Hearing in Drosophila: Development of Johnston's organ and emerging parallels to vertebrate ear development


  • Grace Boekhoff-Falk

    Corresponding author
    1. Department of Anatomy, University of Wisconsin-Madison, Medical School, Madison, Wisconsin
    • Department of Anatomy, University of Wisconsin-Madison, Medical School, Madison, WI 53706
    Search for more papers by this author
    • Dr. Boekhoff-Falk's former surname was Panganiban.


In this review, I describe recent progress toward understanding the developmental genetics governing formation of the Drosophila auditory apparatus. The Drosophila auditory organ, Johnston's organ, is housed in the antenna. Intriguingly, key genes needed for specification or function of auditory cell types in the Drosophila antenna also are required for normal development or function of the vertebrate ear. These genes include distal-less, spalt and spalt-related, atonal, crinkled, nanchung and inactive, and prestin, and their vertebrate counterparts Dlx, spalt-like (sall), atonal homolog (ath), myosin VIIA, TRPV, and prestin, respectively. In addition, Drosophila auditory neurons recently were shown to serve actuating as well as transducing roles, much like their hair cell counterparts of the vertebrate cochlea. The emerging genetic and physiologic parallels have come as something of a surprise, because conventional wisdom holds that vertebrate and invertebrate hearing organs have separate evolutionary origins. The new findings raise the possibility that auditory organs are more ancient than previously thought and indicate that Drosophila is likely to be a powerful model system in which to gain insights regarding the etiologies of human deafness disorders. Developmental Dynamics 232:550–558, 2005. © 2005 Wiley-Liss, Inc.


Most animals have auditory organs. Auditory organs are specialized mechanosensory structures that vary among animal species with respect to frequencies heard and acuteness of hearing. Auditory organ morphologies necessarily vary with their functional attributes. Both vertebrate and invertebrate auditory organs are thought to have evolved from primitive mechanosensors (reviewed in Eberl, 1999; Eberl et al., 2000; Eddison et al., 2000; Fritzsch and Beisel, 2001; Caldwell and Eberl, 2002; Jarman, 2002; Todi et al., 2004), but the nature of the ancestral structure and the evolutionary trajectories followed in distinct animal lineages remain unknown. In particular, we do not know how many types of mechanosensor existed in the protostome–deuterostome ancestor (PDA; Erwin and Davidson, 2002) from which insects and vertebrates evolved or whether the PDA had an auditory organ.

Modern vertebrates possess at least 14 types of mechanosensors (Kandel et al., 2000), whereas Drosophila possess 4 (reviewed in Jarman, 2002). A single ancestral structure could have been modified in the respective lineages to generate the present-day diversity. Alternatively, multiple types of mechanosensors may have existed in the PDA. However, because the evolution of vertebrates and insects involved independent transitions of hearing animals from water to land, adaptation of auditory organs to sense airborne sounds occurred independently at least twice. Consideration of the developmental genetic differences among Drosophila mechanosensors, together with emerging parallels between vertebrate and invertebrate auditory organ development lend increasing support to the hypothesis that the PDA had multiple types of mechanosensors and that a particular class of these gave rise to both invertebrate and vertebrate auditory organs.


The four major types of mechanosensors found in modern insects such as Drosophila are bristle organs, campaniform sensillae, multiple dendritic neurons, and chordotonal organs (CHOs; reviewed in Jarman, 2002). Each follows a distinct developmental genetic program. Studies of bristle organs (reviewed in Jarman, 2002) have provided significant insights into vertebrate as well as invertebrate mechanosensation; bristle organs often are analogized to hair cells of the vertebrate ear. However in most insects, including Drosophila, it is subsets of the CHOs that are specialized for hearing. As discussed in detail below, these CHOs bear some striking genetic and structural similarities to hair cells that are not shared by bristle organs. Studies of Drosophila auditory CHOs, therefore, may enhance our understanding of vertebrate as well as invertebrate hearing.


