The cell biology of touch

A rich variety of somatosensory neurons innervate our skin to initiate the senses of touch and pain (Fig. 1). Discriminative touch is mediated by light-touch receptors, which are activated by innocuous mechanical stimuli. For example, lanceolate endings respond to hair movements, Pacinian corpuscles and Meissner’s corpuscles are vibration receptors that convey textural information, and Merkel cell–neurite complexes encode an object’s spatial features such as edges and curvature. The perception of pain is evoked by nociceptors, which are free nerve endings that respond to noxious stimuli. In addition to these broad categories, numerous classes of somatosensory neurons can be distinguished based on their functional properties and innervation patterns.

Somatosensory neurons share a basic body plan. Their somata are clustered in trigeminal ganglia, near the base of the skull, or dorsal root ganglia (DRG) nestled in each vertebra. Each somatosensory neuron has an axon, called a sensory afferent, which serves as a cellular cable that propagates electrical impulses, or action potentials, from the body to the central nervous system. The peripheral branches of these afferents, which innervate the skin and other organs, transduce sensory stimuli into action potentials.

In the skin, many peripheral afferents terminate in complex end organs whose structures shape their responses to force. For example, Pacinian corpuscles are lamellae-encased rapidly adapting afferents that fire selectively at the onset and offset of a sustained touch (Fig. 1). These lamellae act as mechanical filters to govern adaptation (Loewenstein and Mendelson, 1965); however, recent work suggests that they also release neurotransmitters to shape sensory responses (Pawson et al., 2009). Another rapidly adapting receptor, the Meissner’s corpuscle, is innervated by three distinct types of sensory afferents, which highlights the complexity of touch-sensitive end organs (Paré et al., 2001).

Along with morphology, electrophysiological properties can be used to group touch receptors (Fig. 1). Afferents are broadly classified as Aβ, Aδ, or C-fibers by the speed of action-potential propagation, which is set by myelin thickness. They can be further distinguished by mechanical threshold, adaptation, firing pattern, and modality, or the mechanical stimulus to which they best respond. Most Aβ afferents, which are thickly myelinated, have low mechanical thresholds and are therefore likely to be light-touch receptors. Most unmyelinated C-fiber and thinly myelinated Aδ afferents are thought to be nociceptors based on their high mechanical thresholds and projection patterns to the central nervous system (for review see Smith and Lewin, 2009). Others, including down hair, or D-hair, afferents and low-threshold C-fibers, display mechanical thresholds below the nociceptive range. Although the function of low-threshold C-fibers is not known, they have been proposed to contribute to touch hypersensitivity after injury (Seal et al., 2009) or to an affective, or emotional, component of touch (Olausson et al., 2002; Löken et al., 2009).

Developmental studies have begun to define transcription factors and growth factor pathways that underlie the diversity of touch-receptive afferents (Fig. 1; Luo et al., 2007). For example, some nociceptors require nerve growth factor (NGF) and its receptor TrkA for postnatal survival. Other nociceptors express Runx1, a transcription factor, and Ret, a receptor for glial-derived neurotrophic factor family members. Sensory neurons distinguished by the transcription factor MafA and early Ret expression innervate hair follicles, Pacinian corpuscles and Meissner’s corpuscles (Bourane et al., 2009; Luo et al., 2009).

Most Merkel cell–neurite complexes require neurotrophin-3 (NT-3) and its receptor TrkC for postnatal survival (Airaksinen et al., 1996). These exquisitely sensitive touch receptors mediate slowly adapting type I (SAI) responses (Yoshioka et al., 2001; Woodbury and Koerber, 2007). To properly encode touch stimuli, SAI afferents require the presence of Merkel cells, which are putative sensory cells (Maricich et al., 2009). In striking parallel to hair cells, which are mechanosensory receptors in the inner ear (Schwander et al., 2010), Merkel cells are vertebrate epithelial specializations whose development depends on the transcription factor Atonal 1 (Atoh1; Maricich et al., 2009; Morrison et al., 2009; Van Keymeulen et al., 2009).

