Gap junction Delta-2b (gjd2b/Cx35.1) depletion causes hyperopia and visual-motor deficiencies in the zebrafish

The zebrafish is a powerful model to investigate the developmental roles of electrical synapses because many signaling pathways that regulate the development of the nervous system are highly conserved from fish to humans. Here, we provide evidence linking the mammalian connexin-36 (Cx36) ortholog gjd2b/Cx35.1, a major component of electrical synapses in the zebrafish, with a refractive error in the context of morphological, molecular, and behavioral changes of zebrafish larvae. Two abnormalities were identified. The optical coherence tomography analysis of the adult retina confirmed changes to the refractive properties caused by eye axial length reduction, leading to hyperopic shifts. The gjd2b/Cx35.1 depletion was also correlated with morphological changes to the head and body ratios in larvae. The differential expression of Wnt/ß-catenin signaling genes, connexins, and dopamine receptors suggested a contribution to the observed phenotypic differences. The alteration of visual-motor behavioral responses to abrupt light transitions was aggravated in larvae, providing evidence that cone photoreceptor cell activity was enhanced when gjd2b/Cx35.1 was depleted. The visual disturbances were reversed under low light conditions in gjd2b −/− /Cx35.1−/− larvae. Since qRT-PCR data demonstrated that two rhodopsin genes were downregulated, we speculated that rod photoreceptor cells in gjd2b/Cx35.1−/− larvae were less sensitive to bright light transitions, thus providing additional evidence that a cone-mediated process caused the VMR light-ON hyperactivity after losing Cx35.1 expression. Together, this study provides evidence for the role of gjd2b/Cx35.1 in the development of the visual system and visually guided behaviors.


Introduction
The transient coupling of groups of neurons by electrical synapses (gap junctions) is a phenomenon documented in different regions of the mammalian central nervous system across all life stages (Connors and Long, 2004;Connors, 2017;Pereda, 2019). This coupling is thought to be essential in several developmental events, including neuronal differentiation, cell death, cell migration, synaptogenesis, and neural circuit formation (Belousov and Fontes, 2013;Palacios-Prado et al., 2014;Haas et al., 2016;Pereda, 2016;Nagy et al., 2019). The formation of neuronal circuits and synchronized spontaneous activity are considered a hallmark of gap junction activity in the developing brain (Belousov and Fontes, 2013;Pereda, 2014).
At the molecular level, the principal component of mammalian neuronal gap junctions is connexin-36 (in mouse: Gjd2/Cx36) (Zoidl and Dermietzel, 2002;Connors and Long, 2004;Meier and Dermietzel, 2006). While the roles of Cx36 in the mature nervous system have been well established (Nagy et al., 2018;Nagy et al., 2019), less is known about the functions during mammalian development. It is thought that Cx36 serves canonical functions by coordinating metabolic and electrical activities in developing neurons and neuronal networks, but details of these roles remain to be resolved (Belousov and Fontes, 2013).
The application of gene-editing tools in the zebrafish has allowed for new lines of investigation to provide insight into the diverse roles of Cx36 orthologs during development via loss-of-function studies. Research focusing on the Mauthner neural circuit, which mediates a fast escape behavior, demonstrated that the two Cx36 orthologs, gjd1a/Cx34.1 and gjd2a/Cx35.5, are required for synapse assembly and function at larval stages . The investigation also identified a genetic basis for molecular asymmetry at a vertebrate electrical synapse and showed that they are required for appropriate behavioral performance . Another study demonstrated that gjd2b/Cx35.1 mediated gap junctions regulate glutamatergic synapse formation and dendritic elaboration in Purkinje neurons of 5-8dpf larvae (Sitaraman et al., 2021). In a third study, depletion of gjd2a/Cx35.5 or gjd2b/Cx35.1 in zebrafish caused changes in the biometry and refractive status of the eye (Quint et al., 2021). The gjd2a/Cx35.5 depletion led to a hyperopic shift and electrophysiological changes in the retina, supporting a role for gjd2a/Cx35.5 in regulating ocular biometry. Gjd2b/Cx35.1 had an additional lenticular role, leading to a nuclear cataract that triggered axial elongation.
This study investigated the functional evidence linking gjd2b/ Cx35.1 and refractive error in the context of morphological, molecular, and behavioral changes of 6-7dpf larvae when the visual response system is developed (Niell et al., 2004;Niell and Smith, 2005). We identified two abnormalities. An optical coherence tomography (OCT) analysis of the adult retina confirmed changes to the refractive properties caused by eye axial length reduction, leading to hyperopic shifts. Furthermore, gjd2b/Cx35.1 loss was correlated with morphological changes to the head and body ratios in larvae. Changes to Wnt/ß-catenin signaling, connexin (gjd1b/ Cx34.7, gja3/Cx48.5), and dopamine receptor expression suggested a contribution to the observed phenotypic differences. The alteration of visual-motor behavioral responses to abrupt light transitions was aggravated, providing evidence that cone-cone photoreceptor cell activity was enhanced when gjd2b/Cx35.1 was lost. Visual disturbances were reversed under low/mesopic light conditions in gjd2b −/− /Cx35.1 −/− larvae. Since qRT-PCR data demonstrated that rhodopsin genes were downregulated, we speculated that rod photoreceptor cells (PRCs) in gjd2b −/− /Cx35.1 −/− larvae were less sensitive to bright light transitions, thus providing additional evidence that a cone-mediated process caused the VMR light-ON hyperactivity after losing Cx35.1 expression. Together, the main results of this study provide evidence for distinct roles of gjd2b/ Cx35.1 in the development of the visual system affecting visually guided behaviors.

