Antagonistic Regulation of Escherichia coli Ribosomal RNA rrnB P1 Promoter Activity by GreA and DksA*

The Escherichia coli proteins DksA, GreA, and GreB are all structural homologs that bind the secondary channel of RNA polymerase (RNAP) but are thought to act at different levels of transcription. DksA, with its co-factor ppGpp, inhibits rrnB P1 transcription initiation, whereas GreA and GreB activate RNAP to cleave back-tracked RNA during elongational pausing. Here, in vivo and in vitro evidence reveals antagonistic regulation of rrnB P1 transcription initiation by Gre factors (particularly GreA) and DksA; GreA activates and DksA inhibits. DksA inhibition is epistatic to GreA activation. Both modes of regulation are ppGpp-independent in vivo but DksA inhibition requires ppGpp in vitro. Kinetic experiments and studies of rrnB P1-RNA polymerase complexes suggest that GreA mediates conformational changes at an initiation step in the absence of NTP substrates, even before DksA acts. GreA effects on rrnB P1 open complex conformation reveal a new feature of GreA distinct from its general function in elongation. Our findings support the idea that a balance of the interactions between the three secondary channel-binding proteins and RNAP can provide a new mode for regulating transcription.

Structural and functional studies of bacterial and eukaryotic RNA polymerases (RNAP) 4 reveal an emerging class of transcriptional regulators: proteins that interact directly with the secondary channel, GreA, GreB, DksA, TFIIS, Microcin J25, and Gfh1 (for review, see Refs. [1][2][3][4]. The secondary channel itself is thought to provide NTP access to a stably bound Mg 2ϩ ion in the catalytic center of RNAP when template DNA and the nascent RNA-DNA hybrid occlude the main channel (5,6). Multiple transcription factors acting at a common site on RNAP raise possibilities of new modes of transcription regulation (for review, see Ref. 7).
GreA and GreB are bacterial protein homologs with shared N-terminal ␣-helical coiled-coil finger structures that penetrate the secondary channel and juxtapose two acidic residues near the catalytic center; the C-terminal globular domains are implicated in RNAP binding (8,9). Yeast RNA polymerase II transcription factor TFIIS, although structur-ally distinct from GreA and GreB (10), also positions a pair of conserved acidic amino acid residues near the catalytic center. In all cases, these acidic residues activate an intrinsic RNA phosphodiesterase activity of RNAP paused during elongation, leading to cleavage of nascent RNA, whose 3Ј end has threaded backwards relative to the catalytic center; this cleavage creates a 3Ј-hydroxyl near the catalytic center and restores the possibility of polymerization (11)(12)(13). Despite their structural similarities, GreA and GreB are functionally different. The length of RNA cleavage products differs and GreA, but not GreB, needs to bind before arrest to activate RNA cleavage (14,15). Functions proposed for Gre factors include: facilitating promoter escape by suppressing abortive transcription; suppressing elongational pausing or arrest; and enhancing proofreading (Ref. 16, and for review, see Ref. 1). Although blocking the passage of NTP substrates by binding of Gre factors has not been shown, channel binding by the structurally unique Microcin J25 antibacterial peptide is thought to inhibit transcription elongation directly by obstructing NTP entry (2,17).
Transcription initiation can also be regulated through the secondary channel by DksA and (p)ppGpp, 3Ј-pyrophosphorylated derivatives of GDP (GTP). First, (p)ppGpp, a nearly ubiquitous nucleotide regulator of transcription in eubacteria and in plants, can be localized by co-crystallization with RNAP near the catalytic center (18). Second, DksA, despite a divergent amino acid sequence, is a structural homolog of GreA and GreB. This allows similar modeling in the secondary channel and a proposal of nearly appropriate placement of the key pair of conserved acidic residues (19). Third, DksA and ppGpp can function both in vitro and in vivo as synergistic co-factors for negative as well as positive regulation of transcription initiation (19 -22). Although it requires reorientation of the pair of acidic residues, it has been proposed that DksA might anchor (p)ppGpp in one of two possible orientations near the catalytic center through Mg 2ϩ ions jointly coordinated by ppGpp pyrophosphate residues, thereby stabilizing the bound co-factor and potentiating its regulatory effects (19). Whether the regulatory effect of ppGpp and DksA is positive or negative seems largely specified by the promoter discriminator sequence between the Ϫ10 and ϩ1 region, as predicted long ago (23). DksA and ppGpp exert strong negative regulation of rrnB P1 and other ribosomal RNA promoters. This regulation is argued to result from enhanced destabilization and closure of RNAP-rrnB P1 promoter open complexes together with weakened dependence on the initiating NTP (for review, see Ref. 22). Positive regulation of some promoters for amino acid biosynthetic genes is attributed mainly to increasing the rate of isomerization between closed and open complexes (21). Here, we present experiments designed to test the hypothesis that the three secondary channel binding proteins (GreA/B, DksA) can mutually compete for RNA polymerase despite known differences in the step of transcription at which they function.
