Research ArticlesNotch Signaling

Ligand-activated Notch undergoes DTX4-mediated ubiquitylation and bilateral endocytosis before ADAM10 processing

See allHide authors and affiliations

Sci. Signal.  13 Jun 2017:
Vol. 10, Issue 483, eaag2989
DOI: 10.1126/scisignal.aag2989

Ordering the steps to Notch activation

Signaling by the receptor Notch enables cell position to influence cell fate through information conveyed between neighboring cells. Binding of ligand, which is a transmembrane protein on a cell that neighbors the Notch-bearing cell, stimulates Notch cleavage in multiple stages. One cleavage event is mediated by an ADAM family protease and is followed by the cleavage event that generates the biologically active Notch receptor component in the receptor-bearing cell. Chastagner et al. found that ubiquitylation of Notch1 by the E3 ubiquitin ligase DTX4 promoted its internalization in response to ligand binding. Furthermore, ubiquitylation preceded the processing of Notch1 by ADAM10 in an endosomal compartment, and not at the cell surface as is usual for ADAM10 substrates. These results show that the endocytosis of the ligand-bound Notch1 by the ligand-bearing cell is not necessary for the endocytosis of Notch1 by the receptor-bearing cell, as had been previously thought, but rather that both of these endocytosis events are triggered by Notch1 ubiquitylation in the receptor-bearing cell. Because these results also show that the processing of Notch1 by ADAM10 occurs intracellularly, they suggest that it may be necessary for ADAM10 inhibitors to be cell-penetrant to be effective in treating cancers associated with constitutively activated Notch.


The Notch signaling pathway, which is activated by cell-cell contact, is a major regulator of cell fate decisions. Mammalian Notch1 is present at the cell surface as a heterodimer of the Notch extracellular domain associated with the transmembrane and intracellular domains. After ligand binding, Notch undergoes proteolysis, releasing the Notch intracellular domain (NICD) that regulates gene expression. We monitored the early steps of activation with biochemical analysis, immunofluorescence analysis, and live-cell imaging of Notch1-expressing cells. We found that, upon ligand binding, Notch1 at the cell surface was ubiquitylated by the E3 ubiquitin ligase DTX4. This ubiquitylation event led to the internalization of the Notch1 extracellular domain by the ligand-expressing cell and the internalization of the membrane-anchored fragment of Notch1 and DTX4 by the Notch1-expressing cell, which we referred to as bilateral endocytosis. ADAM10 generates a cleavage product of Notch that is necessary for the formation of the NICD, which has been thought to occur at the cell surface. However, we found that blocking dynamin-mediated endocytosis of Notch1 and DTX4 reduced the colocalization of Notch1 with ADAM10 and the formation of the ADAM10-generated cleavage product of Notch1, suggesting that ADAM10 functions in an intracellular compartment to process Notch. Thus, this study suggests that a specific pool of ADAM10 acts on Notch in an endocytic compartment, rather than at the cell surface.


The Notch signaling pathway plays multiple roles in organisms ranging from worms to mammals, both during development and in adulthood. Deregulation of Notch signaling is associated with many cancers, including leukemia and solid cancers (1), and also with various inherited and acquired diseases (2). Notch activation relies on the contact between neighboring cells, which enables the binding in trans of a membrane-anchored ligand to an adjacent receptor. Drosophila melanogaster has one Notch and two ligands; mammals have four members of the Notch family of receptors and five ligands, yet the activation mechanisms are evolutionarily conserved. Mammalian Notch consists of a transmembrane heterodimeric protein resulting from furin cleavage in the trans-Golgi network (3), the extracellular part of the heterodimer that is attached to the membrane-anchored and intracellular domains through noncovalent linkages. Upon ligand binding to Notch, the first step is the “transendocytosis” event, which creates a mechanical force that dissociates Notch heterodimer and promotes a conformational change (46). The ligand-bound Notch extracellular domain is internalized into the ligand-expressing cell, whereas the remaining portion of the receptor becomes a substrate for a disintegrin and metalloproteinase (ADAM), ADAM10 or ADAM17, which cleaves Notch in the residual extracellular portion 12 amino acids from the transmembrane domain (711). Notch is eventually processed by the γ-secretase activity, and the resulting soluble Notch intracellular domain (NICD) (12) migrates to the nucleus, where it has transcriptional and nontranscriptional functions (13). Despite the progress in the understanding of the molecular mechanisms of this pathway, various questions remain open. What are the first events that occur in the Notch-expressing cell upon activation? What are the accessory factors at these early steps? When and to what cellular compartment is ADAM recruited to Notch?

One of the first genes identified in D. melanogaster as an actor of the Notch pathway was deltex (dx) (14, 15). Mutations in dx produced phenotypes similar to those resulting from loss-of-function mutations in Notch, suggesting that Dx is an activator of Notch signaling. However, the specific role of Dx in fly or its homolog in other organisms has remained controversial, because the effect of inhibiting or mutating Dx varies depending on context (16). Drosophila Dx has been implicated in mediating Notch early endocytosis and trafficking to the lysosomes in the presence or absence of ligand (1721). In mammals, there are five Dx-related proteins (DTX1 to DTX4 and a distantly related protein DTX3L). DTX3 and DTX3L lack the N terminus containing the two WWE (named after the amino acids) modules that mediate the direct interaction with Notch. All DTX family members have a highly conserved C-terminal E3 ubiquitin ligase domain of the RING (really interesting new gene) family (22, 23). In addition to their putative roles in Notch signaling, a few have been described in other processes, such as degradation of TBK1 and MEKK1 or endosomal sorting of CXCR4 (2426).

With the goal of identifying the DTX isoform that plays a role in mammalian Notch signaling, we unraveled the initial steps of activation of mammalian Notch1 (N1). We visualized these early steps and identified a DTX4-dependent ubiquitylation event that contributed to endocytosis of the membrane-anchored subunit of N1 into the receptor-expressing cell, concomitantly to the transendocytosis of the ligand-bound subunit of N1 into the ligand-expressing cell. Finally, we showed that ADAM10 cleaved N1 after the receptor was internalized into the receptor-expressing cell.


DTX4 enhances N1 activation

To monitor N1 activation, we cocultured murine OP9 cells expressing the ligand Dll1 (OP9-Dll1) with human U2OS cells expressing N1 (U2OS-N1) under conditions in which the cells made contact (see Table 1 for a list of the cell lines and fig. S1 for a diagram of the proteins). We chose this coculture system because U2OS-N1 and OP9-Dll1 stable cell lines express N1 and Dll1, respectively, which are the most characterized forms of receptors and ligands of the family and which both harbor extracellular epitope tags in these cell lines [hemagglutinin (HA) and vesicular stomatitis virus (VSV), respectively; fig. S1]. The use of these cell lines allowed us to overcome the lack of good extracellular antibodies by epitope staining. Although U2OS-N1 cells were obtained after retroviral transduction (27), the amount of furin-processed N1 [transmembrane and intracellular domain (TM-IC)] was within an order of magnitude of that in the nontransduced U2OS cells, and the transcript abundance in U2OS-N1 cells was only twofold greater than that in diffuse large B cell lymphoma cells, indicating that the amount of N1 was likely not substantially different from that in other cells (fig. S2, A and B).

Table 1 Cell lines used for analysis.

In some experiments, additional proteins were transiently expressed as indicated by a “+” and the name of the transiently expressed protein. The HA-tagged form of N1 (N1-HA) had the HA tag after the epidermal growth factor (EGF) repeat 22, which was detectable without permeabilization, but did not interfere with activation. The fusion proteins between DTX4 and GFP or Cherry had the fluorescent protein fused to the C terminus of DTX4. N1 fused to GFP had the GFP fused to the amino acid 2097 of N1. Dll1 fused to the VSV tag had the tag on the amino acid 45, which was exposed extracellularly. See fig. S1 for diagrams of the proteins.

