Abstract
Comparative analysis of evolutionarily conserved neuronal circuits between phylogenetically distant mammals highlights the relevant mechanisms and specific adaptations to information processing. The medial nucleus of the trapezoid body (MNTB) is a conserved mammalian auditory brainstem nucleus relevant for temporal processing. While MNTB neurons have been extensively investigated, a comparative analysis of phylogenetically distant mammals and the spike generation is missing. To understand the suprathreshold precision and firing rate, we examined the membrane, voltage-gated ion channel and synaptic properties in Phyllostomus discolor (bat) and in Meriones unguiculatus (rodent) of either sex. Between the two species, the membrane properties of MNTB neurons were similar at rest with only minor differences, while larger dendrotoxin (DTX)-sensitive potassium currents were found in gerbils. Calyx of Held-mediated EPSCs were smaller and frequency dependence of short-term plasticity (STP) less pronounced in bats. Simulating synaptic train stimulations in dynamic clamp revealed that MNTB neurons fired with decreasing success rate near conductance threshold and at increasing stimulation frequency. Driven by STP-dependent conductance decrease, the latency of evoked action potentials increased during train stimulations. The spike generator showed a temporal adaptation at the beginning of train stimulations that can be explained by sodium current inactivation. Compared with gerbils, the spike generator of bats sustained higher frequency input-output functions and upheld the same temporal precision. Our data mechanistically support that MNTB input-output functions in bats are suited to sustain precise high-frequency rates, while for gerbils, temporal precision appears more relevant and an adaptation to high output-rates can be spared.
SIGNIFICANCE STATEMENT Neurons in the mammalian medial nucleus of the trapezoid body (MNTB) convey precise, faithful inhibition vital for binaural hearing and gap detection. The MNTB's structure and function appear evolutionarily well conserved. We compared the cellular physiology of MNTB neurons in bat and gerbil. Because of their adaptations to echolocation or low frequency hearing both species are model systems for hearing research, yet with largely overlapping hearing ranges. We find that bat neurons sustain information transfer with higher ongoing rates and precision based on synaptic and biophysical differences in comparison to gerbils. Thus, even in evolutionarily conserved circuits species-specific adaptations prevail, highlighting the importance for comparative research to differentiate general circuit functions and their specific adaptations.
- auditory system
- comparative neurophysiology
- medial nucleus of the trapezoid body
- membrane properties
- superior olivary complex
- synaptic transmission
Introduction
The medial nucleus of the trapezoid body (MNTB) is embedded in neuronal circuits that rely on temporally precise sound processing, where the MNTB acts as a key element conveying information fast and with high fidelity. Principal MNTB neurons receive one large glutamatergic somatic synapse, the calyx of Held, arising from globular bushy cells of the contralateral cochlear nucleus (Harrison and Warr, 1962; Ryugo and Rouiller, 1988; Smith et al., 1991). The excitatory information is converted to glycinergic inhibition projecting to other ipsilateral auditory nuclei (Glendenning et al., 1981; Borst and Soria van Hoeve, 2012; Altieri et al., 2014; Burchell et al., 2022).
Synaptically evoked input-output functions via the calyx of Held have been extensively investigated in the MNTB of cats, guinea pigs and rodents (Guinan and Li, 1990; Banks and Smith, 1992; Taschenberger and von Gersdorff, 2000; Futai et al., 2001; Stasiak et al., 2018). This synapse elicits large, fast EPSCs (Taschenberger and von Gersdorff, 2000) because of simultaneous transmitter release from numerous active zones (Meyer et al., 2001; Sätzler et al., 2002). These EPSCs are integrated via the cell's capacitive element and voltage-activated conductances (Kaczmarek et al., 2005; Leão, 2019; Kladisios et al., 2020) to produce one-to-one input-output functions at high rates in vivo (Lorteije et al., 2009; Stange-Marten et al., 2017). In vitro, these synapses show frequency-dependent short-term plasticity (STP; von Gersdorff et al., 1997; Wang and Kaczmarek, 1998) with an interplay of short-term depression (STD; for review, see von Gersdorff and Borst, 2002; Neher, 2015) and short-term facilitation (STF; Tsujimoto et al., 2002; Felmy et al., 2003).
Postsynaptically, various potassium channels enable MNTB neurons to respond with only one action potential in response to a single large calyceal EPSC. Thereby, different potassium currents exert shunting and repolarizations on membrane potential dynamics (for review, see Kaczmarek et al., 2005; Leão, 2019). Additionally, sodium current inactivation has been described in MNTB principal neurons (Leão et al., 2005, 2006, 2008). Despite extensive knowledge of ion channels relevant for action potential waveform (Dodson et al., 2002; Macica et al., 2003; B. Yang et al., 2007; Johnston et al., 2008; Leão et al., 2008), the capabilities of the spike generator have not been detailed, especially with focus on maximal firing rate, temporal precision and species-dependent adaptations.
In different mammals, the properties of the calyx of Held synapse and the postsynaptic action potential generator might be differently adapted, since parts of the circuits in the superior olivary complex (SOC) deviate at least between rodents and bats. For example, the medial superior olive (MSO) is innervated by MNTB projections and performs interaural time difference detection in gerbils (Brand et al., 2002), but in bats it serves other computational tasks, such as pattern recognition and binaural filtering (Grothe, 1994; Grothe and Park, 2000). These tasks might rely on differently adapted MNTB mediated inhibition. Comparative approaches can solve questions about species-specific adaptations, and allow a generalized view on neuronal circuits to gain deeper insights into the cellular mechanisms driving different adaptive states. The MNTB is exquisitely suited for comparative analysis, as it is present in all mammals (Kulesza and Grothe, 2015), but its feed forward inhibition might serve species-specific processing tasks.
To gain insights into species-dependent cellular adaptations in the same mammalian circuit, we comparatively assayed the biophysical membrane properties, the synaptic STP and the dynamics of the spike generator in a rodent (Meriones unguiculatus, gerbil) and a bat (Phyllostomus discolor) species. Membrane properties of principal MNTB neurons near rest were similar in both species, but in gerbils larger dendrotoxin (DTX)-sensitive potassium currents were observed. Synaptic inputs are smaller in bats compared with gerbils, but show less frequency-dependent STP composed of an interplay between STF and STD. Finally, the spike generator in bats is more suited to promote high rates of input-output functions compared with gerbils.
Materials and Methods
Animal models
Bats and rodents are well suited for comparative analyses, because, judged by the number of species each group contains, they are the most successful among mammals. From both groups, species have been extensively used for auditory research and despite their long evolutionary separation time (91–101 million years; http://www.timetree.org/; Hedges et al., 2006; Kumar et al., 2017), they share several sound processing related features. Phyllostomus discolor is an echolocating bat with a hearing range between 5 and 120 kHz and the lowest hearing thresholds occur at ∼30 and 75 kHz (Esser and Daucher, 1996; Linnenschmidt and Wiegrebe, 2019). Meriones unguiculatus (gerbil) has well developed hearing abilities with lowest thresholds at 6 kHz and a hearing range between 0.2 and 60 kHz (Ryan, 1976). Both animals live in social structures and communicate with a substantial vocal repertoire, so their hearing abilities are predicted to be highly developed. The overlapping hearing range is extensive and is dominated by high-frequency hearing above 2.5 kHz, a frequency that marks the upper limit for phase locking in mammals. Thus, both species are characterized by very well-developed hearing with similar properties and are therefore exquisitely suited for our comparative analysis.
The colony of Phyllostomus discolor bats was kept at the institute's animal facility with 12/12 h light/dark cycle. The temperature was kept constant between 25°C and 28°C with humidity at 70%. The bats had access to a diet of bananas, small insects, oatmeal, and water ad libitum. All experiments regarding Phyllostomus discolor bats were approved by the university's local authorities under the license number TiHo-T-2019-7 and were in accord with the German law for animal protection. The age range of investigated bats spanned approximately between 1 and 12 years, although animal age could not individually assigned in an open colony. Mongolian gerbils (Meriones unguiculatus), based on Charles River hereditary background, were kept in the institute's animal facility under 12/12 h light/dark cycle, fed ad libitum and were used as matured animals [postnatal day (P)185 ± 30; with a range of P98–P343]. Experiments regarding Mongolian gerbils were approved under the license number TiHo-T-2019-4 and were compliant with German local and federal laws. Data from recordings of gerbil neurons presented in Figure 3 were reanalyzed from previously acquired and published experiments by Kladisios et al. (2020).
