Research ArticleCell MotiIlity

ERK reinforces actin polymerization to power persistent edge protrusion during motility

See allHide authors and affiliations

Sci. Signal.  19 May 2015:
Vol. 8, Issue 377, pp. ra47
DOI: 10.1126/scisignal.aaa8859


Cells move through perpetual protrusion and retraction cycles at the leading edge. These cycles are coordinated with substrate adhesion and retraction of the cell rear. We tracked spatial and temporal fluctuations in the molecular activities of individual moving cells to elucidate how extracellular signal–regulated kinase (ERK) signaling controlled the dynamics of protrusion and retraction cycles. ERK is activated by many cell surface receptors, and we found that ERK signaling specifically reinforced cellular protrusions so that they translated into rapid, sustained forward motion of the leading edge. Using quantitative fluorescent speckle microscopy and cross-correlation analysis, we showed that ERK controlled the rate and timing of actin polymerization by promoting the recruitment of the actin nucleator Arp2/3 to the leading edge. These findings support a model in which surges in ERK activity induced by extracellular cues enhance Arp2/3-mediated actin polymerization to generate protrusion power phases with enough force to counteract increasing membrane tension and to promote sustained motility.


Cell movement is essential to many biological phenomena, including embryogenesis, wound healing, and cancer metastasis. The motility process involves cycles of membrane protrusion and retraction at the leading edge, which are coordinated in space and time with adhesion dynamics and cell rear retraction (1).

In migrating epithelial sheets, the rate of edge protrusion is driven by the rate of F-actin assembly (2). A dendritically branched polymer network grows against the leading-edge plasma membrane and turns over within 1 to 4 μm from the cell edge, which defines the lamellipodium (3, 4). The seven-subunit Arp2/3 protein complex mediates nucleation of this branched actin filament assembly. The WAVE regulatory complex activates Arp2/3 (5, 6) and is recruited along with Arp2/3 to the edge of expanding protrusions (79). Rac and phospholipid binding recruit the WAVE regulatory complex to the plasma membrane (1013).

We have previously proposed a model in which protrusion initiation is followed by a power phase of increased actin filament assembly (we calculated power output from the product of the cell boundary force and the cell-edge motion) (14). We have proposed that as membrane tension increases during edge advancement, the power phase is terminated by a maximal tension level that exceeds the amount of propulsion and adhesion stress produced by the combined assembly of actin filaments and nascent adhesions. In this scenario, protrusion cycle duration is directly related to the efficiency with which actin filament assembly is increased after protrusion initiation. Biochemical mechanisms involving signaling proteins likely contribute to the force and tension-based control. For example, the Rac exchange factor β-PIX and the Rac-recruited Arp2/3 inhibitory molecule Arpin create positive and negative feedback loops for lamellipodial actin polymerization that control protrusion and retraction cycles (15, 16). How extracellular signals feed into and perturb the force and control of protrusion cycle timing is largely unexplored.

Myriad signaling inputs from growth factors, hormones, neurotransmitters, and chemokines feed into the cell migration machinery. One of the chief transducers of signals is extracellular signal–regulated kinase (ERK), a mitogen-activated protein kinase (MAPK) (17, 18). ERK is activated by the small guanosine triphosphatase (GTPase) Ras, which recruits the Ser/Thr kinase Raf to the plasma membrane for activation. Raf phosphorylates and activates the kinases MEK1/2 (MAPK kinase 1/2), which activate ERK1/2 (17, 18). Hereafter, we use MEK to refer to MEK1/2, and ERK to refer to the ERK1/2 isoforms. ERK activity is necessary for epithelial sheet and tubule movement, forms of cell migration common during embryogenesis, wound healing, and cancer metastasis (1921). Reports on ERK’s role in migration include transcription-dependent induction of epithelial-to-mesenchymal transition (22, 23) to direct regulation of actin polymerization and focal adhesions (2426).

We have previously found that ERK phosphorylation of the WAVE regulatory complex promotes the interaction of WAVE with Arp2/3 (25). ERK inhibition for several hours reduces spontaneous protrusion velocity in model migrating epithelial sheets (25). Here, we asked whether the role of ERK in protrusion could be separated from its transcriptional activity by assaying the immediate effects of acute ERK inhibition. We analyzed fluctuations in edge motion during steady-state motility and discovered that ERK promoted a gain in protrusion velocity and duration. We spatiotemporally resolved ERK’s point of action and found that after protrusion initiation, ERK promoted Arp2/3 accumulation at the cell edge, which drove the increase in actin polymerization for protrusion reinforcement. Thus, ERK signaling creates the assembly power needed to overcome increasing membrane tension as protrusions progress and cells move forward.


ERK promotes cell motility

We assessed the effects of MEK1/2 inhibition on cell movements using AZD6244, a noncompetitive MEK1/2 inhibitor with improved potency and specificity over U0126, an older inhibitor that also inhibits MEK5 signaling to ERK5 (2729). Phosphatases constantly remove MEK’s activating phosphorylations on ERK (30) so that treatment of PtK1 epithelial cells with the MEK inhibitor AZD6244 reduced ERK activation within 1 min (fig. S1A). Phosphorylation of the ERK substrates p90 ribosomal S6 kinase (31) and WAVE2 (32) was lost in 2 to 3 min (fig. S1, A and B). Inhibiting MEK with the compound AZD6244 reduced single-cell migration distance in many cell types, including Panc1 pancreatic cancer cells harboring an activating Kras mutation, MDA-MB-468 breast cancer cells harboring amplification of the EGFR gene (which encodes the epidermal growth factor receptor), and Sk-Mel-28 and A375 melanoma lines harboring the activating BrafV600E mutation (Fig. 1A). MEK inhibition also reduced the epithelial sheet migration of untransformed PtK1 cells closing a wound (Fig. 1B), a process driven by edge protrusion and adhesion rather than actomyosin contraction (2). Dimethyl sulfoxide (DMSO)–treated sheets moved at a median velocity of 0.328 μm/min, whereas AZD6244-treated sheets moved at 0.198 μm/min (P = 0.0081).

