Branching morphogenesis depends on the precise temporal and spatial control of epithelial dynamics. In the vertebrate head, endodermal branches, called pharyngeal pouches, form through the transient stratification, collective migration and reorganization of epithelial cells into bilayers. Here, we report novel requirements for the EphrinB ligands B2a and B3b, the Ephb4a receptor and the Pak2a kinase in the development of pouches and the posterior facial skeleton that depends on pouches for its segmentation. Time-lapse imaging in zebrafish shows that EphB-Pak2a signaling is required to stabilize pouch epithelial cells at the end of branching morphogenesis. Transgenic rescue experiments further demonstrate that endodermal Eph-ephrin signaling promotes pouch integrity by targeting Pak2a to the plasma membrane, where subsequent activation by Wnt4a-Cdc42 signaling increases junctional E-cadherin in maturing pouches. Integration of Eph-ephrin and Wnt4a signaling through Pak2a thus signals the end of branching morphogenesis by increasing intercellular adhesion that blocks further epithelial rearrangements.
During development, epithelial tissues undergo morphogenesis to form elaborately branched organs. Whereas epithelial cells are characterized by apicobasal polarity and strong E-cadherin-mediated intercellular adhesion, dynamic changes in polarity and adhesion facilitate the cell rearrangements that drive epithelial morphogenesis (Wang et al., 2012,, 2013). We still know little, however, about what restores polarity and strong adhesion, and hence stops epithelial rearrangements, once branches have formed.
Pharyngeal pouches arise as outpocketings of the foregut endoderm and help to organize later development of the head and neck (Graham, 2008). During their formation in zebrafish, pre-pouch endodermal cells first lose their columnar morphology to become multilayered, which coincides with their collective movement towards the ectoderm (Choe et al., 2013). At the end of outpocketing, cells re-establish columnar morphology and form mature bilayers. These epithelial transitions are guided first by mesodermal Wnt11r, which stimulates the multilayering process, and then by ectodermal Wnt4a, which reorganizes maturing epithelial cells into bilayers. Here, we show that Wnt4a acts together with EphrinB signaling to stabilize mature bilayers.
Eph-ephrin signaling has been implicated in a variety of morphogenetic processes (Dravis et al., 2004; Holmberg et al., 2000; Klein, 2012), with EphB2 and EphB3 being required to stabilize E-cadherin during lumenization of the mouse pancreas (Villasenor et al., 2010). A candidate for mediating these effects of Eph signaling in epithelial morphogenesis is the family of p21-activated kinases (PAKs), as PAKs have been shown to stabilize E-cadherin during zebrafish epiboly (Tay et al., 2010). Whereas these serine/threonine kinases are defined by their activation by small GTPases, such as Cdc42 and Rac (Knaus et al., 1995; Manser et al., 1994; Martin et al., 1995), PAKs are also targets of EphB signaling during nervous system development (Cowan and Henkemeyer, 2001; Kayser et al., 2008). Here, we show that EphB receptors similarly signal through Pak2a in pouch epithelia, with Pak2a terminating pouch outgrowth by stabilizing E-cadherin and inhibiting further cell rearrangements.
RESULTS AND DISCUSSION
efnb2a, efnb3b, ephb4a and pak2a are expressed in developing pouches
In an in situ hybridization screen of zebrafish Eph-ephrin genes, we uncovered expression of one Eph receptor (ephb4a) and two EphrinB ligands (efnb2a and efnb3b) in her5:GFP+ pouch-forming endoderm (Fig. 1A-J). At the onset of pouch formation [18 h post-fertilization (hpf)], efnb2a is expressed in migrating neural crest cells of the second arch, whereas endodermal expression of efnb2a and efnb3b is low. At this time, ephb4a is expressed in the posterior pouch-forming endoderm. As arch development proceeds (24 and 30 hpf), endodermal expression of efnb2a increases in anterior pouches, whereas efnb3b and ephb4a become enriched in more posterior pouches. By 30 hpf, co-labeling with neural crest-specific sox10:GFP and mesoderm-specific nkx2.5:GFP transgenes show that, in addition to endoderm, efnb2a and ephb4a are also expressed in neural-crest-derived cells and arch mesoderm, respectively (supplementary material Fig. S1). Similar to efnb2a transcripts, Efnb2a protein is observed in both maturing pouches and neural-crest-derived mesenchyme, where it localizes to cell membranes (Fig. 1N,O). In addition, we detect dynamic expression of pak2a in developing pouches (Fig. 1K-M). Thus, efnb2a, efnb3b, ephb4a and pak2a are all expressed in the pouch-forming endoderm, although the restriction of ephb4a to posterior pouches suggests the involvement of additional Eph receptors in the anterior pouches.
