Introduction

Intracerebral hemorrhage (ICH) is a life-threatening stroke type with the worst prognosis and few proven clinical treatments. Most patients who survive intracerebral hemorrhage end up with disabilities and are at risk for recurrency, cognitive decline, and systemic vascular issues, making this condition particularly significant among neurological disorders ^1,2. Several rodent models have been used for modeling ICH, including autologous blood injection, collagenase injection, thrombin injection, and micro-balloon inflation techniques ³. Zebrafish (Danio rerio) exhibit a closed circulatory system and regulatory pathways that are highly conserved among vertebrates ⁴ with relatively cheap and easy to maintain disease models. Over the past decades, varieties of zebrafish mutants that spontaneously developed brain hemorrhage have been generated in mutagenesis screens, such as redhead ⁵, reddish ^6,7 and bubblehead (bbh) ^8,9. Elucidating mutated gene function in these zebrafish models enables us to gain insights into the genetic basis and pathophysiological mechanisms of ICH development. The bbh mutant was identified independently in two large-scale ethylnitrosourea-induced mutagenesis screens ^8,9 and positional cloning revealed that p21-activated kinase (Pak)-interacting exchange factor beta (betaPix) gene was mutated in bbh mutants ¹⁰. betaPix contains SH3 domain that binds group I PAKs, and Dbl homology (DH) and pleckstrin homology (PH) domains that function as a RhoGEF to interact with Rac/Cdc42 small GTPases ¹¹, thus participating in multiple cellular pathways to regulate cell polarity, adhesion and migration ¹².

Various studies have reported that brain vessels are highly supported by neurons and glial cells during development and regeneration ^13,14. Glial-specific betaPix have been proposed to interact with αvβ8 integrin and the Band 4.1s in glia, which link to multiple intracellular signaling effectors such as Wnt7a and Wnt7b ¹⁵, but this hypothesis was not experimentally addressed ¹⁶. To this end, betaPix has been shown to interact with αvβ8 integrins and mediate focal adhesion formation and thus modulate cerebral vascular stability in bbh zebrafish mutant ¹⁷. Integrins are the main molecular link between cells and the extracellular matrix (ECM) which serve as scaffolds in the perivascular space. Based on a chemical suppressor screening of bbh hemorrhages, we have previously reported that a small molecule called miconazole downregulated the pErk-matrix metalloproteinase 9 (Mmp9) signaling to reduce ECM degradation, thus improving vascular integrity ¹⁸. However, how betaPix cell-type specifically contributes to vascular integrity and maturation remains unclear. Here, we generated betaPix conditional trap (betaPix^ct) alleles by using a CRISPR/HDR-based Zwitch method, and found that betaPix acted mainly in glia to regulate glial and vascular integrity via Stathmin and Zfhx3/4 signaling. Therefore, this work provides the first experimental evidence that betaPix acts in glia for vascular integrity development, and its functional conservation between zebrafish and human cells may further guide us to decipher the genetic basis of ICH in the future.

Results

Generating betaPix conditional trap (betaPix^ct) allele by an HDR-mediated knockin and knockout method

To study betaPix function in cell-specific manner, we utilized a donor vector with an invertible gene trap cassette via HDR (Zwitch) ^19–21 for generating betaPix conditional trap (betaPix^ct) alleles. Zwitch consists of a splice acceptor conjugated to a red fluorescent protein via a 2A self-cleaving peptide, flanked by two LoxP sites and two Lox 5171 sites in reciprocal orientations. Outside this conditional trap cassette, we included a lens-specific enhanced green fluorescence protein α-crystallin:EGFP for screening knockout founders, as well as inserted both left and right arms for targeting homologous sequences on the genome. We modified the Zwitch vector by adding a glycine-serine-glycine (GSG) spacer in front of the 2A peptide to enhance cleavage efficiency, and adding universal guide RNA (UgRNA)²² target sequences outside the left and right homologous arms for linearizing donor vector (Figures 1A and S1A). We chose CRISPR/Cas9 system to generate double-strand breaks (DSBs) on intron 5 of the betaPix locus, and chose the most efficient guide RNAs based on their efficiency to induce CRISPR indels. We used either long arms such as ∼1000 bp or short arms such as 24 bp that located upstream and downstream the guide RNA sites. We then co-injected the donor vectors with targeting guide RNA, universal guide RNA, and Cas9 mRNA into one-cell-stage embryos. At 4 days post-fertilization (dpf), we selected α-crystallin-EGFP-positive embryos (F₀) for raising to adulthood, and pre-selected F₀ founders by examining inheritable EGFP reporter expression of F₁ embryos. To confirm the correct homologous recombination in the betaPix locus, we identified potential founders by genomic PCR to examine Zwitch insertion in the betaPix locus. Precise knock-in genotypes were further verified using Sanger sequencing (Figures 1B, 1C, S1B). Among 184 adult F₀ founders with α-crystallin-EGFP expression, we found 84 founders had EGFP-positive F₁ embryos, and 7 founders had expected PCR fragments around both 5’ and 3’ arms, which two founders had correct insertions in the betaPix locus by Sanger sequencing (Figure S1C), thus achieving ∼1% efficiency on generating conditional alleles. Briefly, we established an efficient method for generating conditional knockin/knockout mutants in general, and generated a conditional gene trap line Ki(betaPix:Zwitch) in the betaPix locus in particular, which is referred to as betaPix^ct hereafter.

Figure 1.

