Characterization of RNA and DNA synthesis in Escherichia coli strains devoid of ppGpp.

The synthesis rates of DNA, rRNA, bulk mRNA, protein, and RNA polymerase beta- and beta'-subunits were determined as functions of growth rate in a wild-type Escherichia coli strain, which produces guanosine tetraphosphate (ppGpp), and in a delta relA delta spoT mutant which does not produce ppGpp. The rate of stable RNA synthesis per amount of protein depends on three factors: RNA polymerase concentration, RNA polymerase activity, and the distribution of active RNA polymerase between stable and mRNA genes, measured as the stable RNA synthesis rate/total RNA synthesis rate, rs/rt. In the wild-type strain, all three factors increase with growth rate. In the ppGpp-deficient strains, only RNA polymerase synthesis and activity, but not rs/rt, increased with growth rate. Thus, adjustments of rs/rt require ppGpp. In the absence of ppGpp, the synthesis of rRNA and bulk mRNA both varied in direct proportion to the concentration of active RNA polymerase, in contrast to the wild-type strain, in which only rRNA synthesis increased with growth rate, while mRNA synthesis remained constant. Thus, a control specific for rRNA is absent in strains lacking ppGpp. In rich media, the ppGpp-deficient strain synthesized up to 4-fold more mRNA than wild-type bacteria, which was associated with a similarly increased RNA polymerase activity. We propose that RNA polymerase is rendered inactive in wild-type bacteria due to ppGpp-dependent transcriptional pausing during the synthesis of mRNA. Finally, the control of replication initiation was altered in ppGpp-less bacteria, apparently reflecting indirect changes in the cell physiology, rather than a direct effect of ppGpp on replication initiation.

determined as functions of growth rate in a wild-type Escherichia coli strain, which produces guanosine tetraphosphate (ppGpp), and in a ArelA AspoT mutant which does not produce ppGpp. The rate of stable RNA synthesis per amount of protein depends on three factors: RNA polymerase concentration, RNA polymerase activity, and the distribution of active RNA polymerase between stable and mRNA genes, measured as the stable RNA synthesis rateltotal RNA synthesis rate, r.1 rt. In the wild-type strain, all three factors increase with growth rate. In the ppGpp-deficient strains, only RNA polymerase synthesis and activity, but not rJrt, increased with growth rate. Thus, adjustments of re/rt require ppGpp. In the absence of ppGpp, the synthesis of rRNA and bulk mRNA both varied in direct proportion to the concentration of active RNA polymerase, in contrast to the wild-type strain, in which only rRNA synthesis increased with growth rate, while mRNA synthesis remained constant. Thus, a control specific for rRNA is absent in strains lacking ppGpp. In rich media, the ppGpp-deficient strain synthesized up to 4fold more mRNA than wild-type bacteria, which was associated with a similarly increased RNA polymerase activity. We propose that RNA polymerase is rendered inactive in wild-type bacteria due to ppGpp-dependent transcriptional pausing during the synthesis of mRNA. Finally, the control of replication initiation was altered in ppGpp-less bacteria, apparently reflecting indirect changes in the cell physiology, rather than a direct effect of ppGpp on replication initiation.
With regard to the synthesis of ribosomal RNA (rRNA) in bacteria, "stringent control" has been distinguished from "growth rate control." The former refers to the reduction in rRNA synthesis in response to amino acid starvation (Stent and Brenner, 1961), the latter refers to adjustments of rRNA synthesis in response to changes in the nutritional quality of the growth medium (Schaechter et al., 1958). Both kinds of control are associated with changes in the cytoplasmic concentration of guanosine tetraphosphate (ppGpp)' in a manner which has suggested that ppGpp is a negative effector in the control of rRNA synthesis (Lazzarini et al., 1971;Ryals et al., 1982aRyals et al., , 1982bRyals et al., , 1982c; . Stringent control is absent in bacteria with a mutational defect in the relA gene (Stent and Brenner, 1961), which encodes a ribosome-associated guanosine tetraphosphate synthetase (PSI, for (p)ppGpp synthetase I). PSI is present on about 1% of ribosomes and is activated by uncharged tRNA during amino acid starvation (Haseltine et al., 1972;Haseltine and Block, 1973). Overproduction of PSI from a cloned relA gene, so that a large fraction of ribosomes bears PSI (RelA), causes a rapid accumulation of ppGpp and subsequent inhibition of rRNA synthesis in the absence of amino acid starvation (Schreiber et al., 1991;Tedin and Bremer, 1992). These observations are consistent with the proposal that ppGpp is a negative regulator of rRNA synthesis. Furthermore, in vitro studies indicate that ppGpp preferentially inhibits transcription initiation from rRNA promoters (van Ooyen et al., 1975(van Ooyen et al., , 1976Travers, 1976;Kajitani and Ishihama, 1984;Ohlsen and Gralla, 1992), and enhances RNA polymerase pausing at specific sites during the elongation of RNA chains . Whereas these studies have established ppGpp as an inhibitor of rRNA synthesis at high concentrations, the significance of lower ppGpp levels acting as a regulator for the growth rate control of rRNA synthesis has remained controversial (Gaal and Gourse, 1990; and reviews by Lindahl and Zengel, 1986;Cashel and Rudd, 1987;Jinks-Robertson and Nomura, 1987;Bremer and Dennis, 1987).
Generally the control of rRNA synthesis has been considered together with the control of (bulk) tRNA synthesis since both classes of RNA are essentially coregulated (Dennis, 1972;Shen and Bremer, 1977a). Thus, for the following considerations, we divide total bacterial RNA into stable RNA (rRNA + tRNA; synthesis rate rs) and mRNA fractions (synthesis rate rm); the sum of stable RNA and mRNA synthesis rates is the total or "instantaneous" rate of RNA synthesis ( rt = r, + r,,,). In previous work from different laboratories, stable RNA synthesis rates have been measured per cell (rJcell), per genome equivalent of DNA (r,/genome), per rrn gene ( rs/ gene), per amount of protein (rJP), per amount of ribosomal RNA (rs/R), or per total RNA synthesis rate (rJrt). For most genetic systems it is not crucial which reference unit is used, since changes in the synthesis rate of a specific enzyme do not significantly alter the macromolecular composition of the cell. However, when the synthesis of rRNA changes, the reference unit may also be altered, so that an increase in the rate with one unit may show no effect or even a decrease with another (Baracchini and Bremer, 1991). This may be one of the reasons why contradictory conclusions about the growth rate control of rRNA synthesis, and about the role of ppGpp in this control, have appeared in the literature.
The rate of stable RNA synthesis (e.g. rJP) is the product 10851 of several factors which include the RNA polymerase concentration and activity, the RNA chain elongation rate, and the distribution of RNA polymerase over stable RNA and mRNA promoters. Each of these factors is subject to its own control which may or may not involve ppGpp. We assume changes in the RNA chain elongation rates, or in the RNA polymerase concentration and activity, affect the rates of stable and mRNA synthesis alike, and are therefore not specific for stable RNA (Ryals et al., 1982a(Ryals et al., , 1982b. The fraction of the total RNA synthesis rate that is stable RNA (rs/rt) is a parameter which we have singled out as more suited than others to characterize the control of stable RNA synthesis since it represents the specific control of stable RNA synthesis relative to all other transcriptional activities. Changes in rs/ rt were found to be strictly correlated with changes in the cytoplasmic level of ppGpp (Ryals et al., 1982b;Hernandez and Bremer, 1990;Tedin and Bremer, 1992). For example, when a culture is subjected to a temperature upshift, there is a substantial accumulation of ppGpp, but the rate of rRNA accumulation is not reduced therefore, it was originally concluded that ppGpp is not essential for the control of rRNA synthesis (Gallant et al., 1977;Gallant, 1979). However, subsequent measurements showed that rs/rt was indeed reduced under these conditions, but this reduction in rs/rt was compensated, and thus obscured, by an increased RNA chain elongation rate due to the higher temperature .
