Comprehensive bioinformatics analysis of the characterization and determination underlying mechanisms of over-expression and co-expression of genes residing on 20q in colorectal cancer

Abstract

The Long arm of chromosome 20 (20q) is closely related to the development of colorectal cancer, so identifying the expression profile of genes on 20q through a comprehensive overview is indispensable. In this article, preliminar experimental data, several available databases and bioinformatics tools such as the Cancer Genome Atlas, the Encyclopedia of DNA Elements, the JASPAR database and starBase were combined to analyze the correlation between genes and chromosomal aberrations, microRNA and transcription factors, as well as to explore the expression feature and potential regulative mechanism. The results showed that the most frequently unregulated genes in colorectal cancer arelocated on chromosome 20q, present a significant CNA-mRNA correlation.Furthermore, the genes with mRNA overexpression showed co-expression features and tended to be clustered within the same genomic neighborhoods. Then, several genes were selected to carry out further analysis and demonstrated that shared transcription factors, a conserved bidirectional promoter, and competition for a limited pool of microRNAin the 3'UTR of mRNA may be the underlying mechanisms behind the co-expression of physically adjacent genes.Finally, the databases, Lentivirus shRNA, and qPCR were used to find that these adjacent genes with co-expression cooperatively participated in the same biological pathways associated with the pathogenesis and development of colorectal cancer.

Keywords: 20q; CNA; adjacent gene; co-expression; colorectal cancer.

Figure 1. The genes located on chromosome 20q were frequently unregulated in colorectal cancer

(A) The top 20 genes with mRNA over expression in colon adenocarcinoma (COSMIC study Id: COSU376), the matrix on the left side of the figure list the top unregulated 20 genes in colon cancer (more details can be found in Supplementary Figure 1); the distribution of 20 genes on 20q and the symbol, location and gene ID of each gene were showed on the right side of the figure, red dashed box show all genes located in 3 DNA fragments, black rectangle was used to distinguish the genes located in 20q13.33, 20q13.32, 20q13.31, 20q13.13, 20q13.12, 20q11.23, 20q11.22, 20q11.21. (B/C) The representative gene (PLAGL2 and POFUT1) expression in colorectal cancer and normal tissues, which analyzed by Oncomine(GSE9348 and TCGA Colorectal), NC: normal colon, NR: normal rectum, FC: fold change. All other genes demonstrated same result as PLAGL2 (data not show, more detail can be found in Ocomine). (D) The representative protein expression derived from The Human Protein Atlas (PLAGL2 and POFUT1 with moderate to strong immunoreactivity was show here, data derived from antibody-based protein profiling using immunohistochemistry), N: normal tissue, n: negative,h: strong or high. (E) Survival analysis of PLAGL2, POFUT1 and TM9SF4 in colorectal cancer, data derived from TCGA.

Figure 2. 8 genes located on chromosome 20q and frequently unregulated in colorectal cancer showed significant CNA–mRNA correlation

(A) 8 unregulated genes located on 20q13.33 (YYHDF1), 20q13.32 (NELFCD or TH1L), 20q13.13 (STAU1), 20q13.12 (ZWIM3), 20q11.23 (AAR2 or C20orf4), 20q11.22 (PIGU) and 20q11.21 (PLAGL2 and POFUT1) were selected for somatic mutations analysis in CRC. (B) Amplification and mRNA up regulation are the main mutations of 8 genes; the result is evaluated by “The cBioPortal for Cancer Genomics database (bowel-colorectal adenocarcinoma)”. (C) 8 genes showed significant CNA–mRNA correlation (p<0.0001, r>0.65).

Figure 3. The genes located on chromosome 20q with similar expression levels tend to be clustered within the same genomic neighborhoods

(A) The mRNA expression of POFUT1 and PLAGL2 or STAU1 shows extremely obvious positive correlation, 218 TCGA sample data with quantifiable mRNA measurements (Supplementary Table 1) were used for mRNA–mRNA evaluation. (B) The co-expression data between all genes (20436 genes) and STAU1 and POFUT1 in CRC was download from the cBioPortal for Cancer Genomics (Colorectal Adenocarcinoma, 633 samples) and show same obvious positive correlation between POFUT1 and PLAGL2 or STAU1. (C) 86 genes correlated significantly with POFUT1 (Pearson score≥0.6), 96.5% (83/86) gene distributed on 20q and more genes located on 20q11.22,20q11.21 and 20q11.23 (23%, 13% and 19%, respectively, calculated through gene with Pearson score≥0.6/all genes located on this DNA region). (D) 34 genes adjacent with POFUT1 and 17 genes adjacent with STAU2 presented a distinct discontinuous distribution and are clustered into co-regulated groups of 1-7 genes such as POFUT1 co-expressed with its adjacent genes TM9SF4(0.82), PLAGL2(0.87), KIF3B(0.76) and ASXL1(0.78), STAU1 co-expressed with ARFGEF2(0.82),CSE1L(0.71) and DDX27(0.83) (gene cluster presented with different color); the ordinate means Pearson score(genes vs. POFUT1 and POFUT1 was set to 1 in the above diagram, genes vs. STAU1 and STAU1 was set to 1 in the image below), CNV1 and CNV2 means two groups of genes located on same focal amplification fragment. (More detail can be found in Supplementary Table 2)

