a r t i c l e s

Structure of N-terminal domain of ZAP indicates how a zinc-finger recognizes complex RNA

Shoudeng Chen1,2,4, Yihui Xu3,4, Kuo Zhang1,2, Xinlu Wang3, Jian Sun1,2, Guangxia Gao3 & Yingfang Liu1

Zinc-finger antiviral protein (ZAP) is a host factor that specifically inhibits the replication of certain viruses, such as HIV-1, by targeting viral mRNA for degradation. How ZAP recognizes its target RNA has been unclear. Here we report the crystal structure of the N-terminal domain of rat ZAP (NZAP225), the major functional domain. The overall structure of NZAP225 resembles a tractor, with four zinc-finger motifs located at the bottom. Structural and functional analyses identified multiple positively charged residues and two putative RNA-binding cavities forming a large putative RNA-binding cleft. ZAP molecules interact to form a dimer that binds to a ZAP-responsive RNA molecule containing two ZAP-binding modules. These results provide insights into how ZAP binds specifically to complex target RNA.

ZAP is a host factor that specifically inhibits the replication of certain organized to form a previously undescribed structure but also viruses, including HIV-1 (ref. 1), Ebola virus2 and Sindbis virus3. provide insights into how ZAP functions. ZAP does not induce a universal antiviral state, as some viruses grow normally in ZAP-expressing cells3. ZAP binds directly to specific viral RESULTS mRNA sequences4, recruits cellular poly(A) RNase (PARN) to shorten Overall structure of NZAP225 the poly(A) tail and recruits a 3′–5′ exoRNase complex, the RNA exo- Bacterially expressed NZAP225 was purified to homogeneity. Crystals some, to degrade target mRNAs from the 3′ end1. ZAP also recruits the diffracting to a resolution of 1.8 Å were obtained, and the structure of decapping complex through its cofactor RNA helicase p72 to initi- NZAP225 was determined by single-wavelength anomalous disper- ate 5′–3′ degradation of target mRNAs1,5,6. No obvious common ele- sion using the bound zinc atoms within the protein. The structure was ments or motifs have been identified in ZAP target ; their only refined to a final Rwork 19.1% and Rfree 22.8%. The final experimental common feature is that they are more than 500 nucleotides long. electron density map shows that all residues except residues 55–60 © 2012 Nature America, Inc. All rights reserved. America, Inc. © 2012 Nature There are four tandem CCCH-type zinc-finger motifs in the were traced well (Fig. 1a). Data collection, phasing and refinement N-terminal domain of ZAP7. A fusion of the N-terminal 254 amino statistics are presented in Table 1. acids of ZAP with the product that confers zeocin resistance The structure of NZAP225 consists of seven α-helices, three npg (NZAP-Zeo) has antiviral activity comparable to that of the full- β-strands, three short 310α-helices and several loops. This structure length protein7. This suggests that the N-terminal domain is the major resembles a tractor that can be divided into three layers: a ‘cockpit’ functional domain that binds target RNA and recruits the mRNA layer on top, ‘wheel guards’ in the middle and four ‘wheels’ at the degradation machinery. ZAP has recently been reported to interact bottom. The top cockpit layer consists of the first 65 residues and has with retinoic acid–inducible gene-I (RIG-I) and thereby enhance an αβ architecture with a twisted plane formed by three short antipar- RIG-I–mediated interferon production by facilitating RIG-I oligo­ allel β-strands surrounded on top by three α-helices (Fig. 1b). The merization8. Notably, the RIG-I interacting region was also mapped four predicted zinc-finger motifs (ZF) are located at the four corners to the N-terminal domain of ZAP. of a twisted rectangle in the bottom layer (Fig. 1b). Each zinc finger The features of ZAP-responsive RNAs and the mechanism coordinates one zinc ion at the center via three cysteines and one by which ZAP specifically recognizes these RNAs have been histidine (Fig. 1b). The overall folds of the zinc binding domains and elusive. Furthermore, considering the relatively small size of the their sequences are different from each other (Fig. 1c). The top cock- N-terminal domain of ZAP, how ZAP binds a complex target pit layer and the bottom zinc-finger motifs are linked by four helices, RNA and recruits multiple components of the RNA degradation which form the middle layer. Each of these four helices interacts with machinery has been puzzling. To address these questions, we one zinc-finger motif: helix 4 with ZF1, helix 5 with ZF2, helix 6 determined the crystal structure of the N-terminal domain of rat with ZF3 and helix 7 with ZF4, and it is likely that the interactions (Rattus norvegicus) ZAP covering residues 1–225 (termed NZAP225). strengthen the loose structure of the zinc-finger motifs. A detailed Our results not only show how the four zinc-finger motifs are description of the interactions between the cockpit, wheel guards

