Transcription Termination: Primary Intermediates and Secondary Adducts*

In living organisms, stable elongation complexes of RNA polymerase dissociate at specific template positions in a process of transcription termination. It has been suggested that the dissociation is not the immediate cause of termination but is preceded by catalytic inactivation of the elongation complex. In vitro reducing ionic strength can be used to stabilize very unstable and catalytically inactive complex at the point of termination; the previous biochemical characterization of this complex has led to important conclusions regarding termination mechanism. Here we analyze in detail the complexes formed between DNA template, nascent RNA, and Escherichia coli RNA polymerase during transcription through the tR2 terminator of bacteriophage λ. At low ionic strength, the majority of elongation complexes fall apart upon reaching the terminator. Released RNA and DNA efficiently rebind RNA polymerase (RNAP) and form binary RNAP·RNA and RNAP·DNA complexes, which are indistinguishable from binary complexes obtained by direct mixing of the purified nucleic acids and the enzyme. A small fraction of elongation complexes that reach termination point escapes dissociation because RNA polymerase has backtracked from the terminator to an upstream DNA position. Thus, transcription elongation to a terminator site produces no termination intermediates that withstand dissociation in the time scale appropriate for biochemical studies.

In living organisms, stable elongation complexes of RNA polymerase dissociate at specific template positions in a process of transcription termination. It has been suggested that the dissociation is not the immediate cause of termination but is preceded by catalytic inactivation of the elongation complex. In vitro reducing ionic strength can be used to stabilize very unstable and catalytically inactive complex at the point of termination; the previous biochemical characterization of this complex has led to important conclusions regarding termination mechanism. Here we analyze in detail the complexes formed between DNA template, nascent RNA, and Escherichia coli RNA polymerase during transcription through the tR2 terminator of bacteriophage . At low ionic strength, the majority of elongation complexes fall apart upon reaching the terminator. Released RNA and DNA efficiently rebind RNA polymerase (RNAP) and form binary RNAP⅐RNA and RNAP⅐DNA complexes, which are indistinguishable from binary complexes obtained by direct mixing of the purified nucleic acids and the enzyme. A small fraction of elongation complexes that reach termination point escapes dissociation because RNA polymerase has backtracked from the terminator to an upstream DNA position. Thus, transcription elongation to a terminator site produces no termination intermediates that withstand dissociation in the time scale appropriate for biochemical studies.
Structural stability is a trademark of ternary elongation complexes (ECs), 1 consisting of RNA polymerase (RNAP), template DNA, and newly synthesized transcript. Once formed on a promoter, EC can transcribe thousands of nucleotides and survive prolonged pausing and arrest within the transcribed genes (1,2). The current view of EC structure, based on crystallographic and functional studies of eukaryotic, prokaryotic, and phage RNA polymerases, suggests that the enzyme holds the template with a clamp-like domain (3)(4)(5)(6). This interaction does not interfere with advancement of RNAP along the template but makes the complex resistant to high salt concentration and DNA competitors (3)(4)(5)(6). The minimal nucleic acid architecture that can support a stable bacterial EC and may allow the closure of the clamp around the DNA template, consists of 8 -9 bp of an RNA:DNA hybrid at the 3Ј-end of the nascent RNA and a 9 -12-bp DNA duplex located downstream from the hybrid (7)(8)(9).
During transcription termination, processive RNA synthesis is interrupted by dissociating EC into RNAP, RNA, and DNA (1,2). Termination in Escherichia coli occurs at two classes of specific DNA signals; one depends on action of termination factor Rho, and another requires only cis-acting element in DNA. The latter terminators are called intrinsic, or Rho-independent transcription terminators, and they cause spontaneous dissociation of EC (2). Intrinsic terminators have a conserved structure consisting of a region of dyad symmetry followed by a stretch of 7-9 thymidine residues in nontemplate DNA strand (10). Transcription through this sequence results in formation of a stable hairpin in the RNA 7-9 nt upstream from the termination point followed by a run of U residues (2).
The mechanism of intrinsic termination has been studied for decades, and several models have been suggested to explain the disruption of ECs (for a review, see Ref. 11). According to the established thermodynamic model, formation of the hairpin removes RNA from a putative single-strand RNA-binding site located on RNAP near the point where transcript branches from the RNA:DNA hybrid (12)(13)(14)(15)(16). The model suggests that the hairpin formation also disrupts the 5Ј-proximal part of the RNA:DNA hybrid either directly, by displacing the RNA:DNA with the RNA:RNA pairing, or indirectly, by melting the hybrid in the region 3Ј of the hairpin stem (2,16,17). These events, in combination with low stability of oligo(U):dA hybrid at the terminator, provoke EC dissociation (18). Another, recently reported model considers the hairpin formation behind RNAP as the driving force for physical extraction of the 3Ј-end of the nascent RNA from the RNAP active center accompanied by hypertranslocation of RNAP along the DNA and by collapse of the transcription bubble (19). Conformational changes in RNAP, caused by allosteric contacts between the hairpin and the enzyme, were also suggested to trigger termination (20,21). In addition, direct destabilizing contacts between the termination hairpin and the protein clamp have been proposed, based on the results of protein/DNA and protein/RNA cross-linking experiments (15,17). Depending on biochemical tools that the researchers used, each model proposed a distinct sequence of events leading to termination. However, it is not clear yet whether these events constitute true steps in the termination mechanism.
