YxiN Is a Modular Protein Combining a DEx D / H Core and a Specific RNA-binding Domain *

DEx D / H proteins, typically described as RNA helicases, participate in rearrangement of RNA-RNA and possibly RNA-protein complexes in the cell. Aside from the conserved DEx D / H core, members of this protein family often contain N- and C-terminal extensions that are responsible for additional functions. The Bacillus subtilis DEx D / H -box protein YxiN and its Escherichia coli ortholog DbpA contain an (cid:1) 80 amino acid C-terminal extension that has been proposed to specifically interact with a region of 23 S ribosomal RNA including hairpin 92. In this study, the DEx D / H -box core and the C-terminal domain of YxiN were expressed and character-izedasseparateproteins.TheisolatedDEx D / H -boxcore,YxCat,had weak, nonspecific RNA binding activity and showed RNA-stimu-lated ATPase activity with a K m (ATP) that resembled several nonspecific DEx D / H proteins. The isolated C-terminal domain, YxRBD, bound RNA with the high affinity and specificity seen with full-length YxiN. Thus, YxiN is a modular protein combining the activities of the YxCat and YxRBD domains. Footprinting of YxiN and YxRBD on a 172-nucleotide fragment of 23 S rRNA was used to identify the sites of interaction of the C-terminal and helicase domains with the

Members of the DEx D / H helicase family participate in many cellular processes involving rearrangement of RNA-RNA and RNA-protein interactions (1,2). DEx D / H proteins contain a conserved catalytic core consisting of two RecA-like domains that bind ATP and single-stranded RNA (3). Similar to other RecA-like ATPases (4), the cycle of ATP binding, hydrolysis, and release is coupled to a conformational change in the core. In DEx D / H proteins, this results in translocation of the protein along single-stranded RNA and subsequent duplex destabilization and/or ribonucleoprotein disruption (5,6). Outside of the catalytic core, DEx D / H proteins often contain ancillary N-and C-terminal extensions that correlate with their specific cellular functions. In many cases, helicases are part of multiprotein complexes, and the putative role of the ancillary domains is to recruit the helicase to the complex by proteinprotein interactions (7). Alternatively, the ancillary domains can directly bind to a specific RNA substrate, delivering the helicase to its site of action.
Bacillus subtilis DEx D / H protein YxiN and Escherichia coli DbpA define a group of bacterial orthologs with a conserved C-terminal domain of ϳ80 amino acids (Fig. 1A). Both YxiN and DbpA specifically recognize the A-site region of 23 S rRNA, including hairpin 92, suggesting that the RNA specificity is conferred by the C-terminal domain (8 -11). In a domain swap experiment, the C-terminal domain of YxiN was appended to the catalytic core of SrmB, a non-sequence-specific RNA helicase from E. coli (12). The resulting chimera possessed the RNA specificity of YxiN but had catalytic properties similar to those of the parental helicase, SrmB. These results indicate that the C-terminal domain of YxiN is capable of imparting specificity in the context of another DEx D / H box protein, suggesting that the sequence-specific RNA binding and helicase activities are separable and that this group of DEx D / H proteins may be functionally modular. To prove this hypothesis, the helicase and C-terminal domains of YxiN were prepared here as separate polypeptides. This approach permits the intrinsic activities of the catalytic core and the C-terminal domain to be assayed independently and compared with the full-length protein.

Plasmids for Expression of Recombinant YxiN Fragments-Coding
sequences for YxiN protein fragments were PCR-amplified from the native B. subtilis sequence in an existing plasmid (11) and cloned by restriction/ligation into the pTWIN1 vector of an intein-based expression system (New England Biolabs) using the unique NdeI and SpeI sites of the vector. Primers were designed such that the subcloned fragments deleted the first of the tandem inteins of the parent vector and placed the C-terminal residue of each encoded YxiN protein fragment flush with the N-terminal residue of the second intein. In this manner, no extraneous residues remained at the C terminus of the target after selfcleavage of the target-intein fusion. The plasmids used for these studies encoded the DEx D / H core fragment, termed YxCat (residues 1-368) and the C-terminal domain fragment, termed YxRBD (residues 404 -479, which starts at Met-404 of YxiN).
