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

Abstract

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 well appreciated 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 probed 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 towards understanding how Connexins contribute to cellular communication and development throughout vertebrate organogenesis.

Introduction

Animal development and homeostasis requires coordinated cellular communication. One method mediating communication are gap junction (GJ) channels. GJs are intercellular channels that provide a direct path of low resistance for ionic and small molecule exchange between cells¹. These channels are formed by the coupling of two apposed hemi-channels each contributed by adjacent communicating cells^1–3. Additionally, hemi-channels 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^2,3. The proteins that create GJ channels are evolutionarily unrelated in vertebrates and invertebrates⁴. Yet despite little sequence similarity⁵, the vertebrate Connexins and the invertebrate Innexin proteins have similar structure, with both classes creating four-pass, transmembrane-domain proteins that oligomerize to form each hemichannel within the plasma membrane⁴. Moreover, the hemi-channels and intercellular GJs created by Connexins and Innexins have similar structure and function⁴. Outside of these traditional roles, Connexins can also modulate the formation of tunneling nanotubes, that connect non-adjacent cells to facilitate longer distance communication^6–8. These varied functions in cellular communication are likely utilized individually and in combination in all animal tissues⁹, yet are best studied in epithelial¹⁰, neural¹¹, and vascular¹² systems. In these systems, mutations in human Connexin-encoding genes have been linked to defects in the development, regulation, and function including skin disorders^13–16, cataracts^17,18, deafness^19–21, cardiovascular disease^22–24, and gastrointestinal diseases^25–27. 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^28–30. Cell culture and in vitro work suggests that connexin complexity provides functional diversity governed by four general principles: first, hemichannels are created by hexamers of individual Connexin proteins³¹; second, single or multiple Connexin proteins can contribute to hemichannel formation (homo- or heteromeric hemichannels, respectively)^32–34; third, gap junctions 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)^32–34. The combinatorial possibilities of the gene family are restrained by molecular engagement rules that limit which Connexins are compatible to form mixed channels^34–37. 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^37,38. These rules suggest 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 epithelia³⁹, heart^24,40, and glia⁴¹. By contrast, Gjd2/CX36 is found almost exclusively in neurons⁴². 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 points⁴³. 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 development⁴⁴. Heterotypic gap junctions 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^45,46. 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 cannot be simply interchangeable with other Connexins ⁴⁷. While these examples provide a glimpse of functional complexity, understanding the expression of this gene family through vertebrate development remains as 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 post-fertilization (dpf))⁴⁸. 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 towards understanding connexin gene contributions in cellular communication throughout organogenesis and provides a foundation for comparative analysis in vertebrates.

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^28,49 and a recent phylogenetic approach to identify the full teleost connexin family³⁰ 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 (Supp Fig. 1, 2).

Across the family of connexin genes there are seven human connexins for which zebrafish only has a single homolog, eight human connexins for which zebrafish has two homologs, two zebrafish connexins that have no direct homolog but share sequence similarity to human connexins, and sixteen zebrafish connexins that are not present in humans but are conserved in other teleost and mammalian lineages³⁰. 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 HUGO⁵⁰ and ZFIN standards⁵¹ (Table 1, Supp. 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 gjb8³⁰. 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^20,52,53. 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.

The connexin gene family is broadly expressed, but spatially distinct

Next, we examined the spatio-temporal 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 days post fertilization (dpf), and resultant single-cell expression profiles were captured using the 10X platform⁴⁸. 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 (Supp. Fig. 3). To amend this, we used a recently updated gene annotation file that extends gene models¹⁰⁰, evaluated and updated each connexin gene model in reference to bulk RNA-seq data¹⁰¹, and imported the Greek gene names. Using this updated gene annotation file, we processed the scRNA-seq data using Cellranger¹⁰² and evaluated clustering and transcriptional profiles with Seurat¹⁰³. Analysis of the updated scRNA-seq dataset captures transcriptional profiles that appear to represent all major tissues of the developing zebrafish (Fig. 1Ai, Aii) and contains 49,367 cells and 238 clusters. This is 5,355 more cells and 18 more clusters than the original analysis⁴⁸, as expected due to the richer transcriptional information captured from the updated gene model¹⁰⁰. 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 patterns⁴⁸. 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. 1Ai, Aii; Supp. Table 1,2). As a result, we identified all 220 original clusters⁴⁸ and annotated the remaining clusters by analyzing RNA in situ expression information for the most differentially expressed genes (Supp. Table 3). The updated scRNA-seq dataset greatly improves the capture of connexin expression throughout the atlas (Supp. Fig. 3), allowing us to examine their spatio-temporal expression pattern during zebrafish organogenesis.

