Functional Interaction of the Auxilin J Domain with the Nucleotide- and Substrate-binding Modules of Hsc70*

The uncoating of clathrin-coated vesicles requires the DnaJ homologue auxilin for targeting Hsc70 to clathrin coats. This function involves a transient interaction of the auxilin J domain with Hsc70. We have now identified the structural elements of Hsc70 that are responsible for the uncoating activity, and we show that the hitherto accepted view, which implicates the 10-kDa carboxyl-terminal variable domain of Hsc70, is incorrect. A 60-kDa chymotryptic or analogous recombinant fragment of Hsc70, which contains the ATPase- and substrate-binding domains, is sufficient to liberate clathrin from coated vesicles. Consistent with this was the observation that Hsp70 uncoats coated vesicles with the same efficacy as Hsc70 and that DnaK possesses vestigial uncoating activity. Direct binding studies demonstrated that the auxilin J domain undergoes an ATP-dependent reaction only with fragments of Hsc70 that contain both the ATPase- and substrate-binding domains. The individual domains by themselves did not bind to the J domain nor did a recombinant protein that contained the substrate-binding domain attached to the 10-kDa variable domain.

The uncoating of clathrin-coated vesicles requires the DnaJ homologue auxilin for targeting Hsc70 to clathrin coats. This function involves a transient interaction of the auxilin J domain with Hsc70. We have now identified the structural elements of Hsc70 that are responsible for the uncoating activity, and we show that the hitherto accepted view, which implicates the 10-kDa carboxylterminal variable domain of Hsc70, is incorrect. A 60-kDa chymotryptic or analogous recombinant fragment of Hsc70, which contains the ATPase-and substratebinding domains, is sufficient to liberate clathrin from coated vesicles. Consistent with this was the observation that Hsp70 uncoats coated vesicles with the same efficacy as Hsc70 and that DnaK possesses vestigial uncoating activity. Direct binding studies demonstrated that the auxilin J domain undergoes an ATP-dependent reaction only with fragments of Hsc70 that contain both the ATPase-and substrate-binding domains. The individual domains by themselves did not bind to the J domain nor did a recombinant protein that contained the substrate-binding domain attached to the 10-kDa variable domain.
Proteins of the 70-kDa heat shock protein (Hsp70) 1 family participate in numerous cellular functions such as protein folding, protein translocation across membranes, and reorganization of macromolecular complexes (1,2). The underlying basis of these apparently diverse functions is the ability of Hsp70 proteins to interact reversibly with extended polypeptide segments of at least 7 residues generally containing large hydrophobic and basic side chains (3). Hsp70 proteins contain an amino-terminal 44-kDa ATPase domain, a 18-kDa substratebinding domain, and a 10-kDa carboxyl-terminal domain (4). The carboxyl-terminal domain is the least conserved of the three Hsp70 domains and is therefore also referred to as the variable domain (5). The type of nucleotide present in the ATPase domain determines the accessibility of the substratebinding site for substrates. Thus ATP promotes and ADP inhibits substrate release (6,7).
The structure of a recombinant fragment of DnaK (bacterial Hsp70) that includes the substrate-binding domain and most of the 10-kDa variable domain reveals that the first half (Asp 391 -Ile 501 ) has the form of a ␤-sandwich while the second (Glu 509 -His 606 ) folds into five ␣-helical segments (8). Substrates bind in an extended conformation to a channel formed by loops of the ␤-sandwich. The second helix, designated ␣B, covers the binding channel like a lid, thereby presumably preventing the escape of bound substrates (see Fig. 9). It has been suggested that the binding of ATP induces conformational changes that expose the substrate-binding channel by displacement of the ␣␤-helix (8). Hydrolysis of ATP would have the opposite effect.
