Functional antagonism of voltage-gated K⁺ channel α-subunits in the developing brain ventricular system

The brain ventricular system (BVS) is evolutionarily conserved in vertebrates and is formed by a series of inter-connected cavities filled with cerebrospinal fluid (CSF), surrounded by a neuroepithelium of ependymal cells. The BVS plays an essential role in regulating neurogenesis, brain function and homeostasis, and its dysregulation causes hydrocephalus (Zhang et al., 2006; Green et al., 2007; Tully and Dobyns, 2014) and has been linked to neurodegenerative diseases (Sakka et al., 2011). This has led to interest in animal models of hydrocephalus that recapitulate the disturbance of CSF circulation and show changes in neuroepithelium integrity (Chae et al., 2004; Chang et al., 2012; Jiménez et al., 2001). In the adult brain, CSF is produced mainly by the choroid plexus (CP; Johanson, 2003). During human development brain ventricles inflate several weeks prior to CP formation (Bayer and Altman, 2008). Similarly, in zebrafish, the ventricles begin inflating at 19 hours post fertilization (hpf) (Lowery and Sive, 2005), prior to formation of CP at 48 hpf (Bill et al., 2008; Garcia-Lecea et al., 2008). Therefore, an early evolutionarily conserved CP-independent mechanism that produces embryonic CSF (eCSF) must exist for ventricle inflation. Molecular analysis of the fundamental mechanisms of neurulation has shown their conservation across vertebrate species (Copp et al., 2003; Harris and Juriloff, 2010; Korzh, 2014). The formation of the ventricular system begins with the fourth ventricle (hindbrain), followed by the third ventricle (forebrain) and the central canal of the spinal cord (Lowery and Sive, 2005; Garcia-Lecea et al., 2008; Kondrychyn et al., 2013).

Inflation of the BVS requires increased intraluminal pressure. It is thought that this is achieved by a combination of increased production of eCSF by neuroepithelial (ependymal) cells, which also function as neural stem cells (Desmond and Levitan, 2002; Johansson et al., 1999; Lowery and Sive, 2005; Mirzadeh et al., 2008; Meletis et al., 2008; Hamilton et al., 2009; Guo et al., 2011) and/or restriction of neuroepithelial permeability (Chang et al., 2012; Fossan et al., 1985; Lowery and Sive, 2009; Whish et al., 2015). While the mechanisms regulating the ventricular lining remain unclear, the expression of voltage-gated potassium (K_v) channels by ependymal cells might be essential due to their role in cell proliferation (Smith et al., 2008).

K_v channels are K⁺-selective transmembrane proteins with the central pore surrounded by four individual α-subunits that contain six transmembrane segments (S1–S6) and cytoplasmic NH₂- and COOH-termini (Long et al., 2005). Several potassium currents, including both slow-inactivating delayed rectifier (I_K) and fast-inactivating A-type transient outward (I_A) currents, have been detected in neuronal stem cells (NSCs) and neuronal progenitor cells (NPCs) (Aprea and Calegari, 2012; Blackiston et al., 2009; Cai et al., 2004; Li et al., 2008; Liebau et al., 2006; Smith et al., 2008). In vitro-derived NSCs and mesenchymal stem cells (MSCs) express K_v1.3, K_v3.1 and K_v2.1 subunits, implicating these in I_K current (Deng et al., 2007; Smith et al., 2008). However, the composition of K_v channels underlying I_K and I_A currents varies depending on the differentiation and/or proliferation state of the NSCs and NPCs. The relevance of these observations to development of the brain ventricular system remains unclear.

Members of the K_v2 subfamily (i.e. K_v2.1 and K_v2.2 encoded by Kcnb1 and Kcnb2 in mouse, respectively) play an important role in neuronal excitability (Vacher et al., 2008). They assemble into functional heterotetrameric channels with modulatory ‘silent’ K_v (K_vS) subunits (Bocksteins and Snyders, 2012). K_v6.4 is a typical K_vS subunit, which, upon assembly with K_v2.1 at the endoplasmic reticulum, forms a heterotetramer that is transported to the plasma membrane. Similar to K_v2.1 homotetramers, K_v6.4 reduces the current density, shifts the voltage dependence of inactivation ∼40 mV into hyperpolarized potentials and slows the activation and deactivation kinetics (Ottschytsch et al., 2002; Bocksteins et al., 2009a,b).

