Activation of Endogenous Cdc42 Visualized in Living Cells

See allHide authors and affiliations

Science  10 Sep 2004:
Vol. 305, Issue 5690, pp. 1615-1619
DOI: 10.1126/science.1100367


Signaling proteins are tightly regulated spatially and temporally to perform multiple functions. For Cdc42 and other guanosine triphosphatases, the subcellular location of activation is a critical determinant of cell behavior. However, current approaches are limited in their ability to examine the dynamics of Cdc42 activity in living cells. We report the development of a biosensor capable of visualizing the changing activation of endogenous, unlabeled Cdc42 in living cells. With the use of a dye that reports protein interactions, the biosensor revealed localized activation in the trans-Golgi apparatus, microtubule-dependent Cdc42 activation at the cell periphery, and activation kinetics precisely coordinated with cell extension and retraction.

Cdc42, a member of the Rho family of small guanosine triphosphatase (GTPase) proteins, regulates multiple cell functions, including motility, proliferation, apoptosis, and cell morphology (13). In order to fulfill these diverse roles, the timing and location of Cdc42 activation must be tightly controlled. The Cdc42 biosensor used here to examine the spatiotemporal dynamics of Cdc42 activation represents an in vivo application of a dye (I-SO) designed specifically to report protein conformational changes and protein interactions in living cells (4). In this biosensor, a domain from the Cdc42 effector protein WASP that binds only to activated Cdc42 was covalently labeled with the dye. The labeled domain showed a strong increase in fluorescence intensity upon binding to activated, underivatized Cdc42.

On the basis of the nuclear magnetic resonance (NMR) structure of the Cdc42-WASP complex (5), the dye was tested at three positions (I233, D264, and F271 of WASP) to optimize fluorescence response and binding (Fig. 1A). Site-specific labeling was accomplished by inserting a single cysteine and reacting the WASP domain with cysteine-selective iodoacetamide dye (4). The dye showed the greatest response at position 271, undergoing a 2.8-fold increase in fluorescence intensity upon binding to Cdc42-GTPγS, relative to either biosensor alone or GDP-loaded Cdc42 (Fig. 1B). The NMR structure of the Cdc42-WASP complex indicated that the dye at position 271 is inserted into a hydrophobic pocket formed from amino acids of both Cdc42 and WASP (Fig. 1C). The dissociation constant (Kd) for the Cdc42-biosensor interaction, determined by fitting fluorescence changes to the Michaelis equation, was 150 ± 50 nM (Fig. 1D). This was slightly higher than previously reported for a similar, unlabeled WASP fragment (77 ± 9 nM) (6), indicating that the dye minimally perturbed binding of the domain. To determine the specificity of the biosensor, we examined interactions with different activated GTPases. The biosensor distinguished proteins closely related to Cdc42 from other members of the Rho family. It did not interact with RhoA or Rac at concentrations well above physiological levels (Fig. 1D), but responded to both Cdc42 and the closely related protein TC10, which bind WASP with similar affinity (7) (fig. S1).

Fig. 1.

Design of the biosensor. (A) The biosensor was constructed from residues 201 to 321 of WASP, which contained the CRIB domain essential for Cdc42 binding. The dye gave the strongest fluorescence response when attached at position 271. A control biosensor was also generated, in which residues 246 and 249 in the CRIB domain were mutated to eliminate Cdc42 binding (H246D and H249D) (31). (B) The fluorescence excitation and emission spectra of the biosensor showed a 2.8-fold increase in intensity upon interacting with activated Cdc42 (black line, biosensor alone; dark gray line, biosensor with saturating Cdc42-GDP; light gray line, biosensor with saturating Cdc42-GTPγS). (C) Binding of the biosensor to Cdc42 placed the dye attached at position 271 (red) in a hydrophobic pocket formed by amino acids from both WASP (blue) and Cdc42 (orange). The CRIB domain is shown in green. (D) The biosensor showed a concentration-dependent fluorescence increase in response to Cdc42-GTPγS, but did not respond to Rho-GTPγS or Rac-GTPγS. (E) Human neutrophils were lysed at different times after stimulation with fMLP, and biosensor was added to the lysates. Circles, fluorescence intensity in lysates at each time point; triangle, unstimulated lysate equilibrated with GTPγS.

Figure 1E shows that the labeled WASP fragment provides a straightforward means to assay Cdc42 activation in cell lysates. By simply adding the biosensor to the lysate, one can obtain a fluorescence readout of Cdc42 activation. The method was used to determine the kinetics of Cdc42 activation in neutrophils after stimulation with chemoattractant fMetLeuphe (fMLP) peptide. Results paralleled those previously reported using well-established methods (8). Fluorescence of the biosensor was also used to monitor the real-time kinetics of Cdc42 GDP/GTP exchange in vitro (fig. S2D).

