In recent years, the regulation of RNA modification-mediated gene expression has become one of the hotspots in the field of epigenetic research, with m6A RNA modification being one of the mainstream research directions. I happened to come across an article about single cell m6A sequencing recently published in Molecular Cell by Kate Meyer's research group at Duke University, titled "scDART-seq Reveals Distinct m6A Signatures and mRNA Methylation Heterogeneity in Single Cells"

Source《scDART-seq reveals distinct m6A signatures and mRNA methylation heterogeneity in single cells ，Kate Meyer》

"Single-cell m6A" - wouldn't it be fantastic for cell typing? (*^_^*) When recalling the early days of single-cell sequencing being introduced, various famous literatures feature the same research ideas, especially in the developmental biology, and it is just a change in form but not in content.

Existing methods for m6A modification sequencing are mostly unable to be applied directly at the single-cell level. However, the role of m6A modifications in cell development and differentiation is of paramount importance. Therefore, the author innovatively developed a single-cell m6A sequencing technology and conducted relevant exploratory research. Before discussing this paper, let's first delve into the background and introduce some m6A sequencing technologies commonly used.

Research Background

The earliest and most widely used method is the m6A-seq technology based on m6A antibody immunoprecipitation (Dominissini et al., 2012). The method is to capture mRNA using oligo dT, fragment RNA, and then enrich m6A-containing RNA fragments using m6A antibodies, and finally proceed with the RNA library preparation. The shortcomings of this initial approach are quite evident:

1)  Low resolution, with the size of m6A-modified RNA fragments typically limited to around 100 nt;

2) Unquantifiable, making it impossible to accurately determine the specific proportion of methylation sequences;

3）Requires a high-input of RNA for the antibody-based IP system.

Source《Topology of the human and mouse m6A RNA methylomes revealed by m6A-seq，Dan Dominissini》

Dominissini, in his 2012 article, was well aware that the input of RNA was high (400 µg mRNA or 2.5 mg total RNA), and surely understood that such a large amount of RNA starting material is impractical for many experiments. Therefore, Dominissini quickly published another article in 2013 to improve the input (5 µg mRNA or 300 µg total RNA), which greatly expanded the applicability of this method.

In fact, in addition to Dominissini's article published in NATURE in May 2012, there was another article published in CELL in June titled "Comprehensive Analysis of mRNA Methylation Reveals Enrichment in 3' UTRs and Near Stop Codons". The first author of this article is none other than Professor Meyer himself, who is the central figure of conducting single-cell m6A sequencing today, and the technology he used at this time is not very different in principle from Professor Dominissini's method.

Regardless, these three articles can be considered the first instances of large-scale, high-throughput identification of m6A methylation levels in humans and mice at the transcriptional level. The core of these methods involves using m6A antibodies for specific binding to m6A-contained mRNA fragments, followed by high-throughput sequencing. This approach is currently the most successful commercialized technology for m6A sequencing in theory.

There is room for further optimization in the m6A sequencing technologies described in the aforementioned articles. Therefore, in recent years, a series of creative improvements in m6A sequencing technologies have sprung up like mushroom. I have compiled most of the existing m6A modification sequencing methods (thanks to Professor Jia Guifang from the Department of Chemistry at Peking University for her review article "Advances in Biological Functions of RNA Chemical Modification m6A", which has significantly reduced my workload). These methods can be broadly categorized into two approaches.

Approach 1

The first approach is still based on antibody enrichment, same as the initial article:

If you want to publish high-impact articles using antibody enrichment, you need to excel in it. Even the method like the "low-input MeRIP-seq" mentioned in the last article was only published in PLOS Biology. So, it's not surprising that various fancy m6A sequencing methods independent of antibodies have emerged.

Approach 2

The second approach initially utilizes certain sequence motifs associated with m6A or features related to m6A metabolism:

DART-seq

m6A modification motifs exhibit a high degree of conservation: Rm6ACH (R = A/G, H = A/C/U), where the modified base A is followed by a conserved base C. So, if the cytidine deaminase APOBEC1, capable of C-U conversion, is recruited to m6A-modified loci, APOBEC1 can change the base C to U after the m6A modification locus, thus revealing the specific locus of m6A modification.

However, APOBEC1 lacks the ability to specifically recognize m6A. How can it be recruited to m6A modification loci? This is where another protein from the YTH protein family (RNA methylated reading proteins) comes into play. YTH proteins have the ability to specifically recognize m6A.-binding conserved motif.

