Genetic interrogation reveals mechanisms regulating nexus glia development and shaping heart physiology

SUMMARY

Functional organs require precise cellular organization to engage in bidirectional communication with the nervous system. However, the cellular specialization underlying this vital communication remains incompletely understood. Here, we investigated zebrafish cardiac nexus glia and found they are post-mitotic and exhibit dynamic Ca²⁺ transients that mature throughout development and coordinate with cardiomyocyte activity. We demonstrate that mechanical cues are dispensable for nexus glia abundance. Using a large-scale CRISPR screen of genes identified from scRNA-sequencing analysis of embryonic human cardiac cells, we identify nexus glia genetic modifiers. Specifically, perturbation of cdh6 expands the nexus glia population, enhances glial calcium activity, and disrupts cardiomyocyte calcium handling. Crucially, this glial overexpansion protects against cardiac exhaustion via a potassium channel-dependent mechanism. These findings reveal the genetic architecture of neuro-cardiac communication and highlight the role of nexus glia in cardiac resilience.

Graphical Abstract

eTOC Blurb

Light et. al. reveal that cardiac nexus glia, specialized glia in the heart, exhibit dynamic calcium transients that coordinate with cardiomyocytes. The nexus glia population can be expanded, via genetic modifiers, to protect against cardiac exhaustion. These findings reveal cellular specialization underlying vital communication between organs and the nervous system.

INTRODUCTION

Organs need to bidirectionally communicate with the nervous system to respond to physiological demands^1–5. This coordinated regulation requires specialized neural cell types tailored to the unique functional demands of their resident organ. In recent years, glial subtypes have been identified in the heart, spleen, pancreas, kidney, and lungs^6–12. These glia express gfap, a transcript found in glia throughout the central nervous system and plp1, a marker of peripheral glia. Mature glia in these organs also express S100B, another central glial marker^12–14. Further identifying molecular modulators of organ glia could help to elucidate if gfap⁺ cells in other organs can also be reliably identified as glia. While their functional roles are not fully understood, glia in the heart and intestines are important in organ physiology^12,15–19. Robust cardiomyocyte function in the heart depends on precise control of ion kinetics and dysregulation can lead to deadly outcomes like cardiac exhaustion and arrest^20,21. In the brain, similar ion kinetics are mediated by astrocytes²². If astrocyte-like cells in the periphery, such as nexus glia, perform analogous functions, then elucidating the development and function of these organ-resident glia, including their mechanical, electrical, and genetic factors, is critical.

Here, we interrogate cardiac nexus glia, an astroglia-like cell in the heart¹². We used genetically-encoded calcium indicators to reveal dynamic Ca²⁺ transients in nexus glia that correspond with cardiomyocytes. We then identified candidate modifiers from human scRNA-sequencing datasets²³ and performed a high-throughput CRISPR screen. We found cdh6 negatively modulated nexus glia abundance, impacts Ca²⁺ transients, and protects against cardiac failure induced by sympathetic exhaustion. This protective role for nexus glia is dependent on potassium channels. This work is a critical step in uncovering the genetic mechanisms directing organ-resident glia and places a link between expansion of nexus glia and protection from sympathetic exhaustion that could shed light on our understanding of cardiac arrest.

RESULTS

Zebrafish as a model of nexus glia development

To investigate organ-glia, we developed new reagents to label glia within intact tissue, focusing on nexus glia in the heart¹². We utilized zebrafish where nexus glia could be visualized in living, intact tissue. Stable Tg(gfap:nlsGFP-P2A-tdTomato-CAAX) animals labeled both nuclei and cellular membranes, showing astroglia throughout the nervous system, including the heart (Figure S1A–S1C, Table S1). The membrane labeling by tdTomato was present, but dim, in the heart so in most cases, gfap⁺ cells were identified by nuclear GFP (Figure S1D). in situ hybridized chain reaction (HCR) against gfap mirrored the transgenic expression, labeling in the outflow tract and ventricle and sporadically in the atrium (Figure 1G). This is consistent with localization of neural crest cells and their derivative cell populations in the heart^12,24. In contrast, the cardiomyocyte marker, myl7, labeled throughout the heart. A similar labeling pattern in the heart was observed for sox10. Both probes also displayed labeling in the CNS, as expected (Figure S1E).

Figure 1. Zebrafish as a model of nexus glia development.

(A) Average heart rate of AB animals at 2, 3, 4, 5, and 6 dpf; n=24 animals. One-way ANOVA was used. (B) Heart rate quantification of 5 dpf animals treated with BDM compared to untreated controls; n=18 animals. Dashed vertical line indicates BDM washout. Shaded area around each line indicates SEM. Arrow indicates average time to cease contractions. Unpaired t-test to compare groups at each time point was used. (C) Schematic of heart chambers at 2 dpf and 3–6 dpf shown from ventral and lateral views. (D) Max projections of 2–6 dpf Tg(gfap:nlsGFP-P2A-tdTomato-CAAX) hearts with atrium (pink), ventricle (green), and outflow tract (blue) outlined in each panel. (E) Quantification of the total number of gfap⁺ cells in 2–6 dpf hearts; n=40–66 animals. One-way ANOVA was used. (F) Quantification of gfap⁺ cell distribution in 2–6 dpf hearts; n=40–66 animals. (G) Max projections of 4dpf hearts labeled with HCR probes against gfap, sox10, and myl7. Atrium (pink), ventricle (green), and outflow tract (blue) are outlined in each panel. (H) Max projections of 4 dpf Tg(gfap:nlsGFP-P2A-tdTomato-CAAX) hearts injected with myl7:mCherry-NLS. Ventricle is outlined in each panel. (I) Quantification of total number of gfap⁺, myl7⁺, and myl7⁺gfap⁺ cells at 4 dpf; n=13 animals. (J) Max projections of Tg(sox10:Gal4+myl7:GFP);(UAS:Lifeact-GFP) heart at 2, 3, 4, and 5 dpf injected with gfap:Gal4-P2A-mCherry-NLS. Ventricle is outlined in each panel. (K) Quantification of domain size of individual cells tracked between 3 and 5 dpf; n=9 cells in 8 animals. (L) Max projections of Tg(sox10:Gal4+myl7:GFP) hearts injected with gfap:Lifeact-mCherry-P2A-nlsGFP. Error bars represent mean +/− SEM. Scale bar for D, G, H, K, and L = 50 μm. Related to Figure S1.

The heart’s constant mechanical movement throughout larval development complicates the analysis of nexus glia²⁵ (Figure 1A). To visualize nexus glia in intact tissue, we halted heart contractions during imaging by acutely treating animals with 2,3-butanedione monoxime (BDM), a myosin inhibitor^26,27. BDM treatment reduced the heart rate to zero within 90 minutes (Figure 1B). Exposing mounted animals to the drug for 45 minutes (Figure S1F) allowed us to consistently capture high-quality images of nexus glia in vivo (Figure 1C, Figure S1G). During BDM treatment, animals did not show any deleterious effects, and removal of BDM restored heart rates to pre-treatment levels within 30 minutes (Figure 1B, Figure S1H–S1I).

The Tg(gfap:nlsGFP-P2A-tdTomato-CAAX) line clearly labels glia in the outflow tract, ventricle, and atrium of the heart (Figure 1D). This gfap⁺ population expands robustly between 2 dpf and 4 dpf. At 5 dpf, the number of gfap⁺ cells in the heart declined slightly and then leveled out at 6 dpf to an average of 39.23±1.807 cells (Figure 1E). The majority were found in the ventricle from 2–6 dpf (Figure 1F). The observed abundance of nexus glia was higher than previously reported¹². Consistent with previous published transgenics, we observed that most gfap⁺ cells in our new transgenic line are not labeled by a construct driven by the well-established cardiomyocyte marker, myl7 (Figure 1H–1I).

We next investigated whether the gfap⁺ cells in the heart represented a mature and stable cell population. To do this, we developed a mosaic labeling strategy to measure the domain size of individual nexus glia. By imaging on consecutive days, we observed that individual gfap⁺ cells were stable over time, did not move from their original observed position within the tissue, and had relatively consistent domain sizes (Figure 1J–1K, Figure S1J–S1L). In instances where two or more labeled cells overlapped, the collective domain size of these multi-cell groups had a quantal increase in occupied area (Figure S1M). Nexus glia exhibited a bushy-like morphology that resembles astrocytes in the CNS (Figure 1L). Quantifying nexus glia and the total number of nuclei in the heart labeled by DAPI demonstrated that they are a relatively small subset of the total cells in the heart at 4 dpf (Figure S1N). These data collectively suggest that gfap⁺ nexus glia are an anatomically and morphologically stable cell population in the heart during development.

Nexus glia are physiologically active

At these developmental ages, the larval zebrafish heart is not yet fully innervated by neurons or infiltrated by blood vessels^12,28. We hypothesized that nexus glia may be physiologically active and could participate in the emergence of cardiomyocyte activity. To test this, we generated a Tg(gfap:GCAMP6s-CAAX) line to visualize Ca²⁺ transients. To sample nexus glia activity, we collected short time-lapse movies positioned at the midpoint of the ventricle of 3–5 dpf animals (Figure 2A, Figure S2A). Rhythmic and consistent calcium activity was observed at all ages (Figure 2B, Video S1). The fluorescence fluctuations occurred simultaneously in all labeled cells in the ventricle, suggesting that nexus glia may comprise an interconnected network (Figure 2A). The number of individual activity events as well as the total number of active time points decreased between 3 and 5 dpf (Figure 2C–2D). However, the average duration of activity events increased between 3 and 5 dpf (Figure 2E), suggesting that nexus glia activity becomes more coordinated during development.

Figure 2. Nexus glia are physiologically active.

(A) Max projection (of all time points) from time lapse movies of 3, 4, and 5 dpf Tg(gfap:GCaMP6s-CAAX) hearts. Ventricle is outlined in each panel. (B) Representative examples of calcium activity of 3, 4, and 5 dpf Tg(gfap:GCaMP6s-CAAX) hearts. (C) Quantification of total number of active time points in 3, 4, and 5 dpf Tg(gfap:GCaMP6s-CAAX) hearts; n=21–23 animals. One-way ANOVA was used. (D) Quantification of the number of activity events in 3, 4, and 5 dpf Tg(gfap:GCaMP6s-CAAX) hearts; n=21–23 animals. One-way ANOVA with Tukey’s multiple comparisons test was used. (E) Average duration of activity events observed in 3, 4, and 5 dpf Tg(gfap:GCaMP6s-CAAX) hearts; n=21–23 animals. One-way ANOVA with Tukey’s multiple comparisons test was used. (F) Maximum projections of a 4 dpf Tg(gfap:GCaMP6s-CAAX);(myl7:jRGECO1b) heart. Ventricle is outlined in green and atrium in pink in each panel. (G-H) Fluorescence intensity over time of GCaMP6s-CAAX (green line) and myl7:jRGECO1b (pink line) signal from heart shown in E. Dashed boxes outline three activity events shown in greater detail in H. (I) Incidence of animals in which a statistically significant relationship was observed between the gfap:GCaMP and myl7:jRGECO1b activity over time; n=12 animals. (J) Average most significant time point lag between myl7:jRGECO1b and gfap:GCaMP activity in individual animals; n=10 animals. (K) gfap:GCaMP activity predicted from a sampling of myl7:jRGECO1b activity (red line) overlaid with actual observed gfap:GCaMP activity (black line). (L) Average most significant time point lag between myl7:jRGECO1b and gfap:GCaMP activity in individual animals treated with CBX, MK801, or Prazosin as compared to untreated animals; n=7–8 animals. One-way ANOVA with Dunnett’s multiple comparison’s test was used. Error bars represent mean +/− SEM. Scale bar for A, F = 50 μm. Related to Figure S2.

