The Structure of the Zinc Finger Domain from Human Splicing Factor ZNF265 Fold*

Identification of the protein domains that are responsible for RNA recognition has lagged behind the charac-terization of protein-DNA interactions. However, it is now becoming clear that a range of structural motifs bind to RNA and their structures and molecular mech-anisms of action are beginning to be elucidated. In this report, we have expressed and purified one of the two putative RNA-binding domains from ZNF265, a protein that has been shown to bind to the spliceosomal components U1-70K and U2AF 35 and to direct alternative splicing. We show that this domain, which contains four highly conserved cysteine residues, forms a stable, monomeric structure upon the addition of 1 molar eq of Zn(II). Determination of the solution structure of this domain reveals a conformation comprising two stacked (cid:1) -hairpins oriented at (cid:1) 80° to each other and sandwich-ing the zinc ion; the fold resembles the zinc ribbon class of zinc-binding domains, although with one less (cid:1) -strand than most members of the class. Analysis of the structure reveals a striking resemblance to known RNA-binding motifs in terms of the distribution of key surface

The zinc finger architecture has provided a diverse range of structures and functions (1), including recognition of protein, DNA, and RNA targets. Whereas the structural basis for inter-actions between DNA and classical zinc fingers has been well characterized (2)(3)(4), RNA-zinc finger interactions are less well understood, with only limited structural information on the conformation of RNA-binding domains available. Available structures have indicated that in many cases, RNA-binding domains are not completely folded in the absence of their target RNA sequence (5). Conformational changes have been observed to occur in both protein and the RNA target upon binding (5)(6)(7)(8)(9), and even completely unstructured domains have been found to fold upon interacting with RNA (10). It has therefore been suggested that rigid proteins may be unfavorable for RNAbinding and flexible structures may increase specificity (5).
In recent years, the functions of many RNA-binding proteins that regulate the process of RNA splicing have been delineated, enhancing our understanding of splice site recognition, spliceosomal coordination, and alternative splicing. Typically, the accessory proteins involved in splicing (SR proteins) contain an N-terminal RNA-binding domain and a C-terminal proteinbinding or RS domain that is rich in Arg and Ser residues.
The human SR protein, ZNF265 (or ZIS) (11), can bind to the splicing factors U1-70K and U2AF 35 (components of the E-complex that forms early in the splicing process) and is able to stimulate alternative splicing (12). It has also been shown to immunoprecipitate with splicing factors in association with mRNA and to co-localize in the nucleus with components of the transcriptosome (12), indicating that it plays a role in mRNA processing. ZNF265 is a multidomain protein, and is highly conserved in many organisms, including mouse (Zfp265) (13), rat (Zfp265; formerly Zis) (14), and Xenopus (C4SR) (15). Like other SR proteins, ZNF265 contains a C-terminal RS domain, as well as a nuclear localization sequence and a glutamic acidrich domain. However, in contrast to other SR proteins, the N terminus of ZNF265 does not contain any of the known RNA recognition motifs, but rather contains 8 Cys residues spaced appropriately to give rise to two putative zinc finger domains, each with a CX 2/4 CX 10 CX 2 C topology.
Domains with homology to the putative zinc fingers of ZNF265 have been observed in a wide range of proteins, including ubiquitin ligases such as Mdm2 (16,17) and RanBP2 (18,19), as well as RNA-binding proteins such as EWS (20). However, little structural information is available on any of these proteins. In addition, the putative zinc finger domains appear to be capable of carrying out different functions depending on context. In RanBP2, the eight tandem zinc finger domains are essential for exportin-1 binding (21), whereas the double zinc finger domain in C4SR, the Xenopus homolog of ZNF265, binds to RNA (15).