1State Key Laboratory of Biomacromolecules, Institute of Biophysics, Chinese Academy of Sciences, Beijing, China. 2Graduate School of the Chinese Academy of Sciences, Beijing, China. 3Key Laboratory of Infection and Immunity, Institute of Biophysics, Chinese Academy of Sciences, Beijing, China. 4These authors contributed equally to this work. Correspondence should be addressed to Y.L. ([email protected]) or G.G. ([email protected]). Received 14 November 2011; accepted 3 January 2012; published online 11 March 2012; doi:10.1038/nsmb.2243

430 VOLUME 19 NUMBER 4 APRIL 2012 nature structural & molecular biology a r t i c l e s

Figure 1 Overall structure of NZAP225. a b N (a) Ribbon diagram of NZAP225 structure. The �2 �3 cysteine and histidine side chains coordinated �2 to zinc atoms (orange) are shown in a ball-and- β3 Cockpit �1 �3 stick representation. The α-helices and β-sheets �1 β3 β2 β2 are numbered, and the character ‘η’ represents β1 β1 the 310-helix. (b) Topology diagram of the NZAP225 structure. (c) Tertiary structures and �5 Wheel sequence alignment of the zinc-finger motifs. �7 guards �4 �5 �6 �7 c �4 (d) Surface electrostatic distribution generated ZF4 η1 by PyMOL. Positively charged residues are in C H C η1 C C η3 C H C blue, negatively charged residues in red and �6 η3 Wheels η2 Zn Zn η2 Zn Zn neutral residues in white. C C ZF1 C C H C H C ZF2 ZF1 ZF2 ZF3 ZF4 ZF3 infected with a VSVG-pseudotyped MLV-luc vector or transfected with pGL3-Luc-Na, a c Cys96 reporter containing a ZAP-sensitive fragment Cys162 4 Cys174 derived from the Sindbis virus (SINV) , and Cys78 Cys73 Cys182 His110 Cys168 antiviral activity was recorded in terms of fold inhibition. Although the showed His86 Cys108 His191 Cys88 Cys150 different antiviral activities against the two Cys82 His172 Cys187 reporters, presumably resulting from the dif-

ZF1 ZF2 ZF3 ZF4 ferent experimental systems used, the relative activities of the mutant proteins compared ZF1 ZF2 with wild-type ZAP were similar for both ZF3 reporters. Double- or triple-residue mutants ZF4 in the first, second and fourth zinc fingers (R74A R75A K76A, K89A R95A K107A and d R179A R189A, respectively) showed consid- erably reduced activity (Fig. 2c). In contrast, 90° 90° the antiviral activity of the triple mutant in the third zinc finger (R158A K159A Q160A) was only minimally reduced (Fig. 2c). The activities of single-residue mutants were either only marginally (R75A and K151A) Bottom view Side view Top view or moderately affected (R74A, K76A, R103A and R170A) (Fig. 2c). © 2012 Nature America, Inc. All rights reserved. America, Inc. © 2012 Nature and zinc-finger motifs can be found in Supplementary Figure 1. The ability of mutant proteins to bind target RNAs was analyzed These results suggest that this structure is rigid and is not likely to with an in vitro RNA binding assay4. SINV-derived RNA fragment Na undergo significant conformational change. Dali search results reveal was used for the binding assay, and RNA fragment Di, which fails to

npg that the overall folding of NZAP is unlike any that have previously bind to ZAP, was used as a negative control for nonspecific binding. been described (data not shown). The specific RNA-binding ability of the multi-residue mutants (K89A R95A K107A and R179A R189A) was considerably reduced, whereas Roles of positively charged residues in the zinc-finger regions that of single-residue mutants (R74A, K76A, R103A and R170A) was On the bottom surface of the structure formed by the four zinc- little affected (Fig. 2d,e). Reductions in the in vivo antiviral activity finger motifs, 78% (22 of 28) of the basic residues are conserved of the mutants were more dramatic than those in in vitro RNA bind- (Supplementary Fig. 2). Some of these residues contribute to the ing activity, suggesting that, although these mutations did not have formation of a positively charged groove located between the front a major effect on ZAP RNA binding activity in our in vitro system, and the rear wheels (Fig. 1d). In ZF1, residues Arg74, Arg75 and they did strongly influence in vivo antiviral activity. This discrepancy Lys76 are located in the loop region between its first two cysteine between the antiviral activities and RNA-binding abilities of R74A residues. In ZF2, residues Lys89, Arg95, Arg103 and Lys107 extend and K76A may be accounted for by the possibility that the loss of these out from the bottom surface. Residue Arg103 clearly forms a bulge at positively charged residues is compensated for by adjacent positively a corner of the rectangle formed by the tractor wheels. In ZF3, similar charged residues. The discrepancy between the antiviral activity and to ZF1, residues Lys151, Arg158, Lys159 and Arg170 are located in the RNA-binding ability of R103A suggests that Arg103 may be involved in loop region between its first two cysteines. In ZF4, residues Arg179 ZAP’s interaction with protein factors rather than in RNA binding. and Arg189 are located in the short α-helix and loop region, respec- tively (Fig. 2a,b). We speculated that some of these residues might be Structural alignment suggests two putative RNA binding cavities involved in RNA binding and thereby the antiviral function of ZAP. A search in the for CCCH-type zinc-finger proteins To evaluate their contributions to the antiviral activity of ZAP, all identified Muscleblind-like 1 (MBNL1) as an RNA-binding protein that the above residues, either singly or in combinations, were replaced contains four CCCH-type zinc-finger motifs (PDB: 3D2Q9 and 3D2S9). with alanine in the NZAP-Zeo backbone, and mutant proteins were MBNL1 is a splicing factor that has a key role in the development expressed at comparable levels in Rat2 cells (Fig. 2c). Cells were either of myotonic dystrophy (DM) disease. Sequestering of MBNL1 to

