Ambient sound stimulation tunes axonal conduction velocity by regulating radial growth of myelin on an individual, axon-by-axon basis

Significance

Myelin ensures fast and reliable conduction of action potentials along axons. Once myelin in the central nervous system is lost due to disease, injury, or aging, remyelination efforts by proliferating oligodendrocyte precursor cells occur, but often remain incomplete. Therefore, reaching and maintaining optimal myelination through sensory or social experience forms a crucial equilibrium. Using the auditory system as a model, the present work shows that activity-dependent myelination occurs at the level of individual axons and that activity in one input channel, i.e., one ear, contributes to proper myelination of activated axons. However, activity in one input channel cannot compensate for activity deficits in another, thereby causing imbalanced myelination and mismatched auditory temporal processing.

Abstract

Adaptive myelination is the emerging concept of tuning axonal conduction velocity to the activity within specific neural circuits over time. Sound processing circuits exhibit structural and functional specifications to process signals with microsecond precision: a time scale that is amenable to adjustment in length and thickness of myelin. Increasing activity of auditory axons by introducing sound-evoked responses during postnatal development enhances myelin thickness, while sensory deprivation prevents such radial growth during development. When deprivation occurs during adulthood, myelin thickness was reduced. However, it is unclear whether sensory stimulation adjusts myelination in a global fashion (whole fiber bundles) or whether such adaptation occurs at the level of individual fibers. Using temporary monaural deprivation in mice provided an internal control for a) differentially tracing structural changes in active and deprived fibers and b) for monitoring neural activity in response to acoustic stimulation of the control and the deprived ear within the same animal. The data show that sound-evoked activity increased the number of myelin layers around individual active axons, even when located in mixed bundles of active and deprived fibers. Thicker myelination correlated with faster axonal conduction velocity and caused shorter auditory brainstem response wave VI-I delays, providing a physiologically relevant readout. The lack of global compensation emphasizes the importance of balanced sensory experience in both ears throughout the lifespan of an individual.

The development of a bony skull combined with increased brain size during the evolution of the vertebrate nervous system required faster axonal conduction velocities than can be achieved by a simple increase in axon diameter. The role of myelin in supporting such an increase in conduction velocity was originally understood simply as a static physical insulator acting to increase resistance and reduce the time scale of membrane uncharging. However, myelinated axons are highly plastic, shaping firing rates, conduction velocity, and reducing energy consumption of axonal conduction (1, 2). Thus, myelination is now considered an integral part of neuronal processing in the vertebrate brain, but the mechanisms underlying activity-dependent myelination are still elusive.

The function of myelin in sensory processing, cognition, and demyelinating neurological disorders is predominantly addressed by structural analysis of myelination patterns created by the specific sequence of nodes and internodes, as well as by the thickness of the myelin sheaths with respect to axon diameter. While the longitudinal pattern of myelin along axons is defined by neuronal identity (3, 4) and as such, optimizes the network behavior (5, 6), myelin thickness is generally defined by the canonical “g-ratio” (axon diameter/fiber diameter) with species-specific variations between 0.6 and 0.8 (7, 8). However, both myelin thickness and g-ratio change with axonal activity. Increased axonal firing rates following sensory maturation, skill learning, or social engagement (9–11) cause thicker myelination of the involved axons, while sensory or social deprivation results in thinner myelin sheaths (11–13). The mechanisms by which axonal activity levels are communicated to myelinating oligodendrocytes and what degree of change in myelin thickness is required to achieve a physiological objective are still largely unknown.

During development, increased neural activity regulates myelination (14, 15) via the release of soluble axonal factors such as adenosine (16), GABA (17) or glutamate (18), which subsequently lead to calcium-transients in nearby myelin sheaths and in the somata of myelinating oligodendrocytes. This calcium increase triggers enhanced myelination of active axons. Later in development, when myelin sheaths have already formed around their axon, activity-dependent calcium-transients were only observed in the myelin sheaths, but not in the oligodendrocyte somata (19). Therefore, once most parts of the axon are myelinated and vesicle secretion becomes restricted to the synaptic terminal, axon-glia communication needs to take a different form.

Release of ATP, GABA, or glutamate from synaptic terminals or the nodes of Ranvier requires a form of volume transmission to reach their receptors on the oligodendrocyte somata. While such nonspecific mechanisms would be suitable for homogenous fiber tracts where all axons that are myelinated by the same oligodendrocyte are equally active, little is known about whether adaptive myelination can occur at the level of individual axons, particularly within heterogeneous fiber tracts (20). This gap in knowledge most likely arises from the fact that axons within a fiber tract often originate from neurons of similar class-specific genetic identity, which attracts the same subpopulation of myelinating oligodendrocytes (21), conveys similar activity levels and dictates projections to similar targets. Together, this complicates the use of genetic markers for pharmaco- or optogenetics as well as input-specific neural activation. One recent study, using elegant pharmacogenetics, showed that individual active axons within a fiber tract will be preferentially myelinated during development, suggesting local axo-glial communication (22).