A previous work has shown bubblehead (bbh^m292) mutant with a hypomorphic mutation in betaPix, and bbh^fn40a mutant with reduced betaPix transcripts without finding mutations in the coding region and the splicing sites ¹⁰. bbh^fn40a mutants developed cerebral hemorrhages phenotype between 36 to 52 hpf (Figure S1D and S1E). To characterize the betaPix^ct allele, we injected one-cell-stage betaPix^ct/ct embryos with Cre mRNA to induce global inversion of Zwitch and deletion of betaPix. As expected, we found Cre-induced precise inversion in betaPix^m/m as confirmed by Sanger sequencing (Figure 1C) and severe brain hemorrhages developed in the time window similar to bbh^fn40a mutants (Figures 1D and 1E). Consistently, qPCR analysis showed that betaPix decreased and RFP increased in heterozygous betaPix^m/+ and homozygous betaPix^m/m mutants compared with betaPix^ct/ct wild-type siblings, while alphaPix expression was not affected (Figure 1F). Of note, TagRFP expression in betaPix^m/+ (after Cre-mediated recombination in betaPix^ct/+ embryos) reflected endogenous betaPix expression in the boundary of midbrains and hindbrains as well as the hindbrains (Figure S1F and S1G). Together, these results suggest that the conditional trap betaPix^ct/ct zebrafish is functional for visualizing endogenous betaPix expression (knockin) and loss-of-function study (conditional knockout).

betaPix^m/m mutant has brain hemorrhages, central arteries defects and abnormal glial structure that is partially rescued by Pak1 inhibitor IPA3 treatment

The main phenotypes in bbh mutants consist of brain hemorrhage and hydrocephalus as early as 36 hpf, as well as poor endothelial–mesenchymal contacts and defective central arteries (CtAs) sprouting in the hindbrain ^10,17. To demonstrate the practicability of betaPix conditional trap lines, we assessed vasculature and glia in the hindbrain by using light sheet fluorescence microscopy. Global betaPix inactivation in betaPix^m/m mutants resulted in severe cerebral hemorrhages and defective angiogenesis in hindbrain central arteries (Figures 1D, 1E, S2A and S2B), which is consistent with previous studies ^10,18.

It has been suggested that endothelial betaPix might not contribute to the occurrence of brain hemorrhage ¹⁰. During brain development, neurons and glia are important perivascular cells and they orchestrate with endothelial cells in a temporal and spatial pattern. Thus, we studied hindbrain radial glia development and found that radial glial structure had disoriented arrangement and shorten process length in betaPix^m/m embryos compared to siblings (Figures 2A, 2B and S2C). As a result, neuronal and glial precursors marker nestin increased while differentiated neuronal marker pax2a decreased in betaPix^m/m embryos, suggesting delayed neuronal development and differentiation after betaPix inactivation (Figure 2C). By using global betaPix knockout with multiple guide-RNA simultaneously ²³, we found that CRISPR-induced betaPix mutant F₀ embryos had almost identical phenotypes to those in betaPix^fn40a mutants (Figure S3A-S3B), such as cerebral hemorrhages, central arteries defects, and abnormal hindbrain glia and neuronal precursor development (Figures S3C to S3G), further confirming the loss-of-function phenotypes in betaPix^m/m mutants.

Figure 2.

Zebrafish central arteries vascularization in the hindbrains starts at 29 hpf. Tip cells start to form in the dorsal side of the primordial hindbrain channels (PHBC), sprout and migrate into the hindbrain tissue, moving towards the center around 36 hpf to establish connections with the basilar artery. A characteristic arch architecture forms at 48 hpf as boundaries dividing the hindbrain into rhombomeres²⁴. In betaPix knockouts, CtA sproutings were restricted to the bottom of glial processes, thus not being able to migrate upwards to form an arched structure (Figure 2D). The transcription factor SRY box-2 (Sox2) is another marker gene for neuron and glial precursors. Sox2-positive precursor cells tightly wrapped around the central artery in control siblings, but Sox2 precursor cells had loose contacts with the central arteries, with larger perivascular distances in betaPix mutants (Figure 2E). These results implicate the impaired interactions between endothelial cells and neuronal/glial cells during development after betaPix deletion.

p21-activated kinase (Pak) is a binding partner for betaPix. Pak2a has been shown mediating downstream signaling in bbh^m292 mutants. Group I PAK allosteric inhibitor IPA-3 covalently modifies and stabilizes the autoinhibitory N-terminal region of PAK1, PAK2 and PAK3 ²⁵. As expected, IPA-3 treatment significantly decreased incidence and intensity of cerebral hemorrhages in betaPix^m/m mutant embryos (Figure 2F and 2G), while IPA-3 treatment had no effect on hematopoiesis in betaPix^ct/ct control siblings. In addition, central arterial defects were partially rescued in betaPix^m/m mutant embryos after IPA-3 treatment, showing increased endothelial protruding into the hindbrain and more arch structure formation (Figure 2H and 2I). IPA-3 treatment also decreased abnormal glial process arrangements, which the lengths of glial processes statistically reached the levels of the control siblings (Figure 2J). Thus, these data suggest that the betaPix-Pak1/2 signaling regulates brain vascular integrity during development.

