DjlA Is a Third DnaK Co-chaperone of Escherichia coli, and DjlA-mediated Induction of Colanic Acid Capsule Requires DjlA-DnaK Interaction*

DjlA is a 30-kDa type III membrane protein of Escherichia coli with the majority, including an extreme C-terminal putative J-domain, oriented toward the cytoplasm. No other regions of sequence similarity aside from the J-domain exist between DjlA and the known DnaK (Hsp70) co-chaperones DnaJ (Hsp40) and CbpA. In this study, we explored whether and to what extent DjlA possesses DnaK co-chaperone activity and under what conditions a DjlA-DnaK interaction could be important to the cell. We found that the DjlA J-domain can substitute fully for the J-domain of DnaJ using various in vivo functional complementation assays. In addition, the purified cytoplasmic fragment of DjlA was shown to be capable of stimulating DnaK ATPase in a manner indistinguishable from DnaJ, and, furthermore, DjlA could act as a DnaK co-chaperone in the reactivation of chemically denatured luciferase in vitro. DjlA expression in the cell is tightly controlled, and even its mild overexpression leads to induction of mucoid capsule. Previous analysis showed that DjlA-mediated induction of the wca capsule operon required the RcsC/RcsB two-component signaling system and that wcainduction by DjlA was lost when cells contained mutations in either thednaK or grpE gene. We now show using allele-specific genetic suppression analysis that DjlA must interact with DnaK for DjlA-mediated stimulation of capsule synthesis. Collectively, these results demonstrate that DjlA is a co-chaperone for DnaK and that this chaperone—co-chaperone pair is implicated directly, or indirectly, in the regulation of colanic acid capsule.

The Hsp70 family of molecular chaperones, of which DnaK is the major member in Escherichia coli, is regulated by interaction with co-chaperones that function together as a chaperone machine. In E. coli, DnaJ has long been recognized as a partner protein and key regulator of DnaK ATPase activity, together with the nucleotide exchange factor, GrpE (1). The general mechanism of the Hsp70 chaperones, including DnaK, is that binding and release of protein substrates is tightly coupled to their ATPase cycle (2,3).
The Hsp40 family of molecular chaperones is defined by a short, ϳ70-amino acid residue signature sequence, termed the J-domain, which helps direct interaction with partner Hsp70 chaperones (4 -6). Indeed, the J-domain of DnaJ is absolutely essential for its interaction with DnaK and is specifically required for stimulation of the ATPase activity (1,(7)(8)(9). A highly conserved HPD tripeptide, located in an exposed loop of the J-domain, is critical for co-chaperone function because mutations in these residues severely compromise the stimulation of ATP hydrolysis, not only in E. coli, but in many other organisms (5,10,11).
The E. coli genome codes for two Hsp40 proteins, DnaJ and CbpA, and two putative J-domain proteins, DjlA and Hsc20. DnaJ and its close orthologs are characterized by four domains: an N-terminal J-domain, a region rich in glycine and phenylalanine, a zinc finger domain, and a C-terminal domain thought to be involved in substrate binding (4,5). CbpA is 39% identical to DnaJ at its entire length and 55% at its J-domain (12). CbpA is a multicopy suppressor of dnaJ mutations and is thus a functional ortholog of DnaJ (13). A cbpA null mutation shows a synthetic phenotype with a dnaJ null mutation, the doubly mutant strain becoming hypersensitive for growth above 37°C and below 16°C (14). Hsc20 possesses only 20% sequence similarity to the DnaJ J-domain and is unable to interact productively with DnaK (15,16). By sequence alignment methods, DjlA possesses a J-domain signature sequence at its extreme C terminus with 31% sequence identity (17) but otherwise possesses no sequence similarity to other regions of DnaJ and CbpA. Many Hsp70 co-chaperones have only the J-domain signature in their sequence (5). Thus, DjlA could represent a third DnaK co-chaperone in E. coli, but under what circumstances is unclear. A clue to a potential role for DjlA in the cell emerged in previous studies when we and others (17)(18)(19)(20) showed that DjlA overexpression could trigger the synthesis of a colanic acid polysaccharide capsule. In E. coli, a colanic acid capsule can be induced by environmental stress such as osmotic shock, low temperature, and desiccation (21)(22)(23). Because a colanic acid capsule is not normally observed at 37°C, or under physiological conditions prevailing within the host intestinal tract, it has been suggested that this adaptive response may help cells to survive conditions outside the host (24). Interestingly, no link has been shown between colanic acid capsule production and pathogenicity (25,26), although a recent study suggests a role for colanic acid in the architecture of biofilms (27).
