Endogenous Transforming Growth Factor-β Receptor-mediated Smad Signaling Complexes Analyzed by Mass Spectrometry *S

ASmad proteins are the central feature of the transforming growth factor-β (TGF-β) intracellular signaling cascade. They function by carrying signals from the cell surface to the nucleus through the formation of a series of signaling complexes. Changes in Smad proteins and their complexes upon treatment with TGF-β were studied in mink lung epithelial (Mv1Lu) cell cultures. A time course of incubation with TGF-β was carried out to determine the peak of appearance of phosphorylated Smad2. Immobilized monoclonal antibody against Smad2 was then used to isolate the naturally occurring complexes. Three strategies were used to identify changes in proteins partnering with Smad2: separation by one-dimensional SDS-PAGE followed by MALDI peptide mass fingerprinting, cleavable ICAT labeling of the protein mixtures analyzed by LC-MS/MS, and nano-LC followed by MALDI MS TOF/TOF. Smad2 forms complexes with many other polypeptides both in the presence and absence of TGF-β. Some of the classes of proteins identified include: transcription regulators, proteins of the cytoskeletal scaffold and other tethering proteins, motility proteins, proteins involved in transport between the cytoplasm and nucleus, and a group of membrane adaptor proteins. Although some of these have been reported in the literature, most have not been reported previously. This work expands the repertoire of proteins known to participate in the TGF-β signal transduction processes.

Smads are a class of proteins that function as intracellular signaling effectors for the TGF-␤ superfamily of secreted polypeptides. The present picture of TGF-␤ signal transduction process is the following. First, a ligand, TGF-␤, binds to a type II receptor, which recruits and phosphorylates a type I receptor. Second, the type I receptor then associates with a specific receptor-regulated (R-) Smad protein, Smad2, which is phosphorylated by the type I receptor on a serine residue in the carboxyl-terminal domain. Third, the phosphorylated Smad2 (P-Smad2) heterogeneously dimerizes with a common partner (Co-) Smad, Smad4, and together they translocate into the nucleus. Once in the nucleus, the Smad protein may form a complex with other transcription factors and subsequently activate target genes (13- 16).
There is little knowledge obtained from actual physiological conditions to demonstrate the signal transduction processes and proteins involved in regulating gene expression in this system (17). The majority of knowledge on Smads was achieved using non-physiological conditions, such as cloning interesting genes to amplify the expression signals. Conceptually two types of immunoprecipitation (IP) approaches can be applied. One approach is to perform IP using lysates from cells coexpressing candidate interacting proteins that have been introduced by transient cotransfection of expression constructs (18). This method has the advantage that convenient epitope tags can be engineered into the expression constructs with the high level of protein expression obtained in transfected cells facilitating detection of protein interactions. Although this cotransfection approach is often successful, the data may not be physiologically relevant because these Smad proteins are present at far higher levels in cloned cells than would occur under normal physiological conditions. Alterna-tively antibodies against the native protein partners can be used to immunoprecipitate endogenous proteins from cells. This latter approach would provide information on protein interactions under physiological conditions. MS combined with several approaches for the separation of complex mixtures that precedes their mass spectrometric analysis has become the method of choice for the identification of proteins (19). Isotopic labeling combined with MS has also been used extensively to produce accurate quantitation of biological molecules (19). The development of isotopecoded affinity tag reagents allows for quantitation through isotopic labeling and simultaneously achieves a reduction in sample complexity by measuring only the Cys-containing peptides (20). The CID of peptides of interest by MS/MS would give rise to a sequence-specific fragmentation pattern from which the identity of the parent protein can be derived using either database search algorithms or de novo CID spectral interpretation (21).
We report here the results of direct proteomic analysis of Smad2 and its interacting partners from mink lung epithelial (Mv1Lu) cells. We identified not only Smad2 itself but also the interacting partners in both control and TGF-␤-induced samples. Many of these proteins have not been identified previously as part of the Smad signal transduction pathway.

EXPERIMENTAL PROCEDURES
Materials-Unless stated otherwise, all chemical reagents were of the highest purity available and purchased either from Sigma or Fisher Scientific. Protein G-Sepharose (fast flow) and the secondary antibodies used for Western blotting were from Amersham Biosciences. Monoclonal antibody (mAb) against Smad2 was ordered from BD Transduction Laboratories (Lexington, KY), mAb against Smad4 was from Santa Cruz Biotechnology (Santa Cruz, CA), polyclonal antibody anti-P-Smad2 was from Upstate Biotechnology Inc. (Waltham, MA), and TGF-␤1 was from Calbiochem. Cleavable ICAT (cICAT) reagents and kits were obtained from Applied Biosystems (Framingham, MA). Siliconized 0.65-ml tubes from PGC Scientifics (Frederick, MD) were washed with methanol and water prior to use. The polysulfoethyl A strong cation exchanger was obtained from Poly LC through Western Analytical Products (Murietta, CA). BCA protein assay reagent kit and ImmunoPureா immobilized monomeric avidin gel were from Pierce.
