Initiation and transduction of stretch-induced RhoA and Rac1 activation through caveolae: cytoskeletal regulation of ERK translocation.

The Rho family small GTPases play a crucial role in mediating cellular responses to stretch. However, it remains unclear how force is transduced to Rho signaling pathways. We investigated the effect of stretch on the activation and caveolar localization of RhoA and Rac1 in neonatal rat cardiomyocytes. In unstretched cardiomyocytes, RhoA and Rac1 were detected in both caveolar and non-caveolar fractions as assessed using detergent-free floatation analysis. Stretching myocytes for 4 min activated RhoA and Rac1. By 15 min of stretch, RhoA and Rac1 had dissociated from caveolae, and there was decreased coprecipitation of RhoA and Rac1 with caveolin-3. To determine whether compartmentation of RhoA and Rac1 within caveolae was necessary for stretch signaling, we disrupted caveolae with methyl beta-cyclodextrin (MbetaCD). Treatment with 5 mm MbetaCD for 1 h dissociated both RhoA and Rac1 from caveolae. Under this condition, stretch failed to activate RhoA or Rac1. Stretch-induced actin cytoskeletal organization was concomitantly impaired. Interestingly the ability of stretch to activate extracellular signal-regulated kinase (ERK) was unaffected by MbetaCD treatment, but ERK translocation to the nucleus was impaired. Stretch-induced hypertrophy was also inhibited. Actin cytoskeletal disruption with cytochalasin-D also prevented stretch from increasing nuclear ERK, whereas actin polymerization with jasplakinolide restored nuclear translocation of activated ERK in the presence of MbetaCD. We suggest that activation of RhoA or Rac1, localized in a caveolar compartment, is essential for sensing externally applied force and transducing this signal to the actin cytoskeleton and ERK translocation.


Introduction
Mechanical stress is recognized as an important extracellular stimulus facilitating cell growth, one example of which is cardiomyocyte hypertrophy. There is a growing list of signaling pathways that are activated by mechanical stress in cardiomyocytes including protein kinase C (PKC), mitogen activated protein kinases (MAPKs), phosphatidylinositol-3 kinase/Akt, Janus kinase/signal transducer and activator of transcription, calcineurin, and nitric oxide synthase (NOS) (1)(2)(3). There is also evidence that Rho family small GTPases play a central role in mediating the development of stretch-induced cardiac hypertrophy (4,5). Rho family small GTPases regulate a variety of cellular functions such as cytoskeletal rearrangement, cell contractility, and cell migration (6). The cytoskeleton communicates with the extracellular matrix through integrins which have been considered as potential mechanosensors (1,7,8). There is also evidence that integrins transduce signals via Rho dependent pathways (9). However the molecular mechanisms by which mechanical force is transduced and the steps leading to recruitment of Rho family small GTPases in mechanotransduction in cardiomyocytes are poorly understood.
Caveolae are cholesterol-and sphingolipid-enriched plasma membrane invaginations, which are considered to be important platforms for signal transduction (10)(11)(12)(13). There is considerable support for the notion that signaling molecules converge in caveolae (14)(15)(16), and that these play an important role in propagating signals in response to agonist treatment in cardiomyocytes. However the possibility that mechanotransduction signals through caveolar localized molecular pathways in cardiomyocytes has not been previously addressed.
There are two functionally distinct features of caveolae that are critical to their role in signal transduction (17). First, caveolae are considered to function as reservoirs of signaling proteins, with caveolin serving as a negative regulator of these molecules in quiescent cells. Second, caveolae function as clustering sites for ligand-activated Gprotein coupled receptors such as the muscarinic cholinergic or sphingosine 1 phosphate receptor (18,19). In the first regard, several studies indicate that the caveolar scaffolding protein, caveolin-1, can directly interact with signaling molecules (e.g. G-protein subunits, Src family tyrosine kinases, endothelial NOS, PKC, or Ras) within caveolae to negatively regulate their functions (12,(20)(21)(22)(23). Accordingly some of these molecules have been shown to be activated and liberated from caveolae upon hormonal stimulation (24)(25)(26).
In light of the findings cited above we hypothesized that mechanical force might be transduced into cellular signaling pathways in cardiomyocytes via caveolae. In particular, we focused on the localization and activity of Rho family GTPases, RhoA and Rac1, in caveolar microdomains and examined the effect of stretch on their compartmentation and downstream signaling. The data presented here demonstrate that activation of RhoA and Rac1 by stretch requires their proper localization within caveolae and that stretch leads to their dissociation from this compartment. Disruption of caveolae not only prevents RhoA and Rac1 activation but also impairs stretch-induced actin cytoskeletal rearrangement and nuclear localization of ERK.

