Three-Color Confocal Förster (or fluorescence) Resonance Energy Transfer Microscopy: Quantitative Analysis of Protein Interactions in the Nucleation of Actin Filaments in Live Cells

Abstract

Experiments using live cell 3-color FRET microscopy and corresponding in vitro biochemical reconstitution of the same proteins were conducted to evaluate actin filament nucleation A novel application of 3-color FRET data is demonstrated, extending the analysis beyond the customary energy transfer efficiency (E%) calculations.. MDCK cells were transfected for co-expression of Teal-N-WASP/Venus-IQGAP1/mRFP1-Rac1, Teal-N-WASP/Venus-IQGAP1/mRFP1-Cdc42, mCFP-Rac1/Venus-IQGAP1/mCherry-actin or CFP-Cdc42/Venus-IQGAP1/mCherry-actin, and with single-label equivalents for spectral bleedthrough correction. Using confirmed E% as an entry point, fluorescence levels and related ratios were correlated at discrete accumulating levels at cell peripheries. Rising ratios of CFP-Rac1:Venus-IQGAP1 were correlated with lower overall actin fluorescence, whereas the CFP-Cdc42:Venus-IQGAP1 ratio correlated with increased actin fluorescence at low ratios and was neutral at higher ratios. The new FRET analyses also indicated that rising levels of mRFP1-Cdc42 or mRFP1-Rac1 respectively promoted or suppressed association of Teal-N-WASP with Venus-IQGAP1. These 3-color FRET assays further support our in vitro results about the role of IQGAP1, Rac1 and Cdc42 in actin nucleation, and the differential impact of Rac1 and Cdc42 on the association of N-WASP with IQGAP1. Moreover, this work emphasizes the power of 3-color FRET as a systems biology strategy for simultaneous evaluation of multiple interacting proteins in individual live cells.

INTRODUCTION

The nucleation of branched actin filament networks can be accomplished by coordinated activities of IQGAP1, N-WASP, Rac1 and Cdc42. Actin and these three proteins were investigated in different combinations, both in vitro using purfied proteins and by live cell 3-color FRET imaging up to the level of energy transfer efficiencies (E%) (1). The triple expressions included CFP-Rac1/Venus-IQGAP1/mCherry-actin, CFP-Cdc42/Venus-IQGAP1/mCherry-actin, Teal-N-WASP/Venus-IQGAP1/mRFP1-Rac1 or Teal-N-WASP/Venus-IQGAP1/mRFP1-Cdc42, and equivalent empty vector controls (CFP-Venus-mCherry and Teal-Venus-mRFP1). In spite of their 90% homology, our in vitro results suggested differential effects of small G-proteins Rac1 and Cdc42 in the actin nucleation process. Most notably, Cdc42 stabilized binding of IQGAP1 to N-WASP, whereas Rac1 had the opposite effect.

Förster (or fluorescence) resonance energy transfer (FRET), when converted to energy transfer efficiency (E%), represents a powerful tool to investigate and quantify protein-protein interactions and molecular co-localization within distances of ~1–10 nm. This paper assumes that the reader is familiar with the concept of FRET, as documented by an increasing number of applications and publications, mainly in the context of 2-color FRET (2–4). Quantitative analysis of 2-color FRET data requires additional processing steps to deal with spectral bleedthrough corrections applying different algorithms – largely automated by software (5–7). 3-color FRET increases the complexity of correction, but adds greatly to the utility of assaying 3 potential live FRET interactions in one experiment, instead of 3 separate tests with two labeled proteins each in multiple individual experiments (8).

Since 3-color FRET simultaneously establishes the presence of three labeled proteins spatially and temporarily – implying direct interaction – and fluorescence serves as a proxy for molecular quantities and fluorescence ratios as relative concentrations, we sought to gain more insights by correlating fluorescence levels of the proteins engaged in 3-color FRET. Ideally, cellular investigations with labeled components should take into account the unlabeled endogenous species, but so far, we have not employed any strategies to suppress expression of the native protein equivalents of the fluorescent fusion proteins discussed here. Furthermore, we selected only pixels where FRET occurs between all 3 labeled protein pairs, except between actin and Rac1 or Cdc42, where none is expected. The FRET event and minimum E% was used as the entry condition for fluorescence correlations (1).

