A Norrin/Wnt surrogate antibody stimulates endothelial cell barrier function and rescues retinopathy

Abstract The FZD4:LRP5:TSPAN12 receptor complex is activated by the secreted protein Norrin in retinal endothelial cells and leads to βcatenin‐dependent formation of the blood–retina–barrier during development and its homeostasis in adults. Mutations disrupting Norrin signaling have been identified in several congenital diseases leading to hypovascularization of the retina and blindness. Here, we developed F4L5.13, a tetravalent antibody designed to induce FZD4 and LRP5 proximity in such a way as to trigger βcatenin signaling. Treatment of cultured endothelial cells with F4L5.13 rescued permeability induced by VEGF in part by promoting surface expression of junction proteins. Treatment of Tspan12 −/− mice with F4L5.13 restored retinal angiogenesis and barrier function. F4L5.13 treatment also significantly normalized neovascularization in an oxygen‐induced retinopathy model revealing a novel therapeutic strategy for diseases characterized by abnormal angiogenesis and/or barrier dysfunction.


Introduction
Defects in blood-brain barrier (BBB) and blood-retina barrier (BRB) integrity underlie the ontogeny or progression of a number of diseases such as stroke, pathogen infections, diabetic retinopathy, and neurodegenerative diseases (Obermeier et al, 2013;Sweeney et al, 2019). Wnt/bcatenin signaling is required for CNS angiogenesis and endothelial cell barrier function during development (Liebner et al, 2008;Ye et al, 2009;Wang et al, 2012) and for the maintenance of the BRB and BBB during postnatal tissue homeostasis (Chang et al, 2017;Wang et al, 2018). Among the ten vertebrate Frizzled (FZD) receptors, FZD4 is expressed in endothelial cells where it acts as a cell surface receptor for the secreted proteins Norrin (NDP) ) and WNT7A/B (Cho et al, 2017). Activation of FZD4 leads to bcatenin-mediated regulation of contextdependent genes important for barrier functions, including Sox17, Claudin-5, and MFSD2 (Sch€ afer et al, 2009;Ye et al, 2009;Wang et al, 2020) and thus limits paracellular permeability and endothelial cell transcytosis (Ben-Zvi et al, 2014). Norrin plays an especially important role for the development and maintenance of the retinal vasculature (Luhmann et al, 2005;Wang et al, 2012Wang et al, , 2018. In this context, LRP5 and TSPAN12 function as co-receptors with FZD4 and are required for Norrin signaling (Junge et al, 2009). Whereas LRP5 and LRP6 are also obligatory co-receptors for Wnt proteins during bcatenin signaling, TSPAN12 acts as a selectivity gate and signal amplifier exclusively in Norrin-induced signaling (Lai et al, 2017). Highlighting the critical roles of Norrin-FZD4 signaling in the retina, NDP, FZD4, LRP5, and TSPAN12 mutations were identified in a number of related congenital diseases with ocular manifestations, such as Norrie disease, osteoporosis-pseudoglioma syndrome, and familial exudative vitreoretinopathy (FEVR; Warden et al, 2007;Baron & Kneissel, 2013;Gilmour, 2015), each of which are associated with hypovascularization of the retina and severe loss of visual function. Overall, disruption of Ndp (Richter et al, 1998), Fzd4 (Xu et al, 2004), Lrp5 (Xia et al, 2010;Chen et al, 2011), or Tspan12  (Junge et al, 2009;Zhang et al, 2018) genes in mice causes similar retinal vascular phenotypes characterized by abnormal blood vessel development and BRB defects. Importantly, endothelial cell-specific expression of a dominantnegative T-cell factor-4 (TCF4) or a stabilized version of bcatenin is sufficient to, respectively, phenocopy or reverse the Ndp and Fzd4 loss-of-function phenotypes, suggesting that Norrin signaling chiefly functions through regulation of a bcatenin endothelial cell transcriptional program (Zhou et al, 2014). Thus, pharmacological activation of FZD4/bcatenin signaling in endothelial cells may represent a novel strategy to promote or restore barrier function by mimicking Norrin and/or WNT7A/B activity and could represent a novel therapeutic opportunity for the treatment of retinal vascular diseases or neurological diseases driven by barrier defects.

