Functional interaction of heat shock protein 90 and Beclin 1 modulates Toll-like receptor-mediated autophagy

Abstract

Autophagy is one of the downstream effector mechanisms for elimination of intracellular microbes following activation of the Toll-like receptors (TLRs). Although the detailed molecular mechanism for this cellular process is still unclear, Beclin 1, a key molecule for autophagy, has been suggested to play a role. Heat shock protein 90 (Hsp90) is a molecular chaperone that regulates the stability of signaling proteins. Herein, we show that Hsp90 forms a complex with Beclin 1 through an evolutionarily conserved domain to maintain the stability of Beclin 1. In monocytic cells, geldanamycin (GA), an Hsp90 inhibitor, effectively promoted proteasomal degradation of Beclin 1 in a concentration-dependent (EC₅₀ 100 nM) and time-dependent (t₅₀ 2 h) manner. In contrast, KNK437/Hsp inhibitor I had no effect. Hsp90 specifically interacted with Beclin 1 but not with other adapter proteins in the TLR signalsome. Treatment of cells with GA inhibited TLR3- and TLR4-mediated autophagy. In addition, S. typhimurium infection-induced autophagy was blocked by GA treatment. This further suggested a role of the Hsp90/Beclin 1 in controlling autophagy in response to microbial infections. Taken together, our data revealed that by maintaining the homeostasis of Beclin 1, Hsp90 plays a novel role in TLR-mediated autophagy.—Xu, C., Liu, J., Hsu, L. -C., Luo, Y., Xiang, R., Chuang, T. -H. Functional interaction of Hsp90 and Beclin 1 modulates Toll-like receptor-mediated autophagy.

The innate immune system is responsible for early detection of invading microorganisms. In contrast to the high specificity and memory of acquired immunity, the innate immunity initiates host defense responses by recognizing diverse pathogen-associated molecular patterns from microbes through a number of pattern-recognition receptors (PRRs) (1, 2). Toll-like receptors (TLRs) comprise a major group of such PRRs. To date, 13 TLRs have been identified in mammalian cells; 10 of which (TLR1 to TLR 10) are expressed in human cells. These TLRs are located either on the cell surface or in intracellular compartments and detect a wide variety of microbial pathogens with diverse structures from lipids, lipoproteins, glycans, and proteins to nucleic acids (3–6).

On activation, TLR interacts with TIR-domain-containing adapter proteins in the MyD88 family to initiate both an NF-κB signaling cascade and an IRF signaling cascade, leading to the production of proinflammatory cytokines and type I interferons (IFNs) for induction of inflammatory responses and antiviral responses (7, 8). Currently, 5 members have been identified in this MyD88 family: MyD88, TIRAP/Mal, TRIF/TICAM-1, TIRP/TRAM, and SRAM. TLR4 and TLR3 also utilize a MyD88-independent pathway, recruiting TRIF to activate IRF3 and NF-κB. IRF3 activation involves a TBK1 and IKKε/IKKi complex. Activation of NF-κB involves TRAF6 and RIP. With the exception of TLR3, all TLRs signal through a MyD88-dependent pathway. In this pathway, formation of a MyD88/IRAK1/IRAK4/TRAF6 complex activates TAK1, leading to the activation of NF-κB and production of proinflammatory cytokines. In plasmacytoid dendritic cells, TLR7, TLR8, and TLR9 are able to activate IRF7 through the MyD88/IRAK1/IRAK4/TRAF6 complex, leading to the production of type I IFNs (9, 10).

In addition to these two pathways, autophagy has recently been shown to be a downstream effector mechanism by which TLRs eliminate invading microbes (11, 12). Autophagy is a fundamental cellular process for cells to maintain homeostasis. With this process, cells periodically clean their interiors by forming double-membraned organelles called autophagosomes to deliver captured cytosolic constituents to the lysosomes for degradation (13, 14). Recent studies showed that autophagosomes are also able to capture intracellular microbes, including bacteria, viruses, and protozoa, for elimination, and this process can be triggered by activation of TLRs by their cognate ligands (15–17). For example, TLR3-mediated autophagy is induced by natural double-stranded (ds)RNA or its synthetic analog polyinosinic-polycytidylic acid [poly(I:C)]. TLR4 activation by lipopolysaccharide (LPS) induces autophagy in macrophages and increases the capture of Mycobacterium tuberculosis into autophagosomes, illustrating the role of TLR activation in autophagy-mediated microbial elimination. Other TLRs, such as TLR2, TLR7, and TLR9, were also reported to mediate induction of autophagy by their cognate ligands in different cell types (18–21).

The formation of autophagosomes involves multiple steps controlled by multiple protein complexes. Beclin 1 is a key regulatory protein in the early steps. The initiation of phagophore formation is regulated by a protein complex comprising Vps15, Vps34, Beclin 1, and other regulatory proteins, such as Bcl-2. When the Bcl-2 family of proteins is associated with Beclin 1 through its BH3 domain, autophagy is inhibited. In contrast, when Bcl-2 is disrupted from this protein complex, autophagy is initiated (22, 23). Further elongation and closure of phagophores are controlled by the recruitment of LC3-II, a phosphatidylethanolamine lapidated form of LC3 protein, to the docking sites provided by agt5/agt12/agt16 protein complexes (24, 25). Although the detailed signaling cascade leading to induction of autophagy after TLR activation requires further investigation, several reports have shown the requirement for MyD88 and TRIF in TLR-mediated autophagy (19–21). Shi et al. (21) further showed the involvement of Beclin 1 in TLR4-mediated autophagy. TLR4 activation dissociates Beclin 1 from the Beclin 1-Bcl-2 complex and recruits Beclin 1 to a protein complex containing MyD88 and TRIF (21). The detailed function and regulation of Beclin 1 in this signaling complex are still unclear. Given that assembly of this TLR signalsome is crucial for initiating TLR-mediated host defense responses, these results have suggested Beclin 1 as a novel regulator for TLR signaling, particularly in linking TLR activation to induction of autophagy.

