A soluble form of human Delta-like-1 inhibits differentiation of hematopoietic progenitor cells

A human heart λ uni-Zap XR cDNA library (Stratagene, La Jolla, CA) was used to screen hDll1 cDNA using³²P-labeled mDll1 cDNA18 as a probe by standard methods.35 DNA sequencing was performed at Cornell University DNA Sequencing Center (Ithaca, NY) using custom oligonucleotide primers from Gibco-BRL Life Science Technology (Grand Island, NY). DNA and protein sequence analysis was performed using DNA Strider software, and database searches were accomplished by using the BLAST network service at the National Center for Biotechnology Information.36 The full-length hDll1cDNA in pBluescript was in vitro translated in a coupled transcription/translation reaction using the TNT-coupled reticulocyte lysate system (Promega, Madison, WI).

The pTrx/mDll1^NDSL plasmid was constructed by subcloning the polymerase chain reaction (PCR)-amplified mDll1^NDSLregion of the mDll1 cDNA into an Escherichia coliexpression vector pTrxFus (Invitrogen, San Diego, CA). The PCR was carried out with 5' primer (5'- GATCTCTAGACCTCCATACAGACTCT), 3' primer (5'-GATCGTCGACAGATTGGGTCAGTGCA), and mDll1cDNA template.18 The mDll1^NDSLincludes the DSL domain and its adjacent N-terminal 50 amino acids. The fusion protein recovered in the soluble fraction of a liter of bacteria culture was purified by gel filtration chromatography and separated further by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE). The fusion protein was used to immunize chickens. The IgY antibodies were purified from the egg yolk and by the thioredoxin-mDll1^NDSL affinity column.

Western blot analysis was performed using the ECL luminescence detection method (Amersham Life Science, Arlington Heights, IL). Fifteen micrograms protein was analyzed for each sample. The affinity-purified anti-mDll1^NDSL antibody was used at 1:100 dilution. Horseradish peroxidase-labeled rabbit antichicken IgY was then used at 1:5000 dilution. The MS-5 cell line was fixed with cold acetone and incubated with avidin D and biotin blocking solutions (Vector Laboratories, Burlingame, CA), which all contained 2% chicken serum, 1:50 dilution of IgY, and 1:50 dilution of anti-thioredoxin-mNotch1 antibody raised in chicken. Biotinylated anti-mDll1^NDSL antibody was used at 50 μg/mL. Biotinylated IgY was used as negative control. The detection system was carried out using a Vectastain ABC-AP kit (Vector Laboratories).

The hDll1^NDSL region includes the DSL domain and its adjacent N-terminal 50 amino acids (aa 127-225). The cDNA fragment encoding this region was amplified by PCR and was subcloned into the expression vector pGEX-2T to create pGEX-hDll1^NDSL. The PCR was carried out with 5'-primer (5'-CGGGATCCCTCCACACAGATTCTCCTG), 3'-primer (5'-CGGAA TTCTTAGATCGGCTCTGTGCAGTAG), andhDll1 cDNA template. pGEX-hDll1^NDSL and pGEX-2T were induced to express the GST-hDll1^NDSL fusion protein and GST on IPTG induction, respectively.

Bacterial pellet fractions containing GST-hDll1^NDSL were resuspended in 8 mol/L urea buffer A (50 mmol/L Tris/HCl, pH 8, 1 mmol/L EDTA, 50 mmol/L NaCl, and 0.1 mmol/L phenylmethylsulfonyl fluoride) and were renatured by dilution to alkaline buffer without urea. Both GST and the renatured fusion protein were purified through a DEAE Sepharose CL-6B column (Pharmacia Biotech, Piscataway, NJ) with a bed volume of 90 mL. The column was washed by buffer A and eluted with NaCl gradient buffer A (50-500 mmol/L). GST or GST-hDll1^NDSL in the positive DEAE fractions was further purified with affinity chromatography using a glutathione Sepharose 4B column (Pharmacia Biotech). To release hDll1^NDSLfrom the fusion protein, GST-hDll1^NDSL in the DEAE fractions was bound to the glutathione Sepharose 4B column. The column was washed with 4-column volumes of buffer B (50 mmol/L Tris-HCl, pH 8.3, 150 mmol/L NaCl) and then with 2.5 mmol/L CaCl₂in buffer B. Thrombin protease (Pharmacia Biotech) was added at 1 U/100 μg fusion proteins. After digestion overnight, the hDll1^NDSL was recovered from the glutathione column.

