Transcriptional Analysis of Long-Term Adaptation of Yersinia enterocolitica to Low-Temperature Growth

Bacterial strains and plasmids used in this study are listed in Table 1. All cultures were grown in Luria-Bertani (LB) broth (10-g/liter tryptone, 5-g/liter yeast extract, 5-g/liter NaCl), or on LB agar (LB broth supplemented with 1.5% agar). E. coli was grown at 37°C, and Y. enterocolitica was grown at the indicated temperature. If appropriate, the media were supplemented with the following antibiotics: kanamycin (50 μg/ml), nalidixic acid (20 μg/ml), tetracycline (12 μg/ml), or chloramphenicol (20 μg/ml). Strains were frozen in 13% glycerol at −70°C for permanent collection.

Transposon-bearing suicide plasmid pUT mini-Tn5 luxCDABE Km2 (17, 74) was transferred from E. coli donor strain S17.1 λpir (60) to acceptor strain Y. enterocolitica W22703 (14) by liquid mating as previously described (74) with the following modifications: donor and acceptor strains were grown to an optical density at 600 nm (OD₆₀₀) of 0.5, and 100 μl of each culture was mixed and supplied with 800 μl of 10 mM MgSO₄ solution. Following an incubation time of 5 min at room temperature (RT), cells were pelleted (3,000 rpm; 5 min at RT), resuspended in 100 μl LB broth, and spread onto antibiotic-free LB agar plates. The mating was incubated for 24 h at 30°C, and approximately 20% of the bacterial lawn was scraped off carefully and resuspended in 1 ml LB broth containing 13% glycerol. To avoid redundancy of the mutant library, five independent matings were performed, and aliquots of each mating were frozen in liquid N₂ and stored at −70°C. Frozen aliquots were adjusted with LB medium to an OD₆₀₀ of ∼0.3. Subsequently, 100 μl of this suspension was plated on LB selection plates (145-mm diameter) with kanamycin (200 μg/ml) and nalidixic acid and incubated for 24 h at 30°C. From each mating, 2,400 to 3,600 mutants were obtained, and each third colony was individually transferred to 96-well master plates filled with 800 μl LB containing antibiotics, as above, per well. Cells were cultivated for 16 h at 30°C on a shaker for microtiter plates. These overnight cultures were used for a subsequent luciferase-screening assay. In a preliminary test, Y. enterocolitica strain NCTC10460 instead of W22703 was used, leading to the identification of the cspA::luxCDABE mutant.

Random integration of the transposon insertions was experimentally proven by Southern blot analysis using DNA samples of 10 transposants. For that purpose, chromosomal DNA was restricted with Alw44I, gel separated, and hybridized with a 542-bp fragment from luxA, yielding bands of various length (data not shown). The resulting library of 5,700 luxCDABE transposants was used for the bioluminescence measurements. As an example, the genotype resulting from the random insertion of mini-Tn5 luxCDABE into cspB is schematically shown in Fig. 1.

The screening assay to identify temperature-responsive promoters was performed with a set of 60 microtiter plates. For that purpose, 96-well plates with clear bottoms (Matrix Technologies, Hudson, N.H.) were inoculated from the master plates and shaken overnight at 500 rpm and 30°C. Fresh cultures were prepared in duplicate by the dilution of 10 μl of each overnight culture into 190 μl LB medium. One 96-well plate was continuously incubated at 30°C, and the other one was incubated at 30°C for 1 h, cooled to 10°C, and further incubated at this temperature. Of both plates, OD₄₀₅ and bioluminescence (at 490 nm), the latter indicated as relative light units (RLU), were measured in parallel every hour with a Wallac VICTOR² 1420 multilabel counter (Perkin Elmer Life Sciences, Turku, Finland). To avoid interference, high-quality interference filters with an 8-nm bandwidth were used. To allow a direct comparison of 10°C and 30°C measurements, relative light units were referred to the growth phase by dividing the RLU by the respective OD₄₀₅ value (RLU/OD). Following this normalization of data, we identified the start of the temperature-dependent promoter response as the OD value exhibiting the first significantly increased bioluminescence signal that was maintained or enhanced for at least a further 4 h.

