Identification and Mutation of Primary and Secondary Proteolytic Cleavage Sites in Murine Stem Cell Factor cDNAYields Biologically Active, Cell-associated Protein*

Phenotypic abnormalities of melanocytes, germ cells, and hematopoietic cells of Steel mice demonstrate the critical role of stem cell factor (SCF) in development. Production of SCF in the hematopoietic microenvironment as either a membrane-associated or soluble factor leads to pleiotropic effects on hematopoietic stem and progenitor cells and significant effects on the produc- tion of erythroid cells. Although the production of these two forms of SCF is highly regulated, the physiologic role(s) of membrane-associated and soluble SCF remain unclear. We have demonstrated that the generation of soluble murine SCF by murine stromal cells derived from the fetal hematopoietic microenvironment is de-pendent on two distinct proteolytic cleavage sites. The primary site in exon 6 is preferentially utilized in these cells. The secondary site located in exon 7 is utilized only in the absence of the primary site. Proteolytic processing at this secondary site appears to be species-spe- cific, since the human protein sequence diverges at this site, and protein expressed from the human cDNA en- coding this site in murine stromal cells remains largely membrane-associated. Site-directed mutagenesis of the murine SCF cDNA encoding both proteolytic cleavage sites leads to the generation of membrane-associated and biologically active SCF on murine stromal cells. These results suggest that the regulation of processing of the secondary proteolytic cleavage site could play a critical role in the function of membrane-associated SCF protein. expression of murine and human protein in differences in the processing of the human and murine cDNA-encoded protein and the preferential use of the exon 6 cleavage site of SCF24s cDNA-encoded protein in murine

Growth factors and their receptors play important roles in normal development as mediators of intercellular communication by diffusible molecules and direct cell-cell interactions. Many growth factors and some receptors occur in both membrane-bound and secreted forms (1). Several growth factors, such as colony-stimulating factor-1 (2), transforming growth factor-a (TGF-dl (3), and tumor necrosis factor (4) are pro-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
MGF, mast growth cell factor; KL, kit ligand; Sld, Steel-Dickie; hSCF, SCF, stem cell factor; HM, hematopoietic microenvironment; S1, Steel; human stem cell factor; mSCF, murine stem cell factor; rrSCF, recombinant rat stem cell factor; DMEM, Dulbecco's modified Eagle's me-duced as transmembrane proteins that can be released by specific proteolytic cleavage to generate soluble factors. Such growth factors can influence events at a localized site by adhesive interactions or direct contact with receptors, termed juxtacrine stimulation (3). Following proteolytic cleavage, the secreted products of membrane-bound growth factors can also effect other cells in a regional or systemic manner. Alteration in the balance between the diffusible and membrane-bound forms of growth factors may lead to phenotypic abnormalities as seen in Alzheimer's disease (5) and cutaneous mastocytosis (6).
The hematopoietic microenvironment (HM) represents a heterogenous group of cells and matrix molecules located in the bone marrow medullary cavity which play a major role in the proliferation and differentiation of hematopoietic cells by a variety of cell-cell interactions (7). The HM has been proposed to regulate hematopoiesis in several ways, including direct cellcell contact (8), stabilization of growth factors via binding to extracellular matrix proteins (9,101, production of both positive and negative growth regulatory proteins (11)(12)(13)(14), and co-localization of hematopoietic stem and progenitor cells with locally high concentrations of multiple growth factors presented on the surface of cells in a local area network (15).
Defects in the HM associated with the Steel (S1) mutation in mice have recently been demonstrated to be due to abnormalities in the production or presentation of the protein product of the Steel gene, termed stem cell factor or steel factor (also called kit ligand ( K L ) or mast cell growth factor (MGF)), which exists as a locally secreted or membrane-bound protein (16)(17)(18)(19). Stem cell factor (SCF) is presented in the HM as a result of both pre-and post-translational processing events of primary SCF mRNA transcripts. The nucleotide sequence of the cloned SCF cDNA predicts a transmembrane protein with a leader sequence, extracellular domain, membrane spanning region, and short cytoplasmic tail (18,20,21). A major secreted protein, a 164-amino-acid biologically active growth factor, results from proteolytic cleavage at a site encoded by the primary mRNA (SCFZ4') transcript within exon 6. A membrane-bound 220amino-acid protein is produced from an alternatively spliced second mRNA transcript (SCFZz0), which lacks exon 6 sequences (22,23). In some studies, the murine SCFZz0 cDNA also encodes a protein released from the cell (24, 25).
