Mechanical stress is one of the most influential physical factors in biology and one of the least characterized. There are three main sources of energy that animal cells can use: chemical potential, electrical potential and mechanical potential. The latter is the least characterized. While it is obvious from molecular dynamics (Craig et al., 2004a; Kosztin et al., 2002; Marszalek et al., 1999; Ortiz et al., 2005) and force spectroscopy (Brown et al., 2007; Carter and Cross, 2006; Gosse and Croquette, 2002; Li and Fernandez, 2003; Lu et al., 2004; Sarkar et al., 2005; Walther et al., 2006; Wang et al., 1997) that mechanical forces deform molecules, the processes that govern the mechanics of cells are much more complicated (Boey et al., 1998; Discher et al., 1998; Discher et al., 1994). Cells are composed of many proteins that are dynamically linked and coupled to the extracellular matrix (Hocking et al., 2007; Valencik et al., 2006) by integrins and dystroglycans (Michele and Campbell, 2003) that pass through the liquid lipid bilayer (Lele et al., 2007; Matthews et al., 2007).
Mechanical forces alter the physiology and biochemistry of many well-known systems such as muscle contraction, blood pressure (Acher, 2002), hearing (Hudspeth et al., 2000) and touch (Gillespie and Walker, 2001). There are >40,000 publications addressing the influence of mechanical stress on cells and many more addressing the effect on mechanical stress on tissues, organs and organisms.
The most familiar mechanical systems are muscles and sensory nerves, but there are many other less obvious systems such as stress induced remodeling of the cytoskeleton in blood vessel endothelia(Chicurel et al., 1998; Lele et al., 2007; Parker and Ingber, 2007) and the growth of muscles with exercise. Even more subtle are the observations that substrate stress affects stem cell lineage (Engler et al., 2007; Pajerowski et al., 2007). How do mechanical forces affect cells?
One mechanism is the stress induced exposure of cryptic sites (Antia et al., 2006; Baneyx et al., 2002; Brown et al., 2007; Craig et al., 2004a; Craig et al., 2004b; Hocking et al., 2007; Johnson et al., 2007; Smith et al., 2007). Protein folds can conceal enzymatic targets and substrate binding sites (Linke and Fernandez, 2002; Minajeva et al., 2001; Oberhauser et al., 2001; Sarkar et al., 2005). The force that exposes one site can be different from another and a single molecule may have multiple sites. These sites can be activated by stresses of different magnitude and different modality such as stretching or twisting. The cryptic sites coupled to dynamical stresses in the cytoskeleton provide the panoply of potential responses.
The cell bilayer (Bloom et al., 1991; Evans and Yeung, 1994; Evans, 1985; Rawicz et al., 2000) is coupled to the cytoskeleton and the extracellular matrix and contains mechanosensitive ion channels (Bowman et al., 2007; Gottlieb et al., 2007; Hamill and Martinac, 2001; Martinac, 2004; Perozo et al., 2002b; Sachs and Morris, 1998; Suchyna and Sachs, 2007a) and other enzymes (Charras et al., 2006). The ubiquitous cationic MSC (or stretch activated channel, noted SAC) has proven refractory to cloning but recently two candidates have been cloned: Piezo 1 and 2(Coste et al., 2010). The mechanosensitive and K selective 2P channels such as TREK-1 have been cloned and expressed in high density (Chemin et al., 2007a; Chemin et al., 2007c; Honore et al., 2006). These channels are sensitive to amphipaths such as arachidonic acid (Chemin et al., 2007b), and remarkably, they are activated by general anesthetics at clinical doses (Patel et al., 1999). The resulting hyperpolarization of neurons may account for much of general anesthesia. While the idea that MSCs account for anesthesia appears irreverent to many neuroscientists, it contains an intrinsic explanation for the efficacy of a physical knockout punch - transiently hyperpolarize all the neurons.
The 2P channels are expressed in the heart (Kelly et al., 2006; Tan et al., 2004), upregulated in hypertrophy (Tan et al., 2004; Zhao et al., 2007) and more highly expressed in the endocardium than the epicardium (Tan et al., 2004). The channels are the dominant MSC in rat atria (Niu and Sachs, 2003), and their activity has been linked to cardiac arrhythmias (Kohl et al., 2005). The 2P channels have a unique link to the cytoskeleton independent of their role as ion channels (Lauritzen et al., 2005). TREK transfection reorganizes the actin cytoskeleton, but that also occurs if TREK has been mutated to be nonconducting. The basis of this coupling and the potential effects on cytoskeletal stress and TREK kinetics are not yet known.