Transgenic mice have been engineered to express markers in subsets of touch receptor cells, including light-touch receptors (Hasegawa and Wang, 2008; Bourane et al., 2009; Luo et al., 2009), Merkel cells (Lumpkin et al., 2003), low-threshold C-fibers (Q. Liu et al., 2007; Seal et al., 2009), and C-nociceptors (Stirling et al., 2005; Zylka et al., 2005). Most available markers label multiple touch-receptor classes; however, as the list continues to grow, genetically encoded markers will be a gateway to defining the molecular differences that dictate unique responses in touch-receptor subtypes (Zhang et al., 2002; Haeberle et al., 2004).

In the tiny nematode C. elegans, a repertoire of force-evoked behaviors is initiated by mechanosensory neurons, which comprise ∼10% of the entire nervous system (Fig. 2 A; Goodman, 2006). Touch initiates avoidance behaviors, which include speeding up to escape posterior stimuli, and backing up or head turning to avoid anterior touch. Gentle body touch is transduced by touch receptor neurons that extend neurites along the animal’s length (Fig. 2 A, blue; Sulston et al., 1975; Chalfie and Sulston, 1981). Harsh prodding is detected by distinct neurons with elaborate sensory dendrites that tile the body wall (Fig. 2 A, red; Way and Chalfie, 1989; Chatzigeorgiou et al., 2010). Ciliated mechanosensory neurons innervating the nose mediate touch and physically sense food particles, which impacts foraging behaviors (Fig. 2 A, green; Kaplan and Horvitz, 1993; Sawin et al., 2000; Li et al., 2006; Kindt et al., 2007). Mating behaviors rely on ciliated male-specific neurons called sensory rays (Fig. 2 A, orange; Liu and Sternberg, 1995; Barr and Sternberg, 1999; T. Liu et al., 2007).

Thanks to a wealth of tools for analyzing the C. elegans nervous system, mechanotransduction is best understood in this organism. Central to this success, Chalfie and colleagues devised a simple behavioral assay for gentle-body touch and screened for genetic mutations that selectively caused mechanosensory abnormality (mec) without impairing locomotion (Sulston et al., 1975; Chalfie and Sulston, 1981; Chalfie and Au, 1989). Additional studies have dissected responses to nose touch and harsh body touch (Way and Chalfie, 1989; Colbert et al., 1997; Hart et al., 1999; Chatzigeorgiou et al., 2010). To delineate the cellular basis of mechanosensation, laser ablation has been used to pinpoint neurons required for touch-evoked behaviors. Additionally, anatomical reconstructions have mapped connectivity to define neural networks that link touch sensation to behaviors. Moreover, mechanosensory molecules can be assigned to signaling pathways with physiological approaches, including in vivo imaging and electrophysiology (Fig. 2 B).

Mechanosensory genes in Drosophila have been identified through forward genetic screens for insensitivity to gentle touch, noxious stimuli, and mating song (Kernan et al., 1994; Eberl et al., 1997; Tracey et al., 2003). Like worms and mammals, flies have a sizable assortment of mechanosensory neurons essential for survival (for review see Kernan, 2007; Smith and Lewin, 2009). In addition to touch receptors, Drosophila require auditory receptors to distinguish mating songs, wing strain gauges to fly and proprioceptors to coordinate their six legs. Even in a cushy laboratory setting, mechanosensory mutants perish because they are completely uncoordinated.

The fly’s body is studded with type I ciliated mechanosensory receptors such as chordotonal organs and bristles (Fig. 2; Kernan, 2007). Chordotonal organs, which are stretch receptors attached to the skin, or cuticle, make up the fly’s ear and contribute to proprioception. Bristles serve as the principal proprioceptors and touch receptors. These mechanosensory organs are generally innervated by one sensory neuron, which extends a sensory cilium into overlying structures (Fig. 2 C). Bristles act as levers to transmit force to the cilium attached at its base by the dendritic cap. This compresses the ciliary membrane against the tubular bundle, an array of microtubules at the dendrite’s core. The resulting compression is the putative gating stimulus for mechanotransduction channels, which carry cation-selective, adapting currents remarkably similar to those of vertebrate hair cells (Fig. 2 D).

Type II mechanosensory neurons are nonciliated neurons that innervate the cuticle (Kernan, 2007). Their multidendritic morphology is reminiscent of mammalian somatosensory neurons and C. elegans harsh body-touch receptors (Chatzigeorgiou et al., 2010; Oren-Suissa et al., 2010; Smith et al., 2010). Like mammalian nociceptors, at least some of these neurons sense harsh touch and noxious heat (Tracey et al., 2003; Zhong et al., 2010).

