Highly Precise Quantification of Protein Molecules per Cell During Stress and Starvation Responses in Bacillus subtilis

Abstract

Systems biology based on high quality absolute quantification data, which are mandatory for the simulation of biological processes, successively becomes important for life sciences. We provide protein concentrations on the level of molecules per cell for more than 700 cytosolic proteins of the Gram-positive model bacterium Bacillus subtilis during adaptation to changing growth conditions. As glucose starvation and heat stress are typical challenges in B. subtilis' natural environment and induce both, specific and general stress and starvation proteins, these conditions were selected as models for starvation and stress responses. Analyzing samples from numerous time points along the bacterial growth curve yielded reliable and physiologically relevant data suitable for modeling of cellular regulation under altered growth conditions. The analysis of the adaptational processes based on protein molecules per cell revealed stress-specific modulation of general adaptive responses in terms of protein amount and proteome composition.

Furthermore, analysis of protein repartition during glucose starvation showed that biomass seems to be redistributed from proteins involved in amino acid biosynthesis to enzymes of the central carbon metabolism. In contrast, during heat stress most resources of the cell, namely those from amino acid synthetic pathways, are used to increase the amount of chaperones and proteases. Analysis of dynamical aspects of protein synthesis during heat stress adaptation revealed, that these proteins make up almost 30% of the protein mass accumulated during early phases of this stress.

Recently technical approaches in systems biology have become more and more important for the life science community. Successful modeling of biological pathways as part of these approaches strongly depends on quantitative, high-quality, and validated data sets (1). Proteins are an important part of these attempts to uncover the systemic properties of biological systems as they represent the central players in the complex cellular metabolic and adaptational network (2).

Although relative protein quantification methods allow for comparison of protein abundances in samples and to characterize the proteome dynamics in cellular systems, these data are not sufficient for mathematical modeling in systems biology. Furthermore, the availability of protein concentrations at the proteome level can provide new insights in what is going on in the cell upon stress, and thus enable us to better understand how cells adapt to changing conditions. Knowing intracellular protein concentrations is essential in order to obtain a real mass balance leading to evaluation of the costs of running an active metabolic pathway or expressing enzymes for stress responses. In order to provide suitable proteomic data for systems biology, techniques for global absolute quantification of proteins recently emerged. These approaches make use of quantitative Western blotting (3), mass spectrometry (4, 5), or merge traditional two-dimensional polyacrylamide gel electrophoresis (two-dimensional PAGE)¹ and mass spectrometry (6) to determine cellular protein concentrations on a global scale. Entirely mass spectrometry-based strategies have recently convincingly demonstrated the capacity to quantify about half of the predicted proteome of Leptospira interrogans (5) and therefore provide more comprehensive data for systems biology than two-dimensional PAGE based methods (6, 7). Although these gel-based methods are biased toward high-abundant proteins, usage of MS-calibrated two-dimensional gels allows distinguishing even between different protein isoforms caused by post-translational modifications without missing values along a time course experiment. Moreover, two-dimensional PAGE is a well-established method for easy and convenient visualization of main metabolic pathways and the most obvious adaptational responses.

Until now, approaches for large-scale absolute protein quantification have had a strong technological focus. Only a few predominantly physiological applications have been reported (5, 8, 9).

In this study, we provide absolute protein concentrations of the bacterial model organism Bacillus subtilis during stress and starvation adaptation. As model for different stress conditions, we selected the best-studied responses during heat stress and glucose starvation. A comparative analysis of different stress conditions allows to differentiate between general, nonspecific adaptive responses ensuring survival during a wide spectrum of conditions and specific stress adaptation by differential expression of particular regulons facilitating a direct interaction with the stimulus (10).

The general adaptive response of B. subtilis is governed by global regulators such as the alternative RNA polymerase sigma factor Sigma B or the main stringent factor RelA.

Induction of SigB-dependent genes provides cells with a multiple, nonspecific and preventive stress resistance (11, 12). The SigB-dependent general stress response in B. subtilis is induced by a different set of stress and starvation stimuli. Thereby, environmental stresses, like heat shock, activate the phosphatase RsbU via a signal transduction pathway that involves additional regulatory proteins to dephosphorylate the anti-anti-sigma-factor RsbV that subsequently releases SigB (13, 14). During starvation for carbon sources, phosphorus or oxygen dephosphorylation of phosphorylated RsbV followed by the release of active SigB is catalyzed by RsbP (15).

The main (p)ppGpp synthetase RelA mediates the stringent response, which is a crucial component of the regulatory network in B. subtilis cells. The main feature of the stringent response is the down-regulation of genes whose products typically consume energy and building blocks for cell growth, particularly genes encoding components of the transcription and translation apparatus as well as genes coding for proteins involved in nucleotide biosynthesis and DNA replication (16).

During glucose starvation, a substantial reprogramming of protein synthesis pattern is caused by the restricted access to energy and carbon sources. The carbon starvation specific stress response is characterized by repression and degradation of glycolytic enzymes with simultaneous up-regulation of specifically gluconeogenesis and tricarboxylic acid cycle related enzymes (10, 17–19). Additionally metabolic pathways for the utilization of overflow metabolites and other secondary carbon sources such as α- or β-glucosides or amino acids are induced (10, 17).

In addition to starvation B. subtilis has to adjust to various physical stresses in its natural habitat. In order to investigate this adaptational response we have chosen the model heat stress. Already described specific heat-shock induced genes of B. subtilis belong to the HrcA regulon (20), the CtsR regulon (21–23), or the HtpG operon (24).

Moreover, there are heat inducible stress proteins whose regulation does not follow the already mentioned mechanisms. These proteins are for example FtsH (25), ClpX (26), SacB (27), LonA (28), AhpC, AhpF (29), NfrA, YwcH (30), and members of the SigI regulon (31).

In this study, we quantitatively characterize the adaptation of B. subtilis to glucose starvation and heat stress and present concentrations on the level of molecules per cell for more than 700 cytosolic proteins. Furthermore, we analyzed dynamical protein repartition between main processes of the cell during exponential growth and stress providing valuable insights in adaptation to changing conditions. Thus, this comprehensive data set may be suitable for modeling of metabolic pathways.

EXPERIMENTAL PROCEDURES

Pulse-labeling with l-[³⁵S]-methionine

Radioactive pulse-labeling experiments with ³⁵S-methionine were carried out to investigate protein synthesis of B. subtilis during glucose starvation. Therefore three independent biological replicates of B. subtilis 168 trp+ (32) were grown in M9 minimal medium at 37 °C under vigorous agitation. The composition of M9 was: 0.1% (w/v) glucose, 0.1% (w/v) L-malate, CaCl₂ * 2H₂O (0.015 g L⁻¹), MgSO₄ * 7H₂O (0.246 g L⁻¹), Na₂HPO₄ * 2H₂O (8.5 g L⁻¹), and KH₂PO₄ (3 g L⁻¹), NaCl (0.5 g L⁻¹), NH₄Cl (1 g L⁻¹) and trace elements (10 ml stock solution for 1 L medium). The trace element stock solution consisted of (per liter): ZnCl₂ (0.170 g), CuCl₂ * 2H₂O (0.043 g), MnCl₂ * 4H₂O (0.100 g), CoCl₂ * 6H₂O (0.060 g), and Na₂MoO₄ * 2H₂O (0.060 g). Cultures were pulse-labeled for 5 min each with 10 μCi of L-[³⁵S]-methionine per ml at an OD_{600 nm} of 0.4 - 0.5 (for control), at maximal OD and at several time points (15, 30, 60, and 120 min) after the transition to glucose starvation as described previously (10). In combination with protein amounts derived from fluorescence stained two-dimensional gels (see below) it can be investigated which proteins are actually synthesized at a specific time point and which protein amounts have already been accumulated.

