Leucine-rich repeat containing 8A contributes to the expansion of brain ventricles in zebrafish embryos

ABSTRACT The sodium osmotic gradient is necessary for the initiation of brain ventricle inflation, but a previous study predicted that organic and inorganic osmolytes play equivalently important roles in osmotic homeostasis in astrocytes. To test whether organic osmoregulation also plays a role in brain ventricle inflation, the core component for volume-regulated anion and organic osmolyte channel, lrrc8a, was investigated in the zebrafish model. RT-PCR and whole-mount in situ hybridization indicated that both genes were ubiquitously expressed through to 12 hpf, and around the ventricular layer of neural tubes and the cardiogenic region at 24 hpf. Knocking down either one lrrc8a paralog with morpholino oligos resulted in abnormalities in circulation at 32 hpf. Morpholino oligos or CRISPR interference against either paralog led to smaller brain ventricles at 24 hpf. Either lrrc8aa or lrrc8ab mRNA rescued the phenotypic penetrance in both lrrc8aa and lrrc8ab morphants. Supplementation of taurine in the E3 medium and overexpression csad mRNA also rescued lrrc8aa and lrrc8ab morphants. Our results indicate that the two zebrafish lrrc8a paralogs are maternal message genes and are ubiquitously expressed in early embryos. The two genes play redundant roles in the expansion of brain ventricles and the circulatory system and taurine contributes to brain ventricle expansion via the volume-regulated anion and organic osmolyte channels.


INTRODUCTION
The regulation and homeostasis of osmolality is a critical mechanism for all active cells as biological activities are generally in an aqueous environment with various organic and inorganic solutes. In general, cells influx osmolytes to trigger a regulatory volume increase (RVI) in hyper-osmotic environments, whereas efflux osmolytes trigger regulatory volume decreases (RVD) in a hypo-osmotic environment. The existence of a volume-regulated anion channel (VRAC) was first putatively proposed based on the observations of an outwardly rectifying Cl − current and consequently an RVD when cells were exposed to hypotonic solutions (Cahalan and Lewis, 1988;Hazama and Okada, 1988). Later, organic osmolytes such as taurine were considered accountable for at least half of the total RVD, and hence a putative volume-sensitive organic osmolyte/anion channel (VSOAC) was proposed for this activity (Garcia-Romeu et al., 1991;Jackson and Strange, 1993;Strange and Jackson, 1995). Due to the pharmacological similarity and controversial experimental results, the question of whether VRAC is identical to VSOAC was once a highly debated issue (Díaz et al., 1993;Kirk and Kirk, 1994;Lambert and Hoffmann, 1994;Sanchez-Olea et al., 1995;Shennan et al., 1994).
Despite the extensive studies on the properties of this channel for over 20 years, the molecular identity of this channel remained completely elusive until 2014. Two independent groups used a similar strategy and identified leucine-rich repeat containing 8A (LRRC8A) as an indispensable component in constituting VRAC in HEK293 cells (Qiu et al., 2014;Voss et al., 2014). In the human genome, the LRRC8 gene family contains five genes, including LRRC8A to LRRC8E, and shares some homology with pannexins, which constitute hexameric channels (Abascal and Zardoya, 2012). Current evidence suggests that the assembly of LRRC8A and LRRC8D recapitulates all the features of VSOAC, while the LRRC8A and other members of LRRC8 family (LRRC8B, LRRC8C and LRRC8E) mimic most of the features of VRAC except for the taurine efflux (Planells-Cases et al., 2015).
