New Sequencing Approach Examines Specific RNA Epigenetic Modifications

NEW YORK (GenomeWeb) – Researchers from Boston and Los Angeles have devised a new sequencing approach to gauge N⁶-methyladenosine modifications of RNA molecules.

N⁶-methyladenosine (m⁶A) is a reversible chemical modification of mRNAs and long noncoding RNAs. While such modifications have been linked to RNA metabolism, not much is known about how common modified transcripts are and what their regulatory effect is, the researchers noted.

As they described in Nature Methods this week, Harvard Medical School's Cosmas Giallourakis and his colleagues developed an approach they dubbed m⁶A-level and isoform-characterization sequencing, or m⁶A-LAIC-seq, to quantify how many transcripts of a gene are N⁶-methyladenosine (m⁶A) modified and explore the relationship between these modifications and alternative RNA isoforms.

"We envision that m⁶A-LAIC-seq complements standard m⁶A-seq identification of methylation sites and adds to our armamentarium for the study of m⁶A biology," Giallourakis and his colleagues wrote in their paper. "As demonstrated in this work, the combined use of these two technologies provides new insights into the dynamic range and isoform complexity of the m⁶A epitranscriptome."

The m⁶A-LAIC-seq approach relies on anti-m⁶A RNA immunoprecipitation (RIP) followed by sequencing full-length m⁶A-positive and m⁶A-negative transcripts. In contrast, the researchers noted that standard m⁶A-seq fragments RNA prior to RIP. Instead, full-length RNA isolation and sequencing allows for not only the quantification of m⁶A levels, but also of differential isoform usage, they noted.

In a series of analyses, the researchers found that the anti-m⁶A RIP efficiency is independent of the number of modifications per transcript, and they were able to detect as few as one m⁶A-modified transcript among 1,000 unmodified transcripts. In addition, using spike-in controls, they reported not being able to detect the three nonmethylated RNA transcript spike-in controls in the elute, but finding them in the supernatant, suggesting a high specificity. They also noted that they sequenced intact full-length RNAs with no local anti-m⁶A site enrichment.

Giallourakis and his colleagues then applied this approach to study the epitranscriptome of the human embryonic stem cell line H1-ESC. Here, m⁶A levels appeared to follow a nearly bimodal distribution in which most genes had less than 50 percent methylation levels, they noted. As m⁶A RIP efficiency was independent of the number of peaks, they concluded that increasing m⁶A levels with the number of peaks likely reflected m⁶A modifications occurring at different sites on separate transcripts.

By analyzing Gene Ontology terms, the researchers found that the most highly modified genes were mostly involved in transcription regulation and the regulation of RNA metabolic processes. For instance, they reported that genes whose transcripts had high m⁶A levels were enriched in transcription factor domains, especially ones containing the repressive KRAB domain.

The researchers also applied m⁶A-LAIC-seq to a B-cell lymphoblastoid cell line to find that those cells, too, exhibited a bimodal m⁶A distribution, but had higher methylation levels than the stem cell line did, suggesting cell-type specific differences in methylation variability.

Giallourakis and his colleagues also found that methylated and nonmethylated transcript isoforms
of the same genes differed in their use of tandem alternative polyadenylation (APA) sites. Methylated transcripts were more likely to rely on nearby APA sites and therefore also have shorter 3' UTRs than nonmethylated transcripts, they reported. Still, for a portion of genes, they noted that the reverse occurred.

Since m⁶A modifications are associated with RNA degradation, this and other findings led the researchers to describe a model in which m⁶A methylation and the degradation of shorter 3'-UTR isoforms using nearby poly(A) sites led to their increase in the m⁶A-positive fraction and concurrent decrease in the total steady-state RNA population.

In turn, this suggested to the researchers that m⁶A or the m⁶A methyltransferase might affect APA choice. However, when they knocked down or knocked out m⁶A writers in a number of cell lines, they saw limited APA changes, indicating that it cannot explain the link between m⁶A and proximal APA usage.

They also noted that there was differential binding of miRNAs and RNA-binding proteins in the UTR regions and that these transcripts might have different metabolic roles.

"Given the increasing recognition of other RNA modifications, such as 5mC, m1A, and pseudouridylation (ψ), our results may point to an even greater heterogeneity and isoform complexity of the transcriptome as part of an RNA epitranscriptome code," Giallourakis and his colleagues wrote.