Interactions between Kar2p and Its Nucleotide Exchange Factors Sil1p and Lhs1p Are Mechanistically Distinct

Kar2p, an essential Hsp70 chaperone in the endoplasmic reticulum of Saccharomyces cerevisiae, facilitates the transport and folding of nascent polypeptides within the endoplasmic reticulum lumen. The chaperone activity of Kar2p is regulated by its intrinsic ATPase activity that can be stimulated by two different nucleotide exchange factors, namely Sil1p and Lhs1p. Here, we demonstrate that the binding requirements for Lhs1p are complex, requiring both the nucleotide binding domain plus the linker domain of Kar2p. In contrast, the IIB domain of Kar2p is sufficient for binding of Sil1p, and point mutations within IIB specifically blocked Sil1p-dependent activation while remaining competent for activation by Lhs1p. Taken together, these results demonstrate that the interactions between Kar2p and its two nucleotide exchange factors can be functionally resolved and are thus mechanistically distinct.     

Structural basis for the cooperation of Hsp70 and Hsp110 chaperones in protein folding

Protein folding by Hsp70 is tightly controlled by cochaperones, including J-domain proteins that trigger ATP hydrolysis and nucleotide exchange factors (NEFs) that remove ADP from Hsp70. Here we present the crystal structure of the yeast NEF Sse1p (Hsp110) in complex with the nucleotide-binding domain (NBD) of Hsp70. Hsp110 proteins are homologous to Hsp70s and consist of an NBD, a beta sandwich domain, and a three helix bundle domain (3HBD). In the complex, the NBD of Sse1p is ATP bound, and together with the 3HBD it embraces the NBD of Hsp70, inducing opening and the release of bound ADP from Hsp70. Mutations that abolish NEF activity are lethal, thus defining nucleotide exchange on Hsp70 as an essential function of Sse1p. Our data suggest that Sse1p does not employ the nucleotide-dependent allostery and peptide-binding mode of canonical Hsp70s, and that direct interactions of substrate with Sse1p may support Hsp70-assisted protein folding in a cooperative process.     

Nucleotide Binding by Lhs1p Is Essential for Its Nucleotide Exchange Activity and for Function in Vivo

Protein translocation and folding in the endoplasmic reticulum of Saccharomyces cerevisiae involves two distinct Hsp70 chaperones, Lhs1p and Kar2p. Both proteins have the characteristic domain structure of the Hsp70 family consisting of a conserved N-terminal nucleotide binding domain and a C-terminal substrate binding domain. Kar2p is a canonical Hsp70 whose substrate binding activity is regulated by cochaperones that promote either ATP hydrolysis or nucleotide exchange. Lhs1p is a member of the Grp170/Lhs1p subfamily of Hsp70s and was previously shown to function as a nucleotide exchange factor (NEF) for Kar2p. Here we show that in addition to this NEF activity, Lhs1p can function as a holdase that prevents protein aggregation in vitro. Analysis of the nucleotide requirement of these functions demonstrates that nucleotide binding to Lhs1p stimulates the interaction with Kar2p and is essential for NEF activity. In contrast, Lhs1p holdase activity is nucleotide-independent and unaffected by mutations that interfere with ATP binding and NEF activity. In vivo, these mutants show severe protein translocation defects and are unable to support growth despite the presence of a second Kar2p-specific NEF, Sil1p. Thus, Lhs1p-dependent nucleotide exchange activity is vital for ER protein biogenesis in vivo.     

Functional divergence between co-chaperones of Hsc70

The ATPase cycle of the chaperone Hsc70 is regulated by co-chaperones; Hsp40/DnaJ-related proteins stimulate ATP hydrolysis by Hsc70 and can bind unfolded polypeptides themselves. Conversely, various nucleotide exchange factors (NEFs) stimulate ADP-ATP exchange by Hsc70. We analyzed the purified Hsp40-related co-chaperones DJA1 (Hdj2) and DJA2 (Hdj3) and found that they had a distinct pattern of binding to a range of polypeptides. DJA2 alone could stimulate Hsc70-mediated refolding of luciferase in the absence of NEF, whereas DJA1 was much less active. The addition of the Bag1 NEF increased refolding by Hsc70 and DJA2, as did the newly characterized NEF Hsp110, but each NEF had a different optimal concentration ratio to Hsc70. Notably, the NEF HspBP1 could not increase refolding by Hsc70 and DJA2 at any concentration, and none of the NEFs improved the refolding activity with DJA1. Instead, DJA1 was inhibitory of refolding with DJA2 and Hsc70. All combinations of DJA1 or DJA2 with the three NEFs stimulated the Hsc70 ATPase rate, although Hsp110 became less effective with increasing concentrations. A chimeric DJA2 having its Hsc70-stimulatory J domain replaced with that of DJA1 was functional for polypeptide binding and ATPase stimulation of Hsc70. However, it could not support efficient Hsc70-mediated refolding and also inhibited refolding with DJA2 and Hsc70. These results suggest a more complex model of Hsc70 mechanism than has been previously thought, with notable functional divergence between Hsc70 co-chaperones.     