Osmoregulation has been shown to play critical roles in morphogenesis during embryonic development. For example, the cavitation of blastocysts requires the activity of a sodium pump (Wiley, 1984;Burgener-Kairuz et al., 1994). Additionally, the inflation of cerebral ventricles also requires the activity of a sodium pump (Lowery and Sive, 2005). Organic and inorganic osmolytes play equivalently important roles in osmotic homeostasis in astrocytes (Pasantes-Morales et al., 1993). Despite the demonstration that both polyols (such as inositol or sorbitol) and several amino acids and their derivatives (such as aspartate and glutamate) participate in RVD, possibly via VRAC/VSOAC (Banderali and Roy, 1992;Jackson and Strange, 1993), taurine has been the most epitomic organic osmolyte released via VRAC/VSOAC. Previous studies indicate that both taut, the gene coding for the transporter for cellular taurine intake, and csad, the gene coding for the key enzyme for taurine de novo synthesis, are maternal messages in zebrafish embryos (Chang et al., 2013;Kozlowski et al., 2008). A recent study shows two lrrc8a genes in the zebrafish genome, and both protein products act identically to the human LRRC8A protein in VRAC (Yamada et al., 2016). Interestingly, knockdown of either lrrc8aa or csad produces a similar cardiac abnormality (Yamada et al., 2016;Chang et al., 2013). In this study, we aimed to characterize the roles of two paralogous lrrc8a genes in embryonic development and test the hypothesis that Lrrc8a also contributes to the formation and inflation of brain ventricles by modulating organic osmolytes during embryogenesis.

Zebrafish has two lrrc8a genes
We confirmed that there were two genes in the zebrafish genome with protein coding sequences similar to human LRRC8A and these have been annotated as lrrc8aa (ENSDART00000148138) and lrrc8ab (ENSDART0000144732). Despite the similar amino acid sequences of zebrafish Lrrc8aa and Lrrc8ab compared to mammalian LRRC8A, only lrrc8ab has a similar genomic structure to human LRRC8A (Fig. 1A). Phylogenetic tree construction (Waterhouse et al., 2009) using a neighbour-joining method with percentage identity distances showed that both Lrrc8aa and Lrrc8ab are highly conserved throughout evolution (Fig. 1B). Interestingly, one of the LRRC8 family members, lrrc8b, was not found in the zebrafish genome, whereas two lrrc8d paralogs, annotated as lrrc8da and lrrc8db, were identified.
The quantitative expression profiles of the lrrc8 family during zebrafish embryogenesis can be retrieved in the published dataset ( Fig. 2A) (White et al., 2017). As previously described (Yamada et al., 2016), lrrc8aa is much more abundantly expressed during the first 24 h of zebrafish embryogenesis ( Fig. 2A). Among other family members, lrrc8c is the most abundantly expressed gene and the expression of both lrrc8da and lrrc8db were detectable after gastrulation ( Fig. 2A). To validate this result, RT-PCR was performed with β-actin (actb1) as the loading control (Fig. 2B). Interestingly, both lrrc8aa and lrrc8ab were detected in the freshly laid embryos (Fig. 2B). The PCR products of both genes can be detected at all developmental stages through 72 hours post fertilization (hpf ) indicating temporally ubiquitous expressions (Fig. 2B). Whole-mount in situ hybridization further demonstrated spatially ubiquitous expression of both lrrc8aa and lrrc8ab in the early embryos ( Fig. 2C-F). In 24 hpf embryos, both lrrc8aa (Fig. 2G,H) and lrrc8ab (Fig. 2I) are prominently expressed at the ventricular layer along the brain, as well as the cardiogenic region. In 48 hpf embryos, both genes can be detected in brain ventricles, retina, otic vesicles and pectoral fin buds (Fig. 2J,K,L). Interestingly, the expressions of both genes seem predominantly overlapped, suggesting that the two lrrc8a genes in zebrafish might play redundant roles during early embryogenesis.
Both lrrc8aa and lrrc8ab contributed to the development of circulatory system To investigate the physiological roles of lrrc8aa and lrrc8ab during zebrafish embryogenesis, two exon-intron boundary-targeted antisense morpholino oligos (MOs), lrrc8aa-MO e2i2 (aMO-e2i2) and lrrc8ab-MO e2i2 (bMO-e2i2) as well as two translationblocking MOs, lrrc8aa-MO TB (aMO-TB) and lrrc8ab-MO TB Fig. 1. Zebrafish have two lrrc8a paralogs. (A) Two homologous human LRRC8A genes were found in the zebrafish genome (lrrc8aa or swell1a, ENSDART00000148138; and lrrc8ab or swell1b, ENSDART0000144732), but only lrrc8ab has a similar genomic structure to mammalian LRRC8A. The exons are presented as boxes and the coding regions are shaded in green. MOs target sites are indicated by red arrows. CRISPRi guide RNA target sites are indicated by green arrows. (B) Phylogenetic tree constructed by neighbor-joining method with percentage identity distances indicating that both Lrrc8a paralogs are highly conservative throughout evolution. Lrrc8c and two Lrrc8d paralogs are also well-aligned in the phylogenetic tree. The predicted Lrrc8a protein sequence in Callorhinchus milii (Australian ghostshark, a cartilaginous fish) was used as an outgroup.
