YTHDF1 promotes mRNA degradation via YTHDF1‐AGO2 interaction and phase separation

Abstract Objectives YTHDF1 is known as a m6A reader protein, and many researches of YTHDF1 focused on the regulation of mRNA translation efficiency. However, YTHDF1 is also related to RNA degradation, but how YTHDF1 regulates mRNA degradation is indefinite. Liquid‐liquid phase separation (LLPS) underlies the formation of membraneless compartments in mammal cells, and there are few reports focused on the correlation of RNA degradation with LLPS. In this research, we focused on the mechanism of YTHDF1 degraded mRNA through LLPS. Materials and Methods The CRISPR/Cas9 knock out system was used to establish the YTHDF1 knock out (YTHDF1‐KO) cell lines (HEK293 and HeLa) and METTL14 knock out (METTL14‐KO) cell line (HEK293). 4SU‐TT‐seq was used to check the half‐life changes of mRNAs. Actinomycin D and qPCR were used to test the half‐life changes of individual mRNA. RNA was stained with SYTO RNA‐select dye in wild type (WT) and YTHDF1‐KO HeLa cell lines. Co‐localization of YTHDF1 and AGO2 was identified by immunofluorescence. The interaction domain of YTHDF1 and AGO2 was identified by western blot. Phase separation of YTHDF1 was performed in vitro and in vivo. Fluorescence recovery after photobleaching (FRAP) was performed on droplets as an assessment of their liquidity. Results In this research, we found that deletion of YTHDF1 led to massive RNA patches deposited in cytoplasm. The results of 4SU‐TT‐seq showed that deletion of YTHDF1 would prolong the half‐life of mRNAs. Immunofluorescence data showed that YTHDF1 and AGO2 could co‐localize in P‐body, and Co‐IP results showed that YTHDF1 could interact with AGO2 through YT521‐B homology (YTH) domain. We confirmed that YTHDF1 could undergo phase separation in vitro and in vivo, and compared with AGO2, YTHDF1 was more important in P‐body formation. The FRAP results showed that liquid AGO2 droplets would convert to gel/solid when YTHDF1 was deleted. As AGO2 plays important roles in miRISCs, we also found that miRNA‐mediate mRNA degradation is related to YTHDF1. Conclusions YTHDF1 recruits AGO2 through the YTH domain. YTHDF1 degrades targeting mRNAs by promoting P‐body formation through LLPS. The deletion of YTHDF1 causes the P‐body to change from liquid droplets to gel/solid droplets, and form AGO2/RNA patches, resulting in a degradation delay of mRNAs. These findings reveal a previously unrecognized crosstalk between YTHDF1 and AGO2, raising a new sight of mRNA post‐transcriptional regulation by YTHDF1.


