Distinct hsp70 Domains Mediate Apoptosis-inducing Factor Release and Nuclear Accumulation*

Although hsp70 antagonizes apoptosis-inducing factor (AIF)-mediated cell death, the relative importance of preventing its release from mitochondria versus sequestering leaked AIF in the cytosol remains controversial. To dissect these two protective mechanisms, hsp70 deletion mutants lacking either the chaperone function (hsp70-ΔEEVD) or ATPase function (hsp70-ΔATPase) were selectively overexpressed before exposing cells to a metabolic inhibitor, an insult sufficient to cause mitochondrial AIF release, nuclear AIF accumulation, and apoptosis. Compared with empty vector, overexpression of wild type human hsp70 inhibited bax activation and reduced mitochondrial AIF release after injury. In contrast, mutants lacking either the chaperone function (hsp70-ΔEEVD) or the ATP hydrolytic domain (hsp70-ΔATPase) failed to prevent mitochondrial AIF release. Although hsp70-ΔEEVD did not inhibit bax activation or mitochondrial membrane injury after cell stress, this hsp70 mutant co-immunoprecipitated with leaked AIF in injured cells and decreased nuclear AIF accumulation. In contrast, hsp70-ΔATPase did not interact with AIF either in intact cells or in a cell-free system and furthermore, failed to prevent nuclear AIF accumulation. These results demonstrate that mitochondrial protection against bax-mediated injury requires both intact chaperone and ATPase functions, whereas the ATPase domain is critical for sequestering AIF in the cytosol.

Disruption of the outer mitochondrial membrane releases toxic proteins, including cytochrome c and apoptosis-inducing factor (AIF) 3 that activate caspase-dependent and -independent pathways responsible for apoptotic cell death. In a prior study in our laboratory, neither a pancaspase inhibitor nor a specific caspase 3 inhibitor completely prevented apoptosis in renal cells exposed to metabolic inhibitors (1), suggesting that at least a portion of the observed apoptosis is mediated by non-caspase-dependent factors including AIF. Nuclear translocation of AIF activates endogenous endonucleases leading to degradation of native DNA into 50-kilobase fragments, peripheral chromatin condensation, and nuclear shrinkage (2)(3)(4) that is sufficient to cause apoptosis (5,6). AIF mediates apoptosis in multiple cell lines after diverse insults including ischemia (7)(8)(9)(10), oxidant stress (11), ethanol (12), or p53 exposure (8).
AIF is synthesized in the cytosol as a 67-kDa precursor protein that contains a mitochondrial-localizing sequence (13,14). After being imported into mitochondria, the mitochondrial-localizing sequence is cleaved, resulting in the accumulation of the mature, 57-kDa AIF protein (13). Under normal circumstances AIF may facilitate electron transport, since it exhibits robust oxidoreductase activity (14). AIF-mediated cell death is a two-step process that involves initial release from the intramembranous mitochondrial space into the cytosol followed by nuclear uptake and accumulation (5,13,15,16).
hsp70, an inducible cytoprotectant protein, antagonizes apoptosis by interfering with multiple checkpoints in the apoptosis pathway (6,(17)(18)(19)(20)(21). Recent evidence suggests that hsp70 inhibits apoptosis by interfering with events upstream of mitochondrial membrane permeabilization (22,23) that ultimately decrease the activation of bax (22,23), a pro-apoptotic member of the BCL2 family. hsp70 contains two distinct functional regions; that is, a carboxyl-terminal EEVD motif responsible for substrate binding and refolding (referred to as the "chaperone function") and an amino-terminal ATPase domain that facilitates the release of client proteins (24,25). The EEVD motif of the constitutively expressed member of the hsp70 family (hsc70) has recently been shown to interact with several other co-chaperones including hsp40 and hsp90 to mediate the folding and assembly of newly synthesized proteins (26,27). In addition, the ATP binding domain interacts with several regulatory proteins including hip, hop, and bag that either stimulate or inhibit hsp70 chaperone activity (28) and, thereby, provide an elegant complexity to protein-protein interactions. Although the chaperone function of hsp70 (i.e. the EEVD motif) was assumed to mediate cytoprotection (18, 29 -31), recent evidence suggests that the ATPase domain also ameliorates cell injury (30,(32)(33)(34).
