Characterization of a Family of Ranbp2-Type Zinc Fingers That Can Recognize Single-Stranded RNA

doi:10.1016/j.jmb.2010.12.041 J. Mol. Biol. (2011) 407, 273–283

Contents lists available at www.sciencedirect.com Journal of Molecular Biology journal homepage: http://ees.elsevier.com.jmb

Characterization of a Family of RanBP2-Type Zinc Fingers that Can Recognize Single-Stranded RNA

Cuong D. Nguyen†, Robyn E. Mansfield†, Wilfred Leung†, Paula M. Vaz, Fionna E. Loughlin, Richard P. Grant and Joel P. Mackay⁎

School of Molecular Bioscience, University of Sydney, Sydney NSW 2006, Australia

Received 24 September 2010; The recognition of single-stranded RNA (ssRNA) is an important aspect of received in revised form gene regulation, and a number of different classes of protein domains that 24 December 2010; recognize ssRNA in a sequence-specific manner have been identified. accepted 28 December 2010 Recently, we demonstrated that the RanBP2-type zinc finger (ZnF) domains Available online from the human splicing factor ZnF Ran binding domain-containing protein 20 January 2011 2 (ZRANB2) can bind to a sequence containing the consensus AGGUAA. Six other human proteins, namely, Ewing's sarcoma (EWS), translocated in Edited by M. F. Summers liposarcoma (TLS)/FUS, RNA-binding protein 56 (RBP56), RNA-binding motif 5 (RBM5), RNA-binding motif 10 (RBM10) and testis-expressed Keywords: sequence 13A (TEX13A), each contains a single ZnF with homology to the ZRANB2; ZRANB2 ZnFs, and several of these proteins have been implicated in the EWS; regulation of mRNA processing. Here, we show that all of these ZnFs are TLS/FUS; able to bind with micromolar affinities to ssRNA containing a GGU motif. RBM5; NMR titration data reveal that binding is mediated by the corresponding TEX13A surfaces on each ZnF, and we also show that sequence selectivity is largely limited to the GGU core motif and that substitution of the three flanking adenines that were selected in our original selection experiment has a minimal effect on binding affinity. These data establish a subset of RanBP2- type ZnFs as a new family of ssRNA-binding motifs. © 2011 Elsevier Ltd. All rights reserved.

Introduction

*Corresponding author. E-mail address: The processes of transcription, transport, pro- [email protected]. cessing, storage, translation and degradation of † C.D.N., R.E.M. and W.L. contributed equally to this messenger RNAs are all regulated by specific work. proteins, many of which contain discrete domains Abbreviations used: EWS, Ewing's sarcoma; GST, for the recognition of single-stranded RNA (ssRNA). glutathione S-transferase; HSQC, heteronuclear single Although less is understood about the recognition of quantum coherence; NPL4, nuclear protein localization ssRNA in comparison to double-stranded DNA, the protein 4 homologue; RBM5/10, RNA-binding motif 5/10; recent determination of a number of structures has RBP56, RNA-binding protein 56; RRM, RNA recognition begun to provide insight into several classes of motif; SELEX, Systematic Evolution of Ligands by RNA-binding domain (see Ref. 1 for a review). RNA EXponential enrichment; ssRNA, single-stranded RNA; recognition motifs (RRMs; see Ref. 2 for a review), 3 4 TEX13A, testis-expressed sequence 13A; TLS, translocated K-homology domains, Pumilio repeat domains and 5 in liposarcoma; ZnF, zinc finger; ZRANB2, ZnF Ran Tis11d-type zinc fingers (ZnFs) recognize ssRNA binding domain-containing protein 2; ZRANB2-F1, first through a variety of mechanisms. For example, ZnF of ZRANB2; ZRANB2-F2, second ZnF of ZRANB2; RRMs display a substantial amount of diversity PDB, Protein Data Bank. in their interactions with RNA; different family

