Influence of GrpE on DnaK-Substrate Interactions*

The DnaK chaperone of Escherichia coli assists protein folding by an ATP-dependent interaction with short peptide stretches within substrate polypeptides. This interaction is regulated by the DnaJ and GrpE co-chaperones, which stimulate ATP hydrolysis and nucleotide exchange by DnaK, respectively. Furthermore, GrpE has been claimed to trigger substrate release independent of its role as a nucleotide exchange factor. However, we show here that GrpE can accelerate substrate release from DnaK exclusively in the presence of ATP. In addition, GrpE prevented the association of peptide substrates with DnaK through an activity of its N-terminal 33 amino acids. A ternary complex of GrpE, DnaK, and a peptide substrate could be observed only when the peptide binding to DnaK precedes GrpE binding. Furthermore, we demonstrate that GrpE slows down the release of a protein substrate, σ32, from DnaK in the absence of ATP. These findings suggest that the ATP-triggered dissociation of GrpE and substrates from DnaK occurs in a concerted fashion.

The DnaK chaperone of Escherichia coli assists protein folding by an ATP-dependent interaction with short peptide stretches within substrate polypeptides. This interaction is regulated by the DnaJ and GrpE co-chaperones, which stimulate ATP hydrolysis and nucleotide exchange by DnaK, respectively. Furthermore, GrpE has been claimed to trigger substrate release independent of its role as a nucleotide exchange factor. However, we show here that GrpE can accelerate substrate release from DnaK exclusively in the presence of ATP. In addition, GrpE prevented the association of peptide substrates with DnaK through an activity of its N-terminal 33 amino acids. A ternary complex of GrpE, DnaK, and a peptide substrate could be observed only when the peptide binding to DnaK precedes GrpE binding. Furthermore, we demonstrate that GrpE slows down the release of a protein substrate, 32 , from DnaK in the absence of ATP. These findings suggest that the ATP-triggered dissociation of GrpE and substrates from DnaK occurs in a concerted fashion.
Central to the chaperone functions of Hsp70 proteins is the transient interaction of their C-terminal substrate-binding domain with short peptide stretches within substrate polypeptides (1)(2)(3). This interaction is controlled by the nucleotide status of the N-terminal ATPase domain of Hsp70, which in turn is regulated by co-chaperones of the DnaJ family and for some Hsp70s by nucleotide exchange factors (4 -7). It is still an unsolved question as to why for some Hsp70s nucleotide exchange factors are essential for their activity in vivo and in vitro while for others nucleotide exchange factors are not essential or do not exist (8 -11). It has been shown recently that the nucleotide dissociation rates of Hsp70 proteins vary over a range of almost 3 orders of magnitude (12). The exchange rate of the Hsp70 homologue with the highest exchange rate was nearly as high as the exchange rate of the Hsp70 with the lowest basal exchange rate in the presence of saturating amounts of its exchange factor. This demonstrates that the molecular structure of Hsp70s does not have to be limiting for nucleotide exchange rates and suggests that an optimal exchange rate has been evolved for each system. Therefore, the question arises why nucleotide exchange factors exist and whether they have additional functions.
For the essential nucleotide exchange factor of Escherichia coli DnaK (GrpE), it was suggested that it also accelerates substrate release in a nucleotide-independent manner (13). This hypothesis was conceived first on the basis of the unusual structure of GrpE, the N-terminal part of which forms a long ␣-helix of 100 Å in length (amino acids 34 -106) (13). This helix extends well beyond the limits of the ATPase domain, suggesting that it might interact with the substrate-binding domain of DnaK (13). The N-terminal 33 amino acids of GrpE that were unstructured and had to be removed for crystallization were proposed to be responsible for the interaction with the substrate-binding domain of DnaK and for influencing substrate release (13). In a later study (14), it was claimed that even in the absence of added ATP, GrpE was able to stimulate substrate dissociation and association by 200-and 60-fold under saturating conditions. Whereas Mally and Witt (14) found the GrpE-mediated substrate release in the absence of nucleotide and in the presence of ATP, Han and Christen (15) claim that the stimulatory effect only occurs in the presence of ADP and not in the absence of the nucleotide. Finally, it has been suggested that GrpE acts as thermometer, because its activity as a nucleotide exchange factor for DnaK is strongly thermosensitive (16 -20). Accordingly, above 45°C, GrpE would be poorly active, which leads to a stabilization of DnaK-substrate complexes.
