ADF/cofilin family proteins control formation of oriented actin-filament bundles in the cell body to trigger fibroblast polarization

Cell migration is essential for life; it is required throughout embryo development, and for tissue repair and immunity in both the embryo and the adult. It also contributes to several important diseases, including vascular disease, inflammatory diseases and the spread of cancer.

To reach new territory, a cell must polarize, which means that it must form a cell front and rear. For clarity of terms, in this work we define `initiation of cell polarization' as the first visual sign of a break in cell symmetry, `elaboration' as further elaboration of this first break, `fully polarized' when morphologically both the cell front and cell rear have fully formed, and `cell polarization' as the entire process. The cell migrates once it is fully polarized. Cell protrusion at the front of the cell is roughly diametrically opposed to cell retraction at the rear, and cell migration occurs approximately along this axis in the direction of protrusion.

To polarize, there must be coordination between formation of the cell front and cell rear in both time and space. In some migrating cell types, such as neutrophils and amoeba cells, pathways based on phosphoinositide 3-kinases and members of Rho family proteins control formation of the cell front and rear (reviewed in Dormann and Weijer, 2006; Kimmel and Firtel, 2004; Niggli, 2003; Ridley et al., 2003). In these cell types, feedback loops between pathways ensure that the cell front and rear form coordinately (Charest and Firtel, 2006; Schneider and Haugh, 2006; Van Keymeulen et al., 2006; Wong et al., 2006; Xu et al., 2003). However, in other migrating cell types, such as fibroblasts, mathematical models do not invoke feedback loops (Schneider and Haugh, 2005) and coordination is poorly understood in these cells. Furthermore, in all migrating cell types, the very early events in the cell that establish spatial and temporal differences needed in cellular processes for cell polarization are unclear.

Intuitively the cell front might be expected to form first in response to a localized source of chemoattractant. However, all migrating cell types tested have the capacity to polarize in the absence of an externally applied gradient of chemoattractant. This infers that a significant contribution to establishing the cell front and rear must come from self-organization and self-amplification of important molecules within the cell (Bernstein and Bamburg, 2004; Onsum et al., 2006; Verkhovsky et al., 1999b; Wong et al., 2006; Xu et al., 2003). In this regard, it is not obvious whether the cell front or rear should be expected to form first, which has implications for overall mechanism. In some types of already polarized cells, cell retraction precedes cell protrusion (Chen, 1981; Dunn, 1980; Dunn and Zicha, 1995; Small and Resch, 2005). However, whether this also occurs to polarize a cell has not yet been explicitly investigated.

In most cell types, correctly organized actin networks are required for morphological cell polarity as well as providing the driving force needed to move the cell (reviewed by Ridley et al., 2003; Small and Resch, 2005; Wittmann and Waterman-Storer, 2001). Other cytoskeletal proteins, such as microtubules (depending on the cell type) and myosin II, can also play an important role (reviewed by Ridley et al., 2003; Rodriguez et al., 2003; Small et al., 2002; Small and Kaverina, 2003; Wittmann and Waterman-Storer, 2001). Although there is some information in cells (Cassimeris and Zigmond, 1990; Kaverina et al., 2000) and cytoplasmic fragments (Verkhovsky et al., 1999b), how reorganization of the actin cytoskeleton contributes to early events needed to drive local cellular differences to form the front and back of the cell has not been investigated widely experimentally. Particularly unclear is the role of actin within the cell body; previous studies have, for the most part, focused on actin in the leading cell edge. Although actin organization within the cell body has been studied in a few migrating cell types, such as keratocytes (Svitkina et al., 1997), fibroblasts (Cramer et al., 1997) and myoblasts (Swailes et al., 2004), in these studies the cells were already polarized and moving, and thus cell polarization and associated actin reorganization in the cell body were not studied.

To address these issues, here we have studied cell polarization in primary chick heart fibroblasts, a well-established migrating cell system with well-characterized cell polarity (Abercrombie and Heaysman, 1966; Couchman and Rees, 1979; Cramer, 1999b; Cramer et al., 1997). In already polarized and moving fibroblasts, actin comprises an actin-rich lamellipodium at the front of the cell, and oriented actin-filament bundles (graded polarity actin-filament bundles) spanning the cell body and oriented in the direction of migration (Cramer et al., 1997).

Determining regulation of actin organization is important in the identification of actin mechanisms driving cell polarization. It is becoming clear that the ADF/cofilin (AC) family of proteins, which catalyze actin-filament severing and depolymerization in cells (reviewed by Moon and Drubin, 1995; Sarmiere and Bamburg, 2003; Zigmond, 2004), is necessary for cell migration (reviewed by Bamburg and Wiggan, 2002). AC proteins are also known regulators of cell polarity, both in yeast (Drees et al., 2001; Okada et al., 2006) and also in migrating cell types, both to sustain already established polarity (Dawe et al., 2003) and to trigger cell polarization (Helen R. Dawe. The role of ADF/cofilin family proteins in the acquisition and maintenance of cell polarity during fibroblast migration. PhD thesis, University College London, 2003) (Ghosh et al., 2004; Mouneimne et al., 2006; Nishita et al., 2005). Despite the clear role of AC in controlling cell polarity, studies of AC in the context of actin organization during cell polarization are limited; this process is either not the focus of the work (Ghosh et al., 2004; Mouneimne et al., 2006; Nishita et al., 2005) or the studies are preliminary (see PhDcit_addRef.doc). Furthermore, most studies focusing on the function of AC during motility have been limited to actin in the leading cell edge or lamellipodium (Chan et al., 2000; Dawe et al., 2003; Ghosh et al., 2004; Nishita et al., 2005; Zebda et al., 2000). By contrast, any role for AC in regulating actin organization within the cell body of migrating cells during cell polarization has yet to be explored.

