Homeostatic Control of Sensory Output in Basal Vomeronasal Neurons: Activity-Dependent Expression of Ether-à-Go-Go-Related Gene Potassium Channels

Activity-dependent expression of K⁺ channel genes in the VNO

In many neurons, the balance of intrinsic ionic conductances is constantly adjusted as a function of ongoing activity. One mechanism to set a target output range is regulated expression of specific K⁺ channels (Zhang and Linden, 2003). To determine whether transcription of K⁺ channel genes is activity regulated in the mouse VNO, we surveyed expression profiles of adult male C57BL/6 mice that had been singly housed either in an essentially odor-free environment (deprived group) or exposed to rich sources of pheromonal cues (stimulated group). Stimulation cycles (7 d) were composed of 6 h exposure to fresh male urine (2:00–8:00 P.M.), 12 h exposure to female-soiled bedding (8:00 P.M. to 8:00 A.M.), and an additional 6 h without stimulation (8:00 A.M. to 2:00 P.M.) to allow for relief from potential sensory adaptation (Fig. 1a). Vomeronasal K⁺ channel transcription levels were compared between stimulated and deprived tissue using a total of seven Affymetrix microarrays (four experiment and three control replicates), each chip probing pooled mRNA from five animals. Applying conservative quality control and evaluation criteria (see Materials and Methods), 5 K⁺ channel genes were found both consistently detected and differentially expressed (Fig. 1b; see supplemental Table 1, available at www.jneurosci.org as supplemental material, for a summary of results). The most pronounced changes were observed for ERG1 (upregulation) and the modulatory single transmembrane domain β-subunit MiRP2 (downregulation), which was reported to inhibit ERG1 currents when coexpressed in Xenopus oocytes (Schroeder et al., 2000). We next determined vomeronasal expression of ERG1–3 and MiRP2 by RT-PCR. Using specific primers for both ERG1 isoforms (ERG1a and ERG1b), ERG2, ERG3, and MiRP2, we detected transcripts of all genes. Transcript identity was confirmed by nested PCR (Fig. 1c) and sequencing. Bands of the appropriate size were also obtained from cerebellum (Papa et al., 2003; Guasti et al., 2005) and heart (ERG1a, ERG1b, and MiRP2) (Shi et al., 1997) but not from cultured prostate cancer cells (LNCaP).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Activity-dependent K⁺ channel gene expression in the VNO. a, Schematic diagram illustrating the experimental strategy used for pheromone exposure (left) and stimulus deprivation (right) of singly housed C57BL/6 mice. b, Differentially expressed K⁺ channel genes in the VNO. Analysis (see Materials and Methods) of seven microarrays (4 stimulated/3 deprived) reveals both consistent expression (7 of 7 arrays) and significant (p ≤ 0.029) regulation of the following genes: Kcnh2 (ERG1; up), Kcna5 (up), Kcne3 (MiRP2; down), Kcna2 (down), and Kcnj16 (down). For each gene, RefSeq accession numbers, Affymetrix identifications, and average fold changes (mean ± SEM; 12 independent comparisons) are specified. c, Vomeronasal transcripts of both ERG1 isoforms (ERG1a and ERG1b), ERG3, and MiRP2 are detected by RT-PCR and sequencing. Specificity is confirmed by positive [cerebellum; heart (ERG1a, ERG1b, MiRP2)] and negative [heart (ERG3); LNCaP cells] controls. No PCR product is amplified when reverse transcription is omitted (control). d, e, Representative immunoblotting experiments illustrating the regulation of ERG1 and MiRP2 protein expression. Individual blot panels (d or e) correspond to the same SDS-polyacrylamide gel, respectively. d, Western blot analysis demonstrating increased levels of vomeronasal ERG1a/1b expression in stimulated mice versus deprived mice. Blots (n = 4) were additionally probed for β-tubulin and GAPDH as loading controls. e, Western blots (n = 3) probed with MiRP2 antibody reveal downregulated protein expression in stimulated animals.

