Passive and active membrane properties determine the voltage responses of neurons. Within the auditory brain stem, refinements in these intrinsic properties during late postnatal development usually generate short integration times and precise action-potential generation. This developmentally acquired temporal precision is crucial for auditory signal processing. How the interactions of these intrinsic properties develop in concert to enable auditory neurons to transfer information with high temporal precision has not yet been elucidated in detail. Here, we show how the developmental interaction of intrinsic membrane parameters generates high firing precision. We performed in vitro recordings from neurons of postnatal days 9–28 in the ventral nucleus of the lateral lemniscus of Mongolian gerbils, an auditory brain stem structure that converts excitatory to inhibitory information with high temporal precision. During this developmental period, the input resistance and capacitance decrease, and action potentials acquire faster kinetics and enhanced precision. Depending on the stimulation time course, the input resistance and capacitance contribute differentially to action-potential thresholds. The decrease in input resistance, however, is sufficient to explain the enhanced action-potential precision. Alterations in passive membrane properties also interact with a developmental change in potassium currents to generate the emergence of the mature firing pattern, characteristic of coincidence-detector neurons. Cholinergic receptor-mediated depolarizations further modulate this intrinsic excitability profile by eliciting changes in the threshold and firing pattern, irrespective of the developmental stage. Thus our findings reveal how intrinsic membrane properties interact developmentally to promote temporally precise information processing.
- ventral nucleus of the lateral lemniscus
- postnatal development
- neuronal excitability
- cholinergic modulation
the ability of a neuron to generate action potentials with high temporal precision depends on the interaction of passive and active membrane properties (Ammer et al. 2012; Axmacher and Miles 2004; Berntson and Walmsley 2008; Fricker and Miles 2000; Gittelman and Tempel 2006; Kuba et al. 2002). Furthermore, these intrinsic membrane properties define the firing behavior of neurons, as different conductance states support, for example, either coincidence-detector or integrator-operating modes (Ratte et al. 2013). Intrinsic membrane properties depend, in turn, on leak currents (Berntson and Walmsley 2008; Goldstein et al. 2001), cell morphology (Mainen and Sejnowski 1996), and voltage-gated channels (Fricker and Miles 2000; Gittelman and Tempel 2006; Johnston et al. 2010) and are subject to neuromodulatory influences.
Changes in passive and active membrane properties, as well as in neuronal morphology, are widespread among neurons in auditory brain stem nuclei during late postnatal development. Across these nuclei, a reduction in input resistance generates shorter membrane time constants, promoting fast and temporally precise integration (Ahuja and Wu 2000; Ammer et al. 2012; Kandler and Friauf 1995; Kuba et al. 2002; Scott et al. 2005; Wu and Oertel 1987). A reduction in dendritic arborization (Rautenberg et al. 2009; Rietzel and Friauf 1998; Rogowski and Feng 1981; Sanes et al. 1992) decreases the cell capacitance, also leading to faster voltage signaling. Additional changes in active membrane properties result in enhanced excitability, as indicated by the shape and threshold of action potentials (Ahuja and Wu 2000; Ammer et al. 2012; Scott et al. 2005; Taschenberger and von Gersdorff 2000). Thus the regulation of these features during late postnatal development tunes the response properties of neurons in the auditory brain stem.
The ventral nucleus of the lateral lemniscus (VNLL) relays auditory information with exceptional temporal precision, particularly at the onset of sounds (Adams 1997; Covey and Casseday 1991; Recio-Spinoso and Joris 2014; Zhang and Kelly 2006). This temporal precision is attributed to globular cells in the ventral region of the adult rodent and cat VNLL (Adams 1997; Recio-Spinoso and Joris 2014; Zhang and Kelly 2006) and in the columnar region of bats (Covey and Casseday 1991). Around hearing onset, these neurons typically generate onset action potentials in response to prolonged current injections in rats (Zhao and Wu 2001). However, onset firing is lacking at threshold in this cell type before hearing onset in gerbils (Berger et al. 2014), suggesting a change in the firing pattern of globular cells during late postnatal development. Consistent with the mature onset firing pattern, the expression of Kv1.1, a low voltage-activated potassium channel, has been reported in adult bats (Rosenberger et al. 2003). At the circuit level, the temporal fidelity of globular VNLL neurons is critical to generate a prominent, feed-forward onset inhibition to the inferior colliculus (Nayagam et al. 2005; Pollak et al. 2011). Thus globular VNLL neurons constitute an ideal system to evaluate the developing contribution of intrinsic properties to precise action-potential generation. Therefore, we characterized the developmental profile of intrinsic properties of globular VNLL neurons from before hearing onset to maturity to unravel their contribution to action-potential generation and precision.
Our findings reveal that the suprathreshold responsiveness of VNLL neurons emerges from the specific interaction between passive and active membrane properties during late postnatal development. Notably, these alterations generate the emergence of coincidence-detector properties (Ratte et al. 2013), which are crucial to relay auditory information with high temporal precision.
MATERIALS AND METHODS
All experiments complied with institutional guidelines and national and regional laws. Animal protocols were reviewed and approved by the Regierung of Oberbayern (according to the Deutsches Tierschutzgesetz). Mongolian gerbils (Meriones unguiculatus), of either sex and of postnatal day 9 (P9)–P28, were anesthetized with isoflurane and then decapitated. This age range was chosen to evaluate the late postnatal refinement before and after the onset of hearing, which occurs at P12–P13 in gerbils (Finck et al. 1972; Ryan et al. 1982; Smith and Kraus 1987). Brains were removed in dissection solution containing (in mM) 50–120 sucrose, 25 NaCl, 27 NaHCO3, 2.5 KCl, 1.25 NaH2PO4, 3 MgCl2, 0.1 CaCl2, 25 glucose, 0.4 ascorbic acid, 3 myoinositol, and 2 Na-pyruvate (pH was 7.4 when bubbled with 95% O2 and 5% CO2), and 180–200 μm-thick transverse slices containing the VNLL were cut with a VT1200S Vibratome (Leica Microsystems GmbH, Wetzlar, Germany). Slices were incubated for 45 min at 34.5°C in extracellular recording solution (same as dissection solution but with 125 mM NaCl, no sucrose, 1.2 mM CaCl2, and 1 mM MgCl2). All recordings were carried out at near-physiological temperature (34–36°C).
