Structural basis for recognition ofarginine methylated Piwi proteinsby the extended Tudor domainKe Liua,b,1, Chen Chenc,1, Yahong Guob,1, Robert Lamb, Chuanbing Bianb, Chao Xub, Dorothy Y. Zhaoc, Jing Jinc,Farrell MacKenzieb, Tony Pawsonc,d,2, and Jinrong Mina,b,e,2

aHubei Key Laboratory of Genetic Regulation and Integrative Biology, College of Life Science, Huazhong Normal University, Wuhan 430079, People’sRepublic of China; bStructural Genomics Consortium, University of Toronto, 101 College Street, Toronto, ON, M5G 1L7, Canada; cSamuel LunenfeldResearch Institute, Mount Sinai Hospital, Toronto, ON, Canada M5G 1X5; dDepartment of Molecular Genetics, University of Toronto, Toronto, ON, CanadaM5S 1A8; and eDepartment of Physiology, University of Toronto, Toronto, ON, Canada M5S 1A8

Contributed by Tony Pawson, September 2, 2010 (sent for review August 23, 2010)

Arginine methylation modulates diverse cellular processes andrepresents a molecular signature of germ-line-specific Piwi familyproteins. A subset of Tudordomains recognize argininemethylationmodifications, but the binding mechanism has been lacking. Herewe establish that, like other germ-line Tudor proteins, the ancestralstaphylococcal nuclease domain-containing 1 (SND1) polypeptide isexpressed and associates with PIWIL1/Miwi in germ cells. We findthat human SND1 binds PIWIL1 in an arginine methylation-depen-dent manner with a preference for symmetrically dimethylatedarginine. The entire Tudor domain and a bifurcated SN domainare required for this binding activity, whereas the canonical Tudordomainalone is insufficient formethylarginine ligandbinding. Crys-tal structures show that the intact SND1 extended Tudor domainforms a wide and negatively charged binding groove, which canaccommodate distinct symmetrically dimethylated arginine pep-tides from PIWIL1 in different orientations. This analysis explainshow SND1 preferentially recognizes symmetrical dimethylargininevia an aromatic cageand conservedhydrogenbonds, andprovides ageneral paradigm for the binding mechanisms of methylarginine-containing peptides by extended Tudor domains.

Piwi-interacting RNA ∣ Tudor domain-containing proteins

Arginine methylation, the covalent addition of methyl group toarginine residues, is a form of protein posttranslational mod-

ification that is increasingly recognized for its importance in med-iating protein–protein interactions (1, 2). Three types of argininemethylation—namely, monomethylation, asymmetrical dimethy-lation, and symmetrical dimethylation—are catalyzed by type Ior type II protein arginine methyltransferases (PRMTs). Amongthese, PRMT5 is considered the major PRMT that induces sym-metrical dimethylation of arginine residues on target proteins (2).

Piwi proteins, the germ-line-specific clade of Argonaute familyproteins, together with their associated Piwi-interacting RNAs(piRNAs) are evolutionarily conserved in animals and playpivotal roles in transposon silencing during gametogenesis (3, 4).Mammalian Piwi proteins, which include PIWIL1/Miwi, PIWIL2/Mili, PIWIL3 and PIWIL4/Miwi2, contain multiple arginine-gly-cine (RG) and arginine-alanine (RA) repeats at their N termini,and methylation of these arginine sites is important for recruitingthe Tudor domain-containing (TDRD) group of germ-line-en-riched Tudor domain proteins such as Tdrd1, Tdrd2/Tdrkh,Tdrd4/RNF17, Tdrd5, Tdrd6, Tdrd7, Tdrd8/Stk31, and Tdrd9(5–12). Arginine methylation is a conserved posttranslationalmodification of Piwi family proteins of various species includingDrosophila, Xenopus, and mammals (7, 13). Its importance hasbeen demonstrated by null mutants of drosophila PRMT5 (dart5,csul), in which arginine methylation of Piwi family members iscompromised, leading to retrotransposon activation and defectsin germ-line development (7, 14, 15).

The Tudor domain, together with Chromo, MBT, PWWP, andAgenet-like domains, comprise the “royal family” of proteindomains that engage in protein–protein interactions (16, 17).The Tudor domain is characterized by a β-barrel core and in manycases an aromatic cage suitable for docking methylated lysine orarginine (16, 18). The Tudor domain family is largely divided intotwo groups: a methyllysine binding group (i.e., JMJD2A, 53BP1)and a methylarginine binding group (SMN, TDRD group [Tdrd1-Tdrd12]) (19). Crystal structures of 53BP1 tandemTudor domainsin complex with histone H4K20me2 and a JMJD2A hybrid Tudordomain with histone H3K4me3 have provided the structural basisfor Tudor domain recognition of methylated lysine residuesthrough aromatic cages (20, 21). On the contrary, much less isknown about the binding mechanism of methylarginine-specificTudor domains, although several lines of biochemical evidencesuggest they have a preference for binding proteins with symme-trically dimethylated arginine (sDMA) (22). A unique structuralfeature of Tudor domains from the TDRD group is an additionalα-helix and two β-strands N-terminal to the canonical Tudor coredomain (19), which may assist this subfamily of Tudor domain inbinding to its methylated arginine ligands, and are absent frommethyllysine-binding Tudor domains. This extended sequencedomain is annotated as a Maternal Tudor domain (23). Among12 TDRD Tudor proteins, SND1 (TDRD11, staphylococcalnuclease domain-containing 1, Tudor-SN, P100) is the most an-cient molecule to possess this N-terminal extension, and we haveproposed that this Tudor domain may be a precursor for all germ-line TDRD Tudor domains (19). Therefore, elucidating the bind-ing of the SND1 extended Tudor domain to its ligands is of generalimportance in understanding the structural determinants under-lying specific arginine methylation-dependent interactions andthe recognition modes of arginine methylated peptide motifs.