The physiologic roles of the various SCF proteins remain unclear. Animals deficient in SCF as a result of deletion of Steel coding sequences (SLISL homozygotes) die in utero or perinatally with severe deficiencies of germ cells, hematopoietic cells, and melanocytes (26). The presence of pleiotropic abnormalities dium; CS, calf serum; SCFdep, stem cell factor-dependent cells; PBS, phosphate-buffered saline; aa, amino acid.

1237
in viable mice with the Steel-Dickie (SZd) mutation, which encompasses a genomic deletion of the membrane-spanning and cytoplasmic tail of the Steel sequence, has been interpreted as evidence of the importance of the membrane-bound protein in adhesivetmigratory or cell proliferative responses in these systems, since presumably only secreted protein is produced in these mice (22,24,27). Differential effects of several forms of SCF have been demonstrated by our laboratory and other investigators using in vitro culture systems (23, 28-31).
We have previously reported that stromal cells derived from the fetal HM of SZISl mice produce biologically active human SCF when transfected with the human (h)SCF248 or hSCFZz0 cDNAs (23). Expression of the cDNA lacking exon 6 (hSCFZz0) was associated with SCF which remained almost entirely membrane bound, while expression of the hSCF24s cDNAled to high levels of secreted protein. In contrast, Huang et al. (24) have reported that expression of both murine (m)SCF24s and mSCFZ2O cDNAs in COS-7 cells leads to secreted protein, implying a second proteolytic cleavage site exists outside exon 6 which is utilized in these non-HM-derived cells.
In an effort to better characterize the processing of SCF in a HM-derived cell line and to further study the role of various presentations of SCF protein, we have transfected mSCF24s, mSCFZ2O cDNAs and plasmids containing mutations at various sites in exon 6 and exon 7 of mSCF24s cDNA into SCF-deficient SVS14 (16) stromal cells. We have compared expression of murine and human protein in these cells. Our results demonstrate differences in the processing of the human and murine SCFZz0 cDNA-encoded protein and the preferential use of the exon 6 cleavage site of SCF24s cDNA-encoded protein in murine HMderived cells, implying cell or species differences in the protease responsible for the production of secreted protein. Using in vitro mutagenesis, we have mapped the location of the primary proteolytic cleavage site in exon 6 and of a secondary cleavage site in exon 7 close to the membrane-spanning region and have generated cell lines containing multiple mutations of mSCF248 in which SCF remains membrane bound and biologically active.
EXPERIMENTAL PROCEDURES Cell Lines: SCF Factor-dependent Cells-WBB6F1 SVSld (Jackson Laboratory, Bar Harbor) mouse bone marrow was prepared as a cell suspension and placed in liquid culture in RPMI 1640 (Life Technologies, Inc./BRL) supplemented with 10% fetal bovine serum (Life Technologies, Inc./BRL) and 750 ng/ml of recombinant rat stem cell factor (ITSCF'~~) (Amgen, Thousand Oaks, CA). Primary cell cultures were established at varying cell concentrations ranging from lo4 to 5 x lo5 celldml. Control cultures were established in which no ITSCF'~~ was added to the medium. Two weeks after initiation, control cultures had generated sparse cell populations consisting of predominantly macrophages and fibroblastic stromal cells. In contrast, the cultures supplemented with consistently generated populations of cells resembling mast lineage cells similar to those generated in cultures of rat bone marrow (32). Cells generated in cultures initiated at 5 x lo5 celldml were serially passaged in rrSCF" containing medium every 7 days. By this method a factor-dependent cell population, SCFdeP, was established which had a restricted cytokine response profile. SCFdep cells responded only to rrSCF and not to other growth factors including those growth factors known to be produced by stromal cells.
