Mechanosensitive Channels: Multiplicity of Families and Gating Paradigms

R EVIEW

Sergei Sukharev1* and David P. Corey2* (Published 10 February 2004)

Mechanosensitive ion channels are the primary and comprise just a few GTP-binding protein (G protein)–cou- transducers that convert mechanical force into an pled receptor second-messenger pathways (1–3). Given these electrical or chemical signal in hearing, touch, and examples, should we be looking for universal mechanotransduc- other mechanical senses. Unlike vision, olfaction, tion mechanisms within these various systems and presume that and some types of taste, which all use similar kinds the primary receiver elements are essentially the same, but are of primary heterotrimeric GTP-binding protein–cou- tuned differently depending on the setting? If not, are there dif- pled receptors, mechanosensation relies on diverse ferent types of mechanoreceptors that use similar biophysical types of transducer molecules. Unrelated types of principles but dissimilar molecular mechanisms, which arose channels can be used for the perception of various independently many times in the course of evolution? The ex- mechanical stimuli, not only in distant groups of or- amples discussed below suggest that the latter scenario is more ganisms, but also in separate locations of the same likely. Although incomplete, the available molecular data sug- organism. The extreme sensitivity of the transduction gest a multiplicity of structural designs and mechanisms of mechanism in the auditory system, which relies on mechanoreceptors not only across the tree of life, but also with- an elaborate structure of rigid cilia, filamentous in single organisms. This diversity implies that careful analysis links, and molecular motors to focus force on trans- of individual systems will be required before generalizations duction channels, contrasts with that of the bacterial can be made. channel MscL, which is opened by high lateral tension in the membrane and fulfills a safety-valve The Nature of the Primary Transducer rather than a sensory function. The spatial scales of Hearing, vestibular function, and touch—senses that involve conformational movement and force in these two specialized organs, such as the ear and the cutaneous tactile re- systems are described, and are shown to be consis- ceptors, and are present in both vertebrates and invertebrates— tent with a general physical description of mechani- represent the fastest and most sensitive mechanosensory sys- cal channel gating. We outline the characteristics of tems known. In addition to these specialized organs, free so- several types of mechanosensitive channels and the matosensory endings and visceral afferents are diffusely spread functional contexts in which they participate in sig- through many tissues; many possess substantial mechanosensi- naling and cellular regulation in sensory and non- tivity, which allows them to monitor cutaneous stretch and de- sensory cells. formation, limb position, vascular and alveolar distension, and the filling of the stomach or bladder. All of these functions are Introduction attributed to the presence of mechanosensitive ion channels Mechanosensation is ubiquitous and takes on many different (MS channels) embedded in the plasma membranes of sensory forms, as cells, tissues, and entire organisms constantly receive cells (4, 5). In contrast to second messenger–mediated transduc- and generate various mechanical stimuli. Sounds captured by tion in other senses, the ion channels in mechanosensors are be- the ear and static loads on bone represent two extreme examples lieved to be directly activated by mechanical forces, for the fol- of stimuli with vastly different magnitudes and time courses. Yet lowing reasons: (i) An electrical current through the sensory organisms can gauge these stimuli separately by using different cell membrane (receptor current), a consequence of ion channel receptors and integration schemes. Inputs to the central nervous opening, appears to be the earliest recorded event following the system (CNS) from specialized mechanosensory organs initiate mechanical stimulus and it precedes other sequelae of the stim- an array of homeostatic reflexes and behaviors. At the same ulus. (ii) Many mechanosensory responses occur on a submil- time, various physiological responses to force can be observed lisecond time scale, which is faster than that permitted by a sec- at the level of single cells, demonstrating that mechanoreceptors ond-messenger cascade (6). The receptor current in the sensory are indeed ubiquitous. cell or nerve ending causes a receptor potential, which—in all Because mechanosensory phenomena are so diverse, the but the smallest animals—is converted into action potentials in progress of unraveling the underlying molecular mechanisms afferent nerve fibers passing to the CNS. In addition to special has been slower than with studies of other senses. Indeed, in the sensory organs and nerve endings, MS channels are found in mechanisms of olfaction, vision, and some types of taste, the various somatic nonsensory cells, where they are presumed to sets of molecular players appear to be evolutionarily conserved modulate the processes of contraction and secretion in response to mechanical stress, or to regulate cell volume or turgor. 1Department of Biology, University of Maryland, College Park, MD Although they open rapidly and are easy to assay, MS chan- 20742, USA. 2Howard Hughes Medical Institute and Department of nels are not the only primary receptors for mechanical stimuli. Neurobiology, Harvard Medical School, Boston, MA 02115, USA. Neurotransmitter release dependent on mechanical stress (7), *Corresponding authors. E-mail: [email protected] (S.S), slow responses to static load such as cell and tissue remodeling [email protected] (D.P.C.) (8), and gene expression and proliferative reactions (9) may in-

www.stke.org/cgi/content/full/sigtrans;2004/219/re4 Page 1 R EVIEW volve integrins and associated signaling pathways. It has been mechanistic paradigms of their gating: channels to which force proposed that in certain organs, stress-induced adenosine is conveyed by distinct cytoskeletal proteins (often in special- triphosphate (ATP) release may exert a paracrine action on ized sensory cells), and channels gated by tension in the lipid purinergic receptors in sensory fibers (10). Further studies will (best characterized in prokaryotes). clarify whether this purinergic stimulus causes afferent excitation directly or just modulates it. These related but mechanisti- Channel Gating by External Force cally separate processes will not be discussed here. Instead, we At its simplest, a MS channel is a specialized molecule—usual- focus on MS channels, which constitute a heterogeneous group ly a protein—in the cell membrane that can undergo a distortion of membrane proteins in their own right that are unified mostly in response to an external force, which either opens or closes a by function. conductive pathway through the molecule. The force can be ap- There are several systems for classifying MS channels, as plied either through cytoskeletal and extracellular matrix ele- outlined below. Conventionally, families of cloned channels are ments attached to the channel, or through the lipid bilayer itself assigned according to sequence homology. Grouping channels (11). Such a distortion can be described as a conformational according to the tissue of origin (for instance, sensory versus transition between inactive (closed) and active (open) states nonsensory) is historical and reflects two complementary ap- separated by a barrier (12, 13). In the simple case of a two-state proaches: genetic and electrophysiological. Genetic studies be- channel opened by linear force (Fig. 1A), the open probability gin with screens for mutations that affect sensory function and po is described by a Boltzmann relation: lead to identification of the chromosomal loci, genes, and even- tually 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. Elec- trophysiological surveys, in contrast, first identify conductance as a functional manifestation of the channel. This approach led (Eq. 1) to the identification of MS channel activity in neurons and also in numerous nonneural cells. Using electrophysiology, channel where f is the force acting on the channel, b is the distance it activities can be further characterized with pharmacological and moves when it opens (the swing of the gate), ∆u is an intrinsic biochemical tools that may suggest the type of molecule re- energy difference between open and closed states that biases the sponsible for the function. For example, the bacterial MS chan- probability distribution toward the closed state in the absence of −21 nel MscL was identified through a biochemical approach, using applied force, and kBT is thermal energy (~4 × 10 J at room patch-clamp as an assay technique. Genetic and electrophysio- temperature). The equation predicts a sigmoidal dependence of logical approaches complement each other and are often used in po on applied force with the midpoint (po = 0.5) at f = ∆u/b. To combination. The two-pore domain MS channels, for instance, open the channel, the product of f and b must be several times were cloned by homology to potassium channels and were iden- thermal energy; for instance, it requires a change of 6 kBT to go tified as mechanosensitive only upon expression in heterolo- from 5% to 95% open. As an example, a gate swing of 4 nm un- gous cells. If a channel is expressed in both sensory (exterocep- der a force of 1 pN produces an energy bias of about 1 kBT. tor) and nonsensory tissues, it is reasonable to assume that its A similar relation holds for channels activated by lateral ten- behavior in both settings, although it could be multimodal, is sion in the membrane (Fig. 1B). In this instance the free energy based on the same intrinsic mechanism. term (–fb + ∆u in Eq. 1) can be presented as –γ∆a + ∆u, where Without attempting an all-encompassing description, we dis- γ is tension and ∆a is the change in the in-plane area of the cuss several examples of such channels and their established or channel upon opening (13). In this case, one kBT is equivalent putative roles. We also emphasize the coexistence of at least two to an area change of 4 nm2 under a tension of 1 mN/m. An im-

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 (γ∆a).

www.stke.org/cgi/content/full/sigtrans;2004/219/re4 Page 2 R EVIEW portant 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 de- scribing 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 Fig. 2. Schematic diagram of type I mechanosensory organs, from sends a ciliated ending into the hollow bristle, where it is com- Eberl et al. (81). (A) Mechanosensory macrochaete bristle. (B) One pressed by bristle deflection. Other ciliated mechanoreceptors scolopidium of a chordotonal organ. The sensory neurons (ne) are in insects occur in specialized “chordotonal” organs located at white, supporting cells are shaded, and cuticular structures are black. the junctions between body segments, which sense flexion of Structures: bb, basal body; cd, ciliary dilation; ci, cilium; cr, ciliary root- the joints (Fig. 2). Among these are Johnston’s organ, at the let; cu, cuticle; sc, scolopale; and tb, tubular bundle. 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. 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 ar- ranged 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 Fig. 3. (A) Hair cell bundles from a vestibular organ. Each bundle is deflection of stereocilia (toward the tall stereocilia) causes shearing composed of ~60 stereocilia and one kinocilium (identified in this or- between adjacent stereocilia that increases tension on the tip links, gan by a bulbous tip). A gelatinous extracellular matrix normally ex- and this tension is apparently conveyed directly to the channels tends over the entire sensory epithelium and is linked to the tips of hair (15–17). Assuming that the tip links are the morphological correlate bundles through their kinociliary bulbs. (B) Two stereocilia and a tip of the gating springs [which may or may not be true (18)], the com- link extended between them. Deflection of the bundle to the right caus- bined stiffness of the tip links is about the same as the stiffness of es shearing of adjacent stereocilia and tightens tip links. (C) Transduc- the stereociliary pivots, so half or more of the stimulus force goes tion channels (perhaps one to three) are thought to be in the stereocili- toward gating channels. um membrane at each end of a tip link. A collection of myosin 1c Several decades of high-resolution physiology have revealed molecules maintains resting tension on the channels to bias them to a great deal about the mechanics of stimulus delivery. The stiff- the most sensitive part of their activation curve. [Modified from (219)]

www.stke.org/cgi/content/full/sigtrans;2004/219/re4 Page 3 R EVIEW ness of the hair bundle in hair cells of the bullfrog saccule, a fa- DEG/ENaC Channels vorite preparation for mechanosensory physiologists, is about 1 The DEG/ENaC channel family includes the amiloride-sensitive mN/m, so that a force of 100 pN is sufficient to deflect the bun- sodium channels of transporting epithelia (ENaCs), the nematode dle by ~100 nm. A deflection of 100 nm, which is sufficient to degenerins (DEGs), the acid-sensitive channels of vertebrate neu- open most of the channels, stretches the tip links by about 12 rons (ASICs; also known as BNCs and BNaCs), and Drosophila nm. At human auditory threshold, the bundles in the cochlea are PPKs. Proteins of this family have two transmembrane domains and deflected by only a few tenths of a nanometer and the tip links a large extracellular loop that contains two-thirds of the protein. The are stretched by a tenth of that distance! The channels undergo a second transmembrane domain and a small region immediately be- conformational change of about 2 nm when they open, so a fore it are thought to form the ion conduction pathway. The 100-nm deflection must change the tension in each tip link by DEG/ENaC channels are nonselective cation channels of low con- + + 2+ about 12 pN to open most channels (for 6 kBT). The stiffness of ductance that somewhat prefer Na over K and have little Ca per- each gating spring is thus about 1 mN/m. meability, and they are blocked in a voltage-dependent way by Although the stereocilia are superbly designed to convert a de- amiloride and its analogs. Stoichiometry is still unclear, with evi- flection force to tension on the channel proteins, they must them- dence for either four or nine subunits per channel (20–27). selves be coupled to a physical stimulus. This occurs in three MEC-4 and MEC-10 in nematode touch. Touch on the later- ways: Hair cells of the vertebrate saccule and utricle are overlaid al body wall of the nematode C. elegans is sensed by five neu- with an otolithic membrane, a secreted extracellular matrix that is rons that each send a single process along the body in close attached to each hair bundle and carries an inertial mass of calci- contact with the cuticle (28, 29). Worms touched gently on their um carbonate crystals (“otoconia”). The acceleration of the otoco- sides withdraw and move away from the stimulus. Chalfie and nia by movement or by gravity (more specifically, the component Sulston (34) searched for mutants that lacked the touch reflex of acceleration parallel to the epithelium) times the mass of otoco- but were otherwise normal, and found 18 “mec” genes that were nia equals the stimulus force delivered to the combined hair bun- needed for touch sensation. All but one are expressed by the dles of the sensory epithelium. Similarly, hair cells of the semicir- touch neurons (30). Six of these encode genes needed for the cular canals are connected to another extracellular structure, the development of the touch cells (31), but 12 encode structural cupula, which forms a septum across the fluid-filled tube of each proteins that may constitute a mechanotransduction apparatus semicircular canal. Fluid acceleration caused by head rotation ex- (Fig. 4) [see (32) for a review]. Two structural proteins, MEC-4 erts force on the cupula and its attached hair cells. In the cochlea, and MEC-10, are transmembrane subunits of an ion channel; each cycle of sound causes the basilar membrane, a flexible mem- two others, MEC-2 and MEC-6, are apparently accessory sub- brane stretched between two fluid-filled cham- bers 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 Fig. 4. Possible arrangement of nematode MEC proteins in a mechanotransducing channels of transporting epithelia and the de- apparatus. Microtubules formed by MEC-7 and MEC-12 come close to the intracel- generins (DEG/ENaC channels). Both families lular side of the membrane in the mechanosensitive process of a touch neuron. are also found in vertebrates, and both contain MEC-2 might connect them to an ion channel made up of MEC-4, MEC-10, and other ion channels implicated in transduction MEC-6. The large extracellular domains of the channel-forming MECs may link to for a number of sensory modalities. extracellular matrix proteins that can pull on them. [Modified from (220)]

