The Molecular Chaperone gp96/GRP94 Interacts with Toll-like Receptors and Integrins via Its C-terminal Hydrophobic Domain*

Background: gp96 is a molecular chaperone that binds and chaperones TLRs and integrins. But the structural basis of such is unknown. Results: Mutation of a C-terminal loop structure in gp96 abrogates its ability to bind and chaperone both receptors. Conclusion: We have successfully mapped the client-binding domain of gp96. Significance: This study shall facilitate the development of targeted gp96 inhibitors for treatment of inflammation. The structural basis for molecular chaperones to discern misfolded proteins has long been an enigma. As the endoplasmic reticulum paralogue of the cytosolic HSP90, gp96 (GRP94, HSP90b1) is an essential molecular chaperone for Toll-like receptors (TLRs) and integrins. However, little is known about its client-binding domain (CBD). Herein, we provide genetic and biochemical evidence to definitively demonstrate that a C-terminal loop structure, formed by residues 652–678, is the critical region of CBD for both TLRs and integrins. Deletion of this region affects neither the intrinsic ATPase activity nor the overall conformation of gp96. However, without it, the chaperoning function of gp96 collapses. We also find a critical Met pair (Met658-Met662) for the folding of integrins but not TLRs. Moreover, we find that the TLR binding to gp96 is also dependent on the C-terminal dimerization domain but not the N-terminal ATP-binding pocket of gp96. Our study has unveiled surprisingly the exquisite specificity of gp96 in substrate binding and suggests a manipulation of its CBD as an alternative strategy for targeted therapy of a variety of diseases.

Known also as HSP90b1 (1), GRP94 (2), ERp99 (3), endoplasmin (4), and TRA-1 (5), gp96 (6) is the endoplasmic reticulum (ER) 2 paralogue of the cytosolic heat shock protein 90 (HSP90) (7). It has long been implicated in maintaining protein homeostasis in the secretory pathway by participating in the general folding mechanism of glycoproteins and facilitating the ER-associated degradation of misfolded proteins (8). A landmark mutagenesis study in a pre-B cell line in vitro, however, dem-onstrates that the clientele of gp96 is restricted to a relatively small number of proteins, including Toll-like receptors (TLRs) and integrins (9). This conclusion has been largely corroborated by extensive gene knock-out studies in mice (10 -15). gp96 is among the highest conserved molecules along the phylogenic tree. For example, we have demonstrated that the Drosophila gp93, an ortholog of mammalian gp96, can fold both TLRs and integrins in gp96-null murine cells (16).
TLRs are important pattern recognition receptors for microbial products. There are 13 members of the TLR family in the mammalian system (17). TLR3, TLR7, TLR8, TLR9 and TLR13 are localized primarily in the endolysosomes, whereas the rest of TLRs are expressed on the cell surface. All TLRs are type I transmembrane receptors that form dimers in the ER for subsequent subcellular trafficking and function (18). The structure of TLRs includes a leucine-rich repeat (LRR)-containing ectodomain, a transmembrane domain, and a cytoplasmic tail that contains Toll/IL-1 receptor (TIR) domain. TLRs exploit their respective ectodomains to detect and bind to their specific microbial or endogenous ligands, followed by the initiation of signaling events via the TIR domain. Through multiple cytoplasmic signaling adaptors including MyD88, TIRAP, TRIF, and TRAM, TLR ligation can activate several signaling pathways including MAPK, IRF, and NF-B pathways and induce proinflammatory cytokine production (19). While TLRs serve as the first line of defense in innate immunity, they could also be implicated in pathological conditions such as endotoxic shock, autoimmunity, and inflammation-associated malignancy.
