Requirements for Jag1-Rbpj mediated Notch signaling during early mouse lens development


  • Tien T. Le,

    1. Division of Developmental Biology, Cincinnati Children's Hospital Research Foundation, Cincinnati, Ohio
    Search for more papers by this author
  • Kevin W. Conley,

    1. Division of Developmental Biology, Cincinnati Children's Hospital Research Foundation, Cincinnati, Ohio
    Current affiliation:
    1. Museum of Science and Industry, Chicago, IL 60637
    Search for more papers by this author
  • Timothy J. Mead,

    1. The Heart Institute, Division of Molecular Cardiovascular Biology, Cincinnati Children's Hospital Research Foundation, Cincinnati, Ohio
    Current affiliation:
    1. Department of Cell Biology, Cleveland Clinic Lerner Research Institute Cleveland, OH 44195
    Search for more papers by this author
  • Sheldon Rowan,

    1. Division of Genetics, Department of Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston, Massachusetts
    Search for more papers by this author
  • Katherine E. Yutzey,

    1. The Heart Institute, Division of Molecular Cardiovascular Biology, Cincinnati Children's Hospital Research Foundation, Cincinnati, Ohio
    Search for more papers by this author
  • Nadean L. Brown

    Corresponding author
    1. Division of Developmental Biology, Cincinnati Children's Hospital Research Foundation, Cincinnati, Ohio
    • Division of Developmental Biology, Cincinnati Children's Hospital Research Foundation, ML 7007, 3333 Burnet Avenue, Cincinnati, OH, 45229
    Search for more papers by this author


Background: During vertebrate lens development, the lens placode in the embryonic ectoderm invaginates into a lens vesicle, which then separates from the surface epithelium, followed by two waves of fiber cell differentiation. In the mouse, multiple labs have shown that Jag1-Notch signaling is critically required during the second wave of lens fiber cell formation. However, Notch signaling appears to play no obvious role during lens induction or morphogenesis, although multiple pathway genes are expressed at these earlier stages. Results: Here, we explored functions for Notch signaling specifically during early lens development, by using the early-acting AP2α-Cre driver to delete Jag1 or Rbpj. We found that Jag1 and Rbpj are not required during lens induction, but are necessary for proper lens vesicle separation from the surface ectoderm. Conclusions: We conclude that precise levels of Notch signaling are essential during lens vesicle morphogenesis. In addition, AP2α-Cre-mediated deletion of Rbpj resulted in embryos with cardiac outflow tract and liver deformities, and perinatal lethality. Developmental Dynamics 241:493–504, 2012. © 2012 Wiley Periodicals, Inc.


The vertebrate lens is well suited for studies of coordinated cell growth, differentiation and morphogenesis, because experiments can be performed at any age, and tissue loss is not life threatening. Lens formation initiates in a group of embryonic ectodermal cells fated to become the presumptive lens ectoderm (PLE). The PLE overlies the forming optic vesicle, which evaginates outward from the ventral diencephalon. As the PLE and optic vesicle contact one another, the PLE thickens into a placode. Both extrinsic and intrinsic factors orchestrate the rapid invagination of this placode into a lens pit, which occurs simultaneously as the optic vesicle changes into a cup. Subsequently, the lens pit separates from the surface ectoderm, creating a hollow lens vesicle comprised of progenitor cells. Cells at the posterior side of the vesicle, closest to the retina, are the first to become postmitotic. These differentiating cells rapidly elongate into lens fibers that will become the core of the mature lens. Primary fibergenesis displaces proliferating progenitor cells anteriorly, where they coalesce into an anterior epithelial layer (AEL). The juxtaposition of anterior and posterior lens compartments reorients subsequent differentiation to proceed circumferentially. During secondary fibergenesis, AEL cells become postmitotic within the germative zone, pass through the equator or transition zone, then elongate and express markers of terminal differentiation within the fiber cell compartment.

One signal transduction pathway known to regulate lens progenitor cell growth and differentiation is Notch signaling (Jia et al., 2007; Ogino et al., 2008; Rowan et al., 2008). Notch signaling mediates intercellular communication by means of ligands and receptors present on adjacent cells. There are two types of ligands, Deltalike or Jagged, whose encoded proteins bind a Notch receptor through extracellular DSL and EGF-like domains. Among the ligands, Dll1, Dll3, Dll4, and Jag1 are expressed in the prenatal mouse lens (Lindsell et al., 1996; Bao and Cepko, 1997). A further complication for deciphering molecular mechanisms of Notch signaling is the presence of four mammalian Notch genes (Notch1–4), of which Notch1, Notch2, and Notch3 are transcribed during lens development (Lindsell et al., 1996; Bao and Cepko, 1997). Regardless of which ligands and receptors interact with one another, upon binding, the Notch intracellular domain (NotchIC) is released by a series of proteolytic cleavages. The NotchIC then translocates to the nucleus and forms a complex with Rbpj and MAML transcription factor proteins, which directly activates downstream genes (reviewed in Fortini, 2009; Kopan and Ilagan, 2009).

As in many other tissues, Notch signaling performs multiple roles during lens development. In the frog embryo, optic vesicle cells express Delta, which activates Notch within the lens placode, resulting in Su(H)/Rbpj-mediated transcriptional activation of Foxe3 (Ogino et al., 2008). Foxe3 is expressed in the lens placode, vesicle and AEL, where it is essential for lens growth and differentiation (Blixt et al., 2000; Medina-Martinez et al., 2005). By contrast, in the mouse, Jag1, Notch2, and Rbpj are required for lens progenitor cell growth and differentiation (Jia et al., 2007; Rowan et al., 2008; Le et al., 2009; Saravanamuthu et al., 2009), although Cre-mediated deletion of Jag1 or Rbpj does result in Foxe3 down-regulation and a significant loss of AEL proliferation. The end result in this case is a progressive loss of the AEL and transitional zone compartments, with eventual lens aphakia (Jia et al., 2007; Rowan et al., 2008; Le et al., 2009). Conversely, misexpression of the Notch1IC induces abnormal expansion of the AEL and transition zones, as well as ectopic Foxe3 expression (Rowan et al., 2008). Together, these studies indicate that Jag1-Notch-Rbpj signaling is critical for mouse lens progenitor cell growth, particularly during secondary fibergenesis. However, it is unclear whether Notch signaling is active during mouse lens induction and vesicle formation, as it is in the frog eye.

