Actin microdomains on endothelial cells: association with CD44, ERM proteins, and signaling molecules during quiescence and wound healing
- First Online:
- Cite this article as:
- Jensen, P.V. & Larsson, L. Histochem Cell Biol (2004) 121: 361. doi:10.1007/s00418-004-0648-2
- 165 Views
During studies of the actin cytoskeleton in cultured endothelial cells we have observed that the luminal side of many cells contains F-actin microdomains that are rich in the hyaluronan receptor CD44 and in ezrin-radixin-moesin (ERM) proteins. A small subpopulation of the domains are also enriched in tyrosine phosphorylated proteins and signaling molecules. Confocal microscopy of rat aortic endothelial cells in situ demonstrated that similar microdomains occur in vivo. During healing of endothelial wounds, characteristic alterations of the actin cytoskeleton occurred. Thus, in many cells close to the wound, focal F-actin branching points appeared. The branching points were similar to the microdomains in that they colocalized with CD44 and ERM proteins, but, in addition, they formed centers for actin filament branching and were associated with phosphorylated protein kinase C α/βII. These colocalization data are consonant with the view that activated PKC is responsible for activating ERM-mediated crosslinking between CD44 and the actin cytoskeleton. Importantly, inhibition of PKC activity decreased staining for phosphorylated ERM proteins, decreased the frequency of F-actin branching points, and inhibited monolayer wound healing. Together, our data show that endothelial cells contain a novel actin cytoskeletal structure, the F-actin microdomain, and suggest that during wound healing such structures become associated with activated signaling molecules and thereby enhance actin cytoskeletal remodeling.
KeywordsEndothelial cellsWound healingActin cytoskeletonCD44ERM proteinsPKC
Endothelial cells line the inside of blood vessels and sense changes in blood pressure and flow through hemodynamic forces in the form of shear stress and mechanical strain. Such forces act on the actin cytoskeleton and transduce signals regulating endothelial cell nitric oxide synthase expression as well as endothelial cell proliferation and function (Lehoux and Tedgui 2003; Li et al. 2002; Resnick et al. 2003). In addition, a host of soluble signals, including growth factors, cytokines, hormones, and neurotransmitters, affect endothelial cell function (Bernardini et al. 2003; Li et al. 2003; Rubanyi et al. 2002; Vincent et al. 2003; Walch et al. 2001). One of these functions, the formation of new vessels from preexisting vasculature through angiogenesis, is important to many disease states as well as to normal physiology and wound healing (Battegay 1995; Folkman 1995, 2000; Liotta et al. 1991). Angiogenesis depends upon remodeling of the actin cytoskeleton and endothelial cell motility, phenomena which are regulated by many factors including growth factors and chemokines, and interactions with other cells and with the extracellular matrix (Bernardini et al. 2003; Ingber et al. 1995; Li et al. 2003; Nguyen and D’Amore 2001; Uchida et al. 2000).
During studies of the endothelial actin cytoskeleton we observed that endothelial cells contained luminal F-actin microdomains, which were enriched in the hyaluronan receptor CD44. In addition, the microdomains contained activated (phosphorylated) ezrin-radixin-moesin (phospho-ERM) proteins, which have been shown to crosslink CD44 and other receptors to the actin cytoskeleton (Algrain et al. 1993; Bretscher et al. 2002; Legg and Isacke 1998; Tsukita et al. 1994; Yonemura et al. 1998). A subpopulation of preferentially larger F-actin microdomains were also enriched in signaling molecules and in tyrosine phosphorylated proteins. F-actin microdomains appear not to have been described before and could conceivably act to integrate mechanical and soluble signals during processes like angiogenesis and wound healing. During wound healing of endothelial cell monolayers we demonstrate that many cells close to the wound respond rapidly by forming F-actin branching points enriched in CD44, phospho-ERM proteins and phosphorylated protein kinase C α/βII (phospho-PKC α/βII). The F-actin branching points were similar to F-actin microdomains but, in addition, contained phospho-PKC α/βII and showed a radiating pattern of actin fibers that connected them with the rest of the actin cytoskeleton. These colocalization data suggested that activated PKC was responsible for activating ERM-mediated crosslinking between CD44 and the actin cytoskeleton. In support of this view, inhibition of PKC activity decreased staining for phosphorylated ERM proteins, reduced the frequency of F-actin branching points, and inhibited wound healing. These data are the first to examine the localization of these proteins during wound healing and the inhibitor results together with the colocalization of CD44, phosphorylated ERM proteins, and PKC form strong reasons to propose a functional interaction between these proteins during wound healing.
