Proteomic Approaches to the Analysis of Early Events in Colony-stimulating Factor-1 Signal Transduction*

The exposure of cells to growth factors leads to the rapid tyrosine phosphorylation of proteins that play critical roles in initiating cellular responses. These proteins are associated with other nontyrosine-phosphorylated proteins. Together, they represent less than 0.02% of the total cellular protein. To study their functions in growth factor signaling it is necessary to establish their identity, post-translational modifications, and interactions. We have focused on the characterization of this group of proteins during the early response of macrophages to the macrophage growth factor, colony-stimulating factor-1 (CSF-1). We review here the development of approaches to analysis of the rapid CSF-1-induced changes in the CSF-1 receptor tyrosine kinase and phosphotyrosyl signaling complexes. Recent advances in mass spectrometry technology are greatly facilitating the characterization of such complexes. These methods strongly support and enhance genetic approaches that are being used to analyze the function of individual signaling components and pathways.

Colony-stimulating factor-1 (CSF-1), 1 also known as macrophage-CSF, is the primary regulator of the survival, proliferation, and differentiation of mononuclear phagocytes (1)(2)(3)(4)(5). These cells include monoblasts, promonocytes, monocytes, tissue macrophages, and osteoclasts (1)(2)(3)(4)(5). CSF-1 also regulates the function of cells of the trophoblast and other cells in the female reproductive tract (6 -8). CSF-1 is made by a wide variety of cell types (4) and can be secreted as a glycoprotein or proteoglycan (9,10), both of which are found in the circulation (10), or expressed as a biologically active, membrane-spanning, cell-surface glycoprotein (11,12). All of these forms are homodimeric and possess the amino-terminal 149 amino acids that are sufficient for biological activity in vitro. Some target cells are regulated by circulating growth factor, while others are regulated by locally presented growth factor (5,13), presumably the cell surface or sequested proteoglycan forms.
Because undifferentiated precursor cells are rare, CSF-1 signal transduction has mainly been studied in macrophages. The response of macrophages to CSF-1 is pleiotropic, including the morphological changes of membrane ruffling, cell spreading, actin redistribution, and vacuolation. These are followed by increases in cell motility, the rate of protein synthesis, gene expression, and the entry of cells into S phase (14 -17) (Fig. 1).
The CSF-1 receptor (CSF-1R), which mediates all known effects of CSF-1 (18 -21), is a tyrosine kinase encoded by the c-fms proto-oncogene (22,23). A member of the plateletderived growth factor receptor family, it possesses a highly glycosylated extracellular domain of five immunoglobulin-like domains, a transmembrane domain, and an intracellular tyrosine kinase domain that is interrupted by a 73-aa interkinase domain (24 -26). CSF-1 binding to the CSF-1R on cells results in tyrosine phosphorylation of the receptor and many other proteins (27,28). Analysis of CSF-1R tyrosine phosphorylation in transfected fibroblasts or myeloid cells lacking endogenous receptor has shown that, of the 20 mouse CSF-1R cytoplasmic domain tyrosine residues, 8 are phosphorylated in the CSF-1 response (29 -32). Several of these phosphotyrosines are included in binding sites for Src homology 2 (SH2) or phosphotyrosine (PY) binding domain signaling proteins (Fig. 2).
Studies of the interaction of 125 I-CSF-1 with CSF-1-starved bone marrow-derived macrophages at 37°C indicate that they express ϳ5 ϫ 10 4 surface receptors per cell. However, they contain an additional ϳ10 5 mature receptors that can appear at the cell surface within 10 min of addition of CSF-1 (20). CSF-1Rs behave as a single class of high affinity CSF-1 binding sites (K d ϳ 4 ϫ 10 Ϫ10 M at 37°C). Internalization and dissociation of CSF-1 are competing first order processes, and internalization (t 1/2 ϭ 1.6 min) is six times more frequent than dissociation. The internalized CSF-1/CSF-1R complex is destroyed intralysosomally, and there is no receptor recycling (20,42). Thus, addition of CSF-1 leads to down-regulation of cell surface CSF-1Rs due to CSF-1R internalization and destruction. However, cell surface CSF-1Rs are maintained at a reduced level initially by replenishment from the mature CSF-1R pool, then subsequently by de novo synthesis (20). At 2°C, there is no measurable dissociation or internalization of CSF-1 (K d Յ 2 ϫ 10 Ϫ13 M) (20).
In this review, the biochemical and proteomic approaches used by our laboratory and other investigators to characterize the early events in CSF-1 signal transduction are described.

