To study the effects of the effector molecule (p)ppGpp on a genome-wide scale the transcriptomes of a rel-proficient and a rel-deficient strain were compared by hybridization analysis using whole-genome DNA microarrays of C. glutamicum 9. Therefore, both strains were grown in liquid LB complex medium and cell samples were taken at an O.D.₆₀₀ of 0.2. RNA isolation from C. glutamicum, labeling of probes and DNA microarray hybridization, including statistical analysis, were performed as described recently 9. The spot intensities were measured and compared between RES167Δrel (int_rel) and RES167 as control (int_cont). Each experiment was performed four times, using two biological and two technical replicates including a dye-swap. In accordance with the previously determined technical variation of the C. glutamicum DNA microarray employed, only genes with ratios of int_rel/int_cont greater than 2^0.6 or less then 2^-0.6 were considered as being differentially expressed. For these ratios the significance value was determined as 0.95, meaning a 95% chance that the gene is differentially regulated. In addition, a Students t-test was applied to filter the data set and only data with an error probability of p = 0.05 or better were used.

Figure 1 shows the results of this experiment depicted in a scatter plot. A total of 357 genes was found to be differentially expressed in the Δrel mutant, of which 217 were transcribed stronger and 140 weaker than in the rel-proficient strain. Among the most prominent induced genes are uspA3 (putative universal stress protein), and the sigE gene (sigma factor of the extra-cytoplasmic function (ECF) family). The genes sigD (encoding another ECF-type sigma factor) as well as rpf1 and rpf2 (encoding proteins involved in bacterial cell-cell communication, 10) are among the genes with the most strongly reduced expression.

Since global gene expression patterns were very different in the rel-proficient and the rel-deficient strain, and sigma factors responsible for global gene regulation are also affected, it was difficult to dissect primary effects of the rel mutation from secondary effects caused by differentially expressed regulatory genes. In order to overcome these problems, a second experimental strategy was employed to describe rel-dependent and rel-independent effects on global gene expression in which the stringent response was induced by the chemical DL-serine hydroxamate (SHX). The strategy consisted of two steps. First, the transcriptional effects of SHX addition on gene expression of stringently controlled genes were followed in a time course to determine the time point at which a maximal first order effect is exerted on gene expression. Then, microarray hybridizations were performed comparing the transcriptomes of both strains before and after addition of SHX for the dissection of rel-dependent and rel-independent effects during stringent response.

In previous experiments, it was demonstrated that the C. glutamicum RES167Δrel mutant grows poorly in liquid minimal medium without the supplementation of histidine and serine 5. Apparently, the biosynthesis pathways of these amino acids are under control by a positive stringent response mediated by the Rel protein. In order to analyze transcriptional regulation of all genes involved in histidine and serine biosynthesis and to assess the timing of the effects of the stringent response on gene expression, sensitive real-time RT-PCR analyses on all serine and histidine biosynthesis genes were carried out in the rel-proficient C. glutamicum RES167 strain. Therefore cultures of C. glutamicum, growing exponentially in liquid LBG medium, were treated with 50 mM SHX. Samples were taken before addition of SHX (t₀) and in five minute intervals up to a total of 25 minutes (t₅ to t₂₅). Total RNA was isolated from these samples and mRNA levels of the nine known histidine biosynthesis genes hisED, hisDCB and hisHAFI and the three serine biosynthesis genes serA, serB and serC was determined. For each time point after treatment with SHX, the relative mRNA level compared to the t₀ value was calculated (Figure 2).

The results clearly showed an increase of mRNA levels of some of the tested genes already five minutes (t₅) after SHX addition, with all being increased and with a maximum of induction after ten minutes (t₁₀). At this time, mRNA levels of the histidine genes are increased twofold to threefold and serine genes were transcribed stronger by factors of 3 to 7.8. In later stages relative mRNA levels slowly decreased until they reached the initial values at around 25 minutes (t₂₅). It is apparent that the histidine genes, which are localized on the chromosome in operon-like structures, are transcribed in similar patterns, supporting the idea that these genes are organized in the three operons hisEG, hisDCB and hisHA-impA-hisFI. The strongest induction was observed at the hisC and hisH genes (Figure 2). Among the serine biosynthesis genes, serA coding for D-3-phosphoglycerate dehydrogenase, the first enzyme of the serine biosynthesis, was the gene whose mRNA level was induced most strongly.

These elevated mRNA levels after addition of SHX were not detected in the Rel-deficient strain C. glutamicum RES167Δrel (data not shown), supporting the interpretation that this is the cause for its requirement of histidine and serine as culture supplements to achieve normal growth 5.

