A Transcriptomic Analysis of Differential Gene Expression during Chick Periocular Neural Crest Differentiation into Corneal Cells

Abstract

Background:

Multipotent neural crest cells (NCC) contribute to the corneal endothelium and keratocytes during ocular development, but the molecular mechanisms that underlie this process remain poorly understood. We performed RNA-Seq analysis on periocular neural crest (pNC), corneal endothelium, and keratocytes and validated expression of candidate genes by in situ hybridization.

Results:

RNA-Seq profiling revealed enrichment of genes between pNC and neural crest-derived corneal cells, which correspond to pathways involved in focal adhesion, ECM-receptor interaction, cell adhesion, melanogenesis, and MAPK signaling. Comparisons of candidate NCC genes to ocular gene expression revealed that majority of the NCC genes are expressed in the pNC, but they are either differentially expressed or maintained during corneal development. Several genes involved in RA, TGFβ, and Wnt signaling pathways and their modulators are also differentially expressed. We identified differentially expressed transcription factors as potential downstream candidates that may instruct expression of genes involved in establishing corneal endothelium and keratocyte identities.

Conclusion:

Combined, our data reveal novel changes in gene expression profiles as pNC differentiate into highly specialized corneal endothelial cells and keratocytes. These data serve as platform for further analyses of the molecular networks involved in NCC differentiation into corneal cells, and provide insights into genes involved in corneal dysgenesis and adult diseases.

INTRODUCTION

The cornea is a multilayered transparent tissue comprised predominantly of cells derived from a multipotent embryonic cell population, the neural crest cells (NCC). Cranial NCC progenitors of the cornea originate from the neural tube region encompassing the caudal diencephalon and the rostral metencephalon (Lwigale et al., 2004; Creuzet et al., 2005b). This population of NCC forms different streams that migrate into the frontal-nasal, periocular, and maxilla-mandibular regions (Noden, 1975; Johnston et al., 1979; Serbedzija et al., 1992) where they respond to environmental cues and differentiate into region-specific tissues. Periocular neural crest cells (pNC) contribute to various ocular structures including the cornea, eyelids, and connective tissues (Lwigale et al., 2004; Creuzet et al., 2005b). Defects in pNC migration, proliferation, and differentiation are associated with a condition known as anterior segment dysgenesis (ASD) (Cook, 1989). ASD is comprised of spectrum of ocular disorders characterized by malformation of the cornea, iris, lens, and eyelids (Churchill and Booth, 1996; Sowden, 2007). Despite the significance of NCC to eye development, very little is known about the molecular underpinnings of their differentiation into ocular tissues.

During chick corneal development, pNC occupy the mesenchyme surrounding the rudimentary eye for approximately two days prior to their initial migration into the presumptive corneal region at about embryonic day (E)4.5. Migration of pNC into the presumptive corneal region occurs in two waves (Hay and Revel, 1969; Lwigale et al., 2005). The first wave forms the corneal endothelium, an interior monolayer of cells that establishes a barrier and regulates fluid movement between the anterior chamber and the cornea (Waring et al., 1982). The second wave of migration occurs at about E6, when pNC invade the acellular primary stroma and differentiate into keratocytes, which synthesize the extracellular matrix of the corneal stroma (Linsenmayer et al., 1984; Quantock and Young, 2008). The intricate behavior of pNC during chick corneal development is in part regulated by the lens vesicle, given that its ablation results in precocious migration and malformation of the cornea (Beebe and Coats, 2000; Lwigale and Bronner-Fraser, 2009). The molecular signals involved in regulating pNC migration were pinpointed to lens-derived Semaphorin3A and the spatiotemporal downregulation of its receptor, NRP1, by a subset of pNC, which permitted their migration into the presumptive corneal region (Lwigale and Bronner-Fraser, 2009). Studies in mice have identified that retinoic acid (RA) signaling regulates transcription factors expressed by pNC including Pitx2 and Foxc1 that inhibit canonical Wnt signaling via upregulation of Dkk2 (Lehmann et al., 2003; Evans and Gage, 2005; Gage et al., 2008). Knockout of any one of these genes in mice phenocopy the corneal and iridial defects observed in humans with ASD (Gage et al., 1999; Kitamura et al., 1999; Kume and Seo, 2010). Together, these studies have advanced our understanding of early corneal development, but the molecular mechanisms that transform pNC into the diverse progeny of ocular cells remain unclear.

In this study, we take advantage of the stepwise contribution of avian pNC to the nascent cornea and survey their gene expression profile during differentiation into corneal endothelium and keratocytes by RNA-Seq analysis. We evaluated changes in expression profile of candidate NCC markers following aggregation in the periocular region. We studied changes in gene expression profiles of components of the major signaling pathways (RA, TGFβ, and Wnt) associated with ocular development. We identified genes that are likely to be involved in pNC differentiation into corneal endothelium and keratocytes. Altogether, these data serve as a foundation to advance our understanding of the molecular mechanisms underlying pNC migration, proliferation and differentiation.

RESULTS AND DISCUSSION

Generation of a comprehensive transcriptome during pNC differentiation into corneal cells

During early ocular development in chick, the status of pNC differentiation can be categorized into three phases: (1) aggregation of NCC into the periocular region, which occurs between E2-E3; (2) migration of pNC into the presumptive corneal region to form the endothelial layer by E5; and (3) migration of pNC into the primary corneal extracellular matrix to form the stromal keratocytes by E7. At each of the above stages, NCC are subjected to different environmental cues that play important roles in their migration, proliferation, and differentiation (Brugmann et al., 2006; Lwigale and Bronner-Fraser, 2009). Our analysis of NCC contribution to early ocular development using the quail/chick chimera technique indicates that the periocular mesenchyme at E3 is mostly comprised of quail-derived QCPN-positive pNC that give rise to the corneal endothelium at E5, and to the keratocytes and endothelium at E7 (Fig. 1A).

Figure 1. Experimental design for tissue isolation, RNA-Seq analysis, and comparison of genes identified in pNC, En, and KEn.

(A) Localization of neural crest cells in the periocular region and their corneal derivatives revealed by immunostaining cross sections of quail-chick chimera eyes for QCPN (red) at various stages of ocular development. (B) Periocular neural crest cells (pNC), corneal endothelium (En), or the combined keratocytes and corneal endothelium (KEn) were isolated at E3, E5, and E7, respectively. (C) PCA plot showing that the triplicate samples from pNC (Red), En (Green), and KEn (Blue) cluster together. Principal components 1 and 2 summarize 89% of the system variance. (D) Bar graph showing numbers of differentially expressed genes (red indicates upregulated and blue indicates downregulated genes). (E) Histogram representation indicating the number of DEGs between pNC-En, pNC-KEn, and En-KEn belonging to significant KEGG pathways. Abbreviations: ec, ectoderm; oc, optic cup; L, lens; ep, epithelium; en, endothelium; st, stroma. Scale bar: 100 μm.

To identify changes in gene expression during pNC formation corneal cells in chick, we isolated periocular mesenchyme from E3 (pNC), the monolayer of corneal endothelium at E5 (En), and the combined stroma and endothelium at E7 (KEn) (Fig. 1B). Three biological triplicates for each time point were prepared for RNA-Seq analysis as described in the methods section. Principal-components analysis (PCA) was applied to all mapped genes from the 9 samples to determine the reproducibility of biological repeats by Noiseq. The Scatter plots indicate good separation between pNC, En, and KEn, as well as good clustering of the three biological repeats of each sample, with system variance of 89% (Fig. 1C).

