Endosomal entry regulates Notch receptor activation in Drosophila melanogaster

Notch signaling controls a myriad of proliferation and differentiation decisions across animal species, and aberrant Notch signaling is associated with cancers, stroke, and a host of human developmental disorders (Artavanis-Tsakonas et al., 1999). The potency of Notch signaling requires that its activity be under tight cellular regulation, which involves several posttranslational modifications including glycosylation, ubiquitination, and proteolysis. These modifications contribute to the sensitive control of spatiotemporal domains and the degree of activation seen in vivo. In canonical Notch signaling, the signaling-competent Notch receptor exists at the plasma membrane as a heterodimer consisting of the Notch extracellular domain (NECD) and the transmembrane and intracellular domains (Notch extracellular truncation [NEXT]). In signal-sending cells, activated ligands of the Delta, Serrate, Lag-2 (DSL) family bind to Notch. Upon ligand binding, the NECD is separated from NEXT, which then becomes an efficient substrate for an intramembrane cleavage mediated by the γ-secretase enzyme. This cleavage releases the cytoplasmic Notch intracellular domain (NICD) from its transmembrane anchor. Free NICD translocates to the nucleus, where it interacts with transcription factors to relieve transcriptional repression of Notch targets (for review see Schweisguth, 2004).

Because Notch and its ligands are transmembrane proteins, vesicular trafficking could influence signaling activity. Indeed, endocytic internalization and recycling of DSL ligands in the signal-sending cell are needed for productive signaling (for reviews see Le Borgne et al., 2005; Nichols et al., 2007). In contrast, the role of Notch receptor endocytosis in the signal-receiving cell is poorly understood and controversial. In vertebrates, it is unclear whether or not γ-secretase cleavage requires receptor internalization, with evidence supporting either scenario (Gupta-Rossi et al., 2004; Chyung et al., 2005; Kaether et al., 2006). In Drosophilamelanogaster, genetic analysis has suggested that Notch signaling requires internalization in the receiving cell; however, cleavage and activation of NEXT may not (Seugnet et al., 1997; Struhl and Adachi, 2000; Lopez-Schier and St Johnston, 2002). Recently, several D. melanogaster mutations that affect endocytosis after internalization have been described that appear to have disparate consequences for Notch activation, either decreasing, not affecting, or increasing signaling (Giebel and Wodarz, 2006). However, because of the varied approaches used in these studies, it is difficult to correlate the effects of Notch trafficking and signaling in a comparable manner. Hence, the significance of endocytic trafficking of the Notch receptor through various endosomal compartments during signaling remains an open question.

Here, we have used a set of D. melanogaster null mutations that disrupt representative and well-characterized core components of the endocytic machinery to systematically address this question by altering subsequent steps of endocytic trafficking and evaluating Notch localization, processing, and signaling output in vivo.

We first determined the effects of endocytic mutants on in vivo trafficking of the endogenous Notch receptor (see Fig. 1 A and the text below for a description of the mutants used). We generated homozygous mutant imaginal discs and performed a live trafficking assay to follow Notch endocytosis (see Materials and methods). In wild-type (WT) tissue, most surface-labeled Notch is internalized from the apical plasma membrane and degraded within 5 h (Fig. 1 B; Lu and Bilder, 2005; Vaccari and Bilder, 2005). In contrast, in each endocytic mutant analyzed, Notch is not degraded and instead accumulates in distinct subcellular compartments (Fig. 1, C–J), often resembling its steady-state localization seen using standard immunohistochemistry. We monitored Notch accumulation with respect to phalloidin, which marks the cortical actin cytoskeleton, and either Avalanche (Avl) or Hrs, which mark distinct domains of early endosomes (EEs). As expected, in mutants for proteins that regulate cargo entry into EE, such as a dominant-negative form of the dynamin orthologue Shibire (Shi), ShiDN (Moline et al., 1999), the GTPase Rab5, and the endocytic syntaxin Avl (Wucherpfennig et al., 2003; Lu and Bilder, 2005), Notch accumulates at or below the plasma membrane (Fig. 1, C–E). In mutants for Hrs, which is thought to mediate recruitment of ubiquitinated cargo in the EE (Lloyd et al., 2002), surface-labeled Notch reached and accumulated in an Avl-positive EE compartment (Fig. 1 F). In mutants for the endosomal sorting complex required for transport (ESCRT) genes tsg101, vps25, and vps20, which control MVB sorting (Moberg et al., 2005; Thompson et al., 2005;Vaccari and Bilder, 2005; Herz et al., 2006), almost all Notch was present in Hrs- but not Avl-positive puncta, apparently corresponding to a more mature EE compartment (Fig. 1, G–I; and not depicted). Finally, in mutants for proteins that act in a post-MVB, prelysosomal compartment, such as the PIKFYVE orthologue Fab1 (Rusten et al., 2006), Notch accumulated in a structure that is distinct from EEs, as it did not colocalize with either Avl or Hrs (Fig. 1 J and not depicted). The site of Notch trapping in each mutant is consistent with the predicted function of the disrupted endocytic regulator and therefore provides an opportunity to correlate the endocytic localization of Notch with its associated signaling activity.

