Nephronectin-integrin α8 signaling is required for proper migration of periocular neural crest cells during chick corneal development

During development, cells aggregate at tissue boundaries to form normal tissue architecture of organs. However, how cells are segregated into tissue precursors remains largely unknown. Cornea development is a perfect example of this process whereby neural crest cells aggregate in the periocular region prior to their migration and differentiation into corneal cells. Our recent RNA-seq analysis identified upregulation of nephronectin (Npnt) transcripts during early stages of corneal development where its function has not been investigated. We found that Npnt mRNA and protein are expressed by various ocular tissues, including the migratory periocular neural crest (pNC), which also express the integrin alpha 8 (Itgα8) receptor. Knockdown of either Npnt or Itgα8 attenuated cornea development, whereas overexpression of Npnt resulted in cornea thickening. Moreover, overexpression of Npnt variants lacking RGD-binding sites did not affect corneal thickness. Neither the knockdown nor augmentation of Npnt caused significant changes in cell proliferation, suggesting that Npnt directs pNC migration into the cornea. In vitro analyses showed that Npnt promotes pNC migration from explanted periocular mesenchyme, which requires Itgα8, focal adhesion kinase, and Rho kinase. Combined, these data suggest that Npnt augments cell migration into the presumptive cornea extracellular matrix by functioning as a substrate for Itgα8-positive pNC cells.


INTRODUCTION
In vertebrates, the cornea is comprised of three cellular layers: the epithelium, stroma and endothelium. The corneal epithelium is derived from the ectoderm, whereas the stromal keratocytes and corneal endothelium are derived from cranial neural crest cells (Lwigale et al., 2005). In birds and humans, neural crest migration from the periocular region into the presumptive cornea occurs in two waves (Feneck et al., 2020;E. D. Hay, 1980). The first wave forms a monolayer of the nascent corneal endothelium, which is accompanied by a second wave of mesenchymal migration into the acellular matrix between endothelium and ectoderm to form the cornea stroma (Hay and Revel, 1969;Noden, 1978;Lwigale et al., 2005;Creuzet et al., 2005;Feneck et al., 2020). Several mechanisms pertaining to the role of growth factors and guidance cues have been identified (Beebe and Coats, 2000;Saika et al., 2001;Lwigale and Bronner-Fraser, 2009;Choi et al., 2014), but the functional role of ECM/integrin signaling in establishing the cornea remains unclear.
Among the initial effects of the signaling that takes place between the nascent ocular tissues is the synthesis of the corneal extracellular matrix (ECM) by the corneal epithelium in response to inductive signals from the lens vesicle to form the primary stroma (Hay and Revel, 1969;Hendrix et al., 1982;Fitch et al., 1988). In birds and humans, the primary stroma serves an important role as a scaffold for pNC migration throughout the process of corneal development (Hay and Revel, 1969; Bard and Hay, 1975;Quantock and Young, 2008). It consists of multiple ECM proteins including hyaluronan (Toole and Trelstad, 1971), collagen type I and II (Hayashi et al., 1988;Fitch et al., 1988), laminin (Doane et al., 1996), and fibronectin (Fn) (Fitch et al., 1991). Following the second wave of pNC migration into the developing cornea, the primary stroma is concomitantly replaced by the secondary stroma that is synthesized by differentiating stromal keratocytes. The secondary stroma comprises the bulk of the adult cornea and consists of collagens and proteoglycans (Hayashi et al., 1988;Quantock and Young, 2008) that are arranged in patterns that result in transparency (Chen et al., 2015). Over the past decades, the role of collagens and proteoglycans have been the focus of several investigations due to their indispensable functions in corneal transparency. However, most of these studies were conducted in rodents, in which early corneal development does not involve the primary stroma (Pei and Rhodin, 1970;Haustein, 1983;Feneck et al., 2019). Although the ECM plays a critical role during organogenesis by providing cell adhesion substrate, sequestering signaling molecules, providing structural support and mechanical cues (Hynes, 2014), the function of the primary stroma during early corneal development remains to be elucidated.
Our recent RNASeq analysis of pNC differentiation into corneal cells identified novel expression and upregulation of Nephronectin (Npnt) transcripts during corneal development. Npnt is a 70-90 kDa protein that was discovered as a novel ECM ligand for integrin a8b1 (Itga8b1) during kidney development (Brandenberger et al., 2001), and at the same time, as a protein secreted by preosteoblast cells (Morimura et al., 2001). It consists of five epidermal growth factor (EGF)-like domains in the N-terminal, a central region containing an Arg-Gly-Asp (RGD) sequence, and a meprin-A5 protein-receptor protein tyrosine phosphatase µ (MAM) domain in the C-terminal. The EGF-like and RGD domains have been functionally characterized and shown to play critical roles in development and tissue homeostasis. The RGD domain signals through a8b1 receptor during epithelial-mesenchymal interactions involved in kidney development and maintenance (Miner, 2001, Linton et al., 2007Müller et al., 1997, Sato 2009, Cheng et al., 2008, Inagi et al., 2017, Müller-Deile et al., 2017, Zimmerman et al., 2018. These observations in mice were recently confirmed by the identification of recessive mutation in the Npnt gene that caused bilateral kidney agenesis in human fetuses (Dai et al., 2021). The EGF-like domains of Npnt are associated with inducing differentiation and proliferation in osteoblasts and dental stem cells through the activation of mitogen-activated protein kinase (MAPK) pathways (Fang et al., 2010;Kahai et al., 2010;Arai et al., 2017), and to induce vascular endothelial cell migration via phosphorylation of extracellular signal-regulated kinases (ERK) and p38MAPK (Kuek et al., 2016). Recent studies have also implicated Npnt to play a role in heart development (Patra et al., 2011), attachment of the arrector pili muscle to hair follicles (Fujiwara et al., 2011), and during forelimb formation in amphibians (Abu-Daya et al., 2011). Other studies have also demonstrated that Npnt plays a potential role in diseases such as chronic obstructive pulmonary disease (COPD) (Saferali et al., 2020), diabetic glomerulosclerosis (Nakatani et al., 2012), Fraser syndrome (Kiyozumi et al., 2012), and in various cancers (Magnussen et al., 2021). Despite the pleiotropic functions of Npnt in development and disease, its expression and function have not yet been described in the cornea.
Inspired by our observation that Npnt transcripts were upregulated during corneal development, we sought to establish the role of Npnt during avian corneal development. We first characterized the spatiotemporal expression of Npnt mRNA and protein at stages that correspond with pNC migration. Next, we used RCAS-mediated gene knockdown to investigate the role of Npnt. We identified Itga8 as a potential receptor for Npnt during cornea development and confirmed its role in pNC using in vitro migration assay coupled with peptide inhibition, and in vivo using gene knockdown. Lastly, we performed misexpression studies using the either the full-length, RGD mutant, and truncated versions of Npnt to further confirm its role and the signaling mechanism during corneal development. Together, our findings show that Npnt secreted into the ECM of the nascent cornea provides a substrate that promotes migration of Itga8-positive cells, thus segregating the cornea progenitors from other pNC cells.