In order to test whether a full extended Tudor domain isrequired for methylarginine binding, and to establish a prototypicbinding mechanism for arginine methylation-dependent Piwi-Tudor interactions, we have investigated the crystal structures ofthe human SND1 extended Tudor domain in complex with sym-metrically dimethylated arginine peptides derived from PIWIL1.

Author contributions: K.L., C.C., Y.G., T.P., and J.M. designed research; K.L., C.C., Y.G., R.L.,C.B., C.X., D.Y.Z., and F.M. performed research; K.L., C.C., Y.G., R.L., C.B., C.X., D.Y.Z., J.J., T.P.,and J.M. analyzed data; and C.C., T.P., and J.M. wrote the paper.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.

Data deposition: The atomic coordinates, crystallography, and structure factors havebeen deposited in the Protein Data Bank, www.pdb.org (PBD ID codes 3OMC and 3OMG).1K.L., C.C., and Y.G. contributed equally to this work.2To whom correspondence may be addressed. E-mail: pawson@lunenfeld.ca or jr.min@utoronto.ca.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1013106107/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1013106107 PNAS Early Edition ∣ 1 of 6

BIOCH

EMISTR

Y

These data delineate the structural basis for preferential bindingof symmetrically dimethylated arginine by an extended Tudordomain, which is of general relevance to the family of germ-linemammalian Tudor proteins, and identify distinct mechanisms bywhich the SND1 extended Tudor domain recognizes argininemethylated peptide motifs.

ResultsSND1 Associates with Miwi in Vivo, and Its Extended Tudor DomainPreferentially Binds Piwi Peptides with sDMA. Several TDRD pro-teins associate with Piwi proteins in mouse testes or in cotrans-fected cells, and their interactions are most likely argininemethylation-dependent (5). For example, we have establishedthat Tdrkh is a major binding partner of Miwi, and that the Tudordomain is critical in mediating Tdrkh–Miwi interaction (8).Because the crystal structure of the apo-SND1 Tudor domainis highly similar to that of the Miwi-binding germ-line Tudor pro-tein Tdrkh, in terms of overall binding surface and aromatic cageresidue position (8, 24), we asked whether SND1 is capable ofbinding Miwi. Because it is not known whether the SND1 proteinis expressed in male germ cells, we performed immunoblottingto examine its tissue distribution in mice. SND1 was highlyexpressed in pancreas and liver, with moderate to weak levels inother tissues including testis (Fig. S1). We next analyzed SND1expression dynamics during spermatogenesis (Fig. 1A). SND1was undetectable or weakly expressed at postnatal week 1, butexhibited elevated protein levels from postnatal week 2 to adult,correlating with the onset of meiosis and Miwi expression (25).

To examine its potential association with Miwi, we immunopre-cipitated SND1 from testis lysates and immunoblotted with Miwiantibody and found that SND1 coprecipitated with Miwi as com-pared with a negative control IgG immunoprecipitation (Fig. 1B).These results indicate a potential unique involvement of the mul-tifunctional SND1 in the Piwi/piRNA pathway in the germ cells.

SND1 consists of four staphylococcal nuclease-like (SN)domains, and a Tudor domain, which is embedded in a fifth SNdomain (Fig. 2A). We refer to the Tudor domain and the fifthSN domain together as an extended Tudor domain in this study.To pursue the endogenous association with Piwi identified byimmunoprecipitation in the mouse, and to better understand thebinding specificity of SND1, we used a fluorescence polarizationbinding assay tomeasure the affinity of the humanSND1 extendedTudor domain for an N-terminal peptide from human PIWIL1protein, which is 100% conserved with mouse Miwi. This peptidecontains multiple RA/RG repeats, and was synthesized with asymmetrical dimethylation modification at one of three differentarginine sites (R4, R10, or R14) (Fig. 1C–E). The SND1 extendedTudor domain bound all three of these sDMA peptides (Fig. 1 Dand E). Simultaneous methylation of all three arginine residuesdid not markedly affect binding. SND1 bound to the R4me2s pep-tide (symmetrical dimethylation at arginine 4) with aKD of 10 μM.The binding affinity was about twofold weaker for a peptide thatwasmonomethylated at arginine 4 (R4me1), about fourfold weak-er for a peptide with asymmetrical dimethylation at arginine 4(R4me2a), and about ninefold weaker for an unmodified R4peptide at a salt concentration of 50 mM NaCl (Fig. 1 D and E).Hence, the SND1 extended Tudor domain preferentially recog-nizes symmetrically arginine-dimethylated PIWIL1/Miwi pep-tides. At a higher salt concentration of 150 mM NaCl, thebinding affinities were reduced roughly fourfold for each peptide(Fig. 1 D and E), indicating that electrostatic interactions play arole in the SND1 interaction with PIWIL1 N terminus.

