|Home | About | Journals | Submit | Contact Us | Français|
The specificity of biological regulatory mechanisms relies on selective interactions between different proteins in different cell types and in response to different extracellular signals. We describe a bimolecular fluorescence complementation (BiFC) approach for the simultaneous visualization of multiple protein interactions in the same cell. This approach is based on complementation between fragments of fluorescent proteins with different spectral characteristics. We have identified twelve new bimolecular fluorescent complexes that correspond to seven different spectral classes. Bimolecular complex formation between fragments of different fluorescent proteins did not differentially affect the dimerization efficiency or subcellular sites of interactions between the bZIP domains of Fos and Jun. Multicolor BiFC enables visualization of interactions between different proteins in the same cell and comparison of the efficiencies of complex formation among alternative interaction partners.
Networks of protein interactions mediate cellular responses to environmental stimuli and direct the execution of developmental programs. Each protein typically has a large number of alternative interaction partners, and the selectivity of these interactions determines the developmental potential of the cell and its responses to extracellular stimuli. We recently described a new approach for the visualization of protein interactions in living cells designated bimolecular fluorescence complementation (BiFC) analysis 1. The BiFC approach is based on the formation of a fluorescent complex by fragments of the yellow fluorescent protein (YFP) brought together by the association of two interaction partners fused to the fragments. This approach enables visualization of the subcellular sites of protein interactions under conditions that closely reflect the normal physiological environment.
Molecular engineering of the green fluorescent protein (GFP) has produced several variants with altered spectral characteristics 2. These variants allow simultaneous visualization of the distributions of multiple proteins in living cells. Moreover, fluorescence resonance energy transfer (FRET) between different variants enables analysis of interactions between individual pairs of proteins in living cells 3, 4. Thus far, it has not been possible to visualize multiple interactions in the same cell.
Selected fragments of many proteins can associate to produce functional bimolecular complexes. Such bimolecular complementation provides a convenient approach for detection of protein interactions in cells if the protein fragments can associate only when they are brought together by interaction partners fused to the fragments 1, 5-9. The unique characteristic of the BiFC approach is that the bright intrinsic fluorescence of the bimolecular complex allows direct visualization of the complex in living mammalian cells 1. Moreover, the large number of GFP variants with distinct spectral characteristics provided the potential for parallel analysis of multiple protein interactions in the same cell. In the present study, we have realized the promise of multicolor BiFC analysis by characterizing twelve new bimolecular fluorescent complexes, and we have used these complexes to compare the dimerization selectivity and subcellular sites of interactions among basic region leucine zipper (bZIP) family proteins.
The spectral characteristics of bimolecular fluorescent complexes formed by fragments of YFP were virtually identical to those of intact YFP 1. We reasoned that fragments of other GFP variants might support bimolecular fluorescent complex formation, and that such complexes might have distinct spectral characteristics. To identify such complexes, we investigated fluorescence complementation using the corresponding fragments of the enhanced GFP and cyan fluorescent protein (CFP) fused to the bZIP domains of Fos and Jun (bFos and bJun) (Fig. 1A). Each pair of fusion proteins was expressed in mammalian cells and the cells were examined by fluorescence microscopy (Fig. 1B-D). No complementation was detected when fragments of GFP (GN155 and GC155) fused to bFos and bJun were expressed in mammalian cells. However, fragments of CFP (CN155 and CC155) exhibited fluorescence complementation when fused to bFos and bJun. (Fig. 1D). All of the fusion proteins were expressed at comparable levels as determined by Western analysis (Supplementary Fig. 1A online).
To examine the selectivity of bimolecular complex formation, we tested fluorescence complementation between all 9 combinations of fragments (Supplementary Fig. 2 online). YN155 exhibited fluorescence complementation with YC155 and CC155, whereas CN155 exhibited fluorescence complementation only with CC155 when fused to bFos and bJun (Fig. 1B-D). Significantly, the fluorescence spectrum of cells expressing YN155 and CC155 fusions was distinct from those of cells expressing either YN155 and YC155 or CN155 and CC155. GN155 and GC155 did not exhibit detectable fluorescence complementation with any of the other fragments. YC155 and CC155 differ from GC155 by single amino acid residues whereas YN155 and CN155 differ from GN155 by four and three amino acid residues respectively (Fig. 1A). These amino acid substitutions determined the selectivity of bimolecular fluorescence complementation among these fragments.
