DksA potentiates direct activation of amino acid promoters by ppGpp

Abstract

Amino acid starvation in Escherichia coli results in a spectrum of changes in gene expression, including inhibition of rRNA and tRNA promoters and activation of certain promoters for amino acid biosynthesis and transport. The unusual nucleotide ppGpp plays an important role in both negative and positive regulation. Previously, we and others suggested that positive effects of ppGpp might be indirect, resulting from the inhibition of rRNA transcription and, thus, liberation of RNA polymerase for binding to other promoters. Recently, we showed that DksA binds to RNA polymerase and greatly enhances direct effects of ppGpp on the negative control of rRNA promoters. This conclusion prompted us to reevaluate whether ppGpp might also have a direct role in positive control. We show here that ppGpp greatly increases the rate of transcription initiation from amino acid promoters in a purified system but only when DksA is present. Activation occurs by stimulation of the rate of an isomerization step on the pathway to open complex formation. Consistent with the model that ppGpp/DksA stimulates amino acid promoters both directly and indirectly in vivo, cells lacking dksA fail to activate transcription from the hisG promoter after amino acid starvation. Our results illustrate how transcription factors can positively regulate transcription initiation without binding DNA, demonstrate that dksA directly affects promoters in addition to those for rRNA, and suggest that some of the pleiotropic effects previously associated with dksA might be ascribable to direct effects of dksA on promoters involved in a wide variety of cellular functions.

Amino acid starvation in Escherichia coli results in global changes in gene expression called the “stringent response” (reviewed in ref. 1). This regulatory response is initiated by entry of uncharged tRNAs into the A site of the ribosome and activation of the ribosome-associated RelA protein to synthesize the “alarmone” ppGpp (used here to refer to both guanosine 5′-diphosphate 3′-diphosphate and its pentaphosphate precursor). During starvation for amino acids, expression of stable RNA (rRNA and tRNA) is inhibited, whereas expression of enzymes for amino acid biosynthesis and transport is induced. These negative and positive responses are both relA dependent (e.g., refs. 2–6).

ppGpp directly and specifically inhibits the initiation of transcription from stable RNA promoters in vivo and in vitro (7–12). Although the details of the mechanism by which ppGpp exerts its effects on transcription initiation are still ill-defined, it has been shown that ppGpp decreases the lifetimes of competitor-resistant complexes between RNA polymerase (RNAP) and all promoters that have been examined (12). The complex formed by RNAP at rRNA promoters is intrinsically short-lived, and ppGpp further decreases the lifetime of this complex. Thus, we have proposed that ppGpp shifts the equilibrium from a later intermediate on the pathway to RP_O formation to an earlier intermediate in RP_O formation, reducing transcription from those promoters, such as those for rRNA, that are rate-limited at that step.

Although effects of ppGpp on transcription have been observed in purified in vitro transcription systems, they generally have been much smaller than observed in vivo. In some cases, larger effects of ppGpp have been reported in cell extracts (3, 9, 10), consistent with the larger effects observed in vivo and with the recent discovery that an additional protein, DksA, is crucial for negative control of rRNA transcription by ppGpp in vivo (11).

Unlike most transcription initiation factors, DksA does not bind DNA (11). Rather, DksA binds in the secondary (NTP) channel of RNAP (13), thereby decreasing the lifetime of competitor-resistant complexes and the concentration of ppGpp required for inhibiting rRNA transcription. It also greatly enhances inhibition of transcription from rRNA promoters by ppGpp in vitro (11). In one straightforward model, ppGpp and DksA bind to RNAP, increasing the dissociation rate of a competitor-resistant intermediate, leading to a decrease in the number of open complexes before condensation of the initial NTPs can occur (12). Although the mechanistic details responsible for the large increase in the effect of ppGpp on rRNA promoters in the presence of DksA remain to be determined, the clear synergistic behavior between ppGpp and DksA accounts for the large effects of ppGpp on rRNA transcription inhibition observed in vivo.

Deletion and overexpression of dksA have pleiotropic effects in a variety of bacterial genera. Effects of dksA have been reported on gene expression, chaperonin function, cell division, amino acid requirements, quorum sensing, phage sensitivity, and virulence (11, 14–24). Of all these phenotypes, only the effect on rRNA expression has been demonstrated to be direct (11).

