Transcriptome-wide identification of 5-methylcytosine by deaminase and reader protein-assisted sequencing

Reviewer #2 (Public review):

The fledgling field of epitranscriptomics has encountered various technical roadblocks with implications as to the validity of early epitranscriptomics mapping data. As a prime example, the low specificity of (supposedly) modification-specific antibodies for the enrichment of modified RNAs, has been ignored for quite some time and is only now recognized for its dismal reproducibility (between different labs), which necessitates the development of alternative methods for modification detection.

Furthermore, early attempts to map individual epitranscriptomes using sequencing-based techniques are largely characterized by the deliberate avoidance of orthogonal approaches aimed at confirming the existence of RNA modifications that have been originally identified.

Improved methodology, the inclusion of various controls, and better mapping algorithms as well as the application of robust statistics for the identification of false-positive RNA modification calls have allowed revisiting original (seminal) publications whose early mapping data allowed making hyperbolic claims about the number, localization and importance of RNA modifications, especially in mRNA. Besides the existence of m6A in mRNA, the detectable incidence of RNA modifications in mRNAs has drastically dropped.

As for m5C, the subject of the manuscript submitted by Zhou et al., its identification in mRNA goes back to Squires et al., 2012 reporting on >10.000 sites in mRNA of a human cancer cell line, followed by intermittent findings reporting on pretty much every number between 0 to > 100.000 m5C sites in different human cell-derived mRNA transcriptomes. The reason for such discrepancy is most likely of a technical nature. Importantly, all studies reporting on actual transcript numbers that were m5C-modified relied on RNA bisulfite sequencing, an NGS-based method, that can discriminate between methylated and non-methylated Cs after chemical deamination of C but not m5C. RNA bisulfite sequencing has a notoriously high background due to deamination artifacts, which occur largely due to incomplete denaturation of double-stranded regions (denaturing-resistant) of RNA molecules. Furthermore, m5C sites in mRNAs have now been mapped to regions that have not only sequence identity but also structural features of tRNAs. Various studies revealed that the highly conserved m5C RNA methyltransferases NSUN2 and NSUN6 do not only accept tRNAs but also other RNAs (including mRNAs) as methylation substrates, which in combination account for most of the RNA bisulfite-mapped m5C sites in human mRNA transcriptomes. Is m5C in mRNA only a result of the Star activity of tRNA or rRNA modification enzymes, or is their low stoichiometry biologically relevant?
In light of the short-comings of existing tools to robustly determine m5C in transcriptomes, other methods, like DRAM-seq, allowing to map m5C independently of ex situ RNA treatment with chemicals, are needed to arrive at a more solid "ground state", from which it will be possible to state and test various hypotheses as to the biological function of m5C, especially in lowly abundant RNAs such as mRNA.

Importantly, the identification of >10.000 sites containing m5C increases through DRAM-Seq, increases the number of potential m5C marks in human cancer cells from a couple of 100 (after rigorous post-hoc analysis of RNA bisulfite sequencing data) by orders of magnitude. This begs the question, whether or not the application of these editing tools results in editing artefacts overstating the number of actual m5C sites in the human cancer transcriptome.

Remaining comments after resubmission:

(1) The use of two m5C reader proteins is likely a reason for the high number of edits introduced by the DRAM-Seq method. Both ALYREF and YBX1 are ubiquitous proteins with multiple roles in RNA metabolism including splicing and mRNA export. It is reasonable to assume that both ALYREF and YBX1 bind to many mRNAs that do not contain m5C.
To substantiate the author's claim that ALYREF or YBX1 binds m5C-modified RNAs to an extent that would allow distinguishing its binding to non-modified RNAs from binding to m5C-modified RNAs, it would be recommendable to provide data on the affinity of these, supposedly proven, m5C readers to non-modified versus m5C-modified RNAs. To do so, this reviewer suggests performing experiments as described in Slama et al., 2020 (doi: 10.1016/j.ymeth.2018.10.020). Mind you that using dot blots like in so many published studies to show modification-specific antibody or protein binding, is insufficient as an argument because no antibody, nor protein encounters nanograms to micrograms of a specific RNA identity in a cell. This issue remains a major caveat in all studies using so-called RNA modification reader proteins as bait for detecting RNA modifications in epitranscriptomics research and becomes a pertinent problem, if used as a platform for base-editing similar to the work presented in this manuscript.