To further investigate potential novel functions of GrpE in more detail, we analyzed the effect of GrpE on the binding equilibrium and on the association and dissociation kinetics of the interaction of DnaK with peptide and proteins substrates. In addition, we removed the N-terminal 33 amino acids of GrpE (⌬NGrpE) to analyze the effects of the N terminus on peptide binding and release.

EXPERIMENTAL PROCEDURES
Reagents-Unless indicated otherwise, all of the chemicals were of the highest purity available. ATP-agarose (A2767) and luciferase (L9506) were from Sigma. Luciferin was from Fluka.
Substrate Binding and Dissociation Kinetics-Peptide dissociation kinetics using fluorescent-labeled and radiolabeled peptides ( 32 -Q132-Q144-C-AANS, D-NR, and [ 3 H]VSV-GC) were performed as described previously (11,32,33) with the exception that GrpE was added to the reaction mixture or included with the quench peptide at the indicated concentrations using either a stopped-flow apparatus (SX.18MV, Applied Photophysics) or a PerkinElmer LS-55 spectrofluorometer for the fluorescence measurements.  32 were added, and the reaction mixture was incubated further at 30°C for the indicated times before analysis by gel filtration as described above.
Refolding of Thermal-denatured Luciferase-Unfolding and refolding were determined as described previously (36) using HKM buffer with 20 mM K 2 HPO 4 /KH 2 PO 4 , pH 7.6, and a DnaK:DnaJ:GrpE 2 :luciferase ratio of 10:2:0.25:1. Luciferase was denatured in the presence or absence of chaperones for 30 min at 42°C, and refolding was initiated by shifting the reaction mixture to 30°C.
Circular Dichroism-CD measurements were performed using a Jasco J-715 spectropolarimeter and a cuvette with 0.1-cm path length. Thermal-induced unfolding (15-85°C) and refolding (85-15°C) of 1.5 M GrpE 2 or ⌬NGrpE 2 were followed by continuously monitoring the relative ellipticity at 222 nm (bandwidth, 2 nm) with a scan rate of 30°C h Ϫ1 .
GdnCl-induced unfolding of 1.5 M GrpE 2 or ⌬NGrpE 2 was achieved by a stepwise addition of GdnCl up to 5 M final concentration. Refolding experiments with 1.5 M GrpE 2 or ⌬NGrpE 2 were performed by dilution series from 5 to 3 M GdnCl and 3 to 0.28 M GdnCl. All of the protein samples were incubated for 30 min at 30°C before CD measurements took place. For each GdnCl concentration, 10 spectra between 190 and 250 nm (bandwidth, 2 nm) were recorded at a scan rate of 50 nm min Ϫ1 and the average of the molar ellipticity at 222 nm was plotted against its corresponding GdnCl concentration.

Effects of GrpE on DnaK-Peptide Substrate Complexes at
Steady-state Conditions-To investigate the influence of GrpE on the stability of DnaK-substrate complexes, we titrated GrpE to a preformed complex of DnaK⅐ADP with a peptide, which was labeled at a C-terminal cysteine with 2-(4-(iodoacetamido)anilino)naphthalene-6-sulfonic acid ( 32 -Q132-Q144-C-AANS), and measured the fluorescence at steady-state conditions after equilibration of the reaction. As shown in Fig. 1A, the fluorescence of the peptide decreases with increasing concentrations of GrpE. This phenomenon could have two explanations. Firstly, GrpE could quench the fluorescent dye, or secondly, GrpE could destabilize the DnaK-substrate complex. To test the first possibility, we repeated the experiment with a radiolabeled peptide ([ 3 H]VSV-GC) and separated DnaK-bound and non-bound peptides afterward by size-exclusion chromatography. This experiment clearly showed that the amount of DnaK-peptide complex decreases with increasing GrpE concentrations (Fig. 1B).