Cell Culture-The Mv1Lu cell line (CCL 64) was purchased from the American Type Culture Collection and was grown in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum in a water-saturated, 5% CO 2 atmosphere at 37°C in 75-cm 2 polystyrene cell culture flask. Cells grown to ϳ80% confluence were harvested for lysate preparation.
TGF-␤ Stimulation-Mv1Lu cells starved for 12 h were incubated with or without platelet-derived TGF-␤1 (final concentration, 100 pM) for various periods of time (0 -180 min) at 37°C. After the incubation, the cells were quickly washed twice with ice-cold PBS. The cells were lysed with 1 ml of lysing buffer containing 50 mM Tris (pH 8.0), 150 mM NaCl, 1% Nonidet P-40, 2 mM EDTA, 100 M PMSF, 1 mM aprotinin, and 50 mM calyculin A. Calyculin A as an inhibitor of phosphatases was used only for the cells induced with TGF-␤. However, both control and induced samples containing calyculin A had been screened, and there was no difference as compared with control samples without calyculin A. After centrifugation at 1.4 ϫ 10 4 rpm in a 5415 C centrifuge (Brinkmann Instruments) for 15 min at 4°C, the supernatants were collected for immunoprecipitation after clearing with protein G-Sepharose.
Immunoprecipitation of Smad2 Complexes-Typically the lysate from ϳ1.5 ϫ 10 8 cells in 12 75-cm 2 cell culture flasks was extracted using immunobeads prepared from mAb covalently cross-linked to protein G-Sepharose beads with dimethyl pimelimidate as described elsewhere (22). The mAb concentration in the bead is 1.2 g (antibody)/l (drained bead volume). For IP, a 30-l slurry was added to 1 ml of Mv1Lu cell lysate and incubated for 2 h at 4°C with constant gentle rocking. After sedimentation by centrifugation, the beads were washed as follows: 1 ml of ice-cold lysing buffer, 1 ml of ice-cold PBS containing 0.5 M NaCl, 1 ml of PBS containing 0.1% Tween 20, and 1 ml of PBS (three times). Bound proteins were released from the beads with 30 l of nonreducing SDS sample buffer and heated for 5 min in an 85°C water bath. The protein supernatant after centrifugation was separated by SDS-PAGE.
In immunoprecipitation, co-isolation of nonspecifically associated proteins sometimes occurs that will possibly result in false positive identification. Therefore, it is important to remove nonspecific interacting proteins during IP. To achieve this, the following steps were carried out. First, the cell lysates were precleared using adsorbents without coupled antibodies. Second, monoclonal antibodies were used for IP. Third, efficient washing steps were designed with appropriate salt concentration and detergent solutions to remove any nonspecific interacting associations by charge or hydrophobic properties.
Western Blotting-27 l of lysate and 7 l of a nonreducing sample buffer that was concentrated 5-fold were loaded per lane and separated by SDS-PAGE. Proteins were then transferred from the gels to PVDF membrane (Immobilon TM -P, Millipore Corp., Bedford, MA) at 4°C for 90 min using 75 V in 5% (v/v) methanol, 25 mM Tris (pH 8.4) containing 14.4% (w/v) glycine. Dilution of primary antibodies was between 1:100 and 1:1000 depending on the affinity of the antibodies. Protein signals were detected using peroxidase-linked secondary IgGs and enhanced chemiluminescence (Amersham Biosciences).
SDS-PAGE and In-gel Tryptic Digestion-Proteins were released from the beads with a 5-fold concentrated nonreducing sample buffer, and the total 30-l protein supernatant was electrophoresed in a 7.5-17.0% gradient, 1-mm-thick gel using a constant current of 400 mA for 1 h. The resolved proteins were visualized by silver staining (23). The bands corresponding to Smads and the interacting partners were excised. The gel bands were destained with a 1:1 (v/v) solution mixture of 30 mM potassium ferricyanide and 100 mM sodium thiosulfate (24). The proteins in the gels were digested with trypsin using a protocol modified from Hellman et al. (25). Briefly gel pieces were completely dried down in a vacuum centrifuge, rehydrated with a trypsin solution, and allowed to incubate on ice for 45 min. After 45 min, the trypsin supernatant was removed and replaced with ϳ20 l of digestion buffer without trypsin so that the gel pieces were covered. The gel pieces were kept wet at 37°C overnight for digestion with mixing.