Fig. 1 ERK signaling extends membrane protrusion velocity and persistence time.

(A) Displacement distributions of cancer cells migrating after treatment with DMSO or AZD6244 (MEK inhibitor). n = number of cells tracked in three independent experiments. (B) Representative images of the migration of a DMSO- or AZD6244-treated PtK1 monolayer after scratch wounding. (C) Representative overlay of single cell-edge dynamics tracked for 10 to 20 min after treatment. Blue, early time points; red, late time points. (D) Representative profile (one of five cells in each treatment group) of the fraction of the cell edge in protrusion, retraction, and quiescent states. (E and F) Velocity (E) and persistence time distributions (F) of m significant protrusion events in n = 5 DMSO- and AZD6244-treated cells tracked in three to four independent experiments. The “>75th Percentile” graphs plot the distribution of the top 25% of events in the corresponding “All” graphs. For (A), (E), and (F), gray area indicates total smoothed distribution, and the boxes’ upper and lower edges represent the 75th and 25th distribution percentiles, respectively. Central horizontal line indicates the median. Error bars about the median indicate 95% confidence interval. (G) Maximum velocity compared to persistence time of all protrusion events tracked in three to four independent experiments. Encircled areas indicate events within the 1st to 75th percentile of the joint velocity and persistence distribution. Shaded areas indicate protrusions within the 1st to 99th percentile.

ERK signaling increases membrane protrusion velocity and persistence time

We imaged the spontaneous edge protrusion and retraction cycles of steady-state (log-phase) PtK1 epithelial cells expressing Emerald-LifeAct as a generic actin cytoskeleton marker (movies S1 and S2). We identified the cell edge and tracked it over time (Fig. 1C), and from these traces, we calculated the velocity along the entire cell edge (movies S3 and S4) (33). We developed a signal detection algorithm to extract the significant protrusion and retraction events from the noisy velocity profile of each edge segment (Materials and Methods). From each detected event, we extracted the event duration and calculated the mean and maximum velocities of that duration. Quantification of the percentage of cell-edge segments in states of protrusion, retraction, and quiescence (when velocity was at 0) revealed a retraction response induced by the physical perturbation of adding reagent to both control and treated cells. DMSO-treated cells recovered their native motion state within 5 to 10 min after treatment, whereas AZD6244-treated cells did not (Fig. 1D). We removed the retraction artifact from the edge velocity analysis by dividing the data into three time intervals: steady state, 1 to 10 min after treatment, and 10 to 20 min after treatment. We found that short-term MEK inhibition reduced the mean protrusion velocity 10 to 20 min after treatment (Fig. 1E, All, and table S1). Although the protrusion velocity of the control cells was statistically different between steady state and 10 to 20 min after treatment, the distributions overlapped by 70%. In contrast, the distribution overlap between steady-state and MEK inhibitor–treated cells after 10 to 20 min was only 20%. The reduction in protrusion velocity after DMSO treatment was likely a residual effect of the retraction artifact induced by physically perturbing the cells. When compared to the protrusion dynamics of control cells treated with DMSO for 10 to 20 min (median of the mean velocities: 12.4 nm/s), MEK inhibition significantly reduced protrusion velocity (median of the mean velocities: 8.1 nm/s; fig. S1C and table S1). MEK inhibition also reduced protrusion persistence time (Fig. 1F, All, and table S1). Thus, MEK inhibition shifted the distribution of edge protrusions to events with slower velocities and shorter durations.

We hypothesized that ERK activity was particularly necessary for the production of faster, more persistent protrusions that have entered the power phase. Indeed, the reduction in both protrusion velocity and persistence upon MEK inhibition became more evident when only the 25% fastest and longest protrusion events were analyzed (Fig. 1, E and F, fig. S1C, 75th Percentile, and table S1). Cells treated for 10 to 20 min with DMSO exhibited a 75th percentile mean protrusion velocity of 25.6 nm/s, whereas cells treated with the MEK inhibitor exhibited protrusions with a 75th percentile mean velocity of 18.9 nm/s. Similar results were found when the data were analyzed using the maximum protrusion velocity reached per protrusion cycle (table S1) and when cells were treated with the structurally distinct MEK inhibitor U0126 and imaged on an independent microscope configuration (fig. S1, E to G). In the latter case, the overall protrusion velocity was slower, which we attributed to lot-specific variability in the imaging medium and changes in the microscope configuration. Nonetheless, U0126 significantly reduced protrusion velocity and persistence. To test whether ERK was necessary for the maximum range of protrusion velocity and persistence, we plotted the maximum velocity against the duration of each protrusion event in steady-state cells and 10 to 20 min after treatment (Fig. 1G). In control cells, 25% of protrusion events reached velocities greater than 50 nm/s and lasted for more than 90 s. Treatment with the vehicle DMSO did not change this distribution, whereas MEK inhibition caused a shift to slower, shorter protrusions, such that only 1% of the treated events reached peak velocities greater than 50 nm/s. These findings were further confirmed in PtK1 cells stably expressing hyperactive BrafV600E, which causes a moderate increase in ERK activity (fig. S2A). The BrafV600E cells exhibited more frequent protrusion events that reached greater mean and maximum velocities than did cells expressing empty vector (fig. S2, B and C). The persistence time was reduced in the 75th percentile BrafV600E events (fig. S2D), suggesting that the more frequent and faster protrusions might be running out of other necessary actin assembly factors to maintain their persistence.

