Early Transcriptional Arrest at Escherichia coli rplN and ompX Promoters*

Bacterial transcription elongation factors GreA and GreB stimulate the intrinsic RNase activity of RNA polymerase (RNAP), thus helping the enzyme to read through pausing and arresting sites on DNA. Gre factors also accelerate RNAP transition from initiation to elongation. Here, we characterized the molecular mechanism by which Gre factors facilitate transcription at two Escherichia coli promoters, PrplN and PompX, that require GreA for optimal in vivo activity. Using in vitro transcription assays, KMnO4 footprinting, and Fe2+-induced hydroxyl radical mapping, we show that during transcription initiation at PrplN and PompX in the absence of Gre factors, RNAP falls into a condition of promoter-proximal transcriptional arrest that prevents production of full-length transcripts both in vitro and in vivo. Arrest occurs when RNAP synthesizes 9–14-nucleotide-long transcripts and backtracks by 5–7 (PrplN) or 2–4 (PompX) nucleotides. Initiation factor σ70 contributes to the formation of arrested complexes at both promoters. The signal for promoter-proximal arrest at PrplN is bipartite and requires two elements: the extended −10 promoter element and the initial transcribed region from positions +2 to +6. GreA and GreB prevent arrest at PrplN and PompX by inducing cleavage of the 3′-proximal backtracked portion of RNA at the onset of arrested complex formation and stimulate productive transcription by allowing RNAP to elongate the 5′-proximal transcript cleavage products in the presence of substrates. We propose that promoter-proximal arrest is a common feature of many bacterial promoters and may represent an important physiological target of regulation by transcript cleavage factors.

Transcript cleavage factors such as Gre factors in bacteria and the SII factor in eukaryotes are ubiquitous in nature, but their physiological role(s) is poorly understood. Transcript cleavage factors stimulate RNA polymerase (RNAP) 2 intrinsic nucleolytic activity, an evolutionarily conserved function of all multisubunit RNAPs (reviewed in Refs. 1 and 2). In vitro, Gre/ SII factors suppress or prevent transcription pausing and transcription arrest during elongation (3,4), enhance transcription fidelity (5,6), facilitate promoter escape by RNAP (7)(8)(9), and assist in transcription-coupled DNA repair (10). During transcription in vitro and in vivo, RNAP can become arrested upon encountering a roadblock such as a DNA-binding protein or a special DNA sequence (1,11). Some roadblocks induce RNAP to slide backwards along the template ("backtrack") (11)(12)(13)(14), forcing the nascent RNA 3Ј-terminus to extrude through the RNAP secondary channel, typically by 2-18 nt (1-4, 12, 13). Endonucleolytic cleavage of such extruded RNA by the RNAP catalytic center generates a new 3Ј terminus that can be extended in the presence of rNTPs, giving RNAP another chance to read through the roadblock (11)(12)(13)(14). RNA cleavage by RNAP alone is inefficient under physiological conditions but is markedly stimulated by Gre/SII or, in the absence of the factors, by alkaline pH or pyrophosphate (15)(16)(17).
Transcript cleavage factors contribute to transcription proofreading in vitro because incorporation into nascent RNA of nucleotide analogs that disrupt the RNA-DNA hybrid induces backtracking, generating transcription complexes that are subject to Gre-induced RNA hydrolysis (12,18,19). The cleavage activity of Gre factors is also part of the cellular mechanism that resolves conflicts between transcription and the processes of DNA repair/replication. Gre factors help remove stalled or backed up RNAPs from sites of chromosomal lesions caused by UV irradiation or chemicals and thus provide access of DNA repair/recombination apparatus to the lesion sites (20). Gre factors also help resolve collisions between replication forks and RNAP stalled at transcription terminators during co-directional replication and transcription (21).
The transition from transcription initiation to transcription elongation is also facilitated by Gre-induced transcript cleavage (7,9). Gre factors dramatically reduce the amounts of abortive products, both in vitro and in vivo (7,9,(22)(23)(24), and induce substantial cleavage of abortive transcripts (7,22). Recent studies suggested that DNA may undergo "scrunching" during abortive initiation, which could account for long (Ͼ9 nt) abortive products synthesized by RNAP at some promoters. Stress accumulated during scrunching (25,26) is thought to be relieved either through disruption of RNAP-promoter contacts, leading to RNAP escape from promoter or, through a reversal of scrunching, to dissociation of the abortive transcript through the RNAP secondary channel. While inducing endonucleolytic cleavage of abortive transcripts, Gre factors are thought to simultaneously prevent the 5Ј-terminal RNA fragment from dissociating, thus increasing the fraction of RNAP bound to nascent RNA and stimulating promoter escape (7,24,27).
