Determinants of the Interaction of the Spinal Muscular Atrophy Disease Protein SMN with the Dimethylarginine-modified Box H/ACA Small Nucleolar Ribonucleoprotein GAR1*

Deletion or mutation of theSMN1 (survival of motorneurons) gene causes the common, fatal neuromuscular disease spinal muscular atrophy. The SMN protein is important in small nuclear ribonucleoprotein (snRNP) assembly and interacts with snRNP proteins via arginine/glycine-rich domains. Recently, SMN was also found to interact with core protein components of the two major families of small nucleolar RNPs, fibrillarin and GAR1, suggesting that SMN may also function in the assembly of small nucleolar RNPs. Here we present results that indicate that the interaction of SMN with GAR1 is mediated by the Tudor domain of SMN. Single point mutations within the Tudor domain, including a spinal muscular atrophy patient mutation, impair the interaction of SMN with GAR1. Furthermore, we find that either of the two arginine/glycine-rich domains of GAR1 can provide for interaction with SMN, but removal of both results in loss of the interaction. Finally, we have found that unlike the interaction of SMN with the Sm snRNP proteins, interaction with GAR1 and fibrillarin is not enhanced by arginine dimethylation. Our results argue against post-translational arginine dimethylation as a general requirement for SMN recognition of proteins bearing arginine/glycine-rich domains.

The best established role of SMN, the protein implicated in spinal muscular atrophy, is in the assembly of small nuclear RNA (snRNA) 1 -protein complexes that function in pre-mRNA splicing. SMN interacts with the common snRNP proteins Sm B/BЈ, D1 and D3, and LSm 4 through arginine/glycine (RG)-rich domains present in these proteins (1)(2)(3)(4). Several studies have demonstrated that SMN and the other proteins of the SMN complex are required to facilitate the assembly of snRNPs and the generation of active spliceosomes (5)(6)(7)(8)(9)(10)(11).
SMN appears to interact with a series of cellular proteins via a common mechanism that depends on the RG-rich domains in the target proteins (reviewed in Ref. 12). In addition to the Sm and LSm snRNP proteins, SMN has been found to interact with RNA helicase A (13), a protein associated with RNA polymerase II; coilin (14), the signature component of nuclear Cajal bodies; heteronuclear RNP binding proteins Q, R, and U (9,15,16); and fibrillarin and GAR1 (17,18), core protein components of the Box C/D and Box H/ACA snoRNPs, respectively. Most of these proteins contain a single RG-rich region that has been shown to be essential for association with SMN. GAR1 has two RG-rich domains, and interestingly, it has been reported that both of these are necessary for interaction (18). Current evidence supports two different models for the basis of the interaction of SMN with partner proteins. One series of studies indicates that sequences near the carboxyl terminus of SMN mediate the interactions (including the interaction with GAR1) (2,11,13,16,18). This domain of SMN is important for the oligomerization of SMN (11,19,20). Other laboratories have implicated the Tudor domain of SMN in its interaction with some of the same RG domain proteins (3,4,7,14,17). Tudor domains, which have been found in several proteins (21), have been proposed to play a role in protein-protein interaction (4). Each of the models suggests that SMN interacts with multiple different proteins via a similar mechanism but differs in the domain of SMN that mediates the interactions. Based on the common features of the interactions, it has been proposed that SMN may be involved in the assembly of transcription complexes and snoRNPs as well as snRNPs (reviewed in Ref. 12). An additional determinant in the interaction of SMN with snRNP components is the methylation state of the target proteins. Arginines within the RG-rich domains of the Sm and LSm ("like Sm") proteins are post-translationally methylated (3,22). The symmetric dimethylarginines (sDMA) that occur in Sm B/BЈ, D1 and D3, and Lsm 4 have been demonstrated to strengthen the binding of SMN (1,3). A logical hypothesis is that methylation within the RG-rich domains of other target proteins would also enhance SMN binding (1). However, unlike the Sm proteins, most SMN target proteins including fibrillarin (23,24), GAR1 (25), and heterogeneous RNP proteins (26,27) contain the more common asymmetric dimethylarginines (aDMA) (28,29). It is not known whether dimethylation influences the binding of SMN to these other target proteins. GAR1, the primary focus of the studies presented here, is a conserved nucleolar protein that is specifically associated with Box H/ACA family small nucleolar RNAs (snoRNAs) (30 -32) and vertebrate telomerase RNA (33). The Box H/ACA snoRNAs direct site-specific pseudouridylation and processing of rRNA (34 -36). GAR1, along with NHP2, NOP10, and dyskerin (Cbf5p in yeast) comprise a group of core proteins associated with all Box H/ACA snoRNAs (32,37,38). The specific role of GAR1 in Box H/ACA snoRNP function is not yet known. GAR1 is so named because of the presence of two RG-rich (GAR) domains: one at each its amino and carboxyl termini. Studies in yeast indicate that GAR1 is a substrate for asymmetric arginine dimethylation (25).