APPROACHES FOR STUDYING THE EARLY EVENTS IN CSF-1R SIGNAL TRANSDUCTION
Overall Strategy and Rationale-Incubation of CSF-1starved macrophages with CSF-1 leads to the rapid tyrosine phosphorylation of the CSF-1R and other cellular proteins. The intrinsic receptor tyrosine kinase activity is essential for mitogenic activity (43). We therefore focused on the changes that take place in the CSF-1R and in cellular protein tyrosine phosphorylation within the first 2 min of addition of CSF-1 to macrophages that have been starved of CSF-1 for 16 -24 h. The use of such "CSF-1R-up-regulated" macrophages is necessary to increase the signal-to-noise ratios for CSF-1 stimulation of tyrosine phosphorylation and of signaling pathways, because of "down-regulation" of cell surface CSF-1R expression on cells incubated with CSF-1 (20) (Fig. 1). Despite the marked "down-regulation" of the cell-surface CSF-1Rs in cells cultured with CSF-1, signaling via the receptor in its down-regulated state is required continuously for entry of cells into S phase ϳ20 h later (16) (Fig. 1). Thus, it is likely that many of the signaling pathways activated in the initial 2 min of CSF-1 addition to cells are also activated in cells with "downregulated" CSF-1Rs that are responding to CSF-1.
Development of CSF-1-responsive Macrophage Cell Lines-Because of our concern that CSF-1R-initiated signal transduction would be cell context-dependent, we have restricted our analysis to the physiologically relevant and CSF-1R-expressing macrophage. Early analysis of the interaction of CSF-1 with the CSF-1R (20) and the CSF-1-induced changes in protein turnover (15), entry in to S phase (16), survival (14), proliferation, and cell morphology (14,17) were carried out in primary cultures of highly purified bone marrowderived macrophages (BMM) (14,44). Disadvantages of BMM are the time taken to prepare them and their limited prolifer- FIG. 2. Domain organization of the mouse CSF-1R, indicating the cytoplasmic domain tyrosine phosphorylation sites and signaling molecules known to interact with them. The phosphorylated juxtamembrane Y559 associates with Src family members through their SH2 domains (33). The Y544 site is tyrosine-phosphorylated in the v-fms oncoprotein and interacts with an unidentified 55-kDa protein (30). Tyr 697, Tyr 706, and Tyr 721 in the kinase insert domain may function to bring substrates to the kinase domain (34,35). The phosphorylated Y697 site binds the Grb2 SH2 domain (36) and the SH3/SH2 adapter protein, Mona (37). The phosphorylated Y721 site binds the SH2 domain of p85 subunit of the PI-3 kinase (PI3-K) (35,38) as well as the N-SH2 domain of phospholipase C␥2 (PLC␥2) (39). No protein has yet been reported to bind the phosphorylated Y706 site. Y807 is situated within the activation loop or "lip" of the main kinase domain, and there are conflicting results concerning the consequences of tyrosine phosphorylation at this site. While phosphorylation of Y809 (human) or Y807 (mouse) has been shown to be important for the mitogenic response in fibroblasts (38,40), this does not appear to be the case in FDC-P1 myeloid cells (41), where the Y807F mutant is able to proliferate in response to CSF-1 but unable to differentiate. Phosphorylated Y921, at the boundary between the C-terminal tail and the main kinase domain, serves as a second Grb2 binding site (29,31), and phosphorylated Y974 is the Cbl binding site (32). Cbl negatively regulates CSF-1-stimulated macrophage proliferation (42). ative potential. We therefore developed a cloned, splenic macrophage-derived cell line, BAC1.2F5 (45). BAC1.2F5 macrophages require CSF-1 for survival and proliferation. They behave like BMM in culture and are easily obtained in the large numbers required for proteomic studies.
Stimulation of Cells at 4°C-At 4°C, CSF-1-stimulated protein tyrosine phosphorylation gradually increases to a maximum at 120 min and remains constant for the next 2 h (27, 46 -48). The M r distribution of proteins tyrosine-phosphorylated by CSF-1 stimulation at 4°C for 120 min assessed by SDS-PAGE is identical to the M r distribution observed following stimulation at 37°C for 1 min. However, the intensity of tyrosine phosphorylation is significantly higher at 4°C, because protein tyrosine phosphatase (PTP) activity is differentially reduced. Importantly, many subsequent events, observed after the first minute of stimulation at 37°C, that might be difficult to separate from earlier events at this temperature, are eliminated. For example, Raf activation and phosphorylation (49) do not occur at 4°C, despite the fact that there is movement of cytoplasmic protein complexes to and from the receptor, e.g. Grb2-Sos (51) and Grb2-Cbl (50). Stimulation of cells at 4°C enabled us to establish the order of cellular events during the early response and to more conveniently carry out the large-scale cell stimulations required for the proteomic analyses.