Here, we take advantage of the auditory system, which allows differential sensory stimulation of left- and right-side neural circuits. Two monaural fiber tracts were targeted: 1) the auditory nerve, formed by axons of spiral ganglion neurons, and 2) the trapezoid body (TB), formed by axons of globular bushy cells (GBC) which cross the midline and project to nuclei on the opposite side of the brainstem (5, 6, 11). Afferent auditory nerve fibers of the spiral ganglion neurons only contain axons arising from the same ear; however, within the TB, each bundle could contain fibers crossing in one direction only or it could contain a mixture of fibers originating from both sides of the head. An answer to this dilemma will further our understanding of how oligodendrocytes regulate myelination in two ways. First, if all fibers in a bundle arise from the same ear and possess similar activity levels, soluble factors released by active axons and distributed via volume transmission will likely reach oligodendrocyte somata and regulate myelination globally for all axons within the same bundle. Second, if axons within a bundle arise from different ears and possess different activity levels, a differential, local axon-by-axon regulation of myelin might be needed.

Neuronal firing in the auditory pathway consists of spontaneous activity and sound-evoked activity, the latter only starting at about P10 in mice. A mild and reversible (11) sensory deprivation approach was used by raising mice with monaural earplugs between P10 and P20. This allows the degree of myelination of individual axons within mixed-activity fiber bundles to be measured, while additionally creating an internal control by the nonplugged ear to assess the physiological significance of adaptive myelination.

Results

Sagittal brainstem sections show that TB fibers are organized in bundles, isolated from each other by perpendicularly arranged pyramidal tract fibers (Fig. 1C). To test the composition of TB fiber bundles, anterograde tracers of different colors (tetramethylrhodamine and biocytin) were injected into the cochlear nucleus of the control and the deprived ear (Fig. 1A). Coronal brainstem sections show that both tracers labeled TB fibers, approaching and crossing the midline from either side of the brain (Fig. 1B). Many of the fibers form giant calyx synapses upon targeting the medial nucleus of the trapezoid body (MNTB). In addition, some cell bodies are retrogradely labeled, consistent with an efferent projection of the MNTB to the ipsilateral cochlear nucleus (23). Coronal sections were counterstained with CC1, a marker commonly used for detecting mature oligodendrocytes (24). An 800-µm × 400-µm region of interest was defined left and right of the midline in coronal sections and the number of CC1-positive cells was counted. The side receiving active input contained 778 ± 57 oligodendrocytes/mm² (n = 3 mice), which was not significantly different from the deprived side with 751 ± 77 oligodendrocytes/mm² (n = 3 mice; paired t test: P = 0.174). There were no signs of membrane blebbing or nuclear fragmentation in the CC1-positive cells on either side, suggesting that the deprivation did not cause oligodendrocyte death (Fig. 1 B, Inset).

Fig. 1.

Radial growth of myelin occurs on an axon-by-axon basis. (A) Ventral view of a mouse brain illustrating the injection of different colored tracers into both cochlear nuclei to differentially trace fibers originating from either the deprived or the control ear. (B) Coronal brainstem section (reassembled) showing traced TB fibers from both sides approaching and crossing the midline, most of them terminating in calyx synapses in the contralateral MNTB. White arrow heads mark retrogradely labeled green MNTB cells that project to the cochlear nucleus. Inset: CC1-labeled oligodendrocytes are marked by the asterisks and are smaller than MNTB principal neurons (white dashed outlines). Oligodendrocyte numbers were counted in 800 × 400-µm areas (blue dashed boxes) left and right from the midline. Sagittal sections were taken near the midline, at ~200 µm and ~600 µm away from the midline on either side (white perpendicular boxes shown on one side only). (C) Sagittal section of the TB fiber tract shows cross-sections of auditory fiber bundles separated by perpendicularly oriented pyramidal tract fibers. (D) Sagittal TB section shows that fiber bundles contain fibers of both colors. White dotted lines indicate the boundaries of each bundle. (E) Magnification of the section shown in D also reveals fiber cross-sections lacking either tracer (blue MBP circles only). (F) Individual axon diameters (minimum diameter of the inner ellipse; n = 1,565 fibers) were measured for fibers at three lateral locations. Boxes show medians and interquartiles. Gray solid lines indicate the within mouse (n = 11) comparisons and black solid lines, the within bundle comparisons. (G) Same design as in F. Myelin thickness [(minimum outer diameter – minimum inner diameter)/2] is significantly larger in control compared to deprived fibers in all three locations.

