Prediction of Novel Bag-1 Homologs Based on Structure/Function Analysis Identifies Snl1p as an Hsp70 Co-chaperone in Saccharomyces cerevisiae *

Polypeptide binding by the chaperone Hsp70 is regulated by its ATPase activity, which is itself regulated by co-chaperones including the Bag domain nucleotide exchange factors. Here, we tested the functional contribution of residues in the Bag domain of Bag-1M that contact Hsp70. Two point mutations, E212A and E219A, partially reduced co-chaperone activity, whereas the point mutation R237A completely abolished activity in vitro. Based on the strict positional conservation of the Arg-237 residue, several Bag domain proteins were predicted from various eukaryotic genomes. One candidate, Snl1p from Saccharomyces cerevisiae, was confirmed as a Bag domain co-chaperone. Snl1p bound specifically to the Ssa and Ssb forms of yeast cytosolic Hsp70, as revealed by two-hybrid screening and co-precipitations from yeast lysate. In vitro, Snl1p also recognized mammalian Hsp70 and regulated the Hsp70 ATPase activity identically to Bag-1M. Point mutations in Snl1p that disrupted the conserved residues Glu-112 and Arg-141, equivalent to Glu-212 and Arg-237 in Bag-1M, abolished the interaction with Hsp70 proteins. In live yeast, mutated Snl1p could not substitute for wild-type Snl1p in suppressing the lethal defect caused by truncation of the Nup116p nuclear pore component. Thus, Snl1p is the first Bag domain protein identified in S. cerevisiae, and its interaction with Hsp70 is essential for biological activity.

Molecular chaperones of the heat shock protein 70-kDa (Hsp70) 1 family are conserved from Escherichia coli to mammals, and assist the folding of newly synthesized polypeptides as well as the refolding of proteins denatured under stress (1)(2)(3). All family members contain a ϳ44-kDa N-terminal ATPase domain and a ϳ25-kDa C-terminal peptide binding domain. Biochemically, the best characterized Hsp70 proteins are mammalian cytosolic Hsc70 (the constitutively expressed form) and DnaK of the E. coli cytosol. For both of these chaperones, the ATPase cycle has been shown to control the binding and release of substrate peptides by the C-terminal domain (2).
Hsp70 chaperones alternate between two nucleotide-bound states; the ATP-bound form has low affinity for substrate with high on-and off-rates of peptide binding, whereas the ADPbound state has high affinity for substrate with slow on-and off-rates (4 -7). Peptide is first bound by Hsp70 in the ATP state. Hydrolysis of ATP induces conformational rearrangements within the C-terminal peptide binding domain that encloses peptide between a lidlike helical subdomain and a ␤-sheet structure that forms a binding cleft (8,9). Exchange of ADP for a fresh ATP molecule then converts the peptide binding domain to an open state, allowing efficient release of the bound peptide (4,5).
The steady-state ATPase rate (0.02-0.2 min Ϫ1 ) of DnaK and mammalian Hsc70 in isolation is too slow for efficient chaperone function, although peptide binding alone may stimulate the ATPase rate 2-10-fold (7, 10 -13). The ATPase rate of DnaK is strongly stimulated by the activity of two co-chaperone proteins, DnaJ and GrpE. DnaJ stimulates the rate of ATP hydrolysis by DnaK up to several hundredfold, so that exchange of ADP for ATP becomes the rate-limiting step in the ATPase cycle. GrpE reduces the affinity of DnaK for ADP ϳ200-fold and thereby accelerates the nucleotide exchange rate up to 5000-fold. Together, DnaJ and GrpE produce a more than 50-fold increase in the steady-state ATPase rate of DnaK, greatly increasing the protein folding activity of the system (6, 11, 14 -16).
In the eukaryotic cytosol, the Hsp40 family of co-chaperones are homologs of DnaJ and act similarly to stimulate ATP hydrolysis by Hsc70 (17)(18)(19). No homologs of GrpE have been found in the eukaryotic cytosol; consequently, nucleotide exchange factors for Hsp70 in this cellular compartment were thought to be unnecessary because of a sufficiently fast spontaneous exchange rate (12,18). However, a nucleotide exchange stimulatory function has been suggested for the mammalian cytosolic protein Bag-1 (20).