In the present study, we show that the conserved Cys-containing motifs in ZNF265 are genuine zinc-binding domains * This work was supported in part by a grant from the Australian Research Council. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The and we have determined the solution conformation of the first zinc finger of ZNF265, using NMR methods. The structure appears to constitute a variant of the zinc ribbon class of zinc-binding domains (1). In addition, we demonstrate that the ZNF265 zinc finger domain can bind cyclin B1 mRNA, indicating that this fold has apparently been adapted for both RNAbinding and protein-binding. Such adaptability appears to be common among zinc-binding motifs.

EXPERIMENTAL PROCEDURES
Sequence Analysis-All sequence analysis and alignments were performed using the software pileup and prettybox (Australian Genome Information Service). 1 Expression and Purification-The first putative zinc finger region of ZNF265 is located at the N-terminal end of the protein. A construct encoding residues 1-40 of human ZNF265 (accession code NP_005446) was therefore cloned into the Escherichia coli expression vector pGEX 4T-3 (Amersham Biosciences), creating a C-terminal fusion with glutathione S-transferase (GST). 2 Note that DNA sequencing revealed that all 10 isolated clones contained aspartic acid at position 32, rather than asparagine (the residue listed in the GenBank TM sequence). It is likely that the sequence variation indicates either a single nucleotide polymorphism or perhaps a sequencing error in the original paper (11); only one cDNA was sequenced in this initial report. The fusion protein was expressed in the host strain DH5␣ grown in Luria-Bertani broth. Cells were grown at 37°C, and expression of ZNF265-F1 was induced at an A 600 nm of 0.6 by the addition of isopropyl-␤-D-thiogalactopyranoside (0.5 mM). After a further 3 h, cells were harvested by centrifugation, and the cell pellet was stored at Ϫ80°C prior to use.
Cell pellets were resuspended in sonication buffer containing Tris (50 mM, pH 8), NaCl (50 mM), Triton X-100 (1%), phenylmethylsulfonyl fluoride (0.5 mM), and ␤-mercaptoethanol (2 mM). Cells were lysed by sonication and centrifuged. The supernatant was loaded onto a glutathione-Sepharose TM 4B (Amersham Biosciences) column pre-equilibrated in sonication buffer. The column was washed and the 45-residue ZNF265-F1 peptide (containing an additional Gly-Ser at the N terminus, arising from the thrombin recognition site and a Gly-Pro-Ile at the C terminus because of cloning via a pGEM TM -T Easy intermediate) cleaved from the column with thrombin (5 units/1 ml of beads) at room temperature overnight. The eluted peptide was purified to homogeneity by reverse phase high performance liquid chromatography, and its identity confirmed using electrospray mass spectrometry (M theor ϭ 4995.6 Da; M obs ϭ 4996.0 Da).
Analytical Ultracentrifugation-Sedimentation equilibrium experiments were performed using a Beckman model XL-A analytical ultracentrifuge equipped with an An-60 Ti rotor. ZNF265-F1 was prepared by dissolving lyophilized peptide in H 2 O containing 1 mM tris(2-carboxyethyl)phosphine and 1.7 mM ZnSO 4 and the pH was adjusted to 5.8 using 0.1 M NaOH. Protein concentrations were determined in the XL-A by reference to a blank containing this buffer alone. Data were collected in two-sector cells as absorbance (280 nm) versus radius scans (0.001 cm increments, 10 averages) at 36,000, 43,000, and 56,000 rpm. Scans were collected at 3-h intervals and compared to determine when the samples reached equilibrium. Analysis of the data was carried out using the NONLIN software (22) and the best model and final parameters were determined by examination of the residuals derived from fits to several models. The partial specific volume of ZNF265-F1 (0.708 ml g Ϫ1 ) and the density of the buffer (0.999 g ml Ϫ1 ) were determined from the amino acid sequence (23) using the program SEDNTERP (www.bbri.org/ rasmb/rasmb.html) (24).