Sagittal sections were taken at three positions: near the midline, ~200 µm lateral to the midline and ~600 µm lateral to the midline and the two colors of the tracers were then used to identify control or deprived fibers (Fig. 1 D and E). Inspection of eleven monaurally deprived, bilaterally traced brains revealed active and deprived fibers to form interspersed, salt-and-pepper like patterns within each fiber bundle (Fig. 1D). Counterstaining with antibodies against myelin basic protein (MBP) showed that not all TB fibers were traced via the cochlear nucleus injections (Fig. 1E). Based on the circular MBP labeling, axon diameter and myelin thickness were assessed by fitting two ellipses to each axon as previously described (11). To minimize bias, all axon and myelin measures were taken from monochromatic images showing only the MBP staining. Each axon was numbered so that it could be identified as active, deprived, or nonlabeled upon overlaying the tracing colors. Since nonlabeled fibers could potentially originate from the active or deprived side, they were disregarded for further analysis.

In agreement with Sinclair et al. (11), auditory deprivation instigated fewer large-caliber axons in the trapezoid body, particularly at locations further away from the midline (Fig. 1F). GBC axons undergo an increase in axon diameter as they approach the calyx synapse in the MNTB (6, 11). Sagittal sections taken at ~200 µm and ~600 µm from the midline at either side, include the MNTB and therefore contain axons that are close to forming calyx synapses. Our data revealed more large-caliber axons in the 200-µm and 600-µm sections for the control fibers, but not for the deprived fibers, resulting in significantly decreased overall axon diameters. Sagittal sections near the midline did not show a difference in axon diameter (near midline_control: 1.30 ± 0.45 µm vs. near midline_deprived: 1.29 ± 0.50 µm, P ≥ 1.000; 200 µm_control: 1.61 ± 0.59 µm vs. 200 µm_deprived: 1.37 ± 0.39 µm, P ≤ 0.001; 600 µm_control: 1.73 ± 0.86 µm vs. 600 µm_deprived: 1.27 ± 0.54 µm, P ≤ 0.001; n = 1,565 fibers; 11 mice; one-way repeated measures ANOVA followed by the Bonferroni t test; Fig. 1F). In addition to the changes in axon diameter, the present data show significantly thicker myelin in control compared to deprived fibers at all three locations (near midline_control: 0.61 ± 0.19 µm vs. near midline_deprived: 0.52 ± 0.16 µm, P ≤ 0.001; 200 µm_control: 1.04 ± 0.26 µm vs. 200 µm_deprived: 0.95 ± 0.16 µm, P ≤ 0.001; 600 µm_control: 1.04 ± 0.24 µm vs. 600 µm_deprived: 0.87 ± 0.17 µm, P ≤ 0.001; n = 1,565 fibers; 11 mice; one-way repeated measures ANOVA followed by the Bonferroni t test; Fig. 1G). Similar statistical results were achieved when differences in axon diameter and myelin thickness were analyzed within animals (gray lines in Fig. 1 F and G) or within fiber bundles (black lines in Fig. 1 F and G). The within bundle analysis was restricted to bundles containing more than 10 traced fibers of both colors to ensure statistical power (axon diameter_control: 1.43 ± 0.23 µm, axon diameter_deprived: 1.23 ± 0.18 µm, paired t test: P = 0.002; myelin_control: 0.95 ± 0.10 µm, myelin_deprived: 0.87 ± 0.08 µm, paired t test: P = 0.005). Together, these data suggest that activity-dependent adaptive myelination occurs on an individual fiber basis, even when they arise from neurons of an identical cell type, possessing the same target and run within the same fiber bundle.

Radial growth of myelin is achieved by adding new myelin layers starting from the inner tongue. High-magnification electron microscopy was used to measure the exact number of myelin layers in control and deprived fibers and to test for possible differences in individual myelin layer thickness. Since the injected traces were not electron-dense, we measured these parameters in the auditory nerve, which contains fibers solely from one ear. Cross-sections of the proximal auditory nerve from the control and the deprived ear were investigated using electron microscopy. Electron micrographs revealed individual auditory nerve axons, with inner tongues and myelin sheaths (Fig. 2A). Scans of rectangular sections were performed across the myelin sheaths to assess the overall myelin thickness, the number of myelin layers and the thickness of individual myelin layers. The regular peaks and troughs of the resulting histogram represent the extracellular intraperiod line (IPL, lighter shade) and the major dense line (MDL, darker shade), respectively (Fig. 2 B–D). Similar to the findings in TB fibers, auditory nerve fibers of the active, control ear had significantly thicker myelin sheaths (median: 0.205 µm; IQR: 0.175–0.244 µm; n = 42) compared to fibers arising from the deprived ear (median: 0.175 µm; IQR: 0.156–0.203 µm; n = 48; Mann–Whitney rank-sum test: P = 0.003; Fig. 2E). Recognizing the peak-to-peak distance of the scan histograms as one myelin layer, the average number of layers was significantly larger on the control side (30.24 ± 7.02; n = 42) compared to the deprived side (25.48 ± 6.53; n = 48; two-tailed t test: P = 0.001; Fig. 2F). Using the distance between two IPLs as a measure of layer thickness, no significant difference was found between active and sound-deprived fibers (control median: 4.81 nm, control IQR: 4.64–5.35 nm; n = 42; deprived median: 4.82 nm, control IQR: 4.59–5.16 nm; n = 48; Mann–Whitney rank-sum test: P = 0.389; Fig. 2 D and G). These results show that sound-evoked activity increases the number of myelin layers at active auditory axons. There were no signs of myelin swelling or degradation on the sound-deprived side, consistent with the increase in myelination on the active side being physiological rather than representing pathological degradation on the deprived side.

