Search

Supplementary figure 1: A +B. Hierarchical clustering of microdissected stromal and epithelial (A) and FACS sorted endothelial, fibroblast, epithelial and leukocyte (B) cell populations using previously published non-epithelial derived gene list confirms purity of cell populations. C and D. PCA plot of data presented in top panels highlights clear stratification of samples based on cell lineage using this gene signature. E. Expression values for IFNG expression according to cell-of-origin confirms immune-derived nature of signal. Supplementary figure 2: Association of high PD-L1 expression with MSI subtype and high immune infiltrate. Supplementary figure 3: A. Dot and boxplots of PD-L1 mean gene expression values stratified by MSI status highlights significant association between PD-L1 transcription levels and MSI genotype. B. (Left) Hierarchical clustering of stage II/III CRC cohort based on expression profiles of CD274, CTLA4, LAG3 and IDO1. (Right) Significant upregulation of CD274, IDO1, CTLA4 and LAG3 genes in the PD-L1high subgroup. C. (Left) Dot plot and boxplot of IFNG expression levels stratified by PD-L1-subgroup. (Right) Significant upregulation of IFNG gene in the PD-L1high subgroup.

A recent phase II study of patients with metastatic colorectal carcinoma showed that mismatch repair gene status was predictive of clinical response to PD-1–targeting immune checkpoint blockade. Further examination revealed strong correlation between PD-L1 protein expression and microsatellite instability (MSI) in stage IV colorectal carcinoma, suggesting that the amount of PD-L1 protein expression could identify late-stage patients who might benefit from immunotherapy. To assess whether the clinical associations between PD-L1 gene expression and MSI identified in metastatic colorectal carcinoma are also present in stage II/III colorectal carcinoma, we used in silico analysis to elucidate the cell types expressing the PD-L1 gene. We found a statistically significant association of PD-L1 gene expression with MSI in early-stage colorectal carcinoma (P < 0.001) and show that, unlike in non–colorectal carcinoma tumors, PD-L1 is derived predominantly from the immune infiltrate. We demonstrate that PD-L1 gene expression has positive prognostic value in the adjuvant disease setting (PD-L1low vs. PD-L1high HR = 9.09; CI, 2.11–39.10). PD-L1 gene expression had predictive value, as patients with high PD-L1 expression appear to be harmed by standard-of-care treatment (HR = 4.95; CI, 1.10–22.35). Building on the promising results from the metastatic colorectal carcinoma PD-1–targeting trial, we provide compelling evidence that patients with PD-L1high/MSI/immunehigh stage II/III colorectal carcinoma should not receive standard chemotherapy. This conclusion supports the rationale to clinically evaluate this patient subgroup for PD-1 blockade treatment. Cancer Immunol Res; 4(7); 582–91. ©2016 AACR.

This Supplementary Figure indicates the project plan

Clinical relevance of EphA2 expression in CRC tissues.

Supplementary Table 4: Differentially expressed gene list contrasting tumour region of origin between Lymph Node and Invasive Front.

ACE white paper

Purpose: A number of independent gene expression profiling studies have identified transcriptional subtypes in colorectal cancer with potential diagnostic utility, culminating in publication of a colorectal cancer Consensus Molecular Subtype classification. The worst prognostic subtype has been defined by genes associated with stem-like biology. Recently, it has been shown that the majority of genes associated with this poor prognostic group are stromal derived. We investigated the potential for tumor misclassification into multiple diagnostic subgroups based on tumoral region sampled.Experimental Design: We performed multiregion tissue RNA extraction/transcriptomic analysis using colorectal-specific arrays on invasive front, central tumor, and lymph node regions selected from tissue samples from 25 colorectal cancer patients.Results: We identified a consensus 30-gene list, which represents the intratumoral heterogeneity within a cohort of primary colorectal cancer tumors. Using a series of online datasets, we showed that this gene list displays prognostic potential HR = 2.914 (confidence interval 0.9286–9.162) in stage II/III colorectal cancer patients, but in addition, we demonstrated that these genes are stromal derived, challenging the assumption that poor prognosis tumors with stem-like biology have undergone a widespread epithelial–mesenchymal transition. Most importantly, we showed that patients can be simultaneously classified into multiple diagnostically relevant subgroups based purely on the tumoral region analyzed.Conclusions: Gene expression profiles derived from the nonmalignant stromal region can influence assignment of colorectal cancer transcriptional subtypes, questioning the current molecular classification dogma and highlighting the need to consider pathology sampling region and degree of stromal infiltration when employing transcription-based classifiers to underpin clinical decision making in colorectal cancer. Clin Cancer Res; 22(16); 4095–104. ©2016 AACR.See related commentary by Morris and Kopetz, p. 3989