We used a genetic screen in E. coli 1 to identify a second pair of YFP fragments (YN173 and YC173) that exhibit bimolecular fluorescence complementation when fused to bFos and bJun. We examined fluorescence complementation between these fragments and the corresponding fragments of GFP, CFP and the enhanced blue fluorescent protein (BFP) fused to bFos and bJun. The sequences of the C-terminal fragments of GFP, CFP and BFP are identical (Fig. 1A), and thus only YC173 and GC173 were tested. YC173 exhibited fluorescence complementation with YN173, GN173 and CN173 when fused to bFos and bJun (Fig. 1E-G), whereas GC173 did not exhibit detectable fluorescence complementation with any of the fragments tested. The two positions of fragmentation (155 and 173) enabled complementation between distinct combinations of fluorescent protein fragments (Supplementary Fig. 2 online). Thus, a small number of amino acid substitutions can influence the positions where proteins must be fragmented to support bimolecular complementation.
GFP is a β barrel structure containing 11 strands surrounding a central α helix10.The two sites of fragmentation in YFP that support BiFC are separated by one strand of the β barrel that surrounds the fluorophore10. We examined fluorescence complementation between all 24 combinations of fragments truncated at these positions fused to bFos and bJun (Supplementary Fig. 2 online). None of the N-terminal fragments truncated at residue 155 exhibited fluorescence complementation with C-terminal fragments truncated at residue 173. In contrast, all of the N-terminal fragments truncated at residue 173 (YN173, GN173 and CN173) formed bimolecular fluorescent complexes with CC155 and YC155 (Fig. 1H, I, J, data not shown). The corresponding N-terminal fragment of BFP (BN173) also exhibited fluorescence complementation with CC155 (Fig. 1K). Duplication of the segment separating the points of truncation therefore facilitated bimolecular fluorescence complementation. This duplication had no detectable effect on the spectra of the bimolecular complexes (Fig. 2).
To establish whether bimolecular fluorescence complementation between fragments of different fluorescent proteins fused to bFos and bJun required the leucine zipper dimerization interface, we examined complementation by bFos fusions in which the carboxy-terminal half of the leucine zipper was deleted (bFosΔZIPYC155 and bFosΔZIPCC155). We compared the complementation efficiencies of the wild type and mutant fusions with bJun fused to fragments of different fluorescent proteins (Fig. 3, Supplementary Table 1 online). Mutation of the leucine zipper resulted in a more than 10-fold reduction in the efficiencies of fluorescence complementation between all fragments of fluorescent proteins tested. Cells expressing fragments of different fluorescent proteins fused to wild type bFos and bJun produced different intensities of fluorescence emissions at different times following transfection. Cells expressing the same fragments, but with a deletion in the leucine zipper of Fos, produced either undetectable or lower fluorescence emissions at all times following transfection (Supplementary Table 1 online). Similar results were obtained when the fluorescence emissions of the cell populations were measured using fluorescence spectroscopy (data not shown). The mutant proteins were expressed at the same levels and were localized to the same subnuclear sites as the wild type proteins. Thus, efficient fluorescence complementation between all fragments of fluorescent proteins examined required a specific interaction interface.
Multicolor BiFC allows comparison of the efficiencies of complex formation between alternative interaction partners, providing that the fragments do not influence the selectivity of protein interactions. To compare the effects of fragments of different fluorescent proteins on dimerization efficiency, we examined the competition between proteins in which bJun was fused to fragments of different fluorescent proteins for dimerization with bFosYC173 in vitro (Fig. 4). The ratio of bimolecular fluorescent complexes formed by the alternative interaction partners was calculated by fitting the spectrum of the mixture to the weighted sum of the spectra of the two complexes. This ratio exhibited a perfect correspondence with the ratio of bJun fusion proteins that was added to each reaction. Thus, the bZIP domains of Fos and Jun have identical dimerization efficiencies when fused to fragments of different fluorescent proteins, suggesting that the fusions do not have differential effects on interactions between bFos and bJun in vitro.
The spectral differences between bimolecular complexes formed by fragments of different fluorescent proteins enable comparison of the subcellular sites of interactions between different proteins in the same cell providing that the fragments do not differentially affect complex localization. To compare the subcellular locations of bFos-bJun heterodimers fused to fragments of different fluorescent proteins, we co-expressed bJunCN173 and bJunYN173 with bFosYC173 in COS-1 cells and imaged the cells using filters optimized for the detection of CN173-YC173 and YN173-YC173 complexes respectively (Fig. 5A). Both complexes were localized preferentially to the nucleoli, and exhibited perfect co-localization. Similar results were obtained when bJunCN173 and bJunYN173 were co-expressed with bFosYC155 or with bFosCC155. Although the efficiency of nucleolar localization of the complexes varied between different cells in the population, the efficiencies of nucleolar localization of complexes containing fragments of different fluorescent proteins were identical in individual cells. Thus, fragments of different fluorescent proteins did not have differential effects on the subcellular sites of interactions between bFos and bJun.
