Mechanosensitive Channels: Multiplicity of Families and Gating Paradigms

See allHide authors and affiliations

Science's STKE  10 Feb 2004:
Vol. 2004, Issue 219, pp. re4
DOI: 10.1126/stke.2192004re4


Mechanosensitive ion channels are the primary transducers that convert mechanical force into an electrical or chemical signal in hearing, touch, and other mechanical senses. Unlike vision, olfaction, and some types of taste, which all use similar kinds of primary heterotrimeric GTP-binding protein–coupled receptors, mechanosensation relies on diverse types of transducer molecules. Unrelated types of channels can be used for the perception of various mechanical stimuli, not only in distant groups of organisms, but also in separate locations of the same organism. The extreme sensitivity of the transduction mechanism in the auditory system, which relies on an elaborate structure of rigid cilia, filamentous links, and molecular motors to focus force on transduction channels, contrasts with that of the bacterial channel MscL, which is opened by high lateral tension in the membrane and fulfills a safety-valve rather than a sensory function. The spatial scales of conformational movement and force in these two systems are described, and are shown to be consistent with a general physical description of mechanical channel gating. We outline the characteristics of several types of mechanosensitive channels and the functional contexts in which they participate in signaling and cellular regulation in sensory and nonsensory cells.


Mechanosensation is ubiquitous and takes on many different forms, as cells, tissues, and entire organisms constantly receive and generate various mechanical stimuli. Sounds captured by the ear and static loads on bone represent two extreme examples of stimuli with vastly different magnitudes and time courses. Yet organisms can gauge these stimuli separately by using different receptors and integration schemes. Inputs to the central nervous system (CNS) from specialized mechanosensory organs initiate an array of homeostatic reflexes and behaviors. At the same time, various physiological responses to force can be observed at the level of single cells, demonstrating that mechanoreceptors are indeed ubiquitous.

Because mechanosensory phenomena are so diverse, the progress of unraveling the underlying molecular mechanisms has been slower than with studies of other senses. Indeed, in the mechanisms of olfaction, vision, and some types of taste, the sets of molecular players appear to be evolutionarily conserved and comprise just a few GTP-binding protein (G protein)–coupled receptor second-messenger pathways (13). Given these examples, should we be looking for universal mechanotransduction mechanisms within these various systems and presume that the primary receiver elements are essentially the same, but are tuned differently depending on the setting? If not, are there different types of mechanoreceptors that use similar biophysical principles but dissimilar molecular mechanisms, which arose independently many times in the course of evolution? The examples discussed below suggest that the latter scenario is more likely. Although incomplete, the available molecular data suggest a multiplicity of structural designs and mechanisms of mechanoreceptors not only across the tree of life, but also within single organisms. This diversity implies that careful analysis of individual systems will be required before generalizations can be made.

The Nature of the Primary Transducer

Hearing, vestibular function, and touch--senses that involve specialized organs, such as the ear and the cutaneous tactile receptors, and are present in both vertebrates and invertebrates--represent the fastest and most sensitive mechanosensory systems known. In addition to these specialized organs, free somatosensory endings and visceral afferents are diffusely spread through many tissues; many possess substantial mechanosensitivity, which allows them to monitor cutaneous stretch and deformation, limb position, vascular and alveolar distension, and the filling of the stomach or bladder. All of these functions are attributed to the presence of mechanosensitive ion channels (MS channels) embedded in the plasma membranes of sensory cells (4, 5). In contrast to second messenger–mediated transduction in other senses, the ion channels in mechanosensors are believed to be directly activated by mechanical forces, for the following reasons: (i) An electrical current through the sensory cell membrane (receptor current), a consequence of ion channel opening, appears to be the earliest recorded event following the mechanical stimulus and it precedes other sequelae of the stimulus. (ii) Many mechanosensory responses occur on a submillisecond time scale, which is faster than that permitted by a second-messenger cascade (6). The receptor current in the sensory cell or nerve ending causes a receptor potential, which--in all but the smallest animals--is converted into action potentials in afferent nerve fibers passing to the CNS. In addition to special sensory organs and nerve endings, MS channels are found in various somatic nonsensory cells, where they are presumed to modulate the processes of contraction and secretion in response to mechanical stress, or to regulate cell volume or turgor.

Although they open rapidly and are easy to assay, MS channels are not the only primary receptors for mechanical stimuli. Neurotransmitter release dependent on mechanical stress (7), slow responses to static load such as cell and tissue remodeling (8), and gene expression and proliferative reactions (9) may involve integrins and associated signaling pathways. It has been proposed that in certain organs, stress-induced adenosine triphosphate (ATP) release may exert a paracrine action on purinergic receptors in sensory fibers (10). Further studies will clarify whether this purinergic stimulus causes afferent excitation directly or just modulates it. These related but mechanistically separate processes will not be discussed here. Instead, we focus on MS channels, which constitute a heterogeneous group of membrane proteins in their own right that are unified mostly by function.

There are several systems for classifying MS channels, as outlined below. Conventionally, families of cloned channels are assigned according to sequence homology. Grouping channels according to the tissue of origin (for instance, sensory versus nonsensory) is historical and reflects two complementary approaches: genetic and electrophysiological. Genetic studies begin with screens for mutations that affect sensory function and lead to identification of the chromosomal loci, genes, and eventually proteins (not necessarily channels) responsible for the function of a specific sensory organ. The MS channels from the DEG/ENaC and TRP families were identified genetically. Electrophysiological surveys, in contrast, first identify conductance as a functional manifestation of the channel. This approach led to the identification of MS channel activity in neurons and also in numerous nonneural cells. Using electrophysiology, channel activities can be further characterized with pharmacological and biochemical tools that may suggest the type of molecule responsible for the function. For example, the bacterial MS channel MscL was identified through a biochemical approach, using patch-clamp as an assay technique. Genetic and electrophysiological approaches complement each other and are often used in combination. The two-pore domain MS channels, for instance, were cloned by homology to potassium channels and were identified as mechanosensitive only upon expression in heterologous cells. If a channel is expressed in both sensory (exteroceptor) and nonsensory tissues, it is reasonable to assume that its behavior in both settings, although it could be multimodal, is based on the same intrinsic mechanism.

Without attempting an all-encompassing description, we discuss several examples of such channels and their established or putative roles. We also emphasize the coexistence of at least two mechanistic paradigms of their gating: channels to which force is conveyed by distinct cytoskeletal proteins (often in specialized sensory cells), and channels gated by tension in the lipid (best characterized in prokaryotes).

Channel Gating by External Force

At its simplest, a MS channel is a specialized molecule--usually a protein--in the cell membrane that can undergo a distortion in response to an external force, which either opens or closes a conductive pathway through the molecule. The force can be applied either through cytoskeletal and extracellular matrix elements attached to the channel, or through the lipid bilayer itself (11). Such a distortion can be described as a conformational transition between inactive (closed) and active (open) states separated by a barrier (12, 13). In the simple case of a two-state channel opened by linear force (Fig. 1A), the open probability po is described by a Boltzmann relation:

Fig. 1.

Cartoon representation of channel gating (A) by linear force f or (B) by lateral membrane tension γ. The distribution of channels between the closed and open states in the presence of a mechanical stimulus would be biased by the amount of work the external force produces on the channel as the gate swings (fb) or the protein expands (γΔa).

(Eq. 1)

where f is the force acting on the channel, b is the distance it moves when it opens (the swing of the gate), Δu is an intrinsic energy difference between open and closed states that biases the probability distribution toward the closed state in the absence of applied force, and kBT is thermal energy (~4 × 10−21 J at room temperature). The equation predicts a sigmoidal dependence of po on applied force with the midpoint (po = 0.5) at f = Δu/b. To open the channel, the product of f and b must be several times thermal energy; for instance, it requires a change of 6 kBT to go from 5% to 95% open. As an example, a gate swing of 4 nm under a force of 1 pN produces an energy bias of about 1 kBT.

A similar relation holds for channels activated by lateral tension in the membrane (Fig. 1B). In this instance the free energy term (–fb + Δu in Eq. 1) can be presented as γΔa + Δu, where γ is tension and Δa is the change in the in-plane area of the channel upon opening (13). In this case, one kBT is equivalent to an area change of 4 nm2 under a tension of 1 mN/m. An important consequence is that in this model, mechanically gated channels have no intrinsic threshold; instead, their open probability changes continuously with the applied force or tension. With time averaging, these channels can sense stimulus energies much smaller than thermal energy. The expressions describing channel gating by elastic springs, cytoskeletal networks, and membrane distortion are derived and expanded in the Appendix.

Cellular Structures That Convey Force to a Channel

For the many types of sensory cells that use mechanically gated channels to detect external stimuli, a recurring issue concerns how to bring the stimulus force to bear on a few channel proteins without dissipating the stimulus in irrelevant tissue distortion. Various solutions have evolved; many, but not all, involve the bending of a hair or cilium by the stimulus.

In Drosophila, for instance, touch to the body wall is sensed by bristles protruding from the cuticle (Fig. 2). A single neuron sends a ciliated ending into the hollow bristle, where it is compressed by bristle deflection. Other ciliated mechanoreceptors in insects occur in specialized "chordotonal" organs located at the junctions between body segments, which sense flexion of the joints (Fig. 2). Among these are Johnston’s organ, at the juncture between the second and third segments of the antenna, which is caused to vibrate by sound. Both the bristle and the antenna thus focus force on neuronal endings.

Fig. 2.

Schematic diagram of type I mechanosensory organs, from Eberl et al. (81). (A) Mechanosensory macrochaete bristle. (B) One scolopidium of a chordotonal organ. The sensory neurons (ne) are white, supporting cells are shaded, and cuticular structures are black. Structures: bb, basal body; cd, ciliary dilation; ci, cilium; cr, ciliary rootlet; cu, cuticle; sc, scolopale; and tb, tubular bundle.

In mammals, specialized whiskers on the face convey force to several types of mechanosensors arrayed around the whisker follicles. Some neurons of the trigeminal ganglia, which provide the principal sensory innervation of the face, make lanceolate or reticular endings on the follicles; other trigeminal neurons end on Merkel cells--cells associated with some mechanosensory endings which may themselves be mechanosensitive--that are arrayed around the upper follicle. Elsewhere in hairy skin, Merkel cells are clustered at the bases of guard hairs, which deform the Merkel cells when they are bent.

Cilia can convey force to mechanosensors even within single cells. The simplest example may be the solitary cilium protruding from the apical surfaces of epithelial cells, such as those in kidney tubules. Deflection of these cilia by fluid flow somehow conveys force to MS channels within the cilia to allow Ca2+ influx.

