Mouse Jagged1 Physically Interacts with Notch2 and Other Notch Receptors

The Delta/Serrate/LAG-2 (DSL) domain containing proteins are considered to be ligands for Notch receptors. However, the physical interaction between DSL proteins and Notch receptors is poorly understood. In this study, we cloned a cDNA for mouse Jagged1 (mJagged1). To identify the receptor interacting with mJagged1 and to gain insight into its binding characteristics, we established two experimental systems using fusion proteins comprising various extracellular parts of mJagged1, a “cell” binding assay and a “solid-phase” binding assay. mJagged1 physically bound to mouse Notch2 (mNotch2) on the cell surface and to a purified extracellular portion of mNotch2, respectively, in a Ca2+-dependent manner. Scatchard analysis of mJagged1 binding to BaF3 cells and to the soluble Notch2 protein demonstrated dissociation constants of 0.4 and 0.7 nm, respectively, and that the number of mJagged1-binding sites on BaF3 is 5,548 per cell. Furthermore, deletion mutant analyses showed that the DSL domain of mJagged1 is a minimal binding unit and is indispensable for binding to mNotch2. The epidermal growth factor-like repeats of mJagged1 modulate the affinity of the interaction, with the first and second repeats playing a major role. Finally, solid-phase binding assay showed that Jagged1 binds to Notch1 and Notch3 in addition to Notch2, suggesting that mJagged1 is a ligand for multiple Notch receptors.

The gene Notch was originally identified in Drosophila melanogaster as playing an important role in appropriate cell fate specification during embyogenesis (1)(2)(3)(4). Although only one Notch gene has been identified in Drosophila (5), multiple Notch homologs have been described in higher vertebrates, including Notch1 through Notch4 in rodents and humans (6 -12). The basic structure of the Drosophila and mammalian Notch proteins comprises 29 -36 epidermal growth factor (EGF) 1 -like repeats and 3 copies of a Lin-12/Notch/Glp motif in the extracellular region, and cdc10/Ankyrin repeats and a PEST-containing domain in the intracellular region.
On the basis of the finding that Notch family proteins can autonomously transduce a signal in various organisms if most of the extracellular region (and transmembrane region) is truncated, they are considered to represent activated forms of Notch proteins (24 -28). Studies with activated forms of Notch1 provide evidence that this protein can regulate differentiation in various types of cells, including C2C12 myoblasts and 32D myeloid progenitors (29 -31). Activated Notch1 also controls CD4/CD8 cell fate (32) as well as ␣␤/␥␦ T cell lineage (33) decisions during normal T lymphocyte development in mice. An attractive model which ties the activated Notch and Notch ligands was recently proposed wherein the intracellular domain of Notch is processed by stimulation with a Notch ligand and translocates into the nucleus. This suggests that ligand stimulation of Notch switches on the intracellular machinery in a manner similar to that of activated Notch (34 -36). Stimulation with a Notch ligand actually inhibits differentiation of various mammalian cells. Jagged1 prevents granulocyte colony-stimulating factor-induced granulocytic differentiation of 32D cells (23) and horse serum-induced myocytic differentiation of C2C12 cells (20), but in both cases only when transfected with full-length Notch1 cDNA. It has also been reported that mDelta1 and mJagged2 can inhibit myocytic differentiation of wild-type C2C12 cells (22,37).
To better understand the biology of the Notch system in mammals, further investigation of the interaction between DSL ligands and Notch receptors is required. The only information obtained to date has come from Drosophila cell-aggregation assay (38,39). A better quantitative approach to the assessment of ligand-receptor association is required.
In the present study, we isolated a cDNA encoding the mouse homologue of Jagged1 (mJagged1) and obtained various forms of this protein containing an extracellular region. Two methods were developed to characterize molecules interacting with mJagged1 and their binding features, a "cell" binding assay using intact cells and a "solid-phase" binding assay using pu-rified proteins immobilized to a plastic tray. These systems provided important information on the interaction between mJagged1 and the Notch receptors.