0022-2836/$ - see front matter © 2011 Elsevier Ltd. All rights reserved. 274 Characterization of RanBP2-Type Zinc Fingers members recognize between four and eight bases Most of these proteins have been implicated in using a combination of side-chain- and backbone- transcriptional regulation and/or RNA processing. mediated hydrogen bonds, as well as stacking RBM5 is a putative tumor suppressor that associates interactions involving aromatic side chains. In with pre-spliceosomal complexes18 and has recently contrast, Tis11d binds to AU-rich sequences using been shown to regulate splice site pairing in almost exclusively interactions mediated by the apoptosis-related genes such as Fas.19 RBM10 shares protein backbone. ∼50% sequence identity with RBM5 and is therefore ZnF Ran binding domain-containing protein 2 likely to have a similar function. TLS/FUS was (ZRANB2) (ZNF265, Zis) is a splicing factor that discovered by virtue of its N-terminal region contains an arginine–serine-rich domain and two forming part of a fusion oncoprotein associated RanBP2-type ZnFs. It has been shown to interact with liposarcomas,12 and it has been purified as part with other components of the splicing apparatus, of a heterogeneous nuclear ribonucleoprotein com- namely, U2AF65 and U1-70K, and to alter the plex that is involved in both splicing and RNA splicing pattern of a reporter gene (Tra2β1) in transport.20 Most recently, mutations in TLS have splicing assays.6 Recently, we used Systematic been shown to be associated with the motor neuron Evolution of Ligands by EXponential enrichment disease amyotrophic lateral sclerosis.21 EWS and (SELEX) to demonstrate that each of the ZnFs is RBP56, which share a similar domain structure with capable of recognizing ssRNA that carries the TLS (Fig. 1c) and together form the TET family of consensus sequence AGGUAA.7 We also deter- proteins, both co-purify with the TFIID pre-initia- mined the three-dimensional structure of a com- tion complex and both contribute strong transacti- plex formed by the second ZnF of ZRANB2 vation domains to oncogenic fusion proteins.13,22 (ZRANB2-F2) and an oligonucleotide carrying With the exception of TEX13A, these proteins all the consensus sequence. This structure (Fig. 1a) contain RNA-recognition motifs (RRMs) and have revealed that the molecular basis for RNA been shown to bind to ssRNA sequences that are recognition was distinct from that of previously rich in Gua and/or Ura.23-27 In general, however, described RNA-binding domains. the domain(s) responsible for the binding activity In the structure, two sequential guanine bases are have not been delineated. In the case of TLS, SELEX recognized by “double-headed” hydrogen bonds data revealed a preference for GGUG sequences26 from arginine side chains (R81 and R82), and these and 15N-heteronuclear single quantum coherence bases also stack on either side of a tryptophan side (HSQC) titrations showed that the ZnF, but not the chain (W79), forming a unique Gua-Trp-Gua ladder. RRM, was responsible for this binding activity.28 The uridine forms hydrogen bonds with two Despite a growing body of data on each of these asparagine side chains (N76 and N86). It was proteins and evidence of their involvement in a notable that no base-specific interactions were variety of gene regulatory functions, their mecha- observed to any of the three adenines that flank nisms of action are generally not well defined at the the GGU sequence, despite these bases clearly being biochemical level. Here, we demonstrate that the selected in the SELEX experiment. RanBP2-type ZnFs of all of these proteins bind in a RanBP2-type ZnFs are found in organisms as stoichiometric and sequence-specific manner to diverse as fungi, plants and mammals, and analysis ssRNA containing the sequence AGGUAA, forming of the human genome reveals ∼30 other proteins complexes with well-defined structures, and that that contain these domains. In several cases, these their affinities for this sequence are comparable to domains have been identified as protein-binding those observed for ZRANB2:RNA interactions. modules. For example, RanBP2 ZnFs from the Chemical shift perturbation data further reveal nuclear pore complex proteins RanBP2 (Nup358) that the same surface is used by each domain to and Nup153 have been shown to bind RanGDP,8,9 contact RNA. Our data demonstrate that these while RanBP2 ZnF from the nuclear protein domains together constitute a distinct class of localization protein 4 homologue (NPL4) facil- ssRNA-binding domain and should assist in defin- itates binding to ubiquitin.10,11 Alignment of the ing the functions of the associated proteins. sequences of RanBP2 ZnFs reveals that a subset of the residues that are important for RNA recognition in ZRANB2-F2 is also present in at least six other Results human RanBP2 ZnFs (Fig. 1b), suggesting that these RanBP2 ZnFs might bind RNA at related sites. These proteins are translocated in liposarcoma (TLS Assessment of the sequence specificity of or FUS),12 Ewing's sarcoma (EWS),13 RNA-binding ZRANB2 motif 5 (RBM5 or LUCA-15),14 RNA-binding motif 10 (RBM10),15 RNA-binding protein 56 (RBP56, Our SELEX data for ZRANB2 clearly showed a 16 TAF15 or TAFII68) and testis-expressed sequence preference for adenines in the positions flanking 13A (TEX13A).17 the core GGU sequence.7 However, fluorescence Characterization of RanBP2-Type Zinc Fingers 275

