Molecular Strategy for Survival at a Critical High Temperature in Eschierichia coli

The molecular mechanism supporting survival at a critical high temperature (CHT) in Escherichia coli was investigated. Genome-wide screening with a single-gene knockout library provided a list of genes indispensable for growth at 47°C, called thermotolerant genes. Genes for which expression was affected by exposure to CHT were identified by DNA chip analysis. Unexpectedly, the former contents did not overlap with the latter except for dnaJ and dnaK, indicating that a specific set of non-heat shock genes is required for the organism to survive under such a severe condition. More than half of the mutants of the thermotolerant genes were found to be sensitive to H2O2 at 30°C, suggesting that the mechanism of thermotolerance partially overlaps with that of oxidative stress resistance. Their encoded enzymes or proteins are related to outer membrane organization, DNA double-strand break repair, tRNA modification, protein quality control, translation control or cell division. DNA chip analyses of essential genes suggest that many of the genes encoding ribosomal proteins are down-regulated at CHT. Bioinformatics analysis and comparison with the genomic information of other microbes suggest that E. coli possesses several systems for survival at CHT. This analysis allows us to speculate that a lipopolysaccharide biosynthesis system for outer membrane organization and a sulfur-relay system for tRNA modification have been acquired by horizontal gene transfer.


Introduction
Responses of Escherichia coli to high temperatures have been extensively investigated, though previous studies have mainly focused on the response to a temperature up-shift around 42uC, a response known as a heat shock response (HSR) to induce the expression of a set of proteins, heat-shock proteins (HSPs) [1]. The fact that many HSPs are conserved among species indicates that the actions of HSRs are the fundamentally and physiologically important mechanisms in living organisms [2,3]. HSPs play crucial roles not only in the rescue or removal of proteins damaged by environmental stresses, including heat stress and salt stress, but also in the intrinsic folding of proteins under normal growth conditions [4].
It has been shown that 384 genes are up-regulated by short-time exposure to a temperature of 43uC as a heat shock in E. coli [5], and these genes may be directly or indirectly induced by the treatment. The directly induced genes encode HSPs, including the main cellular chaperone machineries of GroEL and DnaK, ATPdependent proteases of Lon, HslUV, Clp and FtsH (HflB), periplasmic protease DegP, and other proteins involved in protein folding, refolding, quality control and degradation [6]. HSPs are under complex regulations and are divided into several regulatory groups by their major stimulons [7]. The control of their expression, however, is highly variable among organisms and even among various bacteria [8].
One of the control elements found in Gram-negative bacteria is a heat shock sigma factor that regulates transcription of the major HSPs. HSR in E. coli is generally mediated by alternative sigma factors, sigma 32 and sigma 24 [4,7,8]. Transcription of the rpoH gene for sigma 32 is induced at elevated temperature via the action of sigma 24 [7]. Sigma 24, which is inactive under non-stress conditions by interaction with anti-sigma factor, is activated by misfolding of outer membrane or periplasmic proteins and by stresses including heat shock [9]. Both sigma factors are further regulated at the translation level and or at the posttranslational level. The factor sigma 24 is in part regulated by a cognate small RNA, and sigma 32 synthesis is regulated by structural change of its own mRNA molecules serving as a cellular thermometer and its activity modulated by phosphorylation [10,11]. Other microorganisms, on the other hand, appear to possess diverged regulatory mechanisms [12].
There is no information on the molecular mechanisms of response to and survival at a critical high temperature (CHT) in organisms, probably due to the limited experimental procedures. Developments of a single-gene knockout library and DNA chip analysis have encouraged us to perform a genome-wide investigation of responses in organisms under extreme conditions. Since several mesophilic bacteria including E. coli can grow and survive at high temperatures compared to other mesophilic bacteria, they are assumed to have acquired the potential for thermotolerance during their evolution. In this study, we utilized new procedures for the first time to obtain information on the molecular mechanisms related to thermotolerance in E. coli at CHT. Screening of thermosensitive mutants at CHT and informatics analysis of the corresponding genes revealed pathways or factors indispensable for survival at CHT. For essential genes, their possible involvement in the response to CHT was examined by DNA chip analysis. Based on the results, we propose novel molecular mechanisms for survival at CHT in E. coli.

