with an overall odds ratio of 2.44, increasing the baseline response rate by 43.47% among predicted responders, without the need for any treatment data for training. Furthermore, its prediction accuracy on these datasets is comparable to a supervised approach predicting the response directly from the images, which needs to be trained and tested on the same cohort. ENLIGHT-DeepPT future application could provide clinicians with rapid treatment recommendations to an array of different therapies and importantly, may contribute to advancing precision oncology in developing countries.

Probing the diabetes and colorectal cancer relationship using gene - environment interaction analyses.

BACKGROUND: Diabetes is an established risk factor for colorectal cancer. However, the mechanisms underlying this relationship still require investigation and it is not known if the association is modified by genetic variants. To address these questions, we undertook a genome-wide gene-environment interaction analysis. METHODS: We used data from 3 genetic consortia (CCFR, CORECT, GECCO; 31,318 colorectal cancer cases/41,499 controls) and undertook genome-wide gene-environment interaction analyses with colorectal cancer risk, including interaction tests of genetics(G)xdiabetes (1-degree of freedom; d.f.) and joint testing of Gxdiabetes, G-colorectal cancer association (2-d.f. joint test) and G-diabetes correlation (3-d.f. joint test). RESULTS: Based on the joint tests, we found that the association of diabetes with colorectal cancer risk is modified by loci on chromosomes 8q24.11 (rs3802177, SLC30A8 - ORAA: 1.62, 95% CI: 1.34-1.96; ORAG: 1.41, 95% CI: 1.30-1.54; ORGG: 1.22, 95% CI: 1.13-1.31; p-value3-d.f.: 5.46 × 10-11) and 13q14.13 (rs9526201, LRCH1 - ORGG: 2.11, 95% CI: 1.56-2.83; ORGA: 1.52, 95% CI: 1.38-1.68; ORAA: 1.13, 95% CI: 1.06-1.21; p-value2-d.f.: 7.84 × 10-09). DISCUSSION: These results suggest that variation in genes related to insulin signaling (SLC30A8) and immune function (LRCH1) may modify the association of diabetes with colorectal cancer risk and provide novel insights into the biology underlying the diabetes and colorectal cancer relationship.

Genome-wide Interaction Study with Smoking for Colorectal Cancer Risk Identifies Novel Genetic Loci Related to Tumor Suppression, Inflammation, and Immune Response.

BACKGROUND: Tobacco smoking is an established risk factor for colorectal cancer. However, genetically defined population subgroups may have increased susceptibility to smoking-related effects on colorectal cancer. METHODS: A genome-wide interaction scan was performed including 33,756 colorectal cancer cases and 44,346 controls from three genetic consortia. RESULTS: Evidence of an interaction was observed between smoking status (ever vs. never smokers) and a locus on 3p12.1 (rs9880919, P = 4.58 × 10-8), with higher associated risk in subjects carrying the GG genotype [OR, 1.25; 95% confidence interval (CI), 1.20-1.30] compared with the other genotypes (OR <1.17 for GA and AA). Among ever smokers, we observed interactions between smoking intensity (increase in 10 cigarettes smoked per day) and two loci on 6p21.33 (rs4151657, P = 1.72 × 10-8) and 8q24.23 (rs7005722, P = 2.88 × 10-8). Subjects carrying the rs4151657 TT genotype showed higher risk (OR, 1.12; 95% CI, 1.09-1.16) compared with the other genotypes (OR <1.06 for TC and CC). Similarly, higher risk was observed among subjects carrying the rs7005722 AA genotype (OR, 1.17; 95% CI, 1.07-1.28) compared with the other genotypes (OR <1.13 for AC and CC). Functional annotation revealed that SNPs in 3p12.1 and 6p21.33 loci were located in regulatory regions, and were associated with expression levels of nearby genes. Genetic models predicting gene expression revealed that smoking parameters were associated with lower colorectal cancer risk with higher expression levels of CADM2 (3p12.1) and ATF6B (6p21.33). CONCLUSIONS: Our study identified novel genetic loci that may modulate the risk for colorectal cancer of smoking status and intensity, linked to tumor suppression and immune response. IMPACT: These findings can guide potential prevention treatments.