Figure 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 days post fertilization (dpf) cells are blue, 2dpf cells are orange, and 5dpf cells yellow. (Bi-Biii) Expression of well-studied connexins in the dataset, where grey 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 cooresponding age are pooled and the percent of cells expressing a given connnexin 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 (Supp. Fig. 4A-OO, Supp. 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. 1B,C, Supp. 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. 1Bi, Ci; Supp. Fig. 5; Supp. Table 4). This is similar to expression reports for gjc4b/Cx43.4 that used RNA in situ and transgenic methods^104–106. 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^62,107–110. We find that expression of gja1b/Cx43 within the updated clusters largely matches these reported expression patterns (Fig.1 Bii; Supp. Fig.4B; Supp. 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^111–114, and in the scRNA-seq dataset we find expression of gja8b/Cx44.1 within clusters with transcriptional profiles consistent with lens cells (Fig. 1Biii; Supp. Figure 5; Supp. Table 4). Further, 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^{64,99,111,114,115} (Supp. Figure 5; Supp. Table 4). Taken together, we conclude that the data represented in the updated dataset provides a useful resource for determining the spatio-temporal 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 four 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; Supp. Figure 4 AF-AI; Supp. Table 4). Within the 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 are restricted to slow muscle clusters, and gje1b/Cx20.3 is restricted to fast muscle clusters (Fig. 2; Supp Fig. 4C, JJ, F, 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 co-expressed at 5 dpf (Fig. 2; Supp. Fig. 4Q, O, Supp. Fig. 6). Finally, we observed that primordial germ cells (PGC) express several different connexins, including gja9a/Cx55.5, gjb8/Cx30.3, gjc4b/Cx43.4, and gjd1b/Cx34.7 (Fig. 2; Supp. Figure 4W, J, GG, EE, Supp. Fig. 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.

Figure 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 connnexin 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)^116,117. These individual cell populations are molecularly identifiable using distinct markers including ppl (periderm)^104,109, tp63 (basal cells)¹¹⁸, foxi3a (ionocytes)¹¹⁹, and sox10 (pigment cells)^116,120 (Fig. 3; Supp. Fig. 8). We used these canonical markers in conjunction with our annotations (Supp. Fig. 8; Supp. Table 2) to identify clusters that represent all four cell types of the integument (Fig. 3Ai). We identified all connexins that are significantly expressed within these presumptive integument clusters (Fig. 3; Supp. 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^49,71 (Fig. 3Aiv). Within ionocyte clusters, we identified novel expression for two connexins, gjb7/Cx28.8 and gjb9b/Cx30.9 (Fig. 3Av). Finally, within presumptive basal cell clusters, we found novel expression for two connexins, gjc4a.1/Cx44.2 and gjc4a.2/Cx44.5 (Fig. 3Avi). These results suggest the integument uses a complex set of connexins throughout organogenesis.

Figure 3:

connexin expression in the zebrafish integument during organogenesis. (Ai) The developing integument includes periderm, pigment cells, ionocytes and basal cells. Relevent 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. Grey 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, 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 primarly 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 uM. (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 1dpf 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 arowhead. Scale bar = 10 uM. (Ciii) Fluorescent RNA in situ in a 1dpf 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 arowhead. Scale bar = 10 uM.

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 five clusters that appear to represent pigment cells, including melanophores^121,122 (dct+, clusters 56, 57, Fig. 3 Bii; Supp. Table 2) and xanthophores¹²³ (aox5+, clusters 52, 53, Fig. 3 Bii; Supp. Table 2). We find that gjb8/Cx30.3 is highly expressed in only the presumptive xanthophore clusters (Fig. 3 Bii). We then performed fluorescent RNA in situ for aox5 and gjb8/Cx30.3 in a 1 dpf embryo and found robust co-localization 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 sub-setting the seven presumptive periderm clusters (ppl+, clusters 40-46, Supp. Table 2) and find expression of gjb8/Cx30.3 in all clusters (Fig. 3 Biv). Indeed, fluorescent RNA in situ for ppl and gjb8/Cx30.3 reveal robust co-localization 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, respectively¹¹⁹. First, we sub-setted all ionocyte clusters¹¹⁹ (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 co-localization 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 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 celltype 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 combinations 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 circuit⁴⁵. The data here shows extensive overlapping expression of these two connexins throughout the central nervous system, suggesting complex hemi-channels 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 as 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 primordial germ cells (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 ^65,66, 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^67–69. 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^86,93,94. 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.

