Journal of Cell Science partnership with Dryad

Journal of Cell Science makes data accessibility easy with Dryad

Tyrosine phosphorylation of annexin A2 regulates Rho-mediated actin rearrangement and cell adhesion
Ursula Rescher, Carsten Ludwig, Vera Konietzko, Alexei Kharitonenkov, Volker Gerke


Cell adhesion and motility require a dynamic remodelling of the membrane-associated actin cytoskeleton in response to extracellular stimuli that are primarily transmitted through receptor tyrosine kinases. In a cellular model system for tyrosine phosphorylation-based growth factor signaling, we observed that annexin A2 is tyrosine-phosphorylated upon insulin receptor activation. The phosphorylation precedes peripheral actin accumulations and subsequent cell detachment. These morphological changes are inhibited by annexin A2 depletion and require Rho/ROCK signaling downstream of tyrosine-phosphorylated annexin A2. A phospho-mimicking annexin A2 mutant is sufficient to drive peripheral actin accumulation and the resulting cell detachment in the absence of insulin stimulation. Thus, a tyrosine phosphorylation switch in annexin A2 is an important event in triggering Rho/ROCK-dependent and actin-mediated changes in cell morphology associated with the control of cell adhesion.


Cell adhesion and motility depend on dynamic responses of the cytoskeleton that are regulated by the precise coordination and integration of multiple signaling cascades transmitting various extracellular signals. Central to many of these processes is a remodelling of the actin cytoskeleton, which requires the coordinated activity of specific actin-binding proteins together with numerous additional cytoskeletal and signaling molecules, such as kinases and GTPases. Extracellular signals that trigger changes in the actin cytoskeleton resulting in altered cell adhesion and motility have been described, and the signaling receptors involved are known. In many cases these belong to the family of receptor tyrosine kinases (RTKs) with prominent examples being the receptors for epidermal growth factor, ephrins and insulin (for reviews, see Heldin, 1996; Schlessinger, 2000; Ullrich and Schlessinger, 1990). Owing to its importance in cellular adhesion and cell motility, RTK signaling and its deregulation are also critically involved in tumorigenesis.

Activation of the insulin receptor, in addition to initiating a wide range of metabolic and nuclear responses, elicits changes in the F-actin distribution, manifesting in increased membrane ruffling, dispersion of focal adhesion proteins and a loss of cell-substratum contact (Moller et al., 1995; Tsakiridis et al., 1999). A number of substrates of the insulin receptor participating in the regulation of these responses have been described, including the insulin receptor substrate p53, a multidomain scaffolding protein implicated in filopodium and lamellipodium formation (Govind et al., 2001; Miki et al., 2000). However, many signaling intermediates and the respective pathways involved in mediating cell morphology changes and loss in substrate adhesion following insulin receptor stimulation remain to be elucidated.

Cells overexpressing the insulin receptor have been introduced as model systems to study tyrosine-kinase-based signaling because experiments with primary cells expressing RTKs, such as adipocytes in the case of insulin receptor, are limited with respect to manipulation by protein knockdown or overexpression. Cell models include baby hamster kidney cells expressing the human insulin receptor (BHK-IR), which respond to insulin treatment with profound alterations in their cellular morphology and a decrease in cell adhesion, ultimately leading to a detachment from the substratum (Andersen et al., 2001; Moller et al., 1995). Using this model system, we here identify annexin A2 as a major insulin receptor substrate that is functionally linked to the insulin-induced changes in the actin cytoskeleton. Annexin A2 is a member of the annexin family of Ca2+-regulated membrane binding proteins, which has been implicated in the organization of membrane domains and membrane-cytoskeleton contacts (for a review, see Gerke et al., 2005). In addition to interacting with Ca2+ and membrane lipids, annexin A2 is able to bind G- and F-actin, it inhibits filament elongation at the barbed ends and it is recruited to sites of actin assembly at cellular membranes (Hayes et al., 2004) (for a review, see Rescher and Gerke, 2004). Hence, the protein has been linked to actin remodelling processes at cellular membranes, such as those occurring during the clustering of certain membrane receptors or during the formation of actin tails on mobile macropinosomes (Merrifield et al., 2001; Oliferenko et al., 1999). These functions are probably linked to the dynamic organization of cholesterol-rich membrane microdomains to which annexin A2 can be recruited via a direct interaction with phosphatidylinositol (4,5)-bisphosphate [PtdIns(4,5)P2] (Hayes et al., 2006; Rescher et al., 2004).

Annexin A2 was initially identified as a major substrate of pp60vsrc (Erikson and Erikson, 1980) and thus has been discussed to play a role in cellular differentiation and/or transformation. The phosphorylation site is located at Tyr23 in the unique N-terminal domain, which is expected to face the cytosol in membrane-bound annexin (for a review, see Gerke et al., 2005). Despite this early identification, the role of annexin A2 tyrosine phosphorylation, in particular in events associated with cell transformation, has not been elucidated so far. Our results now show that this tyrosine phosphorylation of annexin A2 is involved in the Rho/ROCK-mediated generation of contractile force leading ultimately to cell detachment, a morphological hallmark of cell transformation.