To test specifically for Notch signaling requirements during mouse lens induction and morphogenesis, we deleted Jag1 or Rbpj during placode progression into a vesicle. For this, we used AP2α-Cre mice (Macatee et al., 2003), in which Cre activity initiates in the lens ectoderm earlier than in Le-Cre transgenic mice (Ashery-Padan et al., 2000; Song et al., 2007). We observed no obvious requirement for Jag1 or Rbpj during lens induction. Instead, we found that the early loss of Jag1 or Rbpj affects lens vesicle separation from the surface ectoderm, while overexpression of activated Notch1 (NotchIC) also delays vesicle closure and separation. Because AP2α/Tcfap2a is expressed in other tissues of the body (Zhang et al., 1996; West-Mays et al., 1999; Macatee et al., 2003), AP2α-Cre-mediated deletion of Rbpj resulted in embryonic cardiac outflow tract and liver abnormalities. Overall, we conclude that, in the rodent eye, Notch signaling is not required for lens induction, although the appropriate amount of signaling activity is important during lens vesicle formation.


Previously, Ogino and colleagues demonstrated that when frog Delta, Notch, or Su(H) (the Rbpj orthologue) activities were inhibited, Foxe3 transcriptional activation and lens induction were blocked (Ogino et al., 2008). By contrast, loss of mouse Jag1 or Rbpj did not affect lens placode, pit, and vesicle formation, although some mutants failed to fully separate the lens vesicle from the surface ectoderm (Jia et al., 2007; Rowan et al., 2008; Le et al., 2009). Because the developmental timing to block Notch signaling differed in these studies, here we conditionally deleted Jag1 and Rbpj, using the AP2α-Cre driver (Fig. 1A; Macatee et al., 2003), which is active in the surface ectoderm and lens placode 24 hr earlier than the previously used Le-Cre driver (Song et al., 2007).

Figure 1.

Ap2α-Cre deletion of Jag1 or Rbpj during very early lens formation. A: Conditional mutant strategies to create Ap2α-Cre;Jag1CKO/CKO and Ap2α-Cre;RbpjCKO/CKO embryos, as further described in (Fig. 1 of Han et al., 2002; Fig 2 of Macatee et al., 2003; Supp. Fig. S1 of Brooker et al., 2006). For Jag1, the DSL domain that binds to Notch receptors is deleted, creating an out of frame protein; whereas for Rbpj the DNA-binding domain is removed. B,C: Endogenous Ap2α/Tcfap2a protein expression (green) in the forming lens vesicle is unaffected by removal of Jag1. D–F: Complete loss of Jag1 protein (white) in E 10.5 Jag1 mutant lens pit, but not from distal optic cup. Arrows in D,E point to Jag1 expression in cell membranes. G–I: Simultaneous colabeling for Hes1 in sections shown in D–F. For Jag1 heterozygous and homozygous conditional mutants, Hes1 protein (white arrows point to Hes1+ nuclei) is dramatically down-regulated in the lens pit, but unaffected in the adjacent optic cup of all three genotypes. J–L: Loss of Rbpj protein in E 10.5 Rbpj conditionally mutant lens pits, but not the adjacent optic cup (L). Arrows in J,K point to nuclear Rbpj protein expression. M–O: Down-regulation of Hes1 protein in Rbpj conditional mutants (O). Arrows point to Hes1+ nuclei in M,N. P–R: Jag1 protein expression is unaffected in Rbpj conditional mutants. SS, signal sequence; DSL, Delta/Serrate/Lag binding domain; CR, Cysteine repeat; TM, transmembrane domain; LP, lens pit; LV, lens vesicle. Anterior is up in all panels. Scale bar = 20 microns in B; n ≥ 3 embryos per age and genotype.

Ap2α-Cre-mediated Early Deletion of Jag1 or Rbpj in the Embryonic Lens

The AP2α-Cre mouse contains an IRES-Cre cassette homologously recombined into the 3′untranslated region of the AP2α/Tcfap2a gene (Macatee et al., 2003). The diagram in Figure 1A depicts the Cre-mediated recombination events that remove exons 4 and 5 from Jag1, thereby creating a nonfunctional, out-of-frame translated protein (Brooker et al., 2006); or exons 6 and 7 from Rbpj that remove the DNA binding domain (Han et al., 2002). To verify that endogenous AP2α expression is unaffected by the IRES-Cre insertion, we performed AP2α antibody staining of early lens sections from control (Fig. 1B), AP2α-Cre;Jag1CKO/CKO (Fig. 1C) and AP2α-Cre;Jag1CKO/+ embryos. We observed identical spatiotemporal expression patterns within the lens placode, pit, and vesicle for all genotypes. In addition, AP2α expression was unaltered by the removal of Jag1 (Fig. 1C).

