A proteomics strategy to elucidate functional protein- protein ... - Nature

© 2003 Nature Publishing Group http://www.nature.com/naturebiotechnology

TECHNICAL REPORT

A proteomics strategy to elucidate functional proteinprotein interactions applied to EGF signaling Blagoy Blagoev, Irina Kratchmarova, Shao-En Ong, Mogens Nielsen, Leonard J. Foster, and Matthias Mann* Published online 10 February 2003; doi:10.1038/nbt790

Mass spectrometry–based proteomics1 can reveal protein-protein interactions on a large scale2,3, but it has been difficult to separate background binding from functionally important interactions and still preserve weak binders. To investigate the epidermal growth factor receptor (EGFR) pathway4–6, we employ stable isotopic amino acids in cell culture (SILAC)7 to differentially label proteins in EGF-stimulated versus unstimulated cells. Combined cell lysates were affinity-purified over the SH2 domain of the adapter protein Grb2 (GST-SH2 fusion protein) that specifically binds phosphorylated EGFR and Src homologous and collagen (Shc) protein. We identified 228 proteins, of which 28 were selectively enriched upon stimulation. EGFR and Shc, which interact directly with the bait, had large differential ratios. Many signaling molecules specifically formed complexes with the activated EGFRShc, as did plectin, epiplakin, cytokeratin networks, histone H3, the glycosylphosphatidylinositol (GPI)-anchored molecule CD59, and two novel proteins. SILAC combined with modification-based affinity purification is a useful approach to detect specific and functional protein-protein interactions.

The ebb and flow of cellular life depends largely on signaling pathways that are regulated by specific protein-protein interactions. These interactions often involve assembly of large signaling complexes containing many different protein kinases, protein phosphatases, their substrates, and scaffold proteins. A useful method to study protein-protein interactions is affinity purification using a bait molecule, but this is normally confined to verification of suspected interactions by means of western blotting or selective sequencing of the most prominent interaction partners. Here we extend these methods to the study of signaling-dependent interactions. To study the interactions of endogenous, activated EGFR, we devised a strategy based on incorporating isotopically-labelled arginine ([13C6]Arg) into the proteins of a cell population and a posttranslational modification-dependent affinity purification (Fig. 1). This led to the identification of 228 proteins (see Supplementary Table 1 online), of which 28 were specifically enriched by Grb2-SH2, with ratios >1.3 (Table 1 and Fig. 2). Reproducibility was assessed by (i) determining the variability of the ratios for different peptides of the same protein, (ii) determining the variability of the peptide ratio from individual mass spectra over the elution peak, and (iii) doing independent runs for some of the fractions, each of which resulted in an analytical error smaller than the cutoff of 1.3-fold change. Among the 28 proteins, EGFR and Shc are known to bind directly to the SH2 Center for Experimental BioInformatics (CEBI), Department of Biochemistry and Molecular Biology, University of Southern Denmark, Campusvej 55, DK5230 Odense M, Denmark. *Corresponding author ([email protected]). www.nature.com/naturebiotechnology

•

Figure 1. Strategy to study activated EGFR complex. One cell population is grown in normal arginine-containing medium (blue dish), whereas another population is grown in [13C]arginine ([13C]Arg)–containing medium (red dish), encoding all proteins with this 6 Da heavier amino acid (red color). The encoded cells are stimulated with EGF, lysed, and mixed 1:1 with the lysate of unlabeled, untreated cells. The combined cell lysates are affinity-purified with GST fused to the SH2 domain of Grb2, which specifically interacts with the activated, phosphorylated form of the receptor and its associated binders. Proteins eluted from the beads are digested with trypsin and the resulting peptides analyzed by mass spectrometry. Arginine-containing peptides occur in doublets, separated by 6 Da. Proteins that interact with the bait in a stimulation-dependent manner are mostly in their encoded form and therefore have a more intense peak for the labeled arginine-containing peptide (red peaks). These proteins can be distinguished from a large excess of background binding proteins, which manifest themselves by their 1:1 ratio between stimulated and unstimulated states.

