Pax6 organizes the anterior eye segment by guiding two distinct neural crest waves

Summary

Cranial neural crest (NC) contributes to the developing vertebrate eye. By multidimensional, quantitative imaging, we traced the origin of the ocular NC cells to two distinct NC populations that differ in the maintenance of sox10 expression, Wnt signalling, origin, route, mode and destination of migration. The first NC population migrates to the proximal and the second NC cell group populates the distal (anterior) part of the eye. By analysing zebrafish pax6a/b compound mutants presenting anterior segment dysgenesis, we demonstrate that Pax6a/b guide the two NC populations to distinct proximodistal locations. We further provide evidence that the lens whose formation is pax6a/b-dependent and lens-derived TGFβ signals contribute to the building of the anterior segment. Taken together, our results reveal multiple roles of Pax6a/b in the control of NC cells during development of the anterior segment.

Keywords:

Embryos – Eyes – Cornea – Zebrafish – eye lens – Cell migration – Endothelium – Diencephalon

Introduction

Neural crest (NC) cells are a population of pluripotent embryonic stem cells that originate at the boundary between neural and non-neural ectoderm. Subsequently, NC cells delaminate from the neuroectoderm and migrate to remote regions of the body to give rise to a wide variety of different cell types such as the peripheral nervous system, the cranial skeleton and ocular tissues [1].

Anterior segment (AS) dysgenesis (ASD) of the eye refers to a group of developmental disorders of the distal structures of the eye such as the cornea, iris, and trabecular meshwork [2]. These ocular structures as well as the sclera are all NC derivatives [1,3]. ASD is associated with 8 human genes, including the transcription factors PITX2, FOXC1/2 and paired box protein Pax-6 (PAX6) [2,4]. Defective TGFβ signalling also causes ASD: TGFβ2 is required for Pitx2 and Foxc1 expression in NC-derived AS structures in mice [5,6]. The lens which expresses TGFβ2 was shown to act as a signalling centre that organizes AS morphogenesis. Homozygous mutant studies in rodents indicated a non-cell autonomous role of PAX6 in guiding NC cells into the eye [7] and heterozygous conditional mutation of PAX6 in the lens and the cornea led to ASD with malformed trabecular meshwork and Schlemm’s canal [8]. Although the defects are initially evident mainly in the AS, PAX6 mutant mice developed elevated intraocular pressure (IOP) which later led to damage of the whole retina as a result of juvenile glaucoma.

In mammalian eye development, reduced PAX6 levels correlate with the severity of eye defects [9]. Pax6 expression establishes a concentration gradient, with higher levels on the distal side of the optic cup (i.e., the cornea, lens and iris) and lower levels on the proximal side [10,11]. Proximal marker gene expression in Pax6 homozygous mutant eyes was expanded abnormally toward the distal side, with no nasal-temporal and dorsal identities [10,12]. In zebrafish, signals from the midline induce the expression of Pax2, which inhibits Pax6 expression in the optic stalk and thereby regulates the partitioning of the optic primordium [13]. The AS of the zebrafish eye is not only structured similarly to that of mammals [14,15] but also displays the same pax6 dependency. pax6b mutants show a spectrum of ASD phenotypes with misregulated AS marker gene expression [16].

Here, we have examined how neural crest cells contribute to the zebrafish eye, and whether this contribution involves pax6a/b function. We utilised the presence of duplicated pax6a/b genes in the zebrafish by studying how ASD develops in pax6a/b homozygous and compound hetero-/homozygous mutants through live multidimensional imaging of ocular NC cells and the corneal endothelium. Analysis of NC migration in wildtype and pax6a/b single, double and compound mutants revealed two ocular NC populations and a role for Pax6 in correctly guiding their migration. Expression of guidance molecules in the optic cup and its surroundings including NC cells was altered in the pax6a/b double homozygous mutant. Unlike in murine Pax6 mutants, the proximodistal patterning of the optic cup was normal. We demonstrate instead a role of Pax6 in the control of the expression of NC guidance molecules, uncovering a so far unrecognised late function of Pax6 independent of the earlier proximodistal patterning of the optic cup. The double homozygous mutant failed to establish the lens, leading to severe AS defects, with no formation of the corneal endothelium. Similar AS defects were observed in wild type embryos after chemical inhibition of TGFβ signalling or removal of the lens. Conversely, transplantation of a wildtype lens into the double homozygous mutant eye partially restored the AS. The ASD with misguided NC differentiation in pax6a/b mutants underscores the role of the second NC wave in governing distal eye maturation.

Results

Two distinct populations of cranial NC cells contribute to the developing eye

To trace the NC cells entering the eye, two transgenic lines were generated which express the green-to-red photoconvertible Eos fluorescent protein (EosFP [17]) in NC cells under control of sox10 regulatory sequences. Tg(sox10:mem-tdEosFP) expresses a membrane-targeted construct of tandem-dimeric EosFP [18] and Tg(sox10:h2a-tdEosFP) directs tdEosFP to the nucleus. Expression of EosFP became first detectable between 11 and 12 h post-fertilisation (hpf) as two bilateral stripes of NC cells at the lateral border of the anterior neural keel (S1 Video). Between two lateral NC stripes, a temporal medially-located NC cell group was observed during 13–15 hpf. The anterior-most cells of the lateral stripes reached the proximal region of the nascent eye at 14 hpf (S1 Video) and started to envelop the proximal side of the eye (Fig 1A and 1E), eventually forming a monolayer by 18 hpf (Fig 1B and 1F; S1 Fig). These NC cells retain large cell-cell contact planes while enwrapping the proximal side of the optic cup and display a cuboidal cell shape (S1D–S1F Fig). They migrated as bilateral stripes (S1 Video). Importantly, this group of NC cells remained restricted to the proximal half of the eye and did not settle in the distal AS (S1 Fig). In this study, we refer to this early eye-colonizing NC cells as 1°NC cells in contrast to a second group of NC cells, newly defined in this study, which we term 2°NC cells.

These 2°NC cells approach the eye later from the dorsal side at 20–22 hpf (magenta circles in Fig 1C and 1G). At 17–18 hpf, we observed that 2°NC cells formed a transient cell cluster near the dorsal edge of the eye (magenta circles, Fig 1B and 1F; S1 Video). Subsequently around 22 hpf, the 2°NC cluster dispersed into individual cells with a mesenchymal morphology and migrated over the dorsal rim of the eye into the distal region (S2 Video). In contrast to 1°NC cells which migrated in groups with large cell-cell contacts (S1D–S1F Fig), 2°NC migrated as single cells with maximally filopodia-like contacts to neighbouring migrating cells (Fig 1I). At 40 hpf, Sox10 reporter-positive NC cells were found to contribute to the distal and periocular tissues of the eye (Fig 1J–1J”).

Sox10 transcription becomes inactivated after dissolution of the 2°NC cell cluster

At the 40 hpf stage, 2°NC cell derivatives retained only very low levels of expression of the Sox10 reporter and we could not continue to trace their fate in AS structures, suggesting that sox10 expression in 2°NC cells has been turned off by this stage. In agreement, upon migration of 2°NC into the eye, sox10 mRNA expression was detectable only in a few mesenchymal cells in the distal side of the eye, suggesting that sox10 is down-regulated in NC cells once they have entered the differentiation phase [19], in contrast to those of the other eye mesenchymal marker genes pitx2 and foxc1a (S2 Fig). Similar observations were made with the 1°NC cells except that we failed to detect mRNA expression earlier when we examined the RNA expression in these NC cell derivatives in the proximal eye at 18 hpf.