Insulin-induced alterations in cell morphology and adhesion correlate with annexin A2 tyrosine phosphorylation

BHK cells overexpressing the insulin receptor (BHK-IR) were used here as a model system to analyze RTK-based signaling, which leads to actin cytoskeleton rearrangements. These cells respond to insulin activation with significant morphological alterations ultimately leading to cell contraction and detachment from the substratum (Moller et al., 1995) (see also below). The rounded cell morphology is not due to apoptosis because the caspase 3 inhibitor z-VAD-fMK did not inhibit the insulin-induced detachment and insulin-treated cells did not bind the apoptosis marker FITC-annexin V on their surface (not shown). To identify downstream targets of the insulin receptor that possibly participate in the regulation of actin cytoskeleton dynamics, we analyzed protein tyrosine phosphorylation in insulin-treated BHK-IR cells. Whereas a relatively low level of tyrosine phosphorylation was observed in lysates from nontreated BHK-insulin receptor cells, phosphorylation of several proteins was induced upon insulin stimulation (Fig. 1A). In addition to the heavily phosphorylated 90 kDa protein representing the insulin receptor β-subunit, a strongly phosphorylated protein band migrating at ∼36 kDa was detectable within 5 minutes of insulin stimulation, which accumulated further during the next 3 hours of stimulation. Mass spectrometry analysis of several tryptic peptides obtained from this 36 kDa polypeptide revealed a high similarity to human annexin A2, a Ca2+- and phospholipid-binding protein with an apparent molecular mass of 36 kDa (not shown). To confirm that this band corresponded to hamster annexin A2, cell lysates were probed in parallel with annexin A2 and anti-phosphotyrosine (pY) antibodies revealing a co-migration of annexin A2 and the pp36 signal (not shown). Moreover, anti-annexin A2 immunoprecipitation resulted in the time-dependent precipitation of the tyrosine-phosphorylated band at 36 kDa (Fig. 1B, upper panel). Reprobing with annexin A2 antibodies showed that the total amount of annexin A2 remained unchanged during the induction of its tyrosine phosphorylation (Fig. 1B, lower panel). Dose-response and time-course experiments revealed that significant annexin A2 phosphorylation was observed when cells were stimulated with 5 nM insulin (not shown), and that annexin A2 tyrosine phosphorylation increased significantly within the first 3 hours (Fig. 1C). Thus, insulin induced a strong and sustained tyrosine phosphorylation of annexin A2 in BHK-IR cells.

To investigate whether annexin A2 phosphorylation required insulin receptor kinase activity and/or that of downstream kinases such as Src, cells were insulin treated in the presence of the specific insulin receptor tyrosine kinase inhibitor HNMPA-(AM)3, or the Src inhibitor PP2, and tyrosine phosphorylation of annexin A2 was monitored as above. PP2 did not affect annexin A2 phosphorylation, whereas HNMPA-(AM)3 completely inhibited it (not shown). Furthermore, co-immunoprecipitation analyses revealed that annexin A2 associated with the insulin receptor in nonstimulated BHK-IR cells (not shown). Altogether, these data demonstrate that in BHK-IR cells, annexin A2 is a major substrate of the insulin receptor, and that annexin A2 phosphorylation probably occurs in a direct manner.

Fig. 1.

Insulin induces rapid and sustained tyrosine phosphorylation of annexin A2 in BHK cells stably expressing the human insulin receptor. (A) Time course of insulin-induced tyrosine phosphorylation of BHK-IR cell proteins. Serum-starved BHK-IR cells were treated with 100 nM insulin for the indicated times and total cellular lysates were probed by immunoblotting with anti-phosphotyrosine antibodies. Arrows indicate the positions of the insulin receptor β-subunit (IRβ) and the heavily phosphorylated 36 kDa protein (pp36). (B) The tyrosine-phosphorylated polypeptide of 36 kDa is annexin A2. Lysates of cells stimulated for the indicated periods of time with 100 nM insulin were subjected to immunoprecipitation (IP) with the monoclonal anti-annexin A2 antibody HH7. The precipitated protein was immunoblotted (WB) for phosphotyrosine (α-pY) and the blot subsequently reprobed with anti-annexin A2 antibodies (α-Anxa2). c, control lane (lysates were incubated with beads alone). (C) Annexin A2 phosphorylation increases during the first 3 hours of insulin treatment. Serum-starved cells were stimulated for the indicated times with 100 nM insulin. The intensity of the pp36 band was measured by densitometric scanning and is given as arbitrary units (mean values ± s.e.m.).

