The structure of irisin reveals a novel intersubunit β-sheet fibronectin type III (FNIII) dimer: implications for receptor activation

Abstract

Irisin was recently identified as a putative myokine that is induced by exercise. Studies suggest that it is produced by cleavage of the FNDC5 (fibronectin domain-containing protein 5) receptor; irisin corresponds to the extracellular receptor ectodomain. Data suggesting that irisin stimulates white-to-brown fat conversion have led to the hypothesis that it does so by binding an unknown receptor, thus functioning as a myokine. As brown fat promotes energy dissipation, myokines that elicit the transformation of white to brown fat have potentially profound benefits in the treatment of obesity and metabolic disorders. Understanding the molecular basis for such exercise-induced phenomena is thus of considerable interest. Moreover, FNDC5-like receptors are highly conserved and have been shown to be critical for neuronal development. However, the structural and molecular mechanisms utilized by these proteins are currently unknown. Here, we describe the crystal structure and biochemical characterization of the FNDC5 ectodomain, corresponding to the irisin myokine. The 2.28 Å structure shows that irisin consists of an N-terminal fibronectin III (FNIII)-like domain attached to a flexible C-terminal tail. Strikingly, the FNIII-like domain forms a continuous intersubunit β-sheet dimer, previously unobserved for any FNIII protein. Biochemical data confirm that irisin is a dimer and that dimerization is unaffected by glycosylation. This finding suggests a possible mechanism for receptor activation by the irisin domain as a preformed myokine dimer ligand or as a paracrine or autocrine dimerization module on FNDC5-like receptors.

Keywords: Exercise; Hormones; Muscle; Receptors; X-ray Crystallography.

FIGURE 1.

Crystal structure of irisin. A, irisin is a proteolytic product of FNDC5. The domain organization of the FNDC5 receptor is shown (upper). It contains an N-terminal signal sequence, which provides proper membrane insertion of the receptor and is subsequently cleaved. This is followed by the irisin domain, which contains an N-terminal FNIII-like region and a flexible C-terminal tail. The irisin domain is connected to a short transmembrane region, which is followed by the cytosolic region. The irisin domain is putatively produced following proteolytic cleavage of mature FNDC5 (with the signal sequence removed) (9). The FNDC5 schematic is the irisin subunit structure, showing only the FNIII domain and a topology diagram (lower). B, superimpositions of eight subunits in the crystallographic asymmetric unit showing the regions of flexibility that are found on the same face and that may be candidates for protein-protein interaction sites. Also indicated are the FNIII domain and the C-terminal tail, which is observed in two subunits. C, structure of the irisin dimer. Figs. 1 (B and C), 2A, 3 (A and B), and 4A were made using PyMOL (35).

FIGURE 2.

Irisin dimer contacts and mutagenesis experiments. A, ribbon diagram showing key cross-strand-specific salt bridges that fasten the ends of the irisin intersubunit β-sheet dimer and the location of Ile-77, which is positioned in the center of the dimer interface and was selected as a site for mutagenesis to disrupt the dimer. Also shown are the locations of the asparagine residues (magenta). Asterisks denote the two asparagines contained within NXT motifs and modified by glycosylation. The locations of these residues are notably surface-exposed and in positions in which modification would not be predicted to hinder dimerization. The magnified images are of the locations (modeling) where mutations were made to disrupt the dimer (R75E and I77W). The R75E mutation resulted in a clash with the cross-strand Glu-79, whereas the I77W mutation was predicted to prevent the formation of the hydrophobic interface in the dimer as well as disrupt the Arg-75–Glu-79 salt bridge due to its large size. B, size exclusion chromatography experiments showed that bacterially expressed (non-glycosylated) and glycosylated irisin proteins are dimers, whereas the R75E mutation is monomeric. The I77W mutant was unstable and could not be produced in soluble form. The y axis is the elution volume normalized for column volume, and the x axis is the log of the molecular weight (MW).

FIGURE 3.

Superimposition of irisin (red) with the third FNIII domain of tenascin (TNfnIII3, cyan) and the 10th FNIII domain of fibronectin (fnIII10, yellow). A, overlay of TNfnIII3 onto irisin showing the highly twisted nature of the C′ strand of TNfnIII3 at the dimer interface, making dimer formation by this domain impossible. B, overlay of fnIII10 onto irisin highlighting that not only does fnIII10 have a highly twisted and bulged structure but that it also contains two prolines that prevent optimal dimeric β-sheet H-bonding potential.

FIGURE 4.

Speculative models for irisin signaling. A, location of the flexible loop regions in the irisin dimer. Left, the irisin dimer, with the N terminus colored salmon, the flexible region of residues 55–58 colored red, and the flexible loop at residues 106–108 colored magenta. Right, electrostatic surface representation of the irisin dimer (blue and red indicate electropositive and electronegative regions, respectively) (upper) and the dimer rotated by 90° showing the hydrophobic face (white) of the loop-containing regions (lower). B, model for irisin functioning as a myokine cleaved from FNDC5. This model shows how a preformed irisin dimer acting as a myokine could facilitate dimerization and activation of an as yet unidentified receptor, leading to signaling events that stimulate white-to-brown fat conversion. C, model for FNDC5 receptor function. Shown are possible modes of signaling and cell adhesion processes affected by dimerization of the extracellular irisin domain of the FNDC5 receptor. Left, scenario in which FNDC5 molecules exist in the same cell. Here, irisin domains may aid in dimerization of the receptors, leading to signaling events, or a dimerized receptor may subsequently bind a ligand that would induce structural changes within the dimer, leading to downstream signaling. Right, FNDC5 molecules on proximally located cells could dimerize either to initiate signaling programs or to facilitate cell adhesion processes.