Zebrafish lines and husbandry
Zebrafish (Danio Rerio) of the strain Tubingen long fin (TL) were maintained in groups with mixed sex in a recirculation system (Aquaneering Inc., San Diego, CA) at 28°C on a 14h light/10h dark cycle. All animal work was performed at York University's zebrafish vivarium and in a S2 biosafety laboratory in accordance with the Canadian Council for Animal Care guidelines after approval of the study protocol by the York University Animal Care Committee (GZ#2019-7-R2). The Cx35.1 mutant line was generated and characterized in-house.

Generation of the
The Tupfel longfin (TL) strain were used to generate the gjd2b −/− /Cx35.1 −/− line in-house using the CRISPR-Cas9 system (Jinek et al., 2012;Li et al., 2016). Potential CRISPR target sites of the gjd2b/Cx35.1 gene were identified using the E-CRISP tool (www.e-crisp.org). A CRISPR sgRNA (Synthego, Redwood City, CA, United States of America) targeting a XhoI restriction site in exon one of the gjd2b gene was selected ( Figure 1A). The sequence was: 5′-CUCUUAACAGGUAAGGGGGU-3.' CRISPR sgRNA/Cas9 complexes were formed at a 1:2 ratio supplemented with 0.05% phenol red prior to microinjections. For microinjection one-stage TL embryos were placed in a chilled chamber (2% agarose supplemented with E2 embryo medium) and subsequently injected with 2 nL of the Cas9:sgRNA duplex (400 pg). At 3dpf, injected larvae were genotyped. The remaining embryos of the founding generation (F0) were raised to adulthood. Founders were outcrossed to TL fish for two subsequent generations to reduce the risk of CRISPR off-target effects. Germline transmission was tested in each generation (F1-F2) and heterozygous progeny were raised to adulthood and in-crossed to generate the homozygous gjd2b −/− / Cx35.1 −/− zebrafish line.

DNA extraction, PCR amplification, and genotyping
Zebrafish larvae (3dpf) or adult caudal fins were used for genotyping purposes. Adult zebrafish were anesthetized in pHbuffered 0.2 mg/mL tricaine methane sulfonate (MS-222) (Sigma-Aldrich, St. Louis, MO, United States of America) prior to caudal amputations with surgical scissors. Total DNA was extracted by incubation in 100 mM NaOH for 15 min at 95°C using with vigorous shaking. DNA samples were diluted in a 1:2 ratio with TE buffer (pH 8.0) prior to polymerase chain reaction (PCR) analysis. The gjd2b exon one region was amplified in a Mastercycler Nexus X1 (Eppendorf, Mississauga, ON, CA) using the primer information as follows: forward primer: 5′-GGTTCTCTGTGTTACATTCGC CTCC-3′ and reverse primer: 5′-CAATCATAGTAGAGTGCT GTTGGACAGC-3'. PCR primer pairs were ordered from IDT (Coralville, IA, United States of America) and optimized at an annealing temperature of 58°C. The PCR reaction was carried out using the Q5 High Fidelity DNA Polymerase PCR Kit (New England BioLabs, Ipswich, MA, United States of America) in a final volume of 20 µL consisting of 1 µL DNA-template and 19 µL Taq master mix. PCR thermal cycling conditions were initiated at 94°C for 15 min, followed by 38 cycles of denaturing at 94°C for 15 s, annealing at 58°C for 15 s, and extension at 72°C for 30 s. The final extension was carried out at 72°C for 10 min. To determine the presence of indel mutations, PCR amplicons were digested with XhoI using the Fast Digestion Top Fermentas Kit (Thermo Fisher Scientific; Burlington, ON, Canada) as per the manufacturer's protocol. Amplicons were subsequently visualized on a 2% agarose gel supplemented with 0.005% ethidium bromide using the Alpha Imager HP System gene structure and partial DNA sequence. The position of forward and reverse primers targeting exon one flanked the region containing CRISPR mutations. The sgRNA target with PAM sequence is outlined by the purple box; The predicted CRISPR-Cas9 binding region targeting a XhoI restriction site is indicated by the purple arrow. The CRISPR-Cas9 genome engineering strategy produced a 1bp substitution at position 12 of exon 1, resulting in the substitution of G to A in Cx35.1 −/− animals. The mutation is outlined with a green box. (B) Cx35.1 partial protein sequence. The mutation resulted in an early stop codon, as indicated by the purple asterisk (*). (C) Immunohistochemistry of gjd2b −/− /Cx35.1 −/− larvae. (C1-left) A monoclonal anti-Cx36 primary antibody (Invitrogen) was applied to 6dpf wild type (TL) to determine Cx35 immunoreactivity in a frontal section of a zebrafish larva. The retina of gjd2b −/− /Cx35.1 −/− fish showed fluorescence in the inner plexiform layer (IPL) and photoreceptor cell layer (PRL). The optic nerve (ON), optic chiasm (OC), and arborization fields of the retinotectal tract (AF) were immunoreactive. Scale bar 100 µm. (D, E) The immunoreactivity in the IPL and PRL of wild type TL (top) was reduced in gjd2b −/− /Cx35.1 −/− larvae (bottom). Arrows indicate the IPL. Scale bars 50 µm.

Quantitative real-time PCR (qRT-PCR) analysis
Total RNA (1 µg) was extracted and purified from pools of 30 zebrafish larvae at six to seven dpf using the RNeasy Plus Mini Kit (Qiagen, Germantown, MD, United States of America) as per the manufacturer's protocols. When appropriate, larvae were collected immediately following exposure to behavioral tasks, and the tissue was frozen at −80°C until further analysis. Fish were homogenized by bead beating in 1xTE buffer (pH 8.0). RNAs were reverse transcribed into cDNA using the iScript Reverse Transcription Supermix (Bio-Rad, Mississauga, ON, Canada) as per the manufacturer's instructions. Gene expression was analyzed by qRT-PCR using the SsoAdvanced universal SYBR Green Supermix (Bio-Rad, Mississauga, ON, Canada) in the CFX96 ™ Real-Time PCR Detection System (Bio-Rad, Mississauga, ON, Canada). Thermal cycling was carried out for 42 cycles of the following: 94°C for 30s, 52°C for 30s, and 72°C for 1min. Three housekeeping genes (18S, tuba1a, actb2, Supplementary Table S7) were used to determine the quality of the samples and for normalization purposes. CT values were averaged and exported from the CFX Manager Software (Bio-Rad, Mississauga, ON, Canada). The fold-difference for the relative gene expression was calculated in the Relative Expression Software Tool VS. 2009 (Pfaffl et al., 2002). Three technical replicates per gene were performed from three biological replicates. Primers are listed in Supplementary Tables S1, S3-5.