Glial-specific betaPix knockouts recapture its global knockout phenotypes

To investigate betaPix function in different types of perivascular cells, we generated glial- or neuronal-specific transgenic lines using either glial fibrillary acidic protein (gfap) ²⁶ or huC ²⁷ promoters to drive EGFP-Cre fusion gene expression under betaPix^ct/ct background, respectively (Figure S4A). EGFP signals reported Cre expression while RFP signals efficiently reported Cre-induced gene trap cassette inversion and disruption of betaPix expressions (Figure S4B to S4F). Interestingly, glial-specific betaPix knockouts developed severe cerebral hemorrhages and abnormal central arteries vascularization, which were partially rescued by IPA-3 treatment (Figure 3A to 3C), highlighting betaPix function in glia via Pak1/2 signaling. On the other hand, neither vascular-, neuronal-, nor mural-specific deletion of betaPix had evident betaPix mutant phenotypes (Figure S5A to S5D). These results suggest that glial betaPix plays crucial roles in zebrafish embryonic vascular integrity and glial development via regulating Pak activities.

Figure 3.

Single-cell transcriptome profiling reveals that stathmins in gfap-positive progenitors were affected in betaPix knockouts

To investigate the interplays between glial and hemorrhagic pathology caused by betaPix loss-of-function, we profiled the cranial tissues of CRISPR-edited zebrafish at 1 and 2 dpf using the 10X Chromium single-cell RNA sequencing (scRNA-seq) platform (Figure 4A). Multi-guide targeting was able to generate almost 100% F₀ null mutants, allowing us to select mutant embryos before brain hemorrhages from 36 hpf. Low-quality cells were excluded based on the numbers of genes detected and percentages of reads mapped to mitochondrial genes per sample. A total of 38,670 cells passed quality control and were used for subsequent analyses. Uniform manifold approximation and projection (UMAP) identified 71 cell clusters, which represented 24 zebrafish cranial cell types based on known marker gene expression profiles (Figure 4B and S6A). Enriched gene markers were compared with previously annotated gene markers in the ZFIN database and literatures^28,29. By comparing the proportion of cells in each sample, we found that most neuronal clusters had increased trends, while the numbers of progenitors, endothelial cells, erythrocytes, neural crests, muscles, cartilages, retinas (photoreceptor precursor cells), olfactory bulbs, or epidermis and pharyneal arches were reduced in betaPix knockout heads at 2 dpf (Figure 4C and S6B). This is consistent with the open public databases of single-cell transcriptome atlas where betaPix weakly expressed in a broad range of cells including glia, neurons and neuronal precursors, retinas, endothelial cells, pharynx, exocrine pancreas, olfactory cells and heart cells ³⁰. Given that betaPix is critical in glia during brain vascular integrity development (Figures 3 and S5), we examined the gfap expression among each cluster, and found relatively high gfap expression in clusters including progenitors, hindbrain, ventral diencephalon, ventral midbrain and floor plate (Figure S6A, arrow indicates). We next focused on the progenitor cluster, as of its enriched gfap expression and relatively more decreased cell numbers in betaPix mutants. Furthermore, this progenitor cluster exhibited high-level expression of cell proliferation and cell cycle related genes (mki67, pcna, ccnd1, rrm1 and rrm2) as well as key glial-associated genes (gfap, fabp7a, her4.1, cx43, id1, fgfbp3, atp1a1b and mdka) (Table S1). From differentially expressed genes (DEGs) in this progenitor cluster between controls and betaPix knockouts at 2 dpf, enriched gene ontology (GO) terms revealed three major categories: epigenetic remodeling including chromatin silencing, negative regulation of transcription, chromatin packaging and nucleosome positioning; microtubule organizations including microtubule depolymerization, regulation of microtubule polymerization and microtubule cytoskeleton organization; and neurotransmitter secretion/transportation (Figure 4D). Of these signaling genes, we were particularly interested in microtubule genes for further studies.

Figure 4.

Stathmin acts downstream of betaPix in glial migration via regulating tubulin polymerization

Microtubules are essential cytoskeletal elements composed of α/β-tubulin heterodimers. The regulation of microtubule polymerization affects important cellular functions such as mitosis, motor transport, and migration ³¹. Based on the above scRNA-seq data (Figures 4E and S6C), we then examined if microtubule polymerization genes stathmins were affected in betaPix mutants. We found that stathmin1a (stmn1a), stathmin1b (stmn1b), and stathmin4l (stmn4l) were relatively abundant in both the whole brain and the progenitor cluster, and decreased specifically at 2dpf in betaPix knockout progenitor cells (Figure 4E). Subsequent qRT-PCR revealed that these stathmins expression decreased in both global betaPix knockouts (betaPix^m/m) and glial-specific betaPix knockout glia (gfap:EGFP-Cre;betaPix^ct/ct) (Figure 4F and 4G), which was further verified by RNA in situ hybridization analysis (Figure 4H). We next investigated whether Stathmins act downstream to betaPix signaling on regulating vascular integrity development. To this end, we found that Pak1/2 inhibition enabled to rescue down-regulated expressions of stmn1a, 1b and 4l in either global or glial-specific betaPix knockout zebrafish (Figures 5A, 5B and S7A), supporting previous findings that stathmin-1 acts downstream to betaPix-Pak signaling in neurite outgrowth in mice³².

Figure 5.

