Organs are generated from collections of cells that coalesce and remain together as they undergo a series of choreographed movements to give the organ its final shape. We know little about the cellular and molecular mechanisms that regulate tissue cohesion during morphogenesis. Extensive cell movements underlie eye development, starting with the eye field separating to form bilateral vesicles that go on to evaginate from the forebrain. What keeps eye cells together as they undergo morphogenesis and extensive proliferation is unknown. Here, we show that plexina2 (Plxna2), a member of a receptor family best known for its roles in axon and cell guidance, is required alongside the repellent semaphorin 6a (Sema6a) to keep cells integrated within the zebrafish eye vesicle epithelium. sema6a is expressed throughout the eye vesicle, whereas plxna2 is restricted to the ventral vesicle. Knockdown of Plxna2 or Sema6a results in a loss of vesicle integrity, with time-lapse microscopy showing that eye progenitors either fail to enter the evaginating vesicles or delaminate from the eye epithelium. Explant experiments, and rescue of eye vesicle integrity with simultaneous knockdown of sema6a and plxna2, point to an eye-autonomous requirement for Sema6a/Plxna2. We propose a novel, tissue-autonomous mechanism of organ cohesion, with neutralization of repulsion suggested as a means to promote interactions between cells within a tissue domain.
During development, cells are specified to become part of a given tissue, and then sorted so that like cells coalesce and are kept segregated from non-like cells in adjacent tissues. Over time, physical barriers, such as extracellular matrix or epithelium, develop to prevent heterotypic interactions between cells of neighboring domains. A period of extensive change in both cell and tissue shape, however, usually precedes the establishment of a physical barrier. For example, eye formation involves several distinct phases of integrated cell movements (Graw, 2010). Eye precursors are separated away from non-retinal cells through a motile sorting mechanism, ultimately to collect together as a single contiguous eye field (Moore et al., 2004; Cavodeassi et al., 2005). The eye field separates to produce the bilateral optic vesicles, which evaginate from the forebrain. A key question then is how do cells of a specified tissue retain close associations through the complex morphogenetic events of embryogenesis in the absence of a physical barrier?
Two simple molecular mechanisms that result in the formation of borders between like and non-like tissues help explain how tissue integrity is promoted (Krens and Heisenberg, 2011). The first involves selective adhesion, whereby two distinct cell populations express different levels or types of adhesion molecules. Homotypic interactions between like cells then separate and sort the two cell types (Nandadasa et al., 2009). The second mechanism relies on one population of cells expressing a repellent membrane-associated ligand and the other expressing the receptor (Krens and Heisenberg, 2011). For example, transmembrane ephrins and their Eph receptors mediate repellent interactions that prevent mixing between the cells of adjacent hindbrain rhombomeres (Cooke et al., 2005; Kemp et al., 2009). Ephrin/Eph signaling also segregates somitic cells (Watanabe and Takahashi, 2010), ectodermal and mesodermal tissue (Rohani et al., 2011), and anterior neural plate domains (Cavodeassi et al., 2013). Other than these few examples, however, we know little of the molecular mechanisms that mediate tissue cohesion during development.
In this regard, members of the semaphorin (Sema) family and their plexin (Plxn) receptors are interesting (Zhou et al., 2008). They serve as repellents for migrating neurons, axonal growth cones and endothelial cells (Yazdani and Terman, 2006). Furthermore, certain Sema proteins are membrane associated and thus would be ideal candidates to mediate signaling between cells in contact (Zhou et al., 2008). Indeed, Sema6a helps organize chick cardiac development (Toyofuku et al., 2004) and Plx2 controls C. elegans epidermal morphogenesis (Nakao et al., 2007).
In this paper, we provide evidence that Sema6a and Plxna2 promote cohesion of the zebrafish eye primordia, though not primarily through a boundary mechanism between adjacent tissues. Sema6a is expressed by progenitors throughout the eye vesicle, whereas cells in the ventral eye vesicle express its known receptor Plxna2. With Sema6a or Plxna2 knockdown, many morphant eye progenitors fail to associate with the eye epithelium and exit the eye vesicle into either the vesicle lumen or the surrounding mesenchyme. We propose a novel and third mechanism of tissue cohesion, which involves dampening of tissue-autonomous repellent signaling. Our data argue that Sema6a/Plxna2 act eye autonomously to establish and maintain the position and integration of cells within the eye vesicle epithelium, processes we postulate involve both repulsive signaling and selective neutralization of Sema6a activity within the ventral eye vesicle.
plxna2 is required for eye vesicle development
Several class 6 semaphorin genes (Ebert et al., 2012) and plxna2, a known receptor for Sema6a in mammals (Suto et al., 2007; Renaud et al., 2008; Tawarayama et al., 2010), are expressed in the 72 hours post fertilization (hpf) zebrafish retina (supplementary material Fig. S1). To investigate plxna2 expression earlier in zebrafish eye development, we used RNA in situ hybridization. Interestingly, plxna2 is expressed transiently as the eye vesicles first form. At 4-12 somites, plxna2 is present in the ventral (future temporal) eye vesicle and the surrounding mesenchyme (Fig. 1A-D). Expression continues through 18 somites and is absent from the optic cup at 24 hpf (data not shown).