By using the green-to-red photoconversion property of EosFP upon exposure to 400 nm light [20], we estimated more precisely when sox10 expression is silenced. The green form of the EosFP reporter in Tg(sox10:mem-tdEosFP) embryos was converted to the red form at different time points from 16 to 24 hpf and the recovery of green fluorescence was monitored at 36 hpf (S2 Fig). When photoconverted at 16 hpf, the 1°NC cells did not recover green EosFP (S2 Fig, asterisks in D’), suggesting that the 1°NC had stopped to express the sox10 reporter already as early as 16 hpf (S2 Fig; summarized in Fig 1L). In contrast, the 2°NC cells recovered green EosFP expression when converted during 16–20 hpf, but not beyond 20 hpf (S2 Fig). This time point coincided with the dissolution of the 2°NC cell cluster and the arrival of individual 2°NC cells at the distal eye segment. Photoconversion at 17 hpf is thus an effective way to visually distinguish between 1°NC and 2°NC populations (Fig 1K1-4 and 1M).

Wnt reporter is consecutively activated in 1° and 2°NC

Previously, Wnt signalling has been implicated in the control of NC cell differentiation [21]. To assess canonical Wnt signalling activity in 1° and 2°NC cells, Tg(7xTCF-siam:nls-mCherry) [22], in short Tg(Wnt-rep), was crossed into the NC reporter line Tg(sox10:h2a-tdEosFP) and imaged by using a custom-built, multi-channel, high-resolution digital scanned laser light sheet fluorescence microscope (DSLM) [23]. During 13–15 hpf, transient expression of the Wnt reporter was observed in NC cells; expression became barely detectable by 15 hpf (Fig 2A–2C’ and 2F; S3 Video). Starting at 18 hpf, we found intense Wnt reporter activation in the 2°NC cells (Fig 2D–2E’ and 2F; S3 Video). This second activation of the Wnt reporter in 2°NC cells occurred in approximately 50% of NC cells and was maintained after entry of 2°NC cells into the distal side of the eye (Fig 2G, S3 Video, S3I Fig). Thus, 2°NC cells represent a heterogeneous group of cells with respect to Wnt activation.

We next tested whether Wnt signalling plays a role in 2°NC specification, migration or differentiation. Treatment of embryos with the canonical Wnt inhibitor IWR-1-endo [24] showed a reduction of sox10-expressing cranial NC cells (S3D Fig). However, neither migration of 2°NC cells into the eye nor formation of a coherent endothelial layer were affected by the treatment (S3 Fig). Irrespective of the function of Wnt signalling in 2°NC, the combination of Wnt reporter and sox10 gene expression provides an unequivocal marker for a subpopulation of 2°NC cells in situ.

NC cells destined for the distal side of the eye originate from the medial dorsal neural keel

Next, we assessed the relationship between the origin and destination of the two NC cell populations contributing to the eye quantitatively. To this end, we imaged Tg(sox10:h2a-tdEosFP) embryos by DSLM and employed our recently developed interactive image analysis framework [25] to trace back the origin of NC cells that arrived at the proximal and distal sides of the eye (S4 Video). NC cells for each destination were selected at two stages (17 and 24 hpf) for precise group selection based on the final location (Fig 3A). Since 1°NC and 2°NC cell populations are well separated at 17 hpf (Fig 1L), we combined a forward and backward strategy to track cells of the proximal and distal populations after continuous recordings from 11 to 31 hpf (Fig 3A). By backward and forward tracking of selected nuclei, we established the route, origin and destination of individual cells of proximal and distal NC populations (n = 2 embryos; S4 Fig).

For the analysed cells, we compiled the results in tracking diagrams by superimposing the backward and forward tracks of selected cells covering developmental time intervals ranging from 17 to 11 hpf and 17 to 31 hpf, respectively (Fig 3B–3E’, S5 Video). This in-depth analysis confirmed that the cells occupying the proximal side of the eye were all formed at 12–14 hpf in the lateral region of the neural keel (Fig 3B–3B’, S4A and S4B Fig), and are thus derivatives of 1°NC cells. The NC cells destined for the distal side of the eye mainly originated medially near the diencephalon and mesencephalon at 16–17 hpf (asterisk, Fig 3C), with a minor contribution of NC cells that originated in the lateral regions of the neural anlage (S5 Fig). Forward tracking showed that distal NC cells reached the eye via the dorsal rim of the optic cup (arrow, Fig 3E’). The lineage tree analysis on manually corrected data from an embryo identified comparable lineage sizes for NC groups destined to the proximal (n = 57 lineages) and distal eye (n = 46 lineages), respectively (Fig 3F and S7 Fig). Most of the analysed lineages (n = 113; 91%) produced a homogeneous NC population that exclusively consisted of NC cell derivatives in either the proximal or distal eye, showing that these NC cells with different destinations mostly belong to distinct lineages. Ten cells (9%) displayed a mixed lineage suggesting that NC cells with different destination are not totally distinct in their potential to migrate to and differentiate into AS structures. The tracking analysis furthermore confirmed the predominant migration of mediodorsally formed NC cells at 16–17 hpf into the distal side of the eye (82%, n = 39 NC cells), which was further supported by local photoconversion and time lapse analysis (Fig 3G–3H; S6 Video). In contrast, 1°NC cells rarely ended up in the distal part of the eye (3.6%, n = 112 NC cells). Thus, the destinations of NC cell groups with different origins are predominantly distinct; laterally located early migrating 1°NC cells are destined for the proximal side of the eye, whereas medially located 2°NC cells colonize predominantly the distal eye.

We next assessed whether 1°NC derivatives residing in the proximal eye could contribute to distal structures at later stages by assessing the orientation of migration after 22 hpf, systematically. The orientations of migration of most of the NC derivatives in the proximal eye after 22 hpf (S6A and S6C Fig) do not support the notion that 1°NC derivatives (S6E, S6F, S6H and S6I Fig) contribute largely to the distal segment of the eye. In contrast, the migration of the 2°NC cells (S6G, S6J and S6K Fig) is predominantly oriented along the proximodistal axis with long tracks heading to the distal segment (S6B and S6D Fig).

The 2°NC cells arise at the dorsal diencephalon and mesencephalon

DSLM imaging lacks the bright field view for precise anatomical registration of the origin of 2°NC. We thus mapped the origin of 2°NC cells relative to those of the expression domains of known marker genes in the diencephalon and midbrain. At 18 hpf, sox10 and Wnt reporter transcripts showed extensive colocalization in the two bilateral clusters of 2°NC cells (Fig 4A–4A”). In contrast, transcripts of sox9b whose expression was suggested to mark pre-migratory NC cells [26] and the Wnt reporter did not colocalize (Fig 4B–4B”). Only very few cells expressed sox9b and sox10 at the boundary between the expression domains of the two genes (Fig 4C–4C”). Thus, 2°NC cells expressing sox10 and Wnt reporter transcripts are located next to sox9b expressing diencephalic and mesencephalic regions. sox10 did not show extensive colocalization with pax6b (Fig 4D–4E; S7A and S7B Fig), which is expressed similarly as pax6a in the presumptive thalamic portion of the diencephalon at 18 hpf [27]. sox9b showed almost complete colocalization with otx5, a marker of pineal photo-receptor cells (Fig 4F; S7C Fig) [28] in the dorsal-most region of the diencephalon, where pax6b is not expressed (Fig 4E’–4F’). Taken together, the 2°NC cells arise from the dorsal mesencephalon and diencephalon close to the pineal anlage (Fig 4G–4I).