Similar to TLRs, integrins are type I transmembrane receptor dimers formed by ␣ and ␤ subunits. There are 18 ␣ subunits and 8 ␤ subunits, which heterodimerize to form 24 distinct integrin members (20). Integrins collectively play multi-faceted functions in cell migration, adhesion, survival, and apoptosis as well as in cell polarization and chemotaxis (20 -22). Although TLRs and integrins do not share apparent structural homology, the proper folding and subcellular distribution of a majority of TLRs and integrins is unexpectedly and absolutely dependent on gp96 (9, 10). Both gp96 and HSP90 contain a N-terminal ATP-binding pocket with intrinsic ATPase activity, followed by the middle charged domain and the C-terminal homodimerization domain (23,24). However, the client-binding domain of both molecules and the respective chaperoning mechanism still await elucidation. The crystal structure of the C-terminal dimerization domain of the Escherichia coli HSP90 homolog, HtpG, showed that helix 2 of this domain (residues 548 -557) formed a potential client binding surface (25). Subsequently, a mutagenesis study of residues along the same helix in mammalian HSP90 abolished its binding to the glucocorticoid receptor (26). Despite these intriguing early studies, however, it remains unclear if this region is important for chaperoning other client proteins and how client protein specificity is determined.
A corresponding C-terminal loop structure, formed by residues 652-678, is present in gp96 (24), raising an intriguing possibility of such a loop as a part of the client-binding domain (CBD) of gp96. We have addressed this hypothesis both genetically and biochemically. We found indeed that residues 652-678 of gp96 are critical for client binding and for chaperoning both TLRs and integrins, but not for ATPase activity, C-terminal dimerization or overall stabilization of the protein. We have also identified the critical residues in this region that are selectively required for folding integrins but not TLRs.

EXPERIMENTAL PROCEDURES
Constructs and Site-directed Mutagenesis-Wild type murine gp96 cDNA in a pGEM-T Easy vector (Promega) was used as templates for all PCR. Primers for gp96 mutants were listed in supplemental Table S1. ⌬CBD, NTDϩCBD, NTD⌬CBD, CTDϩCBD, CTD⌬CBD were constructed by fusion PCR utilizing respective primers with HF Pfu (invitrogen). AA1, AA2 as well as all gp96 Met mutants were generated using the QuikChange II XL Site-directed Mutagenesis kit (Stratagene). All constructs were subcloned into MigR1 retroviral vector for retrovirus production. MigR1-TLR4HA and MigR1-TLR9HA were described in Ref. 10.
Cloning and Purification of Recombinant Protein-The coding sequence of mouse gp96 73-754 ⌬CBD (652-678) was amplified by PCR and cloned into the NdeI/BamHI sites of pET15b. BL21 Star (DE3) cells bearing this plasmid were grown at 37°C in LB media under Ampicillin selection to an A 600 of 1, and protein production was induced by the addition of IPTG to a final concentration of 0.5 mM. gp96 was purified according to the published method (24), concentrated to 30 mg/ml and stored frozen at Ϫ80°C. His-tagged canine gp96 73-754 and E. coli HtpG were purified in an identical manner.
Analytical Gel Filtration Chromatography-6 -8 mg of purified gp96 ⌬CBD or gp96 73-754 were diluted to 1 ml with 10 mM Tris, pH 7.6, 150 mM NaCl, 1 mM DTT, and applied to a Superdex S200 16/60 column equilibrated in the same buffer. The elution was monitored by absorbance at 280 nm.
ATPase Measurements-ATP hydrolysis measurements were carried out using the PiPer assay kit (Stratagene), which measures the amount of free phosphate liberated in a hydrolysis reaction. gp96 73-754, gp96 ⌬CBD, and HtpG were buffer exchanged into assay buffer (40 mM Hepes, pH 7.4, 150 mM KCl, 5 mM MgCl 2 ) and diluted to 11 M in assay buffer prior to use.
