Aha1 Is Composed of Two Distinct Domains
Aha1 is composed of two domains connected by a degenerate an ~30-amino acid linker (approximately residues 163-200; A). Previous studies have reported that the N-terminal domain of Aha1 binds to the middle domain of Hsp90 and stimulates its otherwise very low intrinsic ATPase activity (Panaretou et al., 2002
; Meyer et al., 2004
). Other studies have concluded that the C-terminal domain of Aha1 is required for ATPase stimulation (Lotz et al., 2003
) and confers higher binding affinity for Hsp90 (Meyer et al., 2004
). Consequently, the precise role of the C-terminal domain of Aha1 in regulating the adenine nucleotide-dependent conformational cycle of Hsp90 remains controversial.
Figure 1. Domain structure of Aha1. (A) Schematic of Aha1 domains. Full-length (FL) human Aha1 (amino acids 1-338 shown in dark green), Aha1 N-terminal domain (NTD; amino acids 1-162 shown in light green), and Aha1 C-terminal domain (CTD; amino acids 163-338 shown (more ...)
To study the biochemical properties of the N- and C-terminal domains of Aha1, we created plasmid constructs of residues 1-162 (N-terminal domain) and 163-338 (C-terminal domain) of Aha1 for expression and purification from E. coli (see Materials and Methods; A). Circular dichroism suggested folded structures (B). When the spectra of the N- and C-terminal domains observed alone were added together, they were equal to the spectrum of the full-length Aha1 protein (B). Thus, the two domains retained the same overall secondary structure as when in the context of the full-length protein. To further confirm this conjecture, we performed NMR analysis on full-length Aha1 as well as the N- and C-terminal domains. The NMR spectrum for the full-length Aha1 protein was well dispersed and the N- and C termini gave equally well dispersed spectra further supporting that the two domains are folded properly on their own (C). However, the NMR spectrum for the C-terminal domain suggested the presence of an unstructured portion of the polypeptide. Given that full-length Aha1 shows a region of extensive degeneracy (a possible flexible linker) separating the N- and C-terminal domains (D; see Supplemental Figure S1 for enlarged image) that corresponded to the residues 163-200 of our C-terminal construct, we generated a new C-terminal construct (residues 201-338) that had this linker peptide removed. The NMR signal of the 201-338 construct was identical to that for the 163-338 construct except for the loss of the poorly dispersed signal contained in the 163-338 data set (E). Together, our results suggest that Aha1 is composed of two independently folded domains joined by a flexible linker.
Only Full-Length Aha1 Robustly Stimulates the Hsp90 ATPase Activity
We tested our full-length, N- and C-terminal constructs for their ability to stimulate the ATPase activity of the β isoform of Hsp90 (to be referred to hereafter as simply Hsp90), the house-keeping isoform common to all mammalian cell types. In all experiments reported below, the C-terminal construct of Aha1 contains the linker region. Consistent with previous reports, we found that the full-length protein gave robust ATPase stimulation (A). Although we found that the N-terminal domain failed stimulate Hsp90, the C-terminal domain weakly but reproducibly stimulated ATPase activity (A). This was surprising because, although there are conflicting reports regarding the ability of the N-terminal domain to stimulate the Hsp90 ATPase activity (Panaretou et al., 2002
; Lotz et al., 2003
; Meyer et al., 2004
), there has never been any activity attributed to the C terminus of Aha1. We also analyzed the ability of Aha1 to interact with Hsp90 using SPR. Full-length Aha1 showed a strong interaction that was fit to a 1:1 Langmuir model and was determined to be 0.5 μM (see Supplemental Table S1 for Kon
values). Because the off rates of the N- and C-terminal domains to Hsp90 were too fast to be measured using this model, binding constants were calculated using a steady-state model applicable for weak interactions (B). Using this approach, the N-terminal domain had an affinity for Hsp90 of 2.5 μM, whereas the C-terminal domain containing the linker region showed a slightly reduced affinity of 3.8 μM. This reduction in the binding of the N- and C-terminal domains of Aha1 compared with the intact full-length protein is consistent with the inability of either the N- or C-terminal domains to stimulate Hsp90 ATPase activity in a manner comparable with the full-length protein.