Next, we aimed to examine activity-dependent changes in ERG1a, ERG1b, and MiRP2 expression on the protein level. In ERG1 immunoblot analyses, the antibody detects two bands at ∼165 and ∼100 kDa in VNO protein extracts (supplemental Fig. 1a, available at www.jneurosci.org as supplemental material), corresponding to N-glycosylated ERG1a and ERG1b (Pond et al., 2000; Hirdes et al., 2005). Equivalent bands were not detected in eye or LNCaP protein extracts. Probing blots of VNO and brain protein with an anti-MiRP2 antibody revealed bands of the expected size (∼20 kDa) (McCrossan et al., 2003) in both tissues (supplemental Fig. 1b, available at www.jneurosci.org as supplemental material). An anti-ERG3 antibody detected a single band of ∼105 kDa in blots of both cerebellum and VNO (supplemental Fig. 1c, available at www.jneurosci.org as supplemental material). An anti-ERG2 antibody, however, stained multiple bands in the expected 100–110 kDa range in the cerebellum but not the VNO (supplemental Fig. 1d, available at www.jneurosci.org as supplemental material), suggesting no or only marginal translation of ERG2 transcripts in vomeronasal tissue. Using the exact same stimulation/deprivation protocol as described above (Fig. 1a), pooled VNO protein from stimulated (n = 10) and deprived (n = 12) mice, respectively, was subjected to immunoblot analysis (Fig. 1d,e). β-Tubulin and GAPDH were found at constant levels in microarray experiments and thus served as loading controls. All antibodies detected distinct bands of the expected molecular size. Comparing stimulated and deprived mice, however, revealed substantial differences in band intensity for ERG1a, ERG1b (n = 4), and MiRP2 (n = 3). For stimulus-exposed animals, prominent bands were detected for ERG1a and ERG1b (Fig. 1d), whereas labeling intensity was much weaker in samples from deprived mice. An opposite regulation was observed for MiRP2 (Fig. 1e). These results were confirmed by semiquantitative densitometry measurements of scanned immunoblot band intensities (supplemental Fig. 1e). Thus, the activity-dependent regulation of ERG and MiRP2 K⁺ channel transcripts we identified by microarray profiling is effectively translated into upregulated (ERG1 isoforms) and downregulated (MiRP2) protein levels in the adult mouse VNO.

Coexpression of ERG1, ERG3, and MiRP2 in basal VSNs

To analyze the spatial distribution of vomeronasal ERG and MiRP2 channel subunits, we performed immunocytochemistry in coronal VNO sections from both C57BL/6 and transgenic OMP-GFP mice. In the gene-targeted animals, all VSNs express GFP under the control of the omp promoter (Potter et al., 2001). Double-labeling experiments using antibodies against ERG1–3 or MiRP2 together with layer-specific markers of apical neurons [phosphodiesterase 4a (PDE4a) (Lau and Cherry, 2000)] or basal VSNs [V2R2 (Martini et al., 2001)] (supplemental Fig. 2a, available at www.jneurosci.org as supplemental material) revealed coexpression of ERG1, ERG3, and MiRP2 in basal neurons (Fig. 2a,b; supplemental Fig. 2c, available at www.jneurosci.org as supplemental material). To control for nonspecific staining, we performed immunocytochemical experiments in which primary antibodies were preadsorbed with their corresponding antigenic peptides (supplemental Fig. 2d,e, available at www.jneurosci.org as supplemental material). In addition, we routinely performed parallel experiments in which primary antibodies were omitted (supplemental Fig. 3d, available at www.jneurosci.org as supplemental material).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Coexpression of ERG1 and ERG3 in basal VSNs. a, Confocal images of immunostainings in VNO cryosections from OMP-GFP mice (GFP fluorescence; gray scale; 1) show costaining of ERG1 (red; 2) and V2R2 (green; 3). High-magnification images (right) reveal overlapping staining in the VSN somata and dendritic tips. b, Representative high-magnification confocal images of the VNO sensory epithelium in a coronal section (25 μm) from an OMP-GFP mouse. GFP fluorescence in mature VSNs is depicted in gray scale (1), ERG1-immunopositive cells are shown in green (2), and ERG3-positive VSNs are shown in red (3). Merged images clearly reveal coexpression of ERG1 and ERG3 in basal VSN somata. In contrast to ERG1 immunoreactivity, ERG3 staining is not observed in the microvillous layer. c, Confocal images of ERG2 immunostainings in VNO sections from OMP-GFP mice (GFP fluorescence, gray scale, 1; anti-ERG2 signal is shown in red, 2). Merged photomicrographs reveal essentially no ERG2-positive VSNs. BV, Blood vessel; L, lumen; ML, microvillous layer; SCL, sustentacular cell layer.

In sections from OMP-GFP mice, merged images of anti-ERG1 and anti-V2R2 labeling show overlapping staining in both the somata and dendritic tips of basal VSNs (Fig. 2a). In contrast, a spatially restricted pattern was observed by anti-ERG1/anti-PDE4a double labeling (supplemental Fig. 2b, available at www.jneurosci.org as supplemental material). PDE4a staining was confined to apical cells, whereas essentially nonoverlapping ERG1 labeling was detected in basal neurons. The same results were obtained using another ERG1 antibody (Santa Cruz Biotechnology; data not shown). Costainings of ERG1 and MiRP2 (supplemental Fig. 2c, available at www.jneurosci.org as supplemental material) as well as ERG1 and ERG3 (Fig. 2b) indicate coexpression of all three subunits. Immunostainings against ERG2, however, showed no VSN-specific labeling (Fig. 2c). As positive controls, we immunostained sagittal sections of the cerebellum in which expression of all ERG subunits had previously been reported (Papa et al., 2003; Guasti et al., 2005). Immunoreactivity for ERG1, ERG2, and ERG3 was detected in Purkinje neurons (supplemental Fig. 3a–c, available at www.jneurosci.org as supplemental material). Consistent with semiquantitative immunoblotting results (Fig. 1d), when parallel ERG1 immunostainings were performed on VNO cryosections from stimulated (n = 3) and deprived (n = 3) mice under the exact same conditions, an increase in both fluorescence intensity and the amount of labeled neurons became apparent (supplemental Fig. 3e, available at www.jneurosci.org as supplemental material).