Overall Structures of the SND1 Extended Tudor Domain in Complexwith Two PIWIL1 Peptides, R4me2s and R14me2s. To gain insight intothe molecular mechanism of selective binding of the SND1extended Tudor domain to symmetrically arginine-dimethylatedproteins, we determined the crystal structures of the SND1extended Tudor domain in complex with R4me2s and R14me2speptides (Fig. 2 and Fig. S2). The apostructure of the SND1extended Tudor domain has been determined before (24) andis almost identical to that in our structures of the domain com-plexed with peptides. Based on our structures of these complexes,we can divide the SND1 extended Tudor domain into three mod-ules: the canonical Tudor domain colored in blue, the fifthSN-like domain colored in cyan and green, and the linker α-helixcolored in yellow. The SN-like domain is split into two segmentsby the Tudor domain: the N-terminal two ß-strands colored incyan and the C-terminal three α-helices and three ß-strandscolored in green (Fig. 2). The canonical Tudor domain, the linkerα-helix, and the N-terminal two ß-strands of the SN domain formthe maternal Tudor domain sequence.

In order to test the significance of the N- and C-terminalsegments of the SN domain in Piwi protein binding, we madetwo deletion mutants by separately truncating the N- and C-term-inal parts of the SN domain. Binding data obtained from thesemutants showed that deleting the N-terminal two ß-strands ofthe SN domain (construct covering residues 700–910) totallydisrupted binding, and deleting the C-terminal part of the SNdomain (residues 650–800) severely diminished binding (KD ¼300 μM) (Table 1). Therefore, both segments of the SN domainare critical for binding of the extended Tudor domain to itsligands. TDRKH is a germ cell-specific Piwi-binding Tudor pro-tein, which has a 25% sequence identity and 43% sequence simi-larity with the extended Tudor domain of SND1, and is thereforepredicted to possess an extended Tudor domain (Fig. S2C). To test

IgG

IP

SN

D1

IP

Miw

i IP

WB: Miwi

Miwi

B

IgG

D

E

ES 1wk 2wk 3wk 8wk DLD1

Testis

13095 72

SND1

Tubulin

APIWIL1 N terminus

C

Fig. 1. SND1 associates with Miwi in vivo and its extended Tudor domainpreferentially binds Piwi peptides with sDMA. (A) Expression kinetics ofSND1 in mouse testes of different developmental ages. ES, mouse embryonicstem cells; DLD1, a human colon carcinoma cell line. (B) Association ofSND1 with Miwi in adult testis. Testis lysates was immunoprecipitated usinganti-SND1, anti-Miwi, and IgG control antibodies and subject to anti-Miwiantibody immunoblotting. (C) The N terminus sequence of human PIWIL1or mouse Miwi. The methylation of those arginine residues colored in redis studied in this work. (D) The binding curves of fluorescence polarizationmeasurements of different PIWIL1-R4 peptides to SND1 extended Tudordomain. (E) Binding affinities of SND1 extended Tudor domain to differentPIWIL1 peptides.

2 of 6 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1013106107 Liu et al.

if the binding characteristics of the extended Tudor domain areconserved, we measured the binding affinities of TDRKH Tudordomain with different boundaries to the R4me2s peptide andfound that TDRKH (residues 290–561) binds R4me2s with a KDof 4 μM. As predicted, binding was severely diminished (KD ¼310 μM) following C-terminal truncation of the TDRKH Tudordomain (residues 290–430), and was undetectable for a constructcovering just the canonical Tudor domain (residues 327–420)(Table 1). These truncation mutants are well behaved during pur-ification and one truncation mutant was crystallized previously(TDRKH, residues 327–420) (8). Therefore, both the N- andC-terminal parts of the split SN domain of SND1, in additionto the Tudor core domain, are essential for ligand binding, and thisarrangement is likely conserved in the extended Tudor domains ofgerm-line-specific Tudor proteins such as TDRKH and Tudor, anotion that is supported by a recent study published during thepreparation of our manuscript (26).

Because the SN domain contains five β-strands, which resem-bles a Tudor domain, we overlaid the fifth SN domain and thecanonical Tudor domain of SND1, and found that the ß-barrelcore of the SN domain can be superimposed with the Tudordomain (Fig. S2B), although the Tudor-like ß-barrel core of theSN domain lacks an aromatic cage. Therefore we propose thatthe extended Tudor domain harbors unique interdigitated tan-dem Tudor domain folds.