ACs are regulated in cells by several mechanisms (reviewed by Moon and Drubin, 1995; Sarmiere and Bamburg, 2003; Zigmond, 2004). In chick fibroblasts, AC is at least regulated by phosphorylation on serine 3 by LIM kinase (Dawe et al., 2003), which inactivates AC. Here, we employ mutant AC proteins, including an AC mutant that acts as a dominant negative, and constitutively active LIM kinase 1 to explore the function of AC during cell polarization in chick heart fibroblasts.

We found that, during cell polarization in chick fibroblasts, AC activity and actin-filament disassembly are required for the formation of oriented actin-filament bundles within the cell body of these cells. This is an early event that consequently triggers coordinate formation of the cell front and rear around opposing ends of the bundles. We conclude that formation of oriented actin-filament bundles within the cell body requires AC and is a required, early step in triggering fibroblast polarization.

We developed three types of cell-polarization assays, all in primary fibroblasts obtained from chick heart explants from 7- to 8-day-old embryos. These assays were: latrunculin A (Lat A) washout of pre-treated cells growing out of explant tissue, plated for 24-36 hours; cell plating of individual, untreated, dissociated cells prepared directly from tissue explants; and spontaneous cell polarization of untreated cells growing out of explant tissue, plated for 24-36 hours. Because we obtained similar data in each different assay, we think that the process we describe here is a fundamental property of fibroblasts and the mechanism we uncover important for cell polarization. In each assay, cells spontaneously polarized in cell culture medium, indicating that polarization was driven by self-organization of the actin cytoskeleton with no externally applied chemoattractant gradient. The fact that the cytoskeleton can self-organize to propagate cell polarization has been recognized for some time and might be a basal property of migrating cells (Verkhovsky et al., 1999b).

Because the Lat-A-washout cell-polarization assay was the most robust and experimentally manipulable, we routinely used this method (presented throughout the paper). Motility and polarity in chick fibroblasts require an organized actin cytoskeleton (Cramer et al., 1997). Pre-treating live cells with Lat A, which sequesters actin monomers, causing filament disassembly (Ayscough et al., 1997), abolishes cell polarity. Upon subsequent washout of Lat A, we found that cells spontaneously repolarized.

We first tested empirically and found that, by 1-hour pre-treatment with 1-5 μM Lat A (depending on different batches of Lat A from the manufacturer, see Materials and Methods), normal actin organization and cell-adhesion patterns were abolished (supplementary material Fig. S1). Thus, after subsequent washout, cell repolarization occurs by formation of newly polymerized and newly organized actin structures rather than from any pre-existing organized structures. Consistent with this, after Lat A washout, cells repolarized and had a cell front and back apparently unrelated to their original position prior to Lat A treatment (e.g. for the individual cell in Fig. 1A, compare –1:18:05 and 1:06:30). This is also represented as a schematic for this cell (Fig. 1B) and for the entire cell population (Fig. 1C).

In the presence of Lat A, cells appeared rounded while remaining adherent to the Matrigel substratum. At around 1-hour total pre-treatment, Lat A was thoroughly and rapidly washed out (see Materials and Methods). By 1 hour after washout, approximately 70% of cells were fully repolarized (quantified in Fig. 1D) – the expected proportion that are polarized in a given population of untreated chick embryo heart fibroblasts – and showed the expected cell shapes (Cramer, 1999b; Cramer et al., 2002; Cramer et al., 1997; Dawe et al., 2003). As soon as cells fully polarize, they begin to migrate at an average speed of 1 μm per minute, the same speed as in untreated cells (Cramer, 1999b; Cramer et al., 1997; Dawe et al., 2003). Actin organization (see below) in migrating cells, once repolarized, appeared no different to control untreated migrating cells. Together, these data indicate that the Lat A cell-polarization assay faithfully reproduces the expected cell behaviour of these cells.

To study cell polarization, we analyzed fibroblasts between 0-1 hours after washing out of Lat A, and then typically beyond this to continue to monitor cell migration. Over this time, using phase-contrast time-lapse microscopy of live cells, we observed a reproducible sequence of shape-change events as fibroblasts repolarized (example of typical individual cell in Fig. 1A from 0:00:00 minutes).

Immediately after washing off Lat A, cells re-spread to form a discoid or ovoid shape with the nuclear mass roughly in the cell centre and delocalized cell protrusions around the entire cell periphery (Fig. 1A, 0:15:00 minutes). In the cell population, 98% of cells had re-spread by 15 minutes after washing out Lat A. Delocalized protrusions in re-spread cells were dynamic – protruding and retracting (see Movie 1 in the supplementary material). Because these protrusions/retractions fluctuated around the entire cell periphery, they did not alter overall cell shape and thus re-spread cells are overall symmetrical with a non-polarized cell shape.

The first visual sign of a break in cell symmetry in these cells was a small, discrete, localized zone of cell-edge retraction (Fig. 1A, black arrow), which formed by 35 minutes after washing out Lat-A for the individual cell shown in Fig. 1A,E and on average by 31 minutes in the cell population (ranging from 26 to 35 minutes for the majority of cells) (quantified in Fig. 1F). Within the cell population, initiation of cell polarization is thus fairly synchronous. This localized cell-edge retraction gave the appearance of a small `bite' taken out from the cell edge and was approximately 5-15 μm across and 3-5 μm deep, taking about 5 minutes to form. In contrast to the earlier fluctuations, this retraction was very stable and permanently broke the symmetrical shape of the cell. This initial break in cell symmetry then elaborated with further cell-edge retraction for about another 20-30 minutes (varying with individual cells) to form the cell rear (Fig. 1A, 0:58:30 minutes, two black arrows, Fig. 1E, 35-60 minutes).

