In vitro methylation of the U7 snRNP subunits Lsm11 and SmE by the PRMT5/MEP50/pICln methylosome

Recombinant Lsm10/11 heterodimer interacts with endogenous PRMT5 methylosome from mammalian extracts

Our recent biochemical and structural studies with recombinant U7 snRNP uncovered key interactions within this multisubunit endonuclease and demonstrated how it cleaves histone pre-mRNAs (Sun et al. 2020; Yang et al. 2020). We next focused on the biogenesis of the U7 snRNP, including the poorly understood process of assembling the U7-specific Sm ring on U7 snRNA. To facilitate this study, we generated rabbit antibodies against the baculovirus-expressed heterodimer of Lsm10-MBP fusion protein and His-tagged Lsm11. The resultant serum from immunized rabbits (α10/11) precipitated from the cytoplasm of HeLa cells very small amounts of endogenous Lsm10 and Lsm11 that could not be visualized by silver staining (Fig. 1A, lane 1), but were readily detectable by western blotting (data not shown). Interestingly, when the same HeLa cytoplasm was supplemented with Lsm10-MBP/Lsm11 heterodimer, the α10/11 serum in addition to the two recombinant proteins precipitated three endogenous proteins strongly stained with silver (Fig. 1A, lane 2). They were identified by mass spectrometry and western blotting as subunits of the PRMT5 methylosome: PRMT5 methyltransferase of ∼70 kDa (Stopa et al. 2015), MEP50 of ∼50 kDa (Friesen et al. 2002), and ∼35 kDa pICln (Pu et al. 1999; Meister et al. 2001). PRMT5 and MEP50 form a heterooctamer consisting of four copies of each protein (Antonysamy et al. 2012; Ho et al. 2013; Timm et al. 2018), and pICln binds stoichiometrically to PRMT5 (Friesen et al. 2001b; Pesiridis et al. 2009; Guderian et al. 2011; Mulvaney et al. 2021), explaining the abundance of the three methylosome subunits relative to the small amounts of Lsm10 and Lsm11 used in the assay.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 1.

Lsm11 interacts with the PRMT5 methylosome. (A) Immunoprecipitation of HeLa (lanes 1–3) or mouse proteins (lanes 4–7) bound to recombinant heterodimers of Lsm10 with either wild-type Lsm11 (labeled as 10/11), Lsm11 lacking the internal loop (10/11ΔL), or Lsm11 lacking both the amino-terminal fragment and the loop (10/11ΔNL). The heterodimers were incubated with indicated extracts and precipitated with the α10/11 rabbit serum. Immunocomplexes were collected on protein A beads, resolved by SDS-PAGE and protein bands stained with silver. Proteins precipitated by the same serum from HeLa or mouse extracts lacking recombinant Lsm10/11 heterodimer are shown in lanes 1,6. HC and LC denote heavy and light chains of immunoglobulins, respectively. (B) Sequence of human Lsm11 (amino acid 1–360). The boundaries of the deleted regions are indicated with arrowheads. Arginine residues neighboring glycines potentially targeted for methylation by PRMT5 are indicated with arrows. Structural elements within the Sm fold are indicated with a horizontal bar (α-helix) or thick arrows (β-stands). The portion of the internal loop between β-stands 3 and 4 that was deleted in Lsm11ΔL and Lsm11ΔNL is underlined. (C–E) Purification of PRMT5 methylosome on amylose beads via maltose-binding protein (MBP) tag attached to Lsm10 or to the amino-terminal fragments of Lsm11. HeLa or mouse cytoplasmic extracts, as indicated, were incubated with variants of Lsm10/11 heterodimers or amino-terminal fragments of Lsm11 fused to MBP, and proteins collected on amylose beads were analyzed by silver staining (panels C and E) or western blotting, using specific antibodies (panel D). In panel E, asterisks indicate full-length amino-terminal Lsm11 fragments fused to MBP.

Human Lsm11 with its 360-amino acids is the largest member of the Sm/Lsm family and contains an extended amino-terminal region of 154 residues that plays a critical role in the function of U7 snRNP in 3′ end processing (Fig. 1B). Within this region, amino acids ∼20–50 bind FLASH (Yang et al. 2009) to form a platform for recruiting the CPSF73 endonuclease and two other components of the cleavage and polyadenylation machinery: CPSF100 and symplekin (Yang et al. 2013; Aik et al. 2017). Between amino acids 68 and 79, Lsm11 contains a GR cluster that may serve as a primary methylation target for PRMT5. Amino acids 155–360 of Lsm11 fold into a canonical Sm domain consisting of an amino-terminal α helix (amino acids 155–165) and five β strands (Sun et al. 2020). The β3 and β4 strands are separated by a large loop (amino acids 207–337) that is unique to Lsm11 and contains four separate GR or RG dipeptide motifs (Fig. 1B).

