During directed cell migration (chemotaxis), cytoskeletal dynamics are stimulated and spatially biased by phosphoinositide 3-kinase (PI3K) and other signal transduction pathways. Live-cell imaging using total internal reflection fluorescence (TIRF) microscopy revealed that, in the absence of soluble cues, 3′-phosphoinositides are enriched in a localized and dynamic fashion during active spreading and random migration of mouse fibroblasts on adhesive surfaces. Surprisingly, we found that PI3K activation is uncoupled from classical integrin-mediated pathways and feedback from the actin cytoskeleton. Inhibiting PI3K significantly impairs cell motility, both in the context of normal spreading and when microtubules are dissociated, which induces a dynamic protrusion phenotype as seen by TIRF in our cells. Accordingly, during random migration, 3′-phosphoinositides are frequently localized to regions of membrane protrusion and correlate quantitatively with the direction and persistence of cell movement. These results underscore the importance of localized PI3K signaling not only in chemotaxis but also in basal motility/migration of fibroblasts.
Cell migration is central to wound healing, immune surveillance, development and cancer. It is characterized by cyclic protrusion, adhesion and contractile processes that are regulated through a complex network of intracellular signaling and cytoskeletal reorganization pathways that are influenced, and in many cases directed, by soluble and adhesion-based factors (Lauffenburger and Horwitz, 1996; Ridley et al., 2003). During wound healing, fibroblasts residing in the nearby tissue sense a gradient of platelet-derived growth factor (PDGF) and exhibit a chemotactic response that accelerates their invasion of the fibrin clot, where they also encounter dramatic changes in the content of the extracellular matrix (ECM) (Singer and Clark, 1999). An established requirement for PDGF-stimulated chemotaxis is the phosphoinositide 3-kinase (PI3K) signal transduction pathway (Anand-Apte and Zetter, 1997; Rönnstrand and Heldin, 2001), which we have previously characterized in the context of its intracellular localization in fibroblasts and sensitivity to PDGF gradients of varying magnitude and steepness (Haugh et al., 2000; Schneider and Haugh, 2005). PI3K signaling has been broadly studied in the context of chemotaxis, and the finer points of its dynamics and roles in other cells are still emerging (Ferguson et al., 2007; Kay et al., 2008; Loovers et al., 2006; Takeda et al., 2007).
Ligated PDGF receptors are among the most potent activators of Type IA PI3Ks (Auger et al., 1989; Jackson et al., 1992), and ECM-bound integrins have also been implicated in PI3K signaling (Chen et al., 1996; Khwaja et al., 1997; King et al., 1997). Upon translocation to the plasma membrane, type I PI3Ks phosphorylate phosphatidylinositol(4,5)-bisphosphate [PtdIns(4,5)P2] to generate PtdIns(3,4,5)P3 and its breakdown product, PtdIns(3,4)P2 (Hawkins et al., 2006; Vanhaesebroeck et al., 2001). These lipid second messengers control a wide range of cellular responses, including cell survival and proliferation, in addition to their roles in cell migration and chemotaxis (Engelman et al., 2006). Downstream effectors such as adaptor proteins, protein kinases, guanine nucleotide exchange factors and GTPase-activating proteins are localized and in some cases activated through phosphoinositide binding, which is generally mediated by their pleckstrin-homology domains (Lemmon et al., 2002). PI3K signaling is thought to affect motility through local recruitment of guanine nucleotide exchange factors for the Rho-family GTPases, Rac and Cdc42, which drive the formation of protrusive structures through recruitment of WAVE/WASP and Arp2/3 (Burridge and Wennerberg, 2004; Etienne-Manneville and Hall, 2002; Pollard and Borisy, 2003), and activation of effector kinases such as Pak, LIM kinases, PDK1 and Akt (del Pozo et al., 2000; Edwards et al., 1999; Primo et al., 2007; Sells et al., 1997; Sumi et al., 1999).
