ERK Regulates Lamellipodial Protrusion
We determined if ERK activity was required for EGF (epidermal growth factor)-stimulated protrusion in normal HMECs (human mammary epithelial cells) using time-lapse microscopy. In HMECs, EGF is a strong activator of the ERK-MAPK pathway and a weak activator of PI(3)K, as assessed by phospho-ERK and phospho-Akt levels (). We manually quantified the EGF-induced change in HMEC cell surface area as a read-out of cell protrusion and retraction. Unmodified HMECs exhibited minimal protrusive activity (, V). Because the small GTPase Rac is required for lamellipodial protrusion (Heasman and Ridley, 2008
), we stably infected and selected the HMECs with virus expressing PuromycinR
and constitutively-active Rac mutant Q61L. RacQ61L
expression did not affect the levels of active ERK (phospho-ERK), excluding Rac regulation of ERK via the PAK–MEK pathway (). EGF induced a protrusion and retraction cycle in HMEC/RacQ61L
cells that closely mimicked the published dynamics of murine mammary adenocarcinoma cells (, (Segall et al., 1996
)). Pre-treatment of HMEC/RacQ61L
cells with a MEK inhibitor (U0126) completely abrogated their protrusion (). Thus, ERK activity is required for cell protrusion.
ERK-MAPK Activity is Required for Protrusion
We further dissected ERK's role in protrusion by quantifying the effects of ERK inhibition on the exploratory dynamics of growing PtK1
cells (Machacek and Danuser, 2006
). We used GFP as a cytoplasmic marker and edge tracking software to calculate the percentage and speed of protrusion and retraction. U0126 pre-treatment abrogated protrusion and caused the cell edges to exhibit overall retraction throughout a 10 min timelapse (). At any given time, 44% of control-treated PtK1
cell edges were protruding, while only 7% of the U0126-treated membranes protruded. Strikingly, the remaining protrusions in MEK-inhibited cells exhibited significantly reduced instantaneous protrusion velocity (7.8 +/− 0.8 nm/s for control cells versus 2.2 +/− 0.4 for U0126-reated cells, means and SEM for 5 cells, p=0.0003). In contrast, retraction velocity was not greatly affected (5.5 +/− 2.3 nm/s for control and 6.2 +/− 2.4 for U0126, means and SEM for 5 cells, p=0.7). Thus, though ERK inhibition results in overall retraction, this is due to down-regulation of protrusive activity rather than an increase in contractility.
We hypothesized ERK might signal through the Arp2/3 machinery to promote protrusion. We first determined if ERK promotes cell protrusion by augmenting ERK activity in HMECs. Cells expressing hyperactive B-RafV600E
or over-expressing the EGFR (EGF receptor) exhibited enhanced EGF-stimulated cell protrusion, similar to constitutively active RacQ61L
(Figures S1A and S1B
). We used these cells to address the requirement for Rac and the WRC in ERK-driven protrusions. In both HMEC/B-RafV600E
and HMEC/EGFR cells, dominant-negative RacT17N
blocked EGF-stimulated protrusion without affecting phospho-ERK levels, although RacT17N
did reduce EGFR expression (). When we reduced the levels of individual WRC components with pooled or individual siRNAs, the abundance of the other WRC complex members was also reduced ( and S2A
), likely via proteasomal degradation (Kunda et al., 2003
). Loss of these WRC components blocked HMEC/EGFR protrusion similar to reduction of ERK (). We also repeated the experiment with Alexa555-labeled siRNA and analyzed only the Alexa555-labeled transfected subpopulation (Figures S2B and S2C
). Again, loss of WRC components significantly inhibited protrusion. Thus, we conclude these WRC subunits are required for ERK-driven protrusion in the HMEC system.
