Processing of the Notch Ligand Delta by the Metalloprotease Kuzbanian

See allHide authors and affiliations

Science  01 Jan 1999:
Vol. 283, Issue 5398, pp. 91-94
DOI: 10.1126/science.283.5398.91


Signaling by the Notch surface receptor controls cell fate determination in a broad spectrum of tissues. This signaling is triggered by the interaction of the Notch protein with what, so far, have been thought to be transmembrane ligands expressed on adjacent cells. Here biochemical and genetic analyses show that the ligand Delta is cleaved on the surface, releasing an extracellular fragment capable of binding to Notch and acting as an agonist of Notch activity. The ADAM disintegrin metalloprotease Kuzbanian is required for this processing event. These observations raise the possibility that Notch signaling in vivo is modulated by soluble forms of the Notch ligands.

The Notch (N) signaling pathway defines an evolutionarily conserved cell interaction mechanism that controls cell fate by modulating the cell's response to developmental signals (1, 2). The N receptor is cleaved in the trans-Golgi network as it traffics toward the plasma membrane and eventually forms a ligand-competent heterodimeric molecule (3). Both known ligands, Delta (Dl) and Serrate (Ser), are thought to act as transmembrane proteins that interact via their extracellular domains with N receptors that are expressed on adjacent cells (2, 4). Given the similar phenotypes produced by loss of Notch signaling and loss-of-function mutations in thekuzbanian (kuz) gene [a gene encoding a putative member of the ADAM family of metalloproteases (5)], it has been suggested that Kuz may be involved in the cleavage of N (6). This hypothesis is not corroborated by recent biochemical studies, indicating that the functionally crucial cleavage of N in the trans-Golgi network is catalyzed by a furinlike convertase (7).

A genetic screen to identify modifiers of the phenotypes associated with the constitutive expression of a dominant negative transgene of kuz (kuzDN) in developing imaginal discs identified Delta as an interacting gene (8). Flies expressing this dominant negative kuzconstruct, despite carrying a wild-type complement of kuz, became semi-lethal when heterozygous for a loss-of-functionDelta mutation (8). In contrast, Deltaduplications rescued the phenotypes associated with kuzDN(Fig. 1). The kuzDN flies display extra vein material (especially deltas at the ends of the longitudinal veins), wing notching (observed with a low penetrance), extra bristles on the notum, and have small rough eyes (Fig. 1, A and E) (6). When kuzDN flies carried three, as opposed to the normal two, copies of wild-type Notch, the bristle and eye phenotypes were not affected (8), nor were the vein deltas altered (Fig. 1D). However, the kuzDNphenotypes were effectively suppressed by Delta duplications (Fig. 1, B and F), indicating that a higher copy number of Dl molecules is capable of overriding the effect of the kuzDN construct.

Figure 1

Modifiers of thekuz phenotype. A genetic modifier screen was carried out to identify genes that interact with kuz. In the screen, a strain was used that constitutively expresses a kuzDNconstruct (18) in developing imaginal discs. Adult mutant phenotypes (19) of these flies included extra wing vein elements, mostly notably deltas at the ends of the longitudinal veins [arrowheads in (A)], small and rough eyes, and extra bristles on the notum [arrows in (E)]. Flies that carried three copies of the Delta gene with thekuzDN background (B and F) showed an almost complete suppression of the kuzDN phenotypes. Three copies of Notch, introduced by a transgene (20), yielded an essentially normal phenotype (C) but showed negligible suppression of the kuzDN phenotype inkuzDN flies (D).

The interaction between Delta and kuz was further explored through their respective protein products. Dl antigen was expressed in a stably transfected S2 cell line and was examined with an extracellular domain–specific antibody (9) (Fig. 2A). A fragment migrating faster than Dl was observed exclusively in the medium. The size of this fragment, about 67 kD (Fig. 2C), is consistent with the extracellular domain of Dl, estimated to be 65 kD (Fig. 2D). This fragment was subsequently affinity-purified from the culture medium, and the NH2-terminal sequence was determined (Fig. 2E). The sequence revealed a putative propeptide processing site that is conserved in all the Delta homologs (Fig. 2E). Thus, Dl may be cleaved at the cell surface to release a soluble fragment, designated as DlEC (Delta extracellular domain). Protein immunoblot analysis of Drosophila embryos revealed the existence of both Dl and a fragment with the same mobility as DlEC, which implies that the same Dl-derived product is present in vivo (Fig. 2B). Between Dl and DlEC, additional potentially transient proteolytic products were detectable (Figs. 2B and3D; kuz +/–).