Thermosensitive mutants and thermotolerant genes
In order to identify genes required for survival at CHT in E. coli, we screened for thermosensitive mutants from a single-gene knockout library [13], which had been constructed according to the one-step gene disruption method with an aph cassette [14] and for which each construct had been confirmed extensively [15]. In the disrupted gene of each mutant strain, the region between the 1 st codon and the last 6 codons was displaced with the aph cassette, so that most of the coding region of the gene was deleted. Our experiments indicated that the parental strain used for construction of the disrupted library is able to grow at temperatures up to 47uC, this temperature thus being its CHT.
After three successive screening steps of the library, including 3,908 disrupted-mutant strains, 51 strains were found to be sensitive to CHT. Their growth curves at 37uC, 45uC and 46uC were then compared to those of the parental strain ( Figure S1). The growth profiles suggest that most mutants selected are significantly sensitive to 46uC and some even to 45uC. Such a disrupted gene responsible for the thermosensitive phenotype was designated as a thermotolerant gene (Table 1 and Table 2).
The gene organization generated by construction of the disrupted mutants might give rise to a polar effect of the inserted aph gene on transcription of downstream genes that are intrinsically transcribed by read-through from the promoter or the region upstream of the disrupted gene. Such an organization was found in 42 of the 51 mutants. Sensitivity was not due to a polar effect in 29 of those 42 mutants because disruption of genes just downstream from the disrupted gene by the same method caused no thermosensitive phenotype. The remaining 13 mutants have either an essential gene or a thermotolerant gene as an immediate downstream gene ( Figure S2). Their possible polar effects were thus tested by RT-PCR with total RNA prepared from cells exposed to a temperature of 37uC or 47uC ( Figure S3). The results suggest that the transcription level of the immediate downstream gene in the mutant was almost the same as that in the parent in all cases except for the cases of mutants of aceF, tolQ, dnaK and rpsF. Most of these downstream genes would thus have their own promoters or the transcription level by read-through would be nearly the same as that of the aph promoter. However, the transcription levels of lpd, tolR and dnaJ located downstream of aceF, tolQ and dnaK, respectively, were increased and the level of rpsR located downstream of rpsF was decreased compared to those of the parental strain at both temperatures. Although the expressional alteration of the 4 genes was nearly the same at both temperatures, growth of the corresponding mutant strains at 37uC was not significantly changed from that of the parental strain. Taken together, the results suggest that the thermotolerant phenotype in the 51 mutants is due to disruption of the targeted gene and not due to a polar effect on its downstream genes. Out of the 51 thermotolerant genes, 8 genes, cydB, degP, dnaJ, dnaK, dnaQ, nlpI, rfaD and rfaC, had been reported as genes supporting growth at a high temperature [16][17][18][19][20][21][22], and thus we newly identified 43 thermotolerant genes in this organism.
Effects of supplements and oxidative stress on growth of thermosensitive mutant strains Since LB was utilized as a medium for the screening of thermosensitive mutants, limitation of carbon source might cause sensitiveness to CHT. We thus examined the effect of glucose as a supplement for growth of the thermosensitive mutant strains ( Table 1, Table 2 and Figure S4). We also tested the effect of MgCl 2 because Mg 2+ somehow protects against cell damage under stress conditions [23,24]. The growth of 20 and 37 mutants was improved at CHT by the addition of 0.5% glucose and 20 mM MgCl 2 , respectively. The growth of sixteen mutant strains was improved by supplementation of not only glucose but also MgCl 2 .
Next, the effect of exogenous oxidative stress on the thermosensitive mutant strains was tested since a higher temperature causes more oxidative stress (Noor et al, 2009; unpublished data). We exposed thermosensitive mutant strains to 0.5 mM H 2 O 2 in LB liquid medium at 30uC. Twenty-nine mutants were found to be sensitive to H 2 O 2 ( Table 1, Table 2 and Figure S4), corresponding to approximately 60% of the thermosensitive mutants. Moreover, out of the 10 thermosensitive mutants for which glucose and MgCl 2 supplementation had no effect, 9 mutants showed sensitivity to H 2 O 2 . These results suggest that the mechanism of thermotolerance at CHT partially overlaps with that of oxidative stress resistance.

