The majority of stem cell factor exists as monomer under physiological conditions. Implications for dimerization mediating biological activity.

Soluble Escherichia coli-derived recombinant human stem cell factor (rhSCF) forms a non-covalently associated dimer. We have determined a dimer association constant (Ka) of 2-4 × 108 M−1, using sedimentation equilibrium and size exclusion chromatography. SCF has been shown previously to be present at concentrations of approximately 3.3 ng/ml in human serum. Based on the dimerization Ka, greater than 90% of the circulating SCF would be in the monomeric form. When 125I-rhSCF was added to human serum and the serum analyzed by size exclusion chromatography, 72-49% of rhSCF was monomer when the total SCF concentration was in the range of 10-100 ng/ml, consistent with the Ka determination. Three SCF variants, SCF(F63C), SCF (V49L,F63L), and SCF(A165C), were recombinantly expressed in Escherichia coli, purified, and characterized. The dimer Ka values, biophysical properties, and biological activities of these variants were studied. Dimerization-defective variants SCF(F63C)S-CH2CONH2 and SCF(V49L,F63L) showed substantially reduced mitogenic activity, while the activity of the Cys165-Cys165 disulfide-linked SCF(A165C) dimer was 10-fold higher than that of wild type rhSCF. The results suggest a correlation between dimerization affinity and biological activity, consistent with a model in which SCF dimerization mediates dimerization of its receptor, Kit, and subsequent signal transduction.

Stem cell factor is a cytokine that is active toward early hematopoietic cells and also plays roles in gametogenesis, melanogenesis, and mast cell function. Its biological and other properties have been extensively reviewed (1,2). It is found in both membrane-bound and soluble forms, with the latter being derived from a membrane-bound form by proteolytic cleavage. The soluble SCF has 165 amino acids.
Both soluble Escherichia coli-derived and CHO 1 cell-derived recombinant human SCF have been reported to be non-co-valently associated dimers, as determined by sedimentation equilibrium and size exclusion chromatography at protein concentrations above 0.4 mg/ml (3). In a previous paper (4), we demonstrated that SCF dimer is dissociable under non-denaturing conditions and the dissociation rate constant (k d ) of E. coli-derived rhSCF dimer is approximately 1.35 ϫ 10 Ϫ4 s Ϫ1 at pH 4.8, 25°C. In the present work, we arrive at a value of 2-4 ϫ 10 8 M Ϫ1 for the dimer association constant (K a ) of E. coli-derived rhSCF, based on several approaches including ultracentrifugation and size exclusion chromatography at low SCF concentrations. Since the SCF concentration in human serum has been found previously to be a few nanograms/ml (5), the K a value suggests that the majority of SCF in serum may be monomeric. We use 125 I-SCF as a tracer added to serum to show that this does in fact appear to be the case.
The binding of ligands to cell receptors, followed by receptor dimerization, is essential for signal transduction by the family of transmembrane receptor tyrosine kinases (6 -8). The receptor for SCF on target cells is Kit (see Refs. 1 and 2 for reviews). Kit is a member of the type III receptor tyrosine kinase subfamily, which also includes the M-CSF receptor (Fms), the PDGF receptor, and Flk2 (6,7). The ligands for each of these receptors can exist as dimers, non-covalently associated in the cases of SCF (3,9) and Flk ligand (10), and disulfide-linked in the cases of M-CSF (11,12) and PDGF (13). It is reasonable to assume that ligand dimerization helps to mediate receptor dimerization for each of these ligand/receptor pairs, but a requirement for ligand dimerization has not been definitively shown in any of these cases. For Kit in particular, dimerization was suggested to be independent of the bivalency of the dimeric SCF ligand (14). However, sedimentation equilibrium and size exclusion chromatography indicated that SCF and Kit form 2:2 complexes in vitro (15). The SCF concentrations used in the latter study were higher than those found in vivo.
As an additional approach to testing whether dimerization of SCF helps to mediate dimerization of Kit, three E. coli-derived SCF variants were designed. Based on the genetic and structural homology between SCF and M-CSF (16), Phe 63 of SCF appeared comparable to the corresponding Phe 67 of M-CSF, which is located in a loop between the B-helix and C-helix. (M-CSF is a four-helix bundle cytokine (16), and SCF almost certainly is as well (17).) The x-ray crystallographic structure of the M-CSF disulfide-linked dimer (16) has shown that Phe 67 resides in the M-CSF dimer interface. Phe 63 involvement in the dimer interface of SCF was assumed, and two variants with Phe 63 substitutions were therefore prepared. The variant SCF(A165C) was also prepared. Characterization of dimer association constant K a , biophysical properties, and biological activities of these variants give results consistent with the hypothesis that SCF dimerization is important for Kit dimer-* 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.
ization. Models for SCF-induced Kit dimerization are discussed in light of the results.

