Studies on epidermal growth factor receptor signaling in vertebrate limb patterning

Abstract

The epidermal growth factor receptor (EGFR) regulates multiple patterning events in Drosophila limb development, but its role in vertebrate limb morphogenesis has received little attention. The EGFR and several of its ligands are expressed in developing vertebrate limbs in manners consistent with potential patterning roles. To gain insight into functions of EGFR signaling in vertebrate limb development, we expressed a constitutively active EGFR in developing chick limbs in ovo. Expression of activated EGFR causes pre- and postaxial polydactyly, including mirror-image–type digit duplication, likely due to induction of ectopic expression and/or modulation of genes involved in anterior–posterior (AP) patterning such as Sonic hedgehog (Shh), dHand, Patched (Ptc), Gli3, Hoxd13, Hoxd11, bone morphogenetic protein 2 (Bmp2), Gremlin, and FGF4. Activation of EGFR signaling dorsalizes the limb and alters expression of the dorsal–ventral (DV) patterning genes Wnt7a, Lmx, and En1. Ectopic and/or extended FGF8 expressing apical ectodermal ridges (AERs) are also seen. Interdigital regression is inhibited and the digits fail to separate, leading to syndactyly, likely due to antiapoptotic and pro-proliferative effects of activated EGFR signaling on limb mesoderm, and/or attenuation of interdigital Bmp4 expression. These findings suggest potential roles for EGFR signaling in AP and DV patterning, AER formation, and cell survival during limb morphogenesis. Developmental Dynamics 233:288–300, 2005. © 2005 Wiley-Liss, Inc.

INTRODUCTION

Patterning of the vertebrate limb along the proximal–distal (PD), anterior–posterior (AP), and dorsal–ventral (DV) axes is controlled by the coordinated activity of three organizing centers (reviewed in Capdevila and Izpisua Belmonte, 2001; Tickle, 2003). The apical ectodermal ridge (AER), a thickened distal epithelium, maintains the proliferation of underlying distal limb mesoderm and promotes PD limb outgrowth. The AER also maintains the activity of the zone of polarizing activity (ZPA), a region of posterior mesoderm that controls the number and identity of the structures of the limb along the AP axis. Nonridge ectoderm controls limb DV patterning, which determines limb flexion and specifies formation of integument derivatives such as scales, claws, or feathers. Finally, localized regression of specific regions of limb mesoderm refines the limb contours and liberates the digits (Chen and Zhao, 1998; Zuzarte-Luís and Hurlé, 2002). These patterning processes are mediated by key signaling pathways using secreted factors belonging to the fibroblast growth factor (FGF), hedgehog, bone morphogenetic protein (BMP), and Wnt families (Chen and Johnson, 1999; Kawakami et al., 2001; Church and Francis-West, 2002; Niswander, 2002; Barrow et al., 2003; Panman and Zeller, 2003; Boulet et al., 2004; Wang et al., 2004).

The epidermal growth factor receptor (EGFR; also known as ErbB1 or HER1) is an archetypal tyrosine kinase receptor belonging to a family of four closely related cell surface receptors (ErbB1-4) and 11 ligands (reviewed in Normanno et al., 2001; Jorissen et al., 2003). The EGFR signaling family has been implicated in invertebrate and vertebrate limb development. The Drosophila EGFR and its ligands are required for fly wing formation (Clifford and Schüpbach, 1989; Simcox, 1997) and mediate multiple aspects of fly wing development, including determination of limb/body wall primordial and regulation of AP and DV wing disc patterning and growth (Baonza et al., 2000; Wang et al., 2000a; Zecca and Struhl, 2002a, b) and distalization and PD patterning of the leg (Campbell, 2002; Galindo et al., 2002). We previously have reported that components of the EGFR network (EGFR, EGF, and transforming growth factor-alpha [TGF-α]) are present in developing chick limb buds during the time at which the limbs are forming and undergoing major patterning events (Dealy et al., 1998). EGF receptor transcripts are expressed throughout the mesoderm and are abundant in the nonridge ectoderm of the limb and immunoreactive EGF and TGF-α are present in specific, localized domains in early limb bud mesoderm and ectoderm, consistent with multiple potential functions for the EGFR signaling network in limb development. Exogenous application of EGF and TGF-α to limb bud explants promotes proliferation and outgrowth of distal limb mesoderm, induces AER-like thickening of nonridge ectoderm, and enhances the morphogenesis of mutant limbless limb buds that lack an AER (Dealy et al., 1998). Although these observations provide support for roles for EGFR signaling in vertebrate limb outgrowth and patterning, limb defects have not been reported in transgenic mice lacking the EGFR (Miettinen et al., 1995; Sibilia and Wagner, 1995; Threadgill et al., 1995) or several EGFR ligands (Luetteke et al., 1993; Mann et al., 1993; Troyner et al., 2001). Others have suggested that functional redundancy within the EGFR signaling network explains the absence of predicted developmental phenotypes in these knockout mice (Troyner et al., 2001; Xian et al., 2001). Because of the potential for functional compensation, loss-of-function studies in transgenic mice may not be sufficient to fully evaluate roles of EGFR signaling in limb development.

As an alternate approach, and one that avoids functional redundancy, in this study, we have used gain-of-EGFR function to explore the role of EGFR signaling in limb development, using an replication-competent retroviral vector (RCAS) retroviral vector to express a constitutively active chicken EGFR (caEGFR) in developing chick limbs in ovo. Constitutively active receptors have proven useful for exploring roles of signaling factors in vertebrate limb patterning and skeletogenesis (Zou et al., 1997; Henderson et al., 2000; Zhang et al., 2000; Pizette et al., 2001). The activated EGF receptor used here signals ligand-independently but activates intracellular pathways used by the normal receptor, such as Ras/MAPK, and does not transform fibroblasts, thus mimicking normal, ligand-dependent EGFR signaling (Boerner et al., 2003). Expression of the activated EGFR in developing limbs causes polydactyly and syndactyly, dorsalizes the limb, and induces the formation of ectopic AERs. These results suggest potential unappreciated roles for EGFR signaling in AP and DV patterning, cell survival, and AER formation during limb morphogenesis.

