Connexinplexity: the spatial and temporal expression of connexin genes during vertebrate organogenesis.

Animal development requires coordinated communication between cells. The Connexin family of proteins is a major contributor to intercellular communication in vertebrates by forming gap junction channels that facilitate the movement of ions, small molecules, and metabolites between cells. Additionally, individual hemichannels can provide a conduit to the extracellular space for paracrine and autocrine signaling. Connexin-mediated communication is widely used in epithelial, neural, and vascular development and homeostasis, and most tissues likely use this form of communication. In fact, Connexin disruptions are of major clinical significance contributing to disorders developing from all major germ layers. Despite the fact that Connexins serve as an essential mode of cellular communication, the temporal and cell-type-specific expression patterns of connexin genes remain unknown in vertebrates. A major challenge is the large and complex connexin gene family. To overcome this barrier, we determined the expression of all connexins in zebrafish using single-cell RNA-sequencing of entire animals across several stages of organogenesis. Our analysis of expression patterns has revealed that few connexins are broadly expressed, but rather, most are expressed in tissue- or cell-type-specific patterns. Additionally, most tissues possess a unique combinatorial signature of connexin expression with dynamic temporal changes across the organism, tissue, and cell. Our analysis has identified new patterns for well-known connexins and assigned spatial and temporal expression to genes with no-existing information. We provide a field guide relating zebrafish and human connexin genes as a critical step toward understanding how Connexins contribute to cellular communication and development throughout vertebrate organogenesis.

Introduction

Animal development and homeostasis require coordinated cellular communication. One method of mediating communication is gap junction (GJ) channels. GJs are intercellular channels that provide a direct path of low resistance for ionic and small molecule exchange between cells (Evans and Martin 2009). These channels are formed by the coupling of 2 apposed hemichannels each contributed by adjacent communicating cells (Evans and Martin 2009; Orellana ; Xing ). Additionally, hemichannels can work independently within a single cell’s membrane, where they can release small molecules such as ATP and glutamate into the extracellular space for paracrine and autocrine signaling (Orellana ; Xing ). The proteins that create GJ channels are evolutionarily unrelated in vertebrates and invertebrates (Beyer and Berthoud 2018). Yet, despite little sequence similarity (Alexopoulos ), the vertebrate Connexins and the invertebrate Innexin proteins have a similar structure, with both classes creating 4-pass, transmembrane-domain proteins that oligomerize to form each hemichannel within the plasma membrane (Beyer and Berthoud 2018). Moreover, the hemichannels and intercellular GJs created by Connexins and Innexins have similar structure and function (Beyer and Berthoud 2018). Outside of these traditional roles, Connexins can also modulate the formation of tunneling nanotubes that connect nonadjacent cells to facilitate longer distance communication (Soares ; Okafo ; Tishchenko ). These varied functions in cellular communication are likely utilized individually and in combination in all animal tissues (Oyamada ), yet are best studied in epithelial (Chanson ), neural (Rozental 2000), and vascular (Figueroa and Duling 2009) systems. In these systems, mutations in human Connexin-encoding genes have been linked to defects in the development, regulation, and function including skin disorders (Richard , 2000, 2002; Richard 2005), cataracts (Willoughby ; Wei ), deafness (Kelsell ; Xia ; Grifa ), cardiovascular disease (Jongsma and Wilders 2000; Yeh ; Li ), and gastrointestinal diseases (Temme ; Maes , 2015b). While Connexin channels serve as an essential form of cellular communication, the temporal and cell-type-specific expression patterns of connexin genes largely remain unknown.

A major challenge in characterizing connexin expression is the complexity of the gene family. In humans, there are 20 distinct connexin genes, and in other vertebrate lineages, the number of Connexin-encoding genes is similarly large and varies widely (Eastman ; Cruciani and Mikalsen 2007; Mikalsen ). Cell culture and in vitro work suggest that connexin complexity provides functional diversity governed by 4 general principles: first, hemichannels are created by hexamers of individual Connexin proteins (Tarzemany ); second, single or multiple Connexin proteins can contribute to hemichannel formation (homo- or heteromeric hemichannels, respectively; Brink ; He ; Koval ); third, GJs form intercellular channels via hemichannel docking at cell–cell junctions; fourth, each contributed hemichannel can contain the same or different Connexin proteins (homo- or heterotypic channels, respectively; Brink ; He ; Koval ). The combinatorial possibilities of the gene family are restrained by molecular engagement rules that limit which Connexins are compatible to form mixed channels (Bruzzone ; Elfgang ; Koval 2006; Koval ). These diverse possibilities culminate in each hemichannel having its own unique permeability properties, dependent upon the pore-lining amino acids and channel gating properties of the individual Connexins (Elfgang ; Weber ). These rules suggest that animals might take advantage of Connexin-based complexity in vivo to generate unique functional outcomes, but given the large number of genes, we know little about how vertebrates deploy this gene family.

