Retinal Pigment Epithelium and Neural Retinal Progenitors Interact via Semaphorin 6D to Facilitate Optic Cup Morphogenesis

Abstract Cell movement propels embryonic tissues to acquire shapes required for mature function. The movements are driven both by acto-myosin signaling and by cells interacting with the extracellular matrix (ECM). Unknown is whether cell-cell interactions within a tissue are also required, and the molecular mechanisms by which such communication might occur. Here, we use the developing visual system of zebrafish as a model to understand the role cell-cell communication plays in tissue morphogenesis in the embryonic nervous system. We identify that cell-cell-mediated contact between two distinct cell populations, progenitors of the neural retina and retinal pigment epithelium (RPE), facilitates epithelial flow to produce the mature cupped retina. We identify for the first time the need in eye morphogenesis for distinct populations of progenitors to interact, and suggest a novel role for a member of a key developmental signaling family, the transmembrane Semaphorin6d, as mediating communication between distinct cell types to control tissue morphogenesis.


Introduction
During development individual tissues undergo morphogenesis driven by cell movements (for review, see Cavodeassi, 2018;Moreno-Marmol et al., 2018). Morphogenesis ensures that the tissue takes on the necessary shape with specific cell types in their proper locations so that the organ can carry out its appropriate function. This is true of the eyes, where during embryogenesis eye vesicles emerge from the forebrain and undergo extensive morphogenesis to produce a cupped retina (optic cup). The outer leaflet of the optic cup comprises the neurons of the retina, while the inner leaflet consists of the retinal pigment epithelium (RPE; Heermann et al., 2015) that supports the photoreceptors (for review, see Bharti et al., 2011). The cell interactions and molecular mechanisms that drive optic cup morphogenesis, however, are poorly understood.
With the zebrafish model, the speed of development, the transparency of the embryos, and fluorescent cell specific transgenic lines, have allowed for visualization of eye progenitor movements in the live embryo over the period of eye morphogenesis (Picker et al., 2009;Kwan et al., 2012;Ivanovitch et al., 2013;Heermann et al., 2015;Cechmanek and McFarlane, 2017). These studies revealed the cellular movements that drive both vesicle evagination and elongation (Ivanovitch et al., 2013), and the second phase of eye morphogenesis, where neural retina progenitors of the inner leaflet of the eye vesicle move around the distal rim of the forming optic cup into the outer leaflet to become the temporal neural retina (Picker et al., 2009;Kwan et al., 2012;Heermann et al., 2015;Sidhaye and Norden, 2017). As a consequence, the RPE stretches around the back of the neural retina (Kwan et al., 2012;Cechmanek and McFarlane, 2017). The cupping of the neural retina depends on the apical constriction of retinal progenitors and Integrin signaling (Nicolás-Pérez et al., 2016;Sidhaye and Norden, 2017), while movement around the rim requires interactions of neural progenitors with Laminin on the basal surface of the eye neuroepithelium (Sidhaye and Norden, 2017). Unclear is whether the movements of the cells of the inner eye vesicle leaflet depend on contact between RPE and neural retina progenitors, and if so, what molecules mediate the interactions.
Semaphorins (Semas) are a large family of secreted, membrane-associated and membrane-tethered ligands, of which classes 3-7 are present in vertebrates (for review, see Raper, 2000). Semas play repellent roles, and direct cells and/or their neurites away from the Sema source, but can also control proliferation, neurite outgrowth, and act as attractants (for review, see Battistini and Tamagnone, 2016;Alto and Terman, 2018). In the developing visual system, different members of the secreted Sema3 family control the growth of retinal ganglion cell axons and dendrites (Sakai and Halloran, 2006;Atkinson-Leadbeater et al., 2010;Kita et al., 2013), while transmembrane Sema6a controls eye progenitor movements in the early eye vesicle (Ebert et al., 2014). Here, we use the zebrafish model to explore a novel role for transmembrane Sema6s as regulators of tissue morphogenesis in the developing nervous system. We find the transmembrane ligand, Sema6d, and its known receptor, Plexina1a (Plxna1a), are expressed in a complementary pattern in the inner leaflet of the eye vesicle before the onset of optic cup morphogenesis; with transcripts for sema6d and plxna1a in neural retinal and RPE progenitors, respectively. Loss-of-function of sema6d by CRISPR/Cas9-mediated mutagenesis causes defects in temporal but not nasal optic cup morphogenesis, a phenotype that is reproduced by transient knock down of Plxna1a, and by blockade of a known Sema6d signaling effector, c-Abl (Kang et al., 2018). With loss of Sema6d signaling, morphogenesis of the temporal eye initiates, but then stalls: neural progenitors of the inner leaflet fail to move properly into the retina, and RPE progenitors fail to come to abut the lens. These data reveal for the first time the importance of cell-cell interactions in optic cup morphogenesis, and identify transmembrane Sema6s as functioning in neural progenitors to facilitate tissue morphogenesis during nervous system development.

Animals
Tupfel Longfin (TL; ZIRC) embryos were raised at 28°C in E3 medium supplemented with 0.25 mg/l methylene blue as described previously (Westerfield, 2000). Embryos were staged by number of somites up to the 18 somite stage (ss), and by hours postfertilization (hpf) thereafter (Kimmel et al., 1995). The Tg(rx3:GFP) fish were gifted from the Wittbrodt lab (Rembold et al., 2006) and the Tg(tfec:EGFP) line was regenerated using a Tol2 construct provided by Brian Link (Miesfeld et al., 2015). Tg (vsx1:GFP) fish were generated by the Higashijima lab (Kimura et al., 2008). Zero hpf is designated as the time dividers were removed for breeding, shortly after lights on in the fish facility. The University of Calgary Animal Care Committee approved the animal protocols.

Confocal time-lapse imaging
For in vivo imaging, 12-ss embryos were dechorionated and mounted (dorsal head down at 12 ss) in 0.8% low melting point agarose (Invitrogen) in 1Â E3 medium, then covered with 1Â E3 containing 0.16 mg/ml tricaine (Research Organics). A stage heater was used to keep embryos at 29°C. Images were acquired by using a Zeiss LSM700 confocal microscope and Zen Black software (Zeiss): 24-30 z-sections, 3-to 5-mm z-step, 5-min intervals, 8-to 12-h time frame, 20Â objective. NIH ImageJ (Schneider et al., 2012) software was used to manually track cells over time. Embryos that moved significantly in the recording period were not used for analysis.

Eye explant cultures
Wild-type Tg(tfec:EGFP) eyes were dissected out at the 18 ss and plated on 0.01% poly L-lysine (Sigma) plus 25 mg/ml Fibronectin (Sigma) coated slides and grown in 1Â MBS (88 mM NaCl, 1 mM KCl, 1 mM MgSO 4 , 5 mM HEPES, pH 7.8, and 2.5 mM NaHCO 3 ) containing 10 mg/ml Research Article: New Research gentamycin (Sigma), with or without 300 ng/ml recombinant mouse Sema6d-Fc chimera (R&D Systems). Explants were cultured at 28°C for 20 h then live imaged on a Zeiss 40 CFL compound fluorescent microscope and, in a blinded fashion, both the number of tfec:EGFP1 cells that had exited the explants, and the percent coverage of the eye by RPE-GFP1 signal, quantified.