Hearing in Drosophila is carried out by the antenna (Fig. 1A; reviewed in Eberl, 1999; Eberl et al., 2000; Göpfert and Robert, 2001, 2002; Caldwell and Eberl, 2002; Jarman, 2002; Todi et al., 2004). Near-field sound waves of the appropriate amplitude (∼80–95 dB) and frequency (∼160 Hz) cause vibration of the arista (ar), which is asymmetrically positioned on the third antennal segment (a3). Movement of the arista exerts a torque on a3, which then rotates by means of a flexible joint connecting it to the second antennal segment (a2). A large CHO called Johnston's organ (JO) housed within a2 senses movement at the a2/a3 joint. In this review, I refer to JO as an “auditory CHO” and other CHOs in the Drosophila larval and adult body, which are used for proprioception, as “nonauditory CHOs.” The building blocks of all CHOs are the scolopidia. The scolopidia of auditory and nonauditory CHOs are similar, but not identical, in morphology and cell type composition (discussed below and in Dambly-Chaudiere and Ghysen, 1986; Bodmer and Jan, 1987; Eberl et al., 2000; Caldwell and Eberl, 2002; Todi et al., 2004).

Figure 1.

Anatomy of a wild-type antenna. A: Wild-type adult antenna with first antennal segment (a1), second antennal segment (a2), third antennal segment (a3), and arista indicated. Also labeled is the flexible joint between a2 and a3 (arrow). B: Schematic of the interior of a wild-type second antennal segment (from Caldwell and Eberl, 2002). a2 contains more than 150 scolopidial subunits that together constitute Johnston's organ. Only six of these subunits are shown.

JO is the largest CHO in the Drosophila body, consisting of more than 150 scolopidia (Fig. 1B). Each scolopidium possesses an apical attachment to the a2/a3 joint cuticle and a basal attachment to more proximal a2 cuticle (reviewed in Eberl, 1999; Eberl et al., 2000; Caldwell and Eberl, 2002; Jarman, 2002; Todi et al., 2004). The scolopidia are deformed in response to joint rotation. This deformation is thought to result in the opening of ion channels in scolopidial neuron dendrites, an influx of ions, and, ultimately, in the propagation of an electrical signal by means of the antennal nerve to the auditory center of the brain. The composition of the fluid bathing scolopidial neuron dendrites in JO or other CHOs has not been analyzed. By analogy to bristle organ and campaniform sensilla lymph (Küppers, 1974; Thurm and Küppers, 1980), it is speculated to be potassium-rich (reviewed in Eberl, 1999).


Adult Drosophila antennae, like other adult Drosophila cephalic and thoracic cuticular structures, arise from primordia termed imaginal discs (reviewed in Cohen, 1993). Cells constituting the imaginal discs are allocated and internalized during embryogenesis, proliferate, and undergo extensive patterning during the three larval instars and terminally differentiate during pupal stages. The Drosophila antennal imaginal disc forms as a cluster of approximately 10 cells and reaches approximately 10,000 cells before differentiation.

More is known developmentally and structurally about nonauditory CHOs and their scolopidia than about JO. For instance, each scolopidium in a nonauditory larval body wall CHO consists of four cell types: neuron, scolopale cell, cap cell, and ligament cell (reviewed in Eberl, 1999; Jarman, 2002). The four cells of a larval scolopidium are clonally related, deriving from a single sense organ precursor (SOP) that expresses the proneural gene atonal (ato; Jarman et al., 1993). Each scolopidium is attached both apically and basally to cuticle. The basal attachment is mediated by a ligament cell. In contrast, JO scolopidia typically possess two neurons instead of one, whether the basal attachment is mediated by a ligament cell remains unclear (Caldwell and Eberl, 2002; Todi et al., 2004), and it is not known whether the cells within a JO scolopidium are clonally related.