Originally discovered as a molecular chaperone to prevent protein unfolding, heat shock protein 90 (Hsp90) has been demonstrated to regulate diverse signaling proteins involved in various biological processes. Hsp90 forms a protein complex to maintain the stability of its client proteins. Disruption of this protein complex with specific Hsp90 inhibitors leads to proteolytic degradation of the client proteins, usually through the ubiquitin-proteasome pathway (26–28). Inhibitors such as geldanamycin (GA), 17-allylamino-17-demethoxygeldanamycin, radicocol, and ansamycin bind tightly to the ATP/ADP pocket of Hsp90 and inhibit its interaction with client proteins and are used to probe the biological functions of Hsp90 (29, 30). In the present study, we investigated the interaction between Hsp90 and Beclin 1, and the role of Beclin 1 in TLR signaling. Our results demonstrated that Hsp90 regulates TLR-mediated autophagy by maintaining the stability of Beclin 1.

MATERIALS AND METHODS

Reagents and antibodies

Lysosome inhibitor E64, KNK437/hsp inhibitor I, and proteasome inhibitor lactacystin were purchased from Calbiochem (San Diego, CA, USA). GA, cycloheximide, polymyxin B, LPS from Salmonella minnesota R595, and poly(I:C) were purchased from Sigma (St. Louis, MO, USA). LC3 rabbit IgG antibody was purchased from Cell Signaling (Danvers, MA, USA). Ubiquitin mAb was purchased from BD Biosciences (Mountain View, CA, USA). Ubiquitin Lys-48-specific mAb was purchased from Millipore (Temecula, CA, USA). Beclin 1 mAb was purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). HRP-conjugated anti-mouse and anti-rabbit secondary antibodies were purchased from Jackson ImmunoResearch (West Grove, PA, USA). FITC- or Rhodamine-conjugated goat anti-mouse IgG were purchased from Biosource (Camarillo, CA, USA). Alexa 594-conjugated transferrin and Alexa 594-conjugated goat anti-mouse IgG were purchased from Molecular Probes (Eugene, OR, USA).

Plasmid constructs and RT-PCR analysis

The GFP-LC3 construct was a kind gift from Dr. Tamotsu Yoshimori (Osaka University, Osaka, Japan; ref. 31). Mammalian expression vectors for ubiquitin and K48R ubiquitin mutant were from Addgene Inc. (Cambridge, MA, USA). Full-length Beclin 1 and Bcl-2 cDNAs were PCR amplified from first-strand cDNA libraries, as described previously (32). cDNAs for Beclin 1 truncations were generated by PCR amplification with the full-length Beclin 1 cDNA as template. These cDNAs were subcloned into PRK5 mammalian expression vectors containing an N-terminal Myc epitope tag. The numbers of each construct for truncated Beclin 1 indicate the N-terminal and C-terminal amino acid residues. For RT-PCR analysis of gene expression, total RNA was isolated from cells using a RNA isolation kit (Qiagen, Valencia, CA, USA). First-strand cDNA synthesis, PCR amplification with gene-specific primers, was performed as described previously (32).

Cell culture and transfection

RAW 264.7 cells and HEK293 cells were cultured in DMEM medium supplemented with 10% FBS. HEK293 cells were transfected by using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA). RAW 264.7 cells were transfected with TransIT-Jurkat (Mirus Bio, Madison, WI, USA), according to the manufacturer's instruction.

Luciferase reporter assays

For luciferase reporter assays, ELAM-1 luciferase reporter plasmid or IFN-β luciferase reporter plasmid (100 ng) and β-galactosidase plasmid (100 ng), were cotransfected into the cells with other vectors as indicated. Sixteen hours later, the cells were treated as indicated or lysed, and luciferase activity was determined using reagents from Promega Corp. (Madison, WI, USA). Relative luciferase activities were calculated as fold of induction compared with an unstimulated vector control. The data are presented as means ± sd (n=3).

Immunoblotting

Cells with or without treatment were collected and lysed in lysis buffer containing 150 mM NaCl, 50 mM Tris (pH 8.0), 0.5 mM EDTA, and 1% Nonidet P-40, plus complete protease inhibitor cocktail (Roche Applied Science, Indianapolis, IN, USA). Following brief vortexing and rotation, cell lysates were separated by SDS-PAGE and transferred to PVDF membranes. These membranes were blocked with 5% fat-free milk in PBS for 30 min and incubated with indicated antibody in PBS with 0.5% fat-free milk for 2 h. The membranes were then washed in PBS and incubated for 1 h with HRP-conjugated secondary antibody. After subsequent washes, the immunoreactive bands were detected with ECL plus immunoblotting detection reagents (Amersham Pharmacia Biotech, Piscataway, NJ, USA).

Immunoprecipitation

For immunoprecipitation, cell lysates were incubated with the indicated antibody plus protein G Sepharose at 4°C overnight to form immunocomplexes. After extensive washing with lysis buffer, the immunocomplexes were analyzed by immunoblotting as described.