The nucleotide exchange factors Grp170 and Sil1 induce cholera toxin release from BiP to enable retrotranslocation

Cholera toxin (CT) intoxicates cells by trafficking from the cell surface to the endoplasmic reticulum (ER), where the catalytic CTA1 subunit hijacks components of the ER-associated degradation (ERAD) machinery to retrotranslocate to the cytosol and induce toxicity. In the ER, CT targets to the ERAD machinery composed of the E3 ubiquitin ligase Hrd1-Sel1L complex, in part via the activity of the Sel1L-binding partner ERdj5. This J protein stimulates BiP''s ATPase activity, allowing BiP to capture the toxin. Presumably, toxin release from BiP must occur before retrotranslocation. Here, using loss-and gain-of-function approaches coupled with binding studies, we demonstrate that the ER-resident nucleotide exchange factors (NEFs) Grp170 and Sil1 induce CT release from BiP in order to promote toxin retrotranslocation. In addition, we find that after NEF-dependent release from BiP, the toxin is transferred to protein disulfide isomerase; this ER redox chaperone is known to unfold CTA1, which allows the toxin to cross the Hrd1-Sel1L complex. Our data thus identify two NEFs that trigger toxin release from BiP to enable successful retrotranslocation and clarify the fate of the toxin after it disengages from BiP.     

Isolation of proteins that speifically interact with the ATPase domain of mammalian ER chaperone,BiP

BiP, immunoglobulin binding protein, is an ER homologue of Hsp 70. However, unlike other Hsp70 proteins, regulatory protein(s) for BiP has not been identified. Here, we demonstrated the presence of potential regulatory proteins for BiP using a pull-down assay. Since BiP can bind any unfolded protein, only the ATPase domain of BiP was used for the pull-down assay in order to minimize nonspecific binding. The ATPase domain was cloned to produce recombinant protein, which was then conjugated to CNBr-activated agarose. The structural conformation and ATP hydrolysis activity of the recombinant ATPase domain were similar to those of the native protein. Eight proteins from metabolically labeled mouse plasmacytoma cells specifically bound to the recombinant ATPase protein. The binding of these proteins was inhibited by excess amounts of free ATPase protein, and was dependent on the presence of ATP. These proteins were eluted by ADP. Of these proteins, Grp 170 and BiP where identified, while the others were not identified as known ER proteins, from Western blot analyses. The presence of the ATPase-binding proteins for Bip was first demonstrated in this study, and our data suggest similar regulatory machinery for BiP may exist in the ER, as found in prokaryotes and other cellular compartments.     

Hsp110 is a nucleotide-activated exchange factor for Hsp70

Hsp110 proteins constitute a subfamily of the Hsp70 chaperones and are potent nucleotide exchange factors (NEFs) for canonical Hsp70s of the eukaryotic cytosol. Here, we show that the NEF activity of the yeast Hsp110 homologue Sse1 itself is controlled by nucleotide. Nucleotide binding results in formation of a stabilized conformation of Sse1 that is required for association with the yeast Hsp70 Ssa1. The interaction triggers release of bound ADP from Ssa1, but nucleotide persists bound to Sse1 in the complex. Surprisingly, removal of this nucleotide does not affect the integrity of the complex. Instead, rebinding of ATP to the Hsp70 prompts the dissociation of the complex. Our data demonstrate that in contrast to previously characterized NEFs for Hsp70 chaperones, the NEF activity of Sse1 requires nucleotide binding and let us propose a new model for Hsp110 function.     