Similarities between the scolopidia of both JO and nonauditory CHOs include the presence of scolopale cells. Scolopale cells contain actin-rich rods (“scolopales”) proposed to provide a straightening force along the length of the scolopidia (Todi et al., 2004). Scolopale cells ensheathe the neuronal dendrites and secrete the extracellular dendritic cap that mediates apical attachment. One component of the cap is NompA, a tectorin-like molecule (Chung et al., 2001). Other similarities between scolopidia of both JO and nonauditory CHOs include the long ciliated dendrites of constituent neurons and the expression of ato in scolopidial precursors.

While the architectural differences between auditory and nonauditory CHOs noted here may reflect differences in functionality, there are several potential caveats. First, the distinctions between audition and other types of mechanosensation are not clear-cut and animals lacking canonical auditory organs are often sensitive and responsive to sound (Gillot, 1995). Second, the comparisons are of adult auditory CHOs to larval nonauditory CHOs. More useful for understanding how an auditory organ differs from other CHOs would be comparisons of serial homologs such as JO and the adult femoral CHO. However, parallel data sets do not yet exist for these CHOs.


Specification and Differentiation

Several developmental genetic differences between nonauditory CHOs and JO have been described. For example, three genes that inhibit nonauditory CHO differentiation, spalt (sal), spalt-related (salr), and cut (ct), are required for JO differentiation. JO scolopidia in sal/salr double mutants form but then degenerate (Dong et al., 2003); ct mutants exhibit defects in JO architecture (Ebacher and Boekhoff-Falk, unpublished observations); and both sal/salr and ct mutants are completely deaf by electrophysiological criteria (Dong et al., 2003; Ebacher, Todi, Eberl, and Boekhoff-Falk, manuscript in preparation).

CHO precursor cells, like those of other neural structures in Drosophila, arise from “proneural clusters” of cells. Each proneural cluster is characterized by the expression of one or more proneural genes. Within each proneural cluster, one or a few cells become the neural precursor or SOP (reviewed in Skeath, 1999). The nascent SOP maintains proneural gene expression and sends out inhibitory signals that extinguish proneural gene expression in the surrounding cells. Proneural gene function in the SOP typically is required for specification of neural fates and for specification of particular neural subtypes. In the case of CHOs, including JO, the relevant proneural gene is atonal (ato; Jarman et al., 1993). Loss of ato function results in the failure of CHOs, including JO, to differentiate. Thus, whereas Ato is critical for CHO differentiation, the presence or absence of Ato protein per se does not distinguish auditory from nonauditory CHOs.

However, the context in which Ato protein functions is different between auditory and nonauditory CHOs, and Ato could have distinct suites of target genes in the two types of CHOs. For instance, ato is coexpressed with distal-less (dll) and homothorax (hth) in the developing JO (Dong et al., 2002). Nonauditory CHOs do not express dll (Cohen, 1990) or require hth function (Kurant et al., 1998). Thus, Ato could cooperate with Dll and Hth to regulate JO-specific targets.

The relationship between sal/salr and ato also differs between auditory and nonauditory CHOs. In the developing JO, these genes are coexpressed and function in parallel (Dong et al., 2002). For nonauditory CHOs, sal and salr are inhibitory and loss of sal/salr in cells surrounding the CHOs results in an increased number of scolopidia (Elstob et al., 2001; Rusten et al., 2001). Although it has not been tested directly, it is likely that the absence of sal/salr surrounding these nonauditory CHOs leads to an increased number of ato-expressing SOPs.

Other unique features of JO development are that the homeodomain transcription factor encoded by ct is coexpressed with ato (Dong et al., 2002) and that ct is required for JO differentiation (Ebacher, Todi, Eberl and Boekhoff-Falk, unpublished observations). This finding is in marked contrast to nonauditory CHOs, where ato but not ct is expressed and required during development, and ectopic ct expression precludes CHO fate specification (Bodmer et al., 1987; Blochlinger et al., 1991; Jarman et al., 1993; Jarman et al., 1995; Merritt, 1997; Jarman and Ahmed, 1998). ct instead is needed for the differentiation of a second type of mechanosensor, the external sense organs or bristles (Bodmer et al., 1987).