Artz and coworkers (25, 26) identified promoter mutants that abolished relA-dependent increases in activation of the hisGBD operon. Additional amino acid promoters respond to amino acid limitation in a relA-dependent fashion, increasing activity with decreasing growth rate, correlating with increases in ppGpp concentrations (12). However, only very small effects of ppGpp on the amino acid promoters (and on a few other promoters) have been observed in purified in vitro transcription systems (12, 27).

These observations led to the model that ppGpp might play primarily an indirect role in activation of amino acid promoters by inhibiting rRNA transcription, increasing the availability of free RNAP in vivo (28, 29). Consistent with this model, positively controlled amino acid promoters require high concentrations of RNAP for transcription both in vitro and in vivo (29). However, these promoters make a long-lived competitor-resistant complex with RNAP and, thus, are resistant to inhibition by ppGpp (29). The model was consistent with a variety of in vivo observations: the amino acid promoters were unusually sensitive to changes in the available pool of Eσ⁷⁰ holoenzyme, their dependence on ppGpp was partially alleviated by fusion of an UP element to the core promoter (increasing the binding constant for RNAP), and RNAP mutants were identified that coordinately decreased rRNA and increased amino acid promoter activity (28, 29).

Identification of DksA as a transcription factor promoting effects of ppGpp on negative regulation of rRNA promoters, together with reports of ppGpp increasing transcription from amino acid promoters in cell extracts (3, 10), prompted us to reevaluate a role for ppGpp in direct activation of amino acid promoters. Here, we show that ppGpp can activate amino acid promoters directly in a purified system but only in the presence of purified DksA. This activation occurs subsequent to initial binding of RNAP to the promoter. Deletion of the dksA gene eliminates activation of the hisG promoter in vivo after amino acid starvation. We propose that ppGpp/DksA both directly and indirectly stimulates transcription from amino acid promoters.

Materials and Methods

In Vitro Transcription. Multiple-round in vitro transcription assays were performed by using RNAP (Eσ⁷⁰), a generous gift from R. Landick (University of Wisconsin, Madison), purified by standard procedures (30). N-terminally hexahistidine-tagged DksA was purified by affinity chromatography, and functioned indistinguishably from native DksA in vivo and in vitro (11). Plasmid templates were derivatives of pRLG770 containing promoter regions with the endpoints indicated in Table 2, which is published as supporting information on the PNAS web site. Linear templates (≈460 bp) were created by PCR amplification of promoter regions from the plasmids by using vector-specific primers 3392 and 3073 (Table 2). Reactants (DNA template, ppGpp, and DksA at the concentrations indicated in the Fig. 1 legend and in Table 1) were incubated at 30°C for 10 min in 40 mM Tris·HCl, pH 7.9/NaCl (at the concentration indicated)/5% glycerol/10 mM MgCl₂/1 mM DTT/0.1 μg/μl BSA/500 μMATP/200 μM CTP and GTP/10 μMUTP/[α-³²P]UTP (2.5 μCi; 1 Ci = 37 GBq). Transcription was initiated by the addition of RNAP, terminated after 10 min by the addition of an equal volume of formamide loading buffer, electrophoresed on 7 M urea-6% polyacrylamide gels, and visualized and quantified by phosphorimaging.

Fig. 1.

ppGpp and DksA can activate amino acid promoters in vitro. (A) In vitro transcription by using 0.5 nM supercoiled templates (upstream and downstream promoter endpoints are provided in Table 2), 5 nM RNAP, and 165 mM NaCl. Lane 1 in each panel, buffer only; lane 2, 100 μM ppGpp; lane 3, 2 μM DksA; lane 4, 100 μM ppGpp plus 2 μM DksA. Transcripts of interest are indicated by arrows. Fold activation is indicated for each lane (test transcript plus ppGpp, DksA, or both relative to addition of buffer only). (B) In vitro transcription by using 2 nM linear templates, 10 nM RNAP, and 55 mM NaCl. Linear templates, ≈460 bp amplified by PCR from the same plasmids as in A, contained the test promoter and the rrnB t1 terminator. Lanes contained the concentrations of ppGpp and/or DksA indicated in A. In addition, lanes 5–7 in PargI contained 0.2 μM DksA, 0.2 μM DksA plus 100 μM ppGpp, and 5 μM DksA, respectively. The transcripts migrating more slowly than the test transcripts likely resulted from readthrough of the terminator, transcripts initiating at the fragment ends, or from “template hopping.” Longer exposures than displayed in the figure were used for quantitation purposes. Transcription experiments were performed at least twice. Increasing the concentration of ppGpp to 400 μM resulted in similar effects on transcription (ref. 12; data not shown). See Materials and Methods for more details.