Preparation
Animals of both sexes were deeply anesthetized with isoflurane, decapitated and the brains were removed in slice solution (room temperature) consisting of the following (in mm): 120 sucrose, 25 NaCl, 25 NaHCO3, 2.5 KCl, 1.25 NaH2PO4, 3 MgCl2, 0.1 CaCl2, 25 glucose, 0.4 ascorbic acid, 3 myo-inositol, and 2 Na-pyruvate, which was bubbled with 95% O2 and 5% CO2 resulting in pH 7.4. Slices of 200-μm thickness containing the MNTB were taken with a VT1200S vibratome (Leica) and were incubated for 45 min at 34°C in recording solution containing the following (in mm): 125 NaCl, 25 NaHCO3, 2.5 KCl, 1.25 NaH2PO4, 1 MgCl2, 1.2 CaCl2, 25 glucose, 0.4 ascorbic acid, 3 myo-inositol, and 2 Na-pyruvate, bubbled with 95% O2 and 5% CO2 with pH 7.4.
Electrophysiology
Whole-cell recordings were performed at 34–36°C or 20–22°C for sodium current measurements on slices containing MNTB neurons, which were identified and visualized with a Retiga 2000DC camera connected to a TILL Photonics System (FEI) and a monochromator (Polychrome V), or with a pco.edge 3.1 camera, both mounted on a chamber where heated recording solution was continuously perfused. Recordings were acquired with EPC10 USB amplifiers controlled by Patchmaster (HEKA). All potentials recorded and presented have been corrected for liquid junction potential (LJP), calculated with custom written IGOR Pro scripts according to previously published work (Barry, 1994; Ammer et al., 2015). The internal solution for current and dynamic clamp recordings contained the following (in mm): 145 K-gluconate, 15 HEPES, 4.5 KCl, 7 Na2-phosphocreatine, 2 Mg-ATP, 2 K2-ATP, 0.3 Na2-GTP and 0.5 K-EGTA, or 0.5 Na-EGTA for recording potassium currents, resulting in a LJP of 15.95 mV. The internal voltage clamp solution for fiber stimulation experiments contained the following (in mm): 130 Cs-gluconate, 20 TEA-Cl, 10 HEPES, 5 Cs-EGTA, 5 Qx-314, 5 Na2-phosphocreatine, 4 Mg-ATP, 0.3 Na2-GTP, and 0.1 spermine, and had a LJP of 13.3 mV. For recording sodium currents the internal voltage clamp solution did not contain QX-314 or spermine. Data were acquired with a sample rate of 20 or 5 µs (Fig. 8) and filtered at 3 kHz. Electrode resistance ranged from 3 to 5 MΩ and access resistance was compensated to a residual of 2 or 5 MΩ (Fig. 8) during voltage clamp experiments. In current and dynamic clamp, the access resistance was bridge balanced to 100% and no holding current was applied.
Membrane properties
To quantify the membrane time constant (τmem) of MNTB neurons, we injected hyperpolarizing currents of −5 pA for at least 50 repetitions and fitted an exponential function at onset on the average traces. Input resistance (Rin) was calculated by hyperpolarizing the neurons with −5 mV and applying Ohm's law [Rin = (V/Iss)], where Iss is the steady state current deflection. To estimate the somatic capacitance (C), a biexponential function was fitted from the maximum deflection, and a new monoexponential curve was created by using the fast decay time of the biexponential fit. Capacitance was quantified as the integrated area from the start of the stimulus, till three times the fast decay time constant. To estimate the voltage threshold, we injected ramp currents with 0.3-ms rise and 0.8-ms descending duration with 100-pA intervals. For the first suprathreshold event, we plotted a phase plane graph of dV/dt against V and assessed the voltage where the slope rapidly increased as voltage threshold. Its voltage difference from the peak of the action potential was defined as action potential amplitude, and the duration between rise and fall level equal to half the distance between voltage threshold and action potential peak as halfwidth. Current-voltage relationships were acquired on the same neurons, by injecting 500-ms-long square steps between 900 and −500 pA, decreasing with 100-pA intervals. The maximum voltage deflection (Vmax) at the beginning of the pulse was subtracted from the steady state voltage to estimate the sag potential.
Potassium currents
Whole-cell potassium currents were pharmacologically isolated with SR95531 (10 μm), strychnine (0.5 μm), DNQX (20 μm), D-AP5 (50 μm), ZD7288 (50 μm), Cd2+ (100 μm), and TTX (1 μm). DTX-sensitive Kv1 currents were blocked with 80 nm dendrotoxin (DTX). DTX was dissolved and applied in Cytochrome C to reduce surface absorption. Between sweeps cells were held at −76 mV and before the voltage step at −86 mV for 500 ms. Step potentials of one second from −86 to 24 mV with 10-mV steps were introduced and currents recorded.
Synaptic transmission
To stimulate the calyx of Held synapse, a glass pipette of up to 5 MΩ, filled with external recording solution, was placed in the vicinity of MNTB neurons and 200 μs long biphasic voltage pulses of varying strength were triggered via the amplifier and delivered by a Model 2100 Isolated Pulse Stimulator (A-M Systems). Only all-or-none responses, validated by stepwise increase of the stimulation voltage, indicative of calyx of Held terminals were recorded. Inhibitory currents were blocked with 1 μm strychnine and 10 μm SR95537. After detecting a calyx of Held input the stimulated EPSCs were recorded at step potentials between −93 to +67 mV in 10-mV intervals and EPSC size, kinetics and current-voltage (IV) relationships were measured. The EPSC decay time (τEPSC) was calculated by fitting an exponential function from the current peak. Since EPSCs showed a prominent inward rectification at positive voltages, we calculated rectification indices (RIs), according to Scheuss and Bonhoeffer (2014) as follows: a straight line was fitted from −93 to −23 mV and projected at positive voltages. The RI was defined as the ratio of the measured to the projected current value at 47 mV. Additionally, 20-pulse trains of 10, 50, 100, 300, and 500 Hz stimulated the synapse. STP was quantified by normalizing the envelope EPSC peaks to the first pulse. Steady state depression was determined as the average STP ratio of the last three pulses. To compare bat STP results with gerbils, we reanalyzed data from previously published fiber stimulation experiments (Kladisios et al., 2020). Specifically, the τEPSC, RI and STP of gerbils were analyzed as described above. Since in gerbils responses to 500-Hz train stimulations could not reliably been elicited, we extrapolated their STP at high input rates by fitting a linear function between 10 and 300 Hz for each stimulation step.
Dynamic clamp
To assess the excitability of MNTB neurons to physiological EPSCs, we applied the dynamic clamp interface introduced previously (Y. Yang et al., 2015). We produced trains of excitatory postsynaptic conductances (EPSGs) with 10, 50, 100, 300, 500, and 800 Hz, where each pulse decayed with the median EPSC decay time. Two sets of EPSG templates were created, one with equal conductance amplitudes for each pulse and another with STP, where the amplitudes facilitated or depressed according to the average ratios measured previously. Since no STP data were collected from 800-Hz stimulation train paradigms, we projected their STP ratio for each step from the STP of lower stimulation frequencies. Injected linear, AMPAR-mediated currents were calculated online, according to: IAMPA = G(t) (Vm – Vrev). For the dynamic clamp experiments, the LJP was corrected online. After determining the conductance threshold for action potential generation, neurons were first injected with conductance amplitudes equal to a single average EPSG, and then with 10% above conductance threshold. Each template was presented 10 times, and the average number of action potential events was calculated. Latency, i.e., the time difference between stimulus to action potential peak, was calculated and averaged.
To assess the absolute refractory period, we injected two conductance pulses with equal amplitude and interstimulus intervals ranging from 0.5 to 100 ms at the conductance equivalent of a single average input, as well as at 10% above conductance threshold. The cutoff frequency, where neurons could no longer segregate the second input, was measured for both conditions. To reciprocally compare bat and gerbil MNTB neurons, species-specific 100-pulse templates of frequencies between 10 and 800 Hz and with maximum amplitude equal to steady state conductances were injected in both species, and the success rate and temporal fidelity were analyzed, as described above.
Sodium currents
Sodium current inactivation, recovery and train stimulations were recorded in the presence of strychnine (0.5 μm), DNQX (20 μm), ZD7288 (50 μm), Cd2+ (200 μm), TEA (10 mm), and 4-AP (2 mm). To analyze sodium current inactivation, cells were held for 50 ms at −93 mV, followed by a 50-ms holding potential between −113 and −38 mV, incremented at 5-mV steps, and succeeded by a stable voltage step of −33 mV for 30 ms. Recovery from sodium current inactivation was recorded for intervals ranging between 0.5 and 33.8 ms between control and test depolarizations, while neurons were held at a holding potential of −68, −73, −93, and −113 mV. The recovery time was extracted by fitting biexponential functions to the average data. Finally, we normalized rate and steady state fraction of maximum sodium currents, recorded in response to train stimulations of 100, 300 and 500 Hz at −68-, −73-, −93-, or −113-mV holding potential.