RESULTS

Distribution of Retroviral Infection Correlates With Perturbation of Patterning Genes

Our retroviral infection protocols would be expected to result in activated EGFR expression as early as stage 17, coincident with the onset of limb budding (Morgan and Fekete, 1996). Distribution of retroviral protein by gag immunohistochemistry at day 5 (approximately stage 25) reveals that retroviral protein is detected in variable patterns, including nearly uniform distribution throughout the limb mesoderm and/or localized patches of intense immunostaining, and/or occasional expression in limb ectoderm or AER (see Fig. 2H). Comparison of adjacent sections subjected to gag staining and in situ hybridization reveals that, in many cases, limbs in which retroviral protein is widely or intensely expressed are also those in which patterning gene expression is perturbed (not shown).

Figure 2. Activation of epidermal growth factor receptor (EGFR) signaling induces ectopic domains of Shh expression, and ectopic FGF8-expressing apical ectodermal ridges (AERs). Day 5. A,B: Control (CT, A) and contralateral RCAS-caEGFR (B) legs (posterior to the right) showing an ectopic domain of Shh in the anterior mesoderm of the RCAS-caEGFR leg (*Shh in B). C,D: Control (C) and contralateral RCAS-caEGFR (D) legs (distal view, posterior at the top). In addition to the normal domain of Shh expression in posterior mesoderm (C), a diffuse ectopic Shh domain (*Shh, dashed line) is present in the distal mesoderm of the RCAS-caEGFR leg (D). Multiple ectopic FGF8 expressing apical ectodermal ridges (AERs) have also formed in the ventral and dorsal ectoderm perpendicular to the endogenous AER (*FGF8, arrows). Note also the extension of the FGF8 expression domain of the endogenous AER into the nonridge dorsal and ventral ectoderm (arrowhead). E,F: Control (E) and contralateral RCAS-caEGFR (F) legs (anterior view). The distal leg tips are aligned by the top dashed line. The anterior limits of FGF8 expression in the AERs (FGF8, arrows) do not align (middle dashed line), revealing that the domain of FGF8 expression in the AER of the RCAS-caEGFR leg is considerably extended anteriorly (F). In F, an ectopic FGF8-expressing AER has also formed in the ventral ectoderm perpendicular to the endogenous AER (*FGF8). G: Another RCAS-caEGFR leg showing a thickened, ectopic FGF8-expressing AER (*FGF8) that has formed parallel to the endogenous AER. H: RCAS-caEGFR limb showing localization of viral gag protein not only in mesoderm but also in ectoderm, including the duplicated AER structure (*AER).

Expression of Activated EGFR Causes Polydactyly and Syndactyly

Living day 10 RCAS-caEGFR limbs are broad, short, thick and syndactylous, and possess extra anterior and/or posterior tissue masses (Fig. 1B,C,I). Subsequent Alcian blue staining to reveal skeletal element patterning shows that the limbs are polydactylous, with two to four extra digits forming from posterior (Fig. 1E,J) or anterior (Fig. 1F,K–M) mesoderm with about equal frequency. Notably, in the most dramatic cases of preaxial polydactyly, multiple digit structures identified by phalange number as posterior are present, indicating they are mirror-image–type duplications (Fig. 1K–M).

Figure 1. Activation of epidermal growth factor receptor (EGFR) signaling causes syndactyly and polydactyly. Day 10. A,D: Control living (A) and Alcian blue-stained (D) wing has three digits (numbered 2–4 in D). B,C,E,F: RCAS-caEGFR living (B,C) and Alcian blue-stained (E,F) wings are short, broad, and possess ectopic posterior or anterior tissue masses (arrows in B,C) that contain ectopic postaxial or preaxial cartilage elements resembling digits (*? in E,F). G,H: Control living (G) and Alcian blue-stained (H) leg in which the interdigital mesoderm (im in G) has regressed and the digits separated. The digits (numbered 1–4 in H) are identified by counting phalanges (2, 3, 4, and 5 in each, respectively). There is a single fibula (f). I,J: RCAS-caEGFR living (I) and Alcian blue-stained (J) leg in which interdigital mesoderm (im in J) has failed to regress, leading to syndactyly. In addition, an ectopic posterior tissue mass is present (arrow in I) that contains an extra postaxial digit (*? in J). K–M: Alcian blue-stained RCAS-caEGFR legs with preaxial mirror-image–type polydactyly. Counting of phalange numbers of the digital structures identifies the digit pattern in K as 4321*1*2, in L as 4321*2*3*4*?, and in M as 432*?*2*3*?. In L,M, extended and/or ectopic fibulae are present (*f), and in M the tibia (*t) is also partially bifurcated.

Analysis of the pattern of the skeletal elements of all day 10 limbs examined, which had received RCAS-caEGFR injection or implantation of RCAS-caEGFR expressing cells, shows that 27% (61 of 225) had alterations in limb patterning, and 22% (50 of 225) had the polydactylous and syndactylous phenotype described above. Injection of RCAS vector alone or implantation of RCAS vector-expressing cells produced only a very low frequency of defective limbs (3% or 2 of 64), and these were relatively mildly affected (one wing with a missing digit 2 and one leg with a missing digit 1). We did not obtain ectopic limbs in response to microinjection of RCAS-caEGFR virus or implantation of RCAS-caEGFR -infected fibroblasts into the prospective flank (0/109, data not shown), although this procedure did produce a low frequency of typical abnormal wings and/or legs indicating presence of functional virus in the flank in these experiments.