Prior reports emphasize that hsp70 inhibits nuclear AIF accumulation solely by binding to and sequestering it in the cytosol (6,15). Garrido and co-workers (35) substantiated this protective mechanism by identifying the AIF sequence responsible for binding hsp70. Our laboratory demonstrated that hsp70 ameliorates apoptosis after exposure to metabolic inhibitors at least in part by inhibiting mitochondrial AIF release (36). Together, these reports suggest that hsp70 limits nuclear injury by antagonizing the effect of AIF at multiple checkpoints in the apoptotic pathway.
Although hsp70 reduces both AIF (36) and cytochrome c (30) release from mitochondria, the mechanism is incompletely characterized. Members of the BCL2 family, including bax, have been intensely scrutinized as potential candidates of mitochondrial membrane injury. Bax has been shown to release mitochondrial cytochrome c (37)(38)(39) and AIF (40) and is activated by exposure to metabolic inhibitors (38), and bax activation is sufficient to induce apoptosis (41). Activation and membrane insertion of bax requires a conformational change that exposes both the amino and carboxyl termini (42) and is antagonized by antiapoptotic members of the BCL2 family, including bcl2 itself (41). Our laboratory has shown that exposure to metabolic inhibitors, an insult that released both AIF (36) and cytochrome c (7), altered the ratio of bax:bcl2 in a manner that promotes apoptosis (43).
The present study examines the hypothesis that two distinct hsp70 domains (each with independent functions) inhibit nuclear accumulation of AIF. Both the hsp70 protein refolding and ATPase domains are required to prevent bax activation and mitochondrial AIF release. In contrast, deletion of the ATPase domain disrupts hsp70-AIF interaction and permits leaked AIF to accumulate in nuclei. These results show that hsp70 exerts domain-specific effects that collectively antagonize AIFmediated injury and identify potential therapeutic targets in the apoptotic pathway.

EXPERIMENTAL PROCEDURES
Materials-All reagents were obtained from Sigma unless otherwise indicated.
Cell Culture-Proximal tubule opossum kidney (OK) cells obtained from the American Type Culture Collection (ATCC CRL-1840) were grown in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% bovine calf serum. To measure active bax, previously characterized proximal tubule cells derived from the immortalized mouse (MPT cells (44)) were grown under similar conditions. Cells were used within 72 h of achieving confluence.
ATP Depletion-Cells were incubated for 1-2 h at 37°C in glucosefree medium (Dulbecco's modified Eagle's medium, Invitrogen 23800-014) that contained 5 mM sodium cyanide, an insult sufficient to cause apoptosis (7,45,46,47). This maneuver results in rapid, severe ATP depletion and is largely reversible with the removal of cyanide and the addition of exogenous glucose (48). In control, parallel medium changes were performed using Dulbecco's modified Eagle's medium containing 10 mM glucose.
AIF Release-To assess the release of mitochondrial AIF into the cytosol, intact cells were incubated with digitonin (1.5 mg/ml ϫ 10 -20 min at 4°C) and a mixture of protease inhibitors (5 M 4 -2(-aminoethyl)benzenesulfonyl fluoride hydrochloride (AEBSF-HCl), 10 nM leupeptin, 1.5 nM aprotinin, 10 nM E-64, and 5 M EDTA, pH 7.40; Calbiochem-Novabiochem) as described by our laboratory (7). This procedure distinguishes mitochondrial from cytosolic AIF (7) and does not itself cause mitochondrial dysfunction (47). A recent study from our laboratory showed that AIF release positively correlates with the duration of ATP depletion (1-2 h) and readily persists for more than 90 min after injury occurs (36). Furthermore, this maneuver does not release the majority of cytochrome c (7) that binds to cardiolipin and resides within the mitochondrial matrix (49). The identity of the immunoreactive AIF bands was confirmed in whole cell lysates exposed to 2% SDS as previously described (7).