The intrinsic ATPase activity of Hsp70 is very low but is synergistically stimulated by substrates and by members of the DnaJ protein family. Other factors, such as GrpE in bacteria (9) or Hip in eukaryotes, respectively, enhance or inhibit nucleotide exchange (10). DnaJ proteins consist of highly conserved elements, together with unique domains that distinguish the proteins from each other (11). The feature common to all DnaJ homologues is the J domain, a conserved module of about 70 residues, which is essential for the interaction with Hsp70 proteins (12)(13)(14). Nuclear magnetic resonance spectroscopy of the J domain of bacterial DnaJ and human Hdj-1 showed that it folds into four helices that contribute side chains to a hydrophobic core (15,16). Two of the helices are engaged in coiled-coil interactions. The absolutely conserved and functionally essential tripeptide, "HPD," forms part of a connecting loop between the two antiparallel helices. Less conserved charged residues on the helix exteriors may recognize particular Hsp70 isoforms (15,16). Other unrelated domains in DnaJ proteins are believed to be involved in recognition of substrates for Hsp70s (11). Thus, the combination of a given J domain with a substrate recognition domain allows the targeting of a Hsp70 protein to a specific substrate or class of substrates. DnaJ and auxilin appear to be representatives of two types of DnaJ class proteins; the former displays affinity for a broad range of denatured polypeptides (11) wheras auxilin acquired a high specificity for assembled clathrin (17)(18)(19).
Clathrin is the major structural component of the regular polygonal lattice that encases a certain type of intracellular transport vesicle (20). Binding of auxilin to assembled clathrin initiates Hsc70-dependent release of clathrin protomers (triskelia) from the vesicle membrane (18). Although the auxilin J domain differs considerably from that of DnaJ or Hdj-1, key residues, known to be engaged in the formation of the hydrophobic core, and the essential HPD segment are conserved (18). Moreover, the auxilin J domain binds to Hsc70 and stimulates its ATPase (19).
Nothing less than intact Hsc70 has previously been reported to uncoat clathrin-coated vesicles (4,21). A chymotryptic 60-kDa fragment lacking the 10-kDa carboxyl-terminal domain would bind only to free clathrin triskelia and to unfolded peptide substrates. More recently a DnaK mutant with a deletion of 94 residues at the carboxyl terminus was reported to interact poorly with immobilized DnaJ (22), and similarly, a human Hsp70 lacking the conserved carboxyl-terminal motif EEVD could no longer be stimulated by the DnaJ homologue Hdj-1 and failed to refold denatured luciferase (23). Taken together, these observations seemingly imply that the carboxyl-terminal domain of Hsp70 interacts with DnaJ homologues and is essential for a complete Hsp70-mediated reaction cycle.
The ATP-induced complex between the auxilin J domain and Hsc70 is relatively stable (19,24). This made it possible to isolate complexes containing both proteins. It should therefore be in principal possible to identify the Hsc70 domains that participate in this interaction. Here we demonstrate that the variable 10-kDa carboxyl-terminal domain of Hsc70 is not required for the binding of auxilin or for the in vitro uncoating of clathrin-coated vesicles. Our results suggest instead that in the ATP-bound state of Hsc70 the ATPase domain interacts with the substrate-binding domain to create a binding site for the J domain.

MATERIALS AND METHODS
Plasmid Constructions-The construction of plasmids used for the expression of GST-auxilin-(547-910), GST-auxilin-(813-910), and GSTauxilin-(547-814) has been described in detail elsewhere (19). BovHsc70.pRSET which was used for the construction of recombinant Hsc70 fragments was a kind gift from Dr. McKay (Stanford). The construct no longer contained a mutation present in the original clone (25), which changed the glutamic acid at position 543 to a lysine. 2 For the construction of PGEXHsc70-(1-540) BovHsc70.pRSET was digested with NdeI and EcoRI. The 1621-bp fragment was treated with Klenow enzyme and inserted into the SmaI site of the expression vector PGEX-4T-1. PGEXHsc70-(373-650) was obtained by digestion of BovHsc70.pRSET with Nar1 and HindIII. The resulting 876-bp fragment was blunted using Klenow enzyme and inserted into the SmaI site of the expression vector PGEX-4T-1. PGEXHsc70-(373-540) was obtained by EcoRI digestion of PGEXHsc70-(373-650). The excised 368-bp EcoRI fragment was removed by agarose gel electrophoresis and the vector with the remaining insert re-ligated. PGEXHsc70-(540 -650) was constructed by ligation of the 368-bp EcoRI fragment into the EcoRI site of PGEX-4T-3. All restriction enzymes were obtained from Boehringer Mannheim. The pGEX-vectors were from Pharmacia Biotech Inc.