Identification of differentially expressed genes between pNC and corneal cells

To identify differentially expressed genes, we performed further analysis using a cutoff of abs(log2)≥1 and FDR<0.05 on genes that overlapped between NoiseqBio and EdgeR analyses (Supplementary Table 1). Based on these analyses, 3160 genes were differentially expressed between pNC and En (Fig. 1D). The 2101 upregulated genes represent En and the 1059 downregulated genes represent pNC. Likewise, 3800 genes were differentially expressed between pNC and KEn (Fig. 1D), of which 2307 upregulated genes represent KEn and the 1493 downregulated genes represent pNC. Given that both En and KEn contain corneal endothelial cells, we analyzed the difference between these two groups. Differentially expressed genes between En and KEn (2096) may either be involved in keratocyte differentiation or associated with further development of the corneal endothelium between E5 and E7. In this group, 846 genes are upregulated and 1250 genes are downregulated. KEGG pathway analysis indicated enrichment of similar pathways between pNC/En and pNC/KEn, with the top five corresponding to focal adhesion, ECM-receptor interaction, cell adhesion, melanogenesis, and MAPK signaling (Fig. 1E). A complete KEGG pathway list that includes the major signaling pathways during ocular development (RA metabolism, Wnt, and TGFβ signaling) is provided in Supplementary Table 2.

Changes in NCC gene expression in the periocular region and cornea

Given that NCC reside in the periocular region for approximately two days before their initial migration into the presumptive cornea (Hay and Revel, 1969), we investigated whether they change their molecular signature prior to, and during their differentiation into corneal cells. A list of 47 genes was generated from candidate chick NCC genes (Simoes-Costa et al., 2014; Simoes-Costa and Bronner, 2015), and we examined their expression in pNC, En, and KEn. From this list, 43 NCC genes were expressed in pNC, of which 20 were downregulated, 7 were upregulated, and 10 were constitutively expressed in both En and KEn. Interestingly, some genes (6) were either downregulated or maintained in only En or KEn (Fig. 2A and Supplementary Table. 3). Of the 4 NCC genes that were not expressed in the pNC (ADAM11, DLX6, ZIC3/4), ADAM11 was upregulated during corneal development. These results indicate that the molecular signature of NCC is maintained in the periocular region, and partially during early corneal development. This is supported by our previous study, showing that keratocytes isolated from E10 quail corneas were capable of differentiating into other NCC-derived tissues when grafted into the migratory stream of stage 9 chick embryos (Lwigale et al., 2005). To validate the expression of NCC genes, we chose one gene from each category based on whether they were downregulated (MSX2), maintained (SNAI2), or upregulated (RHOB) during corneal development and analyzed their spatiotemporal expression by in situ hybridization. Our results revealed that MSX2 and SNAI2 are expressed in the periocular region (Fig. 2B and 2C). Unlike MSX2, which is restricted to the periocular region, expression of SNAI2 is maintained in the corneal endothelium (Fig. 4C, arrowhead) and stroma. In mouse, Msx1 and Msx2 are involved in NCC survival, proliferation, and differentiation (Liu et al., 1999; Ishii et al., 2005), and it is possible that some of the early functions are maintained only in the chick pNC. SNAI2 is involved in the epithelial-mesenchymal transition required for NCC delamination from the neural tube (Taneyhill et al., 2007). We show that expression of SNAI2 is maintained when pNC undergo mesenchyme-endothelial transition to form the corneal endothelium, which strongly expresses N-cadherin (Reneker et al., 2000; Lwigale et al., 2005). Inhibition of E-cadherin by SNAI2 requires LMO4 (Ochoa et al., 2012), which is downregulated in our RNA-Seq data (Fig 2A and Supplementary Table. 3), suggesting that SNAI2 may play a different role that does not affect cell adhesion molecules in the corneal endothelium.

Figure 2. Changes in the molecular identity of NCC during their localization in the periocular region and differentiation into corneal cells.

(A) Analysis based on 47 candidate NCC genes indicates that 43 genes were expressed and 4 were not expressed in the periocular region. Of the 43 expressed genes, 20 were downregulated, 10 were constitutively expressed, and 7 were upregulated during corneal development. (B-D) Section in situ hybridization of E3, E5, and E7 anterior eyes indicating that: (B) MSX2 is strongly expressed in the periocular region but undetectable in the corneal endothelium and stroma; (C) SNAI2 is maintained at all three time points; and (D) RHOB is minimal in the periocular region, but it is strongly expressed in the corneal endothelium and stroma. Arrow indicates periocular region and arrowheads indicate corneal endothelium. Abbreviations: pNC, periocular neural crest; ec, ectoderm; oc, optic cup; L, lens; ep, epithelium; en, endothelium; st, stroma. Scale bars: 100 μm.

Figure 4. TGFβ Signaling pathway.

(A) Black boxes indicate constitutively expressed genes, red boxes indicate upregulated genes, and blue boxes indicate downregulated genes. Double colored boxes indicate partial upregulation, downregulation, or maintenance during corneal development. (B-D) Section in situ hybridization of E3, E5, and E7 anterior eyes showing that: (B) LTBP1 is lightly expressed in the periocular mesenchyme at E3 and it is strongly expressed in the corneal endothelium and stroma. (C) BAMBI is expressed in the optic cup and low in the periocular mesenchyme and it is not detectable in the corneal endothelium and stroma. (D) Expression of TGFβi is also low in the periocular mesenchyme at E3, but it is vividly expressed in the corneal endothelium at E5 (arrowhead) and stroma at E7. (E) Heatmap showing the relative upregulation and downregulation of TGFβ and BMP target genes between En/pNC and KEn/pNC. Font color of genes indicates TGFβ- or BMP-upregulated (red) or TGFβ- downregulated (blue) genes. Abbreviations: pNC, periocular neural crest; ec, ectoderm; oc, optic cup; L, lens; ep, epithelium; en, endothelium; st, stroma. Dotted lines represent the cell and nuclear membranes. Scale bars: 100 μm.

From our list of NCC genes, only CDH11, CSRNP1, ERG, LIMS1, RHOB, TFAP2β, and MYC are upregulated during corneal development. As an example from this group, we validated the spatiotemporal expression of RHOB. Our data indicate that RHOB is expressed by a few cells in the periocular region (Fig. 2D, arrow), then it is strongly expressed in the corneal endothelium (Fig. 2D, arrowhead), and in both the corneal endothelium and stroma at E7. During early development, RHOB is expressed in the dorsal neural tube and transiently in migratory NCC, and it is involved in the delamination process (Liu and Jessell, 1998). The function of RHOB during corneal development remains unclear, but it has been shown to play a critical role during barrier formation in vascular endothelial cells (Marcos-Ramiro et al., 2016). Our data also show upregulation of TFAP2β, which suggests its potential role during chick corneal development. Conditional knockout of TfAp2β in the NCC lineage in mice caused several defects in NCC-derived ocular tissues including malformation of the corneal endothelium and stroma (Martino et al., 2016).

