Domain organization of Escherichia coli transcript cleavage factors GreA and GreB.

The GreA and GreB proteins of Escherichia coli induce cleavage of the nascent transcript in ternary elongation complexes of RNA polymerase. Gre factors are presumed to have two biologically important and evolutionarily conserved functions: the suppression of elongation arrest and the enhancement of transcription fidelity. A three-dimensional structure of GreB was generated by homology modeling on the basis of the known crystal structure of GreA. Both factors display similar overall architecture and surface charge distribution, with characteristic C-terminal globular and N-terminal coiled-coil domains. One major difference between the two factors is the "basic patch" on the surface of the coiled-coil domain, which is much larger in GreB than in GreA. In both proteins, a site near the basic patch cross-links to the 3' terminus of RNA in the ternary transcription complex. GreA/GreB hybrid molecules were constructed by genetic engineering in which the N-terminal domain of one protein was fused to the C-terminal domain of the other. In the hybrid molecules, both the coiled-coil and the globular domains contribute to specific binding of Gre factors to RNA polymerase, whereas the antiarrest activity and the GreA or GreB specificity of transcript cleavage is determined by the N-terminal domain. These results implicate the basic patch of the N-terminal coiled-coil domain as an important functional element responsible for the interactions with nascent transcript and determining the size of the RNA fragment to be excised during the course of the cleavage reaction.

Two closely related Escherichia coli proteins, GreA and GreB, participate in RNA polymerase (RNAP) 1 transcription elongation by preventing and/or suppressing the condition of elongation arrest (1,2). In addition, both factors have been shown to facilitate the transition of RNAP from the stage of abortive initiation to productive elongation (3). GreA and GreB may also have a proofreading role in transcription (4). It is thought that the Gre activity is accomplished by endonucleolytic cleavage of RNAs within the ternary elongation complexes (TECs) (2). The cleavage is followed by dissociation of the 3Ј-terminal fragment and restart of elongation from a newly generated 3Ј-OH terminus (2). Similar reactions are induced in eukaryotic TECs of RNA polymerase II by transcription elongation factor TFIIS (5)(6)(7)(8), which performs the same functions as GreA and GreB but lacks any sequence similarity.
GreA and GreB have almost the same molecular mass and share substantial amino acid sequence homology (2). However, there are several differences in their functional and biochemical properties in vitro. First, in all studied TECs that are susceptible for cleavage reactions, GreA induces hydrolysis of short RNAs 2 or 3 nucleotides long from the RNA 3Ј terminus ("type A" cleavage activity), whereas GreB stimulates hydrolysis of RNAs that are 2-18 nucleotides long, depending on the stage of transcription elongation ("type B" cleavage activity) (2,9). Second, GreA can only prevent the formation of arrested TEC (read-through activity), whereas GreB, besides displaying read-through activity, can convert the preformed arrested TEC to a productive complex by a cleavage and restart mechanism (2). Finally, we have recently shown that Gre factors form reversible complexes with RNAP core or holoenzyme with apparent K d values that differ by at least 2 orders of magnitude: 2-3 ϫ 10 Ϫ5 M (GreA) and 1-2 ϫ 10 Ϫ7 M (GreB). These values were estimated by size exclusion high pressure liquid chromatography of RNAP complexes with 35 S-labeled Gre proteins and were further supported by affinity chromatography of 35 S-labeled Gre factors on immobilized His-tagged RNAP (10). Similar K d values were obtained for the complexes of Gre proteins with purified TECs (prepared on a ribosomal rrnB P1 promoter DNA) carrying 6-mer and 12-mer transcripts (10).
TFIIS and Gre factors do not cleave free RNA or RNA⅐DNA duplexes (2,10) and require the formation of a ternary complex between RNAP, DNA, and RNA transcript for their activity (10) or a specific binary complex between RNAP and RNA (11). On the other hand, both E. coli RNAP and eukaryotic RNAP II show a weak ability to cleave RNA in the absence of factors, and this activity is stimulated by mild alkaline pH (12) or pyrophosphate (13). These data suggest that the transcript cleavage factors are not nucleases per se but cofactors that stimulate an intrinsic nucleolytic activity of RNAP. A working model that emerged from these data envisages the RNAP catalytic center as the performer of the cleavage reaction (12,13). In this model, the catalytic center disengages from the RNA 3Ј terminus, slips back along the transcript, hydrolyzes or pyro-phosphorolizes an internal phosphodiester bond and reengages with the newly generated 3Ј terminus. The model is consistent with the notion of conformational flexibility of the TEC that is central to the discontinuous mechanism of RNA polymerization (14,15).
Understanding the mechanism by which Gre proteins induce these perturbations within the TEC and stimulate cleavage would help unravel the mechanism of elongation. In a previous paper, we reported the 2.2-Å resolution crystal structure of GreA (16), which suggested a way this protein might interact with the TEC. The GreA structure consists of the C-terminal globular and N-terminal extended coiled-coil domains, resembling a fist with an outstretched index finger. One surface of the molecule is uniformly negatively charged, while the opposite surface is essentially hydrophobic with a small basic patch near the tip of the finger. We suggested that during interaction with the TEC, GreA's acidic side is oriented away from the acidic surface of RNAP, while the basic patch contacts RNA in the complex.