Protein Purifications-Coated vesicles and clathrin were purified from frozen bovine brains and adrenal gland (Pel-Freez, Rogers, AR) as detailed elsewhere (26,27). Adrenal clathrin coats were prepared by extracting clathrin and adaptors from adrenal coated vesicles with 0.5 M Tris-Cl (28). Soluble proteins were reassembled into coats by dialysis against buffer A (0.1 M Mes, 1 mM EGTA, 0.5 mM MgCl 2 , 3 mM CaCl 2 , pH 6.5). The coats were harvested by ultracentrifugation and resuspended in buffer C (20 mM HEPES, 25 mM KCl, 2 mM MgCl 2 , 10 mM NH 4 SO 4 , pH 7.0). Bovine Hsc70 was purified as described previously (29,30). DnaK was expressed in the Escherichia coli strain TG1 which was transformed with plasmid pJMII (31) and purified as described elsewhere (30). Human Hsp70 was expressed in BL21 which was transformed with the construct pETWTHsp70 (32), kindly supplied by Dr. Evan Eisenberg (National Institutes of Health). The bacteria were grown for 3 h at 36°C and after induction with 1 mM isopropyl-1-thio-␤-D-galactopyranoside for another 3 h at 28°C. Hsp70 was isolated from the bacterial lysate following a procedure developed for the purification of recombinant rat Hsc70 (33). GST-Hsc70 fusion proteins were expressed in E. coli Bl21. The cells were grown for 3 h at 36°C in the absence of isopropyl-1-thio-␤-D-galactopyranoside and then for 3 h at 22°C in presence of 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside. Bacteria were lysed by sonication as described elsewhere (32). Fusion proteins were purified by affinity chromatography on glutathione-Sepharose 4B (Pharmacia) following the manufacturer's instructions. In most cases the fusion proteins were further purified by gel filtration in buffer C on Superose 12 (Pharmacia). GST-Hsc70-(1-540) used for ATPase determinations was further purified by hydroxyapatite adsorption chromatography. Approximately 2 mg of protein were applied to a 1-ml hydroxyapatite column connected to a fast protein liquid chromatography system (Pharmacia). The protein was eluted with a 20-ml gradient ranging from 0 to 500 mM phosphate, pH 7.2, at a flow rate of 0.5 ml/min and then once more affinity purified on GSH-Sepharose.
Protein concentrations were determined spectroscopically. Extinction coefficients were calculated from the amino acid compositions. In some instances the GST moiety of GST-auxilin-(547-910) and GSTauxilin-(547-814) was cleaved by digesting 1 mg of fusion protein with 1 unit of thrombin (number 154163 from ICN, Costa Mesa, CA) for 16 h at 20°C in 20 mM Tris-HCl, pH 8.4, 150 mM NaCl, 2.5 mM CaCl 2 . The reaction was terminated with 1 mM phenylmethylsulfonyl fluoride. The GST moiety was removed by affinity chromatography and thrombin by gel filtration on Superose 12. The purified fusion proteins were concentrated in Centricon microconcentrators and stored frozen in small aliquots at Ϫ80°C.
Digestion of Hsc70 -To convert Hsc70 into its ATP-or ADP-bound state, the protein was preincubated for 10 min at 25°C in buffer C supplemented with either 3 mM ATP or 3 mM ADP. Digestions with chymotrypsin were carried out as detailed in the legend to Fig. 3. The reaction was terminated with phenylmethylsulfonyl fluoride at a final concentration of 1 mM.
