Isolation and characterization of a disulfide-linked human stem cell factor dimer. Biochemical, biophysical, and biological comparison to the noncovalently held dimer.

Distinct from the noncovalently linked recombinant human stem call factor (rhSCF) dimer, we report here the isolation and identification of an SDS-nondissociable dimer produced during folding/oxidation of rhSCF. Experimental evidence using various cleavage strategies and analyses shows that the isolated dimer is composed of two rhSCF monomers covalently linked by four disulfide bonds. The cysteines are paired as in the noncovalently associated dimer except that all pairings are intermolecular rather than intramolecular. Other structural models, involving intertwining of intramolecular disulfide loops, are ruled out. The molecule behaves similarly to the noncovalently associated dimer during ion-exchange or gel permeation chromatography. However, the disulfide-linked dimer exhibits increased hydrophobicity in reverse-phase columns and in the native state does not undergo spontaneous dimer dissociation-association as seen for the noncovalent dimer. Spectroscopic analyses indicate that the disulfide-linked and noncovalently associated rhSCF dimers have grossly similar secondary and tertiary structures. In vitro, the disulfide-linked dimer exhibits approximately 3-fold higher biological activity in supporting growth of a hematopoietic cell line and stimulating hematopoietic cell colony formation from enriched human CD34+ cells. The molecule binds to the rhSCF receptor, Kit, with an efficiency only half that of the noncovalently associated dimer. Formation of intermolecular disulfides in the disulfide-linked dimer with retention of biological activity has implications for the three-dimensional structure of noncovalently held dimer and disulfide-linked dimer.

other cell lineages, including melanocytes and germ cells (7,8). SCF is initially synthesized as membrane-bound forms of 248 or 220 amino acids, depending on alternative splicing of exon 6. A soluble SCF form of 165 amino acids is biologically functional, apparently arising by proteolytic release from the extracellular domain of the membrane-bound 248-amino acid SCF (9 -11). The naturally occurring soluble SCF is glycosylated at both N-linked and O-linked sites (12,13).
In a companion paper (23), we described the isolation and characterization of intermediates derived during folding and oxidation of the reduced and denatured rhSCF and the assignment of a predominant in vitro folding pathway. The major folded SCF is the noncovalently linked dimer (SDS-dissociable) and a small fraction is SDS-nondissociable dimer. In the present study, we isolate the nondissociable dimer to apparent purity and demonstrate that it is biologically functional and is covalently linked by four intermolecular disulfide bonds involving all cysteinyl residues. The biological, biochemical, biophysical, and structural properties of the noncovalently and covalently linked dimers are compared, and the results provide some insights to the structure and function of SCF.

EXPERIMENTAL PROCEDURES
Materials-E. coli-derived rhSCF (SDS-dissociable dimer) was purified according to methods described previously (17,18). The recombinant molecule contains 165 amino acids plus an N-terminal methionine at position Ϫ1. Iodoacetic acid was purchased from Sigma. HPLC solvents and water were purchased from Burdick and Jackson. Sequencing reagents and solvents were supplied by Applied Biosystems (Foster City, CA) and Hewlett Packard (Mountain View, CA). All other reagents were of the highest quality available. SeaPlaque Agarose low gelling temperature was obtained from FMC BioProducts (Rockland, ME). X VIVO-15 medium and fetal calf serum were purchased from BioWhittaker Inc. (Walkerville, MD) and Hyclone Labs (Logan, UT), respectively.
