We have identified a new folding catalyst, PpiD, in the periplasm of Escherichia coli. The gene encoding PpiD was isolated as a multicopy suppressor of surA, a mutation which severely impairs the folding of outer membrane proteins (OMPs). The ppiD gene was also identified based on its ability to be transcribed by the two‐component system CpxR–CpxA. PpiD was purified to homogeneity and shown to have peptidyl–prolyl isomerase (PPIase) activity in vitro. The protein is anchored to the inner membrane via a single transmembrane segment, and its catalytic domain faces the periplasm. In addition, we have identified by site‐directed mutagenesis some of the residues essential for its PPIase activity. A null mutation in ppiD leads to an overall reduction in the level and folding of OMPs and to the induction of the periplasmic stress response. The combination of ppiD and surA null mutations is lethal. This is the first time two periplasmic folding catalysts have been shown to be essential. Another unique aspect of PpiD is that its gene is regulated by both the Cpx two‐component system and the σ32 heat shock factor, known to regulate the expression of cytoplasmic chaperones.
Bacteria have evolved many adaptive systems in order to cope with various environmental stresses. Among these, heat shock produces a highly conserved response which in Escherichia coli is regulated by two alternative sigma factors, σ32 and σE. These two factors govern the transcription of two heat‐shock regulons which, respectively, are specialized in coping with protein misfolding in the cytoplasm and the extra‐cytoplasm, i.e. periplasm and outer membrane. Protein misfolding in the extra‐cytoplasm induces an additional stress response (other than the σE‐dependent stress response), which is regulated by the two component system CpxR–CpxA. The periplasmic protease HtrA, involved in the degradation of misfolded polypeptides, is encoded by a gene which is transcribed both by the EσE polymerase and the CpxR activator (Danese et al., 1995; Raina et al., 1995). However, fkpA which encodes a peptidyl–prolyl isomerase of the FKBP family in the periplasm (Missiakas et al., 1996), belongs to the σE regulon (Danese and Silhavy, 1997), whereas dsbA and ppiA, which encode periplasmic thiol:disulfide oxido‐reductase and peptidyl–prolyl isomerase respectively, are under the control of the Cpx system (Danese and Silhavy, 1997; Pogliano et al., 1997).
Unlike the cytoplasm which contains many ATP‐dependent chaperones with wide substrate specificity, the periplasm appears to contain two defined types of folding catalysts. Folding of most translocated proteins encounters two types of rate‐limiting steps, which are overcome by two classes of catalysts: protein disulfide isomerase (PDI) and peptidyl–prolyl cis–trans isomerase (PPIase). In E.coli, there are at least six known Dsb proteins (reviewed by Missiakas and Raina, 1997a). These proteins are involved in the oxidation of disulfide bonds or the rearrangement of wrongly paired disulfides (Bardwell et al., 1991; Missiakas et al., 1994; Zapun et al., 1995; Rietsch et al., 1996). PPIases are found in both the cytoplasm and the periplasm. They catalyse the rapid interconversion between the cis and trans forms of the peptide bond Xaa–Pro. Three distinct families of PPIases have been identified so far; these include the cyclophilins (PpiA in the periplasm), FKBP‐like proteins (FkpA in the periplasm) and the newly discovered parvulin family. The first member of the parvulin family was identified as a cytosolic E.coli protein (Rahfeld et al., 1994), whose gene was later designated ppiC. Interestingly, the yeast PpiC homologue, ESS1, is the only essential PPIase for yeast. Surprisingly, yeast mutants lacking all the other 12 PPIases are viable, although some of them are heat‐shock inducible (Dolinski et al., 1997). In E.coli, a second member of the parvulin family, known as SurA, was identified in the periplasm. SurA is involved in the correct folding of outer membrane protein (OMP) monomers. We isolated this gene as a multicopy suppressor of htrM (Missiakas et al., 1996). htrM mutants synthesize altered lipopolysaccharide (LPS), and as a consequence lack most of the OMPs in their outer membrane. This suppression effect is due specifically to correction of the proper outer membrane profile. Also, the accumulation of misfolded OMPs was shown to induce the σE‐dependent response to protein misfolding in the extra‐cytoplasm. Overexpression of surA was shown to dampen this response in many cases where misfolded polypeptides accumulate, suggesting a general chaperone‐like function for SurA. We and others have shown that mutations in surA confer severe defects in the amounts or maturation of OMPs (Lazar and Kolter, 1996; Missiakas et al., 1996; Rouvière and Gross, 1996). Bacteria lacking surA turn on the σE stress response constitutively (Missiakas et al., 1996; Rouvière and Gross, 1996). Finally, surA mutants are extremely sensitive to detergents and hydrophobic drugs such as novobiocin, phenotypes reminiscent to the leakiness of the outer membrane.
In the present study, we took advantage of the hypersensitivity of surA mutants to novobiocin and selected for multicopy suppressors. We identified a new gene and based on its sequence homology to members of the parvulin family, we have designated this gene ppiD. ppiD was also identified in an independent genetic screen aimed at identifying new members of the CpxR–CpxA regulon.
We show here that ppiD encodes a membrane‐anchored polypeptide of 623 amino acids, of which the last C‐terminal 589 residues are located in the periplasm. Transcriptional analysis of ppiD revealed that the gene belongs to two stress regulons: the CpxR–CpxA regulon and the σ32 regulon. This is the first member of a periplasmic folding catalyst to be regulated by the classical heat‐shock sigma factor σ32.
ppiD is a multicopy suppressor of surA
We have shown previously that a null mutation in surA leads to a highly pleiotropic phenotype. One of the defects is hypersensitivity to antibiotics such as novobiocin, due to a higher permeability of the outer membrane. surA mutant bacteria do not form colonies in the presence of novobiocin concentrations >10 μg/ml, while isogenic wild‐type bacteria grow in the presence of 40 μg/ml of novobiocin (Missiakas et al., 1996). We constructed a chromosomal DNA library lacking surA and used it to transform the surA null mutant. Candidates resistant to 30 μg/ml of novobiocin were selected.