Bag-1 was first identified as an interaction partner of the apoptosis inhibitor protein Bcl-2 (21), and was later shown to directly interact with the N-terminal ATPase domain of Hsc70 (22,23). Importantly, Bag-1 stimulates the Hsc70 ATPase rate only in the presence of Hsp40, suggesting that it acts at the step of nucleotide exchange and not ATP hydrolysis (20). Although Bag-1 was originally characterized as an inhibitor of the protein refolding activity of Hsc70 in vitro and in vivo (23)(24)(25), substoichiometric levels of Bag-1 relative to Hsc70 can in fact enhance Hsc70 chaperone activity (19).
Bag-1 exists as several isoforms generated by alternative translation initiation, all of which share the C-terminal sequences that mediate binding to Hsc70 (26). All of the Bag-1 isoforms regulate the Hsc70 ATPase activity similarly, but the originally identified isoform Bag-1M is much less effective at stimulating Hsc70-mediated protein refolding than the shorter and longer isoforms Bag-1S and Bag-1L (27). A family of cochaperone proteins has been identified having sequences homologous to the C-terminal domain of Bag-1, the so-called Bag domain, which mediates binding to Hsc70 (26). However, unlike the Hsp40 co-chaperone family, the Bag domain proteins appear to be more specialized and it is not yet clear if they are universally present in eukaryotes.
We recently solved the crystal structure of the ATPase domain of Hsc70 in complex with the ϳ13-kDa Bag domain of Bag-1 (28). Together with recent biophysical measurements and nuclear magnetic resonance experiments (29,30), the structure clearly established Bag-1 as a nucleotide exchange factor for Hsp70. Binding of the Bag domain induces a conformational change in Hsp70 identical to that induced in the bacterial Hsp70 chaperone DnaK by the co-chaperone GrpE. However, the Bag domain is structurally unrelated to GrpE; although the Bag domain binds Hsc70 as a monomeric threehelix bundle, GrpE is a homodimer of ϳ20-kDa subunits joined by a long ␣-helical stretch, and a subdomain formed largely by ␤-strands contacts DnaK (28,31). Thus, the residues of Bag-1 involved in Hsp70 binding are different in both identity and structural position from their counterparts in GrpE.
Here, we identify key functional residues of Bag-1 by structure-based mutagenesis. Mutant proteins show significantly reduced affinity for Hsp70, impaired stimulation of the Hsp70 ATPase, and a decreased ability to dissociate Hsp70-substrate complexes. Structural alignments and positioning of conserved functional residues revealed the existence of putative Bag proteins in various eukaryotic genomes. Biochemical analysis of the putative Bag protein Snl1p from Saccharomyces cerevisiae confirmed these predictions and established Snl1p as a nucleotide exchange factor for Hsp70 proteins in yeast. The Hsp70regulatory function of Snl1p was found to be essential for its activity in vivo.
Isothermal Titration Calorimetry-Binding constants (K D ) in solution and stoichiometries of complex formation between Bag-1 and Hsc70 proteins were measured by isothermal titration calorimetry using a MicroCal MCS calorimeter (MicroCal Inc., Northampton, VA). Measurements were performed at 25°C by titration of 100 M Hsc70-N with 1 mM various Bag-1M proteins in buffer G, using serial 10-l injections and a 400-s interval between injections. K D values were determined by integrating heats of binding normalized to the amount of injected protein, and the data were fitted to a 1:1 binding model using the Origin software package.
ATPase Activity Assays-To determine ATP hydrolysis rates, reactions containing the various proteins at 3 M each in 50 mM KCl, 20 mM HEPES-KOH, 5 mM MgAc 2 , 2 mM ATP, pH 7.5, and 1 Ci of [␣-32 P]ATP (400 Ci/mmol; Amersham Biosciences) were incubated at 30°C. Samples from time points in the linear range of the assay were analyzed by thin layer chromatography on polyethyleneimine-cellulose (Merck) developed in 0.5 M formic acid and 0.5 M LiCl. The amounts of ADP produced were quantified by phosphorimager analysis on a Fuji Film FLA-2000 using the MacBas software package and ATPase rates calculated.