Npnt mRNA and Protein are expressed during early ocular development
Our previous RNA-seq analysis of gene expression during pNC differentiation into corneal cells identified upregulation of Npnt transcripts during chick corneal development (Bi and Lwigale, 2019). To define the expression of Npnt mRNA and protein distribution during this process, we performed section in situ hybridization and immunohistology, respectively. Consistent with our RNA-seq data (Bi & Lwigale, 2019), we found that Npnt is undetectable in the pNC by embryonic day (E) 4 and in the presumptive corneal endothelium at E5, but it is primarily expressed in the retinal pigment epithelium (RPE) and presumptive lens fiber cells at these time points (Fig. 1A, B). However, by E6, we found that in addition to the RPE and lens, Npnt is vividly expressed by the second wave of migratory pNC that eventually differentiate into the stromal keratocytes ( Fig.   1C, arrowhead). Expression of Npnt persisted in the corneal stroma through E12 (Fig. S1A-C), but it was undetectable at E15 (data not shown).
Because Npnt is a secreted protein that plays a role in cell migration (Kuek et al., 2016;Magnussen et al., 2021;Yamada & Kamijo, 2016), we examined its localization at similar time points during cornea development. At E4, Npnt is localized in the optic cup, lens epithelium, surrounding the pNC cells adjacent to the nascent cornea ( Fig. 1D-F, arrow), and in the primary stroma of the cornea ( Fig. 1D-F, asterisk). At E5, strong Npnt staining is observed in the primary stroma ( Fig.   1G-I, asterisk). By E6, Npnt is localized to the basement membrane of the corneal epithelium and diffusely throughout the primary stroma ( Fig. 1J-L). At this time, the migratory pNC also stain positive for Npnt (Fig 1L, arrowheads). Strong expression of Npnt persists in the basement membrane of the corneal epithelium through E9 (Fig. S1 C,D), and it is faintly detectable in the corneal epithelial cells by E12 (Fig. S1 E). These data show that expression of Npnt coincides with pNC ingression into the cornea, implicating a potential role during development.

Knockdown of Npnt disrupts corneal development
To investigate the function of Npnt during early corneal development, we generated several RCAS-GFP-Npnt-shRNA (Npnt kd ) constructs in chick DF-1 cells ( Supplementary Fig. S2). The construct with the highest knockdown of Npnt was used for the experiments. Constructs containing only GFP (RCAS-GFP) or scrambled shRNA (RCAS-GFP-Scr-shRNA) were used as control. The viral constructs were injected to cover the entire cranial region of HH7-8 (Hamburger & Hamilton, 1951) chick embryos ( Fig. 2A). Using this approach, both the neural progenitors of the optic cup and neural crest cells, as well as the ectodermal progenitors of the lens, where Npnt is expressed showed significant reduction in corneal thickness in Npnt kd corneas (Fig. 2F). These data suggest that Npnt plays a critical role during normal formation of the cornea.

