Glycogen synthase kinase-3 regulates cytoskeleton and translocation of Rac1 in long cellular extensions of human keratinocytes Leeni Koivisto, a Lari Ha ¨kkinen, a Kazue Matsumoto, b Christopher A. McCulloch, c Kenneth M. Yamada, b and Hannu Larjava a, * a Department of Oral Biological and Medical Sciences, Faculty of Dentistry, University of British Columbia, Vancouver, BC, Canada V6T 1Z3 b National Institute of Dental and Craniofacial Research, National Institutes of Health, Bethesda, MD 20892-4370, USA c University of Toronto, Toronto, ON, Canada M5S 3E2 Received 3 July 2003 Abstract Wound keratinocytes form long cellular extensions that facilitate their migration from the wound edge into provisional matrix. We have previously shown that similar extensions can be induced by a long-term exposure to EGF or rapidly by staurosporine in cultured cells. This morphological change depends on the activity of glycogen synthase kinase-3 (GSK-3). Here, we have characterized the cytoskeletal changes involved in formation of these extended lamellipodia (E-lam) in human HaCaT keratinocytes. E-lams contained actin filaments, stable microtubules and keratin intermediate filaments. E-lam formation was prevented by cytochalasin D, colchicine and low concentrations of taxol and nocodazole, suggesting that actin and microtubule organization and dynamics are essential for E-lam formation. Staurosporine induced recruitment of filamentous actin (F-actin), cortactin, filamin, Arp2/3 complex, Rac1 GTPase and phospholipase C-g1 (PLC-g1) to lamellipodia. Treatment of cells with the GSK-3 inhibitors SB-415286 and LiCl 2 inhibited E-lam formation and prevented the accumulation of Rac1 and Arp2/3 complex at lamellipodia. The formation of E-lams was dependent on fibronectin-binding integrins and normally regulated Rac1, and expression of either dominant-negative or constitutively active forms of Rac1 prevented E-lam formation. Overexpression of either RhoA or Cdc42 GTPases suppressed E-lam formation. We conclude that extended lamellipodia formation in keratinocytes requires actin and tubulin assembly at the leading edge, and this process is regulated by Rac1 downstream of GSK-3. D 2003 Elsevier Inc. All rights reserved. Keywords: Lamellipodia; Actin; Tubulin; Rho GTPases Introduction During re-epithelization of wounds, keratinocyte migra- tion on the fibronectin-rich, provisional wound matrix starts with the formation of long lamellipodia that extend into the wound [1–4]. Coordinated changes in the organization of the cytoskeleton are required for lamellipodia formation and cell migration. In keratinocytes, microtubules are responsible for the orientation and polarization of the cell [5]. Actin is involved in regulating cell shape and motility, whereas keratin inter- mediate filaments provide structural integrity and also participate in organization of actin filaments and micro- tubules [5,6]. Cell migration requires interplay between the actin and microtubule cytoskeletal systems and modulation of their organization [7,8]. The organization of filamentous actin (F-actin) during cell motility can affect microtubule assembly and conversely, microtubules may provide a vector for directing sites of F-actin nucleation and assembly [9]. The small GTPases of the Rho family, namely RhoA, Rac1 and Cdc42, have been identified as key molecules regulating morphological changes as they differentially reg- ulate the formation of actin stress fibers, lamellipodia and filopodia, respectively [10,11]. Only a limited amount of 0014-4827/$ - see front matter D 2003 Elsevier Inc. All rights reserved. doi:10.1016/j.yexcr.2003.09.026 Abbreviations: E-lam, extended lamellipodium; F-actin, filamentous actin; GSK-3, glycogen synthase kinase-3; PLC-g1, phospholipase C-g1; PP1, protein phosphatase 1; VSV, vesicular stomatitis virus. * Corresponding author. Department of Oral Biological and Medical Sciences, Faculty of Dentistry, University of British Columbia, 2199 Wesbrook Mall, Vancouver, BC, Canada V6T 1Z3. Fax: +1-604-822-3562. E-mail address: [email protected] (H. Larjava). www.elsevier.com/locate/yexcr Experimental Cell Research 293 (2004) 68 – 80
13
Embed
Glycogen synthase kinase-3 regulates cytoskeleton and translocation of Rac1 in long cellular extensions of human keratinocytes
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
www.elsevier.com/locate/yexcr
Experimental Cell Research 293 (2004) 68–80
Glycogen synthase kinase-3 regulates cytoskeleton and translocation of
Rac1 in long cellular extensions of human keratinocytes
Leeni Koivisto,a Lari Hakkinen,a Kazue Matsumoto,b Christopher A. McCulloch,c
Kenneth M. Yamada,b and Hannu Larjavaa,*
aDepartment of Oral Biological and Medical Sciences, Faculty of Dentistry, University of British Columbia, Vancouver, BC, Canada V6T 1Z3bNational Institute of Dental and Craniofacial Research, National Institutes of Health, Bethesda, MD 20892-4370, USA
cUniversity of Toronto, Toronto, ON, Canada M5S 3E2
Received 3 July 2003
Abstract
Wound keratinocytes form long cellular extensions that facilitate their migration from the wound edge into provisional matrix. We have
previously shown that similar extensions can be induced by a long-term exposure to EGF or rapidly by staurosporine in cultured cells.
