Competition between Delta and the Abruptex domain of Notch

Background Extracellular domains of the Notch family of signalling receptors contain many EGF repeat domains, as do their major ligands. Some EGF repeats are modified by O-fucosylation, and most have no identified role in ligand binding. Results Using a binding assay with purified proteins in vitro, it was determined that, in addition to binding to Delta, the ligand binding region of Notch bound to EGF repeats 22–27 of Notch, but not to other EGF repeat regions of Notch. EGF repeats 22–27 of Drosophila Notch overlap the genetically-defined 'Abruptex' region, and competed with Delta for binding to proteins containing the ligand-binding domain. Delta differed from the Abruptex domain in showing markedly enhanced binding at acid pH. Both Delta and the Abruptex region are heavily modified by protein O-fucosylation, but the split mutation of Drosophila Notch, which affects O-fucosylation of EGF repeat 14, did not affect binding of Notch to either Delta or the Abruptex region. Conclusion The Abruptex region may serve as a barrier to Notch activation by competing for the ligand-binding domain of Notch.


Background
The Notch mutation in the fruit fly Drosophila was the first mutation of embryonic development ever described. Notch mutant embryos die from neural hyperplasia because Notch is a negative regulator of neurogenesis [1,2]. In mammals, Notch signaling is involved in a wide array of other developmental processes including somitogenesis, angiogenesis, germ cell proliferation, immune development, axonal pathfinding, proliferation, and intestinal patterning [3][4][5][6]. Notch signaling probably contributes to the development, homeostasis and pathology of most organs [7].
Notch proteins are conserved cell surface receptors. Both Drosophila Notch and human Notch1 have extracellular domains containing 36 tandem EGF repeats, as well as other sequences [8]. A ligand binding domain comprising the two EGF repeats 11 and 12 has been defined using a cell adhesion assay. In this assay, Drosophila cells transfected to express Notch derivatives adhere to cells transfected to express either of two similar transmembrane ligands, Delta or Serrate, only if Notch EGF repeats 11-12 are present [9]. Genetic studies of Notch mutant flies confirm the importance of the EGF repeat 11-12 region in vivo [10]. EGF repeats 11-12 are far distant from the S2 proteolytic cleavage site that is thought to be the liganddependent step in Notch activation ( Figure 1). The juxtamembrane S2 site is protected by a structure involving 3 Lin-12/N repeats that lie C-terminal to the EGF repeats [11]. How ligand binding at EGF repeats 11-12 is communicated to these distant regions of N is not known. Figure 1 Proteins expressed in this study. The entire N ECD structure is shown at bottom, for comparison to the expressed proteins arranged above. Rectangular blocks represent EGF repeats and shaded circles, predicted O-fucosylation. Delta contains two domains, NT and DSL, conserved amino-terminal to its 9 EGF repeats. EGF repeats 11-12 from Notch, constituting the Ligand-Binding domain, are checquered. All Notch EGF repeats are shaded, and EGF repeats 24-29, affected by Abruptex mutations, are shaded darkly. The Fc domain (striped) is carboxy-terminal in all cases. Similar proteins lacking the Fc domain were also expressed (not shown). The LIN-12/N repeats are represented by three rhomboids, disulfide bonds linking the two Notch chains by a vertical line, the ligand-sensitive S2 site by an arrow, and the trans-membrane domain by a solid bar.

Proteins expressed in this study
Little is known about the contribution of most of the EGF repeats to Notch function. The EGF repeat regions of Notch are predicted to form several rod-like arrays due to Ca-coordination by the inter-repeat linkers that precludes flexibility between repeats, with occasional linkers that do not bind Ca perhaps allowing localized flexibility [12]. The overall tertiary structure, and whether it involves interactions between inflexible EGF repeat regions, is not known.
Mutations affecting the Notch extracellular region point to functional importance for many EGF repeats in addition to the minimal ligand binding domain [13,14]. Similarities between Notch from Drosophila, humans, and other species further suggests functional conservation of EGF repeat regions. Some of these regions might interact with other ligands, or affect receptor clustering [15][16][17][18][19][20]. It has also been suggested that some EGF repeats might interfere with ligand binding [21].
Many EGF repeats are the substrate of an enzyme, OFUT-1, that that transfers fucose to Ser or Thr residues preceeding the third Cys of EGF repeats. The C 2 X 4-5 (S/T)C 3 consensus is found on 23 of the 36 EGF repeats of N, and 7 of the 9 EGF repeats of Dl [22]. In addition, the O-fucosyltransferase 1 enzyme acts independently of its catalytic activity, as a chaperone or trafficking factor [23,24]. A specific mutant allele of Notch, called split, is caused by an I578T substitution in EGF repeat 14 [25,13]. split results in increased Notch activity in some tissues, and introduces a site for O-fucosylation where one is normally lacking [26].
A specific allele of Dl that is a dominant suppressors of spl, called Dl sup5 , corresponds to the G305R substitution [27]. It has been suggested that this mutation might prevent or reduce O-fucosylation of EGF repeat 3 on Thr306 [27,28]. This suggested increased fucosylation of Notch on EGF repeat 14 might be suppressed by reduced fucosylation of repeat 3 of Delta. We noticed that O-fucosylation sites are predicted on many EGF repeats of Delta, but few on EGF repeats 11-20 of Drosophila Notch.
Here we used in vitro binding studies to detect specific binding between the ligand-binding region of Drosophila Notch and other EGF-repeat protein regions. Both Delta and the Abruptex region of Notch showed high affinity binding. We did not identify any effect of the split mutant, but the studies suggest that the Abruptex domain could oppose Notch activation by competing with Delta for the ligand binding site.