Differentiation of pNC is regulated by multiple signaling pathways

Signals from the neural plate, ectoderm, and adjacent mesoderm mediate NCC formation at the neural plate border with the ectoderm. These include growth factors such as Wnt (Saint-Jeannet et al., 1997), bone morphogenetic protein (BMP; Liem et al., 1995), fibroblast growth factor (FGF; Monsoro-Burq et al., 2003), and RA signaling (Villanueva et al., 2002; Martinez-Morales et al., 2011). These signaling pathways regulate NCC formation at all embryonic axial levels, except for RA signaling, which is absent in the rostral cranial NCC streams that contribute to the eye (Maden et al., 1998). pNC experience similar signals from the adjacent ocular tissues (optic cup, lens, and ectoderm) prior to their migration into the presumptive corneal region. RA signaling is from the optic cup and ectoderm (Li et al., 2000; Matt et al., 2005; Molotkov et al., 2006), whereas TGFβ and BMP are expressed in the lens and optic cup (Trousse et al., 2001; Saika et al., 2002). Several WNTs are expressed in the ectoderm, optic cup, and lens (Jin et al., 2002; Fokina and Frolova, 2006). Interestingly, disruption of these major signaling pathways lead to dysgenesis of NCC-derived ocular tissues (Gage et al., 1999; Kitamura et al., 1999; Kume and Seo, 2010), but the mechanisms of how they are regulated and their action on pNC are not clearly understood. Our KEGG pathway analysis also indicates that the above signaling pathways change during corneal development (Supplementary Table 2), thus, we analyzed the expression of RA, TGFβ, and Wnt signaling components and their downstream targets.

RA signaling pathway

Genetic studies in mice showed that three retinaldehyde dehydrogenase enzymes (Raldh1/2/3) responsible for RA synthesis are all required for proper development (Dupe et al., 2003; Molotkov et al., 2006; Maden, 2007). Our data indicate high-level expression of components of the RA signaling pathway during pNC differentiation into corneal cells (Fig. 3A and Supplementary Table 4). These include transporter genes (STRA6, CRABP1/2 and RBP5, FABP5), enzymes (DHRS3, RDH5/12/14, and ALDH1A2), regulator (CYP26B1), and nuclear receptors (NR2F1/2, NR2C1, RXRΑ, RARΑ, and RARΒ, PPARD). Of these genes, ALDH1A2, DHRS3, and CYP26B1, are significantly upregulated in En and KEn compared to pNC (Fig. 3A and Supplementary Table 4). Upregulation of ALDH1A2 and DHRS3 was confirmed by in situ hybridization to determine the specific localization of RA synthesis and inhibition, respectively, in the cornea. Although ALDH1A2 is strongly expressed in the optic cup at E3, it is not expressed in the pNC. Consistent with our RNA-Seq data, ALDH1A2 is vividly expressed in the corneal endothelium at E5 and E7 (Fig. 3C, arrowheads), with sparse staining in the stroma. Similarly, expression of DHRS3 is minimal in the pNC at E3, but robust in the corneal endothelium at E5 and E7 (Fig. 3D, arrowheads), and also expressed in the stroma. Our data show for the first time that the chick corneal endothelium expresses ALDH1A2 and thus, acts as a potential source for RA signaling, which may act in either an autocrine or paracrine fashion to regulate corneal development. Interestingly, our results also suggest that the RA inhibitors DHRS3 and CYP26B1 (Sakai et al., 2001; Feng et al., 2010; Kam et al., 2013) may act at the same time to modulate the levels of signaling in the corneal endothelium and stroma. This is in conjunction with previous observation that RA signaling intensified in various ocular tissues including the cornea of Cyp26a1 null mice (Sakai et al., 2004).

Figure 3. RA signaling pathway.

(A) Black boxes indicate constitutively expressed genes, red boxes indicate upregulated genes, and blue boxes indicate downregulated genes. Double colored boxes indicate partial upregulation, downregulation, or maintenance during corneal development. (B-D) Section in situ hybridization of E3, E5, and E7 anterior eyes showing that: (B) ALDH1A2 is expressed in the optic cup but not the periocular mesenchyme at E3, but it is vividly expressed in the corneal endothelium (arrowhead), and sparsely expressed in the stroma at E7. (C) Expression of DHRS3 is ubiquitous but low at E3, but becomes strong in the corneal endothelium (arrowhead) and stroma. (D) NR2F2 is expressed in the periocular mesenchyme at E3 (arrow), but it is not detectable in the corneal endothelium and stroma at E5 and E7. (E) Heatmap showing the relative upregulation and downregulation of RA target genes between En/pNC and KEn/pNC. Font color of genes indicates RA-upregulated (red) or RA-downregulated (blue) genes. Abbreviations: pNC, periocular neural crest; ec, ectoderm; oc, optic cup; L, lens; ep, epithelium; en, endothelium; st, stroma. Dotted lines represent the cell and nuclear membranes. Scale bars: 100 μm.

Downregulated RA genes in En and KEn include CRABP1, NR2F1 and NR2F2 (Fig. 3A and Supplementary Table 4). Downregulation of CRABP1, which promotes RA degradation (Dong et al., 1999; Michalik and Wahli, 2007), may cause an increase in nuclear RA mediated by CRABP2 that is constitutively expressed. NR2F1 and NR2F2 (also known as COUP-TF1 and COUP-TF2) are transcription factors that function as RA nuclear receptors (Kruse et al., 2008; Pickens et al., 2013), but they have also been shown to inhibit RA signaling by binding to retinoic acid response elements (RAREs) of downstream genes (Kliewer et al., 1992; Tran et al., 1992). Our in situ hybridization analysis confirms that NR2F2 is expressed in the pNC (Fig. 3E, arrow), but it is not detectable in the corneal endothelium or stroma (Fig. 3E). A similar pattern of downregulation is expected for the expression of NR2F1, which may indicate a potential increase in expression of RA-target genes in the corneal endothelium and stroma.

Activation of RA signaling leads to upregulation or downregulation of target genes (Balmer and Blomhoff, 2002; Matt et al., 2005). Examples of RA-target genes in the cornea include transcription factors that are either upregulated (PITX2, DKK2, PAX6, TFAP2β) or downregulated (SOX9, MSX1, OTX2, LMO2) (Fig. 3A and 3E, and Supplementary Table 4). Pitx2, Foxc1, Pax6 play important roles during mouse corneal development (Smith et al., 2000; Kanakubo et al., 2006; Gage et al., 2014), and they are among the ASD genes linked to dysgenesis of the human cornea (Sowden, 2007).

TGFβ pathway

The TGFβ superfamily comprises of TGFβs, BMPs, and other ligands, which regulate differentiation, proliferation, migration, and apoptosis in numerous cell types including the NCC (Shah et al., 1996; Chai et al., 2003; Wurdak et al., 2005). Comparison of En and KEn to pNC indicated that TGFβ1/2/3 and BMP4/7 are expressed during corneal development. TGFβ2 was upregulated and BMP5 was the only gene downregulated in En and KEn (Fig. 4A and Supplementary Table 5). In mice TGFβ2 signaling from the lens induces expression of Pitx2 and Foxc1 in the NCC mesenchyme of the presumptive cornea, and its absence causes severe ocular defects including corneal thinning and absence of the corneal endothelium (Saika et al., 2001; Ittner et al., 2005). Ectopic expression of BMP5 in the chick neural retina transforms it into retinal pigment epithelium (Steinfeld et al., 2017), but its function in the pNC remains unclear.

