Many primary vestibular afferents form large cup-shaped postsynaptic terminals (calyces) that envelope the basolateral surfaces of type I hair cells. The calyceal terminals both respond to glutamate released from ribbon synapses in the type I cells and initiate spikes that propagate to the afferent's central terminals in the brainstem. The combination of synaptic and spike initiation functions in these unique sensory endings distinguishes them from the axonal nodes of central neurons and peripheral nerves, such as the sciatic nerve, which have provided most of our information about nodal specializations. We show that rat vestibular calyces express an unusual mix of voltage-gated Na and K channels and scaffolding, cell adhesion, and extracellular matrix proteins, which may hold the ion channels in place. Protein expression patterns form several microdomains within the calyx membrane: a synaptic domain facing the hair cell, the heminode abutting the first myelinated internode, and one or two intermediate domains. Differences in the expression and localization of proteins between afferent types and zones may contribute to known variations in afferent physiology.
In vestibular sensory epithelia of mammals and other amniotes, primary afferents form expanded calyx terminals around type I hair cells in addition to conventional bouton terminals on type II hair cells (Wersäll, 1956). Unlike well known calyces of the auditory brainstem and ciliary ganglion, the vestibular calyx is postsynaptic: its inner face is postsynaptic to the type I hair cell. Its outer face is continuous with the initial segment leading to a heminode where spikes are initiated and is postsynaptic to cholinergic terminals of efferent neurons in the brainstem. This combination of functions provides an opportunity to investigate how nodal proteins interact with synaptic proteins in a compact ending, in contrast to such nodal models as motor neurons and sciatic nerve (Schafer and Rasband, 2006; Lorincz and Nusser, 2008). We find that the calyx expresses novel combinations of ion-channel and tethering proteins, including some that are characteristic of axonal nodes or initial segments (Salzer, 2003; Lai and Jan, 2006).
Type I hair cells release glutamate onto calyces from vesicles arrayed around synaptic ribbons (Bonsacquet et al., 2006; Holt et al., 2007; Dulon et al., 2009). A single calyx ending may receive input from tens of ribbons; in contrast, each bouton ending on a type II cell is generally driven by vesicular release from a single ribbon (Lysakowski and Goldberg, 1997, 2008; Holt et al., 2007). Type I calyx transmission also includes a nonquantal component of unknown mechanism (Yamashita and Ohmori, 1990; Holt et al., 2007), and retrograde vesicular transmission has been suggested (Scarfone et al., 1988; Devau et al., 1993; Scarfone et al., 1996; Chen and Eatock, 2000; Lysakowski and Singer, 2000). Cholinergic efferent terminals on calyx outer faces may provide positive feedback on afferent activity (Holt et al., 2011).
In mammals, afferents from central and peripheral zones of vestibular epithelia form two physiologically distinct populations. Central-zone, “irregular” afferents have variable spike timing and comparatively phasic response dynamics; peripheral-zone “regular” afferents have highly stereotyped spike timing and more tonic dynamics (reviewed in Goldberg, 2000; Eatock and Songer, 2011). The two populations do not relate in a simple way to the two types of hair cells and afferent terminals: type I and II hair cells are distributed throughout both zones and supply calyces and boutons of both populations. More subtle zonal differences must be at work. Here we describe zonal variations in calyceal ion channels that may contribute.
Ion channels of interest include three potassium channel families [KCNQ (KV7), erg (KV11), and KV1] that include low-voltage-activated channels, which may help set spike timing (Iwasaki et al., 2008; Kalluri et al., 2010). KCNQ and erg channels are present in calyces (Kharkovets et al., 2000; Hurley et al., 2006; Rennie and Streeter, 2006), together with an unusual Na channel, NaV1.5 (Wooltorton et al., 2007). We report localization of these channels to specific calyceal subdomains. To investigate how subdomains are established, we also localized contactin-related cell adhesion molecules, scaffolding proteins, and an extracellular matrix protein.
Materials and Methods
Procedures involving animals were approved by the Institutional Animal Care and Use Committees at the University of Illinois at Chicago.
Choice of antigens.
We chose ion channels from those known to be expressed at nodes and initial segments (KV1, KCNQ, NaV1) and/or implicated by previous studies on vestibular calyces or eighth-nerve neurons (KV1, KCNQ, erg, NaV1.5, and NaV1.6). KV1.1 and 1.2 subunits form channels that carry subthreshold currents near nodes (Arroyo and Scherer, 2000), and KV1 channels are expressed by vestibular ganglion neurons (Iwasaki et al., 2008; Kalluri et al., 2010). KCNQ (KV7) channel subunits contribute to M currents, are expressed at nodes of Ranvier and initial segments of central neurons (Arroyo and Scherer, 2000), and are present in vestibular calyces and neuronal cell bodies (Kharkovets et al., 2000; Hurley et al., 2006; Rennie and Streeter, 2006). Evidence for ether-a-go-go-related (erg, KCNH, KV11) channel subunits has been obtained from immature calyceal terminals (Hurley et al., 2006). Because nodes express NaV channels at high density, we used a pan-NaV antibody. We also chose antibodies to two specific NaV channels: NaV1.5, known to be present in immature vestibular calyces (Wooltorton et al., 2007), and NaV1.6, a nodal ion channel that Hossain et al. (2005) localized to the initial segments of auditory afferents.