Itga8 is the receptor for Npnt during cornea development and plays a role in pNC migration
Next, we reasoned that because knockdown of Npnt causes a significant reduction in corneal thickness, one or more of its receptors should be expressed by the migratory pNC during cornea development. In this study, we focused on integrin a8b1 because of its strong affinity for Npnt (Brandenberger et al., 2001). Although integrin a8b1 was observed to promote spreading of trunk neural crest in vitro (Testaz et al., 1999), little is known about its expression and function in the cranial neural crest. Our analysis by in situ hybridization revealed that Itga8 was expressed in the leading edge of the periocular mesenchyme prior to pNC migration into the primary stroma of the nascent cornea (Fig. 3A, arrow). Subsequently, Itga8 was maintained in the periocular mesenchyme but also expressed in the corneal endothelium at E5 (Fig. 3B, arrows) and stroma at E6 (Fig. 3C).
Given the striking expression of Itga8 by the migratory pNC, we first assessed the potential for Itga8-Npnt signaling in vitro using explanted periocular mesenchyme from the leading edge adjacent to the presumptive cornea ( Fig. 3A and 3D). Mesenchyme explants were cultured on slides coated with Npnt in the presence or absence of a peptide inhibitor previously shown to specifically inhibit binding of a8b1 with Npnt ( Fig. 3D) (Sato et al., 2009). In the absence of the inhibitor, explants attached to the Npnt-coated slides and numerous migratory cells formed a halo around the explant within 12-17 hours of incubation ( Fig. 3E; Movie 1). In contrast, we observed that treatment with the Itga8 inhibitor significantly decreased the density of migratory cells from the explant (Fig. 3F, 3G; Movie 2). These results indicate that a8b1 functions as a receptor for Npnt signaling in the presumptive corneal pNC, and that it is required for their migration.

Knockdown of Itga8 disrupts corneal development
Given that Itga8-expressing pNC appear to be in direct contact with Npnt secreted into the primary stroma, we wanted to characterize the role of Itga8 during corneal development. For this analysis, we generated and tested several RCAS-GFP-Itga8-shRNA constructs and validated one that showed highest knockdown of Itga8 in DF-1 cells and in vivo (Supplementary Fig. S2 and Fig.   S3). As a first step, we analyzed the effect of Itga8 knockdown on the first wave of pNC migration that forms the corneal endothelium at E5. We found that at this timepoint relatively fewer GFP expressing cells appeared to migrate into the Itga8 kd cornea (Fig. 4B, arrows) compared to RCAS-GFP control, which showed robust occupation of GFP-positive cells in the corneal endothelium Furthermore, we performed a BrdU assay and observed no significant differences in cell proliferation between Itga8 kd (Fig. 4G) and control corneas (Fig. 4F). These results indicate that Itga8 is required for pNC migration into the cornea. Given that the corneal defects following knockdown of Itga8 correlate with Npnt knockdown, our results suggest that Npnt-Itga8 signaling plays an important role during pNC migration into the cornea.