| INTRODUC TI ON
N 6 -methyladenosine (m 6 A) is the most prevalent internal messenger RNA (mRNA) modification and regulates stability of targeting mRNAs. 1 The m 6 A methylation in mRNAs is mainly installed by a writer complex composed of METTL3 and METTL14 2,3 in a sequence context of RRACH (R = A or G; H = A, C, or U). 3 The m 6 A reader protein utilizes different mechanisms to recognize and bind to m 6 A-containing RNAs. For example, the YTH-domain-containing proteins (YTHDF1/2/3 and YTHDC1/2) use YTH domain to recognize m 6 A methylation. 1 Cytoplasmic YTHDF2 promotes degradation of its targeting transcripts partially through recruiting the CCR4-NOT deadenylase complex. 4,5 By contrast, the other two cytoplasmic m 6 A readers, YTHDF1 and YTHDF3, are suggested to promote translation of targeting transcripts in HeLa cells by recruiting translation initiation factors. 6,7 Recent researches showed that the combined activity of YTHDF1/2/3 proteins leads to degradation of m 6 A-modified mRNA. 8 When solutions of macromolecules such as proteins or nucleic acids undergo LLPS, they condense into a dense phase that often resembles liquid droplets, and this dense phase coexists with a dilute phase. The driving force underlying LLPS is the exchange of macromolecule/water interactions for macromolecule/macromolecule and water/water interactions under conditions for which this process is energetically favorable. 9-12 Importantly, whether a solution undergoes phase separation depends strongly on the concentration and identities the macromolecules and the solution and also the environmental conditions including temperature, salt type, and concentration, co-solutes, pH, and the volume excluded by other macromolecules. 12,13 The physicochemical properties of LLPS suggest a variety of possible functions of phase separation processes in cells, such as transcriptional or translational programs, [13][14][15] RNA metabolism, 16 and signal transduction. 17 LLPS can locally concentrate molecules in condensates to activate reactions, signaling processes, and nucleation of cytoskeletal structures. 16,18 Increasing the local concentration of a key enzyme or protein complex in a condensate could accelerate biochemical reactions. 18 The miRISCs contain the Argonaute (AGO) family protein AGO2, which plays important roles in the function of miRNA. [19][20][21] Previous studies have reported the localization of miRISC components to the P-body, 19 during this process, GW182 can interact with AGO2 to form P-body. [22][23][24][25] P-body is a membrane-free organelle and is implicated in post-transcriptional regulation of mRNAs. 26,27 Mounting evidences have shown that liquid-liquid phase separation (LLPS) underlies the formation of membraneless structures. 18 However, the mechanism of phase separation in RNA degradation has yet to be understood.
Here, we found that YTHDF1 could promote mRNA degradation. Mechanistically, YTHDF1 could recruit AGO2 through the YTH domain, and YTHDF1 played important roles in P-body formation.
Deletion of YTHDF1 led to P-body from liquid droplets convert to gel/solid, massive RNA patches deposited in the cytoplasm and delay mRNAs degradation.

| Data collection
Raw data of CLIP-Seq datasets for AGO2 and YTHDF1 were was acquired from GEO database (GSM1339401 and GSM1339402).

| Distribution of m 6 A and RBP peaks
MetaPlotR 28 was used to identify the distribution of m 6  was deleted. As AGO2 plays important roles in miRISCs, we also found that miRNAmediate mRNA degradation is related to YTHDF1.
Conclusions: YTHDF1 recruits AGO2 through the YTH domain. YTHDF1 degrades targeting mRNAs by promoting P-body formation through LLPS. The deletion of YTHDF1 causes the P-body to change from liquid droplets to gel/solid droplets, and form AGO2/RNA patches, resulting in a degradation delay of mRNAs. These findings reveal a previously unrecognized crosstalk between YTHDF1 and AGO2, raising a new sight of mRNA post-transcriptional regulation by YTHDF1.

| Analysis of m 6 A reader binding
We obtained the CLIP-seq data of m 6 A readers (IGF2BP1, IGF2BP2,   IGF2BP3, YTHDF1, YTHDF2, and YTHDF3) from POSTAR2 database. Peaks were annotated with ChIPseeker (v1.5.1). 29 Genes with m 6 A peaks in the 3' UTR region and genes with AGO2 binding were overlapped and selected as the candidate gene for further analysis.

| Gene enrichment analysis
We selected genes that have at least 1.5 times prolonged half-life times in YTHDF1 deleted cells and performed gene enrichment through the Metascape website (http://metas cape.org/gp/index. html#/main/step1). Top 20 significant categories were selected and displayed using the ggplot2 package. 30

| Immunofluorescence Staining
Cells were grown on acid-treated glass coverslips. Treated cells were fixed with ice-cold 4% paraformaldehyde (MKCB4217, sigma) for 20 min, washed 1-3 times with phosphate-buffered saline (PBS), and permeabilized with 0.5% Triton X-100 in PBS for 20 min. After washed 1 to 3 times with 0.05% Tween-20 in PBS, the samples were blocked in PBS containing 1-2% BSA for 1 h at room temperature. Cells were incubated with primary antibody for 16-24 h at 4°C. After washed 1-3 times with 0.05% Tween-20 in PBS, and secondary antibody for 1-2 h at 37°C. The slides were washed three times with 0.05% Tween-20 in PBS and sealed the slides with DAPI. Finally, observed under a fluorescence microscope. Images were analyzed using the Image J software.
The antibodies used in this research are listed in Table S1.