Cell Culture and Protein Preparation

For starvation experiments B. subtilis 168 trp+ (32) was cultivated as mentioned above. Samples were harvested in exponential phase (OD_{600 nm} 0.5), in transient phase, at maximal OD and 60, 120, 180, and 240 min after entering the stationary phase triggered by glucose exhaustion (Fig. 1). For heat stress experiments, cells were cultivated in minimal medium (33) and stressed by a sudden temperature shift to 52 °C. Cells were harvested shortly before (control) and at 10, 30, and 60 min after continuous heat shock (Fig. 1). For every experiment, three independent biological replicates were analyzed.

Fig. 1.

Representative growth curves of B. subtilis during glucose starvation and heat stress. Growth of bacterial cells was followed by measuring optical density (OD) at 600 nm. Sampling points are indicated by dots. Sampling occurred as following: glucose starvation: exponential growth, transient, maximal OD, 60, 120, 180, and 240 min stationary phase, heat stress: control (OD_{600 nm} = 0.5), 10, 30, and 60 min 52 °C heat. Phases of starvation or heat stress are indicated by the gray area.

After harvesting and centrifuging cells were resuspended in TE-Buffer (10 mm Tris, 1 mm EDTA, pH 7.5) and mechanically disrupted using the Precellys 24 homogenizator (PeqLab, Germany; 3 × 30 s at 6.5 m s⁻¹). Protein concentration of extracts was determined using a ninhydrin-based assay (6, 34).

Preparation of Two-dimensional Gels

Two-dimensional PAGE was performed as previously described (35) in five technical replicates. 100 μg protein was loaded onto 18 cm IPG strips (pH 4–7, GE-Healthcare). After two-dimensional PAGE gels were fixed with 40% (v/v) ethanol and 10% (v/v) acetic acid for 1 to 2 h and subsequently stained with Flamingo^TM. Stained gels were scanned (Typhoon 9400, GE-Healthcare) and their images analyzed employing Delta2D 4.2 software (Decodon GmbH, Germany). For all spots detected on the gel the spot volume was assigned to proteins, exported from the software and subsequently used for calibration of two-dimensional gels as described earlier (6).

Sample Preparation for MS Analysis

Protein samples were reduced, alkylated and digested as previously described (6). Samples were spiked with heavy peptides of anchor proteins used for calibration of two-dimensional gels to a final concentration of 5–25 fmol μl⁻¹. A detailed list of used peptides and their optimized transition parameters has been published elsewhere (6).

Targeted MS Analysis and Global Absolute Protein Quantification

LC-MS analyses was performed as described previously (6). All raw files were processed using MultiQuant^TM 1.1 software (Applied Biosystems, Foster City, CA). A peptide ratio of native and heavy species was based on three transitions that were weighted according to their signal to noise (S/N) ratios before being averaged. Accordingly, S/N weighted peptide ratios were combined to the final protein ratio. Based on the added amount of heavy peptides, the absolute quantity of target anchor proteins could be calculated.

Absolute amounts of targeted anchor proteins obtained by SRM were used to calibrate two-dimensional gels of the same sample in order to obtain absolute abundance of all proteins visible on this gel. Amounts of multiple spots of the same protein were added up. Standard deviations for proteins represented by multiple spots were calculated using error propagation. Final standard deviation was calculated using a random effect model (36).

Efficiency of Cell Disruption and Determination of Cell Size

Bacterial cell size and cell disruption efficiency were determined as previously published (6). Volumes of the rod-shaped cells were calculated assuming a cylinder and two hemispheres without subtracting any values for the cell envelope. At least 100 cells were dimensioned for every sample. Therefore, a standard deviation could be calculated for each analyzed population (supplemental Table S6).

RESULTS

Determination of absolute protein abundances by combining accurate targeted mass spectrometry with the resolving power of two-dimensional PAGE provides a global view on the concentrations of a large number of proteins. Although the derived quantification data across biological replicates can be at most a highly accurate determination of molecular counts we will use in the following the common term “absolute quantification.”

In this study we present a comprehensive proteomic data set applicable for bioinformatic modeling of B. subtilis' stress and starvation responses. As models of starvation and stress responses we chose glucose starvation and heat stress. Therefore we determined protein concentration at seven different time points along the bacterial growth curve for cells under glucose starvation and at four time points during heat stress (Fig. 1). Whereas glucose starvation causes a complete stop of cell growth, 52 °C heat stress leads to a drop of growth rate from 1.2/h during exponential growth to 0.5/h during heat stress.

Absolute Quantification of Proteins

After preparation of a cell count calibrated protein sample by determination of cell titer, cell disruption efficiency and protein content according to previously published protocols (6, 34), isotopically labeled peptides of anchor proteins were spiked in the sample in known concentrations. Digestion and targeted SRM acquisition led to determination of protein concentration for all anchor proteins in the total protein sample. Hence, the absolute amount of anchor protein on fluorescently stained two-dimensional gels prepared from the same sample could be calculated and was used for calibration of these two-dimensional images by relating spot intensities to the ones of anchor proteins (6). This enabled determination of protein concentration for all proteins detectable on the two-dimensional gels. Detectable soluble cytosolic proteins have an isoelectric point between 4 and 7 and a molecular weight of 10–150 kDa. In order to provide reliable data, only anchor proteins with less than 15% CV among four technical replicates of MS analysis and less than 35% CV between five technical two-dimensional gel replicates were used for calibration of two-dimensional gels.

Noncovalently binding fluorescent dyes (e.g. Flamingo, Krypton) bind proteins in an amount proportional to the protein size (6) so that the same molecular count of a small protein correlates with a lower signal compared with that of a larger protein molecule during detection. Taking this into account the detection of the same count of large protein molecules should be more sensitive compared with smaller ones, which is supported by experimental data (Fig. 2). Accordingly, the molecular weight of every protein has to be considered in the calculation of the number of molecules per cell on the basis of protein concentrations.

Fig. 2.

Limit of detection is a function of molecular weight. Because of sequence-unspecific noncovalent binding of fluorescent dyes the same molecular count of a small protein generates a smaller signal compared with that of a larger protein. Molecules per cell for every protein quantified in this study were plotted against their molecular weight to calculate the limit of detection. The equation as well as the coefficient of determination are indicated.

In this study we were able to identify 783 proteins (465 proteins in starvation and 701 proteins in heat stress experiments), of which 773 proteins (465 proteins in starvation and 691 in heat stress experiments) could be reliably quantified for at least one time point (supplemental Table S1, S2). The amount of 219 and 287 proteins changed significantly (p = 0.05, one-way ANOVA) during glucose starvation and heat stress, respectively (Fig. 3, supplemental Figs. S1, S2). Supplemental Fig. S3 shows the distribution of these proteins according to different regulation thresholds.

Fig. 3.