(bMO-TB) (Fig. 1A) were designed. To confirm the perturbation of mRNA splicing, RT-PCR flanking exon 1 to 3 of both lrrc8a genes showed that aMO-e2i2 and bMO-e2i2 effectively disrupted the splicing of lrrc8aa (aa*, Fig. 3A) and lrrc8ab (ab*, Fig. 3B) mRNA, respectively. Cloning and sequencing of the mis-spliced RT-PCR products showed that both MOs led to the elimination of The expression levels of lrrc8 family during zebrafish embryogenesis in transcripts per million (TPM) were obtained from an Expression Atlas. Lrrc8aa is much more abundant than lrrc8ab in the first 24 h of development. While lrrc8c is the most abundant family member during gastrulation; the expression levels are comparable after 24 hpf ( prim-5). (B) The sequencing result did not detect lrrc8ab in zygotes and at the two-cell stage, but RT-PCR with β-actin (actb1) as the loading control showed that both lrrc8a genes are maternal messages and can be detected at all developmental stages through to 72 hpf. (C-L) Whole-mount in situ hybridization of lrrc8aa (C,E,G,H,J,K) and lrrc8ab (D,F,I,L) showed that both genes are ubiquitously presented in the zebrafish embryo at 6 (C,D) and 12 (E,F) hpf. In 24 hpf embryos (G-I), both genes could be clearly detected at the ventricular layer of the brain as well as the cardiogenic region. In 48 hpf embryos (J-L), both genes could be detected at brain ventricles, retina, otic vesicles and pectoral fin buds. Scale bars: 200 µm. entire exon 2s, which contain the start codons for both genes and hence their protein products were presumably knocked-down.
Both lrrc8aa and lrrc8ab contributed to the morphogenesis of brain ventricle To elucidate the possible role of lrrc8a genes in the formation of brain ventricles, the same fluorescent TRITC-dextran (20 mg/ml) used for angiography was injected into the newly formed fourth brain ventricle in alive non-monstrous 24 hpf zebrafish embryos. The overlaid micrographic images clearly depicted smaller ventricular areas in the aMO-e2i2 and bMO-e2i2 morphants compared to the control morphants ( Fig. 4A).
To further characterize the abnormality in brain ventricle, 24 hpf morphants or untreated embryos were documented and the area representing the diencephalic/mesencephalic ventricle (DMv) was measured in ImageJ ( Fig. 4B,C). Injection of every lrrc8a MO resulted in smaller DMv area compared to control morphants and untreated controls dose-dependently ( Fig. 4B,C). The Shapiro-Wilk normality test indicated that the data for DMv of untreated controls are a normal distribution (P=0.5706). Statistically, the range of two standard deviations from the mean should include more than 95% of the population with a normal distribution. To calculate the phenotypic penetrance, we therefore arbitrarily defined any DMv area with a size smaller than the mean of the untreated control (0.006339 mm 2 ) minus two standard deviations (2×0.0009943) as significantly smaller (∼0.004350 mm 2 ) and phenotypically abnormal. The result showed that, among the alive and nonmonstrous embryos, the injection of lrrc8a MOs dose-dependently caused increased abnormal rate compared to the cMO (Fig. 4D,E).
