Underproduction of ς70Mimics a Stringent Response

When Escherichia coli cells enter stationary phase due to carbon starvation the synthesis of ribosomal proteins is rapidly repressed. In a ΔrelAΔspoT mutant, defective in the production of the alarmone guanosine tetraphosphate (ppGpp), this regulation of the levels of the protein synthesizing system is abolished. Using a proteomic approach we demonstrate that the production of the vast majority of detectedE. coli proteins are decontrolled during carbon starvation in the ΔrelA ΔspoT strain and that the starved cells behave as if they were growing exponentially. In addition we show that the inhibition of ribosome synthesis by the stringent response can be qualitatively mimicked by artificially lowering the levels of the housekeeping ς factor, ς70. In other words, genes encoding the protein-synthesizing system are especially sensitive to reduced availability of ς70 programmed RNA polymerase. This effect is not dependent on ppGpp since lowering the levels of ς70 gives a similar but less pronounced effect in a ppGpp0 strain. The data is discussed in view of the models advocating for a passive control of gene expression during stringency based on alterations in RNA polymerase availability.

When Escherichia coli cells enter stationary phase due to carbon starvation the synthesis of ribosomal proteins is rapidly repressed. In a ⌬relA ⌬spoT mutant, defective in the production of the alarmone guanosine tetraphosphate (ppGpp), this regulation of the levels of the protein synthesizing system is abolished. Using a proteomic approach we demonstrate that the production of the vast majority of detected E. coli proteins are decontrolled during carbon starvation in the ⌬relA ⌬spoT strain and that the starved cells behave as if they were growing exponentially. In addition we show that the inhibition of ribosome synthesis by the stringent response can be qualitatively mimicked by artificially lowering the levels of the housekeeping factor, 70 . In other words, genes encoding the protein-synthesizing system are especially sensitive to reduced availability of 70 programmed RNA polymerase. This effect is not dependent on ppGpp since lowering the levels of 70 gives a similar but less pronounced effect in a ppGpp 0 strain. The data is discussed in view of the models advocating for a passive control of gene expression during stringency based on alterations in RNA polymerase availability.
The alarmone guanosine tetraphosphate (ppGpp) 1 of the stringent response network in Escherichia coli affects ribosome production by specifically lowering the transcription of ribosomal RNA (rRNA) (1) and some of the genes encoding ribosomal proteins (2)(3)(4). The production of ribosomal proteins is also post-transcriptionally feedback-regulated to match the rRNA production (reviewed in 5). Two different ppGpp synthetases (PS) exist in E. coli, the ribosome-associated PS I, encoded by relA, and the cytoplasmic PS II (6), encoded by spoT (7,8). The protein PS II is also responsible for ppGpp hydrolysis (1).
Increased levels of ppGpp not only down-regulate ribosome production but also induce transcription from several promoters (9 -14). As suggested by Schreiber et al. (9), the E. coli promoters can be divided into three groups depending on their response to ppGpp: promoters specifically induced or repressed by the stringent response and promoters that are unaffected by the rise in ppGpp levels. Promoters dependent on the housekeeping factor, 70 , exist in at least two of these groups; one group is positively regulated by ppGpp, e.g. PuspA (11) and Phis (8), whereas the other group is repressed during stringency, e.g. rrnP1. At-tempts to explain this dual effect of ppGpp have involved considerations of RNAP (RNA polymerase) availability and intrinsic differences in the kinetic properties of the promoters affected (15). Several lines of evidence suggest that the availability of transcriptional and/or translational machinery play a role in global gene regulation in concert with classical activators and repressors (13)(14)(15)(16)(17). One well known example of this type of regulation, where the level of RNAP is involved, is factor competition in both Bacillus subtilis (18) and E. coli (19,20).