To investigate whether Plxna2 plays a role in early eye development, we designed an antisense morpholino oligonucleotide (MO) against plxna2 (plxna2e3i3) that alters mRNA splicing and, as seen by RT-PCR, excludes exon3 to create a frameshift mutation (Fig. 1F). Zebrafish eye fields are specified during gastrulation, and eye-field separation into the bilateral eye vesicles initiates during neurulation (Rembold et al., 2006; Fuhrmann, 2010) (Fig. 1G). To determine whether these events occur normally in plxna2 morphants, we examined the expression of key eye-field transcription factors: Rx3 is required for vesicle evagination (Loosli et al., 2003) and precursor proliferation (Stigloher et al., 2006); Six3 for specification (Liu et al., 2010); and Meis1 for patterning (Erickson et al., 2010). Induction of the eye field, as marked by rx3 and six3 at 4-6 somites, appears normal in plxna2 morphants (Fig. 1H-K). By 12 somites, in both control and plxna2 morphant embryos, the eye field separates into bilateral vesicles that evaginate, as seen by meis1 expression (Fig. 1L,M). The size of the meis1 domain, however, is smaller in the plxna2 morphants, and lagging meis1+ cells that fail to segregate into the eye vesicles form a bridge of cells in the midline (Fig. 1L,M,P,Q). One possibility is that cells destined to become eye tissue are reallocated to a hypothalamic fate, as these two tissues derive from the same forebrain region (Wilson and Houart, 2004). Yet, the size of the expression domain of the hypothalamic marker nkx2.1 (Marin et al., 2002) is unaltered in the plxna2 morphants (Fig. 1N,O) (control 194.4±8 µm, n=16; plxna2 MO 183.2±7.6 µm, n=17; P=0.45 unpaired Student's t-test). Not surprisingly, given the brain expression of plxna2, the meis1 domain in the plxna2 morphant midbrain is reduced slightly when compared with control (91% of control length along the anterior-posterior axis, n=20). These data point to possible subtle defects in development of brain structures; however, we focused our study on understanding the unexpected early role for Plxna2 in eye development.
Plxna2 is required for eye vesicle integrity
To visualize eye vesicle development, we used the Tg(rx3:GFP) transgenic line, in which GFP is expressed by developing eye and hypothalamic progenitors (Rembold et al., 2006). In a 20-somite control, GFP+ cells are closely packed in the eye epithelium, whereas in plxna2 morphants the packing of GFP+ cells is less tight (Fig. 2A-D). In fact, significant numbers of GFP+ cells are ectopic to plxna2 morphant eyes as the eye vesicles evaginate and elongate from the neural keel (8-12 somites) (Fig. 2E,F,H). By 12 somites, the GFP+ eye vesicles of plxna2 morphants are smaller than control (Fig. 2E-G), consistent with the decrease in the meis1+ eye domain (Fig. 1Q). Importantly, a similar loss of GFP+ cells from the eye vesicles is seen with a second plxna2 MO directed against the ATG start site, showing specificity (Fig. 2I,J). Subsequent analyses are performed with the e3i3 MO. Moreover, the small eye vesicle observed with the plxna2 e3i3 MO is partially rescued by co-injection of human PLXNA2 mRNA, which is not targeted by the splice MO (Fig. 2M-O). To verify that the loss of eye cells and small eye size did not occur as a result morpholino off-target effects due to p53 activation (Robu et al., 2007), we compared eye size in plxna2 morphants injected with or without a p53 MO. In both groups, eye size is reduced when compared with control eyes (data not shown). These data argue strongly for the specificity of the loss of eye integrity to a reduction in Plxna2.
