Amiloride-sensitive channels are a major contributor to mechanotransduction in mammalian muscle spindles

Abstract

We investigated whether channels of the epithelial sodium/amiloride-sensitive degenerin (ENaC/DEG) family are a major contributor to mechanosensory transduction in primary mechanosensory afferents, using adult rat muscle spindles as a model system. Stretch-evoked afferent discharge was reduced in a dose-dependent manner by amiloride and three analogues – benzamil, 5-(N-ethyl-N-isopropyl) amiloride (EIPA) and hexamethyleneamiloride (HMA), reaching ≥85% inhibition at 1 mm. Moreover, firing was slightly but significantly increased by ENaC δ subunit agonists (icilin and capsazepine). HMA's profile of effects was distinct from that of the other drugs. Amiloride, benzamil and EIPA significantly decreased firing (P < 0.01 each) at 1 μm, while 10 μm HMA was required for highly significant inhibition (P < 0.0001). Conversely, amiloride, benzamil and EIPA rarely blocked firing entirely at 1 mm, whereas 1 mm HMA blocked 12 of 16 preparations. This pharmacology suggests low-affinity ENaCs are the important spindle mechanotransducer. In agreement with this, immunoreactivity to ENaC α, β and γ subunits was detected both by Western blot and immunocytochemistry. Immunofluorescence intensity ratios for ENaC α, β or γ relative to the vesicle marker synaptophysin in the same spindle all significantly exceeded controls (P < 0.001). Ratios for the related brain sodium channel ASIC2 (BNaC1α) were also highly significantly greater (P < 0.005). Analysis of confocal images showed strong colocalisation within the terminal of ENaC/ASIC2 subunits and synaptophysin. This study implicates ENaC and ASIC2 in mammalian mechanotransduction. Moreover, within the terminals they colocalise with synaptophysin, a marker for the synaptic-like vesicles which regulate afferent excitability in these mechanosensitive endings.

Introduction

Mechanotransduction is a process of fundamental importance to all organisms, allowing them to detect mechanical events arising from their environment or within themselves, and thus make appropriate contextual responses to those events (Kung, 2005). Ultimately it must depend on the particular mechanical sensitivity of certain proteins that are likely to include ion channels, several examples of which are now known (Garcia-Añoveros et al. 1997; Hamill & Martinac, 2001; Martinac, 2004; Nicolson, 2005). They may be mechanically gated, or may show mechanical sensitivity in addition to being ligand- or voltage-gated (Calabrese et al. 2002; Lyford et al. 2002; Goodman & Schwarz, 2003; Peng et al. 2004, 2005). The simplest expression of a mechanotransduction system of this kind would presumably be a plasmalemmal ion or water channel gated by intermolecular forces (tension) in the lipid bilayer. Channels like this are probably present in prokaryotes at least (Hamill & Martinac, 2001; Corry & Martinac, 2008). However, metazoa require very diverse and specialised sensory systems of receptor cells and neurons, responsive to mechanical stimuli, in order to accommodate the large spatio-temporal range of mechanical events relevant to their lives (Ernstrom & Chalfie, 2002; Goodman, 2003; Bianchi, 2007). In many cases the receptor cells of multicellular animals, or the sensory terminals of mechanically sensitive neurons, are incorporated into sense organs. In mammals, examples include the hair cells of the cochlea and vestibule, and the sensory terminals of Pacinian and Meissner corpuscles, tendon organs and muscle spindles (Meyers et al. 2003).

The complete process of transduction, from input stimulus to frequency (or rate)-coded nerve impulses as output, is undoubtedly very complex in these mechanosensory organs of animals. For example, there is in general a component of mechanical filtering provided by accessory elements of the sense organ, such as the intrafusal muscle fibres of the muscle spindle (Banks, 2005) or the outer capsule of the Pacinian corpuscle (Mendelson & Lowenstein, 1964). What is more surprising is the occurrence of small, clear vesicles (‘synaptic-like vesicles’) in the sensory terminals of primary mechanosensory neurons, resembling the synaptic vesicles of chemical synapses (Bewick et al. 2005). Since the direct mechanical gating of an ion channel in the sensory terminal membrane could be expected to be sufficient to produce a receptor potential, these vesicles, although long recognised, have been largely ignored. We have now shown, however, that at least in the muscle spindle they do indeed play an important functional role in mechanosensory transduction since they appear to release glutamate in an activity-dependent manner, the glutamate having a self-excitatory action on the sensory terminals that is mediated by a non-canonical metabotropic glutamate receptor. The importance of this mechanism is clearly demonstrated by the powerful inhibition of the output of the spindle following application of PCCG-13, a specific blocker of the metabotropic glutamate receptor (mGluR) concerned (Bewick et al. 2005), yet its functional role remains unclear.

In order to clarify the relationship between the system of synaptic-like vesicles and the primary events of mechanotransduction, we are investigating candidate ion channels in the sensory terminals of the muscle spindle that may be directly gated by mechanical stimulation. Primary mechanosensory ion channels have yet to be identified definitively in any mammalian sense organ, but candidates include members of the DEG/ENaC and transient receptor potential channel (TRP) superfamilies (Ismailov et al. 1997; Satlin et al. 2001; Althaus et al. 2007). Here we present physiological, pharmacological and immunocytochemical evidence for the presence of epithelial sodium channels (ENaCs) and of their importance as at least one component of the primary mechanotransducer in the muscle spindle.

