Single amino acid substitutions confer the antiviral activity of the TRAF3 adaptor protein onto TRAF5

Abstract

The TRAF [tumor necrosis factor receptor-associated factor] family of cytoplasmic adaptor proteins link cell-surface receptors to intracellular signaling pathways that regulate innate and adaptive immune responses. In response to activation of RIG-I (retinoic acid-inducible gene I), a component of a pattern recognition receptor that detects viruses, TRAF3 binds to the adaptor protein Cardif [caspase activation and recruitment domain (CARD) adaptor-inducing interferon-β (IFN-β)], leading to induction of type I IFNs. We report the crystal structures of the TRAF domain of TRAF5 and that of TRAF3 bound to a peptide from the TRAF-interacting motif of Cardif. By comparing these structures, we identified two residues located near the Cardif binding pocket in TRAF3 (Tyr(440) and Phe(473)) that potentially contributed to Cardif recognition. In vitro and cellular experiments showed that forms of TRAF5 with mutation of the corresponding residues to those of TRAF3 had TRAF3-like antiviral activity. Our results provide a structural basis for the critical role of TRAF3 in activating RIG-I-mediated IFN production.

Fig. 1

Structure of the TRAF3-Cardif complex. (A) Representation of TRAF3 homotrimers, shown from both side (left) and top (right) views. The Cardif peptide is in lavender. (B) Electron density map of Cardif bound to TRAF3. The blue meshwork represents the 2F_o – F_c electron density map contoured at 1.2. TRAF3 is gray; the Cardif peptide backbone is lavender. Cardif residues ¹⁴³PVQDT¹⁴⁷ are labeled. (C) Different peptides binding to TRAF3. The surface of TRAF3 is gray; peptides shown are from Cardif (lavender), CD40 (green), TANK (cyan), LTβR (yellow), LMP1 (blue), and BAFF-R (orange). (D) Molecular TRAF3-Cardif interactions. Key TRAF3-Cardif interactions are shown by yellow dotted lines. Conserved hotspot residues—the hydrophobic patch, serine fingers, and polar residues—are in gray, cyan, and blue, respectively. For clarity, only one of the alternative confirmations of Cardif is shown in (A) and (D).

Fig. 2

Structural comparison of TRAF5 and TRAF3. (A) Representation of TRAF5 homotrimers. (B) Sequence alignment of the TRAF domains of TRAF3 and TRAF5. The hotspot residues in the hydrophobic patch, serine fingers, and polar residues are labeled as stars, dots, and squares, respectively. The amino acids that play crucial roles in “switching” TRAF5 to TRAF3 are highlighted by black arrows. (C) Superposition of TRAF3 and TRAF5 structures. TRAF3 is gray, and TRAF5 is cyan. Conserved hotspot residues, Phe⁴⁷³ and Tyr⁴⁴⁰ in TRAF3 and Tyr⁴⁶² and Phe⁴²⁹ in TRAF5, are shown within the magnified insert.

Fig. 3

Tyr⁴⁴⁰ and Phe⁴⁷³ in TRAF3 are necessary for interaction with Cardif. (A) Immunoblot analysis of interactions between a 13-residue Cardif peptide (residues 138 to 150) and the TRAF domains of wild-type (WT) TRAF3, WT TRAF5, TRAF3 mutants (left), and TRAF5 mutants (right). (B) Lysates from cells expressing FLAG-tagged WT TRAF3, WT TRAF5, TRAF5 Y462F, and TRAF5 F429Y constructs were immunoprecipitated with GST-Cardif TIM and immunoblotted for the FLAG epitope. (C) The presumed interactions (orange dotted line) between TRAF3 (left) and Cardif are absent in TRAF5 (right). Results in (A) and (B) are confirmed by three independent experiments.

Fig. 4

TRAF3 Y440F and F473Y lose antiviral function. (A) Immunoblot of TRAF3−/− MEFs reconstituted with WT TRAF3, WT TRAF5, TRAF3 F440Y, TRAF3 Y473F, or TRAF3 F440Y/Y473F. (B) IFN-β production was assessed in reconstituted TRAF3−/− MEFs infected with VSV-GFP. (C) Viral titer was assessed in reconstituted TRAF3−/− MEFs infected with VSV-GFP (D). VSV-GFP fluorescence was quantified and calculated by the product of the percent GFP-positive cells and the geometric mean fluorescence intensity of GFP. Results represent an average of three independently analyzed biological replicates and were confirmed by three separate experiments. Error bars indicate the SD. Unpaired t test (two-tailed), n = 3. *P < 0.05; **P < 0.01.

Fig. 5

The TRAF5 Y462F and F429Y mutants gain TRAF3-like antiviral function. (A) Immunoblot of TRAF3−/− MEFs that were reconstituted with WT TRAF3, WT TRAF5, TRAF5 F429Y, TRAF5 Y462F, or vector. (B) IFN-α production by Sendai virus–infected reconstituted TRAF3−/− MEFs and IFN-β production by VSV-GFP–infected reconstituted TRAF3−/− MEFs. (C) mRNA was collected at 0, 4, 8, and 12 hours after infection with VSV-GFP and assayed by quantitative polymerase chain reaction for induction of IFN-β, IP10, and MX1. (D and E) Viral titer was assessed in reconstituted TRAF3−/− MEFs infected with VSV-GFP (D). VSV-GFP fluorescence was quantified (E). (F) Single-cell clones were generated from stably reconstituted TRAF3−/− MEFs. GFP fluorescence was determined in cells infected with VSV-GFP. Single-cell clones from cells reconstituted with TRAF5 F429Y (left) and Y462F (right) were tested in separate experiments. Plots represent average GFP fluorescence of three single-cell clones per construct. All results represent an average of three biological replicates and were confirmed by three separate experiments. Error bars indicate the SD. Unpaired t test (two-tailed), n = 3. *P < 0.05; **P < 0.01.