Determination of in Vivo Dissociation Constant, K_D, of Cdc42-Effector Complexes in Live Mammalian Cells Using Single Wavelength Fluorescence Cross-correlation Spectroscopy^,

Abstract

The RhoGTPase Cdc42 coordinates cell morphogenesis, cell cycle, and cell polarity decisions downstream of membrane-bound receptors through distinct effector pathways. Cdc42-effector protein interactions represent important elements of cell signaling pathways that regulate cell biology in systems as diverse as yeast and humans. To derive mechanistic insights into cell signaling pathways, it is vital that we generate quantitative data from in vivo systems. We need to be able to measure parameters such as protein concentrations, rates of diffusion, and dissociation constants (K_D) of protein-protein interactions in vivo. Here we show how single wavelength fluorescence cross-correlation spectroscopy in combination with Förster resonance energy transfer analysis can be used to determine K_D of Cdc42-effector interactions in live mammalian cells. Constructs encoding green fluorescent protein or monomeric red fluorescent protein fusion proteins of Cdc42, an effector domain (CRIB), and two effectors, neural Wiskott-Aldrich syndrome protein (N-WASP) and insulin receptor substrate protein (IRSp53), were expressed as pairs in Chinese hamster ovary cells, and concentrations of free protein as well as complexed protein were determined. The measured K_D for Cdc42V12-N-WASP, Cdc42V12-CRIB, and Cdc42V12-IRSp53 was 27, 250, and 391 nm, respectively. The determination of K_D for Cdc42-effector interactions opens the way to describe cell signaling pathways quantitatively in vivo in mammalian cells.

Over the last 2 decades, we have been successful in describing a myriad of cell signaling pathways that regulate the biology of cells. These pathways are made of elements incorporating protein-protein, protein-lipid and protein-ligand interactions. With the advent of GFP2 (1, 2) and its variants (3), it is now possible to genetically encode fluorescent probes into any protein of interest. GFP fusion proteins can be used in live cells giving spatial and temporal resolution to cell signaling pathways (4). To gain mechanistic insights into cellular processes, it is crucial that we measure quantitative parameters to describe cell signaling. In this study, we present an approach based on fluorescence cross-correlation spectroscopy (FCCS) (5, 6) and Förster resonance energy transfer (FRET) to determine quantitative parameters of cell signaling pathways, including the determination of the K_D for Cdc42-effector interactions in live CHO-K-1 (hereafter referred to as CHO) mammalian cells.

The RhoGTPase Cdc42 (7, 8) regulates pathways that coordinate cell cycle, morphogenesis, and polarity. Cdc42 is a molecular switch that cycles between an inactive (GDP-bound) and active (GTP-bound) state. The V12 Cdc42 point mutation freezes the protein in an activated GTP-bound form, which binds effectors strongly. In contrast, Cdc42N17 is a dominant negative protein that is GDP-bound and interacts with effectors weakly if at all (9). A major Cdc42 binding site/domain in effector proteins is known as Cdc42- and Rac-interacting binding region (CRIB)3 and was originally found in activated Cdc42 kinase, p21 activated kinase (PAK), and neural Wiskott-Aldrich syndrome protein (N-WASP) (10). The inverse Bin-amphiphysins-Rvs domain adaptor protein IRSp53 is also an effector but binds Cdc42 through a partial CRIB domain (11, 12). Cdc42 interaction with its effectors has two main consequences, which are not mutually exclusive: (i) unfolding of effector to expose the active site and (ii) relocalization of effector to membrane compartments. Thus Cdc42-effector interactions serve as a good model for cell signaling as a whole.

Fluorescence correlation spectroscopy and FCCS measure fluctuations in fluorescence of a small number of molecules as they pass through a defined confocal volume, respectively (13, 14, 15). Since the number of molecules in the confocal volume and the confocal volume itself can be determined, concentrations of protein can be measured by fluorescence correlation spectroscopy. Single wavelength fluorescence cross-correlation spectroscopy (SW-FCCS) is an FCCS variant in which excitation of two or more probes is achieved by single wavelength one-photon excitation. To date SW-FCCS has been used successfully to follow receptors and receptor-ligand interactions in vitro and in vivo (6, 16, 17).