Imaging of fixed tissue
Whole embryos were imaged in 3% methylcellulose using a Zeiss Stemi SV 11 microscope with an AxioCam HRc (Zeiss) camera and AxioVision (Zeiss) software. For sectioned samples, embryos were dehydrated with an ethanol work-up and embedded in JB4 medium (Polysciences Inc.) and 7-mm-thick sections were cut with a Leica microtome. Sections were collected on glass slides and mounted with a coverslip in Aquapolymount (Polysciences Inc.) and imaged using a Zeiss Axioplan2 microscope with an MRc camera (Zeiss) and AxioVision (Zeiss) software.

Adobe Photoshop and Illustrator suites
Adobe CS5 Photoshop was used to crop and adjust brightness of images. Adobe CS5 Illustrator was used to compile and annotate data photographs as well as generate illustrations.
Experimental design and statistical analyses ( Table 2) For characterization of the sema6d mutants we employed non-random experimental design statistical methods. For plxna1 knock-down and pharmacological manipulations (Sema6d-Fc and dasatanib) we employed random experimental design statistical methods. Statistical analysis was performed by using Prism8 software (GraphPad). For analysis of mutants, sample size calculations were not performed because we had no prediction as to the likely size of effects to use in power calculations. Wild-type and mutant embryos were chosen randomly from clutches. All data were tested for normality by using the Shapiro-Wilk test. When the data were normally distributed, unpaired, two-tailed t tests were used to compare data between two samples, and a one-way ANOVA with post hoc test used for multiple comparisons. For the time lapse analysis, we used a non-parametric Mann-Whitney U test as sample sizes were too small to determine normality. In the text, we indicate the number of the number of biological replicates (n) and the number of independent replicates (N). Analysis of whole-mount RNA ISH embryos and quantitative measurements (except for time-lapse imaging cell speeds) were performed with the researcher blinded to genotype or condition.

Results
sema6d is expressed by neural retina progenitors and its receptor, plxna1, by adjacent RPE progenitors We investigated the expression of genes encoding transmembrane Sema6 proteins in early zebrafish eye development (data not shown), which suggested possible roles for Sema6s in the control of eye progenitor movements. For instance, sema6a is expressed in the early eye vesicles as they evaginate bilaterally from the forebrain and elongate, and is required for proper entry of progenitors in a spatially regulated manner into the eye vesicles (Ebert et al., 2014). Here, we explore a role for Sema6d, whose mRNA is expressed by eye progenitors at the onset of optic cup morphogenesis. To understand the spatial and temporal expression of sema6d, we performed whole-mount RNA ISH with antisense riboprobes for sema6d and its receptor, plxna1 (Toyofuku et al., 2004a,b), on zebrafish embryos ranging from the 8 ss to 24 hpf. Of note, there are two plxna1 genes in zebrafish that are encoded on separate chromosomes. We refer to the plxna1a gene as that encoded on Danio rerio chromosome 23 (NCBI reference sequence XM_017353686.2), and to plxna1b as that encoded on chromosome 6 (NCBI reference sequence NM_001110010.1). The timing of optic cup morphogenesis in zebrafish is well described; eye vesicles evaginate from the neural keel as early as the 4 ss (Ivanovitch et al., 2013), and invaginate around the forming lens starting at about the 17 ss (Li et al., 2000), with optic cup morphogenesis mostly complete by 24 hpf (Kwan et al., 2012;Heermann et al., 2015). Transcript for sema6d was first detected at approximately the 16 ss, with expression in the eye vesicle, brain and ventrally along the anteroposterior axis of the embryo (Fig. 1A). In plastic transverse sections through the eye, sema6d was expressed more highly in the ventral portion of the inner (Fig. 1D, arrows; future temporal retina) and outer leaflets of the eye vesicle. Notably, sema6d mRNA was absent from the presumptive RPE domain, which sits within the dorsal inner leaflet (Fig. 1D, bar). By 24 hpf, staining for sema6d transcript was localized specifically to the temporal optic cup (data not shown). Of note, between 16 and 24 hpf, eye morphogenesis alongside flexing of the brain changes the early dorsoventral axis of the retina to that of nasotemporal, such that "early ventral" becomes "mature temporal" tissue, and "early dorsal" becomes "mature nasal" tissue ( Fig. 1K). In dorsal whole mount, transcript for plxna1a was detected at approximately the 14 ss in the neural keel (Fig. 1B), and in the presumptive RPE domain (shown at the 16 ss; Fig.  1G, arrows), as indicated by the RPE marker pmel1a (Cechmanek and McFarlane, 2017;Fig. 1H). This expression was confirmed in section, where plxna1a ISH signal and Plxna1-like immunoreactivity, with a rabbit polyclonal antibody against human PLXNA1, were visible in the through the brain and eye. sema6d transcript is expressed in the ventral (future temporal) domain of the inner eye vesicle leaflet (arrows) and neural retina (outer leaflet), but is absent from the dorsal (future nasal), presumptive RPE progenitor (pRPE) domain (bar; D). plxna1a mRNA is present in the pRPE domain (bar) and faintly in the dorsal neural retina (asterisk), but absent from the ventral inner and outer eye vesicle leaflet (E). plxna1b is expressed in scattered cells of the outer leaflet of the developing optic cup (F). G, H, Eye vesicle (viewed dorsally) at the 16 ss processed for whole-mount ISH with antisense riboprobes for plxna1a (arrows in G) or the RPE marker, pmel1a (H). I, Plxna1-like immunoreactivity is present at the 16 ss in the pRPE domain (arrows). J-L, Schematics of the eye vesicle (J), the early embryo axes with respect to the eye (K), and optic cup morphogenesis (L); note that as the eye rotates alongside brain development, early ventral retina tissue (pink) becomes mature temporal tissue. M, Schematic of plxna1a and sema6d mRNA expression in the 16-ss eye vesicle. Scale bars: 300 mm (A) and 50 mm (D). br: brain, D: dorsal, e: eye, il: inner leaflet, N: nasal, nk: neural keel, nr: neural retina, ol: outer leaflet, os: optic stalk, pRPE: presumptive RPE, RPE: retinal pigment epithelium, T: temporal, V: ventral, ve: ventricle.
Research Article: New Research presumptive RPE progenitor domain of the dorsal inner leaflet of the eye vesicle ( Fig. 1E,I), with faint signal in the dorsal (future nasal) neural retina (Fig. 1E, asterisk). By 24 hpf, plxna1a transcript is only weakly expressed by the RPE (data not shown). RNA ISH for the plxna1a homolog, plxna1b, revealed minimal expression at the 14 ss (data not shown), and transcript in the presumptive neural retina in a spattering of cells at 18 hpf (Fig. 1C,F). In summary, the expression of sema6d and plxna1a in complementary domains over the period of optic cup morphogenesis ( Fig. 1M) suggests they may interact to control the cell movements that result in the production of an optic cup.

