Semaphorin3f as a cardiomyocyte derived regulator of heart chamber development

During development a pool of precursors form a heart with atrial and ventricular chambers that exhibit distinct transcriptional and electrophysiological properties. Normal development of these chambers is essential for full term survival of the fetus, and deviations result in congenital heart defects. The large number of genes that may cause congenital heart defects when mutated, and the genetic variability and penetrance of the ensuing phenotypes, reveals a need to understand the molecular mechanisms that allow for the formation of chamber-specific cardiomyocyte differentiation. We used in situ hybridization, immunohistochemistry and functional analyses to identify the consequences of the loss of the secreted semaphorin, Sema3fb, in the development of the zebrafish heart by using two sema3fb CRISPR mutant alleles. We find that in the developing zebrafish heart sema3fb mRNA is expressed by all cardiomyocytes, whereas mRNA for a known receptor Plexina3 (Plxna3) is expressed preferentially by ventricular cardiomyocytes. In sema3fb CRISPR zebrafish mutants, heart chamber development is impaired; the atria and ventricles of mutants are smaller in size than their wild type siblings, apparently because of differences in cell size and not cell numbers. Analysis of chamber differentiation indicates defects in chamber specific gene expression at the border between the ventricular and atrial chambers, with spillage of ventricular chamber genes into the atrium, and vice versa, and a failure to restrict specialized cardiomyocyte markers to the atrioventricular canal (AVC). The hypoplastic heart chambers are associated with decreased cardiac output and heart edema. Based on our data we propose a model whereby cardiomyocytes secrete a Sema cue that, because of spatially restricted expression of the receptor, signals in a ventricular chamber-specific manner to establish a distinct border between atrial and ventricular chambers that is important to produce a fully functional heart. 4CiLbvKt57C6wmJ-E7Wy_5 Video abstract Video abstract


Background
The heart is the first organ to form and become functional in an embryo, and congenital heart defects affect 9 per 1000 live births [1]. Collectively, congenital heart defects involve the irregular formation of heart chambers and/or valves. In vertebrates, chamber morphogenesis follows the twisting, expansion and septation of the linear heart tube into a two-(teleost), three-(amphibian) or four-(mammalian/avian) chambered heart [2]. These cardiac chambers, the atrium and ventricle, differ in their biochemical properties, electrophysiological and contractile capacities, as well as in their transcriptional profiles [3,4]. What maintains the segregation of the cell populations in the two discrete chambers as the embryonic heart differentiates and undergoes morphogenesis is unknown.
During development, cells within tissues acquire unique identities, and need to be segregated from one another within subdivisions of the tissue [5,6]. This is true for example of the developing hindbrain, where the initial expression of genes within adjacent rhombomeres is initially imprecise, and is sharpened as embryogenesis proceeds, producing domains with straight borders [6]. Key in this process are interactions between neighbouring hindbrain cells that involve contact-mediated repellent signaling via Ephrin ligands and Eph receptors. Whether a similar process of refinement of the border between ventricular and atrial chambers of the heart occurs is unknown. Chamber-specific development depends on extrinsic factors, which include Nodal, Notch, Bone morphogenetic protein (Bmp), Fibroblast growth factor (Fgf ) and retinoic acid [7][8][9]. These signals then pattern the early heart tube to establish secondary localized signaling that spatially restricts the expression of differentiation genes to the specific chambers and valves. Intrinsic regulators, such as transcription factors, are also necessary for chamber-specific cardiomyocyte development. For example, ventricle specific expression of Irx4 is necessary in chick for the expression of ventricular Myosin Heavy Chain I and suppression of atrial Myosin Heavy Chain [10]. Further, PITX2 within the murine left atrium inhibits the expression of Shox2 and the specification of the pacemaker cells of the sinoatrial node [11]. An unanswered question is whether active mechanisms are required to maintain segregation of the two populations of cardiomyocytes once established, especially considering the disruptive forces of cardiac chamber morphogenesis and heart looping.
One group of molecules known to act as repellents for developing cells of the nervous and cardiovascular systems are the secreted class 3 Semaphorins (Sema3s), which are used as guidance cues for vessels, axons and neural crest cells [12][13][14][15]. Sema3 signaling is mediated through the canonical Plexin (Plxn) receptors and their coreceptors, the Neuropilins (Nrp) [16]. Sema3s are implicated in murine cardiovascular development via their regulation of the directed movement of associated cells [17]. Sema3s pattern embryonic vessels by directing migrating endothelial cells [18], and play various roles in heart development: Sema3a regulates angioblast migration to form the dorsal aorta in zebrafish [19] and sympathetic innervation of the mouse heart [20], murine neural crest cells and zebrafish epicardial cells require Sema3c and Sema3f, respectively, to move into the heart outflow tract [21,22], Sema3d promotes migration of neural crest cells into the primary heart field in zebrafish [23], and Sema3e-mediated PlxnD1 signaling in endothelial cells is required for ventricular compaction in the mouse heart [24]. Whether Sema3s play roles in heart formation via the direct regulation of cardiomyocyte development is unknown.
Here, we take advantage of the zebrafish model as a powerful tool to study the development of the heart; the embryos do not require a functional cardiovascular system during embryogenesis [25], which allows for the extended characterization of manipulations. We characterized sema3 expression in zebrafish embryos and found mRNA for secreted Sema3fb in cardiomyocytes during chamber differentiation. This expression was of particular significance given that we find nrp2b and plxna3 receptor mRNAs are expressed preferentially by ventricular cardiomyocytes, suggesting spatially restricted Sema3fb signaling within the developing myocardium. To investigate a role for ventricular cardiomyocyte Sema3fb signaling, we generated CRISPR/ Cas9 zebrafish sema3fb mutants. With loss of either Sema3fb, or signaling through its Plxna3 receptor via morpholino-oligonucleotide knockdown, chamber specific gene expression is either lost from or expands beyond the border between the atrium and ventricle. Accompanying the loss of segregated gene expression between the cardiac chambers is a decrease in the size of both chambers, and significantly impaired cardiac output and edema of the early post-embryonic heart. We propose a model whereby Sema3fb signaling ensures segregation of atrial and ventricular cardiomyocyte chambers, a feature that appears critical for chamber development and heart function.

