Normalization of non‐canonical Wnt signalings does not compromise blood‐brain barrier protection conferred by upregulating endothelial Wnt/β‐catenin signaling following ischemic stroke

Abstract Background Endothelial canonical (Wnt/β‐catenin) and non‐canonical Wnt signalings (Wnt/PCP and Wnt/Ca2+) promote blood‐brain barrier (BBB) development and antagonize each other. However, the effects of ischemic stroke on endothelial canonical and non‐canonical Wnt signalings are unclear. Further, how non‐canonical Wnt signalings are influenced by upregulation of endothelial Wnt/β‐catenin signaling and subsequently affect BBB function following ischemic stroke have not been studied. Methods First, we determined the levels of Wnt signaling markers including TCF/LEF1 transcription activity, Axin2 mRNA, phospho‐JNKThr183/Tyr185, and NFAT in brain endothelial cells (ECs) with the deletion of Wnt receptor Frizzled (Fzd)4 or Fzd6, the two most abundant Fzds in brain ECs. Next, we observed the effect of ischemia/reperfusion injury on Wnt signalings in brain ECs and adult mice. Last, we assessed the changes of non‐canonical Wnt signalings and BBB injury in the early stage of ischemic stroke in mice with endothelial β‐catenin activation (β‐cat mice). Results Fzd4 or Fzd6 deletion dampened both Wnt/β‐catenin and Wnt/PCP signalings but enhanced Wnt/Ca2+ signaling in brain ECs. Both canonical and non‐canonical Wnt signalings in brain ECs were downregulated after ischemia/reperfusion injury in vitro and in vivo. Upregulating endothelial Wnt/β‐catenin signaling in β‐cat mice normalized the downregulated non‐canonical Wnt signalings, which did not compromise its protective effects on BBB integrity and endothelial tight junction following ischemic stroke. Conclusions The BBB protection induced by upregulation of endothelial Wnt/β‐catenin signaling may be not interfered by the normalization of non‐canonical Wnt signalings in the early stage of ischemic stroke.


| INTRODUC TI ON
The blood-brain barrier (BBB) is comprised of endothelial cells (ECs), pericytes, basement membrane, and astrocyte, among which, the barrier function is mainly attributed to the ECs and regulated by endothelial Wnt/β-catenin signaling (Wnt/β-cat), 1,2 also named as canonical Wnt signaling. The canonical Wnt signaling is involved in many embryonic development processes, and abnormal canonical Wnt signaling causes various diseases, including ischemic stroke, neurodegenerative diseases, and various kinds of cancers. [3][4][5] In addition, as shown by plenty of evidences, Wnt proteins can also activate signaling pathways independent of β-catenin, mainly including Wnt/planar cell polarity (PCP) and Wnt/Ca 2+ signaling, which are together called non-canonical Wnt signalings. 4 The Wnt/β-cat signaling is involved in cell differentiation and proliferation; Wnt/PCP signaling regulates cytoskeleton and cell polarization; Wnt/Ca 2+ signaling is associated with inflammation and neurodegeneration. 4 Previous studies have shown that Wnt/β-cat and Wnt/ PCP signalings in ECs both played crucial roles to control the integrity of ECs and the barrier function of BBB, 1,2,6 and Wnt/ Ca 2+ signaling is involved in the EC growth and migration, 7 indicating synergistic effects can be induced by different types of Wnt signaling. Interestingly, nevertheless, some clues have hinted the existence of antagonism between canonical and noncanonical Wnt signalings. 8,9 Our previous study showed that after ischemic stroke, the BBB integrity was disrupted in mice with conditional deletion of endothelial Gpr124, a Wnt7-specific coactivator of Wnt/β-cat signaling. 10 The disruption of BBB function was fully rescued by genetic activation of endothelial β-catenin, 10 suggesting that the BBB breakdown in the early stage of ischemic stroke could be treated by manipulating endothelial Wnt/β-cat signaling. Besides, some other studies including ours showed that the treatment with glycogen synthase kinase 3β (GSK3β) inhibitors (IM-12, TWS119, and lithium) significantly improved the outcomes of mice with ischemic stroke by upregulating Wnt/β-cat signaling. [11][12][13] Since upregulating endothelial Wnt/β-cat signaling has been considered as a novel therapeutic method for protecting BBB after ischemic stroke, and the antagonisms between canonical and non-canonical Wnt signalings have been suggested to exist, 8,9 it is necessary to determine the changes of non-canonical Wnt signalings and their consequences while manipulating endothelial Wnt/β-cat signaling in the treatment of ischemic stroke.
