The Structural Basis of IRF-3 Activation upon Phosphorylation

The innate immune system detects the pathogen-associated molecular patterns, such as nucleic acids, LPS through pattern recognition receptors, which trigger the induction of a variety of cytokines, including IFN-I, to initiate host defense against pathogens (1–7). Viral or bacterial DNA in cytosol are recognized by cyclic GMP-AMP synthase (cGAS), which catalyzes the synthesis of a cyclic dinucleotide cyclic GMP-AMP (cGAMP). cGAMP binds to the adaptor stimulator of IFN genes (STING) and mediates the recruitment and activation of TANK-binding kinase 1 (TBK1) and IFN regulatory factor (IRF) 3 (5, 8). Activated IRF-3 translocates to the nucleus and initiates the transcription of the IFN-β gene with other transcription factors, such as NF-κB (7, 9–13). dsRNA and LPS can be recognized by retinoic acid-inducible protein 1 (RIG-I)–like receptors (RLRs) and TLRs, respectively, to induce the expression of IFN-Is through the adaptors mitochondrial antiviral signaling (MAVS) and TIR domain-containing adaptor-inducing IFN-β (TRIF). Interestingly, all three of these signaling pathways converge at the recruitment of IRF-3 via a conserved pLxIS motif (p, hydrophilic residue, x, any residue, S, phosphorylation site) within the adaptor proteins (9–13).

The IRF family transcription factors contain nine members (IRF-1 through IRF-9). These transcription factors contain a highly conserved N-terminal DNA-binding domain and a relatively divergent C-terminal regulatory domain, suggesting that most members function nonredundantly (14). IRF-3 is a key transcription factor that regulates the expression of IFN-I genes. Under resting conditions, IRF-3 adopts an autoinhibited conformation and is ubiquitously accumulated in the cytoplasm. Bacterial or viral infection triggers the activation of IRF-3 through various innate immune sensing pathways. Phosphorylated IRF-3 binds to p300/CREB-binding protein (CBP), translocates to the nucleus, and initiates the transcription of IFN-I genes (14–16). Interestingly, IRF-7, which is closely related to IRF-3 in terms of the conserved regulatory domain, also plays an important role in regulating the expression of IFN-Is (17).

The mechanism of IRF-3 activation has been extensively studied, showing that IRF-3 is activated through phosphorylation of the C-terminal serine-rich repeat (16). Previous studies by the Fujita laboratory (18–20) showed that the phosphorylation site 1 (residues Ser³⁸⁵ and Ser³⁸⁶ of human IRF-3 [hIRF-3]) plays a key role in IRF-3 activation. The phosphorylation of Ser³⁸⁶ induced by viral infection has been detected by a specific Ab. They also observed that the dimerization of IRF-3 was abolished by the mutation of Ser³⁸⁶. In addition, the Hiscott laboratory (21, 22) observed that the phosphorylation site 2, which includes residues Ser³⁹⁶, Ser³⁹⁸, Ser⁴⁰², Thr⁴⁰⁴, and Ser⁴⁰⁵ of hIRF-3, plays a critical role in IRF-3 activation. The phosphomimetic mutation of these residues (IRF-3 5D) results in a constitutively active phenotype. Moreover, they observed that the S396D mutation alone induces IFN-I expression, suggesting that Ser³⁹⁶ also plays a critical role in IRF-3 activation. Another study by the Harrison laboratory (23) proposed a two-step phosphorylation and activation model, which suggests that phosphorylation at site 2 leads to the alleviation of IRF-3 autoinhibition that facilitates the phosphorylation at site 1 and eventually leads to the activation of IRF-3. Later on, studies by the Lin group (24) showed that IRF-3 mutant S386D/S396D bound to CBP forms a stable oligomer, suggesting that the phosphorylation of both Ser³⁸⁶ and Ser³⁹⁶ is essential for hIRF-3 activation. Based on these studies, we mutated Ser³⁸⁶ and Ser³⁹⁶ to glutamic acid in a truncated form of IRF-3 (residues 189–398) and determined the structure of the phosphomimetic IRF-3 in complex with CBP, revealing that the phosphomimetic IRF-3 mutant forms a dimer (12). However, the higher affinity between phosphorylated hIRF-3/CBP (phIRF-3/CBP) complex compared to that of the S386/396E mutant indicates that the phosphomimetic mutation does not fully recapitulate the interactions between phosphorylated IRF-3.

To elucidate the exact mechanism of IRF-3 activation upon phosphorylation by TBK1, we coexpressed truncated forms of hIRF-3 and mouse IRF-3 (mIRF-3) C-terminal domains with a CBP fragment, phosphorylated these complexes with TBK1, and determined the crystal structures of the phosphorylated IRF-3/CBP dimers. These structures reveal the molecular basis of IRF-3 activation upon phosphorylation. Biochemical and functional studies based on the structures show that mutations of the key residues mediating IRF-3 dimerization upon phosphorylation dramatically impair IRF-3 dimerization and IRF-3–mediated signaling.