MALDI-TOF Mass Spectrometric Analyses-The method developed by Hellman et al. (25) was used to extract the tryptic peptides. Aliquots of 0.5 l were eluted from a C 18 ZipTip with 10 mg/ml ␣-cyano-4-hydroxycinnamic acid in 50% acetonitrile/water containing 0.1% trifluoroacetic acid, applied directly onto a target plate, and allowed to air dry. The tryptic peptide masses were obtained using a Voyager-DE TM STR MALDI-TOF mass spectrometer (Applied Biosystems, Foster City, CA) at a resolution of 5000 (full width at halfmaximum) and used for peptide mass fingerprinting (PMF). The database used was NCBInr (June 1, 2005). Known trypsin autocleavage peptide masses (842.51, 1045.56, and 2211.1 Da) were used for a three-point internal calibration for each spectrum. In the searches, methionine residues were assumed to be modified to methionine sulfoxide, and cysteine residues were assumed to be reduced and alkylated by iodoacetamide to carboxyamidomethyl cysteine wherever necessary. The database-fitting program Profound (26) was used to interpret MS spectra of protein digests. A protein was considered to be identified with p value of 0.05 and a 95% confidence level when the spectrum of its measured peptide mass met these previously established criteria for positive identification of proteins using MALDI-TOF mass spectrometry and automated database analysis (27)(28)(29)(30). First, a minimum of five measured peptide masses must match tryptic peptide masses calculated for an individual protein in the database with a mass tolerance of 0.1 Da for monoisotopic mass and 1 Da for average mass. Second, the peptides identified by these matches must provide at least 15% sequence coverage of the identified proteins. PMF by MALDI MS is gradually reaching a 95% confidence level compared with sequencing techniques due to improved mass accuracy and sample preparation methods (31). As an example with Profound, proteins are ranked by the probability that "the candidate protein is the single protein" (26) (see Fig. 3). Profound uses the expectation values where the smaller the expectation value, the more likely it is a true match rather than a random match. An expectation value of 1.0 ϫ 10 Ϫ4 would indicate a similar match once in every 10,000 similarly sized databases and thus confirm the protein that matches the MS data. Profound also uses the Z score as an indicator of the quality of the search results. For example, a Z score of 1.65 indicates a 95th percentile result or a 5% random result having a higher Z score. The expectation value of Profound makes it easier to discriminate between scored sequences of top ranked proteins and other predicted results (26). For those proteins with a probability of less than 1 in PMF, repeating experiments under different experimental conditions (e.g. using a different enzyme) can significantly increase the confidence of protein identification.
cICAT Labeling-Protein samples were assayed with the BCA protein assay reagent kit (Pierce). Protein samples of 15-20 g were labeled with cICAT reagents using a modified protocol. Briefly protein samples were denatured in 8 M urea, 50 mM Tris, 5 mM EDTA, 0.05% SDS, pH 8.4. Reduction with 5 mM tributylphosphine was carried out for 30 min at 37°C. cICAT reagents in acetonitrile were mixed together with reduced protein samples. Labeling was allowed to proceed for 2 h at 37°C. Samples labeled with 12 C and 13 C cICAT reagents were combined and diluted to below 1.5 M urea. Tryptic digestion was initiated with the addition of 1% (w/v) of side chainmodified, tosylphenylalanyl chloromethyl ketone-treated porcine trypsin and allowed to proceed at 37°C for 12 h.
Cation Exchange Chromatography-Cation exchange chromatography was used to fractionate the peptide mixture. Tryptic digest samples were adjusted to below pH 3.0 with formic acid. The SCX separation was carried out using an off-line spin column (Bio-Rad). The cation exchanger (30 mg) was equilibrated with buffer A consisting of 5 mM KH 2 PO 4 , 25% ACN, pH 3.0. Samples were loaded onto columns and gently mixed for 15 min. The columns were washed with 500 l of buffer A followed by a step gradient of different ratios of buffers A and B (500 mM KCl plus buffer A). Fractions (2 ϫ 60 l) were collected in 0.65-ml siliconized tubes.
Avidin Affinity Chromatography-The SCX fractions were neutralized with NH 4 OH, and the pH was adjusted up to 8.0. The above mentioned spin column was also utilized for the avidin affinity chromatography. 120 l of ImmunoPure immobilized monomeric avidin gel (Pierce) was suspended in a spin column. The column was blocked with 2 mM D-biotin, 50 mM NH 4 HCO 3 , pH 8.0. The column was then primed using 0.5 ml of 0.4% TFA in 30% ACN followed by 1 ml of 100 mM NH 4 HCO 3 , pH 8.4. Samples were loaded, and the flow-through was collected for LC-MS/MS analysis. The column was washed with 1 ml of 100 mM NH 4 HCO 3 , pH 8.0, followed by the same solution containing 10% methanol (v/v) and then by 1 ml of HPLC grade water. Labeled peptides were eluted with 3 ϫ 60 l of 0.4% TFA in 30% ACN.