F4L5.13 activates bcatenin signaling in endothelial cells
We previously described an induced proximity tetravalent antibody platform called FLAg (Frizzled and LRP5/6 Agonist), which promotes the clustering of Frizzled and LRP5/6 co-receptors and thereby mimics the activity of Wnt proteins to activate bcateninmediated transcription (Tao et al, 2019). Given the genetic evidence supporting the role for FZD4 and LRP5 for the development of the retinal vasculature and maintenance of barrier function, we used antibody fragments to assemble F4L5.13, a highly potent and selective FLAg mediating FZD4 and LRP5 activation. F4L5.13 consists of a diabody formed by two identical paratopes recognizing FZD4 fused to the N-terminus of a heterodimeric Fc, and a diabody composed of two distinct paratopes, respectively, recognizing the first two propellers (E1E2; Wnt-1 binding site) and the membrane proximal two propellers (E3E4; Wnt-3 binding site) of LRP5 fused to the C-terminus of the Fc (Fig 1A). F4L5.13 interacts with FZD4 with pM affinity, displays single digit nM binding to LRP5, and importantly, does not interact with any of the 9 other vertebrate FZD receptors and is selective for LRP5 over LRP6 (Fig EV1A-C). Thus, in contrast to pervasive activators of bcatenin signaling such as GSK3b inhibitors or purified Wnt proteins, F4L5.13 should lead to bcatenin-mediated transcriptional regulation in only those cells that co-express FZD4 and LRP5.
We first compared F4L5.13 and NDP for their ability to activate bcatenin-signaling using the pBAR reporter, which faithfully monitors LEF/TCF-mediated transcription. Treatment of HEK293T cells (which express only trace amounts of FZD4) with F4L5.13 or NDP required the presence of ectopically expressed FZD4 to detect reporter activation, whereas treatment with WNT3A led to activation in both wild-type cells and when FZD4 was expressed (Figs 1B and EV1D). Supporting these results, flow cytometry experiments detected F4L5.13 binding only in HEK293T cells overexpressing either FZD4 or LRP5 (Fig EV1E). Consistent with the NDP-specific functions described for TSPAN12, co-expression of FZD4 with TSPAN12 led to potentiation of the NDP response but did not affect the activation mediated by F4L5.13 ( Fig 1C). We further identified the murine immortalized brain microvasculature cell line bEnd.3 as a highly NDP-responsive cell line. In these cells, F4L5.13 activated expression of the bcatenin target gene Axin2 with similar potency but with higher efficacy than NDP ( Fig 1D). Knockdown of Tspan12 led to blunting of the NDP response as previously described (Junge et al, 2009;Lai et al, 2017), but did not affect signaling promoted by F4L5.13 (Figs 1E and EV2A). In contrast, bEnd.3 cells treated with Fzd4 or Lrp5 siRNA were largely unresponsive to either NDP or F4L5.13 . Both NDP and F4L5.13 similarly led to Dishevelled phosphorylation and bcatenin stabilization ( Fig 1F). We conclude that FZD4 and LRP5 clustering triggered by F4L5.13 is sufficient to activate bcatenin signaling in endothelial cells, and as such, F4L5.13 defines a novel class of FZD4-specific agonists.
A Top: Molecular architecture of tetravalent F4L5.13. Bottom: Schematic for activation of Wnt/bcatenin signaling by F4L5.13. B Activation of bcatenin signaling by F4L5.13 or recombinant NDP (30 nM each) in HEK293T cells transfected with FZD4 and/or LRP5. Values represent fold activation of LEF/TCF reporter gene. Data are presented as mean AE SEM, n = 3. C Dose-response curves for the activation of a LEF/TCF reporter gene in HEK293T cells transfected with plasmids encoding FZD4, LRP5 and with either GFP or TSPAN12 by serial dilutions of F4L5.13 or NDP proteins (x-axis). Data are presented as mean AE SEM, n = 3. D RT-qPCR of Axin2 expression in bEnd.3 cells treated with serial dilutions of F4L5.13, NDP, or isotype control for 24 h. Data are presented as mean AE SEM, n = 3. E RT-qPCR of Axin2 in bEnd.3 cells treated with NDP (200 ng/ml), isotype control or F4L5.13 (1,200 ng/ml) and transfected with control, Tspan12 or Fzd4 siRNAs. Data are presented as mean AE SD, n = 3 technical replicates. Data are representative of two independent experiments. F Time course of phosphorylated Dishevelled-3 (p-DVL3) and bcatenin protein levels in bEnd.3 cells treated with 30 nM of F4L5.13 or NDP. Histogram represents the ratio of the DVL3 phosphorylation levels over total DVL3 protein and bcatenin levels over b-Tubulin measured by densitometry of independent experiments. Data are presented as mean AE SEM, n = 4-6 (*P < 0.05 as compared with NT). Significance was calculated by one-way ANOVA with Bonferroni's multiple comparisons test (*P < 0.05 as compared with NT). A Volcano plot of gene expression changes in bEnd.3 cells following 24 h of treatment with 30 nM F4L5.13. Genes with significant changes in expression level are highlighted in blue and genes co-regulated by treatment with 30 nM NDP for 24 h are highlighted in red. Statistical analysis was performed using the DESeq2 R package and adjusted P-value (Wald test) threshold less than 0.1 was used. Labeled genes are associated with enriched GO biological processes indicated below. B Enrichment Map of overrepresented GO biological processes and pathways in genes regulated by F4L5.13. Node sizes represent the number of genes belonging to individual terms (g:SCS FDR adjusted P-value < 0.05, gProfiler). C Venn diagrams showing overlap in differentially expressed genes following treatment with F4L5.13 or NDP. D Heat maps of differentially expressed genes mapping to the indicated biological processes and pathways across treatment groups in RNA-seq experiment. Mean of n = 3 samples per group.