We also found that MEK inhibition reduced and MEK activation promoted retraction velocity (fig. S3, A, B, and E). This is in contrast to our previous data (25), in which we treated cells with the MEK inhibitor for 4 hours and did not remove noise in the velocity analysis. ERK may also promote retraction persistence. Retractions in control cells exhibited an insignificant reduction in persistence time from 32.5 to 30.9 s upon treatment with DMSO for 10 to 20 min, whereas retractions in MEK inhibitor–treated cells exhibited a significant reduction in persistence from 34.6 to 27.4 s (fig. S3C and table S1). However, because of cell-to-cell heterogeneity, the steady states in control and treated groups were not equivalent and the retraction persistence after DMSO and AZD6244 treatments was not significantly different (fig. S3D). Increased ERK activation by BrafV600E expression reduced retraction persistence (fig. S3F), similar to the BrafV600E-induced reduction in protrusion persistence. The regulation of retraction by ERK could be secondary to reduced protrusion, because protrusion and retraction forces are in equilibrium during motility (2) or through direct modulation of the retraction machinery, such as myosin light chain kinase or the adhesion component paxillin (3437).

ERK promotes actin assembly dynamics in protrusions

We hypothesized that ERK regulates protrusion by directly regulating actin assembly in the power phase. Because ERK activation at adhesions contributes to cell migration velocity (34), we aimed to directly assess actin dynamics in situ. We microinjected fluorescent actin and used quantitative fluorescent speckle microscopy (qFSM) to identify speckles of assembled actin monomers and track their motion (38). We imaged cells for 9 min to obtain a measure of steady-state dynamics, added the MEK inhibitor, and imaged the cells again after a 10-min recovery period (movies S5 and S6). We confirmed that these cells had reduced protrusion velocity and persistence time upon MEK inhibition, similar to the cells labeled with LifeAct (39) (fig. S4, A to C).

Using actin qFSM software, we identified speckles in the raw images (Fig. 2A), and we tracked their displacements (Fig. 2B and movies S7 and S8) and used them to identify local events of actin filament polymerization and depolymerization (40). In this method, speckles are assigned assembly and disassembly scores, which are a function of the rate of change in speckle intensity (net number of fluorescent actin monomers that are incorporated or unincorporated into the actin filament) and background fluorescence. In our PtK1 cells, heterogeneous and rapidly changing cell protrusion and retraction states along the cell edge meant that actin assembly and assembly-driven actin retrograde flows were spatially segregated and transient. As previously reported, averaging the actin dynamics over time concealed spatiotemporal definition (41) but revealed typical lamellipodial actin characteristics: rapid retrograde flow (600 to 800 nm/min) toward the cell center and a 1- to 2-μm band of assembly followed by a 1- to 2-μm band of disassembly (4, 42, 43) (fig. S4, D and E). Fluorescence recovery after photobleaching (FRAP) of enhanced green fluorescent protein (EGFP)–actin in B16-F1 melanoma cells has also shown that actin turnover is restricted to within ~2 μm of the cell edge (7). We then removed the nonspecific spatial and temporal averaging and averaged the actin assembly and disassembly scores as a function of distance from the cell edge. We found that during protrusion events, flow rates peaked at the cell edge and continuously decreased over the 2- to 3-μm width of the lamellipodium (Fig. 2C). Actin polymerization peaked at, but was not restricted to, the cell edge, and depolymerization peaked 0.5 μm inside the edge. Treatment with the MEK inhibitor reduced actin retrograde flow and actin assembly and disassembly rates at the cell edge in protruding edges (Fig. 2, C to F). The effect of MEK inhibition on actin flow was likely secondary to its effect on actin polymerization, because less polymerization against the plasma membrane would generate less force converted into retrograde flow (14). The MEK inhibitor–induced reduction in actin disassembly is likely a combination of a shift in the steady-state balance of lamellipodial actin assembly and disassembly and reduced incorporation of Arp2/3, because adenosine triphosphate hydrolysis of incorporated Arp2 and Arp3 contributes to actin network disassembly (44).

Fig. 2 ERK promotes actin flow and assembly in protrusions.

(A) Raw image of fluorescent actin speckles. The red outline defines the computationally segmented cell mask. (B) Instantaneous flow vectors of speckles identified in (A), calculated with qFSM software (38). Longer and red vectors indicate faster flow rate. Scale bar, 5 μm. (C to F) Normalized instantaneous rates of actin retrograde flow (C), polymerization (D and E), and depolymerization (F) calculated as average rate over all identified speckles as a function of distance from the cell edge. Data from a single representative control cell and a MEK inhibitor–treated cell before and after treatment. Shaded areas are 95% confidence interval. On average, within a micrometer from the cell edge, DMSO treatment reduced flow by 18 ± 1.1%, actin assembly by 2 ± 3.3%, and actin disassembly by 9 ± 7.3% (n = 6 cells tracked in four independent experiments). AZD6244 treatment reduced flow rates by 28 ± 0.8%, actin assembly by 31 ± 5.7%, and actin disassembly by 38 ± 3.1% (n = 4 cells tracked in four independent experiments).