To investigate the effects of Notch accumulation on receptor activation, we examined expression of mβ–lacZ, a transcriptional reporter of Notch signaling (Fig. 2).mβ–lacZ can be reliably assayed in mutant imaginal discs, even in shi, avl, Rab5, and ESCRT mutants that behave as neoplastic tumor suppressor genes. In the latter mutants, epithelial organization is lost and tissue overgrows, but these cell-autonomous phenotypes are independent of Notch signaling (Lu and Bilder, 2005; Vaccari and Bilder, 2005). As compared with the stereotyped levels and patterns seen in WT discs (Fig. 2, A and J; Nellesen et al., 1999), expression is uniformly reduced in shiDN-expressing disc cells and also in avl and Rab5 mutant discs (Fig. 2, K, B, and C; Lu and Bilder, 2005). In contrast, the level of mβ–lacZ in hrs and fab1 mutant discs was unchanged compared with the WT (Fig. 2, D and H). Finally, expression of mβ–lacZ is elevated in ESCRT-I, -II, and -III mutants (Fig. 2, E–G). In light of our Notch localization data, these results suggest a correlation between Notch endosomal access and its degree of signaling activation.

Because loss-of-function mutants across the endocytic panel indicate that access of Notch to endosomes and MVBs correlates with signaling activation, we tested this idea by using a gain-of-function approach. We overexpressed dominant active Rab5 (Rab5DA), which enhances endosomal fusion (Entchev et al., 2000), as well as Hrs, which enhances MVB transport (Seto and Bellen, 2006), in the dorsal compartment of the wing disc and monitored Notch signaling.

As expected from our analysis of Rab5 mutant eye discs, wing discs that express a dominant-negative form of Rab5 (Rab5DN) that inhibits endosomal fusion (Entchev et al., 2000) show reduced mβ–lacZ, mimicking the disruption of Shi function (Fig. 2, K and L). In striking contrast, wing discs that express Rab5DA show strong up-regulation of mβ–lacZ (Fig. 2 M). We obtained similar results, albeit with less potent enhancement of mβ–lacZ upon Hrs overexpression (Fig. 2 N). Consistent with enhanced endosomal fusion in Rab5DA- and Hrs-expressing discs, Notch localization is mostly endosomal (Fig. 2 P and not depicted), whereas Notch is largely cortical in Rab5DN-expressing discs (Fig. 2 O). Collectively, with our endosomal mutant studies, these ectopic expression studies provide evidence for a sharp transition in signal potentiation between the “active” and “inactive” states that depends critically upon Notch entry into endosomes.

A potential explanation for the signaling profile observed in endocytic mutant tissue would assign the differences to trafficking of DSL ligands for Notch rather than the receptor itself. To specifically assess the role of Notch receptor trafficking in signaling activation, we used D. melanogaster ovaries. In these organs, the Notch ligand Delta is up-regulated in the germ cells during stages 6–7 of oogenesis to activate Notch signaling in the adjacent somatic follicle cells (FCs; Fig. 3 A; Deng et al., 2001; Lopez-Schier and St Johnston, 2001). This signaling up-regulates the transcription factor Hindsight (Hnt) and simultaneously down-regulates Cut (Fig. 3 B; Sun and Deng, 2005, 2007) We made genetic mosaic ovaries in which some FCs were deficient for the function of shi, Rab5, hrs, or the ESCRT component tsg101 (Fig. 3, C–F). In these ovaries, internalization of Notch in certain signal-receiving cells is altered but internalization of Delta in the signal-sending cell remains normal. We found a consistent reduction of Hnt expression at stage 6 in FCs expressing shiDN or mutant for Rab5 (Fig. 3, C and D). In contrast, we found no reduction in hrs and tsg101 mutant FCs (Fig. 3, E and F). Reciprocal results were obtained analyzing Cut, and the differences were not caused by disrupted cell polarity because Notch signaling was unaffected in comparable FCs mutant for the nonendocytic neoplastic tumor suppressor gene lgl (Fig. S1; Bilder et al., 2000).