GrpE Has No Effect on the Dissociation of DnaK-Peptide Substrate Complexes in the Absence of ATP-The
GrpE-induced decrease in the equilibrium concentration of the DnaKsubstrate complex could result either from an increase in the complex dissociation rate as claimed (13)(14)(15) or a decrease in the association rate. We therefore determined the dissociation rate constants of the complex of DnaK with 32 -Q132-Q144-C-AANS in the absence and presence of GrpE wild type protein and the ⌬NGrpE variant and nucleotides. Surprisingly, neither GrpE nor ⌬NGrpE did accelerate the substrate dissociation rate in the absence of nucleotide or in the presence of ADP, which had been purified from contaminating ATP by high pressure liquid chromatography (Fig. 2, A and B). Only when the DnaK⅐ADP-substrate complex was mixed with GrpE ϩ ATP ϩ quench peptide, a significant acceleration of the substrate release was measured ( Fig. 2A). Identical results were obtained when we used a peptide that was labeled at its N terminus with a dansyl moiety (D-NR) or the radiolabeled peptide [ 3 H]VSV-GC (data not shown).
To directly demonstrate the stability of the ternary complex of a GrpE dimer (GrpE 2 ) with DnaK and a peptide, we performed size-exclusion chromatography on a Superdex 200 HR column, which is able to separate DnaK-peptide and GrpE 2 ⅐DnaK-peptide complexes (Fig. 2C). When GrpE was added to a preformed complex of DnaK with [ 3 H]VSV-GC and then after a 1-min incubation at 30°C analyzed by gel filtration, comparable amounts of [ 3 H]VSV-GC were found in the ternary complex in the presence of GrpE as in the binary DnaK-peptide complex in the absence of GrpE. When GrpE was present during the incubation of the peptide with DnaK, only minor amounts of radioactivity eluted at the elution volumes of the binary and ternary complexes.
GrpE but Not ⌬NGrpE Prevents the Association of Peptide Substrates with DnaK-In contrast, the association of the fluorescent 32 -Q132-Q144-C-AANS peptide with DnaK was influenced significantly by the presence of GrpE but not by the presence of ⌬NGrpE. As shown in Fig. 3, the absolute amount of fluorescence that was reached in the plateau phase of the association reaction was decreased with increasing GrpE concentrations. At a 0.5:1 molar ratio of GrpE 2 to DnaK, the plateau of fluorescence was ϳ50% of the value in the absence of GrpE, and at a 1:1 ratio of GrpE 2 to DnaK, almost no fluorescence increase was measured. Interestingly, the association rates were not affected by the presence of GrpE (Table I). Similar experiments were repeated with radiolabeled peptides with identical results (not shown). This indicates that GrpE bound to DnaK prevents the association of peptide substrate. It explains why increasing concentration of GrpE led to a decrease in the absolute amount of fluorescence and radioactivity, respectively, observed at equilibrium (see Fig. 1). Together with the gel filtration experiments, these results demonstrate that GrpE significantly slows down peptide association to DnaK and that a ternary GrpE 2 ⅐DnaK-peptide complex only forms when peptide binds first to DnaK.
Effects of GrpE on the Interaction of DnaK with Protein Substrates-Our data suggest that in the DnaK⅐GrpE 2 complex, the N termini of the GrpE dimer are located in close vicinity to the DnaK substrate-binding cavity. Because the N-terminal 33 amino acids of GrpE are considered to be unstructured, they may in fact bind like a peptide into the substrate-binding cavity of DnaK. If GrpE binds into the substrate-binding pocket, it may inhibit the association of protein substrates to DnaK as well. To test this hypothesis, we incubated radiolabeled E. coli heat shock transcription factor 32 with DnaK in the absence and presence of increasing concentrations of GrpE 2 and analyzed the formed complexes by sizeexclusion chromatography. In the absence of GrpE, two peaks of radioactivity were observed, one at the position of free 32 and one at the position of the 32 ⅐DnaK complex. In the presence of increasing concentrations of GrpE, the peak at the elution volume of the 32 ⅐DnaK complex decreased and an additional peak at lower elution volume appeared (Fig. 4A). This peak corresponds to the ternary complex of a GrpE dimer with DnaK and 32 . The peak representing free 32 eluted at exactly the same position in the presence and absence of GrpE. Integration of the radioactivity in all of the peaks showed that the total amount of bound 32 (DnaK⅐ 32 and GrpE 2 ⅐DnaK⅐ 32 complex) decreased albeit not as dramatically as in the case of the peptides (compare Figs. 1B and 4B). We concluded from these results that the affinity of DnaK for 32 is reduced as well in the presence of GrpE, although not to the same extent as found for peptide substrates.
