Mutant analysis of Kcng4b reveals how the different functional states of the voltage-gated potassium channel regulate ear development

The voltage gated (Kv) slow-inactivating delayed rectifier channel regulates the development of hollow organs of the zebrafish. The functional tetramer consists of an electrically active subunit (Kcnb1, Kv2.1) and a modulatory silent subunit (Kcng4b, Kv6.4). The two mutations in zebrafish kcng4b - kcng4b-C1 and kcng4b-C2 (Gasanov et al., 2021) - have been studied during ear development using electrophysiology, developmental biology and in silico structural modelling. kcng4b-C1 mutation causes a C-terminal truncation characterized by mild Kcng4b loss-of-function (LOF) manifested by failure of kinocilia to extend and formation of ectopic otoliths. In contrast, the kcng4b-C2-/- mutation causes the C-terminal domain to elongate and the ectopic seventh transmembrane (TM) domain to form, converting the intracellular C-terminus to an extracellular one. Kcng4b-C2 acts as a Kcng4b gain-of-function (GOF) allele. Otoliths fail to develop and kinocilia are reduced in kcng4b-C2-/-. These results show that different mutations of the silent subunit Kcng4 can affect the activity of the Kv channel and cause a wide range of developmental defects.


Introduction
Development is regulated by biochemical gradients, transcriptional networks, and bioelectrical signals generated by ion channels and pumps (Levin, 2014). Voltage-gated potassium (K v ) channels play an important role in the maintenance of the membrane potential.
Kv2 channels consists of the electrically active (KCNB1) and modulatory (or silent, KCNG4) α -subunits. Subunit monomers assemble functional channel heterotetramers. The stoichiometry of the subunits in any given tetramer is variable, which adds to the functional complexity of K v channel (Bocksteins, 2016;Möller et al., 2020). The genes encoding the subunits are expressed in the cell lineage and tissue specific pattern. Antagonizing activities of different subunits of K v channels establish a regulated balance in vitro (Bocksteins and Snyders, 2012;Ottschytsch et al., 2005) and in vivo (Jedrychowska et al., 2021;Jędrychowska and Korzh, 2019;Shen et al., 2016).
Most human KCNB1 mutants are heterozygotes carrying a single amino acid substitution.
kcng4b is expressed only during development in the brain ventricular system (BVS), ear, etc. (Shen et al., 2016). The developmental role of Kcng4 in the ear is still not fully understood.
To address this issue, we generated two C-terminal kcng4b mutants (Gasanov et al., 2021).
Unexpectedly, these cause opposite developmental defects ear development, likely because they represent the gain-of-function and loss-of -function mutant versions of Kcng4.

Animals
The wild-type (AB) and mutant zebrafish (Danio rerio) lines were maintained in the Zebrafish Facility of the International Institute of Molecular and Cell Biology in Warsaw (license No. PL14656251) according to the standard procedures and ethical practices (Westerfield, 2007). All experiments with zebrafish embryos/larvae were performed in accordance with the rules of the Polish Laboratory Animal Science Association and European Communities Council Directive (63/2010/EEC). Developmental stages (in hours, hpf, and days post fertilization, dpf) were described according to (Kimmel et al., 1995). The transgenic line, (Xiao et al., 2005), which expresses cytosolic EGFP in inner ear and lateral line mechanosensory hair cells, was used in this study. The generation of two kcng4b mutations, C1 and C2, by CRISPR-Cas9 has been described previously (Gasanov et al., 2021).
Xenopus laevis frogs (Nasco, Chicago, IL) were maintained in the animal facility of the Conductance-voltage relations for wild-type and mutant K v channels were determined from tail currents. The Pulsfit (HEKA), Igor Pro 6.0 (WaveMetrics, Lake Oswego, OR, USA) and Kaleidagraph 4.0 (Synergy Software, Reading, PA, USA) software for data analysis were used.
Non-specific leak current was estimated by averaging the currents of 4-7 water-injected oocytes in parallel to each experiment. The average current was subtracted for GV and steady-state inactivation analysis. Normalized data points were fitted with Boltzmann function of the form G/G max =1/(1+e( (V1/2-V)/K)*F/RT ) for GVs and I test /I cont =((1-offset)/(1+e( (V1/2-V)/K)*F/RT )) +offset for steady-state inactivation, V 1/2 is the midpoint voltage of activation and inactivation, respectively, K is the slope constant, F, R and T have their usual meanings.
Computer modelling of Kcng4b/K v 6.4b mutant proteins The isolated S7 helix model was generated by Tasser web server (Yang et al., 2014) using a sequence of 34 amino acids predicted to form a TM helix. The structure of the K v 1.2-2.1 paddle chimera channel (PDB id: 2R9R) was used as a template for the tetramer model after discarding beta-2 chains. Two opposing monomers served as the basis for modelling Kcng4b, whereas the other two were left unchanged. When constructing the Kcng4b monomer structures, the S7 helix was added based on the isolated helix model with constraints on the distances between the ends of S7 and S1 helices. The Modeller software (Webb and Sali, 2016)  Finally, the model was simulated without restraints for 1.2 μ s. We used Gromacs for position restraints and plumed for RMSD restraints (Tribello et al 2014).
Mean square displacement (MSD) and angle calculations were performed using the MDTraj library for Python (McGibbon et al., 2015). Only backbone atoms were used for both analyses.