PDF file - 235K, Clinical relevance of AXL in CRC tissues.

PDF file - 223K, Univariate and multivariate analyses for OS, DSS, DFS and AXL-LOW and AXL-HIGH in stage II/III CRC group (publicly-available CRC microarray dataset).

PDF file - 179K, Effect of AXL siRNA on migration, proliferation and survival of CRC cells.

PDF file - 94K, Sensitivity of parental and invasive CRC cells to chemotherapy and EGFR targeted therapies.

PDF file - 213K, Migration, proliferation rates and RTK profiling of HCT116 parental and invasive cells.

PDF file - 133K, Clinicopathological features of the colorectal cancer patient cohort.

PDF file - 138K, Experimental overview for generation and characterization of invasive CRC subclones.

Supplementary table 1. Clinical-pathological features of the resection-only and chemotherapy/resection stage II/III colorectal cancer patient cohorts with mature survival data in the Singapore dataset. Supplementary table 2. EphA2 and clinical-pathological correlates in CRC. Correlation between EphA2 and CD44, LGR5, CD133, Ki-67 and AXL in 509 stage I-IV CRC cases of the Singapore dataset. Supplementary table 3. Statistical associations between pairs of factors in the surgery only (A) and surgery/chemotherapy (B) stage II/III groups in the Singapore dataset. Supplementary table 4. A. Univariate and multivariate analysis (Cox proportional hazards regression) of resection-chemotherapy stage II/III patient group. B. Final multivariate model (Cox proportional hazards regression) of resection-chemotherapy stage II/III patient group. Supplementary table 5. Statistical association between expression levels of TGF-α and EphA2 using Spearman's Rank Correlation in GSE17536, GSE39582, GSE14333 and The Cancer Genome Atlas (TCGA). Within each dataset the value of Rho and associated p-value is reported.

PDF file - 53K, Statistical analysis.

Modern methods of acquiring molecular data have improved rapidly in recent years, making it easier for researchers to collect large volumes of information. However, this has increased the challenge of recognizing interesting patterns within the data. Atlas Correlation Explorer (ACE) is a user-friendly workbench for seeking associations between attributes in The Cancer Genome Atlas (TCGA) database. It allows any combination of clinical and genomic data streams to be searched using an evolutionary algorithm approach. To showcase ACE, we assessed which RNA sequencing transcripts were associated with estrogen receptor (ESR1) in the TCGA breast cancer cohort. The analysis revealed already well-established associations with XBP1 and FOXA1, but also identified a strong association with CT62, a potential immunotherapeutic target with few previous associations with breast cancer. In conclusion, ACE can produce results for very large searches in a short time and will serve as an increasingly useful tool for biomarker discovery in the big data era.Significance:ACE uses an evolutionary algorithm approach to perform large searches for associations between any combinations of data in the TCGA database.

PDF file - 189K, Univariate and multivariate analyses in stage I-IV CRC group (Singapore dataset) using Cox proportional hazards regression.

This Supplementary Table indicates the sample source for the study

PDF file - 223K, Effect of 5-FU treatment on migration and proliferation of CRC cell lines.

PDF file - 165K, Univariate and multivariate analyses in stage II-III CRC group (Singapore dataset) using Cox proportional hazards regression.