In summary, ato is coexpressed with dll, hth, sal, salr, and ct in the developing JO, and Ato protein may cooperate (directly and/or indirectly) with the products of these other transcription factor-encoding genes to regulate JO differentiation. In nonauditory CHOs, the expression and interactions among these genes are distinct.

Intriguingly, sal, dll, and ato homologs have been implicated in vertebrate ear development (Kohlhase et al., 1998, 2002; Acampora et al., 1999; Bermingham et al., 1999; Depew et al., 1999; Merlo et al., 2002; Robledo et al., 2002; Solomon and Fritz, 2002), and human deafness syndromes are caused by mutations in sal homologs (Kohlhase et al., 1998, 2002) and dll homologs (Crackower et al., 1996; Ignatius et al., 1996; Mishra et al., 2000; Tackels-Horne et al., 2001). There are no known deafness syndromes associated with mutations in human ato homologs, but mouse knockouts lack both the mechanosensory hair cells of the ear and the mechanoreceptive Merkel cells of the skin (Bermingham et al., 1999).


For mechanosensors such as the JO to receive and transduce information, a wide variety of molecules are required. These include extracellular, membrane, and intracellular proteins, as well as ions and neurotransmitters. Here, I focus on a handful of cytoskeletal components, channel and extracellular proteins whose vertebrate homologs also are essential for audition and discuss parallels between Drosophila and vertebrate auditory physiology.

One such cytoskeletal component is encoded by the Drosophila crinkled (ck) gene. ck encodes the unconventional myosin VIIA (Todi et al., 2004, and references therein). ck mutants are ataxic, a condition often indicative of deficits in nonauditory CHO function, and the actin bundles in their scolopales exhibit a splayed phenotype (Dong and Boekhoff-Falk, unpublished observations; Fig. 2). ck mutants are deaf by electrophysiological criteria (Todi et al., 2004, and references therein). ck expression is not limited to CHOs and JO but can be detected in bristle organs and olfactory sensilla that also possess ciliated neurons and in retinal neurons that contain dense apical arrays of microvilli (rhabdomeres; Andrews and Boekhoff-Falk, unpublished observations). Like Ck, vertebrate myosin VIIA is found in a variety of ciliated cells, including those of the kidney, lung, testis, retina, and cochlea as well as in the apical microvilli of pigmented retinal cells (Hasson et al., 1995; Wolfrum et al., 1998). Mutations in human myosin VIIA cause Usher syndrome type 1B (Weil et al., 1995), which is characterized by deafness and retinitis pigmentosa (reviewed in Keats and Corey, 1999).

Figure 2.

crinkled/myosin VIIA is required for scolopale architecture. A: The expression of crinkled/myosin VIIA in a pupal second antennal segment (a2) as visualized using a ck-lacZ enhancer trap line. crinkled is expressed primarily in scolopale cells at this stage. B,C: Wild-type (B) and crinkled mutant (C, ck[13]/ck[13]) second antennal segments stained with rhodamine–phalloidin to visualize the scolopales. The basal ends of the scolopales (arrows) are splayed in the mutant (C) compared with wild-type (B).