In the present analysis, we take a two-step approach to determining the K_D of Cdc42 binding to CRIB (domain of PAK), N-WASP, and IRSp53. First, we show that the proteins under investigation are indeed interacting with each other directly in vivo by FRET analysis. Here we use acceptor photobleaching (AP)-FRET as well as changes in lifetime (through fluorescence lifetime imaging microscopy (FLIM)) as indicators of FRET. Second, we use SW-FCCS to determine the K_D of Cdc42 interacting with its effectors by measuring the concentration of free protein versus complexed protein. Thus, the combined use of FRET and FCCS allows quantitative analysis of cell signaling pathways in vivo.

MATERIALS AND METHODS

Plasmid Preparation—GFP-N-WASP4 was a kind gift from Dr. Silvia Lommel (Helmoholtz Centre for Infection Research, Germany). mRFP-IRSp53 was made by subcloning IRSp53 from HA-IRSp53 into mRFP-pXJ40 vector between the BamHI and NotI restriction site (12). mRFP-N-WASP was made by subcloning the full-length N-WASP in mRFP-pXJ40 vector between the HindIII and NotI restriction site. mRFP-pXJ40 was created by inserting the mRFP between the EcoRI and BamHI restriction sites in pXJ40 vector. mRFP is a kind gift from Prof. R. Tsien. GFP-Cdc42V12, GFP-Cdc42N17, and GFP-CRIB were generated by cloning Cdc42V12, Cdc42N17 and CRIB into pXJ40-GFP vector. mRFP-GFP tandem fusion was made by inserting GFP between BamHI and NotI in the mRFP-pXJ40 vector. mRFP-Cdc42N17 was created by subcloning the Cdc42N17 between the BamHI and BglII restriction sites in pXJ40 vector. All of the restriction enzymes were obtained from NEB Biolabs.

Tissue Culture and Transfection— CHO cells were obtained from ATCC (Manassas, VA) and grown in a 75-cm² tissue culture flask up to 90% confluence in the complete growth medium, in 1× F-12 nutrient mixture (Kaighn's modification) medium containing 10% fetal bovine serum and 1% antibiotics (penicillin and streptomycin). All of the tissue culture reagents were obtained from Invitrogen (Singapore). Cells were seeded in a 6-well tissue culture plate containing the 30-mm/18 × 18-mm prewashed and sterilized coverglass for FCCS and FRET, respectively, at 1.5 × 10⁵ cells/well 1 day before the actual experiment was done. For FRET measurement, samples were made by transient transfection using Fugene 6 reagent (Roche Applied Bioscience) as per the manufacturer's protocol with the respective pair of plasmids (0.5–1 μg) in 6-well plates with necessary controls. 24 h after transfection, cells were washed with 1× phosphate-buffered saline three times and fixed with 4% para-formaldehyde (Sigma) for 15 min; further cells were washed with 1× phosphate-buffered saline containing 10 mm glycine. The fixed coverslips containing cells were mounted on coverslides using Hydromount (National Diagnostics). For SW-FCCS measurements, cells were either transfected with Fugene or electroporated, and live cells were used without fixing.

Fluorescence Spectroscopy—SW-FCCS was carried out on a modified Zeiss Axiovert 200 inverted microscope, as described elsewhere (16). For SW-FCCS measurements of fluorescent proteins fused with Cdc42 and its effectors, an excitation wavelength of 514 nm (argon ion laser; Spectra-Physics, Mountain View, CA) at 40 microwatts was used. The excitation light was reflected by a dichroic mirror, 525DRLP (Omega Optical, Brattleboro, VT), and measurement was carried out using a water-immersion objective (×63, numerical aperture 1.2; Carl Zeiss). Fluorescence light was spatially filtered through a 50-μm pinhole (Linos, Heidelberg, Germany) and was further split by a second dichroic mirror, 560DCLP (Omega Optical), into two detection channels, green and red. Two band pass filters, 545AF35 and 615DF45 (Omega Optical), were placed in front of the avalanche photodiode detectors (SPCM-AQR-14; Pacer Components, Berkshire, UK), to further restrict the emission wavelengths for the green and red channels, respectively. Auto- and cross-correlations were simultaneously calculated by using a hardware correlator (Flex-02–12D; Correlator.com, Zhejiang, China), and fitting of the correlation functions was carried out with Igor Pro 4.0 (Wavemetrics, Portland, OR). SW-FCCS measurements were done 24 h after transfection in a POC minichamber (Carl Zeiss) containing 1× phosphate-buffered saline (1 ml) after washing the cells thoroughly with 1× phosphate-buffered saline. All of the measurements were done at room temperature.