Transcriptomic Response to Calcium in Normal Colon Organoids is Impacted by Colon Location and Sex.

Observational studies indicate that calcium supplementation may protect against colorectal cancer. Stratified analyses suggest that this protective effect may differ based on anatomic subsite and sex, but these hypotheses have been difficult to test experimentally. Here, we exposed 36 patient-derived organoid lines derived from normal colon biopsies (21 right colons, 15 left colons) of unrelated subjects (18 female, 18 male) to moderate (1.66 mmol/L) or high (5.0 mmol/L) concentrations of calcium for 72 hours. We performed bulk RNA-sequencing to measure gene expression, and cell composition was inferred using single-cell deconvolution in CIBERSORTx. We tested for significant differences in gene expression using generalized linear models in DESeq2. Exposure to higher levels of calcium was associated with changes in cell composition (P < 0.05), most notably increased goblet and reduced stem cell populations, and differential expression of 485 genes (FDR < 0.05). We found that 40 of these differentially expressed genes mapped to genomic loci identified through colorectal cancer genome-wide association studies, suggesting a potential biologic overlap between calcium supplementation and inherited colorectal cancer risk. Stratified analyses identified more differentially expressed genes in colon organoids derived from right sided colon and male subjects than those derived from left sided colon and female subjects. We confirmed the presence of a stronger right-sided effect for one of these genes, HSD17B2 using qPCR in a subset of matched right and left colon organoids (n = 4). By relating our findings to genetic data, we provide new insights into how nutritional and genetic factors may interact to influence colorectal cancer risk. PREVENTION RELEVANCE: A chemopreventive role for calcium in colorectal cancer is still unclear. Here, we identify mechanisms through which calcium supplementation may reduce risk. Calcium supplementation increased differentiation and altered expression of colorectal cancer-related genes in a large study of patient-derived colon organoids. These findings were influenced by colon location and sex.

DNA methylation analysis of normal colon organoids from familial adenomatous polyposis patients reveals novel insight into colon cancer development.

BACKGROUND: Familial adenomatous polyposis (FAP) is an inherited colorectal cancer (CRC) syndrome resulting from germ line mutations in the adenomatous polyposis coli (APC) gene. While FAP accounts for less than 1% of all CRC cases, loss of APC expression is seen in > 80% of non-hereditary CRCs. To better understand molecular mechanisms underlying APC-driven CRC, we performed an epigenome-wide analysis of colon organoids derived from normal-appearing colons of FAP patients versus healthy subjects to identify differentially methylated regions (DMRs) that may precede the onset of CRC. RESULTS: We identified 358 DMRs when comparing colon organoids of FAP patients to those of healthy subjects (FDR < 0.05, |mean beta difference| = 5%). Of these, nearly 50% of DMRs were also differentially methylated in at least one of three CRC tumor and normal adjacent tissue (NAT) cohorts (TCGA-COAD, GSE193535 and ColoCare). Moreover, 27 of the DMRs mapped to CRC genome-wide association study (GWAS) loci. We provide evidence suggesting that some of these DMRs led to significant differences in gene expression of adjacent genes using quantitative PCR. For example, we identified significantly greater expression of five genes: Kazal-type serine peptidase inhibitor domain 1 (KAZALD1, P = 0.032), F-Box and leucine-rich repeat protein 8 (FBXL8, P = 0.036), TRIM31 antisense RNA 1 (TRIM31-AS1, P = 0.036), Fas apoptotic inhibitory molecule 2 (FAIM2, P = 0.049) and (Collagen beta (1-0)galactosyltransferase 2 (COLGALT2, P = 0.049). Importantly, both FBXL8 and TRIM31-AS1 were also significantly differentially expressed in TCGA-COAD tumor versus matched NAT, supporting a role for these genes in CRC tumor development. CONCLUSIONS: We performed the first DNA methylome-wide analysis of normal colon organoids derived from FAP patients compared to those of healthy subjects. Our results reveal that normal colon organoids from FAP patients exhibit extensive epigenetic differences compared to those of healthy subjects that appear similar to those exhibited in CRC tumor. Our analyses therefore identify DMRs and candidate target genes that are potentially important in CRC tumor development in FAP, with potential implications for non-hereditary CRC.