The subcellular distribution of ERG1 and MiRP2 was assessed in dissociated VSNs from OMP-GFP mice (supplemental Fig. 4a,b, available at www.jneurosci.org as supplemental material). In immunopositive neurons (n = 5), ERG1 reactivity was evident in the entire cell body. In contrast, MiRP2 staining was most prominent in the soma and dendrite (n = 4) but absent in the knob and microvilli. This pattern confirmed results obtained in whole-organ sections (supplemental Fig. 2c, available at www.jneurosci.org as supplemental material). Together, these data show that both ERG1 isoforms (1a and 1b) as well as ERG3 and MiRP2 are coexpressed in basal neurons of the mouse VNO.

Biophysical “fingerprint” of vomeronasal ERG channels: current activation

We next asked whether ERG channel expression is reflected by voltage-dependent K⁺ currents (I_K(ERG)). Therefore, we obtained whole-cell recordings from optically identified neurons in acute VNO slices (Kelliher et al., 2006). After break-in, physiological saline was replaced by S₃. Exchange was monitored by the gradual block of voltage-gated Na⁺ inward currents in response to step depolarizations [supplemental Fig. 4c, available at www.jneurosci.org as supplemental material; note the increase in tail current amplitude (arrowhead)]. In all recordings, step protocols were applied before and after incubation with highly specific ERG inhibitors: either the methanesulfonanilide E-4031 (10 μm) (Spector et al., 1996) or the scorpion toxin ErgTx-1 (100 nm) (Gurrola et al., 1999). Digital subtraction of currents recorded in the presence of either drug from control recordings revealed the ERG-mediated component.

To investigate the voltage dependence of quasi steady-state activation, a 750 ms prepulse to variable potentials (−70 to 60 mV) was followed by a step to −120 mV (100 ms). After hyperpolarization, rapid recovery from inactivation and slower deactivation induced an E-4031-sensitive tail current in 16 of 27 VSNs (59%) (Fig. 3a). During depolarization, we observed a substantial E-4031-sensitive steady-state outward current, likely attributable to activation of ERG3. This subunit shows fast activation kinetics and slow steady-state inactivation (Shi et al., 1997; McKay and Huizinga, 2006), resulting in weak inward rectification (Wimmers et al., 2002). Eleven of 27 neurons showed no drug-sensitive current (Fig. 3b). To ensure that diminished tail currents did not result from current rundown, we performed controls in the absence of inhibitors (supplemental Fig. 4d, available at www.jneurosci.org as supplemental material). Up to 10 min, essentially no rundown was observed.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Activation of ERG-mediated K⁺ currents in vomeronasal neurons. a, b, Representative whole-cell patch-clamp recording traces. Tail currents were elicited in the absence (control) and presence of E-4031 (10 μm) by hyperpolarization to −120 mV (inset, pulse protocol). ERG-mediated currents (n = 16) are computed as drug-sensitive components after digital off-line subtraction (control − E-4031). Eleven neurons showed no drug-sensitive currents (b). c, d, Quasi steady-state activation curves of both E-4031-sensitive (filled circles; n = 16) and ErgTx-1-sensitive (100 nm; filled triangles; n = 14) currents. Curves illustrate average current densities (c) or normalized current amplitudes (d; I/I_max) versus prepulse potential. Data points are fitted by the Boltzmann equation (see Results). With one exception (c; −40 mV), no significant differences in current isolation were observed between inhibitors. e–g, Time course of vomeronasal ERG current activation (inset, “envelope of tail currents” pulse protocol). e, Representative original recording after digital current subtraction (control − E-4031). On expanded coordinates (dotted square area), note the increase in current amplitude during prolonged depolarization. f, Average tail current densities increased with prepulse duration (monoexponential fit extrapolated to 0 pA/pF: τ_act = 194 ms; n = 13). g, On a logarithmic time scale, normalized tail currents (I/I_max) increase in a linear manner (half-maximal activation at a prepulse duration of 25.8 ms).

When peak tail current density is plotted as a function of prepulse depolarization (Fig. 3c), a similar voltage dependence is seen for E-4031-sensitive VSNs (n = 16 of 27) and ErgTx-1-sensitive neurons (n = 14 of 18). Normalization of tail currents (Fig. 3d) provides a reasonable estimate for the activated channel fraction (open or inactivated state) before repolarization. Using the Boltzmann equation to fit the activation curves, we obtained half-maximal activation values (V_1/2) of −13.8 mV (E-4031) and −7.1 mV (ErgTx-1) and slope factors k = 17 (E-4031) and k = 16 (ErgTx-1), respectively. These data show that, although channels begin to activate at subthreshold potentials, a considerable membrane depolarization is required to evoke substantial channel activation.

