ABSTRACT
Heregulins bind directly to ErbB3 and ErbB4 receptors, leading to multiple dimerization possibilities including heterodimerization with the ErbB2 receptor. We have generated ErbB3-, ErbB2- and heregulin-deficient mice to assess their roles in development and differentiation. Heregulin-- and ErbB2−/− embryos died on E10.5 due to a lack of cardiac ventricular myocyte differentiation; ErbB3-- embryos survived until E13.5 exhibiting cardiac cushion abnormalities leading to blood reflux through defective valves. In ErbB3−/− embryos, the midbrain/hindbrain region was strikingly affected, with little differentiation of the cerebellar plate. Cranial ganglia defects, while present in all three nulls, were less severe in ErbB3−/− embryos. The cranial ganglia defects, along with a dramatic reduction in Schwann cells, enteric ganglia and adrenal chromaffin cells, suggests a generalized effect on the neural crest. Numerous organs, including the stomach and pancreas also exhibited anomalous development.
INTRODUCTION
The heregulins (HRG, neuregulins) are a family of structurally diverse glycoproteins sharing a common EGF-like domain, which are produced as a consequence of alternative splicing of a single gene (review, Lemke, 1996). Family members have been identified and cloned by a number of groups (Marchionni et al.,1993; Falls et al., 1993; Peles et al., 1992; Wen et al., 1992; Orr-Urtreger et al., 1993; Holmes et al., 1992; Ho et al., 1995). All HRG family members have a common EGF-like domain which is characterized by the position of six cysteine residues (Falls et al., 1993). The EGF-like domain is a critical part of the molecule, as it alone can stimulate tyrosine phosphorylation of ErbB2 and acetylcholine receptor synthesis in cultured myotubes (Holmes et al.,1992; Corfas et al., 1995). HRG is expressed transiently during development by several cell types, including migrating cranial neural crest cells and embryonic neurons (Meyer and Birchmeier, 1994; Moscoso et al., 1995), and HRG has been shown to affect the in vitro proliferation and fate choice of pluripotent neural crest cells, inducing them to adopt a glial rather than neuronal phenotype (Shah et al., 1994). HRG is also a survival factor for neural-crest-derived Schwann cells (Morrissey et al., 1991; Dong et al., 1995; Syroid et al., 1996).
HRG signals through a family of protein tyrosine kinases of the class 1 EGFR family, namely ErbB2 (Shih et al., 1981; Schechter et al., 1984), ErbB3 (Kraus et al., 1989; Kraus et al., 1993; Guy et al., 1994; Pinkas-Kramarski et al., 1996) and ErbB4 (Plowman et al., 1990; Plowman et al., 1993a,b; Culouscou et al., 1993). ErbB2, however, fails to bind HRG and depends upon HRG binding to ErbB3 or ErbB4; the heterodimers ErbB2/ErbB3, ErbB2/ErbB4 and the homodimer ErbB4/ErbB4 are the active HRG signaling complexes (Carraway et al., 1994; Carraway and Cantley, 1994; Tzahar et al., 1994). While ErbB3 on its own lacks any biological activity, transactivation with ErbB2 generates a potent signal (Sliwkowski et al., 1994). Both ErbB2 and ErbB3 also heterodimerize with the EGFR, being transactivated by EGFR ligands and potentiating the EGF signal (Earp et al., 1995; Graus-Porta et al., 1995; Karunagaran et al., 1995).
In addition to this diversity in signaling potential based on ligand isoforms and dimer composition, restricted expression of cellular receptors during development and throughout adulthood results in the unique ability of a single ligand, heregulin, to generate a highly complex, regulated pattern of signals (Altiok et al, 1995; Marikovsky et al., 1995). ErbB receptors are expressed during embryonic development, are involved in cell fate determination, and, through amplification and overexpression, are implicated in a variety of carcinomas in adulthood (e.g. breast, ovarian, pancreatic, gastric) (Slamon et al., 1987; Slamon et al., 1989; Lemoine et al., 1992; Hynes and Stern, 1994; Tahara, 1995). This evidence suggests a critical role for HRG and its ErbB receptors in the fundamental mechanisms involved in cell growth, differentiation and control of cell division.
To assess their roles in development and differentiation, and to examine interactions between ErbB family members, we have generated mice lacking either the ErbB3 or ErbB2 receptor or the entire heregulin ligand family.
MATERIALS AND METHODS
Cloning and targeting vector construction
100-mer overlapping oligos to the 5′ and 3′ ends of rat ErbB3, ErbB2 and HRG cDNA sequences were used to probe a mouse 129 library. For HRG targeting, a 2 kb deletion 5′ to the alternative splice region of HRG, containing the coding region for amino acids 172–211 and including the first 4 cysteines in the conserved EGF-like domain, was made (Fig. 1A); for ErbB2 targeting, a 4 kb SmaI-SmaI deletion was made containing the coding region for amino acids 256 to 304 in the first cysteine-rich domain (Fig. 1B) and, for ErbB3 targeting, a 2 kb BstEII-BstEII deletion was made containing the coding region for amino acids 73–106 (Fig. 1C). Deletions in receptor targeting vectors were designed to remove or precede any putative ligand interaction site; for HRG, the biologically active EGF-like domain, common to all isoforms, was deleted to allow no active alternate transcript to be made. A neomycin-resistance gene under control of the Pgk promoter replaced the deleted regions. For the HRG targeting vector, a thymidine kinase gene under control of the HSV promoter was added to the 5’ end.
Gene disruption and expression analysis. (A) Restriction map encompassing the coding region of the HRG EGF-like domain located 5′ to the alternative splice region. (1) Deleted amino acids 172-211 encode the first four cysteines of the conserved EGF-like domain, common to all isoforms. The 3′ black line is the area removed from the targeting vector that includes the probe. (2) The neomycin (neo) targeting vector that includes a 5′ TK gene for secondary selection. (3) The disrupted HRG allele identifying the 8 kb BamH1 fragment generated on Southern blots. (B) Restriction map of ErbB2. (1) Amino acids 256-304 are deleted in the ErbB2 targeting vector. The 3′ black line is the area removed from the targeting vector that includes the probe. (2) The neomycin (neo) targeting vector. (3) The disrupted ErbB2 allele identifying the 10 kb EcoRI fragment generated on Southern blots. (C) Restriction map of ErbB3. (1) The region deleted in the ErbB3 encodes amino acids 73 to 106. The 3′ black line is the area removed from the targeting vector that includes the probe. (2) The neomycin (neo) targeting vector. (3) The disrupted ErbB3 allele identifying the 5.5 kb XhoI fragment generated on Southern blots. (D) Southern blots of: HRG mouse tail DNA digested with BamHI; the 3′ external probe detects a single >23 kb wild-type band in lane 1 indicating a wild-type genotype and both the wild-type band and the 8 kb disrupted HRG allele corresponding to a heterozygous genotype in lanes 2 and 3; of ErbB2 targeted ES cells with EcoRI digestion; the 3′ external probe detects a > 23 kb wild-type band and an approximately 10 kb disrupted allele. Lane 2 (arrow) depicts a single-targeted ES cell clone; the other lanes are non-targeted wild-type clones; and of ErbB3 targeted ES cells withXhoI digestion; the 3′ external probe detects an approximately 23 kb wild-type band and a 5.5 kb disrupted allele (arrow). (E) Western blot of 30 pg/lane E10.5 whole embryo lysates stained with anti-wew C18. A p185 ErbB2 wild-type band was present in the wild-type (+/+) and heterozygote (+/−) lanes. A reduced amount of protein was evident in the het (+/−) lane and no ErbB2 protein was detected in the ErbB2−/− lane. ErbB3 western blot of 48 p g/lane of whole E13.5 embryo lysate stained with anti-ErbB3 C17. A wild-type band is present in both the wild-type (+/+) and heterozygote (+/−) lanes. As with ErbB2, there was a noticeable reduction in ErbB3 in ErbB3+/−. No ErbB3 protein was detectable in the ErbB3−/− lane. (F) In situ hybridization of sagittal sections from E9.5 wild-type (+/+) and homozygous null (−/−) embryos screened with a HRG EGF-like domain antisense probe illustrating the presence of HRG in the wild-type embryo neuroepithelium (left, +/+) and the complete absence of HRG in the corresponding neuroepithelium of the homozygous null embryo (right, −/−). Sense probe (not shown) was negative.