Light-sheet and confocal fluorescence microscopy imaging
In vivo imaging and imaging of fixed specimens were performed as previously described (García-Lecea et al., 2017;Jedrychowska et al., 2021). For imaging of fixed material, 0.8 % lowmelting agarose in phosphate buffer saline (PBS) was used instead of E3 0.02 % tricaine medium. Zeiss Lightsheet Z. 1 with W Plan-Apochromat 20x/1.0 UV-VIS (for in vivo imaging), 40x/1.0 UV-VIS or 63x/1.0 UV-VIS (for fixed embryos) objectives were used, transmitted LED light was also used to obtain the high-resolution bright-field images of zebrafish ear. Confocal imaging was performed using the Zeiss LSM 800 inverted microscope with Airyscan (Carl 1 0 detected using emission filters (425-430, 514-530 and 592-625 nm BP), respectively. Data were saved in the LSM or CZI format and processed using ZEN (Zeiss) or ImageJ 1.51n (Fiji) software. Maximum intensity or sum slices projections were generated for each z-stack.

Immunohistochemistry and in situ hybridization
Whole-mount immunohistochemistry was performed according to a previously with sequences of kcnb1, kcng4b, otop1, slc12a2 and atp2b1a cloned in 3'-5'direction cDNA using Digoxigenin-labelled uridine triphosphate (DIG-UTP). The probe against kcnb1 was synthesized from a previously described plasmid (Shen et al., 2016). The synthesis of the otop1 and atp2b1a probes is described previously (Jedrychowska et al., 2021), while the kcng4b and slc12a2 probes were generated from pKcng4b-probe/pTnT and slc12a2-probe/pTnT vectors.
1 1 template. The PCR product was cloned into a pTnT vector using the EcoRI restriction site (underlined in the primer sequences) in both orientations to get the sense and antisense probes.

Transcriptome analysis by RNAseq and bioinformatics.
150 embryos for each sample were collected at 24 hpf [wild type control and Kcng4b-deficient crispants and morphants (Shen et al., 2016)] and RNA preparation was performed as previously published (Shen et al., 2017;Winata et al., 2013). mRNA levels were quantified using a standard pipeline based on TopHat-cufflinks59 with the Danio rerio gene annotation file (assembly GRCz10) from the Ensembl database. For differential expression analysis, gene-level FPKM values were converted into log ratios (KO/WT, base 2) and the threshold for differential expression to meet 1% FDR was determined using a mixture model-based approach called EBprot60. The transcripts whose largest FPKM values were below 25 percentiles of all FPKM values across the sample were removed from further analysis as unreliable.
For genes whose expression changed in crispants and morphants compared to WT, a gene was considered differentially expressed if the quantified FPKM value (Trapnell et al., 2010) was above the 25 percentiles of the whole transcriptome in the given sample. Functional enrichment test (Gene Ontology) was performed using an in-house program that calculates the significance of enrichment by hypergeometric test, using the GO annotation of genes from ZFIN database (http://zfin.org). The RNAseq data have been deposited online (GEO submission no. GSE194272).