This morphological change depends on the activity of glycogen synthase kinase-3 (GSK-3). Here, we have characterized the cytoskeletal
changes involved in formation of these extended lamellipodia (E-lam) in human HaCaT keratinocytes. E-lams contained actin filaments,
stable microtubules and keratin intermediate filaments. E-lam formation was prevented by cytochalasin D, colchicine and low
concentrations of taxol and nocodazole, suggesting that actin and microtubule organization and dynamics are essential for E-lam
formation. Staurosporine induced recruitment of filamentous actin (F-actin), cortactin, filamin, Arp2/3 complex, Rac1 GTPase and
phospholipase C-g1 (PLC-g1) to lamellipodia. Treatment of cells with the GSK-3 inhibitors SB-415286 and LiCl2 inhibited E-lam
formation and prevented the accumulation of Rac1 and Arp2/3 complex at lamellipodia. The formation of E-lams was dependent on
fibronectin-binding integrins and normally regulated Rac1, and expression of either dominant-negative or constitutively active forms of
Rac1 prevented E-lam formation. Overexpression of either RhoA or Cdc42 GTPases suppressed E-lam formation. We conclude that
extended lamellipodia formation in keratinocytes requires actin and tubulin assembly at the leading edge, and this process is regulated by
was able to induce E-lams to the same extent in spreading
cells and in cells that had already spread [4]. This charac-
teristic allowed us to investigate signaling mechanisms of E-
lam formation that were independent of cell spreading.
Typical staurosporine-induced keratinocyte morphology is
shown in Fig. 1B.
Both actin and tubulin organization are critical for E-lam
formation
As changes in cell shape require rearrangement of the
cytoskeleton, we examined the distribution of cytoskeletal
elements in E-lams. Microtubules, F-actin and keratin inter-
mediate filaments were immunolocalized in cells spread on
fibronectin. Microtubules were detected by immunostaining
with anti-h-tubulin antibody. In untreated control keratino-
cytes, an extensive microtubule network was found through-
out the cell including the lamellipodia (Fig. 2A). In cells
treated with 50 nM staurosporine, long microtubule bundles
extended through the stem of the E-lam and separated in the
lamellar tip to cover the entire lamellipodium (Fig. 2B).
Fig. 1. Staurosporine-induced E-lam formation in HaCaT keratinocytes. HaCaT cells were allowed to spread on fibronectin for 2 h and were then treated with
50 nM staurosporine for 1 h (B) or left untreated (A). The cells were fixed, stained with crystal violet and photographed. Stem regions of extended lamellipodia
are marked with arrows.
L. Koivisto et al. / Experimental Cell Research 293 (2004) 68–80 71
Acetylation of a-tubulin is linked to microtubule stabiliza-
tion [34]. To investigate whether staurosporine could induce
alterations in microtubule stability, HaCaT cells were immu-
nostained with an antibody recognizing acetylated a-tubu-
lin. This antibody strongly recognized the microtubules that
were present in the extended part of the E-lam while those at
the lamellar tips were less pronounced (Fig. 2D). This
staining pattern suggests that the subset of microtubules
that is present in the stem of an E-lam may be more stable
than those that are present in lamella.