Activation of EGFR Signaling Alters the Expression of AP Patterning Genes

Shh, a secreted signaling molecule expressed by the ZPA that controls digit number and identity (te Welscher et al., 2002a; Litingtung et al., 2002), is ectopically expressed in RCAS-caEGFR limbs (9% or 5 of 59). Ectopic Shh may be highly localized to the anterior mesoderm, indicating the presence of a duplicated ZPA (Fig. 2B, see also Fig. 3B) or may be present as a diffuse expansion of the normal posterior Shh domain (Fig. 2E).

Figure 3. Expression of anterior–posterior patterning genes is altered by activation of epidermal growth factor receptor (EGFR) signaling. Posterior is to the left. Day 5. A–F: Adjacent sections of control (CT, A,C,E) and contralateral RCAS-caEGFR (B,D,F) wings. In the RCAS-caEGFR wings, there is ectopic anterior Shh (arrow in B), and in the same location, ectopic anterior Ptc (*bar in D), and loss of anterior Gli3 (*dashed bar in F). G–J: Adjacent sections of control (G,I) and RCAS-caEGFR (H,J) wings showing ectopic anterior dHand expression (H) even in the absence of ectopic anterior Shh expression (J). K,L: Control (K) and contralateral RCAS-caEGFR (L) wings showing mild expansion of the domain of Hoxd13 expression (dotted line) in the RCAS-caEGFR wing (L). M–R: Adjacent sections of control (M,O,Q) and contralateral RCAS-caEGFR (N,P,R) wings showing ectopic anterior and distal Bmp2 (arrows in N) and Gremlin (arrow in P) and expansion of the domain of FGF4 expression in the AER (compare posterior limit of FGF4 expression in Q and R, arrows, and note extended and/or ectopic anterior expression of FGF4 in R, *arrow). [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Ectopic, bifurcating, or branching AER structures, which express FGF8, a molecular marker of the AER (Crossley et al., 1996), are also present in the ectoderm of RCAS-caEGFR limbs (Fig. 2D,F,G, 15% or 3 of 20). The extra AERs form parallel or perpendicular to the endogenous AER and maintain connection to it. They arise from either dorsal (Fig. 2D) and/or ventral (Fig. 2D,F,G) ectoderm. In addition, the endogenous AER is thickened with regions of FGF8 expression that extend into dorsal and/or ventral nonridge ectoderm (Fig. 2D,F). The domain of FGF8 expression in the endogenous AERs of RCAS-caEGFR limbs is also extended anteriorly (Fig. 2F, 25% or 5 of 20; as is the domain of FGF4 expression, see Fig. 3R and below). Note that viral gag protein is present in the duplicated AER structure of the RCAS-caEGFR limb shown in Figure 2H, as well as in the mesoderm.

Expression of genes regulated by Shh is altered in RCAS-caEGFR limbs. Patched (Ptc) is an Shh receptor whose expression is up-regulated in the presence of Shh signaling (Marigo et al., 1996). In the RCAS-caEGFR limb, domains of intense Ptc expression are present not only adjacent to the normal Shh-expressing ZPA but also in the anterior mesoderm (Fig. 3D), in the same region that expresses ectopic anterior Shh (Fig. 3B). Ectopic expression of Ptc is detected in RCAS-caEGFR limbs only in the presence of ectopic Shh. Gli3 is a factor required for digit patterning whose expression in posterior mesoderm is suppressed by Shh (te Welscher et al., 2002b). Gli3 expression is attenuated in same region of the RCAS-caEGFR limb that also expresses ectopic Shh and Ptc (Fig. 3F). Like Ptc, alterations in Gli3 expression are not detected in any of 23 additional RCAS-caEGFR limbs analyzed that do not also express ectopic Shh.

dHand is a factor expressed by posterior limb mesoderm that restricts Gli3 expression and/or Gli3 repressor function to anterior mesoderm (te Welscher et al., 2002b). dHand is strongly ectopically expressed in the anterior mesoderm of the majority of RCAS-caEGFR limbs analyzed (Fig. 3H, 72% or 8 of 11). Of interest, in none of these limbs is ectopic Shh also present (see Fig. 3J). The posterior expression of 5′ members of the HoxD genes is essential for digit development (Zakany et al., 1997). The domains of expression of the 5′HoxD genes Hoxd13 (Fig. 3L) and Hoxd11 (not shown) are consistently but only modestly expanded anteriorly in RCAS-caEGFR limbs (75% or 3 of 4 for Hoxd13 and 100% or 4 of 4 for Hoxd11).

Bmp2 and its antagonist Gremlin are present in posterior and distal mesoderm and function to balance the Shh/FGF signaling loop that promotes limb outgrowth and patterning (Khokha et al., 2003). The anterior mesoderm of RCAS-caEGFR limbs ectopically expresses Bmp2 (Fig. 3N) and Gremlin (Fig. 3P; 20% each or 4 of 20). In addition, the expression domain of FGF4 in the AER is considerably extended posteriorly, and an extended or ectopic FGF4 expression domain is also present anteriorly (Fig. 3R).

EGFR Activation Dorsalizes the Limb Structures and Alters Expression of DV Patterning Genes

The dorsal and ventral integument structures of the normal chick foot are morphologically distinct (Prin et al., 2004): the dorsal surface is covered by large overlapping scales, the ventral surface by tiny round scales, and the claws are ventrally curved and contain a single dorsal nail bed (Fig. 4A,C). RCAS-caEGFR limbs are dorsalized. At day 14/15, ectopic dorsal scales are present on the ventral surface of the autopod (metatarsal and/or digit III) of RCAS-caEGFR legs, the digits are markedly dorsiflexed, and the claw of digit III is conical and possesses a duplicated nail bed (Fig. 4B,D). Subsequent analysis of internal structures confirms the presence of ectopic dorsal scales and loss of ventral ones, and also shows that a prominent ventral tendon is lost (Fig. 4F). These phenotypes are present in 43% (3 of 7) of the day 14/15 limbs analyzed. Analysis of the expression of Lmx1, Wnt7a, and En1, genes that pattern the limb DV axis (Chen and Johnson, 1999; Prin et al., 2004) confirms dorsalization in response to activation of EGFR signaling as early as day 5, the earliest day examined (Fig. 4G–L). RCAS-caEGFR wings express ectopic ventral Wnt7a (Fig. 4H, 24% or 10 of 42) and Lmx (Fig. 4J, 21% or 9 of 43), genes that are normally restricted to dorsal ectoderm and mesoderm, respectively (Fig. 4G,I). In addition, expression of En1, a marker of ventral limb ectoderm, is lost in the ventral ectoderm adjacent to the region of ectopic Lmx expression (Fig. 4L, 17% or 7 of 42).

