S-SCAM inhibits Axin-dependent synaptic function of GSK3β in a sex-dependent manner

S-SCAM/MAGI-2 gene duplication is associated with schizophrenia (SCZ). S-SCAM overexpression in the forebrain induces SCZ-like phenotypes in a transgenic (Tg) mouse model. Interestingly, S-SCAM Tg mice show male-specific impairments in synaptic plasticity and working memory. However, mechanisms underlying the sex-specific deficits remain unknown. Here we report that S-SCAM Tg mice have male-specific deficits in synaptic GSK3β functions, as shown by reduced synaptic protein levels and increased inhibitory phosphorylation of GSK3β. This GSK3β hyper-phosphorylation was associated with increased CaMKII activities. Notably, synaptic levels of Axin1, to which GSK3β binds in competition with S-SCAM, were also reduced in male S-SCAM Tg mice. We demonstrated that Axin-binding is required for the S-SCAM overexpression-induced synaptic GSK3β reduction. Axin stabilization using XAV939 rescued the GSK3β deficits and restored the temporal activation of GSK3β during long-term depression in S-SCAM overexpressing neurons. Interestingly, synaptic Axin2 levels were increased in female S-SCAM Tg mice. Female sex hormone 17β-estradiol increased Axin2 expression and increased synaptic GSK3β levels in S-SCAM overexpressing neurons. These results reveal the role of S-SCAM in controlling Axin-dependent synaptic localization of GSK3β. Moreover, our studies point out the pathological relevance of GSK3β hypofunction found in humans and contribute to understanding the molecular underpinnings of sex differences in SCZ.


Scientific Reports
| (2022) 12:4090 | https://doi.org/10.1038/s41598-022-08220-1 www.nature.com/scientificreports/ Axin is a critical scaffolding protein for forming the GSK3β-β-catenin destruction complex in the canonical Wnt signaling 22 . Axin is encoded by two genes producing highly homologous proteins, Axin1 and Axin2 (also known as Axil/conductin) 23 . Axin is enriched at dendritic spines 7,24 and is important for the axonal localization of GSK3β in developing neurons 25,26 . S-SCAM binds to the middle region of Axin1 protein (containing GSK3βinteracting domain [GID]) through its guanylate kinase (GK) domain 7 . The GIDs of Axin1 and Axin2 are highly conserved (~ 70% similarity at the amino acid sequence level). Since S-SCAM and GSK3β bind the same region in Axin, S-SCAM was shown to compete with GSK3β for Axin-binding 7,26 . This competitive binding inhibits β-catenin phosphorylation by GSK3β in vitro 7 . Therefore, it is likely that GSK3β interaction with Axin is hampered in S-SCAM Tg mice, which may disrupt the proper synaptic localization of GSK3β in neurons. However, these possibilities have not been addressed.
As a first step to understand the molecular bases of the sex differences exhibited in S-SCAM Tg mice, we investigated whether aberrant GSK3β signaling is associated with these phenotypes, as well as the potential role of Axin and female sex hormone in the process. These studies reveal the role of Axin in maintaining proper GSK3β signaling at synapses and provide mechanisms underlying the sex differences found in S-SCAM Tg mice.

Results
Synapse-specific reduction of GSK3β protein levels and its activity in male S-SCAM Tg mice. To study the role of S-SCAM in GSK3β-mediated signaling at synapses (Fig. 1a), we first examined GSK3β protein levels in synaptosomal fractions (P2; biochemical correlates of synapses 27 ) of the forebrain tissues obtained from male Tg mice (Tg-M; 3 ± 0.5-month-old). GSK3β protein levels in P2 fractions were greatly reduced (66.7% ± 4.3% of WT; Fig. 1b, c). On the other hand, GSK3β levels in total homogenate fraction (H) of male Tg mice show no difference from male WT mice (Fig. 1b, d), suggesting that the reduced synaptic GSK3β protein levels are unlikely due to the decreased expression of GSK3β. This synaptic GSK3β reduction is accompanied by increased synaptosomal phospho-Ser9-GSK3β (pS9-GSK3β) levels (normalized to GSK3β levels; 165.9% ± 11.4% of WT; Fig. 1e, f). In contrast, pS9-GSK3β levels in total homogenate (H) of male Tg mcie showed no difference from male WT mice (Fig. 1e, g). These results suggest that S-SCAM overexpression inhibits synaptic targeting of GSK3β and impairs synaptic GSKβ activity in the brain of male S-SCAM Tg mice.

