THE LEVELS OF RETINOIC ACID INDUCIBLE GENE-I ARE REGULATED BY HEAT SHOCK PROTEIN 90-α

Abstract

Retinoic acid-inducible gene-I (RIG-I) is an intracellular pattern recognition receptor that plays important roles during innate immune responses to viral double-stranded RNAs. The mechanisms and signaling molecules that participate in the downstream events that follow activation of RIG-I are incompletely characterized. In addition, the factors that define intracellular availability of RIG-I and determine the steady-state levels of this protein are only partially understood but are likely to play a major role during innate immune responses. It was recently reported that the antiviral activity of RIG-I is negatively regulated by specific E3 ubiquitin ligases, suggesting participation of the proteasome in the regulation of RIG-I levels. In this study, we utilized immunoprecipitation combined with mass spectrometry to identify RIG-I-interacting proteins and found that RIG-I forms part of a protein complex that includes heat shock protein 90-α (HSP90-α), a molecular chaperone. Biochemical studies using purified systems demonstrated that the association between RIG-I and HSP90-α is direct but does not involve participation of the CARD domain. Inhibition of HSP90 activity leads to the dissociation of the RIG-I/HSP90 complex, followed by ubiquitination and proteasomal degradation of RIG-I. In contrast, the levels of RIG-I mRNA are unaffected. Our studies also show that the ability of RIG-I to respond to stimulation with poly (I:C) is abolished when its interaction with HSP90 is inhibited. These novel findings point to HSP90-α as a chaperone that modulates the activity of RIG-I and shields it from proteasomal degradation and they identify a new mechanism whose dysregulation may seriously compromise innate anti-viral responses in mammals.

Introduction

Viral infection leads to the initiation of complex innate immune responses that result from recognition of the viral nucleic acid by cellular receptors including Toll-like receptors (TLRs) (1), followed by host cell secretion of antiviral factors such as cytokines (2, 3). Additional mechanisms are responsible for the activation of the interferon (IFN) response during viral infections (4). We recently reported that a cytoplasmic RNA helicase, retinoic acid-inducible gene I (RIG-I), is an essential regulator of double-stranded RNA (dsRNA)-induced signaling (5). RIG-I is also known as Ddx58 owing to the fact that this protein belongs to the DExH box-containing helicase family; it harbors two caspase recruitment domains (CARD) at the amino terminal end and an RNA helicase motif at the carboxyl terminus (5). The CARD domains are responsible for activating subsequent downstream signaling events through interactions with IPS-1/MAVS/VISA/Cardif, resulting in the induction of IFNs and the activation of antiviral responses (6–9). In previous work we reported that, aside from its recognized role as a viral sensor, RIG-I may have additional functions. We found that RIG-I is a transcriptional activator of the cyclooxygenase-2 gene and that its expression is enhanced following stimulation of endothelial cells with inflammatory agents (10).

RIG-I is the subject of active investigations aimed at dissecting its precise function and mechanism of action in viral immunity. Factors likely to play a key role in these responses include those that affect expression levels, location, stability, and post-translational modifications, all of which can potentially modulate the ability of RIG-I to affect cellular functions. Zhao and co-workers recently reported that RIG-I becomes conjugated to IFN-regulated gene 15/ubiquitin cross-reacting protein (ISG15/UCRP), a 15-kDa ubiquitin-like protein expressed following cellular stimulation with IFN (11). ISG15 becomes conjugated to a wide array of cellular proteins. Several of the targets, including RIG-I, are IFN-α/β-induced antiviral proteins, but most are constitutively expressed proteins that function in diverse cellular pathways, including RNA splicing, chromatin remodeling/polymerase II transcription, cytoskeletal organization and regulation, stress responses, and translation (11). The precise functional consequences of ISG15 modification remain to be established but, unlike ubiquitin, ISG15 does not appear to target proteins for proteasomal degradation (12, 13). In addition to ISG15-mediated derivatization, RIG-I has been shown to undergo two types of ubiquitin-dependent modifications that result in remarkably different effects on functional properties. Gack and co-workers recently showed that RIG-I is robustly ubiquitinated at its amino-terminal CARD domains by interacting with the E3 ubiquitin ligase TRIM, a member of the tripartite motif protein family, and that this process is necessary for RIG-I-mediated IFN-β production and antiviral activity in response to infection (14). This finding indicates that ubiquitination is a RIG-I modification that can modulate its signaling properties. Conversely, Arimoto et al. reported that conjugation of RIG-I with ubiquitin targets RIG-I for degradation by the proteasome, thus effectively acting as a negative regulator of RIG-I (15). These combined observations point to the existence of at least three tightly-regulated protein conjugation mechanisms that control both the stability and signaling functions of RIG-I.

The goal of the present study was to obtain additional insights on the mechanisms that contribute to the regulation of RIG-I function and expression levels. We found that in resting cells RIG-I is a component of a protein complex that also includes the molecular chaperone heat shock protein 90-α (HSP90-α). IN addition, our studies showed that the association between RIG-I and HSP90 is direct, and that inhibiting the interaction between these proteins leads to ubiquitination and proteasomal degradation of RIG-I. Importantly, disrupting the integrity of the RIG-I/HSP90 complex also inhibited key functional responses mediated by RIG-I. We conclude that partnership between RIG-I and HSP90 is likely to affect the function of RIG-I in vivo.

