Nicastrin controls aspects of photoreceptor neuron specification and differentiation in the Drosophila eye


  • Tianyi Zhang,

    1. Department of Ophthalmology, Harvard Medical School and Massachusetts Eye and Ear Infirmary, Boston, Massachusetts
    Search for more papers by this author
    • T. Zhang and S. Trau contributed equally to this work.

  • Susan Tran,

    1. Department of Ophthalmology, Harvard Medical School and Massachusetts Eye and Ear Infirmary, Boston, Massachusetts
    Search for more papers by this author
    • T. Zhang and S. Trau contributed equally to this work.

  • Christopher Clouser,

    1. Department of Ophthalmology, Harvard Medical School and Massachusetts Eye and Ear Infirmary, Boston, Massachusetts
    Search for more papers by this author
  • Francesca Pignoni

    Corresponding author
    1. Department of Ophthalmology, Harvard Medical School and Massachusetts Eye and Ear Infirmary, Boston, Massachusetts
    • Department of Ophthalmology, Harvard Medical School and The Massachusetts Eye and Ear Infirmary, Boston, MA 02114
    Search for more papers by this author


Nicastrin is a component of the Notch signaling pathway involved in proteolytic release of the Notch receptor intracellular domain. It has been postulated that intracellular Notch is required within the nucleus of fly eye progenitor cells to enhance (proneural enhancement) and then repress (lateral inhibition) transcription of proneural genes. We present here an analysis of Nicastrin function during eye development and find that Nicastrin is essential to early photoreceptor neuron development. Nicastrin mutant tissue displays neuronal loss or hyperplasia; these phenotypes can be rescued by targeted expression of an intracellular form of Notch. Thus, nuclear translocation of Notch and its direct regulation of gene expression appear to be critical to proneural enhancement as well as lateral inhibition. In addition, we show that Nicastrin as well as Notch are required to maintain normal R-cell morphology, because the nuclei of mutant photoreceptor neurons cannot maintain their proper position. Thus, Notch signaling plays a role, not only in cell fate specification, but also in differentiation of photoreceptor neurons. Developmental Dynamics 234:590–601, 2005. © 2005 Wiley-Liss, Inc.


The Notch (N) signal pathway is conserved in metazoans and functions in a broad range of developmental processes (reviewed by Artavanis-Tsakonas et al.,1999). The Notch receptor is a large single-transmembrane glycoprotein that is cleaved at multiple sites upon ligand binding. The last cleavage occurs at an intramembranous site (S3 cleavage) and results in the release of the intracellular region of the protein. This domain then translocates to the nucleus, binds to the nuclear transcription factor Suppressor of Hairless (Su(H)), and regulates downstream target gene expression (reviewed by Lai,2004).

The S3 cleavage depends on a large protein complex called γ-secretase that was first identified in the processing of the amyloid precursor protein (APP) and later implicated in the processing of Notch (reviewed in Selkoe and Kopan,2003). The protease activity of the fly γ-secretase is provided by an eight transmembrane protein called Presenilin (Psn; Struhl and Greenwald,2001). Additional factors contribute to the activity of the γ-secretase, including a single transmembrane protein encoded by the nicastrin (nct) gene (Chung and Struhl,2001; Hu et al.,2002; Lopez-Schier and St. Johnston,2002). Studies in mammalian and Drosophila cells show that Nct and Psn regulate each other's stability and determine the activity of the γ-secretase (see reviews by Kopan and Goate,2002; Lai,2002; Takasugi et al.,2003).

Arguably the most studied role of the Notch pathway is its negative regulation of neurogenesis in a process called “lateral inhibition.” In the early stages of neurogenesis, a group of cells, called an equivalence group, acquire competence to differentiate as neurons and engage in reciprocal signaling through the Notch receptor. Because more cells than needed achieve this “proneural” state, a process of lateral inhibition occurs whereby one cell commits to the neuronal fate and inhibits surrounding cells from doing so (see Lai,2004). Modulation of Notch pathway activity is central to sorting out these competing fates: the acquisition of a neuronal fate requires low Notch activity, whereas the suppression of neuronal development requires high Notch activity.

In Drosophila, lateral inhibition through Notch signaling is required during development of both central and peripheral nervous system in embryos and in imaginal discs (progenitor epithelia giving rise to adult fly structures). Thus, loss-of-function mutations in several components of the pathway result in the so called “neurogenic” phenotype, leading to the differentiation of supernumerary neurons (see Lai,2004). A requirement for Nct function during lateral inhibition has been documented during wing and leg development (Chung and Struhl,2001; Lopez-Schier and St. Johnston,2002; Hu et al.,2002). Embryos lacking both maternal and zygotical nct display hyperplasia of the nervous system, and nct mutant leg or wing discs contain expanded clusters of sensory neurons. These phenotypes result from the failure of lateral inhibition and are also seen after loss of N, psn, or Su(H) function (Lecourtois and Schweisguth,1995; Ye et al.,1999; Struhl and Greenwald,1999,2001).

The role of the Notch pathway in eye development is considerably more complex (Baker et al.,1996; Baker and Yu,1997; Li and Baker,2001). By the end of the second larval stage (L2), a region of the eye–antennal discs epithelium, marked by the coexpression of the transcription factors Eyeless (Ey), Eyes absent (Eya), Sine oculis (So), and Dachshund (Dac), has acquired “eye identity.” Early in L3, additional transitions in cell specification or competence occur before overt differentiation of photoreceptor neurons. Eye progenitor cells transition first through a preproneural state characterized by the transient expression of hairy (h) and then into a proneural state marked by the induction of atonal (ato; Fig. 1A,B; Greenwood and Struhl,1999). It is at this stage that signaling by Notch enhances competence to acquire a neuronal fate by raising the expression level of the proneural protein Ato. Thus, the first role of the Notch pathway in the development of photoreceptor neurons is not to antagonize but to promote neurogenesis, a role referred to as “proneural enhancement” (Fig. 1B; Baker and Yu,1997; Baonza and Freeman,2001). Soon after, Notch signaling is required during lateral inhibition to limit the number of competent cells differentiating as R8, the founder neuron of the eye unit or ommatidium. Notch pathway activity during lateral inhibition results in loss of ato expression and suppression of neuronal development in cells surrounding the presumptive R8 neuron (Fig. 1B). Expression of Ato persists only in the R8 cell and results in the induction of Senseless (Sens), a transcription factor that promotes R8 cell identity (Figs. 1A, 2A; Frankfort et al.,2001). In this study, we use Sens expression as a molecular marker for Ato function and the establishment of R8 identity. Finally, additional cells are recruited to form a cluster of eight photoreceptor neurons (R-cells) marked by the expression of the pan-neural nuclear marker Elav.