Ion channels are key components of the transduction cascades that convert stimulus energy into membrane potential changes. In most cases, the resulting receptor potential triggers action potentials that transmit sensory information with high fidelity (Fig. 3). Although transduction channels are known for most mammalian sensory modalities, those that underlie touch and hearing have proven astonishingly difficult to identify. In invertebrates, leading candidates fall into the degenerin/epithelial Na⁺ channel (Deg/ENaC) and transient receptor potential (TRP) families.

In mechanosensory cells, transduction channels are thought to be directly gated by force because of their sub-millisecond response times (Corey and Hudspeth, 1979; Walker et al., 2000; O’Hagan et al., 2005; Kang et al., 2010). Like bacterial osmosensitive channels (Box 2), eukaryotic mechanotransduction channels might be stretch-sensitive channels gated by membrane forces (Kung, 2005; Lumpkin and Caterina, 2007). In some mechanosensory cells, transduction channel gating is proposed to require links to the cytoskeleton or extracellular matrix (ECM). Such tethers might couple directly to transduction channels, as they do for mechanosensitive integrins. Alternatively, tethers could control membrane deformation around a stretch-sensitive channel. The tether model is supported by a wealth of genetic and biophysical studies in hair cells (Assad et al., 1989; Vollrath et al., 2007; Schwander et al., 2010).

The best characterized eukaryotic mechanotransduction channel is the MEC-4 complex, which transduces gentle body touch in C. elegans (Fig. 4; Chalfie, 2009). The Deg/ENaC subunits MEC-4 and MEC-10 form the core of this multiprotein complex (Goodman et al., 2002). Channel activity is dramatically enhanced by the accessory subunits MEC-6 and MEC-2, stomatin domain proteins that bind cholesterol (Chelur et al., 2002; Goodman et al., 2002; Huber et al., 2006; Brown et al., 2008). These are essential components, as mutations in each disrupt touch-evoked behaviors. The unc-24 gene encodes a second stomatin domain protein that colocalizes with the MEC-4 complex. Because unc-24 mutations only impair touch responses on a sensitized genetic background, this molecule participates in, but is not required for, touch reception (Zhang et al., 2004).

Although heterologously expressed MEC-4 complexes have not been shown to be force sensitive, compelling evidence argues that they mediate native mechanotransduction currents. Electrophysiology and in vivo imaging showed that mutations in mec-4, mec-6, and mec-2 abolish transduction without disrupting other cellular functions (Suzuki et al., 2003; O’Hagan et al., 2005). Moreover, point mutations in mec-4 and mec-10 alter ion selectivity of native channels (O’Hagan et al., 2005). With a bona fide transduction channel in hand, the next challenge is to develop a mechanistic understanding of force gating.

Invertebrate Deg/ENaC channels are also required for responses to harsh touch. In Drosophila, pickpocket is expressed in type II multidendritic neurons that serve as nociceptors (Adams et al., 1998; Hwang et al., 2007). Disrupting pickpocket expression impairs harsh touch–evoked behaviors (Zhong et al., 2010). Similarly, the expression of mec-10 and degt-1 in C. elegans multidendritic neurons is required for avoidance of harsh prodding (Chatzigeorgiou et al., 2010). Collectively, these findings indicate that Deg/ENaC channels function in touch sensation across species and modalities.

A number of mec-related molecules are expressed in mammalian DRG neurons and their possible roles in touch reception have been examined in knock-out mice (Fig. 3). Disruption of a distant mec-2 relative, stomatin-like protein-3 (SLP3), causes behavioral deficits in texture discrimination and loss of mechanosensitivity in a subset of mouse touch receptors (Wetzel et al., 2007). Although these SLP3-dependent sensory neurons have yet to be identified, this intriguing result points to a conserved role for stomatin domain proteins in touch. In contrast, only subtle changes in touch-evoked responses result from disrupting mammalian DEG/ENaC isoforms called acid-sensing ion channels (ASICs; encoded by amiloride-sensitive cation channel [ACCN] genes; Price et al., 2000, 2001; Drew et al., 2004). Thus, these channels might modulate rather than transduce mechanosensory information in mammals (Fig. 3). Alternatively, these modest phenotypes may reflect redundant gene function.