Previous microarray studies identified 31 Escherichia coli operons that are up-regulated by GreA in vivo (23). In vitro analysis of 10 randomly selected promoters that controlled genes whose expression was increased by GreA revealed that the factor stimulated E. coli RNAP transcription from these promoters at the stage of transcription initiation and/or promoter escape (23). The exact mechanism of Gre-induced transcription stimulation was not defined. To reveal this mechanism, we chose two promoters previously shown to be stimulated by GreA in vivo and in vitro: the outer membrane protein gene ompX promoter PompX and ribosomal proteins operon rplN-rpmJ promoter PrplN, and characterized in detail in vitro transcription products formed on these promoters, both in the presence and in the absence of transcript cleavage factors. Our data show that on these promoters and, by extension, on other Gre-dependent promoters, a novel type of transcription complex, a promoter-proximal arrested complex, is formed with high efficiency in the absence of Gre factors. Initiation factor contributes to formation of promoter-proximal arrested complexes at both promoters studied. At least in the case of PrplN, the signal for promoterproximal transcriptional arrest is bipartite and requires the presence of elements upstream and downstream of the transcription start site. Because RNAP is unable to escape from promoter-proximal arrested complexes, their formation leads to occlusion of the initial transcribed sequence of the promoter and, ultimately, to cessation of productive transcription. Through their transcript cleavage activity, Gre factors either rescue promoter-proximal arrested complexes or prevent their formation, thus allowing efficient transcription and gene expression. Our analysis, along with the previous transcript profiling data, suggests that rescue of promoter-proximal arrested complexes may be one of the main physiological function of Gre factors.
Reporter plasmids prplN-lacZ and pompX-lacZ were derivatives of pBAD/Myc-His/LacZ (Invitrogen) with the following modifications. The original pBAD/Myc-His/LacZ vector carrying ColEI origin of replication and ampicillin resistance marker was digested with SphI and NcoI to remove araC and ParaBAD promoter. The resulting 5923-bp fragment was gel-purified and ligated with SphI-NcoI PCR fragment containing PrplN [Ϫ72; ϩ46] followed by XbaI site and ϳ30 bp-long sequence of Shine-Dalgarno element from the phage X174 gene E (used to decrease toxicity of LacZ expression under strong PrplN (31), yielding prplN-lacZ. The pompX-lacZ was obtained by replacing the SphI-XbaI PrplN fragment in prplN-lacZ with a SphI-XbaI PCR fragment containing PompX [Ϫ175; ϩ129]. Gre expression plasmids (pBAD-GreAwt and pBAD-GreA-D41E) compatible with reporter plasmids described above were constructed using pBAD33 vector (Invitrogen) carrying the P15 origin of replication and spectinomycin resistance. Expression of wt and mutant GreA was under the control of arabinoseinducible Para promoter. The sequences of primers used for construction of PCR amplification and cloning of PrplN and PompX DNA fragments into reporter plasmids and of wt greA and greA-D41E genes into compatible expression vectors are available upon request.
The  Fig. 6 for details).  32 ]GTP (4500 Ci/mmol) in the absence of Gre factors for 10 min at 37 C. The reaction mixture was applied on G-50 Quick-Spin gel filtration columns to remove NTPs and abortive RNA. For transcript cleavage and NTP-chase experiments, a 4-l aliquot of purified arrested TCs was incubated in 10 l of TB alone or together with 4 M GreA, (or 0.4 M GreB) in the absence or presence of 100 M NTPs for 10 min at 37°C (see Fig. 1A). The reactions were terminated by the addition of SB and analyzed as described above.

In Vitro Transcription Reactions
TCs arrested at PompX were prepared using 10 l of nickelnitrilotriacetic acid-agarose beads carrying 10 pmol of immobilized His-tagged RNAP holoenzyme in 120 l of TB containing 30 M NTPs, 0.165 M [␣-P 32 ]ATP (3000 Ci/mmol), and 3 pmol of template DNA [Ϫ188;ϩ65]. After incubation for 15 min at 37°C, a 10-l aliquot was withdrawn, and the reaction was stopped with SB. The remaining 110 l of bead suspension were washed with 3 ϫ 1 ml of TB to remove NTPs and abortive RNA products, and the suspension was divided into three portions and treated with TB alone, 4 M GreA, or 0.4 M GreB. After incubation for 5 min at 37°C, 10-l aliquots were withdrawn, and the reactions were terminated by SB. The remaining beads were washed again with 3 ϫ 1 ml of TB to remove dissociated 3Ј-terminal RNA cleavage products, and the cleaved TCs were chased with 100 M NTPs (see Fig. 1B).

DNA Footprinting and RNA Mapping
Localized Fe 2ϩ -mediated Hydroxyl Radical Mapping-Mapping of the position of the catalytic center in promoter-proximal arrested TCs at PrplN, by localized Fe 2ϩ -mediated hydroxyl radical cleavages was performed as described (12). Arrested TCs were prepared under same conditions as described above, purified by Quick-Spin Sephadex G-50 (see Fig.  1A), and equilibrated in chelated TB without Mg 2ϩ . Hydroxyl-radical RNA cleavage reactions were carried out by incubation of arrested TCs in 10 l of hydroxyl cleavage buffer (20 mM Na-HEPES, pH 7.5, 150 mM NaCl, 0.1 mg/ml bovine serum albumin) containing 0.1 mM (NH 4 ) 2 Fe(SO 4 ) 2 and 1 mM dithiothreitol for 20 min at 30°C. In control reactions, arrested TCs were incubated as above but in the presence of 10 mM MgCl 2 or 50 mM EDTA (see Fig. 2). The reaction was stopped by the addition of 3 volumes of 7 mM thiourea solution and washed by chloroform, and the RNA products were precipitated by ethanol, dissolved in SB, and resolved on denaturing 23% PAGE.