Overexpression of Inducible hsp70 (hsp72kDa) and bax-Wild type or well characterized hsp70 deletion mutants were individually overexpressed ( Fig. 1). Wild type hsp70 content was increased either by transient heat stress (42.5 Ϯ 0.5°C for 45 min) in a temperature-regulated incubator followed by incubation at 37°C for 16 -18 h as previously described (47) or by co-infection with adenoviruses containing human hsp70 and green fluorescent protein (AdTR5/hsp70-GFP; 40 m.o.i.) expressed on separate cistrons as well as a tetracycline-regulated pro-moter (AdCMV/tTA; 20 m.o.i.) as previously reported (30,50). Control cells were co-infected with AdTR5/GFP and AdCMV/tTA. Induction of hsp70-⌬EEVD, a deletion mutant of hsp70 lacking the carboxyl-terminal EEVD sequence that is essential for efficient peptide binding and re-folding, was achieved by adenoviral co-infection as previously described and does not alter the content of native hsp70 (20). Infection efficiency was Ͼ90 -99% as determined by direct visualization of GFP in cells infected with increasing doses of adenovirus with an m.o.i. that varied between 40 and 100. Cells were co-infected with hsp70-containing adenovirus for 24 h at 37°C followed by removal of the virus and a 24-h recovery period. The hsp70-⌬ATPase mutant was constitutively overexpressed in renal epithelial cells in our laboratory as previously described for wild type hsp70 (46). Selective overexpression of inducible hsp70 (72kDa), hsp70-⌬EEVD, and hsp70-⌬ATPase was assessed by immunoblot analysis.
Bax was selectively overexpressed by co-infecting cells with an 40 m.o.i. of adenovirus (AdV) containing human bax and 20 m.o.i. AdV containing Cre that have been previously characterized (51). Cells were exposed to an equivalent m.o.i. of adenovirus containing empty vector or human bax plus Cre for 16 h at 37°C. The viruses were then removed, and the cells were permitted to incubate for an additional 24 h. In preliminary studies, this maneuver increased steady state bax content as assessed by immunoblot analysis and caused morphologic evidence of apoptosis.

hsp70 Domains Antagonize AIF
To assess AIF-hsp70 interaction in the cytosol, immunoprecipitation was performed in digitonin-treated samples using a polyclonal rabbit antibody directed against AIF (Exalpha Biologicals, Watertown, MA, catalog number X1108P; 2 g/mg protein/ml of immunoprecipitation buffer) as recently described by our laboratory for cytochrome c (7). Apyrase (10 units/ml), a compound that causes ATP hydrolysis, was used to prevent the ATP-mediated release of AIF from hsp70 (1). Protein interaction was assessed by probing the membranes with monoclonal antibodies directed against the inducible member of the hsp70 family (hsp72kDa; StressGen) and AIF (Santa Cruz, catalog number sc-13116). To validate direct assessment of active bax by immunoblot analysis, bax was also immunoprecipitated from whole cell lysates using the 6A7 antibody specific for active bax under non-denaturing conditions in the presence of 2% CHAPS as previously described (23,52,53). After SDS-PAGE and transfer, total bax content was assessed using the 5B7 antibody as described above. To minimize basal activation of bax, cells were incubated for 14 h in 2% fetal bovine serum before exposure to metabolic inhibitors.
Nuclear Isolation-Intact nuclei were separated from "non-nuclear" cell components (cytosol, other organelles, and membrane fragments) using a commercial kit (BioVision Research Products, Mountain View, CA; catalog number K266-100) as previously reported by our laboratory (48). Adequacy of the nuclear purification was assessed by measuring the content of fibrillarin, a nucleolar protein (1:1000 dilution, Abcam catalog number ab4566; Cambridge, MA), and the non-nuclear protein tubulin (1:10,000 dilution, catalog number T6199) in representative samples using immunoblot analysis.