The contractility of the heart is driven by calcium waves in the cardiomyocytes, which begin spontaneously at 22 hpf²⁵ and become highly coordinated by 4 dpf²⁹. To investigate whether the calcium activity of nexus glia was related to the calcium activity of the cardiomyocytes, we developed a transgenic line which expresses a red-shifted calcium indicator³⁰ in cardiomyocytes (Tg(myl7:jRGECO1b)). We then simultaneously imaged activity in nexus glia and cardiomyocytes, both of which showed robust rhythmic calcium activity at 4 dpf (Figure 2F–2H, Video S2). Closer inspection showed activity events beginning in cardiomyocytes slightly before nexus glia activity (Figure 2G–2H). To probe the possibility of a coordinated relationship between nexus glia and cardiomyocytes, we utilized time-series modeling. In 83% of animals, we detected a statistically significant relationship between the Ca²⁺ transients of the cardiomyocyte and nexus glia populations, with peaks in cardiomyocyte activity just prior to peaks in nexus glia activity (Figure 2I–2J). Using the time series models we built, we could predict future nexus glia calcium activity based on prior nexus glia activity and a small sampling of the cardiomyocyte activity (Figure 2K). Notably, this prediction was more accurate than one generated from a sampling of the nexus glia activity alone (Figure S2B). Collectively, these data suggest a direct and coordinated relationship between cardiomyocytes and nexus glia.

To investigate the underlying mechanism of this functional relationship, we performed pharmacological manipulations. We did not detect differences in calcium activity patterns between animals treated with the gap junction inhibitor, carbenoxolone (CBX), the NMDA receptor inhibitor, MK801, or the adrenergic signaling inhibitor, Prazosin compared to control animals (Figure 2L). Future studies will need to investigate additional signaling pathways that control the precise coordination of nexus glia and cardiomyocyte activity.

Nexus glia do not require mechanical tissue movement for their abundance

We next sought to understand whether the mechanically active environment of the heart could provide cues to direct nexus glia development. Such a hypothesis is challenging to address in mammals that require active heart contraction for animal viability but is possible to test in zebrafish. We observed no significant difference in the number of nexus glia found in the heart of 5 dpf animals that had experienced a previous halt of heart movement at 2 dpf (achieved with temporary BDM exposure) compared to those that experienced no disruption to their heart rate (Figure 3A–3C, Figure S2C).

Figure 3. Nexus glia do not require mechanical tissue movement for their abundance.

(A) Heart rate of 5 dpf zebrafish treated with 20 mM BDM for 60 minutes compared to untreated animals; n=12–22 animals. Unpaired two-tailed t-test: p<0.0001. (B) Max projections of brightfield timelapse movies of 5 dpf untreated and BDM-treated hearts. A single frame is shown as well as a maximum projection of 50 time points (collapsed). The atrium (pink), ventricle (green), and outflow tract (blue) are outlined in each panel. (C) Quantification of gfap⁺ cells seen in 5 dpf Tg(gfap:nlsGFP-P2A-tdTomato-CAAX) hearts with and without prior exposure to BDM at 2 dpf; n=9–10 animals. Unpaired two-tailed t-test: p=0.2630. (D) Heart rate of 3 dpf zebrafish injected with tnnt2a gRNA compared to uninjected animals; n=12–23 animals. Unpaired two-tailed t-test: p<0.0001. (E) Max projections of brightfield timelapse movies of 3 dpf hearts injected with tnnt2a gRNA or uninjected. The atrium (pink), ventricle (green), and outflow tract (blue) are outlined in each panel. (F) Quantification of gfap⁺ cells seen in 3 dpf Tg(gfap:nlsGFP-P2A-tdTomato-CAAX) hearts injected with tnnt2a gRNA compared to uninjected animals; n=18–22 animals. Unpaired two-tailed t-test: p=0.3134. (G) Max projection (of all time points) and single time points from a movie of a 5 dpf Tg(myl7:jrGECO1b) heart. (H-I) myl7:jRGECO1b fluorescence in ventricle (green) and atrium (pink) over time shown by z-score. Dashed line indicates a z-score threshold of 1.5. Orange line indicates time points shown in greater detail in I. Error bars represent mean +/− SEM. Scale bar for B, E, and G = 50 μm.

To further explore this idea, we prevented all mechanical movement of the heart by knocking down the cardiac troponin gene, tnnt2a, which had been shown to inhibit cardiac contractility, producing a “silent heart” phenotype^31,32. Although animals injected with tnnt2a gRNAs had lethargic or entirely absent heartbeats (Figure 3D–3E) and later developed cardiac edema, they were otherwise morphologically normal and viable. There was no difference in the number of nexus glia at 3 dpf in the hearts of tnnt2a gRNA injected animals and uninjected controls (Figure 3F, Figure S2D), suggesting that cessation of heartbeat does not impact developmental nexus glia abundance. It is possible such mechanical disruption impacts cardiomyocyte activity^26,27 but short movies of the hearts of 5 dpf animals showed rhythmic calcium activity in the heart that originated in the atrium and moved to the ventricle (Figure 3G–3I, Figure S2E–2F), which follows the known trajectory of calcium waves in the heart^29,33.

Large-scale genetic screen reveals candidate modifiers of nexus glia

We next investigated the genetic factors that may drive nexus glia development by re-analyzing a single cell RNA sequencing dataset of embryonic human heart tissue²³. This analysis led to the identification of a specific cell cluster characterized by the expression of several genes known to mark nexus glia, including SOX10, METRN, and S100B¹² (Figure 4A–4C, Figure S3A). Notably, S100B has been shown to colocalize with GFAP⁺ cardiac cells in zebrafish, mouse, and human models^12,23,34,35, while metrn is recognized as a crucial modulator of nexus glia in zebrafish¹². Furthermore, this subcluster expressed regulators associated with peripheral glia, such as ADGRG6 (GPR126)^36–38, and astroglial modifiers like EDNRB³⁹. Given that nexus glia exhibit astrocyte-like characteristics, we compared this cardiac subcluster with a published astrocyte dataset⁴⁰, and found 14 differentially expressed genes highly expressed in both datasets, including ANXA2, PTPRZ1, ID2, and ID3 (Figure 4D). Previous research has confirmed the specific expression of ID2 in zebrafish nexus glia¹².

Figure 4. Identification of genes of interest for genetic interrogation.

(A) UMAP representing clusters identified in re-analysis of Asp et al., 2019. (B) Expression of known CNG markers Sox10, S100b, and Metrn across clusters shown in UMAP plots (left) and violin plots (right). Cluster 17 is outlined and insets show a higher magnification view. (C) Heat map representing expression of the top 167 differentially expressed genes found in Cluster 17. For the full list of genes in order, please see Figure S3. (D) Venn diagram and list of the 14 genes expressed in both cluster 17 of the Asp et al., 2019 embryonic human heart sc-seq dataset and the Frazel et al., 2023 astrocytes. Related to Figure S3.

Focusing primarily on genes coding for transcription factors, cytoskeletal proteins, and cell-surface proteins, we chose 89 genes of interest. For each gene, we injected a pool of gRNAs into Tg(gfap:nlsGFP-P2A-tdTomato-CAAX) animals at the 1 cell stage (Table S2, Figure 5A). At 4 dpf, these animals were imaged automatically with a custom-built robot that combines microfluidics with confocal microscopy (Figure 5A).

Figure 5. High-throughput genetic screen reveals candidate modifiers of nexus glia.

(A) Experimental timeline for CRISPR knockdown screen. (B) Max projections of Tg(gfap:nlsGFP-P2A-tdTomato-CAAX) hearts in injected knockdown animals as compared to uninjected animals and control animals injected with Cas9 + scrambled gRNA. Images shown are representative examples of hits with an increased number of gfap⁺ cells (adgrg6 and cdh6) and fewer gfap⁺ cells (cnpy2 and psip1a/b). Ventricle is outlined in each panel. (C-D) Volcano plot displaying the mean difference of the total amount of gfap⁺ cells (C) and the number of gfap⁺ cells in the atrium (D) in injected knockdown animals as compared to controls injected with a scrambled gRNA; n=6–28 animals. Gene knockdowns that produced an increased (p < 0.05) gfap⁺ cell number are colored blue while gene knockdowns that produced a decreased (p < 0.05) gfap⁺ cell number are colored red. (E) Heat map displaying screen results for each gene as a function of the adjusted p value for the comparison of total number of gfap⁺ cells (top row), number of gfap⁺ cells in the ventricle, (middle row) and atrium (bottom row) in injected knockdown animals as compared to controls injected with a scrambled gRNA. Knockdowns causing increased number of gfap⁺ cells are colored blue and knockdowns producing fewer gfap⁺ cells are colored red. (F) Quantification of total number of gfap⁺ cells seen in knockdown animals and control animals injected with Cas9 + scrambled gRNA; n=27–56 animals. For information on statistical tests and p values, see Figure S5C. (G) Cluster breakdown of the embryonic human heart and sub-cluster breakdown of the neuronal cluster, as shown by sc-seq analysis in Farah et al., 2024. Sub-clusters identities were manually annotated. (H) Expression profile of known glial markers Sox10, Plp1, Ascl1, and S100b within the neuronal cluster of the Farah et al., 2024 dataset. (I) Expression profile of validated screen hits within the neuronal cluster of the Farah et al., 2024 dataset. aCM, atrial cardiomyocytes; BEC, blood endothelial cells; fibro, fibroblasts; LEC, lymphatic endothelial cells; ncCM, non-chambered cardiomyocytes; P-RBC, platelet-red blood cells; SMC, smooth muscle cells; vCM, ventricular cardiomyocytes; WBC, white blood cells. Scale bar in B = 50 μm. Related to Figure S4–5.

The initial round of screening identified 17 potential genetic modifiers whose knockdown disrupted the size of the nexus glia population at 4 dpf (Figure 5B–E, Figure S3B–S3E). Interestingly, we identified genes whose knockdown produced an overabundance of nexus glia and genes that limited the size of the nexus glia population (Figure 5B–5E, Figure S3B–S3E). These genes included ADGRG6 (GPR126) and DBI, which both function in peripheral glia^{36–38,41,42}. Genes with known roles in astrocytes (EDNRB³⁹ and ID3⁴³), transcription factors (PSIP1 and HOXB4), cell-surface receptors (CDH6), and cytoskeletal modulators (ARPC1B, BCAS3) were also represented. Uniquely, we found that arpc1b and nono impacted only the abundance of nexus glia in the atrium while having no effect on the ventricle population (Figure 5D–5E, Figure S3C–S3E).

