Interplay of the Norrin and Wnt7a/Wnt7b signaling systems in blood–brain barrier and blood–retina barrier development and maintenance

Extensive Redundancy Between Norrin and Wnt7a in BBB Maintenance.

The most direct approach for assessing interactions between the Norrin and Wnt7a/Wnt7b systems in BBB and BRB maintenance involves analyzing the phenotypes associated with loss-of-function mutations in the genes coding for their respective ligands. However, this analysis presents three technical challenges: (i) the likely redundancy between the highly homologous Wnt7a and Wnt7b genes, as suggested by their redundancy in the context of embryonic CNS angiogenesis (9); (ii) the sterility of Wnt7a^−/− males and females; and (iii) the early lethality of Wnt7b^−/− embryos, which die before midgestation from a placental defect, thereby necessitating the use of a conditional Wnt7b allele for postnatal analyses.

To circumvent the third challenge, we began by investigating the effect of eliminating Wnt7a, either alone or in combination with Ndp. (Ndp is X-linked, and we will use “Ndp^KO” to refer both to males with hemizgous loss of Ndp and to females with homozygous loss of Ndp.) Fig. 2 shows sagittal brain sections—the standard format for the postnatal analyses described herein—from representative Wnt7a^−/−, Ndp^KO, Ndp^KO;Wnt7a^+/−, and Ndp^KO;Wnt7a^−/− mice at P30, stained for: (i) GLUT1, an EC marker for the BBB state; (ii) plasmalemma vesicle-associated protein (PLVAP; a structural protein that comprises the diaphragm in EC fenestrae and transcytotic vesicles), an EC marker for the high permeability state; and (iii) covalent Sulfo-NHS-long-chain-biotin (“Sulfo-NHS-biotin”) adducts. Sulfo-NHS-biotin is a marker for leakage of low molecular weight serum components into the CNS parenchyma; it was delivered to each mouse by intraperitoneal injection.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Redundancy between Ndp and Wnt7a in the cerebellum and olfactory bulb. Coronal sections of ∼P30 brains from representative examples of the indicated genotypes, immunostained for GLUT1 and PLVAP. Four-digit mouse identification numbers are in the upper corners of panels in this and other figures. The second and third rows show sections through the cerebellum from two representative brains for each genotype. The fourth row shows a representative region of the cerebellar vasculature at higher magnification. Arrows in the second and third rows indicate the choroid plexus. (Scale bars: 1 mm for rows 1–3; 200 μm for row 4.)

Wnt7a^−/− brains are indistinguishable from WT controls with respect to the distribution of GLUT1, PLVAP, and Sulfo-NHS-biotin. (See SI Appendix, Fig. S1 for a representative WT brain.) In particular, conventional CNS ECs are uniformly GLUT1⁺/PLVAP⁻, and the only locations within the CNS where Sulfo-NHS-biotin accumulates outside of the vasculature is the choroid plexus (arrows in second and third rows of Fig. 2) and the circumventricular organs, small midline structures that are variably present in the sections shown here. Choroid plexus and circumventricular organ ECs are GLUT1⁻;PLVAP⁺ and are highly permeable.

As previously reported (7), Ndp^KO brains show a low level of BBB breakdown in the cerebellum, centered on the occasional ECs that have converted from GLUT1⁺/PLVAP⁻ to GLUT1⁻;PLVAP⁺ (Fig. 2, column 2). Ndp^KO;Wnt7a^+/−and Ndp^KO;Wnt7a^−/− brains show progressively greater extents of BBB breakdown and EC marker conversion in the cerebellum. In Ndp^KO;Wnt7a^−/− brains, this phenotype is also observed in the olfactory bulb. Fig. 3, which quantifies the extent of GLUT1⁺/PLVAP⁻ to GLUT1⁻;PLVAP⁺ conversion of cerebellar ECs for all of the genotypes studied here, confirms the visual impression that, with loss of additional Wnt7a alleles in the Ndp^KO background, there is a step-wise increase in EC conversion from a barrier-competent state to a permeable state.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Quantification of phenotypic conversion of cerebellar ECs from GLUT1⁺/PLVAP⁻ to GLUT1⁻/PLVAP⁺. For each of the indicated genotypes, the fraction of the cerebellar vasculature that was GLUT1⁻/PLVAP⁺ was determined as described in Materials and Methods. Each data point represents a different mouse. Bars indicate mean ± SD. P values for pairwise comparisons were calculated using the two-tailed Student’s t test.