EXPERIMENTAL PROCEDURES
Oligonucleotides and Probes-A probe to screen a cDNA library by low-stringency hybridization was obtained by polymerase chain reaction using degenerate oligonucleotide primers and mouse embryo cDNA as a template. The primers were synthesized based on the peptide sequences FCRPRDD (amino acid (aa) 199 -205) and PWQCLCE (aa 279 -285), corresponding to the DSL domain and second EGF-like repeat of mDelta1 (19), respectively. Amplified fragments of around 260 bp were subcloned into a TA-cloning vector. Sequencing of these the inserts revealed the amplification of three kinds of cDNA. The inserts were mixed and used as a probe. Northern blot analysis of Notch mRNA was done using the 588-bp 5Ј EcoRV-HindIII fragment and 532-bp SacI fragment, corresponding to similar intracellular domains of mouse Notch1 (mNotch1) (6) and mNotch2, respectively. Mouse Notch2 sequence data is from GenBank under accession number D32210. The 660-bp fragment at the 5Ј end of a newly obtained mJagged1 cDNA was used as a probe to detect Jagged1 mRNA in mouse tissues. Each probe was 32 P-labeled with the Megaprime TM DNA labeling system (Amersham Pharmacia Biotech) according to the manufacturer's instructions.
Isolation, Sequence, and Analysis of cDNA Clones-A mouse cDNA library in gt11 made from an 11-day postcoitum embryo (CLON-TECH) was plated out at 5 ϫ 10 4 plaque forming units on Luria broth/Mg 2ϩ agar according to the manufacturer's instructions. Following incubation for 6 h, plaques were transferred to a nylon membrane (Hybond-N, Amersham Pharmacia Biotech), denatured, neutralized, and hybridized at 42°C in a solution containing 5 ϫ SSC, 1% SDS, 5 ϫ Denhardt's solution, and 30% formamide. Following hybridization, filters were washed three times with 2 ϫ SSC, 1% SDS for 15 min each. Positive clones were isolated through a second and third cycle of hybridization under the same conditions. The cDNA inserts in the isolated clones, DSL20-1 and DSL47-1, which harbored a novel sequence, were subcloned into the EcoRI site of the pBluescript SK-vector (pBS, Toyobo). The 500-bp ApaI fragment in DSL47-1 was replaced with the 1400-bp ApaI fragment from DSL20-1 (Fig. 1A) and the resulting plasmid (pBS-20/47) used to sequence the 4.2-kb cDNA clone by a dyeterminator sequencing method with appropriate internal primers according to the manufacturer's instructions (ABI).
Plasmid Construction-Mouse Notch1 cDNA was the kind gift of Dr. Jeffrey S. Nye. The entire coding region of mNotch2 was obtained by assembling library isolated clones and a 5Ј-rapid amplification of cDNA end clone. The extracellular portion of mNotch3 cDNA was originally obtained by polymerase chain reaction using an 11-day postcoitum mouse embryo cDNA library (CLONTECH) as a template. The sequence of the polymerase chain reaction-derived regions of mNotch2 and mNotch3 was verified by a dye-terminator sequencing method with appropriate internal primers according to the manufacturer's instructions (ABI). A cDNA for the Fc portion of hIgG (hIgG Fc) (40) and a Flag(His) 6 sequence (5Ј-GACTACAAAGACGATGACGATAAACATCA-CCATCACCATCACTAG-3Ј) were fused in-frame to the 3Ј end of the cDNAs encoding each partial extracellular region of mNotch1, mNotch2, and mNotch3, in expression vectors pTraserCMV (CLON-TECH) and pME18S, respectively (41). The mNotch1, mNotch2, and mNotch3 cDNAs were truncated at the codon GAA corresponding to glutamic acid (606th aa for soluble Notch1-Fc (sN1-Fc), 610th aa for sN2-Fc and sN2-Flag(His) 6 , and 586th aa for sN3-Fc). The mJagged1 cDNA was truncated at the codon GAT corresponding to aspartic acid (1067th aa for FE-J1-Fc (for full-length extracellular Jagged1) and ⌬DSL-J1 (for the entire extracellular domain lacking only the DSL domain)); GAT corresponding to aspartic acid (296th aa for EGF1, 2-J1 (for the N terminus through the second EGF-like repeat)); and GCT corresponding to alanine (232th aa for DSL-J1 (for the N-terminal region including the DSL domain but lacking all the EGF-like repeats and cysteine-rich region)). For ⌬DSL-J1, the sequence between nucleotide numbers 821 and 964 corresponding to amino acids 185 through 229 was deleted. These extracellular portions of mJagged1 were again tagged with Flag(His) 6 (FE-J1-Flag(His) 6  Antibodies-For Western blot analysis and cell binding assay using the Flag(His) 6 -tagged proteins, an anti-Flag monoclonal antibody (M2, Eastman Kodak) was used at a dilution of 1:1500. A PE-conjugated anti-mouse secondary antibody (Dako) was then used at a dilution of 1:200. For cell binding assay using the Fc-fused proteins, a PE-conjugated anti-human Fc antibody (Chimicon) was used at a dilution of 1:200. For Western blot analysis and cell-binding and solid-phase binding assays for the Fc-fused proteins, a horseradish peroxidase (HRP)conjugated anti-human Fc antibody (Dako) was used at a dilution of 1:5000. A rabbit polyclonal antibody was raised against the intracellular domain of Notch2 fused in-frame to glutathione S-transferase and used at a dilution of 1:1000 for immunoprecipitation. For Western blot analysis of Notch2, an anti-Notch2 monoclonal antibody (bhN6) (gift from Dr. Spyros Artavanis-Tsakonas) (42) was used at a dilution of 1:20.
