RNA passes through the hole of the protein hexamer in the complex with the Escherichia coli Rho factor.

Escherichia coli transcription termination factor Rho is a ring-shaped hexameric protein that uses the energy derived from ATP hydrolysis to dissociate RNA transcripts from the ternary elongation complex. To test a current model for the interaction of Rho with RNA, three derivatives of Rho were made containing single cysteine residues and modified with a photo-activable cross-linker. The positions for the cysteines were: 1) in part of the primary RNA-binding site in the N terminus (Cys-82 Rho); 2) in a connecting polypeptide proposed to be on the outside of the hexamer (Cys-153 Rho); and 3) near the proposed secondary RNA-binding site in the ATP-binding domain (Cys-325 Rho). Results from the cross-linking of the modified Rho proteins to a series of lambda cro RNA derivatives showed that Cys-82 Rho formed cross-links with all transcripts containing the Rho utilization (rut) site, that Cys-325 Rho formed cross-links to transcripts that had the rut site and 10 or more residues 3' of the rut site, and that Cys-153 did not form cross-links with any of the transcripts. From a model of the quaternary structure of Rho, which is largely based on homology to the F(1)-ATPase, amino acid 82 is located near the top of the hexamer, and amino acid 325 is located on a solvent-accessible loop in the center of the hexamer. These data are consistent with binding of the rut region of RNA around the crown, with its 3'-segment passing through the center of the Rho hexamer.

Transcription termination factor Rho in Escherichia coli consists of six identical protein subunits arranged in a ring structure (1,2). For its function in termination, Rho binds to the nascent transcript and acts to dissociate the transcript from RNA polymerase and the DNA template (3,4). Rho can also act as a helicase to dissociate a DNA molecule that is base paired to a 3Ј-segment of an RNA with a Rho attachment site (5,6). The motive power for these actions comes from the hydrolysis of ATP to ADP and P i . The mechanism for coupling ATP hydrolysis to these dissociation reactions is not known.
The Rho polypeptide contains two major domains: an Nterminal RNA-binding domain (residues 1-130) and a C-terminal ATP-binding domain (residues 131-419) (7,8). The RNAbinding domain contains an oligosaccharide/oligonucleotidebinding domain (OB-fold) (9). Rho forms strong binding interactions with single-stranded C-rich RNA molecules, and RNA segments with these characteristics form the attachment sites (known as the rut or Rho utilization sites) used by Rho to mediate termination of a transcript. The structure of a complex of the RNA-binding domain of Rho with oligo(rC) 9 , determined by x-ray crystallography, shows that a cytidylate residue forms H bonds with Arg-66 and Glu-78 and a -bond stacking interaction with Phe-64 (10).
Evidence for a second RNA-binding site distinct from that in the RNA-binding domain has come from a detailed analysis of the polynucleotide requirements for activation of ATP hydrolysis by Rho (11) and from the finding that mutant forms of Rho with changes in certain residues in the ATP-binding domain are defective in their interactions with RNA that are coupled to ATP hydrolysis (12)(13)(14).
Large parts of the Rho polypeptide are similar to the corresponding parts of the ␣ and ␤ subunits of the F 1 -ATPase (15). In addition, the hexameric Rho protein is morphologically similar to the ␣ 3 ␤ 3 sub-component of that ATPase (1,2,16). These similarities have led to proposed models of a tertiary structure for the ATP-binding domain of the Rho polypeptide and a quaternary structure of the hexamer, both based on the structure of the bovine mitochondrial F 1 -ATPase (13, 16 -18). A striking feature of these models is that mutational changes in Rho that affect secondary RNA-binding site interactions are found primarily in residues that face into the center of the hexameric hole (13). These observations have led to the proposal of a model for the interactions of Rho with RNA that cause transcript termination (19). In this model, the nascent transcript first binds to an extensive site composed of the RNA-binding domains on the six subunits arranged around one pole of the hexamer. This binding occurs in a way that allows the RNA that is on the 3Ј-side of the attachment segment to pass through the hole of the hexameric ring, perhaps by assembly of the hexamer on the RNA (20). Once bound, concerted cycles of ATP binding, hydrolysis, and release can cause structural motions in the hole of the hexamer that could act to pull the transcript through and away from its biosynthetic complex with RNA polymerase.
Previously, Rho was identified as being related to a subclass of DNA helicases (superfamilies IV and V) (21,22). All the proteins in this group contain seven common sequence motifs, possess a hexameric quaternary structure, and couple NTP hydrolysis to 5Ј-to 3Ј-translocation along a nucleic acid. Biochemical evidence obtained with two members of this subclass, DnaB (23) and the T7 gene 4 product (24), indicate that they migrate along single-stranded DNA by pulling the DNA through the center of their respective hexameric ring structures. Because Rho is structurally related to these proteins, it could readily have a similar mechanism for nucleic acid translocation.