The major structural genes for colanic acid synthesis in E. coli K-12 are found in the wcaABCDE locus, formerly called cps (28 -30). Detailed analysis of colanic acid induction has demonstrated that the wza-wca operon (hereby referred to as wca) is regulated by the RcsC/RcsB two-component histidine kinase signaling system (24, 31,32). Induction of wca by DjlA is dependent on both RcsC and RcsB, suggesting that DjlA modulated the activity of this phosphorelay sensor.
Previous genetic analyses of parameters affecting capsule induction revealed that activation of a wcaB-lacZ reporter by DjlA was abrogated in strains containing mutations in genes encoding the molecular chaperones DnaK and GrpE but was not adversely affected by null mutations in genes encoding DnaJ, CbpA, HtpG (Hsp90), or Hsc66 (Hsp70 homolog) (17,18). In the present work, we have used both biochemical and genetic analyses to address whether DjlA is a bona fide DnaK co-chaperone and whether direct DjlA-DnaK interaction is necessary for wca activation.
Plasmid Constructions-Plasmid pWCS19 (37), a pTrcHisA vector (Invitrogen) contains the gene encoding for the DnaK(R167H) mutant (a kind gift from Dr. Carol Gross and Dr. Won-Chul Suh, University of California at San Francisco). The wild type dnaK ϩ version of pWCS19, pKG9, was prepared by the Quickchange TM method (Stratagene) using the primers 5Ј-CTGGAAGTAAAACGTATCATCAACGAAC-3Ј and 5Ј-GTTCGTTGATGATACGTTTTACTTCCAG-3Ј. The dnaK ϩ wild type sequence was verified and tested for functional complementation of the dnaK103 mutant strain CG800, which is otherwise unable to form colonies at 42°C and cannot support growth of bacteriophage at any temperature. The wild type and mutant dnaK genes were re-engineered for DnaK expression using a p15A-derived replicon that is compatible in cells harboring the ColE1-derived pBAD vectors (38). Briefly, pA-CYC184 (39) was digested with HincII, and the 3,178-base pair fragment was purified and self-ligated in the presence of NsiI linkers (5Ј-CATGCATG-3Ј). The resulting plasmid, pWKG59, was then digested with HindIII and NsiI, and the 2,556-base pair ori p15A-cam r fragment, containing the ori of replication, was ligated with the 3,509base pair HindIII-NsiI fragment containing lacI q , pTrc promoter, and downstream dnaK sequence from pWCS19, or pKG9, to yield plasmid pKG7 (dnaKR167H) or pKG8 (dnaK ϩ ), respectively.
Polymerase Chain Reaction Construction of J-domain Chimeras-Plasmid pWKG50 (17) expression phagemid was mobilized with VCSM13 (Stratagene) in strain CJ236 (40) and mutagenized to introduce the D235N mutation using the method of Kunkel (40) and the oligonucleotide DjlA D235N: 5Ј-CGCCACCAGCTTATTGGGATGGT-GTTC-3Ј. The resulting plasmid, pWKG55, was sequence verified. Plasmids pWKG90 and pWKG100 (33) encode E. coli dnaJ and dnaJ12, respectively, with a phenotypically silent H71T mutation that was engineered to introduce a KpnI site at the carboxyl junction of the J-domain. Plasmids pWKG90 and pWKG100 were digested with EcoRI-KpnI to remove the E. coli J-domain and then exchanged with a 220base pair polymerase chain reaction fragment derived from the plasmid templates pWKG50 (djlA), pWKG54 (djlAH233Q), or pWKG55 (djlAD235N) (17) that had been digested with the same enzymes. Primer sets 5Ј-GGGAATTCACCATGGAAGATGCCTGTAATG-3Ј, and 5Ј-CCGGTACCTTTAAACCCTTTCTGCTGCTT-3Ј were used to amplify the djlA J-domains and introduced the appropriate EcoRI and KpnI restriction sites. All constructs were sequence verified using appropriate primers. Plasmids pKG1, pKG2, and pKG3 contain djlA, djlA(H233Q), and djlA(D235N) J-domains, respectively, in the pWKG90 vector background. Plasmids pWKG4, pWKG5, and pWKG6 contain the same respective djlA J-domains in the pWKG100 vector background.