Perhaps the most elegant solution for stimulus delivery occurs in the inner ear, where the mechanosensory cells--the hair cells--use a bundle of stereocilia to apply force to the channels (Fig. 3A). Stereocilia in different inner ear organs have lengths from 1 to 60 μm, although 3 to 10 μm is most common. They are always arranged with heights increasing in order toward the single kinocilium, and the tips are connected by 150-nm-long "tip links" that run only along the short to tall axis of the bundle (Fig. 3, B and C). The mechanically gated transduction channels are located at the tips of stereocilia, most likely at both ends of the tip links (14). A positive deflection of stereocilia (toward the tall stereocilia) causes shearing between adjacent stereocilia that increases tension on the tip links, and this tension is apparently conveyed directly to the channels (1517). Assuming that the tip links are the morphological correlate of the gating springs [which may or may not be true (18)], the combined stiffness of the tip links is about the same as the stiffness of the stereociliary pivots, so half or more of the stimulus force goes toward gating channels.

Fig. 3.

(A) Hair cell bundles from a vestibular organ. Each bundle is composed of ~60 stereocilia and one kinocilium (identified in this organ by a bulbous tip). A gelatinous extracellular matrix normally extends over the entire sensory epithelium and is linked to the tips of hair bundles through their kinociliary bulbs. (B) Two stereocilia and a tip link extended between them. Deflection of the bundle to the right causes shearing of adjacent stereocilia and tightens tip links. (C) Transduction channels (perhaps one to three) are thought to be in the stereocilium membrane at each end of a tip link. A collection of myosin 1c molecules maintains resting tension on the channels to bias them to the most sensitive part of their activation curve. [Modified from (219)]

Several decades of high-resolution physiology have revealed a great deal about the mechanics of stimulus delivery. The stiffness of the hair bundle in hair cells of the bullfrog saccule, a favorite preparation for mechanosensory physiologists, is about 1 mN/m, so that a force of 100 pN is sufficient to deflect the bundle by ~100 nm. A deflection of 100 nm, which is sufficient to open most of the channels, stretches the tip links by about 12 nm. At human auditory threshold, the bundles in the cochlea are deflected by only a few tenths of a nanometer and the tip links are stretched by a tenth of that distance! The channels undergo a conformational change of about 2 nm when they open, so a 100-nm deflection must change the tension in each tip link by about 12 pN to open most channels (for 6 kBT). The stiffness of each gating spring is thus about 1 mN/m.

Although the stereocilia are superbly designed to convert a deflection force to tension on the channel proteins, they must themselves be coupled to a physical stimulus. This occurs in three ways: Hair cells of the vertebrate saccule and utricle are overlaid with an otolithic membrane, a secreted extracellular matrix that is attached to each hair bundle and carries an inertial mass of calcium carbonate crystals ("otoconia"). The acceleration of the otoconia by movement or by gravity (more specifically, the component of acceleration parallel to the epithelium) times the mass of otoconia equals the stimulus force delivered to the combined hair bundles of the sensory epithelium. Similarly, hair cells of the semicircular canals are connected to another extracellular structure, the cupula, which forms a septum across the fluid-filled tube of each semicircular canal. Fluid acceleration caused by head rotation exerts force on the cupula and its attached hair cells. In the cochlea, each cycle of sound causes the basilar membrane, a flexible membrane stretched between two fluid-filled chambers that carries the sensory epithelium, to move up and down. Hair cells riding on the basilar membrane are attached to a third extracellular matrix, the tectorial membrane, which moves with the basilar membrane but shears relative to the hair cells, thereby exerting force on the bundles.

What is the transduction channel for vertebrate hair cells? Despite two decades of physiological research that have defined the selectivity, pharmacology, speed, and molecular mechanics of this channel, its molecular identity is not known. Mutations in the channel would be expected to cause deafness, and a large number of deafness genes have been identified. Many of them are ion channels; some of these channels are expressed by hair cells, but none are good candidates for the transduction channel itself (19). Over the same time period, genetic screens in Caenorhabditis elegans and Drosophila have identified a number of ion channels needed for touch sensitivity and hearing. Many of these are suspected to be directly gated by mechanical stimuli. For the most part, these channels fall into two large superfamilies, the transient receptor potential (TRP) channels and the amiloride-sensitive sodium channels of transporting epithelia and the degenerins (DEG/ENaC channels). Both families are also found in vertebrates, and both contain other ion channels implicated in transduction for a number of sensory modalities.

DEG/ENaC Channels

The DEG/ENaC channel family includes the amiloride-sensitive sodium channels of transporting epithelia (ENaCs), the nematode degenerins (DEGs), the acid-sensitive channels of vertebrate neurons (ASICs; also known as BNCs and BNaCs), and Drosophila PPKs. Proteins of this family have two transmembrane domains and a large extracellular loop that contains two-thirds of the protein. The second transmembrane domain and a small region immediately before it are thought to form the ion conduction pathway. The DEG/ENaC channels are nonselective cation channels of low conductance that somewhat prefer Na+ over K+ and have little Ca2+ permeability, and they are blocked in a voltage-dependent way by amiloride and its analogs. Stoichiometry is still unclear, with evidence for either four or nine subunits per channel (2027).

MEC-4 and MEC-10 in nematode touch

Touch on the lateral body wall of the nematode C. elegans is sensed by five neurons that each send a single process along the body in close contact with the cuticle (28, 29). Worms touched gently on their sides withdraw and move away from the stimulus. Chalfie and Sulston (34) searched for mutants that lacked the touch reflex but were otherwise normal, and found 18 "mec" genes that were needed for touch sensation. All but one are expressed by the touch neurons (30). Six of these encode genes needed for the development of the touch cells (31), but 12 encode structural proteins that may constitute a mechanotransduction apparatus (Fig. 4) [see (32) for a review]. Two structural proteins, MEC-4 and MEC-10, are transmembrane subunits of an ion channel; two others, MEC-2 and MEC-6, are apparently accessory subunits of the channel (33). Other MEC proteins include tubulins (MEC-7 and MEC-12) and matrix proteins (MEC-1, 5, and 9) that could convey force to a channel. Certain mutations in MEC-4 and MEC-10, as well as in the related nematode channels DEG-1 and UNC-105, bias the channels open and cause degeneration of the cells that express them (34, 35); hence, this group of channels was termed the degenerins. When cloned, the vertebrate ENaCs (20) were found to be homologs of the degenerins; together, the two groups defined the DEG/ENaC channel family. Worms and flies have apparently found DEG/ENaC channels particularly useful, expressing 28 and 25 genes of this family, respectively, whereas humans have 9 (29).

Fig. 4.

Possible arrangement of nematode MEC proteins in a mechanotransducing apparatus. Microtubules formed by MEC-7 and MEC-12 come close to the intracellular side of the membrane in the mechanosensitive process of a touch neuron. MEC-2 might connect them to an ion channel made up of MEC-4, MEC-10, and MEC-6. The large extracellular domains of the channel-forming MECs may link to extracellular matrix proteins that can pull on them. [Modified from (220)]

Evidence that MEC-4 and MEC-10 form a mechanically gated channel comes largely by elimination. A screen for touch-insensitive mutants was saturated, producing a number of alleles of these two genes, but no other ion channel was found. A MEC-4/MEC-10 heteromeric channel has been expressed in oocytes, but has not yet been shown to be mechanosensitive (33). Similarly, receptor currents have not been recorded in the touch neurons that express these channels. Nonetheless, genetic identification of the MECs has produced the most complete cast of candidates in any organism, for what is likely to be a mechanosensory complex.

ASIC channels in vertebrate cutaneous touch

A third branch of the DEG/ENaC family was defined by the discovery, by three groups, of a channel expressed in neurons of brain and dorsal root ganglia. It was named MDEG, BNC1, and BNaC1 (22, 23), beginning an unfortunate tradition of multiple names for this branch of the family that was only compounded by alternative splicing (Table 1). The first identified splice form is now known as ASIC2a.

Table 1. Nomenclature for the ASIC branch of the DEG/ENaC channel family. Established gene names for human (ACCNn) are shown along with protein names (ASICnx). Most ACCN genes are expressed in multiple splice forms and most splice forms have had multiple names. ASIC5a is somewhat less related and has not been implicated in mechanosensitivity.

ASIC2a is expressed by many large-diameter neurons of the dorsal root ganglia, which are largely mechanosensory. The protein is not transported to the spinal cord, but is transported along the nerves to their distal terminals in skin (36). Antibodies to ASIC2 label terminals associate with a number of mechanosensory structures in skin, including Merkel cells, Meissner corpuscles, and hair follicles. Some free nerve endings are labeled, but unmyelinated intraepidermal endings mostly are not (36, 37). A targeted deletion of the ASIC2 gene reduced the sensitivity of low-threshold mechanoreceptors, suggesting a role in light touch sensation, although it did not eliminate the response altogether (37).

ASIC3 (DRASIC) is expressed in most sensory neurons, including both those with large and those with small diameters. The protein is located in many of the same sensory endings in skin as ASIC2a, and also in nociceptive intraepidermal endings (38). The physiological effects of gene deletion are mixed; rapidly adapting mechanoreceptors become more sensitive, but mechanical nociceptors lose sensitivity. Heat and acid nociception are affected as well (38).

The striking localization of ASIC channel proteins in sensory endings and the significant (although mixed) results of gene deletion suggest a role in cutaneous mechanosensation. Much work remains to determine whether the ASICs are transducers or play a supporting role, and to sort out the functions of alternative splice forms and heteromultimeric subunit assembly.

ENaC channels in vertebrate baroreception and touch

Amiloride-sensitive epithelial sodium channels (ENaCs) have been extensively studied in transporting epithelia such as lung, kidney, and colon. They are located in the apical membranes of these epithelia, and are constitutively open. Conductance of the ENaCs is regulated by hormonal signals (aldosterone, vasopressin, insulin) that regulate the rapid insertion and removal of the channel from the membrane (39). The ENaC channel in transporting epithelia is composed of three similar subunits: αENaC, βENaC, and γENaC (21). Most vertebrate genomes also have a fourth: δENaC. It is not surprising, therefore, that the ENaC channel has been implicated in salt taste in the tongue [(40) and references therein]: If channels in apical membranes of taste receptor cells are also open, the depolarizing inward Na+ flux should depend on Na+ concentration at the outer surface of the taste bud.

