(Received for publication, December 11, 1996, and in revised form, January 28, 1997)
From the Third Department of Internal Medicine, University of Tokyo, Hongo, Bunkyo-ku, Tokyo 113, Japan and the § Third Department of Internal Medicine, Yamaguchi University School of Medicine, Kogushi, Ube, Yamaguchi 755, Japan
The Rho family GTP-binding proteins have been
known to mediate extracellular signals to the actin cytoskeleton.
Although several Rho interacting proteins have been found, downstream
signals have yet to be determined. Many actin-binding proteins are
known to be regulated by phosphatidylinositol 4,5-bisphosphate in
vitro. Rho has been shown to enhance the activity of
phosphatidylinositol-4-phosphate 5-kinase (PI4P5K), the
phosphatidylinositol 4,5-bisphosphate synthesizing enzyme. Recently we
isolated several isoforms of type I PI4P5K. Here we report that PI4P5K
I induces massive actin polymerization resembling "pine needles"
in COS-7 cells in vivo. When truncated from the C terminus
to amino acid 308 of PI4P5K I
, both kinase activity and actin
polymerizing activity were lost. Although the dominant negative form of
Rho, RhoN19, alone decreased actin fibers, those induced by PI4P5K were
not affected by the coexpression of RhoN19. These results suggest that
PI4P5K is located downstream from Rho and mediates signals for actin
polymerization through its phosphatidylinositol-4-phosphate 5-kinase
activity.
The stimulation of cells by growth factors and other agonists is mediated by intracellular signal transduction pathways. In response to environmental signals, a cell changes both its shape and its degree of attachment to the substratum. These changes are caused, at least in part, by the polymerization and rearrangement of the actin into sheet-like structures known as lamellipodia and bundle-like stress fibers (1, 5). The Rho family GTP-binding proteins (Cdc42, Rac, and Rho) play key roles in transmitting signals to the cytoskeleton (1, 6, 7). They share a similar mode of action in that they cycle between the GTP-bound active and GDP-bound inactive form. Whereas constitutively active Rac induces lamellipodia, Rho induces stress fibers (5). When the dominant negative form of each protein is introduced, specific morphological change is blocked (1, 5). Downstream elements of the Rho signaling pathway to the cytoskeleton are not well defined, although several candidates have been reported including Rho GTPase-activating protein (1), protein kinase N (8, 9), Rho kinase (10, 11), and PtdIns4P1 5-kinase (7). Some of these interact with the GTP-bound active form of Rho; therefore they are good candidates for Rho effectors. However, because downstream signaling from Rho is not only involved in cytoskeletal organization but takes diverse pathways including gene transcription, it is not clear which molecule(s) is responsible for cytoskeletal changes.
Actin polymerization is dependent both on the availability of free barbed ends and that of monomeric G-actin, which are regulated by multiple actin-binding proteins (ABPs). Capping proteins bind to barbed ends to mask them, and monomer-binding proteins bind G-actins to sequester them. Several ABPs can bind PIP2 and release the free barbed end or free G-actin to polymerize (2). PIP2 is synthesized by PI4P5K, and Rho has been known to stimulate PI4P5K activity (3). Therefore PI4P5K is a suitable link between the Rho family and actin polymerization. Mammalian type 2 PI4P5K has been cloned (12). We recently cloned several isoforms of type I PI4P5K (4). Here we report its important role in actin polymerization in vivo.
The wild type PI4P5K I construct with
hemagglutinin (HA) epitope used was described previously (4).
C-terminal deletion mutants were constructed by appropriate
endonuclease digestion (BamHI and NcoI),
designated 5K456S and 5K308M, and subcloning into the pAdex1CA vector
(13). C-terminal sequences were confirmed by DNA sequencing. The amino
acids of constructs are: WT, 1-539; 5K456S, 1-456 with extra Asn; and
5K308M, 1-308 with extra KLIKLV due to the polylinker sequence of the
vector. Partial OCRL cDNA was isolated by polymerase chain reaction
using primers corresponding to nucleotides 806-828 (sense) and the
complement of nucleotides 1530-1550 (antisense) based on the published
sequence and tagged with HA (14). The RhoN19 cDNA (threonine at
codon 19 of Rho A was replaced with asparagine) was constructed by
polymerase chain reaction, tagged with the Myc epitope at their N
terminus and cloned into the pAdex1CA vector.
Recombinant adenovirus methods and these techniques were the same as described previously (4, 13). Antibodies used were 12CA5 (anti-HA), 9E10 (anti-Myc), and 4G10 (anti-phosphotyrosine).