Bacteriophage and Colony Forming Assays-Bacteriophage laboratory stocks b2cI Ϫ , a clear plaque former, and dnaJϩ transducing phage were prepared from liquid lysates following infection of the host strain MC4100 using Luria-Bertani (LB) broth supplemented with 50 mM MgSO 4 . 5-ml lysates were treated with chloroform to ensure cell lysis and debris removed by low speed centrifugation. Bacteriophage were serially diluted in SM buffer (10 mM Tris-HCl, pH 7.5, 10 mM MgCl 2 , 50 mM NaCl). Plaque forming assays and spot test viability assays were performed essentially as described previously (33).
Protein Purifications-E. coli DnaJ protein was prepared from lysates of strain WGK190 containing pWKG90, a L-arabinose inducible promoter vector driving expression of wild type DnaJ (33) essentially as described (41). E. coli GrpE was prepared from strain DA262 containing plasmid pOD1, a L-arabinose inducible promoter vector driving expression of wild type GrpE, and using methods essentially as described (42). E. coli DnaK was purified from strain B178 (43) containing dnaK ϩ in pTTQ19 (Amersham Pharmacia Biotech) using the described procedures (44).
DjlA Purification-E. coli DjlA soluble cytoplasmic fragment was purified from strain WKG190 harboring the vector pWKG52 (17), in which the DNA segment encoding the N-terminal 31-amino acid transmembrane-spanning region had been removed. The use of a host strain devoid of both DnaJ and CbpA assured that no other J-domain-containing protein will contaminate the DjlA preparation. Full-length DjlA encoded by the chromosomal copy of djlA is expressed poorly in this genetic background, in addition to being membrane-associated, and thus not detectable by immunoblot analysis in the soluble fraction (17). For purification of DjlA cytoplasmic fragment, a fresh overnight culture of pWKG52 in WKG190 was diluted 1:100 into 12 liters of LB broth supplemented with 50 g/ml ampicillin and grown at 30°C with vigorous shaking. At an A 600 ϭ 1.0, L-arabinose inducer was added to a final concentration of 0.1% (w/v) and shaking continued for an additional 3 h at 30°C. Cells were harvested at 10,000 rpm for 10 min in a Sorvall GSA rotor and resuspended in a minimal volume of 50 mM Tris-HCl, pH 7.6, 10% (w/v) sucrose. All subsequent steps were performed at 4°C. Cells were lysed with 250 ml of buffer A (50 mM Tris-HCl, pH 7.2, 1 M NaCl, 2 mM MgCl 2 , 2 mM DTT, 1 0.4 mg/ml lysozyme) for 60 min on ice.
To aid lysis and to reduce viscosity, the cell suspension was briefly heat shocked in a 45°C immersion bath for 3 min then sonicated with 20 2-s pulses. The lysate was then centrifuged at 25,000 rpm for 1 h 45 min in a Beckman 35 Ti rotor. Ammonium sulfate (0.226 g/ml) was added slowly with stirring to the cleared supernatant, and the precipitate was collected after centrifugation at 25,000 rpm for 1 h in a 35 Ti rotor. The pellet was resuspended and dialyzed against 2 liters of buffer B (50 mM HEPES, pH 7.6, 150 mM KCl, 2 mM DTT, 0.05% Triton X-100, 10% (v/v) glycerol). Insoluble material was removed by centrifugation at 20,000 rpm for 30 min in a 35 Ti rotor. The supernatant was then applied to a fast flow Q Sepharose column (2.5 ϫ 20 cm) that had been pre-equilibrated with buffer B. The void volume was collected, and DjlA fractions were pooled and dialyzed against 2 liters of buffer C (50 mM potassium phosphate, pH 6.8, 150 mM KCl, 2 mM DTT, 0.05% Triton X-100, 10% (v/v) glycerol). The dialysate was applied to a P-11 phosphocellulose column (2.5 ϫ 20 cm) that had been pre-equilibrated in buffer C. The column was washed with 5 column volumes of buffer C, then developed with a 0.15-0.6 M KCl linear gradient in buffer C (800 ml). The 27-kDa DjlA cytoplasmic fragment eluted first, whereas the truncated 25-kDa fragment was well separated, eluting at higher KCl. The two peak fractions were pooled separately, dialyzed against buffer D (50 mM Tris-HCl, pH 7.6, 150 mM KCl, 2 mM DTT, 0.05% Triton X-100, 10% (v/v) glycerol), and concentrated separately by application on an HTP hydroxyapatite (Bio-Rad) column (1.5 ϫ 15 cm) pre-equilibrated in buffer D. Bound proteins were eluted with a 0 -300 mM linear potassium phosphate, pH 7.2, gradient. Peak fractions were pooled, dialyzed against storage buffer E (50 mM HEPES, pH 7.6, 150 mM KCl, 2 mM DTT, 10% (v/v) glycerol), then quickly frozen in liquid nitrogen and stored at Ϫ80°C until use.