Insulin-triggered changes in cell morphology require the presence of annexin A2

Insulin-induced detachment from the substratum and the concomitant appearance of nonadherent cells in the cell culture medium can be detected as early as 1 hour after insulin stimulation. Statistically more reliable data, however, require longer incubation periods. As shown in Fig. 2A, insulin stimulation for 4 hours led to a significant detachment, even when only completely nonadherent cells were quantified. To detect insulin-induced differences in cellular morphology, we examined the actin cytoskeleton. This also enabled us to match the temporal and spatial alterations in cellular morphology and F-actin distribution more precisely with the induction of annexin A2 phosphorylation. Paralleling the onset of annexin A2 tyrosine phosphorylation, F-actin remodeling was already detectable within 5 minutes of insulin stimulation, although rather subtle and to varying degrees (not shown). As demonstrated in Fig. 2B, pronounced alterations in the F-actin distribution were reliably observed after 30 minutes of insulin stimulation. Nonstimulated control cells showed a flattened, well-spread morphology with distinct actin stress fibers (Fig. 2B,C). By contrast, insulin treatment rapidly (within 30 minutes) caused the cells to round up and increase in height. Actin stress fibers disappeared and the cells instead displayed massive accumulations of actin in the form of domes or sheaths at the cell periphery (Fig. 2B,C). These structures were positive for phosphotyrosine and annexin A2 (Fig. 2D). This strongly suggests that an insulin-mediated actin rearrangement generates the force for cell contraction, enabling the cells to move upwards and finally to detach from the substratum.

Fig. 2.

Insulin induces massive actin rearrangements in BHK-IR cells. (A) Insulin-induced detachment of BHK-IR cells. Cells grown in 24-well plates were serum-starved overnight and stimulated with 100 nM insulin for 4 hours. Detached cells in the supernatant were counted. Bar graphs represent mean ± s.e.m. from a representative experiment with five parallel samples. *P<0.05. (B) Insulin-induced changes in the actin cytoskeleton of BHK-IR cells. Serum-starved cells grown on coverslips were stimulated for 30 or 60 minutes with 100 nM insulin, fixed and stained for F-actin with TRITC-phalloidin. (C) y-z and x-y sections of representative noninduced cells or cells induced for 30 minutes with insulin are shown, with their typical F-actin stain. Note the increase in height and the concurrent appearance of actin domes or sheaths in the insulin-treated cells. (D) Serum-starved cells were stimulated for 60 minutes with 100 nM insulin, fixed, stained for F-actin with TRITC-phalloidin and co-stained with anti-annexin-A2 (Anxa2) or anti-phosphotyrosine (pY) antibodies, respectively. Scale bars: 10 μm.

We next sought to determine whether insulin-induced annexin A2 phosphorylation is a crucial event in triggering the actin rearrangements and the resulting loss of adhesion by using RNAi to specifically reduce the level of intracellular annexin A2. RNAi resulted in a marked reduction in the annexin A2 content of total cellular lysates (Fig. 3A). As judged by immunofluorescence, this corresponded to greatly reduced annexin A2 levels in the majority of the siRNA-treated cells (not shown) indicative of an almost complete annexin A2 depletion in positively transfected cells. Specific annexin A2 reduction did not result in a difference in the insulin-induced tyrosine phosphorylation state of the IRβ subunit (Fig. 3A), nor did it cause any obvious alterations in the level of phosphotyrosine-containing proteins other than annexin A2 (Fig. 3A). This indicates that suppression of annexin A2 expression did not affect the insulin-induced tyrosine phosphorylation levels in general. To assess the effect of annexin A2 siRNA on insulin-induced changes in cell morphology and F-actin distribution, we determined the number of cells showing the typical insulin-triggered morphology changes in spite of reduced annexin A2 levels. Therefore, cells transfected with the effective siRNA duplex or a control RNA duplex were stimulated for 30 minutes with insulin, and subsequently fixed and stained for F-actin. As shown in Fig. 3B, insulin induced the appearance of cells with the `insulin-responsive' actin phenotype described above – actin domes and loss of stress fibers – in about 80% of mock-transfected cells, whereas knocking down intracellular annexin A2 drastically reduced the ability of the cells to respond by rearranging their actin cytoskeleton. To confirm that the inability of actin remodelling was a direct consequence of annexin A2 depletion, the insulin-induced changes in F-actin were assessed in annexin-A2-depleted cells re-expressing annexin-A2–GFP (Anxa2-GFP). Cells that had been annexin A2 downregulated for 2 days were transfected with either Anxa2-GFP or GFP alone. Western blotting of the cell lysates 24 hours after transfection revealed that the cells treated with annexin A2 siRNA were still depleted of endogenous annexin A2 but expressed Anxa2-GFP to a level comparable with that in mock-transfected cells (Fig. 3C). Examination of the insulin-induced actin phenotype revealed that GFP-expressing cells remained unable to undergo insulin-triggered F-actin changes, whereas cells re-expressing Anxa2-GFP now responded similarly to nondepleted cells and displayed the insulin-induced actin phenotype. These results indicate that annexin A2 is critically involved in the formation of insulin-induced actin domes.

Fig. 3.