Zebrafish strains and maintenance
Zebrafish (Danio rerio) were maintained on a 14-h light/10-h dark cycle at 28 °C. Embryos were obtained by natural spawning, raised in E3 medium supplemented with 0.25 mg/L methylene blue, and staged by convention [26,27]. To permit exact developmental staging, adult zebrafish spawned for 10 min only before embryo collection. Dishes were screened at 6 h post fertilization (hpf ) to remove unfertilized eggs or embryos that showed delayed embryogenesis. We used sema3fb ca305 (28) and sema3fb ca306 genetic mutants. These lines were maintained as heterozygotes for all initial experiments, and thereafter maintained as homozygotes and wild type (WT) siblings for future embryo collection. The Tg(-6.5kdrl:mCherry) ci5 line was used as an outcross for sema3fb ca305 mutants to label endothelial cells [29], Tg(flk:EGFP) to label endothelial cells, and Tg(myl7:EGFP) [30] Casper fish for identification of cardiomyocytes. Sex was not a variable in these studies as zebrafish cannot be sexed as embryos.

sema3fb mutant lines
To generate genetic knock-outs by using the CRISPR/ Cas9 system a sgRNA (5'-GAA GGA CAA GAA GAC CCG CG) targeting sema3fb exon 2 was selected following CHOPCHOP query [31] and analysis of secondary structure using Vienna RNAfold Prediction (rna.tbi.univie.ac.at). sgRNA template generation and transcription was carried out in accordance to the protocol described previously [32]. One-cell stage embryos were injected with a 1 nl mix of 56-60 pg sgRNA and 190 pg cas9 mRNA (Addgene plasmid #47,322). Mosaic embryos were raised to adulthood and crossed with Tupel Long fin fish to identify founders. To genotype, caudal fin clippings were collected from tricaine methanesulfonate (MS-222; 160 mg/L) anesthetized adults. Genomic DNA extraction was performed [33] and amplified by PCR using primers (Forward: 5'-ATT GCC CCA CAA AAT AAC ATTC; Reverse: 5'-GTC TAC TCT GTG AAT TTC CCGC) around the expected mutation site. Amplicons were sequenced at University of Calgary DNA Core facility for definitive genotype confirmation.

Immunolabeling and imaging
Embryos were fixed overnight in 4% PFA/PBS. Samples were blocked in PBST with 10% bovine serum albumin and 2% normal sheep serum for 30 min and transferred to primary antibody solution in 1/10 blocking buffer overnight at 4 °C. Samples were washed with PBST and incubated in secondary antibody (Alexa Fluor 488/555 rabbit/mouse) supplemented with 5 mg/ml Hoechst in PBST for 45 min. Embryos were fixed and imaged by confocal microscopy. The following primary antibodies were used: MF20 (1:500, Developmental Studies Hybridoma Bank (DSHB)), pHH3 (1:500, Millipore), S46 (1:10, DSHB), zn8 (DM-GRASP; 1:500, DSHB). ApopTag Peroxidase In Situ Apoptosis Detection kit (Millipore) was used to detect apoptotic cells in embryos, with overnight diaminobenzidine staining. For confocal microscopy, embryos were imaged with a 10 × or 20 × objective on a Zeiss LSM 700 microscope after mounting in 0.8-1% low melting point agarose (Invitrogen) on a glass bottom dish. Optical slices were taken at intervals from 1-5 μm and subjected to 2 times averaging. Z-stacks were processed in Zen Blue as maximal projections and compiled using Adobe Photoshop 2020. For recovery of the myl7:EGFP or flk:eGFP transgene following ISH, embryos were put into acetone at -20 °C for 20 min followed by 3 × PBST washes. Embryos were blocked for 1 h with 10% normal sheep serum in PBST then incubated overnight at 4 °C with JL-8 (1:500, Clontech Laboratories Inc). Secondary antibody was added as above before embedding the samples in JB4 medium (Polysciences) and sectioning at 7 µm with a Leica microtome (Leica).

Quantitation of in situ hybridization expression domains
WT and mutant hearts were imaged at 48 hpf from a ventral view of the embryo. Images were blinded as to genotype, and brightness and contrast adjusted on individual images to maximize the resolution of the domain above background. A line drawn across the morphological constriction between the atrium and ventricle was used as the atrioventricular border. Either FIJI [36] or Axiovision (Carl Zeiss Microscopy) software were used to encircle the expression domain and measure the domain area. Schematics (Fig. 4, Additional file 1: Fig. 4) indicate how myh7 positive ventricular length and width were measured.

Heart function and morphology analysis
Embryos (72 hpf ) were anaesthetized with minimal MS-222 and positioned on their right side laterally on a glass slide in a water droplet. Video images were taken 30 frames/second for 15 secs using a Leica DM5500 B microscope equipped with a Leica DFC365 FX. Heart rate was determined in a 15 s window. Measurements [37] were made from frame stills during systole and diastole across three heart beats using ImageJ. Briefly, the following equations were used [2]

Hematoxylin and eosin (H&E) staining
H&E staining of JB-4 plastic sections was carried out as described previously [38] with the following modifications: Acid and base washes were omitted, Harris Modified Hematoxylin Solution (Sigma) and Eosin Y (EMD) were used, and tap water replaced prescribed substitute. Slides were air dried for 15 mins [39] and coverslipped using Permount (Fischer Scientific). Myocardial ventricle wall thickness for each embryo was the average of measurements made at two points along the anterior-posterior axis from 1-2 sections stained by H&E.

