Research Article

Transcription Regulation by Initiating NTP Concentration: rRNA Synthesis in Bacteria

See allHide authors and affiliations

Science  19 Dec 1997:
Vol. 278, Issue 5346, pp. 2092-2097
DOI: 10.1126/science.278.5346.2092


The sequence of a promoter determines not only the efficiency with which it forms a complex with RNA polymerase, but also the concentration of nucleoside triphosphate (NTP) required for initiating transcription. Escherichia coli ribosomal RNA (rrn P1) promoters require high initiating NTP concentrations for efficient transcription because they form unusually short-lived complexes with RNA polymerase; high initiating NTP concentrations [adenosine or guanosine triphosphate (ATP or GTP), depending on the rrn P1 promoter] are needed to bind to and stabilize the open complex. ATP and GTP concentrations, and thereforerrn P1 promoter activity, increase with growth rate. Because ribosomal RNA transcription determines the rate of ribosome synthesis, the control of ribosomal RNA transcription by NTP concentration provides a molecular explanation for the growth rate–dependent control and homeostatic regulation of ribosome synthesis.

Protein synthesis is the dominant activity of the bacterial cell (1). Ribosome synthesis rates increase approximately with the square of the growth rate to increase protein synthesis at higher growth rates and to conserve biosynthetic energy at lower growth rates. The relation between growth rate and ribosome synthesis rate, referred to as growth rate–dependent control, was described almost 40 years ago and has been the subject of intensive investigation ever since (2, 3). Models have been proposed to explain the phenomenon, but the molecular mechanism or mechanisms responsible have not been determined (4).

Ribosomal RNA (rRNA) transcription is the rate-limiting step in ribosome synthesis, because ribosomal protein synthesis rates are regulated by feedback mechanisms sensitive to the rRNA concentration (5). In each of the seven rrn operons in E. coli, rRNA is transcribed from two promoters, P1 and P2 (Fig.1A). Most rRNA transcription at moderate to high growth rates originates from the P1 promoters, whose activities increase with growth rate and are thus responsible for regulation (6). Multiple systems affect transcription by rrnP1 promoters. Positive effectors include (i) a promoter upstream (UP) element that increases rrn P1 activity by binding the α subunit of RNA polymerase (RNAP) (7-9); (ii) a transcription factor, FIS, that binds to sites upstream of the UP element and interacts directly with RNAP (10, 11); and (iii) antitermination factors that bind to the BoxA region in the precursor RNA downstream of rrn P2 and prevent premature transcription termination (12). In addition, a negative effector, ppGpp, inhibits transcription from both rrn P1 and rrnP2 during amino acid starvation, a phenomenon referred to as the stringent response (13-15). Overlapping mechanisms influencing rRNA transcription have complicated efforts to identify the major system (or systems) contributing to growth rate–dependent control.

Figure 1

(A) TherrnB promoter region. Transcripts from promoters P1 and P2 are represented by arrows. DNA regions corresponding to –10 and –35 hexamers (the core promoter), UP elements, FIS binding sites, and the BoxA antitermination region are indicated. (B) P1 core promoter sequences (–41 to +6) from the rrnB andrrnD operons. The –10 and –35 hexamers and the transcription start sites are underlined.

Previously, we evaluated the contributions of the above mechanisms to growth rate–dependent control of the rrnB P1 promoter, using promoter or gene mutations to systematically eliminate specific input signals. Transcription from a “minimal” (core)rrnB P1 promoter (lacking rrnB sequences upstream of –41 with respect to the transcription start site, +1) (Fig. 1B) was growth rate–dependent even in strains unable to make ppGpp (16), which implied that some other mechanism—independent of FIS, the UP element, antitermination factors, or ppGpp—is responsible for growth rate–dependent regulation. However, there was no evidence for the binding of potential regulatory proteins (other than RNAP) to the rrnB P1 core promoter region (17). Therefore, we considered the possibility that the concentration of nucleoside triphosphates (NTPs), the substrates of RNAP, might serve as a signal that differentiates rrn P1 from other promoters in a manner that changes with growth rate.