Materials and methods
Human umbilical vein endothelial cells (HUVECs; ATCC, Manassas, VA) were grown in Kaighn’s F-12 K medium supplemented with 10% fetal calf serum, 100 U/ml penicillin, 100 μg/ml streptomycin, 2 mM l-glutamine (Gibco, Grand Island, NY), and 0.4% endothelial cell growth supplement/heparin (PromoCell, Heidelberg, Germany) at 37°C in a humidified 5% CO2 atmosphere. HUVECs from passages 19–25 were used for all experiments. For studies of intact endothelial cell monolayer cultures, HUVECs were trypsinized, seeded on sterile glass slides (800,000 cells/ml), grown overnight or to confluence, and fixed. Additionally, confluent monolayers were longitudinally wounded using a pipette tip. GF-109203X (bisindolylmaleimide I HCl; Alexis Biochemicals, San Diego, CA) dissolved in dimethylsulfoxide (DMSO; Sigma-Aldrich, St. Louis, MO; final concentration 0.1%) was added to the media to a final concentration of 3 μM or 10 μM. Controls were treated with 0.1% DMSO. Cells were preincubated with inhibitor or vehicle for 1 h prior to wounding and fixed 20 or 60 min after wounding. For studies of effects on wound healing, cultures were fixed 24 or 48 h after wounding. Fixation was in equal parts of methanol/acetone (Baker, Deventer, The Netherlands) or in 3.7% paraformaldehyde (Merck, Darmstadt, Germany). The latter was followed by permeabilization in 1% Triton X-100 (Merck).
Non-cultured endothelial cells
Two adult male Wistar rats received a lethal dose of sodium pentobarbital intraperitoneally and were intracardially perfused with 60 ml of 0.9% sodium chloride followed by 60 ml of 3.7% paraformaldehyde. Specimens from the aorta were postfixed in the same fixative and permeabilized in 1% Triton X-100. Before staining, part of the muscle wall was stripped away with a pair of forceps.
Immunofluorescence and phalloidin staining
To visualize F-actin, cells were stained with Alexa 594-labeled phalloidin (Molecular Probes, Eugene, OR). Primary antibodies included mouse monoclonal antibodies to CD44 (Dako, Glostrup, Denmark), ezrin, phosphatidylinositol 3-kinase (PI3 K), protein tyrosine phosphatase 1B (PTP1B) (Transduction Laboratories, Lexington, KY), phosphotyrosine, insulin receptor substrate 1 (IRS-1), gastrin (Upstate Biotechnology, Lake Placid, NY), VEGF receptor 1 or 2 (Sigma-Aldrich), and affinity-purified rabbit antibodies specific for phosphorylated ezrin (Thr567)/radixin (Thr564)/moesin (Thr558) (phospho-ERM) or phosphorylated protein kinase C α/βII (Thr638/641) (phospho-PKC α/βII) (Cell Signaling Technology, Beverly, MA). The site of antigen–antibody reaction was revealed with goat anti-mouse Ig or goat anti-rabbit Ig antibodies labeled with Alexa 488 or Alexa 594 (Molecular Probes). Controls included use of type-matched monoclonal antibodies, immunoblotting (vide infra), as well as conventional staining controls (Larsson 1988). Double labeling employed primary antibodies raised in different species combined with species-specific second antibodies labeled with contrasting fluorochromes. Analysis of the effects of different treatments on actin and protein immunoreactivities was assessed in blind-coded specimens. Specimens were analyzed either by conventional fluorescence microscopy or in a Molecular Dynamics confocal laser scanning microscope (MultiProbe 2001). All experiments were repeated three to five times. In each experiment, a minimum of two slides were stained with each primary antibody.