Temporal optic cup shows defects in sema6d mutants
To study Sema6d loss-of-function in vivo we used CRISPR/Cas9 gene editing techniques (Hwang et al., 2013;Auer et al., 2014;Gagnon et al., 2014; to generate two mutant fish lines. Suitable sgRNAs could not be targeted to exons upstream of the Sema domain, and so we generated sgRNAs to target exons 6 and 10. The sema6d exon6 (sema6d ca302 ) and exon10 (sema6d ca303 ) mutant lines are predicted to generate truncated proteins of 137 and 307 amino acids in length, respectively. The truncated proteins would both lack the transmembrane and intracellular domains ( Fig.  2A), and contain an incomplete Sema domain, which is required in whole for homodimerization and binding of Plxna1 to induce forward signaling (Janssen et al., 2010;Nogi et al., 2010).
Transcript for sema6d is not detected readily until the 14 ss, and thus, we expected that early embryogenesis and eye vesicle development should be unaffected in sema6d mutants. To verify that this was indeed the case, we assayed early development of the sema6d mutant embryo and eyes. fibroblast growth factor-8 (fgf8a) expression at the 10 ss in the anterior forebrain, at the midbrain-hindbrain boundary, in the somites and in the tail bud argues that the general body plan is normal in sema6d mutants (Fig. 3A,B). When assessed in a blinded manner, the morphology, size (N = 2; sema6d ex6 1/1 185.8 6 11.9 mm, n = 12; sema6d ca302 182.0 6 10.8 mm, n = 12; p c = 0.417, unpaired t test, df = 22; Fig. 3K), and patterning of the early eye vesicles were also similar in sema6d mutants and their wild-type siblings ( Fig. 3C-H). Bilateral eye vesicles evaginated from the forebrain and had elongated. Further, whole-mount ISH at the 12 ss revealed proper early eye patterning had occurred in both sema6d mutants and wild-type siblings; dorsal (future nasal) tissue was present and appropriately labeled with foxg1a (Picker et al., 2009;Hernández-Bejarano et al., 2015), ventral (future temporal) tissue likewise with foxd1 (Picker et al., 2009;Hernández-Bejarano et al., 2015), and anterior (future ventral) tissue with vax2 (French et al., 2009;Fig. 3C-H). RPE progenitors were specified, as assayed by tfec expression ( Fig. 3I,J), and the RPE domain expanded anterowards in an appropriate manner (N = 2; wild-type 60.9 6 5.5% coverage of eye vesicle by RPE, n = 12; sema6d ca302 58.8 6 6.3%, n = 12; p d = 0.381, unpaired t test, df = 22; Fig. 3L; Cechmanek and McFarlane, 2017). Together, these data indicate early molecular and morphologic development of the eyes is normal in the sema6d mutants.
sema6d transcript appears just before optic cup morphogenesis. Therefore, we next asked whether Sema6d was involved in optic cup formation after the 14 ss. First, we assayed embryos at 24 hpf to see whether patterning of the optic cup occurred normally in the mutants. We processed embryos for RNA ISH against markers of the nasal (early dorsal; foxg1a), dorsal (early posterior; tbx5a; French et al., 2009), temporal (early ventral; foxd1), and ventral (early anterior; vax2) mature neural retina (Fig. 4). Whole-mount eyes were imaged and assessed in a blinded manner. foxg1a and tbx5a were expressed similarly in the nasal and dorsal eye, respectively, of mutant embryos and their wild-type siblings (Fig. 4A,B,D,E), as indicated by the similarity in the angles that span the expression domains across genotypes (foxg1a angle: sema6d ex6 1/1 vs sema6d ca302 p e = 0.357, unpaired t test, tbx5a angle: sema6d ex6 1/1 vs sema6d ca302 p f = 0.175, unpaired t test, sema6d ex6 1/1 vs sema6d ca303 p g = 0.905, unpaired t test; Fig. 4C,F). In contrast, the gross appearance of foxd1 and vax2 staining suggested a defect in development of the temporal eye ( Fig. 4G-L): temporal domains of expression were aberrant in size (red bars), shape, and pattern (black arrowheads). Importantly, defects presented in both mutant alleles and in the tissue that expresses sema6d during optic cup morphogenesis; the ventral region of the early eye vesicle that contains the neural progenitors that go on to populate the temporal neural retina.
To better understand the defect in the temporal eye at 24 hpf, 7-mm-thick transverse sections were made of wild-type and mutant whole-mount foxd11 and vax21 embryos (Fig. 4M,N,P,Q). During optic cup morphogenesis, the neural retinal progenitors of the ventral inner leaflet of the eye vesicle move around the distal rim of the eye into the outer leaflet to become the temporal neural retina of the 24-hpf optic cup (Picker et al., 2009;Kwan et al., 2012;Fig. 1K,L). Concurrently, the RPE cells of the dorsal eye vesicle are stretched in both ventral and dorsal directions and end up as a single cell layer thick epithelium that lines the back of the neural retina (Heermann et al., 2015;Cechmanek and McFarlane, 2017). In wild-type siblings, this process of morphogenesis occurred normally, so that at 24 hpf the foxd11 and vax21 domains sat within the neural retina ( Fig. 4M,P, bar). Further, labeling of these markers was absent from the thinned inner leaflet, now comprising the RPE. In mutant eyes (Fig. 4N,Q), fewer foxd11 and vax21 cells sat within the neural retina of the outer leaflet (bars), and a significant proportion of cells aberrantly populated the inner leaflet (arrowheads). To quantitate the temporal eye phenotype, we compared the width of the vax2 ISH domain in the temporal versus nasal eye, in lateral whole-mount images (  . sema6d mutants do not display gross morphologic defects. A, Schematic of the two sema6d mutant alleles; sem-a6d ca302 mutants contain a five-base pair deletion and two-base pair mismatch in exon6, and sema6d ca303 mutants contain a 16-base pair insertion in exon10, both predicted to generate proteins that truncate prematurely within the SEMA domain and result in the loss of the transmembrane domain (TMD) and the intracellular domain (ICD). B-E, Lateral views of wholemount embryos processed for bhlhe40 ISH, revealing normal gross embryonic morphology and somite labeling in both mutant alleles with respect to their wild-type siblings at 24 hpf. F, G, Eye area measured from lateral images of sema6d mutants and their respective wild-type siblings at 24 hpf [unpaired t test, df = 26 (F), df = 21 (G); error bars are SD]. Scale bar: 200 mm. the temporal vax2 signal was found aberrantly in the inner eye vesicle leaflet of mutants (Fig. 4K,L, arrowheads), hidden behind the portion of the vax2 domain in the outer leaflet (Fig. 4K,L, red bar), so that in whole mount the outer leaflet vax21 temporal domain was often more narrow than that observed in wild-type. As such, the ratio of the temporal versus nasal width of the vax2 domain was decreased significantly in sema6d mutant embryos as compared with wild-type siblings (N = 2; ex6 1/1 0.85 6 0.08, n = 11; sema6d ca302 0.56 6 0.1, n = 18, p h , 0.0001; sema6d ca303 0.62 6 0.1, n = 27; p , 0.0001, one-way ANOVA, Dunnett's multiple comparisons test, df = 53; Fig.  4O). The specificity of the defect to temporal (early ventral) and not nasal (early dorsal) eye tissue agrees with the localization of sema6d mRNA to neural progenitors of only the ventral domain of the early eye vesicle.