To probe the points of interaction of Rho with RNA, several mutational derivatives of Rho-containing single cysteine residues were prepared. The positions of the cysteines were chosen based on the known structure of the RNA-binding domain and the proposed structural model of the hexameric ATP-binding domains of Rho (17). The mutants were then modified with the cysteine-specific, photo-activable cross-linker N- [4-(p-azidosalicylamidobutyl]-3Ј-(2Ј-pyridyldithio)propionamide (APDP), 1 and the interactions of these modified proteins with various forms of cro RNA was analyzed by measuring yields of crosslinked complexes.

EXPERIMENTAL PROCEDURES
Materials-All restriction enzymes, T4 DNA ligase, T4 RNA ligase, T4 polynucleotide kinase, and T4 DNA polymerase were purchased from New England BioLabs Inc. APDP and 5,5Ј-dithiobis(2-nitrobenzoate) (DTNB) were purchased from Pierce Chemicals. The oligonucleotides were purchased from either Life Technologies, Inc. or Integrated DNA Technologies. All deoxyribonucleotides, ribonucleotides, and Proteinase K were purchased from Roche Molecular Biochemicals. Radioactive nucleotides were purchased from ICN Radiochemicals. Bio-Rex 70 resin was purchased from Bio-Rad. T7 RNA polymerase was provided by Lislott Richardson (Indiana University). MGP1-2 helper phage was a gift from Stanley Tabor (Havard University Medical School). The p39-AS plasmid containing the C202S mutation was a gift from Terry Platt (University of Rochester). pA4 and pB3 were gifts from Bill Scott (University of California, Santa Cruz).
Construction of rho Variants-pCB111, a plasmid containing the E. coli rho gene under control of the T7 promoter (25), and a p39-AS derivative, containing the E155K and C202S mutations in the rho gene (26), were partially digested with PstI. The resulting 5926-base pair fragment from pCB111 and the 567-base pair fragment from the p39-AS derivative were ligated overnight at 17°C with T4 DNA ligase. The resulting plasmid, pBB-1, contains rho with a C202S mutation. This and all subsequent mutations were confirmed by sequence analysis.
To create a plasmid that can be used to express C202S,S325C Rho (Cys-325 Rho), pBB-1 was transformed into competent CJ236 (dut Ϫ , ung Ϫ ) cells, and these cells were infected with helper phage M13KO7. The uracil-containing single-stranded DNA template was recovered by polyethylene glycol precipitation followed by phenol/chloroform (1:1) extraction, ethanol precipitation, and resuspension in water. The mutagenic primer (3Ј-CGATACCGGTTGTAAAATGGACG-5Ј) was annealed to the purified uracil-containing single-stranded DNA, followed by addition of 1 unit of T4 DNA polymerase, 3 units of T4 DNA ligase, 0.4 mM each of dGTP, dCTP, dATP, and dTTP, and 0.75 mM ATP, which completed synthesis of the second strand. The newly synthesized DNA was transformed into the dut ϩ , ung ϩ strain DH5␣FЈ, and the cells were plated on LB-ampicillin (50 mg/ml) plates and grown overnight at 37°C. Plasmid DNA was isolated from several of the plate colonies via standard plasmid mini-prep procedure. A 341-base pair BstXI to KpnI restriction fragment from a plasmid with the desired mutation was ligated to the 6152-base pair BstXI to KpnI restriction fragment from the parental pBB-1 plasmid. The resulting plasmid, pBB-2, contained the rho gene with both the C202S and S325C mutations.
Expression and Purification of the Mutant Rho Proteins-DH5␣FЈ cells containing one of the mutant plasmid rho genes were infected with the M13 phage derivative mGP1-2 (27). mGP1-2 contains a gene encoding T7 RNA polymerase under control of the E. coli lac promotor. Production of high levels of mutant Rho protein was achieved by induction of a 500-ml LB culture at an A 600 nm ϭ 0.6 with 1.0 mM isopropylthiogalactoside and mGP1-2 (multiplicity of infection of 20 -30). The culture was grown at 37°C for 3 h after induction. The cells were harvested by centrifugation at 7000 rpm for 20 min in a Sorvall RC2-B centrifuge (GS3 rotor). The cells were washed with ice-cold STE (10 mM Tris⅐HCl, pH 8.0, 0.5 mM EDTA, pH 8.0, and 100 mM NaCl), pelleted as above, and stored at Ϫ20°C.
The mutant Rho proteins were purified as described by Nowatzke et al. (28), with one modification. The purification of the protein was stopped after the Bio-Rex 70 column. Analysis of Coomassie Bluestained SDS-polyacrylamide gels indicated that the protein(s) were greater than 95% pure. Proteins were kept in Rho storage buffer (10 mM Tris⅐HCl, pH 8.0, 0.1 M EDTA, pH 8.0, 0.1 mM dithiothreitol, 0.1 M KCl, and 50% glycerol) and stored at Ϫ20°C. Protein concentrations were determined by the Bradford assay (29).