Here, we investigated the role of K⁺ channels in the development of the BVS. Using a Tol2 transposon-mediated ‘gene-breaking’ insertional mutagenesis of zebrafish (Sivasubbu et al., 2006; Clark et al., 2011), we identified the kcng4b mutant, which is due an over-inflated IVth ventricle. We demonstrated that Kcng4b, a zebrafish homologue of the ‘silent’ K_v6.4 subunit, regulates proliferation of neuroepithelium lining the ventricular cavity. Kcng4b is also crucial for ventricle inflation as it maintains neuroepithelium integrity by modulating K_v2.1 activity, which opposes K_v6.4 during formation of the BVS. Taken together, these results establish the Kcng4-Kcnb1 axis as a developmental regulator of the BVS in vertebrates.

To determine novel molecular determinants of the BVS, we performed a forward genetic screen in zebrafish using Tol2 transposon-based insertional mutagenesis with a ‘gene-breaking transposon’ (GBT) cassette (Sivasubbu et al., 2006) (Fig. S1). Once inserted into the intronic sequence by Tol2 transposition, it disrupts splicing, with nearly all tested fish lines displaying a >95% reduction in native transcript abundance (Sivasubbu et al., 2007; Clark et al., 2011). Using this strategy, we generated 61 zebrafish lines bearing germline-inactivating integrations; one of these that mapped to the second intron of kcng4b (Chr. 7, Zebrafish Information Network, http://zfin.org/ZDB-GENE-130530-634; Fig. 1H) was characterized by the appearance of cells in the third ventricle (24 hpf, N=525/525; 100%), hydrocephalus (48 hpf, N=411/525; 78%), cardiac oedema (Fig. 1A-G) and a reduction of intersomitic vessels (not shown).

Apart from kcng4b, the zebrafish genome also harbours kcng4a (Chr.18, ENSDARG00000062967, Ensembl Zv9). Comparison of Kcng4a and Kcng4b with K_v6.4 and K_v2.1 demonstrated an evolutionary conservation of functional domains characteristic of mammalian Shaker-related K_v subunits: proteins from both species possess six transmembrane segments (S1-S6), a pore loop containing the K⁺ signature sequence GYG, several positively charged residues on the S4 segment that function as voltage sensor and a tetramerization domain, T1, on the cytoplasmic N-terminus (Long et al., 2005; Fig. S2A). However, compared with mammalian K_v6.4, the two putative zebrafish Kcng4 proteins have longer C-termini. The mutation causes the premature termination of kcng4b transcript at R249, resulting in a transcript similar to the short splice variant predicted for human KCNG4 and zebrafish kcng4a. Based on sequence homology, this transcript encodes a truncated polypeptide Kcng4b*, which contains a segment of Kcng4b with the N-terminus, S1 segment and partial S1-S2 linker (Fig. 1I and Fig. S2B). Kcng4b* is structurally similar to the truncated K_v1.1 and K_v1.5 that exert a relatively weak dominant-negative effect in vitro (Babila et al., 1994). Hence, the zebrafish mutant provides a unique opportunity to understand the developmental impact of such a truncation of a modulatory K⁺ channel subunit.

Reverse transcription-PCR (RT-PCR; Fig. 1J) on 30 hpf wild-type (kcng4b^+/+), heterozygous (kcng4b^+/−) and homozygous (kcng4b^−/−) embryos revealed a reduction of full-length kcng4b transcripts in kcng4b^+/− and their absence in kcng4b^−/− (Fig. 1J). Thus expression of native kcng4b was effectively ablated by GBT cassette insertion. To confirm that the mutant phenotype was due to kcng4b loss of function (LOF), we injected wild-type embryos with antisense morpholino oligonucleotides (MOs), inhibiting splicing of kcng4b at the intron 2-exon 3 junction (Fig. S3A,B) and observed the same phenotype (Fig. S3C). Hence, LOF of kcng4b caused by insertional mutagenesis or MO-mediated knockdown caused the same defects in the BVS.

To define the developmental expression profile of kcng4b, we used RT-PCR at time points preceding emergence of the mutant phenotype. kcng4b is expressed in a biphasic manner, with the first transient expression phase at the end of epiboly (8-9 hpf) and the second phase starting at 21 hpf (Fig. 2A,A′). Full-length transcripts were not detected in the adult tissues examined (Fig. 2B), indicating that kcng4b functions only in development. Whereas the first phase of kcng4b expression (8-9 hpf) might be linked to formation of the Kuppfer vesicle, the second phase of kcng4b expression correlates with formation of the brain ventricular system at 18-22 hpf (Lowery and Sive, 2005). Whole-mount in situ hybridization (WISH) using antisense RNA probes revealed expression of kcng4b in the ventricular zone and central canal (24-48 hpf, Fig. 2C,E-H), extended yolk and otic vesicle (30 hpf, Fig. 2G,G″; 48 hpf, Fig. 2H′), ciliary retina, lenses (48 hpf, Fig. 2H′, inset) and intersegmental vessels (48 hpf, Fig. 2H″). The sense probe showed no staining (Fig. 2D).