Endothelial cells emerge in a concentric ring pattern in the nascent cornea

The corneal endothelium is derived from NC cells [3] and its establishment is a hallmark of the formation of the anterior chamber. We found that the Tg(-3.1mnx1.1:GFP)^ml4 transgenic line [29] expresses GFP in the corneal endothelium. Although the DNA construct was originally developed to study spinal motor neurons, this particular insertion, which we mapped to the minus strand at the end of chromosome 9 (S9 Fig), is ectopically expressed in the corneal endothelium (Fig 5C–5C”‘). We employed this line, which we call Tg(corneaEndo:GFP), for live imaging. The first corneal endothelial cells emerge as a patch of less than 10 cells on top of the lens around the 60–72 hpf stage (Fig 5A). Corneal endothelial cells proliferate until they occupy the entire inner surface of the cornea (Fig 5B and 5C). The total number of corneal endothelial cells is 27 ± 4 cells (n = 47 embryos) at 120 hpf.

Based on this imaging analysis and the other data presented so far, we propose a model of five developmental phases of AS formation (Fig 5D). After formation of the 2°NC cell cluster at the diencephalon/midbrain level (Phase 1, 18–22 hpf), between 20–30 2°NC cells migrate into the distal side of the eye (Phase 2, 22–28 hpf), attenuating expression of the NC marker sox10. The 2°NC cells differentiate into ocular mesenchymal cells (Phase 3, 28–60 hpf). The first corneal endothelial cells emerge on top of the lens around the 60–72 hpf stage (Phase 4, Fig 5A). The number of corneal endothelial cells increased and they eventually occupied the entire inner surface of the cornea (Phase 5, >72 hpf, Fig 5B and 5C).

It was unclear where sox10 is required for corneal endothelium differentiation. We thus analysed the structure of the cornea of homozygous colourless mutants (sox10^t3/t3) encoding a prematurely truncated Sox10 protein [30]. Whereas wild type or heterozygous siblings showed the normal structure of the cornea at 144 hpf (S10A nad S10B Fig), the homozygous sox10 (n = 4) mutants showed irregularly shaped cells with large vesicular structures in the space between the lens and the stroma of the cornea (S10C and S10D Fig), indicating the requirement of Sox10 for proper corneal endothelium formation at this stage.

To assess whether the corneal marker Tg(corneaEndo:GFP) required Sox10 activity, a sox10 morpholino directed against the sox10 transcription start site [31] was injected into Tg(corneaEndo:GFP) zygotes. Expression of the corneal endothelial marker was assessed at 5 dpf. Control embryos injected with a control morpholino developed 28.6 ± 3.6 GFP-positive endothelial cells (n = 25), whereas sox10 knockdown embryos show 12.8 ± 6.7 GFP-positive cells lining the inner surface of the cornea (n = 28; Welch Two sample t-test p-value = 1.3 x10^-13; S10G–S10I Fig). These cells lining the interior surface of the cornea appear more irregular as seen also in the electron micrographs of the mutant corneae. Like the sox10 mutant, sox10 knock-down embryos have a smaller eye. The sox10 morphants showed a smaller anterior chamber (S10G’–S10H’ Fig, stippled double-head arc with arrows) and annular ligament cells expanded toward the centre of the cornea (GFP-positive cells outside of the circle, AL; S10H Fig), which is normally confined to the peripheral regions (S10G Fig). These results are consistent with a requirement of Sox10 for proper development of the anterior chamber.

Pax6a/b dose correlates with ASD severity

pax6b mutant zebrafish embryos develop ASD [16]. Zebrafish have duplicated Pax6 genes. Pax6a and Pax6b share 92.5% amino acid sequence identity and have overlapping patterns of expression [32]. To assess the impact of pax6a/b on development of the anterior chamber, we embarked on a single and compound mutant analysis. We utilised the previously identified ENU (N-ethyl-N-nitrosourea)-induced pax6b mutant sunrise (tq253a allele) [32]. A pax6a mutant line was created by deleting exons 8–12, the C-terminal half encoding the homeobox and proline/serine/threonine-rich (PST) domains using CRISPR/Cas9 gene editing (Fig 6; S11A–S11D Fig). We first analysed the AS phenotype of various pax6a/b mutant combinations at 5 dpf when the AS is fully established [14,15]. The pax6a zygotic homozygous mutant showed a marginal ASD at 5 dpf with a slightly smaller eye and lens, and an underdeveloped iridocorneal angle (Fig 6C–6C’, arrows). In contrast, the pax6b maternal zygotic homozygous mutants presented a range of severe ASD classified into either keratoconus (Fig 6–6D’) or hypotonic deflated anterior chamber often associated with corneal hyperplasia (Fig 6E–6E’, arrow). The lens, smaller in size and occasionally with gaps between the epithelial cells and fibre cells (arrowheads, Fig 6D’–6E’), localizes at an abnormally distal position. Largely, AS phenotypes of pax6a/b single and double mutants supported an overall correlation between the degree of reduced pax6a/b gene copy number and the severity of ASD (Fig 6B–6I), as observed in mouse studies [11][33]. The distal side of the double mutant eyes was severely affected: it had no lens and was occupied with mesenchymal cells (Fig 6G and 6I, arrows; S7 Video), whereas the proximal side appeared less disturbed, with the exception of the overall smaller size of the eye. The corneal endothelium reporter was not detected in corneas of double mutants generated by injection of pax6b-/- embryos with pax6a gRNAs/Cas9 (Fig 6J–6K’; S8 Video). Thus, a differentiated endothelial layer has not been formed. pax6a gRNA-injected pax6b-/- embryos showed lensless eyes with 100% penetrance and were phenotypically comparable to pax6a-/-;pax6b-/- double homozygous mutants. The periocular mesenchymal cells observed in the AS of pax6a gRNA-injected pax6b-/- embryos retained abnormal expression of the mesenchymal marker gene pitx2 (Fig 6L–6M) but lacked expression of the mature corneal epithelium marker zgc:92380 (Fig 6N–6O) [16]. Taken together, this suggests that migration and differentiation of NC cells may not have occurred normally in the pax6a/b double mutants.

Pax6a/b genes are required for correct migration of 1° and 2°NC

We next investigated the effects of Pax6 gene knockout on NC cell migration. We analysed pax6a/b compound mutants carrying the NC reporter line Tg(sox10:h2a-tdEosFP) in the background. Photoconversion of the EosFP reporter at 17 hpf to the red colour allowed unambiguous identification of 1°NC and 2°NC cells by their maintained red and recovered green fluorescence, respectively (Fig 7A’; S9 Video). Normal NC behaviour was observed with pax6a+/+;pax6b+/+ (wild type), pax6a+/+;pax6b+/- (Fig 7A–7C; magenta and green lines for the 1°NC and 2°NC trajectories, respectively) and pax6a+/-;pax6b+/+ (see also Table 1). In contrast, in embryos with genotypes carrying a homozygous pax6b mutation, i.e., pax6a+/+;pax6b-/-, pax6a+/-;pax6b-/- and pax6a-/-;pax6b-/-, the 2°NC cell cluster did not dissolve but rather persisted at the distal-dorsal periocular site with high penetrance (Fig 7E–7E’, solid circle; Table 1). As a result, these cells became excluded from the eye (Fig 7E–7E’, stippled circle; S9 Video). Morpholino-mediated knockdown of pax6a/b mimicked this abnormal 2°NC cluster behaviour (S12 Fig). 2°NC cells failing to enter the eye maintain their sox10 expression (S12E and S12H Fig). In double heterozygous mutants, pax6a+/-;pax6b+/-, the 2°NC entry into the distal eye compartment was affected in nearly half of the embryos (39%, Table 1), suggesting that this pax6 gene number provides a borderline dose of Pax6 protein. In contrast to 2°NC cells, 1°NC cells were not affected except in the double homozygous pax6a-/-;pax6b-/- mutant: The 1°NC cells migrated far into the distal half of the eye of these double mutants (Fig 7F, magenta tracks in the dashed circle). Taken together, our data are consistent with a gene number dependent effect of pax6a/b on the migration of the two distinct waves of NC cells.