Protein Assay-Protein concentrations were determined with a colorimetric dye binding assay (BCA Assay, Pierce). Results are expressed as mg of protein/ml.
Statistical Analysis-Data are expressed as the mean Ϯ S.E. Multiple experimental groups were compared using two-tailed analysis of variance with standard statistical software (Excel, Microsoft Corp., Santa Monica, CA). A p value of Յ0.05 was considered significant.

RESULTS
Under basal conditions, minimal mitochondrial AIF could be detected in the cytosol of cells overexpressing only empty vector (GFP). Progressive leakage of mitochondrial AIF accompanied transient exposure to a metabolic inhibitor, whereas wild type hsp70 completely inhibited mitochondrial AIF release ( Fig. 2A). Selective overexpression of an hsp70 mutant lacking either chaperone function (hsp70-⌬EEVD) or ATPase activity (hsp70-⌬ATPase) failed to prevent leakage of mitochondrial AIF after cell injury. Infection with adenovirus containing either wild type or hsp70 mutants per se did not alter AIF leakage under base-line conditions compared with empty vector (lane 1 versus lanes 4, 7, and 10) nor did differences in loading account for the observed changes in AIF as indicated by the content of ␤-actin (lower panel). Densitometric analysis revealed that only wild type hsp70 afforded significant protection against mitochondrial AIF leakage (p Ͻ 0.05 versus base line; Fig. 2B).
To investigate a potential mechanism of mitochondrial protection, cells infected with wild type hsp70, hsp70-⌬EEVD, or empty vector were compared for their ability to inhibit the activation of bax, a proapoptotic member of the BCL2 family that causes mitochondrial membrane injury. Active bax could not be adequately detected in proximal tubule cells derived from opossum kidney (data not shown). In mouse proximal tubule cells, however (Fig. 3A), exposure to a metabolic inhibitor resulted in marked bax activation in the presence of empty vector or hsp70-⌬EEVD using an antibody specific for the active form of the protein (Refs. 23, 53, and 55; Fig. 3A, upper immunoblot). Compared with empty vector or the ⌬EEVD mutant, wild type hsp70 markedly inhibited bax activation immediately after and during recovery from cell injury. None of these maneuvers altered total bax content (lower panel).
To confirm that sample preparation (including exposure to detergent and dithiothreitol) did not interfere with direct detection of active bax by the conformation-specific antibody (6A7), aliquots of each sample were subjected to immunoprecipitation under non-denaturing conditions, as described by others (23,52,53) and separated by SDS-PAGE, and the membrane was probed with the 5B7 antibody directed against total bax (Fig. 3B, upper panel). After confirming adequate stripping hsp70 Domains Antagonize AIF MARCH 24, 2006 • VOLUME 281 • NUMBER 12 (Fig. 3B, middle panel), the membrane was re-probed for active bax with the 6A7 antibody. Both direct immunoblot (Fig. 3A) and immunoprecipitation followed by immunoblot analysis (Fig. 3B) revealed that exposure to a metabolic inhibitor markedly activates bax. Compared with empty vector, exposure to adenovirus containing human bax (and Cre) increased steady state bax content (Fig. 4A) and caused mitochondrial AIF leakage in intact cells (Fig. 4B).
Exposure to a metabolic inhibitor was associated with progressive AIF accumulation in isolated nuclei (Fig. 5A), an event that precedes the onset of DNA degradation or characteristic morphologic features of apoptosis (1,56). In contrast, prior heat stress, a nonspecific stimulus that induces multiple stress proteins including hsp70 (7), was associated with a marked decrease in nuclear AIF uptake during ATP depletion and recovery. The content of fibrillarin, a 33.8-kDa nucleolar protein, remained unchanged (Fig. 5A, lower panel). In addition, cell injury did not alter non-nuclear AIF content (data not shown), supporting prior reports that AIF leakage represents only a small fraction of the total mitochondrial AIF pool (7,35). Nuclear AIF detected under base-line conditions (i.e. before injury) in the presence of empty vector or hsp70 is attributable to a small amount of contamination of the nuclear fraction (36). Adequate nuclear purification was confirmed by immunoblot analysis through relative enrichment of fibrillarin and de-enrichment of tubulin, a non-nuclear, structural protein that resides exclusively in the cytoplasm (Fig. 5B).
