Plant Molecular Biology 50: 871–883, 2002. © 2002 Kluwer Academic Publishers. Printed in the Netherlands.
871
Imaging protein-protein interactions in living cells Mark A. Hink1 , Ton Bisseling1,∗ and Antonie J.W.G. Visser1,2 1 MicroSpectroscopy
Centre, Wageningen University, Dreijenlaan 3, 6703 HA Wageningen, The Netherlands (∗ author for correspondence; e-mail
[email protected]); 2 Department of Structural Biology, Faculty of Earth and Life Sciences, Vrije Universiteit, De Boelelaan 1087, 1081 HV Amsterdam, The Netherlands Received 5 November 2001; accepted in revised form 23 June 2002
Key words: cells, FCS, fluorescence spectroscopy, FRET, protein-protein interactions
Abstract The complex organization of plant cells makes it likely that the molecular behaviour of proteins in the test tube and the cell is different. For this reason, it is essential though a challenge to study proteins in their natural environment. Several innovative microspectroscopic approaches provide such possibilities, combining the high spatial resolution of microscopy with spectroscopic techniques to obtain information about the dynamical behaviour of molecules. Methods to visualize interaction can be based on FRET (fluorescence detected resonance energy transfer), for example in fluorescence lifetime imaging microscopy (FLIM). Another method is based on fluorescence correlation spectroscopy (FCS) by which the diffusion rate of single molecules can be determined, giving insight into whether a protein is part of a larger complex or not. Here, both FRET- and FCS-based approaches to study protein-protein interactions in vivo are reviewed.
Introduction Several elegant biochemical methods have been developed to study protein-protein interactions. In general, such biochemical studies are carried out in vitro and have provided valuable information about the properties of the studied molecules. However, to what extent these properties reflect their behaviour in living cells is not clear. The complex organization and the compartmentalization of plant cells make it probable that molecular behaviour in the test tube and in the cell are not identical and therefore it is essential to study molecules in their natural environment. Optical microscopy has been very useful to obtain information about the sub-cellular location of proteins. However, classical light microscopy, for example, cannot reveal whether proteins interact with one another. At best, optical microscopy can demonstrate that two proteins occur in the same region within a cell. It must be realised, however, that the spatial resolution of light microscopy (about 300 nm in a standard confocal microscope) is orders of magnitude larger than the average size of a protein molecule (diameter about
3 nm for a globular protein of 30 kDa). Therefore, it is unclear whether molecules observed in the same region in a light microscopic image interact or not. For example, proteins located in the nucleus may colocalize, but, of course, not all nuclear proteins interact with each other. So how can interactions be imaged in a living cell? The integration of fluorescence spectroscopy in light microscopy adds a new dimension to microscopy since in addition to spatial resolution now also information about the molecular behaviour of molecules can be obtained. Methods to visualize interaction can be based on FRET (fluorescence detected resonance energy transfer), for example in fluorescence lifetime imaging microscopy (FLIM) by which the fluorescence lifetime of a fluorescent dye as a function of intracellular space can be determined. Another method is fluorescence correlation spectroscopy (FCS) that allows determining the diffusion rate of single molecules, providing insight into whether a protein is part of a larger complex or not. Here, both FRET- and FCSbased approaches to study protein-protein interactions in vivo are described.
872 To study the behaviour of proteins with spectroscopic methods it is essential to label the molecules with a fluorescent group. The availability of genes encoding intrinsically fluorescent proteins has made it possible to genetically tag a protein of interest. In addition, smaller amino acid based tags are designed by which proteins can be rendered fluorescent. We will start this review with a short overview of different (genetic) methods by which proteins can be made fluorescent, then we will describe the microspectroscopic methods referred to above.
Fluorescent tags Imaging protein-protein interaction with microspectroscopic approaches requires that the proteins of interest be fluorescently labelled. When such studies are done in living cells it is, in principle, possible to introduce into the cell – by microinjection, for example – a purified protein to which a fluorescent tag has been added. However this requires specialized skills and is technically only possible in very few cell types, since in general plant cell walls form a large physical obstacle to microinjection. Therefore it is far more attractive to add fluorescent tags by genetic approaches. Intrinsic fluorescent proteins like GFP (green fluorescent protein) are especially suitable for this purpose since biologically functional translational fusions have been made in many cases. At present GFP or colour variants thereof are most often used to fluorescently tag proteins. GFP has been identified in the jellyfish Aequorea victoria; it is a 27 kDa protein that forms a barrel-like structure and in its centre a fluorophore is formed by three amino acids (Tsien, 1998). The sequence of the GFP gene of A. victoria has been optimized for plant codon usage and in this way a cryptic splice site has been eliminated (Haselhoff et al., 1997). By site-directed mutagenesis several colour variants of GFP have become available and their properties are summarized in Table 1. Generally speaking, GFP is a monomeric protein and therefore it is very suitable for investigation of protein-protein interactions. In contrast, the more recently identified red fluorescent protein (DsRed) forms aggregates and is therefore not yet suitable for such studies. DsRed has been recently cloned from a coral of the genus Discosoma (Matz et al., 1999). It has very attractive properties since it has an emission maximum around 600 nm by which it can be imaged without interference of autofluorescence and it has
a rather high quantum yield of fluorescence. However, the currently available DsRed proteins still have major drawbacks since they mature slowly and form oligomers (tetramers) (Baird et al., 2000). Although GFP is a rather small protein it still represents a rather bulky group in fluorescent chimeras and in some cases it affects the activity of its host protein. Therefore it will be a challenge to search for smaller tags that genetically can be introduced in proteins. A promising tag was designed by the group of Roger Tsien (Griffin et al., 1998). They designed a receptor domain of only 6 amino acids. It contains 4 cysteine residues. When introduced in a helix the spatial separation used allows the 4 thiol groups to form a receptor side on one side of the helix. This receptor side can bind organo-arsenicals. These are membrane-permeant (non-fluorescent) molecules that, upon binding to the thiol-based receptor side, first form a fluorescent group. This is a very promising approach to label proteins since it is probable that insertion of 6 amino acids in a helix will have less effect on the activity of the host protein than a fusion with GFP. The labelling of recombinant protein molecules with trivalent arsenic compounds has been used successfully in animal systems (Griffin et al., 1998), but we are not aware of its use in plant systems so far.