Purification of caveolin-rich membrane fractions
Caveolin-rich microdomains were isolated using the methods of Song et al. with some modifications (20). Briefly, myocytes were washed with ice-cold phosphate buffer saline (PBS) and lysed directly with 600 µl of 0.5M Na 2 CO 3 (pH 11). Cells were passed 20 times through a 25-gauge needle and sonicated three 20 sec bursts with minimal output of Micro Ultrasonic Cell Disrupter (Kontes). A 0.54 ml aliquot of cell lysates was adjusted to 45% sucrose by the addition of 0.54 ml of 90% sucrose prepared in MBS [25 mM 4-morpholineethanesulfonic acid (MES), pH 6.5, 75 mM NaCl] and placed at the bottom of an ultracentrifuge tube. The 45% sucrose-cell lysate mixture was overlaid with 1.08 ml of 35% sucrose and 1.08 ml of 5% sucrose in MBS containing 0.2 M Na 2 CO 3 .
The gradients were centrifuged at 48,000 rpm for 20 hours in an SW50.1 rotor (Beckman Instruments). Twelve 0.27-ml fractions were collected from the top of the gradient. The proteins in each fraction were precipitated with 20% trichloroacetic acid.
The precipitated proteins were washed once with 95% ethanol and then resolved with Laemmli sample buffer. The proteins were boiled, separated by 12% SDSpolyacrylamide gel (PAGE), transferred to polyvinylidene difluoride (PVDF) membranes (Millipore), and subjected to immunoblotting. Enhanced chemiluminescence was then performed using SuperSignal chemiluminescent detection system (Pierce). with 1 µg of anti-RhoA or anti-Rac1 polyclonal antibody in the presence of Protein G-Sepharose (10 µl) overnight at 4°C. The precipitated proteins were separated by 12% SDS-PAGE, transferred to PVDF membrane, and probed with anti-RhoA, anti-Rac1, or anti-caveolin-3 antibodies.

Immunocytochemistry
Myocytes on stretchers were fixed for 20 minutes in 3.7% paraformaldehyde in PBS, permeabilized for 3 minutes in 0.4% Triton X-100, and blocked in 3% bovine serum albumin for 20 minutes The silicon membrane was cut and then stained with primary antibodies overnight at 4°C followed by secondary antibody and rhodamine-phalloidin for 1 hour at room temperature. The membrane was mounted on a coverslip and images were captured with a DeltaVision deconvolution microscope system (Applied Precision, Inc., Issaquah, WA.) The system includes a Photometrics CCD mounted on a Nikon TE-200 inverted epi-fluorescence microscope. In general between 30 and 80 optical sections spaced by~ 0.1-0.3 um were taken. Exposure times were set such that the camera response was in the linear range for each fluorophore. The data sets were deconvolved and analyzed using SoftWorx software (Applied Precision, Inc) on an Silicon Graphics Octane workstation. When applicable, image quantitation was performed using the Data Inspector program in SoftWorx.

Characterization of isolated caveolae from cardiomyocytes
Caveolin-3 is a heart and muscle cell predominant isoform of caveolin. We isolated cardiac caveolae from neonatal rat ventricular myocytes (Fig. 1A) using caveolin-3 as a specific marker for cardiac caveolae. Non-caveolar components were defined by the absence of caveolin-3 and by the presence of focal adhesion kinase (FAK) and of Rho-GDI, an endogenous inhibitor of RhoA and Rac1 that sequesters them in the cytoplasm.
The caveolin-3 positive bands were concentrated in fractions 5 and 6 in preparations made by the detergent-free purification procedure detailed in Methods. In contrast, FAK and Rho-GDI were detected in fractions 9 to 12 (Fig. 1A). The protein concentration profile obtained using this isolation method was also consistent with that of previous reports ( Fig. 1B) (29).

Activation of RhoA and Rac1 in response to stretch
We next determined whether stretch led to the activation of RhoA and Rac1 in cardiomyocytes. Pull-down assays revealed rapid and transient increases in activated (GTP-bound) RhoA and Rac1 in response to 20% equibiaxial stretch ( Fig. 2A). The activation of both RhoA and Rac1 peaked at 3 to 5 minutes after the onset of stretch and returned to the basal level by 10 minutes of stretch (Fig. 2B). RhoA and Rac1 activation at 4 minutes of stretch is also shown quantitatively in Figure 5.

Segregation of RhoA and Rac1 from caveolae in response to stretch
Changes in the subcellular localization of RhoA and Rac1 in neonatal rat ventricular myocytes subjected to stretch were assessed using the fractionation procedure described in Figure 1 to purify caveolae. RhoA and Rac1 were distributed in both caveolar and non-caveolar fractions in unstretched cardiomyocytes. Equibiaxial stretch for 4 minutes did not affect the distribution of RhoA or Rac1 in caveolar and noncaveolar fractions (data not shown). However, when equibiaxial stretch was applied to the myocytes for 15 minutes, redistribution of RhoA and Rac1 from the caveolar to a non-caveolar fraction was observed (Fig. 3A). Densitometric analysis of the distribution of caveolin-3, RhoA, and Rac1 in caveolar (fractions 5 and 6) versus non-caveolar (fractions 9 to 12) fractions revealed a 47% reduction in the proportion of RhoA and a 50% reduction in the fraction of Rac1 in caveolae. Concomitant increases in noncaveolar fractions were observed. Importantly, caveolin-3 distribution was not affected by stretch (Fig. 3B).