The mechanistic basis for the importance of the SN domain inligand binding was revealed by the structures of the extendedSND1 Tudor domain with complexed peptides. These showed

that all three modules (the N- and C-terminal SN segmentsand the Tudor core) are involved in binding the methylargininepeptides (Fig. 2). Together the three modules form a wide nega-tively charged binding groove that accommodates the methylar-ginine containing peptides. This finding is consistent with the lossof binding displayed by the deletion mutants, because the struc-tures predict that deleting any part of the extended Tudor domainwould severely diminish peptide binding.

PIWIL1 R4me2s and R14me2s Peptides Bind the Extended TudorDomain of SND1 in Reverse Directions. Surprisingly, the R4me2sand R14me2s peptides bound the SND1 extended Tudor domainin reverse directions, although the methyl-guanidinium group ofthe symmetrical dimethylarginine could be well superimposed(Fig. 3A). Both R4me2s and R14me2s peptides form extensiveinteractions with SND1 extended Tudor domain (Fig. 2). Inthe crystal structure of the SND1-R4me2s complex, the Ala7 ofthe peptide is inserted into a shallow pocket and Ala5 is pointedto the sidewall of the binding groove (Fig. 2B and Fig. S3). Arg6forms electrostatic and water-mediated hydrogen bonding inter-actions with SND1 (Fig. 2B and Fig. S3). This conformation canexplain why the RA/RG motif is preferred by SND1. However, inthe case of the SND1-R14me2s complex, if the R14me2s stilladopted the same orientation as the R4me2s peptide, theGlu17 residue would occupy the pocket filled by Ala7 in the caseof the SND1-R4me2s structure, which would be too shallow forGlu17, and more importantly would cause charge expulsion(Fig. 2 and Fig. S3). Therefore, from an energetic point of view,

A

B

C

Fig. 2. Crystal structures of human SND1 extendedTudor domain in complex with PIWIL1 peptides. (A)Domain structure of SND1. SN, staphylococcal nucle-ase-like domain. (B) Complex structure of SND1 andR4me2s peptide: The SND1 extended Tudor domainis shown in cartoon, and the R4me2s peptide and re-sidues interacting with the R4me2s peptide from SNDare shown in stick models. Hydrogen bonds are shownin red dotted lines. (C) Complex structure of SND1 andR14me2s peptide: The SND1 extended Tudor domainis shown in cartoon, and the R14me2s peptide and re-sidues interacting with the R14me2s peptide fromSND are shown in stick models. Hydrogen bondsare shown in red dotted lines.

Liu et al. PNAS Early Edition ∣ 3 of 6

BIOCH

EMISTR

Y

it is favorable for the R14me2s peptide to reside in a reverse di-rection in the wide binding groove (Fig. 3A), reminiscent of SH3domains, which used a conserved binding surface to recognizeproline-rich peptides in opposite directions (27, 28).

By superimposing the R4me2s and R14me2s structures, wenote that the peptides do not overlay well except for the methy-larginine group (Fig. 3A). It is conceivable that the wide bindinggroove of the SND1 extended Tudor domain provides plasticity toaccommodate methylarginine peptides in different bindingmodes (Fig. 3A). In addition to the first cluster of RA/RG repeatsin PIWIL1, Piwi proteins harbor methylarginines in other se-quence contexts (5). The wide binding groove of the extendedTudor domain in SND1 potentially enables it to recognize multi-ple different methylarginine motifs, a prediction that warrantsfurther investigation.

SND1 Extended Tudor Domain Preferentially Recognizes Symmetri-cally Arginine-Dimethylated Peptides. Like other royal family mem-bers, the binding of the symmetrical dimethylated arginine by theextended Tudor domain involves an aromatic cage. The aromaticresidues F740, Y746, Y763, and Y766 form a rectangle cuboidcage with the aromatic rings of F740 and Y766 parallel to eachother, and those of Y746 and Y763 approximately perpendicularto those of F740 and Y766 (Fig. 3 B and C). The parallel aromaticrings of F740 and Y766 flank the side chain of the methylarginine.Besides the aromatic cage, the methylarginine forms a hydrogenbond with N768 through the NH1/2 group, and interacts withN823 and T688 through water-mediated hydrogen bonds (Fig. 3B and C). Residues F740, Y746, Y763, Y766, and N768 are fromthe canonical Tudor domain, and T688 and N823 are from the SNdomain (Fig. 3 B andC). These interactions are conserved in bothR4me2s and R14me2s peptides. Therefore, both the Tudor andSN domains are involved in methylarginine binding. The intactextended Tudor domain is critical in binding methylated argininepeptides and the methylarginine is bound through cation-π,hydrophobic, van der Waals, and hydrogen bond interactions.