Plasmid DNA from seven such candidates was isolated and shown to breed true by retransforming a surA null mutant. The DNA was used to probe the ordered E.coli DNA library (Kohara et al., 1987). All seven plasmids hybridized to bacteriophage λ 148(3B6) and λ 149(7E2) clones. This corresponds to a genetic map position in the 9.5–10 minute region on the E.coli chromosome. This area includes the known serine protease, Lon, and the DNA‐binding protein, HupB. DNA from one such plasmid, pSR3239, was used to construct a minimal subclone 2.4 kbp BsaBI–SmaI DNA fragment in the p15A‐based vector, pOK12. This subclone, pCD51, was found to be sufficient to restore novobiocin resistance to the surA null mutant (Figure 1; Table I). Since our initial clone pSR3239 also carries the intact hupB gene, we subcloned the 1 kbp EcoNI–KpnI DNA fragment which carries the intact hupB gene (pCD56), and found that it does not complement a surA null mutant (Figure 1). Sequence analyses of pCD51 revealed the presence of a single complete open reading frame (ORF) which encodes a polypeptide of 623 amino acids. Sequence examination also revealed homology to PPIase. Hence, this gene was designated ppiD. It carries a putative parvulin‐like domain located between amino acids 227 and 357. This predicted amino acid sequence shares 34% identity with an ORF of unknown function from Haemophilus influenza (SwissProt accession No. P44092). The next closest homologue is a very recently sequenced ORF of unknown function from Acinetobacter (PID entry e1173385), with a sequence identity of 29% over its entire length. PpiD also has sequence similarity to the SurA protein. More precisely, residues 86–121 of PpiD align with residues 65–100 of SurA (44% identity). A second region lying between residues 334 and 357 of PpiD seems to correspond to the parvulin‐like domain of SurA (residues 251–274). It is known that SurA has two very similar parvulin‐like domains at its C‐terminus. These residues include the motif VGFHIL and are highly conserved among the members of the PpiC (parvulin) family. However, it should be pointed out that the highly conserved residues H and L are replaced by L and V, respectively, in the PpiD amino acid sequence. Also, the predicted amino acid sequence of PpiD shows the presence of a single parvulin‐like domain, whereas SurA carries two such domains.
Identification of ppiD as a new member of the CpxR–CpxA regulon
In our effort to understand the nature of the Cpx‐dependent stress response, we have constructed a library of transcriptional fusions using the single copy promoter probe vector pFZY (see Materials and methods). We have shown previously that overexpression of the prpA gene leads to the constitutive induction of the Cpx pathway (Missiakas and Raina, 1997b). Hence, we looked for promoter fusions, which are highly induced (deeper blue colonies) on X‐gal‐containing plates, in the presence of a plasmid carrying the prpA gene under control of the inducible lac promoter (pDM1574). We isolated ∼70 such clones and transduced a genetically well characterized cpxR null allele (Danese et al., 1995) into them. Only those clones which showed a decline in β‐galactosidase activity were retained (Figure 2). Fifteen such isolates were retained and plasmid DNA prepared from them was sequenced. Among them, 12 clones were found to contain sequences from the promoter region of the ppiD gene.
ppiD is also a bona fide heat shock gene under the control of σ32
To analyse the transcriptional activity and promoter usage of ppiD, we constructed a single copy ppiD–lacZ fusion using the promoter region of ppiD and the λRS45 and pRS550 vectors (Simons et al., 1987). Using this fusion, we observed that ppiD transcription is heat‐shock inducible (Figure 3), which is reminiscent of classical heat‐shock gene induction. The single copy ppiD–lacZ‐carrying strain (CD212) was transformed with a vector carrying either the rpoH gene under the control of an inducible promoter (pSR1332; Missiakas et al., 1993a) or the rpoE gene (pSR1628; Raina et al., 1995). As shown in Figure 3, induction of the rpoH gene product from pSR1332 produces a 10‐fold increase in β‐galactosidase activity using the ppiD–lacZ fusion, whereas no induction is observed with the rpoE+ plasmid (pSR1628). Furthermore, no reduction is observed in the presence of the rseA+‐carrying plasmid (pSR2661; Missiakas et al., 1997) which encodes the specific anti‐σE factor, which in high dosage represses the transcription from σE‐dependent promoters (Table II).
It is known that mutation in dnaK, dnaJ, grpE or htrC leads to a constitutive elevated induction of the σ32 heat‐shock response. The ppiD–lacZ‐carrying strain was transduced into the dnaK103 mutant. As can be seen in Figure 3, the β‐galactosidase expression is highly induced.
Taken together, the data clearly show that ppiD has another promoter in addition to its CpxR–CpxA‐dependent promoter, and that this promoter is under σ32 control. This is the first example so far of a periplasmic folding catalyst placed under the control of σ32.
Identification of the heat‐shock protein PpiD on two‐dimensional gels
The PpiD protein has not up to now been identified as a heat‐shock protein on two‐dimensional gel isoelectric focusing gels. The reasons could be that its molecular weight coincides with that of HtpG and DnaK (all are ∼70 kDa in size). In addition, the predicted pI of HtpG is 5.09 and that of PpiD is 4.95. PpiD, being an inner membrane protein, may migrate differently somehow, perhaps overlapping HtpG on these gels. In fact, in many of our earlier two‐dimensional heat‐shock gels, we did observe a protein induced by a 50°C heat shock which runs close to HtpG and DnaK. Hence, we constructed two isogenic strains (ppiD::ΩTet htpG+ and ppiD::ΩTet ΔhtpG) and compared the global heat‐shock profiles of cultures of these strains after a sudden 50°C heat shock for 5 min. Samples were analysed by two‐dimensional gel electrophoresis. As can be seen in Figure 4, there is a clear induction of an ∼70 kDa polypeptide in the ppiD+ htpG+ sample (Figure 4A) but not in that of ppiD::ΩTet ΔhtpG (Figure 4C). As a control for the spot identified as PpiD, we also added [35S]methionine‐labelled PpiD to heat‐shock extracts from ppiD::ΩTet ΔhtpG (Figure 4B. This clearly indicates that PpiD is a heat‐shock protein and accumulates at a temperature of 50°C. The positions of other heat‐shock proteins such as GroEL and DnaK are also shown in Figure 4B.