Histology
Bats were euthanized by inhalation of carbon dioxide and declared dead after at least 1 min of breathing arrest. Subsequently, they were transcardially perfused with cold Ringer solution containing heparin, followed by fixation with 2% paraformaldehyde (PFA) containing 15% picric acid. Brains were removed, postfixed for 4 h or overnight at 4°C in PFA and washed with PBS consisting of the following (in mm): 138 NaCl, 2.7 KCl, 10.2 Na2HPO4, and 1.76 KH2PO4, adjusted to pH 7.4. Brains were embedded in 4% agarose using a custom-made positioning device enabling a standardized brain orientation (Radtke-Schuller et al., 2020) and transversal slices of 50 μm were obtained with a vibratome Ci 7000 (Campden Instruments). Free-floating slices were washed in PBS, treated with sodium borohydride to reduce autofluorescence and subsequently with blocking solution (0.5% Triton X-100, 1% bovine serum albumin and 0.1% saponin, diluted in PBS) for 30 min at room temperature. Afterwards, they were incubated for 3 d at 4°C in blocking solution containing primary antibodies (Kv1.1: Alomone, rabbit, polyclonal IgG, 1:2000, catalog #APC-009; MAP2: Origene, chicken, polyclonal IgY, 1:1000, catalog #TA336617). Sections were washed in 0.5% Triton X-100 and 0.1% saponin diluted in PBS and incubated in blocking solution containing secondary antibodies (Cy3: Dianova, donkey anti-rabbit, 1:400, catalog #711-165-152; Alexa488: Dianova, donkey anti-chicken, 1:300, catalog #703-546-155) for 4 h at room temperature. After washing, slices were mounted on microscope slides embedded in fluorescence mounting medium (VECTASHIELD, Vector Laboratories). Fluorescent images were obtained with a confocal laser microscope (Leica TCS SP5, Leica Microsystems GmbH). Overview images were taken with a 20× 0.7 NA objective, and insets with a 63× 1.4 NA objective. Combining of confocal images was accomplished in FIJI (Schindelin et al., 2012; Schneider et al., 2012).
Antibody specificity
The Kv1.1 antibody is produced against amino acids 416–495 from C terminus of mouse Kv1.1 and recognizes the appropriate bands on Western blottings of rat brain. This antibody was verified by immunofluorescence in knock-out mice (Zhou et al., 1998). Labeling was eliminated by preincubation of the diluted antiserum with 10 µg/ml of the recombinant protein in rat (manufacturer's datasheet) and gerbil in our lab. Kv1.1 labels neurons and neuropils in the SOC and lateral lemniscus of bat (Rosenberger et al., 2003). We used a higher dilution as suggested by the manufacturers to improve specificity. The MAP2 antibody is generated against amino acids 377–1505 of recombinant human MAP2 and recognizes a single band of 280 kDa on Western blotting of bovine brain (manufacturer's datasheet). MAP2 labels the soma and dendrites of neurons in all major SOC nuclei of mice (Rajaram et al., 2019) and at least in the MSO of gerbil (Callan et al., 2021).
Data and statistical analysis tests
Data were analyzed with IGOR Pro 8 (Wavemetrics) using custom-written procedures and Microsoft Excel. Data were tested for normal distribution with the Kolmogorov–Smirnov test, and are presented as average ± SEM. We used Student's and paired t test, and one-way ANOVA with post hoc Tukey's test for statistical comparison. The significance level was set at 0.05. In total, we recorded electrophysiological data for analysis from 75 bat and 79 gerbil MNTB neurons. The results of Figure 1 were acquired from seven bats and three gerbils, with overall 23 and 13 MNTB neurons, respectively. For the fiber stimulation experiments of Figure 3, we acquired eight MNTB recordings from six bats, and for the dynamic clamp experiments shown in Figures 4–6, six bats and eight gerbils with overall 22 and 23 MNTB neurons, respectively, were recorded. Additionally, 12 neurons acquired from five bats and 17 cells from six gerbils were analyzed for the experiments displayed in Figure 7. To isolate the potassium currents presented in Figure 2, we recorded six neurons from five bats, and eight neurons from four gerbils. Sodium current inactivation, recovery and responses to train stimulations were recorded from five to nine neurons of three gerbils, as well as one to four neurons from three bats.
Data accessibility
The data that support the findings to this study are available from the corresponding authors upon request.
Results
Membrane properties of bat MNTB neurons
To gain insights into the basic properties of bat MNTB principal neurons, we first assessed the membrane characteristics of 23 bat neurons, and compared them with data collected from 13 gerbil neurons. The input resistance (Rin) and cell capacitance (C) were quantified from the steady state current and charging transient of a −5-mV hyperpolarizing step (Fig. 1A). The average Rin of bat and gerbil MNTB neurons (66.3 ± 3.8 and 68.9 ± 13 MΩ), as well as the C (17.6 ± 0.8 and 18.9 ± 1.1 pF) did not significantly differ (Student's t test, Rin: p = 0.81; C: p = 0.34; Fig. 1A). Additionally, the resting potential (Erest) and the membrane time constant (τmem) were extracted from the average voltage responses before, and during a −5-pA hyperpolarization step (Fig. 1B). The average resting membrane potential was −69 ± 0.8 mV, and −70.9 ± 1 mV for bat and gerbil MNTB neurons, respectively (Student's t test, p = 0.178). Bat τmem was not significantly faster than τmem in gerbils (2.13 ± 0.16 and 1.61 ± 0.39 ms for gerbil and bat, respectively; Student's t test, p = 0.16; Fig. 1B). Overall, the obtained membrane properties of adult bat MNTB principal cells are similar to those described before in gerbils (Kladisios et al., 2020) and other rodents like mice (Wu and Kelly, 1991) and rats (Banks and Smith, 1992).
Basal parameters of principal MNTB neurons in the bat Phyllostomus discolor and Mongolian gerbil Meriones unguiculatus. A, Top, Average current response to a −5-mV hyperpolarization step, used to calculate the input resistance (Rin) and capacitance (C). Inset highlights the time course that was used to calculate the membrane charging transient. Bottom, Rin (left) and C (right) of individual bat and gerbil MNTB neurons, filled and open circles, respectively. Colored markers show average ± SEM. B, Top, Average membrane response to a −5-pA current injection. The resting potential (Erest) was calculated from the average voltage before the stimulus. The red line highlights an exponential fit, used to extract the membrane time constant (τmem). Bottom, Erest (left) and τmem (right) of bat and gerbil neurons. C, Subthreshold and suprathreshold responses to a ramp current of 100-pA increasing steps. The first suprathreshold response (black trace) was used to determine the voltage threshold (Vthres, magenta cross), halfwidth (magenta line), and AP amplitude. D, Distribution of Vthres (left), halfwidth (center), and AP (action potential) amplitude (right) of bat and gerbil MNTB neurons. E, Input-output current–voltage responses of a bat and gerbil MNTB neuron to a square current injection exhibit characteristic onset responses and rectification at positive current injections. F, Average maximum (square) and steady state (triangle) voltage deflection of bat and gerbil MNTB neurons (blue filled and red open symbols), quantified from the input-output responses of E. Maximum voltage deflection after hyperpolarizing current injection is comparable between the species, but gerbils show stronger rectification. Asterisks show significant differences on maximum deflection (top) and steady state (bottom). G, The sag potential is plotted against the maximum voltage deflection for bat and gerbil MNTB neurons (blue and red, respectively). Individual data shown as broken lines, symbols show mean ± SEM. Bats: n = 23; gerbils: n = 13.
Action potential parameters of bat MNTB neurons were determined from the first suprathreshold event to an increasing triangular current injection of 0.3 ms rise and 0.8 ms decay time (Fig. 1C). The current size eliciting the first suprathreshold event was similar between bats and gerbils (800 ± 68 and 916.7 ± 105 pA, Student's t test, p = 0.353). The average bat voltage threshold of −51.7 ± 0.8 mV was significantly lower than the gerbil threshold of −46.6 ± 3 mV (Student's t test, p = 0.045; Fig. 1D). The resulting voltage difference between action potential threshold and resting potential was significantly smaller in bat MNTB neurons than in gerbils (bats: 17.3 ± 0.9 mV, n = 23, gerbils: 24.3 ± 3.1 mV, n = 12, Student's t test, p = 0.0091). Measured between voltage threshold and peak, the halfwidth (bats: 208 ± 12.8 μs, gerbils: 183.5 ± 19.8 μs, Student's t test, p = 0.29; Fig. 1D) and amplitude (70.4 ± 3.6 and 59.1 ± 4.7 mV for bats and gerbils; Student's t test, p = 0.067; Fig. 1D) of the action potential were similar for both species. Overall, the described action potential parameters are similar to values obtained from other mammals (Banks and Smith, 1992; Brew and Forsythe, 1995; Kladisios et al., 2020), but bat MNTB neurons' action potential voltage threshold is closer to the resting potential compared with gerbils.