www.stke.org/cgi/content/full/sigtrans;2004/219/re4 Page 4 R EVIEW units of the channel (33). Other MEC proteins include tubulins ASIC3 (DRASIC) is expressed in most sensory neurons, in- (MEC-7 and MEC-12) and matrix proteins (MEC-1, 5, and 9) cluding both those with large and those with small diameters. that could convey force to a channel. Certain mutations in The protein is located in many of the same sensory endings in MEC-4 and MEC-10, as well as in the related nematode chan- skin as ASIC2a, and also in nociceptive intraepidermal endings nels DEG-1 and UNC-105, bias the channels open and cause (38). The physiological effects of gene deletion are mixed; degeneration of the cells that express them (34, 35); hence, this rapidly adapting mechanoreceptors become more sensitive, but group of channels was termed the degenerins. When cloned, the mechanical nociceptors lose sensitivity. Heat and acid nocicep- vertebrate ENaCs (20) were found to be homologs of the de- tion are affected as well (38). generins; together, the two groups defined the DEG/ENaC The striking localization of ASIC channel proteins in senso- channel family. Worms and flies have apparently found ry endings and the significant (although mixed) results of gene DEG/ENaC channels particularly useful, expressing 28 and 25 deletion suggest a role in cutaneous mechanosensation. Much genes of this family, respectively, whereas humans have 9 (29). work remains to determine whether the ASICs are transducers Evidence that MEC-4 and MEC-10 form a mechanically or play a supporting role, and to sort out the functions of alter- gated channel comes largely by elimination. A screen for touch- native splice forms and heteromultimeric subunit assembly. insensitive mutants was saturated, producing a number of alleles ENaC channels in vertebrate baroreception and touch. of these two genes, but no other ion channel was found. A Amiloride-sensitive epithelial sodium channels (ENaCs) have MEC-4/MEC-10 heteromeric channel has been expressed in been extensively studied in transporting epithelia such as lung, oocytes, but has not yet been shown to be mechanosensitive kidney, and colon. They are located in the apical membranes of (33). Similarly, receptor currents have not been recorded in the these epithelia, and are constitutively open. Conductance of the touch neurons that express these channels. Nonetheless, genetic ENaCs is regulated by hormonal signals (aldosterone, vaso- identification of the MECs has produced the most complete pressin, insulin) that regulate the rapid insertion and removal of cast of candidates in any organism, for what is likely to be a the channel from the membrane (39). The ENaC channel in mechanosensory complex. transporting epithelia is composed of three similar subunits: ASIC channels in vertebrate cutaneous touch. A third branch αENaC, βENaC, and γENaC (21). Most vertebrate genomes al- of the DEG/ENaC family was defined by the discovery, by so have a fourth: δENaC. It is not surprising, therefore, that the three groups, of a channel expressed in neurons of brain and ENaC channel has been implicated in salt taste in the tongue dorsal root ganglia. It was named MDEG, BNC1, and BNaC1 [(40) and references therein]: If channels in apical membranes (22, 23), beginning an unfortunate tradition of multiple names of taste receptor cells are also open, the depolarizing inward for this branch of the family that was only compounded by al- Na+ flux should depend on Na+ concentration at the outer sur- ternative splicing (Table 1). The first identified splice form is face of the taste bud. now known as ASIC2a. It was more surprising, however, to find ENaC channel sub- ASIC2a is expressed by many large-diameter neurons of the dor- units in arterial baroreceptor neurons. Baroreceptor cell bodies sal root ganglia, which are largely mechanosensory. The protein is are located in the nodose ganglia and send axons to innervate not transported to the spinal cord, but is transported along the nerves the carotid sinus and aortic arch. An increase in arterial pressure to their distal terminals in skin (36). Antibodies to ASIC2 label ter- stretches the vessel wall, which increases firing of the nerve. minals associate with a number of mechanosensory structures in Drummond et al. (41) found that nodose ganglia express skin, including Merkel cells, Meissner corpuscles, and hair follicles. βENaC and γENaC, but not αENaC. The absence of αENaC is Some free nerve endings are labeled, but unmyelinated intraepider- reassuring, because ENaC channels are only constitutively open mal endings mostly are not (36, 37). A targeted deletion of the if this subunit is present (21), and a mechanoreceptor channel ASIC2 gene reduced the sensitivity of low-threshold mechanorecep- might be expected to be closed in the absence of a stimulus. An- tors, suggesting a role in light touch sensation, although it did not tibodies to γENaC specifically labeled the sensory endings of eliminate the response altogether (37). nodose ganglion neurons in the aortic arch (41), and also labeled

Gene Splice form 1 Other names Alternative Other names (human) splice forms ACCN1 ASIC2a MDEG, BNC1a, BNaC1α ASIC2b MDEG2, BNC1b, BNaC1β

ACCN2 ASIC1a BNaC2, ASIC, ASICα ASIC1b ASICβ

ACCN3 ASIC3a DRASIC, SLNAC1, TNaC1 ASIC3b, ASIC3c

ACCN4 ASIC4a BNaC4, SPASIC, ASIC4b

ACCN5 ASIC5a BLINaC, INAC

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.

www.stke.org/cgi/content/full/sigtrans;2004/219/re4 Page 5 R EVIEW sensory endings in carotid sinus. In vivo experiments on carotid si- shorter, with the conventional six transmembrane domains. nus showed that the amiloride analog benzamil in the lumen PKD1 and PKD2 are both widely expressed, but their function blocked the pressure-induced firing of the carotid sinus nerve. has been studied most thoroughly in kidney. PKD1 and PKD2 Similarly, βENaC and γENaC, but not αENaC, are ex- interact through coiled-coil domains in their C termini (50) and pressed in dorsal root ganglion neurons. Antibodies to both form a nonselective cation channel when coexpressed in Chi- βENaC and γENaC labeled sensory endings in the skin that in- nese hamster ovary (CHO) cells (51). Within kidney cells, the nervate Merkel cells, Meissner corpuscles, and small lamellated two proteins are located most notably in the single, nonmotile corpuscles (42). cilium that protrudes from the lumenal surface (52, 53). Bend- Location in sensory endings does not define function, and ing of this cilium by fluid flow increases intracellular Ca2+ in amiloride-like drugs are not terribly specific, so the evidence for single Madin-Darby canine kidney (MDCK) cells (54). Kidney ENaCs as a MS channel subunit is largely circumstantial. But the cells lacking cilia, or lacking PKD1, do not show the response apparent colocalization in dorsal root ganglion (DRG) endings of to fluid flow, nor do cells treated with a blocking antibody to ENaCs with ASICs, and the growing involvement of the PKD2 (55). PKD1 and PKD2 together may form a MS channel, DEG/ENaC channels with mechanoreceptors in many species, are or PKD1 may activate a channel formed by PKD2. PKD1 and certainly consistent with a mechanoreceptor function. PKD2 are also necessary for normal function in pancreas, blood PPK channels in Drosophila sensory neurons. The vessels, and heart, and for left-right axis determination in em- Drosophila genome contains at least 21 genes of the bryos, and it will be interesting to see whether they function as DEG/ENaC family; these channels constitute a branch separate mechanoreceptors in any or all these tissues. from both the MEC branch in worms and the ASIC branch in The nematode C. elegans has a PKD1 homolog known as vertebrates (29). The first such channel identified in flies, LOV-1 (location of vulva) and a PKD2 homolog called PKD-2. PPK1, is expressed in a subset of mechanosensory neurons Both are located in ciliated mechanosensory neurons that inner- (43). Insects have two general types of mechanosensory neu- vate the head, the male copulatory spicules, and the male senso- rons: Type I neurons innervate external sensory organs such as ry rays at the tail (56, 57). Both LOV-1 and PKD-2 are required bristles and chordotonal organs such as the sound-sensing John- for the ability of males to locate the hermaphrodite vulva during ston’s organ; they are bipolar, with cell bodies near the end or- mating, although their absence causes no apparent morphologi- gan, and ciliated, with mechanotransduction occurring in the cal defect and no difficulty in movement, consistent with a di- cilium. Type II neurons have cell bodies closer to the neuraxis rect role in mechanosensation (56–58). and have extensively arborizing dendrites tiling the inner epi- Yvc1p in yeast volume regulation. The yeast genome has on- dermal space. PPK is in a subset of multiple dendritic neurons ly one TRP-like channel gene, YVC1P, which encodes a distant with arborizing dendrites, and is especially located at varicosi- member of the TRP superfamily. The Yvc1p channel is located ties along dendrites that are thought to be the site of mechan- at the limiting membrane of intracellular vacuoles, and is acti- otransduction (43). A PPK null mutant exhibits an altered pat- vated by hypertonic stimuli delivered to whole yeast (59, 60). tern of locomotion (fewer stops and turns) that may be consis- Patch-clamp recording from isolated vacuoles revealed a 400- tent with diminished proprioception (44). As with ASICs, local- pS nonselective cation channel, which is absent in yeast that ization of PPK suggests that it may be a MS channel, but much lack the YVC1P gene (59). The channel is activated by in- physiological research remains to demonstrate function. creased cytoplasmic Ca2+ and is also activated by increasing hy- drostatic pressure across the vacuolar membrane [in the range TRP Channels of 5 to 50 mm Hg (61)]. The channel is thought to act in yeast The founding member of the TRP ion channel family was iso- volume regulation by allowing Ca2+ efflux from vacuoles, lated as the product of a gene defective in Drosophila vision which further activates the channels; the consequent rise in cy- (45). Unlike wild-type flies, trp flies have a transient receptor toplasm Ca2+ may activate compensatory salt transport across potential that decays in 1 to 2 s, even with sustained light (46). the plasma membrane. AlthoughYvc1p is clearly an ion chan- The trp protein is an ion channel with six predicted transmem- nel, it will be important to determine whether it is the pressure brane domains and a pore loop between the fifth and sixth sensor as well. transmembrane domains (47), an architecture similar to that of NOMPC in Drosophila bristle receptors and zebrafish later- a large group of voltage-gated K+ channels and cyclic nu- al line. To find proteins needed for mechanotransduction in cleotide–gated channels. Since then, a large number of TRP Drosophila, Kernan et al. (62) screened mutagenized larvae for channels have been identified in various animal species; there defects in a withdrawal response to touch. Mutations in five are 33 TRP genes, for instance, in the mouse genome (48). Al- genes caused specific defects in bristle mechanotransduction, as though some are expressed by CNS neurons, many appear in assessed by recording extracellular receptor potentials through sensory receptor cells, including cells that mediate vision, the hollow bristle. Several of these nomp genes (no mechanore- pheromone sensation, thermal sensation, taste, and touch. TRP ceptor potential) have been cloned; one of them, nompC, en- channels also occur in several transporting epithelia, where they codes an ion channel of the TRP superfamily (63). nompC is a may mediate cell volume regulation. particularly long TRP channel, with 29 ankyrin domains in an PKD channels in vertebrate kidney and nematode sex organs. extended N terminus. nompC also has up to eight predicted Polycystic kidney disease (PKD) is an inherited disorder caused transmembrane domains, so it is not clear whether the N termi- most often by mutations in two members of the TRP superfami- nus is intracellular or extracellular. Three of the four mutant ally, PKD1 and PKD2 (49). PKD1 [also called polycystin 1 leles of nompC have stop mutations in the N terminus that (PC1)] is a large protein, with 10 to 12 transmembrane domains would certainly act as nulls, but the fourth is a missense muta- and a long extracellular N terminus that contains various bind- tion between the third and fourth transmembrane domains. Flies ing domains. PKD2 [also called polycystin 2 (PC2)] is much carrying this allele showed more rapid adaptation to sustained

www.stke.org/cgi/content/full/sigtrans;2004/219/re4 Page 6 R EVIEW bristle deflections than did wild-type flies, a rather subtle phe- whether TRPV4 is directly or indirectly activated by mechanical notype that is the best evidence that nompC is in fact the trans- stimuli remains unclear. duction channel itself. nompC is expressed in cells within the Osm-9 and OCRs: multimodal receptors in nematode. The bristle organ, but it has not yet been localized to the sensory osm-9 gene was first identified in a screen for mutants defective neuron itself, nor has the protein been localized within cells. in olfaction in C. elegans (76, 77). Like many TRP channels, Walker et al. (63) have identified a nompC ortholog in nema- OSM-9 has three N-terminal ankyrin domains. These might me- tode. In transgenic C. elegans expressing a GFP gene driven by diate association of OSM-9 with the cytoskeleton to convey the nematode nompC promoter, fluorescence was observed in force, or they may simply bind other proteins to mediate precise ciliated mechanosensitive neurons. channel localization within sensory neurons. More extensive There is also a nompC ortholog in the zebrafish genome (64, characterization of the gene product revealed that it is expressed 65). Sidi et al. found that the zebrafish nompC is expressed in in various sensory neurons that mediate olfaction, osmosensa- hair cells of the otic vesicle, and that zebrafish embryos inject- tion, and mechanosensation, and that osm-9 mutants are defec- ed with morpholino antisense oligonucleotides were deaf (as as- tive in osmosensation and in certain types of olfaction and sessed by an acoustic startle reflex) and exhibited circular mechanosensation (78). Nematodes touched lightly on the nose swimming behavior consistent with a vestibular defect (65). stop moving forward and start backing, a behavior mediated by Fish also have hair cells in their lateral line organs, structures the ASH, FLP, and OLQ cells, which are ciliated neurons that comprising clusters of 10 to 20 hair cells in the skin whose send processes to the nose (79). Mutants of osm-9 are defective stereocilia are exposed to the water and which detect water cur- in response to light nose touch, and the ASH, FLP, and OLQ rents. Although nompC transcript was not detected in lateral neurons express osm-9 (78). line hair cells by in situ hybridization, morpholino-injected ze- The osm-9 gene encodes a TRP channel that is part of the in- brafish failed to accumulate the dye FM1-43 (65), a fluorescent vertebrate branch of the TRPV family. The nematode genome marker of functional transduction that enters cells through also contains four other genes in this branch, which have been mechanosensitive transduction channels (66). Nor did they named ocr-1 to ocr-4, for osm-9/capsaicin-receptor related (80). show a stimulus-evoked “microphonic” potential, an extracellu- ocr genes are expressed in a partially overlapping set of sensory lar potential that is a simple physiological measure of functional neurons in nematodes, but osm-9 is also expressed in any cell channels (65). Thus, zebrafish nompC is required for hair cell that expresses an ocr gene. For instance, both osm-9 and ocr-2 transduction, and its similarity to fly nompC is further evidence are expressed in ASH neurons, which can be activated by that this channel constitutes at least part of the transduction mechanosensory and other stimuli, and mutation of ocr-2 abol- channel for zebrafish hair cells. On the other hand, there is not a ishes sensitivity of ASH neurons to nose touch (as well as other nompC ortholog in mouse or human genomes (67), so mam- stimuli). Both osm-9 and ocr-4 are expressed in the malian hair cells must use something else. mechanosensory OLQ neurons (80). TRPV4: An osmosensor and more. A search for vertebrate TRPV channels in Drosophila hearing. The Drosophila nom- homologs of Osm-9 and TRPV1 led several groups to clone the pC is a strong candidate for a bona fide mechanically gated mammalian TRPV4 channel [also called OTRPC4, VR-OAC channel, and so it was gratifying to find it expressed in chordo- (vanilloid receptor osmotically activated channel), and VRL-2 tonal organs like Johnston’s organ, which is the fly auditory or- (vanilloid receptor–like 2)] (68–71). This channel, like other gan (63). Whereas other nomp genes are also needed for fly TRPs, is a nonselective cation channel with significant Ca2+ hearing, however, nompC mutants have only partially compro- permeability. The sequence similarity to Osm-9 suggested that mised hearing (81). Two other TRP channels have now been TRPV4 may also have osmotic sensitivity; indeed, when ex- identified as required for auditory sensation by Johnston’s or- pressed in cultured cells, it was found to potentiate Ca2+ entry gan; both are in the invertebrate branch of the TRPV family. in response to hypo-osmotic stimuli. A decrease in osmolarity The channel encoded by the inactive gene (CG4536) is most of 10% reliably evokes a signal from an intracellular Ca2+- closely related to the nematode OSM-9, whereas that encoded indicator dye (68). TRPV4 is widely expressed, appearing in by nanchung (CG5842) is an ortholog of the nematode OCR distal convoluted tubule of kidney, liver, heart, vascular en- channels (82, 83). Just as OSM-9 interacts with OCR-2 to me- dothelial cells, testis, spleen, salivary gland, lung, and trachea, diate mechanosensation in nematodes, perhaps inactive and and in the stria vascularis of the cochlea (68–71). Its appearance nanchung together form a MS channel that mediates fly hear- in many transporting epithelia, which must balance large apical ing. Neither gene has direct orthologs in vertebrates. and basal fluxes of salts and water, suggests a role in cell vol- Painless: A TRPA channel in Drosophila nociception. In ad- ume regulation. Within the nervous system, TRPV4 is found in dition to the TRPC, TRPM, TRPV, PKD, and NOMPC branches osmosensory cells of circumventricular organs in brain and in of the TRP superfamily, there is a sixth branch (TRPA) that has the Merkel cells surrounding hair follicles (69), so that it may been largely neglected by scientists (and by evolving verte- have a sensory role as well. Indeed, mice lacking TRPV4 have brates). Human ANKTM1, the founding member of the TRPA reduced sensation to noxious tail pressure (72) and slower re- family, is the single mammalian member of this group, but ne- covery of normal serum osmolarity after salt loading (73). matodes have two TRPA channels, Drosophila have four, Understanding the function of TRPV4 has become more mosquitoes have seven, and the sea squirt Ciona has six (48). complicated (74), because it is also activated by heating to 25° One of the Drosophila TRPA channels, encoded by the painless to 30°C, but then inactivates rapidly when heated above ~35°C gene, is necessary for both thermal and mechanical nociception (75). Finally, it can be activated by a phorbol ester (4α-phorbol in fly larvae (84). Multidendritic type II (nonciliated) sensory 12,13-didecanoate) that does not activate protein kinase C neurons express painless and fire when the temperature exceeds (PKC) (74). Responses to all of these stimuli are relatively slow 38°C, but not if painless is disrupted. Similarly, normal larvae (seconds), perhaps because stimulus delivery is slow, but exhibit a withdrawal reflex to strong mechanical stimuli that is