LC-MS Analysis-The labeled tryptic peptides were cleaved according to the standard protocol supplied by Applied Biosystems. The cleaved peptides were subjected to LC-MS/MS analysis on a QSTAR Pulsar i mass spectrometer (Applied Biosystems, Foster City, CA). Chromatographic separation of peptides was performed on a capillary and nano-HPLC system (LC Packings, San Francisco, CA). The LC eluent from a 75-m-inner diameterϫ 15-cm PepMap C 18 column (Dionex, Marlton, NJ) was directed to a microion spray source. Throughout the LC gradient, MS and MS/MS data were recorded continuously using a 6-s cycle time. With each cycle, MS data were accumulated for 1 s followed by two CID acquisitions of 2.5 s each on ions selected by preset selection parameters of the information-dependant acquisition method. In general, the ions selected for CID were the two most abundant obtained from the survey MS spectrum except that singly charged ions were excluded and dynamic exclusion was used to prevent repetitive selection of the same ions within a preset time. Rolling collision energies were used to adjust automatically for the charged state and the mass/charge value of the precursor ion. Searches were performed using MASCOT of the NCBInr database. In all searches, methionine sulfoxide was selected as a variable modification, and cysteine residues was selected as a fixed modification (ICAT heavy and light). The peptide precursor mass tolerance was Ϯ100 ppm, and the MS/MS product ion mass tolerance was Ϯ0.1 Da.
Protein Quantitation-Protein expression ratios were calculated from the peak areas corresponding to ICAT-labeled peptides in the MS spectra. Ratios for each quantitation result were obtained for complete pairs based on the accuracy of the measured mass difference and the similarity of light ( 12 C) and heavy isotopic ( 13 C) profiles.
MALDI MS TOF/TOF data were acquired in an automated mode using the 4700 Proteomics Analyzer (Applied Biosystems, Foster City, CA). Initially a MALDI MS spectrum was acquired from each spot (1000 shots/spectrum), and peaks with a signal-to-noise ratio greater than 15 in each spectrum were automatically selected for MS/MS analysis (7500 shots/spectrum). A collision energy of 1 keV was used with air as the collision gas. Peak lists from all MS/MS spectra were submitted for database searching using an in-house copy of MAS-COT, Version 1.8 (Matrix Science Inc., Boston, MA). The following criteria were used for all database searches: a minimum signal-tonoise ratio threshold of 5-10; masses of 0 -60 Da and masses within 20 Da of the precursor ion were excluded; and a maximum of 60 peaks per spectrum were included as product ions. The mass tolerance was Ϯ75 ppm for MS data, Ϯ200 ppm for MS/MS precursor ions, and Ϯ250 ppm for MS/MS product ions. All samples were searched against the NCBInr (May 1, 2005).

Time Course of Smad2
Phosphorylation-The Mv1Lu cell line has been used to study the biology of TGF-␤. In these cells, TGF-␤ arrests cell proliferation and produces an apo-ptotic response (32). However, this response is context-dependent. To determine the best time window to analyze complexes involving phosphorylated Smad2 in the TGF-␤ signal transduction of the pathway, a time course of stimulation with TGF-␤ was carried out. As seen in Fig. 1, antibodies specific for phosphorylated Smad2 detected only one band in immunoblots. By 5 min, phosphorylated Smad2 was quickly detected and was only faintly visible after 3 h. Based on this result, 20 min of stimulation time was chosen for subsequent experiments.
Immunoprecipitation with Anti-Smad Antibodies-To prevent contamination of isolated protein complexes with immunoglobulins, all antibodies were immobilized on protein G-Sepharose using dimethyl pimelimidate. Immobilization and precycling just prior to use were found to be important for minimizing nonspecific elution of heavy and light immunoglobulin chains. Elution of the bound proteins by 5ϫ sample buffer (0.3 M Tris, pH 6.2, 5% SDS, and 40% glycerol) in an 85°C water bath for 5 min was also useful. Monoclonal anti-body targeted to the Smad2 amino acid sequence (residues between 142 and 263 from mouse Smad2) was used for immunoprecipitation. As seen in Fig. 2, A and B, use of this mAb resulted in isolation of many polypeptides both in unstimulated control cells and in TGF-␤1-stimulated cells. In contrast, direct isolation of Smad4 with monoclonal antibodies against itself suggests that this protein is possibly present in the cell primarily free of associations with other proteins (Fig. 2C). As will be shown below, Smad4 is present among the polypeptides isolated by the Smad2 antibody.
PMF and Protein Identification of Polypeptides in Complex with Smad2-Despite the progress of genomic sequencing, many important model organisms have not been covered adequately as yet. When this experiment was carried out, no cDNA sequences from mink species were present in public databases. Therefore, PMF was limited to searches for homologous proteins that might show conserved tryptic peptides. The success rate of protein identification by sequence similarity searches depends on the molecular properties of analyzed proteins, the evolutionary conservation between analyzed proteins, and their homologues in a database (33). However, there are many successful examples in using PMF for homologous identification of proteins (34 -38). Within the mammalian subkingdom, over 80% of proteins could be positively identified by sequence similarity searches because orthologous proteins share substantial sequence identity (36). Many proteins in signal transduction are highly conserved in evolutionary processes. In fact, in the Smad pathway, there are many proteins with a high degree of identity in their sequences. Moreover LC-MS/MS and nano-LC-MALDI MS TOF/TOF were used to further confirm the homologous identification of proteins using PMF. LC-MS/MS was also used for quantitation, measuring the ratios of proteins in the TGF-␤ signal transduction pathway between control and treated samples.
