Hakai is required for stabilization of core components of the m6A mRNA methylation machinery

N6-methyladenosine (m6A) is the most abundant internal modification on mRNA which influences most steps of mRNA metabolism and is involved in several biological functions. The E3 ubiquitin ligase Hakai was previously found in complex with components of the m6A methylation machinery in plants and mammalian cells but its precise function remained to be investigated. Here we show that Hakai is a conserved component of the methyltransferase complex in Drosophila and human cells. In Drosophila, its depletion results in reduced m6A levels and altered m6A-dependent functions including sex determination. We show that its ubiquitination domain is required for dimerization and interaction with other members of the m6A machinery, while its catalytic activity is dispensable. Finally, we demonstrate that the loss of Hakai destabilizes several subunits of the methyltransferase complex, resulting in impaired m6A deposition. Our work adds functional and molecular insights into the mechanism of the m6A mRNA writer complex.

important for in vivo m6A methylation as evidenced by reduced and altered methylation of RNA and by altered slicing of the Sxl gene, 5) that Hakai is needed to stabilize other MACOM components and that it does so without need for its reputed E3 ubiquitin ligase activity. The claims are all well supported by the experiments presented and represent a significant gain in our understanding of the role that Hakai plays in RNA methylation.
My only reservations about the paper related to the descriptions and presentation of the genetic experiments which make the arguments hard to follow, obscure some data, and suggest things are simpler than the data allow. The actual experiments are fine. All the following comments relate to the experiments presented in Fig. 3.
The experiment presented in Fig. 3E uses a sensitized genetic background with a low-level of femalespecific lethality in which any defects in m6A components would be expected to exacerbate female lethality. This is exactly what is observed with Hakai and with the control Mettl3 control. The problem is the simplified description of the logic of the experiment. The female lethality depends on reduced expression of the Sxl isoforms that initiate Sxl autoregulation (caused by both the loss of maternal da, and by the loss of one copy of zygotic Sxl) and by reduced dose of the products that complete and maintain autoregulated splicing (a consequence of Sxlf7bo). I think the solution here is to avoid the details as they distract rather than help. What really matters is that this is genetic system that is expected to make Sxl splicing hypersensitive to reductions in m6 complex function. I also think it would be more clear to drop the mention of dosage compensation. The proximal cause of the lethality is that reduced hakai dose causes a reduction in Sxl expression that leads to femalelethality.
In the experiment in Fig. 3F it is impossible to determine the actual number of flies analyzed. The cross was between hak2/Balancer and Df(hak)/Balancer flies, but the controls are not specified. We are told in the the text that 67 hak2/Df flightless males emerged and that a much smaller, but unspecified (in the text), and illegible (in the figure), number of females emerged. We are thus left with trying to figure how 67 experimental males (with about 90% viability) relates to 339 control flies. To fix the problems, the # of experimental females observed should be stated in the text, and the control flies precisely defined. If my hunch is correct, and the authors did not distinguish the control hak2/Balancer and Df(hak)/Balancer flies, they should define the viability standard as one half the number of heterozygous Balancer-bearing males or females (but not both). That way the reference value would presumably be about 75 flies (67/0.90).
In the description of experiments in Fig. 3G, the vir2F allele is described as a dominant negative. In fact, as experiment in Fig. 3H illustrates, and the published literature shows, the vir2F allele is recessive. What the experiment in Fig. 3G documents is that females bearing only the vir2F allele (vir2f/Df) are inviable but that they are largely rescued if the flies are also heterozygous for hak1. This is a solid result that strengthens the paper, but it's very hard to interpret with respect to mechanism. What the authors want to argue is that the "dominant negative" [sic] vir2F allele prevents Sxl recruitment (itself a vague notion) and that "weakening" (also vague) the m6A complex allows Sxl better access and thus rescues female viability. It may well be that the mechanism of rescue by reducing hakai dosage is by destabilizing the the vir2F-containing MACOM complex but that is not easily squared with the published finding that reducing the dose of the MAC complex component, mettl3 (Ime4), also rescues vir2F-dependent female viability. I think the solution is for the authors is to take the powerful result as a fact, but resist the speculative interpretation.
The experiment in Fig. 3H are powerful and appropriately described. (They also show that vir2F is, at best, sparingly dominant to the partially defective virts allele.) I would, however, encourage two modifications to the figure. First, correct the typo on the y-axis "apperence", and second, modify the genotypes so that they more clearly reflect the fact that hakai and vir are on the same chromosome. My suggestion is to label the genotypes of the two right hand sections as + vir / + vir and hak1 vir/+ vir. Doing so provides an additional advantage in that would show how the cross was actually done.