ACE Video demonstration

PDF file - 190K, Effect of AXL small molecule inhibition on downstream signalling pathways and migration in CRC cell lines.

Supplementary Data. Exported ACE outputs

Supplementary Table 3: Differentially expressed gene list contrasting tumour region of origin between Central Tumor and Lymph Node.

Supplementary Table 2: Differentially expressed gene list contrasting tumour region of origin between Central Tumor and Invasive Front. Supplementary Table 3: Differentially expressed gene list contrasting tumour region of origin between CT-LN. Supplementary Table 4: Differentially expressed gene list contrasting tumour region of origin between LN-IF.

Purpose: Despite the use of 5-fluorouracil (5-FU)–based adjuvant treatments, a large proportion of patients with high-risk stage II/III colorectal cancer will relapse. Thus, novel therapeutic strategies are needed for early-stage colorectal cancer. Residual micrometastatic disease from the primary tumor is a major cause of patient relapse.Experimental Design: To model colorectal cancer tumor cell invasion/metastasis, we have generated invasive (KRASMT/KRASWT/+chr3/p53-null) colorectal cancer cell subpopulations. Receptor tyrosine kinase (RTK) screens were used to identify novel proteins that underpin the migratory/invasive phenotype. Migration/invasion was assessed using the XCELLigence system. Tumors from patients with early-stage colorectal cancer (N = 336) were examined for AXL expression.Results: Invasive colorectal cancer cell subpopulations showed a transition from an epithelial-to-mesenchymal like phenotype with significant increases in migration, invasion, colony-forming ability, and an attenuation of EGF receptor (EGFR)/HER2 autocrine signaling. RTK arrays showed significant increases in AXL levels in all invasive sublines. Importantly, 5-FU treatment resulted in significantly increased migration and invasion, and targeting AXL using pharmacologic inhibition or RNA interference (RNAi) approaches suppressed basal and 5-FU–induced migration and invasion. Significantly, high AXL mRNA and protein expression were found to be associated with poor overall survival in early-stage colorectal cancer tissues.Conclusions: We have identified AXL as a poor prognostic marker and important mediator of cell migration/invasiveness in colorectal cancer. These findings provide support for the further investigation of AXL as a novel prognostic biomarker and therapeutic target in colorectal cancer, in particular in the adjuvant disease in which EGFR/VEGF–targeted therapies have failed. Clin Cancer Res; 20(1); 164–75. ©2013 AACR.

Cancers presenting at advanced stages inherently have poor prognosis. High grade serous carcinoma (HGSC) is the most common and aggressive form of tubo-ovarian cancer. Clinical tests to accurately diagnose and monitor this condition are lacking. Hence, development of disease-specific tests are urgently required. The molecular profile of HGSC during disease progression was investigated in a unique patient cohort. A bespoke data browser was developed to analyse gene expression and DNA methylation datasets for biomarker discovery. The Ovarian Cancer Data Browser (OCDB) is built in C# with a.NET framework using an integrated development environment of Microsoft Visual Studio and fast access files (.faf). The graphical user interface is easy to navigate between four analytical modes (gene expression; methylation; combined gene expression and methylation data; methylation clusters), with a rapid query response time. A user should first define a disease progression trend for prioritising results. Single or multiomics data are then mined to identify probes, genes and methylation clusters that exhibit the desired trend. A unique scoring system based on the percentage change in expression/methylation between disease stages is used. Results are filtered and ranked using weighting and penalties. The OCDB's utility for biomarker discovery is demonstrated with the identified target OSR2. Trends in OSR2 repression and hypermethylation with HGSC disease progression were confirmed in the browser samples and an independent cohort using bioassays. The OSR2 methylation biomarker could discriminate HGSC with high specificity (95%) and sensitivity (93.18%). The OCDB has been refined and validated to be an integral part of a unique biomarker discovery pipeline. It may also be used independently to aid identification of novel targets. It carries the potential to identify further biomarker assays that can reduce type I and II errors within clinical diagnostics.