To further validate the specificity of this smaller ventricle phenotype in two sets of aMO and bMO morphants, CRISPRi was used to knockdown the transcriptions of lrrc8aa and lrrc8ab mRNA. A pool of gRNA targeting the non-template strand near the transcriptional/translational starting sites was selected from CRISPRscan (Moreno-Mateos et al., 2015) (Fig. 1A). The mRNA encoding the catalytically de-activated Cas9 (dCad9) was coinjected with each gRNA targeting lrrc8aa or lrrc8ab with gRNA targeting green fluorescent protein (gRNA-eGFP) as control (Liao et al., 2015;Shalem et al., 2014). Quantitative real-time PCR indicated that two of the gRNA against lrrc8aa (a2 and a3 in Fig. 5A) and lrrc8ab (b1 and b4 in Fig. 5B) effectively reduced the targeted genes. With the absence of significant gross phenotype (Fig. 5C), the mixture of these gRNA targeting lrrc8aa (gRNA-aa, P<0.0001) or lrrc8ab (gRNA-ab, P=0.009) indeed significantly reduced the area size of DMv compared to the gRNA-eGFP control (Fig. 5D). However, the phenotypic penetrance was significantly observed only in the gRNA-aa group (28/61, P<0.0001), but not in gRNA-ab group (6/40) compared to the gRNA-eGFP control (1/51) suggesting that lrrc8ab might play a relatively minor role in ventricular morphogenesis compared to lrrc8aa.
Lrrc8aa and lrrc8ab play a redundant role in brain ventricle inflation To further confirm that the phenotypes observed in lrrc8a morphants were due to the decrease of the corresponding lrrc8a gene products, eGFP, lrrc8aa-IRES-eGFP or lrrc8ab-IRES-eGFP mRNA was injected with or without lrrc8a MOs. The mRNA used in this experiment did not result in any abnormality in brain ventricle when injected alone. In aMO-e2i2 morphants, the areas of DMv were significantly rescued with the addition of lrrc8aa-IRES-eGFP (a: 100 pg in Fig. 6A) as well as lrrc8ab-IRES-eGFP (b+: 150 pg in Fig. 6A). On the other hand, the addition of lrrc8aa-IRES-eGFP and lrrc8ab-IRES-eGFP mRNA partially and fully rescued the small DMv phenotype resulting from bMO-e2i2, respectively (bMO with a++ and b++ mRNA in Fig. 6A; a++: 200 pg, b++: 200 pg). The lrrc8aa or lrrc8ab mRNA reciprocally rescued the small DMv phenotype caused by the other paralogous lrrc8a gene (Fig. 6A) suggesting that these two genes act in a redundant fashion in the morphogenesis of brain ventricle.
LRRC8A is an indispensable component of VRAC (Qiu et al., 2014;Voss et al., 2014), which mediates organic osmolytes such as taurine to efflux, while taurine efflux has long been a signature for VRAC/VSOAC activity (Jackson and Strange, 1993;Lambert and Hoffmann, 1994;Shennan et al., 1994). Our previous study showed that the homeostasis of taurine in zebrafish embryos predominantly depends on de novo synthesis via csad (Chang et al., 2013). Reduced csad level resulted in pericardial effusion, which is similar in lrrc8a morphants reported previously and in this study (Yamada et al., 2016), and can be rescued by taurine supplementation (Chang et al., 2013). Additionally, embryonic taurine deficiency also led to PBI accumulation (unpublished data) and smaller brain ventricle (Fig. 6B), it is tempting to speculate that both lrrc8a genes contribute to ventricular inflation via VRAC activity by modulating the distribution of organic osmolytes such as taurine.
To test whether an organic osmolyte such as taurine plays a role in lrrc8a-mediated brain ventricle inflation, taurine was supplemented to the embryo medium to rescue the phenotypes of lrrc8a morphants. Indeed, supplementation of 6.25 mM taurine in the embryo medium successfully rescued the small brain ventricle phenotype in bMO morphants (Fig. 6C,E). Interestingly, although the same concentration of taurine failed to rescued aMO morphants, increasing the supplementation concentration to 50 mM successfully rescued the small brain ventricle phenotype in aMO morphants (Fig. 6C,D).
To further validate the role of taurine in brain ventricle inflation, a transgenic zebrafish line that ubiquitously overexpress csad was generated and lrrc8a or lrrc8b MOs were injected into these csad transgenic embryos. Consistent with the taurine supplementation, morphants with csad transgenic background moderately resist the small brain ventricle phenotype induced by lrrc8a or lrrc8b MOs compared to the wild-type background (Fig. 6F). Co-injection of Csad-IRES-eGFP mRNA with lrrc8a or lrrc8b MOs fully rescued the small DMv phenotype (Fig. 6F) indicating that taurine participates in the brain ventricle inflation via VRAC/VSOAC during zebrafish embryogenesis.