Figure 2. Effect of increasing Aha1 on Hsp90 binding and ATPase. (A) Hsp90 ATPase assay with increasing amount of full-length (FL), N-terminal domain (NTD), and C-terminal domain (CTD) of Aha1. We incubated 4 μM final Hsp90 with the indicated amounts of (more ...)
Molecular Footprinting Reveals a Site of Aha1 Interaction with N- and Middle Domains of Hsp90
To develop an understanding of the interaction between Aha1 and Hsp90, we developed a molecular footprinting strategy involving chemical derivatization followed by mass spectrometry analysis (see Materials and Methods
). Here, the rapid covalent modification of a reactive group of a derivatizing agent with the side chain containing terminal amine (such as Lys or Arg) found in Aha1 or Hsp90, or the Aha1–Hsp90 complex, was followed by quenching of the reaction, enzymatic digestion of the proteins, and identification of the sites of modification using mass spectrometry and MudPIT (Washburn et al., 2001
; Washburn et al., 2003
). Theoretically, residues that are protected from derivatization by the interaction of Aha1 with Hsp90 would be evident compared with the sites of derivatization when the Aha1 monomer or the Hsp90 dimer are incubated alone with the derivatizing agent.
After evaluation of a number amine reactive probes, we made use of the short methyl PEG4-NHS ester as the reactive group. This gave adequate mass differences, as well as comparable sequence coverage compared with the unlabeled proteins (our unpublished data). We reasoned that this adduct would not change significantly the surface properties of the proteins as compared with more hydrophobic reagents such as NHS-(CH2)x-biotin, for example, which we found to decrease the MS sequence coverage (our unpublished data). With the methyl-PEG4-NHS-ester we achieved routine labeling of ~75% of all Lys and Arg in Hsp90 and Aha1 that were evenly distributed across the surface of these proteins.
The results of the footprinting experiments in the absence of nucleotide to capture all of the intermediate cycling states are presented in and are illustrated in , A–D. We observed complete protection from covalent modification of the residues positioned within the footprint of the interaction between the middle domain of Hsp90 and the N-terminal domain of Aha1 observed in the cocrystal structure (Meyer et al., 2003
; A). Thus, the footprinting technique recapitulates the previously known interaction site and validates our methodology carried out in solution under physiological conditions. Our analysis, as expected, also revealed a number of protected residues (B) located within the dimerization interface between the N-, middle, and C-terminal domains of Hsp90 (D). Strikingly, several additional Lys and Arg residues outside the dimerization interface were located in both the N-terminal and middle domains of Hsp90 were protected from modification in the presence Aha1 (, B and D). One Lys residue in the C-terminal domain of Aha1 (K273) was protected by Hsp90 (C). The presence of a second and new binding site on Hsp90 for the C terminus of Aha1 in addition to the known N-terminal domain interaction of Aha1 with Hsp90 provides an explanation for the observation that only the full-length Aha1 protein can maximally bind and stimulate the Hsp90 ATPase activity ().
Figure 3. Molecular footprinting of the interaction between Hsp90 and Aha1. (A) Footprinting was performed as described in Materials and Methods. Protected residues on Hsp90 (shown in red) and Aha1 (shown in orange) are depicted on the crystal structure of the (more ...)
Cross-Linking Reveals an Interaction of the Aha1 C Terminus with Hsp90
To capture the site(s) of protein–protein interactions between Aha1 and Hsp90, we used a stringent, zero-length cross-linker, EDC. This cross-linker forms a peptide bond between side chains of amino-containing amino acids (such as Lys or Arg) and an acidic residue (Glu or an Asp), residing in immediate proximity in the protein complex. Under physiological salt conditions, we detected higher order complexes (~5% of total Aha1/Hsp90 added) that reflect interaction of Aha1 full-length or each of the C- and N-terminal domains with either Hsp90 monomer and dimers (A, lanes c–h, black arrowheads). These were not detected in control incubations containing either purified Hsp90 (A, lane l), full-length Aha1 (A, lane k), or the individual C- or N-terminal domains in the presence of EDC (A, lanes i and j).