Reviewer #3 (Remarks to the Author): I don't understand the interpretation of Fig 5A by the authors and, honestly, I think that they may underestimate the value of their data set.
On p14 the authors state: "Cells isotopically labelled with heavy amino acids were depleted for Hakai and cells isotopically labelled with light amino acids served as a control in the forward experiment. A vice versa depletion was performed in the reverse experiment ( Supplementary Fig. 8C).
[…] We found over 3000 ubiquitination sites, but unexpectedly not a single site was reduced in response to Hakai depletion and only one site in SesB was 1.5-fold increased (Fig. 5A, Supplementary Table 3). Therefore, this experiment strongly suggests that Hakai does not act as a general E3 ubiquitin ligase in Drosophila cells. Alternatively, it is possible that its ubiquitination activity depends on specific external stimuli." On p32, however, they state that "For ubiquitinome and proteome analysis 50 mg of proteins of heavy and light replicate were joined in a 1:1 ratio as follows: for forward experiment heavy labelled cells with control KD and light labelled cells with Hakai KD were joined, and vice versa for reverse experiment." First of all, the number of diGly peptides obtained from 50 (!) mg of starting material seems somewhat low and at first I was under the impression that the limited coverage may be the problem here. However, upon reanalyzing the source data set using Suppl Table 3, I noticed something else and I think that the x-axis in Fig  This cloud is typical for ubiquitinome assays upon e.g. proteasome inhibition or knockout of a ubiquitin ligase. In conclusion, this means that, in contrast to the authors' own conclusion, in fact there seems to be (as expected?) a huge effect on the ubiquitinome upon Hakai KD.
Unless I'm missing something, I would suggest the authors to reanalyze this data set.
We thank the reviewers for their suggestions that help us improving the quality of our manuscript. A point-by-point response addressing each of the comments is presented below.
Overall no major experiments were requested. In Figure 2 we nevertheless add a new information regarding the binding of Human WTAP to VIRMA. We now show that instead of one, two VIRMA domains are involved in the interaction with WTAP, of which one is conserved in Drosophila. This new information extends the analysis of our interaction assays and does not modify the overall conclusions of the manuscript.
Reviewer #1 (Remarks to the Author): The m6A RNA methyltransferase complex containing METTL3 as the catalytic enzyme is a topic of great scientific interest. Previous studies from the authors have shed light on the composition, structure and biology of this complex. The METTL3 complex is composed of several components, including an E3 ubiquitin ligase called Hakai, whose molecular function is poorly understood. Here the authors show that the protein forms part of fly and human complexes, while its depletion in flies result in reduced RNA modification levels and methylation-dependent functions in vivo. Interestingly, they find that while the ubiquitination activity is not essential for in vivo function, the catalytic domain serves to dimerize the protein. Pointing to its critical role in complex integrity, loss of Hakai results in destabilization of several subunits of this complex, with dramatic consequences for methylation deposition in vivo. In my opinion, this is the most comprehensive study on the intercomponent interactions of this very important complex, examining it in both flies and human cells. It will be a valuable reference material for the m6A field. I support its immediate publication. I have only a few minor suggestions for text changes.
We like to thank the reviewer for his enthusiasm about our work and to support its publication.
1. Figure 4C. Why are there more differentially spliced events in KD of non-catalytic components than KD of the METTL3 itself? Please explain clearly in the text.
We made similar observations in our previous studies in Lence et al, 2016 andKnuckles et al, 2018. The stronger impact of MACOM components on splicing is consistent with the genetic studies showing that the loss of function of MACOM components induces fly lethality while flies mutant for Mett3 and Mettl14 are viable. This indicates that MACOM components bear additional role(s) beyond their involvement within the MAC complex. Whether this function also involves m6A through interaction with another methyltransferase or whether this is an m6A independent function remains to be investigated.
We have now included this paragraph: Note that depletion of MACOM components has a stronger impact on gene expression and splicing compared to the loss of MAC components. This is consistent with previous genetic data showing that Mettl3 and Mettl14 are dispensable for fly viability while MACOM subunits are not, supporting additional function(s) for MACOM components.
We amended the text as followed: "The Hakai ubiquitination domain but not its activity is required for m6A biogenesis" was replaced by "The Hakai ubiquitination domain but not its activity is required for maintaining MACOM integrity" 3. Figure 1 can definitely use a cartoon with the MACOM complex with names of the fly and human proteins indicated in it. It allows the reader to follow the interactions identified in Fig1-2 and Fig5-6.