Plasmid pBW, containing the rrnB P1P2 region spanning from Ϫ235 to ϩ237 bp (using P1 as a reference) was constructed by removing a 370-bp rrnB T1T2 terminator fragment from plasmid pPS1 (26) and ligating it to an EcoRI and HindIII fragment containing the rrnB P1P2 region from Ϫ235 bp to ϩ237 bp (obtained by PCR using chromosomal DNA templates). The rrnB P1 promoter template (Ϫ180 to ϩ109 bp) was generated by PCR using pBW plasmid as template and the following primers: rrnBϪ180 (5Ј-TGCCTTTTGTATGGCAATGAC-3Ј) and rrnBϩ109 (5Ј-AATACGCCTTCCCGCTACA-3Ј). The ensuing 289-bp fragment was purified using the High Pure PCR Product Purification Kit (Roche Diagnostics) before use as template. We elected to use a linear template with a single promoter for all assays because inhibitory functions of DksA on rrnB P1 have been reported to be independent of superhelicity of the DNA template (21).
Plasmid pHM1506 was derived from pGB2 with a PSC101 origin conferring spectinomycin resistance (27) and was modified to contain lacI q and used for expressing native dksA from the Ptac promoter. For purification of C-terminal His-tagged DksA, pHM1501 was constructed by PCR cloning of the dksA open reading frame from pJK537 (28) in pET21 (Novagen). The inability of the ⌬dksA strain CF9239 to grow on glucose minimal medium is complemented by pHM1506. Plasmids used for in vivo experiments as well as purification of native GreA and native GreB were pDNL278 and pGF296, respectively (29).
In Vitro Transcription-Transcription assays were performed in a reaction volume of 20 l at 30°C, using 10 nM template and 30 nM RNAP (E. coli holoenzyme from Epicenter Technologies) in buffer containing 50 mM Tris acetate, pH 8.0, 10 mM MgAc, 10 mM ␤-mercapthoethanol, 10 g/ml bovine serum albumin, 90 mM potassium glutamate, 100 M ATP, GTP, and CTP, and 10 M UTP (10 Ci/reaction [␣-32 P]UTP, Amersham Biosciences), and either 250 M GDP or 250 M ppGpp. RNAP was preincubated (25°C) with GDP or ppGpp for 7 min prior to the addition of potassium glutamate, and unless stated otherwise, this was followed by a 7-min incubation at 30°C with DNA and the indicated GreA and/or DksA concentrations (0 -600 nM). The reactions were initiated by adding NTP substrates and terminated after 10 min by the addition of an equal volume of stop solution (95% formamide, 20 mM EDTA, 0.05% bromphenol blue, and 0.05% xylene cyanol). In experiments varying the order of addition of GreA and DksA, the protein added first was preincubated with RNAP and DNA, and the second protein was added together with NTPs. In single round experiments, 100 g/ml heparin was added. Samples were analyzed on 7 M urea, 6% polyacrylamide sequencing gels and quantified by phosphorimaging on a GE Healthcare imaging system.

Promoter Escape and Open Complex Stability Assays-Promoter
escape and open complex stability assays were as described in Ref. 30, except that the buffer conditions, RNAP and DNA concentrations were as specified above.
DNase I Footprinting Assay-End-labeled DNA fragments were generated by PCR using 32 P-end-labeled rrnBϪ180 and unlabeled rrnBϩ109 primers, and purified using the High Pure PCR Product Purification Kit (Roche Diagnostics). RNAP (0 -100 nM) was preincubated with 500 M GDP, or ppGpp, at room temperature for 7 min in transcription buffer, then labeled DNA template (3 nM final) was added together with potassium glutamate (90 mM final), 600 nM GreA, and/or 600 nM DksA. The final reaction volume was 20 l. Reaction mixtures were incubated at 37°C for 15 min, treated with DNase I (7.5 ng/reaction; Promega Corp.) for 6 min, terminated by addition of EDTA (25 mM final), concentrated by vacuum evaporation, and resuspended in 20 l of loading buffer (80% formamide, 6 M urea, 10 mM NaOH, 0.05% bromphenol blue, and 0.05% xylene cyanol). The samples were resolved on 7 M urea, 8% polyacrylamide sequencing gels, run in parallel with sequencing reactions obtained by using unlabeled DNA fragment as a template with 32 P-end labeled rrnBϪ180 primer.
KMnO 4 Footprinting Assay-KMnO 4 footprinting of the open complexes was performed as described in Ref. 30, except that the open complexes were allowed to form for 15-30 min before incubating under reaction conditions as described above for the DNase I footprinting. When specified, ATP and CTP were added at 100 M each.