Fluorescence confocal microscopy

Cells grown on coverslips were fixed in BD Cytofix/Cytoperm solution (BD Biosciences) at room temperature for 15 min. These coverslips were incubated with the primary antibody, followed by fluorochrome-conjugated secondary antibody, before mounting. For fluorescence analysis, cell samples were visualized with single-line excitation at 555 nm for Alexa 594 on an Olympus Fluoview confocal microscope with appropriate emission filters (Olympus, Tokyo, Japan).

Autophagy analyses

Autophagy was analyzed by immunoblotting or fluorescence microscopy, as described previously (33). In the immunoblotting analysis, cells were treated as indicated, and cell lysates were immunoblotted with anti-LC3 antibody to monitor the LC3-II generated during the formation of autophagosomes. In the fluorescence confocal microscopy analysis, cells were transfected with a GFP-LC3 construct and treated as indicated. These cells were imaged by fluorescence confocal microscopy, with single-line excitation at 488 nm for GFP, for the formation of punctuate dots in autophagic cells. A minimum of 50 cells were analyzed for each treatment, and each experiment was performed at least 3 times.

Salmonella infection

The Salmonella typhimurium strain used was SL1344, and bacterial concentration was determined by serial dilutions of bacteria on Lenox broth (LB) agar plates. Salmonella infections were performed as described previously (34). Briefly, RAW 264.7 cells were untreated or treated with GA or polymyxin B as indicated and allowed to absorb bacteria at a multiplicity of infection of 2:1. To analyze the growth of S. typhimurium in these cells, the infected cells were exposed to 100 μg/ml of gentamicin in the culture medium for 1 h and washed with PBS to remove the extracellular bacteria. The cells were then cultured in fresh medium with 10 μg/ml of gentamicin and agents at the concentrations and durations indicated. The cells were then washed 3 times with PBS and lysed with 0.1% Triton X-100 to harvest the surviving intracellular bacteria. The recovered bacteria were serially diluted with PBS and grown on LB plates. The number of bacterial colonies for each sample were counted and expressed as colony-forming units (CFU). To determine the association of S. typhimurium to autophagosomes, S. typhimurium were stained with Salmonella O antiserum group B; autophagosomes were stained with monodansylcadaverine (MDC) and analyzed by fluorescence confocal microscopy.

Statistical analysis

The 2-tailed Student's t test was used for all statistical analyses in this study. A value of P < 0.05 is defined as statistically significant.

RESULTS

Ubiquitination of Beclin 1 in TLR-stimulated and unstimulated cells

To investigate the regulation of Beclin 1 in TLR activation, monocytic RAW264.7 cells were treated with poly(I:C) or LPS to activate TLR3 and TLR4, respectively, and cell lysates were analyzed by immunoblotting for changes in Beclin 1 expression. In immunoblots, a laddering pattern was observed for Beclin 1 in control cells, and this pattern became more intense in TLR-stimulated cells, suggesting that Beclin 1 was post-transcriptionally modified in these cells. To investigate whether the laddering pattern resulted from ubiquitination of Beclin 1, Beclin 1 was immunoprecipitated from cell lysates and immunoblotted with antibodies specific for ubiquitin or Lys48-linked ubiquitin. Beclin 1 was basally ubiquitinated in resting cells, and TLR3 and TLR4 activation further enhanced Lys48-linked ubiquitination of Beclin 1 (Fig. 1A). Generally, Lys48-ubiquitination targets protein for proteasomal degradation (35, 36). To determine whether TLR activation promotes proteolytic degradation of Beclin 1, RAW264.7 cells were treated with poly(I:C) or LPS for different periods of time (Fig. 1B), or treated with poly(I:C) or LPS in the presence or absence of a protein synthesis inhibitor, cycloheximide, for 12 h (Fig. 1C), and analyzed by immunoblotting for Beclin 1 protein levels. The results showed that while both TLR3 and TLR4 activation enhanced ubiquitination of Beclin 1 (Fig. 1A), TLR activation did not promote proteolytic degradation of Beclin 1 (Fig. 1B, C).

Figure 1.

Expression and ubiquitination of Beclin 1 in TLR-stimulated and unstimulated cells. A) RAW264.7 cells were treated with poly(I:C) (10 μg/ml) or LPS (100 ng/ml) for 2 h and then lysed. The lysates were subjected to SDS-PAGE and analyzed by immunoblotting with antibody against Beclin 1 or actin. To determine the ubiquitination of Beclin 1, cell lysates were subjected to immunoprecipitation with an anti-Beclin 1 antibody and then analyzed by immunoblotting with antibodies specific to Ub or Lys48-linked Ub. B) RAW264.7 cells were treated with poly(I:C) or LPS for different periods of time, and then lysed. C) RAW264.7 cells were treated with poly(I:C) or LPS in the presence or absence of cycloheximide (CHX, 10 μg/ml) for 12 h and then lysed. Lysates were subjected to SDS-PAGE and analyzed by immunoblotting with anti-Beclin 1 antibody.