Although the role of PI3K in promoting cell motility is appreciated, especially during chemotaxis, as is its influence on cell adhesion and random cell migration on ECM (King et al., 1997; Pankov et al., 2005; Reiske et al., 1999), the nature of PI3K signaling triggered by adhesion remains largely obscure. Here, we show using total internal reflection fluorescence (TIRF) microscopy that PI3K is robustly activated in fibroblasts during their initial spreading on adhesive surfaces. Its lipid products are generated equally well in the presence or absence of classical integrin-mediated signaling, with dynamics that are consistently localized in regions of transient membrane protrusion. We further show that membrane protrusion is largely PI3K dependent. This is seen most dramatically when microtubules are depolymerized using nocodazole, which induces rapid protrusion-retraction events as seen by TIRF. In the context of random fibroblast migration, formation of branched lamellipodia and turning events were found to coincide with localized enrichment of 3′-phosphoinositides.
PI3K lipid products accumulate in a dynamic fashion during fibroblast spreading
To assess the activation of PI3K signaling and its possible relation to adhesion-based motility, we established stable expression of the 3′-phosphoinositide-specific Akt pleckstrin-homology domain, fused with enhanced green fluorescent protein (EGFP-AktPH) in NIH3T3 fibroblasts and monitored these cells by TIRF microscopy as they attached and spread on glass coated with fibronectin or poly-D-lysine (Fig. 1). Glass coated with bovine serum albumin, which was also present in our imaging buffer, does not promote adhesion of these cells (results not shown). After allowing the cells to spread for 30-50 minutes, a saturating dose of PDGF was added to evaluate the maximal activation of PI3K in each cell, followed by a large dose of PI3K inhibitor to evaluate the fluorescence intensity associated with EGFP-AktPH in the cytosol; the latter is used to normalize the PI3K-dependent response (Schneider and Haugh, 2004).
Cells spreading on fibronectin consistently showed dynamic enrichment of 3′-phosphoinositides, with transient bursts of signaling that were often localized in actively protruding regions at the cell periphery and other times showed a more global pattern (Fig. 1A; supplementary material Movie 1). These patterns of PI3K signaling are distinct from the more stable, ring-like patterns seen in response to PDGF stimulation, which we have previously explained in quantitative detail (Schneider and Haugh, 2004). Analysis of average TIRF intensity as a function of time shows that the 3′-phosphoinositide density generally increases dramatically during the first few minutes of cell spreading, indicating that PI3K is activated after cell attachment (supplementary material Fig. S1). Somewhat surprisingly, fibroblasts spreading on poly-D-lysine exhibited similar PI3K signaling dynamics (Fig. 1B; supplementary material Movie 2). Despite the expected differences in contact area morphology (Price et al., 1998) and a significantly higher mean spreading rate on fibronectin (117±90 μm2/minute versus 65±50 μm2/minute, mean ± s.d.) (Fig. 1C), both surfaces promoted 3′-phosphoinositide accumulation to similar extents. The average, PI3K-specific fluorescence intensity recorded for each cell prior to the addition of PDGF was compared with that of the cell periphery after PDGF stimulation (Fig. 1D; mean value 0.53±0.23 for fibronectin and 0.42±0.20 for poly-D-lysine). An intensity ratio of 0.5 roughly translates to a ratio of 0.1-0.2 in terms of 3′-phosphoinositide density if PDGF stimulation elicits ∼80% recruitment of EGFP-AktPH from the cytosol, as previously estimated (Schneider and Haugh, 2004). Control experiments using a cytoplasmic marker confirmed that the observed changes in fluorescence were the result of EGFP-AktPH recruitment to the contact area and not an artifact of the contact area topology (Fig. 1E).
Spontaneous PI3K activation during fibroblast attachment and spreading is not coupled with classical integrin-mediated signaling
Although poly-lysine has long been used to promote `non-specific' cell adhesion, one might attribute the similarity of the PI3K signaling responses on the two adhesive surfaces to the possibility that integrins are activated in both cases. It is conceivable that poly-D-lysine activates integrins or that the cells rapidly secrete ECM proteins that deposit on the surface. To test those possibilities, lysates of cells spreading on the two surfaces were prepared and probed for phosphorylation of focal adhesion kinase (FAK) and paxillin, two hallmarks of integrin signaling. In each of three independent experiments, robust and sustained phosphorylation of FAK and paxillin was elicited in cells spreading on fibronectin, compared with the cell suspension control, but not in cells spreading on poly-D-lysine (Fig. 2A); densitometry analysis indicated that FAK phosphorylation on poly-D-lysine was not significantly stimulated above the level of the suspension control and paxillin phosphorylation on poly-D-lysine was not detected (Fig. 2B). At the same time, both adhesive surfaces generated comparable levels of Akt phosphorylation, an indicator of PI3K signaling, albeit with different kinetics (Fig. 2A,B).