In the present work, this model was further explored through comparative analysis of GreA and GreB domain organization and function. To this end, we sought to relate the similarities and differences between biochemical functions of the two factors to the common and divergent features of their structures.
Strains and Plasmids-E. coli greA Ϫ greB Ϫ strain AD8571 (12) was used for overexpression of wild type (WT) and mutant Gre factors and for preparation of GreA/GreB-free RNAP. E. coli strain XL1-Blue (Stratagene) was used for selection of all plasmids after ligation. pTRC99A (Pharmacia Biotech Inc.) was used as an expression vector for construction of plasmids overproducing WT and mutant Gre proteins. Plasmid pCF3 is a derivative of pUC19 containing an insertion of the cloned greB gene (2). All sequences of WT and mutant gre genes described in this work were verified by dideoxy chain termination sequencing of the double-stranded plasmid DNA.
Oligonucleotides-The following oligonucleotides were used for construction of overexpressing plasmids and PCR mutagenesis (italicized nucleotides correspond to greA and greB genes, as indicated in parentheses; underlined nucleotides define the restriction sites of the enzymes, shown in parentheses; boldface nucleotides correspond to the region coding the His 6 tag): 1) 5Ј-pCAAGCTATTCCGATGACCTTA-3Ј (GreA 5Ј terminus); 2) 5Ј-CGCGCGGATCCTTACAGGTATTCCACCT- Construction of Plasmids Overproducing WT GreA, GreB, and Histagged Gre Proteins--Oligonucleotides 1 and 2, synthesized according to the DNA sequence of the greA gene (17), were used for PCR amplification of greA from E. coli DNA. The PCR product was purified by electrophoresis and digested with BamHI. The expression vector pTRC99A was digested with NcoI, filled in by Klenow fragment of DNA polymerase I, and then cleaved by BamHI. The greA fragment was inserted into the expression vector pTRC99A between the blunt end and BamHI sites. Ligation of these fragments yielded plasmid pMO1.1, which contained the greA gene under the control of isopropyl-1-thio-␤-D-galactopyranoside-inducible trc promoter. For engineering of the plasmid overproducing GreB, oligonucleotides 3 and 4 were used for PCR amplification of the greB gene from the plasmid pCF3. The purified PCR product was digested with BspHI and BamHI and inserted into linearized pTRC99A between the NcoI and BamHI sites. Ligation of these fragments yielded pMO1.4, which contained the greB gene under the control of the trc promoter. Plasmids overproducing His 6 -tagged GreA (pMO1.1His) and GreB (pMO1.4His) were obtained similarly, except that oligonucleotides 5 and 6 carrying His 6 codons at the end of each gre gene were used for PCR.
Construction of GreA/GreB Hybrids-We engineered a set of 10 chimeric GreA/GreB molecules that carry N-terminal polypeptide fragments of one Gre protein fused to the C-terminal part of the other at amino acid positions 6, 12, 31, 81, and 107 of GreA. Ten hybrid primers were synthesized: the 5Ј-proximal half of each primer corresponded to the DNA fragment of one gre gene immediately upstream of the fusion and the 3Ј-proximal half corresponded to the fragment of the other gre gene immediately downstream of the fusion. All greA/greB hybrids were obtained by a conventional two-step PCR procedure.
Protein Purification--WT GreA and GreB were purified to apparent homogeneity as described (18). For isolation of His 6 -tagged GreA, greA Ϫ greB Ϫ E. coli (AD8571) cells transformed with pMO1.1His were grown and induced with isopropyl-1-thio-␤-D-galactopyranoside as described (18). The cells were pelleted by low speed centrifugation, homogenized in 5 ml of buffer A (7 M guanidine HCl, 40 mM Tris-HCl, pH 7.5, 0.8 M NaCl, 1 mM EDTA, and 1 mM dithiothreitol) and centrifuged at 25,000 ϫ g for 20 min. The supernatant was added to 0.8 ml of Ni 2ϩ -chelating NTA-agarose beads (Qiagen) preequilibrated in buffer A. The suspension was incubated for 20 min at 25°C on a rotary shaker, and the beads were pelleted by low speed centrifugation and washed with 3 ϫ 10 ml of buffer A. The adsorbed GreA was renatured by five subsequent washings of the beads with 5 ml of buffer B (the same as buffer A but without guanidine HCl) and eluted with 2 ml of buffer B, containing 0.6 M imidazole. The eluate was desalted and concentrated in Centricon-10 (Amicon). The resulting electrophoretically homogeneous His 6 -tagged GreA (1.2 mg in 200 l) was stored at Ϫ20°C in buffer B containing 50% glycerol. His 6 -tagged GreB and GreA/GreB hybrids were purified essentially as described above.
In Vitro Transcription Assays-The read-through, antiarrest, and transcript cleavage assays were performed using a 202-base pair E. coli rrnB P1 DNA fragment as described previously (1, 2, 18, 30) (see ''Results'' and legends to Figs. 3 and 7 for details). The NTP concentrations for chain extension reactions were 5 M except for 8-N 3 ATP, which was 100 M. The RNA products were analyzed, autoradiographed, and quantified by PhosphorImager (Molecular Dynamics) as described (12,18).