The border of the developing eye vesicle is normally continuous and smooth (Fig. 2A,E). In contrast, the border of the plxna2 morphant eye vesicle is often disrupted, with either cell processes or somata extruding from the vesicle (Fig. 2B,F). To visualize the movements of eye progenitors, we performed time-lapse microscopy of control (Fig. 3A, supplementary material Movie 1) and plxna2 morphant (Fig. 3B, supplementary material Movie 2) rx3:GFP embryos from 6-14 somites. Over this period, control eye vesicles evaginate and elongate, driven by the migration of eye progenitors (England et al., 2006; Kwan et al., 2012), and cells remain integral to the eye domain. By contrast, in plxna2 morphants, several aberrant cellular behaviors are observed. First, some GFP+ eye progenitors fail to enter the eye vesicles and remain in the region separating the two eye vesicles (Figs 2F and 3B, supplementary material Movie 2). These cells are also evident as ectopic meis1 label at the anterior midline of plxna2 morphants (Fig. 1M). Second, GFP+ cells within the eye vesicle move from the epithelium either into the vesicle lumen (Fig. 3D,D″) or the surrounding mesenchyme (Fig. 3B,D,D′). In summary, Plxna2 knockdown leads to some retinal progenitors failing to integrate into the eye vesicle epithelium.
To determine whether GFP+ cells continue to express an eye identity upon leaving the eye vesicle, we assessed meis1 expression. Ectopic meis1 is seen at the midline between the eye vesicles of plxna2 morphants (Fig. 1M), presumably labeling eye progenitors that fail to enter the eye vesicles. In contrast, most of the GFP+ cells that exit from the eye vesicle itself (based on time-lapse data) do not express meis1 (arrowheads, Fig. 4B). In section (Fig. 4C,D), the uneven boundary to the meis1+ morphant eye vesicle is particularly evident. We find that a laminin-immunoreactive basal lamina forms in both control and plxna2 morphants embryos (Fig. 4E-L), suggesting that failure to develop the basal lamina that borders the eye vesicle does not explain the ragged edges of the morphant vesicles. In plxna2 morphants, ectopic GFP+ cells are present in the laminin-rich mesenchyme, both close to and at a distance from the GFP+ eye vesicle (arrowheads, Fig. 4H,J,K). Also evident are GFP+ cells filling an expanded ventricle (Fig. 4K,L), a phenomenon explained by the arrival of progenitors from the epithelium, as seen by time-lapse imaging (Fig. 3D,D″). Finally, although the dorsal and ventral vesicle epithelia are tightly apposed in controls, in many plxna2 morphants this juxtaposition of epithelial layers is lost (Fig. 4H,K,L). Ectopic cells often show hallmark features of apoptosis, including GFP exclusion from subcellular compartments, crenated nuclei and activated caspase 3 immunoreactivity (Fig. 4M-Q). Caspase 3+ cells are virtually absent from control eyes, but in plxna2 morphants, immunopositive cells are present in the ventricle and epithelium of the eye, and the mesenchyme (Fig. 4N-Q). Time-lapse microscopy indicates that cells can acquire an apoptotic appearance prior to leaving the neuroepithelium or shortly thereafter (Fig. 3D′,D″). Cell death may occur as a result of the loss of contact of some progenitors in the plxna2 morphants with the eye epithelium, a phenomenon known as anoikis.
Loss of Sema6a causes defects in eye vesicle cohesion
Sema6a was a candidate ligand for Plxna2 in regulating eye vesicle integrity, as it serves as a Plxna2 ligand in other systems (Zhou et al., 2008) and is expressed in the optic cup (Ebert et al., 2012). We find that sema6a mRNA is expressed throughout the eye vesicle between 4-14 somites (Fig. 5A-D). Although sema6a is co-expressed with plxna2 in the ventral vesicle (Fig. 1C), there is no co-expression in the dorsal vesicle (Fig. 5D,E). If Sema6a and Plxna2 function as a ligand-receptor pair, knockdown of either protein should produce a similar eye phenotype. To inhibit Sema6a, we designed a sema6a antisense MO to splice out exon 1 and introduce a frameshift mutation and premature truncation of the protein (sema6ae1i1) (Fig. 5F). Similar to the plxna2 morphant (Fig. 2B,F), knockdown of Sema6a results in smaller GFP+ eye vesicles at 12 somites when compared with controls (Fig. 5G,I,N), and the presence of GFP+ cells ectopic to the vesicles (Fig. 5I,O). Furthermore, activated caspase 3+ cells are observed within and around the eye vesicle (supplementary material Fig. S2). The specificity of the eye phenotype for Sema6a knockdown is confirmed by us being able to rescue the sema6a morphants by injection at the one-cell stage of a full-length zebrafish sema6a mRNA not targeted by the sema6a splice MO: loss of progenitors from the eye vesicle and eye size are both partially rescued (Fig. 5J,N). The similarity of the eye phenotypes in plxna2 and sema6a morphants argues that Sema6a is the Plxna2 ligand.