To determine the genes directly affected by (p)ppGpp in C. glutamicum, exponentially growing cultures of the rel-proficient strain C. glutamicum RES167 and the rel-deficient mutant strain RES167Δrel carrying a deletion in the (p)ppGpp synthase gene were treated with SHX. The comparison of global gene expression patterns in the rel-proficient and the rel-deficient mutant strain at inducing and non-inducing conditions should allow to define the regulatory networks involved in the stringent response. In addition, comparison of the data sets enabled the identification of those genes that are controlled by the action of the Rel protein (Rel-dependent stringent control) and of those which are not dependent on the action of the Rel protein (Rel-independent control).

Two C. glutamicum cultures were grown in liquid LBG medium up to an O.D.₆₀₀ of 0.2, at which SHX was added to a final concentration of 50 mM. For each strain, untreated samples (t₀) and treated samples were taken ten minutes after SHX addition (t₁₀) and the transcriptomes were compared as described above.

In Figure 3 the results of the microarray experiments are shown in ratio/intensity (m/a) plots for both the rel-proficient and the rel-deficient strain. In the C. glutamicum rel-proficient strain transcription of 60 genes was found to be downregulated and of 31 genes to be upregulated under inducing conditions. The most prominent upregulated transcription was detected in genes which are involved into stress response, among them katA (catalase) and dps (DNA protection during starvation protein). The genes with the strongest downregulation are the tagA1 gene (DNA-3-methyladenine glycosylase) involved in DNA repair or belonging to nitrogen metabolism including hmp (putative NO-detoxification flavohemoprotein), glnK (nitrogen regulatory protein PII), gltB (glutamate synthase large subunit) and amt (ammonia transporter).

In the m/a plot of the rel-deficient strain (Figure 3B) it is shown that the genes with the strongest upregulation are again katA, gnd (6-phosphogluconate dehydrogenase) and thiol stress-related genes. Among the latter, the sufD and nifS2 genes are involved in the repair of damaged iron-sulfur clusters in proteins whereas cysK encodes cysteine synthase. The strongest downregulation is observed for genes encoding ribosomal proteins (e.g. rpsE, rplO, rpmD).

The data obtained on the differentially regulated genes were grouped into three classes: class A genes are under Rel-dependent stringent control and were found to be differentially expressed only when the untreated and SHX-treated cultures of the rel-proficient RES167 strain were compared and therefore are regulated by (p)ppGpp. Class B genes are under positive or negative rel-independent control and show a change in expression in RES167 and its rel-deficient derivative RES167Δrel. In class C those genes were placed which showed a change in expression only when comparing the untreated and the SHX-treated rel-deficient mutant strain. By this classification scheme, the total number of 243 genes found to be differentially expressed in one or both experiments were ordered as follows: 42 genes in class A, 49 in class B and 152 genes in class C (Figure 3C). For reasons of clarity only the classes A and B are presented here in the form of tables (Table 1, 2) and, within the tables only genes which are annotated at least in the form of a COG classification (Clusters of Orthologous Groups of proteins, 11) are listed. A full list is available as supplementary Table S1 [see Additional file 1].

The hallmark of the stringent response is the downregulation of transcription of the genes encoding ribosomal RNA in a (p)ppGpp-dependent manner. In the microarray experiments, a number of genes displayed a similar change in expression dependent on the presence of the alarmone (p)ppGpp synthesized by the Rel protein. These genes can be grouped to functional complexes according to their annotation and the COG classification scheme (Table 1). As the dominant complex, nitrogen metabolism is apparently affected in C. glutamicum in a negative, Rel-dependent fashion. In particular the transcription of genes belonging to the AmtR repressor regulon 12 were found to be downregulated (codA, gdh, glnA, glnK, gltB, amt-ocd-soxA, ureA, urtB-urtC) (Table 1). In addition, two other genes encoding oxidoreductases (betB, cg3370 ; COG class C) are found to be repressed in the presence of (p)ppGpp. Others are three genes encoding ribosomal proteins (rpmG, rpsF, rpsS) and one gene cluster involved in nucleotide biosynthesis and DNA repair (pyrG-cg1607, Figure 4).

The group of genes the transcription of which is positively affected in a Rel-dependent manner also comprises those encoding proteins involved in redox processes (qcrB-qcrC-qcrA1-ctaE, Figure 4). Other genes are involved in carbon metabolism (ptsF, pyk, ldh) or regulatory processes on protein (ptpA, clpP1, clpC) or mRNA level (lexA, sigB). Especially sigB is interesting since it encodes the alternative sigma factor (σ^B) in C. glutamicum which is involved in global gene regulation at the transition from exponential growth to the stationary phase and in the general stress response 1314.

The most important members of this class that are regulated in a negative Rel-independent manner are genes coding for ribosomal proteins (COG class J: translation), of which 26 were found to be downregulated under inducing conditions (Table 2, Figure 5). Additionally, transcription of one other gene from the translational apparatus (infA) and of three genes involved in transcription or DNA replication (COG classes K and L) was repressed in both strains. As a further functional complex, the transcription of the putative nitrate reductase and transport genes (narK-narG-narH-narJ-narI) was found repressed.

