D-type K⁺ current rules the function of electrically coupled neurons in a species-specific fashion

Electrical synaptic transmission is a modality of communication mediated by gap junctions, which represent areas of close apposition between the plasmatic membrane of neurons characterized by the presence of clusters of intercellular channels (Bennett, 1997). These junctions establish pathways of low resistance for the flow of ionic currents supporting the bidirectional and fast communication of both depolarizing and hyperpolarizing signals between neurons (Connors and Long, 2004; Bennett and Zukin, 2004; Pereda et al., 2013). These characteristics allow electrical synapses to contribute to relevant functional operations by neural circuits in many brain areas. Among these operations, lateral excitation results from the ability of electrical coupling to spread excitation within neural circuits, thus creating functional compartments. In this way, the activity in some cells promotes the activation of neighboring coupled ones, thus, operating as a network-boosting mechanism. Although this operation might degrade spatial specificity within neural circuits, lateral excitation between primary afferents tuned to qualitatively similar stimuli acts to boost sensory responses involved in the organization of motor outputs as was shown in invertebrates and lower vertebrates (El Manira et al., 1993; Pereda et al., 1995; Herberholz et al., 2002; Antonsen et al., 2005; Curti et al., 2022). In mammals, this mechanism has also been implicated in the enhancement of excitability in circuits of electrically coupled neurons of the olfactory bulb (Christie and Westbrook, 2006), the cerebellum (Vervaeke et al., 2012), and the dorsal cochlear nucleus (Apostolides and Trussell, 2013, 2014). Moreover, in the retina, lateral excitation provides a mechanism for the precise detection of the spatial location of a moving stimulus, disregarding its velocity (Trenholm et al., 2013).

Mesencephalic trigeminal (MesV) neurons are a special class of primary afferents, whose cell bodies, instead of being located in peripheral ganglia, are distributed in the brainstem (Weinberg, 1928). The peripheral branches of these pseudounipolar neurons innervate the spindles of jaw-closing muscles and mechanoreceptors of periodontal ligaments, whereas the central processes supply sensory input to neurons of the trigeminal motor nucleus, the rostral parvocellular reticular formation, and the nucleus supratrigeminalis (Dessem and Taylor, 1989; Liem et al., 1991). On the other hand, MesV neurons receive synaptic input at their somata from the hypothalamus and various brainstem structures (Lazarov, 2002; Verdier et al., 2004), supporting the notion that these afferents not only relay peripheral information but are also an integral part of the central circuit involved in the generation of masticatory patterns (Morquette et al., 2012). Previous work in rats and mice has shown that MesV neurons are electrically coupled by means of large connexin36 (Cx36) containing somato-somatic contacts characterized by one to three clusters of multiple Cx36-puncta (Curti et al., 2012; Nagy and Lynn, 2018). Additionally, rather than extensive as observed in most structures of the mammalian brain (Connors and Long, 2004; Bennett and Zukin, 2004), coupling in this nucleus is restricted to pairs or small clusters of neurons (Baker and Llinás, 1971; Curti et al., 2012). Moreover, the dynamic interaction of electrical coupling with their intrinsic electrophysiological properties supports the strong synchronization of pairs of MesV neurons and allows them to operate as coincidence detectors (Curti et al., 2012; Davoine and Curti, 2019).

A previous study based on dye transfer experiments and electrophysiological recordings showed that the incidence of coupling in the MesV nucleus is about three times higher in mice compared with rats (Curti et al., 2012). This contrasting circuital organization in homologous circuits subserving the same function in highly related species is surprising, raising the possibility that electrical contacts also present functional differences between both species. Particularly, we focused on the ability of spikes in active neurons to recruit inactive electrically coupled ones, as it represents a critical aspect for operations like synchronization and lateral excitation. Strikingly, here we show that despite similar coupling strength, recruitment of coupled neurons in rats is dramatically more efficient than in mice. Such difference does not result from dissimilarities of the gap junctions themselves, but instead from the properties of the non-synaptic membrane of coupled neurons between species. More specifically, by combining electrophysiological and pharmacological evidence with a comparative approach, we show that the observed interspecific difference results from a significantly higher functional expression of the D-type K⁺ current in MesV neurons from mice. These results identify a role for the subthreshold K⁺ currents in determining the efficacy of postsynaptic recruitment at electrical synapses, and hence the mode of operation of networks of coupled neurons. This emphasizes the role of voltage-dependent membrane conductances in circuits of electrically coupled neurons and its possible contribution to the early stages of sensory processing by primary afferents involved in the organization of relevant behaviors.

Sprague-Dawley rats (age: P7–P16) and C57BL mice of either sex (age: P12–P18) were obtained from the university animal facility accredited by the local authorities (CHEA: Comisión Honoraria de Experimentación Animal of Universidad de la República, Uruguay). All animal care and experimental procedures were performed in accordance with national guidelines and laws, with a minimization of the number of animals used. Protocols were approved by the CNEA (Comisión Nacional de Experimentación Animal) and the School of Medicine CEUA (Comisión de Ética en el Uso de Animales, protocols nos. 070153-000128-20 and 070153-000396-17).