Optical coherence tomography assay
A custom-developed spectral-domain optical coherence tomography (SD-OCT) system was built in Michelson configuration (Supplementary Figure S1), employing a superluminescent laser diode centered at 1,310 nm ( ± 75 nm at 10 dB; Exalos, Switzerland) and a 2048-pixel line scan camera spectrometer with a maximum acquisition rate of 147 kHz (Wasatch Photonics; United States of America). A 50/50 fiber coupler splits the source light into the reference and sample arms. In the sample arm, the output light illuminates the sample surface after passing through a reflective beam collimator (Thorlabs; United States of America), a 2-DOF galvo mirror, and an objective lens (LSM02, Thorlabs; United States of America). The 2-DOF Galvo mirror allows for collection of reflected light from the sample while raster scanning sample surface. In the reference arm, a polarization controller, a dispersion compensation block, and a gold-coated reference mirror were installed. The backreflected light of these two arms is subsequently merged after passing through the beam splitter and redirected to the spectrometer by the optical circulator. The formed interference pattern in the spectrometer is captured by a line scan camera. The captured signal is digitized and sent to the computer for processing. To form an A-line (i.e., depth profile of sample reflections at a given point on sample surface), the tomograms of the sample, the spectrometer signal is background subtracted and mapped to k-space before applying Fourier transformation to calculate depth profile of reflectors in the z-space (physical depth space). Aforementioned processes are repeated for data acquired during raster scanning to eventually form 3-D OCT tomograms of zebrafish eye. The axial and lateral resolutions of the system in tissue were measured as 8.5 µm and 10 μm, respectively.
Before imaging age-matched adult zebrafish were humanely euthanized using MS222 and fixed in 4% paraformaldehyde (PFA) in 1xPBS overnight at 4°C. Fixed zebrafish (n = 26/genotype) were placed in a silicon mold to orient the eye toward OCT system's objective lens. To minimize specular reflections from sample surface, a thin layer of PBS (~80 µm) was placed over the eye before imaging. All captured OCT tomograms were interrogated in ImageJ software which enabled measurement of various geometrical parameters (see Supplementary  Table S3) through a standard calibration process.

Zebrafish body length and head measurements
Zebrafish embryos were raised in 10 cm ϕ Petri-dishes in population-matched groups. At 7dpf, individual larvae were

Behavioral assays
At 6dpf, larvae were acclimated to the behavioral room for 24 h. On the morning of testing (7dpf), individual larvae were transferred into 48-or 96-well plates (Greiner) (for free-swim, VMR, and FSTR) filled with egg water. Transparent plates were utilized for all experiments. The cross-maze to assess color vision was generated in-house. Behavioral assays were performed using the Zebrabox ® system and the Zebralab analytical software (Viewpoint Life Technology, Lyon, Fr) to obtain automated tracking data from video files. The recording chamber holding the plates was equipped with a water heating and flow system to maintain the temperature at 28°C. Illumination was from the bottom with LED and infrared lights. Unless otherwise stated, larvae underwent a dark-adaptation period for 2 hours prior to behavioral analyses. All experiments utilizing light were performed at 30% light intensity unless otherwise stated, where the Zebrabox ® system was capable of 8,000 lux at 100%. Data was acquisitioned in 1s, 10s, 1 min, or 5min intervals for time periods ranging from 5 min to 4.5 h as outlined in figures. Data was acquisitioned in 1s, 10s, 1 min, or 5min intervals for time periods ranging from 5 min to 4.5 h as outlined in figures. The experimental designs for VMR (Emran et al., 2008), Color vision (Park et al., 2016), were adapted from previous reports. Activity was categorized by inactivity (0.0 mm/s), coast swimming (between 0.0 mm/s and 20.0 mm/s) and burst swimming (>20 mm/s). Where applicable, results were normalized in a three step process to account for differences in larval activity caused by 1) variation in light-intensity received by each well, 2) the batch effect from the inclusion of experimental replicates and 3) variation in the baseline activity measured during the final 5 min of light or dark adaptation based on a previous report (Xie et al., 2019).

Spontaneous/free swim assay
Spontaneous zebrafish larval locomotor activity was measured under constant illumination (30%) or darkness (0%). For constant illumination, larvae were acclimated in light (30%) for 2 h prior to behavioral analysis. Locomotor activity was tracked for 5 min or 1 h, depending on the type of data acquisition. Data were normalized accordingly.

Visual Motor Response (VMR) assay
Responsiveness to light changes was measured and under mesopic conditions as described (Emran et al., 2008). Larvae were dark-adapted for 2 h prior to the experiment commencing. Locomotor activity was tracked for 3.5 h (3 h of the experimental protocol and 30 min prior to illumination) for the standard set-up. For mesopic VMR, locomotor activity was tracked for 2.5 h (2 h of the experimental protocol and 30 min prior to illuminations). For both assays, two to three trials of alternating lighting conditions (30 min Light-ON; 30 MIN light-OFF) were utilized. Light-ON intensity for mesopic VMR was set at 5% as indicated in the results text. Mean activity, quantified by the burst duration s), was measured 1s following light transition. Data were normalized accordingly.