Bioinformatics analysis and classification of thermotolerant genes
To understand the molecular mechanism of E. coli survival at CHT, bioinformatics analysis with various public databases including the KEGG PATHWAY database was performed. Out of the 51 thermotolerant genes, 29 genes were successfully mapped on E. coli pathways in the KEGG PATHWAY database. Interestingly, many genes were found to be involved in the same metabolic pathway, suggesting that the organism possesses indispensable pathways at CHT. The remaining 19 genes except for 3 unknown genes were extensively analyzed by using the DDBJ or GenBank database. On the basis of results of these analyses and the effects of the supplements, the 51 thermotolerant genes were classified into 7 groups (Table 1 and Table 2).
Group A consists of genes concerned with energy metabolism for production of ATP. The gene products of aceE, aceF, lpd and ackA are mapped in the pyruvate metabolism pathway from pyruvate to acetyl CoA [25][26][27][28] and that of rpe is located in the pentose phosphate pathway. cydB and yhcB encode subunits of cytochrome d terminal oxidase, which generates the membrane potential responsible for ATP synthesis [16,29]. lipA, which encodes LipA to produce lipoate required for pyruvate dehydrogenase reaction, also contributes to pyruvate metabolism [30]. Based on the results showing that disrupted mutations of these genes caused a thermosensitive phenotype, we assumed that the cells require more ATP at a higher temperature. This assumption was supported by the finding that the phenotype of most mutants in this group was partially suppressed by the addition of glucose (Table 1 and Figure S4).
Group B consists of genes related to biosynthesis of the cell wall or organization of the outer membrane. The products of gmhB, lpcA(gmhA), rfaC (waaC), rfaD (waaD/htrM), rfaE (gmhC), rfaF (waaF) and rfaG (waaG) were mapped into the lipopolysaccharide (LPS) biosynthesis pathway [17,18,31,32]. The products of these genes are involved in synthesis of the heptose unit of ADP-L-glycero-Dmanno-heptose from sedoheptulose-7phosphate or encode early heptosyl transferases for KDO-lipid A (rfaC and rfaF) and to further extend the inner core of LPS with glcosyltransferase (rfaG). ydcL, yfgL (bamB), ynbE, nlpI and ycdO encode peptidoglycanassociated outer membrane lipoproteins, and the products of pal, tolQ and tolR are components for a complex structure forming a biopolymer transporter [33,34]. yciM encodes a protein possibly required for integrity of the outer membrane [35]. The thermosensitive phenotype caused by disrupted mutants of all of these genes was significantly suppressed by the addition of Mg 2+ (Table 1 and Figure S4). Since Mg 2+ is known to stabilize the outer membrane structure by binding extracellularly [36], it is assumed that YdcL, YfgL, YnbE, NlpI, YcdO, Pal, TolQ, TolR and YciM act as components or scaffold proteins of the membrane to maintain outer membrane integrity, especially at a high temperature. Similarly, our data suggest that Mg 2+ is able to stabilize the outer membrane structure when the LPS biosynthesis pathway becomes defective.
Group C consists of dnaQ, holC, priA, ruvA and ruvC for DNA double-strand break repair (DSBR) [37]. DnaQ and HolC are epsilon and chi subunits, respectively, of DNA polymerase III [38,39], which is required for homologous recombination in DSBR [19]. RuvA and RuvC act as DNA helicase and endonuclease, respectively [19,40], before the replication restart in the DSBR process, and PriA functions as DNA helicase after the replication restart [41]. The requirement of DSBR for survival at CHT suggests that DNA molecules are subjected more to doublestrand breaks at a higher temperature. Interestingly, mutants of all members in this group exhibited sensitivity to oxidative stress at 30uC. Therefore, it is thought that there is a strong connection between oxidative stress and DNA double-strand breaks.