The hydrolysis experiment was carried out in 96-well fluorescent assay plates. Each well contained a mixture of 5 l of ATP (0 -16 mM, in assay buffer), 45 l of protein, and 50 l of PiPer reagent (100 M Amplex Red, 4 units/ml maltose phosphorylase, 0.4 mM maltose, 2 units/ml glucose oxidase, 0.4 units/ml horseradish peroxidase, 100 mM Tris, pH 7.5). The final concentration of protein in each well was 5 M, and the final ATP concentration varied from 0 -800 M. Reactions were carried out in duplicate. Plates were incubated at 37°C, and the reaction was stopped after 100 min by burying the plastic wrapped plate in ice. Fluorescence was measured at 544 nm/590 nm (excitation/emission) on a SpectraMax Gemini XS plate reader (Molecular Devices) with 30 readings per well. Data were corrected by first subtracting the reading where the ATP concentration was 0. This was followed by subtracting the average of the readings carried out in the absence of protein for each ATP concentration tested. Kinetic parameters were determined using the program Prism according to the one site binding hyperbola equation Y ϭ V max ϫ X/(K d ϩ X).
Cell Lines-All gp96 mutant-transduced PreB cell lines were generated from parental gp96-null E4.126 PreB cell line, which was a kind gift from B. Seed (Harvard). Phoenix Eco (PE) packaging cell line from ATCC was used for retrovirus production. All culture conditions have been previously described in Ref. 16.
Retrovirus Production and Transduction-Various MigR1-gp96 mutants were transfected into PE cell line using Lipofectamine 2000 (Invitrogen). Six hours after transfection, medium was replaced by pre-warmed fresh culture medium. Virus-containing medium was collected at 24 h and 48 h after transfection. To facilitate the virus adhesion, spin transduction was performed at 1800 ϫ g for 1.5 h at 32°C in the presence of 8 g/ml hexadimethrine bromide (Sigma).
Blasticidin Selection-A blasticidin resistant gene was bicistronically expressed downstream of the target gene in the MigR1 vector. All transduced PreB cells were selected for a week in RPMI culture medium containing 0.5 g/ml blasticidin to ensure a relatively homogenous population and comparable expression level between all mutants.
Flow Cytometry-All staining protocol, flow cytometry instrumentation as well as data analysis were performed as described without significant modifications (12,16). For cell surface staining, single cell suspension of live cells was obtained and washed in FACS buffer twice. FcR blocking with or without serum blocking was performed depending on individual primary antibody used for staining. Primary and secondary anti-bodies staining were performed stepwise, with FACS buffer washing in between. Propidium iodide (PI) was added right before FACS instrumentation to gate out dead cells. For intracellular staining, cells were fixed in 4% paraformaldehyde at room temperature, then permeabilized with ice-cold methanol at Ϫ20°C. Blocking and staining procedures were similar to surface staining, except all steps were done at room temperature. Stained cells were acquired on a FACS Calibur (BD Biosciences) and analyzed using the FlowJo software (Tree Star).
SDS-PAGE, Western Blot, and co-IP-Essentially all procedures were performed as described in Ref. 16, without significant changes.
Statistical Analysis-Error bars represent the standard error of the arithmetic mean (S.E.). Student's t test was used for statistical analysis. Values of p Ͻ 0.05 (*) were considered to represent statistically significant differences.

RESULTS
Deletion of Amino Acid 652-678 Abolishes the Function of gp96-The 27 amino acid C-terminal loop structure of gp96 includes several highly hydrophobic residues including Met, Trp, and Tyr. Yet, this structure appears to be flexible and pointing out to the solvent, providing a potential site for client binding (24) (Fig. 1A). To test this possibility, we made a deletion mutant of gp96 without this region, named gp96 ⌬CBD (Fig. 1A). The 652-678 region is hereafter referred to as the putative CBD. To examine the impact of this deletion on the overall conformation of the protein, we expressed the protein recombinantly. We found that deletion of the CBD does not affect the ability of gp96 to dimerize as revealed by size exclusion chromatography (Fig. 1B). Moreover, the ATPase activity of ⌬CBD is completely preserved (Fig. 1C). To determine the functional impact of ⌬CBD on gp96 in mammalian cells, we took an advantage of a gp96 mutant PreB cell line, E4.126, which lacks gp96 (9). We were able to express both wild type (WT) and ⌬CBD in the mutant PreB cell line (Fig. 1D). More importantly, by immunoprecipitation (IP), we found that the mutant remains capable of binding to a recently discovered gp96 co-chaperone CNPY3 (Fig. 1E). CNPY3 binding site has been mapped to the N terminus of gp96 (11). These data indicate that deletion of 652-678 does not affect the overall conformation of gp96.