The luciferase activity in cells growing at 10°C has been shown to decrease 12.3 fold in comparison to growth at 30°C (11). This correlates with other experimental data (41) and with the Arrhenius equation predicting a 1.5- to 4-fold reduction in enzyme reaction rate, following a temperature decrease of 10°C, given that the light production by the luciferase genes implies four independent catalytic steps (68). The corrected light emission at 10°C was thus determined as the ratio RLU/OD, multiplied by the factor 12.3. The response ratio of each mutant was calculated as the quotient of maximal RLU/OD at 10°C times 12.3 and maximal RLU/OD at 30°C.

Chromosomal DNA of each cold-induced transposon mutant was isolated and completely digested with ClaI, HindIII, SphI, SspI, or DraI (MBI Fermentas, Vilnius, Lithuania). Fragments were treated with T4 DNA ligase (Gibco, California) to allow self ligation resulting in circular molecules, and subsequent inverse PCR (50) was performed using transposon-specific primers derived from the O end or the I end of mini-Tn5 (Table 2). The PCR fragments obtained were sequenced by SequiServe (Vaterstetten, Germany) and MWG (Ebersberg, Germany) with primers hybridizing within the transposon at a distance of 26 to 97 bp to the O end or the I end. Each obtained sequence was screened for ClaI, HindIII, SphI, SspI, or DraI restriction sites and for the partial transposon sequence. Mapping of the mini-Tn5 luxCDABE insertions was conducted by using the Y. enterocolitica BLAST Server from the Sanger Institute (http://www.sanger.ac.uk/Projects/Y_enterocolitica/ ). Sequences without similarities to sequenced strain Y. enterocolitica 8081 were classified as specific for strain W22703. Promoter sequences located upstream of the transposon insertions were identified using a promoter prediction program (http://www.fruitfly.org/seq_tools/promoter.html ).

To generate knockout mutants of uhpC, ybcM, and YE2575 by plasmid insertion via homologous recombination, short intragenic fragments from the target genes were amplified from Y. enterocolitica chromosomal DNA using the primers listed in Table 2. Fragments were digested with XbaI (MBI Fermentas, Vilnius, Lithunia) and SacI (New England Biolabs, Beverly, Mass.) and ligated into XbaI/SacI-restricted suicide plasmid pKRG9. The recombinant plasmids were transformed into E. coli S17.1 λpir by electroporation and transferred into each individual Y. enterocolitica strain by plate mating (see above). Transformants were selected on plates containing nalidixic acid, kanamycin, and chloramphenicol. To exclude illegitimate recombination, the correct insertion of the recombinant plasmid was confirmed by PCR using a gene-specific test primer and a plasmid-derived primer. Y. enterocolitica transposon mutants with and without the insertional knockouts were grown in parallel at 30°C and 10°C to measure bioluminescence and optical density as described above. To complement the double mutants W22703-YE0480::lux-YE2575′ and W22703-YE0480::lux-uhpC′, the coding sequences of YE2575 and uhpC, the first gene together with 362 bp of its upstream sequence, were amplified with the appropriate oligonucleotides listed in Table 2. The resulting fragments were cloned via ScaI (YE2575) and EcoRI (uhpC) into pACYC184. In the resulting recombinant plasmids, pACYC-YE2575 and pACYC-uhpC, the direction of transcription of both genes corresponds to that of the disrupted gene encoding chloramphenicol acetyltransferase. The constructs were verified by PCR and restriction analysis and transformed by electroporation into W22703-YE0480::lux-YE2575′ and W22703-YE0480::lux-uhpC′, respectively.