Protein Sequencing-For N-terminal protein sequencing, sample aliquots were resolved in 12% (w/v) SDS-polyacrylamide gels, electroblotted to polyvinylidene difluoride membranes (Bio-Rad), and bands were visualized by staining with 0.1% Amido Black in 25% (v/v) 2-propanol, 10% (v/v) acetic acid. Bands were excised, rinsed in sterile water, dried, and applied directly to an Applied Biosystems model 473 Sequencer equipped with a blot cartridge.
Protein Concentration Determination-Protein concentrations were determined using published or derived molar extinction coefficients (45,46) or by Bio-Rad dye reagent assay using bovine serum albumin as a standard.
Steady-state ATPase Assay-The assay was performed essentially as described previously (1). The standard reaction conditions were 50 mM HEPES, pH 7.6, 40 mM KCl, 50 mM NaCl, 7 mM magnesium acetate, 2 mM DTT, 300 M disodium ATP, 1 M DnaK, 1 M GrpE, and DnaJ or DjlA titrated over the range 0 -1 M. [␥-32 P]ATP, specific activity of 3,000 Ci/mmol (Amersham Pharmacia Biotech) was used as tracer in the reaction mixture. Aliquots were removed from the reaction at the specified times, spotted on polyethyleneimine thin layer chromatography plates (Merck), and resolved with a 1:1 solvent mixture 1 M formic acid, 1 M LiCl. Plates were dried and autoradiographed, and the region corresponding to spots was excised and counted in a Beckman liquid scintillation counter. Pilot assays were used to determine ATP concentrations necessary for the concentration of DnaK used in the standard assay. For determination of kinetics, linear regression analysis was applied and data points accepted only when correlation coefficients exceeded 0.98. In the analysis presented, all data were obtained in the linear range of the assay, and each data point represents the average rate derived from regression lines compiled from five independent experiments. Data points shown at the zero time point should be considered as subject to a relatively large experimental error because of the sample preparation time.
Luciferase Refolding Assay-Reactivation of denatured firefly luciferase was performed essentially as described (47)  The resulting luciferase activity was measured at different time points after incubation at 22°C by withdrawing aliquots and using the Promega luciferase assay kit (E1500) followed by liquid scintillation counting.
Luciferase Aggregation Assay-The kinetics of luciferase aggregation were followed by measuring light scattering at 320 nm with a Uvikon 940 spectrophotometer using a thermostated cell holder essentially as described (48).
Immunoblot Analysis-Anti-DjlA antibodies, the kind gift of Dr. David Clarke and Dr. I. B. Holland (Institut de Génétique et Microbiologie, Université de Paris Sud), were used at a dilution of 1:20,000. Goat anti-rabbit horseradish peroxidase-conjugated IgG secondary antibodies were used at a dilution of 1:10,000. Blots were developed with enhanced chemiluminescent reagents according to the manufacturer's recommendations (Amersham Pharmacia Biotech).

RESULTS
The DjlA J-domain Can Functionally Replace the J-domain of DnaJ in Vivo-Inspection of DjlA J-domain sequence alignment of residues 208 -271 (the native C terminus) revealed 31% identity and 44% similarity to the DnaJ J-domain, including the highly conserved HPD tripeptide (17, Fig. 1). To determine whether the J-domain of DjlA was capable of functional interaction with DnaK, we first constructed plasmids coding for chimeric DnaJ proteins. Conditional expression of the DnaJ chimeras in a dnaJ cbpA double null strain permits measurement of J-domain activity (33).