TRP channels are a diverse class of cation channels implicated in a wide variety of physiological processes, including numerous sensory modalities. In C. elegans, the TRP vanilloid (TRPV) isoforms osm-9 and ocr-2 are expressed in ciliated mechanoreceptors, and their mutations impair nose-touch avoidance, hypertonicity, and chemical stimuli (Colbert et al., 1997; Tobin et al., 2002; Kahn-Kirby and Bargmann, 2006). These proteins colocalize in sensory cilia. Their ciliary localization is interdependent; therefore, these isoforms likely form heteromers. A role for osm-9’s mammalian orthologue TRPV4 in mechanotransduction has also been proposed; however, TRPV4 disruption has only modest effects on touch thresholds (Liedtke and Friedman, 2003; Suzuki et al., 2003). Interestingly, the related Drosophila TRPV isoforms, nanchung (nan) and inactive (iav), are required for hearing but not touch (Kim et al., 2003; Gong et al., 2004).

PKD-2 is another C. elegans TRP channel that localizes to sensory cilia (Barr et al., 2001). This channel is expressed in male-specific neurons. Mating defects result from mutations in this gene and its partner, the PKD-1 homologue lov-1.

TRPN1, which is encoded by the no mechanorecptor potential C (nompC) gene, was the first candidate mechanotransduction channel identified in Drosophila. Mutant flies exhibit defects in touch, hearing, and proprioception (Kernan et al., 1994; Walker et al., 2000). Because these mutants have severely reduced mechanotransduction currents in bristles, TRPN1 is an excellent candidate for a touch transduction channel (Walker et al., 2000). Consistent with this hypothesis, TRPN1 localizes to the distal tips of sensory dendrites in chordotonal and bristle mechanoreceptors (Cheng et al., 2010; Lee et al., 2010). This channel is also expressed in a subset of multidendritic neurons that may function as proprioceptors (Cheng et al., 2010). In a noteworthy parallel, the C. elegans TRPN1 orthologue TRP-4 is expressed in putative proprioceptors and in ciliated mechanosensory neurons involved in foraging behavior (Li et al., 2006). An exciting recent study demonstrates that mechanotransduction currents in these ciliated neurons require functional TRP-4. Furthermore, TRP-4 pore mutations alter the biophysical properties of native transduction currents. Together, these findings strongly support the notion that TRPN1 is a mechanosensory transduction channel in invertebrates. Mammals have apparently adopted a different molecular strategy for touch transduction: TRPN1 homologues are expressed in mechanosensory cells in C. elegans, zebrafish, and amphibians, but they are not found in mammalian genomes (Walker et al., 2000; Sidi et al., 2003; Shin et al., 2005; Li et al., 2006).

TRPA isoforms are also involved in touch reception. For example, Drosophila painless is required for behavioral responses to harsh prodding and noxious heat (Tracey et al., 2003). C. elegans trpa-1 mutants display defects in nose-touch and foraging behaviors (Kindt et al., 2007). Mammalian TRPA1 is likely to modulate the responsiveness of touch-sensitive nociceptors during inflammation (Fig. 3; Lumpkin and Caterina, 2007; Kwan et al., 2009).

Although the molecular identities of mammalian mechanotransduction channels remain mysterious, touch-evoked currents have been studied in cultured DRG neurons (McCarter et al., 1999). In subsets of DRG neurons these currents have different ion selectivities, which suggests that they are carried by discrete ion channel isoforms (Drew et al., 2002; Hu and Lewin, 2006; Rugiero et al., 2010). Like touch receptors in vivo, cultured DRG neurons display a variety of adaptation profiles (Rugiero et al., 2010). Unlike hair cells, adaptation in cultured DRG neurons is Ca²⁺ independent. Collectively, these studies suggest that mechanotransduction in mammalian cells is mediated by distinct molecular mechanisms. An exciting recent study has identified a novel ion channel class, the piezo family, which is required for touch-evoked currents in cultured DRG neurons (Coste et al., 2010). How these touch-evoked responses in vitro relate to mechanotransduction in vivo is an important, open question.