Our previous work separated cell periphery sites (lamellipodia) from cell-cell boundaries, because of the possibility that actin filament nucleation mechanisms differ in those two locations (1). Here we concentrate is exclusively on data from the cell periphery to demonstrate further utility of the assay. In another related, but different three-color labeled protein study (Teal-N-WASP/Venus-IQGAP1/mCherry-actin) we compared occurence of FRET among all three labels (as in this paper) with pixels where only two of the three where interacting – an additional analysis opportunity afforded by 3-color FRET (9).

MATERIALS AND METHODS

Transgene Expression

Triple-labeled MDCK cells, including those expressing control fluorescent proteins (CFP-Venus-mCherry or Teal-Venus-mRFP1) were transfected using Lipofectamine 2000 immediately after being plated and were imaged 24 hours later. This transfection protocol was also used to drive expression of each fluorescent fusion protein individually to facilitate spectral bleedthrough correction with our proprietary PFRET (processed FRET) software (copyright, University of Virginia), which functions as an ImageJ plugin. In all cases, the cells were plated onto 25 mm round coverslips in 6-well dishes (35 mm diameter per well) using 4 μg total DNA per well. DNA amounts for each vector in triple-label transfections were as follows: CFP/Venus/mCherry (as either fluorescent unfused or fusion proteins), 1 μg CFP + 2 μg Venus + 1 μg mCherry; Teal/Venus/mRFP1 (as either fluorescent unfused or fusion proteins), 1.25 μg Teal + 1.25 μg Venus + 1 μg mRFP1. Most other pertinent materials and methods, including descriptions of the fluorescent fusion proteins, have been described in thorough detail previously (1).

3-Color Confocal FRET Microscopy

The 3-color FRET method (8), and PFRET software were employed throughout this study. As illustrated in Figure 1, the following multi-step process was used for image acquisition and analysis; examples of related images are shown in Figure 2. Each fluorescent fusion protein experiment shown in the figures was performed a minimum of 3 times, and representative results from one such experiment is illustrated. The empty vector control data were obtained from one experiment.

Figure 1. Image acquisition, processing and analysis steps.

Step1. Raw images are generated for triple-label, single-label control (for spectral bleedthrough [SBT] correction) and unlabeled specimens (for background noise correction) at identical imaging settings. Sequential excitation with one wavelength at a time in the line-scan mode captures donor, acceptor and uncorrected (for spectral bleedthrough, or SBT) FRET in the triple-label and reference fluorescence and SBT levels in the single-labels. Step2. The PFRET ImageJ plugin removes average background noise identified in unlabeled specimens, specific for each emission channel. Step3. In ImageJ, the IQGAP1 reference images for Regions of Interest (ROI) are manually modified to only show cell-peripheral or cell-cell boundary areas. Step4. The automatic ROI selection plugin allows to specify ROI size (here 3×3 pixels) and a lower threshold for the average gray-level units in the ROI (here 10) and applies this to the IQGAP1 images. These one-dimensional regional selections are applied to all images during the analysis; only those meeting all thresholds including the requirement of all FRET pairs having to interact (note exception in the text). Approximately 12% of original ROIs meet all thresholds. Step5. Based on pixel-by-pixel processing, the PFRET software will correct for SBT, generate E%, fluorophore distance, SBT-corrected images and 26 data categories for each ROI meeting processing and threshold specifications. We consider the PFRET algorithm to be a most exacting approach to correct non-linear SBT, take into account Förster distances, fluorophore quantum yields and the ability to isolate discrete populations relevant to investigating biological questions, such as the requirement for all fluorophores having to interact to be included in the data analysis. Note: this figure has been published previously (1), was modified to include additional information and is reproduced here with the publisher’s permission.

Figure 2. First Row.

Following the steps of the imaging protocol, using one of the Teal-N-WASP/Venus-IQGAP1/mRFP1-Rac1 images as an example, the images represent each of the fluorophores excited with their specific wavelength and emitted in in an optimal nanometer range (details in Materials and Methods). They were background-noise corrected. ROIs meeting thresholds are shown in this example from an inset. Second Row: Single-label control images used for spectral bleedthrough (SBT) correction. Third Row: At the same time images of the first row are taken, the uncorrected FRET images are produced, still containing background noise and SBT. Fourth Row: After processing all images with the PFRET software, inter alia, the all-important corrected FRET (PFRET) images are generated forming the basis of all the data for E%, distance, unquenched donor and SBT% correction. Note the much lower intensity of the PFRET images compared with their uncorrected versions, highlighting the importance of SBT correction for quantitative evaluations. All of the charts are based on the ROIs applied to the images. Note: some of the individual panels in this figure have been published previously (1), and are reproduced here with the publisher’s permission.