qRT-PCR
The SsoAdvanced Universal SYBR Green Supermix and CFX Connect real-time PCR system (Bio-Rad, USA) was used for the Quantitative Real-Time Polymerase Chain Reaction (qRT-PCR) according to the manufacturer's instructions. The TRIzol-chloroform RNA extraction protocol (Sigma-Aldrich, USA) was used to extract total RNA from 20-50 zebrafish embryos at 28, 48, and 72 hpf. RNA concentration was assessed using a NanoDrop™ 2000 spectrophotometer (Thermo Scientific, USA), and cDNA was synthesized from 1 µg of RNA using the iScript™ Reverse Transcription kit (Bio-Rad, USA). qRT-PCR was performed using gene-specific primers for eef1a1l1, the reference reaction for the housekeeping gene, and other 1 2 primers (Suppl. Table 1) were used to amplify mRNAs of interest selected based on results of the RNAseq analysis described above.
The threshold cycle of each target and reference gene amplification in control and mutant embryos was determined automatically by LightCycler® 96 Software (Roche Diagnostics, USA). Fold change in mutants versus controls was calculated with delta-delta-C(t) method and Student's two tailed t-test with respect to the mismatch control was used to determine statistical significance. Statistical analysis including standard deviation calculation, was performed using Microsoft Excel (Microsoft, USA) and GraphPad Prism 5 (GraphPad, USA) software. Melting curve analysis and agarose gel electrophoresis were performed as product specificity controls, while the reaction efficiency (E) was calculated separately for each gene analyzed. All RNA samples used as the templates were extracted independently from three or four different groups of embryos, and each was analyzed by qRT-PCR at least twice after cDNA synthesis.
The value above 1.0 corresponds to an increase of mRNA level, the value below 1.0 corresponds to a decrease of mRNA level and "n.c." -"not changed" is the same value as in the controls.

Results
kcng4b plays the developmental role in the brain ventricular system (BVS) (Shen et al., 2016). Here we focused on the analysis of Kcng4b function during ear development, where the third (ectopic) otolith was previously detected in the insertional mutant kcng4b-trunc (kcng4b sq301 ) (Suppl. Fig. 1). In absence of this mutant line, which has been lost, to study the developmental role of Kcng4b, we generated two novel Kcng4b mutant alleles (kcng4b-C1, kcng4b-C2) (Fig. 1) both of which were lethal as homozygotes.