Signaling pathways downstream of transduction govern touch sensitivity by altering membrane excitability (Fig. 3; Foulkes and Wood, 2008). Loss-of-function mutations in SCN9A, which encodes Na_V1.7, a nociceptor-specific voltage-activated Na⁺ channel, cause a dramatic loss of sensitivity to painful stimuli in humans and mice (Nassar et al., 2004; Cox et al., 2006). Conversely, gain-of-function mutations in this gene lead to pain hypersensitivity. Increased touch responsiveness has also been observed in mice lacking two-pore K⁺ channels, which set resting membrane potentials (Nöel et al., 2009). Related two-pore K⁺ channels are proposed to be the molecular targets of sanshool, a compound found in Schezuan peppercorns that activates touch receptors and induces a tingling sensation in humans (Lennertz et al., 2010). Although intensive efforts remain focused on identifying mammalian mechanotransduction channels, these studies underscore the possibility that other sensory molecules may be targets for therapeutic development.

Like hair cells, many touch receptors have prominent cytoskeletal specializations. In invertebrates, these specializations are microtubule based. Modified cilia serve as sensory dendrites in Drosophila type I mechanosensory neurons, as well as in C. elegans nose-touch and male-specific neurons (Goodman, 2006; Kernan, 2007). Thus, sensory defects result from mutations in genes that disrupt ciliogenesis, intraflagellar transport, or ciliary protein localization (Perkins et al., 1986; Kernan, 2007; Bae et al., 2008). Although C. elegans body-touch receptor neurons lack sensory cilia, their mechanosensitive processes are filled with highly cross-linked, 15-protofilament microtubules (Fig. 4; Chalfie and Thomson, 1979, 1982).

Analysis of two mec genes demonstrates that these unique structures are essential for touch-evoked behaviors. Mec-12 and mec-7 encode α- and β-tubulins that form 15-protofilament microtubules (Savage et al., 1989; Fukushige et al., 1999). These genes are highly expressed in touch receptor neurons, consistent with the observation that 15-protofilament microtubules are exclusive to these cells (Chalfie and Sulston, 1981; Chalfie and Thomson, 1982; Hamelin et al., 1992; Fukushige et al., 1999). Many mec-7 and mec-12 mutant alleles cause 15-protofilament microtubules to be replaced by typical microtubules (Chalfie and Thomson, 1982). Mutations in these two genes also render worms touch insensitive (Chalfie and Sulston, 1981; Chalfie and Au, 1989). Based on such genetic evidence, early tether models posited that attachments between the MEC-4 complex, the cytoskeleton, and the ECM are necessary for mechanotransduction (Gu et al., 1996).

Recent physiological and structural data indicate that this model must be revised (Fig. 4). Importantly, transduction currents are attenuated, but not abolished, by mec-7 and mec-12 mutations. These data demonstrate that 15-protofilament microtubules, although necessary for touch-evoked behaviors, are not required for transduction channel activation (O’Hagan et al., 2005; Bounoutas et al., 2009). Moreover, functional MEC-4 complexes are unlikely to be attached to microtubules because the densities of MEC-4 puncta and juxtamembrane microtubules are not correlated, and these structures do not colocalize at the plasma membrane (Emtage et al., 2004; Cueva et al., 2007). Instead, microtubule bundles link to the plasma membrane at sites distinct from MEC-4 puncta. One candidate for these links is Echinoderm microtubule-associated protein-like protein-1 (ELP-1), which is expressed in cells that adhere to the cuticle, including body-touch receptor neurons, nose-touch neurons, and male-specific neurons (Hueston et al., 2008). Notably, disrupting elp-1 impairs touch sensitivity.

Together, these findings suggest that the cytoskeleton impacts force sensitivity without direct attachments to the MEC-4 complex. Alternative models posit that the MEC-4 complex is a stretch-activated channel and that microtubule bundles indirectly participate in gating by altering membrane forces during channel activation or adaptation (Cueva et al., 2007; Bounoutas et al., 2009).