Most of our knowledge of connexin expression in vivo comes from only a handful of well-characterized genes. These examples support the idea that connexins can be expressed in distinct tissues, such as in mouse where gap junction a1/Connexin 43 (Gja1/CX43) is expressed extensively in non-neuronal cells, including epithelia (Grueterich ), heart (Beardslee ; Li ), and glia (Contreras ). By contrast, Gjd2/CX36 is found almost exclusively in neurons (Srinivas ). Within the same tissue, connexin expression can have distinct temporal patterns, such as Gjb2/CX26 and Gjb1/CX32 that are both found in the developing mouse neocortex at distinct developmental time points (Nadarajah ). Within the group of well-studied Connexins, there are also a few enticing examples that suggest the rules of Connexin functional complexity found in vitro are relevant to in vivo function. For example, heteromeric channels formed by Gjb1/CX32 and Gjb2/CX26 are found in the mammary gland and the composition of channels changed during development (Locke ). Heterotypic GJs composed of gjd2a/Cx35.5 and gjd1a/Cx34.1 are found at electrical synapses of zebrafish Mauthner cells where each Connexin was required for the localization of the other in the adjacent cell and both were necessary for synaptic transmission (Miller ; Lasseigne ). Finally, replacing the coding region of Gja1/CX43 with either Gja5/CX40 or Gjb1/CX32 results in sterility, cardiac malformations and arrhythmias, and mothers unable to nourish their pups, suggesting that each Connexin has unique properties that contribute to cellular homeostasis that are not interchangeable with other Connexins (Plum ). While these examples provide a glimpse of functional complexity, understanding the expression of this gene family through vertebrate development remains the critical first step to decoding the complexity of connexin usage in vivo.

Here, we set out to examine the expression of all connexins in a vertebrate model system, the developing zebrafish, using single-cell RNA-sequencing (scRNA-seq) of cells derived from the entire animal during organogenesis [1–5 days postfertilization (dpf); Farnsworth ]. Our analysis of connexin expression patterns revealed several trends, including that few connexins are broadly expressed, but rather, most connexins are spatially restricted to tissue- or cell-type-specific expression patterns. Most cells contain combinatorial signatures of connexins with unique profiles within distinct tissues. Finally, connexin expression is dynamic with temporal changes across the organism, tissue, and cell type. Our results reveal the complexity of spatiotemporal connexin control, highlighting novel aspects of well-studied connexins and revealing patterns for connexin genes with no prior expression information. We provide a field guide to relate zebrafish and human connexins genes, based on evolutionary homologies and expression similarities. Collectively, this represents an important step toward understanding connexin gene contributions in cellular communication throughout organogenesis and provides a foundation for comparative analysis in vertebrates.

Materials and methods

Single-cell RNA-sequencing

Embryo dissociation and cDNA library prep

As described by Farnsworth , larvae from the Tg(olig2: GFP)vu12 and Tg(elavl3: GCaMP6s) backgrounds were pooled (n = 15 per replicate), with 2 replicates at each sampled timepoint (1, 2, and 5 dpf). Cells from entire larvae were dissociated using standard protocols (Farnsworth ). Dissociated cells were then run on a 10X Chromium platform using 10x v.2 chemistry aiming for 10,000 cells per run.

Alignment

To ensure that the full transcripts of the Connexin-encoding genes were represented in the dataset, we used gene models with lengthened 3′ UTRs across the zebrafish genome generated and validated by the Lawson Lab (Lawson ). We ensured that the connexin genes were annotated properly by comparing pooled deep-sequencing information and extended the 3′ UTR regions as needed. Using this updated GTF file, we aligned reads to the zebrafish genome, GRCz11, using the 10X Cellranger pipeline (version 3.1). The updated GTF and other materials can be found at https://www.adammillerlab.com/.

Computational analysis

Cells were analyzed using the Seurat (V3.1.5) software package for R (V4.1.0) using standard quality control, normalization, and analysis steps. We performed principal component analysis (PCA) using 115 PCs based on a Jack Straw-determined significance of P < 0.01. Uniform manifold approximation and projection (UMAP) analysis was performed on the resulting 49,367 cells with 115 principal components (PC) dimensions and a resolution of 15.0, which produced 238 clusters. Code for this analysis and other materials can be found at https://www.adammillerlab.com/.

Cluster annotation

The unique barcode assigned to each cell was extracted from the original Farnsworth dataset (Farnsworth ) and identified in our updated dataset. For each updated cluster, we analyzed the percentage of cells contributing which were associated with the original Farnsworth’s clusters. Frequently, we found the updated dataset contained clusters with a significant proportion of cells (>80%) from a single Farnsworth cluster, and in such, we transferred the annotation from the original cluster to the updated cluster. We also found instances of a single Farnsworth cluster breaking nearly evenly across 2 of the updated clusters—for example, the original dataset had a single “photoreceptor” cluster (cluster 115), whereas the updated data had 2 clusters (clusters 13 and 14) with cells from original photoreceptor cluster. Further analysis revealed that these 2 new clusters represented likely rods and cones. Finally, we also found updated clusters that did not have a clear previous annotation. In these instances, we analyzed the most differentially expressed genes from that cluster and compared them with canonical markers.