View this table:

To investigate which DTX isoforms were involved in N1 activation, we transfected the U2OS-N1 cells with small interfering RNAs (siRNAs) targeting each of the DTX isoforms that contain the WWE domain, which mediates the interaction with Notch (23), and then monitored Notch transcriptional activity by quantitative reverse transcription polymerase chain reaction (qRT-PCR) of an endogenous target, Hey1, which was the best responding gene in this coculture system (Fig. 1A, left). Knockdown efficiency and specificity at the level of DTX transcripts were verified for the siRNAs (Fig. 1A, right). Contact between cocultured U2OS-N1 and OP9-Dll1 cells stimulated Hey1 transcription. Exposing the cells to the γ-secretase inhibitor DAPT blocked N1-mediated Hey1 expression. Knockdown of DTX4 reduced Hey1 transcript abundance compared with that in the control cells or the cells in which DTX1 or DTX2 was knocked down. Silencing the three DTXs had no additional effect on Hey1 expression, suggesting that DTX4 was the major DTX involved in N1 activation in this cell culture model. To control for off-target effects, we tested another siRNA targeting DTX4 using a reporter system based on a synthetic Notch-regulated promoter controlling the expression of luciferase (CSL-LUC) (Fig. 1B). We observed a similar reduction of CSL-LUC activity in the DTX4 knockdown cells contacting OP9-Dll1 cells and DAPT inhibition of activity. Overexpressing murine DTX4 in the DTX4 knockdown U2OS-N1 cells restored N1-dependent CSL-LUC activation to that of control cells (fig. S2C). In addition, we generated U2OS-N1 cells in which DTX4 was abolished by clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated 9 (Cas9)–mediated genome editing (N1/DTX4-KO cells; Table 1). N1-dependent activation was decreased by 70% in these cells (Fig. 1C), although N1 abundance was not affected (Fig. 1D and fig. S2D). Other E3 ubiquitin ligases involved in Notch signaling (2830) were also not affected (fig. S2E). These results indicate that DTX4 enhanced, but was not required, for N1 activation.

Fig. 1 DTX4 is a membrane-associated positive regulator of Notch signaling.

(A) Effect of knockdown of DTX isoforms on induction of the Notch target gene Hey1. RNA was extracted from U2OS-N1 cells transfected with the indicated siRNAs (NT, nontargeting) before coculture with OP9-Dll1 (with DAPT when indicated) or control OP9 cells, and transcripts were quantified by qRT-PCR and normalized to UBCH5B. Left: The amount of Hey1 transcript in the cells cultured with OP9-Dll1 cells was set to 100%. Right: Abundance of DTX transcripts. Data in both graphs are means ± SD of three independent experiments. (B) Effect of knockdown of DTX4 on Notch reporter gene induction. U2OS-N1 cells were transfected with both CSL-LUC (Notch reporter) and TK-renilla (internal control) 24 hours before coculture with OP9-Dll1 or OP9 cells. Data are means ± SEM of five experiments, analyzed by one-way analysis of variance (ANOVA) and Bonferroni post hoc test (****P < 0.0001). Reporter activity in cells cocultured with OP9-Dll1 with nontargeting siRNA was set to 100%. The siRNA used in (B) to target DTX4 was different from the siRNA used in (A). (C) Effect of knockout (KO) of DTX4 on Notch reporter gene induction. U2OS-N1 cells or genome-edited DTX4 KO cells (N1/DTX4-KO) were transfected with both CSL-LUC and TK-renilla 24 hours before coculture with or without OP9-Dll1. Data in the graph are means ± SEM of three independent experiments, analyzed by one-way ANOVA and Bonferroni post hoc test (****P < 0.0001). (D) Western blotting (WB) analysis of lysates from U2OS-N1 and N1/DTX4-KO cell, using a Notch antibody that recognizes the intracellular domain of N1 or α-tubulin antibody. p300 is the proform of N1 (not furin-processed); TM-IC is the nonactivated mature form of N1. Molecular weights are indicated on the right. Data are representative of three independent experiments. (E) Localization of N1 and DTX4 in permeabilized (P) and nonpermeabilized (NP) N1/DTX4-GFP cells alone or transiently transfected with DTX4-Cherry or mCherry. For the N1/DTX4-GFP cells, N1 was detected with the antibody recognizing HA, followed by fluorophore-conjugated secondary antibody (red) and DTX4 by GFP fluorescence (green). For the transiently transfected cells, DTX4-GFP is green, and DTX4-Cherry and mCherry are red. Hoechst labeling is in blue in the merged pictures. Scale bar, 10 μm. Insets correspond to threefold enlargements of the boxed regions. Arrows indicate colocalized vesicles. Data are representative of 60 cells analyzed from three experiments. (F) Live-cell imaging (movie S1) of DTX-GFP, mCherry, and DTX4-Cherry. Confocal images are from movies of N1/DTX4-GFP, N1/Cherry, and N1/DTX4-Ch cells. Data are representative of 10 cells analyzed from three experiments. (G) Analysis of the interaction between N1 and DTX4 by coimmunoprecipitation. Whole-cell extracts (WCE) or proteins that immunoprecipitated with the antibody recognizing GFP (IP GFP) from the indicated cell lines were analyzed by Western blotting for the indicated proteins. TM-IC is the mature furin-cleaved form of N1. The lower left blot has residual N1 signal (indicated with gray Notch1). The apparent molecular weights are indicated. Data are representative of three experiments.

DTX4 interacts with N1 at the plasma membrane and in vesicles

To identify the step in N1 activation involving DTX4, we generated stable cell lines from the U2OS-N1 cells that also expressed chimeras of DTX4 with green fluorescent protein (DTX4-GFP) or mCherry (DTX4-Cherry). Each of the DTX4 fluorescent protein fusions complemented DTX4 knockdown in the Notch reporter assay (fig. S2C). We observed a partial colocalization of DTX4-GFP with full-length N1 tagged with HA on the extracellular domain by immunofluorescence of permeabilized and nonpermeabilized cells (Fig. 1E). Both proteins were detected at the plasma membrane and in intracellular vesicles. To avoid fixation-induced artifacts in localization (31), we also assessed protein distribution by live-cell imaging. DTX4-GFP appeared as diffuse and punctate structures, some of which were mobile, especially along the cell edges (Fig. 1F and movie S1). We noticed that DTX4-GFP and DTX4-Cherry exhibited similar, but not identical, patterns (Fig. 1, E and F). These different fluorescence patterns could be due to differences in the abundance of the fusion proteins or could reflect two distinct populations of DTX4, which were revealed by the different chimeric proteins. Distinct populations of Notch have been detected for Notch fused to different fluorescent proteins (32).

We also established a stable cell line from U2OS cells expressing DTX4-Cherry and a fusion protein between N1 and GFP (N1-GFP/DTX4-Ch cells). As a control, we generated cells that expressed GFP and DTX4-Cherry (GFP/DTX4-Ch cells) (Table 1). We confirmed that the N1-GFP/DTX4-Ch cells exhibited a typical pattern of N1 cleavage products in response to activation (fig. S2F). To determine whether N1 and DTX4 interacted either directly or as part of a complex, N1 coimmunoprecipitated with DTX4-GFP from N1/DTX4-GFP cells (Fig. 1G, left), and DTX4-Cherry coimmunoprecipitated with N1 from the N1-GFP/DTX4-Ch cells (Fig. 1G, right). In addition, the N1-DTX4 interaction was not detected when mixing extracts from cells that expressed either N1-HA (U2OS-N1 cells) or DTX4-Cherry (DTX4-Ch cells), showing that the interaction happened before cell lysis (fig. S2G). These coimmunoprecipitation data indicated that DTX4 and N1 form a complex either through direct or indirect mechanisms.