NMR Spectroscopy-ZNF265-F1 was prepared for NMR by dissolving ϳ2.5 mg of peptide in H 2 O/D 2 O (95:5, 500 l) containing tris(2carboxyethyl)phosphine (5 mM) and ZnSO 4 (1.5 mM), giving a ϳ1 mM sample. The sample was prepared at low pH to avoid protein aggregation, and the pH was then increased to 5.8 using 0.1 M NaOH. The sample was placed in a 5-mm outer diameter, susceptibility matched microcell (Shigemi). For experiments in D 2 O, the sample was lyphophilized and reconstituted in a microcell in 500 l of 99.96% D 2 O.
All NMR experiments used for determination of the structure of ZNF265-F1 were carried out at 298 K on a 600-MHz Bruker DRX600 spectrometer, equipped with a 5-mm triple resonance probe and three-axis pulsed field gradients. Water suppression was achieved using pulsed-field gradients and the WATERGATE sequence (25). The following homonuclear two-dimensional spectra were recorded on the ZNF265-F1 sample: DQFCOSY (26), TOCSY (27) ( m ϭ 35 and 70 ms), and NOESY (28) ( m ϭ 50 and 200 ms). Acquisition times were 33 and 131 ms for each of these experiments in t 1 and t 2 , respectively. Spectra were processed using XWINNMR (Bruker) on a Silicon Graphics O 2 work station. The 1 H frequency scale of all spectra was directly referenced to sodium 3-trimethylsilyl propionate-2,2,3,3-d 4 at 0.00 ppm. Two-dimensional NMR spectra were processed using Lorentz-Gauss window functions in the directly detected dimension and shifted square sine bell functions in the indirectly detected dimension. Polynomial baseline corrections were applied to the processed spectra in F 2 . Spectra were analyzed using the program XEASY (29).
Structure Calculations-NOE-derived distance restraints were obtained from intensities of cross-peaks in the two-dimensional 1 H NOESY spectra and calibrated using the CALIBA module of DYANA (30). Dihedral angle restraints for angles were set to Ϫ60 Ϯ 40 o for 3 J HN␣ Ͻ 6 Hz and Ϫ120 Ϯ 40 o for 3 J HN␣ Ͼ 8 Hz, based on the use of the INFIT algorithm (31) that is part of the XEASY package. In addition, a restraint of Ϫ100 Ϯ 80 o for angles was applied for residues in which the intraresidue H ␣ -H N NOE was clearly weaker than the NOE between H ␣ and H N of the following residue (32). The ␤ methylene protons of the two Pro residues were stereospecifically assigned from NOESY spectra on the basis that H ␣ is always closer to H ␤3 than H ␤2 (33). Both X-Pro bonds were identified as being in the trans-conformation on the basis of strong NOEs between the H ␦ values of each Pro residue and the H ␣ protons of the preceding residues (34). No hydrogen bond restraints were used in the structure calculations.
Initially, structure calculations were performed in DYANA using a distance geometry/simulated annealing protocol (30) and additional NOEs were assigned iteratively based on earlier sets of structures. No assumption about zinc coordination was made at this stage and preliminary structures indicated the four cysteines (Cys 15 , Cys 20 , Cys 31 , and Cys 34 ) were possible zinc ligands. To include additional distance and angle constraints that maintain the tetrahedral bonding geometry and appropriate bond lengths with the zinc atom (35), as well as to incorporate ambiguous restraints, subsequent structural refinement was performed in CNS (36) using the package ARIA1.2 (37,38). Manually assigned NOEs in combination with the remaining ambiguous NOEs were included in the ARIA structure calculations and the latter NOEs were iteratively assigned by the program.