Dissecting the early steps in the termination pathway would be extremely helpful for the understanding of the detailed mechanism of termination. Several studies suggest that RNAP may stop at a terminator before it dissociates from the DNA and RNA. Indeed, RNA hairpins similar to those found in the terminators often induce pausing of transcription at 10 -12 nt downstream from their stems (22). In the absence of hairpin, RNAP pauses at the end of 7-9-nt-long oligo(T) track (17,19). Thus, both the RNA hairpin and the oligo(T) track would be expected to cause pausing. Also, kinetic competition between dissociation and elongation at the termination point was shown to affect the efficiency of termination (23). Single-molecule light microscopy studies further suggested that pausing of E. coli RNAP for about 1 min precedes dissociation at the intrinsic terminator (24).
The high rate of elongation and EC instability at the terminator hinder isolation and study of termination intermediates (17,20,25). Moreover, the study of the termination mechanism was complicated by the absence of a characterized biochemical system to isolate catalytically inactive intermediates before they dissociate. Reduction of the ionic strength below physiological level has been used as a tool to "freeze" such intermediates (17,26). The complexes that remained stable at terminator at low salt conditions were immobilized in solid phase on nickel-NTA-agarose through histidine-tagged RNAP and were biochemically characterized (17). Footprinting analysis of this complex showed that the normally opened transcription bubble had closed. These findings suggested that the closure of the transcriptional bubble represented an initial step in the termination pathway that accompanies catalytic inactivation of EC prior to its dissociation. Other evidence argued that in low ionic strength conditions, transcription termination and transcript release occurred, but nonspecific rebinding of the dissociated nucleic acids to the enzyme formed separate binary RNAP⅐RNA and RNAP⅐DNA post-termination complexes (27). The immobilization through RNAP cannot separate ternary complexes from the binary adducts between the nucleic acids and the enzyme, imposing the problem of structural discrimination between a population of rebound complexes and an initial ternary intermediate in the termination process. Therefore, the conclusion that was made about the ternary nature of the termination complex rested heavily upon some assumptions that we believed needed to be fully verified.
Here, we performed systematic analysis of the complexes formed between RNA, DNA, and E. coli RNAP in the course of termination in low salt conditions. To immobilize these complexes in solid phase, we utilized affinity tags introduced either to the protein or to the template DNA. We found no ternary termination complex. Instead, we found a small fraction of the polymerase in a ternary complex, which escaped dissociation by backtracking from the point of termination; all other polymerase complexes resulted from post-termination binding of the released template or transcript back to the enzyme. These findings are discussed in light of the current models of the mechanism of intrinsic transcription termination.

EXPERIMENTAL PROCEDURES
Transcription Template-The standard template for transcription was a 156-bp DNA fragment (DNA 156 ) carrying the A1 promoter of bacteriophage T7 and the tR2 terminator of bacteriophage . It was a product of PCR amplification and was purified by PAGE. The template for transcription was 3Ј-biotinylated in the template DNA strand as follows. The PCR product contained a unique BamHI site 70 nt upstream from the ϩ1-position of the T7A1 promoter. Following digestion with BamHI, the 3Ј-end was filled in by Klenow enzyme in the presence of 1 mM dGTP, dCTP, TTP, and 100 M biotinylated dATP (Sigma) for 15 min at 25°C. The biotinylated DNA 156 was purified by phenol extraction, precipitated with ethanol, dissolved in transcription buffer (TB40 (numerical index indicates concentration of KCl in the buffer expressed in mM); 20 mM Tris-HCl, pH 7.9, 40 mM KCl, 5 mM MgCl 2 , 1 mM ␤-mercaptoethanol), and used for EC11 formation in solution as described below.
Preparation of EC Halted at Position ϩ11 of the Template (EC11)-E. coli RNAP bearing a hexahistidine tag genetically fused to the carboxyl terminus of the ␤Ј subunit was purified as described (28) from extracts of the RL916 strain (obtained from Dr. R. Landick). Stable EC stalled at position ϩ11 (EC11) was obtained by preincubating 2 pmol of the template with 2 pmol of the enzyme (unless indicated otherwise) in regular TB40 for 5 min at 37°C and subsequently adding 10 M trinucleotide RNA primer ApUpC and 20 M rATP and rGTP for 5 min at 25°C. The EC11 was next immobilized either on Ni 2ϩ -NTA-agarose or on sreptavidin-agarose beads (see below).