A series of mutagenesis experiments were performed to verifythe importance of these residues in SND1 for methylargininebinding. Mutating F740, Y746, Y763, or Y766 to leucine in SND1reduced the binding to the R4me2s peptide by 11–22-fold(Table S1). Mutating N768 to aspartic acid increased the bindingaffinity twofold, presumably by introducing extra electrostaticinteractions with the methylarginine; in contrast, mutating N768to glutamine or arginine disrupted binding (Table S1). N768R in-troduces an unfavorable positive charge and the N768Qmutationpossibly disrupts the hydrogen bond with the methylarginine.These data show that the methylarginine binding pocket observedin the crystal structures is essential in binding a methylatedarginine peptide.

Our structures can also explain why the SND1 extended Tudordomain preferentially binds symmetrically dimethylated argininepeptides. Monomethylation would reduce hydrophobic interac-tions with SND1, and based on our SND1-R4me2a model(Fig. 3D), arginine asymmetrical dimethylation would disruptthe hydrogen bond between the NH1 group of the methylarginineand N768.

The extended Tudor domain has a similar aromatic cage to thatobserved in the 53BP1 and L3MBTL1 structures (20, 29). How-

ever, 53BP1 and L3MBTL1 both preferentially bind low methyla-tion states of lysine in histone tails (20, 29). The methyllysine-binding pocket consists of three to four aromatic residues andan aspartic acid, and interacts with the dimethylammonium groupof the methyllysine. By comparing the pocket dimensions, espe-cially the parallel aromatic residues, we found that the distancebetween the F740 and Y766 in SND1 is 1.2 Å narrower than thatbetween the Y1502 and Y1523 in 53BP1 (Fig. 3 B, C, and E). Thepocket size limits SND1 so that it can only accommodate theplanar methyl-guanidinium group. This finding is confirmed byour binding results, which showed that SND1 did not bind anyof the lysine-methylated histone peptides we tested (Fig. 1E).On the other hand, histone N-terminal tails also harbor a fewmethylatable arginine sites, such as H3R2 and H4R3. We haveshown that SND1 exhibits some degree of plasticity to accommo-date various methylarginine peptides. Our binding results showthat SND1 bind symmetrically arginine-dimethylated histone pep-tides, albeit a few times more weakly than PIWIL1 peptides(Fig. 1E and Fig. S4).

DiscussionProtein argininemethylation, catalyzed by PRMTs,modulates cel-lular processes such as mRNA splicing, DNA repair, transcriptionregulation, and signal transduction. These functions are believedto be mediated by methylarginine binding proteins. Thus far, asubset of Tudor domains from proteins, such as SMNandTDRDs,are the only modules identified as binding arginine methylationmarks. Here we reported the crystal structures of the extendedTudor domain of SND1 in complex with two PIWIL1 N-terminalpeptides harboring single symmetrically dimethylated arginines,and provide the structural basis for the binding preference ofthe extended Tudor domain for symmetrically dimethylated argi-nine. Like methyllysine-binding domains, the SND1 extendedTudor domain uses an aromatic cage to recognize methylatedresidues. Interestingly, in their unliganded forms, the Tudordomains of methylarginine-binding SMN and TDRD3 also havesuch an aromatic cage, although in theSMNstructure the aromaticcage is not preformed because W102 occupies the binding pocket(Fig. S5). Presumably, W102 in SMN will flip out to allow themethylarginine residue to reside in the aromatic cage. Hence,our complex structures of SND1 with PIWIL1 peptides providea general binding mechanism for methylarginine recognition.

Arginine methylation sites on the N termini of Piwi proteinsare evolutionarily conserved. The RG/RA-rich clusters providedocking sites for proteins with Tudor domains (8), resemblingthe recognition of histone modifications by methyllysine-bindingdomains. Given that there are six arginine sites in the first RG/RA-rich cluster on the PIWIL1/Miwi N terminus, multiple argi-nine residues may be methylated simultaneously. Our bindingdata showed that a peptide with three symmetrically dimethylatedarginines (R4, R10, and R14) only showed slightly increased af-finity toward the extended Tudor domain of SND1, indicatingthat one methylarginine is adequate for binding a single extendedTudor, and that there is no obvious requirement for multivalentinteractions. However, it is conceivable that multiple methylationsites may increase the local concentration of available methylatedligands and enhance the possibility of initial recruitment ofthe extended Tudor domain. On the other hand, some germ-lineTudor domain proteins contain multiple Tudor domains. Forinstance, TDRD6 has eight different Tudor domains, which po-tentially could bind multiple distinct arginine methylation marksat the same time (30). Regardless, our finding that the bindingsurface of the SND1 extended Tudor domain can bind methylatedpeptides in opposite orientations suggests a surprising plasticity inits ability to accommodate arginine methylated sites. During thepreparation of our manuscript, a significant study has shown thecrystal structure of extended Tudor domain of Drosophila Tudorwith arginine methylated peptides of Drosophila Aubergine (26).