Requirements for Eph-ephrin and Pak2a signaling in pouch development
Next, we used loss-of-function approaches to determine requirements for Eph-ephrin and Pak2a signaling in pouch development. In efnb2a mutants, the normal complement of five pouches formed by 34 hpf, yet pouches were short and disorganized (Fig. 2B). This hu3393 allele of efnb2a is probably a complete null allele, as we detected no Efnb2a protein in mutants (Fig. 2R). Similarly, a hypomorphic mi149 allele of pak2a (previously called redhead) resulted in disorganized pouches, as well as in variable loss of one to two pouches (Fig. 2E). In contrast, no pouch defects were observed in an ephb4ahu3378 allele (which results in a premature stop codon that deletes the tyrosine kinase domain), probably reflecting redundancy with other Eph receptors. Similarly, reduction of efnb3b with an efficient splice-blocking morpholino (MO) did not affect pouches (supplementary material Fig. S2A-C). However, efnb2a mutant pouch defects were enhanced by either ephb4a loss or efnb3b reduction with splice- or translation-blocking MOs (but not a control MO), thus revealing cryptic requirements for Ephb4a and Efnb3b in pouch formation (Fig. 2C,D; supplementary material Fig. S2B).
To confirm requirements of Eph-ephrin and Pak2a signaling in pouch development, we also analyzed the ceratobranchial cartilages (CBs) that depend on pouches for their development. As predicted, loss of pak2a or pairwise (but not individual) reductions of efnb2a, efnb3b and ephb4a resulted in losses and/or fusions of CBs (Fig. 2J-M; supplementary material Fig. S2C). In addition, efnb2a and pak2a single mutants displayed fusions of CB1 to the ceratohyal cartilage (CH) of the second arch. As efnb2a is expressed in neural-crest-derived cells of the second arch, we examined whether these CB1-CH fusions might be due to roles of Eph signaling in neural crest migration (Smith et al., 1997). However, analysis of dlx2a expression at 16.5 hpf revealed no intermixing of second and third arch neural crest streams (which generate CH and CB1) in efnb2a mutants, either alone or in combination with reduced efnb3b and ephb4a (supplementary material Fig. S2E). Furthermore, restoration of Efnb2a to the neural crest in sox10:Efnb2a transgenics failed to rescue the CB1-CH fusions and pouch/CB defects of Ephrin-deficient embryos (supplementary material Fig. S2B-D,G). Instead, live imaging of efnb2a−/− arches, followed by cartilage staining in the same individuals, revealed a precise correlation of CB1-CH fusions with earlier shortening of the her5:mCherryCAAX+ second pouch that separates second and third arch sox10:GFP+ neural-crest-derived cells (supplementary material Fig. S2F). Hence, both CB defects and CB1-CH fusions can be explained by earlier requirements of Eph-ephrin and Pak2a signaling in pouch development.