For scaffolding proteins, we chose those known to tether NaV and KV channels at or near nodes of Ranvier (Hedstrom et al., 2007; Ogawa and Rasband, 2008): ankyrins B and G, βIV spectrin, dystrophin, ezrin, and neurofascin 186. We investigated the cell adhesion molecules (CAMs), contactin and Caspr1 and 2, because they are paranodal or juxtaparanodal in other neurons (Lai and Jan, 2006). We also used an antibody to tenascin-C, an extracellular matrix molecule (ECM) associated with synaptic clefts (Dityatev and Schachner, 2003, 2006) and type I hair cells (Swartz and Santi, 1999; Warchol and Speck, 2007).
For details on primary antibody provenances, see Table 1. Antibodies were obtained from several vendors, including Millipore Bioscience Research Reagents, Alomone, Sigma, United States Biological, Covance, and University of California, Davis/NIH NeuroMab facility. We also obtained antibodies as gifts from Matthew Rasband and Edward Cooper of Baylor College of Medicine, Elior Peles of the Weizmann Institute of Science, Thomas Jentsch of the Max-Delbrück-Centrum for Molecular Medicine, S. Rock Levinson of University of Colorado at Denver, Vann Bennett of Duke University, Álvaro Villarroel of Instituto Cajal and Universidad del País Vasco, and Michel Roux of Université de Strasbourg. The rabbit polyclonal antiserum against pan-NaV was made to an epitope common to all subunit subtypes of the voltage-gated Na+ channel. Alexa dye-labeled secondary antibodies, made in donkey against rabbit (catalog #A21206 and A21207), goat (A11055 and A21447) and mouse (A21202 and A21203), were obtained from Invitrogen. Goat anti-mouse IgG isotype-specific Alexa dye-labeled secondary antibodies were also obtained from Invitrogen (IgG2a, A21131; IgG1, A21125; IgG2b, A21242).
Immunohistochemistry and microscopy.
Adult Long–Evans rats of both sexes were deeply anesthetized with Nembutal (80 mg/kg), then perfused transcardially with 100 ml of physiological saline containing heparin (1000 IU), followed by 2 ml/g body weight of fixative [4% paraformaldehyde, 1% acrolein, 1% picric acid, and 5% sucrose in 0.1 m phosphate buffer (PB) at pH 7.4]. Vestibular epithelia were dissected in PB. Some end organs were fixed by immersion and dissection of the temporal bones was done in 100% EtOH. In some experiments with NaV antibodies, rats were perfused with only 4% paraformaldehyde and 5% sucrose in 0.1 m PB. Otoconia were dissolved with Cal-Ex (Fisher Scientific) for 10 min. Background fluorescence was reduced by incubating organs and ganglia in a 1% aqueous solution of sodium borohydride for 10 min, and tissue was cryoprotected in 30% sucrose-PB overnight at 4°C. Frozen sections (35 μm) were cut with a sliding microtome. Immunocytochemistry was performed on free-floating sections. Tissues were first permeabilized with 4% Triton X-100 in PBS for 1 h at room temperature (RT), then incubated with 0.5% Triton X-100 in a blocking solution of 0.5% fish gelatin and 1% BSA in PBS for 1 h at RT. Ethanol-fixed tissues had no permeabilization pretreatment, although 0.5% Triton X-100 was sometimes included in the antibody diluent. Sections were incubated with a mixture of two, or in some cases three, primary antibodies for 72 h at 4°C. Most primary antibodies were diluted to 1:200 in the blocking solution, except for Caspr1 (1:250), βIV spectrin (1:400), ankyrinB (1:2000), and the Sigma pan-NaV (1:100). We used calretinin antibody as a marker of type II hair cells and calyx afferents. Specific labeling was revealed by incubating sections in a mixture of two secondary antibodies (fluorescein-conjugated donkey anti-goat IgG and rhodamine-conjugated donkey anti-rabbit IgG, diluted 1:200 in blocking solution) for 24 h at 4°C. If a third primary antibody was used, then typically Cy5-conjugated donkey anti-mouse was used with fluorescein- or Alexa 488-conjugated donkey anti-goat and rhodamine- or Alexa 594-conjugated donkey anti-rabbit. To be able to combine two different mouse monoclonal antibodies in the same experiment, we used IgG subtype-specific secondary antibodies (Invitrogen). Sections were rinsed with PBS between and after incubations and mounted on slides in Mowiol (Calbiochem).
In addition to the various controls listed in Table 1, we performed “no primary antibody” controls, which revealed a lack of nonspecific staining. We also confirmed expression of each of the NaV subunits and erg subunits by performing RT-PCR or by quantitative RT-PCR on whole vestibular ganglia and end organs.
