Supplementing traditional genetic approaches, the study of protein interactions has emerged as a valuable method for discovering novel components of signaling pathways. Recent advances in mass spectrometry (MS) made it possible to identify the components of multisubunit protein complexes isolated from cell lysates with high sensitivity and accuracy, thereby facilitating analysis of protein interactions (Bauer and Kuster, 2003). A suitable method of isolating relatively pure native complexes from cells is a key requirement for this type of proteomic analysis.
Comparative studies of different isolation procedures based on affinity chromatography have led to a conclusion that purification of a complex by means of a protein bait tagged with the tandem affinity purification (TAP) tag is optimal for subsequent analysis by MS (Rigaut et al., 1999; Bauer and Kuster, 2003). The TAP tag is a double-affinity module composed of a calmodulin binding peptide and two IgG binding domains of Staphylococcus aureus protein A, separated by a TEV protease cleavage site (Rigaut et al., 1999). The use of the TAP tag allows rapid and efficient purification of complexes from cell extracts under native conditions. TAP in combination with MS now has been validated by many studies in yeast and higher eukaryotes and has led to the identification of multiple protein complexes and large-scale protein interaction networks (Puig et al., 2001; Gavin et al., 2002; Westermarck et al., 2002; Forler et al., 2003; Bouwmeester et al., 2004; Dziembowski and Seraphin, 2004).
We have constructed vectors for inducible expression of TAP-tagged proteins in Drosophila melanogaster cultured cells and in vivo. Using these vectors, we tagged several components of the Notch pathway and expressed these baits in two widely used Drosophila cell lines and embryos. Functionality of TAP-tagged constructs in the context of an organism was validated, among other criteria, by the rescue of corresponding mutant phenotypes. Protein complexes containing the baits and associated proteins were isolated by TAP, and the composition of these complexes was analyzed by liquid chromatography–tandem mass spectrometry (LC-MS/MS). Our results confirmed several known interactions involving Notch pathway elements and uncovered novel interactions, some of which we validated biologically. Thus, TAP–LC-MS/MS should prove to be a useful tool for proteomic analysis in Drosophila.
METHODS AND RESULTS
Vectors for Inducible Expression of TAP-Tagged Proteins in Drosophila
pMK33-NTAP and pMK33-CTAP vectors for use in cell culture were constructed on the basis of the pMK33 (pMt/Hy) vector (Koelle et al., 1991; Fig. 1). To generate these vectors, TAP cassettes were amplified by polymerase chain reaction (PCR) from plasmids pBS1761 (for amino terminal TAP tagging, NTAP) and pBS1479 (for carboxy terminal TAP tagging, CTAP) and cloned into pMK33, preserving the unique sites in the polylinker and allowing in-frame fusions of the TAP tag with a protein of interest. The sequences of the source plasmids pBS1761 and pBS1479 (Puig et al., 2001) are available at http://www.uni-frankfurt.de/fb15/mikro/euroscarf/cz_plas.html. pMK33-NTAP and pMK33-CTAP provide copper-inducible expression of the fusion proteins driven by metallothionein (Mt) promoter, reducing potential toxicity problems associated with expression from constitutive promoters. In addition, the “on-board” hygromycin phosphotransferase gene provides Hygromycin B resistance and facilitates selection of stable cell lines by eliminating a need to cotransfect protein expression constructs with a helper plasmid (Fig. 1).
To construct a vector for inducible expression of TAP-tagged proteins in vivo, we used the pUAST plasmid, which allows flexibility in terms of both the amounts and patterns of expression, using a multitude of available GAL4 drivers (Brand and Perrimon, 1993). NTAP and CTAP modules were amplified by PCR from pBS1761 and pBS1479 plasmids described above, and inserted into pUAST (Fig. 1). Complete sequences of Drosophila TAP vectors were submitted to GenBank (accession nos. AY727853-AY727856).