Briefly, eight cycles of 20 structures each and a last cycle of 500 structures were performed in the final refinement. Manually assigned NOEs were included in iteration zero as soft restraints. In each cycle the seven lowest energy structures from the previous iteration were used to extract additional NOE restraints with a tolerance of 0.02 ppm in the F1 dimension and 0.02 ppm in the F2 dimension. If a restraint was violated by more than a predefined target value in over 50% of the seven structures, it was discarded. The violation target value was progressively reduced from 1000 Å in iteration zero to 0.1 Å in iteration eight. Ambiguous distance restraints were treated as described (38) and the peak volume cut off was gradually reduced from 1.01 in the first iteration to 0.80 in the last. Dihedral angle restraints and stereospecific assignments as detailed above were also included in the calculations. The final assignments made by ARIA were checked and corrected manually where necessary. Overall, 4 (1.2%) non-sequential inter-residue NOEs for which no consistent assignment could be determined were excluded from the final calculations.
Calculations were carried out in the simplified all-hydrogen PAR-ALLHDG5.3 force field with non-bonded interactions modeled by PROLSQ force field (39); floating chirality assignment (40) was used for all methylene and isopropyl groups for which no stereospecific assignment could be made. Finally, the 100 lowest energy structures were refined in a 9-Å shell of water molecules (41). The 20 conformers with the lowest value of E tot were visualized and analyzed using the programs MOLMOL (42), PROCHECK (43), and WhatIf (44). The family of 20 low-energy structures (accession code 1N0Z) have been deposited in the Protein Data Bank. Chemical shift and restraint lists that were used in the structure calculations have been deposited in BioMagRes-Bank (accession code 5667).
RNA Electrophoretic Mobility Shift Assay (REMSA)-Expression plasmids harboring the full-length cyclin B1 cDNA were used to transcribe the cyclin RNA probe in vitro. Following linearization with BamHI, 1 g of template was incubated for 2 h at 37°C in the presence of 50 Ci of [␣-32 P]UTP (DuPont), 10 M UTP, 0.5 mM ATP, 0.5 mM GTP, 0.5 mM CTP, 20 units of RNase inhibitor (Promega), and 50 units of T7 RNA polymerase. RNA probes were purified on a 4% polyacryl-amide-urea denaturing gel, eluted overnight at room temperature in a solution containing 500 mM NH 4 CH 3 COO and 1 mM EDTA, ethanol precipitated at Ϫ80°C, and resuspended in water (100 cps/l) as previously described (45).
For the binding assays, 0.1-5.0 g of affinity purified GST-ZNF265-F1 protein was preincubated with 150 g of heparin (Sigma) for 10 min at room temperature in CEB buffer before addition of the RNA probe (200 cps). After 30 min incubation at room temperature, samples were subjected to RNase T1 digestion (1 l; 2 units; Roche Applied Science) for 15 min at room temperature. Samples were then electrophoresed through a 4% native polyacrylamide gel and protein-RNA complexes were visualized by autoradiography.

RESULTS
Homology of the Zinc Finger Domain of ZNF265-To investigate the conservation of the zinc finger domains of ZNF265, a Blast analysis was conducted using the amino acid sequence of the first zinc finger domain. Fig. 1 shows that in addition to being nearly identical to each other, both of the zinc fingers of human ZNF265 are completely conserved across a number of eukaryotic species, including rat and mouse ZFP265 and the RNA-binding protein C4SR from Xenopus laevis. Furthermore, these domains show homology to the zinc fingers found in Drosophila sol (46), as well as in RNA-binding proteins from Arabidopsis thaliana, suggesting ancient origins for this zinc finger motif, and therefore an important function(s). More than 100 proteins carry domains of this type, and these proteins exhibit a variety of domain structures and functions. Proteins may contain between one and eight copies of these domains, which may be arranged either in tandem clusters or spread throughout the protein. In some cases, no other recognizable sequence motifs have been identified.
Despite this diversity, two functional themes are apparent. Several of the proteins (such as RanBP2) are involved in either ubiquitination or SUMOylation, whereas many others appear to be involved in RNA metabolism. Thus, in a number of cases, the putative zinc-binding motif is directly C-terminal to a known RNA recognition motif, such as an RRM domain (47), and proteins such as ZNF265 contain other elements associated with splicing, such as RS domains (48).