Animals were decapitated without anesthesia and the brains were quickly removed. Transverse brainstem slices (180–250 µm thick) were prepared using a vibratome (Leica VT 1000s or DSK DTK-1000) in cold sucrose solution containing 248 mM sucrose, 26 mM NaHCO₃, 10 mM glucose, 2 mM MgSO₄, 2.69 mM KCl, 1.25 mM KH2PO₄, and 1 mM CaCl₂ for rats, and 213 mM sucrose, 26 mM NaHCO₃, 10 mM glucose, 2 mM MgSO₄, 2.69 mM KCl, 1.25 mM NaH₂PO₄, 1 mM CaCl₂, 0.35 mM ascorbic acid, and 0.3 mM pyruvic acid for mice. In both cases, solutions were bubbled with 95% O₂ and 5% CO₂ (pH ∼7.4). The slices were then transferred to an incubation chamber filled with a physiological solution containing 124 mM NaCl, 2.69 mM KCl, 1.25 mM KH₂PO₄, 26 mM NaHCO₃, 10 mM glucose, 2 mM CaCl₂, and 2 mM MgSO₄ bubbled with 95% O₂ and 5% CO₂ (pH ∼7.4) at 34°C for 30 min. Afterward, the slices were kept at room temperature in the physiological solution until they were transferred into the recording chamber. The recording chamber, mounted on an upright microscope stage (E600; Nikon Eclipse), was continuously perfused with physiological solution (1–1.5 ml/min) at room temperature. Whole-cell patch recordings were performed under visual control using infrared differential interference contrast optics (IR-DIC). MesV neurons were identified on the basis of their location, large spherical somata, and characteristic electrophysiological properties in response to both depolarizing and hyperpolarizing current pulses (Curti et al., 2012). Recording pipettes pulled from borosilicate glass (4–8 MΩ) were filled with intracellular solution containing 148 mM K-gluconate, 3 mM MgCl₂, 0.2 mM EGTA, 4 mM Na2-ATP, 0.3 mM Na-GTP, and 10 mM HEPES (pH ∼7.2). The seal resistance between the electrode tip and the cell membrane was higher than 1 GΩ and pipette capacitance was compensated before breaking the seal. Simultaneous recordings from pairs of MesV neurons whose cell bodies lie in close apposition were made using a Multiclamp 700B amplifier (Molecular Devices). Under the current clamp configuration, the voltage drop across the microelectrode resistance was eliminated by means of the bridge balance control of the amplifier. In voltage-clamp configuration, the membrane capacitance and series resistance were compensated (80%) and continuously monitored. Membrane capacitance was obtained from the readout provided by the Multiclamp 700B amplifier during this procedure. Only cells displaying resting membrane potential more negative than −50 mV or spike amplitude above 70 mV for rats or 60 mV for mice were included in this study. Recordings were low-pass filtered at 5 kHz and acquired by means of an analog-to-digital converter connected to a computer, sampled at 10–40 kHz depending on the experiment.

During simultaneous whole-cell recordings of pairs of coupled MesV neurons in current clamp, a series of hyperpolarizing current pulses of different amplitudes (−50 to −450 pA) were applied to one cell, whereas the voltage changes were recorded in the same injected neuron (presynaptic, Vpre) and in the coupled one (postsynaptic, Vpost). Plots of Vpost as a function of Vpre (measured at the peak of hyperpolarizing responses, vertical dashed lines in Fig. 2, A and B) were constructed and CC was estimated from the slope of linear regressions. CC values estimated by this method are reported as directions (two directions per coupled pair).

During simultaneous whole-cell recordings of pairs of coupled MesV neurons in current clamp, a series of hyperpolarizing current pulses of different amplitudes (−50 to −450 pA) were alternatively applied to one or the other cell, whereas the voltage changes were recorded in the same injected neuron (Vm). Plots of Vm as a function of the intensity of injected current were constructed and the Rin was estimated from the slope of linear regressions. This value represents the equivalent resistance of two parallel branches, one corresponding to the non-junctional resistance of the injected neuron and the other to the gap junction plus the non-junctional resistance of the coupled neuron (Bennett, 1966).

The transfer properties between pairs of electrically coupled MesV neurons were determined by injecting frequency-modulated (2–600 Hz or 0.5–100 Hz) sine waves of current into one of the coupled cells (ZAP protocol) while recording the resulting membrane voltage deflections in both cells (see Fig. 4 A, inset). Peak-to-peak intensity (50–300 pA) was adjusted to induce subthreshold voltage deflections. The magnitude of fast Fourier transform (FFT) was calculated for presynaptic and postsynaptic membrane responses, and the frequency–transfer property was determined as the ratio of the postsynaptic FFT magnitude over the presynaptic FFT magnitude. The population frequency–transfer function for both species was determined by averaging single transfers in each recorded direction. Average transfer functions were low-pass filtered by applying a smoothing algorithm and expressed in decibels (dB). The apparent cutoff frequency was determined as the intersection of the slope of the attenuation observed at high frequencies and a horizontal line representing the value observed in DC.

During current clamp recordings, a series of depolarizing current pulses of 200 ms in duration were applied, whose intensities ranged from 50 to 600 pA, in steps of 50 pA. From these recordings, curves of the number of spikes versus current intensity were constructed and the firing gain, defined as the slope of linear regressions forced through the origin, was determined. This parameter reflects the ability of the neuron to produce repetitive discharges as well as its rheobase (minimal current for eliciting firing), representing a valuable indicator of neuronal excitability (Davoine and Curti, 2019).

From these fits, maximum conductance (gmax), half activation voltage (Vhalf), and slope at Vhalf were obtained. To avoid bias by the difference in cell size, gmax was also reported normalized by the cell’s membrane capacitance. The IA and ID half activation voltages were significantly different both in rats (IA Vhalf: −27.9 ± 4.4 mV [SD], n = 17; ID Vhalf: −39.2 ± 3.8 mV [SD], n = 20; P = 1.42 × 10⁻⁹) and mice (IA Vhalf: −28.9 ± 5.2 mV [SD], n = 14; ID Vhalf: −36.3 ± 4.2 mV [SD], n = 13; P = 0.0004), indicating that these currents were successfully separated following the procedure described above.

To isolate the INap, currents through K⁺ channels and HCN channels were reduced by adding a combination of blockers to the extracellular solution (10 mM TEA-Cl, 1 mM 4-AP, and 5 mM CsCl). In this condition (control), a series of step-like voltage commands of 500 ms in duration, from 0 to 70 mV in steps of 5 mV, starting from a holding potential of −70 mV, were simultaneously applied to both recorded neurons. These voltage commands were repeated after the addition of TTX (0.5 μM). Current traces obtained in TTX were subtracted from those obtained in control, and current intensity was measured after 100 ms of the initiation of the voltage commands. To construct steady-state activation curves, current values were transformed to conductance by dividing by the calculated driving force for Na⁺, and the parameters characterizing this process (gmax, Vhalf, and slope at Vhalf) were determined following the same procedure described for the study of K⁺ currents.

