Abstract
The medial nucleus of the trapezoid body (MNTB) in the auditory brainstem is the principal source of synaptic inhibition to several functionally distinct auditory nuclei. Prominent projections of individual MNTB neurons comprise the major binaural nuclei that are involved in the early processing stages of sound localization as well as the superior paraolivary nucleus (SPON), which contains monaural neurons that extract rapid changes in sound intensity to detect sound gaps and rhythmic oscillations that commonly occur in animal calls and human speech. While the processes that guide the development and refinement of MNTB axon collaterals to the binaural nuclei have become increasingly understood, little is known about the development of MNTB collaterals to the monaural SPON. In this study, we investigated the development of MNTB-SPON connections in mice of both sexes from shortly after birth to three weeks of age, which encompasses the time before and after hearing onset. Individual axon reconstructions and electrophysiological analysis of MNTB-SPON connectivity demonstrate a dramatic increase in the number of MNTB axonal boutons in the SPON before hearing onset. However, this proliferation was not accompanied by changes in the strength of MNTB-SPON connections or by changes in the structural or functional topographic precision. However, following hearing onset, the spread of single-axon boutons along the tonotopic axis increased, indicating an unexpected decrease in the tonotopic precision of the MNTB-SPON pathway. These results provide new insight into the development and organization of inhibition to SPON neurons and the regulation of developmental plasticity in diverging inhibitory pathways.
SIGNIFICANCE STATEMENT The superior paraolivary nucleus (SPON) is a prominent auditory brainstem nucleus involved in the early detection of sound gaps and rhythmic oscillations. The ability of SPON neurons to fire at the offset of sound depends on strong and precise synaptic inhibition provided by glycinergic neurons in the medial nucleus of the trapezoid body (MNTB). Here, we investigated the anatomic and physiological maturation of MNTB-LSO connectivity in mice before and after the onset of hearing. We observed a period of bouton proliferation without accompanying changes in topographic precision before hearing onset. This was followed by bouton elimination and an unexpected decrease in the tonotopic precision after hearing onset. These results provide new insight into the development of inhibition to the SPON.
Introduction
The superior olivary complex (SOC) in the mammalian brainstem is an important early processing station in the ascending central auditory pathway. The two major binaural nuclei of the SOC, the lateral superior olive (LSO) and the medial superior olive (MSO), play essential roles in encoding the direction of incoming sound via the processing of interaural sound level and interaural time differences, respectively (Tollin, 2003; Grothe et al., 2010). In contrast, the monaural superior paraolivary nucleus (SPON) contains neurons that are sensitive to the offset of sounds (Felix et al., 2011; Kopp-Scheinpflug et al., 2011) and to rhythmic modulations in sound amplitudes, which are typically found in animal calls and human speech (Gómez-Álvarez et al., 2018; Magnusson and Gómez-Álvarez, 2019). To extract and encode these diverse acoustic features, all three nuclei critically depend on strong and precisely timed synaptic inhibition, which is provided by the medial nucleus of the trapezoid body (MNTB; Kandler et al., 2020). Mature MNTB neurons are glycinergic and send axon collaterals into each of the three target nuclei in a tonotopically organized manner (Banks and Smith, 1992; Sommer et al., 1993; Kim and Kandler, 2003). Despite using the same source of presynaptic inhibition, neurons in the LSO, MSO, and SPON integrate MNTB-derived synaptic inhibition differently to achieve their specific computational goals (Kandler et al., 2020).
Many studies have investigated the structural and functional processes by which the tonotopic organization of MNTB afferents is established during development and by which their synaptic action on target cells becomes physiologically fine-tuned. Anatomical investigations revealed that with the beginning of hearing (in rodents at the end of the second postnatal week), MNTB axons in both the LSO and MSO are pruned to create a more precise structural tonotopic organization (Sanes et al., 1992; Kim and Kandler, 2003; Werthat et al., 2008). Physiologic studies, however, revealed significant target-specific differences in the time course of refinement on the functional level. In the LSO, functional elimination (synaptic silencing) and strengthening of MNTB inputs occur before hearing onset, thus preceding the period of structural pruning (Kim and Kandler, 2003; Walcher et al., 2011). This early functional reorganization depends on spontaneous, patterned activity generated in the prehearing cochlea from where it propagates along the central ascending pathway (Tritsch et al., 2007; Clause et al., 2014; Babola et al., 2018). In contrast, in the MSO, functional refinement of MNTB inputs occurs simultaneously with structural pruning after hearing onset and is sensitive to disturbances in early auditory experience (Kapfer et al., 2002; Magnusson et al., 2005; Werthat et al., 2008).
In contrast to the LSO and MSO, the development of MNTB afferents to the SPON has received little attention, despite the crucial role of MNTB-generated inhibition in generating sound offset detection and encoding rhythmic sound modulation by SPON neurons. A recent study found no evidence for a refinement of MNTB inputs to the SPON after hearing onset (Rajaram et al., 2020), suggesting that MNTB-SPON connections are optimized before hearing onset. Here, we combined anatomic and functional approaches to investigate the development of MNTB-SPON connections in mice during the first three postnatal weeks, which covers the periods before hearing onset, when spiking in MNTB neurons is driven by spontaneous, cochlear-generated activity (Tritsch et al., 2007, 2010; Clause et al., 2014), and after hearing onset, when MNTB neurons are driven by sound (Sonntag et al., 2009, 2011). Our results demonstrate an extensive increase in the number of MNTB axonal boutons in the SPON, but a complete absence of topographic refinement before hearing onset, both on a structural and functional level. Unexpectedly, following hearing onset, we found an increase in the spread of boutons along the tonotopic axis, indicating a decrease in the tonotopic precision of the MNTB-SPON pathway. These results shed new light on the development and organization of inhibition to SPON neurons and the regulation of developmental plasticity in diverging inhibitory pathways.
Materials and Methods
Animals
Experiments were performed in mice of both sexes of the strain 129S6/SvEv at postnatal day (P)2–P4, P12–P14, and P19–P21. All experimental procedures were in accordance with National Institutes of Health guidelines and were approved by the Institutional Animal Care and Use Committee at the University of Pittsburgh.
Acute slice preparation and electrophysiological recordings
Brain slices
Brainstem slices were obtained as previously described (Bach and Kandler, 2020). Mice were anesthetized with vaporized isoflurane (Piramal Healthcare), decapitated, and the brain was removed and submerged in artificial CSF (aCSF; composition in mm: 124 NaCl, 26 NaHCO3, 10 glucose, 5 KCl, 1.25 KH2PO4, 1.3 MgSO4, 2 CaCl2, pH 7.4 when aerated with 95% O2/5% CO2). During slice preparation the aCSF also contained 1 mm kynurenic acid to decrease excitotoxicity. Coronal brain slices (350 µm thick for biocytin labeling and 300 µm for glutamate uncaging) were prepared with a vibratome (Leica VT 1200). Slices containing the MNTB and SPON were transferred to an interface chamber with aerated aCSF and incubated for 30 min at 32°C after which slices were maintained at room temperature until used for biotin filling or electrophysiological recordings and functional mapping.