nature structural & molecular biology VOLUME 19 NUMBER 4 APRIL 2012 431 a r t i c l e s

Table 1 Data collection, phasing and refinement statistics similarities in structure-based sequence alignment between ZAP NZAP225 ZF2 and MBNL1 ZF3 (Fig. 3a). Superposition of ZAP ZF2 (Cys88– Data collection His110) and the MBNL1 ZF3 (Cys185–His204)-RNA complex results Space group P3(2)21 in a pairwise r.m.s. deviation (all protein atoms) of 1.5 Å, suggesting Cell dimensions that ZAP ZF2 and MBNL1 ZF3 adopt similar folds. The orientation a, b, c (Å)a a = b = 52.894, c = 138.275 of Tyr108 in ZAP is almost completely the same as that of Phe202 α, β, γ (°) α = β = 90, γ = 120 in the MBNL1–RNA complex. The location of Lys107 in ZAP ZF2 Wavelength (Å) 1.25 is very similar to that of Arg201 in MBNL1 ZF3, although the side Resolution (Å) 50 (1.83) – 1.8 chain orientations are slightly different (Fig. 3a). Furthermore, protein Rsym or Rmerge 6.5 (41.3) vacuum electrostatic analysis shows that Tyr108, Val72 and Phe144 I / σI 44.05 (6.55) form a cavity, named cavity 1. Likewise, ZAP ZF4 shows similari- Completeness (%) 99.9 (100) ties to MBNL1 ZF4 (Fig. 3b). Superposition of ZAP ZF4 (Cys174– Redundancy 20.3 (19.7) His191) and the MBNL1 ZF4 (Cys221–His238)-RNA complex results Refinement in a pairwise r.m.s. deviation (all protein atoms) of 0.6 Å (Fig. 3b). Resolution (Å) 32.5–1.8 Residues His176, Tyr184 and Arg189 in NZAP225 form another cav- No. reflections 21,522 ity, named cavity 2. The side chain orientations of Y184 and R189 in Rwork / Rfree 19.1 / 22.8 NZAP225 are very similar to those of Arg231 and Tyr236 in MBNL1, No. atoms respectively. Alignment of ZAP proteins from different species reveals Protein 213 Ligand/ion 4 that Val72, Tyr108, Phe144, His176, Tyr184 and Arg189 are con- Water 200 served (Supplementary Fig. 2). On the basis of these observations, B-factors we speculated that these residues might be important for RNA binding Protein 33.33 and, thus, the antiviral activity of ZAP. Ligand/ion 30.01 To probe the roles of these residues in the antiviral activity of Water 39.45 ZAP, each one was replaced with alanine in NZAP-Zeo. The mutant R.m.s. deviations proteins were assayed for their activity against MLV-luc and pGL3- Bond lengths (Å) 0.006 Luc-Na. Mutants V72A, Y108A and F144A all showed considerably Bond angles (°) 0.907 reduced activity (Fig. 3c), whereas the activity of mutant H176A was a Values in parentheses are for highest-resolution shell. little affected and the activities of Y184A and R189A were only mod- erately affected (Fig. 3c). Consistent with these effects on activity, the RNAs containing defective CUG or CCUG repeats in DM results binding ability of V72A, Y108A and F144A mutants was significantly in splicing defects. MBNL1 also contains four CCCH-type zinc- reduced (Fig. 3d). In contrast, substitution of His176 and Tyr184 finger motifs, but these are located in two regions separated by a long with alanine had little effect, and substitution of Arg189 with alanine segment of 110 residues. Structures of MBNL1 fragments covering only moderately affected RNA binding. These results indicate that the the first two zinc-finger motifs (ZF1 and ZF2), and covering the last contributions of the two cavities to the protein’s RNA binding ability two zinc-finger motifs alone or in complex with a hexanucleotide and antiviral activity are different. These residues are exposed on the (CGCUGU), have been reported9. Structural comparison of NZAP225 bottom surface of the wheel layer, and thus alanine substitution is not © 2012 Nature America, Inc. All rights reserved. America, Inc. © 2012 Nature and the MBNL1 ZF3/4–r(CGCUGU) complex reveals notable expected to significantly affect the overall structure. This notion is