Methods

scRNA-seq

Embryo dissociation and cDNA library prep

As described by Farnsworth et al., 2020⁴⁸, 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, 5dpf). Cells from entire larvae were dissociated using standard protocols. 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 Lab¹⁰⁰. 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_93, using the 10X Cellranger pipeline (version 3.1).

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 PCA using 115 PCs based on a Jack Straw-determined significance of P < 0.01. UMAP analysis was performed on the resulting 49,367 cells with 115 PC dimensions and a resolution of 15.0, which produced 238 clusters.

Cluster annotation

The unique barcode assigned to each cell was extracted from the original Farnsworth dataset⁴⁸ 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 two of the updated clusters – for example, the original dataset had a single ‘photoreceptor’ cluster (cluster 115), whereas the updated data had two clusters (clusters 13, 14) with cells from original photoreceptor cluster. Further analysis revealed that these two 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 the fluorescent in situs, we used a modified RNAscope protocol¹²⁴. Briefly, 1 dpf embryos were fixed for 2 hours at room temp in 4% PFA and then stored in 100% methanol at −20C. The tissue was then exposed to protease plus for 30 min, washed with PBS with 1% Triton X (PBSTx), and then hybridized with the 1x probe overnight at 40C. Standard RNAscope V2 multiplex reagents and Opal fluorophores were used, with the modification that PBSTx which was used for all wash steps. Stained tissue was either mounted (whole mount) or immediately cryo-sectioned and mounted with ProLong Gold Antifade (ThermoFisher).

Zebrafish Husbandry

Fish were maintained by the University of Oregon Zebrafish Facility using standard husbandry techniques¹²⁵. 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)⁴⁸, 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.

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 and analysis, can be found at https://www.adammillerlab.com/.

Supplemental Figure Legends

Supplemental Figure 1: Phylogeny of human (hs) and zebrafish (dr) Connexin proteins.

Supplemental Figure 2: Protein similarities and phylogeny of the novel gjz1/Cx26.3. (A) Zebrafish (dr.) Gjz1/Cx26.3 aligned with the most similar human (hs.) Connexin protein, GJB3. The predicted intercellular portions (coral), transmembrane domains (green) and extracellular loops (blue) for GJB3 are denoted above the sequence. (B) Ensembl-generated phylogenetic tree of gjz1/Cx26.3 (si:rp71-1c10.10 in red). Bony fishes are highlighted in pink, while lobed-fin lineages from coelacanth to mammals are in other colors. Protacanthopterygii and Euacanthomorphacea sub-trees are collapsed for visual purposes (grey triangle). Each related gene is represented on a node and colored based on Ensembl predications, including a gene node (white box), speciation nodes (dark blue box), duplication nodes (red box), and ambiguous nodes (teal box). Genomic sequence alignment similarity is denoted on the right with black boxes (representing 66-100% alignment), moderate sequence alignment is denoted on the right with green boxes (representing 33-66% sequence alignment) and gaps in alignments are denoted in white.

Supplemental Figure 3: Gene model updates to capture connexin expression. (A) Aligned sequencing view showing coverage pileups for scRNA-seq (1, 2, 5 dpf) and bulk RNA-seq, as well as bulk RNA-seq reads (grey boxes). The Ensembl GTF (light green) captures few reads associated with gjd4/Cx46.8. Extending the gene model (dark green) in the updated GTF captures these missed transcripts. (B) Comparisons between the Farnsworth et al., 2020 dataset and the updated dataset. (Bi) Farnsworth cluster 57 has robust expression of slow muscle marker smyhc2, but poor expression of gjd4/Cx46.8. In the updated dataset, the related cluster 4 has similar expression of slow muscle marker smyhc2 but a significant increase of gjd4/Cx46.8 representation (right). (Bii) Farnsworth cluster 205 has robust expression of cardiac muscle marker myl17 and gjd6/Cx36.7. The corresponding cluster in the updated dataset did not significantly alter either of these expression patterns (right). (Biii) Farnsworth cluster 173 has low expression of Schwann Cell marker, mbpa, as well as gjb1a/Cx27.5. The corresponding cluster in the updated dataset captures robust expression for both mbpa and gjb1a/Cx27.5.

Supplemental Figure 4: connexin expression throughout the atlas. Note that this figure extends across 41 pages, one for each connexin gene, labeled A-OO. Expression of each connexin is plotted on the scRNAseq atlas and visualized through color intensity on UMAP plots and by violin plots for each gene and cluster.