EXPERIMENTAL PROCEDURES
Materials-Acetonitrile and HPLC-grade water were purchased from Burdick & Jackson. All protein sequencing reagents were obtained from Applied Biosystems Inc. and Hewlett Packard. Trifluoroacetic acid and other chemicals were purchased from J.T. Baker. Asp-N endoproteinase was purchased from Boehringer Mannheim. 4-Hydroxy-␣-cyanocinnamic acid was obtained from Aldrich. Molecular weight standards for size exclusion chromatography were purchased from Bio-Rad and those for SDS-PAGE from Novex. CHO-derived recombinant human soluble Kit was expressed and purified as described (18).
Expression of rhSCF and Its Variants in E. coli-The construction of the expression vector for human SCF, and the fermentation conditions for expression in E. coli were reported previously (19,20). The expressed gene contains an initiating methionine codon followed by the codons for SCF 1-165. The purified rhSCF retains Met at the N terminus (position Ϫ1). The SCF variants were prepared by procedures including site-directed mutagenesis of the parent gene with standard polymerase chain reaction techniques, essentially as described (21).
Purification of E. coli-derived rhSCF, Its Variants, and Derivatives-The purifications of rhSCF and its variants were performed essentially as described (20). Briefly, the inclusion bodies were harvested after E. coli cell breakage, solubilized in 8 M urea, and the SCF protein refolded and oxidized by diluting into Tris buffer containing 2.5 M urea and reduced glutathione. Subsequently, purification was conducted incorporating the following steps: S-Sepharose, reversed-phase C4, Q-Sepharose, and size exclusion chromatography. SCF(V49L,F63L) was purified by the same procedures as for the wild type rhSCF, except that the reversed-phase C4 step was eliminated. Following the same procedures, partially purified mis-disulfide-linked SCF(F63C) was obtained and subsequently reduced and denatured in 40 mM Tris, pH 8, 8 M urea, 0.5 M dithiothreitol at 37°C, for 30 min; the reduced SCF(F63C) was then purified by RP-HPLC. The preparation was dried in a vacuum centrifuge, redissolved in 8 M urea, and refolded by diluting 10-fold using 20 mM Tris, pH 8.0 with 5 mM reduced glutathione and 1 mM oxidized glutathione. Monomeric SCF(F63C) was isolated by a final RP-HPLC step, and was then dried down and reconstituted in PBS. The disulfidelinked SCF(F63C) dimer was formed by diluting the purified SCF(F63C) monomer into 20 mM Tris pH 8.0, 1 mM copper sulfate with stirring for 5 min at room temperature. The preparation was concentrated and loaded onto a RP-HPLC column. The disulfide-linked SCF(F63C) dimer was recovered, dried, and redissolved in PBS. SCF(F63C)S-CH 2 CONH 2 was made by adding iodoacetamide to SCF(F63C) monomer at a molar ratio of 5 to 1 and was purified by RP-HPLC. The purification of SCF(A165C) was similar to that of SCF(F63C) up to the RP-HPLC step after oxidation in the mixture containing reduced and oxidized glutathione. It was isolated as a disulfide-linked dimer (see "Results").
Size Exclusion Chromatography-Size exclusion chromatography of E. coli-derived wild type rhSCF, SCF variants, and the derivatives was performed using Superdex 75 or Superose 12 columns (Pharmacia Biotech Inc.; 10 ϫ 30 mm) equilibrated in 10 mM Tris-HCl, 0.1 M sodium chloride, pH 7.0, or PBS. The chromatography was carried out using a Hewlett Packard 1050 liquid chromatograph. The detection wavelength was set at 215 nm, and the elution flow rate was 0.7 ml/min.