Whether or not RNAP availability is involved in the actual mechanism of the stringent response has been discussed, and several models for this have been suggested (13)(14)(15)(16)20). Arguments have been raised both for an increased and diminished availability of RNAP during entry into stationary phase. In addition, different models exist on how changes in the availability of RNAP would affect the stringent promoters. Barker et al. (13,14) and Zhou and Jin (15) suggest an increased availability of free RNAP in stationary phase since ppGpp destabilizes the open complex resulting in RNAP drop off at stable RNA promoters, which form intrinsically unstable open complexes. As a consequence of the increased availability of RNAP, positively regulated promoters are then induced since they are relatively poor at recruiting RNAP and are subsaturated during normal growth according to this model. Thus, this model encompasses a direct and active role for ppGpp on stringently controlled promoters (e.g. rrn) and a passive role on the positively regulated promoters (through RNAP availability). Jensen and Pedersen (16), on the other hand, have argued for a diminished availability of RNAP during stringency. The "stringent" promoters, e.g. stable RNA promoters, are dependent on high concentration of RNAP to transcribe at their maximal rate and are therefore argued to be repressed when the RNAP availability is diminished. Krohn and Wagner (21) showed that ppGpp increases pausing of RNAP during transcription in general but more at the stringently controlled genes, which could be another reason for inhibition of expression of these genes. Jensen and Pedersen (16) suggested that one way through which the RNAP availability could be diminished during stringency is through ppGpp-dependent pausing and therefore sequestering of RNAP in transcription. However, Vogel and Jensen (22) have demonstrated that ppGpp-induced pausing is not required for the stringent response since the stringent response was still observed when ppGpp-induced pausing was abolished and the transcriptional elongation speed was made constant by the introduction of the nusAcs10 allele. Thus, it is presently unclear if and how the availability of RNAP changes during the stringent response.
There are also models for how the stringent response works that do not involve RNAP availability. For example Barrachini and Bremer (23) suggested that RNAP can exist in two forms, one with and one without ppGpp bound to it. The RNAP without ppGpp can only transcribe genes encoding stable RNA, * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
while RNAP with ppGpp bound can transcribe from both stable RNA and mRNA promoters but prefers mRNA promoters. The fact that cells totally deficient in making ppGpp can still grow shows that there is no absolute requirement for ppGpp for the transcription of mRNA promoters, but no evidence presented so far discredits the idea that there are direct effects of ppGpp on promoter selection.
The questions that need to be solved are whether variations in the level of free RNAP play a physiologically relevant role in gene regulation and if this mechanism is involved in the stringent response. The specific question we set out to answer in this work was whether variation in the levels of free 70programmed RNAP does affect gene expression and, if so, what genes are sensitive to the altered RNAP availability? We used two-dimensional gel analysis to elucidate the global effects of underproduction of the housekeeping factor, 70 , during exponential growth and to study the role of ppGpp in this effect. We found that lowering the amounts of 70 -programmed RNAP mimics the proteome of stringent cells and that this effect is not ppGpp-dependent. We speculate on the physiological relevance of these results with respect to the availability of RNAP during the stringent response.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Growth Conditions-The E. coli strains used in this work are listed in Table I. The markerless ⌬relA ⌬spoT strain was constructed by K. Kvint and co-workers (20) by the method of Datsenko and Wanner (24). 70 levels were manipulated by using a system where 70 (RpoD) is under control of the trp-promoter, P trp , (25). Levels of 70 were controlled with IAA (Indole-3-acrylic acid, an antagonist of the Trp repressor). The (CamR) P trp -rpoD allele was transduced into the different strains by standard P1 transduction (26). The IAA concentration giving wild type levels of 70 was determined previously (0.2 mM IAA, (20)), and a 2-fold underproduction of 70 was accomplished by diluting to a final concentration of 0.02 mM IAA.
Cultures were grown in liquid M9-defined medium (27) aerobically in Erlenmeyer flasks in a rotary shaker at 37°C. The M9 medium was supplemented with a limiting concentration of glucose (0.08%), all the amino acids in excess (28) and thiamine (10 mM). IAA (0.2 or 0.02 mM) and chloramphenicol (30 g/ml) were added when appropriate.