RPE temporal expansion is disrupted in sema6d mutant embryos
Given the disruption of the normal redistribution of neural eye progenitors from the inner leaflet of the eye vesicle to the temporal neural retina, we asked whether the neighboring RPE cells were affected in the sema6d mutant by assaying for expression of the RPE markers bhlhe40 and tfec (Cechmanek and McFarlane, 2017). In the eyes of whole-mount wild-type embryos at 24 hpf, bhlhe401 RPE cells abutted the lens (Fig. 5A). In both mutant alleles, however, bhlhe40 signal was often seen through the transparent zebrafish lens where in wild-type eyes signal was not observed (Fig. 5B,C, arrowhead). The difference between wild-type and sema6d mutants was particularly evident in transverse sections (Fig. 5A'-C',F). In both wild-type and sema6d mutant embryos, bhlhe401 RPE cells stretched normally toward the nasal lens (black arrows). In sections of the occasional mutant (11%, n = 18 eyes), we found as expected, a complete failure of bhlhe401 RPE cells to extend around the temporal optic continued Figure 3. sema6d mutants display normal eye vesicle development and patterning before optic cup morphogenesis. Wholemount RNA ISH for patterning genes indicates early development occurs normally in sema6d mutants (B, D, F, H, J) relative to wild-type siblings (A, C, E, G, I). In the bottom right of panels are the number of embryos of the total analyzed that exhibited the normal wild-type expression pattern. A, B, fgf8a expression in 10-ss embryos viewed laterally. C-H, Dorsal whole-mount images of the eye vesicles (dashed outlines) and neural keel of 12 ss processed by whole-mount ISH for the expression of early patterning genes. C, D, Early dorsal (future nasal) tissue with foxg1a. E, F, Early ventral (future temporal) tissue with foxd1. G, H, Early anterior (future ventral) tissue with vax2. I, J, RPE progenitors and neural crest cells express tfec at the appropriate time in wild-type siblings and sema6d mutants. cup (early ventral), with the bhlhe40 ISH label present only within the inner eye vesicle leaflet (Fig. 5F, red arrow). More common (72%, n = 18 eyes), however, was the unexpected presence of ectopic bhlhe401 cells present in the distal portion of the temporal neural retina of mutants (Fig. 5B',C', arrowheads). Importantly, these phenotypes were present in both mutant alleles (wild-type, normal RPE morphogenesis n = 12/12; aberrant morphogenesis, sema6d ca302 n = 12/34 and sema6d ca303 n = 5/22). Unlike the expanded RPE cells of wild-type eyes, in both sema6d mutant alleles RPE cells adjacent to the temporal optic cup were often cuboidal in appearance (compare Fig. 5A'-C',F', red asterisk). We crossed the sema6d ca302 line onto a Tg(tfec:EGFP) reporter line (Miesfeld et al., 2015) that we regenerated, to examine EGFP1 RPE cells in transverse sections at 24 hpf (Fig. 5D,E). In agreement with the bhlhe40 data, a thin monolayer RPE backed the central retina of both wild-type and mutant siblings (Fig.  5D,E, asterisk). Extension of the RPE around the temporal (early ventral) retina, however, only occurred in wild-type eyes (Fig. 5D, arrows). In the mutants, weak EGFP (tfec) label was present ectopically within the neural retina (Fig.  5E, arrowheads), and cells within the inner leaflet were often cuboidal in shape (Fig. 5E, arrows).
The ectopic bhlhe401/tfec1 cells within the distal temporal neural retina indicated that progenitors that expressed the temporal retinal markers vax2 and foxd1 also expressed an RPE marker (compare Figs. 4N,Q and 5B', C',E). While initially surprising, we found that it was normal for temporal neural retina progenitors to transiently co-express neural and RPE markers in wild-type eyes younger by a few hours (Fig. 5G-G'', arrowheads). Indeed, in both wild-type and sema6d ca302 Tg(tfec:EGFP) 22-hpf eyes, foxd11 temporal progenitors were weakly EGFP-positive (Fig. 5G,H, arrowheads), indicating that these cells either expressed tfec, or recently turned off tfec expression. By 24 hpf, when optic cup morphogenesis was complete, wild-type siblings did not express bhlhe40 or the tfec: EGFP reporter in cells within the temporal domain of the neural retina, instead only the flattened RPE cells expressed label (Fig. 5D). The wild-type data suggest that a population of phenotypically plastic neural retina Research Article: New Research progenitors, at the border between temporal neural and RPE domains, transiently express both RPE and temporal neural retina markers, before halting expression of RPErelated transcription factors. In mutants, these foxd11/ vax21 neural progenitors (Fig. 4N,Q), both within the outer leaflet and those aberrantly located in the inner leaflet, appeared to fail to turn off RPE specific transcription factors (Fig. 5B',E, arrowheads). The result was that the temporal neural progenitors continued to co-express RPE markers inappropriately at 24 hpf.
Potentially as a consequence of cells in the temporal neural retina of sema6d mutants inappropriately expressing RPE-specific genes, the morphology of these cells was affected. sema6d ca302 and their respective sema6d ex6 1/1 siblings were crossed onto a Tg(rx3:GFP) background, where initially all eye progenitors express GFP.
