NEW YORK (GenomeWeb) – Researchers from the Genome Institute of Singapore have applied a ChIP-seq method that works on small amounts of input DNA to study epigenetic changes in gastric cancer, identifying hundreds of somatically altered promoters and predicted enhancers, many of which have previously not been seen.
The study was published this month in Nature Communications.
Now, the group is using the method to study epigenetic changes in larger numbers of tumor samples from gastric and other cancers and also to profile histone modifications in other diseases.
According to Patrick Tan, a senior investigator at the Genome Institute of Singapore and senior author of the paper, the nano-ChIP-seq protocol is a modified version of the standard ChIP-seq method that enables starting with fewer cells. ChIP-seq requires around 10 million cells, Tan said, but patient samples rarely contain that many cells. By contrast, nano-ChIP-seq has been validated down to 1,000 cells.
The team published the modified ChIP-seq protocol last year in Developmental Cell, demonstrating it on purified mouse germ cells. For this method, researchers first do cell sorting to isolate the cells of interest. They then subject those cells to chromatin immunoprecipitation, after which they do whole-genome amplification of the reverse-crosslinked immunoprecipitated DNA. The researchers next remove the amplification adapters and prepare ChIP-seq libraries.
In the recent Nature Communications study, the team applied the method to five gastric adenocarcinoma tumor and matched normal patient samples from the SingHealth Tissue Repository.
While many cancer sequencing studies focus on alterations to key oncogenes and tumor suppressors, Tan said studying regulatory differences is just as important, because many genes also display differences in gene expression. "As such, it is important to study the altered regulatory elements, such as promoters and enhancers, that drive these aberrant transcriptional patterns," he said.
Using nano-ChIP-seq, the team generated more than 45 million Illumina reads for three different types of chromatin marks for each sample — marks associated with transcribed regions, those associated with promoters and predicted enhancers, and marks associated with repressed regions.
Looking at the chromatin marks, they found that regions of active transcription were exclusive to repressive chromatin. From the five tissue pairs, they predicted more than 21,000 putative promoters and more than 125,000 predicted enhancers. They noted that the method to predict enhancers using ENCODE data had a specificity and sensitivity range of 37 percent to 67 percent and as such, they refer to the putative enhancers as "predicted enhancers rather than true enhancers."
Comparing the gastric cancer samples to the matched normal tissues, they identified 639 promoters that were differentially modified and 975 predicted enhancers that showed somatic alteration. Further analysis revealed that 270 of the somatically altered promoters could be associated with a nearby somatically altered predicted enhancer, and of those, 81 percent showed concordant alterations. In addition, of the 975 altered predicted enhancers, 321 were associated with at least one altered promoter within 500 kb, and of those, 261 showed concordant deregulation. "These results suggest a strong and potentially functional link between somatic alterations in promoters and associated predicted enhancers in [gastric cancer]," the authors wrote.
Of the cancer-associated promoters, the majority were gained promoters rather than lost promoters. And somewhat surprisingly, the majority of those new promoters were "cryptic promoters," meaning they localized to transcription start sites that are not annotated in the RefSeq database.
"We were quite surprised by the level of alternative promoter usage," Tan said, especially given the "vast amount of projects that have been done analyzing transcripts," he said. "We thought we'd have a fairly good idea of the transcription landscape of the cell," but this finding indicates that "there's still quite a bit left to understand with regards to the transcription landscape than we may have appreciated."
Tan said that when the team looked at promoters that were unchanged between the cancer and normal samples or promoters that were present in normal but lost in the cancer samples, those did map to annotated transcription start sites.
Not only was the "widespread, pervasive genome-wide nature of these alternative start sites" surprising, Tan said. But the finding also "suggests that there must be some global mechanism for initiating these cryptic promoters, and if we can identify this mechanism, it may perhaps be targetable."
To try and assess whether the promoters were associated with RNA transcripts, the team next did RNA-seq of 12 tumor/normal pairs including the five original pairs. The majority of promoters, 380 out of 639, or 59 percent, were associated with detectable RNA transcripts. The team identified 192 transcripts that had a greater than 4-fold expression change in gastric cancer compared to normal and about half of those were associated with cryptic promoters. The team then used targeted qPCR to validate 10 of the cryptic promoter-associated transcripts.
The RNA-seq data was invaluable, Tan said. Not only did it allow the researchers to validate the cryptic promoters and demonstrate that they were associated with "bona-fide RNA transcripts," Tan said. But also, "by examining RNA-seq reads bridging across multiple exons, we were able to determine that one consequence of the cryptic promoters was to act as alternative transcript start sites for adjacent genes."
The team also wanted to examine whether underlying genomic changes such as SNVs were associated with the gain of a promoter or enhancer. By reanalyzing the chromatin mark data to look for SNVs in those regions, they found several promoters and enhancers that were only present if a specific somatic SNV or germline SNP was also in that region. However, Tan said that the approach should be validated on a larger number of samples.
Tan said that this study demonstrated a proof of concept, showing that the method can yield valuable insights into epigenetic changes associated with cancer. Going forward, he said his team would like to apply it to larger numbers of cancer samples, including those that are endemic to Asia and also those linked to environmental exposures, "given the impact of the environment on the epigenome," he said. In addition, he said the group would like to use the method to study changes in the epigenome of cancer patients before and after treatment. There is "preclinical data suggesting that changes in the epigenome are a strong contributor to chemotherapy resistance," he said.
The method could also be applied to study epigenetic changes involved in other diseases, he said.
And finally, he said the group would continue to tweak the protocol to decrease input requirements further and improve the bioinformatics pipeline.