Gre-RNAP Binding Assay-15 g of each His 6 -tagged Gre factor were incubated for 10 min at 4°C with 7 l of Ni 2ϩ -chelating NTAagarose beads in 20 l of buffer C (40 mM Tris-HCl, pH 7.5, 30 mM KCl, 0.1 mM EDTA, and 0.1 mM dithiothreitol) containing 0.8 M NaCl. The beads were then washed in 3 ϫ 500 l of the same buffer. The yield of immobilized Gre factors was ϳ95% for all proteins. 5 g of RNAP holoenzyme in 20 l of buffer C containing 0.1 M NaCl, 10 mM MgCl 2 , and 0.5 mg/ml of BSA were added to the beads and incubated with gentle shaking for 15 min. Free RNAP was removed by centrifugation, and the beads were washed twice with 100 l of the same buffer. The beads carrying adsorbed RNAP were suspended in 20 l of transcription buffer (buffer C, supplemented with 0.5 mg/ml BSA and 10 mM MgCl 2 ) containing 0.5 g of rrnB P1 DNA template, 0.5 mM CpA, 2 M ATP, and 1 M [␣-32 P]CTP. After incubation at 37°C for 10 min, the beads were washed with 4 ϫ 200 l of buffer C containing 0.2 mg/ml of BSA and 0.2 M NaCl. The adsorbed proteins and TEC carrying radiolabeled hexameric transcript CpApCpCpApC (6C-TEC) (here and elsewhere, boldface type symbolizes radioactive phosphates) were eluted from the beads with 50 l of buffer C containing 20 mM EDTA and 0.8 M NaCl. An aliquot of each eluate (5 l for all samples except 0.5 l for WT GreB) was analyzed by denaturing 23% polyacrylamide gel electrophoresis (PAGE) followed by autoradiography and quantitation by PhosphorImager. In the control experiment, 6C-TEC was prepared as above but in the absence of beads, and His 6 -tagged GreA or GreB (100 g/ml) were mixed with 2 g of RNAP in 20 l of transcription buffer. The resulting 6C-TEC was purified by gel filtration using Quick-Spin G-50, and an aliquot ( 1 ⁄200 of the total material) from each reaction was analyzed by PAGE.
RNA-Protein Photocross-linking-Photocross-linking experiments were performed using radioactively labeled 9A*-TEC carrying 8-N 3 AMP at the RNA 3Ј terminus essentially as described previously for the cross-linking of GreA (16).
Proteolytic Degradation of the Cross-linked Proteins-The crosslinked ␤Ј-subunit (ϳ3 ϫ 10 4 cpm) carrying radioactive RNA was purified by SDS-4% PAGE as described (19) and mixed with 5 g of purified unlabeled ␤Ј-protein. The resulting material was precipitated by cold acetone (1:1 (v/v)), redissolved in 4 l of buffer D (40 mM Tris-HCl, pH 8.2, 40 mM NaCl, 1 mM dithiothreitol, 1 mM EDTA) containing 1% SDS, and incubated for 5 min at 37°C. The mixture was diluted with 200 l of buffer D containing 0.15 g of thrombin and incubated at 37°C for 5 h. The reaction was terminated by the addition of 10 l of 10 mM diisopropylfluorophosphate and analyzed by Tris-glycine SDS-14% PAGE followed by autoradiography. For proteolytic cleavage of GreB, the reaction mixture containing cross-linked GreB (1 ϫ 10 5 cpm) was mixed with 4 g of unlabeled GreB and incubated with 0.1 g of AspN protease in 50 l of buffer D containing 5 mM CaCl 2 for 1 h at 30°C. Under these conditions, the cross-linked ␤Ј was resistant to proteolysis by AspN (data not shown). The reaction was terminated by the addition of EDTA (20 mM final concentration) and analyzed by Tris-Tricine SDS-14% PAGE and autoradiography. The 16-kDa proteolytic fragment carrying cross-linked RNA was purified as described above for ␤Ј and digested with 1 g of endoproteinase ArgC in buffer D containing 10 mM CaCl 2 and 10 mM dithiothreitol for 4 h at 37°C. The reaction was terminated and analyzed as described above for AspN.
Chemical Degradation of the Cross-linked Proteins-The purified cross-linked species were mixed with 5 g of the same purified unlabeled protein prior to chemical degradation. For cleavage at Cys, the proteins were treated with NTCB for 4 h at 37°C as described (20). For cleavage at Trp, the proteins were incubated with BNPS-skatole (10 mg/ml) in 70% acetic acid for 2 h, at 37°C according to Ref. 21.
CD Spectrometry-CD spectra were obtained using an AVIV circular dichroism DS-62 spectrophotometer at 4°C. The buffers were 50 mM sodium phosphate, pH 7.5, containing 50 mM NaCl for GreA and 250 mM NaCl for GreB. Protein concentrations used for experiments were in the range of 15-40 M. The path length of the cuvette was 0.1 cm. Data were collected every 1 nm with a bandwidth of 1.5 nm, and the CD signal was averaged over 2 s. Each spectrum was averaged over four scans from 190 to 300 nm. The temperature dependence of the CD profiles was studied at 222 nm using a cuvette with the path length of 0.1 cm, and the temperature was varied from 5 to 60°C in 2°C increments. The samples were equilibrated for 2 min at each temperature point, and the CD signal was averaged over 30 s.