To visualize the subcellular sites of interactions between different proteins in the same cell, we co-expressed bJunCN173 and JunYN155 with bFosYC155. bJunCN173-bFosYC155 exhibited nucleolar localization, but JunYN155-bFosYC155 was localized to the nucleoplasm (Fig. 5B). The two complexes exhibited distinct distributions in the same nucleus, indicating that regions outside the bZIP domain of Jun affected the subcellular localization of the complex. Similar results were obtained when bJunCN173 and JunYN155 were co-expressed with bFosCC155. Thus, multicolor BiFC can be used to compare the subcellular sites of interactions between different proteins in the same cell.
The bZIP domains of Fos, Jun and ATF2 can interact with each other in all pairwise combinations in vitro, and activate different genes in cells 11-14. To investigate the competition between alternative interaction partners in cells, we examined the relative efficiencies of bJun interactions with bFos and bATF2. We expressed bJunCC155 with bFosCN173 and bATF2YN155 separately as well as together, and compared the ratio between the fluorescence emissions of the complexes (Fig. 6A, Supplementary Table 2 online). Cells expressing the two pairs of proteins separately exhibited cyan and yellow fluorescence respectively. Co-expression of a limiting amount of bJunCC155 with bFosCN173 and bATF2YN155 together resulted in cyan fluorescence that was comparable to that observed in cells transfected with bJunCC155 and bFosCN173, but yellow fluorescence that was 100-fold lower than that of cells transfected with the same amounts of bJunCC155 and bATF2YN155 (Supplementary Table 2). The ratio between yellow and cyan fluorescence was therefore comparable for cells expressing all three proteins and those expressing only bJunCC155 and bFosCN173 (Fig. 6A). Similar results were obtained when bJunCC155 was expressed with bFosCN155 and bATF2YN155 (data not shown). When excess bJunCC155 was expressed with bFosCN173 and bATF2YN155, both cyan and yellow fluorescence was observed in the cells. It is therefore likely that the inhibition of bimolecular fluorescence complementation between bATF2YN155 and bJunCC155 in the presence of either bFosCN155 or bFosCN173 was caused by a higher efficiency of complex formation by bJunCC155 with either bFosCN155 or bFosCN173 than with bATF2YN155.
To confirm that the identity of the fragments fused to bJun, bFos and bATF2 did not affect the preference for bimolecular fluorescent complex formation, we exchanged the fragments between bFos and bATF2 (Fig. 6B). Cells expressing a limiting amount of bJunCC155 with bFosYN155 and bATF2CN173 produced yellow fluorescence comparable to that of cells transfected with bJunCC155 and bFosYN155, but cyan fluorescence that was 10-fold lower than that of cells transfected with the same amounts of bJunCC155 and bATF2CN173 (Supplementary Table 2). Again, the ratio between yellow and cyan fluorescence for cells expressing all three proteins was similar to that of cells expressing only bJunCC155 and bFosYN173 (Fig. 6B). The fluorescent protein fragments were fused to the same positions in bJun and bATF2 relative to the leucine zipper, and the same linker sequences were used for all fusions. The competing proteins were expressed at comparable levels (Supplementary Fig. 1B online). Consequently, it is likely that the higher efficiency of complementation between bJun and bFos fusions reflected preferential heterodimerization of bJun with bFos than with bATF2 in cells.
To examine the relative efficiencies of bFos interactions with bJun and bATF2 in cells, we expressed limiting bFosCC155 with bJunCN173 and bATF2YN155. The cells exhibited cyan fluorescence comparable to cells transfected with bFosCC155 and bJunCN173, but markedly lower yellow fluorescence than cells transfected with the same amounts of bFosCC155 and bATF2YN155 (data not shown). A similar inhibition of bFos-bATF2 and bJun-bATF2 heterodimer formation was observed in cells transfected using a limiting concentration of bATF2YC155 with bFosYN155 and bJunCN173 (data not shown). The higher efficiency of bFos-bJun heterodimer formation therefore did not require bimolecular complex formation since bFosYN155-bJunCN173 heterodimers could not form a bimolecular fluorescent complex. Thus, the bZIP domains of Fos and Jun favor dimerization with each other over dimerization with the bZIP domain of ATF2 in cells.