Incidentally, in cells spreading on poly-D-lysine and subsequently stimulated with PDGF, phosphorylation of FAK but not paxillin was observed (Fig. 2A). The ability of PDGF receptors to at least indirectly associate with and promote phosphorylation of FAK is well established (Rankin and Rozengurt, 1994; Sieg et al., 2000); although the precise mechanism remains unclear, a PDGF receptor-PI3K-FAK linkage has been implicated (Chen and Guan, 1994; Rankin et al., 1996).
Compared with the kinetics of GFP-AktPH translocation observed in individual cells, the aggregate Akt phosphorylation within a population of cells is affected by the variability in time needed for each cell to attach or initiate spreading; it was subsequently verified that this process takes longer on poly-D-lysine compared with fibronectin, consistent with the difference in Akt phosphorylation levels at early time points (Fig. 2C and results not shown). Fig. 2C also shows that spreading on fibronectin, but not poly-D-lysine, is strongly inhibited by a function-blocking antibody against β1 integrin; this treatment does not grossly affect 3′-phosphoinositide dynamics in cells spreading on poly-D-lysine (Fig. 2C). Although we cannot yet rule out the possibility that other integrins are (weakly) activated on poly-D-lysine, these results do further indicate that the mechanisms of fibroblast adhesion on poly-D-lysine and fibronectin differ dramatically, whereas 3′-phosphoinositides are generated to similar extents and with qualitatively similar spatiotemporal dynamics on the two surfaces.
More strikingly, whereas the blocking antibody strongly inhibited FAK and paxillin phosphorylation on fibronectin as expected, it did not significantly affect Akt phosphorylation on fibronectin (Fig. 2D). Under these conditions at least, activation of PI3K is apparently uncoupled from cell spreading, suggesting that some aspect of cell attachment to the surface is sufficient for activation of PI3K. This result further demonstrates that the PI3K pathway can be decoupled from classical integrin signaling.
Finally, to further exclude the possibility that the cells rapidly produce their own ECM during spreading, we also confirmed that cell spreading and 3′-phosphoinositide generation on poly-D-lysine were normal when protein synthesis was blocked by the inclusion of cycloheximide (25 μg/ml) in the cell suspension (supplementary material Fig. S2).
PI3K signaling is required for efficient spreading of fibroblasts
To assess the role of PI3K signaling in fibroblast motility during active spreading, the EGFP-AktPH-expressing cells were treated in suspension with varying doses of the PI3K inhibitor LY294002, and the rate of cell spreading on fibronectin was quantified and correlated with the normalized fluorescence intensity during the same period (Fig. 3). Experiments on poly-D-lysine showed qualitatively similar results (data not shown). It is known that high concentrations of LY294002 (>100 μM) are needed to completely block PI3K activation in NIH3T3 cells (Haugh et al., 2000; Watton and Downward, 1999), and accordingly, the degree of PI3K inhibition and the inhibition of cell spreading varied significantly from cell to cell. In some cells, 50 μM LY294002 was sufficient to block PI3K activation, whereas other cells required concentrations as high as 200 μM. Nevertheless, the two responses show a definite LY294002 dose dependence (Fig. 3A). GFP-AktPH translocation was, on average, more sensitive to the inhibitor than the observed spreading rate, suggesting that PI3K must be strongly inhibited to affect motility. Relating the two responses for each cell confirmed that the rate of cell spreading is largely responsive to PI3K signaling. Regardless of the LY294002 concentration used, cells in which PI3K was inhibited well (normalized fluorescence intensity <0.1) were far more likely to exhibit a spreading rate less than 60 μm2/minute (49 of 92 cells), which is approximately one s.d. below the mean of the DMSO control condition, compared with cells in which PI3K signaling was less inhibited (15 of 74 cells) (Fig. 3B). The contrast in spreading rates between cells exhibiting low and high PI3K signaling is statistically significant (P=4×10–5).