Biocytin filling and histologic processing
Slices were transferred to a submerged-type recording chamber mounted to an upright microscope and continuously perfused with aerated aCSF. Whole-cell patch-clamp recordings were made from individual neurons for ∼30 min using 8–15 MΩ pipettes filled with an internal solution containing (in mm) 100 K-gluconate, 11 EGTA, 10 KCl, 1 MgCl2, 1 CaCl2-H2O, 10 HEPES, 0.3 Na-GTP, 2 Mg-ATP, 0.1 Alexa Fluor 568 hydrazide (Invitrogen), and 0.5% biocytin (pH 7.2, 280 mOsm/l). Alexa 568 was included in the pipette for online identification of MNTB axons that were cut during slicing. Cell viability during filling was monitored by eliciting spikes in response to depolarizing current injections. After fillings, slices were incubated in the interface chamber for approximately 1 h before being fixed in 4% paraformaldehyde in 0.01 m PBS for 1–7 d. Slices were cryoprotected by infiltration with 30% sucrose in PBS. Slices from P12–P14 and P19–P21 animals were sectioned approximately in half using a freezing microtome (Microm HM 430). Sections went through three freeze-thaw cycles before incubation in 10% methanol and 3% H2O2 in PBS for 30 min. Sections were washed in PBS, incubated in 2% normal goat serum and 0.2% Triton X-100 in PBS for 4 h, and reacted with an avidin-biotin reagent (ABC Elite kit, Vector Laboratories) for 2 h at room temperature and then overnight at 4°C. Sections were then washed in PBS and reacted with 0.05% diaminobenzidine-tetrachloride (DAB; Sigma) in a solution containing 1% CoCl2, 1% Ni(NH4)2SO4, and 0.3% H2O2 in PBS. Sections were washed in PBS, mounted on gelatinized glass slides, dehydrated in an ethanol/xylene series, and coverslipped.
Electrophysiological recordings
Whole-cell patch-clamp recordings from SPON neurons were performed as previously described for LSO neurons (Weisz et al., 2016). Recordings were aimed at round/oval-shaped neurons in the SPON, visually identified under infrared illumination and differential interference contrast (AxioExaminer A1, Zeiss). Recording pipettes (resistance 3–7 MΩ) were filled with a Cs-based and high chloride internal solution containing the following (in mm): 67.6 D-gluconic acid, 49 CsOH, 56 CsCl, 1 MgCl2, 1 CaCl2, 10 HEPES, 11 EGTA, 2 Mg-ATP, 0.3 Na-GTP, 3 Na2-phosphocreatine, adjusted to pH 7.2 and 280 mOsm. Pipette solution also contained QX-314 (5 mm, Tocris Bioscience) and 0.3% biocytin (MilliporeSigma). Intrinsic and synaptic currents were recorded using a MultiClamp 700B amplifier (Molecular Devices) and a Digidata 1440A digitizer (Molecular Devices) controlled by Clampex 10.4 software. Data were filtered at 3 kHz (Bessel filter) and digitally sampled at 10 kHz. Access resistances were between 6 and 28 MΩ and series resistance was compensated by 50%. Membrane voltage was corrected for a liquid junction potential of 7.8 mV calculated by Clampex 10.7 (Molecular Devices).
Experimental design, data analysis, and statistics
Axon reconstruction and analysis
Only cells without visibly cut axon branches within the SPON, as identified by a bulbous ending at the slice surface, were included for analysis. The analysis includes some axons (P12–P14: two axons, P19–P21: five axons) whose collateral projections to the LSO were included in a previous study (Clause et al., 2014). Axonal arbors within the SPON were reconstructed using a Neurolucida system (MBF Bioscience) coupled to a Zeiss Imager M1 with 60× and 100× oil objectives. Putative synaptic boutons were identified based on their characteristic round shape and diameter greater than 2× the width of the axon. Reconstructions and analysis were performed blind to age, although slices at P2–P4 generally were noticeably smaller than those at P12–P14 or P19–P21. For analysis, three-dimensional reconstructions were flattened rostro-caudally into a two-dimensional, mediolaterally and dorsoventrally defined Cartesian coordinate system.
Neurolucida Explorer was used for reconstructions and basic analysis, including determining the location of the soma in the MNTB and the location of the bouton cloud centroid (center of mass) in the SPON, measuring the length of axons in the SPON and the cross-sectional SPON area and determining the length of axons and the number of boutons within the SPON. To estimate the area covered by boutons of individual MNTB axons and their tonotopic spread in the developing SPON, we used the ellipse fitting method (Clause et al., 2014). With this approach, we captured the elongated shape of the termination zones using custom-written LabView programs (National Instruments, TX) and the Khachiyan Algorithm in MATLAB (The MathWorks) to fit an ellipsoid to the bouton coordinates that maximized the density of boutons within the confines of the ellipse. As the SPON increases in size during prehearing development, bouton areas and tonotopic spread were normalized to the cross-sectional area or medio-lateral width of the SPON, respectively, to allow comparisons across ages.
Statistical significance of age group differences was assessed using One-way ANOVA followed by Tukey's HSD post hoc comparisons. We accessed the validity of using ANOVAs, by verifying that the assumptions of parametrical inferential analyses are met, including that the residuals (errors) of the linear model are Normally distributed (normality assumption; Q-Q plots using 'qqplot' function in MATLAB, Anderson–Darling test using 'adtest' function in MATLAB) and with constant variance (homoscedasticity assumption; Bartlett test using 'vartestn' function in MATLAB). To evaluate the topographic organization of the MNTB-SPON projection at each age studied, the Pearson coefficient of correlation (r) between soma location (MNTB) and bouton centroid location (SPON) was measured (a high correlation indicates a high topographic organization. Differences in topographic organization between the ages studied were assessed by ANCOVA analysis. Data are presented as scatter plots raster plots, polar plots, or mean or median ± 95% confidence interval. Figures were created and data analyzed using MATLAB (The MathWorks).
Density heat maps
The boutons of each reconstructed axon were overlaid with a two-dimensional grid comprised of 59.3 μm2 (at P2–P4) or 100 μm2 (at P12–P14 and P19–P21) bins (Neurolucida Explorer). The bin sizes encompass a constant proportion of SPON area (∼0.14%) at each of the ages studied. The grid covered the total area of the SPON and was positioned consistently in relation to the bouton centroid of each axon. As a result, the data from each axon were aligned at the bouton centroid, to account for differences in the location of the termination within the SPON and along its frequency axis. Custom LabView programs were used to sort both the number of boutons per bin and the percent of the total bouton number per bin into vector and matrix formats that maintained the geographical position of the boutons in mediolaterally and dorsoventrally defined space. The vector data for all the axons in each age group were averaged, and the means were converted into matrix form, which was then plotted as a heat map of average bouton density using MATLAB). For presentation purposes, the heat maps were positioned medially within a representative outline of the SPON, to reflect the position of the majority of termination regions, though the actual location of individual axon terminations within the SPON varied.
Density of boutons along the SPON tonotopic axis
The Cartesian coordinates of each axon's boutons were rotated around the centroid so that the horizontal axis of the coordinate system corresponded to the putative SPON frequency axis, rather than mediolateral space, with an origin at the bouton centroid. The frequency axis was then divided into 2.21 μm-wide (at P2–P4) or 2.85 μm-wide (at P12–P14 and P19–P21) bins that corresponded to 1% of the mediolateral length of the SPON. For each axon, counts of the number of boutons per bin along the SPON frequency axis were computed and data for all axons in each age group were then averaged.
To visualize the differences in bouton distributions between ages, their histograms (superimposed with their normal distribution fits) and their empirical cumulative distribution functions were plotted. The histograms (bin width, 1) and the superimposed fitted normal distributions (bandwidth, 5) were created using the function 'histfit' in MATLAB and the empirical cumulative distribution functions were created using the function 'ecdf' in MATLAB. Statistically, the distributions of the total number of boutons along the SPON frequency axis were compared between the three ages studied using the two-sample Kolmogorov–Smirnov test ['kstest2' function in MATLAB; P2–P4 vs P12–P14; P2–P4 vs P19–P21; P12–P14 vs P19–P21; the Kolmogorov–Smirnov test statistic (D) is the maximum difference between their distributions].