Requirement for high concentrations of the initiating NTP for efficient rrn P1 transcription in vitro and in vivo. We used in vitro transcription to test whether varying the concentrations of NTPs, singly or in combination, would affect transcription from rrn P1 promoters differently than from control promoters. Control (RNA I or lacUV5) orrrnB P1 promoters were fused to the plasmid vector at position +1 such that each promoter made a transcript of identical sequence (Fig. 2A) (18, 19). When the concentration of adenosine triphosphate (ATP, the initiating NTP for each promoter) was varied and the concentrations of guanosine, uridine, and cytidine triphosphate (GTP, UTP, and CTP) were kept constant, maximal transcription from rrnB P1 required about 10 times as much ATP as did transcription from control promoters (Fig. 2, B and C) (19). The absolute concentration of ATP required for maximal transcription from rrnB P1 varied with solution conditions, increasing with increasing salt concentration or on linear (rather than supercoiled) templates. However, the ATP concentration needed for maximal transcription from rrnB P1 was greater than for control promoters under all solution conditions (20). Varying the amounts of the other NTPs, individually or together, had no selective effect on rrnB P1 activity (19).

Figure 2

Effect of ATP or GTP concentration on transcription of rrn P1 promoters in vitro. (A) Plasmid DNA templates containing different promoters made the same 170-nucleotide (nt) transcript terminated at rrnB T1 (18). (B) Transcription was performed at increasing ATP concentrations with constant GTP, UTP, and CTP (18), using either an rrnB P1 promoter template (left panel) or a control promoter template (RNA I, right panel). The 170-nt transcripts are indicated. The 108-nt RNA I transcript, originating from the plasmid's native RNA I promoter, is visible below the experimental transcript (7, 18). (C) The relative amounts of transcript from Fig. 2B (rrnB P1, •; cloned RNA I promoter, ○) at each ATP concentration are expressed as a fraction of the plateau value [1.00 (18)]. (D) Transcription from rrnD P1 (18,19) in the presence of varying GTP and constant ATP, UTP, and CTP (•), or with varying ATP and constant GTP, UTP, and CTP (○). Data from representative experiments are shown in (B) to (D); each experiment was performed at least three times, and differences in the apparent K m for ATP or GTP were ≤5%.

Six of the seven E. coli rrn P1 promoters begin transcription with ATP, but the rrnD P1 transcript starts with GTP (Fig. 1B). Maximal transcription of rrnD P1 in vitro was not selectively affected by varying ATP, UTP, or CTP concentrations, but was highly sensitive to GTP concentration (Fig. 2D) (18, 19). Moreover, substitution of G for A at position +1 of rrnB P1 also resulted in a requirement for high GTP, rather than ATP, concentrations (19). Thus, the concentration of the initiating NTP, rather than ATP concentration per se, affects the transcription efficiency of rrn P1 promoters in vitro.

To address whether variation in NTP concentration could account forrrn P1 regulation in vivo, we cultured cells in media supporting different growth rates and analyzed them for NTP content by reversed-phase ion pair high-pressure liquid chromatography (HPLC) (21). ATP and GTP concentrations increased by a factor of about 4 when growth rate increased by a factor of 2 (Fig.3A) (22), correlating with the increase in rrn P1 promoter activity observed from anrrnB P1 promoter fused to lacZ in the same cells (Fig. 3B).

Figure 3

rrnB P1 promoter activity correlates with ATP concentration in vivo. (A) ATP and GTP were measured by HPLC from cultures of RLG3492 with different growth rates (doublings per hour) (21). The concentration of ATP (•) differed from the concentration of GTP (○) at each growth rate, but the relative increase in concentration between the lowest and highest growth rates was almost identical for the two nucleotides. (B) rrnB P1 promoter activity [β-galactosidase units from an rrnB P1 promoter–lacZ fusion (21)] in the cultures used in (A). (C) ATP concentration and (D)rrnB P1 promoter activity in RLG3493, acar::Tn10 derivative of RLG3492, at different growth rates generated by varying the degree of pyrimidine limitation, which uncouples purine NTP concentration from growth rate (25). Symbols in each panel represent averages of three different samples of two different cultures for each growth rate.

This correlation suggested but did not prove that the increase in purine nucleotide concentrations with growth rate is responsible for regulation of rrn P1 transcription in vivo, because NTP concentrations could be saturating even at low growth rates (22). Therefore, we uncoupled purine NTP concentrations from growth rate by partially starving cells for pyrimidines, which reduces UTP and CTP concentrations (and growth rate) but increases the amounts of ATP and GTP (23-25). Under these conditions,rrnB P1 transcription increased with the ATP concentration rather than with the growth rate (Fig. 3, C and D). This observation indicated that the concentrations of purine NTPs, rather than the growth rate per se, regulate rrn P1 promoter activity in vivo.