Size Exclusion Chromatography of Human Serum Containing 125 I-rhSCF-125 I-rhSCF was supplied by Amersham International Inc. in sodium phosphate buffer containing 0.5 mg/ml bovine serum albumin. It was prepared by direct iodination with sodium [ 125 I]iodide using the chloramine-T method (22) and desalted/buffer-exchanged by size exclusion chromatography. Its specific activity was 56 mCi/mg and the concentration 250 Ci/ml. Fifty l of the preparation was further purified using a Superdex 75 column. The eluate (12.6 -14.2 min) corresponding to rhSCF dimer was collected. The concentration of the radiolabeled rhSCF (0.2 g/ml) was based upon an assumed recovery of 100%. Three l of the radiolabeled rhSCF was added to 50 l of human serum alone or along with 0.6, 1.8, or 5.4 ng of unlabeled rhSCF. These preparations were incubated overnight at 4°C and then loaded onto a Superdex 75 column equilibrated in PBS (flow rate 0.7 ml/min). Fractions were collected at 0.4-min intervals and measured for radioactivity using a ␥ scintillation counter.
RP-HPLC of rhSCF-Ten to 15 g of E. coli-derived wild type rhSCF, SCF variants, or the derivatives were loaded onto a reversed-phase column (Vydac C4, 4.6 ϫ 250 mm) on a Hewlett Packard 1090 HPLC system as described previously (23).

Asp-N Endoproteinase Peptide Mapping of rhSCF Preparations-
Fifty or 100 g of each SCF sample was digested overnight at 37°C with Asp-N endoproteinase in 20 mM Tris-HCl, pH 7.0, or in PBS, using an enzyme to substrate ratio of 1:75. The digests were then injected onto a reversed-phase column (Vydac C4, 4.6 ϫ 250 mm). The chromatography was carried out using a Hewlett Packard 1090 HPLC system equipped with a Chemstation and diode array detector as described previously (23).
Equilibrium Ultracentrifugation-Sedimentation equilibrium experiments were performed using a Beckman Optima XL-A analytical ultracentrifuge. The system measures and records the concentration of sample as a function of the radial position at various times during centrifugation. The samples were dialyzed against PBS at 4°C. Aliquots of the dialyzed solution were diluted to various final concentrations with the dialysate. For each protein solution, the absorbances at 280, 230, and 215 nm were determined using a Beckman DU 650 spectrophotometer. The solutions were placed in a 12-mm-thick, sixchannel cell assembly, using dialysate as reference, and centrifuged at 26,000, 20,000, and 16,000 rpm at 25°C. For E. coli-derived wild type rhSCF, several protein concentrations ranging from 0.02 to 0.5 mg/ml were studied. The radial absorption scans (280, 230, and 215 nm), which represent the protein concentration at each radial position of the solution, were taken after the equilibrium had been reached at each centrifugation speed. An absorption scan at 400 nm was used as a blank run for each speed. Data analyses were carried out using the KDAL-TON program (24), which fits the data to various association models by nonlinear regression. For data analysis, the following parameters were used. The partial specific volume, v, of rhSCF was calculated from its amino acid composition to be 0.7429 ml/g, the molecular weight of the SCF monomer is 18,657, and the solvent density is 1.004 g/ml. The extinction coefficient of SCF at 280 nm is 0.62 ml/mg cm (3). For the SCF variants, a single concentration of 0.4 mg/ml was used and the radial absorption scans at 280 nm were taken.
Sequence Analysis and Mass Spectrometry-Amino acid sequence analysis of peptides was performed on an automated Applied Biosystems Inc. (ABI) protein sequencer (models 477A or 470A) equipped with a model 120A on-line phenylthiohydantoin-derivative analyzer and a model 900A data collection system. The sizes of the purified peptides were determined by using matrix-assisted laser desorption mass spectrometry on a Kratos MALDI III instrument as described previously (23).
Circular Dichroism Spectroscopy-Circular dichroism (CD) was performed using a Jasco J-710 spectropolarimeter on samples of about 1 mg/ml in PBS as described previously (23). The spectra in the near UV region (240 -340 nm) and in the far UV region (185-250 nm) were recorded. The data are expressed as molar ellipticity (), assuming a mean residue weight of 112.
Fluorescence Spectroscopy-The fluorescence spectra were determined at ambient temperature on an Amico-SLM 500 spectrofluorimeter using a 0.5-cm pathlength as described previously (23).
Biological Assay-The biological activity of SCF was determined using an in vitro cell mitogenesis bioassay involving [ 3 H]thymidine incorporation by the human megakaryoblastic leukemia cell line UT-7 as described previously (25).
Receptor Binding Assay-The receptor binding assay utilized membranes prepared from the human erythroleukemic cell line OCIM-1, which expresses high levels of Kit, and was carried out as described (25).