Gene disruption and expression analysis. (A) Restriction map encompassing the coding region of the HRG EGF-like domain located 5′ to the alternative splice region. (1) Deleted amino acids 172-211 encode the first four cysteines of the conserved EGF-like domain, common to all isoforms. The 3′ black line is the area removed from the targeting vector that includes the probe. (2) The neomycin (neo) targeting vector that includes a 5′ TK gene for secondary selection. (3) The disrupted HRG allele identifying the 8 kb BamH1 fragment generated on Southern blots. (B) Restriction map of ErbB2. (1) Amino acids 256-304 are deleted in the ErbB2 targeting vector. The 3′ black line is the area removed from the targeting vector that includes the probe. (2) The neomycin (neo) targeting vector. (3) The disrupted ErbB2 allele identifying the 10 kb EcoRI fragment generated on Southern blots. (C) Restriction map of ErbB3. (1) The region deleted in the ErbB3 encodes amino acids 73 to 106. The 3′ black line is the area removed from the targeting vector that includes the probe. (2) The neomycin (neo) targeting vector. (3) The disrupted ErbB3 allele identifying the 5.5 kb XhoI fragment generated on Southern blots. (D) Southern blots of: HRG mouse tail DNA digested with BamHI; the 3′ external probe detects a single >23 kb wild-type band in lane 1 indicating a wild-type genotype and both the wild-type band and the 8 kb disrupted HRG allele corresponding to a heterozygous genotype in lanes 2 and 3; of ErbB2 targeted ES cells with EcoRI digestion; the 3′ external probe detects a > 23 kb wild-type band and an approximately 10 kb disrupted allele. Lane 2 (arrow) depicts a single-targeted ES cell clone; the other lanes are non-targeted wild-type clones; and of ErbB3 targeted ES cells withXhoI digestion; the 3′ external probe detects an approximately 23 kb wild-type band and a 5.5 kb disrupted allele (arrow). (E) Western blot of 30 pg/lane E10.5 whole embryo lysates stained with anti-wew C18. A p185 ErbB2 wild-type band was present in the wild-type (+/+) and heterozygote (+/−) lanes. A reduced amount of protein was evident in the het (+/−) lane and no ErbB2 protein was detected in the ErbB2−/− lane. ErbB3 western blot of 48 p g/lane of whole E13.5 embryo lysate stained with anti-ErbB3 C17. A wild-type band is present in both the wild-type (+/+) and heterozygote (+/−) lanes. As with ErbB2, there was a noticeable reduction in ErbB3 in ErbB3+/−. No ErbB3 protein was detectable in the ErbB3−/− lane. (F) In situ hybridization of sagittal sections from E9.5 wild-type (+/+) and homozygous null (−/−) embryos screened with a HRG EGF-like domain antisense probe illustrating the presence of HRG in the wild-type embryo neuroepithelium (left, +/+) and the complete absence of HRG in the corresponding neuroepithelium of the homozygous null embryo (right, −/−). Sense probe (not shown) was negative.
Generation of HRG, ErbB2 and ErbB3 targeted mice and embryos
Targeting vectors were electroporated into ES D3-C12 cells and cells were selected with G418 at 300 g μml for 10 days; for HRG, FIAU at 0.2 mM was also added. Southern blot analysis using probes designed from sequences 3’ to those present on the targeting vectors was performed to identify targeted clones (Fig. 1D, ErbB2, ErbB3). HRG and ErbB2 chimeric mice were mated with either C57BL/6J or Balb/c females; ErbB3 were bred only on C57BL/6J. Offspring were screened by PCR for the neomycin and targeted gene and confirmed by Southern blot analysis of tail DNA (Fig. 1D, HRG) (Thomas and Capecchi, 1987; Mansour et al., 1988; McMahon and Bradley, 1990). In all three cases, no full-term null mice were born; embryos were recovered from timed-pregnant matings and genotyped by PCR (Lu et al., 1996). For histology, embryos were fixed in 10% neutral buffered formalin for 2 hours (E8.5 to E11.5) or overnight for older embryos.
Histology and immunohistochemistry
Serial 6 μm paraffin sections, affixed to poly-lysine-coated slides were used for histological analysis (hematoxylin/eosin or cresyl violet) or immunohistochemistry. Sections for immunohistochemistry were stained using Vector Lab’s Vectastain ABC-AP Kits (alkaline phosphatase, mouse IgG; AK-5002 for mouse monoclonals and rabbit IgG; AK-5001 for rabbit polyclonals) with minor modifications. Vector Lab’s avidin-biotin blocking kit (SP-2001) and alkaline phosphatase substrate kit II (SK-5200) were used as per instructions. Primary antibodies were anti-ErbB2/v (C-18) rabbit polyclonal IgG (sc-284, Santa Cruz Biotechnology; 1:50), anti-ErbB3 (C-17) rabbit polyclonal IgG (sc-285, S.C.B.; 1:20), anti-S100a rabbit polyclonal IgG (Z0628, Dako, 1:200), anti-tyrosine hydroxylase rabbit polyclonal IgG (AB152, Chemicon, 1:50) and anti-peripherin rabbit polyclonal IgG (AB1530, Chemicon, 1:300). Controls were negative.
Western blot analysis
Equal amounts of protein (E10.5 or E13.5 whole embryo lysates) were loaded in each lane of 8–16% Novex gels (EC60452) and transferred to Novex nitrocellulose (LC2001). Staining and detection were performed using Amersham’s ECL method following instructions provided, using antibodies listed above.
Whole-mount immunohistochemistry
Embryos were fixed for 2 hours in DMSO:MeOH (1:4) followed by bleaching for 48 hours in 20% DMSO:10% hydrogen peroxide:70% MeOH followed by incubation at 4°C overnight in either TUJ1 antibody (Lee et al., 1990) (1:1000) or aEnhb-1 antisera (Davis et al., 1991) (1:500). Embryos were then rinsed and incubated overnight at 4°C in 1:300 goat anti-mouse IgG (H+L) horseradish peroxidase conjugate (172-1011; Biorad) for TUJ1 or 1:300 goat anti-rabbit IgG (H+L) horseradish peroxidase conjugate (172-1013; Biorad) for aEnhb-1. After rinsing, embryos were incubated 1.5 hours with 0.25 mg/ml DAB (170-6535; Biorad) in TBS: 0.02% hydrogen peroxide, rinsed in TBS then MeOH, followed by final clearing and storage in benzyl alcohol:benzyl benzoate (1:2) (Dent et al., 1989).
In situ hybridization
PCR amplification was used to generate a 120 bp murine heregulin EGF-like domain; simultaneously T7 and T3 RNA polymerase promoter sequences were attached to flank the gene fragment. For heregulin in situ, formalin-fixed mouse E9.5 embryos were cryosectioned; for Wnt-1 in situ, formalin-fixed E13.5 mouse embryo paraffin sections were used; generation of g-33P-UTP-labeled antisense and sense RNA hybridization probes and in situ hybridization were performed as previously described (Lu and Gillett, 1994).