Here we present evidence that SMN interacts directly and specifically with GAR1. The SMN/GAR1 interaction is mediated by the Tudor domain of SMN and requires only one of the two RG-rich domains found in GAR1. Furthermore, we show that single missense mutations in the Tudor domain of SMN, including a mutation isolated from an SMA patient (E134K), have a negative effect on the ability of SMN to bind GAR1 in vitro. The panel of Tudor domain mutations affected the binding of GAR1 and Sm proteins in a similar manner, suggesting that similar contacts are made in the binding of SMN to the snoRNP and snRNP proteins. In addition, we have tested the role of arginine dimethylation on the interaction of SMN with both fibrillarin and GAR1. Our results indicate that whereas methylation of snRNP proteins facilitates SMN binding, and whereas snoRNP proteins are also substrates for methylation, arginine dimethylation is not a key determinant in the interaction of SMN with fibrillarin and GAR1. These findings indicate that dimethylation of arginine may only play a role in the interaction of SMN with a subset of its RG-rich target proteins and reveal an interesting difference between the interactions of SMN with snRNPs and snoRNPs, which otherwise appear to occur via similar mechanisms.
To confirm that GAR1 and fibrillarin proteins underwent arginine methylation during translation in the lysates (as has been previously observed for the Sm proteins) (3), 1.0 Ci of 3 H-labeled S-adenosylmethionine (73.0 Ci/mmol; Amersham Biosciences) was added to the lysate prior to the addition of template DNA. Arginine methylation of the proteins was inhibited by preincubating the lysates with increasing amounts of S-adenosylhomocysteine (SAH) at room temperature for 10 min prior to the addition of the template DNA. Incorporation of [ 3 H]methyl groups was detected by SDS-PAGE and fluorography.
In Vitro Methylation of Recombinant Proteins by the PRMT1 Enzyme-Purified fibrillarin and GAR1 GST fusion proteins (ϳ1 g each) bound to glutathione-agarose beads were incubated with 1 g of PRMT1 enzyme and 1.25 mM S-adenosylmethionine in 100 mM Tris, pH 8.0, 200 mM NaCl, 2 mM EDTA, and 1 mM dithiothreitol at 37°C for 1 h (40-l total reaction volume). The bead-bound methylated proteins were washed (3ϫ 1 ml with Ipp-250T (50 mM Tris (pH 7.5), 250 mM NaCl, 0.05% Triton X-100)) and used in subsequent protein binding assays. Methylation of the proteins was directly verified in a parallel reaction containing 1.0 Ci of 3 H-labeled S-adenosylmethionine (73.0 Ci/mmol; Amersham Biosciences) followed by analysis by SDS-PAGE and fluorography.