To investigate this theory in more detail, we determined the dissociation rate of the DnaK⅐ 32 complex in the absence and presence of GrpE. As shown in Fig. 5A, the reduction in affinity of DnaK for 32 32 . The only difference observed was that the starting amount of bound 32 was lower when GrpE was added at the beginning of the preincubation, consistent with the previous titration experiment. When ATP was added together with the unlabeled quench 32 , no 32 or very little bound 32 could be detected independent of whether GrpE was present or absent (Fig. 5B).
Substrate Binding to DnaK Does Not Affect the Activity of GrpE as Nucleotide Exchange Factor-Together with the results of the ⌬NGrpE variant, these data suggest that the N terminus of GrpE is in close vicinity of the substrate-binding pocket. Therefore, we hypothesized that GrpE would be able to monitor the loading status of the substrate-binding domain and that bound substrate could in this way have an influence on the nucleotide exchange activity of GrpE. To test this hypothesis, we determined the GrpE-or ⌬NGrpE-accelerated nucleotide exchange rates in the absence and presence of a peptide or protein substrate. For this purpose, we preincubated DnaK with the fluorescently labeled ADP analogue N 8 -(4-NЈ-methylanthraniloyl)adenosine-5Ј-diphosphate-ADP (29) in the absence or presence of the peptide substrate 32 -Q132-Q144-C or the protein substrate 32 for 2 h and then rapidly mixed these complexes in a stopped-flow instrument with unlabeled ATP. No significant differences in the nucleotide exchange rates were observed (data not shown), clearly indicating that the loading status of the substrate-binding domain does not influence the nucleotide exchange function of GrpE.
The close vicinity of the N terminus of GrpE and the substrate that is bound in the substrate-binding domain of DnaK could have an influence on the chaperone function of DnaK. Therefore, we determined the efficiency of DnaK and DnaJ in refolding the thermally denatured firefly luciferase in the presence of GrpE or ⌬NGrpE. Luciferase was denatured for 30 min at 42°C in the presence of DnaK, DnaJ, and either GrpE or ⌬NGrpE and subsequently shifted to 30°C. The loss of activity during denaturation and the regain of activity due to refolding at 30°C were monitored. No significant differences between GrpE and ⌬NGrpE could be detected (Fig. 6).
Expression of grpE- (34 -197) Complements the ⌬grpE Temperature Sensitivity Phenotype-To test the functional importance of the N-terminal 33 amino acids in vivo, we tried to complement the temperature-sensitive phenotype of a ⌬grpE strain by ectopically expressing wild type GrpE or ⌬NgrpEencoding genes from an IPTG-inducible promoter (Table II). At 30°C, the ⌬grpE strains containing the vector alone or the plasmid encoding ⌬NGrpE grew well at all of the IPTG concentrations. In contrast, the strain containing the wild type GrpEencoding gene did not grow at IPTG concentrations of 100 M and above. At 42°C, the ⌬grpE strain containing the vector alone did not grow. The strain containing the wild type GrpEencoding gene grew in the absence of IPTG and not at concentrations of IPTG of 100 M or above. The strain containing the ⌬NgrpE-encoding gene only grew in the presence of 100 M or more IPTG. To better understand these results, we determined the in vivo concentrations of GrpE and ⌬NGrpE in the ⌬grpE strain at the different IPTG concentrations through quantitative Western blotting using GrpE-specific antiserum. We found that GrpE was present even in the absence of IPTG, probably because of contaminations of the media with lactose as previously observed in our laboratory on several occasions. At 100 M IPTG and above, the GrpE concentrations were increased dramatically by Ͼ1000-fold over the GrpE level of the wild type strains. This strong overproduction of GrpE explains the lethal phenotype because an excess of GrpE is detrimental to the chaperone function of the DnaK system (34). In contrast, the level of ⌬NGrpE at 500 M IPTG was even still lower than the level of wild type GrpE in the absence of IPTG. Therefore, the reason for the difference in their ability to complement the temperature-sensitive phenotype of the ⌬grpE strain was attributed to the difference of the amount of produced protein and not due to the differences in their biochemical properties.