In vitro interaction between extracellular domains of Notch and Delta proteins from Drosophila
An in vitro binding assay was explored as a means to detect interactions between EGF repeat regions of Notch and its ligand Delta. Hirai's group found that purified extracellular domains of mouse Notch2 and its ligand Jagged2 could interact independently of the cell surface when Notch2 extracellular-domains were adsorbed to an ELISA plate, and Jagged2 extracellular domains expressed as ligand-Fc fusion proteins [33]. The Fc domain from human IgG dimerizes the ligand, also serves as an affinity tag for purification, and is readily detected using secondary antibodies. The method is semi-quantitative and unlike cell surface expression experiments is less affected by variation in transgene expression level since the extracellular domains are partially purified before use [33,34].
We tested whether similar methods could detect interaction between Drosophila Notch and its ligand Delta. Drosophila Schneider cells were transfected to express secreted portions of the extracellular domains of Dl and Notch, tagged with a His 6 tag and, where appropriate, the Fc domain from a human IgG (see Materials and Methods). All proteins expressed and purified in the course of our experiments are listed in Figure 1.
To test whether purified extracellular domains of Dl and N from Drosophila could interact, we assessed the binding of purified Dl:Fc fusion proteins to V5-tagged, His-tagged EGF repeats 11-20 of the N extracellular domain. Henceforth we refer to this protein as N11-20. Figure 2 shows binding data for the interaction between Dl:Fc and N11-20 or N21-30. Fc-tagged proteins were detected using HRP-conjugated anti-Fc antibodies and a colorimetric assay.
Dl:Fc showed saturable binding to the N11-20 protein that contained the known Dl-binding site at EGF repeats 11-12 ( Figure 2A). Human IgG alone bound negligibly to N11-20 ( Figure 2B). Therefore, Dl sequences were responsible for the binding by Dl:Fc. Specific binding depended on the N11-20 protein, and was not evident when only BSA carrier was bound to the plate ( Figure 2C). Dl:Fc also did not bind specifically to a N21-30 protein, which contained the same number of EGF repeats as N11-20 but lacked the known Dl-binding site ( Figure 2D). Thus, in this assay Dl:Fc bound a portion of Notch that contain the Dl-binding site, and not to another EGF repeat portion of Notch that lacked the ligand binding domain.
In the experiment shown in Figure 2, Dl:Fc binding was half-maximal at 1.87 nM, compared to 0.7 nM measured for Jagged:Fc binding Notch2 in similar experiments [33]. It should be noted that half-maximal binding data do not correspond to true Kd values, because the stoichiometry of binding is unknown and because these may not be equilibrium binding measurements. In addition, whereas measurements were so reproducible in the short term that many error bars are invisibly small, they varied over weeks, perhaps depending on purity and storage of protein preparations. Therefore, quantitative comparisons have been made only between simultaneous, parallel experiments.
In the experiment shown in Figure  The stoichiometry of Dl/N binding is not known, but seems unlikely to be 1:20. More likely is that much of the adsorbed Notch is not available for binding, a typical finding when protein is adsorbed in non-oriented fashion.
In all remaining experiments, specific binding data are presented from which measured binding to BSA has been subtracted, although such background binding was always low.