Group D includes genes for tRNA modification. Products of iscS, yheL (tusB), yheM (tusC), yheN (tusD) and yhhP (tusA) have been demonstrated to compose the sulfur-relay system [42][43][44]. IscS is a widely distributed cysteine desulfurase that catalyzes desulfuration of L-cysteine by transfer of the sulfur to its active-site cysteine to According to the data in Figure S4, ratios of growth in the presence of glucose to that in the absence of glucose at 46uC wer estimated. ''++'' and ''+'' represent more than 2.0 and 1.5-2.0, respectively. b According to the data in Figure S4, ratios of growth in the presence of MgCl 2 to that in the absence of MgCl 2 at 46uC were estimated. ''++'' and ''+'' represent more than 2.0 and 1.5-2.0, respectively. c According to the data in Figure S4, ratios of growth in the presence of H 2 O 2 to that in the absence of H 2 O 2 at 30uC were estimated. ''S'' represents less than 0.5. doi:10.1371/journal.pone.0020063.t001 form a persulfide group (-SSH), being responsible together with YheL, YheM, YheN and YhhP for biosynthesis of the 2-thio modification of 5-methylaminomethyl-2-thiouridine (mnm 5 s 2 U) [43] and five different thio modifications in bacterial tRNAs [45]. IscS also works as a general sulfur donor in various metabolic pathways [46] including biosynthesis of iron-sulfur (Fe-S) cluster [47], thiamine [48], nicotinic acid and branched-chain amino acids [49]. Additionally, miaA, trmU and truA in this group are involved in tRNA modification. The mutations of genes related to sulfer modification cause the phenotype of sensitivity to antioxidation stress [44]. Consistently, our study provided evidence that mutants of this group exhibited hypersensitivity to oxidative stress. YheL, YheM, YheN and YhhP, which mainly function in tRNA modification [45], are conserved in thermotolerant bacteria in mesophiles (see Table S3), whereas iscS, a general sulfur donor, is widely conserved in mesophiles. These findings suggest that tRNA modifications presented here are indispensable for growth at CHT.
Group E genes encode chaperones and a protease and thus contribute to the cellular process of regulating heat shock response: dnaK and dnaJ encode a chaperone and co-chaperone, respectively, for maturation of protein folding or refolding of unfolded proteins [20,50], and degP encodes a chaperone/serine protease located in the periplasm [21]. The indispensability of these genes at CHT suggests that DnaK/DnaJ play a crucial role in dealing with unfolded proteins caused by CHT and that DegP plays an important role in the removal of damaged proteins that have accumulated at such a temperature. rseA in this group encodes an anti-sigma factor to keep sigma 24 inactive under non-stress conditions. The thermosensitivity caused by rseA disrupted mutation suggests that fine tuning of the intracellular level of active sigma 24 that regulates expression of chaperone or protease genes is somehow crucial for adaptation to the CHT condition. Alternatively, the defective mutant of rseA increased sigma 24 activity, which in turn decreased the production of outer membrane proteins via MicA or RybB as a sigma 24 regulon gene [51], resulting in membrane unstability and thermosensitiveness at CHT.
Genes in group F belong to the translation control apparatus. S6 encoded by rpsF interacts with the central domain of 16S rRNA According to the data in Figure S4, ratios of growth in the presence of glucose to that in the absence of glucose at 46uC wer estimated. ''++'' and ''+'' represent more than 2.0 and 1.5-2.0, respectively. b According to the data in Figure S4, ratios of growth in the presence of MgCl 2 to that in the absence of MgCl 2 at 46uC were estimated. ''++'' and ''+'' represent more than 2.0 and 1.5-2.0, respectively. c According to the data in Figure S4, ratios of growth in the presence of H 2 O 2 to that in the absence of H 2 O 2 at 30uC were estimated. ''S'' represents less than 0.5. doi:10.1371/journal.pone.0020063.t002 and has been demonstrated to play a regulatory rather than a structural role in the ribosome [52]. L36 encoded by rpmJ is a component of the 50S subunit of the ribosome, and its disruption decreases the expression of secY [53], which encodes a proteinconducting channel in the cytoplasmic membrane. DksA encoded by dksA functions as a negative regulator for rRNA genes [54]. Overexpression of DksA has been shown to be a suppressor for a dnaK deletion mutation [55] and ensures replication completion by removing transcription roadblocks [56]. SmpB encoded by smpB is a component of the trans-translation process and performs rescue of stalled ribosomes with its binding partner, transfer-messenger RNA [57]. These lines of evidence suggest that several constituents in translation pathways are crucial for survival at CHT. Finally, genes in group G are related to cell division. A xerCencoded protein is a site-specific recombinase [58] and is essential for conversion of chromosome dimers to monomers during cell division. envC encodes a component of the cell division machinery that is a direct regulator of the cell wall hydrolase responsible for cell separation that is required for cell division [59]. DedD encoded by dedD is a membrane-anchored periplasmic protein involved in septation [60] and has been shown to participate in cytokinesis [61].