Generation of Stable S2 and Kc Cell Lines and Transgenic Flies
Using these vectors, we TAP-tagged selected molecules known to be involved in Notch signaling, representing membrane-bound, cytoplasmic and nuclear subcellular compartments. We made two versions of each bait: pMK33-based (for expression in cultured cells) and pUAST-based (for expression in flies). Full-length Notch (NFL), known to be localized at the membrane and in the cytoplasm (Fehon et al., 1990), was tagged at the carboxy terminus, to make NFL-CTAP. Notch intracellular domain (NICD) and full-length Mastermind, both nuclear proteins (Fortini et al., 1993; Bettler et al., 1996), were tagged at the amino terminus to generate NTAP-NICD and NTAP-Mam, respectively. Full-length Deltex, a cytoplasmic regulator of Notch (Busseau et al., 1994), was tagged at either end to make NTAP-Dx and Dx-CTAP fusion proteins.
Drosophila S2 and Kc cells were grown in Schneider's medium (JRH Biosciences) in the presence of 10% fetal calf serum. Cells were transfected with 2 μg of pMK33-based constructs in six-well plates by using Effectene transfection reagent (Qiagen), according to manufacturer's instructions. After 2 days of incubation with transfection mixes, cells were subjected to selection in the presence of Hygromycin B (Sigma). For S2 cells, the optimal concentration of Hygromycin B was found to be 300 μg/ml, whereas Kc cells tolerated up to 600 μg/ml. Stable lines were established after approximately 1 month. Transgenic fly lines were generated using standard methods (Spradling and Rubin, 1982), with at least three homozygous viable lines established for each construct. pUAST-NTAP and pUAST-CTAP vectors were also used to create control fly lines.
Verification of Construct Functionality
We probed the functionality of the TAP-tagged constructs according to the following criteria: (1) protein size on Western blots; (2) subcellular localization; (3) rescue of loss of function mutant phenotypes by the transgenes; and (4) examination of phenotypes resulting from overexpression of the transgenes in flies.
To verify expression levels and the size of TAP-tagged proteins, S2 and Kc cell lines stably transfected with pMK33-based constructs were induced overnight with 0.35 mM CuSO4, after which cells were harvested and resuspended in sodium dodecyl sulfate (SDS) sample buffer. pUAST-based constructs were induced by crossing the transformants with an arm-GAL4 driver, which is expressed ubiquitously at medium levels (Sanson et al., 1996). The 0–16 hr embryos were collected, dechorionated with bleach, and lysed in SDS sample buffer. After electrophoresis and blotting on Immun-Blot PVDF membrane (Bio-Rad), proteins were visualized in a one-step Western detection procedure using anti–Protein A antibody conjugated to horseradish peroxidase (HRP; Rockland Immunochemicals). All of the stably transfected S2 and Kc lines, and all of the fly lines expressed proteins of expected sizes, taking into account the additional 20 kDa contributed by the presence of the TAP tag (Fig. 3 and data not shown).
Immunocytochemical analysis demonstrated that the TAP tagged proteins were localized in the expected subcellular compartments. NTAP-NICD and NTAP-Mam were exclusively nuclear (Fortini et al., 1993; Bettler et al., 1996), NFL-CTAP was localized on the cell surface and in the cytoplasm (Fehon et al., 1990), and NTAP-Dx and Dx-CTAP were exclusively cytoplasmic (Busseau et al., 1994; Fig. 2).
Finally, the functionality of the TAP-tagged constructs was tested in vivo for their ability to rescue relevant loss of function mutants, as well as to elicit overexpression phenotypes, using pUAST-based transgenes. Thus, NFL-CTAP, NTAP-Mam, and Dx-CTAP provided complete rescue of the corresponding mutant wing phenotypes and gave expected overexpression phenotypes (Fig. 2 and data not shown). NTAP-Dx gave a partial rescue and was somewhat toxic upon overexpression. NTAP-NICD could not be tested in a rescue experiment due to lethality with most drivers, but produced phenotypes indistinguishable from those obtained with untagged NICD when ectopically expressed with GMR-GAL4 (data not shown). Overexpression of NTAP or CTAP tag alone resulted in no observable phenotype, using a variety of drivers, suggesting that the TAP tag does not lead to general toxicity in living animals.
Therefore, using the criteria outlined above, we found that NFL-CTAP, NTAP-Mam, Dx-CTAP, and NTAP-NICD are functional and, thus, could be incorporated into relevant protein complexes in vivo, while in the case of NTAP-Dx, addition of the TAP tag compromised some, but not all of the functional activity of Dx.