Transcription on Ni 2ϩ -NTA-Agarose-20 l of Ni 2ϩ -NTA agarose (Qiagen) prewashed in TB40 was added to EC11 formed in solution as described above. After a 5-min incubation at 25°C the immobilized EC11 was repeatedly (five times) washed with TB40. Then EC11 was either chased with the four rNTPs (100 M rATP, rCTP, rGTP, and 10 M of UTP unless indicated otherwise) or walked to a desired position by repeated alterations of washing (four washing steps with 1 ml of TB40 each) and incubation with a subset of rNTPs (10 M each) for 5 min at 25°C. The KCl concentration was adjusted to 500 mM during the walking (TB500) to suppress pausing and arrest of RNAP. The transcripts were labeled by the incorporation of an appropriate [␣-32 P]rNTP (PerkinElmer Life Sciences; 40 Ci of the labeled rNTP (3000 Ci/mmol), 5 min at 25°C)). The position of labeling was cytosine ϩ12 unless stated otherwise.
All reactions were stopped with an equal volume of gel loading buffer (50 mM EDTA, 10 M urea) unless indicated otherwise, and the products were separated on denaturing PAGE.
Transcription on Streptavidin-Agarose-20 l of streptavidin-agarose suspension (Sigma) were washed with TB40 and mixed with EC11 for 10 min at 25°C with constant shaking. The DNA-immobilized complex was washed with 1 ml of TB40. All other manipulations were the same as described for transcription on Ni 2ϩ -NTA-agarose.
Test for EC Stability and EC Fractionation-Ten microliters of washed immobilized EC, containing either full-size or GreB-cleaved RNA, was incubated for 1 min (unless indicated otherwise) in TB containing 1 M KCl (TB1000). After 20 s of centrifugation (VWRbrand Mini Centrifuge; 6000 rpm, 2000 g), 5 l of the supernatant was removed and combined with an equal volume of gel loading buffer ("S" fraction). The remaining pellet was washed with 1 ml of TB40, the volume of the sample was adjusted to 5 l and combined with an equal volume of gel loading buffer ("P w " fraction). The "Total" fraction contained 10 l of the nonfractionated immobilized EC combined with 10 l of gel loading buffer.
Potassium Permanganate Footprinting and GreB-induced RNA Cleavage-The DNA template for KMnO 4 footprinting was labeled in 20 l of immobilized EC11 with 10 units of T4 polynucleotide kinase (New England Biolabs) and 50 Ci of [␥-32 P]dATP (7000 Ci/mmol; ICN Biomedicals, Inc.) for 10 min at 25°C. After washing with TB40, the complex was used as starting material for the chase or walking reaction. The 32 P end-labeled complexes were treated with 1 mM KMnO 4 at 25°C for 1 min. The reaction was stopped by adding 1 l of ␤-mercaptoethanol. The complexes were eluted from Ni 2ϩ -NTA-agarose with 50 mM EDTA, the DNA in the supernatants was precipitated in the presence of 4 g of the carrier plasmid DNA, cleaved with 10% piperidine for 15 min at 90°C, reprecipitated, lyophilized, dissolved in gel loading buffer, and separated on 6% denaturing PAGE.
GreB protein was purified as described (29). The cleavage reaction was performed in 10 l of TB40 with 0.5 g of GreB for 10 min at 25°C.
Reconstitution of Nonspecific Complexes between RNA and DNA with RNAP-0.5 pmol of RNAP core were immobilized on 3 l of Ni 2ϩ -NTAagarose beads and washed with TB40, and the volume was adjusted to 90 l. In a separate tube, EC11 was obtained as described above, labeled in both the RNA and the DNA, and then chased with four rNTPs in TB1000 for 10 min. The total volume of the reaction was 40 l. Following fractionation, 10 l of the supernatant that contained labeled RNA 64 -65 (termination products of 64 and 65 nt long), RNA 86 (run-off product), and DNA 156 were combined with 90 l of preimmobilized RNAP, making the final KCl concentration 140 mM. After 5 min of incubation at 25°C with intense shaking, the mixture was washed with TB40. GreB cleavage and RNA extension were then performed as described above.

Isolation of Termination Complex by Immobilization through
RNAP-To study potential intermediates formed in the course of transcription termination, we used a synthetic template carrying the well studied Rho-independent tR2 terminator of bacteriophage (14,17,19,25). It was cloned downstream from the A1 promoter from bacteriophage T7 (Fig. 1A). The hexahistidine-tagged RNAP, which can be bound to Ni 2ϩ -NTA-agarose beads, was used in transcription reactions (28,30). By immobilizing RNAP on beads, the enzyme can be walked in steps along the template by adding subsets of rNTPs alternated with washing of the beads. It also allowed the isolation of the polymerase from the DNA and RNA dissociated during termination and therefore could provide a way to isolate in the solid phase putative termination intermediates that retained stability. This approach has been exploited before for the study of the termination mechanism (17,25,31).