Mutagenesis-Murine SCFZz0 or SCF248 (21) c D N h were subcloned into pGEM-7 (Promega) and transformed in Escherichia coli DHllS (Life Technologies, Inc./BRL). Single-stranded template DNA was punfied (33) and oligonucleotide-directed mutagenesis was camed out using the in vitro mutagenesis system (Amersham Corp.). Mutant clones were identified by sequencing (34) and were ligated into V19.8 expression vector (21) and the final plasmid construct was confirmed again by sequencing.
Generation of Stable Dansfectants of S1tSl4 Stromal Cells with mSCF cDNAs-Transfection of plasmid DNA into SVS14 cells was performed using DOTAP (Boehringer Mannheim) according to the manufacturer's recommendations. WS14 cells were grown to subconfluence in DMEM with 10% CS. Plasmid mixtures containing c D N h of mSCF and hygromycin-resistance gene in p48 (35) (1O:l ratio) were incubated with DOTAF' at room temperature for 10 min, then diluted with fresh medium. Medium was removed from SVS14 cells and replaced with the transfection mixture, and the cells were allowed to grow overnight. The following day, fresh medium with hygromycin (Calbiochem) (300 unitd ml) was added to the cells. Well isolated hygromycin-resistant colonies arising in 10-14 days were transferred to 24-well plates and expanded.
'Ibtal cellular RNA was prepared from each cloned cell line using Trireagent (Molecular Research Center, Cincinnati, OH) and clones were screened for mSCF expression by reverse transcriptase-polymerase chain reaction (Perkin-Elmer).
Metabolic Labeling and Immunoprecipitation-For protein analysis, subconfluent cells were labeled overnight in methionine-free DMEM, 10% dialyzed CS with 0.1 mCi/ml of [35S1Met-Cys (ICN). Conditioned medium was collected from each cell line 12 h later, filtered through a 0.45-pm filter, and stored at -80°C until used. Labeled cells were washed with phosphate-buffered saline (PBS) and lysed in lysis buffer (0.5% sodium-deoxycholate, 0.5% Nonidet P-40, 50 m NaC1, 25 m Tris-C1 (pH KO), and 1 nm phenylmethylsulfonyl fluoride). Immunoprecipitation was performed using a rabbit polyclonal antibody raised against rat-SCF (32) (kindly supplied by Dr. Larry Bennett, Amgen). 2 ml of polyclonal sera was conjugated to 1 ml of protein A-agarose using the AfKnica purification kit (Schleicher and Schuell), according to the manufacturer's recommendations. The antibody-conjugated protein Aagarose was resuspended in 3 ml of PBS with 0.05% sodium azide and stored at 4 "C. For immunoprecipitation labeled samples were thawed overnight at 4 "C and passed over DEAE-Sepharose fast flow (Pharmacia LKB Biotechnology Inc.) equilibrated with PBS at 4 "C. The flowthrough was collected (except for cell lysates of mSCFZz0, see below), concentrated by a Centriprep-10 (Amicon) device according to the manufacturer's instructions, and transferred to 1.5-ml microfuge tubes.
For mSCF220 cell lysate, bound protein was eluted with PBS with 0.1 M NaCl. 20 pl of the conjugated antibody was added to each sample and allowed to incubate overnight at 4 "C. The samples were centrifuged for 3 min at 4 "C, supernatant discarded, and the immunoprecipitates were washed three times with wash buffer ((0.5 M NaCI, 20 nm Tris-C1 (pH 7.5), and 1% Triton)) and once with 20 nm Tris-C1 (pH 7.5). The protein was released from the beads by boiling for 5 min. For glycosidase treatment, half of the protein samples were digested with neuraminidase (0.8 unitdml), 0-glycanase (0.12 unitdml), and N-glycanase (20 unitd ml) (all Genzyme) in 10 nm Tris-C1 (pH 7.5), 0.1% SDS and 0.1 M 2-mercaptoethanol, overnight at 37 "C. An equal volume of 2 x sample loading buffer ((100 nm Tris-C1 (pH 6.8),200 nm dithiothreitol, 4% SDS, 0.02% bromphenol blue, and 20% glycerol)) was added to each sample, boiled for 5 min, and the proteins analyzed on 12 or 15% SDS-PAGE (36). Biological Activity of Dansfected Cell Lines-SCF biological activity produced by each stable transfectant of the murine SCF cDNh were assayed using the SCF-dependent murine bone marrow-derived SCFdep cell line and a thymidine incorporation procedure. To analyze biological activity present as soluble protein, conditioned media was collected from subconfluent stromal cell transfectants after incubation overnight in RPMI with 10% fetal bovine serum. Each conditioned medium was passed through a 0.45-pm filter and concentrated using a Centriprep-10 device. The concentrated conditioned medium was filtered again, aliquoted, and frozen at -80 "C until further use. For analysis of SCF activity, 5 x lo4 SCFdep cells were cultured overnight in 100 pl of concentrated conditioned medium in 96-well plates. 1.0 pCi of r3H1thymidine was added to each well and incubated at 37 "C for an additional 4 h. The cells were then harvested using an automated cell harvester (96-well Harvester, Brandel, Gaithersburg, MD) and thymidine incorporation was determined in a scintillation counter.