The protein preparations of GST, GST-hDll1^NDSL, and hDll1^NDSL were dialyzed in phosphate-buffered saline (PBS) and filtered through 0.2-μmol/L filters. Protein concentrations were assayed with the Bio-Rad protein assay solution (Bio-Rad, Hercules, CA). Endotoxin levels in the final protein preparations were measured using an endotoxin kit (Sigma, St Louis, MO).

A StemSep system (Stem Cell Technologies, Vancouver, Canada) was used to isolate bone marrow progenitor cells from the Balb/C mice. Monoclonal antibodies to the following murine cell-surface antigens were included: T lymphocytes (CD5), B lymphocytes (CD45R), macrophages (Mac-1), granulocytes (Gr-1), and erythroid (TER 119). The purity of the cells was verified by fluorescence-activated cell sorting (FACS) of Gr-1 and Mac-1 expression, which was less than 5%.

Lin⁻ cells were cultured in serum-free medium (Ex-vivo 15; Biowhittaker, San Diego, CA) supplemented with cytokine cocktails. The cytokines used in this study included interleukin-3 (IL-3; 10 ng/mL), c-kit ligand (KL; 20 ng/mL), GM-CSF (1000 U/mL), IL-11 (100 ng/mL), Flt3 ligand (FL; 100 ng/mL), G-CSF (1000 U/mL), IL-1β (100 U/mL), IL-6 (20 ng/mL), and erythropoietin (Epo; 10 U/mL). Cytokine cocktails, GST, and GST-hDll1^NDSL proteins were added to the cultures according to individual experiments (see “Results”). Cultures were incubated for 7 to 12 days at 37°C in a fully humidified 5% CO₂atmosphere.

Hematopoietic cells were washed with 2% fetal calf serum (FCS) in PBS at 4°C. The cells were resuspended in 0.1 mL 2% FCS in PBS with 1 μL Fc blocker (Pharmingen, San Diego, CA). Antibodies used were fluorescein isothiocyanate (FITC)-conjugated rat antimouse Mac-1 and antimouse Gr-1 (Pharmingen). One microliter antibody was added to the cells. After 30 minutes of incubation, the cells were washed twice in 2% FCS/PBS and fixed in 1% formalin/PBS.

Cell-cycle analysis was performed using DNA binding dye propidium iodide (PI). Hematopoietic cells were fixed in 50% ethanol and resuspended to 0.2 mL of 10 mg/mL RNAaseA and 50 μg/mL PI. Hematopoietic cells stained with PI in the sub-G₁ phase of the cell cycle represented the apoptotic bodies and were measured as apoptosis events by the method of Fraker et al..37 Analysis of Mac-1/Gr-1 expression and cell-cycle kinetics was performed by our Institute Core Facility using the FACSCalibur flow cytometer (Becton Dickinson, San Jose, CA).

Cells were plated in 1 mL IMDM medium containing 1.2% (wt/vol) methylcellulose, 30% FCS, 2 mmol/L glutamine, 0.1 mmol/L 2-mercaptoethanol, and 4 mmol/L hemin plus cytokines (IL-3, GM-CSF, KL, and Epo). Triplicate cultures were incubated for 7 days at 37°C in a fully humidified 5% CO₂ atmosphere, and colonies were counted under a microscope. CFU-S was measured by the injection of progenitor cells into irradiated Balb/C mice at 5 mice/group. Spleen colonies were counted after 12 days as CFU-S_day12.

Results are presented as the mean ± standard error (SE). Significance was determined using the two-tailed paired Studentt-test.

To clone a human Delta homologue, full-length mouseDelta-like-1 (mDll1) cDNA, generously provided by A. Gossler,18 was used as a probe to screen a human heart cDNA library (Stratagene). Six positive clones were isolated after the tertiary screening. The sequences of these clones showed the highest similarity to mDll1. We concluded that these clones contained partial cDNA fragments of the human homologue of mDll1(hDll1). A 2.1-kb clone was found to contain the 3' end ofhDll1 with 848 bp of untranslated region and 1252 bp of coding region by sequence analysis. The other 1.2-kb clone contained 511 bp of 5' untranslated region and 689 bp of 5' coding region. There was a gap of 228 bp between the two clones. To obtain a complete cDNA clone of hDll1, primers were designed according to the sequences from the two clones, and RT-PCR was performed using total mRNA isolated from human bone marrow endothelial cells, generously provided by S. Raffi (Weil Medical College, Cornell University). A cDNA fragment of 1080 bp was obtained and cloned into plasmid pBluescript. A full-length cDNA ofhDll1 (3039 bp) was assembled by combining the PCR fragment and the 2.1-kb clone.