Kinetics of CSF-1-induced Changes in the CSF-1R in Macrophages-Our current understanding of the early CSF-1-induced changes in the CSF-1R in macrophages are summarized in Fig. 3. Two key approaches in our kinetic analysis of these changes were the analysis of the reactions at 4°C, expanding the events of the first 1-2 min to 120 min, thereby permitting clear resolution of their sequence (27,47) and the stimulation of cells with very high concentrations of CSF-1 to rapidly saturate receptor sites (1 min at 4°C) so that CSF-1 binding was not rate-limiting for subsequent events (27,46). In addition, the following methods were important in this kinetic analysis: 1. a technique for the quantitative recovery of cell-surface CSF-1Rs from unstimulated and stimulated cells by reacting the cells with anti-CSF-1R antibody prior to cell lysis and processing (46,48); Early CSF-1-induced changes in the CSF-1R and some signaling proteins in macrophages at 37°C and 4°C. Prior to CSF-1 addition, CSF-1Rs are clustered or are undergoing a rapid dimer-monomer transition (46). CSF-1 binding initially leads to the rapid noncovalent dimerization and tyrosine phosphorylation of the CSF-1R (step 1). This is followed by the association of Grb2/Sos with the receptor (step 2), the tyrosine phosphorylation of other signaling molecules (step 3), including the Cbl ubiquitin ligase, Shc, and the regulatory subunit of PI-3 kinase (PI3K) (p85), the dissociation of Sos (step 4), and the association of Cbl, Shc, and p85 with the CSF-1R and Grb2 in a complex (46 -48, 50 -53) (step 5). This complex is independent of the more transient CSF-1R/Sos/Grb2 complexes formed earlier (51). The period of active nonreceptor protein tyrosine phosphorylation (step 3) is followed by the extracellular disulfide linkage of monomers within the CSF-1R dimer that permits a second wave of CSF-1R tyrosine phosphorylation, increased CSF-1R serine phosphorylation, coordinate ubiquitination of the CSF-1R and membrane-associated Cbl (step 5), and internalization of the CSF-1/CSF-1R complex (42, 46, 48, 50, 51) (step 6). Tyrosine phosphorylation and ubiquitination of the cell-surface CSF-1Rs are stoichiometric (46,47,51). Once internalized (step 6), both CSF-1 (20) and the CSF-1R (42), are degraded intralysosomally, and it is assumed that the endocytosed receptor-ligand complex enters a multivesicular body (MVB) prior to entry into the lysosomal system. In contrast, the ubiquitinated Cbl ubiquitin ligase is not degraded but recovered in deubiquitinated form in the cytosol 3-10 min after stimulation (42,50,51). Time scales for these events as they occur at 37°C and 4°C are shown at the top of the figure. The number and position of the phosphorylation sites are not those of the actual sites. Ub, ubiquitin; X, proteins tyrosine phosphorylated in the CSF-1 response; pY, phosphotyrosine; pS, phosphoserine.

dual-labeling of cells with [ 3 H]-leucine and 32 P i in order
to obtain the relative specific activity of phosphate incorporated into the CSF-1R tyrosine and serine residues (48); 3. cross-linking immediately after cell lysis (highly efficient), in conjunction with cross-linking prior to lysis (not very efficient), to establish the multimeric state of the receptor (46); 4. the use of low concentration gradient (3-10% acrylamide) SDS-PAGE in order to observe the higher molecular mass dimeric and ubiquitinated forms of the CSF-1R (46); and 5. shifting temperature from 4°C to 37°C to study the internalization of CSF-1-CSF-1R complexes and subsequent events (46).