Slides were examined on a laser scanning confocal microscope (LSM 510 META, Carl Zeiss). Final image processing and labeling were done with Adobe Photoshop. Some whole macular organs were processed for various combinations of antibodies; Z stacks of optical sections were then obtained on the confocal microscope, with the thickness of the optical sections varying from 0.4 to 1.0 μm. The Z-series were deconvolved and reconstructed in 3D with VOLOCITY software (v. 3.7, Improvision).
Immunogold electron microscopy.
We investigated ultrastructural localization of some markers with immunogold EM, and we present results here for NaV1.5, tenascin-C, and βIV spectrin. Vestibular epithelia were sectioned at 40 μm with a Vibratome 2000 (Technical Products International). Free-floating sections were permeabilized with 0.5% Triton X-100 for 1 h, then blocked in a solution consisting of 0.5% fish gelatin and 1% BSA for 1 h. Sections were incubated in primary antibody (1:50 dilution) for 72 h, rinsed, then incubated for 24 h in secondary antibody (1:40 dilution), tagged with ultra-small (0.8 nm) colloidal gold-labeled F (ab) goat anti-mouse IgG and goat anti-rabbit IgG (Aurion Cat. No. 25413, distributed in the United States by Electron Microscopy Sciences). Colloidal gold staining was silver-enhanced (4–8 min; IntenSE M kit; GE Healthcare). Sections were dehydrated in a graded series of alcohols and propylene oxide, embedded in Araldite (Fluka Durcupan) on glass slides with plastic coverslips, and polymerized at 55°C for 48 h. The section of interest was cut free from the slide, glued on top of a blank Araldite block, sectioned with a diamond knife (DiATOME), and stained with uranyl acetate and lead citrate. Sections were examined and photographed with a JEOL 1220× transmission electron microscope.
Colocalization of various ion channels and candidate tethering proteins revealed three to four contiguous domains of the calyx. We have labeled these as domains 1–4, progressing smoothly from the inner face opposite the presynaptic ribbons (domain 1) up and around the apex of the calyx (domain 2, not always present) to the outer face (domain 3) and ultimately the heminode adjacent to the first myelin wrapping (domain 4). We present results from each domain in sequence. Throughout, we also note zonal variations in immunolocalization because afferent physiology and synaptic and afferent morphology vary with zone. Note that we often used calretinin to mark afferent type and epithelial zone. Among afferents, calretinin selectively labels calyx-only afferents, which have distinctive physiology and whose calyceal terminals delineate the central zones of cristae and the striolar zones of the utricular and saccular maculae (Desai et al., 2005a,b). In rats and mice, calretinin also labels type II hair cells (Desai et al., 2005a,b). Because striolar and extrastriolar zones of maculae resemble the central and peripheral zones, respectively, of cristae, we have simplified the text by referring to zones in both cristae and maculae as “central” and “peripheral.” Bouton-only afferents, which innervate only type II cells, are restricted to peripheral zones, and dimorphic afferents, which make both calyceal endings on type I cells and bouton endings on type II cells, are found throughout the epithelia.
KCNQs, KV1.1, Caspr1, tenascin-C, and NaV1.5 delineate domain 1 on the calyx inner face
In type I hair cells, synaptic ribbons and their associated vesicles are arrayed next to the basolateral cell membrane in the basal part of the cell and extending part way up the sides. The number and extent of the ribbons are larger in central type I cells than in peripheral type I cells (Lysakowski and Goldberg, 1997). Previous work on immature rat epithelia (Hurley et al., 2006) showed KCNQ4 (KV7.4) immunoreactivity on the calyx inner face opposite the synaptic ribbons (RS) of the hair cell (domain 1), illustrated in the schematic of Figure 1. Here we show that KCNQ4 labeling of domain 1 is more intense in adult calyces (P60–P180) and that it overlaps with strong immunoreactivity for other voltage-gated K and Na channel subunits.
Figure 1 (A–C) shows labeling of the calyx inner face with antibodies to KCNQ2 (Fig. 1A) and KCNQ5 (Fig. 1B,C) in approximately the same location as previously reported for KCNQ4 (Kharkovets et al., 2000; Hurley et al., 2006). KCNQ4 immunoreactivity shows zonal variation that mirrors the extent of presynaptic ribbons: it is both more intense and extends farther apically in central calyces than peripheral calyces (Lysakowski and Price, 2003). In contrast, KCNQ2 appeared not to vary with zone or afferent type, whereas KCNQ5 immunoreactivity did not vary with zone but did vary with afferent type. As shown in Figure 1, B and C, KCNQ5 staining was seen in the calyx endings of dimorphic afferents, which are found in both zones and are calretinin-negative. KCNQ5 was not localized to calretinin-positive calyx-only afferents, which are found only in central zones and have distinctive physiology (for review, see Lysakowski and Goldberg, 2004).