The development of gene signatures is key for delivering personalized medicine, despite only a few signatures being available for use in the clinic for cancer patients. Gene signature discovery tends to revolve around identifying a single signature. However, it has been shown that various highly predictive signatures can be produced from the same dataset. This study assumes that the presentation of top ranked signatures will allow greater efforts in the selection of gene signatures for validation on external datasets and for their clinical translation. Particle swarm optimization (PSO) is an evolutionary algorithm often used as a search strategy and largely represented as binary PSO (BPSO) in this domain. BPSO, however, fails to produce succinct feature sets for complex optimization problems, thus affecting its overall runtime and optimization performance. Enhanced BPSO (EBPSO) was developed to overcome these shortcomings. Thus, this study will validate unique candidate gene signatures for different underlying biology from EBPSO on transcriptomics cohorts. EBPSO was consistently seen to be as accurate as BPSO with substantially smaller feature signatures and significantly faster runtimes. 100% accuracy was achieved in all but two of the selected data sets. Using clinical transcriptomics cohorts, EBPSO has demonstrated the ability to identify accurate, succinct, and significantly prognostic signatures that are unique from one another. This has been proposed as a promising alternative to overcome the issues regarding traditional single gene signature generation. Interpretation of key genes within the signatures provided biological insights into the associated functions that were well correlated to their cancer type.

Background Liver cancer (Hepatocellular carcinoma; HCC) prevalence is increasing and with poor clinical outcome expected it means greater understanding of HCC aetiology is urgently required. This study explored a deep learning solution to detect biologically important features that distinguish prognostic subgroups. A novel architecture of an Artificial Neural Network (ANN) trained with a customised objective function (LRSC) was developed. The ANN should discover new data representations, to detect patient subgroups that are biologically homogenous (clustering loss) and similar in survival (survival loss) while removing noise from the data (reconstruction loss). The model was applied to TCGA-HCC multi-omics data and benchmarked against baseline models that only use a reconstruction objective function (BCE, MSE) for learning. With the baseline models, the new features are then filtered based on survival information and used for clustering patients. Different variants of the customised objective function, incorporating only reconstruction and clustering losses (LRC); and reconstruction and survival losses (LRS) were also evaluated. Robust features consistently detected were compared between models and validated in TCGA and LIRI-JP HCC cohorts. Results The combined loss (LRSC) discovered highly significant prognostic subgroups (P-value = 1.55E−77) with more accurate sample assignment (Silhouette scores: 0.59–0.7) compared to baseline models (0.18–0.3). All LRSC bottleneck features (N = 100) were significant for survival, compared to only 11–21 for baseline models. Prognostic subgroups were not explained by disease grade or risk factors. Instead LRSC identified robust features including 377 mRNAs, many of which were novel (61.27%) compared to those identified by the other losses. Some 75 mRNAs were prognostic in TCGA, while 29 were prognostic in LIRI-JP also. LRSC also identified 15 robust miRNAs including two novel (hsa-let-7g; hsa-mir-550a-1) and 328 methylation features with 71% being prognostic. Gene-enrichment and Functional Annotation Analysis identified seven pathways differentiating prognostic clusters. Conclusions Combining cluster and survival metrics with the reconstruction objective function facilitated superior prognostic subgroup identification. The hybrid model identified more homogeneous clusters that consequently were more biologically meaningful. The novel and prognostic robust features extracted provide additional information to improve our understanding of a complex disease to help reveal its aetiology. Moreover, the gene features identified may have clinical applications as therapeutic targets.