Substrate Release Assays-The release of Hsc70 from chaperonesubstrate complexes was assayed as described (32). Briefly, purified Myc-tagged LBD was denatured in 1% SDS and 50 mM Tris-Cl, pH 7.5, exchanged into buffer containing 1% Nonidet P-40, 1% sodium deoxycholate, 0.1% SDS, 100 mM KOAc, and 25 mM HEPES-KOH, pH 7.5, then recovered with anti-Myc monoclonal antibodies covalently coupled to Protein G-Sepharose (Amersham Biosciences). Immune pellets were washed with 5% glycerol, 100 mM KOAc, and 25 mM HEPES-KOH, pH 7.5 (buffer B). Chaperone binding was initiated by adding 50% desalted reticulocyte lysate (Green Hectares) in buffer B containing 5 mM MgOAc 2 , 2 mM ATP, and 35 S-labeled Hsc70 produced by in vitro translation using the TNT T7 system (Promega). After 10 min at room temperature, immune pellets were recovered and washed twice with buffer B. Release of Hsc70 from the immune pellets was assayed by immediate resuspension in buffer B containing 5 mM MgOAc 2 and additions as indicated followed by a 10-min incubation at room temperature. The beads and supernatant were separated and proteins released into the supernatants precipitated with trichloroacetic acid. The amount of released and bound radiolabeled Hsc70 was analyzed by SDS-PAGE and phosphorimager quantitation.
Two-hybrid Screen-The two-hybrid host strain PH69 -4A harboring pSW584, encoding residues 36 -160 of Snl1p fused to the Gal4p-DNA binding domain (Gal4bd) in pGBT8, was transformed with three pools of an S. cerevisiae genomic library containing sequences fused to the Gal4p activation domain (35). Expression of the Gal4bd-Snl1 fusion protein in PH69 -4A was verified with immunoblotting using antibodies against Snl1p. The Gal4bd-Snl1p alone, or co-transformed with a plasmid encoding the control protein Snf1p fused to the Gal4p activation domain, did not activate the GAL7-lacZ reporter gene. Approximately 9 ϫ 10 6 transformants of library pool pGad-C1, 1.5 ϫ 10 6 of pGad-C2, and 1.4 ϫ 10 6 of pGad-C3 were screened for a coverage that included all possible candidates, within a 99% confidence interval. Colonies were sequentially screened for the three reporter genes HIS3, ADE2, and lacZ, and library plasmids were isolated from positive colonies. The specificity of the Snl1p interaction was verified by lack of interaction with the control protein lamin C fused to Gal4bd. 30 positive isolates were obtained, and 12 analyzed by DNA sequencing. In the SSB1 isolates where the fusion points with Gal4p activation domain could be determined, three were fused at residue 164 of Ssb1p, one at residue 202, two at residue 210, and one at residue 227.
Yeast Protein Binding-To prepare yeast cytosol, S. cerevisiae strain YPH499 (36) was grown in YPD at 30°C and harvested in mid-log phase. Cells were washed with water, resuspended in buffer G containing Complete protease inhibitor mix without EDTA (Roche), and lysed by agitation with glass beads. Lysates were centrifuged at 30,000 ϫ g for 10 min at 4°C, separated from the beads, and centrifuged at 100,000 ϫ g for 30 min at 4°C to remove membranes. For Snl1p pull-downs, 24 g of His 6 -tagged Snl1-C1 or Snl1-C1-E112A/R141A was bound to 12 l of nickel-NTA-agarose in buffer B containing 2 mg/ml ovalbumin (Sigma) for 30 min at room temperature, then the beads were recovered and washed with buffer B. The beads were resuspended in 400 l of 1 mg/ml yeast cytosol in buffer G containing 1 mg/ml ovalbumin, 20 mM imidazole, 0.1% Nonidet P-40, and protease inhibitors, either in the presence of 4 units/ml apyrase (Sigma), 5 mM ATP or ADP. Reactions were incubated for 30 min at 4°C and then washed with buffer G containing 20 mM imidazole and 0.1% Nonidet P-40, or the same buffer with 5 mM ATP or ADP in parallel to the binding conditions. Bound proteins were eluted with 25 mM EDTA and analyzed by SDS-PAGE and either Coomassie Blue staining or immunoblotting. Antibodies recognizing the C-terminal 80 and 56 amino acids of Ssa1p and Ssb1p, respectively, were a generous gift of E. A. Craig (Madison, WI) (37). Blots were developed using goat anti-rabbit antibodies conjugated to horseradish peroxidase (Sigma), and the ECL detection system (Amersham Biosciences). SNL1 Complementation-Yeast strain SWY2000 (MAT␣ nup116 -5::HIS3 ade2-1 ura3-1 his3-11,15 trp1-1 leu2-3,112 can1-100 pSW171 (GAL/GST-Nup116-C)) (34) was transformed with pSW807 (yeast library plasmid encoding SNL1), pSW534 (GAL/SNL1), or p426GAL/Snl1-E112A/R141A. Cells were grown in media for 3 days at 23°C in medium with 2% glucose, washed with medium containing 2% galactose, adjusted to identical cell densities, serially diluted in 5-fold steps, and seeded on medium containing either 2% glucose or 2% galactose. Growth was observed after 3-4 days.
