Fig 2: Tspan12 binds directly to Norrin with nanomolar affinity via the large extracellular loop (LEL).(A) Schematic of biolayer interferometry (BLI) setup for Tspan12-Norrin binding: Tspan12 lacking the C-terminal tail (∆C), inserted into biotinylated nanodiscs, is immobilized on a streptavidin-coated biosensor, and Norrin association and dissociation are monitored in real time. (B) BLI traces of Norrin at indicated concentrations binding to and dissociating from Tspan12. (C) Steady-state binding curve fit to Norrin-Tspan12 binding (mean ± SD from three independent replicates at each concentration of Norrin) gives a KD of 10.4±1.2 nM (mean ± SEM). (D) Observed association rate constant (Kobs), determined from fitting BLI association traces (mean ± SD in three independent experiments), is linearly dependent on Norrin concentration with a slope Kon = 0.00019 ± 0.00003 nM –1 s–1 (mean ± SEM). When combined with the Koff = 0.0014 ± 0.00016 s–1 (mean ± SEM) determined from fitting the dissociation traces, we obtain a kinetic KD of 7.4±1.4 nM (mean ± SEM). (E) BLI traces of the soluble MBP-tagged Tspan12 LEL domain, at the indicated concentrations, associating to and dissociating from a biosensor loaded with MBP-tagged Norrin. Kinetic fitting gives an apparent affinity of 16±3 nM (mean ± SEM). (F) BLI traces of 10, 32, or 100 nM Norrin show no binding to a biosensor loaded with a nanodisc-embedded chimeric Tspan12 with the LEL replaced by that of Tspan11. Figure 1—source data 1.Steady-state interference shift and Kobs values used to generate Figure 1C and D.

Fig 3: Norrin-Tspan12 binding is competitive with heparan sulfate proteoglycans but compatible with Fzd4 binding.(A) The AlphaFold-predicted structure of Tspan12 (purple) bound to a Norrin dimer (yellow/orange), aligned to the crystal structure (5BQC) of a Norrin dimer bound to the Fzd4 CRD (blue) and sucrose octasulfate (SOS) (red). A zoomed view of the indicated region shows an overlap in the predicted binding site of Tspan12 with that of SOS, suggesting that Tspan12 and SOS cannot bind simultaneously to Norrin. (B) Displacement of Norrin (32 nM) from immobilized Tspan12 by increasing concentrations of SOS, as measured by biolayer interferometry (BLI) (see Figure 3—figure supplement 1A). The equilibrium binding signal is plotted as a percent of signal in the absence of SOS (mean ± SD of three independent experiments), yielding a Ki of 34±4 µM. (C) Side view of structures in A. A zoomed view of the indicated region shows that Tspan12 is predicted to occupy a site on Norrin adjacent to, but not overlapping with, the Fzd4 binding site; adjacent residues are shown. Norrin from the Tspan12-bound AlphaFold model (yellow) and the Fzd4 CRD-bound crystal structure (5BQC, gray) are overlaid. (D) Fzd4 CRDL does not fully compete with Tspan12-Norrin binding, as shown by equilibrium binding of 32 nM Norrin to Tspan12 immobilized on paramagnetic particles in the presence of increasing concentrations of purified Fzd4 CRDL. Bound Norrin and Norrin in the supernatant were both quantified by western blot (anti-Rho1D4; see Figure 3—figure supplement 1B) and used to calculate bound Norrin as a percentage of total Norrin. The expected competition curve, assuming fully competitive binding sites, was simulated (gray dashed line) given starting concentrations of 50 nM Tspan12 and 32 nM Norrin, and binding affinities of 10.4 nM for Tspan12-Norrin and 200 nM for Fzd4 CRDL-Norrin. However, the data better fit a model in which CRDL binding to Norrin shifts Norrin affinity for Tspan12 (blue line). Data represent mean ± SD of three replicates. (E) BLI traces of a ternary Fzd4-Norrin-Tspan12 large extracellular loop (LEL) complex. Biosensors loaded with nanodisc-embedded Fzd4 were first saturated with 100 nM Norrin, then bound to 32, 100, or 320 nM Tspan12 LEL. (F) BLI traces of ternary complex formation. Biosensors loaded with maltose binding protein (MBP)-tagged Norrin were pre-incubated in buffer or saturated with Fzd4 CRDL (5 µM), then bound to 100 nM MBP-tagged Tspan12 LEL (±5 µM CRDL). Tspan12 LEL did bind to Norrin in the presence of the Fzd4 CRDL (dark purple; apparent KD = 27 ± 2.8 nM), albeit more weakly than it bound to Norrin alone (light purple; apparent KD = 16 ± 1.8 nM; see also Figure 1E). Binding affinities were obtained from kinetic fits (black dotted line) to association and dissociation traces of MBP-LEL (100 nM) from three independent experiments. Figure 3—source data 1.Interference shift and band quantification values used to generate Figure 3B and D.