Enzymatic Dissociation of Coated Vesicles-Coated vesicles (0.4 M) from bovine brain in buffer C were supplemented with GST-auxilin-(547-910) (0.2 M) before they were used as substrate in the uncoating assay. This was done to make sure that the endogenous auxilin was not a limiting factor in the uncoating reaction. Dissociation experiments were performed in volumes of 75-100 l in buffer C supplemented with 2 mM ATP, 5 mM phosphocreatine (Sigma), and 5 units/ml creatine phosphokinase (Sigma) (34). The incubation time was 15 min and the temperature 25°C. Protein concentrations are stated in the legends to the figures. Dissociated clathrin was separated from baskets by ultracentrifugation (20 min at 109,000 ϫ g) using a Beckman TLA-45 rotor in a TL-100 benchtop centrifuge. The amount of clathrin present in the supernatant and pellet fractions was determined by densitometry after SDS-PAGE using a Molecular Dynamics instrument (35). For preparative uncoating experiments with digested Hsc70, brain coated vesicles (0.52 mg of protein) were incubated with 19 g of GST-auxilin-(547-910) and 0.28 mg of chymotrypsin-digested hsc70 (composed mainly of 60-and 45-kDa fragments) for 20 min at 25°C in the presence of 4 mM ATP. The reaction was terminated by adding hexokinase to a final concentration of 10 units/ml and 5 mM glucose. Released clathrin was separated from intact baskets by ultracentrifugation. The supernatant was applied directly to a Superose 6 gel filtration column equilibrated with buffer C containing 0.1 mM ADP. The column was eluted at room temperature with a flow rate of 0.5 ml/min. Fractions were analyzed by SDS-PAGE.
Interaction between Hsc70 Fragments and the Auxilin J Domain-A chymotryptic digest of Hsc70, prepared in ADP or ATP, respectively, or recombinant Hsc70 fragments without their GST-moieties, were incubated in buffer C with GST-auxilin-(813-910) for 10 min at 25°C. As indicated in the figures, the buffer contained either 3 mM ATP or 3 mM ADP, respectively. Incubations with GST served as controls. At the end of the incubation period, 20 -40 l of glutathione-Sepharose beads were added, and the incubation was continued for 15 min on ice to allow the fusion proteins to attach to the beads. The beads were pelleted by a 30-s spin at maximum speed in an Eppendorf benchtop centrifuge at 4°C. The beads were extensively washed in buffer C containing either 0.1 mM ATP or 0.1 mM ADP, as detailed elsewhere (19), and then resuspended in 100 l of SDS sample buffer to extract the bound protein which was subsequently analyzed by SDS-PAGE.
ATPase Assays and Nucleotide Analysis-To determine ATPase activities, 1. Unbound nucleotides were removed on a NAP-5 column as described before (19). Hsc70 ATP or GST-Hsc70-(1-540) ATP was then incubated in the presence and absence of GST-auxilin-(547-910) or GST-auxilin-(547-814) for 5 min at 25°C as indicated in the figure legends. The ratio of ATP to ADP was determined as described above.
Miscellaneous Techniques-SDS-PAGE and immunoblotting procedures were performed as described previously (35)

Clathrin Coat Dissociation by Hsp70 and DnaK-Although
the members of the Hsp70 family are generally highly conserved proteins, the sequence of the 10-kDa carboxyl-terminal domain can vary considerably (5). This prompted speculations about a possible involvement of the 10-kDa domain in the interaction with other cellular components, notably DnaJ homologues (21)(22)(23). To explore how different Hsp70 proteins function in the auxilin-dependent uncoating of clathrin-coated vesicles, we compared the activity of bovine Hsc70 with those of human Hsp70 and bacterial DnaK. Human Hsp70 was chosen, because its 10-kDa domain has 62% identity with that of bovine Hsc70 (Fig. 1A), and both proteins have the EEVD motif at their carboxyl termini. In DnaK the EEVD motif is replaced by EEVK and is followed by the sequence DKK. Overall there is only 19% identity between the carboxyl-terminal domains of DnaK and Hsc70. The degrees of identity between the substrate-and nucleotide-binding domains of Hsc70 and DnaK are 56 and 50%, respectively (Fig. 1A). Both, Hsp70 and DnaK were purified from overexpressing bacteria (Fig. 1B).