Figure 4. Activation of epidermal growth factor receptor (EGFR) signaling dorsalizes the limb. A,B: Living day 15 control (CT, A) and contralateral RCAS-caEGFR (B) legs. Large, overlapping dorsal scales (ds) are present on the dorsal surface of the autopod of the control leg (A) and small rounded ventral scales (vs) are present on the ventral surface. In the RCAS-caEGFR leg (B), ectopic dorsal scales (*ds) are present on the ventral surfaces of the metatarsal region and digit III. C,D: Higher magnification view of the claws of digit III of the limbs in (A,B). The control claw (C) is ventrally curved and has a single nail bed (nb), whereas the RCAS-caEGFR claw is symmetrical in shape and the nail bed is duplicated (*nb in D). E,F: Cross-sections of digit III of the legs shown in (A,B). In the control digit (E), dorsal scales (ds) are limited to the dorsal surface and ventral scales (vs) to the ventral surface, a prominent ventral tendon (vt) is present and the cartilage element has an asymmetrical shape. In the RCAS-caEGFR digit (F), ectopic dorsal scales (*ds) are present on the ventral surface, ventral scales are reduced in number, the ventral tendon is lost or rudimentary, and the cartilage element is round. G,H: Day 5 control (G) and contralateral RCAS-caEGFR (H) wings showing ectopic expression of Wnt7a in the ventral ectoderm of the RCAS-caEGFR wing (arrow in H). I–L: Adjacent sections of control (I,K) and contralateral RCAS-caEGFR (J,L) wings showing ectopic expression of Lmx (arrow in J) and in the same region, loss of expression of En1 (arrow in L) in the RCAS-caEGFR wing.

Perturbation of dorsal/ventral patterning is also indicated by the presence of small “bumps” of tissue on the dorsal surface of 5% (4 of 86) of RCAS-caEGFR limbs at day 7 (see Fig. 5D). These bumps of tissue do not appear to develop into dorsal digits, because we do not observe ectopic dorsal digits in limbs analyzed at day 10 for the presence of cartilage elements.

Figure 5. Epidermal growth factor receptor (EGFR) signaling in interdigital regression. Day 7. A,B: Endogenous expression of ErbB1 (EGFR) and immunolocalization of EGF in the normal leg autopod. ErbB1 is expressed throughout the interdigital mesoderm (im in A), whereas EGF is limited to the tissue at the digit tips and adjacent to the digit rays (arrows in B). C,D: Nile blue-stained control (C) and RCAS-caEGFR (D) legs showing numerous dying cells in the central interdigital mesoderm (im) and posterior necrotic zone (pnz) of the control leg (C) but few dying cells in the RCAS-caEGFR leg (D), which is also broad and has excess tissue at the anterior and posterior margins (arrowheads in D). E,F: Control (CT, E) and RCAS-caEGFR (F) legs showing loss of expression of Bmp4 in the interdigital mesoderm (im) and posterior necrotic zone (arrowhead) of the RCAS-caEGFR leg (F), but no change in Bmp4 expression by the perichondrium (pc). G,H: Adjacent sections to those shown in (E,F) showing no change in expression of Msx2 in interdigital mesoderm (im) and in the region corresponding to the pnz (arrowhead) of the RCAS-caEGFR leg (H).

Digit/Interdigit Region Is a Potential Site of Endogenous EGFR Signaling, and Interdigital Cell Death and Expression of Bmp4 Are Suppressed in Response to Expression of Activated EGFR

Regression of interdigital mesoderm is necessary for digit separation and requires a balance of cell death and cell proliferation (Salas-Vidal et al., 2001). In normal legs, ErbB1 (EGFR) is expressed throughout the interdigital mesoderm, including the central interdigital mesoderm and the mesoderm adjacent to the digit rays (Fig. 5A). In contrast, the EGFR ligand EGF is present only in a subset of the EGFR expression domain, as it is not present in central interdigital mesoderm but is limited to the mesoderm adjacent to the digit rays (Fig. 5B), a region that does not undergo cell death. Nile blue staining confirms the restriction of dying cells to the central interdigital mesoderm and region corresponding to the posterior necrotic zone (pnz) of normal legs (Fig. 5C). However, in RCAS-caEGFR legs, few or no dying cells are present in these regions (Fig. 5D, 100% or 13 of 13). In addition, expression of the proapoptotic factor Bmp4 (Zuzarte-Luís and Hurlé, 2002) is attenuated in the central interdigital mesoderm of RCAS-caEGFR legs (Fig. 5F, 75% or 3 of 4), although expression of Msx2, a factor also implicated in interdigital cell death, is not affected (Fig. 5H, 0 of 4). Effects of activation of EGFR signaling on apoptosis at stages earlier than day 7 were not examined.