Promoter regions were amplified from purified Y. enterocolitica DNA by PCR using the primers listed in Table 2. PCR products were digested with EcoRI and SalI (MBI Fermentas, Vilnius, Lithunia), ligated (T4 DNA ligase; Gibco, Hudson, N.H.) into the SalI/EcoRI site of pPROBE-NT (42) and transformed into DH5αMCR. Plasmids containing the correct transcriptional gfp fusions as verified by PCR were isolated (GenElute Plasmid Mini Prep Kit; Sigma-Aldrich, Taufkirchen, Germany) and transformed into Y. enterocolitica. Bacterial cells bearing the recombinant gfp promoter probe vectors were grown in parallel at 30°C and 10°C in LB medium supplemented with kanamycin. From both cultures, 5-ml samples were harvested at an OD₆₀₀ of approximately 1.8, centrifuged (8,000 rpm for 2 min at RT), shock frozen, and stored at −70°C. Pellets were washed once with 500 μl phosphate-buffered saline, centrifuged again, and resuspended in 500 μl phosphate-buffered saline. Quantitative fluorescence assays were performed using a Wallac VICTOR² 1420 multilabel counter (Perkin Elmer Life Sciences, Turku, Finland) with sterile, untreated, black 96-well microplates (Nunc, Wiesbaden, Germany). Suspensions (each, 200 μl) were measured at wavelengths of 485 nm for excitation and 520 nm for emission. Intensity readings are presented as relative fluorescence units.

Out of 5,700 transposon mutants, 4,454 (78%) mini-Tn5 luxCDABE insertions into Y. enterocolitica strain W22703 showed a response ratio of <2.0 or an average light emission at background level of 200 RLUs and were classified as not differentially expressed at both temperatures. Finally, 1,238 clones were identified to be steadily upregulated at 10°C with a response ratio of at least 2.0. Although a twofold difference in relative transcript levels is a threshold commonly used for analysis and interpretation of microarray data (32), we used a threshold of 5.0 to further analyze only mutants with strongly elevated activity of the respective promoter. From the 222 mutants matching this criteria, about 40% showed light emission profiles congruent to those of mutants already identified, indicating a redundant transposon insertion, or exhibited unsteady, unreproducible profiles with strong data fluctuation. Those perturbations were especially derived from promoters with transcriptional activity slightly above background. Another 11% were withdrawn because they were false positive, due to neighboring mutants with strongly up-regulated light production in the microtiter plate. In total, we selected 109 clones for further characterization, representing approximately 1.9% of the 5,700 transposon mutants, plus 5 clones as examples for strongly inhibited luciferase activity at low temperature. The DNA of those mutants for which we observed response ratios of >5.0 or <0.6 and a steady expression pattern was digested and used for inverse PCR as described. The sequences of chromosomal fragments flanking the transposon insertion were compared to the Y. enterocolitica DNA assembly, and the results of the BLAST analysis allowed the precise localization of the transposon insertion and the orientation of the inserted reporter genes with respect to the Y. enterocolitica genome sequence, except for strain-specific fragments. This analysis revealed 42 different genes or open reading frames (ORFs) with enhanced (37 genes) or reduced (5 genes) promoter activity at 10°C compared to 30°C, 13 genes of which are probably localized within putative operon structures. This finding suggests that all proteins encoded by these putative operons are present in the cell in higher or lower copy numbers in response to low temperature. It should be noticed that the 37 genes represent all mutants characterized by a response ratio of the luciferase reporter of >5.0, whereas the five strains with down-regulated promoter activity are examples only of a larger group of mutants not further characterized. For two transposon insertions, no promoter responsible for luciferase induction could be identified, due to preliminary annotation of the respective genome region. An accumulation of independent transposon insertions was found within three ORFs or their surroundings, namely 13 insertions in both tcaB and tcaA and 10 insertions in a putative alkaline serine protease (YE2922). All transposon insertions are listed in Table 3.