So, by creating a fusion protein of YTH and APOBEC1, would it be possible to specifically convert the C to U after m6A modification loci, thereby obtaining the specific locations of m6A modifications? This is the strategy employed by DART-seq: Introduce the APOBEC1-YTH fusion protein into target cells, convert C to U after m6A modification within the cells, and then align with genomic information to obtain the m6A modification loci.

Source《DART-seq: an antibody-free method for global m6A detection，Kate Meyer》

In fact, the drawbacks of this technology are obvious, and the scDART-seq, which is built upon this technology, also faces similar issues:

1) DART-seq relies on the efficiency of cell transfection, requiring stable overexpression of the fusion protein inside the cell, samples are limited to in vitro experiments, and it is not suitable for samples that are difficult or impossible to transfect;

2）Only 60% of the RNA substrates using the YTH domain contain m6A, and not all m6A-containing RNA molecules are bound by the YTH domain;

3) The steric hindrance formed by the YTH domain binding to m6A affects the efficiency of C to U conversion.

scDART-seq

scDART-seq is an upgraded version of DART-seq, and seems like to simply conduct single-cell sequencing on cells with overexpression APOBEC1-YTH fusion protein. Let's take a look at how the authors did it

1.Stable Transfected Cell Lines and Sequencing

1) To determine whether DART-seq can be used to locate m6A loci in individual cells, the authors established stable cell lines expressing APOBEC1-YTH in the classical HEK293T cell line, and simultaneously used APOBEC1-YTHmut (lacking m6A binding domain) as a control for off-target analysis.

2）Following inducible expression, a large number of cells were subjected to DART-seq. The Bullseye tool, which recognizes C to U conversions, was used to identify m6A loci throughout the transcriptome. Simultaneously, the identification of m6A loci in single cells was compared using 10×Genomics and Bullseye.

2.Comparison of DART-seq and scDART-seq Results

1) The authors discovered 16,934 high-confidence m6A modification loci in 3,844 RNA molecules among 10,352 cells, and the results exhibited good reproducibility. These m6A loci were highly consistent with the results obtained from bulk m6A-seq, indicating that scDART-seq feature a high level of accuracy and reproducibility.

2) Single-cell m6A loci exhibited characteristics similar to bulk-level, including strong enrichment near the termination codon and within the conserved RAC (R = A/G) sequences.

Source《scDART-seq reveals distinct m6A signatures and mRNA methylation heterogeneity in single cells ，Kate Meyer》

However, the proportion of individual cells in which RNA is methylated exhibits a high degree of heterogeneity:

1) For instance, both TPT1 and RPL34 are highly expressed in most cells, but TPT1 is methylated in most cells, whereas RPL34 is rarely methylated.

2）The scDART-seq libraries based on 10xGenomics were further validated using SMART-seq2. Of course, the authors also conducted additional experiments to confirm the feasibility of scDART-seq.

Source《scDART-seq reveals distinct m6A signatures and mRNA methylation heterogeneity in single cells ，Kate Meyer》

3.Cell Clustering Using m6A Sequencing Results

Last but not least, analyze methylation patterns in different cellular states using scDART-seq. To determine whether mRNA exhibits differential methylation across the entire cell cycle, the authors identified cells in the G1, S, and G2/M phases:

Overall, there was no significant difference in m6A levels among cells in different cell cycles. However, some transcripts exhibited cell cycle-dependent changes in m6A levels that were unrelated to gene expression. For example, TET1 and TOPBP1 showed no significant differences in expression levels across different cell cycles, but their methylation levels increased with the progression of the cell cycle in an increasing number of cells.

Source《scDART-seq reveals distinct m6A signatures and mRNA methylation heterogeneity in single cells ，Kate Meyer》

So, since RNA expression is no longer able to distinguish these cells, can the differences in m6A modification levels be used to distinguish cell subsets?

To confirm this, the authors conducted clustering analysis on the cells using the results from scDART-seq:

The results revealed two distinct cell clusters (although the clustering effect is not very pronounced), and most encoding RNA-regulated and binding proteins at the differential methylation loci in these two cell clusters, and suggested that the methylation of transcripts encoding RNA factors may contribute to cell typing.

Source《scDART-seq reveals distinct m6A signatures and mRNA methylation heterogeneity in single cells ，Kate Meyer》

I indeed came across an article that used m6A methylation status for cell clustering, but there's always a sense of unfulfilled potential. After all, based on the results from scDART-seq, this method still has some way to go before it becomes truly applicable. Hopefully, we'll see more articles of this kind in the future, delving deep into the biological mysteries of cell development and differentiation.

In recent articles related to m6A-seq, Chinese scholars have made a significant contribution as well. Nothing more to say, keep up the good work!