The presence of SCF-like activity on the surface of stable transfectants was analyzed using a co-culture assay. On the day prior to assay, stromal cells were treated with 5 pg/ml mitomycin C (to inhibit further cell proliferation of the stromal cells), then washed three times in PBS, trypsinized, counted, and seeded at 3 x lo4 celldwell in 0.1% gelatincoated 96-well plates. These cultures were incubated in DMEM, 10% CS at 37 "C. After 24 h, 5 x lo4 SCFdep cells were added to the stromal cells and cultured for an additional 24 h. Subsequently, 1.0 pCi of PHIthymidine was added to each well for 4 h at 37 "C. Cells were then harvested and thymidine incorporation was determined as above.  Murine SCFzo and SCF4' cDNAs in S1 lS14 Cells: Comparison with Expression of Human SCFzo cDNA-Murine SCF248 and SCF220 cDNAs encoding the 248 aa and 220 aa proteins (Fig. 1) were cotransfected into the SVS14 cell line with the hygromycin expression plasmid, p48. Hygromycinresistant colonies were selected and analyzed for expression of the introduced cDNAs using reverse transcriptase-polymerase chain reaction. Multiple clones expressing mSCF248 and mSCF220 were generated. To determine the biosynthesis of mSCF24s and mSCF220 proteins, SVS14 stable transfectants expressing each cDNA at similar levels by Northern blot analysis were labeled with [35S]Met-Cys overnight, and conditioned medium and cells were harvested for immunoprecipitation using rabbit polyclonal antibody raised against rat recombinant SCF. The immunoprecipitated proteins were analyzed on SDS-PAGE gels. As seen in Fig. 2A, expression and processing of both murine SCF248 and SCF220 cDNAs leads to secretion of significant amount of SCF protein into the conditioned media of stable transfectants. After glycosidase digestion (lanes marked +), major protein species of apparent molecular masses of 21 kDa and 18-20 kDa are secreted from the mSCF248 and mSCF220 cDNA transfectants, respectively (arrowheads). This difference in apparent molecular mass is consistent with the lack of exon 6 sequences in the secreted protein generated from the mSCF220 cDNA. In comparison to the amount of protein immunoprecipitated from the conditioned media (*), 3-5-fold more cell-associated SCF protein (**) at -40 and 33 kDa (mSCF248) and 27-33 kDa (mSCF220) is immunoprecipitated from these cell lines (after glycosidase treatment, arrowheads, Fig. 2 B ) . In sharp contrast, the amount of protein secreted by SVS14 stromal cells transfected with the hSCF220 cDNA (*) (Fig.  2C) is minimal with an apparent molecular mass of 23 kDa after glycosidase treatment. The difference in the processing of murine and human SCFZ2O is even more evident when the relative amount of cell-associated (**) and secreted (*) protein is compared for each cDNA (5:l versus -20:1, murine versus human, Fig. 2). Pulse-chase experiments demonstrate the accumulation of soluble protein expressed from the mSCFZ2O cDNA in stromal cells is delayed with respect to the primary secreted product derived from the mSCF248 cDNA (data not shown). These data are consistent with reports by other investigators of a secreted SCF protein derived from the murine cDNA lacking exon 6 (24) and reports from our laboratory demonstrating largely membrane-bound protein derived from the human SCFZ2' cDNA in these stromal cells (23).