The following regulators of TGFβ (CRM1, THBS1, LTBP1/2, DCN, TGIF1/2, SMURF1/2, RBX1, SKP1 AND CUL1) are expressed during corneal development. Of these, THBS1, LTBP1/2, and DCN are upregulated in En and KEn (Fig. 4A and Supplementary Table 5), indicating their potential function during corneal development. Thbs1 activates TGFβ and inhibits angiogenesis (Good et al., 1990; Schultz-Cherry et al., 1994; Tolsma et al., 1997). In the adult cornea, THBS1 is expressed in the corneal endothelium and stroma (Hiscott et al., 2006), where it plays a role as an antiangiogenic agent during wound healing (Hiscott et al., 1999; Matsuba et al., 2011; Blanco-Mezquita et al., 2013). LTBP1 is expressed in a few cells in the periocular region, but it is robust in the corneal endothelium and stroma (Fig. 4B). LTBP1 plays an important role during secretion, deposition, and activation of TGFβ in glioma cells (Tritschler et al., 2009) and it is possible that it plays similar roles during corneal development. DCN inhibits TGFβ signaling by binding directly to its active form (Yamaguchi et al., 1990; Border et al., 1992), and it has also been shown to prevent fibrillogenesis of corneal collagen and to cause fibrosis during wound healing (Rada et al., 1993; Mohan et al., 2010). Therefore, upregulation of DCN in En and KEn indicates potential modulation of TGFβ signaling, which might be required for activation of specific downstream genes during corneal development.

Our data indicate that BAMBI is the only TGFβ regulator that is downregulated in En and KEn (Fig. 4A). BAMBI is a pseudo-receptor with similar structure to TGFβR1 that inhibits TGFβ signaling and activates canonical Wnt signaling (Onichtchouk et al., 1999; Lin et al., 2008). We confirmed that BAMBI is expressed in the pNC (Fig. 4C, arrowhead) where it may play a role in Wnt signaling, but it is not detectable in the corneal endothelium and stroma, which correspond with the upregulation of TGFβ signaling in the cornea.

The TGFβ superfamily receptors TGFβR1/2, BMPR1A/2, and downstream components including SMAD1/2/4/5/6 are all constitutively expressed in pNC, En, and KEn (Fig. 4A and Supplementary Table 5). Several transcription factors downstream of TGFβ signaling are either upregulated (MYC, TWIST2, FOS) or downregulated (SOX9, LIN28A). Surprisingly, some TGFβ activated transcription factors such as SNAI1, TBX3, SOX9, and LIN28A are downregulated, whereas some that are inhibited by TGFβ (IRX3 and MYC) are upregulated (Fig. 4E and Supplementary Table 5). Our data also indicate that majority of the transcription factors activated by BMP signaling such as ID1/2/3, MSX1, and HEY1 are downregulated during corneal development (Fig. 4A and 4E and Supplementary Table 5). An example of a TGFβ downstream target that is upregulated in En and KEn is TGFβ-induced (TGFβI, also known as BIGH3), which was confirmed to be expressed in the corneal endothelium (Fig. 4D, arrowhead) and stroma. TGFβI encodes an ECM protein induced by TGFβ that interacts with collagen (Skonier et al., 1992; Hashimoto et al., 1997). Recent studies have shown that mutations in TGFβI in humans cause a condition known as granular corneal dystrophy, which is characterized by opaque deposits in the corneal stroma (Kattan et al., 2017; Nielsen et al., 2017; Chao-Shern et al., 2018). Therefore its upregulation in our data suggests that TGFβI may play a role in organizing collagen fibrils synthesized by the corneal endothelium and keratocytes during early development.

Wnt signaling

Defects in Wnt signaling pathway cause dysgenesis of the anterior ocular tissues in humans and mice (Ittner et al., 2005; Reis and Semina, 2011; Bankhead et al., 2015). Analysis of the components of the Wnt signaling pathway indicates that despite the expression of several ligands in chick ocular tissues (Fokina and Frolova, 2006), only WNT2B, WNT4, WNT5A, WNT6, WNT9A, and WNT9B are highly expressed in the pNC and its corneal derivatives during development (Fig. 5A and Supplementary Table 6). Genes representing regulators of Wnt signaling at multiple levels including FRZB, DACT1/2, DKK3, SFRP1/2, BAMBI, AXIN1/2, NLK, GROUCHO, WLS, AES, and CTBP1 are also constitutively expressed in pNC, En and KEn. In this group, only WLS and BAMBI mediate Wnt signaling and the rest are inhibitors. We confirmed by in situ hybridization that both WLS and AES are constitutively expressed in the pNC, corneal endothelium, and stroma (Fig. 5B and 5C). Wnt receptors and co-receptors including FRZ1/2/6/7/9 and LRP5/6 are also constitutively expressed. FRZ8/10 are upregulated in En but downregulated in KEn, and FRZ4 is downregulated during corneal development (Fig. 5A and Supplementary Table 6). Downstream components of Wnt signaling such as DVL1/3, GSK3Β, and CTNNB1 are either upregulated or constitutively expressed, whereas TCF7, TCF7L1 and LEF1 are downregulated. Combined, our data suggest that the pNC, corneal endothelium and keratocytes have the potential for canonical Wnt signaling. However, the expression of multiple inhibitors including the upregulation of DKK2 suggest that Wnt signaling is regulated at multiple levels during corneal development. We therefore examined the expression of Wnt target genes in En and KEn and observed that some Wnt-upregulated genes are downregulated (SOX17, FST, JAG1, STRA6, SALL4, RET, TBX3), whereas others are upregulated (NOS2, MYC, AXIN2, FN1, CYR61, PITX2, MMP2, LBH, WISP1) in our data (Fig. 5E and Supplementary Table 6). However, CLDN1 is upregulated in En but downregulated in KEn, and SP5 is downregulated in En but upregulated KEn. Our data also show that some Wnt-downregulated genes (TNFRSF1, CDH1) are upregulated during corneal development. As an example of a Wnt target gene, we confirmed the expression of PITX2 and observed vivid expression in the pNC, En, and KEn (Fig. 5D). Mouse studies have shown that Pitx2 is activated by Wnt signaling (Kioussi et al., 2002; Briata et al., 2003), but also indirectly acts as a Wnt inhibitor via upregulation of Dkk2 (Gage et al., 2008).

Figure 5. Wnt Signaling pathway.

(A) Black boxes indicate constitutively expressed genes, red boxes indicate upregulated genes, and blue boxes indicate downregulated genes. Double colored boxes indicate partial upregulation, downregulation, or maintenance during corneal development. (B-D) Section in situ hybridization of E3, E5, and E7 anterior eyes showing that: (B) WLS and (C) AES are both expressed in the periocular mesenchyme at E3, and in the corneal endothelium and stroma. (D) PITX2 is strongly expressed in the periocular mesenchyme, corneal endothelium, and stroma. (E) Heatmap showing the relative upregulation and downregulation of Wnt target genes between En/pNC and KEn/pNC. Font color of genes indicates Wnt-upregulated (red) or Wnt-downregulated (blue) genes. Abbreviations: pNC, periocular neural crest; ec, ectoderm; oc, optic cup; L, lens; ep, epithelium; en, endothelium; st, stroma. Dotted arrows indicate implied connection in the literature. Scale bars: 100 μm.