Estimation of the number of input fibers and strengths
Postsynaptic currents (PSCs) were elicited by electrical stimulation of MNTB axons at the medio-lateral border of the MNTB using gradually increasing stimulus intensities (0.1 ms, 20–800 µA, Isoflex, AMPI; Kim and Kandler, 2003). The number of MNTB fibers synaptically connected to the recorded SPON neuron was estimated using the number of steps in the stimulus-response plot. To determine the number of steps objectively, a Gaussian mixture model (GMM) was applied to the postsynaptic peak amplitudes (MATLAB, The MathWorks) with the assumption that the distribution of PSC peak amplitudes from multiple fibers follows a linear combination of multivariate Gaussian distribution of PSC peak amplitudes from each fiber. Distributions of PSC peak amplitudes were test-fitted with a set of 1–20 Gaussian component(s), and each set was tested for 100 iterations with random initialization. Bayesian information criteria (BIC) were used to find the optimal model (Schwarz, 1978), in which the lower BIC value provides the stronger evidence for the model. The optimal model was selected by the set that consists of the most Gaussian components among the sets within a BIC score difference of 20 from the minimum, given that a minimum BIC tends to result in underestimated complexity of data (Burnham and Anderson, 2004; Dziak et al., 2020). The model with the difference in BIC score of 20 from the minimum was considered to provide sufficiently strong evidence for the models yet statistically different from the model of minimum BIC (Raftery, 1995). From the optimal model, the number of MNTB fibers was estimated from the number of Gaussian components (the number of steps in peak amplitudes), and fiber strength was measured by the distance between the mean of neighboring Gaussian components. Maximum amplitude was measured by the mean of the five largest peak amplitudes. Individual fiber strength was evaluated by collecting all fiber strength (the distance between the mean of Gaussians) for each SPON neuron). Strongest fiber was determined by the largest distance between Gaussians. The coefficient of variation (CV) was calculated for individual fiber strengths for each SPON neuron. To determine the relative contributions of fiber inputs for each SPON neuron, individual fiber strength for each neuron was normalized to its strongest fiber.
Development of MNTB-derived synaptic inputs and membrane properties of SPON neurons
Postsynaptic currents were aligned and scaled to PSC peaks and the decay time constant was measured by fitting a single exponential to decay currents 20–80% from peak to baseline. For the reversal potential for MNTB-evoked IPSCs, PSCs were elicited by electrical stimulation of MNTB axons at maximum intensity while SPON neurons were held at different holding potentials from −88 to +22 mV. Reversal potential was estimated from the linear regression of peak synaptic amplitude as a function of holding potentials. Intrinsic properties of SPON neurons were determined by changing membrane potentials in the voltage-clamp recordings (the voltage step size of −50 mV for Ih, −10 mV for membrane resistance, 5 mV for membrane capacitance).
Functional connectivity mapping
The spatial distribution of MNTB neurons providing synaptic inputs to a single SPON neuron was determined using focal photolysis of caged glutamate (MNI-caged L-glutamate, 200 μm, Tocris Bioscience; Fig. 11) as previously described (J. Sturm et al., 2014; J.J. Sturm et al., 2017). A custom-built uncaging system and LabVIEW programs (by Tuan Nguyen) were used to control the position of UV (355 nm) light spots (diameter 20 µm, duration 1 ms, intensity 2 mW, measured at the plane of the slice). Light pulses were delivered at 1 Hz in an array of spots spaced 30 µm apart over the MNTB in a randomized order. For each neuron, three stimulus iterations were performed. Data were analyzed with custom-written LabVIEW programs and MATLAB scripts.
Postsynaptic charge (PSQ) in a window of 100 ms after photostimulation was measured. In order to be included as a photostimulation-elicited synaptic response, the following requirements had to be fulfilled: (1) PSQs had to be larger than the mean ± 3 SD of the baseline charges without photolysis, which was determined for each neuron, (2) response latency had to be ≤25 ms to exclude responses mediated by a potential indirect pathway between the MNTB and SPON, and (3) PSQs that meet the first and second requirements had to occur in at least two out of three stimulation iterations. MNTB and SPON boundaries were defined from an image taken with a 2.5× objective of the slice during uncaging. Only input sites within the boundaries of the MNTB were analyzed.
Results
Development of structural MNTB-SPON connectivity
At all ages investigated (P2–P4, P12–P14, P19–P21), intracellular filling of MNTB neurons with biocytin appeared to completely label all axonal arbors, including fine branches and the round swellings that mark putative synaptic boutons (Fig. 1). Labeled MNTB cell bodies were distributed along most of the tonotopic, mediolateral axis and the distribution of the tonotopic location of filled MNTB neurons or bouton centroids did not differ between age groups (Fig. 2A). Consistent with previous reports (Banks and Smith, 1992; Rajaram et al., 2020), we found a topographic organization of the projection from the MNTB to the SPON, as indicated by a positive correlation between the location of the MNTB neuron's mediolateral location and the mediolateral location of the bouton centroid in the SPON (the mean position of all boutons of an axon in the SPON; Fig. 2B,C). This topographic organization was already present at P2–P4, and the correlation between MNTB soma location and SPON bouton centroid, as measured by the slope of the linear regression line, showed no age-dependent change (ANCOVA: no significant interaction between age-range and soma location, F(2,19) = 0.05, p = 0.954).
Examples of MNTB axons and boutons in the SPON over development. A, Upper row, Low-magnification photograph and reconstruction of ventral brainstem with biocytin-filled cell body and axonal projection to the SPON and LSO. Dorsal is to the top, medial is to the left. Scale bar, 100 μm. Lower row, High-level magnification photograph and reconstruction of a portion of the MNTB axon and its boutons in the SPON within a single focal plane. The axon appears discontinuous as it moves in and out of the focal plane. Boutons are marked in red. Lower right panel, Plot of all boutons of this axon in the SPON, with the centroid marked in red. Scale bar, 100 μm. B, C, Examples cells form the age group P12–P14 (B) and P19–P21 (C). Scale bars are the same as in A. MNTB, medial nucleus of the trapezoid body; SPON, superior paraolivary nucleus; LSO, lateral superior olive. Scale bar, 20 μm.
The MNTB-SPON projection is topographically organized throughout early development. A, The median soma location of filled neurons within the MNTB (normalized to the mediolateral length of the MNTB), did not change over the first three weeks of postnatal development (no significant differences between P2 and P4, P12 and P14, and P19 and P21: F(2,22) = 1.1, p = 0.352, one-way ANOVA; number of cells (n) and animals (N) per age group: P2–P4: n = 8, N = 7, P12–P14: n = 9, N = 7, P19–P21: n = 8, N = 7). Data plotted as mean ± 95% CI. B, The median location of the bouton centroid of MNTB axons in the SPON (normalized to the mediolateral length of the SPON) is also not different between the three postnatal ages (F(2.22) = 0.51, p = 0.609, one-way ANOVA). Data plotted as mean ± 95% CI. C, Correlation of soma location in the MNTB with bouton centroid location in the SPON. At each age group, the MNTB-SPON projection shows a topographic organization, with a positive correlation between the location of cell bodies and bouton centroids along the mediolateral dimension of the MNTB and SPON, respectively (Pearson correlation coefficient: r = 0.64, p = 0.090, for P2–P4; r = 0.70, *p = 0.035, for P12–P14; r = 0.84, **p = 0.009, for P19–P21). The slopes of the three fitted lines were not different from each other, indicating that the topographic relationship did not change over the first three weeks of postnatal development (no significant interaction between age-range and soma location, F(2,19) = 0.05, p = 0.954, ANCOVA).