Expression of
Purification of SCF from conditioned media of buffalo ratliver cells and amino acid sequence determination has previously demonstrated the carboxyl-terminal amino acid of the secreted rat SCF protein to be either Ala164 or Ala165 (37). The region of exon 6 surrounding positions 164-165 contains small nonpolar amino acids which are similar to protein cleavage sites located in the colony-stimulating factor-1 (38) and TGF-a proteins (25). In an effort to better analyze this putative cleav-age site, site-directed mutagenesis was utilized to substitute or delete nucleotide sequences corresponding to amino acids surrounding positions 164-165 (Fig. 3). The changes introduced were chosen to facilitate screening and to prevent creation of new protein cleavage sites. The mSCFXg cDNA represents Ala164 + Leu and Ala165 + Glu substitutions with deletion of amino acids at position 163 and 166 of mSCF248 cDNA and introduction of the XhoI-recognition sequence (CTCGAG) in the cDNA. The mSCFD1 cDNA represents deletion of nucleotide sequences corresponding to amino acids 163-166 in mSCF248 cDNA.
These mutated cDNAs were co-transfected into SVS14. [35SlMet-Cys-labeled proteins from clones expressing the mutated cDNAs were immunoprecipitated and analyzed on SDS-PAGE (Fig. 4). Surprisingly, SVS14 cells transfected with mSCFXS and mSCFD1 secrete large amounts of proteins of apparent molecular mass of -30 kDa after glycosidase treatment. Results from pulse-chase experiments show that accumulation of secreted protein derived from the mSCFXS and mSCFD1 cD-NAs are delayed in comparison to both mSCF248 and mSCF220 cDNAs (data not shown). The data are consistent with the utilization of an alternative proteolytic cleavage site in the absence of the exon 6 primary cleavage site. The size of this mutated and secreted protein suggests that the alternative cleavage site must reside in amino acids carboxyl-terminal to the primary site in exon 6 (i.e. closer to the membrane-spanning region).
Examination of the amino acid sequence in this region of the SCF protein revealed a tetrapeptide of Lys-Ala-Ala-Lys at amino acid positions 178-181 in the murine sequence which is similar to the proteolytic cleavage site in exon 6 (Fig. 3). In an effort to further map the alternative proteolytic cleavage site utilized after mutation of the primary site in exon 6, additional mutations were performed. The mSCFXSm3 and mSCFDlD3 cD-NAs represent deletions of 12 nucleotides around amino acid 180 in mSCFXg and mSCFD1, respectively. These mutated cD-NAs were co-transfected with p48 into SVS14. [35SlMet-Cyslabeled proteins from hygromycin-resistant clones expressing the mutated cDNAs were immunoprecipitated and analyzed on SDS-PAGE (Fig. 5). In contrast to mSCFXg, mSCFXgm3 encodes a protein of apparent molecular mass of 36-43 kDa after glycosidase treatment which remains almost entirely cell-associated. Similar results were obtained with mSCFDlm3 (data not shown). These data are consistent with the utilization of a secondary proteolytic cleavage site at or near the tetrapeptide Lys-Ala-Ala-Lys located in exon 7 of mSCF when the primary proteolytic cleavage site in exon 6 is not available. Since the human nucleotide sequence in this area predicts the tetrapeptide Lys-Ala-Lys-Asn, the data suggest that the difference in processing of the human and murine SCFZ2O proteins could be related to sequence differences in this area.