WNT11, the main ligand for the Wnt/PCP pathway, is upregulated in KEn, and WNT5A, which functions in both the Wnt/calcium and Wnt/PCP pathways (Yamanaka et al., 2002; Sato et al., 2010), is maintained at all stages of corneal development. Components of the Wnt/PCP pathway (DAMM2, MAPK10, ROCK2, PRICKLE2, RHOA, RAC1, CDC42) and Wnt/calcium pathway (NFACTC1, PRKCA, CAMK2D, RYK, CAMK2B) are either upregulated or constitutively expressed. These results raise the possibility that non-canonical Wnt signaling occurs during corneal development and it may be involved in inducing cell migration and polarity.

Differential expression of transcription factors during early corneal development

Our data indicate that differential expression of components of the major signaling pathways (RA, Wnt, TGFβ) correspond with the transformation of pNC into corneal endothelium and keratocytes. To identify the genes involved in the process of pNC differentiation into corneal cells, we examined the changes in expression of transcription factors (Table 1 and Supplementary Table 7). Upregulated genes with known ocular functions include PAX6 and ZNF469. PAX6, which is induced by RA signaling, is expressed in various ocular tissues and it has been shown to play a major role in eye development (Hill et al., 1991; Fujiwara et al., 1994; Baulmann et al., 2002). Analysis of NCC-derived cells in Pax6^sey/+ mice showed abnormal migration into the eye (Kanakubo et al., 2006). ZNF469 is involved in ECM synthesis and defects in this gene cause connective tissue disorders in humans including an ocular defect known as brittle cornea syndrome (Zlotogora et al., 1990; Al-Hussain et al., 2004; Vincent et al., 2014). Based on these examples, the upregulated genes may represent transcription factors involved in key processes of corneal development including cell migration, proliferation, and differentiation.

Table 1.

Top 20 differentially expressed transcription factors in En/pNC and KEn/pNC

Gene ID	Symbol	Log2FC (En/pNC)	q-value	Gene ID	Name	Log2FC (KEn/pNC)	q-value
Upregulated				Upregulated
101750712	ZNF750	9.86	2E-18	100857422	POU3F4	8.77	1E-121
418669	SHOX	8.22	2E-224	100858556	DMRT2	7.09	2E-135
396546	HMX1	8.03	2E-19	426886	TWIST3	6.69	5E-142
396346	SOHO-1	7.85	5E-20	418669	SHOX	6.67	5E-128
100858556	DMRT2	7.59	3E-160	101748749	ZNF469	6.12	6E-108
419863	IRF6	6.99	2E-30	101750712	ZNF750	5.46	7E-05
395943	PAX6	6.69	4E-14	107051987	POU3F3	5.29	3E-36
100857422	POU3F4	5.65	2E-58	396158	MYBL1	5.28	2E-226
424652	GLIS1	5.34	1E-167	395405	TWIST2	5.12	0
777244	SHOX2	5.30	1E-91	396346	SOHO-1	4.86	1E-08
395714	BARX2	4.69	9E-19	423315	NPAS3	4.56	2E-156
101748749	ZNF469	4.66	3E-71	777244	SHOX2	4.34	2E-58
395590	DLX3	4.53	6E-19	373990	TBX2	4.10	1E-240
421416	FOSL2	4.45	6E-61	768789	FOXE1	4.08	1E-43
374101	SCX	4.05	2E-194	100859610	CREB3L1	3.86	2E-120
429506	GRHL2	3.98	3E-22	396210	NFIA	3.79	3E-159
373990	TBX2	3.97	8E-228	419631	TFAP2E	3.61	3E-82
428651	GRHL1	3.71	3E-41	395713	TFAP2B	3.50	2E-240
423923	EMX2	3.64	9E-114	427171	PIK3R1	3.44	3E-113
419631	TFAP2E	3.42	1E-74	374101	SCX	3.41	1E-139
Downregulated				Downregulated
418313	TBX15	−10.53	7E-38	418313	TBX15	−10.53	2E-41
373932	NKX3–2	−8.69	6E-65	107053847	DLX2	−10.44	5E-65
396383	ISL1	−7.84	2E-15	428534	SOX17	−10.25	4E-37
396391	GATA5	−7.35	8E-12	419106	GATA3	−9.68	6E-123
396129	SIM1	−7.33	2E-30	724086	SIX2	−9.55	3E-137
396109	FOXD1	−7.00	2E-119	428021	ZIC2	−9.30	4E-94
419106	GATA3	−6.89	2E-100	395573	SOX10	−8.93	5E-147
503512	FOXL2	−6.75	1E-90	396109	FOXD1	−8.18	1E-138
395752	NKX6–1	−6.60	9E-50	395191	OTX2	−7.86	2E-19
395176	EBF2	−6.12	5E-72	395633	MYF5	−7.81	4E-16
395794	FOXD3	−5.98	3E-49	395794	FOXD3	−7.69	1E-61
374137	MEOX2	−5.93	2E-36	421786	LIN28B	−7.65	0
374127	PAX3	−5.74	2E-81	374127	PAX3	−7.46	7E-103
107053847	DLX2	−5.68	3E-47	503512	FOXL2	−7.46	4E-102
395795	FOXD2	−5.57	5E-133	396391	GATA5	−7.35	1E-13
419388	PRDM16	−5.56	7E-155	395752	NKX6–1	−7.19	5E-57
100857760	FOXF2	−5.54	5E-100	395596	EPAS1	−7.05	8E-174
395961	NR5A2	−5.42	5E-25	373932	NKX3–2	−6.98	8E-67
724086	SIX2	−5.35	4E-94	395977	GFI1B	−6.97	1E-25
428021	ZIC2	−4.68	4E-57	374137	MEOX2	−6.93	8E-44

Bold values indicate differentially expressed genes between the two groups (abs(log2)≥1 and q < 0.05). Minus signs indicate downregulation. The individual TPM and standard deviations for the RNAseq replicates are listed in Supplementary Table 7.

The downregulated genes represent transcription factors that are only expressed in the pNC (Table 1 and Supplementary Table 7). Among these are some of the bona fide NCC genes such as FOXD3, SOX9, and SOX10 (Fig. 2A). FOXD3 is required for self-renewal and maintenance of NCC multipotency (Labosky and Kaestner, 1998; Dottori et al., 2001; Mundell and Labosky, 2011), thus its downregulation may be required during pNC differentiation into corneal cells. SOX9 plays an early role during NCC induction (McKeown et al., 2005), but it is later required for chondrogenesis (Zhao et al., 1997; Sahar et al., 2005). Therefore, it is possible that the expression of SOX9 in the pNC may promote the formation of orbital bones, whereas its downregulation in En and KEn might be involved in preventing ossification of the cornea. Similarly, SOX10 is required during NCC induction, but it is later involved in their differentiation into non-mesenchymal derivatives including melanocytes and peripheral neurons (Southard-Smith et al., 1998; Dutton et al., 2001). Downregulation of SOX10 in En and KEn raises the possibility that both the melanocyte and neural lineages are prevented in the cornea.