Elevated CaMKII activity in the brains of male S-SCAM Tg mice.
To investigate the mechanisms associated with the enhanced inhibitory phosphorylation of synaptic GSK3β in the brains of male S-SCAM Tg mice, we examined the activity of two known upstream protein kinases of GSK3β, CaMKII and Akt 19,28 . Activation of these kinases were assessed by specific antibodies recognizing phosphorylated (and thus activated) forms of CaMKIIa and CaMKIIb (pThr286 and pThr287 for CaMKIIa and CaMKIIb, respectively) and Akt (pSer473). We found that total pCaMKII levels were significantly increased in both P2 and H fractions of male Tg mice (168% ± 13.6% and 188% ± 8.6% of WT, respectively; Fig. 2a-c). On the other hand, both pAkt and Akt protein levels were not significantly changed in either P2 or H fractions (Fig. 2a-c). These results suggest that elevation of inhibitory phosphorylation of GSK3β is most likely caused by increased CaMKII activity in male S-SCAM Tg mice. These data are consistent with elevated glutamatergic activity at the synapse found in S-SCAM Tg mice 14 . Reduced synaptic GSK3β protein levels in male S-SCAM Tg mice is associated with decreased synaptic Axin1 levels. Axin1 is one of the S-SCAM-interacting proteins and is known to be involved in the recruitment of GSK3β to the plasma membrane 29 . To investigate the potential role of Axin in the altered synaptic localization of GSK3β in male S-SCAM Tg mice, we first explored the possibility that Axin protein levels are altered in the S-SCAM Tg mice. We examined both Axin1 and Axin2 proteins, since they can be functionally interchangeable 23 . Total protein levels of both Axin1 and Axin2 in the forebrain tissues of male Tg mice were indistinguishable from male WT mice (Fig. 3a, b). In contrast, Axin1 protein levels in the P2 fraction were significantly reduced in male S-SCAM Tg mice (66% ± 4.2% of WT; Fig. 3a, b). On the other hand, there was no such difference in Axin2 protein levels in the P2 fraction (Fig. 3a, c). Therefore, male S-SCAM Tg mice have Axin1 deficits in synapses of the forebrain.
S-SCAM and GSK3β competitively bind the same region in Axin 7 . Therefore, S-SCAM overexpression may hamper the Axin-mediated synaptic targeting of GSK3β. To address this possibility, we took advantage of the fact that Axin binds to the GK domain of S-SCAM 7 . Therefore, a S-SCAM mutant lacking the GK domain (DGK) 13 should not interfere with the Axin-GSK3β interaction. The DGK mutant is indistinguishable from WT S-SCAM in synaptic targeting, as well as its ability to increase spine sizes and synaptic AMPA receptor levels 13 . Consistent with our hypothesis, while the overexpression of WT S-SCAM greatly reduced synaptic GSK3β levels (59.7% ± 6% of GFP control; Fig. 3d, e), the overexpression of DGK mutant did not hamper the synaptic targeting of GSK3β (127% ± 15.2%; Fig. 3d, e). These results strongly suggest S-SCAM overexpression inhibits Axin-mediated synaptic targeting of GSK3β.
Stabilization of Axin1 restores synaptic targeting and temporal regulation of GSK3β activity during LTD. To further study the role of Axin1 deficits in the impairment of GSK3β function at synapses, we thought to rescue the deficits using pharmacological approaches. A small molecule tankyrase inhibitor XAV939 stabilizes Axin by preventing ADP-ribosylation and subsequent degradation 30,31 . Incubation of cultured hippocampal neurons transfected with GFP-expressing Sindbis virus with XAV939 (5 mM) for 24 h greatly increased the total amount of Axin1 (171.3% ± 21% of GFP control; Fig. 4a www.nature.com/scientificreports/ neurons (186% ± 39% of S-SCAM control; 117.4% ± 25% of GFP control; Fig. 4a, b). Unexpectedly, we did not see significant changes in the amount of total Axin2 after XAV939 treatment in both GFP and S-SCAM overexpressing neurons (Fig. 4a, c).