Summary-
The results demonstrate that CSF-1-induced stablilization of the noncovalent dimerization of the CSF-1R activates its kinase and is associated with a first wave of CSF-1R tyrosine phosphorylation, probably due to autophosphorylation in trans of one cytoplasmic domain of the dimer by another (54). Covalent dimerization by extracellular domain disulfide bonding is important for subsequent CSF-1R modifications. These include a second wave of tyrosine phosphorylation of the cell-surface receptor, receptor serine phosphorylation, ubiquitination, and internalization of the receptor-ligand complex. Finally, the receptor-ligand complex undergoes intralysosomal degradation. The determination of the sequence of phosphorylations at the tyrosine and serine phosphorylation sites and the effects of site-directed mutagenesis of these sites and the cysteines involved in the disulfide bonding are now being carried out, because defined steps in cell-surface receptor behavior that they might regulate are known. This kind of kinetic analysis provides a solid base from which to look for causal relationships between alterations in the receptor structure and its function.

IDENTIFICATION AND CHARACTERIZATION OF THE CELLULAR PROTEINS THAT ARE TYROSINE PHOSPHORYLATED IN RESPONSE TO CSF-1
Rationale and Strategy-Because CSF-1R tyrosine kinase activity is required for the mitogenic response, focusing on the identification and characterization of proteins that are tyrosine-phosphorylated in the CSF-1 response and the other proteins associated with them is a reasonable approach. At 37°C, the CSF-1-stimulated macrophage protein tyrosine phosphorylation increases to the maximum at 1 min and subsequently declines to almost basal level by 30 min (27) (Fig. 1). The total tyrosine-phosphorylated protein quantitatively recovered by anti-PY antibody affinity purification from cells at the peak of the tyrosine phosphorylation response plus the associated nontyrosine-phosphorylated protein is less than 0.02% of the total cellular protein (55). For proteomic studies, we therefore developed CSF-1 stimulation conditions that maximized the yield of phosphorylated pro-teins and that facilitated the processing of stimulated cells on a large scale.
Conditions for Stimulation-For proteomic studies, it is desirable to optimize the conditions of growth factor stimulation so as to maximize the yield of tyrosine-phosphorylated proteins. Preincubation of cells with iodoacetic acid (IAA) increased by severalfold the CSF-1-induced tyrosine phosphorylation of cellular proteins (46). IAA could increase the level of tyrosine phosphorylation by inhibiting PTP activity, because IAA can inactivate PTPs by carboxymethylating the free sulfhydryl residue of the active site cysteine of the PTP catalytic domain. However, IAA can also carboxymethylate the free sulfhydryl groups of the extracellular domain cysteine residues of the CSF-1R, and this has been shown to inhibit CSF-1-induced CSF-1R covalent dimerization, ubiquitination, and internalization and increase cellular tyrosine phosphorylation (46). The tyrosine phosphorylation-enhancing effect of IAA in the absence of CSF-1, in contrast to other PTP inhibitors, e.g. pervanadate, is minimal, preserving the effect of increased tyrosine phosphorylation in the presence of CSF-1 that is required to establish regulation by growth factor. Therefore, to increase the yield of tyrosine-phosphorylated proteins for proteomic studies, we incubate with and without CSF-1 in the presence of IAA at 4°C (55,56).
Subcellular Fractionation-Subcellular fractionation of CSF-1-stimulated cells at 4°C revealed that the majority of tyrosine-phosphorylated proteins are in the cytosolic fraction (27) (Fig. 4). The CSF-1R is the major tyrosine-phosphorylated protein in the detergent soluble membrane fraction, and many of the other proteins in this fraction appear to be also present in the cytosolic fraction. Hence, we initially focused on the purification and identification of the soluble anti-phosphotyrosine reactive proteins (␣PYRPs).
Optimization of the Recovery of PY Proteins-Initially, our major losses of PY protein recovery were due to dephosphorylation by PTPs and absorption to surfaces. Prior to dena- turation, to prevent dephosphorylation during cell homogenization, we observed that 5 M Zn 2ϩ was an effective inhibitor of many PTPs. Our conditions also excluded EDTA from the commonly used buffer system, which contains sodium fluoride and sodium pyrophosphate as nonspecific phosphatase inhibitors, sodium orthovanadate, and a protease inhibitor mixture (47,48). It was subsequently shown that EDTA counteracts the action of the PTP inhibitor, orthovanadate, by chelating the vanadate ion (57). This buffer was effective because we showed that ϳ100% of the cell-surface CSF-1Rs from CSF-1-stimulated macrophages were tyrosine-phosphorylated and recoverable by anti-phosphotyrosine affinity chromatography (47).