The djlA region encoding the putative DjlA J-domain corresponding to amino acids 206 -271 (the native C terminus) was amplified by polymerase chain reaction, and chimeras were constructed as described under "Experimental Procedures." The plasmids were transformed into strain WKG190 at the permissive temperature of 30°C, then tested in a colony forming assay. The results are shown in Fig. 2. We observed that the chimeric plasmid pKG1 complemented for bacterial growth at 14°C or 40°C as efficiently as wild type DnaJ in the presence, but not in the absence, of the L-arabinose inducer. In contrast, the vector alone did not complement for bacterial growth nor did either of the chimeric plasmids encoding for mutant DjlA J-domains, pKG2 and pKG3. Similar results were observed using plasmids pKG4, pKG5, and pKG6, which encode the truncated DnaJ12 derivatives (data not shown).
The failure of the DjlA H233Q and D235N mutant J-domain chimeras to complement was not caused by altered steady-state protein expression levels or stability, as judged by immunoblot analysis or inspection of Coomassie Blue-stained SDS gels. We observed that the levels of the DjlA J-domain replacement chimeras were about 10-fold lower than DnaJ but were of comparable level when compared with each other (compare the  second and fourth lanes from the left, Fig. 3). We conclude that the mutations in DjlA altering the highly conserved HPD tripeptide were disrupting the otherwise productive J-domain interaction with DnaK. Restoration of Bacteriophage Plaque Formation by J-domain Chimeras-As a second, independent assay of J-domain function, we exploited the dependence of bacteriophage upon the host bacterial chaperones DnaK, GrpE, and DnaJ for plaque formation. Table I shows the results of a bacteriophage plaque forming assay performed on strain WKG190, harboring the indicated dnaJ or dnaJ12 expression plasmids. The dnaJ ϩ transducing bacteriophage served as a control because it was expected to form plaques with or without L-arabinose, whereas b2cI Ϫ was expected to grow only on plates where both the L-arabinose inducer and a complementing plasmid encoding dnaJ or dnaJ12 were present. We observed that pKG1 and pKG4 could complement as efficiently as pWKG90 or pWKG100 as a source of DnaJ for bacteriophage plaque formation. Plasmids pKG2 and pKG5 could not complement for bacteriophage plaque formation in this assay, consistent with the observation that the well characterized analogous dnaJ259 mutation (encoding for the corresponding H33Q change) cannot support bacteriophage growth either when in single chromosomal copy or when expressed from an L-arabinose-inducible vector (7,33). In contrast, pKG3 but not pKG6 allowed the formation of small turbid plaques in the presence of the Larabinose inducer, indicating partial complementation for bacteriophage growth. It should be noted that the b2cI Ϫ bacteriophage employed in this study forms very clear plaques on the wild type host, thus the appearance of very small turbid plaques is most likely caused by extremely low bacteriophage progeny yields. Control experiments using a pBAD expression plasmid carrying dnaJ encoding the equivalent mutation (D35N) did not support the growth of b2cI Ϫ bacteriophage under any conditions tested (data not shown). Taken collectively, these results show that DjlA harbors a bona fide Jdomain that is capable of replacing the J-domain of DnaJ in two independent in vivo assays of DnaJ function.
Purified DjlA Cytoplasmic Fragment Can Stimulate DnaK ATPase in Vitro-A well studied consequence of DnaJ-DnaK interaction in vitro is the ability of DnaJ to stimulate the intrinsically weak DnaK ATP hydrolysis rate. Because our in vivo assays indicated that the DjlA J-domain could functionally replace the DnaJ J-domain and therefore was likely to interact directly with DnaK, we examined whether the purified cytoplasmic fragment of DjlA (32-271) might by itself engage DnaK and stimulate its ATPase, as predicted if indeed DjlA could function as a DnaK co-chaperone.