F4L5.13 mimics NDP treatment in bEnd.3 cells
We next used RNA-seq analysis to identify differentially expressed genes upon treatment of bEnd.3 cells with F4L5.13 or NDP for 8 or 24 h (Fig 2A, Table EV2). Supporting the results above, F4L5.13 treatment strongly induced Axin2 expression indicating robust activation of bcatenin signaling. Genes differentially expressed following F4L5.13 treatment were enriched for biological processes such as vasculature development/angiogenesis, surface-protein signaling, cell adhesion, and epithelium development (Figs 2B and EV3), consistent with the previously described role of the FZD4-LRP5 signaling axis in CNS vascular development. Importantly, treatment of bEnd.3 cells with NDP for either 8 or 24 h led to differential expression of an overlapping set of genes confirming that F4L5.13 effectively mimics NDP function in endothelial cells (Fig 2A-C). Indeed, following treatment for 24 h, 91% (20/22) of the genes regulated by NDP were also regulated by F4L5.13. A closer examination of expression changes revealed genes linked to GO-Biological Process terms Cell surface receptor signaling pathway (GO:0007166), Cell adhesion (GO:0007155), and Vascular development (GO:0001944) as well as the Wnt signaling pathway (KEGG:04310). When compared with NDP, treatment with F4L5.13 consistently led to increased magnitude of gene induction/repression both by RNA-seq and qPCR validation, which possibly reflects its increased efficacy (Figs 2D and EV4A). Among the genes regulated by both NDP and F4L5.13, Angpt2 was previously shown to be induced by NDP treatment of cultured human retinal microvascular endothelial cells and to mediate, at least in part, the NDP-mediated increase in proliferation (Ohlmann et al, 2010). Another gene strongly induced by both NDP and F4L5.13 was the membrane protein MAL that was previously identified as the receptor for the Clostridium perfringens epsilon toxin (ETX) on endothelial cells and shown to be required for the ETX effects on blood-brain barrier permeability (Linden et al, 2019).