Methods

Animals and dissection

Adult male rats (350–600 g) were killed humanely by stunning and cervical dislocation in accordance with both the UK Schedule 1, Animals (Scientific Procedures) Act, 1986 and the ethical regulations and policies of The Journal of Physiology (Drummond, 2009). Fourth lumbrical nerve–muscle preparations from hind paws and saphenous nerves from the hind legs were dissected, cleaned and mounted in culture dishes lined with silicon rubber (Sylgard, Dow Corning, Stade, Germany) under constantly gassed (95% O₂–5% CO₂) saline containing (mm): 138.8 NaCl, 4 KCl, 12 NaHCO₃, 1 KH₂PO₄, 1 MgCl₂, 2 CaCl₂ and 11 glucose (Liley's solution; Liley, 1956), pH adjusted to 7.4. Muscle spindles from all four deep lumbricals were visualized by staining with the synaptic vesicle marker FM1-43 (10 μm dye, 2 h; Bewick et al. 2005) and dissected free from surrounding extrafusal muscle (see Fig. 7A) for Western blot analysis. Dissected spindles were then kept at 4°C and processed within 24 h for addition to the gel. All electrophysiology experiments were performed at room temperature (RT, 19–21°C).

Figure 7. Expression of ENaC subunits in spindles and extrafusal muscle fibres.

A, an FM1-43-labelled muscle spindle after dissecting free from extrafusal muscle fibres. 80–120 of these were prepared, extracted and used per gel in the Western blots. B, extracts of sensory muscle spindles (S) and extrafusal muscle (M) were fractionated by SDS-PAGE and analysed by immunoblotting using antibodies against α, β and γ ENaC. Bands of appropriate size for all three subunits were clearly identified. Another set of antibodies to the same subunits, generated independently by Professor Lawrence Palmer, gave bands at identical positions. In contrast, immunoblotting for the δ subunit was faint and hard to discern above background (data not shown). The positions of molecular weight markers are indicated.

Drug preparation

Amiloride, ethyl isopropyl amiloride (EIPA) and hexamethylene amiloride (HMA) (Sigma-Aldrich, Gillingham, UK) solutions were prepared daily at 1 mm concentration in Liley's saline. Benzamil (10 mm; Sigma-Aldrich) stock solution was prepared daily in 100% methanol. Capsazepine (100 mm) and icilin (240 mm; Tocris) stock solutions were prepared daily in DMSO. A 1 mm stock solution of tetrodotoxin (TTX; Alomone Labs, Israel) was prepared in distilled water, aliquoted and stored at −20°C. All dilutions were made with fresh Liley's solution each day.

Electrophysiology

Fourth lumbrical muscle

The second phalanx of the fourth toe (distal insertion of 4th lumbrical) was pinned to the Sylgard-lined base of the organ bath and the other tendon hooked to a 3-axis micromanipulator (Prior, UK). Electroneurograms (ENGs) were recorded en passant from the nerve in paraffin oil using silver wire electrodes. Signals were amplified (A 103, Isleworth Electronics, UK and 8102, CFP, UK preamplifiers in series), displayed on an oscilloscope (DSO 400, Gould, UK) and captured on computer hard-drive (WinWCP software, John Dempster, University of Strathhclyde, UK). Muscle length was adjusted so the firing was minimal, the muscle was then manually stretched by 1 mm, using the vernier calibration on the micromanipulator, and left at this length (for 5 s) before returning to the original length (for 5 s). All ENGs were recorded for three consecutive stretch, hold and release cycles. At least three consistent ENGs were recorded as pre-drug controls before increasing drug concentrations were applied (1, 10, 100 and 1000 μm) for at least 1 h each, recording every 20 min. Parallel no-drug control preparations were made over the same time course.

Saphenous nerve

The entire nerve was excised from the medial aspect of the leg and placed under constantly gassed Liley's solution. One end was stimulated by a suction electrode and 12 compound action potentials (CAPs) were recorded every 5 min with silver wire electrodes (as above). Amiloride analogues were applied at the highest concentration used in the recording of stretch-evoked responses from muscle spindles (1 mm). TTX was applied at 1 μm. The purely sensory composition of the saphenous nerve and its greater length made it a suitable subject for these experiments, which tested the drugs for nerve conduction block. Time-matched control preparations were run in parallel.

Data analysis

Afferent mean firing frequency was determined for the first 0.5 s of the ‘hold’ phase, immediately after the dynamic phase, for three consecutive occasions for each time point. The mean CAP amplitude in the saphenous nerve was determined for 12 repetitions at each time point. All electrophysiology data are expressed as mean ±s.e.m. The significance of differences between means was assessed by Student's t test, as indicated in the text. Unpaired t tests for samples of equal or unequal variance were performed according to a prior F test.

Western blotting

Tissue extraction

Spindles were labelled with FM1-43 and dissected free from surrounding extrafusal muscle fibres (e.g. Fig. 7A). For each sample, 80–120 spindles were resuspended in RIPA buffer (10 mm Tris, 150 mm NaCl, pH 7.4, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS), sonicated (3 × 20 s, 50% power; Vibra-cell, Sonics & Materials Inc., CT, USA) and incubated on ice for 15 min, prior to centrifugation at 15 000 g for 10 min at 4°C. Cell lysates were mixed with reducing sample buffer (Laemmli, 1970), to give a final concentration of 2% SDS and 100 mm dithiothreitol, and incubated at 100°C for 5 min.

SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblotting

Cell lysates were fractionated by SDS-PAGE, transferred to nitrocellulose membranes and blocked with 5% non-fat milk powder in Tris-buffered saline (10 mm Tris, 150 mm NaCl), pH 7.4, containing 0.1% Tween 20 for 1 h at RT. Blots were incubated overnight with primary antibody at 4°C, washed and incubated with HRP-conjugated secondary antibody for 1 h at RT. Two sets of primary antibodies were compared. Commercially available antibodies against α, β, γ and δ ENaC were obtained from Santa Cruz Biotechnology Inc. (Santa Cruz, CA, USA) and antibodies against α, β and γ ENaC (Masilamani et al. 1999; Ergonul et al. 2006) were the generous gift of Professor Lawrence G. Palmer, Cornell University, USA. Immunoreactive species were visualized using enhanced chemiluminescence.

Immunofluorescence

Teased muscles

Twenty-seven Adult rats (PVG or Wistar strains) were killed humanely using a Schedule 1 method (Animals (Scientific Procedures) Act, 1986). Deep lumbrical muscles were isolated from the hind feet, and fixed in 4% formaldehyde in 0.1 m sodium phosphate buffer, pH 7.4. While still immersed in fixative, the muscles were teased and squashed between a pair of microscope slides to facilitate penetration of the antibodies into the sensory region of the spindles. The muscles were then rinsed thoroughly in the buffer and blocked for 1 h at room temperature in PBS (pH 7.4) containing 0.25% BSA, 0.3% Triton X-100 and 0.02% sodium azide. The blocking buffer was also used as the antibody diluent. Preparations were double-stained with anti-ENaC subunit or anti-ASIC2 (rabbit polyclonal antibodies as follows: anti-α, β and γ ENaC subunits from Santa Cruz Biotechnology Inc., raised against recombinant proteins using sequences close to the N terminus, an intermediate portion or the C terminus of the respective subunits so as to minimise the possibility of cross-reactivity; anti-δ ENaC from Millipore (Watford, UK), raised against a 19 aa peptide near the N terminus; goat polyclonal anti-ASIC2 from Santa Cruz Biotechnology Inc., raised against a peptide mapping at the N terminus of human ASIC2 (all used at 5 μg ml⁻¹) and mouse monoclonal anti-synaptophysin (Millipore, 1 μg ml⁻¹) primary antibodies. Synaptophysin is often regarded as a synaptic vesicle marker, but is also present in the sensory endings of muscle spindles (De Camilli et al. 1988), presumably associated with the synaptic-like vesicles in these endings (Bewick et al. 2005). We were therefore able to use anti-synaptophysin labelling as a spindle marker. For negative controls, non-immune IgG was used in place of the anti-ENaC/ASIC2 primary antibodies. Blocking of anti-ASIC2 immunoreactivity by prior incubation of the antibody with an excess of the antigenic peptide (Santa Cruz Biotechnology Inc.) was included as an additional control. Primary incubation was for 48 h at 4°C, followed by 3 × 10 min washes in PBS–BSA–azide. Secondary incubation used Alexa Fluor (AF)-conjugated antibodies (AF 594 goat anti-rabbit IgG and AF 488 goat anti-mouse IgG for ENaC primaries; AF 594 donkey anti-goat IgG and AF 488 donkey anti-mouse IgG for ASIC2; all from Invitrogen, Eugene, OR, USA), diluted 1: 250, for 1 h at RT followed by 3 washes in PBS. Muscles were mounted in glycerol–PBS (Citifluor AF1, Agar Scientific, Stansted, UK) for confocal microscopy (Biorad Microradiance confocal microscope, Ar and HeNe lasers, NEOFLUAR plan-objective 40×/1.30; AF 488, emission filter 530/60BP, AF 594, emission filter 570LP, Biorad, Hemel Hempstead, UK).

Data analysis

Pairs of images were acquired sequentially at single confocal planes to ensure direct comparability between the spatial relationships of the sources of fluorescence for the specific anti-ENaC or -ASIC2 subunit and the anti-synaptophysin immunoreactions in each spindle. In order to control for variation in the accessibility of individual spindles to antibodies the software package ImageJ (public domain Java image processing program from NIH, USA –http://rsbweb.nih.gov/ij/) was used to calculate the ratio of raw intensities of anti-ENaC or anti-ASIC2 to anti-synaptophysin immunofluorescence for defined regions of interest (ROIs), including the sensory terminals. ROIs were uniquely defined for each confocal plane, and the same ROI was used to collect data from the red (subunit-specific) and green (synaptophysin) channels. This allowed statistical comparisons to be made between the intensities of immunostaining for the ENaC or ASIC2 subunits and the various controls, when expressed as a ratio of the corresponding intensity of anti-synaptophysin staining in each spindle. Ratiometric data were log₁₀ transformed for analysis. Comparisons between means were analysed in SPSS (version 14.0) using a one-way ANOVA and post hoc Tukey test. The absolute values of the ratios cannot simply be interpreted in terms of biological significance; rather, they reflect most strongly the relative efficiency of fluorescence using the particular combinations of fluorochrome, laser and emission filters employed. Furthermore, comparisons between the immunofluorescence ratios for particular ENaC subunits can only be made by assuming similar affinities of the various antibodies for the target antigens. On the other hand, the dual-labelling technique allowed us to compare the localisation of each specific subunit with that of synaptophysin.