To test whether Stathmin family genes are important for vascular stability, we utilized CRISPR-induced F₀ knockout system simultaneously targeting the three stathmin genes. While singular stathmin gene knockout led to no evident phenotypes (data not shown), we found that simultaneously inactivating stmn1a, stmn1b and stmn4l caused brain hemorrhages and hydrocephalus in the brain, similar to that from betaPix^fn40a and betaPix^m/m mutants (Figures 5C, 5D, S7B to S7E). By labeling both glia and vasculature with Tg(gfap:GFP; kdrl:mCherrry) transgenic embryos, we found that triple stathmin gene knockouts had impaired CtAs, abnormal glia and neuronal development in the hindbrain region (Figure 5E, 5F, S7F to S7H). Furthermore, overexpressing stmn1b by the gfap promoter partially rescued brain hemorrhages and delayed neuronal development in bbh^fn40a mutants (Figures 5G, 5H, S7I and S7J), of which glial processes elongation also improved, but CtAs defects failed to restore (Figures 5I, 5J and S7K). Thus, these data support the critical role of stathmins in glia.

To examine the direct role of betaPix-stathmin in glia, we turned to use human glioblastoma cell lines and manipulated the betaPix levels by siRNA knockdown or transgenic overexpression. We found that betaPix knockdown attenuated glial migration (Figures 5K, 5L, S7L to S7N), which is consistent with previous studies on betaPix function in the endoderm ³³, epithelial cells ³⁴, neurons ³⁵ and multiple carcinoma cell lines including glioblastoma ^36–38. Either betaPix or Stmn1 overexpression led to a partial restoration of the impaired glial migration, suggesting Stathmins as downstream effectors of betaPix. Furthermore, immunofluorescence analysis revealed that betaPix knockdown led to abnormal accumulation of microtubules at the cell periphery, whereas ectopic expression of betaPix and Stathmin partially rescued these microtubule defects (Figures 5M and 5N). Together, these results suggest that betaPix depletion disrupts microtubule dynamics and leads to impaired glial development acting upstream to the Pak-Stathmin signaling, subsequently disrupting vascular integrity and resulting in brain hemorrhages. While the brain hemorrhages were only partially rescued and abnormal CtAs defects failed to restore, indicating that additional pathways may act downstream of betaPix.

Zfhx3/4 acts downstream of betaPix in regulating vascular integrity development

Previous studies have identified multiple signaling pathways important for zebrafish hindbrain angiogenic sprouting, such as VEGF, Notch, Wnt, Integrin β1, angiopoietin/TIE, and insulin-like growth factor signaling ^39–42. To investigate whether angiogenic signal is disrupted by betaPix depletion, we examined several angiogenic gene expression pattern by performing RNA in situ hybridization of embryos at 36 hpf. The expression level of Vegfaa in sprouting CtAs decreased in bbh^fn40a mutants (Figure 6A). Consistently, betaPix knockdown significantly reduced VEGFA expression in cultured glial cells, which was rescued by ectopic expression of betaPix (Figure 6B; right panel). To explore how betaPix regulate Vegf signaling in glia, we also found a significant enrichment of epigenetic and transcriptional regulation in progenitor cluster indicated by the GO analysis of single-cell sequencing data (Figure 4D). Among these candidate genes, transcription factor Zinc Finger Homeobox 3 (ZFHX3) is known to mediate hypoxia-induced angiogenesis in hepatocellular carcinoma via transcriptional activation of VEGFA ⁴³. Furthermore, ZFHX3 has been associated with stroke in multiple genome-wide association studies ^44–46. Interestingly, we found that transcription factors Zfhx3 and Zfhx4 significantly downregulated after betaPix inactivation in cultured glial cells (Figure 6B), CRISPR-induced betaPix mutants at 2 dpf (Figure S8A), and glia-specific betaPix mutants (Figure S8B), thus suggesting that betaPix acts upstream to Zfhx3/4-VEGFA signaling in regulating angiogenesis.

Figure 6.

Figure 7.

To determine whether Zfhx3/4 is critical for vascular integrity by interacting with betaPix in zebrafish, we depleted Zfhx3 and Zfhx4 simultaneously via CRISPR-mediated F₀ knockouts (Figure S8C to S8E). We found that zfhx3/4 knockouts decreased Vegfaa expression at sprouting CtAs at 36 hpf (Figure 6C). Zfhx3/4 knockouts at 48 hpf had under-developed CtAs and reduced penetration of endothelial cells into the hindbrain (Figure 6F and 6G), developed cerebral hemorrhages during 36 to 52 hpf, which are comparable with betaPix mutant phenotypes (Figure 6D-6G). Despite up-regulated expression level of nestin, glial arrangement and processes elongation remained statistically unaffected after Zfhx3/4 inactivation (Figures 6F, 6G and S8F). We next investigated whether increasing expression level of Vegfaa in betaPix mutants was able to rescue impaired vascular integrity. By using Zfhx4 mRNA microinjection into one-cell stage embryos in either bbh^fn40a mutants or CRISPR-induced betaPix knockouts, we found that ectopic expression of Zfhx4 increased Vegfaa expression level (Figure 6L), thus leading to a drastic reduction of both percentages and volumes of brain hemorrhage shown in these mutants (Figures 6H, 6I, S8G to S8I), as well as improved CtAs and glial development (Figures 6J and 6K). In addition, scRNA-seq data showed that CRISPR-induced betaPix knockouts had down-regulated angiogenic gene expression in a cluster of endothelial cells (Figure 6M). Together, these data support the critical role of Zfhx3/4 in vascular integrity and glial development probably acting upstream to VEGFA and downstream to betaPix.