Voronoi-treemaps of B. subtilis during glucose starvation and heat stress. Protein abundances (copies/cell * molecular weight) during control (left) and stress conditions (right). Each cell in the graph displays a protein that belongs to other functionally related elements in parent convex-shaped categories. These are again summarized in higher-level categories (see legend on the right side). Functionally related elements are depicted in close neighborhood to each other and colored similarly. Gene functional data are based on KEGG-orthology. Area size in the graph encodes protein abundance.

During exponential growth cellular concentrations of proteins spans from very few molecules per cell (11 molecules per cell for the DNA mismatch repair protein MutL) to about 250,000 molecules per cell (major cold-shock protein CspB). Considering protein abundances during stress adaptation the proteins determined to have the lowest cellular concentration are the unknown protein YobO under starvation conditions (87 kDa, 13 molecules per cell after 120 min stationary phase) and the DNA exonuclease SbcC under heat stress (128 kDa, 7 molecules per cell after 60 min heat stress). The chaperonin GroES was found to be the most abundant protein under heat stress condition (500,000 molecules per cell after 30 min heat stress), whereas IlvC, an enzyme involved in biosynthesis of branched-chain amino acids, and the elongation factor Tu (TufA) represent the most abundant proteins during glucose starvation (about 20,000 molecules per cell after 60 min stationary phase). Hence, as reported earlier with the method used here a range of cellular protein abundances of 3 to 4 orders of magnitude can be covered (6) (Fig. 3, supplemental Fig. S4).

The availability of absolute quantification data for proteins at a large scale allows calculation of stoichiometries of known oligomeric protein complexes. The already reported ratio of 1:2 for the components of the 2-oxoglutarate dehydrogenase OdhA and OdhB, was verified (6, 37). The stoichiometric ratio for the core complex of the tricarboxylic acid cycle consisting of isocitrate dehydrogenase Icd, malate dehydrogenase Mdh, and citrate synthase CitZ (38) has been determined to 4 molecules Icd : 4 molecules Mdh : 1 molecules CitZ. However, because of Mdh and Icd performing multiple interactions with other proteins of the TCC and coupled pathways, calculated stoichiometries derived from total cytoplasmic protein extracts may not necessarily reflect stoichiometries in the core complex alone.

Sustainment of Basal Metabolic Functions and Cellular Processes during Stress Adaptation

During exponential growth in defined medium the metabolic situation in a bacterial cell is comparatively simple. The bacterium uses its main resources to degrade carbohydrates in order to produce energy and building blocks necessary to build up biomass. Although in nature this situation is a rare exception there are metabolic and regulatory pathways that need to stay active even under changing conditions. Absolute quantification results obtained in this study indicate that bacterial cells sustain basal functions of metabolism and cellular processes needed during exponential growth even after exposure to stress and starvation. About 60% of the top 100 abundant proteins (2.5 × 10⁶ molecules/cell) are present in all conditions examined (Fig. 4, dark blue). Although proteins functioning in the acquisition of iron and the carbon core metabolism are regulated in response to glucose starvation and heat stress, these pathways always stay active. Only few quantified proteins of these respective pathways (less than 2.5%) were found in decreased amounts. Therefore we suggest that these enzymes are strongly needed to ensure the cell's supply with carbon intermediates and iron during all phases of growth. Additionally, the protein amounts of enzymes involved in biosynthesis of serine, glycine, and alanine do not change. Notably, also no enzyme needed for the utilization of branched amino acids was found in lowered amounts in this study, indicating that this pathway, using the most abundant amino acid in proteins, needs to stay active during all conditions tested.

Fig. 4.

100 most abundant proteins at exponential growth, heat stress, and glucose starvation. Protein amounts (molecules/cell) of the 100 most abundant proteins during exponential growth (first bar), after 60 min heat stress (second bar), and after 120 min glucose starvation (third bar) are used to calculate the relative amount of stable and newly accumulated proteins within the 100 most abundant proteins in the cell.

Moreover, no protein with functions in DNA condensation, segregation, repair, or combination was found in lowered amounts indicating that these functions are strongly needed to keep the genetic integrity and enable the cell to respond to stress. Dealing with stress is also ensured by keeping translation and transcription mechanisms active. Hence, more than 80% of the involved proteins quantified in this study were found with stable or even increased amounts.

When cells get stressed the situation described above changes rapidly. Now the focus of the cell is no longer growth, integrity, and supply with resources but survival. Analysis of protein quantification data set presented here revealed that the most pronounced regulons involved in general adaptive responses are the negative stringent response under starvation and the SigB response for cells under heat stress, indicating that quality and intensity of the general stress response differs for stress and starvation conditions. During glucose starvation the cells down-regulated glycolysis and up-regulated gluconeogenesis which is in good agreement with previously published data (10). The specific response to heat stress is mostly characterized by accumulation of cytosolic chaperones and proteases whose genes are controlled by HrcA and CtsR.

However, the impact of the stress stimulus on the protein amount is surprisingly small, only 10.0% and 9.4% of the 100 most abundant proteins (6.1 × 10⁵ molecules/cell and 2.6 × 10⁵ molecules/cell) accumulated in response to heat stress and glucose starvation, respectively (Fig. 4, red, yellow, orange), suggesting an exceptionally high functional efficiency during stress adaptation. The impact of the general stress proteins on the total protein amount is even lower. Only 1% of the 100 most abundant proteins accumulated under both stress conditions (Fig. 4, orange).

General Adaptive Response under Glucose Starvation

When exponential growing cells begin to starve, the needs of the cells change dramatically within a very short time. Although producing biomass has been the main purpose during growth, it is now the demand for energy and metabolic intermediates. In this study, we select glucose starvation as model for starvation conditions. Hierarchical clustering of protein expression patterns during glucose starvation revealed that the cell handles the changed situation by decreasing the amounts of almost 50% of all proteins exerting functions in protein biosynthesis or in biosynthesis and acquisition of amino acids, cofactors, or nucleotides emphasizing the important role of the negative stringent response and related responses under glucose starvation (supplemental Fig. S5). The main feature of the stringent response is the down-regulation of genes typically expressed in growing cells (16). Those genes are involved in transcription and translation, nucleotide biosynthesis, and DNA replication that was also reflected by reduced molecular counts in stationary phase for the ribosomal protein RpsB, the elongation factor FusA, adenylate kinase Adk, phosphoribosylpyrophosphate synthetase Prs, and the single-strand DNA-binding protein SsbA. Additionally, a decreasing protein concentration was observed for cell-shape determining proteins like Mbl or proteins involved in ATP synthesis and respiration, like AtpA and AtpD (Table I). The lower level of proteins necessary in starved cells is caused not only by repression of the corresponding genes, but also by degradation of vegetative proteins no longer active in nongrowing cells (10, 18, 19).

Table I. Determined protein amounts (in molecules per cell) of selected proteins after 240 min stationary phase due to glucose starvation. Absolute amounts for all quantified proteins can be found in supplementary Table S2. *Provided data at time point of maximal induction.