Annexin A2 depletion inhibits insulin-induced actin rearrangements. (A) Downregulation of annexin A2 does not affect the general tyrosine phosphorylation in BHK-IR cells. Cells were nontransfected, transfected with control (mock) or annexin A2-specific siRNAs, serum-starved and insulin-stimulated for 30 minutes as indicated. Total cellular lysates were then immunoblotted for phosphotyrosine (α-pY). The blot was stripped and reprobed for the IRβ-subunit (α-IRβ) and annexin A2 (α-Anxa2). (B) Effect of RNAi-mediated downregulation of annexin A2 on insulin-induced actin rearrangements. BHK-IR cells, grown on coverslips, were mock transfected or transfected with annexin A2 siRNA, serum-starved and either nonstimulated or stimulated with 100 nM insulin for 30 minutes. The fixed cells were stained for F-actin and morphological changes in the actin cytoskeleton (appearance of the `insulin-responsive' actin phenotype) were quantified as described in the Materials and Methods. (C) Re-expression of annexin A2 in annexin-A2-depleted cells restores the ability to insulin-dependently remodel F-actin. Western blotting of total cell lysates shows comparable levels of annexin A2-GFP expression in annexin A2 siRNA and mock-treated cells. GFP-expressing cells remained unable to undergo insulin-induced F-actin changes, whereas cells re-expressing annexin A2-GFP responded similar to nondepleted cells and displayed the insulin-induced actin phenotype. **P<0.0005.

Fig. 4.

The Rho/ROCK pathway is involved in mediating insulin-induced morphological changes in BHK-IR cells. (A) The active RhoA mutant RhoQ63L induces an insulin-like phenotype. BHK-IR cells were transfected with a Myc-tagged RhoQ63L construct and co-stained with anti-Myc-antibodies and TRITC-phalloidin or anti-annexin-A2 antibodies. Note the appearance of the typical `insulin phenotype', i.e. cell rounding and the formation of actin domes upon RhoQ63L expression. Scale bars: 10 μm. (B) Rho/ROCK activity is required for insulin-induced morphological changes. Serum-starved cells grown on coverslips were left untreated or preincubated either with the cell-permeable Rho-inhibiting C3 exotoxin or with two different, structurally unrelated ROCK inhibitors, Y27632 and H-1152, prior to insulin stimulation for 30 minutes. The fixed cells were stained for F-actin and morphological changes in the actin cytoskeleton were quantified. **P<0.0005.

Annexin A2 phosphorylation is linked to Rho/ROCK activation

Members of the Rho family of small GTPases and their effectors are key intracellular signaling molecules coordinating cytoskeletal remodeling required for cell spreading, motility and cell-shape changes (for a review, see Ridley, 2001). To analyze the potential involvement of Rho family GTPases in insulin-induced actin rearrangements, we introduced the permanently active RhoA mutant RhoQ63L into BHK-IR cells. Virtually all cells expressing this mutant showed the changes in the actin cytoskeleton described above even in the absence of insulin (Fig. 4A, upper panel), with annexin A2 localizing to the actin-rich structures (Fig. 4A, lower panel). This underscores the important role of Rho in mediating the observed changes in actin cytoskeleton organization after insulin receptor activation. Furthermore, we specifically inactivated Rho by treating BHK-insulin receptor cells with the C3 exotoxin of Clostridium botulinum. This treatment completely abrogated the insulin effect, reducing the number of cells showing the insulin-induced actin rearrangements to levels found in the absence of insulin (Fig. 4B). A rounded cell morphology and the appearance of bleb-like membrane protrusions had been linked to activation of the Rho/ROCK signaling pathway (Riento and Ridley, 2003). Since both polymerization of actin and actin-myosin contractility are regulated by active Rho via the serine/threonine kinases ROCKI and ROCKII, which act as Rho effectors (Riento and Ridley, 2003), we examined whether Rho signaling to ROCK participated in mediating the insulin-induced actin rearrangements. To exclude possible nonspecific effects of the pharmacological inhibition, cells were treated with two different, structurally unrelated ROCK inhibitors, Y27632 (Uehata et al., 1997) and H-1152 (Rattan and Patel, 2008). Blocking ROCK function through each of the inhibitors almost completely prevented the insulin-induced change in morphology (Fig. 4B). This suggests that the rounded morphology and the observed actin-rich domes are a result of insulin-induced Rho/ROCK signaling, which probably generates the contractile force responsible for detachment.

Fig. 5.

Insulin-triggered Rho activation requires annexin A2. (A) BHK-IR cells were transfected with either control (mock) or annexin A2 siRNA or with the RhoQ63L expression construct. Cells were treated with 100 nM insulin as indicated, fixed and probed for Rho GTP loading state with GST-RBD as described in the Materials and Methods. Bound GST-RBD was detected by staining with anti-GST-antibody. F-actin was visualized with TRITC-phalloidin. Note the insulin-induced increase in GST-RBD binding that is not seen in annexin-A2-downregulated cells. Scale bars: 10 μm. (B) Bound GST-RBD was quantified using the MetaMorph software (mean values ± s.e.m.). *P<0.05.