ERK regulates actin dynamics during the protrusion power phase

Because cells treated with ERK inhibitor exhibit reduced protrusion velocity and persistence and reduced actin filament assembly rates, we postulated that these cells generated a weaker actin assembly–driven “power phase” and their filament propulsion was outcompeted by plasma membrane tension earlier in the protrusion cycle than in cells with normal ERK signaling. To test whether ERK controlled the timing of actin dynamics with respect to edge motion, we computed the cross-correlation of these two waveforms. Cross-correlation scores indicated the strength of coupling between two waveforms, and the lag of the peak in the score indicated the timing of the strongest correlation. Consistent with our previous results (14), we found that in DMSO-treated cells, actin flow negatively correlated with edge dynamics at time lag = 0 (Fig. 3A, mark 1). Furthermore, actin flow positively correlated with edge motion in a MEK inhibitor–sensitive manner when flow was shifted 50 to 60 s forward in time (Fig. 3A, mark 2 in left panel; this correlation disappears in the right panel).

Fig. 3 ERK regulates actin dynamics during protrusion power phase.

(A) Temporal cross-correlation of actin flow with edge velocity along entire edge, including retractions and protrusions. Marks 1 and 2 indicate negative and positive extrema, respectively. (B and C) Temporal cross-correlation of actin flow specifically in retractions (B) and protrusions (C). Marks 1′ and 2′ indicate that the negative and positive extrema of the combined cross-correlation in (A) are associated with retraction and protrusion events, respectively. (D and E) Temporal cross-correlation of actin assembly with edge velocity in retractions (D) and protrusions (E). Retraction velocity was computed with negative values. Positive time lags in the x axis indicate that actin dynamics were delayed relative to edge motion. The corresponding model diagrams interpret the cross-correlation scores as readouts of temporal coordination between actin dynamics and edge velocity. E.V., edge velocity; R.F., retrograde flow; A.A., actin assembly. Orange arrows indicate how the indicated parameter of actin dynamics is altered relative to the edge velocity upon MEK inhibition. In control DMSO-treated cells, actin assembly correlated with protrusive edge motion with a cross-correlation score of 0.4 and a delay of 20 s. In cells treated with AZD6244 and lacking ERK activity, actin assembly dynamics were less correlated with edge motion (cross-correlation score of 0.2), and this correlation occurred with a delay of only 10 s. (F) Revised model of the sequence of mechanical processes during protrusion and retraction events. Actin retrograde flow correlates with the cell edge twice during a protrusion-retraction cycle: once when peak retraction velocities are attained and once 20 s after protrusion initiation, which coincides with peak actin assembly.

We then deconvolved the correlation plots by calculating the correlation scores in retraction against protrusion events separately. We found that the negative correlation between edge motion and flow was present in retractions, but not in protrusions (Fig. 3, B and C), indicating that retrograde flow was fastest when retraction velocity was peaking (Fig. 3B, mark 1′). In contrast, edge motion and flow positively correlated in the protrusions of DMSO-treated cells, but with a shorter time delay (time lag = 20 s) than when total edge motion was analyzed (Fig. 3C, mark 2′, compared to Fig. 3A, mark 2). We analyzed a synthetic data set (fig. S5A) to test whether edge motion in retractions artificially delayed the positive correlation between actin flow and edge motion in protrusions when both forms of edge motion were included in the analysis. The synthetic total edge motion and flow and retraction motion and flow were negatively correlated at lag = 0 (fig. S5, B and C). However, the positive correlation between edge motion and flow was shifted toward longer time lags when total edge motion was correlated with protrusion motion (fig. S5, B and D). Thus, because the negative correlation was stronger than the positive correlation, its inclusion in the analysis artificially delayed the positive correlation between protrusion motion and actin retrograde flow. We concluded that the rate of retrograde flow peaked 20 s after the maximal protrusion velocity, at the end of the power phase (Fig. 3C).

Because our previous model of the protrusion and retraction cycle was based on correlations with total edge motion (14), we computed the cross-correlation between edge motion and actin assembly rates separately during protrusion and retraction events. No correlation was found during retraction events (Fig. 3D). In protrusion events, the correlation maximum occurred at lag = 20 s (Fig. 3E, left panel), indicating that the highest rate of actin assembly occurred 20 s after maximal protrusion velocity is reached, consistent with our previous report (14). Our new analysis indicated that maximum retrograde flow occurred at the same time as the peak actin assembly during protrusion events. Both were delayed relative to the peak protrusion velocity. Retrograde flow then peaked again when retraction velocity was maximal, likely as a result of a release of adhesions and an increase in actomyosin contraction (14, 45) (Fig. 3F).

We next tested whether ERK activity controls the timing of actin dynamics in the protrusion cycle. MEK inhibition had no effect on the negative correlation between edge motion and actin flow during retraction (Fig. 3B, right panel). In contrast, MEK inhibition abrogated the correlation between edge dynamics and actin retrograde flow during protrusions (Fig. 3C). AZD6244 treatment also reduced the correlation score and shortened the time delay between edge dynamics and actin assembly during protrusion (Fig. 3E). Thus, ERK specifically coordinated the timing of actin polymerization and retrograde flow during edge protrusion, but not retraction. Actin assembly and retrograde flow were reduced in cells with ERK inhibition (Fig. 2), and actin assembly peaked sooner in the protrusion cycle and was less correlated with edge motion than in cells with intact ERK signaling (Fig. 3). We conclude that ERK was necessary for reinforcing actin assembly against increasing membrane tension during the protrusion power phase.