Along with a structural role in force sensing, wild-type microtubules are required for proper trafficking of mechanotransduction proteins. Strong loss-of-function mutations in mec-7 and mec-12 disrupt overall protein levels and the distribution of MEC-4 puncta (Emtage et al., 2004; Bounoutas et al., 2009). Although transduction channels must insert into the plasma membrane to activate neurons, immunoelectron microscopy indicated that about half of the MEC-4 complexes are linked to intracellular microtubules in the absence of membrane-bound vesicles (Cueva et al., 2007). This intriguing observation suggests that membrane proteins might traffic along microtubules via nonvesicular transport in touch receptor neurons.

Compared with hair cells and C. elegans touch receptors, little is known about the role of the cytoskeleton in vertebrate touch reception. One study of cultured DRG neurons found that cytochalasin B, which inhibits actin polymerization, attenuates mechanosensitive currents (Drew et al., 2002). Whether this effect is through alterations in the cortical cytoskeleton or microfilament-based specializations is unclear. In fact, cytoskeletal specializations have not yet been described in mammalian somatosensory afferents.

In contrast, Merkel cells have conspicuous microvilli, which are coupled to overlying epidermal cells by electron-dense filaments (Toyoshima et al., 1998). Intriguingly, these processes are enriched in espin, an actin-binding protein found in hair cell stereocilia and other sensory microvilli (Sekerková et al., 2004). Based on their structural similarity to stereocilia, the Merkel cell’s microvilli have been proposed to be sites of mechanotransduction (Iggo and Findlater, 1984); however, functional support for this model is lacking.

Forces exerted between metazoan cells and the ECM play a fundamental role in the development and function of complex tissues (for review see Ingber, 2006). Different tissues express numerous ECM components and their receptors, most notably integrins. The ability of cells to respond appropriately to their local matrix environment is essential for cell migration, differentiation, and survival (for review see Legate et al., 2009).

The importance of ECM proteins to touch sensation is best understood in C. elegans touch receptor neurons. Their mechanosensitive neurites are embedded in an electron-dense ECM and surrounded by epidermal cells, which are attached to the cuticle at periodic hemidesmosomal-like structures (Emtage et al., 2004). In these neurons, touch responses require specialized ECM components as well as integrin signaling (Calixto et al., 2010).

Three mec genes encode essential ECM components (Fig. 4; Chalfie and Sulston, 1981; Du et al., 1996). Touch receptor neurons express MEC-1 and MEC-9, which are secreted proteins containing multiple epidermal growth factor (EGF)–like domains and Kunitz-like repeats (Du et al., 1996; Emtage et al., 2004). A unique collagen encoded by mec-5 is produced by adjacent epidermal cells (Du et al., 1996). Interestingly, these proteins are distributed in puncta that overlap with MEC-4 complexes (Emtage et al., 2004). Mutations in these matrix-component genes disrupt the subcellular distribution of MEC-4 complexes. In contrast, the punctate localization of ECM components is unaffected by mec-4 mutations. Together, these data suggest that the ECM properly localizes transduction channels. Based on incomplete colocalization of MEC-5 and MEC-4 complexes at the ultrastructural level, Cueva et al. (2007) have argued that MEC-5 is unlikely to function as a gating tether. Whether integrins or other linking proteins play a direct role in transduction channel activation remains to be determined.

In Drosophila type I sensory organs, connections between mechanosensory dendrites and the cuticle are essential for transduction (Kernan, 2007). As in C. elegans, mechanosensory processes are encircled by supporting cells that secrete an electron-dense ECM, termed the dendritic sheath or cap (Fig. 2 C). One component of this matrix is encoded by the NompA gene (Chung et al., 2001). Behaviorally, NompA mutants are touch insensitive and deaf. They also lack bristle mechanoreceptor responses (Kernan et al., 1994).

Three lines of evidence argue that NompA is a structural element of the dendritic cap (Chung et al., 2001). First, NompA mutants have disorganized dendritic caps and detached mechanosensory dendrites. Second, NompA includes a large, secreted domain that localizes to the cap. Third, this region contains a zona pellucida domain, which is commonly found in ECM proteins. In a noteworthy parallel, zona pellucida domain proteins called tectorins are major components of the tectorial membrane, which is essential for mechanical stimulation of cochlear hair cells (Killick et al., 1995). Because dendrite attachment is disrupted in NompA mutants, it is clear that NompA, like tectorins, plays a key structural role in mechanotransduction. Whether it also directly participates in transduction channel gating is still unclear.