We reasoned that because MOs may only partially knockdown protein levels, tandem knockdown of both Plxna2 and Sema6a should exacerbate the eye phenotype. Surprisingly, however, injection of both MOs rescues the defects observed in single morphants, including the small eye vesicle, the escape of GFP+ cells and the presence of activated caspase 3+ cells (Fig. 5K-M,O,P, supplementary material Fig. S2). These data argue strongly that the location and levels of Sema6a and Plxna2 activity must be tightly regulated: a decrease of one gene or the other leads to a reduction in eye vesicle integrity and size, whereas loss of both genes in tandem produces no overt defect.
Plxna2 and Sema6a act primarily within the eye vesicle
To shed light on where and how Sema6a/Plxna2 are required to promote eye cohesion, we developed an eye explant assay (Fig. 6A). We dissected eyes from 6-somite rx3:GFP transgenic embryos and plated them in culture for 6 h. We found that a small number of GFP+ cells lose integral association with the control explants within 6 h (Fig. 6B), a fact that provided us with a means to investigate whether Sema6a/Plxna2 could signal in a repellent fashion to block their escape (Toyofuku et al., 2004; Mauti et al., 2007). We examined whether soluble Sema6a-Fc (Haklai-Topper et al., 2010) would prevent cells from leaving the explant. This scenario would only exist if the emerging eye cells expressed a functional Plxn receptor. This appears to be the case, because almost no ectopic cells are seen in Sema6a-Fc cultured explants (Fig. 6B,C,G). Thus, Sema6a can act as a repellent for zebrafish eye progenitors.
With plxna2 mRNA in the mesenchyme and sema6a in eye progenitors, we hypothesized that Sema6a and Plxna2 mediate repellent signaling at the eye vesicle boundary to prevent eye cells from leaving (Figs 1D and 5D). To test this possibility, we cultured eye explants with and without the adherent plxna2-expressing mesenchyme. Only a handful of GFP+ cells lose integral association with the control eye vesicle explants within 6 h [5.8±1 (s.e.m.) cells, n=10 explants], and although more do so when the mesenchyme is removed (13.6±1.6 cells, n=10 explants; unpaired Student's t-test P<0.001), the numbers are still small. Thus, the surrounding mesenchyme likely plays only a small role in maintaining a boundary for the eye vesicle.
An alternative model is that Sema6a and Plxna2 function in a tissue-autonomous manner to control eye vesicle integrity. In support of this, simultaneous knockdown of Plxna2 and Sema6a largely mitigates the ectopic cell phenotype (Fig. 5M,O), whereas loss of both signaling partners would be expected to result in a robust escape of cells from the eye vesicle if a mesenchyme-derived Plxna2 repellent boundary mechanism is key. To provide further support for a tissue-autonomous role, we cultured sema6a and plxna2 morphant eye vesicles (Fig. 6D-F), postulating that morphant explants should exhibit comparable numbers of ectopic cells to a mesenchyme-removed wild-type explant if a Sema6a/Plxna2 repellent boundary is important. This is clearly not the case, because extensive disaggregation of morphant eye explants occurs (Fig. 6D-F,H). Thus, our data argue for an additional mechanism(s), apparently eye-autonomous, by which Sema6a and Plxna2 mediate eye vesicle cohesion. In support of this, in hanging-drop cultures (Foty, 2011) of dissected and disaggregated wild-type and plxna2 18-somite rx3:GFP eye vesicles (n=3; 16 drops/condition/experiment), wild-type, but not plxna2 morphant, cells come together to form cellular aggregates (data not shown).
Plxna2 required for regionalization of the developing eye
In other systems, the interaction of Sema6a and Plxna2 in cis (same membrane) apparently inhibits the ability of Sema6a to interact with Plxna in trans (membrane of different cell) (Suto et al., 2007; Renaud et al., 2008; Haklai-Topper et al., 2010; Yaron and Sprinzak, 2012). Because plxna2 and sema6a mRNAs are both present in the ventral vesicle, we postulated that Plxna2 neutralizes Sema6a repulsion in this domain. Neutralization would divide the eye vesicle into repellent and non-repellent domains. If true, we expected to see disorganization of regionally expressed eye markers in the morphants. We evaluated the expression of several transcription factors at 12 somites. fogx1a, a transcription factor expressed in the dorsal (future nasal) eye vesicle (Picker et al., 2009), expands into ventral domains in plxna2 and sema6a morphants (Fig. 7A-C). vax2 and tbx5 mark tissue that gives rise to future ventral and dorsal retina, respectively (Koshiba-Takeuchi et al., 2000; Take-uchi et al., 2003). In wild type, vax2 is expressed in the anterior eye vesicle (future ventral retina; Fig. 7E,E′), whereas tbx5 is in the posterior eye vesicle (future dorsal retina; Fig. 7K,M). The tight domain of vax2 expands in both plxna2 and sema6a morphants into the posterior/future dorsal region (Fig. 7F-G), as seen by an increase in the ratio of the height of the vax2 expression domain relative to the eye vesicle height (Fig. 7I,J). The tbx5 domain appears less distinct and intense in the plxna2 (compare Fig. 7I,J) and sema6a (data not shown) morphants. Interestingly, although eye vesicle cohesion is rescued in the plxna2 and sema6a double morphants (Fig. 5), regionalization of the eye vesicles is still disrupted (Fig. 7D,H,N). These data indicate a failure of regionalization of the eye vesicle in the absence of Plxna2 and/or Sema6a.