In sum, these data reveal a role for Wnt7a in BBB development and maintenance in the cerebellum and olfactory bulb, and they demonstrate that Wnt7a and Norrin are largely redundant in this context.

Bergmann Glia Are the Principle Source of Wnt7a for Cerebellar BBB Maintenance.

RNA sequencing (RNA-seq) on FACS-sorted cerebral cortical cells shows Wnt7a transcripts in a variety of cell types, and most abundantly in astroglia (12). Based on the expression of an alkaline phosphatase reporter knocked into the Ndp locus, the principal site of Ndp expression in the cerebral cortex is also astroglia, and in the cerebellum it is Bergmann glia (35). To functionally assess whether Bergmann glia are also the principal source of Wnt7a in the cerebellum, we generated mice lacking Ndp in all cells and lacking Wnt7a only in Bergmann glia. In these mice, recombination of a conditional Wnt7a allele by a Gfap-Cre driver [Tg(GFAP-cre)25Mes] (36) was judged to be specific to Bergmann glia because, in separate experiments, the combination of Gfap-Cre and a loxP-stop-loxP reporter at the Rosa26 locus (R26-LSL-mtdT-2A-H2B-GFP) showed colocalization of reporter expression with S100-β, a marker for Bergmann glia, but not with NeuN or Calbindin, markers for granule cells and Purkinje cells, respectively (SI Appendix, Fig. S2). We note that multiple Gfap-Cre lines exist (see www.informatics.jax.org/recombinase/summary?driver=GFAP), and these lines exhibit a range of glial and neuronal expression patterns.

As seen in Fig. 4, Ndp^KO;Wnt7a^CKO/+;Gfap-Cre mice exhibit a mild phenotype of cerebellar BBB breakdown and EC conversion from GLUT1⁺/PLVAP⁻ to GLUT1⁻/PLVAP⁺ that closely resembles the phenotype of Ndp^KO;Wnt7a^+/− mice (compare Fig. 2, column 3 to Fig. 4, column 1; see quantification in Fig. 3). Ndp^KO; Wnt7a^CKO/−;Gfap-Cre mice show a more severe phenotype of cerebellar BBB breakdown and EC conversion from GLUT1⁺/PLVAP⁻ to GLUT1⁻/PLVAP⁺ that closely resembles the phenotype of Ndp^KO;Wnt7a^−/− mice (compare Fig. 2, column 4 to Fig. 4, column 2; see quantification in Fig. 3). The near identity between the phenotypes resulting from constitutive or Bergmann glia-specific elimination of Wnt7a imply that most, and perhaps all, of the Wnt7a that drives BBB maintenance in the cerebellum is derived from Bergmann glia.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

In an Ndp^KO background, conditional knockout of Wnt7a with Gfap-Cre impairs cerebellar BBB maintenance. Coronal sections of ∼P30 brains from representative examples of the indicated genotypes. The second and third rows show sections through the cerebellum from two representative brains for each genotype. The fourth row shows a representative region of the cerebellar vasculature at higher magnification. (Scale bars: 1 mm for rows 1–3; 200 μm for row 4.)

Loss of Fz4 Sensitizes the CNS Vasculature to Loss of Wnt7a.