In Vitro DNA Footprinting by KMnO 4 -Binary and ternary transcription complexes formed on PrplN and PompX were probed with KMnO 4 as described (33). 1 l of 10 mM KMnO4 was added to 10 l of transcription complex. After incubation for 40 s, the reaction was stopped with 1.5 l of ␤-mercaptoethanol followed by chloroform extraction and DNA precipitation by ethanol. The resulting DNA was dissolved in 10 l of 10 mM Tris-HCl, pH 7.5, and used as a template for primer extension reaction using radiolabeled primers (same as those used for PCR and cloning of PrplN and PompX into pES-rplN or pES-ompX) in 10 linear PCR cycles. The reaction was stopped by SB, and the modified thymidines were visualized by 6% sequencing PAGE.
KMnO 4 Reactivity in Vivo-The reactivity of plasmid DNAs carrying PrplN and PompX to KMnO 4 in vivo was determined essentially as described (29). The reactions were conducted in the E. coli greA Ϫ /greB Ϫ strain BW32645 or the isogenic greA ϩ greB Ϫ strain BW32613 (23) carrying reporter plasmid pES-rplN or pES-ompX. The cells were grown in 9 ml of LB medium to A 600 ϭ 0.5, then 1 ml of 100 mM KMnO 4 was added, and the cells were shaken for 3 min at 37°C. The treated cells were pelleted, placed on ice, resuspended in 300 l of STET buffer (50 mM Tris-HCl, pH 8.0, 50 mM EDTA, 8% sucrose, 5% Triton X-100) containing 200 g of lysozyme. After 10 min of incubation, the cell suspensions were heated at 95°C for 2 min and centrifuged for 10 min. The supernatants were treated with RNase A, and the modified pDNA was purified by QIAquick spin columns (Qiagen; QIAquick nucleotide removal kit). For pDNA quantity normalization in all samples, dot blot hybridization was carried out along with ethidium bromide staining of linearized pDNA after agarose gel electrophoresis. The modified thymidines were detected by primer extension reactions using radiolabeled primers (same as those used for PCR and cloning of PrplN and PompX into pES-rplN or pES-ompX) and analyzed by 6% sequencing PAGE as above.
␤-Galactosidase Assay-For measurements of ␤-galactosidase activity, cultures of E. coli greA Ϫ /greB Ϫ strain BW32645 co-transformed with reporter plasmid prplN-lacZ or pompX-lacZ together with Gre expression plasmids pBAD-GreAwt or pBAD-GreA-D41E or control vector pBAD33 were grown in LB to A 600 ϭ 0.5. 100 l of cell culture were centrifuged and resuspended in 100 l of permeabilization solution (100 mM Na 2 HPO 4 , 20 mM KCl, 2 mM MgSO 4 , 0.8 mg/ml hexadecyltrimethyl-ammonium bromide, 0.4 mg/ml sodium deoxycholate, 5.4 l/ml ␤-mercaptoethanol) containing 100 g/ml of chloramphenicol (to terminate translation). After 30 min of incubation at 37°C, 300 l of prewarmed substrate solution (60 mM Na 2 HPO 4 , 40 mM NaH 2 PO 4 , 1 mg/ml o-nitrophenyl-␤-D-galactoside, 2.7 l/ml ␤-mercaptoethanol) was added, and the reactions were incubated at 37°C until solution developed a stable pale yellow color. The reactions were terminated by the addition of 500 l of 1 M Na 2 CO 3 , and the incubation time was recorded. The samples were centrifuged to remove cell debris, the A 400 was measured for the supernatant, and the Miller units were calculated as described (30).

Stable Inactive Promoter-proximal Transcription Elongation Complexes Are Formed on rplN and ompX Promoters during in
Vitro Transcription in the Absence of Gre Factors-In experiment shown in Fig. 1A, the effect of GreA and GreB on in vitro transcription from an rplN promoter-containing linear DNA fragment was investigated. Transcription reactions were performed under conditions described in the previous work (23) but with two important modifications (see also "Experimental Procedures"). First, the length of transcription unit was much shorter than before, which allowed simultaneous detection of abortive and full-length transcripts on the same gel. The second difference was that transcripts initiated from PrplN were 5Ј-terminally labeled with [␥-32 P]GTP, which allowed direct comparison of the relative amounts of short and long transcripts synthesized. As can be seen, almost no full-sized rplN transcript was formed by RNAP in the absence of cleavage factors (Fig. 1A, lane 3). Instead, a series of transcripts whose lengths ranged from 9 to 14 nt were formed, with 11-nt-long transcript being the most prominent. The absence of full-sized transcripts was not caused by a general inability of the RNAP holoenzyme (which was purified from greA Ϫ greB Ϫ cells) to transcribe productively, because the enzyme efficiently produced fulllength run-off products on several control templates harboring promoters such as T7 A1, T5 N25, and rrnB P1 (data not shown). In the presence of either GreA or GreB, the abundance of short transcripts significantly decreased (lanes 2 and 1, respectively); concomitantly, the full-sized rplN transcript appeared.