The functions of the remaining genes, ybgH, yciM and yhhH, are unknown. Notably, the thermosensitiveness of their mutations was partially suppressed by the addition of Mg 2+ . It is thus likely that their gene products are related to cellular activities similar to those in group B, C or G.

Possible acquisition of some thermotolerant genes by horizontal gene transfer
Two groups for outer membrane integrity and tRNA modification are almost completely conserved in limited bacterial species with optimal growth at a relatively high temperature (Table S1). Of these group members, genes for the LPS biosynthesis pathway, some lipoproteins and the sulfur-relay system are distributed in very limited bacterial species including Enterobacteriaceae (Tables S2 and Table S3). The sulfur-relay system classified in tRNA modification has been demonstrated to modify a few nucleotides of tRNA molecules, contributing to stabilization of their structure, and to be required for survival at an extremely high temperature in Thermous thermophilus [62] and it is also conserved in Thermoanaerobacter tengcongensis (Table S3). The mature LPS biosynthesis pathway for assembly of the outer membrane consists of many enzyme reactions, which was found to be dispensable at a lower temperature. Interestingly, this pathway is mostly conserved in Thermodesulfovibrio yellowstonii and Thermanaerovibrio acidaminovorans (Table S2). Enzymes in the LPS biosynthesis and sulfur-relay system in E. coli share about 40% sequence identity and about 50% sequence similarity to the corresponding enzymes in thermophilic bacteria. E. coli and its closely related bacteria would thus have acquired these genes of the two groups presumably by horizontal gene transfer during their evolution. Since the other five groups are widely conserved not only in thermotolerant mesophilic bacteria but also in other mesophilic bacteria, they would be intrinsically present in E. coli. This is consistent with the conserved nature of essentiality of the lipid A part of LPS and essentiality of synthesis of lipid IV A but dispensability of enzymes involved in extension of Kdo 2 -lipid A by various glycosyltransferases. This draws support from Re (rfaC) mutants with only tetraacylated lipid A exhibiting a very narrow growth range with ability to grow only under slow growth conditions on minimal medium around 23uC [18], suggesting overall importance of outer membrane integrity at CHT.

Expressional change caused by heat shock at CHT
None of the thermotolerant genes identified in this study were found to encode HSPs previously identified in E. coli except for dnaJ, dnaK, degP and dnaQ. To examine whether the thermotolerant genes were up-regulated at CHT or not, we examined transient change in expression of the genomic genes at CHT by DNA chip analysis. The results showed that 42 genes and 111 genes were significantly up-regulated and down-regulated, respectively (Table  S4). The up-regulated genes were classified mainly into genes involved in the cellular process, and the down-regulated genes were classified into genes involved in energy metabolism, transport/binding protein and translation. However, none of the thermotolerant genes including degP and dnaQ as a heat-shock gene were identified as up-regulated genes except for dnaJ and dnaK. Taken together with data shown above, it is possible that the chaperone systems except for DnaJ/DnaK and GroEL/GroES are not neccesarily involved in thermotolerant mechanisms acquired at CHT. Therefore, it is likely that most products of thermotolerant genes are not HSPs and that the organism possesses a specific set of genes required for survival at CHT.