To distinguish developmental roles of the two kcng4 genes, we cloned kcng4a. At 24-48 hpf, kcng4a was expressed in the lens, similar to kcng4b, and in the cranial sensory ganglia and Rohon-Beard sensory cells of the spinal cord (Fig. 2F). Therefore, with the exception of the developing eye, kcng4a and kcng4b exhibit distinct expression patterns: importantly, only kcng4b is expressed in the brain ventricular system. The mutant phenotype of kcng4b^−/− correlates with expression of kcng4b (Fig. 2). The divergence of expression patterns of the two kcng4 genes supports a specific role of kcng4b in development of the BVS.

The primary kcng4b^−/− phenotype is the presence of cells in the forebrain ventricle. It prompted detailed evaluation of the ventricular zone of mutant embryos. ZO-1, αPKC and F-actin identify the apical surface of the neuroepithelium (Munson et al., 2008). The distribution of ZO-1 and αPKC at 24 hpf on confocal sections (Fig. 3A-H′) as well as αPKC at 24 hpf (Fig. S4) and F-actin at 48 hpf (Fig. S4) on cryosections was examined by fluorescence microscopy. In wild-type embryos, anti-ZO-1 and anti-αPKC immunolabelling revealed a smooth apical surface, while in kcng4b morphants it was distorted (Fig. 3E′,G′; Fig. S4C,D). In the forebrain ventricle cells were found in the lumen (Fig. 3A′-H′; Fig. S4A,B). At 48 hpf the dorsal wall of the neural tube is very thin in mutants and lacks organized F-actin staining (Fig. S5A-D). Hence loss of function of kcng4b affects cell polarity in the ventricular zone.

Neuroepithelial cells lining the brain ventricles retain eCSF within the lumen (Lowery and Sive, 2009) and the integrity of this barrier could be measured using a dye retention assay (Chang et al., 2012) (Fig. 3I-P′). Following injection of 70 kDa FITC-Dextran into brain ventricles at 24 hpf, the distance that dye front spread from the forebrain was analysed (Fig. 3K,L, blue bar). It was found that within 60 min the dye had spread in mutants much more than in controls (Fig. 3M,N). This indicated that the integrity of mutant neuroepithelium was compromised. We investigated this further by detailed confocal imaging of the neuroepithelium of kcng4b morphant embryos in Tg(h2b:EGFP) background (Fig. 3R-Z). Here, Dextran-Texas Red penetrated into the mutant forebrain neuroepithelium 30 min after its injection into the fourth ventricle (Fig. 3V,V′), but not in midbrain and hindbrain neuroepithelium (Fig. 3Z,Z′; n=6). Therefore, disruption of neuroepithelium integrity mainly happens at the forebrain region in kcng4b-deficient embryos.

The secondary effect of kcng4b LOF is hydrocephalus. We quantified this phenomenon by measuring the volume of the BVS in 48 hpf controls and mutants. Three-dimensional reconstruction of confocal Z-scans showed that the BVS volume was almost three times greater in the mutant compared with the volume in wild-type embryos (Fig. 4A-C). To show an effect of Kcng4b gain of function (GOF) on ventricular development, kcng4b mRNA was injected into 1- to 2-cell-stage wild-type embryos, and at 21-24 hpf, they were soaked in FITC-BODIPY (Lowery and Sive, 2005). The hindbrain ventricle forms at 21 hpf and expands at 24 hpf, when the midbrain lumen (optocoele) forms. Hence, initial brain inflation and subsequent lumen expansion were monitored at 21-24 hpf (Fig. 4F-J). According to analysis of the body axis, somite number and formation of the midbrain-hindbrain boundary (MHB), the Kcng4b GOF embryos developed relatively normally (Fig. 4D,E) except that at 21 hpf the hindbrain ventricle did not form (Fig. 4G) and at 24 hpf it did not form and/or expand and the optocoele failed to form (Fig. 4I,J). Thus, Kcng4b GOF embryos displayed a phenotype opposite to that of Kcng4b LOF (Fig. 4A-C).