Expression and Purification of PRMT1-The rat PRMT1 gene was subcloned from the original GST-PRMT1 construct (25) into a modified pET28b (Novagen) vector, which contains an NH 2 -terminal tag of MGHHHHHH. Cultures of BL21(DE3) containing the PRMT1 construct were grown at 37°C to an OD ϭ 0.4, shifted to 22°C, and induced with 0.4 mM isopropyl-1-thio-␤-D-galactopyranoside overnight at 22°C. The proteins were purified to homogeneity using nickel chelating, Mono Q, and Sephacryl S300 columns (Amersham Biosciences). The protein concentration was estimated by A 280 absorption using an extinction coefficient of 52,090 M Ϫ1 cm Ϫ1 .
In Vitro Protein/Protein Interaction Assays-For protein binding assays, similar amounts (based on Coomassie staining) of GST-tagged proteins were coupled to glutathione-agarose beads equilibrated in Ipp-250T. Equal amounts of in vitro translated proteins (typically 1-3 l) were added and incubated for 1 h at 4°C in a final volume of 500 l. Following 5ϫ 1-ml washes with Ipp-250T or Ipp-1000T (1 M NaCl, all other components the same), bound proteins were eluted with SDS sample buffer and analyzed by SDS-PAGE and fluorography (using Amplify (Amersham Biosciences)). In experiments that are not shown, the binding assays in Fig. 2C were also performed as previously described by another laboratory (11,18). The primary difference in this case was the use of a binding buffer composed of 50 mM Tris-HCl (pH 7.5), 200 mM NaCl, 2 mM EDTA, 0.1% Nonidet P-40, and protease inhibitors (versus Ipp-250T with 50 mM Tris (pH 7.5), 250 mM NaCl, 0.05% Triton X-100). The data shown in Fig. 3 were quantitated by densitometric scanning of two separate experiments using a Bio-Rad Fluor-S multi-imager with Quantity 1 software.

SMN Interacts with GAR1 but Not Other Box H/ACA snoRNP
Proteins-We have investigated the interaction of SMN with snoRNP proteins in in vitro binding assays. GST-tagged human SMN and GST tag alone were expressed in E. coli, purified using glutathione-agarose, and incubated with 35 S-labeled, in vitro translated proteins. Following thorough washing, bound proteins were analyzed by SDS-PAGE and fluorography. Like the spliceosomal SmB protein, GAR1 specifically bound GST-SMN in the presence of both 250 mM and 1 M NaCl (Fig. 1A). GAR1 did not interact with GST alone. In addition, the other core Box H/ACA proteins dyskerin, NHP2, and NOP10 failed to bind to SMN (Fig. 1A). Similar amounts of GST-SMN and GST alone were used in the binding assay (Fig. 1B). Since GAR1 interacts with SMN but not other components of the SMN complex (18), these results suggest that interaction between the SMN complex and the Box H/ACA snoRNP is mediated by direct interaction of the SMN and GAR1 proteins.
The Tudor Domain of SMN Mediates Interaction with GAR1-In order to assess which regions of SMN are involved in interaction with GAR1, various truncation mutants of SMN ( Fig. 2A) were tested for the ability to bind in vitro translated GAR1. Binding similar to wild type levels (residues 1-294) was observed with an SMN fragment consisting of the NH 2 -terminal 160 amino acids (residues 1-160) as well as with a frag-ment corresponding to amino acids 90 -294 (Fig. 2B). The amino acids that are common to these two active fragments (residues 90 -160) comprise the Tudor domain of SMN, which was previously shown to be necessary and sufficient to mediate association with other RG-rich proteins including Sm proteins and fibrillarin (4,7,17). However, the Dreyfuss laboratory has found that deletion of amino acids 278 -294 (SMN⌬Ex7) or point mutation at amino acid 272 (SMN-Y272C) disrupts interaction with GAR1, fibrillarin, and Sm B, suggesting that the COOH-terminal region of SMN mediates binding (11,18). We found that a fragment comprising the COOH-terminal amino acids 159 -294 failed to associate with GAR1 (residues 159 -294; Fig. 2B). We directly tested the interaction of the previously described SMN⌬Ex7 (residues 1-278) and SMN-Y272C proteins with fibrillarin and Sm B as well as GAR1 and did not observe a significant reduction in binding (Fig. 2C). These mutations, which have been found in SMA patients, appear to disrupt other properties essential for the function of SMN (11,19,20) (and therefore may affect sn(o)RNP assembly in vivo) but do not directly affect binding to these RG-rich target proteins. In Fig. 3, we show that a fragment of SMN consisting of the Tudor domain (amino acids 83-173) exhibits binding to GAR1 comparable with full-length SMN in vitro (binding that was within 10% of that of full-length SMN). Our results indicate that SMN interacts with GAR1 via the Tudor domain.