To provide a more complete picture of KCNQ4 immunolabeling, we generated three-dimensional reconstructions of confocal stacks of optical sections (Fig. 1D). These pictures reveal “holes” in the intense KCNQ4 label: 1–3 large holes (1–3 μm diameter) at the bottom of each calyceal cup and smaller holes (200–400 nm) at calyceal invaginations, where the calyx membrane pushes into the hair cell and the synaptic cleft narrows (Spoendlin, 1966). The invaginations are especially numerous in the large synaptic domains of central-zone calyces (Lysakowski and Goldberg, 1997) and may serve to fasten the hair cell and calyx together, like punctae adherens in the calyx of Held. The small holes in KCNQ4 immunolabel are reminiscent of freeze-fracture material showing that invaginations are devoid of membrane particles (Gulley and Bagger-Sjöbäck, 1979).
The low-voltage-activated channel, KV1.1, is, like KCNQ2, expressed in domain 1 in all calyces, both central and peripheral (Fig. 1E). KV1.1 is thus a candidate for the dendrotoxin-sensitive conductance of some vestibular ganglion somata, which has a strong impact on firing pattern (Iwasaki et al., 2008; Kalluri et al., 2010).
NaV1.5, a tetrodotoxin-insensitive subunit best known as the cardiac Na channel, is also concentrated in domain 1 (Fig. 1F). The localization strongly overlaps that of KCNQ4 and, like KCNQ4 immunoreactivity, is seen in all calyces but is more intense in central zones.
What molecules bind these voltage-gated ion channels to domain 1? Sousa et al. (2009) showed that localization of KCNQ4 subunits in the calyx inner face requires the cell adhesion molecule Caspr1. Caspr1 is strongly expressed in domain 1 (Fig. 1A2), as is a related CAM, contactin (see domain 4, below). Although both KCNQ2 and NaV subunits have a binding motif for the scaffolding protein ankyrinG (Cooper, 2011), we found no ankyrinG in domain 1. This negative finding was strengthened by staining for ankyrinG in the same tissue sections at heminodes (see domain 4, below), where it did colocalize with KCNQ2 and NaV subunits. The ECM protein, tenascin-C, has been described as a marker of type I hair cells (Swartz and Santi, 1999; Warchol and Speck, 2007). In confocal material (Fig. 1G,H), tenascin-C is closely associated with domain 1. In ultrastructural material, however, we show that tenascin-C is in the synaptic cleft, congruent with a dense cleft substance (Fig. 1I) that may contain other proteins. Like the presynaptic ribbons and the calyceal immunoreactivity for NaV1.5 and KCNQ4, tenascin-C label showed regional variation in extent, reaching farther up from the base of the calyceal cup in central/striolar than peripheral/extrastriolar zones (Fig. 1, compare panels G and H). Tenascin-C has been linked in neurons to synaptic plasticity involving L-type calcium channels (Dityatev et al., 2010). L-type channels are expressed by vestibular hair cells (Bao et al., 2003; Dou et al., 2004), where they mediate transmitter release, and by primary vestibular neurons (Chambard et al., 1999). Alternatively or additionally, tenascin-C might help stabilize the NaV channels (Srinivasan et al., 1998).
In summary, domain 1, which is coextensive with synaptic ribbons in the type I hair cell, includes multiple KCNQ and KV1 subunits in addition to NaV1.5. KCNQ2, KV1.1, and KV1.2 are expressed in calyces regardless of type (calyx-only and dimorphic) or zone (central or peripheral). KCNQ4 and NaV1.5 are also expressed in all calyces, but staining is most intense and extensive in central calyces. KCNQ5 is restricted to dimorphic afferents of both zones. Caspr1, contactin, and tenascin-C are all candidates for localizing specific proteins and functions to this domain. Differential expression of these proteins with zone (KCNQ4, NaV1.5, and tenascin-C) or afferent type (KCNQ5) may contribute to known physiological differences between afferent fiber populations.
Caspr2 demarcates domain 2, an apical domain of the calyx membrane
The high density of KV channel subunits and Caspr1 prompted us to probe the calyx inner face for additional membrane proteins found near nodes of Ranvier. On the calyces of dimorphic afferents, Caspr2, a CAM typically expressed in axonal juxtaparanodes (Peles and Salzer, 2000; Salzer, 2003), labeled the apical part of the inner face, extended around the apex and approximately one-third to half way down the outer face (Fig. 2A, right hair cell). We refer to this domain of intense Caspr2 labeling as domain 2. Less intense label was seen on the outer face at the base. High-magnification views of cross-sections through the apical half of the calyx revealed two concentric circles, indicating surface labeling on both the inner and outer membranes (Fig. 2B, inset). The labeling is much stronger in the peripheral (extrastriolar) zone (Fig. 2B), partly because many of the calyces of central, calyx-only afferents lack the apical part, or neck, of domain 2. Figure 2A illustrates this difference: domain 2 is labeled by Caspr2 in the right calyx but is truncated in the left calyx. Calretinin labeling shows that the left calyx belongs to a calyx-only (calretinin-positive) afferent, and the right calyx belongs to a dimorphic (calretinin-negative) afferent. In a count of calretinin-positive calyces, approximately two-thirds (57/86) were truncated in this way, and most of the remainder had small neck regions. This striking anatomical difference may also play a role in the physiological differences between calyx-only and dimorphic afferents.