domain of Grb2 in their phosphorylated state. This direct binding was reflected in the ratios for these two proteins, in that their magnitude clearly differentiated them from the remaining proteins in the list (Table 1). Because the binding of Grb2 to Shc occurs only after the latter has already been recruited to and phosphorylated by the activated EGFR, this strategy specifically enriches only proteins involved in EGF signaling. Even if the SH2 domain of Grb2 bound other tyrosine-phosphorylated proteins, these would have to increase as a result of EGF stimulation to register in this experiment. To explore further the activation state of EGFR isolated in this manner, we measured receptor autophosphorylation in the unlabeled, unstimulated versus the labeled, EGF-stimulated states (Fig. 2). Two phosphorylation sites, Tyr1086 and Tyr1173—known binding sites of Grb2 and Shc, respectively—were quantified by SILAC. Indeed, autophosphorylated peptides were only observed in the labeled (activated) state, proving that only activated receptor is enriched in this scheme. Endogenous Grb2 and Sos-1 were associated with the activated complex, as expected (Table 1). In the well-established model of EGFR signaling5, Grb2 is recruited to the receptor, bringing Sos-1 close to Ras at the plasma membrane. Sos-1 then activates Ras by GTP exchange, which in turn leads to the activation of the canonical

MARCH 2003

•

VOLUME 21

•

nature biotechnology

315


TECHNICAL REPORT

Figure 2. Quantification of protein ratios from peptide doublets. The panels show a section of the mass spectrum of peptides from different identified proteins. The lower-mass peak cluster is due to the peptide with normal arginine, whereas the higher-mass peak cluster derives from the same peptide from the labeled and stimulated cell population. The top left and top middle panels show peptide quantification doublets from specifically enriched proteins due to direct binding (EGFR and Shc), as seen by the increased peak intensities of the labeled peptide (solid dot) over the unlabeled peptide (unfilled dot) in the mass spectrum. The top right panel shows a doublet from cytokeratin-17, which is enriched to a lesser degree, indicating secondary binding to the bait by means of activated EGFR complex. The bottom left panel demonstrates the typical 1:1 ratio observed in peptide doublets from proteins that are not specifically enriched. Protein abundance ratios are the average of the ratios observed for several mass spectra, similar to the panels shown here, and the average of several peptides identifying the same protein. The bottom right panel shows identification of the peptide containing the EGFR autophosphorylation site Tyr1173 (Y1173). Only the [13C6]Arg-labeled form of the phosphopeptide is detected (see intensity of peak marked by solid dot versus unfilled dot), indicating that activated receptor was purified exclusively.

Raf-MAPK cascade. Another consequence of Ras activation is the stimulation of other small GTPases (Rac, Rho, Cdc42) that direct cytoskeletal rearrangements. In this context we identified Vav-2, a Rho-regulating protein that binds to phosphorylated EGFR8, and Vav-3, which is also reported to associate with the activated receptor9. The IQ motif–containing GTPase-activating protein-1 is a known binding partner for both Rac1 and Cdc42. This binding is stimulated by EGF and has importance for cellular polarization and cytoskeletal remodeling10. To date, Rac GTPase–activating protein-1, another modulator of small GTPases, has mainly been described in hematopoietic and germ cell development11, but the fruit fly homolog, DRacGAP (34% identity), is a negative regulator of EGFR signaling12. Signaling effects are produced as a result of a balance of activating and deactivating steps, and shortly after stimulation, EGFR is internalized and deactivated through a complex process involving multiple mechanisms5. Among the proteins specifically recruited to activated EGFR, we found all four subunits of the adapter complex AP-2, an essential component of receptor-mediated endocytosis by means of clathrin-coated vesicles13. We have also identified members of the ubiquitination pathway, including ubiquitin itself and Cbl, a signaling molecule that has an E3 ligase function and directs proteasome-mediated destruction of EGFR5. Ubiquitin and ubiquitin-like molecule C were identified in the same molecular-weight fractions that contained EGFR, and at high ratios, suggesting that a 316