Single Amino Acid Changes in the Tudor Domain of SMN Affect GAR1 Binding-The structure of the Tudor domain of SMN has been determined by NMR spectroscopy, and the structural basis for interaction with the RG-rich tails of Sm proteins has been described (4). The SMN Tudor domain adopts a barrel-like structure containing five ␤ strands, three loop structures, and a hydrophobic core (4) (shown in Fig. 3A). Selenko et al. (4) found that the RG domains of Sm proteins interact with a binding site formed by loops 1 and 3 and proximal ␤ strands of the Tudor domain. Electrostatic interactions between specific, negatively charged surface residues in this region of the Tudor domain and positively charged arginine residues of the Sm tails contribute to the interactions (4,7).
To further assess the importance of specific amino acid residues in the Tudor domain, various single missense mutations were tested for their effect on SMN binding to GAR1 (Fig. 3). The location of the amino acids tested is shown in Fig. 3A. As has been observed with SmB and fibrillarin (17), mutations of the negatively charged residues E104K and D105K (of loop 1) as well as of Q136A (of ␤ 4 ) and L142A (of ␤ 5 and linking ␤ 5 to ␤ 4 ) reduced GAR1 binding to the Tudor domain by 70 -95% (Fig. 3B). Mutation of residues K97E (of ␤ 1 ), R133E, E135K, N137A (of ␤ 4 ), and S143Q (of ␤ 5 ) did not significantly affect the binding of the Tudor domain to GAR1 (binding of these mutants was within 25% of that observed for the wild type Tudor domain). The amino acids that we found were important for interaction with GAR1 are located in the region of the Tudor domain identified by Selenko et al. to be involved in Sm protein interactions. Our results further define the residues critical for the interaction between the Tudor domain of SMN and GAR1 and indicate that the same amino acids support the interaction of SMN with snRNP and snoRNP proteins (Fig. 3B) (17).
Deletion or mutation of the SMN1 gene is correlated with the occurrence of spinal muscular atrophy (41). A single amino acid substitution (E134K) that maps to the ␤ 4 strand of the Tudor domain has been shown to be present in SMA patients with a severe form (type 1) of the disease (42). Whereas NMR studies have demonstrated that this substitution does not alter the structure of the Tudor domain, it does change the charge distribution in the RG domain binding site (4). In Fig. 3C, we show   FIG. 2. The Tudor domain of SMN mediates GAR1 binding. A, schematic representation of SMN truncation mutants used to determine the region required for GAR1 binding. Amino acids included in each fragment are indicated. The Tudor domain (encoded by exon 3) is shown in black, and the position of the Y272C mutation is indicated. GAR1 binding is indicated for each SMN construct as ϩ (comparable with GST-tagged wild type) or Ϫ (comparable with GST alone). B, GST-tagged SMN fragments were incubated with in vitro translated, 35 S-labeled GAR1. An NH 2 -terminal fragment of SMN (amino acids 1-160) interacts with GAR1 similar to full-length SMN (amino acids 1-294), whereas the COOH-terminal region (amino acids 159 -294) fails to bind GAR1. A fragment including the COOH-terminal region and the Tudor domain (amino acids 90 -294) interacts similarly to full-length SMN. C, GST-tagged SMN proteins were incubated with in vitro translated, 35 S-labeled GAR1 fibrillarin, and SmB. The previously described COOH-terminal SMN mutations (residues 1-278 (or SMN⌬Ex7) and SMN-Y272C) do not impair binding of SMN with GAR1, fibrillarin, or SmB. FIG. 1. SMN interacts with GAR1 in vitro. A, full-length recombinant GST-tagged SMN was incubated with in vitro translated, 35 Slabeled GAR1, dyskerin, NHP2, NOP10, and SmB and washed at either 250 or 1000 mM NaCl. Like SmB, GAR1 specifically associates with SMN, and the interaction is stable at 1 M NaCl. The