Beyond GWAS of Colorectal Cancer: Evidence of Interaction with Alcohol Consumption and Putative Causal Variant for the 10q24.2 Region.

BACKGROUND: Currently known associations between common genetic variants and colorectal cancer explain less than half of its heritability of 25%. As alcohol consumption has a J-shape association with colorectal cancer risk, nondrinking and heavy drinking are both risk factors for colorectal cancer. METHODS: Individual-level data was pooled from the Colon Cancer Family Registry, Colorectal Transdisciplinary Study, and Genetics and Epidemiology of Colorectal Cancer Consortium to compare nondrinkers (≤1 g/day) and heavy drinkers (>28 g/day) with light-to-moderate drinkers (1-28 g/day) in GxE analyses. To improve power, we implemented joint 2df and 3df tests and a novel two-step method that modifies the weighted hypothesis testing framework. We prioritized putative causal variants by predicting allelic effects using support vector machine models. RESULTS: For nondrinking as compared with light-to-moderate drinking, the hybrid two-step approach identified 13 significant SNPs with pairwise r2 > 0.9 in the 10q24.2/COX15 region. When stratified by alcohol intake, the A allele of lead SNP rs2300985 has a dose-response increase in risk of colorectal cancer as compared with the G allele in light-to-moderate drinkers [OR for GA genotype = 1.11; 95% confidence interval (CI), 1.06-1.17; OR for AA genotype = 1.22; 95% CI, 1.14-1.31], but not in nondrinkers or heavy drinkers. Among the correlated candidate SNPs in the 10q24.2/COX15 region, rs1318920 was predicted to disrupt an HNF4 transcription factor binding motif. CONCLUSIONS: Our study suggests that the association with colorectal cancer in 10q24.2/COX15 observed in genome-wide association study is strongest in nondrinkers. We also identified rs1318920 as the putative causal regulatory variant for the region. IMPACT: The study identifies multifaceted evidence of a possible functional effect for rs1318920.

Novel insights into the molecular mechanisms underlying risk of colorectal cancer from smoking and red/processed meat carcinogens by modeling exposure in normal colon organoids.

Tobacco smoke and red/processed meats are well-known risk factors for colorectal cancer (CRC). Most research has focused on studies of normal colon biopsies in epidemiologic studies or treatment of CRC cell lines in vitro. These studies are often constrained by challenges with accuracy of self-report data or, in the case of CRC cell lines, small sample sizes and lack of relationship to normal tissue at risk. In an attempt to address some of these limitations, we performed a 24-hour treatment of a representative carcinogens cocktail in 37 independent organoid lines derived from normal colon biopsies. Machine learning algorithms were applied to bulk RNA-sequencing and revealed cellular composition changes in colon organoids. We identified 738 differentially expressed genes in response to carcinogens exposure. Network analysis identified significantly different modules of co-expression, that included genes related to MSI-H tumor biology, and genes previously implicated in CRC through genome-wide association studies. Our study helps to better define the molecular effects of representative carcinogens from smoking and red/processed meat in normal colon epithelial cells and in the etiology of the MSI-H subtype of CRC, and suggests an overlap between molecular mechanisms involved in inherited and environmental CRC risk.

Transcriptome-wide In Vitro Effects of Aspirin on Patient-derived Normal Colon Organoids.