•

VOLUME 21

•

proportion of the receptor is ubiquitinated within 10 minutes of stimulation. Finally, the receptor tyrosine phosphatase SHP-2 also showed an increased ratio after EGF stimulation14,15. An additional class of proteins identified in this study are involved in growth factor–induced cytoskeletal rearrangement. Association of EGFR and Shc with actin microfilaments is already known16, but we also find the cytoskeletal keratins 7, 8, 17, and 18 in specific association with EGFR complexes—all with similar ratios. Note that specifically interacting keratins can easily be distinguished from contaminating hair or skin keratins, because the latter do not result in any [13C6]Arg-labeled peptides. In addition, we find the intermediate filaments binding proteins plectin and epiplakin, both from the same protein family17. Confocal immunofluorescence microscopy showed co-localization between Shc and the cytokeratin network (Fig. 3, top row). Together these findings strongly suggest association of EGFR complexes with intermediate filaments. Identification of all of the proteins just mentioned shows that our strategy was effective in pinpointing known, functional members of the EGFR complex in a large background of >200 proteins. Two novel proteins were also identified in this extensively studied pathway (Table 1). The sequence of one has homology to Rho-interacting protein-3 (RIP-3), suggesting that it may be involved in EGFRmediated signaling through small GTPases. The other protein has homology to Caenorhabditis elegans smu-1 and rat brain–enriched

MARCH 2003

•

www.nature.com/naturebiotechnology

TECHNICAL REPORT

Table 1. Proteins recruited to activated Grb2-EGFR complex


Gi no.a

s.d.b

No. of peptides

6.3

23

7.9 1.46 2.01 1.46 1.45 1.75 2.54

2.2 0.12 0.03 0.18 0.19 0.21 0.13

22 1 1 5 4 2 1

Proteins involved in signal attenuation 19913416 AP-2, α1 subunit 4557469 AP-2, β1 subunit 11038643 AP-2, σ1 subunit 18554450 AP-2, µ1 subunit 8928017 Cbl-Bc 7445542 Polyubiquitin 17481486 Similar to ubiquitin C 464494 SHP-2c

2.73 2.11 2.77 1.85 1.73 1.64 3.21 2

0.4 0.27 0.1 0.16 0.04 0.07 0.38 0.15

4 3 1 1 3 2 2 5

Proteins involved in cytoskeletal functions 4501889 Actin 14783596 Keratin-7 2506774 Keratin-8 4557701 Keratin-17 12653819 Keratin-18 4505877 Plectin-1 13876386 Epiplakin-1

1.48 1.47 1.54 1.43 1.36 1.48 1.43

0.19 0.18 0.19 0.21 0.13 0.24 0.11

5 16 20 20 11 21 9

Other proteins with no previously reported involvement in EGFR signaling 515118 CD59 4504281 Histone H3, member A 4759158 snRNP D2 14770022 Novel (similar to RIP-3) 8922679 Novel (with WD repeats)

1.48 1.52 1.35 1.4 1.34

0.27 0.09 0.05 0.09 0.1

4 2 1 1 3

Protein name

Ratio

Proteins directly involved in signaling 4885199 Epidermal growth factor receptor (EGFR) 284403 Shc 4504111 Grb2 476780 Sos-1 4506787 IQ motif containing GAP1 7019433 Rac GAP1 11429601 Vav-2 3928847 Vav-3

15.8

aNCBI

GenBank accession number. deviation determined from quantification of several peptides or from the variance of the ratios obtained from individual mass spectra in the case of proteins quantified with a single peptide. cProteins that also have direct signaling function. bStandard

WD-repeat protein18,19. It contains several putative WD40 repeats thought to be involved in mediating protein-protein interactions, a LisH domain, and an associated CTLH domain (C-terminal to LisH), suggested to be involved in cytoskeletal dynamics. No functions have been described for the homologs except C. elegans smu-1, mutations of which affect alternative splicing events18. EGFR has been found in the nucleus, and there is increasing, if controversial, evidence of direct nuclear function20,21. In this respect it is interesting that we found association of activated complex with snRNP D2 and histone H3. Confocal immunofluorescence microscopy of histone H3 and EGFR showed a punctate nuclear localization of histone H3 and partial nuclear localization of EGFR. In the nucleus EGFR showed extensive co-localization with histone H3 (Fig. 3, middle row). Further study will be necessary to address the potential relevance of these findings with respect to the role of histone H3 in regulating gene expression and reports of EGFRmediated histone H3 phosphorylation22. Surprisingly, we also identified CD59, an extracellular GPI-anchored molecule, as a component of the activated complex (Table 1). Functional characterization of CD59, a typical marker for lipid rafts, has centered on its inhibitory role in the complement system and in www.nature.com/naturebiotechnology