Longitudinal next-generation sequencing of cancer patient samples has enhanced our understanding of the evolution and progression of various cancers. As a result, and due to our increasing knowledge of heterogeneity, such sampling is becoming increasingly common in research and clinical trial sample collections. Traditionally, the evolutionary analysis of these cohorts involves the use of an aligner followed by subsequent stringent downstream analyses. However, this can lead to large levels of information loss due to the vast mutational landscape that characterizes tumor samples. Here, we propose an alignment-free approach for sequence comparison—a well-established approach in a range of biological applications including typical phylogenetic classification. Such methods could be used to compare information collated in raw sequence files to allow an unsupervised assessment of the evolutionary trajectory of patient genomic profiles. In order to highlight this utility in cancer research we have applied our alignment-free approach using a previously established metric, Jensen–Shannon divergence, and a metric novel to this area, Hellinger distance, to two longitudinal cancer patient cohorts in glioma and clear cell renal cell carcinoma using our software, NUQA. We hypothesize that this approach has the potential to reveal novel information about the heterogeneity and evolutionary trajectory of spatiotemporal tumor samples, potentially revealing early events in tumorigenesis and the origins of metastases and recurrences. Key words: alignment-free, Hellinger distance, exome-seq, evolution, phylogenetics, longitudinal.

Modern methods of acquiring molecular data have improved rapidly in recent years, making it easier for researchers to collect large volumes of information. However, this has increased the challenge of recognizing interesting patterns within the data. Atlas Correlation Explorer (ACE) is a user-friendly workbench for seeking associations between attributes in The Cancer Genome Atlas (TCGA) database. It allows any combination of clinical and genomic data streams to be searched using an evolutionary algorithm approach. To showcase ACE, we assessed which RNA sequencing transcripts were associated with estrogen receptor (ESR1) in the TCGA breast cancer cohort. The analysis revealed already well-established associations with XBP1 and FOXA1, but also identified a strong association with CT62, a potential immunotherapeutic target with few previous associations with breast cancer. In conclusion, ACE can produce results for very large searches in a short time and will serve as an increasingly useful tool for biomarker discovery in the big data era. Significance: ACE uses an evolutionary algorithm approach to perform large searches for associations between any combinations of data in the TCGA database.

Background The presence of hypoxia is a poor prognostic factor in prostate cancer and the hypoxic tumor microenvironment promotes radioresistance. There is potential for drug radiotherapy combinations to improve the therapeutic ratio. We aimed to investigate whether hypoxia-associated genes could be used to identify FDA approved drugs for repurposing for the treatment of hypoxic prostate cancer. Methods Hypoxia associated genes were identified and used in the connectivity mapping software QUADrATIC to identify FDA approved drugs as candidates for repurposing. Drugs identified were tested in vitro in prostate cancer cell lines (DU145, PC3, LNCAP). Cytotoxicity was investigated using the sulforhodamine B assay and radiosensitization using a clonogenic assay in normoxia and hypoxia. Results Menadione and gemcitabine had similar cytotoxicity in normoxia and hypoxia in all three cell lines. In DU145 cells, the radiation sensitizer enhancement ratio (SER) of menadione was 1.02 in normoxia and 1.15 in hypoxia. The SER of gemcitabine was 1.27 in normoxia and 1.09 in hypoxia. No radiosensitization was seen in PC3 cells. Conclusion Connectivity mapping can identify FDA approved drugs for potential repurposing that are linked to a radiobiologically relevant phenotype. Gemcitabine and menadione could be further investigated as potential radiosensitizers in prostate cancer.

Graphical abstract, RNA-sequencing (RNA-seq) is a relatively new technology that lacks standardisation. RNA-seq can be used for Differential Gene Expression (DGE) analysis, however, no consensus exists as to which methodology ensures robust and reproducible results. Indeed, it is broadly acknowledged that DGE methods provide disparate results. Despite obstacles, RNA-seq assays are in advanced development for clinical use but further optimisation will be needed. Herein, five DGE models (DESeq2, voom + limma, edgeR, EBSeq, NOISeq) for gene-level detection were investigated for robustness to sequencing alterations using a controlled analysis of fixed count matrices. Two breast cancer datasets were analysed with full and reduced sample sizes. DGE model robustness was compared between filtering regimes and for different expression levels (high, low) using unbiased metrics. Test sensitivity estimated as relative False Discovery Rate (FDR), concordance between model outputs and comparisons of a ’population’ of slopes of relative FDRs across different library sizes, generated using linear regressions, were examined. Patterns of relative DGE model robustness proved dataset-agnostic and reliable for drawing conclusions when sample sizes were sufficiently large. Overall, the non-parametric method NOISeq was the most robust followed by edgeR, voom, EBSeq and DESeq2. Our rigorous appraisal provides information for method selection for molecular diagnostics. Metrics may prove useful towards improving the standardisation of RNA-seq for precision medicine.

