Nature Methods
- 3, 981 - 983 (2006)
Published online: 29 October 2006; | doi:10.1038/nmeth972
Protein interaction screening by quantitative immunoprecipitation combined with knockdown (QUICK)Matthias Selbach & Matthias MannMax Planck Institute of Biochemistry, Department of Proteomics and Signal Transduction, Am Klopferspitz 18, D-82152 Martinsried, Germany.
Correspondence should be addressed to Matthias Mann mmann@biochem.mpg.de Present screening methods for protein-protein interactions (PPIs) rely on the overexpression of artificial fusion proteins, making it difficult to assess in vivo relevance. Here we combine stable isotope labeling with amino acids in cell culture (SILAC), RNA interference (RNAi), coimmunoprecipitation and quantitative mass-spectrometry analysis to detect cellular interaction partners of endogenous proteins in mammalian cells with very high confidence. We used this screen to identify interaction partners of  -catenin and Cbl.
The analysis of PPIs yields important insights into cell signaling and is key for unraveling protein function in the post-genomic era. The traditional yeast two-hybrid (Y2H) screen detects interactions of 'bait' and 'prey' proteins fused to transcriptional activators in the yeast nucleus. Although this method is sensitive, scalable and can be used to identify transient interactions, a major limitation is its high false positive and false negative rates1,
2. Moreover, the Y2H assay detects PPIs outside their normal cellular environment and does not take into account the specific subcellular localization, post-translational modification and dynamic changes in the interaction between bait and prey proteins. Other successful methods rely on overexpression of tagged proteins in relevant cell lines3. Interaction partners that copurify with the tagged protein can be identified by mass spectrometry–based proteomics4. Yet both tagging and overexpression can alter the properties of the protein of interest and yield false results. The gold-standard assay for PPIs is still the coimmunoprecipitation of untagged proteins at their endogenous level. This approach, however, depends on immunoblotting with antibodies against suspected bait proteins and is thus hypothesis-driven.
Several recent methods have allowed for the quantification of proteins by mass spectrometry5. In one approach, named SILAC, the proteome is labeled by metabolically incorporating either a normal or a heavy isotope–substituted amino acid, such as 13C-labeled lysine6. Peptides derived from the two samples can be differentiated in a mass spectrometer owing to their mass difference. The ratio of signal intensities for such peptide pairs accurately indicates the abundance ratio for the corresponding proteins. We have previously used SILAC to detect interactions of proteins with short sequence motifs by relative quantitation of proteins binding to synthetic bait and control peptides7.
RNAi is an emerging technology that makes it possible to specifically deplete proteins from living cells8. We reasoned that by combining SILAC, RNAi and coimmunoprecipitation, it could be possible to develop a new screening method for endogenous PPIs. In principle, four classes of proteins can be found in an immunoprecipitate (Fig. 1a): (i) the target protein itself, (ii) interaction partners of the target protein, (iii) proteins nonspecifically binding to the beads or the antibodies and (iv) proteins with cross-reactivity to the antibody (for example, homologs of the target protein). Removal of the target protein by RNAi would remove categories (i) and (ii) and only leave the background categories (iii) and (iv) (Fig. 1b). Thus, relative quantitation of eluates from cells with and without target protein should result in a high ratio for proteins from categories (i) and (ii), whereas background proteins would be found in a 1:1 ratio.
 | | |
Following this idea we devised a screening method to detect endogenous PPIs with very high confidence (Fig. 1c). Cells are metabolically labeled via the SILAC method by growing them in medium containing either light or heavy stable isotopes of lysine and arginine. The protein of interest is knocked down by RNAi in one of two samples. Cell lysates are incubated with an immobilized antibody against the protein of interest. The precipitated proteins are combined, eluted, digested and analyzed by mass spectrometry. The target protein itself and its interaction partners should be more abundant in the heavy than in the light form and contaminating proteins should be present in similar amounts in both heavy and light forms. Notably, the method only requires depletion of the target sufficient to yield significant peptide ratios rather than complete absence of the target. We have termed this screen QUICK for quantitative immunoprecipitation combined with knockdown.
We chose  -catenin to evaluate our assay because it is a well-characterized protein that has several known interaction partners (see the Science STKE Database of Cell Signaling)9. To knock down expression of -catenin, we used the colon carcinoma cell line LS174T carrying a stably integrated, inducible short hairpin RNA (shRNA) vector10. After SILAC labeling, we induced -catenin knockdown in the 'light cells' by adding doxycycline, and we left the 'heavy cells' untreated. As another control, we performed a crossover experiment (knockdown in the heavy cells). Two days later, we collected the cells, lysed them and incubated them with monoclonal anti–-catenin covalently coupled to protein G agarose. We combined the precipitated proteins from the light and heavy cells, washed them, eluted them at low pH and digested the eluted proteins with trypsin. We analyzed the peptides by online liquid chromatography–tandem mass spectrometry (LC-MS/MS) on a Finnigan LTQ-FT or LTQ-Orbitrap mass spectrometer (Thermo Electron), assigned peptides to proteins using Mascot (Matrix Science) and quantified them with MSQuant (see Supplementary Methods online).