Next, we assessed and compared the suprathreshold and subthreshold membrane potential behavior with 500-ms square current injections. For either species, the first suprathreshold event occurred at considerably smaller current injections compared with the ramp injections, a waveform dependence that has been elucidated in the VNLL before (Franzen et al., 2015). Thresholds of square current injections did not differ between the species (bats: 338.5 ± 33 pA, gerbils: 427.3 ± 57.4 pA, Student's t test, p = 0.18). The subthreshold maximum voltage deflection and the steady state voltage response displayed rectification at positive values and a voltage sag at negative values (Fig. 1E,F). Both the maximum and steady state voltage deflections between the two species were similar for negative current injections (multiple t test, with Holm–Sidak correction, p > 0.05), but the rectification of gerbil neurons was significantly more prominent at positive current injections (multiple t test, with Holm–Sidak correction, maximum: 100 pA; p = 0.044, 200 pA; p = 0.043; steady state: 100 pA; p = 0.33, 200 pA; p = 0.049). The difference in outward rectification might indicate differences in the proportion of low voltage-gated potassium channels between the two species. Taking into account the similar dependence of the sag and maximal hyperpolarizing voltage deflection of bats and gerbils (Fig. 1G), it can be suggested that hyperpolarization-activated and cyclic nucleotide-gated (HCN) channels are similar in both species and match the description in other rodents (Brew and Forsythe, 1995; Dodson et al., 2002; Hassfurth et al., 2009).
Another hallmark feature of many auditory brainstem neurons is the presence of DTX-sensitive, low voltage-activated potassium currents (Brew and Forsythe, 1995; Grigg et al., 2000; Dodson et al., 2002; Svirskis et al., 2004; Scott et al., 2005; Cao and Oertel, 2017; Nabel et al., 2019), which are typically associated with Kv1 channels. To assess the presence of DTX-sensitive potassium currents in adult bat MNTB neurons, we pharmacologically isolated whole-cell potassium currents (bat: n = 6; gerbil: n = 8) and applied 80 nm DTX to block Kv1 channels (Fig. 2A). After DTX wash-in a noninactivating outward current remained (Fig. 2A). To quantify the inactivating DTX-sensitive Kv1 current contribution to whole-cell potassium currents, we subtracted control currents from currents evoked during DTX wash-in (Fig. 2B), and plotted their peak and steady state values against the step potential (Fig. 2C). In both species, we observed that DTX-sensitive currents begin to open between −66 and −56 mV. In bat, the peak and steady state current reached 2.74 ± 0.42 and 1.47 ± 0.38 nA, respectively, at 4-mV step potential. In gerbils at 4-mV step potential the DTX-sensitive currents amounted to 6.39 ± 0.8 nA at the peak and 3.08 ± 0.84 nA in steady state (Student's t test, peak: p = 0.0029, steady state: p = 0.105). The fraction of inactivation [(peak – steady state)/peak] at 4 mV was similar between the two species and ranged between 0.27 and 0.73 in bats, with an average of 0.55 ± 0.08 and from 0.28 to 0.77 in gerbils, averaging 0.5 ± 0.1 (Student's t test, p = 0.77). In order to corroborate the molecular identity of DTX-sensitive currents as Kv1.1 channels in bats, immunofluorescence labeling was performed, revealing somatic Kv1.1 labeling (Fig. 2D,E), as has been indicated before (Rosenberger et al., 2003; Pätz et al., 2022).
Low voltage-activated potassium currents in bat and gerbil principal MNTB neurons. A, Whole-cell potassium currents before (left) and after application of 80 nm DTX (right). Currents were evoked between −86 and 24 mV with 10-mV step intervals. B, Isolated Kv1 channels, obtained from subtracting residual outward currents after DTX application from whole-cell potassium currents (control). C, Peak (filled circles) and steady state (open circles) DTX-sensitive outward potassium currents extracted from bats (blue) and gerbils (red) as a function of step potentials (bats: n = 6; gerbils, n = 8). Note that DTX-sensitive currents were larger in bats. D, Immunofluorescent labeling of bat MNTB neurons against MAP2 (green) and Kv1.1 (magenta). Scale bar: 100 μm. E, Inset from D, taken with higher magnification. Kv1.1 (magenta) is present on the soma of MNTB neurons. Scale bar: 10 μm.
Synaptic transmission at the bat and gerbil calyx of Held synapse
Synaptic transmission in the MNTB is dominated by a strong somatic excitation delivered by the calyx of Held, and has been studied in detail in rodents (for review, see von Gersdorff and Borst, 2002; Schneggenburger and Forsythe, 2006; Kopp-Scheinpflug et al., 2011; Borst and Soria van Hoeve, 2012; Neher, 2015). The large number of release sites of a single calyx of Held (Meyer et al., 2001; Sätzler et al., 2002) supports the physiological one-to-one transmission to generate faithful suprathreshold information transfer during ongoing stimulations in vitro and in vivo (Kopp-Scheinpflug et al., 2003b; Hermann et al., 2007; Mc Laughlin et al., 2008; Lorteije et al., 2009; Stasiak et al., 2018; Kladisios et al., 2020). It is less clear how this synapse operates in other mammalian systems. In vivo data from cats (Guinan and Li, 1990; Mc Laughlin et al., 2008), gerbils (Kopp-Scheinpflug et al., 2003b), and guinea pigs (Stasiak et al., 2018) suggest that faithful transmission is the rule. However, biophysical data on synaptic transmission and postsynaptic input-output functions in mammals other than rodents, and especially for echolocating bats, are missing.
We used afferent fiber stimulation to quantify the calyx of Held inputs to MNTB neurons in Phyllostomus discolor. The large and fast EPSCs showed strong inward rectification (Fig. 3A). Under physiological extracellular calcium concentrations of 1.2 mm (Borst, 2010), the average EPSC size was 5.6 ± 1.05 nA (n = 8) recorded at −73 mV (Fig. 3A). The low rectification index of 0.28 ± 0.05, which was similar to the one recorded in gerbils (0.284 ± 0.03; Student's t test, p = 0.99), indicated a dominance of GluR4 subunits in the postsynaptic AMPA receptors (Washburn et al., 1997; Koike-Tani et al., 2005; Lujan et al., 2019; Franzen et al., 2020; Fig. 3B, left). The EPSCs decayed mono-exponentially with an average time course of 0.26 ± 0.03 ms (Fig. 3B, right). A single EPSC decay time course and the lack of an apparent second component indicated a mainly AMPAR-mediated EPSC, similar to mature rodents (Taschenberger and von Gersdorff, 2000; Joshi et al., 2004; Kladisios et al., 2020). This bat EPSC phenotype was significantly different in size (gerbil: −10.01 ± 1.33 nA at −73 mV; t test, p = 0.022) and decay kinetics (gerbil: 0.17 ± 0.01 ms, Student's t test, p = 0.021) compared with gerbils.
Calyx of Held synapses are smaller and depress less in bats, compared with gerbils. A, Left, exemplary current responses of calyceal EPSCs from a bat and gerbil MNTB neuron recorded from −93 to 67 mV with 10-mV voltage steps. Right, Average EPSC peaks of bats (blue) and gerbils (red), stimulated at step potentials between −93 and 67 mV. Gerbil EPSCs are comparatively larger, and both gerbil and bat neurons show inward rectifying currents at positive voltages. Data shown as average ± SEM (bats: n = 8; gerbils: n = 9). B, Left, Rectification index of MNTB neurons for inward rectifying AMPA currents, measured as the fraction of a projected straight line, fitted from −93 to −23 mV, and the recorded current at 47 mV (bats: filled circles, gerbils: open circles). Right, Fast EPSC decay time (τEPSC) of individual bat and gerbil MNTB neurons. Colored circles show average ± SEM. C, Bat calyx of Held was stimulated with different frequencies to study STP. Exemplary 20-pulse stimulation trains of 50, 300, and 500 Hz are shown. Scale bars: 2 nA, 50 ms (top), 10 ms (middle), 5 ms (bottom). D, STP, normalized to the first EPSC for train stimulations of frequencies up to 500 Hz for bats (filled circles), and up to 300 Hz for gerbils (open circles). Data shown as average with one-sided SEM. E, Top, paired-pulse ratio (PPR) of bats and gerbils demonstrates the facilitation of the second EPSC for high frequencies in bats. For gerbils, projection to 500-Hz PPR (dotted line) suggests that calyces may facilitate. Bottom, Steady state depression, shown as the average STP of the last three pulses. The dotted line shows the projected steady state depression at 500 Hz. Data shown as average ± SEM. F, Average EPSC peak of the first pulse (left) is shown for comparison to the steady state EPSCs size at 10- to 500-Hz stimulation trains (right) of bats and gerbils. The dotted line shows the projected steady state depressed EPSC of gerbils at 500 Hz. G, Synapses with larger initial EPSCs tend to lose their steady state intersynapse variability at high stimulation frequencies, probably because of depletion of superprimed synaptic vesicles. Exemplary large (open triangles) and small bat synapse (filled triangles) stimulated at 10 (left) and 500 Hz (right) show that at high frequencies, steady state intersynaptic variability decreases. Inset, Coefficeint of variation (CV) of steady state depression decreases in higher frequencies.