Acetyl phosphate and the cpxA* mutations influence the transcription of ppiD
We further examined the CpxR–CpxA transcriptional dependence of the ppiD gene. It is known that the level of acetyl phosphate modulates the activation or repression of two‐component systems such as CpxR–CpxA. Thus we determined the β‐galactosidase activity of ppiD–lacZ under different growth conditions which affect the level of acetyl phosphate in vivo. It is known that the use of pyruvate as a carbon source results in the maximum level of intracellular acetyl phosphate. Indeed, a 3‐fold activation of the ppiD promoter can be obtained when cultures are shifted from glycerol to pyruvate medium (Figure 5).
We have also used the SR3570 strain carrying a cpxA* mutation in which T252 is changed to R252 (Missiakas and Raina, 1997b). This chromosomal allele leads to a gain of function of CpxA, presumably a hyperkinase activity (Missiakas and Raina, 1997b). This mutation is similar to a gain‐of‐function mutation in envZ12 (Aiba et al., 1989). We transduced the ppiD–lacZ fusion into cpxA* mutant SR3570, resulting in CD220. This strain again showed a 2‐ to 3‐fold induction of ppiD–lacZ activity (Table III). In control experiments we transduced a htrA–lacZ fusion into the same cpxA* mutant (Table III). The increased ppiD–lacZ expression is in fact more than that observed for the htrA–lacZ fusion. These results further confirm that the transcription of ppiD is in part regulated by the two‐component system, CpxR–CpxA.
Mapping of ppiD transcriptional start sites
In order to understand the transcriptional regulation of ppiD, we determined the transcriptional initiation site(s) under different growth conditions and in different genetic backgrounds. RNA was extracted from bacterial cultures grown either at 30°C or following a brief shift to 50°C. Using the primer extension technique, the transcriptional start sites were mapped. As can be seen in Figure 6, lane 2, using RNA from those cultures subjected to the 50°C shift, two transcriptional start sites designated as Phs1 and P* could be mapped 75 and 83 nucleotides, respectively, upstream of the translational initiation start site. No corresponding signal is seen from RNA extracted at 30°C, which clearly indicates that transcription from the start site designated as Phs1 is heat‐shock inducible. The −10 and −35 regions upstream of this start site show a perfect match with the σ32‐regulated promoters (for the −10 box CCCC and for the −35 box CTTGTG, to be compared with the consensus CTTGAA). In addition, the spacing between the two canonical boxes is 15 nucleotides, a feature common to most of the σ32‐regulated promoters. We do not know at present which sigma factor is responsible for the transcription initiating at nucleotide 83 (P*). The two additional start sites located at nucleotides 85 and 93 appear only in the RNA extracted at 30°C (lane 3) and not at 50°C. Thus, they are not heat‐shock induced and may correspond to σ70‐coupled CpxR–CpxA‐dependent transcription. Examination of the sequence shows at least three conserved CpxR boxes at positions −261 (GGTAAAGAG), −221 (GGTAAGC) and −209 (GGTAACT) (Figure 7) upstream of the translational initiation codon.
Deletion of the putative CpxR boxes leads to a decline in ppiD transcription
As mentioned above, there are three putative CpxR‐binding boxes (Figure 7) similar to what has been described for dsbA (Pogliano et al., 1997). To assay their role in vivo, we removed the first 147 nucleotides from plasmid pCD189 by digestion with BamHI and FokI, resulting in plasmid pCD191 (Table IV). This removes all of the three putative CpxR‐binding boxes. This promoter fusion still carries the σ32‐regulated promoter Phs1, as well as 43 nucleotides upstream of it. This promoter fusion was analysed and shows an overall decline in promoter strength (Figure 2). It no longer responded to changes in the level of acetyl phosphate (compare ppiD–lacZ and ppiD–lacZ CpxR boxes− in Figure 2). Also, introduction of the cpxR null allele did not change the ppiD promoter activity. Furthermore, this promoter fusion is not induced by overexpression of the prpA gene product, but is still subject to regulation by σ32, like the original fusion construct CD212. This further confirms that ppiD transcription is in part regulated by the two‐component system CpxR–CpxA, and that the whole nucleotide region deleted contains the CpxR‐binding boxes which are important in its regulation.
Identification, purification and localization of the PpiD protein
We constructed a minimal subclone of ppiD, pCD52, by PCR in the pET24a expression vector with a T7 promoter. This plasmid was first sequenced and then shown to complement ppiD fully and to suppress surA mutant bacteria. Expression of the protein was induced with isopropyl‐β‐d‐thiogalactopyranoside (IPTG) and, consistent with the predicted mass of the deduced amino acid sequence, was shown to be a protein of ∼69 kDa (Figure 8). Examination of the PpiD amino acid sequence predicted it to be an inner membrane protein, with a single transmembrane domain spanning amino acid residues 16–34 with its N‐terminus in the cytoplasmic membrane. The rest of the polypeptide (amino acids 35–623) is predicted to be in the periplasm. The overexpressed PpiD protein was indeed found to localize to the inner membrane (Figure 8A) and could be released either upon Triton X‐100 (1%) or sarkosyl (0.2%) treatment. The solubilized membrane fraction containing PpiD was first purified by gel filtration followed by an anion exchange chromatography. The purified protein was eluted with a linear gradient of KCl (0.05–1 M) containing 0.2% NP‐40.