Next, we asked how early does AP2α-Cre delete Jag1 or Rbpj proteins in mutant cells, and how quickly is the Notch pathway effector Hes1 down-regulated (Jarriault et al., 1995). Previously Le-Cre;Jag1CKO/CKO lens pit cells were shown to exhibit no Jag1 protein, and down-regulate Hes1 beginning at embryonic day (E) 10.5. But the complete loss of Hes1 did not occur until E12.5 (Fig. 6 of Le et al., 2009). In this study, AP2α-Cre;Jag1lensCKO/CKO lenses had a partial loss of Jag1 protein at E9.5 (n = 6 mutants) and a complete loss at E10.5 (compare Fig. 1D–F). At E10.5, there was also a dramatic reduction of Hes1 expression, two days earlier than in the Le-Cre experiments (Fig. 1G–I). Normally, the nuclear Rbpj protein is ubiquitously expressed throughout the forming eye (arrows in Fig. 1J,K). We observed that Rbpj expression is unaffected in AP2α-Cre;Jag1CKO/+ and AP2α-Cre;Jag1CKO/CKO lenses. However, E10.5 AP2α-Cre;RbpjCKO/CKO mutants show the specific removal of Rbpj nuclear protein from lens pit cells (compare Fig. 1J–L). Furthermore, Hes1 protein is also lost from E11 AP2α-Cre;RbpjCKO/CKO mutant lens pit cells (compare Fig. 1M–O), indicating a 2–3 day earlier block in Notch signaling activity than in Le-Cre; RbpjCKO/CKO lenses (Rowan et al., 2008). Finally, we observed normal Jag1 expression in E11 AP2α-Cre;RbpjCKO/CKO lens vesicles (Fig. 1P–R), consistent with our idea that Notch signaling does not feedback onto Jag1 ligand expression in the prenatal lens (Rowan et al., 2008; Le et al., 2009).

Jag1 or Rbpj Removal Causes Prenatal Lens Microphthalmia

At E11.5 and beyond, AP2α-Cre; Jag1CKO/CKO and AP2α-Cre;RbpjCKO/CKO embryos had grossly malformed eyes, as highlighted by the pigmented RPE (Fig. 2C,L). Histologic sections from these embryos showed that loss of Jag1 or Rbpj resulted in lens microphthalmia (Fig. 2F,O). In most cases, the smaller sized lenses were surrounded by an abnormally folded retina and RPE (Fig. 2I,O). These types of defects are also found in other mutations causing lens microphthalmia, such as conditional or germline deletion of Pax6, Foxe3, or Pygo2 genes (Ashery-Padan et al., 2000; Blixt et al., 2000; Medina-Martinez et al., 2005; Song et al., 2007). To compare the phenotypes produced by AP2α-Cre vs. Le-Cre drivers, we asked whether postnatal lens aphakia (Rowan et al., 2008; Le et al., 2009) also occurred in AP2α-Cre-induced lens mutants. However, AP2α-Cre; RbpjCKO/CKO embryos began dying between E14.5 and E15.5 (n = 8 mutants (2 necrotic) from 11 E14.5 litters, containing 85 embryos; n = 5 mutants (3 necrotic) from 4 E15.5 litters, containing 34 embryos). No mutants were recovered beyond this age. Normal Mendalian ratios of AP2α-Cre;Jag1CKO/CKO embryos were present at postnatal day (P) 0 (Fig. 2G–I), after which we were also unable to recover Jag1 conditional mutants (n = 0 mutants from 4 P3 or P21 litters containing 23 animals). P0 Jag1 mutant lenses were very small (Fig. 2I; n = 5 mutants from 4 litters containing 34 pups). Each embryonic or early postnatal lethality observed presumably stems from widespread expression of AP2α-Cre, which blocked Notch signaling in other, more vital organs (Zhang et al., 1996; Nottoli et al., 1998).

Figure 2.

Perinatal eye deformities in the absence of Jag1 or Rbpj. A–C,J–L: Both Ap2α-Cre;Jag1CKO/CKO and Ap2α-Cre;RbpjCKO/CKO embryos have reduced eyes with abnormal external morphology. D–I: Histologic H&E stained sections show the progressively smaller size of Jag1 mutant lenses. At P 0, Jag1 conditional mutants with very severely reduced lenses, also exhibit abnormally folded retinas. In both F and I, lens fibers are inappropriately present in the forming anterior chamber (arrows in F). M–O:Rbpj conditionally mutant lenses are also obviously smaller, with poorly organized lens compartments. Scale bar = 500 microns in A,J, 20 microns in D,G,M; n = 3 per age and genotype.

The distribution of lens progenitor and fiber cell markers was then assessed from E9.5 onward, to understand if the lens phenotypes in AP2α-Cre;Jag1CKO/CKO and AP2α-Cre; RbpjCKO/CKO embryos are the same, or more severe, than those from Le-Cre deletion of Jag1 or Rbpj (Rowan et al., 2008; Le et al., 2009). We assessed the same lens progenitor, transition zone and differentiated fiber cell markers examined in the two previous Le-Cre studies, from E9.5 to E15.5 in litters containing AP2α-Cre;RbpjCKO/+ and AP2α-Cre; RbpjCKO/CKO embryos, plus and from E9.5 to P0 in litters containing AP2α-Cre;Jag1CKO/+ and AP2α-Cre;Jag1CKO/CKO animals. We found that the earlier-acting Cre induced more severe AEL and transition zone phenotypes, particularly in AP2α-Cre; RbpjCKO/CKO lenses (Fig. 3I,L). AP2α-Cre deletion of Jag1 or Rbpj resulted in smaller Cdh1/Ecad and Foxe3 AEL progenitor cell domains (Fig. 3C,I,M). However, these defects were not apparent until the onset of secondary fibergenesis, as in Le-Cre induced mutants (Rowan et al., 2008; Le et al., 2009). Likewise, the fiber cell differentiation markers Cryg/γ Crystallin (Fig. 3D–F) and Maf (Fig. 3J–L) were activated on schedule during primary fiber cell genesis, although there were fewer differentiation marker+ cells, proportional to an overall reduction in lens size (Fig. 3L). We concluded that the lens progenitor growth and differentiation phenotypes were largely the same using either Cre driver. However, the microphthalmic lens phenotype in AP2α-Cre;RbpjCKO/CKO eyes was more severe, probably due to the earlier block in Notch signaling, as shown by loss of Hes1 expression (Fig. 1).