Cardiomyocyte measurements
Morphometries (area and perimeter) of confocal imaged, Zn8-immunolabelled cardiomyocytes were measured in ImageJ at 48 hpf. Cells measured had clearly visible outlines. Of note, only cells central to the outer curvature of the ventral ventricle were assessed. Circularity was used to distinguish cellular morphologies in which 1 represents a perfect circle. This was calculated by the following formula: Circularity = 4 πArea/Perimeter [2]. Measurements were averaged from 8-12 cells per embryo. Additionally, projected confocal MF-20-immunolabeled images were analyzed using 'RGB Measure' tool on ImageJ across the entirety of the heart, values averaged for all samples, and represented as an intensity plot across distance.

Sema3f Overexpression and Morpholino-mediated Knockdown
For heart specific overexpression, sema3fa was amplified from Image Clone 9,038,816 using primers that incorporate attb1/b2 recombination sites (5'-GGG GAC AAG TTT GTA CAA AAA AGC AGG CTT CAC CAT GCA GGG AGC CGG GAC TTT GGTG-3' and 5'-GGG GAC CAC TTT GTA CAA GAA AGC TGG GTT TGT CTC AGC CAT GCT GCT CTG CTC -3') and inserted into pDONR221 to create a pME-sema3fa vector. Three-way Tol2 gateway cloning [40] was used to insert sema3fa downstream of the cardiomyocyte myl7 promoter [30] giving myl7:sema3fa:p3E-MTpA in Tol2 with pDest:mKate [41] as the transgenesis marker. One cell-stage zebrafish embryos were injected with a solution consisting of 12.5-25 ng/µl plasmid with or without 25 ng/µl transposase mRNA, and embryos checked at 48 hpf for mKate eye expression. For knockdown of sema3fb we injected 1-3 ng of an antisense morpholino oligonucleotide [42] that targets the ATG start site of sema3fb (CAT AGA CTG TCC AAG AGC ATG GTGC, Gene Tools LLC, Corvalis, OR, USA) into 1-cell zebrafish embryos. For knockdown of plxna3 we injected 2-4 ng of an antisense morpholino oligonucleotide that targets the first initiation codon of plxna3 [42]. For morpholino injections, control embryos were injected with water. 48 hpf embryos were scored for edema and fixed for ISH and/or anti-myc immunohistochemistry.

Statistics
Data sets for qualitative scoring or quantitation were analyzed in a blinded fashion. Sample size calculations were not performed because we characterized mutants, and thus had no prediction as to the likely size of effects to use in power calculations. Results are expressed as mean ± standard error of the mean (SEM) unless otherwise indicated. Statistical analyses were performed by using Prism 8 software (Graph Pad). An unpaired, nonparametric Mann Whitney U test was used for comparisons of two samples, and for comparisons between multiple samples a One-Way ANOVA was used for normally distributed data and a Kruskal-Wallis test used if data failed to pass a normality test.

sema3fb is expressed by developing cardiomyocytes
To elucidate the spatial and temporal expression of sema3fb in the zebrafish heart we performed whole mount RNA in situ hybridization (ISH) on embryos from 10-72 hpf. In zebrafish, cardiac progenitors of the atrial and ventricular chambers are bilaterally specified by 5 hpf [43]. The progenitors condense at the midline by 21 hpf and migrate to produce a linear tube of differentiating myocardium that surrounds the underlying endocardium [43,44]. This tube starts pumping between 24-26 hpf [45]. By 48 hpf, the two-chambered heart is discernable by the constriction of the atrioventricular canal (AVC) [43,46]. At 24 hpf, sema3fb mRNA was present through the linear heart tube (Fig. 1A). To confirm expression of sema3fb in cardiomyocytes, we cut JB4-plastic sections of sema3fb ISH on Tg(flk:EGFP) and Tg(myl7:EGFP) fish, where GFP labels the endocardial cells that line the ventricles and cardiomyocytes, respectively. Within the heart tube, sema3fb ISH signal was present adjacent to the GFP + endocardium (Fig. 1B) in cardiomyocytes throughout the myl7:EGFP heart tube, as visualized in whole mount (Fig. 1C,C') and a sagittal section (Fig. 1D,D' ,D''). sema3fb mRNA remained in the ventricle and the atrium as the heart undergoes rightward looping and chamber morphogenesis (24-48 hpf ), with expression at 48 hpf highest at the atrioventricular border (Fig. 1E, arrows). Weak expression persisted throughout the myocardium of the 72 hpf heart (data not shown). ISH of early stage embryos suggested that sema3fb mRNA was present in the bilateral heart fields at earlier stages, as marked by the cardiomyocyte progenitor marker hand2 (Additional file 1: Fig. 1). These data indicate that mRNA for a secreted Sema is expressed by cardiomyocytes at the time they undergo differentiation, and agree with reported RNAseq analysis of FACs-isolated zebrafish cardiomyocytes, where sema3fb mRNA is enriched in the heart over the rest of the embryo [47]. Sema3s signal canonically through Nrp and Plxn receptors [13]. To identify potential receptors via which Sema3fb could influence cardiac chamber development we performed whole mount ISH for plxn and nrp genes.
Expression of zebrafish nrp genes was reported previously [48][49][50]. nrp2b and plxna3 mRNAs were present alongside sema3fb mRNA in the bilateral heart field at 18 hpf (Additional file 1: Fig ; data not shown). In summary, mRNAs for both Sema3fb and its known receptors are expressed by newly formed cardiomyocytes, arguing that Sema3fb plays a role in cardiomyocyte development. Interestingly, while sema3fb is expressed throughout the embryonic ventricle and atrial myocardium, the receptor data at 48 hpf suggest that Sema3fb signaling becomes spatially restricted mainly to the ventricular myocardium.