Computational Methods-Representations of the Bag-C structure and its properties were obtained with the programs LIGPLOT, BOB-SCRIPT, and RASTER3D (38 -40). The buried surface within the Bag-C-Hsc70-N complex structure was calculated using CNS (41). The conserved residues on the molecular surface of the Bag domain were mapped with ConSurf (42) using a CLUSTAL W (43) alignment (Fig. 4) as input. Surface representations were generated using GRASP (44). Alignments were carried out with PSI-BLAST (45).

Mutational Analysis of the Bag-1M-Hsc70
Interaction-We recently solved the crystal structure of the Hsc70 ATPase domain (Hsc70-N) in complex with the C-terminal Bag domain of Bag-1M (Bag-C) at 1.9-Å resolution (28). Based on this struc-ture, a mechanism for the nucleotide exchange activity of Bag domain proteins on eukaryotic cytosolic Hsp70 proteins has been proposed. The interaction with Bag-1M induces a conformational change in Hsc70 identical to the change in DnaK, the major bacterial Hsp70, induced by its specific nucleotide exchange factor GrpE (31). However, the contacts between the Bag domain and Hsc70 are substantially different from those between GrpE and DnaK, in both the position and nature of the residues involved. We therefore undertook a mutational analysis of Bag-1M to test the importance of specific residues for the functional interaction with Hsc70, and to aid in the identification of novel Bag domain proteins.
The structure of the Bag-C-Hsc70-N complex is shown in Fig.  1A, and a schematic representation of the contact residues in Fig. 1B. Among other contact points, Glu-212 and Glu-219 of Bag-1M interact extensively with Arg-261 and Met-61 of Hsc70. In addition, Arg-237 of Bag-1M forms a close contact with Glu-283 and Tyr-294 of Hsc70. Interestingly, although Glu-212 and Arg-237 of Bag-1M are absolutely conserved in other known Bag domain proteins, a leucine or isoleucine is usually found in the place of Glu-219 (28). Therefore, to test the functional importance of these residues, the point mutations E212A, E219A, and R237A were introduced into the sequence of Bag-C. The mutant Bag-C proteins were purified, and their binding affinity for Hsc70-N was measured using isothermal titration calorimetry.
Full-length Bag-1M and the Bag-C domain had similar affinities for Hsc70-N (K D values of 2.8 and 3.6 M, respectively) ( Fig. 2A). As predicted from the structure, the mutations E212A and E219A in Bag-C resulted in a substantial reduction in binding affinities for Hsc70-N (K D values 31 and 33 M, respectively), and the mutation R237A affected the binding more strongly, showing a more than 10-fold reduction in affin- Reductions in the affinity of the Bag domain for Hsc70 should result in a reduced regulation of the Hsc70 ATPase activity. Therefore, the mutant Bag-C proteins were tested for their activity as nucleotide exchange factors for Hsc70, measured by stimulation of the Hsc70 steady-state ATPase rate under conditions where ADP-ATP exchange is rate-limiting. These conditions are met when ATP hydrolysis by Hsc70 is stimulated to its maximum rate by Hsp40, in which case the further addition of Bag-C significantly increased the overall ATPase rate roughly 10-fold (Fig. 2B) similar to full-length Bag-1M (20,28). Under identical conditions, the E212A and Glu-219 Bag-C mutants stimulated the ATPase activity only poorly, whereas the R237A mutant showed no significant ATPase stimulation (Fig. 2B). Importantly, the ATPase stimulation observed with the E212A and E219A Bag-C mutants was absolutely dependent on Hsp40, consistent with a residual activity of these proteins as nucleotide exchange factors.