Figure 4. Adjacent genes with co-expression may be regulated in part by common transcription factors

(A) Interaction networks predicted by GeneMANIA also identified TM9SF4, POFUT1, PLAGL2, KIF3B and ASXL1 showed co-expression. (B) Venn diagrams indicated that the promoters of these 5 genes share 60 transcription factors supported by ChIP-Seq evidence (the transcription factors of each gene and intersections can be found in Supplementary Table 3). (C) The representative enrichment analysis results for ENCODE gene-set library found that 2-4 adjacent genes may share same transcription factor (more details can be found in Supplementary Table 3). (D) The representative enrichment analysis results for PWMs from TRANSFAC and JASPAR gene-set library found that 2-4 adjacent genes may share same transcription factor (more details can be found in Supplementary Table 3). C and D analyzed by software Enrichr.

Figure 5. PLAGL2 and POFUT1 are classic bidirectional gene pair

(A) Schematic representation of a bidirectional promoter between PLAGL2 and POFUT1 genes, black arrows denote transcription direction of both genes. (B) We employed UCSC genome browser(alter tracks displayed) to analyze the genomic organization, the human PLAGL2 and POFUT1 are located on 20q11.21 in a head-to-head orientation, the start of both genes are separated less than 150bp and overlapped with a putative CpG island (87bp), according to ENCODE v24 annotation determined by a ChIP-seq assay, the intergenic region displayed high DNase 1 hypersensitivity and was enriched with histone hallmark associated with transcription like H3K4me3, H3K4me1 and H3K27Ac, furthermore the nucleotide sequence highly evolutionarily conserved (more details can be found in Supplementary Table 4) and contains multiple potential DNA motifs including SP1,TFAP2C, E2F4, EGR1 and Ets transcription family, which have been verified overrepresented in bidirectional promoter.

Figure 6. The optimal bidirectional promoter and transcription factor

(A) Because 3 gene database including NIBI RefSeq, ENCODE v24 and Ensembl genome browser 87 displayed different positionfor PLAGL2 and POFUT1, so we draw a schematic diagramto find out the optimal nucleotide sequence (89bp between32207791 to 32207880) for bidirectional promoter, two-way arrow denote the base number between the different start of both genes (the start derived from NIBI RefSeq was marked red and show here, and more descriptions can be found in Supplementary Table 4). (B) Base pair alignment between human and several other species in the region between 32207791 to 32207880 (89bp), the predicted transcriptions factor binding sites are marked by horizontal line with different color (SP1 and SP3 with blue, SP2 with black, E2F4 and E2F6 with red, EGR1with green, TFAP2C with purple). Two features have been observed, first and foremost, the base sequence including transcription factor binding sites are highly conserved during evolution, besides all predicted transcription factor binding sites mainly focused on 3 regions, especially TFAP2C can better binding to positive-sense(+) and negative-sense(-) strand. Among these transcriptions factor, E2F4 and SP3 were selected for co-expression analysis and labeled with rectangular box. (More conserved nucleotide sequence and details can be found in Supplementary Table 4 and transcriptions factor binding sites analyzed by The JASPAR database can be found in Supplementary Figure 2.) (C) Transcriptions factor E2F4 and SP3 correlated significantly with POFUT1 and PLAGL2 in colon and rectal cancer.

Figure 7. Adjacent genes with co-expression, especially adjacent gene pair, may be mediated by ceRNA

(A) Venn Diagrams demonstrated that 3’UTR of five genes shared 9 miRNAs predicted by TargetScan. (B) The intersections of experimentally validated microRNAs of PLAGL2 and POFUT1 are 12 and 11 in TarBase v7.0 and miRTARBASE6.0, and the intersections of microRNAs in Starbase v2.0 is 38, the union of 3 intersections is 52, which indicated both genes share common 52 microRNAs with experimental evidence. (C) The representative anti-correlation (pearson correlation: r<0, p-value<0.05) between miRNA and target genes across colon and rectal cancer. (D) Starbase v2.0 predicted that POFUT1 is a perfect ceRNA of PLAGL2. (More details about Figure 7 can be found in Supplementary Table 5.)

Figure 8. PLAGL2 and POFUT1 cooperatively participated in SRC biological pathways

(A) 63 genes which can maintain correlation with POFUT1and PLAGL2 in 14 human tumor tissues were selected for protein-protein interactions by STRING v10.5, the oncogenes SRC is the most important intersection point. (B) The knock-out of gene PLAGL2 (sh-PL-PLAGL2) in colorectal cancer cell SW620 led to inhibition of SRC (sh-PL-SRC). (C) The knock-out of gene POFUT1 (sh-PO-POFUT1) in colorectal cancer cell SW620 led to inhibition of SRC (sh-PL-SRC).