Table 1. Binding affinities of SND1 and TDRKH deletion mutants toa PIWIL1 peptide (TGRme2sARARARGRARGQE)

SND1 mutant name KD, μM

SND1-deltaN No bindingSND1-deltaC 300 ± 50TDRKH (290–530) 4 ± 1TDRKH (290–430) 310 ± 50TDRKH (327–420) No binding

4 of 6 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1013106107 Liu et al.

As predicted (19), the extended Tudor domain of the DrosophilaTudor protein displays similar domain folds and conserved aro-matic cage residues for binding sDMA to the ancestral SND1polypeptide. Together, these findings underscore a previouslyunderappreciated requirement of the flanking sequences of cano-nical Tudor core domain in assisting the binding of methylargi-nine-containing peptides, a feature that is likely conservedacross germ-line Tudor proteins (19).

A role for SND1 in small RNA pathways was first suggested bythe discovery of SND1 as a component of RNA induced silencingcomplex of the siRNA pathway. Subsequently, SND1 has beenshown to bind and cleave inosine-containing hyperedited mi-cro-RNAs through its SN domains (31). However, the bindingsubstrates for SND1 Tudor domain in these physiological contextsare still unclear. In this study, we showed the coexpression andinteraction of PIWIL1/Miwi and SND1, suggesting a previouslyundescribed function for SND1 in regulating piRNA pathways. Itwill be of considerable interest to investigate whether SND1 mayaffect piRNA biogenesis, processing, or stability. A mouse knock-out model is necessary to further understand the biological role ofSND1 in these processes.

Experimental ProceduresProtein Expression and Purification. The extended Tudor domain ofhuman SND1 (residues 650–910) was cloned into a modified

pET28-MHL vector to generate N-terminal His-tagged fusionprotein. The recombinant protein was overexpressed in Escheri-chia coli BL21 (DE3)-V2R-pRARE2 at 16 °C and purified byaffinity chromatography on Ni-nitrilotriacetate resin (Qiagen)followed by TEV protease treatment to remove the tag. For crys-tallization experiments, purified protein was further treated withEndoproteinase Glu-C. All of them were further purified byanion-exchange column. Finally, the protein was concentratedto 15 mg∕mL in a buffer containing 20 mM Tris •HCl, pH7.5, 0.2 M NaCl, and 1 mM DTT.

Mutant proteins were also overexpressed in Escherichia coliBL21 (DE3)-V2R-pRARE2 and purified using the same proce-dure as described above, except further purification using size-exclusion chromatography (Hi Load 16∕60 Superdex 75, GEHealthcare Biosciences). The molecular weight of all proteinsamples was checked by mass spectrometry.

Binding Assays.All peptides used for fluorescence polarization andisothermal titration calorimetry (ITC) measurements weresynthesized by TuftsUniversity Core Services. Fluorescence polar-ization binding assay was performed in 10 μL at a constant fluor-escence labeled-peptide concentration of 40 nM and increasingamounts of SND1 (residues 650–910) at concentrations rangingfrom low to high micromolar in a buffer of 20 mM Tris •HCl,pH 7.5, 50 mM NaCl or 150 mM NaCl, 1 mM DTT, and 0.01%

R14me2s

N768

T688

Y763

F740N823

Y746Y766

SND1-R14me2s

C

T688R4me2s

N768

D742

Y763

F740

N823Y746 Y766

SND1-R4me2s

B

C

N

C

N

Rme2s

AR4me2s

R14me2s

W1495D1521

Y1523

Y1502

F1519

H4K20me2

H4R19

53BP1-H4K20me2: PDB 2IG0

E

SND1-Rme2a model

R4me2aF740

Y763

Y746

Y766

D

N768

Fig. 3. Methylarginine binding pockets in SND1. (A)Superposition of SND1-R4me2s and SND1-R14me2scomplexes. R4me2s peptide is colored in yellow,and R14me2s peptide is colored in aquamarine.Symmetrically dimethylated arginines from bothcomplexes are shown in stick models. (B) Methylargi-nine binding pocket residues and R4me2s peptide areshown in stick models. (C) Methylarginine bindingpocket residues and R14me2s peptide are shown instick models. (D) Model of asymmetrical dimethy-lated arginine bound by SND1. Note the absenceof a hydrogen bond between Rme2s and N768.(E) Methyllysine-binding pocket in the 53BP1-H4K20me2 complex.

Liu et al. PNAS Early Edition ∣ 5 of 6

BIOCH

EMISTR

Y

Tween-X-100. The assay was performed in 384-well plates, using aSynergy 2 microplate reader (BioTek). An excitation wavelengthof 485 nm and an emission wavelength of 528 nm were used. Thedata were corrected for background of the free-labeled peptides.To determine the KD values, the data were fit to a hyperbolicfunction using Sigma Plot software (Systat Software, Inc.). ITCmeasurements were performed as reported previously (32).