Stabilization of rrn P1 open complexes by the initiating NTP. During transcription initiation, RNAP (R) forms a binary “closed” complex (RPc) with the promoter (P), isomerizes to form an “open” complex (RPo) in which the double-stranded DNA in the vicinity of the transcription start site is melted, and ultimately binds the initiating nucleotide: R + P ⇌ RPc ⇌ RPo ⇌ RPNTP(26). At most characterized promoters, RPo is relatively stable, with a half-life of 30 min to several hours under typical conditions in vitro (27). However, open complexes atrrn P1 promoters are exceptionally unstable (28-30), with half-lives generally one to two orders of magnitude shorter than those at more typical promoters under comparable solution conditions. At rrn P1 promoters, initiating NTP concentration dependence and open complex stability are strongly affected by salt concentration and template conformation in vitro (20, 28-30), which suggests that the requirement for high NTP concentrations might be related to complex instability.

Direct evidence for a role of NTPs in stabilizing rrnP1 promoter–RNAP complexes was obtained by measuring the half-life of complexes in the presence and absence of the initiating NTP (Fig.4) (31, 32). AtrrnB P1, which initiates with ATP, 2 mM ATP increased the half-life of the complex by a factor of about 3 (Fig. 4A), whereas GTP had little or no effect. At rrnD P1, which initiates with GTP, 2 mM GTP (but not ATP) increased the half-life, again by a factor of about 3 (Fig. 4B). These data suggest that the initiating NTP concentration influences the fraction of rrn P1 promoters present as open complexes, thereby affecting the amount of transcription.

Figure 4

Stabilization of rrn P1 promoter–RNAP complexes by initiating NTPs in vitro. Complexes ofrrnB P1 (A) or rrnD P1 (B) were formed with RNAP in the presence of ATP (▾) or GTP (○), or in the absence of NTPs (control; •). Symbols represent the fraction of complexes remaining at times after heparin addition (32). A representative experiment is shown, but differences between observed half-lives in the presence or absence of the initiating NTPs were highly reproducible (≤10% error).

Mutations that alter growth rate–dependent regulation were identified in the rrnB P1 promoter (16) and inrpoB and rpoC, encoding the β and β′ subunits of RNAP (33). The properties of complexes formed with the mutant promoter or the mutant RNAPs confirmed the importance of NTP concentration and open complex stability for rrn P1 regulation in vivo. A single base substitution at position –1 inrrnB P1 (rrnB P1 C-1T) that resulted in high transcription at all growth rates (Fig.5A) (16) drastically altered the ATP concentration dependence of the promoter: Maximal transcription of the mutant promoter required about one-tenth the amount of ATP required for maximal transcription of the wild-type promoter in vitro (Fig. 5B) (34). Furthermore, the complex containing the mutant promoter was about 5.5 times as stable as the wild-type complex (Fig. 5C) (34). The simplest interpretation of these results is that the mutation allows high rrnB P1 activity at low growth rates, because the promoter is transcribed efficiently even at low ATP concentrations.

Figure 5

Effects of promoter or RNAP mutations on growth rate–dependent regulation, NTP concentration dependence, and open complex stability. (A) Activities (in β-galactosidase units) of the rrnB P1 (–46, +1) wild type (RLG2663) or C-1T mutant (RLG2665) promoters were measured using promoter-lacZ fusions in cultures grown at different growth rates (16). (B) In vitro transcription of wild-type or C-1T mutant rrnB P1 promoters with wild-type RNAP at different ATP concentrations (34). Promoter activities are expressed as fractions of the plateau values (1.0). (C) Decay of open complexes containing wild-type RNAP or wild-type or C-1T mutant rrnB P1 promoters. Decay curves depict the fraction of open complexes remaining at different times after heparin addition, as described for Fig. 4 (32, 35). (D) Growth rate–dependent regulation was measured as in (A), using strains with wild-type RNAP (RLG3950) orrpoCΔ215–220 mutant RNAP (RLG3951), and anrrnB P1 (–61 to +50) promoter–lacZ fusion (33). Because the promoter-lacZ fusions used to monitor transcription activity are different in (A) and (D), the absolute activities should not be compared directly (8). (E) In vitro transcription with wild-type orrpoCΔ215–220 RNAPs and the wild-type promoter at different ATP concentrations (34). (F) Decay of open complexes containing wild-type rrnB P1 promoter (–61 to +50) and wild-type or rpoCΔ215–220 mutant RNAP. The lower salt concentration used in (F) resulted in a slightly slower decay rate for the promoter–wild-type RNAP complex than was observed in (C) (35).