To determine the nature of short rplN transcripts, reactions carried out in the absence of Gre factors were passed through a Sephadex G-50 column, which removes unincorporated NTPs and RNA not associated with RNAP from transcripts that are part of the transcription complex (TC). As can be seen (Fig. 1A, lane 4), most of NTPs were retained by column (as judged by the disappearance of [␥-32 P]GTP), whereas most of the 9 -14-nt-long transcripts eluted in the void volume, indicating that they are stably associated with RNAP. Nevertheless, transcripts in purified TCs could not be chased by the addition of NTPs (lane 5), indicating that rplN TCs containing the 9 -14-ntlong transcripts are transcriptionally inactive. The addition of GreB to inactive complexes led to a substantial decrease in the amount of 9 -14-nt-long transcripts (lane 8) and accumulation of radioactive (and, therefore, 5Ј-terminal) 4 -6nt-long cleavage products. The effect of GreA was similar (lane 6) but much less pronounced. Radioactive cleaved RNAs remained associated with RNAP and could be extended into run-off transcripts by the addition of NTPs (lanes 7 and 9). Again, the stimulatory effect of GreA on full-sized transcript synthesis was much less than that of GreB (and appeared to correlate with the extent of accumulation of transcript cleavage products).
Similar analysis was carried out for transcripts generated from an ompX promoter-containing linear DNA fragment (   DECEMBER 18, 2009 • VOLUME 284 • NUMBER 51 from terminally labeling PompX transcripts. Instead, as in the previous work, we used internal label [␣ 32 -P]ATP, which is first found in the ϩ3 position of the transcript. Both Gre factors strongly stimulated production of full-sized transcripts from the ompX promoter at conditions of a steady-state multipleround transcription assay (Ref. 23 and data not shown). In the absence of Gre factors, several short transcripts were synthesized, among which 3-, 5-, and 12-nt-long transcripts were most prominent. However, unlike in the case of transcription from PrplN, the full-sized ompX transcript was also produced (Fig. 1B, lane 1). Because of the presence of adenines in multiple positions of the ompX transcription unit, specific radioactivity of the run-off product is much higher than that of the 12-ntlong RNA (lane 1). For this reason and because of our inability to cleanly separate the run-off product from ompX TCs by Sephadex G-50 gel filtration, we identified transcripts associated with RNAP by performing immobilized in vitro transcription reactions using hexahistidine-tagged RNAP and nickelnitrilotriacetic acid-agarose (35). As can be seen (lane 2), washed immobilized transcription reactions lacked unincorporated [␣-32 P]ATP and 3-10-nt-long RNAs that must correspond to true abortive transcripts. In contrast, 11-13-nt-long RNAs remained stably associated with RNAP; a substantial amount of full-sized transcripts also remained bound to RNAP, which we attribute to the formation of non-native complexes at the end of the DNA template (36). The addition of Gre factors to washed immobilized TCs led to the appearance of cleavage products ranging in length from 9 to 2 nt (lanes 4 and 5). Unlike the situation observed with PrplN, both GreA and GreB were similarly active in ompX transcripts cleavage, although the pattern of cleavage products differed. Complexes containing the full-sized transcript and presumably stalled at the end of the template were also subject to factor-dependent cleavage, although this was not investigated further.

Promoter-proximal Transcriptional Arrest
An additional round of washing of immobilized PompX TCs after Gre-induced cleavage revealed that short cleavage products were removed, whereas longer, 7-9-nt-long products remained associated with TCs (lanes 7 and 8). Thus, in the case of PompX, the cleavage occurs within 2-4 nucleotides from the 3Ј end of RNA in stalled transcription complexes (recall that PompX transcripts are labeled with [␣-32 P]AMP in positions ϩ3, ϩ6, and ϩ10, as indicated in the scheme shown at the top of Fig. 1B).
The addition of NTPs to washed TCs containing 11-13nt-long transcripts did not lead to transcript elongation (lane 9). In contrast, TCs containing 7-9-nt-long RNA cleavage products could be chased (lanes 10 and 11). Chasing of washed TCs obtained after GreA cleavage led to very efficient formation of complexes stalled at positions ϩ12/ϩ13 (lane 10). No such complexes were formed when washed TCs formed upon GreB cleavage were chased (lane 11, only full-sized transcripts were observed). Although this was not further investigated, we attribute this difference to the lower affinity of GreA to RNAP, which leads to its complete removal during the washing step. In contrast, GreB, which binds RNAP much stronger (37,38), is apparently able to at least partially withstand the washing procedure. Residual GreB then allows RNAP to read through the block at positions ϩ12/ϩ13 of the ompX transcription unit.
In summary, the results of our analysis indicate that on the two Gre-dependent promoters studied here, promoter-proximal inactive TCs are formed. These inactive/stalled TCs can be reactivated by the addition of Gre factors and may thus correspond to previously described dead end or arrested complexes (3,12,13,39). When analysis similar to the one described above was performed with several templates driven by well studied E. coli RNAP promoters such as T7 A1, rrnB P1, and T5 N25, no stalled or arrested promoter-proximal TCs that were sensitive to Gre factors could be detected (data not shown). Thus, formation of promoter-proximal stalled TCs may be a common feature of those promoters whose activity is stimulated by Gre factors.