Fig. 2.

Ambient sound increases the number of myelin layers, not their thickness. (A) Electron micrograph of an auditory nerve fiber (ANF) visualizes the axon (green), mitochondria (mi), myelin sheath (red), and the inner tongue of myelin (it; yellow). (B) Section of myelin sheath showing only 6 (out of >20) individual myelin layers with the cytoplasmatic major dense line (MDL, dark lines) and the extracellular intraperiod line (IPL, lighter lines). (C) Histogram of all the peaks and troughs through the whole myelin sheath acquired by a line scan using ImageJ. (D) Schematic of MDL and IPL in a myelin sheath. (E) Thickness of ANF myelin excluding the inner tongue. (F) Number of myelin layers assessed from peaks in histograms such as shown in C. (G) Thickness of individual myelin layers calculated by trough-to-trough distances in histograms as in C.

It is conceivable that the earplug caused a developmental delay in downstream auditory neurons, so four established neurophysiological markers of maturation were tested for active and sound-deprived neurons (Fig. 3): 1) synaptic delay, 2) pre-, and 3) postsynaptic action potential (AP) half-width (25) and 4) lack of regular bursting of spontaneous activity (26, 27). With respect to the monaural earplug, cochlear nucleus neurons activated by the control ear send sound-evoked activity through TB fibers, across the midline toward their targets in the contralateral superior olivary complex (SOC). In contrast, cochlear nucleus neurons receiving input from the ear-plugged ear transmit only spontaneous APs along the TB fiber toward their SOC targets. In the SOC, neurons of the MNTB receive exclusive excitatory input from contralateral cochlear nucleus neurons and as such serve as an ideal within-subject control of activity changes following the monaural sensory deprivation. Due to the crossed inputs, we refer to MNTB neurons contralateral to the control ear as active and vice versa, MNTB neurons contralateral to the ear-plugged ear as deprived (Fig. 3A). Single-unit activity was recorded from the active and deprived MNTB within the same mice immediately following earplug removal. MNTB neurons were identified by their excitatory response to contralateral sound and the typical complex waveform comprising both presynaptic and postsynaptic components (28). This was used to obtain the half-width of the pre- and postsynaptic APs as well as the synaptic delay (Fig. 3B). The time between the peak and trough of extracellular APs is a recognized marker for AP half-width (29, 30). Presynaptic AP half-widths, synaptic delays and postsynaptic AP half-widths did not differ significantly between active and deprived MNTB neurons (preAP-HW_active: median: 0.18 ms; IQR: 0.14–0.19; n = 28; preAP-HW_deprived: median: 0.16 ms; IQR: 0.14–0.19 ms; n = 22; Mann–Whitney rank-sum test: P = 0.501; Fig. 3C, synaptic delay, SD_active: 0.45 ± 0.06 ms; n = 28; SD_deprived: 0.42 ± 0.04 ms; n = 22; two-tailed t test: P = 0.091; Fig. 3D and postsynaptic AP half-width, postAP-HW_active: median = 0.43 ms, IQR = 0.39–0.47, n = 28; postAP-HW_deprived: median: 0.45 ms, IQR: 0.37–0.49 ms, n = 22; Mann–Whitney rank-sum test: P = 0.672; Fig. 3E). Regularity of spontaneous firing was quantified by the coefficient of variation, which is high (~2) for prehearing activity and ~1 for firing activity in post-hearing-onset animals (27). Spontaneous activity of MNTB neurons from neither control nor deprived side showed any signs of bursting activity (Fig. 3 F and G) and their coefficients of correlation clustered around 1 with no significant difference between the groups (CV_active: median: 0.76; IQR: 0.66–0.83; n = 28; CV_deprived: median: 0.79, IQR: 0.72–0.87; n = 22; Mann–Whitney rank-sum test: P = 0.233; Fig. 3H).

Fig. 3.