Hsp90 protects Beclin 1 from ubiquitination-associated proteolytic degradation

Hsp90 binding protects ubiquitinated proteins from proteolytic degradation (37–40). The stability of Beclin 1 in TLR ligand-stimulated and unstimulated cells led us to speculate that Beclin 1 may be a client protein of Hsp90. Using GA, an Hsp90-selective inhibitor, we showed that GA treatment promoted proteolytic degradation of Beclin1 in RAW264.7 cells in a concentration- and time-dependent manner. The EC₅₀ for GA was 100 nM, and the t₅₀ for degradation of Beclin 1 was 2 h, with nearly complete degradation by 16 h, when the cells were treated with 1 μM GA. In contrast, proteolytic degradation of Beclin 1 was not seen when the cells were treated with KNK437/Hsp inhibitor I, an inhibitor primarily targeting Hsp70 (Fig. 2A). Similarly, in HEK293 cells, GA promoted degradation of Beclin 1 (Fig. 2B). Unlike Beclin 1, the protein levels of its binding protein Bcl-2 were unaffected by GA treatment in both RAW264.7 cells and HEK293 cells (Fig. 2A, B).

Figure 2.

Inhibition of Hsp90 causes ubiquitination-associated and proteasome-dependent degradation of Beclin 1. A, B) RAW264.7 cells (A) or HEK293 cells (B) were treated with GA or KNK437 at the indicated concentrations for 16 h (left panels), or 1 μM GA for different periods of time (right panels) and then lysed. C) HEK293 cells were transfected with expression vector for Ub-K48R or Ub, and the cells were treated with 1 μM GA for 16 h. D) HEK293 cells were treated with GA (0.5 μM), lactacystin (10 μM), or E64 (10 μM) for 6 h and then lysed. Lysates were subjected to SDS-PAGE and analyzed by immunoblotting with antibodies against Beclin 1, Bcl-2, or actin. Data in graphs are relative intensity of Bcl-2 and Beclin 1 normalized to β-actin; values are presented as means ± sd (n=3). Blots shown are representative of 3 independent experiments.

We next investigated whether the GA-induced proteolytic degradation of Beclin 1 was associated with Lys48-linked ubiquitination. HEK293 cells were transfected with an expression vector for ubiquitin or K48R ubiquitin mutant that acts as a dominant-negative by premature termination of ubiquitin chains and inhibition of proteasomal degradation (41) and then treated with GA. Degradation of Beclin 1 was observed in the ubiquitin-expressing cells but not in the K48R ubiquitin mutant-expressing cells (Fig. 2C). To further investigate whether Beclin 1 was targeted to the proteasome for degradation, HEK293 cells were pretreated with an irreversible proteasome inhibitor, lactacystin, or a lysosome inhibitor, E64, followed by GA treatment. Degradation of Beclin 1 was blocked by lactacystin treatment but not E64 treatment (Fig. 2D). These results suggested that Beclin 1 is a client protein of Hsp90, and after GA-induced dissociation from Hsp90, Beclin 1 was targeted for degradation through an ubiquitination-associated and proteasome-dependent pathway.

Hsp90 interacts with Beclin 1 through an evolutionarily conserved domain

Beclin 1, a 448-aa residue protein, contains a BH3 domain, a central coiled-coil domain, and an evolutionarily conserved domain (22). To confirm a physical interaction between Hsp90 and Beclin 1, as well as to map the Hsp90 binding site on Beclin 1, we generated a series of Myc-tagged truncation mutants for Beclin 1. The full-length and truncated Beclin 1 proteins were expressed in HEK293 cells and analyzed by coimmunoprecipitation and immunoblotting to determine their binding with Hsp90. The results showed coimmunoprecipitation between Beclin 1 and Hsp90. Truncated mutants containing the evolutionarily conserved domain, such as Beclin 1 (1–337), Beclin 1 (150–448), and Beclin 1 (244–448), retained their ability to bind Hsp90, whereas those lacking this domain, such as Beclin 1 (1–150), Beclin 1 (88–224), Beclin 1 (1–224), and Beclin 1 (337–448), were unable to bind to Hsp90 (Fig. 3). These results suggested that the binding site for Hsp90 was located within the evolutionarily conserved domain of Beclin 1, and further confirmed that Hsp90 forms a protein complex with Beclin 1 in cells.

Figure 3.

Hsp90 binds with Beclin 1 through an evolutionary conserved domain. A) Schematic diagram of Beclin 1 and truncation mutants used in this study. BH3, BH3 domain; CCD, coil-coil domain; ECD, evolutionarily conserved domain. B) Myc-tagged full-length Beclin 1 or truncates were transiently transfected into HEK293 cells. Cell lysates were analyzed by immunoblotting with anti-myc antibody for the expression of Beclin 1 protein or by immunoprecipitation with anti-Hsp90 antibody following by immunoblotting with the anti-myc antibody for the association between Hsp90 and Beclin 1 proteins.

Bcl-2 does not compete with Hsp90 for binding to Beclin 1

Bcl-2 associates with Beclin 1 through the BH3 domain to inhibit induction of autophagy (22, 23). To investigate whether Bcl-2 competed with Hsp90 for binding to Beclin 1, we first cotransfected HEK293 cells with expression vectors for Bcl-2 and Beclin 1. Beclin 1 and Hsp90 were detected in the protein complex immunoprecipitated from cell lysates with an anti-Bcl-2 antibody, suggesting an interaction between Bcl-2 and the Beclin 1 and Hsp90 protein complex (Fig. 4A). The cells were then cotransfected with an expression vector for Beclin 1 and increasing amounts of an expression vector for Bcl-2. Cell lysates were analyzed for interaction between Hsp90, Beclin 1, and Bcl-2. Increased expression of Bcl-2 did not affect the binding between Hsp90 and Beclin 1 (Fig. 4B). We further investigated the effect of GA on disruption of the Hsp90/Beclin 1 protein complex in the presence of increasing amounts of Bcl-2. The Bcl-2-overexpressing cells were treated with GA and analyzed by immunoblotting for the protein levels of Beclin 1. Increased expression of Bcl-2 had no effect on the stability of Beclin 1 in the GA-treated cells (Fig. 4C). These results suggested that the binding of Bcl-2 did not interfere with the interaction between Hsp90 and Beclin 1 and were consistent with the findings that Bcl-2 and Hsp90 use distinct domains for binding to Beclin 1.