In addition to voltage-dependent gating characteristics, ion channel function is determined by gating kinetics. To investigate ERG channel activation kinetics in VSNs, we applied an envelope of tail current protocol (Schonherr et al., 1999; Sturm et al., 2005). Depolarizing prepulses (40 mV) of increasing duration (5–2310 ms; increment factor, 1.55) were followed by hyperpolarization to −120 mV (Fig. 3e). When current densities were plotted as a function of prepulse duration (Fig. 3f,g), I_K(ERG) amplitude increased exponentially with a time constant τ_act = 194 ms (n = 13). Similar values were reported for recombinant rERG channels [τ_act = 95 ms (rERG3) to 215/123 ms (rERG1a/rERG1b)] (Hirdes et al., 2005). However, we observed a significant current (−37.3 ± 11.2 pA/pF) even at 5 ms depolarizations (Fig. 3f), corresponding to 34.4 ± 5.3% of open channels [current normalization (I/I_max) on a semilogarithmic scale] (Fig. 3g). Half-maximal activation was observed after 25.8 ms. These results indicate that, provided a substantial membrane depolarization, a considerable fraction of vomeronasal ERG channels displays a relatively rapid time course of activation.

Biophysical fingerprint of ERG channels: current availability and deactivation

The combination of fast recovery from inactivation and relatively slow deactivation is a characteristic feature of ERG channels (Schwarz and Bauer, 2004) underlying their role in membrane repolarization and modulation of excitability. Therefore, we next studied vomeronasal ERG current availability after repolarization. VSNs were depolarized (40 mV) for 500 ms to fully activate/inactivate ERG channels. After repolarization to variable potentials (−10 to −120 mV, 100 ms), drug-sensitive tail current amplitudes and deactivation kinetics are clearly voltage dependent (Fig. 4a). Forty-nine percent (25 of 51; E-4031) and 53% (8 of 15; ErgTx-1) of all neurons, respectively, exhibited ERG-mediated currents (Fig. 4a), whereas no currents were detected in the remaining 26 (E-4031) and 7 (ErgTx-1) VSNs (Fig. 4b). I_K(ERG) quasi steady-state inactivation curves were derived from peak current densities and fitted by the Boltzmann equation. No significant amplitude differences (one exception at −30 mV), similar V_1/2 values for half-maximal recovery from inactivation (E-4031, −62.3 mV; ErgTx-1, −63.8 mV), and slope factors (E-4031, k = 19; ErgTx-1, k = 17) illustrate the effective isolation of vomeronasal ERG currents using either blocker (Fig. 4c,d). Using S₃, the calculated Nernst equilibrium potential for K⁺ is −32.5 mV. Individual normalized I_K(ERG) availability curves reversed, on average, at −30.5 ± 1.5 mV (ErgTx-1; n = 6) and −26.7 ± 1.3 mV (E-4031; n = 10; p = 0.078), respectively, confirming the K⁺ selectivity. Notably, at potentials of −60 mV or less, I_K(ERG) deactivation was not complete after 100 ms (Fig. 4a, expanded traces). When plotting average amplitudes, at 95 ± 5 ms after repolarization, versus membrane potential (Fig. 4e), this “steady-state” current becomes apparent (E-4031, n = 13; ErgTx-1, n = 8).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Voltage dependence of ERG channel deactivation. a–d, ERG current availability depends on the degree of repolarization. a, Representative original tail currents recorded during repolarization steps to variable membrane potentials (−10 to −120 mV; inset, pulse protocol). Recordings are shown under control conditions, in the presence of E-4031 (10 μm) and after digital off-line subtraction (control − E-4031). On expanded coordinates (dotted square area; left), the voltage dependence of current availability becomes obvious (n = 25). b, In 26 VSNs, essentially no drug-sensitive currents could be isolated. c, d, Quasi steady-state I_K(ERG) inactivation curves derived from peak current densities (c) or normalized current amplitudes (d; I/I_max) and plotted as a function of the repolarization voltage [filled circles, E-4031, n = 25; filled triangles, ErgTx-1 (100 nm), n = 8]. Data points are fitted by the Boltzmann equation. With one exception (c; −30 mV), no significant differences are observed between curves derived using either inhibitor. e, Steady-state current density after 95 ms of repolarization (E-4031, n = 13; ErgTx-1, n = 8). At membrane potentials of −60 mV or less, a nondeactivating or slowly deactivating current is recorded (a; expanded dotted square area; right). Trend lines were extrapolated using a Gauss function. f, g, Voltage dependence of deactivation kinetics (filled circles, E-4031, n = 16; filled triangles, ErgTx-1, n = 5). τ_deact was determined by single-exponential fits during a fixed time period (2–25 ms after repolarization). On a semilogarithmic scale (f), τ_deact versus repolarization potential shows a linear relationship. When τ_deact values are normalized to data obtained at −60 mV (g), no differences in deactivation kinetics are observed using either ERG inhibitor.