To further establish the functional role of PI3K in fibroblast spreading, NIH3T3 cells were co-transfected with EGFP-AktPH and a dominant-negative PI3K (Δp85) (Dhand et al., 1994), and this cohort was compared with cells transfected with EGFP-AktPH only (Fig. 3C). The vast majority of cells transfected with Δp85 (38 of 52) showed normalized EGFP-AktPH fluorescence values less than 0.1 during spreading, the same criterion used to qualify robust PI3K inhibition in Fig. 3A,B; subsequent stimulation with PDGF also failed to elicit significant PI3K activation in the majority of those cells (results not shown). By contrast, only 4 of 33 control cells, transfected with EGFP-AktPH only, exhibited low PI3K signaling by this measure. The spreading rates of cells with low versus higher PI3K signaling, and of dominant-negative p85-expressing versus control cells (without consideration of signaling levels) are statistically lower (P<0.01) (Fig. 3C).
When administered during the course of active cell spreading, the addition of LY294002 (100 μM) or another PI3K inhibitor, wortmannin (100 nM), caused a dramatic reduction in the rate of spreading in half (9 out of 18) of the cells observed, although some of those cells still showed changes in the morphology of the contact area (Fig. 3D). In the other cells, an effect on the rate of spreading was less apparent. In some cases, this was because the contact area versus time curve was such that the change in slope could not be evaluated unambiguously; in others, the slope appeared to be unchanged, indicating that PI3K-independent motility pathways were active or that PI3K was not inhibited to a sufficient extent to affect spreading in those cells.
Inhibition of actin polymerization uncouples PI3K signaling from the spreading response
In the context of leukocyte chemotaxis, PI3K signaling is amplified by a positive-feedback loop involving the actin cytoskeleton and Rac (Xu et al., 2003), prompting us to ask whether or not PI3K activation during spreading is both a contributor to and a consequence of changes in actin dynamics (Fig. 4). Actin polymerization was inhibited after the initiation of active cell spreading using cytochalasin D (1 μM), which binds and prevents elongation of actin filaments, and latrunculin B (2 μM), which sequesters actin monomers. The rate of cell spreading consistently showed a sharp reduction, but typically not a complete ablation, after the addition of cytochalasin D at this concentration, yet dynamic enrichment of 3′-phosphoinositides persisted (Fig. 4A). By comparison, the treatment with latrunculin B yielded a more dramatic inhibition of spreading, but again PI3K signaling was maintained (Fig. 4B). In both cases, the cells also showed the normal pattern and extent of PI3K signaling upon subsequent stimulation with PDGF. Quantification of the fractional PI3K activation levels after treatment with the inhibitors, assessed as in Fig. 1, confirmed that they were similar to DMSO controls (Fig. 4C). These results indicate that PI3K signaling is an upstream regulator of, but is not affected by, actin-based motility.
Dissociation of microtubules induces transient PI3K-dependent protrusion events on fibronectin
Microtubules influence cell migration, at least in part, by targeting focal adhesions and relaxing actomyosin-mediated contractility, which at least partially explains why fibroblasts in which microtubules have been dissociated cannot polarize or migrate effectively (Kaverina et al., 1999). Hence, we set out to test whether or not microtubule disruption, by treatment with nocodazole (10 μM), affects PI3K signaling during fibroblast spreading, and in the process we made a surprising observation (Fig. 5). Based on the role of microtubules in fibroblast migration, the initial response to nocodazole addition was expected; EGFP-AktPH-expressing fibroblasts on fibronectin- or poly-D-lysine-coated surfaces ceased spreading and, within 5-10 minutes, showed a dramatic shrinkage of the contact area as seen by TIRF. This effect was not accompanied by a loss of PI3K activity. Thereafter, half of the cells plated on fibronectin (36 of 72 observed) exhibited multiple protrusion events (Fig. 5A; supplementary material Movie 3). Most of the protrusions were rapid and unstable, forming and retracting within ∼5 minutes. Equally striking was the effect of PDGF under these conditions, which stimulated a rapid expansion of the contact area. This response to PDGF was completely reversed by the subsequent inhibition of PI3K (Fig. 5A). By comparison, we observed that none of the cells plated on poly-D-lysine exhibited the same protrusion phenotype after addition of nocodazole; instead, they exhibited significant detachment from the surface, and they did not re-establish the contact area in response to PDGF stimulation thereafter (results not shown).