It is possible that some of the essential genes are crucial for growth at CHT. Such genes, however, could not be examined in this analysis because no disrupted mutants for these genes are available other than the conditional mutants. We thus listed essential genes with significant fluctuation in expression at CHT (Table 3). groEL (groL) encoding HSP was up-regulated, indicating the possibility that the gene product contributes to survival at CHT. Consistently, it was reported that GroEL appears as a mediator of evolution of extremely heat-resistant E. coli cells [63]. On the other hand, 90% of the down-regulated genes were mapped into the translation pathway ( Figure S5), encoding for components of ribosomal proteins. It is thus possible that downregulation of ribosomal genes is one of the strategies for survival at CHT in E. coli. Noteworthily, Alix et al. reported that ribosome biogenesis in E. coli is high temperature-sensitive and DnaKdependant and predicted that high temperature causes a severe limitation in DnaK/DnaJ to hamper ribosome assembly because Table 3. Essential genes significantly up-regulated and down-regulated at CHT. heat-induced misfolded proteins would titrate out all the free DnaK/DnaJ [64,65].

Further consideration on mechanisms for survival at CHT
Two groups of DNA double-strand repair and chaperone/ proteinase genes may contribute to endurance against oxidative stress in addition to CHT. Evidence that a higher temperature results in accumulation of more oxidative stress [24] and the finding that mutants of all members in both groups exhibited sensitivity to oxidative stress allow us to speculate that oxidative stress is a main cause of DNA double-strand breaks and of damage to proteins at CHT. Interestingly, oxidative stress is involved in heat-induced cell death in Saccharomyces cerevisiae [66], which is supported by the findings that overexpression of catalase and superoxide dismutase genes could increase the degree of thermotolerance and that the thermotolerance is increased under anaerobic conditions. We thus assume that CHT somehow causes intracellular oxidative stress to elicit harmful effects on cells as a secondary stress.
Significant suppression of the thermosensitive phenotype by a defect in the group of energy metabolism (Group A) by the addition of glucose suggests the limitation of energy level at CHT in the organism. The limitation seems to be resolved by alternative pathways that may generate ATP by glucose assimilation. The requirement of ATP at CHT may be consistent with expression of ribosomal genes. Many genes for ribosomal proteins were found to be down-regulated by exposure to CHT, and the disrupted mutant of dksA that encodes a negative regulator for rRNA genes became thermosensitive to CHT. These findings and evidence that translation as a ribosomal activity utilizes much energy, up to about 90% of energy consumed in cells [67], suggest that cells manage to reduce energy consumption under a severe condition at CHT. Such saved energy would be utilized for other crucial activities such as repair or degradation of damaged DNA or protein molecules. A smooth translational process at CHT might also save energy, for which S6 and L36 of ribosomal proteins in addition to SmpB may have important functions.
Several strategies for E. coli to survive at CHT were discovered. Most of them may also be responsible for other stresses and are conserved even in mesophilic bacteria. Early glycosyltransferases for LPS core biosynthesis for proper outer membrane assembly and permeability barrier function and the sulfur-relay system for tRNA modification might have been acquired for the organism to perform a main task to survive at CHT. Considering the genetic conversion of non-thermotolerant to thermotolerant bacteria, the two strategies might be applicable.

Materials
Oligonucleotide primers for polymerase chain reaction (PCR) were purchased from FASMAC Co, Ltd (Atsugi, Japan). Other chemicals were all of analytical grade.

Screening of thermosensitive mutants
The Keio collection consisting of 3,908 mutant strains was used for screening. In the 1 st screening, mutant strains were grown on LB plates at 30uC overnight. A colony of each strain was patched on LB plates and incubated at 47uC for 48 h to find sensitive strains. The sensitive strains were subjected to the 2 nd screening of spotting tests on plates. Cells were cultured in LB medium for 18 h and then diluted with LB medium to adjust turbidity to OD 600 of 0.5, 0.05 and 0.005. The diluted samples (10 ml) were spotted on LB plates and incubated at 47uC for 48 h. The thermosensitive strains selected by the 2 nd screening were subjected to the 3 rd screening in liquid culture. After 8-h preculture, cells were diluted to a turbidity corresponding to OD 600 of 0.1 and inoculated into LB medium at the final OD 600 of 0.001. Samples were then incubated at 47uC for 18 h under a shaking condition. Thermosensitivity was determined by measuring OD 600 . Thermosensitive strains were defined to be ,0.1 at OD 600 . The experiments were performed three times, and the results were confirmed to be reproducible.