Protein Expression and Purification-Full-length YxiN from a previous study was used (11). YxiN fragments were expressed in E. coli BL21(DE3) either by growing in LB to A 600 of ϳ0.6 and inducing with 0.4 mM isopropyl 1-thio-␤-D-galactopyranoside or by growing in a selfinducing medium (13). Cells were harvested, resuspended in 20 mM Tris-HCl, 500 mM NaCl, pH 7.9, at 20°C (Buffer A) and lysed by sonication. Protein purification was carried out at 4°C. After removal of cell debris by centrifugation, the clarified supernatant was brought to 0.1% (w/v) in polyethyleneimine by dropwise addition from a 10% (w/v) stock solution, pH 7.2, over a period of 10 min. Precipitate was removed by centrifugation, and the supernatant was applied to a chitin column and washed with several column volumes of Buffer A. Then, the column was loaded with 50 mM dithiothreitol in Buffer A, and flow was stopped overnight to allow on-column self-cleavage of the target-intein fusion. The next day, protein was eluted from the column in 2 column volumes of wash, concentrated to a volume of a few milliliters, and loaded onto a gel filtration column (Superdex-200 (Amersham Biosciences) for fragment 1-368; Superdex-75 for fragment 404 -479) pre-equilibrated with 20 mM Tris-HCl, 100 mM NaCl, pH 7.9, at 20°C (Buffer B). Protein from the peaks corresponding to the YxiN fragments were concentrated to Ͼ5 mg/ml and stored at Ϫ70°C in Buffer B.
* 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. 1 To whom correspondence should be addressed. Tel.: 847-491-5139; E-mail: o-uhlenbeck@northwestern.edu.
RNA Substrates-RNAs HPϩ15M and HPϩ15 through HPϩ2 were purchased from Dharmacon, Inc. Poly(A) RNA was purchased from Roche Diagnostics. 172-nt 2 RNA was prepared by run-off transcription as described previously (10).
Gel Mobility Shift Binding Assay-The binding assay was based on the previously described method (14). A low concentration (0.4 nM) of 5Ј-32 P-labeled RNA was mixed in 16-l reactions with a series of protein concentrations in 50 mM HEPES, pH 7.5, 50 mM KCl, 5 mM MgCl 2 , 100 mM dithiothreitol, 70 mM poly(A), 5% (v/v) glycerol, and 0.1% (v/v) Tween 20. Poly(A) was omitted from the YxCat binding experiments. Reactions were incubated for 10 min to reach equilibrium, and 10-l aliquots were loaded onto 5% native acrylamide gels (29:1 ratio of acrylamide to bis-acrylamide) in 0.33ϫ TBE (1ϫ TBE is 90 mM Tris borate, 1 mM EDTA, pH ϳ8.3). Gels were run in 0.33ϫ TBE at 200 V, dried, and exposed to a phosphorimaging screen. The fraction of unbound material was quantitated as the ratio of counts in the free RNA band to the combined counts of the free RNA and the shifted species above. The fraction of bound material (1 Ϫ unbound fraction) was fit to a noncooperative binding curve to derive the dissociation constant. Reported ⌬G and K d values resulted from averages of three experiments. Typical standard deviations of ⌬G values were below 0.7 kcal/mol (ϳ3-fold in K d ), and individual standard deviations are depicted as error bars in Fig. 3.