It was more surprising, however, to find ENaC channel subunits in arterial baroreceptor neurons. Baroreceptor cell bodies are located in the nodose ganglia and send axons to innervate the carotid sinus and aortic arch. An increase in arterial pressure stretches the vessel wall, which increases firing of the nerve. Drummond et al. (41) found that nodose ganglia express βENaC and γENaC, but not αENaC. The absence of αENaC is reassuring, because ENaC channels are only constitutively open if this subunit is present (21), and a mechanoreceptor channel might be expected to be closed in the absence of a stimulus. Antibodies to γENaC specifically labeled the sensory endings of nodose ganglion neurons in the aortic arch (41), and also labeled sensory endings in carotid sinus. In vivo experiments on carotid sinus showed that the amiloride analog benzamil in the lumen blocked the pressure-induced firing of the carotid sinus nerve.

Similarly, βENaC and γENaC, but not αENaC, are expressed in dorsal root ganglion neurons. Antibodies to both βENaC and γENaC labeled sensory endings in the skin that innervate Merkel cells, Meissner corpuscles, and small lamellated corpuscles (42).

Location in sensory endings does not define function, and amiloride-like drugs are not terribly specific, so the evidence for ENaCs as a MS channel subunit is largely circumstantial. But the apparent colocalization in dorsal root ganglion (DRG) endings of ENaCs with ASICs, and the growing involvement of the DEG/ENaC channels with mechanoreceptors in many species, are certainly consistent with a mechanoreceptor function.

PPK channels in Drosophila sensory neurons

The Drosophila genome contains at least 21 genes of the DEG/ENaC family; these channels constitute a branch separate from both the MEC branch in worms and the ASIC branch in vertebrates (29). The first such channel identified in flies, PPK1, is expressed in a subset of mechanosensory neurons (43). Insects have two general types of mechanosensory neurons: Type I neurons innervate external sensory organs such as bristles and chordotonal organs such as the sound-sensing Johnston’s organ; they are bipolar, with cell bodies near the end organ, and ciliated, with mechanotransduction occurring in the cilium. Type II neurons have cell bodies closer to the neuraxis and have extensively arborizing dendrites tiling the inner epidermal space. PPK is in a subset of multiple dendritic neurons with arborizing dendrites, and is especially located at varicosities along dendrites that are thought to be the site of mechanotransduction (43). A PPK null mutant exhibits an altered pattern of locomotion (fewer stops and turns) that may be consistent with diminished proprioception (44). As with ASICs, localization of PPK suggests that it may be a MS channel, but much physiological research remains to demonstrate function.

TRP Channels

The founding member of the TRP ion channel family was isolated as the product of a gene defective in Drosophila vision (45). Unlike wild-type flies, trp flies have a transient receptor potential that decays in 1 to 2 s, even with sustained light (46). The trp protein is an ion channel with six predicted transmembrane domains and a pore loop between the fifth and sixth transmembrane domains (47), an architecture similar to that of a large group of voltage-gated K+ channels and cyclic nucleotide–gated channels. Since then, a large number of TRP channels have been identified in various animal species; there are 33 TRP genes, for instance, in the mouse genome (48). Although some are expressed by CNS neurons, many appear in sensory receptor cells, including cells that mediate vision, pheromone sensation, thermal sensation, taste, and touch. TRP channels also occur in several transporting epithelia, where they may mediate cell volume regulation.

PKD channels in vertebrate kidney and nematode sex organs

Polycystic kidney disease (PKD) is an inherited disorder caused most often by mutations in two members of the TRP superfamily, PKD1 and PKD2 (49). PKD1 [also called polycystin 1 (PC1)] is a large protein, with 10 to 12 transmembrane domains and a long extracellular N terminus that contains various binding domains. PKD2 [also called polycystin 2 (PC2)] is much shorter, with the conventional six transmembrane domains. PKD1 and PKD2 are both widely expressed, but their function has been studied most thoroughly in kidney. PKD1 and PKD2 interact through coiled-coil domains in their C termini (50) and form a nonselective cation channel when coexpressed in Chinese hamster ovary (CHO) cells (51). Within kidney cells, the two proteins are located most notably in the single, nonmotile cilium that protrudes from the lumenal surface (52, 53). Bending of this cilium by fluid flow increases intracellular Ca2+ in single Madin-Darby canine kidney (MDCK) cells (54). Kidney cells lacking cilia, or lacking PKD1, do not show the response to fluid flow, nor do cells treated with a blocking antibody to PKD2 (55). PKD1 and PKD2 together may form a MS channel, or PKD1 may activate a channel formed by PKD2. PKD1 and PKD2 are also necessary for normal function in pancreas, blood vessels, and heart, and for left-right axis determination in embryos, and it will be interesting to see whether they function as mechanoreceptors in any or all these tissues.

The nematode C. elegans has a PKD1 homolog known as LOV-1 (location of vulva) and a PKD2 homolog called PKD-2. Both are located in ciliated mechanosensory neurons that innervate the head, the male copulatory spicules, and the male sensory rays at the tail (56, 57). Both LOV-1 and PKD-2 are required for the ability of males to locate the hermaphrodite vulva during mating, although their absence causes no apparent morphological defect and no difficulty in movement, consistent with a direct role in mechanosensation (5658).

Yvc1p in yeast volume regulation

The yeast genome has only one TRP-like channel gene, YVC1P, which encodes a distant member of the TRP superfamily. The Yvc1p channel is located at the limiting membrane of intracellular vacuoles, and is activated by hypertonic stimuli delivered to whole yeast (59, 60). Patch-clamp recording from isolated vacuoles revealed a 400-pS nonselective cation channel, which is absent in yeast that lack the YVC1P gene (59). The channel is activated by increased cytoplasmic Ca2+ and is also activated by increasing hydrostatic pressure across the vacuolar membrane [in the range of 5 to 50 mm Hg (61)]. The channel is thought to act in yeast volume regulation by allowing Ca2+ efflux from vacuoles, which further activates the channels; the consequent rise in cytoplasm Ca2+ may activate compensatory salt transport across the plasma membrane. AlthoughYvc1p is clearly an ion channel, it will be important to determine whether it is the pressure sensor as well.

NOMPC in Drosophila bristle receptors and zebrafish lateral line

To find proteins needed for mechanotransduction in Drosophila, Kernan et al. (62) screened mutagenized larvae for defects in a withdrawal response to touch. Mutations in five genes caused specific defects in bristle mechanotransduction, as assessed by recording extracellular receptor potentials through the hollow bristle. Several of these nomp genes (no mechanoreceptor potential) have been cloned; one of them, nompC, encodes an ion channel of the TRP superfamily (63). nompC is a particularly long TRP channel, with 29 ankyrin domains in an extended N terminus. nompC also has up to eight predicted transmembrane domains, so it is not clear whether the N terminus is intracellular or extracellular. Three of the four mutant alleles of nompC have stop mutations in the N terminus that would certainly act as nulls, but the fourth is a missense mutation between the third and fourth transmembrane domains. Flies carrying this allele showed more rapid adaptation to sustained bristle deflections than did wild-type flies, a rather subtle phenotype that is the best evidence that nompC is in fact the transduction channel itself. nompC is expressed in cells within the bristle organ, but it has not yet been localized to the sensory neuron itself, nor has the protein been localized within cells. Walker et al. (63) have identified a nompC ortholog in nematode. In transgenic C. elegans expressing a GFP gene driven by the nematode nompC promoter, fluorescence was observed in ciliated mechanosensitive neurons.

There is also a nompC ortholog in the zebrafish genome (64, 65). Sidi et al. found that the zebrafish nompC is expressed in hair cells of the otic vesicle, and that zebrafish embryos injected with morpholino antisense oligonucleotides were deaf (as assessed by an acoustic startle reflex) and exhibited circular swimming behavior consistent with a vestibular defect (65). Fish also have hair cells in their lateral line organs, structures comprising clusters of 10 to 20 hair cells in the skin whose stereocilia are exposed to the water and which detect water currents. Although nompC transcript was not detected in lateral line hair cells by in situ hybridization, morpholino-injected zebrafish failed to accumulate the dye FM1-43 (65), a fluorescent marker of functional transduction that enters cells through mechanosensitive transduction channels (66). Nor did they show a stimulus-evoked "microphonic" potential, an extracellular potential that is a simple physiological measure of functional channels (65). Thus, zebrafish nompC is required for hair cell transduction, and its similarity to fly nompC is further evidence that this channel constitutes at least part of the transduction channel for zebrafish hair cells. On the other hand, there is not a nompC ortholog in mouse or human genomes (67), so mammalian hair cells must use something else.

TRPV4: An osmosensor and more

A search for vertebrate homologs of Osm-9 and TRPV1 led several groups to clone the mammalian TRPV4 channel [also called OTRPC4, VR-OAC (vanilloid receptor osmotically activated channel), and VRL-2 (vanilloid receptor–like 2)] (6871). This channel, like other TRPs, is a nonselective cation channel with significant Ca2+ permeability. The sequence similarity to Osm-9 suggested that TRPV4 may also have osmotic sensitivity; indeed, when expressed in cultured cells, it was found to potentiate Ca2+ entry in response to hypo-osmotic stimuli. A decrease in osmolarity of 10% reliably evokes a signal from an intracellular Ca2+-indicator dye (68). TRPV4 is widely expressed, appearing in distal convoluted tubule of kidney, liver, heart, vascular endothelial cells, testis, spleen, salivary gland, lung, and trachea, and in the stria vascularis of the cochlea (6871). Its appearance in many transporting epithelia, which must balance large apical and basal fluxes of salts and water, suggests a role in cell volume regulation. Within the nervous system, TRPV4 is found in osmosensory cells of circumventricular organs in brain and in the Merkel cells surrounding hair follicles (69), so that it may have a sensory role as well. Indeed, mice lacking TRPV4 have reduced sensation to noxious tail pressure (72) and slower recovery of normal serum osmolarity after salt loading (73).

Understanding the function of TRPV4 has become more complicated (74), because it is also activated by heating to 25° to 30°C, but then inactivates rapidly when heated above ~35°C (75). Finally, it can be activated by a phorbol ester (4α-phorbol 12,13-didecanoate) that does not activate protein kinase C (PKC) (74). Responses to all of these stimuli are relatively slow (seconds), perhaps because stimulus delivery is slow, but whether TRPV4 is directly or indirectly activated by mechanical stimuli remains unclear.