Processing and detailed analysis of the images for colocalisation was carried out using, in particular, the intensity correlation analysis plug-in for MBL-ImageJ (McMaster Biophotonics Laboratory; http://www.macbiophotonics.ca/index.htm), which is based on the method of Li et al. (2004). Individual images were first adjusted by expanding the intensity spectrum to cover the entire 8-bit grey scale range (0–255), so as to compensate for differences in antibody affinity; fluorescence efficiencies; and laser, dye and filter matching for the two fluorophores conjugated to the secondary antibodies. The results were displayed as images, both qualitatively using the familiar merged channels, and quantitatively using product of differences from the means (PDM). PDM is calculated using normalised intensities and is therefore insensitive to the adjustment in grey scale range carried out as above. The limits of the values that PDM can assume depend on the degree of skewness in the intensity spectrum, but in general the values cluster around 0. Positive values of PDM arise when corresponding pixels in the two images co-vary around their respective means, and therefore indicate colocalisation, whereas negative values indicate segregation.

Results

Electrophysiology

Inhibition of muscle spindle afferent discharge by amiloride and its analogues

Amiloride and its analogues inhibit a wide variety of sodium channels including those of the DEG/ENaC family, some of which may have a role in mammalian mechanotransduction (Garty & Palmer, 1997; Drummond et al. 1998; Mano & Driscoll, 1999; Page et al. 2004). To test functionally for the presence of these amiloride-sensitive channels in the sensory endings of muscle spindles, we applied amiloride or one of three analogues to nerve–muscle preparations of deep lumbrical muscles of rat and observed the effects on rate of afferent firing in response to a maintained stretch. The deep lumbrical muscles contain no Golgi tendon organs (R. W. Banks, unpublished observations), so the recorded discharge is due entirely to spindle afferents. Following control measurements in Liley's solution, amiloride (n= 14 preparations), benzamil (n= 12), EIPA (n= 13) or HMA (n= 16) were applied at increasing concentrations (1 μm, 10 μm, 100 μm and 1 mm final concentration in Liley's solution) to the nerve–muscle preparations. All drugs reduced the spiking frequency in a concentration-dependent manner (Fig. 1A and B). In contrast, afferent firing was not changed by incubation in drug-free Liley's solution in any of the time-matched controls (n= 6).

Figure 1. Spindle afferent discharge is sensitive to the ENaC blocker amiloride and its analogues.

A, electroneurograms of static-phase spindle spiking during 1 mm stretches of rat 4th lumbrical muscles. Increasing amiloride concentrations reduced spiking frequency. B, comparison of inhibition of afferent discharge by amiloride and its analogues. Amiloride, benzamil and EIPA produced equipotent inhibition of firing; HMA had little effect below 100 μm, but produced almost total abolition at 1 mm.

For amiloride, benzamil and EIPA, the reduction was already highly significant at a concentration of 1 μm (P < 0.01 in all cases). Higher concentrations progressively inhibited spindle output further. The IC₅₀ for amiloride, benzamil and EIPA was approximately 10 μm (respectively 63.7%, 46.3% and 56.4%, P < 0.0001 in all cases). The highest concentration of each drug used, 1 mm, produced the most robust block, but there were no significant differences in effectiveness between the drugs at any of the concentrations used. Thus, these amiloride analogues and amiloride itself seem to be equipotent in their inhibition of afferent firing.

Compared with amiloride and the other analogues, HMA differed in its pattern of inhibiting afferent firing. At 1 μm and 10 μm, the drug had little, if any, effect on the firing frequency. At higher concentrations, however, the effects were very marked. HMA at 100 μm depressed the afferent discharge significantly (P < 0.002) and 1 mm produced the most profound block of all the drugs used, abolishing the firing in most preparations. HMA totally blocked firing in 12 of 16 preparations after only 40 min. In contrast, amiloride, benzamil and EIPA rarely induced a total abolition at this time. Indeed, even after 2 h at 1 mm concentration, a mean of 14.2%, 11.4% and 9.8%, respectively, of the original firing still remained, and full abolition occurred in only ∼50% of preparations (amiloride, 5/14; benzamil, 5/12 and EIPA, 6/13 preparations).

Amiloride and its analogues do not change saphenous nerve CAP amplitude

Carr et al. (2001) reported that amiloride and its analogues, dimethylamiloride and benzamil, reduced the afferent outputs of airway stretch receptors in guinea pig (Carr et al. 2001). They concluded that this was not due to block of ENaC, but rather to a reduction in the excitability of the afferent neurons, possibly by inhibiting voltage-dependent Na⁺ channels. To test whether the reduction in afferent firing seen in our muscle spindle preparations could be due to blockage of voltage-gated Na⁺ channels, each of these drugs was also applied to isolated saphenous nerve preparations and the CAP amplitude was measured over a 90 min incubation time. After measuring the initial, stable CAP amplitude, amiloride (n= 6 preparations), benzamil (n= 7), EIPA (n= 6) or HMA (n= 5) was applied at 1 mm. This was the highest concentration tested for effects on afferent discharge and produced a robust block of the stretch-induced response. None of the amiloride analogues altered the sensory nerve CAP during a 90 min exposure, neither in amplitude nor shape (Fig. 2A and B). As a control, the selective voltage-gated sodium channel blocker, TTX, was applied at 1 μm concentration (n= 7) using the same protocol as for amiloride and its analogues (Fig. 2C). During TTX incubation, the CAP progressively reduced in amplitude by a mean of 66.8% (P < 0.01). In each case, the CAP amplitude subsequently recovered fully upon washing in drug-free Liley's solution (Fig. 2C).