Cells were lysed in 50 mM HEPES buffer, pH 7.6 (Sigma-Aldrich), containing 100 mM NaCl, 10 mM EDTA, 1% Triton X-100 (Merck), 4 mM sodium pyrophosphate (Sigma-Aldrich), 2 mM sodium orthovanadate, 10 mM NaF (Acros, Geel, Belgium), and a Complete tablet of protease inhibitors (Boehringer Mannheim, Hoersholm, Denmark). Aliquots were removed for protein determination and samples were diluted with 4×NuPage sample buffer and NuPage sample reducing agent was added to a final concentration of 10%. Samples were heated to 70°C for 10 min. Lysates were separated on NuPage 10% BisTris gel using NuPage MOPS SDS running buffer in an EI9001-Xcell II Mini cell apparatus. SeeBlue Plus2 prestained protein standard (Novex, San Diego, CA) was used as marker (191–14 kDa). Electroblotting was performed in a semidry blotting apparatus as recommended by the manufacturer (KemEnTec, Copenhagen, Denmark) using 0.2 μm PVDF immunoblot membranes (BioRad, Hercules, CA).
Immunoblots were stained with antibodies detecting CD44, ezrin, phospho-ERM, and phospho-PKC α/βII. Western Breeze chemiluminescence kit (Novex) and Hyperfilm ECL RPN2103 K (Amersham, Buckinghamshire, UK) were used for detection.
Quantitation of monolayer wound healing
Digital images of coded specimens stained with Giemsa (Merck) were acquired using an Olympus BH-2 microscope equipped with a monochrome Sony CCD camera connected to a PC with ImageProPlus software. Images of the wound area were captured through a 10× objective. A rectangle, 820 μm wide and 1,050 μm long, was placed symmetrically over the wound (570.9±37.0 μm in width). All cells within the rectangle, except those touching the lower or left border, were counted. For each slide, three separate areas were counted. Results were presented as means ± standard deviation (SD) and statistics were computed using the Mann-Whitney U-test. Experiments were repeated three times; in the two initial experiments three control and three experimental slides were analyzed, whereas in the final experiment seven slides of each kind were analyzed.
Quantitation of F-actin branching points
Specimens were stained with Alexa 594-labeled phalloidin, coded, and the number of branching points in 100 cells per slide was counted. Only cells bordering the wound were included. Five slides treated with GF-109203X and five slides treated with vehicle only were analyzed. Results were presented as means ± SD and statistics were computed using the Mann-Whitney U-test. In both types of experiments the level of significance was set at 0.05.
ERM proteins need to unfold in order to crosslink CD44 to the actin cytoskeleton, and this process is catalyzed by phosphorylation on certain critical threonine residues (Thr567 of ezrin, Thr564 of radixin, Thr558 of moesin; Gautreau et al. 2000; Matsui et al. 1998; Nakamura et al. 1995). We therefore double stained HUVECs with the monoclonal CD44 antibody and an affinity-purified rabbit antibody specific for ERM proteins phosphorylated on such threonine residues (phospho-ERM) (Fig. 1E, F). The majority, but not all, CD44-positive areas stained for phospho-ERM. Since, as noted above, CD44 staining also occurred outside F-actin microdomains, we next double stained monolayers with fluorescent phalloidin and the phospho-ERM antibody in order to determine whether the phospho-ERM-immunoreactive areas corresponded to F-actin microdomains (Fig. 1G, H). By this approach we demonstrated that the vast majority of the F-actin microdomains also were positive for phospho-ERM. Thus, by inference, we may conclude that the vast majority of F-actin microdomains were immunopositive for both CD44 and phospho-ERM. In addition, double staining with fluorescent phalloidin and a monoclonal antibody detecting ezrin showed that ezrin-immunoreactive proteins were associated with most of the F-actin microdomains (data not shown). All controls were negative and immunoblotting studies using the monoclonal ezrin antibody revealed a single immunoreactive band (Fig. 2) corresponding in size to that of ezrin. A similarly sized, but much fainter, band was revealed by the phospho-ERM antibody (Fig. 2).
Together, these data show that HUVECs grown in monolayers contain luminal F-actin microdomains, which are enriched in CD44, possibly by a mechanism involving phospho-ERM-mediated crosslinking between F-actin and CD44.
We next examined the cells for the presence of additional receptors. Immunostaining for VEGF receptor types 1 or 2 revealed that these virtually never occurred clustered in dorsal F-actin microdomains, but occurred diffusely distributed in the cell membrane (data not shown). All staining controls were negative and a subtype-matched irrelevant antibody failed to stain the cells.