F4L5.13 treatment promotes endothelial cell barrier function
One of the earliest features underlying multiple retinopathies is the Vascular Endothelial Growth Factor (VEGF)-mediated breakdown of the blood-retina barrier and increased endothelial cell permeability. This underlies the current standard of care use of anti-VEGF therapies such as Lucentis (ranibizumab), Eylea (aflibercept), Beovu (brolucizumab), and Avastin (bevacizumab) that revolutionized the clinical management of vascular and exudative diseases of the retina (e.g. macular edema, retinopathy of prematurity, age-related macular degeneration, diabetic retinopathy, retinal vein occlusion, and myopic choroidal neovascularization). However, incomplete responses to anti-VEGF therapy and therapeutic resistance are common clinically and lead to persistent disease and significant unmet clinical needs (Gross et al, 2018). Moreover, accumulating evidence indicates that anti-VEGF therapy may not be reversing the nonperfusion underlying pathological diabetic retinopathy (Nicholson et al, 2018).
Knowing that FZD4/bcatenin signaling is required for the development and maintenance of CNS endothelial cell barrier properties, we therefore asked whether F4L5.13 treatment could promote barrier functions and oppose the effects of VEGF and determine whether this could represent a novel treatment strategy. Treatment of bEnd.3 cells with F4L5.13 led to increased bcatenin levels and upregulation of ZO-1 expression to comparable levels as NDP ( Fig 3A). Then, we investigated the ability of F4L5.13 to rescue endothelial cell barrier breakdown induced by the endothelial cell permeability factor VEGF. When cells were pretreated with VEGF, the endothelial cell barrier was rapidly disrupted, as observed by marked decrease in cell surface localization of tight junction proteins ZO-1, CLDN3, and CLDN5 ( Fig 3B). F4L5.13 treatment following VEGF stimulation led to a robust rescue of ZO-1, CLDN3, and CLDN5 cell surface expression indicating that stimulation of FZD4/bcatenin signaling was sufficient to restore tight junction organization disrupted by VEGF treatment (Fig 3B and C). Importantly, in the bEnd.3 cell model bcatenin signaling did not lead to increased Cldn5 mRNA levels, as was previously shown in vivo, or to changes in transcript levels of other cell junction components ( Fig EV4B). We conclude that in bEnd.3 cells the effect of F4L5.13 on cell surface levels of junction components is not a result of direct transcriptional regulation. To functionally assess endothelial cell barrier function, we measured the permeability of a confluent endothelial monolayer for 40-kDa FITC-dextran. bEnd.3 cells were grown on transwell filter inserts until confluent and were then treated with VEGF and F4L5.13 simultaneously or were treated with VEGF for 1 h followed by F4L5.13 for 1 h. In both treatment schedules, incubation of bEnd.3 cells with F4L5.13 significantly rescued VEGF-induced permeability ( Fig 3D). We conclude from these results that specific activation of the FZD4-LRP5 receptor complex by F4L5.13 promotes cell barrier function, in part, by promoting the assembly of tight junctions.
F4L5.13 restores retinal vascular development and homeostasis in Tspan12 À/À mice We next tested whether F4L5.13 could induce endothelial cell barrier function in vivo. To do so, we used Tspan12 À/À mice with with or without 100 ng/ml VEGF for 1 h. ZO-1 is shown in green, CLDN3/5 in red, and DAPI in blue. Scale bars: 15 µm. C Quantification of ZO-1, CLDN3, and CLDN5 fluorescence intensity. Each column represents 40 measurements of fluorescence intensity per condition (y-axis). Data are presented as mean AE SEM. Significance was calculated by one-way ANOVA with Bonferroni's multiple comparisons test (*P < 0.05 as compared to VEGF treatment). D Transendothelial permeability assay quantifying the passage of FITC-dextran through a monolayer of bEnd.3 cells. Passage of FITC-dextran was measured after exposure of bEnd.3 cells to 100 ng/ml VEGF, 30 nM F4L5.13 or both or pretreated with VEGF for 1 h before treating with F4L5.13 for 1 h. Data are presented as mean AE SD, n = 5 independent experiments. Significance was calculated by one-way ANOVA with Bonferroni's multiple comparisons test (*P < 0.05 as compared to VEGF treatment). selectively impaired NDP signaling in the retina, which display retinal vascular morphogenesis defects and impaired endothelial cell barrier functions. TSPAN12 is one of the genes mutated in FEVR (Gilmour, 2015), and in many respects, Tspan12 À/À animals recapitulate features observed in FEVR patients. We administered F4L5.13 or an isotype control antibody systemically into developing Tspan12 À/À mice between P5 and P20, i.e., slightly before and during the time when intraretinal vascular development and barrier formation normally occur. P20 retinas were stained with Isolectin B4 (to reveal endothelial cells and microglia) and antimouse IgG (to probe for barrier defects). Stitched high-resolution images ( Fig 4A) and optical sections ( Fig 4B) were used to document the formation or malformation of the three-layered retinal vasculature. Mice injected with isotype control antibody displayed the characteristic phenotypes of Tspan12 À/À mice: massive leakage of endogenous IgG through an impaired BRB, absence of intraretinal capillaries in the deep layer, and glomeruloid vascular malformations instead of a properly formed intermediate capillary layer . Strikingly, BRB function and angiogenesis were virtually completely restored in Tspan12 À/À mice injected with F4L5.13 ( Fig 4A and B), i.e., no substantial IgG extravasation was observed, and intraretinal capillary beds were properly formed. Analysis of retinal cross sections showed that intraretinal capillary formation was restored in the entire retina including central and peripheral regions and that no misdirected angiogenesis occurred (Fig 4C). Quantification of vascular density in the outer plexiform layer (in which the deep capillary bed normally resides) revealed that the FZD4:LRP5 agonist restored intraretinal blood vessel formation in Tspan12 À/À mice to an extent virtually indistinguishable from wild-type mice ( Fig 4D). As expected, the strong BRB defects in Tspan12 À/À mice were associated with reduction of CLDN5 expression and upregulation of the EC fenestration component PLVAP. Interestingly, Tspan12 À/À mice treated with F4L5.13 led to a strong increase in CLDN5 and decreased PLVAP expression ( Fig 4E). These findings indicate that F4L5.13 restored FZD4:LRP5 signaling to a level