ERK controls the protrusion power phase by modulating Arp2/3 recruitment

We wanted to determine whether the effect of ERK on protrusion speed and persistence was through regulation of Arp2/3-mediated actin polymerization. We hypothesized that ERK controls the intensity of activation of prelocalized WAVE regulatory complex to increase the recruitment and activation of Arp2/3. This is because ERK does not regulate WAVE recruitment to the membrane (25), and although ERK is sequestered in different subcellular locales through scaffolding proteins, passive diffusion allows activated Ras to rapidly spread over 10 μm within seconds (18, 46). Thus, ERK is unlikely to be a localization signal for Arp2/3 recruitment or protrusion initiation. In our previous experiments in which serum-starved cells had been stimulated with EGF, we observed localization of phosphorylated ERK with WAVE2 along the cell edge (25). Here, we examined log-phase migrating cells harboring a low ratio of phosphorylated/unphosphorylated ERK. At any given time, these steady-state cells have edges in states of protrusion, retraction, and quiescence, but immunofluorescence did not show a strong cell-edge localization of active ERK. Rather, phosphorylated ERK was found at the edge and throughout the lamella and cell body (fig. S6A). Phosphorylated ERK was not spatially restricted to areas of active protrusion or retraction, as gauged by edge morphology and Arp2/3 levels (fig. S6A). The specificity of the phospho-ERK–488 signal was demonstrated by the lack of staining in cells treated with the MEK inhibitor (fig. S6, B and D) or in cells that were not incubated with the anti–phospho-ERK antibody (fig. S6C).

We used GFP-tagged Arp3 to determine the effects of ERK inhibition on Arp2/3 localization to the cell edge during steady-state motility (Fig. 4A). We quantified the GFP intensity in temporally aligned profiles (Fig. 4, B to E). MEK inhibition reduced the amount of Arp3 recruited along the profiles, but the reduction was only significant in the profile temporally aligned 20 s after maximum protrusion, during the power phase (Fig. 4D). Thus, the presence of active ERK in initiated protrusions increased Arp2/3 recruitment during the remainder of the protrusion cycle. We used immunofluorescence of ARPC2, a distinct Arp2/3 subunit, as an independent method of validating that ERK regulates Arp2/3 recruitment. We did not expect to find a complete abrogation of Arp2/3 localization to the cell edge upon ERK inhibition because the edges of fixed cells are in various states of protrusion initiation, power phase, retraction, and quiescence. However, we did find a trend in which MEK inhibition reduced overall ARPC2 localization to the cell edge (fig. S6, A, B, and E). We propose that in cells with ERK inhibition, reduced Arp2/3 recruitment led to a loss of correlation between actin and edge dynamics during protrusion.

Fig. 4 ERK controls actin assembly and protrusion power phase by enhancing Arp2/3 recruitment.

(A to E) GFP-Arp3 localization at cell edge. Green arrows indicate protrusion events (A). Normalized mean Arp3 intensity (m = 2779 protrusion events from n = 5 cells tracked in three independent experiments) in cells treated with DMSO (blue) or MEK inhibitor (red) at protrusion initiation (B), point of maximum velocity (C), power phase (~20 s after maximal protrusion velocity) (D), and protrusion end (E). GFP intensity at the cell edge (within 0 to 0.5 μm) was normalized to intensity in the lamella (3 to 3.5 μm). Shaded areas indicate 95% confidence intervals. (F) Cross-correlation of Arp2/3 intensity and cell-edge velocity shows reduced correlation and delayed recruitment with MEK inhibitor. (G) Model of the role of ERK in controlling edge protrusion. After protrusion initiation, ERK controls an intensity rheostat of WAVE regulatory complex (WRC) activation and Arp2/3 recruitment to induce actin polymerization power for sustained events.

Cross-correlation of GFP-Arp3 intensity with edge velocity during protrusion events showed two peaks in Arp2/3 recruitment (Fig. 4F). The first peak indicated that a pool of Arp2/3 was recruited to the cell edge at the same time as the peak in edge protrusion velocity. The second peak indicated that a second, weakly correlated pool of Arp2/3 was recruited 20 s later, which coincided with the peak in actin polymerization rates. The timing of the second Arp2/3 peak might depend upon the first. In addition, we expect that both of these pools of Arp2/3 contribute to the peak in actin assembly in the power phase.

In cells treated with the MEK inhibitor, the peak in Arp2/3 correlation was delayed by 20 s (Fig. 4F). A smaller peak coincided with the peak in edge protrusion, a finding that indicated that recruitment of Arp2/3 to the cell edge took longer when ERK was inhibited. Furthermore, the first pool of Arp2/3 that was recruited at the same time as the increase in protrusion velocity was less correlated with edge motion than in uninhibited cells (cross-correlation score was 0.2 in control cells and 0.1 with AZD6244 treatment). The residual Arp2/3 that was recruited during the shortened, weakened power phase was slightly more correlated with edge motion than the power phase peak in uninhibited cells. This could be explained by the homogenization of protrusion behavior under MEK inhibition. Uninhibited cells had a wide range of protrusion behavior, ranging from slower protrusions that lack a “power” phase to faster, longer protrusions with substantial power phases. In cells treated with the MEK inhibitor, the distribution collapsed to a uniform protrusion cycle of weakened, short protrusions. In conclusion, ERK regulated both the extent and the correlation of Arp2/3 recruitment to edge motion.