We took advantage of the dynamic motility phenotype induced by nocodazole to further assess the role of PI3K in fibroblast motility. After nocodazole treatment and the onset of spontaneous protrusions, one of three PI3K inhibitors was added. LY294002 (100 μM), wortmannin (100 nM) and PI3Kα inhibitor IV (3 μM) each abolished the protrusion events while reducing the EGFP-AktPH fluorescence (Fig. 5B). These results provide further evidence that, at least under certain conditions, PI3K signaling is necessary for fibroblast motility.
PI3K signaling is locally activated in lamellipodia in conjunction with random fibroblast migration dynamics
Having established the spatiotemporal dynamics and importance of PI3K signaling during fibroblast spreading, we sought to characterize the 3′-phosphoinositide pattern during random fibroblast migration (Fig. 6). A total of 37 cells were analyzed, 20 on fibronectin and 17 on poly-D-lysine, each monitored for ∼6 hours. Two characteristic behaviors were consistently observed on both fibronectin and poly-D-lysine. During extended periods of persistent migration, cells often exhibited a single lamellipod, or two lamellipodia protruding in tandem, with strong polarization of 3′-phosphoinositide density. The representative cell shown in Fig. 6A (see also supplementary material Movie 4) initially had two opposing regions of PI3K signaling and membrane protrusion, but as the intensities of those regions waned, a new region of localized 3′-phosphoinositides appeared and developed along with the persistent protrusion of that region in a different direction. Other cells achieved movement by executing a zig-zag series of turns, alternately protruding twin lamellipodia; this branching behavior of fibroblasts in culture has been observed for some time (Abercrombie et al., 1970). As shown in the representative cell shown in Fig. 6B (see also supplementary material Movie 5), the appearance and disappearance of locally enhanced PI3K signaling typically foretold, respectively, the acceleration and stalling of these protrusion events. The formation of a new lamellipodial branch, also marked by 3′-phosphoinositide accumulation, often yielded a change in the overall direction of migration (Fig. 6B). In some cases, LY294002 (100 μM) was added during the experiment, and most of the cells showed a marked migration defect once the GFP-AktPH fluorescence drops (results not shown), which is consistent with previous reports (Pankov et al., 2005; Reiske et al., 1999).
Quantification of overall PI3K signaling directionality during fibroblast migration was achieved by construction of a new metric that we call the macroscopic `signaling vector' (Fig. 6C). As explained in the Materials and Methods, regions of interest containing high PI3K signaling (hot spots) are isolated by image segmentation, and then vectors pointing from the cell centroid towards the hot spots are weighted by their respective fluorescence volumes and summed. The resultant signaling vector is static in the sense that it is computed from information contained in a single image. For a cohort of 18 cells that could be continuously tracked as they migrated, the orientation of the signaling vector correlated strongly with the direction of movement of the cell centroid (over 12-minute intervals), expressed in terms of the angle between the two (Fig. 6C).
Finally, activation of PI3K and membrane association of Rac were monitored in the same cells as they migrated on fibronectin, assessed by translocation of co-transfected mCherry-AktPH and EGFP-labeled Rac1. Membrane localization of Rac is to some degree indicative of its conversion to the active, GTP-bound state, because Rac-GDP is predominantly sequestered in the cytoplasm (Bokoch et al., 1994; Moissoglu et al., 2006). As seen in the representative cell shown, Rac1 translocation is most often, but not always, co-localized with 3′-phosphoinositides; however, PI3K inhibition does not markedly alter Rac1 localization (Fig. 6D). These results suggest that, although Rac does not appear to be a major target of PI3K signaling in this context, the two signaling pathways are often active in the same place. The context is important here because even subtle overexpression of Rac can dramatically affect cell migration (Pankov et al., 2005), and in fact we note that our EGFP-Rac1-expressing cells exhibit altered contact area morphology and do not appear to migrate as persistently.