Figure 3.

Lens progenitor cell defects phenocopy later stage deletion of Jag1 and Rbpj. A–C: Ecad/Cdh1 and p57/Cdkn1c colabeling of postnatal day (P) 0 lens sections highlights smaller lens size and reduced anterior epithelial layer (Cdh1+ cells) and transition zone (Cdkn1c+ cells) compartments in Jag1 conditional mutants (C). D–F: Cryg/DAPI colabeling of sections nearby to those in A–C indicate differentiated fiber cells in Jag1 conditional mutants. G–I: E 14.5 lens sections colabeled with Cdh1 and Foxe3, which are both expressed by AEL progenitor cells. Rbpj conditional mutants have a severely reduced lenses at this age, with very few Cdh1+Foxe3+ cells (arrow in I points to remnant AEL). J–L: E14.5 lens sections, colabeled with Maf and DAPI, from a different set of littermates than those in G–I. The Rbpj mutant lens is also dramatically smaller in size, with disorganized Maf+ fiber cell nuclei. In this mutant, the retina is also abnormally folded. M) The percentage of fiber cells (Foxe3neg/total nuclei) is significantly increased at E10.5 and E14.5 after Ap2α-Cre deletion of Jag1, as with Le-Cre;Jag1 (Le et al., 2009) and Rbpj (Rowan et al., 2008) mutants. Scale bar = 20 microns in A (applies to A–L). Anterior is up. n ≥ 3 per age and genotype; **P < 0.01.

To examine the proportion of lens progenitor to differentiating fiber cells, we compared the percentage of Foxe3-neg cells (a proxy for fiber cell differentiation) among AP2α-Cre;Jag1CKO/CKO embryos and control littermates (Fig. 3M). The percentage of fiber cells was significantly increased at E10.5 and E14.5, but not at E12.5 (Fig. 3M). This is the same outcome as Le-Cre deletion of Jag1 (compare Fig. 3M with Fig. 5A of Le et al., 2009). Because some of the E12.5–E14.5 AP2α-Cre;RbpjCKO/CKO embryos were generally necrotic, we did not estimate the percentage of fiber cells (Foxe3-neg) in these mutants, although a substantial loss of Foxe3+ AEL cells was obvious (arrow points to near absence of red nuclei in Fig 3I).

Jag1 and Rbpj Function Are Required for Lens Separation From the Ectoderm

During mouse lens morphogenesis, the lens pit changes shape into a hollow vesicle, which involves cell shape changes, along with a physical separation of the vesicle from overlying ectoderm (McAvoy et al., 1999). We observed that many Ap2αCre;Jag1CKO/CKO and Ap2αCre;RbpjCKO/CKO embryonic lenses had persistent lens stalks, indicating that either lens separation had failed, or was severely delayed (Figs. 2F,I, 4C,F,I,L). The persistent lens stalk phenotype appeared, in most but not all mutants, between E11 and E11.5 (Fig. 4A–F), when lens vesicle separation normally occurs, and was evident as long as mutant lenses could be recovered (E15.5 in Rbpj mutants, P0 in Jag1 mutants). Cells within the persistent lens stalk expressed Cdh1/Ecad (Fig. 4C,F,L), indicating an epithelial identity, but did not coexpress lens progenitor or fiber cell markers such as Foxe3, Maf, Prox1, Cryb, or Cryg (Fig. 4C,I,L). By E14.5, Ap2αCre; Jag1CKO/CKO or Ap2αCre;RbpjCKO/CKO embryonic lenses retained a central opening to the anterior chamber (arrows in Fig. 4I). Cellular material had been extruded from the lens into the forming anterior chamber or the mesenchymal space around the persistent lens stalk (Fig. 4L). The extruded lens cells expressed the fiber cell differentiation markers Cryb (red staining in Fig. 4L) and Cryg. By contrast, Le-Cre;RbpjCKO/CKO and Le-Cre;Jag1CKO/CKO mutant lenses only very occasionally displayed a persistent lens stalk phenotype (Rowan et al., 2008; Le et al., 2009).

Figure 4.

Jag1 and Rbpj are required for normal lens vesicle separation from the overlying epithelium. A–C: Cdh1/Maf colabeling of Rbpj control, heterozygous and homozygous mutants show proper segregation of anterior progenitor cells from posterior postmitotic Maf+ cells. In the absence of Rbpj, the lens vesicle is abnormally small with a persistent connection to the surface ectoderm by means of a lens stalk (arrow in C). D–F: Cdh1/Cdkn1c/DAPI colabeling shows a very similar mutant phenotype in Jag1 mutants, including a persistent lens stalk (arrow in F). G–L: Some Jag1 or Rbpj conditional mutants retained inappropriate adhesion of the lens vesicle to the overlying ectoderm/forming cornea (arrows in I,L). Panel I shows abnormal extrusion of lens tissue into the anterior chamber at E 14.5. In all cases, these cells were Foxe3-neg, Cdh1-neg, but Cryb+ and Cryg+. The weak, nonnuclear appearance of the extruded material was due to nonspecific secondary antibody labeling. Anterior is up in all panels. Scale bars = 20 microns in A,D,G,I,J; n = 3 per age and genotype.

Notch Regulation of Epithelial Cell Apoptosis in the Lens Pit and Vesicle

Apoptotic cell death is a mechanism by which the lens pit and overlying presumptive cornea separate from one another (Silver and Hughes, 1973; Schook, 1980; Mohamed and Amemiya, 2003). Cells targeted for death are largely localized to the lens pit hinge and connecting stalk (Mohamed and Amemiya, 2003). Therefore, we assayed for cleaved PARP (cPARP) expression in E10.5 Ap2αCre;Jag1CKO/CKO and Ap2αCre; RbpjCKO/CKO mutants (Fig. 5A–C). Jag1 conditionally mutant lens vesicles had a significant increase in cPARP+ cells compared with controls (Fig. 5B,C), but the loss of Rbpj did not. The different outcomes between Jag1 and Rbpj lens mutants might be attributed to phenotypic modulation often found in Rbpj mutants, because the encoded protein can activate or repress gene transcription in a context-dependent manner.