Over the first three postnatal weeks, we observed a slight, but statistically not significant, increase in the total length of axon collaterals in the SPON (Fig. 3A). As the cross-sectional SPON area increased ∼2-fold during the first two postnatal weeks (Fig. 3B), the density of collaterals in the SPON decreased slightly but statistically not significantly (Fig. 3C). The most pronounced changes occurred in the number of boutons formed by individual MNTB axons in the SPON, which increased over 2.5-fold during the period before hearing onset (Fig. 3D). Because this increase in the number of axonal boutons was not accompanied by significant axonal growth, the bouton density of individual axons increased significantly from P2–P4 to P12–P14 (Fig. 3E) indicating that new boutons primarily reflect the formation of new en passant synapses. Notably, from P12–P14 to P19–P21, the first week after hearing onset, the bouton density decreased significantly, indicating the presence of synapse pruning in the MNTB-SPO pathway after hearing onset.
Development of individual MNTB axons in the SPON. A, The absolute length of MNTB axon branches in the SPON does not significantly change with development [no significant differences between P2–P4, P12–P14, and P19–P21: F(2,22) = 1.16, p = 0.333; one-way ANOVA, number of cells (n) and animals (N) per age group: P2–P4: n = 8, N = 7, P12–P14: n = 9, N = 7, P19–P21: n = 8, N = 7]. B, The SPON cross-sectional area increases before but not after hearing onset [P2–P4 significantly different from P12–P14 and P19–P21: F(2,22) = 37.04, ***p < 0.001, one-way ANOVA, followed by Tukey's HSD: P2–P4 vs P12–P14, ***p < 0.001, P2–P4 vs P19–P21, ***p < 0.001, P12–P14 vs P19–P21, p = 0.659, number of cells (n) and animals (N) per age group: P2–P4: n = 8, N = 7, P12–P14: n = 9, N = 7, P19–P21: n = 8, N = 7]. C, The length of MNTB axon branches in the SPON normalized to SPON cross-sectional area does not change with development (F(2,22) = 3.29, p = 0.056, one-way ANOVA). D, The number of boutons per axon in the SPON at P2–P4 increases before but not after hearing onset [F(2,22) = 8.63, **p = 0.002, one-way ANOVA, followed by Tukey's HSD: P2–P4 vs P12–P14, **p = 0.002; P2–P4 vs P19–P21, *p = 0.017; P12–P14 vs P19–P21, p = 0.641, number of cells (n) and animals (N) per age group: P2–P4: n = 8, N = 7, P12–P14: n = 9, N = 7, P19–P21: n = 8, N = 7]. E, The Axonal bouton density (# of boutons normalized to axon length) changes with development (F(2,22) = 20.24, ***p < 0.001, one-way ANOVA). Axonal bouton density increases early in development (P2–P4 vs P12–P14, ***p < 0.001, Tukey's HSD) and decreases later in development (P12–P14 vs P19–P21, *p = 0.016, Tukey's HSD), although not returning to earlier development values (P2–P4 vs P19–P21, **p = 0.010, Tukey' HSD). Number of cells (n) and animals (N) per age group: P2–P4: n = 8, N = 7, P12–P14: n = 9, N = 7, P19–P21: n = 8, N = 7). Data plotted as mean ± 95% CI.
Development of topographic organization of MNTB-SPON terminals
To investigate whether bouton pruning changes the topographic precision of developing MNTB-SPON connections, we analyzed the spatial distribution of axonal boutons from individual MNTB-SPON axons. We first determined the area of the SPON covered by the boutons of individual axons by calculating the area of an ellipse that was fitted to the bouton cloud (Fig. 4A; Clause et al., 2014). Over the first three postnatal weeks, the bouton areas increased significantly (Fig. 4B). The spread of boutons along the SPON frequency axis, which is oriented primarily mediolaterally (Kulesza et al., 2003), significantly increased after hearing onset (Fig. 4C). When the bouton areas were normalized to the age-corresponding cross-sectional size of the SPON, which also increased during this time period (Fig. 3B), bouton areas remained unchanged (Fig. 4D). Still, the increase in the tonotopic spread after hearing onset persisted (Fig. 4E), suggesting a decrease in the tonotopic precision after hearing onset.
Area covered by the boutons of individual MNTB axons and tonotopic spread of MNTB boutons in the developing SPON. A, Examples of bouton clouds and fitted ellipses for calculating bouton areas (B, D; solid black curved line, outline of SPON; black dots, location of boutons; dotted gray line, ellipse-circumscribed bouton area; scale bar, 50 μm; D, dorsal; L, lateral). The ellipse was used to calculate the tonotopic spread (C, E) as the maximum distance between boutons along a line oriented parallel to the minor axis of the ellipse field. B, The absolute bouton area increased with development [P2–P4 significantly different from P19–P21: F(2,22) = 5.94, **p = 0.009, one-way ANOVA, followed by Tukey's HSD: P2–P4 vs P12–P14, p = 0.193, P2–P4 vs P19–P21, **p = 0.006, P12–P14 vs P19–P21, p = 0.211, number of cells (n) and animals (N) per age group: P2–P4: n = 8, N = 7, P12–P14: n = 9, N = 7, P19–P21: n = 8, N = 7]. C, The absolute tonotopic spread of boutons in SPON increased with development [P2–P4 and P12–P14 significantly different from P19–P21: F(2,22) = 6.51, **p = 0.006, one-way ANOVA, followed by Tukey's HSD: P2–P4 vs P12–P14, p = 0.989; P2–P4 vs P19–P21, *p = 0.012, P12–P14 vs P19–P21, *p = 0.013, number of cells (n) and animals (N) per age group: P2–P4: n = 8, N = 7, P12–P14: n = 9, N = 7, P19–P21: n = 8, N = 7]. D, The bouton area did not change with development (F(2,22) = 2.26, p = 0.128, one-way ANOVA). E, The relative tonotopic spread of boutons in SPON significantly increased at posthearing age [P12–P14 significantly different from P19–P21: F(2,22) = 4.23, *p = 0.028, one-way ANOVA, followed by Tukey's HSD: P2–P4 vs P12–P14, p = 0.219; P2–P4 vs P19–P21, p = 0.512; P12–P14 vs P19–P21, *p = 0.023, number of cells (n) and animals (N) per age group: P2–P4: n = 8, N = 7, P12–P14: n = 9, N = 7, P19–P21: n = 8, N = 7]. Data plotted as mean ± 95% CI.
Because SPON neurons are proposed to contribute to the directional selectivity of frequency-modulated sweeps of IC neurons (Pollak et al., 2011), we next examined possible developmental changes in asymmetries of bouton density along the tonotopic axis in the SPON, i.e., whether MNTB axon boutons were distributed evenly across frequency space or more concentrated in high or low-frequency portions of the termination region. Polar plots of the location of boutons within the SPON relative to each axon's bouton centroid (Fig. 5) indicate that bouton termination fields were elongated dorsoventrally, with an ∼22° dorsomedial to ventrolateral rotation. The smaller number of boutons per axon at P2–P4 is clearly evident (Fig. 5A), but the relative distribution of boutons among octants did not change from P2–P4 to P19–P21 (Fig. 5B). These results indicate that the orientation of future frequency bands is already present shortly after birth and does not undergo further developmental refinement.