Generation of Stable Cell Lines Expressing Soluble Proteins-To establish stable CHO(r) cells expressing soluble Notch receptors and various forms of soluble Jagged1, 5 g of the pTraserCMV plasmid containing cDNAs for these proteins was transfected into CHO(r) cells by the liposome method (TransIT-1, Takara). After 2 days, cell selection was started in the presence of 250 g/ml Zeocin (Life Technologies, Inc.). Expression of the soluble protein produced by each Zeocin-resistant cell line was evaluated by Western blot analysis using an anti-Flag antibody or an enzyme-linked immunosorbent assay using an antibody against the hIgG Fc. The clone showing the highest expression level was chosen for each protein.
Preparation of Soluble Fusion Proteins-Stable CHO(r) cells expressing each soluble Notch receptor or soluble Jagged1 at the highest level were plated in a 15-cm (diameter) dish at a total of 5 ϫ 10 6 cells and incubated overnight. The cells were washed the following day with phosphate-buffered saline and cultured in 40 ml of Dulbecco's modified Eagle's/Ham's F-12 medium for 5 days. The supernatants were then collected and concentrated through a DIALO membrane (Amicon). Each soluble protein was purified from the concentrated supernatant using Protein G Fast Flow (Protein G, Amersham Pharmacia Biotech) or Ni-bound beads (Probond TM , Invitrogen) according to the manufacturers' instructions. A transient expression system was used to obtain deletion mutants of soluble mJagged1 protein: the pME18S plasmid containing a cDNA for deletion mutant of the soluble Jagged1-Fc protein was transfected into COS1 cells by a DEAE-dextran method (44) or a TransIT-1-based liposome method. The approximate concentration of each fusion protein in conditioned medium was estimated by an enzyme-linked immunosorbent assay detecting the hIgG Fc.
Cell Binding Assay-Cell binding assays were done using FACS and colorimetric methods. For FACS analysis, log phase-grown 32D and BaF3 cells were washed in phosphate-buffered saline and aliquoted at 3 ϫ 10 5 cells in 200 l of binding buffer (phosphate-buffered saline containing 2% FBS, 100 g/ml CaCl 2 , and 0.05% NaN 3 ). After blocking with 5 l of rabbit serum, the cells were incubated in cell-binding buffer containing the Fc-or Flag(His) 6 -tagged soluble mJagged1 proteins at room temperature for 1 h. After washing three times with the binding buffer, the cells were incubated with a PE-conjugated anti-hIgG antibody (for Fc-fused proteins) or an anti-Flag antibody, M2, followed by staining with a PE-conjugated anti-mouse IgG (for the Flag(His) 6 proteins). The cells were then analyzed using a FACS Calibur (Becton Dickinson Immunocytometry Systems). A colorimetric method was adopted for the cell binding assay to determine the affinity of mJagged1 to the surface of BaF3. In this method, a binding procedure similar to those above was performed using 0.1% bovine serum albumin instead of 2% FBS in the blocking buffer and a HRP-conjugated anti-hIgG instead of a PE-conjugated antibody as second antibody. The amount of bound FE-J1-Fc was estimated by measuring cell-bound HRP activity using a HRP development reagent (Sumilon). To determine the absolute molar value of incubated and bound FE-J1-Fc, the optical density (OD) value given by the enzyme-linked immunosorbent assay for a concentrationdefined hIgG (Xymed) was utilized. The intensity of color was then measured using a microplate reader (MR700; Dynatech Laboratories).
