Chapter 90

Regulation of Ribosome Synthesis

JOHN KEENER and MASAYASU NOMURA

INTRODUCTION

In Escherichia coli and Salmonella typhimurium (official designation, Salmonella enterica serovar Typhimurium), as in most other organisms, it has long been appreciated that faster growing cells contain more ribosomes per unit cell mass than do more slowly growing cells (109) because the activity of the ribosomes, that is, the number of amino acids polymerized into protein per second per ribosome, is relatively constant over the range of growth rates usually studied in the laboratory (54, 117). Since about half of the E. coli cell mass is protein, cells that encounter rich nutritional conditions need to increase their ribosome contents in order to grow faster and thus compete effectively with other organisms in a rich environment. Conversely, cells in poor nutritional environments need to dedicate more of their cell mass to nutritional uptake, presumably in direct competition with other organisms, and to biosynthesis of cellular building blocks, rather than to synthesis of translational machinery in excess of that which they can supply with substrates. We will refer to such regulation of ribosome biosynthesis as growth rate control or regulation, although we believe that ribosome biosynthesis is actually regulated by the nutritional quality of the medium and that such regulation adjusts the intracellular concentration of ribosomes to maximize growth rate in a particular medium. This system is clearly of fundamental importance for competitiveness and efficient growth of E. coli, and since ribosomes and associated factors can account for up to 50% of cell mass at high growth rates (14), this regulatory system is the most significant system to E. coli in terms of the magnitude of products regulated.

There is at least one condition in which the regulation of ribosomes apparently does not take place. At very slow growth rates, E. coli cells appear to maintain excess nontranslating ribosomes, and these may allow nearly instantaneous response upon relief of the nutritional limitation (62). Possibly related to this is a stationary-phase phenomenon in which 70S ribosomes are coupled into inactive 100S dimers by a ribosome-associated protein (133). This protein is required for normal viability during stationary phase but not during log phase (140). The rest of this chapter is concerned with regulation of the synthesis of ribosomes.

E. coli ribosomes consist of three rRNAs and more than 50 ribosomal proteins (r-proteins). The RNA molecules found in ribosomes are processed from a larger precursor RNA transcript that contains, in order from the 5' end, the 16S, 23S, and 5S rRNAs, which ensures stoichiometric synthesis of rRNAs. Assembly of r-proteins S1 to S21 onto the 16S rRNA gives rise to the 30S small subunit, and assembly of r-proteins L1 to L36 onto the 23S and 5S rRNAs gives rise to the 50S large subunit. The linear relationship between cellular ribosome content (the number of ribosomes per unit cell mass or total cell protein) and growth rate shown in Fig. 1a means that the ratio of the rates of ribosome accumulation and total protein accumulation must also increase linearly with growth rate. Since the total protein accumulation rate is proportional to the growth rate, then the rate of ribosome accumulation (per unit amount of cell mass or total cell protein) increases roughly in proportion to the square of the growth rate, as shown in Fig. 1b. Experimental measurements demonstrated that synthesis rates of rRNA and individual r-proteins roughly reflect their accumulation rates (except for very slow growth conditions, under which rRNA degradation is observed); that is, the synthesis rates of rRNA and individual r-proteins in fact increase roughly in proportion to the square of the growth rate (46). It is now known that growth rate control of rRNA synthesis is achieved directly and that growth rate control of the synthesis of most, if not all, individual r-proteins is achieved indirectly as a consequence of the regulation of rRNA synthesis. In this chapter, we summarize current understanding of ribosome biosynthesis, with emphasis on more recent findings. Details of earlier work can be found in the previous edition of this book (57) and in other reviews (74, 87). We first discuss regulation of r-protein synthesis and then the regulation of rRNA synthesis.

Fig. 1Ribosome amounts or rates of synthesis plotted as a function of growth rate. (a) The number of ribosomes per unit of cellular protein (or per unit of cell mass) is linearly proportional to the growth rate except at very low growth rates, where it deviates from the extension of the linear function (dotted line). Therefore, the ratio of the rate of ribosome accumulation to the rate of protein accumulation must also be a linear function of growth rate. The synthesis rates of r-proteins are often expressed relative to the rate of total protein synthesis, termed ?r. Thus, ?r is linearly proportional to growth rate as well. Since most RNA is rRNA and tRNA, and most tRNA is coregulated with rRNA, the RNA/protein ratio is also linearly proportional to growth rate. Finally, when ?-galactosidase is used as a reporter for rRNA promoter activity, it is normalized to cell protein (or to cell mass as measured by the optical density of the culture, which is proportional to the cellular protein content). Because the amount of ?-galactosidase enzyme reflects the amount of rRNA expressed from the promoter, after the enzyme activity is normalized to cell protein (or cell mass), it should also be linearly proportional to growth rate if the overall efficiency of translation of mRNA transcripts from P1 does not change with growth rate. (b) Since the ratio of the accumulation rates shown in panel a is linearly proportional to growth rate, and since the protein accumulation rate (per unit cell protein or cell mass) is proportional to growth rate, then by substitution, the rRNA accumulation rate per unit protein (or cell mass) is predicted to be proportional to the square of the growth rate.

Source:

REGULATION OF r-PROTEIN SYNTHESIS

Two purposes for regulation of r-protein synthesis can be envisioned: to achieve stoichiometry between rRNA and r-proteins and to achieve stoichiometry among the various r-proteins. Underexpression of r-proteins relative to rRNA is a disaster for E. coli, as illustrated by the effects of sublethal concentrations of chloramphenicol. Because ribosome assembly is not completely cooperative, the r-proteins that are produced during a shortfall of r-protein synthesis are distributed among all the rRNAs available, resulting in many incompletely assembled ribosomes that are nonfunctional and thereby exacerbating the shortfall in r-protein synthesis (34). On the other hand, overexpression of r-proteins would be wasteful on a particularly large scale for E. coli. As is detailed below, two general mechanisms contribute to stoichiometric synthesis of r-proteins with each other and with rRNA. One mechanism is translational coupling, by which translation of a protein-coding sequence depends on translation of the gene preceding it. Translation of the upstream gene may melt an RNA secondary structure in which the ribosome-binding site of the downstream gene was sequestered and thereby expose it for initiation, or the ribosome itself might be transferred directly from the upstream gene to the initiation codon of the downstream gene. Translational coupling allows coordinate expression of as many as 11 gene products from a single r-protein mRNA. The second mechanism is regulation of translation of such a group of translationally coupled genes by a single repressor r-protein (shown in boldface type in Fig. 2) encoded by that mRNA. The repressor r-protein is usually a "primary" binding protein, which can bind directly to rRNA under conditions of in vitro ribosome reconstitution. When more of it accumulates than can be incorporated into assembling ribosomes, it then (usually) binds an operator sequence or structure in the mRNA (indicated by two asterisks in Fig. 2) and exerts repression. The repression often regulates translation of the first gene on the message directly, and the downstream genes are regulated because their translation is coupled to that of the first gene. However, the repressor r-protein may be targeted elsewhere, for example, between the second and third genes, in which case translation of the first two genes may be decreased by destabilization of the mRNA, termed retroregulation. Surprisingly, diverse mechanisms by which the repressor r-proteins block translation have been discovered. A naive expectation might be that the repressor would act by occluding the ribosome from the initiation site on the mRNA, but this is usually not the case. In some cases, the repressor r-protein causes the ribosome to be trapped in a nonproductive binary complex, perhaps in part because of interaction between the repressor and the ribosome, thus using the ribosome itself as a means of preventing initiation by other ribosomes. In another case, the repressor prevents translational coupling, probably by melting an RNA structure that is required for coupling. Notably, transcription of most r-protein operons is not growth rate regulated (46). We will discuss the most interesting features of r-protein operons, including cotranscription with other translation factors and RNA polymerase subunits. For a more extensive review, see reference 149.