a b Arg75 c 100 MLV 90 80 npg 70 ZF1 ZF4 60 Arg74 Arg179 50 Arg189 Arg158 40 30 90° Lys76 20 Fold inhibition 10 Lys151 0 Lys107 Lys89 7 SIN-Na Arg95 6 Arg179 5 Lys89 Arg189 ZF3 Lys159 4 Arg170 ZF2 Arg170 3 Arg74 Lys159 2 Arg103 Arg95 1 K107 Arg75 Arg158 Fold inhibition 0 Lys151 NZAP-Zeo K76 Arg103 Bottom view Side view β-actin Ctrl WT R74A K89A R158A R179A R74A R75A K76A R103A K151A R170A R75A R95A K159A R189A R74A K76A K107A Q160A K89A R75A R95A R179A d Input Ctrl NZAP R74A R75A K76A K76A e 1/10 Input Ctrl NZAP K107A R103A R170A R189A DiNa Di Na Di Na Di Na Di Na Di Na Di Na Di Na Di Na Di Na Di Na Di Na Di Na Di Na

NZAP-Zeo NZAP-Zeo

Figure 2 Roles of positively charged residues in RNA binding and antiviral activity. (a) Overall distribution of the positively charged residues. Key residues predicted to be involved in RNA binding are shown as stick representations (blue). (b) Bottom view of a. (c) Antiviral activity of NZAP-Zeo-myc mutants against MLV-luc (upper panel) or pGL3-luc-Na (SIN-Na, second panel). Fold inhibition data are means ± s.d. of three independent experiments. Ctrl, Rat2 cells; WT, Rat2 stable cell lines that can inducible express wild-type NZAP-Zeo. (d,e) RNA-binding activity of the mutants. Ctrl, Rat2 cells transduced with empty vector; NZAP, cells expressing wild-type NZAP-Zeo; Di, negative control RNA; Na, RNA segments that bind to NZAP; 1/10 Input, 1/10 of the RNA used for binding.

432 VOLUME 19 NUMBER 4 APRIL 2012 nature structural & molecular biology a r t i c l e s

a b c 100 MLV 80 ZAP_ZF2 ZAP_ZF4 60 90 100 110 180 190 40 ZAP_ZF2 ZAP_ZF4 MBNL1_ZF3 MBNL1_ZF4 20 Fold inhibition 190 200 230 0 MBNL1_ZF3 MBNL1_ZF4 7 6 SIN-Na * * * * 5 4 Tyr184 3 Arg231 2

Fold inhibition 1 0

Phe202 NZAP-Zeo Arg189 -actin Tyr236 β Tyr108 Ctrl WT V72A Y108A F144A H176A Y184A R189A Val72 His176 d 1/10 Input Ctrl NZAP V72A Y108A F144A H176A Y184A R189A Asp223 DiNa Di Na Di Na Di Na Di Na Di Na Di Na Di Na Di Na Phe144

Figure 3 Putative RNA-binding cavities. (a) Structure-based sequence NZAP-Zeo alignment of ZAP ZF2 and MBNL1 ZF3 (PDB: 3D2S). Upper: residues predicted to form hydrogen bonds with RNA bases via their backbone in MBNL1 ZF3 are indicated by black asterisks, and those involved in base- stacking with RNA bases are indicated by black triangles. Residue Tyr108 of ZAP is indicated by a green dot. Lower: structure alignment of ZAP ZF2 (gray) and MBNL1 ZF3 (cyan) generated by PyMOL. (b) Structure-based sequence alignment of ZAP ZF4 and MBNL1 ZF4 (PDB: 3D2S). Upper: sequence alignment of the two zinc-finger motifs labeled as in a. Residues His176, Tyr184 and Arg189 of ZAP are indicated by green dots. Lower: structure alignment of ZAP ZF4 (gray) and MBNL1 ZF4 (magenta) generated by PyMOL. Residues important for RNA binding in MBNL1 and NZAP are shown in stick representation. (c) Antiviral activities of the indicated mutants against MLV-luc (upper panel) and pGL3-Na-luc (SIN-Na, second panel), as described in Figure 2. Fold inhibition data are means ± s.d. of three independent experiments. (d) RNA binding activities of the indicated mutants, as described in Figure 2.

supported by the observation that the Y108 mutant and the wild-type at the interface, involving four salt bridges and 13 hydrogen-bond protein have similar secondary structure compositions, as measured contacts between 28 residues. The intermolecular distances between by circular dichroism (CD) spectrometry (Supplementary Fig. 3). Cys7 and Phe9, Arg31 and Glu133, and Asp3 and Thr43 (main chain interactions) are all less than 3 Å (Fig. 4a), suggesting that these N-terminal tail–mediated intermolecular interactions residues may be important for intermolecular interactions. Recombinant NZAP225 exists as a monomer in solution (data not Furthermore, the structure reveals that the N-terminal 1–9 resi- shown). However, we observed an obvious interaction surface in the dues fit well into a surface groove of the interacting molecule upper layer between adjacent molecules in crystals. Analysis using the (Supplementary Fig. 4). In vitro EGS-mediated cross-linking data PISA program in EBI (http://www.ebi.ac.uk/pdbe/prot_int/pistart. further suggest that the molecules interact and that the N-terminal © 2012 Nature America, Inc. All rights reserved. America, Inc. © 2012 Nature html) showed that interacting molecules share a surface of 979 Å2 tail is required for the interaction (Fig. 4b).