DISCUSSION

Generation of Polydactyly Through Interplay of EGFR Signaling and AP Patterning Genes

Current models for establishment of limb AP patterning involve complex interaction between Shh, Gli3, dHand, and 5′ members of the HoxD genes (reviewed in Panman and Zeller, 2003, see also Chen et al., 2004; Zakany et al., 2004). It is thought that early expression of 5′HoxD genes in posterior mesoderm activates both Shh and dHand (Zakany et al., 2004), which down-regulate Gli3 expression or prevent its conversion to the active repressor form (Litingtung et al., 2002, te Welscher et al., 2002a, b). This process enables subsequent posterior expression of the genes and signals required for elaboration and specification of the digits, as well as the establishment of the Shh/Gremlin/BMP/FGF signal relay that promotes distal progression (te Welscher et al., 2002b, Khokha et al., 2003). It has been suggested that polydactylous phenotypes result from loss of anterior Gli3 expression or Gli3 repressor function (Wang et al., 2000b; Litingtung et al., 2002; Chen et al., 2004), either through genetic perturbation of Gli3 (Litingtung et al., 2002; te Welscher et al., 2002a) or by means of ectopic anterior expression of Shh, dHand, or 5′HoxD genes (Charite et al., 2000; Fernandez-Teran et al., 2000; te Welscher et al., 2002b; Chen et al., 2004; Zakany et al., 2004). The presence of ectopic Shh and dHand in the anterior mesoderm of RCAS-caEGFR limbs suggests polydactyly in response to activated EGFR signaling results from Shh- and dHand-mediated repression of anterior Gli3 expression or function. Consistent with this finding, in limbs in which expression of both Shh and Gli3 were analyzed, down-regulation of Gli3 occurred only in the presence of ectopic Shh. The comparatively mild ectopic expression of Hoxd13/Hoxd11 in the anterior mesoderm of RCAS-caEGFR limbs suggests that this is a later event in generation of the polydactylous phenotype, occurring downstream of ectopic anterior Shh and dHand expression.

Studies place the role of dHand in AP patterning as acting in parallel with Shh (Zakany et al., 2004) or upstream of it (Charite et al., 2000; te Welscher et al., 2002b, Panman and Zeller, 2003). Ectopic anterior dHand achieves the same effects in AP patterning as ectopic Shh: mirror-image digit duplication; induction of anterior Ptc, Bmp2, Gremlin, Shh, and 5′HoxD expression; and suppression of anterior Gli3 transcription (Charite et al., 2000; Fernandez-Teran et al., 2000; Yelon et al., 2000; McFadden et al., 2002; te Welscher et al., 2002b). In our study, the frequency of ectopic Shh expression in RCAS-caEGFR limbs (9% or 5 of 59), is somewhat less than what would be expected based on the frequency of mirror-image–type duplications obtained (14% or 7 of 50). Of interest, the frequency of ectopic dHand expression in RCAS-caEGFR limbs is much higher than that of ectopic Shh, as dHand is present in the anterior mesoderm in nearly three fourths (72% or 8 of 11) of the limbs analyzed simultaneously for dHand and Shh expression, and in no case was ectopic Shh also induced. Although it is possible that Shh is present in these limbs but below our level of detection, ectopic expression of Ptc, an extremely sensitive indicator of Shh signaling (Marigo et al., 1996), is seen in RCAS-caEGFR limbs only in the presence of ectopic Shh. These observations raise the intriguing possibilities that activated EGFR signaling may be capable of inducing either dHand or Shh, and/or that dHand may be the primary target of activated EGFR signaling, with subsequent induction of ectopic Shh.

During normal development, ErB1 (EGFR) transcripts are expressed in the mesoderm of the lateral plate, where prepatterning of the limb AP axis occurs, and in the budding limbs during establishment of the ZPA (Dealy et al., 1998). Thus, endogenous EGFR is present in the right time and place to potentially play a role in AP patterning and/or regulation of Shh or dHand expression. The mechanism whereby EGFR signaling may regulate Shh or dHand expression is not clear, but it has been shown that dHand phosphorylation and transcriptional activity is regulated by means of the Akt pathway, one of the pathways activated by EGFR signaling (Murakami et al., 2004). A role for EGFR signaling in vertebrate limb AP pattering is also consistent with functions of this receptor in the Drosophila wing imaginal disc, where critical interactions between the EGFR and Hh signaling pathways regulate AP patterning (Crozatier et al., 2002), and perturbation of EGFR signaling causes mirror-image wing duplications and alters expression of patterning genes, including Cubitus interruptus (Baonza et al., 2000), the functional equivalent of vertebrate Gli3.

The ligand mediating EGFR activities in limb AP patterning is not obvious. EGF and TGF-α are both present in the limb forming mesoderm of the lateral plate (Dealy et al., 1998) and in distal and peripheral mesoderm of the limb bud, including the posterior limb mesoderm, but are not specifically localized to the ZPA (Dealy et al., 1998). EGF or TGF-α could play a role in AP patterning, even acting at a distance or in low amounts, as amplification of the EGFR signal occurs after ligand binding by means of various mechanisms, including receptor clustering on the cell surface (Ichinose et al., 2004) and formation of autocrine loops with positive feedback (Shvartsman et al., 2002). The availability of ligand for EGFR activation is determined by local protease activity, which releases the ligands from the cell surface (Sahin et al., 2004), and subsequently by the ability of the ligand to reach the receptor by means of diffusion. Notably, direct measurement of the diffusion of EGF, a relatively small peptide, indicates that it diffuses much more readily through cells than larger peptides, and can reach a given concentration hundreds of microns further (Thorne et al., 2004). This finding suggests that an EGF source could be relatively far from target cells and still be able to elicit a biological response (Thorne et al., 2004). Alternately, other ligands for the EGFR also exist, including HB-EGF, Amphiregulin, Betacellulin, and Epiregulin (Normanno et al., 2001), but their expression patterns in chick limbs have not been reported.