To identify regions of Lsm11 that bind the PRMT5 methylosome, we expressed Lsm10 fused to MBP in complex with Lsm11 lacking the internal loop (amino acids 211–322, Lsm11ΔL), or Lsm11 lacking both the loop and the entire amino-terminal region encompassing amino acids 1–154 (Lsm11ΔNL) (Supplemental Fig. S1). We also bacterially expressed four amino-terminal fragments of Lsm11 fused to MBP (Supplemental Fig. S1). These fragments lack the Sm fold region of Lsm11 that interacts with Lsm10 and were expressed alone.

Deletion of the internal loop from Lsm11 had no effect on the interaction of the Lsm10/11 heterodimer with endogenous PRMT5 methylosome. Compared to the wild-type dimer containing full-length Lsm11, approximately the same amounts of PRMT5, MEP50, and pICln were coprecipitated together with Lsm10/11ΔL dimer by α10/11 serum (Fig. 1A, lane 3). Thus, the extensive loop despite containing several GR/RG dipeptide motifs does not bind the methylosome. Readily detectable binding of the methylosome to the Lsm10/11ΔL dimer was also observed in a cytoplasmic extract from mouse myeloma cells (Fig. 1A, lane 4) and was not abolished by additionally deleting the entire amino-terminal region of Lsm11 (Fig. 1A, lane 5). Finally, binding of PRMT5, MEP50, and pICln to Lsm10/11 heterodimers was observed in a mouse myeloma nuclear extract (Fig. 1A, lane 7), indicating that the interaction can occur in both the cytoplasmic and nuclear fractions from different mammalian cell lines.

To further analyze the interaction between the Lsm10/11 dimer and the PRMT5 methylosome, we tested a different approach by directly collecting proteins bound to the dimer on amylose beads via the MBP tag attached to Lsm10 rather than using immunoprecipitation with α10/11 serum. Following a brief incubation of the wild-type or Lsm11ΔNL heterodimer with a HeLa cytoplasmic extract, the bound proteins were collected on amylose beads and visualized by silver staining following their separation by SDS-PAGE. Again, both dimers, one containing full-length Lsm11 and the other lacking the amino terminus and the internal loop copurified with large amounts of the PRMT5 methylosome (Fig. 1C, lanes 1,2, respectively). In the absence of Lsm10/11 dimers, only a background of several protein bands was detected but no trace of the three methylosome components (Fig. 1C, lane 3).

We used the same approach with various amino-terminal fragments of Lsm11 tagged at the amino terminus with MBP: MBP-Δ40 (Lsm11 amino acids 41–169), MBP-105N (Lsm11 amino acids 1–105), and MBP-65N (Lsm11 amino acids 1–65) (Supplemental Fig. S1). As determined by silver staining (not shown) and western blotting, MBP-Δ40 and MBP-105N clearly interacted with PRMT5 and MEP50 (Fig. 1D, lanes 4,5), but compared to the Lsm10/11 dimer, coprecipitated significantly less of pICln (Fig. 1D, lane 3), consistent with this adaptor protein contacting Lsm10 (see below). No trace of the methylosome was detected on amylose beads in the absence of recombinant proteins or in the presence of the first 65 amino acids of Lsm11 fused to MBP (MBP-65N) (Fig. 1D, lanes 2,6, respectively).

We repeated this experiment using a cytoplasmic extract from mouse myeloma cells rather than from Hela cells and an additional Lsm11 protein encompassing all 168 amino-terminal amino acids of the protein. Again, as visualized by silver staining, the heterodimer (in this case Lsm10/11ΔL) was very efficient in binding pICln (Fig. 1E, lane 1). MBP-168N and MBP-105N pulled down only a trace of this subunit while binding efficiently the two remaining proteins of the methylosome, PRMT5 and MEP50 (Fig. 1E, lanes 2,3). Altogether, these results indicate that Lsm10/11 contains at least two binding sites for the methylosome. The first site is located within the region delineated by amino acids 65 and 105. This region contains an RG cluster (amino acids 58–79) that likely interacts with PRMT5 (Friesen et al. 2001b). The second site likely corresponds to the Sm fold of Lsm10 that tightly interacts with pICln, resembling the interaction of this subunit with SmD1 (Chari et al. 2008; Grimm et al. 2013; Paknia et al. 2016).