The Cysteine-rich Region of ZNF265 Forms a Monomeric Zinc-dependent Structure-Sequence analysis performed at the time of the initial discovery of rat ZFP265 ("Zis" (14)) identified the putative zinc finger domains of this protein, although no experimental verification was made. We therefore subcloned DNA encoding the first such domain from ZNF265 (ZNF265-F1) into the pGEX-4T-3 vector. The protein was expressed as a GST fusion and purified using affinity chromatography and reversed-phase high performance liquid chromatography. Onedimensional 1 H NMR spectra, recorded at pH 5.8 in both the absence and presence of 1 molar eq of Zn(II) (Fig. 2A), showed a substantial increase in chemical shift dispersion in the presence of Zn(II). This increase is consistent with the formation of persistent secondary structure and demonstrates that these domains are indeed genuine zinc finger domains.
To determine the aggregation state of ZNF265-F1, we used sedimentation equilibrium methods (Fig. 2B). Data were recorded on two different loading concentrations at three speeds, and these data were well fitted by a model incorporating a single, ideal species with a molecular mass of 4.9 kDa (with 95% confidence intervals of 4.5 and 5.2 kDa; M theor ϭ 5,061 Da). Thus ZNF265-F1 is monomeric in solution under these conditions.
Determination of the Structure of ZNF265-F1-The high quality of the 1 H NMR spectra of ZNF265-F1 in the presence of Zn(II) (Fig. 2A) 2. ZNF265-F1 is a monomeric  zinc-binding domain. A, one-dimensional 1 H NMR spectra of ZNF265-F1 in the absence (top) and presence (bottom) of 1 molar eq of Zn(II). The substantial increase in chemical shift dispersion in the presence of Zn(II) indicates the formation of stable secondary structure. Spectra were recorded at pH 5.8 and 298 K, on a 600 MHz spectrometer (sample concentration was 0.5 mM). B, sedimentation equilibrium data. Data were recorded at 43,000 rpm (25°C). The lower graph displays a plot of absorbance at 280 nm versus radius (r 2 /2; cm 2 ), whereas the upper graph illustrates the residual deviations resulting from the fit of an ideal single species model to the data. were achieved by using standard two-dimensional homonuclear NMR experiments. Using DYANA (30), NOEs were iteratively assigned and a preliminary was fold determined; no zinc ion was included, and the zinc coordination site was not restrained during these calculations. On the basis of these structures, it was clear that the four cysteine thiol groups (Cys 15 , Cys 20 , Cys 31 , and Cys 34 ) comprised the zinc coordination sphere. Subsequent calculations, carried out using the ARIA protocol (37,38) implemented in CNS (36), therefore included a zinc ion, together with constraints defining tetrahedral coordination. The remaining ambiguous NOEs were introduced iteratively in ARIA in an automated manner. The final ensemble of the 20 lowest energy structures is based on a total of 709 unambiguous intra-and inter-residue distance restraints derived from two-dimensional NOESY spectra and 22 backbone dihedral restraints. The total number of experimental restraints was 731. Of these, 347 were meaningful interresidues restraints, representing an average of 12.8 constraints per structured residue (i.e. residues . From the final ARIA calculations, the 20 lowest energy structures were chosen to represent the solution structure of ZNF265-F1 (Fig. 3A). The structures display good covalent geometry and good nonbonded contacts. Analysis of the structures with PROCHECK (43) shows that, for residues that exhibit angle order parameters Ͼ0.6, over 99.6% of non-glycine and non-proline residues falls into the most favored or additionally allowed regions of the Ramachandran plot. Structural statistics for the ensemble are given in