To assess uncoating activity, a constant amount of clathrincoated vesicles was incubated for a fixed time in the presence of ATP with increasing concentrations of Hsp70 proteins. The liberated clathrin was separated from the vesicle membrane by pelleting and quantitated by SDS-PAGE. Bovine Hsc70 and human Hsp70 dissociated bovine brain clathrin-coated vesicles with similar efficiency, but the activity of DnaK was about 10 times lower (Fig. 2). This could result from a lower affinity for the auxilin J domain or for clathrin or, as seems probable, from a combination of both; we observed earlier that DnaK has a lower affinity than Hsc70 for auxilin (19), and current attempts to isolate DnaK-clathrin complexes by gel filtration in the presence of ADP proved unsuccessful, indicating that the complexes that would have formed in the course of the uncoating reaction were unstable (data not shown). In contrast, complexes of Hsc70 with clathrin are readily demonstrated by this procedure (4,19,36).
In sum, when compared with bovine Hsc70, both human Hsp70 and bacterial DnaK, showed uncoating activity, strong in one case and weak in the other, despite extensive sequence differences between the three variable 10-kDa carboxyl-terminal domains and lesser differences between the other domains. While these results fall short of establishing definitively whether or not the variable domain participates, as had been supposed, in an interaction with DnaJ class proteins, they nevertheless place the issue in doubt. To eliminate any possibility that the in vitro uncoating assay may not have been sensitive enough to discriminate between an interaction of the 10-kDa domain with a "cognate" as against a "non-cognate" J domain, we extended our study to fragments of Hsc70.
Uncoating Activity of Hsc70 Fragments-Limited proteolysis of Hsc70 with chymotrypsin progresses through a 66-kDa fragment, which gives rise to two closely spaced bands of about 60 kDa and finally a stable 45-kDa fragment (Fig. 3). On immunoblots of the digest the Hsc70-specific monoclonal antibody 1B5 (37) stained only intact Hsc70 and the fusion protein GST-Hsc70-(540 -650) that was included as a control. This suggested that the 66-kDa fragment was missing a carboxylterminal segment of about 5 kDa. The 60-kDa fragment is known to contain the ATPase and substrate-binding domains of Hsc70, whereas the stable 45-kDa fragment comprises only the ATPase domain (4). As observed previously for DnaK and Hsp70 (23,38,39) the type of nucleotide present in the nucleotide-binding domain influenced the course of the digestion (Fig. 3). With bound ATP, the 60-kDa fragment proved significantly more stable than with ADP. This suggests that the region in between the ATPase and the substrate-binding domains is less accessible to chymotrypsin in the ATP-bound state. In the ADP-bound state the site near the carboxyl terminus, which is cleaved when the 66-kDa fragment is generated, appears to be slightly less accessible. It is conceivable that it forms part of the mechanism that locks the "lid" in a place over the substrate-binding channel. To eliminate as much of the intact Hsc70 as possible, the digestion time was increased to 60 min. This resulted in a mixture of predominantly 60-and 45-kDa fragments (Fig. 3). No intact Hsc70 remained, but a small amount of 66-kDa material, amounting to about 1.5% of the combined staining intensities of the 60-and 45-kDa bands, could be detected. When these fragments were tested in the uncoating assay they were found to be active (Fig. 4A). It could be presumed that the activity was associated with the 66-kDa and especially the much more abundant 60-kDa fragment, because both contain the ATPase-and substrate-binding domains. Based on the concentration of the 60-kDa fragment, which was estimated by densitometry after SDS-PAGE, it appeared to have half the activity of Hsc70 (Fig. 4B). To determine whether the 60-kDa fragment was bound to clathrin at the end of the reaction, coated vesicles were uncoated on a semi-preparative scale. The reaction was stopped by addition of hexokinase and glucose. The supernatant containing the released clathrin was immediately applied to a S6 gel filtration column. Analysis of the clathrin-containing fractions by SDS-PAGE demonstrated the presence of the 60-kDa fragment and some 66-kDa fragment (Fig. 5). This result was in agreement with previously published data which showed that the 60-kDa fragment but not the 45-kDa nucleotide-binding domain can bind to clathrin (4).