Figure 5.

Apoptosis is significantly increased without Jag1 and when Notch1IC is misexpressed. A,B: cPARP/DAPI colabeling of control and Ap2α-Cre;Jag1CKO/CKO E 10.5 lens sections shows Jag1 is required to suppress ectopic cell death. C: The percentages of cPARP+ cells in E10.5 Rbpj and Jag1 control, heterozygous and homozygous conditional mutants. There is a significant increase without Jag1 function. D–I: Comparison of Pax6 expression (red) with TUNEL (terminal deoxynucleotidyl transferase–mediated deoxyuridinetriphosphate nick end-labeling) labeling of apoptotic cells (green) in E10.5–E11.5 lenses. Apoptotic cells are normally localized only to the E10.5 lens margin (arrows in D). At E11.5, P0-3.9-GFPCre;NotchIC lenses have both delayed lens invagination and abnormal numbers of mislocalized apoptotic cells (I). J: Quantification of TUNEL+ cells in control and Notch1IC mutants. K,L: Excess Notch1 signaling causes abnormal lens vesicle patterning and morphology, and an expanded population of Cdh1+Prox1+ double positive lens cells. Anterior is up in all micrographs. Scale bar = 20 microns in A,F, 40 microns in D,K. n ≥ 4 embryos per age and genotype; panels C, J show mean + SEM; *P < 0.05, **P = 0.01.

Previously, we reported that conditional overexpression of activated Notch1 (P0-3.9-GFPCre;Notch1IC) caused abnormal, persistent attachment of lens and cornea (Rowan et al., 2008). Therefore, we also tested for abnormal morphogenesis and/or apoptosis during lens vesicle formation in gain-of-function (GOF) Notch1IC mutants (Fig. 5D–J). At E10.5, control lenses had cell death localized at the lens pit hinges (arrows in Fig. 5D), which subsequently disappeared by E11.5 (Fig. 5F,H). E10.5 P0-3.9-GFPCre;Notch1IC embryos retained unusually thick lens placodes (Fig. 5E), with a scattered distribution, but normal proportions, of apoptotic cells (Fig. 5E,J). At E11, when the lens normally seals and separates from the overlying ectoderm (Fig. 5F), P0-3.9-GFPCre; Notch1IC vesicles exhibited persistent, deeper than normal, lens pits (Fig. 5G). By E11.5, some P0-3.9-GFPCre; Notch1IC lenses had completed morphogenesis into a vesicle, where a burst of apoptosis occurred (Fig. 5I). Quantification of TUNEL+ (terminal deoxynucleotidyl transferase–mediated deoxyuridinetriphosphate nick end-labeling) cells highlighted a 10-fold increase in apoptosis specifically in E12.5 Notch1IC mutants (Fig. 5J). This delay in completing lens morphogenesis was concomitant with the failure to segregate postmitotic, committed lens progenitor cells (Prox1+) posteriorly (compare Figs. 5K,5L). We conclude that excess Notch1 signaling affects cell cycle exit, morphogenesis and A-P patterning (Fig. 5 and Rowan et al., 2008), while loss of Jag1 or Rbpj impacts morphogenesis and cell cycle exit, but not A-P patterning (Figs. 3C,F,I,L, 4C,F, 5L).

AP2α-Cre Deletion of Rbpj Also Causes Embryonic Cardiac Outflow Tract and Liver Defects

The perinatal lethality of Ap2αCre; RbpjCKO/CKO mutants and early postnatal lethality of Ap2αCre;Jag1CKO/CKOmutants is likely due to removal of Jag1 and Rbpj functionality in other tissues, where Ap2αCre is also active. To explore this further, Rosa26R (R26R) Cre-dependent LacZ lineage reporter mice (Soriano, 1999) were bred to those carrying Ap2αCre. In E10.5 Ap2αCre;R26R embryos, β-galactosidase expression was observed in the cranial, cardiac, and trunk neural crest cells, as well as ectodermally derived cells in the lens and limbs (Supp. Fig. S1, which is available online).

We did not recover any Ap2αCre; RbpjCKO/CKO mutant embryos beyond E15.5. Already by E12.5, the Ap2αCre; RbpjCKO/CKO embryos were runted, with malformed trunk organs, including the heart (Fig. 6; n = 3 embryos per genotype). Here, we observed cardiac malignment at E12.5 and a severe ventricular septal defect at E14.5 in Ap2αCre;RbpjCKO/CKO embryos (Fig. 6C–F; n ≥ 3 embryos per genotype). Therefore, Ap2αCre-mediated deletion of Rbpj reaffirms the importance of Notch signaling in neural crest derivatives during prenatal OFT development (Mead and Yutzey, 2011). In addition, E12.5 Ap2αCre; RbpjCKO/CKO embryos had abnormally small and pale livers (Fig. 6D,H), but with a proportional distribution of Prox1+ hepatocytes (Fig. 6I–J; n = 3 embryos per genotype). There was also normal onset of Albumin expression in the absence of Rbpj. Furthermore, we observed decreased vascularization compared with controls, as determined by anti-Flk1 labeling (Fig. 6K,L; n = 3 embryos per genotype). No heart or liver defects were discernable in Ap2αCre;RbpjCKO/+ heterozygotes at any E12.5–E15.5 (n > 3 embryos per age).

Figure 6.