The topology was confirmed further by constructing random TnphoA fusions (Manoil and Bailey, 1997). It is known that alkaline phosphatase is active in the periplasm only under exponential growth conditions in wild‐type bacteria (Derman and Beckwith, 1991). We sequenced the junctions between six randomly chosen AP+ PhoA fusions and found them to be located at residues 115, 223, 229, 309, 437 and 453. As further proof, we inserted the AP+ cassette (Gutierrez and Devedjian, 1989) at the unique PstI site in the ppiD gene. This site is located in the vicinity of the PpiC‐like domain present in PpiD. Like the other six AP+ fusions tested, this one was also found to be AP+. In order to verify that the N‐terminal part is located in the cytoplasm, we inserted an AP+ cassette after the sixth amino residue. This fusion was found to be AP−. Quantitative measurement of alkaline phosphatase activity in this fusion was 55 U versus >2000 U for periplasmic PhoA fusions. Taken together, all these data confirm that PpiD is located in the inner membrane, with the bulk of it facing the periplasm, including its parvulin‐like domain.
More conclusively, the membrane localization of PpiD from wild‐type bacteria was finally confirmed by immunoprecipitation, using total cell extracts, or the soluble or membrane fraction. As can be seen, most of the PpiD could be immunoprecipitated from either the inner membrane fractions or the total cell extracts only (Figure 8B). Thus, PpiD is clearly an inner membrane protein, since these experiments were carried out at the single copy level without any overproducing plasmid.
Construction of null alleles of ppiD
To study the in vivo role of ppiD, we constructed two null alleles, using the ΩKan cassette. One allele was constructed using plasmid pCD57 by inserting either the ΩKan or ΩTet cassette at the unique KpnI site (this site is located within the 5′ region of the ppiD coding sequence). The second allele was constructed by inserting the ΩKan cassette at the PstI site (Figure 1). These alleles were transferred to the chromosome as described by Link et al. (1997), and transferred to various E.coli backgrounds using bacteriophage P1 transduction.
As expected, all these null alleles were found to be >90% linked to the hupB gene which is located just upstream of ppiD. Since ppiD is heat‐shock regulated, we tested whether its product is required for bacterial growth at high temperatures. No noticeable growth difference was observed between the ppiD null mutant and the isogenic wild‐type for temperatures ranging from 30 to 43°C. A weak temperature‐sensitive phenotype was obtained when bacteria were plated at >43.5°C.
Phenotypic defects associated with ppiD null bacteria
Since the ppiD gene was cloned originally as a multicopy suppressor of the surA null mutation and the predicted amino acid sequence analyses show that they share some sequence homology, we examined the possibility that these two genes encode overlapping activities. We found that similarly to surA mutants, ppiD null mutants are hypersensitive to hydrophobic antibiotics such as novobiocin (>20 μg/ml), and to detergents such as SDS (>2%) (Table I). These phenotypes are often associated with membrane defects. The sensitivity to novobiocin and SDS is less drastic than that observed for bacteria carrying a null mutation in surA (Missiakas et al., 1996), but more pronounced than that observed for isogenic bacteria carrying a null mutation in the skp gene. The skp gene product is also needed for membrane integrity and maturation of OMPs (Chen and Henning, 1996; Missiakas et al., 1996). Interestingly, growth of a double skp ppiD null mutant was inhibited in the presence of 15 μg/ml of novobiocin and exhibited an increased sensitivity towards SDS (Table I). Also, all the three null alleles of ppiD accumulated extragenic suppressors at a very high frequency (>10−2).
Synthetic lethality of ppiD and surA null mutations
Because ppiD in multicopy suppresses the phenotypes associated with a surA mutant, and since mutations in either of the genes have defects in the outer membrane, we decided to examine the phenotypes of the double ppiD surA null mutant. However, repeated attempts to construct such a double null mutant failed. Hence, we reasoned that such a combination must be synthetically lethal. To confirm this, we showed that we could only cross the ppiD null allele into a strain CD269 which carries the surA gene cloned onto a plasmid. A converse set of experiments gave the same results, i.e. a surA null allele can be transduced into strain CD270 since it carries ppiD on a plasmid (Table IV). In this plasmid, the ppiD gene is under Plac control and carries the lac repressor gene lacIQ on the plasmid. The surA::ΩKan transductants could be obtained only in the presence of IPTG. Hence, these genetic experiments proved convincingly that the ppiD and surA mutant combination exhibits a synthetic lethal phenotype. These results are interesting when compared with the requirement for PPIases in other organisms. Yeast has 12 PPIases and none of them is essential for growth; in fact yeast remains viable even when all these 12 genes are deleted (Dolinski et al., 1997). However, there is an additional PPIase which is the only member of the PpiC family, ESS1, and its gene is an essential gene in yeast. In E.coli, the presence of two such homologues in the same compartment, SurA and PpiD, would explain why neither of them is essential for growth, although they have individual pleiotropic phenotype defects. These results may also explain the high occurrence of extragenic suppressors, which probably dampens the phenotypic defects of ppiD null mutant bacteria.
The synthetic lethality due to loss of both surA and ppiD function is unique. Escherichia coli has two more PPIases in the periplasm, namely FkpA and RotA. Double null mutants of fkpA::ΩTet ppiD::ΩKan as well as rotA::ΩKan ppiD::ΩTet are easily obtained (Table I). Thus, surA and ppiD gene products must have many overlapping functions as well as complementary functions, and may have some common essential substrate(s) which can account for the synthetic lethal phenotype.
Lack of ppiD leads to a constitutive induction of the σE‐dependent heat‐shock response
It is now well established that the absence of various folding catalysts in the periplasm induces the σE stress regulon, due to the accumulation of misfolded transported proteins (Missiakas et al., 1996; Missiakas and Raina, 1997a,b). If the ppiD gene product is indeed needed for folding of some periplasmic or outer membrane proteins, we should expect a null mutation to have increased EσE activity. The ppiD null alleles were transduced into different EσE promoter fusion‐carrying strains. As can be seen (Figure 9), there is an ∼3‐fold induction of EσE activity as reflected by the induction of β‐galactosidase activity from different promoters tested (htrA–lacZ, rpoEP2–lacZ and rpoHP3–lacZ) as compared with isogenic wild‐type strains. Also, a further additive effect is observed in a double skp and ppiD null combination. Furthermore, a combination of a ppiD null allele with a htrA mutation also leads to a further additive effect (Figure 9). This is quite similar to what we observed earlier by combining mutations of different catalysts of periplasmic folding with a htrA mutation (Raina et al., 1995). Since ppiD is a heat‐shock gene with one of its promoters under control of the rpoH, we also transduced ppiD null alleles into different promoter fusion backgrounds which are under σ32 transcriptional control, such as groEL–lacZ (CD261) and htpG–lacZ (CD259). The results clearly show that neither of the two Eσ32 promoters are induced (Table V). Thus, lack of ppiD gene product induces only the EσE‐dependent periplasmic unfolding response but not that of Eσ32‐transcribed genes. Thus, protein misfolding associated with the ppiD null mutant bacteria should be specific for extra‐cytoplasmic proteins.