Bands from the silver-stained one-dimensional gels (Fig. 2, A and B) were cut into small pieces and digested with trypsin as described under "Experimental Procedures." Similar gels  Table I.

TABLE I Proteins identified from immunoprecipitation and peptide mass fingerprinting
Band numbers are as shown in Fig. 2. Probability, Z value, sequence coverage, and theoretical and experimental masses obtained from Profound peptide mass fingerprinting search results are provided. In most of the gel bands, at least one protein was identified. However, in some gel bands, two or three different proteins were identified. These data are summarized from quadruplicate experiments.  from two separate but identical experiments were also analyzed. It is of note that 10 8 cells were used to generate the immunoprecipitates separated in these lanes. This was necessary because of the low abundance of the Smad-interacting partners present in Mv1Lu cells. Mammalian databases were searched using the parameters described in the legend to Table I and under "Experimental Procedures." Most proteins identified were from the mouse or human genome with a few from bovine. The coverage of the sequences ranged from 15 to 48% having high statistical probabilities of 1. For those identified proteins with a probability of less than 1, the identification confidence was enhanced using repeated experiments and comparing the matched peptides from every experiment. Table I is a summary of results from quadruplicate experiments. Moreover some identified proteins were further confirmed by orthogonal LC-MS/MS or nano-LC-MALDI MS TOF/TOF. Identification and discrimination of the Smad proteins is discussed below. The proteins identified are only those that show a remarkable sequence identity across all species known to express them.
Of particular interest is the presence of some proteins appearing in both unstimulated and stimulated cells. For example, the E1B-55 kDa-associated protein 5 (E1B-AP5) was detected under both conditions. E1B-AP5 is a ubiquitously expressed heterogeneous nuclear ribonucleoprotein (hnRNP) that functions as a regulator of transcription by linking chromatin remodeling and mRNA processing and transport (43). Heat shock protein 70 was also detected in both control and stimulated cells. Cytoskeletal proteins were detected in these experiments and will be discussed below. In addition, other proteins whose functions are implicated in cellular association, such as octamer-binding transcription factor 1, were found. No immunoglobulin heavy or light chains were identified in any experiments reported in this study. Fig. 3 demonstrates the identification of two proteins in one band from one-dimensional SDS-PAGE. Fig. 4 is an example of identification using cICAT LC-MS/MS. Smad2 polypeptides were detected at a location in the gel appropriate to its predicted molecular weight. Smad2 was identified by the presence of 16 peptides (Fig. 3) corresponding to the peptides in the alignment shown in Fig. 5. Smad2 was also detected in bands much higher than its expected molecular mass. Although unusual, we believe it is possible that the association between Smad2 and E1B-AP5 identified by both mass mapping and cICAT is not disrupted by incubation with 1% SDS at 85°C for 5 min. The apparent molecular weight of both Smad2 and E1B-AP5 is at the appropriate location in the gel. Smad4 polypeptides were also identified in the TGF-␤-stimulated cells and in the same gel band with Smad2 ( Fig. 3) but not in control cells. Smad4 was discriminated from Smad2 by nine peptides in PMF. The amino acid homology between Smad2 and Smad4 is 36.9% (Fig. 5). It is of interest that the matched peptides are located in similar positions in aligned sequences shown in colored amino acid sequences (red, Smad2; green, Smad4) (Fig. 5); this implies that for homologous proteins, such as Smad2 and Smad4, the matched peptides in mass mapping are homologous too.
ICAT Labeling and LC-MS/MS Identification of Smad-associated Proteins-Because of the large number of unidentified proteins in these immunoaffinity purification experiments, some of the protein mixtures remaining from these experiments were differentially labeled with the cICAT reagent to permit relative quantitation of the proteins in these cells. Perhaps because of the smaller amount of protein utilized and the two drying steps used in the procedure, fewer proteins were identified in this experiment. Table II shows the identification of proteins that had multiple cysteine-containing peptides identified. Note that several additional peptides (non-cICATlabeled) were also identified. This has been reported previously to occur and is likely due to nonspecific binding of acidic peptides to the positively charged streptavidin used to purify the labeled peptides. These polypeptides are also among the most abundant in these protein mixtures. Smad2 and Smad3 could not be discriminated in these cICAT experiments because only cysteine-containing peptides completely conserved across species were identified. Some proteins such as the DEAE box protein and microtubulins were not identified in PMF but found only in cICAT experiments. Therefore, experiments of PMF and multidimensional LC-MS/MS complement each other for the identification of proteins in complex protein mixtures. The quantitative differences in this peptide from Smad2/3 are shown in Fig. 4. All of the tubulin peptides labeled with the 13 C reagent were present in a lower relative abundance than that with 12 C; this is consistent with a previous observation that microtubulin association with Smad proteins diminishes after TGF-␤ stimulation (39). Most of the proteins identified using MS/MS are represented by single hits (40,41) likely due to a lower cICAT labeling rate. The proteins identified are summarized in Table II. The singly and doubly charged product ions in the MS/MS spectra are checked and found to be real. All identifications listed in Table  II were manually confirmed.