The cDNA encoding hIRF-3 (residues 189–398) and mIRF-3 (residues 184–390) dimerization domains were cloned into a modified pET-28a (+) vector containing an N-terminal His₆-SUMO tag with appropriate primers obtained from Integrated DNA Technologies. SUMO fusion of human CBP (residues 2065–2111) was cloned into the pET-22b (+) vector using appropriate primers from Integrated DNA Technologies. Sequences of all the constructs were confirmed by DNA sequencing. The plasmid containing hIRF-3 or mIRF-3 dimerization domain was cotransformed with CBP plasmid into Escherichia coli BL21 (DE3) cells. The cells were grown on Luria–Bertani agar plates containing both kanamycin and ampicillin. On the next day, the cell colonies from the plates were transferred to 6 liters of Luria–Bertani liquid medium in flasks with kanamycin and ampicillin in an incubator shaker at 37°C under 225 rpm. When OD₆₀₀ reached ∼1.2, BL21 cells were induced with 0.4 mM isopropyl β-D-1-thiogalactopyranoside overnight at 16°C. The cells were harvested by centrifugation at 4000 rpm for 10 min and then suspended in a 200-ml lysis buffer containing 300 mM NaCl and 50 mM Tris-HCl (pH 8). The cells were lysed by sonication for 10 min with a 0.5-s pulse and a 0.5-s rest, and the cell lysate was centrifuged at 16,000 rpm for 30 min. The supernatant was loaded onto a Ni²⁺-NTA column (QIAGEN). Then, 200 ml of washing buffer containing 500 mM NaCl, 20 mM Tris-HCl, 25 mM imidazole (pH 7.5) was used to wash nonspecific binding proteins off the Ni²⁺-NTA column. The target proteins were then eluted with 75 ml of elution buffer containing 150 mM NaCl, 20 mM Tris-HCl, and 250 mM imidazole (pH 7.5). The His₆-SUMO tag was cleaved with SUMO protease at a concentration of 10 μg/ml at 4°C overnight and removed using an Ni²⁺-NTA column. The target proteins in the flow-through were centrifuged to ∼2 ml and further purified by gel filtration chromatography using a HiLoad 16/60 Superdex 75 Column (GE Healthcare) equilibrated with a running buffer containing 150 mM NaCl and 20 mM Tris-HCl (pH 7.5). All the mutations were introduced into a full-length hIRF-3 pET-28a (+) plasmid using the QuikChange Site-Directed Mutagenesis Kit (Agilent Technologies) using proper primers. Sequences of all the plasmids were confirmed by DNA sequencing. The IRF-3 mutants were expressed and purified in the same way as the IRF-3 dimerization domain. Mouse TBK1 (mTBK1) was cloned into the pAcGHLTc vector with an N-terminal GST tag and a His₆ tag. The 2-μg plasmid was transfected together with 2.5 μl of BaculoGold Bright Linearized baculovirus DNA (BD Biosciences) into Spodoptera frugiperda (sf9) insect cells to generate recombinant baculovirus. The recombinant viruses were amplified for at least two rounds (4–6 d/round) before the large-scale protein expression. The insect cells at a density of 2.5 × 10⁶ cells/ml were infected with the TBK1 recombinant baculovirus, cultured at 27°C, and harvested 72 h postinfection by centrifugation at 3000 rpm for 10 min. The cells were lysed in a buffer containing 150 mM NaCl, 0.2 M Tris-HCl, 1% NP-40, and 1 mM PMSF (pH 8) in a shaker at 4°C for 2 h. The cell lysate was centrifuged at 16,000 rpm for 30 min. The GST-TBK1 protein in the supernatant was mixed with 6 ml Ni²⁺-NTA beads and incubated in a shaker at 4°C for 2 h. The beads were then spun down and washed three times using a buffer containing 500 mM NaCl, 20 mM Tris-HCl, and 25 mM imidazole (pH 7.5). The target protein was eluted with a buffer containing 150 mM NaCl, 20 mM Tris-HCl, and 250 mM imidazole (pH 7.5). The eluted protein was further purified by gel filtration chromatography using a HiLoad 16/60 Superdex 200 column.

Purified full-length hIRF-3 proteins were mixed with GST-mTBK1 in a ratio of 10:1 (w/w) in a 1-ml reaction buffer with 20 mM HEPES (pH 7.5), 10 mM MgCl₂, 100 mM NaCl, 5 mM ATP, 0.1 mM Na₃VO₄, 5 mM NaF, and 5 mM DTT at 27°C for ∼24 h. The final concentration of the proteins was ∼1 mg/ml. After ∼24 h of incubation, the phosphorylated IRF-3 proteins were analyzed using a Superdex 200 (10/300 GL) column eluted with a buffer containing 20 mM Tris-HCl and 150 mM NaCl (pH 7.5).

Full-length wild-type (WT) IRF-3 and all the mutant proteins were purified using an Ni²⁺-NTA column followed by gel filtration chromatography, as described above. Ten micrograms of each sample was mixed with 5× loading dye of 250 mM Tris-HCl (pH 6.8), 10% SDS, 30% (v/v) glycerol, 10 mM DTT, and 0.05% (w/v) bromophenol blue and then resolved on 4–20% gradient gels in a buffer containing 25 mM Tris, 192 mM glycine, and 0.1% SDS (pH 8.4) at 100 V for 1 h. The gel was stained with Coomassie blue for 1 h and destained with a solution containing H₂O, methanol, and acetic acid in a ratio of 50:40:10 (v/v/v) until the bands were clearly seen. The gel image was taken using a Bio-Rad imager. For native gel electrophoresis, the purified proteins were phosphorylated with GST-mTBK1 using the method described above, and each of the phosphorylated proteins with nonphosphorylated WT was resolved on 10% native gels running in a buffer containing 25 mM Tris and 192 mM glycine (pH 8.4) at 4°C at 100 V for 30 min. The gels were stained and destained the same way as the gradient gels.