In the experiment of Fig. 1B, RNA-labeled Ni 2ϩ -NTA-immobilized EC11 (the numerical index indicates the RNA length) was obtained as described under "Experimental Procedures." The RNA in the complex was labeled by incorporation of [␣-32 P]rCTP to form EC12 followed by the chase to tR2 terminator with all four rNTPs. Concentration of KCl in the reaction was below physiological level (40 mM) to increase the lifetime of potential termination complex(es). Transcription in TB40 resulted in two RNA products of termination, the major one 64 nt long and the minor one 65 nt long (RNA 64 -65 ), and in run-off RNA 86 nt long (RNA 86 ). To analyze dissociation of the complexes at the terminator, an aliquot of the chase reaction mixture was centrifuged, half of the supernatant was removed (lane 3), and the pellet was washed with TB40 (lane 2, P w fraction). About 60% of the terminator-specific RNA 64 -65 was released, and 40% remained associated with RNAP. The RNAP that reached the end of the template (run-off complex) also was associated with the RNA (RNA 86 , lanes 2 and 3).
Stability of the RNA 64 -65 or RNA 86 complexes with RNAP was much lower than that of the normal ECs (20,32). These RNAs were released after only 1 min of incubation in 1 M KCl (TB1000) (lane 4) except for a small (ϳ10%) fraction of the RNA 64 -65 that remained bound to RNAP even after the wash with TB1000. We address the origin of this minor fraction below.
To test whether DNA was also bound to RNAP after termination, we obtained EC12 with the labeled RNA and labeled template (Fig. 1C) and then chased it with the four rNTPs in TB40. About 10% of the labeled DNA was released into the supernatant (lanes 2 and 3). After a subsequent wash with TB1000, only 20 -30% of the labeled template remained bound to the Ni 2ϩ -NTA-agarose with the RNAP (see Fig. 2B, bottom panel, lanes 5 and 6). Thus, more DNA than RNA 64 -65 was bound to polymerase in both high and low salt conditions. Greater retention of the DNA compared with RNA 64 -65 is most likely explained by a fraction of the stable ternary complexes becoming permanently arrested on the DNA template before reaching the terminator and the presence of the ternary complex containing the run-off RNA 86 , which has not dissociated in low salt (Fig. 1C, lane 3).
We shall define the fraction of EC64 -65, which was retained on the beads after washing with TB40, as termination complex isolated on Ni 2ϩ -NTA agarose (TC/Ni).
Potassium Permanganate Footprinting of TC/Ni-Next, we analyzed the structure of a transcriptional bubble in TC/Ni using potassium permanganate (KMnO 4 ) footprinting, which detects unpaired thymidine residues in DNA (17,33,34). Because of the very low stability of TC/Ni, it was crucial to test whether the complex would withstand the standard footprinting conditions (1 mM KMnO 4 for 5 min (17,33,34)). Under these conditions, EC57, a regular EC, which was used as a control, was stable ( Fig. 2A, lanes 3 and 4). TC/Ni remained intact in TB40 (lanes 5 and 6), but it released most of the RNA after 5 min following incubation with 1 mM KMnO 4 in TB40 (lanes 7 and 8). The half-life of TC/Ni in these conditions (1 mM KMnO 4 ) was less than 2-3 min (data not shown); therefore, more than 80% of the complex would dissociate under standard footprinting conditions. Dissociation of TC/Ni would lead to significant attenuation of the signal from the unpaired thymidines relative to that of regular stable ECs. However, after 1 min of incubation with 1 mM KMnO 4 , TC/Ni remained intact (lanes 9 and 10). We chose this condition to probe for the transcriptional bubble. Fig. 2B shows KMnO 4 footprint of the nontemplate DNA strand in TC/Ni and in a series of ECs halted in the vicinity of the termination site. As expected, in EC46 the bubble was located at a position corresponding to the 3Ј-end of the RNA (lane 2). Elongation of the RNA to 57 nt was accompanied by forward progression of the bubble (lane 3). Extension of the RNA by another 5 nt (to form EC62) halts RNAP after the fifth nucleotide of the oligo(T) stretch of the tR2 terminator. In this position, RNAP becomes arrested and slides backwards along the DNA and the RNA (33,34), causing rearrangement of the bubble and its retreat to the more upstream position (lane 4). The arrest of EC62 is reversible (data not shown), and the RNAP with the DNA bubble returns infrequently to the origi- Lane 5 demonstrates the crucial result of this experiment. Although the amount of TC/Ni withdrawn for the footprinting reaction was the same as the amount of the other complexes (Fig. 2B, top panel), no DNA bubble was detected in the complex as revealed by the resistance of thymidines in TC/Ni to KMnO 4 (Fig. 2B, lane 5). Lanes 5 and 6 of Fig. 2B show that washing with 1 M KCl removed most of the RNA from TC/Ni, but the small fraction that remained generated the same footprint as the total TC/Ni. As we demonstrate below, this footprint originated from the minor salt-resistant fraction of ternary termination complexes described above (see also Fig. 1B,  lane 4). Thus, apart from this minor fraction, there was no transcription bubble detected in the majority of the DNA located in TC/Ni.
GreB-induced RNA Cleavage Testifies that RNA in TC/Ni Belongs to a Binary Complex with RNAP-Normally, the DNA duplex is melted inside the RNAP ternary complex (33)(34)(35).
