Modular and mechanistic changes across stages of colorectal cancer

Abstract

Background: While mechanisms contributing to the progression and metastasis of colorectal cancer (CRC) are well studied, cancer stage-specific mechanisms have been less comprehensively explored. This is the focus of this manuscript.

Methods: Using previously published data for CRC (Gene Expression Omnibus ID GSE21510), we identified differentially expressed genes (DEGs) across four stages of the disease. We then generated unweighted and weighted correlation networks for each of the stages. Communities within these networks were detected using the Louvain algorithm and topologically and functionally compared across stages using the normalized mutual information (NMI) metric and pathway enrichment analysis, respectively. We also used Short Time-series Expression Miner (STEM) algorithm to detect potential biomarkers having a role in CRC.

Results: Sixteen Thousand Sixty Two DEGs were identified between various stages (p-value ≤ 0.05). Comparing communities of different stages revealed that neighboring stages were more similar to each other than non-neighboring stages, at both topological and functional levels. A functional analysis of 24 cancer-related pathways indicated that several signaling pathways were enriched across all stages. However, the stage-unique networks were distinctly enriched only for a subset of these 24 pathways (e.g., MAPK signaling pathway in stages I-III and Notch signaling pathway in stages III and IV). We identified potential biomarkers, including HOXB8 and WNT2 with increasing, and MTUS1 and SFRP2 with decreasing trends from stages I to IV. Extracting subnetworks of 10 cancer-relevant genes and their interacting first neighbors (162 genes in total) revealed that the connectivity patterns for these genes were different across stages. For example, BRAF and CDK4, members of the Ser/Thr kinase, up-regulated in cancer, displayed changing connectivity patterns from stages I to IV.

Conclusions: Here, we report molecular and modular networks for various stages of CRC, providing a pseudo-temporal view of the mechanistic changes associated with the disease. Our analysis highlighted similarities at both functional and topological levels, across stages. We further identified stage-specific mechanisms and biomarkers potentially contributing to the progression of CRC.

Keywords: Biomarkers; CRC stages; Colorectal cancer; Signaling pathways; Stage-specific networks; Stage-unique networks.

Fig. 1

Flow chart of the approach used in our analysis

Fig. 2

Histogram of a permutation test for comparing communities of different stages with degree preservation. A NMI metric between random networks with sizes equal to stage I- and II-specific networks. The actual value of NMI for comparing those stages is 0.0729 (vertical dotted line), corresponding to a p-value of 0.001 (significant for a p-value threshold of 0.05). B NMI metric between random networks with sizes equal to stage II- and III-specific networks. The actual value of NMI for comparing those stages is 0.1501 (vertical dotted line), (p-value of 0.05, significant)

Fig. 3

Topological and functional analysis of the weighted correlation networks. A Heat map for JI values for comparing the communities of stage I-specific network with the communities of other stages. The color-scale is from white for the minimum value of JI (0%) to green for the maximum value (36%). The largest value in each table is selected to perform the functional enrichment. B Functional comparison of the KEGG pathways with p-values ≤ 0.01 and having more than 10 genes for the third community of stage I-specific network with the corresponding communities of other stages. The third community of stage I is very similar to the first community of stage II in terms of the gene counts and p-values of the enriched pathways but less similar to the second community of normal, stages III and IV. C Functional comparison (edge-based enrichment) of the stage-specific and stage-unique networks for 24 cancer-related pathways. The pathways are divided into five categories: cancer related, cell cycle/proliferation/growth, inflammation, angiogenesis and metastasis. Functional enrichment is carried out for both “stage-specific” and “stage-unique” networks. Number of edges related to the genes enriched in each pathway are indicated by the size of the dots. Color scale of the dots indicate p-value with a cut-off of 0.05. D Connectivity of p53 signaling pathway genes across different stage-unique networks. The nodes are colored based on the log2FC values (in a specific stage vs. normal) across the four stages I-IV (dark blue (log2FC of -2) to white (0) to dark red (2)). Each node represents four log2FC values, going from left to right. Edges are colored differently across stages as follows: green for edges in stage I, cyan in stage II, yellow in stage III, and purple in stage IV. E Connectivity of p53 signaling pathway genes in normal

Fig. 4

Biomarkers. A-D Boxplots for 4 biomarkers from STEM analysis and E–F boxplots for 2 stage-specific biomarkers, consistent with GEPIA2 COAD-READ cohort results. Each color indicates one stage and dots show the expressions of biomarker gene for patients in every stage. A HOXB8 with trend (0,1,2,3,4), B WNT2 with trend (0,1,1,1,1), C MTUS1 with trend (0,-1,-2,-3,-4), D SFRP2 with trend (0,-1,-1,-1,-1), E PROCR, stage I-specific biomarker, and F MLXIPL, stage IV-specific biomarker

Fig. 5

Subnetwork of 162 genes in stages I and II. A Stage I-specific network and B Stage II-specific network. Nodes of each subnetwork are grouped together based on the communities they belonged to in the stage-specific networks and colored based on the value of log2FC between that stage and normal: dark blue (log2FC of -2), to white (0) to dark red (2). The width of edges shows the strength of connections based on PCC between them. The thicker the edges are, the larger the PCC between the nodes is

Fig. 6

Drug-Target-PPI network for CRC. Fourteen Drugs approved by FDA for treating CRC (mainly when the cancer metastasize) are used to construct this network; the drug nodes are shown in the center area. The target genes have been found from DrugBank. For each target gene node, four circles are associated with that gene corresponding to four stages I-IV, and are colored based on the log2FC values between a stage and normal (dark blue (log2FC of -1) to white (0) to dark red (log2 FC of 1)). The size of each circle represents the sum of the weights of edges connected to that gene (i.e., gene weights) in each stage. For example, MAPK11 weight is greater in stage I as compared to that in other stages. PPI edges from STRING-db (score threshold ≥ 0.9) are also incorporated in this network by dashed grey-lines between the genes. Some important pathways are also shown. For select functionally important genes, the related functions are listed