•

Figure 3. Subcellular co-localization of selected proteins by confocal microscopy and FRET. HeLa cells stained with antibodies against Shc and keratin (top row) or EGFR and histone H3 (middle row) and their respective overlays. Anti-keratin-18 antibody was conjugated to FITC; Cy5-conjugated anti-rabbit secondary antibodies were used to visualize Shc and histone H3 labeling, whereas Cy3-conjugated anti-goat antibodies were used for EGFR. Confocal slices are between 0.5 and 0.6 µm; for EGFR–histone H3, the image was taken through the center of the nucleus, 3–4 µm from the bottom of the cell, whereas for keratin-Shc, the image was taken near the bottom of the cell. Bottom row: Close association of CD59 and EGFR at the cell surface revealed by FRET. Unpermeabilized HeLa cells were stained as described with antibodies against EGFR (left) and CD59 (middle), and then stained by Cy3-conjugated anti-sheep (left) and Cy5conjugated anti-mouse secondary antibodies (middle). FRET between Cy3 and Cy5 (right) was detected by exciting the sample with 543 nm light and measuring emission intensity in the 645- to 720-nm range (see Experimental Protocol). FRET image is color-coded, with lower intensities colored blue and higher intensities colored red. Confocal slices are between 0.5 and 0.6 µm. Scale bar, 10 µm.

T-cell receptor activation23. Although some studies have reported increased cellular tyrosine kinase activity upon CD59 stimulation24,25, connection with EGFR signaling has not been suggested. Fluorescence resonance energy transfer (FRET) experiments between antibody-labeled CD59 and EGFR indicated that the two proteins are, if not directly associated, at least very closely linked in the plasma membrane (Fig. 3, bottom row). In addition, we observed that a substantial fraction of plasma membrane CD59 disappeared upon EGF stimulation (data not shown). Further functional studies will be necessary to address the precise role of CD59 in EGF signaling. Several advantages of the strategy described here are apparent. Functional protein-protein interactions are revealed through isotope ratios, avoiding a trade-off between false-positive binding and the ability to detect weak interactions. In the SH2 affinity purification reported here, we were able to determine a small number of functional binders among hundreds of nonspecific binders. The strategy is relatively straightforward, involving only standard co-precipitation techniques, and can thus be useful in a wide variety of situations. Endogenous complexes are captured in an activation-dependent manner, reducing the chance of artifacts due to overexpression.

MARCH 2003

•

VOLUME 21

•


317


TECHNICAL REPORT

Dynamics of complexes could also be studied by carrying out the experiment at different time points. Combined with the automation demonstrated in recent large-scale proteomics experiments, the strategy described here can add crucial data for deciphering not only EGFR, but also many other dynamic signaling networks.

analysis (detector gain, pinhole, laser transmission percentage), no signal in the FRET channel was seen for cells labeled only for EGFR (with Cy3) or only for CD59 (with Cy5).

Experimental protocol

Acknowledgments We thank other members of our laboratory for help and fruitful discussions and comments on the manuscript. Akhilesh Pandey provided the GST-SH2 (Grb2) domain construct. Work in the Center for Experimental BioInformatics (CEBI) is supported by a generous grant by the Danish National Research foundation.