The ␣PYRPs tend to avidly bind to plastic or glass surfaces. A nondenaturing neutral detergent (0.5-1% Nonidet P-40 or Triton X-100) is necessary to prevent excessive loss during purification. During sample concentration by ultrafiltration, a detergent with high critical micellar concentration and small micell size, e.g. octylglucoside, should be added to prevent accumulation of the detergent in the concentrate (55). With these improvements and stimulation in the presence of IAA, the yield of protein from the anti-PY affinity purification was increased by more than 10-fold.
Purification of the Tyrosine-phosphorylated and Associated Proteins-A scheme for purification of the tyrosine-phosphorylated and associated proteins is shown in Fig. 5. Such purifications are carried out in parallel with subcellular fractions obtained from macrophages incubated with and without CSF-1. Individual steps in each purification are followed by SDS-PAGE with anti-PY Western blotting to identify those proteins exhibiting increased tyrosine phosphorylation in response to CSF-1 and silver staining to visualize all protein bands (e.g. see Fig. 9).
The first step, nondenaturing anti-PY affinity chromatography (55,56), was used to quantitatively recover the PY proteins together with nontyrosine-phosphorylated proteins complexed with them in the anti-phosphotyrosine reactive fraction (␣PYRF). The ␣PYRPs were then reduced with 2-mercaptoethanol, denatured, and separated according to their molecular weights by size exclusion chromatography (SEC) on Superose 6 (SEC-S6) (Amersham Bioscience) in 6 M guanidine-HCl under reducing conditions (Fig. 6a). Selected SEC-S6 fractions were then purified by microbore C 4 reversephase high-performance liquid chromatography (RP-HPLC). The highly purified proteins from the C 4 separation were digested with either trypsin or endoprotease-RC. The resulting peptides were separated by microbore C 8 RP-HPLC and sequenced by gas-phase Edman degradation chemistry (55,56).
We have also used mass spectrometry (MS) for protein identification, where the protein fractions from the SEC-S6 step are subjected to MS analysis. Improved resolution over SEC-S6 can be obtained by using SDS-PAGE whole-gel elution (Fig. 6b). In this case, because SDS interferes with subsequent HPLC separation and MS analysis, the gel must be washed extensively and eluted into buffer containing 0.4% of octylglucoside (or other dialyzable neutral detergent). A purification scheme used for MS identification is shown in Fig. 7.
Identification of Proteins in the ␣PYRF-Advances in MS technology have enabled the identification of proteins without having to individually purify each one, as is necessary for Edman sequencing. Although the popular approach of in-gel digestion for matrix-assisted laser desorption ionization-time of flight (MALDI-TOF) MS or liquid chromatography-electrospray ionization MS analysis is effective in identifying relatively abundant proteins, this approach did not produce consistent results on proteins present in small amounts in our preparations. Our multi-step all-liquid-phase sample preparation approach coupled with MS peptide fingerprinting and/or tandem MS sequencing (Fig. 7) has been successfully used for the identification of less abundant proteins in our fractions and for the determination of their post-translational modification sites. Table I shows examples of the proteins we have identified in the ␣PYRF of the cytosolic fraction, primarily by Edman sequencing. These proteins in the ␣PYRF include known tyrosine-phosphorylated signaling proteins, signaling proteins associated with tyrosine-phosphorylated proteins, cytoskeletal proteins, and proteins with no obvious link to either growth factor signaling or cytoskeletal reorganization. The presence of the majority of the proteins shown in Table I has been confirmed by SDS-PAGE and Western blotting. They represent one-fifth of the proteins we have so far identified by MS in the cytosolic ␣PYRF. 2 Analysis of Post-translational Modifications by MS-CSF-1induced covalent modificatons play a critical role in CSF-1R signal transduction. The CSF-1R exhibits increases in tyrosine phosphorylation, serine phosphorylation, and ubiquitination as well as CSF-1-induced disulfide binding (42, 46 -48). Several of these modifications have important roles in mediating CSF-1-induced effects and CSF-1R function. Much of the information about the specific tyrosine phosphorylation sites involved (Fig. 2) came from mutating specific amino acids in the receptor. Sites of other post-translational modifications, such as serine/threonine phosphorylation or lysine ubiquitination, are not as easily predicted because they are often not found in consensus sequences. We have used MS to successfully determine post-translational modifications of the CSF-1R and the downstream adapter protein, macrophage actin-associated and tyrosine-phosphorylated protein (MAYP) (58), also known as PSTPIP2 (59). 