Thus, it is extremely unusual if the TC/Ni represents a ternary complex that lacks the bubble. Previous studies had shown that low salt could stimulate formation of binary RNAP⅐RNA and RNAP⅐DNA complexes resulting from reassociation of RNAP with the nucleic acids released in the course of termination (27). Therefore, we performed a series of experiments to determine whether the nucleic acids in TC/Ni belong to ternary complexes or to binary complexes with RNAP.
First, we probed TC/Ni with protein factor GreB, which is known to stimulate cleavage of the RNA in ternary ECs and in binary RNAP⅐RNA complexes (12,33,36). GreB-dependent cleavage is carried out by the catalytic center of RNAP. When RNAP in the ternary complex backtracks on the DNA, its active center slides back from the 3Ј-end of the RNA to an internal position where the cleavage occurs (36 -38). It was suggested that in the binary complex the RNA is bound to the polymerase in a manner similar to that in the ternary complex (12). RNAP can bind the internal segment of the transcript, with the active center located upstream from the RNA 3Ј-end. In both ternary and binary complexes, the 3Ј-proximal end of the cleaved RNA dissociates from RNAP, and the 5Ј-proximal part remains bound to the enzyme and can be elongated in the presence of rNTPs. Importantly, the pattern of the elongation of the cleaved RNA is very different in the two complexes. In binary complexes, the GreB-truncated RNA is extended by no more than 1-3 nt, while in ternary ECs the extension of the product of the cleavage is unlimited and determined only by the length and the sequence of the DNA template (12,33,34,39).
In the experiment of Fig. 3A, we obtained TC/Ni with the RNA labeled in position ϩ12, which is located close to the 5Ј-end of the 64 -65-nt RNA (lane 7). Treatment of the complex with GreB produced two cleavage products (lane 8). To map the sites of the cleavage, the mobility of the products in PAGE was compared with the mobility of RNA markers obtained by walking up to the terminator region (lanes 1-6). Note that elongation of the RNA from 55 to 57 nt led to a significant increase in the RNA mobility in the gel. This mobility shift at the tR2 terminator was reported before (25) and is caused by pairing of the termination hairpin formed in the RNA 57 , which is not denatured in urea PAGE. Note that RNA 55 has the hairpin with the shorter stem, which can be denatured in the gel (lane 4). The two 5Ј-terminal cleavage products in TC/Ni were 55 and 56 nt long (lane 8). Note again that RNA 56 is not denatured and runs abnormally in the gel. Labeling of the RNA in the complex near the 3Ј-end revealed that the cleavage occurred at two distinct sites located 8 and 9 nt upstream from the 3Ј-end, supporting the generation of the RNA 55 and RNA 56 5Ј-labeled transcripts (Fig. 3B).
The incubation of the GreB-treated complex with either rCTP alone (Fig. 3A, lane 9) or with all four rNTPs (lane 10) caused elongation of the 5Ј-terminal products by 1 and 2 nt to form RNA 57 , which has increased mobility due to the hairpin formation. No other elongation products were observed. The limited elongation of the cleaved transcript argued that most of the RNA 64 -65 in TC/Ni belonged to binary but not to ternary complexes with RNAP.
In a separate experiment, we tested the ability of other rNTPs to elongate the cleaved RNA and found that only when rCTP was added did the RNA 55-56 extension occur (data not shown). This observation suggests that in the GreB-treated binary complex, in the absence of DNA, only 1 or 2 CMP nucleotides are added to the 3Ј-end. As suggested by the scheme of Fig. 3C, the RNA hairpin may be used as a template for this elongation. Fig. 3C summarizes the results of the cleavage site mapping in TC/Ni when RNAP binds to the duplex RNA segment of the termination hairpin.  1 and 2 and lanes 5 and 6). B, the mapping of transcription bubble in TC/Ni and surrounding complexes. EC46, -57, and -62 were obtained by walking the RNA-labeled EC12 in several consecutive steps (lanes 1-4). TC/Ni was obtained from EC57 by adding rATP and UTP at 5 M for 5 min followed by washing the pellet with TB40 (lane 5). In the low concentration of rNTPs, the termination efficiency was 90%; thus, only a small fraction of RNAP read through the terminator and reached the position ϩ66. A half of TC/Ni was withdrawn and washed with TB1000 as described under "Experimental Procedures" (lane 6). Bottom panel, the ECs were treated with KMnO 4 for 1 min as described under "Experimental Procedures." The scheme on the left shows the sequence of the nontemplate DNA strand in the region surrounding the tR2 terminator. The positions of all thymidines relative to the start site of transcription are numbered. The region of the dyad symmetry corresponding to the termination hairpin in the RNA is boxed.