Purified hIRF-3 (residues 189–398) and mIRF3 (residues 184–390) bound to the CBP fragment (residues 2065–2111) were phosphorylated by GST-mTBK1. After 24 h of incubation, the proteins were purified using a HiLoad 16/60 Superdex 75 column eluted with 20 mM Tris-HCl and 150 mM NaCl (pH 7.5). The purified phosphorylated proteins were concentrated to a final concentration of ∼5 mg/ml. The crystallization screen was performed by hanging drop vapor diffusion technique at 4°C using Index, Crystal Screen and Crystal Screen 2 reagent kits from Hampton Research. Crystals of hIRF-3 bound to CBP were grown in 0.1 M sodium acetate (pH 5), 0.2 M MgCl₂, and ∼5% polyethylene glycol 3350. Crystals of mIRF-3 in complex with CBP were grown in 0.2 M ammonium citrate tribasic (pH 7) with ∼12% polyethylene glycol 3350. The crystals were flash frozen in liquid nitrogen in the reservoir solution containing 25% (v/v) glycerol. Diffraction data were collected at the Advanced Light Source beamlines 5.0.1, using a Quantum 315r CCD detector. The diffraction data were indexed and integrated with iMosflm and merged with Aimless in the CCP4 package (25). The structures of the pIRF-3/CBP complex were determined by molecular replacement using the structure of our phosphomimetic IRF-3/CBP complex (Protein Data Bank [PDB] identification code: 5JEM) as the search model using Phaser in the Phenix package (26). The structures were manually rebuilt using Coot and refined with Phenix. Details of data quality and structure refinement are summarized in Table I. The structural figures were generated with PyMOL (https://www.pymol.org).

The cDNA encoding WT hIRF-3 was cloned into a pcDNA3.1(−) vector using appropriate primers. Mutants of hIRF-3 were generated using the QuikChange Site-Directed Mutagenesis Kit (Agilent Technologies). Sequences of the mutants were confirmed by DNA sequencing. HEK293T cells were plated in Costar White 96-Well Plates at 4 × 10⁴ cells per well, and each well contained 100 μl DMEM (1×) plus GlutaMAX medium (Life Technologies) supplemented with 10% FBS (Life Technologies). After ∼24 h of incubation at 37°C, the cells were transfected with the IRF-3 plasmids (10 ng per transfection) using Lipofectamine 2000 reagent (Invitrogen) and Opti-MEM (Life Technologies) together with a constant amount of IFN-β firefly luciferase reporter plasmids (20 ng per transfection), phRL-TK–Renilla luciferase plasmids (2 ng per transfection) (Promega), and human STING plasmids (0.2 ng per transfection). Transfections with empty pcDNA3.1(−) and WT hIRF-3 with no STING plasmid were used as controls. The cells were incubated for another 24 h to allow for expression of the genes. The half of the cells in the plates were treated with 30 μg/ml cGAMP dissolved in DMEM (1×) plus GlutaMAX medium, and the other half were treated with the medium only. After ∼16 h of incubation, the cells were analyzed using the Dual-Glo Luciferase Reporter Assay kit (Promega). Luminescence was quantified with the BioTek Synergy HTX Multi-Mode Microplate Reader. The relative firefly luciferase activity was normalized by the Renilla luciferase activity. The relative IFN-β reporter fold of induction represents the ratio normalized to control plasmid values with the same treatment.

HEK293T cells were transfected with empty pcDNA3.1(−), WT IRF-3 or IRF-3 mutants together with human STING plasmid. The cells were stimulated by 30 μg/ml cGAMP added to the culture media 24 h posttransfection. After ∼16 h of incubation, the cells were washed and suspended in PBS and then lysed in 150 mM NaCl, 200 mM Tris-HCl (pH 7.5), 1 mM EDTA, and 1% Nonidet P-40 supplemented with one cOmplete Protease Inhibitor Cocktail EDTA-free Tablet (Roche) and one PhosSTOP Phosphatase Inhibitor Cocktail Tablet (Roche) for each 10 ml of lysis buffer. The proteins were resolved on 10% SDS-PAGE at 100 V for 1.5 h and transferred to a PVDF membrane in a transfer buffer containing 1× Tris-glycine plus 20% methanol for another 1.5 h. Then, 5% milk in 1× PBS with Tween 20 (PBST) solution was used to block the membrane for 1 h, followed by three rinses with PBST. Next, the membrane was incubated with primary Abs dissolved in 1× PBST with 5% BSA overnight at 4°C. The membrane was washed three times using PBST on the next day and further incubated with the corresponding HRP-conjugated secondary Abs dissolved in PBST. The following Abs were used in the Western blot experiment: anti–IRF-3 (1:1000; sc-9082; Santa Cruz Biotechnology) and anti-actin (1: 4000; HHF35; Pierce). The proteins were visualized using the Western Lightening Plus-ECL (PerkinElmer) according to the manufacturer’s protocol. For the detection of phosphorylated IRF-3, half of the cells were transfected with empty pcDNA3.1(−), WT IRF-3 were stimulated by 30 μg/ml cGAMP, and the other half were treated with the medium only. Anti–IRF-3 p-Ser³⁸⁶ (1:2500; ab76493; Abcam) and anti–IRF-3 p-Ser³⁹⁶ (1:1000; 4947S; Cell Signaling Technology) were used.