Fluorescence detected resonance energy transfer Interaction between two proteins can be visualized when they are tagged with appropriate fluorophores. When the fluorophores are in close contact to each other energy can be transferred by a process called FRET (Förster, 1948). This process is based on the phenomenon that excited-state energy from a donor to an acceptor molecule is transferred non-radiatively through space, as is depicted schematically in Figure 1A. FRET takes place if emission and excitation spectra of the fluorophore pair are overlapping and if the distance between the proteins is very small (in general < 7 nm). FRET is a very powerful method for obtaining distance information on macromolecular complexes. The efficiency of transfer (E) is defined by E=
R06 R06
+ R6
where R is the actual distance between the chromophores and R0 is the critical transfer distance (or Förster distance) at which 50% energy transfer takes place. R0 contains other factors such as the overlap
873
Figure 1. Concepts of non-radiative energy transfer. A. The two cylinders represent CFP (donor) and YFP (acceptor) and the arrows the transition dipoles of the donor and acceptor chromophoric groups located in the helix at the centre of the protein. Upon excitation of the donor molecule energy is transferred non-radiatively (edged arrow) to the acceptor that will fluoresce. Note that FRET occurs when the proteins are separated within a distance of 1.5 R0 (R0 is the critical transfer or Förster distance) and that FRET is absent when the distance is more than three times R0 where only emission of the donor molecule is observed. B. Absorbance and fluorescence emission spectra of CFP and YFP and the overlap integral (J) (shaded area) between the emission spectrum of CFP and the absorption spectrum of YFP. R0 is obtained from the following relationship: R0 = [8.785 · 10−5 · κ 2 · Q0D · J/n4 ]1/6 . Substituting for the orientation factor κ 2 = 2/3 (dynamical average limit), the donor fluorescence quantum yield (in absence of acceptor) Q0D = 0.6, J = 1.711 · 1015 M−1 cm−1 nm4 (determined) and the refractive index n = 1.4 gives R0 = 5.0 nm. When the dipoles are parallel (situation sketched in panel A) κ 2 = 4 and R0 = 6.7 nm. When the dipoles are anti-parallel κ 2 = 1 and R0 = 5.4 nm. C. A plot of the efficiency of energy transfer E against R/R0 . Since E scales inversely to the sixth power of the relative distance between acceptor and donor molecule, the dynamic range for E is for a distance 0.5R0 < R < 1.5R0 . At R = R0 the efficiency of energy transfer is 50%. When R = 3R0 , E = 0.0014 implying virtually no FRET.