During exponential growth, basal levels of ppGpp in Escherichia coli are synthesized by a second guanosine tetraphosphate synthetase (PSII, for (p)ppGpp synthetase 11; i.e. the spoT gene product (Hernandez and Bremer, 1991;Xiao et al., 1991)). PSII is apparently inactivated, rather than activated (as is PSI), during amino acid starvation. It has been proposed that the activity of PSII responds to the balance between the supply and consumption of amino acids (Hernandez and Bremer, 1990). In this way the concentration of active ribosomes could control PSII activity by affecting amino acid consumption. This would provide a feedback that adjusts ppGpp levels, which in turn adjust ribosome synthesis, until supply and consumption of amino acids are in equilibrium (Little et al., 1983;Hernandez and Bremer, 1990).
An alternative model of ppGpp-dependent control of rRNA synthesis has been proposed by Jensen and Pedersen (1990). Their model is based on the assumptions that mRNA promoters, but not rRNA promoters, are normally saturated with RNA polymerase and that ppGpp reduces the fraction of free RNA polymerase by slowing the elongation of RNA chains. These workers have proposed that high concentrations of ppGpp (which accumulate during slow bacterial growth) reduce free RNA polymerase by sequestering it in the elongation complex and that this preferentially reduces the activity of unsaturated rRNA and tRNA promoters. Thus, this model assumes that changes in RNA chain elongation or in RNA polymerase concentration affect the rates of stable and mRNA synthesis differently. In contrast, the results below indicate that (bulk) mRNA promoters are not saturated with RNA polymerase and that, in the absence of ppGpp, changes in RNA polymerase concentration equally affect stable and mRNA synthesis.
A third model of rRNA synthesis control, known as the ribosome feedback regulation model, assumes a repressor role of free or translating ribosomes, either directly or via their activity, without involving a control of PSII activity and ppGpp synthesis (Jinks-Robertson and Nomura, 1987;Cole et al., 1987). This model was originally derived from observations with bacteria carrying extra copies of intact or defec-tive rrn genes on plasmids (Jinks-Robertson et al., 1983). No evidence has been reported yet which identifies a ribosomerelated repressor or effector for this control, or a repressor binding site, i.e. rrn operator (Jinks-Robertson and Nomura, 1987). Furthermore, the interpretation of the original observations obtained with artificially increased rrn gene dosage has been put into serious question (Baracchini and Bremer, 1991).
E. coli strains have been constructed in which the genes required for both PSI and PSII activity, relA and SPOT, respectively, are deleted (Xiao et al., 1991). Since these strains are devoid of ppGpp (Xiao et al., 1991;Gaal and Gourse, 1990),' they provide a further opportunity to assess the role of ppGpp. A previous study of such ppGpp-less strains has suggested that ppGpp is not required for the growth rate control of rRNA synthesis (Gaal and Gourse, 1990). The study of ppGpp-less strains presented here confirms that, for a given growth rate, the stable RNA synthesis rate ( r J P ) is the same in wild-type and ppGpp-deficient bacteria. However, it is shown that the control of stable RNA synthesis (rs) cannot be expected to change rs/P at a given growth rate unless it changes average ribosome activity, and a t high growth rates ribosome activity is unaffected by the absence of ppGpp. On the other hand, we found that growth rate-dependent adjustments of rJrt do require ppGpp. In addition, we report several new in vivo effects of ppGpp that occur only at very low levels of ppGpp. Most striking among them is an inhibition of bulk mRNA synthesis that indirectly affects r, and might be caused by ppGpp-dependent transcriptional pausing.
Finally, it has been suggested that ppGpp plays an important role in the control of initiation of chromosome replication and its adjustment to the rate of bacterial growth (Zyskind and Smith, 1992). Our results with ppGpp-less bacteria indicate that ppGpp may play only an ancillary role in this regulation.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Growth Conditions-Bacterial strains used and constructed here are listed in Table I. The strains CF1648 ("21655, prototype E. coli K-12 wild-type strain; B. Bachman), CF1693 (Arekt251 AspoT207), and CF1678 (Arekt251 AspoT209) were obtained from M. Cashel (National Institutes of Health). Different growth rates were obtained (see below) using Luria-Bertani (LB) medium with 0.2% glucose or minimal Medium C (Helmstetter, 1967) with either 0.2% glucose (Glc), glycerol, or succinate (Suc) and supplemented with the nucleic acid bases adenine, xanthine, cytosine, and uracil at 50 pg/ml each; 20 amino acids (aa) at 50 pg/ml each; and serine (Ser) at 500 pg/ml. The average growth rates in doublings/ h (? 5% variation) achieved in various media with the strain CF1648 (wild-type) were: p = 2.85 in LB Glc; p = 2.31 in Glc aaSer + adenine, cytosine, xanthine, uracil; p = 1.95 in Glc aaSer; p = 1.71 in Glc aa; p = 1.58 in glycerol aaSer; p = 1.32 in glycerol aa; p = 1.25 in Suc aa; p = 1.00 in Glc; and p = 0.67 in glycerol. The ppGpp-less strains CF1693, CF1678, VH2734, and VH2375 (Table I) were always grown in the presence of 20 amino acids. The strain VH2733 (Table I) when grown in minimal medium was also supplemented with arginine at 50 pg/ml. Overnight cultures were diluted 1000-2000-fold into fresh media for physiological experiments. All strains were grown at 37 "C in Erlenmeyer flasks with an inital flask-to-culture volume ratio of 5 and aeration provided by rotation at 175 revolutions/min. Growth of cultures was determined turbidometrically at 460 nm (1-cm light path), and samples were taken at an optical density (OD,,) between 0.5 and 1.0.
RNA, DNA, ar.d Protein Determinations-RNA, DNA, and protein determinations of bacterial cultures were performed as previously described (Brunschede et al., 1977) with changes specified below. For the acid precipitation of cells from cultures grown in LB medium, cells were initially collected by filtration on glass fiber filters (Whatman GF/F), washed three times with 5 ml of Medium C (see above), V. J. Hernandez and H. Bremer, unpublished observations. with cat.
e The presence of the AspoT207::cat allele was confirmed by amplification of the chromosomal spo region by the Taq DNA polymerase chain reaction method and checking for the correct amplification and restriction fragment length polymorphisms by agarose gel electrophoresis.
dThe strain VH2733 was transduced to cat+ with a P1 phage lysate grown on CF1678. The presence of the AspoT209::cat allele was confirmed by polymerase chain reaction as described above (footnote c ) .