Pax6a/b genes are required for the expression of guidance molecules in the eye

Pax6 was proposed to convey positional information in the murine eye affecting the establishment of the three axis of the eye [34]. We thus examined markers indicative of the overall coordinates of the eye in pax6a/b compound mutants. For the sake of ease in crossing, we combined the zygotic pax6a-/- and the maternal zygotic pax6b-/- mutation for pax6a-/-;pax6b-/- double homozygous mutants. No change was observed between wild type, maternal zygotic pax6b-/-, and pax6a-/-;pax6b-/- double mutants with the dorsotemporal expression of the aldehyde dehydrogenase 1 family member A2 (aldh1a2, arrows in Fig 8A–8A”), the dorsal expression of T-box 5 (tbx5, arrows in Fig 8B–8B”), the proximoventral expression of pax2a (arrows in Fig 8D–8D”), the ventral expression of ventral anterior homeobox 2 (vax2, Fig 8E–8E”), or nasotemporal expression of foxg1a and foxd1 (S13 Fig). Also expression of pax6a and pax6b itself was not changed (Fig 8C–8C”; S11E and S11F”‘ Fig). As implied by the smaller size of the lens in maternal zygotic pax6b-/- mutants and its absence in pax6a/b double mutants, we observed a reduction of lens-specific expression of crystallin beta A4 (cryba4, arrows in Fig 8F–8F”) with abnormal lenticular patterning (S11E and S11F” Fig). Thus, Pax6a/b proteins are required for lens primordium formation. However, they are not required for dorsal (defined by aldh1a2 and tbx5), ventral (pax2a and vax2), nasal (foxg1a), temporal (aldh1a2 and foxd1) and proximal (pax2a) identity of the eye primordium.

Since the overall positional identities appear to be normal in the pax6a/b double mutant eye, we reasoned that pax6a/b may be required for correct expression of cell guidance molecules. In chick embryos, the expression of the soluble protein, semaphorin 3A in the lens epithelium is known to organize AS-contributing NCs by preventing them from prematurely migrating over the lens. [35]. While semaphorin 3Aa (sema3aa) expression was observed in the lens of pax6b-/- mutants (Fig 8G–8G’), its expression was absent in pax6a/b double mutants (Fig 8G”). Semaphorin 3A is recognized by plexins and their co-receptors neuropilins [35]. neuropilin 2b (nrp2b)-positive ocular mesenchymal cells found in the distal end of the eye (Fig 8H) were present in pax6b-/- embryos (Fig 8H’). They were, however, absent in pax6a/b double mutants (Fig 8H”). The expression of the semaphorin receptor plxna4 was reduced in the eye of both mutants (Fig 8I–8I”). We next examined ephrins and their cognate Eph receptors, both of which are involved in NC guidance. The dorsotemporal expression of the transmembrane ephrin ligand B2a [36] (efnb2a, arrows in Fig 8J) was detected in pax6b-/- mutants (Fig 8J’) but not in pax6a/b double mutants (Fig 8J”). In contrast, dorsonasal expression of a glycosylphosphatidylinositol-anchored ephrin ligand, efna3b [37,38] (Fig 8K), and of Eph receptor B3a (ephb3a, Fig 8L) were reduced in pax6b-/- mutants (Fig 8K’ and 8L’). Both markers were absent in pax6a/b double mutants (Fig 8K” and 8L”). In summary, Pax6a/b proteins are required for correct expression of guidance molecules in the ocular cup and its surroundings.

We hypothesized that this change in expression of guidance molecules affects the migration of the two NC cell populations. To test this notion, we injected morpholinos directed against the transcription start sites of sema3aa [39,40], efnb2a [41] and nrp2b [40], which are mainly expressed in the lens, lens/retina and NC cells, respectively, and monitored the movement of the NC cells. No significant effects on NC migration and AS formation were detectable upon knock-down of sema3aa and efnb2a. Abnormal NC migration and significantly reduced corneal endothelial cells were observed, however, when we knocked-down nrp2b (S14 Fig). The 1°NC cells of nrp2b-knockdown embryos migrated far into the distal half of the eye in comparison to the control knockdown (S14A Fig, n = 14 embryos), phenocopying the behaviour of 1°NC cells in pax6b-/-;pax6a-/- mutants (S14B Fig, arrow; n = 17 embryos). In contrast to the pax6a/b double mutant, 2°NC cells of nrp2b-knockdown embryos migrated normally into the anterior segment of the eye. Thus, altered nrp2b expression in NC cells of the mutants is only one aspect of the defects causing abnormal 1° and 2°NC migration. It also suggests that the two populations use different guidance cues. In accordance with the Pax6 double mutant phenotype the eyes were smaller. The endothelial layer was also reduced in number of cells with the characteristic flat morphology lining the inner surface of the cornea (Welch Two sample t-test p-value<2.2 x10^-16, S14E Fig). This reduction of endothelial cells, however, may reflect just the smaller size of the eye. The presence of the endothelial layer in the cornea is consistent with the migration of the 2°NC into the distal segment of the nrpb2 morphants. Taken together, these results are consistent with the notion that the changes in guidance molecule expression in Pax6 mutants is an important part of the mutant phenotype.

Transplantation of a lens into pax6a/b double mutants partially restores AS morphology

The lens has been proposed as an organising centre of the eye [42]. Pax6a/b double mutants lack the lens suggesting that the more severe defects of the double mutant may be linked to lack of the lens. We first characterized the defects of the anterior chamber of pax6a/b double mutants more closely. To visualize the size of the anterior chamber and the structure of the ocular blood vessels, we injected 2000 kDa FITC-dextran and 3 kDa rhodamine-dextran into the cardinal vein of 5 dpf larvae. The former dye remains inside the vessel while the latter diffuses across the vessel boundaries, allowing one to visualize the anterior chamber space. The pax6a/b double homozygous mutant has no anterior chamber and ocular superficial annular vessels were mislocalised (Fig 9D).

We next assessed the role of the lens in AS formation. When the lens was removed from the wild type eye at 24 hpf, the annular vessels were disorganised (Fig 9B), similar to the pax6a/b mutant eye (Fig 9D). To test whether the presence of a lens was sufficient to rescue AS formation in mutants, we performed lens transplantations [43]. When the wild type lens was transplanted into the eye of pax6a/b double homozygous mutants at 24 hpf, both the anterior chamber and the normal patterning of superficial annular vessels was restored (Fig 9C, n = 3 embryos). Thus, the absence of the lens in pax6a/b double homozygous mutants contributes to the ASD phenotype.

TGFβ signalling is required for corneal endothelium formation

Lens-derived TGFβ2 organizes periocular mesenchyme differentiation to form the AS in mice [44]. Moreover, TGFβ2 was shown to down-regulate key periocular mesenchymal genes (including Foxc1, Pitx2 and Slug) to initiate corneal endothelial differentiation in vitro [45].

We first analysed expression of the tgfb2 gene (Fig 10A–10G). tgfb2 gene was expressed at 22 hpf in the lens and also in mesenchymal cells found at the dorsal edge of the optic cup (arrowhead, Fig 10A). At 24 hpf, tgfb2 mRNA was expressed in mesenchymal cells in both the distal and proximal side of the eye (arrowheads, Fig 10B). Interestingly, at stages after 28 hpf, tgfb2 was not expressed in mesenchymal cells of the distal eye anymore (Fig 10C–10G). However, expression in the lens was detected until 48 hpf (Fig 10C–10G). Thus, tgfb2 expression is highly dynamic.