Since the mutant overexpression and Rho/ROCK inhibitor experiments indicated an involvement of Rho, we next determined the amount of active Rho-GTP following insulin stimulation of BHK-IR cells. To directly visualize Rho activation in cells undergoing insulin-mediated actin rearrangement, we chose to use a novel in situ cell-based Rho activity assay (Flevaris et al., 2007). Insulin receptor induced activation of Rho was shown by increased GST-RBD binding to the fixed and permeabilized cells (Fig. 5A). This Rho activation was also seen in cells expressing the active RhoQ63L mutant and was abrogated in cells transfected with annexin A2 siRNA (see Fig. 5A for representative examples, and Fig. 5B for quantification).

A phosphotyrosine-mimicking mutant of annexin A2 dominantly affects actin cytoskeleton rearrangements and acts upstream of Rho/ROCK

To directly address the possible link between insulin-induced tyrosine phosphorylation of annexin A2 and the observed rearrangements in the actin cytoskeleton that led to loss of cell adhesion, we generated a phospho-mimicking annexin A2 mutant. Tyr23, the site identified as a substrate for Src and the insulin receptor (Glenney, Jr and Tack, 1985; Karasik et al., 1988) was replaced by glutamic acid, and the resulting mutant was expressed as a C-terminal GFP fusion. This type of GFP fusion had been shown previously not to affect biochemical and functional properties of annexin A2 (Zobiack et al., 2001). Anxa2-Y23E-GFP and, as a control, wild-type Anxa2-GFP were expressed ectopically in BHK-IR cells and the actin cytoskeleton was stained with TRITC-phalloidin (Fig. 6). Anxa2-GFP-expressing cells displayed a phenotype similar to that of nontransfected cells or cells expressing GFP alone (not shown); they were well spread and flattened with no alterations in the actin cytoskeleton. By contrast, cells expressing the phospho-mimicking mutant Anxa2-Y23E-GFP were severely impaired in their ability to adhere and easily detached from the coverslips. In the remaining cells, the characteristic morphological changes of the `insulin phenotype' – contraction and increase in height – were apparent even without insulin stimulation (Fig. 4). Thus, tyrosine phosphorylation of annexin A2 at Tyr23 appears to play a critical role in the regulation of actin rearrangements and cell to substratum adhesion. We next addressed the question of whether annexin A2 tyrosine phosphorylation is directly linked to Rho/ROCK signalling. Therefore, we first determined the amount of insulin-induced tyrosine phosphorylation of annexin A2 in the absence or presence of Y27632. Addition of the ROCK inhibitor did not alter the level of tyrosine-phosphorylated annexin A2 (not shown), arguing for a role of annexin A2 upstream of ROCK. To further examine the connection between tyrosine phosphorylation of annexin A2 and Rho/ROCK signaling, we treated BHK-IR cells expressing Anxa2-Y23E-GFP with Y27632. Interestingly, ROCK inhibition abrogated the dominant-active effect of the phospho-mimicking mutant on the formation of actin-containing domes in the cell periphery (Fig. 7A). In the presence of Y27632 the Anxa2-Y23E-GFP-expressing cells maintained a spread and flattened morphology, strongly suggesting a role of tyrosine-phosphorylated annexin A2 upstream of the Rho/ROCK signaling pathway in insulin-stimulated BHK-IR cells (Fig. 7B).

Fig. 6.

The phospho-mimicking annexin A2 mutant Anxa2-Y23E-GFP induces morphological changes. BHK-IR cells grown on coverslips were transfected with either wild type (Anxa2-GFP) or mutant Y23E annexin A2 (Anxa2-Y23E-GFP) and stained for F-actin with TRITC-phalloidin. x-y, x-z and y-z sections of representative transfected and nontransfected cells are shown with their respective typical F-actin stain. Note the increase in height and the concomitant appearance of actin domes or sheaths in cells expressing Anxa2-Y23E-GFP. Scale bars: 10 μm.


In an attempt to identify components acting downstream of receptor tyrosine kinases in the regulation of actin dynamics, we used BHK-IR cells as a model system for such regulations. We show here that the membrane- and actin-binding protein annexin A2 is a direct insulin receptor substrate mediating insulin-induced actin rearrangements that lead to cell rounding and substrate detachment. Annexin A2 functions upstream of the Rho/ROCK pathway that is activated in the insulin-treated cells.