RESULTS
Gene targeting
A null HRG allele was created at the HRG gene locus by homologous recombination (Fig. 1A) deleting amino acids 172-211 (TSTSTTGTSHLIKCAEKEKTKCVNGGECFMV-KDLSNPSRYLCK) which resulted in the deletion of the first four of the six cysteines (bold type) in the conserved EGF-like domain, thereby deleting the active portion of the molecule common to all isoforms. Therefore, this deletion resulted in the disruption of all alternatively spliced isoforms of heregulin (Holmes et al., 1992). A null mutation was created at the ErbB2 gene locus by homologous recombination (Fig. 1B), resulting in the deletion of amino acids 256-304 (CQACLHFNHSG-ICELHCPALITYNTDTFESMHNPEGRYTFGASCVTTCP) in the first extracellular cysteine-rich domain (bold type) and in the removal of the putative ligand interaction site as defined by antagonistic antibody epitope mapping studies (data not shown) (Coussens et al., 1985). Similarly, at the ErbB3 gene locus, a null mutation was created by homologous recombination (Fig. 1C) resulting in the deletion of amino acids 73 through 107 (AVCPGTLNGLSVTGDADNQYQTLYK-LYEKCEVVMG), quite early in the protein and presumably well before a putative ligand interaction site (Hellyer et al., 1995). Gene targeting was carried out in 129/Sv-derived D3 embryonic stem cells (Gossler et al., 1986; Thomas and Capecchi, 1987; Mansour et al., 1988; McMahon and Bradley, 1990). Targeted embryonic stem cell lines were analyzed by Southern blot (Fig. 1E,F), injected into blastocysts and germline chimeric males for HRG and ErbB2 were mated to either C57BL/6J or BALB/c females; ErbB3 mutants were mated only to C57BL/6J females. Animals bearing the respective targeted allele were intercrossed, and the F2 and F3 offspring analyzed by Southern blot (Fig. 1D) and PCR. No full-term ErbB2−/−, ErbB3−/− or HRG−/− animals were born. Therefore, timed pregnancies were carried out and Cesarean sections were performed at E8.5 to E16.5. Viable HRG−/− and ErbB2−/− embryos were present up to and including E10.5. ErbB3−/− embryos, however, were viable up to and including E13.5. HRG embryos were analyzed from backcross litters 1–4 on C57BL/6J and backcross litters 1-5 on Balb/c; ErbB2 embryos were analyzed from backcross litters 1–5 on C57BL/6J and 1–5 on Balb/c; ErbB3 embryos were analyzed on C57BL/6J backcross litters 1–3. No phenotypic differences were noted between the two mouse strains. At least 20 litters were examined for each targeted gene; wild-type controls were all littermates. Detailed analysis was performed on a minimum of six embryos of each genotype (wild-type, heterozygous and homozygous null) at each developmental stage for each targeted gene.
Genotyping and expression analysis
Receptor protein expression in ErbB2 and ErbB3 embryos was examined by western blot using Santa Cruz Biotechnology antibodies: neu (C18) for ErbB2 (epitope corresponding to amino acids 1169-1186, mapping at the carboxy terminus of the precursor form of human ErbB2) and ErbB3 (C17) (epitope corresponding to amino acids 1307-1323, mapping at the carboxy terminus of the precursor form of human ErbB3). Commercially available antibodies for these receptors all have similar carboxy terminus epitopes as this is the region of greatest divergence among ErbB receptors. There was a total absence of ErbB2 protein in the ErbB2−/− embryos and of ErbB3 protein in the ErbB3−/− embryos (Fig. 1E). Also, there was a consistent reduction in levels of ErbB3 protein in ErbB3+/− and a similar reduction of ErbB2 protein in ErbB2+/−.
In situ hybridization was performed to demonstrate the absence of all HRG isoforms in HRG−/− embryos using an antisense probe to the EGF-like domain of the heregulin gene family (Fig. 1F). As this is the only region common to all of the HRG isoforms as well as the active portion of the molecule (Holmes et al., 1992), by necessity, this region was used to probe for the absence of all HRG isoforms. The presence of HRG message was clearly detectable in the neural tube of the E9.5 wild-type embryo and completely absent from the entire E9.5 HRG−/− embryo indicating the absence of the active portion of the heregulin gene family in these embryos.
Cardiac cushion development regulated by ErbB3
Given the high expression of HRG in the endocardium (Meyer and Birchmeier, 1994), the expression of ErbB2 and ErbB4 in developing cardiac myocytes and the presence of ErbB3 transcripts in endocardial cushion mesenchyme (Meyer and Birch-meier, 1995), it is not surprising that gene deletion resulted in significant (but variable) anomalies in the development of this region and in early embryonic lethality. Although cardiac myocytes were present in both HRG−/− (Fig. 2C,D) and ErbB2−/− (Fig. 2E,F) E9.5 embryos, the ventricular wall was thinned and trabeculae were absent from the myocardial layer, leading to an enlarged heart with an irregular heartbeat and poor circulation, presumably accounting for embryonic death at E10.5 (Lee et al., 1995; Meyer and Birchmeier, 1995; Erickson et al., 1996). In contrast, unlike E9.5 wild-type embryos in which cardiac cushions were opposed and contained a thick core of mesenchyme and myocytes (Fig. 2A), ErbB3−/− cushions lacked mesenchyme and were noticeably thinned (Fig. 2G). With continued development, forming valves in ErbB3−/− embryos remained hypoplastic with little additional development by E13.5 (Fig. 3C,D). Myocyte tra-beculation, however, was only marginally affected in ErbB3−/− embryos (Fig 2G,H), although subsequent differentiation and thickening of the myocardium at E13.5 (Fig. 3C) was somewhat reduced compared to wild-type embryos (Fig. 3A). Histologically, ErbB3−/− cardiac valves appeared underdeveloped and unable to support proper cardiac function and, using motion-picture photography of living E13.5 embryos, blood reflux through rudimentary valves was apparent in the ErbB3-/- embryos, likely leading to death at E13.5.
Histological analysis of the developing heart in wild-type, HRG−/−, ErbB2−/− and ErbB3−/− embryos on E9.5. Sagittal section through the cardiac region of a 9.5 day wild-type embryo (A) illustrating the thick core of mesenchyme and myocytes forming the endocardial cushions (EC); V, ventricle, A, atrium. (B) The thick, trabeculated myocardium (M) of the E9.5 wild-type ventricular wall, lined with endocardium (E) and surrounded by the pericardium (P). (C,D) A parasagittal section through the cardiac region of an E9.5 HRG−/− embryo. Note the reduced thickness of the ventricular wall in the null mutants and the significant fluid accumulation beneath the pericardium (P) (BC, bulbus cordis; AS, aortic sac). (E,F) A parasagittal section through the cardiac region of an ErbB2−/− embryo also isolated on day 9.5 of gestation, illustrating its abnormal development. Myoblasts were present in both HRG−/− and ErbB2−/− embryos, but failed to develop a well-organized trabeculated meshwork. Note the lack of development of both endocardial (E) and myocardial (M) layers. (G,H) The ventricular wall of ErbB3−/− E9.5 embryos contains differentiating myocytes but the endocardial cushions (EC) lack mesenchyme and have a thinned appearance compared with the wild-type embryos; TA, truncus arteriosus; BC, bulbus cordis; AS, aortic sac; E, endocardium; M, myocardium; P, pericardium. Scale bar in A,C,G, 153 μm; in B,F, 40 μm; in D, 33 μm; in E,125 μm; in H, 62 μm.
Histological analysis of the developing heart in wild-type, HRG−/−, ErbB2−/− and ErbB3−/− embryos on E9.5. Sagittal section through the cardiac region of a 9.5 day wild-type embryo (A) illustrating the thick core of mesenchyme and myocytes forming the endocardial cushions (EC); V, ventricle, A, atrium. (B) The thick, trabeculated myocardium (M) of the E9.5 wild-type ventricular wall, lined with endocardium (E) and surrounded by the pericardium (P). (C,D) A parasagittal section through the cardiac region of an E9.5 HRG−/− embryo. Note the reduced thickness of the ventricular wall in the null mutants and the significant fluid accumulation beneath the pericardium (P) (BC, bulbus cordis; AS, aortic sac). (E,F) A parasagittal section through the cardiac region of an ErbB2−/− embryo also isolated on day 9.5 of gestation, illustrating its abnormal development. Myoblasts were present in both HRG−/− and ErbB2−/− embryos, but failed to develop a well-organized trabeculated meshwork. Note the lack of development of both endocardial (E) and myocardial (M) layers. (G,H) The ventricular wall of ErbB3−/− E9.5 embryos contains differentiating myocytes but the endocardial cushions (EC) lack mesenchyme and have a thinned appearance compared with the wild-type embryos; TA, truncus arteriosus; BC, bulbus cordis; AS, aortic sac; E, endocardium; M, myocardium; P, pericardium. Scale bar in A,C,G, 153 μm; in B,F, 40 μm; in D, 33 μm; in E,125 μm; in H, 62 μm.