Prenatal cardiac and hepatic abnormalities in Ap2αCre;RbpjCKO/CKO embryos. A,B: Removal of Rbpj resulted in overall runting and a reduced liver (arrows). C,D: Both heart and liver defects were evident in E 12.5 whole-mount organs (white arrows). E,F: A ventricular septal defect (VSD) is apparent (red arrow) in E14.5 Ap2αCre;RbpjCKO/CKO embryonic heart sections, as compared with control. Asterisks indicate the position of the ventricular septum. G,H: H&E sections show that E12.5 Ap2αCre;RbpjCKO/CKO livers are obviously smaller. I,J: Prox1+ hepatocytes (white nuclei) are present and normally distributed in Rbpj conditional mutants. K,L: Diffuse Flk1 expression highlights a vascular defect in E12.5 Ap2αCre;RbpjCKO/CKO livers. Rostral is left in A,B, anterior is up in C–L; OFT, outflow tract; TV, tricuspid valve; MV, mitral valve. n ≥ 3 per age and genotype. Scale bar = 2.5 millimeters in A; 500 microns in C; 20 microns in E,I.

Interestingly, Ap2αCre;Jag1CKO/CKOmutants exhibited a postnatal lethal period between P0 and P3, suggesting that Jag1 is not critically required for embryonic heart valve or liver formation. Moreover, previous examinations of Jag1 expression and function in the mouse liver demonstrated it acts at late prenatal and postnatal ages during bile duct formation (Piccoli and Spinner, 2001; McCright et al., 2002; Loomes et al., 2007; Lozier et al., 2008; Hofmann et al., 2010). Thus, the early defects in the embryonic liver that require Rbpj are either mediated through a different Notch pathway ligand, or by means of a Notch-independent mechanism (Masui et al., 2007; Hori et al., 2008).


Notch Signaling During Lens Induction and Specification

In this study, we asked whether Jag1-Notch signaling is required at very early stages of mouse lens development, to ascertain whether the mouse and frog eye use similar or distinct mechanisms of lens formation. We separately removed Jag1 and Rbpj activity, by Cre-loxP recombination, and found no apparent function for Notch signaling during lens specification and vesicle formation. In Xenopus, a Delta2 signal emanates from the optic vesicle to activate Notch and initiate lens formation. Activated Notch forms a complex with Su(H), which promotes Foxe3 transcription in the lens placode (Ogino et al., 2008). Therefore, in the mouse optic vesicle, Deltalike activity would be predicted to initiate mouse lens development. Intriguingly, there is no evidence for Deltalike mRNA expression (Dll1, Dll3, or Dll4) at this stage of early eye development, nor is there an obvious mouse orthologue of frog Delta2 in the mouse genome. In the rat optic vesicle and lens placode, Dll1 is clearly absent, although robustly transcribed by nearby optic stalk and diencephalon cells (Bao and Cepko, 1997). Jag1 mRNA expression in the optic vesicle and lens placode has been previously reported, implicating Jag1 as the potential mediator of a lens induction signal (Bao and Cepko, 1997; Le et al., 2009). Furthermore, we have not seen any Jag2 mRNA expression within the developing lens (K.W.C. and N.L.B., unpublished observations). Together, the ligand expression patterns implicate mouse Jag1 as the ligand capable of initiating a lens induction signal, yet Jag1 mutants have normal lens vesicle formation, and proper activation of Foxe3 and other lens determination factors.

More convincing however, is that the loss of the Su(H) orthologue, Rbpj, made no impact on lens induction, specification, or placodal expression of Foxe3. Although it is possible that Jag1 in the optic vesicle directs aspects of lens induction, it would be expected to act by means of canonical actNotch-Rbpj activity within lens progenitor cells. Therefore, the loss of Rbpj activity in early lens progenitor cells should be an analogous genetic test to overexpression of dnSu(H) in the frog eye (Ogino et al., 2008). Yet, it remains possible that inherent temporal differences in blocking canonical Notch signaling may not have been overcome by our use of Ap2αCre here, although Jag1 or Rbpj proteins were rapidly deleted, then followed by dramatic down-regulation of Hes1. Together, our data suggest that early expression of Jag1 and Rbpj are not required for Foxe3 activation in the mouse lens placode. Despite these findings it might be worthwhile to test for Rbpj-dependent activation of Foxe3 in vitro, by overexpressing dominant negative Rbpj, or a Rbpj shRNA construct in a mouse embryonic whole head culture, during the timeframe of lens induction (Furuta and Hogan, 1998).

Roles for Jag1-Rbpj Mediated Notch Signaling in Lens Vesicle Separation and Closure

In vertebrates, the lens stalk is a byproduct of placodal morphogenesis into a lens vesicle. Normally, the transitory lens stalk disappears by means of apoptosis by E11 in mice (Silver and Hughes, 1973; Schook, 1980; Mohamed and Amemiya, 2003). Here, we found that early removal of Jag1 or Rbpj function in the forming lens revealed a previously unknown requirement for Notch signaling during lens morphogenesis. At first glance, the higher proportion of apoptotic cells observed seems contradictory to the persistent lens stalk present in Ap2αCre;Jag1CKO/CKOmutants. If canonical Notch signaling normally induces stalk cell death, there should be fewer apoptotic cells in Jag1 mutants. Moreover, Notch1ICD overexpression also results in a persistent connection between the lens and surface ectoderm, along with a delayed burst of apoptosis.

How might Notch signaling orchestrate lens stalk regression? We propose that without Jag1 to initiate a Notch signal, progression from lens pit to vesicle may proceed at a faster rate, thereby disrupting the localization of apoptosis to the lens hinge. As cell–cell contacts become altered, subsequent apoptosis is induced, producing small lens vesicles, with thin, elongated lens stalks. These persistent structures maintain an epithelial identity, but not that of a lens progenitor cell, because they are Cdh1+ but Foxe3-neg. We also found that the abnormal stalks had poor structural integrity, with central openings that allow Crystallin+ lens fibers to extrude inappropriately into the anterior chamber or surrounding mesenchyme.