Another test of Bag-1M function is its ability to dissociate Hsc70-substrate complexes in an ATP-dependent manner, by favoring the ATP-bound state of Hsc70 with low affinity for peptide (27). This activity was determined using the LBD of rat glucocorticoid receptor as a model polypeptide substrate. Chaperone-LBD complexes containing radiolabeled in vitro translated Hsc70 were immune-isolated from rabbit reticulocyte lysate, and the ATP-dependent release of Hsc70 from the LBD was assayed (32). In the presence of ATP, Bag-C produced the maximum level of Hsc70 release, comparable with full-length Bag-1M (Fig. 3A) (28). In agreement with their reduced stimulation of the Hsc70 ATPase, the E212A and E219A Bag domain mutants mediated only a reduced level of Hsc70 release from substrate. The R237A mutant, which was unable to stimulate the Hsc70 ATPase, did not support release of Hsc70 from substrate above control reactions containing a nonspecific protein or performed in the absence of ATP (Fig. 3A).
To further confirm that the substrate release assay reflects the specific interaction of Bag domain proteins with Hsc70, the point mutation R261A was introduced into Hsc70 to disrupt the predicted contact with Glu-212 and Glu-219 of Bag-1M. Radiolabeled Hsc70-R261A bound LBD at a level similar to wild-type Hsc70 (data not shown) but was released from substrate complexes inefficiently by the wild-type Bag-C domain in response to ATP (Fig. 3B). As expected, wild-type Bag-C was approximately as effective as the E212A and E219A mutants at promoting release of Hsc70-R261A, and the R237A Bag-C mutant produced no release of Hsc70-R261A above background (Fig.  3B).
These results demonstrate that residues Glu-212, Glu-219, and Arg-237 of Bag-1M not only contact the Hsc70 ATPase domain, but are important for stabilizing the interaction and efficiently regulating the Hsc70 ATPase activity. The absolutely conserved Arg-237 is particularly important, as mutation of the residue to alanine leads to a complete loss of Bag domain function in all of the assays. Additionally, it is likely that Glu-212 and Glu-219 of Bag-1M contribute redundantly to the interaction with Arg-261 of Hsc70. This would explain the slightly higher affinity of the E212A and E219A mutants for Hsc70 compared with the R237A mutant, and the residual activity of the E212A and E219A mutants in the biochemical assays. In Bag domain proteins where Glu-219 is not con- served, a conserved glutamate residue equivalent to Glu-212 is always found and probably provides the main contact with Arg-261 of Hsc70. Thus, in the search for novel Bag domain proteins, a residue equivalent to Arg-237 of Bag-1M is an essential criterion for Bag domain function, with a residue equivalent to Glu-212 as a secondary criterion.

Identification and Characterization of Snl1p as a Novel Bag Domain Protein-Prediction from genomic data bases of novel
Bag domain co-chaperones was complicated by the relatively poor similarity between known Bag proteins. Therefore, in addition to the homology-based alignment of primary sequences with the Bag domain of Bag-1M using PSI-BLAST (45), the criterion of identity in residues aligned with Arg-237 of Bag-1M was applied. A number of uncharacterized potential Bag domain proteins were identified by this method (Fig. 4), whereas several other proteins having higher overall similarity to Bag-1M were rejected. As expected, putative Bag domain proteins were found only in eukaryotic genomes, and in none of the prokaryotic genomes searched. These predicted Bag domain proteins were from widely diverse species, including Neurospora crassa, S. cerevisiae, Cicer arietinum, Arabidopsis thaliana, Bombyx mori, and Drosophila melanogaster (Fig. 4).
In the sequences of these proteins, several residues in addition to those aligned with Arg-237 of Bag-1M were strictly conserved, including residues equivalent to Glu-212, Asp-222, Lys-238, and Gln-245 of Bag-1M, all of which contact Hsc70 in the crystal structure (Figs. 1B and 4). Other highly conserved residues such as those equivalent to Leu-218 of Bag-1M (Fig. 4) are mainly hydrophobic and are involved in the internal packing of the Bag domain. Beyond these residues, the homology to the Bag domain of Bag-1M is relatively poor, below 22% identity and 50% similarity for all of the selected proteins. However, when the positional conservation between the sequences is mapped onto the surface of the Bag domain structure using the ConSurf algorithm (42), the highest degree of conservation is located on the Hsc70 binding face calculated from the separation between the Bag-1M and Hsc70 surfaces in the crystal structure (Fig. 5). Consistent with a Bag domain function of these proteins, the cytosolic Hsc70 partners are almost absolutely conserved in the residues predicted to interact with Bag domains. Notably, mitochondrial and chloroplast Hsp70 proteins, which are not expected to interact with cytosolic Bag  domains, are divergent in these residues. Thus, the proteins identified here are suggested to function as Bag domain proteins by their predicted structural properties as well as their sequence homology.