ATPase Assay-The rate of ATP hydrolysis was measured using the previously described coupled spectroscopic assay (9) at room temperature with the addition of 0.1% (v/v) Tween 20. The individual cuvette method was modified to a high-throughput plate format using the Spec-traMax Plus 384 spectrophotometer (Molecular Devices). In a single experiment, duplicate 50-l reactions were placed in flat-bottom polystyrene 384-well plates, and absorbance was monitored at 338 nm for 15 min. Individual well path lengths experimentally determined by the instrument (PathCheck) were used to normalize the absorbances to a path length of 1 cm. Data from at least two independent experimental measurements were used in fitting. Typical standard deviations of the reported ATP hydrolysis rates at or above half-maximal ATPase stimulation were below 10%, and individual standard deviations are depicted as error bars in Fig

Gel Mobility Shift Binding of RNA by YxiN, YxRBD, and YxCat-The
RNA binding properties of B. subtilis YxiN are similar to those of its E. coli ortholog DbpA. Previous experiments demonstrated a specific interaction of YxiN with RNAs containing hairpin 92 of 23 S rRNA, such as HPϩ15 (Fig. 1B), by an RNA-stimulated ATPase assay, whereas mutation of two bases in the stem loop of the RNA (RNA HPϩ15M) eliminated the stimulation (12). To directly examine the binding and specificity of YxiN, a gel-shift assay was used. Trace concentrations of 5Ј-32 P-labeled RNA HPϩ15 were mixed with varying concentrations of YxiN and analyzed on a native gel. A distinct, slower migrating band was observed ( Fig. 2A) consistent with complex formation between RNA HPϩ15 and YxiN. The smear below the complex is considered to rep-resent specific complex that falls apart during the electrophoresis and is counted as the bound fraction in quantitation. The plot of fraction bound versus protein concentration fits a sigmoid binding curve with a K d of 4.6 nM (Fig. 4). The curve has a Hill coefficient of ϳ1, indicating noncooperative binding of HPϩ15 with YxiN. In a similar experiment using the nonspecific RNA HPϩ15M (Fig. 2B), no complex band was observed even at 5 M YxiN, a 10-fold higher concentration than the highest used with RNA HPϩ15. Thus, base mutations in the loop of RNA HPϩ15 disrupt its interaction with YxiN, indicating that the binding is sequence-specific. DbpA has shown an analogously specific interaction with RNA HPϩ15 by stimulation of ATPase activity (15) and a similar K d in electrophoretic mobility shifts (9.2 nM) (14).
The interaction between YxiN and its specific RNAs has been proposed to involve sequence specific binding of the C-terminal domain and sequence nonspecific binding of the catalytic core (12). To test this proposal directly, the two portions of the gene were cloned separately as fusions with an intein and a chitin-binding domain (pTWIN1, New England Biolabs). YxCat, the fragment of YxiN containing the catalytic core, extends from residues 1-368, including 35 amino acids after the last highly conserved motif ((H/Q)RXGRXGR) (Fig. 1A). This site was chosen because it aligns close to the end of the minimal DEx D / H protein eIF4A. YxRBD, the C-terminal fragment of YxiN, encompasses a region of conserved sequence between residues 404 -479. YxRBD is rich in basic amino acids and has a calculated pI of 10.0. Amino acids 369 -403 of YxiN that lie between YxCat and YxRBD were not included because this region varies in length among DbpA orthologs (7-35 amino acids) and may represent a flexible linker. Expression of these fragments yielded large amounts of soluble proteins that were purified and cleaved on a chitin column and further purified by gel filtration. 15 N, 1 H-HSQC NMR spectra of 15 N-labeled YxRBD revealed a dispersion of amide proton peaks that is consistent with a folded polypeptide. In addition, crystals of the fragment have been obtained, indicating that it constitutes a well folded domain (data not shown).
The individually expressed fragments of YxiN both show RNA binding activity. The addition of the C-terminal fragment, YxRBD, to RNA HPϩ15 shows a shifted band corresponding to the RNA-protein complex (Fig. 2C). As would be expected for its smaller size, the mobility shift by this protein fragment is less than that of full-length YxiN. The affinity of YxRBD to RNA HPϩ15 (5.2 nM) is similar to that of YxiN (4.6 nM) (Fig. 4). Furthermore, interaction of YxRBD with RNA HPϩ15M is not observed, paralleling the sequence specificity of full-length YxiN (Fig. 2D). The catalytic core fragment, YxCat, interacts with RNA HPϩ15 to produce a faint complex band of intermediate mobility at micromolar protein concentrations (Fig. 2E). The binding is not saturated even at the highest concentrations of YxCat tested, indicating a very weak interaction (K d Ͼ 4 M). YxCat binds RNA HPϩ15M just as well as it does RNA HPϩ15 (Fig. 2F), demonstrating that the catalytic core fragment does not have RNA sequence specificity. Thus, YxRBD interacts with RNA tightly and sequence specifically, whereas YxCat binding to RNA is weak and is not sequence-specific.