Visually, a localized cell front became apparent at the opposing cell edge (Fig. 1A, 0:35:00 and 0:58:30, green dot) as the cell rear formed. However, although recognizable as a cell front, net protrusion at this location only occurred towards the end of formation of the cell rear (Fig. 1A, compare red and green dots; Fig. 1E, at 60 minutes). In the cell population, there was an average 21-minute delay between initial cell-edge retraction and net protrusion at the cell front, ranging from 4 to 37 minutes in individual cells (Fig. 1G), and, in 13/15 cells, the delay was 10 minutes or more. Once both the cell rear and front had fully formed, the cell started to migrate (Fig. 1A, compare 0:58:30 and 1:27:02; Fig. 1E, from 60 minutes). Thus, during cell polarization in this system, formation of the cell rear occurs prior to net protrusion at the cell front.

Having identified the sequence of shape-change events that occur in live cells during cell polarization, we assessed actin organization in each identified shape by two approaches (Figs 2 and 3). In one approach, to obtain a large data set we fixed cells in a time-course series during cell repolarization and then stained them with phalloidin (Fig. 2). In rounded, Lat-A-treated cells prior to repolarization, aggregates of actin were induced by Lat-A treatment (Fig. 2A), as expected. For cell shapes during repolarization upon Lat A washout, in all cell shapes, as expected, F-actin was enriched in protrusions at the cell periphery. For actin in the cell body, the focus of this work, in re-spread cells, two distinct actin populations formed, one earlier after washing out Lat A (by around 15 minutes after washout) and one later (by around 30 minutes after washout). In the earlier population, actin in the cell body formed a non-oriented array, either an actin meshwork (Fig. 2B) or non-oriented actin-filament bundles (Fig. 2B′). However, a dramatic shift in actin organization apparently occurred in re-spread cells: in the later population, actin in the cell body was reorganized into oriented (parallel) actin-filament bundles spanning the length of the cell (Fig. 2C,C′). This organization of oriented actin-filament bundles was also present in cells that had initiated cell polarization, identified as cells with a discrete zone(s) of cell-edge retraction (Fig. 2D), and in fully repolarized cells, identified as cells that appeared similar to already characterized polarized cells of this cell type (Cramer, 1999b; Cramer et al., 2002; Cramer et al., 1997; Dawe et al., 2003) (Fig. 2E). We defined actin-filament bundles as oriented in cells when more than 70% of total visible bundles were parallel (±30°) to each other (e.g. Fig. 2C has roughly 80% oriented actin and Fig. 2C′ has 100% oriented actin).

This dramatic shift in actin organization is clearly noticeable in the cell population, as shown by the quantification of actin organization as a function of time (Fig. 2F) and shape (Fig. 2G). Also, the timing of initiation of cell polarization in live cells from Fig. 1A,F was similar to when there was a shift towards an oriented actin-filament-bundle organization in the fixed cell population (Fig. 2F). From these two analyses, we therefore hypothesize that formation of oriented actin-filament bundles in the cell body is important for triggering cell polarization.

To begin to directly test this hypothesis, we first investigated whether formation of oriented actin-filament bundles in the cell body precedes initial cell-edge retraction, by identifying actin organization in live cells expressing GFP-actin using spinning-disc confocal time-lapse microscopy (see Materials and Methods) during cell polarization (Fig. 3).

Live-cell analysis (Fig. 3) revealed – as in the fixed time-point analysis above (Fig. 2) – that, during cell polarization, non-oriented actin meshwork/bundles is the first actin organization visibly formed within the cell body (Fig. 3A, see 0:02:20-0:06:01 minutes). Strikingly, prior to any break in cell symmetry, actin then appeared to form oriented actin-filament bundles spanning the cell body (Fig. 3A, by 0:29:34), which remained as the cell continued to establish cell polarity (Fig. 3A, 0:29:34-1:00:27) and migrate (Fig. 3A, compare 1:00:27 and 1:06:13) (Movie 2 in the supplementary material). This transition from non-oriented to oriented actin organization as a function of cell polarization occurred in all 22 cells observed in the live-cell population expressing GFP-actin that polarized (quantified in Fig. 3B), similar to data obtained in the fixed time-point assay (Fig. 2G).

Time-history analysis of each of these 22 (22/22) individual cells that polarized showed that oriented actin-filament bundles formed, prior to initiation of cell polarization (individual cell shown in Fig. 3A, compare 0:29:34 and 0:36:39; quantified for the cell population in Fig. 3C, light-grey bar). A further 2 (2/24) cells failed to polarize by up to 2 hours of observation. In these cells, actin remained disorganized (Fig. 3C, dark-grey bar).

Within an average of 5 minutes (average from 22 live cells), ranging from 0.5-9.6 minutes in individual cells (Fig. 3D), of oriented actin-filament bundles forming in the cell body, cell polarization initiated with a small stable retraction event, as documented in Fig. 1 (see representative example in Fig. 3A, 0:36:39, arrow). For most of these cells (19/22), the average timing of this initial retraction event was 36.3 minutes, ranging from 21-47 minutes in the cell population (Fig. 3E, lower panel), similar to live cells analyzed by phase-contrast microscopy (Fig. 1). The other minority of cells expressing GFP-actin (3/22) were delayed up to around 1 hour, due to a delay in forming oriented actin bundles (Fig. 3E, compare upper and lower panels). We suspect the delay was due to an increase in total exposure to UV light and/or higher levels of expression of GFP-actin in a minority of cells.