above the threshold required for angiogenesis and barrier formation. Signaling levels may have exceeded endogenous NDP/FZD4 signaling levels, but this did not induce misdirected angiogenesis into avascular layers of the retina or cause an increase of vascular density, perhaps reflecting the fact that Wnt signaling target genes include potent mediators of negative feedback loops. The findings that F4L5.13 activates FZD4:LRP5 signaling independently of TSPAN12 in bEnd.3 cells ( Fig 1E) and restores the retinal vasculature in Tspan12 À/À mice (Fig 4A-E) together confirm the ligand-specific functions of TSPAN12 in NDP/FZD4 signaling. The striking efficacy of F4L5.13 in the developmental angiogenesis and barrier genesis context prompted us to investigate the effect of F4L5.13 in the adult retinal vasculature characterized by BRB defects. F4L5.13 or isotype control antibody was administered systemically to 2-to 3-month-old Tspan12 À/À mice, which were then subjected to an acute BRB assay based on quantification of 4 kDa FITC-dextran extravasation during a 1 h time window (Fig 5A). Retinal lysates of Tspan12 À/À mice treated with the isotype control antibody contained high levels of FITC-dextran tracer, whereas BRB defects were significantly alleviated in Tspan12 À/À mice treated with F4L5.13 ( Fig 5B). This partial restoration of the BRB was not associated with an increase of vascular density ( Fig 5C). Furthermore, analysis of Tspan12 À/À mice injected systemically with sulfo-NHS-biotin, a tracer that forms covalent bonds with proteins and is detected with fluorescent streptavidin probes, confirmed that F4L5.13 partially restored BRB function but had no discernable effect on the aberrant vascular morphology in adult mutant mice ( Fig 5D). Specifically, no blood vessels in the ◀ Figure 4. F4L5.13 restores intraretinal angiogenesis and blood-retina barrier formation in Tspan12 À/À mice.
A F4L5.13 prevents BRB defects and restores vascular morphogenesis in Tspan12 À/À mice. 10 mg/kg F4L5.13 or isotype control antibody was injected i.p. every 48 h from P5 to P20. Multiple images acquired at 10× magnification were stitched together to create a high-resolution image of a representative part of the retinal flat mounts. Scale bar 500 µm. B Optical sections of IB4-stained P20 retinas showing the mouse three-layered retinal vasculature (superficial, intermediate, and deep layers at 20× magnification) to document blood vessel formation or malformation. In Tspan12 À/À mice, the intermediate capillary layer fails to properly develop and glomeruloid vascular malformations are present. The deep capillary layer is virtually absent. F4L5.13 fully restores intraretinal capillary development in Tspan12 À/À mice. Scale bars: 100 µm. C IB4-stained retinal cross sections show the restored retinal vasculature and the absence of misdirected angiogenesis after F4L5.13 treatment. Boxed areas are shown enlarged in the panels on the right. Scale bars: 500 µm. D Quantification of the vascularized area in the outer plexiform layer. 10 mg/kg F4L5.13 or isotype control antibody was injected every 48 h from P5 to P20. Data are presented as meanAE SD, n = 5 retinas (2-4 fields of view per retina) from 4 to 5 mice per group. Significance was calculated by one-way ANOVA with Bonferroni's multiple comparisons test. E F4L5.13 rescues barrier function defects in Tspan12 À/À mice as shown by an increased expression of the tight junction component CLDN5 and decreased expression of the EC fenestration component PLVAP. IB4-Alexa 647 was used to stain ECs. Scale bars: 100 µm.
A Injection schema for blood-retina barrier assay in adult Tspan12 À/À mice with established retinal vascular defects. B Quantification of extravasated FITC-dextran in retinal lysates from WT and Tspan12 À/À mice injected with isotype control antibody or F4L5.13. Data are presented as mean AE SEM, n = 6 retinas from 3 mice per group. Significance was calculated by one-way ANOVA with Bonferroni's multiple comparisons test (*P < 0.05 as compared to Tspan12 À/À control treatment). C Quantification of the vascularized area in adult Tspan12 À/À mice injected with isotype control or F4L5.13. Data are presented as meanAE SD, n = 4 retinas (4 fields of view per retina) from 2 mice per group. P-value was calculated by Student's t-test. D Flat-mounted retinas of adult Tspan12 À/À mice injected systemically with sulfo-NHS-biotin show that F4L5.13 partially restored BRB function. PECAM was used to stain endothelial cells. Images are representative of 4 retinas per group. Scale bar: 500 µm. E Maximum intensity projection of adult Tspan12 À/À retinas after mice were injected with F4L5.13 or isotype control. Endothelial cells were stained with IB4-Alexa 647.
Images are representative of 4 retinas per group. Scale bar: 100 µm. outer plexiform layer and no vascular cells with tip cell morphology were observed (Fig 5E). Thus, F4L5.13 partially restored BRB function in mature animals, in which barrier induction had failed during development. Administration of F4L5.13 was not sufficient to induce angiogenesis in the quiescent adult vasculature of Tspan12 À/À mice. These results show that blood-CNS barrier function can be modulated by FZD4:LRP5-specific agonists in the mature CNS and that the antibody agonists described herein are a novel class of drug candidates that promote and restore barrier function in pathologically altered CNS vasculatures. A B Figure 6. A FZD4:LRP5 antibody agonist reduces pathological neovascularization in the OIR model.
A Schematic diagram of the OIR model (top) and representative images of P17 OIR retinas are shown (bottom). Neonatal mice were exposed to 75% oxygen from P7 to P12 to induce vessel loss. Then mice received intravitreal injections of PBS vehicle, mouse-specific anti-VEGF (as a positive control) delivered at 0.1 lg/ll, or F4L5.13v2 (a second-generation F4L5.13 modality, in which the N-terminal FZD4-specific diabody was replaced with a FZD4-specific Fab) at 50 nM or 500 nM target vitreous concentrations. Mice were returned to room air from P12 to P17 to induce maximum pathologic neovascularization at P17. At P17, retinas from injected mice were collected, dissected as flat-mounts, and stained with Isolectin B4. Scale bar = 500 µm. B Quantification of the ratio of neovascular and avascular area on total area of vascularization. Data are presented as mean AE SEM, n = 11-13 retinas per group.
Significance was calculated by one-way ANOVA with Bonferroni's multiple comparisons test (indicated P-values are in comparison to Vehicle).