Inactive/Stalled Promoter-proximal TCs Formed on PrplN and PompX Are backtracked-Previously characterized arrested TCs were shown to form when the RNAP catalytic center disengages from the 3Ј end of the nascent transcript and the enzyme backslides along the DNA (12,13). The "extra" RNA proximal to the 3Ј end is likely accommodated in the RNAP secondary channel, which also accepts the transcript cleavage factors (28,40). To determine the position of the catalytic center in promoter-proximal inactive/stalled TCs formed at PrplN, we performed localized Fe 2ϩ -mediated hydroxyl radical cleavage reactions (12) using gel-purified 5Ј-terminally radiolabeled stalled TCs containing Fe 2ϩ instead of Mg 2ϩ in RNAP catalytic center (Fig. 2). Under conditions when hydroxyl radicals were generated by the addition of dithiothreitol, radioactive RNAs of 4, 5, and 6 nt in length started to accumulate (Fig. 2, lane 4). These RNAs were not the products of intrinsic nucleolytic activity of RNAP because no cleavage was observed in the presence of Mg 2ϩ (lane 3). Also, the appearance of these products was decreased when reactions were supplied with Fe 2ϩ chelator EDTA (lane 5), indicating that the observed cleavage products were specifically caused by Fe 2ϩ ions. The efficiency of Fe 2ϩmediated reactions was relatively low (1-2% of the initial inactive/stalled TCs), which is likely to be due to degradation of the RNAP ␤ and ␤Ј subunits and subsequent inactivation of RNAP by hydroxyl radicals (41). The sizes of cleaved 5Ј-radiolabeled RNAs match the sizes of cleavage products observed when stalled PrplN TCs are treated with Gre factors (Fig. 1A). Based on these results, we conclude that stalled promoter-proximal TCs on PrplN are backtracked by ϳ5-7 bp.
To reveal the size and position of the transcription bubble in promoter-proximal stalled TCs, they were probed with KMnO 4 followed by primer extension analysis (Fig. 3). In complexes formed by RNAP on PrplN in the absence of NTPs, little or no KMnO 4 -sensitive bands were detected (Fig. 3A, lane 3; see also supplemental Fig. S1). This was due to instability of PrplN open promoter complexes (data not shown), a situation previously observed for ribosomal RNA promoters (30). The addition of GTP, CTP, and UTP to the reaction allows transcription to proceed to G ϩ7 (position ϩ8 in the PrplN initial transcribed sequence specifies an A; Fig. 1A), which leads to formation of an active promoter-proximal complex carrying a 7-nt-long RNA (TC-7G; supplemental Fig. S1). KMnO 4 probing of TC-7G revealed that in the nontemplate DNA strand there were two sensitive thymidines at positions ϩ5, and ϩ9 and one weakly sensitive thymidine at position Ϫ6 relative to the transcription start point (Fig. 3A, lane 4). Additional bands seen on the gel and corresponding to cytidines at positions ϩ3/ϩ4 and thymidine at ϩ11 were not the result of KMnO 4 modification but were due to stalling of DNA polymerase at and around the modified thymidines during primer extension reaction (see "Experimental Procedures"). Consistent with this interpretation, bands corresponding to these cytidines were not observed during direct visualization of KMnO 4 -induced modification by piperidine treatment of end-labeled DNA template (supplemental Fig. S1). The pattern of KMnO 4 modification in TC-7G did not change in the presence of GreA (lane 5). Probing of reactions supplied with all four NTPs, a condition when inactive promoter-proximal complexes form (Figs. 1A and 2), revealed a similar pattern of modification, although the reactivity of thymidine at downstream position ϩ9 appeared to be decreased (Fig. 3A, lane  7). Importantly, the addition of GreA led to almost total disappearance of KMnO 4 -sensitive bands (lane 7). The addition of GreA-D41E, a mutant that binds transcription complexes normally but is unable to induce transcript cleavage (23,28)

had no such effect (lane 8).
KMnO 4 probing of complexes on PompX was also performed ( Fig. 3B and supplemental Fig. S2). Unlike PrplN, KMnO 4 sensitivity of thymidines within open promoter complexes (i.e. in the absence of NTPs) is readily observed on this promoter (Fig.  3B, lane 6). As revealed by primer extension analysis, thymidines at template strand positions Ϫ11 and Ϫ9 and partially at positions Ϫ5/Ϫ2 are sensitive to the reagent. In the presence of NTPs, a condition when stalled promoter-proximal complexes form (Fig. 1B), the sensitivity of thymidines at positions Ϫ11 and Ϫ9 to KMnO 4 was abolished. Instead, thymidines at positions Ϫ2/Ϫ5 become highly reactive (Fig. 3B, lane 5). The addition of GreA (lane 4) but not the inactive D41E mutant (lane 3) led to disappearance of KMnO 4 sensitivity at positions Ϫ5/Ϫ2. Neither factor appreciably altered the residual sensitivity of thymidines at positions Ϫ11 and Ϫ9, which must be due to formation of unproductive open complexes that fail to leave the promoter in the presence of NTPs.
The summary of KMnO 4 probing on both the template and nontemplate strands for each of the two promoters studied here is shown in the right panels of Fig. 3 (A and B, respectively) (see also supplemental Figs. S1 and S2). On PrplN, upon transition of transcriptionally active complex TC-7G into stalled inactive TCs, the transcription bubble shrinks in size from ϳ18 -19 to ϳ15 bp; the front edge of the bubble moves backward from ϩ8/ϩ9 to ϩ5, whereas the rear edge of the bubble remains at Ϫ10/Ϫ11 (Fig. 3A, right panel). On the other hand, during transition from an open promoter complex to stalled inactive TCs on PompX, the transcription bubble decreases in size from ϳ14 to ϳ12 bp, whereas both the rear and the front edge of the bubble move forward from Ϫ10/Ϫ11 to Ϫ2 and from ϩ1/ϩ3 to ϩ9/ϩ10, respectively (Fig. 3B, right panel). Together, the data presented are consistent with the idea that on both promoters, stalled promoter-proximal TCs are backtracked. This conclusion is further supported by the results of ExoIII nuclease footprinting experiments designed to reveal the positions of the front and rear edges of RNAP on DNA in TCs stalled at PompX and PrplN (data not shown). Because promoter-proximal stalled TCs are 1) unable to elongate nascent transcripts even upon prolonged incubations with high concentrations of NTPs, 2) rescued by transcript cleavage factors, and 3) backtracked, they indeed correspond to previously described arrested elongation complexes (3,12,13,39).