Here, we aimed to investigate the interplay of canonical and non-canonical Wnt signalings in brain ECs under normal state or ischemic stroke in vivo and in vitro, and examine the changes of non-canonical Wnt signalings during the upregulation of endothelial Wnt/β-cat signaling by employing a genetic mouse model with endothelial β-catenin activation.

| Cell culture and OGD/R treatment
The mouse brain endothelial cell line bEnd. 3 Figure S1). Primary brain ECs were obtained from brains of 8-to 10-week-old adult C57BL/6 wild-type mice which were minced and disaggregated. After centrifugation at 400 g for 5 min, cell pellets were resuspended in EGM-2MV medium (LONZA) with 10% FBS and 4 μg/ml puromycin and plated into fibronectinprecoated plates (10 μg/ml for 30 min at 37°C). 3 days later, medium was discarded and cells were washed with PBS twice and medium was replaced with EGM-2MV/10% FBS. All cultures were maintained in a humidified 5% CO 2 incubator at 37°C and routinely passaged when 80%-90% confluent. The optimal durations of Wnt5a for activating Wnt/PCP and Wnt/Ca 2+ signalings by measuring the relative levels of phospho-JNK Thr183/Tyr185 and nuclear factor of activated T cells (NFAT) protein in bEnd.3 cell line had been determined in preliminary experiments (Supplementary Figure S2).
The cells were exposed to oxygen-glucose deprivation (OGD).
Briefly, cultured medium was replaced by Dulbecco's Modified of Eagle's Medium (Solarbio) and the cultured cells were put in a Modular Incubator Chamber (Billups-Rothenberg) with 0.5%-1% O 2 and 99% N 2 , which was monitored with an O 2 analyzer (HNZA). We assessed the effect of different durations of oxygen-glucose recovery (OGR) after 6 hour (h)-OGD for cell vitality and 3 h-OGR was chosen in the following experiments (Supplementary Figure S3). Hence, after 6 h-OGD, cells were returned to normal culture conditions for 3 h-OGR.

| Animal protocol
Male 8-to 10-week-old C57BL/6 mice were obtained from the Beijing Vital River Laboratory Animal Technologies Co. Ltd. To induce endothelial β-catenin (encoded by Ctnnb1) constitutive activation, Cdh5-CreER mice were crossed with mice bearing the Ctnnb1 lox(ex3) allele to generate Ctnnb1 lox(ex3)/+ ; Cdh5-CreER mice (termed β-cat), and Ctnnb1 +/+ ; Cdh5-CreER control mice (termed WT). The recombination efficiency of Ctnnb1 gene in brain ECs of β-cat mice has been confirmed in our previous study 10 and preliminary experiments. In RNA-seq of brain ECs, we found due to only exon3 being deleted while other exons were expressed normally, Ctnnb1 mRNA expression in β-cat mice did not apparently decrease compared with WT mice, but the Wnt/β-cat target genes Axin2, Apcdd1, and Nkd1 significantly increased (data not shown).