It was of interest to determine the minimal length of the 5Ј singlestranded extension in RNA HPϩ15 that is needed to bind YxiN and YxRBD. To this end, a set of RNAs with single nucleotide truncations from the 5Ј end of RNA HPϩ15 (Fig. 1B) were each used in gel-shift binding experiments with YxiN and YxRBD. The resulting free energies of binding are shown in Fig. 3. Surprisingly, neither YxiN nor YxRBD showed a sharp binding "boundary" where the affinity drops dramatically within one or two nucleotides. Instead, the energies of binding to both proteins decreased gradually with decreasing 5Ј extension length. For RNAs with more than seven nucleotides of 5Ј single-stranded sequence, binding affinity to YxRBD is the same as to YxiN. However, shorter RNAs show differences in their binding behavior to YxRBD and YxiN (Fig. 3). For the shortest RNAs, HPϩ2 and HPϩ3 (Fig. 1B), binding to full-length YxiN can still be observed, whereas neither RNA shows detectable binding to YxRBD (Fig. 4). The disparity in binding properties of the two proteins suggests that the catalytic domains of YxiN, which are absent in YxRBD, may contribute to the binding of shorter RNAs. However, the addition of more single-stranded residues to the short RNAs increases their binding affinity to both YxRBD and YxiN (Fig. 3). Because the longer RNAs bind more tightly than short RNAs to YxRBD (which lacks the catalytic domains), in this case the additional binding energy likely arises from nonspecific electrostatic interactions of the single-stranded region with the basic C-terminal domain. Thus, because nonspecific interactions of the RNA can occur both with the C-terminal and the catalytic domains, the resulting affinities of YxRBD and YxiN for longer RNAs become similar to each other.
Footprinting of YxiN and YxRBD on RNA-DbpA and YxiN have been shown to interact with a larger region of 23 S rRNA than RNA HPϩ15. Although the hairpin of RNA HPϩ15 (hairpin 92 in 23 S rRNA) is a major affinity and specificity determinant, DbpA and YxiN show increased binding affinity and ATPase stimulation with larger RNAs that also contain helices 89, 90, and 93 in addition to 92 (14,15). DbpA has an extensive footprint on a 172-nt RNA that includes this region (Fig. 5), indicating the presence of additional contacts of DbpA with RNA elements other than hairpin 92 (10).
The YxRBD fragment is useful in assigning the domains of YxiN to specific interactions with such larger RNAs because it can identify the site of binding of the C-terminal domain devoid of the catalytic core. To this end, the footprint of YxRBD on the 172-mer was obtained using RNases T1, T2, and kethoxal and was compared with analogous footprints of YxiN in the presence of AMPPNP. In addition, the footprint of DbpA with AMPPNP on the same RNA was used as a control. The results of the footprinting experiments are shown on the secondary structure of the 172-mer in Fig. 5. All three proteins showed protections from kethoxal at G2553 in hairpin 92, and from RNase T2 at U2652/ A2564 in the bulge between helices 92 and 90. Protection at G2553 was also confirmed by RNase T1 (data not shown). However, only the fulllength DbpA and YxiN protected the G2502/G2505 residues between helices 89 and 90, and the G2588/G2603 residues in helix 93 from RNase T1 and kethoxal (Fig. 5 and data not shown). Footprinting in the bulge of helix 90 (G2574), reported previously for DbpA (10), could not be assigned unambiguously for either YxiN or YxRBD. Overall, full-length YxiN protected essentially the same set of residues as DbpA, whereas the C-terminal fragment only interacted with hairpin 92 and the singlestranded region connecting it to helix 90. This suggests that the helicase domains of the full-length protein were required for interaction with residues 5Ј and 3Ј of helix 90.