Transient receptor potential (TRP) proteins constitute a superfamily of ion channels. Most TRP family members are nonselective cation channels. A few are selective for monovalent cations, whereas several others are selective for calcium (reviewed in Clapham, 2003; Corey, 2003). Three TRP family members have been implicated in Drosophila mechanotransduction (Kernan et al., 1994; Eberl et al., 2000; Littleton and Ganetzky, 2000; Kim et al., 2003; Gong et al., 2004). The first, encoded by no mechanoreceptor potential C (nompC) belongs to the TRPN subfamily. nompC was identified in a genetic screen for touch insensitivity (Kernan et al., 1994) and subsequently was found to be essential for mechanotransduction in insect bristle organs, while playing a less important role in JO transduction (Kernan et al., 1994; Eberl et al., 2000; Walker et al., 2000). A homolog of nompC recently was shown to be essential for mechanotransduction in zebrafish hair cells (Sidi et al., 2003). However, mouse and human homologs have not been found and may have been lost from the mammalian lineage. Both vertebrate hair cells and the dendrites of bristle organ neurons are bathed in high potassium solutions (Küppers, 1974; Thurm and Küppers, 1980), and TRPN channels such as NompC are closely related to TRPM channels that are selective for monovalent cations (reviewed in Clapham, 2003). However, whether NompC functions as a potassium channel is unknown.

Two TRPV family members have been characterized in Drosophila. They are Nanchung (Nan; Kim et al., 2003) and Inactive (Iav; Gong et al., 2004). TRPV channels function in the transduction of a variety of sensory stimuli in nematodes and mammals, including heat, pain, osmotic stress, olfaction, and mechanosensation (reviewed in Mutai and Heller, 2003). In cultured cells, both Nan and Iav function as calcium channels activated by hypotonic stress (Kim et al., 2003; Gong et al., 2004). In the fly, nan and iav are coexpressed in the neurons of both JO and nonauditory CHOs and both are essential for auditory transduction (Kim et al., 2003; Gong et al., 2004). Intriguingly, a TRPV channel in worms (OSM-9) is expressed in a variety of ciliated sensory structures and is required for transduction of osmotic, proprioceptive, and olfactory information (Colbert et al., 1997), whereas the mouse TRPV4 channel is expressed in both inner ear hair cells and Merkel cells (Liedtke et al., 2000). Together with the Drosophilanan and iav data, these findings suggest that TRPV channels were used by ancestral mechanosensors.

Iav is more closely related to Caenorhabditis elegans OSM-9 than is nan (Adams et al., 2000; Littleton and Ganetzky, 2000; Corey, 2003; Kim et al., 2003). osm-9 is expressed in and required for normal functioning of most mechanosensory and chemosensory amphid neurons of the worm (Colbert et al., 1997) and has been proposed to form different types of heterotetramers with other TRPV proteins whose expression is more restricted (Tobin et al., 2002). Both Iav and Nan proteins are localized to the proximal portion of the ciliated dendrite, and neither protein can be detected in the dendrite in the absence of the other (Gong et al., 2004). This finding is consistent with Iav and Nan forming heteromeric complexes. It, therefore, will be important to evaluate Iav biochemically to test whether it forms functional heteromers with other Drosophila TRP channels such as Nan.

The no mechanoreceptor potential A (nompA) was identified in the same genetic screen for touch insensitivity as nompC (Kernan et al., 1994). nompA subsequently was shown to encode an extracellular protein with zona pellucida (ZP) and plasminogen N-terminal domains (Chung et al., 2001). nompA is expressed by the scolopale cells of both JOs and nonauditory CHOs and in the analogous ensheathing cells of bristle organs (Chung et al., 2001). NompA protein is localized to the dendritic cap that mediates connection between the sensory process of the neuron and the overlying cuticle or bristle (Chung et al., 2001). These attachments are defective in nompA mutants, precluding JO, nonauditory CHO, and bristle function in mechansensation (Chung et al., 2001). Tectorins, which are components of the cochlear extracellular matrix, also possess ZP domains (Killick et al., 1995; Legan et al., 1997). ZP domains have been implicated in the formation of fibrillar extracellular aggregates (Jovine et al., 2002), and tectorin mutations lead to human deafness, most likely due to alterations in the properties of the cochlear extracellular matrix (reviewed in Petersen, 2002). Thus, although there are conspicuous morphological differences between the Drosophila JO and the vertebrate cochlea, there may be similarities in the extracellular environments in which they function.