Innate color preference test
The innate color preference test allowed for color bias and colorblindness to be assessed in zebrafish larvae. Here, a custom-made cross maze was milled from 5 mm transparent acrylic sheets. Each arm of the maze was of equal dimensions (15 × 35 × 10 mm, W × L ×H) and could accommodate removable sleeves of different colors (B-blue, G-green, Y-yellow, R-red, Null-no color). The test was performed in two variations. In the first set larvae were exposed to all four colors simultaneously. In the second test larvae were exposed to two colors and a neutral stimulus (white light). Up to 20 larvae of the same genotype were placed into the center of the cross maze. Larvae were dark-adapted to avoid pre-exposure to the colors, and color sleeves were rotated to avoid location bias. Locomotor activity was tracked for 40 min (35 min of illumination and 5 min prior to illumination). Data was measured manually from tracking videos by counting the number of fish in each arm every 2 min for the duration of the experiment.

Flash stimuli threshold response (FSTR) assay
Responsiveness to light flashes was assessed with the FSTR assay.
Various stimulus lengths were sequentially tested (in ms, 10, 20, 50, 100, 250, 500, 1,000]); however, the final analysis utilized the 10 m and 1000 m experimental endpoints. The experimental design began with 2 h dark adaptation and was followed by alternating a light cue with a 20-min rest period in darkness To avoid habituation to the stimuli, cue lengths were presented in forward and reverse order. Larval locomotor activity was tracked for 2.5hrs; the last 5 min of adaptation period prior to the presentation of the initial light cue was used in normalization calculations. Mean responsiveness, quantified as burst duration s), 1 s following each stimulus was measured. Data were normalized accordingly.

Statistical analysis
Unless otherwise stated, all data is presented as the mean ± SEM. The sample size for behavioral studies varied from n = 48 to n = 144 across a minimum of two experimental replicates; sample sizes are indicated in the figures. Statistical significance was determined by using a two-tailed unpaired Student's t-test or its non-parametric equivalent Mann-Whitney, one-way ANOVA, or its nonparametric equivalent or Kruskal-Wallis, or two-way ANOVA with a Tukey post hoc test where appropriate. Normality was tested with the Shapiro-Wilk normality test. Statistical significance was defined by a p-value of ≤0.05. Statistical analyses were performed in GraphPad Prism 6 (GraphPad Software, La Jolla, CA, United States of America). Gardner-Altman estimation plots were created in a R-statistical program (The R Foundation, Aukland, New Zealand; https://cran.r-project.org/web/packages/dabestr/ vignettes/using-dabestr.html); they show the mean difference between groups. Both groups are plotted on the left or top axes; the mean difference is plotted on floating axes on the right or bottom as a bootstrap sampling distribution. The mean difference is depicted as a dot; the 95% confidence interval is indicated by the Frontiers in Cell and Developmental Biology frontiersin.org ends of the vertical error bar. All remaining figures were created in GraphPad Prism 6.