Protein	Function	Molecules/cell
		Exponential growth	Stress
Stringent response
RpsB	Ribosomal protein	16,000	8600
FusA	Elongation factor	67,000	37,000
Adk	Adenylate kinase	11,000	4700
Prs	Biosynthesis of histidine	3500	1300
SsbA	DNA replication	9000	4200
Mbl	Cell shape determination	3100	960
AtpA	ATP synthase subunit	31,000	16,000
AtpD	ATP synthase subunit	30,000	14,000
σB-regulon*
YdbD	General stress protein	190	4000
YdaD	General stress protein	150	1200
YdaG	General stress protein	7600	37,600
YfkM	General stress protein	1600	4400
ClpP	Clp protease proteolytic subunit	24,200	24,500
YvgN	General stress protein	81,500	67,400
YvaA	General stress protein	15,900	14,400
KatE	General stress protein	1300	2700
Ctc	General stress protein	9200	20,000
McsB	Modulator of CtsR-dependent repression	4300	12,300
Glycolysis
Pgi	Glucose 6-P isomerase	35,600	14,400
PfkA	Phosphofructokinase	8100	4300
FbaA	Fructose 1,6-P aldolase	120,000	69,900
GapA	Glyceraldehyde 3-P dehydrogenase	19,300	13,600
Pgk	Phosphoglycerate kinase	28,900	21,100
Pgm	Phosphoglycerate mutase	12,300	7800
Eno	Enolase	140,000	71,000
Gluconeogenesis and acquisition of secondary carbon sources
GapB	Glyceraldehyde 3-P dehydrogenase	1700	4400
PckA	Phosphoenolpyruvate carboxykinase	3700	5200
AcoA	Acetoin dehydrogenase subunit	6300	24,000
AcoB	Acetoin dehydrogenase subunit	5900	13,800
AcoC	Acetoin dehydrogenase subunit	850	2100
AcsA	Acetyl-CoA synthetase	4100	8400
MalA	6-P-alpha-glucosidase	1200	3100
IolD	Myo-inositol catabolism	160	410
TCC
CitZ	Citrate synthase	21,300	34,300
OdhA	2-oxoglutarate dehydrogenase subunit	4000	7900
SucC	Succinyl-CoA synthetase subunit	17,300	28,700
SucD	Succinyl-CoA synthetase subunit	4900	11,500
SdhA	Succinate dehydrogenase subunit	4500	7600
Amino acid synthesis
MetC	Cystathionine beta-lyase	4900	1600
MetE	Methionine synthase	67,400	21,700
ArgC	N-acetyl-g-glutamyl-P reductase	2800	800
ArgD	Acetylornithine transaminase	4300	1400
IlvA	Threonine dehydratase	2400	180
YwaA	Branched-chain amino acid aminotransferase	15,300	5800
LeuA	2-isopropylmalate synthase	7500	3000

Besides the negative stringent response the alternative sigma factor SigB is expected to be a key player of the general stress response, but only 19 proteins whose regulation is controlled by SigB could be quantified during glucose starvation. Quantitative results for 10 of these were statistically significant (p = 0.05, one-way ANOVA). Seven of these changed more than twofold in at least one of seven time points examined. Thereby, the general stress proteins YdbD, YdaD, YdaG, and YfkM showed highest fold changes (Table I). With 14,400 to 81,500 molecules per cell the protease ClpP and the general stress proteins YvgN, YvaA were the most abundant SigB-dependent proteins. Five out of seven significantly changed proteins accumulated not until late stationary phase, namely the general stress proteins YdbD, YdaD, YdaG, YfkM, and the catalase KatE. In contrast, the general stress protein Ctc and the SigB- and CtsR-dependently expressed protein arginine kinase McsB accumulated only transiently. Although Ctc was present in highest amounts during transient phase, McsB accumulated most after 120 min stationary phase caused by glucose exhaustion.

Supporting analysis by ³⁵S-pulse-labeling during glucose starvation allowed relative quantification of the protein synthesis of 37 SigB-dependent proteins. 27 of these changed significantly (p = 0.05, one-way ANOVA) more than threefold in at least one time point examined (supplemental Table S3, supplemental Fig. S6). The induction of general stress proteins peaked 30 min after entry into starvation-triggered stationary phase and reached almost control levels after 120 min of glucose exhaustion. Despite this transient induction only a slight accumulation of a few stress proteins could be observed (Fig. 5).

Fig. 5.

Changes in protein abundance for selected SigB-dependent proteins in growing and stressed cells. Protein patterns of exponential growing (green) and stressed cells (red) of selected SigB-dependent proteins during different growth stages (columns correspond to sampling points mentioned in Experimental Procedures). On the bottom of spot tiles the protein amount is given (molecules per cell). Bar charts show log2 ratios of protein amounts compared with control sample (exponential growth) for starvation (blue) and heat stress (orange). For the starvation experiment ratios for accumulated proteins (light blue) and synthesized proteins (dark blue) are given.

This raises the question how long it takes until changes in protein synthesis become detectable on the level of protein amounts. Therefore, we compared absolute protein quantities during glucose starvation of this study with previously published data on protein synthesis (10). As only proteins with changed expression pattern are of interest for this kind of analysis only common proteins of both studies that show significant changes in protein synthesis (>twofold) were compared leading to 41 induced and 100 repressed proteins. For about half of the induced proteins an increase in protein amount could be detected in the same time point indicating an immediate protein translation and accumulation (Fig. 6). These proteins are involved in genetic information processing like the ribosomal protein paralog Ctc, transcriptional elongation factor GreA and the sigma factor SigB, or are required for the utilization of alternative carbon sources (AcsA, LicH, and AcoB). For other induced proteins accumulation takes at least 60 min (Fig. 6). Functions of proteins that accumulate much later (after at least 240 min) are very diverse and do not follow an obvious direction. 80% of the repressed proteins are stable for more than 240 min after repression of protein synthesis (Fig. 6). Most of these proteins (62.5%) function in carbon core metabolism or biosynthesis of nucleotides, amino acids, and cofactors. Only three proteins were found to be very unstable as their amount decreases at the same time point where the repression was detected. These proteins are Tgt, functioning in translation, CarB, involved in biosynthesis of arginine, and Sat, an enzyme of the sulfur metabolism.

Fig. 6.

Distribution of duration of protein accumulation and depletion after induction and repression of synthesis after glucose starvation. Common proteins of this study and a work on protein synthesis (10), which show significant regulation in protein synthesis (at least twofold) in one of the time points (maximal OD, 30 min, 60 min, and 240 min glucose starvation normalized to exponential growth) were compared. The time point of first regulation on the level of synthesis and the first detection of changes in protein amount were compared and the period of time needed was calculated. Calculated durations were color-encoded in a Voronoi-Treemap. Induced/accumulated proteins are shown in orange, repressed/depleted proteins are colored blue. The faster the change in protein abundance occurred, the darker the color appears.

Specific Response to Glucose Starvation

Growth of B. subtilis in its natural environment, the upper layers of soil, is characterized by alternating phases of glucose supply and limitation. Adaptation to starvation for glucose and other carbon sources is the key for survival of the cells. Hence, after exhaustion of glucose the glycolytic pathway is repressed because of the need of a high glucose concentration for gapA operon expression (39–41). As repression in nongrowing cells would mean nonchanging protein amounts, an additional proteolytic degradation of unemployed glycolytic enzymes probably occurs. This is supported by halved protein concentrations after 180 min of starvation triggered stationary phase in nongrowing cells (Table I, Fig. 3, supplemental Fig. S5). In general, numbers of protein copies per cell differ in metabolic pathways, which is most probably caused by different enzyme efficiencies as a result of varying binding coefficients and metabolic rates (42) (supplemental Fig. S7). This is also true for glycolytic enzymes for which molecules per cell differ with an average factor of 15 (Table I). However, average stoichiometries of constitutively expressed glycolytic enzymes catalyzing reversible reactions seem to remain stable during all time points examined (5 molecules Pgi : 19 molecules FbaA : 5 molecules Pgk : 2 molecules Pgm : 20 molecules Eno, Table I). In contrast, measured stoichiometries between key players of glycolysis (phosphofructokinase PfkA, glyceraldehyde 3-phosphate dehydrogenases GapA and GapB) change during transition into stationary phase caused by starvation (exponential growth: 20 molecules Pgi : 4 molecules PfkA: 10 molecules GapA : 1 molecule GapB; stationary phase: 20 molecules Pgi: 4 molecules PfkA : 4 molecules GapA : 2 molecules GapB_, Table I, supplemental Fig. S7).

Dealing with glucose starvation also requires new accumulation of proteins specifically needed to react to the changed supply of carbon sources. Hence, after glucose exhaustion gluconeogenesis becomes necessary because cells start to use secondary carbon sources like for example overflow metabolites produced during excess of the preferred carbon source. Hence, the protein amount of the gluconeogenic glyceraldehyde-3-phosphate dehydrogenase GapB increased more than twofold in late stationary phase. Moreover, for phosphoenolpyruvate carboxykinase PckA, feeding into gluconeogenesis by converting oxalacetate to phosphoenolpyruvate, an increased amount could be detected (Table I). Additionally, amounts of enzymes for the utilization of secondary carbon sources like the subunits of acetoin dehydrogenase AcoABC increased significantly. This also applies to acetyl-CoA synthetase AcsA, 6-phospho-alpha-glucosidase MalA, and IolD, necessary for the catabolism of acetate, maltose and myo-inositol, respectively, indicating that some CcpA-dependent catabolic genes also seem to be derepressed in glucose-starved cells without any obvious external inducer (Table I, Fig. 3, supplemental Table S4). For AcsA this can be explained by a possible internal inducer as lipid degradation during stationary phase could provide an additional source of acetyl-CoA (17).

During glucose starvation down-regulation of glycolysis and induction of gluconeogenesis occur simultaneously with induction of TCC enzymes (10, 17, 18). Induction of this metabolic pathway allows for utilization of organic acids and free amino acids as energy sources, which might become available because of protein degradation. As expected protein amounts for these enzymes revealed an increased need for citrate cycle intermediates. Hence, molecules per cell for citrate synthase CitZ, 2-oxoglutarate dehydrogenase subunit OdhA, both subunits of the succinyl-CoA synthetase (SucC, SucD), and succinate dehydrogenase subunit SdhA increased at least 1.7-fold when cells starve for glucose (Table I, Fig. 3, supplemental Fig. S5).

Accumulation of new proteins needed for the specific reaction to starvation conditions requires a lot of energy. However, the availability of energy is restricted in a starved cell. Therefore the decreased growth rate and the reorientation of protein synthesis is necessary when energy becomes limited. Investigation of the repartition of protein amounts among the main processes in B. subtilis during adaptation to glucose starvation detected significantly lowered protein amounts in amino acid biosynthetic pathways already at entry into stationary phase. After 180 min stationary phase caused by glucose exhaustion this becomes only more pronounced (supplemental Fig. S5). Most drastic changes in protein amount could be detected for synthetic pathways of methionine, arginine, and branched amino acids. Hence, the amount of cystathionine beta-lyase MetC and methionine synthase MetE lowered more than threefold during glucose starvation. This was also the case for proteins involved in biosynthesis of arginine like N-acetyl-g-glutamyl-phosphate reductase ArgC and acetylornithine transaminase ArgD. Similar results could be obtained for enzymes that function in synthesis of branched amino acids like threonine dehydratase IlvA, aminotransferase YwaA, and 2-isopropylmalate synthase LeuA (Table I). This supports the assumption that under starvation conditions the degradation of unemployed enzymes not protected in functional metabolic complexes can help to provide the amino acids necessary for de novo protein synthesis. However, the arrest of biomass production is most probably the major actor of this repression. An integrated view of protein repartition is provided in Fig. 7. A considerable portion of protein mass dedicated to amino acid biosynthesis pathways seems to be allocated to the central carbon metabolism. The increased need of protein mass is most probably caused by induction of the TCC, gluconeogenesis, and pathways for acquisition of secondary carbon sources and cannot be covered by lowered protein amounts of glycolytic enzymes alone (supplemental Fig. S5).

Fig. 7.

Integrated view of functional class assigned protein abundances of B. subtilis during glucose starvation. The relative distribution of protein amounts among the main processes of B. subtilis during adaptation to glucose starvation are shown in different bars (from left to right) for exponentially growing cells (exp), cells during transient phase (trans) and at maximal optical density (max. OD) as well as for cells after 60 min D, 180 min E, and 240 min F, stationary phase caused by glucose depletion.

General Adaptive Response under Heat Stress

In this work, heat stress was selected as well described model for physical stresses. In contrast to glucose starvation here the availability of carbon sources is not limited and protein damage is the main challenge the cell has to face. For the growth-restricting heat stress (μ_exp = 1.2/h, μ_heat = 0.5/h) analyzed in this work hierarchical clustering revealed that 40% of all proteins with increasing amounts are SigB-dependent stress proteins or mediate stress resistance. Altogether 44 members of the SigB regulon could be absolutely quantified from which 33 were found to be accumulated more than twofold. Indeed, heat stress seems to elicit the induction of the SigB regulon, but it could be a substantial burden for the cell, as it can occupy up to 20% of the translation capacity (43). Accordingly, only a transient transcription of genes of the SigB regulon is reported for both conditions (17, 44). As cells starving for glucose and heat stressed cells exhibit considerably different cell volumes (differences around factor 2, supplemental Table S6), absolute protein abundances in heat stressed cells have been corrected for the differences in cell size by normalizing to cell volumes of glucose starved cells in order to allow comparison of protein concentration per cell under both conditions. For comparison of the SigB response in both conditions tested, corrected protein abundances will be given in molecules per size-corrected cell in the following. In contrast to glucose starvation, a clear accumulation of induced SigB-dependent proteins can be measured only during heat stress (Fig. 5). Thereby, the general stress proteins YhdN, YvyD, and GsiB showed highest increase in protein amount (more than 19-fold in size-corrected cells, Table II). With 97,800–151,000 molecules per cell (49,400–110,000 molecules per size-corrected cell) most abundant SigB-dependent proteins during heat stress were the relatively small general stress protein GsiB, the anti-anti-SigmaB-protein RsbV, and the protease ClpP. These increases indicate an important functional role of the SigB-dependent proteins during heat stress, which legitimates the high translation capacity needed for them.

Table II. Determined protein amounts (in molecules per cell) of selected proteins after 60 min heat stress. Absolute amounts for all quantified proteins can be found in supplementary Table S1.

Protein	Function	Molecules/cell
		Exponential growth	Stress
σB-regulon
YhdN	General stress protein	100	5800
YvyD	General stress protein	200	9500
GsiB	General stress protein	3900	151,000
RsbV	Control of SigB activity	47,400	135,000
Chaperones and proteases
DnaK	Molecular chaperone	11,000	51,000
GrpE	Activation of DnaK	2000	8400
GroEL	Chaperonin	20,000	190,000
GroES	Chaperonin	37,000	500,000
HtpG	Molecular chaperone	1200	17,000
LonA	Protease	250	680
ClpC	ATPase subunit of the ClpC-ClpP protease	1100	11,200
ClpE	ATP-dependent Clp protease-like	850	900
ClpP	ATP-dependent Clp protease proteolytic subunit	12,500	97,800
Heat inducible proteins
NfrA	Stress protein	400	2000
AhpC	Alkyl hydroperoxide reductase subunit	56,000	250,000
AhpF	Alkyl hydroperoxide reductase subunit	7800	33,000
Negatively altered proteins
HisH	Tyrosine transaminase	1000	600
MetE	Methionine synthase	70,800	42,600
ThiF	Biosynthesis of thiamine	370	200
HisF	Biosynthesis of histidine	4200	2700
ArgB	N-acetylglutamate 5-phosphotransferase	470	150
PatB	Cystathione-beta-lyase	3800	2600
MoeA	Molybdopterin biosynthesis protein	730	290

Further analysis of the distribution of protein amounts to different regulons in the cell additionally revealed that proteins whose expression is controlled by PerR (response to peroxide) and Spx (response to thiol specific oxidative stress) are also enriched (supplemental Table S5), emphasizing the overlap between heat shock response and reaction to oxidative stress. As secondary oxidative stress is described to occur after different environmental stresses (45, 46) quantitative data of proteins with function in adaptation to oxidative and electrophile stress derived from this study were checked. Indeed, the amount of 25 out of 35 quantified oxidative stress proteins increased at least twofold (Table III). The molecules per cell of the general stress proteins YvyD, OhrB, SigB, and Dps increased more than 10-fold after 60 min of heat stress (supplemental Table S1, supplemental Fig. S8). Protein concentrations of the nitro/flavinreductase NfrA, the probable thiol peroxidase Tpx, the alkyl hydroperoxide reductase AhpC/AhpF, and superoxide dismutase SodA were amplified at least fourfold.

Table III. Proteins with function in adaptation to oxidative and electrophile stress and their amounts after heat stress. For all proteins listed SigB-dependent regulation or already known induction by heat and oxidative stress (45) are indicated by x. For each protein molecules per cell and the relative quantitative change is given. Protein names in boldface are subject to quantitative changes of a factor of 4 or higher.

Protein	BSU number	SigB-dependent	Induced after heat and oxidative stress (45)	Molecules per cell				Relative foldchange
				Control	10 min	30 min	60 min	10 min	30 min	60 min
YvyD	BSU35310	x	x	207	681	3262	9540	3.29	15.76	46.09
OhrB	BSU13160	x	x	1457	12,252	26,988	20,433	8.41	18.52	14.02
SigB	BSU04730	x	x	53	227	459	676	4.28	8.66	12.75
YsnF	BSU28340	x	x	115	986	509	256	8.57	4.43	2.23
Dps	BSU30650	x	–	1657	11,213	14,860	17,867	6.77	8.97	10.78
SodA	BSU25020	x	–	50,924	110,342	179,084	224,860	2.17	3.52	4.42
TrxA	BSU28500	x	–	9490	23,432	36,874	36,691	2.47	3.89	3.87
YraA	BSU27020	x	–	1373	4051	5061	4852	2.95	3.69	3.53
KatE	BSU39050	x	–	570	1028	1269	1855	1.80	2.23	3.25
YvgN	BSU33400	x	–	22,012	38,419	44,612	54,906	1.75	2.03	2.49
YdbD	BSU04430	x	–	2088	3282	3989	3890	1.57	1.91	1.86
KatX	BSU38630	x	–	414	577	361	242	1.39	0.87	0.58
NfrA	BSU38110	–	–	394	1102	1623	2015	2.80	4.12	5.11
Tpx	BSU29490	–	–	7703	16,365	26,057	35,678	2.12	3.38	4.63
AhpC	BSU40090	–	–	56,404	122,829	179,490	253,482	2.18	3.18	4.49
AhpF	BSU40100	–	–	7753	14,286	21,777	33,379	1.84	2.81	4.31
OhrA	BSU13140	–	–	502	877	1239	1958	1.75	2.47	3.90

Protein	BSU number	SigB-dependent	Induced after heat and oxidative stress	Control	10 min	30 min	60 min	10 min	30 min	60 min
AzoR1	BSU19230	–	–	3013	5211	11,391	11,585	1.73	3.78	3.85
AzoR2	BSU33540	–	–	4564	7377	8379	15,738	1.62	1.84	3.45
MsrA	BSU21690	–	–	164	249	481	562	1.52	2.93	3.43
MhqA	BSU12870	–	–	519	986	1398	1658	1.90	2.69	3.19
YgaF	BSU08720	–	–	1278	2110	2346	4000	1.65	1.84	3.13
HypO	BSU07830	–	–	979	1402	1631	3062	1.43	1.67	3.13
YwbC	BSU38370	–	–	3854	6778	8080	11,618	1.76	2.10	3.01
MrgA	BSU32990	–	–	808	1210	1279	2321	1.50	1.58	2.87
CatR	BSU33680	–	–	105	287	316	300	2.73	3.01	2.86
YodC	BSU19550	–	–	3942	5772	6104	10,996	1.46	1.55	2.79
YqjM	BSU23820	–	–	957	1532	1805	2209	1.60	1.89	2.31
MhqD	BSU19560	–	–	2156	2889	2322	4354	1.34	1.08	2.02
BshA	BSU22460	–	–	312	444	678	612	1.42	2.17	1.96
MsrB	BSU21680	–	–	775	1264	1520	1154	1.63	1.96	1.49
BshB1	BSU22470	–	–	406	416	349	557	1.02	0.86	1.37
BshC	BSU15120	–	–	894	1093	973	1193	1.22	1.09	1.33
KatA	BSU08820	–	–	2359	2485	1720	2330	1.05	0.73	0.99
HxlA	BSU03460	–	–	180	267	87	36	1.48	0.48	0.20

Specific Response to Heat Stress

Adaptation to growth-restricting heat stress was mainly realized by a strong accumulation of proteins belonging to the HrcA and CtsR regulons. Under this condition for six out of nine proteins of the HrcA regulon, which mainly function in protein folding, increased amounts could be determined. The molecular chaperon DnaK and its activator GrpE accumulated four-fivefold (Table II, Fig. 3, supplemental Fig. S8). Amounts of chaperonins GroEL and GroES even increased more than 10-fold (Table II, Fig. 3, supplemental Fig. S8). With that GroES was the most abundant protein under stress conditions constituting more than 7% of the total molecule amount detected.

Besides proteins of the HrcA regulon, products of genes controlled by CtsR were clearly enriched. On protein level 7 of 12 members of the regulon could be absolutely quantified and were found to be accumulated. A significant (p = 0.05, one-way ANOVA) increase in protein concentration of the ATPases ClpE, ClpC, and protease ClpP was detected. The amount of ClpE increased about sevenfold after 10 min of heat stress. However, after 30 min of stress ClpE concentration had already reached basal level of about 800 molecules per cell again indicating low protein stability. Similar observations have been made by Gerth and co-workers (47). In contrast, enrichment of ClpC and ClpP (Table II) was determined to be four-eightfold already after 10 min of heat stress, but remained stable during all time points of stress examined.

Although the already described induction of the HrcA and CtsR regulons represent the main response to heat stress, various other heat inducible proteins could be absolutely quantified in this study. HtpG was found to be induced 14-fold during heat stress. LonA accumulated about threefold, but failed to reach significance level (p = 0.05, one-way ANOVA). NfrA was found to be accumulated fivefold whereas protein amounts of AhpC and AhpF increased about fourfold (Table II).

Under heat stress most altered protein amounts were positively influenced, but there were also enzymes whose amount decreased. The majority of these proteins is involved in biosynthesis and acquisition of amino acids and cofactors (supplemental Fig. S8). Examples for negatively controlled enzymes are tyrosine transaminase HisH, methionine synthase MetE, and ThiF, involved in thiamine biosynthesis (Table II). The protein amount of these proteins was lowered constantly during heat stress (supplemental Table S1). In contrast, the concentrations of cyclase-like protein HisF, N-acetylglutamate 5-phosphotransferase ArgB, cystathione-beta-lyase PatB, and the molybdopterin biosynthesis protein MoeA were decreased until 30 min of heat stress to reach control levels at 60 min of stress again (Table II, supplemental Table S1), emphasizing the global coordination of various regulation in response to the growth rate adaptation during heat stress.

The results of this study show that B. subtilis needs to reorientate protein synthesis during heat stress adaptation in order to accumulate general and stress-specific proteins although the main metabolic processes seem to be compromised because of the stress. The graphical representation of the repartition of protein amounts among main pathways of B. subtilis during heat stress depicts the very pronounced accumulation of chaperones and proteases during heat stress (Fig. 8, left). These proteins constitute 4% of the total protein mass during exponential growth at 37 °C, but their fraction is already increased to 13% after 30 min heat stress at 52 °C. Furthermore, the left part of Fig. 8 indicates that protein amounts dedicated to chaperones and proteases are most probably derived from resources won by turnover of enzymes responsible for amino acid synthesis (22% of total protein amount under control conditions, 13% after 30 min heat stress).

Fig. 8.

Integrated view of protein abundance and protein production assigned to functional processes of B. subtilis during heat stress. In the left part of the Fig. the relative distribution of protein amounts among the main processes of B. subtilis during adaptation to heat stress are shown in different bars for (from left to right): exponentially growing cells (control) and cells after 10 min, 30 min, and 60 min (D) heat stress. The right part of the Fig. illustrates dynamical aspects of protein repartition (protein production) in B. subtilis during heat stress. The relative distribution of accumulated protein amounts among the main processes during adaptation to heat stress is shown in different bars for (from left to right): exponentially growing cells (exp.) and accumulated proteins between 0 and 10 min, 10 min and 30 min, 30 min and 60 min heat stress. The size of the circles represents the relative amount of proteins accumulated between single sample points (values are given in the respective circles).

Dynamical Aspects of Heat Stress Adaptation

The availability of large-scale absolute protein concentrations allows analysis of dynamical aspects of protein synthesis during stress adaptation. In order to calculate protein production rates during heat stress we assume that proteins synthesized during exponential growth under control conditions and still increasing during heat stress are stable. Studies on glucose starvation revealed that this is quite reasonable at least for vegetative enzymes (19). Regulatory proteins may have a much higher turnover rate because they are required for temporary reaction of the living cell to changing surroundings. However, in our representative data set regulatory proteins make up less than 1.5% of the total protein amount in a cell and their mass will therefore not essentially influence calculation on protein production.

The optical density of a bacterial culture reflects cell growth and therewith the “dilution” of protein amounts by cell division. In this study, the optical density increases 1.62-fold between the control sample of exponentially growing cells and the sample after 60 min of heat stress at 52 °C (Fig. 1). Hence, we would estimate that 62% of the total protein mass in heat stressed cells was already present in the control sample. In reverse this could mean that 38% of the total protein amount present in the stressed cells was newly produced during the 60 min heat stress phase (Fig. 8, circles). In order to illustrate this point the distribution of protein mass accumulated between the sample points to functional groups was analyzed. This led to a figure that shows the repartition of protein production in a given phase (Fig. 8 [left, bars], supplemental Fig. S9). In the first phase, between 0 and 10 min heat stress, 22% of newly accumulated proteins are chaperones. In contrast, the portion of protein amount dedicated to amino acid synthesis is strongly reduced. During the second phase, between 10 and 30 min of stress, the main portion of all protein amount accumulated is devoted only to chaperones, the portion of accumulated proteins with functions in amino acid synthesis is now smaller than 1%. In the third phase, between 30 and 60 min of heat stress, the heat adaptational response seems to be finished. A significant amount of accumulated proteins is now again dedicated to amino acid synthesis.

DISCUSSION

In the study presented here we compared for the first time the differential adaptation of B. subtilis to heat stress and glucose starvation on the basis of absolute protein concentrations at a large scale. Using two-dimensional PAGE with a pH range of 4–7 we were able to determine protein amounts on the single cell level for 773 cytosolic proteins at seven time points of glucose starvation and four time points in a heat stress experiment including controls. Hence, only 26.7% of all cytosolic proteins (locateP (48)) could be identified by this gel-based approach. However, assuming that only 80% of all proteins are expressed at the same time (49), the protein coverage of the presented approach increases to 33.4%, covering most main metabolic pathways and processes in B. subtilis (supplemental Fig. S7, supplemental Fig. S10).

427 proteins quantified during glucose starvation in this study have been relatively quantified elsewhere (18). Whereas the study presented here is limited to cytosolic proteins that can be detected by two-dimensional PAGE Otto and coworkers (18) could provide quantitative data for additional 890 cytosolic proteins using mass spectrometry-based techniques. However, they were only able to provide relative quantification data whereas the study presented here for the first time reports physiologically relevant absolute protein concentrations on B. subtilis under glucose starvation. For the proteomic analysis of the heat stress adaptation this is even more pronounced. Until now relative protein quantification data for 246 cytosolic proteins have been available (50). This could be extended by additional 455 proteins. For all 701 proteins identified during heat stress in this study absolute quantification data are available.

The quantification of protein molecules per cell renders the possibility to compute the redeployment of resources in the bacterium during stress and to accurately estimate the associated energy costs. However, such applications will require the accurate estimation of the costs associated to the mRNAs redeployment. This could be achieved if access to the absolute quantification of mRNAs during the stress period and their half-life during the same stress period was provided. Additionally the integration of nonprotein parameters, such as known polysaccharides or lipids would be of great interest in order to gain new insight on, for example, the membrane composition during different stress conditions. Unfortunately, until now the absolute quantification of membrane proteins is still a challenge because of the need for a complex sample preparation (51). Furthermore, absolute quantification of nonproteinogenic components would be inevitable in this context.

Adaptation of B. subtilis cells to various stresses is the key to survival in the natural habitat. Thereby it is crucial to balance saving of energy and resources and accumulating of inevitably needed stress proteins. This study illustrates different characteristics of general and specific stress responses to distinct stresses as exemplarily shown for glucose starvation and heat stress. Although general, nonspecific adaptive responses ensure survival during various conditions, the specific stress adaptation allows a direct interaction with the stimulus (10).

The general response to starvation is mainly marked by the negative stringent response whereas nonspecific response to heat stress is dominated by activation of promoters controlled by the alternative sigma-factor SigB. Although the SigB regulon is activated under stress and starvation, proteins of SigB-dependent genes accumulate only after heat stress. During glucose starvation these proteins are transiently synthesized, but only five accumulated significantly more than twofold. On average SigB-dependent proteins accumulated 1.8-fold after 240 min of glucose starvation, but 2.6-fold after 60 min of heat stress. Concurrently, gelfree relative quantification data on salt stress in B. subtilis (52) revealed an average accumulation of 3.2-fold. Maximal accumulation ratios were 5.4 for salt stress, 8.0 for glucose starvation, and 26.5 for heat stress indicating pronounced differences in the SigB-dependent regulation in response to various stresses.

Secondary oxidative stress is a phenomenon recently described to occur after ethanol treatment, hyperosmotic and cold stress (45, 46). For four proteins recently described to be induced after heat and oxidative stress (45), namely SigB, OhrB, YsnF, and YvyD, a significantly increased amount after heat stress could be detected in this study. Furthermore, 21 additional proteins with functions in the resistance against oxidative and electrophile stress were found to be accumulated (Table III). The molecules per cell for four proteins (YvyD, OhrB, SigB, and Dps) increased more than 10-fold. As these proteins are controlled by SigmaB this high accumulation is most probably caused by additive effects of the general stress response and the response to oxidative stress. This is supported by the observation that concentrations for proteins without additional regulation by the general stress response increased not more than fivefold. It has been shown that oxidative stress also induces genes otherwise repressed by MgsR controlling a subregulon within the general stress response (53). In this study, 23 target genes of MgsR could be absolutely quantified after heat stress. Thirteen of these proteins show an expression pattern similar to that after ethanol treatment suggesting a regulatory function of MgsR after heat stress.

In contrast, secondary oxidative stress does not seem to play an important role in adaptation to glucose starvation. In this study the amount of only two proteins with functions in the resistance against oxidative and electrophile stress were increased more than twofold (KatE, YbdD). Both proteins are also under control of SigB and their accumulation might rather be an effect of this regulatory mechanism. Additionally, 16 proteins with functions in adaptation to oxidative stress were quantified, but not found in higher copy numbers during stationary phase. This is also supported by recent data from Otto et al. (18). They quantified 24 SigB-independent cytosolic proteins involved in oxidative stress resistance. Only three of these proteins were found to be significantly accumulated. However, this observation might have been caused rather by other direct or indirect regulatory effects than by secondary oxidative stress.

In addition to the general responses the specific stress response to various conditions ensures a direct interaction with the stimulus. After exhaustion of the preferred carbon source, the substantial reprogramming of cellular metabolism is characterized by down-regulation of glycolysis and simultaneous mostly CcpA-dependent induction of metabolic pathways for the utilization of overflow metabolites and other secondary carbon sources (10, 17–19). However, accumulation of TCC enzymes, proteins catalyzing gluconeogenetic reactions as well as of enzymes for the metabolism of secondary carbon sources was less significant in this study when compared with recently published relative quantification data of Otto et al. (18). Relative quantification data from Otto et al. (18) compare protein abundances to calculate changes in protein accumulation without considering protein concentration in the cell. In contrast, absolute quantification data derived from this study consider the cell size and cell count of a sample in order to calculate copy numbers per cell. Hence, if cell size is reduced, like it is the case during starvation (supplemental Table S6), a stable protein concentration in a sample may result in negative changes of copy numbers per cell compared with the control sample leading to enhanced quantitative effects of negative regulations and reduced quantitative effects of positive regulations (Fig. 2). This fact may explain the lowered protein accumulation of TCC enzymes, gluconeogenetic proteins and proteins for the utilization of secondary carbon sources when considering copy numbers per cell.

Specific heat stress response is marked by induction of class I and class III heat stress genes controlled by HrcA and CtsR, respectively.

On transcriptional level it is described that after moderate heat stress dnaK, grpE, and hrcA are strongly induced whereas downstream genes (dnaJ, yqeT, yqeU, and yqeV), groEL, and groES are not induced more than twofold (44). Under growth-restricting heat stress condition, as described here, amounts of all detectable proteins of the regulon increase at least fivefold.

In 1997 Schulz et al. found htpG to be induced 10-fold after transition from 37 °C to 48 °C both at the level of transcription and translation (54). The even higher accumulation rates of HtpG in this study are therefore most probably caused by the higher temperature of 52 °C.

Although it is described to be heat inducible, no member of the CssRS regulon could be detected in this study. This is most probably because these proteins are, with exception of CssR, membrane-anchored and therefore not accessible with two-dimensional PAGE.

Comparisons of our data with recent relative protein quantification (50) revealed a good overlap of results. Of 46 heat-induced proteins 23 were also found to be induced by Wolff and coworkers (50). A broad overlap also applies to proteins with reduced amounts that function in biosynthesis and acquisition of amino acids and cofactors or in transcription and translation. Of 38 proteins negatively influenced in either of the studies 27 show a similar regulation pattern in both studies.

The availability of a comprehensive data set of absolute protein concentrations allows calculation of accumulated proteins during a given phase of stress adaptation without “classical” quantitative synthesis data at hand. However this calculation makes use of the assumption that examined proteins remain stable during the stress phase analyzed. As this may be a reasonable assumption for the comparable short heat stress study presented here (up to 60 min stress) this may not be transferred to the glucose starvation data, which became available within this work (up to 240 min starvation). The amount of a protein is a balance between production and degradation. In exponential phase, almost all reduced protein amounts are a direct consequence of the so-called dilution effect caused by cell growth. That means that the protein production in balanced systems can be easily estimated because it represents the protein amount needed to compensate the dilution effect. Consequently, protein production for each protein is equal to its steady-state amount times the growth rate. In a nonbalanced system, e.g. in transient phase during a stress, it is also possible to deduce the protein production program during a given period, but only under the assumption that the protein degradation is restricted to the dilution. In this case possible proteolysis of some proteins will lead to underestimation of their production. Hence, the presented results can only be an estimation of the distribution of produced and accumulated proteins to functional categories. However, our data can give valuable insights in regulatory aspects during adaptation of B. subtilis to changing growth conditions.

The high amount of proteins with decreased amount in this study (especially after glucose starvation) leads to the question how the proteins are selected to be degraded. The assignment of proteins with lowered amounts to functional categories (supplemental Table S7) revealed that most detected enzymes involved in biosynthesis of cofactors are degraded and not newly accumulated under both, glucose starvation and heat stress. Notably, the enzymes catalyzing the first committed steps in biosynthesis of branched amino acids, pyrimidines, and purines, namely IlvB, CarA, CarB, and Prs, have been found in decreased amounts only during response to glucose starvation pointing to an energy saving mechanism behind this regulatory effect. Hence, we suggest that starving cells need to save energy because of the limited availability of resources and therefore restrict metabolism to minimal activity to guarantee survival. Consequently, in a first instance, cells degrade biosynthetic enzymes, in many cases after regulation of their expression by the stringent response. Although proteins that are active and integrated into functional complexes are protected against a proteolytic attack, these enzymes may be damaged or structurally pertubated in the absence of their (co-) substrates leading to the recognition of these proteins by the degradation machinery.

After the introduction of quantitative Western blotting (3), flow cytometry (55), and MS-based or MS-coupled strategies (4–6, 56) for a global determination of absolute protein abundance, we herewith present first physiologically relevant data for stress adaptation in the model bacterium B. subtilis. Because of a large number of examined time points with high coverage of reliably quantified proteins we are confident to provide suitable data for modeling of cellular regulation under altered growth conditions not only for the systems biology community.