Table 1. ppGpp and DksA activate transcription by increasing the rate of an isomerization step on the pathway to open complex formation.

	Control	0.2 μM DksA plus 100 μM ppGpp	5 μM DksA
ka × 10-5, M-1s-1	1.04 ± 0.03	15.94 ± 1.03	4.25 ± 0.22
ki × 102, s-1	0.9 ± 0.1	7.1 ± 1.2	3.4 ± 0.5
KB × 10-7, M-1	1.17	2.26	1.25
kd × 104, s-1	0.20 ± 0.08	13.00 ± 0.14	6.92 ± 0.31
t1/2, min	>240	10.7 ± 0.5	16.7 ± 0.9

Association rate constants were from nonlinear analysis of the data (Fig. 6), which provides a more accurate estimate of k_i and experimental error. t_1/2, competitor-resistant complex lifetimes. The value of t_1/2 for Pargl in the absence of ppGpp or DksA is a minimum estimate, because only an ≈20-30% decrease in the fraction of competitor-resistant complexes was obtained after 4 h.

Association and Dissociation Kinetics. ³²P-end-labeled promoter fragments were prepared, and filter binding assays were performed and analyzed essentially as described in refs. 12, 31, and 32 and in the legend for Figs. 5 and 6, which are published as supporting information on the PNAS web site. Kinetic constants were determined by using nonlinear analysis by fitting k_obs = k_ak_i[RNAP]/(k_a[RNAP] + k_i), where k_a is the composite second order association rate and k_i is the isomerization rate to a heparin-resistant complex. The initial equilibrium binding constant, K_B, was derived from k_a = K_B·k_i.

To measure the kinetics of dissociation, 10 nM RNAP was incubated at 30°C for 20 min with ≈0.5 nM radioactively labeled DNA fragment and DksA and/or ppGpp as indicated by using the concentrations and conditions indicated in the Fig. 5 legend (see also ref. 31). Irreversible dissociation was initiated by adding heparin (10 μg/ml, final); samples were removed at intervals, filtered, washed, and analyzed as in the association experiments. k_d was determined from the first-order decay equation, cpm_corr = (cpm₀)e^–k_d^t, where cpm_corr represents radioactivity corrected for background retention and cpm₀ represents radioactivity at time 0.

RNA Extraction and Primer Extension. Wild-type and ΔdksA strains containing promoter-lacZ fusions were grown at 30°C in 4-morpholinepropanesulfonic acid minimal medium supplemented with 0.4% glycerol/0.4% casamino acids/40 μg/ml tryptophan/10 μg/ml thiamine (see ref. 11 and Fig. 3 legend). Serine hydroxamate (0.24 mg/ml final concentration; Sigma) was added to induce amino acid starvation when indicated. Boiling lysis extraction of RNA and reverse transcription were performed essentially as described in ref. 11 (see Table 2 for primers). An RNA with the same primer-binding site as that from the rrnB P1-lacZ fusion, but at a different position in the transcript so as to create a primer extension product of a different size, was added at the time of lysis. Normalization to this “recovery marker” allowed correction for differential losses between samples that potentially could occur at any step subsequent to cell lysis.