Fig. 1. Sequence and structure of RanBP2-type ZnF proteins. (a) Three-dimensional structure of the ZRANB2-F2:RNA complex (PDB ID: 3G9Y). The ZnF is shown as a light-gray ribbon, and the RNA is shown as dark-gray sticks. Protein side chains that form hydrogen bonds with the RNA are shown in yellow, as are the side chains of W79 (which stacks between Gua2 and Gua3) and D68 (which forms a water-mediated hydrogen bond with Gua2). (b) Sequence alignment of several RanBP2-type ZnFs from human proteins. Boxes are drawn around the residues that align with the RNA- interacting residues of ZRANB2-F2, and asterisks indicate zinc-ligating cysteine residues. Sequences were aligned using Clustal W.36 GenBank accession codes are as follows: ZRANB2, CAH70058; TLS/FUS, NP_004951; EWS, CAA51489; RBP56, NP_631961; RBM5, NP_005769; RBM10, NP_005667; TEX13A, NP_112564; NUP153, NP_005115; RANBP2, CAI12246; and NPL4, NP_060391. (c) Domain architectures of human proteins containing RanBP2 ZnFs that are predicted to bind RNA. anisotropy binding data for both the first ZnF of sequence to each of the other three possible bases. ZRANB2 (ZRANB2-F1 and ZRANB2-F2 (carried The data shown in Fig. 2 indicate that no single- out using fluorescein-tagged oligonucleotides) base mutants bind to ZRANB2-F2 more tightly indicated that changing each of these adenines than the AGGUAA consensus sequence. These to cytosine had a minimal effect on the affinity of data also confirm that a significant preference the ZnF for RNA.7 In order to obtain a more exists for guanine at positions 2 and 3 and for comprehensive view of the binding specificity of uridine or guanine at position 4. In contrast, there ZRANB2, we assessed the binding of ZRANB2-F2 is very little sequence specificity for the three to a series of oligonucleotides in which we altered flanking adenine bases that were selected in our each of the first five bases in the AGGUAA SELEX experiment.7 276 Characterization of RanBP2-Type Zinc Fingers

labeled versions of EWS, RBP56, RBM5, RBM10 and TEX13A, cleaving each domain from GST and purifying it by gel-filtration chromatography. Each domain yielded a high-quality 15N-HSQC spectrum. We then carried out chemical shift perturbation experiments, recording 15N-HSQC spectra, as either a 6-nt (AGGUAA) or a 10-nt (CCAGGUAAAG) ssRNA was titrated into the protein. Figure 4 shows that each titration induced substantial chemical shift changes and that these changes were restricted to a subset of residues on each ZnF. The spectra obtained following the addition of 1–1.25 molar equivalents of RNA were in general sharp and well dispersed, consistent with the formation of well-ordered monomeric complexes. Fig. 2. RNA-binding preferences of ZRANB2-F2. Asso- ciation constants are shown for ZRANB2-F2 binding to In each case, the titration indicated that formation 17-nt ssRNA oligonucleotides containing the consensus of the complex was in intermediate to fast exchange binding sequence AGGUAA in which single-base muta- on the chemical shift timescale, consistent with the tions have been made. Each of the first five bases in the 6-nt measured binding affinities. Thus, some signals consensus has been changed to each of the other bases in simply shifted in position over the course of the turn, and the original bases are underlined. Affinities were titration, whereas others disappeared during the measured by fluorescence anisotropy titrations, and values titration. In each case, approximately the same shown are means and standard deviations from three or number of signals that had disappeared was more measurements. observed to reappear following the addition of 1 molar equivalent of RNA, and no significant changes were observed following the further addition of 1.25 The RanBP2-type ZnFs of several human molar equivalents of RNA (full sets of titration proteins can bind ssRNA spectra for each ZnF are provided in Fig. S1–5).

We next expressed the ZnF domains from EWS, The RNA interaction surface is conserved TLS, RBP56, RBM5, RBM10 and TEX13A as gluta- between each RanBP2 ZnF thione S-transferase (GST)-fusion proteins and purified the fusion proteins using glutathione- In order to compare the mode of binding of each affinity chromatography. A portion of each protein of these ZnFs with that observed previously for was cleaved from the GST, and a one-dimensional ZRANB2, we prepared uniformly 15N,13C-labeled 1H NMR spectrum was recorded in order to assess ZnFs and recorded backbone triple-resonance the folding of each domain. These spectra (not experiments to make assignments of backbone α β shown) confirmed that each domain took up a well- nuclei (Fig. 4). Assignments of HN,N,C and C ordered conformation in solution. resonances were made for all but 1–3 residues for We then used fluorescence anisotropy titrations to each ZnF alone (a small number of residues showed test the ability of each domain to bind to a 17-nt no assignable signal in the corresponding 15N- ssRNA containing a single AGGUAA motif. The HSQC most likely due to local dynamic processes) ubiquitin-binding RanBP2 ZnF from NPL4,10 which and for 32 of the expected 36 residues for the EWS lacks the conserved RNA-binding residues (Fig. 1b), ZnF:RNA complex (6 residues in intermediate was examined in parallel. Figure 3 shows that, with exchange did not reappear following the addition the exception of NPL4, all of the domains bound to of 1 molar equivalent of RNA). Assignments of HN RNA with affinities that were similar to those of the − and N resonances were also made for the majority individual ZRANB2 ZnFs (0.4–3.0×106 M 1). In of residues in the RNA-bound state for the ZnFs of contrast, no binding was observed when a 17-nt RBP56, RBM5, RBM10 and TEX13A by following oligonucleotide comprising only cytosines was used, the migration of fast-exchange signals throughout indicating that the interaction is sequence specific. the RNA titrations. From these assignments, we Furthermore, GST alone did not bind to either of the calculated weighted average chemical shift changes RNA sequences with an appreciable affinity. (for HN and N) for each residue (on the left of Fig. 5). On the right of Fig. 5, changes that are greater Each domain forms a specific, well-ordered 1:1 than 1 SD above the mean change are mapped onto complex with RNA the homology models of the EWS, RBP56, RBM5, RBM10 and TEX13A ZnFs generated using SWISS- We next used NMR spectroscopy to examine these MODEL29 and the structure of ZRANB2-F2.7 ZnF:RNA interactions. We prepared uniformly 15N- Template structures for the homology models were Characterization of RanBP2-Type Zinc Fingers 277