Step 1 collects 6 images for each field of view of a triple-labeled specimen and corresponding single labeled specimens, using microscopy settings optimized for each fluorophore. An unlabeled specimen is also imaged at these settings for background noise subtraction. Starting with the lowest, wavelength, the three fluorophores are named F1, F2 and F3, their corresponding excitation wavelengths Ex1, Ex2, Ex3, and emission as Em1, Em2 and Em3. We verified the absence of any back-bleedthrough, i.e. Ex2 did not excite F1 and Ex3 did not excite F1 or F2; this allowed us to collect images at Ex1 in three channels (Em1-3), at Ex2 in two channels (Em2-3) and at Ex3 in one channel (Em3). Since we were imaging live cells, we chose a line scan, where at one wavelength at a time (2.5 milliseconds/line) a line is scanned at a total of 7.5 milliseconds for the three wavelengths, so as to capture protein kinetics with minor time delay. A frame scan would complete 3 consecutive scans by each laser and return to any particular line in the image with greater delay. In either case, the FRET signal would not be affected as the emissions in the donor and FRET channel are captured simultaneously at donor excitation. We also did not observe any pixel shifts between images of the three excitation wavelengths. All images were taken on a Leica TCS SP5 X confocal microscope equipped with a temperature-controlled stage (10). Argon laser lines at 458nm were used for F1 and 514nm for F2, while F3 was excited at 581nm with a tunable white light laser, all at 400 Hz scan speed. For each excitation wavelength, laser power was controlled through acousto-optical tunable filters (AOTF). The three photo multiplier tubes (PMT) emission channels were set at 468–515nm (Em1) for F1, 525–585nm (Em2) for F2 and 595–750nm (Em3) for F3 using an acousto-optical beamsplitter (AOBS). Images were acquired as 8-bit TIFF files of 512×512 or 1024×1024 pixels using a 60x 1.2NA water-immersion objective.

Step 2 establishes the average background noise for each channel and excitation wavelength by generating several random Regions of Interest (ROIs) on the unlabeled specimen’s images. The PFRET software subtracts this average channel-specific value from all images and later will generate background-corrected images.

Step 3 manually isolates cell-peripheral and cell-cell boundary regions of the Venus-IQGAP1 triple-label images and uses those as a template for 3×3 pixel ROI selection for Step 4.

Step 4: Figure 1 shows an automatic ROI selection plugin (part of PFRET software) where variable ROI size and lower threshold are specified to meet selection criteria. These ROIs will be processed in the next steps below and only those that meet the criteria/thresholds of ALL three fluorophores will be included in the final data, the most important one being the requirement of FRET interactions between all three (note exception). In this manuscript, only cell-peripheral data is presented.