To identify the specific protein responsible for inhibiting nuclear AIF accumulation in cells subjected to transient heat stress, hsp70 was selectively overexpressed in a dose-dependent manner to a level equivalent or greater than that achieved by hyperthermia (Fig. 6A). Overexpression of wild type hsp70 at a dose of 60 m.o.i. reproduced the protective effect of prior heat stress on nuclear AIF uptake in ATP-depleted cells (Fig. 6B,  upper panel). Immunoblot analysis of fibrillarin content confirmed that each nuclear preparation contained equivalent amounts of protein (Fig.  6B, lower panel).
Because leaked mitochondrial AIF co-immunoprecipitates with wild type hsp70 after ATP depletion (36), the domain (and hsp70 function) responsible for preventing nuclear AIF accumulation was investigated by selectively overexpressing well characterized hsp70 deletion mutants. Infection with an adenovirus containing either hsp70-⌬EEVD or wild type hsp70 resulted in a similar degree of protein expression as detected by a rabbit polyclonal antibody that detects both hsp70-⌬EEVD and the wild type protein (Fig. 7A). Importantly, overexpression of hsp70-⌬EEVD did not alter the steady state content of native hsp70, as determined by a mouse monoclonal antibody specific for the wild type protein only (Fig. 7B). Although hsp70-⌬EEVD failed to prevent mitochondrial AIF release (Fig. 2), this hsp70 mutant was capable of interacting with cytosolic AIF, especially during ATP depletion and early recovery (Fig. 7C). As a result of this interaction (and despite its failure to prevent AIF release), the ⌬EEVD mutant reduced nuclear AIF accumulation after exposure to a metabolic inhibitor (Fig. 7D). These The same membrane was re-probed for total bax using a 5B7 antibody (lower panel). Each lane contains 40 g of total protein separated by 10% SDS-PAGE. ATP depl, ATP depletion. B, detection of active bax before and after cell injury was compared using immunoprecipitation (IP) of 400 g of total protein with an antibody directed against active bax (6A7) under non-denaturing conditions followed by immunoblot (IB) analysis using an antibody directed against total bax (5B7; upper panel). The membrane was chemically stripped before re-probing the same membrane with the active bax-specific antibody (6A7; lower panel), and adequacy of the membrane stripping was assessed by exposing the membrane to secondary antibody alone (middle panel).  hsp70 Domains Antagonize AIF results show that distinct mechanisms of protection by hsp70 prevent nuclear AIF accumulation after cell injury.
To assess the role of the ATPase domain in mediating cytoprotection against AIF, a deletion mutant lacking the ATP hydrolytic domain was overexpressed in a stable cell line. Transfection markedly increased the steady state content of the 35-kDa ATPase mutant (Fig. 8A, left-hand  panel) without altering the content of native hsp70 (right-hand panel). In contrast to either wild type or hsp70-⌬EEVD, hsp70-⌬ATPase failed to prevent nuclear AIF accumulation after exposure to a metabolic inhibitor (Fig. 8B). Unlike wild type or hsp70-⌬EEVD, the ⌬ATPase mutant did not co-immunoprecipitate with AIF before, during, or after exposure to a metabolic inhibitor (Fig. 8C, upper panel). As reported for non-transfected cells (36), minimal interaction between AIF and endogenous hsp70 could be detected before or after injury in cells infected with empty vector (GFP; first through third lanes 1-3). Each of the immunoprecipitates contained similar amounts of AIF (lower panel).