DISCUSSION
VRAC/VSOAC is currently considered constituted by LRRC8 family members with Lrrc8a as the indispensable core component (Qiu et al., 2014;Voss et al., 2014). We confirmed that there are two paralogous lrrc8a genes in the zebrafish genome (Yamada et al., 2016). The human chromosome 9 (Hsa9) includes multiple putative orthologs to the zebrafish linkage group 5 (LG5) and LG21 (Postlethwait et al., 2000). The loci of the zebrafish lrrc8aa on LG21 and lrrc8ab on LG5 were not evident in the previously reported locus of the human LRRC8A on Hsa9 (q34.11). The genomic structures of all human LRRC8 family members are similar to zebrafish lrrc8ab that the majority of the coding sequence is included within a single large exon; this was not the case for lrrc8aa (Fig. 1A). As a previous investigation suggests orthologs tend to retain similar genome structures (Xu et al., 2012), zebrafish lrrc8ab, compared to lrrc8aa, is probably evolutionarily closer to human LRRC8A.
We failed to find lrrc8b paralog in the zebrafish genome but two lrrc8d paralogs were found. Additionally, both lrrc8db and lrrc8c are located on LG6 immediately next to each other, whereas human LRRC8B, LRRC8C and LRRC8D locate on Hsa1 close to one another. Firstly, contrary to the proposal that LRRC8A and E arose by duplicating LRRC8B and C (Abascal and Zardoya, 2012), it is possible that LRRC8B was derived from lrrc8c or lrrc8d paralogs later in evolution. Secondly, as only LRRC8D is considered to contribute to VSOAC activity together with the essential channel component LRRC8A (Planells-Cases et al., 2015), the fact that only these two LRRC8 family members retain duplications in zebrafish might suggest the importance of organic osmoregulation in aquatic vertebrates such as zebrafish.
It is believed that, although polyploidy due to gene/genome duplication might be important for speciation and diversification, evolution tends to lead its way back to the diploid state through gene silencing and loss unless the duplicated genes are somehow mutated to introduce differences in temporal/spatial expressions or biochemical functions between paralogs (Gu et al., 2002;McLysaght et al., 2002;Force et al., 1999). The evidence provided previously and in this report demonstrate that lrrc8aa and lrrc8ab not only have similar temporal and spatial expression patterns (Fig. 2) and identical biochemical and cellular activities (Yamada et al., 2016), but also have similar biological function, as knocking down either one of the genes phenocopies the knockdown of the other (Figs 3-5). The overexpression of either gene can rescue the other one (Fig. 6A). In addition to expression and function, a later theory suggests that gene expression dosage might also be a reason for the retention of duplicated genes (Adams et al., 2003). Therefore, it is likely that the sum of total expression dosages of both lrrc8a paralogs is critical for early embryogenesis in zebrafish. Accordingly, as lrrc8aa is more abundantly expressed than lrrc8ab, the results of our knockdown and rescue experiments suggest that lrrc8aa plays a more dominant role than lrrc8ab (Fig. 6A). Moreover, the smaller brain ventricle phenotypes in lrrc8aa morphants could only be rescued by a higher dose of taurine supplementation than in lrrc8ab morphants (Fig. 6A). However, we cannot exclude the possibility that lrrc8aa and lrrc8ab began to express differently in specific tissues and cells later in the development or in the adult zebrafish.
Consistent with the previous report (Yamada et al., 2016), the most observable phenotype in lrrc8aa or lrrc8ab knocked-down morphants was pericardial effusion, that can be seen after about 28 hpf (data not shown). In addition, many of these morphants were defective in the extension of blood circulation (Fig. 3D). In principle, hemodynamics were modulated by two major factors: the distribution of blood volume and the regulation on the circulation such as by cardiac output. Although the initiation of both cardiogenesis and vasculogenesis seems to be genetically programmed (Isogai et al., 2003 al., 2008), it is clear that hemodynamic force feeds back and affects the continuation of cardiogenesis and vasculogenesis (Broekhuizen et al., 1999;Kowalski et al., 2012). In our study, we observed a significantly slower heart rate in both lrrc8aa morphants and lrrc8ab morphants. Since we did not observe significant alteration in early cardiac development markers such as cmlc2 at 24 hpf lrrc8aa morphants (data not shown), it is likely that lrrc8aa or lrrc8ab are not required for initial cardiogenesis.