By 24 hpf the cells in the temporal eye expressed higher levels of GFP, allowing us to look at cell morphology. Eye sections at 24 hpf were immunolabeled with an antibody that recognizes the apically located tight junction associated protein atypical protein kinase C (aPKC; Suzuki et al., 2001; Fig. 6A,B). In transverse microtome sections, the ventricle of the temporal optic cup in affected sema6d mutants did not exhibit the "zippering-up" that occurs in wild-type siblings (compare Fig. 6A,B, arrowheads). The aPKC label of the neural retina/RPE border in the nasal eye, however, appeared unaffected. Also of note was that while wild-type embryos had GFP1 cells in the temporal neural retina with typical elongated morphologies and radial alignments (Fig. 6C,E,E'), the progenitors in this domain in sema6d ca302 eyes appeared disorganized and were not all aligned radially (Fig. 6D,F,F'). Indeed, the Figure 5. Disrupted RPE morphogenesis in sema6d mutants. A-C, Morphogenesis of the RPE is disrupted in sema6d ca302 and sem-a6d ca303 eyes as compared with wild-type, with mutants displaying ectopic bhlhe40 staining apparent through the transparent lens (white arrowhead in B, C). In the bottom right of panels are the number of embryos of the total analyzed that exhibited either a WT (A) or a disrupted (B, C) expression pattern. A'-C', F, Transverse sections reveal full expansion of the bhlhe401 RPE to abut the nasal lens (black arrows), while the RPE of the temporal eye fails to reach the lens in mutants (compare red arrows). RPE cells are expanded in the inner leaflet of wild-type embryos, but cuboidal in mutants (compare red asterisks). In most mutants the bhlhe401 RPE not only failed to abut the lens, but bhlhe40 was ectopically expressed in the lateral portion of the temporal neural retina (white arrowheads in B', C'), while in some sema6d mutants, bhlhe40 expression was restricted entirely to the inner eye vesicle leaflet (red arrow in F). D, E, Wild-type (D) and sema6d ca302 mutant (E) embryos bred on a Tg(tfec:EGFP) background to label RPE progenitors. Mutants display ectopic EGFP expression in the temporal neural retina (arrowheads in E) and EGFP1 RPE cells in the temporal (early ventral) inner leaflet are cuboidal and not extended as in wild-type (compare arrows in D, E). G, H, Transverse sections of 22hpf eyes of wild-type (G-G'') and sema6d ca302 (H-H'') mutant embryos on a Tg(tfec:EGFP) background that were processed for GFP immunohistochemistry (G, H) and RNA ISH for the temporal neural retina marker foxd1 (G', H'). Blended images (G'', H''). EGFP1 cells that co-express foxd1 are present in the distal temporal neural retina of both wild-type and mutant eyes (arrowheads in G, G'', H, H''). A few foxd11 cells in the wild-type inner leaflet (asterisk in G') have not yet moved around the distal rim, with many more present in the mutant inner leaflet (asterisk in H'). Scale bars: 50 mm (A, A'). D: dorsal, il: inner leaflet, L: lens, N: nasal, ol: outer leaflet, T: temporal, V: ventral.

Temporal eye defects caused by loss of Sema6d signaling are because of aberrant progenitor cell movements
To understand the cell behaviours that lead to temporal disorganization in mutant eyes, we embedded Tg(rx3: GFP) and Tg(tfec:EGFP) wild-type and sema6d mutant embryos in low melting temperature agarose and imaged them over the period of optic cup morphogenesis by confocal microscopy (approximately from the 14 ss to 23 hpf). We used the Tg(rx3:GFP) background to label neural retina progenitors, while on the Tg(tfec:EGFP) background we identified presumptive RPE cells as the brightest EGFP1 cells, which we found via ISH analysis did not express foxd1 or vax2. For these preparations, embryos were embedded dorsally on their heads and z-sections were acquired in an anteroposterior plane (future nasaltemporal plane) and z-stacks through the central retina were used to manually track cells over time traveling around the distal rim of the optic cup. Of note, like-cells (RPE or temporal neural retina) kept their nearest neighbor relationships over the period of optic cup morphogenesis. Figure 7 shows an example of a sema6d ex6 1/1 Tg(rx3: GFP) eye vesicle undergoing morphogenesis, where cells at the ventral-most point of the early eye vesicle can clearly be seen to migrate around the rim of the optic cup ( Fig. 7A,C, pink asterisk follows one cell). Cells in mutant eyes began to move toward the distal rim of the optic cup (t = 120' for example), but by 19-to 20-hpf GFP1 (rx31) neural progenitors (based on location in the ventral inner eye vesicle) stalled and showed little lateral progress. The cells could be seen to bulge out of the ventral (future temporal) inner leaflet of the eye vesicle (Fig. 7B,D, follow blue asterisk that labels one cell), and "clump" together ( Fig.  7D).
To quantitate these data, we measured the average speed of movement of the cells in the inner leaflet over the period of optic cup morphogenesis in individual sema6d ex6 1/1 and sema6d mutant eyes. All embryos imaged in this study were genotyped postimaging. We first assessed the movement of presumptive neural progenitors (four to seven cells per embryo) of the ventral (future temporal) inner leaflet in the Tg(rx3:GFP) background. Presumptive neural progenitors in 17-hpf sema6d mutant eyes moved at ;62% of the speed observed in controls [sema6d ex6 1/1 , 21.8 6 1.4 mm/h (SEM), n = 4 embryos; sema6d ca302 13.7 6 1.6 mm/h, n = 5 embryos, p k = 0.032, unpaired, Mann-Whitney U test, df = 7; Fig. 7E]. We then asked whether similar effects were observed with the movements of RPE progenitors (two to five cells per embryo) by assaying their behavior in a Tg(tfec:EGFP) background. Note that only two homozygotes were identified, thus we focused our analysis on RPE progenitors in sema6d heterozygous eyes (n = 7 embryos; Fig. 7F), and include the data we obtained by the tracking of cells in the two homozygous eyes for comparison purposes. We found initially at 16 hpf (14 ss) that RPE cells in both heterozygous [22.0 6 3.4 (SEM) mm/h, n = 5 embryos] and homozygous (27.4 6 3 mm/h, n = 2) eyes moved toward the distal rim at speeds comparable to those of sema6d ex6 1/ 1 cells (27.3 6 1.7 mm/h, n = 7 embryos; Fig. 7F). The RPE Figure 7. Ventral inner leaflet cells fail to move appropriately around the distal rim during optic cup morphogenesis. A, B, Confocal optical sections, and their corresponding schematics, acquired from the nasal/temporal plane between the 14 ss and 23 hpf of wild-type and sema6d mutant Tg(rx3:GFP) eyes (labels all eye progenitors). Dotted yellow line(s) indicates the separation between the inner and outer leaflets. C, D, Higher magnification of inset boxes (early ventral/future temporal retina) in A, B, respectively. In wild type (A), ventral progenitors move around the distal rim of the optic cup and come to reside in the temporal neural retina at 23 hpf. In contrast, sema6d mutants display temporal defects (follow arrow in B). While a ventral inner leaflet cells in wild type (C) moves around the distal rim (pink asterisk), a corresponding cell in the sema6d mutant (D) begins to move toward the distal rim, but stalls and, together with other progenitors, protrudes from the inner leaflet (blue asterisk). E, F, Mean of the speed of individual cells in each wild-type and sema6d mutant embryo tracked over the period of optic cup morphogenesis. Tracked were ventral neural progenitors located at the distal eye vesicle tip in the Tg(rx3:GFP) background (E) and brightly EGFP labeled presumptive RPE cells in the Tg(tfec:EGFP) background (F). In both reporter lines, sema6d mutant cells move more slowly than their wild-type counterparts, and stall at ;18-20 hpf. ns are number of embryos assayed (2-7 cells/embryo), and error bars are SEM. Mean speeds at individual time points were compared statistically between wild-type and heterozygous embryos by a Mann-Whitney U test (p , 0.05; E, F). A: anterior, il: inner leaflet, L: lens, N: nasal, nr: neural retina, ol: outer leaflet, P: posterior, pRPE: presumptive RPE, RPE: retinal pigment epithelium, T: temporal, ve: ventricle. progenitors in the heterozygote eyes slowed significantly to around half that of wild-type as optic cup morphogenesis proceeded (18 hpf, sema6d ex6 1/1 22.4 6 2.5 mm/h, n = 8 eyes; sema6d 1/ca302 12.7 6 1.7 mm/h, n = 7 embryos; p l = 0.029, Mann-Whitney U test, df = 13). A similar slowing was observed in the homozygous mutant (Fig. 7F). These data argue that in sema6d mutants, cells of the temporal inner leaflet of the eye vesicle initially move toward the rim of the optic cup and then slow dramatically and fail to move around the distal rim.