ErbB3 is required for normal cerebellar development
Until now, little has been known of the role of ErbB receptors in early brain development, although HRG is known to be expressed in the fetal brain. The deletion of the ErbB3 gene results in severe anomalies of brain development, especially in the midbrain/hindbrain region including the cerebellum. The E9.5 wild-type cerebellar plate is pseudostratified neuroepithelium located just ventral to the thinned roofplate of the fourth ventricle and, unlike other regions of the forming brain, its dorsal surface is surrounded by significant amounts of mesenchymal tissue. With additional development, the neuroepithelium thickens and stratifies, giving rise to a dorsal, differentiating field (vermis), a ventral neuroepithelium and a region of migrating deep nuclei. Beginning on day 12.5, a stream of migrating (external) granule cells leaves the rhombic lip to populate the external (pial) surface of the cerebellar cortex. By day 13.5, there is a thick, well-developed vermis and the beginning of a medial deep nucleus in wild-type embryos (Fig. 4B) (Altman and Bayer, 1985).
Histological analysis of the developing heart in wild-type and ErbB3−/− embryos on E13.5. By E13.5 atrioventricular valves (circled) remain poorly developed in the ErbB3−/− embryo (C,D) compared to wild type (A,B). The lack of adequate connective tissue in the atrioventricular valves of the ErbB3−/− embryos allows blood reflux and presumably death at E13.5. Scale bar in A,C, 250 μm; in B,D, 100 μm.
Histological analysis of the developing heart in wild-type and ErbB3−/− embryos on E13.5. By E13.5 atrioventricular valves (circled) remain poorly developed in the ErbB3−/− embryo (C,D) compared to wild type (A,B). The lack of adequate connective tissue in the atrioventricular valves of the ErbB3−/− embryos allows blood reflux and presumably death at E13.5. Scale bar in A,C, 250 μm; in B,D, 100 μm.
Cranial region of (A-F) wild-type and (G-L) ErbB3 null embryos on E13.5. (A) Low magnification view of the wild-type telencephalic region illustrating the normal topography of the area. CTX, cortex; OB, olfactory bulb; CP, choroid plexus; HIP, hippocampus (B) Developing cerebellum illustrating its normal stratified appearance. A wave of granule cells (arrow) is migrating from the rhombic lip. The neuroepithelial layer is continuous along the ventricular lumen and deep nuclei are developing at the vermis. The choroid plexus (CP) consists of an outer epithelial layer with a core of mesenchyme. (C) Tectal neuroepithelium illustrating its normal differentiation and stratification. (D) Section of the trigeminal ganglion illustrating the normal, well organized posterior region (semilunar ganglion), anterior projections (right arrow) and projection from the hindbrain. (left arrow, H, hindbrain) (E,F) Sections through the ventral tegmentum illustrating its normal mesenchyme-filled base (M) and dopaminergic neurons migrating to their adult locations (arrows). (G) In the telencephalon of ErbB3−/− embryos, the ventricle has not expanded normally with resultant infolding of the subicular neuroepithelium. The pallium is malformed and the choroid plexus of the lateral ventricle is abnormally elongated, and its epithelium is clubbed. The hippocampal and thalamic regions are less severely affected. (H) In the cerebellum, the neuroepithelial plate is not thickened in the most medial extent, there is very little granule cell migration (lower arrow), and no indication of a forming vermal region. Note the abnormal infolding of the tectum and isthmic neuroepithelium (upper arrows) and the abnormal appearance of the choroid plexus (CP) and subjacent mesenchyme. (I) In ErbB3−/− embryos, there is little differentiation or stratification of the tectal neuroepithelium, retaining a primitive, condensed appearance. (J) High magnification view of a portion of the trigeminal ganglion (V) illustrating its small size and rudimentary axonal projections from the hindbrain (large arrow, H, hindbrain; small arrows, anterior projections of the trigeminal ganglion). (K,L) Sections through the midbrain in the region of the ventral tegmentum illustrating the abnormal topography and organization of the mesenchyme and significantly delayed migration of dopaminergic neurons (arrow) compared with wild-type embryos. Scale bar in A, 357 μm; in B,D,E,K,I, 143 μm; in C,J, 80 μm; in F,H,L, 200 μm; in G, 313 μm.
Cranial region of (A-F) wild-type and (G-L) ErbB3 null embryos on E13.5. (A) Low magnification view of the wild-type telencephalic region illustrating the normal topography of the area. CTX, cortex; OB, olfactory bulb; CP, choroid plexus; HIP, hippocampus (B) Developing cerebellum illustrating its normal stratified appearance. A wave of granule cells (arrow) is migrating from the rhombic lip. The neuroepithelial layer is continuous along the ventricular lumen and deep nuclei are developing at the vermis. The choroid plexus (CP) consists of an outer epithelial layer with a core of mesenchyme. (C) Tectal neuroepithelium illustrating its normal differentiation and stratification. (D) Section of the trigeminal ganglion illustrating the normal, well organized posterior region (semilunar ganglion), anterior projections (right arrow) and projection from the hindbrain. (left arrow, H, hindbrain) (E,F) Sections through the ventral tegmentum illustrating its normal mesenchyme-filled base (M) and dopaminergic neurons migrating to their adult locations (arrows). (G) In the telencephalon of ErbB3−/− embryos, the ventricle has not expanded normally with resultant infolding of the subicular neuroepithelium. The pallium is malformed and the choroid plexus of the lateral ventricle is abnormally elongated, and its epithelium is clubbed. The hippocampal and thalamic regions are less severely affected. (H) In the cerebellum, the neuroepithelial plate is not thickened in the most medial extent, there is very little granule cell migration (lower arrow), and no indication of a forming vermal region. Note the abnormal infolding of the tectum and isthmic neuroepithelium (upper arrows) and the abnormal appearance of the choroid plexus (CP) and subjacent mesenchyme. (I) In ErbB3−/− embryos, there is little differentiation or stratification of the tectal neuroepithelium, retaining a primitive, condensed appearance. (J) High magnification view of a portion of the trigeminal ganglion (V) illustrating its small size and rudimentary axonal projections from the hindbrain (large arrow, H, hindbrain; small arrows, anterior projections of the trigeminal ganglion). (K,L) Sections through the midbrain in the region of the ventral tegmentum illustrating the abnormal topography and organization of the mesenchyme and significantly delayed migration of dopaminergic neurons (arrow) compared with wild-type embryos. Scale bar in A, 357 μm; in B,D,E,K,I, 143 μm; in C,J, 80 μm; in F,H,L, 200 μm; in G, 313 μm.
In ErbB3−/− embryos, the cerebellar region was quite abnormal, often appearing hypoplastic and, by E13.5, there was very little differentiation of the cerebellar plate, although there was some stratification within it. The neuroepithelial layer lining the fourth ventricle remained rudimentary and the entire cerebellar plate was misshapen (Fig. 4H). In some ErbB3−/− embryos, streams of migrating granule cells were present, although there was no evidence of a vermal region. Anti-TUJ1 immunohistochemistry indicated a significant reduction in axonal development in the cerebellum compared to the controls. Purkinje cells were present but reduced in number in the ErbB3−/− cerebellum. There was a significant lack of differentiated mesenchyme dorsal to the cerebellum and the pontine flexure was abnormal in these embryos. In addition, the choroid plexus of the IV ventricle was malformed; the normal fingerlike projections were absent and the base projection was much longer than in wild-type embryos (Fig. 4B,H).
engrailed and wingless expression are altered in the cerebellar region
Given the severe defects in midbrain/hindbrain junction derivatives, we examined the expression patterns of two genes known to be developmentally expressed in this region, the mouse orthologues of the Drosophila genes engrailed (Davis and Joyner, 1988) and wingless, (Wnt-1, Wilkinson et al., 1987; Bally-Cuif et al., 1992). engrailed expression in wild-type embryos was present in (not restricted to) the isthmic and cerebellar neuroepithelium on E10.5 and E11.5 (Fig. 5A,F). In all three null embryos, engrailed expression was greatly reduced in the cerebellar neuroepithelium; in E10.5 HRG−/− (Fig. 5B) and E10.5 and E11.5 ErbB3−/− embryos (Fig. 5D,H), engrailed expression was similarly reduced to a dorsal strip of isthmic neuroepithelium, with no expression in the cerebellum. In the E10.5 ErbB2−/− embryos, there was very slight diffuse expression in both the isthmic and cerebellar neuroepithelia (Fig. 5C). Engrailed, like Wnt-1, is required for the normal histogenesis of the cerebellar cortex, and the reduction in engrailed-expressing cells in all three nulls may indicate either the deletion or lack of proliferation of cerebellar cell precursors (McMahon and Bradley, 1990; Wurst et al., 1994; Grove et al., 1996). Little cell death, however, was apparent in the CNS neuroepithelium/tectum.