Next, we assessed the specificity of candidate gene expression with re-analysis of the existing embryonic human heart scRNA sequencing dataset²³. PCSK2 and SLC35F1 were exclusively enriched in the neural crest cluster. In contrast, PSIP1, ID3, DBI, NONO, CNPY2 and others were also detected in many other cardiac cell clusters. Finally, some genes like EDNRB, GAS7, BCAS3, TTYH1, and HOXB4 were expressed in the neural crest cell cluster along with a smaller subset of other clusters (Figure S4A).

Additional rounds of screening validation confirmed that knockdown of several of the 17 candidate genes showed a significant difference in the size of the nexus glia population at 4 dpf (Figure 5F, Figure S5A–S5C), providing a list of candidate genetic modifiers of nexus glia development. We hypothesized that these genetic modifiers may be expressed in distinct phases of nexus glia differentiation, so we explored a second scRNA-sequencing dataset collected from embryonic human heart that contained a cluster expressing known glial genes⁴⁴ (Figure 5G–5H). We found a subcluster expressed intermediate glial progenitor gene ASCL1⁴⁵ and early neural crest transcript TUBB3⁴⁶, as well as several proliferation markers consistent with a proliferating intermediate progenitor. In contrast, SOX10, which labels the precursor population of nexus glia and neural crest cells¹², along with FOXD3⁴⁷ and TFAP2A⁴⁸, were enriched in a subcluster that also contained proliferative transcripts, consistent with an immature proliferating glia. Finally, mature glial markers PLP1¹³ and S100B¹⁴ were enriched in a distinct subcluster that did not contain proliferative transcripts, consistent with more mature glia (Figure 5G–5H).

Of our candidate genes, NONO, AKAP12, and PSIP2 were comparably enriched in early differentiation states; ADGRG6, EDNRB, PCKS2, TTYH1, and ID3 were enriched in more mature subclusters and had overlapping expression of PLP1 and S100B (Figure 5I, Figure S5D). ADGRG6 and EDNRB were mostly devoid of detection in the immature glial subclusters, while others like DBI and ID3 were enriched in mature subclusters but also detectable in immature subclusters (Figure 5I, Figure S5D). Some modifiers were expressed homogeneously throughout the subclusters, like CDH6 (Figure 5I). Of the full list of candidate genes, we discovered, by utilizing HCR, that a subset of these genes was expressed in the larval zebrafish heart chambers or outflow tract (Figure S5E).

cdh6 is a genetic modifier of nexus glia development

One gene of interest among the potential genetic modifiers was cdh6. Its knockdown produced overabundance and impacted the nexus glia population in both the atrium and the ventricle (Figure 5B–5F, Figure S5A–S5C). To begin to probe the role of cdh6 in nexus glia development, we first explored the known expression patterns of cdh6 within developing heart cells. In our re-analysis of the embryonic human heart scRNA-seq data from Asp et al., 2019²³, we found that CDH6 is enriched primarily within the cluster previously identified to contain nexus glia (Figure 6A, Figure 4B). In Farah et al., 2024⁴⁴ dataset, CDH6 is highly expressed ubiquitous throughout the “neuronal” cluster (Figure 6B, Figure 5I). To examine in vivo expression, we performed HCR on whole-mount embryos at 4 dpf, which detected cdh6 expression in the heart (Figure 6C) that matches the localization of gfap reporters, being most concentrated in the ventricle (Figure 1D,1G). cdh6 was also observed at the junctions between heart chambers (Figure 6C).

Figure 6. cdh6 directs nexus glia development.

(A-B) Expression of CDH6 across clusters of the Asp et al., 2019 sc-seq dataset shown in UMAP and violin plots. Cluster 17 is highlighted and inset shows higher magnification view. (B) Spatial expression profile of CDH6 in embryonic human heart, as shown in Farah et al., 2024. (C) Max projections of 4 dpf hearts labeled with HCR probe against cdh6. Atrium (pink), ventricle (green), and outflow tract (blue) are outlined. (D) Max projections of 2, 3, and 4 dpf Tg(gfap:nlsGFP-P2A-tdTomato-CAAX) hearts with atrium (pink) and ventricle (green) outlined in each panel. (E) Quantification of total number of gfap⁺ cells in 2, 3, and 4 dpf Tg(gfap:nlsGFP-P2A-tdTomato-CAAX) hearts of cdh6 crispants compared to controls; n=29–55 animals. One-way ANOVA with Šídák’s multiple comparisons test was used. (F) 3 dpf Tg(gfap:nlsGFP-P2A-tdTomato-CAAX) hearts stained with Sox10. Ventricle is outlined. (G) Quantification of percent of gfap⁺ cells labeled with Sox10 at 2, 3, and 4 dpf; n=38–54 animals. One-way ANOVA with Tukey’s multiple comparisons test was used. (H) 3 dpf Tg(gfap:nlsGFP-P2A-tdTomato-CAAX) hearts stained with PCNA. Ventricle is outlined. (I) Quantification of percent of gfap⁺ cells labeled with PCNA at 2, 3, and 4 dpf; n=41–47 animals. One-way ANOVA with Tukey’s multiple comparisons test was used. (J) 4 dpf Tg(gfap:nlsGFP-P2A-tdTomato-CAAX);Tg(sox10:nlsEos) hearts stained for GFP. White arrowhead indicates a cell that is Pc-Eos⁺ and GFP⁺ while yellow arrow indicates a cell that is Pc-Eos⁺ and GFP⁻. Ventricle is outlined in each panel. (K) Quantification of PcEos⁺gfap⁺ double labeled cells seen 48h after photoconversion; n=15 animals. (L) Quantification of percent of gfap⁺ cells labeled with Sox10 at 2, 3, and 4 dpf in Tg(gfap:nlsGFP-P2A-tdTomato-CAAX) cdh6 crispants compared to controls; n=14–34 animals. One-way ANOVA with Šídák’s multiple comparisons test was used. (M) Quantification of the total number of sox10⁺ cells at 30, 48, and 72 hpf in Tg(sox10:nlsEos) hearts of cdh6 crispants compared to controls; n=7–4 animals. One-way ANOVA with Šídák’s multiple comparisons test was used. (N) Quantification of percent of sox10⁺ cells labeled with PCNA at 2 and 3 dpf in Tg(sox10:nlsEos) hearts of cdh6 crispants compared to controls; n=15–23 animals. One-way ANOVA with Tukey’s multiple comparisons test was used. Error bars represent mean +/− SEM. Scale bar in C, D = 50 μm. Scale bar in F, H, and J = 20 μm. Related to Figure S6.

Since cardiac cdh6 was largely limited to neural cells, we utilized our cdh6 crispants to probe the developmental mechanisms of nexus glia. We designed a single gRNA targeting the third exon of cdh6, which generated lesions in 87.9% of injected animals (Figure S6A–S6C). Quantitative PCR confirmed that cdh6 RNA is reduced in cdh6 crispants (Figure S6D–S6E). Using these new reagents, we again observed that knockdown of cdh6 increased nexus glia abundance at 4 dpf (Figure 5C, Figure 6D–6E). To investigate the mechanism underlying this overexpansion, we first determined when it occurred. Nexus glia abundance at 2 and 3 dpf was comparable between cdh6 crispants and control animals but diverged at 4 dpf, suggesting that the population increase is the result of an aberration later in the maturation process of nexus glia (Figure 6D–6E).

To test this idea more specifically, we explored the developmental trajectory and specification of nexus glia. Previous work revealed that nexus glia are derived from neural crest that seeds the heart during embryonic development¹² and contributes to multiple cell populations⁴⁹. We hypothesized that nexus glia may be indirectly produced via an intermediate progenitor capable of differentiating into several cell types. To investigate this, we first examined the expression of sox10, a well-known marker of neural crest, with antibody labeling. In Tg(gfap:nlsGFP-P2A-tdTomato-CAAX) hearts, we observed few Sox10⁺ cells at 2, 3, or 4 dpf (Figure 6F, Figure S6F). Additionally, there was almost no overlap of expression between nlsGFP expression and the Sox10 antibody (Figure 6G), suggesting that Sox10 is rapidly down-regulated. These results are consistent with the co-existence of a sox10 intermediate progenitor and nexus glia that are gfap⁺ and sox10⁻.

As neural crest migration to the heart is completed by 48 hpf¹², we hypothesized that nexus glia expand by proliferating within the heart after seeding the tissue and tested this with an antibody against PCNA, a marker of replicating cells⁵⁰. Although a relatively high number of cells are PCNA⁺ within the heart, only a small percentage of the nexus glia population are PCNA⁺ (Figure 6H–6I, Figure S6G). Nexus glia proliferation peaks at 3 dpf (Figure 6I). This small proliferative capacity likely does not fully account for the rapid expansion of the nexus glia population, which triples in size between 2 and 4 dpf (Figure 1E). As a complementary approach to assess proliferation, we used time lapse imaging of Tg(gfap:nlsGFP-P2A-tdTomato-CAAX) animals, which showed no nexus glia divisions (Figure S6H–S6I). We therefore hypothesized that a neural crest-derived intermediate progenitor population undergoes asymmetric cell divisions to produce progeny that later become gfap⁺ mature glia.

To more directly test our hypothesis that nexus glia rely on an intermediate progenitor, we developed a fate mapping strategy to track cells over time. We utilized a transgenic line expressing a photo-convertible protein Eos, which switches its fluorescent profile from green to red when exposed to a 405 nm laser. At 2 dpf, individual labeled cells in Tg(gfap:nlsGFP-P2A-tdTomato-CAAX); Tg(sox10:nlsEos) ventricles were photoconverted (Figure S6J–S6K). 48 hours later, animals they were fixed and stained with an antibody against GFP (Figure S6J). This labeling approach identifies gfap:nlsGFP⁺ cells in a separate channel to distinguish them from cells expressing unconverted nlsEos. At 4 dpf (48 hours post-conversion), an average of 2.0±0.2 cells are marked with photoconverted Eos (Figure 6J, Figure S6L). This data suggests that the initial photoconverted sox10:nlsEos⁺ cell divided at least once. However, only an average of 0.5 +/− 0.2 pcEos⁺ cells are also labeled with the GFP antibody (Figure 6K), revealing that, of the progeny produced from the sox10⁺ cell, only one of those daughter cells becomes a nexus glia.

The mosaic Lifeact labeling technique described previously (Figure 1) provided a complementary approach to observe the expansion of nexus glia. Often, a single heart contained more than one labeled cell group, and as these groups were regionally distinct from each other, they could be reasonably identified as arising from distinct cells of origin. In most animals, a single clonal group was observed that contained only 1 labeled cell (Figure S6M–S6N). This data is consistent with gfap⁺ nexus glia being post-mitotic and expanding via proliferation and/or specification of an intermediate progenitor, supported also by our observations that nexus glia are stable and non-migratory (Figure 1K, Figure S6H–S6I).