Tissue Immunoprecipitation and Western Blot Analysis. For tissueextract, mouse tissues and testes from different aged animalswere lysed in 150 mM NaCl/50 mM Tris •HCl, pH 7.2/1%Triton X-100/1% sodium deoxycholate/0.1% SDS using Polytrontissue homogenizer. Immunoprecipitation was performed as pre-viously described (8). Briefly, adult testes were homogenized inNP40 lysis buffer (10 mMTris •HCl, pH7.5, 150 mMNaCl, 0.5%NP40, 1 mM PMSF, 10 μg∕mL leupeptin, 10 μg∕mL aprotinin,10 μg∕mL pepstatin) using a Dounce homogenizer. After centri-fugation, the supernatant was filtered through a 0.45 μm filter,precleared with Protein G Sepharose for 2 h, and incubated with1 μg of antibodies overnight at 4 °C. The immunoprecipitateswere recovered by incubation with 80 μL 10% Protein A slurryfor 3 h. After extensive washing with the lysis buffer and elutedusing 1% SDS sample buffer for Western blotting. Antibodiesused for immunoprecipitation and Western blotting were Anti-SND1 (Abcam), Anti-SND1 (C-17, Santa Cruz), Anti-Hiwi(Abcam), and Anti-Miwi (G82, Cell Signalling).

Protein Crystallization. Purified protein treated with V8 proteasewas mixed with PIWIL1 peptides by directly adding a three- orfivefold molar excess of peptide to the protein solution. The crys-tals of SND1 appeared in a condition containing 30% PEG 1500at 18 °C overnight by the sitting drop vapor diffusion method.Prior to the data collection, the SND1 crystals were freshlysoaked in the mother liquor containing 15% glycerol and flash-frozen in liquid nitrogen.

Structure Determination.Diffraction data collected at 100 °K usingCuKα radiation generated on a Rigaku FR-E SuperBright rotat-ing anode system equipped with a Saturn A200 CCD detectorwere integrated and scaled using HKL2000 (33). All peptide-bound SND1 structures were solved by molecular replacementmethod as implemented by MOLREP in the CCP4 program suite(CCP4, 1994) using the apostructure of human SND1 extendedTudor domain (Protein Data Bank ID code 2O4X) as a searchmodel (24). Following several alternate cycles of restrained re-finement and manual rebuilding using COOT (34), the improvedmodels revealed clear electron densities allowing placement ofthe bound peptides. All refinement steps were performed usingREFMAC (CCP4, 1994) in the CCP4 program suite. During thefinal cycles of model building, TLS parameterization (35, 36)was included in the refinement of all models, which comprisedprotein and solvent molecules. Data collection and refinementstatistics are summarized in Table S2.

ACKNOWLEDGMENTS. We thank Aled Edwards and Cheryl Arrowsmith forcritical reading of the manuscript and Rick Bagshaw for reagents. Thisresearch was supported by the Structural Genomics Consortium, a registeredcharity (number 1097737) that receives funds from the Canadian Institutesfor Health Research (CIHR), the Canadian Foundation for Innovation,Genome Canada through the Ontario Genomics Institute, GlaxoSmith-Kline, Karolinska Institute, The Knut and Alice Wallenberg Foundation,the Ontario Innovation Trust, the Ontario Ministry for Research and Innova-tion, Merck & Company, Inc., the Novartis Research Foundation, the SwedishAgency for Innovation Systems, the Swedish Foundation for StrategicResearch, and the Wellcome Trust. T.P. is supported by grants from the CIHR(MOP-6849), the Canadian Cancer Society Research Institute, and the OntarioResearch Foundation. C.C. and J.J. are recipients of CIHR fellowships. K.L. issupported by a Central China Normal University scholarship. Open accesspublication costs were defrayed by the Ontario Genomics Institute GenomicsPublication Fund.

1. Bedford MT, Richard S (2005) Arginine methylation an emerging regulator of proteinfunction. Mol Cell 18:263–272.

2. Bedford MT, Clarke SG (2009) Protein arginine methylation in mammals: Who, what,and why. Mol Cell 33:1–13.

3. Malone CD, Hannon GJ (2009) Small RNAs as guardians of the genome. Cell136:656–668.

4. Thomson T, Lin H (2009) The biogenesis and function of PIWI proteins and piRNAs:Progress and prospect. Annu Rev Cell Dev Biol 25:355–376.

5. Siomi MC, Mannen T, Siomi H (2010) How does the royal family of Tudor rule the PIWI-interacting RNA pathway? Genes Dev 24:636–646.

6. Wang J, Saxe JP, Tanaka T, Chuma S, Lin H (2009) Mili interacts with tudor domain-containing protein 1 in regulating spermatogenesis. Curr Biol 19:640–644.

7. Kirino Y, et al. (2009) Arginine methylation of Piwi proteins catalysed by dPRMT5 isrequired for Ago3 and Aub stability. Nat Cell Biol 11:652–658.

8. Chen C, et al. (2009) Mouse Piwi interactome identifies binding mechanism of TdrkhTudor domain to arginine methylated Miwi. Proc Natl Acad Sci USA 106:20336–20341.