Because the labeled WASP fragment responded to activated Cdc42 through fluorescence intensity modulation, a ratiometric imaging approach was used to correct for effects of varying cell thickness, uneven illumination, and other factors that could affect imaging of dye intensity (9). The biosensor was fused to enhanced green fluorescent protein (EGFP) to provide a fluorescence signal insensitive to Cdc42 binding, but with the same subcellular distribution as the sensitive dye. The dye image could be divided by the EGFP image to normalize changes in dye intensity not originating from Cdc42 binding. A proline-rich region of WASP (amino acids 315 to 321) was also deleted to preclude possible binding to proteins containing SH3 domains. Fluorescence response and Cdc42 binding of the biosensor remained intact after these modifications (fig. S2A). It was named MeroCBD, for the combination of the Cdc42 binding domain with a merocyanine dye.

The biosensor was injected into living fibroblasts, where it showed localized Cdc42 activation even in unstimulated cells, which was highest at cell extensions (Fig. 2A). In cells expressing constitutively active Cdc42-Q61L (10), the overall levels of activity shown by MeroCBD were much higher (Fig. 2, A and B), and activation was distributed throughout the cell. It was important to show that the dye was not binding nonspecifically to membranes or to other hydrophobic cell components that could produce spurious fluorescence intensity increases. When MeroCBD was compared to a control biosensor with severely reduced Cdc42 binding (Fig. 2B; fig. S2, B and C), the mutant biosensor showed no localized activation for either endogenous or dominant positive Cdc42 (Fig. 2A), and showed only slightly increased total activity in cells expressing Cdc42-Q61L (Fig. 2B). Simple localization of the CBD-EGFP was not sufficient to reveal Cdc42 activation (Fig. 2C).

Fig. 2.

Activation of endogenous Cdc42 in living cells. (A) Distribution of activated Cdc42 in MEF/3T3 cells. Cells were injected with MeroCBD (upper panels) or with mutant control biosensor (lower panels). MeroCBD showed localized activation of endogenous, wild-type Cdc42 in unstimulated cells (left column). Overexpression of constitutively active Cdc42-Q61L (right column) led to increased activation signal throughout the cell. Results were unaffected by biosensor concentrations over the broad range examined (EGFP fluorescence per unit area = 50 to 400, normalized for exposure time). (B) Average Cdc42 activity in cells with and without expression of Cdc42-Q61L (with Cdc42-Q61L = mean ± SEM: 12 MeroCBD cells, 7 control cells; without Cdc42-Q61L = mean ± SEM: 7 MeroCBD cells, 11 control cells). Although overall activation was similar in unstimulated cells loaded with real versus control biosensor, differences could be discerned because activation was localized. (C) EGFP and dye images used to produce the ratio in the cell shown at upper left in (A).

The ability to detect endogenous protein with the high sensitivity provided by the dye was important in studying Cdc42. High sensitivity enabled detection of protein activation at native concentrations, unlike previous fluorescence resonance energy transfer (FRET) biosensors that required overexpression of Cdc42 (11), and showed more uniform activation. MeroCBD did not require modification of the Cdc42 terminus with a GFP mutant for FRET (12), thus maintaining normal regulation by guanosine dissociation inhibitors (GDIs) (13).

Cdc42 is known to be important for maintaining cell polarity in motility (14), but the role of localized Cdc42 activation is poorly understood. Cdc42 promotes leading-edge extension through activation of Rac and of WASP, which causes Arp2/3 to nucleate actin filaments (15, 16). It also induces the fine cell extensions known as filopodia (17, 18). The relative spatiotemporal dynamics of Cdc42 activation, protrusion, and filopodia formation were examined in fibroblasts as they attached and spread on fibronectin (19). At 30 to 45 min after plating, Cdc42 was activated in a thin band at cell edges extending filopodia (Fig. 3A). No activation was observed within the filopodia themselves (Fig. 3C). Regions of lower activation sometimes extended into the cell body at the base of filopodia (Fig. 3C, lower cell), consistent with studies showing that actin bundles in filopodia extend into the cell body (20). At 90 to 120 min after attachment, activity became localized within larger dynamic protrusions, and overall activity increased (Fig. 3B); protrusions had more than twice the average activity of any other region (n = 22 cells) (Fig. 3, A and D) (movies S1 to S3; see also motion control fig. S6). Controls showed that biosensor levels did not perturb spreading or motility (figs. S3 to S5).

Fig. 3.