Generation and primary characterization
.1 −/− ) model was created utilizing a CRISPR-Cas9 genome engineering strategy and screening for indel mutations generated by the non-homologous end joining (NHEJ) repair system. Exon one of the gjd2b gene was targeted with a sgRNA located upstream of the start codon ( Figure 1A). A founder (F0) with a minimal 1bp substitution (position G12 →A) was chosen because this mutation altered amino acid 4 (tryptophan 4) to an early stop codon ( Figures 1A,B). The F0 mutation was germline transmitted and after outcrossing with TL for two generations the homozygous gjd2b −/− /Cx35.1 −/− line was established in the F3 generation.
Previous reports have demonstrated the expression of Cx35 in the zebrafish and perch retina (McLachlan et al., 2003;O'Brien et al., 1998); however, these results were likely confounded by the coexpression of closely related Cx35 proteins. With no access to antibodies discriminating the four zebrafish Cx35 orthologs gjd1a/Cx34.1, gjd2b/Cx35.1, gjd1b/Cx34.7, and gjd2a/Cx35.5, we investigated the Cx35.1 −/− phenotype by visualizing changes in the immunofluorescence localization patterns detectable with a monoclonal anti-Cx36 antibody in 6-day post fertilization (6dpf) larvae ( Figure  1C). The antibody detected Cx35/ 36 immunoreactivity in the retina, the optic nerve, and the arborization fields where axons from retinal ganglion cells are expected to contact dendrites from neuronal cells in the retinotectal pathway. A reduction in the fluorescence signal was observed in the inner plexiform layer (IPL) and between photoreceptor cells (PRCs) of gjd2b −/− /Cx35.1 −/− larvae. The residual signals in the remaining PRCs and throughout the retina was attributed to the expression of other Cx35 orthologs that we could not positively identify by immunohistochemistry.
Following the confirmation of the Cx35.1 −/− genotype and partial loss of Cx35/36 immunoreactivity, we tested whether targeted ablation of the gjd2b gene causes a compensatory regulation of connexins in larvae. A previous RNA-seq analysis of 6dpf WT larvae had demonstrated the expression of two of the four zebrafish Cx36 orthologs; gjd1a/Cx34.1 and gjd2b/Cx35.1 were detected, but not gjd1b/Cx34.7 or gjd2a/Cx35.5 (GEO database entry GSE181853; (Whyte-Fagundes et al., 2022). Having a key interest in the role of Cx35.1 in vision, the expression of all four Cx36 orthologs and nine other connexin genes expressed in the eye were quantified (Supplementary Table S1) (Cheng et al., 2003;Zoidl et al., 2008;Yoshikawa et al., 2017). The RNA transcript levels of this representative subset of connexins expressed in the retina was compared between WT and Cx35.1 −/− larvae using qRT-PCR (Supplementary Figure S2). In comparison to WT larvae, the expression of gjd1b/Cx34.7 was reduced (0.680-fold, p = 0.014). Both gjd1b/Cx34.7 and gjd2b/Cx35.1 are co-expressed in cone photoreceptors (O'Brien et al., 1998;O'Brien et al., 2004), forming heterotypic gap junctions (Rash et al., 2015) which form the molecular basis of functional asymmetry of electrical synapses (Rash et al., 2013). The loss was evidence for altered visual functions. Furthermore, a reduction of gja2/ Cx39.9 (0.314-fold, p = 0.005) and gja3/Cx48.5 (0.608-fold, p = 0.022) expression was detected. The expression of other connexins were indistinguishable from the WT. The regulation of gja2/Cx39.9 (Cx46 in human and murine models), and gja3/ Cx48.5 (Cx50 in human and murine models) was of interest due to the role of both connexins in ocular lens homeostasis (Cheng et al., 2003;Brink et al., 2020). In summary, the initial characterization of gjd2b −/− /Cx35.1 −/− larvae showed evidence that the depletion of Cx35.1 protein expression affected the expression of other connexins with roles in the communication between PRCs and in the zebrafish lens.
Statistical analysis of quantified parameters in Figure 2 C-I revealed no significant difference between ratio of proximal-distal and anterior-posterior lens diameters (i.e., lens ratio) of TL controls and gjd2b −/− /Cx35.1 −/− (in micrometer/micrometer, TL:1.01 ± 0.02; Cx35.1 −/− : 1.02 ± 0.02; p = 0.150, n = 26/genotype); similar trend was observed in the measured retinal thicknesses of TL controls and  (Collery et al., 2014), Relative Refractive Error (RRE) was calculated for TL controls and mutants as 1-(retinal radius/ F); here, parameter F is an idealized focal length equal to the measured lens radius×2.324. The significance of RRE values is that they offer a normalized/unitless metric with values lower than zero for myopic eyes (i.e., when distance from lens center to RPE is greater than the expected retinal radius) and values greater than zero for hyperopic eyes (i.e., when distance from lens center to RPE is less than the expected retinal radius). Accordingly, an increase in RRE value highlights a hyperopic shift, while a decrease in RRE value suggests myopic shift (Collery et al., 2014). Statistical comparison of calculated RREs demonstrated a statistically meaningful increase in RRE values of gjd2b −/− /Cx35.1 −/− over the TL controls (in micrometer/ micrometer, TL: 0.19 ± 0.08; Cx35.1 −/− : 0.26 ± 0.08; p = 0.006, n = 26/genotype); the positive values of RRE for the TL controls were due to the fixation process which makes the eye slightly hyperopic as reported by others (Collery et al., 2014). The above analysis confirmed our hypothesis that loss of gjd2b/ Cx35.1 function contributed to a hyperopic phenotype in adult fish.
These findings led us to investigate whether the differences in size were balanced by measuring the head-to-body and head-tomidbrain ratios. No significant differences in the head-to-body ratio were detected, suggesting that at 7dpf the gjd2b/Cx35.1 −/− larvae were objectively larger in their body and head length (WT: 0.18  In contrast, the head-to-midbrain ratio of gjd2b/Cx35.1 −/− larvae was smaller (WT: 0.74 ± 0.71, Cx35.1 −/− : 0.69 ± 0.01, p = 0.008) ( Figure 4G). This suggested that gjd2b/Cx35.1 expression affected larval growth and development. The reduction of the midbrain diameter detected in gjd2b −/− /Cx35.1 −/− larvae suggested functional consequences to visual sensory integration and motor output pathways in knockout larvae.

Loss of gjd2b/Cx35.1 altered larval dominant swimming competence in darkness
The structural and genetic changes detected in gjd2b −/− / Cx35.1 −/− larvae suggested visual-motor deficiencies causing testable behavioral phenotypes. To address this question, spontaneous locomotor activity was quantified in 7dpf zebrafish in light-ON or light-OFF conditions. Both WT and gjd2b −/− / Cx35.1 −/− larvae increased the swimming activity in light-ON conditions relative to light-OFF (WT: p < 0.001; Cx35.1 −/− : p < 0.001; both n = 72). However, the swimming distance of Next, differences in swim bouts were investigated. Larvae utilize a combination of slow rhythmic and fast burst bouts, known as the burst-glide swimming style, during free swim (Muller et al., 2008;Voesenek et al., 2018). Mechanisms underlying slow rhythmic (swim speeds less than 20 mm/s) and fast burst (speeds ≥20 mm/s) bouts represent distinct motor neuron circuits and subsequent muscle recruitment. Therefore, both dominant (coast) and subordinate (burst) swimming competence, mediated by slow motor neurons and Mauthner neurons, respectively, were quantified separately during free swim.  3.6 Cone photoreceptor cell activity is enhanced in gjd2b −/− /Cx35.1 −/− larvae Since we previously confirmed that the loss of gjd2b −/− /Cx35.1 −/− altered retinal expression patterns, which was consistent with genetic profiles alluding to dysregulated eye development, we tested whether the retina was functionally intact using the standardized Visual Motor Response (VMR) assay ( Figure 7A). Due to the nature of the startle/escape behavior teleost species exhibit in response to a sudden change in light intensity, we measured the burst activity as an indication of changes in retinal signaling. Both genotypes demonstrated an immediate and robust increase in burst duration in response to the light-ON VMR transitions. Gjd2b −/− /Cx35.1 −/− larvae were hyperactive in comparison to the WT (in s, WT: 0.16 ± 0.01, Cx35.1 −/− : 0.18 ± 0.01, n = 48, p < 0.001, Mann-Whitney) (Figures 7B,C). In the light-OFF condition, the VMRs were indistinguishable (in s, WT: 0.28 ± 0.02, Cx35.1 −/− : 0.21 ± 0.01, p = 0.754) (Figures 7D,E). The results showed that gjd2b −/− /Cx35.1 −/− larvae readily detected light changes and were most sensitive to photopic light stimuli. This suggested that retinal signaling routed to cone photoreceptors, and downstream acting pathways were altered in gjd2b −/− /Cx35b −/− larvae. We hypothesized, however, that signaling to rod photoreceptors would be preserved.
Next, we modified the color preference test and restricted the color choice to two colors and a neutral choice. Here, the blue color remained attractive when compared to a neutral stimulus (in %,  Figure S6). In any other combination of green, red, and yellow stimuli, larvae of both genotypes preferred the neutral stimulus. All the above suggested that Cx35.1 depletion did not alter color vision despite the upregulation of opn1mw2 expression.