Normal development of basic physiological properties of active and deprived MNTB neurons. (A) Schematic showing the monaural innervation pathway to the MNTB and semantics used with respect to the active and deprived ear. (B) Extracellularly recorded MNTB neuron, consisting of the presynaptic calyx of Held AP (preAP) and the postsynaptic AP (postAP), separated by the synaptic delay (SD). (C–E) Average data for (C) preAP half-width, (D) synaptic delay, and (E) postAP half-width show no significant differences between the control and the deprived group. (F) Patterns of spontaneous firing activity for an active (Top, blue plot) and a deprived (Lower, orange plot) neuron for 20 s. No regularity of interspike intervals (ISIs) was observed, which was corroborated by the absence of (G) multipeaked ISI histograms. (H) Regularity of spontaneous activity was quantified by the coefficient of variation (CV = SDISI/meanISI). CVs around 1 (dashed line) indicate a near-Poisson distribution of the spontaneous rates.

Together, the similarity in synaptic delay, AP half-width and patterns of spontaneous activity supports the notion that cellular and circuit development has not been delayed during the 10-d deprivation period. However, thicker myelin of active auditory nerve and TB fibers could lead to temporal mismatches in downstream binaural processing centers.

Latency differences between auditory brainstem responses (ABR) wave IV and wave I serve as a compelling marker for signal conduction speed through the auditory brainstem and are known to correlate with the extent of myelination of auditory axons in rodents (11, 31) and humans (32, 33). Auditory thresholds were measured in response to brief broadband sounds (clicks) and were compared between the control ear, the closed ear with the earplug still in place, and the deprived ear immediately following earplug removal (Fig. 4A). Thresholds were elevated by about 47 dB due to the earplug, yielding a significant difference in auditory thresholds between the control ear (median: 24 dB SPL, IQR: 22–26 dB SPL) and the closed ear (median: 68 dB SPL, IQR: 61–80 dB SPL; P ≤ 0.001). Immediately after earplug removal, thresholds partially recovered (median: 52 dB SPL, IQR: 46–62 dB SPL; P = 0.127; Kruskal–Wallis one-way ANOVA on ranks; n = 22 ears/11 mice, Fig. 4 A–C). Full recovery of thresholds was observed about 25 d after earplug removal in a previous study (11). In the present study, ABR measurements were used to assess latency differences and relate them to changes in myelination. Since latencies of neural responses are negatively correlated to stimulus intensities, we corrected for the observed differences in auditory thresholds after earplug removal by assessing the latencies 20 dB above the respective thresholds of either ear (Fig. 4D). This correction resulted in an alignment of ABR waves I, but the latency for ABR wave IV was still longer when the deprived ear was stimulated (Fig. 4E). The average latency difference (ΔIV-I) revealed a neuronal processing delay of about 458 µs for the sound-deprived ear (median_control: 2.82 ms, IQR: 2.62–2.97 ms; median_deprived: 3.08 ms, IQR: 3.86–3.62 ms; n = 30 ears; 15 mice; Wilcoxon signed-rank test: P ≤ 0.001; Fig. 4F).

Fig. 4.

Monaural deprivation leads to a mismatch in neuronal processing speed. (A) ABR thresholds in response to click stimulation of the control ear, the deprived ear with the earplug in place (“closed ear”) and the deprived ear immediately after earplug removal. (B) ABR waveform examples of one mouse stimulating either the control or the deprived ear. (C) Both datasets show four characteristic ABR peaks and enhanced thresholds (bold trace) of the deprived ear. (D) Latency-intensity functions depict the typical increase of latencies with decreasing intensities by example of wave I stimulated by either control or deprived ear. The black arrow indicates the correction of wave I latencies with respect to thresholds. At 20 dB above threshold (black circled symbols), wave I latencies are the same. (E) Following latency correction for wave I, the time to wave IV was still longer when stimulating the deprived ear. (F) Average data comparing within animal differences of wave IV-I latencies. (G) The Sinclair et al. (11) model was used to acquire the conduction velocity of the active (blue circles) and deprived (orange circles) fibers. Different velocities according to axon diameter (μm on the vertical color scale) are shown in the background. (H) Histogram of conduction velocities (bin width 0.1 m/s) shows active dataset to have no slow fibers. (I) Model ABRs. Black: Wave I assuming synchronous spiking in the brainstem. Blue/orange: Wave IV derived from the velocity distributions in H, for an ABR filter constant a = 2.5 ms. Inset: IV-I latency difference only slightly increases for filter constants a > 2 ms.

Together, these data suggest that the neuronal processing delay observed in the ABR recordings following 10 d of sound deprivation was caused by an additive change in myelination of fibers along the auditory pathway.