Next, we asked whether the function of Kcng4b is conserved in evolution. We injected wild-type embryos with a construct encoding C-terminal EGFP-tagged human K_v6.4. This affected inflation of brain ventricles similar to that of Kcng4b GOF (Fig. S6) suggesting that human K_v6.4 may regulate formation of ventricles.

During neurogenesis in mammals and zebrafish, the nuclei of dividing neuroepithelial cells prior to mitosis undergo interkinetic nuclear migration (INM) towards the apical surface, where round cells appear (Merkle and Alvarez-Buylla, 2006; Gutzman and Sive, 2010; Leung et al., 2011). Such cells were often found close to the midbrain lumen in 24 hpf wild-type embryos (Fig. 4H″; arrowheads; n=22 per confocal section). In contrast, upon kcng4b GOF, proliferating cells were less numerous (n=6 per confocal section; Fig. 4I″). These results suggest that Kcng4b modulates proliferation of neuroepithelial cells of the BVS.

To extend this analysis, the impact of Kcng4b on cell proliferation was analysed by quantifying phospho-histone 3 (PH3)-positive (M-phase) cells. The majority of the PH3-positive cells were near the lumen (Fig. 5A,D,G). The number of the PH3-positive cells and the proliferative index (the number of PH3-positive cells per thousand neuroepithelial cells) in 24 hpf kcng4b GOF embryos was significantly lower than that in controls (Fig. 5B,C). In contrast, analysis of kcng4b^−/− mutants demonstrated a significant increase in PH3-positive cells (Fig. 5D,I). Whereas during development cell proliferation decreases, in kcng4b^−/− mutants this trend is less obvious as reflected by the difference in proliferative index between kcng4b^−/− mutants and wild-type controls, which increased from 2-fold (24 hpf) to almost 3-fold (48 hpf). Taken together, these results demonstrate that Kcng4b modulates cell proliferation.

To analyse Kcng4b function further, we expressed full-length Kcng4b in human cell lines for electrophysiological characterization. Similar to human K_v6.4 (Ottschytsch et al., 2002), zebrafish Kcng4b alone did not form electrically active channels at the plasma membrane (data not shown). However, upon co-expression with K_v2.1, the zebrafish Kcng4b probably assembled into functional K_v2.1/Kcng4b heterotetramers since this caused modulation of biophysical properties compared with K_v2.1 homotetramers (Fig. 6). Typical current recordings of K_v2.1/Kcng4b heterotetramers are shown in Fig. 6A. K_v2.1/Kcng4b channels display a voltage-dependence of activation and of inactivation that is characterized by V_1/2=−22.1±1.5 mV (n=9) and V_1/2=−52.8±2.5 mV (n=5), respectively (Fig. 6B). Furthermore, K_v2.1/Kcng4b heterotetramers displayed two components in their time course of activation (Fig. 6C). In addition, Kcng4b also reduced the K_v2.1 current density; K_v2.1 and Kcng4b co-expression displayed a current density of 1115±214 pA/pF (n=9, Fig. 6D) at 40 mV, whereas the same amount of transfected K_v2.1 cDNA (co-expressed with CFP to exclude possible effects of DNA dilution and/or an overload of the translational machinery) resulted in current too large to clamp at 40 mV without any voltage errors (not shown). The biophysical properties of these K_v2.1/Kcng4b heterotetramers are very similar to those of K_v2.1/K_v6.4 channels, indicating that zebrafish Kcng4b modifies the biophysical properties of K_v2.1 in a similar way as human K_v6.4; human K_v6.4 shifts the K_v2.1 voltage dependence of inactivation almost 40 mV to hyperpolarized potentials, introduces a second time constant in the K_v2.1 activation time course and reduces the K_v2.1 current density (David et al., 2015). In contrast to full-length Kcng4b, co-expression of K_v2.1 and Kcng4b* did not result in the typical K_v6.4-induced modulations of K_v2.1 activation, inactivation and deactivation properties (Fig. 6). Typical current recordings obtained after co-expression of K_v2.1 with Kcng4b*, are shown in Fig. 6A. Co-expression of K_v2.1 with Kcng4b* resulted in a voltage dependence of activation and of inactivation that is characterized by V_1/2=6.7±1.6 mV (n=5) and V_1/2=−15.0±3.1 mV (n=5), respectively, which are both significantly different (P<0.05) compared with that of K_v2.1/Kcng4b heterotetramers (Fig. 6B). Furthermore, co-expression of K_v2.1 with Kcng4b* yielded an activation time course characterized by only one component and the time course of deactivation was accelerated compared with that of K_v2.1/Kcng4b heterotetramers (Fig. 6C). The biophysical properties obtained upon co-expression of K_v2.1 with Kcng4b* are very similar to those of K_v2.1 homotetramers (David et al., 2015), suggesting that Kcng4b* did not tetramerize with K_v2.1. However, co-expression of K_v2.1 and Kcng4b* reduced the current density to an even greater extent than co-expression of native Kcng4b; Kcng4b* reduced the current density to 606±184 pA/pF (n=9, Fig. 6D) at 40 mV, where the open probability of both K_v2.1/Kcng4b and K_v2.1/Kcng4b* channels is 1, excluding the possibility that the differences in current density reflect the difference in open probability. This suggests that Kcng4b* does interact with K_v2.1 and exerts a dominant-negative effect, as has been demonstrated for the dominant-negative mutants of K_v1.5 and K_v2.1 (Babila et al., 1994; Xu et al., 1999). Taken together, the developmental and electrophysiological analysis of the kcng4b mutant demonstrated that the mutant phenotype resulting from the LOF of Kcng4b has been further compounded by the dominant-negative effect of Kcng4b*.