Supplemental Figure 5: Tissue and cell type markers with corresponding connexin expression. (Ai) Central nervous system clusters express snap25a and pcna, and gjc4b/Cx43.4. (Aii) Lens clusters (193, 112 and 237) express lens markers crybb1 and crybab2, and gja8b/Cx44.1. (Aiii) Skeletal muscle clusters express slow muscle marker smyhc1 (4, 5, 234), fast muscle marker myhz2 (7, 8, 9, 10, 11, 233), and gja2/Cx39.9. (Aiv) Cardiac muscle cluster 233 expresses markers nppa and myl17, and gjd6/Cx36.7. (Av) Retinal horizonal neurons, cluster 203, express marker mdka and gja9b/Cx52.9 and gja10b/Cx52.6.

Supplemental Figure 6: Temporal expression patterns of connexins within the intestine. (Ai) Intestinal clusters (3, 60, 67, 140) selected for expression of canonical markers cldnc, fabp2, and foxa3. (Aii) These clusters display an evolution of connexin expression at different developmental time points. gjc4b/Cx43.4 (left) is expressed at 1 and 2 dpf, gja13.1/Cx23.3 is expressed at 2 and 5 dpf, and gja12.1/Cx28.9 is expressed at 5 dpf.

Supplemental Figure 7: Primordial germ cells express several connexins. Putative primordial germ cell (PGC) cluster (128) expressing PGC markers like ddx4 and nanos3, in addition to connexins including gja9a/Cx55.5, gjb8/Cx30.3, gjc4b/Cx43.4, and gjd1b/Cx34.7.

Supplemental Figure 8: connexin expression in the integument. Clusters classified into cell types are grouped and labelled accordingly. (Ai) gjb3/Cx35.4, gjb8/Cx30.3, gjb10/Cx34.4, and gjc4b/Cx43.4 are all expressed broadly through the integument clusters. (Aii) ppl, krt4, and evpla are expressed highly in periderm clusters, with gjb9a/Cx28.6 being primarily expressed in periderm clusters. (Aiii) sox10 and aox5 expression identifying pigment cell clusters with gja4/Cx39.4 and gja5b/Cx41.8. (Aiv) foxi3a expression identifying ionocyte clusters with gjb7/Cx28.8 and gjb9b/Cx30.9. (Av) tp63 expression identifying basal cell clusters with gjc4a.1/Cx44.2 and gjc4a.2/Cx44.5.

Supplemental Table 1: The zebrafish connexin family.

Supplemental Table 2: Transferring Farnsworth et. al, 2020 cluster annotations to the updated dataset. Updated cluster number (Column A) and the corresponding Farnsworth cluster (Column B). The count of cells from the Farnsworth cluster that ended in the corresponding updated cluster (Column C), and the proportion of cells from a given Farnsworth cluster that contribute to the updated cluster (Column D, E). Total cell counts are colored in blue and bolded. All previous Farnsworth annotations were transferred over (columns H-AE), and update cluster markers are included in orange. The most significant Farnsworth cluster contributor is denoted in black font.

Supplemental Table 3: List of differentially expressed genes for each cluster. (Sheet1) For each cluster (Column A), annotations at germ layer, tissue, cell type, and subtype (Columns B – E) level are listed. For each cluster, the top 16 most differentially expressed genes are listed (Column F - U). (Sheet 2) All differentially expressed genes for each updated cluster, generated using the FindAllMarkers command of Seurat, using the Wilcoxon rank sum test. Pct.1 (Column D) and Pct.2 (Column E) reflect the fraction of cells expressing each marker gene (Boolean) within each cluster and for all other cells, respectively.

Supplemental Table 4: The proportion of cells within each cluster that express each connexin.

Supplemental Table 5: All reagents used for fluorescent RNA in-situ and other immunohistochemistry in this study.

Acknowledgments

We thank the entire Miller Lab ongoing support, comments, and discussions on this manuscript. We thank Clay Small for discussions and expertise in regards to data handling, statistics, and annotation transfer from the original to updated atlas. We thank the University of Oregon AqACS facility for superb animal care. We thank the ZFIN team, and Svein-Ole Mikalsen, for communication regarding the connexin gene names. This work was supported by the NIH National Institute of General Medical Sciences, Genetics Training Grant T32GM007413 to R.M.L., and the NIH Office of the Director R24OD026591 and the NIH National Institute of Neurological Disorders and Stroke R01NS105758 to A.C.M.