Caspr2 staining overlaps with staining for erg1 (KV11.1) subunits (Fig. 2C) and erg2 (KV11.2; data not shown). The same concentric pattern of circles shown for Caspr2 was seen with erg staining in peripheral-zone calyces (data not shown), indicating that erg channels are present on both faces. Thus, Caspr2 is a candidate for tethering erg subunits in place, just as Caspr1 does for KCNQ4 subunits in domain 1 (Sousa et al., 2009). Domain 2 in dimorphic afferents also has an apical band of immunoreactivity for βIV spectrin (Fig. 2D,E), a nodal scaffolding protein (Schafer and Rasband, 2006).
Domain 2 is defined not just by the presence of Caspr2, but also by the absence of other proteins. For example, dystrophin expression in domains 1 and 3 stops abruptly at the border with domain 2 (Fig. 2F). Dystrophins are cytoskeletal glycoproteins frequently associated with NaV channels and certain KV subunits, for example, KCNQ3 (Byers et al., 1991, 1993; Saito et al., 2003). The absence of dystrophin from domain 2 may account for the lack of pan-NaV labeling in this domain (Fig. 3A; see below). KCNQ3 immunoreactivity was strongly present in domains 3 and 4, as described below, but was absent from domain 2.
NaV channels, dystrophin, KCNQ3, and ankyrinB define domain 3 on the calyx outer face
Domain 3 comprises the basal outer face of the calyx and most of the unmyelinated part of the fiber immediately below the calyx (Fig. 3, inset). This domain was labeled by pan-NaV antibody made against an epitope common to all NaV α subunits (Fig. 3A). The complex pattern of labeling, which also included domains 1 and 4 and hair cells, is likely to reflect expression of multiple NaV isoforms. For example, pan-NaV labeling of domain 1 may correspond to NaV1.5 given its similarity to the NaV1.5-like immunoreactivity in Figure 1F. The pan-NaV staining in hair cells may correspond to NaV1.5 and/or NaV1.2, based on reports from younger rats (Chabbert et al., 2003; Wooltorton et al., 2007) and our unpublished observations with subunit-specific antibodies in adult rats (S. Gaboyard and A. Lysakowski). We saw significant differences between afferent types (dimorphic vs calyx-only) in labeling of domain 3 for certain ion-channel-associated proteins, which again suggests differences in their ion channel expression. Pan-dystrophin antibody labeled both faces (domains 1 and 3) of the calyces of dimorphic afferents (Fig. 3B) but not the calyces of calyx-only afferents (Fig. 3C). Thus, NaV subunit expression may differ between calyx-only and dimorphic calyces. Furthermore, given the strong staining for NaV1.5 in domain 1 of calyx-only afferents (Fig. 1F and Wooltorton et al., 2007), it appears that NaV1.5 is not coupled to dystrophin.
Expression in domain 3 also differed between zones. Staining with an antibody to KCNQ3, a marker of nodes of Ranvier and axonal initial segments (Chung et al., 2006; Schwarz et al., 2006), was much more intense in domain 3 of peripheral calyces than in central calyces (Fig. 3D). Domain 2 was not labeled (see dashed outlines of domain 2 in Fig. 3E). Weak KCNQ4 immunoreactivity was also present in domain 3, but was much less intense than in domain 1 (Fig. 1D).
An antibody to ankyrinB, a paranodal scaffolding protein associated with KV channels (Scotland et al., 1998; Ogawa et al., 2006), labeled the unmyelinated terminal segments of dimorphic afferents but not calyx-only afferents (Fig. 3D). In dimorphic calyces, ankyrinB immunofluorescence formed a thin shell along the outer face, continuing below the calyx (Fig. 3E, long arrows) and becoming more intense near the heminode. For calyx-only afferents, ankyrinB labeling was only seen at the heminode (Fig. 3E, short arrows).
In summary, domain 3 in dimorphic afferents is defined by expression of NaV and specific KV channels and associated proteins. The staining for pan-dystrophin, ankyrinB, and KCNQ3 differs between zones, in part because domain 3 of the calyx-only afferents is not stained and these constitute a significant fraction of central afferents (≈50% in crista, 25% in macula of chinchilla; Fernández et al., 1988, 1990). These zonal variations in domain 3 ion channel expression may influence firing pattern differences between afferent populations.