We next examined the chaperoning function of gp96 based on the cell surface expression level of its client proteins. As expected from the previous studies (9 -12, 27), WT gp96 can restore the cell surface expression of all gp96 clients examined including integrin ␣L, ␤2, and ␣4, TLR2 and TLR4 (Fig. 1F). However, ⌬CBD mutant was unable to rescue the expression of any of these clients. In addition, consistent with the functional restoration of TLRs, WT gp96-transduced cells responded well to Pam3CSK4 (a TLR2 ligand), LPS (a TLR4 ligand) and CpG oligonucleotide (a TLR9 ligand), measured by a NF-B-GFP reporter assay (Fig. 1G). However, ⌬CBD transduced cells were unable to activate NF-B in response to any of the TLR ligands tested despite a similar reporter expression level as demonstrated by PMA/ionomycin stimulation. We conclude thus that ⌬CBD is a novel functional null gp96 mutant despite its intact ability to dimerize, hydrolyze ATP, and to interact with its co-chaperone.
Essential Role of GWXGNMER Motif within the CBD in the Function of gp96-Comparison of gp96 with Drosophila gp93 (16) demonstrates a high sequence homology within the putative CBD region, particularly in the first half of sequence where a conserved GWXGNMER motif is present (Fig. 2A). We next performed an Ala-scanning mutagenesis and generated two mutants: AA1 with all the conserved residues in the first half of CBD mutated to Ala and AA2 with Ala substitution of all the conserved residues in the second half of CBD. Retroviral mediated transduction was performed to introduce these mutants into gp96-null PreB cells. Surface expression of integrins and TLRs on the transduced cells were examined by flow cytometry. Whereas both AA1 and AA2 gp96 mutants could be expressed to a similar level to WT gp96, only WT but not AA1 was able to rescue the cell surface expression of clients (Fig. 2B). The client expression was reduced but not abrogated completely in AA2expressing cells. Thus, the most severe defect is ⌬CBD mutant, followed by AA1 mutant and AA2 mutant. As expected, AA1 failed to functionally restore TLR responsiveness of gp96-null cells (Fig. 2C). These data demonstrated the essential roles of GWXGNMER motif in the function of gp96.
Critical Roles of Met 658 and Met 662 for Chaperoning Integrins but Not TLRs-A pair of Met residues in HSP90 is essential for chaperoning the glucocorticoid receptor (26). In gp96, Met 658 and Met 662 of the putative CBD form a similar pair that, together with the neighboring hydrophobic resides (Tyr 575 , Tyr 652 , Trp 654 , Tyr 677 , Tyr 678 ) form a hydrophobic patch on the bottom of the "V" shaped structure of the gp96 dimer (Fig. 1A). To define the role of this Met pair, we mutated both Met (M2A) as well as each of the two Met (M658A, M662A) separately into Ala using site-directed mutagenesis. Unexpectedly, the M2A mutant was able to fully restore the expression of TLR2 and TLR4, but was defective in chaperoning integrins (Fig. 3A, p Ͻ 0.01). While the phenotype of the single M658A or M662A mutants resembles that of M2A, the M2A mutant suffered the greatest defect in chaperoning integrins. Quantification of client expression after normalization for gp96 expression levels confirmed the substrate specificity of the M2A mutants (Fig.  3B). None of the Met mutants had any problems in rescuing gp96-null cells for TLR responsiveness (supplemental Fig. S1, A  and B). In fact, the TLR2-chaperoning function of M658A is even better than that of wild type gp96, although its ability to restore cell surface expression of ␣L, ␤2, and ␣4 integrins is significantly impaired. The M658/M662 region of the gp96 CBD thus appears to confer client specificity to the chaperone.