(2) Using sodium arsenite treatment of cells as a means to change the m5C status of transcripts through the downregulation of the two major m5C writer proteins NSUN2 and NSUN6 is problematic and the conclusions from these experiments are not warranted. Sodium arsenite is a chemical that poisons every protein containing thiol groups. Not only do NSUN proteins contain cysteines but also the base editor fusion proteins. Arsenite will inactivate these proteins, hence the editing frequency will drop, as observed in the experiments shown in Figure 5, which the authors explain with fewer m5C sites to be detected by the fusion proteins.

(3) The authors should move high-confidence editing site data contained in Supplementary Tables 2 and 3 into one of the main Figures to substantiate what is discussed in Figure 4A. However, the data needs to be visualized in another way then excel format. Furthermore, Supplementary Table 2 does not contain a description of the columns, while Supplementary Table 3 contains a single row with letters and numbers.

Author response:

The following is the authors’ response to the original reviews.

Reviewer #1:

In this manuscript, Zhou et al describe a deaminase and reader protein-assisted RNA m5C sequencing method. The general strategy is similar to DART-seq for m6A sequencing, but the difference is that in DART-seq, m6A sites are always followed by C which can be deaminated by fused APOBEC1 to provide a high resolution of m6A sites, while in the case of m5C, no such obvious conserved motifs for m5C sites exist, therefore, the detection resolution is much lower. In addition, the authors used two known m5C binding proteins ALYREF and YBX1 to guide the fused deaminases, but it is not clear whether these two binding proteins can bind most m5C sites and compete with other m5C binding proteins.

Thank you for your kind suggestion. RNA affinity chromatography and mass spectrometry analyses using biotin-labelled oligonucleotides with or without m5C were performed in previous reports (doi:10.1038/cr.2017.55 and doi: 10.1038/s41556-019-0361-y), and the results showed that ALYREF and YBX1 had a more prominent binding ability to m5C -modified oligonucleotides. Moreover, these two m5C -binding proteins are also responsible for mRNA m5C binding, so we chose to use their ability to bind targeted m5C to construct a DRAM detection system in anticipation of transcriptome-wide m5C detection. We hope to propose a suitable detection strategy for RNA m5C, and there will certainly be room for optimization of the DRAM system in the future with more in-depth studies of m5C binding proteins. We have discussed the above issue in lines 75-82 and 315-318.

It is well known that two highly modified m5C sites exist in 28S RNA and many m5C sites exist in tRNA, the authors should validate their methods first by detecting these known m5C sites and evaluate the possible false positives in rRNA and tRNA.

Thank you for your kind suggestion. We attempted PCR amplification of sequences flanking m5C sites 3782 and 4447 on 28S rRNA, as well as multiple m5C sites on tRNA, including m5C48 and m5C49 on tRNAVal, m5C48 and m5C49 on tRNAAsp, and m5C48 on tRNALys.

However, Sanger sequencing revealed no valid mutations, which was implemented in Figure S3. We believe this outcome indicates that the DRAM system is more suited for transcriptome-wide m5C detection of mRNAs. This is supported by current reports that ALYREF and YBX1 are responsible for the m5C-binding proteins of mRNAs (doi:10.1038/cr.2017.55 and doi: 10.1038/s41556-019-0361-y). The above results and descriptions were added to lines 136-143.

In mRNA, it is not clear what is the overlap between the technical replicates. In Figures 4A and 4C, they detected more than 10K m5C sites, and most of them did not overlap with sites uncovered by other methods. These numbers are much larger than expected and possibly most of them are false positives.