Defects in proliferation are present in plxna2 and sema6a morphants
We focus here primarily on a role for Sema6a/Plxna2 in eye tissue integrity, but evaluated whether defects in proliferation, arising either directly from a loss of Sema6a/Plxna2 or secondary to a loss of eye vesicle integrity (Gao et al., 1991; Temple and Davis, 1994), might also contribute to the small morphant eyes. The proliferation regulator rx3 (Stigloher et al., 2006) is expressed normally in the separating eye vesicles at eight somites (supplementary material Fig. S3A,B), but by 16 somites rx3 levels are reduced, specifically in the eye vesicle and not the hypothalamus (supplementary material Fig. S3C,D). Additionally, at 20 somites, fewer mitotically active phospho-histone (pHH3) immunopositive cells are present in both plxna2 and sema6a morphant eyes relative to controls (supplementary material Fig. S3E-K). Thus, the reduction in morphant eye vesicle size probably arises from both a loss of cells from the eye vesicle and defects in proliferation.
Sema6a/Plxna2 are not required for optic cup integrity
Although the loss of Plxna2 signaling transiently compromises eye vesicle cohesion, optic cups form as development proceeds, which is not unexpected given that plxna2 mRNA is absent from the optic cup after 24 hpf (data not shown). The cells that remain in the plxna2 morphant eye organize into an apparently cohesive, apicobasal polarized neuroepithelium by 21 hpf (Fig. 8A-D), as shown by the appropriate apical localization of immunoreactivity for ZO1, a tight junction-associated protein. The pigment epithelium is present in both control and morphant eyes at 30 hpf (supplementary material Fig. S4A,B), and few TUNEL+ apoptotic cells are evident at 24 hpf (data not shown).
Not surprisingly, plxna2 (control, 6.5% small eyes, n=155; plxna2 morphants, 93%, n=233; Fisher's exact test P<0.001) and sema6a morphants have smaller eyes than wild type at 72 hpf (Fig. 8E-H,O). Nonetheless, plxna2 morphant eyes show the appropriate proportion of cells in the three cell layers (Fig. 8P), and contain several distinct cell types: ZN8+ retinal ganglion cells (RGCs), Pax6+ amacrine cells (Pax6 also weakly labels RGCs) and rhodopsin+ rod photoreceptors (Link et al., 2000) (Fig. 8I-N). Reflecting eye size, fewer RGCs are present in the central retina of a transgenic in which differentiating RGCs express GFP [Tg(isl2b:GFPzc7) (Pittman et al., 2010)], in morphants compared with control (control=83.4±4.1 cells (s.e.m.); plxna2 morphant=47.5±3.1 cells; n=3 experiments, n=15; Mann-Whitney, P<0.001). A delay in eye development does not account for the reduction in RGCs, in that GFP expression initiates normally in both control and morphant isl2b:GFP transgenics at 28-30 hpf (supplementary material Fig. S4A,B). Together, these data indicate that subsequent events of retinogenesis occur normally, albeit in the context of the defects that arise from the early role of Plxna2 in eye development.
Here, we provide evidence that precise spatial control of Sema6a/Plxna2 function is required for progenitors to remain within the eye primordium. In the early embryo, the transmembrane Sema6a is expressed throughout the eye vesicle, whereas cells in the ventral eye vesicle express its partner receptor, Plxna2. In the absence of normal Plxna2 signaling, progenitors escape from the eye vesicle over time. We put forward a novel, tissue-autonomous mechanism of organ cohesion involving local control of repellent signaling, which ultimately allows the proper positioning of cells and organization of the eye vesicle into regionalized subdomains.
Our data argue that carefully regulated Sema6a and Plxna2 interactions are required to maintain eye cohesion. Similar defects with regards to eye vesicle size, ectopic eye cells, marker expression and proliferation are present with loss of either Plxna2 or Sema6a. Yet, loss of both Sema6a and Plxna2 results in no obvious epithelial integrity phenotype, arguing that Plxna2 and Sema6a need to be appropriately expressed relative to one another, with no interaction actually less harmful than too little of one or the other protein. The double-morphant data also support the idea that the two proteins function in concert.