Impaired Binding of CBD Mutants to TLRs-We next examined if mutation of CBD region impairs the interaction between gp96 and its clients. To do this, we co-expressed C-terminal HA-tagged TLR9 (TLR9HA) with gp96 in PreB cells (Fig. 4A) and performed a co-immunoprecipitation (co-IP) assay. As expected, we could readily co-IP TLR9HA with WT gp96. The FEBRUARY 24, 2012 • VOLUME 287 • NUMBER 9 AA2 mutant, in which the conserved residues in the second half of the gp96 CBD is mutated to Ala, retains some binding capacity to TLR9HA, while the ⌬CBD and AA1 mutants completely lost their ability to interact with TLR9HA (Fig. 4B). In agreement with the functional data presented above, the M2A, M658A, and M662A mutants could be stably expressed (Fig.  4C), and were able to interact with TLR9HA (Fig. 4D). Consistent with the ability of M2A mutant to be fully functional in folding TLRs, we found that the M2A mutant remains capable of binding to CNPY3 (supplemental Fig. S2).

Definition of the Client-binding Domain of gp96
Soluble CBD Peptide Inhibits gp96-Integrin Interaction-We next examined the possibility of direct binding between gp96 and integrin. Because the Met pair is located within the CBD, we hypothesized that the defect of chaperoning integrins by M2A mutant, as well as single Met mutants, is due to a problem in binding to integrins. To examine this hypothesis, we co-expressed ␣L-HA with the gp96 M2A mutant in HEK293 cells by transient transfection. As shown in Fig. 5A, ␣L-HA was equally expressed in the EV, WT and M2A mutants. However, when we used co-IP to examine the interaction between ␣L-HA and gp96, we found that the M2A mutant did not interact with ␣L as strongly as WT gp96 (Fig. 5A).
To further characterize the interaction between integrins and gp96, we performed a competition experiment with a synthetic peptide that corresponds to CBD. We incubated cells with increasing concentration of a cell-permeable TAT-CBD peptide 24 h prior to cell lysis. We then performed IP analysis to examine the interaction between gp96 and ␣L-HA. We found The blow-up shows the CBD as a helix-loop structure in the C-terminal of gp96. Right, ⌬CBD mutant is modeled to preserve the overall structure of gp96. B, ⌬CBD and WT gp96 exhibit identical behavior on gel filtration chromatography. 5-8 mg of purified protein was injected for each run, and the elution was monitored by absorbance at 280 nm. The peak at 73 ml contains gp96 dimer (ϳ200 kDa). C, ⌬CBD and WT gp96 exhibit identical ATP hydrolysis rates. ATP hydrolysis was measured using the PiPer assay system, which monitors free phosphate. The protein concentration in the reaction was 5 M, and was carried out at 37°C for 100 min. D, ⌬CBD mutant can be stably expressed in the gp96-null E4.126 cells. gp96, ⌬CBD mutant and CNPY3-Flag were introduced into E4.126 cells by MigR retrovirus. Expression level was determined by SDS-PAGE. Empty virus (EV) was used as a control. E, both gp96 and ⌬CBD mutant are able to interact with CNPY3. CNPY3-Flag was immunoprecipitated followed by immunoblot (IB) for gp96 or CNPY3. F, intracellular staining of gp96 and surface expression of integrins and TLRs (solid line histogram) in gp96-null pre-B cells transduced with WT or ⌬CBD mutant. Gray histograms are isotype control antibody stain. Number represents mean fluorescence intensity (MFI) of TLRs or integrins. G, NFB-GFP reporter activation (green histogram) of cells in E after overnight (16 -18 h) stimulation with Pam3CSK4 (10 g/ml), LPS (10 g/ml), CpG (5 M), or P/I, which contains PMA (100 ng/ml) and ionomycin (2 g/ml). Gray histograms are GFP profile of unstimulated cells.
that TAT-CBD dose-dependently inhibited the ability of gp96 to interact with ␣L-HA (Fig. 5).