Figure 4.

Bcl-2 does not interfere with the interaction between Hsp90 and Beclin 1. HEK293 cells were transiently transfected with expression vectors for Bcl-2 and Beclin 1 as indicated. A) Cell lysates were immunoprecipitated with control IgG or antibody against Bcl-2, followed by immunoblotting for the presence of Bcl-2, Beclin 1, and Hsp90 in the immunocomplex. B) Cell lysates were analyzed by immunoprecipitation with antibody against Hsp90, following by immunoblotting for the presence of Beclin 1 and Bcl-2 in the immunocomplex and by immunoblotting for the expression levels of each protein as indicated. C) Cells were treated with GA (1 μM) for 8 h and then lysed. Cell lysates were analyzed by immunoblotting with antibodies for the expression levels of each protein as indicated.

Hsp90 interacts specifically with Beclin 1 among the proteins in the TLR signalsome

Assembly of a TLR signalsome comprising MyD88 family proteins is essential for initiation of TLR-mediated inflammatory and antiviral responses (7, 8). Recently, this signaling complex has been shown to also include Beclin 1 for TLR-mediated autophagy (21). To further investigate whether Hsp90 regulated any other proteins in the TLR signalsome besides Beclin 1, HEK293 cells were transfected with expression vectors for Myc-tagged Beclin 1, Bcl-2, MyD88, TRIF, TIRAP, or TRAM. The cells were treated with GA for different periods of time, and then analyzed by immunoblotting for their protein levels. Consistent with the results observed for endogenous Beclin 1 and Bcl-2 in RAW264.7 cells and HEK293 cells (Fig. 2), proteolytic degradation was seen only for overexpressed Beclin 1 but not for Bcl-2 (Fig. 5A, B). Further, no proteolytic degradation of MyD88, TRIF, TIRAP, and TRAM was observed in the GA-treated cells (Fig. 5C–F), indicating that Hsp90 interacts specifically with Beclin 1 among the proteins in the TLR signalsome.

Figure 5.

Hsp90 specifically interacts with Beclin 1 among the proteins in the TLR signalsome. HEK293 cells were transiently transfected with expression vectors for myc-tagged Beclin 1 (A), Bcl-2 (B), MyD88 (C), TRIF (D), TIRAP (E), and TRAM (F). Cells were treated with GA (1 μM) for different periods of time as indicated. Cells were then lysed, and lysates were analyzed by immunoblotting with antibodies against myc.

Hsp90/Beclin 1 protein complex is not involved in TLR-mediated NF-κB activation and IFN-β induction

To assess the regulatory effect of Hsp90 on TLR signaling through its control of Beclin 1, we went on to study the role of Beclin 1 in the TLR signalsome. NF-κB and IFN-β activation are the major signaling pathways leading to TLR-activated inflammatory and antiviral responses. We investigated whether Beclin 1 acts as a positive regulator or a negative regulator to control TLR-mediated NF-κB activation or induction of IFN-β. To investigate whether Beclin 1 acts as a positive regulator, HEK293 cells were cotransfected with a luciferase reporter gene driven by a NF-κB-controlled ELAM-1 promoter and an expression vector for Beclin 1, or expression vectors for signaling molecules in the MyD88 protein family. These cells were lysed to measure luciferase activities for NF-κB activation. Unlike the signaling molecules in the MyD88 protein family, Beclin 1 was not able to activate NF-κB (Fig. 6A, left panel). In similar experiments with cells transfected with an IFN-β promoter controlled luciferase reporter gene, Beclin 1 was unable to induce IFN-β either (Fig. 6A, right panel). Consistent with these results, RT-PCR analysis with gene-specific primers showed that overexpression of Beclin 1 in cells did not induce expression of NF-κB-regulated genes, including ELAM-1, TNF-α, and IL-6. Nor did it induce IFN-β (Fig. 6B). To investigate whether Beclin 1 is a negative regulator, a dominant-negative ΔMyD88, containing amino acid residues 152 to 296 and a dominant-negative ΔTRIF, containing amino acid residues 387 to 566, were used as controls (42, 43). HEK293 cells were cotransfected with a ELAM-1 luciferase reporter gene, an expression vector encoding a constitutively active form of TLR4 (mCD4/TLR4) generated by fusion of the mCD4 extracellular domain to the intracellular domain of TLR4, and increasing amounts of vector encoding Beclin 1. The mCD4/TLR4-induced NF-κB activities were measured. While ΔMyD88 blocked NF-κB activation, Beclin1 did not (Fig. 6C, left panel). In a similar experiment, Beclin 1 did not block IFN-β induction triggered by TLR3 activation, while ΔTRIF did (Fig. 6C, right panel). These results suggested that Beclin 1 is not involved in TLR-mediated NF-κB and IFN-β activation as either a positive or a negative regulator. Thus, Hsp90 is unlikely to regulate these two effector mechanisms of TLR activation through its interaction with Beclin 1.

Figure 6.