These results suggest that Kcng4b modulates activity of K_v2.1, which in zebrafish is encoded by a single kcnb1 gene (Chr. 6, ENSDARG00000060095, Ensembl Zv9). WISH demonstrated that kcnb1 is expressed in the BVS, central canal and otic vesicles at 24-30 hpf (Fig. 7A,B), similar to kcng4b (Fig. 2C-E). At 48 hpf, the expression domains of these genes include the BVS, eyes and ears (Fig. S7). To explore a developmental role of kcnb1, we performed GOF analysis by injecting kcnb1 mRNA into zebrafish embryos, which in a fraction of 24 hpf embryos, resulted in a clump of cells in the forebrain ventricle (Fig. 7C,D; N=218/325; 67%) and caused hydrocephalus at 48 hpf (Fig. 7E,F; N=42/218; 19%). Thus, the phenotype of Kcnb1 GOF phenotypically mimics that of the kcng4b mutant (Fig. 1A-D). This again pointed to the functional antagonism of members of the Kcng4b-Kcnb1 axis during formation of the BVS.

To further validate the functional role of Kcnb1 during ventricle formation, we generated a kcnb1 mutant by CRISPR-Cas9 mutagenesis (Hwang et al., 2013). The single-guide RNA (sgRNA) targeted zebrafish kcnb1 downstream of the sequence encoding the NAB domain (Fig. 7G). Following Cas mRNA injection, T7E1 assay documented an induction of molecular lesions at the targeted site. Adult F₀ animals were screened for stable mutations inherited by F₁ progeny. We isolated three different kcnb1 indel (insertion/deletion) mutants, including two with premature stop codons (Fig. 7G, #1, #2). kcnb1^−/− displayed two distinct phenotypes: the early one of delayed epiboly and gastrulation failure (Fig. S8B) and some kcnb1 mutants completed gastrulation and displayed the late phenotype of reduced brain ventricles (Fig. 7H,I). As shown by FITC-Dextran labelling, 24 hpf kcnb1^−/− failed to inflate the midbrain and partially inflated the hindbrain (Fig. 7H′,I′).

These results further support the functional antagonism of members of the Kcng4b-Kcnb1 axis during formation of the BVS. Embryos with epiboly and brain ventricle phenotype constitute about 9-14% of the progeny of kcnb1^−/+ parents of both alleles (#1, #2) (Fig. S8B), illustrating incomplete mutation penetrance. Taken together, these results demonstrated a novel function of the voltage-gated K⁺ channels in regulating development of the BVS, which involves modulation of Kcnb1 (K_v2.1) activity by Kcng4b (K_v6.4).

Several electrically active subunits of voltage-gated potassium (K_v) channels contribute to the neuronal delayed rectifier K⁺ current (I_DR) regulating cell proliferation and excitability in the CNS (Wonderlin and Strobl, 1996; Pardo, 2004; Vacher et al., 2008). Kcnb1-encoded K_v2.1 contributes into K⁺ channels that form a major I_DR component in hippocampal and DRG neurons (Murakoshi and Trimmer, 1999; Du et al., 2000; Bocksteins et al., 2009a,b). Moreover, it has been suggested that the electrically silent K_v (K_vS) subunits modulate the K_v2-mediated current; K_v2/K_vS heterotetrameric channels carry approximately 30% of the K_v2-containing I_DR current in DRG neurons (Bocksteins et al., 2009a,b).

