Opposing regulation of the Nα-trimethylase METTL11A by its family members METTL11B and METTL13

N-terminal protein methylation (Nα-methylation) is a posttranslational modification that influences numerous biological processes by regulating protein stability, protein–DNA interactions, and protein–protein interactions. Although significant progress has been made in understanding the biological roles of Nα-methylation, we still do not completely understand how the modifying methyltransferases are regulated. A common mode of methyltransferase regulation is through complex formation with close family members, and we have previously shown that the Nα-trimethylase METTL11A (NRMT1/NTMT1) is activated through binding of its close homolog METTL11B (NRMT2/NTMT2). Other recent reports indicate that METTL11A co-fractionates with a third METTL family member METTL13, which methylates both the N-terminus and lysine 55 (K55) of eukaryotic elongation factor 1 alpha. Here, using co-immunoprecipitations, mass spectrometry, and in vitro methylation assays, we confirm a regulatory interaction between METTL11A and METTL13 and show that while METTL11B is an activator of METTL11A, METTL13 inhibits METTL11A activity. This is the first example of a methyltransferase being opposingly regulated by different family members. Similarly, we find that METTL11A promotes the K55 methylation activity of METTL13 but inhibits its Nα-methylation activity. We also find that catalytic activity is not needed for these regulatory effects, demonstrating new, noncatalytic functions for METTL11A and METTL13. Finally, we show METTL11A, METTL11B, and METTL13 can complex together, and when all three are present, the regulatory effects of METTL13 take precedence over those of METTL11B. These findings provide a better understanding of Nα-methylation regulation and suggest a model where these methyltransferases can serve in both catalytic and noncatalytic roles.

