Remote-Url: https://advances.sciencemag.org/content/7/34/eabh2307
Retrieved-at: 2021-08-19 18:00:17.152465+00:00

RESULTSWnt/PCP signaling components are essential for glutamatergic synapse maintenance and regulate synapse numbers in the adult nervous systemRecent studies showed that PCP signaling components are localized to glutamatergic synapses, interact with multiple key synaptic proteins, and function as essential regulators of synapse formation during early postnatal development (6). Biochemical fractionation showed that Celsr3 is localized on the plasma membrane of both pre- and postsynaptic sides in the synaptic cleft. Frizzled3 is localized on the presynaptic membrane, whereas Vangl2 is localized on the postsynaptic membrane (6). To test whether PCP signaling plays a role in the adult brain, we evaluated whether the PCP components are also present in the glutamatergic synapses of the adult hippocampus and localized in similar ways using confocal microscopy and super-resolution microscopy, the 3D STORM (fig. S1A andFig. 1, A and B) (20,21). We observed that the PCP components, Celsr3 and Vangl2, and the Wnt receptor, Ryk, are present in the adult glutamatergic synapses (at 2, 4, and 6 months of age). Super-resolution images showed that Celsr3 is colocalized with both the presynaptic marker (Bassoon) and the postsynaptic marker (PSD-95) (Fig. 1Band movie S1). Using the same strategies, Frizzled3 was shown colocalized with Bassoon, and Vangl2 was found colocalized with PSD-95 (Fig. 1Band movies S2 and S3). Ryk is a coreceptor of Frizzled and also interacts with Vangl2, regulating Vangl2 function. Using 3D STORM, we found that Ryk is present on both the pre- and postsynaptic sides (Fig. 1Band movie S4). Thus, the PCP components maintain their expression and display similar subcellular localization in both developing and adult glutamatergic synapses (Fig. 1C). The specificity of the antibodies used here has been confirmed using knockouts or cKOs in our previous studies (6,22–24).Fig. 1Localization of Wnt/PCP signaling components in glutamatergic synapses in adult hippocampus and role of Celsrs and Vangl2 in synapse maintenance.(A) Schematic diagram showing the hippocampal areas. (B) Representative 3D STORM images showing the localization of PCP proteins and Ryk (red) with synaptic markers. (C) Schematic of the distribution of Wnt/PCP components in glutamatergic synapses. mo, months. (D) Experimental design of viral injection. (EandF) Representative images and quantification of synaptic puncta in the stratum radiatum (SR) (circles indicate colocalization).n= 4 animals in each group. (GandH) Representative images and quantification of dendritic spine density. Thirteen dendrites from three control animals and 10 dendrites from four animals with Celsr2/3 sgRNAs. (I) Experimental design of viral injection. (JandK) Representative images and quantification of synaptic puncta in the SR.n= 5 control animals;n= 3 Vangl2 cKO animals. (LandM) Representative images and quantification of dendritic spine density. Twenty-one dendrites from three control animals and 17 dendrites from fourVangl2cKO animals. Student’sttest. *P< 0.05, **P< 0.01, and ***P< 0.001. Scale bars, 100 nm (B) and 1 μm (E, G, J, and L). Error bars represent SEM.To assess whether PCP signaling continues to regulate synapse maintenance, we first knocked outCelsrs2andCelsr3, which are the two abundantly expressed homologs in the adult brain, in adult hippocampus using the CRISPR-Cas9 system. CRISPR-induced genome editing was first tested in Neuro2A cells (fig. S2, A and B), and knockout efficiency was evaluated by Western blots of protein extracts from cultured hippocampal neurons (fig. S2C). We then injected both adeno-associated virus (AAV)–sgCelsr2-Cre and AAV-sgCelsr3-Cre into the CA1 region of the adult hippocampus of the Cre-dependentCas9mice and analyzed the synapse numbers first by costaining with synaptic markers 2 months later. We used Image J plugin called “synapse counter” to measure the puncta that are pre, post, or colocalized. When the pre and post are overlapped by 33 to 100%, they are considered colocalized. We found that knocking outCelsr2andCelsr3significantly reduced the glutamatergic synapse numbers (colocalized) in the stratum radiatum (Fig. 1, D to F). To assess the change of synapse number using a different approach, we then sparsely labeled the dendrites with a lower titer of the AAV–human synapsin (hSyn)–mCherry to visualize the dendritic spines. Knocking outCelsr2andCelsr3significantly reduced the number of the dendritic spines of the CA1 pyramidal neurons (Fig. 1, G and H). We next conditionally knocked outVangl2by injecting AAV1-hSyn–enhanced green fluorescent protein (eGFP)–Cre (AAV that harbors the hSyn promotor driving the Cre recombinase) to the CA1 region of the adult hippocampus (at the age of 2 months) of theVangl2fl/flmice (fig. S2D) and then analyzed the synapse numbers 2 months later (Fig. 1I). We observed a 27% increase of synapse numbers inVangl2cKO by constaining with synaptic