Osm-9 and OCRs: multimodal receptors in nematode

The osm-9 gene was first identified in a screen for mutants defective in olfaction in C. elegans (76, 77). Like many TRP channels, OSM-9 has three N-terminal ankyrin domains. These might mediate association of OSM-9 with the cytoskeleton to convey force, or they may simply bind other proteins to mediate precise channel localization within sensory neurons. More extensive characterization of the gene product revealed that it is expressed in various sensory neurons that mediate olfaction, osmosensation, and mechanosensation, and that osm-9 mutants are defective in osmosensation and in certain types of olfaction and mechanosensation (78). Nematodes touched lightly on the nose stop moving forward and start backing, a behavior mediated by the ASH, FLP, and OLQ cells, which are ciliated neurons that send processes to the nose (79). Mutants of osm-9 are defective in response to light nose touch, and the ASH, FLP, and OLQ neurons express osm-9 (78).

The osm-9 gene encodes a TRP channel that is part of the invertebrate branch of the TRPV family. The nematode genome also contains four other genes in this branch, which have been named ocr-1 to ocr-4, for osm-9/capsaicin-receptor related (80). ocr genes are expressed in a partially overlapping set of sensory neurons in nematodes, but osm-9 is also expressed in any cell that expresses an ocr gene. For instance, both osm-9 and ocr-2 are expressed in ASH neurons, which can be activated by mechanosensory and other stimuli, and mutation of ocr-2 abolishes sensitivity of ASH neurons to nose touch (as well as other stimuli). Both osm-9 and ocr-4 are expressed in the mechanosensory OLQ neurons (80).

TRPV channels in Drosophila hearing

The Drosophila nompC is a strong candidate for a bona fide mechanically gated channel, and so it was gratifying to find it expressed in chordotonal organs like Johnston’s organ, which is the fly auditory organ (63). Whereas other nomp genes are also needed for fly hearing, however, nompC mutants have only partially compromised hearing (81). Two other TRP channels have now been identified as required for auditory sensation by Johnston’s organ; both are in the invertebrate branch of the TRPV family. The channel encoded by the inactive gene (CG4536) is most closely related to the nematode OSM-9, whereas that encoded by nanchung (CG5842) is an ortholog of the nematode OCR channels (82, 83). Just as OSM-9 interacts with OCR-2 to mediate mechanosensation in nematodes, perhaps inactive and nanchung together form a MS channel that mediates fly hearing. Neither gene has direct orthologs in vertebrates.

Painless: A TRPA channel in Drosophila nociception

In addition to the TRPC, TRPM, TRPV, PKD, and NOMPC branches of the TRP superfamily, there is a sixth branch (TRPA) that has been largely neglected by scientists (and by evolving vertebrates). Human ANKTM1, the founding member of the TRPA family, is the single mammalian member of this group, but nematodes have two TRPA channels, Drosophila have four, mosquitoes have seven, and the sea squirt Ciona has six (48). One of the Drosophila TRPA channels, encoded by the painless gene, is necessary for both thermal and mechanical nociception in fly larvae (84). Multidendritic type II (nonciliated) sensory neurons express painless and fire when the temperature exceeds 38°C, but not if painless is disrupted. Similarly, normal larvae exhibit a withdrawal reflex to strong mechanical stimuli that is lost in painless mutants, whereas responses to light touch are normal in the mutant. It may be that painless is a MS channel that is also temperature sensitive, or it may be indirectly activated by tissue damage produced by either thermal or mechanical stimuli. Similarly, the mouse ANKTM1 is activated by very cold temperatures, but is expressed in a small number of nociceptive DRG neurons that also express heat-activated TRPV channels and may be multimodal nociceptors (85).

Mechanosensitive Channel Activities in Nonsensory Cells

Channels sensitive to local pressure gradients across the patch membrane have been recorded in many somatic nonsensory cells (8688). Phenomenologically, channels observed in such experiments can be divided into stretch-activated (SA) or stretch-inactivated (SI) channels and can be further classified by ion selectivity and sensitivity to Ca2+, voltage, etc. Many of these channel molecules have not been identified to date and are known only through characteristic currents that are generally blockable by Gd3+ ions (89). Despite the scarcity of molecular information, such channels are clearly involved in intracellular signaling and, in some instances, in the development of pathologies.

Skeletal muscle

SA channel activities were first documented in embryonic chick skeletal muscle cells (90). The 70-pS currents were moderately K+-selective and were more easily activated after treatment of cells with cytochalasin B, suggesting a redistribution of the stress from the disrupted actin cytoskeleton to another stress-conveying component coupled to the gating machinery. Subsequent studies of dystrophic (mdx) muscles with a compromised membrane-cytoskeleton attachment revealed that stretch channels in dystrophic myoblasts, myotubes, and isolated fibers are constitutively more active than in nondystrophic muscles (91). The channels in mdx animals also exhibit irreversible mechanically induced switching to a deregulated open state, making such muscles prone to continuous Ca2+ poisoning (92). Recordings made in muscles from animals deficient in delta-sarcoglycan, a part of dystrophin-glycoprotein complex, led to similar conclusions (93). The SA channels in skeletal muscle have yet to be cloned and their normal physiological role remains to be clarified.

The heart

In the heart, MS channels are involved in feedback processes coupled to the mechanics of the cardiac cycle. Mechanoelectrical feedback normally accelerates the heart in response to excessive distension of the muscle (94). However, under certain circumstances it may also cause arrhythmias that arise from an infarction zone (95), or deadly fibrillation triggered by a sudden blow to the chest (96). Patch-clamp examination of cardiomyocytes from snail, frog, chicken, mouse, rat, guinea pig, and human shows a range of stretch-activated currents [see references in (87, 97)]; the most frequently observed are a 90- to 100-pS K+-selective current (98) and a 20- to 25-pS nonselective cationic current (Erev ≈ –20 mV), which carries Ca2+ (99,100). These currents are activated by suction in a patch pipette (101), by stretching cells (102), or by applying pressure to the cell surface with a blunt glass probe (103105). Channel opening permits sufficient Ca2+ influx to initiate Ca2+-induced Ca2+ release from the sarcoplasmic reticulum and subsequent contractions (99). Cardiac MS channels are blocked by venom from the spider Grammostola spatulata (106). The 28–amino acid active peptide isolated from the venom is antiarrhythmic and its structure has recently been solved using nuclear magnetic resonance (NMR) (107). The very recent cloning of SAKCa, a stretch-activated splice-variant of the BK Ca2+-dependent potassium channel (108), should help dissect the molecular mechanisms of cardiac mechanoelectrical feedback.

The mechanism underlying the massive fibrillation that follows a sudden impact to the chest known as commotio cordis implicates both types of cardiac MS channels mentioned above (96). The likely candidate for the K+-selective species, the KATP channel, is substantially modulated by membrane stretch (109). Glibenclamide, which blocks KATP, reduces the probability of post-impact ventricular fibrillation, possibly by suppressing impact-produced repolarization, which shortens the cell refractory period in the zone of impact. Such a desynchronized tissue, in terms of refractoriness, is predisposed to ectopic contractions, especially when the impact arrives during the period of repolarization (96). The early depolarizations attributed to the cation-selective (depolarizing) current trigger ectopic excitation, which poses a high risk of evolving into a ventricular fibrillation.

Cardiac stretch-activated channels are also proposed to mediate the release of the atrial natriuretic peptide (ANP) in response to atrial distension (100). The complete mechanism of ANP secretion has not been elucidated, but activation of KATP channels have been implicated, as the stretch-dependent component of ANP release is blocked by glibenclamide (110).

Smooth muscles

Smooth muscles are capable of autoregulation and actively contract in response to stretch. This feedback reaction, known as myogenic reflex, is believed to be initiated by stretch-activated channels. Early studies (111) have shown that stretch channels in the urinary bladder muscles are abundant, indiscriminately conduct cations (including Ca2+), and are more active when hyperpolarized. Longitudinal stretch of a cell increases the channel activity proportionally to the elongation (112). Recent studies using patch-clamp combined with Ca2+ imaging have shown that Ca2+ influx through single stretch-activated channels is sufficient to activate large Ca2+-activated K+ channels, trigger massive Ca2+ release from intracellular stores (113, 114), and in turn, activate phospholipase C (PLC) (115).


The endothelial lining, like vascular smooth muscle, participates in regulation of vessel diameter. Endothelial cells respond to pressure, distension, and fluid shear stress by releasing vasodilators such as nitric oxide (NO) and prostacyclin, most likely in response to intracellular Ca2+ mobilization (116). Stretch-activated Ca2+-permeable channels have been reported in cell preparations from the umbilical vein, aorta (117), and mesenteric arteries (118). When cells cultured on an elastic silicon membrane are subjected to stretch, extracellular Ca2+ enters cells not through voltage-gated Ca2+ channels but through stretch-activated channels (119). The calcium entry leads to extensive rearrangement of stress fibers and focal adhesion zones involving c-src and focal adhesion kinases (120). A peculiar channel, responsive to positive pressure and blockable by suction in the patch pipette, is found in intact endocardial endothelium (121). Substantial Ca2+ permeability and the up-regulated state of endothelial MS channels in hypertensive animals points to their possible role in counterregulation of elevated blood pressure (118).

Fluid shear stress acting on endothelial or epithelial cells may exert a different type of deformation compared to tensile stress caused by vascular wall distension. The involvement of SA channels in shear-stress responses has been questioned on the basis of the absence of Gd3+ inhibition of stress fiber remodeling in bovine aortic endothelium (122). Nonetheless, SA channels are up-regulated after stimulation by shear stress in endothelial culture and are present at higher density in endothelia of spontaneously hypertensive animals (123). In kidney epithelium, as mentioned above, fluid flow sensation is ascribed to the primary cilium present on the luminal surface of principal cells. PKD2, a high-conductance Ca2+-permeable channel from the TRP family, has been recently localized to the cilium together with its PKD1 counterpart (55) (see above).

Volume regulation

Volume regulation under changing osmotic conditions is an intrinsic property of all cells. It involves mechanosensation in that swelling or shrinkage puts strain on the membrane and intracellular structures. Two types of cationic MS channels have been shown to be involved in volume regulation for the renal proximal tubule epithelium (124). The phenomenon of regulatory volume decrease (RVD) following hypotonic swelling was ascribed to a coordinated action of volume-sensitive anionic (ClC-2) and cationic (maxi K) channels, as well as osmolyte transporters (125127). None of the anion channels are directly sensitive to membrane stretch, which suggests that RVD is not triggered directly by membrane tension. It has been inferred that volume-sensitive TRPV channels (see above) activate first, providing Ca2+ influx, followed by secondary activation of maxi K and other channels (69).