This cDNA had one open reading frame of 2169 bp and was predicted to encode a protein product having 723 amino acids (Fig.1). The full-length cDNA was successfully translated in vitro to produce a 79-kd protein at the predicted molecular weight (Fig. 2D). Analysis of the amino acid sequence indicated that hDll1 is a transmembrane protein with a large extracellular domain (aa 1-543) and a short intracellular domain (aa 573-723). The hDll1 protein shares more structural features with the Delta family than with the Serrate/Jagged family, including a DSL motif and 8 EGF-like repeats within the extracellular domain.

Figure 1 also shows the alignment of the amino acid sequences of hDll1 and mDll1. The hDll1 protein has 90% overall amino acid identity with mDll1, and it is longer than mDll1 by 1 amino acid. The extracellular domains have higher degrees of conservation (89%) than the intracellular domain (82%), and the N-terminal plus DSL sequences show the highest conservation (95%). In addition, comparison of hDll1 withDrosophila Delta shows that the best-matched sequences are in the extracellular domain, with 48% amino acid identity. However, there was only 41% identity when hDll1 was compared with human Jagged1, a member of Serrate/Jagged family, even in the most conserved extracellular domain. It is clear that members of the same Notch ligand family, ie, Delta or Serrate/Jagged family, in different species, share higher sequence identity than the members of the ligands in different families in the same species.

To assess the expression of Dll1 proteins, cell extracts prepared from multiple tissues were analyzed by Western blot using a chicken polyclonal antibody that was raised against mDll1^NDSL. A negative control was obtained using the preimmune antibody (Fig. 2A). No signals were seen in this blot in the predicted position of mDll1. A major band of molecular weight 79 kd, the correct molecular weight for mDll1 and hDll1, was detected in the blot with anti-mDll1 antibody (Figs. 2B, 2C). It is expressed at high levels in the mouse lung, thymus, and bone marrow and at lower levels in small intestine, brain, spleen, and liver (Fig. 2B). Because hDll1 shares high homology with mDll1 (95% identity in the region used to raise the antibody; Fig.4A), the antibody also detected the protein in human umbilical vein endothelial cells (HUVEC) and in HUVEC treated with interferon-α and IL-1 for 24 hours (Fig. 2C). No major difference was observed in the expression of hDll1 in untreated HUVEC versus cytokine-activated HUVEC. The sizes of mDll1 and hDll1, detected by the Western blot, matched the size of the in vitro translated hDll1 protein (Fig. 2D).

We examined the expression of mDll1 in MS-5 stromal cell line and primary stromal cells using immunohistochemical staining with our biotinylated anti-mDll1^NDSL antibody. MS-5 is a mouse bone marrow stromal derived fibroblast cell line that supports both mouse and human long-term hematopoiesis in vitro.38 39 mDll1 was detected in the MS-5 cells (Fig. 3B), whereas no staining was observed with the biotinylated isotype IgY control (Fig. 3A). To confirm the specificity of the staining, the anti-mDll1^NDSL antibody raised against thioredoxin-mDll1^NDSL was preabsorbed with GST-hDll1^NDSL-coupled agarose and also with the control GST agarose. We found that the anti-mDll1^NDSL antibody staining in MS-5 cells was completely abolished with the GST-hDll1^NDSL preabsorbed antibodies (Fig. 3C). We also detected mDll1 expression using a similar approach in primary bone marrow stromal cells cultured for 2 weeks in 20% FCS in RPMI medium (Figs. 3E, 3F), whereas no staining was observed in the negative control (Fig. 3D). The mDll1 staining appeared mostly in the cytoplasm, with stronger staining in the perinuclear areas of the cells but not in the nuclei (Fig. 3E). We concluded that mDll1 was expressed in a stromal-derived fibroblast cell line and in primary bone marrow stromal cells.

