Structural requirements for the ectodomain cleavage of human cell surface macrophage colony-stimulating factor.

One form of human macrophage colony-stimulating factor (CSF-1256, M-CSFα) is a member of a restricted set of cell surface transmembrane proteins, which is selected to undergo proteolytic ectodomain cleavage. To determine the substrate requirements for this cleavage, we have constructed a series of mutations in the cytoplasmic tail, transmembrane domain, and juxtamembrane region of CSF-1256 and stably expressed the mutated genes in NIH 3T3 cells. Our results demonstrate that membrane association of the CSF-1 precursor is required for cleavage of its growth factor ectodomain and furthermore that the juxtamembrane region Pro161-Gln162-Leu163-Gln164-Glu165 (PQLQE) (residues 161–165 of the ectodomain) is an essential determinant of cell surface CSF-1256 cleavage and that the cleavage site is partially sequence-specific. Furthermore, a mechanism of steric hindrance, which likely involves interference with protease accessibility, is postulated to explain the observed decreases in the cleavage efficiency in certain CSF-1 mutants. Finally, our results strongly suggest that the CSF-1 ectodomain is cleaved at or very near the cell surface by a membrane-associated proteolytic system.

For example, the failure to properly cleave ␤-APP may increase the production of amyloid ␤-proteins, which are invariantly found in the senile plaques of Alzheimer's disease (Haass and Selkoe, 1993;Selkoe, 1994).
CSF-1 256 (M-CSF␣) is one of the growth factors that is cleaved from the cell surface to yield biologically active soluble growth factors Halenbeck et al., 1988;Stein and Rettenmier, 1991). CSF-1 is a disulfide-linked dimeric glycoprotein regulating the growth, differentiation, and survival of cells of monocytic lineage (Stanley et al., 1983;Rettenmier and Sherr, 1989). It also plays an important role in the regulation of placental development and in bone osteoclast survival (Arceci et al., 1989;Wiktor-Jedrzejczak et al., 1990). Alternative splicing of the primary transcript yields multiple mRNAs encoding multiple forms of membrane-bound CSF-1 precursors (Kawasaki et al., 1985;Ladner et al., 1987). The largest human CSF-1 precursor of 554 amino acids (CSF-1 544 , CSF-1␤) is not expressed on the cell surface but is proteolytically cleaved within the cell and released into the extracellular compartment as both a soluble glycoprotein and a proteoglycan form with glycosaminoglycan chains (Rettenmier and Roussel, 1988;Price et al., 1992;Suzu et al., 1992). In contrast, the smallest CSF-1 precursor of 256 amino acids (CSF-1 256 ) is stably expressed on the cell surface , and it is cell surface CSF-1 256 that is biologically active in stimulating adjacent cells that express CSF-1 receptors (Stein et al., 1990). Furthermore, the growth factor domain of CSF-1 256 is released from the cell surface into the extracellular medium by a regulated process of proteolytic cleavage. The rate of this cleavage is slow in unstimulated cells, but it is accelerated when cells are stimulated by certain agonists such as protein kinase C activators (Stein and Rettenmier, 1991).
Thus far, no protease(s) responsible for ectodomain cleavage of cell surface proteins has been identified. Furthermore, the structural determinants that specify such cleavage have not been clearly elucidated. We have therefore used the cell surface CSF-1 256 cleavage system as a model to define the structural determinants that are required for cell surface protein ectodomain cleavage. In this study, we show that cell membrane association of CSF-1 256 is required for its cleavage and that the CSF-1 juxtamembrane region (residues 150 -165) is essential in determining the cleavage. In addition, a mechanism of steric hindrance, which likely involves interference with protease accessibility, might explain the observed decreases in the cleavage efficiency in certain CSF-1 mutants. Finally, we show that the cleavage of CSF-1 256 occurs at or very near the cell surface and is mediated by a membrane-bound proteolytic enzyme system.