Endodermal roles of Eph-ephrin and Pak2a signaling in pouch development
To test whether Eph-ephrin and Pak2a function in the endoderm for pouch and cartilage development, we next blocked the activity of these pathways just in the endoderm, using dominant-negative transgenes. Specifically, we used an endodermal nkx2.3:Gal4VP16 driver (Choe et al., 2013) to express kinase-dead (KD) forms of EphB4a (Zisch et al., 1998) and Pak2a (Sells et al., 1997; Tang et al., 1997). Similar to loss-of-function embryos, endoderm-specific disruption of EphB or Pak2a activity resulted in reduced numbers of dysmorphic pouches, as well as losses and/or fusions of CB cartilages (Fig. 2F,N; supplementary material Fig. S2B,C). In addition, an nkx2.3:Efnb2a transgene, which selectively restored Efnb2a protein to endoderm, partially rescued the pouch and CB cartilage defects of efnb2a−/−; efnb3b-MO embryos (Fig. 2G,O,S,T). A hallmark of transmembrane B-type Ephrins is their ability to both activate Eph receptors in neighboring cells (forward signaling) and signal cell-autonomously (reverse signaling) (Kullander and Klein, 2002). However, a truncated version of Efnb2a only competent for forward signaling (Xu and Henkemeyer, 2009) also rescued pouch and CB defects (Fig. 2H,P,T). These results suggest that forward signaling within the endoderm regulates pouch development, although we cannot rule out additional roles of EphB signaling in other arch tissues. One possibility then is that distinct subsets of endodermal cells express either the ligand or receptor, thus allowing signaling in trans rather than cis-attenuation, with expression and signaling shifting as pouches mature from anterior to posterior.
Disintegration of pouch epithelia in the absence of Eph-Pak2a signaling
To investigate the requirements of Eph-Pak2a signaling in pouch cell behaviors, we performed time-lapse imaging of her5:mCherryCAAX-labeled pouches in signaling-deficient embryos. In wild-type siblings, pouches developed in an anterior-to-posterior progression, with subsets of cells becoming multilayered, collectively migrating and rearranging into bilayers as migration ceases (Fig. 3A-D; supplementary material Movie S1A). In contrast, we observed common pouch cell behavior defects in embryos lacking EphrinBs, EphB activity in the endoderm or pak2a (Fig. 3E-P; supplementary material Movie S1B-D). Although pouch initiation was unaffected, these embryos all displayed progressive detachment of cells from the epithelium during the branching process, with cells breaking off individually or as small clusters. Anti-Alcama staining of fixed embryos similarly revealed disorganized pouch bilayers and clusters of dissociated epithelial cells (Fig. 3Q-T; supplementary material Fig. S3). In particular, pouches deficient for Eph-ephrin or Pak2a displayed misaligned apical membranes, abnormal multi-layering and/or monolayers where bilayers should be present. Whereas we interpret this as a function of Eph-Pak2a signaling in increasing intercellular adhesion to stabilize pouches, Eph-Pak2a signaling might also help to maintain adhesion within the transitional epithelium.
EphrinBs and Wnt4a cooperatively activate Pak2a
As EphrinB-deficient and pak2a mutant embryos display similar pouch defects, we investigated whether Eph-ephrin signaling might activate Pak2a. PAK activation involves autophosphorylation of a conserved serine residue in an amino-terminal inhibitory domain (S143 in Pak2a) (Tay et al., 2010). During normal development, we observed pPAK-S143 staining within pouch epithelia, which increased as pouches matured (Fig. 4A; supplementary material Fig. S4A). Consistent with forward Eph-ephrin signaling activating PAK, pPAK-S143 staining was decreased in KD-EphB4a-expressing endoderm and efnb2a mutants, and further decreased in efnb2a−/−; efnb3-MO embryos (Fig. 4B,E,S; supplementary material Fig. S4D). PAK phosphorylation was also reduced upon endodermal expression of dominant-negative Cdc42 (supplementary material Fig. S4D), consistent with the known roles of small GTPases in activating PAK (Manser et al., 1994). As Wnt4a activates Cdc42 during pouch formation (Choe et al., 2013), we explored whether EphrinB and Wnt4a signaling might function together to maximally activate Pak2a. Indeed, we observed a reduction of pPAK-S143 staining in wnt4a mutants and a near-complete loss in compound efnb2a; wnt4a mutants (Fig. 4C,D,S). efnb2a; wnt4a mutants also displayed much more severe pouch and CB cartilage defects than either single mutant alone (Fig. 4D,J,S). This synergism is unlikely due to Wnt4a regulation of Efnb2a protein localization or ephb4a expression, as these were unaffected in wnt4a mutants (supplementary material Fig. S4B,C). Instead, the strong genetic interactions between Efnb2a and Wnt4a are consistent with Eph-ephrin and Wnt4a-Cdc42 signaling coordinately activating Pak2a.