HEK293T cells were grown on poly-l-lysine–coated coverslips, placed in 12-well plates for 24 h, then cotransfected with 20 ng of WT IRF-3–HA and 1 ng of STING plasmids or 20 ng of mutant IRF-3–HA and 1 ng of STING, respectively, using Lipofectamine 2000 reagent mixed with Opti-MEM (Life Technologies). Twenty-four hours posttransfection, the medium was replaced with fresh DMEM (1×) plus GlutaMAX medium with or without 30 μg/ml cGAMP. After 12 h of incubation, cells were washed using PBS, then fixed by 4% paraformaldehyde in PBS for 15 min at room temperature and permeabilized with PBST containing 0.5% Triton X-100 in PBS. Cells were washed and blocked with 5% FBS in PBST and then incubated with anti–HA tag primary Ab (3724; 1:100; Cell Signaling Technology) overnight. The cells were washed three times with PBS and incubated with Alexa Fluor 488 Goat anti-Rabbit IgG (A11034, 1:1000; Thermo Fisher Scientific) at room temperature for 1 h. The coverslips were washed with PBS, mounted on slides with ProLong Gold Antifade Reagent with DAPI, then imaged under an Olympus FV1000 fluorescence microscope. The scale bars in the images correspond to 20 μm in length. To quantify the amount of nuclear translocation, 12 IRF-3–HA highly expressed cells or cell clusters were randomly selected in each field. The region of interest of nuclear or total IRF-3–HA fluorescence in the same cell or cell cluster was manually drawn. The area of integrated fluorescence intensity was calculated by ImageJ (Version 1.51n).

Molecular mass of WT IRF-3 and phosphorylated IRF-3 were determined by MALDI-TOF mass spectrometry (MS) using a Bruker UltrafleXtreme TOF/TOF mass spectrometer (Protein Chemistry Laboratory, Texas A&M University). The samples were solid-phase extracted using Protea C4 LithTips and analyzed using α-cyanohydroxycinnamic acid as a matrix and using the dried-droplet method. The mass spectrometer was operated in reflector mode and calibrated with angiotensin II, fibrinopeptide, renin substrate, and adrenocorticotropic hormone (18–39 fragment).

HEK293T cells (American Type Culture Collection, CRL-3216) were cultured in DMEM (1×) plus GlutaMAX medium (Life Technologies) supplemented with 10% FBS (Life Technologies), streptomycin (100 μg/ml), and penicillin (100 U/ml) at 37°C in a humidified atmosphere containing 5% CO₂.

Statistical analyses for the luciferase reporter assay and quantification of fluorescence were carried out by Microsoft Excel and Prism, respectively. All of the data are presented as mean ± SEM. Two-way ANOVA with a Tukey multiple comparisons test was used to compare different groups. The statistical significance between the indicated samples and the control is designated as *p < 0.05, **p < 0.01, ***p < 0.001, or NS (p > 0.05).

The atomic coordinates and structural factors of the phosphorylated human and mouse IRF-3/CBP complexes have been deposited in the Worldwide Protein Data Bank, www.wwpdb.org (PDB: 7JFL and 7JFM, respectively).

To investigate how phosphorylation activates IRF-3, we expressed and purified a truncated hIRF-3 (residues 189–398) in complex with a CBP fragment (residues 2065–2111) and phosphorylated the complex with TBK1. MS analysis showed that only one residue of IRF-3 is phosphorylated (Fig. 1A). Next, we crystalized the phIRF-3/CBP and determined the structure at 1.8 Å resolution (Table I). Consistent with the MS analysis, the IRF-3/CBP structure showed that only Ser³⁸⁶ was phosphorylated (Fig. 1B–E). Overall, phIRF-3/CBP forms a dimer through phosphorylated Ser³⁸⁶ and the downstream DLHIS sequence (known as pLxIS motif) (Fig. 1B, 1C). Superposition of phIRF-3/CBP complex and autoinhibited IRF-3 (PDB indentification code: 1QWT) structures reveal a dramatic conformational change of the C-terminal tail upon phosphorylation (Supplemental Fig. 1A, 1B). In autoinhibited IRF-3, the C-terminal tail is folded and blocks the binding of CBP. In the phIRF-3/CBP complex, the C-terminal tail of IRF-3 unfolds and interacts with another IRF-3 molecule through extensive hydrogen bonds, hydrophobic interaction, and electrostatic interaction (Fig. 1E, Supplemental Fig. 1C, 1D, 1F). In addition, the IRF-3 dimer is also stabilized via hydrophobic interactions and hydrogen bonds in the central core region of the dimer (Supplemental Fig. 1E).