Actin filaments were not well organized and were prac-
tically undetectable in untreated control cells, as keratino-
cytes do not typically form prominent actin stress fibers (Fig.
2E) [5]. Treatment of keratinocytes with staurosporine in-
duced accumulation of F-actin in lamellipodia, and actin
became clearly visible in the stems of E-lams (Fig. 2F).
Cytokeratin 14 is a major component of keratin intermediate
filaments in basal epithelial keratinocytes and is also
expressed by keratinocytes in culture [35,36]. Extensive
networks of keratin filaments containing cytokeratin 14 were
present in both untreated and staurosporine-treated cells
except for the outer lamellipodia (Figs. 2G and H). Keratin
filaments were, however, concentrated in the stem of the E-
lam (Fig. 2H). Thus E-lam extension seemed to involve the
assembly of all major cytoskeletal components—actin fila-
ments, stable microtubules and intermediate filaments.
To test the importance of actin filaments and micro-
tubules for keratinocyte morphology and E-lam formation,
we treated the cells with a microtubule-destabilizing agent,
colchicine, and an F-actin-disrupting agent, cytochalasin D,
on fibronectin in the presence and absence of staurosporine.
Untreated and staurosporine-treated cells were differentially
sensitive to these agents. Control cells rounded up after
treatment with colchicine, whereas they were relatively
resistant to disruption of actin filaments and remained
spread (Fig. 3A). In contrast, cells treated with 50 nM
staurosporine were sensitive to cytochalasin D treatment
but resistant to the treatment with colchicine (Fig. 3A). The
results suggest that the spread phenotype of untreated
keratinocytes on fibronectin is dependent mainly on their
microtubule network, while staurosporine-treated were more
dependent on actin filaments. Both colchicine and cytocha-
lasin D totally prevented E-lam formation (Fig. 3B), indi-
cating that both actin and microtubule organization were
needed for E-lam formation.
Since E-lam formation depends on the activity of GSK-3
[4], we tested whether the inhibition of GSK-3 with a
specific inhibitor, SB-415286 [37], affects the staurospor-
ine-induced change in the cytoskeleton. SB-415286 alone
did not affect cell spreading on fibronectin, but strongly
inhibited colchicine-induced cell contraction to a rounded
morphology in control keratinocytes (Fig. 3C). SB-415286
also blocked staurosporine-induced sensitization of kerati-
nocytes to cytochalasin D (Fig. 3C). These observations
suggest that GSK-3 participates in the regulation of cyto-
skeleton in keratinocytes.
Colchicine (1 AM) totally breaks down the assembled
microtubule networks in cells impeding the interpretation of
the importance of microtubule assembly for E-lam forma-
tion [8]. Therefore, we studied the effects of low concen-
trations of taxol and nocodazole, which dampen microtubule
dynamics without disrupting the existing microtubule net-
works on E-lam formation [8]. In agreement with not
breaking down the existing microtubules, neither agent
affected the spread morphology of untreated control cells
(Figs. 4A and B). Both agents, however, caused a dose-
dependent, statistically significant inhibition in staurospor-
ine-induced E-lam formation (Figs. 4A and B), supporting
the importance of dynamic microtubule assembly for E-lam
hair-like cellular extensions consistent with Cdc42-induced
filopodia (Fig. 5C, panels c, g and h). These results suggest
that the activities of these two GTPases might be mutually
exclusive in keratinocytes. The expression level of Rac1 did
wed to spread on fibronectin for 120 min and then treated with 50 nM
l components, the cells with (B, D, F, H) or without staurosporine treatment
, D), actin (E, F) and cytokeratin 14 (G, H). The accumulated F-actin in
are marked with arrowheads and arrows, respectively. To help visualize the
of lamellar edges are shown in the insets. Scale bar = 10 Am.