Fig. 3.

dksA is required for activation of PhisG in vivo after amino acid starvation. PhisG (A) and rrnB P1 (B) activities were assayed by primer extension (by using primers hisL and 1159, see Table 2). The relevant regions of gels from a representative experiment are displayed above the graphs. The steady-state levels of the transcripts from PhisG and rrnB P1 were reduced and increased, respectively, in the ΔdksA strain (see text). In the graphs, the activities of promoters are plotted relative to their activities at time 0 (before addition of serine hydroxamate). Wild-type (RLG5950, •) and ΔdksA (RLG7062, ○) strains were grown for >4 doublings to an OD₆₀₀ 0.2 before amino acid starvation was elicited by the addition of serine hydroxamate (see Materials and Methods for details). RM, recovery marker. Graph shows band intensities corrected for optical density and culture volume, normalized to the recovery marker. Error bars indicate the standard deviation from four independent experiments.

Results

ppGpp Can Activate Amino Acid Promoters Directly in the Presence of DksA. We examined whether the addition of DksA to a purified in vitro transcription system would recreate effects of ppGpp on amino acid promoters previously observed only in vivo or in cell extracts. Four amino acid promoters, PhisG, PargI, PlivJ, and PthrABC, were chosen for study because their activities are reduced in a strain lacking ppGpp (ΔrelAΔ spoT) (12). The wild-type rrnB P1 promoter was tested in parallel as a control promoter inhibited by ppGpp/DksA, and rrnB P1(dis) (an rrnB P1 promoter with mutations that greatly increase the lifetime of the complex it forms with RNAP) was used as a control promoter relatively unaffected by ppGpp/DksA in vivo or in vitro (refs. 11 and 12; data not shown).

Transcription was performed first by using supercoiled plasmid templates, each containing one of the test promoters (Fig. 1A). The addition of saturating concentrations of ppGpp alone (as defined from effects on rrnB P1) (12) resulted in only slight positive effects on transcription from the amino acid promoters (≈15–35% increases), had no effect on rrnB P1(dis) and decreased wild-type rrnB P1 promoter activity by ≈40% (Fig. 1 A, lane 2 in each panel). The addition of 2 μM DksA alone also had little or no positive effect on transcription from the amino acid promoters and rrnB P1(dis), whereas it inhibited transcription from rrnB P1 by ≈50% (Fig. 1 A, lane 3 in each panel). In contrast, the addition of 100 μM ppGpp and 2 μM DksA together (Fig. 1 A, lane 4 in each panel) stimulated transcription from the amino acid promoters 4- to 8-fold, whereas it increased transcription from rrnB P1(dis) only slightly and inhibited transcription from wild-type rrnB P1 5- to 10-fold.

Effects of ppGpp/DksA were also tested by using linear templates containing the same promoter fragments as in the plasmids. As in Fig. 1 A, there was little or no stimulation of the amino acid promoters or of rrnB P1(dis) by either 100 μM ppGpp alone or 2 μM DksA alone, and wild-type rrnB P1 was inhibited by only ≈40–60% (Fig. 1B). However, ppGpp and DksA together had large synergistic effects on transcription (Fig. 1B, lane 4 in each panel), increasing amino acid promoter activities 3.5- to 11-fold and decreasing wild-type rrnB P1 activity by ≈20-fold, whereas it affected the control promoter rrnB P1(dis) little or not at all. Effects on transcription were also observed with ppGpp and lower DksA concentrations or with higher concentrations of DksA alone (Fig. 1B, lanes 5–7; see also discussed below and Fig. 2).

Fig. 2.

Two-step mechanism for open complex formation. Free RNAP (R) binds to promoter DNA (P) to form the closed complex (RP_c), which then isomerizes to form the open complex (RP_o). k_a is the composite second-order association rate constant. K_B is the initial equilibrium binding constant. k_i is the isomerization rate constant. k_d is the overall dissociation rate constant.

ppGpp and DksA Activate Transcription by Increasing the Isomerization Rate. We performed kinetic analyses to quantify effects of ppGpp/DksA on PargI and to determine the step(s) affected by ppGpp/DksA. We chose the argI promoter, because previous kinetic characterization (12) had indicated that RNAP displayed high promoter occupancy, facilitating analysis by our methods.