We next examined the expression of the TGFβ reporter construct (Fig 10H–10N). The TGFβ signalling reporter line Tg(smad3:GFP) utilizes Smad3 binding elements to drive GFP expression [46]. TGFβ signalling dependent gfp transcription was first observed at 24 hpf in the dorsal side of the retina (arrow, Fig 10I) and in mesenchymal cells on the proximal side of the eye (arrowheads, Fig 10I). At 28 hpf, the TGFβ reporter is expressed in the lens (arrow, Fig 10J) and the retina, however, it is not observed in the distal side of periocular mesenchyme until 72 hpf. At 72 hpf, when distally located periocular mesenchymal cells start to differentiate into Tg(corneaEndo:GFP) positive cells (Fig 10O; Fig 5D), the TGFβ signalling reporter became expressed in the distally located mesenchymal cells (arrowheads, Fig 10M). This distal mesenchymal TGFβ signalling was preceded by activation of lenticular TGFβ signalling at 48 hpf (arrow, Fig 10L), which ceased at 96 hpf (Fig 10N).

Next, we examined whether inhibition of TGFβ signalling affects the development of the AS in zebrafish. During 46–87 hpf (stages 3–5 in Fig 5D), when NC cells start to differentiate, Tg(corneaEndo:GFP) embryos were treated with either 1% DMSO or the TGFβ signalling inhibitor SB-505124 at 30 μM. SB-505124 abolishes ALK4/5/7 kinase activity, thereby blocking Smad2/3 phosphorylation [47]. Reporter expression for corneal endothelial cells was abolished in treated embryos (Fig 10P, n = 4 embryos; S10 Video). Thus, TGFβ signalling is required for corneal endothelial differentiation. We did not observe prolongation of sox10 expression in TGFβ inhibitor-treated embryos in contrast to the effects seen on mammalian NC cells in vitro [48].

Discussion

Here we have identified two distinct groups of cranial NC cells that contribute to the development of the eye. The 1°NC cells, which envelop the proximal side of the eye, originate from the lateral premigratory NC cells. The 2°NC cells emigrate form more dorsomedial positions at the level of the diencephalon and mesencephalon a couple of hours later. They colonize the AS of the developing eye. Multidimensional live imaging revealed that 2°NC can be distinguished from 1°NC by their origin, their timing of delamination, target area in the eye, migration route, Wnt-reporter activity and maintenance of sox10 expression. The entry of 2°NC into the eye is selectively blocked in pax6b-/- mutants. In pax6a/b double homozygous mutants, the behaviour of both NC populations was altered. The severity of ASD in pax6 mutants appears to be correlated with the reduction of the pax6 gene number. However, in contrast to the mouse, the proximodistal patterning of the eye was not affected in the zebrafish pax6 mutants. NC migration phenotypes are correlated with lack or reduction of guidance molecules as well as the lack of the lens in the compound mutants. Severe ASD in pax6a/b double homozygous mutants can be alleviated by transplantation of a wild type lens which is, as in mammals, the source of TGFβ-like signals. In agreement, we found that chemical inhibition of TGFβ signalling inhibited differentiation of endothelial cells.

Distinct waves of neural crest cells contribute to the morphogenesis of the eye

We traced the origin of the NC cells entering the eye to two distinct locations in the premigratory NC. The 1°NC cells colonising the proximal eye originate from the two lateral assemblies of premigratory NC cells described previously [21,49]. The 1°NC cells enter the eye posteriorly to cover the back side of the developing eye cup. The cells of the 2°NC group, which enter the eye dorsally and migrate to the distal side of the eye, originate from cells that reside close to the anlage of the pineal gland (dorsal diencephalon) and the dorsal mesencephalon. This region was mapped previously in uncaging experiments in zebrafish to contribute to the eye [50] and corresponds to the rostral limit of NC formation mapped in the newt Pleurodeles [51], chicken embryos [1], mouse [52] and human [53]. In mice, a group of periocular NC cells with a reduced neuropilin-1 (Npn-1) expression emerges from Npn-1 expressing NCs to migrate over the lens, i.e. into the distal end of the eye, and establish the corneal endothelium [54]. The 2°NC cell population is possibly related to this mouse NC population.

Both 1°NC and 2°NC cells share sox10 gene expression. However they differ in the maintenance of sox10 expression. sox10 mRNA expression is not detectable in NC cells once they have entered the eye. sox10:mem-tdEosFP transgene expression had ceased in 1° and 2°NC cells with their immigration into the eye at 13–14 hpf and 22–28 hpf respectively, However, EosFP fluorescent protein still persisted until around 48 hpf leaving a memory of former sox10 driven gene expression in the NC cell derivatives.

There are two peaks of Wnt reporter expression in ocular NC cells. The early peak (12–14 hpf) includes the precursors of 1°NC cells as previously described [21]. The second peak of Wnt reporter activity from 18 hpf onwards was noted in 2°NC cells. Prior to migration into the eye, half of sox10:mem-tdEosFP-positive 2°NC cells co-express the Wnt reporter. They continued to express the Wnt reporter after 2°NC cells entered the nascent anterior segment. The role of Wnt signalling in 2°NC cells remains to be determined. Blocking Wnt signalling with the Wnt signalling inhibitor IWR-1-endo led to a 50% reduction of premigratory sox10-positive 2°NC cells suggesting a role of Wnt signalling in the control of the number of premigratory 2°NC cells.

Both NC cell groups show distinct transformations of cell behaviour: 1°NC cells retain cell-cell contact over large cellular areas while enwrapping the back side of the eye and display a cuboidal cell shape. The 2°NC cells initially accumulate as a group of cells at the dorsal rim of the eye before entering the eye as individually migrating cells. This switch of 2°NC cells from clustering to migration as individual mesenchymal cells is correlated with their entry into the eye and requires pax6a/b as 2°NC cells failed to enter the eye in pax6a/b morphants/mutants. Similarly, aggregation of NC cells was observed at the dorsal edge of the nascent eye in rat Pax6-/- embryos [55]. Pax6a and Pax6b are not significantly expressed in 2°NC cells at this stage, indicating that Pax6a/b expression in the surrounding tissues is responsible for guidance, as previously also suggested for the abnormal migration of NC cells in Pax6 heterozygous mutant mice [7] and Pax6 homozygous mutant rats [55]. In agreement with these findings, we found an altered expression of cell guidance molecules in ocular and surrounding tissues in the zebrafish pax6a/b mutants.

We have focused on the development of the corneal endothelium as a well-defined NC-derived tissue of the eye. By 26–28 hpf, around 20 mesenchymal-shaped 2°NC cells have reached the nascent AS. Corneal endothelium formation and reporter expression commences only by 60 to 70 hpf, as judged by expression of the Tg(corneaEndo:GFP), with the first ten cells expressing the transgene in the corneal area juxtaposed to the lens. These cells then proliferate until on average ca. 27 cells cover the entire inner surface of the cornea by 120 hpf. Curiously, endothelial cells form concentric rings, which may be a reflection of their patterns of division and migration over the inner surface of the cornea.