Nano-LC-MALDI MS TOF/TOF Identification of Smad-associated Proteins-MALDI has been shown to be complementary to ESI with respect to the population of peptides and proteins that may be detected (42). The ability to sequentially couple orthogonal LC techniques has made it possible to analyze complex peptides mixtures derived from multiple proteins. We further applied nano-LC-MALDI MS TOF/TOF to analyze the peptide mixtures from Smad-associated proteins. Here we only analyzed the control sample for comparing with those from PMF and LC-MS/MS. One of the MS spectra obtained with LC-MALDI MS TOF/TOF is shown in Fig. 6. From this MS/MS spectrum, the amino acid sequence is identified as Smad2 or Smad3. The corresponding sequence is shown in Fig. 5 in blue and underlined. The identified proteins including its identified peptide amino acid sequence are summarized in Table III :17887367). Highlighted in red and green are the peptides obtained from MALDI mass mapping, respectively, and in blue are peptides obtained from tandem mass spectrometry. Protein sequences were aligned using ClustalW version 1.82 (www.ebi.ac.uk/clustalw). *, identical residues; :, conserved residues; ., semi-conserved residues.
the different properties of the peptides and their product ions (40).
Validation-To verify interactions of Smad2 protein with its partners, co-immunoprecipitation and Western blot were used. Due to the limitation of antibody sources, we were only able to validate hnRNP. The rat monoclonal antibody is against hnRNP E1B-AP5 and has been characterized by Kzhyshkowska et al. (43). The Western blot analysis is shown

identification from LC-MS/MS analysis of cICAT-labeled isolates
Matched results of proteins obtained from LC-MS/MS experiment are shown with relative concentration of control/TGF-␤1-induced ( 12 C/ 13 C). All MS/MS spectra corresponding to every identification using MASCOT are provided in supplemental materials with all the generated DTA files in Table S1. TLS, translocated in liposarcoma.

NBCI accession no.
Protein identified 12 Table S1, and C* is cICAT-labeled Cys. c The score is obtained from MASCOT database based on a single peptide MS/MS matching. d "ϩ" represents cICAT-labeled; "Ϫ" represents non-cICAT-labeled.  in Fig. 7. In both TGF-␤1-induced and control samples, there is one band corresponding to hnRNP molecular weight, thus proving that hnRNP is an interacting partner of Smad2. This is consistent with the MS identification. However, the quantitation from Western blot is not able to easily demonstrate the type of quantitative difference seen in the MS experiments.

DISCUSSION
Transforming growth factor-␤ family members regulate a broad spectrum of biological responses on a large variety of cell types. TGF-␤ family members initiate their cellular responses by binding to distinct receptors with intrinsic serine/ threonine kinase activity and activate specific downstream intracellular effectors, Smad proteins. Smads relay the signal from the cell membrane to the nucleus where they affect the transduction of target genes. Smad activation, subcellular distribution, and stability have been found to be intricately regulated. A broad array of transcription factors has also been primarily identified as Smad partners (44). In the present study we identified Smad proteins and their interacting partners in Mv1Lu cells. Our identifications show only partial overlap with previous reports in the literature.
TGF-␤ Receptor-induced Smad Activation-The first step in the intracellular TGF-␤/Smad pathway is the recruitment of Smads to the TGF-␤ receptor complex. Several proteins with anchoring, scaffolding, and/or chaperone functions that regulate and facilitate this process have been identified in this study. In the current hypotheses of Smad interacting processes, Smad anchor for receptor activation (SARA) has been shown to regulate the subcellular distribution of Smad2 and Smad3 (45). SARA is associated with the inner leaflet of the plasma membrane via its FYVE domain, which interacts with phospholipids (45). Hrs/Hgs, another FYVE domain-containing protein, was also shown to participate in Smad presentation to the receptor and to synergize with SARA in stimulating TGF-␤/Smad signaling (46). The adaptor molecule disabled-2 (Dab-2) was also found to serve as a bridging function between TGF-␤ receptor complex and Smads (47). However, SARA, Hrs/Hgs, and Dab-2 were not detected from our mass spectrometry measurement. One possibility for this discrepancy is that the interactions between Smad2 and those anchoring proteins are transient. In our MS experiments, a closely related adaptor protein is a FERM-containing protein.