Here, we have deduced the mechanisms by which the ERK signaling pathway altered the protrusion and retraction cycles to promote productive migration. We found that the ERK pathway was not essential for protrusion initiation but rather promoted protrusion reinforcement by Arp2/3-mediated actin polymerization. We uncovered this mechanistic insight by cross-correlating signaling readouts with the mechanical processes of steady-state cell migration in defined protrusion and retraction events.

Phospholipids and locally activated Rac recruit and basally activate the WAVE regulatory complex, which activates Arp2/3 (1012). Arp2/3 nucleates actin dendritic branches to accelerate membrane protrusion. Meanwhile, a positive feedback loop between Rac and the actin cytoskeleton recruits more Rac, WAVE regulatory complex, and basally active Arp2/3 (15, 47). We have previously found that peak Rac recruitment occurs 20 s after protrusion initiation (48), which corresponds to our finding here that actin polymerization peaked 20 s after protrusion initiation. Yet, we also found here that a relatively small percentage of protrusion events were “stabilized” with increased protrusion velocity and duration to compel cell displacement. We propose that in these events, ERK localized to the cell edge phosphorylates and activates the WAVE regulatory complex for further Arp2/3 recruitment and activation (Fig. 4G) (25). This effect generates the actin polymerization power needed to overcome increasing membrane tension and to augment protrusion velocity and persistence (Fig. 3F) (14). We speculate that the role of ERK-mediated enhancement of actin polymerization may be more important in three-dimensional migration, a process in which protrusion events tend to experience resistance from the surrounding matrix in addition to membrane tension.

We have previously found that upon growth factor stimulation, active ERK colocalizes with WAVE2 at the cell edge at the time of active protrusion but does not control WAVE2 localization (25). Here, we found that in steady-state migrating cells, active ERK was distributed throughout the cell edge and cytoplasm. These results are consistent with the diffuse cytoplasmic localization of active ERK in migrating epithelial sheets (20). Thus, growth factor stimulation induces active ERK to transiently accumulate at the cell edge before moving to the nucleus, whereas steady-state conditions maintain low levels of active ERK localized at the cell edge and throughout the cell (18, 25, 4951). Our finding that ERK regulated Arp2/3 recruitment is consistent with our previous study showing that ERK-mediated phosphorylation of WAVE2 and of Abi1 controls the ability of Arp2/3 to interact with the WAVE regulatory complex (25). We propose that ERK signaling to the WAVE regulatory complex and Arp2/3 modulates an intensity rheostat of actin polymerization to increase the volume of treadmilling branched actin at the cell edge. In the setting of oncogenic BrafV600E or an influx of growth factor signals, a global increase in ERK activity will lead to an increase in cell edge–localized active ERK. In this model, the presence of active ERK at the cell edge when Rac and the WAVE regulatory complex are recruited would increase the WAVE-mediated recruitment of Arp2/3 and the likelihood that protrusions are stabilized into productive, persistent events that generate motility. In this manner, the WAVE regulatory complex functions as a coincidence detector for ERK and Rac signals to recruit sufficient Arp2/3 for protrusion power.

Our finding that inhibition of a central signaling pathway during steady-state motility causes specific effects on protrusion mechanics suggests that protrusion is controlled by multiple, overlapping points of regulation by additional pathways. Indeed, when these other regulatory pathways are suppressed by starvation, ERK inhibition completely blocks EGF-induced protrusion, including protrusion initiation (25). In the human mammary epithelial cells used in those experiments, EGF is not expected to activate phosphatidylinositol 3-kinase (PI3K)/Akt, Abl, or Src signaling (32), which also participate in WAVE activation, or Rho GTPases, which activate the formin family of actin nucleators and elongators (52). In that less physiological situation involving serum starvation and EGF stimulation, ERK signaling drives protrusion. We found here that in log-phase migrating cells, steady-state ERK signaling specifically regulated protrusion reinforcement. We observed migration defects upon MEK inhibition in various cancer cell lines, and embryonic ERK inhibition causes neural crest cell migration defects that result in craniofacial and heart abnormalities (53, 54). However, the weak, short protrusions that remain after ERK inhibition may be sufficient for motility in some cell types and environments. For example, Arp2/3-deficient fibroblasts lack lamellipodia and undergo filopodia-based protrusion with reduced migration speed (55).

In conclusion, this work elucidates the role of ERK’s direct activity on the actin polymerization machinery: to reinforce the progression of initiated protrusions into rapid, sustained forward motion of the leading edge. ERK-mediated control occurs specifically in protrusion events, not retractions, and involves the recruitment of Arp2/3 for actin polymerization and treadmilling at the cell edge. This point of regulation explains the previously predicted and observed timing sequence in which actin assembly peaks after protrusion onset and after maximal protrusion velocity is reached (14).


Cell culture, imaging medium, and inhibitor treatment

Panc1, MDA-MB-468, Sk-Mel-28, and A375 cells were cultured and imaged in Dulbecco’s modified Eagle’s medium with 10% fetal bovine serum (FBS), 20 mM Hepes. PtK1 cells were cultured in F-12 with 10% FBS and imaged in L-15 with 10% FBS, 20 mM Hepes. AZD6244 (5 μM) was used for all treatments.