As no obvious difference in cellular excitability was observed between coupled and uncoupled neurons from both species, Na⁺ and K⁺ membrane currents were recorded in coupled and uncoupled neurons, and data were pooled. However, in the case of coupled neurons, voltage commands were always simultaneously delivered to both neurons to improve the space clamp.

Data were analyzed using the following software: Axograph X, Igor Pro7 (Wave Metrics), and Python scientific development environment Spyder (libraries: Numpy, Scipy, Axographio, Stfio, Pandas, and Matplotlib). Results were expressed as average value ± standard deviation (SD). The significance of quantitative data was determined by using Student’s t test of Igor Pro7 (Wave Metrics). To avoid problems related to pseudoreplication when comparing data from rats and mice (which typically involved recordings obtained in multiple cells from the same animal), two different approaches were adopted (Eisner, 2021). First, for the large data sets presented in Figs. 1, 2, 3, 4, 5, and 9, additional analysis was performed by averaging data values obtained from the same animal and running standard Student’s t test (Table S1). For smaller data sets presented in Figs. 6 and 7, additional hierarchical analysis was performed by using GraphPad Prism 9, which considers the structure of the data (number of cells per animal), and results are presented in corresponding supplementary figures. All data for the conclusions of this study are reported in the article.

Fig. S1 and Fig. S2 show the results of statistical hierarchical analysis (nested Student’s t test) for data shown in Figs. 6 and 7, respectively. Fig. S3 characterizes the persistent Na⁺ current of MesV neurons in both rats and mice, along with the corresponding statistical hierarchical analysis. Fig. S4 displays the contribution of ID to the RMP and Rin in mouse MesV neurons. Table S1 summarizes results from statistical analysis (Student’s t test) for data sets presented in Figs. 1, 2, 3, 4, 5, and 9, in which data from the same animal was averaged.

Previous work had shown that the incidence of electrical coupling among MesV neurons from rats and mice is dramatically different. Both electrophysiological and tracer coupling experiments revealed that coupling incidence in rats is about 21–23%, whereas in mice it is considerably more prevalent, about 60–63% (Curti et al., 2012). Such findings suggest contrasting principles of network organization in homolog rodent circuits involved in the organization of orofacial behaviors. To determine if such rat–mouse difference in network organization is accompanied by functional dissimilarities, under identical recording conditions, the properties of electrical synaptic transmission between MesV neurons from these species were systematically compared. Most specifically, the ability of presynaptic spikes to recruit postsynaptic neurons was taken as a readout of functional specializations on the basis of two considerations. First, action potentials triggered in response to sensory stimuli at the periphery most probably constitute the main source for electrical coupling potentials at these primary afferents. Second, relevant functional operations of electrical coupling, like synchronization and lateral excitation, rely on the ability of presynaptic neurons to recruit electrically coupled neighbors (Connors and Long, 2004; Pereda et al., 2013; Connors, 2017; Curti et al., 2022). Thus, the efficacy of postsynaptic recruitment was assessed in pairs of electrically coupled MesV neurons from rats and mice. For this, just-suprathreshold depolarizing current pulses were alternatively injected into each neuron of an electrically coupled pair, whereas the membrane potential was simultaneously monitored in both cells (Fig. 1, A–D). From these experiments, the probability of postsynaptic recruitment was estimated separately for each direction (two directions per pair) as the ratio of the number of postsynaptic spikes over the number of presynaptic ones and represented as a percentage (Fig. 1 E). Strikingly, postsynaptic recruitment is significantly more efficient in rats compared with mice, averaging 43.8 ± 48.5% [SD] (n = 70 directions, N = 31 animals) and 2.4 ± 15.4% [SD] (n = 125 directions, N = 53 animals), respectively (P = 1.09 × 10⁻⁹, unpaired, two-tailed t test; see also Table S1). Moreover, this revealed that rat’s MesV neurons drive spiking of its postsynaptic coupled neurons in half of the tested directions (35 out of 70 directions, probability > 0%), consistent with previous work showing strong spiking synchronization in coupled pairs (Curti et al., 2012). In striking contrast, the recruitment of postsynaptic neurons by presynaptic spikes occurs only in 2.4% of tested directions (3 out of 125, probability > 0) in mice (Fig. 1 F), despite the coupling strength of these two populations being matched (see below).

The higher efficacy of presynaptic spikes to drive spiking in postsynaptic coupled MesV neurons from rats compared with mice might result from differences in coupling strength between these two species. To avoid any bias due to such differences, the populations of coupled pairs from rats and mice were matched in terms of their CC. Therefore, coupled pairs from mice with very low CC were not included in this study. Accordingly, the CC determined by a series of hyperpolarizing current pulses (Fig. 2, A and B, see Materials and methods) showed no statistical difference between these two species, averaging 0.44 ± 0.18 [SD] (n = 70 directions, N = 31 animals) and 0.44 ± 0.11 [SD] (n = 128 directions, N = 53 animals) for rats and mice, respectively (P = 0.981, unpaired, two-tailed t test; see also Table S1). Histograms in Fig. 2 C show that CC from mice display a roughly symmetric distribution, while CC from rats is slightly skewed to the right, although this minor difference most probably cannot explain the dramatic dissimilarity in postsynaptic recruitment between these two populations. Consistently, neither of the determinants of the CC, the junctional conductance (Gj), and the Rin of the postsynaptic neuron (Bennett, 1966; Curti and O’Brien, 2016) displayed any statistical difference. The Rin averaged 92.3 ± 25.3 MΩ [SD] (n = 70 cells, N = 31 animals) and 89.9 ± 27.1 MΩ [SD] (n = 128 cells, N = 53 animals) in rats and mice, respectively (P = 0.538, unpaired, two-tailed t test; see also Table S1; Fig. 2 D), whereas the Gj estimated from current clamp recordings (see Materials and methods), averaged 7.54 ± 5.64 nS [SD] (n = 70 directions, N = 31 animals) in rats and 6.64 ± 2.56 nS [SD] (n = 128 cells, N = 53 animals) in mice (P = 0.210, unpaired, two-tailed t test; see also Table S1; Fig. 2 E). Also, in the age range employed in this study, neither CC nor Rin showed any correlation with age as indicated by linear regression analysis (rat: CC vs. age slope = 0.011, R² = 0.022; mouse: CC vs. age slope = 0.0036, R² = 0.0013; rat: Rin vs. age slope = 0.76, R² = 0.0047; mouse: Rin vs. age slope = −1.8, R² = 0.0057). Moreover, it is worth mentioning that the resting membrane potential of MesV neurons from these two species does not show any statistical difference (rat: −55.6 ± 4.6 mV [SD], n = 70 cells, N = 31 animals; mouse: −55.3 ± 4.0 mV [SD], n = 128 cells, N = 50 animals; P = 0.696, unpaired, two-tailed t test; see also Table S1; Fig. 2 F). Thus, the disparity in the efficacy of postsynaptic recruitment in rats versus mice cannot be explained by a difference in the coupling strength or the resting membrane potential between these species.