| Western blotting
Cells were lysed using RIPA buffer (Tris 20 mM, NaCl 150 mM, KCl 20 mM, Mgcl 2 1.5 mM, glycerol 10%, Triton X-100 1%, PH 7.5), and the protein concentration was measured by BCA Protein Assay Kit (C503021; Sangon Biotech). Equal amounts of proteins were separated by 10%-12% sodium dodecyl sulfate-polyacrylamide gel electrophoresis and were transferred onto polyvinylidene fluoride membranes, and the target proteins were finally detected using standard western blotting protocols and visualized using the Super Signal West Pico Plus Luminol/Enhancer Solution (UC279012, Thermo). Detailed information of primary antibodies and secondary antibodies are listed in Table S1.

| RNA staining
The labeling solution consists of a 500 nM solution of RNA Select was added to 900 μl of medium. The labeling solution (500 nM) was prewarmed at 37°C prior to application and used immediately. A sufficient amount of the prewarmed 500 nM labeling solution was applied to cover the cells adhered to a coverslip and incubated for 20 min at 37°C. When labeling was completed, the labeling solution was removed; cells were rinsed twice in cell-culture medium or PBS and rested for 5 min in medium at 37°C ready for imaging.

| Knockout cell line establishment
We used the lentiviral CRISPR/Cas9 system to establish knockout cell lines. After ligation of T4 DNA Ligase (EL0014, Thermo Scientific™), the ligated plasmid was transformed into E. coli. LentiCRISPRv2, and packaging plasmids pMD2.G and psPAX2 were co-transfected into

| Construction of stable overexpressing cell lines
We used pCDH-CMV-AGO2-coGFP and pLVX-EF1α-YTHDF1-mCherry to package virus for targeting cell infection. Monoclonal cells for culture were picked for culture by limiting dilution. Validated by fluorescence to ensure successful construction.

| Stranded-specific mRNA sequencing
Total RNA was extracted with Trizol (Invitrogen) following the manufactory's instruction. The extracted RNA was digested with DNase I (NEB) to remove the genomic DNA before RNA cleanup. About

| Protein purification
The pET28a-His-EGFP-AGO2 or pET28a-His-mCherry-YTHDF1 plasmids were transformed into bacterial strain E. coli BL21 for protein production. A single bacteria colony from each transformation was selected, inoculated in liquid Luria-Bertani (LB) medium, and grew at 37°C overnight while being shaken at 220-250 rpm.   showed uniformly distributed RNA with small particles aggregation ( Figure 1E), whereas massive RNA patches were found in cytoplasm of YTHDF1 deficient HeLa cells ( Figure 1F). Consistently, 3D imaging results showed uneven distribution and significantly increased density of the RNA patches in YTHDF1 deficient cells. Previous study was reported that YTHDF1 is suggested to promote translation of targeting transcripts in HeLa cells. 7 To explore whether those RNA patches preserved translational activity, we co-stained eIF3B and RNA in WT and YTHDF1-KO cell lines. The results showed that RNA patches were co-localized with eIF3B, it is suggested that the RNA patches have translation initiation capabilities ( Figure 1G).