then treated with 5 ml of 0.5 M perchloric acid, which was allowed to drip through with no filtration pressure. This was done to wash away ribo-and deoxyribonuceotide oligomers which are present in LB medium and prevent their inclusion upon perchloric acid precipitation. Cultures in other media were likewise treated for consistency. Duplicate 5-ml samples were assayed for RNA/protein ratio determination. Following alkali hydrolysis, perchloric acid-soluble RNA nucleotides were quantitated spectrophotometrically at 260 nm assuming one A2@ unit at pH 2 corresponds to 93 nmol of ribonucleotides. Protein concentrations were measured by the Lowry method (Lowry et al., 1951) relative to bovine serum albumin and assuming the average molecular weight of an E. coli amino acid residue to be 108 daltons. The amount of DNA in a culture was determined by a modified diphenylamine colorimetric assay (Burton, 1956;Shay et al., 1990). Duplicate 2.5-ml culture samples were collected on glass fiber filters, as above, each was then hydrolyzed in 0.5 ml of 1.6 M perchloric acid at 67 "C for 30 min. After cooling to room temperature, 1 ml of diphenylamine reagent was added (0.4 g of diphenylamine, reagent ACS, Kodak, in 20 ml of glacial acetic acid, 0.3 ml of concentrated &SO,, and 0.1 ml of 32 pg/ml solution of acetylaldehyde), and the mixture was incubated for 16 h at 30 "C. Solutions were then clarified by filtration through glass fiber filters (Whatman GF/F), and DNA was quantitated spectrophotometrically at 600 nm (1-cm light path). A DNA standard calibration curve was obtained using purified X DNA. DNA concentrations were calculated in genome equivalents/1O8 amino acid residues in protein assuming that one AZM) unit of A DNA at pH 12 corresponds to 28.2 pg (618 g/mol per base pair; GC content of 0.5) and that 1 pg of E. coli DNA corresponds to 2.49 X 10' genome equivalents (3900 kb/genome). When using rifampicin to halt initiation of DNA replication, it was added to a final concentration of 200 pglml.
The nonradioactive methods used to measure RNA and DNA accumulation are not subject to effects of specific radioactivity changes, precursor pool expansion, hybridization efficiency, etc. Systematic errors of the data were previously estimated to be less than 10% (Brunschede et al., 1977).
Determination of the Relative Rate of Stable RNA Synthesis-The instantaneous rate of stable RNA synthesis as a fraction of the total RNA synthesis rate (rs/rt) was determined by hybridizing total 3Hpulse-labeled RNA to a denatured double-stranded rDNA probe (X dilv5) as previously described . Hybridization mixtures included "C-labeled rRNA for determination of the hybridization efficiencies which varied between 65 and 93%.
Determinations of 6-Galactosidase Specific Activity-6-Galactosidase specific activities were determined in triplicate, as previously described (Miller, 1972;Hernandez and Bremer, 1990). Specific activities were expressed as the rate of change in absorbance at 420 nm (1-cm light path) per min of assay time per A4M) unit of culture. In addition, specific activities were determined per amount of total protein.
Determination of RNA Polymerase Concentrations and Activity-RNA polymerase concentrations were determined by SDS-polyacrylamide gel electrophoresis as previously described (Shepherd et al., 1980), with some modifications. Briefly, 10-ml culture samples were taken at an of approximately 0.8 and chilled on ice for 15 min. Cells were harvested by centrifugation at 12,000 X g for 10 min at 4 "C. Cell pellets were then resuspended in 0.1 ml of ice-chilled 2 X protein sample buffer (0.25 M Tris-HC1, pH 6.8, 2.85 M B-mercaptoethanol, 10% glycerol, 0.01% bromphenol blue). After complete resuspension, 0.1 ml of boiling 10% SDS was added and the mixtures were boiled for 5 min.
The Laemmli discontinuous gel electrophoresis system (Laemmli, 1970) was used to resolve proteins. A 27-cm-length 5% polyacrylamide gel was used to separate 6and 6"subunits of RNA polymerase which were visualized by staining with Coomassie Brilliant Blue R250. A 50-pl volume of each sample lysate was loaded and electrophoresed at 250 v constant current for 7 h. A second 50-pl application of each sample was loaded in unoccupied lanes, and electrophoresis was carried out for an additional 2 h. The percent of protein in a given 6"subunits band relative to the total amount of protein loaded (2 h of electrophoresis) was determined by relative densities quantitated with a laser microdensitometer (LKB Ultroscan XL). The absolute concentration of total proteins in the lysate was determined by the Lowry method (Lowry et al., 1951). The area of the 6' peak was converted to micrograms of 6' protein/lysate volume by multiplying the ratio of densities of the P'-band to that of total protein by the actual amount of total protein. The amount of p'-subunit was then converted to number of RNA polymerase molecules assuming molecular masses of 155 and 379 kDa for the P'-subunit and core RNA polymerase, respectively. The reproducibility of this method, using different cultures started on different days from different overnight cultures, including variations in lysis, gel scanning, etc., was found to be 6% (standard deviation; see Table I1 of Shepherd et al., 1980). The values shown in Fig. 7 are the averages from two to three cultures.
Formulas and Calculations-Most of the following formulas are derived from the exponential growth equation N = No X 2'Ir, where N may represent any extensive property of the culture and T is the culture doubling time. The derivations employed simple mathematical manipulations such as substitutions, expansion, differentiation, resolution for a different parameter, etc. These formulas follow from an appropriate definition of the parameters used (Bremer, 1982) and are thus based on a single assumption for their application: that the cultures used are in steady-state exponential growth. Since the time constant for changes in the macromolecular composition due to changes in growth conditions is the generation time of the bacteria, it is necessary that cultures have grown for at least four to five generations under steady-state conditions before their properties can be described by the formulas below. Routinely, fresh overnight cultures were diluted at least lOOO-fold, so that they had grown at least 100-fold before any samples were taken. References for the formulas used are given (reviewed by Bremer and Dennis, 1987).
The rRNA synthesis rate per amount of protein was calculated as the number of rRNA transcripts/min/amino acid residue in total cell protein from the amount of RNA per protein (RIP, Fig. 2, A and B) using the formula (see "Discussion"), (drRNA/dt)/P = 0.84(R/P)/4566 X In 2/7 (Eq. 1) where RIP = RNA nucleotides/amino acid residue; 0.84 = fraction of total RNA which is rRNA; 4566 = RNA nucleotides/ribosome; and 7 = culture doubling time in minutes.
TO calculate r,/P (rate of stable RNA synthesis per amount of protein) in nucleotides/min per amino acid residue (aa), we used the formula, where the factor 1.2 corrects for the degradation of unstable spacers in the stable RNA precursors, assumed to correspond to 20% of precursor transcript lengths, and 0.98 is the fraction of total RNA that is rRNA and tRNA. The mRNA synthesis rate/protein (r,,,/P) was calculated as mRNA nucleotides/min/aa residue in total cell protein from r8/P and r8/rt as follows.
The number of active RNA polymerases (N,) was determined by calculating the number of growing RNA chainslkg protein from the ratio RNA/protein (RIP) and from rr/rt, using the following formula (Shepherd et al., 1980), where 0.98 = fraction of total RNA that is stable RNA 1.2 = correction for unstable rRNA spacers; 6 xlOZ3 = Avogadro's number; 108 = average molecular weight of an amino acid residue in E. coli protein; lo6 = pg/g; 60 = s/min; 85 = stable RNA chain elongation rate in nucleotides/s (Molin, 1976;Shen and Bremer, 1977;Ryals et al., 1982a); 50 = mRNA chain elongation rate in nucleotides/s (Bremer and Yuan, 1968). The rRNA chain elongation rate value of 85 nucleotides/s has been questioned (Gotta et al., 1991). Using electron microscopic visualization of active rrn genes at different times after rifampicin treatment, these authors conclude that the rRNA chain elongation rate is only 42 nucleotides/s at a growth rate of 2.4 doublings/h. However in the same report, the average distance between RNA polymerase molecules on the rDNA template was observed to be 83 bp. At a growth rate of 2.4 doublings/h, rRNA chains are initiated at a rate of about one chainlslrrn gene (Bremer and Dennis (1987) ; Fig. 1OA below). This initiation rate, together with the distance of RNA polymerase molecules on the DNA, implies an rRNA chain elongation rate of 83 nucleotides/s, close to the value of 85 used here. If both the value of 42 nucleotides/s and the distance of 83 bp were correct, then the rate of rRNA chain initiation would be one chain every 2 s, in contradiction to all reported initiation rates (reviewed by Bremer and Dennis, 1987), as well as to the independently determined initiation rate obtained in this work ( Fig. 1OA; Equation 8, below).
