Multi-omics landscapes of colorectal cancer subtypes discriminated by an individualized prognostic signature for 5-fluorouracil-based chemotherapy

Abstract

Until recently, few prognostic signatures for colorectal cancer (CRC) patients receiving 5-fluorouracil (5-FU)-based chemotherapy could be used in clinical practice. Here, using transcriptional profiles for a panel of cancer cell lines and three cohorts of CRC patients, we developed a prognostic signature based on within-sample relative expression orderings (REOs) of six gene pairs for stage II-III CRC patients receiving 5-FU-based chemotherapy. This REO-based signature had the unique advantage of being insensitive to experimental batch effects and free of the impractical data normalization requirement. After stratifying 184 CRC samples with multi-omics data from The Cancer Genome Atlas into two prognostic groups using the REO-based signature, we further revealed that patients with high recurrence risk were characterized by frequent gene copy number aberrations reducing 5-FU efficacy and DNA methylation aberrations inducing distinct transcriptional alternations to confer 5-FU resistance. In contrast, patients with low recurrence risk exhibited deficient mismatch repair and carried frequent gene mutations suppressing cell adhesion. These results reveal the multi-omics landscapes determining prognoses of stage II-III CRC patients receiving 5-FU-based chemotherapy.

Figure 1

Overview of the workflow used in this study. clinically relevant 5-FU-resistant genes, genes correlates with both cells' GI₅₀ value for 5-FU and RFS of CRC patients receiving 5-FU-based chemotherapy; CNV, copy number; Exp, expression; GPS, gene pair signature; Meth, methylation; MSI, microsatellite instability; n, the number of samples; +(−), genes positively (negatively) correlated with the GI₅₀ values of 5-FU or the RFS of CRC patients receiving 5-FU-based chemotherapy.

Figure 2

The performance of the 6-GPS for predicting the RFS of the CRC patients. The Kaplan–Meier curves of RFS for the CRC patients treating with 5-FU-based therapy in the training data set (GSE39582) (a–c) and the validation data sets ((d) GSE14333 and (e) TCGA). The Kaplan–Meier curves of RFS for (f) the CRC patients without 5-FU-based treatment. n, the number of samples.

Figure 3

The performance of the 6-GPS for predicting 5-FU resistance for 58 cancer cell lines from the NCI-60 data. The heat map (a) and box plot (b) of −log₁₀ GI₅₀ values of the 58 cell lines identified by the 6-GPS. Abbreviations: BR, breast; CNS, central nervous system; CO, colon; LC, non-small cell lung; LE, leukemia; ME, melanoma; OV, ovarian; PR, prostate; RE, renal.

Figure 4

The copy number aberration regions and mutation genes characterizing the high- and low-risk patients, respectively. The frequencies of (a) 23 amplification regions, (b) 45 deletion regions and (c) 24 mutation genes in the high- and low-risk groups, respectively. The 24 mutation genes exhibited significantly higher frequencies in the low-risk group compared with the high-risk group (Fisher test, P-value<0.01). ***P-value<0.00001, **P-value<0.0001, *P-value<0.001, ⁺P-value<0.05.

Figure 5

The multi-omic alterations of pathways in the high-risk group. Metabolites of 5-FU: FBAL, a-fluoro-b-alanine; FUrd, fluorouridine; FUMP, fluorouridine monophosphate; FUDP, fluorouridine-5′-diphosphate; FUTP, fluorouridine triphosphate; FdUrd, fluorodeoxyuridine; FdUMP, fluorodeoxyuridine monophosphate; FdUTP, fluorodeoxyuridine triphosphate; FdUDP, fluorodeoxyuridine-5′-diphosphate.

Figure 6

The direct PPI links between the genes with frequent copy number alterations in the high-risk group and 5-FU activity-related genes. 5-FU activity-related genes: genes involved in 5-FU transport, metabolism and other downstream effects (such as DNA repair, apoptosis and cell cycle regulation). The green nodes denoted 5-FU activity-related genes. The red/blue nodes denoted genes with amplification or deletion in the high-risk group. The triangular nodes were the genes overlapped between the genes with frequent copy number alterations and 5-FU activity-related genes.