| YTHDF1 promotes mRNA degradation
To evaluate the RNA dynamics, we performed 4SU-TT-seq in WT and YTHDF1-KO HEK293 cell lines (Figure 2A). After the 1-hour exposure to the nucleoside analog 4-thiouridine (4SU), newly transcribed RNAs were labeled by 4SU, which were isolated and sequenced. The sequencing results revealed that many mRNAs in YTHDF1 deficient cells were more stable than WT cells ( Figure 2B). Previous study shows that m 6 A Peaks are enriched at the mi-croRNA target sites, 31 so we hypothesize that YTHDF1 promotes mRNA degradation may be related to miRNA. The miRISCs contain the Argonaute (AGO) family protein AGO2, which plays important roles in the function of miRNA. 20,21 In line with the aforementioned findings, we observed that m 6 A readers (YTHDF1/2/3 and IGF2BP1/2/3) and AGO2 had the largest number of common targeting mRNAs ( Figure 2F,G) and YTHDF1 showed most significant related. Public CLIP-seq data showed that YTHDF1 binding sites shared distribution pattern with m 6 A modification sites and AGO2 binding sites in HEK293 and HeLa cell lines ( Figure 2H,I). These results suggest that YTHDF1 may regulate mRNA degradation with AGO2.

| YTHDF1 interacts and co-localize with AGO2 in P-body
To confirm the interaction between YTHDF1 and AGO2, we first conducted co-immunoprecipitation (Co-IP) experiments and found that YTHDF1 could interact with AGO2 in RNA independent manner ( Figure 3A,B). In addition, immunofluorescence analysis revealed that YTHDF1 could co-localize with AGO2 ( Figure 3C,D). To characterize the interaction domain of YTHDF1 and AGO2, we overexpressed fragmented and full length of YTHDF1 with FLAG-tag ( Figure 3E) as well as HA tagged full-length AGO2 for Co-IP. Western blot analysis showed that aa1-240 and aa1-388 showed weak signal with AGO2, aa241-331, and aa241-388 had no signal with AGO2, whereas aa389-523, aa389-559, and full length YTHDF1 exhibited strong signal with AGO2 ( Figure 3F-H). These results indicated that YTHDF1 could interact with AGO2 through the 389-523 domain (YTH domain). To observe co-localization of YTHDF1 and AGO2 in vivo, we performed 3D image reconstruction of whole cells and found physical co-localization of YTHDF1 and AGO2 ( Figure 3I).
Since AGO2 is an important component of P-body. 19,20 We hypothesize that YTHDF1 could co-localize with AGO2 in P-body. In order to obtain the precise localization of YTHDF1, we performed immunofluorescence with another two P-body marker proteins (GW182 and DDX6) and found co-localization of YTHDF1 and DDX6/GW182 ( Figure 4A,B). Collectively, our data strongly support an interaction between YTHDF1 and AGO2, and YTHDF1 could localize in P-body with AGO2.
To further confirm that YTHDF1 promotes mRNA degradation, we measured the mRNA decay rate after transcriptional inhibition with actinomycin D treatment in WT, YTHDF1 ( Figure 1A) or METTL14 knock out cell lines ( Figure 4C). We established METTL14 knockout (METTL14-KO) cell line with CRISPR-Cas9 in HEK293 cells. Western blot analysis showed that we significantly knock down the METTL14 expression in HEK293 cells ( Figure 4E).
Our results showed that the half-life of mRNAs was significantly extended after YTHDF1 or METTL14 deletion ( Figure 4F). Then, we used YTHDF1-KO cell line to overexpress YTHDF1 and knock down (siRNA) AGO2. We found that YTHDF1 overexpression promoted RNA degradation, while si-AGO2 treatment inhibited this recuperation effect ( Figure 4G). These results indicated that YTHDF1 likely regulates mRNA degradation through AGO2. We next sought to determine whether m 6 A modification is involved in this process. Knockout of METTL14 significantly decreased m 6 A levels ( Figure 4D) and extended the half-life of the selected mRNAs ( Figure 4F). Rescue of METTL14 expression in METTL14 knockout cells caused the half-life of the selected mRNAs similar to WT level.
Notably, AGO2 knockdown could inhibit the recuperation effect of METTL14 re-expression ( Figure 4H). Together, these data indicate that the YTHDF1 promotes mRNA degradation through interacting with AGO2, and m 6 A modification involved in this process.