Size Exclusion Chromatography of E. coli-derived Wild
Type rhSCF-When E. coli-derived rhSCF was analyzed by size exclusion chromatography (Superdex 75), a single symmetric peak was seen at higher loading concentrations. As the rhSCF loading concentration was lowered to 0.12 mg/ml, a second peak, eluting after the main peak, began to appear (Fig. 1, A  and B). The main peak eluted at 13.6 min, while the second peak eluted at 15.9 min. The molecular weight of the main peak was estimated by light scattering analysis to be 37,600, which corresponds to the molecular weight of the E. coli-derived rh-SCF dimer. The second peak, which eluted in the same position as SCF(F63C)S-CH 2 CONH 2 monomer (see below in Fig. 5), represented the rhSCF monomer. The rhSCF dimer and monomer both elute somewhat earlier than would be expected for globular proteins having these molecular weights (compare elution positions of molecular weight standards). As noted previously (3), this result suggests that SCF has an elongated shape.
Equilibrium Ultracentrifugation of E. coli-derived Wild Type rhSCF-The association behavior of rhSCF in solution was studied by sedimentation equilibrium. Because of the low protein concentration needed for detection of the rhSCF monomers, several protein concentrations ranging between 0.02 and 0.5 mg/ml were studied and the absorption scans were taken at 215, 230, and 280 nm. Data sets collected from various loading concentrations and various centrifugation speeds were normalized and fitted to different self-association models. For rhSCF solutions with concentration equal to or greater than 0.1 mg/ ml, the data fit well to an ideal dimer system. As the protein concentration was decreased, the data fit very well to an ideal (monomer-dimer) system with a dimer association constant (K a ) of 2.4 -4.0 ϫ 10 8 M Ϫ1 (Fig. 2).
Size Exclusion Analysis of Human Serum after Addition of 125 I-rhSCF-Since the concentration of SCF in human serum averages about 3.3 ng/ml (5), it can be calculated from the above estimates of K a for rhSCF dimerization that a substantial amount of the serum SCF could be monomeric. As another approach to determining whether some SCF at physiologic concentrations might exist as monomers, 125 I-rhSCF was added to human serum and then size exclusion chromatography was performed on a Superdex 75 column. The radiolabeled rhSCF was added to human serum alone, or along with increasing amounts of unlabeled rhSCF. The final rhSCF concentrations were approximately 12, 24, 48, or 120 ng/ml, respectively. A significant amount of SCF monomer was detected in the human serum preparations (Fig. 3). The ratio of the integrated area of monomer peak (peak b) to that of dimer peak (peak a) decreased as the total SCF concentration increased. More than 49% of the SCF was detected as monomer at all the concentrations (Table I). The same results were obtained when the experiment was done with human serum albumin, to which 125 I-rhSCF was added. An additional peak, called peak c, which eluted earlier than the dimer peak, also appeared in the experiment. Since there is a considerable amount of soluble Kit extracellular domain (sKit) in human serum (26), it seemed possible that peak c could represent the SCF⅐sKit complex. However, complexes of CHO cell-derived human sKit with rh-SCF (15) elute earlier than peak c (data not shown). Peak c may result from nonspecific binding of rhSCF to human serum components or human serum albumin. These results indicate that the majority of SCF may exist as monomer at the physiological concentration of a few nanograms/ml. The K a derived from the percentage of monomer obtained at various total rh-SCF concentrations is shown in Table I. This K a agrees with the K a obtained by sedimentation equilibrium studies, and by the size exclusion chromatographic studies with purified, unlabeled rhSCF described above.
Characterization of SCF(F63C) and Its Derivatives-SCF(F63C) and its derivatives (Cys 63 -Cys 63 disulfide-linked SCF(F63C) dimer and SCF(F63C)S-CH 2 CONH 2 (carboxyamidomethylation at Cys 63 position)) were obtained as described under "Experimental Procedures." The preparations were essentially pure and homogeneous, as shown by RP-HPLC and SDS-PAGE (Fig. 4A). Comparing the Asp-N endoproteinase peptide maps of the variant and its derivatives to that of wild type rhSCF (Fig. 4B), the peptides d, e, and f each had the same N-terminal sequence except for position 63. Peptide f had twice the theoretical average mass (Table II) Upon analysis by size exclusion chromatography (Superose 12), SCF(F63C) and SCF(F63C)S-CH 2 CONH 2 eluted at 19.4 and 19.5 min (Fig. 5), which corresponds to the position of the wild type rhSCF monomer. The molecular weight of SCF(F63C)S-CH 2 CONH 2 was determined as 20,900 by light scattering analysis. Sedimentation equilibrium analysis demonstrated that SCF(F63C)S-CH 2 CONH 2 has a much lower association constant (K a 1.8 ϫ 10 3 M Ϫ1 ) than wild type rhSCF (data not shown). These results indicate that SCF(F63C) and SCF(F63C)S-CH 2 CONH 2 are monomers. SCF(F63C) easily converted to covalent dimer (see below), such that after 24 h in the ultracentrifuge, half of the SCF(F63C) became covalent dimer; thus, the K a of SCF(F63C) was not determined.