Nodal and ion channel proteins label domain 4, the heminodal membrane next to the myelin
Domain 4 is the heminode, a stretch of approximately 1 μm on the afferent fiber adjacent to the first internode. The first internode can be visualized by staining its myelin sheath with antibody against myelin basic protein (Fig. 4A, blue label). The heminode and nodes farther along the afferent were selectively labeled by antibodies against the nodal proteins NaV1.6 (Fig. 4A), βIV spectrin (Fig. 4B), neurofascin 186 (Fig. 4C), pan-dystrophin (Fig. 4D), ezrin (Fig. 3D, inset), and ankyrinG (data not shown). Labeling for these nodal markers generally did not extend distally along the unmyelinated fiber toward the calyx, justifying the delineation of this portion of the axon as a separate domain. Restriction of NaV1.6 staining to the heminode and nodes (Fig. 4A, inset) in dimorphic, but not calyx, afferents indicates that other NaV subunits are responsible for the pan-NaV staining of domains 3 and 4 (Fig. 3A).
KCNQ staining of domain 4 differed strongly between calyx-only and dimorphic afferents. Heminodes of calyx-only afferents stained intensely for KCNQ4 (Fig. 4E,F). Three-dimensional reconstructions of nerve fibers showed that KCNQ4 immunoreactivity formed a ring around the cytosolic calretinin label of calyx-only fibers (Fig. 4E,F, long arrows), consistent with membrane localization of the channels. In contrast, heminodes below the calyces of dimorphic afferents expressed KCNQ3, as shown by colocalization of KCNQ3 and ezrin staining (Fig. 3D, yellow in inset).
Caspr1 (Fig. 4A), contactin (Fig. 4G), and ankyrinB (Fig. 3D) antibodies labeled the paraheminodal membrane distal to the heminode. The paraheminodal contactin staining (Fig. 4G, arrowheads) shows that the peripheral-zone heminodes are well below the basement membrane of the epithelium, much farther from the hair cells than are central-zone heminodes (Fig. 4F, long arrows).
Antibodies to KV1.1 (Fig. 1E), KV1.2 (data not shown), and Caspr2 (data not shown) labeled the juxtaparaheminodal membrane beyond domain 4 (beneath the myelin), as seen at juxtaparanodes in other neurons.
Functional significance of different microdomains
The vestibular calyx integrates afferent signals from hair cells and efferent signals from the brain into spike activity representing head motions. Calyces in different zones contribute to afferents with distinct discharge patterns and response properties. The localization of ion channels and various associated proteins suggests that these large synaptic cups, previously thought of as uniform, are divided into several discrete domains (Fig. 5), from a postsynaptic domain 1 deep in the calyceal cup to the heminodal domain 4 adjacent to the first myelin wrapping (Table 2, Fig. 5). In addition to ion channels, the domains express a number of scaffolding, CAM, and ECM proteins characteristic of axonal nodes and initial segments. Our results support the hypothesis that differences in molecular organization of these domains, between zones and between calyx-only and dimorphic afferents, may help differentiate the discharge patterns of irregular afferents and regular afferents (Smith and Goldberg, 1986; Baird et al., 1988; Goldberg et al., 1990b; Kalluri et al., 2010).
Domain 1 comprises much of the calyceal inner face and is contiguous with a dense substance in the intervening synaptic cleft, apposing the synaptic ribbons. Indeed, the extent of the immunolabeled cups (Fig. 5A–F) covaries with hair cell membrane devoted to transmitter release, as reflected by adjacent synaptic ribbons. Ribbons cluster in the bottom third in peripheral-zone type I hair cells but in the bottom two-thirds of central-zone type I hair cells (Lysakowski and Goldberg, 1997). Therefore, in addition to the proteins shown here, domain 1 also expresses glutamate receptors—principally AMPA-type, but also NMDA receptors (Matsubara et al., 1999; Bonsacquet et al., 2006).
Domain 1 has an impressively dense expression of KV and NaV channels. KV channels include multiple KCNQ (KV7) and erg (KV11) channel isoforms as well as KV1.1 and KV1.2. Their expression closely matches that of the CAM Caspr1, known to occur at paranodes (Salzer, 2003). Sousa et al. (2009) showed that Caspr1 is necessary for retention of KCNQ4 in the calyx membrane and for maintaining appropriate synaptic cleft width. We report that contactin labeling is comparable to the Caspr1 pattern. Both CAMs may together stabilize other KV channels in this domain, such as KCNQ2, KCNQ5, KV1.1, and KV1.2.
Several proteins localized to domain 1 are reported to associate with NaV subunits in other tissues: contactin (Kazarinova-Noyes et al., 2001; Rush et al., 2005), tenascin-C (Srinivasan et al., 1998; Evers et al., 2002; Dityatev and Schachner, 2006; Ullian and Dityatev, 2006), and dystrophin (Byers et al., 1991, 1993; Kim et al., 1992; Saito et al., 2003; Occhi et al., 2005). Thus, NaV1.5 subunits in domain 1 may associate with any or all of these proteins. Contactin may regulate current density and expression of NaV1.5 channels, as it is thought to do for NaV1.8 and 1.9 subunits (Kazarinova-Noyes et al., 2001; Isom, 2002; Rush et al., 2005). Tenascin-C contributes to the electron-dense cleft material in the synaptic cleft of domain 1 and could interact with NaV β subunits and/or contactin in domain 1 to stabilize the Na channels (Zisch et al., 1992; Srinivasan et al., 1998). Dystrophin forms a complex with dystroglycan that connects the actin cytoskeleton to the extracellular matrix, and is found where NaV channels are present in high density, for example, the deep folds of the neuromuscular junction (Byers et al., 1991), nodes (Byers et al., 1993; Occhi et al., 2005), and postsynaptic densities (Kim et al., 1992).