It is clearly crucial to exclude the possibility that parts of the 10-kDa carboxyl-terminal domain had remained attached to the 60-kDa fragment and contributed to the uncoating function. We therefore expressed a fragment of Hsc70 as a GST fusion protein (GST-Hsc70-(1-540)) in bacteria. GST-Hsc70-(1-540), the sequence of which corresponded closely to that of the 60-kDa chymotryptic fragment, released clathrin from coated vesicles with the same efficacy as the chymotryptic fragment (Fig. 4, A and B). After thrombin digestion, to liberate Hsc70-(1-540) from the GST dimer, the uncoating activity became almost indistinguishable from that of intact bovine Hsc70 (Fig. 4B). We also constructed fusion proteins containing the substrate-binding domain attached to the variable domain (Hsc70-(373-650)), but it could not uncoat clathrin-coated vesicles (Fig. 4A). Moreover, the monoclonal antibody 1B5, which reacts with an epitope in the variable domain, did not interfere with uncoating (data not shown).
Our results taken together suggest strongly that the Hsc70 segment 1-540 suffices for the auxilin-dependent dissociation of clathrin from coated vesicles.  5. Association of the 60-kDa chymotryptic Hsc70 fragment with clathrin. Coated vesicles were incubated in the presence of ATP with digested Hsc70. The reaction was terminated upon addition of glucose and hexokinase. The released clathrin was separated by ultracentrifugation and fractionated by gel filtration on Superose 6. Track 1, digested Hsc70; track 2, supernatant after uncoating; tracks 3-7, clathrin-containing Superose 6 peak fractions. The bulk of the clathrin eluted from the column at the volume as expected for triskelia. Note the presence of the 60-kDa fragment in these fractions.

Interaction between GST-Auxilin-(813-910) with Domains of
Hsc70 -Having excluded any direct role of the carboxyl-terminal 10-kDa domain of Hsc70 in the uncoating reaction, we wished to determine which part of the 60-kDa fragment binds to the J domain. To this end a mixture of intact Hsc70, 66-, 60-, and 45-kDa chymotryptic Hsc70 fragments was prepared and incubated with either GST or with GST-auxilin-(813-910) in the presence of ATP or ADP. Binding was assayed by retrieving GST-containing proteins and protein complexes on GSH-Sepharose. SDS-PAGE analysis of the adsorbed proteins demonstrated that intact Hsc70, as well as the 66-and the 60-kDa fragment, interacted with GST-auxilin-(813-910), whereas the 45-kDa ATPase domain did not (Fig. 7A). Intact Hsc70 appeared to be the marginally preferred substrate. Binding was stimulated by ATP and required the J domain (Fig. 7A). Fragments prepared in the presence of ATP did not behave any differently (Fig. 7A, track 4). These findings were corroborated with recombinant Hsc70 fragments. Like the 60-kDa fragment, Hsc70-(1-540) entered into a complex with the auxilin J domain, whereas the substrate-binding domain by itself (Hsc70-(373-540)) or in association with the carboxyl-terminal domain (Hsc70-(373-650)) did not bind (Fig. 7B). Generally the interaction of the auxilin J domain with Hsc70-(1-540) appeared to be not quite as strong as with intact bovine brain Hsc70 or with the 60-kDa chymotryptic fragment, but this difference did not result in a reduced uncoating activity.
In common with those of other DnaJ proteins the auxilin J domain was shown to stimulate the ATPase of Hsc70 (19). To determine whether the auxilin J domain would also activate the ATPase of the carboxyl-terminally truncated Hsc70, we used highly purified GST-Hsc70-(1-540). To eliminate possible contamination by bacterial ATPases, GST-Hsc70-(1-540), after the initial affinity purification step on GSH-Sepharose, was further purified by hydroxyapatite chromatography and then once more affinity purified on GSH-Sepharose. In contrast to intact Hsc70, we observed that addition of GST-auxilin-(547-910) to GST-Hsc70-(1-540) had only a very small stimulatory effect on the ATPase activity (Fig. 8A). Unexpectedly, the nucleotide associated with the fusion protein had already been largely hydrolyzed to ADP after only 5 min of incubation at 25°C in the absence of a J domain. We noted, moreover, that even before the incubation at 25°C more than 50% of the bound ATP had already been hydrolyzed (Fig. 8B). This result suggested that elimination of the carboxyl-terminal domain increased the basal ATPase activity of GST -Hsc70-(1-540), and rendered it much less responsive to stimulation by the auxilin J domain. The ATPase of mutant human Hsp70 lacking an intact EEVD motif was recently also reported to be unresponsive to stimulation by the DnaJ homologue Hdj-1 (23).