Initially, we postulated that Sema6a/Plxna2 provide a molecular boundary to the eye vesicle, by mediating repulsive interactions between cells in neighboring tissues: mesenchymal Plxna2 would keep Sema6a-expressing cells in the eye vesicle, presumably via reverse Sema6a signaling [cells are known to use Sema6 to respond to repellent Plxn signals from neighboring non-like cells (Toyofuku et al., 2004; Mauti et al., 2007)]. Yet two observations suggest that a boundary mechanism plays only a minimal role in maintaining eye vesicle integrity. First, there is a rescue of eye vesicle size and suppression of delamination when Plxna2 and Sema6a are knocked down simultaneously, whereas with a boundary mechanism we would expect a worsened phenotype. Second, wild-type eye explants missing mesenchyme show a relatively minor loss of cells.
Instead, we suggest that Sema6a/Plxna2 function within the eye vesicle itself. We propose a working model (Fig. 9A), based on our data and work showing that Plxna2 in cis with Sema6a inhibits the function of Sema6a in trans (Suto et al., 2007; Haklai-Topper et al., 2010). In the hippocampus, such a mechanism is important in directing the projections of Plxna4-expressing mossy fibers to the proximal dendritic region of CA3 pyramidal cells: Sema6a is expressed in all layers of the CA3 region, but mossy fibers target only the stratum lucidum, where Plxna2 is present to neutralize Sema6a (Suto et al., 2007). We propose a similar neutralization of repulsive signaling occurs in the ventral eye vesicle, where both plxna2 and sema6a are expressed. By contrast, in the dorsal eye vesicle, where plxna2 is absent, Sema6a is repulsive. Zebrafish eye progenitors migrate laterally and posteriorly into the evaginating eye vesicles, a process that continues through to 14 somites (England et al., 2006; Kwan et al., 2012). We hypothesize that some of these progenitors are sensitive to Sema6a, avoid the repulsive dorsal domain and are directed into the ventral domain. Motor axons entering the limb bud are presented with a similar dorsal/ventral binary choice, mediated by ephrin/Eph repellent signaling (Luria et al., 2008). An additional role for dorsal Sema6a activity could be to restrict dorsoventral mixing of cells within the vesicle. Thus, we propose that Sema6a/Plxna2 control the movement of cells into and within the forming eye.
With knockdown of Plxna2 or Sema6a, we postulate that repellent signaling within the ventral domain is activated aberrantly (Fig. 9B,C), such that Sema6a is active both ventrally and dorsally. Whereas wild-type progenitors entering the eye vesicle are presented with a binary choice, those that are sensitive to Sema6a select the ventral domain. Ventral and dorsal domains become essentially equivalent in morphants with regard to expression of active Sema6a repellent, whether it be low (sema6a morphant), high (plxna2 morphant) or no (double morphant) activity. With no clear preference for one domain over the other, some of the progenitors that normally are directed ventrally will end up inappropriately in the dorsal eye vesicle. In support of this, the plxna2 morphant phenotype is not restricted to the ventral eye vesicle, and expression of regionalized markers is altered in both the single and double morphants. Further support for a neutralization role for Plxna2 comes from the observation that in plxna2 morphants some progenitors fail to enter the eye and collect in the forebrain between the two eye vesicles, as seen by meis1 expression and live imaging, a phenotype that is concordant with high Sema6a throughout the entire vesicle.
Why do progenitors leave the eye vesicle in single morphants? A possible explanation is that once within the control eye vesicle, most progenitors (except possibly those at the boundary between dorsal and ventral domains) do not experience Sema6a repulsion: those located in the dorsal domain presumably lack a mechanism to sense Sema6a, and ventral progenitors sit in a domain where Sema6a is neutralized. In contrast, Sema6a-responsive cells within the morphant eye vesicle, either dorsally or ventrally, are exposed to some level (either low or high) of Sema6a activity. This aberrant repulsion impairs the ability of a cell to integrate into the epithelium. Whether the cells are pushed from the epithelium, or simply fail to associate with neighbors, is unclear. Regardless, the cells that fail to integrate escape the eye vesicle. The double morphant data lend support to the idea that aberrant regulation of repulsion within the eye vesicle is what leads to the loss of eye cells from the forming vesicle. Because virtually no repulsion would occur in the double morphants, mis-positioned progenitors would not be forced out of the eye epithelium. In agreement with this, regionalization but not vesicle cohesion is defective in the double morphants. A similar delamination phenotype is seen with knockdown of Foxg1a (Picker et al., 2009), though the loss of cells appears to occur earlier in plxna2 morphants. Loss of epithelial contacts could in part explain the reduction of mitotically active cells in morphant vesicles, in that cultured neuroepithelial cells prevented from aggregating leave the cell cycle (Gao et al., 1991; Temple and Davis, 1994). Alternatively, Sema6a/Plxna2 signaling could function to directly regulate progenitor proliferation (Kigel et al., 2011).