N-terminal protein methylation (Nα-methylation) is a posttranslational modification that influences numerous biological processes by regulating protein stability, protein-DNA interactions, and protein-protein interactions. Although significant progress has been made in understanding the biological roles of Nα-methylation, we still do not completely understand how the modifying methyltransferases are regulated. A common mode of methyltransferase regulation is through complex formation with close family members, and we have previously shown that the Nα-trimethylase METTL11A (NRMT1/ NTMT1) is activated through binding of its close homolog METTL11B (NRMT2/NTMT2). Other recent reports indicate that METTL11A co-fractionates with a third METTL family member METTL13, which methylates both the N-terminus and lysine 55 (K55) of eukaryotic elongation factor 1 alpha. Here, using co-immunoprecipitations, mass spectrometry, and in vitro methylation assays, we confirm a regulatory interaction between METTL11A and METTL13 and show that while METTL11B is an activator of METTL11A, METTL13 inhibits METTL11A activity. This is the first example of a methyltransferase being opposingly regulated by different family members. Similarly, we find that METTL11A promotes the K55 methylation activity of METTL13 but inhibits its Nα-methylation activity. We also find that catalytic activity is not needed for these regulatory effects, demonstrating new, noncatalytic functions for METTL11A and METTL13. Finally, we show METTL11A, METTL11B, and METTL13 can complex together, and when all three are present, the regulatory effects of METTL13 take precedence over those of METTL11B. These findings provide a better understanding of Nα-methylation regulation and suggest a model where these methyltransferases can serve in both catalytic and noncatalytic roles.
METTL11A (NRMT1/NTMT1) is a distributive trimethylase belonging to the METTL (methyltransferase-like) family of proteins that methylates the N terminus (α-amine) of its target substrates, following cleavage of the initiating methionine (1,2). METTL11A methylates both canonical and noncanonical consensus sequences. The canonical X-P-K sequence allows Ala, Pro, Ser, Gly, or Met in the first position and requires Pro and Lys in the second and third, respectively (1). The noncanonical sequence expands to include one of seven amino acids other than Pro in the second position (A/S/ G/M/E/N/Q) and either Lys or Arg in the third (3). Combined, the two consensus sequences predict over 300 METTL11A targets, and verified substrates include regulator of chromatin condensation 1 (RCC1), zinc fingers and homeoboxes 2 (ZHX2), and the ribosomal proteins RPS25 and RPL12 (1,4,5).
Our lab and others have identified many downstream effects of the loss of METTL11A, including impaired DNA damage repair, altered cell proliferation, abnormal muscle cell differentiation, and premature neural stem cell depletion (6)(7)(8)(9). METTL11A acts as a tumor suppressor in breast cancer cells, as its knockdown promotes DNA damage, cell proliferation, invasive potential, and xenograft tumor growth (6). MET-TL11A acts as an oncogene in colon and cervical cancer, as its loss slows growth and reduces invasion and migration capability, respectively (10,11). CRISPR-Cas9-mediated MET-TL11A knockout (KO) in C2C12 myoblasts prevents their differentiation into myofibers and instead promotes osteoblastic phenotypes (8). KO of Mettl11A in mice results in premature aging phenotypes, including depletion of the neural stem cell pools, neurodegeneration, and cognitive impairments (9). Despite identifying these important biological roles of METTL11A, we still do not have a thorough understanding of its upstream regulation.
One form of METTL11A regulation that we have identified is through physical interaction with its family member, METTL11B (NRMT2/NTMT2) (12). Using analytical ultracentrifugation, METTL11A was found to primarily exist as a dimer, and when combined with METTL11B, formed a heterotrimer consisting of the METTL11A dimer bound to a METTL11B monomer (12). Through this interaction, METTL11B stabilizes METTL11A and specifically promotes METTL11A methylation of noncanonical targets (12). As METTL11B primarily acts as a Nα-monomethylase in vitro, we hypothesized METTL11B could be performing the first methylation event and priming METTL11A substrates for subsequent dimethylation and trimethylation (2). However, a catalytically dead mutant of METTL11B was also capable of activating METTL11A (12), suggesting that METTL11A activation is a noncatalytic function of METTL11B. Similar regulatory interactions are apparent among other methyltransferase pairs, including the RNA methyltransferases METTL3 and METTL14, as well as the DNA methyltransferases DNMT3A, DNMT3B, and DNMT3L (13)(14)(15)(16)(17)(18)(19)(20)(21), where often one member of these pairs is serving a noncatalytic function. METTL3 is the catalytic component of a complex responsible for writing N 6 -methyladenosine (m 6 A), while the primary role of METTL14 is to provide structural support and aid in substrate recognition to promote METTL3 methylation activity (13)(14)(15)(16). METTL14 possesses a relatively occluded active site, suggesting it does not carry out catalytic activity of its own but instead has an important noncatalytic function as an allosteric regulator of METTL3 (14)(15)(16). Interestingly, METTL3 also has a noncatalytic function in translational regulation that is independent of its interaction with METTL14 (22,23). Similar to the METTL3/METTL14 complex, the DNMT3A/DNMT3B/DNMT3L complex has both active and inactive subunits. Compared to DNMT3A and DNMT3B, DNMT3L lacks important motifs in the catalytic site, rendering it catalytically inactive but still able to provide structural support to DNMT3A and DNMT3B to stimulate their methylation activities (17)(18)(19)(20)(21)(24)(25)(26)(27)(28).
METTL11A was reported to co-fractionate with another member of the METTL family, METTL13 (29). METTL13 is a dual-function methyltransferase that methylates eukaryotic elongation factor 1 alpha (eEF1A) on its Nα-amine and on internal lysine 55 (K55) (30,31). eEF1A, which is the collective reference to paralogous proteins eEF1A1 and eEF1A2, is a GTP-binding protein that is vital for the elongation step of translation (32)(33)(34). METTL13 contains two distinct 7β strand methyltransferase domains, with the N-terminal methyltransferase domain targeting K55 of eEF1A, and the C-terminal methyltransferase domain targeting its N terminus (30,31). Aberrant expression of METTL13 is linked to cancer, although like METTL11A, the mechanism of oncogenesis appears to be cell type specific. In clear cell renal cell carcinoma and bladder cancer, METTL13 has been found to inhibit cancerous phenotypes and is associated with more favorable patient outcomes (35,36). However, upregulation of METTL13 and K55me2 eEF1A is associated with poor prognosis in lung and pancreatic cancer patients (31).
Here, we confirm the interaction between METTL11A and METTL13 and demonstrate for the first time the regulatory nature of this interaction. In contrast to METTL11B, METTL13 inhibits METTL11A methylation of both canonical and noncanonical substrates. Reciprocally, we show that METTL11A can both inhibit METTL13 methylation of the eEF1A N terminus and promote methylation of K55, and catalytic activity is not needed for the regulatory roles of either METTL11A or METTL13. Finally, we demonstrate that METTL11A, METTL11B, and METTL13 can complex together, and when this occurs, regulation of METTL11A by METTL13 takes precedence over that of METTL11B. Our results not only describe the first example of a methyltransferase participating in opposing regulatory interactions with close family members but also identify novel noncatalytic, regulatory functions of both METTL11A and METTL13.
These findings on the regulatory mechanisms of Nα-methylation will help us to better understand how dysregulation can lead to the development of disease.