EGF repeat O-fucosylation and its contribution to binding specificities
To assess the effect of the split mutation, N11-14 proteins were prepared with either wild type sequences or contain-Binding of Dl:Fc to a Notch protein including EGF repeats 11-12 Figure 2 Binding of Dl:Fc to a Notch protein including EGF repeats 11-12. A. Dl:Fc binding to adsorbed N11-20, measured using HRP-conjugated anti-human IgG. In these and all subsequent experiments, error bars are standard deviations of parallel replicate experiments. Note that binding is saturable and therefore likely to be specific. Saturation binding and "Kd" (half-maximal binding) were estimated by the Prizm software package. B. Human IgG bound negligibly to N11-20, indicating that Dl sequences were responsible for most binding seen in panel A. C. Dl:Fc did not bind adsorbed BSA, so most of the binding in panel A was to adsorbed N11-20:V5:His. In subsequent figures, non-specific BSA binding by Fc fusion proteins has been measured and subtracted, so that binding specific to adsorbed N proteins is presented. D. Dl:Fc did not bind specifically to N21-30, another EGF repeat protein of equal length to N11-20, but lacking the known ligand binding site. Note that the small amount of signal detected is mostly present in the absence of added Dl:Fc protein, is not saturable, and is therefore likely to be nonspecific.
ing the I578T change from the spl mutation that inserts an O-fucosylation site on EGF repeat 14 ( Figure 1). Previous studies show that the T578 mutant is O-fucosylated in Drosophila S2 cells [26]. Like N11-20, the N11-14 protein was bound by Dl:Fc. The I578T substitution had little effect on binding to Dl:Fc( Figure 3). Thus, the spl mutation appears not to affect N function through a direct effect on Dl binding, consistent with EGF repeat 14 not being part of the known ligand binding site [9].  Figure 4A-B,D). By contrast, N21-30:Fc showed saturable binding to N11-20 ( Figure 4C). Although the absolute levels of binding varied between experiments, saturation binding to N11-20 was often higher for N21-30:Fc than for Dl:Fc ( Figure 4E).

Direct interactions between the ligand-binding region and other regions of the N extracellular domain
If N21-30:Fc and Dl:Fc both interact with N11-20 through the same binding site, then we would expect competition for this common site. To test this, soluble Dl:V5His and N21-30:V5His proteins were prepared and added as competitors. Despite the fact that N21-30 and Dl proteins were expected to be monomeric, but Fc-tagged proteins are dimers [33], N21-30 competed for N21-30:Fc binding, and Dl competed for Dl:Fc binding ( Figure 5A,C). In addition, Dl competed for N21-30Fc binding ( Figure 5B), and N21-30 competed for Dl:Fc binding ( Figure 5D). Thus, purified Dl and N21-30 appeared to interact with N11-20 at the same or overlapping sites.

EGF repeat O-fucosylation and its contribution to Abruptex binding
We hypothesized that although the I578T mutation on split did not affect interactions with Delta, it might affect interactions with N21-30.

pH dependence of interactions
In addition to the Dl sup5 mutation, other second-site suppressors of split map in the scabrous and gp150 genes, which encode proteins found in late endosomes [35,36].
To better mimic the endosomal environment, the effect of low pH on interactions of the ligand binding domain was assessed. This required replacement of the Tris pH8.0 binding buffer. Either citric acid or Tris-maleate buffered binding solutions gave similar results at pH 7.4 to those described so far, and both these buffers were also usable at pH 5.4 ( Figure 8A,C). Binding of Dl;Fc to N11-20 increased markedly at pH 5.4, both in terms of maximum binding, and apparent Kd ( Figure 8B). No such increase was seen with N21-30:Fc binding to N11-20, which might be somewhat reduced ( Figure 8D).  (Figure 9).
It has been suggested that ligands need recycling through an endocytic compartment for activation [37][38][39]. Our data raised the possibility that Dl might be activated by acidity during recycling. If this was the case, then we would expect that the 'activation' of DlFc observed at pH5.4 would be irreversible, and maintained after return to neutral pH. By contrast, if Dl:Fc interacts with N better at low pH, this will be reversed at neutral pH. To distin- guish these possibilities, Dl:Fc was pre-incubated at pH 5.4 for 1 h, then half the sample was assessed for binding to N11-20 at pH 5.4, the other half neutralized and assessed for binding to N11-20 at pH 7.4 ( Figure 10). Neutralized Dl:Fc bound to N11-20 less well than Dl:Fc at pH5.4, and not distinguishable from Dl:Fc protein that was never pre-incubated at pH 5.4 ( Figure 10). The results

Competition between Dl and the Abruptex region
indicate that low pH did not activate Dl:Fc irreversibly, but that pH made a direct, reversible contribution to the binding interaction.

Discussion
In vitro studies were performed to explore how proteins interact with the ligand binding region of Notch, with Abruptex binding to the split mutant 3) The interaction of Dl with N is significantly enhanced at pH's typical of endosomes, more acidic than has previously been used to assess Notch function.