BAP,a mammalian BiP-associated protein,is a nucleotide exchange factor that regulates the ATPase activity of BiP

We identified a mammalian BiP-associated protein, BAP, using a yeast two-hybrid screen that shared low homology with yeast Sls1p/Sil1p and mammalian HspBP1, both of which regulate the ATPase activity of their Hsp70 partner. BAP encoded an approximately 54-kDa protein with an N-terminal endoplasmic reticulum (ER) targeting sequence, two sites of N-linked glycosylation, and a C-terminal ER retention sequence. Immunofluorescence staining demonstrated that BAP co-localized with GRP94 in the endoplasmic reticulum. BAP was ubiquitously expressed but showed the highest levels of expression in secretory organ tissues, a pattern similar to that observed with BiP. BAP binding was affected by the conformation of the ATPase domain of BiP based on in vivo binding studies with BiP mutants. BAP stimulated the ATPase activity of BiP when added alone or together with the ER DnaJ protein, ERdj4, by promoting the release of ADP from BiP. Together, these data demonstrate that BAP serves as a nucleotide exchange factor for BiP and provide insights into the mechanisms that control protein folding in the mammalian ER.     

GrpE-like regulation of the hsc70 chaperone by the anti-apoptotic protein BAG-1.

The BAG-1 protein appears to inhibit cell death by binding to Bcl-2, the Raf-1 protein kinase, and certain growth factor receptors, but the mechanism of inhibition remains enigmatic. BAG-1 also interacts with several steroid hormone receptors which require the molecular chaperones Hsc70 and Hsp90 for activation. Here we show that BAG-1 is a regulator of the Hsc70 chaperone. BAG-1 binds to the ATPase domain of Hsc70 and, in cooperation with Hsp40, stimulates Hsc70's steady-state ATP hydrolysis activity approximately 40-fold. Similar to the action of the GrpE protein on bacterial Hsp70, BAG-1 accelerates the release of ADP from Hsc70. Thus, BAG-1 regulates the Hsc70 ATPase in a manner contrary to the Hsc70-interacting protein Hip, which stabilizes the ADP-bound state. Intriguingly, BAG-1 and Hip compete in binding to the ATPase domain of Hsc70. Our results reveal an unexpected diversity in the regulation of Hsc70 and raise the possibility that the observed anti-apoptotic function of BAG-1 may be exerted through a modulation of the chaperone activity of Hsc70 on specific protein folding and maturation pathways.     

Characterization of the soluble domain of the ABC7 type transporter Atm1

Atm1 is an ABC transporter that is located in yeast mitochondria and has previously been implicated in the maturation of cytosolic iron-sulfur cluster proteins. The soluble nucleotide binding domain of Atm1 (Atm1-C) has been overexpressed in Escherichia coli, purified, and characterized. Dissociation constants (KD) for Atm1-C binding of ATP (KD approximately 97 microm, pH 7.3, and approximately 102 microm, pH 10.0) and ADP (KD approximately 43 microm, pH 7.3, and 92 microm, pH 10.0) were measured by fluorimetry. The higher binding affinity for ADP suggests that the transmembrane-spanning domain may be required to promote a structural change in the nucleotide binding domain to facilitate substrate export and ADP release. ADP also had an inhibitory effect on Atm1-C with an IC50 of 10 mm. The Michaelis-Menten constants Vmax, KM, and kcat of Atm1-C were measured as 1.822 microm min(-1), 513 microm, and 0.055 min(-1), respectively. The metal dependence of Atm1-C ATPase demonstrated a reactivity order of Mn2+ > Mg2+ > Co2+, while Mg2+ and Co2+ were both found to be inhibitory at higher concentrations. The pH profile and structural comparison with HisP are consistent with a role for His and Lys in promoting the ATPase activity. Structural analysis of Atm1-C by CD spectroscopy suggested a similarity of secondary structure to that found for a prokaryotic homologue (HisP), whereas modeling of the Atm1-C tertiary structure using HisP as a template is also consistent with a similarity in tertiary structure. Atm1-C tends to form a dimer or higher aggregation state at higher concentration; however, the concentration dependence of Atm1-C on ATPase activity and the results of a Hill analysis (napp = 1.1) demonstrated that there was essentially no cooperativity in ATP hydrolysis, in contrast to observations for the prokaryotic HisP transporter, which demonstrated full cooperativity for both full-length and the soluble domains. Accordingly, any cooperative response must be mediated through the transmembrane domain in the case of the eukaryotic Atm1 transporter.     