Given that the library is random and that the genome of Y. enterocolitica is composed of approximately 2,500 transcriptional units similar to E. coli (68), our library should allow testing of approximately 70% of all genes or operons for differential expression (38). For this calculation, we assumed that the mini-Tn5 luxCDABE insertion in 50% of the mutants was not under control of a promoter located upstream. Despite this high knockout fraction, none of the insertions affected a gene or gene complex that is dispensable for cell growth at 30°C but specifically required for growth at low temperatures. Of the tested mutants, 0.5% were growth deficient at 10°C, as well as at 30°C.

To prove the differential transcription of the identified genes, we cloned the first assigned or putative promoter regions located upstream of nine mini-Tn5 luxCDABE insertions and the promoter region of cspB as a control into vector pPROBE-NT and measured GFP expression at 30°C and 10°C. Since the autocatalytic protein modification required for fluorescent light emission by GFP proceeds slowly in bacterial cells (2), fluorescence values were taken at an OD₆₀₀ value of 1.8. The background activity derived from a nonrecombinant vector in Y. enterocolitica alone was 3.9-fold (±0.3) higher at 30°C than at 10°C, and this factor was used to normalize the derived gfp reporter activities at 10°C (Fig. 2). This factor describes the reduced synthesis rate of GFP, due to the temperature decrease, and correlates with the Arrhenius equation predicting a factor of 1.5 to 4.0 (57). In all cases but one, the investigated promoters were shown to be induced 1.1 to 102 fold with an absolute range of fluorescence units from 2,536 RFU (urease operon) to 3,119,696 RFU (flagellar operon) at 10°C. An exception was observed when a gfp fusion to a putative promoter region directly upstream of YE0480 was tested. However, further annotation studies revealed that YE0480 is probably transcriptionally coupled to gene YE0479, a finding that might explain the negative result. In another cold-responsive mutant that we examined by GFP fusions, the luxCDABE genes had inserted between the ORFs YE0950 and YE0951, and a putative promoter upstream of YE0951 was found to be opposite to the direction of luxCDABE transcription. However, when the predicted promoter was amplified from this intergenic region and cloned into pPROBE-NT in correct orientation, a 2.4-fold induction of GFP at low temperature was observed (Fig. 2). This result and promoter prediction studies suggest that the selected promoter directs the polycistronic transcription of five genes (YE0951-yeuB-ureC-ureE-ureF) involved in urea metabolism. For all but one promoter probe constructs, the response ratio of gfp was significantly lower than the values deduced from the bioluminescence signal. These discrepancies are probably due to differences in both assays in copy numbers of reporter genes, half-lives, number of genes involved in light emission, reporter sensitivity, and experimental design.

The kinetic data derived from the light emission of transposon recipients grown at 30°C and at 10°C revealed a pattern of varying gene expression profiles due to mini-Tn5 luxCDABE insertions under the control of temperature-responsive promoters (Fig. 3). The interpretation of these profiles allowed (i) determination of luciferase induction with respect to the OD₄₀₅ value, (ii) quantification of the transcriptional response to low temperature by the normalized response ratios of luciferase fusions, and (iii) qualification of the induction as transient or permanent for a certain time interval. The response ratios of the mentioned genes were ranged from 6 to 1,147 (Table 3), with the maximal induction of a hypothetical protein (YE4063) that was not transcribed at 30°C and showed pronounced similarity (E value, 10⁻²⁴) to conserved domains of signaling proteins. A summary of the most important genes induced during long-term response of Y. enterocolitica is provided in Fig. 4. In contrast to cold shock proteins (CSPs), several transcriptional profiles showed no clear peak but steadily increased after temperature downshift and remained at a higher expression level, e.g., the mutants YE2057::luxCDABE and uhpC::luxCDABE (Fig. 3B and I). On the other hand, the change in promoter induction was only transient in transposon mutants like YE2537::luxCDABE (Fig. 3C), indicating a maximal expression of the basal body M-ring protein at an OD₄₀₅ of 0.8.