RESULTS
The Model Three-dimensional Structure of GreB-Diffraction quality crystals of GreB have not been obtained yet. However, the high degree of functional (2) and sequence homology (35% sequence identity) between E. coli GreA and GreB suggests that they have similar structures (25). Moreover, examination of the sequence alignment reveals that key features of the GreA structure are conserved in GreB (16). These features include hydrophobic residues that participate in forming the hydrophobic core of the coiled-coil domain; charged residues that form interhelical salt bridges that stabilize the coiled-coil structure; all of the residues involved in interactions between the N-and C-terminal domains; and the helical bulge at position 24, causing a skip in the coiled-coil heptad repeat pattern.
Ultraviolet CD spectroscopy provides a rough indicator of the  208 for GreA and GreB (1.3 for both) is suggestive of double-stranded coiled-coils, as observed in the GreA x-ray crystal structure. In addition, the CD signals at 222 nm were monitored as a function of temperature to investigate the thermal unfolding of the proteins (Fig. 1B). Both proteins behave similarly, with cooperative thermal unfolding transitions at about 46 and 43°C for GreA and GreB, respectively.
Taken together, these observations strongly support the conclusion that the structure of GreB is very similar to that of GreA. A three-dimensional model of GreB was calculated using comparative protein modeling by satisfaction of spatial restraints as implemented in the program MODELLER 2 (28). First, the relatively high sequence similarity between GreA and GreB resulted in reliable alignment with only one single residue deletion. Second, the alignment was used to derive many distance and dihedral angle restraints on the GreB sequence. Finally, the GreB model was calculated by minimizing the violations of these restraints. The main chain atoms of the model are generally within 0.4 Å of the equivalent GreA atoms. Assuming that the relative orientation of the coiled-coil and the globular domain is conserved, the main chain root mean square error in the model is expected to be approximately 1.5 Å, and about 70% of the side chain rotamers are expected to be modeled correctly (29).
The most striking result from the homology model of the GreB structure is revealed by examining the charge distribution around the water-accessible surface (Fig. 2). As noted earlier (16), the charge distribution around the surface of GreA exhibits a remarkable asymmetry. One face of the molecule is strongly acidic (Fig. 2, top left), whereas the opposite face is neutral except for a small basic region formed by Arg 52 and Arg 37 . The asymmetry in the GreB charge distribution is even more dramatic. As with GreA, one face of GreB is acidic (Fig. 2, bottom left); however, the opposite face is strongly basic and has a much larger basic patch than GreA. These structural features may be related to the different functional properties of the two factors.
Cross-linking of RNA 3Ј Terminus to GreA, GreB, and RNAP in  4 and 7) or 0.5 g of GreB (lanes 3 and 6). TECs-To investigate the orientation of GreA and GreB relative to RNA in the TECs we employed RNA-protein cross-linking using a photoreactive analog of AMP (8-N 3 -AMP) incorporated into the 3Ј terminus of radioactively labeled nascent RNA. On the ribosomal rrnB P1 promoter (starting sequence: CACCACUGA-CACGG . . . ) in the presence of the dinucleotide CpA, ATP, and [␣-32 P]CTP, RNA polymerase forms a stable TEC carrying the hexameric transcript CpApCpCpApC (6C) (30). The 6C-TEC was purified by gel filtration (Fig. 3A, lane 1) and used for stepwise extension of the transcript to 7U, 8G, and 9A (lanes 2, 3, and 4, respectively). The addition of a mixture of UTP, GTP, and 8-N 3 -ATP to radiolabeled 6C-TEC led to the formation of 9A*-TEC carrying transcript CpApCpCpApCpUpGpA* (lane 5) with photoactive azido probe at the RNA 3Ј end.
In the absence of Gre proteins, irradiation of the 9A*-TEC generated a single radioactive band visible on an autoradiograph of SDS-PAGE (Fig. 3B, lane 6), which corresponds to the cross-linked ␤Ј-subunit of RNAP. A control experiment was performed with a mutant RNAP carrying a 127-residue insertion in a nonessential region of the ␤-polypeptide (␤ ins ) (31). The ␤ ins -subunit migrates in SDS-PAGE much slower than the WT ␤ (32) (Fig. 3C, lanes 1 and 2). However, cross-linking experiments conducted with both WT and ␤ ins RNAP produced a single radioactive band of the same mobility (lanes 5 and 8, respectively), eliminating ␤ as the possible cross-linked species. The conclusion that the 3Ј end of RNA cross-links to ␤Ј is in agreement with our earlier result obtained for the 22A*-TEC formed on T7AI promoter (19).
The addition of GreB to the 9A*-TEC followed by UV-irradiation resulted in a 3-fold decrease of ␤Ј cross-linking and the appearance of a second radioactive band, which migrates slightly above GreB with an apparent molecular mass of 19 kDa (Fig. 3B, lane 4). The yield of GreB cross-linking was approximately 18% of the initial 9A*-TEC. Under the same conditions, GreA cross-linked with the yield of about 5% and did not cause decrease in the cross-linking of ␤Ј (Fig. 3B, lane   5). Fig. 3C demonstrates that neither GreA (lanes 4 and 7) nor GreB (lanes 3 and 6) induced the cross-linking of the ␤-subunit. The cross-linked sites on GreB and ␤Ј were mapped by specific chemical and enzymatic degradation as shown in Figs. 4 and 5.