Mechanisms of STP have been studied in detail in the rodent calyx of Held synapse (for review, see von Gersdorff and Borst, 2002; Neher, 2015). In order to compare STP between bats and gerbils, we recorded 20-pulse trains of varying frequencies in physiological calcium concentration. The gerbil data were reanalyzed and replotted from our previous recordings, conducted under the same experimental conditions (Kladisios et al., 2020). In bats, both STD and STF were observed in a frequency-dependent manner (Fig. 3C). At high stimulation frequencies normalized bat EPSCs displayed STD that was led by initial facilitation, while in gerbils a more prominent STD prevailed (Fig. 3D). Analyzing the paired-pulse ratio of the EPSCs from the first two stimulation pulses showed facilitation of the second response at stimulation frequencies above 100 Hz in bats (Fig. 3E, top). Three, six and seven of all eight bat cells facilitated at 100, 300, and 500 Hz, respectively, and one cell showed marginal STF even at 50 Hz. In gerbils, only the linear extrapolation to a stimulation frequency of 500 Hz indicated a possible STF (Fig. 3E, top). In bat, the level of normalized steady state depression significantly decreased between 50 and 500 Hz (ANOVA with post hoc Tukey's test, F = 2.7352, p = 0.0443; 50 vs 500 Hz, p = 0.018; other frequency combinations, p > 0.05; Fig. 3E, bottom), while in gerbil a sustained significant reduction from the lowest to highest stimulation frequencies was detected (ANOVA with post hoc Tukey's test, F = 5.4584, p = 0.0027; 10 vs 300 Hz, p = 0.05; 10 vs 500 Hz, p = 0.0031; 50 vs 500 Hz, p = 0.016; 100 vs 300 Hz, p = 0.04; Fig. 3E, bottom). Finally, the normalized steady state depression was smaller in bats compared with gerbils (Student's t test; 10 Hz, p = 0.0012; 50 Hz, p = 0.0049; 100 Hz, p = 0.032; 300 Hz, p = 0.0089; 500 Hz, p = 0.0051).
The analysis of the STP indicated that the bat calyx of Held generated a smaller first EPSC that depressed less compared with gerbils. Therefore, we compared the absolute steady state currents. Steady state currents of bats and gerbils, derived from the last three pulses of the stimulation trains, were largely overlapping (Fig. 3F) and no significant differences were detected (Student's t test, p > 0.05 for all frequencies between 10 and 500 Hz). In bats, the steady state currents changed from a maximal −3.97 ± 0.45 nA current at 50 Hz to −3.03 ± 0.3 nA recorded at 500-Hz stimulation frequency. In gerbils, the maximal steady state current was −4.81 ± 0.74 at 10 Hz and dropped to −2.96 ± 0.36 nA recorded at 300-Hz stimulation frequency. When extrapolating the reduction of the steady state current in individual gerbil cells, an average of −2.37 ± 0.63 nA was estimated for a stimulation frequency of 500 Hz (Fig. 3F). Therefore, despite the smaller initial EPSC amplitude in bats, the lower STD results in similar steady state currents, with only minor differences.
At the calyx of Held, a small number of vesicles undergo superpriming, a process that augments the release probability of synaptic vesicles (Lee et al., 2013; Taschenberger et al., 2016). Therefore, superprimed vesicles may support rapid onset responses in vitro, since their high release probability mediates large EPSCs at the onset of stimulation. Taschenberger et al. (2016) argue that the steady state EPSC variability in MNTB synapses subsides when synapses are challenged with high-frequency stimulation, because of fast depletion of superprimed synaptic vesicles. We found that across individual bat synapses, depressed steady state currents become equal at high stimulation frequencies. Exemplified in Figure 3G, a 10-Hz stimulation train mediated differently sized steady state currents in small and large inputs. Yet, both synapses depress to the same steady state level at 500-Hz stimulation frequency. In addition, the coefficient of variation (CV) of the responses to the last three pulses of the train decreased with high-frequency stimulation (Fig. 3G, inset). Together, this points to a heterogeneity of frequency-dependent depression between calyx of Held synapses and hence indicates a substantial intersynaptic variability of release probabilities in bat, possibly because of superpriming at individual synapses.
Faithfulness of synaptically evoked input-output functions
The calyx of Held in Phyllostomus discolor showed different STP compared with Meriones unguiculatus. In order to elucidate the functional role of STP, input size and kinetics of the calyx of Held, we generated AMPA conductance templates (EPSGs) of 20 pulses at different frequencies (10, 50, 100, 300, 500, and 800 Hz) for both species, according to the synaptic parameters determined so far. For trains of 800 Hz and in case of gerbils also for 500 Hz, their STP progression was extrapolated from lower frequencies. The peak conductance of a single EPSC input in bats (80.92 nS) and gerbils (124.8 nS) corresponded with the measured values and the pulses decayed with the appropriate median τEPSC (Fig. 3A). To determine the influence of STP on action potential generation, EPSG templates were constructed with and without STP. Additionally, to determine the influence of conductance, we injected EPSGs with initial conductance amplitude equal to a single input, and equal to 10% above conductance threshold (Fig. 4A).
Calyx of Held synapses mediate reliable and accurate postsynaptic excitation. A, Conductance templates (EPSGs) that approximated kinetics and amplitudes of average bat and gerbil EPSCs, are injected in MNTB neurons of either species at excitatory strength resembling one input (top, gray trace), or at 10% above conductance threshold (top, black trace), and voltage response was recorded (bottom). Four sets of EPSGs of varying interstimulus frequencies between 10 and 800 Hz were presented 10 times to 22 bat and 23 gerbil MNTB neurons, with conductance intensities equal to a single calyx-input and near threshold (10% above threshold), as well as with or without STP. Scale bars: 20 nS (top) and 20 mV, 5 ms (bottom). B, Success rate of action potential generation according to template type and frequency for bats (blue) and gerbils (red). Results from EPSGs without STP are shown as triangles, and from EPSGs with STP as squares. EPSGs with initial conductance equal to one synaptic input are depicted as solid lines, and EPSGs near conductance threshold as dotted lines. Data shown as average ± SEM. C, Average success rates of action potential generation plotted against stimulus number for all frequencies, with or without STP of conductance templates equal to single inputs (left, solid lines), or close to conductance threshold (right, dotted lines). Data from bats are shown in the left as filled symbols, and from gerbils on the right side as open symbols. In case of equal success rates between frequencies, only the overlaying lower frequency is visible. D, Action potential latency of MNTB EPSCs of both bats (left) and gerbils (right) plotted against the stimulus number. Color and symbol coded as in C.
First, we averaged the success rate of all evoked action potentials within a stimulation frequency according to the type of EPSG template (Fig. 4B). For an EPSG intensity mimicking single input size and lacking STP the success rate dropped in gerbils at 800-Hz stimulation frequency but not in bats. Including STP led to a drop in success rate in bats from 500-Hz stimulation frequency and in gerbils from 300 Hz to higher stimulation frequencies onwards (Fig. 4B). In direct comparison a single input size with depression yields at 500- and 800-Hz stimulation frequencies only 2.1% and 12.4% failures in bats, but 37.7% and 71.4% in gerbils. To compare the fidelity of MNTB output generation between bats and gerbils for inputs close to threshold, we determined the stimulation frequencies that produced a success rate of 0.5. For EPSGs without STP, frequencies of 652 and 616 Hz (Student's t test, p = 0.085) were extracted for bats and gerbils, respectively. Conversely, for EPSGs including STP the 0.5 success rate frequencies were 359 and 168 Hz, respectively (Student's t test, p = 0.026; Fig. 4B).