Table I. ZNF265-F1 Forms a Zinc Ribbon-Residues 12-38 of ZNF265-F1 adopt a compact, globular fold that incorporates a single zinc ion. The fold consists of two distorted ␤-hairpins that each provide two of the zinc ligating cysteines (Fig. 3B); the Zn(II) ion is effectively sandwiched between the two hairpins, which cross each other at an angle of ϳ80° (Fig. 3C). The first ␤-hairpin is formed by residues Asp 12 -Asn 24 (strands 1 and 2 comprise residues Trp 13 -Cys 15 , and Asn 22 -Asn 24 , respectively) and displays backbone hydrogen bonds between the H N proton of Cys 15 and the carbonyl oxygen of Asn 22 , as well as between the H N proton of Asn 24 and the carbonyl oxygen of Trp 13 . In addition there are two hydrogen bonds to the S ␥ atom of Cys 20 , from the backbone amide protons of Asn 22 and Gly 21 (2.1 and 3.1 Å in the lowest energy structure, respectively); and an additional hydrogen bond to the S ␥ atom of Cys 15 , from the backbone amide protons of Asp 17 (3.6 Å). These hydrogen bonds are characteristic of the rubredoxin knuckle, which is very common in zinc-binding domains (49). Residues Cys 15 -Lys 18 form a type VIII ␤-turn, judging from backbone and angles (although no (i,iϩ3) hydrogen bond is observed), whereas residues Asp 17 -Cys 20 form an interlinked type I ␤-turn (50). The side chain carbonyl oxygen of Asp 17 forms hydrogen bonds with the H N protons of Lys 18 and Lys 19 , as is commonly observed in type I turns (50).  The second ␤-hairpin consists of residues Thr 29 -Lys 38 but is not recognized by the Kabsch and Sander (51) secondary structure prediction algorithm (as implemented in MOLMOL; 42) in any of the final conformers. A closer manual inspection of the structures, however, reveals that many of the hydrogen bonding patterns observed across the first ␤-sheet are also maintained here. For example, hydrogen bonds can be observed between the H ␥ proton of Ser 30 and the side chain carbonyl oxygen of Glu 37 , the H N proton of Lys 38 and the carbonyl oxygen of Thr 29 , and between the H N proton of Cys 31 and the carbonyl oxygen of Arg 36 . This hairpin is recognized in the family of conformers that are obtained prior to water refinement, but the slight loosening of the ensemble that takes place in the water refinement puts the hydrogen bond parameters just outside the tolerances set by the Kabsch and Sander (51) algorithm. Consequently, we have not shown this hairpin in Fig. 3. The two strands of this "hairpin" are connected by residues Cys 31 -Cys 34 , which form a half-knuckle with a hydrogen bond linking the S ␥ atom of Cys 31 to the backbone amide proton of Arg 33 . A small hydrophobic core is formed by Trp 13 , Cys 15 , Pro 16 , Asn 22 , Asn 24 , Cys 31 , and one face of the aliphatic side chain of Lys 40 .
A search of known protein structures, using ZNF265-F1 as a template and the program DALI (52), did not reveal any structural homologs, indicating that ZNF265-F1 constitutes a new protein fold. Because the fold consists of two hairpins crossing each other, we have termed this new fold a crossed finger or CF domain. However, DALI searches can yield false negatives when dealing with small domains (1). Inspection of the structure reveals that it falls into the zinc ribbon class of zincbinding domains, as defined by Krishna et al. (1). Members of the zinc ribbon class all possess two ␤-hairpins, but otherwise display significant sequence and structural variability. Typically, a third ␤-strand adjoins one of the hairpins. Many classical zinc ribbon proteins are involved in the transcriptional/ translational machinery and are capable of interacting with either DNA or RNA (1). Overlays of ZNF-F1 with typical zinc ribbons (e.g. the ribosomal protein L37E; PDB number 1JJ2) give RMSDs of 2-3 Å.