From fixed images of cells stained with phalloidin (Fig. 2), the geometric relationship between actin-filament-bundle orientation and cell polarization was unclear because the initial break in cell symmetry sometimes appeared on one side of the bundles (as in Fig. 2D), yet in fully polarized cells the rear was positioned around one end of the bundles (as in Fig. 2E). To reveal this relationship, we therefore studied precisely how retraction elaborates to form the cell rear by assessing this step in live cells expressing GFP-actin in detail (Fig. 3F). In 14/22 cells that polarized, both initial, stable retraction of the cell edge and subsequent elaboration of the retraction occurred end-on to the oriented actin bundles (Fig. 3Fi). In the remaining 8/22 cells, initiation occurred, side-on to the bundles, either at one (Fig. 3Fii) or both sides (Fig. 3Fiii). In both cases, elaboration then occurred around the end of bundles (as in Fig. 3Fi). Thus, in all cases, elaboration of the retracted cell area always occurred around one end of oriented actin-filament bundles to form the rear of the cell (Fig. 3Fiv) (exemplified in Fig. 3A, and Movie 2 in the supplementary material). A localized protruded area remained at the opposing end of the bundles to form the front of the cell (Fig. 3Fiv, blue band), which then underwent net protrusion after formation of the cell rear (Fig. 3G), as documented in Fig. 1E. We noticed that the exact timing of the onset of migration was either coincident with the onset of net protrusion (as in Fig. 1E) or slightly delayed (as in Fig. 3G). This appears to correlate with the timing of the end-point of forming the cell rear, implying that cell migration is only triggered once both the cell rear and front are fully formed. Once the cell was migrating, migration continued in the direction of the oriented actin-filament bundles (net migration down the page in the direction of the bundles, compare last two frames in Fig. 3A), as documented in previous work (Cramer et al., 1997).

Thus, formation of oriented actin-filament bundles within the cell body appears to coordinate the position of the front and back of the cell, a new finding. This process also thus establishes the correct actin geometry (long length of bundles in the direction of migration) within the cell body required (Cramer et al., 1997) for subsequent cell migration in these cells.

These data provide direct evidence that formation of oriented actin-filament bundles in the cell body precedes and is linked with triggering cell polarization.

Evidently from our inspection of time-lapse movies of GFP-actin in live cells, formation of oriented actin-filament bundles involves a large number of actin dynamic processes, which form a complete body of work in their own right and are not described here in detail. These dynamic processes include actin-filament flow, growth, disassembly, bundling and thickening of bundles (all of which are apparent by close inspection of Movie 2 in the supplementary material), and association with myosin II (data not shown); together, these processes culminate in the straightening and strengthening of bundles into an oriented array spanning the cell body. Here, we focus on testing a role for actin-filament disassembly and severing in this process because it is becoming clear that this is involved in controlling cell polarity (see Introduction for details).

To further test our hypothesis that formation of oriented actin-filament bundles in the cell body is important for triggering cell polarization, and to test whether actin-filament disassembly controls oriented actin-bundle formation, we treated cells with 0.5 μm jasplakinolide during cell polarization. At such low doses, jasplakinolide specifically and rapidly inhibits actin-filament disassembly within 1-2 minutes of its addition to fibroblasts (Cramer, 1999b). Because this is faster than the formation of oriented actin-filament bundles (see Fig. 2F, Fig. 3A), we were able to specifically target a role for these bundles during cell polarization.

To specifically target these bundles, we added 0.5 μm jasplakinolide to cells early in the cell-polarization process: Lat A was washed off cells and cells were incubated in culture medium for a total of 15 minutes, a time when we know (from Figs 2, 3 and Fig. 4A) that actin is not yet oriented, and then jasplakinolide was added. We also know that actin should orient within about the next 15 minutes. We empirically tested in these cells that treating them for up to 45 minutes with 0.5 μm jasplakinolide was specific in blocking actin-filament disassembly, because we did not observe any actin aggregates until after this time (e.g. see Fig. 4C) [aggregates are induced in cells when the weaker activity of jasplakinolide of inducing actin-filament polymerization in cells occurs (Cramer, 1999b)]. This meant that the timing and duration of treating cells with 0.5 μm jasplakinolide during the polarization assay would specifically test a role for actin-filament disassembly in forming oriented actin-filament bundles.

In 8/8 live cells tested, this jasplakinolide treatment blocked both formation of oriented actin-filament bundles within the cell body and cell polarization (Fig. 4, compare A with B). Formation of a cell front and rear was inhibited and the cell remained at the re-spread stage with transient protrusions and retractions around the entire cell periphery (Fig. 4B). The block in polarization was specific to a block in actin depolymerization and not due to any effects of UV light or GFP-actin expression levels, because all 8/8 cells were blocked; less than two would be expected if the effect were non-specific. In addition, we obtained similar data in a separate assay without using GFP-actin; quantification of 600 fixed cells in a fixed time-point assay showed that, at 30 minutes after Lat A washout, only 10% of treated (last 15 minutes in jasplakinolide) cells had oriented actin-filament bundles compared with 68% of controls (Lat A washout only) (Fig. 4D). By 60 minutes after Lat A washout, less than 12% of treated (last 45 minutes in jasplakinolide) cells were polarized compared with 70% of controls (Lat A washout only) (Fig. 4E), the expected proportion in untreated cells (Cramer, 1999b; Cramer et al., 1997; Dawe et al., 2003).