To assess in more detail how neuronal output developed during the stimulation trains, the average success rate of suprathreshold events was plotted against the stimulus number (Fig. 4C). In bat, the stimulus templates with intensities equal to one input and without STP only produced failures for 800-Hz trains, and failure rate was increased when including STP. When EPSGs with a conductance close to threshold were injected, bat MNTB neurons failed more often, compared with a one-input stimulation intensity. For close to threshold templates that lack STP, a very rapid drop in success rate was observed during the stimulation train. When STP was included in these templates, the success rate dropped more gradually and strongly (Fig. 4C, left). When challenging gerbil MNTB neurons with gerbil EPSGs of one input intensity without STP, failures only occurred at 800 Hz. These failures however occurred earlier and were more numerous compared with bats. In gerbils, EPSGs of one input intensity including STP caused a gradual drop of success rate that again was larger compared with bats. Stimulations near threshold without STP led to a rapid reduction of success rate for frequencies above 300 Hz. Adding STP led to a gradual reduction in success rate for all tested frequencies (Fig. 4C, right). Overall, the stimulus number dependent loss of success rate appeared larger in gerbils compared with bats. Taken together, STP led to a gradual reduction in output success rate and bats could outperform gerbils when challenged with their own conductance templates. This discrepancy might be partially attributed to the differences in STP. Finally, action potential generation is secured more strongly and sustained longer during stimulations in bat MNTB neurons.
We investigated the temporal accuracy of action potential generation to the stimulation of the four EPSG paradigms. The action potential latency was calculated as the duration between template and action potential peak. Data from 500 and 800 Hz of EPSGs near threshold were not included in the analysis, because more than half of the stimuli failed to generate output (Fig. 4D). For bat MNTB neurons challenged with one input and near threshold sized EPSGs without STP, the average latency of the first action potential was 131.2 ± 12.3 and 345.2 ± 19.7 µs, respectively (Student's t test, p = 1.2e−11; n = 22). Thus, similar to somata of other auditory neurons (H. Yang and Xu-Friedman, 2009; Couchman et al., 2010; Ammer et al., 2012; Franzen et al., 2015; Cao and Oertel, 2017; Kladisios et al., 2020), biophysical stimulus intensity and shape have a significant effect on action potential threshold and latency. For both EPSG intensities (one input and near threshold), stimulation without STP led to a latency increase above 300 Hz in the first three stimulation pulses, and to a new steady state latency level ranging between 153 and 429 μs in a stimulation frequency-dependent manner (Fig. 4D, left top). This indicates that latency has a brief history-dependent adaptation that is based on the action potential generator itself, since the conductance size of each pulse was constant. The adaptation occurs mainly at high frequencies, indicative of a rapid recovery from its use-dependence.
As expected, by adding STP, the first action potential latencies remained unchanged (129 ± 8 and 338 ± 19 µs for stimulation of one input and near threshold, respectively, Student's t test, EPSGs with STP vs EPSGs without STP; one-input stimulation, p = 0.42; near threshold, p = 0.85). The latencies to the following stimulation pulses gradually increased for both intensities in a frequency-dependent manner (Fig. 4D, left bottom). Action potential latency showed the same pattern in gerbil MNTB neurons. At the beginning of the stimulus, first latencies averaged 199 ± 6 and 343 ± 14 µs for EPSGs without STP, and 207 ± 6 and 363 ± 13 µs for EPSGs with STP. For EPSGs without STP, the latency increased to a higher level within the first pulses that was maintained until the end of the 20-pulse stimulation, while for EPSGs with STP the latency gradually increased (Fig. 4D, right). Importantly, the first action potential latency was significantly longer in gerbils compared with bats when stimulated with intensities mimicking a single calyx input (p = 2.6e−11 for stimulations with STP and p = 4.3e−7 for stimulations without STP). Taken together, latency depended on the history of the spike generator and the synaptic conductance. Both gerbils and bats generate action potentials in respect to their own inputs with temporal precision, and bats appear to be temporally more precise than gerbils.
To compare suprathreshold postsynaptic responses in MNTB neurons from bat and gerbil more directly, we analyzed the average success rate and latency from voltage responses to stimulated peak conductance during trains with STP. Responses to all stimulations independent of stimulation frequency and initial EPSG size were pooled together, to obtain a peak conductance continuum (Fig. 5). Because a response to a single stimulation is either subthreshold or suprathreshold, the average success rate (Fig. 5A) was determined from binned success rate values of single cells (bats: n = 22, gerbils: n = 23). For this purpose, the peak conductance of every pulse was binned in 1-nS conductance increments. This analysis revealed that action potential generation was more sensitive to conductances in bats compared with gerbils and that the half value of a sigmoid fit was shifted leftwards by 11.8 nS (bats, xhalf = 20.498, rate = 4.93; gerbils, xhalf = 32.341, rate = 7.29). From the same data set, we fitted the dependence of the latency on the conductance by exponential functions (Fig. 5B). Larger conductances led to shorter latencies. Based on the similarities of the decay constants obtained from exponential fits (t = 27.8 nS for bats and t = 35. 7 nS for gerbils), we conclude that the conductance-latency correlation does not strongly deviate between those two species. Moreover, the exponential function indicated that the latency was longer in gerbils than in bats in comparable conductances. To assess the temporal precision of action potential generation the jitter was analyzed (Fig. 5C). The jitter was analyzed from the binned responses used in Figure 5A. The precision of action potential generation increased, as the jitter decreases over the conductance increase. The conductance dependent jitter decrease appeared faster in gerbils (decay constant 17.4 nS) compared with bats (34.3 nS); however, the steady state jitter reached at ∼55 nS equals between gerbils and bats and was ∼20 µs. Taken together, this analysis reveals that MNTB neurons in bats generate action potentials with lower conductance threshold, with short latencies and similar temporal precision.
MNTB neurons in bats are more sensitive and precise to simulated calyx of Held inputs compared with gerbils. A, Average success rate of bat (blue) and gerbil (red) MNTB neurons, plotted against binned peak conductance of 1 nS from templates without STP. Sigmoid functions fitted to bat and gerbil data. B, Single action potential latency values according to peak conductance from EPSGs with one input initial conductance (circles) or near threshold (squares). Blue and red colors indicate bat and gerbil data, respectively. Lines show exponential fits for bats and gerbils (bats: n = 22; gerbils: n = 23). C, Average jitter of bat (blue) and gerbil (red) MNTB neurons, plotted against binned peak conductances of 1 nS. Lines show exponential fits for bat and gerbil data.
Next, we comparatively investigated the absolute refractory period of MNTB principal neurons in bats and gerbils. To that end, we stimulated MNTB neurons with paired-pulse conductance templates of equal size and with interstimulus frequencies ranging between 10 and 2000 Hz. Each stimulus was repeated 10 times and the success rate of second action potential generation was calculated. The stimuli were scaled to represent an average EPSG according to each species (80.92 and 124.8 nS for bats and gerbils, respectively), and 10% above conductance threshold (Fig. 6). In Figure 6, the success rate averaged over all neurons is plotted as a function of the interstimulus frequencies. In bats, for a conductance of one input and paired-pulse frequencies up to 666 Hz, a pair of action potentials was always generated. At 1000 Hz, 92% of trials generated pairs of action potentials, and at an extreme of 2000-Hz stimulation frequency, 20.8% of the second pulses elicited an action potential. When both pulses were set to 10% above conductance threshold, all stimuli with paired-pulse frequencies up to 40 Hz generated pairs of action potentials. Up to 666 Hz, a gradual decrease of paired action potential generation to 60.2% was observed, and at 1000 Hz only 20.1% of paired pulses elicited two action potentials. At 2000 Hz, the success rate dropped to 0.8%. Additionally, we determined the cutoff frequency, where the success rate of the second pulse dropped below 75% for each neuron. We noted a significant reduction of this cutoff frequency between conductances of a single input and near threshold, which reached 1238 and 745 Hz, respectively (paired t test, p = 5.74e−6). In gerbils, the same phenotype was observed, however, with lower success rates (Fig. 6). The cutoff frequency for one-input stimulation intensity was 1029 Hz and for threshold intensity 653 Hz (paired t test, p = 1.04e−5). The cutoff frequency of single input intensities between bats and gerbils differed significantly (Student's t test, p = 0.0055), but remained similar when injected with conductances near threshold (Student's t test, p = 0.755). Overall, compared with gerbils, bat MNTB neurons showed a shorter absolute refractory period for paired-pulse stimulations.
MNTB neurons of bats integrate and fire with higher fidelity than gerbils. Inset, Paired-pulse EPSG templates with interstimulus frequencies from 10 to 2000 Hz were injected (top, gray) and the success rate of paired action potentials was estimated (bottom, black). Action potential (AP) success rate of the paired-pulse output for stimulation intensities near threshold (dotted lines) and for a single input size (solid line) are plotted versus interstimulus frequency for MNTB neurons of bats (blue) and gerbils (red). Data shown as average ± SEM (bats: n = 22; gerbils: n = 23).