The role of the lens and TGFβ signalling in shaping the anterior segment

Ectopically located 2°NC cells in pax6a/b mutants/morphants maintain high levels of sox10 transgene expression, suggesting that these cells have to enter the eye for turning off sox10 expression. This behaviour indicates that sox10 is not directly required for endothelial differentiation. However, when we examined the sox10^t3/t3 mutant colourless, the endothelium was not properly formed in this mutant, suggesting that sox10 is involved in corneal endothelial formation. Interference with TGFβ signalling prevents differentiation of the corneal endothelium. The requirement of TGFβ signalling is reminiscent of the effect of TGFβ signals in in-vitro cultures of mouse NC cells [48]. The zebrafish lens strongly expressed tgfb2 between 22 and 40 hpf, covering the time of immigration of NC cells into the AS and their differentiation into the various AS structures including the corneal endothelium. In addition expression was noted in mesenchymal cells in both the proximal and distal compartment of the eye. This was paralleled by the expression of the sensor of TGFβ signalling Tg(smad3:gfp). In this context, it may be of importance that the aqueous humour of the anterior chamber is a source of TGFβ molecules in the mammalian eye [56]. TGFβ molecules may thus be freely diffusible throughout the anterior chamber. Pax6a/b compound mutants lack the lens entirely and may thus influence anterior chamber formation by indirectly eliminating lens derived TGFβ. Our TGFβ inhibitor experiments support a necessary role of TGFβ in corneal endothelium development.

There is evidence of direct regulation of TGFβ2 by Pax6 from studies in mammals. In mouse, Pax6 binding sites were identified in the Tgfb2 promoter [57], and a ChIP-seq/transcriptome study reported up-regulated Tgfb2 transcripts in conditional Pax6 homozygous mutant lenses during embryonic stages [58,59], suggesting that Pax6 negatively regulates Tgfb2 transcription. We found that, when 2°NC cells reached the eye lenticular pax6b expression is polarized to the distal side, while tgfb2 and smad3:gfp reporter transcripts are found on the opposite, proximal side of the lens. This is in agreement with the hypothesis that the distal tgfb2 and smad3 reporter expression were repressed by pax6a/b. In contrast, a mouse study reported down-regulated Tgfb2 transcripts in Pax6(+/-) heterozygous mutant lenses at postnatal day 1 when the AS is already established [57]. Thus, whether Pax6 activates or represses its target might be controlled by stage specific co-regulators in the lens.

Removal of the lens from wild type eyes at 24 hpf phenocopied the defects in the annular vasculature seen in pax6a/b mutants at 5 dpf. When a wildtype lens was transplanted into a pax6a/b double mutant eye, the vasculature formed normally in the mutant, showing that a wildtype lens can direct vessel formation in a mutant background. Whether these effects are mediated by lens derived TGFβ signals remains to be seen. The trabecular meshwork is a key NC-derived anterior segment structure for intraocular pressure control, together with lymph-derived Schlemm’s canal. Although the homologue of the mammalian trabecular meshwork has not been described in zebrafish, the aqueous humour flows out through a ventral canalicular network considered to be of ocular mesenchymal origin [14]. Pax6a/b mutants appear to have defects in the blood/aqueous humour barrier, presumably linked to a failure to establish the annular vasculature correctly.

Pax6 mutations cause ASD in zebrafish without affecting the overall polarity of the eye

In mammalian studies of Pax6 function, it has been proposed that pax6 establishes distal identities such as the lens, the iris and the cornea in a dose-dependent manner within the eye [9,11]. In addition, Pax6 is recently shown to be required for TGFb2 expression in the chick optic vesicle to establish the eye’s first axis [60] and in the mouse iris/ciliary body [61]. The duplication of the genome at the base of teleost evolution has generated two pax6 alleles that have been retained with overlapping functions in the zebrafish genome, offering a unique opportunity to study the influence of pax6 gene number on eye development. In agreement with mouse studies, we noted a tendency of increasing phenotypic penetrance and severity of the anterior chamber defects with decreasing numbers of wild type pax6a/b copies in the zebrafish.

In concordance with mouse homozygous pax6 mutant phenotypes, zebrafish pax6a-/-;pax6b-/- double homozygous mutants showed severe defects in distal eye structures, lacking the lens, iris, corneal endothelium and ganglion cell layer of the retina. With the exception of the missing ganglion cell layer, the proximal side of the eye was relatively unaffected, featuring normal retinal pigment epithelium, photoreceptor layers, outer nuclear layer, outer plexiform layer, inner nuclear layer and inner plexiform layer. In contrast to mouse pax6 homozygous mutants [62], zebrafish compound mutants also formed nasal pits.

The milder phenotype in zebrafish may be the result of the specific mutant alleles. Missense mutations in human Pax6 lead to a variety of phenotypes ranging from mild to severe AS defects [63]. The pax6b mutant allele sunrise used in this study carries a missense mutation L244P in the homeobox DNA-binding domain. The penetrance of its mutant phenotype (smaller lens with elliptic shape and coloboma) was shown to be modulated by the chaperon Hsp90 [64]. In contrast to the pax6b/sunrise allele used here, the pax6a mutant allele harbours a premature termination codon, lacking the homeobox DNA binding domain and the PST-rich transactivation domain. Importantly, the phenotypic expression and, thus, the interpretation of the data with respect to subfunctionalization as well as gene dosage may also be hampered in addition by the presence of maternal pax6a/b mRNAs.

1°NC cells, which normally settle only in the proximal part of the eye, aberrantly invade distal locations in pax6a/b double mutant eyes. During early mammalian eye development, Pax6 appears to be a critical determinant of polarity along the proximodistal axis of the eye. Accordingly, the highest Pax6 levels and dependency are found in the distal part of the eye anlage [10,11]. As a consequence of Pax6 mutation, proximal markers expanded to the distal site of the mouse eye [12]. When we examined the zebrafish pax6a/b double mutants, we did not find consistent evidence for a disturbed patterning of the eye along the proximodistal axis. The reduction in lens specific markers may reflect a requirement for pax6 expression during lens development. Given the similarity in the phenotypes of the Pax6 mouse and zebrafish mutants, we uncovered a later layer of regulation, which had most likely been missed in mammals due to the more severe phenotype. It is tempting to speculate that, through the fortuitous choice of the pax6b/sunrise allele and the presence of maternal contributions of pax6a and pax6b in the zebrafish, the residual pax6 activity may have overcome the early patterning defects, thereby revealing the later direct function of pax6 in the development of the anterior eye segment.

Pax6a/b are required for expression of cell guidance molecules

NC cells are misrouted with high penetrance in pax6a/b double mutants: 1°NC colonise the entire eye instead of colonising the proximal half of the eye only as in wild type embryos. The 2°NC cells, which normally migrate into the nascent AS, are excluded from the eye in pax6a/b double mutants. Abnormal migration of NC cells was also observed in Pax6 heterozygous mouse mutants [7]. Pax6 was not detectably expressed in premigratory and migratory NC cells. Thus, misrouting of NC cell migration appears to be caused by non-cell autonomous factors. In agreement with these findings, we identified changes in the expression of cell guidance molecules in the eye and the surrounding tissue. Some guidance molecules such as efnb2a were affected only strongly in pax6a/b double mutants. Other guidance molecules such as efna3b and ephb3a also showed a reduction in expression in pax6b single mutants. These findings suggest that expression of these guidance molecules is sensitive to the number of wild type pax6 copies. This dependence of the expression of cell guidance molecules on Pax6 is reminiscent of the role of Pax6 in the expression of axon guidance molecules controlling proper axonal connections in the forebrain [65]. In the embryonic lens of mice, Pax6 directly regulates Semaphorin ligands including sema3aa [58,59] that prevent Nrp1-expressing NC cells from migrating over the lens during cornea development [54]. sema3aa expression is absent in pax6a/b double mutants. Abnormal migration of 1°NC cells in double homozygous pax6a-/-;pax6b-/- mutants may therefore be related to the lack of the lens and, therefore, early sema3aa expression, which normally prevents 1°NC cells from entering the distal eye. However, sema3aa cannot be the only factor: In a significant proportion of pax6b-/- embryos, ocular NC cells were also misrouted even though sema3aa was expressed. Moreover, knockdown of sema3aa translation did not show an effect on NC migration in the eye, suggesting either redundancy with other guidance cues or no function. Knocking-down nrp2b mRNA translation, however, lead to an eye phenotype resembling that of pax6a/b mutants. Although we did not observe an effect when we knocked-down sema3aa, the requirement of the semaphorin co-receptor neuropilin nrp2b indicates a role of Semaphorin signalling in the guidance of the NC cells into the eye.