Within the family of crossed finger domains, a number of amino acids, other than the zinc ligands, are highly conserved (Fig. 1). Fig. 3D illustrates the distribution of these conserved amino acids on the structure of ZNF265-F1. Two of these, Trp 13 and Asn 24 , are Ͼ90% buried, and are likely to be important for maintaining the folded structure of the CF domain. Indeed, the buried side chain amide protons of Asn 24 form hydrogen bonds with the backbone carbonyl oxygens of Trp 13 and Ile 14 . Lys 38 also packs against Trp 13 and probably contributes to fold stability, whereas the remaining most highly conserved residues, Asp 12 , Lys 18 , Arg 27 , Arg 28 , and Phe 25 , are largely exposed. These latter residues may therefore serve a functional role.
Could the CF be a Novel RNA-binding Domain?-Analysis of the surface of the ZNF265-F1 structure reveals a number of conserved amino acids that are presumably not involved in specifying the folded conformation of the domain. Two lysines (Lys 18 and Lys 19 ) and three arginine (Arg 27 , Arg 28 , and Arg 33 ) residues that lie within the well defined portion of the domain and are well conserved are highlighted in Fig. 4A together with the exposed phenylalanine residue Phe 25 . Strikingly, the spatial arrangement of a number of these residues is remarkably similar to the RNA-binding surface identified in the double stranded RNA-binding module of PKR (Fig. 4B). A large body of structural (53) and mutagenic (54 -56) data has shown that a solvent-exposed phenylalanine, together with several lysine and arginine residues are used to contact RNA.
The ZNF265 Crossed Finger Domain Can Bind mRNA-The structural homology of the CF domain to a known RNA-binding protein suggested to us that this domain might function in RNA recognition. Interestingly, crossed finger domains are found in a number of other proteins that have either been shown directly to be involved in RNA metabolism or contain domains with RNA-related functions. The double-CF domain of C4SR, the Xenopus homolog of ZNF265, has been shown to stably interact with a cyclin B1 mRNA transcript (15). To determine whether ZNF265 could bind to the same cyclin transcript, a REMSA analysis were performed using an in vitro transcribed 32 P-labeled full-length cyclin B1 transcript and purified ZNF265 protein. As shown in Fig. 5, addition of increasing concentrations of ZNF265 to the cyclin transcript pro- duced a protein-RNA complex (indicated by arrow). This was seen clearly when using 5 g of purified material (lane 4), but was also evident at a lower concentration (i.e. 100 ng) after prolonged exposure of the film. These results indicate that ZNF265 is able to bind to the cyclin transcript.

DISCUSSION
The Structure of ZNF265-F1-The crossed finger domain seems to be a subclass of the zinc ribbon fold, and it adds to the already large family of zinc-binding domains in which the zinc probably serves only a structural (rather than a catalytic) role (1). There are at least eight classes of such domains for which structural information is available (1), and the structures that they form are diverse. A recurring theme among these structures is the fact that the zinc ligands are divided into two pairs in the amino acid sequence. The spacing between the two ligands within a pair is generally 2 to 5 amino acids, whereas the sequence connecting the pairs may be up to 50 amino acids in length. It is clear that these H/C-X 2-5 -H/C motifs take up a limited number of conformations, and presumably form a stable macrocyclic structure when bound to a zinc ion. Motifs containing histidine residues often take up a helical conformation (e.g. in classical zinc fingers), whereas C-X 2-5 -C sequences typically form part of a turn or take up an irregular conformation. Given the diversity of folds that contain these motifs, it is interesting to speculate about why they have apparently arisen so frequently during evolution. It is possible that when two H/C-X 2-5 -H/C units arise by random sequence variation and are spaced less than ϳ50 amino acids apart, there is a relatively high probability that this sequence will be able to ligate a zinc ion, thereby forming some kind of persistent structure (high at least compared with the chance of persistent structure forming in the absence of directing influences like metal-binding motifs).