Finally, we sought to compare the action potential generator of bat and gerbil MNTB neurons directly by crossing the species-specific conductance templates. The templates were composed of 100 pulses and both species-specific stimulation trains were injected to bat and gerbil MNTB neurons. To that end, new templates lacking STP dynamics were constructed, with a constant peak conductance size adjusted to match the steady state of the frequency-dependent STD recorded before, for both species (Fig. 3F). In Figure 7A, a 500-Hz gerbil derived EPSG is injected in both gerbil and bat MNTB neurons, and we observed that gerbils were prone to more failures (Fig. 7A). Overall, templates derived from bat EPSG templates led to lesser frequency-dependent loss of action potentials compared with gerbil templates (Fig. 7B, compare left and right), which might be attributed to the minor EPSC differences in steady state depression at high stimulation frequencies (Fig. 3F). Comparing the same templates in both species showed that gerbil MNTB neurons generated fewer action potentials at frequencies above 500 Hz compared with bats (Fig. 7B). This difference was significant for 800-Hz stimulation frequencies (Student's t test, bats vs gerbils with bat templates, 500 Hz, p = 0.683, 800 Hz, p = 0.0149; bats vs gerbils with gerbil templates, 500 Hz, p = 0.139, 800 Hz, p = 0.0122). Thus, bat MNTB neurons outperformed gerbil MNTB neurons in following high frequency input-output functions.
Crossover comparison of output generation and temporal fidelity between bat and gerbil. A, 100-pulse EPSG trains of interstimulus frequencies ranging between 10 and 800 Hz, and with a conductance maximum corresponding to the respective frequency's steady state (Fig. 3) were created from both bat and gerbil data. Here, a 500-Hz template corresponding to gerbil steady state conductance (top, red trace) was injected to gerbil (middle) and bat (bottom) principal MNTB neurons and their action potential success rate and latency were extracted. B, Number of evoked action potentials in bat (filled circles) and gerbil (open circles) MNTB neurons, in response to injection of bat (blue) and gerbil (red) EPSG trains, as a function of stimulation frequency. Data shown as average and one-sided SEM (bats: n = 12; gerbils: n = 17). C, Average action potential latency values of the 100-pulse EPSG trains for bats and gerbils. Legends as in B. D, Average jitter of action potentials of the EPSG trains for bats and gerbils. Legend as in B.
As bats appeared to outperform gerbils in output generation, we additionally analyzed whether the action potential latencies differ between the two species. Summing all latencies across the stimulation train showed that latency generally increases with higher stimulation frequencies (Fig. 7C). When challenging MNTB neurons with bat derived EPSG templates, a nearly significant shift was observed at 500-Hz stimulation frequency (Student's t test, p = 0.07), where bats showed shorter latencies compared with gerbils (Fig. 7C). When challenging neurons with gerbil derived EPSGs, the latency became significantly larger in gerbils compared with bats from 500-Hz stimulation frequency and above (Student's t test, 500 Hz, p = 0.056, 800 Hz, p = 0.0044, Fig. 7C). This analysis corroborates the findings from short train stimulations presented in Figure 5. Temporal precision was further analyzed as the jitter of action potential generation throughout the trains. Overall, the jitter appeared similar for the suprathreshold output between gerbil and bat MNTB neurons. Only at 500-Hz stimulation frequency was the jitter significantly smaller in bats (Student's t test, bat EPSGs: 500 Hz, p = 0.017, 800 Hz, p = 0.21; gerbil EPSGs: 500 Hz, p = 0.109, 800 Hz, p = 0.293; Fig. 7D). Taken together, bat MNTB neurons followed high frequency stimulations more efficiently and with comparable temporal precision compared with gerbils.
We described a species-specific reduction in temporal precision and output fidelity during EPSG stimulation trains that depends on the action potential generator. A cellular mechanism that might explain this species-specific difference is sodium current inactivation. We therefore probed sodium current inactivation in gerbils and bats. Sodium current inactivation was recorded at room temperature by evoking pharmacologically isolated sodium currents at –33 mV after a prestep potential between –113 to –38 mV (Fig. 8A). In gerbils, the range of sodium current inactivation occurred between −78 and −43 mV (Fig. 8B) with an average half inactivation at −58.4 ± 1 mV (n = 9). Sodium inactivation of bat principal MNTB neurons was indistinguishable to those of gerbils with half inactivation at −58.6 ± 1.3 mV (n = 4). To further characterize sodium currents, we determined the recovery from inactivation (Fig. 8C) at four different holding potentials. Recovery from inactivation of gerbil MNTB neurons followed a biexponential time course that depended on the holding potential (Fig. 8D). At a holding potential of −68 mV, recovery time constants were 4 and 144 ms for the fast and slow components, respectively. At −73 mV the fast and slow time constants were 3 and 19 ms and at −93 mV they decreased to 1 and 10 ms. Finally, at −113 mV, recovery time constants from inactivation were 0.5 and 8 ms (Fig. 8D). In bat MNTB neurons the recovery from inactivation of sodium current for holding potentials of −68 (3 and 23 ms) and −73 mV (2 and 8 ms; n = 2) indicated a slightly faster time course (Fig. 8D). To elucidate whether sodium current inactivation is relevant for stimulation trains and not just a product of long voltage steps, we induced sodium currents with 0.5-ms short steps in a 15-pulse train at various frequencies (Fig. 8E). When holding the gerbil neurons at −73 mV during the step train, the sodium current peak amplitude showed a strong use-dependent and frequency-dependent inactivation (Fig. 8F, n = 6). The amount of the inactivation at the end of the step train depended significantly on the holding potential between the sodium evoking voltage steps (ANOVA, −68 mV: F = 4.5108, p = 0.0293, −73 mV: F = 4.7252, p = 0.0256, −93 mV: F = 2.4523, p = 0.1198, −113 mV: F = 1.1735, p = 0.3361; Fig. 8G). Also, the amount of normalized steady state current depended significantly on the stimulation frequency (paired t test, p < 0.05 for 100, 300, and 500 Hz). In one bat neuron, we were able to obtain the same recordings. This neuron showed a lower inactivation during this stimulation paradigm compared with gerbils (Fig. 8G). Taken together, sodium current inactivation is a suitable mechanism to explain the loss of temporal precision of the spike generator during stimulation trains, because of its strong history, frequency, and holding potential dependence. Moreover, the evidence of the species-dependent recovery from inactivation and the steady state inactivation during trains is an ideal candidate to explain the difference in temporal precision and fidelity of the spike generator in MNTB neurons.
Sodium channel inactivation in MNTB principal neurons. A, Bottom, Example recoding of sodium current inactivation at −33 mV with preceding step potentials between −113 and −38 mV incremented in 5-mV steps. Top, Voltage command protocol. B, Inactivation of sodium currents expressed as I/Imax. Blue symbols show average values from bat, and red symbols from gerbil MNTB principal neurons. C, Bottom, Example recording of sodium current recovery from inactivation. Paired sodium currents were evoked at the onset of a 30-ms depolarizing step (−33 mV), as well as after a hyperpolarizing step with varying holding voltage (−68, −73, −93, and −113 mV) and interval (0.5–33.8 ms) to −33 mV. Top illustrates voltage command protocol with an arrow that encompasses the step interval range. D, Sodium current recovery from inactivation expressed as I/I0, where I0 equals the peak of the first evoked current in the pair. Recovery evoked after −68, −73, −93, and −113 mV is shown with gray, black, cyan, and brown symbols, respectively. Open symbols show average ± SEM from eight gerbil, and closed symbols from two bat MNTB neurons. For clarity, inset shows a magnified recovery after −68 and −73 mV, where bat neurons recover faster. E, Sodium currents in response to train stimulations at 100 Hz (left) and 500 Hz (right). Single current responses were evoked with steps to –68 mV for 0.5 ms. F, Normalized gerbil sodium current (I/I1) as a function of step number of train stimulations at 100 Hz (square), 300 Hz (circles), and 500 Hz (diamonds). Holding potential between steps was −73 mV, n = 6. G, Steady state sodium current inactivation during train stimulations at different frequencies recorded at different holding potentials (gray, black, cyan and pink symbols represent −68, −73, −93, and −113 mV), averaged from six gerbil MNTB neurons (open symbols). Trains obtained from a single bat MNTB neuron are plotted as closed symbols.