Cell culture and affinity purification. For labeling experiments, 2 × 108 HeLa cells (ATCC; American Type Culture Collection, Manassas, VA) were grown in medium containing either normal or [13C6]arginine (Cambridge Isotope Labs, Andover, MA) as described7. After 8 h of serum starvation, the labeled cells were stimulated with 150 ng/ml EGF (Upstate Biotechnology, Lake Placid, NY) for 10 min, whereas the unlabeled cells were left untreated. Cells from both conditions were lysed in lysis buffer containing 1% (vol/vol) Nonidet P-40, 150 mM NaCl, 50 mM Tris, pH 7.5, 1 mM sodium orthovanadate, and protease inhibitors (Complete tablets; Roche, Mannheim, Germany). Lysates were mixed in a 1:1 ratio according to their protein concentrations and incubated at 4 °C with Grb2 fusion protein bound to glutathione-Sepharose beads (Amersham Pharmacia Biotech, Uppsala, Sweden). After 4 h of incubation, the beads were washed extensively with lysis buffer, boiled in LDS sample buffer (Invitrogen, Carlsbad, CA), and resolved on a 4–12% (wt/vol) Novex gel (Invitrogen). The entire lane was excised in ten bands and digested enzymatically with trypsin for liquid chromatography–tandem mass spectrometry (LC-MS/MS) analyses26. The plasmid pGEX-2T (Amersham Pharmacia Biotech), containing a 300 bp fragment encoding the SH2 domain of human Grb2 (amino acids 59–158), was a gift from Akhilesh Pandey (Johns Hopkins University, Baltimore, MD). The GST fusion protein was expressed in Escherichia coli and subsequently purified on glutathione Sepharose beads (purity >95%) following the procedure described by the manufacturer (Amersham Pharmacia Biotech). Mass-spectrometric analysis. Nanoscale LC-MS and LC-MS/MS was done with a quadrupole time-of-flight instrument (QSTAR-Pulsar; ABI-SCIEX, Toronto, Canada) essentially as described27. A linear gradient elution from 95% buffer A (H2O–acetic acid, 100:0.5 vol/vol) to 50% buffer B (H2O– acetonitrile–acetic acid, 20:80:0.5 vol/vol) in 120 min chromatographically separated the peptides. Protein identification was done with the Mascot software package (Matrix Science, London, UK) using the NCBI nonredundant protein database. Arginine-containing peptides (approximately half of all tryptic peptides) appeared in a labeled and unlabeled form separated by 6 Da and co-elute exactly, aiding quantification. High-resolution extracted ion chromatograms (±0.03 Da) for the labeled and unlabeled form were obtained, reducing background contribution. Peptide abundance ratios from the peak intensities of the extracted ion chromatograms were averaged to give protein abundance ratios. Peptide ratios were also verified by quantification directly on the basis of the mass spectra. Microscopy. The following antibodies were used: goat anti-EGFR (Santa Cruz Biotechnologies, Santa Cruz, CA), sheep anti-EGFR (Upstate Biotechnology), rabbit anti–histone H3 (Cell Signaling Technologies, Beverly, MA), fluorescein isothiocyanate (FITC)–conjugated mouse anti-keratin-18 (Sigma, Copenhagen, Denmark), CD59 (Abcam, Cambridge, UK), Shc (BD Transduction Laboratories, Lexington, KY), Alexa-488-conjugated anti-rabbit (Molecular Probes, Eugene, OR), and cyanin-3 (Cy3)- and Cy5conjugated anti-goat, -rabbit, -mouse, and -sheep antibodies (Jackson Immuno Labs, West Grove, PA). Serum-starved HeLa cells were fixed in 4% (vol/vol) formaldehyde for 10 min on ice, permeabilized where indicated with 0.1% (vol/vol) Triton X-100 for 20 min, and blocked for 30 min with 3% (vol/vol) BSA in PBS or, in the case of samples stained for histone H3, with 3% (wt/vol) skim milk powder–3% horse serum in PBS. Primary and secondary antibody incubations were 1 h each, and the coverslips were washed 4× with PBS after each incubation. Coverslips were mounted on glass slides and imaged using a Zeiss Axiovert 200M laser scanning confocal microscope 510 equipped with a META polychromatic multichannel detector. FRET between Cy3 (EGFR) and Cy5 (CD59) was measured as described for sensitized acceptor fluorescence28: Cy3 was excited using the 543 nm laser line of a He-Ne laser, and Cy5 emission was detected with the META system configured to measure wavelengths between 645 and 720 nm. Under conditions identical to those used for FRET

318


•

VOLUME 21

•

Note: Supplementary information is available on the Nature Biotechnology website.

Competing interests statement The authors declare that they have no competing financial interests. Received 12 November 2002; accepted 16 January 2003