Importance of the Dimerization Domain but Not the ATPbinding Pocket in Addition to the 652-660 Region of gp96 for
Client Interaction-We so far demonstrated that the CBD is necessary for gp96 to bind and functionally mature client proteins. We next asked if the putative CBD is also sufficient for interaction with clients. To test this possibility, we made a series of both N-terminal and C-terminal truncation mutants of gp96 with and without the CBD. These mutants contain both the canonical signal peptide for ER insertion and the KDEL ER retention signal to ensure the proper ER targeting (Fig. 6A). We introduced these truncation mutants to gp96-null PreB cells, along with TLR9HA (Fig. 6B). Consistent with the importance of the CBD, in the absence of it, neither the N-terminal domain (NTD⌬CBD) nor the C-terminal domain of gp96 (CTD⌬CBD) was able to efficiently co-IP with TLR9HA (Fig. 6, C and D). Importantly, we found that binding to TLR9 by gp96 is dependent on both the CBD and the CTD (Fig. 6D). The NTDϩCBD mutant was unable to interact with TLR9 (Fig. 6C). We conclude that client interaction is dependent on both CBD and the C-terminal dimerization domain of gp96.

DISCUSSION
Molecular chaperones have an intrinsic ability to discern the unfolded or misfolded conformation of a protein substrate from its folded mature state. The broad substrate specificity of most chaperones suggests that they do not contain just a single specific substrate-binding surface. The ER luminal protein gp96, on the other hand, appears to be a chaperone with selective substrate specificity, making it possible to define a specific client-binding region on gp96. In this study, using biochemical and cell biological approaches, we have demonstrated that a C-terminal loop structure, formed by residues 652-678, is a critical region for gp96 to bind to TLRs and integrins. This region is not required for the overall structural integrity of gp96, since we have shown that deletion of this domain does not alter the hydrodynamic, dimerization, co-chaperone binding, or ATP hydrolysis properties of the molecule. Rather, this region appears to be the locus of a physical interaction with  both integrins and TLRs. This conclusion is further strengthened by the finding that within the 652-678 region of gp96 two residues control substrate specificity: Met 658 and Met 662 are critical for gp96 to bind integrins but not to TLRs.
The crystal structure of gp96 resembles a "twisted V" shaped structure that is dimerized via its paired C-terminal domains. This resembles the structure of the cytosolic counterpart of gp96 known as HSP90. Although the broad client repertoire of HSP90 has been defined, the client-binding domain of HSP90 remains unclear (28). Several regions of HSP90 have been implicated in protein-protein interaction, including a negatively charged region in the N terminus (221-290), another negatively charged region in the C terminus (530 -581) and the middle region with a leucine zipper structure (392-419) (29). Different regions including 381-441 and 601-677 of chicken HSP90 have been mapped to interact with progesterone receptor (PR) (30). The folding of MyoD1, a basic helix-loop-helix (bHLH) protein, appears to depend on the last 194 amino acids of the C-terminal mouse HSP90 (31). These observations suggest that HSP90 possesses disparate binding sites for its diverse clientele.
HSP90 belongs to GHKL superfamily whose members, include GyrB and MutL (23). After comparing the crystal structure of the middle domain of HSP90 to other GHKL members, it was brought to attention of a conserved hydrophobic patch   . TLR9 binding to gp96 is dependent on both CBD and the C-terminal dimerization domain of gp96. A, illustration of different gp96 N-terminal or C-terminal truncation mutants. B, immunoblot analysis of different gp96 truncation mutants as well as TLR9-HA. C, interaction between TLR9-HA and NTDϩCBD or NTD⌬CBD truncation mutants. gp96 or TLR9HA was immunoprecipitated followed by IB with indicated antibodies. D, interaction between TLR9-HA and CTDϩCBD or CTD⌬CBD truncation mutants. Co-IP was performed as in C.

Definition of the Client-binding Domain of gp96
around Trp-300 and an amphipathic loop (327-340) extending out from the middle domain of HSP90 (32). Deletion of a conserved 13 amino acid sequence between HSP90a and HSP90b (327-340) abolished the interaction between HSP90 and AKT in murine 3T3 and human 293T cells (33). This results in a proposal of a "molecular clamp" model similar to other GHKL family member, in that HSP90 binds to client between the M-domain of two monomers, and it releases the client in the more tensed structure induced by ATP hydrolysis (32). This is supported by a HSP90-Cdk4-Cdc37 trimolecular model (34). In this model, the two kinase lobes of Cdk4 bind to the M-domain and N-domain respectively, spanning across the N-M region. The larger lobe of Cdk4 interacts with M-domain via the amphipathic loop (35). As the N-M domain undergo further twisting, the proposed loop extends outwards of the HSP90 dimer, and pushes Cdk4 to the outside of the HSP90 clamp.