markers (Fig. 1, J and K). Dendrites were labeled sparsely by coinjecting AAV1-hSyn-Cre (at a lower titer) and AAV1-CAG-Flex-eGFP into the Vangl2 cKO. We found that conditionally knocking outVangl2led to an increase of dendritic spines of the CA1 pyramidal neurons 2 months later (Fig. 1, L and M), suggesting that similar to development, Vangl2 also negatively regulates synapse numbers in adult mice. Therefore, the functions of Celsrs and Vangl2 extend beyond synapse formation and they continue to regulate synapse numbers in adulthood.Vangl2 is required for Aβ oligomer–induced synapse loss in vitro and in vivoBecause PCP signaling components control the maintenance of a large number of glutamatergic synapses in adulthood, we then hypothesized that PCP signaling components may be involved in synapse loss in neurodegeneration. Oligomeric amyloid-β (Aβ oligomer) is a well-established model for synapse degeneration. We first tested whether Vangl2 is required for Aβ oligomer–induced synapse loss in cultured hippocampal neurons isolated from embryonic day 18 (E18.5) ofVangl2fl/flmice. AAV1-hSyn-eGFP-Cre was added into the culture on the seventh day after the start of culture [day in vitro 7 (DIV7)]. At DIV14, 400 nM monomer equivalent of Aβ oligomers were added, and the cultures were fixed 12 hours later for staining and analysis (Fig. 2A). Figure S3 shows that the effective concentration of the dimers added was 80 nM and that the effective concentration of the tetramers was ~152 nM. The calculation of concentration is described in Materials and Methods (25,26). Neurons isolated from littermate control (Vangl2+/+, control) mice were also treated with AAV1-hSyn-eGFP-Cre. Vangl2 protein levels were successfully reduced inVangl2fl/flneurons treated with the AAV1-hSyn-eGFP-Cre (Fig. 2B). We found that, in control neurons, Aβ oligomers reduced the synapse numbers by 30% as shown by costaining with the Bassoon and PSD-95 antibodies (Fig. 2, C and D). However, the number of glutamatergic synapses inVangl2cKO neurons was unchanged when treated with Aβ oligomers (Fig. 2, C and D). Consistent with our previous finding that Vangl2 inhibits synapse formation,Vangl2cKO itself led to 40% increase of synapse numbers during the 7.5 days of culture of the E18.5 embryonic neurons, suggesting that 1 week was long enough time to observe the increased synapse formation in these cultures from E18.5 neuronal cultures (Fig. 2, C and D) (6).Fig. 2Vangl2 is required for Aβ oligomer–induced synapse loss in vitro and in vivo.(A) Schematic illustrating the experimental design.AAV1-hSyn-eGFP-Crevirus was added to hippocampal neuron cultures on DIV7 for 7 days and then oligomeric Aβ42 was added. Aβ-O, Aβ oligomer. Twelve hours later cultures were fixed for staining with synaptic markers. (B) Western blot showing the level of Vangl2 proteins in cultures infected with theAAV1-hSyn-eGFP-Crevirus. GAPDH, glyceraldehyde-3-phosphate dehydrogenase. (C) Immunostaining for presynaptic (green) and postsynaptic (red) puncta of glutamatergic synapses (arrowheads) in 14-DIV hippocampal cultures from littermateVangl2+/+orVangl2fl/flwith or without oligomeric Aβ42. (D) Quantification of (C).n= 3 forVangl2+/+mice,n= 4 forVangl2fl/flfrom three independent experiments. (E) Schematic illustrating the experimental design.AAV1-hSyn-eGFP-Crevirus was injected into the CA1 region of the hippocampus bilaterally. Two weeks later, Aβ oligomeric was injected intracerebroventricularly (i.c.v.). Five days after Aβ oligomer injection, animals were fixed with perfusion and sectioning and stained with synaptic markers. (F) Vangl2 protein level in the total hippocampal extracts from animals injected with theAAV1-hSyn-eGFP-Crevirus. (G) Representative images of Bassoon (red)– and PSD-95 (green)–immunoreactive puncta (arrowheads) in the stratum radiatum ofVangl2+/+andVangl2fl/flhippocampus (CA1) with or without Aβ oligomeric injection and quantification of synapse numbers. (H) Quantification of (G). One-way ANOVA.n= 8 ofVangl2+/+mice,n= 3 ofVangl2+/+mice with Aβ oligomeric injection,n= 6 ofVangl2fl/flmice, andn= 5 ofVangl2fl/flmice with Aβ oligomeric injection. *P< 0.05 and **P< 0.01. One-way ANOVA. Scale bars, 2 μm (C and G). Error bars represent SEM.To test whether Vangl2 is required for Aβ oligomer–induced synapse loss in vivo in adulthood, we injected AAV1-hSyn-eGFP-Cre into the hippocampal CA1 region ofVangl2+/+(control) andVangl2fl/flmice at the age of 2 months. Two weeks later, we bilaterally injected 5 ng of Aβ oligomers into the cerebral ventricles (Fig. 2E) (27). Five days after injection of Aβ oligomers, animals were perfused, and the synapse numbers were analyzed. Vangl2 protein levels were found significantly reduced inVangl2cKO (Fig. 2F). We observed a significant loss of synapses in control animals injected with Aβ oligomers but not in theVangl2cKO mice injected with Aβ oligomers (Fig. 2, G and H). In contrast to the experiment with cultured neurons from E18.5 embryos,Vangl2cKO in adulthood showed no significant changes in synapse numbers 