A FZD4:LRP5 antibody agonist reduces pathological neovascularization in the OIR model
We next evaluated the efficacy of a NDP mimetic FZD4:LRP5 antibody agonist to reduce the formation of pathologic retinal neovascularization (NV) in the oxygen-induced retinopathy (OIR) model, which exhibit pathological features reminiscent of retinopathy of prematurity (ROP) and other human retinopathies such as diabetic retinopathy. For the OIR model, mouse pups were exposed to 75% oxygen from P7 to P12 followed by a return to room air oxygen until P17 to induce retinal ischemia and proliferative vascular disease in the retinal vasculature, at which point they were sacrificed and their retinas isolated. At P12, mice received intravitreal injections of PBS vehicle, mouse-specific anti-VEGF antibodies (positive control), or the FZD4: LRP5 agonist antibodies. Quantification of pathological neovascularization revealed that both treatment groups significantly reduced the neovascular area when compared to vehicle-injected mice (Fig 6A  and B). No significant effect on the avascular area was observed in any treatment group. These results in combination with the work detailed above demonstrate clear in vivo efficacy of the FZD4:LRP5 agonist in multiple animal models of human retinopathies.

Discussion
Taken together, our results describe the development of F4L5.13, a selective tetravalent antibody agonist of the FZD4:LRP5 receptor complex. F4L5.13 promotes the proximity of FZD4 with LRP5 coreceptors in such a way as to stabilize a conformation that is likely induced by NDP or Wnt proteins and compatible with downstream signaling activated by these endogenous growth factors. Such molecules extend the emerging landscape of multifunctional drugs functioning through "induced proximity" (Deshaies, 2020) such as molecular glue (e.g. immunomodulatory IMiDs drugs) or proteolysis targeting chimeras (PROTACs) that recruit neosubstrates to ubiquitin ligase scaffolds for degradation. Our study also describes the bEnd.3 mouse brain endothelial cell line as being Norrin-responsive, capable of mounting a bcatenin transcriptional response in response to NDP that requires the FZD4-LRP5-TSPAN12 receptor complex. bEnd.3 cell line is a Polyoma middle T transformed line and although these cells may provide a powerful in vitro system to tease apart signaling mechanisms, an important limitation is the difference in the transcriptional responses observed following treatment with NDP or F4L5.13 and the target genes known to be involved in vivo. For example, whereas Cldn5 is a known bcatenin target gene in endothelial cells in vivo (Wang et al, 2012; Fig 4E), it is not modulated by NDP or F4L5.13 in bEnd.3 cells (Fig EV4B).
To our knowledge, F4L5.13 is the first biologic that activates Wnt/bcatenin signaling and displays efficacy in multiple mouse models of human disease. In the context of the retinal vasculature, stimulation of FZD4:LRP5 signaling may provide benefits in FEVR or other rare congenital diseases that harbor mutations within Wnt pathway components but could also help mitigate the vascular leakage associated with neovascularization in various clinical settings including diabetic retinopathy and wet macular degeneration that represent clinical unmet needs. Interestingly, F4L5.13 also suppressed neovascularization in the OIR model, possibly reflecting functional antagonism of VEGF activities. In addition, therapeutic activation of FZD4 could also be beneficial in other vascular diseases of the CNS-such as stroke, traumatic brain injury, multiple sclerosis and Alzheimer's disease-where endothelial cell barrier dysfunction is proposed to contribute to disease initiation and/or progression.

Materials and Methods
Cell culture and treatment bEnd.3 cells, obtained from ATCC (CRL-2299), were grown in DMEM with 4.5 g/l glucose, 3.7 g/l sodium bicarbonate, 4 mM glutamine, 10% FBS, 1% penicillin/streptomycin and cultured in 10 cm dishes until confluency. bEnd.3 cells were passaged using trypsin 0.25%-EDTA solution at a split ratio of 1:4-1:10 and cells were used between p2 and 19 for the experiments. For cell treatment, 200,000 cells were seeded per well of a 6-well plate. After 16 h, cells were around 60-70% confluency and were treated with Norrin (R&D, Cat# 3014-NR), F4L5.13, and isotype control for 24 h. HEK293T cells were authenticated with STR profiling at The Centre for Applied Genomics (Toronto, Ontario, Canada). HEK293T cells were cultured in high-glucose DMEM with 10% FBS at 37°C in the presence of 5% CO 2 . For maintenance, cells were split 1:6 at near confluence using 0.05% Trypsin-EDTA. Cells were routinely tested for Mycoplasma using the MycoAlert Plus detection kit (Lonza, LT07-701). . Fc-tagged fusions of the FZD6 ECD were prepared as previously described (Tao et al, 2019).

Recombinant Proteins and Reagents
F4L5.13 and the isotype control were cloned and produced as previously described (Tao et al, 2019). The Wnt receptor-binding paratopes which comprise F4L5.13 were identified from recombinant selections on mouse LRP5 or human FZD4 using a phage displayed synthetic Fab library (Persson et al, 2013) using methods as previously described (Tao et al, 2019). For F4L5.13v2, the N-terminal diabody targeting FZD4 was replaced with a Fab specific for FZD4. The isotype control molecule is comprised of paratopes that bind to a non-host protein (gaussia luciferase) that were identified in a similar manner using recombinant phage-display selections.

Luciferase reporter assay
HEK293T cells were transduced with lentivirus coding for the pBARls reporter (Biechele & Moon, 2008)  prior to stimulation. The following day, F4L5.13 protein was added, and following 15-20 h of stimulation, cells were lysed and luminescence was measured in accordance with the dualluciferase protocol (Promega) using an Envision plate reader (PerkinElmer). For TSPAN12 dependency, HEK293T-TopFlash cells were seeded at 0.5 × 10 6 cells in a 6-well plate and the next day cells were transfected with FZD4, LRP5 and with either GFP or TSPAN12. Cells were transfected using Lipofectamine 2000 according to the manufacturer's instructions. The next day, cells were passed in a 96-well plate with around 35,000 cells/well, and after 5 h, cells were treated with either F4L5.13 or recombinant NDP overnight. After 16 h, cells were lysed, and luminescence was measured.