In contrast to a block in formation of oriented actin-filament bundles, jasplakinolide did not block delocalized protrusion (see Fig. 4B). This is similar to previous data showing that jasplakinolide does not block delocalized protrusions in non-polarized cells that naturally occur when fibroblasts are cultured long-term (Cramer, 1999b). This suggests that delocalized protrusion does not require actin-filament disassembly irrespective of whether delocalized protrusion occurs during the cell-polarization process or in non-polarized cells. We note, as previously described, that there was an apparent accumulation of F-actin at the rear of the lamellipodium in jasplakinolide-treated cells (Fig. 4B), which we interpret as actin filaments that have not been depolymerized (Cramer, 1999b).

These results show that actin-filament disassembly is required for the formation of oriented actin-filament bundles in the cell body and provides further evidence that oriented actin-filament-bundle formation is necessary for cell polarization.

Because AC catalyzes actin-filament severing and disassembly in cells, we examined the distribution of AC during cell polarization with an antibody that recognizes ADF in chicken cells (Morgan et al., 1993). On initial inspection of cells (images not shown), ADF localized to protrusions, as expected from previous work on ADF and cofilin (Dawe et al., 2003; Svitkina and Borisy, 1999), and dispersed fairly homogeneously throughout the cell body with no apparent localization to actin-filament bundles in this location. However, in studies of cofilin in keratocytes, cells were first extracted live in order to more clearly visualize cofilin bound to actin filaments in the leading cell edge (Svitkina and Borisy, 1999). To test the possibility that a proportion of ADF localized to actin-filament bundles in the cell body but was masked by unbound ADF, we extracted cells in cytoskeleton-stabilizing buffer (to prevent actin depolymerization) prior to fixation using a similar method to the keratocyte work (Svitkina and Borisy, 1999) (Fig. 5). We estimate that we removed 50-70% of total ADF from cells prior to subsequent cell fixation. In this method, ADF localized to actin filaments in the cell body both before (Fig. 5E,E′) and after (Fig. 5F,F′) formation of oriented actin-filament bundles. This is clearly seen in the enlarged views (Fig. 5E′,F′, orange colour). Localization of ADF on the actin-filament meshwork/non-oriented bundles within the cell body prior to formation of oriented actin-filament bundles was striking (Fig. 5E′). The spotty/granular appearance of ADF is similar to previously published images of AC in cells (Aizawa et al., 1996; Chan et al., 2000; Dawe et al., 2003; Mouneimne et al., 2006; Nishita et al., 2005; Svitkina and Borisy, 1999), including actin-colocalization studies (Aizawa et al., 1996; Chan et al., 2000; Svitkina and Borisy, 1999), and is probably indicative of the cooperative F-actin binding of ADF (Hayden et al., 1993), which, at substoichiometric amounts, gives local regions of saturation (McGough et al., 1997). Interestingly, once oriented actin bundles had formed (Fig. 5F-F″), localization of ADF appeared more obvious on thinner/less-bundled actin bundles (Fig. 5F′) than on thicker/more-bundled filament bundles (Fig. 5F″) within the cell body of the same cell. This might infer a difference in age of bundles because we noticed that bundles thicken during cell polarization.

To test AC function (Fig. 6), we expressed either inactive pseudo-phosphorylated Xenopus AC1 E3 (XAC1 E3 mutant), which acts as a dominant negative (Dawe et al., 2003), alone (Fig. 6B), or constitutively active LIMK1 EE508 mutant alone (Fig. 6C) to increase the proportion of inactive phosphorylated-AC in cells, and assessed cells fixed and stained in a time-course assay during cell polarization. We have previously characterized these mutants in live chick fibroblasts (Dawe et al., 2003) (see Materials and Methods).

Cells expressing these mutants to block AC function (Fig. 6B,C) were competent at forming actin-filament bundles in the cell body early during cell polarization (Fig. 6B,C, 15 minutes), similar to control cells (Fig. 6A, 15 minutes). However, cells expressing these mutants failed to subsequently form oriented actin-filament bundles in the cell body (Fig. 6B,C, 25, 30, 60 minutes; and quantified in Fig. 6E, red, yellow lines) compared with control cells (Fig. 6A, 25, 30, 60 minutes; and Fig. 6E, green line). They also failed to polarize; instead, multiple protrusions and retractions occurred that were delocalized around the cell periphery (Fig. 6B,C, 60 minutes). By 60 minutes after Lat A washout 70% of control cells (expressing GFP alone) polarized; by contrast, only 10% of cells that expressed the inactive AC (XAC E3) mutant polarized and 15% of cells that expressed the LIMK EE508 mutant (to inactivate AC) (Fig. 6F). These data are similar to those from cells treated with jasplakinolide (Fig. 4B,D,E).

Failure to form oriented actin-filament bundles in the cell body and to trigger cell polarization was AC-specific, because co-infection of cells with both LIMK EE508 and the active non-phosphorylatable Xenopus AC1 A3 (XAC1 A3) mutant detected with an anti-XAC1 antibody (see Materials and Methods) rescued both formation of oriented actin-filament bundles (Fig. 6D, 25-60 minutes; quantified in Fig. 6E, blue line) and cell polarization (Fig. 6F, blue bar) to expected wild-type (GFP-only, control) levels (Fig. 6E, green line and Fig. 6F, green bar). The time taken for oriented actin-filament bundles to form in rescued cells was indistinguishable from control cells (Fig. 6E, compare green and blue lines). The active non-phosphorylatable Xenopus AC1 A3 mutant used to rescue cells, when expressed alone, had no detectable effect on cells (data not shown).