Electrophoretic Mobility Shift Assays-These studies were performed as described in Ref. 31. When employed, ppGpp was added at 500 M, GreA concentration was 600 nM. The template concentration was 3 nM, and the RNAP concentration was varied.

DksA and GreA/B Effects in Vivo-
The structural study of GreA, GreB, and DksA suggests that they might compete for their interactions with RNAP. We tested this hypothesis in vivo, using an rrnB P1::lacZ ribosomal promoter operon fusion (32,33) known to be negatively regulated by DksA and ppGpp (22). Plasmids allowing IPTG-inducible Ptac driven expression of native GreA (pDNL278), GreB (pGF296), or DksA (pHM1506) (Fig. 1, top, middle, and lower rows) were introduced into strains with or without ppGpp (left two columns and right two columns, respectively). Reporter activities were measured during a cycle of growth in LB broth, with or without IPTG. We also introduced mutations into each series of strains eliminating potential competing proteins (second and fourth columns). Fig. 1 shows differential plots of reporter activities versus turbidity (A600) during growth cycles. For sufficiently low values of A 600 , these curves are approximately linear; the slope of such a curve is the differential rate of synthesis; these values are given in Table 1, expressed as Miller units (24).
Looking first at the untransformed strains in exponential phase, it can be seen ( Table 1) that loss of DksA increased rrnB P1 transcription 2.4-fold over the wild type level, whether ppGpp was present or not. This confirms that DksA is a negative regulator of rrnB P1 transcription. Loss of GreA, in contrast, had no effect. Fig. 1 also shows that the promoter activity of the wild type strain shut down when a critical cell density was reached (A 600 ϳ 2.0). No arrest was observed in the absence of DksA and arrest occurred later in the absence of ppGpp. Despite uncertainties regarding an absolute requirement for ppGpp during entry into stationary phase, this is qualitatively consistent with a requirement for both ppGpp and DksA to restrict P1 transcription late in growth in LB broth (20, 28, 34).
The same arrest was observed at high cell densities when the strains were transformed with the three plasmids, so long as the plasmids were uninduced. Major differences were observed, however, when the proteins were induced with IPTG. Some induced conditions were unexpectedly inhibitory for growth, such as inducing GreA (but not GreB) in a dksA mutant and inducing DksA in all strains. We therefore reduced the inducing concentration of IPTG from the standard 1 to 0.01 mM ( Fig. 1, panels B and D) or 0.1 mM (panels I-L). These adjustments of IPTG also yield roughly similar levels of each protein that approximate the staining intensities of the most abundant E. coli proteins in crude extracts (supplemental Fig. S1).
In wild type cells, increasing GreA levels activated rrnB P1 transcription (Fig. 1A, Table 1). Simply transforming the wild type strain with pgreA mildly elevated the specific activity 1.4 times (ϩ/ϪppGpp) and IPTG induction further activated for a total of 3.4 times. A dksA mutation in the otherwise wild type strain increased its specific activity 2.4 times and transforming with pgreA further activated fusion 1.4 times (in the absence of ppGpp), again for a total of 3.4 times. In this case, however, no further activation was observed after IPTG addition, but rather a slight decrease was observed (from 250 to 211 units). This did not result from a difference in GreA expression in the dksA mutant: protein staining indicated that the mutant did not induce GreA without IPTG but induced it normally when IPTG was added (supplemental Fig. S1). That equal levels of rrnB P1 activation occur at such different levels of GreA may mean that activation has reached a saturation limit. Indeed, this specific activity limit of 251 units was exceeded only once, 294 units, in an uninduced dksA mutant in the absence of ppGpp (Table 1). Nevertheless, the activation of rrnB P1 by GreA (at its uninduced level) is potentiated by the absence of DksA.
Comparing the middle row of panels in Fig. 1 to the top row reveals that the pgreB plasmid had little effect in the absence of inducer. Even with IPTG, activation of rrnB P1 promoter activity was not observed in the wild type strain (Fig. 1, A and E) and was barely noticeable in the dksA ϩ ppGpp 0 strain (Fig. 1G). However, inducing the pgreB plasmid did alleviate the rrnB P1 arrest late in the growth cycle (Fig. 1E). Panels F and H show that the dksA mutant again failed to arrest, as in panels B and D.