Migration and wound-healing assay

Cells were plated in their normal growth medium on glass-bottomed dishes 2 days before assay. Cells were treated with DMSO or AZD6244 and then immediately imaged with phase contrast at 37°C, 5% CO2 using a Nikon Ti inverted microscope with perfect focus, Prior ProScan motorized stage, and incubation chamber. Images were collected every 10 min with MetaMorph software for 12 hours. For migration assay, displacement or the distance migrated over a 6-hour period was calculated using the Track Points function in MetaMorph. To ensure that bias was not introduced into the analysis, every cell that did not divide or move out of the field of view over the 6-hour period was tracked. The results were obtained from three independent experiments. Normality was assessed using the Anderson-Darling test in Matlab. The data were not normally distributed. Therefore, the permutation test was used to calculate statistical significance of changes in the distributions. Error bars about the median were estimated from the 95% confidence interval of the bootstrapped median distribution.

For wound healing, cells were treated with DMSO or AZD6244 immediately after wound scratching. Start and end points of assay were 0 and 11 hours after scratching, respectively. Results were obtained from two independent experiments. Wounds in which the edge did not uniformly migrate were excluded (for example, when only part of the edge migrated or the wound closed from opposite or different part of edge). To avoid issues with the unequal sample size causing type I error inflation (56), we used the Wilcoxon rank sum test to assess the significance of the difference between the control and treated sample medians.

LifeAct imaging and image processing

PtK1 cells were electroporated with Emerald GFP–tagged LifeAct using the Neon transfection system and plated on acid-treated glass coverslips. Cells were cultured for 48 hours at 37°C, 5% CO2 and imaged using a Nikon Ti motorized inverted microscope with Perfect Focus System, Yokogawa CSU-X1 spinning disc confocal and Spectral Applied Research Borealis modification, 491 solid-state laser, 100x 1.45 numerical aperture (NA) objective, Hamamatsu ORCA-AG cooled charge-coupled device (CCD) camera, and MetaMorph software. Cells were in L-15 medium containing 10% FBS, 20 mM Hepes, and Oxyrase (0.03 U/ml). Experiments using U0126 were carried out using a Nikon TE2000 inverted microscope with Perfect Focus System, Yokogawa CSU-10 spinning disc confocal, 488 solid-state laser, 60x 1.4 NA objective, Andor Clara cooled CCD camera, and NIS Elements software.

Cells were imaged for 10 min before perturbation and 20 min after perturbation (30 min total). PtK1 cell boundaries were detected from the fluorescence microscopy data using a threshold method. Once identified, the edge was then subdivided into 1-μm-long segments, and their respective velocities were calculated as described previously (25). For inhibitor-treated experiments, results were obtained from three to four independent experiments. For BrafV600E overexpression experiments, results were obtained from two independent experiments.

Edge velocity analysis

The velocity traces for each segment were processed using a signal detection algorithm on the basis of empirical mode decomposition (57). Velocities within the noise level were classified as insignificant and the respective edge segment as inactive. Once a significant event was detected, it was classified as either protrusion or retraction, and the mean and maximum values were extracted. The event duration was also measured. The Anderson-Darling test indicated that the distributions were not Gaussian. Rather, protrusion and retraction velocities followed a log-normal distribution, which suggested the use of the median value instead of mean for comparisons among different experimental conditions. The median error bars were estimated from the 95% confidence interval of the bootstrapped median distribution.

Protrusion and retraction velocity data exhibited log-normal distributions. We applied the Anderson-Darling test to confirm that the velocity and persistence time for all cells in each protrusion and retraction treatment group came from the same distribution, indicating that they were representative for the same general cell population. The fast and slow, long, and short events were confirmed to be equally distributed among the different cells in all conditions. The permutation test was used to calculate statistical significance of changes in the distributions of the mean velocity and persistence time for the different conditions.

Actin qFSM

Actin microinjection, imaging, and qFSM analysis were carried out as described previously (38). Cells were microinjected with Alexa-568 actin monomers (Molecular Probes) using a Sutter microinjection system. Briefly, cells were allowed to recover for 30 min and were then imaged in L-15 medium with 10% FBS in a 37°C, 5% CO2 incubation chamber on a Nikon Ti motorized inverted microscope with Perfect Focus and Yokogawa CSU-X1 spinning disc confocal with a Spectral Applied Research Borealis modification. Images were acquired every 10 s using a 561-nm (200-mW) solid-state laser excitation and a Hamamatsu ORCA-AG cooled CCD camera controlled by MetaMorph image acquisition software. The actin speckles provided enough out-of-focus light for computational detection of the cell edge and the partitioning of cell-edge segments and associated probing windows. Steady-state actin flow maps depict the raw speckle displacements spatially averaged in 0.7-μm regions and temporally averaged over 4 min.

We identified and tracked >103 speckles per FSM movie. The autocorrelation function of the PtK1 cell-edge dynamics was six frames (60 s). Correlation-based flow tracking revealed the raw speckle displacements spatially averaged in 0.7-μm regions and temporally integrated over 40 s. We then used these flow templates to track the individual speckles and performed statistical processing of the speckle intensity fluctuations to map F-actin flow and assembly and disassembly (kinetics). We averaged 9 frames for steady-state flow analysis and 55 frames for kinetic maps. Average flow at the cell edge varied between 400 and 1000 nm/min between cells, likely because of variability in the fraction of protruding and retracting cell edge and in the protrusion velocity during the captured time series. Reduced flow upon MEK inhibition was reproduced in n = 5 cells, obtained from four independent experiments.