Biolayer interferometry
BLI experiments were performed using an Octet HTX instrument (Forte Bio). FZD selectivity experiments were performed by immobilizing Fc-fused antigen on AHQ BLI Sensors (Cat# 18-5001, ForteBio) to achieve a BLI response of 0.4-0.6 nm. Remaining binding sites on the sensors were then saturated with human Fc (009-000-008, Jackson ImmunoResearch). Antigen or control (Fc alone) coated sensors were transferred to 100 nM of F4L5.13 or isotype control diluted in assay buffer (PBS, 1% BSA, 0.05% Tween-20). Association was monitored for 300 s with a shake speed of 1,000 rpm at a temperature of 25°C. Steady-state response values were determined at 295 s into the association phase. The specific binding signal for each sensor was determined by subtracting the signal from corresponding Fc-coated control sensors. Affinity measurements of F4L5.13 for FZD4 were performed in a similar manner, but FZD4-coated sensors were transferred to a series of five F4L5.13 dilutions ranging from 25 nM to 1.56 nM. Association was monitored for 300 s after which sensors were transferred to wells containing assay buffer and dissociation was monitored for an additional 300 s with a shake speed of 1,000 rpm at a temperature of 25°C. Kinetic constants were determined by fitting the resulting data were to a 1:1 langmuir model using the Octet Data Analysis software (v11.1.0.4).
Affinity measurements of F4L5.13 for mLRP5 were performed by immobilizing mLRP5 on AR2G sensors (18-5092, ForteBio) to a level of 0.3-0.4 nm via aminocoupling. Remaining amino-reactive sites on the sensors were quenched with ethanolamine and allowed to equilibrate in assay buffer for 30 min. Sensors were then transferred to a series of 7 F4L5.13 dilutions ranging from 100 nM to 0.137 nM. Association was monitored for 300 s after which sensors were transferred to wells containing assay buffer and dissociation was monitored for an additional 300 s. Kinetic constants were determined as described above for FZD4 affinity measurements.

Flow cytometry
HEK293T cells were seeded at 0.5 × 10 6 cells in a 6-well plate and the next day cells were transfected with plasmids encoding for FZD4 or LRP5. Cells were transfected using Lipofectamine 3000 according to the manufacturer's instructions. 48 h after transfection, cells were harvested, washed with PBS and blocked with 1% BSA for 30 min on ice. Then, cells were stained with 30 nM of isotype control or F4L5.13 antibody for 1 h on ice. Cells were washed with PBS and stained with Alexa Fluor 488 Goat anti-human IgG, Fc fragment specific (Jackson ImmunoResearch #109-545-008) and incubated for 30 min at 4°C in the dark. Fixable viability dye eFluor 750 was added and incubated for 10 min, followed by fixation using 4% Paraformaldehyde (PFA). Cells were analyzed by flow cytometry using a Beckman Coulter Cytoflex S instrument (Beckman Coulter, Brea, CA, USA).
LRP5/6 ELISA assay LRP5 or LRP6 were coated onto maxisorp 384-well plates at 4°C overnight in PBS at a concentration of 1 lg/ml. Wells were blocked for 1 h at room temperature in PBS supplemented with 0.5% BSA and then washed 6 times in PBST (PBS + 0.05% Tween-20). A series of 8 dilutions of F4L5.13 or a non-targeting antibody (4275) ranging from 100 nM to 0.78 nM in PBST were added to the wells for 1 h at room temperature in a total volume of 30 ll. Unbound protein was removed by washing 6 times with PBST. Thirty microliters of a horseradish peroxidase-conjugated anti-human Fc antibody (Jackson Immuno Cat# 109-036-098) diluted in PBS + 0.5% BSA was added to each well for 30 min at room temperature, after which unbound protein was removed by washing times in PBST. ELISA wells were developed using TMB reagent (Mandel Cat# KP-50-76-03) and quenched with phosphoric acid. Absorbance at 450 nm was determined in an Epoch plate reader (BioTek). Data were analyzed using nonlinear regression in GraphPad Prism (v.9.0.0).

qRT-PCR
Total RNA was extracted from bEnd.3 cells using TRIzol reagent (Invitrogen; Thermo Fisher Scientific, Inc., Waltham, MA, USA) according to the manufacturer's protocol and quantified using a nanodrop. Total RNA (1-2 µg) was reverse-transcribed into cDNA using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Cat#4368813) or Maxima Reverse Transcriptase (Fermentas) according to the manufacturer's protocol. qPCR was performed using the Power SYBR Green PCR Master Mix (Applied Biosystems, Cat# 4368708) using the following thermocycling conditions: 95°C for 10 min, followed by 40 cycles of 95°C for 15 s, 60°C for 1 min, and 72°C for 30 s. Primer sequences indicated in Table EV1 were used, and gene expression was evaluated by the 2 -DDCT method.  (Bray et al, 2016), following default parameters. Differential expression analysis was performed using the DESeq2 R package (Love et al, 2014) using default parameters with the independent filtering option disabled, which thresholds of raw fold change > 1.5 and adjusted Pvalue < 0.1. Testing for enrichment of biological processes and pathways in differentially expressed genes was performed by querying the Gene Ontology, Reactome, and KEGG datasets using gProfiler (Auffray & De Meulder, 2016). A list of the top 20 most representative GO terms for F4L5.13 regulated genes was generated using the REVIGO webtool (Supek et al, 2011). Heat maps of genes mapping to indicate GO-BP or KEGG terms were generated using the pHeatmap R package.