ppiD null mutants exhibit altered OMPs profiles
The increased detergent sensitivity, high basal EσE activity and synthetic lethality observed with surA suggested that there must be a requirement for PpiD for the outer membrane biogenesis or maturation of OMPs. Hence, we analysed the total membrane profiles of wild‐type as compared with isogenic ppiD null mutant bacteria (CD247). The inner and outer membranes proteins were separated using sucrose gradients, as previously described (Missiakas et al., 1996). The protein profiles obtained were quite different (Figure 10). The amounts of all three major OMPs were reduced, i.e. OmpF, OmpC and OmpA. As a control, we also used a surA null mutant under similar experimental conditions. Although the amount of OMPs is reduced in both surA and ppiD mutant bacteria, there are some differences. For example, as shown in Figure 11, the OMP pattern is affected much more severely in a surA null mutant than in a ppiD null, although the overall amount of OMPs is significantly reduced in the ppiD null mutant.
We also examined the state of maturation of LamB (Figure 12, lanes 1–3). There was an overall reduction in the amounts of LamB in the ppiD mutant. As a control, we used a known periplasmic folding catalyst DsbC whose maturation is unaltered in a ppiD mutant (Figure 12).
Complementation of outer membrane defects in surA and htrM mutants by multicopy ppiD
We also analysed the effect of overproduction of PpiD in a surA mutant background. As shown in Figure 11, there is a near complete restoration of normal levels of OMPs upon overexpression of ppiD. This further supports our earlier finding that overproduction of PpiD compensates for the membrane leakiness of surA mutants and restores resistance to novobiocin and SDS.
Also, overexpression of ppiD leads to a decrease in EσE activity which is induced abnormally in surA mutants (Figure 9). The same is true for the ppiD‐induced EσE response: overexpression of surA restores the normal level of EσE activity in a ppiD null mutant. Again, these data suggest that both surA and ppiD perform similar functions in terms of protein folding. Nevertheless, they must have some specific substrates as well, which would account for the synthetic lethality observed.
We previously identified the surA gene as a multicopy suppressor of htrM (rfaD) mutant bacteria (Missiakas et al., 1996). htrM (rfaD) encodes an ADP‐l‐glycol‐d‐mannoheptose‐6‐epimerase (Raina and Georgopoulos, 1991) and participates in LPS biogenesis. LPS is needed for folding of OMPs (Schnaitman and Klena, 1993; Sen and Nikaido, 1993), and a htrM mutant which synthesizes an altered LPS exhibits extremely low amounts of OMPs and, as a consequence, has an altered outer membrane (Missiakas et al., 1996). Because of their leaky outer membrane, htrM mutants are unable to grow on MacConkey agar containing bile salts and are resistant to a variety of bacteriophages such as λ or P1. Overproducing SurA was found to restore both the growth on MacConkey agar, and the normal levels and folding of OMPs including LamB, which serves as a receptor for bacteriophage λ.
In the present study, we examined whether overexpression of ppiD could also complement htrM mutant bacteria. Interestingly, we found that multicopy ppiD (using ppiD under the control of an inducible promoter pCD275 or its own promoter pCD51) allowed the growth on MacConkey agar and restored the sensitivity to different bacteriophages such as λ and P1 (Tables I and VI). This complementation was best observed at 37°C or above. Our previous attempts at looking for multicopy suppressors for htrM mutants were all performed at 30°C using low copy number plasmids and, therefore, did not allow the identification of ppiD. We now know that ppiD is a heat‐shock gene and that its transcription is controlled by Eσ32.
PpiD has peptidyl–prolyl isomerase activity in vitro, and active site mutagenesis
Preliminary examination of the amino acid sequence of PpiD revealed the presence of a single parvulin (PpiC)‐like domain located between amino acid residues 227 and 357. This parvulin‐like domain is repeated twice in SurA. The presence of such a domain in PpiD could explain why ppiD could be isolated as a multicopy suppressor of surA. SurA has been shown to have a very weak PPIase activity in vitro (Missiakas et al., 1996; Rouvière and Gross, 1996). We examined the activity of purified PpiD using various Suc‐Ala‐Xaa‐Pro‐Phe‐p‐Na substrates. These experiments provided good evidence that PpiD can catalyse the rotation around the peptidyl–prolyl bond Ala–Pro of the tetrapeptide with greater efficiency than SurA. The estimated Kcat/Km value varied from 0.4 to 3.4/μM/s, which is quite high compared with the value estimated using purified SurA. The maximal activity was obtained for Glu, followed by Leu and Ala residues N‐terminal to the Pro residue (Table VII).
In order to identify some of the residues critical for PpiD activity, we isolated point mutations linked to hupB, based on their ability to induce the σE response (increased htrA–lacZ activity) as well as hypersensitivity to novobiocin (see Materials and methods). hupB is >95% linked to ppiD. Twenty three such mutants were obtained and sequenced. Two mutations were found to be changes of G312 (GGC to CGC) to R and G313 (GGC to CGC) to R. Two other mutations were changes of L298 to either S (TTA to TCA) or F (TTA to TTT). Another mutation was found to change I350 (ATT to TAT) to F. After completion of this work, the three‐dimensional structure of Pin1, a human PpiC homologue, was solved (Ranganathan et al., 1997). By comparison, it is possible to conclude that G312, G313, G347 and I350 are expected to be part of the active site of PpiD. Since a G to R substitution is rather drastic, we decided to change G312, G313, G347 and I350 to A using site‐directed mutagenesis. All these new mutations led to a loss of PpiD activity in terms of their ability to either complement a surA null mutant or reduce the htrA–lacZ activity induced by periplasmic stresses (Table I). Among these mutant PpiD proteins, we chose to purify the two mutants carrying the G347 to A and I350 to A substitutions, since they were more severely affected in their in vivo activity. These two mutant proteins appeared to lack any detectable PPIase activity (Table VII). This further confirms that these residues are part of the catalytic site, similarly to Pin1, although there is not a complete conservation of the amino acid sequence around this region.