Recently, it has been indicated that a helix structure of the C-terminal dimerization domain could be a substrate-binding site of HSP90. The crystal structure of the dimerization domain of E. coli HtpG observed that Helix 2 (544 -565) resides in the jaw of HtpG dimer, loops out toward the space, and is the most variably positioned region in the C terminus (25). A similar helix-loop structure (548 -557) exists in HSP90 with several hydrophobic residues Met 546 , Met 550 , Ala 551 , Leu 553 , Phe 554 , and Ala 556 lining on the hydrophobic interface of Helix 2. A follow-up mutagenesis study identified that a sequence of ANMERIMKA within the helix 2 region is important in binding to unliganded glucocorticoid receptor, and protecting its hydrophobic groove in the ligand binding domain (26).
Based on the knowledge of HSP90 and the crystal structure of gp96, a similar helix-loop structure (652-678) of gp96 emerged as an attractive candidate as the CBD of gp96. The loop structure of the CBD is potentially more flexible compared with other regions of gp96, and more importantly, it protrudes out toward solvent, making it readily available for mediating client protein interactions. Further support for the identification of this loop as a critical part of the gp96 CBD comes from its hydrophobic character, whereby Met, Tyr, and Trp residues are exposed to solvent, a relatively rare occurrence that implies a potential role in forming protein-protein interactions.
We have presented definitive evidence that identifies a C terminus loop structure of gp96 (652-678) is a critical region of the client binding domain of gp96 in chaperoning both integrins and TLRs. While this work identifies the locus of client-gp96 interactions, it is likely that gp96 may utilize different chaperoning mechanisms to fold TLRs comparing with integrins. This is supported by our findings that integrins but not TLRs are dependent on the Met pairs in the CBD region, based on the following evidence: (i) M2A mutant is able to fully restore the surface expression of both TLR2 and TLR4 (Fig. 3, A  and B), as well as the functional responsiveness to multiple ligands (supplemental Fig. S1). It also retains its ability to bind to a TLR-specific co-chaperone CNPY3 (supplemental Fig. S2).
(ii) However, M2A mutant is significantly impaired in conferring surface expression of multiple integrins (Fig. 3). Strikingly, although single M658A mutant is even more efficient than wild type gp96 in facilitating TLR2 surface expression, its ability to restore ␣L integrin expression is significantly compromised (Fig. 3). (iii) The defect of M2A mutant in facilitating surface expression of integrin correlates directly to its reduced ability to bind to ␣L (Fig. 5). The different folding mechanisms of TLR and integrins is also highlighted by the fact that TLR folding is dependent on both the ATPase activity and ATP-binding property of gp96, while the folding of integrin is only dependent on ATP binding (9,11). In this regard, several questions remain unanswered. It is unclear how the twisted V structure of gp96 is shifted to a "closed" catalytic dimer, which is the ATP hydrolytic conformation of HSP90. It is also unclear how this conformation of gp96 is regulated by the binding or hydrolysis of ATP or ADP. Finally, it remains to be determined whether clients or cochaperones play any role in the conformational change of gp96, and it is unknown if the folding of integrins requires an as yet unidentified co-chaperone.
We believe that the identification of a discrete client-binding region in gp96 represents a major advance in route toward developing a gp96-specific inhibitor. A recent study reported that a peptide sequence targeting part of the charged linker region of gp96 was able to inhibit endotoxic shock when administered in vivo (36). It is noteworthy that CBD-targeted peptide is able to significantly inhibit gp96-integrin interaction. We are pursuing a CBD-targeted strategy to generate gp96-specific inhibitors for the treatment of inflammation and other diseases.