Plasmid pWKG52 encodes DjlA lacking the transmembranespanning region of wild type DjlA. It was used to overexpress and purify the cytoplasmic fragment of DjlA from a strain lacking both dnaJ and cbpA to ensure the absence of contaminating DnaJ activity. The steps of purification are depicted in Fig. 4. Upon cell lysis, approximately half of the DjlA is cleaved rapidly, resulting in the formation of a doublet suggesting the removal of a short stretch of amino acids. The proteolytic activity has been suggested previously to be an artifact of the preparation method (19). Using a variety of lysis methods, including freeze-thaw and detergent lysis as well as a panel of protease inhibitors, we also noted this cleavage except when whole cell lysates were prepared by direct solubilization in reducing SDS sample buffer. Both DjlA products are immunologically reactive with anti-DjlA antiserum and copurifiy during the early chromatographic steps. However, the full-length cytoplasmic fragment can be easily separated from the smaller species, using gradient elution by phosphocellulose P-11 chromatography.
N-terminal protein sequence analysis of the purified smaller proteolytic cleavage fragment revealed that cleavage occurs between residues Arg-39 and Lys-40 of the full-length protein, thus resulting in the removal of 8 residues from the cytoplasmic fragment of DjlA. The full-length cytoplasmic fragment was used for all subsequent studies.
Highly purified DnaK, DnaJ, GrpE, and DjlA⌬TM were prepared and used to analyze steady-state ATP hydrolysis rates in vitro. The results are shown in Fig. 5. We found that DjlA⌬TM could stimulate DnaK ATPase activity in a manner indistinguishable from DnaJ under the range of protein concentrations tested.
Purified DjlA Cytoplasmic Fragment Possesses Co-chaperone Activity-As an additional test of DnaJ co-chaperone activity, we asked whether the purified DjlA cytoplasmic fragment could act as a co-chaperone, together with DnaK and GrpE in the in vitro reactivation of chemically denatured firefly luciferase. Because this assay has been shown to be critically dependent upon DnaJ (47)(48)(49)(50) to promote the efficient refolding of denatured luciferase, we expected that if DjlA possessed the requisite co-chaperone activity, it could fully or partially replace DnaJ in this assay. The results of a representative kinetic analysis of luciferase reactivation are presented in Fig. 6A.
We observed that DjlA⌬TM could fully replace DnaJ in the refolding of guanidinium hydrochloride-denatured luciferase and that the rate and degree of reactivation were comparable between DnaJ and DjlA, using our experimental conditions. Control reactivation of luciferase in the absence of added DnaJ, or DjlA⌬TM co-chaperone revealed no significant reactivation of luciferase and could not be distinguished from the spontaneous refolding rate observed in the absence of added chaperones or DnaK alone. We conclude that the cytoplasmically oriented fragment of DjlA can indeed function as a DnaK co-chaperone.
Purifed DjlA Cytoplasmic Fragment Cannot Act as a Chaperone Alone but Can Help DnaK to Prevent Luciferase Aggregation-DnaJ has been described as a bona fide chaperone because it can bind alone to various denatured substrates and help protect them from aggregation (47)(48)(49)(50). To test whether the purified DjlA cytoplasmic fragment alone behaved as a bona fide chaperone, we chose to study the kinetics of aggregation of denatured luciferase. When DnaJ alone is used in this type of assay, it has been shown to protect luciferase partially from aggregation, whereas DnaK alone is unable to prevent aggregation. However, DnaK and DnaJ together, in the presence of ATP, can efficiently protect luciferase from aggregation, provided that the chaperone proteins are supplied in near stoichiometric amounts. The results of such an experiment are depicted in Fig. 6B.
Our results showed that either DjlA or DnaJ can cooperate with DnaK in the presence of ATP to prevent luciferase aggregation. However, although DnaJ alone could partially protect luciferase from aggregation as expected, we did not observe any such activity for the DjlA⌬TM fragment. Consistent with this result, we were unable to detect binding of DjlA⌬TM to denatured luciferase using a sensitive enzyme-linked immunosor- bent assay (data not shown). DnaK alone had no activity in this assay, thus reinforcing the interpretation that the two pairs of proteins, DnaJ-DnaK, or DjlA⌬TM-DnaK, must collaborate to prevent luciferase aggregation fully. These results suggest that DjlA⌬TM does not have intrinsic chaperone activity but can act as a co-chaperone for DnaK. In a broader context, this result is consistent with the observation that DjlA⌬TM cannot support the replication of bacteriophage either in vivo (17) or in vitro, 2 nor can DjlA or DjlA⌬TM fully substitute for DnaJ for bacterial growth at high or low temperature (17,18,20). Thus, DjlA is clearly distinct 2 A. Wawrzynow, unpublished data. DjlA (D235N) 0.6 -1.0 a Plaque-forming units (pfu) are reported as the mean of at least three independent determinations. Plating efficiencies were normalized to the titer obtained with pWKG90 containing the wild type dnaJ ϩ gene.
b The sign (-) indicates no detectable complementation (pfu Ͻ 10 Ϫ5 ), and (*) indicates tiny turbid plaques. The LB-agar plates were incubated at 30°C overnight. from DnaJ and likely plays a more specific and restricted role in the cell.