Fig. 2Structure of r-protein operons. The operons containing r-proteins are followed by their map positions on the E. coli chromosome and the order of genes in each operon. When the repressor r-protein has been identified, it is shown in boldface type. The positions of the repressor-binding sites are indicated by double asterisks. The standard genetic designations for r-proteins are rpsA through rpsU for S1 through S21, rplA through rplY for L1 through L25, and rpmA through rpmJ for L27 through L36 (1). The following large-subunit protein designations have proven to be redundant: L7 is the acetylated form of L12, L8 represents the L10(L7/12)4 complex, and L26 is identical to S20. The gene sequences for L21 and L27, mapping at 69 min, and for L31, mapping at 89 min, have not yet been published. The standard genetic designations for RNA polymerase subunits are rpoA (?), rpoB (?), rpoC (??), and rpoD (?70). The gene encoding EF-Tu in the str operon is tufA. It is one of two nearly identical genes for EF-Tu in E. coli, the other being tufB, which is cotranscribed with a tRNA operon located at 90 min. EF-Ts is encoded by the tsf gene. att, attenuator; ter, terminator.

Source:

spc Operon

The spc operon, which is named for the spectinomycin resistance locus (spc or rpsE), which encodes S5, encodes 11 r-proteins and a component of the protein secretion apparatus, SecY (18). Upstream lies the S10 operon, and some S10 transcripts read through into the spc operon (72). The repressor protein S8 binds the mRNA between the second and third cistrons, thereby repressing initiation of translation of L5 (19). Translation of the downstream cistrons is coupled to that of L5, so that their translation is also repressed by S8 binding (76). It is not known whether synthesis of the last two proteins encoded by the operon, SecY and L36, is regulated by S8. Translation of the first two genes of the operon, L14 and L24, is repressed indirectly by S8 binding downstream of them (77). This repression is proposed to result from retroregulation, in which repressor binding causes endonucleolytic cleavage of the mRNA (likely due to lack of translation of L5 or genes downstream) followed by 3'-to-5' exonucleolytic degradation. In support of this type of mechanism, S8-mediated repression of L14 and L24 translation was decreased in a mutant lacking two 3' to 5' exonucleases (77).

The S8 binding site on spc mRNA is well characterized and is similar to its binding site on 16S rRNA (19, 51). The site consists of a stem with several bulged-out bases that sequesters the AUG initiation codon for L5 but not the Shine-Dalgarno (S-D) sequence (19, 51, 81). Whether a ribosome can perhaps still bind to the S-D sequence and be trapped there in a nonproductive preinitiation complex as for other r-protein operons (see below) is not known. Presumably, S8 binding stabilizes the stem structure. S8 has about fivefold less affinity for its binding site on mRNA than for that on rRNA (51). Mutations that make the mRNA site more similar to the rRNA site increase affinity for S8 (137).

S10 Operon

The S10 operon encodes 11 r-proteins, all translationally coupled and all regulated by the L4 protein, the product of the third gene of the operon (73, 144, 150). The coupling is leaky, because after elimination of translation of the first gene of the operon, S10, there remains about 20% of the fully coupled translational activity of the second gene of the operon, L3, which is independent of S10 translation (71). L4 appears to regulate both transcription and translation of the S10 operon, and the mechanisms are probably complex. Both levels of regulation are necessary for the full range of repression; each level contributes about one-fifth of the total of about 25-fold repression (41). Although binding has not been demonstrated, target sites for L4 regulatory activity have been localized genetically to the 170-base leader upstream of the S10 gene (41). Sequences necessary for L4-mediated transcriptional and translational control are not identical, but they overlap (146). Control of transcription is by premature termination at position 140 (with respect to the transcription start site) and has been demonstrated in vitro in a purified system consisting of RNA polymerase and the NusA and L4 proteins (147). The termination site lies within a proposed stem (positions 84 to 145) necessary for both transcriptional and translational control. Additional sequence upstream (positions 61 to 83), including a stem-loop structure, is required only for transcriptional attenuation (146). The common stem (positions 84 to 145) is sufficient for a NusA-dependent pause in vitro, and the upstream stem is necessary for the L4-mediated stabilization of the pause (147, 148). Translational regulation requires, in addition to the common stem mentioned above, sequence from 146 to 192, including a proposed stem that would sequester both the S-D sequence and the AUG initiation codon for S10 (146). The current model for translational regulation proposes that a sequence between the two stems acts as a ribosome entry site in the unrepressed state and that ribosomes bound there can then melt the stem that sequesters the initiation site (112). L4 interaction with the mRNA or the mRNA RNA polymerase complex is proposed to favor a structure which sequesters the ribosome entry site, in addition to the S-D sequence and the AUG initiation codon.

α Operon

The α operon encodes four r-proteins and the α subunit of RNA polymerase (7). Translation of the four r-proteins is regulated by S4 (27, 142), the product of the third gene of the operon, which binds in the leader region and within the first gene encoding S13 (30, 31). Translation of S13 is repressed directly, and synthesis of the other three r-proteins is repressed because their translation is coupled to that of S13. Expression of α, whose gene lies between those for the third and the fourth r-proteins of the operon, S4 and L17, is not translationally coupled to them and hence is only weakly regulated by S4 (125). This raises the interesting question, which has not been studied in detail, of how translation of L17 is coupled to that of S4 across the intervening cistron encoding α. Binding of S4 to α-operon mRNA has been well studied and was an early example that a translational repression mechanism could be more complex than simple occlusion of initiating ribosomes. The regulatory region of the mRNA can form a double pseudoknot structure, and binding of S4 stabilizes this structure (31, 123). A loop including the S-D sequence is still available for ribosome binding. When a ribosome does bind, S4 is thought to interact with it, thereby stabilizing a binary complex in which the initiator tRNA has not located the start codon (119, 124). Such a binary complex is distinguished from the initiation-competent ternary complex by toeprinting, a procedure that maps the "downstream edge" of the ribosome by determining where polymerization by reverse transcriptase is stopped. Thus, both repressor and ribosome are coresident in a repressed complex (118).