Polar plots of the distribution of the boutons of individual MNTB axons in the SPON around the bouton centroid. A, Distribution of absolute bouton number around the centroid. Radial coordinates denote the average number of boutons found within the corresponding octant. Median (solid line) and 95% CI (shaded area). B, Normalized distribution of boutons around the centroid. Radial coordinates denote the average percentage of the total number of boutons per axon found within the corresponding octant. There was no significant difference in the relative distribution between ages [one-way ANOVA, number of cells (n) and animals (N) per age group: P2–P4: n = 8, N = 7, P12–P14: n = 9, N = 7, P19–P21: n = 8, N = 7]. D, dorsal; L, lateral; V, ventral; M, medial.
We next analyzed the spatial distribution of boutons in the termination field as the presence or absence of “hot spots” of high bouton density could have implications for the transfer of information between the MNTB and SPON, even in the absence of changes in the overall tonotopic specificity of the projection. To this end, we computed “heat maps” of bouton densities by aligning the bouton centroids of all axons (Fig. 6A). At P2–P4, bouton density was flat over the termination area. In contrast, at P12–P14 and P19–P20, bouton density was highest in the center of the termination area and declined gradually with distance. This indicates that before heating onset, new boutons are added preferentially in the center of the termination regions (Fig. 6B). After the onset of hearing, boutons were added and eliminated across the whole termination field. Still, the greatest bouton elimination occurred near the center, which resulted in a decrease in the central density peak during the third postnatal week.
Heat maps of the density of MNTB boutons in the SPON and distribution of boutons along the SPON frequency axis over development. A, Average bouton density in the SPON at each developmental age. Color bar, average number of boutons per 0.14% SPON area; scale bars, 17.5% of SPON mediolateral length and dorsoventral height; dorsal is to the top and lateral to the right. B, Change in the bouton density over development. Prehearing change: P2–P4 to P12–P14. Posthearing change: P12–P14 to P19–P21. Color bar, average change in bouton number per 0.14% SPON area; scale and orientation same as in A. C, Histograms with superimposed normal distributions fitted to the total number of boutons along the SPON frequency axis (% M-L SPON length; negative distances, medial; positive distances, lateral). Number of cells (n) and animals (N) per age group: P2–P4: n = 8, N = 7, P12–P14: n = 9, N = 7, P19–P21: n = 8, N = 7). D, Empirical cumulative distribution functions evaluated at the tonotopic distances from the bouton centroid (%), using the total number of boutons as in C. The three ages studied have significantly different bouton distributions along the tonotopic axis, with P12–P14 having the largest maximum distribution differences from both P2–P4 and P19–P21 [two-sample Kolmogorov–Smirnov test: D = 0.172, ***p < 0.001, for P2–P4 vs P12–P14; D = 0.151, ***p < 0.001, for P12–P14 vs P19–P21, number of cells (n) and animals (N) per age group: P2–P4: n = 8, N = 7, P12–P14: n = 9, N = 7, P19–P21: n = 8, N = 7], and P2–P4 having a smaller but still significant distribution difference with P19–P21 (two-sample Kolmogorov–Smirnov test: D = 0.061, *p = 0.023).
To analyze the tonotopic distributions quantitatively, we collapsed the data in the dorsoventral dimension and plotted bouton numbers with respect to distance from the centroid along the SPON frequency axis (Fig. 6C,D). The distributions along the tonotopic axis shared the same general shape at all ages, exhibiting a peak located in the center and a gradual decline with distance from the centroid. At all ages, bouton distributions were well-fit with Gaussian functions. Like the heat maps, the distribution is flattest at P2–P4, and the addition of boutons before hearing onset occurs predominantly near the center of termination bands, causing the center peak to increase almost 4-fold between P2–P4 and P12–P14 (Fig. 6C). As a result, the tonotopic spread at half the maximum peak height of the Gaussian was reduced by ∼30%. Cumulative plots of the distribution of boutons along the frequency axis indicate a significant left shift of the P12–P14 distribution, indicating higher bouton densities near the centroid at P12−P14 than at both P2–P4 and P19–P21 (Fig. 6D). These results show that the structural development of the MNTB-SPON pathway during the first three postnatal weeks is characterized by the addition of new boutons in the center of the termination field that resulted in a transient peak at the center of the termination field at hearing onset, which subsequently is reduced to prehearing levels during the first week of hearing.
Absence of functional elimination and strengthening of MNTB-SPON connection
Previous studies of the development of connectivity between the MNTB and LSO demonstrated the presence of functional synapse elimination before hearing onset despite a continuous growth of MNTB axons and the formation of new boutons (Kim and Kandler, 2003; Clause et al., 2014). To test whether such structural-functional disparity also occurs at MNTB axon collaterals terminating in the SPON, we performed whole-cell recordings from SPON neurons in slices to estimate the number of functional MNTB axon inputs that are received by individual SPON neurons. Electrical stimulation of presynaptic MNTB axons at the lateral border of the MNTB with gradually increasing stimulation intensities resulted in a gradual increase of postsynaptic current amplitudes, consistent with the recruitment of an increasing number of axonal inputs (Fig. 7A,B). To objectively identify discrete peaks corresponding to the successive recruitment of individual fibers, we fitted the frequency histograms of response amplitudes using Gaussian mixture models with the Bayesian information criterion (see Materials and Methods). The distance between peaks depicts the strength of individual inputs (Fig. 7B). This analysis revealed that, on average, each SPON neuron was innervated by approximately six MNTB axons and, importantly, that this number did not change significantly from P2–P4 to P19–P21 (H(3) = 7.14, p = 0.067, Kruskal–Wallis test; Fig. 7C). Furthermore, we found no significant change in the mean strength of single fiber inputs (H(3) = 3.75, p = 0.290, Kruskal–Wallis test; Fig. 7D) or the maximal amplitudes elicited by maximal stimulus intensities (F(3,35) = 1.12, p = 0.354, one-way ANOVA; Fig. 7E). These results argue against the functional refinement of the MNTB-SPON pathway by elimination and strengthening of connections.
Absence of functional elimination and strengthening of MNTB-SPON connections. A, Schematic of experimental approach with electrical stimulation (stim) and recording electrode (rec). Increasing the electrical stimulation intensity increases the number of recruited MNTB axons. B, Example traces of evoked IPSCs (top) and IPSC amplitude-stimulation intensity plots with frequency histograms fitted by Gaussian mixture models (right; *, Gaussian center). The number of Gaussians indicates the number of input fibers and the distance between Gaussian peaks represents individual fiber strength. C, The number of input fibers (H(3) = 7.14, p = 0.067, Kruskal–Wallis test). D, Individual fiber strength (H(3) = 3.75, p = 0.290, Kruskal–Wallis test). E, Maximal response amplitude (F(3,35) = 1.12, p = 0.354, one-way ANOVA). Number of cells (n) and animals (N): P2–P4 (n = 10, N = 5), P7–P9 (n = 9, N = 4), P12–P14 (n = 11, N = 6), P19–P21 (n = 9, N = 7). Data are presented as median with 95% boundaries (C, D) or mean ± 95% CI (E).