Effects of glucose and MgCl 2 and sensitivity to H 2 O 2
To examine effects of supplements, glucose (0.5% (w/v)) or MgCl 2 (20 mM) was added to the LB liquid culture. After 8-h preculture, cells were diluted to a turbidity corresponding to OD 600 of 0.1 and inoculated into LB medium with or without the supplement at the final OD 600 of 0.001. Samples were then incubated at 47uC for 18 h under a shaking condition. After 18 h, turbidity at OD 600 was measured. To test the sensitivity to oxidative stress, H 2 O 2 was added to the culture medium at the final concentration of 0.5 mM. After 8-h preculture, cells were diluted to a turbidity corresponding to OD 600 of 0.1 and inoculated into LB medium with or without H 2 O 2 at the final OD 600 of 0.001. Samples were then incubated at 30uC for 8 h under a shaking condition. The experiments were performed three times, and the results were confirmed to be reproducible.

RT-PCR analysis
Cultures were grown in LB medium at 37uC until the exponential phase, and then the temperature was up-shifted to 47uC and incubation was continued for 8 min. Total RNA was immediately prepared from the heat-stressed cells by the hot phenol method [68]. RT-PCR analysis was performed using an mRNA-selective RT-PCR kit (TAKARA BIO Inc, Otsu, Japan) to examine the expression of immediate downstream genes of disrupted genes as described previously [69]. The primer set used for each gene is shown in Table S5. The RT reaction was carried out at 42uC for 15 min, 85uC for 1 min, 45uC for 1 min and extension at 72uC for 2 min using the two specific primers for each gene. After the completion of 15, 20, 25 and 30 cycles, the PCR products were analyzed by 0.9% agarose gel electrophoresis and stained with ethidium bromide. The relative amounts of RT-PCR products on the gel were compared by measuring the band density after the color of the image taken had been reversed using a model GS-700 Imaging Densitometer (Bio-Rad Laboratories, Inc, Tokyo, Japan) [70].

DNA chip analysis
W3110 cultures were grown in LB medium at 37uC until the exponential phase, and then the temperature was up-shifted to 47uC and incubation was continued for 8 min. A control culture was incubated in parallel at 37uC for 8 min. Total RNA was immediately prepared from the heat-stressed cells by the hot phenol method [69]. Preparation of cDNA, fragmentation and the end-labeling of DNA fragments were performed according to the instruction manual from Affymetrix. The ENZO Bioarray terminal labeling kit (Enzo Life Sciences, Inc, New York, USA) was used to end-label DNA fragments. DNA hybridization, data capture and analyses were performed as described in the protocol supplied by Affymetrix and GCOS software (Affymetrix, Inc, California, USA). Two independent experiments were performed and four data sets (two data sets at 37uC: 37uC-1 and 37uC-2, two data sets at 47uC: 47uC-1 and 47uC-2) per gene were obtained. The expression ratio used here indicates the average of the ratios obtained in the two independent experiments. Spots with a significantly lower (,0.50; i.e., a negative fold difference) or higher (2.; i.e., a positive fold difference) fluorescence ratio of the heated sample to the control sample were considered to represent a real significant difference. Physiological function and functional classification of the genes were derived from the Genobase database (http://ecoli.aist-nara.ac.jp/). Array data are accessible through ArrayExpress accession number E-MEXP-3191. Figure S1 Growth of thermosensitive mutants in LB liquid culture at different temperatures. Each 51 thermosensitive mutant strain (opened symbols) and the parental strain, BW25113 (closed symbols), were grown in 30 ml LB medium at 37uC (circles), 45uC (squares), or 46uC (triangles). At the times indicated, turbidity at OD 600 was measured. A, group A; B, group B; C, group C; D, group D; E, group E; F, group F; G, group G; H, others.