The previous studies suggest that the formation of plasma during embryonic vasculogenesis and angiogenesis is via the formation and fusion of intracellular vacuoles (Kamei et al., 2006;Folkman and Haudenschild, 1980). As the fundamental driving force for the vacuole formation is not well understood, it is intriguing whether organic osmoregulation and VRAC contribute to the formation of hemodynamics by participating in the formation of embryonic plasma. Nonetheless, according to the previous study of North et al. (2009), it is possible that knockdown of either lrrc8aa or lrrc8ab perturbed the extension of blood circulation to PBI and in turn affected definitive hematopoiesis (Fig. 3). Interestingly, LRRC8A was first found in a human patient with congenital agammaglobulinemia due to the lack of peripheral B cells (Sawada et al., 2003), whereas a later study in mice also indicates that the LRRC8A participates in the homeostasis of lymphocytes (Kumar et al., 2014). It is not understood whether or not these results are due to the early effect over the fate commitment of definitive hematopoiesis.
Initiation of brain ventricle inflation depends on Atp1a1a.1, a sodium-potassium pump that is thought to shape an osmotic gradient to drive fluid flux into the brain ventricle (Lowery and Sive, 2005). In this study, we showed that knockdown of either one of the lrrc8a paralogous genes also resulted in deflated brain ventricles (Figs 4 and 5), but most of the morphants were with initial formation of the brain ventricle. Therefore, it is likely that lrrc8a paralogs are required for continual expansion of the brain ventricle but not initial formation. It is proposed that the continual expansion of the brain ventricle might be dependent on blood circulation, as most of mutant zebrafish with brain ventricle phenotypes also have heart or circulation phenotypes (Schier et al., 1996). In line with this reasoning and as previously discussed, knockdown of either one of the lrrc8a paralogs resulted in the abnormality in cardiac and circulatory phenotypes. However, as both lrrc8a paralogs are expressed at ventricular walls and the cardiac output track, it is also possible that the expansion of both circulatory system and brain ventricle partly require common mechanisms such as VRAC and other ionic/osmotic regulatory mechanisms, and the sum of multiple mechanisms accounts for the complete morphogenesis. Hence the expansion of both spaces are affected when one or more of these mechanisms was defective.
Taurine efflux is considered one of the epitomic features of VRAC/ VSOAC activities (Qiu et al., 2014;Voss et al., 2014). Although a volume-insensitive taurine efflux pathway via TauT is also proposed (Lambert et al., 2015), LRRC8A/LRRC8D constituted VRAC/ VSOAC is the only channel that is proven to efflux intracellular taurine to the extracellular compartment (Planells-Cases et al., 2015). Our previous study showed that the homeostasis of taurine in zebrafish embryos predominantly depends on de novo synthesis via csad (Chang et al., 2013). Reduced csad level results in pericardial effusion, which is similar in lrrc8a morphants reported previously and in this study (Yamada et al., 2016) and can be rescued by taurine supplementation (Chang et al., 2013). Additionally, embryonic taurine deficiency also led to PBI accumulation (unpublished data) and a smaller brain ventricle (unpublished data). It is possible that both lrrc8a genes contribute to ventricular inflation via VRAC/ VSOAC activity by modulating the distribution of organic osmolytes such as taurine. In line with this speculation, taurine supplementation ameliorated the smaller brain ventricle phenotype resulted from lrrc8aa and lrrc8ab knockdown (Fig. 6D,E).
Taken together, two zebrafish lrrc8a paralogous genes showed similar temporal and spatial expression patterns during early zebrafish embryogenesis, and contributed to the expansion of circulation and brain ventricle. It is likely that these two paralogs play redundant roles in brain ventricle expansion and organic osmolytes such as taurine contribute to this developmental procedure via VRAC/VSOAC.