Thus, a complex involving contactin, tenascin-C, dystrophin, and Caspr1 may hold both KV and NaV channels at high density and in specific subdomains within the calyceal synaptic cleft. Analogous roles are proposed for agrin and laminin at the neuromuscular junction and for reelins and integrins at forming synapses (Dityatev and Schachner, 2003; Ullian and Dityatev, 2006).
The between-zones differences we report in ion channel expression in domain 1 may play a role in differentiating afferent firing patterns between epithelial zones (Smith and Goldberg, 1986; Baird et al., 1988; Goldberg et al., 1990a; Kalluri et al., 2010). Recent work has specifically proposed that low-voltage-activated channels (KV1 and KCNQ) help make the firing of central afferents irregular (Iwasaki et al., 2008; Kalluri et al., 2010). KCNQ4 immunolabeling correlates with irregular spike timing: it is more extensive and intense in the central-zone calyces of irregular afferents than in the peripheral-zone calyces of regular afferents. Furthermore, M-current through KCNQ channels in calyces might be modulated by G-protein-coupled receptors activated by acetylcholine, ATP, CGRP, opioid peptides, or GABA released from efferent terminals onto calyx outer faces. In that case, the greater expression of KCNQ4 in central calyces could contribute to the greater effect of efferent activation on irregular afferents (Holt et al., 2011).
KCNQ4 is present at high density on the calyx inner face (domain 1) and much lower density on the outer face (domain 3; see Fig. 1D and Lysakowski and Price, 2003). In heterologous expression systems, KCNQ4 can form homomeric channels (Søgaard et al., 2001; Xu et al., 2007) and can heteromultimerize with KCNQ2, KCNQ3, and KCNQ5 (Howard et al., 2007; Xu et al., 2007; Bal et al., 2008). Our results suggest that KCNQ4 could partner with KCNQ2 in domain 1 of any calyx (Fig. 1A). In peripheral dimorph calyces, KCNQ4 could also partner with KCNQ5 on the inner face (domain 1; Fig. 1C) and KCNQ3 on the outer face (domain 3; Fig. 3D).
Immunoreactivity for erg subunits was much stronger in calyces of dimorphic afferents than in calyces of calyx-only afferents. This difference explains the reduced erg staining previously noted in central zones in immature epithelia (Hurley et al., 2006). Again, such zonal differences in M-like channel expression have the potential to shape afferent discharge regularity and/or efferent responses.
NaV1.5, localized to domain 1, includes an amino acid sequence that shifts activation and inactivation voltage ranges negative relative to tetrodotoxin-sensitive NaV isoforms (Camacho et al., 2006). In this way, the voltage range of NaV channel activation in calyces may be matched with that of nearby KCNQ and erg channels (Hurley et al., 2006). Either NaV1.5 or NaV1.6 could contribute to a persistent NaV current (Holt et al., 2007; Wooltorton et al., 2007), which could shape spike timing. In the rat utricular macula, NaV1.5 immunoreactivity is strongest in the striola, at least up to P21 (Wooltorton et al., 2007); thus, NaV1.5 is also a candidate to contribute to zonal differences in discharge regularity.
Domain 2, the apical calyx membrane, is defined both by what is expressed and what is not expressed. Antibodies against βIV spectrin, Caspr2, erg1, and erg2 labeled domain 2 more than other zones. Antibodies that labeled domain 3 and not domain 2 include those against NaV channels, dystrophins, KCNQ3, and ankyrinB. Because βIV spectrin associates with NaV channels at nodes, and Caspr2 associates with KV channels at juxtaparanodes (Arroyo and Scherer, 2000), domain 2 may express yet-to-be-identified voltage-gated channels.
Domain 2 has been proposed as a possible source of retrograde feedback from calyces to hair cells, based on immunolocalization of such presynaptic proteins as rab3A (Dechesne et al., 1997), synaptophysin (Scarfone et al., 1988; Dechesne et al., 1997), synapsin I (Scarfone et al., 1988), syntaxin, SNAP25, and synaptotagmin (Demêmes et al., 2000). Calyces might release neuropeptides (Scarfone et al., 1996) or glutamate (Devau et al., 1993) onto the type I hair cell. Detailed ultrastructural studies in rodents have revealed no postsynaptic densities in hair cell membranes adjacent to domain 2, but clusters of dense-cored vesicles are seen in domain 2 in squirrel monkey calyces (A. Lysakowski, personal observations). Dense-cored vesicles are consistent with the neuropeptide hypothesis, and because they release their contents away from active zones (Lysakowski et al., 1999; Shakiryanova et al., 2005), postsynaptic densities are unnecessary.