Although previous work demonstrated that ppGpp alone had no effect on escape of RNAP from the argI promoter (12), the structural similarity between DksA and Gre factors (small proteins shown previously to affect promoter clearance) (33) prompted us to test the effect of DksA and ppGpp together on this step. After preforming open complexes, a single round of transcription was initiated by the addition of NTPs and heparin. Promoter clearance was assayed by measuring the kinetics of appearance of transcripts and abortive initiation products. No significant effects were observed of ppGpp/DksA on the rate of promoter clearance (Fig. 7, which is published as supporting information on the PNAS web site). In addition, no increase in transcription was observed if open complexes were preformed (data not shown), suggesting that transcription activation of PargI by ppGpp/DksA does not occur after this step in initiation.

Therefore, we next characterized the effect of ppGpp/DksA on steps leading to open complex formation. Previous work demonstrated that ppGpp alone increased the rate of association of RNAP with PargI only slightly (12). Effects of ppGpp and DksA together were assessed by using the same solution and temperature conditions as in the in vitro transcription assay on the linear template (Fig. 1B), except with 10-fold less DksA. The lower concentration of DksA (0.2 μM) minimized effects on dissociation, thereby increasing overall promoter occupancy and facilitating accurate measurement of the association rate.

The kinetics of association on PargI were measured either with 100 μM ppGpp alone, with 0.2 μM DksA alone, or with both together. A filter-binding assay was used to measure the rate of formation of competitor (heparin) resistant complexes. At a single RNAP concentration (10 nM), ppGpp alone increased the RNAP association rate, k_obs, only ≈40%, and DksA alone increased the association rate only ≈2-fold, whereas DksA and ppGpp together increased the association rate ≈13-fold (data not shown). (For comparison, 100 μM ppGpp and 0.2 μM DksA together activated PargI 2.9-fold in the multiple-round transcription assay; Fig. 1B, lane 6). Association rates were therefore measured at multiple RNAP concentrations to determine the effect of ppGpp/DksA on the overall second order association rate (k_a) and on the rate of isomerization (k_i) from RP_c to a heparin-resistant complex (Fig. 5). The results are shown graphically as a double reciprocal (Tau) plot (Fig. 6A), and the kinetic constants were calculated by using a nonlinear analysis (Table 1; Fig. 6 B and C) that provides a more accurate estimate of k_i and of experimental variation (31, 32). One hundred micromolar ppGpp/0.2 μM DksA increased k_a 15.9-fold, with most of the effect resulting from an increase in k_i. The derived equilibrium binding constant, K_B, increased ≈2-fold; this increase most likely can be accounted for by experimental error in measuring the rapid rate of association under these solution conditions in the presence of ppGpp/DksA. A more detailed statistical analysis of the data based on the method of Stewart et al. (34) reduced the apparent effect attributable to K_B (P. Suthers, B.J.P., and R.L.G., unpublished data).

Effects of ppGpp and/or DksA on the dissociation rate constant, k_d, were also measured. Dissociation of RNAP from the argI promoter was extremely slow in the absence of ppGpp and/or DksA. Although 0.2 μM DksA/100 μM ppGpp together increased k_d ≈65-fold, the PargI competitor-resistant complex was still relatively long-lived at >10 min. (By comparison, the competitor-resistant rRNA promoter complex is so short-lived that it could not even be measured under these solution conditions.) Thus, the potential for ppGpp/DksA to inhibit transcription is realized only when the competitor-resistant complex is sufficiently short-lived.

Effects of DksA alone on association and dissociation rates of PargI competitor-resistant complexes were also measured, because slight activation was observed by in vitro transcription at higher concentrations of DksA in the absence of ppGpp. At 5 μM DksA, where in vitro transcription of PargI was activated ≈2-fold (Fig. 1B, lane 7), k_a increased ≈4-fold, and an increase in the isomerization rate accounted for the entire effect (Table 1; see Discussion). Although 5 μM DksA was saturating for its positive effect on transcription (data not shown), the effect of DksA alone was not as large as observed in the presence of ppGpp and DksA together. DksA (5 μM) alone also substantially increased k_d of the PargI competitor-resistant complex (≈35-fold) but, as indicated above, this result had no effect on transcription.