Our study demonstrated that kcng4b and kcnb1, which encode homologues of mammalian K_v6.4 and K_v2.1, respectively, are expressed in the ventricular neuroepithelium (Figs 2, 7). Kcng4 expression was also detected in the ventricular neuroepithelium of developing mice (http://developingmouse.brain-map.org/gene/show/42576) (Lein et al., 2007). Coincidentally, the kcng4b mutation causes formation of Kcng4b*, similar to polypeptides with a dominant-negative effect on K_v activity in vitro (Babila et al., 1994). The native short transcripts of the human KCNG4 and zebrafish kcng4a might be involved in development of sensory neurons, etc. Hence several elements of expression of Kcng4 seem to have been conserved in evolution, suggesting a conservation of its developmental roles in vertebrates.

The interplay between activities of different K_v channels determines the proliferative state of various cells/tissues (Wonderlin and Strobl, 1996; Pardo, 2004; Deng et al., 2007). While selective blockage of K_v1.3 and K_v3.1 channels increases the proliferation rate of mesencephalic NPCs, pharmacological blockage of K_v2.1 inhibits cell proliferation in rat MSCs (Pardo, 2004). The combined electrophysiology and genetic analysis revealed the likely interaction between Kcng4b and Kcnb1; however, Kcng4b may interact with other electrically active subunits of K_v channels. Recently, we demonstrated that the N-terminus of K_v6.4 interacts with that of K_v3.1 without the formation of functional heterotetrameric channels (Bocksteins et al., 2014). This suggests that Kcng4b* may interact with K_v3.1 and that Kcng4 LOF could increase cell proliferation by reducing the K_v3.1 current via non-functional K_v3/Kcng4* ‘channels’. Further developmental analysis of the relevant K_v channels will be required in future.

An increase in the surface area of the neuroepithelial layer further increases the proliferation of neuroepithelial cells and where the ventricular lining becomes overcrowded, cells extrude to preserve barrier function (Rosenblatt et al., 2001; Gu and Rosenblatt, 2012). This may explain the appearance of cells in the third ventricle (Fig. 1C′,D′; Fig. 3A-D; Fig. 7D). Conversely, overexpression of Kcng4b would reduce K_v channel activity at the plasma membrane, which could alter the membrane potential of neuroepithelial cells towards depolarization and decrease cell proliferation (Fig. 5A-C) in parallel with deficient inflation of brain ventricles (Fig. 4G,I). Therefore, our data support a role for voltage-gated K channels in the modulation of neuroepithelial cell proliferation coupled to ventricle inflation.

Formation of brain ventricles requires both production of eCSF and its retention. The latter function may be accomplished by the neuroepithelial barrier lining the ventricles (Chang et al., 2012; Fossan et al., 1985; Lowery and Sive, 2009; Whish et al., 2015). In the kcng4b morphant, this layer is compromised (Fig. 3A′-D′; Figs S4,S5) and fails to maintain an expansion of ventricular system within physiological limits and retain FITC-Dextran (Fig. 3E′-N′) coincidental with the abnormal ventricular system induced by changes in Kcnb1 activity (Fig. 7). This further supports the idea that the voltage-gated K⁺ channels regulate formation of the polarized and continuous neuroepithelial layer, which acts as a barrier to restrict a flow of eCSF. This hypothesis has been presented in more detail in Fig. 8, although the exact molecular mechanisms will probably remain elusive until transcriptome analyses of the kcng4b mutant are performed.

Judging by an increased cell proliferation and extrusion of cells, the integrity of the neuroepithelial layer is compromised very early, at least a day before hydrocephalus. Why does cell extrusion happens in the third ventricle and not in the fourth one? It could be that the thin roof of the hindbrain ventricle better accommodates superfluous cells compared with the forebrain ventricle, which is surrounded by the thick walls of the neural tube. These inherent differences in ventricle mechanics manifest in a much greater degree of hydrocephalic expansion in the hindbrain (Fig. 1F,G). The cell extrusion is a much earlier and more common phenomenon compared with hydrocephalus. This suggests that a defect of integrity of neuroepithelium is probably a primary defect and is reminiscent of other mutants. Chang et al. (2012) described a discontinuous and non-polarized cell layer lining the ventricles in Na⁺/K⁺-ATPase (atp1a1)-deficient zebrafish embryos. The delamination (extrusion) of ventricular cells (as a primary defect) and hydrocephalus (as a secondary one) has been also described in Napa mutant (hyh) mice. The gene Napa encodes soluble N-ethylmaleimide sensitive factor (NSF) attachment protein α (αSnap), which is involved in SNAP receptor (SNARE)-mediated vesicle fusion in many cellular contexts (Chae et al., 2004). Similar to α-Snap K_v2.1 also plays a role in membrane vesicle trafficking and SNARE-mediated membrane fusion in both neurosecretory cells and sensory neurons (Feinshreiber et al., 2009, 2010) and is required for transport of K_v6.4 to the plasma membrane (Ottschytsch et al., 2002). This functional connection might underlie the phenotypic similarity of the hyh and kcng4b mutants.