19 days after the AAV1-hSyn-eGFP-Cre injection (Fig. 2, G and H). It is probably because the synapse turnover is not as rapid in adulthood such that 19 days is not long enough to observe changes in synapse numbers.Binding of Aβ oligomers to Celsr3 is required for Aβ oligomer–induced synapse lossTo understand how PCP signaling may mediate Aβ oligomer–induced synapse loss, we investigated a potential direct interaction between Aβ oligomers and the PCP components. PCP components are localized in glutamatergic synapses in a similar fashion to the asymmetric epithelial cell-cell junctions during PCP signaling (Fig. 1, B and C) (6). Among the six core PCP components, Celsrs, Frizzleds, and Vangls are present on the plasma membrane. To determine whether Aβ oligomers target any one of those proteins, we measured binding of biotin-Aβ42 oligomers to human embryonic kidney (HEK) 293T (HEK293T) cells expressing mouse Vangl2 (Vangl2-Flag), Frizzled3 (Frizzled3-HA), Celsr3 (Celsr3-Flag), or control vector (pCAGEN). We found that Aβ oligomers only bound to Celsr3 but not to Vangl2 or Frizzled3 (Fig. 3A), with an apparent dissociation constant (Kd) of ~40 nM equivalent of total Aβ peptide (Fig. 3B). Aβ monomers did not bind to Celsr3 (fig. S4).Fig. 3Celsr3 is a binding protein for Aβ oligomeric, and the binding of Aβ oligomeric to the EGF7 and EGF8 domains of Celsr3 is required for synapse loss.(A) Staining of the HEK293T cells transfected with Vangl2-Flag, Frizzled3-HA, Celsr3-Flag, or control vector (pCAGEN) and incubated with oligomeric Aβ42 (200 nM total peptide, monomer equivalent). Bound oligomeric Aβ42 (green) was visualized using Alexa-488-conjugated streptavidin. (B) Binding curve of oligomeric Aβ42 to Celsr3-expressing HEK293T cells (concentration shown as total peptide, monomer equivalent). (C) Schematic of mouse Celsr3 extracellular domain with nine cadherin domains, eight EGF domains, and three laminin domains. (D) Staining of bound Aβ42 (200 nM total peptide, monomer equivalent) with Celsr3-Flag–transfected or truncated Celsr3-Flag–transfected HEK293T cells. Bound oligomeric Aβ42 (green) was visualized using Alexa-488-conjugated streptavidin. Scale bar, 10 μm. (EandF) Immunoprecipitated assays testing the interaction between Frizzled3 and Celsr3 or with truncated Celsr3 and quantification data of the expression level of co-IPed Frizzled3. A.U., arbitrary units. *P< 0.05. One-way ANOVA. Means ± SEM. (G) Pull-down assay for purified GST function proteins mixed with biotinylated Aβ42 oligomers by streptavidin. (HandI) Aβ oligomer–induced synapse loss in the presence of EGF7-GST or EGF8-GST fusion proteins. Arrowheads indicate the Bassoon/PSD95 colocalized puncta. *P< 0.05 and **P< 0.01. Student’sttest. Scale bars, 10 μm (A and D) and 1 μm (H). Error bars represent SEM.Celsr3 belongs to the family of adhesion G protein–coupled receptors with a large extracellular region, which contains nine cadherin domains, eight EGF repeats, and three laminin domains (Fig. 3C). Cadherin domains are required for homophilic binding of Celsr3. To determine the domain(s) of Celsr3 that bind to Aβ oligomers, we first made a deletion construct that lacks all the EGF repeats and laminin domains and tested binding in HEK293T cells. We found that Aβ oligomers did not bind to this truncated protein, suggesting that Aβ oligomers do not bind to the cadherin domains but rather bind to the EGF repeats and the laminin domains (Fig. 3Dand fig. S5A). We then made a series of Celsr3 constructs that lack the individual EGF and laminin domains (fig. S5B) and tested for binding to Aβ oligomers in HEK293T cells. We found that two EGF domains, EGF7 and EGF8, and one laminin domain, laminin G1, are required for binding of Aβ oligomers (Fig. 3Dand fig. S5C). The human homolog of murine Celsr3 also contains nine cadherin domains, eight EGF repeats, and three laminin domains. The laminin G1 and EGF7 domains ofhCelsr3 align closely with that ofmCelsr3 with a homology of 98.537 and 80%, respectively. The amino acid sequence of the EGF8 domain ofhCelsr3 is 100% homologous with that of the EGF8 domain ofmCelsr3 (fig. S6A). We found that Aβ oligomers also bound tohCelsr3 with an apparentKdof ~70 nM equivalent of total Aβ peptide (fig. S6B). Similar tomCelsr3, EGF7 and EGF8 and laminin G1 ofhCelsr3 are required for binding with Aβ oligomers (fig. S6B).In PCP signaling, protein-protein interaction is essential for the establishment of cell and tissue polarity along the tissue plane (2). Celsr3 forms a complex with Frizzled3 on the plasma membrane of one cell, and Celsr3 forms a complex with Vangl2 on the plasma membrane of the neighboring cells. We then tested whether the Aβ oligomer–binding domains of Celsr3 are involved in the protein-protein interactions among PCP components. We expressed Frizzled3 or Vangl2 together with WT Celsr3 or mutant Celsr3 (with domain deletions) in HEK293T cells. We found that deletion of all eight EGF repeats and three laminin domains caused a 68% reduction of the interaction between Celsr3 and