Quantitative Analysis and K_D Plot—The theory of quantitative analysis on molecular interactions by SW-FCCS has been introduced previously (6, 16, 17). The details of the present method and K_D evaluation are given in the supplemental material. Briefly, the free/bound concentrations of c^G (free green), c^R (free red), and c^GR (complex) are obtained after fitting the autocorrelation function (ACF) and cross-correlation function (CCF) curves and solving the FCCS equations (supplemental material) for co-expression of Cdc42V12/N17 and effectors. If the GFP fusion protein interacts with mRFP-fusion protein, it should be a linear line while plotting c^G × c^R versus c^GR, since K_D is a constant for each binding pair at equilibrium. If there is no interaction between the two fluorescent fusion proteins, there should be no linear relationship between c^G × c^R and c^GR, as discussed in the supplemental material. These concentrations are relative, since they depend on the expression levels in each cell and dark states of both of the fluorescent proteins.

Cdc42-IRSp53 in vitro K_D was determined in a competition assay using Cdc42 binding to CRIB as described by Govind et al. (11). Purified IRSp53 (amino acid residues 203–305) was added as a competitor, and the concentration giving 50% inhibition was determined.

K_D Is an “Apparent Dissociation” Constant—By measuring free and bound proteins, we derive a K_D (as described in the supplemental material), which we define here as an apparent dissociation constant. A number of technical issues and assumptions have to be considered in these measurements. First, there is endogenous expression of protein, which may alter the absolute concentration of protein. Second, we assume that the complexes that form are one to one; however, it may be the case that more than two proteins are involved. Third, our mRFP-GFP tandem construct should show 100% cross correlation but only gives ∼32%. This may be explained partly by dark states of fluorescent proteins (18).

FRET Analysis—Two methods were used to follow FRET. AP-FRET was measured as described in Ref. 12 using a Zeiss LSM 510 confocal microscope with a C-Apochromat ×63, numerical aperture 1.2 objective. The fusion proteins of GFP-mRFP were excited using a 488- and 561-nm laser line as an excitation source and by selecting a 405/488/561-nm dichroic mirror and 490- and 565-nm secondary dichroic mirrors for GFP and mRFP emission, respectively. The emission was monitored by selecting GFP (BP 505–550) and Red (LP 575) emission filters to record the fluorescence intensity (for details, see Ref. 12). FLIM was performed with the LIFA system (Lambert Instruments) on an inverted wide field fluorescence microscope (Olympus IX71, Center Valley, PA) with a 60 × 1.35 oil immersion objective and software. GFP was excited by a sinusoidally modulated 4-milliwatt 470 nm light-emitting diode at 40 MHz. The GFP filter set was used for excitation and emission signals. The emission was collected by an intensified CCD camera. Fluorescein isothiocyanate was used as a standard lifetime reference of 4 ns. 12 phase- and modulation-shifted images were taken, fitted with a sinus function, and used for the calculation of lifetime (19). The lifetimes from 30–40 regions of interest were taken from different cells and averaged to give the average lifetime and S.D. The experiments were repeated three times, and data from a single representative experiment are shown.

RESULTS

Cdc42-Effector Interactions as a Model for Cell Signaling— Cdc42 is a molecular switch that is activated and converted into the GTP-bound form, through GDP-GTP exchange factors acting downstream of membrane-bound receptors. Once activated, Cdc42 interacts with effectors such as IRSp53 and N-WASP through the CRIB domain. A schematic of the domain structure and size of IRSp53, N-WASP, and CRIB (domain of PAK) is shown in Fig. 1. GFP or mRFP pairs of these proteins were used in the analysis with N-terminal tagging in all cases. Previous work has shown (12) that GFP or mRFP fusions of Cdc42, IRSp53, N-WASP, and CRIB are functional. Two point mutations of Cdc42, V12 and N17, were used to demonstrate high and low affinity protein-protein interaction, respectively.

FIGURE 1.

Schematic of protein domains. Shown are domain structures of IRSp53, N-WASP, and PAK-CRIB. Schematics of the proteins (with amino acid numbers) used in this study are shown. Fusion proteins of GFP and mRFP were made by linking the N terminus of the respective interacting proteins and the CRIB domain of PAK.