Mapping of the RNA-cross-linking Site in GreB-First, the cross-linked GreB was subjected to limited proteolysis with endoproteinase AspN, which specifically hydrolyzes peptide bonds at the N-terminal side of Asp (33). Under nondenaturing conditions, this enzyme quantitatively cleaves free GreB only at Asp 47 generating two polypeptide fragments, Met 1 -Ala 46 and Asp 47 -Pro 158 , as revealed by N-terminal amino acid sequencing (data not shown). The reaction yielded a radioactive product with an apparent molecular mass of 16 kDa (Fig. 4A, lane 1), which is the expected mass of the C-terminal fragment Asp 47 -Pro 158 plus the RNA probe. Next, the 16-kDa fragment was chemically cleaved at Cys residues by NTCB (20). Since GreB contains only one Cys at position 68, the only expected cleavage products are the 2.5-kDa N-terminal and 10.5-kDa C-terminal fragments. Two radioactive bands were detected by autoradiography after SDS-PAGE analysis: the uncleaved GreB and a peptide with an apparent molecular mass of ϳ5.0 kDa (Fig. 4B, lane 2), which we interpreted as the N-terminal fragment carrying an additional mass of 2.5 kDa contributed by the cross-linked 9A-RNA. These results map the cross-link within the fragment Asp 47 -Lys 67 . This conclusion was confirmed by a chemical cleavage of the 16-kDa proteolytic fragment at Trp residues with BNPS-skatole (21). This cleavage resulted in a single radioactive product with an apparent mass of ϳ6.5 kDa (Fig. 4C, lane 2), which could only correspond to Asp 47 -Trp 91 . For further mapping, the 16-kDa fragment was digested by endoproteinase ArgC, which specifically cleaves peptide bonds at the C-proximal side of Arg (34). The major radioactive product observed near the 3.5-kDa marker (Fig. 4C 1, 2, 3, and 4,  respectively). B, an autoradiograph of Tris-Tricine gradient (10 -20%) SDS-PAGE shows the initial purified 19-and 20-kDa fragments (lane 1) and the products of chemical degradation by BNPSskatole (lane 2). C, an autoradiograph after SDS-PAGE (same as in B) shows the initial cross-linked ␤Ј (lane 1) and the products of chemical degradation by NTCB (lane 2). D, schematic presentation of the mapping procedure (for explanation, see Fig. 4D). The bar at the top represents 1407-amino acid-long ␤Ј polypeptide, with lettered boxes symbolizing highly conserved regions (42). The Arg-Gly, Cys, and Trp cleavage sites that follow from the sequence of ␤Ј are indicated by arrows.
Arg 63 , which coincides with the location of the large basic patch of GreB.
In our previous paper (16), we described the localization of the cross-linked site on GreA in 9A* TEC between Asp 41 and Phe 57 . Thus, the cross-linking of the RNA 3Ј terminus to both Gre factors occurs within an overlapping 22-amino acid-long fragment between positions 41 and 63, suggesting that this region is responsible for the interactions with the nascent RNA 3Ј terminus.
Mapping of the RNA-cross-linking Site in the ␤Ј-subunit of RNAP-For localization of the cross-link in ␤Ј, the initial radioactive material was isolated from SDS gel and subjected to exhaustive degradation by thrombin. This enzyme recognizes the consensus sequence Leu-Val-Pro-Arg-Gly-Ser. However, it is also highly specific for the sequence Arg-Gly and cleaves peptide bonds at the C-proximal side of Arg (35). After a prolonged incubation in the presence of SDS, the reaction yielded a single prominent radioactive band visible on an autoradiogram after SDS-PAGE (Fig. 5A, lanes 1-4) with apparent molecular mass of ϳ20 kDa. Further analysis of the isolated 20-kDa fragment by Tris-Tricine SDS-PAGE revealed two closely migrating bands with masses of ϳ20 and ϳ19 kDa (Fig.  5B, lane 1). Taking into account the additional mass of RNA these could only correspond to fragments Gly 733 -Arg 907 and Gly 746 -Arg 907 . Next, the mixture of the two Gly-Arg fragments was subjected to chemical cleavage at Trp residues by BNPSskatole. The reaction yielded two radioactive cleavage products: a band corresponding to the detached RNA and a band with a mass of ϳ6 kDa (Fig. 5B, lane 2), which we interpreted as the C-terminal fragment Trp 869 -Arg 907 . Finally, the initial cross-linked ␤Ј was chemically cleaved at Cys residues by NTCB. The major radioactive product with a mass of ϳ10 kDa, representing about 70% of the total radioactivity (Fig. 5C, lane  2) was observed, which was also the smallest fragment visible on the autoradiograph. According to the distribution of Cys residues in ␤Ј (see Fig. 5D), the 10-kDa product corresponds to a fragment flanked by Cys 814 and Cys 898 . The other radioactive bands above the 10-kDa product visible on the autoradiograph presumably correspond to the products of incomplete digestion, since the yield of cleavage reactions with NTCB may vary from 90 to 30% (20), depending on the susceptibility of Cys residues. However, the possibility of minor cross-linking sites (representing less than 10% of the total radioactivity in cross-linked ␤Ј) outside of the localized Cys 814 -Cys 898 fragment cannot be excluded. Thus, the major cross-link site is mapped to the 29amino acid-long segment between Trp 869 and Cys 898 . This segment is located near the conserved "region G" of ␤Ј, where we have previously mapped the cross-link site for the RNA 3Ј terminus in the 22A*-TEC formed on a T7A1 DNA (19).