We initially created a thioredoxin-mDll1^NDSL protein containing the DSL domain of mDll1 and its adjacent N-terminal 50 amino acids using a pTrxFusion vector-mediated expression system. We observed that the fusion protein enhanced the expansion of primitive hematopoietic precursors, inhibited differentiation of myeloid progenitor cells, and suppressed apoptosis of hematopoietic cells (data not shown). However, the control thioredoxin protein showed inhibitory effects in the colony formation by progenitor cells, which might have compromised the function of the fusion protein. We, therefore, made a GST-hDll1^NDSL fusion protein containing the DSL domain of hDll1 and its adjacent N-terminal 50 amino acids (Fig.4A). GST-hDll1^NDSL was expressed and purified from the E. coli lysate. The fusion protein was estimated to be more than 95% pure by Coomassie-blue staining after purification by DEAE- and glutathione-affinity columns.

To remove GST from the fusion protein, the fusion protein bound to the glutathione beads was digested with thrombin protease. The released hDll1^NDSL protein was recovered from the glutathione beads with limited GST contamination (Fig. 4B). It was difficult to remove the contaminated GST from hDll1^NDSL even when the digested protein was passed through the glutathione-affinity column twice, suggesting that GST is probably in close association with hDll1^NDSL and that protein–protein interactions may help to stabilize hDll1^NDSL in solution.

The sequence of GST-hDll1^NDSL in the pGEX-2T vector was verified by DNA sequencing, and the nature of the recombinant protein was examined in Western blot with anti-mDll1^NDSL antibody (Fig. 4C). Thioredoxin-mDll1^NDSL, GST-hDll1^NDSL, GST, and hDll1^NDSL were separated by SDS-PAGE in a 12% nonreducing polyacrylamide gel. Thioredoxin-mDll1 was detected as a 34-kd band, GST-hDll1^NDSL as a 43-kd band, and hDll1^NDSL as a 17-kd band. No band was detected in the GST lane. We noticed that both thioredoxin-mDll1^NDSL and GST-hDll1^NDSLformed polymers in solution, but not hDll1^NDSL.

To test the function of the GST fusion protein, we first examined the possible hematopoietic effects of GST. Hematopoietic progenitor cells were isolated from adult mouse bone marrow by immunomagnetic negative selection. Lineage-negative (Lin⁻) progenitor cells that had not bound the antibodies against the lineage-associated cell surface markers were isolated. More than 95% of the isolated Lin⁻ cells were Mac-1 negative, the marker for murine monocytes and granulocytes, as analyzed by flow cytometry; 10⁴ Lin⁻ cells/mL was cultured in 24-well plates in serum-free medium supplemented with IL-3, IL-1, and GM-CSF. Bovine serum albumin bovine serum albumin and GST were added to the above cultures, as indicated in Table1. After 7 days, the cultures were analyzed for their cellularity, total CFU, Mac-1 expression, and cell cycles. Data collected from three independent experiments were analyzed using overall F test. The differences in the four parameters among 6 tested protein groups were not statistically significant (Table 1). We concluded that GST used at concentrations of up to 0.4 μmol/L in the liquid culture of Lin⁻ cells with serum-free medium was without effect on the hematopoietic cells.

To determine the effective dosage of GST-hDll1^NDSLin the regulation of hematopoietic differentiation, we titrated the fusion protein in the hematopoietic assays. GST and the fusion protein were added to the cytokine-supported cultures of Lin⁻ cells at different concentrations, as indicated in Figure 5. Cultures were analyzed for the expression of Mac-1. Data from three independent experiments showed that Mac-1 expression was inhibited by the addition of the fusion proteins but not by the GST controls (Fig. 5). The inhibitory effect was protein concentration dependent, with the most effective dosage at 0.3 μmol/L.

To examine the hematopoietic effects of GST-hDll1^NDSL, 3 × 10⁴ Lin⁻ cells were added to a single well of a 6-well plate containing 3 mL serum-free medium plus cytokines (IL-3, GM-CSF, and IL-1). GST-hDll1^NDSL and control protein GST were added in 0.1 μmol/L every 3 days. After culture for 7 days, the cells were evaluated for their Mac-1 and Gr-1 expression, progenitor cell expansion, apoptosis, and cell-cycle kinetics. Data from five independent experiments are summarized in Table 1 and Figure 6.

Cells were counted using the trypan blue exclusion method. Live and dead cells were counted separately after the Lin⁻cells were cultured for 7 days. Cellularity analysis from five independent experiments are summarized (Table2). Progenitors cultured with GST-hDll1^NDSL gave rise to a similar number of total cells compared with controls. The viable cells in the GST-hDll1^NDSL culture represented 82% of the total cells compared with 67% in the controls (Table 2). Thus, GST-hDll1^NDSL appeared to inhibit cell death.