Figure 2. Amiloride and its analogues do not inhibit action potential propagation.

A, amiloride changed neither the amplitude nor the shape of the compound action potential (CAP) recorded in the saphenous nerve. In contrast TTX, a selective blocker of voltage-gated Na⁺ channels, reduced CAP amplitude very effectively. B, none of the ENaC inhibitors reduced the CAP amplitude of the purely sensory saphenous nerve at 1 mm, the maximum concentration tested on afferent endings. C, lack of tissue penetration for the drugs should not have been an issue, since 1 μm TTX very effectively reduced CAP amplitude in the saphenous nerve under the same conditions.

ENaC δ subunit agonists produce small significant increases in stretch-evoked afferent discharge

Classical ENaC channels are thought to be heteromers, including α, β and γ subunits, with the α subunit being necessary for pore-forming abilities. However, some evidence suggests a δ subunit can substitute for α in some cases (Waldmann et al. 1995). Therefore, in a further series of experiments, spindles were tested for the functional presence of the δ ENaC subunit, using the agonists icilin (n= 5) and capsazepine (n= 7) at 1 μm, 10 μm, 100 μm and 1 mm concentration (Yamamura et al. 2004, 2005). At around 10 μm both agonists produced a small but significant (P < 0.03) increase in the rate of afferent discharge in response to the standard stretch protocol, without affecting saphenous nerve CAP amplitude or shape (Fig. 3).

Figure 3. δ ENaC-subunit agonists caused small increases in stretch-evoked afferent discharge.

A, electroneurograms showing that the δ ENaC-subunit agonists capsazepine and icilin increased afferent firing. B, at 10 μm, both capsazepine and icilin produced small, significant increases in the firing (*P < 0.03) but effects were reduced or abolished at higher concentrations. C, capsazepine and icilin (1 mm) did not change saphenous nerve CAP shape or amplitude.

Immunofluorescence

ENaC subunits

Subjectively the immunoreactivity (IR) of antibodies against each of the four ENaC subunits shown by the sensory endings of muscle spindles exceeded that of control preparations, particularly for α, β and γ subunits (Fig. 4A). Subunit-specific staining was not confined to the sensory endings, however, and in particular there was some staining associated with intracellular structure in both intrafusal and extrafusal muscle fibres. For a more quantitative comparison of the IR for each subunit, intensity data were collected from regions of interest defined using ImageJ so as to include the sensory terminals (see Methods). There was significant difference in variation among the log-transformed ratios (intensity of anti-ENaC: anti-synaptophysin immunofluorescence) for the data from the four types of subunit and the controls (one-way ANOVA; F= 26.9, P < 0.001). The log-transformed ratios for each of the α, β and γ subunits was significantly greater than that of the controls (post hoc test, P < 0.001, in each case), indicating highly significant labelling for these three subunits, but there was no significant difference between transformed ratios for the δ subunit and the controls (Fig. 4B).

Figure 4. Anti-ENaC subunit immunoreactivity localises to sensory terminals of rat muscle spindles.

A, double immunofluorescent labelling of the sensory regions of rat muscle spindles, comparing anti-ENaC subunit with anti-synaptophysin reactivities. Upper panels (red): anti-ENaC α, β, γ or δ immunoreactivity; lower panels (green): anti-synaptophysin immunoreactivity of the corresponding spindles. Control: anti-ENaC antibody replaced with non-immune rabbit IgG. Immunoreactivity was clearly visible with antibodies to the α, β and γ subunits, but there was little reaction with the anti-ENaC δ antibody; this is in contrast to the control where no specific reactivity was discernible. The intensities of each of the red-channel images have been increased by the same amount to enable clearer reproduction in print. B, mean intensity ratios of anti-ENaC subunit: anti-synaptophysin immunofluorescence and controls in double-labelled muscle spindles. The log-transformations of the ratios of labelling intensities were used for statistical analysis. The ratios for the α, β and γ subunits were all significantly greater than the controls (post hoc Tukey test, P < 0.001 in each case, following one-way ANOVA, F= 26.9, P < 0.001).

ASIC2

Another member of the channel family is the acid-sensing ion channel ASIC (also called the mammalian brain sodium channel, BNaC), and particularly ASIC2, which is also amiloride sensitive (Garcia-Añoveros et al. 2001). It has been localised to peripheral terminals of most large DRG neurons (low-threshold mechanosensors), particularly in the specialised cutaneous mechanosensory terminals. However, its presence has not previously been examined in muscle spindle primary afferent terminals. There are no selective agonists/antagonists for functional testing for the presence of ASIC2 subunits. However, anti-ASIC2 immunoreactivity was demonstrated as shown in Fig. 5A. ASIC2-specific staining (red) was clearly evident compared with either of the controls (non-immune goat IgG or peptide block). Quantitative analysis confirmed significant variation between transformed ratios (one-way ANOVA; F= 11.3, P < 0.001) with ASIC2: synaptophysin staining intensity significantly higher than either non-immune goat IgG (post hoc test, P < 0.002) or the peptide block (post hoc test, P < 0.003) (Fig. 5B).

Figure 5. Anti-ASIC2 immunofluorescence localises to sensory terminals of rat muscle spindles.