EGFR Signaling in AER Formation and Position

The presence of ectopic AERs in response to activated EGFR signaling suggests a role for EGFR signaling in AER formation. The AER is induced from presumptive AER-forming ectoderm (Altabef et al., 1997; Michaud et al., 1997) in response to signals from the prospective limb forming mesoderm (Kieny, 1960; Douailly and Kieny, 1972; Saunders and Reuss, 1974). Mesodermal signals in induction of the AER include FGF10 (Ohuchi et al; 1997) and, more recently, BMPs (Ahn et al., 2001; Pizette et al., 2001; Soshnikova et al., 2003), which are required for induction of β-catenin/Wnt3a in the overlying ectoderm, which subsequently induce the AER and activate expression of AER genes, including FGF8 (Kawakami et al., 2001; Barrow et al., 2003; Soshnikova et al., 2003). Our study does not allow us to distinguish between mesodermal or ectodermal functions of EGFR signaling in AER formation. EGFR signaling could regulate expression or activity of a mesodermal or ectodermal factor involved in AER formation, similar to its ability to regulate Bmp2 or FGF8 (as shown in Figs. 2, 3). An endogenous role for EGFR signaling in AER induction is supported by our previous observations that the EGF receptor and its ligands, particularly TGF-α, are colocalized in prospective limb-forming mesoderm and AER-forming ectoderm (Dealy et al., 1998) and that exogenous TGF-α induces the formation of a thickened AER-like ectoderm in mutant limbless and wingless limbs (Dealy et al., 1998), which lack an AER.

Mispositioned or excessively widened AERs form in limb ectoderm in response to manipulation of ectodermal Wnt3a/B-catenin (Kengaku et al., 1998; McQueeney et al., 2002; Barrow et al., 2003; Soshnikova et al., 2003), FGF10/FGF8 (Ohuchi et al., 1997), radical fringe (Laufer et al., 1997; Rodriguez-Esteban et al., 1997), En1 (Loomis et al., 1998; Pizette et al., 2001), and the BMP antagonist Noggin (Ahn et al., 2001; Pizette et al., 2001; Khokha et al., 2003; Wang et al., 2004), suggesting roles for these pathways in positioning of the AER and/or in defining its boundaries. The boundaries of the AER are defined by lineage restrictions between the AER and nonridge ectoderm (Kimmel et al., 2000). Activation of EGFR signaling in the ectoderm in the present study could disrupt these lineage boundaries, leading to ectopic AER formation. A role for EGFR signaling in defining lineage boundaries in limb ectoderm would be consistent with the role of EGFR signaling in the Drosophila wing imaginal disc to direct the cell fate decisions that specify the wing-forming from non–wing-forming cells (Baonza et al., 2000; Zecca and Struhl, 2002a) and that establish the DV cell lineage-restricted compartment boundary (Zecca and Struhl, 2002b) along which forms the wing margin, an outgrowth-promoting signaling center analogous to the vertebrate AER (Dahman and Basler, 1999).

Maintenance of the ectodermal lineage boundaries that define the AER is suggested to require BMPR signaling by the AER itself (Wang et al., 2004), as antagonism of BMPR signals by means of ectodermal Noggin expression expands the AER boundaries and induces ectopic AERs (Pizette et al., 2001; Wang et al., 2004). EGFR signals are antagonistic to BMPR signals (Kretzschmar et al., 1997). Opposing activities of the EGFR and BMPR are involved in patterning of the Drosophila head and wing (Croatzier et al., 2002; Chang et al., 2003), and in vertebrate heart and skeletal development (Nonaka et al., 1999; Jackson et al., 2003). Thus, as in the case of Noggin (Wang et al., 2004), EGFR-mediated antagonism of BMPR signals within the AER could disrupt the AER/nonridge ectoderm boundary, resulting in ectopic AER formation. That ectopic expression of the activated EGF receptor is indeed present within the duplicated AERs of an RCAS-caEGFR limb (as determined by viral gag distribution) supports this idea (see Fig. 2H).

We have shown previously that, during normal development, ErbB1 (EGFR) is expressed by nonridge ectoderm but not by the AER and have suggested that exclusion of EGFR expression from the AER may be required for some aspect of AER formation and/or function (Dealy et al., 1998). Based on the observations above, exclusion of EGFR signaling from the AER may be necessary to permit BMPR-mediated repression of the conversion of nonridge ectoderm to AER. In turn, the repressive effects of AER-BMPs might be counterbalanced by EGFR signaling in nonridge ectoderm, which may promote conversion of nonridge ectoderm into AER, perhaps by means of interaction with other ectodermal factors such as Wnt3a/β-catenin, radical fringe, or FGFs.

EGFR Signaling in PD Outgrowth

Limb outgrowth and patterning is coordinated by reciprocal signals between the distal mesoderm, AER, and ZPA (Zwilling, 1961; Saunders, 1977; Laufer et al., 1994; Niswander et al., 1994; Boulet et al., 2004). BMP signals in distal mesoderm suppress the expression of FGFs by the overlying AER, and antagonism of this activity is required for maintenance of the AER and subsequent maintenance of Shh expression by the ZPA (Zuniga et al., 1999). Gremlin has been identified as the BMP antagonist expressed by posterior distal mesoderm that maintains this FGF/Shh signaling loop (Khokha et al., 2003). The ectopic anterior and distal expression of Bmp2 and Gremlin in RCAS-caEGFR limbs likely reflects their induction by ectopic Shh (Merino et al., 1999, Zuniga et al., 1999; Drossopoulu et al., 2000) and/or ectopic dHand (te Welscher et al., 2002b) and is consistent with the existence of an expanded and/or additional outgrowth and patterning center in the anterior distal mesoderm of RCAS-caEGFR limbs. Induction of ectopic Gremlin is also consistent with subsequent expansion of overlying FGF8 and FGF4 expression domains in response to activated EGFR signaling (Zuniga et al., 1999; Khokha et al., 2003). Expansion of FGF8 and FGF4 could also result from direct antagonism of BMPR signals by means of activated EGFR signaling in distal mesoderm.