Figure 2

A soluble Delta fragment is released constitutively in Delta-S2 cell culture and in vivo. (A) Expression of Dl antigen in stably transfected S2 cells (17) is detected by SDS-PAGE and protein immunoblotting with monoclonal antibody 9B (9) in nonreduced cell extracts (c) and culture medium (m). A product consistent with Dl is detected in the cell extract. A product of greater mobility is seen in the medium that is consistent in size with the extracellular domain of Dl and is referred to as DlEC. (B) Bands of the same mobility are seen in extracts of 16-hour wild-typeDrosophila embryos. The number of embryos loaded on the gel is shown above the lanes. (C) Affinity-purified DlEC (21) migrates with a molecular mass of about 62 and 67 kD in nonreducing (lane 1) and reducing (lane 2) conditions, respectively, on Coomassie blue–stained SDS-PAGE. (D) A schematic of Drosophila Dl illustrates the conserved Delta Serrate, Lag-2 domain (DSL), the epidermal growth factor (EGF)–like repeats, and the transmembrane domain (TM). Amino acid numbering of the NH2-terminus, the beginning of the TM domain, and the COOH-terminus are shown. (E) Thirteen cycles of NH2-terminal amino acid sequence analysis of DlEC (DIEC) are shown with alignment to the sequences ofDrosophila (dDl), Xenopus (xDl), and human (hDl) Delta proteins. The arrow indicates the conserved serine in the position of the NH2-terminus of DlEC and the putative signal peptide processing site for Dl.

Figure 3

Kuz plays a direct role in Delta processing in vitro and in vivo. (A) Dl and DlEC were visualized by protein immunoblotting with the 9B antibody in the cell pellet (c) and the medium (m) in S2 cells transiently transfected with Dl alone [pMTDl (22) lanes 1 and 2], cotransfected with Kuz (6, 23) (lanes 3 and 4) or cotransfected with Kuz DN (6, 23) (lanes 5 and 6). (B) Cotransfection of Kuz and KuzDN with N [pMTNMg (22)], done under identical experimental conditions as for Dl and protein immunoblotted with an intracellular domain–specific antibody (22), yielded a negligible effect on the processing of N as seen by the invariant levels of NTM, the constitutively processed form of N (3). (C) The metalloprotease inhibitors EDTA and 1,10-phenanthroline (1,10-ph) inhibited the endogenous S2 cell proteolytic activity that produced DlEC. The left panel shows the accumulation of DlEC at various time points up to 60 min in the medium of S2 cells stably expressing Dl. The right panel shows the accumulation of DlEC at 60 minutes in the presence of EDTA and 1,10-phenanthroline. Both of these reagents, which are well-documented metalloprotease inhibitors, inhibited accumulation of DlEC in the medium. (D) Dl processing was inhibited in kuz /– embryos. Nine kuz +/– andkuz /– embryos were identified by morphology, and the extracts were analyzed by SDS-PAGE and protein immunoblotting with 9B. DlEC was absent inkuz /– embryos and demonstrated a higher level of Dl as compared tokuz +/– embryos.

The possibility that the generation of DlEC can be influenced by Kuz was examined by cotransfection experiments in S2 cells that express wild-type Kuz endogenously (6). Cotransfection of Dl with Kuz showed an increase in the DlEC fragment as compared to Dl transfection alone (Fig. 3A). The corresponding decrease in Dl suggests that Dl is the precursor of the DlEC product. These data also indicate that transfection of Kuz acts additively to the endogenous Kuz in the S2 cells. Supporting this hypothesis, cotransfection with KuzDN had an inhibitory effect on DlEC production (Fig. 3A). Under identical experimental conditions, cotransfection of Kuz or KuzDN had no effect on the proteolytic processing of N (Fig. 3B). Thus, Kuz functions in the processing of Dl but not of N. In agreement with this conclusion, DlEC production was markedly inhibited by the metalloprotease inhibitors EDTA and 1,10-phenanthroline (Fig. 3C), whereas no effect was observed with serine protease inhibitors (phenylmethylsulfonyl fluoride and aprotinin), cysteine protease inhibitor (leupeptin), or aspartyl protease inhibitor (pepstatin) (10).

The role of Kuz in generating this product in vivo was examined in kuz mutants. kuz maternal null embryos with either one (kuz +/–) or no (kuz /–) zygotic copies ofkuz were created by crossing female flies carryingkuz germline clones (5). Thekuz /– embryos were distinguished from kuz +/– embryos by the absence of malpighian tubules and the lack of movement in thekuz /– embryos. Extracts prepared from a collection of nine of each type of embryo showed the distinct absence of DlEC and higher levels of Dl in thekuz /– embryos as compared to kuz +/– embryos (Fig. 3D). Reprobing of the same membrane with antibody to N showed no difference in the processing of N in the kuz /– andkuz +/– embryos (10). Furthermore, analysis of 14 randomly selected individual embryos showed 8 embryos with high levels of Dl (10), analogous to thekuz /– embryos (Fig. 3D) and consistent with the predicted numerical outcome of the cross. Together, these observations indicate that Kuz mediates the proteolytic processing of Dl in vivo.

Although kuz mutants have multiple defects, indicating an involvement in several different processes (5), their phenotypes partially overlap with that of Delta mutants. Inactivation of kuz during embryogenesis causes a more extensive neurogenic phenotype than do Delta mutations; nevertheless, it is clear that in the ventrolateral region the neural hypertrophy in the two mutations is identical. Similarly, due to the pleitropy of kuz and Delta the phenotypes associated with mosaic clones are complex. Yet they are also partially overlapping, compatible with the hypothesis that the processing of the Dl protein is mediated by Kuz (5, 6, 11).