Effect of methyl β-cyclodextrin (MβCD) on stretch-induced RhoA and Rac1 distribution
The observation that activation of RhoA and Rac1 by stretch appeared to precede their dissociation from caveolae suggested that caveolar compartmentation might be necessary for their initial activation. To address this issue, we used cholesterol depletion with MβCD to disrupt caveolae. Cardiomyocytes were incubated with 5 mM MβCD for 1hour prior to stretch. The impact of this intervention on the activation and subcellular distribution of RhoA and Rac1 was then examined. Although the destruction of caveolae was not complete (as evidenced by retention of some caveolin-3 immunoreactivity in fractions 5 and 6), this treatment was sufficient to dissociate both RhoA and Rac1 from caveolae (Fig. 4A). The association of RhoA and Rac1 with caveolin-3 was also examined in immunoprecipitation studies. As expected, caveolin-3 co-precipitated with RhoA and with Rac1 in unstretched myocytes (Fig. 4B). Following fifteen minutes of stretch, the amount of caveolin-3 associated with RhoA and Rac1 was diminished consistent with their dissociation from the caveolar fractions. In addition, the amount of caveolin-3 that co-precipitated with RhoA and Rac1 was decreased by MβCD treatment, confirming dissociation of these small GTPases from caveolar microdomains ( Fig. 4B).

Effect of MβCD on stretch-induced RhoA and Rac1 activation
The ability of stretch to activate RhoA and Rac1 was then assessed in cells treated with MβCD. As shown in Figure 5, four minutes of stretch activated both RhoA and Rac1 in untreated cells. Pretreatment with MβCD did not significantly affect the basal activity of RhoA and Rac1, however stretch failed to activate either RhoA or Rac1. These observations suggest that the initiation of stretch-induced RhoA and Rac1 activation requires intact caveolar microdomains.

Effect of MβCD on ERK activity during stretch
Recent findings from studies of caveolin-3 and -1/3 null mouse demonstrated that chronic loss of caveolae resulted in hyperactivation of ERK in the heart (30,31). We, therefore, determined whether ERK was activated in myocytes when caveolar structure was acutely disrupted with MβCD. As shown in Figure 6, MβCD treatment alone did not lead to significant activation of ERK, nor was ERK activation by stretch impaired by MβCD treatment (Fig. 6).

Cytoplasmic retention of active ERK by MβCD in stretched myocytes
When we investigated the subcellular localization of active ERK by immunostaining with a phospho-specific ERK antibody, a different pattern emerged. As shown in Figure 7, in unstretched myocytes, weak phospho-ERK fluorescence (green) and disorganized striated actin filaments (red) were detected in the cytoplasm. One hour of stretch increased the intensity of phospho-ERK staining and nuclear staining was evident in ~90% of the cells. Stretch concomitantly elicited actin fiber alignment. In contrast, in cells treated with MβCD stretch increased phospho-ERK fluorescence but phospho-ERK staining was observed only in the cytoplasm and actin fibers remained disorganized. These data suggest that failure to activate RhoA and Rac1 and to organize the actin cytoskeleton, may prevent nuclear translocation of active ERK.

Jasplakinolide restores nuclear translocation of active ERK
To further examine the involvement of the actin cytoskeleton in nuclear translocation of ERK we tested the effect of, an actin polymerizer jasplakinolide, on the translocation of active ERK by stretch. Actin fibers could not be detected with phalloidin staining in these studies because jasplakinolide competes with phalloidin for binding to actin fibers. We, therefore stained for α-actinin, an actin-binding protein. Following jasplakinolide treatment, the actin cytoskeleton was organized and stretch led to the nuclear translocation of phospho-ERK in a significant fraction (~40%) of the MβCD treated cells (Fig. 7). Furthermore we determined that treatment of myocytes with cytochalasin-D, an actin depolymerizer, did not affect ERK activation but prevented stretch induced organization of the actin cytoskeleton and led to retention of active ERK in the cytoplasm of most cells (Fig. 8). This is consistent with a previous study with cytochalasin-D which suggested that nuclear translocation of active ERK is facilitated by an organized actin cytoskeleton (32).
The precise molecular relationship between RhoA and Rac1 activation, actin cytoskeletal reorganization, and ERK nuclear translocation remains to be explored.
However, ERK signaling in cardiomyocytes has been associated with the development of hypertrophy, characterized by increases in cell size, actin myofilament organization and ANF expression. Cardiomyocytes stretched continuously for 24 hours show clear increases in each of these parameters (Fig. 9). Concomitant treatment with MβCD prevents these changes. While long term MβCD treatment could disrupt myriad signaling pathways that regulate hypertrophy, the data are consistent with a requirement for caveolae in transduction of stretch to hypertrophic growth, possibly mediated via RhoA, Rac1 and ERK signaling pathways.