Fluorescent RNA in situ

Custom RNAscope probes to target connexin genes were designed and ordered through ACD (https://acdbio.com/; for probes, please see reagent table in Supplementary Table 5). For the fluorescent in situs, we used a modified RNAscope protocol (Gross-Thebing ). Briefly, 1 dpf embryos were fixed for 2 h at room temperature in 4% paraformaldehyde (PFA) and then stored in 100% methanol at −20°C. The tissue was then exposed to protease plus for 30 min, washed with PBS with 1% Triton X (PBSTx), and then hybridized with the 1× probe overnight at 40°C. Standard RNAscope V2 multiplex reagents and Opal fluorophores were used, with the modification that PBSTx that was used for all wash steps. Stained tissue was either mounted (whole mount) or immediately cryo-sectioned and mounted with ProLong Gold Antifade (ThermoFisher). Full protocol can be found at doi.org/10.17504/protocols.io.b47vqzn6.

Zebrafish husbandry

Fish were maintained by the University of Oregon Zebrafish Facility using standard husbandry techniques (Westerfield 2000). Embryos were collected from natural matings, staged, and pooled. Animals used in the original Farnsworth data were: Tg(olig2: GFP)vu12 and Tg(elavl3: GCaMP6s) (Farnsworth ), and animals used for RNAscope in situs were ABC-WT. Animal use protocol AUP-18-35 was approved by the University of Oregon IACUC committee and animal work was overseen by Dr. Kathy Snell.

Results

Zebrafish have 41 connexin genes

To understand connexin expression throughout organogenesis, we first set out to ensure the entire connexin gene family in zebrafish was identified. Previous efforts (Eastman ; Watanabe 2017) and a recent phylogenetic approach to identify the full teleost connexin family (Mikalsen ) captured 40 individual connexin genes. Through reciprocal BLAST analysis between the zebrafish genome and (1) human and (2) other teleost Connexin sequences, coupled with phylogenetic analysis, we identified the 40 previously noted connexins and one previously unreported connexin, gjz1, which is conserved in mammals but forms an outgroup with the rest of the Connexin proteins (Supplementary Figs. 1 and 2).

Across the family of connexin genes, there are 7 human connexins for which zebrafish only has a single homolog, 8 human connexins for which zebrafish has 2 homologs, 2 zebrafish connexins that have no direct homolog but share sequence similarity to human connexins, and 16 zebrafish connexins that are not present in humans but are conserved in other teleost and mammalian lineages (Mikalsen ). We summarize these relationships in Table 1, listing zebrafish connexin genes and their closest relationship with their human counterparts, providing known human and zebrafish expression patterns and phenotypes for comparison. For clarity, we denote connexins by their Greek name and by their predicted molecular weight, a naming structure consistent with HUGO (Bruford ) and ZFIN standards (Sprague ; Table 1, Supplementary Table 1). The table is organized to emphasize Connexin similarities based on evolutionary homology, protein sequence, and expression, in alphabetical order of zebrafish connexin genes and denotes human similarity across merged rows. There are a limited number of rows where the zebrafish connexin gene resembles its human counterpart(s), but the genes are not direct homologs. For example, the human GJB2/GJB6 genes are duplicated in the human lineage while having only a single similar gene in zebrafish called gjb8 (Mikalsen ). Despite not being direct homologs, expression and mutant analyses have found that zebrafish gjb8 and human GJB2/GJB6 genes are all involved in inner-ear support cell function and loss of these genes in their respective systems causes deafness (Grifa ; Snoeckx ; Chang-Chien ). The comprehensive list of 41 zebrafish connexin genes provided a basis to examine the expression patterns of this gene family.

Table 1.

A field guide to zebrafish connexins.