Identifying robust predictive biomarkers to stratify colorectal cancer (CRC) patients based on their response to immune-checkpoint therapy is an area of unmet clinical need. Our evolutionary algorithm Atlas Correlation Explorer (ACE) represents a novel approach for mining The Cancer Genome Atlas (TCGA) data for clinically relevant associations. We deployed ACE to identify candidate predictive biomarkers of response to immune-checkpoint therapy in CRC. We interrogated the colon adenocarcinoma (COAD) gene expression data across nine immune-checkpoints (PDL1, PDCD1, CTLA4, LAG3, TIM3, TIGIT, ICOS, IDO1 and BTLA). IL2RB was identified as the most common gene associated with immune-checkpoint genes in CRC. Using human/murine single-cell RNA-seq data, we demonstrated that IL2RB was expressed predominantly in a subset of T-cells associated with increased immune-checkpoint expression (P < 0.0001). Confirmatory IL2RB immunohistochemistry (IHC) analysis in a large MSI-H colon cancer tissue microarray (TMA; n = 115) revealed sensitive, specific staining of a subset of lymphocytes and a strong association with FOXP3+ lymphocytes (P < 0.0001). IL2RB mRNA positively correlated with three previously-published gene signatures of response to immune-checkpoint therapy (P < 0.0001). Our evolutionary algorithm has identified IL2RB to be extensively linked to immune-checkpoints in CRC; its expression should be investigated for clinical utility as a potential predictive biomarker for CRC patients receiving immune-checkpoint blockade.

RAS mutant (MT) metastatic colorectal cancer (mCRC) is resistant to MEK1/2 inhibition and remains a difficult-to-treat group. Therefore, there is an unmet need for novel treatment options for RASMT mCRC. RALA and RALB GTPases function downstream of RAS and have been found to be key regulators of several cell functions implicated in KRAS-driven tumorigenesis. However, their role as regulators of the apoptotic machinery remains to be elucidated. Here, we found that inhibition of RALB expression, but not RALA, resulted in Caspase-8-dependent cell death in KRASMT CRC cells, which was not further increased following MEK1/2 inhibition. Proteomic analysis and mechanistic studies revealed that RALB depletion induced a marked upregulation of the pro-apoptotic cell surface TRAIL Death Receptor 5 (DR5) (also known as TRAIL-R2), primarily through modulating DR5 protein lysosomal degradation. Moreover, DR5 knockdown or knockout attenuated siRALB-induced apoptosis, confirming the role of the extrinsic apoptotic pathway as a regulator of siRALB-induced cell death. Importantly, TRAIL treatment resulted in the association of RALB with the death-inducing signalling complex (DISC) and targeting RALB using pharmacologic inhibition or RNAi approaches triggered a potent increase in TRAIL-induced cell death in KRASMT CRC cells. Significantly, high RALB mRNA levels were found in the poor prognostic Colorectal Cancer Intrinsic Subtypes (CRIS)-B CRC subgroup. Collectively, this study provides to our knowledge the first evidence for a role for RALB in apoptotic priming and suggests that RALB inhibition may be a promising strategy to improve response to TRAIL treatment in poor prognostic RASMT CRIS-B CRC.

Combining alignment-free methods for phylogenetic analysis with multi-regional sampling using next-generation sequencing can provide an assessment of intra-patient tumour heterogeneity. From multi-regional sampling divergent branching, we validated two different lesions within a patient’s prostate. Where multi-regional sampling has not been used, a single sample from one of these areas could misguide as to which drugs or therapies would best benefit this patient, due to the fact these tumours appear to be genetically different. This application has the power to render, in a fraction of the time used by other approaches, intra-patient heterogeneity and decipher aberrant biomarkers. Another alignment-free method for calling single-nucleotide variants from raw next-generation sequencing samples has determined possible variants and genomic locations that may be able to characterize the differences between the two main branching patterns. Alignment-free approaches have been applied to relevant clinical multi-regional samples and may be considered as a valuable option for comparing and determining heterogeneity to help deliver personalized medicine through more robust efforts in identifying targetable pathways and therapeutic strategies. Our study highlights the application these tools could have on patient-aligned treatment indications.