In summary, the insertional mutagenesis screen in zebrafish led to identification of a kcng4b mutant that develops cell extrusion in the BVS followed by hydrocephalus, indicating a novel function of K⁺ I_DR channels in regulating brain ventricle development. This further led to identification of the functional interplay between K_v2.1 and K_v6.4 subunits during brain morphogenesis and suggests that Kcng4b could be a novel risk factor of integrity of ependyma and hydrocephalus. By extension, other negative regulators of K_v2.1 channel activity represent a reserve pool of factors that could similarly contribute to hydrocephalus. In contrast, deficiency of K_v2.1 has been linked with epileptic encephalopathy (Torkamani et al., 2014). It is rather remarkable that the two proteins involved in the same developmental mechanism underlying formation of the BVS, could be linked to two different hereditary diseases. The intrinsic functional interplay of the ‘silent’ K⁺ channel subunit Kcng4b and the electrically active K⁺ channel subunit Kcnb1 during formation of the BVS highlights the dysregulation of K_v channels as the shared origin of these diseases and warrants further investigation of this complex problem.

Wild-type zebrafish (AB, ZIRC) were maintained as described (Westerfield, 1995). All animal experiments were carried according to the regulations of Institutional Animal Care and Use Committee (Biological Resource Center of Biopolis, license no. 120787), which approved this study. Developmental stages are in hours post fertilization (hpf) at 28.5°C (Kimmel et al., 1995).

The GBT cassette and Tol2 transposase mRNA were injected at one-cell stage (Sivasubbu et al., 2006). By 24 hpf, 697/4347 embryos showed mosaic GFP expression. A total of 350 adults were mated to wild-type fish, screened for stable F1 transgenics and 61 founders with germline integrations identified. In founders of F1 subfamilies, the germline transmission was 5-62%. Two fish from this subfamily were randomly chosen for mapping GBT integration sites using the TAIL-PCR method (Liu and Whittier, 1995; Parinov et al., 2004). Fish with multiple insertions were out-crossed to a single insertion state. pDR274 and pMLM3613 (Addgene, USA) were used to generate the gRNA that targets the GGAGCTGGACTACTGGGGAG sequence in kcnb1 exon1 and Cas9 mRNA, respectively, and to generate the kcnb1 mutant (Hwang et al., 2013).

cDNA sequences of kcng4a and kcng4b were obtained using Expand Long reverse transcriptase kit (Roche Diagnostics) and RNA isolated as a template from pooled 24 hpf embryos using the RNeasy kit (Qiagen). Full-length kcng4b was cloned into the pTNT vector (Promega) using the primers: A, 5′-CCG CTCGAG GCCACC ATGCCCATCATCAGCAATG-3′ and B, 5′-TGC TCTAGA TCAGATATCTTTGCAACATGC-3′ with XhoI/Xba restriction sites (underlined). The same primer pair (A and B) was used for RT-PCR detection of kcng4b transcripts. Plasmid pSP64-K_v6.4-EGFP contains the full-length C-terminal EGFP tagged human KCNG4 (Ottschytsch et al., 2005). Full-length kcnb1 was cloned into pCS2+ vector using primers: 5′-CGC GGATCC GCCACC ATGGAGAAACCCTCGGCA-3′ and 5′-CCG CTCGAG TCA AAG GCC CTT ATC AAA AG-3′ with BamHI/XhoI restriction sites (underlined). mRNA for microinjection was in vitro-transcribed using the plasmids linearized at the 3′-end as templates. 5′-capped RNA was synthesized using the mMessage T7/SP6 kit (Ambion). Transient knockdown of kcng4b was performed using a splice site-blocking MO (5′-TGCATTCGCCCTGTAAAAGAACAAA-3′) targeting kcng4b intron 2-exon 3 (I2E3) junctions. Standard MO with the antisense sequence 5′-CCTCTTACCTCAGTTACAATTTATA-3′ was used as a negative control. Morpholinos were designed by and purchased from GeneTools.