Having defined the likely developmental progression of nexus glia, we next determined how cdh6 could impact abundance. Assessment of the proliferative status of nexus glia revealed no difference in the percent of gfap⁺ cells co-labeled with PCNA at 2, 3, or 4 dpf between cdh6 crispants and controls (Figure 6L). Similarly, there was no difference in the number of sox10⁺ cells present within the heart at 30, 48, or 72 hpf when compared to controls, confirming that cdh6 knockdown does not limit neural crest seeding (Figure S6O, Figure 6M). We also considered that the nexus glia population expansion in cdh6 crispants could result from alterations to individual cell morphology. Quantifying the domain size of nexus glia mosaically labeled with gfap:Lifeact-GFP-P2A-mCherry-NLS showed an average nexus glia domain size in cdh6 crispants that was indistinguishable from control animals (Figure S6P–S6Q), consistent with the idea that morphological changes are unlikely to account for nexus glia overabundance.

Overabundance could also be generated by increased proliferation of either neural crest or intermediate progenitor cells, so we used the Tg(sox10:nlsEos) transgenic and labeled proliferating cells with the PCNA antibody. No difference in the percent of cells that co-labeled with proliferation marker PCNA or apoptotic marker Caspase3 were detectable (Figure 6N, Figure S6R). Collectively, these data suggest that cdh6 is likely regulating differentiation decisions of an intermediate progenitor population that exists transiently between sox10⁺ neural crest and gfap⁺ nexus glia.

Perturbation of nexus glia population influences heart activity and function

We observed that nexus glia exhibit rhythmic calcium activity (Figure 2) and previous work demonstrated that their ablation causes ventricular fibrillation and tachycardia¹². The cdh6-mediated overabundance phenotype provided a unique opportunity to further investigate how nexus glia might contribute to heart physiology. We first evaluated changes to the physiological activity patterns of the nexus glia population with short time lapse movies of 4 dpf Tg(gfap:GCaMP6s-CAAX) animals. We scored the number and average duration of activity events seen in the first 600 time points of each movie. To assess the activity more holistically, we calculated these outcome measures at a z-score threshold of 2.0 and 1.5, representing a more (2.0) and less (1.5) stringent cutoff for activity. We observed that cdh6 crispants exhibited a significant increase in the number of activity events at a z-score threshold of 2.0 as compared to controls (Figure 7A, 7D).

Figure 7. Nexus glia overabundance affects heart physiology (A-B).

Quantification of number of events (A) and average event duration (B) observed in 4 dpf Tg(gfap:GCaMP6s-CAAX) hearts of cdh6 crispants and controls at z-score thresholds of 2.0 or 1.5; n=20–24 animals. One-way ANOVA with Šídák’s multiple comparisons test was used. (C) Quantification of the distribution of active time points binned by z-score in 4 dpf Tg(gfap:GCaMP6s-CAAX) hearts of cdh6 crispants and controls; n=20–24 animals. Chi-square test: p=0.7169. (D) Dot plot representing the activity patterns in all analyzed Tg(gfap:GCaMP6s-CAAX) hearts of cdh6 crispants and controls at z-score threshold of 1.5. Each row represents a single animal, and all active time points are marked with a diamond. (E-F) Quantification of number of events (E) and average event duration (F) observed in 4 dpf Tg(myl7:jRGECO1b) hearts of cdh6 crispants and controls at z-score thresholds of 2.0 or 1.5; n=19–22 animals. One-way ANOVA with Šídák’s multiple comparisons test was used. (G) Quantification of the distribution of active time points binned by z-score in 4 dpf Tg(myl7:jRGECO1b) hearts of cdh6 crispants and controls; n=19–22 animals. Chi-square test: p=0.0182. (H) Dot plot representing the activity patterns in all analyzed Tg(myl7:jRGECO1b) hearts of cdh6 crispants and controls at a z-score threshold of 1.5. Each row represents a single animal, and all active time points are marked with a diamond. (I-J) Quantification of heart rate observed in 3 dpf (I) and 4 dpf (J) cdh6 crispants compared to controls injected with Cas9 only and untreated controls after chronic exposure to isoproterenol (1 μM); 20–48 animals. One-way ANOVA with Tukey’s multiple comparisons test was used. (K) Quantification of heart rate observed in 4 dpf cdh6 crispants compared to controls injected with Cas9 only and untreated controls after chronic exposure to isoproterenol + DMSO or isoproterenol + HMR1556; n=45–73 animals. One-way ANOVA with Tukey’s multiple comparisons test was used. (L) Quantification of heart rate observed in 5 dpf cdh6 crispants compared to controls injected with Cas9 only and untreated controls after chronic exposure (0–4d) to isoproterenol + DMSO or isoproterenol + Chr293B; n=10–47 animals. One-way ANOVA with Tukey’s multiple comparisons test was used. (M-N) Quantification of average atrial area (M) and atrial wall thickness (N) of 4dpf cdh6 crispants compared to controls injected with Cas9 only and untreated controls after chronic exposure to isoproterenol, n=24 animals. One-way ANOVA with Tukey’s multiple comparisons test was used. (O) Quantification of heart rate observed in 5 dpf cdh6 crispants compared to controls injected with Cas9 only and untreated controls after chronic exposure (0–4 dpf) to isoproterenol + DMSO or isoproterenol + HMR1556 and then allowed to recover for 24 hours; n=30–44 animals. One-way ANOVA with Tukey’s multiple comparisons test was used. (P) Incidence of ventricular fibrillation in 5 dpf hearts of cdh6 crispants compared to controls injected with Cas9 only and untreated controls after chronic exposure (0–4 dpf) to isoproterenol + DMSO or isoproterenol + HMR1556 and then allowed to recover for 24 hours; n=30–44 animals. Fisher’s exact test: p = 0.0332. Error bars represent mean +/− SEM. Related to Figure S7.

Correspondingly, when the total number of active time points is binned according to z-score threshold, we see a slight increase in activity frequency at the higher threshold in cdh6 crispants as compared to controls although there is no difference in the total number of active time points (Figure 7C, Figure S7C). The average duration of these activity events was not different between groups at either threshold (Figure 7B). We observed no difference in the average ventricular area or fluorescence intensity between groups, confirming that these alterations in activity are not the result of more brightly labeled animals (Figure S7A–S7B). These data suggest that the over-abundance of nexus glia, via cdh6 knockdown, increases overall activity levels but not the dynamics of individual events.

Since the activity of nexus glia is intertwined with cardiomyocytes (Figure 2), we evaluated whether nexus glia over-abundance altered cardiomyocyte activity. In 4 dpf Tg(myl7:jRGECO1b) hearts, we observed that animals injected with cdh6 gRNA exhibited no difference in the number of activity events at a z-score threshold of 2.0 as compared to controls, but they displayed fewer activity events at a z-score threshold of 1.5 (Figure 7E, 7H). However, the average duration of these activity events was not different between groups at either threshold (Figure 7F). Binning the active time points reveals that although there is no difference in the total number of active time points (Figure S7F), there is a significant shift in the activity profile of cdh6 crispants as compared to controls, with crispants displaying significantly more time points at higher thresholds and significantly fewer time points at lower thresholds (Figure 7G). We confirmed there was no difference in the average ventricular area between groups (Figure S7D). cdh6 crispants also had similar average fluorescence intensities and total numbers of active events as Cas9 controls, giving us confidence that overall levels of activity are comparable between groups (Figure S7E–S7F). These data suggest that an over-abundance of nexus glia impacts the activity of cardiomyocytes.

Limiting nexus glia abundance causes tachycardia and arrhythmias and affects the heart’s autonomic responses¹². To test if nexus glia overabundance produced functional consequences, we challenged the functional capacity of the heart with exposure to sympathetic (isoproterenol) and parasympathetic agonists (carbachol)^12,51–54. In 5 dpf animals, short exposures to carbachol (60 minutes) significantly decreased the heart rate of animals injected with Cas9 only and cdh6 crispants (Figure S7G), whereas isoproterenol (30–60 minutes) caused a significant increase in heart rate in cdh6 crispants or Cas9 controls (Figure S7H).

We next considered that responses to pathogenic stress may be altered if nexus glia are increased. To explore that, we systemically treated with isoproterenol to induce heart failure via sympathetic hyper-activity^55,56. We observed that chronic isoproterenol exposure from 0–3 dpf caused a reduction in heart rate in all treatment groups (Figure 7I). However, in 4 dpf animals, when the nexus glia population is statistically larger, cdh6 crispants maintained a significantly higher heart rate after treatment. While Cas9 injected animals experienced a 33% reduction in heart rate as compared to untreated animals, the heart rate of cdh6 knockdown animals only decreased by 26% (Figure 7J). Similar results were seen in all our treatment groups at a higher isoproterenol concentration (10mM) in both 3 and 4 dpf animals (Figure S7I–S7J). Although some animals developed edema, the overall health and morphological appearance of these animals was normal. Consistent with previous reports, staining of heart tissue with Masson’s trichome confirmed that zebrafish heart exposed to chronic isoproterenol were not fibrotic like mammalian hearts (Figure S7K)⁵⁵. The size and tissue thickness of the atrium and ventricle, proxy measures for hypertrophy, a common feature of heart damage and cardiomyopathy⁵⁷, showed no differences between groups (Figure 7M–7N, Figure S7L). These results suggest that knockdown of cdh6 had a protective effect against the heart rate reversal phenotype caused by chronic isoproterenol.

The expanded nexus glia population may protect the heart from functional decline by increasing buffering capacity. To test this, we repeated our chronic exposure and concurrently treated with HMR1556, a potassium channel blocker⁵⁸. Interestingly, we observed that in animals injected with Cas9, there was no difference seen between groups exposed to isoproterenol and those exposed to isoproterenol and HMR1556 (Figure 7K). However, in cdh6 crispants, the protective effect that we observed with isoproterenol treatment alone was abolished with the addition of HMR1556 (Figure 7K). We validated these results using animals exposed to Chr293B, a second inhibitor of potassium channels⁵⁹ (Figure 7L). These data are consistent with the hypothesis that the protective role of cdh6 perturbation is at least partially mediated by potassium ions.

To assess the potential long-term consequences of this chronic sympathetic stress, we repeated our exposure experiments, allowed animals to recover for 24 hours, and then quantified heart rates at 5 dpf. While all treatment groups displayed significantly decreased heart rates compared to untreated animals, all groups showed modest recovery of heart rate after 24 hours. No difference was seen between heart rates of animals injected with Cas9 alone and cdh6 crispants, independent of whether animals were co-treated with HMR1556 or with isoproterenol alone (Figure 7O). After recovery, some animals treated with chronic isoproterenol displayed intermittent fibrillation and the occurrence of this phenotype was lower in cdh6 crispants compared to controls (Figure 7P, Video S3–S5). Although this difference is not statistically significant, the heart rates of cdh6 crispants after 24 hours recovered to a slightly higher level as compared to Cas9 controls (Figure 7O). This result, in combination with the lower incidence of ventricular fibrillation, suggests that the protective effect of cdh6 knockdown seen during chronic exposure translates to better maintenance of long-term function after periods of high sympathetic stress.