Frizzled3. Deletion of laminin G1 alone led to a 66% reduction of the interaction between Celsr3 and Frizzled3, whereas deleting EGF7 or EGF8 did not affect the interaction between Celsr3 and Frizzled3 (Fig. 3, E and F). The interaction between Vangl2 and Celsr3 did not require the EGF repeats or the laminin domains (fig. S7). To determine whether the binding of Aβ oligomers on these domains is important for synapse loss, we tested the role of EGF7 and EGF8 domains. We performed binding assays with purified EGF7–glutathioneS-transferase (GST) or EGF8-GST fusion proteins and found that both domains can bind to Aβ oligomers, as pulling down biotinylated Aβ oligomers with streptavidin (NeutrAvidin agarose) can pull down the EGF7-GST and EGF8-GST fusion proteins (Fig. 3G). Last, we added EGF7-GST or EGF8-GST fusion proteins to neuronal hippocampal culture and found that they both blocked Aβ oligomer–induced synapse loss (Fig. 3, H and I). Therefore, the EGF7 and EGF8 domains of Celsr3 are likely direct targets of Aβ oligomers.Aβ oligomers enhance the function of Vangl2 in disrupting the Celsr3/Frizzled3-Celsr3 intercellular complex essential for PCP signalingPCP signaling is known to be mediated by a set of dynamic protein-protein interactions (2). One such essential interaction is an asymmetric intercellular complex made of Frizzled and Celsr (17). The Frizzled/Celsr complex on the plasma membrane on the distal side of a cell forms an intercellular complex with Celsr on the plasma membrane on the proximal side of the neighboring cell (distal to the first cell), bridging the two cells via the homophilic interaction of the cadherin repeats of the two Celsr proteins. Such an asymmetric intercellular bridge is sufficient to polarize both cells even in the absence of Vangl and, therefore, is thought to be essential for PCP signaling (17).Because the Aβ oligomers bind to the laminin G1 domain of Celsr3, which is also required for Frizzled3 binding, we hypothesized that Aβ oligomers may weaken the interaction between Celsr3 and Frizzled3, allowing Vangl2 to more efficiently disrupt the asymmetric intercellular complex of Celsr3/Frizzled3-Celsr3 and thus disassemble synapses. We established an assay to test the intercellular PCP complex, similar to the “transcellular interaction assay” that we published previously (28). We cotransfected Frizzled3 (HA-tagged) and Celsr3 (untagged) in one dish of HEK293T cells and transfected Celsr3 (Flag-tagged) and Vangl2 in another. After culturing them separately for one day, we mixed them together and cultured for one more day and then performed coimmunoprecipitation to test the protein-protein interactions (Fig. 4A). To test whether Vangl2 disrupts the intercellular bridge, we pulled down Frizzled3 and measured how much Celsr3 from the neighboring cell was coimmunoprecipitated. We found that Vangl2 disrupted this intercellular complex as much less Flag-tagged Celsr3 was pulled down by HA-tagged Frizzled3 (fig. S8, A to C). We think that Vangl2 is likely disrupting this intercellular complex by weakening the interaction between Ceslr3 and Frizzled3, because the presence of Vangl2 alone from the neighboring cell caused the reduction of the interaction between Frizzled3 and Celsr3 by 30 to 40% (fig. S8, D to F). Celsr3 from the neighboring cell does not affect the complex between Frizzled3 and Celsr3 (fig. S8, G to I).Fig. 4Aβ oligomer enhances Vangl2’s function in disrupting the intercellular complex of Celsr3/Frizzled3-Celsr3.(A) Schematic illustrating the experimental design of the “transcellular interaction” assay. (B) Testing the intercellular complex by immunoprecipitating Celsr3 in one cell and detecting Frizzled3 in the neighboring cell with or without Vangl2. (C) Quantification of (B). (D) Representative 3D STORM images showing the endogenous Celsr3 (red), Frizzled3 (green), and Bassoon (blue) on the secondary dendrite of the cultured hippocampal neuron. (d1tod3) Representative 3D STORM images of synapses in the cultured hippocampal neuron that were labeled for Celsr3 (red), Frizzled3 (green), and Bassoon (blue). (E) The distribution of localization points along the trans-synaptic axis was measured and fit with Gaussian functions, and the distances between the centroid positions of the two Gaussians were defined as the distances between the secondary antibodies detecting Celsr3 and Frizzled3 antibodies, which reflect the distances between Celsr3 and Frizzled3. (F) The histogram of the distances between the secondary antibodies detecting Celsr3 and Frizzled3 antibodies, which reflect the distances between Celsr3 and Frizzled3. (G) The distribution of localization points along the trans-synaptic axis was measured and fit with Gaussian functions, and the distances between the centroid positions of the two Gaussians were defined as the distances between the secondary antibodies detecting Celsr3 and Frizzled3 antibodies with different concentration (monomer equivalent) of Aβ oligomer. (H) The histogram of the distances with different concentration (monomer equivalent) of Aβ oligomer between the secondary antibodies detecting Celsr3 and Frizzled3 antibodies, which reflect the distances between Celsr3 and Frizzled3. *P< 0.05 and ***P< 0.001. One-way ANOVA. Western blot results are representative of four biological replicates. Scale bars, 1 μm (D) and 100 nm (d1 to d3). Error bars represent SEM.To determine how Aβ oligomers cause synapse loss, we tested whether Aβ oligomers facilitate the function of Vangl2 in disrupting the intercellular complex. First, we found that Aβ oligomers did not disrupt the interaction between Celsr3 and Frizzled3 that were transfected and expressed in the same cell, suggesting that Aβ oligomers themselves are not sufficient to disrupt the Ceslr3-Frizzled3 complex (fig. S9, A to C). As shown previously, Vangl2 expressed in the neighboring cell can decrease the interaction between Celsr3 and Frizzled3 (fig. S8, D to F). The interaction between Frizzled3 and Celsr3 was reduced to a greater extent when Aβ oligomers were added to the culture (fig. S9, D to F), indicating that Aβ oligomers may enhance the function of Vangl2 in disrupting the Celsr3-Frizzled3 complex across the cell-cell junction. Furthermore, we found that Aβ oligomers can also disrupt the intercellular complex, as the HA-tagged Frizzled3 in one cell pulled down significantly less Flag-tagged Celsr3 from a neighboring cell when Aβ oligomers were added to the mixed culture (fig. S9, G to I). This suggests that the intercellular interaction of Celsr3 between two neighboring cells may weaken the intracellular interaction between Celsr3 and Frizzled3 within the same cell, allowing Aβ oligomers to more efficiently disrupt the entire intracellular interaction. To test the role of Aβ oligomers in enhancing the function of Vangl2, we reduced the amount of transfected Vangl2 by half (at 0.5 μg per well) so that Vangl2 is at suboptimal concentration (Fig. 4, A to C). We found that adding Aβ oligomers to mixed cultures with Vangl2 led to the greater disruption of this intercellular complex (Fig. 4, A to C).To further test our model that Aβ oligomers may promote synapse disassembly by interfering with the interactions between Celsr3 and Frizzled3, we performed super-resolution microscopy and characterized the distance between Celsr3 and Frizzled3 on cultured hippocampal neurons using 3D STORM. We labeled Celsr3, Frizzled3, and a presynaptic marker Bassoon using the corresponding primary antibodies and secondary antibodies conjugated with fluorophore and collected the STORM signals in the synapses on the secondary dendrite (Fig. 4D). We observed three types of colocalization: Celsr3 and Frizzeld3 colocalized with Bassoon (Fig. 4D, d1), Celsr3 colocalized with Bassoon (Fig. 4D, d2), and Frizzled3 colocalized with Bassoon (Fig. 4D, d3). We then rotated the three-dimensional (3D) images so that Celsr3 and Frizzled3 are axially distributed and measured the distance between the centroids of the two Gaussian areas (Fig. 4D, d1 to d3) (21). We found that the distance between the signals that represent Celsr3 and Frizzled3 was 181.2 ± 5.2 (±SEM) nm in the control group (192 pairs quantified in control group). Thirty minutes after the treatment of Aβ oligomers (400 nM), the distance became larger (236.5 ± 5.9 nm, 174 pairs). One hour after the treatment, the distance was 245.8 ± 5.4 nm (179 pairs). Two hours after the treatment, it became shorter (213.1 ± 5.3 nm, 177 pairs) (Fig. 4, E and F). These distances represent the two secondary antibodies (with a size of ~15 nm of each antibody) that correspond to Celsr3 and Frizzled3 and, therefore, are greater than the actual distances between Celsr3 and Frizzled3. Nonetheless, they do reflect the distances between Celsr3 and Frizzled3. In addition, we observed a dose response. The Celsr3-Frizzled3 distance was slightly increased when the neurons were treated with 100 nM Aβ oligomers for 1 hour (224.3 ± 3.4 nm, 50 pairs) and further increased when the neurons were treated with 400 nM Aβ oligomers for 1 hour (256.3 ± 5.3 nm, 29 pairs). The distance became shorter with 800 nM Aβ oligomer treatment (238.9 ± 4.3 nm, 38 pairs) (Fig. 4, G and H). The increased distance may represent the change of Celsr3 and Frizzled3 interaction with the synapses. It may also reflect the change of the histogram of distances because the disassembly of synapses increases the average distances. Future experiments will be needed to distinguish these possibilities. The fact that the average distance of Celsr3 and Frizzled3 is changed within 30 min after the addition of Aβ oligomers supports our model that Aβ oligomers directly target the PCP proteins in the synapses and disrupt the intercellular complex of Celsr3/Frizzled3-Celsr3 by interfering with the interactions between Celsr3 and Frizzled3.The Wnt/Ryk/Vangl2 signaling axis mediates Aβ oligomer–induced synapse loss in vitro and in vivoTo further address how Vangl2 mediates Aβ oligomer–induced synapse loss and identifies potential therapeutic targets to protect synapses, we explored regulators