Mechanisms underlying aspirin chemoprevention of colorectal cancer remain unclear. Prior studies have been limited because of the inability of preclinical models to recapitulate human normal colon epithelium or cellular heterogeneity present in mucosal biopsies. To overcome some of these obstacles, we performed in vitro aspirin treatment of colon organoids derived from normal mucosal biopsies to reveal transcriptional networks relevant to aspirin chemoprevention. Colon organoids derived from 38 healthy individuals undergoing endoscopy were treated with 50 µmol/L aspirin or vehicle control for 72 hours and subjected to bulk RNA sequencing. Paired regression analysis using DESeq2 identified differentially expressed genes (DEG) associated with aspirin treatment. Cellular composition was determined using CIBERSORTx. Aspirin treatment was associated with 1,154 significant (q < 0.10) DEGs prior to deconvolution. We provide replication of these findings in an independent population-based RNA-sequencing dataset of mucosal biopsies (BarcUVa-Seq), where a significant enrichment for overlap of DEGs was observed (P < 2.2E-16). Single-cell deconvolution revealed changes in cell composition, including a decrease in transit-amplifying cells following aspirin treatment (P = 0.01). Following deconvolution, DEGs included novel putative targets for aspirin such as TRABD2A (q = 0.055), a negative regulator of Wnt signaling. Weighted gene co-expression network analysis identified 12 significant modules, including two that contained hubs for EGFR and PTGES2, the latter being previously implicated in aspirin chemoprevention. In summary, aspirin treatment of patient-derived colon organoids using physiologically relevant doses resulted in transcriptome-wide changes that reveal altered cell composition and improved understanding of transcriptional pathways, providing novel insight into its chemopreventive properties. PREVENTION RELEVANCE: Numerous studies have highlighted a role for aspirin in colorectal cancer chemoprevention, though the mechanisms driving this association remain unclear. We addressed this by showing that aspirin treatment of normal colon organoids diminished the transit-amplifying cell population, inhibited prostaglandin synthesis, and dysregulated expression of novel genes implicated in colon tumorigenesis.

A Functional Variant on 9p21.3 Related to Glioma Risk Affects Enhancer Activity and Modulates Expression of CDKN2B-AS1.

Genome-wide association studies have identified SNPs associated with glioma risk on 9p21.3, but biological mechanisms underlying this association are unknown. We tested the hypothesis that a functional SNP on 9p21.3 affects activity of an enhancer, causing altered expression of nearby genes. We considered all SNPs in linkage disequilibrium with the 9p21.3 sentinel SNP rs634537 that mapped to putative enhancers. An enhancer containing rs1537372 exhibited allele-specific effects on luciferase activity. Deletion of this enhancer in GBM cell lines correlated with decreased expression of CDKN2B-AS1. Expression quantitative trait loci analysis using non-diseased brain samples showed rs1537372 to be a consistently significant eQTL for CDKN2B-AS1. Additionally, our analysis of Hi-C data generated in neural progenitor cells showed that the bait region containing rs1537372 interacted with the CDKN2B-AS1 promoter. These data suggest rs1537372, a SNP at the 9p21.3 risk locus, is a functional variant that modulates expression of CDKN2B-AS1.

Colon Crypts of Subjects With Familial Adenomatous Polyposis Show an Increased Number of LGR5+ Ectopic Stem Cells.

INTRODUCTION: Familial adenomatous polyposis (FAP) is a hereditary colorectal cancer (CRC) syndrome characterized by accelerated adenoma development due to inherited (or de novo) mutations in the APC regulator of WNT signaling pathway (APC) gene. The mechanism underlying this accelerated polyp development in subjects with FAP has not been defined. Given that LGR5+ stem cells drive crypt cell proliferation, we hypothesized that FAP crypts would demonstrate aberrant leucine-rich repeat-containing G-protein-coupled receptor 5 (LGR5) staining patterns. METHODS: Biopsies were taken from 11 healthy subjects, 7 subjects with Lynch syndrome, 4 subjects with FAP, and 1 subject with MUTYH-associated polyposis syndrome during routine screening or surveillance colonoscopy. Crypt staining was evaluated by immunohistochemistry of paraffin-embedded tissue sections. Stem cell numbers were estimated by immunofluorescence staining of isolated crypts using antibodies against LGR5 and other proteins. RESULTS: Subjects with FAP exhibited a greater number of LGR5+ stem cells in their crypts than healthy subjects and subjects with Lynch syndrome and MUTYH-associated polyposis syndrome. Most crypts of subjects with FAP harbored LGR5+ cells located above the lower third of the crypts. DISCUSSION: These findings support a model in which inactivation of one copy of APC leads to increased numbers of LGR5+ stem cells, many of which are ectopic, in colon crypts of subjects with FAP. Overabundant and ectopic LGR5+ stem cells could lead to an expanded proliferative zone of dividing cells more likely to develop mutations that would contribute to the accelerated adenoma development observed in FAP.