Our model requires that a subset of eye progenitors expresses a Plxn in order to respond to ‘activated’ Sema6a. It is possible that Plxna2 is the receptor. The double morphant data argue that some residual Plxna2 remains in the single plxna2 morphants. Alternatively, eye progenitors may express a different Plxn with which to respond to the Sema6a. Indeed, Plxna4 allows hippocampal axons to sense Sema6a+ in all CA3 layers, except where Plxna2 is also expressed (Suto et al., 2007). Future experiments will need to address the involvement of additional Plxn receptors.
Time-lapse imaging suggests that cells in different domains of the zebrafish eye vesicle do not intermingle (Kwan et al., 2012). Thus, tissue cohesion is likely to be important not only for an entire organ, but also for subdomains within a tissue. The cellular and molecular mechanisms that underlie such cohesion are unknown. Here, we propose that Sema6a/Plxna2 promote the formation of cohesive collections of cells within subdomains of the eye vesicle by establishing within the vesicle a repulsive and a non-repulsive domain. The latter arises from neutralization of repulsion, which may be a key mechanism to promote interactions of cells within a tissue domain by allowing adhesive mechanisms between like-cells to predominate. Thus, anti-repellent and adhesive mechanisms between like cells may cooperate to induce and maintain tissue cohesion. Eventually, the maintenance of these cellular domains by dynamic cellular signaling would be unnecessary as a tight epithelium becomes established.
Eye formation depends on the successful completion of a series of segregation events that involve distinct molecular mechanisms. Ephrin/Eph signaling mediates cell dispersal required for retinal progenitors to access the Xenopus eye field (Moore et al., 2004), and allows the zebrafish eye field to remain segregated from adjacent neural plate regions (Cavodeassi et al., 2013). Subsequent morphogenesis of the eye field may involve non-canonical Wnt signaling (Cavodeassi et al., 2005). We propose that the Plxn signaling pathway then promotes positioning of cells within the eye vesicle to help establish and maintain early regionalization of the developing eye. Understanding how repellent signals promote tissue subdomain organization will be important in determining how developing organs form.
MATERIALS AND METHODS
Zebrafish embryos were developmentally staged as described previously (Kimmel et al., 1995). Pigmentation was blocked by addition of 0.003% phenylthiourea at 24 hpf. rx3:GFP and isl2b:GFP transgenic fish were provided by Drs Wittbrodt (University of Heidelberg, Germany) and Chien (University of Utah, USA), respectively. All procedures were approved by the University of Calgary Animal Care Committee.
Antisense morpholino oligonucleotides (Gene Tools), mRNA or combinations of the two, were injected at the one-cell stage (Kimmel et al., 1995). Morpholino concentrations and sequences were as follows: plxna2, 2 ng (AAAAGCGATGTCTTTCTCACCTTCC); sema6a, 4 ng (TGCTGATATCCTGCACTCACCTCAC). To isolate full-length zebrafish sema6a, cDNA was obtained from total mRNA of 20 hpf zebrafish using Trizol and reverse transcriptase (Invitrogen), according to the manufacturer's protocols. The PCR product generated with specific primers GATCGGATGTGATCGGGATTT (forward) and GATTATGT-GCAGGAGTCTCC (reverse) was subcloned into the pCRII topo vector. Sequencing of the full-length coding sequence of sema6a showed 100% identity with data previously published (NM_199992) between nucleotides 20 and 3026. sema6a mRNA and human PLXNA2 mRNA (Origene) were injected at 175-200 pg/embryo.
RNA in situ hybridization
RNA in situ hybridization was performed as described previously (Thisse and Thisse, 2008). Antisense riboprobes were made using the following primers: plxna2 forward, ATGTGATACAGGAGCCGAGG; plxna2 reverse, AGAG-TCAGAAGGCTGTCGGA; sema6a forward, ATGCGAGCGCAGGCCC-TGC; sema6a reverse, CACGTTGGCGTGTTTGGCGT; foxg1a forward, GCAGGAAGAAAAACGGGACGC; foxg1a reverse, GATGGGTGAGGG-ACATGGGG. rx3, six3 and nkx2.1 constructs were provided by Dr Kurrasch (University of Calgary, Canada), and the meis1 construct by Dr Waskiewicz (University of Alberta, Canada).