The phIRF-3/CBP complex structure reveals that pSer³⁸⁶ reaches into a highly positively charged pocket surrounded by residues Arg²¹¹, Arg³⁸⁰, Arg³⁴¹, and Lys³⁶⁰ and interacts with these residues via electrostatic interactions and an extensive network of hydrogen bonds (Fig. 1E). Specifically, pSer³⁸⁶ interacts with Arg²¹¹ from another IRF-3 molecule through a network of three hydrogen bonds. In addition, Arg²¹¹ also interacts with Lys³⁶⁰ and Glu³⁸⁸ through hydrogen bonds, thus making a critical contribution to the formation of IRF-3 dimer. Arg³⁸⁰ forms a hydrogen bond with pSer³⁸⁶ within the same IRF-3 molecule via its side chain guanidinium group. A water molecule forms a network of three hydrogen bonds with Arg³⁸⁰, pSer³⁸⁶, and Asp²⁵⁴, making additional contributions to the interactions between pSer³⁸⁶ and Arg³⁸⁰. In addition, Asp²⁵⁴ of the other IRF-3 molecule interacts with Ser³⁸⁵ upstream of pSer³⁸⁶ via two hydrogen bonds. Arg³⁴¹ is within 4.0 Å from pSer³⁸⁶ and interacts with pSer³⁸⁶ through electrostatic interactions. Moreover, Arg³⁴¹ also interacts with the phosphate group of pSer³⁸⁶ via a solvent-mediated hydrogen bond. Similarly, the sidechain of Lys³⁶⁰ from the other IRF-3 in the dimer is within 4.0 Å from the phosphate group and interacts with pSer³⁸⁶ via electrostatic interaction and a solvent-mediated hydrogen bond. In addition, Ser³³⁹ forms a hydrogen bond with pSer³⁸⁶ through its side chain hydroxyl group. Ser³³⁹ also interacts with the phosphate group of pSer³⁸⁶ via a solvent-mediated hydrogen bond through its main chain amine group. Based on these structural analyses, it is obvious that Arg³⁸⁰, Arg²¹¹, Ser³³⁹, and Asp²⁵⁴ play key roles in promoting IRF-3 dimerization by interacting with pSer³⁸⁶ through electrostatic interactions and a network of hydrogen bonds (Fig. 1E). Strikingly, every polar atom of pSer³⁸⁶ contributes to one or more hydrogen bonds in the phosphorylated IRF-3 dimer.

Although the overall structure of the phosphomimetic S386/396E IRF-3/CBP dimer is similar to the phIRF-3/CBP dimer (Supplemental Fig. 2A), Glu³⁸⁶, which mimics pSer³⁸⁶, contributes much less significantly to IRF-3 dimerization (Fig. 1F). Similar to the phIRF-3/CBP dimer, Arg³⁸⁰ is <4 Å from Glu³⁸⁶, and they likely interact with each other via electrostatic interactions. By contrast, the closest distance between the side chains of Arg²¹¹ and Glu³⁸⁶ is over 5 Å, and Arg²¹¹ forms no hydrogen bonds directly with Glu³⁸⁶. Instead, Arg²¹¹ stabilizes the phosphomimetic IRF-3 dimer mainly through its interaction with the side chain of Glu³⁸⁸. In addition, Arg³⁴¹ and Lys³⁶⁰ are farther away from Glu³⁸⁶ and do not interact with Glu³⁸⁶ directly. Moreover, Lys³⁶⁰, Gln³⁵⁶, and Ser³⁵¹ are not involved in the interaction with any residues downstream of Glu³⁸⁶ in the phosphomimetic dimer (Supplemental Fig. 2B). Based on this structural comparison, the phosphomimetic dimer does not fully recapitulate the extensive intermolecular interactions observed in the phIRF-3/CBP dimer, explaining why the phosphorylated IRF-3 dimer is more stable compared with the phosphomimetic dimer.

In addition to Ser³⁸⁶, Ser³⁹⁶ within the pLxIS motif is also involved in IRF-3 activation and can be phosphorylated by TBK1 (6, 21, 23, 24, 27–29). Western blot showed that both Ser³⁸⁶ and Ser³⁹⁶ are phosphorylated upon cGAMP stimulation in HEK293T cells transfected with WT IRF-3 (Supplemental Fig. 2C). In the phIRF-3/CBP complex structure, the electron density for Ser³⁹⁶ and surrounding residues was well defined. However, we did not observe the phosphorylation of Ser³⁹⁶ in the structure, likely because of the truncation at residue Ser³⁹⁸ that prevents the phosphorylation of Ser³⁹⁶ by TBK1. To investigate how phosphorylation of Ser³⁹⁶ contributes to IRF-3 activation, we superimposed the structure of pSTING/IRF-3 (PDB identification code: 5JEJ) complex over the phIRF-3/CBP complex structure (Fig. 1G). The pLxIS motif of the phosphorylated STING is well aligned with the pIRF-3 pLxIS motif (Fig. 1G). Thus, it is likely that phosphorylated Ser³⁹⁶ could reach into the positively charged pocket surrounded by Arg²⁸⁵, His²⁸⁸, His²⁹⁰, and Lys³¹³ and interact with them through electrostatic interactions in a similar fashion as pSer³⁶⁶ of pSTING (Fig. 1G).