DTX4 participates in an early step of Notch activation

Because our data indicated that DTX4 was localized to the plasma membrane and vesicular membranes, and of previous work suggesting a role for Drosophila Dx in Notch endocytosis (18, 20, 21), we monitored the effect of DTX4 silencing on the transendocytosis of the Notch extracellular domain into the ligand-expressing cell (4), which is considered the first step in Notch activation. N1 cells were transfected with siRNAs to silence DTX4 (Fig. 2A) or proteins necessary for clathrin-mediated endocytosis, dynamin 2 (Dyn2), adaptor protein 2α (AP2α), or AP2μ (Fig. 2, B and C), and then cocultured with OP9-Dll1 cells for 4 hours before fixation and permeabilization for immunofluorescence analysis with antibodies that recognize the HA tag on the extracellular domain of N1 and the VSV tag on the extracellular domain of Dll1. Compared with control cells, less N1 colocalized with Dll1 in the OP9-Dll1 cells when DTX4 was knocked down in the U2OS-N1 cells, indicating that transendocytosis of N1 into the Dll1-expressing cells was impaired (Fig. 2, A and C). A similar effect was observed with a different siRNA targeting the 3′-untranslated region (3′UTR) of DTX4, but not by targeting DTX1, and was partially complemented by DTX4 overexpression (fig. S3A). DTX4 knockdown did not impair the N1-Dll1 interaction, because HA and VSV staining on nonpermeabilized cells colocalized at sites of cell contact (Fig. 2B), nor did it affect N1 abundance at the cell surface (fig. S3B). We also checked that whole N1 or Dll1 proteins were not pulled into opposing cells and wrongly quantified as transendocytosis events (fig. S3, C and D). These results are consistent with the reduced N1-mediated transcriptional activity in DTX4-deficient cells (Fig. 1) and indicated that DTX4 functioned at a step after ligand binding to Notch and before transendocytosis.

Fig. 2 Transendocytosis is affected by events occurring in Notch-expressing cells.

(A) Effect of knockdown of DTX4 on transendocytosis. U2OS-N1 cells transfected with the indicated siRNAs were cocultured with OP9-Dll1 cells for 4 hours before immunofluorescence analysis using fluorescent anti-HA (green) and anti-VSV (red). The images are representative of five independent experiments. (B) Effect of knockdown of DTX4 or Dyn2 on transendocytosis. As in (A), except that cells were permeabilized or not before immunofluorescence. The images are representative of five independent experiments. Scale bars, 10 μm. (C) Quantification of transendocytosis events in (A). The percentage of colocalizations of the extracellular portions of Notch and Dll1 was determined, using the Colocalizer protocol under Icy software, as the ratio between HA- and VSV-positive spots over the total VSV spots. Equivalent acquisition settings were used within each experiment. The graphs represent mean ± SEM of five independent experiments (100 to 160 cells from 30 to 60 fields). Statistically significant differences are compared to the NT conditions (****P < 0.0001, **P < 0.01, *P < 0.05).

Endocytosis of Dll1 exerts a mechanical strain on the Notch extracellular domain as it pulls part of Notch into the Dll1-expressing cell (4). Endocytosis events in the Notch-expressing cell may also contribute to the force necessary to dissociate the Notch heterodimer. Knockdown of Dyn2 or AP2 (α or μ2 subunits) in the U2OS-N1 cells decreased the number of transendocytosis events (Fig. 2, B and C) without impairing the N1 and Dll1 interaction (Fig. 2B, nonpermeabilized). Expression of a dominant-negative form of Dyn2 in U2OS-N1 cells also prevented the internalization of the cell surface HA-tagged N1 and VSV-tagged Dll1, resulting in an apparent increase in N1-Dll1 colocalization at sites of cell contact area (fig. S3E). These results indicated that, in addition to endocytosis in ligand-expressing cells, in the Notch-expressing cells, DTX4 and endocytosis of Notch contribute to the dissociation of Notch heterodimer.

DTX4 functions before ADAM10 in the Notch activation process

The production of the NICD requires two successive proteolytic events, the first by a protease of the ADAM family (10, 11), which, in most cells, is ADAM10 (7, 33), and a second by the γ-secretase proteolytic complex (12). ADAMs are thought to function at the extracellular side of the plasma membrane to catalyze the ectodomain shedding of their substrate (34), implying that ADAM10-mediated cleavage could occur before Notch endocytosis. However, we found that U2OS-N1 cells in which ADAM10 was knocked down exhibited a similar colocalization of the HA-tagged N1 and VSV-tagged Dll1, indicating that ADAM10 silencing in the Notch-expressing cells did not impair N1 transendocytosis into the Dll1-expressing cell (Fig. 3A). Consequently, this result indicated that ADAM10 acts after separation of the Notch heterodimer, as has been suggested previously (4), and therefore that ADAM10 may cleave Notch after endocytosis. We confirmed that ADAM10 was the relevant metalloproteinase in the U2OS cells by analyzing the Notch-driven luciferase activity in response to either coculture with the OP9-Dll1 cells or activation by coating the culture dishes with a fusion protein consisting of the extracellular portion of Dll1 and the Fc portion of an antibody (Dll-Fc) (35). Dll-Fc coating experiments showed that ADAM10 inhibition affected Notch signaling by targeting the N1-expressing cells, not the Dll1-expressing cells. Knocking down ADAM10 in U2OS-N1 cells or exposing the cells to a selective inhibitor of ADAM10 activity, GI 254023X (GI) (36, 37), reduced the induction of the reporter, with GI producing greater inhibition than ADAM10 knockdown (Fig. 3B). Knockdown of both DTX4 and ADAM10 produced greater inhibition that was achieved by single knockdowns in the reporter assay. Two models could explain these results: Either DTX4 functions before ADAM10 does or DTX4 functions in an activation pathway that is independent from ADAM10. To discriminate between these two possibilities, we exposed the cocultured cells to increasing concentrations of GI (Fig. 3C). At the maximal GI concentration, Notch reporter activity was reduced to background for the control and DTX4-knockdown cells, consistent with Notch activation requiring ADAM10 activity or the activity of another GI-sensitive ADAM in U2OS-N1 cells and indicating that DTX4 and ADAM10 function in the same activation pathway. We confirmed that GI specifically targeted ADAM cleavage of Notch without off-target effects because GI exposure did not affect Notch processing in cells expressing a constitutively active, membrane-anchored form of N1, N1ΔE-GFP (fig. S4, A and B). Therefore, the data support a model in which DTX4 acts upstream of ADAM10 on Notch.

Fig. 3 ADAM10 acts downstream of DTX4 and transendocytosis in the Notch pathway.

(A) Effect of knockdown of ADAM10 on transendocytosis, which was measured as in Fig. 2. The picture is representative of three independent experiments used for quantification (more than 60 cells from 30 fields) in the middle graph (mean ± SEM, statistical analysis by an unpaired t test; n.s., P > 0.05). The right graph shows ADAM10 expression normalized to HPRT expression, as quantified by qRT-PCR (unpaired t test, ****P < 0.0001). (B) Effect of knockdown of DTX4 and ADAM10 (A10) on Notch reporter gene induction. CSL-LUC activity was obtained as in Fig. 1, either after coculture with OP9-Dll1 cells or after culture on Dll-Fc–coated plates. DAPT or GI (at a suboptimal concentration) was applied all along the coculture. No stimulation was in the absence of Dll1. Data are means ± SEM of four experiments for Dll1-expressing cells (dark gray bars), analyzed by one-way ANOVA and Bonferroni post hoc test (****P < 0.0001, **P < 0.01). Data are means ± SEM of two experiments for coated Dll-Fc (light gray bars), each point in triplicate. (C) Effect of increasing doses of GI on Notch reporter gene induction. Curves giving the variation in luciferase activity (means ± SD of three independent experiments) as a function of the GI concentration, depending on the transfected siRNAs. 100% is control condition (siNT) without GI.