Fig. 5. The role of GTPases in E-lam formation. Cells were transiently transfected with plasmid constructs expressing wild type (WT), constitutively active
(QL) and dominant-negative (N) forms of Rac1, RhoA and Cdc42 GTPases. The cells were allowed to spread on fibronectin in the presence or absence of
staurosporine, immunostained with Cy3-conjugated anti-VSV antibody to identify the transfected cells, and analyzed for cell spreading (A), E-lam formation
(B) and cellular morphology (C). The percentage of VSV-positive cells that were spread and had formed E-lams was determined and compared to the
percentage for the negative cells on the same coverslip. Mean F SD of the ratios of these percentages is presented (n = 16). No difference between VSV-
positive and -negative cells would be calculated as 1. (C) The typical morphology of staurosporine-treated transfected cells expressing wild type (a, d, g),
constitutively active (b, e, h) and dominant-negative (c, f, i) forms of Rac1 (a–c), RhoA (d– f) and Cdc42 (g– i) GTPases on fibronectin.
L. Koivisto et al. / Experimental Cell Research 293 (2004) 68–8074
not seem to be critical for the process of cell spreading and
E-lam formation. Overexpression of RacWT, however, in-
duced branched E-lams (Fig. 5C, panel a). The expression
levels of RhoA and Cdc42 appeared to be more critical,
since overexpression of these two exogenous wild-type
GTPases was detrimental to cell spreading and E-lam
formation. We conclude that Rac1 activity that is under
normal cellular control is needed for keratinocyte cell
spreading and E-lam formation, whereas activities of RhoA
and Cdc42 are suppressive.
Staurosporine can also inhibit Rho-associated kinase
(Rho-kinase), a downstream effector of RhoA [38,39].
When the Rho-kinase inhibitor Y-27632 was added to the
cells together with staurosporine, the numbers of E-lams
were strongly enhanced (data not shown). These results
support the notion that RhoA activity is suppressive for E-
lam formation.
Staurosporine-induced membrane ruffling and the
accumulation of Rac1 to the cell periphery can be
prevented by inhibition of GSK-3
Next, we examined the localization of several molecules
that are involved in or regulate F-actin assembly, namely,
cortactin, filamin, Arp2/3 complex and Rac1 GTPase [40–
42], in E-lams, and how the inhibition of GSK-3 affects the
cellular localization of these molecules. Additionally, we
analyzed whether the localization of tyrosine-phosphorylat-
ed and presumably activated phospholipase C-g1 (p-PLC-
g1), which we have previously shown to colocalize with the
L. Koivisto et al. / Experimental Cell Research 293 (2004) 68–80 75
sites of F-actin in E-lams [4], is affected by the inhibition of
GSK-3. F-actin, cortactin, Arp2/3 complex, Rac1 and p-
PLC-g1 did not localize to the cell periphery in control cells
(Figs. 6A, 6E, 6M and 7A, 7E), but some positive staining
for filamin was detected in lamellipodia of control cells (Fig.
6I). Staurosporine induced the accumulation of F-actin,
cortactin, filamin, Rac1 and p-PLC-g1 to the lamellipodial
ruffles (Figs. 6B, 6F, 6J and 7B, 7F). Arp3 staining was
punctate, but it was also enriched in peripheral regions of E-
lams (Fig. 6N). Treatment of HaCaT cells with the GSK-3
Fig. 6. The effect of GSK-3 inhibition on actin assembly. To determine the effect
cells were allowed to spread on fibronectin for 120 min and then treated with
combination of both agents (D, H, I, P), or left untreated (A, E, I, M) for 60 min. Th
G, H), filamin (I, J, K, L) or Arp3 (M, N, O, P). Enlarged details of lamellar edg
inhibitor SB-415286 did not alter the localization of any of
these molecules compared to control cells (Figs. 6C, 6G,
6K, 6O and 7C, 7G). In the presence of SB-415286, E-lam
formation by staurosporine was blocked. Interestingly, how-
ever, the cells expressed hair-like filopodia that contained
actin, cortactin, filamin and p-PLC-g1 (Figs. 6D, 6H, 6L
and 7H). Significantly and in agreement with the importance
of Rac1 in E-lam formation, SB-415286 prevented staur-
osporine-induced accumulation of Rac1 and Arp2/3 com-
plex into lamellipodia (Figs. 7D and 6P, respectively). Thus,
of GSK-3 inhibition on staurosporine-induced actin filament assembly, the
50 nM staurosporine (B, F, J, N), 30 AM SB-425286 (C, G, K, O) or a
e cells were fixed and immunostained for actin (A, B, C, D), cortactin (E, F,
es are shown in the insets. Scale bar = 10 Am.