Also, among the downregulated genes are representatives of transcription factors not expressed by cranial NCC, therefore we considered them specific for the periocular mesenchyme. pNC that do not form the cornea give rise to various skeletomuscular derivatives and also contribute pericytes to the ocular vasculature (Creuzet et al., 2005a). This is reflected in the pNC transcription factors that play important roles in osteogenesis, myogenesis, and vasculogenesis. For example, HEY1 is a downstream target of Notch signaling that is involved in vasculogenesis and osteoblastic differentiation (Fischer et al., 2004; Salie et al., 2010). DACH2 is required during eye and limb development in drosophila (Mardon et al., 1994), and it synergizes with EYA2 during myogenesis in chick (Heanue et al., 1999). CITED2 is a negative regulator of Hif-1 that is required for normal lens development and regression of the hyaloid vasculature (Chen et al., 2009).

Identification of genes involved in pNC differentiation into corneal endothelium

The corneal endothelium is comprised of polarized cells that form tight junctions and transport fluids between the cornea and anterior chamber, which is critical for maintaining stromal deturgescence and transparency (Waring et al., 1982). Our data indicate that most genes involved in membrane transport, cell polarity, and cell junctions are upregulated in both En and KEn (Table 2 and Supplementary Table 8). Examples of membrane transporter genes that are upregulated in the corneal endothelium include AQP1 and ABCB5. AQP1 is a bi-directional water transporter (Sui et al., 2001), which indicates potential increase in fluid absorption during pNC differentiation into corneal cells. ABCB5 is an ATP-dependent transporter that is also a well-known marker for limbal stem cells (Chen et al., 2005; Ksander et al., 2014).

Table 2.

Differential expression of corneal endothelium related genes in En/pNC and KEn/pNC.

Gene ID	Symbol	Log2FC (En/pNC)	q-value	Log2FC (KEn/pNC)	q-value
Membrane transporters
420384	AQP1	9.24	0	10.34	0
420606	ABCB5	8.04	8E-114	7.12	5E-89
417849	BEST3	7.29	4E-34	6.78	1E-27
776258	SLC35F4	6.21	4E-60	7.21	1E-91
769904	FAM26E	5.65	3E-167	7.19	2E-268
395443	SLC1A3	−5.16	7E-135	−4.38	7E-121
428653	KCNS3	−4.43	2E-11	−4.76	7E-14
396532	SLC4A1	−4.34	1E-19	−7.32	5E-35
770954	KCNK2	−4.26	1E-70	−5.07	1E-90
414871	KCNH6	−3.27	4E-29	−0.78	0.0012
Cell polarity
378897	THY1	3.70	6E-180	4.87	1E-293
Tight junction
395144	CLDN4	7.16	6E-20	2.57	0.0308
374029	CLDN3	5.76	1E-18	0.72	0.3202
429506	GRHL2	3.98	3E-22	−0.13	0.8131
424910	CLDN1	3.00	1E-05	−1.79	0.0112
769245	CLDN19	2.93	1E-07	2.80	2E-07
420022	RAMP2	−3.27	2E-32	−5.47	3E-59
374028	CLDN5	−2.73	2E-29	−3.35	3E-41
396047	SNAI1	−1.61	9E-10	−0.85	0.0011
422858	PPP2R2C	−1.38	6E-05	3.56	2E-54
420481	MPP7	−1.23	7E-18	−1.49	7E-26
Gap junction
374196	PDGFA	4.78	7E-35	3.18	5E-17
395771	GJB6	3.61	9E-11	0.53	0.4574
374128	PDGFB	1.93	9E-12	0.86	0.0035
395581	HTR2B	1.81	0.001	3.40	6E-11
408035	EGF	1.75	5E-04	−2.32	0.0024
430991	ADRB3	−3.71	3E-11	−3.08	1E-09
771678	GUCY1A2	−3.47	8E-26	−4.19	6E-35
396502	GJA5	−3.04	2E-11	−7.27	2E-35
404529	GJA4	−2.11	1E-12	−5.78	2E-43
422407	GUCY1A3	−2.08	1E-14	−5.46	2E-50
Adherens junction
414745	CDH2	3.31	1E-100	1.59	4E-27
414849	CDH13	3.14	0.444	3.61	2E-15
423718	CDH23	2.37	0.107	1.68	3E-10
419222	CDH4	1.90	9E-13	−0.45	5E-09
415797	CDH11	1.49	5E-36	0.01	1E-36
374068	CDH5	−8.37	4E-32	−4.73	3E-97
374007	CDH7	−6.90	9E-61	−1.72	1E-79
Expressed in adult corneal endothelium*
770869	MGARP	6.35	0	5.75	0
424391	MYOC	5.00	5E-23	1.07	0.0282
428221	COL8A2	3.12	2E-69	3.69	1E-92
396221	LDHA	1.88	2E-24	3.26	1E-65
418425	CD200	−3.08	3E-42	−4.47	3E-71
396092	ALCAM	−2.00	1E-41	−1.57	5E-28
374110	PTGDS	−1.44	0.009	−0.41	0.4664

Bold values indicate differentially expressed genes between the two groups (abs(log2)≥1 and q < 0.05). Minus signs indicate downregulation. The individual TPM and standard deviations for the RNAseq replicates are listed in Supplementary Table 8.

^*Chng et al., 2013.

The characteristic hexagonal shape of the corneal endothelium is due to forces that straighten the actomyosin fibers on the apical surface that faces the anterior chamber. The apical surface is also marked by ZO-1, whereas the basal surface that is in contact with the Descemet’s membrane is labeled by α3β1 (He et al., 2016). Interestingly, although ZO-1 (TJP1) is a major component of the adult corneal endothelium (Petroll et al., 1999; Ramachandran and Srinivas, 2010), it was not upregulated in our data sets (Supplementary Table 8). Suggesting that its expression and function may arise at a later time point than the stages of corneal development used for this study. Our data indicate that THY1, which is involved in cell polarity, is upregulated (Table 2). THY1 is predominantly expressed in neurons, where it is involved in cell attachment, migration and polarization (Tiveron et al., 1994; Leyton et al., 2001; Kong et al., 2013). THY1 is also localized on apical surfaces of epithelial cells (Powell et al., 1991), but it has not been studied in the corneal endothelium.

Genes involved in cell junctions include CLDN4, CLDN3, AND CDH2 (N-CADHERIN). Of these genes, CLDN3 is involved in tight junction barrier function in the submandibular glands (Mei et al., 2015; Yokoyama et al., 2017), and it is possible that it plays a similar role in the corneal endothelium. CDH2 mediates cell adhesion in neural cells and it is a well-known marker for corneal endothelial cells (Reneker et al., 2000). Its upregulation indicates the potential for formation of intercellular bonds at the onset of endothelial differentiation.

We also detected genes that are highly expressed in the adult corneal endothelium (Chng et al., 2013). These were either upregulated (MGARP, MYOC, COL8A2, LDHA) or constitutively expressed at all three time points (VIM, TSPAN6, TPJ2, ATP1B1, TPT1, PRDX6) (Table 2 and Supplementary Table 8).