Osmosensation is also critical for whole organisms in the homeostatic maintenance of the fluid and electrolyte balance, and it may use mechanisms similar to those in regulatory cell volume changes. One of the genes needed for osmotic avoidance in C. elegans, osm-9, encodes a TRP channel (see above), which is also involved in mechanosensation by ciliated touch neurons in the nose of the worm and in olfaction (78). A mammalian homolog of OSM-9, the hypotonically activated TRPV4, is localized in brain, sensory neurons, ear, lung, and kidney, which suggests that this mammalian channel mediates broad tissue responses to changes in systemic osmolarity (69). Because the deletion of ankyrin repeats providing cytoskeletal attachments to the channel did not abolish hypotonic activation, it has been proposed that the channel must be gated directly by membrane stretch; however, no other evidence has been presented.

By regulating drinking behavior and water reabsorption in the kidney, the hypothalamus imposes central control over the osmotic balance in the entire body. The magnocellular neurosecretory cells, located in the paraventricular and supraoptic nuclei in the hypothalamus, are intrinsically sensitive to the osmolality of extracellular fluid (128). When the osmolality increases above 295 mosm, the magnocellular cells depolarize and release vasopressin and oxytocin from branched neurosecretory terminals that project to the posterior pituitary. This response is directly related to the volume decrease in the bodies of magnocellular neurons, which behave almost as ideal osmometers (129). The accompanying depolarization is ascribed to stretch-inactivated cationic channels that activate upon cell shrinkage. The putative channels have a conductance of 32 pS and a ratio of K+ to Na+ permeabilities of 5:1 (130). The hypothalamic osmosensory channel has not yet been identified.

Two-pore domain potassium channels

TREK and TRAAK represent a group of four–transmembrane domain channels found in high density in several areas of the CNS (131). These regulated "leakage" channels are believed to be responsible not only for maintaining the resting potential in neurons, but also for regulating it in response to different cues such as mechanical or osmotic stress, intracellular pH, or temperature. Phosphorylation by intracellular protein kinases in response to increased levels of cyclic adenosine monophosphate inhibits the channels (131). These leakage channels are truly mechanosensitive: They increase their po by orders of magnitude in response to membrane stretch applied through the patch pipette. TREK-1 is also modulated by osmotic cell swelling (132). Besides direct membrane stretch, TREK-1 is readily activated by elevated levels of polyunsaturated fatty acids, lysophospholipids, or volatile anesthetics. The sensitivity of TREK-1 to stretch, anesthetics, and intracellular acidosis requires an intact C-terminal domain (133, 134). The pharmacological and functional properties of TREK channels closely resemble those of S-channels in Aplysia sensory neurons (135), which are responsible for serotonergic modulation and plasticity. It is presumed that the neuroprotective role of TREK and TRAAK channels in mammalian neurons is hyperpolarization and prevention of Ca2+ poisoning in the event of ischemic brain damage, which is typically accompanied by acidosis and osmotic cell swelling.

What Happens Under the Patch Pipette?

The questions of how stretch channels gate and whether they obligatorily require the cytoskeleton for their function have been discussed in reviews by Sachs and Morris (87) and by Hamill and Martinac (88). It remains possible that in some instances stress is conveyed to the sensory molecule through the network of submembrane filaments (discussed in the Appendix). The frequent observation that pretreatment of cells by cytochalasins makes the channels more sensitive to stretch suggests, however, that the cytoskeleton is not a force-transmitting element but rather a restraining element. The MS channels that become more active when the membrane is partially liberated from the cytoskeletal scaffold or completely decoupled from it are likely gated directly by lateral tension in the lipid bilayer. Consistently, lysolipids, fatty acids, and anesthetics, which increase membrane area and induce curvature stress in the lipid bilayer, sensitize some types of stretch channels to mechanical stimuli (98, 134, 136). With a few exceptions (132, 137), MS channels can be activated either by suction or by positive pipette pressure in inside-out patches; this illustrates that lateral tension, not bending in a specific direction, is the key parameter. In many systems, it was also possible to find correspondence between the properties of single-channel currents activated by suction in cell-attached patches and whole-cell currents evoked by positive pressure (6, 87, 88). Two exceptions, where whole-cell mechanosensitive currents are practically impossible to evoke, are Aplysia neurons (138) and Xenopus oocytes (139). Both types of cells are characterized by highly developed microvilli and dense networks of cortical cytoskeleton. Unwrinkling of microvilli and partial detachment of the membrane from the cytoskeleton by strong suction under tight seal conditions alters MS channel behavior in these cells (140, 141). An interesting phenomenon of changes of patch geometry produced by depolarizing voltages, accompanied by activation of MS channels, has been reported in Xenopus oocytes (142). Measurable movements of patch membrane were observed in response to prolonged depolarizations (exclusively with borosilicate, but not soft glass pipettes). The questions of how profoundly the structure of a patch membrane is perturbed by the tight seal and whether channel activities produced by stretch or voltage under such conditions are physiologically relevant must be addressed for each particular case.

Why Do Walled Cells Have Mechanosensitive Channels?

At first glance, it seems puzzling that organisms with strong extracellular walls would need MS channels. Indeed, the static shape of a plant is maintained in part by turgor pressure that the cytoplasm exerts on the cell wall. As long as cell expansion is restrained, the plasma membrane, typically present in excess, experiences no lateral tension (143). But in growing plants, the cell wall undergoes constant remodeling and, in certain areas, very fast expansion (144), and, therefore, new membrane must be inserted at the same pace. Bacterial cell walls are also distensible (145, 146), and when swelling under strong osmotic shock occurs, they may not guarantee complete mechanical protection to the plasma membrane. Indeed, turgor-operated osmolyte efflux systems have evolved in many unicellular organisms (147). MS channels would perfectly fit the role of membrane tension reporters that provide necessary signaling feedback as well as simple permeation pathways for osmolytes. MS channel activity has been documented in plant protoplasts (148, 149) and vacuoles (150), fungi (151), yeast (152), yeast vacuoles (61), archaea (153). and bacteria (154, 155).

Besides turgor regulation, MS channels in plants may participate in gravitropism (149), which makes roots grow downward and shoots grow upward. Mechanistically, gravitropic responses are initiated by intracellular sedimentation of dense starch granules (amyloplasts) in stems and root caps and are ultimately effected by a directional transport of the growth hormone auxin (156). The intermediate events in this signaling cascade, which may involve MS channels, have not yet been elucidated.

Bacterial Mechanosensitive Channels Are Osmotic Safety Valves

The osmolyte efflux system in Escherichia coli comprises at least three known inner membrane tension-operated channels of high conductance, namely MscL (157), MscS [formerly known as YggB (158)], and MscK [formerly KefA (159)]. Homologs of each of these three proteins are found in almost all Gram-positive and Gram-negative bacteria. BLAST searches also identify MscS homologs in fission yeast and higher plants (Arabidopsis thaliana), whereas a distant MscL homolog was found in fungus Neurospora crassa (155). A group of channels that resembles MscS and MscL, including MscTA and MscMJ, has been isolated from archaea (153).

When mscL-null mutants of E. coli were first generated, no visible growth or survival phenotype was observed under various osmotic conditions (154). The double MscL/MscS knockout displayed an osmotically fragile phenotype, proving that a tension-dependent increase in membrane permeability is indeed critical for cell survival under strong hypo-osmotic shock, and indicating some functional redundancy of the system (158). MscK alone, although it opens before MscS and MscL, is unable to rescue cells under acute osmotic shock, which could be due to either low-copy-number protein expression or specific regulation by permeant ions (160). The MscS and MscL channels, although they share no sequence homology, are functionally similar in many respects. They are both active when purified and reconstituted in liposomes, indicating that they gate directly by tension in the lipid bilayer (161163). MscS is moderately voltage sensitive (164) in that it is more active at depolarizing membrane potentials. Both MscL and MscS are sensitive to membrane curvature perturbations induced by amphipathic compounds such as lysolipids or local anesthetics (165, 166). The successful crystallographic work done in the laboratory of Rees (167, 168) makes MscL and MscS good models for basic studies of molecular mechanisms of channel gating by stretch.

According to the crystal structure, MscS is a homoheptamer of three–transmembrane domain subunits (168). C-terminal segments of each subunit contribute to a large cytoplasmic domain that apparently acts as a molecular sieve. The transmembrane domains (TM1 and TM2) bear positive charges and are oriented at about a 30° angle to the pore axis. The structure shows that the MscS pore, formed by the largely hydrophobic third transmembrane domain (TM3), is open. However, it is unclear whether the size of the pore and the character of the lining can support the 1-nS conductance observed in experiments. More studies are required to verify whether the crystal structure represents the closed, open, or some other (desensitized) state, and what the gating transition looks like. We next discuss the gating mechanism of the large MS channel MscL, which at this time is better understood than that of MscS.

MscL, a Simple Model

MscL has been identified as a 136-residue protein that forms a large stretch-activated channel upon reconstitution into asolectin liposomes (157). The protein has been localized to the inner membrane of E. coli and is predicted to cross the membrane twice, with both the N and C termini in the cytoplasm (169). Deletions show that the short N-terminal segment is critical for channel function or assembly, whereas the absence of 25 C-terminal residues could be tolerated (170, 171). The extremely large (3.2 nS) nonselective conductance of the fully open state suggests a water-filled pore 30 to 40 Å in diameter (172, 173). The midpoint tension of channel activation is 10.4 dynes/cm in spheroplasts (174) and 11.8 dynes/cm in large liposome patches (173), respectively, which is close to the lytic limit for many bilayers.

The crystal structure of the MscL homolog from Mycobacterium tuberculosis (TbMscL) was solved to 3.5 Å resolution for the closed state (167), revealing a tight assembly of five two–transmembrane domain subunits. Slightly tilted M1 helices are inside the channel barrel, whereas the M2 helices are on the periphery facing the lipid. The C-terminal domains, connected to M2 helices with flexible linkers, are helical and form an additional pentameric bundle in the cytoplasm. The M1 helices, which are tightly packed at their intersection near the middle of the membrane, create a hydrophobic pore constriction proposed to be the gate. It was initially hypothesized that the opening of the pore involved straightening of the M1 helices and the formation of a cylindrical barrel with parallel M1 and M2 helices acting as "staves" (but see below) (167, 175).

Random mutagenesis of mscL combined with screens for growth-suppressing mutations gave the first "functional" clues as to where the MscL gate might be (176). The isolated mutations indicated that the integrity of the cytoplasmic half of M1, specifically the constriction region where the helices are tightly packed, is critical for maintaining the channel in the closed conformation at subthreshold tensions. Most of the toxic mutations were either polar or charged substitutions for hydrophobic residues occluding the pore, or bulky side-chain substitutions for regularly spaced glycines in M1 (apparently involved in tight helical packing). The majority of these mutations produced easy-to-open ("gain-of-function") channel phenotypes. Further site-specific mutagenesis studies demonstrated a direct dependence of the activation pressure threshold on the degree of hydrophobicity of a substitution for Gly22 near the constriction point, indicating that hydration of the pore interior favors the open state (177179).