Tandem Affinity Purification
The general procedure for TAP has been discussed elsewhere (Puig et al., 2001; Cox et al., 2002; Bouwmeester et al., 2004; Gould et al., 2004). We have followed the established protocols with only minor modifications. For protein extraction, we used, in most cases, a detergent-based lysis buffer that was a modified version of the buffer described in Bouwmeester et al. (2004; 50 mM Tris pH 7.5, 125 mM NaCl, 5% glycerol, 0.2–0.4% NP-40, 1.5 mM MgCl2, 1 mM dithiothreitol, 25 mM NaF, 1 mM Na3VO4, 1 mM ethylenediaminetetraacetic acid, Complete protease inhibitor from Roche). For NTAP-NICD, which is a nuclear protein, we also used hypotonic lysis followed by high-salt extraction (Kamakaka and Kadonaga, 1994), and obtained qualitatively similar results (data not shown). However, for other protein complexes, an extraction procedure that specifically targets the compartment where the complex is localized may increase the recovery of the bait and associated components.
Using extracts from S2 and Kc cells and embryos, we found that 1–2 × 109 S2 or Kc cells (0.5–1 L of culture volume), or 3–6 g of embryos provide sufficient starting material for the TAP to obtain a visible band after colloidal Coomassie staining (Fig. 3). Kc and S2 cells stably transfected with pMK33-based TAP constructs were induced overnight by addition of CuSO4 to a final concentration of 0.07 mM, which provides a medium-level induction of the proteins. Cells were pelleted at 500 × g, washed three times with cold phosphate buffered saline, and lysed by homogenization with 5–10 volumes of lysis buffer. pUAST-based fly transgenes were expressed in Drosophila embryos by crossing or recombining the lines with the arm-GAL4 driver (Sanson et al., 1996). The 0–24 hr embryos were collected in population cages, dechorionated with bleach, washed, and immediately homogenized on ice in Lysis Buffer, by using a Wheaton dounce homogenizer with tight pestle. After a 20-min incubation on ice, both cell and embryo extracts were cleared by centrifugation at 20,000 × g for 15 min, followed by ultracentrifugation at 100,000 × g for 45 min, to obtain S100 supernatants that were snap-frozen in liquid nitrogen and stored at −80°C. We found that ultracentrifugation reduced clogging of the columns during subsequent binding and washing steps.
Using S100 lysates, we performed TAP essentially as described in Puig et al. (2001). At each step during TAP, we collected analytical samples, and probed these fractions in a Western blot with bait-specific antibodies. Two examples of purification profiles from Kc cells are shown in Figure 3A,B, for NTAP-NICD and Dx-CTAP. Densitometry measurements of band intensity showed that the yield of the final bait protein was approximately 6% for NTAP-NICD and 2% for Dx-CTAP, relative to the amount of the bait in the S100 extract. Although recovery rates were somewhat different for different bait proteins, in all cases presented here, they were sufficient for an unambiguous identification of the baits and interacting partners.
Purified protein complexes were treated with iodoacetamide at a concentration of 20 mg/ml for 30 min at room temperature and resolved on 4–12% Novex NuPage denaturing protein gels (Invitrogen). Gels were fixed and stained with colloidal Coomassie reagent (Sigma). Estimation of protein amounts in Coomassie-stained gels relative to standards showed that the amount of purified bait protein was between 0.1 and more than 5 μg (Fig. 3 and data not shown). In addition to the bait proteins, in all cases, we also detected other bands that corresponded to putative associated components (detection limit of colloidal Coomassie stain is approximately 10 ng per band). These amounts are sufficient for unambiguous protein identification by LC-MS/MS, which is effective down to subpicomolar range (Bauer and Kuster, 2003). Band patterns were very reproducible in repeated purifications using the same bait/cell type combination (Fig. 3 and data not shown). In our experience, purifications from cultured cells generally resulted in higher yields and less overall background, compared with embryos.
Mass Spectrometry, Protein Identification, and Functional Validation of Identified Interactors
Mass-spectrometric identification of proteins was carried out as described in Bouwmeester et al. (2004). Complete lanes from protein gels were cut into slices (narrow for specific bands and wider for interband areas) and analyzed by LC-MS/MS. We have carried out a total of 105 TAPs using TAP-tagged Notch pathway components as well as several other TAP-tagged proteins and control samples. Protein complexes from 41 purifications were analyzed by LC-MS/MS, and a total of approximately 400 proteins were identified. We would like to emphasize that it is essential to perform several purifications for a given bait protein to obtain a reliable view of its interaction network, as significant interactors are expected to be reproducibly identified in more than one experiment.