ImmunofluorescenceCOS-7 cells were maintained in Dulbecco's modified Eagle's medium with 10% fetal bovine serum. Cells were plated on coverslips and infected with the virus on the next day. They were incubated about 18 h, fixed with 3% paraformaldehyde, and incubated with the appropriate antibody in phosphate-buffered saline containing 0.2% gelatin (PBS-G) at room temperature for 30 min. After washing three times with PBS-G, coverslips were incubated with FITC-conjugated anti-mouse IgG (DAKO) and rhodamine-conjugated phalloidin (Molecular Probe Inc.) at room temperature for 30 min. Slides were observed under a Bio-Rad confocal microscope system (MRC 1024).
C-terminal
truncations of PI4P5K I were made by endonuclease digestion using
unique site with BamHI (amino acids 1-456) or NcoI (amino acids 1-308) and designated 5K456S and 5K308M,
respectively (Fig. 1A). Wild type and
truncated constructs were tagged with the HA epitope and expressed in
COS-7 cells. Control cells were infected with the vector containing
lacZ insert. The size of overexpressed proteins revealed by
anti-HA antibody (12CA5) were: WT, 68 kDa (Fig. 1B,
lane 2); 5K456S, ~58 kDa (lane 3); and 5K308M,
~38 kDa (lane 4). Although WT and 5K456S retained PtdIns4P
5-kinase activity, 5K308M revealed little kinase activity (Fig.
1B). This was observed in several independent kinase assays
and was not due to low expression levels.
Because 12CA5 and anti-PI4P5K I antibody (4) yielded
identical patterns in overexpressed cells under
immunofluorescence, we used 12CA5 in subsequent investigations. Wild
type PI4P5K I
showed cytosolic as well as punctate distribution, and
the plasma membrane also appeared to be stained (Fig.
2A). This is consistent with the results
reported previously (15). Although 5K456S showed a pattern similar to
WT cells, 5K308M showed diffuse but little marginal staining, which is
not typical for plasma membrane localization (Fig. 2B).
Between amino acids 309 and 456, there is a sequence (YRXXXXXXXSWK) similar to the putative
PIP2-binding site of
-actinin (16). The detailed
analysis of the relationship between kinase activity and the plasma
membrane association will be published elsewhere.2
Effect of PI4P5K Overexpression on Actin Polymerization
If
PI4P5K I is an effector for actin polymerization, increased levels
of PI4P5K would be expected to result in increased actin
polymerization. In control cells, actin filaments were organized in
parallel bundles at the bottom of the cells, and peripheral actin
filaments formed lamellipodia (Fig. 3A).
However, the amount of F-actin in PI4P5K WT (and 5K456S; data not
shown) cells increased as shown in Fig. 3C. Typical stress
fibers and lamellipodia appeared to decrease, but relatively short
actin fibers were observed in random array even in aggregated fashion
like "pine needles." In kinase-negative 5K308M cells, actin fibers
did not show such a pattern but instead were arranged as typical stress
fibers as in control cells (Fig. 3E). With double labeling
of 12CA5 and rhodamine-phalloidin, we confirmed this result at single
cell level even in 5K308M high expressing cells. Therefore the lack of
the effect was not due to low level expression. We then overexpressed partial OCRL protein (amino acids 264-506), known to retain a PtdIns
5-phosphatase activity, in COS-7 cells (OCRL cells). This protein was
found in the cytoplasm unlike the Golgi pattern of full-length OCRL
(14). In contrast to the cells expressing PI4P5K, there were few actin
fibers in OCRL cells (Fig. 3F). There was little change in
the total amount of actin between control and PI4P5K and OCRL
overexpressed cells by Western blotting (data not shown). These results
strongly suggest that PI4P5K induces actin polymerization through its
kinase activity by elevating levels of PIP2. These results
were also observed in other cell lines (data not shown).
Rho has been known to regulate actin-based cytoskeleton (1, 7). The
dominant negative form of RhoA, RhoN19, was made by replacing the
threonine in a position analogous to codon 17 of Ras with asparagine.
When RhoN19 was expressed in COS-7 cells (Fig.
4B) stress fibers decreased (Fig.
4A). We then expressed RhoN19 and PI4P5K (WT) simultaneously
in COS-7 cells. Unlike RhoN19 cells, the actin fiber pattern was very
similar to that in PI4P5K WT cells (Fig. 4C). Thus PI4P5K is
probably located downstream from Rho, consistent with previous results
(3).