Binding of recombinant PRMT5 methylosome to Lsm10/11 heterodimer and methylation of Lsm11

To determine whether the interaction between PRMT5 methylosome and Lsm10/11 can be reconstituted from recombinant components, we coexpressed PRMT5 and MEP50 in the baculovirus system, and purified pICln as a separate protein from bacteria. The PRMT5/MEP50 complex was mixed with the Lsm10/11 heterodimer in molar ratio 1:1 either in the absence or in the presence of pICln, and the amount of the methylosome components immobilized on amylose beads via the MBP tag attached to Lsm10 was assessed by SDS-PAGE and silver staining. As seen in Figure 2A, both PRMT5 and MEP50 were readily detected and their amount increased in the presence of pICln, consistent with its role as the adaptor subunit for recruiting the Lsm10/11 substrate. We conclude that the interaction between Lsm10/11 heterodimer and the methylosome can be recapitulated using recombinant components.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 2.

Binding and methylation activities of recombinant PRMT5 methylosome. (A) PRMT5/MEP50 complex without (lane 1) or with pICln (lane 2) bound to Lsm10/11 heterodimer was purified on amylose beads via MBP attached to Lsm10 and analyzed by silver staining. (B) Lsm10/11ΔNL heterodimer was tested for the ability to stably bind PRMT5/MEP50 complex either in the absence (lane 1) or in the presence of pICln (lane 2). Proteins purified on amylose beads via MBP attached to Lsm10 were visualized by silver staining. Lanes 3–4 represent 5% input of proteins used in the binding assay. (C) Various Sm or Lsm heterodimers were incubated overnight in solution with recombinant complex of PRMT5 and MEP50 in the presence of ³H SAM. Proteins were resolved by SDS-PAGE and visualized by staining with Coomassie Blue (lanes 1–7). Dried gel was used for fluorography to detect proteins labeled with radioactive methyl group (lanes 8–14). (D) Various amino-terminal fragments of Lsm11 fused at the amino terminus with MBP were tested for their ability to undergo methylation, as described in panel C. Coomassie Blue stained gel and fluorogram are shown at the top and bottom, respectively. Asterisks in the top panel indicate full-length amino-terminal Lsm11 fragments fused to MBP. Lower bands correspond to proteolytic fragments.

Our studies with cytoplasmic extracts suggested that binding of endogenous methylosome to the Lsm10/11 heterodimer with the deletion of both the internal loop and the amino-terminal region (Lsm10/11ΔNL), hence lacking the RG cluster between residues 58 and 79, likely depends on pICln, which is predicted to bind the Sm fold of Lsm10 (Paknia et al. 2016). With the availability of recombinant PRMT5/MEP50 heterodimer and pICln as two separate entities, we directly tested this interpretation by either adding or omitting pICln in the binding assay with the Lsm10/11ΔNL heterodimer. Only background amounts of PRMT5 and MEP50 bound to Lsm10/11ΔNL heterodimer when pICln was excluded and the interaction was greatly stabilized in its presence (Fig. 2B, lanes 1,2, respectively). Thus, as with endogenous methylosome, recombinant methylosome recognizes two independent binding sites in the Lsm10/11 heterodimer, one likely being the RG cluster (amino acids 58–79) that interacts with PRMT5, and the other being the Sm fold of Lsm10 that interacts with pICln. These two contacts collectively contribute to the strong interaction between the methylosome and the Lsm10/11 heterodimer.

The strong association of the PRMT5 methylosome with the Lsm10/11 heterodimer and with the amino-terminal region of Lsm11 containing a GR cluster raised the possibility that either Lsm11 alone or both subunits of the heterodimer are methylated. This would be consistent with the catalytic activity of PRMT5 on the RG-rich carboxy-terminal tails of SmD1, SmB, and SmD3. To test this possibility directly, we carried out an in vitro methylation assay using all recombinant components and ³H-labeled SAM as a source of radioactively labeled methyl group. Initially, the methylation reaction was carried out only with PRMT5 and MEP50, in the absence of pICln. We tested methylation of the three variants of Lsm10/11 heterodimers, in which Lsm11 was either full length or lacked the amino-terminal region and/or the internal loop, as described above (Supplemental Fig. S1). As a control, we used spliceosomal heterodimers SmD1/D2 and SmB/D3 that are known substrates for in vitro symmetric arginine dimethylation by the PRMT5 methylosome (Friesen et al. 2001b, 2002; Meister et al. 2001). As an additional control, we used a variant of SmB/D3 heterodimer (SmB1-95/D3) in which the SmB subunit encompassed amino acids 1–95, lacking the entire unstructured GR-rich carboxy-terminal tail.