Zebrafish Connexin gene/protein	Phenotypes associated with zebrafish Connexin gene/protein	Updated scRNA-seq tissue/cluster	Human Connexin gene/protein	Diseases associated with human Connexin gene/protein
gja1a/Cx40.8	No known phenotype	Neural crest, connective tissue, and nervous system	GJA1/CX43	Bone, skin, eye, teeth, heart, and digit abnormalities (Dasgupta et al. 2001; Paznekas et al. 2003; Paznekas et al. 2009; Brice et al. 2013; Hu et al. 2013)
gja1b/Cx43	Smaller body shape and shortened fins, shorter vertebrae, disrupted regeneration, and diminished motile cilia (Haffter et al. 1996; Henke et al. 2017; Misu et al. 2016; Hoptak-solga et al. 2008; Zhang et al. 2020)	Broadly expressed
gja2/Cx39.9	Decreased skeletal slow muscle contractability (Hirata et al. 2012)	Skeletal muscle	—	—
gja3/Cx46	Heart abnormalities (Chi et al. 2008, 2010)	Lens, heart	GJA3/CX46	Cataracts (Mackay et al. 1999; Burdon et al. 2004; Yang et al. 2010; Yao et al. 2011)
gja4/Cx39.4	Disrupted pigment patterns (Watanabe et al. 2016)	Endothelial and pigment cells	GJA4/CX37	Cardiovascular abnormalities (Yeh et al. 2001)
gja5a/Cx45.6	Faster vessel growth (Denis et al. 2019)	Muscle and endothelial	GJA5/CX40	Cardiovascular abnormalities (Groenewegen et al. 2003; Makita et al. 2005; Gollob et al. 2006; Yang et al. 2010; Wirka et al. 2011)
gja5b/Cx41.8	Leopard pigment patterns and faster vessel growth (Frohnhöfer et al. 2016; Watanabe et al. 2016; Henke et al. 2017; Watanabe 2017; Denis et al. 2019)	Pigment cells and endothelial
—	—	—	GJA6P/CX43px	—
gja8a/Cx79.8	No known phenotype	Lens	GJA8/CX50	Cataracts (Berry et al. 1999; Polyakov et al. 2001; Willoughby et al. 2003; Hansen et al. 2007)
gja8b/Cx44.1	Cataracts (Ping et al. 2021)	Lens
gja9a/Cx55.5	Disrupted perception of light stimulation (Klaassen et al. 2011)	Nervous system and integument	GJA9/CX58	No known implications
gja9b/Cx52.9	No known phenotype	Retina
gja10a/Cx52.7	No known phenotype	Low expression in this dataset	GJA10/CX62	No known implications
gja10b/Cx52.6	No known phenotype	Retina
gja11/Cx34.5	No known phenotype	Low expression in this dataset	—	—
gja12.1/Cx28.9	No known phenotype	Liver, intestine, and kidney	—	—
gja12.2/Cx28.1	No known phenotype	Intestine	—	—
gja13.1/Cx32.3	No known phenotype	Liver, intestine, and kidney	—	—
gja13.2/Cx32.2	No known phenotype	Macrophage	—	—
gjb1a/Cx27.5	No known phenotype	Schwann cell	GJB1/CX32	Neuropathy (X-linked Charcot-Marie-Tooth; Bergoffen et al. 1993; Ionasescu et al. 1996)
gjb1b/Cx31.7	Disrupted spacing of Muller glia cells (Charlton-Perkins et al. 2019)	Schwann cell
gjb3/Cx35.4	No known phenotype	Integument	GJB3/CX31	Deafness and skin abnormalities (Xia et al. 1998; López-Bigas et al. 2001; Richard et al. 1998, 2000; Richard 2005)
gjb7/Cx28.8	No known phenotype	Integument, hair cell, and olfactory neurons	GJB7/CX25	No known implications
gjb8/Cx30.3	Disrupted inner-ear development (Chang-Chien et al. 2014)	Integument, pigment cell, endothelial, and hair cell	GJB2/CX26	Deafness and skin abnormalities (Kelsell et al. 1997; Willems 2000; Richard et al. 2002; Richard 2005; Iossa et al. 2011)
			GJB6/CX30	Deafness and skin abnormalities (Grifa et al. 1999; del Castillo et al. 2002; Lamartine et al. 2000; Richard 2005)
gjb9a/Cx28.6	No known phenotype	Integument	—	—
gjb9b/Cx30.9	No known phenotype	Integument, macrophage
gjb10/Cx34.4	Impaired cardiac function (Okamoto et al. 2020)	Integument, neural crest, and nervous system	GJB4/CX30.3	Skin abnormalities (Macari et al. 2000; Richard 2005)
			GJB5/CX31.1	No known implications
gjc1/Cx52.8	No known phenotype	Muscle, neural crest, and nervous system	GJC1/CX45	No known implications
gjc2/Cx47.1	No known phenotype	Schwann cell	GJC2/CX47	Myelin disorders and lymphatic abnormalities (Uhlenberg et al. 2004; Orthmann-Murphy et al. 2007; Ferrell et al. 2010)
—	—	—	GJC3/CX29	No known implications
gjc4a.1/Cx44.2	No known phenotype	Vasculature and integument	—	—
gjc4a.2/Cx44.5	No known phenotype	Integument
gjc4b/Cx43.4	Disrupted left/right symmetry and abnormal Kupffer's vesicle development (Hatler et al. 2009)	Broadly expressed
gjd1a/Cx34.1	Loss of electrical synapses and disrupted startle response (Miller et al. 2017)	Nervous system and retina	GJD2/CX36	Epilepsy associated (Wang et al. 2017)
gjd1b/Cx34.7		Nervous system, retina, and muscle
gjd2a/Cx35.5	Myopia, loss of electrical synapses, and disrupted startle response (Miller et al. 2017; Quint et al. 2021)	Nervous system and retina
gjd2b/Cx35.1	Myopia (Quint et al. 2021)	Nervous system and retina
—	—	—	GJD3/CX31.9	No known implications
gjd4/Cx46.8	No known phenotype	Skeletal muscle	GJD4/CX40.1	No known implications
gjd5/Cx40.5	No known phenotype	Low expression in this dataset	—	—
gjd6/Cx36.7	Abnormal cardiac muscle tissue development (Sultana et al. 2008)	Heart	—	—
gje1a/Cx23.9	No known phenotype	Lens	GJE1/CX23	No known implications
gje1b/Cx20.3	No known phenotype	Skeletal muscle and nervous system
gjz1/Cx26.3	No known phenotype	Nervous system	—	—