The ΔdksA Mutation Eliminates Regulation of an Amino Acid Promoter in Vivo. Effects of amino acid starvation on amino acid promoters in vivo have been studied in the most detail by using the hisG promoter (3, 25, 35). Therefore, we investigated effects of dksA on PhisG after amino acid starvation, which induces ppGpp synthesis in both wild-type and ΔdksA strains (11, 24). Promoter activity was estimated by direct measurement of his RNA by primer extension with an oligonucleotide primer (Table 2) annealing near the beginning of hisL, the RNA leader in this operon. hisG promoter activity increased 2-fold by ≈12–20 min after starvation of wild-type cells, whereas there was no increase in the ΔdksA strain (promoter activities normalized to their respective activities at time 0; Fig. 3A). RNA measurements by using a primer (Table 2) annealing to the coding region of the first structural gene, hisG, yielded similar results (data not shown), in agreement with reports in refs. 25 and 26 that increases in transcription initiation account for the predominant effect of ppGpp on hisG expression. For comparison, rrnB P1 promoter activity was also measured. Consistent with previous results, rrnB P1 transcription in the wild-type strain was barely detectable by 12 min after induction of the stringent response, whereas it was relatively unaffected in the ΔdksA strain (Fig. 3B). Furthermore, we observed that the ΔdksA mutation decreased the steady-state level of hisG transcripts and increased the steady-state level of rrnB P1 transcripts (RNA bands at time 0; Fig. 3).

The results with hisG and the previous observation (12) that transcription from several amino acid promoters was reduced at lower steady-state growth rates in ΔrelAΔspoT cells, relative to wild-type strains, encouraged us to test the effects of dksA on lacZ fusions to three other amino acid promoters at a range of steady-state growth rates. The thrABC promoter displayed reduced activity at lower growth rates in the mutant compared with the wild-type strain (Fig. 8, which is published as supporting information on the PNAS web site). Small but reproducible effects of dksA on steady-state transcription were also observed for PargI (Fig. 8) and PlivJ (data not shown).

Discussion

ppGpp Activates Transcription in a Purified System Containing DksA. Effects of amino acid starvation on gene expression in bacteria have been under investigation for >50 years (36). Although a requirement for ppGpp as a positive regulator of certain amino acid biosynthetic and transport genes was established many years ago (3), the mechanism(s) responsible have remained elusive. Some recent investigations have favored indirect models for activation (28, 29), because large effects of ppGpp on transcription from the amino acid promoters were not observed in purified systems in vitro, but increases in amino acid promoter activity were observed in vivo under conditions that disfavored rRNA transcription. We demonstrate here that ppGpp, in conjunction with DksA, can activate transcription from amino acid promoters directly in a purified system and that this activation occurs by increasing the rate of an isomerization step on the pathway to open complex formation.

Activation by ppGpp and DksA in vitro is highly synergistic, as was also the case for negative regulation of rRNA promoters (11). We describe DksA as “potentiating” the regulation of amino acid promoters by ppGpp, because DksA concentrations are relatively constant under all conditions tested to date (ref. 11; B.J.P., C. Gu, and R.L.G., unpublished data), whereas ppGpp concentrations change significantly with varying growth conditions (37). In this view, ppGpp serves as the regulator responding to internal signals, and DksA serves as a coregulator.

Effects of dksA on Positive Regulation of Amino Acid Promoters in Vivo. Strains lacking dksA displayed defects in relA-dependent transcription from several amino acid promoters, but these promoters were not affected uniformly. In a wild-type strain, PhisG activity increased ≈2-fold during a stringent response, a magnitude in close agreement with previous promoter measurements observed by using other methods in refs. 26 and 35, whereas activation of PhisG was eliminated in ΔdksA strains. Furthermore, thrABC promoter activity increased much less at low steady-state growth rates in the ΔdksA strain than in the wild-type strain (Fig. 8). Smaller effects of dksA were also observed for PargI (Fig. 8).

More broadly, these results indicate that the magnitude of regulation of different amino acid promoters by ppGpp/DksA is not identical and will require quantitation on a promoter-by-promoter basis. Eliminating ppGpp had a larger effect on amino acid promoters during steady-state conditions than eliminating dksA (i.e., compare effects of ΔdksA reported here with those of ΔrelAΔspoT reported in ref. 12). We suggest that detection of defects in regulation in ΔdksA strains could potentially be compromised by compensatory effects of other DksA-like proteins, such as GreA and GreB, that can inhabit the secondary channel of RNAP (38, 39) or by other compensatory mechanisms, especially given the extended time intervals inherent in steady-state assays.