A, double immunofluoresent labelling of the sensory regions of rat muscle spindles, comparing anti-ASIC2 with anti-synaptophysin reactivities. Upper panels (red): anti-ASIC2 immunoreactivity; lower panels (green): anti-synaptophysin immunoreactivity of the corresponding spindles. Anti-ASIC2 immunoreactivity was evident on sensory terminals in contrast to controls (anti-ASIC2 antibody blocked with peptide, or replaced with non-immune goat IgG) where specific immunoreactivity of the sensory terminals was not visible. The intensities of each of the red-channel images have been increased by the same amount to enable clearer reproduction in print. B, mean intensity ratios of ASIC2: anti-synaptophysin immunofluorescence and controls in double-labelled muscle spindles. The log-transformations of the ratios were used for statistical analysis. There was significant variation among the ratios (one-way ANOVA; F= 11.3, P < 0.001) with ASIC2: synaptophysin staining intensity significantly higher than either non-immune goat IgG (post hoc Tukey test, P < 0.002) or peptide block (post hoc test, P < 0.003) controls.

Colocalisation

Spindle primary afferent endings express synapsin I and synaptophysin (De Camilli et al. 1988), presumably associated with the regulated secretion of the synaptic-like vesicles in these endings (Bewick et al. 2005). Synaptophysin, therefore, acts as a useful marker for a quantitative analysis of the colocalisation of ENaCs and ASIC2 to primary terminals. Double-labelling of teased spindles with antibodies against synaptophysin and ENaC or ASIC2 subunits allowed us to examine the extent to which the antibodies were colocalised on binding their respective antigens (see Methods). Immunofluorescent labelling of each of the ENaC subunits, α, β and γ, and of the ASIC2 subunit, was clearly colocalised with that of synaptophysin, as illustrated by examples shown in Fig. 6A and B. Notice that the sensory terminals are the only features that show a strong colocalisation signal, indicated by the positive intensity differences from the means (PDMs) in corresponding pixels in both images. Positive values of PDM arise when corresponding pixels in the two images co-vary around their respective means, and therefore indicate colocalisation, whereas negative values indicate segregation. From these confocal images, which produce optical sections through the terminal, notice also that most of the ENaC and ASIC2 subunit-specific immunofluorescence associated with the sensory terminals arises from within the terminals.

Figure 6. Intensity correlation analysis shows ENaC and ASIC2 colocalisation with synaptophysin.

A, representative images of a spindle primary ending double labelled for anti-ENaC and anti-synaptophysin immunofluorescence to demonstrate intensity correlation analysis after the method of Li et al. 2004 using ImageJ software. The top two panels show ENaCα (red) and synaptophysin (green) immunoreactivity, respectively. The third panel is the merge of the first two with colocalisation appearing as yellow/orange. The fourth panel is a new image where each pixel is equal to the PDM value at that location. The PDM value is the product of the differences from the mean for each pixel, calculated as follows: (red intensity – mean red intensity) × (green intensity – mean green intensity). The image is pseudocoloured according to PDM value (blue = modest colocalisation, red = higher, white = highest); pixels that are below average in both channels are excluded (black). Colocalisation is especially high within the sensory terminals, as shown by white pixels. Immunolabelling of each of the four ENaC subunits colocalised with synaptophysin staining in a similar manner. B, representative images of a spindle primary ending double labelled for anti-ASIC2 and anti-synaptophysin immunofluorescence to demonstrate intensity correlation analysis (as for A). The top two panels show ASIC2 (red) and synaptophysin (green) immunoreactivity, respectively. The third panel is the merge of the first two with colocalisation appearing as yellow/orange. The fourth panel is constructed from PDM values as in A. As with the ENaC subunits, ASIC2 immunoreactivity is most clearly colocalised with that of synaptophysin in the sensory terminals.

Western blotting for ENaC subunits

Given the evidence from pharmacology and teased-tissue immunofluorescence for the presence of ENaC subunits in spindle primary sensory endings, it should be possible to extract the relevant proteins from isolated muscle spindles and detect them by immunolabelling from Western blots. Muscle spindles were isolated by first labelling with FM1-43 and dissecting them free of surrounding extrafusal muscle fibres (Fig. 7A). Spindles (80–120 per gel) were isolated in this way and the proteins extracted by RIPA buffer incubation and sonication. Similar quantities of extrafusal muscle fibres were processed in the same way for comparison. The presence of ENaCs in both muscle spindles and extrafusal muscle was supported by the presence of bands of appropriate molecular weight (85–90 kDa) for α, β and γ subunit antibodies (Fig. 7B), a result that is consistent with the localisations of the subunits shown by immunocytochemistry (see above). Similar results were found for both commercially produced antibodies used for teased tissue labelling and independently produced antibodies (generous gift from Lawrence Palmer, Cornell University, USA). Labelling for the δ ENaC subunit was difficult to discern unambiguously above background.

Discussion

Specialised mechanosensory nerve endings and end-organs are very diverse, but they probably share a common underlying mechanism of sensory transduction (Benos, 2004). At the molecular level, mechanotransduction is thought to rely on the gating of a channel protein by mechanical deflections transmitted to the channel, perhaps via an extracellular tether. The channel is probably also connected to the cytoskeleton by an intracellular link (Hamill, 2006), perhaps through ankyrin and fodrin from the actin-binding family (Smith et al. 1991).