GrpE Folding Is Positively Influenced by Its N-terminal 34
Residues-Because both grpE variants were expressed from the same promoters in identical plasmids and RNA-secondary structure prediction programs did not predict any significant differences in the secondary structure, we hypothesized that the N terminus of GrpE may influence the folding and/or stability of GrpE. To test this hypothesis, we performed thermal and chemical denaturation and renaturation experiments monitoring the folded state of GrpE by circular dichroism and fluorescence spectroscopy. Both methods yielded identical re-

TABLE I
The rate of DnaK-peptide association is independent of GrpE DnaK alone or in complex with GrpE or ⌬NGrpE at the indicated ratios was mixed with 32 -Q132-Q144-C-AANS Ϯ GrpE. The increase in fluorescence was fitted to single or double exponential equations. F-Pep, 32 -Q132-Q144-C-AANS; A, amplitude.   sults. The denaturation kinetics of GrpE and ⌬NGrpE did not exhibit any significant differences. The renaturation kinetics, however, showed clear differences. Whereas full-length GrpE renatured efficiently under the conditions tested, ⌬NGrpE renatured much more inefficiently and only a fraction of the original circular dichroism and fluorescence signal could be regained ( Fig. 7 and data not shown).

DISCUSSION
Our findings that GrpE and not ⌬NGrpE prevents the association of peptides to DnaK while it is capable of binding to a DnaK⅐peptide complex to form a ternary complex could be explained in two ways. First, the N terminus of GrpE interacts with the substrate-binding domain in a way that the opening of the substrate-binding pocket is prevented. Second, the 33 Nterminal residues are either located in close proximity or directly interact with the substrate-binding pocket. The first hypothesis is unlikely because GrpE had no influence on the dissociation of prebound peptides, which would be expected if GrpE would hamper the substrate-binding domain in its opening movement. The second hypothesis is supported by the fact that the N terminus of GrpE is unstructured and thereby could mimic an unfolded protein. The published algorithm for the identification of DnaK-binding sites does not identify a high affinity-binding site within the N-terminal region of GrpE, in particular, because this region of the protein is rich in negatively charged residues (37). However, a high affinity-binding site might not be necessary because GrpE binds with a K d of 35 nM to the ATPase domain, and when the N terminus comes in close vicinity of the substrate-binding pocket, even a low affinity-binding site may enter the substrate-binding pocket because of the local high effective concentration of this peptide stretch. Consistent with this hypothesis are the results of a recent publication by Witt and co-workers (38) in which they demonstrate that GrpE but not ⌬NGrpE interacts with low affinity with a substrate-binding domain fragment of DnaK (DnaK-(393-507); K d ϭ 67 M). This interaction with the substrate-binding domain, however, does not contribute significantly to the overall affinity of the GrpE-DnaK interaction because ⌬NGrpE binds with similar affinity to DnaK as wild type GrpE and deletion of an exposed loop in the ATPase domain strongly affects the binding of GrpE to DnaK (25,38). Whether other nucleotide exchange factors of Hsp70s also interact with the substrate-binding domain of their respective partner protein has not been investigated so far.
We and others (35,39,40) show that DnaJ also interacts with the substrate-binding pocket of DnaK, thereby competing with substrates, whereas the main ATPase-stimulating interaction occurs in the ATPase domain (41)(42)(43)(44)(45). It is intriguing that both co-chaperones that regulate the ATPase cycle of DnaK also interact with the substrate-binding domain. Because ATP binding and hydrolysis induces conformational changes not only in the ATPase domain of DnaK but also in the substrate-binding domain, an interaction of both regulatory cofactors with both domains of DnaK may contribute to the coordination of these conformational changes. Vice versa, the conformational change of the substrate-binding domain of DnaK upon association of ATP to its ATPase domain may contribute to the rapid release of GrpE. This might be important given the fact that GrpE binds with high affinity to the ATPase domain and, in the ADP state, has a very low dissociation rate (38). The dissociation rate of GrpE from DnaK upon the addition of ATP has not yet been determined.