Kinetic and thermodynamic analyses of MscL in liposomes provided important information on the character of the gating transition. Recordings showed that MscL opening occurs through a series of short-lived substates (173). The relative dependencies of transitions between the substates on tension suggested that gating proceeds through a pre-expanded low-conducting intermediate state. A molecular model of the transition of MscL between the closed and the open state through an expanded intermediate state is depicted in Fig. 5. The homology model of E. coli MscL, built from the TbMscL crystal structure, represents the initial closed conformation. The "parallel barrel-stave" model of the open conformation failed to satisfy several criteria (180). Expansion of the transmembrane barrel is achieved not by straightening of the helices, but rather by their tilting with a simultaneous outward movement in an iris-like manner. The N-terminal segments (S1), unresolved in the crystal structure of TbMscL, were hypothesized to form a short bundle that acts as a secondary cytoplasmic gate (181). In the short-lived pre-expanded conformation, this gate is closed and the channel remains in a nonconductive or, more likely, low-conductive state. Further expansion of the barrel pulls on short S1-M1 linkers, resulting in disruption of the N-terminal bundle and full opening of the channel. The C-terminal (S3) helices were initially modeled as separating (180); however, recent data have strongly suggested that they remain stably associated in all conformations and form a prefilter on the cytoplasmic side of the pore (174).

Fig. 5.

Molecular models of the large bacterial MS channel MscL in the closed, pre-expanded, and open conformations [after Sukharev et al. (180)]. The conformation of the S3 cytoplasmic bundle is depicted according to Anishkin et al. (174). The transition from the closed state (left) is accompanied by tilting and radial movement of the transmembrane helices. The pore constriction formed by M1 helices apparently acts as a primary gate. When the barrel expands and the M1 gate opens, the S1 domains may still hold together, forming a short-lived low-conducting state (middle). S1-M1 linkers pull apart the S1 bundle, which results in complete opening of the channel (right).

Different aspects of this model of the open state have been tested using disulfide cross-linking, and also in patch-clamp experiments on cysteine mutants under reducing or oxidizing conditions (174, 181, 182). The barrel was trapped in the expanded state with highly tilted M1 helices by pairs of cysteines in positions 20 and 36 on different subunits. These cysteine cross-links stabilized the open state of the channel. Cross-links between adjacent M1 and M2 helices of neighboring subunits did not prevent gating, indicating that these helices tilt together during the transition; in other words, they do not change their relative position toward one another substantially (182). The proximity of highly conserved phenylalanines (Phe7 or Phe10) in the N-terminal gate (S1) was supported by spontaneous intersubunit cross-linking of cysteines in these corresponding positions, which prevented the channel from opening. At the same time, a random mutagenesis study by Maurer and Dougherty (183) suggested that, despite a high degree of conservation, the N terminus may tolerate drastic substitutions without affecting channel function in in vivo cell viability assays. The length of the linkers that connect S1 domains (N-terminal gate) with M1 nonetheless had a strong effect on the structure of the subconducting states, suggesting a functional role as tension-transmitting elements between the domains (181). Disulfide cross-links between the C-terminal (S3) helices did not affect gating parameters, consistent with their stable association in all conformations (174).

Similar to the activation of MscS by amphipathic substances (165), E. coli MscL reconstituted in liposomes can be stabilized in the open state by adding lysolipids to the outside of the liposome (166). This made it possible to perform a site-directed spin labeling and EPR (electron paramagnetic resonance) study of MscL in the open state, which independently confirmed not only that the pore is lined primarily by the M1 helix, but also that the M2 helix is not exposed to the aqueous phase in any state (184). The parameters of EPR spectra predicted separation of the M1 helices by about 25 Å. EPR-derived constraints produced a model with a highly tilted conformation of helices, similar to that depicted in Fig. 5. The iris-like expansion of the channel barrel was also observed in molecular dynamic simulations (185187), especially when the steering force mimicking membrane tension was applied specifically to the residues facing the layer of phospholipid glycerols, the most rigid and densely packed part of the bilayer (186). The residues that are in contact with lipid head groups have been shown to be functionally important for sensing tension in in vivo assays (183).

The current model of the open state, under constraints imposed by the cysteine cross-link and EPR data, predicts that both M1 and M2 helices tilt by about 40° and move outward by about 12 to 13 Å. The inner surface of the pore is lined mostly with the polar face of M1. The total in-plane expansion of the transmembrane barrel assessed from models is 22 to 23 nm2, which is highly consistent with the initial slope of tension-activation curves (188). The flattening of the structure reduces the hydrophobic thickness of the protein, which may substantially distort annular lipids around the channel complex. Indeed, the open state of MscL was shown to be more stable when the channel was reconstituted in short-chain lipids rather than long-chain lipids (166). The energy of the closed-to-open transition is estimated as 50 to 55 kT (188), corroborating with extremely high activation tension (10.4 dynes/cm) and a large-scale conformational rearrangement. Such a high energy cost may include contributions from conformational stress in the barrel and periplasmic loops (189), from unfavorable pore hydration (177), and from perturbation of the surrounding lipids (166). Further studies are required to quantify these energetic contributions.

Mechanosensitive Channels: Do They Have Anything in Common?

Several families of MS channels that function in different contexts in the same organism clearly coexist. In animals, channels from the TRP family are implicated in osmosensation, gauging of fluid flow, and mechanosensation in certain neurons. Other groups of sensory neurons use DEG/ENaC-type channels. There are stretch-activated potassium channels from the seven–transmembrane domain (7TM) (SAKCa) and 2TM (KATP) families in the heart, as well as mechano-gated two-pore (4TM) potassium leakage channels in the brain. Even a simple prokaryote, E. coli, has at least two types of functionally similar MS channels (MscL and MscS) that share no homology or structural similarity.

In terms of gating mechanisms, the MEC channel of C. elegans critically relies on specific cytoskeletal attachments. The auditory channel, which is possibly a member of the TRP family, requires intact tip links for gating and a connection to the actin cytoskeleton through myosin for opening. It is unclear at present whether other mechanosensitive TRP channels require the cytoskeleton for gating, but it seems likely. SAKCa (136) and the two-pore (TREK) potassium channels (134), however, seem to be gated directly by tension in the lipid bilayer.

In addition to strongly mechano-activated channels with a steep dependence of open probability on the stimulus, there have been reports of mechanical modulation of well-characterized channels with established functions other than mechanosensation. These include ATP-dependent (KATP) potassium channels (109), G protein–modulated (KACH) inward rectifiers (190), and NMDA receptors (191). Calcium-dependent BK channels (192), voltage-gated K+ channels (193), L-type Ca2+ channels (194), and N-type Ca2+ channels (195) are also stretch-modulated. Because many types of non-MS channels are modulated by stretch, one may speculate that the entire spectrum of MS channels could be the result of independent evolution of several distinct predecessors to acquire mechanical sensitivity, by more efficient linkage to cytoskeletal elements or by an increased conformational flexibility or area change upon opening. This notion is consistent with the existence of MS channels in various channel superfamilies that contain channels activated by other stimuli than force. Examples include the two-pore TREK and TWIK channels (131), SAKCa and BK (108), or osmosensitive and non-osmosensitive TRPVs (196). Despite attempts to find a common structural motif underlying the mechano-gated behavior (197), a common origin of MS channels is not evident.

Concluding Remarks

Various putative MS channels and their functional contexts have been described. No unifying common origin of MS channels is evident; rather, evolution to mechanosensitivity has apparently occurred many times, producing several types of MS channels based on various channel architectures.

Two mechanistic paradigms exist: gating by force conveyed through elements tethered to the cytoskeleton, and gating by force conveyed through lateral tension in the membrane. "Tethered" channels tend to reside in specialized sensory cells and organs, which typically bear flexible extensions such as stereocilia, specialized dendrites, or hairs. Deflection of these extensions in certain directions focuses the stress on the channel, which acts as a transducer and undergoes a gating transition at the nanometer scale. In the auditory system and probably elsewhere, molecular motors adjust the set point for gating of such channels.

For many channels found in nonsensory cells, the plasma membrane itself constitutes the receptive structure for mechanical stimuli. Interactions with the cortical cytoskeleton or the external cell wall define the fraction of stress borne by the lipid bilayer, and therefore the amount of tension sensed by the channel. MscL is a simple prokaryotic channel gated by tension in the lipid bilayer; a detailed understanding of its gating serves as an excellent prototype for channels of this type. During MscL gating, a large iris-like transition driven by lateral membrane tension is accompanied by a strong tilting of transmembrane helices. The spatial scale of transition predicted by molecular models corresponds well to results of thermodynamic analysis, which supports the model and the way the stress is transmitted from the bilayer to the channel gate.

Appendix: Energetics of Mechanical Gating

Gating by Linear Force

We present a simple theory for the gating of mechanically sensitive ion channels, which assumes just two conformational states. Although it only approximates channel behavior, this theory nonetheless represents a consistent viewpoint that can be used to understand and compare different systems. As such, it provides a framework for understanding the molecular entities involved in mechanically sensitive channel gating. The theory was presented initially by Corey and Hudspeth (198) and has been expanded by Howard and Hudspeth (17), Howard et al. (199), Hudspeth (14), and Denk and Webb (200). It is based on Eyring’s (201) theory of reaction rates, and parallels the theory for voltage-dependent gating of ion channels.

Any gated ion channel has at least two stable conformations, an open state that conducts ions and a closed state that does not (Fig. 6). Although the "gate" is understood perhaps only for the MscL channel, it is useful to think of some part of the protein moving like a gate and opening the channel. The free energy profile for a channel is drawn as a continuous function of distance along a reaction coordinate that represents the movement of the gate in Fig. 6, which indicates the stable wells of the open and closed conformations and an energy barrier that limits transitions between the states. Because the continuous energy profile cannot be known with any certainty, simple theories generally consider just three points: the closed state, the open state, and the peak of the barrier. The energy difference between the two states, Δg, dictates the likelihood of the channel occupying each state, as described by a Boltzmann distribution:

Fig. 6.

Force-dependent gating. (A) Cartoon of a channel with a gate that moves by a distance b when sufficient force is applied. In many cases, force is applied by stretching a gating spring with stiffness κg by a certain distance. (B) Energy profiles of the channel for different applied forces as the gate moves, if the total swing is 4 nm. The lower lines indicate the force-dependent component. The energy barrier for transition between states is positioned a fraction, d, between open and closed states, so the term db determines the force sensitivity of the opening rate.