2 Analysis of in-gel tryptic digests from gel slices containing 2 g (ϳ20 pmol) of CSF-1R purified from CSF-1-stimulated BAC1.2F5 macrophages, using MALDI-TOF, did not yield enough material to determine the phosphorylation, ubiquitination, and disulfide bonding sites. The final yield of peptides from the tryptic digest of gel slices was low, and the coverage of the CSF-1R sequence was unacceptably low (less than 20%). An alternative strategy was then developed for sample preparation to improve the peptide yield and coverage in MS ( Fig. 8). This strategy reduces copurification of receptor-associated proteins by denaturing the membrane proteins in SDS and performing CSF-1R affinity purification in a dissociating detergent mixture (0.1% SDS and 1% Nonidet P-40), eliminating the need for subsequent SDS-PAGE. The purified receptor is subjected to tryptic digestion, the tryptic peptides are fractionated by RP-HPLC, and the RP-HPLC fractions are subjected to MS analysis. If separation of different modified forms of the receptor is necessary, preparative SDS-PAGE followed by whole-gel elution can be performed before the trypsin digestion. Using this method, we obtained peptides covering more than 35% of the CSF-1R sequence. Only ϳ0.6 pmol of the peptide fraction after RP-HPLC is required to obtain strong, clear peptide mass signals on MALDI-TOF MS analysis. The purified CSF-1R used for the tryptic digestion contains a small amount of SDS. This results in a significant number of the more hydrophobic peptides co-eluting with the detergent at high solvent concentration in the RP-HPLC. This occurred despite the use of a precolumn to remove SDS. Greater than 50% receptor sequence coverage can be achieved if the SDS-containing peptide fractions are pretreated with a detergent removal Zip-tip (Millipore, Bedford, MA) prior to MALDI-TOF MS. Alternatively, SDS and other neutral detergents can be removed by using more effective detergent removal approaches, e.g. cation exchange batch fractionation, prior to the final RP-HPLC step. We now routinely recover phosphopeptides with masses of those containing the eight known CSF-1R tyrosine phosphorylation sites, two serine sites, as well as several ubiquitination sites.
Summary-Stimulation of cells at 4°C in the presence of IAA increases the yield of tyrosine-phosphorylated proteins. Important additional information about these proteins can be obtained by isolating them separately from membrane and cytosolic fractions prepared from the cells. The inclusion of 5 M Zn 2ϩ , the avoidance of EDTA, and the inclusion of detergent (even with the cytosolic fractions) prevented loss of recovery of PY proteins in the initial step of nondenaturing anti-PY affinity chromatography. Further purification of proteins uses highly resolving SDS-PAGE with whole-gel elution. Proteins in the eluted fractions are then directly identified by MS. Approximately 150 proteins, including tyrosine-phosphorylated proteins and the non-PY proteins associated with them, have been identified in the ␣PYRF from CSF-1-stimulated macrophages. Considerable information concerning the post-translational modification of these proteins may be gleaned during their MS identification. Specific approaches to obtain the high coverage needed are detailed for the CSF-1R.

DEMONSTRATION AND ANALYSIS OF SIGNALING PROTEIN COMPLEXES
To investigate the occurrence, size, and composition of cytosolic phosphotyrosyl protein complexes and changes in these complexes with CSF-1 stimulation, the cytosolic ␣PYRFs from CSF-1-stimulated and unstimulated macrophages were separately fractionated by SEC-S6 under nondenaturing conditions (Fig. 9). In contrast to the separation of these same fractions by SEC-S6 under denaturing conditions, where the elution volume of individual proteins was proportional to their mobility on SDS-PAGE (Fig. 6a), the proteins of high electrophoretic mobility were often eluted at a low elution volume or in a broad band, indicating that they were associated with other proteins and/or oligomerized. Interestingly, ␣PYRF proteins like F-actin, EF-1␣, GAPDH, and Vav (Fig. 6c, middle bracket) exhibited a CSF-1-stimulated increase in the SEC-S6 higher M r fractions and a CSF-1-stimulated decrease in SEC-S6 lower M r fractions, indicating that CSF-1 causes the redistribution of these proteins from smaller complexes to larger complexes. In contrast, this phenomenon was not encountered with proteins possessing SH2 or PY binding domains, including Grb2, Shc, SHP-1, the Doks, the STATs, and Cbl (Fig. 6c, upper bracket), which showed a marked CSF-1stimulated increase in all fractions in which they were detected. Consistent with both these observations, separation of the ␣PYRF from 32 P-labeled cells by nondenaturing gradient gel electrophoresis in the first dimension, followed by SDS-PAGE in the second dimension, demonstrated a CSF-1-induced increase in the size of the 32 P-labeled protein complexes (55). Demonstration of these phosphotyrosyl protein complexes requires isolation of the ␣PYRF (0.02% of total cellular protein) so that the complexed proteins are enriched enough to detect. Fractionation of the ␣PYRF also allows identification of the non-PY proteins associated with the PY proteins. Many of these signaling complexes may exist transiently, their composition varying with time during the course of signaling. To characterize the function of each complex, it is necessary to separate them, determine their composition, kinetics of assembly/disassembly, as well as the function, arrangement, and stoichiometry of their individual component proteins.