ATPase Activity of YxiN and YxCat-The ATPase activity properties of YxiN and its fragments were compared using a coupled ATP hydrolysis assay (Fig. 6 and TABLE ONE). The background ATP hydrolysis rate of YxiN in the absence of RNA was determined to be 0.39 min Ϫ1 (data not shown). Upon addition of the specific RNA HPϩ15 and the nonspecific RNA HPϩ15M and poly(A), YxiN shows a stimulation of ATPase activity with the expected selectivity (Fig. 6A). RNA HPϩ15 elicits full activation of 55 min Ϫ1 at the lowest RNA concentration in this experiment (200 nM). Using a lower range of RNA concentrations, the apparent activation constant of RNA HPϩ15 (K app ) was determined to be 31 nM (TABLE ONE). In contrast, stimulation by RNA HPϩ15M occurs with a much higher K app of 3500 nM and saturates at a lower maximal activity (25 min Ϫ1 ). Poly(A) also activates YxiN weakly, increasing activity linearly up to a concentration of 10 M without reaching saturation (the concentration of poly(A) is expressed in 32-nt units). To gauge the effectiveness of such nonsaturating substrates, the apparent second order rate constant (k max /K app (RNA)) was derived from the initial slope of ATPase activity versus RNA concentration (TABLE ONE). Under these conditions, YxiN exhibits a 250-fold specificity for RNA HPϩ15 over RNA HPϩ15M and a 1000-fold specificity over poly(A). Previous experiments with YxiN at the more stringent salt concentration of 150 mM showed a specificity of 170,000 for RNA HPϩ15 over RNA HPϩ15M (12).
The isolated catalytic core maintains the capacity for RNA-stimulated ATPase activity (Fig. 6B). Similar to full-length YxiN, the RNAindependent ATP hydrolysis rate of YxCat was found to be 0.40 min Ϫ1 (data not shown). Upon addition to 1 M YxCat, RNAs HPϩ15, HPϩ15M, and poly(A) stimulate the ATPase activity linearly with RNA concentration and show similar second order rate constants between 5.4 ϫ 10 Ϫ5 and 6.9 ϫ 10 Ϫ5 min Ϫ1 nM Ϫ1 (Fig. 6B and TABLE ONE). The lack of saturation up to 10 M RNA concentrations may reflect the more acidic pI of YxCat (5.8 compared with 7.7 for YxiN), making it less likely to interact with RNA at the assay pH of 7.5. In fact, titrations with poly(A) remain linear up to 50 M (in 32-nt units), eliciting rates up to 3.5 min Ϫ1 (data not shown). Because saturation of ATPase activity was not attained, the k max of YxCat can only be estimated to be between 3.5 min Ϫ1 and the full-length YxiN value of 55 min Ϫ1 . Assuming this range of k max , the calculated apparent RNA binding affinities (K app ) for RNAs in TABLE ONE are in the range of 65 to 1000 M. Thus, YxCat has very weak apparent RNA binding affinity and no intrinsic specificity among the tested RNAs. Finally, because the C-terminal fragment does not contain any ATPase motifs, it does not hydrolyze ATP on its own, as expected (data not shown).
Two experiments suggest that the ATPase activity of YxCat is functionally independent from the C-terminal fragment. First, the apparent affinity of YxCat for ATP was compared with that of YxiN. The ATPase activity of both proteins was assayed at varying concentrations of ATP⅐Mg with 50 M poly(A) (in 32-nt units) (Fig. 6, C and D).   and YxRBD (gray). Binding of HPϩ2 and HPϩ3 to YxRBD was not detected and is considered to be more than Ϫ6 kcal/mol. amounts of the specific 23 S ϩ 16 S rRNA substrates (11). Thus, YxCat does not require the C-terminal fragment for proper formation of the ATP site and for ATP binding. Second, when 1 M YxRBD was added to the YxCat ATPase assay with either RNA HPϩ15 or HPϩ15M, no additional stimulation or preference for RNA HPϩ15 was observed (data not shown). This indicates that YxRBD is incapable of selectively presenting the specific RNA to YxCat in trans, suggesting a lack of interaction between the catalytic core and the C-terminal domain.

DISCUSSION
The present study demonstrates that the C-terminal region of YxiN is a domain that can independently bind RNA in a sequence-specific manner. Expressed as a 76-amino acid fragment, YxRBD folds well and specifically binds RNAs containing hairpin 92 of 23 S rRNA with low nanomolar affinity (Fig. 2C and D). This domain is similar in size to the currently known RNA-binding domains of the RRM (RNA recognition motif), dsRBD (double-stranded RBD), and OB (oligonucleotide/ oligosaccharide binding)-fold families, as well as to the small, singledomain ribosomal proteins. Although assignment to the OB-fold family cannot be made based on sequence (16), alignments of YxiN, DbpA and other bacterial homologs (Fig. 1A) indicate that the C-terminal domain defines a distinct subfamily of RNA-binding domains. However, in a Pfam search, a few members of the subfamily show very weak similarity with the eukaryotic RRM family (expectation values, Ͼ0.05) in a 20-amino acid region starting at RNP-1 (17). A closer examination of a possible alignment indicates that half of the six most conserved positions of the RRM (Leu-16, Val-38, and Ala-49, numbering by Birney et al. (18)) have analogous positions in the DbpA C-terminal domain, whereas the other half does not (Leu-7, Phe-20, and Phe-40) (18). Thus, the precise classification of this sequence specific RNA-binding domain awaits further studies.