Fig. 3. Other RanBP2 ZnFs can bind ssRNA. (a) Fluorescence anisotropy titration data showing that each of the ZnFs from EWS, TLS, RBP56, RBM5, RBM10 and TEX13A can bind to ssRNA containing the sequence AGGUAA. The binding of NPL4 and GST is also shown. (b) Association constants derived from the data in (a). Means and standard deviations from three measurements are shown. (c) Fluo- rescence anisotropy data for a subset of RanBP2 ZnFs and GST binding to a 17-nt oligonucleotide comprising only cytosines (C17). The binding of ZRANB2-F2 to AGGUAA-containing RNA is also shown. chosen automatically by SWISS-MODEL and corre- point mutations in the RNA sequence on the sponded to ZRANB2-F1 [Protein Data Bank (PDB) binding affinity of TLS ZnF. Figure 6a shows that ID: 1N0Z] in the case of RBM5, RBM10 and TEX13A changes in either of the two guanines (Gua→Cyt and to ZRANB2-F2 in complex with RNA (PDB ID: mutations) reduce the association constant by 3- to 3G9Y) in the case of EWS and RBP56. Overall, the 4-fold, consistent with the data for ZRANB2-F1 RNA-binding surface defined by these data resem- and ZRANB2-F2. Given the preference for GGUG bles that observed for ZnFs from ZRANB2 (Fig. 5f). sequences demonstrated for TLS in published It is notable, however, that TEX13A ZnF displayed SELEX data,26 we also tested the binding of this some significant chemical shift changes on the domain to a sequence containing an AGGUGA “back” side of the domain, perhaps indicating that motif. As shown in Fig. 6a, the affinity for this this ZnF undergoes some conformational change interaction was approximately twofold lower than upon binding to RNA (Fig. 5e). that of the interaction with AGGUAA. Mutations of the flanking adenines did in two cases also The RNA-binding specificity of TLS and RBM5 result in approximately twofold reductions in affinity (A1C and A5G mutations), suggesting that In order to probe the specificity of the interaction there might be some protein contacts made by these for these proteins, we measured the effect of single- nucleotides. 278 Characterization of RanBP2-Type Zinc Fingers

120 (a) 105 N531 V515 110 R537 122 D521 N533 E549 115 C529 Q532 124 Q516 EWS 120 Q523

125 R518 126 A519 130 N541 128 10 98 7 8.8 8.6 8.4 8.2 8.0 7.8

120 K354 (b) 105 R373

N367 D386 110 122 D352 D357 115 E385 V359 124 M368 RBP56 120

125 N369 126 C365 130 N377 128 10 98 7 8.8 8.6 8.4 8.2 8.0 7.8

120 F195 (c) 105 D207 F182

E183 D210 110 K181 122 R179 115 L186 RBM5 124 E212 D184 120 N193

125 126 A206 Q213 130 N chemical shift (ppm)

15 128 10 98 7 8.8 8.6 8.4 8.2 8.0 7.8

120 (d) 105 R230 E215 I213 110 K228 E242 122 D216 115 N214 Q225 L218 K212 K210 K240 124 RBM10 120 A243 D208 V238 125 126 C233 K235 Q245 130