Discussion
In this study, we aimed to provide insight into the potential roles of gjd2b/Cx35.1 in the developing visual system of zebrafish larvae. We describe a new mutant line generated with Cas9/CRISPR technology, which differs from the gjd2b fh454  and gjd2b ncb215 (Sitaraman et al., 2021) mutant lines generated by TALEN technology. The general strategy to obtain the three lines was similar; all targeted exon one of the gjd2b gene close to the start codon. A difference was the use of multiple zebrafish strains; the transgenic line M/CoLo:GFP (Et (Tol-056:GFP)) with an AB/ Tu , Indian WT (Sitaraman et al., 2021), or TL background in this study. Another difference is the minimal 1bp substitution (position G12 →A) altering amino acid 4 (tryptophan 4) to an early stop codon (after amino acid 3) in our study. In contrast, the gjd2b ncb215 line has an insertion after the translation start site causing a frame shift and insertion of a premature stop codon, with a predicted truncated protein of 55 amino acid residues. How the different genetic backgrounds or the formation of a truncated protein contributed to phenotypic differences is presently unknown.
The depletion of Cx35.1 was investigated using an anti-Cx36 antibody (#37-4000; Thermofisher). This antibody was different to the anti-Cx35/36 antibody used previously to demonstrate Cx35.1 depletion (Quint et al., 2021;Sitaraman et al., 2021). Here, the immunoreactivity found in the photoreceptor and inner plexiform layers of the zebrafish retina at 6dpf was comparable to the rodent retina (Li et al., 2008) and previous reports showing staining of the neuropil in the optic tectum of larvae (Jabeen and Thirumalai, 2013), and at adult stages of the zebrafish (Li et al., 2009;. Further, the observed Cx36 immunoreactivity overlapped with the single cell RNA sequencing data from Quint et al. (2021), which showed expression in all neuron subtypes (Quint et al., 2021). Consistent with the introduction of an in-frame translation stop signal, the Cx36 immunoreactivity was diminished in gjd2b/Cx35.1 KOs with residual fluorescence attributed to other Cx36 orthologs. The characterization of the mutant line suggested that the loss of Cx35.1 function led to various morphological phenotypes, including biometrical changes to the eyes and head. The molecular changes were consequential, affecting connexin gene expression, two signaling pathways, and behavioral outcomes in response to temporal and luminescence changes of visual stimuli.