A deletion of amino acids 215 to 220 in the RNAP β′ subunit (rpoCΔ215–220) resulted in low activity ofrrnB P1 promoters lacking FIS sites at all growth rates (Fig. 5D) (33). Relative to wild-type RNAP, the mutant RNAP required 8 to 11 times as much ATP for comparable rrnB P1 transcription in vitro (Fig. 5E), but not for RNA I transcription (33). Moreover, rrnB P1–mutant RNAP complexes were about 8% as stable as wild-type complexes (Fig. 5F) (35). These data suggest that rrn P1 expression in rpoCΔ215–220 mutant strains is altered because the NTP concentrations present even at the highest growth rates are insufficient to stabilize rrn P1 open complexes.

Model for homeostatic control of ribosome synthesis by NTP sensing. These data support a model (Fig.6) in which purine NTP pools control the rate of rRNA transcription—and thereby the rate of ribosome synthesis and the amount of translation—by stabilizing rrn P1–RNAP complexes in vivo. Intracellular ATP and GTP concentrations are determined by their rates of synthesis and consumption; synthesis rates are determined by nutritional conditions (which influence the efficiency of fermentation and respiration), and consumption is determined to a large extent by the amount of protein synthesis [ATP for amino acid biosynthesis and tRNA charging, GTP for tRNA binding to the ribosome and ribosome translocation (36)]. Transient imbalances between NTP generation and consumption thus create a feedback signal to readjust the rRNA synthesis rate to the translation rate and the nutritional state of the cell.

Figure 6

Model for homeostatic regulation of rRNA transcription and ribosome synthesis by the initiating NTP concentration. ATP and GTP, whose concentrations vary with growth rate (nutrient availability), regulate rRNA transcription by stabilizing RNAP (R)–rrn P1 promoter (P) open complexes (RPo). rRNA transcription determines the rate of ribosome synthesis and therefore the amount of translation. ATP and GTP are consumed during the process of translation, resulting in a feedback signal affecting rrn P1 transcription. Initiating NTP pools reflect the balance between protein synthesis rates and nutritional conditions.

Many previous observations are consistent with the predictions of this model. Conditions that inhibit translation—for example, chloramphenicol treatment or mutations in the translation apparatus, which would be expected to reduce the drain on purine NTP pools—result in overproduction of rRNA (37, 38). Conversely, conditional mutants of the glycolytic enzyme fructose 1,6-diphosphate aldolase have reduced amounts of ATP at the restrictive temperature, which might explain the observed transcription inhibition of rrn P1 promoters (39).

The model also provides a molecular explanation for the feedback control of rRNA synthesis previously observed. Overall rRNA expression remains relatively constant in situations that might be expected to perturb it (3). For example, total rRNA transcription remains roughly the same when cells contain as few as 4 or as many as 21 functional rRNA operons (38, 40), when rrnantitermination is defective because of nus mutations (41), or when rrn P1 transcription activation is defective because of fis or rpoA mutations (7, 10). In addition, transcription from rrn P1 promoters is decreased when cells overproduce rRNA from a λPL promoter (42). The adjustments inrrn P1 promoter activity in each of these situations can be attributed to over- or underproduction of translating ribosomes, resulting in changes in ATP and GTP pools.

Mechanism of NTP sensing by rrn P1 promoters. The effect of purine NTP concentration on rrn P1 promoter activity involves stabilization of the RNAP–rrn P1 complex. NTPs are the substrates of transcription, but we emphasize that the initiating NTP affects the rrn P1 promoter complex before catalysis occurs. The initiating NTP most likely functions as a ligand that binds to the open complex, presumably at the active site, leading to an increase in the observed half-life of the complex and a greater chance for productive transcription before the complex dissociates (31). The higher the NTP concentration, the greater the fraction of promoters in open complexes with RNAP, and the higher the extent of transcription.

The promoter sequences responsible for the instability of therrn P1 open complex are not understood. The determinants are likely to be complex, involving multiple aspects of promoter architecture. The sequence just upstream of the transcription start site is a likely determinant, because the C → T change at position –1 (43) increased the stability of the rrnB P1 open complex (Fig. 5C). It is striking that all seven rrn P1 promoters initiate transcription at the unusual distance of 9 base pairs (bp) from the –10 hexamer and contain the least preferred NTP for initiation, CTP, at the preferred positions 7 and 8 bp from the –10 hexamer (3, 44). However, the C-1T mutation increased stability without altering the transcription start site (13), indicating that start site position alone apparently is insufficient to account for this instability.