Although the CC estimated at the peak of hyperpolarizing voltage responses to current pulses is informative about the coupling strength in the passive regime, it might not faithfully reflect the efficacy of spike transmission. On one hand, transmission at electrical synapses typically behaves as a low-pass filter, meaning that high-frequency signals (like action potentials) are considerably more attenuated in comparison to signals with a lower frequency content. This implies that besides the Gj and the Rin of the postsynaptic cell, the waveform of the presynaptic signal (i.e., its frequency content) also represents a critical determinant of the coupling strength (Bennett, 1966; Connors and Long, 2004; Alcamí and Pereda, 2019; Curti et al., 2022). On the other hand, active subthreshold mechanisms can crucially shape postsynaptic responses. For example, it has been well established that the persistent Na⁺ current (INap) is able to significantly increase the efficacy of spike transmission at electrical contacts (Mann-Metzer and Yarom, 1999; Curti and Pereda, 2004; Dugué et al., 2009; Curti et al., 2012).

Thus, to assess the efficacy of spike transmission in rats and mice, spike characteristics were determined first and their impact on the postsynaptic cell was evaluated thereafter. Fig. 3, A and B display representative spikes (top) recorded in MesV neurons from rats and mice, respectively, with their corresponding time derivatives (bottom), illustrating the method employed to measure several spike parameters. While spike duration does not show statistical difference between rats and mice, averaging 0.48 ± 0.14 ms [SD] (n = 69 cells, N = 31 animals) and 0.50 ± 0.15 ms [SD] (n = 117 cells, N = 50 animals), respectively (P = 0.316, unpaired, two-tailed t test; see also Table S1; Fig. 3 C), spike amplitude is significantly larger in rats (88.7 ± 11.9 mV [SD], n = 70 cells, N = 31 animals) compared with mice (68.2 ± 12.3 mV [SD], n = 118 cells, N = 50 animals; P = 1.29 × 10⁻²¹, unpaired, two-tailed t test; see also Table S1; Fig. 3 D). Moreover, spike after-hyperpolarization (AHP) is also of larger amplitude in rats compared with mice (rat: −5.8 ± 2.5 mV [SD], n = 68 cells, N = 31 animals; mouse: −4.9 ± 2.1 mV [SD], n = 118 cells, N = 50 animals; P = 0.019, unpaired, two-tailed t test; however, see also Table S1; Fig. 3 E), most probably due to the stronger activation of repolarizing mechanisms during rat action potentials, whose peak levels attain more positive values compared to mice. Thus, although similar in duration, MesV neurons’ spikes from rats are on average ∼20 mV bigger in amplitude than those from mice. Correspondingly, the maximum value of the spike time derivative is significantly higher in rats versus mice (rat: 310.8 ± 96.6 mV/ms [SD], n = 66 cells, N = 31 animals; mouse: 183.5 ± 89.6 mV/ms [SD], n = 118 cells, N = 50 animals; P = 9.3 × 10⁻¹⁵, unpaired, two-tailed t test; see also Table S1), indicating that spikes from rat MesV neurons present a faster time course. Consistently, spike phase plots (first-time derivative of membrane voltage as a function of the membrane voltage) show that trajectories from mice lie almost completely within those of rats’ MesV neurons (Fig. 3 F). Also, this analysis revealed that the threshold, defined as the value of membrane voltage at which the rate of change reaches 10 mV/ms, is significantly more hyperpolarized in rats compared with mice (rat: −47.5 ± 5.5 mV [SD], n = 66 cells, N = 31 animals; mouse: −40.0 ± 5.3 mV [SD], n = 118 cells, N = 50 animals; P = 4.9 × 10⁻¹⁵, unpaired, two-tailed t test; see also Table S1; Fig. 3 G).

The above result raises the possibility that the higher rate of postsynaptic recruitment in rats might be a consequence of presynaptic spikes larger in amplitude. However, their faster time course, corresponding to larger high-frequency components in the frequency domain, might result in more dramatic attenuation due to the filter properties of electrical synaptic transmission. Consistent with a previous study (Curti et al., 2012), frequency–transfer characteristics determined by ZAP protocols and FFT analysis (see Materials and methods) revealed that despite some degree of frequency preference, electrical transmission between MesV neurons from both species essentially obeys a low-pass filter, with an apparent cutoff frequency of 51 and 40 Hz in rats and mice, respectively (Fig. 4 A). According to these results, predicting the relative efficacy of transmission in these two species is not straightforward as spikes of larger amplitude might result in larger postsynaptic coupling potentials, whereas attenuation of high-frequency components due to low-pass filter properties is expected to have the opposite effect. To directly assess the efficacy of spike transmission in these two species, the spike-related CC (CC Spike) for the population of electrically coupled MesV neurons was compared. For this, the CC Spike was determined as the ratio between the postsynaptic coupling potential (spikelet) amplitude and the amplitude of the presynaptic spike from recordings like those depicted in Fig. 4, B and C. Noteworthy, this analysis revealed that the CC Spike from mice is almost 30% larger than that from rats, averaging 0.094 ± 0.028 [SD] (n = 122 directions, N = 50 animals) and 0.070 ± 0.034 [SD] (n = 57 directions, N = 28 animals), respectively (P = 7.66 × 10⁻⁶, unpaired, two-tailed t test; see also Table S1; Fig. 4 D), indicating that spike transmission is more efficient in mice compared with rats. This difference most probably results from the fact that spikes from rats seem to present larger high-frequency components in comparison to mice (see above). However, despite this difference in CC Spike, spikelet’s amplitudes from these two populations show no statistical difference as they averaged 5.95 ± 2.91 mV [SD] (n = 57 cells, N = 28 animals) in rats and 6.13 ± 1.41 mV [SD] (n = 122 cells, N = 50 animals) in mice (P = 0.645, unpaired, two-tailed t test; see also Table S1; Fig. 4 E). These results indicate that the higher efficiency of spike transmission in mice is compensated by the lower amplitude of presynaptic spikes, resulting in postsynaptic coupling potentials of similar amplitude.