The Abruptex domain competes with Delta binding
The EGF repeat 21-30 region of Notch bound to the Notch ligand binding region in vitro, and competed with the Dl extracellular domain for binding( Figures 5, 6). The binding regions were mapped approximately to EGF repeats 11-14 and EGF repeats 22-27, respectively (Figure 7). Other EGF repeat regions of Notch did not bind to the ligand binding region in this assay (Figure 4).
Another recent study has suggested that the Abruptex domain participates in Notch-Notch interactions, based on indirect evidence [21]. The specific interactions pre-dicted are not the same as that which we directly demonstrate here. Our conclusions are conceptually similar, however. It should be noted that our experiments might not detect Notch-Notch interactions that required other stabilizing interactions in the context of the entire Notch molecule, or ternary complexes also involving Delta, should such interactions exist.

O-fucosylation of the split mutant protein does not affect direct Notch-Delta binding
Historically, we sought to evaluate Notch-Notch interactions because of the spl mutation of Notch, and its suppression by the Dl sup5 mutation [26]. If the spl mutation interfered with binding of Notch to the Abruptex domain more than to Delta, this could account for increased Notch activity, which could be suppressed by a further mutation of Dl. We did not detect any effect of the spl mutation on binding to Dl or to the EGF repeats 21-30 domain, however. Protein O-fucose is itself the substrate of Fringe glycosyltransferases, so that EGF repeats can be pH effects the carriers of a short oligosaccharide at this site [22]. Fringe modifies binding interactions and functions of Notch and its ligands, although not all Notch functions depend on Fringe [40,20,[41][42][43][44][45]. It would be interesting to investigate whether modification of O-fucose groups by Fringe enzymes affects interactions between Notch domains, but we did not investigate this because in vivo the spl mutation affects Notch signaling independently of Fringe [26]. Although it is possible that spl does not affect protein-protein interactions of Notch, we cannot exclude an effect on an interaction not studied here, perhaps involving other proteins, should such interactions exist.

pH effects on Notch-Delta interactions
Low pH favored interaction of Dl with N( Figure 8). The findings raise the possibility that Dl might displace Abruptex-domain more effectively after endocytosis. A role for endocytosis in Notch activation would be distinct from the now well-established requirements for Epsin, Neuralized, and Mind-bomb to endocytose Notch ligands, both because these latter molecules act in the ligand-producing cell, and because they are hypothesized to generate an active ligand recycled to the cell surface [46]. By contrast, we found the effect of pH on Dl to be reversible, and therefore not likely to affect ligands after recycling to the cell surface ( Figure 10).
Although genetic evidence supports the idea that endocytosis is required in the receptor-activated cell in addition to the ligand-presenting cell [47], its specific role in Notch signalling is uncertain. It is not certain whether liganddependent Notch cleavage occurs inside the cell [8,48,49].
Dl is also believed to "cis-inactivate" Notch function when co-expressed in the same cells [50][51][52][53]. Recent studies indicate that cis-inactivation occurs at the cell surface, however, and that endocytosis is not required [54].

Possible roles of the Abruptex domain in vivo
Could the Abruptex domain interact with the ligand binding domain in intact Notch molecules in vivo, as in in vitro experiments with protein fragments? Because mutations within the Abruptex domain enhance Notch signaling [13,25,[29][30][31][32], it is reasonable to propose that the Abruptex domain normally interferes with Notch activation. Competition between the Abruptex domain and Delta for the ligand binding site of Notch provides a plausible mechanism by which this could occur. It has also been suggested that the Abruptex domain may be involved in cis-inactivation [31,32]. The structural basis for cis-inactivation is not known.
Notch activation is thought to require 'opening' of a receptor structure that protects the juxtamembrane S2 cleavage site [11]. It is not known how ligand binding to the EGFR11-12 region achieves this, because the S2 cleavage site is in the distant juxta-membrane region of the Notch extracellular domain, separated by the Abruptex domain and by other sequences (Figure 1). Ligand binding to the ligand-binding region could initiate a conformational change by displacing the Abruptex domain [21]. It is not known how this would deprotect the S2 cleavage site, however, especially as Notch molecules lacking EGF repeats 1-18 are not active [55]. One alternative model is that mechanical force unfolds the S2 region in response to ligand binding [11,56] EGF repeats 11-14 and 22-27 from the same molecule seem unlikely to be in proximity if EGF repeats 10-21 form a stiff linear array, as is now thought [12]. Intermolecular Notch-Notch interactions might be possible, however. Cell adhesion assays have not detected homophilic interactions between Notch proteins [57], but it is possible that Notch molecules on different cells cannot interact, for example if inter-molecular interactions have saturated before cells are mixed. In addition, Dl binding to N in such assays requires sequences amino-ter- minal to EGF repeats [58]. No such sequences are present in the EGFR21-30 segment of Notch, which was not able to replace the extracellular domain of Dl in cell adhesion and Notch signalling assays (our unpublished results). Should they occur, inter-molecular Notch-Notch interactions could cluster Notch molecules into dimers or into chains. It will be interesting to determine whether such structures occur and contribute to Notch regulation.