128 10 98 7 8.8 8.6 8.4 8.2 8.0 7.8 120 (e) 105 D381

I403 110 122 V388 D379 115 K401 Q406 124 TEX13A 120 K407 D398 H409 125 Y374 R375 126 R376

130 C396 A387 128 10 98 7 8.8 8.6 8.4 8.2 8.0 7.8 1H chemical shift (ppm)

Fig. 4 (legend on next page) Characterization of RanBP2-Type Zinc Fingers 279

Comparison of the amino acid sequences given in data further showed that although these domains Fig. 1b reveals that, in a subset of RNA-binding can interact with single-stranded DNA, the interac- RanBP2-type ZnFs (RBM5, RBM10 and TEX13A), the tion is likely to be substantially weaker than the two asparagines that contact Ura4 (N76 and N86 in interaction with ssRNA. Based on these results and ZRANB2-F2, Fig. 1a) are changed to hydrophobic our current data, we conclude that it is more likely residues. To assess whether these changes have a that the RanBP2 domains from each of the proteins significant impact on sequence specificity, we mea- that we have examined function to recognize ssRNA. sured the binding of RBM5 to oligonucleotides It is interesting to note that S439 of TLS can be containing each of the four bases at position 4 phosphorylated.31 This residue sits in the center of (AGGNAA). The data in Fig. 6b(andcomparison the RNA-binding surface of this domain, suggesting with Fig. 2) show that these amino acid substitutions that phosphorylation might act as a mechanism by significantly affect the sequence specificity of RBM5, which RNA binding is regulated in this protein. which displays a 1.5-fold preference for guanine over uracil at position 4. The preferences for guanine over Conservation of RNA-binding RanBP2-type ZnFs adenine and cytosine are 2- and 8-fold, respectively. Comparison of the RBM5 amino acid sequence BLAST searches carried out using each of the ZnFs with that of ZRANB2 also reveals that the arginine in this study revealed that the domains are present that contacts Gua2 in the latter protein (R81 in in the apparent paralogues of these proteins in a ZRANB2-F2) is substituted with lysine in RBM5. number of other organisms (Fig. S6). The RNA- Surprisingly, measurement of the affinity of RBM5 binding residues of TLS, EWS and RBP56 are for RNA sequences containing substitutions at conserved in mammals, birds, amphibians, bony position 2 (ANGUAA) shows that the sequence fish and insects, indicating an ancient and highly specificity imparted by lysine is similar to that conserved RNA-binding activity. Highly related achieved by arginine in this position (Fig. 6b and domains are also observed in organisms as diverse comparison with Fig. 2). as the parasite Schistosoma japonicum and the nematode Caenorhabditis elegans, as well as in the plants Arabidopsis thaliana and Oryza sativa. It is also Discussion notable that all of these related proteins additionally contain an RRM, one of the best-characterized ssRNA-binding domains, suggesting that these A family of RNA-binding RanBP2-type ZnFs two domain classes can combine to fine-tune target specificity. We show here that the RanBP2-type ZnFs from the Close examination of the sequences of the human proteins TLS, EWS, RBP56, RBM5, RBM10 domains in these other organisms reveals several and TEX13A are bona fide ssRNA-binding proteins changes in the putative RNA recognition residues— and that they can bind with micromolar affinity to some conservative and some relatively nonconser- sequences containing a central GGU motif. These vative—and it will be of interest to establish whether findings corroborate published biochemical data these changes subtly alter RNA sequence specificity. that point toward a nucleic acid binding capability for most of this set of proteins. Thus, full-length The design of novel RNA-binding proteins RBP56,23 RBM524 and RBM1025 were shown to bind in vitro to ssRNA containing a poly-G stretch, The recent development of ZnF proteins with whereas SELEX and NMR data revealed that TLS tailored DNA-binding specificity has resulted in a can recognize GGUG sequences.26,28 wide range of research and clinical applications (see The TET family proteins TLS, EWS and RBP56 Refs. 32–34 for reviews). Tandem repeats of three or share similar domain structures, and it has previ- more of these ZnF domains have been fused to a ously been suggested that RanBP2 ZnFs in these variety of effector domains to create new proteins proteins might function as DNA-binding domains.30 that can act, for example, as transcriptional regula- We previously showed that ZRANB2 ZnFs do not tors or site-specific nucleases. The success of these interact appreciably with double-stranded DNA proteins has been built on the properties of the containing the AGGUAA consensus motif.7 Our classical ZnF, including small size, stability to

Fig. 4. NMR analysis of the formation of RanBP2-type ZnF complexes with ssRNA. NMR analysis of the formation of RanBP2-type ZnF complexes with ssRNA 15N-HSQC spectra for the ZnFs of (a) EWS, (b) RBP56, (c) RBM5, (d) RBM10 and (e) TEX13A showing the addition of a 6- or 10-nt RNA containing the sequence AGGUAA. Spectra are shown for the proteins alone (red) and following the addition of 1.25 (or 1.0 in the case of RBM5) molar equivalents of RNA (blue). Full spectra are shown on the left. On the right, partial expanded spectra for each protein with resonance assignments are shown. Full sets of titration spectra are also provided in Fig. S1−5 for the ZnFs of EWS, RBP56, RBM5, RBM10 and TEX13A, respectively. 280 Characterization of RanBP2-Type Zinc Fingers

Fig. 5 (legend on next page) Characterization of RanBP2-Type Zinc Fingers 281

target ssRNA in the cell. The addition of effector domains to an RNA-binding domain with tailored specificity could yield designer translation factors, splicing factors or probes of RNA localization, and it will be of great interest to see whether an application of this type can be realized in the future.