Isolation of SDS-nondissociable Dimer-Recovery of rhSCF expressed in E. coli includes solubilization of rhSCF-containing inclusion bodies, oxidation, and folding and subsequent chromatographic steps (18). After cation-exchange chromatography using an S Sepharose column, pooled SCF (approximately 1 liter containing 600 mg of rhSCF) was further subjected to C-4 reverse-phase chromatography performed with a BioCat liquid chromatographic system (Perceptive Inc., NJ). Sample in 10 mM sodium acetate buffer, pH 4.5 was loaded onto a column (2.6 ϫ 7.8 cm) packed with C4 silica gel (100 Å, wide pore; Vydac, San Jose, CA) preequilibrated with 12 mM HCl. After loading, the column was washed with mobile phase A (30% ethanol in 12 mM HCl). The separation was accomplished by a linear gradient from 0% mobile phase A to 100% mobile phase B (80% ethanol in 12 mM HCl) at a flow rate of 20 ml/min for 10 h. The column effluent was monitored continuously by a UV detector set at 280 nm for protein detection. SDS nondissociable SCF dimer elutes later than the bulk of the SCF (18) (see "Results"); it was pooled separately and immediately diluted into 10 volumes of 10 mM sodium acetate buffer, pH 5.0, concentrated by ultrafiltration, and buffer-exchanged by diafiltration using the same sodium acetate buffer. All purification steps were carried out at 5°C. The final material at 7 mg/ml was stored at Ϫ80°C.
Analytical HPLC-Reverse-phase HPLC was performed using TFAacetonitrile gradient elution. A Vydac C4 column (4.6 mm ϫ 25 cm; 300 Å) was equilibrated with 97% solvent A (0.1% TFA), 3% solvent B (0.1% TFA in 90% acetonitrile) with 215 and 280-nm UV detection at a flow rate of 0.7 ml/min. After samples were injected into the column, the following elution program was used: a linear gradient to 20% solvent B in 5 min and to 70% B in 60 min, then isocratic elution at 70% B for 20 min.
Structural Characterization-Endoproteinase Asp-N digestion of SCF samples and peptide mapping procedures were essentially identical to those described earlier (18). N-terminal amino acid sequence analysis of peptides was performed on an automatic protein sequencer (Applied Biosystems models 477A and 470 or Hewlett Packard HPG1000A) as described elsewhere (12). Procedures used to sequence peptides recovered from gel bands electroblotted onto PVDF membranes were described in a previous report (24).
Hydrogen Peroxide Oxidation of SCF Dimer-The conditions used were found to completely oxidize all Met residues except Met 48 . SDSnondissociable rhSCF dimer at 1 mg/ml in 10 mM sodium acetate, pH 5.0, was incubated with 0.5% (w/v) H 2 O 2 at 25°C for 3 h. After reaction, the mixture was purified by analytical reverse-phase HPLC as described above. A major oxidized peak eluting earlier than the unoxidized SCF dimer was obtained. Peptide mapping using Asp-N endoproteinase digestion (below) showed that all Met residues but Met 48 were completely converted to the sulfoxide derivatives. Only a small fraction (about 10%) of Met 48 was oxidized. The HPLC purified sample was subjected to CNBr cleavage as described below.
Chemical Cleavages and Analysis of Cleavage Products-Cleavage of rhSCF dimer species by BNPS-skatole (Pierce) was carried out in 50% acetic acid as described elsewhere (12). This reaction allows the cleavage of the sample at Trp 44 , which is the lone Trp in each monomer.
A complete CNBr cleavage at the Met residues of SCF dimer species or H 2 O 2 -oxidized SCF dimer species was performed as follows. Vacuumdried samples were redissolved in 70% formic acid (0.2 mg in 150 l) and then incubated with freshly prepared CNBr (400 molar ratio to SCF) at 25°C for 24 h in the dark. For partial CNBr cleavages, 50-fold molar excess of CNBr was used and the incubation times were shortened to 2-8 h. All the cleaved samples were immediately vacuum dried for further analysis.
Aliquots of dried samples (5-20 g) were loaded onto individual lanes of precast 16% Laemmli polyacrylamide gels (10 wells; Novex Inc., San Diego, CA) and electrophoresed (25) under nonreducing and reducing conditions. After Coomassie Blue staining and destaining, protein band intensity in each gel lane was measured using an image scanner (PDI Inc., New York); images were integrated using PDQuest software (PDI Inc.). In separate analyses, gel bands were also electrophoretically transferred onto PVDF membrane and the Coomassie Blue-stained bands were excised for N-terminal sequence analysis (24).