In intact cells the hsp70-⌬ATPase mutant was less abundantly expressed than either wild type or hsp70-⌬EEVD when using an antibody that detects both protein species (Fig. 8C). To exclude the possibility that insufficient expression of the ⌬ATPase mutant accounted for the apparent lack of co-immunoprecipitation with AIF, hsp70-AIF interaction was also examined in a cell-free system using purified proteins (described under "Experimental Procedures"). hsp70 co-immunoprecipitated with AIF in a dose-dependent fashion in the cell-free assay (Fig. 8D), mirroring the interaction observed between these two proteins in the intact cell. Unlike wild type protein, equimolar amounts of purified hsp70-⌬ATPase did not interact with AIF, confirming that the ATPase domain facilitates AIF-hsp70 interaction and is important for preventing nuclear translocation of leaked AIF.

DISCUSSION
Although prior studies emphasize cytosolic retention of AIF as the major mechanism of hsp70-mediated protection (6,35), we show that mitochondrial protection by hsp70, rather than cytosolic AIF retention, is the earliest checkpoint at which hsp70 inhibits caspase-independent apoptosis. A recent report by our laboratory showed that hsp70 reduced mitochondrial AIF leakage in a dose-dependent manner in cells exposed to a metabolic inhibitor (36). The present study suggests that hsp70 reduces AIF leakage, at least in part, by inhibiting bax activation, thereby  A, wild type (wt) and hsp70-⌬EEVD content was examined in whole cell lysates obtained from cells infected with empty vector alone, hsp70-⌬EEVD, or wild type hsp70 using a rabbit polyclonal antibody (Ab) that recognizes both proteins. Each lane contains 10 g of total protein. B, hsp70 content after infection with adenovirus-containing empty vector, hsp70-⌬EEVD, or wild type hsp70 in whole cell lysates before and after ATP depletion using a mouse monoclonal antibody that recognizes only wild type hsp70. In contrast to infection with wild type hsp70, infection with empty vector or hsp70-⌬EEVD minimally affected the content of endogenous hsp70. Each lane contains 10 g of total protein. C, AIF was immunoprecipitated (IP) from samples containing 200 g of total protein obtained from digitonin-exposed cells using a rabbit polyclonal anti-AIF antibody (described under "Experimental Procedures"). After separation by 10% SDS-PAGE, hsp70 content in cells infected with empty vector or hsp70-⌬EEVD AdV was examined by immunoblot (IB) analysis using an anti-hsp70 antibody that recognizes both the wild type and ⌬EEVD limiting pro-apoptotic changes in membrane permeability. Inhibition of bax by hsp70 requires both the EEVD and ATPase domains, since only the intact protein prevented bax activation in cells subjected to transient ATP depletion (Fig. 3). Our data regarding the protective effect of hsp70 in reducing bax activation and AIF release could explain prior observations that intact (but not mutant) hsp70 inhibited mitochondrial cytochrome c release (30) and preserved mitochondrial membrane potential in cells transfected with hsp70 or injected with intact, purified protein before stress (6). In a recent study by Stankiewicz et al. (23), wild type hsp70, not the ATPase mutant, prevented the oligomerization and translocation of bax to the mitochondrial membrane and decreased apoptosis after heat stress (23). The present study extends these observations by demonstrating that hsp70 inhibits bax activation after pharmacologic ischemia, requires the hsp70-EEVD domain, and prevents the leakage of AIF, a key mediator of apoptotic cell death. Because bax releases both cytochrome c (39,57) and AIF (40) and its activation in ischemia-induced apoptosis in the brain (58), heart (59), and kidney (60) has been suggested, maneuvers that inhibit bax may have therapeutic potential.
How does hsp70 inhibit bax? Although interaction between hsp70 and bax could regulate bax and was observed in a single report (61), other investigators have not (1,23). Alternatively, hsp70 could inhibit the activation of Jun-N-terminal kinase (23,62), potentially reducing its ability to recruit bax to the mitochondrial membrane (63). In addition to preventing bax activation, bcl2 may be an attractive target for protection by hsp70, since it antagonizes the effect of bax on the mitochondrial membrane and reduces both mitochondrial AIF release (64,65) and cytochrome c-dependent caspase activation (30,64,66). Furthermore, hsp70 interacts with bcl2, especially during stress (1) and may facilitate the anti-apoptotic effect of bcl2 on bax (67,68). Last, by regulating stress kinases, hsp70 could indirectly alter the phosphorylation of bax at one or more recently identified serine sites that mediate its activation (69,70).