Meanwhile, non-canonical Wnt signaling target genes Pfn2, Vangl2, Celsr3, DAAM1, CamkII, and NFATc1/c3 expression in brain ECs of β-cat mice had no significant changes (Supplementary Figure S4). 8-to 10-week-old mice were treated with tamoxifen (2 mg/10 g body weight) in corn oil through an oral feeding needle every other day for 5 days with a total of three doses per mouse. Mice were allowed to recover from tamoxifen treatment-related toxicities (un- Randomized animals were used in the sham operation and ischemia/ reperfusion (I/R) group. For the surgical procedures, anesthesia was induced with 4% isoflurane in an induction chamber; anesthesia was maintained with 2% isoflurane delivered through a face mask (RWD).
A heating pad was used to maintain each mouse's core temperature at 37 ± 0.5°C throughout the surgical procedure. A modified intraluminal filament model was used to induce transient middle cerebral artery occlusion (MCAO) as previously described. 14 After 60 min of MCAO, reperfusion was established by retracting the filament. Animals had free access to food and water throughout the 24 h reperfusion period. Neurological deficit scores were evaluated at 24 h post-MCAO by a blind observer according to the following scoring system: 0 = normal extension of both forelimbs; 1 = adduction of affected forelimb; 2 = the grip strength of affected forelimb decreased significantly; 3 = leaning to the opposite side; 4 = circling to the contralateral side; and 5 = no autonomic activity.

| Infarct size analysis
The infarct region in sections with H & E staining was defined as the area with reduced staining or areas containing eosinophilicnecrotic cell bodies. The boundaries between regions of infarct and adjacent normal brain were clearly delineated. Tissue sections were photographed using a ZEISS microscope (Axio Imager 2 Pol), and the infarct area was measured using ImageJ software

| Immunofluorescence staining
Frozen 7μm-thickness sections were allowed to dry on Adhesion Microscope slides (CITOTEST) at room temperature before being rehydrated in PBS. Sections were blocked in 10% normal goat serum (Thermal Fisher) in PBS +0.2% Triton X-100 for 1 h at room temperature. Samples were incubated at 4°C with the following antibodies in PBS +5% goat serum +0.2% Triton X-100: hamster anti-mouse CD31 for 1 h at room temperature. Excess antibody was removed by rinsing in PBS for 5 min, 3 times. Slides were mounted in antifading mounting medium with DAPI (Solarbio) and imaged with an Olympus microscope to obtain 20× or 40× images. The whole-brain image was taken under low-power microscope and adjusting white balance through healthy hemisphere. The optical density (OD) value was measured by an experimenter blinded to group allocation with ImageJ. Immunofluorescence signal area or density was quantified and normalized by CD31 signal area in 5 to 8 random ischemic areas of cortex. Pericyte coverage was quantified by measuring the staining signal length of pericytes or ECs with ImageJ. All tests were performed by an experimenter who was unaware of the identity of experimentations.

| FACS sorting of brain ECs
Brains of adult mice were minced and disaggregated in Type

| RNA-seq
For mouse brain ECs, the total RNA was extracted using the Arcturus picopure RNA Isolation Kit (Applied Biosystems).
According to our previous study, 10 we performed quantitative 3′ end SEQ. The sequencing was performed on Illumina-hiseq-2500 platform (Illumina) in single-ended 50 bp format. TopHat was used to map the original sequence reading to the mouse genome (mm10), and customized R scripts were used to calculate the frequency of Refseq genes. Then, the original counts were standardized by using the pruning average (TMM) method of M value, and compared with Bioconductor software package "edgeR." Readings per gigabase per million (RPKM) are also calculated from raw counts. In at least one sample, if RPKM ≥1, fold change ≥1.5, p ≤ 0.05, the differentially expressed genes can be identified.