Beclin 1 does not participate in TLR signaling leading to activation of NF-κB and induction of IFN-β. A) HEK293 cells were transfected with expression vectors for the indicated signaling proteins, together with an ELAM-1 promoter-controlled luciferase-reporter gene (left panel) or an IFN-β promoter-controlled luciferase-reporter gene (right panel). Blots show expression of Beclin 1. B) HEK293 cells were transfected with expression vectors for the indicated signaling proteins. Expression of the indicated genes was analyzed by RT-PCR. Data presented are relative intensity of PCR product for indicated gene normalized to GAPDH; values are presented as means ± sd (n=3). Blots show a representative RT-PCR. C) Left panel: HEK293 cells were transfected with increasing amounts of expression vector for Beclin 1 or dominant-negative ΔMyD88, as indicated, together with an expression vector for mCD4/TLR4 and an ELAM-1 promoter controlled luciferase-reporter gene for 16 h. Right panel: HEK293 cells were transfected with increasing amounts of expression vector for Beclin 1 or dominant-negative ΔTRIF, as indicated, together with an expression vector for TLR3 and an IFN-β promoter-controlled luciferase-reporter gene for 12 h and then treated with poly(I:C) (10 μg/ml) for 6 h. Cells were lysed, and relative luciferase activities were determined. Blots show expression of Beclin 1. Data represent means ± sd (n=3). *P < 0.05 vs. control cells.

Hsp90/Beclin 1 protein complex regulates TLR-mediated autophagy

We then focused on the function of Hsp90 in regulating TLR-mediated autophagy, using GA as a probe to disrupt the Hsp90/Beclin 1 protein complex. The induction of autophagy was assessed by monitoring LC3. In resting cells, LC3 exists as an 18-kDa cytosolic LC3-I form. Following cell activation to induce autophagy, some of the LC3 is lipidated to produce LC3-II, which migrates in SDS-PAGE as a 16-kDa protein (33). RAW264.7 cells were pretreated with or without GA, and then stimulated with poly(I:C) or LPS to activate TLR3 and TLR4, respectively, for induction of autophagy. The cells were lysed, and the lysates analyzed by immunoblotting for LC3-II. TLR3 and TLR4 activation significantly increased the level of LC3-II in cells. Treatment of the cells with GA blocked the formation of LC3-II (Fig. 7A). The inhibition of TLR-mediated autophagy by GA was further quantitated using RAW 264.7 cells transfected with expression vector for green fluorescence protein fused LC3 (GFP-LC3). The cells were treated with GA, followed by poly(I:C) or LPS stimulation. When autophagosomes form, the GFP-LC3-II is recruited from the cytosol to autophagosome membranes, which can be visualized as punctate dots by confocal microscopy. GFP-LC3-positive cells that have punctate dots can be counted to quantitate the induction of autophagy. Consistent with the immunoblot results (Fig. 7A), activation of TLR3- and TLR4-induced autophagy in 40–50% of the cells, and the autophagy was markedly decreased by GA treatment (Fig. 7B). To ensure that this effect resulted from proteolytic degradation of Beclin 1, we first determined the titer of Beclin 1 expression vector sufficient to compete with the destabilizing effect of GA on Beclin 1 in cells (Fig. 7C, top panel). Beclin 1 was then overexpressed in cells in excess to rescue the effect of GA treatment. As shown in Fig. 7C (bottom panel), overexpression of Beclin 1 enhanced TLR-mediated autophagy in the GA-treated cells. These results suggested that Hsp90 could regulate TLR-mediated autophagy through its interaction with Beclin 1.

Figure 7.

GA treatment attenuates TLR-mediated autophagy, and expression of Beclin 1 reverses this effect of GA. A) RAW264.7 cells were treated with poly(I:C) (10 μg/ml) or LPS (100 ng/ml), with or without GA (1 μM) for 16 h, and then lysed. Lysates were analyzed by immunoblotting with antibodies against LC3 for the conversion of LC3-I to LC3-II. B) Top panels: cells were transiently transfected with GFP-LC3 overnight, followed by treatment with poly(I:C) (10 μg/ml) or LPS (100 ng/ml) for 16 h, and then fixed. Digital images were captured by confocal microscopy. Scale bar = 10 μm. Bottom panel: autophagy induced by different treatments, as indicated, was quantified as the percentage of GFP-positive cells with punctate structures. C) Top panel: cells were transfected with different amounts of expression vector for Beclin 1, as indicated, and treated with GA. Cell lysates were analyzed by immunoblotting for proteolytic degradation of Beclin 1. Bottom panel: cells were transiently transfected with GFP-LC3 together with control vector or with expression vector for Beclin 1 to overexpress Beclin 1. Cells were treated and quantitated for autophagy as in B. Data represent means ± sd of 3 experiments. *P < 0.05.

Hsp90/Beclin 1 complex regulates TLR-mediated antimicrobial responses via control of autophagy

Autophagy has been shown to restrict the growth of S. typhimurium in cells (44). We further investigated the function of Hsp90 in controlling TLR-mediated uptake of bacteria into autophagosomes and restriction of their growth in cells, using these gram-negative bacteria. Association of S. typhimurium with autophagosomes in RAW 264.7 cells was evident by the localization of Salmonella O antiserum group B-stained bacteria in the MDC-stained autophagosomes (Fig. 8A). The roles of Hsp90 and TLR4 in controlling S. typhimurium infection-induced autophagy was investigated by pretreatment of RAW264.7 cells with GA or polymyxin B, an inhibitor for neutralization of LPS. The cells were then treated with S. typhimurium, and induction of autophagy was assessed by LC3-II immunoblots. S. typhimurium infection-induced autophagy was blocked by GA and polymyxin B treatment (Fig. 8B), suggesting a role for the Hsp90/Beclin 1 protein complex in this cellular process. The inhibitory effect of GA on autophagy was further evidenced by reduced colocalization of S. typhimurium with autophagosomes in cells (Fig. 8C), and increased growth of bacteria isolated from the cells (Fig. 8D). These results suggested a role of the Hsp90/Beclin 1 complex in TLR-mediated antimicrobial responses through control of autophagy.