Multiple Mechanisms Affecting Positive Control. Although the results presented here demonstrate a clear role for DksA and ppGpp in direct activation of at least some positively regulated amino acid promoters, we emphasize that positive regulation by ppGpp/DksA need not all be direct. Deletion of the dksA gene completely eliminated the response of the hisG promoter to amino acid starvation, but this result does not differentiate between effects of dksA mediated directly versus indirectly. We suggest that ppGpp/DksA increases activities of some amino acid promoters indirectly by increasing the concentration of free RNAP liberated from reduced transcription of stable RNA promoters (28, 29) and also directly by increasing the association rate of RNAP. Quantifying the relative contributions of these two mechanisms to overall increases in amino acid promoter activity during nutritional limitations and assessing their contributions relative to ppGpp-independent operon-specific mechanisms (e.g., attenuation and/or repression) is likely to be experimentally challenging.

Kinetic Basis for Control by ppGpp/DksA. We show here that DksA and ppGpp synergistically increase the isomerization rate constant, k_i, on the pathway to open complex formation, likely accounting for their direct role in positive control of amino acid promoters in vivo. However, at all promoters studied to date (11), DksA and ppGpp decrease the lifetime of the competitor-resistant complex. The seemingly paradoxical observations that ppGpp and DksA together activate transcription from amino acid promoters, whereas they inhibit rRNA promoters, can be rationalized by considering the microscopic rate constants that contribute to initiation at these promoters. The rate of collapse of the open complex (or of some other kinetic intermediate that contributes to the dissociation constant in our half-life assays) appears to be rate-determining for rRNA transcription initiation (12, 40, 41). In contrast, the dissociation rate is slow enough at amino acid promoters that it is not rate-determining for transcription even when it is enhanced by ppGpp/DksA. Unlike at amino acid promoters, apparently the forward reaction is efficient enough at rRNA promoters that it does not require enhancement by ppGpp/DksA.

A thermodynamic model (42) depicting open complex formation at two different promoters provides an intuitive framework for understanding the opposing effects of ppGpp/DksA on these promoters (Fig. 4). In this energy diagram, RNAP-promoter association is depicted at an amino acid promoter and at an rRNA promoter progressing from an earlier intermediate RP₁ to a later intermediate RP₂, two complexes on the pathway to open complex formation with a single transition state RP^‡. The energy barrier that must be overcome for formation of the later intermediate is large at the amino acid promoter, but this complex is stable once achieved, because its free energy is lower than that of the earlier intermediate (Fig. 4A). ppGpp/DksA is proposed to decrease the free energy of the transition state (dashed line), thus facilitating the forward reaction. (In theory, ppGpp/DksA could affect more than one transition state.) The ΔG required for the backward reaction is still large enough, even in the presence of ppGpp/DksA, that the later intermediate does not decay before transcription initiates. In contrast, the energy barrier that needs to be overcome for formation of the later intermediate at an rRNA promoter is relatively small and occurs readily (Fig. 4B). However, in this case, the complex is unstable, because its free energy is higher than for the earlier intermediate. Thus, when ppGpp/DksA decreases the free energy of the transition state(s), decay of the later intermediate is facilitated, and transcription is inhibited.

Fig. 4.

Thermodynamic representation of promoter activation versus inhibition by ppGpp/DksA. Free energy diagrams for an amino acid promoter (A) and an rRNA promoter (B) showing the progression of transcription initiation. The diagram illustrates two steps, RP₁ and RP₂, on the pathway to open complex formation with a single transition state ([RP]^‡). A two-step mechanism is depicted here for simplicity, but RP₁ and RP₂ could refer to any two intermediates on the pathway to open complex formation, and ppGpp/DksA conceivably could affect any transition state between the two intermediates. ΔG₁ and ΔG₂ represent the changes in free energy required to transition from RP₁ to [RP]^‡ and [RP]^‡ to RP₂, respectively. The dashed line represents the free energy of the transition state in the presence of ppGpp/DksA. ΔG_1D and ΔG_2D represent the changes in free energy in the presence of ppGpp/DksA. ppGpp/DksA is proposed to decrease the free energy of [RP]^‡, allowing for a more rapid conversion from RP₁ to RP₂ and from RP₂ to RP₁. The ground state of both RP₁ and RP₂ at a particular promoter would dictate whether DksA would facilitate activation or inhibition of the promoter.