Overexpression of Npnt causes corneal thickening
Given that knockdown of both Npnt and Itga8 resulted in corneal thinning, combined with our observation that Npnt is expressed during the second wave of migration at E6, we investigated whether overexpression of the Npnt affects cornea development. We generated and tested viral constructs containing the full-length Npnt gene (Npnt oe ) in DF1 cells (Fig. S2). Control and Npnt oe constructs were injected in HH7-8 embryos as described (Fig. 2), and the corneas were collected for analysis at E7, E9, and E15. First, we confirmed the overexpression of Npnt mRNA and protein in the corneas. At E9, expression of Npnt was localized in the corneal stroma during normal development (Fig. 5A). However, Npnt oe corneas showed robust expression of Npnt in the stroma, as well as ectopic expression in the corneal epithelium and endothelium, and the lens (Fig. 5D).
At this stage, strong staining for Npnt protein was only observed in the basement membrane of the corneal epithelium and low diffused staining in the stroma of control corneas (Fig. 5B). In contrast, the Npnt oe corneas showed strong ectopic staining for Npnt in the corneal epithelium and endothelium, and a substantial increase in protein expression in the stroma compared to the control ( Fig. 5E).
Next, we performed histological analysis on E7, E9, and E15 corneas to determine whether there were any morphological differences between the control and Npnt oe corneas. Our analysis did not reveal any significant differences in corneal thickness and cell number at E7 (Fig. 5G, H). By contrast, cross-sections of E9 Npnt oe corneas revealed significant thickening and increase in the cell number (Fig. 5F, G, H) compared with control ( Fig. 5C, G, H). Similar increases in corneal thickness and cell number were observed at E15 (Fig. 5G, H). Given that the keratocyte density is higher in the anterior stroma than the posterior stroma (Patel et al., 2001;Berlau et al., 2002), we quantified the cell densities in these regions. Overall, we observed relatively higher cell densities in the anterior stroma of E7, E9, and E15 corneas and there were no significant differences between Npnt oe and control corneas (Fig. 5I). Posterior cell density was not affected in E7 Npnt oe corneas, but there was significant reduction at E9 and E15.
To determine whether elevated cell proliferation was a contributing factor to the increased corneal thickness in Npnt oe , we examined corneas at E7, E9, and E15 by performing BrdU labeling and immunofluorescent detection. Overall, we observed cell proliferation in all corneal layers, and there was a decreasing trend in the percentage of labeled cells in the stroma as development progressed from E7 to E15 (Fig. 5J, K). However, our results revealed no significant increase in cell proliferation at E7 and E15. Surprisingly, there was a significant reduction in cell proliferation in E9 Npnt oe corneas despite their increase in thickness (Fig. 5K). Combined, our data indicate that overexpression of Npnt increases transcript and protein expression in the cornea, which results in increased cell number in the stroma and corneal thickness. Since these increases were not correlated with increased cell proliferation, our data suggest that the changes in corneal thickness are due to augmented pNC migration into the cornea caused by excessive expression of Npnt.

The RGD domain mediates Npnt signaling in pNC during corneal development
Npnt mediates signal transduction in development and cancer by binding to various receptors through the EGF-like and RGD domains (Arai et al., 2017;Kahai et al., 2010;Kuek et al., 2016;Linton et al., 2007). The marked reductions in corneal thickness observed following knockdown of both Npnt and Itga8 ( Fig.2 and Fig.3) suggest that Npnt functions via the RGD domain during pNC migration. Given that overexpression of the full-length Npnt protein caused corneal thickening, we generated an Npnt-RAE version of the protein in which the RGD domain was mutated by substitution with RAE sequence, and an Npnt-EGF version in which both the RGD and MAM domains were truncated (Fig. 6A). Constructs were injected in HH7-8 embryos and E9 corneas were collected, sectioned, and analyzed following DAPI staining. Contrary to our observations following overexpression of the full-length protein (Fig.5), neither the Npnt-RAE or Npnt-EGF versions caused a significant change in corneal thickness compared with the control (Fig. 6B, C). Interestingly, there was a significant reduction in cell count and density in Npnt-RAE corneas although no difference in these values was observed for Npnt-EGF (Fig. 6D, E). The absence of corneal thickening following overexpression of either Npnt-RAE or Npnt-EGF constructs further suggest that the RGD domain plays an essential role in this process. Our results also indicate that the EGF-like domain does not play an essential role during for pNC migration into the cornea.

DISCUSSION
Very little is known about the function of the corneal ECM during early development. Here, we focus on Npnt, which we recently found in our RNA-seq study (Bi & Lwigale, 2019) to be upregulated during early development of the chick cornea. Specifically, we have identified novel expression of Npnt and its Itga8 receptor at critical timepoints of corneal development. Our data reveal that disruption of Npnt/Itga8 signaling either in ovo or in vitro perturbs pNC migration that subsequently result in corneal thickness defects, but formation of the cellular layers is not affected.
Our model (Fig. 7) provides a context in which Npnt functions in the presence of Fn during corneal development. Fn is a well-known substrate for neural crest cell migration (Alfandari et al., 2003;Bronner-Fraser, 1986;Newgreen & Thiery, 1980) that is robustly expressed in the periocular mesenchyme and cornea (Doane et al., 1996;Kurkinen et al., 1979). In comparison, Npnt is expressed in an increasing gradient from the edge of the periocular region towards the cornea (Fig.   7A). We posit that all pNC cells express a5b1 and previously showed that they robustly migrate on Fn substrate in vitro (Lwigale & Bronner-Fraser, 2009). However, in vivo, it is most likely that the pNC which express both a5b1 and a8b1, and thus can respond to Npnt and Fn, migrate into the cornea. Based on our current study, we conclude that Npnt/Itga8 signaling plays an essential role in pNC migration during corneal development.