Shortly after addition of IPTG to growing cells containing pdksA, reporter expression abruptly stopped in all strains ( Fig. 1, I-L); this arrest in rrnB P1 activity did not require ppGpp or GreA. Note that the ordinate scale in the last row of panels (I-L) is amplified 4 times relative to the upper row panels to contrast inhibition of rrnB P1 transcription with low activity levels. The block of rrnB P1 transcription when DksA was overexpressed was both more rapid and more complete than the growth limitation observed under these conditions. FIGURE 1. Overproduction of GreA, GreB, or DksA alters rrnB P1::lacZ activity. Using ⌬lacZ strains carrying the rrnB P1::lacZ transcriptional fusion, we evaluated the effects of overproduction of GreA (panels A-D, strains transformed with pDNL278 ϭ pgreA), GreB (panels E-H, strains transformed with pGF296 ϭ pgreB), or DksA (panels I-L, strains transformed with pHM1506 ϭ pdksA) on rrnB P1::lacZ activity. All strains were grown at 37°C; IPTG was added when the culture reached an A 600 of 0.10 -0.15; ␤-galactosidase activity was monitored periodically. Squares, untransformed strains cultivated in LB. Filled diamonds, transformed strains cultivated in LB supplemented with ampicillin (panels A-H) or with spectinomycin (panels I-L). Circles, transformed strains cultivated in the same media supplemented with IPTG at 10 Ϫ3 M (panels A, C, and E-H), 10 Ϫ4 M (panels I-L), or 10 Ϫ5 M (panels B and D). Although lines showing arrest of activity for A 600 Ն 2.0 seem defined by a single point in panels A, E, I, and J, the lines as drawn actually depend on more extensive data not shown. The ⌬relA spoT ϩ strains are termed wild type here; equivalent results were obtained with relA ϩ spoT ϩ strains (data not shown).

TABLE 1 Data tabulated are differential rates of synthesis of ␤-galactosidase expressed in Miller units
These specific activities were calculated for the growth curve regions where absolute activities increase linearly with turbidity ( Fig. 1). In parentheses are the relative rates, normalized to the wild type rate. These results are compatible with the hypothesis that DksA and GreA compete with each other, the latter activating rrnB P1 transcription, the former inhibiting it. They can be viewed as non-physiological responses because of the abnormally high level of IPTG-induced proteins. If physiological function requires a balanced ratio of proteins then simultaneous induction of putative protein competitors should eliminate changes in promoter regulation. To test this, we constructed strains with two compatible plasmids allowing IPTG induction of both DksA (pHM1506) and GreA (pDNL278). Co-induction of DksA and GreA activated rrnB P1 much the same as inducing GreA alone, overriding inhibitory effects of DksA alone (supplemental Fig. S2). Co-induction of GreB and DksA protected cells from inhibitory effects of DksA, but did not activate like GreA (supplemental Fig. S2). Evidently, balanced levels of these proteins have physiological effects regardless of their absolute concentration.
These in vivo experiments confirm the negative effect of DksA on rrnB P1 activity and suggest that excess GreA, and to a lesser extent excess GreB, can counteract it. We now ask whether these observations can be reproduced and understood in a purified transcription system.
GreA Activation, Not DksA Inhibition, Requires Its Presence during Open Complex Formation-We first asked whether GreA stimulation and DksA inhibition of rrnB P1 transcription occurs during multiround transcription using purified GreA and DksA proteins (see "Experimental Procedures"). All transcription experiments employed GDP as a control for the effects of ppGpp. Initially, RNAP (30 nM) was preincubated with 250 M ppGpp (or 250 M GDP) together with increasing amounts of GreA, and the reaction was initiated by addition of 10 nM DNA template and NTP substrates. We reasoned that because GreA is implicated in binding to the RNAP secondary channel, preincubation with RNAP might increase its ability to interact with RNAP and activate. However, preincubating with even a 10-fold molar excess of GreA over RNAP did not stimulate transcription ( Fig. 2A). Preincubating RNAP with only ppGpp (or GDP) and initiating the reaction with GreA, template, and substrates also failed to reveal GreA activation (Fig. 2B). A modest activation at higher amounts of GreA was noted when the reaction was initiated with GreA and NTPs (Fig. 2C). It became clear that to achieve maximum activation (2 to 2.5 times), GreA had to be present during open complex formation, i.e. in the presence of template, starting the reaction with NTPs (Fig. 2D). As observed in vivo, activation did not require ppGpp but was seen using a 5-10-fold molar excess of GreA over RNAP (Fig. 2D). Fig. 2 also shows the results of similar assays with DksA that affirm its ability to inhibit rrnB P1 transcription, again at a 5-10-fold molar excess over RNAP. Unlike in vivo inhibition by DksA, in vitro inhibition   Fig. 2C; reactions for panels C-F were single round under the conditions of Fig. 2D: C and D, GreA was added to the RNAP/DNA mixture and the reaction was initiated by addition of NTPs with DksA; E and F, DksA was added first and the reaction was initiated by the addition of NTPs with GreA. In panels A, C, and E, assays were carried out in the presence of ppGpp; in panels B, D, and F, GDP was substituted.
requires ppGpp (Fig. 2, A-D). Clear ppGpp-dependent inhibition was seen under all conditions tested (Fig. 2, A-D), with a somewhat stronger effect when the reactions were started with NTPs. Evidently GreA activates and DksA inhibits rrnB P1 transcription in vitro. Experiments similar to those shown in Fig. 2 were performed with GreB and slight activation was observed with or without ppGpp (data not shown), consistent with data in Fig. 1.