Materials and Methods

Cloning, expression and purification of RanBP2 ZnF constructs

The RanBP2 ZnF domain from the human proteins TLS (residues 422–453; NM_004960), EWS (514–550; NM_013986), RBP56 (349–388; NM_139215), RBM5 (176–213; NM_005778), RBM10 (207–245; NM_005676), TEX13A (370–409; NM_031274) and NPL4 (580–608; NM_017921) was amplified from a K562 cDNA library and inserted into the plasmid pGEX-2T (GE Healthcare). The length of each construct corresponds to the struc- tured region in the second RanBP2 ZnF of ZRANB2.7 Each construct was expressed in Escherichia coli BL21 cells, as a fusion with GST and cell pellets were lysed by sonication in buffer containing 1% (v/v) Triton X-100, 3 mM phenylmethanesulfonyl fluoride and phosphate- buffered saline, pH 7.4. 15N- and 15N,13C-labeled RanBP2 ZnFs were prepared following the protocol of Cai et al.35 Proteins used for fluorescence anisotropy experiments were purified by glutathione-affinity chromatography as a fusion with GST, while proteins used for NMR experiments were cleaved from GST using thrombin and further purified by size-exclusion chromatography Fig. 6. RNA-binding specificity of TLS and RBM5. (a) (HiLoad 16/60 Superdex 30). Affinities of the TLS RanBP2-type ZnF for AGGUAA- based oligonucleotides containing single-base mutations. (b) Affinities of the RBM5 RanBP2-type ZnF for RNA AGGUAA-based oligonucleotides containing single-base mutations. Affinities were measured using fluorescence RNA oligonucleotides were purchased from Dharma- anisotropy titrations, and values shown are means and con and were deprotected prior to use according to the standard deviations from three or more measurements. instructions provided by the manufacturer. For fluorescence anisotropy experiments, 17-nt oligonucleotides that were based on the sequence GCAACCAGGUAAAGUCU mutagenesis and the fact that interactions with the were used. For NMR measurements, 6- or 10-nt sequences DNA target are mediated exclusively by side-chain were used (AGGUAA or CCAGGUAAAG). contacts, allowing a range of binding specificities to be generated on the same scaffold. RanBP2-type Fluorescence anisotropy titrations ZnFs exhibit similar properties and as such display potential as a scaffold for the design of sequence- GST-tagged RanBP2 ZnFs and RNA were dialyzed into specific RNA-binding proteins that could be used to buffer containing 10 mM Tris (pH 8), 50 mM KCl, 5 mM

Fig. 5. RNA-binding surfaces of the RanBP2-type ZnFs. Histograms showing weighted chemical shift changes (HN, N) following the addition of 1 molar equivalent of RNA containing an AGGUAA motif are provided on the left for (a) EWS, (b) RBP56, (c) RBM5, (d) RBM10 and (e) TEX13A. Horizontal dotted lines indicate 1 SD above the mean chemical shift change. Resonances shown in red were deemed to undergo significant chemical shift changes. Resonances shown in green underwent even larger changes but could not be traced during the RNA titration or did not reappear in the 15N-HSQC of a 1:1 EWS:RNA complex due to intermediate exchange. On the right, space-filling homology models created using SWISS-MODEL29 are shown in two orientations, separated by a 180° rotation. Residues that are perturbed by RNA binding are shown in red and green (color scheme as for the histograms described above). Orientation is preserved throughout (a) to (e). (f) RNA-binding surface of ZRANB2-F2 (red) α deduced from weighted average chemical shift changes between free and RNA-bound F2 for backbone (HN,H,N α and C ) and side-chain atoms.7 Orientations are the same as in (a) to (e). 282 Characterization of RanBP2-Type Zinc Fingers