Since the azido probe is directly attached to the adenine at the RNA 3Ј end and has a reactivity radius of less than 5 Å, the simultaneous cross-linking of ␤Ј and Gre proteins strongly suggests that they are situated in close proximity to each other.
Swapping of Gre Domains in GreA/GreB Hybrids-To delineate functional domains in Gre proteins, we constructed a set of chimeric GreA/GreB molecules carrying the N-terminal fragment of one Gre protein fused to the C-terminal fragment of the other. Each GreA/GreB hybrid carried six His residues at the C terminus in order to facilitate their purification. His-tagged GreA, GreB, and GreA/GreB hybrids (purified as described under "Experimental Procedures") were compared in their ability to activate the transcript cleavage reaction in TECs formed on a ribosomal E. coli rrnB P1 DNA fragment (2,18).
The transcript cleavage assays used in these experiments are based on two properties of Gre proteins: first, the "catalytic" nature of their activity and, second, the type specificity of their respective activities. Gre proteins are able to induce complete transcript cleavage in TECs in substoichiometric amounts even when they are present at a Gre/TEC molar ratio of 0.01:1 (1). This catalytic property of Gre factors enables us to quantitate their specific activity in enzymological terms. Thus, in quantitative assays, one unit of cleavage activity is defined as the amount of Gre protein required for the hydrolysis of 50% of the RNA in 7U-or 9A-TECs under standard reaction conditions (18). The specific cleavage activity is expressed as units/g of Gre protein. In 9A-TEC carrying the radiolabeled transcript CpApCpCpApCpUpGpA, GreA stimulates cleavage and release of dinucleotides pGpA and pCpU (type A cleavage activity), whereas GreB induces the cleavage of the tetranucleotide pCpUpGpA (type B cleavage activity) (2). Thus, in the qualitative assay performed on the 9A-TEC, the type of products generated define the GreA-or GreB-type activity. The results obtained from comparative analyses of hybrid and WT Gre proteins using these criteria are summarized in Fig. 6.
For hybrids 1 and 6, in which only the first six N-terminal residues were exchanged, the type of cleavage activity was determined by the C-terminal part of the resulting polypeptide. Thus, hybrid 1 retained the GreA-type activity, while hybrid 6 retained the GreB-type activity. Both hybrids exhibited decreased cleavage activity toward 7U-and 9A-TECs compared with the corresponding WT Gre factors. The extent of the decrease varied from 7-to 1000-fold, depending on the combination of N-and C-terminal fragments and the type of TECs. Hybrid 1 and GreA displayed, for example, 140 and 1000 units/g cleavage  5, 6, 7, and 8, respectively).  6, 7, 8, and 9, respectively). activity toward 7U-TEC, or 0.1 and 100 units/g toward 9A-TEC, respectively. Hybrid 6 and GreB displayed 20 and 300 units/g activity toward 7U-TEC or 50 and 1000 units/g toward 9A-TEC, respectively. Thus, the changes within the short N-terminal segment of GreA and GreB affected the specific activity (units/g) of these proteins and their selectivity toward TECs but did not affect the type of cleavage reaction.
Increasing the N-terminal portion of GreB in GreA to 12 residues (hybrid 2) restored the original cleavage activity of GreA. However, similar substitution in GreB (hybrid 7) led to a total loss of cleavage activity toward both TECs. Apparently, this effect was due to incorrect protein folding, since hybrid 7, when expressed in E. coli, was found in the inclusion bodies, less than 1% of which was recovered in soluble form after renaturation from 7 M guanidine HCl. The same results were obtained when the first 12 N-terminal residues in GreA and GreB were deleted (data not shown). According to the threedimensional structure of GreA and GreB, four nonconserved residues at positions 4, 5, 8, and 9 and conserved Thr 6 are buried inside the Gre molecule and, presumably, participate in forming the hydrophobic core of the N-terminal coiled-coil domain (16). Our results together with crystallographic data suggest that the N-terminal 12 residues form an essential structural element for initiating the proper folding of the coiled-coil domain of Gre molecules.
The exchange between GreA and GreB of the N-terminal 31 residues (hybrids 3 and 8) as well as 44 residues (data not shown) yielded insoluble proteins that refolded poorly and displayed no detectable cleavage activity. The N-terminal residues 1-44 comprise the first ␣-helix of the coiled-coil domain, and most of the residues in this region that are critical for the formation of correct coiled-coil structure are highly conserved in both Gre molecules (2,16). The only exceptions are those that comprise two of the four interhelical bonds that stabilize the overall structure of N-terminal domain: salt bridge Lys 22 -Glu 66 and hydrogen bond Arg 25 -Asn 54 in GreA and salt bridge Glu 25 -Arg 54 and hydrophobic bond Trp 22 -Val 62 in GreB (2,16). Presumably, the disruption of these two interhelical bonds in GreA/GreB hybrids led to destabilization of coiled-coil structure and resulted in proteins that failed to assume their native conformation.