Limited Proteolysis by Endoproteinase Lys-C-A sample (1 mg/ml) was reconstituted in 20 mM Tris-HCl buffer and digested with endoproteinase Lys-C (enzyme-to-substrate ratio ϭ 1:100) at 25°C. At 15 min and 2 h, sample aliquots (100 l each) were taken, and digestion was stopped by adding 5 l of 20% TFA. Samples of 5-20 g were dried completely and subjected to SDS-PAGE as described.
Partial Reduction of SCFs-One mg/ml solutions of SCF dimer species were incubated in the presence of 1.24 mg/ml DTT in 0.1 M Tris-HCl buffer (pH 8.5) containing 2.5 M urea, 60 mM NaCl, 2 mM EDTA. Aliquots of the reaction mixture were removed at selected time intervals, and unreacted thiols were blocked by the addition of 1 M iodoacetic acid (10:1 molar ratio to the thiol) in 0.3 M Tris, pH 8.0, for 2 min at room temperature. Samples were then quickly frozen in a methanol/dry ice bath and subsequently analyzed by reverse phase-HPLC using conditions described previously (23).
Biophysical Studies-Circular dichroism (CD) and fluorescence spectroscopic studies and thermostability analysis of SCF forms were performed as described in a companion study (23). Samples were in 20 mM Tris-HCl, pH 7.5, for all biophysical analyses.
In Vitro Mitogenic Activity and Radioreceptor Assays-[ 3 H]Thymidine incorporation by the human megakaryoblastic leukemia cell line UT-7 was monitored in the cell proliferation assay as described previously (26). Receptor binding assays were also performed using whole cell lysate containing membrane-associated Kit receptor as reported elsewhere (26).
Purification of CD34 ϩ Cells-Human leukapheresis products were obtained from consenting healthy donors and patients with approval of the Institute review committee. Isolation of CD34 ϩ cells from peripheral blood and bone marrow cells was carried out using Miltenyi Mini-Macs columns as suggested by the manufacturer (Miltenyi Biotec. Inc., Sunnyvale, CA). Briefly, Ficoll-Paque medium (Pharmacia Biotech Inc.) was used to fractionate mononuclear cells. The mononuclear cells were subsequently stained with mouse anti-human CD34 antibody (QBEND/ 10) and colloidal magnetic microbeads recognizing mouse IgG. The stained cells were then applied to the MiniMacs column with magnetic separator and washed with 500 l of buffer four times. After removal of the column from the magnet, the retained cells were eluted with 500 l of phosphate-buffered saline with 0.5% bovine serum albumin. Approximately 80% pure CD34 ϩ cells were routinely isolated by this method as judged by fluorescein-activated cell sorting (27).

RESULTS
Isolation of SDS-nondissociable SCF Dimer-Expression of rhSCF in bacteria has resulted in the production of insoluble, inactive SCF accumulated in inclusion bodies. Solubilization and in vitro folding and oxidation are therefore necessary for the recovery and chromatographic purification of active SCF (17,18,23). The rhSCF isolated in this way is a noncovalently linked, SDS-dissociable dimer (17), like naturally occurring SCF (1). However, during the cationic exchange column after folding and oxidation, we have noticed that rhSCF bands of 18.5 and 37 kDa co-elute, as analyzed by nonreducing SDS-PAGE analysis (Fig. 1A, lanes 1 and 2). Sequence analysis of these two bands electrophoretically transferred onto a PVDF membrane revealed that both have the expected N-terminal sequence of rhSCF. When SDS-PAGE was performed under reducing conditions, the 37-kDa SCF band disappeared and the 18.5-kDa SCF band intensity increased, suggesting that the 37-kDa band is SCF dimer dissociable upon reduction (data not shown).