WISH, immunohistochemistry and cryosections were performed following established protocols (Korzh et al., 1998; Kondrychyn et al., 2013). Anti-phospho-histone H3 (mouse, #06-570, Millipore), anti-αPKC (rabbit C20, sc-216, Santa Cruz), anti-ZO-1 (mouse, #33-9100, Invitrogen) primary antibodies were used for immunohistochemistry (all at 1:200 in PBS). Secondary antibodies were Alexa Fluor 488 or Alexa Fluor 555 (Invitrogen) (1:500 in PBS) conjugated or Alexa Fluor 555-conjugated Phalloidin and DAPI (1:500 in PBS). To quantify cell proliferation, the PH3-labelled cells in each z-series were counted, averaged and measured using ImageJ software. The approximate area occupied by each cell, total cell number and the percentage of labeled cells in each region were calculated.

FITC-conjugated BODIPY ceramide (Invitrogen) was dissolved in DMSO (5 mmol/l). Embryos were soaked in 50 nmol/l BODIPY ceramide in egg water overnight in the dark at 28.5°C, washed, dechorionated and placed in 1% agarose wells (Lowery and Sive, 2005). Imaging was performed using a Zeiss LSM700 laser-scanning microscope. Images were processed by ImageJ software (NIH).

For dye retention assays (Chang et al. 2012), 70 kDa FITC-Dextran (2.5 ng/ml in water) was injected into the brain ventricles at 22 hpf and imaged at 15 min intervals for 60 min. Neuroepithelial permeability was quantified using ImageJ to measure the distance of the dye front from respective forebrain ventricle hinge-points at a 20° angle relative to the embryonic midline. The net distance of the dye front diffusion was calculated by subtracting the distance at t=0 from other time points. Statistical analysis was performed and data plotted with Prism4 software (GraphPad).

Electrophysiological characterization of Kcng4b and Kcng4b* was performed using the whole-cell patch clamp. Human embryonic kidney (HEK293) cells obtained from the ATCC are yearly tested for contamination. The new vial (originating from the original culture) was thawed and cultured in enriched modified Eagle's medium supplemented with 10% fetal bovine serum, 1% non-essential amino acids and 1% penicillin-streptomycin (Invitrogen) at 37°C and 5% CO₂. HEK293 cells were transiently co-transfected with K_v2.1 and Kcng4b or Kcng4b* cDNA using Lipofectamine2000 according to the manufacturer's instructions. GFP was co-transfected as a transfection marker. Current recordings were obtained 24 h post transfection using an Axopatch-200B amplifier (Axon Instruments) and were low-pass filtered and sampled at 1-10 kHz with a Digidata 1440 data acquisition system (Axon Instruments). The pClamp10 software (Axon Instruments) was used to command voltages and store data. Patch pipettes (with 1.5-2.5 MΩ resistance) were pulled with a P2000 laser puller (Sutter Instruments) from borosilicate glass (World Precision Instruments). After heat polishing, patch pipettes were filled with an intracellular solution (ICS) containing 110 mM KCl, 5 mM K₄BAPTA, 5 mM K₂ATP, 1 mM MgCl₂ and 10 mM HEPES. Cells were continuously perfused with an extracellular solution (ECS) containing 145 mM NaCl, 4 mM KCl, 1.8 mM CaCl₂, 1 mM MgCl₂, 10 mM HEPES and 10 mM glucose. The pH of the ICS and ECS were adjusted to 7.2 and 7.35 with KOH and NaOH, respectively. Current recordings with voltage errors exceeding 5 mV after series resistance compensation were excluded from analysis. The voltage-dependence of activation and of inactivation was fitted with the Boltzmann equation y=1/{1+exp[−(V−V_1/2)/k]} with V the voltage applied, V_1/2 the voltage at which 50% of the channels are activated or inactivated and k the slope factor. Activation and deactivation kinetics were fitted with a single or double exponential function.

We thank Sridhar Sivasubbu and Stephen Ekker for the pX vector, Mr. Melvin Sin, Ms. Yi-chuan Chen and Ms Shelvi Ganda for collecting data, the personnel of the IMCB Fish Facility for maintenance of zebrafish, Dr Lucy Robinson of Insight Editing London for editing the manuscript and three anonymous reviewers for constructive comments.