The genes of 37 transcriptional units significantly up-regulated during growth at 10°C were grouped into three classes according to the growth phase at which transcriptional induction was detected. At acclimatization phase (OD₄₀₅ < 0.3), two transposon insertions were located within the genes encoding CspA and CspB, showing a strong normalized response ratio of 84 and 268 after cold shock. CspA was transiently expressed with a peak 4 h after temperature downshift (Fig. 3J), while light emission driven by the cspB promoter decreased only slightly during prolonged growth (Fig. 3E). The light emission profile of the cspA::luxCDABE mutant correlated with Northern blot analysis, showing inhibited cspA transcription 120 min after cold shock (46) and with the relative amount of the major CSPs as determined by two-dimensional gel electrophoresis (Fig. 3J) (45), thus demonstrating the validity of chromosomal luxCDABE promoter fusions as a tool to measure transcriptional activities. The temporal distance between the maximal amount of CspA and the corresponding luciferase activity was calculated as 180 to 195 min. The sustained luciferase activity of the cspB::luxCDABE knockout mutant might be explained by an autoregulatory effect known to occur for E. coli cspA (4). Parallel to both CSPs, mini-Tn5 luxCDABE insertions within the promoter region of ymoA, which plays a role in thermoregulation of virulence functions (15), and within the ORF of gltP, encoding a glutamate-aspartate symport protein, were shown to be upregulated within the acclimatization phase.

Seventy percent of the transcriptional units identified in our approach were induced at 10°C before they reached the late growth phase, most of them involved in sensoring-regulation, motility, and virulence. Besides several as-yet-uncharacterized genes, we found arcA (YE0595), which is necessary to survive prolonged starvation (59), uhpABC, controlling the expression of the hexose phosphate transporter UhpT (70), and two obviously strain-specific loci without homology to known genes. The most prominent response to temperature decrease was the enhanced transcription of motility and chemotaxis genes such as the flagellin operon or cheA. Other differentially expressed genes are YE2057, which encodes a homologue of the putative virulence factor SrfA of Salmonella, mgtC (YE2586) involved in magnesium uptake (44, 65), and YE2463 encoding an outer membrane porin. Surprisingly, we identified two mini-Tn5 luxCDABE insertions in the genes tcaA and tcaB encoding two subunits of Tca, a homologue of the insecticidal toxin complex proteins from Photorhabdus luminescens, and we measured response ratios of 12 and 81 when cultivating these mutants at 10°C (11). We also observed the induction of two putative operons involved in urease activity. One was YE0951-yeuB-ureC-ureE-ureF, encoding urea amidohydrolase, an accessory protein, and three urease subunits. The other operon was YE0959-YE960, which encodes a urease accessory protein and a urea transporter.

During late exponential growth, we observed the stimulation of the following genes: YE0480, encoding a predicted hemolysin transporter with homology to the accessory protein FhaC from Bordetella pertussis (identity, 28%; E value, 2⁻⁴⁷); the monocistronically organized gene YE2922, encoding a homologue of subtilisin (identity 33%; E value, 6⁻²¹), which is an alkaline serine protease associated with the onset of sporulation in B. subtilis (67); two genes encoding a histidine ammonia-lyase (hutH) and a putative amylase (YE4013); and one gene (YE3697), encoding a permease of the drug/metabolite transporter (DMT) superfamily.