Previous studies have shown that developing MNTB connections to the LSO express activity-dependent strengthening (Kotak and Sanes, 2014; Bach and Kandler, 2020) and weakening (Kotak and Sanes, 2014), which is thought to result in long-lasting changes in the strength of individual inputs. Such strengthening and weakening of individual MNTB fiber inputs could be balanced across all inputs received by individual SPON neurons, which would not be revealed in the mean fiber strength. Plotting the response amplitudes of only the strongest fiber input for each SPON neuron, however, showed no significant changes between age groups (F(3,35) = 1.28, p = 0.295, one-way ANOVA; Fig. 8A). We also did not observe age-dependent changes in the coefficient of variation of input strengths of all individual fibers (F(3,35) = 0.89, p = 0.458, One-way ANOVA; Fig. 8B). In addition, the distribution of individual fiber strengths, normalized to the strongest fiber's amplitude for each SPON neuron, showed a uniform distribution of input strengths for all ages (Fig. 8C), and there was no difference in the distributions between age groups (H(3) = 4.12, p = 0.249 p = 0.458, Kruskal–Wallis test). These results argue against a developmental adjustment of the strength of specific MNTB fiber inputs to SPON neurons and provide further evidence against the presence of developmental strengthening and weakening of MNTB-SPON connections from shortly after birth to three weeks of age.
The strength of MNTB-SPON connections is uniformly distributed at all ages groups. A, No developmental change of strongest fiber amplitudes (F(3,35) = 1.28, p = 0.295, one-way ANOVA). B, Coefficient of variation of all fiber strength per SPON neuron did not change over development (F(3,35) = 0.89, p = 0.458, one-way ANOVA). C, Top, Distribution of individual fiber strengths normalized to the highest fiber's strength for each SPON neuron. Fibers for each SPON neuron were plotted in the same row. Bottom, Cumulative frequency plots for normalized individual fiber strengths. The strongest fiber of each cell was excluded from the analysis. Normalized fiber strengths were uniformly distributed and indistinguishable among groups (H(3) = 4.12, p = 0.249, Kruskal–Wallis test). Data in A and B are presented as mean ± 95% CI. Number of cells (n) and animals (N): P2–P4 (n = 10, N = 5), P7–P9 (n = 9, N = 4), P12–P14 (n = 11, N = 6), P19–P21 (n = 9, N = 7). Data are presented as mean ± 95% CI.
During the first two postnatal weeks, onset latencies of synaptic currents decreased from 2.3 ms at P2–P4 (median, 95% CI bounds 2.1, 2.5, n = 10, N = 5) to 1.6 ms at P12–P14 (median, 95% CI bounds 1.5, 1.8, n = 11, N = 6) with no further decrease after hearing onset (P19–P21, 1.5 ms median, 95% CI bounds 1.4, 1.6, n = 9, N = 7; H(3) = 31.2, p < 0.001, Kruskal–Wallis test, followed by Dunn's test; P2–P4 vs P12–P14, p < 0.001, P12–P14 vs P19–P21, p > 0.999). Likewise, decay times of synaptic currents became significantly shorter during the first two postnatal weeks with no further shortening of their decay times thereafter (Fig. 9A,B). Because SPON neurons maintain a very low intracellular Cl- concentration because of a high expression and activity of the K+ -Cl– cotransporter, KCC2 (Löhrke et al., 2005; Kopp-Scheinpflug et al., 2011), the experimentally imposed high intracellular chloride concentration in our whole-cell patch clamp recordings may be counteracted by a developmental increase in KCC2 activity. This may have resulted in a lower driving force and hence underestimated fiber strength in older animals. However, we observed no changes in the reversal potential of MNTB-elicited currents, which at all ages remained close to the value predicted by the Nernst equation for our intracellular and extracellular chloride concentrations (Fig. 9C,D). We also observed no developmental changes in the whole cell capacitance (Fig. 10), which may indicate that the cell bodies of SPON neurons already reached their mature sizes shortly after birth. However, like the maturation of synaptic responses, the developmental increase in the amplitude of the hyperpolarizing-activated cation current (Ih; F(3,37) = 5.62, p = 0.003, one-way ANOVA; P2–P4 vs P7–P9, p = 0.024, P2–P4 vs P12–P14, p = 0.002, P12–P14 vs P19–P21, p = 0.166, Tukey's test; Fig. 10B) and the decrease in the membrane resistance (Rm; H(3) = 10.31, p = 0.016, Kruskal–Wallis test; P2–P4 vs P12–P14, p = 0.036, P2–P4 vs P19–P21, p = 0.033, P12–P14 vs P19–P21, p > 0.999, Dunn's test; Fig. 10C) were completed around hearing onset with no significant changes occurring thereafter.
Maturation of synaptic MNTB-SPON responses. A, Averaged traces of MNTB-evoked IPSCs in SPON neurons from P2–P4 to P19–P21. Averaged traces were normalized and aligned to the peak current. Responses with multiple peaks were excluded from the analysis. B, Age-dependent decrease in decay time constant (τ; F(3,30) = 20.00, p < 0.001, one-way ANOVA; P2–P4 vs P7–P9, p = 0.11, P2–P4 vs P12–P14, p < 0.001, P2–P4 vs P19–P21: p < 0.001, P7–P9 vs P12–P14, p = 0.009, P7–P9 vs P19–P21: p = 0.004, P12–P14 vs P19–P21, p = 0.999, Tukey's test). Number of cells (n) and animals (N): P2–P4 (n = 10, N = 5), P7–P9 (n = 7, N = 4), P12–P14 (n = 8, N = 6), P19–P21 (n = 9, N = 7). Data are presented as mean ± 95% CI. C, IPSC peak amplitudes at different membrane holding potentials. D, Reversal potentials of MNTB-evoked IPSCs were not age dependent (H(3) = 5.65, p = 0.130, Kruskal–Wallis test). Number of cells (n) and animals (N): P2–P4 (n = 8, N = 4), P7–P9 (n = 9, N = 3), P12–P14 (n = 11, N = 6), P19–P21 (n = 9, N = 7). Data are presented as mean ± 95% CI (A–C) or median with 95% CI boundaries. **p < 0.01, ***p < 0.001 in post hoc Tukey's test.
Development of intrinsic membrane properties of SPON neurons. A, Top, Example current traces to steps to negative voltage steps (from −68 to −118, −98, or −78 mV; dark to light color traces). Bottom, Averaged I-V plots (n = 10 for P2–P4, 10 for P7–P9, 11 for P12–P14, 10 for P19–P21). Current amplitudes were measured at 19 ms after the start (open circle) and 12 ms before the end (closed circle) of voltage steps. Vertical bars indicate IH at −118 mV. B, Developmental changes in IH amplitude for voltage steps to −118 mV (F(3,37) = 5.62, p = 0.003, one-way ANOVA; P2–P4 vs P7–P9, p = 0.024, P2–P4 vs P12–P14, p = 0.002, P12–P14 vs P19–P21, p = 0.166, Tukey's test). C, Developmental decrease in membrane input resistance (H(3) = 10.31, p = 0.016, Kruskal–Wallis test; P2–P4 vs P12–P14, p = 0.036, P2–P4 vs P19–P21, p = 0.033, P12–P14 vs P19–P21, p > 0.999, Dunn's test). D, Membrane capacitance (Cm), remained unchanged during development (F(3,37) = 0.27, p = 0.847, one-way ANOVA). Data presented as mean ± 95% CI (A, B, D) or median with 95% CI boundaries C. Number of cells (n) and animals (N; B–D): P2–P4 (n = 10, N = 5), P7–P9 (n = 10, N = 4), P12–P14 (n = 11, N = 6), P19–P21 (n = 10, N = 7). *p < 0.05, **p < 0.01 in post hoc test.