To further assess the role of Wnt7a in BBB development and maintenance, we asked how the severity of the BBB phenotype changes with simultaneous reductions in Wnt7a and Fz4. Removing one copy each of Fz4 and Wnt7a (Fz4^+/−;Wnt7a^+/−) did not alter BBB integrity (Fig. 5A, Left column). Removing one copy of Fz4 and both copies of Wnt7a (Fz4^+/−;Wnt7a^−/−) produced a very low level of cerebellar EC conversion from GLUT1⁺/PLVAP⁻ to GLUT1⁻/PLVAP⁺ and a correspondingly low level of vascular leakage (Fig. 5A, second and third columns; quantified in Fig. 3). This phenotype was substantially milder than that associated with loss of Fz4 alone (Fig. 5A, fourth column) or Ndp alone (Fig. 2, second column). Removing both copies of Fz4 and one copy of Wnt7a (Fz4^−/−;Wnt7a^+/−) produced a phenotype similar to that observed in Fz4^−/− mice, with vascular leakage largely confined to the cerebellum, olfactory bulb, and interpeduncular nuclei (Fig. 5A, Right column; quantified in Fig. 3).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Loss of Fz4 synergizes with loss of Wnt7a in BBB maintenance. (A) Coronal sections of ∼P30 brains from representative examples of the indicated genotypes, immunostained for GLUT1 and PLVAP. Arrows in the Right two panels of the Top row point to Sulfo-NHS-biotin leakage in the interpeduncular nuclei. The second and third rows show an enlargement of the boxed region from each Top row image encompassing the ventral-most folia of the cerebellum and the adjacent choroid plexus. In the two Fz4^+/−;Wnt7a^−/− panels, arrows in the second and third rows point to regions with mild Sulfo-NHS-biotin leakage and partial conversion of ECs, and boxed regions in the second row are enlarged as Insets and show capillary ECs that have converted from GLUT1⁺/PLVAP⁻ to GLUT1⁻/PLVAP⁺. (Scale bars: 1 mm in row 1; 500 μm in rows 2 and 3. Insets in row 2 are enlarged 3.7×.) (B) Coronal sections of E17 WT or Fz4^−/−;Wnt7a^−/− brains, immunostained for CLDN5, GLUT1, and PLVAP. (Scale bars, 500 μm.)

Eliminating both Fz4 and Wnt7a resulted in perinatal lethality, with no viable Fz4^−/−;Wnt7a^−/− progeny among 175 that were genotyped from a Fz4^+/−;Wnt7a^+/− intercross in which one-sixteenth should have been Fz4^−/−;Wnt7a^−/− (P = 0.0022; two-tailed χ² test). Among embryonic day (E)17 embryos from the Fz4^+/−;Wnt7a^+/− intercross, Fz4^−/−;Wnt7a^−/− embryos were distinctive in exhibiting intracranial bleeding, a phenotype that had been seen previously in association with severe defects in CNS angiogenesis (e.g., ref. 21). Fz4^−/−;Wnt7a^−/− brains showed bleeding in multiple locations, and these locations were associated with mild-to-moderate defects in vascularization, as seen in SI Appendix, Fig. S3 for the tectum. Overall, CNS vascularization was sparser in Fz4^−/−;Wnt7a^−/− brains (Fig. 5B). In E17 Fz4^−/−;Wnt7a^−/− brains, PLVAP is present in many ECs in the olfactory bulb and striatum, and lower but still detectable levels of PLVAP are present in ECs in the cerebral cortex. Additionally, in E17 Fz4^−/−;Wnt7a^−/− brains, CLDN5 and GLUT1 were only partially anticorrelated with PLVAP, with many ECs expressing both PLVAP and GLUT1 and a smaller number expressing PLVAP and CLDN5 (Fig. 5B). This pattern contrasts with the adult pattern, in which CNS ECs with reduced levels of β-catenin signaling are generally either CLDN5⁺/GLUT1⁺/PLVAP⁻ or CLDN5⁻/GLUT1⁻/PLVAP⁺, and few or no ECs show an intermediate phenotype (e.g., Figs. 2 and 4). In E17 WT, Fz4^+/−;Wnt7a^−/−, and Fz4^−/−;Wnt7a^+/− CNS ECs, PLVAP was undetectable or barely detectable and CLDN5 and GLUT1 were uniformly expressed.