RNA isolation and RT-PCR
Total RNA from 72 hpf embryos was prepared using Trizol/chloroform (Invitrogen). First-strand cDNA was made using the Superscript II RT-PCR protocol (Invitrogen) and amplified by PCR using Taq polymerase (Fermentas) and the following primers for plxna2: forward, CTTTGAACCACTCAG-CACCA; reverse, CGTATTCCAGTCGACCCTGT.
rx3:GFP transgenic embryos were imaged with a 10 or 20× objective using step sizes of 3 µm after mounting in 0.5% low melt agarose on glass bottom dishes (Mat-Tek) on a Zeiss LSM 510 Meta or LSM 700 inverted microscope. Time-lapse stacks were taken every 5 min over a 3.5 h window. Stacks were subjected to a Kalman stack filter in Image J or processed in Zen Lite and are presented as maximum intensity projections.
Embryos were fixed in 4% PFA, infiltrated with 35% sucrose (EM Science) and embedded in optimal cutting temperature (OCT; Tissue Tek). Transverse sections were performed on a Microm HM 500 OM cryostat at 12 µm. The following antibodies were used: activated caspase 3 (G7481, 1:500; Promega), laminin (L9393, 1:200; Sigma), Pax6 (PRB-278P, 1:500; Covance), ZN8 (1:1000; Developmental Hybridoma Studies Bank), rhodopsin (MAB5356, 1:1000; Chemicon), pHH3 (pHH3-06-570, 1:500; Millipore), ZO-1 (339100, 1:150; Invitrogen), and anti-mouse or anti-rabbit Alexa Fluor 546 (mouse-A11030, rabbit-A11030, 1:1000; Invitrogen). Images were taken on a Leica DMR compound microscope using a Magnafire camera (Optronics). Images were processed in Adobe Photoshop for brightness and contrast. pHH3 immunostaining was performed in wholemounts, and pHH3+ cells within the GFP+ eye vesicle counted under a fluorescent Zeiss compound stereomicroscope. Statistical analyses throughout the study were performed with GraphPad software. For H&E staining, slides were bathed in consecutive 5 min washes of aqueous Eosin and Hematoxylin (Sigma-Aldrich). Slides were imaged on a Zeiss StemiSV11 stereomicroscope using Axiovision imaging software.
Eye size measurements
Fixed 12-somite rx3:GFP embryos were photographed dorsal side upwards by using a Zeiss StemiSv11 stereomicroscope and Axiovision software. The area (not including the optic stalk) and/or length of the GFP or meis1+ left eye vesicle was measured using Axiovision software. Alternatively, embryos were grown to 72 hpf, embedded in JB4 (Polysciences) and sectioned, and imaged and measured using SPOT imaging software. Retinal height was measured from dorsal to ventral surfaces, through the lens.
Six somite control and morphant eyes were dissected and each placed in an individual well of a 96-well plate in modified Barth's solution with 10 mg/ml gentamycin. Select explants were incubated with human recombinant Sema6a-Fc (300 ng/ml) (R&D Systems). Explants were incubated at 28°C for 6 h. Images were taken on a Zeiss Axiovert 40 CFL compound fluorescent microscope, and the numbers of GFP+ progenitors that exited the eye vesicle were quantified.
The authors thank Drs Schuurmans and Kurrasch for helpful suggestions, A. Zuba for technical assistance, Dr Bertolesi and J. Johnston for identification of zebrafish sema6a, Drs Kurrasch and Waskiewicz for constructs, and Drs Chien and Wittbrodt for transgenic lines.
The authors declare no competing financial interests.
A.M.E. contributed Figs 1-3 and 5-9 and supplementary material Figs S1 and S3, analyzed data (Figs 8 and 9, supplementary material Fig. S3), and participated in writing. S.J.C. provided tools, generated data (Fig. 3), edited the manuscript, and provided intellectual input. C.L.H. contributed data (Figs 2-5 and 7) and edited the manuscript. P.B.C. generated data for Figs 4O-Q and 5G-N and supplementary material Fig. S2. S.M. contributed to experimental design, performed data analysis (Figs 1,2,5,6 and 7) and wrote the manuscript.
A.M.E. was supported by fellowships from Alberta Innovates-Health Solutions (AI-HS), Canadian Institutes of Health Research (CIHR) and the Foundation for Fighting Blindness. S.M. was a Tier II Canada Research Chair (CRC) in Developmental Neurobiology and is an AI-HS Scientist. S.J.C. was a Tier II CRC in Angiogenesis and Genetics and is an AI-HS Senior Scholar. CIHR operating grants to S.M. [MOP#14138] and S.J.C. [MOP#97787] supported this research.
Supplementary material available online at http://dev.biologists.org/lookup/suppl/doi:10.1242/dev.103499/-/DC1
- Received October 3, 2013.
- Accepted April 24, 2014.
- © 2014. Published by The Company of Biologists Ltd