Identification of residues that regulate the METTL11A/ METTL11B interaction
We have previously identified that METTL11A and METTL11B interact to form a heterotrimer composed of a METTL11A dimer and a METTL11B monomer (12). Through this interaction, METTL11B provides stability to METTL11A and enhances METTL11A trimethylation activity on noncanonical substrates (12). Using computational modeling, we previously published a model of the METTL11A/METTL11B interaction and proposed 12 pairs of residues from MET-TL11A and METTL11B that are predicted to be important for mediating this interaction (12). Here, we first aimed to verify this model by mutating predicted residues and assaying the effect on METTL11A binding to METTL11B.
From the list of 12 predicted pairs, we prioritized mutations that also have biological relevance. Using the Catalogue of Somatic Mutations in Cancer, we found three METTL11B cancer-associated mutations (Q67H-lung, F68L-colon, D232N-prostate) that affected residues predicted to be important for binding from our previous model (12,37). These three mutations, and a fourth cancer-associated METTL11B mutation (V224L-breast) that is catalytically inactive but does not mediate binding to METTL11A (10,12), were all introduced individually into human METTL11B-GFP. Either WT or mutant METTL11B-GFP was transiently transfected into human embryonic kidney (HEK293T) cells along with human METTL11A-FLAG, and co-immunoprecipitations (co-IPs) were performed. Western blots were used to compare the ability of WT and mutant METTL11B-GFP to co-IP with METTL11A-FLAG (Fig. 1A). Only METTL11B-GFP possessing the cancer-associated D232N mutation showed a significant decrease in its ability to co-IP with METTL11A-FLAG as compared to WT METTL11B-GFP (Fig. 1, A and B). No other mutations had a significant effect on the METTL11A/ METTL11B interaction (Fig. 1B). These results suggest that D232 of METTL11B is an important residue for the interaction between METTL11A and METTL11B.
Having positively identified the importance of METTL11B D232, we next aimed to update our computational model with D232 included as a required element of the interaction surface. Docking-based modeling was used to predict the new proteinprotein interaction interface that results from this added constraint. Computational models were generated with the ZDOCK server, and graphics were viewed and analyzed using Chimera (38,39). This method was first used to generate a model of the METTL11A (PDB 5E1B) dimer that closely correlated with our previously published METTL11A dimer model (12). The METTL11A dimer model was then docked to a METTL11B (PDB 6DUB) monomer using the ZDOCK server, with METTL11B D232 identified as a contacting residue (Fig. 1C). The resulting new model of the METTL11A/ METTL11B interaction shared many similarities with our previously published model (12). The updated model suggested METTL11B D232 interacts with METTL11A R78 (Fig. 1D). To verify if METTL11A R78 is also important for the METTL11A/METTL11B interaction, we next repeated the METTL11A-FLAG/METTL11B-GFP co-IPs with METTL11A R78G alone or in combination with METTL11B D232N. Similar to METTL11B D232N alone, METTL11A R78G alone also significantly reduced the interaction between METTL11A and METTL11B (Fig. 1, E and F). When both mutants were expressed together, there was an even greater decrease in the interaction (Fig. 1, E and F), confirming the importance of this residue pair, showing that both members play important regulatory roles and further refining our interaction model.
METTL11A co-fractionated with another METTL family member, METTL13 (29). To verify METTL11A and METTL13 interact, we again performed co-IP experiments. HEK293T cells were transfected with METTL11A-FLAG and METTL13-GFP, and Western blots determined that METTL13-GFP did co-IP with METTL11A-FLAG ( Fig. 2A). This was interesting to us, as we have previously shown by immunofluorescence that METTL11A is predominantly nuclear and inactive toward RCC1 in the cytoplasm (1, 3), and METTL13 is thought to be primarily cytoplasmic (45).
To better understand the cell compartment localization of METTL11A and METTL13, we performed nuclear/cytoplasmic fractionations and found that while METTL13 is predominantly cytoplasmic, METTL11A is found in both the nucleus and cytoplasm (Fig. 2B). These data indicate MET-TL11A and METTL13 are interacting in the cytoplasm. To confirm these findings, HEK293T cells were transfected with METTL13-FLAG, cell lysates were fractionated into cytoplasmic and nuclear fractions, METTL13-FLAG was immunoprecipitated out of the cytoplasmic fraction, and interactors were identified using liquid chromatography-mass spectrometry. Six proteins were identified as having a 100% abundance ratio in the METTL13-FLAG transfected cells as compared to the untransfected control cells (Fig. 2C). Of these six proteins, METTL11A had the highest percent coverage and number of representative peptides (Fig. 2C). Finally, to confirm direct binding of METTL11A and METTL13, both were cloned into a bacterial dual expression vector. Untagged, recombinant METTL11A co-purified with His-tagged METTL13, and untagged, recombinant METTL13 co-purified with His-tagged METTL11A (Fig. S1A).