Regardless of the mechanism by which hsp70 antagonizes bax activation, we show that bax is sufficient to cause AIF leakage (Fig. 4B), the primary mediator of caspase-independent cell death (56). Although the event(s) responsible for mitochondrial AIF release are controversial (71), AIF leakage appears to be independent of cytochrome c, at least in some circumstances (72,73). AIF release is inhibited by bongkrekic acid, a membrane pore transition inhibitor (7,64), suggesting that AIF traverses the mitochondrial membrane pore complex. Interestingly, bax (as well as other members of the BCL2 family) has been implicated in membrane pore transition dysregulation (73,74).
A unique aspect of the present study is the observation that distinct hsp70 domains mediate two major anti-AIF effects. Only wild type hsp70 inhibits mitochondrial AIF release (Fig. 2), suggesting that interaction between the ATP binding and chaperone domain is required to prevent mitochondrial injury (see below). In contrast, deletion of the ATPase domain completely disrupts hsp70-AIF interaction in the intact cell (Fig. 8C) as well as in a cell-free system (Fig. 8D), suggesting that this domain is required to retain AIF in the cytosol. Because mitochondrial AIF release is virtually eliminated by overexpressing wild type hsp70, it is not possible to determine whether cytosolic binding/sequestration of AIF also contributes to decreased nuclear AIF accumulation. To address this limitation, the effect of an hsp70 mutant that fails to prevent AIF release was examined. Although hsp70-⌬EEVD mutant did not significantly reduce mitochondrial AIF release (Fig. 2), it interacted with AIF in the cytosol after mitochondrial injury (Fig. 7C) and decreased nuclear AIF accumulation (Fig. 8B). These observations suggest that hsp70 lim- Overexpression of the ⌬ATPase mutant did not alter the content of native, inducible hsp70 (right-hand panel). hsp70 appears relatively abundant in both lanes due to the large amount of total protein loaded (30 g). The rabbit polyclonal antibody detects both the wild type and hsp70-⌬ATPase proteins. B, AIF accumulation in nuclei isolated from cells infected with AdV-containing empty vector, wild type (wt) hsp70, or hsp70-⌬EEVD and from a stable cell line transfected with human hsp70-⌬ATPase at base line (B), immediately after ATP depletion (ATP Depl), and after 30 min of recovery (R 30 ). Samples were run on a single membrane and then separated by experimental groups to improve clarity. Fibrillarin content is shown to control for loading. Each lane contains 40 g of total protein. C, AIF was immunoprecipitated (IP) from samples containing 200 g of total protein obtained from digitonin-exposed cells using a rabbit polyclonal anti-AIF antibody. After separation by 10% SDS-PAGE, hsp70 content was assessed before and after injury in cells that overexpress either the wild type or hsp70-⌬ATPase mutant using an anti-hsp70 antibody that recognizes both proteins (upper panel). IB, immunoblot. hsp70 and hsp70-⌬ATPase (ϳ36 kDa) were localized by combining whole cell lysates obtained from cell lines that overexpress one of these proteins (right hand lane). Lower panel, total AIF in each sample. Interaction between wild type hsp70 and AIF was greatest after ATP depletion and 30 min of recovery. Results are representative of at least two separate experiments. D, interaction between wild type or hsp70-⌬ATPase and AIF in a cell-free system (described under "Experimental Procedures"). Equimolar amounts of purified wild type or hsp70-⌬ATPase protein were incubated with 200 g of whole cell lysate obtained from normal cells and then immunoprecipitated with rabbit polyclonal antibody directed against AIF. After separation by SDS-PAGE, membranes were probed for hsp70 content using a mouse monoclonal antibody that recognizes both the wild type and hsp70-⌬ATPase proteins. hsp70 and hsp70-⌬ATPase were identified from combined whole cell lysates that contained one of these two proteins (right-hand lane).
hsp70 Domains Antagonize AIF its nuclear AIF accumulation by reducing mitochondrial AIF release as well as by sequestering AIF in the cytosolic compartment.