9. Vagin VV, et al. (2009) Proteomic analysis of murine Piwi proteins reveals a role forarginine methylation in specifying interaction with Tudor family members. GenesDev 23:1749–1762.

10. Reuter M, et al. (2009) Loss of the Mili-interacting Tudor domain-containing protein-1activates transposons and alters the Mili-associated small RNA profile. Nat Struct MolBiol 16:639–646.

11. Vasileva A, Tiedau D, Firooznia A, Muller-Reichert T, Jessberger R (2009) Tdrd6 isrequired for spermiogenesis, chromatoid body architecture, and regulation of miRNAexpression. Curr Biol 19:630–639.

12. Nishida KM, et al. (2009) Functional involvement of Tudor and dPRMT5 in the piRNAprocessing pathway in Drosophila germ lines. EMBO J 28:3820–3831.

13. Kirino Y, et al. (2010) Arginine methylation of Aubergine mediates Tudor binding andgerm plasm localization. RNA 16:70–78.

14. Anne J, Ollo R, Ephrussi A, Mechler BM (2007) Arginine methyltransferase Capsuleenis essential for methylation of spliceosomal Sm proteins and germ cell formation inDrosophila. Development 134:137–146.

15. Gonsalvez GB, et al. (2007) Two distinct arginine methyltransferases are required forbiogenesis of Sm-class ribonucleoproteins. J Cell Biol 178:733–740.

16. Maurer-Stroh S, et al. (2003) The Tudor domain “Royal Family”: Tudor, plant Agenet,Chromo, PWWP and MBT domains. Trends Biochem Sci 28:69–74.

17. Kim J, et al. (2006) Tudor, MBTand chromo domains gauge the degree of lysine methy-lation. EMBO Rep 7:397–403.

18. Adams-Cioaba MA, Min J (2009) Structure and function of histone methylationbinding proteins. Biochem Cell Biol 87:93–105.

19. Jin J, et al. (2009) Eukaryotic protein domains as functional units of cellular evolution.Sci Signal 2(98):1–18.

20. Botuyan MV, et al. (2006) Structural basis for the methylation state-specific recogni-tion of histone H4-K20 by 53BP1 and Crb2 in DNA repair. Cell 127:1361–1373.

21. Huang Y, Fang J, BedfordMT, Zhang Y, Xu RM (2006) Recognition of histone H3 lysine-4 methylation by the double tudor domain of JMJD2A. Science 312:748–751.

22. Cote J, Richard S (2005) Tudor domains bind symmetrical dimethylated arginines. J BiolChem 280:28476–28483.

23. Bateman A, et al. (2004) The Pfam protein families database. Nucleic Acids Res 32:D138–141.

24. Shaw N, et al. (2007) The multifunctional human p100 protein “hooks” methylatedligands. Nat Struct Mol Biol 14:779–784.

25. Deng W, Lin H (2002) Miwi, a murine homolog of piwi, encodes a cytoplasmic proteinessential for spermatogenesis. Dev Cell 2:819–830.

26. Liu H, et al. (2010) Structural basis for methylarginine-dependent recognition ofAubergine by Tudor. Genes Dev 24:1876–1881.

27. Alexandropoulos K, Cheng G, Baltimore D (1995) Proline-rich sequences that bind toSrc homology 3 domains with individual specificities. Proc Natl Acad Sci USA92:3110–3114.

28. Lim WA, Fox RO, Richards FM (1994) Stability and peptide binding affinity of anSH3 domain from the Caenorhabditis elegans signaling protein Sem-5. Protein Sci3:1261–1266.

29. Min J, et al. (2007) L3MBTL1 recognition of mono- and dimethylated histones. NatStruct Mol Biol 14:1229–1230.

30. Jozic D, et al. (2005) Cbl promotes clustering of endocytic adaptor proteins. Nat StructMol Biol 12:972–979.

31. Li CL, YangWZ, Chen YP, Yuan HS (2008) Structural and functional insights into humanTudor-SN, a key component linking RNA interference and editing. Nucleic Acids Res36:3579–3589.

32. Guo Y, et al. (2009) Methylation-state-specific recognition of histones by the MBTrepeat protein L3MBTL2. Nucleic Acids Res 37:2204–2210.

33. Otwinowski Z, Minor W (1997) Processing of X-ray diffraction data collected in oscilla-tion mode. Method Enzymol 276:307–326.

34. Emsley P, Cowtan K (2004) Coot: Model-building tools for molecular graphics. ActaCrystallogr, Sect D: Biol Crystallogr 60:2126–2132.

35. Winn MD, Isupov MN, Murshudov GN (2001) Use of TLS parameters to modelanisotropic displacements in macromolecular refinement. Acta Crystallogr, Sect D: BiolCrystallogr 57:122–133.

36. Winn MD, Murshudov GN, Papiz MZ (2003) Macromolecular TLS refinement inREFMAC at moderate resolutions. Method Enzymol 374:300–321.

6 of 6 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1013106107 Liu et al.