Annexin A2 is found at the membrane-actin cytoskeleton interface, showing a preference for membrane regions enriched in cholesterol and PtdIns(4,5)P2. Such microdomains, also termed lipid rafts, are thought to function as signaling platforms for the regulation and transduction of extracellular signals in processes often involving dynamic rearrangements of the actin cytoskeleton (for a review, see Simons and Toomre, 2000). The recent identification of annexin A2 as a PtdIns(4,5)P2-binding protein (Hayes et al., 2004; Rescher et al., 2004), its ability to interact with F-actin (Gerke et al., 2005) and its function as a regulator of actin polymerization at the barbed ends (Hayes et al., 2006) are in line with this localization, and support the view that the protein participates in providing and/or regulating links between certain membrane domains and the actin cytoskeleton, thereby affecting the state of membrane-associated actin. A specific association of annexin A2 with sites of actin remodeling has indeed been reported in several experimental models. They include actin-rich pedestals induced by infection with noninvading enteropathogenic E. coli (EPEC) (Zobiack et al., 2002) and submembranous actin rearrangements initiated by clustering of the hyaluronate receptor CD44, which resides in membrane microdomains (Oliferenko et al., 1999). Annexin A2 is also found on actin tails, which propel motile macropinosomes showing a specific enrichment at the membrane-actin interface. In addition, expression of a dominant-negative annexin A2 mutant protein inhibits the formation of these motile pinosomes (Merrifield et al., 2001). Moreover, a subset of secretory vesicles in epithelial cells that rely on an actin-dependent transport for apical membrane delivery require annexin A2 for this process (Jacob et al., 2004). However, annexin A2 is not found on actin tails of the motile intracellular pathogen Listeria monocytogenes (Merrifield et al., 2001) or on stress fibers in cultured cells. This strongly suggests that annexin A2 is not a general F-actin-binding protein but only acts at the interface between certain cellular membranes and the actin cytoskeleton. At such sites, annexin A2 appears to function by supporting the assembly of membrane lipids and/or recruiting factors initiating actin-remodeling events, such as Rho family GTPases. In support of the latter, annexin A2 was recently shown to recruit Cdc42 to the PtdIns(4,5)P2-rich apical membrane domain of polarized epithelial cells, thereby assisting the induction of the apical actin belt that drives epithelial morphogenesis (Martin-Belmonte et al., 2007). Thus, annexin A2 seems to regulate actin remodelling at cellular membranes by acting upstream of different Rho family GTPases, Cdc42 in polarizing epithelial cells and, as shown here, Rho in insulin-triggered BHK-IR cells.

In insulin-stimulated BHK-IR cells, annexin A2 is one of the major tyrosine-phosphorylated proteins showing a strong and sustained level of tyrosine phosphorylation upon induction of insulin signaling. This phosphorylation pattern is rather unusual for a receptor substrate because it increases for a prolonged period of time. However, the insulin-induced tyrosine phosphorylation in BHK-IR cells, which is caused directly by the kinase activity of the insulin receptor, is generally high and persists for a long time. This is probably due to the high ratio of tyrosine kinase versus counteracting protein tyrosine phosphatase (PTP) activity. Indeed, expression of PTPs has been shown to rescue the cells from insulin-induced detachment and those PTPs that were able to downregulate phosphorylation of annexin A2 were most efficient in generating an insulin-resistant adhesion phenotype (Andersen et al., 2001).

The insulin-induced morphological changes in BHK-IR cells can be linked directly to phosphorylation of Tyr23 in annexin A2. Depletion of intracellular annexin A2 by RNAi prevents these changes from occurring, and re-expression of annexin A2 restores the observed actin phenotype almost completely. Furthermore, the dominant-active phospho-mimicking mutant transforms the cell morphology, bypassing any insulin stimulation. These results not only identify annexin A2 as an insulin receptor substrate required for mediating insulin-induced actin rearrangements, but also show for the first time that tyrosine-phosphorylated annexin A2 can cause the remodeling of membrane-associated actin leading to a loss of cell-substratum contact. This provides the long sought after link between annexin A2 phosphorylation and cell transformation, which was postulated when annexin A2 was identified as a major Src substrate more than 25 years ago (Erikson and Erikson, 1980).

Fig. 7.

ROCK inhibition abrogates the dominant effect of the phospho-mimicking annexin A2 mutant. (A) BHK-IR cells were transfected with the phospho-mimicking mutant Anxa2-Y23E-GFP and either left untreated or were treated for 1 hour with 20 μM of the ROCK inhibitor Y27632. Cells were then fixed and stained for F-actin. Note the induction of the `insulin-responsive' actin phenotype in Anxa2-Y23E-GFP expressing cells (lower panels), which is abrogated by ROCK inhibition (upper panels). Scale bars, 10 μm. (B) Several hundred cells expressing Anxa2-Y23E-GFP were inspected and the relative number of cells showing the normal F-actin distribution (rescued phenotype) was quantified for experiments carried out in the presence (+) or absence (–) of Y27632. **P<0.0005.