Materials and Methods

Materials

Cycloheximide (CHI), a protease inhibitor cocktail, FLAG-agarose, anti-FLAG antibodies, the expression vector p3XFLAG-CMV7.1, polyinosinic-polycytidylic acid [poly (I:C)], His-tagged human high-mobility-group protein-1 (HMG-1), creatine kinase, and creatine phosphate were purchased from Sigma-Aldrich (St. Louis, MO). We obtained a metal affinity purification system (TALON), the transfer vector pBacPAK8 and pBacPAK6 viral DNA from Clontech Laboratories, Inc. (Mountainview, CA). The β-galactosidase expression vector pSV40-β-gal was obtained from Promega (Madison, WI) and the bacterial expression vector pGEX-5X-1 and Glutathione Sepharose 4B were from Amersham Biosciences (Pittsburgh, PA). Factor X-a was from Qiagen (Valencia, CA) and Lipofectamine™ Reagent, fetal bovine serum (FBS), Sf21 insect cells, DNAse I, Platinum Pfx DNA polymerase and TRIZOL were from Invitrogen™ (Carlsbad, CA). Polyvinylidene fluoride (PVDF) membranes and chemiluminescence detection reagents were from PerkinElmer Life Sciences (Waltham, MA). Protein content was assessed using a Pierce protein assay kit (Pierce, Rockford, IL). Reagents for cDNA synthesis were from Fermentas (Hanover, MD) and iQ™ SYBR® Green Supermix was from BioRad (Hercules, CA). ICN (Costa Mesa, CA) provided an anti-actin mouse monoclonal antibody. Anti-HSP90 and anti-rpS6 antibodies and protein A/G-agarose were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Cell Signaling Technology (Boston, MA) provided anti-JNK1/2 antibodies. We obtained HRP-conjugated secondary antibodies from BioSource (Camarillo, CA); the anti-RIG-I antiserum was previously described (10) and the anti-ubiquitin antibody was from Novagen (Madison, WI). Geldanamycin was from Biomol (Plymouth Meeting, PA) and MG-132 and HSP90 were from Calbiochem (La Jolla, CA). Alexa 488–conjugated anti–rabbit and Texas red–conjugated anti–mouse IgGs were from Molecular Probes (Eugene, OR). We obtained rabbit reticulocyte lysates from Hokudo Co., Ltd. (Sapporo, Japan).

Cellular studies

HeLa and HEK293T cells were maintained in a 5% CO₂ atmosphere at 37°C in Dulbecco’s modified Eagle’s medium supplemented with 10% FBS. HT-29 cells were maintained in a 5% CO₂ atmosphere at 37°C in McCoys’ 5A medium supplemented with 1% FBS. HSP90 inhibition was accomplished by treating monolayers with geldanamycin (0–400 nM) for the indicated time periods. MG-132 (up to 10 µM, 6 h) was utilized to inhibit proteasomal activity. Unless otherwise stated, we washed treated cells twice with PBS and solubilized the monolayers in cell lysis buffer. After one cycle of freezing and thawing, the lysates were cleared by centrifugation at 12,000 g for 2 min at 4°C. We subjected 50 µg of protein extracts to electrophoresis, as described below.

Quantitative RT-PCR

We subjected DNAse I-treated RNA (1 µg) to reverse transcription using a cDNA synthesis kit, following the instructions provided by the manufacturer. A Chromo4™ Real-Time PCR Detection System (Bio-Rad) was used for quantitative assessment of RIG-I and GAPDH, as previously described (16).

Construct generation, expression, and purification of recombinant proteins

We extracted total RNA from HEK293T cells and subjected the samples to reverse transcription, as previously described (16). For amplification, we used Platinum Pfx DNA polymerase and specific primers harboring a Not I site (shown in lower case font) or a Sal I site (shown in lower case font), the sequences of which were: 5’-GAC AAG CTT gcg gcc gcG ATG ACC ACC GAG CAG CGA CGC-3’ [RIG-FL-F (sense) and 5’-TCC TCT AGA gtc gac TCA TTT GGA CAT TTC TGC TGG-3’ (RIG-FL-R (antisense)]. The products were purified, digested with Not I and Sal I and then cloned into a p3XFLAG-CMV7.1 expression vector. Positive clones were subjected to automated DNA sequencing analyses. We next generated various RIG-I deletion mutants by amplification of full-length RIG-I with the following primers pairs: Δ3 (1-640), RIG-FL-F (sense) and 5’-AGA gtc gac CAC AAG TGC TCT GGT TTT CAC-3’ (antisense); RIG-CARD (1-238), RIG-FL-F (sense) and 5’-AGA gtc gac GTA CAA GTT TGT ATC AGA CAC-3’ (antisense); ΔCARD (239-925), 5’-CTT gcg gcc gcG AGC CCA TTT AAA CCA AGA AAT-3’ (sense) and RIG-FL-R (antisense); Δ10 (452-925), 5’-CTT gcg gcc gcG GTT TAT AAG CCC CAG AAG TTT-3’ (sense) and RIG-I-FL-R (antisense). The PCR products were purified by electrophoresis on agarose gels, digested with Not I and Sal I, cloned into the p3XFLAG-CMV7.1 expression vector, and analyzed by automated sequencing.