874 Table 1. Adapted from Tsien (1998). Fluorescent protein
Amino acid substitutions
Absorbance/emission
GFP (green fluorescent protein) EGFP (enhanced GFP) CFP (cyan fluorescent protein)
– Ser65Thr, Phe64Leu Phe64Leu, Ser65Thr Tyr66Trp, Asn146Ile Met153Thr, Val163Ala Asn212Lys Ser65Gly, Ser72Ala Thr203Tyr Ser65Gly, Val68Leu, Gln69Lys, Ser72Ala Thr203Tyr Tyr66His, Tyr145 Ser –
396, 488/505 nm 488/509 nm 434, 452/476, 505 nm
YFP (yellow fluorescent protein) (less pH-sensitive version of YFP) (Miyawaki) BFP (blue fluorescent protein) DsRed
of the fluorescence spectrum of the donor with the absorption spectrum of the acceptor and the relative orientation of donor and acceptor molecules (Figure 1B). The critical transfer distance R0 scales with the inversed sixth power of the distance between donor and acceptor molecule. This implicates that when the distance increases from R0 to 2*R0, E would decrease from 50% to 1.5%. For this reason FRET is such a powerful tool to determine if molecules are close to each other (reviewed, among others, by Stryer, 1978; Herman, 1989; Wu and Brand, 1994; Clegg, 1996; Bastiaens and Jovin, 1998; Selvin, 2000; Wouters et al., 2001). Further, it also demonstrates the limitations of FRET in visualizing protein-protein interactions. The maximum distance over FRET can take place is ca. 7 nm. Hence, interactions between proteins in large complexes or binding of molecules to large transmembrane proteins might be difficult to observe. Fusion constructs of proteins of interest and variants of fluorescent proteins can be genetically prepared and intracellular interactions can be sensed through FRET. Suitable donor-acceptor couples are cyan fluorescent protein (CFP) as donor and yellow fluorescent protein (YFP) as acceptor. When monomeric variants of DsRed become available, enhanced GFP and DsRed will be an ideal FRET pair. So how would FRET be visualized? Interaction of a protein pair that results in FRET causes quenching of the donor fluorescence and enhanced acceptor emission. Therefore excitation of the donor will result in emission of the acceptor if both molecules are in close proximity of each other. In case the protein
514/527 nm 516/529 nm
434, 452/476, 505 nm 558/583 nm
pair is present in constant stoichiometry, FRET can be determined by measuring the ratio of intensities of acceptor and donor emissions. For example, the calcium sensor cameleon has a fixed stoichiometry as donor and acceptor proteins are fused to the same calmodulin module (Miyawaki et al., 1997, 1999). The conformation of calmodulin changes upon calcium binding in such a way that donor and acceptor molecules come closer. The increase in FRET efficiency becomes evident in an increase in the ratio between acceptor and donor fluorescence intensities. However, in most cases donor and acceptor proteins are not covalently linked. Then it is possible that the stoichiometry is not constant in each position of the cell. As a consequence ratio measurements are not suitable to determine FRET and therefore other methods have to be used to reliably measure the FRET efficiency. Methods that do not depend on the local concentration of both the acceptor and the donor molecules are based upon a reduction of the quantum yield (QD ) and/or lifetime (τ D ) of donor fluorescence: QD τD E = 1− 0 = 1− 0 QD τD where the superscript 0 denotes the quantum yield or lifetime of the fluorescence of the donor without FRET (i.e. without acceptor). Detection of the quantum yield in a microscopic image is not straightforward, since one should have knowledge on the fraction of donor molecules having an acceptor in the vicinity as compared to those having no acceptor. Xia et al. (2001) have summarized the protocols required
875
Figure 2. Continued.
Figure 2. Information from fluorescence anisotropy decay, r(t). A. Experimental (dots) and fitted (solid line) fluorescence anisotropy decay curves of 56 nM EGFP in aqueous buffer pH 8 and at 20 ◦ C. The experiment was obtained with the technique of time-correlated single photon counting. Pulsed excitation was at 480 nm and fluorescence photons passed a band-pass filter transmitting between 500 and 550 nm. Analysis yielded a single rotational correlation time φ = 14 ns. The initial anisotropy (β = 0.385) approaches the limiting value of 0.4. B. Simulated fluorescence anisotropy decay curves where the sole depolarization mechanism is assumed to be homo-FRET. The equation is: r(t) = 0.1 · [(3−z) · exp(−2kT t) + z + 1], with z = 3 cos2 θ where kT is the transfer rate constant and θ is the angle between the transition dipoles of both GFPs. C. Two situations are presented where the separation between both GFPs is 7.0 nm and the transfer rate constant kT has been set equal to 0.2 (ns)−1 . For curve 1 the transition dipoles are parallel (θ = 0◦ ) not leading to depolarization by homo-FRET. Curve 2 describes the anisotropy decay for θ = 45◦ leading to distinct depolarization. The latter case has been found for thymidine kinase-GFP dimers in living cells (Gautier et al., 2001; see also this reference for more details).
to obtain FRET efficiencies both from quenching of donor fluorescence and for sensitized acceptor emission. Despite the problems it turned out to be possible to image protein-protein interactions in Arabidopsis, by measuring the ratio of donor and acceptor emission. Más et al. (2000) showed that the GFP and DsRed fusion proteins of Arabidopsis photoreceptors phyB and cry2 interact in nuclear speckles that are formed in a light-dependent fashion. In these experiments the ratio changes observed are most likely due to FRET since the fluorescence intensity of the donor increased after acceptor photobleaching, inducing irreversible destruction of the acceptor molecules. Although FRET studies can be performed by measuring the ratio donor and acceptor fluorescence, detection of FRET in a microscope using the shortening of donor fluorescence lifetimes is a more direct method and will be discussed in the next section.
Fluorescence lifetime imaging microscopy FRET measurements in a microscopic object can be conveniently carried out with fluorescence lifetime imaging microscopy (FLIM) (Gadella et al., 1999;
876
Figure 3. Principles of FCS. A. A focused laser beam creates an excitation volume (solid-lined cone) wherein fluorescent molecules will be excited, resulting in the emission of photon bursts. B. Molecules will, due to the Brownian motion, move through the confocal detection element (grey), causing a fluctuation of the detected fluorescence intensity over time, I(t). C. The intensity is correlated over lag time τ , according to: )> G(τ ) =