These two different approaches to block AC function show that AC activity is required for the formation of oriented bundles in the cell body and for the triggering of cell polarization in fibroblasts.

We discovered that formation of oriented actin-filament bundles within the cell body requires the AC family of proteins and is an early event required for subsequent cell polarization (summarized in Fig. 7). The fact that actin within the cell body is needed for cell polarization is consistent with the already appreciated role for actin in this cellular location for cell migration (Cramer et al., 1997; Small et al., 1998; Svitkina et al., 1997; Verkhovsky et al., 1999a) and with the known role for actin-filament cables in the cell body for yeast cell polarity (reviewed by Moseley and Goode, 2006). Although our work shows that oriented actin-filament bundles within the cell body are necessary for cell polarization, it does not address whether they are sufficient, an important direction for future studies.

One apparent function of oriented actin-filament bundles in the cell body in fibroblast polarization is coordinating the formation of the cell rear and front around opposing ends of long lengths of the bundles (summarized in Fig. 7C-E). How might this occur? It is unlikely that these bundles generate power to drive initial shape-change events per se because, when oriented bundles were absent (Fig. 4B and Fig. 6B,C), delocalized protrusion and retraction could still occur around the cell periphery (the cell is not polarized and does not translocate under these conditions). More likely is that oriented actin bundles actively determine at least the position of the cell rear, because this event occurred only after bundles were formed. Positioning of the cell front then could simply be a consequence of the cell rear forming first, and subsequent net protrusion at the cell front could perhaps be promoted by retraction-mediated protrusion (Chen, 1981; Dunn, 1980; Dunn and Zicha, 1995; Small and Resch, 2005). Because protrusion is driven by a separable actin-based mechanism, however, it is likely that additional cellular factors are also locally activated at the cell front to ensure that net actin polymerization is locally maintained.

An important area for future work is to identify precise details of how oriented actin-filament bundles function in determining the position of the cell rear and front in fibroblasts. Plausible ideas are that oriented actin bundles provide a track for actin-based nuclear movement, a movement that is known to play a role in NIH 3T3 fibroblast polarization (Gomes et al., 2005), or that they act as a guide for polarity determinants, perhaps similar to that discussed within the leading cell edge of already polarized cells (reviewed by Ettienne-Manneville, 2004; Gundersen and Bretscher, 2003).

Any role for actin in the cell body for positioning the cell rear and front in other migrating cell types is yet to be uncovered. In myoblasts, graded polarity actin-filament sheets span the cell body under the plasma membrane and are aligned in the direction of migration (Swailes et al., 2004), comparable to the alignment of actin bundles that we describe in the cytoplasm here, and they might perhaps have a similar role in polarizing these cells. During polarization of cytoplasmic fragments of keratocytes, as with fibroblasts, oriented actin bundles also form in the cell body, but, unlike fibroblasts, are aligned transversely to the direction of migration (Verkhovsky et al., 1999b), and, in migrating amoeba and neutrophils, most polymerized actin in the cell body is a meshwork localized in the cell cortex. Thus, the precise mechanism of positioning the front and back of the cell during cell polarization might vary with distinct types of actin organization in the cell body in different migrating cell types. Such variation is not unexpected, because the precise mechanism of cell migration is itself probably varied among distinct cell types (Cramer, 1999a; Lauffenburger and Horwitz, 1996; Mitchison and Cramer, 1996; Sheetz, 1994; Small and Resch, 2005; Verkhovsky et al., 1999a). This highlights the importance of advancing knowledge on detailed mechanisms in all types of migrating cells in order to derive any global principals.

In this work in chick fibroblasts, formation of the cell rear during polarization occurred first in the absence of any externally applied chemoattractant gradient, implying a basic property of self-organization in the cell. This does not appear to be restricted to this cell type. From close inspection of the images presented in the primary literature in polarizing cells, formation of the cell rear can also precede local net protrusion at the opposing end of the cell in a variety of other cell types, including mouse fibroblasts, neutrophils, keratocytes and dictyostelium cells (Gomes et al., 2005; Swanson and Taylor, 1982; Verkhovsky et al., 1999b; Wong et al., 2006). This not only occurs in a uniform concentration of applied chemoattractant (Wong et al., 2006) but also in the presence of a localized source (Swanson and Taylor, 1982) or mechanical stimulus (Verkhovsky et al., 1999b). Although this needs to be explicitly studied in other cell types and cell polarization systems, it seems, then, that a primary control of cell polarization can be governed by activation of cell-rear formation first, perhaps related to the observation in some cells that retraction occurs first in already polarized and migrating cells (Chen, 1981; Dunn, 1980; Dunn and Zicha, 1995; Small and Resch, 2005). This might explain – apparent from the images shown – why some cells do not initially protrude a localized cell front in the direction of a supplied localized source of chemoattractant (Gerisch and Keller, 1981; Mouneimne et al., 2006; Ridley et al., 1999).

Local membrane events govern protrusion at the cell front (Niggli, 2003; Ridley et al., 2003) and these are also important. Whether the cell front or rear forms first might vary with precise cellular situation. Activating the cell front first might be sufficient for short-range excursions, for example. Allowing pathways to activate cell-rear formation first, by contrast, might be more important in situations in which individual cells completely change direction of migration, mediated by the front and back of the cell reversing their position, as in reverse chemotaxis, for example (Keizer-Gunnink et al., 2007; Mathias et al., 2006), or in response to chemorepulsion during embryogenesis.