For studies that required the use of purified, full-length RIG-I we amplified cDNA isolated from HeLa cells using primers RIG-I-Xho I for BacPAK8-F (5’-GC ctc gag GCA GAG GCC GGC ATG ACC ACC GAG-3’) and RIG-I-6xHis-Not I for BacPAK 8-R (5’-AAg cgg ccg cTC AGT GAT GGT GAT GAT GAT GTT TGG ACA TTT CTG CTG GAT CA-3’); the reverse primer included a 6xHis tag for purification purposes. The product was digested with Xho I and Not I, and then cloned into the BacPAK8 expression vector to generate pBacPAK8-6xHis-RIG-I. We co-transfected pBacPAK6 viral DNA and pBacPAK8-6xHis-RIG-I into Sf21 insect cells, cultured the cells for three days at 27°C, collected the virus-containing supernatant, and plated this fraction on a fresh monolayer of Sf21 cells to generate individual plaques that were subsequently amplified in Sf21 cells. We harvested the cellular proteins using equilibration/wash buffer [50 mM sodium phosphate, 300 mM NaCl (pH 7.0)] containing protease inhibitors, and purified RIG-I by metal affinity chromatography on a TALON resin, according to the manufacturer’s instructions. We eluted 6xHis-tagged RIG-I with 20 mM Tris-HCl (pH 8.0) containing 100 mM NaCl and 75 mM imidazole, dialyzed the preparation against 10 mM Tris-HCl (pH 8.0) and subjected it to immunoblot analysis.

We also expressed various domains of RIG-I and the full-length recombinant protein in the bacterial expression vector pGEX-5X-1, following the instructions provided by the manufacturer. We then amplified cDNA isolated from HeLa cells using primers BamHI-Card-F (5’CGT Ggg atc ccC ATG ACC ACC G AG CAG CGA-3’) and XhoI-Card-R (5’-CG ctc gag GTG ATG ATG ATG ATG ATG CAA GTT TGT ATC AGA CC); EcoRI-DexH-F (CC gaa ttc AGC CCA TTT AAA CCA AGA) and XhoI-DexH-R (CG ctc gag GTA ATG ATG ATG ATG ATG AAC TTG CTC CAG TTC CTC); EcoRI-452-641-F (CC gaa ttc TAT AAG CCC CAG AAG) and XhoI-452-641-R (CG ctc gag GTG ATG ATG ATG ATG ATG CAC AAG TGC TCT GGT TTT); and EcoRI-Helicase-F (CC gaa ttc GAC GCT TTA AAA AAT TGG) and XhoI-Helicase-R (CG ctc gag GTG ATG ATG ATG ATG ATG TTT GGA CAT TTC TGC TGG). Bacterial pellets were resuspended in PBS containing 0.5% Triton X-100 and then incubated with Glutathione Sepharose 4B beads for 30 min at 4°C. The beads were washed twice with PBS and twice with Factor Xa buffer [20 mM Tris (pH 7.4), 50 mM NaCl, 1 mM CaCl₂]. After resuspension in 500 µl of Factor Xa buffer we added 5 µl of Factor Xa and incubate the slurry for 3 h at room temperature. The supernatants were subjected to chromatography on a TALON resin, as decribed above.

In vitro binding and ubiquitination assays

To assess whether direct interaction(s) accounted for the association between HSP90 and RIG-I we mixed 1 pmol each of recombinant HSP90 and either His-tagged RIG-I or His-tagged HMG-1 as a control, and incubated the mixtures in binding buffer [200 mM NaCl, 50 mM Hepes, 1% BSA (pH 7.4) for 30 min at 30°C in a total volume of 20 µl. We then added a suspension of TALON resin in PBS containing 0.5% Triton X-100 (final volume: 500 µl) and rocked the mixtures for 60 min at 4°C. The beads were extensively washed with PBS containing 0.5% Triton X-100 and treated with 2x SDS-sample buffer. We subjected the eluted proteins to electrophoresis on SDS-PAGE, followed by immunoblot analyses using anti-HSP90 and anti-RIG-I antibodies.

To assess geldanamycin-related effects on ubiquitination of RIG-I we incubated 1 pmol of His-tagged RIG-I in 50 mM Tris-HCl (pH 7.4), 20 mM MgCl₂, 2 mM ATP, 10 mM creatine phosphate, 10 IU/ml creatine kinase, with a rabbit reticulocyte lysate (33.3%) in the presence or absence of geldanamycin (400 nM) for 60 min at 30°C. The total volume was adjusted to 20 µl. We then added a suspension of TALON resin in PBS containing 0.5% Triton X-100 (final volume: 500 µl) and rocked the mixtures for 60 min at 4°C. The beads were extensively washed with PBS containing 0.5% Triton X-100 and treated with 2× SDS-sample buffer. We subjected the eluted proteins to electrophoresis on SDS-PAGE, followed by immunoblot analyses using anti-ubiquitin and anti-RIG-I antibodies.

Immunoprecipitation studies

HeLa cells subjected to various treatments were washed twice with ice-cold PBS, and then lysed in immunoprecipitation buffer [20 mM Tris-HCl (pH 7.4), 100 mM NaCl, 1.5 mM MgCl₂, 1% Triton-X 100] containing 0.5% protease inhibitor cocktail, for 30 min on ice. The lysates were centrifuged at 12,000 g for 10 min at 4°C, and the pellets were discarded. The supernatant fractions were pre-cleared by incubation with normal IgG and protein A/G-plus for 2 hrs. Immunoprecipitations were performed by overnight rocking of the pre-cleared supernatants with 1 µl of the primary antibody indicated in each case, at 4°C. The antigen-antibody complexes were harvested by incubation with protein A/G-agarose for 1 hour at 4°C. FLAG-tagged proteins were harvested by overnight incubation with FLAG-agarose beads at 4°C. After extensive washing with ice-cold PBS, the beads were boiled for 5 min in 2x SDS-PAGE sample buffer to dissociate the immunoprecipitated proteins. These fractions were analyzed by electrophoresis on SDS-PAGE and immunoblot analyses, as described below. Studies involving co-immunoprecipitation of endogenous proteins were conducted on the human colon cancer cell line HT-29 using essentially the same approach, except that immunoprecipitations were performed by overnight rocking of the pre-cleared supernatants with 1 µl of an anti-RIG-I antiserum at 4°C. The complexes were harvested by incubation with protein A/G-plus for 1 hr at 4°C. After extensive washing with ice-cold PBS containing 0.5% Triton X-100, the beads were boiled for 5 min in 2X SDS-PAGE sample buffer and the eluted proteins were subjected to SDS-PAGE.