T43 Input IP: Anti- ag npg a D3 b c Molecule A 50-kDa NZAP-myc – WT WT F9A E133A D1-9 – WT WT F9A E133A D1-9 dimer NZAP- ag WT – WT F9A E133A D1-9 WT – WT F9A E133A D1-9 C7 F9 25-kDa Anti-myc R31 monomer E133 F9 C7 Anti- ag T43 E133 WT WTD1-9 R31 EGS (0.2 mM) Molecule B D3 Model: one site d e 2 7 100 ) � /DoF = 1.158 × 10

80 –1 300 N = 2.02 ± 0.0140 Sites 60 6 5 Figure 4 Intermolecular interactions of ZAP molecules. 40 K = 2.34 × 10 ± 1.0 × 10 per mol 5 20 200 ∆ H = 3.81 × 10 ± 3,859 cal per mol (a) Ribbon representation of two ZAP molecules. Molecules A 3 Fold inhibition 0 S = 1.33 10 cal per mol per deg (green) and B (pink) interact with each other via residues Asp3, ∆ × 100 Cys7, Phe9, Arg31, Thr43 and Glu133. The major interactions NZAP-Zeo -actin Injectant (Kcal mol between two adjacent molecules are shown by dashed lines. β 0 (b) In vitro EGS cross-linking assay. WT, wild-type NZAP225; 0 1 2 3 4 5 Ctrl WT F9A D1-9 Molar ratio D1-9, NZAP225 mutant with residues 1–9 deleted. E133A (c) Co-immunoprecipitation of NZAP254 proteins. IP, f immunoprecipitation assays; WT, wild-type NZAP254; F9A, NZAP254-F9A; E133A, Input Ctrl NZAP-Zeo Y108A-Zeo Di Na Nf Nc Di Na Nf Nc Di Na Nf Nc Di Na Nf Nc NZAP254-E133A; D1-9, NZAP254-D1-9. (d) Antiviral activity of NZAP-Zeo-myc mutants against MLV-luc as described in the legend to Figure 2. Fold inhibition data are means ± s.d. of three independent experiments. (e) Characterization of NZAP-Zeo binding to Na RNA using isothermal titration calorimetry. (f) RNA binding activities of the indicated mutants, as described in the legend to Figure 2. Nc, fragment 6062–6298 of SINV genomic RNA lgG– fragment Na; Nf, fragment 6298–6580 of SINV genomic RNA fragment Na4. NZAP-Zeo–

nature structural & molecular biology VOLUME 19 NUMBER 4 APRIL 2012 433 a r t i c l e s

ZF4 has been shown to be the minimum sequence a ZF4 b Lys76 Arg74 4 His176 Tyr184 that confers sensitivity to ZAP . When Na is Arg75 Arg179 ZF1 split into two fragments, Nf and Nc, neither is ZF1 Arg189 Arg158 4 Lys151 functional . Our in vitro RNA binding assay

Val72 shows that both Nf and Nc bound to ZAP Tyr108 Arg170 (Fig. 4f). However, neither Nf nor Nc bound Lys159 Phe144 Lys89 to the Y108A mutant, suggesting that binding Lys107 of Nf or Nc to ZAP is specific. Taken together, Arg95 these results support the notion that one ZAP ZF3 ZF3 dimer binds to one RNA molecule. Arg103 ZF2 ZF2