Figure 4. Analysis of binding of Aha1 cross-linked to Hsp90. (A) A 4–12% gradient SDS-PAGE of purified Hsp90 (lane a) and Aha1 (lane b). Full-length (lanes e, h, and k), N-terminal domain (lanes c, f, and i), or C-terminal domain (lanes d, g, and j) were (more ...)
To further characterize the interaction of Aha1 and Hsp90, we mapped the chemically cross-linked sites by using mass spectrometry. In brief, MS/MS were generated from trypsin-digested non–cross-linked and cross-linked samples from incubations containing full-length Aha1 or N- and C-terminal fragments incubated with Hsp90 (see Materials and Methods). To identify cross-linked residues, we used the schema outlined in Supplemental Fig. S2. MS/MS from non–cross-linked samples were searched against the EBI IPI human database, version 3.23, with its reverse decoy to identify peptides that can be obtained from the individual proteins (e.g., Aha1 or Hsp90). The identified Aha1 peptides and Hsp90 peptides were then cross-linked in silico to create a database of artificial cross-linked peptides. We only used the experimentally identified peptides to create the artificial cross-linked peptides database because the cross-linked peptides from experimentally identified non–cross-linked peptides are probably a more accurate reflection of peptides relevant to the cross-linking experiment. MS/MS from cross-linked samples were also searched against the EBI-IPI human database. MS/MS matched to the database were subtracted from the data set. The remaining MS/MS were then searched against the EBI-IPI human database plus the artificial cross-linked peptides database with their reverse decoys for the discovery of cross-linked peptides. These experiments were performed in the absence of presence of ATP nucleotide to access all possible intermediate folding states, or in the presence of ADP or AMP-PNP, to favor populating the Hsp90 dimer to the open or closed states, respectively.
In the absence of nucleotide using either low-ionic strength buffer (25 mM HEPES, pH 7.4) to favor the interactions between Hsp90 and Aha1 (Richter et al., 2008
) or in a physiologically relevant ionic strength (50 mM HEPES, pH 7.4, and 100 mM NaCl) to minimize nonspecific charge–charge interactions that might occur in the low ionic strength buffer, we found that only one binding site in Aha1 is the preferred site for cross-linking given prominence of this peptide in both MS/MS analyses ( and B). This site involves the C terminus of Aha1 (residues 277-289; 275
) with the N terminus of Hsp90 (residues 70-82; 70
). Identical results were obtained in the presence of ATP (), demonstrating that consistent with previous results (Richter et al., 2008
), the interaction of Aha1 with Hsp90 occurs with similar efficiencies in the presence or absence of nucleotide. In the absence of nucleotide, we also detected a second cross-linking event between residues 277-289 (275
) and a peptide (residues 320-329; 320
) found in the middle domain of Hsp90, suggesting additional conformational flexibility. In contrast, when cross-linking was performed at physiological salt in the presence of ADP to stabilize the open state, we observed no cross-linking (), suggesting that interaction of Hsp90 with Aha1 is more restricted in an ADP-conformation favored incubation condition. These results provide an important control for specificity of interactions observed in the presence of absence of ATP.