Stable MNTB-SPON input maps before hearing onset
Although our electrophysiological experiments found no evidence for a postnatal refinement of MNTB-SPON connections, it is still possible that topographic sharpening occurs via a concentration of presynaptic MNTB neurons to more topographically restricted MNTB areas. This scenario could be achieved by the simultaneous formation and elimination of connections, which would have remained undetected by estimating input numbers with electrical stimulation, which does not provide information about the location of input neurons in the MNTB. To test for this possibility, we used laser scanning photostimulation (LSPS) with caged glutamate (Kim and Kandler, 2003; J. Sturm et al., 2014) in slices from P2–P4 and P12–P14 animals (Fig. 11A) to determine the location of neurons in the MNTB that provide synaptic input to individual SPON neurons (MNTB-SPON input maps). In agreement with our anatomic results, we observed a clear relationship between the mediolateral location of the recorded SPON neuron and the mediolateral location of presynaptic MNTB neurons (Fig. 11B,C). This topographic relationship was already present at P2–P4 and did not change subsequently (Fig. 11C), consistent with our anatomic data (Fig. 2C). Importantly, the size of MNTB-SPON input maps did not change between P2–P4 and P12–P14, both in absolute value and when normalized to the MNTB cross-sectional area (Fig. 11D–F). Similarly, the absolute or normalized mediolateral width (spread along the tonotopic axis) of input maps showed no change during the period before hearing onset (Fig. 11G–I). The synaptic response strength (charge density) along the tonotopic axis was well fit by a Gaussian curve and, importantly, did not change its width during development (P2–P4, R2 = 0.65, FWHM = 18.9%; P12–P14, R2 = 0.49, FWHM = 16.32%; Fig. 11J). Taken together, these functional mapping experiments demonstrate that the topographic relationship between the MNTB and SPON on a functional level is already established shortly after birth (P2–P4) without refinement during the first two postnatal weeks.
The topographic organization of functional MNTB-SPON input maps remains unchanged before hearing onset. A, Schematic illustration of mapping MNTB-LSO connections with focal glutamate uncaging. B, Examples of MNTB input maps (left) and their surface plot (right) for neurons recorded in the medial, central, and lateral SPON (top to bottom) at the age of P2–P4 (Bi) and P12–P14 (Bii). Surface plots illustrate a region within the MNTB (black outline) eliciting responses to the recorded SPON neuron. A red circle indicates the recording site in the SPON. Scale bar, 200 µm. C, Topographic relationship between the locations of MNTB input hotspots against recorded SPON neurons. Location is normalized to the lateral border of the MNTB and dorsolateral border of the SPON (∼15°; slope: F(1,10) = 0.594, p = 0.459, y-intercept: F(1,11) = 0.823, p = 0.384, ANCOVA). D, Area of MNTB input maps remains unchanged between P2–P4 and P12–P14 (t(12) = 0, p > 0.999, two-tailed unpaired t test). E, Area of the MNTB cross-sectional area (t(12) = 1.95, p = 0.074, two-tailed unpaired t test). F, Normalized MNTB input areas (t(12) = 1.02, p = 0.328, two-tailed unpaired t test). G, Mediolateral length of the inputs maps (t(12) = 1.29, p = 0.221, two-tailed unpaired t test). H, Mediolateral length of the MNTB (U = 4.00, p = 0.006, Mann–Whitney test). I, Normalized tonotopic width of input maps (t(12) = 0.40, p = 0.695, two-tailed unpaired t test). J, Normalized synaptic charge density along the proportional mediolateral distance from the location of the largest synaptic response. Synaptic charge density was collapsed along the mediolateral tonotopic plane and normalized to the maximum in each neuron. Data were fitted with a Gaussian function. Black lines indicate the full-width at half maximum (FWHM) of the Gaussian fits (P2–P4, R2 = 0.65, FWHM = 18.9%; P12–P14, R2 = 0.49, FWHM = 16.32%). Data are shown as mean ± 95% CI (D–G, I) or median with 95% CI boundaries (H). Number of cells (n) and animals (N; C–I): P2–P4 (n = 7, N = 6), P12–P14 (n = 7, N = 5), **p < 0.01.
Discussion
In this study, we characterized the anatomic and functional connectivity from the MNTB to the SPON in mice from shortly after birth until three weeks of age. This developmental period includes the time before hearing onset when activity in the MNTB is driven by cochlea-generated spontaneous activity (first two postnatal weeks; Tritsch et al., 2007; Sonntag et al., 2009; Clause et al., 2014) and the first week of hearing when MNTB neurons are driven by sound-evoked activity (third postnatal week; Sonntag et al., 2009). Our anatomic and physiological studies revealed the absence of topographic refinement during the period before hearing onset and, unexpectedly, a decrease in the topographic precision during the first week of hearing. These results provide new insight into the organization and development of inhibitory synaptic inputs to the SPON and further demonstrate that divergent axon collaterals of MNTB neurons projecting to functionally different nuclei exhibit distinct forms and timelines of refinement.
MNTB projections to the binaural nuclei LSO and MSO exhibit a highly precise tonotopic organization that is aligned with excitatory glutamatergic projections from the cochlea nucleus (Sanes and Rubel, 1988; Tollin, 2003; Pecka et al., 2008). In contrast, the tonotopic organization of the SPON has been somewhat controversial, with some in vivo recordings finding a tonotopic organization (Kulesza et al., 2003) while others reported the absence of any tonotopy in the SPON (Dehmel et al., 2002). Our anatomic results from three-week-old mice confirm and extend previous tracing studies (Banks and Smith, 1992; Sommer et al., 1993; Saldaña and Berrebi, 2000) that found a topographic organization from the MNTB to the SPON with an axis that runs from dorsolateral (low frequency) to ventromedial (high frequency). Because synaptic inhibition by MNTB inputs constitutes the major spike-generating drive to SPON neurons via the generation of rebound action potentials (Felix et al., 2011; Kopp-Scheinpflug et al., 2011), the topography of the MNTB-SPON organization results in a corresponding tonotopic axis in the SPON. Our results also reveal that this tonotopic organization is much less precise than that in the LSO or MSO. In the LSO, individual MNTB axons innervate only ∼14% of the tonotopic axis (Clause et al., 2014), while in the SPON, individual MNTB axons innervate on average ∼60% of the tonotopic axis (Fig. 4E). Likewise, MNTB neurons that are presynaptic to an individual LSO neuron cover only ∼ 14% of the tonotopic axis, while presynaptic MNTB neurons to individual SPON neurons are distributed along ∼35% of the MNTB tonotopic axis. The convergence of approximately six, sharply tuned MNTB neurons with different frequency tuning provides a structural explanation for the resulting broad frequency tuning of MNTB-mediated offset responses in the SPON (Behrend et al., 2002; Dehmel et al., 2002) and provides a functional match for the innervation by only one or two, broadly tuned excitatory input from octopus cells in the cochlear nucleus (Felix et al., 2017).
Our developmental studies revealed that the topographic organization of the MNTB-SPON connectivity is already present shortly after birth, similar to what has been described for the MNTB pathways to the LSO (Sanes and Siverls, 1991; Kim and Kandler, 2003). During the following prehearing period, MNTB-SPON axons exhibited extensive growth and a 3-fold increase in the number of boutons. However, this growth was matched by the expansion of the SPON, which resulted in an unchanged topographic precision. New boutons were not added uniformly over an MNTB axon's termination area but were added preferentially to the center, resulting in the build-up of a bouton-dense center area. This pattern mirrors the formation of new boutons in the MNTB-LSO pathway before hearing onset (Clause et al., 2014).