Gjd2b/Cx35.1 uncoupling causes biometrical changes to eyes and head
In humans, GJD2 (Cx36) is associated with refractive error (Kunceviciene et al., 2022). As discussed previously by Quint et al. (2021) two common genetic variants downstream of GJD2 (Cx36) have consistently been associated with refractive error (rs634990 and rs524952) and myopia. However, the coding region of GJD2 (Cx36) has not revealed coding variants explaining the genome wide association study results; it has been suggested that the identified variants might play a role through gene regulation. As pointed out by Quint et al. (2021) the identified risk variants for myopia might upregulate the expression of GJD2, facilitating the coupling of Cx36 (GJD2) gap junctions in the eye. Following this idea, the uncoupling of gap junctions could explain the opposite outcome.
In our study the Cx35.1 depletion caused no cataracts when the genetic background was TL. This phenotype differs from Quint et al. (2021) which describes lenticular cataract formation as the main outcome of Cx35.1 depletion. Instead, loss of Cx35.5 caused a hyperopic phenotype when the genetic background was AB. We do not disagree with the principal results of Quint et al. but like to point out that previously reported RNA-seq and the qRT-PCR results shown in this study demonstrate a robust expression of gjd2b/Cx35.1 in the TL strain. The robust expression of gjd2b/Cx35.1 is consistent with the reported single-cell RNA-seq data showing the expression of this gene in all major cell types of the developing retina. Interestingly, we found that gjd2b/ Cx35.1 depletion caused a downregulation of gjd1b/Cx34.7. Previous work has demonstrated that both genes are expressed in cone photoreceptors (O'Brien et al., 1998;O'Brien et al., 2004). Gjd2b/Cx35.1 and gjd1b/Cx34.7 form heteromeric gap junctions with asymmetric function in mixed electrochemical synapses (Rash et al., 2013;Rash et al., 2015). The concurrent depletion of the two connexins adds to the accumulating evidence for the molecular and functional asymmetry of electrical synapses (O'Brien, 2014;Palacios-Prado et al., 2014). The molecular mechanisms explaining the relationship between Cx35.1 mediated gap junction uncoupling and the observed changes to eye growth and head abnormalities are unknown. However, our results provide the first evidence that links Cx35.1 depletion with the Wnt/ß-catenin signaling pathway. Wnt/ß-catenin signaling is known to affect brain development and the formation of camera eyes at embryonic stages. Genome-wide association studies in humans have identified loci causing refractive errors which are linked to Wnt signaling (Cheng et al., 2013). In the zebrafish we identified a downregulation of wnt8b, wnt11, and the receptor fdz5 when gjd2b/Cx35.1 was lost. Wnt8a and wnt8b are required to activate the Wnt/ß-catenin signaling pathway and the integration of wnt11, fdz5, and Wnt/β-catenin signaling coordinates cell fate determination and morphogenesis of the developing brain (Cavodeassi et al., 2005). Wnt8b has roles in biological processes like eye, forebrain and hypothalamus development (Kim et al., 2002;Lee et al., 2006), and neuronal differentiation, including the cluster size of dopaminergic neurons (Russek-Blum et al., 2008). We propose that the Cx35.1 KO models generated in different research programs are assets for in-depth investigations into how Cx35.1 modulates Wnt/β-catenin signaling through gap junction (un)coupling during development and maturation.
Another point of interest in Cx35.1-deficient larvae was motivated by the regulation of the D2-receptor genes drd3 and drd4b, and vmat2 (alternative gene name: slc18a2). In mammals, five distinct types of dopamine receptors have been identified and are classified into two families; D1-like (D1R and D5R), which are stimulatory subsets to adenyl cyclase, and D2-like (D2R, D3R, and D4R) which inhibit adenyl cyclase (Girault and Greengard, 2004). D4R subtypes are expressed in rod and cone PRCs of the outer retina of different vertebrates (Alfinito and Townes-Anderson, 2001;Ostergaard et al., 2007;Ma et al., 2013), and drd3 gene transcripts have been found in teleost retinal ganglion cells (Boehmler et al., 2013). Literature provides well-described evidence of similar expression patterns for D1-like and D2-like subtypes across the retina of different vertebrates (Callier et al., 2003;Yamamoto et al., 2013). Typically, photopic conditions increase dopamine production (Witkovsky, 2004) and Vmat2 plays a role in vesicular dopamine transport (Esposito et al., 2012). PRC coupling has been linked to dopamine and the D4R subtype (Li et al., 2013;O'Brien et al., 2004). Li et al. (2014) has concluded that adenosine and dopamine coregulate photoreceptor coupling through opposite action on the protein kinase A (PKA) pathway and Cx36 phosphorylation (Li et al., 2013). Therefore, under photopic light conditions, gap junctions will be primarily de-phosphorylated, resulting in the uncoupling of PRCs; this is critical in supporting high contrast vision. The opposite occurs under scotopic vision where PRCs are coupled in a phosphorylation-dependent manner (Kothmann et al., 2007;Kothmann et al., 2009;Li et al., 2009). Considering this, we propose that the reduction of drd3 and drd4b receptor subtypes, together with the upregulation of vmat2, modulates the phosphorylation cascade. Although this provides evidence for a dopaminergic-dependent compensatory mechanism in Cx35.1 −/− larvae, it remains unclear which Cx36 ortholog(s) or other retinal connexins could be phosphorylated. An alternative explanation could derive from the functional antagonism of dopamine and adenosine receptors, specifically the A 2A and D2-like subtypes. In such circumstances, activation of A 2A receptors reduces in dopamine affinity (Franco et al., 2000). In this regard, the actions of A 2A receptors on the available D2-like receptors would complement proposed mechanism by reducing phosphorylated Cx36 orthologs in our gjd2b −/− /Cx35.1 −/− model. However, the relative expression of A 2A was not explored in this study and deferred to follow up studies. Further, we detected a downregulation of two connexins associated with expression in the eye (gja2/Cx39.9, gja3/Cx48.5) (Zoidl et al., 2008). The Cx48.5 protein is involved in post-embryogenic eye development (Cheng et al., 2003) and the gja3 s213/s215 and gja s226/s226 lines display a decreased eye size (Chi et al., 2008). Interestingly, experimental depletion of gja3/Cx48.5 with morpholinos caused cataracts, which we have not observed here. We conclude that the moderate downregulation detected in our fish model was insufficient to cause a similar phenotype. However, in mammalians and humans, the expression of mutants of the orthologous gene Cx50 is associated with cataract formation, smaller lenses, and eyes, indicating the importance of gap junctions in the maintenance of lens fiber homeostasis in mammals (Beyer and Berthoud, 2014;Berthoud et al., 2020;Retamal and Altenberg, 2022). Hirata et al. (2012) showed that morpholino-based depletion of gja2/Cx39.9 in the zebrafish does not affect the locomotory behavior but caused abnormal thigmotaxis and changes to muscle contraction (Hirata et al., 2012). It remains to be determined how this gene contributes to the observed phenotype of our model.