The connexin gene family is broadly expressed, but spatially distinct

Next, we examined the spatiotemporal expression patterns of the zebrafish connexin genes through organogenesis using scRNA-seq. We used our recent scRNA-seq atlas dataset in which cells were dissociated from whole embryos at 1, 2, and 5 dpf, and resultant single-cell expression profiles were captured using the 10X platform (Farnsworth ). In our initial analysis of the data, we found that many of the connexin genes lacked expression information. An examination of the connexin gene models generated by Ensembl (GRCz11_93) that were used for mapping single-cell reads revealed that most annotations were truncated at or near the end of the protein-coding sequence, with most lacking 3′UTRs leading to a failure in capturing the 3′-biased 10X sequencing information (Supplementary Fig. 3). To amend this, we used a recently updated gene annotation file that extends gene models (Lawson ), evaluated and updated each connexin gene model in reference to bulk RNA-seq data (Miller ), and imported the Greek gene names. Using this updated gene annotation file, we processed the scRNA-seq data using Cellranger (Zheng ) and evaluated clustering and transcriptional profiles with Seurat (Satija ). Analysis of the updated scRNA-seq dataset captures transcriptional profiles that appear to represent all major tissues of the developing zebrafish (Fig. 1, ai and aii) and contains 49,367 cells and 238 clusters. This is 5,355 more cells and 18 more clusters than the original analysis (Farnsworth ), as expected due to the richer transcriptional information captured from the updated gene model (Lawson ). In our original analysis, we extensively annotated each cluster, assigning the most likely anatomical annotation based on comparing the differentially expressed genes for each cluster to RNA in situ patterns (Farnsworth ). We transferred these previous annotations to our updated analysis by identifying cell-specific barcodes from the original dataset, identifying them in the updated clusters, and transferring the cluster annotations (Fig. 1, a; Supplementary Tables 1 and 2). As a result, we identified all 220 original clusters (Farnsworth ) and annotated the remaining clusters by analyzing RNA in situ expression information for the most differentially expressed genes (Supplementary Table 3). The updated scRNA-seq dataset greatly improves the capture of connexin expression throughout the atlas (Supplementary Fig. 3), allowing us to examine their spatiotemporal expression pattern during zebrafish organogenesis.

Fig. 1.

scRNA-seq dataset of zebrafish organogenesis and connexin expression. ai) Clustered cell types, where each dot represents a single cell and each color represents a set of transcriptionally related cells. aii) The age of animals from which cells were dissociated denoted by color—1 dpf cells are blue, 2 dpf cells are orange, and 5 dpf cells yellow. bi–biii) Expression of well-studied connexins in the dataset, where gray represents low expression and red represents the highest level of expression. bi) gjc4b/Cx43.4 is expressed broadly across the dataset. bii) gja1b/Cx43 is expressed in a large number of clusters, with notable patterns in liver, endothelial, macrophage, neural crest, spleen, retina, kidney, epiphysis, osteoblast, mesoderm, tailbud, pigment cells and lens clusters. biii) gja8b/Cx44.1 is expressed in lens clusters. ci) Broadly expressed connexins, gja1b/Cx43 and gjc4b/Cx43.4 and (cii) the remaining connexin family shown for each sampled time point. Here, all cells from the corresponding age are pooled and the percent of cells expressing a given connexin are represented through dot size while the relative expression level is denoted through color intensity.

Using the updated scRNA-seq organogenesis dataset, we examined the expression of each connexin related to its clusters, its correlation with marker gene expression, and with cluster annotations (Supplementary Fig. 4, a–oo, Supplementary Table 4). Overall, connexin genes had a variety of expression patterns, varying from nearly ubiquitous to cluster-specific and showing a variety of temporal profiles, including constant expression over time or temporal specificity (Fig. 1, b and c, Supplementary Fig. 4, a–oo). To begin to evaluate the dataset’s utility, we first turned our attention to several well-studied connexin genes. First, gjc4b/Cx43.4 displayed the broadest expression, with particularly high levels in the nervous system, and with diminishing expression from 1 to 5 dpf (Fig. 1, b; Supplementary Fig. 5; Supplementary Table 4). This is similar to expression reports for gjc4b/Cx43.4 that used RNA in situ and transgenic methods (Thisse ; Baxendale ; Wierson ). gja1b/Cx43 is another well-described connexin, with broad expression in the cardiovascular system, non-neuronal cells of the retina and central nervous system, mesenchymal cells such as chondrocytes, and within the digestive system including the pancreas (Thisse and Thisse 2004; Chatterjee ; Iovine ; Hoptak-solga ; Yang ). We find that the expression of gja1b/Cx43 within the updated clusters largely matches these reported expression patterns (Fig.1 bii; Supplementary Fig. 4b; Supplementary Table 4). We also find expected patterns for connexins that have well-known, spatially restricted expression. For example, gja8b/Cx44.1 is expressed almost exclusively in the early developing lens(Cason ; Thisse and Thisse 2005; Yoshikawa ; Farnsworth ), and in the scRNA-seq dataset, we find expression of gja8b/Cx44.1 within clusters with transcriptional profiles consistent with lens cells (Fig. 1biii; Supplementary Fig. 5; Supplementary Table 4). Furthermore, we find gja2/Cx39.9 expression in presumptive skeletal muscle cells, gjd6/Cx36.7 specifically in presumptive cardiac muscle, and both gja9b/Cx52.9 and gja10b/Cx52.6 in presumptive horizontal cells, all well-matching published reports on the expression of these genes (Sultana ; Hirata ; Yoshikawa ; Greb ; Farnsworth ; Supplementary Fig. 5; Supplementary Table 4). Taken together, we conclude that the data represented in the updated dataset provide a useful resource for determining the spatiotemporal patterns of connexin expression during zebrafish organogenesis.