Gene richness was analyzed by David database. A plethora of evidences have suggested that Fzd4 serves as the principal receptor for Wnt/β-cat signaling in brain ECs, 15 while the role of Fzd6 in this signaling is yet unclear. As shown in Figure 1A and B, deleting Then, we detected the phosphorylation of JNK, the key downstream process in Wnt/PCP signaling. 16 We found that with  Data are mean ± SE, *p < 0.05, **p < 0.01. One-Way ANOVA followed by LSD t test. Fzd, frizzled; WT, wild type; P-JNK, phospho-JNK Thr183/Tyr185 ; NFAT, nuclear factor of activated T cells; ns, not significant F I G U R E 2 Canonical and non-canonical Wnt signalings in brain endothelial cells were both downregulated by OGD/R. (A) After OGD/R, the relative active β-catenin, P-JNK, and nucleus NFAT protein levels in bEnd.3 cell line with 100 ng/ml Wnt3a for 24 h, 200 ng/ml Wnt5a for 30 min, and 200 ng/ml Wnt5a for 60 min, respectively. (B) Quantitation of the bands in (A). n = 3 samples per condition. One-way ANOVA followed by Dunnett T3 test for active β-catenin and LSD t test for other proteins. (C) Expression of Axin2 mRNA in bEnd.3 cell line after OGD/R, with 100 ng/ml Wnt3a protein treatment for 24 h, as assessed by RT-qPCR. n = 3 biological replicates per condition. Oneway ANOVA followed by LSD t test. (D) Expression of Axin2, Nkd1, and Apcdd1 mRNA in primary brain ECs of adult mice after OGD/R, with 100 ng/ml Wnt3a protein activation for 24 h, as assessed by RT-PCR. n = 4 biological replicates per condition. One-way ANOVA followed by LSD t test. (E) After OGD/R, P-JNK, JNK, P-CamkII, and CamkII protein levels in primary brain ECs were measured by Western blot, with 200 ng/ml Wnt5a protein stimulation for 60 min. (F) The ratios of P-JNK and JNK, P-CamkII, and CamkII from the protein bands in (E). n = 3 samples per group. One-way ANOVA followed by LSD t test. Data are mean ± SE, *p < 0.05, **p < 0.01. OGD/R, oxygen-glucose deprivation and recovery; P-JNK, phospho-JNK Thr183/Tyr185 ; P-CamkII, phospho-CamkII Thr286 ; NFAT, nuclear factor of activated T cells; ECs, endothelial cells; ns, not significant cells (F = 18.33, p = 0.0359; Figure 2A,B) and in primary brain ECs (F = 6.093, p = 0.0139; Figure 2E,F) with Wnt5a stimulation, while it was just a slight trend of increase in nucleus NFAT protein expression (F = 6.904, p = 0.6947; Figure 2A,B). After OGD/R, apparent decreases of P-JNK and nucleus NFAT protein were observed in bEnd.3 cells and phospho-CamkII Thr286 (P-CamkII) in primary brain ECs with Wnt5a stimulation (F = 18.33, p = 0.0011; F = 6.904, p = 0.0253; F = 9.114, p = 0.0240; Figure 2A,B,E,F). The decrease of P-JNK in primary brain ECs after OGD/R was not apparent with or without

| Upregulating endothelial Wnt/β-cat signaling after I/R resulted in the normalization of downregulated non-canonical Wnt signalings
Given that upregulating endothelial Wnt/β-cat has been demonstrated as a novel potential treatment for BBB protection following ischemic stroke, [10][11][12][13] we wondered if the treatment after ischemic stroke would induce any changes of non-canonical Wnt signalings. The mice with F I G U R E 3 The downregulated non-canonical Wnt signalings in mouse brain following I/R were normalized by endothelial-specific β-catenin activation. (A) Breeding schema for endothelial cell conditional Ctnnb1 exon3 deletion by crossing to Cdh5-CreER mice, deleting exon3, and converting the Ctnnb1 lox(ex3) allele into a Ctnnb1 ex3Δ allele. (B) After I/R, P-JNK, JNK, P-CamkII, and CamkII protein levels in brains of mice were measured by Western blot. (C, D) The ratios of P-JNK and JNK, P-CamkII, and CamkII from the bands in (B). n = 3 mice per group. (E, F) Expression of Pfn2 and CamkII, as