Immunoblot analyses

We subjected the protein extracts indicated in each case to electrophoresis on 8.5% SDS-PAGE gels, and then transferred the proteins to PVDF membranes that were blocked for 60 min at room temperature in 1x TBST buffer [50 mM Tris-HCl (pH 7.5), 250 mM NaCl, 0.1% Tween 20] containing either 3% nonfat dry milk or 3% BSA. The membranes were incubated for 60 min at room temperature with the primary antibody indicated in each case. After 5 washes with blocking solution, we added an HRP-conjugated secondary antibody, incubated the membranes for 60 min at room temperature, repeated the washes using TBST, and then visualized the immunoreactive bands using chemiluminescence detection reagents.

Immunofluorescence analyses

HeLa cells grown on glass coverslips were rinsed in PBS, fixed with 4% formaldehyde for 20 min, permeabilized with 0.1% Triton X-100 for 10 min, and blocked with 5% BSA for 1 h. All solutions were prepared in PBS. The cells were then incubated for 1 h with a polyclonal anti-RIG-I antibody (1:250-fold dilution) and a monoclonal anti-HSP90 antibody (1:100). After a washing step, we incubated the coverslips with Alexa 488–conjugated anti–rabbit and Texas red–conjugated anti–mouse IgGs. Excess reagents were washed with PBS and the expression levels of RIG-I and HSP90 were analyzed by confocal microscopy (Olympus FV300).

Mass spectrometry

HeLa cell proteins that co-immunoprecipitated with RIG-I were resolved on SDS-PAGE and identified by staining, and individual lanes were subjected to in-gel trypsin digestion at the Mass Spectrometry Core Facility, University of Utah. The tryptic peptides were identified by MS/MS studies and analyzed using MASCOT software (Matrix Science, London, UK).

Reporter assays

To generate an IFN-β reporter vector that spanned nt −299 to +19, we amplified HeLa cell genomic DNA with a sense primer harboring a 5’-XhoI site (XhoI-IFN-β-299, 5’-GGctcgagAGTGTTTTGAGGTTCTTGA-3’) and an antisense primer designed with a 5’-HindIII site (HindIII-IFN-β+19, 5’-GCCaagcttAAGGTTGCAGTTAGAATGT-3’). The product was cloned in the pGL3-basic vector and 300 ng of the plasmid were co-transfected with 50 ng of pSV40-β-gal and 100 ng of a p3XFLAG7.1-expression vector (RIG-I full length or RIG-CARD) using Lipofectamine. The following day we transfected the cells with 200 ng of poly (I:C) using Lipofectamine 2000 and incubated the cells in the presence or absence of 200 nM of geldanamycin for 24 h. We harvested the cells in 200 µl of Reporter Lysis Buffer and assessed luciferase and β-galactosidase activities in the extracts (n = 4).

Results

Identification of heat shock proteins as partners of RIG-I

Our initial goal was to identify protein partners that associated with RIG-I under basal conditions. We utilized HT-29, a human colonic adenocarcinoma cell line that expresses high levels of RIG-I (our unpublished observations). We subjected HT-29 lysates to immunoprecipitation using anti-RIG-I polyclonal antibodies or control IgG. After electrophoretic separation of the immunoprecipitated proteins, we noted the presence of multiple species that specifically associated with RIG-I (Fig. 1A). We initially focused our attention on the identification of RIG-I-interacting proteins that ranged in size between 70 and 110 KDa, as these represented a considerable proportion of the immunoprecipitated material (Fig. 1A). We analyzed various species using mass spectrometry and identified the two largest proteins (100 and 110 KDa) as RIG-I (not shown). These results validated the specificity of our immunoprecipitation approach and suggested that the smaller, less abundant 100 KDa species was a proteolytic fragment derived from full-length RIG-I (Mr = 110 KDa). We next analyzed the composition of the co-immunoprecipitated 90 KDa and 70 KDa proteins (Fig. 1A, arrow and asterisk, respectively). We found that the 90 KDa RIG-I-interacting protein harbored peptides that precisely matched sequences present in human HSP90-α (Table I). Our studies also revealed the presence of a 70 KDa protein in immunoprecipitates generated using both control IgG and anti-RIG-I antibodies (Fig. 1A). The amount of 70 KD protein in the latter appeared to be much higher than that observed in the control lane, suggesting that RIG-I may specifically interact with this protein. Additional studies revealed that the 70 KDa protein was heat shock 70 KDa protein 8 (also known as HSPA8 and HSP70, Table I). This conclusion was drawn based on the sequence of seven peptides identified by mass spectrometric analyses and that precisely matched sequences found in the 70 KDa human chaperone (Table I). Our data supported the existence of in vivo partnerships between RIG-I and select heat shock proteins.

Fig. 1. RIG-I associates with HSP90.