Discussion

Conditional knockout technology in mice has been widely used to modify and determine targeted gene function in a cell-specific manner. Spatio-temporal specific modification is beneficial for studying embryonic lethality genes or investigating the function of targeted genes in specific cell populations. Despite being broadly utilized in mice, conditional knockouts in other animal models are limited due to technical difficulties. To this end, multiple research groups have designed different strategies for establishing floxed-mutant alleles, gene traps, and inducible Cas9 in zebrafish ⁴⁷. In particular, the Zwitch gene trap-based approach has been established successfully in zebrafish, enabling to simultaneously produce knockin reporter driven by the endogenous gene promoter and disrupting endogenous gene function ^19–21. We adapted this strategy with the aid of CRISPR/Cas9 technology, and added a GSG spacer to enhance cleavage efficiency and universal guide RNA target sites on both homologous arms for producing linearized targeting vector. Moreover, we also found that utilizing either short or long homologous arms enables to achieve precise targeted integrations with this donor vector. Our short arms-based method can be broadly adapted because of less restrictions on arm sequences and easily engineering the donor vector. From the betaPix locus and other five unpublished loci that we already generated Zwitch alleles with ∼1% correct insertions based on homologous recombination, we achieved Cre-induced, cell-specific deletion and functional analysis of targeted genes in both embryos and adult hearts. Therefore, our CRISPR-based Zwitch method can be widely applied for generating conditional mutant zebrafish in particular and possibly extended for other animal mutants in general.

Development of the vasculature in the central nervous system can be classified into three phases: vasculogenesis, angiogenesis and blood-brain barrier formation. The hallmark of CNS vasculature maturation is acquisition of BBB properties, establishing a stable environment crucial for brain homeostasis in response to extrinsic factors and physiological changes. The core features of the BBB include specialized tight junctions, highly selective transporters and limited immune cell trafficking ^48,49. In zebrafish, BBB can be functionally characterized from 3 dpf ^50,51. While bubblehead mutants acquire vessel rapture phenotype early before BBB maturation, presenting a good hemorrhagic model for studying early BBB development. Multiple types of cells contribute to diverse aspects of BBB development and maintenance, leading to proper vasculature function across developmental timeline. Pericytes are perivascular cells embedded within the same basement membrane with endothelial cells ⁵². Pericyte-deficient mice with disrupting PDGF-B signaling displayed BBB leakiness, reflecting altered pericyte coverage and endothelial gene expressions ^53–55. Pericyte-secreted signals promoting the endothelial survival, proliferation and sprout formation via multiple pathways such as transforming growth factor-β (TGF-β) signaling ^56–59, VEGF-VEGFR signaling ^60,61, Angiopoietin-Tie2 signaling ^62–64, Sphingosine-1-phosphate (S1P) signaling ⁶⁵ and Notch signaling ^66,67. Apart from endothelial cells, pericytes also regulate development of other members of the neurovascular unit. Pericytes regulate astrocyte aquaporin-4 polarization and guide end-foot processes toward endothelial tubes ^53,68. In addition, pericytes secrete ECM molecules supporting the basement membrane and regulation of the angioarchitecture ^69–72.

Endothelial cells form the continuous inner layer of blood vessels with junctional protein and selective transporters, which are the main subject of permeability regulation in BBB. It is well accepted that endothelial cells do not show a predetermined role and that the brain environment is sufficient to induce endothelial barrier ⁷³. Accumulative evidences also point to critical role of endothelial signals in regulating several aspects of BBB development. In particular, endothelial cell-derived PDGF-B directly binds to surface receptor PDGFRβ on pericytes that are critical for pericyte recruitment and function ^74,75. Although their barrier properties are non-assessed, angiocrine signals such as leukaemia inhibitory factor (LIF) and TGF-β induce both astrocyte ⁷⁶ and oligodendrocyte differentiation ⁷⁷, respectively. Wnt7a/b ligands secreted by neural progenitor cells bind to Frizzled receptors on endothelial cells, activating canonical Wnt signaling and downstream genes. WNT ligands knockouts lead to angiogenic defects and BBB leakages specifically in the CNS ⁷⁸. Endothelial GPR124, as a coactivator of canonical Wnt signaling, has important function in CNS-specific angiogenesis and BBB establishment ^79–81. Other signaling ligands such as angiopoietins and Ephrin family ligands orchestrate with Tie or EphB receptors on endothelial cells, respectively⁸². Briefly, complex interactions between BBB cell types orchestrate proper angiogenesis and CNS development in a spatio-temporal manner.