Figure 1.

Morphogenesis of the photoreceptor neurons array in the L3 eye disc epithelium. A: Schematic diagram of eye disc development in L3. In this diagram and all stainings, discs are shown with posterior to the left and dorsal up. The antennal disc is continuous with the eye disc, but only eye discs development is addressed in this work. Early in L3, a depression or furrow appears at the center–posterior margin of the epithelium. This furrow marks the front of a wave of neuronal morphogenesis that sweep across the disc from posterior to anterior (arrow) and is called morphogenetic furrow (MF; bracket). Posterior to the MF, clusters of neurons corresponding to each eye unit (ommatidium) form, starting with the specification of the R8 neuron and followed by the progressive recruitment of R2–R5, R3–R4, R1–R6, and finally R7. Anterior to the MF, eye progenitor cells continue to proliferate and are marked by coexpression of Ey, Eya, So, and Dac. Expression of h, ato, and dpp is linked to the migrating MF: h is expressed ahead of the MF, ato, and dpp within the MF. Just posterior to the MF, ato expression becomes restricted to a single cell, fated to become the R8 neuron. In this cell, Ato induces expression of Sens. Staining with the pan-neural marker Elav highlights the assembly of neuronal clusters by progressive recruitment of photoreceptor neurons (R cells). B: Notch signaling is required multiple times during photoreceptor neurons development: first to establish competence to undergo neuronal development (proneural enhancement) and second to suppress neuronal development of cells surrounding the presumptive R8 neuron (lateral inhibition). Just ahead of the MF, cells are found in a preproneural state marked by the expression of H. These cells then become proneural, a state marked by down-regulation of h and induction of ato. It is at this stage that Notch signaling leads to further up-regulation of Ato expression (proneural enhancement). As cells emerge from the MF, signaling by Notch results in loss of Ato expression in all cells but the founder R8 neuron (lateral inhibition). Due to its diametrically opposite function in the first vs. the second step, lack of Notch function in proneural enhancement causes failure to initiate neurogenesis (no neurons form), whereas lack of Notch function in lateral inhibition causes failure to suppress neurogenesis (too many neurons form). C: Schematic drawing of a cross-section through an eye disc epithelium at the time of neurogenesis. Anterior to the MF, the nuclei of progenitor cells are found at various positions along the apical–basal axis. At the MF, cells drop their nucleus and constrict to generate the visible indentation on the surface of the epithelium. Posterior to the MF, as cells are recruited into a neuronal fate, their nuclei undergo stereotypical movements as they move first apically and then adjust to specific apical and subapical positions typical of each R-cell type within the ommatidium. For simplicity, only one of each pair of corresponding R-cells (2/5, R3/4, and R1/6) is shown. Dividing cells are shown in dark gray. Divisions are unpatterned anterior to the MF. Posterior to the MF, a single round of division occurs in the vicinity of the five-cell cluster (R2,3,4,5,8). In a mature neuronal cluster, i.e., a cluster contains all R-cells, the R8 nucleus occupies the lowest subapical position. Below the R8 lie the nuclei of uncommitted cells.

Figure 2.

ST96B mutations map to the nicastrin locus. Here and in all figures, discs are oriented posterior to the left and dorsal up. A: Late L3 wild-type and nctST1/nctST3 eye–antennal discs stained for Sens (green) and Elav (red). These proteins localize to the nucleus of the R8 cell (Sens) or to the nuclei of all R-neurons (Elav) in each cluster. In addition, Sens is also expressed in other sensory organ precursors (SOP) in the eye disc (e) and the antennal disc (a). In the nct mutant, Sens is seen in supernumerary sensory organ precursors within the antennal discs. In the eye disc, neither Sens nor Elav are present in the eye region, reflecting the lack of photoreceptor neurons formation. However, two expanded patches of sensory organ precursor cells (arrowheads) are present in the eye disc. These Sens-expressing cells arise from the peripodial cell layer (i.e., cells of the eye disc that does not give rise to eye photoreceptors). In the wild-type, these groups of sensory organs precursors are harder to see because they are composed of only a few cells. Notice also that the nct mutant eye disc is smaller that wild-type. This finding is due to a failure to grow during L3. This effect is linked to loss-of-Notch signaling (Dominguez and de Celis,1998; Papayannopoulos et al.,1998) and is also commonly observed in mutants that fail to initiate neurogenesis. B: Late L3 wild-type and nctST1/nctST3 mutant leg discs stained for Sens (green) and Elav (red). Staining is seen in supernumerary SOPs and neurons in the mutant disc, whereas it is restricted to fewer cells in the wild-type disc. C: Genetic mapping of the ST96B locus, as per flybase at The Df(3R)96ST was generated during this study by imprecise excision of a P-element in the CycB3 locus (marked by the green arrowhead). D: Schematic diagram of Nicastrin protein showing location of the transmembrane domain (TM) and sites mutated in the nct alleles. DNA and amino acid changes in the four alleles of nct isolated in this study are described below. The insertion (*) in the 5′ region of the ST4 allele was detected by polymerase chain reaction but was not characterized in detail.

All of these stages of eye development (eye specification, preproneural state, proneural state, neuronal fate specification, and neuronal differentiation) can be visualized within each single disc as eye progenitor cells are progressively recruited into the morphogenetic process beginning with the most posterior cells and ending with the most anterior progenitors (Fig. 1A). Moreover, cells transitioning through the proneural stage are morphologically distinguishable as they constrict their apical surface and drop their nuclei basally (Fig. 1C). This phenomenon results in a visible indentation within the epithelium traditionally referred to as the morphogenetic furrow (MF). Cells just anterior to the MF are in a preproneural state, whereas groups of cells posterior to the MF acquire a specific neuronal identity and initiate neuronal differentiation (Fig. 1A). Developing neurons undergo stereotypical nuclear movements as their nuclei move first apically and then adjust to specific apical and subapical positions typical of each R-cell type within the ommatidium (Fig. 1C).

We present here an analysis of nicastrin loss-of-function during these early stages of eye morphogenesis. We find that nct is indeed required for both proneural enhancement and lateral inhibition and that a dominant active receptor consisting of the intracellular domain of Notch (Nintra) rescues neurogenesis in nct mutant eye discs. These findings, in addition to the requirement for the psn encoded component of the γ-secretase (Li and Baker,2001), strongly support the view that proteolytic processing of Notch is essential in proneural enhancement as well as lateral inhibition within the eye.