The mechanosensitive ion channels have not been identified definitively in mammals. However, in invertebrates, there are a number of likely candidates (Goodman & Schwarz, 2003). These are the amiloride-sensitive channels (ASCs) and the transient receptor potential (TRP) channels. Recent studies have also explored whether acid-sensing ion channel 2 (ASIC2) might have a role in auditory transduction (Peng et al. 2004) and whether there is mechanical gating of N- and L-type Ca²⁺ channels (Calabrese et al. 2002; Lyford et al. 2002; Peng et al. 2005). However, the mechanosensitive channels of many proprioceptors, including mammalian muscle spindles, are unknown.

The main candidates in mammals are members of the amiloride-sensitive degenerin/epithelial Na⁺ channel (DEG/ENaC) superfamily (Ismailov et al. 1997; Satlin et al. 2001; Althaus et al. 2007). For example, the ASIC2 brain sodium channel (now known to be the same protein as BNC1/BNaC1) is important for normal mechanosensitivity in both rapidly and slowly adapting mechanosensory endings in glabrous and hairy skin (Price et al. 2000; Garcia-Añoveros et al. 2001). The COOH terminus of ENaCs has been shown to contribute to the modulation of the channel activity by actin (Jovov et al. 1999; Copeland et al. 2001) and the COOH terminus is physically and functionally linked to the cellular cytoskeleton through F-actin (Mazzochi et al. 2006). Thus, the ENaC channels exhibit at least some of the characteristics expected of a mechanosensitive ion channel. Moreover, other ENaC family members, those of the ASC family, have been shown by immunocytochemistry to be expressed in baroreceptors (Drummond et al. 1998), various mechanoreceptor structures in the rat foot pad (Drummond et al. 2000) and in sensory nerve endings of vibrissae (Fricke et al. 2000). Their distribution on sensory nerve endings in the rat larynx has been taken to suggest a dual role in both mechano- and chemo-transduction (Yamamoto & Taniguchi, 2006).

However, ENaC family members functioning as stretch-activated channels in mammals is controversial. Attempts to activate mechanically the DEG/ENaC channels, expressed in heterologous systems, have failed (Awayda & Subramanyam, 1998) and not everyone agrees that ENaCs can act as biological mechanosensors (Rossier, 1998), but this disagreement may relate to the fact that the channel can be blocked by intracellular ATP (Ma et al. 2002). There is growing evidence that laminar shear stress represents an adequate stimulus for the mechanical activation of ENaC (Althaus et al. 2007; Carattino et al. 2007; Fronius & Clauss, 2008). Furthermore, in vivo imaging studies of C. elegans mechanosensory neurones revealed a specific role for the MEC-4 isoform of the DEG family in the process of gentle touch sensation (Suzuki et al. 2003), and electrophysiological studies show that this protein is capable of transducing mechanical signals (O’Hagan et al. 2005). The structural similarity between mammalian ENaC genes and the C. elegans MEC/DEG genes indicates that they encode similar cation channels and support the hypothesis that mammalian ENaCs are involved in mechanotransduction.

In terms of amiloride sensitivity, ENaCs can be divided into low- and high-affinity types (Smith & Benos, 1991). They have also been differentiated on the basis of their single-channel conductance in reabsorptive tissue (Palmer, 1992). These early classifications together with the phylogenetic and structural comparison of DEG/ENaC family members (Price et al. 1996; Garcia-Añoveros & Corey, 1997) predict the presence of diverse ENaCs comprising various combinations of subunits with different biophysical and pharmacological properties. There are at least four subunits (α, β, γ and δ), different combinations of which can occur. The functional relevance of subunit composition is not known in detail, but different combinations may confer different transduction mechanisms (Weisz & Johnson, 2003). Homology between the four, cloned, human ENaC subunits is relatively low (ranging from 27% to 37%; Waldmann et al. 1995), but structurally they all have two transmembrane domains (M1 and M2), a large extracellular loop and intracellular amino- and carboxy-terminals in common (Benos & Stanton, 1999). The amiloride binding sites are on the extracellular-loop either on the α or on the δ subunit (Kellenberger et al. 2002). Despite their low homology, the α and δ-subunits can compose functional ion channels either in αβγ (Duc et al. 1994) or δβγ combination (Waldmann et al. 1995). Originally it was proposed that they are inserted into the membrane with a stoichiometry of 2α: 1β: 1γ (Firsov et al. 1998; Kosari et al. 1998), and even a nonameric structure of three of each subunit has also been hypothesised (Eskandari et al. 1999). More recently, a cloned ASIC channel was found to have a stoichiometry of 1: 1: 1, and it was suggested this almost certainly meant ENaCs are obligate trimers (Jasti et al. 2007). Channels with αβγ profile have been characterized in urinary bladder, renal collecting duct, distal colon, sweat and salivary glands, lung, and taste buds (Garty & Palmer, 1997) and the ENaC δ subunit is expressed mainly in pancreas, testis, ovary and brain but can be found in smaller amounts in heart, skeletal muscle, kidney, thymus, lung, liver and in colon as well (Waldmann et al. 1995). The reports of ENaC subunits in skeletal muscle support our own findings from both Western blotting and immunofluorescence.