assessed by RT-qPCR, in ischemic hemisphere of mice. n = 4 mice per group. Data are mean ± SE, *p < 0.05, **p < 0.01. One-way ANOVA followed by LSD t test. TM, tamoxifen; MCAO, middle cerebral artery occlusion; WT, wild type; I/R, ischemia/reperfusion; P-JNK, phospho-JNK Thr183/Tyr185 ; P-CamkII, phospho-CamkII Thr286 endothelial β-catenin constitutive activation were generated, shown in Figure 3A. As shown in Figure 3B-F, the relative P-JNK (F = 4.615, p = 0.0410) and P-CamkII (F = 8.664, p = 0.0070) protein, and the mRNA levels of Pfn2 (F = 37.36, p < 0.0001) and CamkII (F = 15.17, p = 0.0001) in WT mice with I/R were apparently reduced, compared with WT mice with sham operation; after I/R, P-JNK (p = 0.0310), P-CamkII (p = 0.005), Pfn2 (p = 0.0127), and CamkII (p = 0.0231) mRNA levels in βcat mice were higher than that in WT mice. In order to confirm whether endothelial non-canonical Wnt signalings were also normalized after the upregulation of endothelial Wnt/β-cat signaling, we determined the changes of non-canonical Wnt signaling markers in primary brain ECs with Wnt/β-cat signaling activated by Wnt3a protein after OGD/R. We found that after Wnt3a upregulated Wnt/β-cat signaling, only the relative P-CamkII protein level increased significantly (F = 5.687, p = 0.0326; Supplementary Figure S6B,C), while Pfn2, Vangl2 mRNA, and relative P-JNK protein level did not change significantly (Supplementary Figure   S6A-C). These data suggest that the upregulation of endothelial Wnt/βcat signaling after I/R would synergistically normalize inhibited Wnt/ PCP and Wnt/Ca 2+ signalings in brain tissues, but only Wnt/Ca 2+ signaling in brain ECs. It is worth mentioning that with sham operation, the Pfn2 mRNA level was lower in β-cat mice than that in WT mice (F = 37.36, p = 0.0207; Figure 3E). Also, for Wnt/Ca 2+ signaling, the ratio of P-CamkII to CamkII protein (F = 8.664, p = 0.015; Figure 3D) and the mRNA level of CamkII (F = 15.17, p = 0.0029; Figure 3F) were lower in β-cat mice than that in WT mice, supporting that both Wnt/PCP and Wnt/Ca 2+ signalings in brain tissues were inhibited by upregulating endothelial Wnt/β-cat signaling. (C) Co-immunofluorescence staining for ZO-1, Claudin-5, Laminin, and Desmin (green) with CD31 (red) in brain infarcted regions. The yellow signals showed positive signals on blood vessels. Scale bar, 100 µm. (D) Relative positive signal densities or areas were normalized to the CD31 signal area. 5 -8 low-power fields per mice were randomly selected. n = 4 mice per group. Data are mean ± SE, *p < 0.05, **p < 0.01. One-way ANOVA followed by LSD t test. I/R, ischemia/reperfusion; ZO-1, zonula occludens-1; β-cat, Ctnnb1 lox(ex3)/+ ; Cdh5-CreER mice; WT, Ctnnb1 +/+ ; Cdh5-CreER mice; ns, not significant endothelial tight junctions, basement membrane, and pericyte coverage, were examined. First of all, our data showed that smaller infarct areas were found in β-cat mice (37.0 ± 5.6%) than that in WT mice (54.1 ± 2.2%; t = 2.870, p = 0.0208; Figure 4A,B), and the neurological deficit scores in β-cat mice were also lower than that in WT mice (t = 2.220, p = 0.0434; Figure 4C). The BBB leakage was assessed by mouse blood IgG extravasation from blood vessels in infarct areas. Our results showed that blood IgG leakage was rescued in β-cat mice compared with WT mice (F = 31.09, p = 0.0012; Figure 4D,E), indicating that upregulating endothelial Wnt/β-cat signaling significantly improved the BBB function after I/R. Then, the protein levels of the markers of BBB components in ischemic hemisphere of mice were detected by Western blot. As shown in Figure 5A and B, there were no obvious differences in these markers between β-cat and WT mice with sham operation.