Plxna1 is a candidate receptor for Sema6d signaling during optic cup morphogenesis Given the expression of plxna1a in RPE progenitors in the domain adjacent to early ventral eye vesicle neural progenitors, we asked whether Plxna1 serves as a receptor for Sema6d in optic cup morphogenesis. To do so, we perturbed Plxna1 function by injecting an antisense MO against the exon2-intron2 splice site of plxna1a (NCBI reference sequence XM_017353686.2) into wild-type TL embryos at the one cell stage (Fig. 8A). To verify the efficacy of the morpholino, we collected mRNA at 24 hpf and performed RT-PCR to confirm mis-splicing of plxna1a transcript (Fig. 8A). Perturbing Plxna1 signaling did not disrupt eye vesicle formation or elongation (Fig. 8C), RPE morphogenesis (Fig. 8D), or patterning of the early eye vesicle (data not shown). Further, the nasal retina was appropriately patterned at 24 hpf, as assessed by vsx1 (Fig.  8E,F,K) and foxg1a (data not shown) expression. However, at 24 hpf, there was a strong phenocopy of the temporal retina defects observed with Sema6d loss-offunction: vax21 (N = 2, 1/12 control embryos, 19/23 morphants; Fig. 8G,H) and foxd11 (N = 2, 0/15 control embryos, 12/16 morphants; Fig. 8I,J) tissue failed to move completely around the distal rim of the temporal optic cup in the majority of morphants as compared with controls, and remained partially in the inner leaflet (Fig. 8J, red asterisk). We quantified this phenotype by assessing the ratio of the width of the temporal to nasal vax2 domain, and found a significant decrease in this ratio in the plxna1a morphant embryos (N = 2; control 0.95 6 0.1, n = 13; plxna1MO 0.61 6 0.2, n = 23; unpaired t test, p m , 0.0001, df = 34; Fig. 8L). Further, as assessed in wholemount (data not shown), ectopic bhlhe40 expression within progenitors of the distal portion of the temporal neural retina was observed in 18% (N = 2, n = 17) of plxna1a morphants, as compared with 0% in control injections (N = 2, n = 11). The bhlhe40 labeling phenotype was more readily assessed in transverse section, where RPE cells in the temporal region of morphant eyes were cuboidal and did not elongate properly (compare Fig. 8M,N, arrows), and ectopic bhlhe40 label was present in the temporal neural retina (Fig. 8N, arrowheads). A second method of plxna1a mRNA knock-down via CRISPRi Larson et al., 2013;Qi et al., 2013) corroborated these findings. A sgRNA targeting exon-5 of plxna1a was co-injected with dead-cas9 mRNA at the one cell stage. mRNA was collected from control and CRISPRi-injected embryos at 24 hpf, and RT-PCR verified partial knock down of plxna1a mRNA transcript levels in the latter (Fig. 8B).
We observed temporal eye defects at 24 hpf in a portion of the plxna1a CRISPRi-injected embryos. The ratio of the temporal to nasal vax2 expression domain width was significantly smaller in CRISPRi embryos than control (N = 1; control 0.98 6 0.05, n = 5; CRISPRi 0.72 6 0.2, n = 7; unpaired t test, p n = 0.012, df = 10), indicative of defects in morphogenesis of the temporal (early ventral) eye. In support, in CRISPRi-injected embryos (19%, N = 2, n = 21) bhlhe401 RPE progenitors in the inner leaflet were cuboidal in shape (compare Fig. 8O,P, arrows) and aberrant expression of bhlhe40 was present in the distal temporal neural retina (Fig. 8P, arrowheads), phenotypes not observed in any control-injected embryo (N = 2, n = 15). The fact that perturbing Plxna1a function produces similar eye defects as those observed with Sema6d loss-of-function is supportive of Plxna1a acting as a receptor for Sema6d signaling during optic cup morphogenesis.

Optic cup morphogenesis may depend on Sema6d reverse signaling
Sema6d can signal bi-directionally, acting as either a ligand for Plxna1 or by transducing a Plxna1 signal. To determine the mode by which Plxna1 and Sema6d expressing cells communicate, we first asked whether forward signaling was possible by determining whether Sema6d is repulsive for Plxna1-expressing RPE progenitors. We explanted eye vesicles from Tg(tfec:EGFP) sema6d ex6 1/1 embryos (one/embryo), at a time when RPE cells express plxna1a mRNA (16-18 ss), and grew them in culture for 20 h in the presence or absence of soluble recombinant mouse Sema6d (Sema6d-Fc) that should be able to activate forward signaling by the Plxna1 receptor in RPE cells (Fig. 9E), as is suggested for chick endocardial cells (Toyofuku et al., 2004b). Cultured eye vesicle explants generally form optic cups and undergo similar morphogenesis as in vivo (Fig. 9A,B). The Tg(tfec: EGFP) label was used to identify RPE progenitors that leave the eye explant when cultured ( Fig. 9A'',B'', arrows). Significantly fewer EGFP1 RPE cells exited the eye onto the Fibronectin substrate in the presence of Sema6d-Fc [N = 3; Sema6d-Fc(-) 10.8 6 3.7 cells, n = 9 explants; Sema6d-Fc(1) 4.8 6 2.6 cells, n = 9 explants; unpaired t test, p o = 0.001, df = 16; Fig. 9C]. These data suggest that similar to what happens for a truncated soluble form of the Sema6d receptor in inhibiting the migration of chick endocardial cells (Toyofuku et al., 2004b), Sema6d-Fc can inhibit the migration of RPE progenitors.
While these data support the idea that Sema6d can forward signal through Plxna1, they do not address by which signaling mode Sem6d participates in optic cup morphogenesis. Thus, we tested the importance of Sema6d reverse signaling in temporal retinal progenitors for optic cup morphogenesis. We rejected taking a molecular approach using a dominant negative Plxna1 receptor because of difficulties in targeting expression of the mutant protein selectively to the RPE so that it would not impact other aspects of early eye development (Ebert et al., 2014). Instead, we took a pharmacological approach, where the timing of the inhibition of Sema6d signaling could be controlled tightly so as to occur just before optic cup morphogenesis. The kinase c-Abl is the only known effector of Sema6d reverse signaling, working downstream of Sema6d in macrophage polarization and embryonic chick myocardial cells (Toyofuku et al., 2004b;Kang et al., 2018).