Having verified the restoration of Axin1 levels in S-SCAM overexpressing neurons, we next examined the effect of XAV939 on synaptic GSK3β levels in S-SCAM overexpressing neurons. Incubation of hippocampal neurons with XAV939 greatly increased synaptic GSK3β staining intensity (178% ± 19% of GFP control; cf. 62% ± 5% for S-SCAM control; Fig. 4d, e), indicating increased synaptic GSK3β protein levels. In contrast, XAV939 did not change synaptic GSK3β levels in GFP-transfected neurons (Supplemental Fig. 2).
We showed previously that the overexpression of S-SCAM blocks LTD formation in both cultured hippocampal neurons and hippocampal slices 1 . To evaluate the restoration effect of Axin1 and GSK3β protein levels at synapses on synaptic plasticity, we examined the temporal changes of GSK3β activity during LTD which is critical for LTD formation 19 . We used a chemically (NMDA)-induced LTD (chem-LTD) protocol, which mimics the low frequency stimulation-induced LTD mechanistically 21,32 . We first examined whether XAV939 by itself had an effect on GSK3β activation during LTD. XAV939 alone did not affect NMDA-induced activation of GSK3β in GFP overexpressing neurons, as shown by the rapid reduction in the inhibitory phosphorylation of S9 (Fig. 4f, g). This is similar in temporal profile to GFP control and to previous reports obtained from naïve hippocampal neurons 17,21,33 . Having confirmed that XAV939 does not affect the rapid activation of GSK3β upon Representative results (e) and quantification of relative pS9-GSK3β levels normalized to GSK3β protein levels (f, g). n = 5-7 mice per group. ***p < 0.001, unpaired t-test with Welch's correction. n.s., not significant. www.nature.com/scientificreports/ NMDA treatment, we next examined the changes of GSK3β activity during chem-LTD in S-SCAM overexpressing neurons. Surprisingly, pS9-GSK3β levels were not significantly changed upon NMDA treatment in S-SCAM overexpressing neurons, indicating dysregulated GSK3β activity. Remarkably, XAV939 restored the temporal activation of GSK3β in S-SCAM overexpressing neurons, which is indistinguishable from XAV939-treated GFP control neurons (Fig. 4f, g). These results collectively suggest that S-SCAM overexpression hampers the temporal regulation of GSK3β activity and blocks LTD formation by inducing Axin1 deficits at synapses.   (Fig. 5a, b) and Axin1 protein levels (Fig. 5c,  d). Surprisingly, synaptic Axin2 protein levels were significantly increased in the female Tg mice (154% ± 7% of WT; Fig. 5c, d). These results suggest that female-specific alteration in Axin2 may have a protective effect in preserving synaptic GSK3β function. Axin2, while functionally similar to Axin1, shows distinguished expression patterns from Axin1 23 . For example, it was shown that Axin2 expression is inducible by Wnt signaling 34,35 . To investigate the effect of female sex hormone E2 on Axin2, we treated cultured hippocampal neurons with E2 (100 nM) and examined its effect on total Axin protein levels. As shown in Fig. 5e, E2 significantly increased the amount of Axin2 protein (149% ± 13% of control; Fig. 5f), while Axin1 protein levels were not affected much (Fig. 5g). This E2-induced Axin2 protein increase was blocked by the transcription inhibitor actinomycin D (ActD, 2 mM; Fig. 5e, f), suggesting that E2 induces Axin2 expression. Notably, Axin1 protein levels were decreased by ActD (61.5% ± 13% of control; Fig. 5g), although the data did not reach statistical significance (p = 0.051).
Since Axin stabilization increases synaptic GSK3β protein levels in S-SCAM-overexpressing neurons (Fig. 4e), we next examined the effect of E2 on synaptic GSK3β protein levels in S-SCAM overexpressing neurons. Compared to control neurons, E2-treated neurons showed a great increase in synaptic GSK3β staining intensities www.nature.com/scientificreports/ (217% ± 21%; Fig. 5h, i). E2 did not have significant effect on synaptic GSK3β levels in GFP-transfected neurons (Supplemental Fig. 2). These results suggest that E2 preserves synaptic GSK3β function by increasing Axin2 expression in female Tg mice.