Compared with pericytes and endothelial cell signaling, we have less knowledge on glia signaling in the BBB development and maturation. Astrocytes ensheathe capillaries through polarized end feet that are enriched with aquaporin-4 (Aqp4) proteins, colocalized with inwardly rectified K⁺ channels^83,84. Functional coupling of ion and water fluxes plays a critical role in regulating local osmotic equilibrium. During development, radial glia is a neuroepithelial origin with heterogeneous populations that are able to generate neurons, astrocytes, and oligodendrocytes^85–88. In zebrafish, radial glia has long been considered serving an astrocytic role. Radial glia in zebrafish enriches with several key molecular markers for astrocytes and tight junctions while Aqp4⁺ radial glial processes rarely contact the vasculature⁸⁹. A more recent article has reported that zebrafish Glast-expressing radial glia transform into astrocyte-like cells, displaying dense cellular processes, tiling behavior and proximity to synapses ⁹⁰. Whether end feet of these astrocytes enwrap capillary blood vessels and resemble mammalian astrocytes warranty future investigations. In agreement with mammalian astrocytes, two independent groups have shown that ablation of pan-glia results in progressive brain hemorrhage^91,92 or deficiency of spinal cord arteries development in zebrafish ⁹³, demonstrating important role of zebrafish glia in BBB formation. Multiple glia/astrocyte-derived signals have been shown to contribute to endothelial barrier properties development. Astrocyte-expressed sonic hedgehog (Shh) bind to hedgehog (Hh) receptors on endothelial cells and contribute to the BBB functions by promoting junctional protein expression and the quiescence of the immune system ^94–97. Before BBB formation, notochord-derived Shh activity promote arterial cell fate on developing endothelial cells ⁹⁸. Shh stimulates the production of angiogenic cytokines, angiopoietins and interleukins, via downstream transcription factor Gli or non-canonical pathways crosstalks with iNOS/Netrin-1/PKC, RhoA/Rock, ERK/MAPK, PI3K/Akt, Wnt/β-catenin, and Notch signaling pathways ⁹⁹. VEGF is the main pro-angiogenic growth factor produced by various cell types that controls multiple vascular development steps, such as proliferation, migration, permeability and survival via interacting with VEGF receptors on endothelial cells. Inactivation of VEGFA in astrocytes leads to BBB disruption in multiple sclerosis models ^100,101. In early development of avascular neural tube, VEGF derived from neuroectoderm serve as a central role in inducing tip cells sprouting from the perineural vascular plexus (PVNP) and formation of vasculature lumens from stem cells ^102,103. Ablating neuroglia reduces vegfab signals and leads to ectopic intersegmental vessels sprouting and vertebral arteries in zebrafish ⁹³. VEGF signaling orchestrates elaborately with other signaling pathways such as delta-like 4 (Dll4)/Notch signaling ¹⁰⁴ and activates downstream components including the Ras/Raf/MEK, PI3K/AKT, and p38/MAPK/HSP27 pathways ¹⁰⁵. Our data suggest glial betaPix as an important factor in vascular integrity development. Together with previous studies, betaPix-Pak1-stathmin signaling regulates glial growth and differentiation³² and betaPix regulates Vegfaa secretion via downstream transcription factors Zfhx3/4. Thus, this work presents a novel glial betaPix signaling in regulating zebrafish BBB development.

Basement membrane is a non-cellular component consisting of several extracellular matrix (ECM) proteins, which laminins are the most abundant ones, while collagen IV, fibronectin, and perlecan serve as hub for structural supports, anchoring and intercellular communication. Components of basement membrane are synthesized in surrounding cells with diverse expression patterns ¹⁰⁶. For example, mice with deletion of Collagen IV in either brain microvascular endothelial cells (BMEC)- or pericytes lead to fully penetrant intracerebral haemorrhage, while COL4A1 deletion in astrocytes results in mild haemorrhage phenotypes ⁶⁹. Differential cell-specific expression pattern of laminin isoforms contribute differently to BBB formation ¹⁰⁷. As astrocytic laminins mediate pericyte differentiation and regulate capillary permeability ¹⁰⁸, intercellular interactions with the ECM occur by cell surface integrin receptors. Integrin β1 on endothelial cells are critical during angiogenesis¹⁰⁹. On the other hand, integrin αvβ8 has glial-specific enrichment which mediate focal adhesion to modulate vascular integrity^17,110,111. Notably, degradation of ECM and focal adhesions at the perivascular space is responsible for the hemorrhagic phenotypes of bbh mutants ^17,18. It would be of great interest to determine whether glial betaPix account for ECM breakdown in bbh mutants in future studies.

Materials and Methods

Fish maintenance

Zebrafish were raised and handled in accordance with the guidelines of the Peking University Animal Care and Use Committee accredited by the AAALAC. The bbh^fn40a mutant was isolated from a large-scale mutagenesis screen of the zebrafish genome at Massachusetts General Hospital, Boston⁹. Tg(acta2:GFP-Cre) was purchased from Xinjia Pharmaceutical Technology Co., Ltd (Nanjing, China); Tg(neurod:EGFP) were kindly provided by Dr. Bo Zhang (Peking University, China); Tg(huC:GFP) were kindly provided by Dr. Liangyi Chen (Peking University, China); Tg(kdrl:GFP) ¹¹² and Tg(kdrl:mCherry) ¹¹³ were described previously.

Cell culture and transfection

U251 and A172 human glioblastoma cell lines, kindly provided by Dr. Jian Chen at Chinese Institute for Brain Research, Beijing, were cultured in high glucose DMEM medium (Hyclone) supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin, at 37°C in 5% CO₂ humidified incubator. The betaPix siRNAs and STMN1 siRNAs were purchased from GenePharma (Shanghai, China). The sequences of the siRNAs are listed in Table S2. Transfection was conducted with Lipofectamine 3000 (Invitrogen) at a final siRNA concentration of 30 nM.

mRNA/gRNA synthesis and microinjections

mRNA and gRNA were synthesized as described previously ¹¹⁴. In brief, linearized pT3TS-nls-zCas9-nls ¹¹⁵, pXT7-Cre, and pXT7-Zfhx4 plasmid DNA were purified as templates using TIANquick Mini Purification Kit (TIANGEN, DP203, China). Next, in vitro transcription reactions were performed using the mMESSAGE mMACHINE T3 (for pT3TS-nls-zCas9-nls) or T7 (for pXT7-Cre and pXT7-Zfhx4) kits (Life Technologies) according to manufacturer’s instructions. Templates for gRNA were generated by complementary annealing and elongation of two oligos. Forward oligo contained a T7 promoter and gRNA target sequence, and reverse oligo contained the universal gRNA scaffold ²³. The resulting double-stranded DNA served as the templates for in vitro transcription using HiScribe® T7 High Yield RNA Synthesis Kit (NEB). The sequences of the gRNAs are listed in Table S3. For global betaPix inactivation, Cre mRNA with a working concentration of 200 ng/µL were injected into the betaPix^ct/ct embryos at one-cell stage. For CRISPR-mediated knockouts, 300 ng/µL Cas9 mRNA and 20-50 ng/µL gRNAs were injected into zebrafish embryos at one-cell stage. For Zfhx4 rescue experiments, Zfhx4 mRNA with a working concentration of 400 ng/µL were injected into the bbh^fn40a mutants or in combination with CRISPR-mediated betaPix knockout system at one-cell stage. Injected embryos were scored for mutant phenotypes at 36 or 48 hpf.