Simultaneous Versus Serial Addition of GreA and DksA-We next performed reactions similar to those of Fig. 2C, but asked whether GreA and DksA compete for RNAP at the rrnB P1 promoter. First, GreA and  1-3, neither GreA, DksA, GDP, nor ppGpp were present; lanes labeled 600 nM D, only DksA was present; lanes labeled 600 nM A, only GreA was present; A 3 D, samples were preincubated with GreA and then DksA was added; A ϩ D, both proteins were added simultaneously. Lanes labeled G, A, C, or T are sequencing ladders. Numbers at the right represent nucleotide positions relative to the start site. Panel A footprints were performed in the absence of NTP substrates, whereas Panel B footprints were in the presence of ATP and CTP. The panel A autoradiograph is lightly exposed to better visualize the differences between the footprints. DksA at concentrations varying from 30 to 600 nM were presented simultaneously to a fixed concentration of 30 nM RNAP and 10 nM DNA with ppGpp or GDP, with reactions initiated by substrate addition (Fig. 3, A and B). The first set of reactions showed that DksA with ppGpp could overcome stimulation by GreA even when present in significantly lower concentrations than GreA (Fig. 3A). In the absence of ppGpp, however, DksA was no longer able to overcome GreA activation even at a 20-fold molar excess of DksA over RNAP (Fig. 3B). This is consistent with a strong ppGpp co-factor requirement in vitro for DksA inhibition but not with the ppGpp-independent inhibition observed in vivo when DksA was overproduced (Fig. 1, J and L). Similar results were obtained when these experiments were repeated but allowing only a single round of transcription (data not shown).
We then asked whether allowing the first protein to be present during open complex formation could be reversed by a brief exposure to the second protein. This was accomplished under conditions similar to Fig.  2D, but utilizing single round assays and initiating the reaction by the addition of the second protein together with NTP substrates (Fig. 3, C-F). When GreA was added first in the presence of ppGpp, the activation by even a 20-fold molar excess of GreA over RNAP was progressively abolished by initiating the reaction with increasing amounts of DksA, NTP substrates, and heparin (Fig. 3C). Again, this reversal required the presence of ppGpp (Fig. 3D). In contrast, even a 20-fold molar excess of GreA over RNAP did not reverse DksA inhibition in the presence of ppGpp (Fig. 3E). This was not simply because the stimulatory effect of GreA was abolished when it was added after open complexes had formed since control experiments with GDP show that GreA was able to activate rrnB P1 under similar conditions (Fig. 3F), although to a lesser extent than in Fig. 3, B and D. Taken together, these experiments support the conclusion that, in the presence of ppGpp, transcription inhibition by DksA is dominant over in vitro activation of the rrnB P1 promoter by GreA.
GreA and DksA Effects on RNAP-DNA Complex Conformation-We next wished to document that the GreA-and/or DksA-sensitive steps in rrnB P1 transcription do indeed both occur at a stage of initiation, as implied by Fig. 3, by asking whether there are conformational changes in open complexes formed in the presence of these transcription regulators. Open binary complexes of rrnB P1 promoters and RNAP previously have been reported as too unstable for footprinting unless ternary complexes are created by a limited number of phosphodiester bonds in the presence of ATP and CTP (35). Nevertheless, we were able to obtain DNase I footprints with RNA polymerase on double-stranded DNA labeled on the 5Ј end of the nontemplate strand. Footprints in the absence of initiating substrates are shown in Fig. 4A using varying amounts of RNAP, 500 M ppGpp or GDP, with or without 600 nM GreA and DksA added alone, simultaneously, or serially. We found protection with RNAP alone, with ppGpp or with GDP extending from Ϫ57 to about ϩ30 (Fig. 4A, lanes 1-3). This protection pattern was not appreciably altered by adding DksA, whether alone, with GDP, or with ppGpp (Figs. 4A, lanes 4 -6, 7-9, and 19 -21; and 5, A and B). Adding GreA, especially in the presence of ppGpp (as compared with GDP) reduced DNase I protection in the region from Ϫ35 to ϩ30 (Figs. 4A,  lanes 10 -12 and 22-24; and 5, A and B). Adding DksA after preincubating GreA with RNAP and ppGpp largely eliminated the reduced protection by GreA, thereby restoring the DNase I footprint to one characteristic of RNAP alone (Figs. 4A, lanes 2-3, 13-15, and 25-27;  and 5, A and B). Adding DksA simultaneously with GreA (with ppGpp) seemed slightly less effective at reversing the GreA footprint change than adding DksA last (Figs. 4A, lanes 25-27 and 28 -30; and 5B).