Prestin is a member of the solute carrier 26 (SLC26) family. SLC26 proteins serve diverse functions as sulfate transporters, chloride–iodide transporters, and chloride–carbonate exchangers (reviewed in Alper et al., 2001; Meech and Holley, 2001; Dallos and Fakler, 2002). In the mammalian cochlea, prestin, which also is called SLC26A5, is expressed in outer hair cells where it serves as a molecular motor that drives mechanical amplification of auditory signals (Zheng et al., 2000). It was thought that this type of amplification did not occur in lower tetrapods, let alone in more distantly related taxa. It, therefore, came as a surprise to find prestin homologs expressed in the ears of zebrafish, which are nontetrapod vertebrates (Weber et al., 2003). At least as astonishing was the identification of SLC26 family members in insect arthropods such as mosquitoes and Drosophila and the localization of their transcripts to their JOs (Weber et al., 2003). Although the functions of zebrafish prestin and insect SLC26 remain to be elucidated, the presence of these molecules in auditory structures is consistent with an ancestral role for SLC26 family members in mechanotransduction.

In addition to simply amplifying auditory inputs, the vertebrate cochlea has the ability to regulate amplification such that it can respond nonlinearly. This permits the detection of faint sounds (reviewed in Walker, 2003). The mechanisms of amplification vary between mammals and lower tetrapods and can involve either the contraction of entire hair cell bodies or the movement of apical hair cell bundles (reviewed in Dallos, 1992; Ruggero, 1992; Hudspeth, 1997; Nobili et al., 1998; Robles and Ruggero, 2001). The existence of the amplifying feedback mechanisms results in spontaneous oscillations within the vertebrate cochlea (reviewed in Kemp, 2002). Because of both the morphological differences between vertebrate and Drosophila auditory organs and the prevailing notion that these auditory organs evolved independently, amplification and nonlinear responses were not predicted to exist in Drosophila. Nevertheless, the Drosophila auditory system has now been found to exhibit nonlinear tuning and to show spontaneous oscillation in the absence of sound (Göpfert and Robert, 2003). Several Drosophila genes already implicated in hearing, including nompA, nompC, touch insensitive larvae B and beethoven (Eberl et al., 1997, 2000; Kernan et al., 1994), are required for this response (Göpfert and Robert, 2003). Because the genes underlying vertebrate nonlinear responses remain unknown and those underlying Drosophila nonlinear responses are only partially elucidated, it is unclear whether these responses share an evolutionary origin or whether they evolved multiple times independently.


The distribution of auditory chordotonal organs is not restricted to the antenna, or even to the head, in insects. For instance, auditory organs are found on the legs or first abdominal segment in Orthoptera (e.g., bush crickets, locusts, and grasshoppers), on the third thoracic or first abdominal segments or on the wing in Lepidoptera (e.g., moths and butterflies), and on the abdomen in Hemiptera (e.g., cicadas; Gillot, 1995; Hoy, 1998). The locations of these thoracic and abdominal auditory structures are places where nonauditory chordotonal organs are found in other insects and may be serially homologous with nonauditory chordotonal organs in the same animal. For instance, the auditory Vogel's organs on the wings of some butterflies correspond to nonauditory chordotonal organs on Drosophila wings, and nonauditory chordotonal organs are present in posterior abdominal segments in Orthoptera. Thus, it may be possible to coopt a nonauditory chordotonal organ to auditory function (Boyan, 1993; reviewed in Fullard and Yack, 1993; Hoy and Robert, 1996; Yager, 1999; Stumpner and von Helversen, 2001). Whether there is a common developmental genetic mechanism that is used to transform a nonauditory chordotonal to an auditory function remains unknown. With a better molecular genetic understanding of what distinguishes chordotonal organs specialized for audition from those that are not, we will be able to investigate the evolutionary trajectories of these diverse hearing organs.