There is a broad agreement that ABR wave I corresponds to activity entering the cochlear nucleus (CN) and wave IV to activity of the inferior colliculus (IC). The path length between those two nuclei in mice was estimated from coronal brain sections to be 9.82 mm. Including these data on adaptive myelination along the TB fibers into our reported computational model (11) provides estimates for the conduction velocity of active and sound-deprived fibers (Fig. 4G). The distribution of conduction velocity was narrower in the control population, and the median conduction velocity of the controls was significantly faster (velocity_active: median = 4.46 m/s, IQR = 3.78–5.30 m/s, n = 314) compared to the deprived ear (velocity_deprived: median = 4.39 m/s, IQR = 3.42–5.28, n = 257; Mann–Whitney rank-sum test: P = 0.047; Fig. 4H). This result was then used to predict how ABR latencies would change based on adaptive myelination alone. The model shows that the wave ΔIV-I latency in controls was around 2.5 ms (gray curve in Fig. 4I), assuming each action potential would make an alpha function-shaped contribution to the ABR with the filter length a of the alpha function taken as 2 ms (approximating synaptic currents in the inferior colliculus; ref. 34). This wave ΔIV-I latency is in accordance with the actual ABR measurements (Fig. 4F), validating the model. Assigning the peak of wave I as time point zero and using a filter length of a = 2 ms (Fig. 4E), the model predicts the peak of wave IV to occur 0.355 ms earlier when stimulating the control ear (blue curve in Fig. 4I) compared to stimulating the sound-deprived ear (orange curve in Fig. 4I). Since the filter length a is not well constrained, increasing it to 2.5 ms gave an upper boundary of 0.45 ms for the wave IV latency-changes (Fig. 4 I, Inset). This delay closely approximates the mean wave IV-I differences between both ears seen in the ABR measurements of 0.45 ± 0.10 ms (n = 14), consistent with the temporal mismatch measured at the level of the inferior colliculus after monaural sensory deprivation being fully attributed to adaptive myelination.

Discussion

The present study reveals that radial growth of myelin is driven by axonal activity rather than being an exclusive product of the neuron’s (or oligodendrocyte’s) genetic identity. We show that fiber bundles carrying axons originating from the same neuronal cell type on either side of the brain form a mosaic of interspersed active and passive axons. This raises the question of how individual myelinating oligodendrocytes can identify active from passive axons in such a mixed population.

One of the main fiber types in the trapezoid body are GBC axons. In early postnatal development (long before earplugs were inserted in the present study), GBC axons are guided across the midline to find their targets in the contralateral MNTB by the expression of the receptor tyrosine kinases Ephs and their ephrin ligands, a large family of contact-mediated guidance molecules (35). Particularly ephrin-A5 is strongly expressed in the MNTB and may cause the concentration of axons into bundles within 600 µm left and right of the midline (36, 37). One might imagine that axons growing in the same direction are facilitated in reaching their targets by being in bundles. However, as our salt and pepper patterns of traced fibers show, each fiber crosses the midline without the support of “like-minded neighbors.” The precise temporal pattern of initial myelination at newly formed axons and the rate at which myelin sheaths are generated will vary with the identity of the neuron and the brain region, with the brainstem typically being myelinated earlier than the cortex (38). On initiation, the wrapping of individual axonal segments might be completed within only a few hours (39), and this process mainly determines internodal patterns. In the TB, myelinated axons were found as early as postnatal day eight (P8), based on the expression of myelin basic protein and electron microscopic visualization of myelin sheaths in axonal cross-sections (11, 40).

The ear canal of rodents is still closed with connective tissue at P8, preventing the passage of airborne sound. Over this developmental period until the onset of hearing at P10, the cochlea generates spontaneous activity, which occurs in bursts and propagates throughout the auditory system (27, 41). Interestingly, MNTB neurons that develop without sound-evoked activity for 10 d also change their spontaneous activity from developmental burst-firing to mature Poisson-distributed firing, suggesting that the change is triggered intrinsically and independently of hearing onset. Therefore, the large increase in the number of action potentials due to sound stimulation rather than the change in firing pattern likely underlies the rapid radial growth of myelin following hearing onset (11).

Earplugs are a widely established method to mimic mild and reversible conductive hearing loss (11, 42–46). It is important to point out that reversibility of this hearing loss relates to the removal of earplugs and the subsequent restoration of peripheral conductive hearing, rather than to central auditory processing, which can show persistent adaptations in temporal and spatial processing (42, 43, 45, 46). We have employed monaural earplugs to prevent the typical increase in neuronal activity that occurs with the onset of hearing (27). Blocking the sound-evoked activity while maintaining spontaneous firing enables titration of firing rates required for normal development of axonal conduction.