Npnt and Itga8 in the cornea
While Npnt is associated with kidney, teeth, and bone development (Arai et al., 2017;Kuek et al., 2016;Linton et al., 2007), only one reference was made to its expression in the mouse lens during ocular development (Brandenberger et al., 2001). First, we confirmed that Npnt is expressed in the migratory pNC and showed that it is also localized in the optic cup and lens vesicle. Our immunohistochemistry analysis revealed that Npnt protein staining consistently corresponded with the mRNA expression during corneal development. In addition, we observed that Npnt was localized in the acellular primary stroma suggesting that prior to its expression by the migratory pNC in the cornea, the optic cup and lens vesicle are the primary sources of Npnt in the nascent corneal ECM. The link between the primary stroma and the presumptive corneal epithelium and endothelium was established in classical studies, which showed that the lens and corneal endothelium induce the corneal epithelium to synthesize collagen and glycosaminoglycans into the underlaying space (Fitch et al., 1994;E. Hay et al., 1979;E. D. Hay & Revel, 1969;Toole & Trelstad, 1971). Our expression analyses suggest that the primary stroma sequesters Npnt initially secreted by the optic cup and the lens, and exposes it to pNC during corneal development.
Itga8 is expressed in the spinal cord, optic nerve, retina, urogenital and digestive systems, and in the head epidermis (Bossy et al., 1991;Ogawa et al., 2018). a8b1 is known to modulate epithelialmesenchymal interactions during kidney and pharyngeal development (Müller et al., 1997;Talbot et al., 2016) and to promote migration of mesangial and smooth muscle cells (Bieritz et al., 2003;Zargham & Thibault, 2006). We report for the first time that Itga8 is expressed by pNC during ocular development. Since Itga8 heterodimerizes with Itgb1 subunit (Bossy et al., 1991), which is robustly expressed in the periocular mesenchyme and migratory pNC (Doane & Birk, 1994), we infer that a8b1 signaling functions during corneal development in response to Npnt secreted in the primary stroma. Following delamination from the neural tube, cranial neural crest migrate in response to Fn and laminin (Duband & Thiery, 1982;Newgreen & Thiery, 1980;Sternberg & Kimber, 1986) via integrin receptors such as a1b1, a4b1, a5b1, aVb1 (Alfandari et al., 2003;Delannet et al., 1994;Desban & Duband, 1997;Kil et al., 1996;Lallier et al., 1994). Although neural crest cells are initially highly migratory, they remain relatively stationary upon their localization in the periocular region. Given that Itga8 expression coincides with the onset of pNC migration, our results indicate its potential role in their ingression into the developing cornea.

Npnt promotes pNC migration via RGD domain and Itga8
The three major factors that contribute to corneal thickness during development are pNC migration, cell proliferation, and synthesis of the secondary stroma by the keratocytes. We showed that disruption of Npnt expression in the anterior ocular tissues via RCAS-mediated knockdown resulted in decreased corneal thickness. In addition to functioning in epithelial-mesenchymal interactions, Npnt has been shown to promote migration in various tissues including vascular endothelial cells during osteogenesis (Kuek et al., 2016), infiltration of immune cells into the liver (Hong et al., 2020;Inagaki et al., 2013), and cancer metastasis (Magnussen et al., 2020;Mei et al., 2020;Wang et al., 2018). Given that Npnt is localized in the primary stroma during corneal development, we hypothesized that it may play a potential role in pNC migration. In agreement with our hypothesis, our in vitro migration assays confirmed that robust pNC migration from mesenchyme explant occurred on Npnt substrate but was abrogated in the presence of the Itga8 inhibitor. We also observed that knockdown of Itga8 reduced pNC migration into the cornea and phenocopied the reduction of corneal thickness observed following Npnt knockdown. In addition, we found that reduction in corneal thickness was accompanied by decrease in cell count, but cell density and proliferation were not affected, suggesting that Npnt/Itga8 signaling mediates pNC migration during corneal development. Given that the primary function of keratocytes is to synthesize the corneal ECM comprised of collagens and proteoglycans, and represents approximately 90% of the corneal thickness (Fini, 1999;Funderburgh et al., 2003;Kao, 2010), our observation that there was no change in stromal cell density raises a possibility that matrix synthesis by the pNC that differentiated into keratocytes is not a major contributing factor to the reduction in corneal thickness.
Furthermore, our overexpression studies showed the opposite effect whereby the full-length construct caused increased corneal thickness at E9. We also found that increased corneal thickness persisted at E15 when Npnt appears to be downregulated in the cornea. Surprisingly, the E7 corneas were not affected in the overexpression studies. This could be caused by limited pNC migration potential due to their expression of Neuropilin1, which prevents them from entering the corneal environment that contains a repulsive Semaphorin3A signal (Lwigale & Bronner-Fraser, 2009). As observed in the knockdown studies, there was no differences in cell proliferation between control and Npnt oe corneas, suggesting that there was an increase of pNC migration. The mutant Npnt-RAE construct overexpresses approximately the same size protein as the endogenous Npnt, while maintaining its MAM domain to support ECM-ECM interactions, but it did not impact cornea thickness. This indicates that expression of high levels of protein does not affect cornea thickness, further implicating that Npnt functions in directing cell migration specifically through its RGD sequence during early cornea development. The reduction in cell count and density could be attributed to the mutant RAE domain outcompeting the expression of the endogenous RGD by the pNC. Similarly, the truncated version of Npnt containing only the EGF-like repeats did not affect cornea thickness, further conforming that Npnt signals through the RGD domain and Itga8 to modulate pNC migration into the cornea.
Precise coordination of multiple signals from surrounding ocular tissues orchestrate the spatiotemporal migration and differentiation of multipotent pNC during the formation of the avian corneal endothelium and stromal keratocytes. Disruptions in the sequence of these early events can lead to significant defects in corneal development. Our study provides the first characterization of Npnt/Itga8 function in pNC that contribute to the cornea. Despite being surrounded by an ECM rich in Fn, pNC remain relatively immobile in the periocular region until they are competent to respond to Npnt cues via the expression of a8b1. This transformation may be crucial for the timely induction of directional migration of pNC towards the gradient of additional cues generated by Npnt in the primary stroma (Fig.7A, B), and therefore plays a vital role in segregating the cornea progenitors from the rest of the periocular mesenchyme. Our results also show that Npnt subsequently localizes to the basement membrane of the corneal epithelium. Future studies will investigate the potential role of Npnt in epithelial-mesenchymal interactions between the corneal epithelium and stroma, potentially as a result of signaling through the EGF-like domain to EGFR receptors in the epithelial cells.