Site-directed Mutagenesis-Human WT CSF-1 256 cDNA (Kawasaki et al., 1985) was subcloned into pBluescriptIISK(PD), a vector modified from pBluescriptIISK(Ϫ) (Stratagene) by deleting part of the polylinker region (674 -719) to facilitate subcloning. A polymerase chain reaction mutagenesis technique was used for generation of each mutation (Steffan et al., 1989). Two internal mutagenizing primers and two external nonmutagenizing primers were used to generate two mutant DNA fragments having overlapping ends. These fragments were purified and used to serve as the template for a secondary polymerase chain reaction with the two external primers to produce the appropriate length of mutant DNA fragment. The secondary polymerase chain reaction products were subcloned back into pBluescriptIISK(PD) CSF-1 256 . Only one internal mutagenizing primer containing the restriction enzyme site and one external nonmutagenizing primer were used if there was an appropriate restriction enzyme site near the site to be mutated. All sequences produced by polymerase chain reaction were confirmed by DNA sequencing using Sequenase (U.S. Biochemical Corp.).
Expression of CSF-1 Mutants in NIH 3T3 Cells-The mutant forms of CSF-1 cDNA in the pBluescriptIISK(PD) vectors were subcloned into an expression retroviral vector PSM, which is a replication-defective vector derived from the SM strain of feline sarcoma virus . Restriction mapping was used to confirm the correct orientation of the inserts for expression. To generate stable transfectants, NIH 3T3 cells were cotransfected with circular plasmid DNA and pSV2neo by the CaPO 4 coprecipitation technique. G418-resistant cell lines were screened by immunoprecipitation to identify stable cell lines that express mutant CSF-1 forms.
Metabolic Labeling-Cells expressing WT or mutant CSF-1 forms were metabolically labeled with [ 35 S]methionine for 20 min and chased with complete medium containing 20 ϫ methionine for 1 h for CSF-1 256 to express on cell surface. Then the 35 S label was chased in the same medium for variable amounts of time in the absence or presence of 0.5 M phorbol 12-myristate 13-acetate (PMA) (Sigma) as indicated. The medium was collected, and the cells were lysed in radioimmune precipitation lysis buffer (50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 20 mM EDTA, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS) with 2% aprotinin and 1 mM phenylmethylsulfonyl fluoride as protease inhibitors for immunoprecipitation. For metabolic labeling under tunicamycin treatment, cells were preincubated for 4 h in medium with tunicamycin (2 g/ml) and labeled and chased in the continued presence of the drug.
Cell Surface Radioiodination-Cells grown in 100-mm tissue culture plates were washed once with phosphate-buffered saline and once with phosphate-buffered saline containing 5 mM glucose. The cells were incubated in 1.0 ml of phosphate-buffered saline containing 5 mM Dglucose, 1.0 mCi of carrier-free Na 125 I (Amersham Corp.), 75 g of lactoperoxidase, and 8.8 g of glucose oxidase (Sigma) for 5 min at 22°C and another 5 min at 4°C with occasional gentle agitation. After washing with cold phosphate-buffered saline twice, the cells were incubated at 37°C for variable amounts of time in complete medium with or without 0.5 M PMA as indicated (Stein and Rettenmier, 1991). The medium was collected, and the cells were lysed in radioimmune precipitation lysis buffer with 2% aprotinin and 1 mM phenylmethylsulfonyl fluoride as protease inhibitors for immunoprecipitation.
Immunoprecipitation-The primary antibody used was the YYG-106 rat monoclonal antibody (Stanley, 1985). Immunoprecipitation was carried out with protein A-Sepharose (Pharmacia Biotech Inc.) precoated with a goat anti-rat IgG (Cappel) as immunoadsorbant. The control for nonspecific precipitation was performed using an isotype-matched rat myeloma protein. Washed immune precipitates were analyzed by SDSpolyacrylamide gel electrophoresis (SDS-PAGE) (Anderson et al., 1984). The labeled proteins were detected by fluorography and quantitated with Betascope 603 (Betagen). Murine CSF-1 produced by parental NIH 3T3 cells was not detected by immunoprecipitation . The apparent molecular weights of the labeled proteins were determined by comparison with the mobilities of protein molecular weight standards.