www.stke.org/cgi/content/full/sigtrans;2004/219/re4 Page 7 R EVIEW lost in painless mutants, whereas responses to light touch are and its structure has recently been solved using nuclear magnet- normal in the mutant. It may be that painless is a MS channel ic resonance (NMR) (107). The very recent cloning of SAKCa, a that is also temperature sensitive, or it may be indirectly activat- stretch-activated splice-variant of the BK Ca2+-dependent potas- ed by tissue damage produced by either thermal or mechanical sium channel (108), should help dissect the molecular mecha- stimuli. Similarly, the mouse ANKTM1 is activated by very nisms of cardiac mechanoelectrical feedback. cold temperatures, but is expressed in a small number of noci- The mechanism underlying the massive fibrillation that fol- ceptive DRG neurons that also express heat-activated TRPV lows a sudden impact to the chest known as commotio cordis channels and may be multimodal nociceptors (85). implicates both types of cardiac MS channels mentioned above + (96). The likely candidate for the K -selective species, the KAT P Mechanosensitive Channel Activities in Nonsensory channel, is substantially modulated by membrane stretch (109). Cells Glibenclamide, which blocks KAT P , reduces the probability of Channels sensitive to local pressure gradients across the patch post-impact ventricular fibrillation, possibly by suppressing im- membrane have been recorded in many somatic nonsensory pact-produced repolarization, which shortens the cell refractory cells (86–88). Phenomenologically, channels observed in such period in the zone of impact. Such a desynchronized tissue, in experiments can be divided into stretch-activated (SA) or terms of refractoriness, is predisposed to ectopic contractions, stretch-inactivated (SI) channels and can be further classified especially when the impact arrives during the period of repolar- by ion selectivity and sensitivity to Ca2+, voltage, etc. Many of ization (96). The early depolarizations attributed to the cation- these channel molecules have not been identified to date and selective (depolarizing) current trigger ectopic excitation, which are known only through characteristic currents that are general- poses a high risk of evolving into a ventricular fibrillation. ly blockable by Gd3+ ions (89). Despite the scarcity of molecu- Cardiac stretch-activated channels are also proposed to me- lar information, such channels are clearly involved in intracellu- diate the release of the atrial natriuretic peptide (ANP) in re- lar signaling and, in some instances, in the development of sponse to atrial distension (100). The complete mechanism of pathologies. ANP secretion has not been elucidated, but activation of KAT P Skeletal muscle. SA channel activities were first documented channels have been implicated, as the stretch-dependent compo- in embryonic chick skeletal muscle cells (90). The 70-pS cur- nent of ANP release is blocked by glibenclamide (110). rents were moderately K+-selective and were more easily acti- Smooth muscles. Smooth muscles are capable of autoregula- vated after treatment of cells with cytochalasin B, suggesting a tion and actively contract in response to stretch. This feedback redistribution of the stress from the disrupted actin cytoskeleton reaction, known as myogenic reflex, is believed to be initiated to another stress-conveying component coupled to the gating by stretch-activated channels. Early studies (111) have shown machinery. Subsequent studies of dystrophic (mdx) muscles that stretch channels in the urinary bladder muscles are abun- with a compromised membrane-cytoskeleton attachment re- dant, indiscriminately conduct cations (including Ca2+), and are vealed that stretch channels in dystrophic myoblasts, myotubes, more active when hyperpolarized. Longitudinal stretch of a cell and isolated fibers are constitutively more active than in increases the channel activity proportionally to the elongation nondystrophic muscles (91). The channels in mdx animals also (112). Recent studies using patch-clamp combined with Ca2+ exhibit irreversible mechanically induced switching to a deregu- imaging have shown that Ca2+ influx through single stretch-acti- lated open state, making such muscles prone to continuous Ca2+ vated channels is sufficient to activate large Ca2+-activated K+ poisoning (92). Recordings made in muscles from animals defi- channels, trigger massive Ca2+ release from intracellular stores cient in delta-sarcoglycan, a part of dystrophin-glycoprotein (113, 114), and in turn, activate phospholipase C (PLC) (115). complex, led to similar conclusions (93). The SA channels in Endothelium. The endothelial lining, like vascular smooth skeletal muscle have yet to be cloned and their normal physio- muscle, participates in regulation of vessel diameter. Endothe- logical role remains to be clarified. lial cells respond to pressure, distension, and fluid shear stress The heart. In the heart, MS channels are involved in feed- by releasing vasodilators such as nitric oxide (NO) and prosta- back processes coupled to the mechanics of the cardiac cycle. cyclin, most likely in response to intracellular Ca2+ mobilization Mechanoelectrical feedback normally accelerates the heart in (116). Stretch-activated Ca2+-permeable channels have been re- response to excessive distension of the muscle (94). However, ported in cell preparations from the umbilical vein, aorta (117), under certain circumstances it may also cause arrhythmias that and mesenteric arteries (118). When cells cultured on an elastic arise from an infarction zone (95), or deadly fibrillation trig- silicon membrane are subjected to stretch, extracellular Ca2+ en- gered by a sudden blow to the chest (96). Patch-clamp examina- ters cells not through voltage-gated Ca2+ channels but through tion of cardiomyocytes from snail, frog, chicken, mouse, rat, stretch-activated channels (119). The calcium entry leads to ex- guinea pig, and human shows a range of stretch-activated cur- tensive rearrangement of stress fibers and focal adhesion zones rents [see references in (87, 97)]; the most frequently observed involving c-src and focal adhesion kinases (120). A peculiar are a 90- to 100-pS K+-selective current (98) and a 20- to 25-pS channel, responsive to positive pressure and blockable by suc- nonselective cationic current (Erev ≈ –20 mV), which carries tion in the patch pipette, is found in intact endocardial endothe- Ca2+ (99,100). These currents are activated by suction in a patch lium (121). Substantial Ca2+ permeability and the up-regulated pipette (101), by stretching cells (102), or by applying pressure state of endothelial MS channels in hypertensive animals points to the cell surface with a blunt glass probe (103–105). Channel to their possible role in counterregulation of elevated blood opening permits sufficient Ca2+ influx to initiate Ca2+-induced pressure (118). Ca2+ release from the sarcoplasmic reticulum and subsequent Fluid shear stress acting on endothelial or epithelial cells contractions (99). Cardiac MS channels are blocked by venom may exert a different type of deformation compared to tensile from the spider Grammostola spatulata (106). The 28–amino stress caused by vascular wall distension. The involvement of acid active peptide isolated from the venom is antiarrhythmic SA channels in shear-stress responses has been questioned on

www.stke.org/cgi/content/full/sigtrans;2004/219/re4 Page 8 R EVIEW the basis of the absence of Gd3+ inhibition of stress fiber re- cal or osmotic stress, intracellular pH, or temperature. Phospho- modeling in bovine aortic endothelium (122). Nonetheless, SA rylation by intracellular protein kinases in response to increased channels are up-regulated after stimulation by shear stress in levels of cyclic adenosine monophosphate inhibits the channels endothelial culture and are present at higher density in endothe- (131). These leakage channels are truly mechanosensitive: They lia of spontaneously hypertensive animals (123). In kidney ep- increase their po by orders of magnitude in response to mem- ithelium, as mentioned above, fluid flow sensation is ascribed to brane stretch applied through the patch pipette. TREK-1 is also the primary cilium present on the luminal surface of principal modulated by osmotic cell swelling (132). Besides direct mem- cells. PKD2, a high-conductance Ca2+-permeable channel from brane stretch, TREK-1 is readily activated by elevated levels of the TRP family, has been recently localized to the cilium togeth- polyunsaturated fatty acids, lysophospholipids, or volatile anes- er with its PKD1 counterpart (55) (see above). thetics. The sensitivity of TREK-1 to stretch, anesthetics, and Volume regulation. Volume regulation under changing os- intracellular acidosis requires an intact C-terminal domain (133, motic conditions is an intrinsic property of all cells. It involves 134). The pharmacological and functional properties of TREK mechanosensation in that swelling or shrinkage puts strain on channels closely resemble those of S-channels in Aplysia senso- the membrane and intracellular structures. Two types of cationic ry neurons (135), which are responsible for serotonergic modu- MS channels have been shown to be involved in volume regula- lation and plasticity. It is presumed that the neuroprotective role tion for the renal proximal tubule epithelium (124). The phe- of TREK and TRAAK channels in mammalian neurons is hy- nomenon of regulatory volume decrease (RVD) following hypo- perpolarization and prevention of Ca2+ poisoning in the event of tonic swelling was ascribed to a coordinated action of volume- ischemic brain damage, which is typically accompanied by aci- sensitive anionic (ClC-2) and cationic (maxi K) channels, as dosis and osmotic cell swelling. well as osmolyte transporters (125–127). None of the anion channels are directly sensitive to membrane stretch, which sug- What Happens Under the Patch Pipette? gests that RVD is not triggered directly by membrane tension. It The questions of how stretch channels gate and whether they has been inferred that volume-sensitive TRPV channels (see obligatorily require the cytoskeleton for their function have been above) activate first, providing Ca2+ influx, followed by sec- discussed in reviews by Sachs and Morris (87) and by Hamill and ondary activation of maxi K and other channels (69). Martinac (88). It remains possible that in some instances, stress is Osmosensation is also critical for whole organisms in the conveyed to the sensory molecule through the network of sub- homeostatic maintenance of the fluid and electrolyte balance, and membrane filaments (discussed in the Appendix). The frequent it may use mechanisms similar to those in regulatory cell volume observation that pretreatment of cells by cytochalasins makes the changes. One of the genes needed for osmotic avoidance in C. ele- channels more sensitive to stretch suggests, however, that the cy- gans, osm-9, encodes a TRP channel (see above), which is also in- toskeleton is not a force-transmitting element but rather a re- volved in mechanosensation by ciliated touch neurons in the nose straining element. The MS channels that become more active of the worm and in olfaction (78). A mammalian homolog of when the membrane is partially liberated from the cytoskeletal OSM-9, the hypotonically activated TRPV4, is localized in brain, scaffold or completely decoupled from it are likely gated directly sensory neurons, ear, lung, and kidney, which suggests that this by lateral tension in the lipid bilayer. Consistently, lysolipids, fat- mammalian channel mediates broad tissue responses to changes in ty acids, and anesthetics, which increase membrane area and in- systemic osmolarity (69). Because the deletion of ankyrin repeats duce curvature stress in the lipid bilayer, sensitize some types of providing cytoskeletal attachments to the channel did not abolish stretch channels to mechanical stimuli (98, 134, 136). With a few hypotonic activation, it has been proposed that the channel must exceptions (132, 137), MS channels can be activated either by be gated directly by membrane stretch; however, no other evidence suction or by positive pipette pressure in inside-out patches; this has been presented. illustrates that lateral tension, not bending in a specific direction, By regulating drinking behavior and water reabsorption in the is the key parameter. In many systems, it was also possible to kidney, the hypothalamus imposes central control over the osmotic find correspondence between the properties of single-channel balance in the entire body. The magnocellular neurosecretory cells, currents activated by suction in cell-attached patches and whole- located in the paraventricular and supraoptic nuclei in the hypotha- cell currents evoked by positive pressure (6, 87, 88). Two excep- lamus, are intrinsically sensitive to the osmolality of extracellular tions, where whole-cell mechanosensitive currents are practically fluid (128). When the osmolality increases above 295 mosm, the impossible to evoke, are Aplysia neurons (138) and Xenopus magnocellular cells depolarize and release vasopressin and oxytocin oocytes (139). Both types of cells are characterized by highly de- from branched neurosecretory terminals that project to the posterior veloped microvilli and dense networks of cortical cytoskeleton. pituitary. This response is directly related to the volume decrease in Unwrinkling of microvilli and partial detachment of the mem- the bodies of magnocellular neurons, which behave almost as ideal brane from the cytoskeleton by strong suction under tight seal osmometers (129). The accompanying depolarization is ascribed to conditions alters MS channel behavior in these cells (140, 141). stretch-inactivated cationic channels that activate upon cell shrink- An interesting phenomenon of changes of patch geometry pro- age. The putative channels have a conductance of 32 pS and a ratio duced by depolarizing voltages, accompanied by activation of of K+ to Na+ permeabilities of 5:1 (130). The hypothalamic os- MS channels, has been reported in Xenopus oocytes (142). Mea- mosensory channel has not yet been identified. surable movements of patch membrane were observed in re- Two-pore domain potassium channels. TREK and TRAAK sponse to prolonged depolarizations (exclusively with borosili- represent a group of four–transmembrane domain channels cate, but not soft glass pipettes). The questions of how profoundly found in high density in several areas of the CNS (131). These the structure of a patch membrane is perturbed by the tight seal regulated “leakage” channels are believed to be responsible not and whether channel activities produced by stretch or voltage un- only for maintaining the resting potential in neurons, but also der such conditions are physiologically relevant must be ad- for regulating it in response to different cues such as mechani- dressed for each particular case.