As expected, -catenin was identified as a major component of the precipitates with sequence coverage of about 70%. The protein was over 8 times more abundant in the heavy than in the light form (Fig. 2a), indicative of a knockdown efficiency of 90% (Supplementary Table 1 online). We identified about 140 additional proteins from at least two peptides. Most peptides had abundance ratios of around 1 (Fig. 2b), corresponding to proteins that precipitate irrespective of the presence of -catenin in the cells, and we therefore considered these proteins to be nonspecific contaminants. This was the case for  -catenin, a -catenin homolog with sequence similarity to the epitope detected by the antibody against -catenin. This protein would have been incorrectly identified as a -catenin interaction partner in a standard immunoprecipitation experiment. The ability of our screen to eliminate cross-reacting proteins provides an important advantage compared to classical coimmunoprecipitation.
| | |
All proteins with significantly increased abundance ratios (P < 0.01, Grubbs outlier test; see Supplementary Methods) were well-known interaction partners of -catenin (Fig. 2c,d,f): the T cell–specific transcription factor 4 (TCF-4),  -catenin and the -catenin interacting protein (ICAT). Therefore, the screen did not yield any false positive results. In contrast, we did not identify all reported interaction partners. One possible explanation for this is that the specific cell line used for the analysis, LS174T, expresses an oncogenic -catenin allele that is not phosphorylated by glycogen synthase kinase-3 and is associated with neither axin nor E-cadherin11,
12. Thus, it is not surprising that we did not see significantly differing abundance ratios for these proteins in our screen. Although this characteristic impedes the identification of all PPIs in a single cell line, it allows for the detection of interaction partners that are relevant in a specific cellular context. These results demonstrate that our method can be used to correctly identify several cellular interaction partners of -catenin at different subcellular locations.
The shRNA strategy described above requires establishment of a stable cell line. We also tested the QUICK screen in a format that only requires an antibody and small interfering RNA (siRNA) reagent against the protein of interest. To find interaction partners of Cbl, we transfected light SILAC-labeled 293T cells with siRNA and analyzed mixed immunoprecipitates. We treated both SILAC cell populations with ortho-vanadate to allow capturing phosphorylation-dependent interactions. In addition to Cbl itself, four proteins had substantial abundance ratio changes that made them statistically significant outliers in the population of identified proteins (Fig. 3a and Supplementary Table 2 online). Three of these proteins had already been identified as members of the Cbl interactome13. The fourth protein was sorting nexin 18. Like Cbl, sortin nexins are known to be involved in receptor-mediated endocytosis, pointing to the biological process in which this interaction may be involved. Although they were not statistical outliers of the distribution, several other proteins had increased abundance ratios (Supplementary Table 3 online) and could be functionally relevant interaction partners. For example, the known interaction partners Nck and flotillin 1 are in this group, as are three actin capping proteins that yielded similar abundance ratios, indicating that they may be involved in direct or indirect interactions with Cbl. We also tested whether the siRNA transfection itself led to global changes in the proteome. SILAC quantitation of the 526 most abundant proteins in 293T cells did not show any gross alterations after RNAi knockdown (Fig. 3b and Supplementary Table 4 online).
| | |
In conclusion, the QUICK method assesses interactions between untagged endogenous proteins at their normal cellular levels within the appropriate cell type. It is thus a new, high-confidence screening method that offers substantial advantages over other approaches. As researchers interested in a specific protein usually have both antibodies and knockdown constructs at hand, the availability of reagents should not be a hurdle. Although QUICK is presently not a high-throughput method, improved shRNA libraries covering the whole human genome14 and ongoing efforts to generate antibodies against every human protein15 could make available the necessary reagents for a larger set of proteins.
Note: Supplementary information is available on the Nature Methods website.
Received 23 August 2006; Accepted 4 October 2006; Published online: 29 October 2006.
REFERENCES
- Bader, J.S., Chaudhuri, A., Rothberg, J.M. & Chant, J. Nat. Biotechnol. 22, 78–85 (2004). | Article | PubMed | ISI | ChemPort |
- Li, S. et al. Science 303, 540–543 (2004). | Article | PubMed | ISI | ChemPort |
- Zhu, H., Bilgin, M. & Snyder, M. Annu. Rev. Biochem. 72, 783–812 (2003). | Article | PubMed | ISI | ChemPort |
- Aebersold, R. & Mann, M. Nature 422, 198–207 (2003). | Article | PubMed | ISI | ChemPort |
- Ong, S.E. & Mann, M. Nat. Chem. Biol. 1, 252–262 (2005). | Article | PubMed | ChemPort |
- Ong, S.E. et al. Mol. Cell. Proteomics 1, 376–386 (2002). | Article | PubMed | ISI | ChemPort |
- Schulze, W.X. & Mann, M. J. Biol. Chem. 279, 10756–10764 (2004). | Article | PubMed | ISI | ChemPort |
- Gunsalus, K.C. & Piano, F. Curr. Opin. Cell Biol. 17, 3–8 (2005). | Article | PubMed | ChemPort |
- Reya, T. & Clevers, H. Nature 434, 843–850 (2005). | Article | PubMed | ISI | ChemPort |
- van de Wetering, M. et al. EMBO Rep. 4, 609–615 (2003). | Article | PubMed | ISI | ChemPort |
- Kucerova, D., Sloncova, E., Tuhackova, Z., Vojtechova, M. & Sovova, V. Int. J. Mol. Med. 8, 695–698 (2001). | PubMed | ChemPort |
- Kim, I.J. et al. Clin. Cancer Res. 9, 2920–2925 (2003). | PubMed | ISI | ChemPort |
- Schmidt, M.H. & Dikic, I. Nat. Rev. Mol. Cell Biol. 6, 907–919 (2005). | Article | PubMed | ISI | ChemPort |
- Silva, J.M. et al. Nat. Genet. 37, 1281–1288 (2005). | Article | PubMed | ISI | ChemPort |
- Uhlen, M. et al. Mol. Cell. Proteomics 4, 1920–1932 (2005). | PubMed | ChemPort |
Acknowledgments
We thank B. Blagoev from the Center of Experimental Bioinformatics for fruitful discussions, J.V. Olsen and other members of our department for discussion and assistance, H. Clevers for generously providing us with the inducible -catenin knockdown cell line and the Interaction Proteome project of the European Union for funding.
Competing interests statement:
The authors declare that they have no competing financial interests.
|