As suggested we included a small table in Figure 1 with the names of MAC and MACOM components of both organisms. In addition we added a new Figure 7 with a model for the methyltransferase complex organization and the impact of Hakai on its integrity.

Reviewer
#2 (Remarks to the Author): The manuscript, "Hakai is required for stabilization of core components of the m6A mRNA methylation machinery" by Praveen Bawankar et al. is an interesting and important contribution to RNA Biology and of the role of RNA modification in gene regulation. The paper makes several key claims: 1) that Hakai gene product and its mammalian counterpart is a MACOM subunit, 2) the Hakai protein physically interacts with a subset of other MACOM complex members, 3, 4) that Hakai is important for in vivo m6A methylation as evidenced by reduced and altered methylation of RNA and by altered slicing of the Sxl gene, 5) that Hakai is needed to stabilize other MACOM components and that it does so without need for its reputed E3 ubiquitin ligase activity. The claims are all well supported by the experiments presented and represent a significant gain in our understanding of the role that Hakai plays in RNA methylation.
My only reservations about the paper related to the descriptions and presentation of the genetic experiments which make the arguments hard to follow, obscure some data, and suggest things are simpler than the data allow. The actual experiments are fine. All the following comments relate to the experiments presented in Fig. 3.
We are delighted by the reviewer appreciation of our progress in understanding the m6A mRNA methylation machinery. As detailed below, we have taken care to make the genetic analysis accessible to a broad readership by improving the presentation of the data. In addition, we have put all the details about the genetic crosses and the data from them into a supplementary source data file.
The experiment presented in Fig. 3E uses a sensitized genetic background with a low-level of female-specific lethality in which any defects in m6A components would be expected to exacerbate female lethality. This is exactly what is observed with Hakai and with the control Mettl3 control. The problem is the simplified description of the logic of the experiment. The female lethality depends on reduced expression of the Sxl isoforms that initiate Sxl autoregulation (caused by both the loss of maternal da, and by the loss of one copy of zygotic Sxl) and by reduced dose of the products that complete and maintain autoregulated splicing (a consequence of Sxlf7bo). I think the solution here is to avoid the details as they distract rather than help. What really matters is that this is genetic system that is expected to make Sxl splicing hypersensitive to reductions in m6 complex function. I also think it would be more clear to drop the mention of dosage compensation. The proximal cause of the lethality is that reduced hakai dose causes a reduction in Sxl expression that leads to female-lethality.
We have simplified the description of this genetic interaction experiment according to the reviewer suggestion. However, the cause of female lethality is mis-regulation of dosage compensation from reduced Sxl levels and we felt that dosage compensation needs to be mentioned in this context to understand the assay. We have tried our best to explain this better.
We have replaced: To test whether Hakai is required for Sxl autoregulation, we made use of a genetically sensitized background whereby one copy of daughterless (da), which is involved in Sxl transcription, and one copy of Sxl required for Sxl autoregulation were removed. As for Mettl3 null mutants, also removal of one copy of Hakai killed females (Fig. 3E), due to aberrant dosage compensation.

By
To test whether Hakai is required for Sxl autoregulation, we made use of a genetically sensitized background based on reduced Sxl levels by removal of one copy of daughterless (da), which is involved in Sxl transcription, and one copy of Sxl required for Sxl autoregulation. In the progeny of a cross between da Df /+; Mettl3 null /+ females and Sxl 7B0 null males, most females died (Fig. 3E). Likewise, also removal of one copy of Hakai killed females (Fig. 3E).
In the experiment in Fig. 3F it is impossible to determine the actual number of flies analyzed. The cross was between hak2/Balancer and Df(hak)/Balancer flies, but the controls are not specified. We are told in the the text that 67 hak2/Df flightless males emerged and that a much smaller, but unspecified (in the text), and illegible (in the figure), number of females emerged. We are thus left with trying to figure how 67 experimental males (with about 90% viability) relates to 339 control flies. To fix the problems, the # of experimental females observed should be stated in the text, and the control flies precisely defined. If my hunch is correct, and the authors did not distinguish the control hak2/Balancer and Df(hak)/Balancer flies, they should define the viability standard as one half the number of heterozygous Balancer-bearing males or females (but not both). That way the reference value would presumably be about 75 flies (67/0.90).
We have now explained the cross in the text and given the numbers. In addition, we have put all the details about the genetic crosses and the data from them into a supplemental file.