In this work, we have identified a novel heat‐shock protein which belongs to the classical heat‐shock σ32 regulon. This protein is a catalyst which is required for the folding of exported proteins. This is the first σ32 heat‐shock protein which has been shown to participate in the folding of non‐cytoplasmic proteins. The gene encoding this protein was identified while looking for additional components of protein folding in the extra‐cytoplasm. Most of the known periplasmic folding catalysts so far have been found to be under the control of either the alternative sigma factor σE or the two‐component system CpxR–CpxA (Danese and Silhavy, 1997; Pogliano et al., 1997). Both regulons have evolved to respond to stresses occurring in the extra‐cytoplasm, and globally in the cell envelope, and we have used this knowledge as a tool to determine the presence of additional periplasmic folding catalysts.
The folding of OMPs has been shown to be dependent on LPS, the SurA and Skp proteins (reviewed by Missiakas and Raina, 1997a). Among the folding agents, SurA requirement has been found to be acute, since the maturation of many OMPs is affected in surA mutant bacteria. Consequently, a surA deletion leads to the induction of the periplasmic heat‐shock response, which is EσE dependent. In this work, we tried to determine whether other proteins are also required for OMP folding. To do so, we directly selected for genes which can bypass the requirement for the surA gene product, in vivo.
In parallel, we also looked for genes transcribed by the Cpx two‐component system. The CpxR–CpxA system is induced as a parallel pathway which monitors misfolding events occurring in the envelope. We took advantage of our recent finding that CpxR–CpxA‐dependent promoters such as htrA (a periplasmic protease) are constitutively induced upon overexpression of prpA in a strictly CpxR–CpxA‐dependent manner (Missiakas and Raina, 1997b). Hence, an E.coli promoter fusion library was transformed with a prpA plasmid and promoters strongly induced upon PrpA overproduction were selected. However, we retained only those promoters which were turned down in a cpxR null background. All these genetic approaches led to the identification of a new gene, ppiD. Based on the predicted amino acid sequence, ppiD encodes a membrane‐anchored PPIase with resemblance to members of the PpiC family, such as surA.
Why did these approaches identify only the ppiD gene?
First, our data clearly demonstrate that overexpression of ppiD can functionally compensate for lack of SurA. Secondly, we selected those promoters specifically induced upon overexpression of prpA. We had shown earlier that overexpression of prpA leads to the induction of the CpxR–CpxA regulon as well as the classical heat‐shock response regulon under the control of σ32. As a consequence, only a promoter region carrying both CpxR‐binding boxes and a σ32‐dependent promoter has been selected. In addition, none of the other known catalysts of folding can bypass the requirement for surA. For example, neither skp nor fkpA in multicopy restore normal OMP profiles in surA mutant bacteria (our unpublished data). Also, our earlier attempts to target putative periplasmic chaperones by searching for mutants which induce the periplasmic heat‐shock response (EσE‐dependent) were all performed at 30°C or below. A null mutation in ppiD induces only a marginal induction of the σE response at 30°C (2‐ to 3‐fold), as compared with 37°C (5‐fold), meaning that the misfolding defect of a ppiD null mutant can be screened far more easily at 37°C than at 30°C.
PpiD is a member of the PpiC family
It has been shown that members of the PpiC family (parvulin‐like) are also well conserved across evolution. First, the sequence examination showed that PpiD contains a parvulin‐like domain between amino acids 277 and 357. Consistent with this, we clearly demonstrated that PpiD has a PPIase activity in vitro. This activity is in fact much higher than the one measured for SurA, another member of the PpiC family. Residues G347 and I350 of PpiD were found to be essential for the PPIase activity. By sequence homology, these residues can be placed in the active site of the structure of PinI, the parvulin human homologue (Ranganathan et al., 1997). These are the first residues shown to be essential for PPIase activity, based on site‐directed mutagenesis. Thus, the PPIase activity seems to be essential for its in vivo role as a folding catalyst.
Synthetic lethality of the surA ppiD null combination
One of the most important findings of this study is that a double null combination of surA and ppiD genes is lethal under all the growth conditions tested. This is unique for surA and ppiD, since a ppiD null mutation can be combined with either ppiA (rotA), fkpA or even skp null mutations. Such a lethality can be explained by an overlap of function between surA and ppiD gene products and independent and different transcriptional regulation of each gene. This may also explain why neither of these genes is essential. The overlap in their functions is based on the following observations: (i) overexpression of ppiD can restore the normal membrane profile in surA mutant bacteria as well as suppress other membrane defects such as sensitivity to novobiocin; (ii) PpiD overproduction restores the elevated σE activity observed in surA mutant bacteria; (iii) overexpression of ppiD can restore membrane defects of htrM mutant bacteria which lack proper LPS; and (iv) phenotypes exhibited by ppiD mutants can be corrected by overexpressing the surA gene.