Connexins exhibit complex and combinatorial patterns of expression

To examine the relationship of connexin gene expression relative to one another, we organized the scRNA-seq clusters by their tissue annotations and plotted both expression levels and percentage of cells within each cluster (Fig. 2). When arranged in this fashion, the complexity of connexin expression within putative tissues and cell types is revealed. In particular, unique combinatorial patterns of connexins are observed within tissues developing from all germ layers. For example, within neural clusters (ectoderm), we find that there are 4 broadly expressed connexins, yet each displays bias to either the retina, gjd1b/Cx34.7 and gjd2b/Cx35.1, or central nervous system, gjd1a/Cx34.1 and gjd2a/Cx35.5 (Fig. 2; Supplementary Fig. 4, af–ai; Supplementary Table 4). Within the skeletal muscle clusters (mesoderm), a unique set of connexins are expressed and display a nested hierarchy of expression, with gja2/Cx39.9 in all skeletal muscle clusters, gja5a/Cx45.6 and gjd4/Cx46.8 restricted to slow muscle clusters, and gje1b/Cx20.3 restricted to fast muscle clusters (Fig. 2; Supplementary Fig. 4, c, jj, f, and nn). We also observed temporally complex patterns of expression. For example, within presumptive intestinal epithelial cells (endoderm), we find that gjc4b/Cx43.4 expression diminishes from 1 to 5 dpf, while gja13.1/Cx32.3 begins expression at 2 dpf and continues at 5 dpf and gja12.1/Cx28.9 becomes coexpressed at 5 dpf (Fig. 2; Supplementary Figs. 4, q and o and 6). Finally, we observed that primordial germ cells (PGCs) express several different connexins, including gja9a/Cx55.5, gjb8/Cx30.3, gjc4b/Cx43.4, and gjd1b/Cx34.7 (Fig. 2; Supplementary Figs. 4, w, j, gg, and ee and 7). These observations highlight aspects of the complexity of connexin spatial and temporal expression patterns within and across tissues and cell types during zebrafish organogenesis.

Fig. 2.

Connexin expression during zebrafish organogenesis. Clusters are organized by annotations and grouped into tissues and germ layers denoted on the y-axis. Along the x-axis, connexins are arranged based on spatial expression patterns. Each dot represents a single cluster. The percent of cells expressing a given connexin are represented through dot size while the relative expression level is denoted through color intensity. Diff. Neuron, differentiating neuron; Oligo, oligodendrocyte; Phar. Endoderm, pharyngeal endoderm; Arch, pharyngeal arch; PGC, primordial germ cell.

Cell-type-specific expression of connexins in the integument in vivo

To validate that the connexin expression identified in the updated atlas related to in vivo tissues and cell types, we examined the integument, or the embryonic skin, as it represented one of the most striking trends of combinatorial expression (Fig. 2). Throughout zebrafish organogenesis, the integument is composed of distinct cellular populations including the periderm (the outermost epidermal layer), the basal cells (a keratinocyte stem cell population), the ionocytes (epithelial cells that maintain osmotic homeostasis), and the pigment cells (neural crest-derived cells that provide pigmentation; Guellec ; Eisenhoffer ). These individual cell populations are molecularly identifiable using distinct markers including ppl (periderm; Thisse ; Thisse and Thisse 2004), tp63 (basal cells; Lee and Kimelman 2002), foxi3a (ionocytes; Jänicke ), and sox10 (pigment cells; Budi ; Eisenhoffer ; Fig. 3; Supplementary Fig. 8). We used these canonical markers in conjunction with our annotations (Supplementary Fig. 8; Supplementary Table 2) to identify clusters that represent all 4 cell types of the integument (Fig. 3ai). We identified all connexins that are significantly expressed within these presumptive integument clusters (Fig. 3; Supplementary Fig. 8). We found that gjb3/Cx35.4, gjb8/Cx30.3, gjb10/Cx34.4, and gjc4b/Cx43.4 are expressed broadly across these clusters (Fig. 3aii). We then looked for connexins enriched in subsets of clusters and found unique and specific patterns of expression. Within periderm clusters, we discovered gjb9a/Cx28.6, which has not previously been documented in the skin (Fig. 3aiii). Within the presumptive neural crest-derived pigment clusters, we found gja4/Cx39.4 and gja5b/Cx41.8, which are both known to contribute to adult zebrafish skin patterns (Watanabe ; Watanabe 2017; Fig. 3a). Within ionocyte clusters, we identified novel expression for 2 connexins, gjb7/Cx28.8 and gjb9b/Cx30.9 (Fig. 3av). Finally, within presumptive basal cell clusters, we found novel expression for 2 connexins, gjc4a.1/Cx44.2 and gjc4a.2/Cx44.5 (Fig. 3avi). These results suggest that the integument uses a complex set of connexins throughout organogenesis.