Biological Activity of Murine SCF Proteins Expressed in HMderived Stromal Cells-To assess the biological activity of proteins expressed from wild-type and mutated mSCF cDNAs, we examined both conditioned medium and cell-associated proteins for the ability to stimulate proliferation of the SCFdep cell line. Medium conditioned overnight from representative stable transfectants of mSCF cDNAs and SVS14 were concentrated and used in L3H1thymidine incorporation studies with the SCFdep cell line. As seen in Fig. 6 A , conditioned medium from both mSCF248 and mSCFZ2O stimulated thymidine incorporation in the SCFdep cell line at levels significantly above SVS14. No stimulation of thymidine incorporation could be detected with mSCFXg, in spite of the presence of immunologically reactive protein in the conditioned medium. Similar results were obtained with mSCFD1 (data not shown). The apparent antagonistic effect of mSCFXs-conditioned medium was not seen con-

human SCF cDNAs in murine SVS14
FIG . 2. Expression of murine and stromal cells. Parental Sl/S14 and stable transfectants were labeled overnight with ["5Slmethionine, conditioned media coltated as described under "Experimental lected, cells lysed, and immunoprecipi-Procedures." A, supernatant of cultured cell lines S1/S14, SVS14-mSCF248, and SI/ SI4-mSCFZz0 treated without (-1 and with (+) glycosidases. B , cell lysates from the cell lines described in A. C, conditioned medium and cell lysate from S1/S14-  Fig. 3. Stable transfectants expressing mutated cDNAs were labeled with [36Slmethionine overnight, conditioned media collected, and immunoprecipitated as described under "Experimental Procedures." The results from two mutations are shown without (-) and with (+) glycosidase treatment. 5. Expression of murine SCF cDNAs mutated both in exon 6 and exon 7 in SVSl' stromal cells. Abbreviations used for the mutated cDNAs are as in Fig. 3. Stable transfectants expressing mutated cDNAs were labeled with [35Slmethionine overnight, conditioned media collected, cells lysed, and immunoprecipitated as in Experimental Procedures." The results of one mutation is shown without (-) or with (+) glycosidase treatment.

FIG.
inhibited by neutralizing antibody to murine SCF protein (data not shown).
Analysis of cell-associated SCF-like activity was accomplished using thymidine incorporation assays carried out while SCFdep cells were co-cultivated with each stromal transfectant cell line. Although the stromal cell lines were treated with sistently. As expected, no soluble SCF growth stimulatory activity was present in medium conditioned with stable transfectants of the mSCFXSm3 cDNA. These data are in agreement with the observed lack of secreted protein in the SVS14-mSCFXgm3 cell line. The bioactivity demonstrated in the conditioned medium of mSCF248 and mSCF220 was specifically mitomycin C prior to cultivation to inhibit thymidine incorporation by these cells, some increase in background incorporation was noted on SVS14 cells. However, Fig. 6B demonstrates that mSCF248 and mSCFZZ0 proteins stimulates significantly more thymidine incorporation by SCFdep cells than SVS14 in cocultivation. In spite of the lack of active protein in conditioned medium, the expression of mSCFXgm' cDNA in SVS14 cells is associated with the presence of biologically active protein on the cell surface. No biologically active protein was detected by co-culture using the mSCFXg transfectant cell line. These results demonstrate that the mSCFXgml protein is present on the cell surface of stromal cells and is biologically active. In experiments utilizing primary, adherent cell-depleted murine bone marrow cells, increased numbers of progenitorderived colonies are seen in cultures utilizing the concentrated conditioned medium from mSCF248 and mSCFZz0 cell lines. In addition, short co-culture (48 h) experiments demonstrate stimulation of myeloid and mixed primary progenitor colony formation on mSCF248, mSCFZz0, mSCFXgm3, and mSCFXg cells (data not shown). These data demonstrate that the biological activity assayed in the SCFdep cell line assays is also active on primary bone marrow cultures. consists of a heterogeneous group of cells, including stromal cells, which have been shown to produce SCF in a regulated fashion (24). The production of SCF in the HM as either a soluble or membrane-associated protein leads to pleiotropic effects on a variety of hematopoietic stem and progenitor cell types as well as significant effects on the production of erythroid-lineage cells both in vitro and in vivo (39). The physi-ologic roles of membrane-associated and soluble SCF remain unclear. The presence of major phenotypic abnormalities in S1 lSld mice, which express a mutated form of SCF lacking the membrane-spanning domain and cytoplasmic domains (22,24, 27), suggests that the membrane-bound protein is important for many aspects of hematopoietic stem cell biology. Several investigators have demonstrated differential effects of membrane and secreted proteins on germ cell and hematopoietic cell adhesion, survival, and proliferation using in vitro culture systems (23, 28, 29, 40). A major goal of current research is to examine this issue in vivo.