Although the importance of the ECM in development of mammalian somatosensory neurons has long been recognized, a possible role in sensory transduction is only now being explored. In DRG neurons, interactions between specific integrins and the ECM promote neurite extension during development and injury-induced regeneration (Tomaselli et al., 1993; Andrews et al., 2009). An intriguing recent study implicates molecularly distinct extracellular contacts in mammalian touch reception (Hu et al., 2010). In DRG neurons in vitro, Hu et al. (2010) observed that touch-sensitive neurites are connected to laminin substrates via 100-nm proteinaceous filaments. Reminiscent of pioneering studies that revealed the Ca²⁺ sensitivity of hair cell tip links (Assad et al., 1991), a battery of treatments was tested to define those that disrupted 100-nm filaments and abolished mechanosensitivity in putative light-touch receptors. These filaments are sensitive to furin proteases but are resistant to treatments that disrupt integrins, cadherins, and glycosyl phosphatidylinositol-anchored proteins. This unique sensitivity profile indicates that 100-nm tethers are distinct from integrins as well as cadherin-based tip links in hair cells. Defining the molecular identity of these junctional proteins and determining their role in mechanotransduction will be exciting next steps.

Although the epidermis’s mechanical properties are key for transmitting force from the skin’s surface to touch receptors, several lines of evidence suggest that epidermal cells play more than a mere structural role. In C. elegans touch receptor neurons, structural attachment to the cuticle is not required for touch sensitivity, as demonstrated by the touch responsiveness of him-4 mutants that lack hemidesmosomal connections to the cuticle (Vogel and Hedgecock, 2001). Instead, the epidermis secretes essential ECM molecules that are proposed to position transduction channel complexes in sensory neurons (Emtage et al., 2004).

Mammalian epidermal cells are ideally poised to participate in somatosensory signaling (Lumpkin and Caterina, 2007). The epidermis is innervated by sensory afferents that transduce noxious mechanical stimuli (Zylka et al., 2005; Cavanaugh et al., 2009) and by Merkel cell–neurite complexes (Johnson, 2001). Keratinocytes, which are the principal cells of the epidermis, Merkel cells, and the lamellae of Pacinian corpuscles express neurotransmitters that have the potential to tune the touch sensitivity of afferents (Halata et al., 2003; Lumpkin and Caterina, 2007; Pawson et al., 2009). Although keratinocytes and sensory afferents do not form synapses, their proximity could allow rapid paracrine signaling.

Notably, mammalian epidermal cells express sensory ion channels implicated in mechanotransduction, such as TRPV4 and TRPA1 (Liedtke et al., 2000; Lee and Caterina, 2005; Kwan et al., 2009). Mechanically evoked firing properties of light-touch receptors are altered in TRPA1 knockout mice, leading Kwan et al. (2009) to propose that TRPA1 influences touch sensitivity through a modulatory role in keratinocytes. This model is readily testable with tissue-specific knock-outs.

In 1875, Merkel posited that his eponymous cells act as touch receptors and several lines of evidence support this notion (Merkel, 1875). Merkel cells form synaptic contacts with sensory neurons, express numerous presynaptic proteins, and are intrinsically force sensitive in vitro (Haeberle et al., 2004, 2008; Lumpkin and Caterina, 2007; Boulais et al., 2009). Moreover, Merkel cells are required for touch-evoked SAI responses (Maricich et al., 2009). In Atoh1 knock-out mice, Merkel cells fail to develop, but SAI sensory afferents still innervate their proper receptive fields. Electrophysiological analysis of these mice revealed a complete loss of SAI responses in the absence of Merkel cells (Maricich et al., 2009). As predicted, light-touch receptors that innervate other end organs displayed normal mechanosensitivity, demonstrating that Atoh1 is selectively required for touch responses in Merkel cell–neurite complexes. The effects of postnatal loss of Merkel cells are less clear. Some studies report that Merkel cell loss impaired SAI responses (Ikeda et al., 1994; Senok et al., 1996), whereas others found little impact on slowly adapting responses (Mearow and Diamond, 1988; Mills and Diamond, 1995; Kinkelin et al., 1999). Further studies are needed to determine whether this discrepancy is due to methodological differences. Alternatively, Merkel cells might be required for proper development but not maintenance of functional SAI afferents.