The mild CNS vascularization defects in Fz4^−/−;Wnt7a^−/− embryos are in contrast with the severe defects in Gpr124 and Reck mutants and in EC-specific Ctnnb1 (the gene coding for β-catenin) mutants (7), suggesting that additional Frizzled receptors and additional ligands—presumably Norrin and Wnt7b—were partially compensating for loss of Fz4 and Wnt7a. Similarly, the incomplete conversion of ECs from CLDN5⁺/GLUT1⁺/PLVAP⁻ to CLDN5⁻/GLUT1⁻/PLVAP⁺ may reflect residual β-catenin signaling.

In sum, these data imply that: (i) Wnt7a and Fz4 function broadly in the prenatal CNS; (ii) this function is partially covered by redundant ligands and receptors; (iii) altered expression of CLDN5, GLUT1, and PLVAP—especially in the olfactory bulb and striatum—is a sensitive indicator of reduced β-catenin signaling during embryogenesis; and (iv) the all-or-none nature of the EC choice between CLDN5⁺/GLUT1⁺/PLVAP⁻ to CLDN5⁻/GLUT1⁻/PLVAP⁺ states is a feature of more mature CNS ECs.

Genetic Interactions Between Tspan12 and Other β-Catenin Signaling Components in the Brain.

To assess the role of Tspan12 in BBB development and maintenance, we first examined Tspan12^−/− mice. Postnatal day (P)12 and adult Tspan12^−/− brains are indistinguishable from WT control brains in showing no Sulfo-NHS-biotin leakage into regions that are normally protected by the BBB and no detectable conversion of ECs from GLUT1⁺/PLVAP⁻ to GLUT1⁻/PLVAP⁺ (Fig. 6, column 1). The absence of a BBB phenotype in Tspan12^−/− mice and the mild cerebellar leakage in Ndp^KO mice (Fig. 2, second column) parallels the relative severities of the retinal angiogenesis defects associated with these two genotypes, and is consistent with the current model in which Tspan12 augments Norrin signaling but is not absolutely required for it (15, 16).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 6.

Anatomic distribution of BBB breakdown with combined loss of Tspan12 and Gpr124 in CNS ECs. Coronal sections of P12 brains from representative examples of the indicated genotypes. Boxed regions in the second and third rows are enlarged in the fourth and fifth rows, and show GLUT1⁺/PLVAP⁻ to GLUT1⁻/PLVAP⁺ conversion and Sulfo-NHS-biotin leakage in the anterior cerebellum (left part of each image) and the superior colliculus (right part of each image). The two Tspan12^−/−;Gpr124^CKO/−;Pdgfb-CreER brains show the stereotyped regional distribution (with some variability in severity) of the BBB breakdown phenotype. 4HT was administered at P8 and brains were analyzed at P12. These experiments used a short time interval between 4HT injection and sacrifice because mice with combined postnatal loss of the Norrin and Wnt7a/Wnt7b systems do not survive more than several days to several weeks, depending on the 4HT dose and the extent of BBB loss. (Scale bars: 1 mm for rows 1–3; 200 μm for rows 4 and 5.)

Because Ndp^KO;Wnt7a^−/− mice exhibit a massive loss of BBB integrity in the cerebellum and olfactory bulb, we asked whether Tspan12^−/−;Wnt7a^−/− mice exhibit BBB defects. Fig. 3 and SI Appendix, Fig. S4 show that Tspan12^−/−;Wnt7a^−/− mice have a normal BBB, implying that loss of Tspan12 produces only a partial reduction in Norrin-dependent signaling so that the β-catenin signal remains above the critical threshold for BBB maintenance in this double mutant combination.