The RNA polymerase activity (fraction pp) was calculated by dividing the number of active RNA polymerase molecules/pg protein (N,/P), calculated as above (Equation 4), by the observed number of total RNA polymerase molecules/gg protein.
The chromosome replication time (C-period) was determined by standard methods for comparison with previously published data, as described (Pritchard and Zaritsky, 1970;Churchward and Bremer, 1977) from the ratio (AG) of the amount of DNA which accumulates after inhibition of initiation of rounds of DNA replication and the amount of DNA prior to inhibition. The time required for completion of a single round of DNA replication (C in min) was calculated relative to the culture doubling time (7) from AG (oriC/genome) using the following formula (Pritchard and Zaritsky, 1970;Bremer et al., 1979).
The amount of DNA present after completion of ongoing rounds of replication, measured in genome equivalents, gives directly the number of functional replication origins present at the time of inhibition of initiation, as has been confirmed by flow cytometry (Skarstad et al., 1986). In contrast, hybridization methods with an oriC probe give only relative, but not absolute, values because of uncertainties in the hybridization efficiency, specific radioactivity, efficiency of counting radioactivity, recovery of oriC sequences, etc. The average replication velocity (DNA nucleotides/s per replication fork) was obtained by dividing the number of base pairs/genome (3900 kbp) by 2C (two replication forks; C expressed in s).
The initiation mass (Po) in amino acid residuesloric was calculated from AG (oriC/genome) and the DNA concentration (G/P = genomes/amino acid residue) as follows.
The average number of rRNA genes/genome ( N J G ) in actively replicating chromosomes at a particular culture doubling time ( 7 ) was calculated as follows, where m' = map location of the r m genes (A, 86.5; B, 89.7; C, 84.5; D, 72.1; E, 90.4; G; 56.5, and H, 5.1) relative to oriC (see Bremer and Dennis (1987) for conversion of map location into m' values).
The rrn gene activity (i,,"; initiation frequency) was then calculated as transcripts/min per rrn gene from rJP, DNA concentration (G/ P), and N J G using the following formula: i , The mRNA gene activities, defined as transcription initiations/min per genome (imRNA/G), were calculated from r,/P and the DNA concentration as follows, assuming that mRNA genes are equally distributed throughout the chromosome and an average mRNA transcript length of 1000 nucleotides. The calculations assume a chromosome branching pattern, which, although different for every individual cell, is accurately described for the whole cell population as long as replication forks move at constant speeds (Bremer et al., 1979). For chromosome replication from oriC, a constant replication velocity has been shown by marker frequency measurements under various growth conditions (Bird et al., 1972).

Growth of ppGpp-less Strains
E. coli strains in which both relA and SPOT genes have been deleted, are devoid of ppGpp (see Introduction). Two such ArelA AspoT strains, CF1693 and CF1678, bearing the AspoT207 and AspoT209 alleles, respectively, differ only in the extent of the deletion within the spo operon (Table I; Xiao et al., 1991). Strains bearing these sets of deletions will be here designated AAspoT207 and AAspoT209, respectively.
These ppGpp-less strains have multiple amino acid auxotrophies (Xiao et al., 1991) and require a full complement of amino acids for unrestricted growth. Growth media containing all 20 amino acids were used and supplemented with different carbon sources and nucleotide bases to achieve growth rates between 1.2 and 2.9 doublings/h for the wild-type control strain. In a given medium, the ppGpp-less strains grew 10-30% more slowly than the isogenic wild-type; the poorer the medium, the greater the growth rate reduction ( Fig. 1 and Table 11). This indicates that, whereas ppGpp is not essential for growth, at least not in the presence of amino acids, it is required for optimal growth. The strain AAspoT209 grew more slowly than AAspoT207 in all growth media, presumably due to the deletion of additional genes in the spa operon adjacent to SPOT; this includes the gene for the w protein which copurifies with RNA polymerase but whose function is unknown (Table I; Gentry and Burgess, 1989).
Amounts and Synthesis Rates of Stable RNA and mRNA Ribosome Concentration and Activity-The amount of RNA per amount of protein (RNAIprotein, R I P ) is a measure for the ribosome concentration; RIP is expected to increase in proportion to the growth rate if the ribosome activity (protein synthesis rate/average ribosome = rate of protein synthesis, divided by number of ribosomes) remains constant (see "Discussion"). At a given growth rate (achieved in different media for ppGpp-less and wild-type strains), RIP was nearly iden-  a LB Glc = Luria-Bertani medium with 0.2% glucose; Glu aaSer = Medium C with 0.2% glucose, 19 amino acids (50 pg/ml each), serine (500 pg/ml); adenine, xanthine, cytosine, uracil (50 pg/ml each); Gly aa = Medium C with 0.2% glycerol, 20 amino acids (50 pg/ml each).
* Ribonucleotide residues in total RNA/amino acid residues in total protein.
Ribosome activity in amino acid residues/s per average ribosome (see Equation 16 under "Discussion").
ND, not determined (radioactive labeling in LB medium is too inefficient). e Value of 0.90 was calculated based on ppGpp concentration extrapolated from the culture doubling time and the ppGpp concentration ( K ) at which RNA polymerase is equipartitioned between ppGpp bound and unbound forms, K = 5 pmol/ODm unit of culture, determined from previous data for E. coli K-12 (Baracchini et al., 1988;Hernandez and Bremer, 1990).
tical in the ppGpp-less and wild-type strains, increasing in proportion to the growth rate (Fig. 2, A and B) as previously reported (Gaal and Gourse, 1990). However, in a given medium (producing different growth rates for ppGpp-less and wild-type strains), R I P was reduced in the ppGpp-less strains at all except the lowest growth rates (Table 11).
FIG. 2. RNA/Protein and ribosome activity at different growth rates. The same wild-type, AAspoT207, and AAspoT209 strains as in Fig. 1 were assayed for RNA and protein, see "Experimental Procedures." Cultures of the wild-type strain were grown in the same media listed in Fig. 1, and in addition: Glc aa ( p = 1.71); glycerol aaSer ( p = 1.58); Glc ( p = 1.00); and glycerol ( p = 0.67), p = doublings/h. The ppGpp-less strains were grown in the same media with the exception of those not supplemented with amino acids. The ribosome activity was calculated from the RNA/protein ratio (see "Discussion").  T 2 0 7 ) . The values were calculated from the RNA/protein ratios in Fig. 2 as explained in the text ("Discussion").

l s ) and ppGpp-less bacteria ( o p e n s y m b o l s ; A A s p o T = A A s p o
Ribosome activities in amino acid residues/s/ribosome were calculated from RIP at different growth rates. The absence of ppGpp reduced the ribosome activity up to 30% at growth rates below 1.5 doublingslh (Table 11) and accounted for the 20-30% reduction in growth rates in the poorer growth media (Fig. 1). At higher growth rates ribosome activity became constant, equal to about 15 amino acid residuesls per average ribosome, and was independent of ppGpp ( Fig. 2, C and D, and Table 11).