Neuronal activity triggers high-frequency calcium transients in individual myelin sheaths which subsequently promotes myelin sheath elongation (19, 47). Although it is unclear whether radial growth of myelin is also promoted by these high-frequency calcium transients, both processes depend on axonal activity and require local translation of myelin basic protein. The activity-dependent local increase in calcium can be caused by either calcium influx through calcium channels or by calcium release from intracellular stores. To our knowledge, voltage-gated calcium channels have not been described in the membrane of myelin sheaths. Therefore, an intracellular calcium release via voltage-induced second messenger activation, which has been described in neurons, could act as an alternative origin of activity-dependent calcium transients (48). One mechanism by which axons could communicate their activity levels directly to their myelin sheaths would be for action potentials to provide the depolarization required to increase intracellular calcium and initiate MBP synthesis. The depolarization can be achieved by the efflux of potassium ions from highly active axons into the inner tongue of myelin via composite channels consisting of a low-voltage activated Kv1 channel on the axonal surface and a connexin 29 hemichannel on the inner tongue of myelin (49). The subsequent depolarization of the myelin sheath could then trigger calcium-transients and the translation of MBP at the inner tongue of individual myelin sheaths. Remote translation of MBP, directly at the active axon–myelin sheath interface (50, 51) and away from the oligodendrocyte soma, is well recognized and provides the basis for an axon-by-axon adaptive myelination process. It is likely that such a potassium flux mechanism enables activity-dependent fine-tuning of axonal conduction velocity, but is not required for general myelination. Connexin-29 null mutants show no major myelination deficits and inconspicuous ABRs (52). As suggested by our current results, activity-dependent myelin plasticity taking place after the initial myelin formation may lead only to small temporal mismatches of about half a millisecond. However, since the time window during which spiking activity in the lateral superior olive is suppressed by MNTB inhibition only lasts a few hundreds of microseconds (53, 54), a ~500-µs delay of the inhibitory input will most certainly cause a deficit in the sound localization ability. Indeed, deficits in sound localization performance and the effects on neural plasticity following temporary monaural occlusion by either earplugs or Otis media have been investigated for many decades (55, 56). These studies agree that monaural occlusion causes an initial deficit in sound localization, which over time is followed by adaptive neural plasticity. Our present results add to this body of work by showing that the change in myelination is an additional factor which contributes to these adaptive processes.

In the present study, we observed increased myelination and faster overall processing speed through the auditory brainstem for the controls compared to sound-deprived ears. However, the earplug-induced delay at each processing stage was small, suggesting compensatory mechanisms. In rats that received earplugs at 30 d of age, a lasting change in AMPA receptor composition toward GluA3 subunits in cochlear nucleus neurons was described, a subunit known to be responsible for ultrafast excitatory synaptic transmission also at the calyx of Held (42, 57, 58). Even though we corrected for earplug-induced differences in auditory thresholds, additional changes or compensations could occur at different levels of the auditory pathway. However, the observation that the ABR wave IV-I difference between plugged and control sides was still nearly 500 µs suggest a large influence of myelination. The computational model only considered changes in conduction velocity and could still predict at least 75% of the temporal mismatch based on stronger myelination in the control side. Such a temporal mismatch during early days after hearing onset is likely to disturb the experience-dependent refinement of the sound localization circuit (59).

Materials and Methods

Experiments were approved in accordance with the stipulations of the German animal welfare law (Tierschutzgesetz) (ROB-55.2-2532.Vet_02-18-118). Male and female CBA/Ca mice born at the LMU Faculty of Biology vivarium to breeders acquired from Charles River laboratories were housed with 12-h light/dark cycles and food and water ad libitum.

Depending on normality of the distribution, population average data are given by the mean ± SD, or the median and 25 and 75% interquartile ranges (IQR) as box edges with full data ranges depicted by individual data points. Accordingly, parametric or nonparametric tests were used to determine statistical significance.

Ear Plugging.

Small pieces of an “E.A.R Classic II” human foam earplug were compressed and inserted into the external auditory meatus before being sealed with dental cement (Paladur; Heraeus-Kulzer) at P10 under MMF anesthesia (Medetomidin: 0.5 mg/kg BW, Midazolam: 5.0 mg/kg BW, Fentanyl: 0.05 mg/kg BW). Earplugs were examined daily and replaced every third day to avoid uncontrolled loss as the mice grew.

ABRs were recorded in MMF-anesthetized mice, placed on a temperature-controlled heating pad (ATC1000, WPI) in a soundproof chamber (Industrial Acoustics). Subdermal needle electrodes (Rochester Electro-Medical, Inc.) were placed at the vertex of the mouse’s head (reference), ventral to the pinna (active), and near the base of the tail (ground). Electrodes were attached to a low-impedance headstage (RA4LI, Tucker Davis Technologies: TDT), and a preamplifier (RA16PA, TDT; amplification factor: 250) and connected to the auditory processor (RZ6, TDT) via optical cables. SPIKE software (Brandon Warren, UW) was used for speaker calibration (MF1, TDT), stimulus (100 µs clicks; 10 to 90 dB SPL) generation and waveforms recording. Thresholds were determined as the sound intensity at which at least two peaks could be distinguished in the average (1,000×) ABR waveform. When switching between control ear and ear-plugged ear, only the active electrode was swapped to the other ear. All other electrodes were kept in place during earplug removal. At the end of ABR experiments, animals were either kept in anesthesia for subsequent in vivo single-unit recordings or killed with an overdose of MMF for tracing and histology experiments.