engrailed expression in the midbrain/hindbrain region of E10.5-E11.5 embryos. Side views of E10.5 (A) wild-type, (B) HRG−/−, (C) ErbB2−/− and (D) ErbB3−/− embryos stained in whole mount with an antibody to the engrailed gene product (aEnhb-1) recognizing both En-1 and En-2. engrailed expression in E10.5 wild-type embryos was present in (not restricted to) the isthmic and cerebellar neuroepithelium (A, arrow). In HRG--(B) and ErbB3−/− embryos (D), engrailed expression was reduced to a dorsal strip of isthmic neuroepithelium, with no expression in the cerebellum (arrows). In ErbB2−/− embryos (C), there was very slight expression in both the isthmic and cerebellar neuroepithelia (arrow). Side views of E11.5 wild-type (E) and ErbB3−/− (G) embryos at autopsy illustrate the normal external appearance of these ErbB3−/− mutants at this stage, and even by E13.5, they still externally appear quite normal. In E11.5 wild-type embryos, engrailed was expressed in the hindbrain in two stripes corresponding to the isthmic and cerebellar (arrow) neuroepithelium (F); M, midbrain. In contrast, in E11.5 ErbB3−/− embryos (H), engrailed expression was reduced to a dorsal strip of isthmic neuroepithelium with no expression in cerebellum. These views also clearly illustrate the abnormal development of the hindbrain/midbrain region, the abnormal pontine flexure and abnormal development of the ventricular system in the null embryos. Scale bar in A–D,F,H, 1 mm; in E,G, 2 mm.
engrailed expression in the midbrain/hindbrain region of E10.5-E11.5 embryos. Side views of E10.5 (A) wild-type, (B) HRG−/−, (C) ErbB2−/− and (D) ErbB3−/− embryos stained in whole mount with an antibody to the engrailed gene product (aEnhb-1) recognizing both En-1 and En-2. engrailed expression in E10.5 wild-type embryos was present in (not restricted to) the isthmic and cerebellar neuroepithelium (A, arrow). In HRG--(B) and ErbB3−/− embryos (D), engrailed expression was reduced to a dorsal strip of isthmic neuroepithelium, with no expression in the cerebellum (arrows). In ErbB2−/− embryos (C), there was very slight expression in both the isthmic and cerebellar neuroepithelia (arrow). Side views of E11.5 wild-type (E) and ErbB3−/− (G) embryos at autopsy illustrate the normal external appearance of these ErbB3−/− mutants at this stage, and even by E13.5, they still externally appear quite normal. In E11.5 wild-type embryos, engrailed was expressed in the hindbrain in two stripes corresponding to the isthmic and cerebellar (arrow) neuroepithelium (F); M, midbrain. In contrast, in E11.5 ErbB3−/− embryos (H), engrailed expression was reduced to a dorsal strip of isthmic neuroepithelium with no expression in cerebellum. These views also clearly illustrate the abnormal development of the hindbrain/midbrain region, the abnormal pontine flexure and abnormal development of the ventricular system in the null embryos. Scale bar in A–D,F,H, 1 mm; in E,G, 2 mm.
In Drosophila, Wingless protein is secreted and internalized by nearby cells and is thought to maintain the transcription of engrailed (Danielian and McMahon, 1996). Wnt-1 expression was also altered in the E13.5 ErbB3−/− embryos (Fig. 6D-F). In wild-type E13.5 embryos, Wnt-1 was expressed in a discreet patch in the anterior cerebellar neuroepithelium at the rhombic lip, and again at the caudal transition of choroid epithelium and neuroepithelium (Fig. 6A-C). In the ErbB3 mutants, cerebellar expression of Wnt-1 appeared normal in both the anterior position and at the transition between cerebellar neuroepithelium and choroid epithelium. It is between these two areas of Wnt-1 expression that the neuroepithelium of the ErbB3-/- cerebellum does not differentiate and where engrailed expression is lost. However, ectopic expression of the boundary molecule Wnt-1 in the transition zone between the choroid epithelium and neuroepithelium of the spinal cord is dramatically extended appearing as a long stripe lining the choroid epithelium (Fig. 6D–F).
In situ hybridization localization of Wnt-1 gene expression in the midbrain/hindbrain region of E13.5 wild-type and ErbB3−/− embryos. In sagittal sections of wild-type E13.5 embryos labeled by in situ hybridization with a probe for Wnt-1, expression was present in a discreet patch in the anterior cerebellar neuroepithelium at the rhombic lip, and again at the caudal transition of choroid epithelium and neuroepithelium (A–C, arrows). In sagittal sections of the E13.5 ErbB3−/− embryos, cerebellar expression of Wnt-1 appeared normal in both the anterior position and at the transition between cerebellar neuroepithelium and choroid epithelium (D,E). However, expression of Wnt-1 in the transition zone between the choroid epithelium and neuroepithelium of the spinal cord was dramatically extended in null embryos, appearing as a long, ectopic stripe lining the choroid epithelium (D,F). Scale bar in A,D, 2.6 mm; in B,C,E,F, 100 μm.
In situ hybridization localization of Wnt-1 gene expression in the midbrain/hindbrain region of E13.5 wild-type and ErbB3−/− embryos. In sagittal sections of wild-type E13.5 embryos labeled by in situ hybridization with a probe for Wnt-1, expression was present in a discreet patch in the anterior cerebellar neuroepithelium at the rhombic lip, and again at the caudal transition of choroid epithelium and neuroepithelium (A–C, arrows). In sagittal sections of the E13.5 ErbB3−/− embryos, cerebellar expression of Wnt-1 appeared normal in both the anterior position and at the transition between cerebellar neuroepithelium and choroid epithelium (D,E). However, expression of Wnt-1 in the transition zone between the choroid epithelium and neuroepithelium of the spinal cord was dramatically extended in null embryos, appearing as a long, ectopic stripe lining the choroid epithelium (D,F). Scale bar in A,D, 2.6 mm; in B,C,E,F, 100 μm.
Rostrally, the tectal neuroepithelium failed to differentiate and stratify and was occasionally pyknotic, as was the neuroepithelium in the isthmus region. The aqueduct lining these regions was also malformed and occasionally the tectal neu-roepithelium/future inferior colliculus was infolded near the cerebellum (Fig. 4H). Basically, the ventricular system of the brain failed to enlarge, likely resulting in/from the lack of development of the pontine flexure and a lack of normal production of cerebrospinal fluid by defective choroid plexes. Cerebrospinal fluid, in combination with neural lumen occlusion, is thought to play a role in flexure development and enlargement of the ventricular system (Fig. 4A,G) (e.g., Desmond and Schoenwolf, 1986).
While forebrain regions, such as the striatum and the olfactory bulb, appeared morphologically normal, other midbrain regions were affected by deletion of the ErbB3 gene. The tegmental neuroepithelium was developmentally delayed, and many dopaminergic neurons, which migrate from this location to their adult positions in the substantia nigra, the retrorubral field and the ventral tegmental area (Fig. 4F) (Choi et al., 1992), failed to do so in ErbB3−/− embryos, resulting in clumps of ectopic neurons in the mesencephalic tegmental neuroepithelium (Fig. 4L) at E13.5. Anti-TH immunohistochemistry carried out on day 12.5 detected few migrating neurons in this region compared to controls, consistent with the developmental delay in cell migration observed on day 13.5. As in the hindbrain region, the midbrain flexure had developed abnormally in ErbB3−/− embryos; midbrain mesenchyme separating hypothalamus and ventral tegmentum and pons (Fig. 4E) was thickened and maloriented in the ErbB3−/− embryos (Fig. 4K). At its base, the choroid plexus of the lateral ven-tricles was abnormal with a highly elongated base and clubbed projections, an appearance identical to the choroid plexus of IV The pallium was abnormally elongated into the wall of the ventricle (Fig. 4A,G). The neocortex was somewhat thinned; only the subicular area was abnormally positioned.