We have suggested previously that endogenous EGFR signaling may be involved in reciprocal interactions required for PD limb outgrowth, as EGF and TGF-α are present in the AER and distal limb mesoderm, can suppress the differentiation and promote the proliferation of subridge mesoderm, and can maintain the thickness and activity of the normal AER in vitro (Dealy et al., 1998). Recent studies in Drosophila reveal that EGFR signaling is also required for PD outgrowth of the fly leg (Campbell, 2002; Galindo et al., 2002) and suggest that the tip of the leg provides a source of an EGFR ligand that promotes PD outgrowth in a manner similar to the vertebrate AER (Campbell, 2002).

The extension of the domains of FGF4 and FGF8 expression in the AER of RCAS-caEGFR limbs may explain some of the more mild cases of polydactyly, and/or the postaxial polydactyly, produced in response to activated EGFR signaling. An ectopic FGF signal produced by an extended AER has been suggested to cause non–mirror-image pre- and postaxial polydactyly due to recruitment of normally non–limb-forming cells from flank mesoderm located anterior and posterior of the limb bud (Tanaka et al., 2000). Postaxial polydactyly is also caused by expansion of the AER and of the FGF domains within it as a result of inhibition of BMPR signaling by Noggin (Guha et al., 2002; Wang et al., 2004). Finally, because one major function of the AER is to promote the survival and proliferation of underlying mesoderm (Kosher et al., 1979; Rowe et al., 1982), the extended AERs induced by activated EGFR signaling may promote local accumulation of limb bud mesoderm with intrinsic digit-forming potential (Omi et al., 2002).

EGFR Signaling in Regulation of DV Pattern

The formation of ectopic dorsal structures (ectopic dorsal scales and duplicated nail beds) and the loss of ventral ones (missing ventral scales and ventral tendon) in limbs in which activated EGFR is expressed indicate that EGFR signaling dorsalizes the limb. Dorsalization of RCAS-caEGFR limbs is confirmed by the early presence of ectopic Lmx and Wnt7a expression in ventral mesoderm and ectoderm, respectively, and by the loss of expression of En1. BMPR signaling is upstream of Lmx, Wnt7a, and En1 signals in limb DV patterning, as loss of ectodermal BMPR signaling achieved by conditional knockout of BMPRIA or misexpression of the BMP antagonist Noggin results in limb dorsalization (Ahn et al., 2001; Pizette et al., 2001; Barrow et al., 2003; Soshnikova et al., 2003; Wang et al., 2004). Accordingly, in the current model for limb DV patterning, ectodermal BMPs positively regulate expression of En1 by the ventral ectoderm, whereas En1, in turn, represses Wnt7a expression and limits it to the dorsal ectoderm, where it subsequently induces expression of Lmx by underlying dorsal mesoderm (see Fig. 8 of Pizette et al., 2001). In the present study, expression of activated EGFR in ventral ectoderm could dorsalize the limb by means of antagonism of BMPR signals, similar to the dorsalized limbs obtained after misexpression of the BMP antagonist Noggin.

We have shown previously that, in the normal limb, ErbB1 (EGFR) is expressed by both dorsal and ventral nonridge ectoderm but EGF is only present in ventral ectoderm (Dealy et al., 1998). The presence of an endogenous EGFR signal in ventral ectoderm could serve to counterbalance the ventralizing signals provided by BMPs, similar to our suggestion that nonridge EGFR signals may counterbalance AER-BMPs (see above). Remarkably, EGFR signaling directs dorsal cell fate decisions in the Drosophila wing imaginal disc, where it regulates expression of the Lmx homolog apterous by providing a permissive signal, active only in the presence of the fly EGFR ligand Vein (Zecca and Struhl, 2002a, b). A role for EGFR signals in limb dorsalization is also supported by our observation that the EGFR ligand EGF is ectopically localized throughout the bidorsal ectoderm of mutant limbless limb buds (Dealy et al., 1998).

EGFR Signaling Negatively Regulates Interdigital Regression

In the normal autopod during the time at which interdigital regression is occurring, ErbB1 (EGFR) and EGF are colocalized by the mesoderm adjacent to the digit rays, which does not participate in the cell death occurring in the adjacent central interdigital mesoderm (Ganan et al., 1996). This finding suggests that ligand-activated EGFR may function endogenously as a survival signal to prevent apoptosis of tissue not scheduled to regress. Endogenous EGFR signaling in the tissue adjacent to the digit ray may also suppress the pro-apoptotic activities of BMPs present in the adjacent interdigital mesoderm, thus confining regression to the central interdigital mesoderm. That activation of EGFR signaling prevents cell death and regression in the central interdigital mesoderm while suppressing Bmp4 expression is consistent with this idea. The loss of cell death and Bmp4 expression in the posterior necrotic zone of the RCAS-caEGFR limbs also suggests that its gross broadening may result from perturbation of the cell death and proliferative processes that normally refine the limb contours (Chen and Zhao, 1998). Inhibition of cell death and stimulation of proliferation as a consequence of activated EGFR signaling is consistent with the known involvement of ErbB survival pathways in development, tissue remodeling, and tumorigenesis (Danielson and Maihle, 2002), as well as in the regulation of cell growth and cell cycle progression (Grant et al., 2002).