Using TAP-MS, we were able to confirm the existence of a molecular complex composed of the Notch intracellular domain, Su(H) and Mam proteins, previously characterized by other approaches (Wu et al., 2000; Fryer et al., 2002; Jeffries et al., 2002). We observed this association in Kc cells using three independent TAP-tagged entry proteins (Notch ICD, Mam and full-length Notch; Fig. 3C, lanes 1, 2, and 4). For example, endogenous Su(H) and Mam proteins were copurified with NTAP-NICD, used as a bait, in three independent experiments. Significantly, we were able to obtain very similar results from Drosophila embryos using UAS-NTAP-NICD transgene crossed to the arm-GAL4 driver (Fig. 3C, lane 3). To our knowledge, this is the first demonstration of a protein complex identified by TAP-MS from a whole organism.
It has been shown previously that Drosophila S2 cells expressing Notch protein aggregate with S2 cells expressing its ligand Delta and that the two proteins can be coimmunoprecipitated (Rebay et al., 1991). We have identified Delta among the proteins associated with TAP-tagged full-length Notch, after mixing Kc cells stably transfected with pMK33-NFL-CTAP with S2 cells expressing Delta (data not shown). Thus, TAP-MS can be used to study interactions of a receptor with its ligands expressed on adjacent cells.
Comparing the interacting partners of NTAP-Dx with proteins that copurified with Dx-CTAP, we found that the overlap between the two sets was not extensive, which was also reflected in different band patterns observed in Coomassie-stained gels (Fig. 3D). Differences in the sets of interacting partners of these two constructs may explain their different abilities to rescue the dx mutant phenotype.
The specificity of the protein bands in the final TAP fractions can be assessed by comparing experimental band patterns with those of controls. Purifications using extracts from untransfected Kc cells and arm-GAL4 embryos, performed in parallel with experimental samples, gave a reproducible pattern of low-level contaminating bands that were also observed in experimental lanes (Fig. 3E, lane 2 is a typical example). Common contaminants were represented by abundant cytoplasmic proteins, such as myosins, heat-shock proteins, tubulin, actin, and ribosomal proteins.
Our TAP-MS approach has identified several potential novel interactors involved in the regulation of Notch signaling. We currently are validating the functional significance of these interactions by using conventional genetic and biochemical approaches. One of such validated interactors, identified as a binding partner of NTAP-Dx, is a Drosophila homolog of mammalian nonvisual β-arrestins, Kurtz. We have confirmed a direct interaction between Kurtz and Dx proteins, demonstrated their genetic and molecular interactions in vivo, and have shown that the formation of a trimeric complex between Notch, Kurtz, and Deltex is involved in down-regulation of Notch signaling (Mukherjee, Veraksa et al., manuscript submitted for publication). In another study, we used the TAP-MS approach to uncover an interaction between Drosophila Myc and an E3 ubiquitin ligase Archipelago and showed that this interaction results in down-regulation of Myc in vivo (Moberg et al., 2004).
It is worth noting that our analysis does not directly provide an estimate of the minimum abundance of the TAP-tagged protein bait that is required to give results with this approach. In all cases presented here, TAP-tagged proteins were overexpressed in cultured cells and in embryos at a level sufficient for the identification of the baits and interacting partners. We do note that, based on our results and other TAP-MS studies, controlled overexpression of the baits does not interfere with reliable identification of complexes whose functional significance has been validated extensively by other approaches. We conclude that the TAP-MS approach can be successfully applied to the study of protein interaction networks in Drosophila, both in cultured cells and in vivo.
We thank the scientists at Cellzome for sharing their expertise in TAP technology. Constructs pBS1479 and pBS1761 were provided by B. Seraphin. The C96-GAL4 UAS-MamH line was a gift from B. Yedvobnick. This work was supported by National Institute of Health grants NS26084, GM62931, and CA098402 to S.A.-T. A.V. was supported by the MGH Fund for Medical Discovery postdoctoral fellowship.