Other promoters (including the other rrn P1 promoters and many tRNA promoters) may also be controlled by variation in initiating NTP concentration, because they are subject to growth rate–dependent regulation, make unstable open complexes, or share other characteristics with rrnB P1 (38,45). Initiating NTP concentrations potentially could affect any promoter whose expression is limited by the stability of its open complex with RNAP and is poised such that physiologically relevant NTP concentrations could affect its lifetime. Regulation of rRNA transcription by purine NTP concentration apparently is not limited to bacteria: ATP and GTP pools control mammalian rRNA synthesis as well, although the mechanism responsible is not understood (46).

Nucleotide concentrations affect the expression of many operons by mechanisms different from that reported here for rrn P1 promoters. For example, changes in amounts of pyrimidine NTPs can alter expression of pyrimidine biosynthesis and salvage operons by affecting start site selection, reiterative transcription, transcription elongation, transcription attenuation, and translation initiation efficiency (24, 44, 47). Adenine nucleotides modulate transcription by phosphorylation or dephosphorylation of components of transcription complexes (48). Adenine nucleotides have also been proposed to affect anti–σ factor function and thereby control transcription by at least two RNAP holoenzymes in Bacillus subtilis (49).

Overlap in rRNA regulation mechanisms. The NTP-sensing mechanism need not account for all aspects of rRNA regulation. Ribosomal RNA promoters integrate multiple input signals: RNAP α and σ subunit interactions with promoter DNA, FIS, ppGpp, antitermination factors, and rrn P2 all contribute to rRNA synthesis in vivo (3), and regulatory mechanisms affecting rRNA expression may partially overlap. For example, deletion of the fis gene does not decrease overall rRNA transcription or alter growth rate–dependent regulation, because there are compensating increases in rrn P1 core promoter activity, presumably through feedback signaling involving the NTP-sensing mechanism described above (10). Conversely, RNAP mutations (such asrpoCΔ215–220; Fig. 5) that decrease rrnP1 core promoter activity are compensated for, in part, by FIS: FIS activates the β′ mutant RNAP more than wild-type RNAP and restores growth rate–dependent regulation (33, 50), because FIS concentrations vary with nutritional conditions (17, 51).

Overlapping regulatory mechanisms appear to be an intrinsic feature of rRNA synthesis, perhaps because of the central role played by rRNA transcription in cell physiology. Additional systems may also contribute to rRNA regulation, either independently or by influencing the NTP-sensing or FIS-dependent activation mechanisms. In fact, any condition that alters rrn P1 promoter–RNAP stability could potentially play a role in rRNA regulation (52). For example, the mediator of stringent control, ppGpp, which is produced in large amounts after amino acid starvation, inhibits transcription by decreasing the half-life of the rrn P1 open complex (19, 30, 53), perhaps making the complex unable to be “rescued” by normal intracellular NTP concentrations. ppGpp is dispensable for growth rate–dependent control (16, 54), but even the low concentrations of ppGpp that are present during steady-state growth could conceivably supplement the NTP-sensing mechanism.

Models for the control of rRNA synthesis involving substrate limitation were considered, and discarded, previously (55, 56). In particular, after an upshift it was found that both NTP pools and rRNA transcription ultimately reached higher steady-state levels, but NTP concentrations dropped transiently while stable RNA synthesis increased almost immediately (56). These data were interpreted to mean that NTP concentrations do not correlate with rRNA transcription rates. The apparent conflict with the NTP-sensing model proposed here might reflect the contribution of additional mechanisms to regulation during growth transitions. Consistent with this hypothesis, rrnP2–derived transcripts are responsible for most rRNA transcription immediately after an upshift, and P1-derived transcripts become dominant only after about 30 min (57). Because transcription from the rrnB P2 promoter, which initiates with several C residues (58), is insensitive to reduced CTP concentrations in vitro (19), rrn P2 promoter activity could account for the reported transient inverse correlation between NTP concentrations and rRNA synthesis during an upshift. It is also possible that the increase in rRNA transcription that occurs immediately after upshift could result, in part, from loss of inhibition by ppGpp, because ppGpp concentrations drop quickly after shifts before attaining new steady-state levels (59).

In summary, the sequence of a promoter determines the concentration of the initiating NTP required for maximal transcription efficiency. At rrn P1 promoters, unstable open complexes serve as sensors of the concentration of the initiating NTP (ATP or GTP). Purine NTP concentrations reflect the nutritional state as well as the translational activity of the cell, and they satisfy the role of a feedback effector of rRNA transcription. NTP sensing thus provides a molecular explanation for the growth rate–dependent regulation that is observed even in the absence of all other systems known to affect rRNA transcription.

  • * To whom correspondence should be addressed. E-mail: rgourse{at}


View Abstract

Navigate This Article