Figure 5 Model of ZAP-RNA interaction. (a) Residues involved in RNA binding. The positively DISCUSSION charged residues important for ZAP’s function are shown in blue and those less important in The mechanism by which ZAP recognizes light blue. The residues in cavity 1 and cavity 2 are in green and pink, respectively. (b) Cartoon its target RNA is largely unknown. Here, representation of RNAs bound to ZAP based on structural alignment with MBNL1. we determined the crystal structure of the N-terminal domain of ZAP, which is sufficient To analyze whether ZAP molecules interact in vivo, three mutations to bind specifically to target RNA and promote RNA degradation. Our were introduced into NZAP. Phe9 was replaced by alanine (F9A) to results show that this domain possesses a large RNA binding surface disrupt the intermolecular interactions between Phe9 and Cys7, and involving many positively charged residues and some residues from Glu133 was replaced with alanine (E133A) to disrupt the intermolecular two cavities in its zinc-finger motifs. Substitution of some of these interactions between Glu133 and Arg31. Considering that the intermo- residues dramatically reduces ZAP’s antiviral activity, whereas substi- lecular interactions are mainly mediated by the N-terminal 1–9 residues tution of others has only a minimal or moderate effect. Interestingly, and that a single mutation of any residue may thus not have a dramatic those residues whose substitution with alanine dramatically affects effect, we constructed a mutant in which residues 1–9 were deleted ZAP’s activity are located along a cleft (Fig. 5a, residues in blue). We (D1-9). Immunoprecipitation of Flag-tagged NZAP co-precipitated built a ZAP-RNA binding model by superimposing ZF3 and ZF4 from myc-tagged NZAP, suggesting that wild-type NZAP molecules also the MBNL1 complex with its target RNA respectively to ZF2 and ZF4 interact in vivo. In contrast, interactions of NZAP-F9A were moderately in NZAP structure (Fig. 5b). In this model, three RNA fragments reduced, those of NZAP-E133A were more strongly reduced, and those derived from the MBNL1 ZF3/4–RNA complex are docked onto the of NZAP-D1-9 were abolished (Fig. 4c). These results confirmed that proposed ZAP RNA binding surface: two RNAs are docked on the ZAP molecules interact in vivo through the N-terminal tail. two long boundaries of the rectangular-like surface and one RNA in To investigate the role of intermolecular interactions in ZAP’s anti- the groove of the surface formed between ZF1/2 and ZF3/4. These viral activity, mutations were introduced into NZAP-Zeo. Mutant three RNAs fit well into the surface cleft constituted by the residues proteins were expressed in Rat2 cells and assayed for their antiviral identified to be important for RNA binding. We propose that the activity against MLV-luc. Expression levels of the mutant proteins were positively charged residues are involved in interactions of the protein comparable (Fig. 4d, middle panel). Compared with wild-type NZAP- with phosphate groups in the RNA backbone and that the residues Zeo, the antiviral activity of the F9A mutant was reduced by about in the cavities are involved in interactions of the protein with base © 2012 Nature America, Inc. All rights reserved. America, Inc. © 2012 Nature 50%, that of the E133A mutant was reduced by about 90% and that of groups of the RNA. Binding of the two RNA fragments to MBNL1 the D1-9 mutant was almost abolished (Fig. 4d, upper panel). ZF3/4 was antiparallel, suggesting that MBNL1 may bind a looped To determine whether the mutations dramatically disrupt the over- RNA9. Regarding the high similarity of ZAP zinc-finger motifs ZF2

npg all structure of the protein, the D1-9 mutant of NZAP-Zeo was exam- and ZF4 with those of MBNL1 ZF3 and ZF4, we propose, on one ined for its ability to bind to the exosome component Rrp46 (ref. 10), hand, that ZAP may bind its target RNA in a similar manner but may deadenylase PARN1 and RNA helicase p72 (ref. 6). Co-immunopre- be more complicated because more zinc-finger motifs are involved cipitation data showed that the D1-9 mutant interacted with Rrp46, in RNA binding. On the other hand, the protein may bind different PARN and p72 (Supplementary Fig. 5) with slightly reduced affini- target RNAs slightly differently, and thus the contribution of each ties. In vitro RNA binding results revealed that specific binding of the residue to the binding of different RNAs may be different. This would mutants to target RNA was little affected (Supplementary Fig. 4). explain why mutations in residues such as Lys76, Tyr184 and Arg189 These results suggest that the D1-9 mutation does not substantially have different effects on different target RNAs. disrupt the overall structure of the protein and that the intermolecular Results from our structural and functional studies suggest that N-terminal interaction is not required for the protein to bind to its ZAP molecules form a dimer (Fig. 4), consistent with a recent report target RNA and the RNA degradation machinery. that ZAP molecules interact with each other, as determined using the split-Gaussia assay11. Because ZAP must first bind RNA and then One ZAP-responsive RNA binds two ZAP molecules recruit cellular machinery to degrade the RNA, it is conceivable that The above results suggest that ZAP forms a dimer. To investigate whether a dimerized protein would provide a larger platform than a mono- ZAP dimers bind one or two ZAP-sensitive RNA molecules, we per- mer to coordinate such a complex process. ITC results show that two formed isothermal titration calorimetric (ITC) analysis. Recombinant ZAP molecules bind to one ZAP-responsive RNA, suggesting that NZAP precipitated at the high concentrations required for this assay, ZAP functions as a dimer (Fig. 4e and Supplementary Fig. 6). This but recombinant NZAP-Zeo was stable and was thus used for this notion is further supported by the observation that there are two assay. Results show that one Na RNA binds two NZAP-Zeo molecules ZAP-binding modules within the minimum ZAP-responsive RNA (Fig. 4e and Supplementary Fig. 6). We reasoned that if one dimeric fragment Na (Fig. 4f). ZAP binds one RNA molecule, there should be two ZAP-binding mod- The structure of ZAP also provides clues about some features of ules within one ZAP-sensitive RNA molecule. The RNA fragment Na its target RNA. Residues of the putative RNA-binding surface occupy