Synthesis Rate of rRNA per Amount of Protein-The rate of rRNA synthesis per amount of protein was calculated from RIP (Equation 2 in "Experimental Procedures" ; Fig. 2, A and   B). It increased in a parabolic fashion with growth rate with no prominent differences between the wild-type and AAspoT207 strains (Fig. 3), as was to be expected (see "Discussion").
Rate of Stable RNA Synthesis per Total Instantaneous RNA Synthesis Rate-The rate of stable rRNA and tRNA synthesis relative to the total instantaneous rate of RNA synthesis (r,/ rt) was determined by hybridization of pulse-labeled RNA to an rDNA probe. This ratio measures the distribution of transcriptional activities between stable RNA and mRNA genes. In the wild-type strain r8/rt increased continuously from 0.4 to 0.9 as the growth rate increased from 0.6 to 1.8 doublingsl h (Fig. 4, filled circles), in agreement with previous values obtained in E. coli B/r and K-12 strains (Ryals et al., 1982a;. In contrast, in the ppGpp-deficient strains r8/rt was nearly constant, about 0.6 in AAspoT207 and 0.5 in AAspoT209, a t all growth rates (Fig. 4, open symbols). However, in a given medium rs/rt was reduced in the ppGppless strains relative to the wild-type strain (Table 11).
Expression of lacZ From a n rRNA P1 Promoter-The ArelA and AspoT alleles of strains CF1693 and CF1673 were recombined into a "ppGpp reporter" strain bearing a single chromosomal copy of an rrnB PI-lac2 fusion (Table I; Bremer (1990, 1991)). In the ArelA SPOT strain, VH2733, @-galactosidase specific activity increased with growth rate, as previously observed in a relA1 background (Hernandez and Bremer, 1990). However, in the ppGpp-deficient strains, the specific activity remained nearly constant (Fig. 51, consistent with the growth rate invariance of rJrt in ppGpp-less strains (Fig. 4). When the specific activity was determined per amount of protein rather than per cell mass, the same results were obtained (data not shown).
Synthesis Rate of mRNA-The mRNA synthesis rate/protein, calculated as described (Equation 3 under "Experimental Procedures"), was nearly constant in the wild-type strain, but it increased dramatically with growth rate for the AAspoT207 strain so that in rich media the ppGpp-less strain synthesized 4-fold more mRNA/protein than the wild-type (Fig. 6). This suggests that mRNA synthesis is severely inhibited at very low, basal levels of ppGpp.

RNA Polymerase Concentration and Activity
The RNA polymerase concentration, determined as p'subunits/total protein, increased with growth rate in the AAspoT207 ppGpp-less strain, but less than in the wild-type strain (Fig. 7A, uppermost curves). On the other hand, the : : I :  Figs. 1 and 2). The wild-type strain was grown in the following media: Glc aaSer; glycerol aaSer; glycerol aa; Glc; and glycerol. For the ppGpp-less strains the media were: Glc aaSer + adenine, xanthine, cytosine, uracil; Glc aaSer; glycerol aaSer; Glc aa; and glycerol aa. Symbols are the same as in Fig. 2. number of active RNA polymerase molecules, determined as described (Equation 4 under "Experimental Procedures"), was greater in the AAspoT207 than in the wild-type strain (Fig.  7A, lower curves). The fraction of total RNA polymerase engaged in transcription a t any one time (RNA polymerase activity) increased with growth rate from 15 to 30% in the wild-type strain ( Fig.  7 B ) , as previously reported (Shepherd et al., 1980a). In the AAspoT207 strain, RNA polymerase activity increased twice as much in this range of growth rates, from 20 to 60% (Fig.  7B). Thus, the increased RNA polymerase activity in rich media more than compensates for the decreased total RNA polymerase concentration in the AAspoT207 strain.

Chromosome Replication and Gene Activities
To evaluate the possibility that rrn genes might be limiting for transcription in the ppGpp-less strain, thus forcing RNA polymerase to overexpress mRNA genes, the DNA replication time and rrn gene dosage at different growth rates were determined in the ppGpp-less AAspoT207 strain.
DNA Replication Time-In the presence of rifampicin, initiation of replication ceases, but ongoing rounds of replication proceed to completion (Skarstad et al., 1986). The increase in DNA after cessation of initiation, AG (Fig. 8, A and B ) , corresponds to the average number of oriC sites per genome equivalent of DNA, oriC/genome (Pritchard and Zaritsky, 1970). The number of oriC sites per genome is a measure for  (Bremer and Dennis, 1987). A , (lower curves) number of active RNA polymerase molecules/pg protein, calculated from RNA/protein ratios (Fig. 2) and rJrt (Fig. 4; Table 11) as described under "Experimental Procedures." Filled symbols, wild-type strain; open symbols, AAspoT207 strain. C, RNA polymerase activity (ratio of lower to upper curves in A ) . AAspoT = AAspoT207 strain. the extent of chromosome branching due to overlapping rounds of replication and allows determination of the DNA replication time (C-period (Pritchard and Zaritsky, 1970;Churchward and Bremer, 1977)).
The exact time after addition of rifampicin at which initiation of replication stops was determined from the kinetics of DNA accumulation after rifampicin addition  as illustrated graphically in Fig. 8, C and D. There was an 8-min delay in the time at which rifampicin stopped initiation in the wild-type strain (Fig. 8, A and C), while in the ppGpp-less mutant rifampicin action on initiation was instantaneous (Fig. 8, B and D ) . The time delay in rifampicin action occurred at all growth rates in the wild-type and was negligible for the AAspoT207 strain.
In the AAspoT207 strain AG (oriC/genome) was 10-20% greater than in the wild-type strain (Fig. 9B). This indicates more chromosome branching resulting from a 20-30% longer replication time, corresponding to a lower average replication velocity in the absence of ppGpp (Fig. 9C). At growth rates between 1.0 and 2.5 doublings/h, the average replication velocity in the wild-type strain increased with growth rate from 600 to 800 bp/s per replication fork, in good agreement with values reported for E. coli B/r . In the AAspoT207 strain, the average replication velocity was lower and increased from 500 to 600 bp/s/fork over the same range of growth rates. The longer C-period in the ppGpp-less strain was associated with a reduced initiation mass (Fig. 9D, see below). Previously, the same correlation has been observed after overproduction of DnaA protein and was thought to be due to stalled replication forks when initiation has occurred prematurely (Lobner-Olesen et al., 1989).  . 8. (A and B ) DNA accumulation after treatment of cultures with rifampicin. T = culture doubling time before rifampicin addition; AG = ratio of the amount of DNA after cessation of DNA accumulation relative to the amount of DNA at the time when initiation of DNA replication ceases; y = difference between the amounts of DNA which accumulate with and without rifampicin treatment, respectively. C and D, evaluation of the kinetic curves in A and B, 4 plotted as a function of the time after rifampicin addition.
The intersection of the y-curve with the abscissa is the time after rifampicin addition when replication initiation ceases. Filled symbols, wild-type; open symbols, AAspoT207.
DNA Concentration-In the wild-type strain the DNA concentration (DNA/protein; Fig. 9A) decreased a t higher growth rates as previously reported (Brunschede et al., 1977;Bremer and Dennis, 1987). Conversely, in the AAspoT207 strain, the DNA concentration increased with growth rate and was higher than in the wild-type strain at high growth rates.