str Operon

The str operon, named for mutations to streptomycin resistance found in the first gene (str or rpsL) of the operon and coding for S12, also includes genes for (in order) S7, elongation factor G (EF-G), and one of the two genes in E. coli coding for elongation factor Tu (EF-Tu; the gene being tufA). S7 is the repressor r-protein for the str operon, and it binds str operon mRNA between the genes for S12 and S7, repressing its own synthesis (28). Synthesis of EF-G is partially repressed by S7 overproduction, and EF-Tu is only very weakly repressed (105). Synthesis of S12 is clearly regulated by S7 protein, even though the S7 binding site is downstream of the gene for S12. The regulation is presumably by retroregulation, because the kinetics of repression appear to be similar to those observed for L14 and L24 in the spc operon. Translation of S7 is coupled to translation of S12 but not absolutely; if the start codon for S12 is mutated to UAG, residual S12-independent translation is about 20% of the coupled level (105). Interestingly, S7 represses only coupled translation and not S12-independent translation of S7. A structure of the intergenic region proposed on the basis of chemical modification experiments in vitro was largely confirmed by mutational analysis in vivo. By the same methods, the S7 binding site was localized (106). Mutational analysis also identified parts of the structure required for efficient coupling of S7 translation to S12 translation, leading to a model in which S7 binding is thought to cause denaturation of a helix (or helices) whose integrity is essential for translational coupling (106). This example of translational coupling, in which coupled and independent translation appear to occur by distinct mechanisms, is one of the few cases for which there is evidence that the ribosome translating the upstream cistron is likely to be physically transferred to the downstream cistron without mixing with free ribosomes in the cellular pool.

L11 and L10 Operons

The L11/L1 operon is regulated by L1 binding to the leader region upstream of L11 and secondarily by coupling of L1 translation to that of L11 (5, 6, 27, 142). Although some deletions of the gene for L11 allow uncoupled translation of L1, that translation is inefficient, suggesting that ribosomes that have translated L11 may be more efficient at initiating translation of L1 (5, 116). L1 binding to mRNA has not been directly demonstrated, but analogous mutations in the proposed binding sites in both the L11 leader region and 23S rRNA relieved L1-mediated repression or decreased L1 affinity for 23S rRNA, respectively, indicating that similar features are recognized by L1 on both RNAs (104, 126). The mechanism by which L1 binding inhibits translation of L11 has not been determined.

The L10 operon lies immediately downstream of the L11 operon, and in fact, about 50% of L11 transcripts read through into it (16). Cotranscribed with L10 are the gene for L12 and the genes for the β and β' subunits of RNA polymerase, but transcription is attenuated about 80% between the coding sequences for L12 and β, and expression of β and β' is otherwise independently regulated (32, 35, 122, 141). L12, often referred to as L7/12, indicating the acetylated and unacetylated forms, is present in four copies per ribosome, and yet its translation is coupled to that of L10, which is present at one copy per ribosome. Translation of both L10 and L12 is repressed by L10 (or by a complex of L10 and L12 [15, 42, 143]). Although translation of L12 is dependent on translation of L10, L12 translation must involve initiation by ribosomes that have not translated L10. The intercistronic region between L10 and L12 is 66 bases long, and L12 translation also seems to depend on translation of the 5' end of the L10 gene (93). Whether the ratio of translation rates of L10 and L12 reflects the 1:4 stoichiometry found in the ribosome and exactly how the ratio might be achieved await further study (see discussion in reference 143). The complex of L10-(L12)₄ binds to the L10 leader 120 to 160 bases upstream of the L10 start codon, but how it represses translation of L10 is not known (21, 22, 58).

S15 Operon

The S15 gene lies upstream of the pnp gene encoding polynucleotide phosphorylase, an exonucleolytic RNase (97). About half of the transcripts terminate after S15, and pnp appears to be regulated independently of S15 (97). S15 binds to a regulatory site that extends from –60 to +13 relative to the initiation codon. This interaction has been well studied. RNA within this region is in equilibrium between a double-stem structure that only weakly sequesters the S-D sequence and a pseudoknot structure in which the S-D sequence is in a loop available for binding but the initiation codon is partly sequestered (82). S15 protein stabilizes the pseudoknot structure, but the 30S ribosomal subunit can still bind to the S-D sequence (95). In the presence of S15, though, the binary complex of 30S and the S-D sequence is stabilized, and even the initiator tRNA can bind to this complex, but the complex is unable to make a transition to an initiation-competent ternary complex (94). Presumably, this stabilized inactive complex also prevents binding of other ribosomes. Thus, autogenous repression of S15 translation is by a ribosome-trapping mechanism similar to S4-mediated regulation of the α operon.

Other Known Translationally Regulated r-Proteins

The genes for L35 and L20 lie in a complex operon downstream of genes coding for threonyl-tRNA synthetase and translation initiation factor 3 (IF3), both of which are regulated at the translational level independently of each other and of L35 and L20 (70, 134). Four promoters read toward the r-protein suboperon, but about half the transcripts are terminated between the IF3 and L35 cistrons (134). Translation of L20 is coupled to that of L35 (69). The mechanism proposed involves disruption of an RNA structure that sequesters the initiation site for L20 by ribosomes translating L35. Maximal L20-mediated inhibition of L35 translation requires sequences immediately upstream of L35 and also within the IF3 gene, suggesting a long-range interaction to form a repressing structure (70).

S20 is expressed from a monocistronic operon and is autogenously regulated posttranscriptionally (75, 91, 135, 136). Genetic studies revealed no evidence for S20 interaction with the leader region of its own mRNA (90). The initiation codon for S20 is UUG, and changing it to AUG eliminates autogenous regulation (90). The mechanism of regulation is unknown.

S1 is partially cotranscribed with a gene of unknown function located upstream and the gene coding for an integration host factor subunit (hip) downstream (92). As is true for other r-proteins, synthesis of S1 is regulated posttranscriptionally by S1, although the mechanism has not been determined (20, 115).

Given the variety of mechanisms found for regulation of the r-protein operons studied thus far, we can expect further surprises as some of the other operons listed in Fig. 2 are investigated.

REGULATION OF rRNA AND tRNA SYNTHESIS

E. coli has seven rRNA operons (termed rrn operons) in its genome, and the rRNAs encoded by them are all very similar, if not identical. After correction for position effects in the E. coli chromosome, these operons are all similarly expressed (25). tRNAs are found within the rRNA precursor transcripts, between the 16S and 23S rRNAs, and at the end of the transcripts following the 5S rRNA. The identities of these tRNAs vary from operon to operon (Table 1). Their synthesis is obviously coregulated with that of rRNA by cotranscription. In addition, synthesis of many tRNAs not encoded by rRNA operons is coregulated with synthesis of rRNA (Table 2).