METTL11A and METTL13 exhibit reciprocal regulation
As the interaction between METTL11A and METTL11B results in the activation of METTL11A methylation activity, we next determined if METTL13 could similarly regulate METTL11A activity. In vitro methyltransferase assays were performed using combinations of recombinant human MET-TL11A and METTL13 enzymes and an RCC1 N-terminal peptide as substrate. We found that METTL13 inhibited METTL11A trimethylation activity of RCC1 (Fig. 3, A and B). Specifically, METTL13 was able to both significantly increase the Km of METTL11A and significantly lower its Vmax (Fig. 3, C and D). These data indicate METTL13 is a mixed inhibitor of METTL11A, and it regulates METTL11A in an opposing manner to METTL11B.
We were next interested in examining if METTL11A can alter METTL13 activity. METTL13 is a dual-function methyltransferase that can methylate both the N-terminus and internal lysine 55 (K55) of eEF1A. We first used in vitro methyltransferase assays with an N-terminal eEF1A peptide substrate to measure the effect of METTL11A on the ability of METTL13 to Nα-methylate eEF1A. We found that MET-TL11A exhibited a partial, though significant, inhibition of METTL13 Nα-methylation of eEF1A at high concentrations (Fig. 4, A and B). Unlike the effect of METTL13 on MET-TL11A, we found a significant decrease in Vmax but no significant difference in Km (Fig. 4, C and D), indicating METTL11A is acting as a noncompetitive inhibitor of METTL13 Nα-methylation of eEF1A.
We then tested if METTL11A affects the ability of METTL13 to methylate K55. Unfortunately, it has been shown that peptides containing K55 are not viable substrates for

Regulation of METTL11A by METTL11B and METTL13
in vitro methyltransferase assays (31). It has been suggested that the methyltransferase domain responsible for methylating K55 may require the substrate to be in its fully folded state rather than an unstructured short peptide (31). Thus, we used full-length, recombinant eEF1A protein substrate and K55 methyl-specific antibodies to assay METTL13 activity by Western blot (Fig. 4E). We found a significant increase in eEF1A K55 dimethylation by METTL13 at both 5 and 10 min when METTL11A is present compared to METTL13 alone (Fig. 4F). Together, these results suggest that while METTL13 decreases the methylation activity of METTL11A, METTL11A causes METTL13 to favor eEF1A K55 methylation over Nαmethylation. Further functional studies will be needed to further delineate the implications of this shift, as the distinct roles of each methylation remain unknown.

Noncatalytic activities of METTL11A and METTL13
The regulation of METTL11A is a noncatalytic function of METTL11B, and another family member, METTL16, has also recently been shown to have a noncatalytic role facilitating assembly of the translation-initiation complex, in addition to its role as an m 6 A methyltransferase (46). To determine if the regulatory roles of METTL11A and METTL13 are dependent on their catalytic activities, we repeated the previous in vitro methyltransferase assays with catalytically inactive mutant METTL13 and METTL11A. The METTL13 mutations G58R and E524A were previously found to abrogate methylation activity of the METTL13 N-terminal methyltransferase domain and C-terminal methyltransferase domain, respectively (30,31). We constructed a G58R/E524A double mutant and tested its ability to inhibit METTL11A methylation of RCC1. Although lacking significant catalytic activity of its own (Fig. S1, B and C), the G58R/E524A METTL13 double mutant was able to inhibit METTL11A methylation of RCC1 at a level similar to that of WT METTL13 (Fig. 5A). Mutation of Asp180 to Lys or Asn (D180K/D180N) in METTL11A has also been shown to inhibit its catalytic activity (1,47). Similar to the METTL13 double mutant, D180K METTL11A was able to inhibit METTL13 Nα-methylation of eEF1A at levels comparable to WT METTL11A (Fig. 5B), despite lacking catalytic activity of its own (Fig. S1D). D180K METTL11A was also able to significantly increase eEF1A K55 dimethylation by METTL13 (Fig. 5, C and D). Together these results demonstrate novel, noncatalytic regulatory functions for both MET-TL11A and METTL13.