On expanded coordinates (Fig. 4a), the voltage dependence of deactivation becomes evident. Tail current decay was fitted by a single exponential function (gray dashed line) to determine deactivation time constants (τ_deact). Currents elicited by strong hyperpolarization (Fig. 4a, −120 mV trace) deactivated faster than currents evoked by step repolarization to less negative potentials (Fig. 4a, −90 mV trace). This characteristic of ERG channels was quantified by plotting τ_deact versus the repolarization potential (Fig. 4f,g). On a semilogarithmic scale, the voltage dependence of τ_deact was linear (Fig. 4f). Normalization to values recorded at −60 mV (E-4031: 21.5 ± 2.6 ms, n = 16; ErgTx-1: 20.3 ± 5 ms, n = 5) reveals no inhibitor-related differences (Fig. 4g). Together, substantial ERG-mediated K⁺ current, consisting of both a deactivating tail and a steady-state current component, is available in basal VSNs during fast repolarization to subthreshold membrane potentials.

The ERG1 channel subunit is a major contributor to vomeronasal I_K(ERG)

K⁺ channel profiling and semiquantitative immunochemistry revealed activity-dependent upregulation of ERG1 expression (Fig. 1). Therefore, we aimed to dissect the role of this specific subunit in generation of vomeronasal ERG currents using a pharmacological approach. In contrast to the less selective agents E-4031 and ErgTx-1, the recombinant peptide rBeKm-1 completely blocks ERG1 at nanomolar concentrations without affecting other channels including ERG3 (Korolkova et al., 2001; Restano-Cassulini et al., 2006). We, thus, repeated various experiments shown in Figures 3 and 4 using rBeKm-1 (100 nm) to study activation, recovery from inactivation, and deactivation of isolated ERG1-mediated currents (Fig. 5). Eight of 15 VSNs (53%) showed rBeKm-1-sensitive currents. Quasi steady-state activation was compared with data obtained when blocking all subunits using E-4031 (Fig. 5a). We found significantly smaller rBeKm-1-sensitive current amplitudes at positive potentials. Activation curves showed similar V_1/2 values for currents isolated by rBeKm-1 (−18.6 mV) and E-4031 (−13.8 mV). However, the slope of the rBeKm-1 curve (k = 23.8) was considerably more shallow than for currents isolated by E-4031 (k = 17). In experiments addressing current availability after repolarization (Fig. 5b), significantly decreased amplitudes are observed in rBeKm-1-treated VSNs after repolarization to membrane potentials of −80 mV or less. Moreover, half-maximal recovery from inactivation and slope factors were different (E-4031: −60.3 mV, k = 19; rBeKm-1: −71.5 mV, k = 25). Notably, no differences in the nondeactivation current component were found using either treatment (Fig. 5c), suggesting that the steady-state current is predominantly, if not exclusively, mediated by homomeric or heteromeric ERG1 channels. In summary, these data indicate that the ERG1 subunit(s) contributes substantially to the total ERG-mediated current. However, a small but significant fraction is carried by rBeKm-1-insensitive subunits.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Substantial contribution of the ERG1 channel subunit to vomeronasal I_K(ERG). a, Quasi steady-state activation of vomeronasal ERG currents (see Fig. 3c) [filled circles, E-4031 (10 μm), n = 16; filled squares, rBeKm-1 (100 nm), n = 8]. Current densities are shown as a function of prepulse potential and fitted by the Boltzmann equation. At depolarization to positive voltages (10–40 mV and 60 mV), rBeKm-1-sensitive amplitudes are significantly smaller than currents isolated using the nonsubunit-selective agent E-4031 (p < 0.05). b, Significant differences in channel availability after repolarization (filled squares, rBeKm-1, n = 8; filled circles, E-4031, n = 25). Inactivation curves show significantly decreased peak densities for ERG1-specific currents at potentials of −80 mV or less (p < 0.05). c, Similar steady-state current characteristics are detected using either pharmacological treatment. Trend lines are calculated using a Gauss function.