Genetic Effects on Transcriptome Profiles in Colon Epithelium Provide Functional Insights for Genetic Risk Loci.

BACKGROUND & AIMS: The association of genetic variation with tissue-specific gene expression and alternative splicing guides functional characterization of complex trait-associated loci and may suggest novel genes implicated in disease. Here, our aims were as follows: (1) to generate reference profiles of colon mucosa gene expression and alternative splicing and compare them across colon subsites (ascending, transverse, and descending), (2) to identify expression and splicing quantitative trait loci (QTLs), (3) to find traits for which identified QTLs contribute to single-nucleotide polymorphism (SNP)-based heritability, (4) to propose candidate effector genes, and (5) to provide a web-based visualization resource. METHODS: We collected colonic mucosal biopsy specimens from 485 healthy adults and performed bulk RNA sequencing. We performed genome-wide SNP genotyping from blood leukocytes. Statistical approaches and bioinformatics software were used for QTL identification and downstream analyses. RESULTS: We provided a complete quantification of gene expression and alternative splicing across colon subsites and described their differences. We identified thousands of expression and splicing QTLs and defined their enrichment at genome-wide regulatory regions. We found that part of the SNP-based heritability of diseases affecting colon tissue, such as colorectal cancer and inflammatory bowel disease, but also of diseases affecting other tissues, such as psychiatric conditions, can be explained by the identified QTLs. We provided candidate effector genes for multiple phenotypes. Finally, we provided the Colon Transcriptome Explorer web application. CONCLUSIONS: We provide a large characterization of gene expression and splicing across colon subsites. Our findings provide greater etiologic insight into complex traits and diseases influenced by transcriptomic changes in colon tissue.

Ethanol exposure drives colon location specific cell composition changes in a normal colon crypt 3D organoid model.

Alcohol is a consistently identified risk factor for colon cancer. However, the molecular mechanism underlying its effect on normal colon crypt cells remains poorly understood. We employed RNA-sequencing to asses transcriptomic response to ethanol exposure (0.2% vol:vol) in 3D organoid lines derived from healthy colon (n = 34). Paired regression analysis identified 2,162 differentially expressed genes in response to ethanol. When stratified by colon location, a far greater number of differentially expressed genes were identified in organoids derived from the left versus right colon, many of which corresponded to cell-type specific markers. To test the hypothesis that the effects of ethanol treatment on colon organoid populations were in part due to differential cell composition, we incorporated external single cell RNA-sequencing data from normal colon biopsies to estimate cellular proportions following single cell deconvolution. We inferred cell-type-specific changes, and observed an increase in transit amplifying cells following ethanol exposure that was greater in organoids from the left than right colon, with a concomitant decrease in more differentiated cells. If this occurs in the colon following alcohol consumption, this would lead to an increased zone of cells in the lower crypt where conditions are optimal for cell division and the potential to develop mutations.

A functional variant on 20q13.33 related to glioma risk alters enhancer activity and modulates expression of multiple genes.

Genome-wide association studies (GWAS) have identified single-nucleotide polymorphisms (SNPs) associated with glioma risk on 20q13.33, but the biological mechanisms underlying this association are unknown. We tested the hypothesis that a functional SNP on 20q13.33 impacted the activity of an enhancer, leading to an altered expression of nearby genes. To identify candidate functional SNPs, we identified all SNPs in linkage disequilibrium with the risk-associated SNP rs2297440 that mapped to putative enhancers. Putative enhancers containing candidate functional SNPs were tested for allele-specific effects in luciferase enhancer activity assays against glioblastoma multiforme (GBM) cell lines. An enhancer containing SNP rs3761124 exhibited allele-specific effects on activity. Deletion of this enhancer by CRISPR-Cas9 editing in GBM cell lines correlated with an altered expression of multiple genes, including STMN3, RTEL1, RTEL1-TNFRSF6B, GMEB2, and SRMS. Expression quantitative trait loci (eQTL) analyses using nondiseased brain samples, isocitrate dehydrogenase 1 (IDH1) wild-type glioma, and neurodevelopmental tissues showed STMN3 to be a consistent significant eQTL with rs3761124. RTEL1 and GMEB2 were also significant eQTLs in the context of early CNS development and/or in IDH1 wild-type glioma. We provide evidence that rs3761124 is a functional variant on 20q13.33 related to glioma/GBM risk that modulates the expression of STMN3 and potentially other genes across diverse cellular contexts.