In order to determine if the position 63 variant and its derivatives had substantially altered structure, the solution structures of rhSCF and its variants were compared by CD and fluorescence spectroscopy. The far UV CD signals arise from peptide bonds and reflect secondary structure. The two minima at 208 and 222 nm in the far UV CD spectra (Fig. 6B) indicate that wild type rhSCF, SCF(F63C) monomer, and its derivatives contain substantial ␣-helix (27). Using the Greenfield-Fasman equation (27), these spectra suggested more than 40% ␣-helix, as reported previously for wild type rhSCF (3). The near UV CD signals arise from the asymmetric environments of the aromatic residues and thus reflect tertiary structure (28). SCF(F63C) monomer and SCF(F63C)S-CH 2 CONH 2 had CD spectra identical to that of wild type rhSCF in both the near and far UV regions (Fig. 6, A and B). SCF contains only one Trp, located in a fairly hydrophobic environment, with a fluorescence maximum of 320 nm (3,23). The fluorescence maxima of the variants were also identical to that of wild type rhSCF (Fig. 6C), indicating that the hydrophobic environment of Trp 44 remained unchanged by the mutation. CD and fluorescence results suggest that replacing Phe at position 63 with Cys does not grossly alter the protein fold. The change may weaken hydrophobic interaction in the interface of the SCF dimer, resulting in monomer formation. SCF(F63C)S-CH 2 CONH 2 had 1,300-fold lower activity than wild type rhSCF in the UT-7 mitogenic assay. SCF(F63C) monomer had 10,000-fold lower mitogenic activity (Fig. 7). 50% inhibition of 125 I-SCF binding to Kit in the competitive receptor binding assay required 3-7 ng/ml wild type rhSCF, 0.4 g/ml SCF(F63C) monomer, and 0.9 g/ml SCF(F63C)S-CH 2 CONH 2 , respectively (Table III). Thus these two dimerization-defective species had 130 -300-fold lower receptor binding affinity than wild type rhSCF. SCF(F63C) monomer rapidly formed covalent dimer via disulfide linkage at Cys 63 in the presence of copper ion in Tris-HCl, pH 8.0. The formation of SCF(F63C) disulfide-linked dimer also occurred during storage or upon agitation. The dimeric SCF(F63C) species eluted 0.4 min later than the noncovalently associated wild type rhSCF dimer upon size exclusion chromatographic analysis (Fig. 5). The two far UV CD peaks at 208 nm and 222 nm for SCF(F63C) disulfide-linked dimer were significantly decreased, compared to wild type rh-SCF (Fig. 6B), indicating 10% less ␣-helix based on calculation by the Greenfield-Fasman equation (28). These results suggest that Phe 63 resides at the dimer interface, but that the location of the Phe 63 in the two SCF subunits may not be ideally situated for covalent linkage, such that disulfide bond formation causes structural distortion and the loss of biological activity.
Characterization of SCF(V49L,F63L) and SCF(A165C)-SCF(V49L,F63L) was purified essentially to homogeneity, as determined by SDS-PAGE. Upon size exclusion chromatographic analysis, the variant eluted in the monomer position at a loading concentration of 0.01 mg/ml (load volume 100 l), but eluted near dimer position at a loading concentration of 0.3 mg/ml. Equilibrium ultracentrifugation analysis revealed that the association constant K a of SCF(V49L,F63L) was 1.47 ϫ 10 5 M Ϫ1 . The variant had 30-fold lower mitogenic activity than wild type rhSCF (Fig. 7).