Preparative reverse-phase column chromatography resolved these two species (Fig. 1B). Peak 1 eluting earlier represents the rhSCF which is 18.5 kDa on nonreducing SDS-PAGE and on reducing SDS-PAGE (Fig. 1A, lanes 4 and 6; note that the reduced protein migrates slightly faster than the nonreduced, as described previously (17) and as is typical for proteins containing intramolecular disulfide bonds). This form corresponds to the active noncovalently associated, SDS-dissociable rhSCF dimer (17). Peak 2 represents the SCF dimer which is 37 kDa on nonreducing SDS-PAGE and about 18.5 kDa upon reducing SDS-PAGE (Fig. 1A, lanes 7 and 5, respectively). Analytical reverse-phase HPLC using a TFA-acetonitrile gradient elution is shown in Fig. 1C (bottom chromatogram). This analysis provides a full resolution of the two species and accurately estimates that the disulfide-linked dimer form is 10 -20% of the total (varying somewhat between preparations) (top chromatogram). The middle and bottom chromatograms demonstrate the purity of both peaks obtained from the preparative C4 chromatography.
Peptide Map Analyses- Fig. 2 shows peptide map analyses to compare the primary sequence and disulfide structure between the two forms. The two maps derived from Asp-N endoproteinase digestion are indistinguishable. Since peptides 1 and 2, which contain the disulfide bonds Cys 4 -Cys 89 and Cys 43 -Cys 138 , respectively, are present in both maps and since the monomers of the SDS-nondissociable SCF dimer become dissociable in the presence of reducing agent, as shown above, it follows that there are two types of structural models to explain the lack of dissociation in SDS (Fig. 3).
These two types of models are intermolecularly disulfidelinked dimers and concatenated dimers (Figs. 3, A and B, respectively). There are three disulfide-linked dimers, of which cysteines are involved in the formation of four intermolecular S-S linkages (structure A1), or two inter-and two intramolecular S-S bridges (structures A2 and A3). The five possible concatenated dimers would contain interlocked, but not covalently linked, monomers (Fig. 3B). Structure B1 is a dimer concatenated by N-terminal disulfide loops of the two monomers, while structure B2 is interlocked by two C-terminal disulfide loops. Structures B3 and B4 are dimers with a respective N-and C-terminal disulfide loop of one monomer locked into the other monomer near a sequence region (between residues 44 and 88) shared by both N-and C-terminal disulfide loops. Structure B5 is concatenated between the N-terminal disulfide loop of one monomer and the C-terminal loop of the other. In order to determine which structure corresponds to the isolated SDS-nondissociable dimer, the experiments described in the following five sections were performed.  (Table I).
On reducing SDS-PAGE (Fig. 4A, lane 3), a single 13-kDa band corresponding to sequence from Val 49 to Met 159 was found. As  Table II, all dimer structures shown in Fig. 3 could lead to these data upon complete CNBr cleavage.
In separate experiments, limited CNBr cleavages of the SDS nondissociable rhSCF dimer were performed to help distinguish between the structural models. It can be seen in Fig. 4 (B and C) and in Table I, that the incomplete cleavages lead to bands about 17 kDa and about 35 kDa on nonreducing SDS-PAGE; the relative amounts of these bands depend on the degree of partial cleavage. From the sequence data of Table I, it appears that the 17-kDa bands represent material cleaved at all Met residues. The 35-kDa bands represent material cleaved at Met residues Ϫ1, 27, and 36 (note that the Ile 28 to Met 36 peptide is lost in the gel analysis and therefore not detected).
The key point is that this 35-kDa material is not cleaved at Met 48 since the sequence for the peptide beginning at Val 49 is low or missing. Therefore, the disulfide loop formed by Cys 4 -Cys 89 disulfide bonding would be opened up, but that formed by the Cys 43 -Cys 138 disulfide bond would not. The fact that this material retains the size of 35 kDa on nonreducing SDS-PAGE is inconsistent with structures B1, B3, and B5, but consistent with all the other structures, as indicated in Table II.