We next asked whether a more substantial reduction in Wnt7a/Wnt7b signaling in a Tspan12^−/− background—produced by eliminating Gpr124 in postnatal ECs (with Pdgfb-CreER and a Gpr124^CKO allele)—could reduce the β-catenin signal below the critical threshold for BBB disruption (Fig. 6). While there was minimal conversion of cerebellar ECs from GLUT1⁺/PLVAP⁻ to GLUT1⁻/PLVAP⁺ in Tspan12^+/−;Gpr124^CKO/−;Pdgfb-CreER mice (Fig. 3), there was widespread low-level Sulfo-NHS-biotin labeling in the Tspan12^+/−;Gpr124^CKO/−;Pdgfb-CreER brain (Fig. 6, column 2). Eliminating Gpr124 in postnatal ECs in a Tspan12^−/− background (Tspan12^−/−;Gpr124^CKO/−;Pdgfb-CreER) produced severe Sulfo-NHS-biotin leakage in the hippocampus, superior colliculus, striatum, and brainstem, and milder leakage in the cerebral cortex, cerebellum, and hypothalamus, together with widespread conversion of ECs from GLUT1⁺/PLVAP⁻ to GLUT1⁻/PLVAP⁺ (Fig. 6, columns 3 and 4). This phenotype resembles the phenotype seen in postnatal Ndp^KO;Gpr124^CKO/−;Pdgfb-CreER mice (18), and implies a role for Tspan12 in Norrin signaling in postnatal brain ECs.

Finally, we asked whether loss of both Tspan12 and Fz4 enhances the BBB phenotype compared with the loss of Fz4 alone. Fig. 3 and SI Appendix, Fig. S5 show that the two genotypes exhibit indistinguishable distributions and severities of BBB loss, consistent with a model in which Tspan12 acts exclusively to enhance Norrin-Fz4 signaling.

The Wnt7a/Wnt7b System Plays a Minimal Role in Retinal Vascular Development.

As judged by publicly available RNA-seq datasets and published RT-PCR and in situ hybridization analyses in mice, the developing retina expresses Wnt7a at a low level and Wnt7b at a higher level; the adult retina expresses both Wnt7a and Wnt7b at a low level; and the developing lens expresses Wnt7a (37⇓–39). To assess the possibility that the Wnt7a/Wnt7b system plays a role in retinal vascularization and BRB development and maintenance, we began by looking for a genetic interaction between Wnt7a and Ndp by comparing retina flat mounts from Wnt7a^−/−, Ndp^KO, and Ndp^KO;Wnt7a^−/− mice at P30 (Fig. 7A). In Wnt7a^−/− retinas, ECs were uniformly GLUT1⁺/PLVAP⁻ and the vascular architecture was indistinguishable from that of WT retinas. In Ndp^KO retinas, ECs form numerous small clusters in place of intraretinal capillaries, and ECs in capillaries and veins are uniformly converted from GLUT1⁺/PLVAP⁻ to GLUT1⁻/PLVAP⁺, while ECs in arteries and arterioles remain GLUT1⁺/PLVAP⁻. The Ndp^KO phenotype was unaltered by the additional loss of Wnt7a. While these data imply that Wnt7a plays a far smaller role in retinal vascular development and BRB maintenance than Norrin, the severity of the Ndp^KO phenotype leaves little room for a worsening of the phenotype with the further loss of Wnt7a.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 7.

