Genome-wide DNA Methylation Profiling Of Primary Colorectal Laterally Spreading Tumors Identifies Disease-specific Epi-mutations On Common Pathways

Objective:1.To define the DNA methylation signature of LSTs,especially for genomic regions that are typically neglected.2.To define the main DNA methylation difference between adenomas and LSTs.3.To identifie disease-specific epi-mutations of primary colorectal laterally spreading tumors on common pathways by genome-wide DNA methylation profiling.Methods:1.Patient recruitment and sample collection2.Human Methylation 450K Array3.DMP and DMR identification?-values in LST cases were compared against controls using paired t-tests.A statistical cutoff of Bonferroni corrected/adjusted q-value<0.05 was applied to select LST associated differentially methylated positions?DMPs?from each pairwise comparison.We also investigated potential regional clustering of differential methylation to define.differentially methylated regions?DMRs?by using the algorithm in minfi package.4.Differentially methylated position distribution analysisThe DMPs were stratified per genetic feature/chromosome and compared to the total number of 450k probes associated with the respective genetic feature/chromosome.A Fisher exact test of independence was then used to calculate the probability that the number of DMPs found for a specific genetic feature/chromosome was significantly different from the expected number of DMPs.A second Fisher exact test was then performed on the number of hypermethylated DMPs versus the hypomethylated DMPs to assess whether the distribution was significantly different in any genetic feature/chromosome.The threshold for statistical significance was set to a Bonferroni-adjusted?of 0.05.5.Pyrosequencing validationWe used 10 normal-matched cases as a discovery cohort and 9 additional cases as a validation cohort.We performed pyrosequencing for validation.6.Gene Ontology and pathway enrichment analysis7.Histone modification enrichment8.Public data download and normalizationProcessed DNA methylation data of GSE4868413 were downloaded from the NCBI Gene Expression Omnibus?GEO?database.The DNA methylation data and expression analyses of 11 laterally spreading tumors and paired normal colorectal mucosa were also downloaded from GEO?GSE776357?.We used the RNA-seq TMM normalized read counts to quantify expression levels of nearest genes associated with IGR DMPs&DMRs and transcription factors.9.Calculating observed and expected values of DMPs&DMRs in ChromHMM-18segmentationsChromHMM-18 segmentations by the Roadmap Project14 on hg19 were downloaded from the Roadmap Project data repository.The 18-state model consists of 12 active states and 6 repressed states.For the set of 78 reference epigenomes'ChromHMM-18 annotation bedfiles was used to determine the overlap of DMPs&DMRs with each ChromHMM-18state.The total expected DMPs&DMRs in state m in cell type n was calculated using custom R code.Then the total observed DMPs&DMRs in state m in cell type n was tabulated and compared to expectation.10.TFBS identification enrichmentGenome sequences were obtained for repressed regions DMPs from the hg19 human genome assembly.We extended to 1000bp from the midpoint of DMPs&DMRs and 450k probes.Motif finding analysis was performed using the FIMO tool from the MEME suite using default vertebrate databases15,with a q-value?FDR-corrected p-value?cutoff of 0.05and ratio of target sequences with motif vs background sequences with motif>1.1 defined as being significantly enriched.11.Availability of supporting dataArray data has been deposited in the NCBI's Gene Expression Omnibus repository under the accession number GSE106556.Results:1.Discovery of DMPs and DMRs in LST with validationWe set out to define the DNA methylation aberration of LST disease.Focal analysis identified 2,452 differentially methylated probes?DMPs?in the 10 LST neoplasms using a stringent statistical cutoff?Bonferroni correction of P<0.05,Figure 3A?.We observed a statistically significant difference in the DMP distribution for multiple chromosomes,especially chrX(P<3.37×10^-18,Fisher's exact test)and chr8(P<1.85×10^-16,Fisher's exact test).Interestingly,we found significantly more hypomethylated DMPs than hypermethylated DMPs.Of all the DMPs,1,925?78.51%?were