Limited Endoproteinase Lys-C Digestion- Fig. 4D shows SDS-PAGE of digests generated by limited proteolysis (nonreducing conditions, for 15 min (lane 1) or 2 h (lane 2)) with Lys-C, which cleaves at the C-terminal side of Lys residues. The SCF polypeptide has Lys residues at positions 13, 17, 24, 31, 62, 78, 91, 96, 99, 100, 103, 127, 148, and 156. After the limited proteolysis, bands are still apparent near the 36-kDa position. When these bands were transferred to PVDF membrane and sequenced, two sequences, M Ϫ1 -E-G-I-C . . . and S 104 -P-E-P-R . . . , were detected in equivalent yields, suggesting that there is complete cleavage after Lys 103 . Several small peptides were also isolated from the digest by reverse-phase HPLC and shown by sequence and mass spectrometric analyses to be D 99 -L-K, K 100 -S-F-K, D 149 -S-R-V-S-V-T-K-P-F-M-L-RP-V-A-A, and P 157 -F-N-L-P-P-V-A-A. The latter two are Cterminal peptides not in the disulfide loops. Identification of these small peptides indicates that there was also partial cleavage by Lys-C after Lys 96 , Lys 99 , Lys 148 , and Lys 156 . Verification of these complete and partial cleavages was provided by reducing SDS-PAGE (Fig. 4D, lanes 3 and 4). Only three large peptides, 11, 10, and 7 kDa, were seen; the 11-and 10-kDa bands had the rhSCF N-terminal sequence M Ϫ1 -E-G-I-C . . . , and the 7-kDa peptide had the sequence S 104 -P-E-P-R . . . In this case, the key point is that cleavage after Lys 103 , i.e. within the Cys 43 -Cys 138 disulfide loop, still leaves material which migrates near 35 kDa on nonreducing SDS-PAGE. This finding is inconsistent with structures B2, B4, and B5, but consistent with all of the other structures (Table II).
Trp Cleavage by BNPS-Skatole-BNPS-skatole cleavage of the dimer will generate a single clip, after Trp 44 . Fig. 4E (both  lanes 1 and 2) shows the nonreducing SDS-PAGE analysis of the cleavage products. Approximately 20% of the material becomes "monomer" (about 18 kDa). N-terminal sequence analysis of the 18-kDa band revealed two sequences, M Ϫ1 -E-G-I-C-R-. . . and I 45 -S-E-M-V-V-. . . in identical yield, indicating cleavage after the Trp 44 residue. This result is consistent with structures A1, B3, and B4, but inconsistent with the other structures (see Table II). About 80% of the BNPS-skatole treated material remained at the "dimer" position (about 18 kDa) on nonreducing SDS-PAGE. Sequence analysis showed that most of this material was unclipped, but about 8% was clipped. However, note that for structures A1, B3, and B4 it is possible in each case to have cleavage of one chain but not the other, and still retain the "dimer" size on nonreducing SDS-PAGE.
CNBr Cleavage after H 2 O 2 Oxidation-In another approach, the SDS-nondissociable dimer was reacted with H 2 O 2 under conditions which completely oxidize Met residues Ϫ1, 27, 36, and 159, but only partially (about 10%, as indicated by peptide mapping analyses) oxidize Met 48 (see "Experimental Procedures"). The oxidized material was then subjected to exhaustive CNBr cleavage. Since Met sulfoxide residues are resistant to CNBr cleavage, we expected to get selective cleavage after Met 48 in the 90% of polypeptide chains which were not oxidized at this position, and sequence analysis of the digest confirmed this expectation. Fig. 4F shows SDS-PAGE analyses. Note in lane 4 (reducing condition) that some material remains uncleaved (about 18 kDa), whereas the majority has been cleaved. Sequence analysis showed that the 6-and 13-kDa bands correspond to the peptides generated by cleavage at Met 48 . In lane 2 (nonreducing condition), the material at 18 kDa has two equivalent sequences, corresponding to the rhSCF N terminus and to the peptide sequence starting at Val 49 . Thus, as with cleavage after Trp 44 , cleavage after Met 48 generates "monomer" (18 kDa on nonreducing SDS-PAGE), a finding which is consistent only with structures A1, B3, and B4 (see Table II). Nineteen percent of the material visualized in lane 2 remains at the "dimer" position (about 35 kDa). The material at this band position has sequences corresponding to the rhSCF N terminus and to the peptide beginning at Val 49 , in a ratio of 2:1; this result is expected since 10% of the H 2 O 2 treated material was oxidized at Met 48 and therefore uncleavable with CNBr. As noted above for the Trp cleavage data, structures A1, B3, and B4 all allow for retention of "dimer" if only one chain and not the other is cleaved.