To experimentally test our prediction, the candidate Bag domain protein Snl1p from S. cerevisiae was selected, because of the established biochemical and genetic characterizations of Hsp70 chaperones in this species (37,46). SNL1 interacts genetically with components of the nuclear pore, but ⌬snl1 strains show no obvious phenotype. The 159-residue Snl1p protein has a single transmembrane domain at the extreme N terminus and is localized on the nuclear and endoplasmic reticulum membranes with the C terminus and most of the protein sequence facing the cytosol (34). The Bag domain homology region of Snl1p lies within the cytosolic C-terminal domain, in agreement with an interaction with cytosolic Hsp70.
To identify possible interactions between Snl1p and yeast Hsp70, we began with an unbiased approach by using the cytosolic portion of Snl1p as a bait in a two-hybrid screen. The complete-coverage screen against an S. cerevisiae library returned 30 positive isolates, and of the 12 that were analyzed by sequencing, 7 were identified as SSB1, 2 as SSA1, and 1 as SSA4. These correspond to the constitutive and one of the inducible forms of cytosolic Hsp70 (SSA1 and SSA4, respectively), which are homologous to those in other eukaryotes, and a cytosolic Hsp70 found so far only in yeast (SSB1), which exists as both ribosome-bound and free populations (46). Two isolates encoded SSQ1, an Hsp70 of the mitochondria matrix (47), and represented a biologically irrelevant interaction. Further sequencing of the SSB1 isolates revealed that all contained the complete C-terminal lobe of the ATPase domain (see "Experimental Procedures"), where many of the predicted contact sites with a Bag domain are located. Because no interacting proteins other than the Hsp70 proteins were found, these data supported the hypothesis that Snl1p functions mechanistically as a Bag domain co-chaperone.
As a direct test of this hypothesis, two protein fragments containing the Snl1p cytosolic domain were expressed in bacteria and purified. The resulting proteins contained residues 40 -159 of Snl1p (Snl1-C1), the entire cytosolic portion lacking the transmembrane anchor, and residues 49 -159 (Snl1-C2), lacking the anchor plus a short segment outside the region of Bag domain homology. We first confirmed the two-hybrid interactions with the Hsp70 chaperones by identifying proteins in a yeast lysate which could bind to the purified Snl1p cytosolic domain. Because both ATP and ADP have been shown to greatly reduce the affinity of mammalian Bag-1M for Hsc70 (28), binding was performed either after nucleotide depletion, or in the presence of 5 mM ATP or ADP.
Purified His 6 -tagged Snl1-C1 adsorbed onto nickel-NTA-agarose was incubated with a yeast cytosol extract, and proteins recovered with the beads were detected by Coomassie staining. In the reaction depleted of nucleotides with apyrase treatment, proteins with an apparent molecular mass of ϳ70 kDa coprecipitated with the Snl1p fragment (Fig. 6). Binding of these proteins to Snl1-C1 was strongly inhibited by the addition of ATP, and only a small amount was recovered after ADP addition (Fig. 6). The apparent size of the co-precipitated ϳ70-kDa proteins and their behavior in the presence of nucleotide strongly suggested that they were Hsp70 proteins. The identity of these proteins was tested by immunoblotting with antibodies specific for either the Ssa or Ssb proteins, and consistent with the two-hybrid results, both types of Hsp70 were identified in the fraction bound to Snl1-C1 (Fig. 6). Binding of the Hsp70 proteins was reduced to background levels by mutation to alanine of the conserved residues Glu-112 and Arg-141 of Snl1-C1, equivalent to Glu-212 and Arg-237 of human Bag-1M (Fig.  6). Thus, Snl1p binds yeast cytosolic Hsp70 identically to the mammalian Bag-1M-Hsc70 interaction.