In addition, we identify a requirement for nct function in the maintenance of proper cell morphology. We find that nct mutant cells that develop as neurons are unable to maintain proper position of their nucleus and show that this also occurs when Notch function is removed. These observations show that maintenance of nuclear position within photoreceptor neurons is a Notch-dependent process and potentially link this cell–cell signaling pathway to the regulation of microtubule-associated motors that control nuclear position in postmitotic neurons.


Gene ST96B Encodes Nicastrin

The mutant line ST96B was identified in a histological screen for mutants that affected early stages of eye development. Three additional mutant alleles were obtained by EMS mutagenesis, all three were pupal lethal over the original allele or as homozygous. The loss of function phenotype of all four alleles was characterized by a lack of neurogenesis within the eye disc. The late L3 eye–antennal epithelium shows a reduced eye disc with few, if any, neurons as assessed by expression of the R8 marker Sens and the pan-neural marker Elav (Fig. 2A). Antenna, wing and leg discs are also affected; however, contrary to the eye disc, a neurogenic phenotype is observed in these tissues. This finding is reflected in the presence of expanded groups of Sens-positive sensory organ precursor cells and supernumerary Elav-positive neurons (Fig. 2A,B and not shown).

The mutant locus mapped to the 96B region of the third chromosome by noncomplementation with the deficiencies Df(3)XTA1 (96A17-21; 96D1) and Df(3)96B (96A21; 96B11; Fig. 2C). In a three-point cross with a white+ P-element marker inserted in the 5′ region of the CycB3 gene and the proximal marker ebony (93C7-D1), we established that ST96B maps approximately 0.03 map units to the left of the P-element. Using the same P-element, we generated a small deficiency that removes genomic DNA from the 5′ end of the CycB3 gene to the 5′ end of the gene CG5807. This deficiency, Df(3)96ST, did not complement the ST96B mutation confirming its location to the left of the CycB3 gene (Fig. 2C).

A scan of the genes in the region indicated that the nicastrin gene fell in this genetic interval. Several observations suggested that the ST96B locus corresponded to nct. In fact, the ST96B mutant phenotype is consistent with a loss of Notch signaling. As mentioned in the Introduction section, mutations in the nct locus cause a neurogenic phenotype in leg and wing discs and lead to pupal lethality (Chung and Struhl,2001; Hu et al.,2002; Lopez-Schier and St. Johnston,2002). Sequencing of all four alleles obtained in this study revealed mutations within the nct coding region confirming the identity of Nicastrin and ST96B (Fig. 2D). Hence, we refer to our ST96B alleles as nctST1, nctST2, nctST3, and nctST4. The nctST1 and nctST3 alleles display equally severe mutant phenotypes over Df(3R)96ST or when homozygous. As mentioned above, nct encodes a single transmembrane domain protein. The nctST1 and nctST3 alleles are predicted to encode proteins truncated upstream of the transmembrane domain, with nctST3 generating the shortest truncation among nct alleles reported to date (Fig. 2D; Chung and Struhl,2001; Hu et al.,2002; Lopez-Schier and St. Johnston,2002). We consider nctST3 and nctST1 to be null alleles; the nctST2 and nctST4 alleles instead are hypomorphic but still strong alleles.

Nicastrin Mutant Eye Discs Are Arrested at the Preproneural to Proneural Transition

To better characterize the nct loss-of-function phenotype, we assayed for the expression of genes/proteins involved in various stages of eye development. We used (1) ey to assay for eye disc specification, (2) Eya and Dac to assay for eye primordium specification, (3) h as a marker of preproneural identity, and (4) ato to assay for proneural identity (Fig. 1A). In addition, we also assayed for the expression of decapentaplegic (dpp), a bone morphogenetic protein-4–type factor expressed along the posterior margin of the eye disc before MF formation and within the MF during the migration of the furrow across the epithelium.

Expression of ey was unchanged at the time of MF formation (early L3) and was still present in late L3 discs (not shown). Similarly, expression of Eya and Dac in early L3 discs was essentially indistinguishable from wild-type (Fig. 3A). In late L3 discs, expression of these factors was somewhat reduced but still detectable (Fig. 3A). Hence, eye discs from nct mutant larvae appear normal for eye disc and eye primordium specification.

Figure 3.

Eye development is arrested at the proneural to preproneural transition in nct mutant discs. Before the start of neurogenesis, the wild-type eye disc displays broad coexpression of Ey, Eya, and Dac marking eye progenitor cells, and expression of h and dpp along the posterior margin of the disc. During morphogenesis of the neuronal array, expression of Ey and Dac changes as both genes are down-regulated posterior to the morphogenetic furrow (MF): Ey sharply at the MF, Dac more gradually posterior to it. Eya continues to be broadly expressed anterior and posterior to the MF. Expression of h and dpp also changes at this time. Transcription of both genes, as well as the newly induced expression of ato, is linked to the progress of the MF across the epithelium: h ahead of the MF, dpp and ato within the MF. Arrowheads mark the MF in all wild-type L3 discs. Notice that all nct mutant late L3 discs are smaller that wild-type. This finding is expected, because proliferation of eye disc cells depends on Notch signaling (Dominguez and de Celis,1998; Papayannopoulos et al.,1998). Nonetheless, as shown here, this does not result in a general loss of gene expression (ey, Eya, Dac, Dpp, and H are all easily detected) but in the specific loss of ato expression. A: Expression of Eya and Dac in wild-type and in nct mutant discs. Top panels: early L3 nct mutant discs (nctST1/Df(3R)96B). Middle panels: late L3 wild-type discs. Bottom panels: late L3 nct mutant discs (nctST1/nctST3). Arrow in the wild-type Eya-stained disc (left middle panel) points to non–eye-related expression of Eya (expression marks the ocellar region of the disc). B: Expression of h and ato in wild-type and nctST1/nctST1 late L3 discs. The arrow in the wild-type ato expression panel (top right panel) points to expression is in the ocellar (non-eye) region of the disc. C: Expression of dpp-lacZ in early L3 nctST1/Df(3R)96B (top), late L3 wild-type (middle), and late L3 nctST1/nctST1 discs.

To establish whether nct mutant cells transitioned through the preproneural and proneural stages, we assayed discs for expression of h and ato. We found that mutant discs consistently show robust and expanded expression of the h gene (Fig. 3B). On the contrary, none of the mutant discs shows normal expression of the proneural gene ato. Most mutant discs display weak ato expression (Fig. 3B), and in some discs ato mRNA was not detected (not shown). Lastly, expression of dpp persisted along the posterior margin of the disc in a pattern typical of stages preceding MF movement (Fig. 3C).