Partial Reduction of Dimeric SCFs-Native, SDS-dissociable rhSCF dimer was partially reduced with DTT followed by alkylation with iodoacetate, and the resulting mixture was analyzed (by reverse-phase HPLC and peptide mapping of the HPLC peaks). The Cys 4 -Cys 89 bond was found to be preferentially reduced, with generation of an intermediate (I-2) containing only the Cys 43 -Cys 138 bond (Fig. 5A, top panel); reduction of the Cys 43 -Cys 138 bond follows at later times (see Ref. 23). When the SDS-nondissociable SCF dimer was similarly subjected to partial reduction and alkylation, no I-2 was detected.
Instead, two unique peaks, a and b, were resolved by HPLC at retention times later than that of the SDS-nondissociable SCF dimer (Fig. 5A, bottom panel). By sequence analysis, both species gave a clear PTH-Cys (Cm) signal at position 4; the signal for peak b was about half that for peak a. In addition, both peaks a and b migrate at the 37-kDa "dimer" position on nonreducing SDS-PAGE (Fig. 5B, lanes 6 and 7). These findings indicate that peak a is "dimeric" material in which both Cys 4 -Cys 89 disulfide bonds have been broken, and the Cys 43 -Cys 138 disulfide bonds are intact, while peak b is "dimeric" material in which only one of the Cys 4 -Cys 89 disulfide bonds has been broken. Since there is no detectable "monomeric" material (i.e.   I-2), we conclude that the data are inconsistent with structures A2, B1, B3, and B5, but consistent with the other structures (see Table II). As summarized in Table II, the only structure for the SDSnondissociable rhSCF dimer compatible with all the results of the last five sections is structure A1, with four intermolecular disulfide bonds involving all four Cys residues of each monomer. Therefore, we will subsequently refer to the SDS-nondis-sociable dimer as disulfide-linked dimer.
Other Biochemical Properties-The usual noncovalently associated rhSCF dimer can undergo spontaneous rapid monomer dissociation-reassociation (22). This was shown with the use of an N10D variant of rhSCF which migrates differently from the wild type on ion-exchange HPLC. Upon mixing the N10D and wild type molecules, the appearance of hybrid dimer could be monitored by the ion-exchange HPLC method. Not surprisingly, the disulfide-linked dimer did not undergo such dissociation-reassociation and monomer exchange (data not shown).
The noncovalently associated rhSCF dimer and the disulfidelinked dimer have identical elution times and elution profiles on high resolution ion-exchange HPLC and gel filtration, indicating that they have similar surface charge distribution and molecular size in solution.
Spectroscopic Properties and Thermostability- Fig. 6, A and B, compares the far-UV and near-UV CD spectra of noncovalently associated rhSCF dimer and the disulfide-linked dimer. In solution under the described conditions, CD spectra of the two forms are essentially identical. The differences in the CD magnitudes for the near UV CD (Fig. 6B) are minor. Fig. 6C measures changes in the CD ellipticity at 222 nm as a function of temperature, demonstrating the disappearance of helical structure caused by thermodenaturation. The helix-coil transition occurs similarly for both species, with a temperature of 71-74°C for half-maximal melting. Fig. 6D compares the Trp fluorescence spectra of the two forms. Again, the spectra are essentially identical, suggesting that the local environment near Trp and the energy transfer to Trp from the nearby Tyr residue(s) are similar for the two forms. Thus, the various biophysical analyses indicate that the two dimer forms have grossly similar secondary and tertiary structural folding and conformation.