Step 5, the automatic processing phase of the PFRET software (8). The software contains many options and thresholds described below to isolate data populations relevant to biological processes (such as morphology) and reject marginal or outlier values. Its most important feature is the correction of the frequently non-linear spectral bleedthrough. This is achieved by imaging single-label donor and acceptor specimens at identical settings and record their emissions at donor excitation in the FRET channel. This is followed by matching single and double-label absorption intensities within narrow ranges and subtracting the bleedthrough ratios of the single-labels at donor excitation from the double-label - pixel-by-pixel. In addition several variables are manually entered before processing: fluorophore quantum yields (QY), detector quantum efficiencies (QE), FRET-pair specific Főrster distances, plus many threshold choices on fluorescence levels. While quantum yields of fluorophores conjugated to proteins frequently differ from the pure fluorophore, as a matter of practical application, published QY-values are used. The PFRET software normalizes the differences between the QY of each FRET pair to achieve more comparable intensity levels and D:A ratios; E% calculation is not affected. Once these parameter values have been entered, the PFRET software calculates uncorrected FRET, PFRET, E%, spectral bleedthrough (SBT) percent and donor: acceptor ratios. The thresholds are applied to each pixel in the previously generated ROIs of the images and all thresholds must be met for the pixel to be included in the analysis. Typically, the first processing run is executed at default without any threshold restrictions. In the final processing step, we set the PFRET and E% threshold at 5 to eliminate borderline FRET events. The only exception was made for the G-protein-actin random FRET occurrence, where no biological interaction is expected, which was set at zero; not having done so, would have eliminated a large amount of valid FRET data between Rac1/Cdc42 – IQGAP and IQGAP1-actin in shared ROI locations. The PFRET software output consists of images (background-subtracted-, E%-, fluorophore distance-, SBT images) and data. The data cover a complete account of the SBT ratios at fluorescence ranges and spreadsheet files of all ROIs in each image detailing 26 data categories for each ROI: ROI number; coordinates on the image; number of pixels meeting inclusionary/exclusionary thresholds (e.g. here, FRET interaction between all 3 fluorophores must take place before the pixel is included, exception noted above); quenched fluorescence of fluorophore1 and 2; raw/uncorrected FRET between fluorophores 1-2, 1-3, 2-3; unquenched fluorophore (uF) 1 and 2 and fluorescence of fluorophore 3; processed FRET (corrected for spectral bleedthrough-PFRET) between fluorophores 1-2, 1-3 and 2-3; ratios of uF1:uF2, uF1:uF3 and uF2:uF3; E% of uF1-2, uF1-3 and uF2-3; FRET distance for uF1-2, uF1-3 and uF2-3; and the percent of spectral bleedthrough correction (SBT%) for uF1-2, uF1-3 and uF2-3. Separating ROIs into subpopulations, charts and statistics were performed using Excel spreadsheet software (Microsoft).

RESULTS

We initially used the 3-color FRET assay in live cells to evaluate the physiological significance of the following in vitro reconstitution results obtained using purified proteins: (a) the small G-proteins, Rac1 or Cdc42, cooperatively stimulate actin polymerization with IQGAP1, albeit with distinct kinetics; and (b) in dose-dependent manners, Rac1 suppresses and Cdc42 stimulates binding of IQGAP1 to N-WASP (1). Having established FRET interaction between each pair of the 3 labeled moieties, using the same basic experimental strategy (Figures 1 and 2), we now correlate rising fluorescence of one protein with the fluorescence level of the second (in the presence of the third) and by extension, whether the changing ratio of two proteins affects the third. Since the in vitro data showed dose-dependency, a single correlation coefficient for the whole data set would miss a tipping or saturation point based on fluorescence. We consequently used a stepwise increasing series of correlation coefficients in accumulating cohorts. All charts and correlations also show the empty vector equivalents.

CFP-Rac1/Venus-IQGAP1/mCherry-actin and CFP-Cdc42/Venus-IQGAP1/mCherry-actin combinations

Rac1 and Cdc42 do not interact directly with actin (11), confirmed by low-level random FRET. However, both, increasing fluorescence of Rac1 and Cdc42 correlate with rising actin fluorescence (Fig 3A), as indicated by in vitro results via other effectors – we will show one of them – N-WASP – in the next section. When the overall correlation coefficient is dissected into discrete accumulating cohorts (Fig 3B), Rac1 quickly reaches a steady state ending with a final coefficient of 0.28, with Cdc42 peaking at 0.43 to end with 0.33 for the total set. Another important regulator – IQGAP1 – in this triple specimen provided a clue as to how the small G-proteins exert their differential effect. Correlating actin fluorescence as a function of rising Rac1:IQGAP1 or Cdc42:IQGAP1 ratios (Fig 3C) revealed a progressive lowering of actin fluorescence, as the Rac1:IQGAP1 ratio rose, while rising Cdc42:IQGAP1 ratios had a small, opposite impact at low initial ratios (Fig 3D). T-Tests comparing actin fluorescence data points in the Rac1 vs. Cdc42 triple-label showed a statistical difference of p=9.2E-124 and IQGAP1 a p=2.23E-30; adding the empty vector data of CFP and mCherry, ANOVA produced a p=0 in both cases, with all the other parameters confirming a statistical difference.

Figure 3. Differential effects of Cdc42 vs. Rac1 and Cdc42/Rac1:IQGAP1 Ratios on actin fluorescence.