Identification of genes involved in pNC differentiation into keratocytes

Keratocytes are characterized by their ability to synthesize the stromal ECM, which is comprised of collagens and proteoglycans that constitute more than 90% of the cornea (Knupp et al., 2009; Hassell and Birk, 2010). Our data show that genes for several ECM-related proteins are upregulated during corneal development (Table 3 and Supplementary Table 9). A majority of the upregulated corneal ECM genes belong to collagen (Supplementary Table 9). They represent fibrillar collagens (COL1/2/5/6/11/27) required for the corneal structure, and nonfibrillar collagens, which are either associated with the collagen fibrils (COL9/14/16) or expressed on cell surfaces (COL4/8/17) (Shaw and Olsen, 1991; Kadler et al., 1996; Knupp et al., 2009). Collagens form the bulk of the corneal stroma and associate with ECM proteins and regulatory enzymes that align and space the structural fibrils in configurations that elicit transparency (Rada et al., 1993; Michelacci, 2003; Hassell and Birk, 2010). Our data also show that genes for several proteins expressed in the adult cornea are upregulated in En and KEn. They include glycoproteins (NTN3, VIT, SPON2, OLFML2B, and THSB2), proteoglycans (OGN[Mimecan], KERA, and DCN), other ECM proteins (EGFL6, NPNT, PTN, and LTBP1/2), and regulatory enzymes (MMP27, ADAMTS2/5/8, and COLGALT2). We also identified some ECM genes that are significantly downregulated in the En and KEn (VWF, VTN, HAPLN1, LEPREL1, HPSE2 and CSGALNACT1). Given the functional differences between the corneal endothelium and keratocytes (Waring et al., 1982; Linsenmayer et al., 1984; Quantock and Young, 2008), we were surprised that many of the ECM related genes were expressed by both cell-types. One possibility is that some of these genes are transiently expressed in the corneal endothelium, and this could be resolved by comparing their localization at later stages of development. It is also possible that most of the ECM proteins are intracellularly degraded upon synthesis in the corneal endothelium (Ko and Kay, 2001; Lee et al., 2012), and do not contribute to the corneal matrix.

Table 3.

Differential expression of ECM related genes in En/pNC and KEn/pNC

Gene ID	Symbol	Log2FC (En/pNC)	q-value	Log2FC (KEn/pNC)	q-value
Collagens
428448	COL6A6	2.72	2E-17	11.03	0
396292	COL6A2	6.41	0	10.60	0
396000	COL6A1	4.51	8E-137	7.71	1E-306
416696	LOC416696	8.85	9E-29	7.14	4E-21
101747382	LOC101747382	4.86	6E-120	7.04	2E-218
107056318	LOC107056318	0.19	0.763	-4.18	3E-11
421873	COL19A1	0.16	0.7209	−3.16	9E-14
101751793	LOC101751793	−1.79	5E-13	−2.93	4E-29
422350	COL4A6	−0.11	0.5729	−2.03	2E-32
422348	COL4A5	0.00	0.9913	−1.60	2E-21
Glycoproteins
396387	NTN3	1.75	4E-07	6.03	7E-127
421471	VIT	10.87	5E-30	5.67	2E-11
422905	SPON2	3.62	4E-44	5.61	5E-96
424366	OLFML2B	5.59	0	5.41	0
414837	THBS2	2.97	7E-99	4.69	3E-215
419031	VWF	−2.75	8E-25	−10.87	6E-102
395935	VTN	−5.57	3E-35	−5.70	4E-39
417180	LAMC3	−3.01	4E-67	−5.70	2E-161
427850	RELN	−2.02	5E-14	−5.26	4E-53
395531	NID1	−1.49	1E-48	−3.49	2E-212
Proteoglycans
374039	OGN	6.04	2E-295	8.13	0
373995	KERA	4.86	1E-160	6.42	1E-242
107056545	LOC107056545	4.55	1E-05	6.20	1E-08
417892	DCN	2.58	3E-33	4.25	4E-77
395798	ACAN	0.23	0.6479	3.46	5E-20
414143	LEPREL1	−3.90	7E-71	−6.99	1E-128
396475	HAPLN1	−2.51	8E-18	−6.31	1E-70
770863	GPC3	−1.29	3E-08	−4.51	8E-64
419184	SDC4	0.01	0.9846	−2.37	2E-16
Other ECM proteins
418637	EGFL6	4.31	8E-96	8.18	3E-307
422535	NPNT	0.65	0.0021	6.25	1E-151
418125	PTN	4.32	2E-290	5.52	0
421461	LTBP1	2.17	2E-76	3.21	3E-153
428888	LTBP2	2.23	1E-66	2.82	7E-104
395912	MGP	0.69	0.0939	−3.23	4E-10
Enzymes and Regulators
395850	MMP27	8.15	3E-158	9.07	8E-209
416291	ADAMTS2	0.82	0.0511	7.57	1E-120
769222	ADAMTS8	5.08	8E-271	4.54	8E-219
424447	COLGALT2	2.36	3E-44	4.28	3E-131
427971	ADAMTS5	5.14	1E-174	4.09	3E-112
422488	CSGALNACT1	−1.49	2E-08	−3.95	1E-34
423834	HPSE2	−3.12	2E-19	−1.42	2E-05

Data is arranged to show the top 5 upregulated and downregulated genes in each category. Bold values indicate differentially expressed genes between the two groups (abs(log2)≥1 and q < 0.05). Minus signs indicate downregulation. The individual TPM and standard deviations for the RNAseq replicates are listed in Table 12.

CONCLUSIONS

We have used RNA-Seq analysis to provide an unbiased depiction of gene expression profile of NCC in the periocular region and during their differentiation into corneal endothelium and stromal keratocytes. We confirmed the expression patterns of candidate genes at in the periocular region and during corneal development by in situ hybridization. We highlighted minor deviations in gene expression following aggregation of cranial NCC in the periocular region, suggesting that pNC maintain a level of multipotency that enables their contribution to various ocular structures later in development. We focused on three major pathways (RA, Wnt and TGFβ) involved in corneal development and identified for the first time that several modulators are upregulated during pNC differentiation into corneal cells, which could be critical in directing the expression of downstream transcription factors with potential roles in establishing endothelial and keratocyte identities. Previous studies have shown that there is crosstalk between these pathways during corneal development. The RNA-Seq data provides a useful tool for generating gene regulatory networks for pNC migration, proliferation, and differentiation. These data not only provide a foundation for identifying genes with novel function during corneal development, they also provide a valuable resource for future studies of corneal diseases, stem cell biology, and bioengineering of corneal tissues.

EXPERIMENTAL PROCEDURES

Animals

Fertilized White Leghorn chicken (Gallus gallus domesticus) and Japanese quail eggs (Coturnix coturnix japonica) were obtained from commercial sources and incubated at 38°C under humidified conditions. Quail-chick chimeras were generated to track cells of neural crest origin in ocular tissues by grafting dorsal neural tube explants from HH stage 9 (Hamburger and Hamilton, 1951) quail donors to stage-matched chick embryos as previously described (Lwigale et al., 2005). Chick and chimeric embryos were incubated until E3, E5, or E7. Embryos were manipulated according to protocols approved by the Institutional Animal Care and Use Committee (IACUC) at Rice University.

Tissue collection from chick embryos

Periocular neural crest (pNC) mesenchyme was obtained by dissecting the anterior half of the eyes from E3 embryos. Corneal endothelium (En) tissues were obtained by trimming E5 corneas at boundary with the periocular region. Combined keratocytes and corneal endothelium (KEn) tissues were obtained by trimming E7 corneas at the boundary with the limbus region. All tissues were incubated in dispase (1.5 mg/ml; Worthington) at 38°C for 5 minutes, then rinsed twice in Ringer’s solution. For the pNC, the ectoderm, lens vesicles, and optic cups were removed and discarded, and the periocular mesenchyme from 26 eyes was pooled into each sample. For the En, the corneal epithelium was removed and discarded, and the endothelial layers from 120 eyes were pooled into each sample. For the KEn, the corneal epithelium was removed and discarded, and the stroma and endothelium from 10 corneas were pooled into each sample. The KEn samples included the corneal endothelium to capture any changes in gene expression at this time point. Three biological repeats were made for each group for a total of 9 samples. All samples were immediately immersed in Trizol (Life Technologies), flash frozen in liquid nitrogen, and shipped in dry ice to BGI.US® for library preparation and sequencing.