Allele-specific Suppression Analysis Shows that DjlA-DnaK Interaction Is Necessary for wca Transcriptional Activation-
The E. coli dnaJ236 allele, encoding DnaJ with a point mutation (D35N) in the highly conserved HPD tripeptide of the J-domain, is temperature-sensitive for bacterial growth above 42°C and cannot support replication of bacteriophage at any temperature (39). An allele-specific suppressor of dnaJ236(D35N), but not dnaJ259(H33Q), was isolated in the dnaK gene, dnaK(R167H), changing a residue in a solvent exposed cleft in the ATPase domain (39,51). The dominant dnaK(R167H) suppressor carried on a multicopy plasmid restored productive interaction with DnaJ(D35N) and permitted restoration of bacterial growth above 42°C as well as bacteriophage growth.
We exploited this genetic analysis to study DjlA-DnaK interaction. Specifically, we reasoned that if direct interaction between DjlA and DnaK were necessary for wca transcriptional activation through the RcsC/RcsB two-component system, then engineered point mutations in the DjlA J-domain should abolish wca activation. Importantly, if analogous allele-specific suppression were possible with another J-domain-DnaK pair from the same organism, then simultaneous coexpression of defective DjlA(D235N) with DnaK(R167H) should restore wca-lacZ reporter activity. The allele-specific interaction would also predict that simultaneous coexpression of DjlA(H233Q) with DnaK(R167H) would not restore wca-lacZ reporter activity.
Plasmids encoding wild type DnaK(pKG8) or DnaK (R167H)(pKG7) under the control of the pTrc promoter were constructed. Combinations of DjlA and DnaK expression plasmids, or parental vectors pBAD22 and pWKG59 alone, were transformed into the SG20781 wcaB10-lacZ reporter strain and the activation of the wca operon analyzed. The results are depicted in Fig. 7. The overexpression of DjlA upon addition of the L-arabinose inducer resulted in a strong induction of the wca operon, as judged by the increased levels in ␤-galactosidase activity. Control experiments showed that cells harboring the pKG8dnaK ϩ or pKG7dnaK(R167H) expression plasmids alone did not show any significant wca operon activation.
When plasmids pWKG54 and pWKG55, containing the mutation H233Q, or D235N alone were expressed under conditions identical to wild type DjlA, no wca induction was observed, indicating that mutations within the djlA J-domain HPD tripeptide had abolished DjlA's effect on capsule synthesis. Furthermore, control experiments revealed that steadystate protein levels of wild type DjlA and the two mutant derivatives were indistinguishable under the conditions of the assay (data not shown).
A significant activation of wca-lacZ reporter was observed, however, when plasmid pWKG55 was coexpressed with pKG7 dnaK(R167H), but not with pKG8 dnaK ϩ . In contrast, no significant activation of wca-lacZ was observed when pWKG54 was coexpressed with pKG7 or pKG8, indicating that the observed wca-lacZ activation was indeed the consequence of allele-specific genetic interaction between DjlA(D235N) and DnaK(R167H). We conclude that DjlA and DnaK must interact directly to elicit wca activation after DjlA overexpression. DISCUSSION The major finding of this work is the demonstration that DjlA represents a third regulatory DnaK co-chaperone in E. coli. Several lines of evidence presented here clearly demonstrate that (a) the J-domain of DjlA is indeed functional, as judged by its ability to replace the J-domain of DnaJ in several in vivo assays; (b) purified DjlA cytoplasmic fragment can productively interact with DnaK in vitro, thus behaving as a bona fide co-chaperone for DnaK; (c) DjlA does not apparently possess intrinsic chaperone activity alone as judged by its inability to protect luciferase from aggregation; and (d) direct DjlA-DnaK interaction is necessary for wca activation as revealed by genetic allele-specific suppressor analysis.