Following overnight incubation, the proteins used in the assay were resolved by SDS-PAGE, visualized by staining with Coomassie Blue (Fig. 2C, lanes 1–7) and their methylation status determined by fluorography (Fig. 2C, lanes 8–14). Incubation of PRMT5 and MEP50 with Lsm10/11 and Lsm10/11ΔL heterodimers resulted in methylation of both full-length Lsm11 and Lsm11ΔL, but not Lsm10 (Fig. 2C, lanes 8,10). As expected, radioactive labeling was almost completely blocked by excess of the competitive methylation inhibitor SAH (Fig. 2C, lane 9) (Meister et al. 2001). Thus, Lsm11, but not Lsm10, is a substrate for in vitro methylation by PRMT5, with methylation occurring somewhere outside the internal loop. Methylation was entirely abolished in Lsm10/11ΔNL heterodimer (Fig. 2C, lane 11), demonstrating that the methylation site is located within the deleted amino-terminal portion of Lsm11. The spliceosomal Sm proteins, D1, B, and D3 were methylated as expected, with a deletion of the carboxy-terminal tail abolishing methylation of SmB (Fig. 2C, lanes 12–14).

To map the methylation site more precisely, we tested amino-terminal fragments of Lsm11 fused at the amino terminus to MBP (Supplemental Fig. S1). Again, the reaction was carried out in the presence of ³H SAM and limiting amounts of recombinant PRMT5 and MEP50. Of the four amino-terminal fragments of Lsm11 fused to MBP, 168N, 168NΔ40, and 105N, became radioactively labeled (Fig. 2D, lanes 1–3), whereas the shortest 65N fragment was not methylated (Fig. 2D, lane 4). In contrast to the three other fragments, MBP-65N does not bind the methylosome and lacks the GR cluster, strengthening the notion that the methylosome binding site and the methylation site are located between amino acids 68 and 79 of Lsm11.

Structure of the methylosome–Lsm10/11 complex

To determine how PRMT5 interacts with Lsm11, we mixed all three methylosome components with the Lsm10/11 heterodimer and purified their complex by gel filtration chromatography for cryo-EM studies, which yielded a structure of the complex at 2.86 Å resolution (Fig. 3A–D; Supplemental Table S1). Clear EM density was observed for PRMT5 and MEP50, and for only a few residues of the PRMT5 binding motif (PBM) in pICln (Fig. 3A; Mulvaney et al. 2021). The remaining parts of pICln, most of Lsm11 and the entire Lsm10 were disordered, providing no structural information about the interaction between the Lsm10/Lsm11 heterodimer and pICln. PRMT5 and MEP50 form a heterooctamer of four tightly bound heterodimers (Fig. 3A), consistent with previous structural studies of the methylosome by X-ray crystallography and cryo-EM (Antonysamy et al. 2012; Ho et al. 2013; Timm et al. 2018).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 3.

Structure of human methylosome–Lsm10/11 complex. (A) Schematic drawing of the human methylosome–Lsm10/11 complex. The molecules in the top layer are shown in color, and those in the bottom layer are in gray. Peptide segments from pICln and Lsm11, as well as the adenosine portion of the SAM cofactor, are shown in the sphere model. (B) Cryo-EM density for the Lsm11 peptide in the active site of PRMT5. The fitted atomic model is shown in the stick model. (C) Detailed interactions between the Lsm11 peptide and the active site of PRMT5. The structure of histone H4 peptide bound to the methylosome is shown in the overlay (in gray). (D) Electrostatic surface of the active site region of PRMT5. The Lsm11 peptide is shown in the stick model. The structure figures were produced with PyMOL (www.pymol.org). (E,F) WT Lsm11 encompassing the first 168 amino acids amino-terminally fused to MBP (MBP-168N) and indicated Lsm11 mutants were incubated with a mouse cytoplasmic extract. Proteins bound to Lsm11 were immobilized on amylose beads via MBP. The beads were split evenly into two halves and analyzed by western blotting (panel E) for the presence of methylosome components or resuspended in a buffer containing ³H SAM to test methylation of Lsm11, as detected using fluorography (panel F).

We observed EM density consistent with a peptide segment in the active site of PRMT5. After examining many possibilities, residues 71-GRGRGR-76 in the amino-terminal extension of Lsm11 were chosen as the most likely interpretation of this density (Fig. 3B). The density exists in the active sites of all four PRMT5 molecules, although the quality of the density varied among them. The side chain of Arg74 is inserted deep into the active site and would be located next to the cofactor SAM (Fig. 3C,D), arguing that it is poised for methylation. Since the density for the side chain of Arg72 is relatively weak, we cannot exclude the possibility that the active site of PRMT5 accommodates the overlapping 69-GGGRGR-74 Lsm11 peptide. In this alternative model, methylation would occur at Arg72, a residue that is well conserved among vertebrate Lsm11 orthologs. Plausibly, the observed EM density is derived from a mixture of the two different peptides bound in the active site of different PRMT5 molecules.