434 VOLUME 19 NUMBER 4 APRIL 2012 nature structural & molecular biology a r t i c l e s

a three-dimensional cleft covering a surface area of approximately We thank colleagues at the High Energy Accelerator Research Organization 220 Å2 that is 16 Å in depth. The positions of the nucleotides in the (KEK), Japan, and at the Shanghai Synchrotron Radiation Facility for assistance model (Fig. 5b) suggest that they may not be from a consecutive in the use of the synchrotron resource. This work was supported by grants to Y.L. from the Ministry of Science and Technology (863 Project 2006AA02A314; 973 linear sequence but rather from different regions of the target RNA. Programs 2007CB914303, 2012CB910204 and 2011CB910304) and the National The presence of two cavities on the surface suggests that specific Science Foundation (30925011, 31021062 and 31030024) of China, and by grants nucleotides at specific positions are required. Because each cavity to G.G. from the National Science Foundation (81030030), the Ministry of Science can accommodate only one or two nucleotides, the sequence require- and Technology (973 Program 2012CB910203) and the Ministry of Health (2012ZX10001-006) of China. ment may not be obvious in the primary sequence of a ZAP-binding RNA. Taken together, these structural features of ZAP suggest that the AUTHOR CONTRIBUTIONS target RNA should have some specific tertiary structure, with certain S.C. and K.Z. obtained the crystal of NZAP; S.C. and J.S. determined the structure; nucleotides positioned precisely to fit into the three-dimensional cleft, Y.X., X.W. and S.C. carried out the and functional experiments, and and one or two nucleotides positioned specifically to fit into the cavi- S.C., Y.X., G.G. and Y.L. wrote this paper. ties (Fig. 5b). In addition, the fact that two ZAP-binding modules are COMPETING FINANCIAL INTERESTS required in each ZAP-responsive RNA fragment adds another layer The authors declare no competing financial interests. of complexity to the RNA structure. Taken together, features of the Published online at http://www.nature.com/nsmb/. ZAP-responsive RNA predicted on the basis of the structure of ZAP Reprints and permissions information is available online at http://www.nature.com/ explain why only a small number of ZAP-responsive RNAs have been reprints/index.html. identified and why they are at least 500 nt long and have no easily identified sequence motifs. 1. Zhu, Y. et al. Zinc-finger antiviral protein inhibits HIV-1 infection by selectively The availability of the structure of the N-terminal region of ZAP targeting multiply spliced viral mRNAs for degradation. Proc. Natl. Acad. Sci. USA 108, 15834–15839 (2011). offers a platform for further investigations into the mechanisms of 2. Müller, S. et al. Inhibition of filovirus replication by the zinc-finger antiviral protein. ZAP. The ZAP structure reported here not only provides an example J. Virol. 81, 2391–2400 (2007). 3. Bick, M.J. et al. Expression of the zinc-finger antiviral protein inhibits alphavirus of how multiple tandem zinc-finger motifs can be arranged to build replication. J. Virol. 77, 11555–11562 (2003). up a module for RNA binding, but also helps to elucidate the molecu- 4. Guo, X., Carroll, J.W., Macdonald, M.R., Goff, S.P. & Gao, G. The zinc-finger antiviral lar mechanism underlying ZAP function. protein directly binds to specific viral mRNAs through the CCCH zinc-finger motifs. J. Virol. 78, 12781–12787 (2004). 5. Zhu, Y. & Gao, G. ZAP-mediated mRNA degradation. RNA Biol. 5, 65–67 (2008). Methods 6. Chen, G., Guo, X., Lv, F., Xu, Y. & Gao, G. p72 DEAD box RNA helicase is required Methods and any associated references are available in the online for optimal function of the zinc-finger antiviral protein.Proc. Natl. Acad. Sci. USA 105, 4352–4357 (2008). version of the paper at http://www.nature.com/nsmb/. 7. Gao, G., Guo, X. & Goff, S.P. Inhibition of retroviral RNA production by ZAP, a CCCH-type zinc-finger protein. Science 297, 1703–1706 (2002). 8. Hayakawa, S. et al. ZAPS is a potent stimulator of signaling mediated by the RNA Accession codes. Protein Data Bank: Coordinates for NZAP225 have helicase RIG-I during antiviral responses. Nat. Immunol. 12, 37–44 (2011). been deposited under the accession code 3U9G. 9. Teplova, M. & Patel, D.J. Structural insights into RNA recognition by the alternative- splicing regulator muscleblind-like MBNL1. Nat. Struct. Mol. Biol. 15, 1343–1351 Note: Supplementary information is available on the Nature Structural & Molecular (2008). Biology website. 10. Guo, X., Ma, J., Sun, J. & Gao, G. The zinc-finger antiviral protein recruits the RNA processing exosome to degrade the target mRNA. Proc. Natl. Acad. Sci. USA 104, 151–156 (2007). Acknowledgments 11. Law, L.M. et al. Identification of a dominant negative inhibitor of human zinc-finger © 2012 Nature America, Inc. All rights reserved. America, Inc. © 2012 Nature We thank K. Ye for advice and valuable discussions, L. Wu, H. Liang, H. Wang and antiviral protein reveals a functional endogenous pool and critical homotypic X. Tao for technical assistance, and J. Fleming for critical reading of this manuscript. interactions. J. Virol. 84, 4504–4512 (2010). npg