Cell differentiation was examined by flow cytometry using anti-Mac-1 (detecting monocyte and granulocyte) and anti-Gr-1 (detecting granulocyte) antibodies. In the five independent experiments, Lin⁻ cells cultured with GST-hDll1^NDSLshowed consistently lower percentages of cells expressing Mac-1 and Gr-1. As shown in Table 2, the mean percentage of cells expressing Mac-1 at day 7 was 73% in the population cultured with GST-hDll1^NDSL versus 84% in the control population. Similarly, the mean percentage of cells expressing Gr-1 was 50% versus 76%. These differences in expression of Mac-1 and Gr-1 are statistically significant (P < .01). In addition, the median intensity of fluorescence (MIF) of cells expressing Mac-1 and Gr-1 was significantly decreased when the cells were treated with GST-hDll1^NDSL. A representative experiment is shown in Figure 6A in which at day 7, 78% (MIF, 157) and 67% (MIF, 71) of the cells exposed to GST-hDll1^NDSL ligand expressed Mac-1 and Gr-1 (bottom panel), respectively, compared with 92% (MIF, 253) and 88% (MIF, 1186) of control cells (middle panel).

Cell-cycle profiles were analyzed by cell sorting of the PI-stained hematopoietic cells derived from the Lin⁻ cells cultured for 7 days. In the five independent observations (Table 2), Lin⁻ cells cultured with GST-hDll1^NDSL showed a significantly lower fraction of cells in G₀/G₁ phase (73%) than control GST-treated cells (86%, and a higher fraction of cells in S phase (21%) than control cells (7%). The differences in cell cycles between the two groups were statistically significant (P = .01).

Because GST-hDll1^NDSL appeared to inhibit cell death (Table2), the apoptotic cell population in the sub-G₀/G₁ phase was analyzed during the process of cell-cycle data collection. Cells treated with GST-hDll1^NDSL had a significantly reduced apoptotic cell population (14%) compared with the control cells cultured with GST (46%) (Table 2). The difference between the two groups was statistically significant (P = .005).

The effect of GST-hDll1^NDSL on the expansion of progenitor cells was analyzed by clonogenic assays. The number of colonies generated after the Lin⁻ cells were cultured for 7 days (day 7 CFU) was divided by the input colonies, giving the fold expansion of the colony-forming progenitors during the culture. Data from the five independent experiments are presented (Fig. 6B). Lin⁻ cells cultured with GST-hDll1^NDSLshowed significantly greater (3.3 fold) expansion of progenitors compared with the GST controls (Total CFU). In addition, GST-hDll1^NDSL promoted progenitor cell expansion in each category of progenitors including CFU-mix (4.6 fold), CFU-GM (4.2 fold), early BFU-E (3.9 fold), and late BFU-E (3.2 fold) over the control cultures with GST.

In summary, we investigated the phenotypic and functional changes of Lin⁻ progenitor cells cultured with GST-hDll1^NDSL compared with the cells treated with GST in conditions promoting myelopoiesis. Data from five independent experiments demonstrated that GST-hDll1^NDSL partially suppressed differentiation and promoted expansion of myeloid progenitor cells. We also observed that the fraction of cells in the S phase of the cell cycle was increased and that the percentage of apoptotic cells was decreased.