www.stke.org/cgi/content/full/sigtrans;2004/219/re4 Page 9 R EVIEW

Why Do Walled Cells Have Mechanosensitive Channels? The successful crystallographic work done in the laboratory of At first glance, it seems puzzling that organisms with strong ex- Rees (167, 168) makes MscL and MscS good models for basic tracellular walls would need MS channels. Indeed, the static studies of molecular mechanisms of channel gating by stretch. shape of a plant is maintained in part by turgor pressure that the According to the crystal structure, MscS is a homoheptamer cytoplasm exerts on the cell wall. As long as cell expansion is of three–transmembrane domain subunits (168). C-terminal restrained, the plasma membrane, typically present in excess, segments of each subunit contribute to a large cytoplasmic do- experiences no lateral tension (143). But in growing plants, the main that apparently acts as a molecular sieve. The transmem- cell wall undergoes constant remodeling and, in certain areas, brane domains (TM1 and TM2) bear positive charges and are very fast expansion (144), and, therefore, new membrane must oriented at about a 30° angle to the pore axis. The structure be inserted at the same pace. Bacterial cell walls are also disten- shows that the MscS pore, formed by the largely hydrophobic sible (145, 146), and when swelling under strong osmotic shock third transmembrane domain (TM3), is open. However, it is un- occurs, they may not guarantee complete mechanical protection clear whether the size of the pore and the character of the lining to the plasma membrane. Indeed, turgor-operated osmolyte ef- can support the 1-nS conductance observed in experiments. flux systems have evolved in many unicellular organisms (147). More studies are required to verify whether the crystal structure MS channels would perfectly fit the role of membrane tension represents the closed, open, or some other (desensitized) state, reporters that provide necessary signaling feedback, as well as and what the gating transition looks like. We next discuss the simple permeation pathways for osmolytes. MS channel activity gating mechanism of the large MS channel MscL, which at this has been documented in plant protoplasts (148, 149) and vac- time is better understood than that of MscS. uoles (150), fungi (151), yeast (152), yeast vacuoles (61), archaea (153). and bacteria (154, 155). MscL, a Simple Model Besides turgor regulation, MS channels in plants may partic- MscL has been identified as a 136-residue protein that forms a ipate in gravitropism (149), which makes roots grow downward large stretch-activated channel upon reconstitution into and shoots grow upward. Mechanistically, gravitropic responses asolectin liposomes (157). The protein has been localized to the are initiated by intracellular sedimentation of dense starch gran- inner membrane of E. coli and is predicted to cross the mem- ules (amyloplasts) in stems and root caps and are ultimately ef- brane twice, with both the N and C termini in the cytoplasm fected by a directional transport of the growth hormone auxin (169). Deletions show that the short N- terminal segment is crit- (156). The intermediate events in this signaling cascade, which ical for channel function or assembly, whereas the absence of may involve MS channels, have not yet been elucidated. 25 C-terminal residues could be tolerated (170, 171). The extremely large (3.2 nS) nonselective conductance of the fully Bacterial Mechanosensitive Channels Are Osmotic open state suggests a water-filled pore 30 to 40 Å in diameter Safety Valves (172, 173). The midpoint tension of channel activation is 10.4 The osmolyte efflux system in Escherichia coli comprises at dynes/cm in spheroplasts (174) and 11.8 dynes/cm in large lipo- least three known inner membrane tension-operated channels of some patches (173), respectively, which is close to the lytic lim- high conductance, namely MscL (157), MscS [formerly known it for many bilayers. as YggB (158)], and MscK [formerly KefA (159)]. Homologs The crystal structure of the MscL homolog from Mycobac- of each of these three proteins are found in almost all Gram- terium tuberculosis (TbMscL) was solved to 3.5 Å resolution positive and Gram-negative bacteria. BLAST searches also for the closed state (167), revealing a tight assembly of five identify MscS homologs in fission yeast and higher plants (Ara- two–transmembrane domain subunits. Slightly tilted M1 helices bidopsis thaliana), whereas a distant MscL homolog was found are inside the channel barrel, whereas the M2 helices are on the in fungus Neurospora crassa (155). A group of channels that re- periphery facing the lipid. The C-terminal domains, connected sembles MscS and MscL, including MscTA and MscMJ, has to M2 helices with flexible linkers, are helical and form an ad- been isolated from archaea (153). ditional pentameric bundle in the cytoplasm. The M1 helices, When mscL-null mutants of E. coli were first generated, no which are tightly packed at their intersection near the middle of visible growth or survival phenotype was observed under vari- the membrane, create a hydrophobic pore constriction proposed ous osmotic conditions (154). The double MscL/MscS knock- to be the gate. It was initially hypothesized that the opening of out displayed an osmotically fragile phenotype, proving that a the pore involved straightening of the M1 helices and the for- tension-dependent increase in membrane permeability is indeed mation of a cylindrical barrel with parallel M1 and M2 helices critical for cell survival under strong hypo-osmotic shock, and acting as “staves” (but see below) (167, 175). indicating some functional redundancy of the system (158). Random mutagenesis of mscL combined with screens for MscK alone, although it opens before MscS and MscL, is un- growth-suppressing mutations gave the first “functional” clues able to rescue cells under acute osmotic shock, which could be as to where the MscL gate might be (176). The isolated muta- due to either low-copy-number protein expression or specific tions indicated that the integrity of the cytoplasmic half of M1, regulation by permeant ions (160). The MscS and MscL chan- specifically the constriction region where the helices are tightly nels, although they share no sequence homology, are functional- packed, is critical for maintaining the channel in the closed con- ly similar in many respects. They are both active when purified formation at subthreshold tensions. Most of the toxic mutations and reconstituted in liposomes, indicating that they gate directly were either polar or charged substitutions for hydrophobic by tension in the lipid bilayer (161–163). MscS is moderately residues occluding the pore, or bulky side-chain substitutions voltage sensitive (164) in that it is more active at depolarizing for regularly spaced glycines in M1 (apparently involved in membrane potentials. Both MscL and MscS are sensitive to tight helical packing). The majority of these mutations pro- membrane curvature perturbations induced by amphipathic duced easy-to-open (“gain-of-function”) channel phenotypes. compounds such as lysolipids or local anesthetics (165, 166). Further site-specific mutagenesis studies demonstrated a direct

www.stke.org/cgi/content/full/sigtrans;2004/219/re4 Page 10 R EVIEW dependence of the activation pressure threshold on the degree conductive state. Further expansion of the barrel pulls on short of hydrophobicity of a substitution for Gly22 near the constric- S1-M1 linkers, resulting in disruption of the N-terminal bundle tion point, indicating that hydration of the pore interior favors and full opening of the channel. The C-terminal (S3) helices the open state (177–179). were initially modeled as separating (180); however, recent data Kinetic and thermodynamic analyses of MscL in liposomes have strongly suggested that they remain stably associated in all provided important information on the character of the gating conformations and form a prefilter on the cytoplasmic side of transition. Recordings showed that MscL opening occurs the pore (174). through a series of short-lived substates (173). The relative de- Different aspects of this model of the open state have been pendencies of transitions between the substates on tension sug- tested using disulfide cross-linking, and also in patch-clamp ex- gested that gating proceeds through a pre-expanded low-con- periments on cysteine mutants under reducing or oxidizing conducting intermediate state. A molecular model of the transition ditions (174, 181, 182). The barrel was trapped in the expanded of MscL between the closed and the open state through an ex- state with highly tilted M1 helices by pairs of cysteines in posi- panded intermediate state is depicted in Fig. 5. The homology tions 20 and 36 on different subunits. These cysteine cross-links model of E. coli MscL, built from the TbMscL crystal structure, stabilized the open state of the channel. Cross-links between ad- represents the initial closed conformation. The “parallel barrel- jacent M1 and M2 helices of neighboring subunits did not pre- stave” model of the open conformation failed to satisfy several vent gating, indicating that these helices tilt together during the criteria (180). Expansion of the transmembrane barrel is transition; in other words, they do not change their relative posi- achieved not by straightening of the helices, but rather by their tion toward one another substantially (182). The proximity of tilting with a simultaneous outward movement in an iris-like highly conserved phenylalanines (Phe7 or Phe10) in the N-termi- manner. The N-terminal segments (S1), unresolved in the crys- nal gate (S1) was supported by spontaneous intersubunit cross- tal structure of TbMscL, were hypothesized to form a short linking of cysteines in these corresponding positions, which bundle that acts as a secondary cytoplasmic gate (181). In the prevented the channel from opening. At the same time, a ran- short-lived pre-expanded conformation, this gate is closed and dom mutagenesis study by Maurer and Dougherty (183) sug- the channel remains in a nonconductive or, more likely, low- gested that, despite a high degree of conservation, the N termi-

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).

www.stke.org/cgi/content/full/sigtrans;2004/219/re4 Page 11 R EVIEW nus may tolerate drastic substitutions without affecting channel MS channels (MscL and MscS) that share no homology or function in in vivo cell viability assays. The length of the link- structural similarity. ers that connect S1 domains (N-terminal gate) with M1 In terms of gating mechanisms, the MEC channel of C. ele- nonetheless had a strong effect on the structure of the subcon- gans critically relies on specific cytoskeletal attachments. The ducting states, suggesting a functional role as tension-transmit- auditory channel, which is possibly a member of the TRP fami- ting elements between the domains (181). Disulfide cross-links ly, requires intact tip links for gating and a connection to the between the C-terminal (S3) helices did not affect gating pa- actin cytoskeleton through myosin for opening. It is unclear at rameters, consistent with their stable association in all confor- present whether other mechanosensitive TRP channels require mations (174). the cytoskeleton for gating, but it seems likely. SAKCa (136) Similar to the activation of MscS by amphipathic substances and the two-pore (TREK) potassium channels (134), however, (165), E. coli MscL reconstituted in liposomes can be stabilized seem to be gated directly by tension in the lipid bilayer. in the open state by adding lysolipids to the outside of the lipo- In addition to strongly mechano-activated channels with a some (166). This made it possible to perform a site-directed steep dependence of open probability on the stimulus, there spin labeling and EPR (electron paramagnetic resonance) study have been reports of mechanical modulation of well-character- of MscL in the open state, which independently confirmed not ized channels with established functions other than only that the pore is lined primarily by the M1 helix, but also mechanosensation. These include ATP-dependent (KAT P ) potas- that the M2 helix is not exposed to the aqueous phase in any sium channels (109), G protein–modulated (KACH) inward recti- state (184). The parameters of EPR spectra predicted separation fiers (190), and NMDA receptors (191). Calcium-dependent of the M1 helices by about 25 Å. EPR-derived constraints pro- BK channels (192), voltage-gated K+ channels (193), L-type duced a model with a highly tilted conformation of helices, sim- Ca2+ channels (194), and N-type Ca2+ channels (195) are also ilar to that depicted in Fig. 5. The iris-like expansion of the stretch-modulated. Because many types of non-MS channels channel barrel was also observed in molecular dynamic simula- are modulated by stretch, one may speculate that the entire tions (185–187), especially when the steering force mimicking spectrum of MS channels could be the result of independent membrane tension was applied specifically to the residues fac- evolution of several distinct predecessors to acquire mechanical ing the layer of phospholipid glycerols, the most rigid and sensitivity, by more efficient linkage to cytoskeletal elements or densely packed part of the bilayer (186). The residues that are in by an increased conformational flexibility or area change upon contact with lipid head groups have been shown to be function- opening. This notion is consistent with the existence of MS ally important for sensing tension in in vivo assays (183). channels in various channel superfamilies that contain channels The current model of the open state, under constraints im- activated by other stimuli than force. Examples include the two- posed by the cysteine cross-link and EPR data, predicts that pore TREK and TWIK channels (131), SAKCa and BK (108), or both M1 and M2 helices tilt by about 40° and move outward by osmosensitive and non-osmosensitive TRPVs (196). Despite at- about 12 to 13 Å. The inner surface of the pore is lined mostly tempts to find a common structural motif underlying the with the polar face of M1. The total in-plane expansion of the mechano-gated behavior (197), a common origin of MS chan- transmembrane barrel assessed from models is 22 to 23 nm2, nels is not evident. which is highly consistent with the initial slope of tension-activation curves (188). The flattening of the structure reduces the Concluding Remarks hydrophobic thickness of the protein, which may substantially Various putative MS channels and their functional contexts have distort annular lipids around the channel complex. Indeed, the been described. No unifying common origin of MS channels is open state of MscL was shown to be more stable when the evident; rather, evolution to mechanosensitivity has apparently channel was reconstituted in short-chain lipids rather than long- occurred many times, producing several types of MS channels chain lipids (166). The energy of the closed-to-open transition based on various channel architectures. is estimated as 50 to 55 kT (188), corroborating with extremely Two mechanistic paradigms exist: gating by force conveyed high activation tension (10.4 dynes/cm) and a large-scale con- through elements tethered to the cytoskeleton, and gating by formational rearrangement. Such a high energy cost may in- force conveyed through lateral tension in the membrane. “Teth- clude contributions from conformational stress in the barrel and ered” channels tend to reside in specialized sensory cells and periplasmic loops (189), from unfavorable pore hydration (177), organs, which typically bear flexible extensions such as stere- and from perturbation of the surrounding lipids (166). Further ocilia, specialized dendrites, or hairs. Deflection of these exten- studies are required to quantify these energetic contributions. sions in certain directions focuses the stress on the channel, which acts as a transducer and undergoes a gating transition at Mechanosensitive Channels: Do They Have Anything the nanometer scale. In the auditory system and probably else- in Common? where, molecular motors adjust the set point for gating of such Several families of MS channels that function in different con- channels. texts in the same organism clearly coexist. In animals, channels For many channels found in nonsensory cells, the plasma from the TRP family are implicated in osmosensation, gauging membrane itself constitutes the receptive structure for mechani- of fluid flow, and mechanosensation in certain neurons. Other cal stimuli. Interactions with the cortical cytoskeleton or the ex- groups of sensory neurons use DEG/ENaC-type channels. ternal cell wall define the fraction of stress borne by the lipid There are stretch-activated potassium channels from the bilayer, and therefore the amount of tension sensed by the chan- seven–transmembrane domain (7TM) (SAKCa) and 2TM (KAT P ) nel. MscL is a simple prokaryotic channel gated by tension in families in the heart, as well as mechano-gated two-pore (4TM) the lipid bilayer; a detailed understanding of its gating serves as potassium leakage channels in the brain. Even a simple an excellent prototype for channels of this type. During MscL prokaryote, E. coli, has at least two types of functionally similar gating, a large iris-like transition driven by lateral membrane

www.stke.org/cgi/content/full/sigtrans;2004/219/re4 Page 12 R EVIEW 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 pro- Fig. 6. Force-dependent gating. (A) Cartoon of a channel with a file cannot be known with any certainty, simple theories gener- gate that moves by a distance b when sufficient force is applied. ally consider just three points: the closed state, the open state, In many cases, force is applied by stretching a gating spring with and the peak of the barrier. The energy difference between the stiffness kg by a certain distance. (B) Energy profiles of the chan- two states, ∆g, dictates the likelihood of the channel occupying nel for different applied forces as the gate moves, if the total each state, as described by a Boltzmann distribution: 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, (Eq. 3) which is 4.07 × 10−21 J at room temperature (22°C). By substi- tuting pc = 1 – po we obtain the open probability, a time-averaged value measured in experiments: How is ∆g 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

www.stke.org/cgi/content/full/sigtrans;2004/219/re4 Page 13 R EVIEW changed:

(Eq. 4) (Eq. 7) where go and gc are the free energies of the open and closed states, and ∆u is the intrinsic energy difference between states where uc and uo are the energies of the closed and open state in in the absence of applied force. Then the absence of applied force, and xstim is the stimulus displacement, shown as Xs in Fig. 6. The difference between them is

(Eq. 1) (Eq. 8)

The open probability po is plotted as a function of f in Fig. 7A, 1 2 for values of b equal to 1, 2, and 5 nm. The term ∆u is arbitrari- The term ⁄2kgb is a constant offset energy corresponding to the ly set to 5 kBT for this plot, which makes the channel closed for gate moving against the gating spring. Its sign depends on zero force. It is clear that a larger gate swing means that less whether the open or closed state is taken as zero, and it disap- force is needed to create the same ∆g, and so the activation pears if zero is taken as halfway between (17). Note that the en- curve po (f) is steeper. ergy difference between states is a linear function of the stimu- Note that this is formally equivalent to the gating of volt- lus displacement, xstim. age-dependent channels. For them, the force is the product of Because the physical parameters b and kg are characteristics the electric field strength E and the net charge q on the gating of molecular entities and are not easily measured, it is useful to ≡ region. The field is E = Vm/m, where Vm is the transmembrane define a gating sensitivity, z kgb. z is a characteristic of the potential and m is the distance across which the field drops, combined channel–gating spring complex; it describes the usually approximated as the membrane thickness. Then 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 (Eq. 5) 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. Because of uncertainties in determining both b and m, the term Using the sensitivity z and defining an offset displacement x0 qb/m is collectively referred to as the equivalent gating charge. = ∆u/z, 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 (Eq. 9) stimulus is a pure force, directed to the channel. In others, the stimulus is a displacement of a larger structure, which is con- A plot of po as a function of xstim is shown in Fig. 7D for a value veyed by some elastic element to the channel’s gate (Fig. 1A). If of kg = 1 mN/m. this element, termed the “gating spring,” is a simple Hookean Stiffness of the channel–gating spring complex. Until the spring (which follows Hooke’s law and so has force proportion- gating domains of single channels can be manipulated to mea- al to extension) with a stiffness kg, then the energy in the spring sure displacements and forces directly, we must infer the me- 1 2 is ⁄2kgx . Thus, the energies of closed and open states are chanical 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). (Eq. 6) Consider, for example, the displacement of the far end of the gating spring when a specified force is applied to it. Under and 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,

www.stke.org/cgi/content/full/sigtrans;2004/219/re4 Page 14 R EVIEW

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

complex, consisting of channel and gating spring, as the reciprocal of the slope of this curve,

(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 x as a function of f, for values of b = 1, 2, and 5 nm, stim (Eq. 11) 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 k for the transduction This is plotted in Fig. 2C as a function of f, using Eq. 1. t A somewhat different result is obtained if the experiment is

www.stke.org/cgi/content/full/sigtrans;2004/219/re4 Page 15 R EVIEW done under “length-clamp” conditions—if the end of the gating trin networks in red blood cells (205). spring is moved to a specified position, and the force needed to With the assumption that there is at most one channel per node, hold it there is measured (200). Then the force can be calculated the maximum channel density can be calculated; for all three ge- 2 by considering the average stretch of the gating spring, ometries it is roughly 1/s . 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 (Eq. 12) result is independent of the length of a filament. Gating by Stress in the Lipid Bilayer which includes the reduction in length upon opening of the Most work on mechanically gated channels has used patch- channel. The force as a function of displacement is shown in recording electrodes to apply suction to a patch and thereby Fig. 2E, now using the dependence of po on xstim (Eq. 9). Note generate tension in the membrane. If we consider the patch that Figs. 2B and 2E are not simply a swap of force and dis- membrane as a section of a sphere, as is often the case, the lat- placement axes, as two different expressions for po are used. eral tension γ can be determined by Laplace’s law from the Consequently, the stiffness calculated as a function of xstim is pressure across the membrane, P, and the radius of curvature of different: the membrane, r:

(Eq. 15) (Eq. 13) Because it is almost impossible to measure tension in a membrane This is plotted as a function of xstim in Fig. 7F, using Eq. 9. Note that area of a few square micrometers directly, determining the radius is the stiffness under length-clamp conditions can be negative. The a particularly useful way of estimating membrane tension. term po(1 – po) has a maximum value of 0.25 when po = 0.5, so the Tension in general is the energy excess per unit area created 2 stiffness is negative when kgb /4 > kBT, which occurs when the by any type of stress. A lateral two-dimensional stretch of a swing of the gate or the gating spring stiffness is large (203). Nega- membrane generates expanded area. The tension in this instance tive stiffness has been observed for hair cell transduction (204). is given by 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 (Eq. 16) 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 where ka is the area elasticity constant and ∆A/A is the propor- bilayer is less expandable than a network of filaments; however, if tional area change of the membrane. Note that the spring con- the lipid area is in excess (wrinkled) or can flow, then the cytoskele- stants for two-dimensional materials have units of N/m, as for ton contributes most of the stiffness and conveys force to the chan- one-dimensional springs, but the stretch ∆A/A is a dimension- nel. The relationship between the two-dimensional stress in the net- less ratio, so that lateral tension is also measured in N/m or work and force in individual filaments must be determined. It can J/m2. The elasticity constant of a typical lipid bilayer is about be derived for certain regular arrays in which polygons tile a flat 0.2 N/m. Generally bilayers are considered nonstretchable be- surface, by considering how tension in one dimension is divided cause the relative expansion (∆A/A) at a lytic (mechanical among the filaments oriented in that dimension (Fig. 8). In the case breakdown) tension does not exceed 3 to 5%. of a square array, the force in an individual filament is clearly just Two other forms of stresses that pertain to biological mem- f = γs, where s is the length of a side and γ is lateral tension. For branes, shear stress and bending stress (Fig. 9), can also be pre- trigonal and hexagonal arrays, similar expressions may be derived sented as surface energy excess. Tension generated by shear ex- by considering the force component along one dimension. The area tension of the membrane is expressed as of each polygon may be calculated as well. It can be seen that the force in an individual filament is roughly

(Eq. 14) (Eq. 17)

This general approximation can be used when a filamentous net- where Φ represents the elastic modulus of shear rigidity, and λr work forms an irregular polygonal array, as has been seen for spec- is the extension ratio, in which

www.stke.org/cgi/content/full/sigtrans;2004/219/re4 Page 16 R EVIEW

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 ge- 1/2 ometry, the force is close to γ(Ap) , 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 = –ka∆A/A in the force expression, ∆A/A ≈ 2∆s/s, and f = kg∆s. It is close to 2ka, but (perhaps surprisingly) does not depend on s.

Fig. 9. Representation of three types of stress in a membrane. Lat- eral 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.

www.stke.org/cgi/content/full/sigtrans;2004/219/re4 Page 17 R EVIEW

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.

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 (Eq. 18) contributions from each form of tension and by the time course of the viscoelastic relaxations of the membrane. In general, for extensions λ1 and λ2 in orthogonal directions. Shear stress however, biological membranes are “stiffest” to area expansion; pertains to cell membranes scaffolded by proteins and cy- the free energies associated with shear extension and bending toskeleton. For two-dimensional lipid films in liquid state, the are smaller, at least for the distortions normally experienced by shear rigidity modulus is zero. cell membranes. We therefore focus on the contribution to free Tension generated by bending of the membrane is expressed as 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 ∆a, hence the tension produces the work of γ∆a on the protein. This lowers the free energy difference between the two states, which can now be expressed as –γ∆a + ∆h. We use not the Gibbs but the Helmholtz free energies, because formally the system is not un- (Eq. 19) der constant pressure. Here, ∆h is the Helmholtz energy differ-

www.stke.org/cgi/content/full/sigtrans;2004/219/re4 Page 18 R EVIEW ence between the open and closed states in the absence of ten- easy to see that the forward rate would be more sensitive to sion. In the Boltzmann representation, the tension response of tension when the barrier is positioned closer to the center of the channel can be written as 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 (Eq. 20) Gating A conformational transition in a mechanosensory protein may In practice, po is often measured as time-averaged mean current change the geometry of the protein-lipid boundary. For in- through a single channel or a population relative to the current stance, the predicted flattening of MscL upon opening may at saturating tension. This equation predicts the behavior of produce a mismatch with the boundary lipids around the pro- many MS and mechano-modulated channels (11). Even chan- tein, and the induced stress in the bilayer would make a sub- nels with a number of subconducting states such as MscL (173) stantial contribution to the free energy of the open state. The are reasonably well described by Eq. 20, as long as the occu- observation that MscL gating occurs at lower pipette pressures pancies of main substates remain independent of tension (188). when the channel is reconstituted in shorter-chain lipids that Equation 20 is based on a linear dependence of the free ener- form thinner bilayers (166) supports this view. Gating in sim- gy on tension and implies that the closed and open states are pler peptide channels that form conducting dimers, such as characterized by fixed areas at all tensions. The potential profile gramicidin A, also depends on membrane thickness (209), cur- of such a system along the reaction coordinate (∆a) would be vature stress (210), and tension (211). The effect of tension on represented by two narrow rectangular wells (Fig. 10, middle). the probability of the conducting state for gramicidin channels Accounting for elasticity of each state (harmonic approxima- may depend on the thickness of the bilayer (212). Tension in- tion) invokes a profile of two parabolic wells (Fig. 10, bottom) creases the number of conducting gramicidin dimers in thick and a quadratic term in the free energy expression (6, 13). bilayers (diecosenoyl phosphatidylcholine), whereas in thinner Indeed, the rate constants of unidirectional transitions written in membranes (diolyl phosphatidylcholine) it decreases them Eyring’s form are (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 in- (Eq. 21) trinsic composition of a membrane than to perturbations of the membrane by external force. It originates from the fact that hy- and drophobic 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 (Eq. 22) would be (214, 215). This intrinsic elastic stress not only pre- disposes to the formation of nonbilayer structures, but also may change the conformational equilibrium of membrane- where ko is a scaling factor, acb and aob are the area changes embedded proteins (216). The measure of curvature stress can from the bottoms of the closed or open wells to the top of the be quantified by the lateral pressure profile across the bilayer transition barrier, and Ki is the elastic constant of the corre- (217). Figure 11 depicts a channel that changes its shape inside sponding well. The equilibrium constant po/pc is equal to the ra- the membrane from hourglass-like (closed) to barrel-like tio of the forward and backward rates, which gives the follow- (open). Apparently, the equilibrium depends on the profile of ing expression for po: 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 phos- (Eq. 23) phatidylethanolamine (PE) with a small head group, or in the presence of apolar inclusions such as free fatty acids (AA). When the stiffnesses of the two conformations are equal (Kc = Conversely, the pressure contribution of the polar region would Ko), the quadratic term disappears and the free energy depen- increase in the presence of micelle-forming lipids with a bulky dence becomes linear with γ. When the stiffnesses are differ- head and one tail (lysophosphatidylcholine, LPC), or local ent, the nonzero quadratic term represents the difference in anesthetics such as chlorpromazine (CPZ) residing at the inter- elastic energies stored in each state (6). The difference in state face between polar and non-polar regions. elasticities can be assessed from the dependencies of the for- It is easy to show that a bilayer perturbation that changes the ward and backward rate constants on tension. The slopes of pressure along the z axis by ∆p(z) would produce a change in ln(kon) and ln(koff) on tension would give ∆acb and ∆aob and free energy difference between the states equal to the product of their ratio would estimate the position of the transition state local protein area and pressure changes integrated over the (barrier) between the endpoints on the reaction coordinate. It is membrane thickness (218):