Thank you for your kind suggestion. We observed significant overlap between the technical repeats by comparing the data across biological repeats, as shown in Figure S4C and described in lines 174-175. We considered m5C modification in a region only when editing events were detected in at least two biological replicates, ensuring a high-stringency screening process (details seen in the revised method in lines 448-455 and Figure 3F). With more in-depth research into m5C readers, we aim to achieve more accurate detection in the future.

Besides, it is not clear what is the detection sensitivity and accuracy since the method is neither single base resolution nor quantitative.

Thank you for your suggestion. As shown in Figure 3G, we found that the editing window of the DRAM system exhibited enrichment of approximately 20 bp upstream and downstream of the m5C site. Previous reports identified Type I m5C sites, which tend to have a downstream "NGGG" motif, and Type II m5C sites, which often contain a downstream "UCCA" motif. However, these m5C motifs do not fully characterize all m5C sites, and their presence downstream of an m5C site is not guaranteed (doi: 10.1038/s41594-019-0218-x). This limitation complicates single-base resolution analysis by the DRAM system. Nevertheless, we believe that with further exploration of m5C sequence features, precise single-base resolution detection can be achieved in the future. This point is also discussed in lines 314-322.

Regarding the quantitative level of the assay, we conducted additional experiments by progressively reducing the expression levels of the fusion proteins. Sanger sequencing revealed that the editing efficiency of A-to-G and C-to-U within the m5C region significantly decreased as fusion protein expression diminished (Figure S9). These findings suggest that the DRAM system's transfection efficiency is concentration-dependent and that the ratio of editing efficiency to transfection efficiency could aid in the quantitative analysis of m5C using the DRAM system. The relative results were supplemented in Figure S9 and discussed in lines 263-271.

There are no experiments to show that the detected m5C sites are responsive to the writer proteins such as NSUN2 and NSUN6, and the determination of the motifs of these writer proteins.

Thank you for your kind suggestion. We have performed a motif enrichment analysis based on the sequences spanning 10 nt upstream and downstream of DRAM-editing sites. The relative results of this analysis were supplemented in Figure S4D and lines 168-171. Unfortunately, we did not identify any clear sequence preferences for the m5C sites catalyzed by the methyltransferases NSUN2 and NSUN6, which have previously been associated with “G”-rich sequences and the “CUCCA” motif. This limitation is mainly due to the DRAM detection system’s inability to achieve single-base resolution for m5C detection, which is also explained in the above response.

Reviewer #2:

(1) The use of two m5C reader proteins is likely a reason for the high number of edits introduced by the DRAM-Seq method. Both ALYREF and YBX1 are ubiquitous proteins with multiple roles in RNA metabolism including splicing and mRNA export. It is reasonable to assume that both ALYREF and YBX1 bind to many mRNAs that do not contain m5C.

To substantiate the author's claim that ALYREF or YBX1 binds m5C-modified RNAs to an extent that would allow distinguishing its binding to non-modified RNAs from binding to m5C-modified RNAs, it would be recommended to provide data on the affinity of these, supposedly proven, m5C readers to non-modified versus m5C-modified RNAs. To do so, this reviewer suggests performing experiments as described in Slama et al., 2020 (doi: 10.1016/j.ymeth.2018.10.020). However, using dot blots like in so many published studies to show modification of a specific antibody or protein binding, is insufficient as an argument because no antibody, nor protein, encounters nanograms to micrograms of a specific RNA identity in a cell. This issue remains a major caveat in all studies using so-called RNA modification reader proteins as bait for detecting RNA modifications in epitranscriptomics research. It becomes a pertinent problem if used as a platform for base editing similar to the work presented in this manuscript.

We thank the reviewer for the valuable suggestion. Previous studies have shown that while ALYREF and YBX1 can bind mRNAs without the m5C modification, their binding affinity for m5C-modified oligonucleotides is significantly higher than for unmethylated controls. This has been demonstrated through experiments such as in vitro tractography, electrophoretic mobility shift assay (EMSA) (doi:10.1038/cr.2017.55), and UHPLC-MRM-MS/MS. Additionally, isothermal titration calorimetry measurements and PAR-CLIP experiments have shown that mutations in the key amino acids responsible for m5C binding in ALYREF and YBX1 result in a significant reduction in their ability to m5C (doi: 10.1038/s41556-019-0361-y).