AC functions to sever/depolymerize actin filaments and regulates actin dynamics within cell protrusions of motile cells (Chan et al., 2000; Dawe et al., 2003; Ghosh et al., 2004; Nishita et al., 2005; Zebda et al., 2000). Here, we uncovered a new role for AC, the formation of oriented actin-filament bundles within the cell body during cell polarization. Our work is consistent with the known requirement for AC in the formation of oriented actin cables in the cell body for polarized growth in yeast (Okada et al., 2006). This might imply that AC functions both in cell protrusions and within the cell body in general in migrating cells, although this remains to be directly tested in a single assay in a single cell type.

How, then, does AC function to form oriented actin-filament bundles during cell polarization? The jasplakinolide data (Fig. 4) provides an argument for actin-filament depolymerization being important. With a time resolution of seconds, we did not observe complete depolymerization of the existing non-oriented actin-filament meshwork in live cells (Fig. 7A) and then subsequent formation of oriented actin-filament bundles. This argues against a simple model of complete depolymerization of one actin organization and then temporally distinct repolymerization to form another. Several models are plausible, then, for the function of depolymerization, and more than one is probably required to explain formation of the oriented bundles.

Actin-filament depolymerization might directly contribute actin monomers required to make oriented bundles, perhaps similar to the known function for depolymerization in providing actin monomers for polymerization within the leading cell edge of migrating fibroblasts (Cramer, 1999b; Dawe et al., 2003) and other cell types (Kiuchi et al., 2007). This idea is also consistent with data on actin bundles in cells depleted of AC (Hotulainen et al., 2005). This scenario would require new actin polymerization to make bundles, perhaps at sites of cell adhesions observed during cell polarization (supplementary material Fig. S1), although we do not yet know whether adhesions actively promote actin-bundle formation. In addition to this function, actin depolymerization might have a lesser role in removing off-axis bundles during cell polarization, a phenomenon we have observed for a few bundles towards the end of forming oriented actin bundles in cells (data not shown).

For this theory to work, actin depolymerization of the existing non-oriented actin meshwork must occur nearly simultaneously with the formation of oriented bundles, which is difficult to test. AC severing activity can also form actin bundles (Aizawa et al., 1996; Maciver et al., 1991; Moriyama et al., 1996) and severing would generate shorter actin filaments that could be manoeuvred into a bundled organization by proteins such as a-actinin and myosin II. In support of this idea, we observed both movement of bundles in live cells (data not shown) and association of polymerized actin with myosin II during formation of oriented actin-filament bundles (T.M. and L.P.C., unpublished). Alternatively, or in addition to, a proportion of severed filaments could be used as a source of free-actin barbed ends for new actin-filament growth (Ichetovkin et al., 2002), although see Kiuchi et al. (Kiuchi et al., 2007).

To distinguish between these plausible AC mechanisms for forming oriented actin-filament bundles during polarization, the relative importance of depolymerization and severing activities of AC, respectively, need to be assessed, as well as any separate functions for individual AC family members. High local concentrations of cofilin, but not of ADF, can nucleate assembly of actin filaments (Andrianantoandro and Pollard, 2006; Chen et al., 2004). Such a difference in activity could be important in polarized cell migration. Silencing of ADF, but not cofilin, in a colorectal-tumour cell line blocked their transwell migration through Matrigel, a measure of the metastatic potential of a cell (Estornes et al., 2007). Significantly, these cells express 83% cofilin and 17% ADF, almost identical to the cofilin and ADF percentages in chick cardiac fibroblasts (Tahtamouni and J.R.B., unpublished).

We conclude that AC control formation of oriented actin-filament bundles in the cell body, which in turn coordinates the formation of the cell rear and front during cell polarization. The finding that oriented actin in the cell body contributes to triggering cell polarization is new and potentially opens novel avenues in which to explore the overall mechanism and targets for regulating this important process in migrating cells.

Migrating primary fibroblasts were freshly prepared from explants of embryonic day (E)7 or E8 chick embryo hearts for each experiment and grown in DMEM low glucose (GIBCO) containing 10% chick serum (Sigma) and 10% FCS (CEF medium) at 37°C and 5% CO₂ for 24-36 hours after plating the explants, as previously described (Cramer, 1999b; Cramer et al., 1997). Coated glass coverslips were used for immuno-cell-staining and live-cell imaging. The glass was coated with 1 mg/ml poly-L-lysine (Sigma) and then coated for 30 minutes with Matrigel (Becton Dickinson) on the day of plating explants. Once the explants had adhered and cells migrated out, cells were treated with 1-5 μm Lat A (Calbiochem) for 1 hour then washed thoroughly with CEF medium. Lat A was thawed from a 1000× DMSO stock immediately before treatment. The final concentration of DMSO in the drug mixture was 0.1% and this has no effect on actin organization and cell behaviour when added to fibroblasts alone. Lat A concentration (1-5 μm) was standardized so that all normal organized actin in cells was abolished by 1 hour of treatment; this required periodically titrating Lat A over 1-5 μm because strength varied between batches and with age of drug. To monitor polarization, cells were then analyzed from 0 minutes to 2 hours after washing out Lat A.