To demonstrate the biochemical activity of Snl1p as a Bag domain co-chaperone, we applied the same assays used to characterize the wild-type and mutant mammalian Bag domain proteins. Because of the high conservation between mammalian and yeast cytosolic Hsp70 proteins (ϳ74% identity), it was likely that yeast Snl1p would interact functionally with mammalian Hsc70. In any case, divergence between mammalian and yeast Hsp70 proteins would bias our experiments against a positive result. However, calorimetric analysis showed that the affinities of Snl1-C1 and Snl1-C2 for Hsc70-N were similar to that of Bag-1M, each having dissociation constants of ϳ5 M and ϳ1:1 stoichiometries of binding (data not shown). The Snl1p cytosolic fragments were next tested for nucleotide exchange activity on mammalian Hsc70. In the presence of Hsp40, where nucleotide exchange by Hsc70 is rate-limiting for the ATPase activity (see Fig. 2), the addition of Snl1-C1 and Snl1-C2 strongly stimulated the Hsc70 ATPase rate (Fig. 7A). The effect was comparable with that seen with mammalian Bag-C (Fig. 7A) and full-length Bag-1M (28), and was strictly dependent on Hsp40 (Fig. 7A). Finally, we tested the ability of Snl1-C1 and Snl1-C2 to dissociate Hsc70-substrate complexes in the expected ATP-dependent manner (see Fig. 3). When immune-isolated complexes of LBD bound to reticulocyte lysate chaperones were treated with ATP, both of the Snl1p cytosolic fragments induced efficient release of Hsc70 from the complexes, at a level similar to that obtained with the mammalian Bag domain (Fig. 7B). Again as expected, Hsc70 dissociation was stringently dependent on ATP (Fig. 7B).
Taken together, our results clearly identify Snl1p as a Bag domain co-chaperone interacting with cytosolic Hsp70 proteins in S. cerevisiae, the first such co-chaperone to be characterized in this species. To establish whether the co-chaperone activity of Snl1p is involved in its biological function, we made use of the previous experiments implicating Snl1p in nuclear pore function. In S. cerevisiae, NUP116 encodes a large nucleoporin of the GLFG-repeat family (48,49). Expression of a C-terminal fragment of the protein (Nup116-C) is lethal in ⌬nup116 cells, which are otherwise viable, although temperature-sensitive. SNL1 was first identified in a screen for multicopy suppressors of this lethality and was also found to suppress temperaturesensitive mutants of the nuclear pore genes GLE2 and NIC96 (34).
Because the E112A/R141A point mutations essentially abolished the binding between the Snl1p cytosolic domain and the Hsp70 proteins, full-length Snl1p mutated in the same residues was tested for the ability to suppress the nup116-C phenotype. ⌬nup116 yeast with a plasmid encoding galactoseinducible Nup116-C were viable on glucose, and growth on galactose was supported by co-expression of wild-type Snl1p under a native or galactose-inducible promoter (Fig. 8). However, cells with plasmids for galactose-inducible co-expression of the mutant Snl1-E112A/R141A with Nup116-C were inviable on galactose (Fig. 8), similar to cells expressing Nup116-C alone. Galactose induction of both wild-type and mutant Snl1p was observed in cells without the nup116-C plasmid (data not shown). Thus, Hsp70 regulation by the Bag domain of Snl1p is apparently essential for its function in vivo.

DISCUSSION
The results presented here establish the S. cerevisiae protein Snl1p as the first known Bag domain nucleotide exchange factor for cytosolic Hsp70 proteins in this organism. Because Snl1p is not required for the normal function of the Hsp70 chaperone system and is localized to the nuclear and endoplasmic reticular membranes, it is likely to be involved only in specific cellular processes. One of these processes may be nu-clear pore biogenesis, suggesting a novel cellular function of the Hsp70 system.
The approach taken in this study began with structure-based mutagenesis to test the functional importance of selected residues of Bag-1M, and the characterization of residue Arg-237 as essential for Hsp70 regulation allowed a more accurate prediction of Bag domain proteins than sequence alignment alone. This approach was validated by the successful identification of Snl1p as a Bag domain co-chaperone. As further support for our approach, the predicted Bag domain protein from B. mori (Fig. 4) has recently been identified as the Bag domain co-chaperone Samui (i.e. cold-shock inducible) (50). Interestingly, in the only known exception to the rule, Bag-2 from humans and C. elegans has a glutamine in place of Arg-237 in Bag-1M (51). It is possible that our search for Bag domain proteins overlooked actual family members that also have a variant architecture. Nevertheless, a review of the sequences discarded by the search uncovered no further S. cerevisiae proteins in which other important Bag domain residues, such as Glu-212 and Asp-222 of Bag-1M, were conserved. Therefore, Snl1p is in all likelihood the only Bag domain cochaperone in S. cerevisiae.