Cranial ganglia require ErbB3, ErbB2 and HRG
Cranial ganglia in these null embryos were carefully examined at E10.5 (Fig. 7) and, as reported previously, HRG−/− embryos (Fig. 7E–G) lacked the mandibular portion (Vmn) of the trigeminal nerve (nV) (Meyer and Birchmeier, 1995); the ophthalmic division (Vop), although robust, was unfasciculated. In addition, the vestibulo/cochlear ganglion (VIIIg) was malpos-itioned, as was the geniculate (VIIg), which had only a small, thread-like projection (nVII) emerging from it. The petrosal ganglion (IXg) was disorganized, with individual neurons scattered between it and the nodose. The nodose ganglion (Xg) was typically decreased in size, but vagal (nX) projections appeared normally oriented, although not as advanced toward their target organs as in wild-type controls (Fig. 7A–C). Posterior branches of the hypoglossal (nXII) were rudimentary and failed to coalesce with those from C1 and the vagus (nX).
Cranial and dorsal root ganglia/nerve formation in E10.5 embryos. Side views of day 10.5 embryos stained in whole mount with anti-TUJI antibodies, illustrating the organization of cranial ganglia and nerves, dorsal root ganglia and nerves in wild-type (A–D), HRG−/− (E-H), ErbB2−/− (I–L) and in ErbB3−/− (M-P) embryos. (A–C) Low to high magnification views of a control embryo, illustrating the normal organization of the cranial ganglia. (D) Cervical-thoracic dorsal root ganglia near the forelimb bud (LB) illustrating the normal constriction of the spinal nerve as it leaves the dorsal root ganglion to enter the periphery (arrows). Note the well-organized ganglia and fasciculated axonal projections. (E–G) Low to high magnification views of cranial ganglia in a HRG−/− embryo. The trigeminal ganglion has ophthalmic (op) and maxillary (mx) nerve projections, but not the mandibular (mn) component. The geniculate and vestibulo-cochlear ganglia (VII/VIII) appear fused, and the facial nerve (VIIn) emerging from the geniculate is small. The petrosal ganglion (IX) is also reduced in size, as is the nodose (X) ganglion. (H) The cervical dorsal root ganglia appear disorganized and projections emerging from them are unfasciculated (arrow) compared to those typical of wild-type embryos. (I–K) Low to high magnification views of cranial ganglia and nerve in an ErbB2−/− embryo. The trigeminal ganglion (V) has only the ophthalmic projection (Vop), the geniculate and vestibulo-cochlear ganglia (VII/VIII) are fused, and the facial nerve (VIIn) is quite small. Note the scatter of neurons between the trigeminal (V) and VII/VIII ganglia. The jugular and the nodose (X) ganglia are significantly reduced in size and vagal projections of the nodose extend too far ventrally. Although vagal projections fuse with those from C1, in combination with the lack of projections from XIIn, this bundle is very rudimentary. (L) Axonal projections emerging from dorsal root ganglia in the cervical region are quite unfasciculated, and the ganglia appear disorganized. (M–O) Low to high magnification views of the cranial ganglia and nerves in ErbB3−/− embryos. In general, these appear similar to wild-type controls, with notable exceptions. The trigeminal ganglion (V) is lacking the mandibular projection (mn), the vestibulo-cochlear and geniculate ganglia (VII/VIII) appear normal, the petrosal (IXg) was only slightly malpositioned, although it and the nodose (Xg) were reduced in size. Other projections including those from XIIg, and C1 were also unaffected. (P) ErbB3−/− dorsal root ganglia appeared well organized, but spinal nerve projections emerging from them often failed to constrict as they left the ganglia, resulting in a broad projection into the limb bud (arrows). Scale bar in A,E,I,M, 2000 μm; in B,D,F,H,J,L,N,P, 1000 μm; in C,G,K,O, 333 μm.
Cranial and dorsal root ganglia/nerve formation in E10.5 embryos. Side views of day 10.5 embryos stained in whole mount with anti-TUJI antibodies, illustrating the organization of cranial ganglia and nerves, dorsal root ganglia and nerves in wild-type (A–D), HRG−/− (E-H), ErbB2−/− (I–L) and in ErbB3−/− (M-P) embryos. (A–C) Low to high magnification views of a control embryo, illustrating the normal organization of the cranial ganglia. (D) Cervical-thoracic dorsal root ganglia near the forelimb bud (LB) illustrating the normal constriction of the spinal nerve as it leaves the dorsal root ganglion to enter the periphery (arrows). Note the well-organized ganglia and fasciculated axonal projections. (E–G) Low to high magnification views of cranial ganglia in a HRG−/− embryo. The trigeminal ganglion has ophthalmic (op) and maxillary (mx) nerve projections, but not the mandibular (mn) component. The geniculate and vestibulo-cochlear ganglia (VII/VIII) appear fused, and the facial nerve (VIIn) emerging from the geniculate is small. The petrosal ganglion (IX) is also reduced in size, as is the nodose (X) ganglion. (H) The cervical dorsal root ganglia appear disorganized and projections emerging from them are unfasciculated (arrow) compared to those typical of wild-type embryos. (I–K) Low to high magnification views of cranial ganglia and nerve in an ErbB2−/− embryo. The trigeminal ganglion (V) has only the ophthalmic projection (Vop), the geniculate and vestibulo-cochlear ganglia (VII/VIII) are fused, and the facial nerve (VIIn) is quite small. Note the scatter of neurons between the trigeminal (V) and VII/VIII ganglia. The jugular and the nodose (X) ganglia are significantly reduced in size and vagal projections of the nodose extend too far ventrally. Although vagal projections fuse with those from C1, in combination with the lack of projections from XIIn, this bundle is very rudimentary. (L) Axonal projections emerging from dorsal root ganglia in the cervical region are quite unfasciculated, and the ganglia appear disorganized. (M–O) Low to high magnification views of the cranial ganglia and nerves in ErbB3−/− embryos. In general, these appear similar to wild-type controls, with notable exceptions. The trigeminal ganglion (V) is lacking the mandibular projection (mn), the vestibulo-cochlear and geniculate ganglia (VII/VIII) appear normal, the petrosal (IXg) was only slightly malpositioned, although it and the nodose (Xg) were reduced in size. Other projections including those from XIIg, and C1 were also unaffected. (P) ErbB3−/− dorsal root ganglia appeared well organized, but spinal nerve projections emerging from them often failed to constrict as they left the ganglia, resulting in a broad projection into the limb bud (arrows). Scale bar in A,E,I,M, 2000 μm; in B,D,F,H,J,L,N,P, 1000 μm; in C,G,K,O, 333 μm.
ErbB2−/− embryos (Fig. 7I,J,K) similarly lacked the posterior portion (semilunar ganglion) of the trigeminal (Lee et al., 1995), with scattered neurons occupying the space between Vg and the geniculate (VIIg) (Fig. 7J). The mandibular portion of nV was totally absent and, unlike previous reports, the maxillary portion (Vmx) was also reduced in size. The geniculate ganglion (VIIg) and the vestibulo/cochlear ganglion (VIIIg) appeared fused, and few axons (nVII) projected from the geniculate toward the second branchial arch. Both nIX (glossopharyngeal) and superior nerve fibers emerging from the petrosal (IXg) were significantly stunted in these embryos. The petrosal ganglion itself had a rounded, rather than elongated organization and was positioned a considerable distance from the nodose (Xg); the jugular ganglion was significantly reduced in size. The vagus contacted C1 near the usual position dorsal to the limb bud, but rootlets of nXII were missing from this plexus. The vagal projection in many ErbB2−/− embryos had extended too far ventrally, and had actually made a dorsal turn to contact C1.