In the present study, hsp70 content exceeded that elicited by heat stress (Fig. 5C). Because hsp70 content is lower during physiologic conditions, hsp70 would be insufficient to completely prevent mitochondrial AIF release. Under these circumstances, hsp70 would also limit nuclear AIF uptake by binding and sequestering it in the cytosol. In fact, interaction between wild type or hsp70-⌬EEVD and AIF was greatest during ATP depletion and recovery (Figs. 7C and 8C), precisely the time period when AIF leakage occurs (36). Taken together with the report by Ravagnan et al. (6) that emphasized the role of hsp70 in sequestering cytosolic AIF, we suggest inhibition of AIF release also antagonizes its pro-apoptotic action.
In addition to pharmacologic ischemia (75), domain-specific cytoprotection by hsp70 has been observed after hyperthermia (30,33) and tumor necrosis factor (76) or nitric oxide (61) exposure, supporting our contention that hsp70 domains independently antagonize the proapoptotic effects of AIF. Although the ATPase domain is required for AIF interaction in the present study, AIF might not directly interact at this hsp70 site. AIF could bind to hip, hop, or bag-1 regulatory proteins that associate with the ATPase domain (32,77,78). Although all three regulatory proteins were abundant in renal cell lysates, none was detected in AIF immunoprecipitates (data not shown). Alternatively, the ATPase domain could facilitate interaction between AIF and the carboxyl-terminal peptide binding domain as has been reported for other hsp70 substrates (79). Finally, AIF could bind to one of more of the co-chaperones that directly interact with the hsp70-EEVD domain, including hsp40 or hsp90 (26,27).
Our results regarding the hsp70 domain responsible for binding AIF in the cytosol partially contradict a single prior study. In contrast to the present study, Ravagnan and colleagues reported that deletion of the ATPase domain did not disrupt AIF binding (6). Their study differed from ours in several important respects. First, these investigators simultaneously overexpressed both AIF and hsp70, potentially causing nonspecific protein-protein interactions. This may be important, since prior studies show that only a fraction of mitochondrial AIF escapes after cell injury (35,36). Second, the relatively large lactate dehydrogenase tag used to immunoprecipitate wild type and hsp70 mutants could alter AIF-hsp70 interaction. Third, AIF-hsp70 interaction was examined under conditions of normal cell growth as apposed to physiological injury. The present study avoided many of these potential problems by 1) overexpressing only hsp70 or its deletion mutants, 2) obtaining cytosolic samples by selectively permeabilizing the plasma membrane with digitonin, a maneuver that does not itself release mitochondrial AIF (7,36), 3) expressing hsp70 and GFP as separate proteins rather than as an hsp70 fusion protein and most importantly, 4) assessing dynamic (versus static) protein-protein interaction under conditions of physiologic stress. Despite these apparent differences, both studies confirm that hsp70 sequesters mitochondrial AIF in the cytosol, decreases nuclear AIF accumulation, and inhibits apoptosis.
Because AIF leakage and nuclear translocation accompany diverse cell stressors that result in apoptosis or necrosis (65), the present observations may have broad implications. Apoptosis and necrosis co-exist in injured tissues, and both are ameliorated by hsp70 (75,80). Furthermore, elaboration of heat stress proteins acting as molecular chaperones has been implicated in mediating the resistance of tumor cells to apoptotic stimuli (81). By interfering with multiple checkpoints in the cell death cascade, stress proteins in general and hsp70 in particular may be ideal candidates for promoting cell survival (82)(83)(84). hsp70 affords opti-mal protection against apoptosis since it inhibits both mitochondrial AIF release and retains leaked AIF in the cytosol.