DISCUSSION

Bidirectional communication between the nervous system and internal organs is important. Here, we demonstrate that nexus glia, a specialized glial cell in the heart, are physiologically active, developmentally controlled by precise genetic mechanisms, and can confer protection against pathological heart states. Combined, these findings highlight the important contribution of nexus glia for neural control and maintenance of heart physiology.

Nexus glia contribute to the heart’s response to autonomic stimuli and rhythm maintenance¹². In this work, we further revealed that nexus glia exhibit calcium transients that occur nearly simultaneously throughout the ventricle, suggesting functional coupling. This widespread synchronization of calcium activity is similar to coordinated calcium waves observed in astrocytes and enteric glia^15,60, which is mediated by gap junctions. Calcium activity in astrocytes and satellite glia is critical for their ability to respond to and modulate neuronal activity^60–65, which suggests that nexus glia might share similar roles in cardiac regulation.

Beyond their intrinsic activity, we also observed that calcium activity of nexus glia closely mirrors calcium activity in cardiomyocytes. Our results demonstrate that an over-abundance of nexus glia alters the calcium activity patterns in cardiomyocytes, lending evidence to the idea that nexus glia can directly influence cardiomyocyte function. This bi-directional relationship is reminiscent of glial cells in the enteric nervous system, which interact directly with intestinal smooth muscle cells^16,18. Pharmacological inhibitors targeting gap junctions, NMDA receptors, and adrenergic receptors did not affect this coordinated relationship so while we cannot definitively rule out those mechanisms, they may have a minor role. Modulation of ATP through P2RY receptors could be a potential mechanism⁶⁶, although this seems unlikely given the speed and repetitive nature of the activity patterns. Metabotropic glutamate receptors could also mediate nexus glia Ca²⁺ transients, similarly to their role in astrocytes⁶⁷. Taking into account the temporal constraints of this coordination, it seems likely that ion flux could be a key regulator. The robust depolarization that occurs with each contraction could induce ion concentration changes to which nexus glia respond. Electrophysiological analysis of nexus glia would be insightful to explore.

Regulation of progenitor proliferation and cell fate decisions is critical for producing the correct diversity of cell types. Cadherins are homophilic cell adhesion molecules that influence and regulate differentiation and cell fate decisions through Wnt, BMP, and MAPK signaling pathways⁶⁸. As cdh6 is widely expressed across the presumptive nexus glia lineage, we hypothesize that it may be acting as a molecular sensor to control precursor proliferation and cell fate decisions. Following this model, in our cdh6 crispants, progenitor regulation is lost, and an over-abundance of nexus glia are produced. We theorize that expression of cdh6 in nexus glia may tile them within the heart by mediating cell-cell recognition, halting expansion when appropriate tissue coverage is achieved.

To assess the functional impact of nexus glia over-abundance, we employed an established model of heart failure mediated by chronic sympathetic activation of the β-adrenergic receptor system^55,56. Our results suggest that nexus glia may contribute to resilience against sympathetic stress. Interestingly, the cardioprotective effect of nexus glia over-abundance was abolished by a potassium channel blocker, lending credence to the idea that nexus glia may be modulating potassium in the heart like astrocytes buffer ions⁶⁹. While we observed protection in a heart failure model, it will be critical to further identify the relationship between nexus glia function and heart physiology.

Together, our data demonstrates a critical role for a subtype of glia in the modulation of appropriate heart function. Given that glia have been described in the lungs^8,10, kidney⁷, pancreas¹¹, and spleen^6,9 and that bidirectional communication exists between almost every organ and the nervous system, this work underscores a critical need to identify if nexus glia, or other resident-glia, might modulate organ function. This report could provide entry points for such future discoveries.

Limitations of Study:

The nexus glia population observed in our Tg(gfap:nlsGFP-P2A-tdTomato-CAAX) line is larger than previously reported transgenics¹². Our new transgenic could label nexus glia more consistently and robustly or other gfap⁺ cell populations, which would also be post-mitotic and not cardiomyocytes. Although we verified that calcium transients still occur in cardiomyocytes with BDM exposure, we were unable to investigate how this calcium activity relates to heart rhythm as BDM prevents heart contraction. Additionally, we cannot determine the specific source of this intracellular calcium. Our results do not rule out that mechanical movement contributes to nexus glia development, only that it does not direct abundance. Although we detect coordination of nexus glia and cardiomyocytes, the underlying mechanism of this signaling is unknown. Further investigation of other candidate molecules, specifically metabotropic glutamate and purinergic receptors is needed. Additionally, electrophysical experiments will provide important insight. Finally, as we did not confirm knockdown efficiency in our large-scale genetic screening, it is not appropriate to exclude genetic modifiers based on their absence of a phenotype.

RESOURCE AVAILABILITY

Lead contact:

Cody J. Smith, csmith67@nd.edu

Materials Availability:

All unique/stable reagents generated in this study are available from the lead contact without restriction.

Data and Code Availability:

• All data collected for the study is included in the figures and is available from the lead contact upon request.

• All original code is publicly available at: https://github.com/SmithGliaLab/Manuscripts

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

STAR METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Zebrafish model

All experimental procedures adhered to the NIH guide for the care and use of laboratory animals and were approved by The University of Notre Dame Institutional Animal Care and Use Committee (Protocol #22-07-7322). Notre Dame IACUC adheres to the United States Department of Agriculture, the Animal Welfare Act (USA), and the Association for Assessment and Accreditation of Laboratory Animal Care (International).

Zebrafish stable strains used in this study were AB, Tg(gfap:nlsGFP-P2A-tdTomato-CAAX), Tg(myl7:jRGECO1b), Tg(gfap:GCaMP6s-CAAX)⁷⁰, and Tg(sox10:Gal4 + myl7:GFP)⁷¹;(UAS:Lifeact-GFP)⁷². Zebrafish embryos were produced through pairwise mating and placed in an incubator at 28°C in constant darkness⁷³. Beginning at 24 hpf, embryos were treated with PTU (0.003%) to reduce pigmentation for imaging.

METHOD DETAILS

Generation of constructs

All constructs were generated using the Gateway LR Clonase II Plus System (Thermo Fisher) with zebrafish-compatible Tol2 vectors⁷⁴. Multi-site gateway recombination of p5e, pMe, and p3e vectors was accomplished using pDest394-Tol2 as a destination backbone⁷⁴.

gfap:nlsGFP-P2A-tdTomato-CAAX⁷⁰

A p5e-gfap (Addgene #82401) vector was used to drive expression in glial cells⁷⁵. A pMe-nlsGFP-P2A vector (Addgene #80808)⁷⁵ provided nuclear expression. A p3e-tdTomato-CAAX vector was generated with Hifi DNA Assembly by amplifying tdtomato-CAAX from pMe-tdTomato-CAAX (Addgene #135204)⁷⁶ and inserting the cassette into p3e-MCS1 vector (Addgene #49004) blunt cut with SnaBI.

myl7:jRGECO1b

A p5e-myl7 vector was generated by amplifying myl7 out of the pDest395 vector and inserting the cassette into a pENTR vector with Topo cloning. A pMe-jREGECO1b vector was generated by amplifying jRGECO1b from a pAAV.Syn.NES-jRGECO1b.WPRE.SV40 vector (Addgene #100857)⁷⁷ and inserting the cassette into a PCR8 vector with Topo cloning. These were recombined with a p3e-polyA vector⁷⁸ to drive expression of the calcium indicator in cardiomyocytes.

myl7:mCherry-NLS

A pMe-mCherry-NLS vector was generated by restriction enzyme cloning with SpeI and XhoI to put a mCherry-NLS cassette from a p3e-P2A-mCherry-NLS vector (described below) into a pMe-MCS vector⁷⁸. This was recombined with a p5e-myl7 (see above) and a p3e-polyA vector (see above) to drive nuclear expression in cardiomyocytes.

gfap:Gal4-P2A-mCherry-NLS⁷⁹

A p5e-gfap vector (see above) was used to drive expression in astroglial cells. A pMe-Gal4 vector (Addgene plasmid #53366)⁸⁰ was edited to remove the stop codon using the QuikChange Lightning kit (Agilent). A p3e-P2A-mCherry-NLS vector was generated using Hifi DNA Assembly (New England Biolabs) by amplifying mCherry-NLS from a pUAS-mCherry-NLS vector (Addgene plasmid #87695)⁸¹ and inserting the cassette into a p3e-P2A-MCS vector (Addgene plasmid #80825)⁷⁵ linearized with SmaI.

gfap:Lifeact-GFP-P2A-mCherry-NLS

A p5e-gfap vector (see above) was used to drive expression in astroglial cells. A pMe-Lifeact-GFP vector was generated with Hifi DNA Assembly by amplifying a GFP cassette from the p3e-eGFP vector (Addgene # 140877)⁸² and inserting it into an amplified pMe-Lifeact (Addgene #109545)⁸² vector. These were recombined with a p3e-P2A-mCherry-NLS vector (see above) to drive nuclear and cytoskeletal expression in nexus glia.

gfap:Lifeact-mCherry-P2A-nlsGFP

A p5e-gfap vector (see above) was used to drive expression in astroglial cells. A pMe-Lifeact-mCherry vector was generated with Hifi DNA Assembly by amplifying a mCherry cassette from the p3e-mCherry vector (Addgene # 108884)⁸³ and inserting it into an amplified pMe-Lifeact (Addgene #109545)⁸² vector. A p3e-P2A-nlsGFP vector was generated using Hifi DNA Assembly by amplifying nls-GFP from a pMe-nlsGFP-P2A vector (Addgene plasmid #80808)⁷⁵ and inserting the cassette into a p3e-P2A-MCS vector (Addgene plasmid #80825)⁷⁵ linearized with SmaI. These plasmids were recombined to drive nuclear and cytoskeletal expression in nexus glia.

Generation of transgenics

AB animals were injected with the transgenic plasmid at an approximate concentration of 25 ng/μl in a mixture containing 150 ng Tol2 mRNA⁷⁴. Injected animals were screened for bright fluorescence at 2 dpf and then grown to adulthood. As adults, injected animals were individually screened to identify founders by outcrossing to AB animals: F0 animals producing transgenic F1 progeny with the expected expression pattern were identified as founders and used to produce F1 generations. At minimum, all experiments were performed with animals of the F2 generation.

Heart rate quantifications

To measure heart rates, animals were mounted in 0.8% low melting point agarose without anesthetic and positioned with their ventral surface up. Alternatively, to circumvent the need for mounting, a small amount of 1X Tricaine-S (Pentair Aquatic Ecosystems) was added to the dish just until animals slowed enough for heartbeats to be observed. Under a dissecting microscope, contractions of the ventricle were counted for 15 seconds. These raw numbers were then multiplied by 4 to obtain a final number representing heart rate in beats per minute.