Interestingly, we observed precocious Hnt expression in tsg101 mutant FCs at stage 5 (Fig. 3 F), which suggests that signaling activation in ESCRT mutant FCs is ligand independent. Ligand-independent Notch activation has been demonstrated in cultured cells depleted of ESCRT components by RNAi (Thompson et al., 2005). To test ligand involvement directly in vivo, we assayed mβ–lacZ in imaginal disc cells lacking both the ESCRT component Tsg101 and the Epsin homologue Liquid facets (Lqf), which is required for ligand-dependent Notch signaling (Overstreet et al., 2003; Wang and Struhl, 2004). In lqf tsg101 double mutant discs, mβ–lacZ expression was unchanged compared with that observed in tsg101 mutant discs (Fig. 2 I). Collectively, these data demonstrate that the alterations in Notch signaling in endocytic mutant FCs parallel those seen in imaginal discs and indicate that endosomal entry of Notch itself is required for efficient signaling in epithelia.

How could the entry of the Notch receptor into endosomes promote efficient signaling? Because signaling depends on intramembrane cleavage of Notch by γ-secretase, we asked whether altered Notch endosomal transport might affect its cleavage. We directly measured the amount of Notch in endocytic mutant tissue that can be cleaved by γ-secretase by using an assay that induces ligand-independent Notch cleavage in tissue extracts (see Materials and methods). Notch cleavage efficiency ex vivo is measured in Western blots by quantifying the amount of the lower band (corresponding to the γ-secretase cleavage product NICD) relative to the upper band (corresponding to its immediate precursor, the membrane-anchored γ-secretase substrate NEXT) of the ∼120-kD doublet recognized by an antibody against an NICD epitope. As expected, generation of free NICD in this assay is completely blocked by treatment with the γ-secretase inhibitor N-(N-[3,5-difluorophenacetyl-l-alanyl])-S-phenylglycine t-butyl ester (DAPT; Fig. 4 A, lanes 2 and 6). Strikingly, NICD generation is strongly reduced in avl mutant extracts compared with WT extracts, although it is not completely abolished (Fig. 4 A; lane 3), paralleling the reduced activation profile observed in situ with the mβ–lacZ reporter. To assess whether this reduction in γ-secretase cleavage is specific to avl or common to mutants that block endosomal entry, we tested shiDN extracts and found that they are also severely impaired (Fig. 4 B, lane 7). In contrast, no significant difference in cleavage was observed in extracts from vps25 tissue in which, as in WT tissue, Notch can traffic to endosomes (Fig. 4 A, lane 4). This result is in apparent contrast with the elevation of Notch signaling observed in vps25 mutants in vivo. However, elevated signaling is likely induced by persistence of trapped Notch in cleavage-proficient endosomes, and assessment of this increased cleavage over time is precluded by the ex vivo nature of the assay. These data indicate that access of Notch to endosomes correlates with increased ability to be cleaved by γ-secretase.

Although both Notch cleavage and signaling are substantially reduced in mutants that block endosomal entry, they are not completely abolished. We hypothesized that alternative low-efficiency Notch trafficking routes could account for such residual signaling. To test this hypothesis, we reasoned that any population of Notch that was internalized in endosomal entry mutants would subsequently accumulate and fail to be degraded in the absence of ESCRT activity. We therefore generated avl tsg101 double mutant discs and found that, in contrast to avl discs, avl tsg101 double mutant discs display some Notch immunoreactivity in Hrs-positive intracellular puncta that resemble those found in tsg101 discs (Fig. 5, A, D, and G). This indicates that a small but significant amount of Notch can enter endosomes independent of avl function. We further found that avl tsg101 discs display a level of mβ–lacZ intermediate between avl and tsg101 alone (Fig. 5 H). Finally, to confirm this increase in Notch activation levels, we examined adult eyes resulting from genetically mosaic, rather than entirely mutant, eye discs. ESCRT cells in mosaic discs promote Notch-dependent overgrowth and give rise to bulging adult eyes (Fig. 5 F; Moberg et al., 2005; Thompson et al., 2005; Vaccari and Bilder, 2005; Herz et al., 2006), whereas adult eyes developing from avl mosaic discs are morphologically normal (Fig. 5 C). In contrast, adult eyes developing from avl tsg101 mosaic discs display mild bulging, which is consistent with intermediate levels of Notch signaling hyperactivation (Fig. 5 I). These data suggest that more than one Notch internalization route exists and may account for residual Notch activation seen in mutants that impair cargo recruitment into endosomes.