The best studied MSCs are the bacterial channels that have been functionally reconstituted in lipid membranes (Perozo et al., 2002a; Perozo et al., 2002b; Sukharev and Anishkin, 2004; Wiggins and Phillips, 2005). The channels are activated by lipid tension (Markin and Sachs, 2006; Markin and Sachs, 2004; Wiggins and Phillips, 2005) with a possible contribution of bending energy from hydrophobic mismatch and amphipaths (Andersen et al., 1999). Eukaryotic MSCs have not been reconstituted, so the mechanical environment of the channel is not yet known.
Stress distribution in a heterogeneous cell membrane.
Particularly for the eukaryotes, the term "membrane stress" is misleading. Cortical stresses are distributed between the lipids, cytoskeleton (with tangential and normal components), and extracellular matrix. When a piece of the cell cortex is drawn into a patch pipette, approximately half of the mean stress in the spanning membrane is borne by the lipids and the rest by the cytoskeleton and/or extracellular matrix (Akinlaja and Sachs, 1998).
Cortical stresses are dynamic and plastic, hence affected by the history of stimulation (Hamill and McBride, 1992; Suchyna and Sachs, 2004; Suchyna and Sachs, 2007a). The cortex is inhomogeneous and there are mechanically differentiated domains that can be readily observed using MSCs as probes. Groups of channels can show spontaneous and reversible shifts in sensitivity, so that these domains or corrals (Leitner et al., 2000) contain many channels. While the simplest preparation in which to study MSCs is the patch, patches are complex, being much more a sample of the cell cortex than a bilayer (Ruknudin et al., 1991). Patches contain cytoskeleton, organelles (Sokabe and Sachs, 1990; Sokabe et al., 1991), vesicles and tubules (Suchyna and Sachs, 2007a) and these components share the stresses and thus determine how much stress reaches a channel (Suchyna et al., 2009). However, patches are simpler than cells.
Are channels linked directly to the cytoskeleton? MSCs in the specialized receptors such as cochlear hair cells and touch receptors seem to be pulled on by fibrous proteins (Assad et al., 1991; Emtage et al., 2004; Ernstrom and Chalfie, 2002), but there is no evidence of such functional links in non-specialized cells. Without careful analysis, the dynamic effects of local stress in patches can be confused with intrinsic channel kinetics. The applied stimulus in patch experiments is hydrostatic pressure, but the channels do not respond to pressure, they respond to cortical tension. Pressure is converted to local tension through the constitutive mechanical properties of the patch.
A major perturbation for patch mechanics is the gigaseal itself. The adhesion energy density of the seal is equal to the mean patch tension (Opsahl and Webb, 1994; Priel et al., 2007; Suchyna and Sachs, 2007b). This tension activates MSCs and can even saturate 2P channels (Honore et al., 2006). The inevitable presence of resting tension in a patch means that it is not possible to assay the resting activity of MSCs in a patch. It is worth noting that all the voltage-sensitive channels that have been examined are also stretch-sensitive (Calabrese et al., 2002; Gu et al., 2001; Morris and Juranka, 2007).
To calibrate the sensitivity of an MSC requires estimating the local stress. We did that for MscL (Moe and Blount, 2005; Sukharev et al., 1999), and with less precision for eukaryotic MSCs (Gottlieb et al., 2007; Guharay and Sachs, 1984; Guharay and Sachs, 1985; Honore et al., 2006; Suchyna and Sachs, 2007a; Sukharev et al., 1999). However, the presence of mechanical domains in natural membranes (Bett and Sachs, 2000) complicates the measurement. The bilayer is liquid and cannot sustain any gradients in tension unless there are domains that can be created from aggregates or proteins or lipids.
The cortical cytoskeleton is heterogeneous containing many scaffolding (or organizing), crosslinking and anchoring proteins in addition to the more commonly known actin and tubulin fibers. Mutations causing functional disruption or elimination of one cortical component, such as occurs in many dystrophies (Franco-Obregon and Lansman, 2002; Suchyna and Sachs, 2007a), leads to a reorganization of the remaining components, altered cortex mechanical properties and disease. Stress in the cytoskeleton is dynamic, varying in time and space (Meng and Sachs, 2011a; Meng et al., 2011; Meng et al., 2008; Meng and Sachs, 2011b).