These experiments were repeated with ATP and CTP present to allow formation of initial RNA diesters and stabilization of RNAP-DNA open complexes (35) (Figs. 4B and 5, C and D). Protection patterns were similar except that GreA reversed protection now only in the Ϫ35 to Ϫ10 region (whereas protection persisted from Ϫ10 to ϩ30) (Figs. 4, B,  lanes 10 -12 and 22-24; and 5, C and D). Another difference is that the ppGpp specificity of GreA effects shown in Fig. 4A disappears in Fig. 4B; reversal of protection occurred in the presence of GDP as well as with ppGpp. DksA alone had no effect on the protection pattern of RNA  18). B, corresponding colors have been applied to lanes with ppGpp from Fig.  4A (lanes 21, 24, 27, and 30). C, a similar presentation applies to analogous GDP control lanes from Fig. 4B (lanes 8, 11, 14, and 17) as well as for D, footprint lanes with ppGpp from Fig. 4B (lanes 20, 23, 26, and 29). polymerase (ϩ/ϪppGpp), but DksA in the presence of ppGpp restored the protection diminished by GreA (compare pairs of Figs. 4B, lanes 11-12, 14 -15, and 17-18 with 23-24, 26 -27, and 29 -30; and 5, C and D). This conclusion is in agreement with results presented in Fig. 3 where DksA was unable to overcome effects of GreA in the absence of ppGpp. The fact that GreA induces a change in the footprint in the presence of ppGpp but not GDP (Fig. 4A, compare lanes 11-12 and  23-24) may indicate an additional effect of GreA to overcome the negative effect of ppGpp and to activate transcription. To confirm the promoter specificity of the footprints, similar experiments were performed with a template bearing scrambled rrnB P1 Ϫ10 and Ϫ35 regions; even in the presence of ATP and CTP no DNase I protection was obtained (data not shown). Note that GreA and DksA yielded no observable footprint in the absence of RNAP (Fig. 4, A and B).
We also performed KMnO 4 footprinting to ask whether singlestranded DNA was generated by conditions similar to those presented in Fig. 4A. These footprints were found to be weak without ATP and CTP but unchanged by GreA and DksA. Parallel studies of stable complexes in the presence of ATP and CTP, similar to Fig. 4B, also did not reveal GreA-or DksA-dependent differences (data not shown). This implies that the DNase I footprints in Fig. 4A reflect the formation of open complex intermediates and that the single-stranded DNA bubbles are unchanged in the stable open complexes visualized in Fig. 4B. The effects of GreA on the extent of open complex formation seemed minimal from electrophoretic mobility shift assays without NTP substrates (Fig. 6A). In the presence of ATP and CTP a more striking effect of GreA on the formation of complex III was found (Fig. 6B), suggesting a more lasting conformational effect (as visualized in Figs. 4B and 5, C and D). Complex III is judged to be the only active complex, as seen from its unique disappearance in the presence of all four NTP substrates (data not shown).
We were surprised that DksA did not alter the open complexes in Fig.  4, A and B, although it is a strong inhibitor of rrnB P1 transcription. To visualize the effect of DksA better, we lowered the levels of RNAP to the point where a stable footprint was not well defined and then asked whether DksA or GreA altered the footprint. With RNAP present at 25 nM, adding 600 nM DksA (but not GreA) gave an increased level of protection in the presence of either GDP or ppGpp (Fig. 7). The protection afforded by DksA with 25 nM RNAP seems to be nearly the equivalent of 50 nM RNAP without DksA. This shows that stable open complexes are facilitated by DksA even at more dilute concentrations of RNAP (25 nM) than those in Fig. 4B (50 -100 nM).

DISCUSSION
Cellular Experiments Do Not Support a Simple Competition Hypothesis-A simple prediction of the hypothesis tested here was that secondary channel interactions between GreA/B, DksA, and RNAP dictate functional competition in transcription. The prediction is that changing the abundance of one protein should alter the regulatory effects of each of the others.
Measurements of effects of mutation and/or overproduction of the three proteins on transcription of the rrnB P1 in vivo are not strictly consistent with this hypothesis. First, GreB overexpression during exponential growth does not alter activity, even if DksA is abolished by mutation. In contrast, overexpression of GreA activates this promoter, depending on the abundance of DksA. Second, overexpression of DksA inhibits activity of the same promoter independent of GreA levels. Nevertheless, there exists an antagonistic regulation between GreA and DksA. A high induced level of GreA in a dksA mutant, but not in dksA ϩ , inhibits growth. The wild type level of DksA protects the cell against the potentially inhibitory effect of a high level of GreA. We believe growth inhibition by overproduction of a protein is common, but conditional inhibition depending on both overexpression of one protein and the absence of another protein is not. The cause of growth inhibition by excess GreA in the absence of DksA remains obscure.