I_K(ERG) is selectively expressed in basal VSNs

Immunochemistry revealed that ERG subunits are exclusively found in basal neurons (Fig. 2; supplemental Figs. 2b–e, 3e, available at www.jneurosci.org as supplemental material). If so, ERG-mediated currents should only be expressed in ∼50% of all VSNs. Indeed, we observed I_K(ERG) in 53% (41 of 78; E-4031), 67% (22 of 33; ErgTx-1), and 53% (8 of 15; rBeKm-1) of investigated neurons. To consolidate these findings, we labeled several VSNs (n = 11) during voltage-clamp analysis by adding Alexa488 hydrazide (20 μm) to the pipette solution. This allowed on-line in situ visualization of VSN morphology (Fig. 6a, left and middle), post hoc immunocytochemical analysis (Fig. 6b–f), and 3-D reconstruction (Fig. 6a, right (n = 7) (supplemental Movie 1, available at www.jneurosci.org as supplemental material). All labeled neurons exhibited an intact cytoarchitecture. Given the rather diffuse zonal boundary in the VNO (Fig. 2; supplemental Fig. 2, available at www.jneurosci.org as supplemental material) (Martini et al., 2001; Leinders-Zufall et al., 2004), it was, however, impossible to infer basal or apical identity of a given neuron simply from in situ visualization. We therefore determined cell identity by post hoc immunocytochemistry using anti-V2R2 (Fig. 6b,d), anti-PDE4a (Fig. 6c,e), and anti-ERG1 (Fig. 6f) antibodies. All labeled VSNs that showed ERG-mediated currents (n = 8) were immunopositive for either ERG1 (n = 4) or V2R2 (n = 2) or showed no PDE4a staining (n = 2), respectively (Fig. 6b,c). In contrast, cells that exhibited no current (n = 3) were either negative for V2R2 (Fig. 6d) or ERG1 (Fig. 6f) or strongly stained by the anti-PDE4a (Fig. 6e) antibody. ERG-mediated currents are thus unique to basal V2R-expressing neurons.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Sensory activity controls ERG K⁺ channel expression and function in basal VSNs. a, During whole-cell experiments, cells were fluorescently labeled via diffusion of Alexa488 for in situ visualization, post hoc immunocytochemical analysis, and 3-D reconstruction of VSN morphology. Left, Overlay of two photomicrographs of an acute VNO slice taken under infrared-DIC illumination and epifluorescence, respectively. The soma (black arrowhead) of the VSN under investigation is visualized by dye diffusion from the pipette. Middle, Low-magnification bright-field image of a VNO slice after recording. Under epifluorescence, the recorded neuron is readily identified (inset). Right, 3-D reconstruction of the same fluorescently labeled VSN shown before (middle; inset). An axial transparency projection is generated from confocal z-stacks (supplemental Movie 1, available at www.jneurosci.org as supplemental material). b–f, Layer-specific classification of Alexa488-labeled VSNs by post hoc immunocytochemistry. V2R2, PDE4a, or ERG1 counterstaining shows that VSNs that exhibited ERG currents in previous voltage-clamp recordings (n = 8) are V2R2 positive (b) or PDE4a negative (c), respectively, whereas neurons that showed no ERG-dependent currents are either negative for V2R2 (d) or ERG1 (f) or PDE4a positive (e). g–j, ERG currents are reduced in VSNs of stimulus-deprived mice. g, i, Activation curves of either E-4031-sensitive (g) or rBeKm-1-sensitive (i) ERG currents from stimulus-deprived (filled diamonds; g, n = 9; i, n = 11) and control (g, filled circles, n = 16; i, filled squares, n = 8) animals. At depolarizations of −40 mV or more (g) or −10 mV or more (i), respectively, the reduction is highly significant (p ≤ 0.005). h, j, I_K(ERG) availability curve for stimulus-deprived (filled diamonds; h, n = 8; j, n = 7) and control (h, filled circles, n = 25; j, filled squares, n = 8) mice. Current densities are significantly smaller after repolarization to voltages of −60 mV or less (h) or −30 mV or less (j), respectively (p ≤ 0.05). a, Axon; BV, blood vessel; d, dendrite; k, knob; L, lumen; ML, microvillous layer; p, patch pipette; s, soma; SCL, sustentacular cell layer; SE, sensory epithelium.

Regulated transcription is effectively translated into changed protein functionality

We next asked whether the reduced ERG1 channel expression in stimulus-deprived mice (Fig. 1) is reflected in VSN electrophysiology. To address this question, we recorded ERG-mediated currents in optically identified basal VSNs (Kelliher et al., 2006) from stimulus-deprived animals (n = 9). Using the same pulse protocols as shown in Figures 3a and 4a, we recorded both E-4031- and rBeKm-1-sensitive tail currents as a function of either prepulse (Fig. 6g,i) or repolarization (Fig. 6h,j) potential, respectively. Indeed, using either E-4031 to isolate total ERG currents or rBeKm-1 to specifically isolate the ERG1-dependent component, current densities were significantly reduced in VSNs from deprived mice. When data from deprived animals (E-4031) were fitted by the Boltzmann equation, the slopes of both the I_K(ERG) activation (k = 20) and the availability curve (k = 27) were considerably shallower than observed for control neurons (k = 17/19) (Fig. 6g,h), indicating not only a reduced density of functional channels but also a relative change in subunit composition. Notably, when ERG1-specific currents were isolated from VSNs of control versus deprived mice using rBeKm-1, we observed significantly reduced current densities in deprived mice (Fig. 6i,j), indicating that the substantial reduction in total I_K(ERG) in VSNs from stimulus-deprived mice is, at least in part, attributable to decreased ERG1 expression. This conclusion is supported by the fact that we never observed a nondeactivating current in deprived mice. Because rBeKm-1-sensitive steady-state currents were almost indistinguishable from controls (Fig. 5c), this lack of a steady-state current supports the concept of ERG1 being a critical subunit for incomplete deactivation at hyperpolarized voltages.