Our findings place Notch among the growing number of signaling pathways in which activity is stimulated, rather than attenuated, by receptor endocytosis. Examples of such pathways that have been studied in D. melanogaster include Dpp and Wingless (Wg) signaling (Fischer et al., 2006). Although the effects of endocytosis on Wg and Dpp signaling appear complex, transport to endosomes of Wg- and Dpp-bound receptors can, in some circumstances, enhance activation of the respective pathways (Entchev et al., 2000; Piddini et al., 2005; Marois et al., 2006; Rives et al., 2006; Seto and Bellen, 2006). Interestingly, at least in the case of Wg receptors, signaling appears to be reduced by stimulation of MVB transport (Seto and Bellen, 2006), an effect opposite to the enhancement observed for Notch signaling under similar conditions. This example highlights how each step of endocytic traffic can exert distinct regulatory effects on different activated receptors and emphasizes the importance of defining the influence of endocytic trafficking on all receptor-mediated cell signaling pathways.

Previous studies have concluded that endocytosis of the activated NEXT fragment of Notch is not a requirement for signaling. For example, Struhl and Adachi (2000) showed that an overexpressed NEXT-like construct can be cleaved in shi mutant D. melanogaster embryos. We have performed analogous experiments in imaginal discs and also find that NEXT overexpression induces activation when shi-dependent internalization is disrupted (Fig. S2). We favor the hypothesis that the difference between these experiments and the data presented above involves the use of an overexpressed Notch mimic instead of endogenous Notch. Our data, which point to the existence of more than one route to deliver Notch to endosomes, can reconcile and clarify the above findings: alternative routes may permit sufficient internalization of overexpressed protein to account for the activation observed. A second, not mutually exclusive possibility is that a low and normally subthreshold degree of γ-secretase cleavage of NEXT at the plasma membrane is amplified by overexpression. Consistent with either scenario, in a study of bristle patterning, activation of Notch via overexpression did not require shi but activation of Notch via a missense mutation in the endogenous protein was shi dependent (Seugnet et al., 1997).

How might endosomal entry promote Notch signaling? The strong correlation between γ-secretase cleavage efficiency and the degree of Notch signaling observed in vivo, seen in both loss- and gain-of-function manipulations that alter endosomal traffic, leads us to propose that in WT D. melanogaster cells, γ-secretase cleavage occurs most efficiently in endosomes. This proposition is consistent with the fact that γ-secretase can act in endosomes and has optimal activity at low pH, typical of endocytic organelles (Lah and Levey, 2000; Pasternak et al., 2003; Gupta-Rossi et al., 2004; Urra et al., 2007). Unliganded Notch, which is continually internalized, is normally rapidly transported through such endosomes; thus, the ligand-independent activation seen in ESCRT mutants could result from amplification of a subthreshold propensity of heterodimeric Notch to be cleaved because of altered endocytic kinetics and endosomal accumulation. Alternatively, considering that the ionic environment can affect shedding of the inhibitory extracellular portion of Notch (Rand et al., 2000), ESCRT mutant endosomes might possess a lumenal environment that weakens Notch heterodimeric interactions, leading to ligand-independent cleavage. Future investigation will test the relevance of ligand-independent activation in ESCRT mutants to physiological Notch signaling.

Our finding of an endosomal trafficking route that promotes ligand-independent Notch activation may be relevant to human neoplasias. A subset of leukemia-associated mutations display γ-secretase–dependent Notch hyperactivation and render Notch sensitive to ligand-independent cleavage by weakening the heterodimerization of Notch N- and C-terminal fragments (Malecki et al., 2006). Several γ-secretase inhibitors are currently under evaluation as potential therapeutic agents for Notch-regulated cancers (Kogoshi et al., 2007). The fact that endosomal trafficking of Notch appears closely tied to the γ-secretase–mediated cleavage step raises the possibility that endosomal routing pathways might offer additional prospective targets for clinical intervention.