A. Cell lysates from HT-29 cells were immunoprecipitated with control IgG (left lane) or with anti-RIG-I antibody (right lane) and proteins associated with the RIG-I immune complex were resolved by SDS-PAGE (7.5%). The arrow and asterisk indicate the relative mobilities of HSP90 and HSP70, respectively. B–C. A plasmid encoding FLAG-tagged RIG-I was transfected into HEK293T cells. Protein-protein interactions were monitored by co-immunoprecipitation using anti-FLAG-conjugated agarose followed by immunoblotting with anti-HSP90 antibody (B) and by co-immunoprecipitation using anti-HSP90 antibodies followed by immunoblotting with an anti-FLAG antibody (C). D. HeLa cells were immunostained with anti-RIG-I or control rabbit IgG (green; left) or with anti-HSP90 or control mouse IgG (red; center). The left and center panels represent sequential confocal images of the same field. The right panel depicts an overlay of the images. E. Interaction between endogenous RIG-I and HSP90 was assessed by subjecting lysates from HT-29 cells to immunoprecipitation with anti-RIG-I or a control IgG. The immune complexes were isolated by incubation with protein A/G-plus and eluted from the beads with 2XSDS-PAGE buffer. The eluted proteins were subjected to electrophoresis on SDS-PAGE and immunoblot analysis, using anti-RIG-I and anti HSP90 antibodies. F. Purified recombinant HSP90 was incubated with buffer (left panel), His-tagged RIG-I (center panel), or His-tagged HMG-1 as a control (right panel). The resulting mixtures were subjected to affinity purification, and the bound proteins were eluted and subjected to immunoblot analyses using anti-HSP90 and anti-His tag antibodies.

Table I.

Mass spectrometric identification of HSP90-α and HSP70 protein 8

Protein Size	Tryptic peptide sequences Identified by MS/MS	Identified protein	Location in protein
90 KDa	ADLINNLGTIAK VILHLKEDQTEYLEER NPDDITNEEYGEFYK GVVDSEDLPLNISR	HSP90-α (HSP90AA1, HSP90AA2, HSP90A, HSP90N, HSPC1, HSPCA, HSPCAL1, HSPCAL3, HSPCAL4, HSPN, HSP86, HSP89, HSP90,LAP2)
70 KDa	TTPSYVAFTDTER SFYPEEVSSMVLTK TVTNAVVTVPAYFNDSQR DAGTIAGLNVLR IINEPTAAAIAYGLDKK FEELNADLFR SQIHDIVLVGGSTR	HSP70 protein 8 (HSPA8, HSC54, HSC70, HSC71, HSP71, HSPA10. LAP1, NIP71)

Molecular characterization of RIG-I/HSP90 interaction

We next focused on the characterization of the interaction between HSP90 and RIG-I. To address this issue, we used molecular, immunohistological and biochemical approaches. First, transfection of HEK293T cells with FLAG-tagged RIG-I followed by immunoprecipitation using anti-FLAG or anti-HSP90 antibodies revealed that endogenous HSP90 interacted with exogenous RIG-I (Fig. 1B,C). We next considered whether the interactions observed in transfected cells were the consequence of forced expression of RIG-I and did not represent a naturally- occurring association between endogenous proteins. To address this issue, we tested the ability of endogenous RIG-I and HSP90, both of which have been reported to be expressed in the cytoplasm, to co-localize under basal conditions (14, 17). Although no particular cellular structure was highlighted, we observed that in HeLa cells HSP90 and RIG-I were expressed in the same cytoplasmic region, suggesting that the proteins formed part of the same complex (Fig. 1D). In addition, immunoprecipitation studies confirmed that endogenous RIG-I interacted with endogenous HSP90 (Fig. 1E), indicating that the association between these proteins occurs in intact cells and is not an artifact of overexpression systems.

We next asked whether RIG-I and HSP90 interacted directly or if participation of an additional protein(s) was required for association. To test this, we conducted in vitro experiments with purified recombinant proteins. We found that incubation of purified, His-tagged RIG-I with recombinant HSP90 resulted in the association between the proteins, as shown by the binding of both species to a His-tag purification resin (Fig. 1F). Control experiments revealed no binding of HSP90 to the beads in the absence or presence of a control His-tagged protein (HMG-1, Fig. 1F). These combined studies provide evidence indicating that the association between RIG-I and HSP90 involves direct interaction(s) between the proteins.

Identification of domains required for interaction between HSP90 and RIG-I

Our next goal was to identify RIG-I domains involved in its interaction with HSP90. In previous studies, it was demonstrated that the ability of RIG-I to interact with the adaptor protein IFN-β promoter stimulator 1 (IPS-1) and the ubiquitin ligases RNF125 and TRIM25 depended on the presence of the amino-terminal CARD domain of RIG-I (5, 6, 14, 15). Second, it was recently shown that the CARD domain is the primary binding site of HSP90-β on Apaf-1 (18). To systematically analyze individual contributions, we generated four FLAG-tagged deletion mutants harboring various domains of RIG-I (Fig. 2A). We transfected HEK293T cells with each deletion construct and found robust expression of all species (Fig. 2B, left panel). Next, we subjected cell lysates from transfected cells to immunoprecipitation using anti-HSP90 antibodies and found that the CARD domain lacks the ability to associate with HSP90 (Fig. 2B, right panel). The remaining deletion constructs interacted with HSP90 in an efficient manner, suggesting that a domain common to these constructs was necessary for the interaction. An examination of the deletion mutants tested in this study allowed us to conclude that the region comprised between amino acids 452–641 was involved in the association between RIG-I and HSP90.