Generation of a Sema3fb loss-of-function genetic model
To assess the importance of Sema3fb for heart development and function we used CRISPR/Cas9 gene editing technology [32] to generate sema3fb genetic mutants. A specific guide RNA (sgRNA) targeting exon 2 ( Fig. 2A) was co-injected with cas9 mRNA at the one cell stage and a single founder was recovered that housed two allelic variants; a 19-base pair (bp) deletion and a 10-bp insertion. Both lines were propagated from this founder to generate stable homozygous viable generations, alongside wild type (WT) siblings. RT-PCR analysis of mRNA isolated from 24 hpf embryos indicated that the 19 bp deletion mutant (hereafter sema3fb ca305 ) lacked exon 2 that contains the ATG start site (Fig. 2B), presumably because the deletion at the end of exon 2 disrupts a key splice site. Indeed, sequencing of the PCR product generated with exon 1-exon 3 primers showed exon 1 spliced in frame with exon 3 to give an mRNA that if translated would produce a protein missing the first 62 amino acids (aa), including the signal sequence (amino acids 1-23) required for protein secretion and the first 14 aa of the SEMA domain necessary to elicit intracellular signaling [51] (Fig. 2A). Software could not identify another signal sequence in the predicted translated protein, thus in the sema3fb ca305 allele a C-terminal truncated Sema3fb, if even produced, would not be secreted from cardiomyocytes. RT-PCR and sequencing analysis indicated that exon 2 was included in the mRNA product in the 10 bp insertion allele (hereafter sema3fb ca306 ) (Fig. 2B). The sema3fb ca306 allele is predicted to produce a protein that is 58 aa in length and is prematurely truncated within the SEMA domain ( Fig. 2A). Of note, a commercial antibody against mouse SEMA3F failed to distinguish between multiple Sema3s in zebrafish (data not shown). Importantly, injection of an antisense morpholino oligonucleotide (MO) against the ATG start site of sema3fb produced a similar heart edema when injected at the one-cell stage in WT embryos as seen for sema3fb ca305 (Additional file 1: Fig. 3A). MO injection into sema3fb ca305 embryos produced no greater incidence of severe edema (Additional file 1: Fig. 3A), arguing strongly for the specificity of the heart phenotype in sema3fb mutants being due to Sema3fb loss. Of note, sema3fa mRNA was not detected in the 48 hpf WT and sema3fb ca305 hearts (Additional file 1: Fig. 3B), indicating that Sema3fa upregulation does not compensate for the loss of Sema3fb. ) and primers used to identify the mutation. sema3fb ca305 fish have a 19 bp deletion (dots) and sema3fb ca306 fish have a 10 bp insertion (red sequence). Schematic representation of WT and mutant proteins. In the sema3fb ca305 allele, exon 1 is spliced in frame with exon 3, with loss of 62 amino acids (aa) at the beginning of a protein that would be translated, including the signal sequence (ss) and the first 14 aa of the SEMA domain.
In the sema3fb ca306 allele, the 10 bp insertion introduces a frame shift and a premature stop codon to generate a predicted truncated 58 aa protein. F, forward primer; R, reverse primer. B RT-PCR indicates that exon 2 is skipped in the sema3fb ca305 allele; no PCR product is generated with exon 2-exon 3 primers, but product is generated with exon 3-exon 4 primers. Exon 2-exon 3 primers reveal exon 2 is present in the sema3fb ca306 allele. C Brightfield images of sema3fb +/+ , sema3fb ca305 and sema3fb ca306 72 hpf embryos. Dotted line outlines the pericardial sac. D The percent of embryos that present with cardiac edema at 48 hpf (N = 3, n = 60 for each genotype) and 72 hpf (N = 3, n = 60 for each genotype). E The severe mutant phenotype displaying major cardiac edema and failure to thrive (E), and the abnormal linear arrangement of the heart (E' , inset in E).   2D). Most mutant embryos exhibited edema, though the severity of the defect varied. Of note, other zebrafish mutants show a similar variability in heart phenotype, with fish showing less severe edema able to recover and survive [52]. A small percentage of homozygous embryos exhibited severe cardiac edema and a hypoplastic ventricle and a linear arrangement of the heart tube ( Fig. 2E,E'), and did not survive past 5-7 days post fertilization (6.1%, n = 82 embryos), (Fig. 2E,F). The remaining embryos developed to adulthood and were fertile. To quantitate the edema phenotype, we measured the area of the pericardial sac at 48 hpf: The area of the sac of sema3fb ca305 hearts (35,290 ± 2,251 µm 2 , n = 26 (standard error of the mean (SEM) here and throughout the paper, unless otherwise indicated), n = 27, p < 0.0001, Mann Whitney) was larger significantly than that observed in WT (21,983 ± 1,038 µm 2 ) (Fig. 2G). Heterozygous sema3fb ± embryos also exhibited a spectrum of heart edema, ranging from no edema to severe edema, suggesting that genetic dosage influences the phenotype (data not shown). Of note, we excluded embryos with severe edema from further analysis so as not to confound data interpretation. As the phenotypes between the two allelic variants were highly comparable, we used sema3fb ca305 fish for our analyses, confirming key findings in sema3fb ca306 fish. Hematoxylin and eosin (H&E)-stained sections revealed that the thickness of the ventricle myocardium at 72 hpf was reduced significantly (6.3 ± 0.5 µm; Mann Whitney, p = 0.024, N = 2, n = 6) in sema3fb ca305 as compared to WT (9.4 ± 0.7 µm; N = 2, n = 3) hearts (Fig. 2H), suggesting that myocardial development was impaired.

sema3fb mutants have impaired cardiac output and edema
By 72 hpf, cardiac function is considered rhythmic and the zebrafish circulatory system relatively mature [53].