Interestingly, 1° and 2°NC show differential requirement of nrpb2 function. Migration of 2°NC cells was not affected by nrpb2 knock-down suggesting distinct guidance mechanisms in the two populations. 2°NC cells enter through the dorsal side of the eye. In this context, it may be of relevance that the dorsal expression of efnb2a and efna3b is attenuated in pax6b single mutants and abolished in pax6a/b double mutants. In both genetic backgrounds, the migration of 2°NC cells is affected. Clearly, expression of these markers shows a dependence on the pax6a/b gene dose.

In conclusion, the severe ASD phenotype with lacking lens, corneal endothelium and vasculature in pax6a/b homozygous double mutants is tightly linked to the resultant misguidance of the two distinct ocular NC populations. We showed that Pax6 promotes guidance molecule expression in the ocular cup and its surrounding mesenchymal cells. Later defects in the anterior segment of the eye are attributable at least partially to the lack of the lens as an important source of further signals such as TGFβ. These findings may have implications for glaucoma development in congenital cataract patients whose lens are typically removed in the first few weeks after birth.

We have characterised in depth the development of the AS of the zebrafish eye and showed how these processes are disturbed in a zebrafish model for human ASD. The knowledge on NC function in eye development gained with this model may contribute to future therapeutic approaches. Given the ease with which zebrafish embryos can be exposed to drug candidates, our detailed description of zebrafish AS development and the new genetic and transgenic tools will pave the way to drug development. By exploiting the advantages of a whole organism approach offered by the zebrafish embryo, screening of this system will probably identify new chemical entities and drug targets representing previously unrecognised drugable components in the complex developmental pathways disturbed in ASD and congenital glaucoma.

Methods

Ethics statement

Animal husbandry and experimental procedures were performed in accordance with the German animal protection regulations and were approved by the Government of Baden-Württemberg, Regierungspräsidium Karlsruhe, Germany (AZ35-9185.81/G-137/10, AZ35-9185.81/G-22/15, AZ35-9185.81/G-102/19 and AZ35-9185.81/G-103/19).

Fish, transgenic lines

We used the wildtype line ABO (European Zebrafish Resource Centre [EZRC], Karlsruhe). Zebrafish (Danio rerio) embryos were maintained at 28°C as described [66,67]. The 7.2 kb sox10 regulatory element [19] was used to drive expression of Eos fluorescent protein (tandem dimer tdEosFP [18]). Tg(sox10:mem-tdEosFP) [official ZFIN name Tg(-7.2sox10:gap43-tdEosFP)^ka97] express tdEosFP N-terminally tagged with a zebrafish gap43-derived dual palmitoylation sequence for plasma membrane localization. Tg(sox10:h2a-tdEosFP) [official ZFIN name Tg(-7.2sox10:h2afva-tdEosFP)^ka96] harbour tdEosFP fused N-terminally with h2afva for nuclear localization. Transgenic lines were generated by injection of Tol2kit-based DNA constructs [68] into one-cell stage embryos, followed by screening for positive founder fish in the next generation. The Wnt reporter line Tg(7xTCF-Xla.Siam:nlsmCherry) [22] was crossed with each Sox10 reporter line and double-positive embryos were used for imaging. Homozygous mutant pax6b^{tq253a/tq253a} zebrafish were viable (the tq253a allele is also called sunrise), fertile and maintained as adult fish. pax6a mutants were created by deleting a ~3 kb genomic region across exon 8–12 with CRISPR/Cas9, as described in S11 Fig. Briefly, two short guide RNAs (sgRNAs) were transcribed in vitro by using the MEGAshortscript T7 transcription kit (Ambion) from double stranded DNA templates prepared as described [69], harbouring target sequences of 5’-GGT TGC CAA CAG TCA GAC GG-3’ for pax6a exon 8 and 5’-GGT CTG GCT GGG CTG TGA AG-3’ for exon 12. Two sgRNAs, each at 300 ng/μl final concentration, were incubated briefly with GeneArt Platinum Cas9 nuclease (300 ng/μl final; Invitrogen) and injected into the zebrafish one-cell stage blastodisc.

Genotyping

Genomic DNA was prepared by an adopted HotSHOT method [70], using a base solution (25 mM KOH, 0.2 mM EDTA pH13.8) and a neutralisation solution (40 mM Tris-HCl, pH3.8). A part of the tail fin or the whole embryo was incubated in 75 μl of base solution at 96°C for 20 min and cooled down to room temperature. 75 μl of neutralisation solution were added to the sample before KASP assay or genotyping PCR. KASP assay was performed to genotype pax6b mutants harbouring a single base mutation in the pax6b locus. The following three primers, two-allele specific forward primers (wild type and mutant alleles) and a common reverse primer, were synthesized by LGC Limited (UK): (wild type allele labelled with FAM: 5’-AAA GTT GTG ATC GTT CAC CTT TCT CAA-3’; mutant allele labelled with HEX 5’-AGT TGT GAT CGT TCA CCT TTC TCA g-3’ [mutation in small letter]; common primer: 5’-CGT CCT TCA CGC AGG AGC AGA T-3’). KASP assay mix (0.11 μl), KASP master mix (4 μl), 1 M MgCl₂ (0.64 μl) and genomic DNA (1 μl) were assembled per reaction in a 96-well plate (AB-0800/K, Thermo Fisher Scientific). After PCR, fluorescence was measured by an EnVision plate reader (PerkinElmer) and analysed by KlusterCaller software (LGC Limited, UK). To genotype pax6a mutants, we designed two sets of primers, one for detecting the deletion and another for the presence of the wild type allele. A set of pax6a_Ex8-12_L (5’-GTG GCC GTA GCC TAA TTG AA-3’) and pax6a_Ex8-12_R (5’-GCC ATT GAT TGG CTG TTC AT-3’) was designed to yield a 405 bp amplicon for identifying the mutant allele, and another set pax6a_Ex8_L (5’-GTG GCC GTA GCC TAA TTG AA-3’) and pax6a_Ex8_R (5’-AGT GCG ATT CCT TTG CAG TT-3’) was targeted to yield a 384 bp amplicon for identifying the presence of the wild type allele.