Chick embryos
All experiments were performed using fertilized White Leghorn chicken eggs (Gallus gallus domesticus) obtained from Texas A&M Poultry Center (College Station, TX). Eggs were incubated at 38 ˚C in humidified conditions until the desired stages. Animal studies were approved by the Institutional Animal Care and Use Committee (IACUC) at Rice University.
For Npnt and Itga8 knockdown and overexpression studies, eggs were incubated for approximately 24-26hrs to obtain 3-somite stage or HH8 (Hamburger & Hamilton, 1951), then windowed as previously described (Spurlin & Lwigale, 2013). A few drops of Ringer's solution containing 100 U/mL penicillin and 100µg/mL streptomycin (PenStrep, ThermoFisher Scientific) were added to embryos to maintain hydration. Embryos were injected with viral constructs (see below) using a Picospritzer III pneumatic microinjection system (Parker Hannifin) in the space between the vitelline membrane and cranial region, ensuring that the neural tube and adjacent ectoderm were completely covered ( Fig. 2A). Injected embryos were re-incubated and collected for GFP screening at the desired stages of development ranging between embryonic day (E)5-E15. Only the corneas showing robust GFP expression were used for subsequent analyses.

Histology, Hematoxylin and Eosin staining, and Immunohistochemistry
Embryos were collected at desired stages and eyes were dissected in Ringer's saline solution.
Afterwards, the tissues were dehydrated in ethanol series, cleared in Histosol (National Diagnostics), embedded in paraffin blocks, and sectioned at 10 µm thickness. Sections were stained with Hematoxylin for 15 seconds, counterstained with Eosin for 40 seconds, dehydrated in ethanol series, and mounted with Cytoseal (ThermoFisher Scientific) for imaging. For Npnt immunohistochemistry, eyes were fixed in methanol-acetic acid (MAA) fixative (98% methanol, 2% glacial acetic acid) that was chilled on dry ice. Samples were maintained in MAA fixative at -80 ˚C for at least two days, then were gradually warmed to room temperature before further dehydration in ethanol series and embedding in paraffin blocks. Samples were sectioned at 10 µm, rehydrated, then immunostained with Npnt antibody (1:100; orb221700, Biorbyt) following standard procedures. For all the other immunohistochemistry and phalloidin staining, samples were fixed in 4% PFA overnight at 4 ˚C, embedded in paraffin, and sectioned as described above.

Section In situ Hybridization
Eyes were fixed in modified Carnoy's fixative, embedded in paraffin, and sectioned as described above. Riboprobes were generated from gene fragments using cDNA pooled from E7 anterior eyes and cloned into TOPO-PCRII (Invitrogen). Digoxigenin (DIG)-labeled riboprobes were synthesized following manufacturer's protocol (DIG Labeling Kit, Roche). Primers used to generate the riboprobes are listed in Supplementary Table 1. Sections were hybridized with riboprobe at 52 ˚C (Npnt) and 60 ˚C (Itga8) overnight. Hybridization was detected using anti-DIG antibody conjugated with alkaline phosphatase (Roche) and color was developed with 5-bromo-4chloro-3-indolyl phosphate/nitro blue tetrazolium (BCIP/NBT; Sigma). Following color development, sections were fixed with 4% PFA, mounted in Cytoseal and imaged using an Axiocam mounted on an AxioImager2 microscope (Carl Zeiss).