Figure 8.

GA treatment attenuates S. typhimurium-induced autophagy and increases bacterial survival in cells. A) RAW264.7 cells were infected with S. typhimurium and then stained with MDC and with rabbit polyclonal Ab against Salmonella O antiserum group B, followed by Alexa-594-conjugated anti-rabbit antibody. Digital images were captured by confocal microscopy. Scale bar = 10 μm. B) Cells were pretreated with GA (1 μM) or polymyxin B (PMB) (20 μg/ml) and then infected with S. typhimurium for 12 h. Cells were lysed, and cell lysates were analyzed by immunoblotting with antibodies against LC3 or actin. C) Cells were untreated or treated with GA and then infected with S. typhimurium. Colocalization of MDC-positive autophagosomes and Salmonella O antiserum group B-stained bacteria was quantitated. D) Colony-forming units of S. typhimurium in GA-untreated or -treated cells were quantitated. Data represent means ± sd (n=3).

DISCUSSION

The results presented in this study suggest that Hsp90 is essential for TLR-mediated autophagy by interacting with and maintaining the stability of Beclin 1. Hsp90 is a member of the phylogenetically conserved Hsp family, which includes Hsp110, Hsp70, Hsp60, and Hsp40 in mammalian cells (45). Most Hsps are molecular chaperones that help restore protein native folding and confer protein stability. Similarly, Hsp90 plays an important role in maintaining the stability and function of cellular proteins. However compared to other Hsps, Hsp90 appears to be more selective in interaction with its client proteins, as most of Hsp90's client proteins are involved in signal transduction. For example, Hsp90 interacts with Her/neu, tyrosine kinase Erb2, c-Src, v-Src, BCR-Abl, and p53 in oncogenic signaling pathways, and binds to MEKK3, NIK, RIP1, TAK1, and TBK1 in inflammatory signaling pathways (27, 37, 46–49).

Hsp90 contains 3 domains: a highly conserved N-terminal ATP-binding domain, a middle domain, and a carboxyl terminal. In the canonical model for Hsp90 function, ATP binding is required for Hsp90 to associate with its client proteins to form a mature complex. It is in this state that a client protein is able to be activated, and, in turn, activates a cellular signaling pathway. Inhibition of ATP-binding through inhibitors such as GA prevents client protein maturation and results in ubiquitination and proteolytic degradation of the client protein (26–30). In the canonical model, a client protein is ubiquinated following dissociation from Hsp90. However, increasing evidence has shown that Hsp90 binds to ubiquitinated proteins and maintains their stability for signal transduction. For example, Hsp90 associates with ubiquitinated Her-2/neu, proto-Dbl, RIP1, and LRRK2 to protect them from proteolytic degradation (37–40). Although it is not clear whether ubiquitination of these proteins is required for them to mediate signal transduction, these observations suggest a noncanonical model for the function of Hsp90. In this study, we found basal ubiquitination of Beclin 1 in resting cells and determined that TLR activation enhanced the K48-linked ubiquitination. In general, K48-linked polyubiquitin chains target proteins for proteasomal degradation (35, 36). However, we found that the protein levels of Beclin 1 were unchanged in TLR ligand-stimulated cells. The finding that Beclin 1 is a client protein for Hsp90 provides an explanation for this observation and suggests that Hsp90 regulates Beclin 1 through the noncanonical model.

Beclin 1 is the mammalian homologue of yeast Agt6. In yeast, Agt6 forms a complex with Vps15, a class III phosphatidylinositol 3-kinase (PI3K), and Vps34, a myristoylated and membrane-anchored kinase. In this protein complex, Atg6 acts as an activator for Vps34 to phosphorylate phosphatidylinositol into phosphatidylinositol 3-phosphate (PI3P) for activation of nucleation and retrieval of additional membranes for phagophore formation. In mammalian cells, the interaction between Beclin 1, Vps34, and Vps 15 is conserved, together with additional regulators such as UVRAG, Ambra 1, and Bcl-2. UVRAG and Ambra 1 are indispensable for optimal activation of Vps34 and induction of autophagy. Bcl-2, on the other hand, constitutively binds Beclin 1 through its BH3 domain and inhibits autophagy (22, 23). Induction of autophagy by diverse means, such as nutrient starvation or TLR ligand stimulation, is accompanied by dissociation of Beclin 1 from Bcl-2 (21–23). The results presented in this study suggest that Hsp90 is a novel regulator for Beclin 1 in this protein complex. The unchanged protein levels of Beclin 1 in TLR ligand-stimulated cells suggest that Hsp90 does not dissociate from Beclin 1 following TLR stimulation. Hsp90 interacts with Beclin 1 through an evolutionary conserved domain. The binding of Bcl-2 to Beclin 1 does not interfere with the interaction between Hsp90 and Beclin 1. These clearly show that Bcl-2 and Hsp90 use distinct molecular mechanisms and binding sites to interact with Beclin 1.