www.stke.org/cgi/content/full/sigtrans;2004/219/re4 Page 19 R EVIEW

7. R. Maroto, O. P. Hamill, Brefeldin A block of integrin-dependent mechanosensitive ATP release from Xenopus oocytes reveals a novel mechanism of mechanotransduction. J. Biol. Chem. 276, 23867–23872 (2001). 8. M. Chiquet, Regulation of extracellular matrix gene expression by mechanical stress. Matrix Biol. 18, 417–426 (1999). 9. F. J. Alenghat, D. E. Ingber, Mechanotransduction: All signals point to cytoskeleton, matrix, and integrins Sci. STKE 2002, pe6 (2002). 10. W. Rong, K. M. Spyer, G. Burnstock, Activation and sensitisation of low and high threshold afferent fibres mediated by P2X receptors in the mouse urinary bladder. J. Physiol. 541, 591–600 (2002). 11. O. P. Hamill, D. W. McBride Jr., The cloning of a mechano-gated membrane ion channel. Trends Neurosci. 17, 439–443 (1994). 12. D. P. Corey, A. J. Hudspeth, Analysis of the microphonic potential of the bullfrog’s sacculus. J. Neurosci. 3, 942–961 (1983). 13. F. Sachs, H. Lecar, Stochastic models for mechanical transduction. Bio- phys. J. 59, 1143–1145 (1991). Fig. 11. Effect of lateral pressure profile on the conformational equi- 14. A. J. Hudspeth, Hair bundle mechanics and a model for mechanoelec- librium of a membrane protein. Each leaflet of the bilayer consists of trical adaptation by hair cells. In Sensory Transduction, D. P. Corey, S. three distinct regions: polar head groups and hydrophobic acyl Roper, Eds. (Rockefeller Univ. Press, New York, 1992), pp. 357–370. chains, both exerting positive pressure, and a region in between, 15. J. O. Pickles, Recent advances in cochlear physiology. Prog. Neurobiol. 24, 1–42 (1985). where interfacial tension is present (217). The protein is presumed to 16. J. A. Assad, G. M. Shepherd, D. P. Corey, Tip-link integrity and mechan- change its shape from hourglass-like to barrel-like. A change in the ical transduction in vertebrate hair cells. Neuron 7, 985–994 (1991). balance between lateral pressures in the polar (P) and hydrophobic 17. J. Howard, A. J. Hudspeth, Compliance of the hair bundle associated (H) regions of the membrane would bias the equilibrium one way or with gating of mechanoelectrical transduction channels in the bullfrog’s saccular hair cell. Neuron 1, 189–199 (1988). the other. In a membrane composed predominantly of “bilayer” lipids 18. B. Kachar, M. Parakkal, M. Kurc, Y. Zhao, P. G. Gillespie, High-resolu- such as phosphatidylcholine (PC), the pressures are about equal, tion structure of hair-cell tip links. Proc. Natl. Acad. Sci. U.S.A. 97, meaning no curvature stress. Lipids with smaller hydrophilic moieties 13336–13341 (2000). such as phosphatidylethanolamine (PE) or an admixture of unsatu- 19. K. P. Steel, C. J. Kros, A genetic approach to understanding auditory function. Nat. Genet. 27, 143–149 (2001). rated arachidonic acid (AA) would create larger pressure in the hy- 20. C. M. Canessa, J. D. Horisberger, B. C. Rossier, Epithelial sodium drophobic layer, favoring the hourglass-like conformation. The channel related to proteins involved in neurodegeneration. Nature 361, micelle-forming molecules such as lysophosphatidylcholine (LPC) or 467–470 (1993). chlorpromazine (CPZ) increase pressure in the polar regions and 21. C. M. Canessa, L. Schild, G. Buell, B. Thorens, I. Gautschi, J. D. Horis- berger, B. C. Rossier, Amiloride-sensitive epithelial Na+ channel is produce an opposite effect. made of three homologous subunits. Nature 367, 463–467 (1994). 22. J. García-Añoveros, B. Derfler, J. Neville-Golden, B. T. Hyman, D. P. Corey, BNaC1 and BNaC2 constitute a new family of human neuronal sodium channels related to degenerins and epithelial sodium channels Proc. Natl. Acad. Sci. U.S.A. 94, 1459–1464 (1997). 23. R. Waldmann, G. Champigny, N. Voilley, I. Lauritzen, M. Lazdunski, The mammalian degenerin MDEG, and amiloride-sensitive cation channel activated by mutations causing neurodegeneration in Caenorhabdi- tis elegans. J. Biol. Chem. 271, 10433–10436 (1996). (Eq. 24) 24. M. Price, P. Snyder, M. J. Welsh, Cloning and expression of a novel human brain Na+ channel. J. Biol. Chem. 271, 7879–7882 (1996). 25. D. Firsov, I. Gautschi, A. M. Merillat, B. C. Rossier, L. Schild, The het- This will dictate a redistribution of the protein between the erotetrameric architecture of the epithelial sodium channel (ENaC). states. In the particular case depicted in the figure, the open EMBO J. 17, 344–352 (1998). state is favored by micelle-forming lipids, such as LPC, which 26. P. M. Snyder, C. Cheng, L. S. Prince, J. C. Rogers, M. J. Welsh, Elec- trophysiological and biochemical evidence that DEG/ENaC cation create a “vacuum” in the hydrocarbon core of the membrane. channels are composed of nine subunits. J. Biol. Chem. 273, 681–684 This type of perturbation induces positive spontaneous curva- (1998). ture so that each leaflet would tend to “curl” locally toward the 27. S. Eskandari, P. M. Snyder, M. Kreman, G. A. Zampighi, M. J. Welsh, E. M. Wright, Number of subunits comprising the epithelial sodium chan- midplane of the bilayer. If the protein becomes not only wider nel. J. Biol. Chem. 274, 27281–27286 (1999). in the middle but also shorter, this conformation would also be 28. M. Chalfie, J. E. Sulston, J. G. White, E. Southgate, J. N. Thomson, S. favored. Strong effects of lysolipids have been observed on the Brenner, The neural circuit for touch sensitivity in Caenorhabditis elegans. J. Neurosci. 5, 956–964 (1985). small bacterial MS channel (165) and, more recently, on MscL 29. M. B. Goodman, E. M. Schwarz, Transducing touch in Caenorhabditis (166). elegans. Annu. Rev. Physiol. 65, 429–452 (2003). 30. H. Du, G. Gu, C. M. William, M. Chalfie, Extracellular proteins needed for C. elegans mechanosensation. Neuron 16, 183–194 (1996). References and Notes 31. M. Chalfie, M. Au, Genetic control of differentiation of the Caenorhabdi- 1. M. E. Burns, D. A. Baylor, Activation, deactivation, and adaptation in tis elegans touch receptor neurons. Science 243, 1027–1033 (1989). vertebrate photoreceptor cells. Annu. Rev. Neurosci. 24, 779–805 32. G. G. Ernstrom, M. Chalfie, Genetics of sensory mechanotransduction. (2001). Annu. Rev. Genet. 36, 411–453 (2002). 2. V. Y. Arshavsky, T. D. Lamb, E. N. Pugh Jr., G proteins and phototrans- 33. M. B. Goodman, G. G. Ernstrom, D. S. Chelur, R. O’Hagan, C. A. Yao, duction. Annu. Rev. Physiol. 64, 153–187 (2002). M. Chalfie, MEC-2 regulates C. elegans DEG/ENaC channels needed 3. S. Firestein, How the olfactory system makes sense of scents. Nature for mechanosensation. Nature 415, 1039–1042 (2002). 413, 211–218 (2001). 34. M. Chalfie, J. Sulston, Developmental genetics of the mechanosensory 4. P. G. Gillespie, R. G. Walker, Molecular basis of mechanosensory neurons of Caenorhabditis elegans. Dev. Biol. 82, 358–370 (1981). transduction. Nature 413, 194–202 (2001). 35. J. García-Añoveros, J. A. García, J.-D. Liu, D. P. Corey, The nematode 5. M. J. Welsh, M. P. Price, J. Xie, Biochemical basis of touch perception: degenerin UNC-105 forms ion channels that are activated by degenera- Mechanosensory function of degenerin/epithelial Na+ channels. J. Biol. tion- or hypercontraction-causing mutations. Neuron 20, 1231–1241 Chem. 277, 2369–2372 (2002). (1998). 6. F. Sachs, Stretch-sensitive ion channels: An update. Soc. Gen. Physiol. 36. J. García-Añoveros, T. A. Samad, L. Zuvela-Jelaska, C. J. Woolf, D. P. Ser. 47, 241–260 (1992). Corey, Transport and localization of the DEG/ENaC ion channel

www.stke.org/cgi/content/full/sigtrans;2004/219/re4 Page 20 R EVIEW

BNaC1alpha to peripheral mechanosensory terminals of dorsal root mechanosensitive. Proc. Natl. Acad. Sci. U.S.A. 100, 7105–7110 ganglia neurons. J. Neurosci. 21, 2678–2686 (2001). (2003). 37. M. P. Price, G. R. Lewin, S. L. McIlwrath, C. Cheng, J. Xie, P. A. Hep- 62. M. Kernan, D. Cowan, C. Zuker, Genetic dissection of mechanosensory penstall, C. L. Stucky, A. G. Mannsfeldt, T. J. Brennan, H. A. Drum- transduction: Mechanoreception-defective mutations of Drosophila. mond, J. Qiao, C. J. Benson, D. E. Tarr, R. F. Hrstka, B. Yang, R. A. Neuron 12, 1195–1206 (1994). Williamson, M. J. Welsh, The mammalian sodium channel BNC1 is re- 63. R. G. Walker, A. T. Willingham, C. S. Zuker, A Drosophila mechanosen- quired for normal touch sensation. Nature 407, 1007–1011 (2000). sory transduction channel. Science 287, 2229–2234 (2000). 38. M. P. Price, S. L. McIlwrath, J. Xie, C. Cheng, J. Qiao, D. E. Tarr, K. A. 64. H. Mutai, S. Heller, Vertebrate and invertebrate TRPV-like mechano- Sluka, T. J. Brennan, G. R. Lewin, M. J. Welsh, The DRASIC cation receptors. Cell Calcium 33, 471–478 (2003). channel contributes to the detection of cutaneous touch and acid stimuli 65. S. Sidi, R. W. Friedrich, T. Nicolson, NompC TRP channel required for in mice. Neuron 32, 1071–1083 (2001). vertebrate sensory hair cell mechanotransduction. Science 301, 96–99 39. E. Kamynina, O. Staub, Concerted action of ENaC, Nedd4-2, and Sgk1 (2003). in transepithelial Na+ transport. Am. J. Physiol. Renal Physiol. 283, 66. J. R. Meyers, R. B. MacDonald, A. Duggan, D. Lenzi, D. G. Standaert, 377–387 (2002). J. T. Corwin, D. P. Corey, Lighting up the senses: FM1-43 loading of 40. B. Lindemann, Receptors and transduction in taste. Nature 413, sensory cells through nonselective ion channels. J. Neurosci. 23, 219–225 (2001). 4054–4065 (2003). 41. H. A. Drummond, M. P. Price, M. J. Welsh, F. M. Abboud, A molecular 67. D. P. Corey, unpublished data. component of the arterial baroreceptor mechanotransducer. Neuron 21, 68. R. Strotmann, C. Harteneck, K. Nunnenmacher, G. Schultz, T. D. Plant, 1435–1441 (1998). OTRPC4, a nonselective cation channel that confers sensitivity to ex- 42. H. A. Drummond, F. M. Abboud, M. J. Welsh, Localization of beta and tracellular osmolarity. Nat. Cell Biol. 2, 695–702 (2000). gamma subunits of ENaC in sensory nerve endings in the rat foot pad. 69. W. Liedtke, Y. Choe, M. A. Marti-Renom, A. M. Bell, C. S. Denis, A. Sali, Brain Res. 884, 1–12 (2000). A. J. Hudspeth, J. M. Friedman, S. Heller, Vanilloid receptor-related 43. C. M. Adams, M. G. Anderson, D. G. Motto, M. P. Price, W. A. Johnson, osmotically activated channel (VR-OAC), a candidate vertebrate M. J. Welsh, Ripped pocket and pickpocket, novel Drosophila osmoreceptor. Cell 103, 525–535 (2000). DEG/ENaC subunits expressed in early development and in 70. U. Wissenbach, M. Bodding, M. Freichel, V. Flockerzi, Trp12, a novel mechanosensory neurons. J. Cell Biol. 140, 143–152 (1998). Trp related protein from kidney. FEBS Lett. 485, 127–134 (2000). 44. J. A. Ainsley, J. M. Pettus, D. Bosenko, C. E. Gerstein, N. Zinkevich, 71. N. S. Delany, M. Hurle, P. Facer, T. Alnadaf, C. Plumpton, I. Kinghorn, M. G. Anderson, C. M. Adams, M. J. Welsh, W.A. Johnson, Enhanced C. G. See, M. Costigan, P. Anand, C. J. Woolf, D. Crowther, P. locomotion caused by loss of the Drosophila DEG/ENaC protein Pick- Sanseau, S. N. Tate, Identification and characterization of a novel hu- pocket1. Current Biol. 13: 1557-1563 (2003). man vanilloid receptor-like protein, VRL-2. Physiol. Genomics 4, 45. D. J. Cosens, A. Manning, Abnormal electroretinogram from a 165–174 (2001). Drosophila mutant. Nature 224, 285–287 (1969). 72. M. Suzuki, A. Mizuno, K. Kodaira, M. Imai, Impaired Pressure Sensa- 46. B. Minke, C. Wu, W. L. Pak, Induction of photoreceptor voltage noise in tion in Mice Lacking TRPV4. J. Biol. Chem. 278, 22664–22668 (2003). the dark in Drosophila mutant. Nature 258, 84–87 (1975). 73. A. Mizuno, N. Matsumoto, M. Imai, M. Suzuki, Impaired osmotic sensa- 47. C. Montell, G. M. Rubin, Molecular characterization of the Drosophila tion in mice lacking TRPV4. Am. J. Physiol. Cell Physiol. 285, trp locus: A putative integral membrane protein required for phototrans- C96–C101 (2003). duction. Neuron 2, 1313–1323 (1989). 74. B. Nilius, H. Watanabe, J. Vriens, The TRPV4 channel: Structure- 48. M. Gupta, D. P. Corey, unpublished data. function relationship and promiscuous gating behaviour. Pflugers Arch. 49. M. A. Arnaout, Molecular genetics and pathogenesis of autosomal dom- 446, 298–303 (2003). inant polycystic kidney disease. Annu. Rev. Med. 52, 93–123 (2001). 75. H. Watanabe, J. Vriens, S. H. Suh, C. D. Benham, G. Droogmans, B. 50. F. Qian, F. J. Germino, Y. Cai, X. Zhang, S. Somlo, G. G. Germino, Nilius, Heat-evoked activation of TRPV4 channels in a HEK293 cell ex- PKD1 interacts with PKD2 through a probable coiled-coil domain. Nat. pression system and in native mouse aorta endothelial cells. J. Biol. Genet. 16, 179–183 (1997). Chem. 277, 47044–47051 (2002). 51. K. Hanaoka, F. Qian, A. Boletta, A. K. Bhunia, K. Piontek, L. Tsiokas, V. 76. H. A. Colbert, C. I. Bargmann, Odorant-specific adaptation pathways P. Sukhatme, W. B. Guggino, G. G. Germino, Co-assembly of poly- generate olfactory plasticity in C. elegans. Neuron 14, 803–812 (1995). cystin-1 and -2 produces unique cation-permeable currents. Nature 77. H. A. Colbert, C. I. Bargmann, Environmental signals modulate olfacto- 408, 990–994 (2000). ry acuity, discrimination, and memory in Caenorhabditis elegans. Learn. 52. G. J. Pazour, J. T. San Agustin, J. A. Follit, J. L. Rosenbaum, G. B. Wit- Mem. 4, 179–191 (1997). man, Polycystin-2 localizes to kidney cilia and the ciliary level is elevat- 78. H. A. Colbert, T. L. Smith, C. I. Bargmann, OSM-9, a novel protein with ed in orpk mice with polycystic kidney disease. Curr. Biol. 12, 378–380 structural similarity to channels, is required for olfaction, mechanosen- (2002). sation, and olfactory adaptation in Caenorhabditis elegans. J. Neurosci. 53. B. K. Yoder, X. Hou, L. M. Guay-Woodford, The polycystic kidney dis- 17, 8259–8269 (1997). ease proteins, polycystin-1, polycystin-2, polaris, and cystin, are co- 79. J. M. Kaplan, H. R. Horvitz, A dual mechanosensory and chemosensory localized in renal cilia. J. Am. Soc. Nephrol. 13, 2508–2516 (2002). neuron in Caenohabditis elegans. Proc. Natl. Acad. Sci. U.S.A. 90, 54. H. A. Praetorius, K. R. Spring, Bending the MDCK cell primary cilium in- 2227–2231 (1993). creases intracellular calcium. J. Membr. Biol. 184, 71–79 (2001). 80. D. Tobin, D. Madsen, A. Kahn-Kirby, E. Peckol, G. Moulder, R. 55. S. M. Nauli, F. J. Alenghat, Y. Luo, E. Williams, P. Vassilev, X. Li, A. E. Barstead, A. Maricq, C. Bargmann, Combinatorial expression of TRPV Elia, W. Lu, E. M. Brown, S. J. Quinn, D. E. Ingber, J. Zhou, Polycystins channel proteins defines their sensory functions and subcellular local- 1 and 2 mediate mechanosensation in the primary cilium of kidney ization in C. elegans neurons. Neuron 35, 307–318 (2002). cells. Nat. Genet. 33, 129–137 (2003). 81. D. F. Eberl, R. W. Hardy, M. J. Kernan, Genetically similar transduction 56. M. M. Barr, P. W. Sternberg, A polycystic kidney-disease gene homo- mechanisms for touch and hearing in Drosophila. J. Neurosci. 20, logue required for male mating behaviour in C. elegans. Nature 401, 5981–5988 (2000). 386–389 (1999). 82. C. Kim, Y. Chung, Z. Gong, J. Kim, W. Son, D. Park, D. Shin, S. Choi, 57. T. Kaletta, M. Van der Craen, A. Van Geel, N. Dewulf, T. Bogaert, M. H. Lee, J. Hirsh, M. J. Kernan, Nanchung and Inactive are TRPV chan- Branden, K. V. King, M. Buechner, R. Barstead, D. Hyink, H. P. Li, L. nel subunits required for hearing. Paper presented at the 44th Annual Geng, C. Burrow, P. Wilson, Towards understanding the polycystins. Drosophila Research Conference, Chicago, 5 to 9 March 2003. Nephron Exp. Nephrol. 93, 9–17 (2003). 83. J. Kim, Y. D. Chung, D. Y. Park, S. Choi, D. W. Shin, H. Soh, H. W. Lee, 58. M. M. Barr, J. DeModena, D. Braun, C. Q. Nguyen, D. H. Hall, P. W. W. Son, J. Yim, C. S. Park, M. J. Kernan, C. Kim, A TRPV family ion Sternberg, The Caenorhabditis elegans autosomal dominant polycystic channel required for hearing in Drosophila. Nature 424, 81–84 (2003). kidney disease gene homologs lov-1 and pkd-2 act in the same path- 84. W. D. Tracey Jr., R. I. Wilson, G. Laurent, S. Benzer, painless, a way. Curr. Biol. 11, 1341–1346 (2001). Drosophila gene essential for nociception. Cell 113, 261–273 (2003). 59. C. P. Palmer, X. L. Zhou, J. Lin, S. H. Loukin, C. Kung, Y. Saimi, A TRP 85. G. M. Story, A. M. Peier, A. J. Reeve, S. R. Eid, J. Mosbacher, T. R. Hri- homolog in Saccharomyces cerevisiae forms an intracellular Ca2+- cik, T. J. Earley, A. C. Hergarden, D. A. Andersson, S. W. Hwang, P. permeable channel in the yeast vacuolar membrane. Proc. Natl. Acad. McIntyre, T. Jegla, S. Bevan, A. Patapoutian, ANKTM1, a TRP-like Sci. U.S.A. 98, 7801–7805 (2001). channel expressed in nociceptive neurons, is activated by cold temper- 60. V. Denis, M. S. Cyert, Internal Ca2+ release in yeast is triggered by hy- atures. Cell 112, 819–829 (2003). pertonic shock and mediated by a TRP channel homologue. J. Cell Bi- 86. C. E. Morris, Mechanosensitive ion channels. J. Membr. Biol. 113, ol. 156, 29–34 (2002). 93–107 (1990). 61. X. L. Zhou, A. F. Batiza, S. H. Loukin, C. P. Palmer, C. Kung, Y. Saimi, 87. F. Sachs, C. E. Morris, Mechanosensitive ion channels in nonspecial- The transient receptor potential channel on the yeast vacuole is ized cells. Rev. Physiol. Biochem. Pharmacol. 132, 1–77 (1998).