Production of RCAS virus
Replication-Competent ASLV long terminal repeat with a Splice acceptor (RCAS) (Hughes et al., 1987) virus was used for stable and prolonged expression of short hairpin (sh)RNA constructs used for knockdown and for the overexpression studies. To produce RCAS viral stocks, chick fibroblasts (DF-1 cells; Lot 58217603, ATCC) were cultured in Dulbecco's modified eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS, Invitrogen), 100 U/mL penicillin and 100 µg/mL streptomycin (referred to here as complete DMEM, ThermoFisher Scientific). Cells were transfected at about 70% confluency with plasmid DNA containing the RCAS constructs using Lipofectamine 3000 (Invitrogen). Transfected cells were grown for four days to enable viral replication. Media containing viral particles were collected on subsequent days, pooled, and centrifuged at 21,000 rpm (Beckman) for 1.5 hour at 4 ˚C to concentrate the virus. Virus pellet was resuspended in DMEM and stocks at approximately 1-7x10 6 Ifu/ml were stored in -80 ˚C until use.

Generation of shRNA and viral constructs
shRNA target sequences used for Npnt kd and Itga8 kd (Supplementary Table 2), were designed using the BLOCK-iT RNAi Designer tool (ThermoFisher Scientific) and linked to their reverse complementary sequence by a loop sequence, TTCAAGAGA. A short termination sequence and restriction enzyme sites were added at each end to create fragments that were ligated into a pSLAX shuttle vector (Hughes et al., 1987) and to add a chick U6 (Cu6) promoter and a GFP reporter. The shRNA sequences were cloned into the RCAS vector (Supplementary Fig. 2A,B) by homologous recombination using a CloneEZ kit (GenScript) as previously described (Kwiatkowski et al., 2017;Ojeda et al., 2017). The Npnt oe overexpression constructs were either generated by substitution of the shRNA with a full-length coding sequence of Npnt in the vector above or using a modified vector where the full-length Npnt protein was driven by a promoter within the viral long terminal repeats (LTRs) with eGFP linked by an IRES motif (Supplementary Fig. 2A, B). Constructs containing either mutated Npnt with the RGD sequence changed to RAE (PRGDVFIPRQPGVSNNLFEILEIER to PRAEVFIPRQPGVSNNLFEILAIAR) or the truncated version of Npnt containing only the N-terminal EGF-like repeats domain (Npnt-EGF) were commercially synthesized (Genscript) and cloned into the modified RCAS vector. All constructs were validated using primers for Npnt and Itga8 (Supplementary Table 3).

BrdU staining
Bromodeoxyuridine (BrdU) was used to assay for cell proliferation. The BrdU solution was prepared at a final concentration of 10 µM in complete media. Eyes were collected at desired stages and injected into the anterior chamber between the cornea and lens with BrdU solution (approximately 50-100 µl depending on development stage), then cultured in BrdU solution for 2 hours at 37 ˚C. Eyes were rinsed in phosphate buffered solution (PBS), fixed in 4% PFA, embedded in paraffin, and sectioned as described above. BrdU-positive cells were identified by immunohistochemistry using anti-BrdU antibody (1:30; G3G4, Developmental Studies Hybridoma Bank). Sections were counterstained with DAPI.

In vitro explant culture
Embryos were collected at E4 in Ringer's solution and the anterior eyes were dissected and digested in dispase (1.5mg/mL, Worthington) for 10 minutes at 37 ˚C. The periocular mesenchyme was isolated by physically removing the presumptive cornea epithelium, lens, and the optic cup.
The mesenchyme ring was further trimmed using tungsten needles to obtain cells that are proximal to the presumptive cornea, then dissected into approximately 120x120 µm explants. Nunc Lab-tek II 8-well chamber slides (Sigma) were coated with 1.5 µg/cm 2 with poly-D lysine (MP Biomedicals) for 1 hour at room temperature, followed by recombinant mouse or human Npnt (R&D Systems) at 1.5 µg/cm 2 for 2 hours at 37 °C. The α8β1 peptide inhibitor following the 23mer sequence PRGDVFIPRQPTNDLFEIFEIER (Sato et al., 2009) was commercially generated (Genscript) and used at a 10 µM working concentration. Mesenchyme explants were transferred to the coated slides containing complete media, with and without α8β1 inhibitor and incubated at 37 °C in a humidified tissue culture incubator with 5% CO2 for 12 hours. Explants with migratory cells were imaged with a Rebel T6s camera (Canon) mounted on an Axiovert 40C microscope (Carl Ziess). The magnitude of cell migration from explants was calculated by measuring the density of cells in a 175 µm x 175 µm area located 270 µm from the center of the explant.