three other known Box H/ACA snoRNP proteins dyskerin, NHP2, and NOP10 fail to interact with SMN. No interaction was seen between any of the translated proteins and the tag alone (GST). In this and all subsequent figures, input lanes (input) represent 20% of the total amount of in vitro translated protein used in each binding assay. B, corresponding Coomassie Blue-stained protein gels are included in this and subsequent figures to show that approximately equal amounts of recombinant proteins were used in the experiments. Protein bands of different mobilities were aligned in these panels. that this SMA patient mutation (E134K) disrupts the binding of SMN to GAR1 (the mutation decreased binding by ϳ95%). This substitution also abrogates Sm protein and fibrillarin binding (4,7,17), indicating the general importance of Glu 134 in the interaction of SMN with RG-rich SMN partner proteins. The result also indicates that in SMA patients with the E134K mutation, a decrease in the ability of SMN to interact with RG-rich proteins (rather than, for example, a loss in ability to interact with other SMN complex components) may be responsible for decreases in SMN function.
Both the NH 2 and COOH-terminal GAR Domains Can Mediate GAR1 Binding to SMN-The conserved central core of the GAR1 protein is flanked by GAR domains at both the N and C termini (Fig 4A). We tested the importance of the GAR domains in the interaction with SMN by performing binding studies with GAR1 proteins containing both, one, or no GAR domains (Fig. 4B). Similar binding was observed using the full-length GAR1 protein and truncated GAR proteins lacking either the NH 2 -terminal (amino acids 1-62) or COOH-terminal (amino acids 169 -217) GAR domains. We found that the binding of SMN was disrupted only if both GAR domains were deleted. These results indicate that SMN associates via the RG-rich domains of the GAR1 protein and can interact with either the NH 2 -or COOH-terminal GAR domains.

Asymmetric Arginine Dimethylation of GAR1 and Fibrillarin Does Not Enhance SMN Binding-Arginines within the RGrich domains of SMN binding partners, including the Sm and
LSm snRNP proteins and the GAR1 and fibrillarin snoRNP proteins, are methylated post-translationally in vivo (22)(23)(24)(25). The Sm and LSm snRNP proteins receive the relatively rare sDMA, which enhances the affinity of SMN for these proteins (1,3,22). Interestingly, fibrillarin, GAR1, and several other SMN-binding proteins receive the more common, aDMA modification (28,29). At least in the case of fibrillarin, there is evidence that the protein contains exclusively aDMA (24).
We investigated whether the methylation status of fibrillarin or GAR1 influenced SMN binding by comparing the binding of SMN to methylated and unmethylated forms of the snoRNP proteins. Methylated and unmethylated fibrillarin and GAR1 were prepared by in vitro translation in rabbit reticulocyte lysates in the absence or presence of the methyltransferase inhibitor, SAH (Fig. 5). As was previously reported (3), blocking methylation in this manner inhibited the binding of SMN to the SmB protein (Fig. 5A), indicating that methylation enhances the interaction. In contrast, SMN binding to fibrillarin and GAR1 (and SMN oligomerization) was unaffected by the SAH treatment. To demonstrate that GAR1, fibrillarin, and SmB were indeed methylated in the  35 S-labeled full-length and truncated GAR1 proteins. As depicted in A, the GAR1 fragments lacked GAR domains at either the N terminus or C terminus or at both termini. B, similar levels of SMN binding were observed with full-length GAR1 (amino acids 1-217) and GAR1 proteins containing a single GAR domain (amino acids 62-217 and 1-169). No SMN binding was observed when both GAR domains were deleted from GAR1 (amino acids 62-169). All GAR1 constructs failed to interact with the GST tag alone.