Another key question is whether Merkel cell synapses are excitatory or whether they release neuromodulators that tune the sensitivity of touch-receptive afferents. Functional studies that blocked synaptic transmission have led to conflicting models (Fagan and Cahusac, 2001; Halata et al., 2003; Cahusac et al., 2005; Cahusac and Senok, 2006; Cahusac and Mavulati, 2009). Surprisingly, immunostaining has localized several neurotransmitter receptors to Merkel cells rather than to their SAI afferents (Beiras-Fernández et al., 2004; Cahusac et al., 2005; Tachibana and Nawa, 2005; Tachibana et al., 2005). These findings suggest that neurotransmitters may act on Merkel cells themselves. Thus, the role of Merkel cells in touch reception remains to be determined.

Touch is a complex sense comprising a diversity of modalities, and we have just begun to glimpse the underlying cellular principles. Common themes have emerged from histological, physiological, and behavioral studies of genetically tractable organisms.

First, mechanosensory signaling relies on specialized cellular morphologies. Across invertebrate and vertebrate species, noxious touch is transduced by free nerve endings. In contrast, light-touch receptors display a range of morphologically complex end organs. These are largely microtubule based in invertebrate neurons. Elucidating the role of cytoskeletal proteins in vertebrate touch reception awaits future studies.

Second, tissue mechanics and nonsensory cells shape responses to mechanical stimuli. Recent studies propose that human tactile acuity is influenced by skin mechanical properties such as fingertip size, epidermal stiffness, and the spacing of fingerprint ridges (Gerling and Thomas, 2008; Peters et al., 2009; Scheibert et al., 2009). For many light-touch receptors, elaborate accessory structures govern tuning and sensitivity. Epidermal cells provide structural attachments, secrete specialized ECM components, and might affect touch-evoked responses through the release of neuroactive compounds. The question remains as to whether Merkel cells and keratinocytes actually transduce mechanical stimuli or whether they play a modulatory role.

Third, excitatory ion channels are central to touch reception. All of the invertebrate mechanotransduction channels identified through unbiased genetic screens fall into the Deg/ENaC and TRP channel families. Understanding their mechanisms of force gating will require biophysical insights such as high-resolution protein structures. In vertebrate mechanosensory cells, an emerging picture indicates that the molecular details of transduction differ substantially. Intensive studies of mammalian Deg/ENaC and TRP channels have failed to demonstrate a fundamental role for these channels in cutaneous mechanotransduction, although they modulate touch responsiveness. The modest touch deficits of knock-out mice might reflect the genetic redundancy of transduction channels, the functional overlap of touch-receptor cells that use distinct transduction mechanisms, or the involvement of novel ion channels, such as the piezo family. Distinguishing between these possibilities will require techniques borrowed from the invertebrate playbook, including selective markers for different touch receptors, feasible approaches for recording transduction currents, innovative behavioral assays for tactile discrimination, and unbiased molecular screens in mammals and zebrafish (Granato et al., 1996; Ribera and Nüsslein-Volhard, 1998; Low et al., 2010). These approaches hold promise for unraveling the molecular complexity of touch, which is an essential step in dissecting the neuronal code of this enigmatic sense.

We thank Drs. Martin Chalfie and Jennifer Garrison for comments on the manuscript and Dr. Scott Wellnitz for discussions. Portions of this manuscript were prepared in the Departments of Neuroscience and Molecular Physiology and Biophysics, Baylor College of Medicine, Houston TX.

The authors are supported by NIAMS grant AR051219 (to E.A. Lumpkin) and a McNair Scholar Award (to A.M. Nelson). We apologize to those whose relevant work was not discussed due to space constraints.

The authors declare no conflicts of interest.

Other reviews in this series are: The cell biology of hearing (Schwander et al. 2010. J. Cell Biol. doi:10.1083/jcb.201001138), The cell biology of taste (Chaudhari and Roper. 2010. J. Cell Biol. doi: 10.1083/jcb.201003144), and The cell biology of vision (Sung and Chuang. 2010. J. Cell Biol. doi: 10.1083/jcb.201006020).