Composition of Complexes-CSF-1 stimulates protein tyrosine phosphorylation and assembly of many proteins into large complexes. Phalloidin was able to pull down F-actin with many other tyrosine-phosphorylated proteins from the total cytosolic ␣PYRP, indicating that F-actin is a component of complexes in this fraction (55). The CSF-1R has also been shown to form multimeric complexes with signaling proteins including phosphatidylinositol-3 kinase (PI-3 kinase), SHP-1, Grb2, Shc, c-Src, Cbl, as well as other tyrosine-phosphorylated proteins and structural proteins in CSF-1-stimulated BAC1.2F5 macrophages and myeloid cells (29,50,51,60). These complexes are stable and can be separated by ion exchange chromatography. The receptor kinase and PI-3 kinase activities were shown to vary in different complexes (60). The CSF-1R complexes that contain Sos do not contain Cbl (51), suggesting that different CSF-1R complexes are generated to transduce different cellular signals.
Fractionation of Complexes-The stability of complexes containing tyrosine-phosphorylated proteins obtained from cells both prior to and after CSF-1 stimulation allows fractionation by conventional approaches. To isolate intact signaling complexes from the ␣PYRF for analysis of their composition and stoichiometry, we have used a mild competitive elution with 5 mM phenyl phosphate to dislodge protein complexes from the anti-PY solid-phase affinity matrix. SH2 domainphosphotyrosine interactions are not disrupted in 5 mM phenyl phosphate, ensuring the integrity of protein complexes during the elution. Nondenaturing SEC-S6 of the ␣PYRF eluted with 5 mM phenyl phosphate shows high complexity of protein organization in this fraction (Fig. 9). Fractionation of the ␣PYRF (Fig. 9, Orig) by sucrose gradient nondenaturing isoelectric focusing showed the protein complexes to be distributed over a wide pH range (3-9; Fig. 10a). In contrast, when the larger protein complexes eluting in the SEC-S6 void volume fraction (fraction 1) were subject to the same analysis, the fraction 1 proteins were all contained in a narrow pH range (ϳ3.1-4.3; Fig. 10b). Because the majority of proteins present in the ␣PYRF were also present in fraction 1 (compare Fig. 10a Orig. with Fig 10b Orig.), it is likely that the ␣PYRF contains incompletely assembled or partially disassembled versions of the complexes found in fraction 1. This study confirms that the protein complexes in the ␣PYRF can be recovered efficiently and purified.
The protocol we use to isolate protein complexes and the identify their component proteins is outlined in Fig. 11. Following nondissociating elution from ␣PY immunoaffinity columns, the eluate, containing a mixture of complexes, is further fractionated sequentially by nondenaturing methods, e.g. size exclusion chromatography, density gradient isoelectric focusing, and ion-exchange chromatography. This approach precludes artifacts arising from the use of cells overexpressing tagged signaling molecules.
Analysis of Complexes-Characterization of the protein complexes involved in CSF-1 signaling requires identification of the component proteins, their stoichiometric proportions, the direct protein-protein interactions within the complex, the three-dimensional arrangement of the proteins in the complex, and the hierachy of interaction of the component proteins. Methods that can be used to achieve these goals are listed in Table II. Preference is given to methods that require the least quantity of sample and that can be performed quickly, with high resolution, under physiologically relevant conditions. Analyses of complexes by MS or electron microscopy are rapid and highly resolving. However, techniques involving surface plasmon resonance (biosensor), calorimetry, fluorescence resonance energy transfer microscopy, atomic force microscopy, and nuclear magnetic resonance can be carried out in solution and provide data that are likely to be more relevant to conditions in the living cell. Details of these techniques have be described elsewhere (61) and will not be discussed here.