Biological Activities- Fig. 7A shows a comparison of the two dimer species in stimulating proliferation ([ 3 H]thymidine uptake) of UT-7 human megakaryocytic leukemia cells cultured in vitro. The specific activity (amount of SCF giving half-maximal [ 3 H]thymidine incorporation) of the disulfide-linked rhSCF dimer is reproducibly 3-5-fold higher than that of the noncovalently associated SCF dimer in this system.
It has been repeatedly shown previously that SCF acts synergistically with later-acting hematopoietic cytokines to stimulate in vitro colony formation from hematopoietic progenitor cells (2,26). For example, with purified CD34 ϩ bone marrow cells, BFU-E colony formation is minimal with SCF or EPO alone, but is marked with SCF and EPO together; similarly GM-CFC colony formation is minimal with SCF or rhG-CSF  alone, but marked with SCF and rhG-CSF together. Doseresponse curves for the two rhSCF dimer species are shown in Fig. 7B (BFU-E; fixed concentration of EPO) and Fig. 7C (GM-CFC; fixed concentration of G-CSF). In both assays, the disulfide-linked dimer has a potency severalfold greater than that of noncovalently associated SCF dimer.
In contrast to the situation with biological potencies on cells, the receptor binding affinity of disulfide-linked rhSCF is about 2-fold lower than that of noncovalently associated rhSCF. This is demonstrated in Fig. 7D, in which the abilities of the rhSCF species to compete with 125 I-labeled rhSCF for binding to receptor are determined. DISCUSSION Biologically active rhSCF can be recovered after folding and oxidation of inactive rhSCF in the solubilized extracts prepared from inclusion bodies after recombinant expression in bacteria (17,18,23). The major folded/oxidized form is the SDS-dissociable, noncovalently associated dimer, corresponding to naturally occurring SCF. However, a significant amount of the rhSCF is the SDS nondissociable dimer that is separable from the main rhSCF form by reverse-phase chromatography. Our data allow us to propose and distinguish between eight different structural models that could explain why the minor form gives peptide mapping results (including disulfide-linked peptides) the same as those of the major form, but is SDS-nondis-sociable. As we show, structure A1, with the four intermolecular disulfide bonds, appears to represent the actual material.
As described in our companion study (23), intermediate I-1 with a Cys 4 -Cys 89 disulfide bond is the main intermediate form during rhSCF folding and oxidation. This and other intermediates lead to the noncovalently associated SCF dimer with intramolecular disulfides, but could also undergo disulfide rearrangement to form intermolecular disulfides. For such events to occur, the partially oxidized rhSCF monomers would have to be associated prior to disulfide formation; we have shown that all of the intermediate forms that have been identified are in dimeric state (23).
Many of the biochemical and biophysical properties of the noncovalently associated dimer and the disulfide-linked dimer appear indistinguishable, including surface charge, molecular size, plus secondary and tertiary structure and local environments. The disulfide-linked dimer does behave differently than the noncovalently associated dimer on reverse-phase HPLC at low pH and in the monomer dissociation-reassociation experiments. In each case the differences essentially reflect the covalent attachment of the disulfide-linked dimer. For example, with the reverse-phase HPLC, the noncovalently associated SCF dissociates to monomer at the low pH, 2 but the disulfidelinked SCF dimer is obviously unable to dissociate.