Northern Blot Analysis-Poly(A) ϩ RNA was isolated using the Poly(A)Ttract TM mRNA isolation system (Promega) according to the manufacturer's instructions. Two micrograms of mRNA from different adult mouse tissues was separated in a 1.2% agarose-formamide gel, transferred to a nylon membrane (Hybond-N, Amersham Pharmacia Biotech) and UV cross-linked. The 5Ј-end 660-bp fragment of mJagged1 was radiolabeled by a random priming method and used as a probe. Hybridization was carried out for 24 h at 42°C in 50% formamide, 5 ϫ SSPE, 5 ϫ Denhardt's solution, 1% SDS, and 0.1 mg/ml salmon sperm DNA. The membrane was subjected to stringent washing twice with 0.2 ϫ SSC and 0.1% SDS at 65°C for 30 min and the blot was visualized with a Fuji BioImage Analyzer BAS2000 (Fuji Film).
Immunoprecipitation and Western Blot Analysis-A total of 1 ϫ 10 7 32D or BaF3 cells were subjected to cell binding assay as described above with 1.7 nM FE-J1-Fc. The cells were then solubilized in a TNP buffer containing 20 mM Tris-Cl (pH 7.4), 150 mM NaCl, 1.0% Nonidet P-40, 5 g/ml aprotinin, and 100 g/ml CaCl 2 for 30 min at 4°C. The lysates were precipitated with Protein G beads which were washed four times with TNP buffer and boiled in the SDS sample buffer under reducing conditions. The samples were subjected to 7.0% SDS-polyacrylamide gel electrophoresis and electrophoretically transferred to a nylon membrane (Immobilon, Millipore). The membranes were blocked with 5% nonfat dry milk in TBST (20 mM Tris-HCl (pH 7.4), 500 mM NaCl, 0.1% Tween 20) for 1 h and incubated overnight at room temperature with an anti-Notch2 monoclonal antibody, bhN6. The membrane was washed three times with TBST and incubated with an alkaline phosphatase-conjugated anti-rat IgG antibody (Promega) for 1 h. Following three washes with TBST, the membrane was visualized using BCNP and nitro blue tetrazolium (Promega).
Solid-phase Binding Assay-Two hundred nanograms of purified FE-J1-Flag(His) 6 or sN2-Flag(His) 6 was added to the wells of a 96-well plate (Dynatech). The plate was blocked with phosphate-buffered saline containing 2.5% gelatin at room temperature for 1 h and Dulbecco's modified Eagle's medium containing purified sN2-Fc or FE-J1-Fc was added. After 2 h incubation, the plate was washed three times with TBST containing 100 g/ml CaCl 2 , followed by addition of a HRPconjugated anti-human IgG antibody. After washing with the same buffer, the color was visualized with the HRP development reagent and the degree of binding was measured using a microplate reader.

RESULTS
Isolation and Sequencing of mJagged1 cDNA-In an attempt to isolate a novel mouse Notch ligand, we screened a cDNA library constructed from an 11-day postcoitum mouse embryo using a low stringency hybridization method. The probe used was a mixture of three DSL sequences, which were obtained by polymerase chain reaction using degenerated primers and mouse embryo cDNA as a template ("Experimental Procedures"). The sequences of these fragments corresponded to the DSL region of mouse Delta1, Jagged1, and Jagged2, the latter two of which have to date only been identified in rats and humans. The nucleotide sequence of a number of cDNA clones isolated revealed that two, DSL20-1 and DSL47-1, represented partial cDNA clones encoding a mouse homologue of Jagged1 and which covered the entire coding region when combined. DSL20-1 contained the 5Ј end of mJagged1 with 268-bp noncoding and 3.2-kb coding regions. DSL47-1 contained 3.6-kb coding and 160-bp 3Ј-end noncoding regions (Fig. 1A). The full-length mJagged1 clone contained an open reading frame of 3,654 bp, which predicted a protein consisting of 1,218 amino acids. Amino acid sequence alignment of mouse, rat, and human Jagged1 shows mJagged1 protein having overall amino acid identities with rat and human Jagged1 (rJagged1 and hJagged1) of 97 and 96%, respectively (Fig. 1B). Amino acid conservation in the DSL and EGF-like repeat regions was high at 98 and 97%, respectively. mJagged1 Gene Expression in Various Mouse Tissues-Examination of mJagged1 expression by Northern blot analysis with mRNAs from various adult mouse tissues and a 14.5-day postcoitum embryo showed that mJagged1 was expressed in various tissues. Highest expression was in the embryo (Fig. 2). Expression in adult tissues was highest in brain, heart, muscle, and thymus. An mRNA transcript of 6.5 kb was detected as a major band; tissues showing this transcript also showed two minor bands of about 4.0 and 3.6 kb. These short transcripts may be alternatively spliced forms of mJagged1 mRNA.