In conclusion, the transition through a preproneural state takes place in nct mutant discs, whereas the transition from preproneural to proneural is severely impaired. This finding implicates Nct in the “proneural enhancement” step of photoreceptor neuron morphogenesis.

Nicastrin Mutant Clones Display Defects in Proneural Enhancement and Lateral Inhibition

To further evaluate the requirement for Nct function during photoreceptor neurons formation, we analyzed the phenotype in mutant clones. Clones homozygous mutant for the nctST1, nctST2, or nctST3 allele were generated in late L1 by the flippase-mediated recombination (FLP/FRT) method. Wild-type or Minute genetic backgrounds were used to obtain small or large mutant clones respectively (see Experimental Procedures section). Neuronal development and R8 formation were assayed by monitoring Elav or Sens expression. The mutant phenotype observed in large clones vs. small clones differed in some respects.

When several large mutant clones were present, few if any Elav-positive cells were observed in the eye disc, most often signs of neuronal development were purely restricted to nonmutant tissue (Fig. 4A). In some cases, groups of Elav-positive cells could be seen along the posterior edge of larger clones (Fig. 4A,B). This phenotype is analogous to the complete or nearly complete lack of neurogenesis observed in completely or largely homozygous mutant eye discs (Figs. 2A, 4A) and is consistent with the requirement for Notch signaling in the transition from preproneural to proneural.

Figure 4.

Proneural enhancement and lateral inhibition are impaired in nct mutant clones but can be rescued by expression of the intracellular Notch domain. All panels show discs at the late L3 stage. B–F show total projections from confocal microscopy, i.e., each panel reflects the entire stack of confocal images spanning one entire disc. A: Eye–antennal disc containing large nctST3 mutant clones stained for the neuronal marker Elav (red). Mutant tissue is marked by lack of GFP expression (green). Notice that, when large mutant clones are induced, eye disc size is reduced. This finding is expected, because signaling by Notch at the midline has been shown to nonautonomously induce proliferation across the entire eye disc (Dominguez and de Celis,1998; Papayannopoulos et al.,1998). Hence, whereas small clones do not interfere with eye disc proliferation, large clones do so when they extend across the midline, as in this example. B: Region of an L3 eye disc containing nctST2 mutant clones stained for Elav (red). Mutant tissue is marked by lack of GFP expression (green). Arrowheads mark supernumerary neurons present in the posterior portion of a mutant clone (failure of lateral inhibition). Further anterior (arrow) neurons are no longer formed (failure of proneural enhancement). C: Disc containing an nctST3 mutant clone stained for Sens (blue) to mark R8. Mutant tissue is marked by lack of GFP expression (green). Arrows point to regions of failed lateral inhibition where supernumerary Sens-positive cells are present. D: Rescue of neurogenesis in nctST1/nctST3 mutant discs by Nintra expression along the posterior region of the disc (dpp-Gal4 driver). All panels show R8 neurons, expressing Sens (green), and all photoreceptor neurons, expressing Elav (red). Genotypes are as follows: WT (one disc); UAS-Nintra/+, dpp-Gal4 nctST1/+ (dpp::Nintra, two discs); UAS-Nintra/+, dpp-Gal4 nctST1/nctST3 (dpp::Nintra nct mutant, two discs). Panels displaying staining of the same disc (Sens/Elav double, Elav single, and/or higher magnification) are joined by green fill. Arrowheads in dpp::Nintra nct mutant panels mark examples of clusters containing more than one Sens-expressing cell. As expected, expression of Nintra induces overproliferation in both wild-type and nct mutant discs. Disc enlargement was noticeably more pronounced on the ventral side than the dorsal side. This finding may reflect stronger expression of dpp-Gal on the ventral side or genetic differences along the dorsoventral axis. E: Disruption of cluster assembly and neuronal differentiation in GMR-Gal4 UAS-Nintra discs stained for Sens (green) and Elav (red). Initial expression of Sens (green) is largely undisturbed in these discs. However, Sens expression is lost completely soon after. Moreover, reduced numbers of Elav (red) -positive nuclei are seen where mature clusters should be present. F: Nts1 disc stained for Sens (green) and Elav (red). Discs shifted to 29°C shortly before MF formation retain sufficient Notch function to initiate neurogenesis. However, Notch pathway activity levels are insufficient for lateral inhibition. Hence, most cells begin to develop as R8 neurons and display expression of the R8 marker Sens.

Small mutant clones did not lack neurons. On the contrary, supernumerary Elav-positive cells were present (Fig. 4B). This phenotype is expected when proneural enhancement occurs but lateral inhibition fails. Consistent with this interpretation, extra R8 neurons marked by expression of the Sens transcription factor are present in these small clones (Fig. 4C). Thus, small Nicastrin mutant clones display the neurogenic phenotype associated with the lack of Notch signaling during lateral inhibition.

Activated Notch Rescues Proneural Enhancement and Partially Rescues Lateral Inhibition in Nicastrin Mutant Tissue

As a modulator of γ-secretase activity, Nicastrin is known to affect the proteolytic processing of the Notch receptor (Chung and Struhl,2001; Hu et al.,2002; Lopez-Schier and St. Johnston,2002). Therefore, the requirement for Nct function during proneural enhancement and lateral inhibition likely reflects a failure to release the Notch intracellular domain, thus preventing its nuclear translocation. To test this hypothesis, we chose to rescue the nct mutant phenotype by driving expression of a dominant receptor consisting of the intracellular domain of Notch, Nintra. This protein does not require proteolytic cleavage for activation and translocation to the nucleus (Fortini et al.,1993; Struhl and Adachi,1998). Nintra has been shown not only to rescue the lateral inhibition defect observed in psn and nct mutant wing discs but to also to suppress neurogenesis when expressed in these discs (Struhl and Greenwald,2001; Chung and Struhl,2001; Hu et al.,2002; Lopez-Schier and St. Johnston,2002). In addition, transient expression from an hs-Nintra transgene in a wild-type eye disc results in increased Ato expression during proneural enhancement but loss of Ato expression during lateral inhibition (Baker et al.,1996; Baker and Yu,1997).

The UAS/Gal4 binary system was used to induce expression of a UAS-Nintra transgene (Brand and Perrimon,1993). Under the control of the dpp-Gal4 driver, transgene expression is induced in a broad band along the posterior margin of the eye disc from early L2. Hence, in dpp-Gal4 UAS-Nintra discs, Nintra is already expressed before the start of neurogenesis.