Our assay for J-domain function in E. coli, using chimeric proteins, has revealed a surprisingly broad tolerance for Jdomains from a wide range of sources with limited sequence homology. In contrast to our findings, analogous experimental use of J-domain chimeras in Saccharomyces cerevisiae and SV40 virus shows that these systems have a much more strict tolerance for interchanging J-domains (52)(53)(54). At the molecular level, specificity determinants for Hsp70-Hsp 40 interaction are largely unknown, but they are thought to be mediated largely by the J-domain.
Some features of the DjlA J-domain provide some new insights into J-domain-DnaK interaction. The DjlA J-domain possesses a longer loop between helices II and III as well as a truncation of the residues corresponding to the majority of the region helix IV, including most of the QKRAA motif (see Fig. 1). It is noteworthy that the loop region between HPD and the start of helix III shows considerable sequence and length heterogeneity among J-domains that function in E. coli and thus is unlikely to be by itself a crucial specificity determinant for interaction with DnaK. The QKRAA motif, comprising the helix IV region of DnaJ J-domain, has been described as a possible site of interaction with DnaK (57). Because the DjlA Jdomain lacks this motif, yet can functionally replace the DnaJ J-domain in vivo and in vitro, helix IV cannot be essential for directing interaction with DnaK. In agreement with this conclusion, a DnaJ derivative entirely lacking helix IV can fully substitute for DnaJ in vivo. 3 Of course, these findings do not exclude the possibility that helix IV plays some as yet undefined modulatory role in the biology of DnaJ in vivo.
Despite possessing a functional J-domain, DjlA is clearly not a functional ortholog of DnaJ and CbpA. Previous work showed that DjlA could not adequately complement DnaJ function in a strain lacking both dnaJ and cbpA, even when expressed to the same levels of DnaJ which could functionally complement in this test system (17,20). Additionally, overexpression of the DjlA cytoplasmic fragment could not support bacterial growth of the dnaJ cbpA double null strain above 39°C nor at temper-3 W. L. Kelley, unpublished data. atures below 20°C, where this strain also exhibits a coldsensitive phenotype (17,20). Furthermore, DjlA cannot replace either DnaJ or CbpA in supporting bacteriophage growth under any conditions tested.
The precise cellular role of DjlA has not been determined yet. A clue to a possible function emerged when it was shown that overexpression of DjlA could lead to the induction of a colanic acid capsule (17)(18)(19). Our observation that DjlA⌬TM, the DjlA 27-kDa cytoplasmic fragment, can function as a co-chaperone, in a manner identical to that of DnaJ, now strongly suggests that once produced in the cell in its native form, membraneanchored DjlA can engage and activate DnaK. How might a DjlA-DnaK chaperone interaction participate in the regulation of wca activation? Because it is known that many members of the J-domain family bind specific substrates (5), it is possible that DjlA acting as a co-chaperone helps to direct substrate (or substrates) in the vicinity of DnaK. In this way, the DnaK chaperone machine may act at a specific site. Other possibilities, including that DjlA exerts indirect effects, cannot be excluded, and must await further analysis. In this context, it is interesting that computer data base searches reveal the existence of hypothetical proteins sharing strong sequence similarity to DjlA in many pathogenic Gram-negative bacteria. The high degree of sequence similarity observed suggests that these putative DjlA orthologs may act as regulatory co-chaperones for cognate Hsp70s in these organisms as well.
Activation of the wca operon by DjlA-DnaK interaction in E. coli implies a chaperone machine capable of modulating either directly or indirectly the RcsB/RcsC phosphotransfer signaling pathway, a role for chaperones which has not been described previously in bacteria. However, a GroEL chaperonin homolog found in the intracellular symbiotic bacterium of aphids has been described as a potential histidine kinase with phosphotransferase activity, although no direct evidence exists for its role in signaling to date (58). In contrast, it is well known in eukaryotes that chaperones actively participate in signaling mechanisms involving the steroid receptor family (59), and a variety of kinases including double strand RNA-activated kinase (PKR) in influenza infection (60), c-Raf1 (61), and members of the Src tyrosine kinase family (62). Future studies should address the extent of chaperone involvement, if any, in bacterial signaling systems.