Cdc42 Interacts Directly with IRSp53, N-WASP, and CRIB in Vivo—As a first step to determining K_D values, we used FRET analysis to confirm that the Cdc42-effector pairs were interacting directly with each other in vivo. Co-expression of Cdc42V12 with IRSp53 or N-WASP in CHO cells induces distinct phenotypes (Fig. 2). CHO cells are flat with little if any peripheral morphology. Co-expression of Cdc42V12 with IRSp53 induces the cells to take on a neuronal morphology; they attain a bipolar shape with neurite-like processes and filopodia. Cdc42V12 co-expression with N-WASP induces multiple small protrusions with filopodia. In contrast, the Cdc42V12/CRIB co-expression did not affect cell phenotype (Fig. 2). Both FRET and subsequent FCCS analysis were carried out at the cell peripheries.

FIGURE 2.

FRET analysis of Cdc42V12 interaction with CRIB, N-WASP, and IRSp53. CHO cells (a, b, and d) or fibroblasts (c) were transfected with cDNAs encoding (GFP/mRFP)-Cdc42 and either CRIB (a), N-WASP (b and c), or IRSp53 (d). Regions of interest (ROI) were then marked, and AP-FRET was carried out as described under “Materials and Methods.” The line graphs on the right of the cell images show changes in the GFP (green) and mRFP (red) intensity during the AP-FRET experiment in the regions of interest (scale bar 10 μm).

Two FRET indicators were used: (i) change in donor fluorescence upon acceptor photobleaching (AP-FRET; Fig. 2) and (ii) change in donor lifetime measured by FLIM (Table 1). Detailed descriptions of these methods can be found in Refs. 12, 20, and 21. Controls for the FRET analysis included a tandem fusion mRFP-GFP. For the AP-FRET, we also measure a correlation coefficient that reflects whether changes in donor fluorescence are correlated with changes in acceptor fluorescence upon bleaching. In the FRET scenario, there should be high negative cross-correlation. The FRET efficiency (FE) values obtained with controls set the upper and lower limits for protein-protein interaction. For Cdc42V12-N-WASP and Cdc42V12-IRSp53, we measured AP-FRET in neurites, cell protrusions, and filopodia, giving us spatial information (Fig. 2). Similarly, a distinct lifetime change was observed for IRSp53 (1.93 ± 0.04 ns), N-WASP (2.08 ± 0.09 ns), and CRIB (1.76 ± 0.10 ns) along with Cdc42V12 in comparison with Cdc42V12 (2.30 ± 0.09 ns) alone. The FE values presented in Table 1, obtained from both AP-FRET and FLIM, demonstrate that Cdc42V12 is interacting directly with IRSp53, N-WASP, and CRIB in vivo.

TABLE 1.

Cdc42 FRET analysis The table shows FRET efficiency obtained for Cdc42V12/N17-effector interactions by AP-FRET (first column) and changes in lifetime obtained with FLIM (third column). Data in the second column represent cross-correlation values for changes in mRFP/GFP intensity monitored during bleaching in AP-FRET. Data are expressed as averages ± S.D. (n = 7–12).

Construct	AP-FRET efficiency ± S.D.	Cross-correlation value ± S.D.a	FLIM-FRET efficiency ± S.D.
	%		%
Positive control
mRFP-GFP (tandem fusion)	28.6 ± 3.7	–0.99 ± 0.01	25.4 ± 7.4
Negative controls
Cyto-mRFP/GFP	1.9 ± 1.5	0.17 ± 0.63	1.1 ± 1.2
mRFP-IRSp53 + cyto-GFP	2.1 ± 1.5	–0.16 ± 0.55
GFP-N-WASP + cyto-mRFP	2.7 ± 1.9	–0.63 ± 0.42
Experimental
mRFP-Cdc42V12 + GFP-CRIB (domain)	18.4 ± 3.6	–0.99 ± 0.01	22.7 ± 5.5
mRFP-Cdc42N17 + GFP-CRIB (domain)	2.3 ± 2.3	–0.09 ± 0.75	4.1 ± 1.8
mRFP-N-WASP + GFP-Cdc42V12	10.2 ± 2.4	–0.97 ± 0.02	11.7 ± 2.4
mRFP-N-WASP + GFP-Cdc42N17	2.4 ± 1.7	–0.47 ± 0.49	6.0 ± 2.2
mRFP-IRSp53 + GFP-Cdc42V12	9.8 ± 3.5	–0.94 ± 0.06	13.8 ± 4.2
mRFP-IRSp53 + GFP-Cdc42N17	2.7 ± 2.7	–0.10 ± 0.69	3.5 ± 2.6