Gjd2b/Cx35.1 un-coupling alters visually evoked behaviors
To gain insight into how gjd2b/Cx35.1-mediated gap junctions regulate brain functions, we focused on testing visually evoked behaviors based on a significant reduction in Cx36immunoreactivity in the IPL and in a subpopulation of photoreceptor cells, which was attributed to the loss of Cx35.1. The IPL houses the synaptic connections between ganglion cells and amacrine and bipolar cells and is thought to be the region in which motion, brightness, contrast, and hue are processed (Wassle and Boycott, 1991;Calkins and Sterling, 1999). Pan et al. (2010) reported a partial or complete loss of ganglion-to-amacrine cell coupling in Cx36 knock-out mouse models concluding that Cx36 expression was critical in driving the connectivity between most ganglion cell subtypes (Pan et al., 2010). In our gjd2b −/− / Cx35b −/− larvae, the atypical pattern of fluorescence observed at the PRC layer suggests that gjd2b/Cx35.1 is localized to specific PRC subtype(s); the lack of access to a Cx35.1 specific antibody did not allow us to resolve this question. However, we also found Cx36 immunoreactivity in the optic nerve and the arborization fields (AF) where RGC axons contact dendrites of tectal neurons (Robles et al., 2013;Baier and Wullimann, 2021;Guggiana Nilo et al., 2021). Previous work also has demonstrated that Cx35.1 depletion altered dendrites and synaptic plasticity in Purkinje neurons of the zebrafish (Sitaraman et al., 2021). We Frontiers in Cell and Developmental Biology frontiersin.org reasoned that gjd2b/Cx35.1 could have similar roles in other brain regions and that altered dendrites and synaptic plasticity could affect processing of visual information by impairment of sensory integration and disturbances in motor outputs that manifest as measurable changes in locomotion and visual acuity. Considering the gjd2b/Cx35 expression in the visual system previously noted (Kothmann et al., 2007;Marsh et al., 2017;Miller et al., 2017), and the changes observed in our knock-out model, we explored several visually guided locomotor behaviors. Firstly, we assessed their free swim activity and found that gjd2b −/− /Cx35.1 −/− larvae were largely comparable to the wild-type. Due to the zebrafish's innate avoidance of darkness (Mansur Bde et al., 2014), we anticipated a reduction in overall swimming activity in Light-OFF (dark) conditions regardless of genotype. We did note that gjd2b −/− / Cx35.1 −/− larvae had a reduced duration of bout and glide behavior (defined as swim speeds >0 and <20 mm/s) that was exclusive to darkness, however, it is unclear whether this can be attributed to a motor-muscle deficiency, anxiety-like phenotype, or a combination of them both. Nevertheless, this served as our initial basis that processing of luminescence requires functional gjd2b/Cx35.1 expression.
The Visual Motor Response (VMR) assay (Emran et al., 2008;Liu et al., 2015) was used to better characterize how luminescence changes impacted the behavioral responses of larvae of both genotypes. The expectation that both rods and cone photoreceptor cells retained some functionality despite the loss of gjd2b/Cx35.1 was observed; gjd2b −/− /Cx35.1 −/− larvae responded to sudden changes in luminescence, although this was met with an increased sensitivity to light-ON transitions. Under this condition, cone photoreceptor stimulation and subsequent downstream ONbipolar cell activation is the primary signaling mechanism, providing evidence that gjd2b/Cx35.1 contributes to cone PRC activity. In support of this, exposing gjd2b −/− /Cx35.1 −/− larvae to a mesopic visual range (5% light intensity) reversed the light-ON VMR hyperactivity, indicating that the recruitment of rod photoreceptor cells was able to reverse this response. Li et al. have previously demonstrated that Cx35 electrical coupling facilitates both cone-cone and rod-cone photoreceptor crosstalk, the latter of which allows for saturated rod signaling to reroute to cones when exposed to the mesopic visual range (Li et al., 2009). Taken together, our results support previous research regarding the role of Cx35.1 in mediating cone PRC coupling (Li et al., 2009). We propose that gjd2b/Cx35.1 mediates cone PRCs crosstalk with a primary role in fine-tuning communication between nerve cells. Since larvae had altered D2-receptor family and vmat2 expression, we suspect that pre-and/or postsynaptic dopamine signaling may contribute to the observed cone-mediated sensitivity. However, this hypothesis requires further study of how changes in dopaminergic signaling alter cone functions and its similarity to the known effects on visual acuity following depletion of the retinal dopaminergic interplexiform cells (Li and Dowling, 2000).

4.4
Gdj2b/Cx35.1 contributes to a spatiotemporal processing but not chromaticity An alternative explanation to light-ON VMR hyperactivity takes into consideration the upregulation of the opsin gene op1mw2. Berry et al. (2019) have found that viral-delivered medium-wave opsin restored visual sensitivity and speed in blind rd1 mice to allow for spatiotemporal discrimination under dim-light and room-light (photopic) conditions; a result rhodopsin was unable to achieve (Berry et al., 2019). As such, we cannot discount that op1mw2 upregulation may have provided additional visual sensitivity, though we believe this is unlikely since we provide evidence that gjd2b/ Cx35.1 uncoupling does not alter the spectral sensitivity. Previous reports have revealed the innate color preferences of adult zebrafish (Avdesh et al., 2012;Bault et al., 2015) and larvae as early as 5dpf (Park et al., 2016). Although the reported evidence for red, green, and yellow color preferences is inconsistent in the literature, zebrafish larvae in this study consistently presented a robust preference for blue. The preference was identical in both genotypes, which led us to conclude that the differential expression of the green cone opsin gene op1mw2 in gjd2b/Cx35.1 −/− larvae was insufficient to impact innate color preferences and thus spectral sensitivity.
The quality of visual processing can be quantified by the measurement of luminescence, chromaticity, and spatiotemporal properties of light stimuli. Our VMR studies established roles of gjd2b/Cx35.1 in the discrimination of luminescence and the color preference test addressed chromaticity; to overcome the limited information regarding spatiotemporal properties, the FSTR assay was performed. Here, our results revealed that gjd2b −/− /Cx35.1 −/− larvae responded to stimuli short in duration (10 m). In support of this result, Quint et al. (2021) demonstrated that electroretinograms of gjd2b −/− / Cx35.1 −/− measuring the B-wave responses in zebrafish were indistinguishable from the wildtype (Quint et al., 2021). Though we concluded that the visual processing speed remained intact in gjd2b −/− /Cx35.1 −/− larvae, the overall behavioral responses were more robust during VMR (i.e., with longer durations of light stimulation) than during FSTR. As such, this revealed an unexpected contribution of gjd2b/Cx35.1 expression to vision in mediating behavior based on the temporal features of the stimuli.
This study found that gjd2b/Cx35.1 depletion leads to changes in the refractive properties of the eye, leading to hyperopic shifts, and morphological changes to the head and body ratio of zebrafish larvae. The changes to Wnt/ß-catenin and dopaminergic signaling pathways, as well as retinal immunostainings, identify novel roles for gjd2b/Cx35.1. As such, the results provide evidence for the contribution of gjd2b/Cx35.1 in the development of the visual system and visually guided behaviors in zebrafish. To our knowledge, we are the first to describe these distinct behavioral responses in gjd2b/Cx35.1 deficient zebrafish. While the details surrounding junctional coupling in gjd2b −/− /Cx35.1 −/− zebrafish remain unclear at present, we speculate that they are caused at the synaptic level by changes to the plasticity and functional asymmetry of electrical synapses. In conclusion, the evidence provided from phenotypic, molecular, and behavioral analysis support that significant functions of gjd2b/Cx35.1 in visual processing through the differential discrimination of luminescence and temporal properties, mediated by cone photoreceptors, start at the early stages of development.
Frontiers in Cell and Developmental Biology frontiersin.org