www.stke.org/cgi/content/full/sigtrans;2004/219/re4 Page 21 R EVIEW

88. O. P. Hamill, B. Martinac, Molecular basis of mechanotransduction in 116. P. F. Davies, Flow-mediated endothelial mechanotransduction. Physiol. living cells. Physiol. Rev. 81, 685–740 (2001). Rev. 75, 519–560 (1995). 89. O. P. Hamill, D. W. McBride Jr., The pharmacology of mechanogated 117. W. J. Sigurdson, F. Sachs, S. L. Diamond, Mechanical perturbation of membrane ion channels. Pharmacol. Rev. 48, 231–252 (1996). cultured human endothelial cells causes rapid increases of intracellular 90. F. Guharay, F. Sachs, Stretch-activated single ion channel currents in calcium. Am. J. Physiol. 264, H1745–H1752 (1993). tissue-cultured embryonic chick skeletal muscle. J. Physiol. 352, 118. J. Hoyer, R. Kohler, W. Haase, A. Distler, Up-regulation of pressure- 685–701 (1984). activated Ca2+-permeable cation channel in intact vascular endothelium 91. A. Franco-Obregon Jr., J. B. Lansman, Mechanosensitive ion channels of hypertensive rats. Proc. Natl. Acad. Sci. U.S.A. 93, 11253–11258 in skeletal muscle from normal and dystrophic mice. J. Physiol. 481, (1996). 299–309 (1994). 119. K. Naruse, M. Sokabe, Involvement of stretch-activated ion channels in 92. A. Franco-Obregon, J. B. Lansman, Changes in mechanosensitive Ca2+ mobilization to mechanical stretch in endothelial cells. Am. J. channel gating following mechanical stimulation in skeletal muscle Physiol. 264, C1037–C1044 (1993). myotubes from the mdx mouse. J. Physiol. 539, 391–407 (2002). 120. M. Sokabe, K. Naruse, S. Sai, T. Yamada, K. Kawakami, M. Inoue, K. 93. T. Y. Nakamura, Y. Iwata, M. Sampaolesi, H. Hanada, N. Saito, M. Art- Murase, M. Miyazu, Mechanotransduction and intracellular signaling man, W. A. Coetzee, M. Shigekawa, Stretch-activated cation channels mechanisms of stretch- induced remodeling in endothelial cells. Heart in skeletal muscle myotubes from sarcoglycan-deficient hamsters. Am. Vessels (suppl. 12), 191–193 (1997). J. Physiol. Cell Physiol. 281, C690–C699 (2001). 121. R. Kohler, A. Distler, J. Hoyer, Pressure-activated cation channel in in- 94. J. R. Blinks, Positive chronotropic effect of increasing right atrial pres- tact rat endocardial endothelium. Cardiovasc. Res. 38, 433–440 (1998). sure in the isolated mammalian heart. Am. J. Physiol. 186, 299–303 122. A. M. Malek, S. Izumo, Mechanism of endothelial cell shape change (1956). and cytoskeletal remodeling in response to fluid shear stress. J. Cell 95. D. E. Hansen, C. S. Craig, L. M. Hondeghem, Stretch-induced arrhyth- Sci. 109, 713–726 (1996). mias in the isolated canine ventricle. Evidence for the importance of 123. S. Brakemeier, I. Eichler, H. Hopp, R. Kohler, J. Hoyer, Up-regulation of mechanoelectrical feedback. Circulation 81, 1094–1105 (1990). endothelial stretch-activated cation channels by fluid shear stress. Car- 96. P. Kohl, A. D. Nesbitt, P. J. Cooper, M. Lei, Sudden cardiac death by diovasc. Res. 53, 209–218 (2002). Commotio cordis: Role of mechano-electric feedback. Cardiovasc. Res. 124. H. Sackin, Mechanosensitive channels. Annu. Rev. Physiol. 57, 50, 280–289 (2001). 333–353 (1995). 97. H. Hu, F. Sachs, Stretch-activated ion channels in the heart. J. Mol. 125. K. Strange, F. Emma, P. S. Jackson, Cellular and molecular physiology Cell. Cardiol. 29, 1511–1523 (1997). of volume-sensitive anion channels. Am. J. Physiol. 270, C711–C730 98. D. Kim, A mechanosensitive K+ channel in heart cells. Activation by (1996). arachidonic acid. J. Gen. Physiol. 100, 1021–1040 (1992). 126. D. E. Clapham, The list of potential volume-sensitive chloride currents 99. W. J. Sigurdson, A. Ruknudin, F. Sachs, Calcium imaging of mechani- continues to swell (and shrink). J. Gen. Physiol. 111, 623–624 (1998). cally induced fluxes in tissue-cultured chick heart: Role of stretch-acti- 127. J. M. Fernandez-Fernandez, M. Nobles, A. Currid, E. Vazquez, M. A. vated channels. Am. J. Physiol. Heart Circ. Physiol. 262, H1110–H1115 Valverde, Maxi K+ channel mediates regulatory volume decrease re- (1992). sponse in a human bronchial epithelial cell line. Am. J. Physiol. Cell 100. D. Kim, Novel cation-selective mechanosensitive ion channel in the atri- Physiol. 283, C1705–C1714 (2002). al cell membrane. Circ. Res. 72, 225–231 (1993). 128. C. W. Bourque, S. H. Oliet, Osmoreceptors in the central nervous 101. A. Ruknudin, F. Sachs, J. O. Bustamante, Stretch-activated ion chan- system. Annu. Rev. Physiol. 59, 601–619 (1997). nels in tissue-cultured chick heart. Am. J. Physiol. 264, H960–H972 129. Z. Zhang, C. W. Bourque, Osmometry in osmosensory neurons. Nat. (1993). Neurosci. 6, 1021–1022 (2003). 102. N. Sasaki, T. Mitsuiye, A. Noma, Effects of mechanical stretch on mem- 130. S. H. Oliet, C. W. Bourque, Mechanosensitive channels transduce os- brane currents of single ventricular myocytes of guinea-pig heart. Jpn. mosensitivity in supraoptic neurons. Nature 364, 341–343 (1993). J. Physiol. 42, 957–970 (1992). 131. A. J. Patel, E. Honore, Properties and modulation of mammalian 2P do- 103. H. Hu, F. Sachs, Mechanically activated currents in chick heart cells. J. main K+ channels. Trends Neurosci. 24, 339–346 (2001). Membr. Biol. 154, 205–216 (1996). 132. A. J. Patel, E. Honore, F. Maingret, F. Lesage, M. Fink, F. Duprat, M. 104. G. C. Bett, F. Sachs, Activation and inactivation of mechanosensitive Lazdunski, A mammalian two pore domain mechano-gated S-like K+ currents in the chick heart. J. Membr. Biol. 173, 237–254 (2000). channel. EMBO J. 17, 4283–4290 (1998). 105. G. C. Bett, F. Sachs, Whole-cell mechanosensitive currents in rat ven- 133. F. Maingret, A. J. Patel, F. Lesage, M. Lazdunski, E. Honore, Mechano- tricular myocytes activated by direct stimulation. J. Membr. Biol. 173, or acid stimulation, two interactive modes of activation of the TREK-1 255–263 (2000). potassium channel. J. Biol. Chem. 274, 26691–26696 (1999). 106. T. M. Suchyna, J. H. Johnson, K. Hamer, J. F. Leykam, D. A. Gage, H. 134. A. J. Patel, M. Lazdunski, E. Honore, Lipid and mechano-gated 2P do- F. Clemo, C. M. Baumgarten, F. Sachs, Identification of a peptide toxin main K+ channels. Curr. Opin. Cell Biol. 13, 422–428 (2001). from Grammostola spatulata spider venom that blocks cation-selective 135. D. H. Vandorpe, C. E. Morris, Stretch activation of the Aplysia S- stretch-activated channels. J. Gen. Physiol. 115, 583–598 (2000). channel. J. Membr. Biol. 127, 205–214 (1992). 107. R. E. Oswald, T. M. Suchyna, R. McFeeters, P. Gottlieb, F. Sachs, Solu- 136. Q. Zhi, K. Naruse, M. Sokabe, Ionic amphipaths affect the gating of a tion structure of peptide toxins that block mechanosensitive ion chan- stretch activated BK channel (SAKca) cloned from chick heart. Biophys. nels. J. Biol. Chem. 277, 34443–34450 (2002). J. 84, 234a (2003). 108. Q. Y. Tang, Z. Qi, K. Naruse, M. Sokabe, Characterization of a function- 137. C. L. Bowman, J. W. Lohr, Mechanotransducing ion channels in C6 ally expressed stretch-activated BKca channel cloned from chick ven- glioma cells. Glia 18, 161–176 (1996). tricular myocytes. J. Membr. Biol. 196, 185-200 (2003). 138. C. E. Morris, R. Horn, Failure to elicit neuronal macroscopic 109. D. R. Van Wagoner, Mechanosensitive gating of atrial ATP-sensitive mechanosensitive currents anticipated by single-channel studies. potassium channels. Circ. Res. 72, 973–983 (1993). Science 251, 1246–1249 (1991). 110. S. H. Kim, K. W. Cho, S. H. Chang, S. Z. Kim, S. W. Chae, Gliben- 139. Y. Zhang, O. P. Hamill, On the discrepancy between whole-cell and clamide suppresses stretch-activated ANP secretion: Involvements of membrane patch mechanosensitivity in Xenopus oocytes. J. Physiol. K+ATP channels and L-type Ca2+ channel modulation. Pflugers Arch. 523, 101–115 (2000). 434, 362–372 (1997). 140. Y. Zhang, F. Gao, V. L. Popov, J. W. Wen, O. P. Hamill, Mechanically 111. M. C. Wellner, G. Isenberg, Properties of stretch-activated channels in gated channel activity in cytoskeleton-deficient plasma membrane myocytes from the guinea-pig urinary bladder. J. Physiol. 466, 213–227 blebs and vesicles from Xenopus oocytes. J. Physiol. 523, 117–130 (1993). (2000). 112. M. C. Wellner, G. Isenberg, Stretch effects on whole-cell currents of 141. X. Wan, P. Juranka, C. E. Morris, Activation of mechanosensitive cur- guinea-pig urinary bladder myocytes. J. Physiol. 480, 439–448 (1994). rents in traumatized membrane. Am. J. Physiol. 276, C318–C327 113. M. T. Kirber, A. Guerrero-Hernandez, D. S. Bowman, K. E. Fogarty, R. (1999). A. Tuft, J. J. Singer, F. S. Fay, Multiple pathways responsible for the 142. Z. Gil, S. D. Silberberg, K. L. Magleby, Voltage-induced membrane dis- stretch-induced increase in Ca2+ concentration in toad stomach smooth placement in patch pipettes activates mechanosensitive channels. muscle cells. J. Physiol. 524, 3–17 (2000). Proc. Natl. Acad. Sci. U.S.A. 96, 14594–14599 (1999). 114. H. Zou, L. M. Lifshitz, R. A. Tuft, K. E. Fogarty, J. J. Singer, Visualiza- 143. C. E. Morris, U. Homann, Cell surface area regulation and membrane tion of Ca2+ entry through single stretch-activated cation channels. tension. J. Membr. Biol. 179, 79–102 (2001). Proc. Natl. Acad. Sci. U.S.A. 99, 6404–6409 (2002). 144. D. J. Cosgrove, Expansive growth of plant cell walls. Plant Physiol. 115. H. Matsumoto, C. B. Baron, R. F. Coburn, Smooth muscle stretch-acti- Biochem. 38, 1–16 (2000). vated phospholipase C activity. Am. J. Physiol. 268, C458–C465 145. A. L. Koch, S. Woeste, Elasticity of the sacculus of Escherichia coli. J. (1995). Bacteriol. 174, 4811–4819 (1992).