| Both YTHDF1 and AGO2 can activate phase separation in vitro
YTHDF1 had Prion-like domain (PrLD) and multiple disorder sequences ( Figure 5A), which were found to be drivers of phase separation. 18,32 This observation is consistent with previous reports showing that YTHDF1 can undergo phase separation. [33][34][35] In contrast, AGO2 had only a few weak disorder sequences and no PrLD domain ( Figure 5B), although AGO2 was reported to aggregate miRISCs formation through LLPS. 16 To gain more insight into the role of YTHDF1 and AGO2 in phase separation, we performed both in vivo and in vitro assays. E. coli was used to over-express AGO2 EGFP and YTHDF1 mCherry . We found that both AGO2 EGFP and YTHDF1 mCherry could undergo phase separation alone in vitro ( Figure 5C,D), and the ability of AGO2 to drive phase separation appears to be weaker than YTHDF1. Fluorescence recovery after photobleaching (FRAP) is often performed on the droplets as an assessment of their dynamics.
FRAP results showed that the P-body fluorescence foci labeled with AGO2 coGFP and YTHDF1 mCherry were able to recover after bleaching ( Figure 5E,F). Additionally, we also found that these liquid droplets could undergo diffusion and fusion in vivo ( Figure 5G,H). These data suggest that YTHDF1 and AGO2 co-localized P-body is formed by phase separation.

| YTHDF1 is more important in driving YTHDF1/AGO2 complex LLPS in vitro
To determine whether YTHDF1 or AGO2 plays a more critical role in the formation of P-body, we performed phase separation in vitro. We found that in the YTHDF1 mCherry and AGO2 EGFP mixed solution, increasing the concentration of AGO2 EGFP had little effect on driven LLPS ( Figure 6A). In contrast, raising the concentration of YTHDF1 mCherry substantially increased solution LLPS, in the YTHDF1 mCherry and AGO2 EGFP mixed solution ( Figure 6B). In addition, a higher concentration of RNA m6A (A synthetic RNA with four m 6 A modification sites) did not significantly effect on driven LLPS when solution only with AGO2 EGFP (Figure 6C) but largely increased LLPS in the presence of YTHDF1 mCherry ( Figure 6D). When RNA m6A , AGO2, and YTHDF1 mCherry were mixed, increased concentration of RNA m6A could markedly increase LLPS of solution ( Figure 6E). Taken together, our data suggest that YTHDF1, rather than AGO2, that is likely a more important factor in driven LLPS-process, and m 6 A modification may participate in this process.
Interestingly, the AGO2 coGFP foci in WT cells were uniformly fluorescent spots, whereas in the YTHDF1 or METTL14 knock out cell lines, many large and hollow AGO2 coGFP foci were observed. The results indicated abnormal formation of AGO2 coGFP foci ( Figure 7A,B).
We performed FRAP in HEK293 YTHDF1-KO (overexpress AGO2 coGFP ) and found that the large and hollow fluorescent foci failed to restore ( Figure 7C). These results indicated that deletion of YTHDF1 leaded to the conversion of AGO2 liquid droplets to gels/solids. Furthermore, we explored the state of endogenous AGO2 after YTHDF1 deletion by immunofluorescence experiments. The results showed that compared with the WT cells, the size of AGO2 foci significantly increased after YTHDF1 deletion, which is consistent with the findings in HEK293 YTHDF1-KO (overexpress AGO2 coGFP ) and HEK293 METTL14-KO (overexpress AGO2 coGFP ) ( Figure 7D). Additionally, we used the Pbody marker GW182 to confirm the defect in P-body formation by YTHDF1 deletion. The results showed that the GW182 foci size showed increased which is similar to the results of AGO2 ( Figure 7E).
Taken together, our data demonstrate that deletion of YTHDF1 increases the aggregation of AGO2 and causes abnormal P-body formation.
As we found YTHDF1 deletion led to the deposition of RNA patches in cytoplasm ( Figure 1F) and impaired P-body formation ( Figure 7A-C), we co-stained those RNA patches with AGO2/GW182 in YTHDF1-KO cells. The results showed that the RNA patches and AGO2/GW182 were co-localized ( Figure 7F,G), indicated that the RNA patches were been able to localizing into P-body, although the P-body is gels/solids but not liquid character.
Previous studies reported that AGO2 is an important component of miRISCs and plays important roles in the function of miRNA. [19][20][21] We found that YTHDF1 could interact with AGO2 and co-localize in P-body. So, whether miRNA participates in the process of mRNA degradation driven by YTHFD1. We choose let-7 as the representative miRNA to confirm our hypothesis. 36,37 We successfully over-