The Strength of the 70 Subunit Interaction with the RNAP Core Affects the Formation of Promoter-proximal Arrested TCs-Previously described promoter-proximal RNAP pausing (29,32,(43)(44)(45)(46) is caused by "hopping," a process when the subunit region 2, as a part of the RNAP holoenzyme, recognizes a region in the initial transcribed sequence of the promoter that is similar to the Ϫ10 promoter element consensus sequence (29). During hopping, the contacts of with the bona fide Ϫ10 promoter element are broken, and new contacts are established with the downstream "Ϫ10-like" element, which is present in a single-stranded form because it is part of the transcription bubble. The interaction of with the "sticky" downstream Ϫ10 like element leads to pausing. Inspection of the PrplN and PompX initial transcribed sequences did not reveal any similarities with the Ϫ10 promoter element consensus sequence. Likewise, analysis of the initial transcribed sequences of other GreA-responsive promoters (23) revealed no similarities to the Ϫ10 consensus sequence (we also failed to detect any common sequences for these promoters other than promoter consensus elements, data not shown). To elucidate the role of the subunit in the formation of promoter-proximal arrested TCs on Gre-stimulated promoters, we performed in vitro transcription experiments using RNAP holoenzyme reconstituted with 70 harboring a L402F substitution. Previous work demonstrated that this substitution decreases the strength of 70 interaction with the core enzyme (29). As can be seen from Fig. 4A, during multiround transcription on the PompX template in the absence of Gre factors, RNAP holoenzyme harboring L402F produced considerably fewer transcripts corresponding to promoterproximal arrested TCs (and more run-off products) than the wt holoenzyme ( compare lanes 1 and 3). On PrplN, the amount of RNAs corresponding to promoter-proximal arrested TCs was similar for both holoenzymes; however, ϳ3.5 times more of the full-sized transcripts were produced by the mutant RNAP (compare lanes 5 and 7). The effect was specific, because both holoenzymes transcribed equally well from the T7 A1 promoter (data not shown). The addition of GreA decreased the amount of arrested TCs formed and increased the production of full-sized transcripts by wt RNAP on both templates, as expected (compare lanes 1 and 2  with lanes 5 and 6). The addition of GreA prevented transcriptional arrest of L402F-RNAP on PrplN and increased the amount of run-off (compare lanes 7 and 8), although not as much as in the case of the wt RNAP (compare lanes 5 and 6). The significant amount of short (10 -13-nt-long) transcripts formed by the L402F-RNAP on the PrplN template (Fig. 4A, lane 7) could, in principle, be due to decreased stability of promoter-proximal arrested TCs formed by the mutant holoenzyme. Such decreased stability would lead to increased production of full-length transcripts by allowing faster RNAP recycling. If promoter-proximal arrested TCs formed by the mutant RNAP were less stable, then in a single-round transcription assay, open promoter complexes formed by the mutant enzyme would have produced fewer fullsized transcripts than the same amount of open complexes formed by the wt RNAP. Further, in this scenario, RNA in stalled complexes formed by the wt RNAP should remain bound to the enzyme, whereas corresponding RNAs formed by the mutant enzyme should dissociate from the complex. In Fig.  4 (B and C), the results of a single-round transcription experiment on PrplN are presented. A mixture of NTPs and rifampicin, a drug that prevents RNAP escape into elongation, was added to preformed PrplN open complexes formed by the wt or the mutant holoenzyme. RNAP molecules that manage to elongate RNA past trinucleotide become resistant to rifampicin and can complete a single round of transcription; however, further rounds of productive transcription are prevented. As can be seen, the mutant RNAP produced fewer transcripts corresponding to promoter-proximal arrested TCs and more fullsized transcripts than the wt RNAP (compare lanes 1 and 4). Quantitative analysis showed that although only ϳ12% of the wt RNAP molecules reach the end of the template with 88% becoming arrested, more than 50% of transcripts initiated by the mutant RNAP holoenzyme became full-sized (Fig. 4C). These results suggest that by weakening 70 -core interactions, the L402F mutant improves the readthrough efficiency (i.e. transcripts initiated by the mutant RNAP fall into the arrested state at promoter-proximal sites less often), thereby accelerating promoter escape and increasing productive transcription from Gre-dependent promoters studied here. Indeed, in the presence of NTPs, KMnO 4 modification of thymidines at nontemplate strand positions ϩ5/ϩ9 of PrplN and ϩ8/ϩ9 of PompX indicative of arrested TCs was much less efficient in reactions containing the mutant holoenzyme than in the wt RNAP-containing reactions. Nevertheless, in both cases, the KMnO 4 -reactive thymidines were sensitive to GreA and GreB (supplemental Fig. S3). Thus, we conclude that formation of promoter-proximal arrested TCs is determined, at least in part, by the strength of 70 interaction with the core enzyme and may therefore involve hopping despite the apparent absence of sequences that could be recognized by in arrested TCs.