Conclusion
We demonstrate binding in vitro between the Abruptex domain of the Notch extracellular domain and its ligandbinding domain, and propose that a similar interaction in vivo creates a barrier to Notch activation that is overcome by ligand binding, To replace Ile464 with Thr in EGF repeat 11 from Notch, the oligonucleotide ggcctACCtgcgtgaacacaccgggcag was used for oligonucleotide-mediatedmutagenesis. To replace Ser502 with Ile in EGF repeat 12 from Notch, the oligonucleotide cgggATCtgcctggatgatccggggaacg was used for oligonucleotide-mediatedmutagenesis. To generate fucosylated or un-fucosylated 4-EGF repeat proteins, EGF repeats 11-12 containing either a Thr464 mutations to introduce a fucosylation site to EGF repeat 11, or a Ile502 mutation of the fucosylation site on EGF repeat 12, were amplified and each cloned as tandem repeats. In the first reaction the forward primer (from Ser447) was gaagatcttcagaggacatagatgaatg, the reverse primer (to Asp526) was ccatcgatttcgcactgtgtgcccgtg. In the second reaction the forward primer (from Asp449) was ccatcgatatagatgaatgcgatcaggagtc, the reverse primer (to Glu529) wasccgctcgagttcgtccaatgtcgatttcgcac. The two PCR products were digested with BglII/ClaI and ClaI/XhoI respectively and subcloned into BglII/XhoI disgested pMT/BiP/ V5 using a three-way ligation.

Cell culture
Schneider cells were kept at 25°C in Shields and Sang M3 medium (Sigma) supplemented with 10% heat-inactivated fetal bovine serum (Sigma) and 50 U/ml penicillin, 50 U/ml streptomycin (Invitrogen). Cells were transfected using lipofectin as described [59]. Proteins were harvested from serum-free M3 medium 48 h after induction with 0.5 mM CuSO 4 . Secreted His-tagged proteins were purified from conditioned media using ProbondTM resin (Invitrogen), eluted in 50 mM NaH 2 PO 4 , 500 mM NaCl, 300 mM imidazole, pH7.5 and stored at -80°C prior to serial dilution in TBS or other buffers as described for use(TBS: 10 mM Tris-Cl pH 8.0 150 mM NaCl, 1 mM CaCl). Fc-Tagged proteins were purified using Protein A beads (Amersham), eluted with 0.1 M Glycine HCl pH 2.6 into 10 mM Tris pH 7.4 and stored at -80°C prior to serial dilution in TBS or other buffers as described for use. Purity and yield were assessed using a Protein-Assay Kit from Pierce and Coomassie Blue staining of SDS-PA gels. Fctagged proteins were usually recovered ~95% pure, Histagged proteins ~85-90% pure. Storage for more than a few weeks reduced activity in binding assays, although no degradation was apparent.
Binding Assay 200 ng of His-tagged protein eg N11-20V5H was added to each well of a 96-well plate (Apogent) in 50 μL TBS at 4°C overnight. After 3 TBS washes, the plate was blocked with 10 mM Tris-Cl (pH 8.0), 150 mM NaCl, 1 mM CaCl 2 , 3% BSA at 4°C overnight. After 3 washes with TBS containing 0.05% Tween20, known amounts of purified Fc fusion protein (eg Dl:Fc) were added in 50 microlitres of TBS, and after 2 h incubation the plate was washed 3x with TBS + 0.05% Tween20. After incubation with HRP-conjugated anti-human IgG antibody (1:5000 in TBS), the plate was washed three more times in TBS + 0.05% Tween20, and bound antibody detected with the HRP development reagent and quantified with a microplate reader (Perkin Elmer Wallac Victor2). Other pH values were achieved using citric acid/Na2HPO4 or Tris maleate/NaOH buffers in place of Tris pH8.0 [60]. Tris maleate/NaOH appeared to show better Ca solubility.

Binding curves
Data were compiled using Prizm 4 software (GraphPad). Errors (standard deviations from replicate experiments) are shown for all datapoints (but are too small to see in some figures). Molarities bound are expressed on a monomer basis.