S-SCAM overexpression did not change the amount of β-catenin. Axin is considered as a key lim-
iting factor for canonical Wnt signaling since it controls β-catenin levels by promoting the assembly of β-catenin destruction complex. In neurons, however, β-catenin also plays a structural function at synapses, independent of its role in the transcriptional regulation of Wnt signaling 26 . To evaluate the effect of S-SCAM overexpression on these processes, we examined the total and synaptic β-catenin levels in male S-SCAM Tg mice. We found that both total (H) and synaptic (P2) amounts of β-catenin were not significantly changed (Fig. 6a-c). Therefore, the reduction in synaptic Axin1 amounts did not cause β-catenin stabilization in the Tg mice. These results suggest that synaptic Axin1 plays a role in GSK3β recruitment and S-SCAM overexpression does not directly affect the canonical Wnt signaling.

Discussion
In this paper, we described the intriguing interplay of S-SCAM and Axins in the assembly of the GSK3β signaling complex at synapses. We also uncovered the role of Axins in the sex differences displayed in S-SCAM Tg mouse model of SCZ. There are three main findings: First, we showed that Axin plays a pivotal role in the synaptic localization of GSK3β and the proper assembly of the GSK3β signaling complex involved in synaptic plasticity. S-SCAM inhibits these processes by binding to Axin in competition with GSK3β. Notably, Axin mediates the synaptic localization of GSK3β in a manner independent of the canonical Wnt signaling, which is reminiscent of Axin's www.nature.com/scientificreports/ role in the axonal localization of GSK3β 25 . At this point, the molecular mechanism(s) responsible for the reduced Axin protein levels at synapses in S-SCAM overexpressing neurons remains unidentified. However, we suspect that increased glutamatergic activity at synapses might contribute to the reduction of synaptic Axin1 levels. S-SCAM Tg mice showed impaired synaptic GSK3β activity (shown by both reduced GSK3β protein levels and increased inhibitory phosphorylation), which is associated with increased CaMKII activity at synapses. Reduced GSK3β activity could lead to Axin reduction, since GSK3β stabilizes Axin via direct phosphorylation 36,37 and dephosphorylated Axin is degraded 38,39 . Moreover, it is well known that increased glutamatergic activity promotes poly-ubiquitination of synaptic proteins 40,41 . Axin is a target for poly-ubiquitination and subsequent degradation by proteasomes 29 . Interestingly, under our experimental conditions, XAV939 did not increase Axin2 levels, unlike Axin1. In addition to preventing Axin degradation, XAV939 blocks the mRNA expression of Axin2 but not Axin1 23,42 . Therefore, it is plausible that Axin2 protein levels did not change by XAV939 because there is no new Axin2 protein synthesis to increase total Axin2 levels. Further studies await to verify this possibility. Second, we demonstrated that GSK3β hypofunction is associated with the S-SCAM Tg mouse model. GSK3β is a key protein kinase strongly implicated in the pathogenesis of SCZ 43,44 . GSK3 hyperfunction is found in individuals with SCZ 45 . Consistently, genetic and molecular studies confirmed the key contribution of GSK3β hyperfunction in the pathogenesis of SCZ 44,46,47 . Paradoxically, hypofunction of GSK3β is found in patients with SCZ. Postmortem studies of SCZ patients revealed low GSK3β activity 48 , reduced protein levels 49,50 (but see also 51 ), and immunoreactivity 52 . Moreover, genetic association studies uncovered a single-nucleotide polymorphism that reduces GSK3β mRNA and protein levels in human subjects 53 . However, due to small sample sizes and high variability of the data, the pathophysiological significance of GSK3β hypofunction has remained unclear. To our knowledge, our studies provide the first preclinical evidence supporting the causal relationship of GSK3β hypofunction and the pathogenesis of SCZ.
Third, we identified mechanisms responsible for sex differences observed in S-SCAM Tg mice, which might be potentially relevant to SCZ. Sex differences are well documented in the humans with SCZ, which include disease risk, course, and outcome 15 . Men have 1.4-fold higher incidence of SCZ, display more severe symptoms and worse cognitive impairments, and are generally less responsive to antipsychotic treatments when compared to women 54 . Moreover, SCZ incidence in women greatly increases around menopause 55 . Based on sex differences of SCZ, the "estrogen hypothesis" was proposed in which the powerful female sex hormone estrogen plays a protective role against the development and severity of the disease. In clinical trials, supplemental estrogen treatment administered in conjunction with antipsychotics is beneficial for SCZ 55 . Our results suggest that Axin deficits at synapses are responsible for the male-specific GSK3β hypofunction in S-SCAM Tg mice. In female S-SCAM Tg mice, estrogen seems to buffer the Axin deficit at synapses by increasing Axin2 expression in neurons. Remarkably, Axin2 levels are increased to a level comparable to the elevated amounts of synaptic S-SCAM proteins (~ 1.5-fold) 14 . Therefore, the increase in Axin proteins seems sufficient to compensate for the increased S-SCAM levels and maintains GSK3β levels at synapses. These results are consistent with previous findings that, unlike Axin1, Axin2 shows an inducible expression pattern that is dependent on β-catenin/Tcf 34 . Moreover, it was shown that E2 activates β-catenin-dependent transcription in neurons 56 . Therefore, it is highly conceivable that E2 protects Axin levels at synapses in female S-SCAM Tg mice (and thereby GSK3β signaling complex) by promoting Axin2 transcription. Overall, our findings provide strong support for the estrogen hypothesis.
The S-SCAM Tg mice were generated by using the CaMKII promoter to drive the expression of S-SCAM transgene. Therefore, the Tg mice have elevated S-SCAM levels primarily in excitatory neurons of the forebrain area and most highly in the hippocampus 57 . Therefore, the sex differences observed in the Tg mice are likely caused by the excitation/inhibition imbalances in these principal neurons. Consistently, it is well documented that hippocampal functions are highly influenced by estrogen [58][59][60] . However, it remains to be determined whether sex differences of SCZ are also driven by the alterations of the glutamatergic function in the hippocampus and/ or other brains regions.
We used cultured rat hippocampal neurons, which are reliable and amenable for molecular-genetic and pharmacological manipulations. These advantages allowed us to perform initial studies on the molecular mechanisms underlying sex-differences in S-SCAM Tg mice. We do not expect significant differences in rat vs mouse hippocampal neurons, since the major phenotypes of S-SCAM overexpression found from rat neurons were replicated in mouse neurons in vivo 1,14 . Further studies using the S-SCAM Tg mouse model will strengthen the role of Axin and E2 in sex-differences found in this study.
Finally, our studies provide a potential new therapeutic target for SCZ. Recent reports demonstratedthat chronic intraperitoneal XAV939 treatment of mice has an impact on the Wnt signaling in the brains, suggesting www.nature.com/scientificreports/ that XAV939 passes the blood-brain barrier and stabilizes Axin 61 . Therefore it would be interesting to investigate the effect of XAV939 on the SCZ-like behavioral deficits displayed in S-SCAM male Tg mice, especially on synaptic plasticity and working memory deficits.