^an = 7–12 cells for each pair. Cross-correlation values between –0.7 and –1.0 are considered as positive FRET (12)

Parameters Obtained from SW-FCCS—Fig. 3A shows a typical output obtained from an SW-FCCS analysis: two ACFs, signals from GFP (green line) and mRFP (red line) channels, respectively, and one CCF (black line), which is the relationship between the two ACFs. Fig. 3B shows SW-FCCS traces obtained from the mRFP-GFP tandem fusion (a) and the free cyto-GFP and cyto-mRFP (b), in which maximum and minimum CCFs, respectively, would be expected. When two fluorescent fusion proteins diffuse together through the confocal volume, the amplitudes of the CCF curve are significantly higher than the cross-correlation expected from cross-talk alone. If the two fluorescent proteins diffuse randomly through the confocal volume, the amplitude of CCF should be very low compared with those of the ACF curves, indicating the absence of cross-correlation. The complex percentages obtained for a and b are 32.3 ± 3.7 and 3.7 ± 3.4, respectively. The observed cross-correlation amplitude for b implies that there is limited interaction between green and red channels. We consider the value of b as a background cross-correlation signal.

FIGURE 3.

SW-FCCS methodology and data analysis. A, the diagram shows the configuration of the SW-FCCS system with typical fluorescence fluctuation signals (in boxes below the system diagram). Green, GFP; red, mRFP. The graph shows the ACFs (green, GFP; red, mRFP) and CCF (black) lines. The axis terms in the graph G(τ) and τ(s) represent the correlation function and the lag time, respectively. B, ACF and CCF signals measured from CHO cells expressing fluorescent proteins. ACF of GFP (green) and mRFP (red) and CCF (blue) with their fits (line below main curves). Black dashed lines (base) represent the limit of cross-correlation levels due to cross-talk between the two channels. a, signals from mRFP-GFP (tandem fusion) expressed in CHO cells as positive control. The blue line above the dashed base line represents the positive cross-correlation. b, signals from cyto-GFP and cyto-mRFP proteins serving as negative control and showing low cross-correlation amplitude.

ACF and CCF Signals for Cdc42 and Its Effectors—We chose CHO cells for this analysis, since these cells have little or no expression of endogenous Cdc42 (22). Fig. 4 shows the ACF and CCF curves for GFP-Cdc42V12 or GFP-Cdc42N17 and mRFP-fused effectors of IRSp53, N-WASP, and mRFP-Cdc42V12/N17 with GFP-CRIB. For Cdc42V12 (a–c), but not for Cdc42N17 (d–f), the CCF curves are approaching the ACF curve, suggesting a significant interaction among these Cdc42-effector pairs. The percentage interaction determined from the correlation functions of Cdc42V12 with N-WASP, CRIB, and IRSp53 was 84.8 ± 16.2, 51.6 ± 20.0, and 30.9 ± 19.4, respectively, and for Cdc42N17, the percentage interaction values stood as 8.6 ± 4.8, 5.0 ± 4.8, and 10.8 ± 8.8, respectively (Table 2). The diffusion times of Cdc42V12 were similar to those of CRIB, N-WASP, and IRSp53 when expressed together. In contrast, Cdc42N17 had a diffusion time distinct from that of CRIB, N-WASP, and IRSp53 (Table 2). These differences in diffusion times reflect the lack of complex formation with Cdc42N17.

FIGURE 4.

ACF and CCF plot for Cdc42-CRIB, -N-WASP, and -IRSp53 protein pairs. CHO cells were transfected with GFP/mRFP-fused Cdc42V12 or Cdc42N17 with GFP/mRFP-fused CRIB, N-WASP, or IRSp53. Selected cells were then used for measurements based on intensity. Shown are the ACF and CCF for Cdc42V12/N17 with CRIB (a and d), N-WASP (b and e), and IRSp53 (c and f).

TABLE 2.

SW-FCCS analysis of Cdc42-CRIB, -N-WASP, and -IRSp53 pairs The table summarizes data derived from SW-FCCS experiments shown in Figs. 4 and 5 on Cdc42 interacting with its effectors. The first column gives the concentration of individual proteins. The second column gives the diffusion times of proteins (see supplemental material). The third column gives percentage of complex formed. The fourth column gives estimated K_D. SW-FCCS data are averages ± S.D. (n = 7–28). K_D data are expressed as averages ± S.E. (n = 7–28). n = 7–28 cell measurements for each set of experiments.