In vivo imaging with BDM

To perform in vivo imaging, animals were mounted in glass-bottomed dishes with 0.8% low melting point agarose containing 20 mM BDM. After agarose had set, dishes were covered with 20 mM BDM in solution and left on the benchtop to incubate for 45 minutes before imaging to allow the heart sufficient time to slow and stop contractions. Animals were mounted lateroventrally (2 dpf) or ventrally (3–6 dpf) with tails pulled up to position the heart as close to the glass as possible. Animals needing to be kept for days of successive imaging were immediately removed from BDM agarose after imaging sessions and transferred to dishes with fresh egg water + PTU to recover.

Images were acquired on a custom built 3i spinning disc confocal microscope as previously described with a Zeiss Axio Observer Z1 Advanced Mariana Microscope; X-cite 120LED White Light LED System; filter cubes for GFP and mRFP; a motorized X,Y stage; a piezo Z stage; 20× Air (0.50 NA), 63× (1.15 NA), 40× (1.1 NA) objectives; CSU-W1 T2 Spinning Disk Confocal Head (50 μM) with 1× camera adapter and a Teledyne Prime 95B sCMOS camera; dichroic mirrors for 446, 515, 561, 405, 488, 561, and 640 excitation; laser stack with 405 nm, 445 nm, 488 nm, 561 nm, and 637 nm with laser stack FiberSwitcher; photomanipulation from vector high-speed point scanner ablations at diffraction limited capacity; and Ablate Photoablation System (532 nm pulsed laser; pulse energy 60 J at 200 Hz)⁸⁴. Confocal images were acquired using a 40–80μm stack with a 1 μm step size. Images were processed with ImageJ. Only brightness and contrast were adjusted and enhanced for images represented in this study.

Nexus glia quantification

To track the nexus glia population at different developmental ages, 70–80 μm stacks with a 1 μm step size were obtained of the hearts of Tg(gfap:nlsGFP-P2A-tdTomato-CAAX) animals between 2 and 6 dpf. Individual nuclei were counted in each image z-stack in ImageJ using the Cell Counter plugin (https://imagej.net/ij/plugins/cell-counter.html) and recorded based on their location in the heart (ventricle, atrium, or outflow tract).

Cellular ratios quantification

To quantify the ratio of nexus glia to cardiomyocytes, in vivo imaging of 4dpf Tg(gfap:nlsGFP-P2A-tdTomato-CAAX); Tg(myl7:jRGECO1b) animals was performed to quantify nexus glia numbers. Following the live imaging, each animal was individually fixed and stained with DAPI (i.e, live imaged animal no.1 was the same animal as DAPI no.1). In brief, animals were fixed in 4% paraformaldehyde (PFA) in PBS with 0.1% Triton X-100 for 1hr at 25°C. Following 3 washes in 1x PBS, animals were permeabilized using 100% acetone for 5 minutes at 25°C followed by 100% acetone for 10 minutes at −20°C. Animals were washed 3×5 minutes in 1xPBST followed by addition of 1ug/mL DAPI diluted in 1xPBS and incubated for 10 minutes at 25°C. Animals were washed 3 times with 1xPBS and ventrally mounted in 0.8% low melting agarose for imaging in a glass bottom dish. DAPI images were collected in 70um stacks with a 1um step size. To quantify the ratio of nexus glia and cardiomyocytes, individual nuclei from matched animals were counted in each image z-stack in ImageJ using the Cell Counter plugin (https://imagej.net/ij/plugins/cell-counter.html) and recorded based on their location in the heart (ventricle, atrium, or outflow tract). This process was carried out for both nexus glia and DAPI images. The ratio of nexus glia to cardiomyocytes was recorded by subtracting the total number of nexus glia from the total DAPI count, then dividing the nexus glia number by the DAPI-nexus glia number to provide the percent of cells that are nexus glia within the total heart cell population.

Calcium imaging and analysis

Animals were mounted and pre-treated with BDM as described above. Confocal movies were acquired at a single optical plane positioned in the approximate center of the ventricle using an exposure time of 10–30 ms. 1800–2500 time points were collected, generating a 4–5 minute movie of each animal. Movies were exported as 16-bit TIFF files and imported into ImageJ. The ventricle was outlined using the freehand drawing tool. The area and integrated density of fluorescence of this region was extracted using the Multi-Measure function within the ROI manager menu. A z-score was calculated for each time point of a movie and activity events were identified by the presence of at least 1 time point that exceeded a particular z-score threshold. Average activity durations were calculated by dividing the total number of time points in a movie that exceeded the z-score threshold by the number of discrete instances in which the z-score exceeded the threshold. For all experiments, these outcomes were measured for the first 600–750 time points of each movie, since the overall activity decreased as the capture progressed, possibly due to photobleaching. To capture a more holistic picture of a wide sampling of activity, calcium data was analyzed individually at thresholds of 2.0 and 1.50.

Time series modeling

The calculated myl7 and GCaMP fluorescence z-scores were imported into R for each animal, and each series was converted into a time series format for analysis. For each animal, z-scores were plotted on a graph to visually determine whether a relationship may exist between the two time series. A dataframe for each animal’s myl7 and GCaMP z-scores was created, and the Cross-Correlation Function (CCF) was performed to determine whether there were significant relationships between the time series lags or leads. In each case, the CCF revealed that pre-whitening of the data was needed to more accurately identify the significant relationships. A large-order AR model was applied to the myl7 time series, which calculated the model’s residuals and coefficients. These coefficients were used to filter the GCaMP data for the same animal, and a time series was created from the filtered residuals. CCF was then applied to the residuals of both data series, which identified significant leads and lags between the two. A new dataframe was created with the GCaMP time series, as well as the relevant myl7 lags and leads. A linear regression model was created to determine whether the myl7 lags and leads were significant predictors of the GCaMP z-score; if there were insignificant lags and leads in the model, the most insignificant predictor was removed, and a new linear regression model was created. This process was completed until only statistically significant myl7 lags and leads remained as predictors.

Next, a SARIMAX model was created and fitted with the relevant myl7 lags and leads as predictors. A second SARIMAX model was created and fitted with only the biologically relevant myl7 lags and leads as predictors (lags of 1–3 timepoints). To determine whether these relationships had the potential to be predictive, a forecasting model was created for one representative animal. The initial model was created with the sarima.for function with all myl7 z-scores and the first 400 GCaMP z-scores with the goal of predicting the final 300 GCaMP z-scores (the SARIMA(3,0,0)x(0,1,0)₄₇ model was created with lag 2 of the myl7 data and a plot showing the predictions, prediction confidence intervals, and true values was created). To determine whether or not the myl7 lag 2 meaningfully contributed to the predictive power of the model, a second SARIMA(3,0,0)x(0,1,0)₄₇ model was created with only the GCaMP data as a predictor, and the resulting plot showing the predictions, prediction confidence intervals, and true values was evaluated as being significantly less accurate than the model including the myl7 lag 2 predictor.

CRISPR/Cas9 knockdown

Guide RNAs (Integrated DNA Technologies) against genes of interest were injected at the one cell stage. The injection mixture also contained 5 μM AltR S.p. Cas9 Nuclease V3 (Integrated DNA Technologies)⁸⁵. For the high throughput genetic screen, control animals were injected with a mixture containing a scrambled gRNA. For all other experiments, control animals were injected with Cas9 alone. To achieve heartbeat disruption, a combination of 3 guide RNAs targeting the tnnt2a gene were utilized. For cdh6 knockdown, a single gRNA targeted to exon 3 was used. The guide RNA sequences used are as follows:

tnnt2a

• 5’ GGACATCCACCGTAAGCGCATGG 3’

• 5’ CTTCGGCGGTTACATGCAAAAGG 3’

• 5’ AATCCGCAGCGAACGCGAAAGGG 3’

cdh6

• 5’ TGTCACTCAAGTGACCGCCC 3’

T7 Endonuclease Assay

To assess the lesion efficiency of gRNAs, a T7 endonuclease assay (New England Biolabs) was performed according to the manufacturer’s instructions. In brief, genomic DNA was individually extracted from animals injected with gRNA and a small region around the gRNA target site was amplified by PCR. This PCR product was briefly annealed in a reaction containing NEB Buffer 2, and then T7 Endonuclease was added to the reaction. Digested products and original PCR products were run side-by-side using gel electrophoresis. Lesion efficiency was determined by calculating the percent of digested products with smaller bands not seen in PCR product lanes for the corresponding samples.

Quantitative Reverse Transcription Polymerase Chain Reaction (qRT-PCR)

To quantify cdh6 knockdown, 24 hpf animals injected with Cas9 only or cdh6 sgRNA were collected and RNA was extracted using Direct-zol RNA Kit from Zymo Research (Cat#R205) per the manufacture’s recommendation. cDNA was generated using SuperScript^® IV Reverse Transcriptase Kit (Thermo Cat#18090050). qPCR reactions were set up in a 96-well plate (Applied Biosystems MicroAmp Fast Optical 96-well reaction plate with barcode Cat#4346906) and the master mix was prepared using SYBR Green (Applied Biosystems PowerTrack SYBR Green Master Mix Cat#A46012) with primers against cdh6 or tubulin as an endogenous control (primer sequences below). In brief, a comparative Ct program was used on the QuantStudio3 as follows: 40 cycles of 95°C 15 seconds, 60°C 1 minute and a melt curve protocol of 95°C for 15 seconds (1.6C/s), 60°C 1 minute (1.6C/s), 95°C 15 seconds (0.075C/s) with ROX as passive dye reference. cDNA PCR was prepared using GoTaq Green Master Mix (VWR Cat# PAM7123) and the same primers for cdh6 and tubulin were used as noted above. PCR parameters included an annealing temperature of 50°C for 30 seconds and a 30 second extension at 72°C for both genes. Quantification of relative expression was calculated using the 2^^−ΔΔC_t value for each group using tubulin as the reference gene for normalization. The ΔΔC_t values used to calculate the fold change were calculated using the uninjected group as reference.

• cdh6 F: 5’ GCTGAGATGTCCTACAGCAT 3’

• cdh6 R: 5’ GCTCATATAGTGCGAGTATG 3’

• tubulin F: 5’ GAGACCGGAGCTGGAAAACA 3’

• tubulin R: 5’ GGAAACCCTGGAGACCTGTG 3’

Photoconversion of Eos to track clonal expansion

At 2 dpf, single labeled nuclei in the hearts of Tg(sox10:nlsEos); Tg(gfap:nlsGFP-P2A-tdTomato-CAAX) animals were focally exposed to 405 nm UV light with the 3i vector system to photoconvert the Eos protein from green to red⁸⁶. Successful photoconversion was assessed by comparing the 488 and 561 channels in pre- and post-photoconversion images. Animals were then fixed at 4 dpf and stained with anti-GFP to discern between Eos⁺ and GFP⁺ nuclei. Slidebook and ImageJ were used to quantify the number of Eos⁺ and GFP⁺ nuclei in the heart and assess their co-localization. The following parameters were used for photoconversion: laserstack power: 8, double-click size: 8, laser pulse duration: 5 ms, raster block size: 8.