Methods
Animals. S-SCAM Tg mice (C57BL/6J-Tg(Camk2a-Magi2)1Shlee/J; Jackson stock No: 027306) were maintained as described 14 . Timed pregnant Sprague-Dawley female rats were obtained from Envigo. All experimental procedures involving the mice and rats were performed in accordance with the relevant guidelines and regulations and approved by the Institutional Animal Care and Use Committee in the Medical College of Wisconsin. All methods are reported in accordance with ARRIVE guidelines.
Cultured rat hippocampal neurons and transfection. Dissociated rat hippocampal neuron culture was prepared from E18 embryos (both sexes were used) of Sprague Dawley rats and maintained in Neurobasal medium supplemented with B27 and Pen/Strep (ThermoFisher Scientific) as described previously 62 . Hippocampal neurons were transfected at div14 using Lipofectamine 2000 as described previously 63 .
Immunocytochemistry. Transfected neurons were fixed at 2 days post-transfection. Immunocytochemistry was performed as described 1,63  Immunocytochemical image acquisition and analyses. Images were acquired by using a Nikon C1 plus laser scanning confocal microscope and 60 × objective (NA1.4). Acquired images (z-series stacks) were first converted to projection images (with maximal projection option) for analyses. Both image acquisition and analyses were done in a blind manner. To measure synaptic GSK3β intensities, 15-30 dendritic spine regions were randomly selected from the dendrites of transfected neurons based on GFP or myc fluorescence (overexpressed S-SCAM is highly enriched in dendritic spines 1 ) using SynPAnal software 64 . After applying threshold, integrated intensity values from individual spines were obtained and average values of them were calculated per neuron basis in Excel. All data were transferred to GraphPad Prizm software for computation and graphical representation.

Statistical analyses.
All neuron experiments were performed at least in triplicate using independent batches of hippocampal neuron cultures. All data values represent means ± s.e.m. For multiple group comparisons, one-way ANOVA with Tukey's multiple comparison post hoc test were performed using the GraphPad Prizm software. Welch's t test (unpaired) was used to determine the statistical significance for two groups. p < 0.05 was considered significant.

Data availability
All materials, data, and associated protocols will be promptly made available to readers without undue qualifications in material transfer agreements.