The Type of Gre-specific Cleavage Activity Is Determined by the N-terminal Domain-Exchanging the N-terminal 81 residues of GreA with the corresponding residues of GreB (hybrid 4) resulted in the switch of cleavage activity from type A to type B (Fig. 6, hybrid 4; Fig. 7A, lanes 5 and 6). The specific transcript cleavage activity of hybrid 4 toward 7U-and 9A-TECs decreased 1500-and 500-fold, respectively, in comparison with GreB. (Because the cleavage activity of hybrid 4 switched to type B, its specific activity can only be compared with GreB.) Similarly, the hybrid 9 displayed type A cleavage activity (Fig.  6, hybrid 9; Fig. 7A, lanes 7 and 8); however, the specific activity of this hybrid toward 7U-and 9A-TECs decreased only moderately (50-and 3-fold, respectively) compared with GreA. Hybrids 5 (Fig. 7A, lanes 9 and 10) and 10 ( Fig. 7A, lanes 11  and 12) displayed the same types of cleavage activity as hybrids 4 and 9, respectively. However, the specific cleavage activity of hybrid 5 showed substantial improvement over that of hybrid 4, whereas the activity of hybrid 10 toward 9A-TEC was lower than that of hybrid 9 (Fig. 6). These data demonstrate that the N-terminal coiled-coil domain of Gre protein dictates the type of cleavage activity characteristic for each factor. These results also suggest that the C-terminal globular domain does not affect the type of cleavage reaction but is required for full transcript cleavage activity toward TECs.
The Antiarrest Activity of Gre Factors Is Determined by Their N-terminal Coiled-coil Domain-The hybrids 4, 5, 9, and 10 were further compared with the WT factors for their ability to suppress the formation of arrested TECs during transcription elongation on a ribosomal rrnB P1 promoter (read-through assay) and to induce the transcript cleavage in preformed arrested TECs (antiarrest assay) (2,18).
For the read-through assay (Fig. 7B), the radiolabeled 6C-TEC (Fig. 7B, lane 1) was incubated with four NTPs alone or in the presence of GreA, GreB, or hybrid proteins. In the absence of Gre factors, about 50% of the TEC was arrested at positions ϩ12 and ϩ13 (lane 2). The addition of GreA, GreB, or hybrids 5 and 9 to the initial 6C-TEC, prior to the addition of NTPs, reduced the formation of arrested TECs and increased the total amount of the full-length run-off transcription product (lanes 3, 4, 7, and 8, respectively). The hybrids 4 and 10 had little or no effect on the formation of arrested TECs (lanes 5 and 6, respectively).
For the antiarrest assay (Fig. 7C), the arrested 12C-and 13G-TECs obtained by chasing the initial 6C-TEC with four NTPs were further purified by gel filtration (lane 2), and then were exposed to either hybrid or WT factors. Of all the proteins tested, only GreB and hybrid 5 induced the cleavage of arrested TECs (lanes 4 and 9, respectively), yielding TECs carrying 5A and 6C transcripts that could be converted into full-length run-off products upon the addition of NTPs (data not shown). Under the same conditions, GreA and hybrids 9, 10, and 4 did not cleave any arrested TECs (lanes 5, 6, 7, and 8, respectively). The fact that GreA was inactive toward arrested 12C-and 13G-TECs is consistent with our earlier observation that, in contrast to GreB, GreA can act only before the arrest took place and not after the fact (2).
Thus, the functional activity of GreA/GreB hybrids in the read-through and the antiarrest assays was determined by their N-terminal domain and correlated with their specific transcript cleavage activity on 9A-TEC (see Figs. 6 and 7A). Indeed, hybrid 9 carrying the coiled-coil domain of GreA and globular domain of GreB acted as GreA displaying only the read-through activity, and hybrid 5 composed of the coiled-coil domain of GreB and two-thirds of the globular domain of GreA behaved as GreB, exhibiting both the read-through and the antiarrest activities. At the same time, hybrids 4 and 10 were inactive in both assays apparently due to their low specific transcript cleavage activity (see Fig. 6).
Both the Coiled-coil and the Globular Domains Contribute to Specific Binding of Gre Factors to RNAP-The functional roles of Gre domains were further explored by comparing the binding affinities of GreA/GreB hybrids and WT Gre proteins with RNAP. The addition of six His residues to the C terminus of Gre proteins does not alter the binding properties of GreB but causes a ϳ10-fold decrease in the binding affinity of GreA (10). When Gre factors are immobilized on Ni 2ϩ -NTA-agarose through His tags, they are still able to bind RNAP from solution in accordance with their binding affinities. Moreover, RNAP molecules bound to immobilized Gre factors are able to form a stable 6C-TEC on rrnB P1 promoter, and neither GreA nor GreB affects the formation and the stability of 6C-TEC. 3 Therefore, the amount of 6C transcript generated by RNAP on the beads is proportional to the amount of RNAP molecules adsorbed to immobilized Gre proteins.
To assess the contribution of each Gre domain to the binding of Gre molecules to RNAP, Ni 2ϩ -NTA-agarose beads carrying WT or hybrid factors were incubated with RNAP holoenzyme at a molar ratio of 10:1 in the presence of BSA as a carrier protein.