Construction of a Cell Binding Assay System: Specific Binding of Extracellular Region of mJagged1 Protein to Hematopoietic Cell Lines-Two kinds of soluble protein comprising the full-length extracellular region of Jagged1, one with hIgG Fc and the other with Flag(His) 6 tagged to the C terminus, designated FE-J1-Fc and FE-J1-Flag(His) 6 , respectively, were stably produced by cDNA-transfected CHO(r) cells. Coomassie Brilliant Blue staining of Protein G-purified FE-J1-Fc and Nipurified FE-J1-Flag(His) 6 revealed that purity was over 95 and 90%, respectively (Fig. 3A). FE-J1-Fc was found at positions of about 210 kDa under reducing and over 400 kDa under nonreducing conditions, whereas FE-J1-Flag(His) 6 migrated to similar positions of about 140 -180 kDa under both conditions (Fig.  3A). This indicates that, as expected, FE-J1-Fc is dimerized at the Fc portion.
We first investigated the binding of FE-J1-Fc to various cell lines by FACS analysis. FE-J1-Fc bound to all of the cell lines screened, including BaF3 (early B lineage cells), 32D (myeloid lineage cells), CTLL2 (T lineage cells), NSF60 (myeloid leukemia cells), NIH3T3 (embryonic fibroblasts), and C2C12 (myoblastic cells) (data not shown), with binding to 32D and BaF3 greater than to the other cells ( Fig. 3B; other data not shown). FE-J1-Flag(His) 6 also bound to these hematopoietic cells (Fig.  3B). With regard to the characteristics of the Notch/Notchligand interaction, it is known that both Delta and Serrate interact with Notch in a Ca 2ϩ -dependent manner in the Drosophila cell system (39). To determine whether this feature was also conserved in a mammalian system, EGTA, a Ca 2ϩ -chelat-ing reagent, was added to the binding mixture. Binding of FE-J1-Fc and FE-J1-Flag(His) 6 to either the 32D or BaF3 cells was clearly abolished by the addition of EGTA (Fig. 3B).
We further found that the amount of BaF3-bound FE-J1-Fc increased in accordance with the incubation concentration of FE-J1-Fc, reaching a plateau at 1 nM (Fig. 3C). Competition assay in the same experimental system but with an excess of purified FE-J1-Flag(His) 6 added to the cell-binding mixture showed complete abrogation of FE-J1-Fc binding to 32D and BaF3 at an approximately 500-fold molar excess of FE-J1-Flag(His) 6 (Fig. 3D). These results indicate that the binding observed represents specific interaction of the extracellular domain of mJagged1 with a Notch receptor on the surface of the hematopoietic cells.
Determination of the dissociation constant (K d ) of mJagged1 binding to its receptor on the BaF3 cell surface was done using a colorimetric rather than FACS method. Specific colorimetric activity was calculated as 0.092 OD/fmol of hIgG or FE-J1-Fc. Re-evaluation of a saturation binding assay and associated Scatchard plot (Fig. 3E) gave a K d value of about 0.4 nM. The number of binding sites was estimated at 5,548/BaF3 cell (Fig. 3E).
Involvement of Jagged1 DSL Domain and EGF-like Repeats in Binding to Hematopoietic Cells-The roles of the DSL domain and EGF-like repeats in the binding capacity of Jagged1 were examined with three deletion mutants of Jagged1. We first confirmed that FE-J1-Fc derived from COS1 possessed binding activity equivalent to that derived from CHO(r) (data not shown). When COS1-derived purified FE-J1-Fc and three deletion mutants (Fig. 4, A and B) were allowed to bind to 32D at identical molar concentrations, binding activity was greatest for FE-J1-Fc, slightly lower for EGF-1,2-J1-Fc and profoundly lower for DSL-J1-Fc (Fig. 4C). In all cases, binding activity was dependent on Ca 2ϩ (data not shown). ⌬DSL-J1-Fc did not bind at all (Fig. 4C). Essentially identical results were obtained when the same experiment was performed using proteins tagged with Flag(His) 6 instead of hIgG Fc (data not shown). These observations indicate that the DSL domain of Jagged1 confers the minimum binding capability to these hematopoietic cells and that the EGF-like repeats of Jagged1, in particular the first and second, substantially stabilize the interaction.