Preceding results show that in both species presynaptic spikes evoke postsynaptic coupling potentials similar in amplitude, which arise from similar resting membrane potential levels, suggesting that differences in postsynaptic recruitment should result from dissimilarities in membrane excitability. To test this possibility, a detailed characterization of firing properties was performed by using experimental protocols consisting of a series of depolarizing current pulses of increasing intensity (50–600 pA) applied from the cells’ RMP (Fig. 5, A and B). From these experiments, plots of the number of spikes as a function of the injected current intensity were constructed (Fig. 5 C). The average behavior of the population of rats and mice can be seen in Fig. 5 D. This graph shows that rat MesV neurons respond with more spikes in almost the entire range of current intensity tested. Consistently, the slope of linear regression to spikes vs. current relationships (firing gain) in rats is significantly higher than in mice, averaging 0.011 ± 0.015 spikes/pA [SD] (range: 0.0004–0.0664, n = 70 cells, N = 31 animals) and 0.003 ± 0.008 spikes/pA [SD] (range: 0.0004–0.0637, n = 99 cells, N = 50 animals), respectively (P = 7.19 × 10⁻⁵, unpaired, two-tailed t test; see also Table S1; Fig. 5 E), whereas the rheobase for the population of MesV neurons is significantly lower in rats than in mice, averaging 205.7 ± 104.4 pA [SD] (n = 70 cells, N = 31 animals) and 396.2 ± 181.4 pA [SD] (n = 128 cells, N = 50 animals), respectively (P = 1.12 × 10⁻¹⁶, unpaired, two-tailed t test; see also Table S1; Fig. 5 F). Noteworthy, in the age range employed in this study, neither firing gain nor rheobase showed any correlation with age as indicated by linear regression analysis (rat: firing gain vs. age slope = 0.001, R² = 0.04; rheobase vs. age slope = −3.7, R² = 0.007; mouse: firing gain vs. age slope = 2.1 × 10⁻⁵, R² = 9.9 × 10⁻⁶; rheobase vs. age slope = 33.3, R² = 0.04). These results clearly indicate that rat MesV neurons present higher excitability, which is consistent with their lower threshold, and suggest that it might underlie the higher efficacy in postsynaptic recruitment.

To gain insights into the membrane mechanism responsible for such difference in cellular excitability between rats and mice, voltage clamp experiments were designed to characterize the main subthreshold K⁺ currents (IA and ID), whose involvement in membrane resonance, oscillations, and bursting has been well established in these neurons (Del Negro and Chandler, 1997; Wu et al., 2001; Enomoto et al., 2006; Saito et al., 2006; Hsiao et al., 2009; Yang et al., 2009). Accordingly, IA and ID of MesV neurons from both species were characterized following standard protocols (see Materials and methods). To avoid bias by the difference in cell size between species, conductance was normalized by the cell’s membrane capacitance (density). This is based on the finding that membrane capacitance is on average ∼25% larger in rats compared with mice (rats: 55.1 ± 11.4 pF [SD], n = 34 cells, N = 8 animals; mice: 39.9 ± 10.8 pF [SD], n = 36 cells, N = 12 animals; P = 2.65 × 10⁻⁷, unpaired, two-tailed t test; P = 0.0015, unpaired, two-tailed nested t test; data not shown), indicating that MesV neurons from rats present a larger membrane surface area.

Fig. 6 A shows representative results from rats and mice, in which IA was recorded in voltage clamp, during protocols consisting of a series of step commands from 0 to 70 mV in steps of 5 mV and starting from a holding potential of approximately −70 mV. Membrane currents obtained following this procedure showed rapid activation and an inactivation process with kinetics typical of IA (Fig. 6 B; Storm, 1990; Mitterdorfer and Bean, 2002). From membrane current recordings, conductance values were calculated to construct activation curves that were fitted to a Boltzmann function (Fig. 6 C). From these fits, maximum conductance (gmax), half activation voltage (Vhalf), and slope at Vhalf were obtained for the population of recorded neurons. This analysis revealed that IA from rats and mice present similar characteristics and cannot account for differences in membrane excitability between these species. In fact, gmax normalized by the cell’s capacitance averaged 1.40 ± 0.61 nS/pF [SD] (n = 17 cells, N = 5 animals) in rats and 1.42 ± 0.38 nS/pF [SD] (n = 14 cells, N = 6 animals) in mice (P = 0.944, unpaired, two-tailed t test; Fig. 6 D, left), Vhalf averaged −27.9 ± 4.4 mV [SD] (n = 17 cells, N = 5 animals) in rats and −28.9 ± 5.2 mV [SD] (n = 14 cells, N = 6 animals) in mice (P = 0.571, unpaired, two-tailed t test; Fig. 6 D, middle), whereas slope averaged 9.51 ± 1.71 mV [SD] (n = 17 cells, N = 5 animals) in rats and 9.55 ± 2.61 mV [SD] (n = 14 cells, N = 6 animals) in mice (P = 0.969, unpaired, two-tailed t test; Fig. 6 D, right). These results were confirmed by hierarchical statistical analysis which considers possible clustering of data obtained from the same animal (Fig. S1).