Fig 4: Tspan12 does not directly enhance formation of a Norrin-LRP5/6-Fzd4-Dvl signaling complex.(A) Hypothesis: Tspan12 could enhance Norrin signaling by enhancing interactions within the Norrin-LRP5/6-Fzd4-Dvl complex, including Fzd-Dvl binding and Norrin-LRP binding. (B) Representative biolayer interferometry (BLI) traces of the Dvl2 DEP domain associating to and dissociating from Fzd4 in nanodiscs containing 75:20:5 POPC:Ccholesterol:PIP2. (C) Equilibrium binding of the Dvl2 DEP domain to Fzd4 monomer or Tspan12/Fzd4 heterodimer in nanodiscs; affinities ± SEM are 183±24 and 279±46 nM, respectively. (D) Equilibrium binding of the Dvl2 DEP domain to Fzd4 monomer or Tspan12/Fzd4 heterodimer nanodiscs, each pre-saturated with 10 nM Norrin. Binding affinities are 161±21 and 274±39 nM (mean ± SEM), respectively, determined from three independent replicates. Affinities and kinetic constants are reported in Supplementary file 1. (E) The LRP6 E1E2 domain fully competes with Tspan12-Norrin binding, as shown by decreased equilibrium binding of 32 nM Norrin to Tspan12 immobilized on paramagnetic particles in the presence of increasing concentrations of purified LRP6 E1E2 domain. Norrin was quantified by western blot (anti-Rho1D4; see Figure 5—figure supplement 1) and plotted as a percent of bound Norrin in the absence of LRP6 E1E2. The curve was fit to a competitive binding model using known binding affinities of 10.4 nM for Tspan12-Norrin and starting concentrations of 50 nM Tspan12 and 32 nM Norrin; the best fit reported a Norrin-LRP6 E1E2 binding affinity of 1.06 µM (95% CI 0.747–1.51 µM). Data represent mean ± SD of three replicates. (F) β-Catenin transcriptional activity in response to no ligand, 1 nM recombinant Norrin, or Wnt3a conditioned media (Wnt3a CM) in Fzd1/2/4/5/7/8-knockout HEK293T cells transfected with Tspan12 siRNA or indicated amount of Tspan12_pTT5 plasmid, along with Fzd4 and TopFlash luciferase reporter plasmids. Data are plotted as mean ± SD from n=3 replicate wells. Figure 5—source data 1.Interference shift, band quantification, and luciferase activity values used to generate Figure 5C–F.

Fig 5: Diverse co-receptors facilitate growth factor signaling by capturing and delivering ligands to their target receptors.(A) Model: Norrin is captured by Tspan12 or heparan sulfate proteoglycans (HSPGs) and is handed off to Fzd4 for association with LRP5/6 and subsequent signaling. Norrin binding to cell-surface Fzd4 is enhanced when HSPGs concentrate Norrin at the cell surface. In contrast, Tspan12 directly and specifically delivers Norrin to co-localized Fzd4. (B) In the β-catenin signaling pathway, Left: Tspan12 captures Norrin and co-localizes with Fzd4, delivering Norrin to Fzd4. Middle: Likewise, RECK binds Wnt7a/b and co-localizes with Fzd via GPR124, delivering Wnt7a/b to Fzd. Right: Glypican-3 (GPC3) also binds both Fzd and Wnt to deliver Wnt to Fzd and enhance signaling. (C) Structurally diverse co-receptors play a similar role to Tspan12 in various pathways activated by cystine knot growth factors. Left: Neuropilin-1 (Nrp1) captures vascular endothelial growth factor A (VEGF) and co-localizes with the VEGF receptor 2 (VEGFR2) to specifically deliver VEGF to VEGFR2. Middle: The repulsive guidance molecule (RGM) binds bone morphogenic protein 2 (BMP2) as well as neogenin-1 (Neo1) to facilitate BMP signaling. Right: Betaglycan captures transforming growth factor β1 (TGF-β) and presents it to TGF-β receptor type 2 (TGFβR2).