PKC isoforms α and θ are known to phosphorylate ERM proteins on the critical threonine residues needed for unfolding (Gautreau et al. 2000; Matsui et al. 1998; Nakamura et al. 1995; Ng et al. 2001; Pietromonaco et al. 1998; Simons et al. 1998) and, in addition, ligands binding to CD44 have been found to activate PKC signaling (Slevin et al. 2002). We therefore examined the actin microdomains for their association with PKC. An antibody to phospho-PKC α/βII preferentially stained some of the larger (>5 μm) F-actin microdomains and, additionally, stained the vast majority of nuclei (Fig. 1I, J). In contrast, little or no staining of the microdomains was obtained with antisera to PI3 K or PTP1B (data not shown). Again, all staining controls were negative and subtype-matched irrelevant antibodies failed to stain the cells. Immunoblotting studies with the phospho-PKC α/βII antibody revealed an immunoreactive band of the size expected for PKC α (Fig. 2).
Thus, these data show that predominantly some of the larger F-actin microdomains contain phosphorylated PKC α/βII and the immunoblotting results indicate that the isoform detected is PKC α. The involvement of a subpopulation of the F-actin microdomains in active signaling was also underlined by the fact that a subpopulation of preferentially larger domains also stained for phosphotyrosine (Fig. 1K, L) as well as for the signaling intermediate IRS-1 (Fig. 1M, N).
The majority of F-actin branching points also demonstrated staining for phospho-ERM proteins (Fig. 4I–K). Additionally, the phospho-ERM-positive branching points also stained for CD44 (Fig. 4C–E) and some of the cells exhibited membrane ruffles that were immunopositive for CD44 and phospho-ERM (Fig. 4C–E). Antibodies recognizing other cellular proteins like PTP1B reacted with other cellular areas, but not with the F-actin branching points, attesting to the specificity of the immune staining (Fig. 4O–Q). Antibodies detecting phospho-PKC α/βII reacted with all F-actin branching points (Fig. 4F–H). Thus, these results demonstrate that phospho-PKC α/βII associates with the focal F-actin branching points of wounded monolayers.
Our results show that cultured, untransformed HUVECs contain F-actin microdomains that are selectively enriched in certain types of receptors. Since cultured endothelial cells gradually loose their characteristics during passaging, we additionally studied rat aortic endothelial cells in situ and documented that also these cells contained luminally oriented actin microdomains. The presence of actin microdomains seems not to have been noticed previously. The striking colocalization observed between F-actin, CD44, and phospho-ERM proteins in the microdomains may reflect the known interactions between these proteins (Algrain et al. 1993; Bretscher et al. 2002; Legg and Isacke 1998; Tsukita et al. 1994; Yonemura et al. 1998). Thus, ERM proteins can exist both in a folded, inactive form and in an unfolded, active form. In the unfolded form they are capable of crosslinking certain transmembrane proteins, like CD44, to the actin cytoskeleton (Gautreau et al. 2000; Matsui et al. 1998; Nakamura et al. 1995). Importantly, unfolding of ERM proteins requires phosphorylation on critical threonine residues (Thr567 of ezrin, Thr564 of radixin, Thr558 of moesin) and such phosphorylation may be catalyzed by PKC α or θ (Gautreau et al. 2000; Matsui et al. 1998; Nakamura et al. 1995; Ng et al. 2001; Pietromonaco et al. 1998; Simons et al. 1998) or by Rho kinase (Matsui et al. 1998). It is therefore interesting to note that staining for phospho-PKC α/βII was associated with a subpopulation of F-actin microdomains. Moreover, immunoblotting studies with this antiserum revealed a single band migrating at a size corresponding to that of PKC α. Most of the PKC-immunopositive domains were of large size and similar such microdomains also stained for tyrosine phosphorylated proteins and IRS-1, which have been implicated in cell signaling events (reviewed by Hunter 1998). Together these results argue that a small subpopulation of large microdomains were associated with signaling events. This indicates that the latter microdomains are activated, while the majority of small microdomains only contain clustered, unactivated receptors. Importantly, receptor clustering may enhance sensitivity to soluble ligands by virtue of the close proximity of receptors and signaling adaptor molecules (Bray et al. 1998; Kim et al. 2002) and the association with actin microdomains may enhance effects of signaling on the actin cytoskeleton.