We have replaced Furthermore, when we crossed Hakai 2 , which harbors an early stop codon, to Df(2L)Exel8041 to normalize genetic background we observed strong female lethality (Fig. 3F). Although the few females we obtained did not show sexual transformation, all male flies were flightless (n=67), as observed for Mettl3 null and Mettl14 null mutants By Furthermore, when we crossed Hakai 2 /CyO females, which harbors an early stop codon, to Df(2L)Exel8041/CyO males to normalize genetic background, we observed strong female lethality compared to CyO balancer carrying control animals (149 females and 144 males, Fig. 3F). Although the two females we obtained did not show sexual transformation, all male flies were flightless (n=44), as observed for Mettl3 null and Mettl14 null mutants.
In the description of experiments in Fig. 3G, the vir2F allele is described as a dominant negative. In fact, as experiment in Fig. 3H illustrates, and the published literature shows, the vir2F allele is recessive. What the experiment in Fig. 3G documents is that females bearing only the vir2F allele (vir2f/Df) are inviable but that they are largely rescued if the flies are also heterozygous for hak1. This is a solid result that strengthens the paper, but it's very hard to interpret with respect to mechanism. What the authors want to argue is that the "dominant negative" [sic] vir2F allele prevents Sxl recruitment (itself a vague notion) and that "weakening" (also vague) the m6A complex allows Sxl better access and thus rescues female viability. It may well be that the mechanism of rescue by reducing hakai dosage is by destabilizing the the vir2F-containing MACOM complex but that is not easily squared with the published finding that reducing the dose of the MAC complex component, mettl3 (Ime4), also rescues vir2F-dependent female viability. I think the solution is for the authors is to take the powerful result as a fact, but resist the speculative interpretation.
Indeed, the reviewer is correct that per definition, vir2F is recessive and we have removed the notion that the vir2F allele is dominant negative as we don't see any signs of sexual transformation in vir2F/+ females.
We have replaced: To further confirm the involvement of Hakai in Sxl alternative splicing we made use of the dominant negative vir 2F allele, that prevents Sxl recruitment and results in female lethality (Haussmann et al. 2016). We found that weakening the m 6 A complex by removal of one copy of Hakai restored female viability of vir 2F /Df(2R)BSC778 females by correcting Sxl alternative splicing (Fig. 3G).
by To further confirm the involvement of Hakai in Sxl alternative splicing we made use of the femalelethal vir 2F allele (Haussmann et al. 2016). We found that removal of one copy of Hakai restored female viability of vir 2F /Df(2R)BSC778 females by correcting Sxl alternative splicing (Fig. 3G), as shown previously for other components of the methyltransferase complex.
The experiment in Fig. 3H are powerful and appropriately described. (They also show that vir2F is, at best, sparingly dominant to the partially defective virts allele.) I would, however, encourage two modifications to the figure. First, correct the typo on the y-axis "apperence", and second, modify the genotypes so that they more clearly reflect the fact that hakai and vir are on the same chromosome. My suggestion is to label the genotypes of the two right hand sections as + vir / + vir and hak1 vir/+ vir. Doing so provides an additional advantage in that would show how the cross was actually done.
According to the reviewer suggestions, we have corrected the label of the Y-axis in Fig 3H and modified the genotypes in Fig 3H and 3L-N.
Reviewer #3 (Remarks to the Author): I don't understand the interpretation of Fig 5A by the authors and, honestly, I think that they may underestimate the value of their data set.
We thank the reviewer for the careful review of our manuscript and for pointing out the discrepancy in the presented data. We incorrectly described the experimental set-up in the material and methods section of the manuscript but it was correctly described in the main text.
On p14 the authors state: "Cells isotopically labelled with heavy amino acids were depleted for Hakai and cells isotopically labelled with light amino acids served as a control in the forward experiment. A vice versa depletion was performed in the reverse experiment ( Supplementary  Fig. 8C).
[…] We found over 3000 ubiquitination sites, but unexpectedly not a single site was reduced in response to Hakai depletion and only one site in SesB was 1.5-fold increased (Fig. 5A, Supplementary Table 3). Therefore, this experiment strongly suggests that Hakai does not act as a general E3 ubiquitin ligase in Drosophila cells. Alternatively, it is possible that its ubiquitination activity depends on specific external stimuli." On p32, however, they state that "For ubiquitinome and proteome analysis 50 mg of proteins of heavy and light replicate were joined in a 1:1 ratio as follows: for forward experiment heavy labelled cells with control KD and light labelled cells with Hakai KD were joined, and vice versa for reverse experiment." First of all, the number of diGly peptides obtained from 50 (!) mg of starting material seems somewhat low and at first I was under the impression that the limited coverage may be the problem here. However, upon reanalyzing the source data set using Suppl Table 3, I noticed something else and I think that the x-axis in Fig 5A is mislabeled.