Membrane targeting of Pak2a rescues pouches in EphrinB-deficient embryos
As we detected phosphorylated PAK at the apicolateral membranes of maturing pouch cells, we tested whether Eph-ephrin signaling might activate Pak2a by targeting it to the membrane. Consistently, pouch-specific expression of Pak2a with a membrane-targeting farnesylation sequence (nkx2.3:Pak2aCAAX) partially rescued pouch and CB defects and pPAK-S143 staining in efnb2a−/−; efnb3b-MO embryos (Fig. 4F,L,S). This rescue is unlikely due to increased general levels of Pak2a, as a non-membrane-targeted nkx2.3:Pak2a transgene failed to rescue (supplementary material Fig. S4E,G,H,J). Instead, rescue by membrane-targeted Pak2a suggests that Eph-ephrin signaling activates Pak2a by targeting it to the plasma membrane, where it could then be activated by Cdc42.
Eph-Pak2a signaling increases junctional E-cadherin localization in pouches
As the maturation of developing pouch epithelia into bilayers coincides with increased E-cadherin levels (Choe et al., 2013), we examined whether the disorganized bilayers of signaling-deficient embryos might be due to reduced junctional E-cadherin. Consistently, we observed a moderate decrease in junctional E-cadherin in efnb2a mutants, which was further exacerbated by reduction of efnb3b (Fig. 4N,Q,S). These reduced E-cadherin levels might be causative for pouch defects, as expression of a dominant-negative version of E-cadherin specifically in the nkx2.3+ endoderm resulted in similar pouch and CB defects to loss of EphrinBs (supplementary material Fig. S5). Moreover, the loss of junctional E-cadherin is probably due to reduced activation of Pak2a, because (1) similar reductions in E-cadherin were observed in pak2a mutants and dominant-negative Pak2a embryos (supplementary material Fig. S4F); (2) membrane-targeted Pak2a rescued E-cadherin levels in Ephrin-deficient embryos (Fig. 4R); and (3) the synergistic loss of phosphorylated PAK in efnb2a; wnt4a compound mutants was reflected by further reductions of E-cadherin levels compared with single mutants alone (Fig. 4P).
Previously, we had shown that Wnt4a promotes adherens junction stability in part through junctional localization of the Alcama immunoglobulin protein (Choe et al., 2013). However, the role of Wnt4a in activating Pak2a appears to be distinct from its role in localizing Alcama, as we detected no apparent changes in Alcama localization in pak2a mutants or embryos deficient for Eph-ephrin signaling (Fig. 3R-T; supplementary material Fig. S3). In addition, a sub-phenotypic dose of alcama-MO further decreased E-cadherin localization and enhanced the pouch and CB defects of pak2a mutants (supplementary material Fig. S4F-J). These results suggest that Eph-ephrin and Wnt4a signaling increase E-cadherin levels and intercellular adhesion through parallel Pak2a and Alcama pathways, which together would stop pouch outgrowth by inhibiting further epithelial rearrangements.