The tyrosine phosphorylation switch in annexin A2 can affect changes in the actin cytoskeleton by two principal means. Firstly, it can regulate properties of the annexin A2 molecule itself and thereby interfere with a role of the protein as membrane-cytoskeleton linker or membrane domain organizer. In line with this hypothesis, it has been shown that tyrosine phosphorylation of annexin A2 in vitro inhibits the ability of the protein to bind F-actin and to form a higher order complex together with plasma membrane and secretory vesicles (Hubaishy et al., 1995). Secondly, tyrosine-phosphorylated annexin A2 could act in a dominant-active manner by upregulating the Rho/ROCK pathway. Rho/ROCK signaling is induced by insulin treatment and this induction requires the presence of annexin A2. Moreover, Rho/ROCK activation is required to elicit the actin cytoskeleton rearrangements observed in BHK-IR cells and with Rho/ROCK most likely acting downstream of tyrosine-phosphorylated annexin A2 since the dominant effect of the Anxa2-Y23E mutant is abrogated by ROCK inhibition. Whether phosphorylated annexin A2 in this scenario solely acts as a signal transducer or functions by recruiting components of the Rho/ROCK signaling pathway to certain membrane sites remains to be determined.

The activation of Rho-dependent ROCK signaling and subsequent cellular blebbing or dome formation has also been linked to a certain mode of cell motility in 3D matrices, and it seems to correlate with the metastatic potential of tumor cells (Sahai and Marshall, 2003). The generation of contractile force through Rho signaling via ROCK has been described to promote the disruption of junctions thereby causing disintegration of the epithelium (Sahai and Marshall, 2002). Thus, Rho/ROCK signaling can induce a more metastatic and motile cell phenotype. The findings linking annexin A2 to this Rho/ROCK pathway and thereby to the regulation of cell-substrate interaction and cell dynamics are also in line with an effect of annexin A2 on cell motility that can be inferred from a number of recent observations. The plant metabolite withaferin A, a potent inhibitor of cell-substrate adhesion, was found to specifically bind to annexin A2 and this interaction was linked to the withaferin-A-induced actin filament aggregation and an inhibition of tumor cell invasion (Falsey et al., 2006). In epithelial cell layers, annexin A2 was shown to regulate cell spreading and wound closure in line with an increased expression of annexin A2 in migrating versus stationery epithelial cells (Babbin et al., 2007). Furthermore, in endothelial cells the reduction of annexin A2 plasma membrane levels by cholesterol depletion or RNAi inhibited the recruitment of E-cadherin and thereby cell contact formation, which itself negatively regulates migration (Yamada et al., 2005). It will be interesting to learn whether the effect of annexin A2 on cell contact formation is also regulated by its tyrosine phosphorylation, which has been established here as a molecular switch transforming cellular actin from a stationary to a more migratory phenotype.

Materials and methods

Antibodies, inhibitors, annexin A2 siRNA and expression plasmids

Annexin A2 was detected with the mouse monoclonal antibody HH7 (Osborn et al., 1988). The monoclonal anti-phosphotyrosine antibody 4G10 was purchased from Upstate. Monoclonal anti-Myc-antibody (9E10), monoclonal anti-GST-antibody (B-14) and the polyclonal antibody to insulin receptor β chain (IRβ) were from Santa Cruz. The structurally unrelated selective ROCK inhibitors Y27632 and H-1152 were obtained from Calbiochem. The cell-permeable Rho inhibitor, consisting of the Clostridium botulinum Exoenzyme C3 transferase covalently linked to a cell-penetrating moiety, was obtained from Cytoskeleton.

RNAi to silence annexin A2 expression was performed using the silencing and control RNA duplexes described (Zobiack et al., 2003). To generate Anxa2-Y23E-GFP with a mutated tyrosine phosphorylation site, amino acid Tyr23 was replaced by glutamate using site-directed mutagenesis (QuikChange, Stratagene). The wild-type annexin A2-GFP and the dominant-active RhoAQ63L expression constructs have been described previously (Mayer et al., 1999; Rescher et al., 2004).

Cell culture and transfection

The BHK cell line (BHK-IR) stably expressing the human insulin receptor was kindly provided by Axel Ullrich (Max-Planck Institute for Biochemistry, Martinsried, Germany). Cells were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum, 2 mM L-glutamine, 1 μM methotrexate and penicillin/streptomycin in a humidified 5% CO2 atmosphere at 37°C. For transfections with annexin A2 expression plasmids or siRNA duplexes, Effectene (Qiagen) or Oligofectamine (Invitrogen), respectively, were used according to the manufacturers' protocols.

Analysis of cell adhesion and morphology

BHK-IR cells grown in 24-well plates were serum-starved overnight. Cells were then cultivated in the presence or absence of 100 nM insulin for an additional 4 hours. Subsequently, wells were washed three times with 3 ml PBS and the nonadherent cells were pelleted from the combined wash fractions. Remaining adherent cells were trypsinized and the numbers of nonadherent and adherent cells were determined using a Coulter counter. Mean values ± s.e.m. were calculated, and the statistical significance of the results was evaluated by unpaired t-tests.