RESULTS
Ectodomain Cleavage of CSF-1 256 Requires Membrane Association-The CSF-1 256 precursor consists of an amino-terminal signal peptide, a growth factor domain, a small juxtamembrane region, and a transmembrane domain (TM) followed by a short cytoplasmic tail (Fig. 1A). CSF-1 256 is synthesized as a 68-kDa disulfide-linked dimeric glycoprotein whose two identical sub-units undergo sequential cleavage to yield a 56-kDa membrane-associated heterodimeric intermediate and a final soluble 44-kDa dimeric growth factor. The cleavage can be accelerated by PMA (Fig. 1B) (Stein and Rettenmier, 1991). To address whether the cytoplasmic tail is required for the cleavage, NIH 3T3 cells were transfected with mutant CSF-1 256 cDNA encoding CSF-1 256 -TC31 in which the carboxyl terminus (31 amino acids) of the cytoplasmic region (36 amino acids) was truncated. CSF-1 256 -TC31 was expressed on the cell surface ( Fig. 1B). Similar to the cleavage of wild type (WT) CSF-1 256 , it was cleaved slowly under basal conditions, and the cleavage was accelerated by the addition of PMA (Fig. 1B).
To define the role of the transmembrane domain in the cleavage, a CSF-1 mutant with the deletion of 16 amino acids (residues 170 -185) in the transmembrane domain (CSF-1 256 -⌬TM16) was produced and stably expressed in NIH 3T3 cells. CSF-1 256 -⌬TM16 was not expressed on the cell surface as demonstrated by cell surface radioiodination (data not shown). After metabolic labeling of cells expressing CSF-1 256 -⌬TM16, full-length CSF-1 256 -⌬TM16 with a molecular mass of 64 kDa was secreted into the medium without cleavage (Fig. 1C). Furthermore, the secretion of CSF-1 256 -⌬TM16 was not stimulated by PMA (Fig. 1C). Taken together, these results demonstrate that membrane association of the CSF-1 256 precursor is required for the cleavage.
PQLQE  in the Juxtamembrane Region Is an Essential Determinant of CSF-1 256 Ectodomain Cleavage-To define the role of the juxtamembrane region in the cleavage process, four deletion mutants spanning this region were constructed and stably expressed in NIH 3T3 cells. Dele- tion of 156 -160 (CSF-1 256 -⌬156 -160), containing the cleavage site 157-159 (Halenbeck et al., 1988), reduced the rate of cleavage to about 30% of the rate of WT as determined by quantitative fluorography (Fig. 2). Deletion of residues 150 -156 (CSF-1 256 -⌬150 -156), N-terminal to the cleavage site, reduced the rate of cleavage to about 20% of the rate of WT (Figs. 2 and 3). In marked contrast, deletions of 5 or 7 amino acids, Cterminal to the cleavage site and immediately N-terminal to the start of the transmembrane domain, CSF-1 256 -⌬159 -165 or CSF-1 256 -⌬161-165, completely abolished the cleavage (Figs. 2 and 3). Abolished or reduced cleavage of these mutant CSF-1 proteins was not due to transport deficiency of these proteins, since all of these mutants were expressed on the cell surface with similar efficiency to WT CSF-1 256 as demonstrated by cell surface radioiodination (data not shown). These studies strongly suggest that the juxtamembrane region PQLQE (residues 161-165) is essential for cleavage and that region 150 -160 contributes to the efficiency of the cleavage.
A series of substitution mutants were also constructed and stably expressed in NIH 3T3 cells to test the effects of amino acids with different side chain size and/or charge and amino acids that might change the normal secondary structure of the juxtamembrane region in CSF-1 256 protein (Fig. 2). Pro 161 was mutated to Ala, Gly, or Ser. Each of these three substitutions interfered with the cleavage process. Of the three mutations, Pro161Ser reduced the rate of cleavage the most as compared with that of WT (Figs. 2 and 3). Gln 162 was mutated to Glu, Lys, or Pro. Each of the three mutant CSF-1 256 forms was cleaved as efficiently as WT (Fig. 2). Substitutions, including change of Leu 163 to Pro, Gln 164 to Glu, Gln 164 to Lys, and Glu 165 to Gln, did not interfere with cleavage (Fig. 2). However, substitution of Leu 163 -Gln 164 with Ile-Pro substantially reduced the rate of cleavage (Fig. 2). These results are summarized in Fig. 2 and show that the PQLQE region (residues 161-165) is essential for cleavage, that the length of this PQLQE region is critical, and that the amino acid sequence requirement in this PQLQE region is partially specific.