Fig. 3.

Connexin expression in the zebrafish integument during organogenesis. ai) The developing integument includes periderm, pigment cells, ionocytes, and basal cells. Relevant integument clusters were subsetted from the scRNA-seq dataset. Inset shows the age of animals from which cells were dissociated. aii) Four connexins are broadly expressed in integument clusters, gjb3/Cx35.4, gjb8/Cx30.3, gjb10/Cx34.4, and gjc4b/Cx43.4. Gray represents low expression and red represents the highest level of expression. aiii) Periderm marker ppl and gjb9a/Cx28.6 are expressed in clusters 40–46. aiv) Neural crest-derived pigment cell marker sox10 and gja4/Cx39.4 are expressed in clusters 52–57, while gja5b/Cx41.8 is only expressed in clusters 54 and 56. av) Ionocyte marker foxi3a and gjb7/Cx28.8 are expressed in clusters 38, 39, 47, and 48. avi) Basal cell marker tp63 and gjc4a.1/Cx44.2 are expressed in clusters 23, 25–32, 51, 222–224, while gjc4a.2/Cx44.5 is only expressed in clusters 25–29. bi) Fluorescent RNA in situ for gjb8/Cx30.3 in a transverse cross-section of a 1 dpf zebrafish embryo, contrast is inverted for clarity. Dorsal is up, section is from the trunk. Strong expression of gjb8/Cx30.3 in neural crest cells is denoted with arrow and weaker, but distinct, periderm expression is denoted with arrowhead. bii) Within the pigment cell clusters the melanocyte marker dct is expressed in clusters 56 and 57, whereas xanthophore marker aox5 is primarily expressed in clusters 52 and 53. gjb8/Cx30.3 is predominantly expressed in clusters 52 and 53. biii) Transverse cross-section of a 1 dpf zebrafish embryo stained with DAPI (blue) and fluorescent RNA in situ against aox5 (cyan) and gjb8/Cx30.3 (white), with white box denoting the zoomed panels at the right. Scale bar = 10 µM. biv) Expression of ppl and gjb8/Cx30.3 within the periderm clusters. bv) Transverse cross-section of a 1 dpf zebrafish embryo stained with DAPI (blue) and fluorescent RNA in situ against ppl (purple) and gjb8/Cx30.3 (white) with white box denoting the zoomed panels at the right. ci) Within the ionocyte clusters the Na+,K+-ATPase-rich cell and H+-ATPase-rich cell markers atp1b1b and atp6v1aa, respectively, are expressed in conjunction with low expression of gjb7/Cx28.8. cii) Fluorescent RNA in situ in a 1 dpf zebrafish embryo against atp1b1b (yellow), gjb7/Cx28.8 (white), with merged signal (right). atp1b1b expressing cells are outlined with a dashed yellow line, and gjb7/Cx28.8 signal outside of those cells are marked with yellow arrowhead. Scale bar = 10 µM. ciii) Fluorescent RNA in situ in a 1 dpf zebrafish embryo against atp6v1aa (green), gjb7/Cx28.8 (white), with merged signal (right). atp6v1aa expressing cells are outlined with a dashed yellow line, and gjb7/Cx28.8 signal outside of those cells are marked with yellow arrowhead. Scale bar = 10 µM.

We next examined a subset of the identified integument connexins in vivo. We first tested a broadly expressed connexin, gjb8/Cx30.3, to see if it was expressed in the pigment cells and periderm using fluorescent RNA in situ on 1 dpf embryos. Transverse cross-sections through the trunk revealed prominent gjb8/Cx30.3 staining in dorsally located cells near the neural tube and additional dim staining was observed in a single layer of cells surrounding the entire embryo (Fig. 3bi). We first confirmed gjb8/Cx30.3’s expression in pigment cells by subsetting the 5 clusters that appear to represent pigment cells, including melanophores (Kelsh ; Parichy ; dct+, clusters 56, 57, Fig. 3bii; Supplementary Table 2) and xanthophores (Parichy ; aox5+, clusters 52, 53, Fig. 3bii; Supplementary Table 2). We find that gjb8/Cx30.3 is highly expressed in only the presumptive xanthophore clusters (Fig. 3bii). We then performed fluorescent RNA in situ for aox5 and gjb8/Cx30.3 in a 1 dpf embryo and found robust colocalization of these transcripts, confirming that gjb8/Cx30.3 is expressed in xanthophore cells (Fig. 3biii). We then examined gjb8/Cx30.3’s expression in the periderm through subsetting the 7 presumptive periderm clusters (ppl+, clusters 40–46, Supplementary Table 2) and find expression of gjb8/Cx30.3 in all clusters (Fig. 3biv). Indeed, fluorescent RNA in situ for ppl and gjb8/Cx30.3 reveal robust colocalization of these transcripts in the outermost epithelial layer (Fig. 3bv), confirming that gjb8/Cx30.3 is expressed in the developing periderm.