Modification of Mutant Rhos with APDP-The modification reactions were done using 1:1 molar mixtures of the Cys-0 Rho to Cys-82 Rho, Cys-153 Rho, and the Cys-325 Rho mutant Rho polypeptides. A 100-l reaction containing 17 M (monomer) of both the Cys-0 Rho and one of the mutant Rho polypeptides were mixed in TGES buffer (50 mM Tris⅐HCl, pH 8.0, 5% glycerol, 1 mM EDTA, and 50 mM KCl) and allowed to equilibrated on ice for 30 min. This 30-min incubation allows for formation of mixed hexamers (30). The mixed hexamer solutions were then spun through a 1-ml Sephadex G-50 spin column equilibrated in TGES buffer to remove residual dithiothreitol. The eluate from the G-50 column was then mixed with 10 l of a 50 mM APDP solution in 100% dimethyl sulfoxide and incubated at 22°C for 3 h in the dark. The unreacted APDP was removed by centrifugation of the solution through a 1-ml Sephadex G-50 spin column. The protein concentration was determined by Bradford assay. The modified protein was stored at Ϫ80°C and was active for up to a month.
Quantification of APDP Modification-Reaction of free sulfhydryl groups with 5,5Ј-dithiobis(2-nitrobenzoate) (DTNB) results in the release of the 2-nitro-5-thiobenzoate anion, which can be detected by its characteristic absorbance at 412 nm (molar absorptivity of 13,600 M Ϫ1 cm Ϫ1 ) (31). The various proteins were modified with APDP as above, with the exception that no Cys-0 Rho protein was added to the cysteinecontaining mutants. The reactions were done in a total volume of 105 l. Protein solutions containing at least 25 g of cysteine-containing residues (final Cys residue concentration of 3.5 M or higher) in a volume Յ 40 l was filled to 50 l with TGES buffer. 50 l of detection buffer (0.1 M sodium phosphate, 1.34 mM EDTA, and 6 M guanidine hydrochloride, pH 8.0) and 5.0 l of freshly prepared DTNB buffer (10.1 mM DTNB, and 0.1 M sodium phosphate, pH 8.0) were added and mixed. The solution was incubated at 22°C for 25 min. After 25 min, the absorbance of the solution was measured at 412 nm, and the concentration of free sulfhydryl was determined using the Beer-Lambert law.
ATPase Assay-For values in Table III (see below), the assay was performed as described previously by Nowatzke and Richardson (32) with the modification that 20 ng of Rho protein was used per 100-l reaction. For values in Table IV (see below), the ATPase reaction buffer was TGES buffer, 2 mM MgCl 2 , 1 mM ATP, 3.6 nM wild type Rho hexamer, and 10 nM RNA (100-l reaction volume). This buffer was chosen to more closely resemble the buffer conditions of the crosslinking experiments. The concentration of ATP used for these assays was one-fifth the concentration used in the cross-linking studies due to the high background in the colorimetric assay when using 5 mM ATP, thus the MgCl 2 concentration was adjusted to one-fifth (2 mM) the amount used in the cross-linking studies.
Preparation of cro Transcripts-Plasmid pIF2 was described by Faus and Richardson (33), and pB3 was described by Gan and Richardson (20). Plasmid pA4, constructed by Dr. William Scott, contained DNA base pairs 210 -362 of the cro gene flanked by sequences encoding hammerhead ribozymes between the XbaI and HinDIII restriction sites in pUC18. The sequence of the A4 insert (underlined) and the flanking hammerheads is as follows: 5Ј-CTAGAATGCTACTGATGAG-GTTCGCCGAAACGTTCGCGTCTAGCATAAATAACCCCGCTCTTAC-ACATTCCAGCCCTGAAAAAGGGCATCAAATTAAACCACACCTATG-GTGTATGCATTTATTTGCATACATTCAATCAATTGTTATCTAAGG-AAATACTTACATATGGTTCGTGCAAACAAACGCAACGTCTACCGA-AAGGTACTGATGAGGTTCGCCGAAACGTTGCGTTA-3Ј. The plasmid pA4 was maintained in E. coli DH5␣FЈ.
DNA templates were prepared for T7 transcription by linearization of the plasmid with a restriction enzyme as indicated. After digestion with the appropriate restriction enzyme, 25 g of Proteinase K and 0.2 volume of 5ϫ Proteinase K buffer (50 mM Tris-HCl, pH 7.8, 25 mM EDTA, and 5% SDS) were added and incubated for 90 min at 37°C. The reaction mixture was then treated with an equal volume of phenol, then treated with an equal volume of chloroform:iso-amyl alcohol (24:1), and ethanol-precipitated. The DNA was washed with 70% ethanol, dried, and resuspended in water. The DNA concentration was determined by absorbance at 260 nm.