To control the possibility that the effects of GST-hDll1^NDSL fusion protein were caused by the novel fusion polypeptide rather than by the hDll1^NDSL sequences, we examined the function of hDll1^NDSL released from the fusion protein. hDll1^NDSL was released by digestion of the fusion protein bound to the GST-affinity beads with thrombin protease. Because the free GST and thrombin were not totally removed from the hDll1^NDSL fraction (Fig. 4), 4 protein groups were compared: (1) 0.1 μmol/L GST; (2) 0.1 μmol/L GST-hDll1^NDSL; (3) 0.1 μmol/L GST and 30 ng/mL thrombin (the same amount used to digest the fusion protein); (4) 0.1 μmol/L hDll1^NDSL, 0.1 μmol/L GST, and 30 ng/mL thrombin. The Lin⁻ cells were cultured for 7 days at 104cells/mL (total of 3 ml per well in 6-well plates) in the serum-free medium with a cytokine cocktail of IL-3, GM-CSF, and IL-1. Under these culture conditions, the 4 protein groups were added every 3 days. Cell proliferation, differentiation, and apoptosis of the cultured cells were evaluated. Data from three independent experiments are summarized in Table 3. Essentially, the hDll1^NDSL effects were similar to those seen with the GST-hDll1^NDSL fusion protein. The numbers of live cells were increased in the GST-hDll1^NDSL and hDll1^NDSL group over the two control groups. The numbers of progenitor cells detected as CFU were all significantly higher in the GST-hDll1^NDSL and hDll1^NDSL groups than in the controls. The hDll1^NDSL group also showed consistently lower percentages of cells expressing Mac-1 and Gr-1 than the control GST/thrombin group. Like the fusion protein, the hDll1^NDSLgroup also showed a significantly lower fraction of cells in the G₀/G₁ phase than the control GST/thrombin-treated cells and a higher fraction of cells in the S phase than the control cells. The differences between the GST-hDll1^NDSL and the GST groups and between the hDll1^NDSL and the GST/thrombin groups were statistically significant. Those between the GST-hDll1^NDSL and the hDll1^NDSL groups and between the GST and the GST/thrombin groups were not statistically significant. In conclusion, the results demonstrated that the GST-hDll1^NDSL fusion protein retained the function of hDll1^NDSL in regulation of hematopoiesis. The observed hematopoietic effects of the fusion protein reflected the function of hDll1^NDSL.

After defining the role of GST-hDll1^NDSL in the regulation of Lin⁻ cells, a population that included late-stage progenitors (44% of which are in the S phase of the cell cycle) and stem cells, we extended the functional analysis of GST-hDll1^NDSL to the early-stage hematopoietic progenitors and stem cells. These cells were enriched in Lin⁻cells isolated from 5-fluorouracil-treated mouse bone marrow. The specificity of the GST-hDll1^NDSL effects was also tested by the use of the neutralizing anti-thioredoxin-mDll1^NDSLantibody (anti-mDll1). Five thousand cells were added to a single well of 24-well plate containing 1 mL serum-free medium plus cytokines (IL-11, KL, and FL). Four protein groups were tested: (1) 0.1 μmol/L GST; (2) 0.1 μmol/L GST-hDll1^NDSL; (3) 0.1 μmol/L GST-hDll1^NDSL plus 5 μg/mL anti-mDll1; and (4) 0.1 μmol/L GST-hDll1^NDSL plus 5 μg/mL preimmune serum. Proteins were added to the cultures every 3 days. Cells were counted after culture for 12 days, and CFU cells including HPP-CFC (colony greater than 0.5 mm) and CFU-GM were measured in methylcellulose assay. Early-stage progenitors in the suspension population of cells were read out in standard CFU-S_day12 assay by injecting the cultured cells into the irradiated mice. Twelve days later, the spleen colonies were counted. Data from 3 independent experiments are summarized in Table 4. Cells cultured with GST-hDll1^NDSL showed a 2.5-fold increase in HPP-CFC, a 1.6-fold increase in CFU-GM, and a 4-fold increase in CFU-S_day12 compared with the control cells cultured with GST. Furthermore, the GST-hDll1^NDSLeffects were significantly blocked by the presence of anti-mDll1 antibody but not affected by the preimmune serum. The anti-mDll1 was raised against thioredoxin-mDll1^NDSL, which showed 92% identity with the corresponding region of hDll1 (Fig. 4A). Thus, the data in Table 4 demonstrate that hDll1^NDSL promoted the expansion of noncycling primitive progenitors. The hDll1^NDSL effect was blocked by the anti-mDll1 antibody, which further confirmed that the function of the fusion protein was hDll1^NDSL sequence dependent.

In this article, we describe the isolation of a cDNA clone encoding a human Delta ligand (hDll1) for Notch and the creation of a soluble form of the ligand. A role for this soluble hDll1 in the regulation of hematopoiesis in vitro was revealed. Although hDll1 shares all the conserved features with other DSL proteins, it is most closely related to mDll1. Expression of mDll1 was detected in a variety of adult mouse tissues, including hematopoietic tissues of spleen and bone marrow. Together with Jagged1 and Jagged223,30-32 then, 3 DSL ligands are expressed by bone marrow stromal cells. Because Notch1 and Notch2 were detected in hematopoietic cells,30,40 it is likely that multiple DSL ligands and Notch receptors are involved in the regulation of bone marrow hematopoiesis. Recently, Radtke et al41 reported a conditional knock-out of Notch1 in mouse. In a competitive repopulation experiment, Notch-1-deficient bone marrow contributed normally to all hematopoietic lineages, but not to T cells. These observations indicate the redundancy of Notch signaling, probably caused by the expression of multiple members of Notch and its ligand in the hematopoietic system.