Zebrafish husbandry
The AB wild-type zebrafish were housed at a density of two to four fish per 3 l tank in the aquatic facility with an automatic recirculation system. The system was maintained at 28.5°C with a light/dark cycle of 14/10 h, and the fish were fed with live adult brine shrimp twice a day (Wei and Liu, 2014). Embryos were collected after spontaneous spawning, allowed to develop in E3 medium and staged by hpf at 28.5°C using morphological criteria (Kimmel et al., 1995). For the rescue experiment, embryos were cultured in E3 medium with or without the supplementation of taurine (Sigma Chemical Co.) as previously described (Chang et al., 2013). All experimental procedures in this study were reviewed and approved by the Institutional Animal Care and Use Committee of National Taiwan University (NTU104-EL-00085 and NTU105-EL-00147) and were performed in accordance with the approved guidelines.

Molecular cloning
The total RNA of embryonic zebrafish was obtained as previously described . Briefly, 30 zebrafish embryos were homogenized in TRIzol Reagent (Life Technologies), mixed with 1-Bromo-3-chloropropane (Molecular Research Center) and then centrifuged at 12,000×g for 15 min at 4°C. The aqueous phase was collected, mixed with 500 µl of isopropanol, briefly incubated and then centrifuged at 12,000×g for 10 min at 4°C. The pellet was then washed with 75% ethanol, briefly air-dried and dissolved in DEPC-treated water. The single-stranded cDNA was synthesized from 2 µg of total RNA with random primers and High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA, USA) or SuperScript III Reverse Transcriptase (Invitrogen). The resulting cDNAs were used both for cloning and RT-PCR.
To clone zebrafish lrrc8aa and lrrc8ab, primers were designed using the NEBuilder Assembly Tool (New England Biolabs), and the coding sequences of both genes were cloned with the Q5 Hot Start High-Fidelity 2X Master Mix (New England Biolabs). The sequences of primer pairs are listed in Table 1. The resulting DNA fragments were then cloned into pT7-IRES2-EGFP (Chang et al., 2013) with SalI and BamHI to give rise to pT7-lrrc8aa-IRES2-EGFP and pT7-lrrc8ab-IRES2-EGFP, respectively. After amplification (ECOS 101, Yeastern, Taipei, Taiwan) and purification (Presto Mini Plasmid Kit, Geneaid, Taipei, Taiwan), the sequences of cloned genes were confirmed by a sequencing service (Center for Biotechnology, National Taiwan University, Taipei, Taiwan).
To construct a catalytically de-activated Cas9 (dCad9) that lacks endonucleolytic activity, pCS2-nzcas9n was used as template and two point mutations, D10A and H840A, were introduced into zebrafish optimized Cas9 to create the pCS2-nzdcas9n plasmid .

qRT-PCR
To quantitatively analyze the expression levels of lrrc8aa and lrrc8ab and the hematopoietic differentiation markers, cmyb and pu.1, 4 µl of 10× diluted cDNA was mixed with 5 μl iQ SYBR Green Supermix (Bio-Rad) and 1 µl of primer set mix to amplify the fragments of target genes at 95°C for 3 min, and 39 cycles of 95°C for 3 s and 60°C for 30 s followed by 60°C for 1 min with a thermocycler (Thermo Fisher Scientific). The sequences of primer pairs are listed in Table 1.

In situ hybridization
To demonstrate the spatial expression of zebrafish lrrc8a genes during embryonic development, whole-mount in situ hybridization was performed as described previously (Chang et al., 2013). Briefly, the DNA template for antisense digoxigenin-labeled riboprobes for lrrc8aa and lrrc8ab were generated by PCR and then synthesized by in vitro transcription using T7 polymerase. Zebrafish embryos were dechorionated, fixed with 4% paraformaldehyde in PBS and digested with proteinase K (10 μg/ml, Amresco, Solon, OH, USA) if older than 24 hpf. The embryos were then pre-hybridized for 3 h at 65°C without riboprobes and then hybridized with 50 ng RNA probe at 65°C overnight. After washing, hybridized embryos were blocked for 3 h at room temperature and incubated with Anti-Digoxigenin-AP Fab fragments (1:5000 in blocking solution; Roche Applied Science, Mannheim, Germany) with agitation at 4°C overnight. After washing, the hybridization signals were detected by NBT/BCIP solution (Roche Applied Science), observed and documented with a microscope (Leica Z16-APO).