For microscopic analysis of insulin-induced changes in morphology, cells were cultured on coverslips, serum-starved overnight, pre-treated with inhibitors, cultured for 30 or 60 minutes in the presence or absence of 100 nM insulin and then fixed. F-actin was stained using TRITC-phalloidin (Sigma). Images of cells were collected digitally with a 40× or a 63× 1.4 NA objective using appropriate filter settings on a Zeiss LSM 510 META confocal microscope. For quantification of insulin-induced changes in morphology, the very apical and basal part of cells were marked using z-sectioning, and the medial section was automatically scanned and then used for image analysis of the phalloidin staining. Images were taken from different positions on the coverslips chosen randomly and data were collected from more than three independent experiments. In each case, the proportion of cells that showed the typical `insulin-responsive' actin phenotype, i.e. actin domes and loss of stress fibers, was determined. Mean values ± s.e.m. were calculated, and the statistical significance of the results was evaluated by unpaired t-tests.

Cell lysis, immunoprecipitation and immunoblotting

Cells were serum-starved overnight and treated with 100 nM insulin for different times. Subsequently, cells were placed on ice and incubated for 30 minutes in lysis buffer (50 mM HEPES, pH 7.4, 150 mM NaCl, 1% Triton X-100, 10% glycerol, 1 mM EDTA) containing 2 mM vanadate, 10 mM NaF, 2 mM PMSF, 1 μg/ml aprotinin and 1 μg/ml leupeptin. Cell lysates were then centrifuged at 10,000 g for 10 minutes at 4°C and the supernatants were used for immunoprecipitation and immunoblotting. Cleared lysates containing equal amounts of total protein were incubated for 2 hours with the HH7 antibody bound to magnetic Dynal M-450 anti-mouse IgG beads (Invitrogen) at 4°C. After several washing steps, the immunocomplexes were released by boiling in SDS-PAGE sample buffer and loaded onto 12% SDS-PAGE gels. Proteins were then transferred onto PVDF membrane, the membrane was blocked with 3% nonfat dry milk and probed with the respective antibodies and appropriate secondary antibodies conjugated to horse radish peroxidase. Immunoreactive bands were visualized using enhanced chemiluminescence (AppliChem) and signal intensities were measured using a Boehringer Mannheim Lumi-Imager with the appropriate software. The blots shown are typical results of at least three individual experiments.

Immunofluorescence microscopy

After fixation with 4% paraformaldehyde, cells were permeabilized with 0.2% Triton X-100/PBS, quenched with 50 mM NH4Cl in PBS, and incubated with TRITC-phalloidin or the respective primary antibodies for 30 or 45 minutes, respectively. After additional washing with PBS, specimens were treated with the appropriate labeled secondary antibody for 45 minutes. Cells were then mounted in Mowiol containing N-propyl-gallat as anti-fading reagent and inspected in a Zeiss LSM 510 META confocal microscope.

Inhibition of the Rho/ROCK pathway

Serum-starved cells grown on coverslips were pre-incubated with either the cell-permeable Rho-inhibiting C3 exotoxin (2.5 μg/ml, 6 hours) or the ROCK inhibitors Y27632 (20 μM, 1 hour) or H-1152 (5 μM, 1 hour) prior to insulin stimulation for 30 minutes. The fixed cells were stained for F-actin and morphological changes in the actin cytoskeleton were quantified as described above.

Cell-based Rho activity assay

Rhotekin Rho-binding domain (RBD) with the glutathione-S-transferase (GST) at the N-terminus (GST-RBD) was recombinantly expressed in E. coli strain BL21(DE3)pLysS (Stratagene) and purified essentially as described (Ren and Schwartz, 2000). Protein expression was induced with 0.25 mM IPTG (isopropyl β-D-thiogalactoside) for 3 hours at 30°C. To analyze the GTP-loading state of intracellular Rho, Rho activity measurements were carried out in situ as described previously (Flevaris et al., 2007). In brief, cells grown on coverslips were treated as indicated, fixed with 4% paraformaldehyde, 10 mM MgCl2. GST-RBD (30 μg/ml) and TRITC-phalloidin in permeabilization buffer (0.1% Triton X-100, 0.1 M Tris-HCl, pH 7.5, 10 mM EGTA, 150 mM NaCl, 10 mM MgCl2, 1 mM PMSF, 0.1 mM E64, 1%BSA) were added to the cells for 20 minutes. Subsequently, cells were washed, and the bound GST-RBD was visualized by immunofluorescence staining of the GST moiety of the fusion protein. To obtain both the F-actin and the GST-RBD signal, images from at least ten randomly chosen positions were taken using the same acquisition settings to allow for comparison of signal intensities and the multitrack acquisition mode to prevent signal crossover. Digital images were imported in MetaMorph (Universal Imaging) and thresholded in noncellular regions of each image to define background pixel intensity. The outlines of individual cells were defined and all pixels of the GST-RBD signal with intensities above the threshold within a given cell were recorded. Data were collected from more than three independent transfection and labeling experiments and are presented as mean pixel gray value intensity ± s.e.m. of more than 1000 cells inspected. The statistical significance of the results was evaluated by unpaired t-tests.


This study was supported by grants from the German Research Society (DFG, SFB 629) and the Interdisciplinary Clinical Research Centre (IZKF) of the University of Muenster (Re2/039/07).


  • * These authors contributed equally to this work

  • Accepted April 9, 2008.


View Abstract