Table 1rRNA operons and cotranscribed tRNAsa

Table 2tRNA operonsa

Feedback Regulation

Variation in rRNA expression as a consequence of cultivation in media of different nutritional qualities resulting in different growth rates is termed growth rate control and is exerted by a feedback mechanism. As is detailed below, it is the P1 promoter of rRNA operons that is growth rate regulated and is responsible for most rRNA expression. Feedback regulation was observed when the number of rRNA operons normally found on the chromosome was increased by two- or threefold by using plasmids bearing an rRNA operon (56). When an intact rRNA operon was present on the plasmid, expression of the chromosomal operons was repressed according to monitoring of expression of the unique tRNA species they encode or of expression of a β-galactosidase fusion to the P1 promoter, such that total rRNA expression was unchanged by the increase in gene dosage (48, 56). Expression of chromosomal tRNA operons was repressed as well. However, when the plasmid-borne operon was internally deleted so that nonfunctional rRNA was produced from the plasmid, repression of the chromosomal rRNA operons did not occur, and if the defective rRNA expressed from the plasmid is included, the total rRNA synthesis rate increased in proportion to the increased gene dosage (56). The conclusion was that functional rRNA, presumably assembled into ribosomes, was responsible for the observed repression of chromosomal rRNA operons that compensated for the increased gene dosage, and thus the repression was accomplished by a feedback mechanism.

The results of a complementary experiment support a feedback model. The gene dosage of rRNA operons was decreased by deletion of four of seven of the operons present in the E. coli genome (24). Expression of rRNA, as monitored by expression of a chloramphenicol acetyltransferase (cat) gene fused to the rRNA promoter region, was 2.3-fold depressed, thereby compensating for the decreased gene dosage of rRNA operons (24). (Interestingly, the decreased gene dosage also resulted in an increased rate of transcriptional elongation of rRNA. Although this increased rate may be required to achieve the increased production of rRNA per operon that was observed, it would not affect initiation frequency, which is growth rate regulated and was increased 2.3-fold in this case, fully accounting for the 2.3-fold derepression of individual rRNA operons [24]. Elucidation of the mechanism responsible for this phenomenon awaits further investigation.)

As originally formulated, the ribosome feedback model proposed that nontranslating free ribosomes were the feedback affectors. Free ribosomes were thought to be a signal of excess translational capacity, so it was logical to imagine that they might cause repression of rRNA expression. However, an experiment to test this hypothesis actually showed that the feedback mechanism depended on actively translating ribosomes. To experimentally control the amount of ribosomes translating, expression of IF2 was placed under the control of the lac promoter (23). Thus, limitation of isopropyl-β-d-galactopyranoside (IPTG) in the medium limited IF2 synthesis, which in turn limited translational initiation, causing an accumulation of nontranslating ribosomes. This accumulation resulted in an increased rate of rRNA synthesis, which is consistent with the idea that (actively) translating ribosomes exert feedback repression of rRNA expression (23). When a saturating amount of IPTG was added to the translationally limited cells, restoring normal levels of IF2, the number of ribosomes engaged in translation increased and expression of rRNA decreased, showing that the previously nontranslating ribosomes were capable of contributing to feedback repression when they became engaged in translation (23). The conclusion is that repression of rRNA synthesis by the feedback mechanism depends on translating ribosomes in excess of what the nutritional quality of the medium can support, suggesting that cells may use this mechanism to assess how completely they utilize the nutritional quality of the medium.

Another demonstration that feedback regulation requires translating ribosomes employed "specialized" ribosomes (139). These mutant ribosomes have an altered anti-S-D sequence in their 16S rRNAs and therefore exhibit altered translation initiation specificities. Whereas synthesis of excess wild-type (control) rRNA from the λ P_L promoter caused feedback repression of chromosomal rRNA and tRNA operons, comparable synthesis of specialized ribosomes failed to feedback repress (139).

The mechanism by which translating ribosomes exert feedback repression is not known. However, it seems reasonable that the translational apparatus signals the sufficiency or insufficiency of substrates (and/or energy supply) to support translation by the cells’ current level of ribosomes. A signal of substrate sufficiency would indicate that the cells were able to take up from the medium or synthesize translational substrates (or generate energy required for translation) in excess of their ribosomes’ capacities to consume them, and consequently they should increase rRNA synthesis, thereby allowing an increased growth rate. Conversely, a signal of substrate insufficiency (relative to translational capacity) should cause decreased synthesis of rRNA. If rRNA synthesis were decreased, then some cellular resources would be shifted from ribosome biosynthesis toward the provision of translational substrates (or energy supply for translation), thereby alleviating the insufficiency that caused the decreased rRNA synthesis. Thus, the feedback mechanism would act incrementally to balance the allocation of cellular resources between synthesis of ribosomes and synthesis of substrates (or energy supply) for translation.

Stringent Regulation

In addition to regulation in response to the nutritional quality of the medium, rRNA and tRNA are preferentially repressed during severe starvation for an individual amino acid; this repression is termed the stringent response. The product of the relA gene has long been known to be required for the stringent response (121). The RelA protein synthesizes the unusual nucleotides pppGpp and ppGpp (guanosine penta- and tetraphosphates, respectively; ppGpp is also known as "magic spot" [17]) when it interacts with a ribosome containing the cognate uncharged tRNA rather than the cognate amino-acylated tRNA in its acceptor site (52). Thus, the stringent response monitors severe deficiency of a particular aminoacyl-tRNA species. At the high intracellular concentration of ppGpp generated during severe starvation, synthesis of rRNA and most tRNAs ceases. Severe carbon source downshift or starvation also increases the ppGpp concentration, in this case mediated by the spoT gene product, most likely by inhibiting turnover of ppGpp but possibly by stimulating its proposed synthetase activity (66, 80, 138). The mechanism by which carbon starvation is communicated to SpoT protein is unknown. Like growth rate regulation, stringent regulation affects synthesis of both rRNA (from the P1 promoter [49, 108]) and tRNA and, as a consequence of its control of rRNA synthesis, indirectly affects r-protein synthesis.

ppGpp is also present during steady-state growth in E. coli but at much lower levels than during the stringent response. Its concentration varies inversely with the growth rate (103). Although ppGpp is an attractive candidate for the signal that mediates feedback repression and growth rate regulation, recent work, to be discussed in a later section, indicates that growth rate regulation of rRNA synthesis can occur in the absence of ppGpp. Thus, the role of ppGpp in signaling feedback repression is controversial at present. Another function of ppGpp is to stimulate transcription of some amino acid biosynthetic operons in response to a shortfall of aminoacyl-tRNA. For a more complete discussion of ppGpp metabolism and its effects on cell physiology, see chapter 92 of this volume.