The six-residue peptide of Lsm11 bound to the active site of PRMT5 assumes the conformation of a type II reverse turn, with the Arg residue ready for methylation located at the third position of the turn between two glycines in the primary sequence (Fig. 3C). Consistently, there is no indication of EM density for a Cβ atom on either side of the methylated Arg (Fig. 1B). In fact, a Cβ atom on the residue before the methylated Arg would clash with the side chain of Phe580 of PRMT5 (Fig. 3C). For the residue that follows the methylated Arg, a Cβ atom would be located ∼3 Å from the side chains of Gln309 and Val503, and a larger side chain here would clash with PRMT5. Therefore, the current structure suggests that a GRG motif may be preferred for the methylation site, and a GRA motif could also be accommodated.

The binding mode of the Lsm11 peptide resembles that of the SGRGK histone H4 peptide (Antonysamy et al. 2012), in which the methylated arginine is also flanked by two glycines. The overall structures of the two complexes are similar, with rms distance of 0.64 Å for 609 equivalent Cα atoms of PRMT5 and 0.36 Å for 295 equivalent Cα atoms of MEP50. For the cofactor SAM (or SAH), EM density for only adenosine is observed in the four active sites of PRMT5, and the rest of the cofactor (the Met residue) is disordered (Fig. 3C). Possibly related to this disordering, the side chain of Glu435 does not interact with Arg74 of Lsm11 and instead assumes a different conformation that would clash with the main chain of the Met residue of SAM (Fig. 3C).

To determine which of the arginines in the RG cluster of Lsm11 is methylated, we generated three mutants of the amino-terminal Lsm11 region (amino acids 1–168) fused amino-terminally to MBP (MBP-168N): R72A, R74A, and double mutant R72,74A. In these mutants, either one or two arginines predicted by cryo-EM to undergo methylation were replaced with alanines. Both the wild-type MBP-168N and three mutant proteins were incubated with a cytoplasmic extract from mouse myeloma cells and proteins immobilized on amylose beads via the MBP were visualized by silver staining (not shown) and western blotting (Fig. 3E). As expected, no trace of PRMT5, MEP50, and pICln was recovered from the beads in the absence of MBP-168N, and the wild-type protein readily bound all three methylosome subunits (Fig. 3E, lanes 1,2). The R72A mutant was almost as efficient in binding the methylosome, whereas the R74A mutant bound weakly, as illustrated by significantly reduced amounts of MEP50 and pICln (Fig. 3E, lanes 3,4). Importantly, the R72,74A double mutant failed to interact with the methylosome (Fig. 3E, lane 5).

We also tested the ability of each sample to methylate Lsm11. The beads were resuspended in a buffer containing ³H SAM and incubated overnight. As expected, the results of this in vitro methylation reaction mirrored the results of the binding assay, with the strongest radioactive signal observed for the wild-type MBP-168N. The R72A mutant was methylated with a lower efficiency, whereas the radioactive signal for R74A and R72,74A mutants was very weak and undetectable, respectively (Fig. 3F). Based on the cryo-EM reconstruction, binding assay, and in vitro methylation, we conclude that arginine in position 74 of Lsm11 is the primary binding and methylation target for the methylosome. PRMT5 also binds and methylates arginine 72, but with lower efficiency than that observed for arginine 74.

Methylation of SmE

In vivo, SmD1/D2 heterodimer is believed to rapidly associate with pICln, which recruits the heterodimer to the methylosome for methylation of SmD1. The heterodimer subsequently associates with a trimer of SmE, SmF, and SmG, forming a stable 6S intermediate complex consisting of Sm proteins D1/D2/E/F/G and pICln (Chari et al. 2008; Grimm et al. 2013; Paknia et al. 2016). We tested whether methylation of Lsm11 can be affected by pICln. We also tested methylation of Lsm11 in the presence of both pICln and the SmE/F/G heterotrimer that together may form a complex equivalent to the spliceosomal-type 6S complex (Neuenkirchen et al. 2015; Paknia et al. 2016).