We performed an additional cross-linking experiment to analyze whether we gain or lose interactions between full-length Aha1 and Hsp90 by incubation in the presence of the nonhydrolysable analogue AMP-PNP. When AMP-PNP is added at the beginning of the incubation of Aha1 and Hsp90, we observe poor and variable recovery of cross-linked species, suggesting that rapid recruitment of the analogue leading to the closed state may preclude interaction of Aha1 and Hsp90. However, when AMP-PNP is added shortly after reincubation of Aha1 and Hsp90 (5–10 min), we recover the identical C-terminal Aha1 and N-terminal Hsp90 cross-linked peptide observed in the presence or absence of ATP (). Under these conditions we did not detect interactions between the N terminus of Aha1 and the peptide in the middle domain of Hsp90 found in the nucleotide-free condition. These results suggest that Aha1 must interact with Hsp90 in a specific manner as it actively cycles from the open to closed states with Aha1 becoming trapped in a partially or fully closed state in the presence of AMP-PNP. In summary, the limited number of cross-linked species detected had strong overlap with the regions protected by the interaction of Aha1 and Hsp90 using molecular foot printing, supporting the conclusion that Aha1 and Hsp90 have a unique two-domain interaction.
Aha1 Bridges the Hsp90 Dimmer to Facilitate ATP Hydrolysis
The presence of two discrete binding sites for full-length Aha1 on Hsp90 poses several models. One possibility is that Aha1 binds along the surface of one Hsp90 monomer (‘cis'-binding); alternatively, Aha1 may bind across the Hsp90 dimmer interface (‘trans'-binding). To address this question; we performed an additional cross-linking experiment using bifunctional cross-linkers. In brief, we labeled Aha1 with ANB-NOS. This cross-linker has an aminoreactive moiety that can be used to modify surface Lys residues and a UV-activated group that can react nonspecifically with side chains in proximity. Using this reagent, we mildly derivatized Aha1 under substoichiometric labeling conditions to retain its surface properties. After quenching of the reaction and removal of the unreacted ANB-NOS, Aha1 was incubated with Hsp90, and complexes were UV irradiated to cross-link Aha1 to Hsp90 with the photo-reactive moiety of ANB-NOS followed by SDS-PAGE. Because Hsp90 has not been labeled, and as expected, we only recover Hsp90 monomers (C). If one molecule of Aha1 binds to a Hsp90 monomer (~90 kDa), we would expect to observe cis-cross–linked species larger than the Hsp90 monomer (~130 kDa) but always smaller than the migration of the Hsp90 dimer (~180 kDa). These species were not observed. If, in contrast, Aha1 binds across the Hsp90 dimer interface in trans, we would expect to see trans-cross–linked species larger (>220 kDa) than that observed for the migration of the Hsp90 dimer (C, asterisk). Because Hsp90 is not modified with the cross-linking agent, the only way that species larger than the Hsp90 dimer can be generated are through Aha1 interactions. Indeed, we were able to identify a number of high-molecular-weight bands (>220 kDa) that represent Aha1-cross-linked Hsp90 species migrating more slowly than the Hsp90 dimer (C). It remains possible that these slower migrating cross-linked bands reflect an unusual extended state in response to denaturation by the sample buffer before SDS-PAGE. We consider this unlikely given the very low stoichiometry of the cross-linking agent used to covalently modify Aha1; that Aha1 is well folded and compact under these conditions; and that Hsp90 is a native, unmodified protein.
Our combined results demonstrate that the two distinct interaction sites between Aha1 and Hsp90 detected by protection () and cross-linking experiments () reveal that Aha1 functions to bridge the dimer interface of Hsp90 with the N terminus of Aha1 bound to the middle domain of one Hsp90 monomer and the C terminus of Aha1 bound to the other Hsp90 monomer in the dimer pair to stabilize and/or promote ATP hydrolysis.
Overexpression of WT Aha1 Prevents WT-CFTR Folding and Export from the ER in a Dose-dependent Manner
In vitro interactions do not necessarily recapitulate interactions observed in the cell given the involvement of additional unknown factors, particularly in the complex cochaperone interactions with Hsp90. Moreover, the unknown stoichiometries involved in biological reactions often provide an additional level of complexity. For example, Hsp90 is present in large excess over Aha1 (>100-fold; see below), has additional cochaperones that assist Hsp90 in folding, and has numerous clients that it must maintain in a physiological functional state (Panaretou et al., 2002
)(our unpublished data).