We first looked at the effect of Nintra expression on eye neurogenesis in a wild-type genetic background. Discs expressing Nintra in the dpp-Gal4 pattern were larger than wild-type. This effect was expected, because Notch is known to promote proliferation in the early disc (Papayannopoulos et al.,1998; Dominguez and de Celis,1998; Keyon et al.,2003). To evaluate R8 development and cluster formation, we monitored expression of the Sens and Elav proteins. In dpp-Gal4 UAS-Nintra discs, cells initiating morphogenesis of the neuronal array displayed a quite well ordered pattern of Sens expression (Fig. 4D). Thus, the lateral inhibition process appeared to be effective in selecting single R8 neurons distributed at regular intervals.

Following this stripe of Sens-positive cells, a second region of relatively normal R-cells recruitment followed. Clusters included a single Sens-positive cell and many appeared to successfully reach the five-cell stage (Fig. 4D). The nearly normal pattern of Sens expression in dpp-Gal4 UAS-Nintra discs indicates that dpp-Gal4-driven expression of Nintra does not prevent proper R8 neuron specification in an otherwise wild-type genetic background.

Lastly, toward the posterior the disc, we observed a highly disorganized region where expression of both Elav and Sens was sporadic or completely absent. Because this most posterior region of the disc contains older, more mature clusters, the loss of Elav and Sens immunoreactivity suggests that Nintra is not well tolerated at the later stages of neuronal cluster formation. To confirm that this was indeed the case, we expressed Nintra specifically posterior to the MF, essentially after proneural enhancement and lateral inhibition occurred. Expression was induced using the GMR-Gal4 line. Under the control of this driver, high levels of responder transgene expression occur exclusively posterior to the MF. GMR-Gal4 UAS-Nintra discs displayed nearly normal selection of the R8 neuron as assessed by Sens expression but aberrant formation of neuronal clusters (Fig. 4E), with later (more posterior) neuronal clusters completely lacking Sens expression and displaying less Elav-positive nuclei than clusters closer to the MF. This result confirms that Nintra expression is in fact incompatible with the later stages of cluster development.

We next investigated the effect of restoring Notch signaling in an nct mutant background through Nintra expression. In dpp-Gal4 UAS-Nintra nctST1/nctST3 mutant discs, neurogenesis was indeed rescued and expression of Sens and Elav was restored (Fig. 4D). In the most anterior region of the developing neuronal array, many Sens-positive cells were present. However, these cells were not distributed in as regular an array as in wild-type or dpp-Gal4 UAS-Nintra discs. Moreover, more Sens-positive nuclei appeared to be present than expected. Posterior to this region, expression of Elav was present but disorganized and more than one Sens-positive cell could be seen in some clusters (Fig. 4D). Rescue of the proneural enhancement step is shown simply by the presence of substantial neuronal development in these mutant discs. Rescue of lateral inhibition is shown by the expression of neuronal markers in some but not all cells in the disc. In total absence of lateral inhibition, we would expect to see initial induction of Sens expression in nearly all cells of the disc, as seen in Nts1 discs maintained at a nonpermissive temperature (Fig. 4F). However, as described above, the resulting array of Sens-positive cells is not patterned in dpp-Gal4 UAS-Nintra nctST1/nctST3 discs as in wild-type, and many clusters display multiple Sens-positive nuclei; thus, lateral inhibition is rescued somewhat imperfectly.

In conclusion, Nintra expression significantly restores proneuronal enhancement and lateral inhibition in nct mutant tissue. The observation that Nintra is epistatic to nct indicates that the Nct protein functions in these developmental processes through the intracellular domain of Notch. This finding is consistent with the putative role for Nct in the processing of the Notch receptor and the postulated requirement for intracellular Notch during both proneural enhancement and lateral inhibition.

Nicastrin and Notch Mutant Clones Display a “Nuclear Drop” Phenotype

Surprisingly, in our analysis of nct mutant tissue we also observed an additional aberrant phenotype. When supernumerary neurons were present in nct mutant clones, many of the Elav-positive or Sens-positive nuclei were found in basal locations not normally occupied by photoreceptor neurons nuclei (Fig. 5A,B). Mislocalized nuclei were occasionally found at the very bottom of the disc epithelium where axonal tracts extend toward the posterior of the disc and into the optic stalk (Fig. 5A,B). Of interest, these “misplaced” nuclei were always observed in posterior regions of the clones, i.e., in more mature neurons.

Figure 5.

R-cells in nct and Notch mutant clones show mislocalization of their nuclei. A: L3 disc containing nctST2 mutant clones stained for Elav (red). Mutant tissue is marked by lack of green fluorescent protein (GFP) expression (green). Panels show projections of confocal images. Top panels: apical projection; bottom panels: basal projection. A total projection of this same disc can be seen in Figure 4B. Arrowheads point to a small clone with apically positioned R-cell nuclei. None of the Elav-positive nuclei are located basally. Arrows point to a clone with increased neurogenesis along the posterior margin, neurogenesis arrests, however, more anteriorly. Neuronal cells along the posterior edge of this clone are located farther away from the MF and, therefore, are at a more advanced stage of development than cells from the small clone. Many of the most posterior Elav-stained nuclei are found at ectopic basal positions. Occasionally, some nuclei can be seen to drop into the optic stalk, where axonal tracks extend toward their targets in the brain (asterisk). B: L3 disc containing nctST1 mutant clones stained for Elav (red). Mutant tissue is marked by lack of GFP expression (green). Top panel: apical projection; bottom panel: basal projection. Many of the basally located Elav-stained nuclei in this disc have dropped at the very bottom of the epithelium where R-cell axons extend toward the optic stalk. C: Region of an L3 disc containing small homozygous Notch54l9 mutant clones stained for Elav (red). Mutant tissue is marked by a lack of GFP expression. Separate total (all sections), apical, and basal projections are shown. Basally located nuclei are present along the posterior margin of each clone.

Because this phenotype had not been associated previously with lack of Notch signaling, we looked at the position of the nuclei in Notch mutant tissue. We observed essentially the same phenotype in tissue mutant for the null allele N54l9. Large clones showed few if any neurons; many small clones showed extra neuronal cells. As in the case of nct mutant tissue, basally located R-cell nuclei were found along the posterior edge of the clone (Fig. 5C). Hence, this “nuclear drop” phenotype is indeed associated with decreased Notch pathway activity.