Representative recordings of ID from rats and mice are shown in Fig. 7 A. This conductance is characterized by its rapid activation and slow voltage-dependent inactivation kinetics (Fig. 7 B). ID inactivation is considerably slower than that of the IA, confirming that they were successfully separated by the experimental protocols. Strikingly, in contrast to the IA, ID gmax is 63% higher in MesV neurons from mice compared with rats, averaging 2.26 ± 0.90 nS/pF [SD] (n = 13 cells, N = 5 animals) and 1.39 ± 0.56 nS/pF [SD] (n = 20 cells, N = 5 animals), respectively (P = 0.0057, unpaired, two-tailed t test; Fig. 7, C and D, left), whereas Vhalf (rat: −39.2 ± 3.8 mV [SD], n = 20 cells, N = 5 animals; mouse: −36.3 ± 4.2 mV [SD], n = 13 cells, N = 5 animals; P = 0.056, unpaired, two-tailed t test; Fig. 7 D, middle) and slope at Vhalf (rat: 6.92 ± 2.42 mV [SD], n = 20 cells, N = 5 animals; mouse: 8.54 ± 2.70 mV [SD], n = 13 cells, N = 5 animals; P = 0.092, unpaired, two-tailed t test; Fig. 7 D, right) did not exhibit statistical difference. These results were confirmed by hierarchical statistical analysis which considers possible clustering of data obtained from the same animal (Fig. S2). A comparison of average activation curves for the population of recorded neurons confirms the higher functional expression of ID in mice (Fig. 7 E). Moreover, its incomplete inactivation determines a large window current of about 50–60% of gmax at membrane voltages between −40 and −30 mV and about 10% close to the RMP in both species (Fig. 7, F and G), supporting the notion that ID contributes to set the Rin and the RMP. This is consistent with a more depolarized threshold and lower excitability of mice MesV neurons (Bekkers and Delaney, 2001; Guan et al., 2007; Higgs and Spain, 2011; Ordemann et al., 2019). On the other hand, the INap, whose involvement has also been established as a critical determinant of MesV neurons’ excitability in both species (Wu et al., 2001; Enomoto et al., 2006, 2007), showed no significant difference between rats and mice. Indeed, INap gmax averaged 0.096 ± 0.025 nS/pF [SD], (n = 10 cells, N = 3 animals) and 0.12 ± 0.075 nS/pF [SD] (n = 8 cells, N = 4 animals) in rats and mice, respectively (P = 0.445, unpaired, two-tailed t test). The Vhalf averaged −45.9 ± 4.5 mV [SD] (n = 10 cells, N = 3 animals) and −43.2 ± 8.2 mV [SD] (n = 8 cells, N = 4 animals) in rats and mice, respectively (P = 0.412, unpaired, two-tailed t test), whereas the slope at Vhalf averaged 5.1 ± 1.0 mV [SD] (n = 10 cells, N = 3 animals) and 6.0 ± 2.4 mV [SD] (n = 8 cells, N = 4 animals) in rats and mice, respectively (P = 0.356, unpaired, two-tailed t test; Fig. S3, A–E). These results were confirmed by hierarchical statistical analysis which considers possible clustering of data obtained from the same animal (Fig. S3 F).

Taken together, the preceding results strongly suggest that the higher level of ID expression in MesV neurons from mice underlies their lower excitability in relation to rat MesV neurons. To confirm this interpretation, the involvement of ID in regulating cellular excitability was directly assessed by a pharmacological approach. Fig. 8, A and B, shows results from mice MesV neurons in which the addition of 4-AP (30 μM) to block ID resulted in a marked increase in firing for the entire range of injected current intensity. In fact, these neurons that typically respond to depolarizing current pulses with one or two spikes in control conditions, in the presence of 4-AP, respond with robust repetitive responses. Consistently, the firing gain (slope of linear regressions of spikes vs. current relationships) displayed a significant increase from 0.0015 ± 0.0004 spikes/pA [SD] in control to 0.012 ± 0.010 spikes/pA [SD] after 4-AP addition (n = 19 cells, N = 3 animals; P = 0.00019, paired, two-tailed t test; Fig. 8, B and C), whereas the rheobase was significantly reduced from 389.5 ± 117.4 pA [SD] in control to 78.9 ± 41.9 pA [SD] in the presence of 4-AP (n = 19 cells, N = 3 animals; P = 1.89 × 10⁻¹⁰, paired, two-tailed t test; Fig. 8 D). In addition, blockade of this membrane conductance resulted in a small but significant depolarization of the RMP (control: −51.7 ± 1.9 mV [SD]; 4-AP: −50.0 ± 3.1 mV [SD]; n = 34 cells, N = 10 animals; P = 0.0010, paired, two-tailed t test), as well as an increase of the Rin (control: 132.9 ± 43.8 MΩ [SD]; 4-AP: 140.1 ± 39.9 MΩ [SD]; n = 34 cells, N = 10 animals; P = 0.0031, paired, two-tailed t test; Fig. S4), thus confirming the involvement of the ID in setting the cell’s Rin and RMP as suggested in voltage clamp experiments summarized in Fig. 7 G. These results clearly indicate that the ID plays a critical role regulating membrane excitability, and its expression level contributes to defining the electrophysiological phenotype of MesV neurons in a species-specific fashion. To test whether the control of membrane excitability exerted by ID also has an impact on the efficacy of postsynaptic recruitment, the effect of 4-AP (30 μM) was assessed on the ability of presynaptic action potentials to drive spiking in postsynaptic coupled neurons in mice. Results from these experiments are illustrated in Fig. 8 E, in which spiking in the presynaptic neuron fails to activate the coupled neuron in control conditions (left), whereas after the addition of 4-AP (30 μM), there is a dramatic increase in firing of the postsynaptic cell (right). In fact, while in control conditions recruitment of the postsynaptic neuron was not observed in any of the tested directions, after blockade of the ID, recruitment occurred in 11 out of 19 directions, close to the proportion observed in rats in control conditions (see Fig. 1 C). Consistently, the average number of postsynaptic spikes in relation to presynaptic spikes varied from 0.0 ± 0.0% [SD] in control to 19.7 ± 30.4% [SD] in 4-AP (n = 19 directions, N = 3 animals; P = 0.011, paired, two-tailed t test; Fig. 8 E, inset). These results clearly indicate a critical role of the ID in controlling the efficacy of postsynaptic recruitment, and hence in determining the operation mode of electrical synapses between MesV neurons.