As outlined above, molecules acting at distinct levels of the developmental genetic hierarchy regulating auditory system development and function are used by both Drosophila and vertebrates. An important question for the future is how these parallels came to be. One possibility is that the parallels arose by means of convergent evolution. A second possibility is that the protostome–deuterostome ancestor had a primitive mechanosensor whose construction and function used the genes that are shared today. A third possibility is that the protostome–deuterostome ancestor already had an auditory system in which the shared genes were used. In the absence of that ancestor or fossils of it, how might we ascertain which, if any, of these scenarios might be correct?

One reasonable approach would be to use a combination of gene expression and regulatory interaction data. Most genes are pleiotropic, and if they are expressed during the development of multiple tissues or organs, one cannot use those genes individually to infer that particular tissues or organs are homologous. However, if a constellation of genes is uniquely coexpressed in a single tissue or organ in one species, and the homologs of those genes are uniquely coexpressed in a single tissue or organ in a second species, the probability increases that those two tissues or organs are homologous (Fig. 3). One then could investigate the regulatory relationships among the genes in that constellation. If common regulatory interactions are found in the putatively homologous tissues or organs in both species, then the likelihood that those genes interacted in the same way in an ancestor is high, as is the likelihood that the modern tissues or organs are homologous (Fig. 4).

Figure 3.

Identifying constellations of genes expressed in the auditory organs of Drosophila and vertebrates. Most genes are pleiotropic, functioning in multiple tissues. However, if genes can be identified that are uniquely coexpressed in a single tissue, such as the Johnston's organ (JO) of the fly and the vertebrate ear, this finding may indicate that these genes functioned together in the last common ancestor of protostomes and deuterostomes in a mechanosensory cell type or organ that gave rise to both JO and the ear. See text for details.

Figure 4.

A diagram using developmental genetic hierarchies to identify potential evolutionary homology. If, in addition to being uniquely coexpressed in a particular tissue or organ, a constellation of genes exhibit similar regulatory relationships between species, this finding may indicate that both the regulatory relationship and the tissue or organ to which it is specific already existed in common ancestor. Diagrammed here is a hierarchy known in Johnston's organ whose interactions could be tested in the developing auditory system of a vertebrate model organism. See text for details.

How could this methodology be applied to auditory systems? And where are the potential pitfalls? As described in this review, there are multiple homologous genes expressed in both vertebrate and invertebrate auditory systems. However, in many cases their expression patterns are incompletely described, and in most cases, their regulatory interactions are unexplored. In addition, the developmental genetics of vertebrate nonauditory mechanosensors remain an open frontier. The tasks now are to identify groups of genes uniquely coexpressed in both developing JOs and vertebrate ears (Fig. 3) and to ascertain whether they exhibit shared regulatory relationships (Fig. 4). If particular gene networks are shared among multiple classes of mechanosensors, then it is likely that the protostome–deuterostome ancestor possessed a primitive mechanosensory structure that evolved multiple times independently into both auditory systems and other types of mechanosensory organs. On the other hand, if genes can be identified whose coexpression is restricted to developing auditory systems and which interact similarly in those systems across phyla, a strong argument could be made that the protostome–deuterostome ancestor had an auditory system.

With what types of genes might one initiate this analysis? Arguments could be made for genes at various levels of the genetic hierarchy governing auditory system development. Based on what is known about genetic conservation during the evolution of other tissue and organ systems, patterning and selector genes are likely to exhibit conservation across large evolutionary distances, as are tissue-specific transducers. In the case of eye development, selectors such as eyeless/Pax6, sine oculis, eyes absent, and dachshund are conserved from Drosophila to vertebrates as are the phototransducing rhodopsins (reviewed in Arendt and Wittbrodt, 2001; Kumar and Moses, 2001; Gehring, 2002; Pichaud and Desplan, 2002). In the case of developing auditory systems, several essential regulatory genes, including dll/Dlx, ato/ath, and sal/sall are uniquely coexpressed in both invertebrates and vertebrates (Wagner-Bernholz et al., 1991; Jarman et al., 1993, 1995; Scherer et al., 1994; Crackower et al., 1996; Ignatius et al., 1996; Kohlhase et al., 1998, 2002; Acampora et al., 1999; Bermingham et al., 1999; Depew et al., 1999; Dong et al., 2000, 2002, 2003; Mishra et al., 2000; Tackels-Horne et al., 2001; Merlo et al., 2002; Robledo et al., 2002; Solomon and Fritz, 2002). We also know that dll regulates ato during JO development (Dong et al., 2002). Because dll and ato are not coexpressed in other mechanosensory structures, if Dlx regulates ath in the developing vertebrate ear, it would support the hypothesis that the protostome–deuterostome ancestor had an auditory system, as vs. merely a primitive mechanosensor. Similar analyses could be undertaken for other patterning and selector genes as well as for components of the auditory transduction apparatus.