FIG. 3. Tudor domain missense mutations affect GAR1 binding to SMN.
A, diagram of the SMN Tudor Domain indicating the relative positions of the amino acids that were altered to assess binding to GAR1. Circled amino acids represent those that, when mutated, resulted in reduced binding to GAR1. B, GSTtagged, full-length SMN (SMN) and nine single missense mutations in a minimal Tudor domain construct (amino acids  were tested for their ability to interact with 35 S-labeled GAR1 in vitro. C, a point mutation found in an SMA patient impairs SMN binding to GAR1. GSTtagged wild type and mutant (E134K) SMN (amino acids 1-160) were tested for their ability to interact with 35 S-labeled GAR1.
lysate (and that SAH treatment inhibited methylation), the translations were performed in the presence of the 3 H-labeled methyl donor, S-adenosylmethionine (Fig. 5B). As expected, the lysate methyltransferases transferred tritiated methyl groups onto GAR1, fibrillarin, and SmB proteins (but not SMN) in the absence but not in the presence of SAH (Fig. 5B). All proteins were synthesized at comparable levels as judged by translation in the presence of 35 S-labeled methionine (Fig. 5C).
The results shown in Fig. 5 clearly indicated that methylated and unmethylated snoRNP proteins interacted with SMN to comparable extents, suggesting that arginine dimethylation is not an important factor in the interaction of SMN with snoRNP proteins. However, we had not specifically determined if the proteins were asymmetrically (or instead symmetrically) dimethylated by the lysate methyltransferases. To directly test whether asymmetric arginine dimethylation (the physiologically relevant modification) affected the binding of GAR1 and fibrillarin to SMN, we methylated the proteins in vitro using recombinant GAR1, fibrillarin, and the class I protein arginine methyltransferase, PRMT1 (Fig 6). PRMT1 is thought to be primarily responsible for asymmetric dimethylation of RGGrich proteins in the cell (28,29). GST-tagged GAR1 and fibrillarin were expressed in E. coli (which does not contain arginine methylation activity (29)), purified, and incubated with PRMT1 under conditions optimized for maximum methylation of the substrate proteins. Using 3 H-labeled S-adenosylmethionine, we directly confirmed the asymmetric dimethylation of the fibrillarin and GAR1 proteins (Fig. 6). Under these reaction conditions, fibrillarin was found to be fully methylated based on estimation from trichloroacetic acid-precipitable counts (data not shown). Binding of the PRMT1-methylated and unmethylated GAR1 and fibrillarin to SMN was tested and found to be similar (Fig. 6). Our results indicate that methylation, including asymmetric arginine dimethylation, of the snoRNP proteins fibrillarin and GAR1 is not a key factor in their interaction with SMN.

DISCUSSION
Spinal muscular atrophy, one of the leading hereditary causes of infant mortality, results from mutation or deletion of the SMN1 gene and a reduction in the levels of functional SMN (41). SMN is present in both the cytoplasm and nucleus of most or all tissues and has been implicated in a wide array of cellular functions (12,43). The role of SMN in the cytoplasmic assembly of spliceosomal snRNPs (complexes of Sm and other proteins and snRNAs involved in pre-mRNA intron splicing) is the clearest function of the protein (5, 7-11, 44). The work presented here and in other recent studies (17,18) suggests that SMN also mediates the nuclear assembly of snoRNPs (complexes of snoRNP proteins and snoRNAs involved in ribosome biogenesis (45)(46)(47)(48)) and perhaps telomerase (a protein-RNA complex that shares features of the Box H/ACA snoRNPs and includes GAR1 (33, 49 -52)). SMN directly interacts with the core snoRNP proteins GAR1 and fibrillarin in vitro and in vivo (see Refs. 17 and 18; this study). Moreover, SMN appears to interact with the snRNP and snoRNP complexes via a similar mechanism that involves the RG-rich domains of the target proteins and either the Tudor domain or COOH-terminal sequences of SMN (reviewed in Ref. 12). Localization studies also support a role for SMN in snoRNP biogenesis and/or function. SMN, like fibrillarin, GAR1, and snoRNAs (and telomerase RNA) is found within Cajal bodies and nucleoli (19,40,50,51,(53)(54)(55)(56)(57)(58)(59)(60)(61), the proposed intranuclear sites of snoRNP biogenesis and function, respectively (46,60,62).