Rac Activity and the WRC are Required for Lamellipodia Protrusion
ERK Co-localizes with and Directly Phosphorylates the WRC
We hypothesized that molecular cooperation between ERK and the WRC would require them to be in the same place at the same time during lamellipodial protrusion. We assessed their localization in HMEC/EGFR cells 2.5 min after EGF stimulation, when the protrusions are still growing. At this time point, HMECs retained large amounts of cortical actin and exhibited small areas of ruffled edges and cytoplasmic pERK. 76% (19/25) of randomly-selected cells exhibited peaks in the intensity of WAVE2 stain at the edge, suggesting these cells had responded to the EGF and translocated WAVE2 to the membrane. 21% (4/19) of these protruding cells exhibited pERK staining preferentially at the cell edge (, S3A, and S3B
). While these results are not surprising given that ERK regulates many disparate biological processes in different subcellular locales (McKay and Morrison, 2007
; Shaul and Seger, 2007
), we aimed to confirm any WAVE2 and pERK co-localization during protrusion in a system with less heterogeneity. We thus analyzed pERK and WAVE2 localization in Cos-7 cells, which exhibit a more uniform protrusion response to EGF: large lamellipodia with WAVE2 localized to the edge and minimal cortical actin (, 25/25 randomly-selected cells). Measuring the intensity of each stain with respect to distance from the cell edge verified that both WAVE2 and phospho-ERK intensities peaked at the edge and decreased coincidently with the actin intensity peak ( and S3B
, 24/25 cells). pERK intensity then increased further inside the cell, consistent with active ERK localization to the cytoplasmic and nuclear compartments (, S3A, and S3B
, (McKay and Morrison, 2007
Active ERK and WAVE2 Co-localize and Are Required for Lamellipodial Protrusions
We assessed the co-localization of WAVE2 and phospho-ERK via an object-based co-localization method (Lachmanovich et al., 2003
). We calculated the probability of finding WAVE2 stain punctae within 0.2 μm (the resolution of diffraction-limited microscopy) of any given pERK stain puncta. This probability was 25% (+/− 3% SEM) higher than a random distribution of equally dense points for HMECs, and 28% (+/− 4% SEM) higher for Cos-7 cells, indicating they are significantly closer together than would occur by chance. The relative probabilities of the distance between pERK and WAVE2 decreased to 1 (equal to that of chance) and then to less than 1 (less likely than chance) at further distances (). Thus, WAVE2 and pERK indeed co-localize.
In both systems, ERK inhibition caused a re-structuring of actin into stress fibers and large punctate structures reminiscent of adhesions (). However, ERK inhibition did not impact WAVE2 localization to the cell edge. 6/10 HMECs and 9/10 randomly-selected Cos-7 cells pre-treated with U0126 exhibited peaks in WAVE2 intensity at the cell edge (). We further confirmed ERK localizes to the edge of lamellipodial protrusions by co-staining with Paxillin, a component of lamellipodial nascent adhesions. Phospho-ERK stain overlapped with lamellipodial nascent adhesions and also large mature adhesions, consistent with previous studies (Fincham et al., 2000
; Ishibe et al., 2003
; Vomastek et al., 2007
) ( and S3C
). Inhibition or loss of ERK or WAVE2 caused indistinguishable cellular phenotypes: the edges lacked any sign of protrusion and were instead surrounded by large adhesions ( and S3C
We hypothesized ERK cooperates with the WRC in cell protrusion by regulating WRC activation. ERK has been suggested to both phosphorylate (Danson et al., 2007
) and not phosphorylate (Lebensohn and Kirschner, 2009
) WAVE2. We tested whether ERK phosphorylates the WRC by in vivo
labeling and in vitro
kinase assay. When the WRC was immunoprecipitated from 32
P-labeled EGF-stimulated HMEC/EGFR cells, ERK-dependent phosphorylation was detected on WAVE2 and Abi1, but not Nap1 or Sra-1 (). In vitro
, GST-ERK2 phosphorylated WAVE2 in the WRC, but not un-complexed WAVE2 (). Further, ERK2 phosphorylation of WAVE2 and Abi1 increased with increasing amounts of Nap1/Sra-1 in the WRC substrate ().
ERK Phosphorylates WAVE2 and Abi1
In order to identify the sites that ERK phosphorylates on WAVE2 and Abi1, we used mass spectrometry to analyze endogenous WRC purified from EGF-stimulated HMEC/EGFR cells and in vitro
phosphorylated WRC purified from 293F cells (Figures S4A and S4B
). ERK interacts with its scaffolds and substrates via ERK's CD (common docking) and/or DBP (DEF motif binding pocket) binding to D domains and DEF motifs within the interacting protein (Dimitri et al., 2005
; Lee et al., 2004
). We used ERK2 mutants that cannot bind D (D319N) domains or DEF (Y261A) motifs as negative controls for the in vitro
-phosphorylated WRC mass spectrometry analysis, as they exhibited minimal phosphorylation of the WRC (). We counted the number of times the spectra for each site are observed in the different ERK2 samples, as it is suggestive of their relative abundance.