The stress within specific proteins in situ has been cleverly measured by chemically labeling cryptic sulfhydryl sites in the presence of mechanical stress (Johnson et al., 2007). This methodology, however, is not well suited to dynamic measurements nor to patch experiments where there is little material. Rief's lab proposed using GFP as a force sensor (Dietz and Rief, 2004; Mickler et al., 2007), but that has not yet proven operational. To remedy this situation we designed and built genetically encoded FRET based stress sensors that can be inserted into fibrous proteins in cells and animals (Meng and Sachs, 2008; Meng and Sachs, 2011a; Meng et al., 2011; Meng and Sachs, 2011b.)
Atoms to Animals.
If we knew the atomic structure of MSCs we would still not know why cardiac enlargement causes arrhythmias (Kohl et al., 2005). Knowledge of the properties of a molecule is insufficient to describe the properties of an animal. We need to move upward in complexity to utilize our knowledge of molecules. Understanding mechanical stress is essential to understanding animal physiology. The ability of cells to generate and distribute stresses over space and time (Deng et al., 2006) makes them adaptable to changes in their environment.
We can begin to build models of animals by examining the behavior of structural proteins in situ. In which situ? C. elegans is remarkably well characterized, the genetics are known, there are public libraries of mutants, it is simple to create transgenic animals, and they are cheap. If we can understand the stresses in a worm, we are closer to understanding the stresses in humans. The constitutive properties of the worm have been measured using macroscopic deformability (Park et al., 2007) and osmoregulation (Acher, 2002). A recent paper used piezoresistive cantilevers to indent the worms and derived some constitutive properties from the force-distance relationships (Park et al., 2007). They showed that, contrary to the view that worms are hydrostats, puncturing the body to relieve the pressure had little effect upon the stiffness. The stiffness arose from the cuticle itself rather than from surface tension. This result is reminiscent of results we obtained using the AFM to study the effect of hypotonic stress in cells (Spagnoli et al., 2008).
Hypotonic stress does cause cells to swell, but that did not make them stiffer, and some cells, such as astrocytes, became softer (Figure 1). Apparently much of the osmotic pressure is borne by the cytoskeleton in three dimensions, not just the membrane in two. Physically, cells behave like membrane coated sponges – they swell in water and become softer. As water enters a sponge it stretches the walls of the pores. The elastic stress raises the pressure of the water in the gel preventing further influx. Lysis in cells occurs only when the cortex dissociates from the deeper cytoskeleton and membrane blebs form with the stress distributed in two dimensions and then burst. These simple experiments reminded us of how ignorant we are about the distribution of mechanical stress in cells. It is enticing to imagine using a multiphoton microscope to examine the stress in specific structural proteins of the beating heart.
Since the patch clamp is how most data is gathered on mechanosensitive channels it is worth looking at the mechanics of patches. The most detailed work involved dystrophic mdx mouse muscle where the loss of the reinforcing dystrophin protein affects the mechanical and anatomical properties of the patch (Sachs, 2010; Suchyna and Sachs, 2007a). The increased sensitivity of MSCs in patches from dystrophic muscles is the result of changes in the mechanical properties of the membrane and potentially scaffolding protein interactions..
While patches from normal muscles were flat, stretched tight by adhesion to the glass, patches from dystrophic muscle were concave toward the tip of the pipette (Figure 2). The concavity is the result of actin (≈100 molecules/µm2) pulling normal to the membrane. This force from actin could be balanced with a small amount of suction calibrating the force. However, the most significant change in the mdx patch mechanical properties is a 3-fold increase in cortical viscosity, probably due to the breaking of noncovalent bonds. This leads to an extended period of elevated cortical tension, more time for the channels to open and also an increased probability of rupture. In mdx myotubes, the ability of the channels to inactivate with time is compromised for reasons that remain to be determined, but the net effect is that the channels stay open longer letting in more ions.
In tissue culture, developing mdx muscle fibers show spontaneous Ca2+ transients that rarely occur in normal muscle. The transients are reversibly inhibited by GsMTx4 demonstrating that MSCs can be active in a resting membrane, and notably for clinical purposes, in a dystrophic membrane. MSC deregulation has been shown to be a significant, if not the primary, factor responsible for the excess Ca+2 uptake associated with dystrophy. GsMTx4, like the less specific inhibitor streptomycin (Whitehead et al., 2006), also inhibited stretch-induced Ca2+ uptake in intact stressed mdx muscle fibers (Yeung et al., 2005). This result points toward a potential therapy for muscular dystrophy. The targets of GsMTx4 are endogenous cationic MSCs; these may be of the TRP or Piezo family (Coste et al., 2010).