Interaction with RG-rich domains of target proteins appears to be a common mechanism whereby SMN interacts with unrelated components of diverse RNPs, leading to the conception of SMN as a master assembler of a variety of cellular RNPs (reviewed in Ref. 12). In this study, we have investigated the determinants of this key interaction. We have demonstrated that the centrally located Tudor domain of SMN is necessary and sufficient for interaction with GAR1 (Figs. 2 and 3). Pre- vious studies by us and others indicated that the Tudor domain is also the primary binding site for fibrillarin (17) and snRNP proteins (3,4,7). In contrast, investigations from another laboratory suggest that amino acids at the C terminus of SMN are essential for binding to these same proteins (2,18). We did not find a significant effect of the deletion of the C terminus of SMN (residues 1-278 or SMN⌬Ex7) or of the Y272C point mutation on interaction with GAR1, fibrillarin, or Sm B in this study (Fig. 2C) even with the assay conditions described in the previous work (see "Experimental Procedures"; data not shown). Among known SMN target proteins, GAR1 is unique in having two RG-rich domains, and we have shown that either domain can mediate interaction with SMN (Fig. 4). This observation is consistent with the fact that SMN interacts with multiple proteins that contain a single RG domain (2, 4, 11, 13, 16 -18) but again contrasts with a recent report indicating that both of the RG domains of the GAR1 protein are necessary for SMN binding (18). The basis of the differences observed within the field remains to be determined.
For over 30 years, it has been known that arginines of RGrich protein domains are subject to post-translational dimethylation in eukaryotes (28,29). A family of protein arginine methyltransferases catalyzes the transfer of one, or more commonly two, methyl groups from S-adenosylmethionine to the side chain (guanidino) nitrogen of arginine. Most PRMT family members, referred to as type I enzymes and including PRMT1, generate aDMA (63). Asymmetric arginine dimethylation is believed to be the most common arginine methylation of cellular proteins, and PRMT1 (Hmt1/Rmt1 in yeast) appears to execute most of these modifications (28,29,63,64). PRMT1 is localized primarily in the nucleus (65). Proteins known to contain aDMA include nucleolin (66), poly(A)binding proteins (67), and several heterogeneous RNP proteins (26,27,69,70) as well as the snoRNP proteins GAR1 (25) and fibrillarin (23,66). Thus far, PRMT5 (JBP1) (71), which has been shown to add sDMA to RG-rich substrates including a subset of Sm and LSm snRNP proteins (22) and the myelin basic protein (72), is the only class II enzyme that has been identified. PRMT5 functions in the cytoplasm (as a component of the "methylosome") to symmetrically dimethylate snRNP proteins (44,73).
SMN interacts both with proteins that contain sDMA (e.g. Sm and LSm snRNP proteins) and aDMA (e.g. fibrillarin, GAR1, and heterogeneous RNP proteins). Recent studies have shown that binding of SMN to Sm and LSm snRNP proteins is greatly enhanced by symmetric dimethylation of the proteins (1, 3). Accordingly, it was thought that dimethylarginine may promote interaction of SMN with the full array of its RG-rich target proteins (1). However, our results indicate that post-translational arginine methylation, which is believed to be an irreversible protein modification, is not a general requirement for substrate recognition by SMN. Whereas we also observed the substantial enhancement in SMN binding with methylation of SmB ( Fig. 5; see Refs. 1 and 3), we did not detect any comparable effect of methylation on GAR1 or fibrillarin (Figs. 5 and 6).