Immunolabeling for the phosphorylated activated form of c-Abl appeared downregulated in the temporal retina of 18to 20-hpf sema6d mutants (Fig. 9K,L), consistent with the idea that c-Abl works downstream of Sema6d reverse signaling in temporal neural progenitors. To test this idea further, we asked whether inhibition of c-Abl phenocopied the loss of function of Sema6d. Wild-type embryos at the 16-ss stage were grown with or without 100 mM dasatinib, a pharmacological inhibitor of c-Abl (Lombardo et al., 2004), for 7 h until 24 hpf, and then processed by whole-mount ISH for the ventral/temporal domain marker vax2 (Fig. 9F,G). In a significant number of dasatinib-treated embryos there were defects in temporal optic cup morphogenesis (Fig. 9G). We quantitated this effect by comparing the ratio of the width of the temporal versus the nasal retina as measured from lateral images of the vax2 ISH processed embryos (Fig. 9J). Dasatinibtreated embryos exhibited a significantly smaller ratio than wild-type controls (N = 3; wild-type 0.85 6 0.11, n = 14 embryos; wild-type1dasatinib 0.67 6 0.12, n = 17, p p = 0.0003; one-way ANOVA, Dunnett's multiple comparisons test, df = 59) because of the failure of the temporal retinal progenitors to move completely around the distal rim of the optic cup, resulting in a smaller portion of the vax2 domain in the temporal outer leaflet (Fig. 9G). This phenotype is what is observed in sema6d mutants (sema6d ca302 , 0.65 6 0.14, n = 18, p q , 0.0001, one-way ANOVA, Dunnett's multiple comparisons test, df = 59; Fig. 9H). These data support the idea that c-Abl works downstream of Sema6d in temporal retinal progenitors to promote eye morphogenesis. To further test this possibility, we performed an epistasis experiment, and treated 16 ss sema6d ca302 embryos with dasatinib, with the idea being that if c-Abl works downstream of Sema6d, then c-Abl inhibition should have no additional impact on optic cup morphogenesis than the loss of Sema6d signaling itself. Indeed, this is what we observed, with a similar ratio of temporal to nasal eye width observed in sema6d ca302 embryos treated with and without dasatinib (sema6d ca302 , 0.65 6 0.14, n = 18; sema6d ca302 1dasatinib, 0.60 6 0.13, n = 14; df = 59; Fig. 9H-J).
While c-Abl is the only kinase reported to date to work downstream of Sema6d (Toyofuku et al., 2004b;Kang et al., 2018), dasatinib can inhibit other kinases and so we were unable to rule out the possibility that dasatinib impacts forward signaling through a kinase active downstream of Plxna1 (Lombardo et al., 2004). To further explore the participation of Sema6d reverse signaling, we revisited the Sema6d-Fc treatment of wild-type eye explant cultures. In the developing chick heart, Sema6d mediates both forward and reverse signaling (Toyofuku et al., 2004b). In this system, misexpression of soluble Sema6d in the embryonic heart is reported to both promote forward signaling and block reverse signaling (Toyofuku et al., 2004a,b). As such, we expected that Sema6d-Fc should bind the Plxna1a of RPE cells and promote forward signaling via activation of the receptor, but also block reverse signaling, in that the Sema6d-Fc occupied RPE-expressed Plxna1a receptor would no longer be available to interact with Sema6d in temporal neural progenitors. Note that while we assayed above the behavior of RPE cells alone (Fig. 9C), here, we assayed optic cup morphogenesis, which depends on the movements of temporal neural retinal progenitors. We reasoned that if optic cup morphogenesis is controlled by Sema6d activating Plxna1 forward signaling, then Plxna1a signaling would be activated by both temporal progenitor Sema6d and the soluble Sema6d-Fc and optic cup morphogenesis would occur normally. If instead reverse signaling is key, by binding to the Plxna1a of RPE progenitors the soluble Sema6d-Fc would act as an antagonist and block the ability of Plxna1a to interact with and activate endogenous Sema6d reverse signaling. As such, optic cup morphogenesis would be disrupted. As described above, explanted eyes from Tg(tfec:EGFP) wild-type 16-18 ss fish were treated with and without soluble Sema6d-Fc. Eye buds from Tg(tfec:EGFP) embryos injected at the one cell stage with the plxna1a morpholino were used as a positive control, as we expected optic cup morphogenesis defects with impaired Plxna1a signaling. Wild-type explants underwent optic cup morphogenesis to produce an optic cup surrounding the lens, and with GFP1 RPE cells that expanded and covered almost the entire optic cup, as they would have in vivo (Fig. 9A'). The degree of optic cup covered by the GFP1 RPE was measured, to represent the extent to which optic cup morphogenesis proceeded to completion (Fig. 9D). This measure was reduced significantly in the plxna1a morpholino-treated explants, agreeing with the in vivo data indicating that Plxna1a function is required for normal optic cup morphogenesis (N =3; wild type, 88.4 6 9.1%, n = 9 explants; plxna1a MO 47.0 6 15.4%, n = 9; p r , 0.0001, one-way ANOVA, Dunnett's multiple comparisons test, df = 23). Treatment of wild-type explants with Sema6d-Fc produced a similar eye morphogenesis defect, notably with the RPE failing to cover the temporal optic cup (58.6 6 16%, n = 8 explants, p s = 0.0004; Fig. 9B',D). These data, alongside the c-Abl downregulation in temporal progenitors of sema6d mutants, support the idea that Sema6d reverse signaling promotes optic cup morphogenesis.

Discussion
Little is known about the cellular and molecular interactions that drive optic cup morphogenesis. We reveal an important role for Sema6d-mediated communication between retinal and RPE progenitors in the flow of the inner leaflet of the eye vesicle around the distal rim of the invaginating optic cup. Specifically, we find sema6d mRNA in ventral neural retina progenitors (future temporal) of the inner leaflet, while the adjacent RPE progenitors of the dorsal inner leaflet express Plxna1a mRNA. Loss of Sema6d signaling, either by CRISPR/Cas9 mutagenesis of sema6d, transient Plxna1a knock-down, or c-Abl inhibition, impairs the movements of both ventral retinal cells and their RPE neighbors. Our study identifies for the first time that interaction of RPE and neural retina progenitors, via a Sema6d-dependent mechanism, is a key modulator of optic cup morphogenesis.
During optic cup morphogenesis, sema6d1 neural progenitors of the ventral inner vesicle migrate around the distal rim of the vesicle to form the mature temporal neural retina, and plxna1a1 RPE cells elongate and come to comprise the entire inner leaflet of the optic cup. When Sema6d signaling is impaired, these processes are disrupted. Ventral foxd1/vax21 progenitors accumulate in a disorganized fashion at the lateral edge, and significant numbers of these progenitors fail to move normally into the retina proper. In an associated manner, RPE cells of the inner leaflet do not elongate and stretch to the temporal edge of the lens. Time-lapse imaging reveals that morphogenesis initiates properly, with neural progenitors moving toward the rim and RPE cells elongating. Rim movement, however, proceeds more slowly and then stalls as cells of the ventral inner leaflet clump and form a disorganized bulge.
The morphogenesis defects appear specific to Sema6d and potentially involve a known Sema6d partner, Plxna1a (Toyofuku et al., 2004a,b). First, two distinct mutant alleles exhibit the same phenotype, which is restricted to the eye domains that express sema6d and plxna1a: defects are seen in RPE expansion and ventral eye vesicle morphogenesis, but not morphogenesis of dorsal (future nasal) eye. Second, knock down of Plxna1a, via two distinct means, produces temporal eye phenotypes similar to those of sema6d mutants.