The cleavage Site  in the Juxtamembrane Region Is Partially Sequence-specific-Results from a previous study suggest that cell surface CSF-1 256 is cleaved at or near Ser 158 (Halenbeck et al., 1988). A deletion encompassing the cleavage site (CSF-1 256 -⌬156 -160) substantially reduced the rate of cleavage but did not completely abolish the cleavage, suggesting that the cleavage is partially sequence-specific in this region. To test this point further, we changed Ser 158 -160 to amino acids with different side chain size and charge, including CSF-1 256 -Lys 158 -160 , CSF-1 256 -Ala 158 -160 , CSF-1 256 -Asp 158 -160, and CSF-1 256 -Leu 158 -160 . All of these amino acid replacements reduced the cleavage efficiency (Fig. 2). Asp substitution reduced the rate of cleavage to about 20% of the rate of WT. Leu and Lys substitutions reduced the rate of cleavage to about 40% of the rate of WT, whereas Ala substitution had the least effect on cleavage. These studies therefore show that the cleavage site of CSF-1 256 is partially sequence-specific.
Steric Hindrance of Protease Accessibility Reduces Cleavage Efficiency-Deletion of region 150 -156 in CSF-1 256 severely impaired cleavage (Figs. 2 and 3). This deletion is located N-terminal to the cleavage site as well as the essential PQLQE region. One possible explanation for the marked decrease in cleavage when residues 150 -156 are deleted is that this deletion may bring the CSF-1 growth factor domain closer to the plasma membrane, which interferes with the protease accessibility. To test this, we analyzed the effects of reducing the size of the CSF-1 256 globular structure on the cleavage. CSF-1 256 is a heavily glycosylated protein, which contains only N-linked oligosaccharides with terminal sialic acids (Rettenmier and Roussel, 1988). The canonical sugar attachment sites are Asn 122 and Asn 140 , which are located near the cleavage site (Fig. 1A). Therefore, tunicamycin, an inhibitor of asparagine (N)-linked glycosylation (Kuo and Lampen, 1974) was used to reduce the size of the CSF-1 256 globular structure by removing the N-linked oligosaccharides. The ectodomain cleavage of WT CSF-1 256 with tunicamycin treatment is similar to that without tunicamycin treatment. In the presence of tunicamycin, WT CSF-1 256 was synthesized as 46-kDa homodimeric polypeptide, and its two subunits were cleaved slowly into 41-kDa membrane-bound heterodimeric intermediate and 37-kDa soluble CSF-1 (Fig. 4). The cleavage was activated by PMA (Fig. 4). Similarly, the cleavage rate of CSF-1 256 -Pro161Ser was not changed by tunicamycin treatment (Fig. 4). Furthermore, CSF-1 256 -⌬161-165 was still not cleaved in the presence of tunicamycin, which further demonstrates the essential role of the PQLQE region in the cleavage (Fig. 4). In marked contrast, the cleavage rate of CSF-1 256 -⌬150 -156 was increased from 20% to more than 90% of the rate of WT by tunicamycin treatment (Fig. 4). Likewise, the cleavage rate of CSF-1 256 -⌬156 -160 was increased by tunicamycin treatment (data not shown). These results strongly suggest that the N-linked oligosaccharides in CSF-1 mutants with a shorter juxtamembrane region sterically interfere with the accessibility of the protease(s), therefore reducing the cleavage rate.
Effect of Calcium Ionophores on CSF-1 256 Cleavage and Subcellular Location of the Cleavage System-Calcium ionophores have been shown to activate the cleavage of certain cell surface proteins such as pro-TGF␣ (Pandiella and Massagué, 1991). To test their role in CSF-1 256 cleavage, NIH 3T3 cells expressing CSF-1 256 were metabolically labeled and chased with medium containing calcium ionophores. In contrast to PMA, the calcium ionophores A23187 (Fig. 5) and ionomycin (500 nM, data not shown) did not stimulate the cleavage of CSF-1 256 .