Fig. 2. The CARD domain is not required for association of RIG-I with HSP90.

A. Schematic representation of RIG-I mutant constructs utilized in this portion of the study. B. HEK293T cells were transfected with the constructs indicated. Twenty-four hours after transfection, the cells were harvested and the levels of expression of each construct were assessed by immunoblot analysis (left panel). In addition, aliquots of the lysates were subjected to immunoprecipitation using anti-HSP90 antibodies, followed by immunoblotting with an anti-FLAG antibody (right panel).

To further explore this issue and to assess whether additional RIG-I domains also contributed to its interaction with HSP90, we expressed His-tagged constructs harboring various regions of RIG-I (Fig. 3A). We found that all the constructs were expressed, albeit to various extents (Fig 3B, bottom panel). In addition, we found that the CARD domain, which was efficiently expressed, lacked the ability to interact with HSP90, and that deletion of the CARD domain did not affect association with HSP90. These combined studies are in complete agreement with our results using FLAG-tagged constructs (Fig. 2B) and they firmly establish that the CARD domain does not participate in the interaction between RIG-I and HSP90. Interestingly, our studies also showed that all of the remaining RIG-I domains retained the ability to bind HSP90 (Fig. 3B), suggesting that more than one domain participates in the interaction between the proteins.

Fig. 3. Several RIG-I domains can mediate binding to HSP90.

A. Schematic representation of purified RIG-I mutant constructs utilized in this portion of the study. B. Purified recombinant HSP90 and the constructs indicated were incubated with purified, full-length His-tagged RIG-I, or various purified His-tagged RIG-I domains. The resulting mixtures were subjected to affinity purification, and the bound proteins were eluted and subjected to immunoblot analyses using anti-HSP90 and anti-His antibodies.

HSP90 affects the levels of RIG-I protein

HSP90 is primarily known as a molecular chaperone owing to its ability to ensure the stability and correct conformation, activity, intracellular localization and proteolytic cleavage of a range of proteins involved in cell growth, differentiation and survival (19). These include key signaling molecules such as ERBB2, AKT, Raf, hypoxia inducible factor-1α and steroid hormone receptors (20). To investigate the functional consequences of HSP90/RIG-I association, we destabilized the interaction using pharmacological approaches. Geldanamycin is a naturally-occurring ansamycin antibiotic and anti-tumor agent that binds tightly to the HSP90 ATP/ADP pocket and prevents association of client proteins (20, 21). Treatment with geldanamycin prevented association between transfected RIG-I and HSP90 (Fig. 4A) and significantly reduced the levels of endogenous RIG-I protein, in a time-dependent manner (Fig. 4B). To assess the efficacy of geldanamycin treatment we took advantage of a study reported by Piatelli and co-workers who showed that geldanamycin selectively depletes cellular Raf-1, thus interrupting activation of MEK1/2 and ERK (22). In addition, these investigators showed that geldanamycin does not affect the levels of total Erk (22). We found that geldanamycin decreased phospho-Erk levels, thus confirming the efficacy of the treatment in our studies (Fig. 4B). We next investigated whether these effects were the consequence of altered rates of mRNA transcription or decreased mRNA stability, as was shown for other proteins (21). However, quantitative assessment of RIG-I mRNA levels demonstrated that transcriptional effects did not account for the geldanamycin-mediated decrease in RIG-I protein levels (Fig. 4C). Higher concentrations of geldanamycin (> 2 µM) were avoided as these resulted in generalized cytotoxic effects (not shown). Our results indicated that the chaperone activity of HSP90 protects RIG-I from proteolytic degradation, suggesting that this mechanism contributes to the definition of steady-state levels of RIG-I.

Fig. 4. HSP90 stabilizes RIG-I.

A. HEK293T cells were transfected with a plasmid encoding FLAG-tagged RIG-I. Twenty-four hours after transfection, we supplemented vehicle or geldanamycin (100 nM) for 2 h. Protein-protein interactions were monitored by immunoprecipitation using anti-FLAG-conjugated agarose beads, followed by immunoblotting with anti-HSP90 antibodies. B. HeLa cells were treated with geldanamycin (100 nM) for the indicated time periods and cell lysates were then subjected to SDS-PAGE and immunoblotting using antibodies against RIG-I, phospho-ERK, HSP90, and actin. The numbers above this panel depict quantitative assessments of band intensity relative to actin controls, performed using Image-J. C. HeLa cells were treated with geldanamycin (100 nM) for the indicated time periods. Total RNA was then extracted and the levels of RIG-I were assessed by quantitative RT-PCR.

HSP90 protects RIG-I from ubiquitination and proteasomal degradation

Our next goal was to identify the protease system that participates in the degradation of RIG-I following disruption of the RIG-I/HSP90 complex. We hypothesized that degradation of RIG-I might involve the proteasome because most proteins known to be stabilized by HSP90 are degraded through this mechanism once the interaction is inhibited (23). In control experiments, we found that treatment with MG-132, an agent known to specifically inhibit proteasomal activity, led to substantial accumulation of RIG-I (Fig. 5A). To specifically assess the contribution of HSP90 to the stability of RIG-I we investigated whether the effects of geldanamycin on RIG-I levels were affected when proteasomal activity was inhibited with MG132. We found (Fig. 5A) that MG132 prevented geldanamycin-mediated RIG-I proteolysis, thus providing evidence for a role of HSP90 activity in the protection of RIG-I from proteasomal degradation.

Fig. 5. HSP90 protects RIG-I from proteasomal degradation and ubiquitination.

