Swiss-Czech Proteomics Server

Proteomics Database of Caulobacter crescentus

Basic information and experimental procedures

Here you will find basic information about cultivation conditions, sample preparation, and data treatment.
You can also download a lab protocol in the PDF format.


Proteomic Analysis of the Bacterial Cell Cycle

Grünenfelder, Björn1, Rummel, Gabriele1, Vohradsky, Jiri2, Röder, Daniel3, Langen, Hanno3 and Urs Jenal1*

1 Division of Molecular Microbiology, Biozentrum, University of Basel, Klingelbergstrasse 70, CH-4056 Basel, Switzerland
2 Institute of Microbiology, Czech Academy of Sciences, CZ-14220 Praha-4, Czech Republic
3 Roche Genetics, F. Hoffmann-La Roche Ltd, CH-4070 Basel, Switzerland

This chapter was published as an article in the Proceedings of the National Academy of Sciences USA on April 10 2001, Vol. 98 (8), p. 4681-4686.


Abstract

A global approach was used to analyze protein synthesis and stability during the cell cycle of the bacterium Caulobacter crescentus. Approximately one fourth (979) of the estimated C. crescentus gene products were detected by two-dimensional gel electrophoresis, 144 of which showed differential cell cycle expression patterns. Eighty-one of these proteins were identified by mass spectrometry and were assigned to a wide variety of functional groups. Pattern analysis revealed that co-expression groups were functionally clustered. A total of 48 proteins were rapidly degraded in the course of one cell cycle. More than half of these unstable proteins were also found to be synthesized in a cell cycle-dependent manner, establishing a strong correlation between rapid protein turnover and the periodicity of the bacterial cell cycle. This is the first evidence for a global role of proteolysis in bacterial cell cycle control.


Introduction

Bacteria have the potential to rapidly multiply and spread in an environment that provides all the nutrients for growth. Rapid growth of virulent bacteria can be fatal for both plant and animal hosts. To understand the ability of these simple cells to quickly grow and divide while faithfully passing on their genetic information, it is critical to unravel the regulatory circuits that control the bacterial cell cycle. In particular, cells have to control DNA replication, chromosome segregation, and cytokinesis temporally and spatially, and coordinate these events with growth. The periodicity of these cell cycle events is accompanied by oscillations of gene expression both in eukaryotic cells and in bacteria. In the unicellular eukaryote Saccharomyces cerevisiae a total of 800 genes, representing approximately 13% of the yeast genome, are differentially expressed during the cell cycle, many being involved in DNA replication, cell division, mitosis and mating (Cho et al., 1998; Spellman et al., 1998) . Similarly, temporal control of gene expression is an important regulatory element of cell cycle progression and development in the bacterium Caulobacter crescentus (Jenal, 2000) . In accordance with this, a recent study analyzing gene expression on a global scale identified 590 genes, which are differentially expressed during the Caulobacter cell cycle. A large subset of these genes encode proteins involved in the execution of cell cycle and cell differentiation events, whereby the time of gene expression reflected the time of the function of their products (Laub et al., 2000) . This included genes coding for flagellar components (Wu and Newton, 1997) , for the essential DNA-methyl-transferase CcrM (Zweiger et al., 1994) , for cell division proteins (Kelly et al., 1998; Sackett et al., 1998) , and for a number of regulatory proteins involved in cell cycle control (Hecht et al., 1995; Ohta et al., 1992; Quon et al., 1996; Wang et al., 1993) . Thus, studying cell cycle-dependent gene expression in a global manner not only catalogs periodically expressed genes, but can also help to identify genes with novel cell cycle functions. Gene expression studied globally by monitoring changes of the cell’s mRNA levels must be complemented by proteome analysis because the protein rather than the mRNA is the biologically active, and thus more relevant molecule, and because the relationship between mRNA levels and the rates of protein synthesis can be non-linear (VanBogelen et al., 1999) . Moreover, proteomics makes it possible to study critical posttranslational control mechanisms such as modification and protein stability, which may contribute greatly to the ultimate activity of a given protein.

In eukaryotic cells, a complex regulatory network has been elucidated, which authorizes faithful progression and coordination of the cell cycle by acting not only at the level of protein synthesis but also at the level of protein phosphorylation and degradation (Nasmyth, 1996) . Cell cycle progression depends on phosphorylation events performed by cyclin-dependent kinases (Nigg, 1995) . Moreover, proteolysis plays a crucial role by removing key cell cycle regulatory proteins at specific time points, thereby irreversibly driving the cell cycle in one direction (King et al., 1996) . Finally, differential gene expression, usually in combination with protein degradation, enables the cells to alter the levels of regulatory proteins during the cell cycle thereby controlling the time when the threshold level of a specific activity is reached. Recent discoveries in Caulobacter crescentus have revealed similar multi-layered regulatory mechanisms for cell cycle propagation in bacteria (Osteras and Jenal, 2000) . A prime example is the transcriptional regulator CtrA, which acts as timing device for several cell cycle events, including DNA replication, DNA methylation and cell division. CtrA activity is temporally and spatially controlled by differential expression, phosphorylation and protein degradation (Domian et al., 1997; Quon et al., 1996) . CtrA negatively controls initiation of chromosome replication and has to be removed before cells enter S-phase (Quon et al., 1996; Quon et al., 1998) . This is accomplished by subsequent CtrA dephosphorylation and degradation during the G1-to-S phase transition (Domian et al., 1997) . Expression and phosphorylation of CtrA take place after replication has initiated, quickly increasing the level of active CtrA in the dividing cell. This triggers the sequential activation of genes involved in cell cycle progression and development (Reisenauer et al., 1999) . While the activation of CtrA through phosphorylation is carefully fine-tuned by several kinases localized to specific subcellular sites (Jacobs et al., 1999; Wheeler and Shapiro, 1999) , CtrA degradation is catalyzed by the essential ClpXP protease complex, a structural homolog of the 26S proteasome of eukaryotic cells (Jenal and Fuchs, 1998) . ClpXP not only degrades CtrA at the correct time of the cell cycle, but is also required for G1-to-S phase transition. Mutants lacking either the ClpX or the ClpP component of the protease are specifically blocked in G1 (Jenal and Fuchs, 1998) . This indicates that in addition to CtrA, other factors have to be removed by the protease to allow proper cell cycle progression (Jenal and Fuchs, 1998) . However, in contrast to the situation in eukaryotic cells, the significance and scope of specific protein degradation for cell cycle control has not been elucidated in bacteria so far. While the total protein turnover has been estimated to be about 3% of the Escherichia coli protein mass per hour (Pine, 1970) , only a small number of E. coli proteins with a short half life have been identified (Gottesman, 1996; Larrabee, 1980) .

We have used 2-D gel electrophoresis combined with peptide mass fingerprinting to investigate both protein synthesis and degradation during the C. crescentus cell cycle. An earlier study had found that the expression of about 6 % of the protein-spots detected on 2-D gels were under cell cycle control in Caulobacter (Milhausen and Agabian, 1981) . High resolution 2-D gel electrophoresis (Bjellqvist et al., 1993) combined with image analysis (Garrels, 1989) have facilitated the study of complex spot patterns on 2-D gels allowing complete investigation and proper statistical analysis of the image data. In addition, the development of sensitive mass-spectrometry methods (Yates, 2000) to measure exact sizes of peptides derived from digested proteins, combined with the availability of the complete genome sequence has made it possible to rapidly identify proteins isolated from 2-D gels (Wheeler et al., 1996) . To detect differentially expressed proteins, synchronized Caulobacter cells were pulse-labeled at different time points during the cell cycle with 35S-methionine and the extracted proteins were separated by 2-D gel electrophoresis (Fig. 1). Similarly, protein stability was measured by pulse-labeling cells followed by an extended chase period. We found that no less than 15% of the reproducibly detectable protein spots were differentially expressed during the cell cycle and 5% had a half life of one cell cycle or less. Most importantly, comparison of the two data sets revealed that more than half of the unstable proteins were also differentially expressed during the cell cycle, establishing a strong correlation between rapid protein turnover and the periodicity of the cell cycle. The identification of a large fraction of proteins, which are both differentially synthesized and rapidly degraded opens new entry points into analyzing the role of these proteins in directing cell cycle progression through controlled proteolysis.


Materials and Methods
Cell cycle synchronization and protein labeling.

The synchronizable C. crescentus strain NA1000 was grown at 30oC in M2 minimal glucose medium (M2G) (Johnson and Ely, 1977) . Swarmer cells were isolated by density gradient centrifugation (Stephens and Shapiro, 1993) and released into fresh M2G medium. Cell cycle progression was monitored by light microscopy. To monitor differential protein expression, cells were pulse labeled at 0, 0.3, 0.6, 0.8, and 1 cell cycle units (corresponding to G1, early S, late S, G2 phase and cell division) (Marczynski, 1999; Winzeler and Shapiro, 1995) by adding 20 µCi of a 35S-methionine/cysteine mix (NEG 772, NEN) to 1 ml of culture for 4 minutes, followed by a 2 minutes chase with 0.2% tryptone, 1 mM methionine, and 0.02 mM cysteine before harvesting. Cells were collected by centrifugation, frozen in liquid nitrogen and stored at -80°C. To analyze protein stability, asynchronous cultures were labeled as described above and chased for up to 120 minutes. To assess the influence of the synchronization procedure on protein synthesis, expression patterns were compared in asynchronous cultures before and after the synchronization procedure. Swarmer, stalked and predivisional cells were re-pooled after separation by density gradient centrifugation and pulse-labeled as described above. The protein expression pattern was then compared to a labeled culture that had not experienced the synchronization procedure. Proteins which were significantly up- or downregulated as a consequence of the synchronization procedure were subtracted form the cell cycle expression data set (see below).

Protein preparation.

For analytical gels, 1 ml of pulse labeled cells were washed twice in 20 mM phosphate buffer pH 7 and then lysed in 200 µl lysis buffer containing 8M urea, 4% cholamidopropyl-dimethyl-ammonio-propane sulfonate (CHAPS), 0.8% ampholytes pH 3-10 (Pharmalyte, Pharmacia), 65 mM dithiothreitol (DTT), and a few grains of bromophenole blue. The lysate was frozen and thawed three times before it was centrifuged at 14000 rpm in an Eppendorf centrifuge for 2 minutes. The incorporated radioactivity was determined by scintillation counting of trichloroacetic acid precipitated proteins.

For preparative gels 250 ml cells were washed in 20 mM phosphate buffer pH 7 and resuspended in 8 ml breaking buffer (20 mM phosphate buffer pH 7, 5% sucrose) containing a protease inhibitor cocktail (Complete, Roche), 4 µg/ml RNase and 16 µg/ml DNase. The suspension was passed twice through a precooled French pressure cell at 1.000 psi and centrifuged at 120.000 x g. The soluble proteins were concentrated and washed twice with h6O in an Amicon filtration cell using a membrane with a molecular weight cutoff of 10.000. Solid urea and concentrated lysis buffer were added to the protein solution to give the same final concentration as described above.

Two dimensional gel electrophoresis.

Two-dimensional (2-D) gel electrophoresis was performed as described earlier (Bjellqvist et al., 1993) with minor modifications. In the first dimension, the proteins were separated in 18 cm long immobilized pH gradient (IPG) gels with a non linear pH gradient ranging from pH 3 to 10 (Immobiline DryStrip pH 3-19 NL 18 cm, Pharmacia) in a horizontal electrophoresis tray (Multiphore, Pharmacia). For analytical gels, the IPG-gels were rehydrated overnight in 8M urea, 2% CHAPS, 10 mM DTT, 0.8% Pharmalyte 3-10 (Pharmacia). Protein extract corresponding to 106 cpm (counts per minute) was diluted to 100 µl with lysis buffer and loaded in cups at the basic (cathodic) end of the IPG strips. For preparative gels, the IPG-gels were rehydrated in a rehydration tray (Pharmacia) for at least 6 hours with 1-2 mg of protein sample diluted to 500 µl with lysis buffer. A small piece of electrode strip paper was soaked in 15 mM DTT and placed close to the cathode. The gels were focused at thermostatically controlled 20°C in a four step program: Step 1: 300 Volt hours (Vhrs) at 150 V, step 2: 600 Vhrs at 300 V, step 3: linear increase from 300 to 3500 V for 10 kVhrs, step 4: 60 kVhrs at 3500 V for analytical gels or 110 kVhrs at 3500 V for preparative gels. After isoelectric focusing, the IPG strips were washed in equilibration solution (50 mM Tris-HCl pH 6.8, 6 M urea, 30 % glycerol, 2% SDS) containing first 0.02 g/ml DTT for 10 min followed by a second 10 min wash in equilibration buffer containing 0.025 g/ml iodoacetamide and a few grains of bromophenole blue.

In the second dimension, the proteins were separated with standard continuous 12% SDS gels of 18 cm x 18 cm x 1mm size in a PROTEAN II xi multi-cell (Bio-Rad) using acrylamide with 0.8% piperazyldiacrlyamide crosslinker (Acrylogel-PIP, BDH). Analytical gels were poured without SDS, preparative gels with 0.1% SDS. The IPG strips were overlaid with running buffer containing 0.5 % agarose. The gels were run at 10°C with 40 mA per gel. Analytical gels were fixed in 40% ethanol, 10% acetic acid for at least one hour and soaked in 25% ethanol and 2 % polyethylenglycol 4000 for 30 min each before drying on Whatman paper. The radioactively labeled proteins were detected by exposing the gels to storage phosphor screens (Molecular Dynamics) for 48 hours. Preparative gels were stained with colloidal Coomassie Blue (Novex), and preparative reference gels were dried and exposed to storage phosphor screens.

Protein size (10-100kDa) and isoelectric point range (pH 3-10) of the 2-D gels was determined by comigrating 2-D gel marker proteins (Bio-Rad) with a Caulobacter protein sample. The proteins were visualized by silver staining following a standard protocol (Bjellqvist et al., 1993) .

Data processing and analysis.

Samples from six independent labeling experiments were resolved on six 2-D gels for each time point investigated. The 2-D gel autoradiographs were matched and quantified by image analysis using PDQuest (Version 5.0.1, PDI). The quantified data were then analyzed with S-Plus (MathSoft) and Excel (Microsoft) as follows: 1) The spot intensities were converted into ppm (parts per million) of the total gel intensity and normalized as described in (Vohradsky et al., 1997) . 2) Spots were removed from the data set of a given time point if they were detected in less than three of the six repeats. 3) Single outliers were removed by a standard t-significance test. 4) All spots were removed which were present on gels of the non-synchronous cultures but could not be detected on any of the gels of the cell cycle time points and vice versa. 5) Spots with the highest intensity below 200 ppm had high experimental variation and were removed. These spots had the highest proportion of experimental variation with a coefficient of variation of up to 50% and were therefore unreliable. Finally 979 highly reproducible protein spots were left and were used as the minimal reproducible data set for statistical analyses.

Spot intensity changes were considered to be statistically significant and relevant if the ANOVA (analysis of variance) or t-test confidence level was higher than 99% and if the ratio between two mean values was at least two. All spots identified by these criteria were manually investigated on the original gels and spots with low quality or clear mismatches were removed. An additional 162 spots were removed from the cell cycle expression data set that were detected at 0 minutes (swarmer cells) and/or at 45 minutes (stalked cells) but not at 140 minutes (after cell division). These types of expression changes do not make biological sense, and most likely represented analysis artifacts. Spots which were significantly induced or repressed by the synchronization procedure (see above) and whose cell cycle expression patterns could only be attributed to this synchronization effect, were removed from the cell cycle expression data set. Spots with differential expression patterns were sorted by hierarchical clustering with an aglomerative nesting algorithm (Kaufman and Rousseeuw, 1990) (see Fig. 7 Appendix).

Protein identification.

Protein spots were mainly identified by peptide-mass fingerprinting (Fountoulakis and Langen, 1997) . Spots were cut out from Coomassie stained preparative gels, destained and digested with endoproteinase Lys-C or Trypsin (Fountoulakis and Langen, 1997) . The masses of the peptides after the proteolytic digest were determined with a matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrometer (Reflex3, Bruker). All translated putative open reading frames (ORF) larger than 80 amino acids of the Caulobacter pre-release genome sequence version of September 1999 (Nierman, 2001) were used for the peptide size query. Protein sequences were searched that could theoretically give rise to at least three peptides of the measured masses after proteolytic digest as described in (Berndt et al., 1999) . A protein was considered to be identified if it was detected as the only candidate in two independent experiments, or if at least five peptide masses matched the best hit. Protein homology searches were done by BLAST-2 (Altschul et al., 1997) . A function was assigned to an identified protein if the similarity had an expectancy value e < 10-8 and if the function was supported by experimental data for at least one of the homologs.

Several proteins were identified by immunoblotting, performed as described previously (Aldridge and Jenal, 1999) . 35S-labeled protein extracts were separated on 2-D gels as described above. After electrophoresis the gels were transferred to a PVDF membrane (Immobilone-P, Millipore) for 12 hours at 30 V. The membranes were dried and exposed to x-ray films. The same membranes were then used for immunodetection with polyclonal antibodies against the following proteins (the dilutions are shown in parentheses): CtrA (1:10’000) (Domian et al., 1997) , ClpX (1:10’000) (Jenal and Fuchs, 1998) , ClpP (1:20’000) (Jenal and Fuchs, 1998) , PleD (1:5’000) (Aldridge and Jenal, 1999) , CcrM (1:5’000) (Stephens et al., 1996) , Flagellins (1:5’000), FtsZ (1:5’000) (Kelly et al., 1998) , FliF (1:5’000) (Jenal and Shapiro, 1996) , McpA (1:20’000) (Alley et al., 1992) , FliL (1:5’000) (Jenal et al., 1994) , FlgH (1:5’000) (Jenal et al., 1994) , and FliM (1:2’500) (Jenal et al., 1994) . A polyclonal goat-anti-rabbit IgG antibody (1:10’000) (Gibco BRL) coupled to horseradish-peroxidase was used as secondary antibody and was detected by chemoluminescence (Renaissance, NEN) following the manufacturers instructions. The signal of the immunoblot was overlaid with the autoradiogram to identify the corresponding 2-D gel spots. The following proteins could finally be detected and localized on the 2-D gels by immunoblotting: CtrA, ClpX, ClpP, PleD, CcrM, Flagellins, and FtsZ.

Immunoprecipitation of ClpP was performed as described in (Jenal et al., 1994) using an anti-ClpP antibody (Jenal and Fuchs, 1998) . Expression profiles, identities, and 2-D gel coordinates of all protein spots that were differentially expressed during the cell cycle, repressed or induced by the synchrony procedure, or exhibited a change in intensity during chase are shown in Table 2 and Fig. 8 (Appendix).


Results

15% of the C. crescentus Proteins detected are Differentially Expressed During the Cell Cycle.

We used a global proteomics approach to independently analyze the timing of protein synthesis and decay during the C. crescentus cell cycle. First, protein synthesis was monitored by pulse labeling synchronized Caulobacter cultures at five different time points of the cell cycle (Fig. 1). A total of 979 protein spots were reproducibly detected on two dimensional (2-D) gels (see Materials and Methods; Fig. 8A Appendix), corresponding to about 1/4 of the estimated total number of Caulobacter genes (Nierman, 2001) . Since membrane integral proteins generally can not be well separated on 2-D gels and low abundant proteins are difficult to visualize (Wilkins et al., 1998) , these 979 spots mainly represent soluble, highly abundant proteins. Taken into account that about 25% of all bacterial proteins are membrane-integral proteins (Wallin and von Heijne, 1998) these spots represent about 35% of all soluble proteins encoded by the Caulobacter genome. The expression of 234 spots of this minimal reproducible data set oscillated significantly during the progression of synchronized cultures. To assess the influence of the synchronization protocol on protein expression, extracts of pulse-labeled non-synchronous cells were compared before and after the synchronization procedure (see below). The expression of 90 protein spots was significantly affected either positively or negatively by the steps of the synchronization protocol and the corresponding spots were removed from subsequent analysis. The remaining cell cycle-variable data set contained 144 spots (15% of all spots detected) that were randomly distributed among the resolved spots. These spots were then sorted by cluster analysis into 23 groups of proteins with distinct cell cycle expression profiles (Fig. 2). Clustering

 

Fig. 1. Cell cycle of Caulobacter crescentus and cell cycle-dependent protein expression of the CtrA regulator. A) Motile, replication silent swarmer cells (G1 phase) differentiate into stalked cells by shedding the polar flagellum and growing a stalk at the same pole. DNA replication is initiated in stalked cells and continues as cells elongate and increase in mass during S phase. A new flagellum is assembled in the predivisional cells at the pole opposite the stalk. Upon completion of DNA replication, the newly synthesized chromosomes segregate to the poles and an asymmetric cell division generates two new daughter cells (G2 phase). B) Protein synthesis was measured during the cell cycle by pulse labeling cells of a synchronized culture with 35S-methionine in G1 (0 cell cycle units), early S (0.3), late S (0.6), G2 phase (0.8) and immediately after cell division (1.0). The labeled extracts were separated on 2-D gels and fluctuations were determined by quantifying and comparing the spot intensities. The example shows a small area of the 2-D gels with the arrows marking the CtrA protein. C) Oscillation of ctrA expression during the C. crescentus cell cycle. Relative levels of ctrA mRNA (diamonds) were taken from (Quon et al., 1996) and CtrA protein synthesis (bars) was quantified from the 2-D gel spots shown in B) for each time point investigated.

expression data has been shown to compile members of synexpression groups, which represent sets of genes that share a complex expression pattern under different conditions and that function in the same process (Eisen et al., 1998; Niehrs and Pollet, 1999) . In agreement with this we found possible synexpression groups for several proteins involved in riboflavin synthesis (four proteins in cluster 6), energy metabolism (four proteins in the similar clusters 22 and 23), redox reactions (three proteins in cluster 1), amino acid biosynthesis (two proteins in cluster 19), carbohydrate metabolism (three proteins in the similar clusters 1 and 7), protein degradation (five proteins in the similar clusters 1, 2, and 7), and motility and chemotaxis (five proteins in the similar clusters 11 and 13) (Table 1, Fig. 2).

Ninety-one protein spots with differential cell cycle expression patterns were identified by peptide-mass fingerprinting or immunoblotting (see Materials and Methods). Several of the identified spots represented separable isoforms of the same proteins, reducing the total number of identified proteins to eighty-one (Table 1). These belonged to a wide variety of functional groups such as metabolism (23%), redox reactions (7%), transcription and translation (6%), protein folding and degradation (14%), cell envelope and transport (7%), motility and chemotaxis (7 %), DNA synthesis (4%), cell division (1%), and regulation (11%). To 20% of the identified proteins no function could be assigned. For a subset of the corresponding genes, distinct cell cycle transcription patterns have been described as for GroEL and GroES (Avedissian and Lopes Gomes, 1996) , FljL and FljK (Minnich and Newton, 1987) , FtsZ (Kelly et al., 1998) , CcrM (Zweiger et al., 1994) , CtrA (Quon et al., 1996) , PleD (Hecht and Newton, 1995) , and DivK (Hecht et al., 1995) , most of which were consistent with the protein synthesis profiles detected on 2-D gels (for an example see Fig. 1). The only exception was ClpP, the proteolytic subunit of the Clp protease (Osteras et al., 1999) , which, as a member of cluster 12, was strongly

 

 

Fig. 2. Co-expression groups with distinct cell cycle expression profiles are contained within 23 clusters. The bars represent the mean relative synthesis levels (as percentage of the maximum value) of all members of a cluster at the five cell cycle time points indicated in Fig. 1C. The values of each individual member of a cluster were calculated from six independent repeats and the changes were significant at the 99% confidence level. The clusters are grouped according to the timing of the highest or lowest expression value. The number on top of each chart indicates the cluster number. The total number of protein spots assigned to each cluster is indicated at the bottom of each chart with the number of unstable spots in parenthesis.

 


 

Table 1. Identification of C. crescentus proteins with differential cell cycle expression profiles. ORF, Caulobacter genome open reading frame number (see Materials and Methods). An asterisk indicates that similar, a hash that no or inverse cell cycle oscillations were detected for the corresponding  mRNAs in (Laub et al., 2000) ; CL, cluster number of co-expression groups (Fig. 2); Function, putative function determined by BLAST homology searches. Proteins with a half-life shorter than one cell cycle are shown in bold. The references for all known examples of oscillating cell cycle transcription are indicated in parenthesis.

induced in G2 (Fig. 2). Using promoter fusions to the lacZ reporter gene, clpP transcription has recently been shown to be constant throughout the cell cycle (Osteras et al., 1999) . Analysis of clpP mRNA levels, on the other hand, suggested an oscillation of clpP transcription identical to the temporal changes of protein synthesis on 2-D gels (Laub et al., 2000) . To examine this discrepancy we analyzed ClpP protein expression during the C. crescentus cell cycle by pulse labeling and immunoprecipitation with a specific anti-ClpP antibody (Jenal and Fuchs, 1998) . The ClpP expression profile obtained with this method matched the 2-D gel expression as well as the mRNA concentration pattern (Fig. 3), indicating that the expression of clpP is indeed subject to cell cycle control and that the regulation acts on the transcriptional level.

 

 

Fig. 3. Cell cycle expression of clpP measured on mRNA and protein level. mRNA levels (triangles, long dashes) were taken from the study by (Laub et al., 2000) . Protein expression was determined by pulse labeling in combination with either 2-D gel analysis (squares, solid line) or immunoprecipitation (diamonds, short dashes).

 

Most of the genes coding for proteins listed in Table 1 were also identified in a parallel study, which used DNA microarrays to analyze the variation of mRNA levels as a function of the C. crescentus cell cycle (Laub et al., 2000) . While 18 of the corresponding 81 genes were not represented on the microarrays, 49 genes showed cell cycle expression patterns, which were identical or very similar to their pulse labeled products on 2-D gels, and four genes showed an inverse cell cycle expression pattern. 10 of the 81 proteins identified in our proteome study showed no fluctuation on the mRNA level, implying translational or posttranslational cell cycle control (see Fig. 9 Appendix).

Functional Diversity of Differentially Expressed Proteins

Several of the differentially expressed proteins identified in this study were involved in DNA metabolism, cell division, or development (Table 1). Despite the fact that a number of replication or cell division genes have been shown to be under cell cycle control (Kelly et al., 1998; Winzeler and Shapiro, 1995; Wortinger et al., 2000; Zweiger and Shapiro, 1994) only few of the corresponding proteins were identified in this study (SSB, cluster 21; CcrM, cluster 13; FtsZ, cluster 20). While DNA synthesis and cell division proteins were upregulated in early S phase, proteins required for motility and chemotaxis were induced late in the cell cycle. A single flagellum and a chemotaxis machinery are assembled during each division cycle in the C. crescentus predivisional cell (Fig. 1). Several proteins required for directed motility were found to be synthesized predominantly during this phase of development, including chemotaxis proteins CheR, CheB, CheYI and CheD, the flagellins FljL and FljK (Minnich and Newton, 1987) , and a regulator of flagellin expression, FlbT (Mangan et al., 1999) (clusters 4, 10, 11, 13).

A surprisingly large number of proteins were involved in metabolic functions not typically thought of as cell cycle regulated (Table 1). Enzymes of energy metabolism were upregulated as cells entered S phase and initiated growth. Similarly, enzymes involved in the production of several amino acids (tryptophane, arginine, methionine) were synthesized predominantly in S and G2 phase, which are associated with cell mass increase. Several enzymes involved in riboflavin and tetrahydrofolate biosynthesis had a sharp expression peak in G1 (Table 1). Several proteins engaged in oxidative stress response were also under cell cycle control. These are the GTP cyclohydrolase II, thioredoxin, thioredoxin reductase, glutathione-S-transferase, and the ferric uptake regulator Fur (Storz and Imlay, 1999; Touati, 2000) . A large group of proteolytic enzymes (6 out of 13) were predominantly synthesized in the swarmer cell (Table 1, Fig. 2). All of these enzymes have predicted export signal sequences (Nielsen et al., 1997) , implying that they are involved in the degradation of extracellular polypeptides and thereby contribute to nutrient scavenging in the planktonic swarmer cell.

The expression of 16% of all proteins is affected by the synchronization procedure

Calculation of a correlation matrix of the different time points using the average spot intensities (Fig. 4A) showed a strong influence of the synchronization procedure on gene expression. High correlation with coefficients ranging from 0.88 to 0.97 was found between the gels of mixed cultures before synchronization and gels of early S-phase, late S-phase, G2-phase, and after cell division. Lower correlation with coefficients ranging from 0.67 to 0.79 was found when the gels of G1-phase and mixed cultures after synchronization, which both represent time points just after the synchronization procedure, were compared with the gels of all other time points. This together with the observed high correlation coefficient (0.92) between the gels of G1-phase and mixed cultures after synchronization indicated that the synchronization procedure had an effect on protein expression which was independent from cell type or stage of the cell cycle.

To study the effect of the synchronization procedure on the expression of single proteins, the quantified data of the non-synchronous cultures before and after synchronization procedure was statistically analyzed (see Materials and Methods). The expression of 154 spots, which corresponds to 16% of the minimal reproducible data set, was found to be affected. Based on the increase or decrease in spot intensity 76 spots were induced and 78 spots repressed by the steps of the synchronization procedure. These spots were randomly distributed over the entire 2-D gel area (Fig. 8C Appendix). As an example, the expression profiles of two spots with either induced or repressed expression by the synchronization procedure are shown in Fig. 4B. The expression of 24 spots was found to be both influenced by the synchronization procedure and cell cycle regulated. In these cases a significant change was observed in the non-synchronous cultures but the cell cycle expression profile did not resemble the expected pattern of a corresponding synchronization effect as shown for the two example spots in Fig. 4.

Of the 154 spots for which the expression was affected by the synchronization procedure, 16 induced and 25 repressed spots were identified, representing 16 and 21 different proteins, respectively (see Table 2 Appendix). The induced proteins were evenly distributed among several functional groups, whereas a large fraction of the repressed proteins were ribosomal proteins and elongation factors (8 of 21). It thus appears that the cell down-regulates the protein synthesis machinery in response to the synchronization procedure. This conclusion is supported by the observation that the incorporation of radioactive amino acids was always low in swarmer cells (data

 

Fig. 4. Effect of the synchronization procedure on protein expression. A) Table showing the correlation coefficients calculated by pair-wise comparison of the spot intensities between the gels of the mixed cultures before and after the synchronization procedure and the five time points during the cell cycle. The bold open numbers symbolize higher correlation (3 0.88) and the bold filled numbers symbolize lower correlation (L 0.79). B) Expression profile of a typical synchronization-induced protein. C) Expression profile of a typical synchronization-repressed protein. Abbreviations: B: Mixed culture before the synchronization procedure, A: Mixed culture after the synchronization procedure, G1: G1-phase, eS: early S-phase, lS: late S-phase, G2: G2-phase, D: after cell division.

not shown). The finding that the expression of 16% of all reproducibly detectable protein spots was affected by the synchronization procedure underscores the requirement for the proper control experiments for cell cycle studies using the density gradient centrifugation method.

Cell Cycle Expression and Protein Stability

The concentration of several important modulators of the Caulobacter cell cycle progression such as the transcriptional regulator CtrA, the cell division protein FtsZ, and the DNA methyltransferase CcrM fluctuate during the cell cycle as a consequence of timed synthesis and degradation. In all three cases this has important functional implications for the timing and control of cell cycle events (Domian et al., 1997; Kelly et al., 1998; Wright et al., 1996) . To identify similar fluctuations we globally investigated protein stability in a pulse-chase experiment. Exponentially growing, asynchronous cells were pulse labeled and chased for up to 120 minutes, equivalent to one cell cycle length. Proteins degraded in the course of one cell cycle should partially or completely disappear during the chase period. Chased extracts were separated and analyzed on 2-D gels allowing comparison of this data set with the cell cycle expression data. For 72 protein spots a significant change was observed during the chase period (see Materials and Methods), with 48 spots decreasing and 24 increasing in intensity (Fig. 5). Most importantly, 26 of the 48 unstable protein spots were also differentially synthesized during the cell cycle. In contrast, oscillating expression was only found for one of the 24 spots that increased during the chase period (Fig. 5). A chi-square test for independence revealed a p-value of 2x10-18 and thus a very strong correlation between protein instability and cell cycle expression. Interestingly, most of the proteins, which are rapidly degraded,

 

 

Fig. 5. Protein spots showing differential expression or stability during the cell cycle. The total number of spots found in each data set is indicated with the number of identified proteins in parenthesis. The number of spots located in two different data sets is shown in the intersections.

 

group in clusters 4, 10, 11, 13, 15, and 20, all of which are characterized by sharp peaks of expression in G1, S or G2 phase (Fig. 2, Fig. 6).

While 63% of the differentially synthesized protein spots could be identified by mass spectrometry, the unstable proteins (33%), because of their low abundance, were underrepresented in the pool of identified proteins (Table 1). As expected, among the 26 spots found to be differentially expressed and rapidly degraded were CtrA, FtsZ, and CcrM (Table 1, Fig. 6). Two additional proteins, the flagellar anchor protein FliF and the chemoreceptor McpA, have recently been shown to be subject to cell cycle-dependent proteolysis in Caulobacter (Alley et al., 1993; Jenal and Shapiro, 1996) . Both were not found in the pool of unstable proteins identified here, most likely because as membrane-integral proteins, they could not be resolved on the 2-D gels. Concomitant with the degradation of FliF and McpA the polar flagellum is ejected into the supernatant at the swarmer-to-stalked cell differentiation. As expected, the two flagellin proteins FljL and FljK were found to disappear during the chase period. The other identified unstable and differentially expressed protein spots

 

 

Fig. 6. Selected examples of protein spots with distinct cell cycle expression and stability profiles. The bars represent the mean relative protein synthesis levels (as percentage of the maximum value) at the five cell cycle time points indicated in Fig. 1C (left panel) or the mean relative concentrations of the pulse-labeled protein (as percentage of the value at time zero) after chasing for 0, 60, and 120 minutes (right panel). The expression and degradation profiles of the cell division protein FtsZ, the CtrA regulator, and the CcrM DNA methyltransferase are indicated and are in good agreement with patterns reported earlier for these proteins (Domian et al., 1997; Kelly et al., 1998; Quon et al., 1996; Stephens et al., 1996; Wright et al., 1996) . In addition, the expression and degradation profiles for CheYI, CheD and three so far unidentified proteins are shown.

 

represented the two soluble chemotaxis proteins CheYI and CheD (Fig. 6), beta-D-glucoside glucohydrolase, ATP-synthase alpha subunit, a prolyl endopeptidase, and a protein of unknown function (Table 1).

The 24 spots which increased in intensity during the chase period are either modification or processing products that appear very slowly or in response to changing conditions during the chase period. In support of this, isoforms were found for 8 of the 10 identified protein spots of this group, including DnaK, enolase, aconitase and acetyl-CoA-acetyltransferase (data not shown). While most of these isoforms did not decrease in intensity substantially during the chase, a concomitant increase and decrease of isoforms was only detected for two components of the acetyl-CoA-carboxyltransferase complex. These were the biotin-carboxylase (AccC) and the carboxyl-transferase-b-subunit (AccD) (see Table 2 Appendix) and both were converted into isoforms with slightly lower pI, suggesting a modification by phosphorylation. Acetyl-CoA-carboxyltransferase is the key enzyme of fatty acid biosynthesis, its activity being controlled by phosphorylation in eucaryotic cells (Witters and Kemp, 1992) . No evidence for modifications of the bacterial enzyme complex has been reported so far.

In contrast to the spots that increased during the chase period, isoforms were only found for 4 of the 15 identified protein spots that decreased during the chase. Two of them, which were also found to be differentially expressed during the cell cycle, were identified as glucan-1,4-beta-glucosidase and ATP-synthase alpha subunit. Their isoforms remained constant during the chase, did not show a significant expression change during the cell cycle and were slightly larger in size (data not shown).


Discussion

Overview

In this study, protein expression and stability were investigated during the cell cycle of the gram-negative bacterium Caulobacter crescentus using a global proteomics approach. Expression and stability of individual proteins were measured by radioactive labeling, separation of the labeled proteins by two-dimensional (2-D) gel electrophoresis, and quantification of the autoradiogram spots. 15% of the reproducibly detected proteins were found to be differentially expressed during the cell cycle, 16% were induced or repressed by the synchronization procedure, 5% were degraded, secreted, processed or modified, and 2.5% were modification or processing products. When these groups of proteins were compared with each other, the striking result was that 54% of the spots that disappeared during the pulse-chase time course were also found to be differentially expressed during the cell cycle. By using mass-spectrometry and immunodetection we were able to identify 131 (42%) of all the spots that were found in any data set. These spots represented 112 different proteins. 81 of the identified proteins were cell cycle dependently expressed and, as expected, some of them were found to be involved in cell division, DNA metabolism, regulatory functions, and development. More surprising was the large number of differentially expressed proteins involved in metabolism, protein synthesis, and protein degradation. 11 proteins were identified that were both differentially expressed during the cell cycle and disappeared during chase, five of which, CtrA, CcrM, FtsZ, FljL, and FljK, have been identified and characterized before (Domian et al., 1997; Kelly et al., 1998; Minnich and Newton, 1987; Quardokus et al., 1996; Quon et al., 1996; Stephens et al., 1996; Zweiger et al., 1994) .

Limitations of the proteomics approach

Only about one fourth of all putative Caulobacter proteins were detected on 2-D gels in this study. There are several possible reasons for this limitation: Not all genes might be expressed under laboratory conditions. Some of the expressed proteins could not be labeled because they do not contain methionine or cystein; others might be expressed at such low levels that the signal disappeared in the background. Some proteins were probably missed out because of the limitations of the 2-D gel conditions used. Only proteins with an isoelectric point (pI) between 3 and 10 and a molecular weight (MW) between 10 and 100 kDa were resolved on the 2-D gels used in this study. Even though this pI and MW range contains most of the E. coli (Link et al., 1997) and most likely also the majority of Caulobacter proteins, a subfraction of proteins was missed out due to the pI and MW restrictions. For example, many ribosomal proteins and the morphogene CicA have a pI above 10 (Fuchs et al., 2001; O'Farrell et al., 1977) . In addition, minor spots were probably not resolved and detected in regions of the 2-D gels with high spot density, because they were masked by bigger spots or disappeared in the generally higher background. Importantly, most membrane proteins were either not solubilized or did not enter into the 2-D gels. Membrane proteins are known to be difficult to detect on 2-D gels using the immobilized pH gradient isoelectric focusing gels applied in this study (Wilkins et al., 1998) . This is supported by the fact that none of the inner membrane proteins could be detected on the 2-D gels with the help of specific antibodies and only 3 of the 131 identified proteins appear to be membrane proteins (TonB-dependent receptors, see Table 1).

With the state-of-the-art technology used to identify the 2-D gel spots, we were able to identify two thirds of all cell cycle expressed proteins and in total about one third of all spots visible on the 2-D gels (Grünenfelder et al., preliminary data). This number corresponds roughly to the number of proteins that have been identified in one of the largest bacterial 2-D gel databases available to date (Langen et al., 2000) . Several factors contributed to the limited yield in spot identification: A) Coomassie-staining was used to visualize the spots on the preparative gels, reducing the sensitivity compared to radioactively labeled protein spots. B) Only two proteases (Trypsin and LysC) can be used for in-gel digestion. Both cut at a lysine residue, Trypsin additionally cuts at an arginine residue. The MALDI-TOF system used for protein identification allows to detect peptides with a size between 900 and 2500 Da. Thus the number of fragments in the suitable size range that could be generated from a protein was restricted by the frequency of only two amino acids. C) Several peptides were difficult to extract from the acrylamide and solubilize for mass determination. D) Because not all peptides are ionized by the matrix assisted laser desorption ionization only a subset of all fragments of the suitable size range were detected by the MALDI-TOF mass spectrometer E) A pre-release of the Caulobacter genome sequence (Nierman, 2001) was used to search protein sequences matching the measured peptides. Due to sequencing gaps and sequencing errors like frameshifts, several proteins might have been missed during the identification process.

The fact that less than one fourth of all cellular proteins can be investigated by proteomic techniques, constitutes a serious disadvantage for expression studies compared to DNA microarrays. However, since proteomics allows to investigate biologically active proteins rather than mRNAs, not only expression but also steady state levels, stability and modification of proteins can be studied. This permits a more complete approach for the analysis of global temporal and spatial control.

Differential cell cycle expression

C. crescentus cells can be synchronized by a simple density gradient centrifugation where swarmer cells are separated from stalked and predivisional cells. Protein expression was studied during the cell cycle by pulse-labeling synchronized cells in G1-phase, early S-phase, late S-phase, G2-phase and after cell division, and by visualizing the labeled proteins on 2-D gels. Using stringent statistical criteria (at least two-fold change in expression between two time points and a statistical confidence level of 99%) 144 spots were found to oscillate during cell cycle, corresponding to 15% of the reproducibly detected minimal data set. Assuming that proteins with cell cycle-regulated expression patterns have the same probability to be detected on 2-D gels as constitutively expressed proteins, and in the light of the fact that we applied stringent statistical criteria, one can conclude that in C. crescentus the expression of at least 15% of all proteins expressed under the conditions used is oscillating during the cell cycle. It is interesting to note that about the same proportion of genes was found to be cell cycle regulated in a study using DNA-microarrays covering almost all genes of the yeast Saccharomyces cerevisiae (Spellman et al., 1998) .

Correlation between mRNA concentration and protein expression

In contrast to earlier publications claiming that mRNA levels and protein expression often do not correlate (VanBogelen et al., 1999) , most of the genes coding for proteins that were found to be differentially expressed during the cell cycle were also identified in a parallel study, which used DNA microarrays to analyze the variation of mRNA levels as a function of the C. crescentus cell cycle (Laub et al., 2000) . While 18 of the corresponding 81 genes were not represented on the microarrays, 49 genes showed cell cycle expression patterns, which were identical or very similar to their pulse labeled products on 2-D gels, and four genes showed an inverse cell cycle expression pattern. 10 of the 81 proteins identified in our proteome study showed no fluctuation on the mRNA level, implying translational or posttranslational cell cycle control (see Fig. 9 Appendix). This important finding leads to several conclusions. A) The strong correlation between data from microarrays and 2-D gels makes assaying changes in mRNA levels a valuable approximation for changes in protein synthesis. B) Differential cell cycle expression of bacterial genes seems to be regulated mainly on the transcriptional level. C) Comparable kinetics of fluctuating mRNA levels and protein expression throughout the cell cycle suggests that the majority of the corresponding mRNA species has a very short half life. D) Oscillating protein expression during the cell cycle can be regulated posttranscriptionally. This has not been observed in bacteria so far and indicates new regulatory mechanisms for cell cycle control. However, we can not exclude that in some cases a misidentification of a particular protein spot is responsible for this observation. It is possible that a protein was erroneously identified by matching several peptides by chance or, more likely, that a wrong spot was cut out from the Coomassie-stained gel and analyzed.

Why differential cell cycle expression?

What is the functional significance of cell cycle regulated protein expression? A) The cell could save resources by only producing certain proteins when they are mostly needed. This could include proteins whose functions are only needed during part of the cell cycle for processes like DNA replication or cell division. B) The cell cycle-specific expression might be required for fine-tuning the activity of regulatory proteins that act as periodic switches during the cell cycle. An excellent example is the transcriptional regulator CtrA, which is required for cells to proceed through the S- and G1-phases but inhibits the G1-to-S transition (Quon et al., 1996) . The activity of CtrA is temporally regulated by cell cycle-specific expression, modification, and degradation (Domian et al., 1997; Quon et al., 1996) to ensure the proper timing of action under any circumstance. C) Cell cycle regulated expression could be required to build a structure in a highly controlled and entropically favorable way. This is, for instance, the case for the flagellar proteins, which are expressed towards the end of the Caulobacter cell cycle in a three-tired expression cascade resulting in the ordered assembly of a single flagellum at one pole of the predivisional cell. Expression is coordinated with flagellar assembly by two distinct checkpoints of structural intermediates. Thus, the expression of a given class of flagellar genes requires the physical presence and correct assembly of all components encoded by the preceding class (Wu and Newton, 1997) . D) Differential cell cycle expression could be a simple indicator of the progress of the cell differentiation program, as differentiation is intimately connected with cell proliferation in Caulobacter. A good example is the differential expression of the flagellar, pili and chemotaxis proteins (Alley et al., 1991; Skerker and Shapiro, 2000; Wu and Newton, 1997) .

Synexpression groups

We applied cluster analysis (Kaufman and Rousseeuw, 1990) to sort the 144 differentially expressed protein spots. The clustering algorithm used grouped the spots into 23 clusters with similar cell cycle expression. Co-expressed proteins may belong to synexpression groups, which represent sets of proteins that share a complex expression pattern under different conditions and that function in the same process (Eisen et al., 1998; Niehrs and Pollet, 1999) . In S. cerevisiae, for example, about half of the cell cycle-regulated genes could be assigned to nine clusters, members of which not only shared common promoter elements but were also functionally related (Spellman et al., 1998) . This finding supported the suggestion that members of synexpression groups belong to the same physiological pathways.

In agreement with this cluster analysis was indeed able to group co-expressed and functionally related C. crescentus proteins: The chaperones GroEL and GroES, which are co-transcribed from an operon (Avedissian and Lopes Gomes, 1996) , were grouped in one expression cluster with low expression during S-phase. Four enzymes of riboflavin biosynthesis were identified, which appear to be encoded in an operon. All of them were found in expression cluster 6 with high expression in G1-phase. Additionally a protein involved in tetrahydrofolate synthesis was also part of that expression cluster. Riboflavin and tetrahydrofolate are growth factors involved in redox reactions and in the synthesis of building blocks like purines and certain amino acids. Their induction in swarmer cells might reflect the need of this cell type to prepare for the upcoming proliferation phase of the division cycle.

Proteins involved in energy metabolism were found in two similar clusters (clusters 22 and 23) with induction early S-phase just when replication and cell growth is initiated. This could indicate a change in metabolism between the swarmer and the stalked cell state and the increased demand for energy at the onset of cell growth an replication.

Another example for functionally related co-expressed proteins are the chemotaxis proteins CheYI, CheD, CheB and CheR, which are encoded in a large chemotaxis operon (Nierman, 2001) , and the flagellin FljL (clusters 11 and 13). A second flagellin, FljK, was also identified with a slightly different expression pattern (cluster 4). These proteins, however, account for only a small fraction of the motor and chemotaxis components identified in C. crescentus (Ely and Ely, 1989; Wu and Newton, 1997) . This might be due to the overall low abundance of most flagellar components in this uniflagellated organism and to the fact that a large fraction of flagellar components and all chemoreceptors are membrane-integral proteins, which, due to their low solubility, could not be resolved in the first dimension of the 2-D gels. In agreement with this CheB, CheD, CheR, CheYI, FljK, and FljL are all soluble proteins.

Replication and cell division proteins

DNA replication genes share common promoters in C. crescentus and are predominantly transcribed in early S-phase (Winzeler and Shapiro, 1996) . We could only identify a single replication protein, the single strand binding protein SSB, which indeed was mainly expressed in early S-phase. Other replication factors were not identified, either because they are not differentially expressed or because they are present at very low concentrations in the cell (Kornberg and Baker, 1992) . We also expected to find several co-expressed cell division proteins. However, we were only able to identify FtsZ (Kelly et al., 1998) . Since FtsZ is the only identified protein of its cluster 20 it is possible that several of the other co-expressed proteins are also involved in cell division. Plausible explanations for the finding that only one cell division factor was identified are that most cytokinesis proteins, of which FtsZ is by far the most abundant one, are present at very low concentrations in the cell (Rothfield et al., 1999) and that the membrane-integral or membrane-associated cell division components were lost during the extraction for the preparative gels.

Physiology and cell cycle

One of the surprising results was that 19 (23%) of the identified proteins with differential cell cycle expression patterns were assigned to metabolic functions. Four proteins were found to be involved in amino acid metabolism (tryptophane, arginine, and methionine synthesis and formation of glutamine-tRNA by transamidation), three in carbohydrate metabolism, five in cofactor metabolism (riboflavin and tetrahydrofolate), and two in lipid metabolism (beta-oxidation of fatty acids). A transcriptional regulator of the Lrp/AsnC family was upregulated in early S phase. Homologs of this protein act as global metabolic regulators and are involved in the control of amino acid metabolism in a number of different bacteria, making this protein a candidate for cell cycle control of amino acid biosynthesis (Beloin et al., 1997; Inoue et al., 1997; Newman and Lin, 1995; Peekhaus et al., 1995) .

A relatively high number of proteolytic enzymes were found to be differentially expressed during the cell cycle. Nine peptidases and two proteases were identified, most of which (5) were mainly expressed in the swarmer cell. Interestingly, all but one of these proteolytic enzymes have a predicted signal peptide for export from the cytoplasm (see Table 2, Appendix). The only exception was the proteolytic subunit of the cytoplasmic Clp protease, ClpP (Jenal and Fuchs, 1998) . Thus, most of the proteolytic enzymes that were found to be differentially expressed are either periplasmic or secreted out of the cell. This interpretation implies that they are mainly involved in the degradation of extracellular proteins and peptides, and probably required for nutrient scavenging. Because swarmer cells are in the adventurous planktonic phase of the Caulobacter life cycle they are maybe more dependent on elaborate nutrient uptake systems than sessile stalked cells.

The expression of several proteins involved in oxidative stress response, such as GTP cyclohydrolase II, thioredoxin, thioredoxin reductase, glutathione-S-transferase, and the ferric uptake regulator Fur (Storz and Imlay, 1999; Touati, 2000) was also under cell cycle control. Oxidative stress can result from hydroxide radicals, which are generated in the presence of Fe2+ and O2 (Touati, 2000) . To avoid DNA damage the cell tightly controls its iron metabolism. Fur negatively controls iron acquisition and import genes (Touati, 2000) and is induced in early S phase, while two iron uptake proteins were repressed in S phase and induced in G1 or G2 (Table 1, Fig. 2). Thus, one could speculate that the cell needs to keep the iron concentration low in S phase to prevent DNA damage during ongoing replication. Alternatively, induction of iron uptake in G1 and G2 could reflect a metabolic peculiarity of the planktonic swarmer cell type as having a specialized role in nutrient scavenging. Moreover, many of the proteins involved in oxidative stress require FAD as a cofactor. As mentioned above riboflavin (a precursor of FAD) synthesis genes are also differentially expressed during the cell cycle.

It is interesting to note that a large group of nutritional genes involved in amino acid and sugar metabolism as well as iron uptake are also under cell cycle control in yeast (Spellman et al., 1998) . The requirement for periodic induction of key metabolic pathways during growth is apparently conserved in both prokaryotic and eukaryotic cells. In some cases this probably reflects the higher need of certain building blocks during a specific time of the cell cycle. For example, nucleotide synthesis genes are induced before or at the start of replication (Laub et al., 2000; Spellman et al., 1998) . Methionine synthesis genes are induced during and after S-phase (Laub et al., 2000; Spellman et al., 1998) . Methionine is the precursor of S-adenosyl-methionine which acts as a methyl-donor in several reactions. Probably the best studied case is DNA-methylation, which has been shown to be important for processes like replication initiation, transcription regulation, mismatch repair and recombination in many eukaryotic and prokaryotic species (Colot and Rossignol, 1999; Reisenauer et al., 1999) . Restoring the original methylation status after DNA replication is probably critical for cell proliferation which could be one reason for the differential expression of methionine synthesis genes. Interestingly, yeast does not have a detectable DNA-methylation system (Colot and Rossignol, 1999) . This fact indicates that other methylation reactions may also be temporally controlled during the cell cycle.

Protein degradation

Differential protein expression combined with a short half life of the corresponding protein would result in characteristic fluctuations of protein concentration during the cell cycle. Since drastic changes in protein level during the cell cycle could have important functional or regulatory implications we decided to investigate the stability of all detectable protein spots in pulse-chase experiments. A total of 48 spots were significantly decreased and 24 spots increased during the chase period, which corresponded to the length of one cell cycle equivalent. The latter are either modification or processing products that appear very slowly or in response to changing conditions during the chase period. In support of this conclusion, isoforms were found for 8 of the 10 identified protein spots accumulating during chase. In contrast, isoforms were found for only 4 of the 15 identified protein spots that decreased during the chase. This result argues that most of disappearing spots are not substrates for modification or processing reactions. Rather, the majority of these 48 spots represent proteins that are degraded in the course of one division unit. By extrapolating this conclusion to the entire C. crescentus proteome, an estimated 5% of the cell's proteins have a half life of one cell cycle equivalent or shorter. In E. coli a similar fraction of proteins were found to disappear during a pulse-chase time course (Mosteller et al., 1980) . However, only half lifes longer that two hours were detected for cells grown in single strength minimal medium E with a doubling time of 72 minutes. Thus, this is the first global study that defines a comprehensive catalog of unstable bacterial proteins with half lifes less than or one generation time.

It is important to note that a disappearing spot could also reflect the protein being secreted from the cell during the chase period. This is certainly the case for the components of the flagellar filament, among them FljL and FljK, which are released into the medium when the flagellum is shed (Jenal, 2000) . In order to distinguish between protein export and degradation, the proteins secreted by the cells would have to be collected from the medium and separated on 2-D gels. It could then be tried to match the resulting spots with the 2-D gel pattern of the total cellular protein extract. However, finding no matches with the secreted spots would not necessarily mean that none of the disappearing spots would be secreted. Because many proteins, the flagellar components excluded, are processed during their secretion (Izard and Kendall, 1994) , they would probably not migrate to the same positions as their cytoplasmic precursors in most cases. Thus, only the identification of all disappearing and secreted proteins by mass spectrometry would finally allow to identify the secreted members among the disappearing spots.

Comparison of the cell cycle expression, the synchronization and the pulse-chase data sets with each other revealed that several spots were present in more than one data set. 48 of the 144 cell cycle expression spots, 30 of the 154 synchronization procedure affected spots, 2 of the 24 spots increasing during chase and 31 of the 48 spots decreasing during chase were present in at least one other data set (Table 2, Appendix). The fact that a strikingly high number of spots that disappeared during chase were also found in one of the other two data sets is most likely of high biological significance, and might reflect the important regulatory function of selective proteolysis. This seems especially important during the cell cycle because 26 of the 48 chase disappearing spots were also found to be differentially expressed during the cell cycle. Even more interesting was the observation that these spots were concentrated in 9 of the 23 expression clusters and in particular in expression clusters 4, 10, 11, 13, 15, 20, where over 30% of the spots disappeared during chase. The latter are characterized by sharp peaks of expression in G1, S or G2-phase. One can assume that narrow windows of expression, in combination with rapid degradation, result in distinct periodic changes of protein concentrations during the cell cycle. The highest proportion of presumably unstable proteins was found in cluster 11 where 7 of the 11 spots disappeared during the chase period. This accumulation of unstable spots in a few expression clusters probably indicated that specific proteolysis is restricted to a few functional groups of proteins with differential expression during the cell cycle.

The proteins required for motility and chemotaxis belong certainly to one of these functional groups: The chemotaxis proteins CheYI and CheD as well as the flagellin FljL were identified for three unstable spots of expression cluster 13, which is characterized by a sharp expression peak late in cell cycle. The flagellin FljK was identified for both unstable spots of the expression cluster 4, which is characterized by high expression both late and early in cell cycle. The disappearance of these proteins reflects the cell differentiation step from the planktonic to the sessile stage of the Caulobacter life cycle. During this cell differentiation step the flagellar anchor protein FliF and the chemoreceptor McpA are degraded (Alley et al., 1993; Jenal and Shapiro, 1996) and the flagellum is ejected into the supernatant (Jenal, 2000) . The disappearance of the spots representing the flagellins FljL and FljK certainly reflects the flagellar ejection. However, a recent report found indications that the intracellular pool of FljK might be degraded concomitant with the flagellar ejection (Mangan et al., 1999) . Therefore the disappearance of the flagellin spots might be a consequence of both flagellar ejection and degradation of the intracellular pool. The finding that two soluble chemotaxis proteins were unstable is the first indication that the cell degrades also other components of the chemotaxis machinery besides the membrane integral chemoreceptor McpA. However, for both CheD and CheYI, only half of the protein amount was degraded during the chase time of one cell cycle length. Moreover, other soluble chemotaxis proteins, CheB and CheR, were identified that were not degraded at all. This result indicates that the cell only degrades some soluble chemotaxis proteins. The reason why some soluble chemotaxis proteins are unstable and some not is not clear. It might be that the unstable components are harmful for the sessile stalked cell and therefore have to be removed. In contrast, some of the chemotaxis proteins might have another, so far unknown function in the sessile stalked cell and are therefore not removed.

What other protein groups could be subject to both cell cycle regulated expression and specific proteolysis is not clear. The data is not complete enough to draw any conclusions because most of the unstable proteins could not be identified, most likely due to their low abundance. Nevertheless, we speculate that one of these possible groups could be the cell division proteins. The best characterized member of these proteins, FtsZ, was the only identified protein of the expression cluster 20. This cluster is characterized by a sharp peak in early S phase and by a high amount of unstable protein spots, among them the one representing FtsZ. It is therefore possible that other spots of this cluster could also represent cell division proteins.

Conclusion

Studying cell cycle regulated protein expression and degradation using a global proteomics approach has provided new insights into the complexity of the bacterial cell cycle. The unexpected large number of proteins synthesized at a specific stage of the reproductive cycle suggested that periodic protein expression is critical for the cell either to guarantee the optimal utilization of resources or to maintain the proper order and functioning of the cell cycle. Differentially expressed proteins were found to be involved in many different aspects of the cell's metabolism. A large subgroup of periodically synthesized proteins were of unknown function (Table 1) and represent candidates for novel regulators of the C. crescentus cell cycle or development. The strong correlation between protein turnover and differential synthesis indicated that one of the main reasons for rapid protein degradation in bacteria is to maintain the periodicity required for ordered cell cycle progression. This is the first evidence for a global role of proteolysis in bacterial cell cycle control. In eukaryotic cells the controlled proteolysis of key proteins at specific time points plays an essential role in promoting irreversible steps during cell cycle progression (King et al., 1996) . Based on our results we postulate that specific and controlled proteolysis plays a similar role in bacteria. It will be of particular interest to determine the identity of all proteins found to be both unstable and differentially expressed, as some might have critical functions in cell cycle progression. Characterization of these proteins will deepen our understanding of the molecular basis of bacterial growth. In an age of reemerging bacterial diseases, detailed knowledge of all regulatory processes involved in bacterial proliferation will be indispensable for the development of novel antimicrobial strategies.


References

Aldridge, P., and Jenal, U. (1999). Cell cycle-dependent degradation of a flagellar motor component requires a novel-type response regulator. Mol. Microbiol. 32, 379-391.

Alley, M. R., Gomes, S. L., Alexander, W., and Shapiro, L. (1991). Genetic analysis of a temporally transcribed chemotaxis gene cluster in Caulobacter crescentus. Genetics 129, 333-341.

Alley, M. R., Maddock, J. R., and Shapiro, L. (1993). Requirement of the carboxyl terminus of a bacterial chemoreceptor for its targeted proteolysis. Science 259, 1754-1757.

Alley, M. R. K., Maddock, J. R., and Shapiro, L. (1992). Polar localization of a bacterial chemoreceptor. Genes and Development 6, 825-836.

Altschul, S. F., Madden, T. L., Schaffer, A. A., Zhang, J., Zhang, Z., Miller, W., and Lipman, D. J. (1997). Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic Acids Res. 25, 3389-3402.

Avedissian, M., and Lopes Gomes, S. (1996). Expression of the groESL operon is cell-cycle controlled in Caulobacter crescentus. Mol. Microbiol. 19, 79-89.

Beloin, C., Ayora, S., Exley, R., Hirschbein, L., Ogasawara, N., Kasahara, Y., Alonso, J. C., and Hegarat, F. L. (1997). Characterization of an lrp-like (lrpC) gene from Bacillus subtilis. Mol. Gen. Genet. 256, 63-71.

Berndt, P., Hobohm, U., and Langen, H. (1999). Reliable automatic protein identification from matrix-assisted laser desorption/ionization mass spectrometric peptide fingerprints. Electrophoresis 20, 3521-3526.

Bjellqvist, B., Pasquali, C., Ravier, F., Sanchez, J. C., and Hochstrasser, D. (1993). A nonlinear wide-range immobilized pH gradient for two-dimensional electrophoresis and its definition in a relevant pH scale. Electrophoresis 14, 1357-1365.

Cho, R. J., Campbell, M. J., Winzeler, E. A., Steinmetz, L., Conway, A., Wodicka, L., Wolfsberg, T. G., Gabrielian, A. E., Landsman, D., Lockhart, D. J., and Davis, R. W. (1998). A genome-wide transcriptional analysis of the mitotic cell cycle. Mol. Cell 2, 65-73.

Colot, V., and Rossignol, J. L. (1999). Eukaryotic DNA methylation as an evolutionary device. Bioessays 21, 402-411.

Domian, I. J., Quon, K. C., and Shapiro, L. (1997). Cell type-specific phosphorylation and proteolysis of a transcriptional regulator controls the G1-to-S transition in a bacterial cell cycle. Cell 90, 415-424.

Eisen, M. B., Spellman, P. T., Brown, P. O., and Botstein, D. (1998). Cluster analysis and display of genome-wide expression patterns. Proc. Natl. Acad. Sci. U S A 95, 14863-14868.

Ely, B., and Ely, T. W. (1989). Use of pulsed field gel electrophoresis and transposon mutagenesis to estimate the minimal number of genes required for motility in Caulobacter crescentus. Genetics 123, 649-654.

Fountoulakis, M., and Langen, H. (1997). Identification of proteins by matrix-assisted laser desorption ionization-mass spectrometry following in-gel digestion in low-salt, nonvolatile buffer and simplified peptide recovery. Anal. Biochem. 250, 153-156.

Fuchs, T., Wiget, P., Osteras, M., and Jenal, U. (2001). Precise amounts of a novel member of a phosphotransferase superfamily are essential for growth and normal morphology in Caulobacter crescentus. Mol. Microbiol. 39, 679-692.

Garrels, J. I. (1989). The QUEST system for quantitative analysis of two-dimensional gels. J. Biol. Chem. 264, 5269-5282.

Gottesman, S. (1996). Proteases and their targets in Escherichia coli. Annu. Rev. Genet. 30, 465-506.

Hecht, G. B., Lane, T., Ohta, N., Sommer, J. M., and Newton, A. (1995). An essential single domain response regulator required for normal cell division and differentiation in Caulobacter crescentus. EMBO J. 14, 3915-3924.

Hecht, G. B., and Newton, A. (1995). Identification of a novel response regulator required for the swarmer- to-stalked-cell transition in Caulobacter crescentus. J. Bacteriol. 177, 6223-6229.

Inoue, H., Inagaki, K., Eriguchi, S. I., Tamura, T., Esaki, N., Soda, K., and Tanaka, H. (1997). Molecular characterization of the mde operon involved in L-methionine catabolism of Pseudomonas putida. J. Bacteriol. 179, 3956-3962.

Izard, J. W., and Kendall, D. A. (1994). Signal peptides: exquisitely designed transport promoters. Mol. Microbiol. 13, 765-773.

Jacobs, C., Domian, I. J., Maddock, J. R., and Shapiro, L. (1999). Cell cycle-dependent polar localization of an essential bacterial histidine kinase that controls DNA replication and cell division. Cell 97, 111-120.

Jenal, U. (2000). Signal transduction mechanisms in Caulobacter crescentus development and cell cycle control. FEMS Microbiology Reviews 24, 177-191.

Jenal, U., and Fuchs, T. (1998). An essential protease involved in bacterial cell cycle control. EMBO J. 17, 5658-5669.

Jenal, U., and Shapiro, L. (1996). Cell cycle-controlled proteolysis of a flagellar motor protein that is asymmetrically distributed in the Caulobacter predivisional cell. EMBO J. 15, 2393-2406.

Jenal, U., White, J., and Shapiro, L. (1994). Caulobacter flagellar function, but not assembly, requires FliL, a non- polarly localized membrane protein present in all cell types [published erratum appears in J Mol Biol 1995 May 12;248(4):883]. J. Mol. Biol. 243, 227-244.

Johnson, R. C., and Ely, B. (1977). Isolation of spontaneously derived mutants of Caulobacter crescentus. Genetics 86, 25-32.

Kaufman, L., and Rousseeuw. (1990). Finding groups in data (New York: Wiley).

Kelly, A. J., Sackett, M. J., Din, N., Quardokus, E., and Brun, Y. V. (1998). Cell cycle-dependent transcriptional and proteolytic regulation of FtsZ in Caulobacter. Genes Dev. 12, 880-893.

King, R. W., Deshaies, R. J., Peters, J. M., and Kirschner, M. W. (1996). How proteolysis drives the cell cycle. Science 274, 1652-1659.

Kornberg, A., and Baker, T. A. (1992). DNA Replication, 2 Edition (New York: W. H. Freeman).

Langen, H., Takacs, B., Evers, S., Berndt, P., Lahm, H. W., Wipf, B., Gray, C., and Fountoulakis, M. (2000). Two-dimensional map of the proteome of Haemophilus influenzae. Electrophoresis 21, 411-429.

Larrabee, K. L., Phillips, J.O., Williams, G.J., and Larrabee, A.R. (1980). The relative rates of protein synthesis and degradation in a growing culture of Escherichia coli. J. Biol. Chem. 255, 4125-4130.

Laub, M. T., McAdams, H. H., Feldblyum, T., Fraser, C. M., and Shapiro, L. (2000). Global analysis of the genetic network controlling the Caulobacter cell cycle. Science 290, 2144-2148.

Link, A. J., Robison, K., and Church, G. M. (1997). Comparing the predicted and observed properties of proteins encoded in the genome of Escherichia coli K-12. Electrophoresis 18, 1259-313.

Mangan, E. K., Malakooti, J., Caballero, A., Anderson, P., Ely, B., and Gober, J. W. (1999). FlbT couples flagellum assembly to gene expression in caulobacter crescentus [In Process Citation]. J. Bacteriol. 181, 6160-6170.

Marczynski, G. T. (1999). Chromosome methylation and measurement of faithful, once and only once per cell cycle chromosome replication in Caulobacter crescentus. J. Bacteriol. 181, 1984-1993.

Milhausen, M., and Agabian, N. (1981). Regulation of polypeptide synthesis during Caulobacter development: two-dimensional gel analysis. J. Bacteriol. 148, 163-173.

Minnich, S. A., and Newton, A. (1987). Promoter mapping and cell cycle regulation of flagellin gene transcription in Caulobacter crescentus. Proc. Natl. Acad. Sci. U S A 84, 1142-1146.

Mosteller, R. D., Goldstein, R. V., and Nishimoto, K. R. (1980). Metabolism of individual proteins in exponentially growing Escherichia coli. J. Biol. Chem. 255, 2524-2532.

Nasmyth, K. (1996). Viewpoint: putting the cell cycle in order. Science 274, 1643-1645.

Newman, E. B., and Lin, R. (1995). Leucine-responsive regulatory protein: a global regulator of gene expression in E. coli. Annu. Rev. Microbiol. 49, 747-775.

Niehrs, C., and Pollet, N. (1999). Synexpression groups in eukaryotes. Nature 402, 483-487.

Nielsen, H., Engelbrecht, J., Brunak, S., and von Heijne, G. (1997). Identification of prokaryotic and eukaryotic signal peptides and prediction of their cleavage sites. Protein Eng. 10, 1-6.

Nierman, et al. (2001). Complete genome sequence of Caulobacter crescentus. Proc. Natl. Acad. Sci. USA 98, 4136-4141.

Nigg, E. A. (1995). Cyclin-dependent protein kinases: key regulators of the eukaryotic cell cycle. Bioessays 17, 471-480.

O'Farrell, P. Z., Goodman, H. M., and O'Farrell, P. H. (1977). High resolution two-dimensional electrophoresis of basic as well as acidic proteins. Cell 12, 1133-1141.

Ohta, N., Lane, T., Ninfa, E. G., Sommer, J. M., and Newton, A. (1992). A histidine protein kinase homologue required for regulation of bacterial cell division and differentiation. Proc. Natl. Acad. Sci. USA 89, 10297-10301.

Osteras, M., and Jenal, U. (2000). Regulatory circuits in Caulobacter. Curr. Opin. Microbiol. 3, 171-176.

Osteras, M., Stotz, A., Schmid Nuoffer, S., and Jenal, U. (1999). Identification and transcriptional control of the genes encoding the Caulobacter crescentus ClpXP protease. J. Bacteriol. 181, 3039-3050.

Peekhaus, N., Tolner, B., Poolman, B., and Kramer, R. (1995). The glutamate uptake regulatory protein (Grp) of Zymomonas mobilis and its relation to the global regulator Lrp of Escherichia coli. J. Bacteriol 177, 5140-5147.

Pine, M. (1970). Steady-state measurement of the turnover of amino acid in the cellular proteins of growing Escherichia coli: Existence of two kinetically distinct reactions. J. Bacteriol. 103, 207-215.

Quardokus, E., Din, N., and Brun, Y. V. (1996). Cell cycle regulation and cell type-specific localization of the FtsZ division initiation protein in Caulobacter. Proc. Natl. Acad. Sci. U S A 93, 6314-6319.

Quon, K. C., Marczynski, G. T., and Shapiro, L. (1996). Cell cycle control by an essential bacterial two-component signal transduction protein. Cell 84, 83-93.

Quon, K. C., Yang, B., Domian, I. J., Shapiro, L., and Marczynski, G. T. (1998). Negative control of bacterial DNA replication by a cell cycle regulatory protein that binds at the chromosome origin. Proc. Natl. Acad Sci. USA 95, 120-125.

Reisenauer, A., Kahng, L. S., McCollum, S., and Shapiro, L. (1999). Bacterial DNA methylation: a cell cycle regulator? J. Bacteriol. 181, 5135-5139.

Reisenauer, A., Quon, K., and Shapiro, L. (1999). The CtrA response regulator mediates temporal control of gene expression during the Caulobacter cell cycle. J. Bacteriol. 181, 2430-2439.

Rothfield, L., Justice, S., and Garcia-Lara, J. (1999). Bacterial cell division. Annu Rev Genet 33, 423-448.

Sackett, M. J., Kelly, A. J., and Brun, Y. V. (1998). Ordered expression of ftsQA and ftsZ during the Caulobacter crescentus cell cycle. Mol. Microbiol. 28, 421-434.

Skerker, J. M., and Shapiro, L. (2000). Identification and cell cycle control of a novel pilus system in Caulobacter crescentus. EMBO J. 19, 3223-3234.

Spellman, P. T., Sherlock, G., Zhang, M. Q., Iyer, V. R., Anders, K., Eisen, M. B., Brown, P. O., Botstein, D., and Futcher, B. (1998). Comprehensive identification of cell cycle-regulated genes of the yeast Saccharomyces cerevisiae by microarray hybridization. Mol. Biol. Cell 9, 3273-3297.

Stephens, C., Reisenauer, A., Wright, R., and Shapiro, L. (1996). A cell cycle-regulated bacterial DNA methyltransferase is essential for viability. Proc. Natl. Acad. Sci. USA 93, 1210-1214.

Stephens, C. M., and Shapiro, L. (1993). An unusual promoter controls cell-cycle regulation and dependence on DNA replication of the Caulobacter fliLM early flagellar operon. Mol. Microbiol. 9, 1169-1179.

Storz, G., and Imlay, J. A. (1999). Oxidative stress. Curr. Opin. Microbiol. 2, 188-194.

Touati, D. (2000). Iron and oxidative stress in bacteria. Arch. Biochem. Biophys. 373, 1-6.

VanBogelen, R. A., Greis, K. D., Blumenthal, R. M., Tani, T. H., and Matthews, R. G. (1999). Mapping regulatory networks in microbial cells. Trends Microbiol. 7, 320-328.

Vohradsky, J., Li, X. M., and Thompson, C. J. (1997). Identification of procaryotic developmental stages by statistical analyses of two-dimensional gel patterns. Electrophoresis 18, 1418-28.

Wallin, E., and von Heijne, G. (1998). Genome-wide analysis of integral membrane proteins from eubacterial, archaean, and eukaryotic organisms. Protein Sci. 7, 1029-1038.

Wang, S. P., Sharma, P. L., Schoenlein, P. V., and Ely, B. (1993). A histidine protein kinase is involved in polar organelle development in Caulobacter crescentus. Proc. Natl. Acad. Sci. USA 90, 630-634.

Wheeler, C. H., Berry, S. L., Wilkins, M. R., Corbett, J. M., Ou, K., Gooley, A. A., Humphery-Smith, I., Williams, K. L., and Dunn, M. J. (1996). Characterisation of proteins from two-dimensional electrophoresis gels by matrix-assisted laser desorption mass spectrometry and amino acid compositional analysis. Electrophoresis 17, 580-587.

Wheeler, R., and Shapiro, L. (1999). Differential localization of two histidine kinases controlling bacterial cell differentiation. Molecular Cell 4, 683-694.

Wilkins, M. R., Gasteiger, E., Sanchez, J. C., Bairoch, A., and Hochstrasser, D. F. (1998). Two-dimensional gel electrophoresis for proteome projects: the effects of protein hydrophobicity and copy number. Electrophoresis 19, 1501-1505.

Winzeler, E., and Shapiro, L. (1996). A novel promoter motif for Caulobacter cell cycle-controlled DNA replication genes. J. Mol. Biol. 264, 412-425.

Winzeler, E., and Shapiro, L. (1995). Use of flow cytometry to identify a Caulobacter 4.5 S RNA temperature- sensitive mutant defective in the cell cycle. J. Mol. Biol. 251, 346-365.

Witters, L. A., and Kemp, B. E. (1992). Insulin activation of acetyl-CoA carboxylase accompanied by inhibition of the 5'-AMP-activated protein kinase. J. Biol. Chem. 267, 2864-2867.

Wortinger, M., Sackett, M. J., and Brun, Y. V. (2000). CtrA mediates a DNA replication checkpoint that prevents cell division in caulobacter crescentus [In Process Citation]. EMBO J. 19, 4503-4512.

Wright, R., Stephens, C., Zweiger, G., Shapiro, L., and Alley, M. R. (1996). Caulobacter Lon protease has a critical role in cell-cycle control of DNA methylation. Genes Dev. 10, 1532-1542.

Wu, J., and Newton, A. (1997). Regulation of the Caulobacter flagellar gene hierarchy; not just for motility. Mol. Microbiol. 24, 233-239.

Yates, J. R. (2000). Mass spectrometry. From genomics to proteomics. Trends Genet. 16, 5-8.

Zweiger, G., Marczynski, G., and Shapiro, L. (1994). A Caulobacter DNA methyltransferase that functions only in the predivisional cell. J. Mol. Biol. 235, 472-485.

Zweiger, G., and Shapiro, L. (1994). Expression of Caulobacter dnaA as a function of the cell cycle. J. Bacteriol. 176, 401-408.