Sil phosphorylation in a Pin1 binding domain affects the duration of the spindle checkpoint

SIL is an immediate-early gene that is essential for embryonic development and is implicated in T-cell leukemia-associated translocations. We now show that the Sil protein is hyperphosphorylated during mitosis or in cells blocked at prometaphase by microtubule inhibitors. Cell cycle-dependent phosphorylation of Sil is required for its interaction with Pin1, a regulator of mitosis. Point mutation of the seven (S/T)P sites between amino acids 567 and 760 reduces mitotic phosphorylation of Sil, Pin1 binding, and spindle checkpoint duration. When a phosphorylation site mutant Sil is stably expressed, the duration of the spindle checkpoint is shortened in cells challenged with taxol or nocodazole, and the cells revert to a G2-like state. This event is associated with the downregulation of the kinase activity of the Cdc2/cyclin B1 complex and the dephosphorylation of the threonine 161 on the Cdc2 subunit. Sil downregulation by plasmid-mediated RNA interference limited the ability of cells to activate the spindle checkpoint and correlated with a reduction of Cdc2/cyclin B1 activity and phosphorylation on T161 on the Cdc2 subunit. These data suggest that a critical region of Sil is required to mediate the presentation of Cdc2 activity during spindle checkpoint arrest.     

The carboxyl-terminal lobe of Hsc70 ATPase domain is sufficient for binding to BAG1.

The molecular co-chaperone BAG1 and other members of the BAG family bind to Hsp70/Hsc70 heat shock proteins through a conserved BAG domain that interacts with the ATPase domain of the chaperone. BAG1 and other accessory proteins stimulate ATP hydrolysis and regulate the ATP-driven activity of the chaperone complexes. Contacts are made through residues in helices alpha2 and alpha3 of the BAG domain and predominantly residues in the C-terminal lobe of the bi-lobed Hsc70 ATPase domain. Within the C-terminal lobe, a subdomain exists that contains all the contacts shown by mutagenesis to be required for BAG1 recognition. In this study, the subdomain, representing Hsc70 residues 229-309, was cloned and expressed as a separately folded unit. The results of in vitro binding assays demonstrate that this subdomain is sufficient for binding to BAG1. Binding analyses with surface plasmon resonance indicated that the subdomain binds to BAG1 with a 10-fold decrease in equilibrium dissociation constant (K(D) = 22 nM) relative to the intact ATPase domain. This result suggests that the stabilizing contacts for docking of BAG1 to Hsc70 are located in the C-terminal lobe of the ATPase domain. These findings provide new insights into the role of co-chaperones as nucleotide exchange factors.     

Structure of the ABC ATPase domain of human TAP1, the transporter associated with antigen processing

The transporter associated with antigen processing (TAP) is an ABC transporter formed of two subunits, TAP1 and TAP2, each of which has an N-terminal membrane-spanning domain and a C-terminal ABC ATPase domain. We report the structure of the C-terminal ABC ATPase domain of TAP1 (cTAP1) bound to ADP. cTAP1 forms an L-shaped molecule with two domains, a RecA-like domain and a small alpha-helical domain. The diphosphate group of ADP interacts with the P-loop as expected. Residues thought to be involved in gamma-phosphate binding and hydrolysis show flexibility in the ADP-bound state as evidenced by their high B-factors. Comparisons of cTAP1 with other ABC ATPases from the ABC transporter family as well as ABC ATPases involved in DNA maintenance and repair reveal key regions and residues specific to each family. Three ATPase subfamilies are identified which have distinct adenosine recognition motifs, as well as distinct subdomains that may be specific to the different functions of each subfamily. Differences between TAP1 and TAP2 in the nucleotide-binding site may be related to the observed asymmetry during peptide transport.     