I_K(ERG) contributes to the late phase of AP repolarization in VSNs

Several biophysical characteristics of vomeronasal ERG currents suggest a potential function in AP repolarization. First, considerable membrane depolarization (e.g., during AP overshoot) is required to evoke substantial activation (Fig. 3a,d). Second, significant currents are observed even at preceding depolarizations as short as 5 ms (Fig. 3f). Third, little inactivation occurs during depolarization (Fig. 3e). Do ERG channels thus contribute to spike repolarization in VSNs? To clarify this question, we recorded E-4031-sensitive currents using the AP clamp technique. Prerecorded AP trains elicited by positive current injection into VSNs were applied as the command waveforms (black traces). This way, vomeronasal ERG channels experience the same trajectory of voltage as during naturally occurring APs (Bean, 2007). Figure 7a shows a drug-sensitive current that is rapidly activated and shifted from outward to inward during the repolarizing phase (n = 24). On expanded coordinates (Fig. 7a, right), this current reversed close to the calculated Nernst K⁺ equilibrium potential of −32.5 mV (red dotted lines), reached a peak close to the point of maximal repolarization (−70 mV; black dotted lines), and decayed exponentially with a mean time constant of τ_deact = 17.9 ± 4.3 ms.

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

Activation of I_K(ERG) during VSN AP repolarization. a, When prerecorded VSN spike trains are applied as the command voltage (AP clamp; black waveforms), E-4031-sensitive currents are observed after digital subtraction (red trace; n = 24). A single representative event (dotted square) is shown at higher resolution (right). During AP discharge, an outward current is activated that reverses direction during repolarization, reaches a maximum at −70 mV (black dotted lines), and decays exponentially. The observed reversal potential is close to the calculated Nernst K⁺ equilibrium potential of −32.5 mV (red dotted lines). b, Representative AP clamp recording (left; red trace) at a command voltage spike train displaying frequency adaptation (black). Quantitative analysis (right) of ERG current amplitudes versus interspike intervals is shown. Total I_K(ERG) is isolated from control VSNs using E-4031 (10 μm; n = 14; black circles, solid line), whereas ERG1-dependent currents are isolated by rBeKm-1 [100 nm; black squares, dashed line (control, n = 14); red diamonds, solid line (deprived, n = 9)]. Current densities decrease exponentially with prolonged interspike intervals. In VSNs from deprived animals, I_K(ERG) is significantly reduced (p ≤ 0.05).

We next asked whether the duration of interspike episodes affects the ERG current amplitude. Frequency-dependent ERG current accumulation has previously been observed (Chiesa et al., 1997; Schonherr et al., 1999). When a train of VSN APs, showing spike frequency adaptation during constant current injection, was used as a stimulus template, ERG amplitudes decreased with prolonged interspike intervals (Fig. 7b) (control/E-4031, n = 14; control/rBeKm-1, n = 14; deprived/rBeKm-1, n = 9). At discharge intervals longer than ∼150 ms, however, peak current densities remain relatively stable, indicating that at frequencies ≥7 Hz incomplete ERG channel deactivation between spikes could lead to current accumulation. Moreover, AP-driven ERG currents are strongly ERG1 dependent (Fig. 7b, right). Total I_K(ERG) amplitudes (E-4031) and the ERG1-specific current component (rBeKm-1) are not statistically different, although the latter tends to be consistently smaller. In VSNs from deprived animals, however, ERG1-dependent current flow during AP discharge is strongly reduced (red graph). In summary, these data show that considerable ERG1-dependent currents flow during the late repolarizing phase of vomeronasal AP discharge.

I_K(ERG) extends the dynamic range of the vomeronasal stimulus–responsefunction

What is the physiological relevance of ERG channel activation during AP firing? We addressed this question by recording VSN input–output relationships in current-clamp configuration (supplemental material, available at www.jneurosci.org) under physiological ionic conditions. The majority of VSNs were spontaneously active at rest. We therefore injected a constant hyperpolarizing current (−0.5 to −3 pA) to keep V_M as close as possible to −75 mV, thereby abolishing spontaneous firing. First, we investigated VSN output in response to a naturally occurring receptor current profile. We designed a current injection protocol that mimicked a basal VSN response that we recorded after application of diluted urine (Fig. 8a, gray box, top). This protocol was applied for five cycles of increasing duration (Fig. 8a, gray box, bottom, a–e). VSN firing patterns were analyzed in the absence and presence of E-4031 (10 μm). Six of 13 VSNs showed a dramatic effect on AP firing. In these neurons, the number of spikes was significantly reduced after drug incubation (Fig. 8b). Notably, this effect was most pronounced during the initial phase of high-amplitude current injection (25 − 7 pA). Here, we frequently observed that VSNs ceased firing after one or a few initial spikes, but returned to a tonic discharge pattern during small-amplitude current injection (7 − 2 pA) (Fig. 8a, traces d and e). Analog results were obtained from analysis of firing patterns in deprived mice (Fig. 8a,b, red) (n = 36). These data suggest that, by ensuring a sufficient level of membrane repolarization, vomeronasal ERG currents could play a critical role during periods of strong sensory activation.