(A) Correlation of Rac1 (blue), Cdc42 (red) vs. actin fluorescence data points -in the presence of IQGAP1 - plus empty vector CFP (green) vs. empty vector mCherry fluorescence – in the presence of Venus. (B) Gradually segmenting the changing correlation coefficients by accumulating Rac1, Cdc42 or CFP fluorescence to detect at which level change occurs. In Fig. 3(A) & (B), the narrow range of the empty vector control (green) and its virtually unchanged correlation coefficient over that range indicates no biological interrelationship; equally, rising levels of Rac1 (blue) maintain their correlation virtually unchanged with Cdc42 (red) increasing their correlation to a peak between 40–60 gray-level units suggesting a modest concentration response. (C) To include IQGAP1 in the equation, the impact on actin fluorescence is charted by the rising ratio of Rac1:IQGAP1 (blue) and Cdc42: IQGAP1 (red), also including empty vector data (green). (D) At the lowest Cdc42:IQGAP1 (red) ratio is there insignificant positive correlation, while increasing Rac1:IQGAP1 (blue) reaches negative correlation peak at accumulation ratio of 0.1–0.4. This strengthens the idea that the impact on actin nucleation by the small G-proteins is modulated (or not, in the case of Cdc42) by their changing ratio inter alia to IQGAP1 (see also Suppl. Figs 1 & 2).

Supplementary figures provide details on mean fluorescence of each of the triple combinations and their mean ratios (Figure S-1A–B); The percent frequency distribution of the meaningful G-protein:IQGAP1 ratios support the finding that fewer fluorescence units associate with Rac1 than Cdc42; empty vector CFP: Venus ratios show a much narrower range (Figure S-1C). Another approach to demonstrate the impact of the G-prot:IQGAP1 ratio on the G-protein’s correlations to actin, is to sort the data in Fig 2A into discrete cohorts and correlate each by the ratio (Figure S-2A–C).

Returning to the in vitro results the imaging findings not only strongly support the dose-dependent effects of Rac1 and Cdc42, but also the role of IQGAP1, which 3-color FRET is able to correlate, all three proteins being in the same space and time.

Teal-N-WASP/Venus-IQGAP1/mRFP1-Rac1 and Teal-N-WASP/Venus-IQGAP1/mRFP1-Cdc42

The main objective of the above triple combination was to analyze - by 3-color FRET - the in vitro results that Rac1 and Cdc42 modulated the binding of IQGAP1 to N-WASP in a dose-dependent manner, which in turn had a profound effect on actin nucleation. We interpreted the imaging data by targetting the effect of Rac1 and Cdc42 on E% between N-WASP and IQGAP1 and the ratio of N-WASP: IQGAP1, the former being an expression of distance between the two fluorophores, the latter to investigate differential ratios. While overall, the effect on E% between N-WASP and IQGAP1 was independent of increasing amounts of Rac1 as expressed by fluorescence, Cdc42 had a negative effect, i.e. diminishing E%s signaled an increase of the distance between N-WASP and IQGAP1 (Fig 4A). When breaking down the correlation coefficients as described earlier into small steps, Rac only had a small effect at low fluorescence, negative to start; Cdc42 continued the negative trend, reaching a peak 60 gray-level units of fluorescence (Fig 4B). These results support the notion of ‘dose-dependency’, here interpreted by the level of fluorescence. This fluorescence- dependent feature is repeated when correlating the small G-proteins with the N-WASP:IQGAP1 ratio. At low Rac1 and Cdc42 fluorescence levels the impact on this ratio is small, but at increasing amounts their effects diverge in opposite directions, with Rac1 increasing the amount of N-WASP associating with IQGAP1 and Cdc42 increasing that amount modestly. Empty vector data on the effect of RFP on Teal-Venus show in both cases a different distribution.

Figure 4. Differential effects of Cdc42 vs. Rac1 on E% between N-WASP-IQGAP1 and the N-WASP:IQGAP1 Ratio.