To examine the relative efficiencies of Fos-Jun heterodimerization and Jun homodimerization in cells, we expressed bJunCC155 with bFosYN155 and bJunCN173 (Fig. 6C). The cells produced on average 60% of the cyan fluorescence and 35% of the yellow fluorescence produced by the control cells. The fluorescence intensities of individual cells were normally distributed and exhibited a moderate (r=0.83) correlation between cyan and yellow fluorescence. Thus, Fos-Jun heterodimers and Jun homodimers can co-exist in cells, consistent with the similar thermodynamic stabilities of these dimers in vitro 15.
The multicolor BiFC assay provides a unique approach for the simultaneous visualization of multiple protein interactions in a living cell. Although we describe only cases where two interactions were compared in parallel, the number of interactions that can be visualized simultaneously is limited in principle only by the number of spectrally distinct bimolecular fluorescent complexes. In practice, the number of complexes that can be visualized simultaneously depends on the spectral resolution of the imaging system. In typical cases where two different complexes need to be resolved, the broad range of spectral characteristics among the bimolecular complexes identified here enables visualization of the complexes without interference from spectral overlap using standard fluorescence microscopy.
Fragments of different fluorescent proteins did not differentially affect bZIP protein dimerization or the localization of those dimers. These fragments therefore enable comparison of the efficiencies and subcellular locations of complex formation with alternative interaction partners in the same cell. It remains possible that these fragments have distinct effects on interactions between other proteins or the localization of those complexes. It is therefore necessary to compare the efficiencies of complex formation by proteins fused to these fragments, and to establish the effects of the fragments on complex localization in order to use multicolor BiFC for comparison of interactions among other proteins. In addition, since formation of the bimolecular fluorescent complex is essentially irreversible, the efficiency of fluorescence complementation represents the efficiency of the association between the interaction partners at the time of complex formation, and does not reflect subsequent shifts in the equilibrium among alternative interaction partners 1.
The systematic analysis of complementation between fragments of different fluorescent proteins provides several insights into protein complementation. Complementation between fragments of different fluorescent proteins was selective, and single amino acid substitutions had greater than 100-fold effects on the efficiency of fluorescence complementation. These same substitutions had little effect on the quantum yields of the intact proteins 2, and some were located at a distance from the fluorophore. Substitutions that resulted in different efficiencies of fluorescence complementation had only small effects on the kinetics of fluorophore maturation (data not shown). It is therefore likely that these amino acid substitutions influence the efficiency of bimolecular complex formation.
Two different positions were identified where fragmentation of fluorescent proteins enabled bimolecular complementation. These positions are within loops at opposite ends of the β barrel structure. The bFos-bJun interaction and the linker sequences used in these experiments therefore provide sufficient flexibility for the fragments to associate with their N-termini separated by a distance of either 10 or 50 Å. We have used the multicolor BiFC assay to investigate interactions among several structurally unrelated protein families. The results from these experiments confirm that the interaction partners do not need to juxtapose the fragments in a specific orientation, providing that the linkers that tether the fragments have sufficient flexibility to allow the fragments to associate with each other. Multicolor BiFC therefore provides a general approach for the analysis of complex formation among alternative interaction partners in living cells.
The β barrel structure that surrounds the fluorophore in green fluorescent proteins is composed of three segments of contiguous peptide sequence, each of which forms an anti-parallel β-sheet that together form the β barrel10. Both of the truncations that allow fluorescence complementation interrupt one of these β-sheet segments. Thus, all of the fragments that allow fluorescence complementation contain an incomplete β-sheet. Fragments that were truncated at the junctions between the three β-sheet segments did not exhibit fluorescence complementation 1. Other structurally unrelated protein fragments that exhibit complementation also contain incomplete domains that are unlikely to fold in the absence of the complementary fragments 6, 7, 16-18. It is therefore likely that bimolecular complementation by many protein fragments requires at least part of the fragments to be unfolded, which may facilitate their association.
Comparison of the efficiencies of interactions between alternative interaction partners using multicolor BiFC requires that the fusion proteins meet several criteria. First, fragments of different fluorescent proteins must not differentially affect interactions between the proteins. In the case of bFos-bJun heterodimers, identical efficiencies of complex formation were observed between proteins fused to fragments of different fluorescent proteins. Second, the efficiencies of complementation between the fragments must be equivalent once they have been brought together by interactions between the alternative interaction partners. In the case of bJun-bFos and bJun-bATF2 heterodimers fused to the same fragments, the fluorescence emissions of bimolecular complexes formed separately were comparable, consistent with the structural similarity between these complexes. Finally, the alternative interaction partners must be expressed at similar levels and be localized to their normal subcellular compartments. The bZIP domains of Fos, Jun and ATF2 were expressed at comparable levels and were localized to the nucleoli. When these criteria are met, multicolor BiFC enables comparison of the efficiencies and subcellular sites of complex formation between alternative interaction partners in living cells.