A previous study by Qu et al. (50) with heterozygous Beclin 1 gene-deleted mice showed that a 50% reduction in Beclin 1 protein expression in cells causes a severe defect in autophagy, suggesting that the steady level of Beclin 1 protein in cells is crucial for induction of autophagy. In this study, we found that Hsp90 controls the steady-state level of Beclin 1. Disruption of the Hsp90/Beclin 1 protein complex by GA treatment led to the degradation of Beclin 1 through a proteasome-dependent pathway and blocked TLR-mediated autophagy, as well as S. typhimurium-induced autophagy. These results confirmed that a threshold level of Beclin 1 is required for optimal induction of autophagy, and Hsp90 clearly controls the availability of Beclin 1 to the autophagy machinery by maintaining the homeostasis of Beclin 1.

TLRs are the major sensors of microbial pathogens in innate immune cells. TLR activation triggers effector mechanisms, including an inflammatory response and an antiviral response to combat microbial infections. NF-κB activation and IFN induction are the two major signaling pathways to elicit these responses, and both signaling pathways are initiated by assembly of a TLR signalsome, which contains adaptor proteins in the MyD88 family (7–10). Autophagy is a newly recognized effector mechanism for TLR activation to directly eliminate microbes in cells (11–12). The signaling pathway leading from TLR activation to induction of autophagy has not been well established, but several studies have shown that the TLR signalsome is required for induction of autophagy, as well as for the induction of inflammatory and antiviral responses. Shi et al. (21) reported that TLR activation recruits Beclin 1 to the TLR signalsome and dissociates Beclin 1 from the inhibition by Bcl-2. This study suggested that Beclin 1 is a signaling molecule to link TLR activation to induction of autophagy and also suggested that Beclin 1 may be involved in other TLR-mediated cellular responses, since all TLR signaling is initiated from the TLR signalsome. Our results demonstrate that Beclin 1 is not a signaling molecule in TLR-mediated NF-κB activation and IFN-β induction, since, unlike other signaling molecules for TLR activation, overexpression of Beclin 1 in cells did not activate NF-κB or induce production of IFN-β, nor did it block TLR4-induced NF-κB or TLR3-activated IFN-β production. These results suggest that several distinct signaling pathways lead from the TLR signalsome to activation of different effector mechanisms for microbial elimination. In cells other than plasmacytoid dendritic cells, MyD88 mediates NF-κB activation to provoke inflammatory responses, TRIF mediates IFN-β production to elicit antiviral responses, and Beclin 1 mediates induction of autophagy, as illustrated in Fig. 9. Hsp90 regulates TLR-mediated autophagy through regulation of Beclin 1. However, since Hsp90 has been shown to regulate several other proteins in the TLR signaling cascade such as RIP1, TAK1 and TBK1 (37, 47, 49), this does not exclude the possibility that Hsp90 may also control other facets of TLR signaling. RIP1 and TAK1 are essential for NF-κB activation and TBK1 is required for induction of IFN-β in TLR signaling (Fig. 9), and an Hsp90 inhibitor has been shown to inhibit TLR4-induced NF-κB activation and cytokine production (51, 52).

Figure 9.

Schematic of the roles of Hsp90 in controlling TLR-mediated host-defense responses. Ligation of TLR activates inflammatory and antiviral responses, as well as induction of autophagy for host defense to microbial infections. The signaling leading to these responses diverges from the TLR signalsome. MyD88 mediates NF-κB activation, TRIF mediates IFN-β production, and Beclin 1 is involved in signaling for induction of autophagy. Hsp90 controls TLR-mediated autophagy by maintaining the stability of Beclin 1 and also may control the inflammatory responses and antiviral responses through its interactions with TAK1, RIP1, and TBK1, as indicated.

Taken together, the results presented in this study show that Beclin 1 is modified by K48-linked ubiquitin chain in TLR-stimulated and unstimulated cells, and Hsp90 forms a complex with Beclin 1 to maintain the stability of this protein for autophagy. Nevertheless, the detailed mechanism by which the ubiquitination of Beclin 1 is controlled is still unclear. During preparation of this report, Shi et al. (53) reported that K63-linked ubiquitination of Beclin 1 is regulated by TRAF6 and A20. TRAF6 promoted the ubiquitination of Beclin to induce TLR4-mediated autophagy, and A20 deubiquitinated the K63-linked ubiquitin chains (53). A20 is an ubiquitin-editing enzyme containing an N-terminal OUT (ovarian tumor) domain for deubiquitination of K63-linked ubiquitin chains, and a C-terminal zinc finger domain to catalyze K48-linked ubiquitination. This protein has been shown to simultaneously remove K63-linked polyubiquitin chains and add K48-linked polyubiquitin chains to RIP1 following IL-1 and TLR activation (54). Whether Beclin 1 undergoes sequential ubiquitin editing by A20 that results in K48-linked ubiquitination, as shown for RIP1, or whether it is modified by other E3 ubiqitin-protein ligases following deubiquitination by A20 remains to be determined. In addition, the role the K48-linked ubiquitin chain plays in the function of Beclin 1 also awaits investigation. Analysis of LC3-binding proteins has identified adapter proteins, such as p62 (also called A170/SQSTM1), neighbor of BRCA1 gene 1 (nbr1), and histone deacetlyase 6 (HDAC6), that bind ubiquitinated substrates in the autophagosome (55). Of these, the most established adapter, p62, has been shown to bind substrates tagged with monoubiquitin or K63- or K48-linked ubiquitin chains through its ubiquitin-associated (UBA) domain (56). This leads to speculation that, similar to the K63-linked polyubiquitin chain-modified Beclin 1, the K48-linked ubiquitin chain-modified Beclin 1 also represents an active state of Beclin 1 to facilitate induction of autophagy.