METTL11A/METTL11B/METTL13 interactions
Once we determined that the METTL11A/METTL13 interaction exhibits regulatory effects on both members of the pair and the regulatory effects were likely structural (not

Regulation of METTL11A by METTL11B and METTL13
catalytic), we were next interested in beginning to narrow down the interaction interface. As METTL13 is relatively large (79 kDa) and contains two distinct methyltransferase domains, each responsible for targeting one of the methylation sites on eEF1A (30), we wanted to determine which region of METTL13 was responsible for mediating the interaction with METTL11A. We created GFP-tagged N-terminal and C-terminal fragments of METTL13 (Fig. 6A). The N-terminal fragment (residues M1-Y344) contained the methyltransferase domain that targets K55 of eEF1A, and the C-terminal fragment (residues E345-V699) contained the methyltransferase domain that targets the N-terminus of eEF1A (30). Each was expressed in HEK293T cells with METTL11A-FLAG, and co-IPs were performed. Only the N-terminal METTL13 fragment was found to co-IP with METTL11A-FLAG (Fig. 6A), suggesting that residues M1-Y344 interact with METTL11A. This suggests that binding at one region of METTL13 can affect activity at both domains.
Since we had now determined that METTL11A is involved in opposing regulatory interactions with two different METTL Figure 4. METTL11A inhibits the ability of METTL13 to methylate eEF1A at the N-terminus but promotes methylation at K55. A, methylation of an eEF1A N-terminal peptide is significantly lower as METTL11A concentration increases compared to METTL13 by itself. B, activity curves of eEF1A Nαmethylation by METTL13 in the presence or absence of METTL11A. C, Km is not significantly different when METTL11A is present. D, Vmax is significantly lower when METTL11A is present. E, levels of dimethylated lysine 55 (me2K55 eEF1A) are higher at 5 and 10 min when METTL11A is present with METTL13 compared to METTL13 alone. F, quantification of blots showing the ratio of me2K55 eEF1A to total eEF1A. These data indicate METTL11A is a noncompetitive inhibitor of eEF1A Nα-methylation and an activator of K55 methylation. n = 3, * denotes p < 0.05, ** denotes p < 0.01, *** denotes p < 0.001 as determined by unpaired t test. eEF1A, eukaryotic elongation factor 1 alpha.
family members, METTL11B and METTL13, we were next interested in determining if METTL11A, METTL11B, and METTL13 can exist in a complex together or if the interactions are mutually exclusive. We first used co-IPs to determine that METTL13 and METTL11B can interact (Fig. 6B), indicating they could be in a complex together. To determine if this interaction was dependent on METTL11A acting as a bridge between METTL11B and METTL13, we performed similar experiments in control HCT116 cells and HCT116 cells with CRISPR-Cas9-mediated KO of METTL11A (10). We found that the interaction between METTL13 and METTL11B is enhanced when METTL11A is present in control cells compared to in the METTL11A KO cells (Fig. 6, B and C), indicating they can all be in a complex together.
Finally, we were interested in determining if the regulatory effects of METTL13 or METTL11B took precedence when all three enzymes were present. As METTL11B has previously been shown to activate the activity of METTL11A specifically for noncanonical substrates (4, 12), we used in vitro methyltransferase assays to measure if METTL13 also affected the ability of METTL11A to methylate the noncanonical substrate ZXH2. We found that similar to the canonical RCC1 substrate (Fig. 3A), METTL13 inhibited METTL11A trimethylation of a ZHX2 N-terminal peptide (Fig. 6D). We replicated previous findings that METTL11B significantly increased METTL11A methylation activity on the ZHX2 N-terminal peptide (4), but interestingly found that inhibition by METTL13 takes precedence over activation by METTL11B. We found a significant decrease in ZHX2 methylation when both METTL11B and METTL13 are present compared to METTL11A alone (Fig. 6E). As ZHX2 is a nuclear METTL11A substrate, we also wanted to determine if METTL13 could inhibit METTL11A methylation of cytoplasmic targets in the presence or absence of METTL11B. Using the N-terminal peptide of the cytoplasmic, canonical substrate RPS25, we found METTL13 could inhibit METTL11A methylation of RPS25 in both the presence and absence of METTL11B (Fig. 6F). These results suggest that METTL13 inhibits methylation of both nuclear and cytoplasmic METTL11A substrates and that these inhibitory effects are not altered by the presence of METTL11B. Together, these findings suggest a model where METTL13 is able to inhibit METTL11A in the cytoplasm, even if MET-TL11A translocates to the cytoplasm with METTL11B.

Discussion
We had previously shown that METTL11B binds MET-TL11A and activates its methylation activity toward noncanonical substrates (12). Here, we further refine our predicted model of the METTL11A/METTL11B complex and identify mutations that disrupt its interaction. We also verify that METTL11A participates in an additional regulatory