T7 transcription was carried out in a 100-l mixture containing 5 pmol of linearized DNA, 40 mM Tris-HCl, pH 8.1, 30 mM MgCl 2 , 2 mM spermidine, 4 mM each of ATP, CTP, GTP, and UTP, 0.02 N NaOH, and 8 g of T7 RNA polymerase. The reaction mixture was incubated at 37°C for 2 h. The RNA was precipitated by addition of 0.1 volume of 8 M LiCl and 3 volumes of 100% ethanol, followed by 30 min at Ϫ80°C. The RNA was pelleted by centrifugation at 13,000 rpm for 30 min at 4°C, washed with 150 l of 70% ethanol, and recentrifuged. The pellet was dried and subsequently resuspended in 20 l of formamide loading dye (98% formamide, 2 mM EDTA, 0.03% (w/v) bromphenol blue, and 0.03% (w/v) xylene cyanol). The resuspended RNA was loaded onto a 6%/7 M urea polyacrylamide gel (acrylamide:bis, 19:1) containing 1ϫ TBE buffer (45 mM Tris borate, and 1 mM EDTA) and run in 1ϫ TBE buffer at 30 watts. The RNA was detected by UV shadowing, excised using a sterile razor blade, and placed into a 1.5-ml Eppendorf tube containing 200 l of phenol, 200 l of chloroform:iso-amyl alcohol (24:1), and 250 l of passive elution buffer (50 mM Tris-HCl, pH 8.0, 0.3 M sodium acetate, and 1 mM EDTA). The gel slice was incubated for 10 -12 h at 4°C. The RNA was recovered by removing the aqueous layer, performing two chloroform extractions, followed by ethanol precipitation and 70% ethanol wash. The dried pellet was resuspended in 25 l of TE buffer (10 mM Tris-HCl, pH 8.0, and 1 mM EDTA), and stored at Ϫ80°C. The concentration of the RNA was determined by its absorbance at 260 nm.
To label the 5Ј-end of the T7 transcripts from the linearized pB3 and pA4 plasmids, which have a free 5Ј-hydroxyl as a result of cleavage with the 5Ј-hammerhead, 1 M RNA was mixed with 10 units of T4 polynucleotide kinase and 2.8 M [␥-32 P]ATP (7000 Ci/mmol) in 50 l of PNK buffer (70 mM Tris-HCl, pH 7.6, 10 mM MgCl 2 , and 5 mM dithiothreitol), and the mixture was incubated at 37°C for 90 min. The RNA was precipitated with ethanol, washed, and resuspended in 20 l of formamide loading dye. The RNA was loaded onto a 6%/7 M urea polyacrylamide gel and run at 30 watts in 1ϫ TBE buffer. The RNA was isolated as above. All RNAs were stored in TE buffer at Ϫ80°C (see Fig. 1).
Cross-linking of APDP-modified Rho to [ 32 P]RNA-The APDP-modified Rho proteins were diluted to 10 ng/l in Rho dilution buffer (40 mM Tris-HCl, pH 7.6, 0.1% acetylated bovine serum albumin, 50 mM KCl, 5 mM MgCl 2 , 0.1% Nonidet P-40). The cross-linking reactions, prepared under subdued light, contained 20 ng of modified Rho, 10 nM 32 P-labeled RNA, and 5 mM adenine nucleotide (if present) in 20 l of TGES buffer with 10 mM MgCl 2 and were incubated for 2 min at 22°C. The samples were then placed onto a piece of Parafilm, directly under a mid-range UV 302-nm lamp (Model UVM-57, UVP Products). A polystyrene filter was placed over the samples during irradiation; this helped to filter out shorter wavelength UV light, which can induce nonspecific proteinnucleic acid cross-linking. The samples were irradiated for 2 min at 4°C. The samples were then mixed with 0.2 volume of SDS loading dye (150 mM Tris-HCl, pH 6.8, 4% SDS, 0.2% bromphenol blue, and 20% glycerol), and run on a 10% SDS-polyacrylamide gel (acrylamide:bis, 29:1) at 25 mA in Tris-glycine buffer (25 mM Tris, 250 mM glycine, and 0.1% SDS) (34). Rho-RNA cross-linked species were detected by using a 20-ϫ 25-cm Storage Phosphor Screen and PhosphorImager (Molecular Dynamics). The computer program ImageQuaNT (Molecular Dynamics) was used to quantify the cross-linked products.
Molecular Modeling-Coordinates for the N-terminal 125 amino acids of Rho were downloaded from the Brookhaven Protein Data Bank, accession code 1A62 (9). The coordinates for the C-terminal region (amino acids 126 -419) of Rho were downloaded from the web site of the Department of Biotechnology, University of Virginia (17). Modeling of the Rho monomer was performed using the InsightII visualizer software obtained from Molecular Simulations, Inc.

Properties of Rho Factors with Cross-linking Groups-Vari-
ants of Rho protein containing single cysteine residues at different positions were created by site-directed mutagenesis. Wild-type Rho contains a single cysteine at position 202. Dombroski and Platt (26) showed that it could be changed to a serine or an alanine without affecting the function of Rho. Starting with a rho gene containing the C202S mutation, derivatives with single cysteines at other positions were made. The positions were chosen based on the proposed quaternary structure of E. coli Rho (Fig. 2) (17). All the residues changed were predicted to be solvent-exposed serines. The changes were thus expected to have minimal perturbation of wild-type function and allow for a high potential for modification with a cysteine-reactive cross-linking moiety. Position 325 is in a section called the R loop and is believed to be near a proposed secondary RNA-binding site located in the center of the Rho hexamer (13). Position 82 is in close proximity to the RNAbinding cleft in the RNA-binding domain (9,10), and position 153 is a residue on a large loop on the outside of the hexamer (17).