A second intriguing finding of our study was that the dissociation rate of the complex of DnaK with a protein substrate was (approximately 2-fold) slowed down significantly in the presence of GrpE. This means that, under the conditions of   high ATP concentrations, the release of the substrate and the release of GrpE are practically always coordinated. Under stress and starvation conditions when the ATP concentration might drop significantly, substrate dissociation is slowed down further through the action of GrpE. Because under these conditions the ATP driven chaperone cycle is not functioning, the release of a misfolded protein substrate would lead to its aggregation with high probability. Similarly, it was shown recently that substrate release from DnaK at a high temperature above 45°C, when chances for refolding to the native state are also low, is slowed down because of a reversible thermal transition of GrpE, which thereby loses the ability to trigger nucleotide exchange (16 -18). Our findings that the N-terminal 33 amino acids are important for the folding of GrpE are consistent with these studies. Dimerization, which is essential for the nucleotide exchange function of GrpE, may start with an interaction of the N-terminal 33 amino acid of both monomers with each other and then proceed through the long ␣-helices in a zipper-like fashion.
Two recent independent studies claim that GrpE accelerates substrate binding to and the release from DnaK independent of its nucleotide exchange function (14,15). Our findings are in contrast to these results. We first suspected that our GrpE preparation might have experienced a partial loss of function. Therefore, we carefully tested the functionality of our protein preparations. First, the structural integrity was studied using circular dichroism spectroscopy including the monitoring of thermal and chemical denaturation curves. Second, the nucleotide exchange activity was determined in direct measurements of the ADP and ATP dissociation rates with increasing GrpE concentrations as described earlier (34) or indirectly by determining the steady-state ATPase rates of DnaK in the presence of DnaJ and GrpE. Third, the functionality of GrpE also was shown in chaperone assays by preventing the aggregation and refolding thermally and chemically denatured luciferase. Finally, for our study, we used at least four independent protein preparations. We purified GrpE from ⌬dnaK strains to avoid contaminations with DnaK, but we also purified it from a wild type strain in case that DnaK might be important for the folding of GrpE. DnaK was in this case removed by extensive washing with an ATP-containing buffer. To make sure that this GrpE preparation is not contaminated with ATP, we monitored any residual ATP by the very sensitive luciferase assay and found that the contamination is lower than 0.1% of the GrpE concentration. One GrpE preparation was purified in a different laboratory (a kind gift of H.-J. Schönfeld), which also supplied the GrpE that was used in one of the above mentioned studies (15). We could not find any indication that our GrpE preparations are defective and therefore concluded that there must be a different reason for this discrepancy.
The two publications (14,15) reporting the substrate release function of GrpE are not completely concordant. Whereas Mally and Witt (14) show that GrpE stimulated peptide release in the absence of nucleotide as well as in the presence of ADP, Han and Christen (15) could only find the GrpE stimulation of substrate release in the presence of ADP. Because DnaK has a much higher affinity for ATP (K D ϭ 1 nM) than for ADP (K D ϭ 25 nM) (29,46), small contaminations of the ADP solutions with ATP could explain the results because GrpE would accelerate ADP release and ATP would preferentially bind to DnaK triggering substrate release. We found that commercially available ADP is contaminated usually with 3-5% ATP and therefore always purify our ADP by high performance liquid chromatography. Such a contamination of ADP with ATP, however, could not explain the results of Mally and Witt (14), showing that GrpE stimulated the substrate release from DnaK in the absence of nucleotide. Mally and Witt (14) purified their GrpE by affinity chromatography on immobilized DnaK according to Zylicz et al. (47). In this protocol, GrpE is eluted with a buffer containing 10 mM ATP and dialyzed subsequently. If this final dialyzing step is not carried out extensively as described originally, i.e. with three buffer changes of sufficient volume, significant amounts of ATP can remain in the GrpE preparation, which would then be able to trigger substrate release by DnaK. Based on these considerations and our own data presented here, we must conclude that GrpE does not have the claimed function to trigger substrate release from DnaK even in the absence of nucleotide. Instead, GrpE decelerates the dissociation of protein substrates in the absence of ATP, and it is likely that a coordinated dissociation of GrpE and substrate occurs when ATP binds to the ATPase domain of DnaK.