At least three of the classes of zinc-binding domains have been demonstrated to bind specifically to RNA. The retroviral nucleocapsid domain that is essential for RNA packaging comprises a C-X 2 -C-X 4 -H-X 4 -C zinc-binding domain (57). The zinc ribbon fold is found in many ribosomal proteins and recognizes RNA, and a number of proteins containing classical zinc fingers, such as TFIIIA, also recognize RNA (58). Of these, structural alignments indicate that the zinc ribbon resembles most closely the ZNF265 domain.
ZNF265 as an RNA-binding Protein-The REMSA data presented here indicate that the ZNF265 CF domains are capable of interacting with an mRNA transcript from the cyclin B1 gene. These data are in accord with the previous demonstration by Ladomery et al. (15) that the zinc finger region from X. laevis C4SR, a ZNF265 sequence homolog (ϳ98% similarity in the double CF domain region), can stably bind the same RNA message.
Several themes have emerged from studies of RNA-binding proteins to date and at least two of these are reflected in ZNF265. First, the importance of both basic and exposed aromatic amino acids for RNA recognition has become clear (5). The CF domains of ZNF265 contain many basic residues (seven in ZNF265-F1), as well as an exposed phenylalanine (Phe 25 ) and, remarkably, a number of these residues combine to form a surface that is rather similar to that observed in unrelated RNA-binding proteins such as PKR. Indeed, the use of a surface phenylalanine for the recognition of a ribonucleotide base is rather common among RNA-binding proteins; the transcriptional termination factor Rho is another example of this phenomenon (59).
Second, RNA recognition motifs (and double stranded RNAbinding modules) are often found in pairs separated by a flexible linker; and large conformational changes, such as the freezing of a flexible linker between two RNA-binding domains, often occur upon RNA binding (47,60). This is in contrast to DNA-binding zinc fingers that typically have much shorter linkers, usually of seven amino acids between adjacent fingers (61). ZNF265 contains two CF domains separated by a 24residue linker, and interestingly PKR has two double stranded RNA-binding modules that are connected by a 22-residue flexible linker, consistent with the observed RNA-binding function of ZNF265. It has previously been suggested that widely spaced zinc fingers may allow greater flexibility in binding site selection (62,63).
We have shown that the zinc fingers of ZNF265 form an RNA-binding motif, and the protein therefore conforms to the standard architecture of SR protein splicing factors (64,65). It has been shown that ZNF265 has a role in directing alternative splicing and is able to bind to both U1-70K and U2AF 35 (12). However, the binding to these proteins alone is insufficient for directing alternative splicing, because U1-70K and U2AF 35 bind at their respective sites in all introns. Thus, to enable splice site selection the zinc finger domain of ZNF265 may selectively bind RNA targets; we are currently in the process of defining the sequence specificity for the ZNF265-RNA interaction.
The CF Domain in Other Proteins-Sequence analysis of the proteins that contain CF domains indicates a likely role for many CF-containing proteins in RNA metabolism. Whereas a number of these proteins have no known function, at least 30 contain other motifs such as the well characterized RNA recognition motif, or RRM (66), further linking this domain to RNA metabolism. The presence of multiple CF domains in the nuclear transport factor RanBP2 is particularly interesting. RanBP2 is a 358-kDa nucleoporin that is involved in nuclear export (18,67). The presence of CF domains in this protein may indicate that it is involved not only in the export of protein cargo but also in shuttling RNA in and/or out of the nucleus. It should be noted, however, that few of the positively charged residues found in ZNF265-F1 are conserved in either the RanBP2 or the other ubiquitin ligase CF domains. This may argue for the adaptation of the CF fold for more than one function, in the same way that many other classes of zincbinding domains appear now to be capable of a variety of roles (68).
In summary, a combination of structural, REMSA, and sequence analysis data have shown that the zinc-binding domains of the splicing factor ZNF265 are zinc ribbons that bind to RNA, most likely via a mechanism that bears some resemblance to the binding modes of other RNA-binding proteins. These data point to an interesting convergence of function among a wide variety of proteins involved in RNA metabolism.