nature structural & molecular biology VOLUME 19 NUMBER 4 APRIL 2012 435 ONLINE METHODS of the NZAP-zeocin fusion protein for binding Na RNA were determined by ITC Protein expression, purification, crystallization and data collection.The at 20 °C using iTC200 isothermal titration calorimetry (Microcal, GE Healthcare). N-terminal domain of rat zinc-finger antiviral protein Zc3hav1 covering the first The protein was dialyzed against an assay buffer (137 mM NaCl; 2.7 mM KCl; 225 residues (NZAP225) was expressed as a GST fusion protein in Escherichia 10 mM Na2HPO4; 2 mM KH2PO4; 1 mM EDTA; 10 µM ZnCl2) and Na RNA was coli. The GST tag was removed by protease cleavage and NZAP225 was purified dissolved in the same buffer. The exothermic heat of the reaction was measured by successive chromatography methods. Crystals were obtained in a solution by 25 sequential 1.5-µl injections of the protein (41 µM) into 200 µl of the RNA containing 15% (w/v) PEG3350 (Fluka), 0.06 M magnesium formate dehydrate solution (1.87 µM), spaced at intervals of 150 s. The heat of dilution was obtained (Sigma-Aldrich) at 16 °C. For data collection, crystals were flash frozen (100 K) by injecting assay buffer into the RNA solution and was subtracted from the heat in the above reservoir solution supplemented with 10% (v/v) glycerol. Diffraction of reaction before the fitting process. Binding curves were analyzed by nonlinear data were collected on a 17-A beam line at KEK, Japan. SAD datasets were col- least-squares fitting of the data using Microcal Origin software. lected to 1.8-Å resolution at wavelength 1.25 Å. The data were processed by HKL2000 (ref. 12). The crystals of NZAP belongs to space group P3(2)21, with Circular dichroism secondary structure analysis method. Circular dichroism one protein molecule per asymmetric unit (Table 1). (CD) spectra data were collected on a Bio-Logic MOS-450 CD spectrometer at room temperature using a 1-mm path-length cell. Proteins were dissolved in PBS 13 Structure determination and refinement.The program PHENIX AutoSol buffer (137 mM NaCl; 2.7 mM KCl; 10 mM Na2HPO4; 2 mM KH2PO4). Protein was used to calculate SAD initiative phases. Automatic protein model build- concentrations were estimated from the calculated molar extinction coefficient at ing was performed with PHENIX AutoBuild13. The resulting protein models 280 nm. The corresponding concentrations were 2.35 mg ml−1 and 2.61 mg ml−1, were completed manually using COOT14. The final models were refined using respectively. Both samples were scanned three times and the averaged data REFMAC15. Ninety-eight per cent of residues were located in the most favorable were used. region and none in the disallowed region of the Ramachandran plot as evaluated 16 using PROCHECK in CCP4 (ref. 17) (Table 1). The final model had an Rwork factor of 19.1% and an Rfree factor of 22.8%. Atomic coordinates and diffraction data have been deposited in PDB with accession code 3U9G. Structural figures 12. Otwinowski, Z. & Minor, W. Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol. 276, 307–326 (1997). were prepared using PyMOL (http://www.pymol.org). All figures were edited by 13. Adams, P.D. et al. PHENIX: a comprehensive Python-based system for macromolecular using Microsoft PowerPoint (Microsoft Corporation) or Photoshop CS3 (Adobe structure solution. Acta Crystallogr. D Biol. Crystallogr. 66, 213–221 Systems Incorporated). (2010). 14. Emsley, P., Lohkamp, B., Scott, W.G. & Cowtan, K. Features and development of Coot. Acta Crystallogr. D Biol. Crystallogr. 66, 486–501 (2010). EGS cross-linking assays.Ethylene glycol bis(succinimidyl succinate) (EGS; 15. Vagin, A.A. et al. REFMAC5 dictionary: organization of prior chemical knowledge 4.7 µl 0.2 mM EGS in DMSO) cross-linking reactions were performed with 2 µg and guidelines for its use. Acta Crystallogr. D Biol. Crystallogr. 60, 2184–2195 NZAP225 in PBS for 2 h and quenched with 2.7 µl 1 M Tris, pH 7.5. The proteins (2004). were detected by western blotting. 16. Laskowski, R.A., Macarthur, M.W., Moss, D.S. & Thornton, J.M. PROCHECK: a program to check the stereochemical quality of protein structures. J. Appl. Crystallogr. 26, 283–291 (1993). Isothermal titration calorimetric analysis. Recombinant NZAP254 (residues 17. Collaborative Computational Project. N., The CCP4 suite: programs for protein 1–254) fused with zeocin was used in this assay. The thermodynamic parameters crystallography. Acta Crystallogr. D Biol. Crystallogr. 50, 760–763 (1994). © 2012 Nature America, Inc. All rights reserved. America, Inc. © 2012 Nature npg

nature structural & molecular biology doi:10.1038/nsmb.2243