Computational analysis of Kcng4b mutants
The To explore this possibility molecular dynamics simulations of the S7 helix in model lipid membranes were performed (see M&M for details). Since the behavior of TM helices is phasedependent, we used POPC lipids as a model for liquid-disordered (Ld) domains, and DPPC lipids with 40% cholesterol as a model of liquid-ordered (Lo) domains (Fig. 3C). The S7 behaved differently in these cases: in the thinner Ld phase, the helix tilted by 44 degrees to better accommodate its hydrophobic thickness in the membrane hydrocarbon core, while in the much thicker Lo phase, it remained stably upright with average tilt of 11 degrees (Fig. 3C).
To explore whether the S7 helix forms a stable complex within the K v 2.1 (Kcnb1)-K v 6.4 (Kcng4) heterotetramer, we set out to simulate the entire TM portion of the channel. Since a structure of an entire K v 6.4 has not been solved, we used the K v 1.2-K v 2.1 paddle chimera channel (PDB ID: 2R9R) as a basis for the homology model in approach used previously (Matthies et al., 2018).
To show that the S7 helix remains stably bound to the channel, we compared the diffusion coefficients of the S7 helix either in pure lipid bilayers or in the context of the heterotetramer model by plotting mean squared displacement (MSD) against the time interval τ (Fig. 3D). The helices in the heterotetramer model are much less mobile than in the pure bilayer models, indicating that the S7 helix remains bound to the rest of the TM domains. Although we were unable to predict its exact rotational orientation with respect to the tetramer, the fact that its protein-protein interface has not been evolutionarily optimized strongly suggests that there is no single high-affinity binding site, and that the location of the additional helix is largely determined by an entropic tendency to displace lipids from the lateral surface of the protein where they undergo conformational ordering (Fig. 3B).
While the 35 amino acids of Kcng4b-C2 mutant discussed above clearly form a helix, this region is preceded by the 14 amino acids sequence whose secondary structure remains unknown. However, since this sequence results from the same ORF shift, in the absence of strong predictions, we found it reasonable to assume that this region is unstructured. Indeed, in our simulations, this region remained unstructured for a total of 2.25 µs, forming tangled loops with no clearly defined secondary structure (Fig. 3B). shifted the steady-state inactivation curve toward the hyperpolarized potentials (Fig. 4, Table 1).
These findings were consistent with the previous results obtained by co-expression of KCNG4 with KCNB1 in Xenopus oocytes (Möller et al., 2020) as well as in mammalian cells (Shen et al., 2016). Co-injection of KCNB1 and kcng4b-C1 mRNA had a very similar effect on the biophysical properties of Kv2.1 as co-expression with kcng4b-wt (Fig. 4 In contrast, a dramatic reduction of current density was observed upon co-expression of KCNB1 with both Kcng4b-C2 and Kcng4b-trunc (Fig. 4D). Furthermore, the biophysical properties of channels formed by co-injection of KCNB1 with these Kcng4b variants were almost identical to those of KCNB1 homomers ( Fig. 4; Table 1), indicating that the measured currents were generated by KCNB1 homomers. Since the current amplitudes of oocytes coinjected with KCNB1/Kcng4b-C2 and KCNB1/Kcng4b-trunc combinations are approximately 1/16 of the current mediated by homomeric channel (Fig. 4D), we consider it likely that these Kcng4b mutant forms exert a dominant-negative effect on Kv2.1 activity in the Xenopus oocyte system. A similar dominant-negative effect, but with much lesser extent, was previously shown by co-expressing Kcng4b-trunc with KCNB1 in HEK 293 cells (Shen et al., 2016). This inconsistency likely reflects the difference between the Xenopus oocyte and HEK cells expression systems. The former allows better control over the ratio of expressed proteins. Thus, the simplest explanation for the substantial current reduction in oocytes co-injected with the stabilized mRNA of KCNB1/Kcng4b-C2 and KCNB1/Kcng4b-trunc is that heterotetramers containing Kcng4b mutant variants become much less functional or completely non-functional.
Since the Kcng4b-trunc polypeptide lacked the protein domains forming ion permeation pore, but retained the tetramerization domain (Fig. 3), the observed dominant-negative effect can be explained by the formation of ion-conducting defective heterotetramers.

kcng4b mutants develop abnormal ears
The kcng4b-C1 -/embryos exhibited cell delamination in the third ventricle and mild hydrocephaly at 28 hpf (Suppl. Fig. 3A-D) along with abnormality of the heart, peripheral 1 8 We have previously shown that the Kcnb1 LOF causes changes in gene expression (Jedrychowska et al., 2021). Here, we asked whether it is also true for mutations affecting the modulatory "silent" subunit of Kv2.1, Kcng4b. The two LOF mutants kcng4b-trunc and kcng4b-C1 show rather similar phenotypes: enlarged ear and increased number of otoliths. In contrast, the kcng4b-C2 mutants show the small ears and "no otoliths" phenotype. This phenomenon could be faithfully reproduced using the morpholino-or CRISPR-mediated knockdown (Shen et al., 2016). Aiming to address a cause for developmental changes in these mutants, we started with the comparative RNAseq analysis of transcriptomes of the Kcng4b-deficient embryos (crispants and morphants) by RNAseq (Suppl. Fig. 6). This analysis revealed changes in transcription of several genes known to play a role during development of the ear and otoliths.
These results were used as a guide to select the candidate genes for qRT-PCR analysis of the kcng4b-C1 and -C2 mutants at 24, 48 and 72 hpf (Suppl. Table 1). We focused on genes expressed in the mechanosensory hair cells, sensory patches, and otic vesicles. First, genes whose expression was affected in the same way at all three developmental stages (marked in bold black font in the Suppl. Table 2-4). These included atp1b2b, atp2b1a, cldn7b, dachb, myo6b, sox9a, i.e., the genes expressed in the mechanosensory hair cells and otic vesicle.
Third, cldn7b, fgf3, fgf8a, gas8, gsdmeb, hapln1a, jag1 were expressed in two mutants in opposite ways during at least one developmental stage, which seems to be a common theme for changes in expression of all genes.
Previously, it was shown that in Kcnb1 mutants, otop1 expression increased in parallel with reduction in otolith number (Jedrychowska et al., 2021). According to the qRT-1 9 the level of otop1 expression changed dramatically. It was decreased in the kcng4b-C1 mutants ( Fig. 8A, B) and increased in the kcng4b-C2 mutant (Fig. 8A, C), like that in the Kcnb1 LOF mutant.