Discussion
The findings presented here demonstrate by direct biochemical analysis that stretch increases the amount of activated RhoA and Rac1 in cardiomyocytes. This observation extends published evidence demonstrating more indirectly that Rho is activated by stretch in vascular smooth muscle and cardiac muscle cells (4,33).
The precise mechanism by which mechanical force is transduced into activation of downstream signaling cascades remains unknown. The data presented here suggest that the compartmentation of RhoA and Rac1 in caveolae play a critical role in mechanotransduction in cardiomyocytes.
Using a detergent-free purification of caveolae we determined that a significant fraction of RhoA and Rac1 are localized in caveolar microdomains of cardiomyocytes.
Furthermore we demonstrate that mechanical stretch results in the dissociation of RhoA and Rac1 from the caveolar fraction and decreases the amount of caveolin-3 associated with RhoA or Rac1 in immunoprecipitates. We also provide evidence that short term MβCD treatment leads to complete loss of compartmentation of RhoA and Rac1 within caveolae and that this manipulation prevents stretch induced RhoA and Rac1 activation.  (34). Therefore, it might be hypothesized that stretch, by altering caveolar morphology, renders caveolar localized RhoA and Rac1 more accessible to guanine nucleotide exchange factors which catalyze the release of GDP and its replacement by GTP (8).
In addition to affecting RhoA and Rac1 activation, MβCD treatment prevents stretch induced actin cytoskeletal organization. The role of RhoA and Rac1 in activation of molecules responsible for actin cytoskeletal reorganization (e.g. Rho kinase, LIM kinase, p21-activated kinase 1 (PAK), cofilin) is well documented (35,36). It has also been demonstrated that filamin, a protein involved in actin cytoskeletal assembly through PAK, is a caveolin binding protein (37,38). The ability of caveolar disruption to prevent stretch mediated actin cytoskeletal organization could therefore result either from failure to activate RhoA and Rac1 or from impaired binding of cytoskeletal regulatory molecules to caveolin/caveolae. Unexpectedly, our data also suggest that stretch induced actin cytoskeletal rearrangement plays a role in insuring the proper localization of active ERK. Disruption of caveolae with MβCD did not prevent ERK activation by stretch. Indeed the finding that ERK is activated in the presence of MβCD suggests that some stretch-induced signaling pathways remained intact. Possibly integrin signaling pathways are engaged and lead to ERK activation in the absence of intact caveolae (see Fig. 10). Most notably, however, MβCD treatment led to cytoplasmic retention of activated ERK in response to stretch. The possibility that the impaired ERK localization is related to the failed cytoskeletal organization is consistent with another recent report demonstrating that cytochalasin-D, an actin cytoskeleton disrupting agent, impairs integrin stimulated nuclear translocation of active ERK in NIH 3T3 cells (32). Experiments presented here indicate that this also occurs in cardiomyocytes and in response to stretch. Stretchinduced ERK activation and nuclear translocation in mesangial cells has also been reported to be disrupted via destabilization of the actin cytoskeleton resulting from elevating NO or treatment with 8-bromo-cGMP (39). Jasplakinolide reverses the destabilization and allows proper ERK localization in mesangial cells (39). Similarly, in cardiomyocytes, jasplakinolide restores nuclear translocation of active ERK induced by stretch, even in the presence of MβCD.
The mechanism by which an intact actin cytoskeleton participates in the transport of active ERK into the nucleus remains unknown. Nonetheless our data suggest that organization of the actin cytoskeleton is required for mechanotransduction in cardiomyocytes. Mechanotransduction and G-protein coupled agonists are known to utilize ERK signaling pathways to induce cardiomyocyte hypertrophy (1,40,41). We demonstrate here that MβCD treatment prevents stretch induced hypertrophic increases in ANF gene expression and myocyte cell size. While there are many potential molecular events inhibitable by MβCD our findings suggest that the ability of MβCD treatment to block RhoA or Rac1 activation, cytoskeletal organization and ERK translocation affects nuclear responses.
In summary, our findings demonstrate that stretch activates and subsequently dissociates RhoA and Rac1 from a caveolar compartment in rat cardiomyocytes.
Mechanotransduction via this pathway appears to require organization of the actin cytoskeleton, a response that contributes to the nuclear localization of ERK. We        Localization of activated ERK was examined using phospho-specific ERK antibody.