Other Forms of Regulation

Two independent screens for mutants failing to synthesize rRNA and tRNA at high temperature yielded alleles of fda, the gene encoding fructose 1,6-diphosphate (FDP) aldolase (10, 113). In temperature-sensitive mutant cells grown on glucose at high temperature, FDP accumulates to high levels, and activity of the rRNA promoters is blocked (11, 114). Mutations that suppress the temperature-sensitive lethality and prevent FDP accumulation fail to block rRNA synthesis, implicating FDP as the molecule responsible for the inhibition (110). Promoters sensitive to FDP include the growth-rate-regulated and stringently controlled rRNA P1 promoter and the non-growth-rate-regulated but stringently regulated rRNA P2 promoter but not several rRNA P1 mutant promoters that show stringent but not growth rate regulation (114; see below also). Thus, FDP inhibition does not correlate exactly with either the growth rate control system or the stringent-control system. The mechanism is unknown. Interestingly, eda mutants, which have lost the function of an aldolase in the Entner-Doudoroff pathway, also accumulate a toxic diphosphorylated sugar (39, 40).

A recent report pointed out that rRNA P1 promoters contain recognition sequences for RNA polymerase bearing the heat shock sigma factor σ ³² (83). Transcription from the rRNA P1 promoter by this form of RNA polymerase holoenzyme could be demonstrated in vitro (83). The heat shock sigma factor was shown not to be required for growth rate control or for stringent regulation at non-heat shock temperatures in vivo (83).

rRNA Promoter Structure

The structures of the promoter regions for the seven rRNA operons (rrn operons) are very similar (57). Tandem promoters P1 and P2 are responsible for initiation of transcription (Fig. 3). The P1 promoter is responsible for most rRNA transcription (29, 45) and is both growth rate regulated and stringently regulated (48, 49, 108). Downstream of P2 lie antitermination sequences transcribed by RNA polymerases initiating from either promoter (8). These antitermination sequences are thought to allow efficient transcription of the long, highly structured untranslated rRNAs by facilitating interaction of Nus proteins with the elongating polymerase (111, 120). The Nus proteins may be responsible for the more rapid rate of elongation of RNA polymerase transcribing rRNA than of that transcribing mRNA (130). Upstream of P1 lie several transcriptional enhancing elements, the UP element and Fis-binding sites, which together make the P1 promoter among the strongest in E. coli (Fig. 3). The P2 promoter appears to constitutively express rRNA at a low level and may be responsible for the rRNA expressed at very low growth rates. P2 was originally thought to be resistant to the stringent response, but recent data indicate that it is sensitive (45, 59). A minimal core promoter region of P1 extending from –41 to +1 with respect to the start site of transcription is sufficient for growth rate regulation and stringent regulation; neither the P2 promoter nor the antitermination sequences nor the enhancing elements upstream of P1 are required (3, 48, 59). The core promoter includes the –35 and –10 recognition sequences for the σ ⁷⁰ form of RNA polymerase and a G+C-rich sequence between the –10 hexamer and the start site of transcription, 5'GCGCCNCC3' (see the legend to Fig. 4). The P1 –35 and –10 hexamers are nearly perfect and perfect matches, respectively, to the consensus sequences for σ ⁷⁰ RNA polymerase (in boldface type in Fig. 4), and the spacing between them is 16 bp, which is 1 bp less than the consensus 17 bp. The promoter regions of most tRNA operons, especially for the more abundant tRNAs, also have good matches to the consensus –10 and –35 hexamers (Fig. 4), although most tRNA promoters have a few more nonconsensus base pairs than the rRNA P1 hexamers. A reasonable match to the G+C-rich sequence of P1 between the –10 hexamer and the start site is present in many tRNA promoters (underlined in Fig. 4). In addition, probable UP elements and Fis-binding sites can be found upstream of most tRNA promoters (Fig. 4). Similar structural features of the rRNA P1 promoter and many tRNA promoters may reflect their coordinate regulation in response to the nutritional quality of the medium.

Fig. 3Structures of rRNA operons. Numbering of the structural features is from the rrnB operon. Each diagram is drawn to scale.

Source:

Fig. 4Comparison of rRNA and tRNA promoters. The ?35 and ?10 hexamers are in boldface type. Potential Fis-binding sites upstream of the core promoter region are underlined, and the number of mismatches to the consensus sequence, 5?GNT(C/T)(A/G)A(A/T)(T/A)(T/A)T(T/C)(G/A)ANC3? (38), is given in parentheses to the left of the sequence. The matches of each tRNA and P2 promoter region to the G+C box between the ?10 hexamer and the start site of the P1 promoters (5?GCGCCNCC3?) are underlined. The original proposed stringent-box consensus sequence [5?C(C/G)(C/G)CNCC3?; note that the 5? base of the P1 G+C box is not included here] is more degenerate (128) because it was derived from a set of promoters, including ribosomal protein promoters, that are now known not to be growth rate regulated or, for the most part, stringently regulated. The sequences were obtained from the following GenBank files: X14007 (aspV), X00786 (thrW [26]), M15153 (argU), J01713 (metT), X04017 (lysT), X00547 (serT), M10742 (tyrT), M10746 (serU), M21681 (metZ), M54893 (glyU), M28401 (metY), X02434 (proK), K01994 (argX), J01717 (thrU), K00969 (pheU), M10747 (leuX), and J01712 (leuV). The remaining sequences are from reference 63. The matches for the ?10 and ?35 hexamers to the consensus are usually convincing, but in many cases, the start point of transcription has not been experimentally determined. Similarly, in most cases, Fis binding has not yet been demonstrated.

Source:

Promoter Mutants

An extensive mutational analysis of the P1 core promoter of the rrnB operon turned up two classes of mutant promoters that no longer respond to growth rate regulation (33, 43). One class includes alterations of features known to be important for σ ⁷⁰-dependent promoter function that improve the nearly perfect match of the P1 promoter to consensus. These mutations are T–33A, which changes the –35 hexamer from TTGTCA to the consensus TTGACA, and insertions of a single base pair, which change the spacing between the –10 and –35 hexamers to the consensus 17 bp (33). The mutations greatly increase promoter activity and abolish the typical growth rate-dependent pattern of transcription (33). A mutation of the growth rate-regulated promoter of the leuV tRNA operon, which changes the –35 sequence from TTGACG to the consensus TTGACA, has the same effect (4). The second class consists of alterations of the G+G-rich sequence located between the –10 hexamer and the start site of transcription, which covers the region of strand separation prior to initiation of transcription. One mutation changes position –1 from C to T and results in complete loss of growth rate regulation; that is, expression from P1 at low growth rates is not repressed (3). Mutations of C to T at –4 and G to T at –6 partially decrease the growth rate dependency of P1 activity (33). Also, C-to-T mutations at positions –2 and –8 partially decrease the growth rate dependency of the leuV promoter (4). A 3-bp change of CGC to ATA from –5 to –7 of P1 also destroys growth rate regulation (59). Similarly, growth rate regulation of the tyrT tRNA promoter is abolished by a 4-bp substitution in its G+C box (129). Another example of this type of mutation involves the P2 rRNA promoter, which is not normally growth rate regulated. However, the P2 promoter does have a close match to the G+C box, and a single change from A to G at –6 (of the rrnB P2 promoter), which improves the match to the G+C box of the P1 promoter, is sufficient to confer growth rate regulation on the P2 promoter (145). Also, conversion of the tac promoter to match the G+C box at positions –8 through –5 does not confer growth rate regulation, most likely because the tac promoter does not match the G+C box of P1 from –4 to +1 (145). Although it is not known whether simple G+C richness is required for growth rate control, it is tempting to speculate that resistance to strand separation might be involved in the regulation, especially considering the unusual instability of open complexes at the P1 promoter (see below). Probably the most important finding from the mutational analysis of the rRNA P1 core promoter region is the absence of any evidence for a binding site of a potential growth rate regulatory protein (33, 43).

Several of the mutants that are unresponsive to growth rate regulation are also partially insensitive to stringent control. These mutants include those with multiple-substitution mutations in the G+C box of the rRNA P1 and the tyrT promoters mentioned above and a double mutant containing two changes toward consensus, T–33A and a 1-base insertion to increase the spacer to 17 bp (59, 67). The T–33A mutation or the insertion mutation alone did not affect stringent control of P1, nor did the single-base-pair substitutions in the G+C box, demonstrating that (mutated) promoters that have lost responsiveness to growth rate regulation can still be stringently controlled (59). A detailed understanding of the sequences necessary for stringent control awaits further study.

The UP Element and Fis-Binding Sites

Two elements located upstream of the minimal-growth-rate-regulated core promoter, the UP element and Fis-binding sites, are responsible for strong stimulation of promoter activity but are not required for growth rate regulation. These two types of promoter-enhancing elements activate independently of one another (96). Just upstream of the –35 hexamer from –41 to –61 lies the UP element, which binds the carboxy-terminal region of the α subunit of RNA polymerase and stimulates the activity of the core rrnB P1 promoter about 30-fold (9, 96, 100). Consistent with this, the footprint of RNA polymerase on the rrnB promoter bearing the UP element shows protection in this region, which is not seen for promoters in which the UP element has been deleted (100). Because the UP element binds to RNA polymerase itself rather than to an accessory factor, it is viewed as an extended part of the core promoter. The UP element is AT rich, but the exact DNA sequence requirements remain to be established. The UP element has not yet been demonstrated for the other rRNA P1 promoters; however, in view of their AT richness in the same region (Fig. 4) and their comparable strength in vivo, it is very likely present.

Further upstream, between positions –61 and –121 of the rrnB P1 promoter, lie three binding sites for the Fis protein, as demonstrated by footprinting in vitro and in vivo (99, 101). Fis protein bound to these sites, especially to the primary site nearest the core promoter, causes 5- to 10-fold stimulation of rRNA P1 core promoter activity (101). Fis protein enhances the initial binding of RNA polymerase to the promoter 3.5-fold under high-salt conditions (12). It may also facilitate promoter clearance after initiation of transcription (107). Fis-binding sites are also evident in most sequences upstream of the other rRNA P1 promoters and tRNA promoters (underlined in Fig. 4) and have been demonstrated by footprinting in several cases.

Fis is a relatively abundant DNA-binding protein in lag-phase or early-log-phase cells just after inoculation of stationary-phase cells into fresh medium (2, 127). It becomes less abundant in late-log-phase cells (2, 127), but its primary binding site upstream of P1 is still about 30% occupied as the cells enter stationary phase (99). In fact, it appears that after an initial burst of synthesis of Fis protein upon upshift or inoculation of stationary-phase cells into fresh medium, little if any additional Fis protein is made, and preexisting protein is simply diluted by growth of the cells (127). However, it takes 2 days in stationary phase to completely deplete Fis protein (99). Fis is also less abundant on poor medium, and this scarcity appears to cause a decrease in the activity of some growth rate-regulated promoters of tRNA operons (84, 85), but a contribution by Fis to growth rate regulation of the rRNA P1 promoter has yet not been convincingly demonstrated.

The Fis protein is not essential for growth of E. coli, but the consequences of a fis null mutation are interesting. As indicated above, Fis protein bound upsteam of the P1 promoter is responsible for about fivefold stimulation of transcription initiation, a fact demonstrated by comparing the β-galactosidase activities of a P1 promoter-lacZ fusion that lacks the upstream Fis binding sites to those of a promoter in which the Fis-binding sites are intact (101). In a fis null mutant, β-galactosidase activity from the P1 promoter with intact Fis-binding sites is the same as that in fis ⁺ cells. Paradoxically, in the fis null mutant, the activity from the P1 promoter lacking the Fis-binding sites is fivefold higher than activity in fis ⁺ cells. Thus, in the fis null cells, promoters with or without Fis-binding sites are apparently equally expressed, and the level of expression for both promoters is comparable to that of the promoter with Fis-binding sites in the fis ⁺ cells (101). A likely explanation is that in the absence of Fis protein, the feedback mechanism of growth rate control, acting through the core promoter, elevates expression from the P1 promoter, compensating for the loss of Fis-mediated transcriptional activation (101). This phenomenon is reminiscent of the derepression of rRNA synthesis to compensate for decreased rRNA gene dosage (24) or for premature termination of rRNA transcription by a nusB mutant defective in antitermination (111).

Unusual Kinetic Features of rRNA P1 Initiation

Given the strength of the rRNA P1 promoter in vivo, it is surprising that formation of a heparin-resistant complex with RNA polymerase requires addition of the first two nucleotides of the transcript (47). This requirement is not shared by other E. coli promoters except for the only other growth rate-regulated promoter that has been tested, the tyrT tRNA promoter (65). Continuing advances in understanding the mechanisms of transcriptional initiation have allowed a more detailed characterization of the properties that appear to be limited to rRNA P1 and growth rate-regulated tRNA promoters.