Complexity of EGFR Signaling

EGFR signaling activates multiple distinct intracellular signaling cascades (Yarden and Sliwkowski, 2001), depending on which ligand is involved (Sweeney et al., 2001) and whether signaling is occurring from the homodimeric EGFR or from an EGFR-ErbB heterodimer (Olayioye et al., 1998). EGFR signaling is further diversified through interaction with other signaling networks such as those mediated by Wnts (Civenni et al., 2003), BMPs (Nonaka et al., 1999; Jackson et al., 2003), insulin-like growth factors (Hallak et al., 2002; Adams et al., 2004), and FGFs (Wu et al., 2003), which have themselves been implicated in a wide array of roles in vertebrate limb development (reviewed in Capedevila and Izpisua Belmonte, 2001; see also Dealy and Kosher, 1995, 1996; Allan et al., 2000; Dahn and Fallon, 2000; Ahn et al., 2001; Pizette et al., 2001; Church and Francis-West, 2002; Barrow et al., 2003). The diverse limb phenotypes obtained in response to activated EGFR expression in the present study likely reflect the nature of the EGFR signaling network as a paradigm for signal diversification (reviewed in Hackel et al., 1999; Gschwind et al., 2001; Prenzel et al., 2001). Specificity of the cellular responses resulting from activation of various signaling cascades is determined by the spectrum of intracellular mediators expressed within the responding cells (reviewed in Chang and Karin, 2001; Dumont et al., 2002; Tanoue and Nishida, 2003). Recently, it has been shown that overexpression of an FGF-regulated intracellular signaling intermediate that relays signals between the Ras/MAPK and PI(3)K/Akt pathways inhibits limb outgrowth, induces apoptosis, and causes limb truncations (Eblaghie et al., 2003; Kawakami et al., 2003). The markedly different phenotypes observed in the present study support the existence of distinct EGFR and FGFR signaling mechanisms in the limb, and suggest that the limb patterning defects caused by activation of EGFR signaling are unlikely to be mediated by means of alteration of FGFR signaling.

The results of this study suggest novel and previously unappreciated roles for EGFR signaling in vertebrate limb morphogenesis. Questions still to be addressed include identification of the EGFR ligands involved in various limb patterning processes, the potential roles of other ErbB family members in limb development, and relationship of EGFR signaling pathways to other signaling networks, particularly the BMPR network. Further studies in this area are clearly warranted.

EXPERIMENTAL PROCEDURES

Retroviral Vector Preparation and In Ovo Retroviral Infection

The constitutively active avian EGFR receptor (caEGFR) used in this study (a.k.a. E1 vErbB/E1, see Boerner et al., 2003) possesses a complete intracellular domain, including the tyrosine kinase and transmembrane regions, but lacks all but 40 amino acids of the receptor ligand binding domain. Construction and characterization of the RCAS-caEGFR vector (Pelley et al., 1989) and virus propagation, titering, and concentration were performed as described (Ferrari et al., 1998). RCAS-caEGFR virus (approximately 10⁹ titer) was microinjected in ovo into the prospective wing- or leg-forming mesoderm or non–limb-forming flank mesoderm of the lateral plate of stage 11–14 chick embryos (University of Connecticut Agricultural Experiment Station). Alternately, pellets of virus-infected chick embryo fibroblasts were inserted microsurgically into a slit made between lateral plate ectoderm and mesoderm. To obtain limbs suitable for subsequent immunohistochemical detection of retroviral gag protein, pathogen-free chick embryos (SPAFAS) were also used.

Phenotypic and Molecular Analysis

Contralateral pairs of control uninfected and RCAS-caEGFR limbs were collected at day 10, examined for gross morphological defects, and stained whole-mount with 0.5% Alcian blue in 3% acetic acid, followed by clearing in methyl salicylate, to visualize the pattern of the limb cartilage elements. To examine the patterning of dorsal and ventral integument and mesoderm structures, some limbs were also collected at day 14–15, and their autopods were fixed in Bouin's fixative and embedded and sections were stained with hematoxylin and eosin. Whole-mount in situ hybridization was performed on contralateral pairs of 4% paraformaldehyde-fixed stage 22–26 limbs as described by Omi et al. (2000), except using ethanol instead of methanol for dehydration, with digoxigenin-labeled chicken Shh (Riddle et al., 1993) and FGF8 (Crossley et al., 1996) RNA probes. Some whole-mount samples were subsequently embedded in paraffin and sectioned to confirm expression patterns (data not shown). For radioactive in situ hybridization (Ferrari et al., 1998), contralateral pairs of left uninfected and right infected limbs were fixed in Bouin's fixative or 4% paraformaldehyde and embedded together in paraffin; sections were hybridized with ³³P-labeled chicken ErbB1 (EGFR; Lax et al., 1988), Bmp2 and Bmp4 (Francis et al., 1994), dHand (Srivastava et al., 1995), En1 (Logan et al., 1992), FGF4 (Niswander et al., 1994), Gli3 (Wang et al., 2000b), Gremlin (Hsu et al., 1998), Hoxd11 (Rogina et al., 1992), Hoxd13 (Rogina and Upholt, 1993), Lmx (Riddle et al., 1995), Msx2 (Coelho et al., 1991), Patched (Marigo et al., 1996), Shh (Riddle et al., 1993), or Wnt7a (Dealy et al., 1993) cDNA probes. Whenever possible, adjacent sections were hybridized with several different probes to enable direct comparison of expression patterns. For many samples, adjacent sections were also subjected to immunohistochemistry with a monoclonal antibody against the RSV viral gag protein p19 (Developmental Studies Hybridoma Bank) to facilitate correlation of retroviral infection with changes in gene expression. Retroviral gag staining was detected with a Vectastain ABC kit (Vector) using antigen retrieval (Dako). For analysis of cell death, day 7.5 limbs were stained whole with 1:10,000 Nile blue sulfate for 30 min at 37°C and rinsed in phosphate buffered saline. Immunohistochemical detection of EGF protein in normal day 7 chick limbs was performed as described (Dealy et al., 1998), using a polyclonal anti-mouse EGF antibody (Upstate No. 06102).

Acknowledgements

We thank Juan Carlos Izpisua Belmonte, Philippa Francis-West, Richard Harland, Alex Joyner, Gail Martin, Lee Niswander, Eric Olson, Cliff Tabin, William Upholt, and Bjorn Vennstrom for providing chicken cDNAs. We also thank John W. Saunders, Jr., for comments on the manuscript. C.N.D. and N.J.M. were funded by the NIH and M.F. was funded by the UCHC Skeletal, Craniofacial, and Oral Biology Program (NIDCR).