Regulation of METTL11A by METTL11B and METTL13
interaction with METTL13 in the cytoplasm. For both the canonical and noncanonical substrates tested, METTL13 inhibits METTL11A activity, while METTL11A inhibits METTL13 methylation of the eEF1A N-terminus and promotes K55 methylation. These regulatory functions are distinct from their catalytic functions, as catalytically inactive mutants elicit levels of inhibition comparable to WT. Finally, we found that all three enzymes can exist in a complex together, with METTL11A likely acting as a bridge. When METTL11A, METTL11B, and METTL13 are together, METTL13 outcompetes METTL11B to inhibit METTL11A methylation activity. These are the first studies to show a methyltransferase participating in opposing regulatory interactions with two different family members and identify novel, noncatalytic regulatory functions for both METTL11A and METTL13.
Together, these data suggest a model where cellular localization plays an important role in determining the interaction partners and main function of METTL11A. In the nucleus, METTL11A catalytic activity predominates and can be activated by METTL11B. In the cytoplasm, METTL11A is inactive against nuclear and cytoplasmic targets (3), due to inhibition by METTL13. Here, its noncatalytic role of METTL13 regulation predominates. It remains unknown if METTL11B can translocate out of the nucleus with METTL11A, but our data show that if this were the case, it would not affect the inhibitory effects of METTL13 on METTL11A. While we see very little METTL13 in nuclear fractions (Fig. 2B), there are reports of nuclear METTL13 (45). Our data suggest that if it enters the nucleus, it will be able to inhibit METTL11A activity, so it will be interesting to determine under what conditions METTL13 translocates to the nucleus and if it is in concentrations high enough to inhibit METTL11A activity.
Methyltransferases taking on different roles in different cellular compartments has become an increasingly apparent trend. METTL3, the active subunit of the primary complex responsible for the m 6 A modification, typically localizes to the nucleus, where it interacts with METTL14 and additional Values were normalized to 1. n = 4. D, methylation of the ZHX2 N-terminal peptide by METTL11A is significantly lower when METTL13 is present compared to when METTL11A is alone. METTL13 does not methylate the ZHX2 peptide. E, methylation of the ZHX2 N-terminal peptide is significantly higher when METTL11B is present compared to when METTL11A is alone. When both METTL13 and METTL11B are present with METTL11A, methylation activity is significantly reduced compared to when METTL11A is alone. F, there is no significant difference between methylation activities of METTL11A against the RPS25 Nterminal peptide when METTL11B is present compared to METTL11A alone. Methylation activity is significantly lower when METTL13 is present with METTL11A, with or without METTL11B also present. These data indicate METTL11A, METTL11B, and METTL13 can exist in a complex together, and the regulatory effects of METTL13 on the methylation activity of METTL11A take precedence over the effects of METTL11B. Unless otherwise specified, n = 3, * denotes p < 0.05, ** denotes p < 0.01, *** denotes p < 0.001, **** denotes p < 0.0001 as determined by unpaired t test. eEF1A, eukaryotic elongation factor 1 alpha; KO, knockout. proteins to form the m 6 A writer complex (48,49). However, METTL3 has also been found in the cytoplasm where it interacts with different binding partners (translation initiation machinery) to enhance mRNA translation (22,23). Interestingly, METTL14, the principal activator of METTL3 activity, is only found in the nucleus (22). Another member of the METTL family, METTL16 has also recently been found to take on two distinct roles depending on cellular localization (46). Similar to METTL3, METTL16 in the nucleus places the m 6 A modification on a set of distinct RNA targets (50) and in the cytoplasm, promotes translation through direct interactions with translation initiation machinery (46). Relative to METTL3, which is predominantly nuclear, METTL16 has a much larger cytoplasmic population, suggesting its role in the cytoplasm may be more impactful (46). It will be interesting to determine if the cytoplasmic role of METTL11A is also related to promoting translation through its ability to increase methylation of K55 eEF1A by METTL13.
Similar to METTL11A, the distinct nuclear versus cytoplasmic functions found in other METTL proteins can also be divided into catalytic versus noncatalytic functions, respectively. Both METTL3 and METTL16 were reported to maintain their cytoplasmic translation regulatory functions independent of their own catalytic activity (22,46). Again, our findings with METTL11A and METTL13 are consistent with this pattern, in that their regulatory effects were still elicited by catalytically inactive mutants. This pattern highlights a distinct division among the METTL family of proteins, as some methyltransferases have both catalytic and noncatalytic functions like METTL3, METTL16, METTL11A, and METTL13, while others, like METTL14 and perhaps METTL11B (51), only seem to have noncatalytic functions. We suspect additional functions will be identified for methyltransferases that were previously thought to only have one defined role.
Since the regulatory interactions are independent of catalytic activity, it is likely that the described regulatory effects are the result of structural changes elicited through binding. Future studies will use nuclear magnetic resonance (NMR) to delineate structural changes occurring in METTL11A following the binding of either METTL11B or METTL13. By comparing the structural changes that occur in METTL11A when it interacts with its binding partners to the changes that occur with substrate binding, we predict that the binding of METTL11B may promote a METTL11A conformation that is more favorable for non-canonical substrate binding, subsequently activating methylation activity. An interaction with METTL13 on the other hand, may promote a conformational change in METTL11A that makes substrate binding unfavorable, thereby inhibiting its methylation activity.
The identification of these novel, noncatalytic functions of both METTL11A and METTL13 is important because despite the many similarities between the various members of the METTL family, each displays their own unique and complex characteristics surrounding their regulatory mechanism, localization pattern, and functions outside of their methylation activities. By having a more thorough understanding of each METTL protein, we can also learn about how their dysregulation, mislocalization, or misfunction can lead to the development of diseases. Disruptions to noncatalytic, regulatory functions are known to manifest as abnormal catalytic functions, as a mutation in METTL14 (R298P) leads to decreased methylation activity of the METTL3/METTL14 complex and also disrupts the ability of the complex to distinguish between the WT and mutant RNA substrate (14). R298 of METTL14 is located close to the METTL3/METTL14 junction and catalytic site, and this mutation is frequently found in human endometrial cancer, suggesting the biological relevance of a mutation in the inactive activator of the METTL3/METTL14 complex (14-16, 37, 52).
Future studies will attempt to identify important residues mediating the METTL11A/METTL13 interaction, as we have done with METTL11A/METTL11B. We suspect some of these interacting residues will also be found in human diseases, given the large number of mutations that have been found present in various cancer types for METTL11A, METTL11B, and METTL13 (37), which could possibly be eliciting problematic effects through the disruption of their noncatalytic functions. We also suspect that as the interaction sites between MET-TL11A, METTL11B, and METTL13 become more defined, we will pinpoint additional biologically relevant mutations disrupting regulatory interactions and ultimately affecting Nαmethylation patterns. As rational design of peptidomimetic molecules that disrupt protein-protein interactions is a growing trend in drug development (53), targeting these methyltransferase interactions may also be a promising therapeutic option moving forward.
For direct binding experiments, METTL11A and METTL13 were also cloned into pETDuet-1. First, METTL11A was put in the multiple cloning site (MCS) upstream of the His-tag using HindIII and NotI sites, and METTL13 was put in the untagged MCS with NdeI and BglII. METTL13 naturally contains both a HindIII site and a BglII site that were each disrupted in the template vector through the introduction of silent mutations at Lys27 (AAA to AAG) and Ile222 (ATC to ATA), respectively, using Quikchange mutagenesis. Similarly, the BglII site in METTL11A was disrupted through a silent mutation at Ile214 (ATC to ATA) as described previously (4). For the reciprocal construct, METTL13 was subcloned into the MCS upstream of the His-tag using HindIII and NotI, and METTL11A was subcloned into the untagged MCS with NdeI and BglII.
The primers used to create the METTL13-pET15b, pETDuet-1, and eEF1A-pET30a constructs and the forward primers used for Quikchange are as follows: 5 0 NdeIMETTL13: All recombinant His-tagged proteins were purified as described previously (55).