Isolation of the Termination Complex by Immobilization of the DNA Template in a Solid
Phase-Another confirmation of the mostly binary nature of TC/Ni came from an experiment, in which termination was performed in the complexes immobilized on streptavidin-agarose beads via a 5Ј-biotinylated DNA (31). In this system, binary RNAP⅐RNA adducts should be released to the supernatant, while the ternary termination complexes and the binary RNAP⅐DNA adducts should remain in the pellet. EC12 containing both labeled RNA and labeled DNA was formed in solution using biotinylated template and then was immobilized either on Ni 2ϩ -NTA agarose (to RNAP) or on streptavidin-agarose beads (to DNA). After a chase with four rNTPs in TB40, each reaction mixture was centrifuged, the supernatants were removed (Fig. 3D, lanes 2 and 4), and the pellets were washed with TB40 (lanes 1 and 3). When EC12 was immobilized on streptavidin beads through the DNA, the vast majority of terminated RNA 64 -65 was found in the supernatant, and less than 10% of the RNA 64 -65 remained in the washed pellet (lanes 3 and 4). When the complex was immobilized on Ni 2ϩ -NTA-agarose through RNAP, about 40% of RNA 64 -65 remained in the pellet (lanes 1 and 2; see also Fig.  1B). The dramatic difference in retention of RNA 64 -65 in the two systems confirmed that the majority (about 80%) of the terminated transcript in TC/Ni belonged to the binary RNAP⅐RNA adduct, which contained no DNA and thus was not retained on streptavidin-agarose.
This indicated that most of the DNA, which was found in TC/Ni, also belonged to binary complexes with the separate  3) were analyzed on denaturing PAGE. Note that in this experiment the pellet was not washed before loading on the gel. Therefore, the excess of RNA 64 that dissociated from RNAP in a course of incubation with GreB was not removed. C, the scheme illustrates the position of the cleavage in the sequence of the terminated RNA and the result of the limited extension of the cleaved product with rCTP. The transparent oval shape and black circle represent RNAP and the enzyme active center, respectively. D, isolation of termination complex by DNA immobilization. TC was purified using either the RNAP immobilization on the Ni 2ϩ -NTA-agarose beads or the template immobilization on the streptavidin-agarose beads. After transcription, the supernatants (S) and washed pellets (P w ) were analyzed as described under "Experimental Procedures." The scheme below of the autoradiogram illustrates the alternative techniques used for the TC purification.
pool of RNAP molecules. This conclusion explains the failure to detect a transcription bubble in TC/Ni, because simple nonspecific electrostatic binding should not cause melting of the double-stranded DNA.
The minor fraction of EC64 -65 retained on streptavidinagarose beads through biotinylated DNA is now defined as termination complex isolated on streptavidin-agarose (TC/St). As we show below, this small fraction is in a ternary complex, is stable, and does have a transcription bubble in the DNA.
The Binary Complexes Do Not Represent Authentic Intermediates of the Termination Pathway-Binary complexes having the same catalytic properties as those found during termination can be reconstituted artificially from the purified core enzyme and labeled DNA and RNA. In the experiment depicted by Fig. 4A, template DNA 156 , RNA 64 -65 , and RNA 86 purified free from RNAP were collected from the supernatant fraction after Ni 2ϩ -NTA-immobilized EC12 was chased with four rNTPs in high salt (lane 2). Simple incubation of these nucleic acids with core RNAP preimmobilized on Ni 2ϩ -NTA-agarose caused their binding to the enzyme. When the unbound material was removed by washing, ϳ30% of the RNA and the DNA remained in the pellet (lane 3). TC/Ni and the de novo reconstituted complex showed identical patterns of GreB cleavage (compare lanes 1 and 4) and of the subsequent RNA extension with rNTPs (compare lanes 4 and 5 of Fig. 4A and lanes 8 and  10 of Fig. 3A). The apparent difference in the relative mobility of the 55-, 56-, and 57-nt RNAs in the gels of Figs. 3A and 4A is caused by the different percentage of the gels.
Did the binary complexes of TC/Ni derive from the reassociation of the components of the EC disintegrated at the terminator, or could they have been formed in the course of termination, through a selective dissociation of the DNA or RNA component? These two possibilities can be defined by altering the concentrations of the reaction components. If binary complexes are formed during termination by loss of a component (not by rebinding), it should not be dependent on concentration. Thus, the yield of TC/Ni should not depend on the total volume of the termination reaction and on the concentration of RNAP on the beads. To address the issue, we divided EC12 into three aliquots (transcript in EC12 was labeled in the ϩ12-position, to follow the RNA component of TC/Ni). The three samples contained the same amount of the complex immobilized on the same amount of Ni 2ϩ -NTA-agarose. The volume of the samples was adjusted with TB40 to 10, 50, and 200 l. After incubation with four rNTPs, the pellets were washed with TB40, and the yield of TC/Ni was determined in each sample. Lanes 2-4 ( Fig.  4B) demonstrates that retention of the RNA decreased while concentration of components decreased. This result indicated that the RNA in TC/Ni derived from secondary rebinding of the terminated RNA to RNAP.
The yield of TC/Ni was also compared in two samples of equal volume but containing the different amount of RNAP attached to the same amount of the beads (Fig. 4B, lanes 1 and  2 and lanes 5 and 6). As expected, the increased amount of RNAP synthesized a larger amount of transcript. In agreement with the result described above, an increase of the concentration of the two reassociating components (RNAP and RNA) caused almost complete retention of the transcript in TC/Ni.