ErbB3−/− embryos (Fig. 7M–O) similarly lacked the mandibular division (Vmn) of the trigeminal. The vestibulo/cochlear (VIIIg) and geniculate (VIIg) ganglia, however, appeared unaffected. The petrosal (IXg) and the nodose (Xg) were decreased in size; the petrosal was often mal-positioned ventrally, with the superior nerve fused with rootlets emerging from the jugular ganglion (nX), rather than projecting toward the otic vesicle. Unlike HRG−/− and ErbB2−/− embryos, the hypoglossal (nXII) appeared normal, as did vagal and C1 projections.
Interestingly, while hindbrain connections to/from the trigeminal were absent in both the HRG−/− and ErbB2−/− embryos, in the ErbB3−/− embryos these connections were present, albeit rudimentary and disorganized (Fig. 4D,J). Because ErbB4−/− embryos did not have abnormalities of cranial ganglia but instead had aberrant connections to and from the hindbrain (Gassmann et al., 1995), the cranial ganglia abnormalities characteristic of HRG−/−, ErbB2−/− and ErbB3−/− embryos demonstrate the importance of ErbB2/ErbB3 heterodimers for HRG signal transduction in the trigeminal, petrosal and nodose ganglia. However, ErbB3−/− embryos were noticeably less affected than HRG−/− and ErbB2−/− embryos, having normal vestibular/cochlear and geniculate ganglia (VII/VIIIg), a normal hypoglossal (XII) and normal vagal and C1 projections. These unaffected ganglia/nerves suggest that another yet unidentified receptor could be cooperating with HRG and ErbB2 in this region.
Considerable work has been done in chick embryos to examine the relative contribution of cranial neural crest (high expressors of both HRG and ErbB3) and placodal epithelium to the cranial ganglia (D’Amico-Martel and Noden, 1983; Le Douarin et al., 1991). In general, it appears that placodal neurons are present in the more distal portions of these ganglia, with cranial neural crest forming Schwann cells and proximal neurons in these structures. The notable exception to this rule is the ganglion (petrosal) of IX, which is nearly entirely of placodal origin (D’Amico-Martel and Noden, 1983). The previous suggestion that there is a reduction in the cranial neural crest contribution to these structures, particularly the trigeminal (review, Lemke, 1996) is supported by the current investigation. However, following the hypothesis that cells of cranial ganglia origin are abnormal and those of placodal origin are normal, one would predict more severe defects (deletions) of ganglia populated largely by cranial neural crest derivatives and little effect on those ganglia derived largely from placodes. However, in ErbB3−/− embryos, the petrosal ganglia, which receives the majority of its cells from the ectodermal placode of the 2nd branchial arch, is severely affected while facial mesenchyme, derived from cranial neural crest cells, appears normal in the same embryos. Because cranial neural crest contributed normally to cartilage and facial mesenchyme, it appears that at least certain classes of cranial neural crest do migrate properly. In contrast to the CNS where little cell death was apparent, moderate neuronal cell death was present in these ganglia at E10.5, particularly in the trigeminal. There was also moderate cell death in the vestibular/cochlear ganglia and in the dorsal root ganglia, all ganglia composed of neural-crest-derived structures. This neuronal cell death suggests that neuronal differentiation proceeds but that many neurons do not survive in the (abnormal) environment of the ganglia.
Neural-crest-derived Schwann cells, normal expressors of both ErbB2 and ErbB3, were present in reduced numbers in the ErbB3−/− trigeminal ganglia (Vg) as evidenced by S 100a staining (data not shown). Due to the loss of ErbB3 expression, it is possible that these Schwann cells are themselves defective and unable to properly contribute (by either the reduction in Schwann cell number or a loss of function, such as the ability to produce growth factors) to an environment necessary for neuronal survival. It has previously been shown that, in the early embryo, there is widespread expression of ErbB3 in the facial mesenchyme where cranial ganglia reside and where nerves grow to their targets (Meyer and Birchmeier, 1995). Like the Schwann cells, perhaps the absence of ErbB3 in the mesenchyme renders the environment (through a loss of either +/− cues) inadequate for neuronal survival and subsequent nerve outgrowth/pathfinding.
Dorsal root ganglia show few defects
Dorsal root ganglia, of trunk neural crest origin and known to express HRG, ErbB2 and ErbB3, appeared to be of normal size in all three nulls. However, moderate cell death was apparent in the dorsal root ganglia and the ventral root projections in the HRG−/−- (Fig. 7H) and ErbB2−/− (Fig. 7L) groups were strikingly unfasciculated compared with wild-type controls (Fig. 7D); this was particularly obvious in the ErbB2−/− embryos. In HRG−/− embryos, the constriction in the ventral root where it will emerge from the adult intervertebral foramen was missing in the lower cervical ganglia projection to the anterior limb bud, and there was a significant delay in development of projections from the posterior (lower thoracic) ganglia. In ErbB3−/− embryos, dorsal root ganglia appeared more similar to controls in the degree of neurite fasciculation and organization of the ganglion itself. However, like the HRG−/− embryos, the normal constriction of the ventral root was missing (Fig. 7P). As in the cranial ganglia, Schwann cells were present in reduced number in the ErbB3−/− dorsal root ganglia and relatively few supporting cells were present along the surface of axon projections from either the dorsal root ganglia or the cranial nerve. Again, the Schwann cells of the dorsal root ganglia could be abnormal, resulting in an environment inadequate for proper patterning and pathfinding. High ErbB3 expression is found in mesenchyme adjacent to nerves coming from the dorsal root ganglia that innervate the limb bud (Meyer and Birchmeier, 1995) and the lack of proper environmental cues due to the loss of ErbB3 expression in the mesenchyme could contribute to defective guidance and fasiculation. Studies of Schwann cells are ongoing, as are recombination experiments to elucidate further the role of the mesenchymal environment and the ErbB family in these processes.
Stomach, pancreas and adrenals affected by lack of ErbB3
While HRG−/− and ErbB2−/− embryos died prior to the development of most organ primordia, ErbB3−/− embryos were viable through organogenesis. Certain organs were found to be highly affected by the absence of ErbB3. Interestingly, overexpression of ErbB2 has previously been implicated in many adenocarcinomas of these regions in adults (Lemoine et al., 1992; Tahara, 1995). In the stomach, ErbB3 is normally expressed in the epithelium. E13.5 ErbB3−/− embryos exhibited a normal organization of stomach layers; although both the mesenchyme and epithelial lining appeared quite thinned (Fig. 8A,C). In addition, although present, trunk neural-crest-derived enteric ganglia in the forming gut were greatly reduced in number (Fig. 8B,D). Neural-crest-derived adrenal medullary cells also were affected in the ErbB3−/− embryos as no chromaffin cells stained with a tyrosine hydroxylase antibody (Fig. 8E,F), again suggesting a generalized defect in neural crest derivatives. ErbB3 expression was present in the mesenchyme of the adrenal.
Development of stomach, pancreas and trunk neural crest derivatives is affected by the lack of ErbB3. Compared with the wildtype control (A), both mesenchyme and epithelial layers (E) of the E13.5 ErbB3−/− stomach (C) appeared thinned. Anti-peripherin staining of the developing wild-type (B) and ErbB3−/− (D) gut shows a significant decrease in the number of enteric ganglia (arrows) in the ErbB3 null. D, duodenal region. By E13.5, neural crest cells had migrated to their adult location in the adrenal medulla in wild-type embryos and express tyrosine hydroxylase activity typical of chromaffin cells (E). There was a significant reduction in the number of TH-expressing chromaffin cells present in the adrenal of the ErbB3-/- embryo (arrows, F). Interestingly, ErbB3 was highly expressed only in the connective tissue layer of the adrenal. Wildtype E13.5 pancreas (G) is quite organized compared to the rudimentary ErbB3−/− pancreas (H) which appears less developed, with variable penetrance among embryos. Scale bar in A,C,G,H, 50 μm; in B,D, 200 μm; in E, 47 μm; in F, 80 μm.