Time-lapse video microscopy
Mesenchyme explants were prepared and cultured as described above in media containing 0.1 µg/ml Hoechst dye solution (ThermoFisher Scientific) used to label all nuclei. Chamber slides containing attached explants were imaged at 3 minutes and 27 second intervals for 17 hours using an FV1200 laser scanning microscope (Olympus) with a stage top incubator (Tokai Hit). Movies were generated using IMARIS Software (Oxford Instruments).

Corneal thickness
Differences in corneal thickness between control and knockdown samples was determined by averaging measurements taken at three separate locations along the cornea. All measurements were taken perpendicular to the radial curvature of the cornea. Since corneal thickness is not always uniform in overexpression samples, measurements were taken at three locations across the center of the thickened region.

Cornea cell counts and density
Nuclei labeled by the DAPI staining of corneal sections were used to count stromal cells within a width of 200 µm along the entire height of that region. Cells were either counted manually or using the threshold particle analysis in ImageJ Software in which nuclei from the epithelium and endothelium are discounted from the final tally. Cell density was determined by dividing the number of nuclei and the area, which was measured by multiplying the value of the corneal thickness by the width of the selected region. Comparison of cell counts and density between the anterior vs posterior corneal region was determined by taking similar areas for all corneas, which was set at a value of 20% the average thickness of control corneas.

Cell proliferation
Cell proliferation was determined in corneal sections by first utilizing the cell count technique described above. The total number of cells was represented by all the DAPI-positive nuclei, out of which the BrdU-positive cells was quantified. Cell proliferation was calculated as a percentage of total the DAPI-positive nuclei that were BrdU positive.

Software and Statistics
ImageJ Software was used to measure the staining intensity or tissue morphological features, such as cornea thickness and cell density. The number of cornea sections analyzed are summarized in each figure legend. All statistical analyses were conducted using GraphPad Software. Data are presented as scatterplot with mean values. Statistical significance was determined by two-tailed unpaired Student's t-test was used to compare differences between means. Samples with p-values < 0.05 were considered significant. (A-B) Npnt expression in the retina pigment epithelium layer of the optic cup and region of the presumptive lens fiber cells at E4 and E5. (C) Initial expression of Npnt by pNC is observed during the second wave of migration into the stroma (black arrows). (D-F) At E4, Npnt protein was detected in the optic cup, lens epithelium, in the periocular mesenchyme proximal to the presumptive cornea region (arrow), and in the matrix of the primary stroma (asterisk). (G-I) At E5, vivid expression of Npnt protein is localized in the primary stroma adjacent to the corneal epithelium and diffusely throughout the primary stroma (asterisk), and persists at low levels in the optic cup and lens epithelium. At E6, vivid expression of Npnt protein persists in the primary stroma adjacent to the corneal epithelium and it remains diffusely expressed throughout the primary stroma (asterisk). At this time, low expression of Npnt protein is also observed in the migratory periocular neural crest cells invading the primary stroma (arrows in L). Abbreviations: ec, ectoderm; oc, optic cup; pom, periocular mesenchyme; rpe, retinal pigment epithelium; st, stroma; en, corneal endothelium; ep, corneal epithelium. Scale bars: 100 µm.    and protein (B,E) expression, and H&E staining indicating corneal thickness (double sided arrows, C,F). (G,H) Measurements for corneal thickness and cell counts were taken from: E7, N=10 control, N=10 Npnt oe ; E9; N=7 control, N=12 Npnt oe ; E15; N=6 control, N=9 Npnt oe . Bar graphs show no difference at E7, but a significant increase at E9 and E15 in corneal thickness (G), and corneal cells (H). (I) Cell densities were determined from: E7, N=8 control, N=7 Npnt oe ; E9; N=6 control, N=11 Npnt oe ; E15; N=7 control, N=6 Npnt oe . Bar graph shows that there were no differences at E7 and anterior cell densities at E9 and E15, but the posterior cell densities were significantly decreased. (J and K) BrdU analysis and quantification of cell proliferation in corneal sections taken from E7, N=8 control, N=7 Npnt oe ; E9; N=9 control, N=9 Npnt oe ; E15; N=7 control, N=6 Npnt oe . No significant differences were observed at E7 and E15, but there was a significant reduction at E9. ns, not significant; * indicates p < 0.05; ** indicates p < 0.01; and *** indicates p < 0.001. Abbreviations: ep, corneal epithelium; st, stroma; en, corneal endothelium. Scale bars: 100 µm.  Cross-section of E5 eye immunostained for Npnt and Fn. Npnt appears in an increasing gradient from the periocular region into the cornea, whereas Fn stains both the periocular mesenchyme and cornea. (B) All pNC respond to Fn via expression of a5b1, but a subpopulation of pNC that reside in the region adjacent to the presumptive cornea express both a5b1 and a8b1, and become competent to also read the additional gradient of Npnt in the ECM, thus migrating into the corneal region. Abbreviations: pNC, periocular neural crest; oc, optic cup; ps, primary stroma; en, corneal endothelium.