A. HeLa cells were treated with geldanamycin (400 nM) in the absence and presence of MG-132 (0–10 µM) for 16 hours. The cells were harvested and the lysates were subjected to SDS-PAGE and immunoblot analyses using antibodies against RIG-I, HSP90, and actin. The numbers below this panel depict quantitative assessments of band intensity. B. We incubated His-tagged RIG-I with a rabbit reticulocyte lysate in the presence (right panel) or absence (left panel) of geldanamycin (400 nM) for 60 min at 30°C, and then purified RIG-I using affinity chromatography on TALON beads. We eluted the bound proteins and subjected them to electrophoresis on SDS-PAGE, followed by immunoblot analyses using anti-ubiquitin and anti-RIG-I antibodies.

Protein degradation through the proteasome commonly involves conjugation with ubiquitin, and HSP90 has been shown to utilize this mechanism in the stabilization of a variety of proteins. In addition, the degradation of RIG-I requires derivatization with ubiquitin (14, 15). To assess whether HSP90-mediated effects on RIG-I involved the ubiquitin pathway, we developed an in vitro assay that allowed us to assess changes in the extent of ubiquitination of RIG-I. We incubated purified, His-tagged RIG-I with a ubiquitination-competent rabbit reticulocyte system, isolated RIG-I using affinity chromatography, and then assessed the extent of ubiquitination by immunoblot analyses. We found that under basal conditions, a fraction of RIG-I was present as a high molecular weight ubiquitinated species (Fig. 5B, top left panel). In addition, dissociation of the RIG-I/HSP90 complex with geldanamycin stimulated RIG-I ubiquitination, as demonstrated by robust increases in the extent of RIG-I ubiquitination and by attenuated expression of un-derivatized RIG-I (Fig. 5B). We did not detect high molecular weight forms of RIG-I using anti-RIG-I antibodies, even after treatment with geldanamycin, suggesting that our RIG-I antibodies were unable to recognize ubiquitinated forms of the protein. Our combined data are consistent with a model whereby blockade of RIG-I/HSP90 interactions results in ubiquitination and proteasomal degradation of RIG-I.

The interaction between HSP90 and RIG-I is functionally important

Our final goal was to assess whether the interaction between HSP90 and RIG-I had functional consequences on known biological activities mediated by RIG-I. It was previously shown that RIG-I acts as an intracellular dsRNA receptor that regulates interferon regulatory factor 3 activation and IFN-β induction at the transcriptional level (24). Thus, we investigated whether the ability of RIG-I to transcriptionally activate the IFN-β gene required association with HSP90. We transfected 293T cells with a reporter construct harboring nt −299 to +19 of the IFN-β gene combined with either empty vector, a FLAG-tagged RIG-I construct harboring full-length RIG-I, or a FLAG-tagged CARD construct. We stimulated the cells with poly (I:C) which has recently been shown to be recognized by RIG-I (25) in the absence and presence of geldanamycin. We found that geldanamycin significantly inhibited both basal and poly (I:C)-stimulated IFN-β promoter activation (Fig. 6) suggesting that the association between RIG-I and HSP90 was necessary for this response. In addition, we found robust basal IFN-β promoter activation mediated by the CARD domain of RIG-I that was not affected by geldanamycin treatment (Fig. 6). These results indicate that blocking the ability of RIG-I to interact with HSP90 inhibits its biological activity. The fact that IFN-β promoter activation mediated by the CARD domain was not affected by geldanamycin is consistent with our studies (Fig. 2 and Fig. 3) showing that the CARD domain is not required for association of RIG-I with HSP90.

Fig. 6. Inhibition of HSP90/RIG-I association abolishesRIG-I-mediated transcriptional activation of the IFN-β promoter.

HEK293T cells were co-transfected with a reporter construct harboring nt−299/+19 of the IFN-β gene, a β-galactosidase expression vector, and either full-length RIG-I full length or the CARD domain. The following day we transfected half of the cells with poly (I:C) in the presence or absence of geldanamycin for 24 h; the remaining cells were treated with vehicle or geldanamycin. We assessed luciferase and β-galactosidase activity in the extracts (n=4).

Discussion

RIG-I has been described as a pattern-recognition receptor that detects intracellular dsRNA through its helicase domain in a TLR-independent fashion (5, 6). A series of well-characterized downstream signaling events that require the CARD domains of RIG-I lead to the activation of a family of IFN-induced genes such as IRF3, IRF7 and NFkB, through the signaling adaptor protein IPS-1 (6). The ability of RIG-I to elicit downstream signaling events requires the formation of a macromolecular protein complex necessary for gene activation (26). This complex is most likely in dynamic equilibrium depending on the state of cellular activation, viral stimulation, and other factors. To identify novel proteins that associated with RIG-I under basal conditions, we used biochemical approaches that, when combined with mass spectrometric measurements, resulted in the discovery of HSP90, and perhaps HSP70, as partners of RIG-I. HSP90 is a ubiquitous, constitutively expressed, highly conserved molecular chaperone involved in the folding, activation and assembly of key mediators of signal transduction, cell-cycle control and transcriptional regulation (27). This chaperone interacts with native and denatured partners such as protein kinases, transcription factors, viral-replication proteins and a range of intracellular receptors, affecting their turnover, trafficking, cellular localization, and activity (28, 29). Recent studies by Kim and co-workers demonstrated that HSP90 associates with 40S ribosomal components and prevents ubiquitin-dependent ribosomal protein degradation (23). Our data point to a similar mechanism in which partnership between HSP90 and RIG-I serves to stabilize this protein. In addition, we provide novel evidence showing that the interaction between HSP90 and RIG-I is direct and independent of the CARD domain. Interestingly, we found that the Dex-H Box, the Helicase and other domains of RIG-I efficiently associated with HSP90. These results suggest that the interaction between these proteins may involve multiple domains. Similar results were reported by Kim et al. who showed that interaction of HSP90 with ribosomal components involves interaction with two domains of ribosomal protein S3 (23).