Development of stomach, pancreas and trunk neural crest derivatives is affected by the lack of ErbB3. Compared with the wildtype control (A), both mesenchyme and epithelial layers (E) of the E13.5 ErbB3−/− stomach (C) appeared thinned. Anti-peripherin staining of the developing wild-type (B) and ErbB3−/− (D) gut shows a significant decrease in the number of enteric ganglia (arrows) in the ErbB3 null. D, duodenal region. By E13.5, neural crest cells had migrated to their adult location in the adrenal medulla in wild-type embryos and express tyrosine hydroxylase activity typical of chromaffin cells (E). There was a significant reduction in the number of TH-expressing chromaffin cells present in the adrenal of the ErbB3-/- embryo (arrows, F). Interestingly, ErbB3 was highly expressed only in the connective tissue layer of the adrenal. Wildtype E13.5 pancreas (G) is quite organized compared to the rudimentary ErbB3−/− pancreas (H) which appears less developed, with variable penetrance among embryos. Scale bar in A,C,G,H, 50 μm; in B,D, 200 μm; in E, 47 μm; in F, 80 μm.
By E13.5, wild-type pancreas was organized into a branched epithelium surrounded by mesenchyme (Fig. 8G), yet the ErbB3−/− pancreas was far less developed (Fig. 8H); the mesenchyme was thinned, with variable penetrance among embryos. Like the adrenals, ErbB3 is expressed in pancreatic mesenchyme, not epithelium, suggesting either a paracrine role, or that a cascade of mesenchymal factors are required for pancreatic differentiation. Differentiation had proceeded in the ErbB3−/− pancreas as islet cell markers were detectable, albeit at a strikingly reduced level.
DISCUSSION
Heregulin has been described as a mesenchymal factor as it is often expressed in close proximity to neighboring epithelia (Meyer and Birchmeier, 1994). Therefore, it is not surprising that the ErbB3−/−, ErbB2−/− and HRG−/− phenotypes are characterized by a pattern of anomalies many of which are likely associated with a general mesenchymal defect. While neural crest cells appeared to retain their ability to migrate in all three nulls, as shown in development of facial cartilage, aortic arches and dorsal root ganglia, a large number of tissues that require epithelial-mesenchymal interaction including the cardiac cushions and the choroid plexes (required for the cascade of events including pontine flexure and normal expansion of the ventricular system) were abnormal.
The heart originates from precardiac mesoderm and several epithelial-mesenchymal transformations are required prior to the appearance of the myocardium and endocardium. Data suggest that selected endocardial cells, responding to signals from the myocardium, have the potential to undergo yet another transformation to prevalvular (cushion) mesenchyme cells, which are very important in remodeling the tubular heart into its final four-chambered structure (Eisenberg and Markwald, 1995). These prevalvular mesenchymal cells express ErbB3. In the present study, the absence of ErbB3 in these cells ultimately results in defective valve formation characterized by a lack of sufficient connective tissue, resulting in death at E13.5. Paracrine signaling between HRG in the endocardium lining the cushions (Meyer and Birchmeier,1995) and ErbB3, present in endocardial cushion mesenchymal cells (Meyer and Birchmeier, 1994), appears to be required for normal development of this region, with HRG regulating cushion development through an ErbB3-containing heterodimer, probably ErbB2/ErbB3. (Another unknown ErbB receptor could be involved as expression of an ErbB3 coreceptor in the cushions has not yet been shown.) This is similar to the paracrine model proposed by Lemke (1996) and supported by our study, in which myoblast trabecular differentiation is regulated by HRG (or a currently unknown ligand) signaling via the ErbB2/ErbB4 heterodimer.
In the murine brain, there is widespread but tissue-specific expression of HRG (Meyer and Birchmeier, 1994) and ErbB2 (Kokai et al.,1989) throughout embryonic development. Interestingly, ErbB3 is the major ErbB receptor present early in gestation in the chick brain, but disappears entirely at E6 only to reappear transiently at E14 with expression confined only to the cerebellum (Francoeur et al., 1995). A cascade of gene expression, including wingless and engrailed genes, appears crucial for proper cerebellar patterning. This study demonstrates that the ErbB family plays a previously unknown role in the differentiation and patterning of the midbrain/hindbrain junction possibly influencing the cascade of gene expression involved in marking cerebellar boundaries (eg. Davis and Joyner, 1988). While the HRG and ErbB2 nulls die too early in development, the abnormal cerebellum and defective adjacent mesenchyme is apparent in the E13.5 ErbB3−/− embryos. However, the absence of engrailed-expressing cerebellar precursor cells occurs in all three nulls by E10.5, suggesting a similar cerebellar phenotype would likely be observed in the HRG and ErbB2 nulls were they to survive. In addition, the impaired migration of dopaminergic neurons and the defective adjacent mesenchyme in the mesencephalic tegmentum of the ErbB3−/− embryos suggests yet another region where HRG signaling might play a role in neuronal patterning through close CNS-mesenchymal contact. The development of the choroid plexus also requires signaling between epithelium and mesenchyme, with thinning of the overlying neuroepithelium and ingress of mesenchyme into the branching epithelium. Again, signaling between these two populations is impaired in the ErbB3−/− embryos and the extended ectopic expression of the Wnt-1 boundary molecule appears involved in the abnormal elongation of the choroid plexus.
Although it is widely assumed that the trunk neural tube is largely unpatterned in the absence of mesenchyme, the role of mesenchymal cells in patterning of the cephalic neuroepithelium is less well defined. In the current investigation, two regions, the cerebellum and the mesencephalic tegmentum areas, which are in close contact with significant amounts of mesenchyme, were grossly malformed. Although it is possible that the abnormal folding in these regions is a result of mechanical forces (e.g., lack of support via reduced cell number, CSF pressure, abnormalities of extracellular matrix), it is plausible that the HRG family of membrane-bound and diffusible factors may play a role in short-range signaling and gradient formation as has been shown in floor plate/notochord interactions (Yamada et al., 1991) and with retinoic acid and its many receptors and binding proteins (Gudas, 1994; Horton and Maden, 1995) during development. The presence of heparin-binding domains in HRG isoforms may facilitate binding and sequestering of these factors in the largely cell-free space of the mesenchyme, thereby producing a gradient, for the paracrine effects to be realized.
Due to the limited life span of the HRG and ErbB2 gene deletion embryos, only data regarding the earliest developing systems were obtained. The ErbB3−/− embryos provide an opportunity to identify specific organs and regions of the central nervous system where the ErbB family is required. While no defects were noted in the developing E13.5 lung or kidney, several organs dependent upon epithelial-mesenchymal interactions during development were dramatically affected, namely the stomach and pancreas. Interestingly, ErbB3 was found to be expressed in the stomach epithelia while in the pancreas and adrenal gland, ErbB3 expression was confined to the surrounding mesenchyme (data not shown). Perhaps in some instances ErbB3, capable of binding HRG yet incapable of transducing a signal on its own, could itself act in a paracrine manner and/or help facilitate HRG sequestering in the mesenchyme.
The effect of the ErbB gene deletions on neural-crest-derived tissues clearly illustrates that certain derivatives are affected (e.g. cranial ganglia, Schwann cells, enteric ganglia, chromaffin cells) while, significantly, others are not (cartilage, facial mesenchyme). Development of cranial ganglia and cerebellar differentiation, like formation of the mesenchyme of the cardiac cushions, share a requirement for morphogenic cell migration. It is possible that this family plays a role in cell motility, cell-cell, or cell-ECM interactions requisite for this complex series of events. Since HRG is frequently expressed in mesenchyme near epithelial borders, the ErbB3−/− phenotype may reflect a general mesenchymal defect implicating this receptor in the patterning of both the nervous system and certain selected organ systems.
ACKNOWLEDGEMENTS
We are grateful to William E. Holmes for genomic cloning of HRG, John Koland for rat ErbB3 cDNA sequence; Mary Hynes for the Wnt1 probe; Alex Joyner for Engrailed antisera; Xiao-Kang Lu and Roger Premo for frozen sectioning and Sonia Janich and Julio Ramirez for paraffin sectioning of embryos; Jose Zavala for maintaining our mouse colony; Chris Lugo for PCR genotyping; Robin Taylor for immunohistochemistry; Mark Vasser and Peter Ng for oligo synthesis; Sharon Pitts-Meek and Mary Dowd for microinjection of ES cells; Charles Hoffman for preparation of figures; Greg Lemke for critical review of this study and Mark Sliwkowski for extremely helpful discussions, advice and critical review of this manuscript. All animal care was in accordance with NIH guidelines.