After removal of free RNAP, the immobilized Gre⅐RNAP complex was incubated with rrnB P1 DNA, CpA, ATP, and [␣-32 P]CTP to allow the formation of 6C-TEC. The free NTPs were removed, and the radioactivity in 6C-TEC formed on the beads was quantified following urea-PAGE analysis (Fig. 8). According to this semiquantitative assay, the radioactivity of 6C transcript associated with immobilized GreB was ϳ300 times higher than that of GreA (lanes 7 and 4, respectively). In solution, the amount of 6C-TEC formed was the same in the presence of GreA and GreB (Fig. 8, lanes 1 and 2, respectively), indicating that the low recovery of 6C transcript from immobilized GreA (lane 4) was due to a low amount of RNAP bound to GreA. These results are in good agreement with our earlier observation that the apparent K d values for GreB⅐RNAP and GreA⅐RNAP complexes differ by approximately 2 orders of magnitude (10). When Ni 2ϩ -NTA-agarose was used for the assay without immobilized Gre factors, no radioactive 6C transcript was detected (lane 3). In comparison with GreA, hybrids 4, 5, 9, and 10 displayed a 15-, 50-, 8-, and 4-fold increase in the apparent binding affinity toward RNAP (Fig. 8, lanes 5, 6, 8,  and 9, respectively), although none of them reached the binding efficiency of GreB. Both hybrids 5 and 9, carrying, respectively, 3 S. Borukhov, unpublished observation.  (3)); and labels for the elements of secondary structure. The color shading represents the following: amino acid identities (gray); generally conserved hydrophobic residues occupying a and d positions of the coiled-coil heptad repeat and positions 6 and 72 (yellow); residues that participate in interhelical and interdomain interactions (magenta); conserved residues at positions 37 and 52 forming small basic patch of GreA (green); residues forming large basic patch of GreB (blue). the N-and C-terminal domains of GreB, have higher binding affinity than GreA, suggesting that the sites responsible for specific binding to RNAP are equally distributed between the two domains of the Gre molecule. These results also suggest that the binding of Gre factors to RNAP requires cooperative action of their coiled-coil and globular domains. DISCUSSION The transcript cleavage reaction appears to be a ubiquitous and evolutionarily conserved function among multisubunit RNA polymerases. It has been observed in TECs formed by vaccinia virus RNA polymerase (36), prokaryotic RNAP of E. coli (2), and eukaryotic RNA polymerases II (5)(6)(7)(8) and III (37). The elongation factor TFIIS facilitates the transcript cleavage and the read-through/antiarrest activities of RNA polymerase II in such divergent organisms as yeast (8), insects (6), and mammals (5). In prokaryotes, the genes encoding Gre factors have been identified and sequenced from five different organisms including E. coli (2,17), Hemophilus influenzae (38), Rickettsia prowazekii (39), Mycobacterium leprae, 4 and Mycoplasma genitalium (40). In addition, GreA-and GreB-like activities were detected in Pseudomonas sp., Acinetobacter sp., and Bacillus subtilis. 3 The biological importance of Gre factors is also underscored by the fact that the greA gene is present in the smallest known genome of any free-living organism, M. genitalium (40).
The amino acid sequence alignment of seven homologous GreA and GreB proteins is shown in Fig. 9. Although the percentage of identical residues among the members of the Gre family varies, most of the key structural elements of the E. coli GreA (16) are preserved with few exceptions. The conserved structural elements include hydrophobic residues at the expected a and d positions of the characteristic heptad repeat, (abcdefg) 2 , and at positions 6 and 72 that form the hydrophobic core of the N-terminal domain (Fig. 9, yellow shading) and most of the residues participating in interhelical and interdomain salt bridges and/or hydrogen bonds stabilizing the overall Gre structure (magenta and orchid shading). In addition, all five members of the GreA family contain two highly conserved basic residues at positions 37 and 52 (turquoise shading), comprising a small basic patch on the water-accessible surface of GreA (see Fig. 2, top). Similarly, two representatives of the GreB family contain six conserved basic residues at positions 2, 52, 53, 56, 60, and 67 ( Fig. 9, blue shading) that form a large basic patch on the surface of GreB (see Fig. 2, bottom). These observations suggest that all Gre proteins have similar structural organization and spatial arrangement of their N-and C-terminal domains. Comparison of the CD spectra of E. coli GreA and GreB and homology modeling of the GreB three-dimensional structure strongly support this conclusion.
Our results of RNA-protein cross-linking demonstrate that both GreA and GreB interact with the 3Ј end of RNA and presumably ␤Ј-subunit of RNAP in TEC through a 16 -17residue-long peptide segment that overlaps with basic patches (Fig. 9, red underline). This supports our earlier hypothesis that the surface of Gre molecules harboring the basic patch contacts RNAP and RNA in TEC (16). The analysis of the functional roles of Gre domains using Gre hybrids revealed that the N-terminal domain is responsible for specific binding of Gre factors to RNAP and carries the structural determinants conferring the GreA or GreB type of cleavage activity and, presumably, the antiarrest and read-through activities. The C-terminal domain does not have a direct role in transcript cleavage or antiarrest function but participates in binding to RNAP and is required for full activity of Gre factors.