1. Pandey, A. & Mann, M. Proteomics to study genes and genomes. Nature 405, 837–846 (2000). 2. Ho, Y. et al. Systematic identification of protein complexes in Saccharomyces cerevisiae by mass spectrometry. Nature 415, 180–183 (2002). 3. Gavin, A.C. et al. Functional organization of the yeast proteome by systematic analysis of protein complexes. Nature 415, 141–147 (2002). 4. Hunter, T. Signaling—2000 and beyond. Cell 100, 113–127 (2000). 5. Schlessinger, J. Cell signaling by receptor tyrosine kinases. Cell 103, 211–225 (2000). 6. Pawson, T. & Nash, P. Protein-protein interactions define specificity in signal transduction. Genes Dev. 14, 1027–1047 (2000). 7. Ong, S.E. et al. Stable isotope labeling by amino acids in cell culture, SILAC, as a simple and accurate approach to expression proteomics. Mol. Cell. Proteomics 1, 376–386 (2002). 8. Pandey, A. et al. Analysis of receptor signaling pathways by mass spectrometry: identification of Vav-2 as a substrate of the epidermal and platelet-derived growth factor receptors. Proc. Natl. Acad. Sci. USA 97, 179–184 (2000). 9. Moores, S.L. et al. Vav family proteins couple to diverse cell surface receptors. Mol. Cell. Biol. 20, 6364–6373 (2000). 10. Erickson, J.W., Cerione, R.A. & Hart, M.J. Identification of an actin cytoskeletal complex that includes IQGAP and the Cdc42 GTPase. J. Biol. Chem. 272, 24443–24447 (1997). 11. Kawashima, T. et al. MgcRacGAP is involved in the control of growth and differentiation of hematopoietic cells. Blood 96, 2116–2124 (2000). 12. Sotillos, S. & Campuzano, S. DRacGAP, a novel Drosophila gene, inhibits EGFR/Ras signalling in the developing imaginal wing disc. Development 127, 5427–5438 (2000). 13. Robinson, M.S. & Bonifacino, J.S. Adaptor-related proteins. Curr. Opin. Cell Biol. 13, 444–453 (2001). 14. Feng, G.S., Hui, C.C. & Pawson, T. SH2-containing phosphotyrosine phosphatase as a target of protein-tyrosine kinases. Science 259, 1607–1611 (1993). 15. Vogel, W., Lammers, R., Huang, J. & Ullrich, A. Activation of a phosphotyrosine phosphatase by tyrosine phosphorylation. Science 259, 1611–1614 (1993). 16. Thomas, D., Patterson, S.D. & Bradshaw, R.A. Src homologous and collagen (Shc) protein binds to F-actin and translocates to the cytoskeleton upon nerve growth factor stimulation in PC12 cells. J. Biol. Chem. 270, 28924–28931 (1995). 17. Leung, C.L., Green, K.J. & Liem, R.K. Plakins: a family of versatile cytolinker proteins. Trends Cell Biol. 12, 37–45 (2002). 18. Spike, C.A., Shaw, J.E. & Herman, R.K. Analysis of smu-1, a gene that regulates the alternative splicing of unc-52 pre-mRNA in Caenorhabditis elegans. Mol. Cell. Biol. 21, 4985–4995 (2001). 19. Di Benedetto, A.J., Klick Stoddard, J. & Glavan, B.J. Cloning and molecular characterization of a novel gene encoding a WD-repeat protein expressed in restricted areas of adult rat brain. Gene 271, 21–31 (2001). 20. Lin, S.Y. et al. Nuclear localization of EGF receptor and its potential new role as a transcription factor. Nat. Cell Biol. 3, 802–808 (2001). 21. Wells, A. & Marti, U. Signalling shortcuts: cell-surface receptors in the nucleus? Nat. Rev. Mol. Cell. Biol. 3, 697–702 (2002). 22. Cheung, P., Allis, C.D. & Sassone-Corsi, P. Signaling to chromatin through histone modifications. Cell 103, 263–271 (2000). 23. Korty, P.E., Brando, C. & Shevach, E.M. CD59 functions as a signal-transducing molecule for human T cell activation. J. Immunol. 146, 4092–4098 (1991). 24. Stefanova, I., Horejsi, V., Ansotegui, I.J., Knapp, W. & Stockinger, H. GPIanchored cell-surface molecules complexed to protein tyrosine kinases. Science 254, 1016–1019 (1991). 25. Murray, E.W. & Robbins, S.M. Antibody cross-linking of the glycosylphosphatidylinositol-linked protein CD59 on hematopoietic cells induces signaling pathways resembling activation by complement. J. Biol. Chem. 273, 25279–25284 (1998). 26. Pandey, A., Andersen, J.S. & Mann, M. Use of mass spectrometry to study signaling pathways. Sci. STKE 2000, PL1 (2000). 27. Rappsilber, J., Ryder, U., Lamond, A.I. & Mann, M. Large-scale proteomic analysis of the human spliceosome. Genome Res. 12, 1231–1245 (2002). 28. Kenworthy, A.K. Imaging protein-protein interactions using fluorescence resonance energy transfer microscopy. Methods 24, 289–296 (2001).

MARCH 2003

•

www.nature.com/naturebiotechnology