Further insights into the mechanism by which such control is exerted was obtained by comparing the spikelets from these two species. Typical examples from rats and mice are illustrated in Fig. 9, A and B, respectively. Despite similar peak amplitudes, their waveforms present important differences. Rat spikelets display a falling phase characterized by large variability and delayed time to peak, resulting in a significant longer duration (Fig. 9 C). Indeed, spikelet half-amplitude duration averaged 4.6 ± 1.8 ms [SD] (n = 22 cells, N = 13 animals) and 2.4 ± 0.7 ms [SD] (n = 106 cells, N = 47 animals) in rats and mice, respectively (P = 2.03 × 10⁻⁵, unpaired, two-tailed t test; Fig. 9 D; see also Table S1). This protracted duration is consistent with the participation of the INap, whose activation promotes the spiking of the postsynaptic coupled neurons in rats (Curti and Pereda, 2004; Curti et al., 2012). In contrast, in mice MesV neurons, which on average express 63% more ID, the INap is strongly antagonized by the swift activation of this outward current, curtailing the spikelets’ duration and therefore reducing the efficacy of postsynaptic recruitment. Consistently, blockade of the ID in mice significantly increased the spikelets’ time course, whose half-amplitude duration averaged 2.5 ± 0.8 ms [SD] and 4.9 ± 1.4 ms [SD] before and after the addition of 4-AP (30 μM), respectively (P = 2.30 × 10⁻⁹, paired, two-tailed t test; n = 19 cells, N = 6 animals; Fig. 9 E). Noteworthy, this change resulted from the slowing of the spikelet’s falling phase (Fig. 9 E, inset). While this approach to assess spike transmission in current clamp during paired recordings suggest the involvement of the ID in shaping spikelets, the fact that ID also contributes to spike repolarization (Fig. 8 E; Mitterdorfer and Bean, 2002) implies that changes in presynaptic spike waveform would also affect spikelet duration. Thus, to confirm the involvement of the ID in spike transmission, artificial spikelets were generated in single MesV neurons from mice by injecting current waveforms corresponding to postjunctional currents measured in voltage clamp during paired recordings. For this, pairs of electrically coupled MesV neurons were simultaneously recorded, one cell in current clamp and the other in voltage clamp. While spiking was induced in the current-clamped cell by means of short depolarizing current pulses, the resultant membrane current (postjunctional current) was recorded in the postsynaptic voltage-clamped neuron. During independent recordings from single cells, the opposite (sign reversed) of the postjunctional recorded current was injected as a current clamp command, and its intensity was scaled according to the neuron’s Rin to obtain a spikelet of typical amplitude (Fig. 9 F, inset). Such artificial spikelets, which were indistinguishable from real ones (spikelet amplitude: 6.13 ± 1.41 mV [SD], n = 122 cells, N = 50 animals; artificial spikelet amplitude: 6.63 ± 0.95 [SD], n = 17 cells, N = 5 animals; P = 0.073, unpaired, two-tailed t test), were generated in control conditions and in the presence of 4-AP (30 μM) to evaluate the contribution of the ID to its waveform (Fig. 9 F). In those cases, in which the 4-AP application depolarized the neuron’s RMP, it was corrected to near pre-application levels by injecting DC. Confirming our hypothesis, half-amplitude duration of artificial spikelets was significantly increased after ID blockade, averaging 2.0 ± 0.3 ms [SD] and 5.3 ± 5.3 ms [SD] in control and in the presence of 4-AP (30 μM), respectively (P = 0.023, paired, two-tailed t test; n = 17 cells, N = 5 animals). Moreover, recruitment by artificial spikelets increased from 0.0 ± 0.0% [SD] in control to 9.1 ± 16.4% [SD] in the presence of 4-AP (30 μM; P = 0.037, paired, two-tailed t test; n = 17 cells, N = 5 animals). Altogether, these results suggest that the density of ID determines coupling potentials duration and postsynaptic recruitment in electrically coupled MesV neurons.

The comparative study of the MesV nucleus from rats and mice exposed a remarkable species-specific difference in the operation mode of electrical synapses, revealing the role of the intrinsic electrophysiological properties in shaping the behavior of circuits of coupled neurons. In the present study, a naturally occurring difference in the efficacy of postsynaptic recruitment between electrically coupled MesV neurons was used to uncover the critical role of subthreshold K⁺ currents. We showed that while in rats the spiking of one MesV neuron activates its coupled partner in ∼50% of the cases, this rarely occurs between mice MesV neurons, in which spiking results in a subthreshold postsynaptic response in ∼98% of the tested pairs. Noteworthy, the population of connected MesV neurons from both species bears similar strength of electrical coupling as a result of the comparable magnitude of its determinants (Gj and neuronal Rin). Thus, despite spike-evoked postjunctional coupling potentials (spikelets) from both species being of similar amplitude and arising from a comparable resting membrane potential level, postsynaptic recruitment is considerably more efficient in rats than in mice. This striking difference is imposed by the differential expression of the D-type K⁺ current, whose density is significantly higher in MesV neurons from mice compared with rats. Consistently, previous reports suggested that the firing properties of MesV neurons result from the relative expression of the INap and the subthreshold K⁺ currents (Wu et al., 2001, 2005; Hsiao et al., 2009). With regard to this, MesV neurons have been shown to express several types of Kv1 subunits whose pharmacological profile and voltage dependency match those of the ID characterized in the present study. In fact, α-DTX, considered a specific blocker of Kv1 mediated currents, shows an almost complete overlap in the blocking effect with 4-AP in the low micromolar range (50 μM; Saito et al., 2006; Hsiao et al., 2009). Moreover, while membrane capacitance estimates indicate that MesV neurons from rats present a larger membrane surface area, Rin is similar in both species. This suggests that cells’ size difference, which should result in lower Rin values in rats compared with mice, may be compensated by the higher expression level of ID in mice.