Resolution of Proteins by Two-dimensional Polyacrylamide Gel Electrophoresis-After growth for at least five generations, samples were taken in exponential phase, and 1 ml of the culture was mixed with 5 l of [ 35 S]methionine (10 mCi/ml, 1000 Ci/mmol, Amersham Biosciences) for pulse labeling. Incorporation was allowed to proceed for 5 min at 37°C, then non-radioactive methionine (50 l, 0.2 M) was added and a chase was allowed for 3 min. The samples were processed for resolution on two-dimensional polyacrylamide gels according to O'Farrell (29) with modifications (30). Electrophoresis in the first dimension was carried out between pH 3.5 and pH 10, and the gels were run according to the NEPHGE protocol (Non Equilibrium pH Gel Electrophoresis, (31,32)). The NEPHGE protocol was used since it results in two-dimensional gels where the ribosomal proteins (most with very basic pI) are visible. The second dimension was run on 11.5% polyacrylamide gels. Radiolabeled proteins were detected in a phosphorimaging device (Personal FX, Bio-Rad), and the images were analyzed using the PDQuest 6.2 software (Bio-Rad). Alphanumeric designations and/or protein names were assigned to protein spots after matching them to the reference twodimensional images of the gene-protein data base of E. coli (33) or after performance of mass spectrometry.
Measurement of Cellular Components-Western blot analysis was performed as described (20). Cell sampling and measurement of the cellular level of ppGpp with high pressure liquid chromatography was done according to Neubauer et al. (34). For RNA and protein measurements, bacteria were first incubated in 5% trichloroacetic acid on ice and subsequently washed once with 5% trichloroacetic acid and lysed with 1 M NaOH (30 min at 40°C). Protein content in the cell extracts was analyzed using the BCA protein assay kit (Pierce). RNA concentration was measured using the UV absorption assay as described (35). Based on the values obtained (normalized to optical density of the culture at the time of sampling), the RNA/protein ratio was calculated for cells underproducing 70 and compared with the control.

RESULTS
A ppGpp 0 Mutant Fails to Respond to Carbon Starvation-A proteomic analysis demonstrated that the ribosomal proteins are immediately down-regulated 4-to 64-fold when wild type cells enter stationary phase due to carbon starvation (Fig. 1, B  and D). The fold reduction in the synthesis rate is different for each of the ribosomal proteins but eventually the production of all the ribosomal proteins falls below the detection limits (data not shown). As demonstrated previously a set of proteins is induced during entry into stationary phase ( Fig. 2A). Among those are a set of 70 -dependent proteins, e.g. UspA and GlyA (Fig. 1D). In a ppGpp 0 strain the ribosomal proteins are not down-regulated when the cells enter stationary phase (Fig. 1E), and the protein production pattern, as seen on the two-dimensional gels, looks rather like a growing cell even after the entry into stationary phase (Fig. 1C). It should be noted that relA ϩ and relA1 strains behave indistinguishably during the conditions analyzed. The SpoT protein (PS II) is responsible for ppGpp production during the glucose starvation conditions employed (8), and the relA1 allele did not affect the proteome during these conditions. Apart from differences between the rate of production of specific proteins in the wild type and relaxed strain, it is clear that the magnitude of the proteome response is much reduced in the ⌬relA ⌬spoT mutant strain (Figs. 1 and 2). The proteome of the ⌬relA ⌬spoT mutant strain appears to be locked in a growth mode.
70 Underproduction Qualitatively Mimics a Stringent Response-When 70 was underproduced in a wild type strain the ribosomal proteins were found to be specifically affected. When the 70 levels were lowered to half the level of a normal strain, the ribosomal proteins were down-regulated 2-to 3-fold (Fig.  3A). This effect was not as great as the effect of carbon starvation, but this was expected since the cells are still growing in the 70 underproduction experiment. The effect of 70 underproduction on ribosomal proteins has been confirmed on the transcriptional level by the use of macroarrays. 2 The ppGpp levels were not affected by the 70 underproduction (Fig. 3B), which means that the down-regulation of ribosomal protein production is not caused by elevated levels of ppGpp in this experiment. In agreement with the effect on ribosomal proteins the RNA/protein ratio was reduced (Fig.  3C). This measurement can be used as a rough measure of stable RNA since total RNA consists of 98% stable RNA (36). In addition, the levels of RNAP ␣-subunit were not affected by the reduced 70 levels (data not shown).