Interaction of Jagged1 with Notch2 Receptor on the Surface of 32D and BaF3 Cells-We next attempted to identify the Notch receptor interacting with the soluble Jagged1. Candidate receptors were selected through Notch receptor expression in hematopoietic cells by Northern blot analysis. Notch2 mRNA was strongly expressed in both 32D and BaF3. In contrast, full-length (9.5 kb) Notch1 mRNA was not expressed in these cells, although the shorter mRNA species with approximate sizes of 6.0 and 7.2 kb were detected in 32D (Fig. 5A). Furthermore, mRNA for Notch3 or Notch4 was also not detected. Notch2 was therefore selected as the major candidate for receptor interaction with Jagged1. Using an anti-Notch2 monoclonal antibody recognizing a cytoplasmic domain of Notch2, Western blot analysis for the protein G-bound fraction of the lysates of 32D and BaF3 was performed after FE-J1-Fc binding. The results showed that the cytoplasmic domain-containing Notch2 fragments were coimmunoprecipitated with FE-J1-Fc but not with control hIgG (Fig. 5B). One protein species with an approximate molecular mass of 115 kDa was similar in size to that found by other investigators as a fragment of the Notch2 protein (42,45). In contrast, full-length 300-kDa Notch2 was detected in precipitates with an anti-Notch2 polyclonal antibody but not in the Protein G precipitate. These results indicated that the soluble Jagged1 protein binds to Notch2 on the cell surface.

Establishment of a Solid-phase Binding Assay using Purified
Recombinant Soluble Proteins-To further understand the interaction between Jagged1 and Notch, we established a binding assay system (solid-phase binding assay). CHO-derived purified soluble Notch2 proteins were used, namely sN2-Flag(His) 6 and sN2-Fc, consisting of the N terminus through the 15th EGF-like repeat. Purity estimated by Coomassie Brilliant Blue staining using a densitometer was greater than 90 and 95%, respectively (Fig. 6A). As shown in Fig. 6B, direct and specific binding of FE-J1-Fc to the immobilized sN2-Flag(His) 6 was clearly demonstrated. As in the cell binding assay, this interaction was again dependent on the presence of Ca 2ϩ (Fig. 6C). Next, to evaluate the affinity of FE-J1-Fc binding to sN2-Flag(His) 6 , a solid-phase binding assay was performed by adding increasing concentrations of FE-J1-Fc. Results again showed that binding was concentration-dependent and saturable (Fig. 6C). Conversion of the data into a Scatchard plot gave an estimated K d value of about 0.7 nM (Fig. 6C), within a magnitude of that calculated in the cell binding assays. Furthermore, examination of the interaction under the opposite conditions, in which FE-J1-Flag(His) 6 was immobilized and sN2-Fc was used as a probe, showed that sN2-Fc interacted with immobilized FE-J1-Flag(His) 6 but not with the control (Fig. 6D). This interaction was again abolished by the addition of EGTA (Fig. 6D).
Analysis of the functional domain by addition of the deletion mutants of soluble Jagged1 at the same molar concentration (3.3 nM) to the sN2-Flag(His) 6 -immobilized plate showed that both FE-J1-Fc and EGF-1,2-J1-Fc strongly interacted with sN2-Flag(His) 6 , although the amount of EGF-1,2-J1-Fc binding was a little less than that of FE-J1-Fc (Fig. 6E). DSL-J1-Fc bound weakly to sN2-Flag(His) 6 while ⌬DSL-J1-Fc did not bind at all (Fig. 6E). These findings for mutant mJagged1 proteins binding to purified soluble Notch2 were closely similar to those for binding to the 32D cell surface.
Interaction of Jagged1 with Notch1 and Notch3-The solidphase binding assay using purified and immobilized FE-J1-Flag(His) 6 allowed the evaluation of its association with Notch proteins other than mNotch2, namely mNotch1 and mNotch3. Coomassie Brilliant Blue staining of CHO-derived and purified sN1-Fc and sN3-Fc (purity greater than 95%) is shown in Fig.  7A. FE-J1-Flag(His) 6 interacted with sN1-Fc and sN3-Fc as well as sN2-Fc in a dose-dependent manner, although with different binding affinities (Fig. 7B). Among these three soluble Notch proteins, sN3-Fc had highest affinity, followed by sN2-Fc and sN1-Fc in that order. As with Notch2, Ca 2ϩ dependence was maintained (Fig. 7C).

DISCUSSION
In this study, we cloned a cDNA for mouse Jagged1 (mJagged1) and used it to establish two experimental systems for the assessment of the Notch-ligand interaction. Results using these systems have shown that mJagged1 binds to multiple Notch receptors in a saturable and Ca 2ϩ -dependent manner, and that the DSL domain and EGF-like repeats of mJagged1 are critical for binding to Notch2. They have also revealed the affinity of mJagged1 binding to BaF3 cells as well as soluble Notch2 and the number of mJagged1-binding sites on BaF3.