It is puzzling that DksA induction inhibits growth even in the absence of ppGpp (Fig. 1K), because it is a co-factor of DksA for the negative regulation of rRNA synthesis (see Introduction). We also find ppGpp to be required for DksA-mediated transcription inhibition in vitro (Fig. 2). The explanation might relate to observations that very high (200-fold) molar ratios of DksA to RNAP inhibit rrnB P1 transcription in vitro even when ppGpp is absent (20). Nevertheless, the effects noted above on cell growth seem to reflect a need for a balance between the cellular content of GreA and DksA. This implies that an altered hypothesis might be that mutual competition between the three proteins is specified by unique binary combinations. Support for this possibility comes from two observations. First is that simultaneous overexpression of GreB and DksA reverses the growth toxicity of DksA without altering rrnB P1 activity (supplemental Fig. S2). Second, overexpression of both GreA and DksA reverses inhibition of rrnB P1 activity (supplemental Fig. S2).
The balance between these three proteins can also affect the arrest of rrnB P1 activity that normally occurs at A 600 ϳ2 (Fig. 1A). This arrest is bypassed in the absence of DksA (Fig. 1B) or when high levels of GreA are induced in a wild type strain (Fig. 1A). It seems that the normal level of DksA prevents the normal level of GreA from activating rrnB P1 in stationary phase, just as it prevents GreA-mediated growth inhibition during exponential growth. GreB, when overexpressed, shares with GreA the ability to bypass the arrest of rrnB P1 transcription (Fig. 1E). Arrest of rrnB P1 transcription in wild type cells probably reflects entry into early stationary phase, because it requires both DksA and ppGpp (see "Results"). Arrest of rrnB P1 activity is promoter specific because no arrest was observed with a lacUV5 promoter operon fusion in greA, greB, or dksA strains (data not shown). We believe this is the first evidence that the balance of the three secondary channel-binding proteins has functional significance and raises new possibilities for transcription regulation.
Similarities and Differences in Vivo and in Vitro, for GreA and DksA Effects-A direct activating effect of GreA on transcription has been found. Maximal activation requires its presence during open complex formation. Multiple round transcription reactions reproduced the activation of rrnB P1 transcription by GreA to nearly the same extent (2.5 times) as in vivo (3.4 times); the ppGpp independence of this effect was also verified (Fig. 2).
Several parameters associated with initiation were not altered by GreA. The presence of GreA (or DksA) had no effect on the low level of abortive transcripts under all conditions of Fig. 2, A-D (data not shown). Also, measuring promoter escape by back extrapolating closely timed kinetic samples failed to reveal significant differences in the presence of GreA (data not shown). Finally, open complex stabilities, determined with a heparin challenge, were unaffected by the presence of GreA (data not shown).
DksA added in parallel in vitro experiments allowed us to verify published observations of a co-factor requirement for ppGpp to inhibit rrnB P1 transcription (20). Mixtures of GreA and DksA presented simultaneously to RNAP at varying ratios to each other and to RNAP revealed that even brief exposures to DksA could inhibit and prevent or reverse GreA activation (Fig. 3). Like DksA-mediated inhibition, this reversal of GreA activation was ppGpp-dependent.
DksA and GreA Alter RNAP-DNA Complex Conformation-DNase I footprints of RNAP with GreA and/or DksA and the rrnB P1 promoter DNA revealed the existence of a specific rrnB P1 promoter binary open  JUNE 2, 2006 • VOLUME 281 • NUMBER 22 complex without the need for nucleotide substrates or accessory proteins (Fig. 4A). This footprint persisted unchanged in the presence of high DksA levels, with or without ppGpp, as if DksA did not destabilize the open complex. When GreA was substituted for DksA, the protection pattern was diminished, especially in the presence of ppGpp. The GreA-mediated change in the DNase I protection pattern was partially or completely reversed by DksA addition in the presence of ppGpp but not GDP. A change in the intensity of a footprint, without altering the extent of the protected region, could result from a change in RNAP promoter affinity. However, this possibility was ruled out by electrophoretic mobility shift assay, where we did not observe decreased binding of RNAP to the rrnB P1 region (Fig. 6). Instead we found that GreA even enhanced the formation of active RNAP-DNA complexes in the presence of ATP and CTP (Fig. 6B). We therefore interpret the GreAinduced change in the DNase I footprint (Figs. 4A and 5, A and B) to represent a conformational change in the RNAP-promoter binary complex.

Antagonistic Regulation by GreA and DksA
Furthermore, the footprints imply that GreA acts before DksA during initiation, in the absence of NTPs, which is consistent with the order of addition experiments (Fig. 3, C-F). Analogous footprints obtained in the presence of ATP and CTP were similar except that the footprint pattern with GreA was reversed at slightly different sequence positions by DksA. With ATP and CTP, GreA reversal now occurred in the presence of ppGpp and to a lesser extent in the presence of GDP (Figs. 4B and 5, C and D). Here, we are more confident of a conformational change because the extent of the footprint was changed by GreA, not just its intensity. This footprint covered only about 30 bp of DNA, which corresponds to expectations for an elongating complex (36). This difference is interesting, because without GreA present one would expect the first phosphodiester bond to be formed, and yet the footprint has an Alternatively, DksA inhibition may prevail despite a 20-fold molar excess of GreA (Fig. 3) because of a higher affinity of DksA for RNA polymerase than GreA as appears to be the case from Fig. 7.