DTX4 is involved in Notch ubiquitylation early during activation

Because experiments involving N1 activation in response to coculturing with Dll1-expressing cells precluded any kinetic analysis of Notch activation, we performed experiments with soluble Dll1-Fc (35), which we induced to form clusters before using in solution or coating onto culture dishes (see Materials and Methods). When coated on the culture dishes onto which U2OS-N1 cells were plated, the preclustered Dll1-Fc (hereafter named clustered Dll-Fc) activated the Notch reporter to the same extent as OP9-Dll1 cells and exhibited the same sensitivity to DTX4 silencing and ADAM10 inhibition (Fig. 3B). Therefore, the function of DTX4 in N1 activation did not require the presence of ligand-expressing cells and the pulling force provided by transendocytosis of the extracellular part of the receptor and the ligand into these cells. We incubated N1/DTX4-Ch (stably expressing N1-HA and DTX4-Cherry; Table 1) or U2OS-N1 cells with clustered Dll-Fc for 30 min, which led to the formation of N1-HA and Dll-Fc clusters and of DTX4-Cherry and Dll-Fc clusters (Fig. 4A), suggesting the formation of a complex containing the ligand-bound receptor and the E3 ubiquitin ligase. Analysis of permeabilized and nonpermeabilized N1/DTX4-Ch cells showed that N1, DTX4-Cherry, and clustered Dll-Fc colocalized in puncta detectable in nonpermeabilized cells, indicating that they colocalized at the cell surface (fig. S5A). To confirm that the Cherry fluorescent protein was not causing the formation of DTX4-Cherry fusion protein in puncta, we analyzed the localization of Cherry, N1, and Dll-Fc in N1/Cherry cells (expressing HA-tagged N1 and mCherry; Table 1). Cherry remained diffusely spread throughout the cell in the presence of clustered Dll-Fc and did not accumulate in the N1 clusters (fig. S5B), showing that DTX4 accounted for DTX4-Cherry clustering in clustered Dll-Fc–treated cells expressing N1 and DTX4-Cherry. Moreover, GI or DAPT treatments did not affect the formation of clustered Dll-Fc–induced clusters of DTX4-Cherry that colocalized with the ligand (fig. S5C), consistent with cleavage by ADAM10 and γ-secretase occurring after the interaction of the receptor with DTX4. To establish that Dll-Fc, N1, and DTX4 colocalized at the same clusters, we exposed N1-GFP/DTX4-Ch cells (Table 1) to clustered Dll-Fc for 60 min and labeled Dll-Fc with a far-red fluorophore. Many of the Dll-Fc clusters were also positive for N1-GFP and DTX4-Cherry (fig. S5D). Therefore, we concluded that DTX4 was recruited to (or became enriched at) the cell surface ligand-bound Notch, consistent with an early function for DTX4 in the activation process.

Fig. 4 N1 undergoes clustering and ubiquitylation upon activation by Dll-Fc.

(A) Formation of clusters. N1/DTX4-Ch or U2OS-N1 cells were incubated with clustered Dll-Fc for 0 or 30 min at 37°C, fixed, permeabilized, and labeled with antibodies recognizing Fc (green), HA (on U2OS-N1 cells; red), and appropriate secondary antibodies. In N1/DTX4-Ch cells, red fluorescence is the DTX4-Cherry protein. Hoechst is in blue in merged images (first column). Insets are threefold enlargements of the boxed regions; arrows indicate colocalized vesicles. Scale bar, 10 μm. Data are representative of four independent experiments (>200 cells). (B) Ubiquitylation of N1. GFP/DTX4-Ch or N1-GFP/DTX4-Ch cells were transfected with the indicated siRNAs before activation with clustered Dll-Fc and GI for 30 min at 37°C. Whole-cell extracts or proteins that immunoprecipitated with the antibody recognizing GFP (IP GFP) were analyzed by Western blotting for the indicated proteins. White lane indicates that intervening lanes have been spliced out from the same blot. On the left are the molecular weights. The bracket indicates N1-derived ubiquitylated products; TM-IC is the mature form of N1. The table below gives the quantification of the ubiquitylated/total immunoprecipitated N1 reported to control conditions and the quantity of exogenous DTX4 (means from two experiments).

Because DTX4 is a E3 ubiquitin ligase, we monitored the ubiquitylation of N1 from cells exposed to clustered Dll-Fc. The amount of ubiquitylated Notch increased after 30 min of exposure to clustered Dll-Fc in N1-GFP/DTX4-Ch cells (Fig. 4B) or U2OS-N1 cells (fig. S6A). Blocking endocytosis in the N1-expressing cells by knocking down Dyn2 resulted in an increase in the amount of ubiquitylated N1 (Fig. 4B and fig. S6A), consistent with ubiquitylation occurring before endocytosis and the accumulation of ubiquitylated products when endocytosis was impaired. In contrast, silencing DTX4 reduced Notch ubiquitylation (fig. S6A) and prevented the increase observed in response to Dyn2 silencing (Fig. 4B). Finally, we expressed His-tagged ubiquitin in U2OS-N1 cells in which DTX4 or Dyn2 had been knocked down individually or in combination, exposed them to clustered Dll-Fc, isolated ubiquitylated products under denaturing conditions (27), and monitored the amount of Notch-ubiquitylated products with an antibody that recognized the intracellular part of the receptor (fig. S6B). The decrease in ubiquitylated Notch when DTX4 was silenced confirmed that N1 was directly modified by ubiquitin under activation in a DTX4-dependent manner and before Dyn2 requirement. Together, these results suggested that N1 and DTX4 clustering upon Dll-Fc–mediated activation induces DTX4-dependent N1 ubiquitylation before the receptor is internalized.

Visualizing Notch activation in living cells reveals the formation of moving N1- and DTX4-containing clusters

We performed live-cell imaging with the N1-GFP/DTX4-Ch cells to assess the process of N1 activation. N1-GFP clusters formed in less than 1 min after the addition of clustered Dll-Fc directly to the culture medium on top of the cells and continued to form over the course of 40 min (Fig. 5A and movie S2). DTX4-Cherry was also present in these clusters of N1-GFP (see enlargements in Fig. 5A, blue arrows). Some of the clusters were less mobile, and others appeared to detach from the membrane, likely as vesicles, and move within the cytosol (Fig. 5A, yellow arrows). Live-cell imaging 60 min after Dll-Fc addition revealed many moving vesicles that were positive for N1-GFP and DTX4-Ch, along with the less mobile clusters that might remain associated with the plasma membrane (Fig. 5B, Table 2, and movie S3). Imaging of the cells 40 min after Dll-Fc addition in the presence of DAPT showed that both mobile and immobile clusters of N1-GFP were present in Dll-Fc–stimulated cells exposed to DAPT (Table 2 and movie S4). Mobile and immobile N1-GFP clusters were also detected 60 min after Dll-Fc addition in cells in which γ-secretase (DAPT; movie S5) or ADAM10 (GI; movie S6) was inhibited (Table 2). These live-cell imaging data were consistent with Notch and DTX4 clustering before their endocytosis and suggested that DTX4 functions in Notch activation early in the process. Because DTX4-Cherry fluorescence was difficult to detect, we also performed live-cell imaging of N1/DTX4-GFP cells 45 min after the addition of Dll-Fc and observed DTX4 immobile and mobile clusters similar to those observed in the N1-GFP/DTX4-Ch cells after clustered Dll-Fc addition (Table 2 and movie S7).

Fig. 5 Live imaging of Notch-GFP activation.