The biological properties of the covalent dimer are noteworthy. Its activity toward hematopoietic target cells is 3-fold higher than the activity of noncovalently associated dimer. However, in Kit receptor binding experiments, the disulfidelinked dimer if anything displayed slightly lower affinity for Kit in comparison with the noncovalently associated dimer. How can similar (or lower) receptor binding be reconciled with higher biological activities for the disulfide-linked dimer? Doseresponse for SCF in the in vitro biological assays occurs in the 0.05-2 ng/ml concentration range. In the case of the noncovalently associated dimer, depending on the K a for monomer association to dimer, it is possible that much of the SCF could be monomeric at the 0.05-2 ng/ml concentration range. Lev et al. (29) have proposed that monomeric SCF can mediate the 2 Arakawa, T., unpublished results. dimerization and activation of Kit receptor, but it remains possible that SCF dimer is necessary to mediate Kit dimerization, or at least that SCF dimer may be more effective at doing so than SCF monomer. Since the disulfide-linked SCF dimer would of course be dimeric at all concentrations, it follows that it could be more potent, i.e. more active at low concentrations, than the dissociable dimer.
This line of reasoning implies that the overall quaternary structure, including interactions at the dimer interface, would be similar for the disulfide-linked and noncovalently associated dimers. In considering the structure of SCF, there are many reasons to expect similarity to the structure of M-CSF, which is known (30). As mentioned in the Introduction, SCF and M-CSF are noncovalently associated and disulfide-linked dimers, respectively, and their receptors are related members of the type III tyrosine kinase family. As Bazan (31) has pointed out, the exons of the SCF and M-CSF genes can be closely aligned, suggesting evolutionary relatedness. Both SCF and M-CSF are expressed as longer membrane-bound forms from which soluble forms are released by proteolytic cleavage at sites encoded within exon 6, or as shorter membrane-bound forms (lacking the exon 6-encoded cleavage sites as a result of alternative mRNA splicing) which remain membrane-associated (9 -11, 32). The M-CSF monomer has intramolecular disulfide bonds (Cys 7 -Cys 90 and Cys 48 -Cys 139 ) (21,30) that align well with those of the SCF monomer (Cys 4 -Cys 89 and Cys 38 -Cys 143 ). M-CSF has a third intramolecular disulfide (Cys 102 -Cys 146 ) (21,30) and an intermolecular disulfide joining the Cys 31 residues of each monomer in the M-CSF dimer. These disulfide bonds are apparent in the x-ray crystallographic structure of M-CSF dimer (30). The structure includes the four-helix bundle for each monomer which had been proposed by Bazan (31) for both M-CSF and SCF. The two monomers of M-CSF associate in head-to-head fashion, i.e. the top ends of the helix bundles associate leading to a flat and elongated overall shape. The SCF-equivalent intramolecular disulfide bonds (Cys 7 -Cys 90 and Cys 48 -Cys 139 ) of M-CSF are at the ends of the helix bundles distal to the dimer interface, and the Cys 31 -Cys 31 disulfide bond is of course part of the interface.
Given that the disulfide-linked SCF dimer described here is highly active, we suggest the following speculation as to how its structure may compare to that of the noncovalently associated SCF dimer. If the quaternary structure of noncovalently associated SCF is homologous to that of M-CSF, the monomers of the disulfide-linked dimer would need to be inverted in order to accommodate the disulfide bond formation, without an adverse effect on activity. Alternatively, if the quaternary structures of the disulfide-linked SCF dimer and the noncovalently associated SCF dimer were similar to each other, both could be inverted relative to that of the M-CSF dimer. Third, and perhaps most likely, the quaternary structures of the SCFs could be similar to each other and to that of M-CSF if, for the disulfide-linked SCF dimer, the proposed A and D helices (31) were swapped between the monomers within the dimer. Such swapping would be feasible within the constraints of the proposed SCF structure (i.e. similar to M-CSF structure), could conceivably arise during the refolding of the E. coli-derived recombinant molecule, and would allow the observed intermolecular disulfide bond formation. There is precedent for such swapping of helices or other domains between monomers within overall oligomeric structures, e.g. interleukin 5 (33) and many other proteins as well (34).