We have taken a protein biochemical approach to address hDll1 function by creating a soluble form of the newly identified Notch ligand. We made a soluble hDll1 as a GST fusion protein, containing the DSL domain and the adjacent N-terminal region, in a bacterial expression system. Compared to the soluble form of Jagged1 produced in COS cells,23,30 our production strategy achieved higher yield and purity. Our method of ligand production may prove to be a useful general strategy for the production of soluble receptor agonists using other members of the Notch ligand family. In the culture of murine Lin⁻ cells with GST-hDll1^NDSL and cytokines, we found decreased numbers of mature monocytes/macrophages and granulocytes and increased numbers of progenitor cells. We also observed decreased numbers of apoptotic cells and increased numbers of cells in the S-phase of the cell cycle. The function of hDll1 is thus consistent with reports on the action of Jagged1 and Jagged2 in hematopoiesis. In culture of primary human and mouse hematopoietic progenitor cells, both Jagged1 and Jagged2 were shown to increase the number of progenitors.30,31 Jagged2 also decreased the number of mature myeloid cells as measured by the surface marker expression of CD11b and CD14.31 

We believe that the hDll1 effects, including increased numbers of progenitor cells and cells in the S-phase of the cell cycle and decreased numbers of apoptotic cells in cultures with hDll1^NDSL, result from the suppression of hematopoietic progenitor differentiation. It is known that mature granulocytes generated in culture readily undergo apoptosis within a day and that progenitor cells exit the cell cycle as they mature. However, we cannot rule out a mechanism by which hDll1 directly modulates the expansion, apoptosis, and cell cycle of hematopoietic cells. Data supporting the latter model have emerged recently. Jagged1 was shown to affect the proliferation of hematopoietic cells as measured by thymidine incorporation.32 The activated form of Notch1 expressed in HL-60 cells has been shown to enhance cell progression through the G1 phase of the cell cycle, which is associated with Notch1-induced delay of differentiation.31 It is also reported that an activated form of Notch1 provides significant protection of T-cell lines from TCR-mediated apoptosis.42 

Most cytokines and their receptors have been identified at the molecular level.43 Binding of cytokines to their receptors elicits both proliferation and differentiation signals. The receptor functional domains have been mapped.44 The membrane-proximal domains of the Epo,45GM-CSF,46 G-CSF,47 and thrombopoietin48 receptors are sufficient for mediating mitogenic signals, whereas the membrane-distal domains are required only for generating differentiation signals. Cytokine-mediated receptor oligomerization juxtaposes and activates JAK kinases associated with the membrane-proximal box 1 and box 2 motifs of the receptor. Activated JAKs then phosphorylate and induce nuclear translocation of the STAT DNA-binding proteins.49 STATs appear to be involved in functional differentiation (as opposed to mitogenic) responses to cytokines.50 We speculate that the activated Notch receptor could inhibit hematopoietic differentiation in response to cytokine stimulation by interacting with molecules involved in the cytokine differentiation pathways. Recently, Bigas et al34 reported that mammalian Notch intracellular domain contains a Notch cytokine response region (NCR). Notch1 and Notch2 contain different NCR and, therefore, respond to different cytokine signals.

The functional domain of the fusion protein is the region of hDll1 sequences. This is demonstrated by GST control protein having no hematopoietic effect, antibodies against mDll1 blocking the GST-hDll1^NDSL function, and hDll1^NDSL released from the fusion proteins showing hematopoietic effects similar to those of the fusion protein. The potential physiological role of the secreted DSL proteins in vivo has recently been suggested because soluble forms of Delta do exist in Drosophila embryos.27 Given the general role of the Notch pathway in the inhibition of hematopoietic progenitor cell differentiation, the availability of soluble hDll1 ligand should facilitate the development of a new strategy for the ex vivo expansion of hematopoietic stem/progenitor cells.

We thank Dr Achim Gossler for providing the mDll1 cDNA probe and Lisa Wang for outstanding technical assistance.