Lowering the levels of E 70 also affected the growth rate (Fig. 4A), as has previously been shown for ppGpp overproduction (9,10). In the case of 70 underproduction no steady state of growth could be obtained, but rather the growth rate decreased gradually over time (Fig. 4A). The 70 underproduction levels were therefore checked at later times in exponential phase and showed that the level of underproduction did not 2 Miki Jishage, unpublished results. change within the range of the experiment (data not shown). The effects of 70 underproduction were studied at two different absorbance values during growth (A 420 ϭ 0.5 and 0.7), and no significant differences in the protein expression between the two absorbance values were found (data not shown). Finally, the 70 underproduction was performed in both a relA1 and a relA ϩ strain with the same result (data not shown). 70 Underproduction in a ppGpp 0 Gives Repression of Stringently Controlled Proteins-The specific effect on ribosomal proteins is not ppGpp-dependent since the down-regulation of ribosomal proteins was obtained also during 70 underproduction in a ppGpp 0 strain (Fig. 5A). The global effects of 70 underproduction were less pronounced, but the same trend was seen as in the wild type strain (comparison of Fig. 5A with 3A). In addition, the effect of 70 underproduction on growth rate (Fig. 4B) was not ppGpp-dependent. Moreover, the level of underproduction of 70 was the same in the ppGpp 0 as in the wild type (data not shown), and the levels of RNAP ␣-subunit were not affected in the ppGpp 0 strain. In agreement with the effect on ribosomal proteins the RNA/protein ratio was reduced (Fig. 5B). DISCUSSION In this work we demonstrate that a strain unable to produce ppGpp is largely decontrolled upon the entry to stationary FIG. 1. Effects of carbon starvation in a wild type strain and in  a ppGpp 0 strain. Cells were grown into starvation in defined media, and samples were labeled in exponential phase (A 420 ϭ 0.5) and stationary phase (20 min after transition). A, location on the NEPHGE reference two-dimensional gel of the proteins analyzed in this work. The production of all these proteins depends on the housekeeping factor, 70 . B, two-dimensional gels of proteins produced in the wild type strain during growth (Exp.) and in stationary phase (Stat.). C, twodimensional gels of proteins produced in the ppGpp 0 strain during growth (Exp.) and in stationary phase (Stat.). D, the effects on a subset of proteins in response to carbon starvation in a wild type strain. The

FIG. 2. The global effects on protein production by entering stationary phase in a wild type strain (A) and in a ppGpp 0 strain (B).
The graph was made the same way as in Fig. 1. The set of proteins in A and B were sorted depending on the magnitude of the response not on the protein identity. Therefore the proteins are not necessarily in the same order in A and B.
dark-gray bars indicate the ribosomal proteins, and the light-gray bars indicate a set of proteins that are non-ribosomal but under control of 70 . The y-axis shows the fold change in a 2 Log scale which means that 1 is a 2-fold induction, while Ϫ1 is a 2-fold repression of the production of a protein. The fold change is calculated by dividing the quantity of the specific protein produced during 5 min in stationary phase (normalized to total protein production) by the quantity of the specific protein produced during 5 min in exponential phase (normalized to total protein production). E, the effect on the proteins shown in C) in a ppGpp 0 strain entering stationary phase. phase and its proteome appears "locked" in a growth mode. By comparing the effects of a wild type and a ppGpp 0 strain going into stationary phase (Fig. 1) we can define the significant changes in the proteome mediated by the stringent response under these conditions. The ribosomal proteins are, as expected, down-regulated by the entry into stationary phase, and this effect is dependent on ppGpp (Fig. 1, C and D). Some proteins, e.g. UspA, are induced in a ppGpp-dependent manner as has been shown previously (10,11). A few amino acid biosynthesis proteins (e.g. the proteins encoded by carB, gltD, ilvE, metC, metE, metH, and tyrB) are identified on NEPHGE two-dimensional gels (33), but those were not among the proteins significantly affected by the stringent response to carbon starvation. In a recently published study using macroarrays, the only genes encoding amino acid biosynthesis proteins that were induced by the stringent response were the ones involved in the histidine and arginine biosynthetic pathways (37). Thus, it appears that promoters requiring ppGpp do not necessarily behave identically during a stringent response as demonstrated by the fact that the production of amino acid biosynthetic gene products, in contrast to the Usp family proteins, are not induced during carbon starvation induced stringency.