In the SPON, the lack of topographic refinement on the structural level was paralleled on the functional level. During the first two postnatal weeks, both the number and strength of input fibers and the size of MNTB-SPON input maps remained unchanged (Fig. 11). Such topographic stability is in sharp contrast to the topographic refinement that occurs in the MNTB-LSO pathway, where there is extensive synaptic silencing leading to the functional disconnection of the majority of presynaptic MNTB neurons and to an over 2-fold increase in the topographic precision (Kim and Kandler, 2003; Clause et al., 2014). The cellular or synaptic mechanisms that regulate the stark differences in functional refinement exhibited by MNTB axon collaterals in these target nuclei are unknown.
A possible reason for the refinement differences may be differences in the temporal relationship of excitatory and inhibitory activity in the SPON and LSO before hearing onset. In addition to inputs from the MNTB, both nuclei also receive converging excitatory inputs from the cochlear nucleus. In the SPON, activity in both inputs is driven by the contralateral cochlea (Magnusson and Gómez-Álvarez, 2019) and, therefore, is coincident even before hearing onset when cochlear activity is generated spontaneously. This coincident activity among glutamatergic cochlear nucleus and glycinergic MNTB inputs may reinforce MNTB-SPON connections, as has been described for MNTB connections in the MSO (Winters and Golding, 2018), protecting them from being silenced or eliminated. In contrast, in the LSO, glutamatergic and glycinergic activity is driven by the ipsilateral and contralateral cochlea, respectively (Sanes, 1990; Kandler and Friauf, 1995; Tollin, 2003) and thus expected to be less coincident (Babola et al., 2018), making MNTB-LSO synapses more prone to undergo activity-dependent weakening and being silenced (Kotak and Sanes, 2014). It also may be possible that SPON neurons lack the mechanisms required for activity-dependent long-term depression or potentiation. However, many properties of MNTB-LSO synapses involved in activity-dependent synaptic plasticity or tonotopic refinement, such as glutamate co-release (Gillespie et al., 2005; Kalmbach et al., 2010; Noh et al., 2010), expression of neurotrophins and their receptors (Hafidi et al., 1996; Kotak et al., 2001; Tierney et al., 2001), or GABA co-release (Kotak et al., 1998; Chang et al., 2003; Kim and Kandler, 2010; Bach and Kandler, 2020) all seem to be also present in SPON neurons (Hafidi et al., 1996; Tierney et al., 2001; Blaesse et al., 2005; Gillespie et al., 2005) making this a more unlikely explanation for the lack of refinement in the SPON. However, a striking difference during the first postnatal week is how SPON and LSO neurons respond to GABA and glycine. Because of their high intracellular chloride concentration, LSO neurons respond to MNTB stimulation with membrane depolarization, an increase in the intracellular calcium concentration, and triggering action potentials (Kandler and Friauf, 1995; Ehrlich et al., 1999; Kakazu et al., 1999; Kullmann and Kandler, 2001; Kullmann et al., 2002; Lee et al., 2016), which is often linked to activity-dependent synaptic plasticity (Caillard et al., 1999; Deidda et al., 2015; Brady et al., 2018). This is in stark contrast to SPON neurons, in which GABA and glycine elicit hyperpolarizing responses already at birth because of the early expression of functional KCC2 transporters (Löhrke et al., 2005; Kopp-Scheinpflug et al., 2011). Since an excitatory action of GABA and glycine is associated with synaptic plasticity in a number of systems, the early hyperpolarizing response to GABA and glycine (Rivera et al., 1999; Lee et al., 2005; Blaesse et al., 2006) may be a mechanism contributing to the stabilization of the SPON's topographic organization during the prehearing period.
During the first week after hearing onset, we observed both pruning and new formation of MNTB axonal boutons in the SPON. Pruning occurred preferentially in the center of the termination area while boutons were added to peripheral regions, resulting in wider bouton terminal areas, i.e., a decrease in topographic precision (Fig. 6C). Our findings differ from a previous study which did not find significant changes in the topographic precision of MNTB-SPON connections after hearing onset (Rajaram et al., 2020). This discrepancy may be because of differences in the analysis or sample size. The present study analyzed axonal bouton distribution, while the previous study analyzed axon morphology (length, branch points, and hull). Because of changes in bouton density along MNTB-SPON axons during development, an analysis of axon morphology alone may not fully capture changes on the synaptic contact level.
Importantly, the MNTB axon bouton dynamic in the SPON is in stark contrast to the massive bouton elimination occurring on MNTB axons in the LSO during the same period, which occurs nonselectively across the entire termination area, resulting in a sinking ice-berg effect and an increase in topographic precision (Clause et al., 2014). It has been suggested that bouton pruning in the LSO involves the structural elimination of previously silenced (functionally pruned) connections. However, in the absence of such input silencing in SPON, a center-biased pruning of MNTB-SPON boutons likely involves different guiding mechanisms. The decrease in the tonotopic precision of the MNTB-SPON pathway after hearing onset is unexpected, considering that topographic changes in the developing central auditory system generally increase rather than decrease the tonotopic precision (Zhang et al., 2001; Jones et al., 2007; Yu et al., 2007; Werthat et al., 2008; Kandler et al., 2009; J. Sturm et al., 2014). However, the function of the SPON is the detection of sound offsets and the precise encoding of the temporal envelope of complex sounds, for which precise frequency tuning is less important.
A plausible function for the refinement of MNTB axons in the SPON, therefore, could be to increase the strength and to broaden the frequency tuning of inhibition, which arises from six narrowly-tuned MNTB neurons, to match the strength and the broad tuning of excitation, which is generated by one or two strong and broadly tuned octopus cells (Felix and Magnusson, 2016; Felix et al., 2017). SPON neurons typically respond with a transient spike discharge at the onset and/or offset of the sound (Kuwada and Batra, 1999; Behrend et al., 2002; Dehmel et al., 2002; Gómez-Álvarez et al., 2018). Glycinergic inhibition from the MNTB is essential for generating this response pattern by strongly hyperpolarizing SPON neurons during the sound stimulus. This hyperpolarization suppresses tonic spiking during the duration of sound, thus contributing to the generation of ON responses (Kulesza et al., 2007), and provides the hyperpolarization necessary for triggering rebound spiking (Felix et al., 2011; Kopp-Scheinpflug et al., 2011). ON responses are especially prominent for broadband sound stimuli and natural calls, where they encode the sound envelope. The unexpected tonotopic spread of MNTB axons in the SPON after hearing onset may optimize the functional balance of synaptic excitation and inhibition to ensure reliable call encoding soon after hearing onset (Gómez-Álvarez et al., 2018). Future experiments could address whether this tonotopic adjustment is guided by early hearing experience.
Footnotes
This work was supported by National Institute on Deafness and Other Communication Disorders Grants DC004199 (to K.K.) and DC019814 (to K.K.), the National Science Foundation IGERT Training Grant DGE 0549352 (to A.C.), and a Pennsylvania Lions Hearing Research Foundation grant (K.K.). We thank Flora Antunes for help with statistical analysis and feedback on an earlier version of the manuscript. We also thank Srivatsun Sadagopan for suggesting and supporting the implementation of the Gaussian mixture model and members of the Kandler lab for valuable discussions.
The authors declare no competing financial interests.
- Correspondence should be addressed to Karl Kandler at kkarl{at}pitt.edu