The most detailed study of a cloned eukaryotic MSC was done on the K+ selective 2P channel TREK-1 (Honore et al., 2006). These channels express in high density often providing several nA from a cell-attached patch. Being K+ selective, they are easily manipulated and distinguished from background. The channels have pronounced inactivation and using a high speed pressure clamp (Besch et al., 2002) and QuB analysis software that we wrote (Feng et al., 1997) (www.qub.buffalo.edu), we solved the detailed kinetics analysis providing the relevant rates and the pressure sensitivity of each rate. Assuming that the channel is a 5nm diameter cylinder, the observed mechanical sensitivity required that the channel change its diameter by only 0.4nm upon opening. If the movable domains of the channel were smaller than the entire channel, they would have had to move further. The experiments on TREK emphasized the need to understand cortical mechanics if we are to understand MSC gating. The tension created by the gigaseal, ≈ 1mN/m, was strong enough to activate TREK in the patch without an applied stimulus (Opsahl and Webb, 1994; Suchyna and Sachs, 2007b).
A surprising additional complexity appeared at the end of each suction pulse. There was a transient drop in channel activity below baseline. This proved to be a result of patch wrinkling, not complicated intrinsic channel kinetics. Suction can stretch the patch into a spherical cap that has more area than the planar disk that is required to span the pipette (the normal resting geometry). When suction is released, the excess area causes the patch to wrinkle and the mean tension becomes zero causing the open channels to close. The adhesion energy of the seal then reanneals the excess area to the glass restoring the resting tension and restoring resting channel activity. The minimal channel activity during wrinkling shows that curvature at the scale of 0.1µm is not a significant driving force for gating (Markin and Sachs, 2004).
We measured the gigaseal adhesion energy density (Priel et al., 2007) by peeling the patch from the glass using pipette pressure. This energy density was 1-4mN/m while the lytic limit of bilayers is ≈10mN/m (Suchyna and Sachs, 2007b; Suchyna and Sachs, 2008; Suchyna and Sachs, 2007c). The adhesion energy is the magnitude expected from van der Waals attraction (Parsegian, 2006) and is similar to the energy of adhesion of gecko feet(Autumn et al., 2002). Our data showed an additional small repulsive Coulomb contribution that can be titrated with calcium. Our model of seal formation is the "fried egg zipper" where the extracellular domains of membrane proteins denature and stick to the glass, pulling the adjacent membrane close to the glass and causing nearby proteins to also denature. This process continues until the tension in the dome is equal to the adhesion energy.
Based upon the stretch-induced increase in muscle capacity, the possible involvement of caveolae in that change and the colocalization of many proteins to calveolae, we overexpressed caveolin-3 (with a GFP tag) in mouse muscle fibers. Overexpression of Cav-3 increased MSC activity but produced little change in the patch mechanics or capacitance. Since caveolae are enriched in cholesterol, we removed cholesterol with cyclodextrin to exaggerate the effect of Cav-3 knockdown. The effect was dramatic. Caveolin was internalized, MSC currents increased, the membrane became almost invisible and the mechanical relaxation time decreased to the limit of the pressure clamp (Figure 3). Cholesterol depletion seemed to dissociate the cytoskeleton from plasma membrane as evidenced by the uniform optical density across the patch and the increased speed of relaxation. MSC currents also increased suggesting that they receive stress from the bilayer. This result may be clinically relevant since depletion of cholesterol by the statins disrupts T-tubules and the plasmalemma in 100% of asymptomatic patients (Draeger et al., 2006).
The stretch-induced jump in specific patch capacitance shown in Figure 3 is what is expected for thinning and stretching a bilayer. Assuming a resting capacitance of 10fF/μm2 and a constant volume membrane, the increase in capacitance of ~3%, would correspond to a thickness change of ~1.5%. This agrees with measurements of voltage dependent bilayer compression (Alvarez and Latorre, 1978). Thickness changes associated with stretch may modify the hydrophobic mismatch thought to contribute to the energy flow in MSC gating (Lee, 2006; Wiggins and Phillips, 2005). To explore the stress distribution between the cytoskeleton and the rest of the "mechanical membrane", and to simplify the patch mechanics, we tried to erase the cytoskeleton with proteolytic enzymes applied to inside-out patches (Niggel and Sachs, 2007). To our surprise none of them had a significant effect on endogenous MSC activity – they neither digested the channels nor altered the channel kinetics as might be expected for a change in the distribution of stress. Apparently the relevant proteins are inaccessible to proteases.