Excess Notch signaling also delayed lens vesicle formation, producing a shallow, thickened lens pit. Although initially there is a normal proportion of apoptotic cells in Notch1IC GOF mutants, they are randomized across the lens placode/pit. A round lens vesicle does eventually appear; however, it fails to separate from the surface ectoderm (Rowan et al., 2008, and this study). Interestingly, there is a surge of apoptosis in the center of Notch1ICD GOF mutant lens vesicles at E11.5–E12.5, suggesting that although temporally delayed, lens progenitor cell shape changes or contraction still trigger the apoptosis that should normally occur at the hinge. Latent apoptosis, coupled with abnormal tissue morphology resulted in an open lens vesicle phenotype, resembling the lenses of Peter's Anomaly patients (Rowan et al., 2008).

Because every lens pit and vesicle cell expresses Jag1, Rbpj and Hes1 proteins, we conclude that there is likely ubiquitous Notch signaling cross talk among progenitor cells. In this situation, the stoichiometric balance of signaling levels among lens progenitor cells would be critical. We further propose that Notch pathway interaction and integration with other signaling pathways, for example Wnt signaling, may indirectly alter the timing of lens vesicle morphogenesis. This could occur by means of spatial or temporal alteration of normal apoptosis that severs the vesicle-surface ectoderm connection. In the future it will be interesting to explore this idea, because disruption of pygo2 in the adjacent mesenchyme nonautonomously blocks lens vesicle morphogenesis (Song et al., 2007). It may also be important to examine the morphology and integrity of cytoplasmic processes spanning the lens pit and optic vesicle/cup, or ask whether the Notch pathway regulates (directly or indirectly) lens cell cytoskeletal shape changes, by means of Cdc42 or Shroom3-mediated pathways (Chauhan et al., 2009; Plageman et al., 2010).

Heart and Liver Defects in Ap2-Cre-Mediated Deletion of Canonical Notch Signaling

Notch signaling plays well-known, critical roles in cardiac development (reviewed in Jain et al., 2010; MacGrogan et al., 2010), and the phenotypes described here are consistent with those previously reported (High et al., 2007; Mead and Yutzey, 2011). However, in our study the OFT phenotype arose at a much younger age (E14–E15) than in Wnt1-Cre;DN-Maml mutants (P0–P12), where canonical Notch signaling was dominantly blocked (High et al., 2007). This could be attributable to different spatiotemporal activity between Cre drivers, the timing of homozygous removal of endogenous Rbpj protein, versus dominant blockage of active NotchIC/Rbpj/Maml complexes, or complications arising from simultaneous disruption of Notch-dependent and -independent functions for Rbpj and/or Maml1. The cardiac defects observed in the Ap2αCre;RbpjCKO/CKO mice are similar to those observed with Wnt1-Cre mediated loss of Rbpj (Mead and Yutzey, 2011). Finally, all of the Ap2αCre;RbpjCKO/CKO embryos had misaligned OFTs and ventral septal heart defects, while only 2 of 36 Ap2αCre;Hes1CKO/CKO embryos exhibited OFT malformations (van Bueren et al., 2010). The Rbpj cardiac phenotypes seen here also manifested at a younger age than Wnt1Cre;RbpjCKO/CKO mice, which die at birth (Mead and Yutzey, 2011). The less penetrant phenotype of Hes1 conditional mutants suggests that Rbpj normally regulates additional downstream target genes in the neural crest-derived cells of the cardiac OFT.

In the liver, numerous studies have tested the requirements of Notch pathway genes by means of conditional mutations and multiple Cre drivers, the earliest of which (FoxA3Cre) deletes gene activity beginning at E13.5 (Zong et al., 2009). The conditional phenotypes of Jag1, Notch2, Rbpj and Hes1 present a cohesive picture of the requirement for Notch signaling during biliary cell differentiation and tubule morphogenesis for the bile ducts (Sumazaki et al., 2004; Geisler et al., 2008; Zong et al., 2009; Hofmann et al., 2010; Sparks et al., 2010, 2011). These defects recapitulate particular aspects of Alagille Syndrome, which occurs in individuals who are haploinsufficient for JAG1 or NOTCH2 (Li et al., 1997; Oda et al., 1997; McDaniell et al., 2006). Here, we found obviously smaller livers in Ap2αCre;RbpjCKO/CKO embryos by E12.5. This is earlier than any of the known biliary defects reported for Notch pathway mutations. The small liver size in Rbpj conditional mutants might arise from reduced embryonic hepatoblast growth and/or liver vascularization. We also observed mutant livers were pale, suggesting a potential fetal hematopoietic defect. By contrast, E12.5 Ap2αCre;Jag1CKO/CKO embryos had normal livers (T.J.M., personal observations). Assuming this particular Rbpj mutant phenotype is Notch-dependent, a different ligand must be involved. Most surprising however, is the lack of evidence for Ap2Cre deletion of Rbpj within the embryonic liver. Indeed, Ap2α gene expression has only been reported within ectodermal tissues, including neural crest derivatives (Zhang et al., 1996; West-Mays et al., 1999; Macatee et al., 2003; Song et al., 2007). Therefore, we propose that the small livers observed in Ap2αCre; RbpjCKO/CKO embryos are the result of nonautonomous loss of Rbpj, arising within a population of cells surrounding the forming liver. In the future, it will be interesting to discover how early this prenatal liver phenotype arises, and its underlying molecular mechanism.