(A) Activation in N1-GFP/DTX4-Ch cells. Sequential confocal images from movie S2 showing N1-GFP upon activation by clustered Dll-Fc over a 40-min period. Insets are threefold enlargements of the squared boxes at each time point, with Notch-GFP in green and DTX4-Cherry in red. Blue arrows indicate immobile clusters, and yellow arrows indicate mobile vesicles. The image is representative of five independent experiments, with more than 10 cells in randomly selected visual fields recorded in each experiment. (B) Tracking of mobile vesicles. Sequential confocal images (4 s apart) of N1-GFP/DTX4-Ch cells, 60 min after the addition of clustered Dll-Fc, extracted from movie S3. Scale bars, 10 μm. The tracking of four rapidly moving vesicles (four colors) is indicated in green, as well as their position on each image by arrowheads. The image is representative of five independent experiments, with more than 10 cells in randomly selected visual fields recorded in each experiment. Nucl, nucleus.

Table 2 Summary of the live-cell imaging analysis.

Live-cell imaging in three independent experiments was analyzed under Icy software, sequentially using Spot Detector, Spot Tracking, and Motion Profiler plugins. Tracks over 10 s of duration were quantified; immobile vesicles were defined as having a net displacement under 1 μm. Because the speed is not constant over the whole tracking of a given vesicle, we monitored the average speed and the maximal speed of the vesicles. Speeds are expressed as means from the total mobile vesicles. Note that because the movies are acquired using a fixed z-stack, vesicle movements can be underestimated when they become out of focus. The number of analyzed cells for each condition is indicated in the table. Dll-Fc stimulation is with clustered Dll-Fc. n.a., not applicable.

View this table:

ADAM10 acts on Notch after internalization

To visualize the spatiotemporal recruitment of endogenous ADAM10 to Notch, we activated N1 by adding clustered Dll-Fc in the presence of a monoclonal antibody that recognized the ADAM10 extracellular domain (38) and in the presence of GI to prevent N1 cleavage in N1-GFP/DTX4-Ch cells. ADAM10 did not colocalize with N1-GFP and DTX4-Cherry in the receptor clusters detected 15 min after stimulation (Fig. 6A, yellow arrows). ADAM10-labeled endocytic vesicles were detected after 30 to 60 min, and some of the ADAM10 staining colocalized with N1-GFP and DTX4-Cherry (Fig. 6A, white arrows). Colocalization of N1-GFP and ADAM10 was not detected in nonpermeabilized cells (fig. S7A), confirming that these were intracellular structures and suggesting that ADAM10 recruitment to the ligand-bound receptor occurred after internalization of the receptor along with DTX4. We also predicted that the clusters containing N1-GFP and DTX4-Cherry that were negative for ADAM10 were at the cell surface (Fig. 6A, yellow arrows). To test whether ADAM10 recruitment to Notch required endocytosis of both proteins, we treated the N1-GFP/DTX4-Ch cells in the presence or absence of clustered Dll-Fc and of the antibody against the extracellular domain of ADAM10 with Dynasore, a cell-permeable inhibitor of dynamin-dependent endocytosis (39). In the presence of Dynasore, Dll-Fc–induced clustering of N1-GFP still occurred, but N1-GFP was localized at the cell periphery, and the amount of N1-GFP and ADAM10 colocalization decreased (Fig. 6, B and C, for quantification). We obtained a similar colocalization of N1 and ADAM10 when using U2OS-N1 cells, in the presence of clustered Dll-Fc, GI, and fluorescent anti-HA and anti-ADAM10 (fig. S7B). Transfecting U2OS-N1 cells with siRNAs targeting DTX4, Dyn2, or AP2 also reduced the colocalization of N1 and ADAM10 (Fig. 6D and fig. S7B). Finally, comparing the number of HA-positive vesicles per cell colocalized with ADAM10 under the various conditions of activation showed that Notch and ADAM10 colocalization was increased under activation in a dynamin-dependent manner (fig. S7C). Together, these data are consistent with DTX4 promoting dynamin-mediated endocytosis after which ADAM10 encountered Notch.

Fig. 6 ADAM10 is recruited to Notch after internalization.

(A) Colocalization of N1-GFP, DTX4-Cherry, and ADAM10. N1-GFP/DTX4-Ch cells were incubated at 37°C with GI, clustered Dll-Fc, and an antibody recognizing ADAM10 extracellular domain for 15, 30, or 60 min, and then fixed, permeabilized, and incubated with fluorescent secondary antibody against ADAM10 antibody (blue in the merged pictures). Insets correspond to threefold enlargements of the boxed regions; yellow arrows indicate N1-GFP (green) and DTX4-Cherry (red) colocalized vesicles, devoid of ADAM10 staining, and white arrows point to vesicles with the three proteins. The data are representative of five independent experiments, with more than 10 cells in randomly selected visual fields observed in each. Scale bar, 10 μm. (B) Effect of Dynasore treatment. N1-GFP/DTX4-Ch cells were incubated as in (A) for 60 min, in the presence or absence of clustered Dll-Fc and Dynasore, and then fixed and analyzed as in (A). Insets are threefold enlargements of the boxed regions; white arrows point to vesicles with the three proteins. Images are representative of three independent experiments. (C) Quantification of N1-ADAM10 colocalization in activated N1-GFP/DTX4-Ch cells from (B). The percentage of N1-ADAM10 colocalization was determined as the ratio between GFP- and ADAM10-positive spots over the total GFP spots. Equal acquisition settings were used within each experiment used for the graphs (mean ± SEM from two independent experiments; 342, 245, and 143 cells for the three conditions). The results of the third experiment in (B) were not amalgamated with the other two because of different acquisition settings; however, the quantification results were similar. Statistically significant differences are compared to the NT conditions (one-way ANOVA and Bonferroni post hoc test, ****P < 0.0001). (D) Effect of knockdown of DTX4, Dyn2, or AP2 on N1-ADAM10 colocalization. U2OS-N1 cells were transfected with the indicated siRNAs and then incubated at 37°C with GI, clustered Dll-Fc, and a fluorescent antibody recognizing ADAM10 extracellular domain for 60 min before fixation and subsequent staining with fluorescent antibody recognizing HA. The percentage of Notch-ADAM10 colocalization was the ratio between HA- and ADAM10-positive spots over the total HA spots. Equivalent acquisition settings were used within each experiment. The graph represents mean with SEM from two independent experiments (>200 cells from 30 fields), and statistically significant differences are compared to the NT conditions (one-way ANOVA and Bonferroni post hoc test, ****P < 0.0001). Representative images are shown in fig. S7B. (E) Representative blotting of N1-GFP processing products. N1-GFP/DTX4-Ch cells were cultured for 7 hours on coated Dll-Fc in the presence or not of DAPT and Dynasore, as indicated. Cell extracts were analyzed by Western blot with antibodies recognizing GFP, S3 product (V1744), and tubulin. TM-IC (S1) is the nonactivated, mature form of Notch; S2 and S3 result from ADAM and γ-secretase processing, respectively. The apparent molecular masses are indicated. Below is the quantification of S2 abundance, as the ratio of S2 over (S1 + S2) in the presence of DAPT. The graph represents means with SEM from six independent experiments, with statistical analysis by unpaired t test (****P < 0.0001).

To monitor Notch processing by ADAM10, we plated N1-GFP/DTX4-Ch cells on dishes coated with Dll-Fc for 7 hours in the presence or absence of DAPT (to block γ-secretase) or Dynasore (to block dynamin-dependent endocytosis) or both. By Western blotting for the N1-GFP forms produced from the TM-IC (WB GFP) and specifically for the S3 (γ-secretase cleaved form) product (WB V1744), we showed that Dll-Fc–mediated N1 activation produced S3 and that activation in the presence of DAPT stabilized the S2 form (Fig. 6E). Activation in the presence of Dynasore reduced the production of both S3 and S2 (Fig. 6E), in accordance with dynamin-dependent endocytosis of N1 enabling efficient ADAM10 cleavage. Because Dynasore treatment for 7 hours affected cell shape, we checked that N1-GFP was present at the cell surface and matured into the N1-GFP TM-IC (S1 form, resulting from furin cleavage) using a cell surface biotinylation assay (fig. S7D). From the data, we proposed that ADAM10 could cleave the receptor in the endocytic vesicles in which ligand-bound Notch and ADAM10 colocalized. Structures positive for both ADAM10 and N1-GFP also stained with the early endocytic marker Rab5, confirming that these vesicles were endocytic (fig. S7E).