(Eq. 2)

where po is the open probability, pc is the closed probability, and kBT is the Boltzmann constant times the absolute temperature, which is 4.07 × 10−21 J at room temperature (22°C). By substituting pc = 1 – po we obtain the open probability, a time-averaged value measured in experiments:

(Eq. 3)

How is Δg determined? If a force f is exerted on the protein and the gating domain moves a distance b along the force vector, then work is done when the gate moves and the free energy is changed:

(Eq. 4)

where go and gc are the free energies of the open and closed states, and Δu is the intrinsic energy difference between states in the absence of applied force. Then

(Eq. 1.)

The open probability po is plotted as a function of f in Fig. 7A, for values of b equal to 1, 2, and 5 nm. The term Δu is arbitrarily set to 5 kBT for this plot, which makes the channel closed for zero force. It is clear that a larger gate swing means that less force is needed to create the same Δg, and so the activation curve po (f) is steeper.

Fig. 7.

Gating and stiffness of a transduction complex with channel and gating spring. (A) Open probability (Po) as a function of force (f) on a channel, for a gate swing b of 1, 2, or 5 nm. (B) Average movement of the far end of the gating spring as a function of force on the far end. Additional movement occurs when the channel opens. (C) Stiffness (K) of the channel–gating spring complex as a function of force, which is the reciprocal of the slope of the curve in (B). (D) Open probability as a function of gating spring stretch, for a gating spring stiffness of kg = 1 mN/m and gate swing of 1, 2, or 5 nm. (E) Average force on the far end of the gating spring as a function of movement of the far end. In the range where the channel opens, the force may drop because of the relaxation of the gating spring upon channel opening. (F) Stiffness of the channel–gating spring complex as a function of movement of the far end, which is the slope of the curve in (E). pN, piconewtons; xstim, stimulus displacement

Note that this is formally equivalent to the gating of voltage-dependent channels. For them, the force is the product of the electric field strength E and the net charge q on the gating region. The field is E = Vm/m, where Vm is the transmembrane potential and m is the distance across which the field drops, usually approximated as the membrane thickness. Then

(Eq. 5)

Because of uncertainties in determining both b and m, the term qb/m is collectively referred to as the equivalent gating charge. Typical voltage-gated channels have an equivalent gating charge equal to 4 to 6 electron charges per subunit (202), so that an energy difference of 1 kBT is produced by a membrane potential of ~ 5 mV.

The gating spring

What exerts the force on a mechanically gated channel? For some channels, it could be that the relevant stimulus is a pure force, directed to the channel. In others, the stimulus is a displacement of a larger structure, which is conveyed by some elastic element to the channel’s gate (Fig. 1A). If this element, termed the "gating spring," is a simple Hookean spring (which follows Hooke’s law and so has force proportional to extension) with a stiffness kg, then the energy in the spring is ½kgx2. Thus, the energies of closed and open states are

(Eq. 6)


(Eq. 7)

where uc and uo are the energies of the closed and open state in the absence of applied force, and xstim is the stimulus displacement, shown as Xs in Fig. 6. The difference between them is

(Eq. 8)

The term ½kgb2 is a constant offset energy corresponding to the gate moving against the gating spring. Its sign depends on whether the open or closed state is taken as zero, and it disappears if zero is taken as halfway between (17). Note that the energy difference between states is a linear function of the stimulus displacement, xstim.

Because the physical parameters b and kg are characteristics of molecular entities and are not easily measured, it is useful to define a gating sensitivity, zkgb. z is a characteristic of the combined channel–gating spring complex; it describes the steepness of the activation curve and has units of force. A larger z indicates a higher sensitivity; that is, less stimulus displacement xstim is required to change the energy. Note that z has been called the "gating force" and is sometimes misunderstood to be the force needed to open a channel. Instead, z is equal to the change in force in the gating spring when a channel opens. The force needed to open a channel is best defined as a force constant, f0 = kBT/b, which is the force required to change the energy difference by 1 kBT, and is independent of the gating spring.

Using the sensitivity z and defining an offset displacement x0 = Δu/z,

(Eq. 9)

A plot of po as a function of xstim is shown in Fig. 7D for a value of kg = 1 mN/m.

Stiffness of the channel–gating spring complex

Until the gating domains of single channels can be manipulated to measure displacements and forces directly, we must infer the mechanical properties of the channels from less direct methods. Although these indirect methods typically involve a population of channels stimulated through an indirect linkage, such as deflection of a stereocilia bundle, much has been learned about fundamental gating mechanisms from such measurements (17).

Consider, for example, the displacement of the far end of the gating spring when a specified force is applied to it. Under these "force-clamp" conditions, the displacement is a sum of two terms: the elastic stretch of the gating spring, and the additional movement corresponding to the opening of the channel,

(Eq. 10)

where bpo is the time average of the gate position, taking into account the direct dependence of po on f (Eq. 1). Figure 2B shows xstim as a function of f, for values of b = 1, 2, and 5 nm, and kg = 1 mN/m. An experiment of this sort can clearly be used to determine both b and kg.

It is also possible to define a stiffness kt for the transduction complex, consisting of channel and gating spring, as the reciprocal of the slope of this curve,

(Eq. 11)

This is plotted in Fig. 2C as a function of f, using Eq. 1.

A somewhat different result is obtained if the experiment is done under "length-clamp" conditions--if the end of the gating spring is moved to a specified position, and the force needed to hold it there is measured (200). Then the force can be calculated by considering the average stretch of the gating spring,

(Eq. 12)

which includes the reduction in length upon opening of the channel. The force as a function of displacement is shown in Fig. 2E, now using the dependence of po on xstim (Eq. 9). Note that Figs. 2B and 2E are not simply a swap of force and displacement axes, as two different expressions for po are used. Consequently, the stiffness calculated as a function of xstim is different:

(Eq. 13)

This is plotted as a function of xstim in Fig. 7F, using Eq. 9. Note that the stiffness under length-clamp conditions can be negative. The term po(1 – po) has a maximum value of 0.25 when po = 0.5, so the stiffness is negative when kgb2/4 > kBT, which occurs when the swing of the gate or the gating spring stiffness is large (203). Negative stiffness has been observed for hair cell transduction (204).

Gating by a network of filaments

For real cells, the area stiffness of the membrane, ka, is a sum of two components: stiffness attributable to the lipid, and stiffness attributable to the attached cytoskeleton. The relative contribution of these two components may vary among cell types. Moreover, some stretch-sensitive channels may be gated by tension in the lipid itself, whereas others are connected to filamentous proteins of the cytoskeleton. Typically, a lipid bilayer is less expandable than a network of filaments; however, if the lipid area is in excess (wrinkled) or can flow, then the cytoskeleton contributes most of the stiffness and conveys force to the channel. The relationship between the two-dimensional stress in the network and force in individual filaments must be determined. It can be derived for certain regular arrays in which polygons tile a flat surface, by considering how tension in one dimension is divided among the filaments oriented in that dimension (Fig. 8). In the case of a square array, the force in an individual filament is clearly just f = γs, where s is the length of a side and γ is lateral tension. For trigonal and hexagonal arrays, similar expressions may be derived by considering the force component along one dimension. The area of each polygon may be calculated as well. It can be seen that the force in an individual filament is roughly

(Fig. 8.)

Comparison of force and maximum channel density for cytoskeletal networks of different geometry. The micrograph (top left) shows a negatively stained spectrin network from an erythrocyte membrane (221). Such networks may be approximated by regular polygonal models. For each, the area of each polygon Ap is calculated from the length of a side s. The area per channel is shaded. If a uniform lateral tension γ is applied to an extended network, then the force in each filament can be calculated from s. For each geometry, the force is close to γ(Ap)1/2, which suggests that the same would hold for an irregular network. The channel density is calculated by assuming that there is a channel at every node, and is close to 1/s2 in each case. The stiffness of a filament in a network is calculated by using γ = –kaΔA/A in the force expression, ΔA/A ≈ 2Δs/s, and f = kgΔs. It is close to 2ka, but (perhaps surprisingly) does not depend on s.

(Eq. 14)

This general approximation can be used when a filamentous network forms an irregular polygonal array, as has been seen for spectrin networks in red blood cells (205).

With the assumption that there is at most one channel per node, the maximum channel density can be calculated; for all three geometries it is roughly 1/s2. Finally, if the area stiffness ka is a macroscopic manifestation of elastic filaments in a network, the stiffness of the individual filaments can be calculated to be kg ≈ 2ka, where linear and area stiffness have the same units. It is interesting that this result is independent of the length of a filament.

Gating by Stress in the Lipid Bilayer

Most work on mechanically gated channels has used patch-recording electrodes to apply suction to a patch and thereby generate tension in the membrane. If we consider the patch membrane as a section of a sphere, as is often the case, the lateral tension γ can be determined by Laplace’s law from the pressure across the membrane, P, and the radius of curvature of the membrane, r:

(Eq. 15)

Because it is almost impossible to measure tension in a membrane area of a few square micrometers directly, determining the radius is a particularly useful way of estimating membrane tension.

Tension in general is the energy excess per unit area created by any type of stress. A lateral two-dimensional stretch of a membrane generates expanded area. The tension in this instance is given by

(Eq. 16)

where ka is the area elasticity constant and ΔA/A is the proportional area change of the membrane. Note that the spring constants for two-dimensional materials have units of N/m, as for one-dimensional springs, but the stretch ΔA/A is a dimensionless ratio, so that lateral tension is also measured in N/m or J/m2. The elasticity constant of a typical lipid bilayer is about 0.2 N/m. Generally bilayers are considered nonstretchable because the relative expansion (ΔA/A) at a lytic (mechanical breakdown) tension does not exceed 3 to 5%.

Two other forms of stresses that pertain to biological membranes, shear stress and bending stress (Fig. 9), can also be presented as surface energy excess. Tension generated by shear extension of the membrane is expressed as

Fig. 9.

Representation of three types of stress in a membrane. Lateral two-dimensional stretch generates tension that depends on the area increase and the area elasticity constant ka. Tension generated by bending depends on the change in curvature and the bending modulus B. A shear distortion without area expansion generates tension that depends on the extension ratio and the elastic modulus of shear rigidity Φ.

(Eq. 17)

where Φ represents the elastic modulus of shear rigidity, and λr is the extension ratio, in which

(Eq. 18)

for extensions λ1 and λ2 in orthogonal directions. Shear stress pertains to cell membranes scaffolded by proteins and cytoskeleton. For two-dimensional lipid films in liquid state, the shear rigidity modulus is zero.