Jagged 1tm1JLew mice (Jag1CKO) were generated by Brooker and colleagues, maintained on a C57BL/6 background and genotyped as described (Brooker et al., 2006). Rbpjtm1Hon mice (RbpjCKO) were generated by Han et al., maintained on a 129/SvJ background and genotyped as described (Han et al., 2002). The abbreviation CKO throughout this manuscript indicates a “conditional knock-out” allele. Ap2αCre mice were generated by Macatee and colleagues, maintained on a 129/SvJ background and polymerase chain reaction (PCR) genotyped as described (Macatee et al., 2003). Rosa26R mice (Soriano, 1999) were used to monitor neural crest and ectodermal lineages for the Ap2αCre driver. Rosa26Notch1IC mice have been described, along with the details of PCR genotyping (Murtaugh et al., 2003; Rowan et al., 2008). P0-3.9-GFPCre line was maintained on an FVB background and genotyped as described (Rowan et al., 2008). Embryonic day (E) 0.5 was designated after a morning presence of a vaginal plug. Animals were maintained in accordance with the National Institutes of Health Guide for the Care and Use Laboratory Animals, and Cincinnati Children's Hospital Research Foundation Institutional Animal Care and Use Committee. Images of whole-mount embryos were captured with a Leica dissecting microscope and Optronics digital camera.

Tissue Analyses

Embryonic and postnatal heads were fixed in 4% paraformaldehyde/phosphate buffered saline (PBS) for 1 hr at 4°C, processed through a sucrose/PBS series, cryoembedded, and sectioned. Primary antibodies used were anti-AP2α/Tcfap2a (DSHB Clone 3B5 1:500), anti-cleaved PARP (Cell Signaling 1:500), anti-Cyclin D1/Ccnd1 (Neomarkers SP4 1:200; or Santa Cruz 72-13G 1:500), anti-Cyclin D2/Ccnd2 (Santa Cruz 34B1-3 1:200), anti-E Cadherin/Cdh1 (Zymed ECCD-2 1:500), anti-Foxe3 (gift from Peter Carlsson 1:1,000 or Santa Cruz SC48162 1:200), anti-β-Crystallin/Cryb (gift from Richard Lang 1:8,000), anti-γ-Crystallin/Cryg (Santa Cruz 1:1000), anti-GFP (Molecular Probes 1:1,000), anti-Rbpj (Institute of Immunology, Tokyo or Cosmo Bio Co. USA; clone Hyb-K0043, 1:100); anti-Hes1 (1:1,000), anti-Jag1 (Santa Cruz 1:1,000), anti-p27Kip1/Cdkn1b (BD Laboratories Clone 57 1:100), anti-p57Kip2/Cdkn1c (Abcam 1:200), anti-Pax6 (Covance 1:1000), anti-cMaf/Maf (Santa Cruz 1:200), anti-Prox1 (Covance 1:6,000), anti-Sox1 (Affinity BioReagents 1:500), anti-Flk1 (Santa Cruz sc504, 1:200), and anti-Albumin (Sigma, A3293, 1:1,000), all following (Lee et al., 2005) or (Mead and Yutzey, 2009). Secondary antibodies were directly conjugated to Alexa Fluor 488, Alexa Fluor 594 (Invitrogen) or biotinylated (Jackson Immunoresearch) and sequentially labeled with streptavidin Alexa 488 or 594 (Invitrogen). TUNEL staining was performed on Notch1IC and P0-3.9GFPCre;Notch1IC tissue sections using the in situ cell death detection kit, (Roche). Microscopic imaging was performed on a Zeiss fluorescent microscope with a Zeiss camera and Apotome deconvolution device and Axiovision Imaging software.

For heart and liver analyses, embryos and postnatal pups were fixed in 4% paraformaldehyde/PBS overnight and paraffin embedded. Alcian blue stained and nuclear fast red counterstained sections were generated as described (Mead and Yutzey, 2009). Standard histology of paraffin embedded sections was also performed. Images were processed using Axiovision (v6.0) and Adobe Photoshop software (v7.0) and electronically adjusted for brightness, contrast and pseudocoloring.

β-Galactosidase Detection

Whole-mount X-Gal staining for β-galactosidase activity was carried out as described with modifications (Sanes et al., 1986; Lincoln et al., 2004; Mead and Yutzey, 2009). Embryos were developed in staining solution in the dark for 2 hr at 37°C, fixed in 4% PFA in PBS, washed in methanol and isopropanol, and cleared in tetrahydronaphthalene (Fisher Scientific).

Cell Counting

As a proxy for quantifying lens fiber cells, tissue sections separated by at least 30 μm were antibody-stained with anti-Foxe3 and DAPI (4′,6-diamidine-2-phenylidole-dihydrochloride) counterstain, and then quantified using the measurements module within Axiovision (v6.0) software. Three or more animals were analyzed per genotype and age and at least two independent sections through the center of the lens per animal quantified. The percentage of Foxe3+ lens progenitor cells was determined by dividing the number of antibody-positive cells by the total nuclei, and then subtracted from 100% to obtain percentages of Foxe3-neg lens cells.

To determine the percentage of apoptotic cells during lens vesicle morphogenesis for loss of function mutants, the Photoshop CS4 measurements tool was used to count cPARP+ and DAPI+ cells within a 18 mm2 box centered around the lens vesicle within each ×40 image. For each conditional mutant litter four different forming lenses were serially sectioned per genotype, and all sections immunostained and analyzed. In control versus gain of function NotchIC mutants, the percentage of TUNEL+ cells was quantified from four embryos per genotype, using Image J software. An analysis of variance/Bonferroni post hoc test or a Student's t-test was performed using Instat (v3.0) software to obtain P values.


The authors thank Julian Lewis for Jag1CKO mutant mice; Tasuku Honjo for RbpjCKO mice; Mario Cappechi and Richard Lang for Ap2α-Cre mice; Charlie Murtaugh for Rosa26NotchIC mice; Peter Carlsson and Richard Lang for antibody reagents; April Smith, Jason Spence, Aaron Zorn and Joe Brzezinski for technical advice and valuable discussion. N.L.B. and K.E.Y. were funded by the NIH and T.J.M. received an AHA-Great Rivers Affiliate Predoctoral Fellowship.