The Wnt7a/Wnt7b system has only minimal effects on retinal vascular development and BRB maintenance. Flatmounts of P30 (A and B) and P10 (C) retinas from representative examples of the indicated genotypes. (A) The Ndp^KO and Ndp^KO;Wnt7a^−/− vasculature lacks intraretinal capillaries and exhibits numerous EC clusters just beneath the vitreal face of the retina. The Wnt7a^−/− vasculature is indistinguishable from WT. In Ndp^KO and Ndp^KO;Wnt7a^−/− retinas, ECs in capillaries and veins (arrowheads) are converted from GLUT1⁺/PLVAP⁻ to GLUT1⁻/PLVAP⁺, whereas arteries (arrows) and arterioles remain GLUT1⁺/PLVAP⁻. (Scale bar, 500 μm.) (B) The Fz4^−/−;Tspan12^+/− retina is an example of a severe phenotype, with no intraretinal capillaries and many EC clusters just beneath the vitreal face of the retina. In the Tspan12^−/− retina (far Right), ECs in capillaries and veins are converted from GLUT1⁺/PLVAP⁻ to GLUT1⁻/PLVAP⁺, but vascular growth into the retina and intraretinal capillary formation are only partially impaired, with few EC clusters. Tspan12^−/−;Wnt7a^−/− retinas exhibit a retinal vascular phenotype indistinguishable from Tspan12^−/− retinas; two examples are shown. Boxed regions in the Upper images are enlarged in the Lower images. [Scale bars, 500 μm (Upper), 200 μm (Lower).] (C) Postnatal loss of Gpr124 (Gpr124^CKO/−;Pdgfb-CreER with 4HT at P6) synergizes with heterozygosity for Tspan12 to produce a partial conversion from GLUT1⁺/PLVAP⁻ to GLUT1⁻/PLVAP⁺ in the developing (P10) retinal vasculature. Boxed regions labeled a–d are enlarged in the Lower Right images. (Scale bars, 500 μm for low magnification, 100 μm for enlarged images.)

The experiments presented in Figs. 2–6, together with those reported in ref. 7, imply that small decrements in β-catenin signaling—as seen, for example, in Wnt7a^−/− or Tspan12^−/− brains—have little or no effect if the β-catenin signal is not reduced below a critical threshold. To reduce β-catenin signaling via the Norrin system in a manner that might permit the observation of phenotypic effects caused by small additional decrements, we studied genetic interactions in a background of homozygous and heterozygous loss of Tspan12.

Tspan12^−/− retinas resemble Fz4^−/− and Ndp^KO retinas in exhibiting a complete conversion of venous and capillary ECs from GLUT1⁺/PLVAP⁻ to GLUT1⁻/PLVAP⁺ (Fig. 7B, Right). However, as reported previously (15), Tspan12^−/− retinas exhibit an angiogenesis defect that is milder than that of Fz4^−/− and Ndp^KO retinas, with partial vascularization of the inner retina and fewer EC clusters near the vitreal surface [in Fig. 7B, compare Fz4^−/−;Tspan12^+/− (an example of a severe phenotype) with Tspan12^−/−]. Fig. 7B shows that Tspan12^−/−;Wnt7a^−/− retinas exhibit a vascular phenotype that is indistinguishable from the Tspan12^−/− phenotype, suggesting that Wnt7a produced by the retina and lens has little or no effect on retinal vascular anatomy. However, a limitation of this experiment is that the full conversion of Tspan12^−/− vein and capillary ECs to a GLUT1⁻/PLVAP⁺ state precludes an assessment of Wnt7a effects on the BRB phenotype.

In a second approach, we started with a Tspan12^+/− genetic background, which only minimally reduces β-catenin signaling and leaves retinal ECs in a GLUT1⁺/PLVAP⁻ state, and then studied the effect of impairing the Wnt7a/Wnt7b system. Tspan12^+/− mice and Tspan12^+/−;Wnt7a^−/− mice showed no abnormalities in retinal vascularization and no conversion of retinal capillary and venous ECs from GLUT1⁺/PLVAP⁻ to GLUT1⁻/PLVAP⁺ (Fig. 7C). Gpr124^CKO/−;Pdgfb-CreER mice also show no retinal vascular phenotype at P10 (Fig. 7C) or at P30. However, Tspan12^+/−;Gpr124^CKO/−;Pdgfb-CreER mice show a partial conversion of capillary and venous ECs from GLUT1⁺/PLVAP⁻ to GLUT1⁻/PLVAP⁺ (Fig. 7C, Lower row). [The BBB disruption that results from the combined postnatal loss of Tspan12 and Gpr124 (as seen in Fig. 6), limits the time window available between 4HT injection (P6) and phenotypic analysis (P10) for the experimental series shown in Fig. 7C, and this may lead to an underestimation of the full extent of the retinal phenotype.] Fig. 7C also shows that, at P10, Tspan12^−/− retinas have already acquired the GLUT1⁻/PLVAP⁺ phenotype that they exhibit at P30 (Fig. 7B), and that this phenotype and the modestly reduced vascular density characteristic of the P10 Tspan12^−/− retinal vasculature are unaffected by the additional loss of Gpr124.