Initiation Mass-The initiation mass (cell mass per replication origin at initiation of replication (Donachie, 1968)) is given by the amount of protein per oriC site ( Fig. 9 D ) , and was obtained from the amount of protein/genome (Fig. 9A) and oriC/genome (Fig. 9B). In the wild-type strain, the initiation mass was fairly constant, in agreement with previously reported observations, equal to about 5 x108 amino acid residuesloric (Donachie, 1968;Bremer and Dennis, 1987). In the ppGpp-less strain AAspoT207, the initiation mass decreased from 6 XlO' at 1.0 doubling/h to 3.5 XlO' amino acid residuesloric at 2.5 doublings/h (Fig. 9D).
Ribosomal RNA Gene Actiuities-The rrn gene activity was calculated as transcriptional initiations/min per rm gene (Fig.  1OA) by combining rRNA synthesis rates (Fig. 4A) and DNA concentrations (Fig. 9A), and taking into account the replication velocity (Fig. 9C) and chromosomal map locations of the seven rrn genes (using Equation 7 of "Experimental Procedures"). The rrn gene activity increased parabolically with growth rate in the wild-type strain from 10 to about 75 transcripts/min per gene between 1.0 and 2.5 doublings/h (Fig. 1OA). The latter value corresponds to 1.25 transcripts/s per rrn gene, which approaches physical saturation of rrn genes with RNA polymerase. In the ppGpp-less strain, rrn gene activity increased less, from 20 to 60 transcriptions/min/ gene (Fig. 1OA). Relative to the wild-type strain, this rate was reduced above, and increased below, a growth rate of 1.25 doublings/h, which is consistent with the differences in r8/rt 0.20, . . . . . , values observed in these strains by independent measurements (Fig. 4).
Messenger RNA Gene Actiuities-Assuming that mRNA genes are evenly distributed throughout the chromosome, a relative value for the average mRNA gene activity was calculated from mRNA synthesis rates (Fig. 6) and DNA concentrations ( Fig. 9A; Equation 9, "Experimental Procedures"). In the wild-type strain mRNA gene activity decreased 30% from 750 to 500 initiations/min per genome between 1.0 and 2.5 doublings/h, whereas in the AAspoT207 strain it increased almost 4-fold from 550 to 2000 initiations/min per genome over the same growth rate range (Fig. lOB), in parallel to both the rRNA gene activity (Fig. 1OA) and the RNA polymerase activity (Fig. 7B).

DISCUSSION
E. coli K-12 strains that do not produce ppGpp were found to differ from wild-type strains with regard to growth rate regulation of RNA synthesis ( r8/rt), RNA polymerase synthesis, and DNA synthesis, but only slightly with respect to protein synthesis (i.e. ribosome activity). To assess the role of ppGpp from these results, a relationship between bacterial growth rate and ribosome concentration, first reported by Schleif (1967), will be again derived and evaluated here for clarity.

Ribosomal RNA Synthesis at Different Growth Rates
Relationship between Growth Rate and Ribosome Concentration-Bacterial growth may be defined as increase in protein, P , of an exponential culture with the doubling time 7 (in min) as follows.
The rate of protein accumulation is obtained from this expression by differentiation: dP/dt = P X In 2/7 (Eq.

11)
Under most conditions bulk protein turnover is negligible, so that the rate of synthesis equals the rate of accumulation of protein. Substituting the culture growth rate p (in doublingsl h; p = 60 min/r), setting k = ln2/(60 min/h), and rearranging to make ~1 the dependent variable, gives the growth rate in terms of the rate of protein synthesis as follows. k p = (dP/dt)/P (Eq. 12) Expanding this expression by the number of ribosomes, N,, gives, where [N,/P] is the number of ribosomes per amount of protein, i.e. the ribosome concentration, and (dP/dt)/N, is the protein synthesis rate (amino acids polymerized/min) per average ribosome, or ribosome activity, a, as follows. This relationship (Equation 14) defines the bacterial growth rate as the product of ribosome concentration and activity.
(Ribosome activity, defined as protein synthesis rate divided by the number of ribosomes, is to be distinguished from peptide chain elongation rate, defined as protein synthesis rate divided by the number of active ribosomes.) Since 84% of the total amount of E. coli RNA which accumulates is rRNA (2% is mRNA (Baracchini and Bremer, 1987); and 14% is tRNA (Bremer and Dennis, 1987)), and each 70 S ribosome contains collectively 4566 nucleotide residues in its 16 S, 23 S, and 5 S rRNAs, the ribosome concentration, [Nr/P], is found from the ratio RNA/protein, R I P (RNA nucleotides/amino acid residue) as follows. The results in Fig. 2, A and B, show that RIP is nearly proportional to p for both wild-type and ppGpp-less strains. According to Equation 16, this means that the ribosome activity, a,, is approximately constant. Using Equation 16, a, was calculated from RIP and p (Fig. 2, C and D ) and was found to increase from about 12 amino acid residues polymerizedls (aa/s) a t 0.6 doublingslh to a plateau of 15 aa/s at growth rates above 1.5 doublingslh, in agreement with previous estimates (Schleif, 1967;Dennis and Bremer, 1974).
The ppGpp independence of a, at growth rates above 1.5 doublings/h (Fig. 2, C and D ) is consistent with previous observations that the ribosome activity is not altered in spoT strains which have elevated basal levels of ppGpp (Hernandez and Bremer, 1990). At low growth rates the ppGpp-less strains did have somewhat lower ribosome activity than the wild-type strain due to a small overproduction of ribosomes (Fig. 2, A  and B). Thus, the reduction in growth rates of ppGpp-less strains in a given medium (Fig. 1) was mainly due to decreased ribosome concentration at high growth rates and to decreased ribosome activity (up to 30%) at low growth rates (Table 11).
From the observation that the growth rate dependence of R I P is not altered by the absence of ppGpp, Gaal and Gourse (1990) concluded that ppGpp is not required for the growth rate-dependent control of rRNA synthesis. This interpretation is not justified, because the control of the rRNA synthesis rate cannot change R I P a t a given growth rate unless it also changes ribosome activity (Equation 16).
Based on the assumption that ppGpp inhibits ribosome synthesis (see Introduction), one might have expected that ppGpp-less bacteria would overproduce ribosomes at all but the fastest growth rates. The excess of ribosomes, evident as an elevated R I P value at a given growth rate, might then deplete the supply of substrates for protein synthesis and lead to a reduced ribosome activity, as is observed when extra copies of plasmid-borne rrn genes are present (Baracchini and Bremer, 1991). T o a limited extent, our observations agree with this expectation: at low growth rates, ribosomes were 10-20% overproduced in the ppGpp-less strains, seen as a 10-20% greater RIP at a given growth rate, which resulted in a 10-20% reduced ribosome activity. However, one might have expected a much greater effect, such as a constant high R I P value in the absence of ppGpp at all growth rates. In this case the growth rate could only vary as a result of changes in ribosome activity. This is not observed for a variety of reasons. First, a greater overproduction of ribosomes would be expected to waste sufficient energy to lower the growth rate and thus cause a number of secondary changes in global transcription control. Second, the rate of stable RNA synthesis is regulated not only by ppGpp, but also by changes in the RNA polymerase concentration and activity (see below). Third, the expectation that ppGpp only inhibits ribosome synthesis (so that the absence of ppGpp should stimulate ribosome synthesis) must be modified on the basis of observations reported here. Unexpectedly, it was found that there is a significant difference between near zero and absolute zero levels of ppGpp in the global control of transcription. At very low levels, ppGpp preferentially inhibits mRNA synthesis, so that the total absence of ppGpp leads to a stimulation of mRNA synthesis in rich media at the expense of stable RNA synthesis. These phenomena are discussed below. That a reduction in the level of ppGpp stimulates rRNA synthesis in a given medium without amino acid starvation has been shown previously by the isolation of PSII-deficient mutants upon selection for increased expression from an rRNA promoter (Hernandez and Bremer, 1991). Conversely, increasing the level of ppGpp by variable induction of relA from PlacUV5 suffices to reduce both rJP and rJrt (Tedin and Bremer, 1992).