A question related to how the parallels between extant auditory systems came to be is how many distinct types of sensory neurons there were in the protostome–deuterostome ancestor. At this point, based on gene expression, cell morphology, and function, an argument can be made that there were at least two. The first type probably had ciliated neurons and used TRP channels. This type is the class that would have given rise to Drosophila chordotonal organs, bristle organs, and campaniform sensilla, as well as to Drosophila olfactory sensilla and photoreceptors. Derivatives of these ancestral ciliated neurons in nematodes would include the ciliated chemosensory, osmosensory, and mechanosensory neurons, and in vertebrates would include hair cells and Merkel cells. A second type of sensory neuron, which was not ciliated and used degenerins/epithelial sodium channels could have given rise to the nonciliated multidendritic sensory neurons of Drosophila, the touch-sensitive body wall cells of nematodes, and baroreceptors and some cutaneous mechanoreceptors in vertebrates (reviewed in Corey and Garcia-Anoveros, 1996; Mano and Driscoll, 1999; Ernstrom and Chalfie, 2002). Testing this scenario will require more developmental genetic information about each of the modern day sensory structures. Logical starting places include thorough investigations of the relevant proneural genes for each receptor type and continuing analyses of their transduction apparati. From this perspective, it is intriguing that ato has been proposed to regulate the formation of a protosensory structure that gave rise during evolution to multiple types of sense organs of Drosophila (Niwa et al., 2004).

The chemosensory nematocytes of cnidarians arise from cells expressing a proneural achaete-scute homolog (Cnash; Grens et al., 1995) and exhibit structural and physiological properties reminiscent of hair cells of the vertebrate ear (Hausmann and Holstein, 1985; Brinkmann et al., 1996). Cnidarians diverged from other metazoan animals before the protostome-deuterostome split. It, therefore, is possible that nematocytes represent an ancient type of sensory structure of the sort that may have given rise to many or most others. Evaluation of nematocyte transduction apparati, in particular testing whether they use TRP channels, could provide substantial insight into sense organ evolution.

There are limitations to developmental genetic analyses of the sort proposed here. For instance, over large evolutionary distances, genetic programs that were once the same are likely to have diverged. Drosophila and vertebrates diverged at least 600 million years ago. We, therefore, should not expect every gene in a given developmental program to be conserved—even when the structures they regulate or in which they function are homologous. Along the same lines, many genes have been duplicated during the course of metazoan evolution, and although a particular class of molecule may be required for a given developmental process, which family member is used may vary between species (e.g., different Wnt family members are used for dorsoventral patterning of mouse and chick limbs; discussed in Tabin et al., 1999). This “paralog swapping” complicates the evaluation of expression data. Nonetheless, if one looks at a sufficient number of genes and interactions, and divergence time is not too great, underlying homologies between tissues or organs should emerge.

Drosophila clearly is a powerful model system in which to study mechanosensation. It now is emerging as a useful model in which to gain insights into audition and human deafness syndromes.


The author thanks Dan Eberl for the schematic in Figure 1, Duc Dong for the photomicrographs in Figure 2, and Dominic Ebacher, Eric Pueschel, and three anonymous reviewers for helpful comments on the manuscript.