PpiD so far is the only folding catalyst for extra‐cytoplasmic function which is rpoH regulated
Until now, it seemed that protein folding in the cytoplasm of E.coli was assisted exclusively by the classical heat‐shock regulon placed under the control of rpoH (Eσ32), whereas genes responsible for folding in the extracytoplasm belonged to either the EσE regulon or the CpxR–CpxA regulon. Here we report a unique observation that a folding catalyst encoded by the ppiD gene and needed for the folding of OMPs is placed under the control of rpoH, the sigma factor transcribing cytosolic heat‐shock proteins. This is probably quite important for bacterial physiology. It is known that elevated temperatures not only lead to the aggregation of proteins in general, but also to changes in the composition of the outer membrane. At high temperatures, some OMPs are more abundant, e.g. OmpC (reviewed by Pratt et al., 1996), and the lipid composition also changes. This in turn affects the structure of LPS. This may be the reason why E.coli has chosen one of the important catalysts of folding needed for membrane biogenesis to be regulated by the classical heat‐shock sigma factor σ32. It should be pointed out that unlike rpoH, rpoE (encoding the alternate heat‐shock factor σE) transcription is not induced at high temperatures. rpoE transcription is governed by two different promoters which respond differently to temperature changes; however, the global amount of RNA synthesized remains constant under any conditions (Raina et al., 1995; Rouvière et al., 1995). Our unpublished data also show that surA transcription is not heat‐shock induced and is neither rpoE (σE) nor CpxR–CpxA dependent. Consistent with these observations, we found that the best suppression of both htrM and surA mutants by ppiD overexpression occurred at or above 37°C.
Materials and methods
Bacterial strains and plasmids
The bacterial strains and plasmids used in this study are listed in Table IV.
Selection strategies and cloning of the ppiD gene
The ppiD gene was cloned from a DNA library as a multicopy suppressor of a surA mutant using the in vivo mini‐Mu plasmid cloning system (Groisman and Casadaban, 1986). The mini‐Mu lysate was prepared on a MudII 4042‐carrying strain, into which a surA::Tn10 Tet mutation previously was introduced, to prevent the cloning of the surA+ gene during subsequent selection. Multicopy suppressors were selected based on their ability to allow surA mutant bacteria to form colonies at concentrations of novobiocin >30 μg/ml. It should be pointed out that surA mutant bacteria do not form colonies at >10 μg/ml of novobiocin due to defective membranes. Only those candidates were retained which also suppressed the sensitivity of the surA mutant to detergents such as SDS and sodium deoxycholate. The plasmid DNA prepared from 10 such mini‐Mu plasmids was subjected to partial Sau3A digestion to produce DNA fragments of ∼4 kbp in length. These DNA fragments were gel purified and ligated into the BamHI site of the p15A‐based vector pOK12 (Vieira and Messing, 1991). The ligation reaction was used to transform surA mutant bacteria plated on agar plates supplemented with 30 μg/ml novobiocin.
The restriction enzyme digestion pattern of seven such randomly chosen plasmids showed that they all carried the same chromosomal insert with a unique KpnI site. 32P‐labelled nick‐translated DNA from these plasmids was hybridized to the ordered E.coli genomic library (Kohara et al., 1987). All of them were found to map to the same region of the E.coli chromosome. One of these plasmids, pSR3239, was chosen for further subclonings and resulted in the construction of pCD50 which contains a 3 kbp EcoNI–SmaI DNA fragment. A further minimal subclone, pCD51, containing 2.4 kbp BsaBI–SmaI was constructed. Both plasmids were able to suppress surA mutant bacteria for resistance to novobiocin or SDS, to the same extent as the original plasmid pSR3239.
The overexpression of PpiD was achieved by first amplifying the ppiD gene by PCR of the minimal coding region, using primers 5′‐GTTGTTACCATATGGACAGC‐3′ and 5′‐TTGCGAGGCGAATTCAGGATTG‐3′, and the product was digested with NdeI and EcoRI, and cloned into the T7 promoter expression vectors pET24‐a (pCD52). To produce controlled expression, we also cloned the ppiD gene in a tightly regulated expression system under control of Ptrc, having a lacO operator as well as a lacIq repressor pSE420 (pCD275). This plasmid, pCD275, was also used in many complementation studies.
Screening for CpxR–CpxA‐dependent promoters
We were interested first in constructing a complete promoter library. To do this, chromosomal DNA isolated from E.coli wild‐type strain MC4100 was subjected to Sau3A partial digestion. DNA fragments in the range 400–800 bp were gel purified and cloned into the single copy F‐based promoter probe‐based vector pFZY (Koop et al., 1987) at its unique BamHI site. Lac+ clones were pooled en masse and transformed into a dsbD mutant background (Missiakas et al., 1995). We have shown previously that the induction of htrA transcription is fully optimized only when functional CpxR–CpxA and Prp proteins are present (Missiakas and Raina, 1997b). The htrA transcription is partly dependent on CpxR–CpxA. Hence, we pooled all Lac+ deeper blue AmpR colonies from the dsbD mutant background and transformed them further with a prpA+ plasmid wherein prpA expression is inducible with IPTG from the lac promoter. Approximately 30 000 colonies were plated on LB‐agar + glucose to repress prpA induction, and ∼500 colonies per plate were taken and replica plated on LB‐agar + IPTG to induce the expression of prpA. Approximately 70 transformants were selected which were deep blue, and were later transduced with the cpxR null allele (Danese et al., 1995). Fifteen such plasmids which exhibited less β‐galactosidase activity were selected for further analyses and sequenced. Twelve of them were found to have a promoter sequence region corresponding to ppiD.
We also constructed a new single copy promoter fusion to the promoter region of ppiD. The DNA of the promoter region of ppiD, as predicted from the above experiments, was PCR amplified using primers 5′‐GTTAACGAGCGTGGATCCCGTACT‐3′ and 5′‐GATACCGAATTCAATCTTGAG‐3′. This amplified DNA was digested with BamHI and EcoRI and cloned into the corresponding sites of the promoter probing vector pRS550 (Simons et al., 1987), resulting in plasmid pCD189. A single copy lysogen was made using λ RS45 (Simons et al., 1987) for further use. To compare our data with known heat‐shock‐dependent promoters or rpoE‐regulated promoters, we used our previously constructed groEL–lacZ, htpG–lacZ (SR1353), rpoEP2–lacZ (SR2195) and htrA–lacZ (SR1458) fusions (Raina et al., 1995; Missiakas et al., 1996). Also, the already described rpoH+‐, rpoE+‐ and prpA+‐carrying plasmids were used to verify further the transcriptional regulation of ppiD.