In kinetic studies of the major form of E. coli RNA polymerase, the holoenzyme that includes the promoter recognition subunit σ ⁷⁰, two distinct steps in the formation of stable RNA polymerase-promoter DNA complexes can be distinguished: an initial binding step that is bimolecular and reversible and that has a forward rate dependent on the concentration of polymerase and a subsequent isomerization step that is a unimolecular rearrangement of RNA polymerase on the template and that ends in strand separation around the start site of transcription (78). The intermediate complex after initial binding, called a closed complex, can normally be detected only at low temperature (so that subsequent steps are inhibited), and its footprint extends to about –5 (64). Because the initial binding step is reversible, RNA polymerase in the closed complex is in equilibrium with that in solution, and heparin, which binds to RNA polymerase in solution and prevents binding to the template, can sequester RNA polymerase off the template. Isomerization, the kinetically defined second step, has recently been subdivided into two steps distinguishable by thermodynamic analysis (98), a conformational change of polymerase that results in a longer footprint, extending to +20 (64), and strand separation, which can be detected by single-strand-specific chemical modification. The resulting preinitiation complex with the DNA strands separated around the start site is called an open complex, and upon addition of nucleotides, this complex will initiate phosphodiester bond synthesis. The open complex of most promoters is very stable even in the absence of nucleotides and is resistant to heparin, a feature that permits footprinting in order to localize the RNA polymerase-binding site at the promoter. This stability and heparin resistance are due to the almost irreversible nature of the isomerization step and the tendency for the strands to remain separated (at 37°C) at most promoters.

The rRNA P1 and tyrT promoters were first appreciated as being unusual among the σ ⁷⁰-dependent promoters because they failed to form a stable heparin-resistant complex with RNA polymerase (47, 65). However, a stable complex at the rrnB P1 promoter could be detected if the first two nucleotides of the transcript, ATP and CTP, were present (47). (This complex was recently shown to have initiated and produced, by a slippage mechanism, abortive transcripts 5 nucleotides long [5'ACCAC3'], which are not produced if all four nucleoside triphosphates are present [13].) More recently, a supercoiled template was shown to greatly increase the stability of open complexes to a half-life of about 2 min in the presence of heparin, compared to half-lives ranging from 30 min to hours for other E. coli promoters (68). A supercoiled template apparently also increases the efficiency of the initial binding step (68). The UP element causes about a 20-fold increase in the equilibrium constant for initial binding and about a 5-fold increase in the unimolecular rate of isomerization (96). The rate of RNA polymerase association to the extended P1 promoter including the core promoter and the UP element is limited only by the rate of diffusion under low-salt conditions, consistent with its strength in vivo (96), although under higher-salt conditions, the Fis protein stimulates RNA polymerase binding by an additional 3.5-fold (12). Under different conditions in which RNA polymerase is bound to a supercoiled template in a heparin-sensitive complex, the strands are mostly not separated, which is consistent with the instability of open complexes (88). Likewise, in vivo after treatment with rifampin (which traps open complexes at most promoters and results in a large increase in strand separation as detected by the single-strand-specific reagent permanganate), the permanganate sensitivity of the rRNA P1 promoter is unaltered, indicating that most P1 promoters are not in open complexes (89).

In summary, initial bimolecular binding to the extended P1 promoter is diffusion limited at low salt, whereas at higher salt, decreased affinity is apparently offset by the activator protein Fis. Therefore, the binding step appears to be nearly maximally efficient. The subsequent unimolecular isomerization step, however, is unusual in that it is readily reversible in the absence of nucleotides, as indicated by the relative instability of open complexes. Whether the complexes are mostly closed or are intermediates in which conformational change of RNA polymerase but not strand separation has occurred remains to be determined. Since instability of open complexes has been found only with growth rate-regulated promoters, i.e., the rRNA P1 and the tyrT tRNA promoters, an attractive hypothesis is that repression of transcription initiation exerted by the ribosome feedback mechanism operates by taking advantage of this feature, for example, by transiently limiting access of the initiating nucleotide.

ppGpp Is Not Necessary for Growth Rate Control

For many years, the best candidate for the effector of growth rate control was ppGpp. Observations consistent with this idea included the excellent inverse correlation between ppGpp concentration and growth rate (103), the relA-dependent mechanism by which ppGpp was synthesized in response to a shortfall of translational substrates (charged tRNA [52]), and the fact that rRNA and tRNA transcription was preferentially inhibited at the higher ppGpp levels found during the stringent response (17, 121). Recently, strains lacking detectable ppGpp were constructed by disrupting both the relA gene and the spoT gene (138), whose protein product had previously been thought to degrade only ppGpp but now, because of its sequence similarity to relA (79) and the "spotless" phenotype of the double mutants, is appreciated as a likely ppGpp synthetase as well (138). The doubly disrupted strain was multiply auxotrophic for many amino acids but was viable (138) and was tested for growth rate regulation of rRNA expression. The RNA content of the spotless strain (per unit of protein) was normal and increased linearly with increasing growth rate, as previously seen for wild-type strains (44). Feedback inhibition by plasmid-borne rRNA operons occurred in the doubly disrupted strain as well as in the wild-type (44). The rate of synthesis of β-galactosidase from a P1-lacZ fusion, normalized to total protein synthesis rate, also increased linearly with increasing growth rate as in wild type, indicating that the rate of initiation from the P1 promoter was regulated normally in the absence of ppGpp (44).

The conclusion that growth rate regulation does not require ppGpp was challenged by another group of investigators. They showed that the ratio of the rates of synthesis of stable RNA to total RNA, which is normally proportional to growth rate, was unchanged over a range of growth rates (53). This seems to contradict the conclusion that growth rate regulation of stable RNA synthesis is unaffected by the absence of ppGpp. However, the rate and the growth rate dependency of stable RNA synthesis in the double mutant are in fact identical to those for the control wild-type strain (53). The ratio (of stable RNA to total RNA) was abnormal because the mRNA synthesis rate increased with increasing growth rate, as did the stable RNA synthesis rate. The growth rate dependence of RNA content, normalized to protein content (RNA/protein ratio), increased linearly with growth rate in the spotless strain, behavior that was identical to that of the wild-type strain (53). Thus, the defect in the spotless strain is overproduction of mRNA, whereas stable RNA synthesis, including rRNA synthesis, is normal.

Other lines of evidence indicate that growth rate regulation and stringent regulation are not identical. For example, several mutant rRNA P1 promoters that are no longer responsive to growth rate regulation are still stringently regulated (59). In another example, a mutant that could be partially pyrimidine starved in a systematic way exhibited a direct correlation between ppGpp pool sizes and growth rate rather than the inverse relationship normally observed (132). Finally, a strain in which four rRNA operons were deleted, resulting in 2.3-fold derepression of rRNA promoters to compensate for the decreased rRNA gene dosage, showed no change in ppGpp pools (24). The evidence weighs heavily against ppGpp being the sole mechanism for growth rate regulation. Recent reports suggest that in addition to previously defined functions, an important role of ppGpp may be modulation of the chain elongation rate of RNA polymerase (117, 130, 131).