To address the impact of the interaction of Aha1 with Hsp90 as dimer-bridging molecule directing client folding in vivo, we turned to our previous work. This work focused on the role of WT Aha1 and Hsp90 in the stability of the ΔF508 variant of CFTR for trafficking through the exocytic pathway and delivery to the cell surface (Wang et al., 2006
). Here, we demonstrated that small interfering RNA (siRNA) silencing of Aha1 expression reduced ERAD of ΔF508 and promoted conversion to band C and delivery to the cell surface (Wang et al., 2006
). In contrast, overexpression of Aha1 destabilized both WT and ΔF508-CFTR and promoted ER-associated degradation, suggesting that the Hsp90 ATPase cycle regulated by Aha1 plays a critical role in CFTR folding (Wang et al., 2006
). The mechanism responsible for these events remains unknown.
In the current study, we wanted to characterize the effect of Aha1 and N- and C-terminal mutants that disrupt its interaction with Hsp90 on the stability and trafficking of WT-CFTR as a biological measure of domain function. For this purpose, we cotransfected WT-CFTR with increasing amounts of full-length Aha1 by using a vaccinia expression system in human embryonic kidney (HEK)293 cells (see Materials and Methods). Trafficking of CFTR to the cell surface can be readily measured by following the processing of CFTR from the band B, high-mannose glycoform localized to the ER during nascent synthesis by ER-bound ribosomes, to the more slowly migrating band C glycoform that is generated by Golgi-associated enzymes to form complex sugar-containing species before delivery of CFTR to the cell surface (A). Similar to results we observed for ΔF508, increasing amounts of WT Aha1 destabilized WT-CFTR in the ER. This is evident by the observed reduced levels of band B. Moreover, increasing WT Aha1 Aha1 interfered with export from the ER to Golgi as indicated by the partial loss of band C even at the lowest levels of expression tested (A).
Figure 5. Effect of Aha1 mutants on CFTR folding and trafficking. (A) Dose-dependent destabilization of CFTR overexpression with the indicated amount of Aha1 plasmid. Western blotting performed with indicated antibody to CFTR (top) or Aha1 (bottom). The three identical (more ...)
The observation that the steady-state level of band C preceded the loss of band B restricted to ER in response to increasing Aha1 (A) is consistent with the interpretation that the Aha1-sensitive Hsp90-dependent step(s) required for the stability of WT-CFTR in the ER is required to generate a folding competent conformation for export (Wang et al., 2004
). Using absolute quantification (single ion reaction monitoring mass spectrometry; Janecki et al., 2007
) approaches, we determined that Hsp90 constitutes approximately ~1–2% of the total protein content in the lysates prepared from the HEK293 cells, whereas Aha1 is a minor pool (<0.01%). Quantitative Western blotting showed that band C CFTR loss could be detected with as little as approximately threefold overexpression (the lowest plasmid concentration tested; A). This represents a change in stoichiometry with Hsp90 from ~1:100 to ~1:30 but was dose dependent and much stronger with increased expression of Aha1. These results likely reflect competition Aha1 with a large pool of dynamic Aha1–Hsp90–client complexes. These experiments suggest that increased levels of Aha1 relative to its endogenous concentration impair the ability of even WT-CFTR to fold into an export competent state that is protected from ERAD.
The N-Terminal Domain of Aha1 Is Critical for Destabilization of WT CFTR Folding
Because the relationship between the in vivo properties of WT Aha1 and its Hsp90 ATPase activity are controversial and the mechanism(s) facilitating client folding remain largely unknown, we tested whether destabilization of WT-CFTR folding by Aha1 overexpression was dependent on Hsp90 by generating Aha1 mutants. We were particularly interested in whether binding and/or ATPase activity were required for stabilization of WT CFTR folding. Whereas loss of binding would be predicted to result in loss of interaction Hsp90 and hence, loss of ATPase stimulating activity, it is unclear whether binding is sufficient and that the ATPase-stimulating activity conferred by the C-terminal domain is, in addition, required in CFTR folding.