The mutant proteins were purified and tested for retention of normal function by measuring their ATPase activity with cro mRNA. Mutant Rhos that are defective in transcript termination have very reduced ATPase activity with this natural Rhoterminated transcript (35). The results (Table I) show that the specific activities of the pure Cys-82 and Cys-325 Rhos were about half that of the wild-type Rho, whereas the Cys-153 Rho had about normal activity. With poly(C), a super activator for Rho ATPase, Cys-82 had nearly the same activity as wild-type Rho, whereas the Cys-325 still had reduced activity. Thus, changes of residues that are believed to be near or at parts of the protein that interact with RNA do affect the functional interaction of Rho with a test transcript. However, the changes did not eliminate function, which means that the mutants are suitable for probing functional interactions.
The mutant forms of Rho with single cysteine residues were reacted with APDP to make derivatives containing the photoactivable azidosalicylamido group. To determine the extent of modification, the amounts of free sulfhydryl in the protein preparations before and after APDP modification were deter- The rut site region is indicated with dark shading. The asterisk denotes the end of the RNA that was labeled with 32 P. All 3Ј-end-labeled transcripts had a nontemplate C at their 3Ј-ends. mined by reaction with dithiobis(nitrobenzoate) (Ellman's reagent). The results (Table II) show that at least 25% of the Cys residues in the Cys-153 and Cys-325 Rhos were modified and 50% of the Cys residues in Cys-82 Rho were modified. The Cys at residue 202 of wild-type Rho was not substantially modified with APDP. Hence, wild-type Rho was not suitable for use in this cross-linking study.
To reduce the chance that multiple cross-linking groups on a single Rho hexamer might interfere with normal interactions, the fractionally modified derivatives were mixed at a 1:1 ratio with the Cys-0 Rho under conditions that allow exchange of subunits (30). These modified, mixed proteins were then tested for retention of function by measuring their abilities to hydrolyze ATP with cro mRNA and poly(C) ( Table I). The proteins had at least one-third the activity of the unmodified wild-type Rho, indicating that they all retained partial function in their interactions with RNA.
Cross-linking of APDP-modified Rho Factors to the cro Derivative, cro(216 -365)-For the initial studies on the formation of cross-linked complexes with the Rho derivatives, a form of cro RNA called cro(216 -365) was chosen, because it contains both the rut site and sequence 3Ј of rut. Previous filter binding experiments showed that Rho protein bound as well to this RNA as to the full-length cro transcript. 2 Complexes of this RNA with APDP-modified Cys-82 Rho were irradiated with UV light and separated on a 10% polyacrylamide gel in the presence of sodium dodecyl sulfate. Fig. 3 shows that irradiation of the complex for 1 min caused the appearance of 3.7% of the RNA in slower-migrating complexes. Because irradiation for 2 min increased the yield only slightly, the cross-linking reaction is nearly complete by 1 min. This was confirmed by findings (not shown) that longer irradiations did not increase the yield over that with the 2-min incubation. The mobility shift was dependent upon irradiation of APDP-modified Rho protein (Fig. 3, lane 1) and was eliminated by treatment of an irradiated sample with an excess of a reducing thiol, dithiothreitol (Fig. 3, lane 2) prior to electrophoresis, confirming that the linkage is through a disulfide. Irradiation of unmodified, Cys-82-mixed hexamers failed to produce shifted RNA (data not shown). Lastly, irradiation of the RNA alone did not affect its mobility (Fig. 3, lane 5), indicating that the shifted bands are not due to a UV-induced change in the RNA structure itself. Overall, these experiments suggest that the observed shift in radiolabeled RNA is an azide-specific cross-link between RNA and the APDP-modified Rho protein.
The reason for the differing mobilities of cross-linked species is not known. Because the rut segment of the RNA is believed to bind across several subunits of the hexameric Rho, subunits cross-linked at different positions on the RNA would have different structures that might have differing mobilities. This point will be considered in more detail later, because the appearance of multiple bands depends on the RNA used and the position of the cross-linker group in the protein.
To investigate the topological interactions between Rho and RNA, the modified mutant Rho proteins with Cys positioned at different locations on the hexameric structure were crosslinked to cro(216 -365). Fig. 4 shows the mobility-shift evidence for cross-linking in the absence and presence of ATP for the APDP-modified Rho proteins. Lanes 1 and 2 show that the addition of ATP to the complexes greatly enhances the percentage cross-linking with modified Cys-82 Rho (from 0.99 to 4.62%). Lanes 3 and 4 show that very little RNA was crosslinked to the modified Cys-153 Rho in the absence of ATP (0.13%) with even less (Ͻ0.06%) in the presence of ATP. Finally, lanes 6 and 7 show that cro(216 -365) RNA formed crosslinks with modified Cys-325 Rho almost as effectively as to the modified Cys-82 Rho in the presence of ATP. However, in this case the presence of ATP had no effect on the signal. Thus, this evidence indicates that this form of cro RNA does bind to Rho in a way that brings it in close proximity to an amino acid residue on the R loop.