The sequences encoding amino acid residues 1-154, 155-238, 1-172, and 173-238 of the enhanced YFP, GFP, CFP and BFP (Clontech, Palo Alto, CA) were fused downstream of sequences encoding Fos118-210 (bFos) and Jun257-318 (bJun) using linker sequences encoding RPACKIPNDLKQKVMNH and RSIAT respectively. The chimeric coding regions were cloned into pFLAG-CMV2 (Sigma, St. Louis, MO) and pHA-CMV (Clontech, Palo Alto, CA) vectors for expression in mammalian cells and into pDS56 11 for expression in E. coli. As a control, residues 179-193 in the leucine zipper of Fos were deleted to produce bFosΔZIP. Plasmids encoding JunYN155, bATF2YN155, and bATF2YC155 were described previously 1.
COS-1 cells were transfected with plasmids encoding the fusion proteins indicated. Transfected cells were incubated at 37°C for 24 h and then switched to 30°C for 0-24 h to promote fluorophore maturation. To image the fluorescence emissions of cells expressing different combinations of fusion proteins, we used filter pairs centered at 436 nm excitation and 470 nm emission (C) or 500 nm excitation and 535 nm emission (Y) together with dichroic mirrors with transmission windows at 450-490 nm and 520-590 nm. The fluorescence intensities of individual cells were quantified using automated feature recognition software (Compix). The signal in an area of the field containing no cells was used as background and subtracted from all values. There was less than 2% overlap between the signals from bimolecular fluorescent complexes formed by YN173 and CN173 with either YC155 or CC155 using these filtes. Thus, no correction for cross-talk was necessary. To measure the spectra of bimolecular fluorescent complexes, the cells were washed and resuspended in PBS. All spectra were corrected for background signal produced primarily by scatter by subtracting the spectrum of untransfected cells. The level of expression of each fusion protein was quantified by Western blotting using antibodies directed against the FLAG and HA epitopes (Sigma, St. Louis, MO).
Chimeric fusion proteins were purified from E. coli using nickel chelate affinity chromatography in the presence of 6 M guanidine as described 1. The proteins indicated in each experiment were heated to 95 °C for 5 min and diluted to a final concentration of 30 μg/ml in buffer (50 mM NaPO4, pH 8.0, 150 mM NaCl, 5% glycerol (vol/vol), 0.1 mg/ml BSA and 1mM DTT) at room temperature. Excitation and emission spectra were collected after no further change in fluorescence was observed. The fluorescence intensities of the bimolecular fluorescent complexes were comparable to those observed for the intact fluorescent proteins. The spectra were corrected for scatter by subtracting the spectrum of the buffer alone.
Supplementary Figure 1. Comparison of expression of bFos, bJun and bATF2 fused to fragments of different fluorescent proteins. (A) Plasmids encoding the fragments indicated fused to bJun and bFos respectively were co-transfected into cells. The YN155 used in this experiment (but not in others presented in this manuscript) was fused to Jun225-334, resulting in a larger fusion protein (lane 1). (B) Plasmids encoding the proteins indicated above each lane were co-transfected into cells. Cell lysates were analyzed by Western blotting using antibodies directed against the FLAG and HA epitopes.
Supplementary Figure 2. Selectivity of bimolecular fluorescence complementation between fragments of fluorescent proteins fused to bFos and bJun. The excitation/emission maxima of all complexes that exhibited detectable fluorescence complementation are shown. None of the fragments exhibited detectable fluorescence alone. Cells with no detectable complementation (indicated by −) produced less than 1% of the fluorescence emissions of cells expressing YN155 and YC155 or CN155 and CC155 at every wavelength tested.
Supplementary Table 1. Effects of a deletion in the leucine zipper of Fos on complementation between fragments of different fluorescent proteins fused to bFos and bJun. The relative fluorescence intensities of cells transfected with the expression vectors indicated were evaluated by fluorescence microscopy. For a calibration of the qualitative scale, please compare with the quantitative data in Fig. 3.
Supplementary Table 2. Selectivity of dimerization among the bZIP domains of bFos, bJun and bATF2. The fluorescence intensities of cells transfected with plasmids expressing the proteins indicated were measured using 436/470 nm (C) or 500/530 nm (Y) filters. The average fluorescence intensities and the average of the ratios for all cells examined are shown.
We thank members of the Kerppola laboratory for helpful discussions.