(A) Correlation of Rac1 (blue), Cdc42 (red) fluorescence vs. E% (an expression of distance) between N-WASP and IQGAP1. (B) Increasing levels of Rac1 (blue) have little effect on the correlation coefficient, whereas increasing Cdc42 (red) fluorescence levels show concentration-dependent negative correlation coefficients. Rac (blue) has only a modest effect on the distance (binding?) of N-WASP to IQGAP1 at very low fluorescence, whereas rising Cdc42 (red)correlates with increasing that distance (lowering E%) (C) The rising fluorescence of Rac1 (blue) and Cdc42 (red) correlated to the N-WASP:IQGAP1 ratio shows up again differences between the two small G-proteins. (D) Breaking down the coefficients by G-protein levels of fluorescence into discrete cohorts, Cdc42 (red) has a mildly positive effect of more N-WASP associating with IQGAP1 at lower fluorescence concentrations, whereas Rac1(blue) has the opposite effect of causing less N-WASP associating with IQGAP1 as Rac1 fluorescence rises; this confirms our in-vitro results. Empty vector control (green) data shows different, random correlation features.

Figures 5A–B correlate Rac 1 and Cdc42 with N-WASP (in the presence of IQGAP1), basically reflecting their effects already seen in Figure 4C–D, with Cdc42-stimulated increase in N-WASP also driving the increased ratio of N-WASP:IQGAP1. Figures 5C–D reverse the correlations, juxtaposing IQGAP1 and N-WASP (in the presence of either Rac1 or Cdc42) showing similar trends. To capture the interrelationship among all three labeled proteins, Figures 5E–F analyze the impact on N-WASP by the increasing ratios of IQGAP1:Rac1 and IQGAP1:Ccd42. While changes in the former show N-WASP’s independence of that ratio, rising IQGAP1:Ccd42 creates a positive dependency up to a peak ratio of ~1.0 (correlation coefficient ~0.4), declining therafter. T-Tests of the N-WASP populations in the presence of IQGAP1 and/or Rac1/Cdc42 show robust statistical differences at p=3.1E-107, IQGAP1 at p=1.75E-197, N-WASP:IQGAP1 ratio at p=4.76E-08 and E% N-WASP-IQGAP1 at 3.6E-37. Adding empty vector data to the statistical pool resulted in all cases in ANOVA-based p-values of zero

Figure 5. Differential effects of Cdc42 vs. Rac1, IQGAP1 and IQGAP1:Rac1 vs. IQGAP1:Cdc42 Ratios on N-WASP fluorescence.

(A–B) Correlating the small G-proteins (Rac1-blue, Cdc42–red) with N-WASP (in the presence of IQGAP1), the correlation coefficients again show up their differences and opposite effects in a fluorescent-level dependent manner. (C–D) Equally, tracking the effect on N-WASP as a function of increasing IQGAP1 fluorescence in the presence of either Rac1 (blue) or Cdc42 (red) highlights the different influence of the small G-proteins. (E–F) The effect of Rac1 and Cdc42 is best demonstrated by correlating their ratios with IQGAP1 against N-WASP fluorescence. The IQGAP1:Rac1 (blue) ratio changes show N-WASP being independent on that ratio, whereas the opposite is true for IQGAP1:Cdc42 (red) rising ratios, which peak between 0.6 and 1.0. Empty vector (green) data distribution largely differs from labeled protein species. The response in terms of fluorescence as a surrogate measure of protein numbers of the important effector N-WASP is explored with respect to Cdc42/Rac1, IQGAP1 and the ratio of IQGAP1 and the small G-Proteins. In each case, it is the presence of Cdc42 (red) that drives the increases, modulated by the ratio to IQGAP1.

DISCUSSION

We previously confirmed and quantitatively evaluated 3-color FRET interactions among IQGAP1, N-WASP, Cdc42, Rac1 and actin (1). Here, we extended the analysis on the assumption that in pixels exhibiting 3-color FRET, most if not all fluorophore-labeled species were interacting during the millisecond time scale of imaging. To increase the statistical probability, we analyzed several thousand data points. We also did not suppress endogenous protein species in this set of experiments, although there are several strategies that can be employed to accomplish this in the future.