Cell surface radioiodinated CSF-1 256 can be cleaved into soluble CSF-1 (Fig. 1), which demonstrates that cell surface CSF-1 256 can serve as the immediate precursor for cleavage. This raised the question of whether or not the cleavage occurs at or near the cell surface or in the lysosomal compartment, which is the cellular site of general membrane protein degradation. Cell surface detection of the heterodimeric intermediate of 56 kDa that contains one proteolyzed polypeptide chain (22 kDa) assembled through disulfide bonds to an intact subunit (34 kDa) (Fig. 1) favors the first explanation. To further test this, the lysosomotropic agent chloroquine (Kornfeld and Mellman, 1989) was used. Chloroquine did not prevent either constitutive or PMA-induced cleavage of CSF-1 256 (Fig. 5). These studies suggest that the lysosomal compartment is not involved in the cleavage process.
We also asked whether CSF-1 256 is cleaved before it is transported to the cell surface. To answer this question, we used brefeldin A, an agent that blocks Golgi functions by inducing a resorption of the Golgi apparatus into the endoplasmic reticulum, and fusion of the trans-Golgi network with the endosomal system (Klausner et al., 1992). Normal CSF-1 256 maturation was blocked under brefeldin A treatment, and an underglycosylated CSF-1 256 with a smaller molecular mass (62 kDa) was detected (Fig. 6A). Both constitutive and PMA-induced cleavage of CSF-1 256 were blocked if brefeldin A was present from the start of labeling (Fig. 6A). In contrast, brefeldin A did not prevent the cleavage if CSF-1 256 had been chased to the cell surface (Fig. 6B). Resistance of CSF-1 256 to ectodomain cleavage was not due to underglycosylation of this protein by brefeldin A treatment, since nonglycosylated CSF-1 256 can be cleaved (Fig. 4). Thus, brefeldin A prevents the cleavage by blocking the transport of CSF-1 256 to the cell surface, suggesting that the cleavage occurs in subcellular sites post Golgi compartment. Collectively, these results indicate that the cleavage system operates at or near the plasma membrane and outside the Golgi or lysosomal compartments. DISCUSSION We have used CSF-1 256 as a model system for the ectodomain cleavage of cell surface transmembrane proteins. Previous work has shown that CSF-1 256 is expressed on the cell surface, where it contacts and activates CSF-1 receptors on adjacent cells Stein et al., 1990). The growth factor domain is released very slowly by proteolytic cleavage, and this process is activated by phorbol ester Stein and Rettenmier, 1991). The present results identify and characterize the structural determinants required for the release of the CSF-1 256 extracellular growth factor domain and demonstrate that the proteolytic cleavage system operates at or very near the cell surface.
The PQLQE region (residues 161-165) is the essential determinant of the cleavage, and the region 150 -160 encompassing the cleavage site contributes to the cleavage efficiency. The coding regions for all of the three CSF-1 cDNA forms isolated so far share a common amino-terminal domain of 149 amino acids that contains the entire CSF-1 cytokine module, forming a globular structure (Bazan, 1991;Pandit et al., 1992). In addition, CSF-1 256 is cleaved at or near Ser 158 (Halenbeck et al., 1988). Furthermore, native conformation including dimer for- FIG. 4. Effect of steric hindrance of protease accessibility on the cleavage. Cells were metabolically radiolabeled for 20 min and chased for 1 h for maturation in the presence of tunicamycin. Then the cells were chased in medium containing tunicamycin with (ϩ) or without (Ϫ) 0.5 M PMA for the indicated intervals. Cell lysates and harvested culture medium were immunoprecipitated and separated by SDS-PAGE without disulfide reduction. The apparent molecular masses of unglycosylated CSF-1 molecules are indicated in kDa on the left. In addition to the indicated CSF-1 bands, several nonspecific bands such as the bands at the top of the gel and at about 43 kDa were detected in the cell lysates. These nonspecific bands were also detected in precipitates prepared with the control myeloma protein (data not shown).