RNA Sequencing and Mapping

Sequencing library of each sample was prepared according to Illumina Standard Protocols. High-throughput sequencing was performed by Illumina HiSeq™4000 (Single Read, 50bp) and generated an average of 47,074,369 raw sequencing reads from each sample. Raw reads were filtered by removing reads with adaptors, removing reads with more than 10% unknown bases, and removing low quality reads (if more than 50% of base has Phred quality score (Q score) <5). An average of 46,928,426 clean reads with Q30 >95% (Richterich, 1998) were generated for all nine samples for further analysis. Clean reads were mapped to Gallus gallus reference assembly Galgal5 and annotated gene model (GenBank: GCA_000002315.3) using Bowtie 2 (Langmead and Salzberg, 2012) and HISAT (Kim et al., 2015). The average mapping ratio with reference gene was 76.26% and average genome mapping ratio was 93.10%. The mapped reads were then counted by RSEM (Li and Dewey, 2011). Transcripts Per Kilobase Million (TPM) values were generated by normalization of counts with library size and gene length (Wagner et al., 2012).

Data access

RNA-Seq data sets have been deposited in the in the NCBI’s GEO database (https://www.ncbi.nlm.nih.gov/geo/browse/). Accession number (GSE120434).

DEGs and Pathway Analyses

Differentially expressed genes (DEGs) were screened out by EdgeR (Robinson et al., 2010) and NoiseqBio (Tarazona et al., 2011; Tarazona et al., 2015). Each pair (En/pNC, KEn/pNC, or KEn/En) were compared and DEGs chosen based on the following standards: (1) CPM (count per million) greater than 1 in at least three samples; (2) Log2 fold change greater than 1 in both EdgeR and NoiseqBIO; (3) FDR (corrected p-value, q-value) less than 0.05 in EdgeR and Probability (equal to 1-FDR) more than 95% in NoiseqBio. To reduce false positive rate, overlapped (common) genes from EdgeR and NoiseqBio were defined as DEGs for further analysis (Supplementary Table 1). KEGG pathway analysis (Kanehisa et al., 2012) was performed by DAVID Bioinformatics Resources (Huang da et al., 2009). Top significant pathways were ordered by q-value, which is a corrected p-value by multiple hypothesis testing using the Benjamini-Hochberg, with 0.01 as a cutoff.

Neural crest genes with TPM greater than 1 in three pNC samples are considered to be expressed. Genes associated with RA, Wnt and TGF-beta pathways were analyzed with KEGG pathway, Map Database, and published data from Pubmed. Transcription factors, corneal endothelium, and ECM genes were analyzed with various tools including KEGG Orthology database, KEGG pathway database, InterPro database, and published data from Pubmed.

In situ hybridization

Freshly isolated E3 chick heads and either E5 or E7 eyes were fixed overnight at 4°C in modified Carnoy’s fixative (60% ethanol, 30% formaldehyde, and 10% glacial acetic acid). Samples were embedded in paraffin and sectioned between 10–12 μm. Probes were designed by Primer-blast (https://www.ncbi.nlm.nih.gov/tools/primer-blast/) or adapted from Gallus Expression in Situ Hybridization Analysis (Geisha; http://geisha.arizona.edu/geisha/) and cloned into pCR®II-TOPO® vector or pCR®4-TOPO® with dual promoters (Invitrogen). A list of primer sequences used for generating in situ probes is provided in Table 4. Digoxigenin-labeled riboprobes were generated by in vitro transcription with Superscript III®. Section in situ hybridization was performed as previously described (Etchevers et al., 2001). Brightfield images were captured using a Zeiss Axiocam mounted on AxioImager2 microscope (Zeiss).

Table 4.

List of primers used for in situ hybridization.

Gene ID	Symbol	Forward primer	Reverse primer
NCC genes
395245	MSX21	5’-CGAGGAGCACCACAAAGTCAAG-3’	5’-GACAGGAGTAGCATAGAGTCCAACG-3’
432368	SNAI2	5’-CTGTGTGGACTACAACCGGG-3’	5’-CTTCACATCCGAGTGGGTCT-3’
395734	RHOB	5’-TCGTCTTCAGCAAGGACGAG-3’	5’-GCAATTGATGCAGCCGTTCT-3’
RA genes
416880	ALDH2	5’-ATCCCTTGGAACTTCCCCCT-3’	5’-ACTGATGCCCAACAGCAACT-3’
419480	DHRS3	5’-CCTTCCCTTCCTTCGCTTTAT-3’	5’-GTGCTTCCAAACTCCACATTTC-3’
386585	NR2F2	5’-CAAAGTTGGCATGAGACGGG-3’	5’-AGCTTCCCGAATCGTGTTGG-3’
TGF-β
421461	LTBP12	5’-TGCATCAAACCTAACTGTGCA-3’	5’-TCGGAAGTTAGTGGCTGTCA-3’
428413	BAMBI3	5’-GATCGCCATTCCAGCTACAT-3’	5’-TTTGCTGTCGTTGATCTTGC-3’
395897	TGFBI	5’-CACACAGCTCTACTCCGACC-3’	5’-GGCCAACTCAAACAGGGTCT-3’
Wnt genes
100858542	AES	5’-TGTTTCCACAAAGCCGACAC-3’	5’-TTCTCCCCATCGTCGTCTTG-3’
424707	WLS	5’-AGTGATCGCCTTTCTGGTGG-3’	5’-GCTATGGGTCCAACCTGCTT-3’
395862	PITX2	5’-AGCGGACGCACTTCACCAGC-3’	5’-CGCAGCTCAGTCCGTGGCAA-3’

¹Adapted from Geisha (http://geisha.arizona.edu/geisha/search.jsp?entrez_gene=461423)

²Lorda-Diez et al., 2010.

³Adapted from Geisha (http://geisha.arizona.edu/geisha/search.jsp?entrez_gene=461645)

Immunostaining

Paraffin sections of E3, E5, and E7 quail-chick chimeras were rehydrated and prepared for immunostaining following standard protocols. Mouse anti-QCPN monoclonal antibody (1:1, IgG1, DHSB) was used to identify quail neural crest-derived cells during corneal development. Secondary antibody (Alexa 594 Goat anti-mouse IgG1, Invitrogen) was used at a concentration of 1:200. Imaging was performed with AxioImager2 fluorescent microscope with ApoTome (Zeiss).

Key findings:

• RNA-Seq profiling revealed enrichment of genes between periocular neural crest corneal endothelium, and keratocytes.

• The top KEGG pathways of the differentially express genes include focal adhesion, ECM-receptor interaction, cell adhesion, melanogenesis, and MAPK signaling.

• Candidate neural crest genes are expressed in the periocular neural crest, but they are either differentially expressed or maintained during cornea development.

• Modulators of major signaling pathways associated with ocular development (retinoic acid, TGFβ, and Wnt), transcription factors, and differentiation genes were differentially expressed.