• Download figure
• Open in new tab
• Download powerpoint

Figure 8.

ERG channels determine the output range of basal vomeronasal neurons. a, Representative AP discharge patterns induced by current injections. The injection protocol (gray box) mimics a prerecorded VSN receptor current profile in response to focal application of diluted urine (1:100; red trace). High-frequency firing is impaired in response to current injections both after ERG channel inhibition and VSNs of stimulus-deprived animals. Lowercase letters indicate corresponding injection protocols. b, Mean number of spikes versus current injection length (a) in control VSNs (n = 6; black circles, solid line), after ERG channel inhibition (n = 6; black diamonds, dashed line), and in VSNs from deprived animals (n = 36; red diamonds, solid line). Drug incubation and deprivation significantly reduce the total spike count at durations ≥500 ms (p ≤ 0.05). c, Typical firing responses induced by constant 1 s depolarizing current steps of variable amplitude (4–40 pA; 4 pA increments). Note the continuous decline in AP amplitude (deprivation/ERG channel inhibition) as a putative result from reduced membrane repolarization and, consequentially, insufficient recovery from Na⁺ channel inactivation. d, Mean firing frequency versus current amplitude (f-I curve). For amplitudes ≥4 pA (deprived, n = 30) or ≥8 pA (E-4031, n = 11), respectively, the frequency is significantly reduced (p < 0.005). Under control conditions, firing rates approximate a maximum frequency of ∼25 Hz. After deprivation or ERG channel inhibition, the f-I curve shows a more bell-shaped appearance.

If the above hypothesis is true, the effect should become more pronounced with increased spike frequencies. Indeed, when the firing rate is analyzed as a function of current injection amplitude (4–40 pA; 1 s duration), the mean spike frequencies are significantly reduced in the presence of E-4031 at amplitudes ≥8 pA (Fig. 8c,d) (n = 11 of 15 VSNs). This effect of pharmacological ERG channel inhibition mimics the impaired AP firing phenotype observed in VSNs from stimulus-deprived animals (Fig. 8c,d, red) (n = 30). However, maintenance of a constant resting potential (supplemental Fig. 4e, available at www.jneurosci.org as supplemental material) (n = 12) and spontaneous activity (supplemental Fig. 4f, available at www.jneurosci.org as supplemental material) (n = 3) were essentially unaffected by ERG channel inhibition. We thus conclude that basal VSNs express I_K(ERG) to guarantee an appropriate input–output relationship at intermediate to high levels of sensory stimulation.

To substantiate this concept, we analyzed changes in the “anatomy” of single spikes both during ERG channel inhibition (supplemental Fig. 5a–c, available at www.jneurosci.org as supplemental material) and in stimulated compared with deprived mice (supplemental Fig. 5d–f, available at www.jneurosci.org as supplemental material). Single APs were elicited by repeated injection of suprathreshold current (10 pA; 45 ms; 50 cycles at 5 ms intervals). Supplemental Figure 5a (available at www.jneurosci.org as supplemental material) shows representative examples of 50 normalized and averaged spikes of the same neuron in the absence (black trace) or presence (red trace) of E-4031, respectively. Here, spike width is increased from 5.05 ms (control) to 5.85 ms (E-4031). Eleven VSNs were examined for both half-width (supplemental Fig. 5c, available at www.jneurosci.org as supplemental material) and maximal membrane repolarization (supplemental Fig. 5b, available at www.jneurosci.org as supplemental material) during current injection. Because the same cells were also used for experiments shown in Figure 8, c and d, we were able to categorize these VSNs as either drug sensitive (n = 7; red data points) or insensitive (n = 4; black data points). E-4031-sensitive neurons show a significant shift of the maximally repolarized membrane potential toward more depolarized values (control, −67.1 ± 1.4 mV; E-4031, −58.8 ± 1.3 mV; p = 0.008) and a significant increase in half-width (control, 4.8 ± 0.18 ms; E-4031, 5.66 ± 0.44 ms; p < 0.005)). A similar broadening of spike width during ERG channel inhibition has been shown for immature Purkinje neurons (from 5.1 to 6.3 ms) (Sacco et al., 2003). Again, when comparing single VSN spikes from stimulated (n = 13) versus deprived (n = 21) mice, the deprivation phenotype (supplemental Fig. 5d–f, available at www.jneurosci.org as supplemental material) matches the effects of pharmacological treatment observed in control VSNs. Together, these data strongly suggest that vomeronasal ERG channels are important determinants of AP shape and VSN set-point excitability.