FIG. 5. Effect of chloroquine or A23187 on CSF-1 256 cleavage. Cells were metabolically labeled for 20 min and allowed to mature for 1 h. The label was then chased for 2 h in the absence (Ϫ) or presence(ϩ) of chloroquine (100 M). Where indicated, PMA (0.5 M) or A23187 (1 M) was added 1 h before the end of chase. Then the cells were lysed, and the medium was collected for immunoprecipitation. The products were resolved by SDS-PAGE without disulfide reduction and detected by fluorography. The apparent molecular masses of soluble CSF-1 molecules are indicated on the left. Similar amounts of 44-kDa CSF-1 in control medium, medium with A23187, and medium with only chloroquine were detected by fluorography with longer exposure time (not shown). mation of CSF-1 256 growth factor domain is not required for the cleavage. 2 Therefore, we hypothesize that the juxtamembrane region from Gly 150 to Glu 165 might play an important role in the cleavage. To test this hypothesis a series of mutants with systematic structural alterations were constructed: first, deletions to identify the regions required for cleavage; second, substitutions of amino acids with different side chain size and/or charge; and third, amino acid substitutions that might change the local secondary structure. These substitution mutants were constructed to define the amino acid specificity. We show that deletions of PQLQE region 161-165 or region 159 -165 abolish the cleavage, whereas deletions of region 150 -156 or 156 -160 only reduce the rate of the cleavage. Further deletions and substitutions show that the length of this region is critical and that the amino acid requirement in this region is partially sequence-specific. It should be noted that substitution of residue Gln 162 or Leu 163 with proline had no inhibitory effect on the cleavage. Since proline places a higher constraint on the polypeptide backbone than any other amino acid and disrupts the regular secondary structure features of proteins (Yaron and Naider, 1993), these findings suggest that distortion of the native secondary structure of CSF-1 256 in this region does not interfere with the ectodomain cleavage. We propose that the PQLQE region may serve as a binding site for the unidentified protease(s) to CSF-1 256 prior to the cleavage process (Fig. 7).
Our data show that the deletion of region 156 -160 encompassing the cleavage site as well as the substitutions of Ser 158 -160 with different amino acids reduces the rate of the cleavage but does not abolish it completely. This suggests that this region is not essential but contributes to the cleavage efficiency and that there is partial sequence specificity required for cleavage. The fact that the cleavage is only partially dependent on the amino acid side chain identity in the region 156 -165 containing the cleavage site and putative enzyme binding site may reflect the involvement of multiple proteases with distinct sequence specificities. Alternatively, cleavage may be caused by a single protease with a relaxed amino acid sequence specificity.
Experiments utilizing glycosylation inhibitors coupled with experiments using mutants with altered length of the juxtamembrane region suggest that steric hindrance of protease accessibility interferes with the cleavage of certain CSF-1 mutants. This finding may provide an explanation for the mecha-nism by which the extracellular domains of certain cell surface proteins such as low density lipoprotein receptor are rapidly released from the cell surface by the proteolytic system in mutant Chinese hamster ovary cells with a defect in protein O-glycosylation (Kozarsky et al., 1988;Reddy et al., 1989). By inference from our findings, it is possible that the stalk-associated O-linked oligosaccharides in these cell surface proteins do not directly protect them from the ectodomain cleavage but interfere with the cleavage by steric hindrance of protease accessibility.
Existence of critical juxtamembrane determinants might be a general mechanism for cell surface transmembrane protein cleavage. Mutations in the cleavage site or its surrounding region inhibit or abolish the cleavage of certain cell surface proteins such as TGF␣ (Wong et al., 1989), tumor necrosis factor ␣ (TNF␣) (Perez et al., 1990;Decoster et al., 1995), and TNF receptor (Gullberg et al., 1992;Brakebusch et al., 1994). Deletion and substitution mutants in the juxtamembrane region of ␤-APP, including a deletion of amino acids 5-24 (numbered from the start of A␤) within A␤ as well as a deletion of amino acids 11-28, still allow for efficient cleavage (Sisodia, 1992;Sahasrabudhe et al., 1992;Maruyama et al., 1991). By inference from our results, the explanation may be that ␤-APP has a very long exposed juxtamembrane stalk region. After 18 or 20 amino acid deletions in this region, the mutant ␤-APPs may still have the appropriate length of the exposed stalk region, which contains the enzyme binding site and cleavage site and allows for the steric accessibility of the protease as well. Consistent with our assumption, it has been documented that more extensive deletions of juxtamembrane region in a modified ␤-APP (APP770⌬35) abolish the cleavage (Sisodia et al., 1990).