Animal
All animal work was done in accordance with the Association for Research in Vision and Ophthalmology statement for the Use of Animals in Ophthalmic and Vision Research and with approval from the Institutional Animal Care and Use Committee at the University of Minnesota or Vanderbilt University Institutional Animal Care and Use Committee. Animals of both genders were used for all studies. The Tspan12 null allele (Tspan12 tm1.2Hjug ) was previously described . Mice were backcrossed to a C57BL/6J background for more than 10 generations. C57BL/6J mice were purchased from Jackson Labs. WT and Tspan12 À/À C57BL/6J mice of both sexes were used at the ages indicated in the manuscript text. All mice were housed in a specific pathogen-free facility with a 12 h light-dark cycle at room temperature.
Analysis of retinal whole mounts using confocal microscopy F4L5.13 or isotype control antibody (directed against Gaussia luciferase) were administered at 10 mg/kg, i.p., into P5 Tspan12 À/À pups or adult mice. Injections were repeated every 2-3 days until mice were sacrificed. For P20 mice, eyeballs were removed and fixed for 1 h in 4% PFA at RT. Eyeballs were washed three times with PBS, and retinas were dissected and blocked with 5% goat serum, 0.5% Triton X-100 Oxygen-induced retinopathy model C57/BL6 mouse pups were purchased from Charles River (Wilmington, MA). Pups were born and raised with dams in room air (RA; 20.9%) for 7 days (P0-P7), and then moved to a 75% oxygen atmosphere for 5 days (P8-P12). At P12, mice received intravitreal injections of PBS vehicle, mouse-specific anti-VEGF (delivered at 0.1 lg/ ll), or F4L5.13v2 at 50 or 500 nM target vitreous concentrations (0.5 ll injection of 250 and 2.5 lM solutions, respectively). Allocation of mice to treatment/control groups was randomized. Mice were anesthetized with isoflurane and treated with topical proparacaine before injection. Treatment and dosing were randomized between eyes, exposure chambers, and housing cages. Mice were sacrificed on P17 (five days after oxygen removal). Pups were not weighed prior to sacrifice, and however, the weigh, physiology, and behavior of the animals were subjectively observed and no abnormalities in pup health was noticed or attributed to a specific treatment group. Both normal, intraretinal vascular growth and pathological, preretinal neovascularization were assessed in isolectin-B4-stained retinal flat mounts via computer-assisted image analysis. Investigators were blinded during treatment and data collection. Resulting measurements were analyzed by ANOVA and Tukey's multiple comparison test. Data are reported as the ratios of neovascular area to total area (NV/TA) and avascular area to total area (AV/TA).

Statistics
Data were analyzed using GraphPad Prism software in at least three independent experiments. Graphs represent the mean AE SEM or AE SD. Two-tailed Student's t-test was used to compare two groups. Comparisons between multiple groups were made using one-way ANOVA followed by post hoc Bonferroni's multiple comparisons test among groups. P < 0.05 was considered statistically significant. Exact P-values are listed in Appendix Table S1.

Data availability
RNA-seq data have been deposited to Gene Expression Omnibus (accession number GSE166764) and are publicly available. All data are available in the main text or the supporting information. URL: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE166764.
Expanded View for this article is available online.

The paper explained Problem
Retinal homeostasis requires an intact blood-retina barrier (BRB) and BRB dysfunction is associated with retinal diseases. In addition, retinal homeostasis requires a proper balance of angiogenesis vs. vascular quiescence. In the retina, the secreted ligand Norrin (NDP) binds FZD4 and the co-receptors LRP5 and TSPAN12 and leads to bcatenindependent development of the retinal vascular and BRB formation. Mutations in genes important for this pathway lead to rare congenital eye diseases, such as Norrie disease and FEVR, in which the retinal vasculature and BRB is disrupted, causing impaired vision or blindness. More prevalent diseases such as diabetic retinopathy and macular degeneration are also characterized by neovascularization and blood-retina barrier defects. New therapeutic approaches are needed that actively restore BRB function and normalize aberrant retinal vasculatures.

Results
We have engineered F4L5.13, a tetravalent antibody, designed to induce FZD4 and LRP5 proximity in order to activate downstream bcatenin signaling. We demonstrated that it is selective for FZD4 and that it efficiently activates bcatenin signaling in mouse brain endothelial cells (bEnd.3 cells) to levels surpassing NDP. F4L5.13 mimics the activity of NDP by activating FZD4 and LRP5 but without requiring TSPAN12. F4L5.13 promotes barrier function in endothelial cells in vitro. In the Tspan12 À/À mouse model that exhibits several features of FEVR, systemic injection of F4L5.13 rescued the blood vessel morphogenesis and BRB defects. Finally, when tested in the oxygen-induced retinopathy model, which exhibits hallmarks of neovascular diseases such as retinopathy of prematurity and diabetic retinopathy, the FZD4:LRP5 antibody agonist was found to normalize the observed pathological neovascularization.

Impact
F4L5.13 is a novel synthetic tetravalent antibody that allows the precise (FZD4 and LRP5 specific) and potent activation of bcatenin signaling in vitro and in vivo. This antibody therefore represents a first-in-class antibody to treat ocular diseases as well as stroke and neurological disorders where endothelial cell barrier function is impaired.