PI3K signaling is important in a variety of chemotactic systems, and its localization is polarized in response to chemoattractant gradients. Using TIRF microscopy, we have shown here that PI3K is also dynamically activated in fibroblasts during adhesion-based motility, both in the context of cell spreading and random migration, and our results reinforce that PI3K signaling is important for motility in these cells. Many studies suggest that PI3K is activated in response to integrin-mediated adhesion to ECM, and so we were surprised to find that PI3K is activated quite similarly in cells plated on fibronectin and poly-D-lysine, even though the latter does not elicit classical integrin-mediated signaling responses. At least two intriguing possibilities might explain this result. The first is that both fibronectin and poly-D-lysine activate a common receptor that mediates PI3K activation. Indeed, we were initially intrigued by the prospect that fibronectin activates not only integrins but also syndecans via one of its heparin-binding domains, which is rich in the basic amino acids arginine and lysine (Barkalow and Schwarzbauer, 1991; McCarthy et al., 1990) and that syndecans might be responsible for the activation of PI3K. However, treatment of our cells with heparinase I and III, at extremely high concentrations (up to 20 U/ml), did not noticeably affect cell spreading or PI3K activation on poly-D-lysine or fibronectin (data not shown). An alternative possibility that we are currently investigating is that PI3K signaling is responsive to mechanical cues, in which case the interplay between motile forces and PI3K signaling might constitute a positive-feedback loop. Positive-feedback loops have been implicated in spontaneous polarization of PI3K signaling and random migration in cells that exhibit amoeboid motility (neutrophils and Dictyostelium discoideum), although in those contexts F-actin is required for maintaining PI3K activation (Sasaki et al., 2007; Xu et al., 2003). In certain transformed cells, constitutive membrane localization of PI3K has been implicated in spontaneous cell polarization and migration (Wicki and Niggli, 2001).
Although basic motility of our cells was more often than not sensitive to PI3K inhibition, a significant number of cells exhibited PI3K-independent spreading behavior. This is not surprising in light of the growing evidence that PI3K is but one of at least a few pathways capable of transducing receptor-mediated signals to the cytoskeleton (Chen et al., 2007; Kay et al., 2008; van Haastert et al., 2007; van Rheenen et al., 2007). The importance of each of those pathways is proving to be context dependent. For example, it was reported some years ago that the dependence of PDGF-stimulated chemotaxis on PI3K signaling can be overcome if the phospholipase C pathway is overly active (Rönnstrand et al., 1999). Such redundancy might reflect an ability to affect cytoskeletal dynamics in distinct ways, or parallel pathways might converge to activate a common downstream effector. Indeed, actin polymerization in epithelial cells can be stimulated by both Cdc42 and PI3K signaling, each acting through Rac, such that inhibition of either pathway alone is not sufficient to fully abolish actin assembly (Bosse et al., 2007).
The use of TIRF microscopy afforded us the ability to observe novel contact area motility dynamics. Whereas dissociation of actin filaments halted or at least retarded cell spreading, effectively decoupling PI3K signaling from membrane protrusion, dissociation of microtubules using nocodazole resulted in contraction of the contact area and, in half of the cells spreading on fibronectin, a dynamic protrusion phenotype that we found to be PI3K dependent. The transient protrusion-retraction events sometimes appear to be non-random, firing in a radial pattern around the contact area. It would be interesting to investigate whether or not these dynamics seen by TIRF are related to previous observations that nocodazole induces orbital oscillations of the nucleus and of membrane blebs in spreading cells (Cuvelier et al., 2007; Pletjushkina et al., 2001). Microtubules have several roles in cell migration, including the local repression of contractility and maturation of focal adhesions near the leading edge (Small and Kaverina, 2003). During migration, contraction forces stimulate adhesion maturation and also the dissociation of mature adhesions and retraction at the rear of the cell (Galbraith et al., 2002; Palecek et al., 1998). A speculative hypothesis would therefore hold that the dynamics we observe reflect the tenuous relationship between contractility, presumably enhanced in the absence of microtubules, and the stability of focal adhesions. That this behavior is observed on fibronectin and not on poly-D-lysine lends some credence to this hypothesis.