Although Me-RIP analysis was unsuccessful in our laboratory, likely due to the poor specificity of the m5C antibody, we alternatively performed RNA pulldown experiments. These experiments verified that the ability of DRAMmut-expressing proteins to bind RNA with m5C modification was virtually absent compared to DRAM-expressing proteins, while their binding ability with non-modified RNA was not significantly affected. The relative RNA pulldown results were supplemented in Figure S1E, S1F and lines 110-111. Therefore, we believe that by integrating DRAMmut group, our DRAM system could effectively exclude the false-positive mutations caused by unspecific binding of DRAM’s reader protein to non-m5C-modified mRNAs.

(2) Since the authors use a system that results in transient overexpression of base editor fusion proteins, they might introduce advantageous binding of these proteins to RNAs. It is unclear, which promotor is driving construct expression but it stands to reason that part of the data is based on artifacts caused by overexpression. Could the authors attempt testing whether manipulating expression levels of these fusion proteins results in different editing levels at the same RNA substrate?

Thank you for pointing this out. To investigate how different expression levels of these proteins influence A-to-G and C-to-U editing within the same m5C region, we conducted a gradient transfection using plasmid concentrations of 1500 ng, 750 ng and 300 ng. This approach allowed us to progressively reduce the expression levels of the fusion proteins. Sanger sequencing revealed that the editing efficiency of A-to-G and C-to-U within the m5C region significantly decreased as fusion protein expression diminished. These findings suggest that the transfection efficiency of the DRAM system is concentration-dependent and that the ratio of editing efficiency to transfection efficiency may assist in the quantitative analysis of m5C using the DRAM system. The relative results and hypotheses were added and discussed in Figure S9 and lines 263-271 of the revised manuscript.

(3) Using sodium arsenite treatment of cells as a means to change the m5C status of transcripts through the downregulation of the two major m5C writer proteins NSUN2 and NSUN6 is problematic and the conclusions from these experiments are not warranted. Sodium arsenite is a chemical that poisons every protein containing thiol groups. Not only do NSUN proteins contain cysteines but also the base editor fusion proteins. Arsenite will inactivate these proteins, hence the editing frequency will drop, as observed in the experiments shown in Figure 5, which the authors explain with fewer m5C sites to be detected by the fusion proteins.

Thank you for pointing this out. We used bisulfite sequencing PCR to determine that the m5C levels in RPSA and AP5Z1 were significantly reduced after sodium arsenite treatment. This was followed by a significant decrease in editing frequency detected by the DRAM system in sodium arsenite-treated samples compared to untreated samples. This reduction aligns with the decreased editing efficiency observed in methyltransferase-deficient cells (as shown in Figures 2G and 2H), which initially convinced us that these results reflected the DRAM system's ability to monitor dynamic changes in m5C levels.

However, as the reviewer pointed out, sodium arsenite treatment could potentially inactivate the fusion proteins, leading to the observed reduction in editing efficiency. This possibility has not been conclusively ruled out in our current experiments. Optimizing this validation may require the future development of more specific m5C inhibitors. In light of this, we have revised our previous results and conclusions in lines 235-244, and discussed these points in lines 308-315.

(4) The authors should move high-confidence editing site data contained in Supplementary Tables 2 and 3 into one of the main Figures to substantiate what is discussed in Figure 4A. However, the data needs to be visualized in another way than an Excel format. Furthermore, Supplementary Table 2 does not contain a description of the columns, while Supplementary Table 3 contains a single row with letters and numbers.

Thank you for your kind suggestion. We have visualized the data from Supplementary Tables 2 and 3 into Figure 3F, presenting it as a screening flowchart for high-confidence editing sites. In Supplementary Table 3, we have displayed only the DRAM-mutated genes, which is why it contains a single row with letters and numbers. As requested, we have included descriptions of each column and reorganized the Supplementary table 2 and 3 accordingly.