Summary-Pre-formed and CSF-1-induced complexes of PY proteins can be identified in the ␣PYRF of both the cytosolic and membrane fractions. The complexes consist primarily of PY proteins, cytoskeletal components, and signaling molecules. The general effect of CSF-1 stimulation is to increase the size of the complexes by recruiting PY binding proteins and cytoskeletal proteins. The complexes, although they appear transiently in the CSF-1 response, are stable and can be fractionated. A variety of existing techniques can therefore be utilized to determine the composition, stoichiometry, and function of the isolated complexes.

PERSPECTIVE
Our studies delineate three distinct early phases in CSF-1R signal transduction in macrophages. The initial phase is associated with a first wave CSF-1-induced CSF-1R tyrosine phosphorylation (ϳ15 s) that appears to be the active signaling phase for cell proliferation and activation of the Rasmitogen-activated protein kinase pathway. The second phase is associated with CSF-1-induced CSF-1R extracellular domain disulfide bonding, which is permissive for a second wave of tyrosine phosphorylation, increased serine phosphorylation, and ubiquitination of the CSF-1R. This phase appears to involve tyrosine phosphorylation by other tyrosine kinases and prepares the CSF-1/CSF-1R complex for clathrindependent endocytosis and eventual lysosomal destruction. The third phase is characterized by the internalization of the CSF-1/CSF-1R complex.
The tyrosine-phosphorylated CSF-1R forms complexes with a variety of signaling molecules. The bulk of the CSF-1induced tyrosine-phosphorylated proteins are found in the cytosolic fraction. By purification, sequencing, and mass spectrometry, we have identified over 100 signaling, cytoskeletal, and other proteins in the cytosolic ␣PYRF. These proteins can be isolated as complexes. The largest complexes   contain signaling, cytoskeletal, and other proteins, only some of which are tyrosine-phosphorylated. The majority of proteins in the largest complexes are also present in the smaller complexes, suggesting that the ␣PYRF contains fully and partially assembled complexes. Some of the multi-protein components of these complexes are pre-formed prior to CSF-1 stimulation. The challenge remains to define the composition of these complexes as well as the direct proteinprotein interactions between component proteins within them.
Most of our studies on CSF-1R signal transduction have utilized macrophages of the differentiated mouse macrophage cell line, BAC1.2F5. These cells permit somatic cell genetic (62,63) and retroviral infection approaches (64) to be used to obtain stable variant cell lines for signal transduction studies. For CSF-1R structure/function studies, it is desirable to introduce mutant forms of the receptor into receptor-deficient macrophages. A significant problem with macrophages and differentiated macrophage cell lines is that their transfection rate is extremely low. Studies have been carried out with CSF-1R-transfected fibroblasts and myeloid cells or with erythropoietin receptor-CSF-1R hybrids in bone marrow precursor cells (65). While useful, in the former cases the cell context is not appropriate, and in the latter system an inherent disadvantage is that the CSF-1R extracellular domain disulfide bonding has been shown to be required for many events in CSF-1R signaling in macrophages.
We have developed a cloned macrophage cell line, MacCsf1r Ϫ/Ϫ, 3 from CSF-1R Ϫ/Ϫ mouse (21) BMM. MacCsf1r Ϫ/Ϫ cells are maintained by culture in granulocytemacrophage CSF. Retroviral introduction of the wild-type CSF-1R gene into these cells restores the complete phenotype of CSF-1-responsive cell lines derived from wild-type BMM (MacCsf1r ϩ/ϩ cells). MacCsf1r Ϫ/Ϫ cells are therefore suitable for physiologically relevant proteomic studies of CSF-1R structure/function necessitating an analysis of CSF-1R mutations. To confirm results obtained in this system and to study the effects of mutations on differentiation, these mutations should also be studied in retrovirally infected primary BMM from CSF-1R Ϫ/Ϫ mice and by the generation of CSF-1R gene-targeted mouse germline mutations. At present, detailed proteomic studies to identify sites of multiubiquitination, phosphorylation, and other as yet undiscovered post-translational modifications of the wild-type CSF-1R that take place during CSF-1 signaling are needed to identify sites that can be mutated for analysis of their function using the above approaches. Similar genetic approaches to those outlined above for the CSF-1R can also be applied to other signaling components for which targeted mouse mutations have been obtained.
Thus the procedures of CSF-1 stimulation and subcellular and protein fractionation reviewed here, together with the development of macrophage cell lines that allow genetic approaches to structure/function analysis of the CSF-1R in macrophages, can now be coupled with MS approaches to the analysis, assembly, and disassembly of CSF-1R and cytosolic complexes recovered in the ␣PYRF of cells during the early response to CSF-1.