Fig. 7. The effect of GSK-3 inhibition on localization of Rac1 and p-PLC-g1. HaCaT cells were allowed to spread on fibronectin for 120 min and then treated
with 50 nM staurosporine (B, F), 30 AM SB-425286 (C, G) or a combination of both agents (D, H), or left untreated (A, E) for 60 min. The cells were fixed and
immunostained for Rac1 (A, B, C, D) or tyrosine-phosphorylated PLC-g1 (E, F, G, H). Enlarged details of lamellar edges are shown in the insets. Scale bar =
10 Am.
Fig. 8. The effect of inhibitory anti-integrin antibodies on E-lam formation.
HaCaT cells were allowed to spread on fibronectin and then treated with
inhibitory antibodies against a5, h1, av and h6 integrins in the presence of
50 nM staurosporine. The percentage of spread cells (A) or cells forming E-
lams (B) of total cell number (mean F SD, n = 8) was calculated. The
statistical significance of the inhibition in cell spreading and E-lam
formation by anti-integrin antibodies compared to cells treated with
staurosporine only was also calculated (*P < 0.05; **P < 0.01).
L. Koivisto et al. / Experimental Cell Research 293 (2004) 68–8076
inhibition of GSK-3 specifically prevented Rac1 localiza-
tion to lamellipodia and its action in F-actin assembly that is
required for E-lam formation, yet still allowed the formation
of filopodia, which are presumably Cdc42-mediated. These
data collectively suggest that GSK-3 could function up-
stream of Rac1 in E-lam formation. To confirm this novel
finding, Rac1 was also prevented from localizing to lamel-
lipodia by LiCl2 (data not shown), another GSK-3 inhibitor
that blocks E-lam formation [4,43].
E-lam formation is integrin-mediated
Cell–matrix interactions were found to contribute to the
staurosporine-induced morphology, since staurosporine-in-
duced cellular extensions on collagen types I and IV were
shorter, branched and surrounded the whole cell perimeter
instead of being long, straight and directional on fibronectin
(data not shown). We have previously shown that human
keratinocytes express a5h1, avh6 and avh1 fibronectin
receptor integrins in culture and that function-blocking
antibodies against these integrins differentially inhibit cell
spreading on fibronectin [26]. To investigate whether staur-
osporine-stimulated and GSK-3-mediated E-lam formation
on fibronectin was mediated by a specific fibronectin-bind-
ing integrin, spreading assays were conducted in presence
of blocking antibodies against these integrins. To differen-
tiate the inhibitory effect of these antibodies on E-lam
formation from their effect on cell spreading, the antibodies
were added together with staurosporine to cells that had
L. Koivisto et al. / Experimental Cell Research 293 (2004) 68–80 77
already spread. In this experimental model, staurosporine
caused a slight but statistically significant increase in the
proportion of spread cells compared to untreated control cells
(Fig. 8A). Anti-a5 integrin antibody did not affect cell
spreading, whereas the addition of antibodies against av
and h6 integrins restrained cell spreading to control levels
(Fig. 8A). The antibody against h1 integrin subunit reduced
cell spreading by 35% (n = 8, P < 0.01) compared to cells
treated with staurosporine only (Fig. 8A). The combination
of anti-av and h1 integrin antibodies that is capable of
blocking all fibronectin-binding integrins in keratinocytes
caused a loss of cell spreading and rounding up of the cells
(Fig. 8A). The combination of anti-a5 and h6 integrin
antibodies aimed at revealing the remaining adhesive func-
tion due to the avh1 integrin [26] produced an effect similar
to anti-h1 integrin antibody (Fig. 8A). None of the inhibitoryantibodies alone was able to inhibit extended lamellipodia
formation in a statistically significant manner, whereas the
combination of anti-av and h1 integrin antibodies complete-
ly blocked E-lam formation (Fig. 8B). These findings
suggest that while E-lam formation obviously is integrin-
mediated, it is not mediated by any single specific receptor,
but that all of the fibronectin-binding integrins can inter-
changeably contribute to this phenomenon.