The peptide was found in a blind screen of arachnid venoms and was found in only a single species of tarantula showing that it is not a common peptide (Suchyna et al., 2001). We solved the primary structure (Ostrow et al., 2003) and the tertiary structure (Oswald et al., 2002) and synthesized it and its D enantiomer to compare efficacy (Suchyna et al., 2004a). The D and L enantiomers were equally efficacious showing that they acted on the channel outside of a stereospecific binding site. This iconoclastic result seems to conflict with the fact that GsMTx4 is specific for cationic MSCs, but remarkably not the MSCs in differentiated mechanoreceptors such as those involved in touch and hearing (Bowman et al., 2007; Sachs et al., 2004). So when considering the use of GsMTx4 as a drug, expect no sensory side effects. If the MSCs can't tell a right handed drug from a left handed drug, how can the drug can be specific? The key is that the drug only shows an effect on active channels and MSCs are closed in the resting membrane (Suchyna and Sachs, 2007d). If GsMTx4 makes closed channels less sensitive to stress by inhibiting the closed channel, there is no drug effect. Thus, GsMTx4 is only effective on channels in highly stressed regions membranes, and those are likely to be regions of pathology. The physical mechanism of action of GsMTx4 is a gating modifier. A unique amphipath, when applied from the extracellular solution, GsMTx4 dissolves in the outer monolayer and binds near the channels in the domain of the boundary lipids (Suchyna et al., 2004b). However, GsMTx4 is not a universal inhibitor of MSCs. For example, TREK-1 is insensitive to extracellular GsMTx4. We believe that GsMTx4 is effective only channels where the gating region is in the outer monolayer.
We examined the physical properties of GsMTx4 binding to lipids using Trp fluorescence (Posokhov et al., 2007). The affinity for negative lipids was higher than for neutral lipids as expected since GsMTx4 is positive. With GsMTx4 in solution, the tryptophans were quenched with I- and Cs+. When bound to lipid vesicles, the tryptophans were shielded from quenching presumably because they faced the interior of the membrane. Using brominated lipids to quench the Trps, we located the Trps at ~0.9nm from the midline of the membrane. Despite intramembrane localization of the Trps, their spectra were characteristic of a water environment. Apparently, the hydrophobic face of the peptide retains a thin coating of water when bound to the membrane. MD simulations are consistent with this explanation (Posokhov et al., 2008). We have made many GsMTx4 mutants including fluorinated Phe for use in the NMR and a Tyr mutant we iodinated and used for PET scans to examine the pharmacokinetics in mice.
We looked for a high throughput assay we could use to screen GsMTx4 mutants. Since MSCs have been proposed to act as sensors for cell volume regulation (Chen et al., 1996; Christenson and Hoffman, 1992), we built a microfabricated chip to measure cell volume (Ateya et al., 2005) and found a cell type in which volume regulation is controlled by MSCs that can be blocked by GsMTx4 (Hua et al., 2010). We are thus are equipped to screen new drugs and mutants of GsMTx4.
GsMTx4 effects on other cells
GsMTx4 inhibits stretch-induced ET-1 secretion from glia (Ostrow and Sachs, 2005; Ostrow, 2003) and since growing cells create mechanical stress on their neighbors, this will increase ET-1 excretion, and since ET-1 is the most potent mitogen for glia, it closes a positive feedback loop supporting the idea the glial tumors may have a mechanical contribution to the driving force (Ostrow and Sachs, 2005). Related to CNS pathology, GsMTx4 is a potent stimulator of neurite growth in Xenopus spinal cord and PC-12 cells (Barone, 2008; Gottlieb et al., 2010).
In the heart, GsMTx4 can reversibly inhibit atrial fibrillation in a rabbit Langendorff model potentiated by atrial inflation (Bode et al., 2001). But GsMTx4 has no effect on the action potential of resting isolated rabbit heart cells, even at nearly 10X the Kd (Sachs et al., 2004). GsMTx4 also has no effect on human heart tissue removed in surgery (Kockskamper et al., 2007a; Kockskamper et al., 2007b).
GsMTx4 protects sickled red cells from the Ca2+ uptake associated with low oxygen (Vandorpe et al., 2010).
This wide range of effects would seem to warn against the use of GsMTx4 as a drug. However, only the activity of open MSCs is relevant and those channels are in stress regions of cells that are most commonly cells with some form of pathology. The pharmacological manipulation of cell mechanics is a wide open field.