It has been shown that Fzd4 could activate Wnt/β-cat signaling by specifically interacting with Wnt proteins and norrin, an atypical Fzd ligand, to play an essential role in barrier maintenance and plasticity in brain ECs. 15,21 Our study supported that Fzd4 was not the only Fzd receptor in mediation of endothelial Wnt/β-cat signaling. We found that TCF/LEF TOP-flash value and Axin2 mRNA levels were significantly reduced not only in Fzd4 −/− cells, but also in Fzd6 −/− cells, indicating that Fzd6 also had certain functions in Wnt/β-cat signaling.
Fzd6 has been shown to regulate Wnt/PCP signaling as its prevalent role, 20  Previous studies showed BBB breakdown occurred rapidly following reperfusion within several hours and reached to a peak around at 24 to 48 h after ischemia. [24][25][26][27] Furthermore, hemorrhagic transformation in ischemic stroke patients is detected usually within 24 to 36 h following intravenous thrombolysis therapy and thrombectomy. Therefore, in this study we chose to allow the reperfusion to carry on for 24 h in order to fully observe the effect of noncanonical Wnt signalings normalization for BBB protection induced by upregulation of endothelial Wnt/β-cat signaling. Our previous study showed that increasing Wnt/β-cat signaling protected BBB by regulating endothelial tight junctions in the acute stage of ischemic stroke. 10 Consistently, we also found the protection for tight junctions in mice with endothelial β-catenin activation in the early stage of ischemic stroke, and BBB protection was not weakened even though antagonistic non-canonical Wnt signalings were synchronously upregulated, which is the first time to observe the changes of non-canonical Wnt signalings in the process of BBB protection by manipulating endothelial Wnt/β-cat signaling. Interestingly, we found that the changes of non-canonical Wnt signalings in primary brain ECs were not completely consistent with that in brain tissues after upregulating endothelial Wnt/β-cat signaling, which showed that downregulated Wnt/PCP and Wnt/Ca 2+ signalings in brain tissues could be normalized, while only Wnt/Ca 2+ signaling was normalized in primary brain ECs. It must be mentioned that Wnt proteins in brain tissue are secreted principally by neurons and astrocytes, 28,29 so here we observed the changes of Wnt signalings in the whole-brain tissues, not only ECs, because the upregulation of endothelial Wnt/β-cat signaling would affect non-canonical Wnt signalings not only in ECs, but also in other brain cells. In fact, pericytes, astrocytes, and microglia are also related to BBB integrity through PDGFRβ, AQP4, Poldip2 or inflammation-related signaling pathways. [30][31][32][33] At the same time, Wnt signaling is not the only one to maintain endothelial homeostasis through tight junction proteindependent mechanism. [34][35][36] Even so, our data provide the evidence for Wnt signaling as an important link in maintaining BBB integrity.
In conclusion, our findings demonstrate that: (1) After ischemic stroke, canonical and non-canonical Wnt signalings were downregulated, while non-canonical Wnt signalings could be normalized by upregulating endothelial canonical Wnt signaling. However, we did not investigate the effect of targeted manipulation of non-canonical Wnt signalings on BBB injury after ischemic stroke, which may be worthy of further study. (2) The protection of the BBB by upregulation of Wnt/β-cat signaling depends on protecting the endothelial tight junction in the early stage of ischemic stroke, which was not disturbed by normalization of non-canonical Wnt signalings. (3) It is worth mentioning that except Fzd4, Fzd6 also plays an important role in endothelial Wnt/β-cat signaling. Combined with previous studies, 3,15,[18][19][20] Fzd4 and Fzd6 might have dual mediating roles in both Wnt/β-cat and Wnt/PCP signalings, which will be beneficial for a more comprehensive understanding of Wnt signalings in brain endothelial cells.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available in the supplementary material of this article.