Our studies also provide evidence suggesting that one of the functional consequences of the interaction between HSP90 and RIG-I is the shielding of the latter from ubiquitin-dependent proteasomal degradation. The mechanisms that regulate the levels of RIG-I protein under various conditions have been partially characterized and clearly require ubiquitin-mediated functions at different levels. Arimoto and co-workers reported that the ubiquitin ligases RNF125 and UbcH8 are key determinants of RIG-I levels through their ability to catalyze the conjugation of RIG-I to ubiquitin or ISG15, respectively (15). The authors described participation of the E2 and E3 ubiquitin ligases in a regulatory circuit the function of which is to control the levels of RIG-I protein (30). In addition, Lin and co-workers found that A20, a virus-inducible protein that harbors two ubiquitin-editing domains, functions as a negative regulator of RIG-I through indirect mechanisms that require degradation of an as yet unidentified adapter of the RIG-I pathway (31). These combined observations reveal the existence of several control points to insure that appropriate levels of RIG-I are expressed in response to changing cellular needs.

In humans, two distinct genes, HSP90-α and HSP90-β, encode closely related cytoplasmic proteins that display minor sequence differences at the amino termini (32–34). The isoforms differ in their ability to activate certain client proteins and regulate cellular functions of immune cells (34–38). The α and β isoforms of HSP90 often cooperate with other proteins to generate a cytosolic multichaperone machinery that includes HSP70, peptidyl-prolyl isomerases and other co-chaperones, to achieve adequate protein folding (39). Certain cytosolic client polypeptides have been shown to first bind HSP70 and then interact with dimeric HSP90 (39–42). Interestingly, our studies suggest that HSP70 may be a second chaperone that interacts with RIG-I under basal conditions. Our work did not directly address whether HSP70 facilitates the formation or stability of the RIG-I/HSP90 complex, but it is tempting to speculate that HSP70 contributes to this process, and that a co-chaperone(s) may be involved as well. Pratt and Toft have pointed out that many of the client proteins of HSP90 are involved in signal transduction and speculated that HSP90 may facilitate access of its client proteins to diverse intracellular sites and/or ligands (17), thus affecting their ability to elicit biological activities. We hypothesized that the formation of a complex between RIG-I and HSP90 had important functional consequences additional to protecting RIG-I from proteasomal degradation. Indeed, we found that interfering with the ability HSP90 to associate with RIG-I prevented RIG-I-mediated transcriptional activation of the IFN-β promoter. An interesting observation resulting from our studies was the finding that the CARD domain of RIG-I, which does not bind to HSP90, transcriptionally activated the IFN-β promoter in a robust, dsRNA-independent manner. A number of mechanisms could account for this effect, but one possibility is that stimulation of the HSP90/RIG-I complex with dsRNA results in conformational changes that facilitate CARD domain-mediated functions. However, other scenarios include HSP90-mediated regulation of the ability of RIG-I to respond and/or bind to ligands, or to undergo protein modification. These possibilities await further investigation.

While HSP90 seems to play beneficial physiological roles, this chaperone has been shown to stabilize oncogenic proteins, thus contributing to cancer progression (27). Increased expression of HSP90 is associated with tumorigenesis and small molecule HSP90 inhibitors that include geldanamycin and its derivatives selectively kill certain types of cancer cells by promoting apoptosis (43). In addition, HSP90 is recruited in cells undergoing infection as a result of virus-mediated sequestration of the host’s chaperone to facilitate viral assembly and replication (27, 44). Thus, our results, combined with those of Kim and co-workers (23), suggest that the ability of HSP90 to protect endogenous proteins from degradation may represent a physiological regulatory mechanism, and that this function may be hijacked in settings of viral infection or oncogenic transformation, resulting in pathological consequences.

In summary, we have shown that RIG-I exists as a complex with the chaperone HSP90. The association is direct and stabilizes RIG-I by inhibiting its ubiquitination and proteasomal degradation. In addition, the formation of an HSP90/RIG-I complex is essential for expression of functional activities. Our study, combined with other recent reports (15, 30) demonstrate that ubiquitination plays dual roles in the regulation of RIG-I activity and protein levels, thus providing sophisticated mechanisms to control its function and expression in temporally-and spatially-regulated manners. The function of HSP90 may be compromised in settings of infection owing to virus-mediated chaperone sequestration, a process thought to facilitate viral assembly and replication (27, 44). These findings provide new insights into the physiological mechanisms that define expression of a key protein thought to profoundly impact innate anti-viral responses in mammals.

Abbreviations used in this paper

RIG-I

retinoic acid-inducible gene-I

CARD

caspase recruitment domain

HSP70

Heat shock protein 70

HSP90

Heat shock protein 90

CHI

cycloheximide

dsRNA

double stranded RNA

TLR

Toll-like receptor

RNF125

RING finger protein 125

TRIM25

Tripartite motif protein 25

IFN

interferon

ISG15

IFN-regulated gene 15

IPS-1

IFN-beta promoter stimulator 1

UCRP

ubiquitin cross-reacting protein

FBS

fetal bovine serum

PVDF

polyvinylidene fluoride