To investigate whether this drop reflected a shift in position of the nucleus within the cell or a drop of the entire cell from the epithelium, we chose to assay for the localization of PATJ, a component of septate junctions, and Armadillo (Arm), a β-catenin found in cell–cell adherens junctions (Fig. 6). As both types of junctions are located on the apical side of the disc epithelium, these protein serve as apical cell polarity markers. Mislocalization of apical markers was occasionally seen in mutant tissue (Fig. 6A,C); however, it did not occur in every instance of nuclear drop (Fig. 6A–C). Hence, the drop of the nucleus in nct mutant photoreceptors does not reflect the loss of all R-cells from the epithelium, although the occasional mislocalization of junction proteins suggests that epithelial integrity sometimes may be compromised. Thus, the large majority of basally located Elav-positive or Sens-positive nuclei result from a drop of the nucleus within the cell and not a failure of cell–cell adhesion within the epithelium.

Figure 6.

PATJ and Armadillo/β-Catenin are apically localized in nct mutant clones. A: Region of an L3 disc containing nctST3 mutant tissue stained for Sens (red), to mark R8 nuclei, and PATJ (green), an apical polarity marker. Mutant tissue is marked by lack of green fluorescent protein (GFP) expression (blue). Panels show projections of confocal images. Left-most panels: total projections; a, apical projections; b, basal projections. Total projections are same as shown in Figure 4C and are provided here for side-by-side comparison. The asterisk in the bottom middle panel marks an example of failed lateral inhibition: multiple Sens-positive cells are in close proximity. Arrows mark the posterior region of a mutant clone containing supernumerary Sens-positive nuclei located in abnormally basal positions. Mutant Sens-positive nuclei are visible only in the total projections (left panels) and are absent in the apical projection panels (middle). Notice that nuclei in surrounding wild-type cells are visible in both total projections and in the apical projection panels. Right-most panels show apical (top) and basal (bottom) views of PATJ expression. Arrowheads in these panels point to a mutant clone showing normal apical localization of PATJ. Arrowheads point to occasional sites of PATJ immunoreactivity detected at basal locations within mutant clones. B: Cross-section of disc epithelium along the posterior boundary of an nctST3 mutant clone. Orientation of the epithelial layer shown is with apical side up and basal side down. The disc was stained for Elav (green) and PATJ (red). Mutant tissue is marked by lack of GFP expression (blue). Three neuronal clusters appear in this image. The left-most cluster is wild-type; middle and right-most clusters are mutant. PATJ (red) is apically localized in all three ommatidia; immunostaining for Elav (green) is found just below the apical membranes marked by PATJ in the wild-type ommatidium, whereas it is also detected basally in the two mutant ommatidia. C: Disc containing nctST3 mutant clones stained for Elav (green) and Armadillo/β-Catenin (red). Mutant tissue is marked by a lack of GFP expression (blue). Panels show projections of confocal images: a, apical projection; b, basal projection. Arrows mark two clones present in this portion of the disc. Armadillo staining in these clones appears to be less punctate or somewhat diffuse. However, immunoreactivity to the anti-Armadillo antibody is still mostly apical. Only occasionally, sites of staining can be found basally (arrowhead in lower right panel) within mutant tissue. R-cell nuclei are observed in ectopic basal positions in both clones.


The Notch signaling pathway plays a complex role in the development of the eye. Before L3, Notch is required to promote cell proliferation across the eye disc. During morphogenesis of the neuronal array, Notch signaling is first required for proneural enhancement, i.e., to promote neuronal development by enhancing expression of the proneural factor ato. Few hours later, Notch must signal a second time to antagonize neuronal development in cells other than the founder R8 neuron through a process called lateral inhibition. In this study, we have analyzed the requirement for Nicastrin, a component of the Notch pathway, in the development of the eye disc.

We find that Nct mutant tissue displays defects in proneural enhancement and lateral inhibition. These findings further support the view that Notch functions within the nucleus to promote as well as restrict neurogenesis in the eye. We also uncovered an additional requirement for Notch signaling in neuronal morphogenesis. Although Notch or nct function is not required for a cell to initiate neuronal development (after proneural enhancement), signaling through the Notch pathway is still necessary, as differentiation proceeds to maintain proper cell morphology. Thus, Notch or nct mutant photoreceptor neurons cannot maintain proper nuclear position, and their nuclei drop basally within the cell.

Nct Function Is Required for Proneural Enhancement and Lateral Inhibition During Eye Development

As previously reported, nct mutant animals die during pupation. Survival during embryogenesis, a stage also dependent on Notch pathway function, is due to maternally provided mRNA and/or protein (Hu et al.,2002; Lopez-Schier and St. Johnston,2002). Maternally provided Nct is likely responsible for normal proliferation of nct mutant eye discs through the L2 stage. Thus, early L3 mutant discs are morphologically indistinguishable from wild-type. Soon after, however, the lack of protein activity becomes apparent as few, if any, cells initiate expression of neuronal markers, and proliferation within the eye disc is severely impaired. Because robust expression of h but weak or no expression of ato is detected in these discs, the arrest occurs at the preproneural to proneural transition when proneural enhancement by Notch fails to occur (Greenwood and Struhl,1999; Li and Baker,2001; Baonza and Freeman,2001).

Defects in proliferation are not always detected during clonal analysis of Notch pathway components. This finding is expected, because Notch induces proliferation across the disc in a nonautonomous manner by signaling only along the dorsoventral midline (Dominguez and de Celis,1998; Papayannopoulos et al.,1998). Hence, as long as Notch or nct mutant clones do not span a significant portion of the disc midline, proliferation is unaffected. On the contrary, the requirement for Notch during proneural enhancement and lateral inhibition is cell-autonomous and, therefore, should be apparent in mutant clones depending on clone location relative to the MF and irrespective to their location relative the dorsoventral midline. Due to its diametrically opposite functions in the first vs. the second step, lack of Notch function in proneural enhancement should lead to failure to initiate neurogenesis, whereas lack of Notch function in lateral inhibition should lead to a neurogenic phenotype, i.e., more neurons.

These phenotypes were indeed observed in mutant clones. Failure to initiate neurogenesis was observed in large mutant clones, whereas a neurogenic phenotype indicative of a failure in lateral inhibition was observed in small mutant clones or along the posterior edge of large clones. This observation is also true of clones mutant for a null allele of the Notch receptor (Baker and Yu,1997; Li and Baker,2001; this work). As in the case of Notch (Baker and Yu,1997), we do not regard this neurogenic phenotype as indicative of non–cell-autonomy. In fact, lack of neurogenesis is observed along the lateral margins of large clones as well as in the center. This observation disproves the hypothesis that wild-type tissue might be sufficient to rescue neighboring mutant tissue. The most likely explanation is that the difference between large and small clones reflects variations in residual levels of pathway activity due to persistence of the mRNA and/or protein.