After oxidation and refolding of SCF(A165C), the preparation was further purified by RP-HPLC (Fig. 8A). Peak T42 was monomer and the majority of peak T48.7 was dimer, as determined by non-reducing SDS-PAGE. Comparing the Asp-N endoproteinase peptide maps of these two samples to that of wild type rhSCF (Fig. 8B), peptides g, h, and i each had the same N-terminal sequence except for position 165. Peptide i had twice the theoretical average mass (Table II), indicating that peak T48.7 dimer contained an interchain disulfide at position Cys 165 . The covalent SCF(A165C) dimer had a 10-fold higher biological activity than wild type rhSCF (Fig. 7). The mass of peptide h was 303 units higher than the theoretical average mass, indicating that the T42 species was a glutathione derivative. SCF(A165C)S-glutathione had biological activity identical to that of E. coli-derived wild type rhSCF (data not shown).

DISCUSSION
In this paper, we have demonstrated that the dimer association constant (K a ) of E. coli-derived wild type rhSCF is 2-4 ϫ 10 8 M Ϫ1 , as determined by equilibrium ultracentrifugation and size exclusion chromatography. Consistent with this K a value, we have also demonstrated, by using size exclusion chromatography after addition of 125 I-SCF as a tracer, that the SCF monomer is probably the major species in human serum preparations.
Based on Bazan's prediction of the genetic and structural homology between SCF and M-CSF (17), and the x-ray crystallographic structure of M-CSF (16), the Phe 63 in SCF probably resides at the dimer interface. Variants with Phe 63 substituted were therefore studied. Variant SCF(F63C) and its derivative SCF(F63C)S-CH 2 CONH 2 exist primarily as monomers, as indicated by size exclusion chromatography (K a 1.8 ϫ 10 3 M Ϫ1 for the derivative). Thus substitution of Phe 63 with Cys significantly reduces formation of the SCF dimer. Substitution with  (5). Note that the endogenous serum SCF is glycosylated (5). We are assuming that the behavior of 125 I-rhSCF reflects that of total SCF, including the endogenous. Given ability for subunit interchange (4), this assumption is reasonable.
b Percent monomer was calculated as the integrated area of peak b (monomer) divided by the sum of the integrated areas of peak b (monomer) and peak a (dimer).
Leu 63 in the variant SCF(V49L,F63L) resulted in a K a of 1.47 ϫ 10 5 M Ϫ1 . Replacing the bulky Phe by Leu could result in unfavorable packing in the hydrophobic dimer interface. We believe that Leu, replacing Val at position 49, would reside in the hydrophobic side of the amphipathic B helix and would not be in the interface of the dimer, again based on the structural analogy to M-CSF. Leu 49 therefore would be less likely to affect SCF dimerization. The conformation of these two variants, determined by circular dichroism and fluorescence, is indistin-guishable from that of wild type rhSCF. However, these methods are unlikely to detect small structural alterations.
The disulfide-linked SCF(A165C) dimer would be 100% dimer at all concentrations. Since SCF 1-141 is fully active (29) and is a non-covalently associated dimer as judged by size exclusion chromatography, the Ala 165 almost certainly does not reside in the dimer interface. The C-terminal portion, which would include the disulfide bond Cys 165 -Cys 165 of the SCF(A165C) disulfide-linked dimer, is probably flexible. The phenomenon of dynamic association/dissociation at the dimer interface might still occur in the SCF(A165C) disulfide-linked dimer. However, the association at the interface would be much enhanced because of the interchain disulfide bond (Cys 165 -Cys 165 ).
The mitogenic activities of SCF(F63C)S-CH 2 CONH 2 and SCF(V49L,F63L) are 1,300-fold and 30-fold lower, respectively, than that of wild type rhSCF (Fig. 7), while that of SCF(A165C)  is about 10-fold higher. Based on calculation from the K a values listed in Table III, the dimer concentration at 50% maximum in the UT-7 mitogenic bioassay is 0.8 ng/ml for the SCF(A165C) disulfide-linked dimer, 0.8 ng/ml for wild type rhSCF, 0.7 ng/ml for SCF(V49L,F63L), and 18 ng/ml for SCF(F63C)S-CH 2 CONH 2 . Thus the proliferation bioactivity and K a are correlated to some extent, consistent with a hypothesis that dimerization of SCF mediates the mitogenic response. (Models of SCF association with Kit for Kit dimerization and signal transduction are discussed further below.) However, we cannot rule out the possibility that the correlation is coincidental, and that SCF(V49L,F63L) and SCF(F63C)S-CH 2 CONH 2 have lower activity because of other structural alterations that might affect their ability to interact with the receptor.