In sum, our data indicate that the observed interaction between GST-auxilin-(813-910) requires both the substrate-binding domain and the nucleotide-binding domain of Hsc70. DISCUSSION We have presented evidence that in vitro the variable 10-kDa carboxyl-terminal domain of Hsc70 is not essential for the interaction of Hsc70 with the auxilin J domain or for the uncoating of clathrin-coated vesicles. These results were unexpected, because several previous reports had suggested otherwise (4, 21, 23). However, when the experimental details and conclusions of those reports are related to our experimental conditions and results, some of the points of disagreement can be resolved. Rothman and co-workers (4) were the first to report that upon chymotrypsin digestion the decrease of intact Hsc70 with generation of the 60-and 44-kDa fragments paralleled the loss of uncoating activity. The uncoating assay used by these authors differed from ours in many respects. First, reconstituted clathrin cages, which did not contain stabilizing associated proteins, were used as the substrate for Hsc70. Second, the dissociation reaction was performed in the absence of significant amounts of auxilin, which had not yet been discovered. Considering these differences in the assay conditions, it is not possible to compare them directly. However, it should be noted that the data (see Figs. 1 and 3 in Ref. 4) do not necessarily support a close relationship between the amount of dissociated clathrin and loss of intact Hsc70. After 10 min of digestion the concentration of intact Hsc70 had dropped to about 30% of its original concentration, but coat dissociating activity remained unimpaired.
Subsequently a recombinant approach was used to express a 60-kDa fragment from rat Hsc70 in bacteria (21). The fragment was almost identical to our Hsc70-(1-540), except for four additional carboxyl-terminal residues. The purified recombinant protein caused only very little clathrin release from the vesicle membrane. Unfortunately, the intact recombinant Hsc70, used as a positive control, also had low activity and released no more than 10% of the total clathrin (see Fig. 3A in Ref. 21). The experimental conditions are therefore incompatible with ours.
More recently it was shown that mutations within the widely conserved carboxyl-terminal EEVD motif of human Hsp70 rendered its ATPase activity unresponsive to stimulation by the DnaJ homologue Hdj-1 and inhibited Hdj-1-dependent refolding of denatured luciferase (23). Neither of these results necessarily implies that Hdj-1 did not interact with Hsp70. First, some of the mutants already had an elevated ATPase activity which could well have been incapable of further stimulation by Hdj-1. In addition, stimulation of the ATPase could have become uncoupled from Hdj-1 binding, if one of the functions of the EEVD motif is to signal Hdj-1 binding to the ATPase domain. The failure of Hsp70 mutants with defective EEVD motifs to refold luciferase efficiently also could equally result from downstream effects unrelated to the interaction with Hdj-1.
The effects of EEVD mutations and truncations of Hsc70 on substrate binding suggest that the segments between residue Asn 540 and the EEVD motif may be involved in inter-domain interactions. EEVD mutants were shown to have low affinity for substrates, but the additional elimination of the ATPase domain restores the high affinity. Similarly, the removal of the carboxyl-terminal 10-kDa domain, which includes the EEVD motif, by chymotrypsin digestion or recombinant techniques, does not inhibit substrate binding (4,21,23,30). Thus we may conjecture that the segment between Asn 540 and the EEVD motif interacts with the nucleotide-binding domain and that this interaction leads to inhibition of substrate binding. One function of the EEVD motif may be to alleviate this inhibition. This scheme affords an explanation for the increase of high substrate-binding affinity when either the nucleotide-binding domain or the 10-kDa domain which includes the EEVD motif is lost. The relatively high sequence divergence between the carboxyl-terminal domains of Hsp70 proteins suggests that their inter-domain communication might be differentially regulated by specific cellular factors.