GFP-Arp3 analysis

PtK1 cells were infected with retrovirus carrying pBabeNeo-GFP-Arp3 and selected with G418. Cells were imaged by Yokogawa CSU-X1 spinning disc confocal with Spectral Applied Research Borealis modification, in L-15 medium containing 20 mM Hepes and Oxyrase (0.03 U/ml), with a 100× 1.45 NA objective and a Hamamatsu R2 camera with 2 x 2 binning.

Cells were segmented using a thresholding method, and the Arp2/3 fluorescence intensity was then averaged in 0.5-μm-thick bands parallel to the cell edge at increasing distances from the cell edge. To handle the variability of fluorescence intensity among cells, we normalized the average of each band by the average of the band 3 μm away from the cell edge. This normalization allowed us to average the signal from multiple cells without introducing artifacts from cells with too high or too low intensities. The final confidence interval was calculated by propagating the error from the normalized mean calculated for each cell. Images were obtained from three independent experiments.


PtK1 cells were grown on glass coverslips for 48 hours and fixed with 3.7% formaldehyde with 10 nM calyculin A in phosphate-buffered saline (137 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, 1 mM KH2PO4, pH 7). Cells were permeabilized and blocked with 0.2% Triton and 3% bovine serum albumin (BSA). Cells were stained with purified mouse monoclonal anti–phospho-ERK (Sigma), rabbit anti-ARPC2 (Millipore), and phalloidin-647 (Molecular Probes) in 1% BSA, 0.2% Tween 20 in tris-buffered saline. Images were acquired on a Nikon Ti motorized inverted microscope with Yokogawa CSU-X1 spinning disc confocal, using 405-, 488-, 561-, and 647-nm solid-state laser excitation and an Andor Clara cooled CCD camera controlled by Nikon Elements image acquisition software.

Cells were segmented by thresholding in the phalloidin-647 channel. Because the peak Arp2/3 fluorescence was in front of the phalloidin signal, the segmented masks were then dilated out 0.5 μm from the edge. These dilated masks were then applied to the 488 and 568 channels, and the signal intensities were integrated over the cell area (mask). The ARP2C-568 fluorescence intensity was further averaged in a 1-μm-thick band parallel to the cell edge (0.5 to 0.5 μm), and the percentage of total signal present in this band was averaged for each group. The SEM and 95% confidence interval were graphed.


Cross-correlation analysis can indicate relationships and timing between two events (58). As previously described, correlation scores between the edge motion and sampled signals (for example, Arp3 intensity) were calculated for each window of the cell edge for all cells and then averaged using bootstrap and a variance stabilization method (58). Cross-correlation scores were graphed in the y axis, with the score value indicating how well the two activities were correlated with a specific time lag (x axis) between them. The uncertainty on each correlation value was defined as the 95% confidence interval of the bootstrapped distribution at each time lag.

In silico experiments

We performed in silico experiments to confirm the artificial shift on the positive correlation peak between edge motion and retrograde flow. Time series for both variables were generated using a periodic sawtooth waveform with added Gaussian noise (signal to noise = 2). These waves had 180° phase difference, and a Gaussian random sequence of one quarter of the period was introduced between the peaks for the synthetic retrograde flow. A thousand replicas of both time series were generated, and the correlation scores were calculated as described above. Then, the correlation scores were separately recalculated for the negative and positive component of the edge motion.


Fig. S1. ERK inhibition in PtK1 cells.

Fig. S2. ERK hyperactivation increases protrusion velocity.

Fig. S3. ERK increases retraction velocity.

Fig. S4. ERK promotes peak membrane protrusion velocity and persistence time in PtK1 cells microinjected with Alexa-568 actin monomers.

Fig. S5. Simulation of cross-correlation of actin retrograde flow with cell-edge velocity.

Fig. S6. ERK regulates Arp2/3 localization to the cell edge.

Table S1. Event velocity averages.

Movie S1. Spontaneous protrusion and retraction cycles.

Movie S2. MEK inhibition dampens spontaneous edge motion.

Movie S3. Edge motion analysis.

Movie S4. Analysis of spontaneous edge motion upon MEK inhibition.

Movie S5. Actin speckles.

Movie S6. Actin speckles with MEK inhibition.

Movie S7. Flow tracking of actin qFSM.

Movie S8. Actin flow under MEK inhibition.


Acknowledgments: We thank the Nikon Imaging Center at Harvard Medical School and T. Wittman and A. Sil at the University of California, San Francisco, for help with light microscopy. Funding: This work was funded by NIH grants K01 CA168850 to M.C.M., F32 GM103278 to M.V., R01 GM071868 and U01 GM068230 to G.D., and R01 CA046595 to J.B. Author contributions: The project was conceived by M.C.M. and J.B. The research design was regularly adjusted in discussions between M.C.M., M.V., and G.D. M.C.M. performed all live-cell imaging experiments, Western blotting for the AZD6244 treatments, and routine data analysis using speckle analysis software. J.E.J. performed the Western blotting for the U0126 treatments and the ARPC2 immunofluorescence. M.V. extended the speckle analysis software, wrote new programs for correlation analysis, and, together with M.C.M., interpreted the results in the context of the proposed model. M.C.M. and M.V. jointly prepared the figures. M.C.M. drafted the manuscript. All authors contributed to the final editing. Competing interests: The authors declare that they have no competing interests.
View Abstract

Navigate This Article