Disruption of the ppiD gene
In order to construct a null allele of ppiD, we inserted an ΩTet cassette (Fellay et al., 1987), previously digested at its BamHI restriction sites and made blunt, into the unique KpnI site (also made blunt using T4 DNA polymerase prior to ligation) of the ppiD coding region, as shown for plasmid pCD57 (Figure 1). Similarly, an ΩKan cassette from plasmid pUCK4 was inserted at the KpnI site as well as at the PstI site into plasmid pCD57. The resulting DNA fragments carrying the disrupted ppiD gene were digested appropriately from plasmids pCD160 and pCD174, respectively, and were introduced into the unique SmaI site of the vector pKO3 which carries a temperature‐sensitive replicon as well as the sacB gene for counter selection (Link et al., 1997). The resulting plasmids pCD180 and pCD178 were transformed into wild‐type strain MC4100 at 42°C to select for co‐integrate formation, and subsequently streaked at 30°C on LB‐Tet or LB‐Kan plates supplemented with 5% sucrose. Bacteria which had lost plasmid (loss of CmR) and were either TetR or KanR were retained.
RNA isolation and mapping of 5′ termini
Total cellular RNA was isolated by using the RNeasy kit from Qiagen. To prepare RNA produced at different temperatures, cultures were grown at 30°C and aliquots were shifted to 50°C for a period of either 5 or 10 min. Immediately after heat shock, cultures were lysed by guanidinium isothiocyanate, and RNA was purified as per the manufacturer's instructions. To define the transcriptional start site(s) of the ppiD gene, ∼10 ng of the oligonucleotide probe 5′‐CGTAAGCTGTTCCATCATGGTG‐3′, which is complementary to nucleotide positions −5 to 17 of the ppiD coding sequence, was annealed to 10 μg of total cellular RNA. The annealed primer was extended by avian myeloblastosis virus (AMV) reverse transcriptase. The primer extension products were electrophoresed on the same gel as the dideoxy sequencing reactions, using the same primer.
Determination of topology of PpiD
The membrane topology of PpiD was determined by the method developed recently by Manoil and Bailey (1997). We used both λ TnlacZiln and λ TnphoAlin to place these fusions on the plasmid pCD57. We obtained mostly XP+ colonies with λ TnphoAlin. The exact junctions of fusions were sequenced as suggested by Manoil and Bailey (1997). We also used the PhoA+ cassette described by Gutierrez and Devedjian (1989) and inserted it at the unique PstI site of the ppiD coding sequence. This site is located next to the parvulin‐coding domain facing the periplasm. Another PhoA fusion was constructed by first creating an AttII restriction site after the sixth amino acid predicted to be part of domain residing in the cytoplasm and then inserted the PhoA+ cassette.
Since ppiD is >90% linked to hupB, we used this as a linked marker to obtain point mutations. Briefly, a P1 bacteriophage lysate was grown on the hupB::Kan strain and the lysate was mutagenized with hydroxylamine as described earlier (Raina et al., 1995). This mutagenized lysate was transduced into a htrA–lacZ fusion, and Lac colonies were picked. These transductants were tested further for sensitivity to 20 μg/ml of novobiocin and 2% SDS. Twenty three independent isolates were retained, as they exhibited a phenotype close to that of ppiD null mutants. In the present study, we were more interested in the determination of active site residues of the putative parvulin domain. We PCR amplified the minimal parvulin domain from the chromosomal DNA prepared from these mutants. These PCR products were sequenced directly using standard procedures.
Based on the above experiments and results obtained therein, we also carried out site‐directed mutagenesis by changing G312, G313, G316, G347 and I350 to A residues. This was achieved by oligonucleotide‐directed mutagenesis, following the procedure described in the Quick Change kit from Stratagene, using Pfu DNA polymerase. For PCR amplification, 50 ng of template pCD51 DNA was used. The resulting plasmids carrying the mutated ppiD gene were verified by sequencing and by their inability to complement ppiD null mutants. Out of these, pCD271 A347 and pCD273 A350 mutant plasmids were used for purification of mutant proteins.
PpiD purification and peptidyl–prolyl isomerase activity
Escherichia coli carrying plasmid pCD51 (ppiD+) were induced with 1 mM IPTG, at an OD of 0.4 at 595 nm for 4 h. Cells were resuspended in buffer A (50 mM Tris–HCI pH 7.2, EDTA 1 mM, KCl 50 mM, glycerol 10% v/v) and lysed by sonication. The lysate was centrifuged at 35 000 r.p.m. for 45 min at 4°C. Protein recovered in the membrane fraction was resuspended in buffer A containing 0.2% sarkosyl to solubilize the inner membrane and then spun again. The soluble inner membrane fraction was dialysed against buffer A containing the non‐polar detergent NP‐40 (0.2%) and loaded onto a gel filtration column (G75). The PpiD‐containing fractions were loaded on a MonoQ column (Pharmacia). The protein of >95% purity eluted with a KCl gradient of 0.05–1 M.
The PPIase activity was determined by a protease‐coupled assay using chymotrypsin and succinyl‐Ala‐Xaa‐Pro‐Phe‐4‐nitroanilide as the substrates (Fischer et al., 1992). The reaction mixture contained chymotrypsin in HEPES buffer 35 mM, pH 7.8 and the reaction was started by adding sequentially the substrate dissolved in 460 mM LiCl in trifluoroethanol and one of the purified proteins. A pre‐incubation of chymotrypsin and PpiD was avoided to minimize protein degradation. The time course of the reaction was monitored at 390 nm on a Uvikon 940 spectrophotometer. The relative specific constants kcat/Km were calculated as described by Rahfeld et al. (1994) and Fischer et al. (1992).
We thank Dr Dominique Missiakas for many helpful suggestions and Dr C.Georgopoulos for constant encouragement. We are grateful to Dr Meno Kok for the kind gift of LamB antibodies and to Dr U.Hennig for providing the skp mutant. This work was supported by a grant from the Fond National Scientifique Suisse (FN31‐42429‐94) to S.R.
- Copyright © 1998 European Molecular Biology Organization