Constructs	Concentration ± S.D.	Diffusion times (τD) ± S.D.	Complex ± S.D.	∼KD ± S.E. (R2 value)
	nm	ms	%	nm
Positive control
mRFP-GFP (tandem fusion)			32.3 ± 3.7	NDa
GFP	269.6 ± 130.0	0.8 ± 0.1
mRFP	106.6 ± 64.8	1.1 ± 0.2
Negative control
Cyto-GFP/mRFP			3.7 ± 3.4	Failed to fit linear regression
GFP	121.0 ± 45.4	0.6 ± 0.1
mRFP	78.5 ± 35.3	0.5 ± 0.1
GFP-CRIB(domain) and mRFP-Cdc42V12			51.6 ± 20.0	250 ± 25 (0.56)
GFP-CRIB	365.0 ± 206.0	147.3 ± 180.6
mRFP-Cdc42V12	177.0 ± 69.3	88.5 ± 51.1
GFP-CRIB (domain) and mRFP-Cdc42N17			5.0 ± 4.8	Failed to fit linear regression
GFP-CRIB	191.0 ± 139.0	1.3 ± 0.8
mRFP-Cdc42N17	512.0 ± 288.0	10.7 ± 0.7
GFP-Cdc42V12 and mRFP-NWASP			84.8 ± 16.2	27 ± 3 (0.44)
GFP-Cdc42V12	81.8 ± 60.6	179.0 ± 270.0
mRFP-NWASP	252.0 ± 237.0	103.9 ± 47.3
GFP-Cdc42N17 and mRFP-NWASP			8.6 ± 4.8	Failed to fit linear regression
GFP-Cdc42N17	71.6 ± 84.8	71.0 ± 76.2
mRFP-NWASP	162.0 ± 249.0	11.2 ± 15.9
GFP-Cdc42V12 and mRFP-IRSp53			30.9 ± 19.4	391 ± 33 (0.8)
GFP-Cdc42V12	76.3 ± 69.8	15.0 ± 9.2
mRFP-IRSp53	224.0 ± 161.0	8.5 ± 5.8
GFP-Cdc42N17 and mRFP-IRSp53			10.8 ± 8.8	2180 ± 494 (0.09)
GFP-Cdc42N17	221.0 ± 168.0	86.9 ± 154.1
mRFP-IRSp53	145.0 ± 143.0	7.6 ± 4.2

^aND, not determined

For the Cdc42V12-N-WASP and Cdc42V12-CRIB pairs, the complex percentage measured by SW-FCCS was higher than that observed for the tandem fusion (mRFP-GFP). These results are unexpected but could be explained by (i) multimerization of CRIB and/or N-WASP, (ii) mRFP-GFP tandem fusion being less fluorescent than protein fusions, or (iii) a difference in the microenvironment of the fluorescent proteins. A similar differences between tandem and protein pairs has been reported (23) for Bat2-GFP-mCherry tandem protein (complex ∼50%) and GFP-Arc18/mCherry-Arc15 (complex ∼60%).

Derivation of K_D from Plots—The free GFP-mRFP fusion protein concentrations can be derived from the diffusion of complex ACF signal. The overlap spikes observed from the red and green channel generate the CCF, which gives the complex concentration. The apparent K_D values can be calculated from data obtained with ACF and CCF, as described in the supplemental material. Fig. 5, a–c, shows linear regression lines with K_D as slope. The K_D obtained for Cdc42V12 interaction with CRIB, N-WASP, and IRSp53 was 250, 27, and 391 nm, respectively. Table 3 shows a comparison of in vivo and in vitro K_D measurements. The in vitro data use purified Cdc42 and Cdc42 binding domains derived from PAK (CRIB), WASP (24), and IRSp53 proteins. The Cdc42-CRIB data allow direct comparison of the in vivo and in vitro K_D measurements, since the same domain is being used in the two measurements. Taken together, the measured K_D values are similar in vitro and in vivo. Fig. 5, d–f, shows that dominant negative mutant Cdc42N17 with the respective effector proteins failed to generate the linear regression fit passing through the origin with a positive R² value. The failure of Cdc42N17 effector pairs to generate the linear regression fit reflects the specificity of the method.

FIGURE 5.