Tension generated by bending of the membrane is expressed as

(Eq. 19)

in which B is a bending modulus, C is the membrane curvature, and the two Rs are the principle radii of curvature (206, 207). The bending modulus for a typical lipid bilayer is on the order of 10−19 J (208).

In a real cell, the net membrane force will be determined by contributions from each form of tension and by the time course of the viscoelastic relaxations of the membrane. In general, however, biological membranes are "stiffest" to area expansion; the free energies associated with shear extension and bending are smaller, at least for the distortions normally experienced by cell membranes. We therefore focus on the contribution to free energy arising from lateral tension.

Consider a cylindrical channel with an in-plane area of a, residing in a membrane that has a specified lateral tension γ (Fig. 10). Upon opening, the channel increases in area by Δa, hence the tension produces the work of γΔa on the protein. This lowers the free energy difference between the two states, which can now be expressed as –γΔa + Δh. We use not the Gibbs but the Helmholtz free energies, because formally the system is not under constant pressure. Here, Δh is the Helmholtz energy difference between the open and closed states in the absence of tension. In the Boltzmann representation, the tension response of the channel can be written as

Fig. 10.

Two-state representations of MS channel gating by membrane tension. (A) Cartoon illustrating channel expansion upon opening. (B) Energetic profile consisting of two rectangular wells for a channel with "stiff" closed and open states. (C) A harmonic energy profile for the two states characterized by finite but different elasticities. In this particular instance, the wider parabolic well (left) corresponds to a closed state that is "softer" than the open state. The dashed line represents the harmonic profile in the presence of membrane tension γ.

(Eq. 20)

In practice, po is often measured as time-averaged mean current through a single channel or a population relative to the current at saturating tension. This equation predicts the behavior of many MS and mechano-modulated channels (11). Even channels with a number of subconducting states such as MscL (173) are reasonably well described by Eq. 20, as long as the occupancies of main substates remain independent of tension (188).

Equation 20 is based on a linear dependence of the free energy on tension and implies that the closed and open states are characterized by fixed areas at all tensions. The potential profile of such a system along the reaction coordinate (Δa) would be represented by two narrow rectangular wells (Fig. 10, middle). Accounting for elasticity of each state (harmonic approximation) invokes a profile of two parabolic wells (Fig. 10, bottom) and a quadratic term in the free energy expression (6, 13). Indeed, the rate constants of unidirectional transitions written in Eyring’s form are

(Eq. 21)


(Eq. 22)

where ko is a scaling factor, acb and aob are the area changes from the bottoms of the closed or open wells to the top of the transition barrier, and Ki is the elastic constant of the corresponding well. The equilibrium constant po/pc is equal to the ratio of the forward and backward rates, which gives the following expression for po:

(Eq. 23)

When the stiffnesses of the two conformations are equal (Kc = Ko), the quadratic term disappears and the free energy dependence becomes linear with γ. When the stiffnesses are different, the nonzero quadratic term represents the difference in elastic energies stored in each state (6). The difference in state elasticities can be assessed from the dependencies of the forward and backward rate constants on tension. The slopes of ln(kon) and ln(koff) on tension would give Δacb and Δaob and their ratio would estimate the position of the transition state (barrier) between the endpoints on the reaction coordinate. It is easy to see that the forward rate would be more sensitive to tension when the barrier is positioned closer to the center of the open well. This result was obtained for MscL, which predicts that the closed state is "softer" (less stiff) than the open state and the transition state is substantially pre-expanded relative to the closed conformation (173).

Effects of Bilayer Thickness and Curvature Stress on Gating

A conformational transition in a mechanosensory protein may change the geometry of the protein-lipid boundary. For instance, the predicted flattening of MscL upon opening may produce a mismatch with the boundary lipids around the protein, and the induced stress in the bilayer would make a substantial contribution to the free energy of the open state. The observation that MscL gating occurs at lower pipette pressures when the channel is reconstituted in shorter-chain lipids that form thinner bilayers (166) supports this view. Gating in simpler peptide channels that form conducting dimers, such as gramicidin A, also depends on membrane thickness (209), curvature stress (210), and tension (211). The effect of tension on the probability of the conducting state for gramicidin channels may depend on the thickness of the bilayer (212). Tension increases the number of conducting gramicidin dimers in thick bilayers (diecosenoyl phosphatidylcholine), whereas in thinner membranes (diolyl phosphatidylcholine) it decreases them (212). A theoretical treatment of mismatch-induced elastic deformation of the lipid bilayer near a "short" channel can be found in (213).

Curvature stress is a packing stress related more to the intrinsic composition of a membrane than to perturbations of the membrane by external force. It originates from the fact that hydrophobic interactions, which hold the membrane structure in water, pack the lipids into a flat structure irrespective of what the "spontaneous" curvature of a relaxed lamellar assembly of the same lipids defined by sizes of polar and apolar groups would be (214, 215). This intrinsic elastic stress not only predisposes to the formation of nonbilayer structures, but also may change the conformational equilibrium of membrane-embedded proteins (216). The measure of curvature stress can be quantified by the lateral pressure profile across the bilayer (217). Figure 11 depicts a channel that changes its shape inside the membrane from hourglass-like (closed) to barrel-like (open). Apparently, the equilibrium depends on the profile of lateral pressure, which is about equally distributed between the hydrocarbon (H) and polar (P) layers in "bilayer" lipids such as phosphatidylcholine (PC), but may be distorted by addition of "nonbilayer" components. There would be a higher pressure in the apolar core of the membrane made of phosphatidylethanolamine (PE) with a small head group, or in the presence of apolar inclusions such as free fatty acids (AA). Conversely, the pressure contribution of the polar region would increase in the presence of micelle-forming lipids with a bulky head and one tail (lysophosphatidylcholine, LPC), or local anesthetics such as chlorpromazine (CPZ) residing at the interface between polar and non-polar regions.

Fig. 11.

Effect of lateral pressure profile on the conformational equilibrium of a membrane protein. Each leaflet of the bilayer consists of three distinct regions: polar head groups and hydrophobic acyl chains, both exerting positive pressure, and a region in between, where interfacial tension is present (217). The protein is presumed to change its shape from hourglass-like to barrel-like. A change in the balance between lateral pressures in the polar (P) and hydrophobic (H) regions of the membrane would bias the equilibrium one way or the other. In a membrane composed predominantly of "bilayer" lipids such as phosphatidylcholine (PC), the pressures are about equal, meaning no curvature stress. Lipids with smaller hydrophilic moieties such as phosphatidylethanolamine (PE) or an admixture of unsaturated arachidonic acid (AA) would create larger pressure in the hydrophobic layer, favoring the hourglass-like conformation. The micelle-forming molecules such as lysophosphatidylcholine (LPC) or chlorpromazine (CPZ) increase pressure in the polar regions and produce an opposite effect.

It is easy to show that a bilayer perturbation that changes the pressure along the z axis by Δp(z) would produce a change in free energy difference between the states equal to the product of local protein area and pressure changes integrated over the membrane thickness (218):

(Eq. 24)

This will dictate a redistribution of the protein between the states. In the particular case depicted in the figure, the open state is favored by micelle-forming lipids, such as LPC, which create a "vacuum" in the hydrocarbon core of the membrane. This type of perturbation induces positive spontaneous curvature so that each leaflet would tend to "curl" locally toward the midplane of the bilayer. If the protein becomes not only wider in the middle but also shorter, this conformation would also be favored. Strong effects of lysolipids have been observed on the small bacterial MS channel (165) and, more recently, on MscL (166).


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.
  44. 44.
  45. 45.
  46. 46.
  47. 47.
  48. 48.
  49. 49.
  50. 50.
  51. 51.
  52. 52.
  53. 53.
  54. 54.
  55. 55.
  56. 56.
  57. 57.
  58. 58.
  59. 59.
  60. 60.
  61. 61.
  62. 62.
  63. 63.
  64. 64.
  65. 65.
  66. 66.
  67. 67.
  68. 68.
  69. 69.
  70. 70.
  71. 71.
  72. 72.
  73. 73.
  74. 74.
  75. 75.
  76. 76.
  77. 77.
  78. 78.
  79. 79.
  80. 80.
  81. 81.
  82. 82.
  83. 83.
  84. 84.
  85. 85.
  86. 86.
  87. 87.
  88. 88.
  89. 89.
  90. 90.
  91. 91.
  92. 92.
  93. 93.
  94. 94.
  95. 95.
  96. 96.
  97. 97.
  98. 98.
  99. 99.
  100. 100.
  101. 101.
  102. 102.
  103. 103.
  104. 104.
  105. 105.
  106. 106.
  107. 107.
  108. 108.
  109. 109.
  110. 110.
  111. 111.
  112. 112.
  113. 113.
  114. 114.
  115. 115.
  116. 116.
  117. 117.
  118. 118.
  119. 119.
  120. 120.
  121. 121.
  122. 122.
  123. 123.
  124. 124.
  125. 125.
  126. 126.
  127. 127.
  128. 128.
  129. 129.
  130. 130.
  131. 131.
  132. 132.
  133. 133.
  134. 134.
  135. 135.
  136. 136.
  137. 137.
  138. 138.
  139. 139.
  140. 140.
  141. 141.
  142. 142.
  143. 143.
  144. 144.
  145. 145.
  146. 146.
  147. 147.
  148. 148.
  149. 149.
  150. 150.
  151. 151.
  152. 152.
  153. 153.
  154. 154.
  155. 155.
  156. 156.
  157. 157.
  158. 158.
  159. 159.
  160. 160.
  161. 161.
  162. 162.
  163. 163.
  164. 164.
  165. 165.
  166. 166.
  167. 167.
  168. 168.
  169. 169.
  170. 170.
  171. 171.
  172. 172.
  173. 173.
  174. 174.
  175. 175.
  176. 176.
  177. 177.
  178. 178.
  179. 179.
  180. 180.
  181. 181.
  182. 182.
  183. 183.
  184. 184.
  185. 185.
  186. 186.
  187. 187.
  188. 188.
  189. 189.
  190. 190.
  191. 191.
  192. 192.
  193. 193.
  194. 194.
  195. 195.
  196. 196.
  197. 197.
  198. 198.
  199. 199.
  200. 200.
  201. 201.
  202. 202.
  203. 203.
  204. 204.
  205. 205.
  206. 206.
  207. 207.
  208. 208.
  209. 209.
  210. 210.
  211. 211.
  212. 212.
  213. 213.
  214. 214.
  215. 215.
  216. 216.
  217. 217.
  218. 218.
  219. 219.
  220. 220.
  221. 221.
  222. 222.
View Abstract

Stay Connected to Science Signaling

Navigate This Article