A number of possibilities can be envisioned to explain why methylation plays a critical role in the recognition of the Sm and LSm (sDMA-containing) proteins but not GAR1 and fibrillarin (aDMA-containing) proteins by SMN. SMN may recognize the differences between the symmetric and asymmetric dimethylarginines. sDMA and aDMA are structurally very similar, and both modifications add hydrophobic character and bulk but do not alter the net positive charge of the arginine.
The key difference between the sDMA and aDMA is the distribution of the added methyl groups (a single methyl group added to both terminal nitrogens versus two methyl groups added to a single terminal nitrogen, respectively). Alternatively, enhancement of SMN binding by methylation may only be observed in the context of intrinsically weaker interactions. Conceivably, SMN has a greater affinity for unmodified GAR1 and fibrillarin (due perhaps to another feature such as the regularly spaced phenylalanine residues found in the RG domains of GAR1 and fibrillarin), which is not further enhanced by the addition of methyl groups. Taking this view, methylation of the RG-rich domains of the Sm and LSm proteins may be a requirement for SMN interaction because the unmodified forms of these proteins may have a lower affinity for SMN. Finally, it is possible that small changes in affinity that were not detected in vitro are significant in vivo. Detailed structural studies examining the amino acid contacts between SMN and a number of its substrates in their methylated and unmethylated states should help address this important issue.
Whereas our work suggests that methylation is not a critical determinant for association with SMN, arginine dimethylation may influence other important aspects of the biogenesis or function of GAR1 and fibrillarin. RG domains have been proposed to be important for a variety of functions including nucleic acid (RNA) interaction, protein/protein interaction, and intracellular transport and subcellular localization (28,29). In particular, there is evidence to indicate that the RG-rich domain of fibrillarin is important for localization of fibrillarin to the nucleolus (74) and that the RG-rich domains of GAR1 play an auxiliary role in snoRNA binding (75). However, it is not clear what role arginine dimethylation of the RG domains may play in these proposed functions. Few studies have specifically addressed the effect of methylation on RG domain activity. Furthermore, the available data (from this study and others) suggest that the effect of methylation on protein activity is not predictable. Arginine dimethylation has been reported to have slight or no effect on RNA binding (76,77) and to stabilize, weaken, or have no effect on specific protein/protein interactions (see Refs. 1, 3, 78, and 79; this study).
In spinal muscular atrophy, the decreased levels of active SMN protein result in selective death of spinal motor neurons, muscle wasting, and frequently death during childhood. Mutations in various regions of the SMN gene can result in reduced levels of functional SMN protein. It is unclear why reduction in functional SMN leads to selective death of motor neurons, given the ubiquitous expression of SMN and the evidence that the SMN complex performs fundamental cellular functions such as RNP assembly. The new link to snoRNP (and perhaps telomerase) assembly provides some additional avenues to investigate in pursuit of an understanding of the pathogenesis of spinal muscular atrophy. Selective effects on spinal motor neurons could be a result of a higher requirement for snoRNP activity in motor neurons than in other cell types. It is noteworthy that in the cells of the nervous system (including spinal motor neurons) SMN is more strongly localized to nucleoli (sites of snoRNP function) and that SMN-containing Cajal bodies/gems (proposed sites of snoRNP assembly) appear enlarged and more often associated with nucleoli than in most other cell types (53,54,80,81). Alternatively, the susceptibility of motor neurons may reflect their requirements for cell type-specific snoRNPs. snoRNAs have recently been identified whose expression is limited to brain tissue (68,82,83). Lack of expression of the brain-specific snoRNAs is correlated with the etiology of another neurogenetic disease Prader-Willi syndrome (68,82,83). Further studies are required to determine whether SMN is required for snoRNP assembly and function and whether defects in snoRNP activity contribute to the disruption of cellular function in the motor neurons of spinal muscular atrophy patients.