Due to its high dynamic range, the luciferase reporter also allows the assessment of decreased transcriptional activity in long-term response to temperature changes (75). We mapped five mini-Tn5 luxCDABE insertion mutants responding to growth at 10°C with strong reductions in luciferase activity. One of these insertions was found to be located within gene YE4164, encoding a protein identical to heat shock proteins of the IbpA family, probably cotranscribed with YE4163, which encodes an IbpB-like heat shock protein. At 10°C, very low luciferase activity corresponding to the background level was observed for the putative promoters in front of gene YE4121, which encodes a sugar-binding periplasmic protein, and upstream of YE3774, which encodes a putative phosphotrehalase as part of a phosphotransferase system. This is in line with the observed elevated transcription levels of cheA at low temperature, resulting in a tumbling behavior of bacterial cells. The activity of the phosphotransferase system, however, that transports carbohydrate attractants into the cell is known to extend swimming runs (40). Interestingly, the sequence obtained from another transposon insertion was identical to an as-yet-unannotated region of the Yersinia virulence plasmid pYVe8081 immediately downstream of yopT/sycT, and the maximal light emission of 20,489 RLU/OD obtained for the respective mutant at 30°C was reduced by 1 order of magnitude to 3,865 RLU/OD (Table 3).

The above-described determination of induction time points of gene expression should allow potential interdependencies between the cold-responsive genes found to be uncovered. To examine the effect of knockouts of three regulatory genes from Table 3 on selected transposon mutants, insertional duplication mutagenesis of the three regulatory genes uhpC, ybcM, and YE2575 was performed with a set of 24 temperature-responsive luxCDABE mutants, resulting in a total of 72 double mutants. None of the insertion mutants showed deficient growth at 10°C or 30°C, indicating that the double knockout did not affect cell viability (data not shown). The promoter activities of the double mutants were measured at 10°C and 30°C, and the data were compared to the bioluminescence values of the corresponding mutants without additional insertional knockout mutations. In most cases, this analysis did not reveal significant changes in promoter activities of the double knockout mutants in comparison to the single luxCDABE insertion mutants. However, two exceptions were observed: the activity of the lux reporter inserted in YE0480, encoding an FhaC-like transporter, was enhanced by the factor 4.8 and 5.6 at 15°C as a result of the insertional knockouts of YE2575 and uhpC, respectively, and by a factor of approximately 10 at 30°C (Fig. 5). The cloned uhpC and YE2575 genes complemented the mutations; in the case of uhpC, the light signal of the complemented strains exceeded that of W22703-YE0480::lux, probably due to the lack of the native uhpC promoter. These results indicate that the two-component system UhpABC, as well as methyl-accepting chemotaxis protein I (MCPI) involved in chemotaxis, negatively regulates the expression of the putative hemolysin transporter encoded by YE0480.

In many cases, gene induction has been studied immediately after sudden changes, e.g., cold shock, acid shock, or heat shock, while much less effort has been dedicated to analyzing gene expression during prolonged growth under a specific environmental stress. It is a significant result of this work that there is no overlap of the well-described cold acclimatization response directly following cold shock and the genes induced during exponential growth at low temperature. This is supported by our negative results of an in silico search for consensus regulatory regions within promoter regions or translational start regions of cold shock genes, e.g., the putative downstream box or the cold box (22, 43). Similar observations have been reported for the UV-B shock response and subsequent growth under elevated UV-B of the cyanobacterium Nostoc commune (19), and for L. monocytogenes grown to mid-exponential phase (39). The pronounced separation between shock response and long-term adaptation may be a general feature of bacterial gene expression induced by environmental stress.