To distinguish the roles of Ser³⁸⁶ and Ser³⁹⁶ in IRF-3 activation, we expressed and purified both S386A and S396A mutants of hIRF-3, phosphorylated them by TBK1, and analyzed them by native PAGE. Interestingly, the phosphorylated S386A mutant showed a single lower band, which is indicative of a monomer, whereas the phosphorylated S396A mutant exhibited two bands, indicating a mixture of both monomer and dimer (Fig. 1H). Consistent with the native gel result, size-exclusion chromatography (SEC) showed that phosphorylated S386A mutant was eluted at the same position as unphosphorylated S386A and WT IRF-3, whereas the phosphorylated S396A mutant showed two peaks, which correspond to a mixture of IRF-3 monomer and dimer (Fig. 1I, 1J). These results demonstrate that both Ser³⁸⁶ and Ser³⁹⁶ are involved in IRF-3 dimerization but Ser³⁸⁶ plays a more important role in IRF-3 activation compared with Ser³⁹⁶. To further explore how these two residues affect IRF-3–mediated signaling, we conducted IFN-β luciferase reporter assays in cells transfected with STING and IRF-3. We observed that the S386A mutation blocked the IFN-β reporter activation and the S396A mutation reduced the reporter signal by ∼50%, demonstrating that both Ser³⁸⁶ and Ser³⁹⁶ are involved in IRF–mediated signaling but that Ser³⁸⁶ is more crucial (Fig. 1K). Altogether, these extensive structural and functional studies provide critical insights into the detailed mechanism of IRF-3 activation upon phosphorylation.

Based on the structure of the phIRF-3/CBP complex, the C-terminal tail of IRF-3 mediates the IRF-3 dimerization upon phosphorylation. We wondered whether a similar mechanism is involved in IRF-3 activation in other mammalian species. First, we aligned the C-terminal sequences of IRF-3 across different species and observed that Ser³⁸⁶ and the pLxIS motif containing Ser³⁹⁶ of hIRF-3 are highly conserved in other species (Fig. 2A). However, Lys³⁸¹ of mIRF-3 and Lys³⁸³ of rat IRF-3 replace Glu³⁸⁸ and Asn³⁸⁹ of hIRF-3 (Fig. 2A). To elucidate the mechanism of mIRF-3 activation, we phosphorylated the mIRF-3 (residues 184–390)/CBP complex using TBK1 and determined the crystal structure of the phosphorylated mIRF-3/CBP (pmIRF-3/CBP) complex (Supplemental Fig. 3A, Table I). Similar to hIRF-3, MS analysis showed that only one residue of mIRF-3 is phosphorylated (Supplemental Fig. 3B). The overall structures of the pmIRF-3/CBP and phIRF-3/CBP complexes are similar (root mean square deviation 1.3 Å, Fig. 2B). We observed that Ser³⁷⁹, which corresponds to Ser³⁸⁶ of hIRF-3, is phosphorylated in the structure (Fig. 2B, 2C). Because of the replacement of Glu³⁸⁸ and Asn³⁸⁹ of hIRF-3 by Lys³⁸¹, this region of pmIRF-3 is restructured. Unlike Glu³⁸⁸ in human IRF-3, which contributes to hIRF-3 activation by interacting with Arg²¹¹, Lys³⁶⁰, and Gln³⁵⁶, Lys³⁸¹ of mIRF-3 flips into the solvent and does not interact with any residues nearby (Fig. 2C). However, structures of pSer³⁷⁹ and the pLxIS motif downstream are well preserved. Arg³⁷³ is structurally conserved and interacts with pSer³⁷⁹ in a similar fashion as Arg³⁸⁰ in phIRF-3 (Fig. 2C). The side chain of Arg²⁰⁵ adopts a slightly different conformation compared with Arg²¹¹ of phIRF-3 and interacts with pSer³⁷⁹ through electrostatic interaction and hydrogen bonds (Fig. 2C). The interactions between Asp²⁴⁷, pSer³⁷⁹, and Ser³⁷⁸ are also well preserved in both mouse and hIRF-3 (Fig. 2C). By contrast, the side chains of Arg³³⁴ and Lys³⁵³ move away from pSer³⁷⁹ and do not interact with pSer³⁷⁹ directly (Fig. 2C). These structural comparisons demonstrate that Arg³⁷³ and Arg²⁰⁵ of mIRF-3, which correspond to Arg²¹¹ and Arg³⁸⁰ of hIRF-3, are critical for the interactions with pSer³⁷⁹ and contribute to the dimerization of mIRF-3 upon phosphorylation.

In contrast to mIRF-3, Glu³⁸⁸ of hIRF-3 interacts with Arg²¹¹, Lys³⁶⁰, and Gln³⁵⁶ through electrostatic interactions and the solvent-mediated hydrogen bond and likely plays additional roles in hIRF-3 activation (Fig. 2C). Indeed, the luciferase reporter assay showed that mutating Glu³⁸⁸ to alanine in hIRF-3 reduced the IFN-β reporter signal by ∼45% (Fig. 2D). Furthermore, the phosphorylated E388A mutant showed two bands on native gel compared with phosphorylated WT IRF-3 (Fig. 2E). In agreement with these results, SEC shows that the E388A mutant of phIRF-3 elutes as two peaks, indicating the dimerization of hIRF-3 was compromised by this mutation (Fig. 2F). Taken together, these structural analyses reveal that mIRF-3 is activated in a similar but distinct manner compared with hIRF-3.