Promoter-proximal Arrested TCs Are Formed on PrplN and PompX in Vivo-Using plasmid-borne PrplN-and PompX-lacZ fusion reporter constructs, we determined that wt GreA but not the cleavage-defective GreA-D41E mutant stimulates the production of ␤-galactosidase ϳ7and ϳ2-fold, respectively (Fig.  5A). To determine whether GreA-dependent stimulation of PrplN and PompX involves formation of promoter-proximal arrested complexes in vivo, KMnO 4 probing of greA Ϫ greB Ϫ or isogenic greA ϩ greB Ϫ cells harboring PrplN-and PompX-reporter plasmids was carried out (Fig. 5B). As can be seen, on both promoters robust KMnO 4 -sensitive thymidines were only observed in the absence of Gre factors. In the case of PrplN, KMnO 4sensitive thymidine at position ϩ5 of the nontemplate strand (Fig. 5B,  left panel, lane 3) corresponded to the one observed in vitro for promoter-proximal arrested TCs (compare with Fig. 3A). In the case of PompX, KMnO 4 -sensitive thymidines at positions Ϫ2/Ϫ5, Ϫ9, and Ϫ11 of the template strand (Fig. 5B,  right panel, lane 3) 5 and 6). We therefore, conclude that 1) promoter-proximal arrested TCs form on PrplN and PompX in vivo and 2) GreA, through its transcript cleavage activity, reduces the formation of these complexes and thereby stimulates gene expression. By extension, we propose that Gre-dependent stimulation of transcription not only from PrplN and PompX but from at least some other Gre-dependent promoters revealed in our previous work (23) is due to suppression of formation and reactivation of promoter-proximal arrested TCs.
Signal for Promoter-proximal Arrest at PrplN Is Bipartite and Requires Two Elements: the Extended Ϫ10 Box and the Initial Transcribed Region from ϩ2 to ϩ6-To investigate the role of different promoter elements in formation of arrested TCs at PrplN, we constructed a set of 15 promoter variants where the upstream (promoter region) and the downstream (initial transcribed region, ITR) segments of PrplN are replaced either with the corresponding parts of the T7 A1 promoter or with segments of random DNA sequence (Fig. 6). The transcriptional activities of resulting promoter variants were analyzed by single-and multiround run-off transcription assays in the absence or in the presence of Gre factors. The ability of different promoter variants to induce formation of promoter-proximal arrested TCs (and susceptibility of arrested TCs to GreA) was examined by comparing the patterns of short RNA products (5-20 nt long) in transcription reactions before and after puri- fication through Quick-Spin gel filtration columns (see "Experimental Procedures").
First, we found that swapping the upstream (from Ϫ84 to ϩ1, construct 14) or the downstream (from ϩ2 to ϩ45, construct 9) parts of PrplN with the corresponding parts of T7 A1 had only a small (1.7-2.5-fold) stimulatory effect on overall transcriptional activity but effectively eliminated transcriptional arrest at both hybrid promoters (Fig. 6, compare constructs 1, 9, 14, and 15). Thus, the signal for promoter-proximal transcriptional arrest is (at least) bipartite and is located both in the upstream and the downstream portions of PrplN. More detailed analysis of the upstream part of PrplN revealed that deletion of A/T-rich "UP-like" sequence (located around position Ϫ45, construct 3), alone or together with the Ϫ35 element (construct 4), or introduction of the consensus Ϫ35 element sequence into PrpllN (construct 2) had only a moderate effect on promoter activity or efficiency of transcriptional arrest (constructs 2-4). Similar results were observed when we introduced multiple substitutions in the discriminator region (positions Ϫ5 to Ϫ1) or replaced a G at the transcription start site (ϩ1) to an A (constructs 6 and 7, correspondingly). However, substitu-tion of a TG dinucleotide upstream of the Ϫ10 element (positions Ϫ13/Ϫ14) with a CC dinucleotide caused Ͼ20-fold decrease in promoter activity and the disappearance of arrested TCs (construct 5). The result indicates that PrplN belongs to the "extended Ϫ10" promoter class and that the extended TGelement constitutes a part of the signal for promoter-proximal arrest. Other upstream promoter elements of PrplN do not play a significant role in either promoter activity or early transcriptional arrest. Further dissection of the downstream (ITR) part of PrplN using substitutions in 5-bp-long segments encompassing a region from ϩ2 to ϩ21 revealed that changes in four positions (ϩ2, ϩ4, ϩ5, and ϩ6) in the wt ITR sequence led to nearly complete disappearance of arrested TCs (construct 10), whereas sequence alterations between positions ϩ7 to ϩ11 (construct 11), ϩ12 to ϩ16, and ϩ17 to ϩ21 (constructs 12 and 13) had only moderate or negligible effects. Thus, the second part of the signal determining promoter-proximal transcriptional arrest at PrplN is located between positions ϩ2 and ϩ6 of the ITR.