To further demonstrate that the formation of the crosslinked complexes is due to specific Rho-RNA interactions, the effect of the inclusion of various competitor RNAs on the formation of shifted complexes with labeled cro(1-380) RNA was tested (Table III). When poly(C), which has a very high affinity for Rho, was added to the reaction mixture, it eliminated the formation of complexes with modified forms of either Cys-82 or Cys-325. On the other hand, when poly(A), which binds with very low affinity and does not activate ATP hydrolysis, was present, the yields of cross-linking were not greatly affected. These results are further evidence that the formation of shifted complexes is due to specific interactions between Rho and cro .
Requirement for RNA Sequence 3Ј of the cro rut Site for Cross-linking to Cys-325 Rho-To investigate the role of the rut site, the sequences 3Ј and 5Ј of rut and the overall RNA length on the yields of cross-linking with Rho factors modified either at residue 82 or 325, several cro transcripts were made (see Fig. 1) and tested for their ability to form cross-linked products (Table IV). First, cro(1-86) RNA, a cro segment that lacks the rut site, was tested. Although a transcript of the same size carrying the rut sequence binds to Rho with a K d of ϳ0.6 nM (in 0.15 M potassium glutamate with ATP present (20)), cro(1-86) forms complexes that can barely be detected in the filter binding assay (K d Ͼ 50 nM). cro(1-86) gave no detectable crosslinked products with either of the modified Rhos (Table IV). Thus, the formation of cross-links with RNA correlates with the ability of Rho to make a highly stable complex with the RNA.
Second, a 61-nucleotide cro RNA segment (cro(216 -276) RNA) consisting of the rut sequence but ending at residue 276, which is about the end of the rut sequence, was tested. It binds Rho with high affinity (K d ϳ 0.6 nM) and forms cross-links with Rho when the APDP group was attached to residue Cys-82 in the RNA-binding site but not when the APDP group was on residue 325 (Fig. 5A). Thus, this RNA appears to be in the 2 E. Gan, unpublished results. primary RNA-binding site but does not extend into the range of a group attached to a residue in the putative secondary RNAbinding site. This result should be compared with those with cro(216 -365), which formed cross-links with Cys-325 Rho almost as well as with Cys-82 Rho, when ATP was present ( Fig.  4 and Table III).
Other cro RNA fragments starting at residue 216 but ending at positions between 276 and 365 also formed cross-links with APDP-modified Cys-325 Rho. The relative abilities of these RNAs to form cross-links with modified Cys-325 Rho to that with modified Cys-82 Rho are presented in Table IV as ratios of Cys-325/Cys-82 Rho. Although these intermediate-sized cro RNA fragments were not as effective as the cro(216 -365) in forming cross-links with Cys-325 Rho, they were considerably more effective than the cro(216 -276) RNA. To test whether the presence of extra RNA sequences 5Ј of the rut were as effective as the extra sequences on the 3Ј-side of the rut sequence in enhancing an interaction with residue 325, an RNA containing the first 276 nucleotides of the wild-type cro RNA was tested. This RNA formed cross-links to the Cys-325 Rho less than 9% as well as with Cys-82 Rho. Thus, extra sequences on the 5Ј-side of rut were not as effective as extra sequences on the 3Ј-side in allowing interactions with residue 325 in Rho.
A normalization procedure was necessary to compare the relative efficiency of cross-linking to the two forms of Rho, because the efficiency of cross-linking even to the modified Cys-82 Rho varied considerably from one RNA to another with a general trend of producing higher efficiencies as the length of the RNAs increased. This observed increase is likely due to two factors. First, among the RNAs containing the rut sequence, the larger RNAs bind Rho with a higher affinity (20,33) than the smaller RNAs. Second, although cro RNA has only one known rut site, more than one Rho might be binding to the larger RNA molecules. Evidence for binding of more than one Rho to an RNA containing only one rut-site was obtained with transcripts containing the trptЈ rut sequences (6). The crosslinking of more than one Rho subunit per RNA may also explain the appearance of products with very different mobilities on the larger RNAs (see Fig. 5B for example).