Discussion
In order to comparatively characterize membrane and synaptic properties of principal MNTB neurons and to assess their input-output functions, we have applied current, voltage and dynamic clamp recordings in acute brain slices of bat (Phyllostomus discolor) and gerbil (Meriones unguiculatus). We have shown that the basic resting membrane biophysics of MNTB neurons are largely stereotypic between mammals, while whole-cell currents of voltage-gated ion channels, the synaptic transmission and the action potential generator are species dependent. Thereby, the maintenance of high-frequency input-output functions is more prominent in bats compared with gerbils.
Comparative biophysical features of MNTB neurons
The membrane properties at resting levels of bat MNTB neurons were similar to those of gerbils, and matched with other species (Wu and Kelly, 1991; Banks and Smith, 1992). Also, the membrane time constant, relevant for synaptic integration, is similar between bats and gerbils. Only the action potential voltage threshold was lower in bats. The resulting difference in voltage that needs to be crossed for evoking suprathreshold output was smaller in bats, indicating that bat MNTB neurons are easier excited to a suprathreshold level than gerbils. The somatic firing behavior of bat MNTB neurons matches other species (Banks and Smith, 1992; Brew and Forsythe, 1995; Kladisios et al., 2020) and all tested neurons responded with a single action potential to a square pulse current injection. The overall action potential waveform, did not differ between gerbil and bat MNTB neurons indicating that repolarization is identical. This similarity suggests that high voltage-activated Kv3 currents, relevant for repolarizations (Kaczmarek et al., 2005; Leão, 2019) are similar in both species. Again similar to rodents (Brew and Forsythe, 1995; Dodson et al., 2002), somatic Kv1.1 channels are present in bats. The activation voltage of the DTX-sensitive current matched between both species, and with rats (Dodson et al., 2002) and mice (Gittelman and Tempel, 2006). Therefore, we propose that during resting state the general biophysical properties of MNTB neurons are similar across mammalian species, yet variations exist that render bat neurons more excitable. One variation is the different size of the DTX-sensitive potassium current. The reduced DTX-sensitive current in bats likely underlies the differences in rectification of steady state voltage responses to square depolarizations. Also lower amounts of DTX-sensitive currents might influence the spike generator, as they are thought to reduce temporal fidelity (Kopp-Scheinpflug et al., 2003a; Gittelman and Tempel, 2006). However, our bat data do not indicate lower temporal precision. Therefore, we postulate that the change in sodium current, or morphologic structure might counteract the loss of temporal precision during ongoing activity that might stem from lower amounts of DTX-sensitive currents.
Comparative synaptic physiology of MNTB neurons
The strong and fast excitatory input provided by the calyx of Held is the synaptic hallmark of the MNTB and is also present in bats, with distinct differences from gerbils. The size of the EPSC, generated by the bat calyx of Held under resting conditions is significantly smaller and slower compared with the EPSC in gerbils. AMPA current properties, like reversal potential and rectification, are the same between bats and other rodents (Taschenberger and von Gersdorff, 2000; Joshi et al., 2004). In agreement with recordings from adult gerbils under physiological conditions (Kladisios et al., 2020), we could not detect a second synaptic component indicative of an electrogenic NMDAR-mediated current in bats. This is different to recordings from mice and rats, where small NMDAR-mediated currents could be at least detected until postnatal day 35 in elevated extracellular calcium concentrations (Steinert et al., 2010).
Our direct comparison demonstrated differences in STP between bats and gerbils. Almost every bat calyx of Held input initially facilitated at frequencies above 100 Hz followed by depression. In comparison, gerbil inputs depressed from the beginning of the stimulation train and the depression was more severe. Together with the different initial EPSC size, the synaptic depression profiles indicate that bat calyx of Held terminals release vesicles with a lower probability. This lower release probability might also be one mechanism, as to why during steady state depression the input size between gerbils and bats leveled at similar sizes, despite the different onset EPSC amplitudes. The spontaneous activity in vivo (Smith et al., 1998; Kopp-Scheinpflug et al., 2003b; Kadner et al., 2006; Hermann et al., 2007) might indicate that the steady state depression values below 100 Hz represent the physiologically relevant EPSC size. In this case, the differences in the input size delivered by the calyx of Held are minor between gerbils and bats. Nevertheless, the absolute steady state currents during STD indicate that in bats, fewer frequency-dependent alterations are present.
Comparative physiology of MNTB input-output functions
To understand the physiologically generated input-output functions of neurons, it is crucial to take the action potential generator and the synaptic input phenotype into account. Using dynamic clamp we could decompose the effects of the synaptic STP and assign functional differences between the two species. In both species, the action potential generation and precision depends strongly on the presence of STP, similar to endbulb synapses (H. Yang and Xu-Friedman, 2009; Kladisios et al., 2022). This effect is based on the change in conductance during the decrease of synaptic strength. Besides the influence of synaptic STP, the action potential generator also limits output generation in a frequency-dependent manner. In addition to limiting the output, the action potential generator influences temporal accuracy, since an increase in latency is observed at the beginning of a stimulation train without STP. This effect becomes apparent for EPSG stimulations without depression and at full calyx intensity at frequencies of 500 Hz in bats and 300 Hz in gerbils. To explain this increase in latency, investigation of sodium current inactivation during step trains is mechanistically ideal. Because the action potential generator maintains the latency after the initial reduction, it might be relevant that the calyx of Held fires spontaneously. This spontaneous activity might prevent the action potential generator to fully recover. Thereby, this ongoing activity could help to stabilize the temporal accuracy, not only by limiting STP, but also by maintaining the adaptation level of the action potential generator.
Our experiments demonstrate that bat neurons are intrinsically more suited to sustain higher output rates with shorter latencies compared with gerbils. Both parameters indicate a higher excitability in bat neurons. However, bats show the same temporal precision in suprathreshold output generation compared with gerbils, since the action potential jitter was largely similar. This similarity indicates that in bats temporal precision is also relevant, since it matches the gerbil circuit that is tuned to interaural time difference detection and therefore especially relying on temporal precision. Thus, bat MNTB neurons have found a cellular mechanism to accommodate both, sustained high frequency output and temporal precision. In contrast gerbils are relying specifically on temporal precision. We can suggest underlying mechanisms for these functional differences. One explanation is the smaller voltage difference from the resting potential to the action potential voltage threshold in bats compared with gerbils, which allows bats to fire with higher repetition rates. Moreover, the difference in low voltage-activated Kv1.1 currents quenches excitation in MNTB neurons of gerbils more efficiently. Yet, the presence and similar voltage dependence of the Kv1.1 in bats and gerbils appears sufficient for both species to accommodate high temporal precision in suprathreshold output generation. Lastly, the sodium channels of the action potential generator appear to inactivate during high frequency activation in a lesser degree in bats compared with gerbils. Indeed, we present evidence that the time to, and the recovery from inactivation of sodium currents are different between gerbils and bats.
The functional significance of the difference in sustaining higher output rates with shorter latencies in bats is speculative. One view might be in the context of echolocation that requires faithful and strong ongoing inhibitory suppression to tag the end of the sound in respect to next incoming sounds. The sustainability of higher firing rates in the bat MNTB could thereby directly or indirectly contribute to precise delay measurements in higher brain areas, like the inferior colliculus or the thalamus. Also the proposed object recognition and binaural filtering in the bat's superior olivary circuitry (Grothe, 1994; Grothe and Park, 2000) might require both, high output rates with short latencies and temporal precision. Moreover, that temporal precision plays a role in bat sound source localization via the lateral superior olive was demonstrated before (Park et al., 1996). Alternatively, the convergence of MNTB axons onto single target neurons might be lower in bats compared with gerbils and therefore, each output neuron must be more reliable. Overall, our cellular data highlight species-specific adaptations within an evolutionary conserved circuit.
Our comparative approach highlights various differences that would most likely be overlooked by investigating a single species. The importance of comparative neurophysiology has again recently been demonstrated in the cortex (Beaulieu-Laroche et al., 2021), MNTB (Kopp-Scheinpflug et al., 2008), and lateral superior olive (Walcher et al., 2011). Moreover, it is important to continue the use of nonmodel organisms in comparison to standard model-species, to obtain deeper insights into the mechanistic function of neuronal circuits.
Footnotes
This work was supported by Grants DFG FE789/7-1, DFG FE789/8-1 and DFG FE789/12-1. We thank Dr. Elisabeth M. M. Meyer for technical help with the conductance clamp setup. We also thank Dr. Karl-Heinz Esser for introduction to the animal system Phyllostomus discolor, Sönke von den Berg for animal handling, and Alexandra Benn for technical support with the immunofluorescence. We thank Prof. Dr. Christian Leibold for comments and discussion.
The authors declare no competing financial interests.
- Correspondence should be addressed to Prof. Felix Felmy at felix.felmy{at}tiho-hannover.de