Aside from orthologs of DbpA, there is a highly similar, but distinct, C-terminal domain present in E. coli DeaD/CsdA, a DEx D / H protein involved in the ribosomal biogenesis of the 50 S subunit (19). In CsdA and its orthologs, the domain is located significantly further away from the helicase core and is flanked on both sides by arginine-and glycinerich sequences that are known to accompany other RNA-binding domains (20). The C-terminal domain in CsdA homologs differs from the YxiN/DbpA RBD in several conserved sequence positions (the two groups are defined together in Pfam entry PF03880). Interestingly, CsdA does not exhibit any detectable in vitro specificity for hairpin 92, and its ATPase activity is stimulated by several RNAs (21). 3 These results suggest that CsdA either lacks RNA specificity or recognizes a relatively common RNA motif.
Characterization of the YxCat fragment indicates that it possesses the 3 K. Kossen and O. Uhlenbeck, unpublished results.    (26). The individual domains of YxiN appear to function in a modular fashion. The C-terminal domain provides the bulk of the binding energy as well as sequence specificity, whereas the helicase core performs the catalytic functions. The two domains do not require each other for activity, and there appears to be no interaction between them, because the addition of YxRBD to YxCat in trans does not increase the ATPase activity or impart specificity. Thus, the C-terminal domain simply brings the catalytic core to its cellular target. Such a modular design strategy may be common for DEx D / H proteins with ancillary domains. For example, modularity in the yeast splicing DEx D / H protein Prp16 was established by testing domain deletion mutants of Prp16 for viability in  vivo and interactions with the spliceosome in vitro (27). The N-terminal ancillary domain was found to interact with the spliceosome and act as a dominant negative inhibitor of splicing in vitro. Thus, the N-terminal domain functions to bring the protein to the spliceosome. However, in contrast with YxiN, the N-terminal domain of Prp16 in trans with the other domains of Prp16 can rescue a strain devoid of Prp16, indicating that the N-terminal domain interacts with the rest of the protein.
The modular nature of YxiN sheds light on how it interacts with its larger, more biologically relevant 172-mer substrate, as well as potential modes of action on the cellular substrate of YxiN. The footprinting pattern of YxRBD indicates that this domain interacts sequence-specifically with hairpin 92 and the single-stranded sequence connecting it to helix 90, tethering the catalytic core to this location (Fig. 5). The absence of RNA specificity of YxCat indicates that the full-length enzyme has the ability to act on any suitable substrate in the vicinity of its molecular localization. Accordingly, within the 172-mer RNA, the catalytic core interacts with regions encompassing positions 2502 and 2505 5Ј of helix 90, as well as 2588 and 2603 3Ј of helix 90. The flexibility of the catalytic core relative to the tethering point of the C-terminal domain, reflected in the ability of DbpA to unwind duplexes in several locations around helix 92 (28), suggests that YxiN and DbpA may interact with the regions 5Ј and 3Ј of helix 90 in two distinct binding populations. The function of YxiN and DbpA may be to unwind an RNA duplex or rearrange a ribonucleoprotein structure at one or both of these locations during the ribosomal lifecycle. Recently, DbpA has been shown to unwind duplexes in a 3Ј-to-5Ј direction, regardless of the orientation of the duplex relative to hairpin 92, and to require a 3Ј single-stranded loading site, presumably for the binding of the catalytic core (29). A similar requirement of a 3Ј single-stranded RNA loading site may exist for its natural targets. In the context of the entire cellular substrate of YxiN, the 50 S ribosomal subunit or its assembly precursor, the catalytic domains of YxiN may act at the same site(s) as in the 172-mer or rearrange a different target or targets near hairpin 92. Rearrangement of several interactions near hairpin 92, an area crucial to ribosome func-tion, may be required for proper RNA folding and establishment of correct interactions with ribosomal proteins.