MATERIALS AND METHODS
Published lines include efnb2ahu3393 (Cavodeassi et al., 2013), pak2a (rhdmi149) (Buchner et al., 2007), Tg(∼3.4her5:EGFP)ne1911 (Tallafuss and Bally-Cuif, 2003), Tg(sox10:LOXP-GFP-LOXP-Hsa.DLX3)el8 (Das and Crump, 2012); and wnt4afh295, Tg(her5:mCherryCAAX)el72, Tg(nkx2.3:GAL4-VP16,myl7:EGFP)el93 and Tg(UAS:cdc42_T17N,cryaa:Cerulean)el268 (Choe et al., 2013). efnb2ahu3393 and ephb4ahu3378 were identified by the Zebrafish Mutation Project (ZMP) (http://www.sanger.ac.uk/Projects/D_rerio/zmp/). Translation-blocking (5′-ACTCC-CATCAAAGCCGTGTGCGGGA-3′), splice-blocking (5′-TGTGTGTT-ACTCACTCTGTCCCACT-3′) and control (antisense to splice-blocking, 5′-AGTGG-GACAGAGTGAGTAACACACA-3′) efnb3b MOs were obtained from Genetools and 1 nl of a 300-μM solution was injected at the one-cell-stage. alcama-MO is published (Choe et al., 2013). Tg(nkx2.3:Efnb2a), Tg(nkx2.3:Pak2aCAAX), Tg(UAS:KD-Ephb4a) and Tg(UAS:KD-Pak2a) transgenic constructs were generated using the Gateway (Invitrogen) Tol2kit (Kwan et al., 2007). See supplementary materials and methods for primers.
Immunohistochemistry was performed as described (Crump et al., 2004) using antibodies against Alcama/ZN8 (Zebrafish International Resource Center; 1:400), E-cadherin (610181, BD Biosciences; 1:250), pS143-PAK (pPAK144, a gift from Edward Manser, National University of Singapore; 1:250), GFP (TP401, Torrey Pines Biolabs; 1:1000) and Efnb2a (AF1088, R&D Systems; 1:500). Alcian Blue staining and fluorescent in situ hybridizations were as described (Zuniga et al., 2011) using probes against dlx2a (Zuniga et al., 2011) and PCR-generated probes against efnb2a, efnb3b, ephb4a and pak2a. See supplementary materials and methods for primer sequences.
Imaging and scoring
Fluorescent images were acquired on a Zeiss LSM5 confocal microscope. Dissected craniofacial cartilages were photographed on a Leica DM2500 upright microscope. To quantitate pouch defects, misshapen or normal pouches were scored as 1 and missing pouches as 0. For cartilage defects, normal CBs were scored as 1, two fused CBs as 1.5, reduced CBs as 0.5 and absent CBs as 0. Wild types invariantly had five pouches per side at 34 hpf (the sixth pouch forms by 38 hpf) and five CBs per side. For pPAK and E-cadherin, we scored from 3 for wild-type pouch staining to 0 for near complete loss of staining. Embryos were genotyped after scoring for an unbiased analysis. For time-lapse imaging (Crump et al., 2004), ∼80-μm-thick z-stacks were captured every 10 min at 1.5 μm intervals using a Zeiss LSM5 confocal microscope with 40× LD-Plan Neofluar objective. Maximum intensity projections encompassing 15-25 sections were assembled using ZEN software.
We employed the multiple comparison test of Tukey–Kramer to quantify pouch/CB defects and pPAK/E-cadherin staining. See supplementary material Table S1 for numbers of animals employed.
We thank the Hubrecht laboratory and the Sanger Institute Zebrafish Mutation Resource (ZF-MODELS Integrated Project; European Commission and Wellcome Trust grant 077047/Z/05/Z) for efnb2ahu3393 and ephb4ahu3378 alleles, Jordan Shavit for pak2a (rhdmi149) and Edward Manser for phosphoPAK antibody.
C.P.C. designed and performed the experiments. C.P.C. and J.G.C. interpreted results and wrote the manuscript.
Funding was from the National Institutes of Health [R01 DE022572 to J.G.C.] and a California Institute of Regenerative Medicine (CIRM) Postdoctoral Fellowship to C.P.C. Deposited in PMC for release after 12 months.
The authors declare no competing or financial interests.