Interestingly, high concentrations of DksA can facilitate open complex formation at amino acid promoters, even in the absence of ppGpp (Fig. 2). In contrast, there is little or no effect of ppGpp by itself, even at high concentrations (ref. 12; see also ref. 27). One interpretation of this result is that DksA binding favors the conformational change in RNAP that increases k_i, and ppGpp increases the ability of DksA to facilitate this transition. Direct contacts between ppGpp and DksA need not be required for this synergism; conformational changes caused by each molecule independently could, in theory, lead to their synergistic effects on transcription. We note that DksA is not the only example of a transcription factor that activates a promoter without binding DNA (43).

Structural Basis for Control by ppGpp/DksA. The recent high resolution x-ray structures of E. coli DksA (13) and of Thermus thermophilus RNAP holoenzyme in complex with ppGpp (44) are important steps in the development of a structural basis for the effects of ppGpp/DksA on the rate of complex formation and dissociation. ppGpp is positioned in two different orientations in RNAP-ppGpp cocrystals, located near (but not overlapping) the active site, with the diphosphate of ppGpp more distal to the active site facing the outlet to the secondary (NTP) channel (44). It was suggested that the two different orientations of ppGpp in the cocrystal might correlate with positive versus negative effects on transcription (44). Our finding that ppGpp can both stimulate and inhibit transcription directly in vitro is consistent with such a model, although the model illustrated in Fig. 4 could potentially explain the opposing effects of ppGpp without the need to invoke opposing effects of alternative binding modes for ppGpp. Rather, the kinetic parameters of the target promoter could dictate the effect of bound ppGpp/DksA.

The structural similarity between DksA and GreA/GreB (13) and extensive biochemical investigations of DksA's interaction with RNAP indicate that DksA inhabits the RNAP secondary channel, extending its coiled-coil domain toward ppGpp and the active site (ref. 13; I. Toulokhonov, J. Mukhopadhyay, R. H. Ebright, and R.L.G., unpublished data). A model has been proposed in which DksA coordinates the Mg²⁺ ion bound to the diphosphate of ppGpp facing the outlet to the secondary channel (13). Whether such an interaction could be responsible for the reduction in the lifetime of the RNAP-promoter complex and/or reduce the free energy of the proposed transition state remains to be determined.

A preinsertion site for NTP entry into the prokaryotic RNAP elongation complex has been proposed in ref. 45. The position where ppGpp is found in the T. thermophilus cocrystal appears to overlap the position of the NTP entry (preinsertion) site in the structure of the yeast transcription elongation complex (46). Assuming that the NTP entry site occurs at a similar position in bacterial promoter complexes as it does in eukaryotic elongation complexes, and that the ppGpp binding sites defined in the T. thermophilus RNAP structure are the functionally relevant ones, this overlap suggests that ppGpp/DksA might influence NTP binding. However, the opposing direct effects of ppGpp/DksA on different promoters indicate that simple substrate competition (47) (or “trapping” of the complex in a closed state; see ref. 48) is unlikely to be the mechanism of ppGpp action.

A Global Role for DksA in Transcription. Our work suggests that some of the many phenotypes associated with dksA (11, 14–24) might be attributable to a direct role of DksA in the synthesis of products that play roles in these processes. More generally, the action of DksA provides an example of the versatility of mechanisms that have evolved to regulate transcription initiation. In this case, DksA has evolved a direct and promoter-specific mechanism for regulating RNAP holoenzyme activity without employing a transcription factor binding site on DNA that determines its promoter specificify. Promoter sequence-independent interactions of DksA with RNAP result in promoter-specific effects on transcriptional output that derive solely from the intrinsic kinetic characteristics of the promoter. The action of DksA provides a model for how other small proteins, RNAs, and small molecules in addition to ppGpp could conceivably interact in a promoter sequence-independent manner with RNAP but exert promoter-specific effects on gene expression.