A previous study identified a mutation in the N-terminal domain of yeast Hch1p (D53K) that impaired its ability to stimulate yeast Hsp90 (Meyer et al., 2004
). We made the equivalent mutation in human Aha1 (E67K) and tested for stimulation of Hsp90 ATPase activity. No stimulation was observed, similar to that observed previously for yeast D53K (B). Using SPR, we found that the E67K mutant had an extremely slow association rate (at least 15-fold slower than WT) and did not reach equilibrium binding under the range of conditions and concentrations tested (1–30 μM; our unpublished data), limiting our ability to accurately assess the kinetics of binding of the E67K mutant to Hsp90. However, it is clear that binding is severely impaired and this result explains the inability of the E67K mutant of Aha1 to stimulate the Hsp90 ATPase activity. We then tested this mutant by examining its effect on WT-CFTR export from the ER and trafficking to the cell surface. Here, we found that overexpression of the E67K mutant, compared with WT Aha1, was less efficient in promoting WT-CFTR degradation of band B at the highest expression level tested (C, left). We observed statistically significant protection against the loss band C at the lowest concentration tested (C, right). These results demonstrate that interaction of Aha1 through the N-terminal domain with Hsp90 is at least one critical event required for promoting efficient folding and export of WT-CFTR from the ER.
Mutations in the C Terminus of Aha1 Impair Hsp90 ATPase Stimulation without Significantly Altering Binding to Hsp90
To address the role of the C-terminal ATPase activation domain of Aha1 on WT-CFTR export, we made a series of mutants based on primary sequence alignments and conserved residues between Aha1 proteins from different species (D and Supplemental Figure S1). Several of these mutants are found in regions that are likely to form the interaction region of the C-terminal domain of Aha1 with the N terminus of Hsp90 ( and ). Many mutants did not express well in mammalian cells, probably owing to the loss of critical side chains necessary for folding based on the structure of Aha1 C- and N-terminal domains. However, several mutants not only seemed to be expressed to levels within 50% or greater of wild-type Aha1 but were markedly impaired in their ability to destabilize WT-CFTR (Supplemental Figure S3A) and ΔF508-CFTR (Supplemental Figure S3B). Notably, these included C207S, E221A, E267K, D293A, E297A, T298A, and E313A, residues that may contribute to the stability of the Aha1 interaction with Hsp90 ( and B).
Mutants impaired in their ability to destabilize CFTR were selected for purification from E. coli for further testing in vitro. Only the E221A and D293A mutants were soluble when purified, as many of the mutants either did not express well in E. coli or rapidly aggregated into higher molecular weight species during or after purification. Unlike the E67K mutant, both the E221A and D293A Aha1 mutants were capable of interacting with Hsp90 by SPR (A) in a manner similar to that observed for wild-type Aha1 (0.89 and 0.44 μM, respectively; see Supplemental Table S1 for Kon and Koff values). Interestingly, both of these mutants were partially impaired in Hsp90 ATPase stimulation (B). To more carefully assess their ability to influence WT-CFTR export, we coexpressed increasing amounts of E221A and D293A with WT-CFTR in HEK293 cells (C). At an ~5- to 10-fold increased expression relative to endogenous Aha1 (~1:10, Aha1:Hsp90), these mutants were statistically significantly impaired compared with WT Aha1 in their ability to prevent CFTR folding and export.
Figure 6. Analysis of E221A and D293A mutants in vitro. (A) SPR analysis of the indicated amount of E221A and D293A mutants of Aha1 with Hsp90 as described in Materials and Methods. (B) Hsp90 ATPase assay with the indicated amount (micromolar) of WT (black bars), (more ...)
The combined data suggest that the activity of both the N- and C-terminal domains of Aha1 in binding Hsp90 or stimulating the ATPase activity of Hsp90, respectively, are important for Hsp90 regulated folding and export of CFTR. These interactions are probably aggravated by the energetically destabilizing Phe 508 deletion in ΔF508 CFTR leading to cystic fibrosis.