Exclusion of any requirement of the carboxyl-terminal 10-kDa domain for the in vitro interaction of auxilin's J domain with Hsc70 and for dissociation of clathrin coats then raises the question of which of the other Hsc70 domains does interact with the J domain. Direct binding experiments demonstrated that only a fusion protein encompassing both the nucleotidebinding domain and the substrate-binding domains was capable of associating with the auxilin J domain. The failure of either the substrate-binding domain (Hsc70-(373-540)) or the nucleotide-binding domain by themselves to bind to the auxilin J domain suggests that segments from both domains may contribute to the J domain-binding site. Alternatively, interactions between the two domains could result in the exposure of an otherwise cryptic site for the J domain on one of them. The initial interaction of the auxilin J domain with Hsc70 requires the Hsp70 protein to be in the ATP-bound state. The transition of Hsp70 proteins from their ADP-bound state to the ATPbound state is known to be accompanied by large conformational rearrangements and monomerization of Hsp70 proteins, which can be detected by limited proteolysis, by fluorescence spectroscopy, and by small angle x-ray scattering (23,38,39,(41)(42)(43). The scattering data suggest a significant reduction in the radius of gyration of both the intact Hsc70 and the 60-kDa fragment. Although dissociation of the Hsp70 dimers could in principle account for the shape change, an ATP-induced internal rearrangement of domains was considered the more likely interpretation, because the 60-kDa fragment, unlike the intact protein, did not readily dimerize in the presence of ADP (43). , respectively. Track 7 shows the nucleotide composition of GST-Hsc70-(1-540) immediately after the removal of unbound nucleotide. Note, that even without the 5 min incubation at 25°C GST-Hsc70-(1-540) had already hydrolyzed more than 50% of its bound ATP.
In conclusion the available conformational data on Hsc70 are consistent with an ATP-induced inter-domain interaction leading to formation of a binding site for the J domain at the newly formed interface between the interacting Hsc70 domains.
At this stage it remains a matter of conjecture where on Hsp70 molecules the binding site might be. The junction between the substrate-and nucleotide-binding domains appears an attractive choice, because the J domain entering this region could easily be expected to affect both the ATPase and the helices which cover the substrate-binding channel. A model of the corresponding region of Hsc70, based on the calcium coordinates of DnaK (8), shows that the two structures are indeed very similar (Fig. 9). The model of the substrate-binding domain of Hsc70-(1-540) indicates that helix ␣B is now too short to cover the entire substrate-binding channel. While one might suppose that this truncation of the helix might weaken the association of Hsc70-(1-540) with its substrates, it nevertheless appeared to be without any severe consequences for the uncoating reaction.
Our assay was also able to detect a vestigial uncoating activity in bacterial DnaK. Although the auxilin J domain can certainly not be the natural partner for DnaK, the disadvantage of a poorly matching Hsp70 can clearly be partly offset in vitro by simply raising the concentration of the chaperone. On the other hand under the competitive conditions that prevail in the cell a less than perfect match between J domain and Hsp70 isoform would be expected to preclude an efficient interaction. DnaK appeared also to interact poorly with clathrin, since no complex could be demonstrated by gel filtration. Differences in the substrate specificity between Hsc70 and DnaK have been noted previously (3).
The work described here and our previous work have defined the domains of auxilin and Hsc70, respectively, that are essential for the uncoating of clathrin-coated vesicles. It is hoped that this information will be helpful for elucidating the regulatory mechanism that controls the uncoating reaction in the cell.  (8). The procedure of Peitsch (44) was used for constructing the model and the program Rasmol was used to visualize it (45). The arrows indicate substrate-binding channel of DnaK and Hsc70-(1-540), respectively. The substrate-binding channel of DnaK is occupied by a peptide. Note that the truncation of Hsc70 appears to increase the accessibility of the substrate-binding channel. The models do not include the nucleotide-binding domains, which would be located at the respective amino-terminal ends.