To investigate how the residues interacting with pSer³⁸⁶ and pSer³⁹⁶ contribute to IRF-3 activation, we generated 12 mutants of full-length hIRF-3 that include mutations R211A, R380A, S339A, R211A/R380A, and R211A/R380A/S339A of residues interacting with pSer³⁸⁶, and mutations R285A, H288A, H290A, K313A, R285A/K313A, H290A/K313A, and H288A/H290A/K313A of residues that are likely to interact with pSer³⁹⁶. Each of these mutants was expressed and purified for in vitro phosphorylation (Supplemental Fig. 3C, 3D). MS analyses showed that WT full-length IRF-3 can be efficiently phosphorylated by TBK1 (Fig. 3A). Each of the IRF-3 mutants was phosphorylated and analyzed by SEC (Fig. 3B–N). Strikingly, the R211A, R211A/R380A, and R211A/R380A/S339A mutants failed to form dimers upon phosphorylation compared with the WT control (Fig. 3B, 3C, 3F, 3G). In addition, mutation R380A severely impaired the dimerization of IRF-3 upon phosphorylation (Fig. 3D). By contrast, mutations S339A, R285A, R285A/K313A, H290A/K313A, and H288A/H290A/K313A have moderate effects on IRF-3 dimer formation (Fig. 3E, 3H, 3L–N). Moreover, mutations H288A, H290A, and K313A have little effect on IRF-3 dimerization (Fig. 3I–K). To further investigate the effect of these mutations on IRF-3 activation, we analyzed these IRF-3 mutants by native gel electrophoresis (Fig. 3O, 3P). We observed that phosphorylated R211A, R211A/R380A, and R211A/R380A/S339A mutants only showed a single band similar to unphosphorylated IRF-3, suggesting the activation of IRF-3 was disrupted by these mutations. Phosphorylated R380A mutant appeared as two bands, and only a small fraction of this mutant formed dimers, indicating that the activation of IRF-3 was dramatically impaired by this mutation. By contrast, mutations S339A, R285A, R285A/K313A, H290A/K313A, and H288A/H290A/K313A moderately affected the dimerization of IRF-3 upon phosphorylation. Three other mutations H288A, H290A, and K313A, do not individually affect the dimerization of IRF-3 (Fig. 3O, 3P). These results are consistent with the SEC analyses of phosphorylated IRF-3 mutants. Taken together, these studies show that mutations of key residues interacting with pSer³⁸⁶ significantly impair the activation of IRF-3. By contrast, mutations of residues that interact with pSer³⁹⁶ have moderate effects on IRF-3 activation, demonstrating that phosphorylation of Ser³⁸⁶ is more critical for the activation of IRF-3.

It has been reported that IRF-3 dimerizes in the cytosol upon phosphorylation and then translocates to the nucleus to initiate the transcription of the IFN-β gene (14, 30–33). To investigate how IRF-3–mediated signaling is affected by mutations of residues involved in IRF-3 dimerization, we conducted IFN-β luciferase reporter assay in cells transfected with full-length IRF-3 mutants. As is shown in Fig. 4A, mutations R211A, R380A, R211A/R380A, and R211A/K380A/S339A blocked the induction of the IFN-β reporter, whereas mutation S339A reduced the IFN-β reporter signal by ∼50%. As controls, the cells transfected with pcDNA3.1(−) or WT IRF-3 plasmid in the absence of STING plasmid showed almost no signals. The cells cotransfected with pcDNA3.1(−) and STING plasmids also showed very little signal (Fig. 4A). Western blot showed that the expression level of the IRF-3 mutants was similar, whereas only a very faint band was observed in the control vector, which corresponds to endogenous IRF-3 (Supplemental Fig. 3E). Mutations of residues surrounding pSer³⁹⁶, such as mutations R285A, R285A/K313A, H290A/K313A, and H288A/H290A/K313A reduced the reporter signal by 60–70%. By contrast, mutations H288A, H290A, and K313A, individually, barely impacted the activation of the reporter (Fig. 4B). Similarly, Western blot of cells transfected with WT IRF-3 and mutants indicated that these IRF-3 mutants are expressed at similar levels (Supplemental Fig. 3F). These results demonstrate that the mutations of key residues interacting with pSer³⁸⁶ abrogated IFN-β reporter activation, whereas the mutations of residues interacting with pSer³⁹⁶ only partially inhibited IFN-β reporter activation, suggesting that residues mediating IRF-3 dimerization are critical in IRF-3–mediated signaling.

To investigate how these mutations affect the subcellular localization of IRF-3, we conducted confocal microscopy analyses of cells transfected with the IRF-3 mutants (Fig. 4C, Supplemental Fig. 4A). We observed that WT IRF-3 efficiently translocated to the nuclei after cGAMP treatment, whereas the R211A/R380A, R211A/R380A/S339A, R211A, and R380A mutants are mostly localized in the cytosol. The S339A, R285A, R285A/K313A, H290A/K313A, and H288A/H290A/K313A mutants partially entered the nuclei. By contrast, the H288A, H290A, and K313A mutants behaved similar to WT IRF-3 (Fig. 4C, Supplemental Fig. 4A). We used ImageJ to quantify the amount of nuclear translocation of IRF-3 in the cells and observed that mutations R211A, R380A, R211A/R380A, and R211A/R380A/S339A greatly impaired the amount of IRF-3 translocated to the nucleus after cGAMP treatment, whereas mutations H288A, H290A, and K313A did not individually affect IRF-3 nuclear translocation (Supplemental Fig. 4B). These results demonstrate that the mutations that affect IRF-3 dimerization impair IRF-3–mediated signaling and nuclear translocation of IRF-3.