When the methylation reaction was carried out in the presence of either pICln, which binds Lsm10/11 (Paknia et al. 2016) or SmE/F/G heterotrimer, which based on the behavior of SmD1/D2 is unlikely to form a stable complex with a heterodimer of Lsm10/11 (Chari et al. 2008; Zhang et al. 2011; Grimm et al. 2013), methylation of Lsm11 was virtually unaffected (Fig. 4A, bottom panel, lanes 1–3). While the lack of any stimulatory effect by the heterotrimer was expected, the failure of pICln to increase the efficiency of Lsm11 methylation by improving the affinity of PRMT5 to Lsm10/11 was surprising. This was likely due to the high concentration of recombinant PRMT5 used in the assay. Strikingly, when both pICln and the SmE/F/G heterotrimer were included into the reaction, methylation of an additional protein migrating between 10 and 15 kDa size markers was detected (Fig. 4A, bottom panel, lane 4). Alignment of the Coomassie Blue stained gel with the fluorogram tentatively identified the methylated protein as SmE, which migrates at the top of the SmE/F/G triplet (Fig. 4A, top panel, lane 4).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 4.

Methylation of SmE by recombinant methylosome. (A–C) Recombinant PRMT5/MEP50 complex was incubated with indicated Sm or Lsm heterodimers and ³H SAM either in the absence or presence of pICln and SmE/F/G heterotrimer, as indicated. Protein methylation was detected, as described in the legend for Figure 2. Coomassie Blue stained gels and fluorograms are shown at the top and bottom, respectively.

The most obvious difference between the SmD1/D2 and Lsm10/11 heterodimers is the presence of the long amino-terminal region in Lsm11, which does not exist in SmD2, the spliceosomal counterpart of Lsm11. To determine whether this Lsm11 region is important for the methylation of SmE, we carried out an in vitro methylation assay using Lsm10/11ΔNL heterodimer, in which Lsm11 lacks both the amino-terminal extension and the internal loop within its Sm fold (Supplemental Fig. S1). As expected, due to the deletion of the RG-rich region between amino acids 68 and 79, no methylation of Lsm11 was observed (Fig. 4B, lanes 1–3). Yet, methylation of SmE in the presence of SmE/F/G and pICln was even more pronounced, being detected as a strong signal after a relatively short exposure. Thus, the amino-terminal region of Lsm11 may play an inhibitory role in methylation of SmE, perhaps acting as a competitor for binding PRMT5.

We tested whether methylation of SmE can occur if the Lsm10/11 heterodimer is replaced with the spliceosomal SmD1/D2 heterodimer. Only SmD1 (fused to SUMO) was methylated, as expected, and no radioactive signal was detected for SmE or any other protein of the complex (Fig. 4C, lane 4). In a parallel reaction carried out with Lsm10/11, pICln, and SmE/F/G, SmE was readily methylated (Fig. 4C, lane 2). This result is consistent with previous studies demonstrating that only SmD1 becomes methylated by the PRMT5 methylosome in the 6S complex consisting of SmD1/D2 heterodimer, SmE/F/G heterotrimer, and pICln (Neuenkirchen et al. 2015). Thus, methylation of SmE appears to be an event specifically linked to Lsm10/11 heterodimer and may play a role in the assembly of U7 snRNP in vivo.

SmE is methylated by endogenous methylosome

We next tested whether methylation of SmE can be carried out by endogenous methylosome. Heterodimers of Lsm10 with either full length or ΔNL Lsm11 were incubated with a mouse cytoplasmic extract either in the absence or in the presence of the SmE/F/G heterotrimer and bound complexes collected on amylose beads via the MBP tag attached to Lsm10, as described above. Following exhaustive washes, the beads were incubated overnight in a buffer containing radioactive SAM, and the bound material was separated by SDS-PAGE, stained with Coomassie Blue (Fig. 5A, top) and subsequently analyzed by fluorography (Fig. 5A, bottom). Each recombinant Lsm10/11 heterodimer bound a number of proteins, some of which were identified as the components of endogenous methylosome (Fig. 5A, top panel, lanes 1–4, Supplemental Fig. S2), consistent with the data shown in Figure 1C. Importantly, endogenous methylosome in the presence of the SmE/F/G heterotrimer methylated SmE (Fig. 5A, bottom panel, lanes 2,4), with methylation of this subunit typically exceeding the level observed for full-length Lsm11 (Fig. 5A, bottom panel, lane 2, Supplemental Fig. S2, lane 6, and Supplemental Fig. S3, lane 3), supporting the conclusion that SmE is a legitimate PRMT5 methylation target. Lsm11ΔNL that lacks the amino-terminal RG cluster failed to undergo any detectable methylation (Fig. 5A, bottom panel, lanes 3,4). Only background proteins and no components of the methylosome were purified from HeLa (Supplemental Fig. S2) and mouse (Supplemental Fig. S3) cytoplasmic extracts in the absence of recombinant Lsm10/11 heterodimer. As expected, no protein methylation was detected in these samples.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 5.