A comparison of the effect of vanadate on the binding of myosin-subfragment-1.ADP to actin and on actomyosin subfragment 1 ATPase activity

Equilibrium binding studies were used to determine the binding constant of vanadate ion (Vi), to the complex of actomyosin subfragment 1 (S1) with ADP and Vi and of actin to the myosin S1.ADP.Vi complex. The proteins used were obtained from rabbit skeletal muscle. Pre-steady-state measurements were also performed to determine the rates of Vi association and dissociation from the actomyosin S1.ADP.Vi complex. Using these measured values in a simple model, the steady-state actomyosin S1 ATPase activity was predicted over a range of Vi concentrations. This model predicted that Vi would have little effect on the actomyosin S1 ATPase activity. In agreement with this prediction, the measured ATPase activity of actomyosin S1 was not greatly inhibited by Vi, except at high concentrations at which polymeric species of Vi may occur (greater than 900 microM).     

Fes1p acts as a nucleotide exchange factor for the ribosome-associated molecular chaperone Ssb1p

The HspBP1 homolog Fes1p was recently identified as a nucleotide exchange factor (NEF) of Ssa1p, a canonical Hsp70 molecular chaperone in the cytosol of Saccharomyces cerevisiae. Besides the Ssa-type Hsp70s, the yeast cytosol contains three additional classes of Hsp70, termed Ssb, Sse and Ssz. Here, we show that Fes1p also functions as NEF for the ribosome-bound Ssb Hsp70s. Sequence analysis indicated that residues important for interaction with Fes1p are highly conserved in Ssa1p and Ssb1p, but not in Sse1p and Ssz1p. Indeed, Fes1p interacts with Ssa1p and Ssb1p with similar affinity, but does not form a complex with Sse1p. Functional analysis showed that Fes1p accelerates the release of the nucleotide analog MABA-ADP from Ssb1p by a factor of 35. In contrast to the interaction between mammalian HspBP1 and Hsp70, however, addition of ATP only moderately decreases the affinity of Fes1p for Ssb1p. Point mutations in Fes1p abolishing complex formation with Ssa1p also prevent the interaction with Ssb1p. The ATPase activity of Ssb1p is stimulated by the ribosome-associated complex of Zuotin and Ssz1p (RAC). Interestingly, Fes1p inhibits the stimulation of Ssb1p ATPase by RAC, suggesting a complex regulatory role of Fes1p in modulating the function of Ssb Hsp70s in co-translational protein folding.     

Hsp70 Cochaperones HspBP1 and BAG-1M Differentially Regulate Steroid Hormone Receptor Function

Hsp70 binding protein 1 (HspBP1) and Bcl2-associated athanogene 1 (BAG-1), the functional orthologous nucleotide exchange factors of the heat shock protein 70 kilodalton (Hsc70/Hsp70) chaperones, catalyze the release of ADP from Hsp70 while inducing different conformational changes of the ATPase domain of Hsp70. An appropriate exchange rate of ADP/ATP is crucial for chaperone-dependent protein folding processes. Among Hsp70 client proteins are steroid receptors such as the glucocorticoid receptor (GR), the mineralocorticoid receptor (MR), and the androgen receptor (AR). BAG-1 diversely affects steroid receptor activity, while to date the influence of HspBP1 on steroid receptor function is mostly unknown. Here, we compared the influence of HspBP1 and BAG-1M on Hsp70-mediated steroid receptor folding complexes and steroid receptor activity. Coimmunoprecipitation studies indicated preferential binding of Hsp40 and the steroid receptors to BAG-1M as compared to HspBP1. Furthermore, Hsp70 binding to the ligand-binding domain of GR was reduced in the presence of HspBP1 but not in the presence of BAG-1M as shown by pull-down assays. Reporter gene experiments revealed an inhibitory effect on GR, MR, and AR at a wide range of HspBP1 protein levels and at hormone concentrations at or approaching saturation. BAG-1M exhibited a transition from stimulatory effects at low BAG-1M levels to inhibitory effects at higher BAG-1M levels. Overall, BAG-1M and HspBP1 had differential impacts on the dynamic composition of steroid receptor folding complexes and on receptor function with important implications for steroid receptor physiology.     

Structural basis of J cochaperone binding and regulation of Hsp70

The many protein processing reactions of the ATP-hydrolyzing Hsp70s are regulated by J cochaperones, which contain J domains that stimulate Hsp70 ATPase activity and accessory domains that present protein substrates to Hsp70s. We report the structure of a J domain complexed with a J responsive portion of a mammalian Hsp70. The J domain activates ATPase activity by directing the linker that connects the Hsp70 nucleotide binding domain (NBD) and substrate binding domain (SBD) toward a hydrophobic patch on the NBD surface. Binding of the J domain to Hsp70 displaces the SBD from the NBD, which may allow the SBD flexibility to capture diverse substrates. Unlike prokaryotic Hsp70, the SBD and NBD of the mammalian chaperone interact in the ADP state. Thus, although both nucleotides and J cochaperones modulate Hsp70 NBD:linker and NBD:SBD interactions, the intrinsic persistence of those interactions differs in different Hsp70s and this may optimize their activities for different cellular roles.     