Another group has observed that the yield of the termination complex that was isolated at tR2 terminator in low salt through RNAP immobilization on Ni 2ϩ -NTA-agarose varied among different preparations of the enzyme (17). We also found that the efficiency of the binary complex formation was dependent on the amount of core enzyme in the sample of RNAP (data not shown), which agreed with yet another observation that the subunit in the holoenzyme inhibited nonspecific binding of RNA and DNA (27).
In principle, the RNA retention should be the same at a given salt concentration for an RNAP sample with a fixed amount of the core enzyme. On the other hand, the fraction of the retained RNA may also be the same at two different salt concentrations for the two separate preparations of RNAP, containing different amounts of the core. For our standard preparation of the enzyme, the yield of TC/Ni at 40 mM KCl was very high (Ͼ40%), which justified our selection of this condition for the purification of the termination intermediates. The yield of TC/Ni increased even more at 5 mM KCl, but the properties of the binary RNAP⅐RNA complexes remained constant, such as sensitivity FIG. 4. The secondary nature of complexes stabilized at the tR2 terminator. A, RNA cleavage and extension with rNTPs in the binary RNAP⅐RNA complexes obtained on Ni 2ϩ -NTA-agarose by de novo assembly from the core enzyme and mixture of RNA 64 -65 , RNA 86 , and DNA 156 (see "Experimental Procedures"). Lanes 1 and 4 show the products of RNA cleavage with GreB in TC/Ni and in the binary RNAP⅐RNA complex, respectively. B, the dependence of the yield of TC/Ni on the concentration of RNAP and RNA in the transcription reaction. Lanes 1-4, 3 pmol of RNAP was used to obtain EC12 as described under "Experimental Procedures." 30 l of Ni 2ϩ -NTA-agarose suspension was used to immobilize the complex. The reaction mixture was divided into three aliquots (1 pmol of RNAP and 10 l of Ni 2ϩ -NTA-agarose in each); the volume of the samples was adjusted to 10, 50, and 200 l with TB40. After a 5-min incubation with all four rNTPs, supernatants were removed from each sample, and the pellets were washed with TB40. All four P w fractions (lanes 2-4) and S fraction from the 10-l reaction (lane 1) were analyzed on the denaturing PAGE. Lanes 5 and 6, 3 pmol of RNAP was used to obtain EC12 as described under "Experimental Procedures." 10 l of Ni-NTA agarose suspension was used to immobilize the complex. The reaction mixture was separated into S and P w fractions after chase with four rNTPs in a 10-l volume under the conditions described above.
to high salt and the pattern of GreB cleavage (data not shown).

Ternary TC/St Constitutes a Fraction of the Polymerase That Exists in Arrested
Complexes-Although the majority of complexes formed at low salt conditions during termination comprised binary reassociates of the RNAP to the nucleic acids, a small fraction of true ternary complexes (TC/St) was isolated (Fig. 3D, lane 3). The properties of these complexes were different from those of the binary RNAP⅐DNA and RNAP⅐RNA complexes of TC/Ni. First, TC/St was more stable then TC/Ni. It remained stable after 1-min exposure to 1 M KCl (data not shown) and dissociated significantly, but not completely, after 5 min of incubation in 1 M KCl (Fig. 5A, compare lanes 5 and 7). Note that the cleavage with GreB fully stabilized the complex (lanes 6 and 8). In contrast, TC/Ni completely dissociated after 5 min in 1 M KCl, and its stability only slightly increased after GreB treatment (lanes 1-4). Since the yield of TC/St was very small compared with that of TC/Ni (less than 10% of the amount of EC12), lanes 5-8 were exposed for a longer period of time to make the RNA visible.
It is evident from Fig. 5, A and B, that the transcript in both TC/Ni and TC/St was susceptible to cleavage with GreB. This observation testified that TC/St consisted of the ternary RNAP⅐RNA⅐DNA complexes rather than of the direct associates between DNA and RNA and that RNAP in the complex, as in TC/Ni, was active. In both TC/St and TC/Ni, GreB truncated transcripts to the same 55-56-nt length (Figs. 3A and 5, A and  B). However, the ability of the two complexes to elongate the truncated products was different. In the presence of all four rNTPs, the binary TC/Ni extended the cleaved RNA by 1-2 nt (Fig. 3A), while TC/St extended the cleaved product by 9 nt, back to the original termination point (Fig. 5B, lanes 3 and 4). The same 9 nt were added when only rATP, rCTP, and UTP were present, and the RNA was extended by only 2 nt after incubation with rCTP alone, which further confirmed the proper alignment of the 3Ј-end of the RNA and template DNA strand in TC/Ni (data not shown).
The experiment of Fig. 5C shows KMnO 4 footprinting of the nontemplate DNA strand in the complexes analyzed in Fig. 5B. The quantitative analysis of lanes 1 and 2 of Fig. 5B revealed that the yield of TC/St was less than 10% of the initial EC12. This explains substantial decrease of the signal from the bubble in TC/St compared with that in EC12 (Fig. 5C, lanes 2 and  3). TC/St carried RNA 64 and 65 nt long (Fig. 5B, lane 2). However, thymidine residues in the nontemplate DNA strand corresponding to the 3Ј-end of the RNA (Fig. 5, T58 -62 and T64 -65) were resistant to KMnO 4 , signifying that they were paired with the template strand. Instead, thymidines located upstream were modified, suggesting that the transcription bubble in TC/St had been translocated backward (Fig. 5C, lane  3). Isolation of the complex cleaved with GreB did not change the position of the transcriptional bubble (lanes 3 and 4).