We also show that it is possible to mimic the down-regulation of ribosomal proteins seen in the stringent response by under-producing the housekeeping factor, 70 (Fig. 3A). The expression of ribosomal proteins has been shown to depend on the availability of rRNA, and the ribosomal protein promoters have also been shown to be under direct stringent control (2)(3)(4). Thus, the effects seen on ribosomal proteins in the 70 underproduction experiment could be due to effects on rRNA production via transcription of rrn promoters causing a subsequent feedback control of the expression of genes encoding ribosomal proteins. However, we can not rule out the possibility of direct effects on the transcription from ribosomal protein promoters. The effects seen are most likely not post-transcriptional since the ribosomal protein transcripts were similarly affected by a reduction in 70 -programmed RNAP. 2 It should also be noted that the total RNA/total protein ratio decreased during 70 underproduction demonstrating that rRNA levels decreased. The data indicate that among the genes expressed during exponential growth of E. coli the expression of rRNA and ribosomal protein genes is more sensitive than the average to a reduction in RNAP availability. The results support the idea of stringent promoters requiring high levels of free RNAP, as suggested by Jensen and Pedersen (16). In fact, rrn promoters may not work at their maximal capacity even during logarithmic growth since data from Squires and co-workers (38) showed that a cell containing only one rrn operon instead of seven was still able to produce 56% of the wild type rRNA.
Our results of 70 underproduction also go in line with the model of Jishage et al. (20) who suggested that one mechanism of ppGpp-dependent gene regulation is to affect the relative competitiveness of factors, possibly by affecting the affinity of 70 to the core enzyme. The lowering of 70 interaction with RNAP increases the opportunities for alternative factors to bind RNAP core enzyme and redirect RNAP to their respective promoters, and at the same time it lowers the availability of RNAP holoenzyme for the 70 -dependent promoters. If the mechanism behind the stringent control is RNAP availability, either by changing the pool of free RNAP or by affecting factor competition or both, the constitutively stringent mutations in RNAP (e.g. (15)) should also affect these parameters. One interesting feature of the stringent rpoB mutants characterized by Zhou and Jin (15,39) is that they express lower levels of 70 . The stringent rpoB alleles had effects on transcription in vitro so the lower level of 70 expressed by these cells is not the only explanation for their action. But, considering the results presented here, it is possible that the lower 70 levels of the mutants fortify the stringent phenotype in vivo.
The fact that lowering the availability of 70 -programmed RNAP has effects that mimic the stringent response does not, per se, prove that a reduction of 70 -programmed RNAP is part of a bona fide stringent response, but it is possible to speculate on such a mechanism. Even if changing the availability of 70 -programmed RNAP is one effect of ppGpp in the cell, direct positive effects of ppGpp on expression of specific genes/proteins has been shown previously by Choy (40) using a coupled in vitro transcription-translation system. It is possible that the RNAP availability works together with more direct effects of ppGpp on specific promoters.
Lowering the levels of 70 -programmed RNAP also affects growth rate (Fig. 4, A and B), as has previously been shown for cells overproducing RelA and consequently ppGpp (9,10). It could be argued that the effects seen on the protein synthesizing system are merely an effect of the growth rate-dependent regulation of these particular genes. But, then the question remains how 70 underproduction mechanistically causes a reduced growth rate, since most gene products, with the exception of for example the ribosomal proteins, were unaffected by the reduction in E 70 . Maybe the lowered growth rate is a result of the lower levels of the protein synthesizing system due to lower availability of free 70 -programmed RNAP. The rrn promoters and ribosomal protein promoters are affected before the other promoters because they are, in this model, sensitive to changes in free RNAP. It has long been known that the stringent response and growth rate-dependent control have many effectors in common, but it is still controversial whether the one system, ppGpp and stringent control, is involved in the other, growth rate control. RNAP availability could account for growth rate-dependent regulation of gene expression whether or not the signal controlling RNAP levels is ppGpp.
In this article we show that the ppGpp 0 mutant appears locked in a growth state even after entering stationary phase due to carbon starvation. We also show that it is possible to mimic a stringent down-regulation of ribosome production by underproducing 70 . We suggest that the effects on expression of ribosomal proteins by lowering the levels of 70 indicate that the ribosomal protein promoters are sensitive to the availability of RNAP. Thus, it is formally possible that the stringent response could encompass a mechanism of lowered availability of 70 -programmed RNAP. Further experiments will address the question of whether available, transcription-competent, RNAP levels change significantly in response to ppGpp accumulation.