A lipid-binding domain in Wnt: a case of mistaken identity?

Several groups have recently presented convincing evidence for extensive membrane localisation during signalling between the secreted glycoprotein Wnt and its cell-surface receptor Frizzled ([1] and references therein). Aravind and Koonin [2] extended this by identifying a colipase-like domain in the Wnt antagonist Dickkopf. They suggested that this domain might be necessary for lipophilic interactions leading to membrane association during Wnt antagonism. They identified this colipase domain with PSI-BLAST [3], which found several highly conserved cysteine and other residues shared between the cysteine-rich Dickkopf protein and the colipase. This homology was subsequently backed up with a convincing alignment and a consideration of the known structure of the colipase. Recently, Reichsman et al. [4] have proposed a similar lipid-binding and membranelocalisation role in a region of Wnt proteins. Using BLAST, they identified a putative homology between the Wnt carboxyl terminus and the phospholipase A2 (PLA2) lipid-binding domain. Here, we analyse a diverse set of Wnt homologues and the known structures of PLA2s and argue against homology between Wnts and PLA2s. PLA2s are a diverse superfamily of proteins. Three-dimensional structures are known for at least 15 species [5], and analysis of a structure-based sequence alignment [6] shows much variation across vertebrate and invertebrate PLA2s. Nevertheless, all known PLA2s contain a common core (boxes in Figure 1), comprising a calcium binding loop, two alpha helices and three conserved disulphide bonds (solid lines in Figure 1). The alignment of Reichsman et al. [4] corresponds to approximately half of the PLA2 core structure, and there is insufficient sequence in Wnt Magazine R717