Early heart development occurs normally in the sema3fb mutants
The edema of the sema3fb mutants suggested a defect in heart development. To begin to address the underlying biology, we examined whether early cardiac development was defective, since sema3fb appears to be expressed in the bi-lateral heart fields (Additional file 1: Fig. 1). Expression of fgf8a, a heart field marker [54] (Additional file 1: Fig. 3C-E), and cardiomyocyte progenitor markers nkx2.5 and tbx5a [55,56] (Additional file 1: Fig. 3I-K), were not affected in the sema3fb mutants. Note that fgf8a expression was not impacted elsewhere in the embryo, including in the brain, somites and tailbud region (Additional file 1: Fig. 3F-H). Further, the expression of myosin heavy chain 7 (myh7), a marker of ventricular cardiomyocytes [57], was condensed at the midline in similar numbers of WT (N = 2; n = 14/16) and sema3fb ca305 (n = 12/14) 21 hpf embryos (data not shown). These data suggest that the migration of cardiomyocytes to the midline and fusion to form the cardiac disc, and initial differentiation of ventricular cardiomyocytes, were not impacted by Sema3fb loss. Next, we assessed the heart at 24 hpf by whole mount ISH for the cardiomyocyte marker cardiac troponin T type 2a (tnnt2a) [58], and for myh7 and myosin heavy chain 6 (myh6) [59] to label ventricular and atrial cardiomyocytes, respectively (Additional file 1: Fig. 4). These genes were present in both WT and mutant hearts. While 24 hpf WT hearts had elongated into linear tubes [60,61] (Additional file 1: Fig. 4A,D,E,H,L), sema3fb ca305 and sema3fb ca306 hearts were still in a cone/condensed state (Additional file 1: Fig. 4B-D,F-H,L). The delay in heart morphogenesis was minimal as within two hours sema3fb mutant hearts elongated; 100% of sema3fb ca305 (N = 1, n = 7/7) and 90% of sema3fb ca306 (N = 1, n = 9/10) hearts (Additional file 1: Fig. 4N,O), comparable to the numbers observed in WT siblings (N = 1, n = 4/4) at 23 hpf (Additional file 1: Fig. 4M,P). Notably, the cranial ganglia that emerge immediately adjacent to the heart developed normally in aged-matched embryos (Additional file 1: Fig. 4Q,R). Overall, these data argue for largely normal initial formation of the atrial and ventricular chambers in the absence of Sema3fb.
We next assessed cardiomyocyte differentiation, focusing on the ventricle, as at 48 hpf cardiomyocytes in this chamber express the requisite receptors to respond to Sema3fb. We assessed the expression of the ventriclespecific homeobox-containing transcription factor, iroquois1a (irx1a) [62], and T-box transcription factor (tbx5a), which at 48 hpf is expressed in the ventricle and anterior atrium [63]. irx1a was expressed in both the WT and mutant ventricle ( tbx5a was also expressed in WT (N = 1, n = 9) and sema3fb ca305 (N = 1, n = 7) ventricles (Fig. 4Q, R), apparently at similar levels as quantitated by RT-qPCR (-/-0.93 ± 0.3 (standard deviation) fold change vs. WT, N = 3; Mann Whitney, p = 0.7). These data indicate that while the sema3fb ca305 ventricle is smaller than in WT, the ventricular differentiation program is not affected globally by Sema3fb loss.