Significant differences might exist between some components of the proteolytic system responsible for the cleavage of various groups of proteins. The cytoplasmic C-terminal valine of pro-TGF␣ is essential for the cleavage , but the cytoplasmic domains of cell surface CSF-1 256 , ␤-APP (Hung and Selkoe, 1994), and TNF receptor (Brakebusch et al., 1992) are not. In addition, calcium ionophores activate the cleavage of pro-TGF-␣ (Pandiella and Massagué, 1991), KL-1, and KL-2  but not CSF-1 256 and ACE (Ramchandran et al., 1994). Finally, protease inhibitors blocking the cleavage of pro-TGF␣, KL-1, and KL-2 (Pandiella et al., 1992) do not inhibit the cleavage of either CSF-1 256 (data not shown) or ACE (Ramchandran et al., 1994). Recently, it has been reported that TNF␣ protease inhibitor does not affect the proteolytic processing of CSF-1 554 while abolishing the ectodomain cleavage of the interleukin-6 receptor and the p60 TNF receptor (Mü llberg et al., 1995). It would have been FIG. 6. Effect of brefeldin A on the biosynthesis and cleavage of CSF-1 256 . A, cells were preincubated in medium with brefeldin A (10 g/ml) and then metabolically labeled for 20 min and chased for 1 h for maturation. The medium was replaced with fresh medium to continue the chase for 2 h. Where indicated, PMA (0.5 M) was added 1 h before the end of chase. Brefeldin A was included in both the labeling and chase medium. B, cells were metabolically labeled for 20 min and chased for 1 h in the absence of brefeldin A (10 g/ml). The chase continued in fresh medium containing brefeldin A in the absence or presence of 0.5 M PMA. 44-kDa CSF-1 in medium without PMA was detected by fluorography with longer exposure time (not shown). more informative to test the effect of TNF␣ protease inhibitor on the cleavage of CSF-1 256 , since it is CSF-1 256 , not CSF-1 554 , that belongs to the family of cell surface proteins undergoing ectodomain cleavage (Rettenmier and Roussel, 1988;Stein and Rettenmier, 1991).
The results in this study strongly suggest that the extracellular growth factor domain of CSF-1 256 is cleaved and released into the extracellular compartment by a membrane-associated proteolytic enzyme(s). First, a mutant CSF-1 256 precursor lacking a functional membrane-anchoring domain is released intact without being cleaved, although it still contains the essential cleavage determinant PQLQE, the authentic cleavage site, and the appropriate length of juxtamembrane stalk region. Second, direct phosphorylation of the cytoplasmic domain of CSF-1 256 by a PMA-activated protein kinase C appears not to be the mechanism for activation of the cleavage, since deletion of the cytoplasmic domain does not interfere with the cleavage. Therefore, it is possible that the protease(s) involved is a transmembrane protein(s), the cytoplasmic domain of which directly or indirectly interacts with protein kinase C.
We have synthesized the results of this study in the proposed model for CSF-1 256 ectodomain cleavage (Fig. 7). Homodimeric CSF-1 256 of 68 kDa is expressed on the cell surface with the 16-amino acid juxtamembrane region (residues 150 -165) being long enough to allow for the accessibility of the putative membrane-associated proteolytic enzyme(s) to the CSF-1 256 substrate. One domain of the proteolytic enzyme(s) recognizes and binds to the essential determinant PQLQE (residues 161-165). This binding leads to certain conformational changes of the enzyme(s) and brings the cleaving domain of the enzyme(s) close to the cleavage site (residues 157-159) suitable for cleavage and located at a specified distance of approximately 7-9 amino acids from the plasma membrane. Subsequently, the 44-kDa extracellular growth factor domain of CSF-1 256 is cleaved and released into the extracellular compartment. Since the two identical subunits are not always cleaved at the same time, a membrane-bound 56-kDa heterodimeric intermediate is often detectable on the cell surface.
Mutations that change the essential binding determinant of the protease(s), mutations that change the amino acid sequence of the cleavage site, or mutations that sterically interfere with the enzyme accessibility can reduce or abolish the cleavage.