Domain 2 appears to be lacking in most calyx-only afferents (Figs. 1H, 2A–C, 5E), another zonal difference that could influence afferent physiology. In particular, retrograde feedback from the calyx to the hair cell might be missing at calyx-only synapses, which are principally found in central and striolar zones.
In domain 3, we identified many proteins characteristic of axonal initial segments, but there are also some differences. To our knowledge, the combination of ion channels does not match any of those described in central neurons, although some diversity of expression has been reported for axonal initial segments (Lorincz and Nusser, 2008). For example, olfactory mitral cells express KV1.2, layer 5 pyramidal cells also express KV1.1, and Purkinje cells express neither (Lorincz and Nusser, 2008). Domain 3 of the vestibular calyx resembles the axonal initial segment in that it expresses a gradient of ion channels and scaffolding proteins. In addition to ion channels, domain 3 contains ankyrinB, which anchors Na+-K+ ATPases and Na+-Ca2+ exchangers in heart (Mohler et al., 2005) and retina (Kizhatil et al., 2009). Na+-Ca2+ exchangers and the α3β1 isoform of Na+-K+ ATPase have been reported within the vestibular sensory epithelium (ten Cate et al., 1994; Boyer et al., 1999), but resolution was not sufficient to localize either to presynaptic or postsynaptic membranes. Unlike some axonal initial segments, domain 3 does not express βIV spectrin and neurofascin 186.
Finally, domain 4 contains many known nodal proteins (Peles and Salzer, 2000; Salzer, 2003; Dityatev and Schachner, 2006; Ullian and Dityatev, 2006) and resembles in its organization one-half of a full node (Fig. 4A, inset). Many vestibular heminodes were recognized by the juxtaposition of NaV1.6, the most common nodal NaV channel, with the paranodal protein Caspr1 and the internodal protein myelin basic protein. Scaffolding proteins at nodes of Ranvier (βIV spectrin and neurofascin 186) also label vestibular afferent heminodes intensely (Fig. 4B,C). The expression of KV7 subunits (KCNQ3, but not KCNQ2, at dimorphic heminodes and KCNQ4 at calyx-only heminodes) differs from the combination of KCNQ2 and KCNQ3 expression reported at other axonal nodes (Pan et al., 2006; Bennett and Healy, 2009).
The spike trigger zone in eighth nerve afferents has been a subject of speculation (Goldberg, 1996; Hossain et al., 2005). The presence of many known nodal proteins in domain 4 suggests that the heminode immediately below calyces may be a spike trigger zone, but does not rule out a more distant locus (Palmer and Stuart, 2006; Bean, 2007). Furthermore, the location of spike initiation may vary with zone, given that peripheral dimorphic afferents have more extensive arbors with longer unmyelinated portions (Fig. 4A,F,G).
In summary, the calyx terminal combines unique features of dendrites and initial segments in a highly ordered, novel arrangement. The large size of the ending has allowed us to define several regions, or microdomains, each characterized by a distinct set of ion channels, neurotransmitter receptors, and accessory proteins. None of the antibodies we used produced detectable labeling of bouton terminals on type II cells. The much smaller surface areas of bouton endings (5–10 μm2) compared with calyx endings (∼1000 μm2) may be less able to accommodate microdomains.
Supplemental Material, including a movie showing a 3D reconstruction of KCNQ4 immunolabeling in the utricular macula, is available at http://www.uic.edu/depts/mcan/Vestibular_Calyx_Model.htm. Three-dimensional reconstruction of a segment of vestibular sensory epithelium from the utricular macula. An antibody to KCNQ4, a low-voltage activated potassium channel subunit, intensely labels Domain 1, the inner surface of the calyx, while the outer surface (Domain 3, seen at 11–14 s) is much less intensely labeled. A small subset of heminodes in central complex calyces, likely calyx-only afferents, is also labeled. Gaps in immunoreactivity are observed in the bottom of each calyx. This material has not been peer reviewed.
This study was supported by NIH Grants R01 DC-02521, R01 DC-02058, and R01 DC-02290. The Electron Microscopic Facility of the Research Resources Center of the University of Illinois at Chicago provided equipment and assistance to conduct this study. We are grateful to Drs. Matt Rasband (NaV1.6, NaV1.2, Caspr1, βIV spectrin, neurofascin 186, and pan-neurofascin), James Trimmer (pan-NaV), Elior Peles (Caspr3 and Caspr4), Edward Cooper (KCNQ2 and KCNQ3), Thomas Jentsch (KCNQ4 and KCNQ5), Rock Levinson (pan-NaV), Vann Bennett (ankyrinG and ankyrinB), and Michel Roux (H4, pan-dystrophin) for generous gifts of antibodies, and to Marcin Klapczynski for technical assistance. We thank Drs. Jay M. Goldberg, D. Kent Morest, and Matt Rasband for helpful discussions.
The authors declare no competing financial interests.
- Correspondence should be addressed to Dr. Anna Lysakowski, Department of Anatomy and Cell Biology, University of Illinois at Chicago, 808 South Wood Street, Chicago, IL 60605.