The avl¹ (Lu and Bilder, 2005), lqf^l71 (provided from J. Fisher, University of Texas, Austin, TX), and tsg101²(provided by K. Moberg, Emory University School of Medicine, Atlanta, GA) alleles were used to generate recombinant chromosomes. Other chromosomes used include Rab5² UAS-Rab5DA and UAS-Rab5DN (provided by M. González-Gaitán, Département de Biochimie, Sciences II, Genève, Switzerland), hrs^D28 and UAS-Hrs (provided from H. Bellen, Baylor College of Medicine, Houston, TX), vps25^A3 (Vaccari and Bilder, 2005), fab-1^2-1 (provided by T.E. Rusten, University of Oslo, Oslo, Norway) and vps20^I3 (Vaccari and Bilder, 2005), lgl⁴ (Bilder et al., 2000), E(spl)mß-LacZ (provided by J. Posakony, University of California, San Diego, La Jolla, CA), UAS-shiK44A (Bloomington Stock Center), UAS-NotchFL, UAS-NICD (provided from G. Struhl, Columbia University, New York, NY), and MS1096 GAL4 (Bloomington Stock Center). UAS-NEXT expresses a protein consisting of the first 25 amino acids of Notch (including signal sequence) followed by an HA epitope tag, followed by Notch sequences from AAKHQ located ∼35 amino acids N terminal to the transmembrane domain to QQLGG in the intracellular domain, and ending with C-terminal V5 and His tags. Follicle cell clones were generated as described by Bilder et al. (2000). Mosaic eyes were generated as described by Tapon et al., (2001), whereas completely mutant eye discs (referred to in the text as mutant discs) were generated as described by Newsome et al., (2000).

Ovaries and imaginal disc tissues were fixed and stained under standard conditions with rhodamine-phalloidin and primary antibodies against the following antigens: Avl (Lu and Bilder, 2005), Hrs (Lloyd et al., 2002), β-gal (Capell) and NECD, NICD, Wg, Hnt (Developmental Studies Hybridoma Bank). Notch endocytosis assays were conducted as described by Vaccari and Bilder (2005). Secondary antibodies were obtained from Invitrogen. In Fig. 1, subapical cross-sections of epithelium anterior to the morphogenetic furrow (Fig. 1, B, F, and J) or in the wing pouch (C) are shown; other panels show equivalent sections, although these mutant cells with altered polarity do not display a clear apical side. All images are single confocal sections taken at room temperature with a microscope (TCS; Leica) using 16× NA 0.5 or 63× NA 1.4 oil lenses (Leica). Images were edited using Photoshop 7.0 (Adobe) and assembled with Illustrator 10 (Adobe).

For Notch cleavage assays, postnuclear extracts from WT and mutant discs were prepared using a hypotonic lysis buffer containing 20 mM Tris-HCl, pH 7.5, 10 mM KCl, 1 mM EDTA, and Complete protease inhibitor (Roche; Hu et al., 2002). The low salt concentration and absence of detergents promotes preservation of in vivo compartmental localization of transmembrane proteins, whereas EDTA chelates Ca²⁺ and thus promotes ligand-independent Notch processing (Rand et al., 2000). To test γ-secretase cleavage of Notch after harvesting of discs, extracts were incubated ex vivo for 20 min in hypotonic buffer alone or supplemented with 100 μM DAPT to inhibit the γ-secretase enzyme. Western blots were probed with both anti-NICD and anti–β-tubulin (Sigma-Aldrich). Quantification in Fig. 4 C and Fig. 4 D, based on three and four independent experiments, respectively, was obtained by calculating of the intensity ratio of free NICD (L) band to uncleaved NEXT (U) band normalized to WT (normalized L/U). The distinct L/U ratios seen between WT eye and wing disc extracts are consistent and may reflect intrinsic tissue-specific differences in Notch processing. Statistical analysis was performed by applying a Student's t test on the quantified dataset. Image acquisition and quantification was performed with an imaging system (Odyssey; Li-Cor Biosciences).

Fig. S1 shows that the loss of Notch signaling in endocytic mutant FCs is not an indirect effect of cell polarity alterations. Fig. S2 shows that Notch ligand control can be bypassed by overexpression of a truncated Notch mimic in ShiDN-expressing cells. Fig. S3 shows GFP expression controls for the endocytic mutant FC clones shown in Fig. 3.

We are grateful to the individuals listed in Materials and methods for providing reagents. We apologize to colleagues whose work could only be cited in review due to space constraints.

This work was supported by an American Heart Association award (0625181Y) to T. Vaccari, the Intramural Research Program of the National Institutes of Health, the National Cancer Institute, a grant from the Center for Cancer Research (contract 15446611) to M. Fortini, and grants from the National Institutes of Health (RO1 GM068675-01) and the American Chemical Society (FSG-07-040-01-CSM) to D. Bilder.