Effect on Focal Adhesion
When COS-7 cells were infected with kinase active constructs (WT and 5K456S), we noticed that the degree of adhesion to the substratum decreased. This phenotype correlated with the expression levels of PI4P5K as judged by immunofluorescence. This change did not occur with expression of the kinase-negative construct (5K308M). It has been known that introduction of active Rho into Swiss 3T3 cells induces both stress fibers and focal adhesion (FA). This indicates that these two phenomena are located downstream from Rho. We stained control and PI4P5K-overexpressed cells with anti-phosphotyrosine (Tyr(P)) antibody (4G10) and anti-vinculin antibody to localize FA. Control cells showed the typical FA pattern at the end of actin stress fibers as shown in Fig. 3B (arrowheads). In PI4P5K WT cells, the Tyr(P) positive plaques were smaller, although their number was increased (Fig. 3D). Not all of the ends of actin fibers were associated with FA (Fig. 3, C and D), indicating that the increased numbers of actin fibers were not organized into typical stress fibers and did not serve to anchor the extracellular matrix effectively. This may explain why PI4P5K cells show decreased adhesion activity.
In this paper we described the effects of kinase-active and
-negative forms of PI4P5K I on actin polymerization using
immunofluorescence. Although several proteins that interact with Rho
have been found recently (8-11), it is not clear which candidate is
responsible for actin polymerization. Rho kinase is a candidate because
it phosphorylates myosin phosphatase and induces a stress fiber-like pattern (17, 18). Although this indicates a pathway through myosin, it
has been strongly suggested that there exists a mechanism acting
directly on actin (19). Our results showed that PI4P5K plays an
important role in actin polymerization. It is possible that the
polymerization of actin fibers and their organization into stress
fibers are distinct steps and that PI4P5K is important in the former
process and Rho kinase is required in the latter process. The Rho
family GTPases have been known to activate PtdIns4P 5-kinase activity
(19, 20). Moreover, several recent reports have suggested that they
interact with PI4P5K directly. Rac physically interacts with PI4P5K
(20), and Rho binds to 68-kDa PI4P5K (21). It is possible that each
member of Rho protein interacts with a different isoform of PI4P5K.
Actin polymerization is dependent on multiple ABPs. Many ABPs have been known to be regulated by PIP2, including prophillin, thymosin (22), gelsolin, actinin (23), capping protein, and vinculin (24). For example, thymosin releases free G-actin when it binds to PIP2. Binding of vinculin to actin is regulated by PIP2 (24). Rac (and/or Rho) increase free actin barbed ends through PIP2 (19). Most of these results on ABPs in relation to PIP2 were obtained in vitro because it has been difficult to prove their effects in vivo. Our system is suitable for this purpose, and the results were striking.
PI4P5K probably acts to increase PIP2 levels. PIP2 is hydrolyzed by phospholipase C (PLC) to generate inositol (1,4,5)-P3 and diacylglycerol, which are well known second messengers. Our results might include these secondary effects. However, multivalent antigen or phorbol myristate acetate causes the polymerization of actin in basophilic leukemia cells. Although multivalent antigen activates PLC, phorbol myristate acetate does not. However, in both cases, there is good correlation between F-actin levels, PtdIns kinase activity, and the increased production of PIP2 (25). Therefore, activation of PI4P5K itself does not activate PLC. Our result showing that PI4P5K-induced actin polymerization was most likely due to elevated PIP2 levels. Because OCRL overexpression decreases PIP2 levels (14), the results in OCRL cells are consistent with our model in which signals to actin are mediated by PI4P5K through elevated PIP2 levels.
FA is a multimolecular complex, and its formation and maturation are regulated by many factors including Rho, tyrosine kinases, and Src homology 2-containing proteins (26). Compared with the drastic change in actin polymerization, the effect of PI4P5K on FA was not prominent in our study. These results suggest that although integrin regulates the rate of synthesis of PIP2, FA formation needs a signal(s) other than PIP2 elevation, such as recruitment of FA proteins and sequential phosphorylation cascade.
Recent characterization of STM7 revealed that the Friedreich's ataxia
gene is human PI4P5K I (27). This raises the interesting possibility
that the disease is caused by a deficiency in the PtdIns pathway.
Because, in addition to neurological symptoms, hypertrophic
cardiomyopathy is observed in this disease, our results might be
relevant to the pathogenesis. However, PIP2 itself has many
biological activities other than actin polymerization. PIP2 binds many pleckstrin homology domain-containing proteins (28) and
activates phospholipase D activity. Its effects need to be characterized further in our system.