In this study, we have investigated the importance of ENaC channels in mechanotransduction of rat muscle spindles by measuring the sensitivity of stretch-evoked afferent discharge to amiloride and three of its derivatives – benzamil, 5-(N-ethyl-N-isopropyl) amiloride (EIPA) and hexamethyleneamiloride (HMA) – and the δ subunit agonists icilin and capsazepine. Afferent discharge was reduced in an amiloride-sensitive manner, while it was slightly but significantly increased by δ subunit agonists. The relationship between the structure and activity of the amiloride analogues used here is inconsistent with either the high-affinity ENaC, where amiloride ≫ EIPA activity, or the Na⁺/H⁺ antiporter, where EIPA ≫ amiloride activity, (Kleyman & Cragoe, 1988; Rusch et al. 1994). Rather, it is similar to that of the low-affinity mechanosensory Na⁺ current in the Xenopus oocyte (Lane et al. 1992), though in that case benzamil was somewhat more active than amiloride.

Western blot analysis for the detection of ENaC subunits clearly identified antibody binding to the α, β and γ subunits, in both muscle spindles and extrafusal muscle. When interpreting the results of the Western blots, it must be recalled that the sensory terminals represent only very small proportions of the isolated muscle spindles. Fluorescent immunohistochemistry demonstrated that these subunits are expressed at relatively high levels in the sensory terminals, though some reactivity was also seen in intrafusal and extrafusal muscle fibres. If the antibodies used for the immunocytochemical localisation of the ENaC subunits have similar affinities, the relative intensities of immunostaining indicate that the α subunit may be present at lower levels than the β and γ subunits but, given the ASIC2 immunoreactivity detected in spindle sensory terminals, the ASIC2 may substitute in some channels for the ENaC α subunit in mechanotransduction (Garcia-Añoveros et al. 2001). Conversely, labelling for the δ subunit did not significantly exceed control levels and δ subunit agonists had only marginal effects on firing. Further studies with more selective labelling and pharmacological agents are needed to determine its presence definitively.

The immunoreactivity colocalisation analyses comparing the α, β and γ ENaC subunits and the ASIC2 subunit with synaptophysin indicates that much of the labelling for these proteins is strongly colocalised within the terminals. How closely, or otherwise, it is associated with the synaptic-like vesicles (SLVs) previously described by us (Bewick et al. 2005) awaits further study at the ultrastructural level. We have shown that the SLVs recycle in an activity-dependent manner and that they probably release glutamate which has an autocrine-like excitatory action on the sensory terminals via a phospholipase-D-coupled metabotropic receptor (Bewick et al. 2005). If the colocalisation between SLVs (synaptophysin) and the ENaC subunits is borne out at the ultrastructural level, our present analysis raises the further interesting possibility that the primary mechanosensitive cation channel is delivered to its site of action in the sensory terminal plasmalemma by SLV fusion. The inhibitory effect of syntaxin 1A on ENaC current when co-expressed with ENaC in the Xenopus oocyte has been cited as evidence for such a vesicle-mediated delivery mechanism (Qi et al. 1999), and immunogold labelling of ENaC subunits in rat kidney has also shown that much of the protein is indeed intracellular (Hager et al. 2001). Alternatively, the ENaCs may occupy a distinct vesicle population, with separate fusion control mechanisms, although regulation of this fusion would probably include v-SNARE and t-SNARE proteins, including syntaxins.

In conclusion, therefore, we present here consistent evidence using electrophysiology, pharmacology, molecular biology (Western blotting) and immunocytochemistry in favour of a role for members of the DEG/ENaC superfamily as primary mechanotransducers in the mammalian muscle spindle. The major ion channel appears to be a low-affinity amiloride-sensitive channel like that expressed in Xenopus oocyte (Lane et al. 1992), and somewhat different from the high-affinity constitutively active ENaC of transporting epithelia, and also different from the Na⁺/H⁺ antiporter. Its subunit composition may include ASIC2 as well as α, β and γ ENaC. In some channels, δ ENaC may substitute for the ASIC2 and/or α ENaC. These channels would all be expected to have high permeability to Na⁺, but little if any for Ca²⁺. Ca²⁺ permeability is relevant because a small, mechanically sensitive Ca²⁺ current remains when the Na⁺ current is removed (Hunt et al. 1978). It is possible that the entire current is carried by an amiloride-sensitive, non-selective cation channel. However, even at 10 mm, Ca²⁺ was unable to substitute for Na⁺ in the receptor potential despite a Nernst potential under these conditions (even assuming a relatively high intracellular [Ca²⁺] of 1 μm) of more than +100 mV. Conversely, blocking the Ca²⁺ current, or omitting Ca²⁺ from the medium, had little effect on the receptor potential. These amiloride-sensitive channels, therefore, are likely to be responsible for the Na⁺ component of the receptor potential depolarisation of the sensory terminals caused by muscle spindle stretch (Hunt et al. 1978). If this is the case, the ion channel responsible for the Ca²⁺ permeability has yet to be identified.

Glossary

Abbreviations

ASC

amiloride-sensitive channel

ASIC

Acid-sensing ion channel

DEG

amiloride-sensitive degenerin channel

EIPA

5-(N-ethyl-N-isopropyl) amiloride

ENaC

epithelial sodium channel

HMA

hexamethyleneamiloride

TRP

transient receptor potential channel

Author contributions

This work was performed at the University of Aberdeen (electrophysiology, pharmacology and Western blotting) and the University of Durham (immunocytochemistry and confocal imaging). G.S.B. and R.W.B. conceived and designed the experiments, while A.S., F.S. and I.H. collected and analysed the data. All authors contributed significantly to data interpretation and drafting/revising the article for important intellectual content. All authors approved the final version to be published.