In a Rel-independent way, transcription of genes involved in oxidative stress response (katA encoding catalase, dps encoding a DNA protection protein) was upregulated not only in the rel-proficient, but also in the rel-deficient mutant strain of C. glutamicum (Table 2). Additionally, the transcription of genes from the suf/nif gene cluster (sufB-sufD-nifS2) involved in the repair of Fe-S clusters damaged by oxidation 15 was induced under these conditions (Figure 5).

A large number of genes were found to be regulated after treatment with SHX only in the rel-deficient mutant but not in the rel-proficient strain (class C). Within this class additional genes belonging to the translational (COG class J) and transcriptional apparatus (COG class K) were grouped, in total, 18 genes encoding ribosomal proteins, infB encoding translation initiation factor 2 and prfB (peptide chain release factor 2), as well as rpoB (β subunit of RNA polymerase) as well as cg2177-nusA encoding proteins involved in transcription termination. As illustrated in an example (Figure 5) the transcription ratios of several ribosomal genes in the rel-proficient strain experiment are just below the significance value. Taken into account that most of these genes are organized in operons, it can be assumed they rather belong to class B than to class C.

The expression of genes involved in oxidative stress response has been found to be upregulated only in the rel-deficient mutant (sod encoding superoxide dismutase, cg2674 encoding a putative alkyl hydroperoxidase and tpx encoding a thiol peroxidase). Especially thiol oxidation stress is apparent since also the expression of the fpr2-cysI-cysH-cysD gene cluster for assimilatory sulfate reduction 16 as well as cysK encoding cysteine synthase are enhanced.

It is apparent that a number of transcriptional regulatory genes fall into class C. Among these are cg1792 encoding a WhiA homologue, whiB2 and whiB4, cg0725 encoding a MarR-family regulator, clgR encoding a stress-induced proteolysis regulator 17, phoU encoding the putative phosphate uptake regulator, and cg3303 encoding a regulator of the PadR family. All these genes were found to be repressed under stringent conditions in the rel mutant. Another group of genes encoding transcriptional regulators is found to be expressed stronger under these conditions, namely cg1631 encoding a MerR family regulator, dtxR and fur, encoding regulators implicated in metal uptake and homeostasis, cg2500 encoding an ArsR-type regulator, cg2614 encoding a TetR-type regulator, and cg3315 encoding a regulator of the MarR family 1819. This is indicative that cellular metabolism is not well balanced in the mutant but on the other hand this deregulation might cause secondary effects on global gene expression patterns not directly associated with the stringent response.

In this work, C. glutamicum RES167, a restriction and modification-deficient derivative of the wild-type strain ATCC 13032 51 and C. glutamicum RES167Δrel 6 were used. The strains were cultivated at 30°C in Luria-Bertani medium 52 supplemented with 2 gl^-1 glucose (LBG). For transcriptome analysis the cultures were grown to an O.D.₆₀₀ of 0.2 and then harvested. In the induction experiments, the cultures were divided into two aliquots, one of which was treated with DL-serine hydroxamate (final concentration 50 mM) to induce the stringent response.

Approximately 1 × 10⁹ cells were harvested from C. glutamicum cultures and immediately frozen in a dry ice/ethanol bath. The cells were thawed and centrifuged at 6.000 g for 60 seconds. The supernatant was decanted and the pellet was resuspended in 800 μl RLT-buffer (RNeasy Mini Kit, Qiagen) and disrupted by means of a Rybolyser Instrument (Hybaid, Heidelberg, Germany). The following clean-up of RNA along with DNase I digestion was performed as described 9.

The labeled cDNA was synthezised according to 9, as were the DNA microarray experiments and data analysis. The ImaGene 5.0 Software (BioDiscovery, Los Angeles, CA) was used for spot finding, signal-background segmentation and intensity quantification. Normalization and t-test statistics were carried out with the EMMA microarray data analysis software 53. According to the validation experiments 9 all genes with detected gene expression ratios greater than 2^0.6 or smaller than 2^-0.6 were regarded as differentially expressed with a confidence level of at least 95%. The additional Students t-test statistics were performed with a probability cutoff of p = 0.05.

One-step real-time RT-PCR reactions were performed by using the LightCycler instrument (Roche Diagnostics) and the QuantiTect SYBR Green RT-PCR Kit (Qiagen) according to the manufacturer's instructions. To verify the purity of the obtained PCR products, melting curve analyses were performed. Differences in gene expression were calculated by comparing the crossing points of the samples with the help of the LightCycler software (version 3; Roche Diagnostics).