(5) The authors state that "plotting the distribution of DRAM-seq editing sites in mRNA segments (5'UTR, CDS, and 3'UTR) highlighted a significant enrichment near the initiation codon (Figure 3F).", which is not true when this reviewer looks at Figure 3F.

Thank you for your kind suggestion, and we replaced the expression of " near the initiation codon" with "in the CDS" in lines 192-193.

(6) The authors state that "In contrast, cells expressing the deaminase exhibited a distinct distribution pattern of editing sites, characterized by a prevalence throughout the 5'UTR.", which is not true when this reviewer looks at Figure 3F.

Thank you for your kind suggestion. This distribution was actually characterized by a prevalence throughout the "3'UTR", but not "5'UTR". We have also made the necessary changes in lines 193-195.

(7) The authors claim in the final conclusion: "In summary, we developed a novel deaminase and reader protein assisted RNA m5C methylation approach...", which is not what the method entails. The authors deaminate As or Us close to 5mC sites based on the binding of a deaminase-containing protein.

Thank you for your kind suggestion, and we have made the necessary changes in lines 331-334.

(8) The authors claim that "The data supporting the findings of this study are available within the article and its Supplementary Information." However, no single accession number for the deposited sequencing data can be found in the text or the supplementary data. Without the primary data, none of the claims can be verified.

Thank you for pointing this out. The sequencing data from this study has already been deposited to the GEO database (GEO assession number: GSE254194, GEO token：ororioukbdqtpcn), and we will ensure it is made publicly available in a timely manner.

(a) To underscore point (1), a recent publication (https://doi.org/10.1038/s41419-023-05661-y) reported: "To further identify the potential mRNAs regulated by ALYREF, we performed RNA-seq analysis in control or ALYREF knockdown T24 cells. After knockdown of ALYREF, 143 mRNAs differentially expressed, including 94 downregulated mRNAs (NC reads >100, |Fold change | >1.5, P-value <0.05). Functional enrichment analysis using Kyoto Encyclopedia of Genes and Genomes (KEGG) indicated that regulated mRNAs by ALYREF are chiefly enriched in canonical cancer-related pathways (Fig. S4A), including TGF-β signaling, MAPK signaling, and NF-κB signaling, strongly supporting the oncogenic function of ALYREF in tumor progression. Among these 94 downregulated genes, 11 mRNA showed a significant reduction in m5C methylation after NUSN2 silencing in T24 cells, combined with previously transcriptome-wide RNA-BisSeq data of T24 cells [21] (Fig. 4A)."

These results translate into 94 mRNAs are regulated by ALYREF in bladder cancer-derived cells. From those, very few (11) mRNA identities respond to NSUN2-dependent RNA methylation mediated by ALYREF binding.The question then arises, is that number sufficient to claim that ALYREF is a m5C-binding protein?

And if so, how does the identification of 10.000+ edits by DRAM-Seq compare with the 94 mRNAs that are regulated by ALYREF? Were these 94 mRNAs identified by DRAM-Seq.

Thank you for your kind suggestion. Previous reports by Yang et al. ( doi: 10.1038/cr.2017.55), including the literature you refer to, have detailed the close relationship between ALYREF and m5C modification, and the ALY/REF export factor (ALYREF) was identified as the first nuclear m5C reader, and it was demonstrated that many mRNAs are regulated by ALYREF, and is therefore considered to be an m5C-binding protein.

As required, by comparing the DRAM-edited mRNAs with the reported 94 mRNAs, we found that only 55.32% of the 94 mRNAs regulated by ALYREF could be detected by the DRAM system. This indicates that the DRAM system specifically targets certain mRNAs, as illustrated in Figure S4E. The relevant results were described and discussed in lines 175-179.

(b) Line 123:

"The deep sequencing results showed that the deamination rates of RPSA and SZRD1 were 75.5% and 27.25%, respectively. (Fig. 2A, B)."