In vitro methylation of SmE by endogenous methylosome and identification of the SmE methylation site. (A) Recombinant Lsm10/11 (lanes 1,2) and Lsm10/11ΔNL (lanes 3,4) heterodimers were incubated either in the absence or in the presence of the SmE/F/G heterotrimer with a mouse cytoplasmic extract and purified together with bound endogenous methylosome on amylose beads via MBP attached to Lsm10. The immobilized proteins were tested directly on the beads in a buffer containing ³H SAM for the ability to methylate Lsm11 and SmE. Following overnight methylation, proteins in each sample were stained with Coomassie Blue (top panel) and their methylation status analyzed by fluorography (bottom panel). (B) Sequence of human SmE (amino acids 1–92) and mutations made near the amino terminus. The arrows indicate each of the two arginines in SmE neighboring a glycine. (C) Recombinant methylosome complex consisting of PRMT5, MEP50, and pICln was incubated with Lsm10/11ΔNL heterodimer in the presence of ³H SAM and SmE/F/G heterotrimer containing WT, R-A or Δ4aa variants of SmE. Proteins used in the assay were resolved by SDS-PAGE and visualized by staining with Coomassie Blue (top panel) and their methylation status analyzed by fluorography (bottom panel). Lsm10-MBP and SmE were additionally detected by western blotting using αMBP and αSmE antibodies, respectively (middle panels). (D) Two carboxy-terminally biotinylated peptides encompassing the first 27 amino acids of SmE and either lacking or containing symmetric dimethyl group at Arg4 were incubated with a GST-tagged amino-terminal region of SMN (amino acids 1–146) containing the Tudor domain. Proteins immobilized on streptavidin beads were resolved by SDS-PAGE and probed using anti-GST antibody. Lane 1 represents the background observed in the absence of any biotinylated peptide. The input (5%) for all binding reactions is shown in lanes 4–6.

Mutational analysis identifies the site of methylation in SmE

To prove that Lsm10/11 heterodimer promotes methylation of SmE and to identify the site of methylation, we used mutational analysis. The sequence of SmE contains two arginines directly neighboring a glycine residue: an RG dipeptide located near the amino terminus (amino acids 4–5) and a GR dipeptide (amino acids 75–76) located in the fourth β-strand of the Sm fold (Fig. 5B). The amino-terminal arginine is located within the region that is likely unstructured and this residue was considered a primary candidate for methylation in SmE. We substituted this arginine with alanine (R-A) or deleted four amino-terminal amino acids including the RG dipeptide (Δ4aa) and tested a SmE/F/G heterotrimer containing the resultant SmE mutants for the ability to undergo methylation by recombinant PRMT5 methylosome. Importantly, both mutations tested in the context of heterodimers containing ΔL Lsm11 (Fig. 5C, bottom panel, lanes 1–3) or ΔNL Lsm11 (Fig. 5C, bottom panel, lanes 4–6) abolished methylation of SmE. The effect of these two mutations was particularly striking when tested in the presence of Lsm10/11ΔNL, which promotes very efficient methylation of wild-type SmE (Fig. 5C, bottom panel, compare lane 4 with lanes 5,6).

Since the three subunits of the SmE/F/G often migrate as a cluster of weakly separated bands in 15% SDS/polyacrylamide gels and are difficult to identify as individual proteins, we used western blotting and anti-SmE antibody to confirm the presence of SmE in the samples that failed to show SmE methylation. We also used anti-MBP antibody to confirm the identity of the Lsm10 subunit visualized by Coomassie Blue staining (Fig. 5C, top panel). Bands corresponding to SmE and Lsm10-MBP were detected in all six samples (Fig. 5C, middle panels). Based on these experiments, we conclude that the arginine in position four of SmE is the only residue of the SmE/F/G heterotrimer that becomes methylated by PRMT5/MEP50 in the presence of pICln and Lsm10/11 dimer.