FERM-containing proteins might play multiple roles in cellular processes and will be further discussed. After phosphorylation of Smad2 by TGF-␤, Smad2 dissociates from microtubulins and translocates into the nucleus (39). Filamin-1, a protein that connects the actin filament networks to membrane receptors and acts as a scaffold protein for signal transduction molecules, is also bound with Smads (48). Moreover Axin, a negative regulator in Wnt signaling, was recently identified as an adaptor of Smad3 that may facilitate TGF-␤ receptorinduced Smad3 activation (49). Our new identifications include actin-␤, myosin, and F-actin capping protein (Tables I,  II, and III). Some of these proteins interact indirectly with Smads. Interestingly TGF-␤ receptor protein-1 (TRAP-1) was found to specifically interact with Smad4. TRAP-1 associates with the inactive TGF-␤ receptor complex, and upon receptor activation, TRAP-1 dissociates from the receptor complex and associates with Smad 4. TRAP-1 possibly functions as a chaperone to reduce the autoinhibitory interactions between amino-and carboxyl-terminal domains and thereby facilitates the interaction with activated R-Smads (50). However, our work showed that heat shock proteins are involved in TGF-␤ signal transduction processes. Heat shock proteins (HSPs) can be classified according to their molecular mass into four major families, the small, low molecular weight HSP (HSP) family, the HSP60 family, HSP70 family, and HSP90 family (51). Each family is comprised of several members and exhibits a distinctive constitutive and inducible expression pattern. HSP can be induced in a cell upon exposure to environmental stress, including heat shock, oxidative stress, heavy metals, or pathological conditions such as ischemia, inflammation, tissue damage, infection, and neoplastic transformation. Additionally HSPs exert multiple physiological functions, including the regulation of cellular homeostasis and apoptosis, and play an important role in tumor antigenicity (52). Our MS experiments also identified heat shock proteins with molecular masses of both 70 and 90 kDa. The function of HSP70 is to assist in the correct folding and assembly of newly synthesized proteins, the refolding of partially denatured or misfolded proteins, and the participation in the disassembly of protein aggregates (53). Recently Mattaj and Englmeier (54) reported that HSP70 plays a role in nuclear localization sequence protein import. Recently Morimoto (55) found that HSP could dynamically remodel transcription complexes. Little is known about small HSP (51). HSP70 possibly helps assemble or facilitate the formation of the complex of Smad2 with its partners. Because Smad2 plays different roles in different signal transduction pathways, it may exist in the FIG. 7. A, Western blot immunostained with monoclonal rat antibody against hnRNP E1B-AP5. About 20 ng of protein from coimmunoprecipitation extraction of both control and TGF-␤-treated Mv1Lu cell lysates were loaded onto a 5-17% SDS-PAGE gradient gel. Proteins were transferred, and the membrane was immunostained using monoclonal rat antibody against hnRNP E1B-AP5. B, antibody is monoclonal antibody against Smad2 only. All other things are the same as in A.
cytoplasm in different interacting states. In contrast, Smad4 seems to be free in the cytoplasm with few interacting partners as shown in the control Mv1Lu cells (Fig. 2C) from our MS experiments. Only after the cells were induced was Smad4 associated with phosphorylated Smad2.
Nuclear Import and Export of Smad Proteins-Without stimulation of ligand, R-and Co-Smads exist primarily in the cytoplasm, and ligand stimulation induces their nuclear accumulation. The amino-terminal regions of R-and Co-Smads contain a nuclear localization-like sequence (Lys-Lys-Leu-Lys) (56). This motif has been reported to be necessary for Smad3 nuclear accumulation after TGF-␤ stimulation. Smad3 directly binds to importin-␤ via its amino-terminal region but not to importin-␣ (57), and Ran GTPase promotes nuclear translocation of activated Smad3 to the nucleus (58). Smad2, when compared with Smad3, has a unique exon insert in the amino-terminal domain that prevents its association with importin-␤ (58). In contrast with Smad2 and Smad3, Smad4 has a leucine-rich nuclear export signal in its linker region (59) (Fig.  5). We identified three proteins that reportedly facilitate molecular transportation between the nucleus and cytoplasm: hnRNPs, DEAD-box protein, and HSP. hnRNPs are predominantly nuclear RNA-binding proteins that form complexes with RNA polymerase II transcripts. These proteins function in a broad array of cellular activities ranging from transcription and pre-mRNA processing in the nucleus to cytoplasmic mRNA translation and turnover (60). Recent studies suggest that several fundamental characteristics of hnRNPs account for their involvement in regulatory pathways (61). Dreyfuss et al. (62) have reported that hnRNP A2/B1 is involved in mRNA splicing or mRNA localization. In co-immunoprecipitation and Western blot analysis (Fig. 7) we further confirmed that hnRNP directly binds to Smad2 or hnRNP binds to an interacting partner of Smad2. Nevertheless it is expected that hnRNP plays a role in the TGF-␤ signal transduction process together with Smad2. One of the functions of DEAD-box protein is to facilitate the export of mRNA from the nucleus (63). Interestingly it also interacts with the transcription machinery. It has been reported that HSP70 takes part in nuclear localization sequence protein import (54).