While our 3-color FRET assay offers an exponential increase in exploring correlations between the 3 components as compared with 2-color FRET (9) we have concentrated on a limited number of objectives using fluorescence data to analyze protein interactions in live cells beyond the level of E%. What was already known, i.e. the involvement of N-WASP, IQGAP1, Rac1 and Cdc42 in the nucleation of actin was further quantitated, by demonstrating that while both Rac1 and Cdc42 indirectly stimulate actin polymerization, they do so at differing degrees depending on their ratio with IQGAP1. The other important effector in actin nucleation, N-WASP, was also shown to be influenced by the levels of Rac1 and Cdc42, the specific nature of their ratio to IQGAP1 and thereby their apparent ability to affect the N-WASP:IQGAP1 ratio. In spite of their >90% homology, Rac1 and Cdc42 seem to cause subtle, finely-tuned changes on N-WASP, IQGAP1 and their ratios and indirectly actin, as part of the regulation of actin polymerization. While suggested in our in-vitro results, the 3-color FRET assay in live cells lends considerable support to the notion of the small G-protein’s playing a differential role - in concert with other effectors – in actin nucleation. Future applications of this imaging assay could include labeled, constitutively active Rac1 or Cdc42 together with labeled actin, N-WASP or IQGAP1, to explore the effect of changing ratios of Rac1: Cdc42 on the third FRET partner.

There is a paucity of 3-color FRET publications in live cells as the technique is mostly applied to in vitro assays, examining molecules in solutions, mainly by spectroscopy, tracking conformational changes and the like (12–17). We believe that 3-color FRET offers promise for many other systems level, cell biological applications, where capturing simultaneous interactions among three components in the same cell, provides insights not possible with 2-color FRET. Any live-cell FRET imaging -including 3-color FRET- has to consider a number of important variables which can make interpretation challenging. Transient interactions (as in this manuscript) lead to wider ranges of energy transfer efficiencies (E%), E% being an expression of distance between fluorophores. This distance varies as donors and acceptors are in the process of approaching or departing from each other – within a 1–10nm FRET distance – at the instance of imaging. Complex-formations or tracking protein clusters during cellular trafficking produce a somewhat lower range of E%s, but still contain some of the same dynamics. Both need to consider inter-vs. intra-molecular FRET, particularly when fusion-labeled proteins are overexpressed, potentially increasing the number of non-FRET donors, depressing E% levels. Thresholding donor: acceptor ratios and analyzing sub-populations offer solutions to concentrate on specific areas of interest. Because of the heterogeneity of cells and their cellular functions, large number of data points need to be generated for robust conclusions. This in turn requires computational resources, modeling and the like.

Supplementary Material

Figure S-1. Figure S-1. Mean fluorescence of Rac, Cdc42, IQGAP1, actin and their mean ratios including respective empty vector means. % Frequency distribution of G-protein:IQGAP1 ratios. (A–B).

These data summarize the means for Figure 2A and C. (C) The peaks of the percent frequency distribution of G-protein:IQGAP1 ratios show a much lower ratio Rac1:IQGAP1 (blue) at 0.6 in comparison to Cdc42:IQGAP1 (red) at 1.1 with much wider ranges, suggesting again G-protein’s differential roles.

Figure S-2. Figure S-2. Rac1, Cdc42 vs. actin fluorescence by G-protein:IQGAP1 ratio cohorts and the equivalent empty vector data.

(A–C) To highlight the effect of IQGAP1 on the relationship of Rac1/Cdc42 and actin fluorescence in a different way, the data in Fig. 3A was sorted by rising G-protein:IQGAP1 ratio segments. The lower Rac1:IQGAP1 ratios are the result of low levels of Rac1 (rather than high levels of IQGAP1), represent the larger proportion of data and are associated with higher levels of actin. Fig. 3B in contrast, Cdc42:IQGAP1 ratios have the higher proportion of data points in the >1.2 ratios segments, while overall correlating with a narrower range of actin fluorescence. This indicates again the important role of the ratios between effectors. Fig. 3C, empty vector shows a narrower distribution, suggesting a random association pattern

Figure S-3. Figure S-3. Mean fluorescence of N-WASP, IQGAP1, Rac1 and Cdc42, and their mean ratios including respective empty vector means. % Frequency distribution of N-WASP:IQGAP1 and IQGAP1:G-protein ratios.

(A–B) These data summarize the means for Figure 3 and 4. (C) Percent frequency distribution of N-WASP:IQGAP1 in the presence of either Rac1 (blue) or Cdc42 (red) are identical indicating that rather than this ratio, the ratio between IQGAP1: Rac1 or Cdc42 defines the differences between the two G-Proteins. (D) Percent frequency distribution of IQGAP1: Rac1 (blue) and IQGAP1:Cdc42 (red) confirm the point made under (C). Empty vector data repeats the higher peaks in a narrower range reported earlier.