Expression of lac2 from an rRNA Promoter-We have previously reported the construction of a relAl AlacZ strain carrying an rrnB P1-lac2 fusion; this fusion was recombined into the chromosome at a position close to the normal map location of rrnB and oriented so that the direction of transcription coincides with the direction of replication, which simulates the natural disposition of the rrn genes. Under conditions of reduced ppGpp concentration, e.g. in rich media or during the relaxed response, @-galactosidase specific activity expressed from rrnB P1-lac2 is increased, reflecting increased rRNA gene activity (Hernandez and Bremer, 1990).
Here the relAl allele of this strain was replaced with a relA deletion. Again, P-galactosidase specific activity (activity per cell mass) increased with growth rate (Fig. 5, filled symbols). Following deletion of the spoT gene as well, /I-galactosidase specific activity became growth rate-independent (Fig. 5, open  symbols).
The specific activity of @-galactosidase (enzyme per total protein) is proportional to the ratio of the synthesis rates, lac-mRNA/total mRNA, and when lac2 is expressed from a stable RNA promoter, @-galactosidase specific activity reflects rJr,; i.e. the ratio of lacZ-mRNA synthesis rate expressed from an rrnB P1 promoter to the total mRNA synthesis rate (Hernandez and Bremer, 1990). Since rs/rt is constant in ppGpp-less strains (Fig. 4), then rs/rm must also be constant (because rt = rs+rm). Thus, a constant ,&galactosidase activity in the absence of ppGpp (Fig. 5) is expected from the direct measurements of r, and r,.
The P-galactosidase activity expressed from a stable RNA promoter might be expected to increase with growth rate like R I P or rJP. The comparison of Figs. 2, A and B, and 5 shows that this is evidently not the case. The increases in RIP and rJP in the ppGpp-less strain reflect the increasing concentration of active RNA polymerase, which increases r, and rm equally (Figs. 3 and 6), so that the ratio rJr, does not change. Differences between P-galactosidaselp (corresponding to rs/ r,) and RIP, or rJP, respectively, are due to the fact that the total rate of protein synthesis is not limited by mRNA, but by the number of ribosomes (Bremer and Dennis, 1987). That is, overproduction of bulk mRNA reduces the fraction of mRNAs which represent lac2-mRNA produced from an rrnB P1 promoter, and since ribosomes are limiting for protein synthesis, this reduces the translation of lac2-mRNAs. R I P and rJP are not affected in this case because the rate of total protein synthesis remains the same.
The use of r8/rt to characterize the control of r, has been considered inappropriate (Nomura et al., 1984;Jinks-Robertson and Nomura, 1987). Because of the close relationship between rJrt and rJr,, this criticism would equally apply to measurements of @-galactosidase expressed from an rRNA promoter. These authors themselves, however, have used such measurements as a function of p as indicators of rrn promoter activity (e.g. Gourse et al., 1986), assuming @-galactosidaselp expressed from an rRNA promoter as a measure for R I P or r J P .
Using the same AreU and AspoT alleles in combination with a different rrnB P1-lac2 fusion (containing about 1 kbp of trp sequences in the lac2 mRNA leader), Gaal and Gourse (1990) found no difference in the growth rate-dependence of P-galactosidase specific activity between ppGpp-deficient and wild-type strains. Their fusion gene was inserted a t a different location on the chromosome (i.e. at the Xatt site, since it is imbedded within a X transducing phage) and in an orientation opposing the direction of replication fork movement. We do not know if either of these differences might be responsible for the discrepancy between their and our results, The inhibition of rrn P1 promoter activity by ppGpp in uitro can be modulated by small changes in superhelical density of the DNA templates. High negative superhelical density causes maximal rrn P1 promoter activity and insensitivity to ppGpp; upon relaxation of the DNA template up to 80% inhibition by ppGpp is observed (Ohlsen and Gralla, 1992). In addition, it has been shown that RNA polymerase pausing within the rrnB leader is sensitive to the superhelical density of the template (Krohn et al., 1992). Therefore, differences in chromosome location of the rrnB P1-lac2 fusion might affect their control. Besides potential effects of DNA superhelicity on the rrn promoter activity, the expression of lac2 is subject to a number of effects unrelated to promoter activity; these include mRNA stability, intracistronic polarity, and translational efficiency of different mRNAs (Hernandez and Bremer, 1990), which may differ for different lac2 fusions.
Rate of rRNA SynthesislProtein-At a given growth rate, the rRNA synthesis ratelprotein (rRNA transcripts/min per amino acid residue) depends only on the ribosome activity. This is seen by substitution of Equation 16 above into Equation l ("Experimental Procedures"), which yields, where f = 4655k2 = 0.62 (this value o f f assumes a, in amino acids polymerized/min per ribosome). Equation 17 shows, as long as ppGpp does not alter a, (see above), it cannot affect the rate of rRNA synthesis/protein as a function of p, as observed (Fig. 3). The parabolic shape of the curve in Fig. 3 agrees with the expectation that the rate of rRNA synthesis/ protein is proportional to p2 and implies a constant (or nearly constant) ribosome activity. Maaloe (1969) stated that the rate of rRNA synthesis per genome equivalent of DNA increases with p2. The rates of rRNA synthesis per protein and per genome can both increase with p* only if DNA/protein remains growth rate-invariant. In reality this condition is met neither in wild-type nor in ppGpp-deficient strains ( Fig. 9A and Brunschede et al., 1977;Churchward et al., 1982). Thus, only the rate of rRNA synthesis per protein, but not per genome, increases with p2.
Growth Rate-and ppGpp Dependence of Stable RNA Synthesis-Transcriptional activities like r, are generally expressed either per amount of protein ( rs/P), or per amount of total RNA (rs/R), or per total RNA synthesis rate (r8/rt). Whereas r,/P increases with F' and is independent of ppGpp (see above), rJR increases in direct proportion to p (seen by substituting R for P in Equation 11, above), and is also independent of ppGpp; rJrt may increase with p or remain constant (Fig. 4) and is strictly dependent on the concentration of ppGpp (Ryals et al., 1982b ; Fig. 4). These differences in the growth rate and ppGpp dependence are due to the control of the reference unit. Depending on the parameter used to express the stable RNA synthesis rate different properties of r, are being measured. At a given growth rate, r,/P reflects the ribosome activity (see preceding section); rJR reflects the stability of stable RNA (Baracchini and Bremer, 1991); and rs/rt reflects the proportions of mRNA and stable RNA synthesis. Evidently, there is no "clean" measure to define r, , which obscures the role of ppGpp in its control.