Compared with the seemingly chaotic dynamics observed during cell attachment and spreading, the pattern of PI3K signaling during random cell migration seems to be tightly controlled in spite of its stochastic nature. It has been shown previously that polarized fibroblasts often exhibit a non-uniform PI3K-activation pattern (Haugh et al., 2000), and that protrusive structures in these cells are predisposed to higher levels of PI3K signaling in response to PDGF stimulation (Arrieumerlou and Meyer, 2005; Schneider et al., 2005). Here, we characterized the orchestration of PI3K signaling as fibroblasts executed persistent migration and turning behavior on adhesive surfaces. Persistent migration was accompanied by stable polarization of PI3K signaling at the front of the cell, whereas cells exhibiting a more tortuous path typically showed localized regions of PI3K signaling that were more transient. The distinctions between these two modes of migration warrant further study at the molecular level, but in both cases we find that the appearance, persistence, and disappearance of localized PI3K activation correlates with the ensuing membrane protrusion dynamics. For migration of cells such as fibroblasts, inexorably linked to adhesion, both soluble and matrix- or substratum-associated inputs must be integrated at the level of signal transduction pathways to affect cytoskeletal dynamics, and our results suggest that the PI3K pathway is one common medium for doing so.
Materials and Methods
Cell culture and reagents
Stable expression of the 3′-phosphoinositide-specific biosensor construct EGFP-AktPH (Haugh et al., 2000) in NIH3T3 mouse fibroblasts (American Type Culture Collection) was achieved by retroviral infection after cloning into the NotI-BamHI sites of the pBM-IRES-Puro vector (a gift from Steven Wiley and Lee Opresko, Pacific Northwest National Laboratory). The ecotropic ϕNX packaging cell line was transiently transfected, and virus-containing supernatants were used for serial infection of NIH3T3 cells, as described in detail previously (Kaur et al., 2006). This procedure yielded >80% infection efficiency as judged by fluorescence microscopy. EGFP-AktPH-expressing cells were further enriched by selection for 48 hours in the regular growth medium (Dulbecco's modified Eagle's medium supplemented with 10% v/v fetal bovine serum and 1% v/v penicillin-streptomycin-glutamate) supplemented with 2-4 μg/ml puromycin. All tissue culture reagents were purchased from Invitrogen (Carlsbad, CA), and cells were used between passages 10 to 40.
When two vector constructs were introduced, parental NIH3T3 cells were transiently co-transfected as described previously (Schneider and Haugh, 2004). The vectors used were EGFP-AktPH together with dominant-negative PI3K regulatory subunit (Δp85) (Dhand et al., 1994) [a gift from Tobias Meyer, Stanford University (Heo et al., 2006)] or mCherry-AktPH (mCherry DNA provided by Roger Tsien, University of California, San Diego, CA) along with EGFP-Rac1 [a gift from Martin Schwartz, University of Virginia (Moissoglu et al., 2006)].
Human plasma fibronectin was obtained from BD Biosciences (San Jose, CA) and Invitrogen. Human recombinant PDGF-BB was from Peprotech (Rocky Hill, NJ). LY294002 and PI3Kα inhibitor IV were from Calbiochem (San Diego, CA). Poly-D-lysine, wortmannin, latrunculin B, cytochalasin D, nocodazole, cycloheximide, heparinase I and heparinase III were from Sigma (USA). Antibodies against phospho-Akt (Ser473-P), phospho-paxillin (Tyr118-P), beta-actin (13E5) rabbit mAb, and HRP-conjugated anti-rabbit IgG were from Cell Signaling Technology (Beverly, MA). Phospho-specific anti-FAK (Tyr397-P) antibodies were from BioSource International (Camarillo, CA). Function-blocking anti-mouse β1 integrin antibodies were from BD Pharmingen (San Diego, CA).