Several factors support this hypothesis. First, because mosaic clones are a product of somatic recombination, a founder mutant cell is initially genotypically mutant but phenotypically wild-type due to residual mRNA/protein carried over from the heterozygous progenitor cell during mitosis. Over time, decay of protein and mRNA as well as dilution through progressive rounds of cell division eventually leads to protein levels below a minimal functional threshold. Thus, mutant cells from larger clones would be expected to have lower levels of residual activity and display more severe phenotypes. Second, as proposed by Li and Baker (2001), proneural enhancement appears to require lower levels of Notch pathway activity than lateral inhibition. Thus, in small clones, pathway activity could be adequate for proneural enhancement but insufficient for lateral inhibition. Third, significant stability of nct mRNA and/or Nct protein is strongly suggested by the late (pupal) lethality of zygotic mutant animals as opposed to the embryonic lethality resulting from removal of maternal and zygotic Nct. Fourth, in Drosophila, somatic recombination is induced during the larval stages (around late L1), that is only ∼ 3 days before clone analysis in late L3. In conclusion, residual Notch pathway function is a likely explanation for the range of mutant phenotypes observed in nct mutant tissue.

Expression of the intracellular domain of N, Nintra, rescues the failure of neurogenesis observed in nct mutant discs. However, expression of this constitutively nuclear form of Notch (Fortini et al.,1993; Struhl and Adachi,1998) did not permit a proper execution of the lateral inhibition process. In fact, neuronal clusters were abnormally spaced and often more than one Sens-positive cell could be detected within one cluster. This effect cannot be attributed merely to the high level of nuclear Notch, because similar expression of Nintra in a wild-type genetic background did not significantly disrupt the selection of the R8 neuron (discussed below). Because lateral inhibition is based on achieving different levels of Notch pathway activity in neighboring cells (Fig. 1B), a simple explanation for this effects is that the comparable levels of nuclear Notch expression in neighboring cells may be responsible for the disruption in R8 cell selection in a nct mutant background. However, we cannot exclude the possibility that nct mutant discs also lack activity of one or more factors essential for proper lateral inhibition which have not yet been identified and are regulated through proteolytic cleavage by a γ-secretase.

In conclusion, Nct function is required during both proneural enhancement and lateral inhibition. The requirement for Nct (this work) as well as Psn (Li and Baker,2001) confirms that these early steps in neurogenesis are both dependent on γ-secretase activity. This evidence, together with the finding that expression of the intracellular domain of N, Nintra rescues the failure of neurogenesis observed in nct mutant discs, supports the model proposed by Li and Baker (2001) for photoreceptor neurons development. In this model, translocation of the Notch intracellular domain to the nucleus and its direct involvement in regulation of target gene expression are critical steps in proneural enhancement as well as during lateral inhibition.

Neurogenesis Can Tolerate Increased Levels of Nuclear Notch

In addition to showing rescue of nct loss-of-function, our Nintra experiments produced some unexpected results. Based on previously published data, increased levels of nuclear Notch should result in lack of R-cells formation due to excessive lateral inhibition. Defects in neurogenesis are apparent after a 30 min heat shock of hs-Nintra larvae (Baker et al.,1996). Within 90 min of heat shock, hs-Nintra discs show down-regulation of the second phase of Ato expression (Baker et al.,1996). Yet, in dpp-Gal4 UAS-Nintra discs, eye progenitor cells appear to tolerate the exogenous nuclear Notch protein and still generate a remarkably well-ordered array of R8 neurons. This finding is unlikely to reflect lack of Nintra activity because effects on disc proliferation and later stages of neuronal development are readily apparent, indicating that the Notch pathway is indeed hyperactivated in dpp-Gal4 UAS-Nintra discs.

A critical difference between our experiment and the one by Baker and colleagues lies in the mode of Nintra induction. In the experiments by Baker et al. (1996), expression of Nintra was induced transiently at a time when neuronal morphogenesis was in progress. As mentioned above, effects on ato expression were seen within 90 min of heat shock treatment. Hence, these experiments investigated the essentially immediate response to a relatively sudden increase in Notch signaling. On the contrary, in our experiment, Nintra expression was already established before the start of, and then maintained during, neurogenesis. The remarkably normal pattern of Sens expression observed in these discs suggests that eye progenitor cells can somehow “adapt and tolerate a background level” of Notch pathway activity. These cells can then still sense, and respond to, changes in signaling level due to stimulation of the endogenous receptor during lateral inhibition. However, persistently elevated levels of Notch signaling result in failure of neuronal development at later stages, because later steps in cluster assembly and neuronal differentiation following R8 selection are significantly impaired. These results emphasize the importance of differential activity of the Notch pathway, rather than absolute levels of signaling, as the critical factor underlying lateral inhibition.

Notch Function Is Required to Maintain Nuclear Position Within Postmitotic Photoreceptor Neurons

In addition to defects in proneural enhancement and lateral inhibition, nct mutant tissue displays a nuclear mislocalization phenotype not originally associated with Notch signaling. In the early stages of neurogenesis, the R-cell nuclei undergo stereotypical movements: moving first apically as they initiate to express neuronal markers, then adjusting to specific apical and subapical positions as the full complement of R-cells are added to each ommatidium (Fig. 1C). As shown here, in Notch or nct mutant clones, many Elav-positive nuclei are also found at ectopic basal locations.

We considered two possible explanations for this effect. Because this phenotype is associated with the induction of supernumerary neuronal cells, it may result simply from the late recruitment of more and more cells into a neuronal fate. Because uncommitted progenitor cells typically maintain their nuclei at basal positions, their late recruitment may simply cause a failure to adjust the position of their nucleus in relation to other R-cells within the cluster. Alternatively, maintenance of proper nuclear position in R-cells directly or indirectly requires the activity of a factor dependent on Notch signaling, at the MF or sometime after lateral inhibition.