A previous experimental and theoretical work has already shown the relevance of active membrane properties levering the activity of networks of coupled neurons, particularly through the action of boosting mechanisms like the INap (Mann-Metzer and Yarom, 1999; Pfeuty et al., 2003; Curti and Pereda, 2004; Dugué et al., 2009; Curti et al., 2012). The synergic operation of these mechanisms endows neural circuits with the ability to generate synchronous and rhythmic patterns of activity with potential functional relevance for both physiological and pathophysiological processes (Draguhn et al., 1998; Perez Velazquez and Carlen, 2000; Hormuzdi et al., 2001; Mylvaganam et al., 2014). Moreover, other circuit operations supported by electrical synapses like coincidence detection are also critically shaped by the intrinsic excitability of neurons. Indeed, the ability of coupled neurons to discriminate between synchronous and temporally uncorrelated inputs is strongly regulated by the hyperpolarization-activated cationic current (H-current; Davoine and Curti, 2019). In spite of this background, direct experimental evidence for the role of K⁺ currents in the context of electrical synaptic transmission was lacking. Here, we show that the expression level of the D-type K⁺ current, a depolarization-activated subthreshold conductance, whose activation kinetics ranges in the time scale of physiologically relevant signals like action potentials, gates the transfer of spikes between coupled neurons. In fact, the swift activation of this outward current near the resting membrane potential opposes depolarizations, curtailing the duration of spike-evoked coupling potentials and regulating membrane excitability. While the age of the rats and mice employed in this study does not match exactly (rats ranged from P7 to P16, mean: P12.1 d; mice ranged from P12 to P18, mean: P15.1 d), this cannot explain the observed differences. Remarkably, since the excitability of MesV neurons is developmentally regulated (increases) over the first 10 postnatal days (Curti et al., 2012), the opposite findings would be expected in firing properties and efficacy of postsynaptic recruitment if solely based on the age difference between rats and mice. This strongly suggests that our results truly represent phenotypic differences between these species’ MesV neurons.

Previous work has shown that the synergistic interaction between electrical coupling and the intrinsic neuronal properties promotes the strong synchronous activation of MesV neurons (Curti et al., 2012). Interestingly, while the efficacy of postsynaptic recruitment in rats is high, facilitating the spread of excitation among MesV neurons, the incidence of electrical coupling is relatively low (∼23% of apposed pairs are electrically coupled in rats compared to ∼63% in mice; Curti et al., 2012). Thus, cellular excitability and electrical coupling at the MesV nucleus of these species are inversely related, suggesting that the expression of these mechanisms is reciprocally regulated in a sort of homeostatic relationship. In this regard, regulations at the network level that reciprocally operates on the expression of mechanisms underlying cellular excitability and interneuronal connectivity supported by gap junctions might be critical to ensure network function stability, as was proposed in networks of neurons interconnected by chemical contacts (Marder and Goaillard, 2006). Indeed, widespread coupling between highly excitable neurons might lead to hypersynchrony and aberrant spiking across neuronal ensembles, characteristic of diseases like epilepsy (Perez Velazquez and Carlen, 2000).

MesV neurons are primary sensory afferents that originate in the spindles of jaw-closing muscles (masseter) and mechanoreceptors of periodontal ligaments (Corbin and Harrison, 1940). In turn, similar to their spinal cord counterparts, MesV neurons establish monosynaptic excitatory connections with motoneurons controlling the same muscles, contributing to the organization of orofacial behaviors (Luo and Li, 1991; Grimwood et al., 1992; Luo et al., 2001; Stanek et al., 2014). Interestingly, differences in the efficacy of postsynaptic recruitment between rats and mice strongly suggest that circuits of electrically coupled MesV neurons from these species operate in totally different regimes. In a large proportion of electrically coupled pairs from rats, activation of one afferent might lead to the activation of a coupled partner, increasing the number of active afferents that respond coordinately to the sensory input, thus supporting lateral excitation as was shown in many sensory systems (El Manira et al., 1993; Herberholz et al., 2002; Rela and Szczupak, 2004). In this way, this phenomenon enhances or amplifies the influence of sensory input on jaw-closing motoneurons. In striking contrast, spike transmission at MesV contacts in mice rarely results in the activation of the postsynaptic neuron due to the higher expression level of the ID, indicating that these contacts do not support lateral excitation. Instead, as these afferents also receive excitatory synaptic input from hierarchical superior centers (Lazarov, 2002), electrical synapses might promote activation of coupled MesV neurons when inputs are temporally correlated, thus acting as coincidence detectors (Davoine and Curti, 2019).

Finally, these findings illustrate how related species display different cellular and circuital strategies to solve the same biological problem, in this case, the control of orofacial behaviors. Similar diversity in homolog circuits from rats and mice has also been shown in hypothalamic neurons controlling pituitary prolactin secretion whose networks display contrasting behaviors. However, in contrast to our findings, such diversity arises from interspecific differences in the electrical coupling, while neurons from both species present comparable membrane properties (Stagkourakis et al., 2018).

The data underlying all figures are available in the published article and its online supplemental material.

Jeanne M. Nerbonne served as editor.

We thank A. Pereda, G. Budelli, F. Trigo, and V. Abudara for critical discussions and comments.

This work was supported by Agencia Nacional de Investigación e Innovación (ANII), Uruguay (FCE_1_2021_1_166745), Programa de Desarrollo de las Ciencias Básicas (PEDECIBA), and Comisión Académica de Posgrado of Universidad de la República (BDMX_2022_1#48716352).

Author contributions: A. Dapino: Conceptualization, data curation, formal analysis, investigation, methodology, software, visualization, and writing—review & editing. F. Davoine: Conceptualization, data curation, formal analysis, investigation, methodology, software, visualization, and writing—review & editing. S. Curti: Conceptualization, data curation, formal analysis, funding acquisition, project administration, supervision investigation, methodology, visualization, and writing—original draft.

The data underlying all figures are available in the published article and its online supplemental material.