The methylation site in SmE in both its location near the amino terminus and the presence of only one flanking glycine sharply contrasts with the methylation sites in Sm protein D1, B, and D3, which are located within carboxy-terminal GR-rich regions. Symmetrically dimethylated sites in these Sm subunits are bound by the SMN Tudor domain, facilitating the assembly of the Sm ring on spliceosomal snRNAs (Friesen et al. 2001a; Selenko et al. 2001; Cote and Richard 2005; Tripsianes et al. 2011). We tested whether the unusual methylation site in SmE is recognized by the Tudor domain of SMN. We ordered the chemical synthesis of two peptides containing biotin at the carboxyl terminus and encompassing the first 27 amino acids of SmE (Fig. 5D). Of the two peptides, one was unmodified and the other contained symmetric dimethyl group at Arg4. A binding assay was carried out in the presence of each peptide and GST-tagged SMN fragment (amino acids 1–146) containing the Tudor domain. In the absence of any peptide, no GST-SMN protein accumulated on streptavidin beads (Fig. 5D, lane 1). Importantly, the SmE-N(Me₂s) peptide, but not the unmodified peptide, readily bound the SMN fragment (Fig. 5D, lanes 2,3), indicating that symmetrically dimethylated Arg4 of SmE is recognized by the SMN Tudor domain.

Structural studies on the methylosome–U7 6S complex

Methylation of SmE by both recombinant and endogenous methylosome prompted us to assemble an entirely recombinant complex consisting of U7-specific 6S (U7 6S) intermediate and the methylosome, and to test whether it is amenable for structural studies. The U7 6S intermediate was generated by mixing Lsm10/11 heterodimer, SmE/F/G heterotrimer, and pICln, and shown to survive gel filtration chromatography (Fig. 6A), confirming that the hexameric complex is stable. The U7 6S intermediate together with recombinant PRMT5/MEP50 formed a larger complex that eluted as a single peak on gel filtration chromatography (Fig. 6A). Interestingly, gel filtration studies suggest that the methylosome heterooctamer binds only one copy of the U7 6S complex (4:1 molar ratio of PRMT5/MEP50 to U7 6S). The same ratio was observed when a large molar excess of the U7 6S complex was used, suggesting that the methylosome contains only one active binding site for the U7 6S. When applied to the cryo-EM grids, the complex became disrupted, with only the methylosome remaining stable. To stabilize the complex, we used cross-linking with 4 mM BS3 but failed to observe any density for SAM in the EM map. We repeated the cross-linking with 8 mM BS3 in the presence of the SAM analog inhibitor sinefungin (SFG). This approach allowed us to determine a structure of the methylosome at 2.69 Å resolution (Fig. 6B; Supplemental Table S1), yielding well-defined EM density for the adenosine portion of SFG. The rest of SFG had weaker density and assumed different conformations in the four PRMT5 active sites.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 6.

Structural studies of the human methylosome–U7 6S complex. (A) Gel filtration profiles of Lsm10/11 bound to SmE/F/G and pICln (U7 6S complex, green), methylosome bound to U7 6S (blue), and methylosome bound to U7 6S in the presence of sinefungin (SFG, orange). Inserts are SDS gels of the purified U7 6S sample (right) and methylosome–U7 6S sample (left). Lsm10 is amino-terminally fused with MBP, and the amino-terminal region and internal loop of Lsm11 are deleted (Lsm11ΔNL). (B) Schematic drawing of the human methylosome–U7 6S complex. Only the PBM of pICln is included in the atomic model, and the rest of the U7 6S is flexible. (C) Cryo-EM density for the methylosome–U7 6S complex, low pass filtered to 7 Å resolution. The atomic model for the methylosome is shown. The extra EM density that could correspond to U7 6S is labeled. (D) Overlay of two PRMT5 molecules with different conformations in the methylosome–U7 6S complex. The PRMT5 molecule in contact with the putative U7 6S EM density is shown in cyan and the other molecule in gray. The disordered segments, 292–306 and 311–315 or 311–329, are labeled. Panel D was produced with Chimera (Goddard et al. 2007).

Weak density was observed associated with one of the PRMT5 molecules in the EM reconstruction (Fig. 6C). The density could not be conclusively interpreted as the U7 6S, although the stoichiometry was consistent with gel filtration studies. The binding of U7 6S to the methylosome is likely highly dynamic. The density binds one PRMT5 molecule and has no contact with MEP50. It is not clear based on the map why there is only one U7 6S binding site in the PRMT5 tetramer.

The structural analysis indicates that the four PRMT5 molecules in the methylosome assume two distinct conformations, with rms distance of 0.4 Å within each pair of the molecules with similar conformation. In contrast, the rms distance between two PRMT5 molecules with different conformation is 0.9 Å, and the differences are mostly located at the interface between the catalytic and TIM barrel domains of PRMT5 (Fig. 6D). This interface is also where the density for U7 6S is located, and residues 292–306 that link the two domains are ordered in the region where U7 6S contacts PRMT5 but disordered in the other PRMT5 molecule. These residues are also disordered in the structure of the methylosome–Lsm10/11 complex, although they are ordered in the complex with the histone H4 peptide (Antonysamy et al. 2012).