of core PCP signaling components. As Ryk is a coreceptor for Wnt in PCP signaling by directly interacting with Vangl2 and promoting the function of Vangl2, we sought to test whether Wnt/Ryk signaling is involved in Vangl2 function in this context (18,19). Wnt5a causes a reduction of synapse numbers probably by causing Frizzled3 endocytosis (6,29). We first tested whether Ryk mediates Wnt5a function in regulating synapse numbers and whether Ryk does so in a Vangl2-dependent manner. Hippocampal neurons isolated from E18.5 WT mice were treated with Wnt5a on DIV14 for 12 hours or pretreated with a function blocking monoclonal Ryk antibody, which blocks the binding between Wnts and Ryk, for 2 hours (Fig. 5A) (24). Normal mouse immunoglobulin G (IgG) was used as control. Wnt5a caused a 30% reduction in the number of colocalized puncta. In contrast, Wnt5a did not produce a significant difference in synapse numbers when the cultures were pretreated with the function-blocking Ryk monoclonal antibody (Fig. 5, A and B), suggesting that Wnt5a inhibits synapse formation through binding to Ryk. During the 14 hours of culture, Ryk antibody itself did not cause any significant changes in synapse numbers (Fig. 5B). To test whether Vangl2 mediates the function of Wnts downstream of Ryk, we cultured embryonic hippocampal neurons fromVangl2+/+(control) andVangl2cKO (transduced with AAV1-hSyn-eGFP-Cre) and treated neurons with Wnt5a on DIV14 for 12 hours. We found that Wnt5a addition toVangl2+/+caused a 30% reduction in the number of colocalized puncta, whereas Wnt5a addition toVangl2cKO neurons did not produce a significant difference compared with untreatedVangl2cKO neurons (Fig. 5, C and D), suggesting that Vangl2 is required for the inhibitory function of Wnt5a in synapse formation. Consistent with our previous finding (Fig. 2, C and D),Vangl2cKO itself led to more synapses in the cultured embryonic neurons 7.5 days after the introduction of Cre (Fig. 5D). Finally, we pretreated cultured hippocampal neurons with the function-blocking anti-Ryk monoclonal antibody for 2 hours before adding the Aβ oligomers (24). We found that Aβ oligomers failed to reduce synapse numbers in the presence of the Ryk antibody (Fig. 5, E and F). Again, during the 14 hours of time, Ryk antibody itself did not cause any change in synapse numbers in these cultures (Fig. 5F). Therefore, Ryk, together with Vangl2, is also required for Aβ oligomer–induced loss of glutamatergic synapses. As Aβ oligomers do not bind to Ryk (fig. S10), we propose that Wnt/Ryk and Vangl2 function as one signaling axis that is required for synapse disassembly and Aβ oligomers enhance their function in disrupting the synapses.Fig. 5The Wnt/Ryk/Vangl2 signaling axis mediates synapse loss induced by Aβ oligomeric.(AandB) Representative images and quantification of synaptic puncta (arrowheads) testing the effect of the Ryk antibody on Wnt5a-induced synapse reduction in cultured hippocampal neurons.n= 3 experiments (n= 27 neurons in IgG control,n= 22 neurons in Ryk antibody,n= 24 in IgG + Wnt5a, andn= 20 neurons in Ryk antibody + Wnt5a). (CandD) Representative images and quantification of synaptic puncta (arrowheads) testing the role of Vangl2 in Wnt5a-induced synapse reduction in cultured hippocampal neurons.n= 3 WT mice andn= 4Vangl2 cKOmice. (EandF) Representative images and quantification of synaptic puncta (arrowheads) testing the effect of the Ryk antibody in oligomeric Aβ–induced synapse reduction in cultured hippocampal neurons.n= 3 experiments (n= 26 neurons in IgG control,n= 33 neurons in Ryk antibody,n= 34 neurons in Aβ oligomeric, andn= 39 neurons in Ryk antibody + Aβ oligomeric). (G) Schematic illustrating the experimental design.AAV1-hSyn-eGFP-Crevirus was injected into the CA1 and CA3 regions of the hippocampus bilaterally. Two weeks later, Aβ oligomers were injected into the lateral ventricle for 5 days. Five days later, animals were fixed with perfusion and sectioning and stained with synaptic markers. (HandI) Representative images and quantification of synaptic puncta detected by costaining for Bassoon (red)– and PSD-95 (green)–immunoreactive puncta (arrowheads) in the stratum radiatum.n= 4 forRyk+/+mice,n= 3 forRyk+/+mice with Aβ oligomeric injection,n= 3 forRykcKO mice, andn= 3 forRykcKO mice with Aβ oligomeric injection. *P< 0.05, **P< 0.01, and ***P< 0.001. One-way ANOVA. Scale bars, 1 μm. Error bars represent SEM.To further test the role of Ryk in vivo, we injected of Aβ oligomers into theRykcKO mice previously generated in our laboratory and analyzed synapse numbers (24). Because Ryk is likely involved on both the presynaptic and the postsynaptic side, AAV1-hSyn-eGFP-Cre was injected into the CA1 and CA3 regions of the hippocampus of 8-week-old mice to conditionally knock outRykin adulthood. InRyk+/+(control), Aβ oligomers led to a 60% reduction of synapse numbers. However, the synapse numbers were not affected in theRykcKO mice that received intracerebroventricular injection of Aβ oligomers (Fig. 5, G to I).RykcKO prevents