Construction of betaPix conditional knockout zebrafish

The pZwitch plasmid clone was kindly provided by Dr. Kazu Kikuchi (National Cerebral and Cardiovascular Center Research Institute, Suita, Japan). GSG spacer was added to pZwitch+3 between P2A and TagRFP coding sequences by overlapping primer pairs. Next, a highly efficient gRNA was found from the fifth intron of betaPix. Primer pairs were designed to amplify the left and right homologous arms from the gRNA site for 1000 bp and 24 bp with following modifications: 1) The 5’ end of the forward primer of the left homologous arm were added with an Nhe1 site and a universal gRNA site which provided an in vivo linearization cleavage site; a Mlu1 site to the 5’ end of the reverse primer of the left arm. 2) Added EcoR1 site to the 5’ end of the forward primer of the right homologous arm; added Xho1 site and universal gRNA site to the 5’ end of the reverse primer of the right homologous arm. Then we cloned the inverted homologous arms into polyclonal sites of the modified pZwitch+3, and purified plasmid DNA with EndoFree Mini Plasmid Kit (Tiangen, DP118). In vivo knock-in system was composed of 300 ng/µL Cas9 mRNA, 50 ng/µL betaPix-intron5 gRNA, 50 ng/µL Universal gRNA, and 20 ng/µL donor vector. Microinjected in vivo knock-in system into wildtype embryos at one-cell stage. Fish founders were screened for α-crystallin reporter expression, raised to adulthood, and re-screened for germline transmission and precise knock-in by Sanger sequencing.

Construction of transgenic reporters

The glial reporter plasmid gfap-EGFP ²⁶ was kindly provided by Dr Jiulin Du (Shanghai Institute for Biological Sciences, Shanghai, China). The gfap promoter was used to establish Tg(gfap:GFP-Cre) plasmid. The stmn1b coding sequences were amplified from wild-type zebrafish cDNA library, which were used to establish Tg(gfap:GFP-stmn1b) plasmid. Transgenic reporters were generated by using Tol2-based transgenesis ¹¹⁶. In brief, 100 ng/µL Tol2 transposase mRNA in combination with 20 ng/µL donor plasmid DNA were co-injected into the embryos at one-cell stage. Transgenic founders were screened for specific transgenic GFP expression, raised to adulthood, and re-screened for germline transmission.

Inhibitor treatment

IPA-3 (Proteintech, CM05727) were dissolved in DMSO to form a 10 mM stock, and then diluted with E3 medium to 3 μM in 6-well plates. Zebrafish embryos were treated from 24 to 48 hpf, and then washed with E3 medium for phenotype analysis.

o-dianisidine staining

o-dianisidine staining was performed as described previously ¹⁸. In brief, embryos were dechorionated, anesthetized and incubated in fresh staining solution (0.6 mg/mL o-dianisidine, 0.01 M sodium acetate pH4.5, 0.65% H₂O₂ and 40% Ethanol) for 20 minutes in dark. Washed three times with methanol, and performed benzyl alcohol/benzyl benzoate (BABB) tissue clearing before imaging with stereo microscope (Leica, 160F).

Whole-mount in situ hybridization (WISH)

Whole-mount in situ RNA hybridization was performed as described previously ¹¹⁷. Antisense probes were synthesized using a digoxigenin RNA labeling kit (Roche, 11277073910). Primer sequences for all WISH probes used in this paper are provided in Table S4.

Quantitative real-time RT-PCR

Total RNA was isolated from embryos using Trizol (invitrogen) and cDNA was generated using HiScript® III RT SuperMix (Vazyme) according to the manufacturer’s instruction. Quantitative real-time PCR was performed using Lightcycler (Roche) and ChamQ SYBR qPCR Master Mix (Vazyme). Primer sequences are listed in Supplementary Materials Table S5. Gene expression level was normalized against GAPDH level.

Light-sheet fluorescence microscopy imaging

Light-sheet microscopy imaging was performed as described previously ¹¹⁸. In brief, transgenic zebrafish embryos were collected and maintained at 28.5°C. At the required developmental stages, the embryos were carefully dechorionated, paralyzed with tricaine, and transferred into 1% ultrapure low melting point agarose (16520-050; Invitrogen). The embryo was then drawn into a glass tube in top-down position using a 1 mL syringe with an 18G blunt needle. After agarose coagulation, a wire was inserted from bottom to push the agarose with embryo upwards, removed excess agar until the head area is exposed from the top of the glass tube. Finally, the glass tube was fixed in a sample holder for subsequent imaging.

Imaging was carried out with Luxendo Multi-View Selective-Plane Illumination Microscopy (Bruker). Optical calibrations were adjusted according to the instruction manual.