IRF-3 is the key transcription factor regulating the expression of IFN-Is in response to various pathogens. In this study, we determined the crystal structures of phosphorylated human and mIRF-3/CBP complexes, which reveal that IRF-3 forms a dimer upon phosphorylation. Compared with autoinhibited IRF-3, the C-terminal tail of IRF-3 undergoes a dramatic conformational change upon phosphorylation, extending to the binding surface on another IRF-3 molecule, and mediates the dimerization of IRF-3. Phosphorylated Ser³⁸⁶ interacts with several residues in a positively charged pocket through extensive electrostatic interaction and hydrogen bonds. Cell-based studies combined with in vitro phosphorylation assays demonstrate that mutations of Ser³⁸⁶ and the residues surrounding pSer³⁸⁶ abrogate IRF-3 dimerization, block its translocation to the nuclei and abolish IRF-3–mediated signaling. By contrast, phosphorylation of Ser³⁹⁶ within the pLxIS motif likely plays a moderate role in IRF-3 activation. Mutations of Ser³⁹⁶ or residues that may interact with pSer³⁹⁶ only partially impair IRF-3 activation and signaling. Moreover, the structural analyses reveal that Glu³⁸⁸ plays additional roles in the activation of hIRF-3. These structural and functional studies established the molecular basis of IRF-3 activation upon phosphorylation.

In previous studies, Ser³⁸⁶ and the adjacent Ser³⁸⁵ (Ser³⁷⁸ and Ser³⁷⁹ in mouse) were considered as two important phosphorylation sites and were identified as critical residues for IRF-3–mediated signaling (18, 19, 34). However, the structures of the human and mIRF-3 dimer clearly show that only Ser³⁸⁶ is phosphorylated. The phIRF-3/CBP structure reveals that Ser³⁸⁵ interacts with Asp²⁵⁴ via hydrogen bonds to stabilize the IRF-3 dimer. Phosphorylated Ser³⁸⁶ reaches into a large positively charged pocket formed between two molecules of IRF-3 and contributes significantly to IRF-3 dimerization. Mutations of Ser³⁸⁶ and residues interacting with pSer³⁸⁶ dramatically impair IRF-3 dimerization, nuclear translocation, and signaling. These results suggest that phosphorylation of Ser³⁸⁶ but not Ser³⁸⁵ is essential for IRF-3–mediated signaling. Consistent with these results, two previous studies showed that mutation of Ser³⁸⁵ to aspartic acid impairs IRF-3 dimerization, and no phosphorylated Ser³⁸⁵ was detected with specific Abs (20, 24).

Although Ser³⁹⁶ (Ser³⁸⁸ in mIRF-3) within the pLxIS motif is not phosphorylated in the human and mIRF-3 dimer structures presented in this study, these residues can be phosphorylated in vivo (Supplemental Fig. 2C). Mutations of Ser³⁹⁶ and key residues that may interact with pSer³⁹⁶ impair IRF-3 dimerization and its functions, suggesting that phosphorylation of Ser³⁹⁶ also plays an important role in IRF-3 activation. Our previous studies demonstrated that the adaptors STING, MAVS, and TRIF employ the conserved pLxIS motif to recruit IRF-3 upon phosphorylation (6, 12). Interestingly, the phosphorylated pLxIS motifs from the adaptors or IRF-3 itself bind to the same surface on IRF-3, suggesting that the pLxIS motif of IRF-3 itself has two functions. First, the pLxIS motif of pIRF-3 occupies the pLxIS motif-binding surface of another IRF-3 molecule, displacing the adaptors and allowing pIRF-3 to dissociate from the adaptors. Second, the phosphorylated pLxIS motif works together with the phosphorylated Ser³⁸⁶ to facilitate IRF-3 dimerization, thus promoting IRF-3 activation.

The structures of the phosphorylated IRF-3 dimer and the adaptors bound to IRF-3 reveal that Arg²⁸⁵ (Arg²⁷⁸ in mouse) is a key residue interacting with the phosphorylated serine residue within the pLxIS motif. Mutation of Arg²⁸⁵ to alanine moderately impairs IRF-3 activation. These results explain why the R285Q mutation impaired IFN responses to HSV-1 infection and the R285E mutation in either HEK293 cells or fibroblasts of IRF-3–deficient mice showed much less activity upon Newcastle disease virus infection compared with WT IRF-3 (35, 36). Besides Ser³⁸⁶ and the pLxIS motif, Glu³⁸⁸ of hIRF-3 also plays a role in hIRF-3 activation. SEC and cell-based studies confirm the contribution of Glu³⁸⁸ in IRF-3 dimerization and IRF-3–mediated signaling. Interestingly, Glu³⁸⁸ is highly conserved in various species, except for mouse and rat. The replacement of the two residues Glu³⁸⁸ and Asn³⁸⁹ in human by Lys³⁸¹ in mIRF-3 restructures this region and likely reduces the intermolecular interactions of mIRF-3 upon phosphorylation that may affect the kinetics of IFN-I induction by mice. In summary, these extensive structural and functional studies provide critical insights into the molecular basis of IRF-3 activation upon phosphorylation.

We thank The Berkeley Center for Structural Biology for helping to collect x-ray diffraction data. We would also like to acknowledge the members of the Li laboratory for help with revising the manuscript and critical feedback.

The authors have no financial conflicts of interest.