MgCl2 and 1 mM DTT made with diethyl-pyrocarbonate- References treated water. All buffers and equipment were kept RNase free at all times, and RNasin (Promega) was added to the 1. Auweter, S. D., Oberstrass, F. C. & Allain, F. H. (2006). protein solution to inhibit RNase activity. Fluorescence Sequence-specific binding of single-stranded RNA: is anisotropy titrations were performed at 25 °C, with a slit there a code for recognition? Nucleic Acids Res. 34, width of 10 nm and with excitation and detection at 495 4943–4959. and 520 nm, respectively; data were averaged over 15 s. In 2. Clery, A., Blatter, M. & Allain, F. H. (2008). RNA each titration, the fluorescence anisotropy of a solution of recognition motifs: boring? Not quite. Curr. Opin. 50 nM fluorescein-tagged RNA containing 0.05 mg/mL Struct. Biol. 18, 290–298. heparin (to reduce nonspecific binding) was measured as a 3. Valverde, R., Edwards, L. & Regan, L. (2008). function of added protein concentration. Binding data Structure and function of KH domains. FEBS J. 275, were fitted to a simple 1:1 binding model by nonlinear 2712–2726. least-squares regression. Each titration was performed 4. Wang, X., McLachlan, J., Zamore, P. D. & Hall, T. M. three times, and the final affinity was taken as the mean of (2002). Modular recognition of RNA by a human these measurements. pumilio-homology domain. Cell, 110, 501–512. 5. Hudson, B. P., Martinez-Yamout, M. A., Dyson, H. J. NMR spectroscopy & Wright, P. E. (2004). Recognition of the mRNA AU- rich element by the zinc finger domain of TIS11d. Nat. Struct. Mol. Biol. 11, 257–264. For chemical shift perturbation experiments, purified 6. Adams, D. J., van der Weyden, L., Mayeda, A., 15N-labeled RanBP2 ZnFs were concentrated to 70– Stamm, S., Morris, B. J. & Rasko, J. E. (2001). 250 μM. The 15N,13C-labeled RanBP ZnFs were concen- ZNF265—a novel spliceosomal protein able to induce trated to 0.3–1.5 mM. All NMR samples contained 5% alternative splicing. J. Cell Biol. 154,25–32. D Oand20μM 2,2-dimethyl-2-silapentane-5-sulfonic 2 7. Loughlin, F. E., Mansfield, R. E., Vaz, P. M., McGrath, acid as a chemical shift reference. All experiments were A. P., Setiyaputra, S., Gamsjaeger, R. et al. (2009). The run at 298 K on a 600- or 800-MHz Bruker Avance zinc fingers of the SR-like protein ZRANB2 are single- spectrometer equipped with a cryoprobe. For titrations, stranded RNA-binding domains that recognize 5′ proteins and 10-ntRNA were dialyzed into 25 mM splice site-like sequences. Proc. Natl Acad. Sci. USA, sodium phosphate, pH 6.5, containing 100 mM NaCl 106, 5581–5586. and 1 mM DTT. The 6-nt RNA was buffer exchanged into 8. Higa, M. M., Alam, S. L., Sundquist, W. I. & Ullman, water using a mini Quick Spin Oligo Column (Roche), K. S. (2007). Molecular characterization of the ran- lyophilized and resuspended in the same buffer as the binding zinc finger domain of NUP153. J. Biol. Chem, protein. 15N-HSQC spectra were recorded for the ZnF 282, 17090–17100. alone and following the addition of typically 0.25, 0.5, 1 9. Yaseen, N. R. & Blobel, G. (1999). Two distinct classes and 1.25 molar equivalents of RNA (exact titration points of Ran-binding sites on the nucleoporin Nup-358. for each ZnF are specified in Figs. S1−5). In order to assign Proc. Natl Acad. Sci. USA, 96, 5516–5521. backbone resonances of the RanBP2 ZnFs, we recorded 10. Alam, S. L., Sun, J., Payne, M., Welch, B. D., Blake, B. HNCACB and CBCA(CO)NH experiments. For the EWS K., Davis, D. R. et al. (2004). Ubiquitin interactions of ZnF, these spectra were also recorded in the presence of NZF zinc fingers. EMBO J. 23, 1411–1421. the 6-nt RNA. 11. Wang, B., Alam, S. L., Meyer, H. H., Payne, M., Stemmler, T. L., Davis, D. R. & Sundquist, W. I. Accession numbers (2003). Structure and ubiquitin interactions of the conserved zinc finger domain of Npl4. J. Biol. Chem. α β 278 – Chemical shift assignments for HN,N,C and C nuclei , 20225 20234. of EWS, RBP56, RBM5, RBM10 and TEX13A ZnFs have 12. Crozat, A., Aman, P., Mandahl, N. & Ron, D. been deposited in the BioMagResBank under accession (1993). Fusion of CHOP to a novel RNA-binding protein in human myxoid liposarcoma. Nature, 363, IDs 17133 (EWS), 17380 (RBP56), 17387 (RBM5), 17386 – (RBM10) and 17385 (TEX13A). 640 644. Supplementary materials related to this article can be 13. Delattre, O., Zucman, J., Plougastel, B., Desmaze, C., found online at doi:10.1016/j.jmb.2010.12.041 Melot, T. et al. (1992). Gene fusion with an ETS DNA- binding domain caused by chromosome translocation in human tumours. Nature, 359, 162–165. 14. Gure, A. O., Altorki, N. K., Stockert, E., Scanlan, M. J., Old, L. J. & Chen, Y. T. (1998). Human lung cancer antigens recognized by autologous antibodies: defini- Acknowledgements tion of a novel cDNA derived from the tumor suppressor gene locus on chromosome 3p21.3. Cancer 58 – We thank Ann Kwan and Bill Bubb for expert Res. , 1034 1041. 15. Nagase, T., Seki, N., Tanaka, A., Ishikawa, K. & maintenance of the NMR facility. R.E.M., P.M.V. Nomura, N. (1995). Prediction of the coding sequences and F.E.L. were supported by Australian Postgrad- of unidentified human genes. IV. The coding uate Awards. This work was supported in part by a sequences of 40 new genes (KIAA0121-KIAA0160) Project Grant from the National Health and Medical deduced by analysis of cDNA clones from human cell Research Council of Australia (to J.P.M.). line KG-1. DNA Res. 2, 167–174. Characterization of RanBP2-Type Zinc Fingers 283