To confirm that the process of Notch activation that we identified with the clustered Dll-Fc were the same in cells activated by the ligand in its cell-attached configuration, we cocultured N1-GFP/DTX4-Ch cells with OP9-Dll1 cells for 4 hours. We observed that N1-GFP, DTX4-Cherry, and Dll1 accumulated at sites of cell-cell contact, as well as some structures that appeared to be internal vesicles positive for N1-GFP and DTX4-Cherry (Fig. 7A). After 4 hours of coculture, we added the antibody recognizing the extracellular domain of ADAM10 and incubated the cells at 37°C for an additional 30 min. We observed partial colocalization of ADAM10 with N1-GFP but only in permeabilized cells (Fig. 7B), confirming that N1-GFP colocalized with ADAM10 after endocytosis of both proteins.

Fig. 7 ADAM10 recruitment to N1 in coculture.

(A) N1 and DTX4 colocalization upon activation by OP9-Dll1 cells. N1-GFP/DTX4-Ch cells were cocultured without or with OP9-Dll1 (second panel) for 4 hours, fixed but not permeabilized (NP), and incubated with antibody recognizing VSV and secondary antibody coupled to Alexa Fluor 647 (blue). A star indicates the position of the closest murine Dll1-expressing cell of the N1-expressing cell (N). Insets on the right correspond to the boxed region; green arrows point to a probable endocytic vesicle. The second panel is representative of the small proportion of cells in which N1 has not been activated, where N1-GFP was still cell surface–localized together with Dll1-VSV. Images are representative of two independent experiments (>20 cells observed). (B) Colocalization of N1 and ADAM10. N1-GFP/DTX4-Ch cells were cocultured with OP9-Dll1 for 4 hours, incubated with the antibody recognizing ADAM10 for an additional 30 min, fixed, permeabilized (P) or not (NP), and labeled with secondary antibody coupled to Alexa Fluor 647 (red). Insets correspond to threefold enlargements of the boxed regions; white arrows indicate N1- and ADAM10-positive vesicles. Images are representative of two independent experiments (>40 cells observed). Scale bars, 10 μm. (C) Model depicting the early steps of Notch activation. Upon Dll1 binding (left), Notch (green) and DTX4 (blue) are clustered. DTX4-dependent ubiquitylation (orange) of Notch leads to the recruitment of the endocytic machinery, including Dyn2 and AP2 complex (middle), enabling bilateral endocytosis and eventually Notch heterodimer dissociation (right). ADAM10 is internalized separately and cleaves Notch in endocytic vesicles (right).


Using two ways of activating Notch (Dll1-expressing cells and clustered Dll-Fc), we monitored the fate of the two fragments of the N1 heterodimer in U2OS cells. Our work revealed previously unknown steps contributing to Notch activation (Fig. 7C). We identified the clustering of ligand-bound receptor at the cell surface with DTX4 (Fig. 7C, left and middle), N1 ubiquitylation in a DTX4-dependent manner (Fig. 7C, middle), and endocytosis of the receptor before ADAM10 processing (Fig. 7C, right). These events facilitated the transendocytosis of the N1 extracellular domain into the ligand-expressing cell. In addition, our results indicated that the cleavage of N1 by ADAM10 occurred primarily in endocytic vesicles, rather than at the cell surface (Fig. 7C, right), and that the ADAM10 and the N1-DTX4 complex were internalized through distinct mechanisms.

Both clustered Dll-Fc–mediated activation and OP9-Dll1–mediated activation of N1 were enhanced by DTX4 and involved endocytosis before ADAM10 recruitment, and both systems of activation exhibited similar sensitivity to inhibitors of ADAM10 or γ-secretase. The major advantage of the Dll-Fc–mediated activation system was that kinetic analyses were possible. Under both conditions, we first observed a clustering of N1 at the cell surface, followed by the movement of some of the clustered N1 into vesicles that formed even when inhibiting ADAM10 or γ-secretase, suggesting that N1 endocytosis did not involve cleavage by these enzymes. The formation of the clusters and its time course were reminiscent of the cell aggregation and of the movement of clusters of ligand-bound Notch described by others, in Drosophila or mammalian cell coculture experiments (4043). Our live-cell imaging showed that N1- and DTX4-positive mobile vesicles were common in 30 min of stimulation and that some moved rapidly, with maximal speeds of about 1 μm/s, and traveled long distances, in accordance with trafficking along microtubules (44). Nevertheless, it is possible that the clustered Dll-Fc stimulation only recapitulates the first steps of Notch activation (until Notch encounters ADAM10-positive vesicles) because Dll-Fc could remain bound and impair Notch heterodimer dissociation. This could explain the accumulation of moving vesicles along time in our movies and the lack of nuclear N1-GFP staining.

Inhibiting endogenous DTX4 expression by siRNA or by genome editing decreased the N1-dependent transcriptional activation significantly, although not completely. In addition, biochemistry, immunofluorescence, and in vivo imaging showed that DTX4 acted early in the Notch activation process. However, our results do not demonstrate an obligate role for DTX4 in the Notch pathway but indicate that DTX4 facilitates Notch activation in the cellular model that we used. DTX1 and DTX2 had no role in Notch signaling in this model, and the molecular basis of such specificity needs further investigation. In different cell or tissue contexts, other DTX isoforms may perform the function that we identified for DTX4 in the N1-expressing U2OS cells. Although Drosophila Dx has been involved in ligand-independent activation of Notch (18), we do not favor a similar role for DTX4 because, in our coculture system, activation was always dependent on the presence of clustered Dll-Fc or of Dll1-expressing cells. Alternatively, Notch signaling in the absence of DTX4 could be affected by gain of function of Su(dx) homologs (Itch, Nedd4, Nedd4-2, WWP1, or WWP2 in mammals), which regulate Notch internalization in Drosophila and mammals (21, 28). Although we cannot completely rule out this possibility, our results showed that Notch quantity at the cell surface was the same, regardless of whether DTX4 was knocked down or not, suggesting that Itch-dependent Notch degradation was not affected.

We showed here that DTX4 colocalized and coimmunoprecipitated with N1, even in the absence of activation, consistent with possible additional functions in the degradation of nonactivated Notch, such as recruitment of arrestins to Notch (45, 46). Upon Dll1 binding, we found that N1 was ubiquitylated in a DTX4-dependent manner, suggesting that clustering or some other event associated with ligand binding to Notch stimulates the E3 ubiquitin ligase activity of DTX4. We also showed that N1 ubiquitylation preceded Notch internalization because Dyn2 silencing stabilized the N1 ubiquitylated species. DTX4-dependent ubiquitylation of Notch, receptor clustering, or both processes could locally concentrate endocytic adaptors and initiate endocytosis. DTX4 interacts with Grb2 (47), which can bind to dynamin. Thus, DTX4 could also indirectly activate the guanosine triphosphatase activity of dynamin and promote dynamin oligomerization to stimulate the release of endocytic vesicles from the plasma membrane (48). Our results also indicated that DTX4 was associated with N1 after endocytosis and that there was some overlap with ADAM10-positive structures. We have previously reported that activated Notch associates with DTX1 and the deubiquitylase eukaryotic translation initiation factor 3 subunit F (eIF3f) before γ-secretase processing (27). DTX4 may also recruit eIF3f or another deubiquitylase to N1 after endocytosis. This hypothesis is consistent with Notch undergoing successive ubiquitylation events: Our results here place ubiquitylation as an early event before endocytosis, and other studies have described monoubiquitylation after endocytosis (27).