Cellular interactions with Integrin of the basal lamina drive temporal progenitor migration during optic cup morphogenesis (Sidhaye and Norden, 2017). Our data suggest that Sema6d provides an additional mechanism during morphogenesis, and point to Plxna1a in the RPE as the interacting receptor. The fact that loss of Sema6d or Plxna1a alone produce similar morphogenesis defects, argues that RPE and retinal progenitors have to interact for normal morphogenesis. Thus, we propose that RPE expansion does not depend entirely on the cells being stretched passively by the rim movement of retinal progenitors. Instead, RPE and neural progenitors must signal to each other, through a cell-cell contact-mediated process. Of note, the contraction of RPE cells may also contribute to morphogenesis by pushing the neural progenitors around the rim.
Sema6d, and other transmembrane Sema6s, mediate forward signaling through a Plxn, and reverse signaling through their intracellular domain (for review, see Casazza et al., 2007). We expect that no reverse signaling occurs in mutants, even if protein is made, in that mutant Sema6d is predicted to lack transmembrane and intracellular domains. Forward signaling is likely also impacted in the mutants, because even if truncated Sema6d is present it would lack the intact Sema domain for binding Plxna (Janssen et al., 2010;Nogi et al., 2010). Forward signaling could normally be active in RPE progenitors. Indeed, a parsimonious explanation for our culture data is that Sema6d-Fc inhibits RPE migration via activation of Plxna1 forward signaling, although alternative explanations are possible. Yet, our data speak to an important role for Sema6d reverse signaling in optic cup morphogenesis. Soluble Sema6d was shown to block events in the embryonic chick heart that depend on Sema6d reverse signaling, while promoting those that require forward signaling (Toyofuku et al., 2004a,b). Based on these observations, our Sema6d-Fc data, where optic cup morphogenesis is inhibited, argue that morphogenesis depends on Sema6d reverse signaling. In agreement, the activation of a known Sema6d reverse signaling downstream mediator, c-Abl, appears downregulated in mutant temporal retinal progenitors, and a pharmacological inhibitor of this kinase produces a morphogenesis defect similar to that of sema6d mutants. Thus, our data support a role for reverse signaling in eye morphogenesis but do not exclude the involvement of forward signaling.
One could imagine a simple mechanism by which Plxna1a1 RPE cells reverse signal to Sema6d-expressing neural progenitors and help push a front of ventral neural tissue around the distal rim (Fig. 9M). An alternate possibility is a "tissue sliding" model (Fig. 9N). Cells in the inner and outer leaflets slide past each other in opposite directions during optic cup morphogenesis; RPE progenitors move laterally, while neural progenitors travel medially on entry into the retina (Kwan et al., 2012;Heermann et al., 2015). A sliding of the anterior mesoderm and overlying ectoderm occurs during Xenopus gastrulation (for review, see Hammerschmidt and Wedlich, 2008), with Eph-ephrin repulsion preventing the two tissues from adhering tightly (Park et al., 2011). Similarly, Sema6d-Plxna1a repulsion could allow the inner and outer eye vesicle leaflets to move past one another, with repulsion preventing cells in opposing leaflets from sticking to each other too tightly. Time-lapse imaging indicates that neural and RPE progenitors are in close proximity and likely physically interact, as the basal lamina lines the opposite side of the two epithelia (Sidhaye and Norden, 2017).
A "de-adhesion" mechanism described in Drosophila proposes an alternate means to reduce RPE cell "stickiness," where transmembrane Sema-1a binds to PlxnA and blocks Integrin-extracellular matrix (ECM) interactions (for review, see Yang and Terman, 2013). While, Sema6d activation of RPE Plxna1a might similarly prevent tight adherence of the RPE to Laminin (Fig. 9O; Sidhaye and Norden, 2017), this model seems less likely in that our data speak against a significant role for Sema6d forward signaling. Regardless of the exact mechanism, we propose a model whereby Sema6d promotes the efficient movement of cells via reducing adhesive contacts, either between cells or with the ECM.
Interestingly, our data suggest that the neural retinal cells of the inner leaflet normally transiently express RPE markers. Indeed, at 22 hpf, these cells are positive for both foxd1 and EGFP in a Tg(tfec:EGFP) background. tfec/bhlhe40/EGFP are more weakly expressed by these temporal progenitors than by RPE cells, and expression is quickly lost once the progenitors settle within the temporal retina. The continued expression of RPE markers by temporal progenitors in sema6d mutants, both for progenitors that complete their movement into the retina and for those still stuck in the inner leaflet, could reflect a role of Sema6d signaling in cell fate specification. If true, a defect in the movement of temporal progenitors could occur secondary to a failure to acquire the appropriate identity. While possible, our data and the literature support instead that dys-regulated RPE gene expression occurs secondary to the mislocalization of temporal cells. First, Sema6s are known as regulators of cell movements (for review, see Battistini and Tamagnone, 2016;Alto and Terman, 2018), with no reports to our knowledge of direct roles in cell fate specification. Second, wild-type temporal progenitors also exhibit a mixed expression of RPE and neural markers, but lose this expression once they settle in the retina. Thus, marker co-expression is a feature of normal eye development, and not unique to the mutant background. Instead, we propose that on arrival into the retina, temporal retinal progenitors either receive an extracellular signal to inhibit RPE gene expression, or lose a signal that maintains RPE marker expression. The ectopically located temporal progenitors in the sema6d mutant would fail to receive these positional signals.
Other Sema6s are expressed in the developing eye vesicle. Indeed, Sema6a/Plxna2 are required for vesicle extension (Ebert et al., 2014). Thus, distinct Sema6s may coordinate the different cell movements that drive eye morphogenesis. While multiple Sema6 genes are expressed in the postnatal mouse retina , only loss of SEMA6A affects retinal circuits . No morphogenesis phenotype has been reported in Sema6 mouse mutants. Possibly, SEMA6s act redundantly to each, or other repellent signaling pathways function in their place. Alternatively, roles in eye morphogenesis may have been overlooked, given that the loss of Sema6d in zebrafish does not produce a change in eye size.
The cellular movements involved in eye morphogenesis are characterized (Picker et al., 2009;Kwan et al., 2012;Heermann et al., 2015;Cechmanek and McFarlane, 2017;Sidhaye and Norden, 2017), and we know that acto-myosin contractility and cellular interactions with the ECM are required (Nicolás-Pérez et al., 2016;Sidhaye and Norden, 2017). Here, we provide data to support a need for communication between RPE and retinal progenitors via Sema6d/Plxna1a, and implicating cell-cell contact mediated signaling in optic cup morphogenesis. The expression of SEMA6s in other embryonic tissues (Xu et al., 2000;Toyofuku et al., 2004a;Koestner et al., 2008;Shalom et al., 2019) suggests they may play similar roles in mediating cell-cell and/or cell-ECM interactions that underlie morphogenesis of other embryonic tissues.