A cDNA obtained from a mouse embryo library showed 97 and 96% homology with rat and human Jagged1 cDNAs, respectively, at the deduced overall amino acid level (Fig. 1), indicating that this isolated cDNA represents the mouse homo- Domain analysis of mJagged1 for binding to 32D. A, schematic representation of the various deletion mutant cDNAs for mJagged1. C terminus of each deletion mutant was fused to hIgG Fc. DSL, DSL region; CR, cysteine-rich region; TM, transmembrane domain. B, secretion from COS1 cells of Fc-fused soluble Jagged1 proteins with various deletions. Each sample was purified with protein G beads and electrophoresed under reducing conditions. Integrity of these proteins was verified by Western blot analysis using an anti-hIgG Fc antibody. Each Fc-fused protein is indicated by an asterisk. C, binding of various Fc-fused soluble Jagged1 to 32D. Cell binding assays with the deletion mutants of mJagged1 were performed at the same molar concentration (3.3 nM).
logue of Jagged1. The mJagged1 mRNA is expressed in various mouse tissues (Fig. 2).
The physical interaction of DSL ligands with Notch receptors has to date been poorly understood. To better understand the binding features of Jagged1 to its receptors, we established two experimental systems which use the various extracellular regions of mJagged1 after purification, a cell binding assay using various live mammalian cells and a cell-free binding system using the extracellular portion of purified recombinant Notch receptors. We designated the latter a solid-phase binding assay, because either the purified Jagged1 proteins or the purified Notch receptors are immobilized. Although the cell binding assay is not the best system for differential analysis of molecules with which mJagged1 interacts, it is conducted under physiological conditions. In contrast, the solid-phase binding assay is more artificial, but is a powerful tool in the specification and differential analysis of the receptor molecules. Importantly, we observed that in many aspects mJagged1 bound to the cell surface and purified soluble mNotch2 in a remarkably similar manner.
Our initial investigation showed that, among various Notch receptors, mJagged1 interacts primarily with Notch2, at least in the two hematopoietic cell lines 32D and BaF3. That is, whereas full-length mRNAs for Notch1, -3, and -4 were not detected in these cells, strong expression of mRNA for Notch2 was detected. Furthermore, soluble Jagged1 coprecipitated with Notch2 protein expressed on the surface of these cells (Fig.  6). In an immunoprecipitation analysis, the full-length 300-kDa Notch2 species was not coprecipitated with Jagged1, al-though it was precipitated by an anti-Notch2 polyclonal antibody. This is consistent with the recent finding that plasma membrane Notch protein comprises two polypeptide chains as a result of proteolytic cleavage of the full-length Notch protein (45,46). A 115-kDa protein, found by other investigators as a fragment of the Notch2 protein (42,45), and multiple shorter Notch2 species of about 70 -80 kDa were coprecipitated with Jagged1. We suggest that the 70 -80-kDa proteins are the products of the 115-kDa species degraded by metalloproteinases during the immunoprecipitation procedure. EDTA, an inhibitor of metalloproteinases, inhibited the appearance of these fragments (data not shown) but was not added to the ligand-Notch interaction due to the expectation that it would cancel the binding.
Binding assays with added EGTA showed that the interaction of Jagged1 with the cell surface, as well as with purified Notch1, Notch2, and Notch3, depends on the presence of Ca 2ϩ . This characteristic has also been described in the Drosophila Notch system (39), in which it was reported that a Ca 2ϩbinding site exists in each of the 11th and 12th EGF-like repeats, which are necessary and sufficient for the interaction with both Delta and Serrate (39). This 11th and 12th EGF-like repeat sequence which included Ca 2ϩ -binding sites is strongly conserved among all the mammalian Notch receptors (39,47).
Scatchard analysis in the cell-binding and solid-phase binding assays gave K d values of approximately 0.4 and 0.7 nM, respectively, indicating that the affinity of the full-length extracellular domain of mJagged1 for the cell surface of BaF3 is slightly higher than that for the purified partial extracellular portion of mNotch2. As the difference is surprisingly small, however, we suggest that the N-terminal region through to the 15th EGF repeat of mNotch2 plays a major role in binding to Jagged1. Affinity of Jagged1 binding to purified Notch2 might have been as high as that to the cell surface if the full-length extracellular Notch2 protein had been used.