Gene knockdown by MO and CRISPR interference
To knockdown Lrrc8aa or Lrrc8ab, antisense MOs were designed against exon2-intron2 splicing sites or translation start site of respective genes (Fig. 1A) and a standard control morpholino was used as the control (Gene Tools LLC, Philomath, Oregon, USA). All MOs were dissolved in distilled water to make a 2 mM stock and diluted to desire concentration with 0.5% phenol red (Sigma Chemical Co.) before use. The CRISPR interference (CRISPRi) was also used for gene-specific knockdown . A pool of guide-RNA (gRNA) targeting to non-template strands and near the transcriptional start sites of lrrc8aa and lrrc8ab was selected from CRISPRScan (Moreno-Mateos et al., 2015). A previously described gRNA-targeting eGFP was used as gRNA control (Shalem et al., 2014;Liao et al., 2015). Each gRNA was individually cloned into the BsmBI site of pT7-sgRNA, sequenced and linearized by BamHI for in vitro transcription using a MEGAshortscript T7 kit (Invitrogen). Microinjection was performed at the one-cell stage in embryos with 100 pg of nzdcas9n mRNA and 10 pg of gRNA mix per injection as previously described (Chang et al., 2013).

Microangiography and heartbeat
For microangiography, TRITC-dextran (20 mg/ml) was injected into the sinus venosus of the anaesthetized [0.016% of ethyl 3-aminobenzoate methanesulfonate (MS-222, Sigma-Aldrich)] 28∼30 hpf zebrafish embryos, and the fluorescent images of the embryos were documented (Leica DM2500) at 32 hpf (Chen et al., 2006;Zhong et al., 2006). For the analysis of heart rate, 20 embryos of each treatment group were videorecorded under the microscope (Leica Z16-APO) and the heartbeat was manually counted in slow playback.

Micrography and morphometrics for brain ventricle
To evaluate the inflation of the brain ventricle during embryogenesis, TRITC-dextran (20 mg/ml) was injected into the fourth ventricle of the anaesthetized 24 hpf zebrafish embryos, and the fluorescent images of the embryos were documented (Leica DM2500). For morphometric analysis, the area representing DMv (Fame et al., 2016) was manually depicted and calculated in ImageJ (Schneider et al., 2012).

Generation of csad transgenic fishline
The zebrafish ubiquitin B (ubb) promoter (Mosimann et al., 2011) and csad coding sequence with FLAG tag (Chang et al., 2013) were cloned into pDestTol2pA2 (Kwan et al., 2007) using Gibson Assembly Master Mix (New England Biolabs) resulting in ubb:csadΔstop-FLAG destination vector (Gibson et al., 2009). The primers used are listed in Table 1. One-cellstage embryos were injected with 50 pg of DNA constructs and 50 pg of transposase mRNA. The injected embryos were raised and genotyped by fin-clipping upon adulthood. The founder fish carrying transgenes were outcrossed to wild-type fish to generate F1 fish. Adult F1 fish were genotyped and the integration sites were identified by inverse PCR (Kotani et al., 2006). In brief, the genomic DNA of F1 fish was digested with BglII/ BamH1 or XbaI/NheI/AvrII/SpeI and ligated before undergoing nested PCR and sequencing. The ubb:csadΔstop-FLAG sequence was integrated in Chr24:28632092.28632099 (NC_007135.6) in the transgenic line used in this study. The F1 transgenic fish with the same integration site were intercrossed to generate homozygous F2 transgenic fish Tg (ubb:csadΔstop-FLAG).

Statistical analysis
The phenotypic penetrance, mortality, heart rate and gene expression level were subjected to Kruskal-Wallis test with Dunn's multiple comparisons. The area size of brain ventricle was analyzed by one-way ANOVA with Tukey's multiple comparisons test. All statistical analyses were performed using Prism 8 software (GraphPad), and P<0.05 was considered statistically significant. All data were presented as mean±s.e.m.