It has become increasingly clear that membrane-associated proteins may have significant effects modulated via soluble forms (2, 3, 23). Stem cell factor is somewhat unique among these proteins in that the secreted form of this growth factor involves both a separate biosynthetic pathway (differential splicing of pre-mRNA to include or exclude exon 6) and posttranslational release of the extracellular domain via specific hydrolytic cleavage (19, 21, 24, 25). The possible role of membrane-associated SCF, like TGF-a and tumor necrosis factor, may relate to intercellular communication restricted to adjacent cells, termed jwtacrine stimulation (3).
Intercellular communication by direct cell-cell contact has been previously proposed to be a major mechanism of HM regulation of hematopoiesis (8). In addition to the effects on c-kitexpressing hematopoietic cells, it is possible that binding of SCF to its receptor also induces signaling in the stromal cells presenting SCF. Therefore, advantages of proteolytic release of membrane-associated growth factor could include the rapidity with which a local affect can be changed to a regional stimulation and removal of stimulatory affects of ligandreceptor interactions via the cytoplasmic tail in the presenting stromal cell. The function of the cytoplasmic tail of SCF is currently unknown, although some experimental data suggest a role in protein processing (41). Enzymatic cleavage of the proteolytic site in exon 6 of mSCF248 appears to be a highly regulated event with similarities to cleavage of TGF-a (251, including cleavage at a Ala-Val dipeptide and activation via protein kinase C pathways with calcium- dependent mechanisms (24, 42, 43). These characteristics suggest a n elastase-like protease (44) may be involved in cleavage of the primary proteolytic site in exon 6. Previous studies (24, 25) have suggested that additional proteolysis in COS-7 cells must take place outside exon 6 to generate a secreted protein from the alternatively spliced mSCFZz0 mRNA, which lacks exon 6. Protease inhibitor profiles and sequence analysis suggests a distinct protease may be involved in the generation of this secreted form of mSCFZz0 (25). Our results from stromal cell expression of mSCF220 also indicate a second proteolytic cleavage site outside exon 6. Utilization of this site in stromal cells results in delayed appearance of soluble SCF. Similar results have previously been observed in COS-7 cells (24).
Data presented here also map this putative secondary proteolytic cleavage site within exon 7 and demonstrate that in stromal cells derived from the HM this site is clearly used only when the primary site is not present. In addition, the differential processing of human versus murine SCFZz0 protein in murine stromal cells, and the sequence divergence in this region support the suggestion that a distinct protease is responsible for the proteolytic cleavage of the exon 7 site (25).
Secreted forms of mSCF248 and mSCFZz0 were demonstrated to be biologically active. In the studies presented here mSCF expressed from a cDNA containing mutations in both exon 6 and exon 7 (mSCFXgm3) proteolytic cleavage sites remains membrane-associated. No biologically active SCF-like protein activity was demonstrated in concentrated conditioned me-

Stem Cell Factor Proteolytic Cleavage Sites
dium from stromal cells expressing this cDNA, although cellassociated protein functioned when SCF-dependent cells were in direct cell-cell contact with producing stromal cells. These results are similar to data previously reported from our laboratory utilizing the hSCF220 protein, which in the context of murine stromal cells remains membrane-associated and is biologically active. No biological activity could be demonstrated with mSCF protein-containing mutations of the primary proteolytic cleavage site in exon 6. Analysis of SCF secondary structure (45) has shown that exon 6 encodes a variable spacer chain which may be important in the physiologic regulation of dimer formation. Mutation of exon 6 sequences could interfere with dimer formation of this protein resulting in the continued presence in conditioned medium of immunologically reactive SCF protein with diminished or absent biological activity. These data would suggest that inhibition of dimer formation plays a more critical role in the function of soluble versus membrane-associated SCF, since the same mutation when accompanied by additional mutations remains active when presented on the stromal cell surface. These results provide valuable information about the regulation of murine SCF, including confirmation of the primary cleavage site in exon 6 and the mapping of a secondary proteolytic cleavage site in the extracellular domain of the protein.
Mutation of both these sites generates a functional but entirely membrane-associated form of murine SCF. These studies are critical to the design of experiments aimed at analysis of the role of various SCF isoforms in uivo using gene targeting techniques. In addition, the availability of stromal cells derived from the HM which present SCF as secreted or membraneassociated proteins will allow more detailed analysis of the function of these proteins in hematopoiesis using in vitro cultures.