Background DNA methylation plays a vital role in the cell, but loss-of-function mutations of the maintenance methyltransferase DNMT1 in normal human cells are lethal, precluding target identification, and existing hypomorphic lines are tumour cells. We generated instead a hypomorphic series in normal hTERT-immortalised fibroblasts using stably integrated short hairpin RNA. Results Approximately two-thirds of sites showed demethylation as expected, with one-third showing hypermethylation, and targets were shared between the three independently derived lines. Enrichment analysis indicated significant losses at promoters and gene bodies with four gene classes most affected: (1) protocadherins, which are key to neural cell identity; (2) genes involved in fat homoeostasis/body mass determination; (3) olfactory receptors and (4) cancer/testis antigen (CTA) genes. Overall effects on transcription were relatively small in these fibroblasts, but CTA genes showed robust derepression. Comparison with siRNA-treated cells indicated that shRNA lines show substantial remethylation over time. Regions showing persistent hypomethylation in the shRNA lines were associated with polycomb repression and were derepressed on addition of an EZH2 inhibitor. Persistent hypermethylation in shRNA lines was, in contrast, associated with poised promoters. Conclusions We have assessed for the first time the effects of chronic depletion of DNMT1 in an untransformed, differentiated human cell type. Our results suggest polycomb marking blocks remethylation and indicate the sensitivity of key neural, adipose and cancer-associated genes to loss of maintenance methylation activity. Electronic supplementary material The online version of this article (10.1186/s13072-018-0182-4) contains supplementary material, which is available to authorized users.

BackgroundBRAF mutation occurs in 8-15% of colon cancers (CC), and is associated with poor prognosis in metastatic disease. Compared to wild-type BRAF (BRAFWT) disease, stage II/III CC patients with BRAF mutant (BRAFMT) tumors have shorter overall survival after relapse; however, time-to-relapse is not significantly different. The aim of this investigation was to identify, and validate, novel predictors of relapse of stage II/III BRAFMT CC.Patients and methodsWe used gene expression data from a cohort of 460 patients (GSE39582) to perform a supervised classification analysis based on risk-of-relapse within BRAFMT stage II/III CC, to identify transcriptomic biomarkers associated with prognosis within this genotype. These findings were validated using immunohistochemistry in an independent population-based cohort of Stage II/III CC (n=691), applying Cox proportional hazards analysis to determine associations with survival.ResultsHigh gene expression levels of Bcl-xL, a key regulator of apoptosis, were associated with increased risk of relapse, specifically in BRAFMT tumors (HR=8.3, 95% CI 1.7-41.7), but not KRASMT/BRAFWT or KRASWT/BRAFWT tumors. High Bcl-xL protein expression in BRAFMT, untreated, stage II/III CC was confirmed to be associated with an increased risk of death in an independent cohort (HR=12.13, 95% CI 2.49-59.13). Additionally, BRAFMT tumors with high levels of Bcl-xL protein expression appeared to benefit from adjuvant chemotherapy (P for interaction =0.006), indicating the potential predictive value of Bcl-xL expression in this setting.ConclusionsThese findings provide evidence that Bcl-xL gene and/or protein expression identifies a poor prognostic subgroup of BRAFMT stage II/III CC patients, who may benefit from adjuvant chemotherapy.Key MessageUsing a combination of computational biology discovery and immunohistochemistry validation in independent patient cohorts, we show that high expression of the apoptosis regulator Bcl-xL is associated with disease relapse specifically within BRAF mutant stage II/III colon cancer.This data could enable tailored disease management to reduce relapse rates in the most aggressive subtype.