Our data also show that wounding of endothelial cell monolayers brings about a rapid rearrangement of the F-actin cytoskeleton in cells close to the wound edge. This rearrangement is characterized by the gradual rounding up of many of the cells resulting in a more motile phenotype, characterized by numerous membrane ruffles, and in the establishment of characteristic F-actin branching points. Importantly, the focal F-actin branching points stained both for the hyaluronan receptor CD44 and for ERM proteins. The F-actin branching points are similar to microdomains but are associated with branching actin filaments and with phospho-PKC α/βII. These data suggest that F-actin microdomains of cells close to the wound edge become associated with phosphorylated PKC isoforms that activate ERM-mediated crosslinking between CD44 and F-actin. This, in turn, results in increased numbers of connections between the microdomains and the rest of the actin cytoskeleton in the form of F-actin branching points.
Conventional and novel PKC isoforms require phosphorylations on specific threonine or serine residues in order to become activated by diacylglycerol (Liu and Heckman 1998). We therefore used antibodies recognizing PKC α/βII phosphorylated on specific threonine residues. Importantly, this does not necessarily imply that only active PKC α/βII was detected, but shows that molecules that fulfill one requirement for activation were detected. Direct support for a role of PKC in local phosphorylation of ERM proteins and activation of crosslinking between CD44 and F-actin came from use of a PKC inhibitor, which markedly reduced staining for phospho-ERM and reduced the number of F-actin branching points. The inhibitor (GF-109203X) has been reported to be a specific PKC inhibitor and is commonly used in concentrations from 1 to 10 μM, typically with an IC50 in the range of 1–5 μM (Habib et al. 1997; Higaki et al. 1999; Toullec et al. 1991; Villard et al. 1998). This concentration range has been found to be needed in studies of intact cells although studies of purified enzymes show that, in the test tube, the IC50 is considerably lower (Toullec et al. 1991). In agreement with previous studies (Tang et al. 1997; Wang et al. 2002) we found that inhibition of PKC activity markedly inhibited wound healing. However, the dose needed to demonstrate this effect was in the upper range of concentrations commonly used. This could reflect either that the cells studied were relatively poor in taking up the drug or that they metabolized it quickly. Alternatively, it could reflect that the drug also inhibited other enzymatic activities. Whatever the explanation is, it is important to note that the concentrations that were needed for significant inhibition of wound healing were also needed for inhibition of staining for phospho-ERM proteins.
In conclusion, our studies demonstrate a novel actin cytoskeletal structure, the F-actin microdomain. In quiescent cells in vitro and in situ these microdomains associate with CD44 and ERM proteins, but are for the most part not associated with signaling molecules. However, important rearrangements of the localization of the actin cytoskeleton, phospho-ERM proteins, and the hyaluronan receptor CD44 to focal F-actin branching points occur in response to monolayer wounding. To the best of our knowledge this is the first description of such a rearrangement. The colocalization of the F-actin branching points with phospho-PKC α/βII in combination with the inhibitor studies suggest that PKC isoforms are important both for monolayer wound healing, phosphorylation of ERM proteins, and for the very formation of focal F-actin branching points. The morphology of the F-actin branching points was very similar to that observed for F-actin microdomains in HUVEC monolayers. However, the latter did not present extensive connections to the remaining actin cytoskeleton, nor were most of them associated with staining for phospho-PKC α/βII. We therefore suggest that F-actin microdomains on quiescent cells become activated in response to wound healing and that activated PKC α/βII (and possibly additional kinases) serves to induce activation of ERM proteins, so that extensive crosslinks between CD44 and the actin cytoskeleton are formed, resulting in the formation of focal F-actin branching points. This represents an early response to wounding that may be important to the rearrangement of the actin cytoskeleton and, hence, to wound healing per se.
Grant support was from the Danish Medical Research Council, the Danish Veterinary and Agricultural Research Council, and the Danish Cancer Society. P.V. Jensen was supported by a PhD scholarship from the Royal Veterinary and Agricultural University. We thank Mrs. B. Traasdahl for performing image analyses.