Two observations favor the latter explanation. First, the appearance of Elav-positive nuclei in the basal region of the epithelium is often accompanied by a loss of Elav staining from the apical region (Fig. 5A). These “holes” in apical Elav staining are only observed when basally located Elav-positive nuclei are present. Second, basally located neuronal nuclei are also evident when specifically monitoring the position of the R8 nucleus (Fig. 6A). Because R8 is the first R-cell specified during cluster formation, the basally located R8 nuclei are unlikely to result from late recruitment of cells that maintain their lower nuclear position. For these reasons, we believe that the observed mutant phenotype reflects an actual drop of nuclei originally located apically within the cell.

Additional abnormalities in cell morphology, particularly mislocalization of cytoskeleton-associated molecules, have been linked to loss of the γ-secretase components. Psn function has been implicated in the apical localization of Armadillo/β-Catenin in fly embryos (Noll et al.,2000). In addition, Lopez-Schier and St. Johnston (2002) have reported that apical localization of β-Spectrin is disrupted in nct and psn mutant follicle cells. Can these effects be related to the nuclear drop phenotype described here? Two observations argue against this hypothesis. First, although we did observe mislocalization of Armadillo/β-Catenin as well as the apical marker PATJ in nct mutant clones, these phenotypes were observed in only a few mutant cells and did not correlate with most instances of nuclear drop (Fig. 6). Second, as reported by Noll et al. (2000) and Lopez-Schier and St. Johnston (2002), mislocalization of Armadillo/β-Catenin or β-Spectrin is not associated with loss of Notch function, suggesting that they reflect the activity of the γ-secretase on other substrates. As shown here, the nuclear drop phenotype is also seen in Notch mutant clones. Based on this evidence, abnormalities in Armadillo, PATJ, or β-Spectrin localization do not cause or precede the nuclear drop phenotype.

Lastly, is Notch signaling merely required for maintenance of proper nuclear position in R-cells or does it also induce the nuclear movements associated with the MF? In nct mutant clones, R-cells located just posterior to the MF display normal nuclear movements, with their nuclei in apical locations. Because the nuclear drop phenotype is only observed in more mature mutant cells, located some distance from the MF, signaling through Notch may not be required for the initial apical movement of R-cells nuclei but only for the maintenance of nuclear position in more mature differentiating neurons. In fact, the Notch pathway continues to be required during eye development posterior to the MF. Additional Notch signaling events occur after lateral inhibition and are essential for the recruitment of the correct number of R-neurons and accessory cells into each ommatidium (Cagan and Ready,1989; Parks et al.,1995). Hence, signaling by Notch in the later stages may also be involved in maintaining nuclear position in differentiating neurons. However, as discussed above, small Notch or nct mutant clones that display the neurogenic phenotype are likely to proceed through proneural enhancement thanks to residual Notch pathway activity. For this reason, we cannot excluded the possibility that low-level Notch signaling, still present in the early stages, is sufficient to support the nuclear movements associated with the MF and masks a requirement for Notch in inducing the nuclear movements. Due to its role in proneural enhancement, a possible function of Notch signaling in initiating nuclear movements at the MF cannot be readily addressed.

In conclusion, Notch and Nct function is required to maintain nuclear position within postmitotic photoreceptor neurons. Interestingly, recent studies have implicated the plus-end and minus-end directed microtubule motors Dynein and Kinesin as well as their regulators Dynactin and Klarsicht in the maintenance of nuclear position within photoreceptor cells (Patterson et al.,2004; Whited et al.,2004). How signaling at the cell membrane through the Notch receptor is linked to the balance of microtubule motors activity remains to be determined. Further studies of this critical question are complicated by the requirement for Notch signaling in nearly all cell fate decisions within the Drosophila eye and will require the development of reagents that specifically address Notch pathway function in nuclear positioning.



The nctST1 allele was identified in a histological screen of late lethal EMS-induced lines on the third chromosome (gift of Charles Zucker). The alleles nctST2, nctST3, and nctST4 were recovered from ∼5000 EMS-mutagenized flies screened for noncomplementation of the nctST1 allele. To identify the molecular lesions, all nct exons from each of the alleles were PCR amplified and sequenced. Genomic DNA was obtained from homozygous mutant larvae. Df(3R)96ST was obtained as a result of a local hopping experiment using the P element line P{lacW}CycB3L6540. The following fly lines were used in this study: dpp-lacZ, P{lacW}CycB3L6540, dpp-Gal4, pGMR-Gal4, Notch54l9, UAS-Nintra, Ubi-GFP FRT19A; ey-FLP(II), ey-FLP; FRT82B, Ubi-GFP, hs122-FLP; FRT82B Ubi-GFP, hs122-FLP; FRT82B Ubi-GFP RpS3 and are described at The alleles nctST1, nctST2, and nctST3 were recombined onto FRT82B carrying chromosomes and used to generate mutant clones by the FRT/FLP method (Xu and Rubin,1993). Similar phenotypes were observed in nct mutant clones induced with either ey-FLP or hs-FLP (heat shock treated in late L1 after 72 hr at 18°C and then incubated at 25°C). Mutant clones shown in Figures 4–6 were induced by ey-FLP, except for disc in Figure 4A, in which clones were induced by heat shock in a Minute background.


Larvae were dissected, and discs were stained for reporter gene (dpp-lacZ), protein (Elav, Eya, Dac, Sens) or mRNA (ey, h, ato) expression by standard protocols (Sullivan et al.,2000). Antibodies used were as follows: mouse monoclonal antibody (MAb) anti-Elav (1:1,000; Robinow and White,1991), rat MAb anti-Elav (1:20; O'Neill et al.,1994), mouse MAb anti-Dac (1:500; Mardon et al.,1994), mouse anti Eya mAb (1:100; Bonini et al.,1993), mouse MAb anti-LacZ (1:50; Goto and Hayashi,1999), rabbit anti-PATJ (1:2,000; Tanentzapf et al.,2000), mouse anti-Armadillo (1:100; Riggleman et al.,1990), guinea pig anti-Sens (1:1,000; Nolo et al.,2000). Anti-mouse, anti-rat, anti-rabbit, and anti-guinea pig secondary antibodies conjugated to Cy2, Cy3, and Cy5 (Jackson Immuno Research Laboratories), or horseradish peroxidase (Bio-Rad) were used at 1:200.


We thank Paul Garrity and Stephan Heller for comments on the manuscript; the Developmental Studies Hybridoma Bank (Iowa) and the Bloomington Stock Center (Indiana) for antibodies and fly stocks. DNA sequencing and confocal laser scanning microscopy were performed at the HMS/MEEI Ophthalmology Core Services supported by the NIH. F.P. was funded by a Basil O'Connor Starter Scholar Research Award and a NIH grant and is the recipient of a RPB-Career Development Award.