co-IP experiments
Twenty-four hours prior to transfection, 1 x 10 6 HEK293T cells or 4 × 10 6 HCT116 control or METTL11A KO cells were plated in 10 cm tissue culture dishes. HEK293T cells were calcium phosphate transfected with 1 μg each of appropriate constructs, and HCT116 cells were transfected with 8 μg each of appropriate constructs using Lipofectamine 2000 (Thermo Fisher Scientific). Approximately 24 h post-transfection, cells were scraped directly into 200 μl of lysis buffer (50 mM Tris, 300 mM NaCl, 5 mM MgCl 2 , 1% NP-40, 7 mM BME), plus protease inhibitors. Twenty microliter of cell lysate was saved for input controls. The remainder of the lysate was added to 5 μl of Pierce Protein G agarose beads (Thermo Fisher Scientific), and the mixture was rotated 1 to 2 h at 4 C to preclear. Following the preclear incubation, the mixtures were spun quickly, and the super was added to 40 μl of EZ View Red anti-FLAG M2 agarose beads (Millipore Sigma). The mixture was rotated 1 to 2 h at 4 C and washed 3× with PBS + 0.1% NP-40 + 500 mM NaCl. The immunoprecipitated proteins were eluted from the beads in 5× Laemmli buffer and boiled at 95 C for 10 min. The bead-free IP supernatant and the input samples were run on 10% SDS-PAGE gels and analyzed with Western blots as described above.