In Vivo Physiology.

MMF-anesthetized mice were placed on a heating pad in a soundproof chamber and stabilized in a custom stereotaxic device. A craniotomy was performed just anterior to the lambda suture and glass microelectrodes (3M KCl; 5 to 20 MΩ) were lowered into the brainstem. A ground electrode was placed in the muscle at the base of the neck. Signals were amplified (AM Systems, Neuroprobe 1600), filtered (300 to 3,000 Hz; TDT PC1), and sampled (50 kHz) with a Fireface UFX audio interface (RME). AudioSpike software (HoerrTech) was used to calibrate the multifield magnetic speakers, generate stimuli, and record action potentials. Search stimuli to identify auditory nuclei consisted of pure tones (50 to 100 ms duration, 5 ms rise/fall time) at varying intensity (0 to 90 dB SPL) and were presented through short hollow ear bars directly connected to the speakers.

Axon Tracing.

P20 mice were killed with an overdose of MMF and intracardially perfused with Ringer’s solution containing heparin. Brains were removed from the skull into ice-cold ACSF. Crystals of tetramethylrhodamine (TMR, Invitrogen D3308) were grown from a 50 to 100% TMR in H₂O solution on fine tungsten needle tips and stored until used (24 h maximum). Small incisions were made in the meninges corresponding to the ventral cochlear nucleus on each side and TMR crystals or correspondingly small biocytin (TOCRIS, 3349/ Sigma B4261) crystals were inserted. The brains were then incubated for 2 h in carbogenated ACSF at room temperature before being fixed in 4% PFA overnight.

Immunohistochemistry and Confocal Microscopy.

Sagittal brainstem sections (50 µm) including the TB were taken using a vibratome (Leica, VT1200S). Immunostaining was carried out as previously described (11) using primary MBP antibodies (ab7349 abcam; 1:150) and corresponding secondary antibodies (Alexa-647 donkey anti-rat, Dianova 712-605-150, 1:140; biocytin-targeted streptavidin Alexa-488, Dianova, 016-540-084, 1:250). Optical sections were acquired with a confocal laser-scanning microscope (TCS SP5-2, Leica Microsystems, Mannheim, Germany) equipped with HCX PL APO CS 20×/NA0.7 and HCX PL APO Lambda Blue 63×/NA1.4 immersion oil objectives. For each optical section, images were collected sequentially for the different fluorophores. 8-bit grayscale images were obtained with ABC pixel sizes of 120 to 1,520 nm depending on the selected zoom factor and objective. To improve the signal-to-noise ratio, images were averaged from three successive scans.

Electron Microscopy.

P20 mice were killed with an overdose of MMF and intracardially perfused with Ringer’s solution, followed by 2.5% glutaraldehyde plus 2% PFA in 0.1% cacodylate buffer (CB) pH 7.0. Brains were removed from the skull and postfixed in the same fixative overnight at 4 °C. After washing (3×) in CB, control and deprived auditory nerves were dissected and sectioned perpendicular to the fiber direction. The tissue blocks were washed (4×) in CB and postfixed in 1% OsO₄ in CB for 1 to 2 h. After washing and dehydrating in graded series of acetone, the tissue was embedded in Spurr’s resin (60). Before ultrathin (70 nm) sectioning, several semithin (1 µm) sections were cut for light microscopic investigation. The ultra-thin sections were collected on formvar-coated copper slot grids and stained with uranyl acetate and lead citrate. EM images were taken using a FEI Morgagni transmission electron microscope (80 kV, SIS Mega view III camera, 1,375 × 1,032 pixels).

Computational Modeling.

We used the multicompartmental model of an axon developed in previous publications (9, 11) to assess the conduction velocity of control vs deprived fibers for the dataset collected in the present study. Conduction velocities were obtained for measured fibers (Fig. 1 F and G) using the geometrically determined pair of parameters, axon radius and myelin thickness, and leaving all other parameters unchanged. The ABR was modeled by replacing each spike time t_n (at the CN for wave I and at the IC for wave IV) by an alpha function with parameter a:

$$\text{ABR}\left(\mathrm{t}\right)={\mathrm{\Sigma }}_{\mathrm{n}}\left(\mathrm{t}-{\mathrm{t}}_{\mathrm{n}}\right)/\left({\mathrm{a}}^2\right)\text{exp}\left[-\left(\mathrm{t}-{\mathrm{t}}_{\mathrm{n}}\right)/\mathrm{a}\right]\ast \left(\mathrm{t}<{\mathrm{t}}_{\mathrm{n}}\right).$$

Spike times at the CN are assumed to be fully coincident; spike times at the IC are derived from the distribution of simulated conduction speeds (Fig. 4H).

Data, Materials, and Software Availability

All study data are included in the main text.