To stain for actin and XAC1, cells were fixed and stained at 37°C in 4% methanol-free formaldehyde (TAAB) in cytoskeleton buffer (10 mM MES pH 6.1, 3 mM MgCl₂, 138 mM KCl, 2 mM EGTA) with 0.32 M sucrose for 30 minutes, as previously described (Cramer and Mitchison, 1993). In brief, they were then permeabilized in 0.5% Triton X-100 for 10 minutes. For indirect immunofluorescence, cells were incubated with an anti-XAC1 antibody (Rosenblatt et al., 1997) for 45 minutes, washed and then incubated simultaneously with a Cy-3 or Alexa-Fluor-594-conjugated anti-rabbit secondary antibody (Jackson ImmunoResearch Laboratories) and 0.1 μg/ml FITC or Alexa-Fluor-350-conjugated phalloidin (Molecular Probes) for 45 minutes. Cells treated with jasplakinolide were gently fixed in –20°C methanol for 1 minute then stained with anti-actin C4 (ICN Biomedicals) for 45 minutes, then an Alexa-Fluor-594-conjugated anti-mouse secondary antibody (Jackson ImmunoResearch Laboratories) for 45 minutes.

Cells were treated with 0.5-3.0% Triton X-100 (Sigma) in cytoskeleton, extraction buffer (PEM; Pipes 100 mM, EGTA 1 mM, MgCl₂ 1 mM and 4% PEG) at 37°C for 1 minute, similar to previous methodology (Svitkina and Borisy, 1999). The cells were then immediately fixed at 37°C in 4% methanol-free formaldehyde (TAAB) in cytoskeleton buffer (10 mM MES pH 6.1, 3 mM MgCl₂, 138 mM KCl, 2 mM EGTA) with 0.32 M sucrose, for 30 minutes. They were then permeabilized in 0.5% Triton X-100 for 10 minutes. For indirect immunofluorescence, cells were incubated with an anti-ADF antibody (rabbit 12977) (Shaw et al., 2004; Morgan et al., 1993) for 45 minutes, and incubated simultaneously with an Alexa-Fluor-488-conjugated anti-rabbit secondary antibody (Molecular Probes) and 0.1 μg/ml Alexa-Fluor-594-conjugated phalloidin (Molecular Probes) for 45 minutes. The anti-ADF antibody has previously been characterized in chick fibroblasts and specifically recognizes both chicken ADF and chicken phospho-ADF (Dawe et al., 2003).

We used the Ad Easy adenoviral system (Stratagene), consisting of adenovirus containing recombinant expression vectors. Replication-incompetent adenoviruses containing the cDNA for expressing covalently fused actin-GFP (Tanner et al., 2005), LIM kinase (T508EE) with GFP on a separate promoter (Edwards et al., 1999), Xenopus AC1 (XAC1) (S3A) (Meberg and Bamburg, 2000), XAC1 (S3E) (Meberg and Bamburg, 2000) and, as a control, GFP with a lacZ promoter were made, expanded and titred as described (Minamide et al., 2003). These were used to infect CEF as previously described (Dawe et al., 2003). Using this method of infection, we found that chick embryo fibroblasts expressed exogenous proteins efficiently. Control viral infection, containing only GFP, has no effect on either migration speed or cell polarity in chick fibroblasts (Dawe et al., 2003). When cells were infected, we aimed for 60-70% of fibroblasts at the edge of the explant to be expressing the exogenous proteins. This was achieved by using 1.0×10⁷ to 3.0×10⁷ virus particles per E7-E8 chick heart (approximately 0.9-1.4 million cells); variations in transfection levels occurred between different viruses and the age of the virus aliquot. Indirect immunofluorescence with an anti-XAC1 antibody (Rosenblatt et al., 1997) was used to distinguish cells infected by the XAC1 A3 and XAC E3 viruses from uninfected cells because this antibody has minimal cross reactivity with endogenous chick AC (Dawe et al., 2003; Shaw et al., 2004). Matrigel-coated glass-bottomed cell culture dishes (WillCo Wells) were used for live imaging of cells exposed to virus.

Lat A was washed off cells then, 15 minutes later, 0.5 μm jasplakinolide (a gift from M. Sanders, Dep. Chemistry UC, Santa Cruz, CA) was added. Cells were then either imaged live by time-lapse microscopy or fixed at various time points for immuno-cell-staining. Jasplakinolide was diluted to the correct concentration, just before use, from a thawed 1000× DMSO stock.

High-resolution fluorescent images of fixed cells were obtained digitally by a 12-bit cooled CCD camera (SenSys KAF 1400, Roper Scientific) on an upright microscope (Nikon eclipse e800) using a 60×1.4 NA oil objective controlled by Metamorph software (Universal Imaging).

Digitally acquired phase-contrast time-lapse movies of live cells were made using an inverted microscope (Zeiss Axiovert S1000-TV) and a 12-bit cooled CCD camera (MicroMax KAF 1400, Roper Scientific) with a 63×1.40 NA oil Zeiss objective controlled by Metamorph software. The cells on glass coverslips were heated to 36.5°C by a temperature-controlled water bath, custom built into a heating chamber.

Spinning-disc confocal microscopy (Perkin Elmer) controlled by Metamorph was used to obtain time-lapses of actin dynamics in live cells expressing GFP-actin. The images were captured on an inverted microscope (Nikon TE 2000-U) by a high-performance camera (Orca II-ER, Hamamatsu) and a 60× NA1.40 oil Nikon lens. Cells on glass-bottomed dishes (WillCo Wells) were heated to 37°C in an electronically controlled chamber (Heatwave-30, World Precision Instruments) and the objective was also electronically heated to minimize focal drift (Bioptechs large objective heater, Intracell).

We are very grateful to Vania Braga (Imperial College London) for helpful discussion and comments on the manuscript. Helen Dawe, as part of her thesis work, made some preliminary discoveries in a related cell system that AC blocks cell polarization. We gratefully acknowledge grant support from the Cancer Research UK and Wellcome Trust (L.P.C.) and the National Institutes of Health, grant NS40371 (J.R.B.). L.P.C. is a Royal Society University Research Fellow.