Oncogenic Features in Histologically Normal Mucosa: Novel Insights Into Field Effect From a Mega-Analysis of Colorectal Transcriptomes.

INTRODUCTION: Colorectal cancer is a common malignancy that can be cured when detected early, but recurrence among survivors is a persistent risk. A field effect of cancer in the colon has been reported and could have implications for surveillance, but studies to date have been limited. A joint analysis of pooled transcriptomic data from all available bulk RNA-sequencing data sets of healthy, histologically normal tumor-adjacent, and tumor tissues was performed to provide an unbiased assessment of field effect. METHODS: A novel bulk RNA-sequencing data set from biopsies of nondiseased colon from screening colonoscopy along with published data sets from the Genomic Data Commons and Sequence Read Archive were considered for inclusion. Analyses were limited to samples with a quantified read depth of at least 10 million reads. Transcript abundance was estimated with Salmon, and downstream analysis was performed in R. RESULTS: A total of 1,139 samples were analyzed in 3 cohorts. The primary cohort consisted of 834 independent samples from 8 independent data sets, including 462 healthy, 61 tumor-adjacent, and 311 tumor samples. Tumor-adjacent gene expression was found to represent an intermediate state between healthy and tumor expression. Among differentially expressed genes in tumor-adjacent samples, 1,143 were expressed in patterns similar to tumor samples, and these genes were enriched for cancer-associated pathways. DISCUSSION: Novel insights into the field effect in colorectal cancer were generated in this mega-analysis of the colorectal transcriptome. Oncogenic features that might help explain metachronous lesions in cancer survivors and could be used for surveillance and risk stratification were identified.

Genome-wide Modeling of Polygenic Risk Score in Colorectal Cancer Risk.

Accurate colorectal cancer (CRC) risk prediction models are critical for identifying individuals at low and high risk of developing CRC, as they can then be offered targeted screening and interventions to address their risks of developing disease (if they are in a high-risk group) and avoid unnecessary screening and interventions (if they are in a low-risk group). As it is likely that thousands of genetic variants contribute to CRC risk, it is clinically important to investigate whether these genetic variants can be used jointly for CRC risk prediction. In this paper, we derived and compared different approaches to generating predictive polygenic risk scores (PRS) from genome-wide association studies (GWASs) including 55,105 CRC-affected case subjects and 65,079 control subjects of European ancestry. We built the PRS in three ways, using (1) 140 previously identified and validated CRC loci; (2) SNP selection based on linkage disequilibrium (LD) clumping followed by machine-learning approaches; and (3) LDpred, a Bayesian approach for genome-wide risk prediction. We tested the PRS in an independent cohort of 101,987 individuals with 1,699 CRC-affected case subjects. The discriminatory accuracy, calculated by the age- and sex-adjusted area under the receiver operating characteristics curve (AUC), was highest for the LDpred-derived PRS (AUC = 0.654) including nearly 1.2 M genetic variants (the proportion of causal genetic variants for CRC assumed to be 0.003), whereas the PRS of the 140 known variants identified from GWASs had the lowest AUC (AUC = 0.629). Based on the LDpred-derived PRS, we are able to identify 30% of individuals without a family history as having risk for CRC similar to those with a family history of CRC, whereas the PRS based on known GWAS variants identified only top 10% as having a similar relative risk. About 90% of these individuals have no family history and would have been considered average risk under current screening guidelines, but might benefit from earlier screening. The developed PRS offers a way for risk-stratified CRC screening and other targeted interventions.