Backward translocation of the bubble is known to accompany backsliding of RNAP in the course of transcriptional arrest both in vivo and in vitro (33,34,40,41). In a typically arrested EC, the RNA is in the proper register with the DNA, but the 3Ј end of the transcript is unpaired and extruded from the active center because RNAP has retreated to an upstream template position (33,34). GreB cleavage of the transcript in the arrested complex does not change the position of the enzyme and the bubble (33, 34) (see also Fig. 6). The lack of catalytic activity, its relatively high stability, the pattern of GreB cleavage, and the back-shifted transcriptional bubble all suggest that TC/St represents an arrested ternary EC in which RNAP  retreated upstream along the DNA and RNA (33,34). The similar pattern of the thymidine modification between the TC/St fraction (Fig. 5C, lane 3) and the TC/Ni fraction that was washed with TB1000 (Fig. 2B, lane 6) confirmed that after the binary component of TC/Ni was removed by incubation in high salt, the more stable arrested ternary complex remained. Therefore, only three kinds of RNAP complexes were formed in low ionic strength at the terminator: binary RNAP⅐RNA and RNAP⅐DNA complexes and the arrested ternary complex. All three were found in TC/Ni, and only the ternary and RNAP⅐DNA complexes were isolated as TC/St.
We believe that the arrest prevented dissociation of EC on the terminator because (i) the backslid arrested complex occupied the site with a strong RNA:DNA hybrid, with the 3Јterminal oligo(U) tail extruded from the polymerase, and (ii) the backsliding of the enzyme prohibited formation of the termination hairpin because the 3Ј-proximal arm of the hairpin was involved in the hybrid with DNA. We note that the stability of the arrested complex at the tR2 terminator was lower than the stability of elongation complexes arrested at sites outside of the terminator area (20,32). This lower stability may be explained by the fact that RNAP infrequently returns to the original location at the unstable U:dA heteroduplex from which it either dissociates or enters another round of arrest (lateral oscillations; Ref. 33). Low stability of the hybrid should promote dissociation, which can occur faster than the enzyme escapes back to the arrested state. Interestingly, arrest at the tR2 terminator has some unusual properties, different from those of conventional transcription arrest (33,34). Normally, treatment with GreB of the regular arrested EC leads to cleavage that allows RNAP to resume transcription elongation through the original arrest site when supplied with the rNTPs. RNAP arrested at tR2 (TC/St) responds to GreB differently: after the RNA cleavage and restart with rNTPs, TC/St goes back to but not beyond the site where it was arrested originally, which coincides with normal release site of tR2. Instead of dissociating or reading through the termination signal, TC/St becomes stuck and undergoes the second round of the same arrest by sliding back to the arrested position (Fig. 5C, compare  lanes 3 and 5). In this sense, arrest of TC/St is irreversible, and elongation of the GreB-treated complex is restricted to a short segment of the DNA.
A ternary termination complex with similar properties to this was observed previously at the tR2 terminator when RNAP was stopped at the point of termination by a protein roadblock imposed by the mutant EcoRI restriction endonuclease (25). In that system, the roadblock stabilized the termination complex, which remained sensitive to GreB and capable of elongating the cleaved RNA back to but not beyond the terminator. These properties suggest that the roadblocked termination complex escaped the dissociation by backtracking to the upstream DNA. The nature of the roadblock that may stop elongation and termination in TC/St is unknown. It may be a second RNAP molecule bound to the DNA downstream from the terminator. Because only a small fraction of the complexes undergoes this kind of arrest, some chemical damage or property of RNAP itself may make it deficient in termination and prone to arrest.
Conclusions-We conclude that the reduction of the ionic strength during transcription does not stabilize an intermediate of termination at the tR2 terminator. This is true at least to a level sufficient for purification and biochemical study of such an intermediate. Instead, these conditions provoke RNA and DNA that have dissociated at the terminator to rebind to RNAP, forming binary RNAP⅐DNA and RNAP⅐RNA adducts. Fig. 6 summarizes our results and illustrates the events that unfold after RNAP arrives at tR2 terminator. The major pathway, which dominates during termination, is the dissociation of the RNA from RNAP (pathway 1). In vitro, reassociation of the released product with the enzyme may occur (pathway 2). The reassociation can be prevented by diffusion of the nucleic acids away from the protein, by elevating the ionic strength of the solution, or by binding of -subunit to the RNAP core. In the cell, the terminated RNA is likely to be actively excluded from the reaction by translating ribosomes and by association with RNA-binding proteins. A ternary complex at the terminator can be rescued from dissociation reaction only by transcription arrest, which occurs as an alternative to the release of the RNA (pathway 3).