Discussion
We have shown previously that the activity of GSK-3 is
required for the formation of long lamellipodia and for cell
migration in keratinocytes [4]. In the present study, we
demonstrated a novel action for GSK-3 in the regulation
of the cytoskeleton and Rac1 translocation during E-lam
formation in cultured human keratinocytes.
HaCaT cells immunostained with anti-h-tubulin anti-
body showed long microtubule bundles extending through
the E-lam. The elongated part of the E-lam also consisted
of actin and keratin intermediate filaments. The presence
of keratin in the stem of the E-lam together with acety-
lated a-tubulin suggests that this extension may be a
stabilized structure. Staurosporine also induced accumula-
tion of prominent actin filaments to the tips of protruding
lamella. Indeed, E-lam formation required both microtu-
bule and F-actin assembly. E-lam formation was complete-
ly blocked by either colchicine or cytochalasin D, agents
that disrupt microtubules and actin, respectively. In addi-
tion, E-lam formation was inhibited by taxol and nocoda-
zole that prevent dynamic microtubule assembly without
breaking down existing microtubules [8]. Previous studies
have suggested that a factor, possibly Rac1, associated
with microtubules can be released and affects the assem-
bly of F-actin in lamellipodia formation [8]. Rac1 induces
cortical actin polymerization that results in membrane
ruffling and lamellipodia formation [44]. Notably, Rac1
showed a similar distribution as F-actin in staurosporine-
induced E-lams.
The formation of E-lams required Rac1 that was under
normal cellular regulation, that is, Rac1 apparently needs to
be activated and deactivated in a spatiotemporally controlled
fashion to allow the cell to acquire a polarized morphology.
Expression of either the dominant-negative or constitutively
active form of Rac1 caused loss of E-lams. In contrast,
overexpression of either Cdc42 or RhoA prevented E-lam
formation, but cells formed E-lams when they expressed
dominant-negative forms of these GTPases.
In its activated form, Rac1 binds tubulin directly in vitro
and colocalizes with microtubules in vivo, whereas tubulin
does not bind Cdc42 or RhoA [45]. In fibroblasts, Rac1
signaling depends on microtubule assembly, and polymer-
ization of microtubules is required for Rac1 activation [9].
On the other hand, Rac1 and/or Cdc42 may be able to
stabilize microtubules locally [45,46]. Activated Rac1 may
then be localized to the cell membrane where its activity
leads to de novo assembly of actin filaments and also
promotes microtubule growth into leading-edge protrusions
[9,47]. The cellular translocation of Rac1 from cytosol to
cytoskeleton and to cell membrane is likely a key event in
Rac1 function, and tubulin binding may permit translocation
along the microtubule network. Disruption of microtubules
by colchicine prevents Rac1-mediated membrane ruffling
[45]. In the present study, we showed that inhibition of
GSK-3, which specifically mediates E-lam formation in
keratinocytes [4], prevented Rac1 translocation to the cell
membrane, accumulation of Arp2/3 complex and F-actin
assembly. GSK-3 inhibition did not affect the staurosporine-
induced accumulation of various other molecules present in
E-lams, namely, cortactin, filamin and tyrosine-phosphory-
lated PLC-g1. GSK-3 regulates kinesin-based motility and
membrane-bound organelle release along stable microtu-
bules [48–50], providing one potential pathway by which
GSK-3 could regulate Rac1 translocation in keratinocytes.
GSK-3 binds to microtubules in vitro and in vivo and it
aligns with microtubules both in neuronal cells and in
fibroblasts [50,51]. We have previously shown that in
staurosporine-induced E-lams, GSK-3 is localized to the
stem of the E-lam [4]. As shown in the present report, the
stem of the E-lam is an area of extensive microtubule
elongation, bundling and acetylation indicating increased
microtubule stabilization. Stable microtubules may be im-
portant mediators of differentiative events since they are
typically found aligned with developing asymmetries in