4A Insights‎ > ‎

2012.12.13 How Frizzled domains really work

posted Feb 4, 2013, 1:21 PM by Sirid Kellermann   [ updated Nov 14, 2013, 10:43 AM ]

Fernando Bazan shares the fascinating story of how he and his colleagues at Genentech and Stanford University elucidated the true nature of ligand binding to Frizzled (Fz) and Smoothened (Smo). These findings have had significant and continuing impact on the design of next-generation oncology therapeutics. 

Wnt and Hedgehog (Hh) are families of secreted, lipid-tagged proteins that play critical roles in cell development, tissue patterning, and organ development. The evolutionarily conserved intracellular signaling pathways triggered by these morphogens have captured significant research interest, because (1)malfunctions in these pathways are implicated in many types of cancers; (2) emerging regulatory ties to immune cell activation; and (3) mechanistic links to sensory cilia trafficking and signaling.

Wnts impact cell signaling directly by binding a family of Frizzled (Fz) GPCRs. In contrast, Hhs exert their effects indirectly, by engaging a cell surface transporter called Patched and triggering the release of unknown, possibly sterol-like molecules that act as activating ligands for a distant Fz-like GPCR, Smoothened (Smo). Smo in particular has been the target of  vigorous drug development efforts, although with the more recent appreciation of the role of Wnt signaling in cancer, Fz's are increasingly garnering attention as therapeutic targets for intervention.

As Fernando recalls, “I started looking at these receptors soon after arriving at Genentech. The company had generated small molecule antagonists of Smo that showed efficacy against basal cell carcinoma - in fact, Erivedge was approved by the FDA for treatment of BCC earlier this year. I was part of a team that sought to find a structural explanation for how these drugs worked. I pursued the notion that Smo, besides possessing an obvious binding site in its core GPCR cavity, may have additional binding site(s). Identifying these alternative pockets would help us better explain the enigmatic mechanisms of action of diverse Smo-active drugs in the literature (which ran the gamut from full agonists to full antagonists) and allow us to more intelligently design effective compounds.”

Fz's and Smo both possess cysteine-rich Fz ectodomains. Fernando’s work led to the surprising proposal that the Fz module is endowed with a natural binding cavity (or groove) within a cage-like array of  disulfide-linked helices. Moreover, this unique fold is found in several other protein families, including the NPC2 cholesterol transporter as well as folate receptors. As Fernando describes in the accompanying video, this connection between apparently unrelate proteins is not immediately obvious by visual inspection or comparison using standard computational approaches.

These structural and evolutionary links are intriguing, considering that NPC2 and folate receptors are known to shuttle cholesterol and folate-like molecules, respectively, while the proposed ligands for Fz and Smo are lipidated Wnt proteins and oxysterols, respectively. Taking the analysis further, Fz domains appear to be related to pheromone- and lipid-binding secreted proteins found in insects.

This work led to the prediction that the lipid moiety of Wnts is a critical component of their interactions with Fz's. After Fernando and his colleague, Fred de Sauvage, published this insight in 2009 paper in Cell, it provided a tantalizing clue to solving the structure of Wnts, which have historically been intractable to crystallography since no one could find a way to make significant amounts of pure, stable, and soluble Wnt protein. K. Christopher Garcia’s group at Stanford leveraged Fernando’s prediction by cleverly coexpressing the Fz ectodomain with Wnt, thereby stabilizing the protein and creating sufficient amounts to be crystallized.

“In our Cell paper, we hypothesized that all Fz ectodomains would be found to bind lipids, sterols or folates. This is supported by observations that at least some agonists and antagonists of Smo do not appear to bind in the GPCR’s cavity. Might they be acting allosterically, by interacting with Smo’s Fz domain?”
- Fernando Bazan


The Garcia team published the structure of Wnt in 2012 in Science, revealing how Wnt grips the globular Fz ectodomain at two sites (see video at 2:10), somewhat like a ball being pinched between a thumb and a forefinger – where the tip of the Wnt "thumb" has a lipid moiety that docks in Fz's hydrophobic groove, in accord with Fernando's proposal that this was a lipid-sequestering site. 

Listen to Dr. Garcia tell the story in this Science podcast.

Moreover, the structure of Wnt showed that it possesses a remarkably complex, novel fold. “I realized that this odd Wnt structure was actually a molecular tapestry of several modular folds, each conferring a unique functional attribute to the morphogen,” says Fernando. “There are two key folds, as we outlined in a subsequent 2012 paper, coauthored by Dr. Garcia and his group in Developmental Cell: an N-terminal saposin-like module designed to interact with lipid membranes, fused to a C-terminal domain reminiscent of a degenerate cystine-knot cytokine fold. The former module contains the lipid attachment site, while the latter cytokine-like domain is charged with making a second, protein-protein contact with Fz. That second domain also likely engages the LRP family member that completes the ternary Wnt/Fz/LRP complex. Quite likely, Wnts evolved from the ancient fusion of two genes encoding these smaller protein domains, creating a molecule with composite functions.”

New possibilities in drug development
The prediction that Smo, Fz, and other proteins expressing Fz ectodomains interact with their ligands through lipid binding, in addition to insights about the structure of the Fz ligand, Wnt, open new possibilities for the design of small molecule compounds and biotherapeutics (video, 4:09). The Fz ectodomain’s lipid-binding groove is a natural hotspot for drug binding. In the case of biotherapeutics, deconstructing Wnt into its component parts through protein engineering can help create superbinders that interact extremely tightly with receptors that bear Fz ectodomains, while antibodies directed against the lipid binding groove may be effective in blocking the binding of Wnts.

"The (field) has been ignoring that there's this readily druggable domain that will be an excellent antagonist binding site," Fernando concludes (video, 7:08).

And there’s a third therapeutic strategy, as Fernando points out: “We may be able to specifically target Wnt’s interaction with LRP5/6, the second signaling receptor of the Fz supercomplex, through the interaction site found in Wnt’s cytokine-like subdomain. That’s an area of pretty intensive research activity right now.”