With respect to the roles of each domain in Jagged1, deletion mutant analyses of Jagged1 clearly demonstrated in both binding assays that the DSL domain and the EGF-like repeats are critical for binding to Notch2 (Figs. 4C and 6E). The DSL domain is indispensable and the minimal unit for binding to Notch. Our finding that the DSL domain is a receptor-binding site is consistent with previous speculation that an oligonucleotide corresponding to the DSL domain could be biologically active (23,48,49). The DSL domain is essential to the biological activity of LAG-2 and APX-1 in C. elegans (48,49), and it has been speculated that the lack of the DSL domain in LAG-2 or APX-1 impairs binding activity of DSL proteins for Notch. Our results support this idea. Regarding the function of the EGFlike repeats in the DSL ligands, the 4th EGF-like repeat in Drosophila Delta is speculated to be involved in maintaining stable affinity for Notch (50). Results from both the present assays show that the EGF-like repeats, particularly the 1st and 2nd, function in the formation of the high affinity complex with Notch2 (Figs. 4C and 6E).
The solid-phase binding assay showed that Jagged1 bound to Notch3, Notch2, and Notch1 with diverse affinities in this order (Fig. 7). However, it cannot be concluded that this apparent difference invariably reflects the characteristics of interaction between mJagged1 and each Notch polypeptide, due to the possibility that binding affinities between mJagged1 and Notch receptors are modified by Fringe proteins, putatively secreted glycosyltransferases which may confer glycosyl chains to Notch. In this regard, it is known that Fringe modulates the biological activity of Serrate and Delta in Drosophila (51-53). Thus far, stress has been put on Notch1 as a natural ligand for Jagged1 (20,23). However, our observations suggest that  1 and 4) was allowed to bind to 32D or BaF3. Immunoprecipitation was performed with Protein G alone ( lanes  1, 2, 4, and 5). To identify the Notch2 protein fragments, lysates of 32D and BaF3 were precipitated with an anti-Notch2 rabbit polyclonal antibody (lanes 3 and 6). These precipitates were analyzed by Western blot analysis probed with a Notch2-specific monoclonal antibody. Size marker protein positions are shown at the left. Bands of approximately 115, 80, and 75 kDa represent Notch2 intracellular region fragments. f-Notch2, full-length Notch2. IP, immunoprecipitation. Jagged1 is potentially a natural ligand for multiple Notch receptors. This suggestion agrees with the results from in situ hybridization analyses (20,54).
In this paper, we describe a useful experimental approach which provides direct evidence that one of the DSL proteins, mJagged1, directly associates with multiple mouse Notch pro-FIG. 6. Solid-phase binding: binding of soluble Jagged1 to soluble Notch2. A, generation and purification of the Flag(His) 6 -tagged partial extracellular region of mouse Notch2. Integrity of purified sN2-Flag(His) 6 and sN2-Fc was verified by Coomassie Brilliant Blue (CBB) staining and Western blot analysis. B, binding of FE-J1-Fc to immobilized sN2-Flag(His) 6 . The interaction between soluble Jagged1 and soluble Notch2 was examined as described in the "Solid-phase Binding Assay" section under "Experimental Procedures," in the absence (open square) or presence (hatched square) of 2 mM EGTA. The same concentration of hIgG was added as a control. C, Scatchard analysis of FE-J1-Fc binding to immobilized sN2-Flag(His) 6 . A saturation binding curve (inset, mean value of duplicates and standard deviation for each protein) and Scatchard plot are shown. Calculated K d is shown. D, binding of sN2-Fc to immobilized FE-J1-Flag(His) 6 in the absence (open square) or presence (hatched square) of 2 mM EGTA. E, domain analysis of Jagged1 for binding to sN2-Flag(His) 6 . Purified FE-J1-Fc, EGF1,2-J1-Fc, DSL-J1-Fc, or ⌬DSL-J1-Fc at the same concentration (3.3 nM) was allowed to bind to immobilized sN2-Flag(His) 6 6 . A, generation and purification of Fc-fused partial extracellular region of mouse Notch receptors. Integrity of sN1-Fc, sN2-Fc, and sN3-Fc purified with protein G beads was evaluated by Coomassie Brilliant Blue (CBB) staining and Western blot analysis with anti-hIgG Fc antibody. B, binding of increasing concentrations of sN1-Fc, sN2-Fc, and sN3-Fc to immobilized FE-J1-Flag(His) 6 . C, Ca 2ϩ -dependent binding of sN1-Fc, sN2-Fc, and sN3-Fc to FE-J1-Flag(His) 6 . Purified sN1-Fc, sN2-Fc, or sN3-Fc at the same molar concentration (6.7 nM) was allowed to bind to immobilized FE-J1-Flag(His) 6  teins. We believe these findings add to the further understanding of Notch signaling.