Experimentally biased model structure of the Hsc70/auxilin complex: substrate transfer and interdomain structural change

A model structure of the Hsc70/auxilin complex has been constructed to gain insight into interprotein substrate transfer and ATP hydrolysis induced conformational changes in the multidomain Hsc70 structure. The Hsc70/auxilin system, which is a member of the Hsp70/Hsp40 chaperone system family, uncoats clathrin-coated vesicles in an ATP hydrolysis-driven process. Incorporating previous results from NMR and mutant binding studies, the auxilin J-domain was docked into the Hsc70 ATPase domain lower cleft using rigid backbone/flexible side chain molecular dynamics, and the Hsc70 substrate binding domain was docked by a similar procedure. For comparison, J-domain and substrate binding domain docking sites were obtained by the rigid-body docking programs DOT and ZDOCK, filtered and ranked by the program ClusPro, and relaxed using the same rigid backbone/flexible side chain dynamics. The substrate binding domain sites were assessed in terms of conserved surface complementarity and feasibility in the context of substrate transfer, both for auxilin and another Hsp40 protein, Hsc20. This assessment favors placement of the substrate binding domain near D152 on the ATPase domain surface adjacent to the J-domain invariant HPD segment, with the Hsc70 interdomain linker in the lower cleft. Examining Hsc70 interdomain energetics, we propose that long-range electrostatic interactions, perhaps due to a difference in the pKa values of bound ATP and ADP, could play a major role in the structural change induced by ATP hydrolysis. Interdomain electrostatic interactions also appear to play a role in stimulation of ATPase activity due to J-domain binding and substrate binding by Hsc70.     

Crystal structure of yeast V1‐ATPase in the autoinhibited state

Vacuolar ATPases (V‐ATPases) are essential proton pumps that acidify the lumen of subcellular organelles in all eukaryotic cells and the extracellular space in some tissues. V‐ATPase activity is regulated by a unique mechanism referred to as reversible disassembly, wherein the soluble catalytic sector, V₁, is released from the membrane and its MgATPase activity silenced. The crystal structure of yeast V₁ presented here shows that activity silencing involves a large conformational change of subunit H, with its C‐terminal domain rotating ~150° from a position near the membrane in holo V‐ATPase to a position at the bottom of V₁ near an open catalytic site. Together with biochemical data, the structure supports a mechanistic model wherein subunit H inhibits ATPase activity by stabilizing an open catalytic site that results in tight binding of inhibitory ADP at another site.     

Role of Hsp70 ATPase Domain Intrinsic Dynamics and Sequence Evolution in Enabling its Functional Interactions with NEFs

Catalysis of ADP-ATP exchange by nucleotide exchange factors (NEFs) is central to the activity of Hsp70 molecular chaperones. Yet, the mechanism of interaction of this family of chaperones with NEFs is not well understood in the context of the sequence evolution and structural dynamics of Hsp70 ATPase domains. We studied the interactions of Hsp70 ATPase domains with four different NEFs on the basis of the evolutionary trace and co-evolution of the ATPase domain sequence, combined with elastic network modeling of the collective dynamics of the complexes. Our study reveals a subtle balance between the intrinsic (to the ATPase domain) and specific (to interactions with NEFs) mechanisms shared by the four complexes. Two classes of key residues are distinguished in the Hsp70 ATPase domain: (i) highly conserved residues, involved in nucleotide binding, which mediate, via a global hinge-bending, the ATPase domain opening irrespective of NEF binding, and (ii) not-conserved but co-evolved and highly mobile residues, engaged in specific interactions with NEFs (e.g., N57, R258, R262, E283, D285). The observed interplay between these respective intrinsic (pre-existing, structure-encoded) and specific (co-evolved, sequence-dependent) interactions provides us with insights into the allosteric dynamics and functional evolution of the modular Hsp70 ATPase domain.