Cdc42 activation during cell adhesion and spreading. (A) Cdc42 activation was examined at different times after MEF/3T3 cell adhesion to fibronectin. Activity was first distributed in a narrow band around the cell, then concentrated in the larger protrusions that formed at later time points (n = 11 cells at 30 to 45 min, n = 22 cells at 90 to 120 min). (B) Total integrated Cdc42 activity divided by cell area (average cellular activation) was higher during later stages when cells were polarized and producing large extensions (t = 30 to 45 min: **P < 0.01, n = 11 cells; t = 90 to 120 min: ***P < 0.001, n = 12 cells). Error bars represent the SEM. (C) Activation was observed at the cell edges near filopodia, but not in the filopodia themselves (n > 300 filopodia). Areas of low activation sometimes extended into the cell at the base of filopodia (lower panel). (D) Active remodeling of the cell perimeter was associated with high levels of Cdc42 activation (movies S1 to S3). (E) Cells were treated with nocodazole (30 μM, n = 7 cells), cytochalasin D (300 ng/ml, n = 9 cells), or dimethyl sulfoxide (DMSO) control (n = 6 cells). Average whole-cell Cdc42 activity was monitored before and after drug treatment (at the time indicated by the arrow). Error bars represent the SEM. See movies S4 to S6.

Microtubules or actin may direct Cdc42 activation to specific peripheral locations (2124). We explored this possibility by treating cells with the microtubule-depolymerizing agents nocodazole and colchicine, or with cytochalasin D, an inhibitor of actin polymerization. Only nocodazole and colchicine markedly affected peripheral Cdc42 activation (Fig. 3E; fig. S9, movies S4 to S6). Microtubules may localize interactions between Cdc42 and guanine nucleotide exchange factors (25), by directing vesicle trafficking or regulating events at adhesion complexes (26, 27). Cdc42 has been implicated in intracellular trafficking (2830). Our biosensor consistently showed activation in the trans-Golgi apparatus of endothelial cells (Fig. 4, A to C), and sometimes also in fibroblasts. Activation in this major secretory compartment suggests that Cdc42 regulates directional sorting or trafficking of polarity cues, or that microtubules mediate trafficking of activated Cdc42 to specific portions of the periphery (Fig. 3E).

Fig. 4.

Cdc42 activation at trans-Golgi. Activation kinetics synchronized with extension and retraction. (A) Immunofluorescence colocalization of trans-Golgi markers and Cdc42 activation in human umbilical vein–derived endothelial cells (HUVEC). Trans-Golgi fluorescence is in red, Golgi fluorescence is in yellow. EGFP fluorescence is in green. (B) Mutant control biosensor in fixed HUVEC cells. (C) Upon collapsing the Golgi using Brefeldin A, the trans-Golgi and Cdc42 activation remained intact. (D) For individual protrusions, area and Cdc42 activation per unit area were monitored over time. Of 22 cell regions analyzed, 17 showed high synchrony between protrusions and Cdc42 activity during both extension and retraction, as was apparent through inspection of the graphs and using cross-correlation analysis (32). Data from a representative protrusion are shown. Images correspond to the time points indicated in the plots. All other data are shown in fig. S7. (E) When cells were treated with Rho kinase inhibitor Y27632 (30 μM, at the time indicated by the arrow), cells continued to extend even after Cdc42 activation receded (see also fig. S8 and movie S7).

Using the dye's ability to obtain more than a hundred sequential images at low biosensor concentrations, we carried out high-resolution kinetic studies of Cdc42 activation during extension and retraction of individual protrusions (Fig. 4D; fig. S7). Using an algorithm to objectively determine the boundaries of protrusions, we plotted the changing areas of individual protrusions against the protrusions' total activation per unit area. The rise and fall of Cdc42 activity was markedly correlated with both extension and retraction. This close correlation suggested that Cdc42 activation and deactivation could be rate-determining steps for extension and retraction. Alternatively, upstream signals might coordinately inhibit Cdc42 activity while inducing retraction. We distinguished these possibilities by blocking retraction using an inhibitor of Rho kinase (Y27632). This caused protrusions to continue expanding even after Cdc42 activity decreased (Fig. 4E; fig. S8 and movie S7), indicating that upstream signals (possibly regulated by the microtubule cytoskeleton) control Cdc42 activity and retraction in parallel. Cdc42 activity did not remain elevated during protrusion, suggesting that Cdc42 initiates rather than maintains extension.

MeroCBD exemplifies a biosensor approach that combines the ability to sense endogenous molecules with the sensitivity provided by direct excitation of a fluorescent dye. This extends our ability to examine proteins that cannot be derivatized or overexpressed for live cell studies, and enabled detailed kinetic analysis of rapid cellular processes. The biosensor revealed Cdc42 activation in the trans-Golgi compartment, microtubule-dependent activation at the cell periphery but not in filopodia, and tightly coordinated kinetics of cell extension, retraction, and Cdc42 activation.

Supporting Online Material

Materials and Methods

Figs. S1 to S9

Movies S1 to S7


References and Notes

View Abstract

Navigate This Article