Morpholinos

Following antisense morpholino oligos were synthesized by Gene Tools, LLC (OR, USA) and injected into one-cell stage embryos at concentrations indicated in parenthesis: 1ATG_pax6ab (250 μM): 5’-GTTATGGTATTCTTTTTGAGGCATT-3’; 2ATG_pax6ab (250 μM): 5’- ACTGTGACTGTTTTGCATCATGGAC-3’; 1ATG_pax6ab_5-mismatch (250 μM): 5’- GTTAaGcTATTCTTaTTcAcGCATT-3’ (small letters indicate the mutated sites in mismatch morpholinos);

2ATG_pax6ab_5-mismatch (250 μM): 5’- ACTcTcACTcTTTTcCATCATcGAC-3’; MO2-nrp2b (125 μM) [40]: 5’- CGCGTAGAGGAAAAAGCTGAAGTTC-3’; MO1-sox10 (250 μM) [31]: 5’- ATGCTGTGCTCCTCCGCCGACATCG-3’; MO4-tp53 (62.5 μM) [71]: 5’- GCGCCATTGCTTTGCAAGAATTG-3’

In situ hybridization

Whole-mount in situ hybridisation was conducted as described [72]. Antisense digoxigenin-labelled RNA probes for otx5 [73], wnt2 [74], sox10, pitx2, pax6b, foxc1a, zgc:92380 [16], sox9b, tgfb2, gfp, efna3b, epha4b, aldh1a2, efnb2a, vax2, and sema3aa were prepared from cloned DNA constructs. For double fluorescent in situ hybridisation, a set of two probes, each labelled either with digoxigenin-11-UTP (Roche) or DNP-11-UTP (Perkin Elmer), was used during hybridisation (day 1). After stringency washing and blocking in 1% BSA/PBS/0.1% Tween-20, embryos were incubated with a 1:1000 dilution of anti-digoxigenin-POD (poly) Fab fragments (Roche; day 2) and labelled with a 1:250 dilution of TSA Plus Cy3 stock solution (Perkin Elmer) for 45 min, then further incubated in PBS/0.1% Tween-20 (PBT) at 4°C overnight (day 3). After a PBT wash, embryos were treated with 3% H₂O₂ for 2 h to eliminate the activity of probe-bound peroxidase. Embryos were extensively washed in PBT and incubated with a 1:500 dilution of anti-DNP-HRP (Perkin Elmer) at 4°C overnight (day 4). The second fluorophore labelling was done with a 1:250 dilution of TSA plus Cy5 (Perkin Elmer; day 5). For nucleus counterstaining, we applied incubation on a 1:1000 dilution of DAPI in PBS/0.5% Triton-X100 at 4°C overnight (day 6).

Histology

Epoxy resin (EPON) embedding and sectioning were done as described [16]. Briefly, the embryos were dehydrated through a series of increasing concentrations of ethanol (50%, 70%, 95%, 100% [v/v]), followed by 100% propylene oxide and a series of increasing concentrations (33%, 66%, 100%) of EPON (glycid ether 100, Serva) in propylene oxide. Polymerisation was carried out at 65°C in the presence of 20.8% (w/w) DDSA (dodecenylsuccinic acid anhydride, Serva) and 23.3% (w/w) MNA (methylenacid anhydride, Serva), with the accelerator 1.8% (w/w) DMP 30 (2,4,6-tris[dimethyl-aminomethyl]phenol, Serva). The polymerised block was trimmed (Leica EM TRIM) and 6 μm and 350 nm sections were cut with glass and diamond knives (DiATOME ultra 45°) using RM2065 and UC7 microtomes (Leica Microsystems), respectively.

Intravenous dye injection

Zebrafish embryos at 5 dpf were anesthetized in 643 μM MS-222 (tricaine methanesulfonate) and embedded in 0.8% low melting point agarose laterally with the right side up. To minimize the effect of low blood pressure on the aqueous humour dynamics, embryos were immersed in fish water to wash out the anaesthetic after hardening of the agarose. A mixture of 2000 kDa FITC-labelled dextran and rhodamine-labelled 3 kDa dextran was transferred into a glass capillary and 10–20 nl was injected into the cardinal vein in the vicinity of the heart by a micro-injector (Eppendorf FemtoJet) as described [75].

Multidimensional real-time imaging

Measurements were performed on a custom-built digital scanned laser light sheet fluorescence microscope (DSLM) setup that allows multi-channel image acquisition at high temporal and space resolution assisted by bi-directional sample illumination [23]. The spectrum for illumination included lasers of 405 nm (diode laser module, Omicron LuxX, 60 mW), 488 nm (diode laser module, Omicron LuxX, 60 mW) and 561 nm (DPSS laser, Cobolt Jive, 75 mW). With the 5×/0.15 (illumination) and 16×/0.8w (detection) objective lenses (both Nikon, Japan), the lateral and axial full width at half maximum of the point spread function resulted in 1.5 ± 0.3 μm and 6.6 ± 1.4 μm at 0.4 μm and 2.0 μm voxel size in the lateral and axial directions, respectively. Embryos were mounted in cleared fluorinated ethylene propylene tubes in 0.3% (w/v) low-melting point agarose. The pipeline of algorithms for NC cell tracking has been described earlier [25].

Confocal microscopy images were acquired with a Leica TCS SP5 upright microscope with an HCX PL APO 20x/0.70 lambda blue IMM CORR objective. Time lapse imaging was conducted in resonant scanning mode (8 kHz, bidirectional scanning). For photoconversion, filter cube A (excitation bandpass filter, 360 ± 40 nm) was selected and the sample was illuminated for 20 s. The temperature was set to 28°C. Embryos were mounted in 0.5% (w/v) low-melting point agarose in embryo medium (50 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl₂, 0.33 mM MgSO₄, 0.7 mM HEPES pH7.0). Co-localization of fluorescent signals was analysed in ImageJ “Image Calculator” by using the “Multiply” operation between two channels after background subtraction.

Inhibitors

SB-505124 (Sigma-Aldrich), an inhibitor of TGFβ type I activin receptor-like kinase, was used at 30 μM. IWR-1-endo was obtained from Dr. Lawrence Lum. As a negative control, embryos were treated with 0.5% (v/v) DMSO in embryo medium.

Statistical analyses

Statistical significance was assessed by using R [76]. All numerical data underlying graphs and statistics are summarized in S2 Method.

Supporting information

S1 Video [m4v]
Time-lapse movie of a : embryo during 10–33 hpf.

S2 Video [m4v]
Confocal time-lapse of : transgenic embryo during 15–24 hpf.

S3 Video [white]
Wnt reporter activation in NC cells.

S4 Video [m4v]
Overlay of cell trajectories on raw microscopic images.

S5 Video [mp4]
Trajectory path comparison of 1°NC and 2°NC cells during 11–24 hpf.

S6 Video [m4v]
Locally labelled 2°NC cells migrate into the distal compartment of the eye.

S7 Video [rpe]
Histological reconstitution of the eye at 5 dpf.

S8 Video [m4v]
are required for normal corneal endothelium formation.

S9 Video [m4v]
Pax6a/b are required for guiding the two ocular NC populations to their distinct destinations.

S10 Video [m4v]
TGFβ signalling inhibitor blocks the formation of the corneal endothelium.

S1 Fig [oc]
1°NC cells cover the proximal but not the distal side of the optic cup.

S2 Fig [a]
2°NC cells turn off expression upon migration into the eye.

S3 Fig [a]
Wnt signalling is required for the correct number of 2°NC cells but is dispensable for formation of the corneal endothelium.

S4 Fig [pdf]
Backward and forward tracking from a second embryo.

S5 Fig [a]
Early origin of ocular NC cells.

S6 Fig [c]
2°NC cells show distal-oriented movement after dissolving the cluster.

S7 Fig [green]
Lineage of proximal and distal destined NC cells.

S8 Fig [d]
2°NC cells arise from the dorsal diencephalon and mesencephalon.

S9 Fig [pdf]
Mapping of the transgene insertion site of ..:.

S10 Fig [arrow]
Absence or reduced Sox10 activity causes abnormal AS structures.

S11 Fig [a]
The mutant locus created by CRISPR/Cas9.

S12 Fig [a]
antisense morpholino oligonucleotide injection mimics the genetic mutant phenotype.

S13 Fig [pdf]
Nasotemporal polarity of the eye is not affected in pax6a/b double mutants.

S14 Fig [green]
Nrp2b is required for correct guidance of NC cells and corneal endothelium formation.

S1 Method [docx]
Morpholinos and transmission electron microscopy.

S2 Method [txt]
All numerical data underlying graphs and statistics.