We next tested a connexin with more specific expression within the integument clusters, gjb7/Cx28.8, which has expression specific to the presumptive ionocytes (Fig. 3av). Developing foxi3a+ ionocytes form Na+,K+-ATPase-rich (NaR) cells or H+-ATPase-rich (HR) cells, which are characterized by the expression of the specific ATPase genes atp1b1b and atp6v1aa, respectively (Jänicke ). First, we subsetted all ionocyte clusters (Jänicke ; foxi3a+, 38, 39, 47, 48) and found unique expression combinations of atp1b1b and atp6v1aa across clusters and low expression of gjb7/Cx28.8 in 3 of 4 clusters (Fig. 3ci). Fluorescent RNA in situ revealed colocalization of gjb7/Cx28.8 with both atp1b1b (Fig. 3cii) and atp6v1aa (Fig. 3ciii), confirming that gjb7/Cx28.8 is expressed in ionocytes. Together, these data confirm the predictive power of the scRNA-seq dataset for connexin expression in the integument and support the utility of the dataset as a novel tool for the discovery of investigating connexin complexity in vertebrate development.

Discussion

Here, we reveal the details of connexin gene-family expression during zebrafish organogenesis showing that connexin usage is widespread yet displays gene-specific variations across tissue, cell type, and developmental time. The large gene family of connexins in zebrafish (41 genes) is expressed in complex patterns ranging from nearly ubiquitous to cell-type specific, with unique combinatorial and nested expression sets restricted to individual tissues. Temporally, connexins display sustained, increasing, and diminishing expression profiles across development, dependent upon gene and tissue. Together, these data reveal the complexity of expression of this critical gene family in a model vertebrate and demonstrate that this critical form of communication is likely to be used by all tissues during organogenesis. These data provide a critical framework facilitating analysis of how these genes contribute to cellular communication in tissues developing from all germ layers, providing a basis to understand connexins in development and in modeling human disease.

We find that all cells express connexins, but each tissue expresses a unique combination of the gene family with the composition of the expressed set evolves over developmental time. This spatiotemporal complexity of connexin family usage likely contributes to both functional redundancy within tissues as well as functional diversity. The many connexins expressed might allow for a myriad of combinatorial interactions amongst Connexin proteins, which could contribute to heteromeric hemichannels and heterotypic GJs. Importantly, Connexins can only interact with potential partners if they are expressed in the same cell or between interacting cells, thus the work here constrains the combinatorial problem of complex usage by revealing the details of the expression patterns through organogenesis. For example, gjd2a/Cx35.5 and gjd1a/Cx34.1 have been shown to form heterotypic GJs (unique Connexins on each side of the GJ) at electrical synapses of the Mauthner cell neural circuit (Miller ). The data here show extensive overlapping expression of these 2 connexins throughout the central nervous system, suggesting complex hemichannels and GJs could be common throughout the brain. Given that each Connexin-mediated hemichannel has its own unique set of compatibilities and permeability properties, this dataset provides a platform for future research to explore whether connexins expressed within the same tissue or cell type form functional channels, and how the molecular identity of these channels influences function.

This dataset presents a powerful resource for zebrafish and connexin biology. We establish connexin expression in cells previously unknown to express connexins, such as the ionocytes of the skin. Within our dataset, there are numerous other cell types with striking connexin expression patterns that have under-appreciated connexin usage inviting exploration, including macrophages (gja13.2/Cx32.2) and PGCs (gja9a/Cx55.5, gjb8/Cx30.3, gjc4b/Cx43.4, and gjd1b/Cx34.7). Another strength of this dataset is the exploration of expression across multiple cell types, tissues, and timepoints simultaneously. For example, gja3/Cx46 has only been examined in the heart (Chi , 2010), yet, in our dataset, we find robust gja3/Cx46 expression in both heart and lens clusters, which suggests an enticing link to human GJA3/CX46, in which mutations are associated with cataracts (Mackay ; Burdon ; Yao ). Finally, this dataset provides putative expression to many connexin genes that had no previous expression information (22/41 genes). For example, gjb1a/Cx27.5 and gjc2/Cx47.1 are both highly expressed in the Schwann cell cluster. While neither of these genes had previously known expression information, mutations of their human orthologs GJB1/CX32 and GJC2/CX47 contribute to neuropathy and myelin disorders (López-Bigas ; Uhlenberg ; Orthmann-Murphy ). The identification of tissues and cell-type expression patterns for the entire gene family creates a basis to explore connexin-related diseases in zebrafish and provide comparisons to human biology. Through exploring the connexin family expression across diverse cell types and tissues, we can begin to envision a holistic view of Connexins utilization and usage in cellular communication throughout organogenesis.

Data availability

All data generated or analyzed during this study are included in the published article and its supplementary information files. Sequences used in this study were deposited to the NCBI SRA and can be found using the identifier PRJNA564810. Additional files, including the updated GTF, analysis, and code, can be found at https://www.adammillerlab.com/.

Supplemental material is available at G3 online.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.

Click here for additional data file.