The Figure shows exactly the opposite of bisulfite-mediated deamination. These are the cytosines that were not deaminated by the chemical treatment and therefore can be sequenced as cytosines and not thymidines. Hence, the term deamination rate is wrong.

Thank you for your kind suggestion. We have made the necessary change in lines 129-130 to change the deamination rates to m⁵C fraction.

(c) Line 157:

"DRAM-seq analysis further confirmed that DRAM was detected in an m5C-dependent manner, with minimal mutations in AP5Z1 and RPSA mRNAs in methyltransferase knockout cells compared to wild-type cells (Fig. 3C, D)."

There is no indication of what the authors mean by minimal mutation in these Figures. The term "minimal mutation" should be reconsidered as well.

Thank you for your kind suggestion. We intended to express that "Mutations in AP5Z1 and RPSA mRNA are reduced in methyltransferase-deficient cells." There was an issue with the initial formulation, and we have made the necessary changes in lines 165-167.

(d) Line 167:

"To further delineate the characteristics of the DRAM-seq data, we compared the distribution of DRAM-seq editing sites within the gene structure, specifically examining their occurrences in the 5'untranslated region (5'UTR), 3' untranslated region (3'UTR), CDS and ncRNA."

Which part of a coding RNA is meant by "ncRNA"?

Thank you for pointing this out. This was actually the Intergenic or Intron region, but not ncRNA. We have also corrected this labelling in Figure 3G and lines 186-189 of the revised manuscript.

(e) Line 189:

"Subsequently, we assessed the capacity of DRAM-seq to detect m5C on a transcriptome-wide scale, comparing its performance to BS-seq that have been previously reported with great authority."

The term "great authority" is not a scientific term. Please, remove adulation to senior authors.

Thank you for your kind suggestion. We removed this unsuitable expression and made the necessary changes in lines 207-208.

(f) Line 233:

"Several experiments have highlighted the requirement of 100-500 ng of RNA for m5C-RIP-seq, while BS-seq necessitates an even more demanding 500-750 μg of RNA21,25,61."

This reviewer doubts that RNA bisulfite sequencing required half to one mg of RNA input. Please, check these references.

Thank you for your kind suggestion. According to the references, we corrected μg to ng and made the necessary changes in lines 251-252.

(g) Line 247:

"Several experiments have highlighted the requirement of 100-500 ng of RNA for m5C-RIP-seq, while BS-seq necessitates an even more demanding 500-750 μg of RNA21,25,61."

This reviewer doubts that RNA bisulfite sequencing requires half to one mg of RNA input. Please, check these references.

Thank you for your kind suggestion. According to the references, we corrected μg to ng and made the necessary changes in lines 251-252.

(h) Line 292:

"Since m5C lacks a fixed motif, DRAM has an apparent limitation in achieving single-base resolution for detecting m5C."

m5C deposition by NSUN2 and NSUN6 occurs in particular motifs that were coined Type I and II motifs. Hence, this statement is not correct.

Thank you for your kind suggestion. Previous reports identified Type I m5C sites, which tend to have a downstream "NGGG" motif, and Type II m5C sites, which often contain a downstream "UCCA" motif. However, these m5C motifs do not fully characterize all m5C sites, and their presence downstream of an m5C site is not guaranteed (doi: 10.1038/s41594-019-0218-x ). Therefore, we have corrected the expression “fixed motif” to “fixed base composition for characterizing all m5C modification sites” in lines 317.

(i) Line 390:

"1 μl of total cellular RNA was used for sequencing library gene..."

1 uL does not allow us to deduce which RNA mass was used for cDNA synthesis.

Thank you for your kind suggestion. According to our cDNA synthesis protocol, we corrected “1μl” to “1μg” in lines 422-423.

(j) Line 405:

"...was assessed on the Agilent 5400 system (Agilent, USA) and quantified by QPCR (1.5 nM)"

What does the 1.5 nM refer to in this sentence?

Thank you for your kind suggestion. Here, "1.5nM" means that the concentration of the constructed library should be no less than 1.5nM. We have also revised this expression in the methods in lines 436-438.