the loss of synapses and preserves cognitive function in the 5XFAD miceTo determine whether Wnt-Ryk signaling is required for Aβ oligomer–induced synapse loss in a mouse model of Aβ deposition, we crossed theRykcKO with the 5XFAD transgenic mouse. PCP components are present in the glutamatergic synapses of the adult 5XFAD animals (fig. S1B). AAV1-hSyn-eGFP-Cre was injected into the hippocampal CA1 and CA3 regions of the hippocampus of 8-week-old mice to conditionally knock outRykin adulthood. Glutamatergic synapse numbers were analyzed 2 months later (Fig. 6A). We found thatRykcKO in hippocampal neurons prevented the synapse loss (Fig. 6, B and C). We stained for Celsr3 and found that the protected synapses are Celsr3-positive (Fig. 6D).RykcKO itself showed a statistically insignificant (P= 0.107) trend toward increases in synapse numbers 2 months after the cKO. Compared to the 5XFAD animals,RykcKO resulted in a 400% increase of synapse numbers in 5XFAD, preserving the synapse numbers to the control level (Fig. 6, B and C). To validate the protection of synapses, we analyzed spine density in vivo using sparse labeling of dendrites in 4-month-old mice with a virus mixture of AAV1-hSyn-Cre (at a lower titer) and AAV1-CAG-Flex-eGFP (Fig. 6E). The 5XFAD mice showed a decreased spine density compared with age-matched control animals.RykcKO itself led to no significant difference in spine density, whereasRykcKO preserved the dendritic spines in the 5XFAD animals (Fig. 6, F and G).Fig. 6RykcKO prevented synapse loss and preserved cognitive function in the 5XFAD mice.(A) Schematic of timeline and experimental design.AAV1-hSyn-eGFP-Crevirus was injected into the CA1 and CA3 region of the hippocampus bilaterally. Animals were fixed with perfusion and sectioning and stained with synaptic markers at 4 months of age. A separate set of animals were injected and subjected to NOR at 6 months of age. (BandC) Representative images and quantification of synapse numbers. (D) Quantification of Celsr3-positive glutamatergic synapse (Celsr3 colocalized with PSD-95 and bassoon). (E) Schematic illustrating the experimental design for sparse labeling of neurons to visualize dendritic spines. (FandG) Representative images and quantification of dendritic spine density. Nineteen dendrites from four animals in the control group, 15 dendrites from three animals in theRykcKO group, 19 dendrites from three animals in the 5XFAD group, and 17 dendrites from four animals in the 5XFAD/RykcKO group. (H) Schematic showing the design of novel object recognition (NOR). Mice were subjected to an open arena for three trails to evaluate the memory of objects. (I) Trajectories of mice in the NOR test session. (J) Quantification of locomotion. (K) Quantification of total time for exploration of the two objects. (L) Quantification of NOR. *P< 0.05, **P< 0.01, and ***P< 0.001. One-way ANOVA. Scale bars, 1 μm. Error bars represent SEM.To test whetherRykcKO can preserve cognitive function, we performed the novel object recognition (NOR) test with a new set of mice at 6 months of age (Fig. 6A). NOR tests were carried out in an open-field arena measuring 0.4 m by 0.4 m by 0.45 m (Fig. 6H). The total distance was recorded during the 5-min open-field session as locomotor activity; no differences were found among the groups (Fig. 6J). The total time spent exploring the familiar and novel objects was also measured; no differences were found among the groups (Fig. 6K). We found that theRykcKO itself did not cause any behavioral defects, but it rescued the impaired NOR behavior of the 5XFAD mice (Fig. 6, I and L). We asked whether theRykcKO in neurons affected the development of amyloid plaques in 5XFAD mice and found that the number and total areas of plaques were unchanged at the matching 4 months of age when we quantified synapse numbers (fig. S11).Ryk monoclonal antibody infusion prevents the loss of synapses and preserves cognitive function in the 5XFAD miceTo investigate whether Wnt/Ryk/PCP signaling can be a potential therapeutic target, we tested the aforementioned function-blocking anti-Ryk monoclonal antibody by intracerebroventricular infusion into the 8-week-old 5XFAD mice (Fig. 7A) (24). Intracerebroventricular administration is a reliable way to deliver antibodies into the brain parenchyma (30). The purified Ryk monoclonal antibody (at 1 mg/ml in 100 μl of artificial cerebrospinal fluid) was loaded in a subcutaneously implanted osmotic minipump connected to a stainless steel cannula. The cannula was stereotaxically implanted into the lateral ventricle, and the infusion was performed at a flow rate of 0.25 μl/hour for 14 days. Therefore, each animal received a total of 84 μg of the purified Ryk monoclonal antibody. Synapse numbers were then analyzed at the age of 4 months, and behavioral tests were performed at the age of 6 months (with a different group of animals). We found that the Ryk monoclonal antibody infusion was able to block the loss of glutamatergic synapse numbers in the hippocampus of the adult 5XFAD