The border between the ventricle and atrium appears disrupted with loss of Sema3fb
The fact that sema3fb mRNA is expressed by cardiomyocytes of both atrial and ventricular chambers, while key receptors are present predominantly in the 48 hpf ventricle, combined with the known role for Sema3s in providing positional information to cells [64], raised the possibility that Sema3fb might provide spatial information to cardiomyocytes as to their ventricular localization. In this respect, we were intrigued by an apparent disruption of the tight border between the expression of atrial vs. ventricular chamber specific genes in 48 hpf mutant hearts. In most WT hearts (N = 5, 30/31 hearts), the myh7-expression border aligned with the morphological constriction between the ventricle and atrial chambers (dotted line, Fig. 4E). In contrast, in  Fig. 4G) hearts, the posterior border of the high myh7 expression domain sat anterior to the constriction, with myh7 ISH signal low or absent from the atrioventricular band (yellow arrowheads, Fig. 4F, G). To quantitate this phenotype, we measured the area between the atrio-ventricular constriction and the border of high myh7 expression. This area was quite small in most WT hearts, reflecting the alignment of ventricular myh7 expression with the atrioventricular constriction (WT 156.6 ± 18 µm 2 , n = 31 embryos) (Fig. 4T). Because the myh7 border often sat anterior to this morphological constriction in mutants, the myh7(-) area was significantly larger than in WT (sema3fb ca305 502.1 ± 36 µm 2 n = 26; sema3fb ca306 631.7 ± 7 µm 2 , n = 18; p < 0.0001 One Way Ordinary ANOVA, Dunnett's multiple comparisons). The myh6 expression border between the atrium and ventricle was also disrupted in mutant hearts. In most WT hearts the anterior border of myh6 expression aligned with the morphological constriction between the atrium and ventricle (n = 20/26), but expression often extended a small extent past this border into the ventricle in sema3fb ca305 (N = 4, n = 23/31) and sema3fb ca306 (N = 1, n = 6/7) hearts ( Fig. 4I-K; yellow arrowheads). To visualize the border between the ventricle and atrium we marked ventricular cardiomyocytes by myh7 ISH and atrial cardiomyocytes by S46 immunolabeling [65]. Here it was obvious that while the two markers abutted each other in the WT heart ( Fig. 4U; n = 6/6), they were separated from one another in the mutant heart and S46 immunoreactivity extended into the gap (arrows, Fig. 4V; n = 6/6 separated). To ask whether border defects were present earlier in sema3fb mutant hearts we assayed myh6 at 36 hpf. Interestingly, myh6 label anterior to the atrioventricular constriction was present in both the WT and sema3fb ca305 heart (Fig. 5a, b). These data suggest that in WT an initially sloppy border resolves over time as the chambers continue to mature, a process that fails to occur in the absence of Sema3fb.
To further investigate the chamber border, we assessed other markers that show differential expression between the atrium and ventricle. In WT hearts, the border of the irx1a at the AVC is straight, but was imprecise in sema3fb ca305 hearts, as evident by an increase in the length of the ISH border (Fig. 4O, P insets, dotted lines) in mutants vs. WT hearts (N = 2, WT 44.3 ± 3 µm, n = 13; sema3fb ca305 85.3 ± 9 µm, n = 21; p < 0.0001, Mann Whitney). In support, tnnt2a mRNA exhibited a sharp drop in level in the atrium relative to the ventricle in WT hearts (Fig. 5C), but lacked an obvious change in expression between the two chambers in both mutant alleles. To confirm at the protein level that there was disruption of the chamber border in sema3fb mutants at 48 hpf, we used antibodies for proteins expressed at higher levels by ventricular than atrial myocardium: MF20 is a marker in zebrafish that appears more robustly expressed by ventricular than atrial cardiomyocytes [66], and DM-GRASP marks the surface of cardiomyocytes [67]. Whole mount immunostaining, followed by confocal microscopy, revealed separate atria and ventricles in WT hearts at 48 hpf (N = 1, n = 5/5 MF20; N = 3, n = 11/12 DM-GRASP, Fig. 5D, F). In contrast, the distinct expression border between the chambers was disrupted in at least 83% of mutants (N = 1, n = 8/9 MF20; N = 3, n = 10/12 DM-GRASP, Fig. 5E, G). This loss of a sharp ventricle to atria border was evident with the quantitation of the average intensity of MF20 immunolabel over the antero-posterior axis of the heart (WT n = 5, sema3fb ca305 n = 9; Fig. 5N).
The expansion of bmp4a expression in mutant hearts could have occurred because of a direct effect of Sema3fb on gene transcription; Sema3fb may normally inhibit Page 14 of 23 Halabi et al. Cell Communication and Signaling (2022) 20:126 bmp4a expression by ventricular cardiomyocytes that are not part of the AVC. If true, overexpression of Sema3f in ventricular cardiomyocytes might be expected to reduce bmp4a mRNA levels. To test this possibility, we injected a Tol2 myl7:sema3f-myc construct into embryos at the one-cell stage along with transposase mRNA, with injection of the plasmid alone serving as control. While we tried but failed to clone sema3fb, the fact that Sema3fa and Sema3fb paralogs share 83% amino acid identity, and sema3fa or sema3fb mRNA both result in delayed neutrophil recruitment in a zebrafish model of inflammation [69], argued that ectopic Sema3fa could substitute functionally for Sema3fb. Thus, we took advantage of an available full-length cDNA clone of zebrafish sema3fa for this assay. At 48 hpf, the sema3f-expressing fish were processed by whole mount ISH for bmp4a mRNA. Notably, Sema3f was expressed in a mosaic fashion in both the atrium and ventricle, as visualized by immunolabeling for the myc-tag of Sema3fa (Fig. 5L). Interestingly, Sema3f overexpression produced a similar bmp4a phenotype to that seen with loss of Sema3fb. Both control and Sema3fexpressing 48 hpf hearts exhibited enhanced expression of bmp4a at the AVC (Control, 100% n = 20/20; Sema3f overexpression, 97% n = 35/36). While bmp4a mRNA was present only within the AVC in most control hearts (75%, n = 15/20), this was true in a small number of sema3f-overexpressing hearts (25%, n = 9/36), where the bmp4a domain spread beyond the AVC into the atrium and/or ventricle (arrowheads, Fig. 5M). Measurement of the area of the bmp4a expression domain in WT and mutant hearts supports this observation (WT 1652 ± 244 µm 2 , n = 20; Sema3fa overexpression 3346 ± 205 µm 2 , n = 36; Mann-Whitney, p < 0.0001). These data argue that Sema3fb does not inhibit bmp4a transcription. The fact that both loss and gain of Sema3f function produced a similar phenotype argues that normally Sema3fb signaling needs to be strictly regulated, and that either too little or too much signaling disrupts normal development of the border between the ventricular and atrial chambers, such that that the regulation of bmp4a expression is disrupted.
Because cardiomyocyte size is linked intimately to the hemodynamic forces of flow within the heart [67,72], we asked whether sema3fb mutant hearts were still smaller than their WT counterparts when cardiac flow was removed from the equation by growing hearts independent of the circulatory system. WT and mutant hearts were explanted at 24 hpf, cultured for 24 h, and assessed by ISH for the expression of atrial myh6 (Additional file 1: Fig. 6A, B). myh6 + mutant atria were significantly smaller (N = 2 WT 9,982 ± 490 µm 2 , n = 10 explants; sema3fb ca305 6,035 ± 633 µm 2 , n = 13 explants; Mann Whitney, p = 0.0011) in size as compared to WT (Additional file 1: Fig. 6C), arguing that chamber defects present in the Sema3fb mutants independently of hemodynamic forces.