Previous studies on the requirements for various segments of cro RNA for activation of ATP hydrolysis by Rho (36) had shown that both the rut segment and sequences on the 3Ј-side of rut were necessary for strong activation. To determine whether the sequences on the 3Ј-side that are necessary for cross-linking to residue Cys-325 in Rho are similar to those    4. Photo-activable cross-linking of cro(216 -355) to APDPmodified Cys-82 Rho, Cys-153 Rho, and Cys-325 Rho. 3.6 nM Rho hexamer was mixed with 10 nM cro(216 -365) and incubated in the absence (lanes 1, 3, and 6) or presence (lanes 2, 4, and 7) of 5 mM ATP. All samples were irradiated with 302-nm light for 2 min at 4°C. The samples were separated and analyzed as described in Fig. 3. necessary for activation of ATP hydrolysis, we measured the ATPase activity of Rho with most of the RNAs used in this study (Table IV). The results confirm that cro transcripts with rut and extensive sequences to the 3Ј of rut (cro(1-380) and cro(216 -365)) activated ATPase very effectively, whereas the RNA lacking rut (cro(1-86)) did not. The two RNAs that crosslinked to APDP-modified Cys-82 Rho but not as well to the modified Cys-325 Rho (cro(1-276), and cro(216 -276)) gave reduced ATPase activity, as expected. However, two of the shorter RNAs with longer 3Ј-segments than cro(216 -276) gave very low levels of ATPase even though they cross-linked moderately well with modified Cys-325 Rho. Hence, close proximity of that 3Ј-segment of RNA to residue 325 is not sufficient for activating Rho ATPase. Because both cro(216 -313) and cro(216 -388) have the potential to form stable stem structures at their 3Ј-end (data not shown), those structures could prevent the interactions that lead to ATP hydrolysis but allow the RNA to bind in a way for it to cross-link to a group on residue 325. Effect of Various Adenine Nucleotides on Cross-linking Yield-To further investigate the role of nucleotide binding on the efficiencies of cross-linking with Cys-82 Rho and Cys-325 Rho, yields of cross-linking were measured for all the cro derivatives in the absence of nucleotide or in the presence of ATP, ADP, and the nonhydrolyzable ATP analog AMP-PNP. These nucleotides were chosen to determine whether the effects seen were due to hydrolysis of ATP, or just binding of the nucleotide to Rho. Some examples of the various results are presented in Fig. 5. In general, all three nucleotides increased the efficiency of cross-linking with the modified Cys-82 Rho. However, ATP was more effective than AMP-PNP, which was more effective than ADP. This could simply be due to the difference in K d values for the binding of these nucleotides to Rho (0.2 M for ATP, 10 M for AMP-PNP, and 23 M for ADP (37)). However, all nucleotide concentrations were well above the K d values (5 mM nucleotide, 10 mM MgCl 2 ). Thus, the effect is likely due to slightly different Rho conformations in the presence of the different nucleotides. The fact that none of the nucleotides had any significant effect on the yield of cross-linking with the modified Cys-325 Rho is surprising, because that residue is presumably on a loop that is not far from the ATP-binding site, whereas Cys-82 is in a separate domain of the protein.

DISCUSSION
The cross-linking results presented here support a model in which the rut site of cro RNA binds to clefts in the RNA-binding domains arranged around the crown of the hexamer and has the sequence on the 3Ј-side of rut passing through the hole in the center of the hexamer (Fig. 6). When the cross-linking group was placed in an accessible position in the cleft in the RNA-binding domain, all cro derivatives that could bind to Rho with high affinity formed cross-links. However, when the group was placed in the center of the hole of the Rho hexamer, the ability of those RNA molecules to form cross-links depended as well on having an extension on the 3Ј-side of the rut site. In contrast, a cro RNA with a rut site and a 3Ј-segment failed to cross-link efficiently with a group placed in an equatorial position on the outside of the ring. Because the 3Ј-extension on the RNA is necessary for activation of ATP hydrolysis (Table IV  and Ref. 36), this result suggests that the interactions coupled to ATP hydrolysis occur in the center of the hexamer rather than around the outside. Although only one site in the Rho polypeptide was used for attaching APDP groups on the outside, the length of that group from the disulfide to the first nitrogen in the azido group is 21 Å. Hence, this group should be able to form a cross-link with RNA if the RNA had been present within 40-Å wide circles placed in positions around the equator of the hexamer. The extent of reach of the group is indicated on Fig. 2. Thus, the results fail to support a fairly stringent test of a model that has an ATP hydrolysis-dependent wrapping of RNA around the outside.
In the structural model of Rho, each Cys residue that was created by mutagenesis was predicted to be solvent-exposed, whereas the Cys at position 202 in wild-type Rho was predicted to be partially buried. These predictions were confirmed by the relative reactivity of each of these groups; the three created by mutagenesis all reacted readily with the bulky APDP reagent, whereas the Cys at position 202 in wild-type Rho did not react well with that reagent. Hence, these results support the validity of the structural model.
At the concentration used in these experiments (ϳ1 g/ml), Rho is primarily monomeric in the absence of ATP and hexam-  eric in the presence of ATP and other adenine nucleotides (20,39,46). This state of oligomerization was nearly the same in the presence of the 61-nucleotide rut segment, cro(216 -276) (20). Gan and Richardson (20) found from cosedimentation studies that Rho monomers could bind to cro(216 -276), but that the binding affinity increased greatly under conditions when nucleotides enhanced hexamer formation. Thus, the low level of APDP-modified Cys-82 Rho that formed cross-links to cro(216 -276) in the absence of adenine nucleotides (Fig. 5A) could be from complexes between the RNA and a monomer, and the increase of cross-linking yields to that RNA could be a consequence of hexamerization. Although APDP-modified Cys-325 Rho did not form cross-links with that 61-nt rut segment, it did form cross-links with RNAs that had additional 3Ј-sequences. But unlike the Cys-82 derivative, it formed cross-links to those RNAs as well in the absence of nucleotides as in the presence of ATP. This result suggests that RNA is interacting near that site in both monomeric and hexameric forms and that either the process of hexamerization or some other consequence of the binding of nucleotides is reducing the cross-linking yield to an extent that offsets the higher binding affinity of the RNA to the hexamer.