In the presence of GreA, the formation of abortive RNA and arrested TCs (wherever arrest took place) in all of the studied promoter constructs was significantly reduced with concomitant increase in the amount of run-off product, both in singleand multiround transcription assays ( Fig. 7A and data not shown). For most studied templates where arrested TCs were formed, the stimulatory effect of GreA was in the range of 6 -13-fold (Fig. 7B). Noticeably, even on templates where arrest did not occur, GreA increased the amount of run-off 3-7-fold. The only exception was the wt T7 A1, where the effect was only 2-fold (Fig. 7B). These observations are consistent with results of Hsu et al. (9), who found that various deviations from the natural "optimized" sequence of ITR interfere with efficiency of promoter escape through multiple mechanisms, such as increased abortive cycling, formation of "moribund" initiation complexes, etc.

DISCUSSION
The principal observation of this work is that during transcription initiation in the absence of transcript cleavage factors Gre, at promoters rplN and ompX, E. coli RNAP falls into a promoter-proximal arrested state that impedes production of full-length transcripts both in vitro and in vivo. Promoter-proximal transcriptional arrest occurs when RNAP synthesizes 9 -14-nt-long transcripts and backtracks by 5-7 (PrplN) or 2-4 (PompX) nucleotides. GreA and GreB prevent arrest at PrplN and PompX by inducing RNA cleavage at the onset of arrested TC formation and stimulate productive transcription by allowing RNAP to use the 5Ј end proximal transcript cleavage products as primers for productive transcription. Both factors can cleave and reactivate arrested TCs at PompX, but only GreB can reactivate arrested TCs at PrplN. The difference between the action of Gre factors is apparently due to the fact that GreA can cleave RNA only by small 2-4-nt-long fragments, whereas GreB is able to cleave both in small and in large, 4 -18-nt-long increments (1)(2)(3)13).
The two arrested TCs described here are clearly different from promoter-proximal paused TCs that have been previously observed at several bacterial and bacteriophage pro- moters (29,32,(43)(44)(45)(46). The principal difference between promoter-proximal arrested and paused TCs is that the latter exist only transiently and eventually clear the pause site. In contrast, arrested TCs formed on PrplN and PompX are permanently inactive in the absence of Gre factors. In this, they are similar to two previously characterized arrested TCs that were obtained in vitro: one forming during unusual primer shift initiation at the ribosomal rrnBP1 promoter (3,47) and another forming at phage T7 A1 promoter after prolonged incubation of an elongation complex containing 27-nt-long RNA in the absence of a substrate specified by the template position 28 (12,13,39). Unlike these artificially arrested TCs, promoter-proximal arrest at PrplN and PompX takes place both in vitro and in vivo during "natural" initiation in the presence of all NTPs at high concentrations.
Promoter-proximal transcriptional arrest may occur through different mechanisms, which need not be mutually exclusive. First, the arrest may be caused by excessive DNA scrunching during promoter escape. The relief from scrunching, in the form of RNAP backtracking, may reposition short transcripts into an RNAP site that allows tight interaction with the enzyme and therefore prevents transcript dissociation. Second, the RNAP subunit can contribute to promoter-proximal arrest either by directly interacting with short transcripts in the backtracked conformation of the complex or through the hopping mechanism, although the latter must take place in the absence of appropriately located sequences with similarities to the Ϫ10 promoter element consensus in the initial transcribed regions of promoters studied here.
Be that as it may, we propose that formation of promoterproximal inactive/stalled TCs is a common feature of pro-moters whose efficient transcription requires transcript cleavage factors. The molecular mechanisms of promoterproximal stalling are as yet unknown. The signal for transcriptional arrest may be a complex function of the strength of RNAP-DNA/RNA interactions with different promoter elements and the structure of DNA (such as methylation, supercoiling, etc.). The contribution of these factors may be different at different promoters. Clearly, promoter-proximal arrest is a previously unrecognized yet likely common feature of many bacterial promoters and is a physiological target of regulation by transcript cleavage factors.
In fact, our results suggest that suppression of promoterproximal arrest could be one of the main in vivo functions of GreA and GreB. The two other known anti-arrest mechanisms in the cell, including cooperative action of multiple RNAP molecules co-directionally transcribing through an arresting DNA site (48) and ATP-dependent dissociation of arrested TCs by transcription-coupled repair factor Mfd (42), may not work during promoter-proximal arrest. First, in the case of PrplN and PompX, a significant portion of promoter DNA is still occupied by arrested RNAP (Fig. 3), preventing new transcription initiation and therefore rescue of arrested TC by the second RNAP molecule. In support of this, we found that arrested TCs are resistant to reactivation by excess of RNAP holoenzyme (data not shown). Second, the presence of -factor in promoter-proximal arrested TCs excludes Mfd from binding to and acting on arrested RNAP (42). Therefore, transcript cleavage reaction activated by GreA and GreB appears to be the only mechanism available in the cell to prevent and counteract promoter-proximal transcrip- tional arrest, which may explain the high degree of conservation of transcript cleavage factors in evolution.
Very recently, an avalanche of data indicating the previously unrecognized commonality of transcription regulation via "transcription-poised" RNAP arrested early at the transcription unit in eukaryotes became available (reviewed in Ref. 34). Such a mechanism allows very rapid responses to environmental changes by "restarting" stalled transcription complexes and avoiding the need to assemble transcription initiation complexes de novo. Bacterial transcription units on which promoter-proximal arrested complexes responsive to Gre factors form are formally analogous to eukaryotic examples of regulation by transcription poising. Environmental conditions that lead to changes in gene expression by affecting the efficiency of promoter-proximal arrest are currently being investigated in our laboratories.