Heart valve development disrupted slightly with Sema3fb loss
The defects in gene expression observed at the AVC of sema3fb mutants led us to ask if heart valves develop normally, as cells at the boundary between the two chambers give rise to the mature valves [46,73]. Valves were present in sema3fb ca305 hearts when assessed in H&E-labeled 72 hpf sections (Fig. 6A, B). To confirm these data, we bred the sema3fb ca305 allele onto the Tg(kdrl:mCherry) line, in which the endocardial cells that make up the valves [73] are mCherry + at 72 hpf [29]. In both the WT and sema3fb ca305 hearts, valves were present at the border between the atrium and ventricle (Fig. 6C, D), and were of similar lengths (N = 2; WT 61.6 µm ± 2.5 µm, n = 10 embryos; sema3fb ca305 60.1 µm ± 3.8 µm, n = 12 embryos; Mann Whitney, p = 0.72) (Fig. 6N). The intensity of the mCherry label in the sema3fb ca305 valves, however, was reduced as compared to WT (Fig. 6C, D). Further, the distribution of the valves differed between the genotypes, with the WT valves roughly symmetrically distributed on either side of the constriction between the atrial and ventricular chambers, but preferentially located on the ventricular side of this border in mutants (Fig. 6O).

Discussion
Our data argue that embryonic cardiomyocytes secrete Sema3fb that spatially restricts chamber-specific gene expression at the atrioventricular border to ensure proper function of the embryonic heart. We find sema3fb mRNA is expressed by all cardiomyocytes over the period of chamber morphogenesis and differentiation, while expression of mRNAs for the candidate receptors, Nrp2b and Plxna3, is restricted predominantly to ventricular cardiomyocytes by 48 hpf. We propose that Sema3fb informs ventricular cardiomyocytes of their location at the border with atrial cells, and that the neighboring atrial cells likely lack the machinery to sense Sema3fb. The consequence is the proper refinement of chamber-specific gene expression at the AVC, a feature that is disrupted in the absence of Sema3fb. The size of both heart chambers is smaller, we propose because of these defects in gene expression. Our work argues for the first time that a cardiomyocyte-derived secreted Sema3 signals potentially in a spatially-restricted manner to promote the normal refinement of the border between the ventricle and atrium, a process required for proper chamber size and function. We suggest a model whereby cardiomyocyte-derived Sema3fb signals to Plxna3/Nrp2b-expressing ventricular cardiomyocytes. In support, sema3fb mRNA is expressed by early cardiomyocytes in both heart chambers, while at 48 hpf mRNAs for Plxna3 and Nrp2 appear restricted largely to the ventricular cardiomyocytes. Of note, these receptor mRNAs also appear restricted to a subregion of the 28 hpf heart tube, though we did not directly assess their expression by ventricular vs. atrial cardiomyocytes at this stage. Further, the timing of heart phenotype presentation in the sema3fb mutants argues against the causative involvement of non-cardiomyocyte tissues. While Purkinje fibers secrete SEMA3A that patterns sympathetic innervation of the murine pacemaker node [20], Sema3fb in zebrafish likely does not play a similar role, at least at the 48 hpf analysis time point, in that the cardiomyocyte phenotypes arise prior to when the node is innervated [76]. Epicardial cells also arrive in the heart after clear chamber defects are evident in the mutants at 48 hpf: epicardial cells emerge at around 55 hpf and migrate to and attach to the myocardium by 72 hpf [39]. We also saw in mutant hearts the normal arrival of cardiomyocytes from the secondary heart field [56] and CNC [77]. Interestingly, Sema3fb is important in zebrafish epicardial development at later stages: by 5 dpf sema3fb is expressed by tbx18 + epicardial cells and not cardiomyocytes, and Sema3fb regulates epicardial cell numbers in the bulbus arteriosus [22]. Thus, Sema3fb has at least one additional role in heart development than the early function we describe here.
Endocardial cells are important in cardiomyocyte development. For instance, PLXND1 signaling within the murine endocardium promotes cardiomyocytes to form trabeculae [78]. Yet, our data argue that alterations in cardiomyocytes in sema3fb mutants do not occur secondary to defects in the adjacent endocardial cells. First, the notch1b and nfatc1 AVC endocardial markers are expressed at 48 hpf in sema3fb mutants. Second, in both WT and sema3fb mutants the vegfr2/kdrl-expressing endocardium lines both heart chambers, and the endocardium-derived heart valves form. The mutant valves do appear mis-distributed at the atrioventricular border, possibly as a result of altered gene expression at the atrioventricular border, but this is a mild phenotype. Third, the cardiomyocytes and not the endocardium express mRNAs for the known Sema3f receptors, Plxna3 and Nrp2 [79]. Nor do we think Sema3fb acts on endothelial cells via Vegfr2, as SEMA3F inhibition of VEGF signaling occurs via NRP and not VEGFR [80,81], and nrp mRNA appears to be absent from the early zebrafish endocardium [50]. Finally, we saw no sustained defects in heart tube assembly, a process known to be influenced by the endocardium [82]. These data argue that the contribution of the endocardium to the initial heart impairments we observe in sema3fb mutants is minimal. Instead, the literature and our data strongly support the idea that cardiomyocyte defects arise from the loss of Sema3fb signaling within the cardiomyocytes themselves. Nonetheless, it is possible that defects in valve morphogenesis, and subsequent anatomy and function contribute to the cardiac dysmorphogenesis and edema seen in the sema3fb mutants.