S308, S343, T346 and S351 were previously inferred to be ERK phosphorylation sites, as their point mutation abrogates growth factor-induced WAVE2 phosphorylation and bandshift (Danson et al., 2007
; Nakanishi et al., 2007
). S482 and S484 were also originally reported to be ERK sites (Nakanishi et al., 2007
), but have since been shown to be phosphorylated by Casein Kinase 2, not ERK2 (Pocha and Cory, 2009
). S482 and S484 are followed by Asp and Glu residues, creating motifs selected against by ERK2 and preferred by Casein Kinase 2 (Hutti et al., 2004
). Our mass spectrometry analysis identified pS296 in both the wildtype and D319N ERK2 samples. We found WAVE2 pS308 only in the wildtype ERK2 sample (Figure S5 and Table S1). This suggests WAVE2 S296 and S308 are phosphorylated in an ERK-dependent manner. S343, T346 and S351 have not been reported in any mass spectrometry analysis, likely because they reside in a poly-proline rich region refractory to tryptic and chymotryptic digestion. We made antibodies to phospho-S308, -S343, -T346 and -S351 and confirmed they specifically detected their respective sites using alanine mutants as controls (). We used these antibodies to test whether ERK regulated phosphorylation of these sites in vivo
. In HMECs, each site exhibited U0126-sensitive phosphorylation with EGF stimulation (). Cos-7 cells exhibited ERK-dependent phosphorylation of S343, T346 and S351, but not S308. The S308, T346, and S351 sites exhibited relatively high basal phosphorylation, suggesting residual ERK activity or a different Pro-directed kinase maintains their phosphorylation in the starved state. In Cos-7 cells, mutation of the three common WAVE2 sites to Ala residues (3A) abrogated EGF-induced U0126-sensitive in vivo
labeling, although total basal phosphorylation remained high (Figures S4C
). Mutation of S482 and S484 to Ala did not affect EGF-inducible U0126-sensitive phosphorylation. Thus, pS343, pT346, and pS351 account for the majority of ERK-induced phosphate transfer to WAVE2.
Identification of Multiple ERK Phosphorylation Sites on WAVE2 and Abi1
Our mass spectrometry also identified Abi1 S183 and S225 on Abi1 (numbering corresponds to Abi1 isoform 1) as sites phosphorylated on endogenous protein and in the wildtype ERK-dependent in vitro
phosphorylated sample (Figure S6 and Table S1). pS216 was also a suspected, although not definitive, site (data not shown). These sites were not found in the ERK2 CD or DBP mutant samples. Interestingly, 5 additional Abi1 residues are predicted ERK-phosphosites (T265, S267, S392, T394 and S410) (Obenauer et al., 2003
). These additional sites also lie in regions refractory to tryptic and chymotryptic digests. S392 or T394, and S410 were recently identified using Asp-N digestions (Ross Tomaino, personal communication) (Lebensohn and Kirschner, 2009
). We mutated each of these sites individually and assayed their contribution to the in vitro
ERK phosphorylation of GST-Abi1. Although no one site contributed significantly to the overall ERK phosphorylation of Abi1, mutation of some sites did induce a mobility shift in the Abi1 protein that is consistent with reduced phosphorylation (, S216A, T265A and T394A). We combined mutation of the 3 sites identified by our mass spectrometry with mutation of T265 and S410, a site identified in previous mass spectrometry experiments of the WRC (Lebensohn and Kirschner, 2009
). ERK phosphorylation of this 5A mutant was reduced, but not abolished (). When the remaining 3 Scansite-predicted sites were additionally mutated, the Abi1 8A mutant exhibited only residual (0.2% of wild-type) phosphorylation by ERK2 (). To further validate that this sheer number of sites were truly phosphorylated by ERK, we expressed and immunoprecipitated the 5A and 8A mutants from 32
P-labeled Cos-7 cells (). While the 5A retained EGF-inducible U0126-sensitive phosphate transfer, basal and EGF-induced phosphorylation of the 8A mutant were abolished. We made antibodies to phospho-S225 and confirmed its recognition of wildtype but not alanine-mutated Abi1 (). The antibody also detected U0126-sensitive phosphorylation of endogenous Abi1 in both HMECs and Cos-7 cells ().
ERK Phosphorylation of the WRC Regulates Interaction with Arp2/3 and Lamellipodial Protrusion
We sought to understand how ERK phosphorylation of WAVE2 and Abi1 might regulate Arp2/3 activity and cell protrusion. Our immunofluorescence data indicate ERK does not regulate WRC activation by controlling its localization to the membrane (). We therefore hypothesized ERK promotes WRC activation by regulating its allosteric activation. All the ERK phosphorylation sites on WAVE2 and Abi1 reside within their respective Pro-rich domains (PRD), unstructured domains that were genetically deleted in the WRC crystal structure (Chen et al., 2010
). Phosphorylation of the PRDs would disrupt their interaction with SH3 and PLP binding domains, potentially altering WRC activation.