The formation of cross-links between Rho derivatives and RNA yielded multiple products with different mobilities in some cases (see Figs. 3, 4, and 5B) and a product with a single mobility in other cases (Figs. 4 and 5A). In general, the multiple products were found using APDP-modified Cys-82 Rho and RNA molecules longer than 98 nucleotides. One explanation for the multiple products is the formation of cross-links to different points in the RNA. However, when the 61-nucleotide rut segment RNA was used, the product appeared to have a unique mobility. We would expect the subunits to cross-link to different positions on it as well as to the other RNAs, but perhaps the different products with this smaller RNA have smaller differences in their mobility.
In contrast, APDP-modified Cys-325 Rho gave unique products with all but the longest RNA used, suggesting that it is cross-linking to a unique point in most of the RNAs. When cro(1-380) was used, two major products were found. In this case, the RNA was large enough that some Rho molecules could bind at possible rut sequences downstream from the primary rut sequence of cro RNA. Indeed, as the ratio of Rho to RNA was increased from 0.9 to 1.8, the yield of low mobility material saturated when the ratio approached 1.0, whereas the yield of the higher mobility complexes increased steadily (data not shown). Because of the low efficiency of cross-linking, we do not believe the very slow moving complexes could arise from two subunits cross-linked to the same RNA molecule. ATP affected the cross-linking in very different ways depending on the location of the cross-linker group. It caused a 3-fold increase when the group was in the RNA-binding domain (Cys-82), had no effect when the group was on residue 325, and decreased even further the low level cross-linking when the group was attached on the equatorial position. ATP is also known to change the structure of Rho in a way that blocks access of trypsin to a site at Arg-128 (8,38), probably by altering the conformation within a subunit. It also enhances oligomerization (39) and increases the affinity of Rho for RNA (20). However, because the cross-linking experiments were done using concentrations of Rho (21 nM as subunits) that saturate binding even in the absence of ATP, we believe that the changes in cross-linking efficiency are likely due to the ATP-induced changes in the conformation within a subunit. A change in conformation in the RNA-binding domain could bring the group into closer proximity to RNA bound in the clefts in the RNA-binding domains.
AMP-PNP and ADP, two other adenine nucleotides that bind to Rho and affect Rho's structure and RNA-binding activity in a seemingly similar extent (20), also increased the extent of cross-linking to the group on Cys-82 but to lesser degrees than did ATP. Although these nucleotides have lower affinities than ATP for Rho (37), they were used at concentrations that were at least 100-fold above their K d values. Because the order of relative cross-linking was the same with RNAs that activated ATP hydrolysis as with RNAs that bound but did not activate ATP hydrolysis, the lower cross-linking yields with the nonsubstrate nucleotides could be due to differences in their effects on the conformation of Rho, rather than the absence of changes associated with nucleotide hydrolysis.
Hexameric Rho Model-All experimental data from this paper were interpreted using a model of the Rho hexamer based on the structure of the F 1 -ATPase ␣ and ␤ subunits. Additional evidence for the structure of the Rho hexamer, positioning the ATP-binding face (P loop) of the Rho monomer toward the center of the hole, comes from work done with image reconstruction and image superimposition on the hexameric form of the RecA onto the F 1 -ATPase. In work reported by Yu and Egelman (40), the crystal structure of the ATP-binding domain of RecA (similar to that of F 1 -ATPase ␤ subunit, Rho, Bacillus stearothermophilus PcrA RNA helicase (41), and many of the hexameric, ATP-binding helicases) were superimposed on that of the F 1 ATPase with less than 2.0-Å root mean square deviation. The overall orientations of the ATP-binding domains in the RecA hexameric structure are positioned so as to face the hole at the center of the hexamer. Also, Sawaya et al. (42) recently solved the x-ray crystal structure of the helicase domain from the T7 gene 4 protein. They found that the structure of this domain was similar to both the RecA and F 1 ATPase ␣ and ␤ subunits. When the crystal data were imposed on reconstructed EM structures of the T7 gene 4 protein, they found that the arrangement of the hexameric ring was also similar to the F 1 -ATPase hexameric structure. Thus, the proposed orientation of the monomers in the Rho hexameric model has strong evidence based on structural and functional homology.
Further analyses of electron micrographic images of Rho have allowed an orientation of the RNA-binding site domains in a three-dimensional reconstruction (18). The atomic structure or the N-terminal domain of Rho is known (9) and could be readily modeled into a low resolution structure of the Rho hexamer. In this model the cleft for binding RNA is oriented facing outward from the hexameric ring. This orientation was supported by images showing tRNA molecules bound to Rho (18). This model is consistent with the evidence we have presented, because the positions of the clefts are beyond the 21-Å distance to the equatorial residue (residue 153) used for the cross-linking studies. These images further support the consensus that Rho acts like the hexameric DNA helicases. Our crosslinking studies, however, have supplied the first direct experimental evidence that RNA on the 3Ј-side of the primary FIG. 6. Topological model of Rho bound to mRNA. The shape of the hexameric Rho form was based on the three-dimensional reconstruction of electron micrograph images as reported by Yu et al. (18). We have placed the RNA-binding domain(s) with the binding cleft(s) facing out to allow formation of a continuous RNA-binding surface that was suggested previously from RNaseA protection studies (44,45). attachment site does indeed pass through the hole of the hexameric structure.