At least two virulence factors of Y. enterocolitica have been known to be maximally expressed at ambient temperatures, the heat-stable enterotoxin Yst (10) and the outer-membrane protein Inv (53). One of the most important aspects of differential gene expression in terms of the Y. enterocolitica temperature response that emerged from our study is the enhanced transcription of another four genes associated with virulence functions. At low temperature, we observed the upregulation of tcaA and tcaB encoding homologues of the insecticidal toxin subunits from P. luminescens that is found in the gut of entomopathogenic nematodes (72). Extracts of Y. enterocolitica grown at 10°C but not 30°C were shown to be lethal for insect larvae, and tcaA was demonstrated to be essential for this insecticidal activity. This finding gave rise to speculations about an as-yet-undisclosed niche of Y. enterocolitica in insects (11). The transcription of another putative virulence factor, a homologue of the SsrAB-regulated factor SrfA (76), was also shown to be low-temperature responsive. This is in line with the observation that srfA transcription is repressed fivefold during intracellular growth in macrophages at 37°C (21). Interestingly, microarray data of Salmonella typhimurium motility only recently revealed the surface-dependent regulation of several virulence genes, among them srfABC, thus demonstrating the association between swarming behavior and virulence, as observed with many bacteria (28, 71). YE0480 is possibly involved in hemolysin secretion, and hemolysins are predicted determinants of Photorhabdus pathogenicity against insects (12). It is remarkable that neither tcaA, tcaB, srfA, nor YE0480 showed significant transcriptional activity at 37°C (data not shown). The correlation of bacterial pathogenicity with exponential phase is reminiscent of a molecular model that describes factors involved in the association of Photorhabdus with nematodes and insects (24). Taken together, these data suggest a role of these virulence factors outside mammalian hosts, possibly in invertebrates.

The specific expression profile at low temperature discussed here strongly supports the view of temperature as an important stimulus for the long-term adaptation of Y. enterocolitica to environmental changes, mediated by sensory proteins and regulatory networks. In our approach, we observed the induction of several regulatory systems of known (ArcA and UhpABC) and predicted (YE1436, ybcM, YE4063, and YE1324) function, many of them involved in signal transduction (Table 3). The ArcAB system controls the expression of at least 40 operons involved in catabolic gene expression (3, 59), and arcA is suggested to control the rate of utilization of endogenous reserves (49). None of our results points to the presence of a cold regulon during prolonged growth at 10°C. However, YE0480, a gene encoding a homologue of the accessory protein FhaC from B. pertussis, is strongly expressed at 10°C but not at 37°C, and its transcription is negatively influenced by MCPI and UhpC, suggesting that YE0480 expression is connected with nutrient acquisition and motility regulons. Interestingly, a similar regulatory circuit was reported recently for expression and activity of a Xenorhabdus nematophila hemolysin that is required for full virulence against insects (16). These interdependencies confirm the linkage between motility and virulence properties at the level of regulation (18). The induction of flagellae and chemotaxis genes at low temperature is a well-characterized phenomenon that is in line with the finding that loss of motility at 37°C does not affect pathogenesis of yersiniae in humans (33), indicating a role of motility in the colonization of new habitats. The master flagellum regulatory operon, flhDC, is assumed to be required for full Xenorhabdus virulence against insects. Moreover, the flagellar sigma factor of Y. enterocolitica, σ²⁸, controls the phospholipase gene yplA, whose product is exported by the flagellum secretion apparatus (58).

A predominant feature of the derived expression profile at environmental temperatures is the prevalence of genes involved in biodegradative metabolism during late growth and stationary phase. We found the significant upregulation of two urea operons and of a histidine ammonia-lyase (YE3021) with homology to the HutH histidase, all of them involved in nitrogen utilization. In addition, low temperature results in elevated expression of the α-glucan branching enzyme GlgB (YE4013) that, together with the α-amylase GlgX and GlgC, regulates the glycogen metabolism (56). Interestingly, the degradative hut operon, as well as an α-amylase, have been shown to be upregulated at low temperature in several antarctic psychrotrophic bacteria (23, 36). A role of the Y. enterocolitica alkaline serine protease, a hot spot of mini-Tn5 luxCDABE insertions and a member of the extracellular subtilisin family (67), has not yet been described. Proteases, however, represent exoenzymes that might help to degrade tissues and macromolecules of a eukaryotic host. None of the genes mentioned as being involved in degradative metabolism is expressed at 37°C (data not shown), again suggesting a role of these enzymes outside a mammalian host. It is therefore tempting to speculate that these genes play a role in bioconversion of an insect cadaver, providing a nutrient base for bacterial growth. Taken together, the long-term portrait of cold-responsive genes as derived from the luciferase assay is a promising starting point for further analysis of the association of Y. enterocolitica with invertebrates.