hypomethylated and 527?21.49%?were hypermethylated?Figure 3B?.Next,differentially methylated regions?DMRs?were defined using the algorithm in minfi package16.This analysis defined 73 hypermethylated DMRs and 13 hypomethylated DMRs.We also used pyrosequencing to validate four probes located in different regions in our discovery cohorts and in a new validation cohort consisting of 9 additional cases?Figure 4?.All four probes displayed consistent differential methylation levels,confirming our discovery.2.DMPs are highly enriched in intergenic regions and around transcription start sites3.LSTs exhibited hallmark DNA methylation abnormality around TSSs4.Intergenic regions are hotspot for DNA methylation abnormality in LSTs5.Potential functions enriched in IGR-regionsThe top biology function of IGR-hypomethylated DMPs was“digestive system development”,which included many important developmental transcription factors,such as TGFB2,FOXF1,FOXF2,GATA4,FGF10 and FGF9?Figure 8A?.Interestingly,IGR-hypomethylated DMPs were associated with genes involved in histone deacetylation?Histone deacetylation,-log10?P-value?=4.97??HDAC2,HDAC4,HDAC9,NACC2,SALL1 and SUDS3??Figure 8A?.This result highlighted the possibility of a cross-talk between lesions in DNA methylation and histone modification in LST tumorigenesis.6.H3K9me3 are highly enriched in IGR-DMPs7.Repressed regions in normal mucosa are demethylated in LST,while hypomethylation of IGR-DMP in LST may lead to dysregulation of transcription factors8.The enrichment states of signal pathways in DMPs between LSTs,adenomas and CRCsA Venn diagram revealed the overlapping DMPs in the LST,adenomas and CRCs.435 genes were found to be hypermethylated in all three colon neoplasms,while 517 hypomethylated DMPs were shared by three neoplasms?Figure 13A?.Next,we performed a pathway-centric survey of DMPs in different neoplasms.Interestingly,we found PI3K-AKT enrichment in hypermethylated DMPs of CRC comparing to normal controls;both hyper-and hypomethylated DMPs of adenoma caused enrichment in the PI3K-AKT pathway;and LST's hypomethylated DMPs showed significant changes in PI3K-AKT as well?Figure 13B?.Further more,we found eight important pathways with epigenetic lesions,but the patterns of the DNA methylation abnormality were opposite to each other between LSTs and adenomas?Figure 13C?.Genes enriched in Ras and Rap1 signaling pathways were found to be hypomethylated in LSTs whereas in adenomas they were hypermethylated,for example,MAGI2,PRKCB,FGF12,PDGFD,NGF,PIK3R5,GRIN2A and PDGFRA?Figure 13D?.The three related diseases occupy nearby but distinct niches on such a landscape,underscoring the distinct molecular events that disturb common and important pathways,for example,the PI3K-Akt signaling pathway?Figure 14?.The three diseases shared some common epigenetic target genes,including ECM,ITGA,CCDN1,and GF?Figure 15?.The PI3K Class IA factor,P21 and FasL exhibited LST-specific epigenetic changes.In contrast,adenoma exhibited disregulated RTK-ERK pathway which could potentially lead to cell proliferation,angiogenesis and DNA repair problems.Abnormalities in protein synthesis and tumor metabolism highlighted unique features of CRCs.Conclusion:1.DMPs in LST are highly enriched in IGRs and TSSs,and hypomethylation DMPs and DMRs are mainly located in IGRs.2.The hypomethylation IGR-DMPs in LST may result in the loss of specific phenotype of colon cells and the formation of tumors by affecting the non-colon cell type-specific regulatory elements in the repressed regions.3.The DNA methylation expression of DMP enriched by signal pathways such as PI3K-AKT,Ras and Rap1 differs greatly between LST and adenoma,which may be one of the reasons for the difference in the molecular mechanism of tumor transformation between LST and adenoma.4.Both IA factor P21 and FasL of PI3K family showed LST specific epigenetic changes,while adenoma showed abnormal in RTK-ERK pathway.Abnormalities in protein synthesis and tumor metabolism highlighted unique features of CRCs.5.PI3K-Akt pathway has unique epigenetic differences in the occurrence and development of LST,adenoma and CRC,which has very important clinical significance for both LST,adenoma and CRC patients.It may be used as a biomarker for early monitoring,therapeutic targets and prognosis judgement.