Our data argue against a Sema3fb signal promoting directly a ventricular fate or repressing atrial/AVC fates. While both sema3fb and receptor mRNAs are expressed within the early bilateral heart field, ventricular and atrial cardiomyocyte progenitors are specified much earlier, prior to zebrafish gastrulation [83]. Indeed, both the formation of the ventricular and atrial chambers and their subsequent morphogenesis appear to occur normally in the absence of Sema3fb. Furthermore, irx1a and tbx5a, transcription factors implicated in ventricular cardiomyocyte differentiation, remain expressed throughout the mutant ventricle. Thus, if Sema3fb has an early role in heart development, it likely acts redundantly with other mechanisms.
The main impact of Sema3fb loss occurs after the initial specification of the chambers. Interestingly, Tolterodine-mediated blockade of the cardiac conduction system from the 10 ss to 48 hpf results in a small ventricle in the absence of alterations in cell death and ventricular cardiomyocyte numbers [84], similar to what we see in sema3fb mutants. Yet, other aspects of the mutant and Tolterodine-treated hearts differ, including the complete loss of notch1b mRNA at the AVC and a normal sized atrium in the latter, but not the former. Thus, the sema3fb mutant phenotype is not explained simply by the loss of the cardiac conduction system. Both atrial and ventricular chambers are smaller in the absence of Sema3fb, and several genes and proteins whose expression normally obeys the border between the atrium and ventricle are disrupted: 1) bmp4a and tbx2b, which are expressed by the specialized slow-conducting cardiomyocytes of the AVC [85] by 48 hpf, fail to show this restriction in mutants (bmp4a label at 72 hpf argues that ultimately this refinement occurs), 2) myh7 is largely absent from the ventricle immediately proximal to the border, 3) myh6 spreads anteriorly past the atrioventricular morphological constriction, and 4) DM-GRASP, MF20, and tnnt2a mRNA that are normally expressed more robustly in the ventricle show no dramatic change in expression at the border between the ventricle and atrium. These data support the idea that Sema3fb provides a patterned signal to ventricular cardiomyocytes at the border with the atrial chamber. This signal does not establish the border between the chambers, as the AVC does form in the sema3fb ca305 heart. Further, several genes still exhibit the expected profound change in expression at the border, including irx1a, bmp4a and myh6, though the expression border at the AVC for all three genes is imprecise and jagged.
Instead, we propose that Sema3fb allows for the sharpness of the border to emerge in a timely manner. Ventricular cardiomyocytes in the border region can respond to Sema3fb while immediately adjacent atrial cardiomyocytes lack the receptors to do so. Our data suggest that this differential signaling ability is required for the refinement of the AVC region. The overexpression data support the idea that Sema3f tells ventricular cardiomyocytes that they sit at the border with the atrium. Certainly, the fact that ectopic bmp4a expression in the 48 hpf heart is seen with both loss-and gain-of function of Sema3f argues against the possibility that Sema3fb regulates bmp4a transcription. Instead, our data suggest that if the correct spatial information provided by Sema3fb is lost, either because the signal itself is gone (mutant) or the signal is disrupted by cells in the AVC region experiencing different levels of signaling (overexpression), proper refinement of the AVC region fails to occur. An imprecise AVC region leads to disrupted regulation of cardiomyocyte-specific genes. In support, restriction of Bmp signaling to the AVC is necessary to prevent inhibition of ventricular cardiomyocyte differentiation by Tbx2 [86][87][88].
Interestingly, in sema3fb mutants the atrium is also impacted in size and misexpression of ventricular and AVC markers. Given that many atrial cardiomyocytes may lack the machinery to respond to Sema3fb, altered atrial differentiation arises presumably secondarily to disrupted ventricular cardiomyocyte and/or AVC development. One possibility is that disrupted ventricle function impacts atrial cardiomyocytes, as the function of the two chambers shows some inter-dependence [59]. Explanted mutant atria are still smaller than their WT counterparts, however, despite having removed flow as a variable. Potentially, alterations in the secretion of factors from ventricular cardiomyocytes and/or cells of the AVC affects the differentiation of the nearby atrial cardiomyocytes. Alternatively, the presence of plxna3 mRNA at 48 hpf in the anterior pole of the atrium may indicate a potential direct role for Sema3f signaling in atrial development.
The heart chambers are smaller in sema3fb mutants, though the data suggests that fewer cardiomyocytes is not the explanation. The expression of fgf8a, tbx5a and nkx2.5 in the primary heart field, and markers of secondary heart field-derived cells, are comparable in the two genotypes. Further, cell death and proliferation at 48 hpf are not impacted by Sema3fb loss. Indeed, similar numbers of ventricular cardiomyocytes are present in WT and mutant hearts at this stage. Instead, without Sema3fb these cardiomyocytes appear to be smaller than normal. The fact that the atrial chamber was smaller in the absence of changes in proliferation, apoptosis and cell integration also argues for smaller cardiomyocytes within the atria. Heart dysfunction could result in smaller cells, in that zebrafish cardiomyocytes hypertrophy normally at about 45 hpf in response to shear stresses arising from the onset of blood flow through the heart [72]. Yet, the atria of explanted mutant hearts were still smaller than those of WT, even though shear force was removed as a functional regulator. Potentially, Sema3fb signaling causes cytoskeletal rearrangements that induce ventricular cardiomyocyte expansion. For atrial cardiomyocytes,