Mosaic cell analysis in zebrafish

We observed individual nexus glia within the heart relying on a construct that labeled both the nuclei and cytoskeleton using a gfap promoter. This construct, gfap:Gal4-P2A-mCherry-NLS, was injected at the 1 cell stage into Tg(sox10:Gal4 + myl7:GFP);(UAS:Lifeact-GFP) animals. We identified animals that were positive for the UAS:Lifeact transgene and had mosaic expression of our injected gfap-driven construct but did not express the myl7:GFP reporter. Alternatively, a gfap:Lifeact-GFP-P2A-mCherry-NLS construct was injected into AB animals. When applicable, the same animals were imaged over multiple days to track the maturation of individual cells.

Single cell sequencing re-analysis

All analyses were conducted using R (v4.3.0, 2023) in an RStudio (v 2023.06.01) Markdown notebook environment (https://www.r-project.org/, http://www.rstudio.com/) Code was executed using the Seurat (v5) framework for single-cell RNA sequencing (scRNA-seq) analysis⁸⁷. Additional R packages included ggplot2 (https://ggplot2.tidyverse.org), dplyr (https://dplyr.tidyverse.org), Matrix (https://CRAN.R-project.org/package=Matrix), and biomaRt^88,89. This analysis used a scRNA-seq dataset from Asp et al., 2019²³, which profiled embryonic human heart tissue at 4.5–5-, 6.5-, and 9-weeks post-conception. Filtered single-cell RNA sequencing (scRNA-seq) count matrices and corresponding metadata were obtained from Asp et al., 2019²³. A Seurat object was constructed with these inputs. Quality control filtering had been performed by the original authors. Data normalization, identification of variable features, and scaling were performed using the SCTransform function^90,91

Dimensionality reduction was conducted via Principal Component Analysis (PCA) using identified variable features, and an elbow plot guided the selection of 22 components for downstream clustering. Seurat’s FindNeighbors() and FindClusters() functions were used for graph-based clustering at a resolution of 0.7. Uniform Manifold Approximation and Projection (UMAP) was used for dimensionality reduction and visualization. Clusters were originally annotated using ScType, employing heart-specific gene sets obtained from the ScType database⁹². The scaled expression matrix from the SCT assay slot was used for enrichment scoring. The highest-scoring cell type for each cluster was assigned as the presumed identity. Clusters with low enrichment confidence (defined as scores less than 25% of the cluster’s cell count) were labeled “Unknown.” Marker based identification was conducted using Seurat’s FindAllMarkers() function (only.pos = TRUE, logfcthreshold = 0.7).

Expression of known glial markers (Metrn, S100b, Sox10) was assessed using the VlnPlot() and FeaturePlot() functions to confirm marker expression patterns across clusters. Cluster 17 expressed these features at higher levels than other clusters. Differentially expressed genes of cluster 17 were then extracted using FindMarkers() with only.pos = TRUE, and the top 250 markers were visualized using DoHeatmap(). The list of genes selected for screening was loaded from a CSV file. These genes were visualized using a heatmap across all clusters using DoHeatmap(). For each gene, violin and feature plots were generated and saved to evaluate expression using Seurat’s VlnPlot() and FeaturePlot() functions. All Seurat objects and key outputs were saved as .RDS or .csv files. Code is available at: https://github.com/SmithGliaLab/Manuscripts

High throughput genetic screen

For each of the 89 genes chosen from the single cell re-analysis, 4–6 guide RNAs were designed to target the gene using the CHOPCHOP online selection tool (4 guides for unduplicated genes, 3 guides for each of the “a” and “b” forms of duplicated genes; for a full list of gRNA sequences, please see Table S2)⁹³. Selection followed these parameters: guides should ideally be 20bp in length, begin with GA, and have no off-targets with fewer than 3 bp mismatches.

Tg(gfap:nlsGFP-P2A-tdTomato-CAAX) animals were injected with a pool of guides in combination with 2 μM EnGen Spy Cas9 NLS (New England Biolabs). Injected animals were screened for fluorescence at 2 dpf. At 4 dpf, screened animals were prepared for imaging with exposure to 20 mM BDM in solution for 45 minutes and then placed in individual wells of a 96-well plate. Images were captured on a personalized upright spinning disk screening confocal microscope equipped with a Zeiss Axio Examiner Z1 Advanced VIVO microscope, W Plan-Apochromate 20X/1.0NA Water dipping objective, CSU-W1 Spinning disk confocal (50 um disk), 405/488/561/640 excitation with quad emitter and individual emitters, Prime 95B Back Illuminated Scientific CMOS camera, 488/561/637 laser stacks, Ablate 532 photoablation system, Vector2 High-Speed Point Scanner. Attached is a Union BioMetrica VAST BioImager System for whole zebrafish organ-level imaging that contains a large particle sampler, allowing for high-throughput screening of zebrafish animals⁹⁴. A minimum of 6 animals were imaged and analyzed for each gene knockdown group in the original screen. For validation rounds of screening, a minimum of 27 animals and 2 technical replicates were collected.

Immunohistochemistry

The same immunohistochemistry protocol was performed on all tissue samples. The primary antibodies used were anti-GFP (1:500, chicken, Abcam), anti-Sox10 (1:5,000, Sarah Kucenas lab⁹⁵), anti-PCNA (1:500, Millipore Sigma), and anti-Caspase3 (BD Biosciences, 1:200). The secondary antibodies used were Alexa Fluor 488 goat anti-chicken (1:600, Thermo Fisher), Alexa Fluor 594 goat anti-mouse (1:600, Thermo Fisher), Alexa Fluor 647 goat anti-rabbit (1:600, Thermo Fisher), and Alexa Fluor 647 goat anti-chicken (1:600, Thermo Fisher). Modified immunohistochemistry protocols were adapted from the Hibi lab. Animals were fixed using 4% PFA in PBS with 0.1% Triton X-100 at 25°C for 1–3 hours. Animals were then washed with PBST (PBS, 1% TritonX-100) 3 × 5 minutes followed by a 5 minute wash with DWTx (distilled water, 1% Triton). Animals were then incubated with cold acetone and chilled at −20°C for 10 minutes and again washed with PBST 3 × 5 minutes. Next, animals were incubated for an hour in 5% goat serum in PBST at 25°C. Animals were then incubated in 5% goat serum in PBST with the primary antibody for an hour at 25°C and then incubated overnight at 4°C. After 3 washes of PBST for 30 minutes each, the animals were then incubated in 5% goat serum in PBST with the secondary antibody at 25°C for an hour and then transferred to 4°C overnight. The animals were then washed 3 times with PBST for 30 minutes and finally transferred to a 50% glycerol stock and stored in 4°C until imaging was performed.

HCR Probe Design

Each probe set was custom generated and ordered as 50pmol Oligo Pools (oPools) without 5’ end modifications.

Whole Mount Hybridized Chain Reaction In Situ

After chronic isoproterenol exposure at 4 dpf, AB animals were fixed 4% PFA in PBST overnight on a rocking shaker at 4°C. Fixed larvae were washed with 1xPBST using 2 quick washes followed by 2×15 minute washes. Samples were then dehydrated and permeabilized with a series of methanol washes from 25%−100% in 1xPBSTw (PBS, 0.1% Tween-20) for 15 minutes each at 25°C. Samples were then stored in 100% methanol overnight at −20°C. Each sample was rehydrated using a series of graded methanol washes from 100% to 20% in 1xPBST for 15 minutes each at 25°C. Each sample was washed 2 times for 5 minutes with 1xPBST. For probe detection, each sample was pre-hybridized with 40% probe hybridization buffer (Molecular Instruments) for 1 hour at 37°C. Probe solution was prepared using 3 uL of the appropriate 1 uM probe set in 200 uL of 40% probe hybridization buffer and samples were incubated with this prepared probe solution overnight at 37°C. The next day following washes at 37°C (4×15 minutes with probe wash buffer (Molecular Instruments), samples were washed with 5xSSCT (5x sodium chloride sodium citrate with 0.1% Tween 20) 2 × 5 minutes at 25°C. Next, samples were pre-amplified by incubating in 200uL amplification buffer for 1 hour at 25°C followed by the addition of 10 uL of 3 uM stock snap cooled hairpins in 500 uL amplification buffer (Molecular Instruments), protected from light on a rocking nutator overnight at 25°C. Following amplification, samples were washed to remove excess hairpins with 500 uL of 5xSSCT at 25°C 4×15 minutes followed by 2×5 minute washes with 1xPBST. Each sample was then ventrally mounted in 0.8% low melting point agarose for confocal imaging. An average of 10–20 animals were imaged per gene.

Masson’s Trichrome Staining

For histological analysis, 4 dpf Tg(sox10:Gal4 + myl7:GFP) animals that were injected with Cas9 only or cdh6 sgRNA and exposed to 1 mM isoproterenol, or untreated, from 0–4 dpf were collected. The larvae were fixed in 4% PFA in PBST overnight at 4°C. Fixed larvae were quickly washed 3 times with 1xPBS followed by 3×10 minute washes in 1xPBS at 25°C on a rocking nutator. Following the washes, 30% sucrose in 1xPBS was added to each tube of larvae and kept at 4°C on a rocking nutator for 48 hours. Larvae were then laterally embedded in OCT (Fisher Cat#4585) and stored at −80°C until cryo-sectioning. Masson’s trichrome staining was performed on 25 um sections that were dried overnight on a slide warmer at 37°C. In brief, all steps were performed in Coplin jars with each slide slowly dipped in each reagent and immediately pulled out to prevent overstaining of the fragile larval tissue. The Bouin’s fix was omitted as it comprised issue integrity. Staining began with a 1xPBS hydration, followed by 1:1 solution of Weigert’s Iron Hematoxylin stain set (Newcomer Supply Cat#1409A), distilled water rinse, Biebrich Scarlet-Acid (1% aqueous) Acid Fuchsin (0.1%) stain (Electron Microscopy Sciences Cat#26033–02, Acid Fuchsin Sigma Aldrich Cat# F-8129), distilled water rinse, 1:1 solution of 2.5% Phosphomolybdic Acid (Thermo Scientific Cat# 206380250)/2.5% Phosphotungstic Acid (Sigma Aldrich Cat# P4006–100G), 2.5% Aniline blue solution (CHEM-IMPEX, 22900), distilled water rinse, 1.0% acetic acid (Fisher Scientific Cat# A38–500), dehydrated 3× 2 minutes in 100% ethanol, cleared with xylene and cover slipped using Cytoseal and 24×50 mm rectangular cover glass #1.5 (Corning Cat#2980–245).

Drug treatments

Stock solutions were prepared in water (carbachol, isoproterenol) or DMSO (HMR1556), stored at the appropriate temperature, and diluted to the working concentration immediately before experimentation. Working concentrations were as follows: carbachol (500 μM, Sigma-Aldrich, Cat# C4382), isoproterenol (1 mM or 10 mM, Sigma-Aldrich, Cat# I5627), HMR1556 (20 mM, Sigma-Aldrich Cat# SML2354), Chr293B (100 uM, Tocris Cat# 1412). For chronic exposure treatments, working solutions of the drug treatments were prepared and refreshed every 24 hours beginning at 2 hpf. For recovery experiments, animals were exposed to the chronic drug treatment for 0–4 dpf and allowed to recover for 24 hours in a dish containing 0.003% PTU in egg water. For all drug treatment experiments, heart rate counts were only collected from animals that appeared morphologically normal and did not display edema. The procedure for counting heart rates was performed as described above.