Our results support the previously identified role for transendocytosis of Notch in the activation process (4, 49). Gordon et al. (6) have reported that ligand-expressing cells supply sufficient force to induce metalloproteinase sensitivity in the Notch NRR (negative regulatory region) and to enable Notch activation. However, those in vitro experiments may not recapitulate all the in vivo parameters, including the membrane microenvironment, interactions of the receptor with other factors, or ADAM availability. Chapman et al. (43) have shown that contact with ligand-expressing cells induces internalization and intracellular trafficking of N1, but have not concluded that Notch internalization contributes to transendocytosis. In contrast, we found that clathrin-dependent endocytosis of the activated receptor contributes to transendocytosis. Therefore, our results support a model in which the conformational change in Notch that is necessary to open the ADAM cleavage site involves both the mechanical strain provided by the internalization of the ligand and Notch extracellular domain into the ligand-expressing cell (4) and also the internalization of the transmembrane subunit of the separated receptor and DTX4. The respective requirement or importance of each event and whether DTX4 or a different isoform is involved may depend on each tissue context, which defines the mechanical properties of and proteins present in tissues, and needs further study.

Biochemical assays have shown a role for proteases of the ADAM family in Notch signaling (10, 11), and mice, worms, and flies lacking ADAM10 have phenotypes that resemble N1-deficient phenotypes (9, 50, 51). Although the typical mechanism of ADAM action is ectodomain shedding of substrates at the cell surface (52, 53), we found that cleavage of N1 by ADAM10 was diminished when receptor internalization was impaired. In addition, we were unable to detect cell surface colocalization of Notch and ADAM10, whether N1 was activated by clustered Dll-Fc or by contact with OP9-Dll1 cells. These results indicated that ADAM10 functioned after N1 endocytosis. Furthermore, our results indicated that ADAM10 and N1 were internalized separately and then meet after fusion of ADAM10-positive vesicles with N1-positive vesicles. Chen et al. (54) have proposed that, instead of spatial and temporal separation of activity, the α- and γ-secretases, which act on various substrates including β-amyloid precursor protein (APP) and Notch, are linked in high–molecular weight complexes that facilitate sequential processing of substrates. If so, APP and Notch should be targeted by different complexes, at the cell surface for APP, and in the endocytic pathway for Notch, consistent with our data for ADAM10 and with previous results for the γ-secretase (27, 5559). We have shown that ADAM10 associates with several members of a subgroup of tetraspanins referred to as TspanC8, which differentially affect its ability to mediate Notch signaling (60). It will be interesting to determine whether these differences in activity when in association with different members of the TspanC8 family are linked to differences in ADAM10 trafficking. These observations suggest the existence of distinctly composed multisubunit secretase complexes with specialized activity for specific substrates, as well as specific cellular locations of activity.


Plasmids, reagents, and antibodies

The N1-GFP construct was obtained by inserting the GFP-encoding complementary DNA sequence, obtained by PCR amplification, in frame at the ECo RV and Xba I sites of murine N1 (61). GFP was inserted at amino acid 2097 of murine Notch, after the sixth ankyrin domain, and replaced the C terminus (including the PEST domain) (fig. S1). DTX4-Cherry and DTX4-GFP constructs were obtained by inserting murine DTX4 at the Hind III–Eco RI sites of pmCherry-N1 and pEGFP-N1 vectors (Clontech), respectively. DAPT was from Calbiochem, and GI and Dynasore were from Sigma-Aldrich. The rabbit polyclonal antibody recognizing N1 intracellular domain and the monoclonal antibody against human ADAM10 (11G2, IgG1) have been previously described (3, 38). Commercial antibodies used in this study were as follows: monoclonal antibody against HA (HA11, 16B12, Covance; Alexa Fluor–conjugated, Life Technologies), rabbit polyclonal antibody against mCherry (DsRed, Clontech), rabbit polyclonal antibody to cleaved N1 and mouse antibody against ubiquitin (#2421 and #3936, respectively; Cell Signaling), monoclonal antibody against VSV (P5D4 Cy3–conjugated, Sigma-Aldrich), mouse antibody against Rab5 (IgG2a, BD), and antibody against EGF receptor (EGFR; sc-120, Santa Cruz Biotechnology). Alexa Fluor–conjugated secondary antibodies were from Life Technologies.

Cell culture, transfections, and siRNAs

OP9 cells expressing the murine Notch ligand Dll1, containing a VSV epitope in the extracellular domain (OP9-Dll1), and the human osteosarcoma cell line U2OS expressing human N1 with an HA epitope inserted between EGF repeats 22 and 23 have been previously described (fig. S1) (27). These cells are named U2OS-N1 cells. N1/DTX4-Ch, N1/Cherry, and N1/DTX4-GFP cells were obtained by transfecting U2OS-N1 cells with DTX4-Cherry, mCherry, or DTX4-GFP vectors, respectively; N1-GFP/DTX4-Ch (and GFP/DTX4-Ch) cells were established by transfecting U2OS cells successively with vectors expressing DTX4-Cherry (producing DTX4-Ch cells) and N1-GFP (or GFP). Stably expressing cells were sorted by flow cytometry, and N1/DTX4-Ch, N1-GFP/DTX4-Ch, and N1/DTX4-GFP clonal populations were obtained by limiting dilution. Table 1 describes all cell lines.

siRNAs were transfected using jetPRIME transfection reagent (Polyplus-transfection) or Lipofectamine RNAiMAX (Invitrogen) for immunofluorescence. DNA vectors were transfected with Fugene HD (Promega), according to the manufacturer’s instructions.

The siRNA sequences DTX1 (5′-CAGAGAGAACCCAGAGUUAdTdT-3′), DTX2 (5′-AGGGAAAGAUGGAGGUAUU[dT][dT]-3′), and DTX4 targeting the 3′UTR (5′-CAGUAGGGAUCUUGAAUUU[dT][dT]-3′) were purchased from Sigma-Aldrich. The siRNA DTX4 targeting the coding sequence (5′-GGAUCGACCUCACUUCCAU-3′) was obtained from Thermo Fisher Scientific.

Generation of N1/DTX4-KO cells

CRISPR/Cas9-mediated gene editing was performed in U2OS-N1 cells to obtain DTX4 KO cells. Guide RNAs were identified using Crispor online tool ( Two guide RNAs (5′-TTCCATGTCGTAGGGCGTCC and 5′-GTTCAGAGCCAGGCACTAGT), targeting sequences in the second exon and in the second intron of the hDTX4 gene, respectively, were cloned into the sgRNA(MS2) plasmid (Addgene). These plasmids were transfected together with the human codon–optimized Cas9-expression vector (pSpCas9, Addgene) into the U2OS-N1 cells. Three days later, cells were cloned by limiting dilution. Clones in which the intervening DNA between the two targeted guide sites had been deleted on both alleles were selected by PCR analysis of genomic DNA (primers, 5′-TGCAAACCAAGATGTCCCGA and 5′-TCCCTGTACCCAAAATGCCC).

Activation assays

For coculture of N1-expressing cells with Dll1-expressing cells, Notch-expressing cells were seeded at 15,000 or 20,000 cells/cm2 and transfected with siRNAs 8 or 18 hours later and, if necessary, with DNA vectors the next day. Twenty-four hours later, cells were cocultured with OP9 or OP9-Dll1 at 25,000 cells/cm2 for 18 hours for the Notch activity assays. For transendocytosis assays, both cell types were trypsinized, mixed, and seeded at 20,000 cells/cm2 (each) for 4 hours.

For the activation by clustered Dll-Fc, Dll-Fc was obtained from 3-day conditioned medium of human embryonic kidney (HEK) 293 cells expressing the fusion protein [provided by G. Weinmaster, University of California, Los Angeles (35)]. Preclustering was either in solution with goat antibody (13 μg/ml) against the human Fc antibody (Sigma-Aldrich) for 1 hour at 4°C or by applying to plates coated with the antibody. For immunostaining or live imaging, preclustered Dll-Fc was diluted twice with Dulbecco’s modified Eagle’s medium and then used to replace cell culture medium on cells grown in serum-containing medium for 24 hours on coverslips. When indicated, cells were preincubated for 1 hour with GI or DAPT before activation, and the antibody recognizing the extracellular domain of ADAM10 (mAb 11G2), GI, or DAPT was added to the mixture.