We fully agree with the reviewer that we cannot exclude to have not quantified the ubiquitylated sites that are regulated by Hakai due to the limited depth of the analysis. We now tone down our statement and address this point at pages 14-15: Previous text: "Therefore, this experiment strongly suggests that Hakai does not act as a general E3 ubiquitin ligase in Drosophila cells. Alternatively, it is possible that its ubiquitination activity depends on specific external stimuli".
Rephrased text: "Therefore, this experiment suggests that Hakai does not act as an E3 ubiquitin ligase in Drosophila S2R+ cells. Alternatively, it is possible that its ubiquitination activity depends on specific external stimuli or that we have not quantified the ubiquitination sites that are regulated by Hakai due to the limited depth of the analysis." This cloud is typical for ubiquitinome assays upon e.g. proteasome inhibition or knockout of a ubiquitin ligase. In conclusion, this means that, in contrast to the authors' own conclusion, in fact there seems to be (as expected?) a huge effect on the ubiquitinome upon Hakai KD.
Unless I'm missing something, I would suggest the authors to reanalyze this data set.
We thank the reviewer for pointing out the inconsistency in the description of experimental setup and we wish to apologise for this mistake. For the ubiquitylome and proteome experiments we have performed a double labelling SILAC experiment (using light and heavy isotope labelled cells) that included a label-swap as correctly described in main text (p14) but wrongly described in the materials and methods section. We have now corrected the following text on p32 (materials and methods): Previous text: "For ubiquitinome and proteome analysis 50 mg of proteins of heavy and light replicate were joined in a 1:1 ratio as follows: for forward experiment heavy labelled cells with control KD and light labelled cells with Hakai KD were joined, and vice versa for reverse experiment." Corrected text: "For ubiquitylome and proteome analysis 25 mg of heavy and 25 mg of light labelled protein lysates were joined in a 1:1 ratio as follows: for forward experiment heavy labelled lysates with Hakai KD and light labelled lysates with control KD were joined, and vice versa for reverse experiment." For further clarification of double labelling SILAC experiment, we have now included a For clarity, we generated new Supplemental Table3 (ubiquitylome analysis) and 4 (proteome analysis), where we now indicate the ratios not as heavy to light, but as Hakai KD versus Control KD. Accordingly, we provide new graphs in Figure 5A (ubiquitylome) and Figure 6A (proteome) displaying the corresponding data. Please note that in the replicate 2 (Rev) the ratios were inverted (1/ratio) so that in both replicate experiments ubiquitylation sites that are downregulated represent putative Hakai ubiquitylation sites. We have now included this explanation in the method section.
Previous text in Materials and Methods (p34): "For peptide identification in SILAC samples, raw data files were analyzed using MaxQuant (development version 1.5.2.8) to calculate the ratios between the different conditions (Cox and Mann 2008)." Rephrased text Materials and Methods (p34): "For peptide identification in SILAC samples, raw data files were analyzed using MaxQuant (development version 1.5.2.8) to calculate the ratios between the different conditions (Cox and Mann 2008). The ratios in the replicate 2 (Rev) were inverted (1/ratio). The tables with all quantified ubiquitylation sites (di-glycine sites) and protein groups are provided as Supplemental_Table_3 (ubiquitylome analysis) and Supplemental_Table 4 (proteome analysis)." The lysates that were used for ubiquitylome anaylsis were the same as the one used for the proteome analysis where we see downregulation of Vir and Flacc in the Hakai KD cells. In the proteome analysis, we see downregulation of Hakai in replicate 1 (SILAC ratio = 0.41) but was unfortunately not quantified in replicate 2. We wish to note that upon re-inspection of our data, we found a typo mistake in the graph and Heatmap of Fig 6A. The Heatmap was manually created to depict the Fold change (Hakai KD/Control KD) of all MAC, MACOM, YTH proteins and proteins below the defined threshold (as shown by red dashed line in the graph). The value for Hakai was correctly included for replicate 1 (0.41), but was wrongly typed as 0.40 for replicate 2. We wish to apologise for this mistake, which has now been corrected in both the Heatmap as well as in the graph.