In vitro methyltransferase assays
In vitro methyltransferase assays for measuring Nα-methylation were conducted using the MTase-Glo Methyltransferase Assay (Promega) following the manufacturer's protocol. Each assay used 0.2 μM recombinant enzyme unless otherwise specified for the titration of METTL11A, 20 μM S-adenosyl methionine, and either 1 μM of RCC1 (SPKRIAKRRSPPADA), 10 μM of eEF1A (GKEKTHINIVVIGH), 60 μM of ZHX2 (ASKRKSTTPCMVRTS). or 0.75 μM of RPS25 (PPKDDKKKKDAGKS) N-terminal peptides (Anaspec, Fremont, CA) as substrates. Briefly, reactions between the various combinations of enzymes and substrates were incubated in wells on a 96-well plate at room temperature for either 5 or 20 min and stopped with the addition of 0.5% trifluoroacetic acid. The MTase-Glo detection reagents were added according to the manufacturer's protocol, and the luminescence was measured using the Cytation5 Imaging System (BioTek). Background signals were measured through the inclusion of no substrate control reactions and subtracted where applicable.
Nuclear/cytoplasmic fractionations and mass spectrometry HEK293T cells were transfected as described previously to express METTL13-FLAG. Twenty-four hours posttransfection, lysates from METTL13-FLAG expressing cells and untransfected control cells were fractionated to yield both a nuclear and a cytoplasmic fraction. This process was repeated for four 10 cm plates for each condition, either METTL13-FLAG or untransfected. Briefly, 24 h posttransfection, cells were collected directly into cytoplasmic lysis buffer (10 mM Hepes, 1.5 mM MgCl 2 , 10 mM KCl, 0.5 mM DTT, 0.05% NP-40, pH 7.9) plus protease inhibitors. To disrupt the cellular membrane, lysates were vortexed on high 7 s, incubated on ice 10 min, vortexed again, incubated on ice 1 min, and then spun down at high speed (13,200 rpm) for 5 min at 4 C. Supernatant was collected as the cytoplasmic fraction. Remaining nuclei were washed 3× with cold PBS and resuspended in a nuclear lysis buffer (5 mM Hepes, 1.5 mM MgCl 2 , 0.2 M EDTA, 0.5 mM DTT, 26% glycerol (v/v), 300 mM NaCl, pH 7.9). Nuclei were ruptured by vortexing on high for 15 s and incubating on ice for 40 min, with a 15 s vortex every 10 min. Supernatant was collected following a 10min spin at 4 C (13,200 rpm) as the nuclear fraction. Fractionation cleanliness was tested by analyzing input samples (20 μl of lysate) by Western blot with RCC1 (nuclear protein) as a readout of cytoplasmic fraction cleanliness, and tubulin (cytoplasmic protein) as a readout of nuclear fraction cleanliness (Fig. 2B).
The remainder of the cytoplasmic fraction was added to 5 μl of washed Pierce Protein G agarose beads (Thermo Fisher Scientific), and the mixture was rotated 1 to 2 h at 4 C to preclear. Following the preclear, the mixture was spun quickly, and the super was added to 15 μl of washed Pierce anti-DYKDDDDK (FLAG) magnetic agarose beads (Thermo Fisher Scientific). The mixture was rotated 1 to 2 h at 4 C and washed 3× with PBS + 0.1% NP-40 + 150 mM NaCl. Interacting proteins were eluted into buffer (25 mM Hepes-NaOH, pH 7.5, 100 mM NaCl, 0.1 mg/ml FLAG peptide (Anaspec), and 10 μl/ml PMSF, filter sterilized). Elution was sent for label-free quantification mass spectrometry analysis at the Cornell Institute of Biotechnology. Top interactors were those 100% abundant in METTL13-FLAG sample compared to untransfected control.

Molecular modeling
Molecular modeling was produced using a combination of the ZDOCK server and Chimera UCSF (38,39). The METTL11A dimer model was produced using the METTL11A crystal structure of the monomer (PDB: 2EX4, chain A) as both input structures in the protein-protein docking tool, ZDOCK (38). Out of the top ten predicted models, the one most closely resembling the previously published METTL11A dimer model (12) was selected for use in further modeling. To produce the METTL11A/ METTL11B heterotrimer model with the METTL11B D232 constraint, the METTL11A dimer and the METTL11B crystal structure (PDB: 5DUB) were used as inputs in ZDOCK (38), with METTL11B D232 selected as a contacting residue. Molecular graphics and analyses were performed with UCSF Chimera developed by the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco, with support from NIH P41-GM103311 (39).

Statistical analyses
All statistical analyses were performed using GraphPad Prism 9 software (GraphPad). The specific statistical tests used are noted in the respective figure captions, and results are presented as mean ± standard deviation (SD).

Data availability
All data are contained within the manuscript.
Supporting information-This article contains supporting information.