Globular VNLL neurons were visualized and imaged with a 60 × 1 numerical aperture (NA) objective with a microscope (BX51WI or BX50WI; Olympus, Center Valley, PA) equipped with gradient contrast illumination, a 1.4-NA oil-immersion condenser, and a TILL Photonics imaging system (FEI Munich GmbH, Munich, Germany) composed of a charge-coupled device (TILL-Imago VGA or Retiga 2000DC) camera and a monochromator (Polychrome IV or V). Recordings were performed using an EPC 10/2 amplifier [HEKA Elektronik, Lambrecht (Pfalz), Germany]. Data were acquired at 100 kHz for current-clamp and 50 kHz for voltage-clamp recordings and filtered at 3 kHz. In whole-cell, current-clamp recordings, the bridge balance was set to 100% after estimation of the series resistance and was monitored repeatedly during recordings. The series resistance during whole-cell, voltage-clamp recordings was compensated to a constant residual of 2–3 MΩ. The liquid junction potential (LJP) was not corrected for any of the solutions. For current-clamp experiments, the internal recording solution consisted of (in mM) 145 K-gluconate, 4.5 KCl, 15 HEPES, 2 Mg-ATP, 2 K-ATP, 0.3 Na2-GTP, 7.5 Na2-phosphocreatine, 5 K-EGTA, and 20–50 μM Alexa Fluor 568 (pH adjusted with KOH to 7.3; LJP ∼16 mV). To reduce the driving force during recordings of whole-cell potassium currents, the potassium reversal potential was elevated to −71.5 mV by the following internal solution (in mM): 65 K-gluconate, 80 Na-gluconate, 4.5 KCl, 15 HEPES, 2 Mg-ATP, 2 K-ATP, 0.3 Na2-GTP, 7.5 Na2-phosphocreatine, 5 Na-EGTA, and 20–50 μM Alexa Fluor 568 (pH adjusted with NaOH to 7.3; LJP ∼14 mV). In addition, the external potassium concentration was raised to 5 mM and the calcium concentration reduced to 1 mM. Potassium currents were isolated in the presence of 1 μM TTX, 100 μM Cd2+, 50 μM ZD 7288, 10 μM SR95531, 0.5 μM Strychnine, 20 μM 6,7-dinitroquinoxaline-2,3-dione, and 50 μM D-2-amino-5-phosphonopentanoate. The reduction of the extracellular calcium concentration, together with the use of micromolar concentration of Cd2+, largely prevented precipitation. The cell compartment, type, and location were morphologically verified for all recordings.
Conductance-clamp recordings were made with a SM-1 conductance-clamp amplifier (Cambridge Conductance, Royston, UK) with added linear conductances. The reversal potential of the leak current was set to the resting membrane potential of each cell. Leak conductances, ranging from 2 to 20 nS, were applied to evaluate the impact of decreasing the input resistance on cellular properties in globular VNLL neurons of P9 and >P25 animals. The range of applied background conductances was based on the developmental change of the input resistance observed in these neurons.
Cholinergic responses were evoked by local application of carbachol (500 μM) and oxotremorine (0.5 or 1 mM), dissolved in extracellular recording solution, and loaded into a patch-pipette connected to a dispense system (Picospritzer III; Parker, Cleveland, OH). Pipettes of equal size were consistently placed ∼200 μm from the soma of the recorded neuron. The agonist was applied using a 100-ms pulse at 15–20 lb./in2. The resting membrane potential was recorded for ∼2.5 min, following agonist application to test for recovery. Changes in the current threshold and the firing pattern induced by cholinergic agonists were evaluated using a short (0.5 ms) and a long (200 ms) current injection before and after agonist application.
Immunohistochemistry was performed on brain tissue from animals aged P9 and P26. Animals were anesthetized using 200 mg/kg body wt pentobarbital (Narcoren; Merial GmbH, Hallbergmoos, Germany) via intraperitoneal injection. The animals were perfused transcardially with Ringer solution, supplemented with 0.1% heparin (Meditech Vertriebs GmbH, Parchim, Germany) for 5 min, followed by 4% paraformaldehyde for 30 min. The brains were removed and postfixed overnight at 4°C. Brains were washed three times in PBS, and coronal brain stem slices of 50–60 μm thickness were taken using a VT1200S Vibratome (Leica Microsystems GmbH). Standard immunohistochemical procedures were performed after 1 h incubation in blocking solution (0.3% Triton, 0.1% saponin, 1% BSA in PBS) on free-floating slices using the anti-microtubule-associated protein 2 (MAP2) antibody (chicken polyclonal; 1:1,000; Neuromics, Edina, MN). The secondary antibody was applied after 2 days of incubation for 3 h at room temperature and was conjugated with aminomethylcoumarin (anti-chicken; 1:100; Dianova, Hamburg, Germany). Finally, the slices were mounted in Vectashield medium (H-100; Vector Laboratories, AXXORA, Enzo Life Sciences GmbH, Lörrach, Germany).
Data analysis and statistics.
The cell diameter of VNLL neurons was analyzed in ImageJ software on confocal image stacks with a voxel size of 320 × 320 × 335 nm. These images were acquired with a 63× objective (1.32 NA) and a 1.5 zoom on a Leica SP5 system. To obtain an unbiased estimate of the cell diameter, the maximal length and width were measured from maximal projections of all imaged neurons in the ventral VNLL, irrespective of a specific morphological subpopulation. Electrophysiological data were analyzed with custom-written IGOR Pro procedures (WaveMetrics, Lake Oswego, OR) and in Microsoft Excel. Results are reported as mean ± SE. Statistical significance was determined with either a paired or unpaired Student's t-test in Excel, and the significance level was defined as P < 0.05. Linear correlation analysis was performed using Pearson's r in Prism.
Here, we characterized the developmental changes in passive and active membrane properties, as well as in the firing pattern of visually identified, globular VNLL neurons from P9 to P28. Specifically, we addressed how the developing interaction between these membrane properties provides mature VNLL neurons with temporally precise responses characteristic of coincidence-detector neurons. Finally, we examined how depolarizations induced by endogenous cholinergic modulation further promote such response properties.
Development of the resting membrane properties.
Voltage signaling in neurons depends on both passive and active membrane properties. To gain insight into the developmental regulation of voltage signaling in VNLL neurons, we first estimated their passive membrane properties, namely, the resting potential, membrane time constant, resting input resistance, and effective capacitance. These membrane parameters were analyzed from the average voltage response to 40–60 repetitions of 200 ms-long step current injections of −10 pA (Fig. 1A; n = 75) (Ammer et al. 2012; Berger et al. 2014; Couchman et al. 2010; Porres et al. 2011). The use of such a small current amplitude kept the resulting voltage deflection minimal and hence minimized the activation or deactivation of additional voltage-dependent conductances close to the neuron's resting state. These current injections generated voltage responses within the range of the observed resting voltage fluctuations of a cell (Fig. 1A). Therefore, this paradigm allows the passive membrane properties of these neurons to be approximated at rest. Among the intrinsic membrane parameters at rest, only the membrane potential remained similar between P9 (n = 11) and ≥P25 (n = 7; P9: −64.6 ± 1.0 mV; ≥P25: −63.9 ± 1.1 mV, P > 0.05; Fig. 1B). In contrast, the membrane time constant and the resting input resistance decreased significantly from 13.1 ± 1.0 ms and 254.6 ± 30.9 MΩ to 3.6 ± 0.5 ms and 108.3 ± 13.7 MΩ, respectively, between P9 and ≥P25 (all P < 0.05; Fig. 1B). As a result, the estimated effective resting capacitance decreased by 40% from 55.2 ± 5.5 to 33.8 ± 2.5 pF (P < 0.05) over this developmental period (Fig. 1B). This reduction in the effective capacitance could result from morphological alterations or a reduction in spatial charging efficacy, resulting from a decrease in input resistance.
To ascertain whether anatomical changes occur in the VNLL during postnatal development, we quantified the cell size of VNLL neurons from MAP2 stainings at P9 and P26 (Fig. 1, C–E). The average soma diameter decreased significantly from 15.6 ± 0.3 μm at P9 (n = 52; three sections from three animals) to 13.9 ± 0.4 μm at P26 (n = 31; three sections from three animals, P < 0.05; Fig. 1E). The calculation of a spherical surface from the measured soma diameter leads to an estimated reduction in soma size of ∼20%. Thus the developmental reduction in soma diameter explains ∼50% of the decrease in effective capacitance. The remaining reduction in the effective capacitance is likely due to a refinement in the dendritic arbor, a process known to occur in the auditory brain stem (Rautenberg et al. 2009; Rietzel and Friauf 1998; Rogowski and Feng 1981; Sanes et al. 1992), and a reduction in spatial charging efficacy. Importantly, the refinement of passive membrane properties occurs predominantly before, but continues for a short time after hearing onset, as illustrated by the exponential fits (decay time of 3.6 days for the membrane time constant, 4.5 days for the resting input resistance, and 3.5 days for the estimated capacitance; Fig. 1B). Taken together, changes in resting membrane properties, such as the input resistance and the cell capacitance, effectively decrease the membrane time constant, thus shortening the somatic integration time window of VNLL neurons during late postnatal development.
Development of firing properties.
Short integration times are characteristic features of an operating mode of coincidence detection (Ratte et al. 2013), well suited to achieve the temporal precision required for auditory processing. However, the late and sustained firing pattern of VNLL neurons in juvenile gerbils is inconsistent with this operating mode (Berger et al. 2014). To investigate whether the refinement in passive membrane properties is accompanied by a change in operating mode, we characterized the firing pattern of VNLL neurons in response to 500 ms-long current injections at different stages of late postnatal development (n = 75; Fig. 2). In P9 animals, these neurons responded with continuous firing after an initial silent period (Fig. 2A). In P13 animals, the threshold current generated an action potential at stimulus onset. Further increasing the current amplitude led to an action potential at stimulation onset, followed by continuous firing (Fig. 2A). This firing pattern is also found in P12–P16 rat VNLL neurons (Zhao and Wu 2001). Finally, at P26, only the action potential at stimulation onset remained (Fig. 2A). The current threshold in response to these long injections increased during late postnatal development (P9: 0.28 ± 0.04 nA; ≥P25: 0.42 ± 0.04 nA, P < 0.05; Fig. 2B). As the firing pattern changed, the maximal number of action potentials increased between P9 (78.6 ± 7.9, n = 11) and ∼P13 (108.2 ± 13.4, n = 6) before rapidly declining to three action potentials maximally, from ≥P25 onward (1.8 ± 0.3, n = 7; Fig. 2B). The time to the first elicited action potential also decreased from 172.8 ± 28.6 ms to 1.6 ± 0.1 ms (Fig. 2B), between P9 and ≥P25. This decrease was not limited to the switch from continuous firing to an onset response but continued between P13 and P28 for the action potentials at stimulation onset (Fig. 2B). The developmental change in firing pattern and particularly, the generation of a single spike at stimulation onset are consistent with the emergence of intrinsic coincidence-detector properties (Ratte et al. 2013).
Next, we determined the relative contribution of the effective capacitance and the resting input resistance to the change in the firing properties of VNLL neurons during postnatal development. During long current injections, cellular voltage responses reach steady state, and therefore, the input resistance should dominate the neuronal output. Indeed, in response to these long current injections, the current threshold correlated significantly with the input resistance (r = −0.5069, P < 0.0001) but not with the cell capacitance (r = −0.06653, P = 0.057; Fig. 3A). To probe whether the input resistance can directly influence the firing behavior of VNLL neurons, we performed conductance-clamp recordings in juvenile animals. In these experiments, the input resistance of recorded cells was artificially lowered by applying a background leak conductance. We then quantitatively characterized the input-output functions of P9 neurons in response to 200 ms current injections (Fig. 3B). The input resistance during the application of the additional background conductance was estimated at rest, as for control conditions, illustrated in Fig. 1A. At decreased input resistances, P9 cells now generated an action potential at stimulation onset at threshold (Fig. 3B). However, a further increase in the stimulation intensity still resulted in action potentials late during stimulation (Fig. 3B), contrasting the mature firing pattern of VNLL neurons (Fig. 2A). Nevertheless, the current threshold increased considerably upon lowering the input resistance in P9 neurons (n = 7; Fig. 3C). This increase matched the relationship between the current threshold and the input resistance of cells throughout the developmental period (Fig. 3C). The maximal number of action potentials evoked in P9 neurons also decreased upon the application of a background conductance, even leading to the generation of only a single action potential at stimulation onset in two out of seven neurons, similar to mature neurons (Fig. 3D). Furthermore, the imposed reduction in input resistance consistently shortened the time to the first action potential (Fig. 3E). To conclude, an experimental decrease in input resistance in P9 neurons induces a firing pattern similar to that at more mature stages. Thus the developmental alteration in input resistance largely promotes the developmental change in firing pattern. Yet, further refinement of active membrane properties appears required to explain fully the emergence of the mature firing pattern.
Development of whole-cell potassium currents.
Since lowering the input resistance could not fully explain the maturation of the firing pattern (Fig. 3), a developmental change in the active membrane properties might be required to generate additional outward current. Therefore, we analyzed pharmacologically isolated, voltage-dependent, whole-cell potassium currents at P9–P11 (n = 13) and in animals >P22 (n = 9). Potassium currents were elicited with step potentials ranging from −70 mV to +70 mV, following a 500-ms conditioning potential at −80 mV (Fig. 4A). The peak potassium current at a step potential of +70 mV was slightly larger in mature animals (17.6 ± 1.47 nA, n = 9) compared with that in juvenile P9–P11 animals (14.4 ± 0.87 nA, n = 13, P < 0.05; Fig. 4, A and B). However, as the effective capacitance of VNLL neurons decreased during late postnatal development (Fig. 1), the potassium current density provided a more meaningful basis for comparison. We used the compensated capacitance values to normalize the potassium current to the cell size. The compensated cell capacitance was 1.6-fold larger in P9 (23.7 ± 1.6 pF, n = 13) than in >P22 neurons (15.5 ± 1.4 pF, n = 9, P < 0.05), corroborating the 40% developmental reduction observed in current-clamp recordings (Fig. 1). Based on the calculated current densities, the peak and steady-state potassium flux were substantially larger in >P22 neurons compared with those in P9 neurons (peak: 1.18 ± 0.12 nA/pF, n = 9, vs. 0.62 ± 0.04 nA/pF, n = 13, at a step potential of +70 mV, P < 0.05; Fig. 4B). Thus the relative amount of potassium current increased during late postnatal development in VNLL neurons.
Importantly, the inactivation time course of potassium currents differed substantially between the two age groups (Fig. 4A). As these whole-cell potassium currents appeared to be based on a mixed subset of channel types (Fig. 4A), the inactivation kinetics could not be described accurately using exponential decays. Therefore, we quantified the half inactivation time between peak and steady-state current. Our half inactivation time analysis revealed a four-fold difference (P < 0.05) in the maximal inactivation time in P9 of 0.16 ± 0.06 s (n = 13) compared with 0.57 ± 0.05 s (n = 9) in >P22 animals (Fig. 4B). The lack of ongoing activity in the firing pattern of adult animals is consistent with a larger density of slowly inactivating potassium outward currents.
Development of action-potential properties.
So far, we characterized the developmental changes and relative contributions of passive and active conductances to the firing pattern using step protocols between 200 and 500 ms. However, action potentials in the auditory brain stem are usually triggered by 0.5–2 ms fast synaptic α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid conductances (Ammer et al. 2012; Couchman et al. 2010; Porres et al. 2011; Taschenberger and von Gersdorff 2000; Yang and Xu-Friedman 2009) rather than by prolonged depolarization plateaus. Thus the development of the relevant action-potential properties cannot be extracted using long current injections.
To unravel the developmental interplay of cell intrinsic factors toward action-potential generation and precision, we injected currents approximating the excitatory postsynaptic current (EPSC) shape in juvenile VNLL neurons (Berger et al. 2014). This physiologically relevant current injection of 180 μs rise and 500 μs decay time was incremented in steps of 100 pA (Fig. 5, A and B). Parameters describing the shape of the action potential were averaged from the first suprathreshold event over two to three repetitions/cell (Fig. 5). During late postnatal development, the action-potential size, measured from the resting membrane potential, remained constant, from 86.6 ± 2.4 mV at P9 (n = 11) to 88.7 ± 3.2 mV at ≥P25 (n = 7, P < 0.05; Fig. 5C). The duration of the action potential—determined as the action-potential half-width—decreased significantly over the same developmental period (529 ± 34 μs to 189 ± 14 μs, P < 0.05), with a time constant of 4.3 days (Fig. 5C). In line with the reduced half-width, the speed of action-potential depolarization and repolarization increased significantly throughout development (depolarization P9: 416 ± 30 mV/ms, ≥P25: 814 ± 75 mV/ms, P < 0.05; repolarization P9: −176 ± 15 mV/ms, ≥P25: −610 ± 63 mV/ms, P < 0.05), with developmental time constants of 9.3 and 11.7 days, respectively (Fig. 5C). Importantly, the developmental increase of the repolarization speed of the action potential exceeded that of its depolarization, as indicated by a significant decrease in the ratio of depolarization over repolarization speed (P9: 2.42 ± 0.14; ≥P25: 1.35 ± 0.04, P < 0.05; Fig. 5C). This developmental alteration occurred with a time constant of 5 days. The differential acceleration of action-potential kinetics supports a change in active membrane properties during late postnatal development. Specifically, the larger acceleration of action-potential repolarization is consistent with the larger whole-cell potassium current density in mature animals. Notably, the refinement of active membrane parameters clearly persists after hearing onset (Fig. 5) and appears to succeed the refinement of passive properties, as judged from the exponential fits (Figs. 1 and 5).
To address further the mechanism of action-potential initiation, we investigated the developmental regulation of the action-potential current and voltage threshold in response to EPSC-approximating stimuli. In contrast to the increase in the current threshold to long current injections, both the current and voltage threshold decreased significantly in VNLL neurons between P9 (n = 11) and ≥P25 (n = 7; 3.18 ± 0.24 nA to 1.22 ± 0.11 nA, P < 0.05; −35.8 ± 1.6 mV to −43.2 ± 1.4 mV, P < 0.05; Fig. 6A). Furthermore, since the resting membrane potential remained constant (Fig. 1B), the decrease in the voltage threshold effectively reduced the potential difference that needs to be crossed from rest to the voltage threshold from P9 (n = 11) to ≥P25 (n = 7; 28.9 ± 1.6 mV to 20.7 ± 1.2 mV, P < 0.05; Fig. 6A). Taken together, the development of the suprathreshold response of VNLL neurons indicates a change in active membrane properties.
Interplay of active and passive membrane properties during development.
The developmental changes described so far prompted us to examine the interactions and contributions of passive/resting and active membrane properties in regulating changes in the suprathreshold response of VNLL neurons during late postnatal development. We quantified the relation of the action-potential current threshold to the effective capacitance and the resting input resistance (Fig. 6, B and C). The action-potential threshold in response to EPSC-approximating current injections now correlated more strongly with the cell's effective capacitance (r = 0.797, P < 0.0001) than with its input resistance (r = 0.3759, P = 0.0009; Fig. 6, B and C). The strong correlation between the current threshold and the capacitance can be explained by the fact that these short current injections of ∼0.5 ms fall within 5–10% of the membrane time constant. The voltage response in this initial part of the membrane time constant is largely determined by the cell's capacitance, since the charging current dominates over the current flowing through the resistive element of the membrane. Thus the developmental change in effective capacitance potentially mediates the change in action-potential current threshold. To evaluate this hypothesis further, the current threshold was normalized to the estimated capacitance (Fig. 6D). This normalized current threshold only displayed a 1.6-fold decrease (Fig. 6D) compared with the 2.6-fold decrease in the absolute current threshold over the investigated age range (Fig. 6A). Therefore, the change in capacitance, a passive membrane property, partially drives the developmental decrease in current threshold and hence influences the integrational properties of VNLL neurons.
However, the current threshold also depends on a cell's active membrane properties. Thus the developmental decrease in the current threshold might result from concurrent changes in the active membrane properties, more likely represented by the voltage threshold (Fig. 6A). To ascertain how current and voltage thresholds relate to each other during development, the current threshold normalized to the estimated effective capacitance was correlated to the difference in membrane potential required to reach the voltage threshold (Δ to threshold; Fig. 6E). As indicated by the linear fit, a 10-mV change in the Δ to threshold corresponded to a reduction of almost 20 pA/pF. The 8-mV decrease in the Δ to threshold between P9 and ≥P25 is therefore enough to explain the 22 pA/pF drop in the normalized current threshold over the same developmental period. Taken together, the reduction in the effective capacitance and the change in active membrane properties explain the large drop in the current threshold during late postnatal development. Thus the effective capacitance and active membrane properties govern the output of a neuron to short EPSC-approximating stimuli.
If responses to EPSC-approximating stimuli in the soma are controlled by the capacitance, as well as by the active membrane properties of a cell, then the decrease in input resistance during development should not strongly affect the action-potential current threshold in response to these very short stimuli. To assay directly the role of the input resistance in determining the current threshold, the effects of applying a background leak conductance were, once more, investigated in P9–P10 animals (Fig. 6, F and G). As before, the current threshold was assessed using an EPSC-approximating current injection, incremented in 100 pA steps. The current threshold was measured for the different input resistances within a given neuron (Fig. 6F). In agreement with our previous results, the increase of the leak—and thus decreasing the input resistance—only slightly increased the current threshold in P9 neurons (n = 7; Fig. 6G). Furthermore, the decrease in input resistance did not affect the action-potential waveform (Fig. 6G). To conclude, these findings confirm our prediction that the observed developmental reduction in the current threshold in response to short current injections cannot be explained by the decrease in input resistance but is rather largely mediated by the reduction in cell capacitance.
Input resistance controls action-potential precision.
At the end of a current injection, the time course with which a charged membrane relaxes back to the resting potential also depends on the membrane resistance. Thus we hypothesize that the membrane resistance influences the time the membrane potential spends close to the action-potential threshold. Since action-potential precision is influenced by the time the membrane spends close to threshold (Ammer et al. 2012; Rodriguez-Molina et al. 2007), the input resistance should affect the precision of action-potential generation to short current injections. Therefore, we assessed whether action-potential precision is developmentally regulated in the VNLL and how it is affected by changes in input resistance.
Action-potential precision is classically assessed in terms of action-potential latency and jitter. To measure these parameters, the injected, short EPSC-approximating current was scaled to evoke action potentials in ∼40% of the trials (Fig. 7A) (Ammer et al. 2012; Cathala et al. 2003; Fricker and Miles 2000; Lehnert et al. 2014). As in the DNLL (Ammer et al. 2012), action-potential latency and jitter decreased significantly between P9 (n = 11) and P25 (n = 10; latency P9: 1.15 ± 0.05 ms, P25: 1.00 ± 0.04 ms, P < 0.05; jitter P9: 0.17 ± 0.02 ms, P25: 0.12 ± 0.01 ms, P < 0.05; Fig. 7B) in VNLL neurons. This type of stimulation close to threshold confirmed the age dependency of the current threshold to short current injections (P9: 3.13 ± 0.15 nA; P25: 1.00 ± 0.06 nA, P < 0.05; Fig. 7B). To assess whether the observed developmental increase in action-potential precision is driven by changes in input resistance, a constant background conductance was again added to the EPSC-approximating current injections (Fig. 7A) to reduce artificially the input resistance of recorded cells. As indicated by the linear regression analysis, the application of this background conductance reduced the latency and jitter of evoked action potentials within single cells and over the population of juvenile and mature VNLL neurons (latency P9: r = 0.4669, P = 0.0036, P25: r = 0.663, P = 0.0006; jitter P9: r = 0.5864, P = 0.0001, P25: r = 0.4969, P = 0.0159; Fig. 7, A and B). Interestingly, the imposed leak changed the latency and jitter to such an extent that these parameters in P9 neurons adopted the same dependency on the input resistance than at P25 (Fig. 7B). Thus leak appears not only important but also sufficient to regulate action-potential precision. Consistent with our previous results (Fig. 6G), the imposition of a leak conductance increased the current threshold to short current injections rather slightly (P9: r = 0.01229, P = 0.9425; P25: r = −0.08852, P = 0.6880; Fig. 7B). Finally, the lowering of the input resistance did not affect the ratio of the action-potential de- and repolarization speed in either age group tested (P9: r = 0.2582, P = 0.1229; P25: r = 0.2478, P = 0.2542; Fig. 7B), indicating that these properties indeed depend on active properties.
Consistent with the developmental increase in precision, P9 neurons (n = 11), with a higher input resistance, generated membrane deflections that remained close to action-potential threshold for longer than those in P25 neurons (n = 10; P9: 0.49 ± 0.03 ms; P25: 0.36 ± 0.02 ms, P < 0.05; Fig. 7C). The time the voltage deflection spent within 1 mV below threshold was reduced significantly with the addition of a background conductance, hence, by decreasing the input resistance (time at top 1 mV, P9: r = 0.6045, P < 0.0001; P25: r = 0.5983, P = 0.0026; Fig. 7C). As a result, P9 neurons adopted the same relationship between the input resistance and the time the membrane potential spent close to threshold, as found in P25 neurons. Thus the time window for action-potential generation decreased in an input resistance-dependent manner. Taken together, our data show that during development, the action-potential current threshold is strongly affected by the cell's effective capacitance, whereas the precision of action-potential generation is controlled by the cell's input resistance. Thus action-potential generation elicited with brief EPSC-approximating current injections is controlled differentially by the two major passive membrane properties.
Cholinergic modulation enhances excitability.
From our findings, we would expect a depolarization to reduce the current threshold to EPSC-approximating injections through a decrease in the Δ to the action-potential voltage threshold (Fig. 6E). Moreover, predepolarizations were shown previously to change the firing pattern of VNLL neurons (Berger et al. 2014). Thus endogenous neuromodulations that lead to a depolarization should facilitate the generation of action potentials and alter the firing pattern of VNLL neurons. A strong candidate for generating endogenous depolarizations in the VNLL is acetylcholine, as suggested by the expression of acetylcholine receptors in this nucleus (Happe and Morley 2004). Furthermore, cholinergic modulation has been reported previously to alter neuronal excitability (Gulledge et al. 2009; Smith and Araneda 2010; Stiefel et al. 2008) and in the case of the auditory system, has been described in the cochlear nucleus (Fujino and Oertel 2001; He et al. 2014). Therefore, we examined whether cholinergic neuromodulators could generate depolarization-induced changes in excitability in P9–P10 and P19–P21 VNLL neurons, thereby altering the neurons' response profiles. First, we tested whether cholinergic agonists could evoke voltage responses in juvenile VNLL neurons. Puff application of carbachol (500 μM for 4 s), a rather unselective acetylcholine agonist, evoked a clear membrane depolarization in a P9 neuron (Fig. 8A). The same neuron was subsequently challenged with a puff application of external recording solution, which did not elicit a voltage response (Fig. 8A). Therefore, our pressure application itself did not induce voltage responses. Finally, puff application of oxotremorine (500 μM), a muscarinic acetylcholine receptor agonist, elicited a membrane depolarization in VNLL neurons (Fig. 8A). In the next set of experiments, we assessed whether depolarizations elicited by cholinergic agonists can be evoked at mature stages. A brief, 100-ms pressure application of 500 μM carbachol induced a robust depolarization in all four P9–P10 (7.6 ± 1.2 mV) and in all six P19–P21 (5.1 ± 0.5 mV) cells. An even more pronounced depolarization was evoked by the application of 1 mM oxotremorine for 100 ms (P9–P10: 11.9 ± 3.6 mV, n = 4; P19–P21: 9.8 ± 2.7 mV, n = 3). Thus VNLL neurons are likely subject to a developmentally independent form of neuromodulation mediated through metabotropic acetylcholine receptors.
The effect of metabotropic acetylcholine receptor activation on excitability was assayed further by applying short and long current injections before and after agonist application in these P9–P10 (Fig. 8Bi) and P19–P21 (Fig. 8Bii) neurons. The injected current was adjusted to evoke a voltage deflection ∼10% below action-potential threshold in response to the short current injections and a response just above threshold during the long current injections. The subthreshold event in response to short current injections became suprathreshold following agonist application in both P9–P10 and P19–P21 animals (Fig. 8Biii). Furthermore, cholinergic modulation elicited changes in the firing pattern of P9 neurons in response to long current injections. This change was manifested as a decrease in the time to the first action potential and in some cases, as the generation of an action potential at stimulation onset (Fig. 8, Biii and F). The onset voltage response to long current injections remained largely unchanged in P19–P21 neurons (Fig. 8Biii). Thus the generation of action potentials, the firing pattern, and the integrational properties of ventral VNLL neurons can be dynamically adjusted by slow, endogenous neuromodulation.
Next, we evaluated the contribution of the membrane resistance to depolarization-induced changes in firing pattern. To generate a change in excitability similar to that elicited by cholinergic neuromodulation, we introduced a predepolarization using a current injection of +70 pA for 800 ms. The membrane resistance during the predepolarization was measured with a nested, 200-ms hyperpolarization step, induced by a −15-pA current injection (Fig. 8C). The neuron's input-output function was subsequently determined using a 300-ms test injection incremented in 100-pA steps (Fig. 8C). The predepolarization raised the membrane potential by 8.42 ± 0.76 mV in P9–P10 neurons (n = 7) and promoted an action potential at stimulus onset (Fig. 8C). The predepolarization also reduced the membrane resistance of P9–P10 neurons (270 ± 18 MΩ to 156 ± 22 MΩ, n = 7, P < 0.05; Fig. 8D). This decrease in membrane resistance coincided with a slight but insignificant increase in the current threshold induced by the predepolarization (286 ± 18 pA to 299 ± 15 pA, n = 7, P < 0.05; Fig. 8E), in agreement with previous results (Fig. 6). Finally, both the application of acetylcholine agonists and depolarizations induced by current injection appeared similarly effective in decreasing the latency to the first action potential (Fig. 8F). Again, this decrease in action-potential latency was similar to that induced by artificial reductions in membrane resistance (Fig. 6), supporting the contribution of the membrane resistance to depolarization-induced changes in excitability. To conclude, depolarizations induced by cholinergic modulation result in changes in the excitability and firing pattern of VNLL neurons, largely predicted by concomitant changes in membrane resistance.
Here, we examined the interaction of passive and active membrane properties in establishing the temporal precision of mature VNLL neurons during late postnatal development. The maturation of these intrinsic properties leads to shorter integration time windows and enhanced precision of action-potential firing. These developmental changes form the basis of an operating mode characteristic of intrinsic coincidence detectors, optimized for precise information processing. Moreover, the activation of metabotropic acetylcholine receptors depolarizes VNLL neurons in a developmentally independent manner, promoting the generation of suprathreshold responses with high precision. Importantly, our data show that passive and active membrane properties work in synergy during development to establish the physiologically relevant features of auditory neurons.
The role of passive membrane parameters in the development of excitability.
By estimating the passive membrane properties close to the resting membrane potential, we show that during late postnatal development, VNLL neurons become leakier and smaller, the latter being corroborated anatomically. Besides leading to a shorter integration time constant, these changes interact with active membrane properties to determine the threshold and precision of action-potential generation as well as the firing pattern. Interestingly, the input resistance and the capacitance contribute to VNLL excitability in different ways. Our findings also suggest that the relative contribution of these membrane properties depends critically on the stimulus duration. For very short EPSC-approximating stimuli, lasting <10% of the membrane time constant, the capacitance dominates the voltage response that leads to action-potential generation. In contrast, a decrease in input resistance alone is sufficient to shorten the time the voltage response spends close to threshold. Thus the capacitance could control the reaching of the action-potential threshold, while a decrease in input resistance alone was sufficient to enhance action-potential precision by shortening the subthreshold integration time window. This finding is consistent with the view that the shape of subthreshold events controls action-potential precision (Ammer et al. 2012; Axmacher and Miles 2004; Cathala et al. 2003; Cudmore et al. 2010; Fricker and Miles 2000; Rodriguez-Molina et al. 2007; Zsiros and Hestrin 2005). For long-lasting stimuli, the input resistance is the dominating membrane property, as it determines the membrane potential that is reached in steady state. Under these stimulation conditions, VNLL neurons displayed a developmental shift in operating mode from buildup to firing only at stimulation onset. The latter type of firing pattern is inherent to the operating mode of coincidence-detector neurons (Ratte et al. 2013). To conclude, our data illustrate the importance of the stimulation time course, as the response to different stimulus durations differentially depends on the capacitance or the input resistance.
Contribution of potassium currents to the development of action-potential firing.
In the auditory brain stem, developmental changes in voltage-gated potassium channels (Bortone et al. 2006; Garcia-Pino et al. 2010; Hardman and Forsythe 2009; Liu and Kaczmarek 1998; Nakamura and Takahashi 2007; Scott et al. 2005) are well documented. Here, we describe how developmental changes in the whole-cell potassium current interact with alterations in passive membrane properties to mediate this change in firing behavior. Mature VNLL neurons exhibited a larger potassium current density than juvenile neurons. This difference resulted from a slight developmental increase in the absolute current amplitude and most importantly, from a concurrent reduction in the effective capacitance. This particular interaction of passive and active membrane properties can largely explain the developmental change in action-potential kinetics. Notably, the potassium current also acquired slower inactivation kinetics. Together with the developmental decrease in input resistance, the prolonged potassium current could mediate the outward drive required for the firing pattern of coincidence-detector neurons (Ratte et al. 2013). Hence, these findings, once more, illustrate the importance of the interaction of passive and active membrane properties. Thus the developing interaction of these intrinsic properties is crucial in establishing physiologically relevant firing properties.
Cholinergic-induced depolarizations modulate the excitability of VNLL neurons.
Besides being developmentally regulated, neuronal firing properties are tightly controlled by neuromodulations. Here, we provide evidence of a strong excitatory drive provided by the activation of muscarinic acetylcholine receptors. Depolarizations induced by muscarinic acetylcholine receptor agonists facilitated the generation of action potentials. Similar depolarizations induced by current injections also reduced the input resistance, suggesting that cholinergic modulation could enhance temporal precision. Importantly, these cholinergic-mediated depolarizations could be elicited regardless of the developmental stage, similar to glutamatergic modulation in the cochlear nucleus (Chanda and Xu-Friedman 2011). Thus rather than driving developmental processes, this modulation might support the faithful transfer of suprathreshold information. For example, the activation of this excitatory drive during periods of high activity might further enable cells to sustain high firing rates with high precision, possibly even counteracting synaptic depression.
Time course of developmental alterations.
As in the dorsal nucleus of the lateral lemniscus (Ammer et al. 2012) and in the medial superior olive (Chirila et al. 2007; Magnusson et al. 2005; Scott et al. 2005), the maturation of passive membrane properties in the gerbil VNLL precedes that of active membrane properties. The delayed development of active properties could result from a more extensive requirement of sensory experience in regulating channel expression. Whereas the maturation of the input resistance and the capacitance is complete at ∼P15 in the VNLL and in the dorsal nucleus of the lateral lemniscus (Ammer et al. 2012), changes in these features persist until P17–P20 in the medial superior olive (Chirila et al. 2007; Magnusson et al. 2005; Scott et al. 2005). Active membrane properties based on action-potential properties mature until ∼P18 in the VNLL and in the dorsal nucleus of the lateral lemniscus (Ammer et al. 2012) compared with ∼P22 in the medial superior olive (Scott et al. 2005). From the comparison of these developmental profiles (Ammer et al. 2012; Chirila et al. 2007; Magnusson et al. 2005; Scott et al. 2005), it is tempting to speculate that the time course of refinement is related to the complexity of the integrational task of a neuron. Neurons in the medial superior olive integrate excitation and inhibition with active conductances with submillisecond precision (Golding and Oertel 2012; Grothe 2003; Myoga et al. 2014). Although the integration of excitation and inhibition is also of functional relevance in the dorsal nucleus of the lateral lemniscus, the timing is less exquisite, as it operates within tens of milliseconds (Burger and Pollak 2001; Pecka et al. 2007; Siveke et al. 2006; Yang and Pollak 1994). Finally, VNLL neurons predominantly integrate excitation within the millisecond time range (Adams 1997; Covey and Casseday 1991; Recio-Spinoso and Joris 2014; Zhang and Kelly 2006). Thus the differential time course of developmental maturation across auditory nuclei might reflect the amount of sensory experience required to achieve different integration complexities.
Functional implications for globular VNLL neurons.
The emergence of precise action-potential firing and an intrinsic operating mode of coincidence detection observed here is well suited to the functional requirements of mature VNLL neurons. The predominant excitatory input to globular VNLL neurons originates from the octopus cell area and is capable of generating high firing rates (Oertel et al. 2000). In turn, VNLL neurons give rise to rapid-onset inhibition in the inferior colliculus (Nayagam et al. 2005; Pollak et al. 2011). Thus mature VNLL neurons are expected to be capable of following high stimulation rates and generate action potentials with high temporal precision. Here, we show mechanistically how biophysical adaptations allow VNLL neurons to achieve the temporal precision of this signal-processing task.
In juvenile VNLL neurons, the temporal precision of information transfer relies on the coincidental activation of synaptic inputs (Berger et al. 2014). Here, we describe how the development of intrinsic features further supports precise information transfer by shaping the neurons' operating mode to that of an intrinsic coincidence detector. Thus we propose that the basis for temporally precise information transfer, depending initially on the coactivation of synaptic inputs, is complemented by the emergence of intrinsic coincidence-detector properties in mature VNLL neurons.
Funds for this work were provided by the Deutsche Forschungsgemeinschaft (DFG; FE789/3-1). F. Felmy and S. A. Gleiss were funded by the Elisabeth and Helmut Uhl Foundation, C. Berger by the DFG (FE789/3-1), D. L. Franzen by the DFG Research Training Group (GRK) 1091, J. J. Ammer by the Bundesministerium für Bildung und Forschung through the Bernstein Fokus 01GQ0981, and F. S. Kümpfbeck by the DFG SFB870.
The authors declare no competing financial interests.
Author contributions: D.L.F., S.A.G., C.B., F.S.K., J.J.A., and F.F. conception and design of research; D.L.F., S.A.G., C.B., J.J.A., and F.F. performed experiments; D.L.F., S.A.G., C.B., F.S.K., J.J.A., and F.F. analyzed data; D.L.F., S.A.G., C.B., F.S.K., J.J.A., and F.F. interpreted results of experiments; D.L.F., S.A.G., C.B., F.S.K., and F.F. prepared figures; D.L.F., J.J.A., and F.F. drafted manuscript; D.L.F., F.S.K., J.J.A., and F.F. edited and revised manuscript; D.L.F., S.A.G., C.B., F.S.K., J.J.A., and F.F. approved final version of manuscript.
The authors thank Rainer Uhl and Benedikt Grothe for support, Mike Burger and Mike Myoga for comments on the manuscript, and Christian Leibold for discussions.
- Copyright © 2015 the American Physiological Society