DlEC bound specifically to N-expressing S2 cells (Fig. 4A), suggesting that a DlEC-N complex forms on these cells. These results were extended by analysis of the ability of DlEC to compete for Dl binding to N in a cell aggregation assay (Fig. 4B). Preincubation of the N cells with DlEC concentrate (16) resulted in a reduction in the initial rate of aggregation with Dl cells. The competitive effect of DlEC was sensitive to the concentration added and to the time of preincubation with the N cells (10). Furthermore, preincubation of the Dl cells with DlEC had no effect on subsequent aggregation with N cells, indicating that DlEC specifically binds to N in a competitive manner with respect to Dl.

Figure 4

DlEC binds to Notch, competes for Notch-Delta interaction, and acts as an agonist. (A) The DlEC fragment specifically binds to N-expressing S2 (N-S2) cells and does not bind to S2 cells alone. N-S2 cells (lanes 1 and 2), incubated in the absence (lane 1) or presence (lane 2) of DlEC (lane 6), were sedimented through a sucrose cushion, and the extract was protein immunoblotted with antibody 9B (16). Lanes 3 and 4 show parallel incubations with S2 cells in the absence (lane 3) or presence (lane 4) of DlEC. Lane 5 shows DlEC sedimented in the absence of cells. (B) Preincubation of N-S2 cells with DlECconcentrate reduced the subsequent rate of aggregation with Dl-S2 cells as measured turbidimetrically with transmitted light at 320 nm (24). At the concentration shown [1× DlEC, solid circles, (24)] a 60% inhibition in the initial rate of aggregation was seen as compared to control medium concentrate [1× ΔECN, solid squares, (24)]. The error bars show the standard deviation of the mean of triplicate determinations. (C) The effect of DlEC on primary cultured cortical neurons 7 to 10 days old (12) is shown in the representative images as follows: (I) before treatment, (II) cultured in the presence of ΔECN medium, (III) cultured in the presence of DlEC medium, (IV) cultured in the presence of affinity-purified DlEC, and (V) buffer control for purified DlEC. The graph represents the mean length of neurites per neuron. Each bar represents the mean ± SEM of three separate experimental trials. Primary cortical neurons exhibited multipolar morphology and the extensive neurite network in control cultures (I), in cultures in the presence of ΔECN medium (II), and in a buffer control for purified DlEC. A significant decrease in the mean neurite length per neuron and limited neurite branching in cultures treated with DlEC medium (III) and purified DlEC (IV) is seen. Scale bar, 50 μm.

The biological activity of DlEC was examined in a cell culture assay. Neurons develop axodendritic processes (such as neurites) in primary cultures of mouse embryonic cerebrum (Fig. 4C). Sestan et al. (12) have demonstrated that ligand-dependent Notch activation in cortical neurons, which express endogenous Notch receptors, causes morphological changes as well as retraction of neurites. The same effects were observed when the neurons were cultured in the presence of enriched DlEC-containing medium or purified DlEC (Fig. 4C). Thus, DlEChas biological activity and apparently acts as an agonist of Notch activity.

Genetic and biochemical evidence demonstrate that proteolytic processing of Dl produces the soluble DlEC fragment, which is biologically active with an apparent agonistic function in the Notch pathway. Previous studies involving in vivo expression of artificially truncated Notch ligands, in Drosophila and other systems, have demonstrated both agonistic and antagonistic activities (13,14). A soluble form of Delta (DlS) can act as an antagonist in the developing Drosophila eye (13). However, DlEC is not identical to DlS, and therefore it is plausible that the two molecules may be functionally different.

Although Kuz does not appear to be responsible for the constitutive cleavage of N, the possibility that Kuz can cleave N at alternative sites remains. In this regard, it has been claimed that KuzDN is able to inhibit transactivation of a target gene of the N pathway induced by ligand binding to the receptor (7). However, it is possible that this effect does not reflect N cleavage but rather the cleavage of Dl to produce an active ligand. Kleuget al. (15) have recently reported the processing of Dl during normal embryogenesis, demonstrating the existence of Delta fragments, one of which is consistent with DlEC. The intermediate forms detected in embryos 16 to 20 hours old (Figs. 2B and3D; kuz +/–) were not present in kuzmutants (Fig. 3D; kuz /–), raising the possibility that the generation of these products may also be mediated by Kuz. The importance of additional cleavages in Dl, the mode of activity of full-length Dl, and whether the second ligand Ser is also processed are critical questions to resolve. It is now apparent that future analyses of Delta in Notch signaling events must consider its potential as a diffusable ligand.

  • * These authors contributed equally to this work.

  • Present address: Harvard Medical School, Massachussetts General Hospital Cancer Center, Building 149, 13th Street, Charlestown, MA 02129, USA.


View Abstract

Stay Connected to Science

Navigate This Article