| DISCUSS ION
The m 6 A modifications are enriched near stop codons and 3'UTRs, 38 where AGO2 recognition sites are heavily distributed in 3'UTRs. 22 This phenomenon implies a potential correlation between the m 6 A modification and miRNA-AGO2 mediated gene regulation. In our research, we demonstrate that YTHDF1 through YTH domain interacts with AGO2 (AGO2 perhaps along with miRNA), then undergoes LLPS to aggregate more components of miRISCs, leading to P-body formation for mRNA degradation.
Notably, our data show that YTHDF1, rather than AGO2, is the leading factor to drive LLPS. Deficiency of YTHDF1 disrupts the interaction between YTHDF1 and AGO2, leads to the conversion of AGO2 liquid droplets to gels/solids and substantially increases the mRNA stability ( Figure 8F). This mechanism raises a new sight for the mRNA regulation of YTHDF1. The YTH-domain-containing proteins are involved in regulation of mRNA stability. 1,39 For example, YTHDF2 can promote the degradation of mRNAs and is localized in P-body, 5 with the ability to drive phase separation. 33,34 In our study, we found that YTHDF1 could interact with AGO2 through the YTH domain, but YTHDF2 and YTHDF3 also contained YTH domain, and YTHDF1/2/3 were interacts with AGO2 (AGO2 perhaps along with miRNA) through YTH domain, then undergoes LLPS to aggregate more components of miRISCs, leading to P-body formation for mRNA degradation. Deficiency of YTHDF1 disrupts the interaction between YTHDF1 and AGO2, leads to the conversion of AGO2 liquid droplets to gels/solids and substantially increases the mRNA stability reported with ability of driving LLPS. 33 This interesting finding prompted us to explore whether all YTH-domain-containing proteins participate in this process.
In addition, previous studies showed that YTHDF1 is related to mRNA translation. 7 To clarify whether these RNA patches are functional (translationally active) or nonfunctional (dormant or degradation intermediate) in YTHDF1 deletion cells. We found that these RNA patches could co-localize with eIF3B ( Figure 1G), indicated that they may preserve functional translation initiation capabilities. We are not sure whether the translation efficiency is changed or not of RNAs in these patches compared with normal situation. Therefore, we speculate that there may be a balance between translation and degradation, yet the specific molecular mechanism needs further investigation.
In this study, we found YTHDF1 could interact with AGO2 through the YTH domain. Thus, it is worth investigating whether YTHDF1 affects AGO2 binding to its targeting mRNA. To verifying this speculation, we conducted AGO2-CLIP assay. Unfortunately, the reproducibility of the experimental results is poor so reliable conclusion was not achieved. In our results, we found that the RNA patches and AGO2/GW182 were co-localized ( Figure 7F,G), indicating that these RNAs could bind to AGO2/GW182. Whether binding efficiency was affected or not is undetermined. YTHDF1 recognizes m 6 A modification, interacts with AGO2, and then undergoes phase separation to drive P-body formation. In this process, we don't have solid data to prove whether YTHDF1 affects AGO2 binding to targeting mRNA, Further investigation is necessary to validate our observation and speculation.

| CON CLUS IONS
Overall, YTHDF1 promotes mRNA degradation via YTHDF1-AGO2 interaction and phase separation. Our study provides new insights into the mRNA stability regulation of YTHDF1. These findings shed light on important roles of m 6 A in many physiological and pathological processes, such as tumor formation, 39 embryo development, 40 meiosis, 41,42 stem cell differentiation, 43,44 and DNA damage repair. 45

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest.