Identification of sporulation genes by genome-wide analysis of the {sigma}E regulon of Bacillus subtilis

Starvation induces the Gram-positive bacterium Bacillus subtilis to initiate a simple, two-cell developmental process that results in the formation of dormant spores. Early in sporulation, the developing cell divides asymmetrically to produce a smaller compartment, the prespore, which becomes the spore, and a larger compartment, the mother cell, which participates in the maturation of the spore and finally lyses to release it. Differential gene expression in the two cells is governed by the tightly regulated synthesis and activation of a series of sporulation-specific transcription factors. These sigma factors redirect RNA polymerase to specific new sets of promoters. The first σ-factor, σ^F, becomes active only in the prespore and is required also for the activation of the first mother-cell-specific σ-factor, σ^E. Then σ^F and σ^E are later replaced by σ^G and σ^K respectively. Intercompartmental signal transduction pathways link the activation of σ^E and σ^K in the mother cell to the action of σ^F and σ^G in the prespore, respectively, thereby ensuring the correct sequence of gene expression in the two compartments (reviewed by Kroos & Yu, 2000; Piggot & Losick, 2002).

The early mother-cell-specific σ-factor, σ^E, is encoded by the sigE (spoIIGB) gene and is expressed upon the initiation of sporulation from a promoter that is recognized by RNA polymerase containing the housekeeping σ-factor σ^A, in conjunction with SpoOA, the key regulator for entry into sporulation (Kenney & Moran, 1987; Satola et al., 1991). σ^E-dependent gene expression is, however, not observed until 2 h after the initiation of sporulation as the protein is synthesized in an inactive form, which must undergo proteolytic cleavage to become active (LaBell et al., 1987). The protease responsible for the cleavage is probably SpoIIGA (Jonas et al., 1988; Stragier et al., 1988), encoded by the gene spoIIGA upstream of, and co-transcribed with, sigE. Cleavage of pro-σ^E occurs only in the mother cell and requires the SpoIIR protein, which is expressed in the prespore from a σ^F-dependent promoter, thereby coupling the activation of σ^E to σ^F activity (Karow et al., 1995; Londono-Vallejo & Stragier, 1995). As the first σ-factor to become active in the mother cell, σ^E is responsible for the expression of the genes encoding the late mother cell σ-factor, σ^K, and SpoIIID, a transcription factor required for fine-tuning the regulation of many σ^E-dependent genes (reviewed by Errington, 1993). Other σ^E-dependent genes are involved in the formation of the spore coat and cortex and some are necessary for proper germination (reviewed by Piggot & Losick, 2002). At least one σ^E-dependent gene is involved in the regulation of transcription of sigG, encoding the late prespore-specific σ-factor σ^G, because mutations in the spoIIG operon block transcription at the sigG promoter (Partridge & Errington, 1993).

DNA microarrays are increasingly being used for transcriptional profiling in various organisms. This technique enables an overview of which genes are being expressed under particular conditions and can produce vast quantities of informative data. Recently, DNA microarrays have been used to identify genes of Bacillus subtilis dependent on σ^B (Price et al., 2001), σ^H (Britton et al., 2002), SpoOA and σ^F (Fawcett et al., 2000) and CodY (Molle et al., 2003).

Here we used DNA microarrays to compare the transcriptional profile of two different σ^E mutants to the profile found in wild-type cells. In addition to the identification of previously known σ^E-dependent genes, we were able to assign 124 additional genes, of which 88 are organized in operons, to the σ^E regulon. Furthermore, disruption of some of the previously unknown genes resulted in a defect in sporulation.

Plasmids and bacterial strains.
Table 1 lists the plasmids and bacterial strains used. DNA manipulations and Escherichia coli transformations were carried out using standard methods (Sambrook et al., 1989). B. subtilis cells were made competent for transformation with DNA as described previously (Anagnostopoulos & Spizizen, 1961; Jenkinson, 1983). Disruptions of cotZ, ybfJ, ydcC, ydhF, yhbH, yjaV, yndA, yodP, yqfD, yqhV, yrzE, ysnE and ywcB were created by single-crossover (Campbell) integration. The central portion of the coding region (at least 250 bp) was amplified by PCR from SG38 chromosomal DNA and cloned into pSG1164. The following oligonucleotides were used for PCR: COTZ1 (5'-CGGAATTCGCTGCGTGCGTGAAGC-3'), COTZ2 (5'-CCGCTCGAGACACACCGTATCTTTATCGG-3'), YBFJ1 (5'-GGGGTACCTGCGGGTGCAGTTCGC-3'), YBFJ2 (5'-CGGGATCCGTAATCGCCTTTTCTGCCGG-3'), YDCC1 (5'-GGGGTACCATGACCATCGAGACAGGG-3'), YDCC2 (5'-CGGGATCCAGCGGCGTTTTGACCGC-3'), YDHF1 (5'-CGGAATTCCAGGTAACTGATGACCGG-3'), YDHF2 (5'-CCGCTCGAGTATCATGAGCATGGCCGC-3'), YHBH1 (5'-CGGAATTCGGCAAGGAGACGGCG-3'), YHBH2 (5'-CCGCTCGAGCAAACCCCCATACTGCC-3'), YJAV1 (5'-CGGAATTCCGGACGAATTTTTCGGCC-3'), YJAV2 (5'-CCGCTCGAGCACGGTTTTTTCCGAGCC-3'), YNDA1 (5'-GGGGTACCACTTACTCAAGTCCGACGG-3'), YNDA2 (5'-CGGGATCCCATCAGGATTGTCACCGGG-3'), YODP1 (5'-CGGAATTCTATTTGCGCCGGAGGGC-3'), YODP2 (5'-CCGCTCGAGTTGTTAACGAATGGCCGCG-3'), YQFD1 (5'-GGGGTACCCATGCCTTTCGGCGGG-3'), YQFD2 (5'-CGGGATCCGACAATATTGCGCGGGC-3'), YQHV1 (5'-GGGGTACCTGTTTTAACAATGGCGGG-3'), YQHV2 (5'-CGGGATCCAAAATCAAGACAACACCCG-3'), YRZE1 (5'-CGGAATTCGAGGAGACAATCATGGAGG-3'), YRZE2 (5'-CCGCTCGAGGCCGAAAATCCCGCCC-3'), YSNE1 (5'-CGGAATTCAACTGGACGTCAGGTGG-3'), YSNE2 (5'-CCGCTCGAGGATCTTCACCGTAGTCTGC-3'), YWCB1 (5'-GGGGTACCTTTTTAAAGCAGAAACGCGC-3') and YWCB2 (5'-CGGGATCCGAATTGGGCGATGGCG-3'). Wild-type strain SG38 was transformed with the resulting plasmids. patA, yjcA and yloB were disrupted by using plasmid pMUTIN4. The following oligonucleotides were used: PATA1 (5'-GGGGAATTCGTTAAGGTGATGAATTATG-3'), PATA2 (5'-CGGGATCCAACGGCTTAAACGCCG-3'), YJCA1 (5'-GGGAAGCTTGGCAAAGGAGTGAACGAGTG-3'), YJCA2 (5'-CCGGATCCGCCTTATTCTTGCAGCC-3'), YLOB1 (5'-GGGGAATTCGACAAACAGATTTACTAGAGG-3') and YLOB2 (5'-CCGGATCCGACAGCAACGACAAGTACG-3'). Disruptions of ykpC and yhaL were created by double-crossover integration due to the small size of the genes. The 5' and 3' regions of ykpC and yhaL (including the start and stop codons respectively) were amplified independently by PCR from SG38 chromosomal DNA using oligonucleotides YKPC1 (5'-CCGAGCTCGTTGATGAGCCGGTTGCC-3'), YKPC2 (5'-CGGGGTACCGCCTCCAAGAATAATTTCCGC-3'), YKPC3 (5'-GGCATGAGCATTTCTTTGGC-3'), YKPC4 (5'-ACATGCATGCGTTGTCTTCATTTACGCGCG-3'), YHAL1 (5'-CCGAGCTCATGGATTATCCGGCGGC-3'), YHAL2 (5'-CGGGGTACCCTCTTTGGCTGCCGCC-3'), YHAL3 (5'-ACATGCATGCGAGCGAGAACGCCGC-3') and YHAL4 (5'-CCCAAGCTTCATCAAAGTAACAGACGCCG-3'). The PCR products corresponding to the 5' region of each gene were cloned into pBEST501 (upstream of the neo cassette). Then PCR products corresponding to the 3' region of each gene were inserted into the respective plasmid (downstream of the neo cassette). Wild-type strain SG38 was transformed with the resulting plasmids. To measure σ^E and σ^K activities, strains that had a sporulation defect were transformed with chromosomal DNA of strains YJFSd (pdaAlacZ) and NIS8141RT (yabGlacZ).

Table 1. Strains and plasmids

Mutants BFS1631 (yhaX : : pMUTIN), BFS414 (yhbH : : pMUTIN), BFS2843 (yjbE : : pMUTIN), BFS2860 (yjbX : : pMUTIN), BFS3241 (ylbO : : pMUTIN), BFS2827 (yloB : : pMUTIN), BFS2661 (yncD : : pMUTIN), BFS2668 (yndA : : pMUTIN) and BFS2412 (ysnE : : pMUTIN), which were generated by the B. subtilis genome function project, were transformed with chromosomal DNA of strains 901 and 646.

Media and growth conditions.
Nutrient agar (Oxoid) was used as a solid medium for growing B. subtilis. X-Gal (150 µg ml^-1), chloramphenicol (5 µg ml^-1), kanamycin (5 µg ml^-1), erythromycin (1 µg ml^-1) and lincomycin (25 µg ml^-1) were added as required. E. coli strains were grown in 2x TY medium (Sambrook et al., 1989) or on nutrient agar plates (Oxoid) supplemented with ampicillin (100 µg ml^-1). B. subtilis strains were induced to sporulate by the resuspension method as described previously (Partridge & Errington, 1993; Sterlini & Mandelstam, 1969).

β-Galactosidase assay.
β-Galactosidase was assayed using a method described by Errington & Mandelstam (1986); one unit of enzyme is the amount that releases 1 nmol 4-methylumbelliferone min^-1.

Determination of sporulation frequency and microscopy.
The sporulation frequency was determined by counting phase-bright spores and total cells by phase-contrast microscopy in samples taken 7 h after induction of sporulation. Spore morphology was analysed by phase-contrast light microscopy and images were acquired and analysed with a Princeton Instruments Micromax 1300Y/HS CDD camera and METAMORPH version 3.6 software.

mRNA preparation.
Cells were harvested from 50 ml cultures 2 h after the initiation of sporulation, pelleted and frozen in liquid nitrogen. In parallel, a sample was taken and assayed for alkaline phosphatase activity as a means of checking σ^E activity (Errington & Mandelstam, 1983; Partridge & Errington, 1993). The cell pellets were thawed at 37 °C in 1 ml TE containing 25 mg lysozyme ml^-1 for 5 min; 1 ml 150 mM NaCl, 0·1 % SDS, 10 mM EDTA, 10 mg Pronase ml^-1 was then added, mixed gently and the mixture incubated for a further 5 min. Nucleic acids were extracted with acid phenol at 60 °C and precipitated with 2·5 vols absolute ethanol for 2 h at -20 °C. The nucleic acid pellet was washed with 70 % ethanol, inverted and dried for 15 min on the bench, before being dissolved in 180 µl RNase-free H₂O (Ambion). A 75 µl sample of material from this crude preparation was incubated with 3 µl DNase (RNase free; Promega) at 37 °C for 1530 min to remove excess genomic DNA. The RNA was then purified further using the RNeasy mini kit (Qiagen) according to the manufacturer's instructions, the RNase-free DNase set was used at the appropriate step in the protocol. RNA was eluted in two aliquots of 30 µl RNase-free dH₂O. The sample was checked for contaminating DNA by agarose gel electrophoresis.

Transcriptional profiling.
A commercially available microarray was used, consisting of 3925 B. subtilis ORFs spotted on a glass slide (Eurogentec), onto which two fluorescently labelled RNA populations were simultaneously hybridized. RNA (50 µg) was used as a template for first-strand cDNA synthesis using the CyScribe First-Strand cDNA Labelling Kit (Amersham Pharmacia). The volume of each RNA sample was reduced to 5 µl in a Heterovac vacuum desiccator. Then 0·5 pmol B. subtilis-specific RT primers (Eurogentec) and 1 µl RNasin RNase inhibitor (Promega) were added and the reaction was made up to a total volume of 11 µl using RNase free dH₂O. This reaction mix was heated to 70 °C for 5 min, then allowed to cool gradually to 42 °C (over about 15 min) to anneal the primers to the RNA. After a further 5 min at 42 °C, 0·5 µl RNasin, 2 µl 0·1 M DTT, 1 µl dCTP nucleotide mix, 1 µl dCTP CYDye-labelled nucleotide and 1 µl Cyscript reverse transcriptase were added. Strain 2809 and strains 2810 and 2811 were labelled with Cy5 dCTP and Cy3 dCTP, respectively. The reaction was incubated at 42 °C for 2 h in the dark. mRNA was degraded by the addition of 2 µl 2·5 M NaOH to the sample and incubating at 37 °C for 15 min; the reaction was neutralized with 10 µl 2 M HEPES. Labelled cDNA was purified from degraded mRNA and unincorporated nucleotides using DyEx spin columns (Qiagen) according to the manufacturer's instructions.

The slide was prehybridized with 1 µl yeast tRNA (10 mg ml^-1; Sigma) and 1 µl salmon sperm DNA (10 mg ml^-1; Clonetech) in 20 µl 5x SSC, 0·2 % SDS (both RNase free; Ambion) at 42 °C for 2 h in a sealed cassette. The coverslip was removed by dropping the slide into 0·1x SSC and the slide was dried with compressed air. The two cDNA samples were reduced down to <5 µl volume in a vacuum desiccator, mixed together and the total volume was made to 20 µl with 5x SSC, 0·1 % SDS. This was pipetted onto the slide, a new coverslip was put on and the slide was incubated at 42 °C for approximately 20 h in a sealed cassette. The slide was washed twice in 0·1x SSC, 0·1 % SDS, followed by twice in 0·1x SSC, then dried with compressed air and scanned using a GenePix 4000B microarray scanner (Axon Instruments). The GenePix Pro software (Axon Instruments) was used to scan the slide, to overlay the grid of ORF names and to measure the signal intensities; the scanning voltage for each signal was adjusted to eliminate areas where the signal was saturated and to obtain an intensity ratio of as close to 1 : 1 (total red signal : total green signal) over the whole slide as possible. Whilst aligning the grid, any spots showing very low levels of fluorescence were discarded from further analysis. Finally, a list of genes and the corresponding ratios of the two fluorescent signals was generated. To calculate the ratio, for each fluor the background intensity immediately around the spot was subtracted from the intensity of the spot itself, and then the ratio of the two signals was calculated. The whole procedure was carried out twice with two independent RNA preparations and each ORF was spotted twice onto the slide; therefore, the final ratio for a given gene corresponds to the mean intensity of four spots.

Transcriptional profiling of a sigE null mutant
To identify genes under the control of σ^E, the transcriptional profile of a sigE null mutant (strain 2810) was compared with that of a strain wild-type for σ^E (strain 2809). Both strains also contained a mutation in the sigG gene (encoding σ^G) to eliminate differences occurring later in sporulation between the two strains. Cells were induced to sporulate by the resuspension method and harvested 2 h later, at a time when σ^E is fully active (Partridge & Errington, 1993). Total RNA was prepared, used for cDNA synthesis by RT-PCR incorporating Cy3- or Cy5-dCTP and hybridized to DNA microarrays containing around 96 % of the annotated coding genes of the B. subtilis genome.

All genes with at least threefold higher expression in the wild-type than in the σ^E mutant were assumed to be dependent on σ^E and are listed in Table 2. We used a rather high threshold value of 3 to minimize false positives. Genes transcribed by σ^F, another sporulation-specific transcription factor, did not give any detectable signal or had values well below 3 (e.g. gpr, 1·4; spoIIQ, 0·3; dacF, 0·8; spoIIR, 0·1; sigG, 0·1; spoIVB, 0·2). A total of 155 genes fell into this category, of which 43 had been reported previously as being dependent on σ^E. The remaining 112 genes are organized in 88 operons and among these are 23 genes (Table 2; in bold) that were tentatively assigned as being σ^E-dependent on the basis of a previous microarray analysis of σ^F-dependent genes (Fawcett et al., 2000). Three genes, cotZ, spoVFB and spsK, were previously reported to be σ^K-dependent (Daniel & Errington, 1993; Zhang et al., 1994); detecting these in our arrays might be due to the fact that σ^E and σ^K have similar recognition sequences (Haldenwang, 1995).

Table 2. Genes found to be dependent on σE or σE663 by microarray analysis

To verify the data obtained by microarray we analysed the expression pattern of nine so far unknown genes (yhaX, yhbH, yjbE, yjbX, ylbO, yloB, yncD, yndA and ysnE) during sporulation. We took advantage of the mutant collection generated by the B. subtilis genome function project, in which genes were inactivated and thereby fused to lacZ through integration of plasmid pMUTIN (Vagner et al., 1998). Expression of β-galactosidase (seen as development of blue colour on nutrient agar plates supplemented with X-Gal) was analysed in the wild-type, a sigE null mutant and a sigG mutant (encoding σ^G, the transcription factor which is activated after σ^E and whose activation is dependent on σ^E) background. As shown in Fig. 1, expression of β-galactosidase was abolished in all strains in the sigE null mutant but not in the sigG mutant background, supporting the idea that these genes are transcribed by σ^E-containing RNA polymerase. Also, the strong and weak expression of β-galactosidase in yhaX, yhbH, yjbX, ylbO and yjbE, yloB, yncD, yndA, respectively, corresponded well with their high and low transcript ratios found by microarray. While this paper was being reviewed Eichenberger et al. (2003) published a study in which genes under control of σ^E were identified (by microarray) and analysed, and their results are generally in good agreement with our data.

(97K): Fig. 1. β-Galactosidase expression in strains carrying a lacZ-fusion to an unknown σE-dependent gene. Strains in the wild-type (wt), a sigE null mutant (sigE) and a sigG mutant (sigG) strain background were patched onto nutrient agar plates containing X-Gal and grown at 37 °C for 3 days.

Identification of genes recognized by σ^E663
The function of region 4.2 of σ-factors is to interact with the -35 promoter element and mutations in this region can alter promoter specificity. A missense mutation (sigE663) in this region of σ^E, which changes the glutamine at position 218 to glutamic acid, has been shown to virtually abolish expression of spoIVC, spoIIID, spoVD and spoIID, whilst increasing the expression of spoIIIA (Errington et al., 1990; Illing, 1990). Importantly, the mutation allows transcription from the sigG promoter, so that the σ^E-dependent gene(s) required for this transcription should be among the subset of genes not affected by the sigE663 mutation. The transcriptional profile of the sigE663 mutant (strain 2811) was compared to that of the isogenic strain wild-type for σ^E (strain 2809) using a DNA microarray as described above. Again, both strains contained a sigG null mutation to eliminate downstream effects on gene expression. The resulting transcript ratios are listed in column 3 of Table 2. Note that in general the numbers tend to be lower than those in column 2, indicating that the block in transcription in this mutant is not as strong as for a null mutation. We found 16 additional genes which were not detected before and four of these genes are known σ^E-dependent genes. As expected, the transcript ratios found for spoIIID, spoIVC, spoIID and spoVD were high, confirming that these genes are transcribed by Eσ^E and not by Eσ^E663. On the other hand, the transcript ratios found for the spoIIIA operon were generally low, either below or just above 3, confirming that expression of this operon is allowed by σ^E663. In total, transcription of 81 % of the σ^E-dependent genes (138 out of 171 genes) was affected by the sigE663 mutation. Only 30 genes were transcribed normally in the mutant strain (i.e. genes with a value less than threefold different than that of the wild-type strain).

Screening for genes required for efficient sporulation
Most of the σ^E-dependent genes identified by the microarray analysis were genes with unknown function. Some of these are likely to be required for formation of the spore coat, the cortex, for proper germination, and at least one σ^E-dependent gene is postulated to be involved in the regulation of transcription of sigG, encoding the late prespore-specific σ-factor σ^G. We therefore investigated the phenotypic effects of interrupting selected candidate genes. Candidates were prioritized on the basis of two criteria: the presence of predicted membrane-spanning domains in the protein, or recognition of the promoter by σ^E663. The products of several known sporulation genes dependent on σ^E (spoIIB, spoIID, spoIIM, spoIIP) are involved in prespore engulfment and are membrane proteins (Abanes-De Mello et al., 2002; Frandsen & Stragier, 1995; Perez et al., 2000; Smith et al., 1993). It seems likely that other σ^E-dependent genes involved in engulfment would have domains capable of inserting into the membrane. The class of genes dependent on σ^E but not affected by the σ^E663 mutation were screened because they were candidates for the effector responsible for control of sigG transcription (see above).

Eighteen genes of unknown function (cotZ, patA, ybfJ, ydcC, ydhF, yhaL, yhbH, yjaV, yjcA, ykpC, yloB, yndA, yodP, yqfD, yqhV, yrzE, ysnE and ywcB) were disrupted in a strain containing a lacZ fusion to sigG and in the wild-type background. All the deletion strains were screened for activity of the sigG : : lacZ fusion on nutrient agar plates supplemented with X-Gal, and the blueness of the colonies was compared with that of the parent strain. None of the genes tested had an effect on sigG expression that could be detected on plates (data not shown).

Screening the mutants based on colony appearance on nutrient agar plates revealed that 10 of the mutants might be affected in sporulation efficiency. These mutants were further analysed by inducing sporulation in liquid medium and scoring quantitatively for spore frequency by counting phase-bright spores by phase-contrast microscopy (Table 3). For five of the deletion strains (ydcC, yhaL, yhbH, yjaV and yqfD) a significant reduction in sporulation frequency was observed. Surprisingly, the ydcC mutant strain, which was severely affected when grown on nutrient agar plates (<0·5 % spores compared to the wild-type), produced around 23 % spores when sporulation was induced by the resuspension method. The most severe sporulation phenotype was exhibited by the yqfD mutant, which did not produce detectable numbers of spores (as examined by phase-contrast microscopy). Microscopic examination of the sporulating cultures revealed the existence of phase-dark spores that had probably failed to complete development to the phase-bright state. Disruption of yqfD had the strongest phenotype, producing no detectable phase-bright spores. Although the yhbH mutant strain produced 20 % phase-bright spores, phase-dark spores could also be detected readily. Fig. 2 shows images of sporulating cultures of wild-type, yqfD and yhbH mutant strain taken 7 h after the initiation of sporulation.

Table 3. Frequency of sporulation in wild-type and deletion strains

(66K): Fig. 2. Spores formed by wild-type, yqfD and yhbH mutant strains. Images taken by phase-contrast light microscope illustrating the appearance of the spores from sporulating cultures of the wild-type, yhbH and yqfD mutant strains. Arrowheads point to phase-bright (wt) or -dark (yhbH and yqfD) spores.

Next, we examined if any of the mutations had an effect on the activation of the late sporulation-specific transcription factors σ^G and σ^K. The activity of σ^G and σ^K was measured with pdaAlacZ and yabGlacZ reporter genes respectively (Fukushima et al., 2002; Takamatsu et al., 2000). As shown in Fig. 3(a), yhaL, yhbH, yjaV and yqfD mutants displayed normal kinetics of σ^G activation, whereas the ydcC mutant exhibited reduced σ^G activity. The ydcC gene, which encodes a putative membrane protein, may play a role in engulfment as it has been shown before that σ^G activation is inhibited in mutants which block engulfment (Abanes-De Mello et al., 2002; Partridge & Errington, 1993; Perez et al., 2000; Smith et al., 1993). No differences in activation of σ^K were observed in the yjaV, yhaL and yqfD mutants (Fig. 3b). The slightly reduced σ^K activity found in the ydcC mutant is probably a result of the reduced σ^G activity in this strain. Surprisingly, σ^K activity was around threefold higher in the yhbH mutant than in the wild-type, although the timing of activation of σ^K appeared to be normal. We repeated the experiment using a different σ^K-dependent reporter gene (dpaAlacZ; Daniel & Errington, 1993) and again observed increased levels of σ^K activity (data not shown). In summary, three of the further analysed mutants exhibited a sporulation defect, but none of these were affected in the activation of the later-acting σ-factors σ^G and σ^K. Two of the mutants, ydcC and yhbH, affected σ^G and σ^K activity respectively, but caused only modest defects in sporulation efficiency.

(17K): Fig. 3. σG and σK activity in the mutant strains. Strains SG38 (wild-type; squares), 1341 (ydcC; triangles), 1342 (yhaL; crossed squares), 1344 (yhbH; circles), 1345 (yjaV; asterisks) and 1351 (yqfD; diamonds) carrying either the pdaAlacZ (a) or the yabGlacZ (b) reporter gene were induced to sporulate and assayed at intervals for β-galactosidase activity.

The microarray data presented here and in other papers (Fawcett et al., 2000; Britton et al., 2002) demonstrate that this technique is a powerful tool to study differential gene expression. Overall, our data appear to provide a good description of the σ^E regulon as nearly all of the known σ^E-dependent genes and a further 23 genes assigned as being σ^E-dependent on the basis of an analysis done by Fawcett et al. (2000) were detected in the microarrays (Table 2). However, a few genes, cotH spoIIM, spoIIP, spoVE, spoVID, spoVUB and yfhS, which are also known to be under control of σ^E, were missing from the list. Low level and/or the unusual timing of their expression might explain why these genes were not picked up by the array. Also, to use a threefold difference in expression as a cut-off point is arbitrary and is based on examining the transcript ratio of the identified known σ^E-dependent genes and the related work of Fawcett et al. (2000) and Price et al. (2001). We used this rather high threshold to reduce the risk of picking up false positives. A few unknown genes might still be missing as the microarray chip contained only 96 % of the annotated coding genes of the B. subtilis genome. Interestingly, most of the σ^E-dependent genes were not expressed in the sigE663 mutant. The sigE663 effect is certainly compounded by, and possibly entirely due to, the lack of spoIIID transcription by this mutant form of σ^E (Errington et al., 1990). SpoIIID is well known as a major auxiliary transcription factor in the σ^E regulon (Kunkel et al., 1989; Stevens & Errington, 1990).

Our data suggest that at least 178 genes (including the ones which were not detected by our microarrays) belong to the σ^E regulon and 130 of these are organized in operons. Here, we identified 101 genes that had not previously been described as being under the control of σ^E. Among the newly identified genes are several transporters (citH, glnM, glnP and yknV), which could provide additional nutrients to the prespore. SodF (superoxide dismutase) and YocM (similar to small heat-shock protein) might provide protective properties to the sporulating cell. Only two of the identified genes, ykvU (similar to spore cortex membrane protein) and yqfD (similar to stage IV sporulation protein of B. licheniformis), show similarity to known proteins involved in sporulation. The majority of the newly identified genes are genes with unknown function. Interestingly, around 50 % of these genes (65 out of 124) encode putative membrane proteins. This might reflect the possibility that σ^E controls the expression of many genes involved in engulfment, formation of the spore coat and cortex, and possibly nutrient-scavenging functions that help to adapt the cell to starvation.

Disruption of some of the newly identified genes led to the identification of five additional proteins required for efficient sporulation (YdcC, YhaL, YhbH, YjaV and YqfD). Little is known about the function of these genes: ydcC, yhbH and yhaL have no function predicted from their sequences. Deletion of yqfD had the most severe sporulation phenotype as the mutant produced only phase-dark spores that failed to become phase-bright, indicating a block at a late stage in sporulation. yqfD, which encodes a protein with one potential membrane domain, probably plays a role after completion of engulfment in coat or cortex formation. It has been shown before that mutants that block engulfment also block activation of σ^G, but the yqfD mutant was able to activate σ^G normally (Fig. 3). In contrast, the reduced σ^G activity in the ydcC mutant might indicate that the YdcC protein, which also encodes a protein with one potential membrane domain, is involved in engulfment (Fig. 3). The medium-dependent effect on sporulation in this mutant (hardly any spores were produced on nutrient agar plates), suggests that YdcC protein becomes essential for spore formation only under certain environmental conditions. A similar phenomenon has been reported previously for mutation of the spoIIAB and spoIIQ genes (Foulger et al., 1995; Sun et al., 2000). The increased σ^K activity found in the yhbH mutant suggests that the yhbH gene, which encodes a soluble protein, might be involved directly or indirectly in repression of certain σ^K-dependent genes. Alternatively the protein might affect the activity of σ^K by destabilizing the protein or changing its activity.

The challenge for the future is now to characterize more of the newly identified genes from our microarray analysis to get a better understanding of how the mother cell contributes to the formation of the spore and how differential gene expression and morphogenesis are co-ordinated in the two compartments.

We thank Professor Watabe, Dr Sekiguchi and Dr Asai for strains. We thank R. Daniel for his help with the microarrays. This work was supported by grants from the Biotechnology and Biological Sciences Research Council (BBSRC) and Medical Research Council (MRC). L. E. was the recipient of a BBSRC postgraduate studentship.