The optical plates are emitted from two Nikon CFI Plan Fluor 10x W 0.3 NA immersion objectives in opposite directions. The detection is completed by two Olympus 20x 1.0 NA immersion lenses. The parameters are set as follows: the green fluorescence channel uses a laser of 488 nm coupled with BP497-554 filter, the red fluorescent channel uses laser 561 nm coupled with BP580-627 filter, with an exposure time of 100 ms, a delay of 11 ms, a line mode of 50 px, and z-Stack interval of 3 μm. The raw data were stored in H5 file format, processed into TIFF format and merged Multi-View dataset using MATLAB software. After pre-processing, multi-fluorescence channels were merged in ImageJ, and 3D visualization and measurement were performed by Imaris.

Single-cell RNA sequencing

For knockout samples, 300 ng/µL Cas9 mRNA and a mixture of number 1 to 4 betaPix gRNAs with 50 ng/µL each were co-injected into wildtype embryos at one-cell stage. Injection with equivalent concentration of Cas9 mRNA with PBS served as siblings. After microinjection, embryos were collected and maintained at 28.5°C. At 24 or 48 hpf, zebrafish were dechorionated, paralyzed and transferred onto agarose plate. Heads were harvested with dissecting scissors in cold sterile 1X PBS with pooling 200 heads for each group. Heads were then dissociated in 900 μL Accutase cell detachment solution (Sigma-Aldrich, A6964) at 28.5°C for 3 hours and re-suspended by pipetting every 30 min. Once digestion was complete, 100 μL FBS was added to cell suspension and centrifuged at 4℃, 500 g for 3 min. The cells were gently re-suspended in cold 500 μL 2.5% FBS in 1X PBS and filtered through a 40 μm strainer. PI and Hoechst33342 were stained for distinguishing living cells. The single, living cells were sorted by Aria SORP (BD biosciences) into 1.5 ml tubes. The cell counts and vitality were verified by AOPI staining coupled with an Automated Cell Counter (Countstar BioTech). Single cells were barcoded with Chromium Next GEM Single Cell 3 ‘Reagent Kits V3.1 kit (10x Genomics, 1000269) in 10× Chromium Controller (10× Genomics). After qualified by peak shapes, fragment size and tailing with Fragment Analyzer System kit (Agilent Technologies, DNF-915), single-cell transcriptome libraries were sequenced via Illumina High Throughput Sequencing PE150 (Novogene, Beijing). The sequencing data were analyzed using the CellRanger-6.1.1 (10x Genomics) and mapped to reference genome GRCz11-GRCz11.103.

Cells were excluded from subsequent analyses under the following conditions: when the number of expressed genes was less than 500, when there were abnormally high counts of UMIs or genes (outliers of a normal distribution), or when the mitochondrial content exceeded 9%. Unsupervised clustering was performed using Seurat (version: 4.0.2) with a resolution of 2.5, resulted in 71 cell populations and further annotated into 24 zebrafish major cranial cell types. Differential expression analysis in Seurat v4 was used to identify cluster/cell type markers by Wilcoxon rank sum test.

Scratch assay

U251 cells were planted into culture plate and performed transfection at the desired density. At 24 hours post-transfection, we vertically scratched monolayer cells by using a 200 μL pipette tip. Washed the cells three times with PBS, then replaced with serum-free culture medium for further culture. Stereo fluorescence microscopy (Leica, 160F) was used to document the scratch size at 0 hours, 18 and 24 hours. Images were processed by ImageJ.

Immunostaining

For assessing tubulin expression, U251 cells were planted into chamber slides (Saining, 1093000) and transfected at the desired density. At 24 hours post-transfection, washed the cells once on ice with pre-cooled PBS, then fixed with pre-cooled methanol at −20℃ for 20 minutes. The fixed cells were permeabilized with pre-cooled acetone at −20℃ for 1 minute. Removed acetone, incubated 0.5% BSA/PBS blocking solution at room temperature for 15 minutes. Cells were then incubated with 1:200 anti-α-tubulin mouse monoclonal antibody (EASYBIO, BE0031) at 37℃ for 45 minutes. After rinsing once with PBS, cells were incubated with 1:400 Alexa Fluor™ 488 goat anti-mouse IgG (H+L) secondary antibody (Invitrogen, A11029) at 37℃ for another 45 minutes. Washed three times with PBS, and sealed with Mounting Medium with DAPI (ZSGB Bio, ZLI-9557). Images were acquired by upright fluorescence microscopy (Leica, DM5000B) at 40× magnification, and processed by ImageJ.

Statistical analysis

Statistical analysis was performed using Graphpad Prism 6. The statistical significance of differences between the two groups was determined by the independent unpaired Student’s t-test. Among three or more groups, one-way ANOVA analysis coupled with Dunnett’s test were used. All data are presented as the mean ± SEM. P value <0.05 indicates significant, with individual P values mentioned in the figure/figure legends.

Acknowledgements

The authors thank Drs. Bo Zhang, Jiulin Du, Liangyi Chen, Kazu Kikuchi, and Jian Chen for providing fish lines, plasmid clones, and human glial cell lines; the members of Dr. Jing-Wei Xiong’s laboratory for helpful discussions and technical assistance; and the National Center for Protein Sciences at Peking University, particularly Dr. Liying Du at the Flow Cytometry Core for technical help on the Beckman Coulter MoFlo XDP; and Dr. Hua Liang at National Biomedical Imaging Center, Peking University for assistance with the Luxendo Multi-View Selective-Plane Illumination Microscopy (Bruker). This work is supported by grants from the National Key R&D Program of China (2019YFA0801602 and 2023YFA1800600); the National Natural Science Foundation of China (32230032 and 31730061).