Growth Medium Control of Stable RNA Synthesis-Since rJP is fixed a t a given growth rate so long as ribosome activity remains constant, neither ppGpp nor any other factor can alter it. Only if rs/P is compared in the wild-type and ppGppless mutant for a given growth medium (i.e. not for a given growth rate), can it be seen that the absence of ppGpp does affect rJP. Maaloe and coworkers suggested that only the growth rate is important, but not the composition of a particular growth medium used to achieve that growth rate (Schaechter et al., 1958). Therefore, physiological parameters in bacteria are commonly presented as functions of p, which assumes 1. 1 as an independent variable. Actually, only the growth medium can be freely chosen, the composition of which ultimately determines the growth rate. If only wild-type bacteria are considered, there is no need to distinguish between "growth rate control" and "growth medium control" of r,. However, for mutants with altered control of ribosome synthesis this distinction must be made. That the absence of ppGpp alters the growth medium-dependence of rJP is evident from the data in Table 11. In a graph, growth mediumdependent effects of ppGpp can be visualized by using the growth rate of wild-type bacteria to represent the growth medium, as in Fig. 1.
Factors Determining the Rate of Stable RNA Synthesis/ Protein-Even for a given growth medium, the absence or presence of ppGpp does not strongly affect rJP (5-20%; Table  11). The reason for this small effect is that rs/P is the product of four factors (Bremer, 1975) with different ppGpp dependencies which partly compensate one another as follows.

(Eq. 18)
In this expression [RNAP] is the RNA polymerase concentration, given as the number of core RNA polymerase molecules/ protein (Fig. 7A, upper curves), @, is the fraction of RNA polymerase that is active at any instant (Fig. 7B), +s is the fraction of active RNA polymerase synthesizing stable RNA at any instant (a function of rs/rt), and c, is the stable RNA chain elongation rate (nucleotides/min per growing chain). With the exception of c,, all other parameters in this relationship have been determined here in strains with and without ppGpp. Since the stable RNA chain elongation rate is the same during both the relaxed and stringent response, i.e. a t very low and high levels of ppGpp (Shen and Bremer, 1977;, we have assumed c, to be independent of ppGpp (see "Experimental Procedures," Equation 4).
In the wild-type strain, [RNAP] (Fig. 7A 1, P, (Fig. 7B), and (derived from rJrt in Fig. 4) all increased with growth rate. The combined increases in these factors causes the parabolic increase with p2 in r,/P (Fig. 3). In the ppGpp-deficient strain, $s (i.e. rs/rt) was constant (Fig. 4) and r,/P was adjusted at different growth rates entirely as a result of changes in RNA polymerase synthesis and activity (Fig. 7). Thus, in the absence of ppGpp and in rich media, the loss of specific control by ppGpp (i.e. rs/rr) is compensated by a nonspecific increase in global transcription that equally affects both stable and mRNA synthesis. In poor media, rRNA is overproduced in ppGpp-deficient bacteria, which is partly compensated by a reduced ribosome activity (Table 11).
Increased Synthesis of mRNA in the Absence of ppGpp The rate of mRNA synthesis was up to 4-fold higher in PpGpp-deficient bacteria than in wild-type bacteria (Fig. 6). The difference increased with growth rate and was absent at growth rates below 1.25 doublings/h. This increased mRNA synthesis was associated with an increased RNA polymerase activity relative to the wild-type strain (Fig. 71, as if RNA polymerase were inactivated during the synthesis of mRNA in ppGpp-synthesizing wild-type bacteria, but not in the ppGpp-deficient mutants. Independent of whether these effects are caused directly or indirectly by ppGpp, one would expect them to diminish with increasing growth rate when the ppGpp concentration in wild-type bacteria approaches zero. However, contrary to this expectation the effects were found to be maximal during fast growth (Figs. 6 and 7). We suggest this paradox may be attributable to ppGpp-dependent transcriptional pausing of RNA polymerase during the synthesis of mRNA. If gene expression were inhibited by transcriptional pausing, it would depend on the fraction of RNA polymerase that pauses, the duration of the pause, the distance of the pause site from the promoter, and most importantly on the time intervals between successive transcription initiations. If these initiation intervals are longer than the pause time, then pausing would not affect the rate of transcription initiation, no matter how large the fraction of polymerase that pauses. Therefore, transcriptional pausing should have a minimal effect on the rate of gene expression during slow growth when RNA polymerase concentrations are low and the number of derepressed mRNA genes competing for RNA polymerase are high; i.e. the ratio of free RNA polymerase to available promoters is low. Only at high growth rates, when RNA polymerase concentrations are high and the number of remaining derepressed mRNA genes are low, could ppGpp-induced pausing significantly reduce the rate of transcription and transiently inactivate RNA polymerase. Enhanced transcriptional pausing at specific sites in the presence of ppGpp has been observed in vitro . Thus, the growth ratedependent differences in mRNA synthesis and RNA polymerase activity observed here between wild-type and ppGppdeficient bacteria (Figs. 6 and 7) suggest that ppGpp-enhanced transcriptional pausing could also be occurring in vivo.
In addition, this would explain the unexpected observation that at high growth rates rB/rt is lower in ppGpp-deficient than in wild-type bacteria (Fig. 4); i.e., since r,,, is increased. The presence of antitermination sites in the stable RNA genes might explain why the effect appears to be specific for mRNA synthesis.
The lower rJrt values in fast growing ppGpp-less bacteria indicates that stable RNA synthesis rates can also be controlled indirectly, via the control of the rate of mRNA synthesis. This reflects the fact that in any transcription system that is not limited by the concentration of DNA, promoters compete for RNA polymerase. It has been shown in vivo, that transcription in E. coli is not limited by DNA (Churchward et al., 1982). Thus, if two classes of promoters are considered, i.e. mRNA and stable RNA promoters, then the distribution of RNA polymerase over these two classes of genes can be changed by either activating one group, or inhibiting the other group of promoters. For example, an increase in stable RNA synthesis can be achieved by activating stable RNA promoters or by inactivating mRNA promoters. Therefore, the control of (bulk) mRNA synthesis is inseparable from the control of stable RNA synthesis.

Modulation of RNA Polymerase Synthesis and Activity by PPGPP
In wild-type bacteria, changes in RNA polymerase concentration and activity accompany changes in ppGpp levels and contribute to the control of stable RNA synthesis (see above).
For ppGpp-deficient strains, changes in RNA polymerase concentration and activity seem to provide the only means to adjust the rate of stable RNA synthesis in response to different growth media. The lower RNA polymerase activity in fast growing wild-type bacteria in comparison to ppGpp-less bacteria is assumed to reflect ppGpp-dependent transcriptional pausing (see above). The increased RNA polymerase synthesis in wild-type bacteria (Fig. 7A, upper curves) may then be a consequence of the decreased RNA polymerase activity, since a negative correlation between RNA polymerase activity and synthesis has been previously observed under a variety of conditions and appears to reflect a feedback control of the transcriptional attenuator upstream of the rpoBC genes (Downing and Dennis, 1991).

Role of ppGpp in the Control of Chromosome Replication
The promoters of the dnaA and mwC genes, which are involved in the control of replication initiation at oriC, are subject to growth rate and stringent control (Rokeach and Zyskind, 1986;Zyskind, 1989, 1990). Therefore, in the absence of ppGpp, initiation of chromosome replication might be stimulated due to increased transcription of these genes, producing the lower initiation mass in the spoT207 strain at high growth rates (Fig. 9D). However, despite the absence of ppGpp, the initiation mass was normal in the ppGpp-less strain at a growth rate of 1.25 doublings/h when rJrt was also normal. This suggests that ppGpp does not directly affect replication initiation, but that its effect may be indirect, related to the general physiology of the bacteria and rs/rt.