www.stke.org/cgi/content/full/sigtrans;2004/219/re4 Page 22 R EVIEW

146. R. J. Doyle, R. E. Marquis, Elastic, flexible peptidoglycan and bacterial tial parameters for gating of the bacterial large conductance cell wall properties. Trends Microbiol. 2, 57–60 (1994). mechanosensitive channel, MscL. J. Gen. Physiol. 113, 525–540 147. J. M. Wood, Osmosensing by bacteria: Signals and membrane-based (1999). sensors. Microbiol. Mol. Biol. Rev. 63, 230–262 (1999). 174. A. Anishkin, V. Gendel, N. A. Sharifi, C. S. Chiang, L. Shirinian, H. R. 148. D. J. Cosgrove, R. Hedrich, Stretch-activated chloride, potassium, and Guy, S. Sukharev, On the conformation of the COOH-terminal domain calcium channels coexisting in plasma membranes of guard cells of Vi- of the large mechanosensitive channel MscL. J. Gen. Physiol. 121, cia faba L. Planta 186, 143–153 (1991). 227–244 (2003). 149. J. P. Ding, B. G. Pickard, Mechanosensory calcium-selective cation 175. A. F. Batiza, I. Rayment, C. Kung, Channel gate! Tension, leak and dis- channels in epidermal cells. Plant J. 3, 83–110 (1993). closure. Struct. Fold. Des. 7, R99–103 (1999). 150. J. Alexandre, J.-P. Lassalles, Hysrostatic and osmotic pressure activat- 176. X. Ou, P. Blount, R. J. Hoffman, C. Kung, One face of a transmembrane ed channel in plant vacuole. Biophys. J. 60, 1326–1336 (1991). helix is crucial in mechanosensitive channel gating. Proc. Natl. Acad. 151. X. L. Zhou, M. A. Stumpf, H. C. Hoch, C. Kung, A mechanosensitive Sci. U.S.A. 95, 11471–11475 (1998). channel in whole cells and in membrane patches of the fungus 177. K. Yoshimura, A. Batiza, M. Schroeder, P. Blount, C. Kung, Hydrophilici- Uromyces. Science 253, 1415–1417 (1991). ty of a single residue within MscL correlates with increased channel 152. M. C. Gustin, X. L. Zhou, B. Martinac, C. Kung, A mechanosensitive ion mechanosensitivity. Biophys. J. 77, 1960–1972 (1999). channel in the yeast plasma membrane. Science 242, 762–765 (1988). 178. K. Yoshimura, A. Batiza, C. Kung, Chemically charging the pore con- 153. A. Kloda, B. Martinac, Mechanosensitive channels in archaea. Cell striction opens the mechanosensitive channel mscl. Biophys. J. 80, Biochem. Biophys. 34, 349–381 (2001). 2198–2206 (2001). 154. S. I. Sukharev, P. Blount, B. Martinac, C. Kung, Mechanosensitive 179. A. F. Batiza, M. M. Kuo, K. Yoshimura, C. Kung, Gating the bacterial channels of Escherichia coli: The MscL gene, protein, and activities. mechanosensitive channel MscL in vivo. Proc. Natl. Acad. Sci. U.S.A. Annu. Rev. Physiol. 59, 633–657 (1997). 99, 5643–5648 (2002). 155. C. D. Pivetti, M. R. Yen, S. Miller, W. Busch, Y. H. Tseng, I. R. Booth, M. 180. S. Sukharev, S. R. Durell, H. R. Guy, Structural models of the mscl gat- H. Saier Jr., Two families of mechanosensitive channel proteins. Micro- ing mechanism. Biophys. J. 81, 917–936 (2001). biol. Mol. Biol. Rev. 67, 66–85 (2003). 181. S. Sukharev, M. Betanzos, C. S. Chiang, H. R. Guy, The gating mecha- 156. C. Brownlee, F. Berger, F. Y. Bouget, Signals involved in control of po- nism of the large mechanosensitive channel MscL. Nature 409, larity, cell fate and developmental pattern in plants. Symp. Soc. Exp. Bi- 720–724 (2001). ol. 51, 33–41 (1998). 182. M. Betanzos, C. S. Chiang, H. R. Guy, S. Sukharev, A large iris-like ex- 157. S. I. Sukharev, P. Blount, B. Martinac, F. R. Blattner, C. Kung, A large- pansion of a mechanosensitive channel protein induced by membrane conductance mechanosensitive channel in E. coli encoded by mscL tension. Nat. Struct. Biol. 9, 704–710 (2002). alone. Nature 368, 265–268 (1994). 183. J. A. Maurer, D. A. Dougherty, Generation and evaluation of a large mu- 158. N. Levina, S. Totemeyer, N. R. Stokes, P. Louis, M. A. Jones, I. R. tational library from the E. coli mechanosensitive channel of large con- Booth, Protection of Escherichia coli cells against extreme turgor by ac- ductance, MscL. Implications for channel gating and evolutionary de- tivation of MscS and MscL mechanosensitive channels: Identification of sign. J. Biol. Chem. 278, 21076–21082 (2003). genes required for MscS activity. EMBO J. 18, 1730–1737 (1999). 184. E. Perozo, D. M. Cortes, P. Sompornpisut, A. Kloda, B. Martinac, Open 159. D. McLaggan, M. A. Jones, G. Gouesbet, N. Levina, S. Lindey, W. Ep- channel structure of MscL and the gating mechanism of mechanosensi- stein, I. R. Booth, Analysis of the kefA2 mutation suggests that KefA is tive channels. Nature 418, 942–948 (2002). a cation-specific channel involved in osmotic adaptation in Escherichia 185. J. Gullingsrud, D. Kosztin, K. Schulten, Structural determinants of mscl coli. Mol. Microbiol. 43, 521–536 (2002). gating studied by molecular dynamics simulations. Biophys. J. 80, 160. Y. Li, P. C. Moe, S. Chandrasekaran, I. R. Booth, P. Blount, Ionic regu- 2074–2081 (2001). lation of MscK, a mechanosensitive channel from Escherichia coli. EM- 186. J. Gullingsrud, K. Schulten, Gating of MscL studied by steered molecu- BO J. 21, 5323–5330 (2002). lar dynamics. Biophys. J. 85, 2087–2099 (2003). 161. S. I. Sukharev, B. Martinac, V. Y. Arshavsky, C. Kung, Two types of 187. G. Colombo, S. J. Marrink, A. E. Mark, Simulation of MscL gating in a mechanosensitive channels in the Escherichia coli cell envelope: Solu- bilayer under stress. Biophys. J. 84, 2331–2337 (2003). bilization and functional reconstitution. Biophys. J. 65, 177–183 (1993). 188. C. S. Chiang, A. Anishkin, S. Sukharev, Gating of the large 162. K. Okada, P. C. Moe, P. Blount, Functional design of bacterial mechanosensitive channel in situ. Estimation of the spatial scale of the mechanosensitive channels. Comparisons and contrasts illuminated by transition from channel population responses. Biophys. J., in press. random mutagenesis. J. Biol. Chem. 277, 27682–27688 (2002). 189. B. Ajouz, C. Berrier, M. Besnard, B. Martinac, A. Ghazi, Contributions of 163. S. Sukharev, Purification of the small mechanosensitive channel of Es- the different extramembranous domains of the mechanosensitive ion cherichia coli (MscS): The subunit structure, conduction, and gating channel MscL to its response to membrane tension. J. Biol. Chem. 275, characteristics in liposomes. Biophys. J. 83, 290–298 (2002). 1015–1022 (2000). 164. B. Martinac, M. Buechner, A. H. Delcour, J. Adler, C. Kung, Pressure- 190. A. Pleumsamran, D. Kim, Membrane stretch augments the cardiac sensitive ion channel in Escherichia coli. Proc. Natl. Acad. Sci. U.S.A. muscarinic K+ channel activity. J. Membr. Biol. 148, 287–297 (1995). 84, 2297–2301 (1987). 191. P. Paoletti, P. Ascher, Mechanosensitivity of NMDA receptors in cultured 165. B. Martinac, J. Adler, C. Kung, Mechanosensitive ion channels of E. coli mouse central neurons. Neuron 13, 645–655 (1994). activated by amphipaths. Nature 348, 261–263 (1990). 192. J. Mienville, J. L. Barker, G. D. Lange, Mechanosensitive properties of 166. E. Perozo, A. Kloda, D. M. Cortes, B. Martinac, Physical principles un- BK channels from embryonic rat neuroepithelium. J. Membr. Biol. 153, derlying the transduction of bilayer deformation forces during 211–216 (1996). mechanosensitive channel gating. Nat. Struct. Biol. 9, 696–703 (2002). 193. C. X. Gu, P. F. Juranka, C. E. Morris, Stretch-activation and stretch-in- 167. G. Chang, R. H. Spencer, A. T. Lee, M. T. Barclay, D. C. Rees, Struc- activation of Shaker-IR, a voltage-gated K+ channel. Biophys. J. 80, ture of the MscL homolog from Mycobacterium tuberculosis: A gated 2678–2693 (2001). mechanosensitive ion channel. Science 282, 2220–2226 (1998). 194. G. L. Lyford, P. R. Strege, A. Shepard, Y. Ou, L. Ermilov, S. M. Miller, S. 168. R. B. Bass, P. Strop, M. Barclay, D. C. Rees, Crystal structure of J. Gibbons, J. L. Rae, J. H. Szurszewski, G. Farrugia, alpha (1C) (Ca Escherichia coli MscS, a voltage-modulated and mechanosensitive (V)1.2) L-type calcium channel mediates mechanosensitive calcium channel. Science 298, 1582–1587 (2002). regulation. Am. J. Physiol. Cell Physiol. 283, C1001–C1008 (2002). 169. P. Blount, S. I. Sukharev, P. C. Moe, M. J. Schroeder, H. R. Guy, C. 195. B. Calabrese, I. V. Tabarean, P. Juranka, C. E. Morris, Mechanosensi- Kung, Membrane topology and multimeric structure of a mechanosensitivity of N-type calcium channel currents. Biophys. J. 83, 2560–2574 tive channel protein of Escherichia coli. EMBO J. 15, 4798–4805 (2002). (1996). 196. C. Montell, L. Birnbaumer, V. Flockerzi, The TRP channels, a remark- 170. P. Blount, S. I. Sukharev, M. J. Schroeder, S. K. Nagle, C. Kung, Single ably functional family. Cell 108, 595–598 (2002). residue substitutions that change the gating properties of a 197. A. Kumanovics, G. Levin, P. Blount, Family ties of gated pores: Evolu- mechanosensitive channel in Escherichia coli. Proc. Natl. Acad. Sci. tion of the sensor module. FASEB J. 16, 1623–1629 (2002). U.S.A. 93, 11652–11657 (1996). 198. D. P. Corey, A. J. Hudspeth, Kinetics of the receptor current in bullfrog 171. C. C. Hase, A. C. Le Dain, B. Martinac, Molecular dissection of the saccular hair cells. J. Neurosci. 3, 962–976 (1983). large mechanosensitive ion channel (MscL) of E. coli: Mutants with al- 199. J. Howard, W. M. Roberts, A. J. Hudspeth, Mechanoelectrical transduc- tered channel gating and pressure sensitivity. J. Membr. Biol. 157, tion by hair cells. Annu. Rev. Biophys. Biophys. Chem. 17, 99–124 17–25 (1997). (1988). 172. C. C. Cruickshank, R. F. Minchin, A. C. Le Dain, B. Martinac, Estimation 200. W. Denk, W. W. Webb, Forward and reverse transduction at the limit of of the pore size of the large-conductance mechanosensitive ion chan- sensitivity studied by correlating electrical and mechanical fluctuations nel of Escherichia coli. Biophys. J. 73, 1925–1931 (1997). in frog saccular hair cells. Hear. Res. 60, 89–102 (1992). 173. S. I. Sukharev, W. J. Sigurdson, C. Kung, F. Sachs, Energetic and spa- 201. H. Eyring, Viscosity, plasticity, and diffusion as examples of absolute re-

www.stke.org/cgi/content/full/sigtrans;2004/219/re4 Page 23 R EVIEW

action rates. J. Chem. Phys. 4, 283-291 (1936). activation and stretch inactivation depending on bilayer thickness. Proc. 202. B. Hille, Ion Channels of Excitable Membranes (Sinauer, Sunderland, Natl. Acad. Sci. U.S.A. 99, 4308–4312 (2002). MA, 2001). 213. J. A. Lundbaek, O. S. Andersen, Spring constants for channel-induced 203. W. Denk, R. M. Keolian, W. W. Webb, Mechanical response of frog sac- lipid bilayer deformations. Estimates using gramicidin channels. Bio- cular hair bundles to the aminoglycoside block of mechanoelectrical phys. J. 76, 889–895 (1999). transduction. J. Neurophysiol. 68, 927–932 (1992). 214. R. M. Epand, Lipid polymorphism and protein-lipid interactions. 204. P. Martin, A. D. Mehta, A. J. Hudspeth, Negative hair-bundle stiffness Biochim. Biophys. Acta 1376, 353–368 (1998). betrays a mechanism for mechanical amplification by the hair cell. 215. S. M. Bezrukov, Functional consequences of lipid packing stress. Curr. Proc. Natl. Acad. Sci. U.S.A. 97, 12026–12031 (2000). Opin. Colloid Interface Sci. 5, 237–243 (2000). 205. D. Branton, C. M. Cohen, J. Tyler, Interaction of cytoskeletal proteins on 216. R. S. Cantor, The influence of membrane lateral pressures on simple the human erythrocyte membrane. Cell 24, 24–32 (1981). geometric models of protein conformational equilibria. Chem. Phys. 206. E. A. Evans, R. Skalak, Mechanics and thermodynamics of biomem- Lipids 101, 45–56 (1999). branes: Part 1. CRC Crit. Rev. Bioeng. 3, 181–330 (1979). 217. A. Ben-Shaul, Molecular theory of chain packing, elasticity and lipid- 207. E. A. Evans, R. Skalak, Mechanics and thermodynamics of biomem- protein interaction in lipid bilayers. In Handbook of Physics of Biological branes: Part 2. CRC Crit. Rev. Bioeng. 3, 331–418 (1979). Systems, vol. 1: Structure and Dynamics of Membranes, R. Lipowsky 208. W. Rawicz, K. C. Olbrich, T. McIntosh, D. Needham, E. Evans, Effect of and E. Sackmann, Eds. (Elsevier, Amsterdam, 1995), chap. 7, pp. chain length and unsaturation on elasticity of lipid bilayers. Biophys. J. 359–402. 79, 328–339 (2000). 218. R. S. Cantor, Lateral pressures in cell membranes: A mechanism for 209. N. Mobashery, C. Nielsen, O. S. Andersen, The conformational prefer- modulation of protein function. J. Phys. Chem. B 101, 1723–1725 ence of gramicidin channels is a function of lipid bilayer thickness. (1997). FEBS Lett. 412, 15–20 (1997). 219. D. P. Corey, J. Garcia-Anoveros, Mechanosensation and the 210. J. A. Lundbaek, A. M. Maer, O. S. Andersen, Lipid bilayer electrostatic DEG/ENaC ion channels. Science 273, 323–324 (1996). energy, curvature stress, and assembly of gramicidin channels. Bio- 220. J. Garcia-Anoveros, D. P. Corey, The molecules of mechanosensation. chemistry 36, 5695–5701 (1997). Annu. Rev. Neurosci. 20, 567–595 (1997). 211. M. Goulian, O. N. Mesquita, D. K. Fygenson, C. Nielsen, O. S. Ander- 221. T. J. Byers, D. Branton, Visualization of the protein associations in the sen, A. Libchaber, Gramicidin channel kinetics under tension. Biophys. erythrocyte membrane skeleton. Proc. Natl. Acad. Sci. U.S.A. 82, J. 74, 328–337 (1998). 6153–6157 (1985). 212. B. Martinac, O. P. Hamill, Gramicidin A channels switch between stretch 222. We appreciate discussions with G. M. G. Shepherd, who contributed to