Snl1p interacts biochemically with both the Ssa and Ssb classes of yeast Hsp70 chaperones in the expected nucleotidedependent manner (Fig. 6). These two Hsp70 classes are distinct not only in their primary sequence, but also in their biochemical properties and cellular functions. Similar to mammalian Hsc70, Ssa1p has a reasonably strong affinity for ATP (K m value of 0.11 M) and its ATPase rate is stimulated roughly 10-fold by J domain co-chaperones. In contrast, Ssb1p has an unusually poor K m (147 M), is unaffected by J domain cochaperones, and is largely bound to translating ribosomes, although there is also a free pool in the cytosol (37,46). Snl1p may act primarily on Ssa chaperones for the following reasons. First, Snl1p should possess an Hsp70-binding face similar to the mammalian Bag domain (Fig. 5) and may bind more stably to Ssa, which is identical to Hsc70 in all of its contact residues, than to Ssb, which has several conservative substitutions. Second, stimulation of ADP-ATP exchange should have a significant effect on the ATPase rate of Ssa because of its high affinity for nucleotide, whereas the affinity of Ssb for nucleotide may be too low to be modulated by Snl1p. Third, the population of Ssb tightly bound to ribosomes may not be able to contact Snl1p for steric reasons. If, as has been suggested, Ssb chaperones act at earlier stages in the folding of nascent polypeptide chains (46), then Snl1p regulation of Ssa chaperones would more likely influence later stages of protein folding. However, biological interactions between Snl1p and non-ribosomal Ssb may still play a role in some folding processes.
The co-chaperone activity of Snl1p suggests an involvement of Hsp70 in suppression of the nup116-C lethality. This is strongly supported by the behavior of the E112A/R141A mutant Snl1p, which is unable to bind Hsp70 (Fig. 6) and to complement the nup116-C phenotype (Fig. 8). Interestingly, Snl1p overexpression could not suppress the temperature-sensitive phenotype of the ⌬nup116 strain, suggesting that it must operate on the Nup116-C protein fragment (34). The C-terminal region of Nup116p is responsible for its targeting to nuclear pores and association with Nup82p and other components of a nucleoporin subcomplex (52)(53)(54). A model may be proposed where newly synthesized Nup116-C is assembled into nucleoporin complexes in the absence of wild-type Nup116p, but remains only partially folded, and localized stimulation of Hsp70 by Snl1p allows Nup116-C to reach and maintain a functional folded state. Similarly, multicopy Snl1p expression suppressed the temperature-sensitive gle2-1 mutation but not ⌬gle2 (28), again suggesting that Hsp70-mediated folding of the mutant Gle2p nuclear export factor could be responsible for the suppression. Although direct contacts between Snl1p and the Nup116-C or Gle2-1 proteins cannot be entirely ruled out, such interactions would not normally occur as Snl1p is not part of the mature wild-type nuclear pore complexes (55). Moreover, two-hybrid interactions between the C-terminal region of Snl1p and Nup116-C or Gle2p have not been detected. 2 It may be proposed that, under normal conditions, Snl1p aids in the Hsp70-mediated assembly of similar multiprotein complexes on nuclear and endoplasmic reticulum membranes.
Targeting of Bag domains for specific cellular purposes is a common characteristic of this co-chaperone family. Most Bag domain proteins identified so far, whether by biochemical analysis or sequence alignment, have the Hsp70 regulatory element fused to other features which provide additional specific interactions. The originally discovered mammalian Bag-1M contains a ubiquitin-like domain required for the association of Bag-1M with the proteasomal degradation machinery (27). The N terminus of Bag-1M also contains a short DNA-binding sequence involved in transcriptional modulation (56), and the longer isoform Bag-1L has a nuclear localization sequence. Human Bag-6 also contains a ubiquitin-like domain, Bag-3 has a WW repeat region, and Bag-4 interacts specifically with apoptosis-regulatory death domains (26,51). Interestingly, the recently described Bag-1B protein of Schizosaccharomyces pombe (51) contains a hydrophobic sequence near its N terminus, which could serve as a membrane anchor, and may be a direct functional homolog of Snl1 not only in Bag domain cochaperone activity but also in the cellular processes where it acts. Further study of such targeted co-chaperone proteins will expand our understanding of both the Hsp70 chaperone system and the cellular processes dependent on it.