transgenic mice (Fig. 7, B and C), as measured by immunostaining with synaptic markers and the density of the dendritic spines (Fig. 7, D and E). We infused a separate cohort of animals with the Ryk monoclonal antibodies, and LTP of the protected CA3-CA1 synapses was assessed at the age of 4 to 5 months. We found that Ryk monoclonal antibody infusion preserved LTP in the 5XFAD animals (Fig 7, F and G). In the NOR test with another cohort of animals (at 6 months of age), Ryk monoclonal antibody infusion preserved the cognitive function of the 5XFAD mice (Fig. 7, H to K). We asked whether the Ryk monoclonal antibody infusion affected the development of the amyloid plaques in 5XFAD mice and found that the number and total areas of plaques were unchanged at the matching 4 months of age (Fig. 7, L and M). These results are consistent with the observation that amyloid plaques were unchanged in 5XFAD when Ryk was conditionally knocked out (fig. S11). Astrogliosis and microglia activation are observed in the 5XFAD mice and other Aβ models (27,31). We then investigate whether Ryk monoclonal antibody could affect astrogliosis or microgliosis in the 5XFAD mice. We observed a significant increase in glial fibrillary acidic protein (GFAP)+and Iba-1+(Ionized calcium binding adaptor molecule 1) area coverage in the mouse IgG-infused 5XFAD mice. Ryk monoclonal antibody reduced the increase of both in the 5XFAD mice (Fig. 7, N and O). We currently cannot distinguish whether the reduction of astrogliosis or microgliosis is a secondary effect due to the protection of synapses in neurons or a direct effect of the Ryk antibody on astrocytes or microglia. However, the reduced astrogliosis or microgliosis is consistent with the preservation of the cognitive function and supportive of the Ryk antibody as a potential treatment option. To confirm that the Ryk monoclonal antibody is effectively delivered using intracerebroventricular administration, we also stained these tissues using a secondary antibody against mouse IgG and found that the mouse IgG signals can be detected in the animals infused with Ryk monoclonal antibody but not in those animals infused with control mouse IgG (fig. S12). Some signals appear to be distributed on neurites, most likely dendrites, suggesting that the Ryk monoclonal antibody diffused into the hippocampal parenchyma and became enriched on the dendrites where Ryk protein and the synapses are located. In addition to the signals on neurites, the signals appear to have some colocalization/close apposition with Iba-1 immunoreactivity and GFAP immunoreactivity (potentially more with Iba-1 than with GFAP). However, with the current resolution, we are not able to be certain about the exact distribution of the Ryk monoclonal antibody. These observations motivated us to use higher resolution and a more complete set of markers in our future studies to explore whether Ryk monoclonal antibody is engaging the glutamatergic synapses or the processes of the microglia or astrocytes or both.Fig. 7Function blocking Ryk antibody can prevent synapse loss and preserve cognitive function in the 5XFAD mice.(A) Timeline and schematic of intracerebroventricular infusion of the Ryk monoclonal antibody. Cannula and preinfused osmotic minipumps were implanted at ~2 months old. Minipumps were removed 2 weeks later. Synapse number analysis was done at the age of 4 months and NOR was done at the age of 6 months. (BandC) Staining and quantification of glutamatergic synapses in the 5XFAD mice infused with the Ryk monoclonal antibody. (DandE) Representative images and quantification of dendritic spine density. Twenty-one dendrites from four animals in the control_mIgG group, 22 dendrites from four animals in the 5XFAD_mIgG group, and 23 dendrites from four animals in the 5XFAD_Ryk mAb group. One-way ANOVA. (F) Average traces for field excitatory postsynaptic potentials (fEPSPs) in hippocampal slices obtained from four animals in the control_mIgG group, five animals in the 5XFAD_mIgG group, and six animals in the 5XFAD_Ryk mAb group. (G) Mean changes in the EPSP amplitude between 45 and 60 min after HFS (high-frequency stimulation) protocol. (H) Trajectories of mice in the NOR test session. (I) Quantification of locomotion. (J) Quantification of total time for exploration of the two objects. (K) Quantification of NOR. (LandM) Representative images of hippocampi stained for Aβ using antibody 6E10 and quantification of number of Aβ-immunopositive deposits and total Aβ-immunostained area.n= 4 animals in the 5XFAD_mIgG group andn= 6 animals in the 5XFAD_Ryk mAb group. Student’sttest. (N) Representative images of hippocampi stained for GFAP+astrocytes and (O) Iba-1+microglia, and quantification of GFAP+and Iba-1+area coverage.n= 4 to 6 animals per group. (P) Schematic diagram showing the balance of Wnt/PCP signaling in synapse maintenance and the binding site of Aβ oligomeric. *P< 0.05, **P< 0.01, and ***P< 0.001. Scale bars, 1 μm (B and D), 100 μm (L, N, and O), and 10 μm (N and O insets). Error bars represent SEM.