We asked whether ERK phosphorylation of WAVE2 and Abi1 could regulate WRC interaction with Arp2/3 or actin. We immunoprecipitated the WRC from 293T cells and found EGF induced an ERK-dependent increase in Arp2/3 and actin co-immunoprecipitation (). Consistent with these data, mutating all 4 WAVE2 and 8 Abi1 sites to phospho-mimetic Asp/Glu residues (WRC-D/E) increased Arp3 and actin co-immunoprecipitation in the starved state above that seen with wildtype WRC under EGF-stimulation (). The increased WRC-D/E interaction with Arp2/3 and actin was not further induced with EGF stimulation. These data suggest EGF-induced ERK phosphorylation of the WRC promotes WRC interaction with Arp2/3 and actin. Somewhat surprisingly though, mutation of the sites to Ala's (WRC-A) also increased basal WRC co-immunoprecipitation with Arp3 and actin, though not to extent of the phospho-mimetic mutations. We then dissected which phosphorylation sites contribute to the basal and regulated WRC interaction with Arp2/3 and actin by assaying the individual WAVE2 and Abi1 phospho-site mutants. Mutation of the WAVE2 sites alone induced un-regulated WRC interaction with Arp2/3 and actin similar to that of WRC-A (). In contrast, mutation of the Abi1 sites reduced both basal and EGF-stimulated WRC interaction with Arp3 and actin.
ERK Phosphorylation of WAVE2 and Abi1 Regulates WRC Function
These data indicate ERK phosphorylation of Abi1 is required for basal and EGF-induced WRC interaction with the Arp2/3 complex. In addition, the basal phosphorylation of WAVE2 found in the starved U0126-treated states might be required for a conformation that permits a regulated release of Arp3 and actin. We hypothesized the de-phosphorylated WRC-A might function as a dominant negative by forming a more stable and un-regulatable interaction with Arp2/3, thereby preventing WRC and Arp2/3 cycling on and off of the expanding lamellipodial actin meshwork. In contrast, hyper-phosphorylated mimic WRC-D might cycle through binding and releasing Arp2/3 and actin faster than hypo-phosphorylated WRC, producing more productive Arp2/3 activation and actin polymerization.
To test this, we purified wildtype and mutant WRC from log-phase mammalian 293F cells, permitting the complexes to aquire activating phosphorylations found in steady-state cells (Lebensohn and Kirschner, 2009
). The purified wildtype WAVE2 and Abi1 consisted of mixed populations of phospho-proteins (). In contrast, the WRC-A (4 WAVE2 and 8 Abi1 sites mutated to Ala's) and WRC-D/E (4 WAVE2 and 8 Abi1 sites mutated to phospho-mimetic Asp/Glu residues) complexes formed sharp bands of unphosphorylated or phospho-mimetic protein, respectively. WAVE2 concentration is 30–50 nM in wildtype cells (Suetsugu et al., 2006
) and we microinjected a volume of 200 nM WRC that roughly equaled 30% of HMEC/EGFR cytoplasmic volume, for a final concentration of ~67 nM purified WRC. Cells microinjected with buffer or wildtype WRC exhibited EGF-stimulated lamellipodium protrusion that did not retract within the 15 min timecourse (). Strikingly, the cells injected with WRC-A did not protrude. This suggests that while the WRC-A protein can bind Arp2/3 and actin, it functions as a dominant negative by inhibiting proper Arp2/3 activity and actin polymerization cycles. In contrast, the cells injected with WRC-D/E exhibited more extensive protrusion (Figure 7B), suggesting the increased Arp3 and actin binding observed with the phospho-mimetic mutations hyper-activated WRC activation of Arp2/3 and actin polymerization.
We further assayed the function of ERK phosphorylation of the WRC in Cos-7 cell migration, which involves cell polarization, edge protrusion and adhesion and cell body translocation. Depending on the cell type, an increase in instantaneous protrusion may lead to an increase in migration speed. However, excessive protrusion will slow down motility, as the protrusive forces overcome the cells' ability to contract and move the cell body (Ji et al., 2008
). Under growing conditions, Cos-7 cells move with an average speed of 0.26 +/− 0.02 μm/min (means and SD for 30 cells over 10 hours). Inhibition of ERK with U0126 reduces migration speed (0.17 +/− 0.01 μm/min, p=0.0004). We simultaneously reduced endogenous WAVE2 and Abi1 levels with Alexa555-labeled siRNA and rescued their expression with the wildtype and point mutant combinations by co-transfection (Figure S7A
). Wildtype (WT) WAVE2 and Abi1 rescued the reduction in cell speed caused by wave2 and abi1 siRNA (Figure S7B
). As expected, co-transfection of WAVE2 3A and Abi1 8A failed to rescue, while WAVE2 3D and Abi 8D increased migration speed above that of wildtype. Interestingly, WAVE2 3A and 3D also rescue cell motility, indicating that in the context of wildtype Abi1 and residual wildtype WAVE2, WAVE2 phosphosite mutants do not impact cell polarization, protrusion, adhesion, and cell body translocation. In contrast, expression of Abi1 8A does not rescue migration while Abi1 8D increases migration speed (Figure S7B
). Thus, ERK phosphorylation of Abi1 is essential for functional WRC during cell motility.