Analysis of Notch activity

Total RNA extraction was performed using the RNeasy Mini Kit purchased from Qiagen. Reverse transcription was done using the Qiagen Omniscript Reverse Transcription Kit. Quantitative PCR was then performed using the Takyon No Rox SYBR MasterMix dTTP Blue Kit from Eurogentec and analyzed using a CFX96TM real-time PCR detection system under the CFX Manager software (Bio-Rad). Gene expression was normalized to hypoxanthine-guanine phosphoribosyltransferase (HPRT) or to the ubiquitin-conjugating enzyme UBCH5. Oligonucleotides were purchased from Eurogentec, and sequences were as follows: hHPRT, 5′-TAATTGGTGGAGATGATCTCTCAAC-3′ (forward) and 5′-TGCCTGACCAAGGAAAGC-3′ (reverse); hUBCH5, 5′-TGAAGAGAATCCACAAGGAATTGA-3′ (forward) and 5′-CAACAGGACCTGCTGAACACTG-3′ (reverse); hHey1, 5′-GCCTCCTATAGCAGAAAGGTGA-3′ (forward) and 5′-GCAGCTGGTCAGATGGATTC-3′ (reverse); hDTX1, 5′-CTGTTCTCCCATCCTCCCTA-3′ (forward) and 5′-CCTTTGGCTCTCAAAATTGG-3′ (reverse); hDTX2, 5′-CCTGCAAAACCATCTATGGAG-3′ (forward) and 5′-GCGACATCTGGAACCGTAAT-3′ (reverse); hDTX4, 5′-AGCACCAGCAGCCTAGTTTC-3′ (forward) and 5′-GCATCTGACGAGGTTTGTCTC-3′ (reverse).

For luciferase assays, cells were transfected with the CSL-LUC reporter and Renilla plasmid 24 hours before coculture. Cell lysates were prepared using a dual luciferase reporter assay according to the manufacturer’s instructions, and the activities of firefly and Renilla luciferases were determined using a Centro XS luminometer (Berthold).

Ubiquitylation assays

N1-GFP from N1-GFP/DTX4-Ch cells was immunoprecipitated using GFP-trap coupled to agarose beads, according to the manufacturer’s instructions (Chromotek). For purification of His-Ub–containing products, cells were lysed in 8 M urea, 0.1 M NaH2PO4, 10 mM tris (pH 8), 1% Triton X-100, and 20 mM imidazole at room temperature. His-Ub conjugates were purified on Ni-charged chelating Sepharose beads (Pharmacia). Whole-cell extracts and precipitates were analyzed by Western blotting. Image acquisition was performed using myECL Imager, and quantification was performed with myImageAnalysis software (Thermo Fisher Scientific).

Flow cytometry, immunostaining, microscopy, and image analysis

For flow cytometry analysis, U2OS-N1 cells were harvested and suspended in cold phosphate-buffered saline (PBS; pH 7.4) 48 hours after transfection with the siRNAs. Cells were fixed in 4% paraformaldehyde in PBS and immediately labeled with antibodies recognizing HA, ADAM10, EGFR, or preclustered Dll-Fc at 4°C before incubation with Alexa Fluor 488–labeled secondary antibodies. Cell parameters were monitored using a CyAn flow cytometer (Beckman Coulter) and analyzed using the FlowJo software. For immunofluorescence, cells were grown on glass coverslips. Cells were fixed with 4% paraformaldehyde and permeabilized with PBS containing 0.2% Triton X-100 for 5 min before the incubation with appropriate antibodies. Cell preparations were mounted in Mowiol 488 supplemented with DABCO (1,4-diazabicyclo[2.2.2]octane) (Sigma-Aldrich). Images were acquired using an Axio Imager microscope with ApoTome system with a ×63 magnification and using the AxioVision software (Carl Zeiss MicroImaging Inc.) for Figs. 2 to 4 and figs. S3 (A and E), S5, and S7B. For quantifications, images were acquired blind regarding colocalizations, but based on the presence of murine and human cells (which were different by Hoechst staining) in close proximity for transendocytosis events. Post-acquisition image analysis was performed with Icy software (62). The undecimated wavelet transform Spot Detector plugin and the Colocalizer protocol were used to quantify the number of vesicles and the colocalized vesicles, respectively. For Figs. 1D, 6, and 7 and figs. S3 (C and D) and S7 (A, C, and E), image acquisition was performed using a confocal inverted microscope (LSM700) under Zen 2010 software (Carl Zeiss MicroImaging Inc.), and deconvolution was done with Huygens Professional software. For live imaging, cells were grown on glass bottom microwell dishes (MatTek Corp.), and acquisition was done in an incubation chamber at 37°C under 5% CO2 using a spinning-disk UltraView VOX inverted microscope, equipped with a Yokagawa CSUX1 spinning disk, and two cameras to do dual-color simultaneous imaging, controlled by Volocity software (PerkinElmer). The GFP and Cherry signals were always acquired simultaneously; z was fixed along the acquisition. Green and red fluorescence were excited with 488- and 561-nm diode lasers and collected by a dual “green-infrared” 500- to 550-nm and 680- to 750-nm filter. Post-acquisition image analysis was performed automatically with Icy software, using sequentially Spot Detector, Spot Tracking, and Motion Profiler plugins or Manual Tracking for Fig. 5B.

Statistical analysis

Statistical analysis was performed using Prism 6.0 (GraphPad) by one-way ANOVA, followed by the Bonferroni’s multiple comparisons test. Statistically significant differences are indicated with asterisks: ****P < 0.0001; ***P < 0.001; **P < 0.01; *P < 0.05; n.s., P > 0.05.


Fig. S1. Diagram of the constructs used in this study.

Fig. S2. Characterization of the cell lines used in this study.

Fig. S3. Effect of DTX silencing and dynamin K44A overexpression on transendocytosis.

Fig. S4. Specificity of GI on ADAM-mediated cleavage of N1.

Fig. S5. Clustering of N1 with DTX4 at the cell surface upon Dll-Fc activation.

Fig. S6. Ubiquitylation of N1 upon Dll-Fc–mediated activation.

Fig. S7. Colocalization of N1 and ADAM10 upon Dll-Fc–mediated activation.

Movie S1. DTX4-GFP in unstimulated cells.

Movie S2. The first 40 min of N1 activation.

Movie S3. N1 in moving vesicles after 60 min of activation.

Movie S4. Tracking N1 from the cell surface into fast-moving vesicles.

Movie S5. N1 in moving vesicles after 60 min of activation in the presence of DAPT.

Movie S6. Notch in moving vesicles after 60 min of activation in the presence of GI.

Movie S7. DTX4-GFP in stimulated cells.


Acknowledgments: We thank R. Weil and J. Moretti for critical reading and all members of “Signalisation et Pathogenèse” laboratory for the support and discussion. We are thankful for the help of J. Y. Tinevez (Imagopole, Institut Pasteur) in spinning disk microscopy, P. H. Commere (cytometry facility of the Institut Pasteur) in cell sorting, S. Volant for the support for the statistical analyses, the ICY community in image quantifications, and especially V. Meas-Yedid Hardy, S. Dallongeville, and A. Dufour. We thank G. Weinmaster (University of California, Los Angeles) for providing the Dll-Fc–producing cells. Funding: E.R. was supported by grants from Fondation ARC and from INCa (Institut National du Cancer). C.B. was supported by the Institut Pasteur. Author contributions: P.C. and C.B. conducted the experiments. E.R. and C.B. designed the experiments and wrote the manuscript. Competing interests: The authors declare that they have no competing interests.
View Abstract

Navigate This Article