Apparently, GreA and DksA can exist in distinct complexes with RNAP and the rrnB P1 promoter. As stated above, it seems that GreA can modify the conformation of RNAP complexes with DNA. It is important to note that the GreA effects we see on open complex conformation in the absence of NTP substrates are very different from previous reports that in the presence of NTP substrates GreA either stimulates the initiation process after phosphodiester bond formation has begun (16,37) or prevents the formation of inactive moribund complexes at the pR promoter (38).
DksA does not seem to change the conformation of the open complex, but when present with GreA, it restores a conformation that mimics what is seen when GreA is absent. It is interesting in this respect that DksA alone appears to decrease the amount of RNAP necessary to obtain a DNase I footprint corresponding to an open complex (Fig. 7). This could mean that DksA facilitates DNA binding by RNAP, which could in turn occur by increasing the DNA binding constant (K B ), or by increasing the transition rates between closed and open complex or by stabilizing the open complex. Two different approaches for estimating the conformational change (electrophoretic mobility shift assay and footprints), indicate that DksA does not cause the collapse of open complexes under our conditions. These observations conflict with existing views that DksA and ppGpp are thought to increase the instability of rrnB P1 open complexes and simultaneously weaken the binding constant for initiating nucleotides (20). An analogous finding to ours was reported by Maitra et al. (31), for a rpoC⌬-(312-314) mutant that suppresses ppGpp deficiency. Interpretations by Paul et al. (20) as to the instability of open complexes based upon the heparin chase experiments were challenged by the discovery of stable, but inactive promoter-RNAP binary complexes formed by this rpoC mutant.
Considerations of GreA and DksA Actions-The conclusion that GreA participates in initiation goes against existing information on its mode of action (Fig. 8). However, there is evidence that GreA, unlike GreB, must be added before pausing actually occurs to activate cleavage of backtracked RNA (14). Perhaps there is a GreA binding site that is inaccessible in paused RNAP but accessible during initiation. The new finding here is that GreA can be bound during open complex formation (Fig. 8). It could even be imagined that GreA joins RNAP during initiation and is carried along during elongation to later perform its well established functions. This does not necessarily exclude the binding of GreA during elongation. Speculations about assembly of a GreA-containing transcription complex during initiation are reminiscent of eukaryotic transcription. Although the eukaryotic GreA homolog, TFIIS, is thought to bind RNAP during elongation, there is a report that it may play a role during initiation with Gal4-dependent promoters (39).
If GreA assembles in such a complex, then either GreA binding does not obstruct the secondary channel or NTP entry can occur by an alternative route. Only GreB has been co-crystallized with RNAP (9). Modeling of the docking of both GreA and DksA structures (1,19) revealed that the key acidic residues of DksA require rotation to achieve the same orientation as in GreA and GreB. We wonder whether this orientation might determine activation for GreA or GreB and inhibition for DksA. Similar speculations have been raised by Lamour et al. (3) and Symersky et al. (4), in their work on Gfh1, a Thermus thermophilus Gre-like homolog responsible for transcription inhibition. Compared with GreA, there are four acidic residues in the Gfh1 finger domain, which are again in a different orientation. The authors hypothesize that these residues impair transcription by chelating Mg 2ϩ in the RNAP active site. Our finding that GreA activates transcription initiation suggests that additional studies are required before proposing a more detailed mechanism for GreA.
If indeed GreA functions in initiation as argued here, it may be possible to obtain GreA mutants that differentially affect initiation but not elongation, or vice versa. GreA effects on open complex conformation before RNA diesters are formed might suggest that cleavage or the conserved acidic residues of GreA might not be activation requirements. Because GreA seems to act at the stage of transcription initiation, DksA in turn may exert unexpected additional effect on elongation. Although preliminary studies in vitro (19) seem to discourage this idea, a similar investigation in vivo may prove interesting.
Finally, it was reported recently by Trautinger et al. (40) that deleting dksA sensitizes cells to mitomycin C. Intriguingly, introducing a greA deletion in addition to dksA reversed this phenotype, consistent with our proposal that GreA and DksA counteract each others actions. One of the most important implications of our finding is that cellular regulation of transcription can be achieved by a balance of antagonistic activities of RNA polymerase secondary channel binding factors: DksA, GreA, and to a lesser degree GreB.
During the process of revising our manuscript, a publication by Susa et al. (41) appeared describing genes whose transcription was impaired by disrupting GreA and GreB. The mechanism explaining these effects is argued to be reversal of formation of abortive moribund complexes. It will be interesting to see if this process is related to the phenomenon we observed.