Taken together, these experiments imply that Wnt7a/Wnt7b signaling is active at a low level in the retinal vasculature, and that its loss can only be observed if the Norrin system is partially compromised. It is presently unknown whether the lower level of Wnt7a/Wnt7b signaling in the retina reflects a lower level of Wnt7 ligands compared with Norrin, a lower level of Gpr124 or Reck compared with Tspan12, the presence of extracellular Wnt inhibitors, or some other factors.

Specificity of Tspan12 Action in a Cell Culture System.

To date, Tspan12 loss-of-function phenotypes have been interpreted on the assumption that Tspan12 acts only to augment Norrin signaling via its receptor Fz4. This model is based on the observation that Tspan12 has no effect on Wnt3a, Wnt5a, or Wnt7b signaling in the same cell culture system in which it augments Norrin signaling by ∼five- to sevenfold (15, 16). In the present context, it is of interest to systematically determine whether Tspan12 can augment signaling via any of the other Wnts.

We addressed this question by measuring the level of β-catenin signaling in a stable luciferase reporter cell line [Super Top Flash (STF) cells] with or without an N-terminal Rim-epitope tagged Fz4 and Tspan12 and with no coexpressed ligand or with coexpression of each of the 19 mammalian Wnts or Norrin (Fig. 8). We used Rim-tagged Fz4 rather than untagged Fz4, because—compared with untagged Fz4—Rim-tagged Fz4 exhibits greater responsiveness to Tspan12 (∼30-fold) despite an overall decrease of ∼8-fold in signal amplitude (Fig. 8, Inset). Immunoblotting shows that the majority of the Rim-tagged Fz4 proteins have a lower mobility in SDS/PAGE, suggesting that only a minority of the Rim-tagged Fz4 proteins follow the correct pathway of biosynthesis and intracellular trafficking (SI Appendix, Fig. S6). RNA sequencing has shown that STF cells express many Frizzleds at low level, and this likely accounts for the nonzero luciferase signal observed in the absence of Frizzled transfection (7). As seen in Fig. 8, the magnitude of luciferase activation depended on which Wnt was expressed. For each of the 19 Wnts there was either no change or a modest inhibition in β-catenin signaling in the presence of Tspan12. The only ligand that showed an enhanced signal in the presence of Tspan12 was Norrin (Fig. 8, arrow).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 8.

Specificity of Tspan12 for Norrin vs. Wnts in a cell-based assay of β-catenin signaling. STF cells were transiently transfected with plasmids coding for Norrin, the indicated Wnt, or an empty vector together with the indicated combination of plasmids coding for Rim-Fz4, and/or Tspan12, or empty vector, together with Renilla luciferase. The ratio of firefly luciferase activity from the stably integrated STF reporter relative to Renilla luciferase activity is shown. Norrin is the only ligand that shows an enhanced luciferase signal in response to Tspan12 expression (red arrow). (Inset) Comparison between untagged Frizzled4 and Rim-Fz4, with or without cotransfected Tspan12. Mean values ± SD are shown. Statistical significance values for pairwise comparisons with the vector-only controls are shown in SI Appendix, Table S1. The inhibition of some Wnts (e.g., Wnt1 and Wnt3a) by some Frizzleds may represent competition for essential signaling components by overexpressed proteins.

Taken together, these data support the original model that Tspan12 specifically enhances Norrin signaling (15), and they mirror the results of analogous STF experiments showing that Gpr124 and Reck specifically enhance Wnt7a and Wnt7b signaling (14, 18⇓–20, 25).