Passive Model

A model proposing "passive" regulation of rRNA synthesis was introduced some years ago (see p. 375–385 of reference 54) and was reintroduced recently with more elaboration (55). The central concept of passive models is that the rRNA P1 promoters are assumed to compete poorly with the other (mRNA) promoters in the cell for RNA polymerase. In order to achieve the highest rate of initiation from the rRNA P1 promoters, the intracellular concentration of free RNA polymerase (that which is available for binding to promoters) would have to be considerably greater under rich nutritional conditions than under poor nutritional conditions. This increase of free (available) RNA polymerase was originally proposed to result from repression of biosynthetic operons (or a failure to activate them) in rich medium. However, the overall rate of mRNA synthesis per unit cell mass does not appear to change significantly between poor medium and rich medium (14). In the most recent version of the model, the concentration of free RNA polymerase is proposed to be transiently modulated by changes in the elongation rate of RNA polymerase: a decrease in the RNA chain elongation rate would sequester polymerase on the template DNA, whereas an increase in the chain elongation rate would more rapidly release polymerase after completion of transcripts, thus transiently swelling the pool of free polymerase (55). However, it is not known how the free RNA polymerase concentration would be modulated during steady-state growth.

Regarding competition among promoters for RNA polymerase, the ability of a promoter to compete depends not only on its relative affinity for RNA polymerase but also on its availability for binding polymerase. A promoter with rapid rates of isomerization, strand separation, and promoter clearance (taken together, promoter turnover) will be maximally available for binding RNA polymerase. Conversely, a promoter with slow turnover will be mostly blocked by bound polymerase in the process of initiation and will be on average less available for binding RNA polymerase.

Three sets of experimental observations argue against any type of passive model. First, initial binding of RNA polymerase to the P1 promoters is highly efficient in vitro (68, 88, 96). Early evidence to the contrary was obtained in promoter competition experiments conducted in the presence of heparin or rifampin, both of which exacerbate the instability of open complexes at P1 but not at other promoters, and resulted in an erroneously low value for the binding constant (60). The calculated initiation frequency or turnover time for an rRNA P1 promoter in vivo during rapid growth is about 1/s (14), which is nearly as rapid as is possible for the polymerase to clear the promoter, assuming an elongation rate of 60 nucleotides per s. Thus, the P1 promoter is maximally available for RNA polymerase binding and has a high affinity for polymerase, indicating that it should be a strong competitor in vivo. Second, limitation of the total RNA polymerase concentration in vivo sufficiently to decrease mRNA synthesis twofold had no effect on the rate of synthesis of rRNA (86). This result indicates that rRNA promoters are successful competitors for RNA polymerase in vivo. Overexpression of RNA polymerase, resulting in twofold-increased mRNA synthesis, also had no effect on the rRNA synthesis rate (86). Thus, perturbation of the intracellular concentration of RNA polymerase affects mRNA synthesis but not rRNA synthesis, results opposite to those predicted by passive models. Third, an additional observation that is inconsistent with a passive model comes from the rRNA gene dosage experiments mentioned above. In these experiments, feedback repression was observed when the number of rRNA operons was increased twofold by the use of a plasmid carrying an rRNA operon (56). However, this feedback regulation was not observed when the plasmid-borne rRNA operon was internally deleted and the total rRNA synthesis rate (including defective rRNA synthesis) increased twofold in proportion to the increased gene dosage (56). This result demonstrates the presence of "extra" RNA polymerase capable of initiating at the rRNA promoters. These observations contradict the fundamental postulate of the passive model, that regulation of rRNA synthesis depends on the poor affinity of rRNA promoters, compared to other promoters in the cell, for RNA polymerase.

SUMMARY

It has long been appreciated that the ribosome content of E. coli increases linearly with increasing growth rate (109) and that the rate of ribosome synthesis therefore increases approximately with the square of the growth rate (46), a phenomenon termed growth rate regulation. rRNA synthesis is directly regulated, whereas translation of r-proteins is regulated by systems that monitor the availability of rRNA for ribosome assembly. Regulation of rRNA synthesis can compensate for increased or decreased gene dosage of rRNA operons (24, 56) or for genetic loss of the activator protein Fis, indicating a feedback mechanism. Ribosomes not engaged in translation do not contribute to feedback inhibition, indicating that translating ribosomes are involved in the feedback loop (23, 139). This suggests that cells monitor whether the nutritional quality of the growth medium is adequate to supply the current intracellular level of ribosomes with substrates (or energy) for translation. According to this view, if there is a shortfall of substrates (or energy), synthesis of ribosomes is decreased and synthesis of biosynthetic enzymes and/or nutrient uptake systems is increased. Conversely, if the supply of substrates (or energy) for translation is abundant, the synthesis of ribosomes is increased, so that the growth rate can be maximized in a particular medium. Thus, the translational apparatus is proposed to serve as a nutritional sensor for incremental regulation of ribosome biosynthesis.

The targets of growth rate regulation are the P1 promoters of the rRNA operons and many tRNA promoters. Although upstream elements contribute much of the strength of the P1 promoters, the core promoter region, from –41 to +1, is responsible for regulation (3, 48). Mutational analysis of this region failed to provide any evidence for a DNA-binding site of a regulatory protein (33, 43), suggesting that the growth rate regulation factor may interact directly with RNA polymerase when it is in a complex with the P1 promoter. Mutations that conferred growth rate regulation independence either improved the –35 sequence to consensus, increased the spacing between the –35 and –10 hexamers to the optimal 17 bp, or altered the G+C sequence between the –10 hexamer and the start site of transcription (33, 59). The rRNA promoters are probably the strongest in E. coli, and kinetic measurements of RNA polymerase initiation in vitro (in the absence of other factors) are consistent with their strength in vivo (68, 96). Therefore, the feedback mechanism for growth rate regulation is probably repression. Despite their strength, the rRNA P1 and the tyrT tRNA promoters are distinct from other σ ⁷⁰-dependent promoters in E. coli in that open complexes of RNA polymerase are very unstable (65, 68). It is tempting to speculate that this instability is due to the G+C richness of the region in which the DNA strands are separated in the open complex and that the unusual instability is involved in the mechanism of growth rate regulation. For example, repression of initiation at P1 could be achieved by transient occlusion of nucleotides from the polymerase-promoter open complex, allowing it to decay without a productive initiation event.

The identity of the signal generated by translating ribosomes to regulate initiation of the rRNA P1 and many tRNA promoters during steady-state growth is unknown. ppGpp may contribute to growth rate regulation, but it is not essential (44). The identity of the factor that might interact with RNA polymerase at these promoters to control initiation is also unknown, although a mutant σ ⁷⁰ that may fail to interact with that factor has been described (61). Elucidation of the mechanism for growth rate-regulated control of ribosome biosynthesis remains an important challenge in bacterial physiology.

ACKNOWLEDGMENTS

We thank C. Josaitis and R. Gourse for their helpful comments.