Transcriptional Control by Smads-R-Smads (except for Smad2) and co-Smads can recognize specific DNA sequences (5Ј-AGAC-3Ј), termed Smad binding elements (SBEs), via their MH1 domains in the promoters for target genes. The crystal structure of Smad3 MH1 domain with SBE showed that the ␤-hairpin loop in Smad3 is in contact with the region SBE (64). The DNA-binding ␤-hairpin is highly conserved among R-Smads and Smad4, suggesting that the MH1 domain of other Smads might also have the same motif to recognize SBE. Interestingly Smad2 has a 30-amino acid insertion immediately prior to the DNA-binding ␤-hairpin that inhibits the capacity of Smad2 to bind DNA (64). This additional sequence may change the structure, altering the DNA binding characteristics of Smad2. Therefore, Smad2 and Smad3 may have a different subset of target genes and regulate distinct cellular processes. Generally the intrinsic DNA binding ability of Smads is relatively weak (64). Smads must cooperate with other transcription factors to activate or repress target genes. Many of the Smad transcription regulators are reported (65). It is likely that more members remain to be identified. Our experiments suggest that the following proteins should be added to the regulatory pathway: in control samples, C 2 H 2 type zinc finger protein, TATA-binding proteinassociated factor, zinc finger protein 2; in TGF-␤-treated cell cultures, ring finger protein, protein p 241 , zinc finger protein 278, octamer-binding transcription factor 1, PRO2042, and FERM-containing protein. We should point out that although the cellular processes of these TGF-␤ signal transduction regulators are not clear, their role in this pathway should be further investigated.
Degradation of R-Smad Proteins-Smurfs have reportedly been found to interact with R-Smads, thereby directly targeting R-Smads for ubiquitin-mediated degradation via the proteasome pathway (66). Whereas Smurf 1 preferentially interacts with bone morphogenetic protein R-Smads, Smurf 2 can associate with TGF-␤/activin R-Smads as well as bone morphogenetic protein R-Smads (66). Smurf-mediated degradation of R-Smads induces a decrease in the cellular components to TGF-␤ family-induced responses (66). Activated nuclear Smad2 has also been found to be targeted for proteasome degradation by selective multiubiquitination. Proteolysis of activated Smad2 involves E2 enzymes and is not dependent on carboxyl-terminal phosphorylation of Smad2 but requires nuclear localization of Smad2 (67). Ubiquitinspecific protease 9 was found in TGF-␤ signal transduction processes using our approaches. Therefore, ubiquitin-specific protease 9 appears to participate in regulating Smad2 in TGF-␤ signal transduction pathways in physiological conditions.
The DEAE box protein (DEAEp) was only found by cICAT LC-MS/MS. However, this protein was not identified in PMF, although experiments were repeated four times. This suggests the necessity of combining PMF with LC-MS/MS in protein identification. DEAEp was identified in both control and TGF-␤-induced samples. This protein was less abundant in TGF-␤-induced sample than in control. DEAEp is well characterized, and it has multiple cellular functions (63). Many processes have been shown to require DEAEp. Several DEA-Eps have been found to interact physically with the transcription machinery or colocalize with it. There is no prior report about the function of DEAEp in TGF-␤ signal transduction.
Smad Protein Cellular States-In the immunoprecipitation experiments, both monoclonal antibodies against Smad2 and Smad4, respectively, were used, and both of them were chemically cross-linked to protein G immobilized in Sepharose beads. In SDS-PAGE, there was only one band when mAb anti-Smad4 was used, whereas there were many bands when mAb anti-Smad2 was used. These results imply that Smad4 is possibly not associated with other proteins but rather is present free in cytoplasm, whereas Smad2 was not free. The binding partners of Smad2 are mostly cytoskeletal proteins. Some of these partners have been reported previously, and we identified some of these. In the TGF-␤ signal transduction processes, Smad2 will bind with Smad4, and the complex of Smad2 and Smad4 will associate with other proteins, such as scaffolding and adaptor proteins, or with signal regulating proteins. Although new regulating proteins have been identified, the detailed cellular functions of these protein complexes need to be clarified. Importantly what appears to be a strong binder of Smad2, E1A55 protein, has only been reported in this work. Finally the relative quantitative results of Smad2 and its interacting partners in control and TGF-␤induced samples were determined by Mv1Lu cells in various states because the cells were not synchronized.
Conclusion-Western blots showed the presence of only a small amount of phosphorylated Smad2 in the samples treated with TGF-␤1. This indirectly indicates the multiple functions of Smad2 in signal transduction and the difficulty in capturing Smad2-Smad4-interacting complexes. Most Smad2 proteins appear to be bound to the cytoskeletal scaffold. However, Smad4 proteins are possibly present free in the cytosol and bind with Smad2 only after the cells are treated with TGF-␤1. Many new signal regulators were identified for TGF-␤ Smad2 signal transduction; cICAT labeling experiments clearly demonstrated that the ratios ( 12 C/ 13 C) of tubulin-␣ and tubulin-␤ are 3.7 and 2.5, respectively. This indirectly demonstrates that Smad2 proteins bound to cytoskeletal scaffold were significantly decreased after cells were induced with TGF-␤1. These data further confirm the bound state of Smad2 proteins. One important cellular functional protein, DEAE box protein, was only found by cICAT LC-MS/MS. * This study was supported by National Institutes of Health Grants CA055011 (to R. H. A.) and T32DK07218 (to Q. L.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.