K_D plots for Cdc42-CRIB, -N-WASP, and -IRSp53 interactions. The graphs show the best fit analysis for concentration of GFP and mRFP fusion proteins in the complex against concentration of free GFP fusion proteins and mRFP fusion proteins. Shown is Cdc42V12/N17 with CRIB (a and d), N-WASP (b and e), and IRSp53 (c and f). The Cdc42N17 mutant with effectors CRIB and N-WASP failed to generate a linear regression fit producing a negative R² value when forced to pass through the origin. The R² value represents the “quality of fit” (for a perfect fit, the value should be close to 1). The K_D values were obtained for Cdc42V12 and its effectors from the slope of the fitted linear line, briefly by plotting the concentration of c^G (concentration of free green) and c^R (concentration of free red) against c^GR (concentration of complex). If the GFP-fusion protein interacts with mRFP-fusion protein, it should be a linear line while plotting c^G × c^R versus c^GR, since K_D is a constant for each binding pair at equilibrium. If there is no interaction between the two fluorescent fusion proteins, there should be no linear relationship between c^G × c^R and c^GR.

TABLE 3.

Comparison of in vivo and in vitro K_D measurements of Cdc42-effector interactions The Cdc42-CRIB, -N-WASP, and -IRSp53 interactions were measured in vivo using FCCS, and data are taken from the plots presented in Fig. 5. Cdc42-CRIB and WASP interactions were in vitro, and data are taken from Owen et al. (24). Cdc42-IRSp53 binding was measured as described under “Materials and Methods,” and K_D is an average ± S.D. (n = 6). N-WASP and WASP are isoforms with almost identical Cdc42 binding sequences.

Proteins	CRIB in vivo	CRIB in vitro	N-WASP in vivo	WASP in vitro	IRSp53 in vivo	IRSp53 in vitro
KD	250 ± 25	620 ± 138	27 ± 3	55 ± 6	391 ± 33	880 ± 86
Source	This study	Ref. 24	This study	Ref. 24	This study	This study

We also carried out similar K_D determinations for the Cdc42V12-N-WASP pair in N-WASP KO cells. The N-WASP KO background did not alter the derived K_D (data not shown), suggesting that endogenous N-WASP protein does not influence the interaction.

DISCUSSION

Quantitative or semiquantitative analysis of protein-protein interactions has been made possible by in vitro studies using tagged proteins with the two-hybrid method and co-immunoprecipitation. These methods, although powerful, suffer from the absence of physiological relevance. In other words, they are carried out, for example, in the absence of cell compartmentalization or steady-state conditions. In addition, they do not give spatial or temporal information. The application of FRET and FCCS provides a means to address these problems, since these methods can be applied in vivo (15, 26, 27).

In a recent analysis of the Ste5-mitogen-activated protein kinase pathway in yeast, Maeder et al. (28) derive K_D values by FCCS. However, FCCS cannot determine whether two proteins interact directly with each other, and this is why parallel FRET measurements are important. FRET gives definitive evidence for two proteins interacting with each other (25) by way of measuring the distance between two fluorophores attached to the proteins (only proteins binding each other will allow the fluorophores to transfer energy). AP-FRET gives spatial information by showing where exactly in the cell two proteins are interacting. We found that Cdc42V12 interacts with both N-WASP and IRSp53 in filopodia, with N-WASP in cell extensions and IRSp53 in neurite-like processes.

FCCS is carried out in vivo/in live cells and measures protein concentrations of free and bound proteins and by so doing can be used to determine the K_D of protein-protein interaction. The determination of free protein concentration is in itself an important parameter that can be linked with cell phenotype. FCCS also gives diffusion parameters of the proteins, which can be used to assess the molecular weight of the complex. We have measured the dissociation constant for Cdc42 interacting with two of its effectors and a discrete domain. The values obtained in vivo are similar to values reported from purified proteins in vitro. Future work quantitating the formation of Cdc42-effector complexes (as well as other complexes) will give us novel mechanistic insight into cell and developmental biology.

CONCLUSION

We have presented for the first time an approach using FRET and FCCS to determine in vivo K_D values of protein-protein interactions in live mammalian cells. This approach will allow cell signaling and cell biology work to shift from generating qualitative to quantitative data with spatial resolution. The generation of quantitative data is essential if we are to gain mechanistic insight of cellular processes in vivo. Furthermore, this analysis has utility for drug screening of mammalian cells.