Since the dissociation rate constant, k d , for wild type SCF has been determined previously (4), we can estimate the association rate constant k a from the K a determined in the present work: k a ϭ K a ϫ k d ϭ 4 ϫ 10 4 M Ϫ1 s Ϫ1 . Given a soluble SCF concentration of 3.3 ng/ml in human serum (5), the K a for SCF dimerization also allows the calculation that about 90% of the FIG. 6. Circular dichroism spectra and fluorescence spectra of E. coli-derived wild type rhSCF and its variants. Panel A represents the near UV spectra of wild type rhSCF (2) and SCF(F63C) monomer (1). Panel B represents the far UV spectra of wild type rhSCF (2), SCF(F63C) monomer (1), SCF(F63C)S-CH 2 CONH 2 (3), and SCF(F63C) disulfide-linked dimer (4). Panel C represents the fluorescence spectra of wild type rhSCF (2) and SCF(F63C) monomer (1).
K a ϭ ͑A Ϫ X͒ ϫ 10 Ϫ6 g/liter 2 ϫ 18,649 g/mol ͫ X ϫ 10 Ϫ6 g/liter 18,649 g/mol ͬ ϭ ͑A Ϫ X͒ X 2 ϫ 9.32 ϫ 10 9 M Ϫ1 K a X 2 ϩ9.32ϫ10 9 XϪ9.32ϫ10 9 Aϭ0 X ϭ Ϫ9.32 ϫ 10 9 ϩ ͱ͑9.32 ϫ 10 9 ͒ 2 ϩ 4 ϫ K a ϫ 9.32 ϫ 10 9 2K a ng/ml SCF would be monomeric, and the size exclusion chromatographic experiments with 125 I-SCF added to human serum support this calculation. However, we have shown in vitro that at concentrations where rhSCF is dimeric, a complex is formed with sKit that includes two rhSCF monomers and two sKit molecules (15), and we consider it most likely that SCF dimer mediates Kit dimerization. Models by which SCF could induce receptor dimerization and consequently activate signal transduction are shown in Fig. 9. In these models, some correlation between K a for rhSCF dimerization and biological activity, such as that noted above, would be anticipated. The models differ from each other primarily in that each SCF monomer interacts with only one receptor molecule in model I, whereas each SCF monomer interacts with both receptor molecules in model II. In this sense, model II is analogous to the 1:2 growth hormone-growth hormone receptor complex (30), in which a single growth hormone molecule mediates receptor dimerization by interacting with both receptor molecules. It is important to note (e.g. with regard to monomeric SCF which may be present in human serum) that each model allows that an SCF molecule be initially monomeric and still wind up as part of a 2:2 SCF⅐sKit complex.
It is also noteworthy that reported K a values for SCF binding to Kit (ranging from 0.6 ϫ 10 8 to 2 ϫ 10 9 M Ϫ1 , depending on the system; see Refs. 14, 15, 18, 31, and 32) are close to the K a reported here (2-4 ϫ 10 8 M Ϫ1 ) for rhSCF dimerization. This could be considered consistent with a model in which SCF⅐sKit interactions were relatively strong and sKit⅐sKit interactions relatively weak or non-existent, such that SCF⅐SCF interactions determine the apparent K a for formation of the 2:2 SCF⅐Kit complex (15). We do not attempt to address the issues FIG. 8. A, RP-HPLC and SDS-PAGE analysis of the rhSCF (A165C) preparation. The two peaks, T42 and T48.7, were collected, vacuum-dried, and then reconstituted in PBS. The samples were analyzed by SDS-PAGE under non-reducing conditions. The majority of peak T48.7 appeared as disulfide-linked dimer. Peak T42 appeared as a monomer, but its molecular weight is slightly higher than that of unmodified rhSCF. B, Asp-N endoproteinase peptide maps of SCF species. Peaks g, h, and i are Cys 165 -containing peptides. The mass and sequence data of these peptides are listed in Table II. of conformational changes and cooperative interactions in SCF⅐Kit complex formation (14,15,33).
In summary, the details of SCF⅐Kit interactions will require further elucidation, but it is apparent that the association and dissociation among SCF monomer, SCF dimer, and Kit are dynamic, and play an important role in the regulation of the development and function of hematopoietic cell lineages and other cells such as mast cells, germ cells, and melanocytes.