Discussion
The plasma membrane (PM) potential ( where Kcng4b LOF or Kcnb1 GOF increases and Kcng4b GOF or Kcnb1 LOF decreases Kv channel activity (Shen et al., 2016). The data presented illustrate a significant difference in inner ear development caused by mutations affecting the Kcng4b C-terminal.
The difference may be due to the structural difference between the putative mutant proteins. Kcng4b-C1 represents a milder version of the mutation causing Kcng4b LOF, which according to electrophysiological data (Table 1) is slightly less efficient in modulating the channel compared to the wildtype Kcng4b (Table 1), likely resulting in the mild channel GOF.
Surprisingly, such a small increase in channel activity triggers a rather distinct developmental effect (Fig. 5), in indication of importance of precise tuning of channel activity by the modulatory subunits. Moreover, the mild LOF effect in vitro is further enhanced by the progressive reduction of Kcng4b-C1 transcript in vivo (Suppl.  (Table 1), which alone can strongly block Kv channel activity. This effect seems to persist even at reduced transcript level (Suppl. Table 2, 3). Furthermore, the level of this transcript seems to increase in larvae up to 72 hpf (Suppl. Table 4). Of interest is a difference in the functional effects of the two potentially inhibitory forms of Kcng4b -Kcng4b-trunc and Kcng4b-C2 (Table 1), which cause opposite developmental effects ( Fig. 6; Suppl. Fig. 1). This is most likely because the Kcng4b-trunc is unstable resulting in the Kv channel GOF (Shen et al., 2016) unlike dominant-negative Kcng4b-C2, which causes Kv channel LOF.
The electrophysiology data (Fig. 4 These data suggested that a fraction of Kcnb1 homomers remains active in the kcng4b-C2 mutant. It may explain the lack of phenotype in the brain (Fig. 5), where the activity of the Kcnb1 homomer fraction may be just sufficient to avoid cell delamination. In contrast, in the ear, which is known to have a much higher concentration of K ions, this activity may not be sufficient to maintain otoliths development.
Previously, our study demonstrated a reduction in otic vesicle size and number of otoliths associated with hearing and balance defects in zebrafish kcnb1 LOF mutants (Jedrychowska et al., 2021). In the absence of otoliths in Kcng4b-C2, the developing zebrafish is likely to lack proper balance, as it was shown for embryos deficient in otoliths/otoconia (Hughes et al., 2004;Jedrychowska et al., 2021;Lundberg et al., 2015;Söllner et al., 2004). This illustrates that the role of Kcnb1-Kcng4b axis in the inner ear development is conserved in evolution between teleosts and mammals.
Language development is closely linked to hearing (Anne et al., 2017). In some human heterozygotic patients carrying a KCNB1/Kv2.1 mutation, speech development is impaired (Latypova et al., 2017;Marini et al., 2017;Bar et al., 2020). The current study of the two kcng4b mutants shows that different mutations affecting the silent α -subunit of the Kv channel can cause both LOF and GOF of the subunit, like what has been shown for the human electrically active subunit KCNB1. Given the antagonistic interaction between the two types of subunits (Shen et al., 2016), the LOF and GOF mutations of Kcng4 may result in the GOF and LOF effects on the functional heterotetramers of Kv, respectively.