16. Morohoshi, F., Arai, K., Takahashi, E. I., Tanigami, A. of an RNA binding specificity for the potential & Ohki, M. (1996). Cloning and mapping of a human splicing factor TLS. J. Biol. Chem. 276, 6807–6816. RBP56 gene encoding a putative RNA binding protein 27. Ohno, T., Ouchida, M., Lee, L., Gatalica, Z., Rao, V. N. similar to FUS/TLS and EWS proteins. Genomics, 38, & Reddy, E. S. (1994). The EWS gene, involved in 51–57. Ewing family of tumors, malignant melanoma of soft 17. Wang, P. J., McCarrey, J. R., Yang, F. & Page, D. C. parts and desmoplastic small round cell tumors, codes (2001). An abundance of X-linked genes expressed in for an RNA binding protein with novel regulatory spermatogonia. Nat. Genet. 27, 422–426. domains. Oncogene, 9, 3087–3097. 18. Hartmuth, K., Urlaub, H., Vornlocher, H. P., Will, C. L., 28. Iko, Y., Kodama, T. S., Kasai, N., Oyama, T., Morita, Gentzel, M., Wilm, M. & Luhrmann, R. (2002). Protein E. H., Muto, T. et al. (2004). Domain architectures composition of human prespliceosomes isolated by a and characterization of an RNA-binding protein, tobramycin affinity-selection method. Proc. Natl Acad. TLS. J. Biol. Chem. 279, 44834–44840. Sci. USA, 99, 16719–16724. 29. Arnold, K., Bordoli, L., Kopp, J. & Schwede, T. (2006). 19. Bonnal, S., Martinez, C., Forch, P., Bachi, A., Wilm, M. The SWISS-MODEL workspace: a web-based envi- & Valcarcel, J. (2008). RBM5/Luca-15/H37 regulates ronment for protein structure homology modelling. Fas alternative splice site pairing after exon definition. Bioinformatics, 22, 195–201. Mol. Cell, 32,81–95. 30. Tan, A. Y. & Manley, J. L. (2009). The TET family of 20. Calvio, C., Neubauer, G., Mann, M. & Lamond, A. I. proteins: functions and roles in disease. J. Mol. Cell. (1995). Identification of hnRNP P2 as TLS/FUS using Biol. 1,82–92. electrospray mass spectrometry. RNA, 1, 724–733. 31. Guipaud, O., Guillonneau, F., Labas, V., Praseuth, D., 21. Kwiatkowski, T. J., Jr, Bosco, D. A., Leclerc, A. L., Rossier, J., Lopez, B. & Bertrand, P. (2006). An in vitro Tamrazian, E., Vanderburg, C. R., Russ, C. et al. (2009). enzymatic assay coupled to proteomics analysis Mutations in the FUS/TLS gene on chromosome 16 reveals a new DNA processing activity for Ewing cause familial amyotrophic lateral sclerosis. Science, sarcoma and TAF(II)68 proteins. Proteomics, 6, 323, 1205–1208. 5962–5972. 22. Sjogren, H., Meis-Kindblom, J., Kindblom, L. G., 32. Camenisch, T. D., Brilliant, M. H. & Segal, D. J. (2008). Aman, P. & Stenman, G. (1999). Fusion of the EWS- Critical parameters for genome editing using zinc related gene TAF2N to TEC in extraskeletal myxoid finger nucleases. Mini Rev. Med. Chem. 8, 669–676. chondrosarcoma. Cancer Res. 59, 5064–5067. 33. Carroll, D. (2008). Progress and prospects: zinc-finger 23. Bertolotti, A., Lutz, Y., Heard, D. J., Chambon, P. & nucleases as gene therapy agents. Gene Ther. 15, Tora, L. (1996). hTAF(II)68, a novel RNA/ssDNA- 1463–1468. binding protein with homology to the pro-oncoproteins 34. Sera, T. (2009). Zinc-finger-based artificial transcrip- TLS/FUS and EWS is associated with both TFIID and tion factors and their applications. Adv. Drug Delivery RNA polymerase II. EMBO J. 15, 5022–5031. Rev. 61, 513–526. 24. Drabkin, H. A., West, J. D., Hotfilder, M., Heng, 35. Cai, M., Huang, Y., Sakaguchi, K., Clore, G. M., Y. M., Erickson, P., Calvo, R. et al. (1999). DEF-3 Gronenborn, A. M. & Craigie, R. (1998). An efficient (g16/NY-LU-12), an RNA binding protein from and cost-effective isotope labeling protocol for pro- the 3p21.3 homozygous deletion region in SCLC. teins expressed in Escherichia coli. J. Biomol. NMR, 11, Oncogene, 18, 2589–2597. 97–102. 25. Inoue, A., Takahashi, K. P., Kimura, M., Watanabe, T. 36. Thompson, J. D., Higgins, D. G. & Gibson, T. J. (1994). & Morisawa, S. (1996). Molecular cloning of a RNA CLUSTAL W: improving the sensitivity of progressive binding protein, S1-1. Nucleic Acids Res. 24, 2990–2997. multiple sequence alignment through sequence 26. Lerga, A., Hallier, M., Delva, L., Orvain, C., Gallais, I., weighting, position-specific gap penalties and weight Marie, J. & Moreau-Gachelin, F. (2001). Identification matrix choice. Nucleic Acids Res. 22, 4673–4680.