For drug treatments performed for calcium imaging experiments, animals were exposed to drugs in solution for 2 hours, mounted in agarose containing 20% BDM (as described above). After the agarose had solidified, a solution containing 20% BDM and the experimental drug was placed on top to cover the agarose for another 45 minutes before imaged commenced. Working concentrations were as follows: carbenoxolone (CBX) (100 uM, ApexBio Cat#A8389), MK801 (50 uM, Sigma-Aldrich Cat#M107), Prazosin (50 uM, Sigma Aldrich Cat #P7791).

Software

Adobe Illustrator, Slidebook, Prism, and ImageJ were used in this study to analyze, acquire, and compile figures.

QUANTIFICATION AND STATISTICAL ANALYSIS

Scientists were blind to the identity of the experimental versus control group during quantifications whenever possible to do so. The normality distribution of experiments was assessed using the “Normality and Lognormality Test” feature on Prism, and the presence of Gaussian distributions was assessed followed by a D’Agostino–Pearson omnibus test. When applicable, two-tailed t-tests were performed. When applicable, a one-way ANOVA was performed, followed by Dunnett’s, Tukey’s, or Šídák’s post-hoc test to account for variance. Figure legends indicate n number and statistical tests used for all experiments panels that present comparisons of data. Statistical details of all experiments can be found in Table S1.

Supplementary Material

2

Video S1. Nexus glia calcium activity, related to Figure 2.

Timelapse of a 4 dpf Tg(gfap:GCaMP6s-CAAX) heart. The timelapse displays the unstimulated calcium activity of nexus glia in the ventricle. The timelapse consists of 600 captures (0.117 sec intervals) for ~70 seconds. Video rate is 30 frames per second.

3

Video S2. Simultaneous cardiomyocyte and nexus glia calcium activity, related to Figure 2.

Timelapse of a 4 dpf Tg(gfap:GCaMP6s-CAAX);(myl7:jRGECO1b) heart. The timelapse displays the unstimulated calcium activity of nexus glia (green) and cardiomyocytes (magenta) in the ventricle and atrium. The timelapse consists of 700 captures (0.415 sec intervals) for ~290 seconds. Video rate is 20 frames per second.

4

Video S3. Normal heart rhythm, related to Figure 8.

Timelapse of a 5 dpf untreated AB animal. The timelapse displays the normal heart rhythm and contraction pattern of the atrium and ventricle. The timelapse consists of 1,000 captures (0.125 sec intervals) for 125 seconds. Video rate is 7 frames per second.

5

Video S4. Mild fibrillation phenotype, related to Figure 8.

Timelapse of a 5 dpf AB animal injected with Cas9 only and exposed to 1 mM isoproterenol from 0–4 dpf and then allowed to recover for 24 hours. The timelapse displays a mild heart rhythm phenotype where the ventricle occasionally stutters and fails to maintain a smooth contraction. The timelapse consists of 1,000 captures (0.125 sec intervals) for 125 seconds. Video rate is 7 frames per second.

6

Video S5. Severe fibrillation phenotype, related to Figure 8.

Timelapse of a 5 dpf AB animal injected with Cas9 only and exposed to 1 mM isoproterenol from 0–4 dpf and then allowed to recover for 24 hours. The timelapse displays a severe abnormal heart rhythm phenotype where contraction of the heart completely freezes for several seconds and then resumes beating as normal. The timelapse consists of 1,000 captures (0.125 sec intervals) for 125 seconds. Video rate is 7 frames per second.

7

Table S1. Descriptive Statistics. Table that presents the descriptive statistics for every figure panel that presents quantifications.

Document S1. Figures S1–S7, Table S2

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Chicken polyclonal anti-GFP	Aves Labs	Cat#GFP-1020; RRID:AB_10000240
Rabbit anti-Sox10 (Ab2-sox10)	Binari et al.93	ZDB-ATB-130417–3
Mouse monoclonal anti-PCNA	Millipore Sigma	Cat#MABE288; RRID:AB_11203836
Rabbit monoclonal anti-Caspase3	BD Biosciences	Cat#559565; RRID: AB_3661884
Goat anti-Chicken IgY (H+L) Secondary Antibody, Alexa Fluor™ 488	Thermo Fisher	Cat#A-11039; RRID:AB_2534096
Goat anti-Mouse IgG (H+L) Cross-Adsorbed Secondary Antibody, Alexa Fluor™ 594	Thermo Fisher	Cat#A-11005 RRID: AB_2534073
Goat anti-Rabbit IgG (H+L) Cross-Adsorbed Secondary Antibody, Alexa Fluor™ 647	Thermo Fisher	Cat#A-21244 RRID:AB_2535812
Goat anti-Chicken IgY (H+L) Secondary Antibody, Alexa Fluor™ 647	Thermo Fisher	Cat#A-21449 RRID:AB_2535866
Chemicals, peptides, and recombinant proteins
2,3-butanedione monoxime (BDM)	Millipore Sigma	Cat#B0753
Tricaine-S	Pentair Aquatic Ecosystems	Cat#NC0589195
Carbamoylcholine chloride (carbachol)	Tocris	Cat#2810
Isoprenaline hydrochloride (isoproterenol)	Thermo Fisher	Cat#AC437210050
HMR1556	Sigma Aldrich	Cat#SML2354
Chromanol 293B	Tocris Biosciences	Cat#1412
Carbenoxolone (CBX)	Apex Bio	Cat#A8389
MK801	Sigma Aldrich	Cat#M107
Prazosin	Sigma Aldrich	Cat#P7791
Critical commercial assays
HCR™ Amplifier (v3.0)	Molecular Instruments	B1–488
HCR™ Amplifier (v3.0)	Molecular Instruments	B3–564
HCR™ Amplifier (v3.0)	Molecular Instruments	B5–647
HCR™ Probe Hybridization Buffer (v3.0)	Molecular Instruments	N/A
HCR™ Probe Wash Buffer (v3.0)	Molecular Instruments	N/A
HCR™ Amplifier Buffer (v3.0)	Molecular Instruments	N/A
Deposited data
Single-cell sequencing dataset of developing human heart	Asp et al.23	https://pubmed.ncbi.nlm.nih.gov/31835037/
Single-cell sequencing dataset of developing human heart	Farah et al.44	https://cells.ucsc.edu/?ds=hoc+allheart&cell=09W0D_LA_AGGTCATGTGTATGGG&pal=tol-sq-blue
Experimental models: Organisms/strains
Zebrafish: AB	ZIRC	ZDB-GENO-960809–7
Zebrafish: Tg(gfap:nlsGFP-P2A-tdTomato-CAAX)	This paper	N/A
Zebrafish: Tg(myl7:jRGECO1b)	This paper	N/A
Zebrafish: Tg(gfap:GCaMP6s-CAAX)	Koh et al.70	ZDB-PUB-251223–7
Zebrafish: Tg(sox10:Gal4 + myl7:GFP)	Hines et al71	ZDB-LAB-020513–1
Zebrafish: Tg(UAS:Lifeact-GFP)	Helker et al72	ZDB-ALT-130624–2
Oligonucleotides
tnnt2a gRNA1: 5’ GGACATCCACCGTAAGCGCATGG 3’	This paper	N/A
tnnt2a gRNA2: 5’ CTTCGGCGGTTACATGCAAAAGG 3’	This paper	N/A
tnnt2a gRNA3: 5’ AATCCGCAGCGAACGCGAAAGGG 3’	This paper	N/A
cdh6 gRNA: 5’ TGTCACTCAAGTGACCGCCC 3’	This paper	N/A
gRNAs for high-throughput genetic screen: See Supplemental Table 1	This paper	N/A
cdh6 qPCR F: 5’ GCTGAGATGTCCTACAGCAT 3’	This paper	N/A
cdh6 qPCR R: 5’ GCTCATATAGTGCGAGTATG 3’	This paper	N/A
tubulin qPCR F: 5’ GAGACCGGAGCTGGAAAACA 3’	This paper	N/A
tubulin qPCR R: 5’ GGAAACCCTGGAGACCTGTG 3’	This paper	N/A
Recombinant DNA
myl7:mCherry-NLS	This paper	psSL17
p5e-myl7	This paper	psSL03
pMe-MCS	Kwan et al.74	N/A
p3e-P2A-mCherry-NLS	This paper	psSL14
pAAV.Syn.NES-jRGECO1b.WPRE.SV40	Dana et al30	Addgene#100857; RRID:Addgene_100857
p3e-polyA	Kwan et al.74	N/A
gfap:Gal4-P2A-mCherry-NLS	This paper	psSL15
p5e-gfap	Fowler et al.75	Addgene#82401; RRID:Addgene_82401
pME-Gal4	Fujimoto et al.78	Addgene#53366; RRID:Addgene_53366
pUAS-mCherry-NLS	Zhang et al.79	Addgene#87695; RRID:Addgene_87695
gfap:Lifeact-GFP-P2A-mCherry-NLS	This paper	psSL19
pME-Lifeact-mCherry	This paper	psSL26
p3e-mCherry	Ariotti et al.81	Addgene#108884; RRID:Addgene_108884
pMe-nlsGFP-P2A	Fowler et al.75	Addgene#80808; RRID:Addgene_80808
p3e-P2A-nlsGFP	This paper	psSL27
p3e-P2A-MCS	Fowler et al.75	Addgene#80825; RRID: Addgene_80825
gfap:Lifeact-mCherry-P2A-nlsGFP	This paper	psSL28
pMe-Lifeact	Hall et al.80	Addgene#109545; RRID:Addgene_109545
p3e-eGFP	Hall et al.80	Addgene#140877; RRID:Addgene_140877
gfap:nlsGFP-P2A-tdTomato-CAAX	Koh et al.70	N/A
pMe-tdTomato-CAAX	Oehlers et al.76	Addgene#135204; RRID:Addgene_135204
p3e-MCS1	Moore et al.96	Addgene#49004; RRID:Addgene_49004
Software and algorithms
ImageJ (v1.54)	National Institutes of Health; Schneider et al94	RRID:SCR_003070; https://imagej.net/
Slidebook	Intelligent Imaging Innovations	RRID:SCR_014423; https://www.intelligentimaging.com/slidebook.php
Prism	GraphPad	RRID:SCR_002798; http://www.graphpad.com/
RStudio	Posit Software, PBC	RRID:SCR_000432; https://posit.co/
Single cell sequencing re-analysis code	This paper	https://github.com/SmithGliaLab/Manuscripts

Highlights.

• Nexus glia exhibit dynamic Ca²⁺ transients

• Nexus glia Ca²⁺ transients coordinate with cardiomyocyte Ca²⁺ transients

• Manipulation of cdh6 modifies nexus glia abundance

• Modulation of nexus glia abundance can be protective against cardiac exhaustion