Cell spreading and migration experiments
For spreading experiments, EGFP-AktPH-expressing cells were serum-starved for 2.5 hours and then detached with a brief trypsin-EDTA treatment and suspended in imaging buffer (20 mM HEPES pH 7.4, 125 mM NaCl, 5 mM KCl, 1.5 mM MgCl2, 1.5 mM CaCl2, 10 mM glucose and 2 mg/ml fatty-acid-free bovine serum albumin). After centrifugation at 100 g for 3 minutes, the cells were resuspended in imaging buffer and incubated at a density of 104 cells/ml for 30 minutes prior to the experiment; pre-incubation with inhibitors, where applicable, was also carried out during this step. Adhesive surfaces were prepared on clean, sterile glass cover slips, which were coated with poly-D-lysine (100 μg/ml) for 2 hours at room temperature or fibronectin (20 μg/ml) for 1 hour at 37°C, washed with deionized, sterile water and dried within 30 minutes of the experiment.
The methods described above were altered for random migration experiments as follows. The coating concentration of fibronectin was 10 μg/ml, the incubation of the cells in suspension was omitted, and the cells were allowed to spread for 1 hour prior to imaging. The imaging buffer was supplemented with 1% fetal bovine serum to maintain cell viability, and mineral oil was layered on top of the buffer to prevent evaporation during the 6-7 hours duration of each experiment.
TIRF microscopy is used to selectively excite fluorophores within ∼100 nm of the substratum-buffer interface, which illuminates the plasma membrane contact area and, in fibroblasts, ∼5-10% of the cytoplasm directly above it (Schneider and Haugh, 2004). Our prism-based TIRF microscope has been described in detail previously (Schneider and Haugh, 2005). EGFP was excited using a 60 mW 488 nm line from a tunable wavelength argon ion laser head (Melles Griot, Irvine, CA), whereas CellTracker Red (Molecular Probes) was imaged using a 100 mW diode-pumped 561 nm line (Crystalaser, Reno, NV). The relevant filter sets are 515/30 nm for EGFP and 630/60 nm for CellTracker Red (Chroma, Brattleboro, VT). A ×20 water immersion objective (Zeiss Achroplan, 0.5 NA) and ×0.63 camera mount were used. Digital images, with 2×2 binning, were acquired at regular intervals (10-15 seconds for spreading experiments and 2 minutes for migration experiments) using a Hamamatsu ORCA ER cooled CCD (Hamamatsu, Bridgewater, NJ), with a fixed exposure time×gain of 1000-1200 milliseconds for EGFP and 500 milliseconds for CellTracker Red, and Metamorph software (Universal Imaging, West Chester, PA).
Analysis of spreading experiments
The following cell-selection criteria were imposed for the analysis of TIRF images. Cells must have spread to a contact area of at least 200 μm2 within 30 minutes of the start of image acquisition, and they needed to have an average, whole cell fluorescence that was 100 gray levels above the background throughout the experiment; the background was consistently between 250-300 out of 4096 gray levels. Cells that were partially out of the field of view or in significant contact with another spreading cell were also omitted from the analysis.
Analysis of random migration experiments: determination of the macroscopic signaling vector
EGFP-AktPH-expressing NIH3T3 fibroblasts were prepared as described above for spreading experiments, except that a higher cell density (2.5×105 per ml) was needed. The adherent cells were washed once with ice-cold Dulbecco's phosphate-buffered saline and then scraped into ice-cold lysis buffer (50 mM HEPES, pH 7.4, 100 mM NaCl, 1% v/v Triton X-100, 1 mM sodium orthovanadate, 10 mM sodium pyrophosphate, 50 mM β-glycerophosphate, pH 7.3, 5 mM sodium fluoride, 1 mM EGTA, and 10 μg/ml each aprotinin, leupeptin, pepstatin A and chymostatin). The lysates were vortexed briefly, incubated on ice for 20 minutes, and clarified by centrifugation. The samples were subjected to SDS-PAGE and immunoblotting using standard techniques. The blotted membranes were incubated with enhanced chemiluminescence substrates (Pierce, Rockford, IL) and imaged using a high sensitivity camera (BioRad Fluor S-Max, Hercules, CA), which gives a linear output with respect to light intensity.
This work was supported by grants to J.M.H. from the National Institutes of Health (R21-GM074711) and the Office of Naval Research (N00014-03-1-0594). J.J.R. was supported through an award to J.M.H. from the Camille & Henry Dreyfus Foundation (TC-05-022). We thank the laboratories of H. Steven Wiley, Tobias Meyer, Roger Tsien, and Martin Schwartz for providing reagents. Deposited in PMC for release after 12 months.