Refinement of axonal conduction and myelination in the mouse optic nerve indicate an extended period of postnatal developmental plasticity

Abstract

Retinal ganglion cells generate a pattern of action potentials to communicate visual information from the retina to cortical areas. Myelin, an insulating sheath, wraps axonal segments to facilitate signal propagation and when deficient, can impair visual function. Optic nerve development and initial myelination has largely been considered completed by the fifth postnatal week. However, the relationship between the extent of myelination and axonal signaling in the maturing optic nerve is not well characterized. Here, we examine the relationship between axon conduction and elements of myelination using extracellular nerve recordings, immunohistochemistry, western blot analysis, scanning electron microscopy, and simulations of nerve responses. Comparing compound action potentials from mice aged 4–12 weeks revealed five functional distinct axonal populations, an increase in the number of functional axons, and shifts toward fast-conducting axon populations at 5 and 8 weeks postnatal. At these ages, our analysis revealed increased myelin thickness, lower g-ratios and changes in the 14 kDa MBP isoform, while the density of axons and nodes of Ranvier remained constant. At 5 postnatal weeks, axon diameter increased, while at 8 weeks, increased expression of a mature sodium ion channel subtype, Na_v 1.6, was observed at nodes of Ranvier. A simulation model of nerve conduction suggests that ion channel subtype, axon diameter, and myelin thickness are more likely to be key regulators of nerve function than g-ratio. Such refinement of axonal function and myelin rearrangement identified an extended period of maturation in the normal optic nerve that may facilitate the development of visual signaling patterns.

1 |. INTRODUCTION

Effective neuronal signaling is dependent on axonal properties, including axon diameter (Hursh, 1939; Rushton, 1951), and ion channel distribution (Pumphrey & Young, 1938; Tasaki, 1939) which influence action potential propagation. In the central nervous system, oligodendrocytes can dramatically increase conduction speed through the generation of myelin that enwraps axons, resulting in a concentration of ion channels at nodes of Ranvier. By providing distinct levels of insulation, the arrangement of myelination including the number of myelin wraps, the length of the internode, and myelin thickness can alter the speed of signal propagation along the axon allowing for the coordination of signals from multiple neuronal sources at their targets (Ford et al., 2015; Lang & Rosenbluth, 2003; Seidl et al., 2010; Waxman, 1975). While unmyelinated axons can have slow, but effective signaling, demyelination of previously myelinated axons can induce severe deficits in higher order functionality, including learning, socialization, and sensory processes (Barrera et al., 2013; Bengtsson et al., 2005; Eluvathingal et al., 2006; Etxeberria et al., 2016; Hill et al., 2014; Keller & Just, 2009; Liu et al., 2012; Makinodan et al., 2012; McKenzie et al., 2014; Sánchez et al., 1998; Scholz et al., 2009; Zhao et al., 2012), as a result of delayed or blocked conduction (Hardmeier et al., 2017).

The optic nerve, a white matter tract extending from the eye to the optic chiasm, represents an ideal central nervous system (CNS) region to investigate the role of myelination in axonal function. In rodents, the vast majority of optic nerve axons are derived from retinal ganglion cells (RGCs) that integrate visual information from the retina as patterns of action potentials. These signals travel along the axon through both unmyelinated (retina to optic nerve head) and myelinated (optic nerve head to optic chiasm) segments before synapsing at nuclei in the thalamus and superior colliculus (Ellis et al., 2016; Huberman et al., 2008; Martin, 1986). RGCs form connections to their cortical targets during late embryogenesis and after a pruning process, achieve final numbers around the second postnatal week (Gordon & Stryker, 1996; Miller, 1996; Osterhout et al., 2014; Sefton et al., 1985). During this period, oligodendrocyte precursors differentiate and begin to myelinate RGC axons (Antonini et al., 1999), continuing until a majority (80%–95%) of axons in the nerve are myelinated. Myelination in the optic nerve has been considered largely completed by the fifth postnatal week (Ellis et al., 2016; Huberman et al., 2008); however, the current study indicates a greatly protracted and more complex maturation period of axonal function. The mature optic nerve has axons of varied diameter and levels of myelination that generate a range of potential conduction velocities and signaling patterns to code visual information. Notably, the development of RGCs and myelination is concurrent with periods of experience-dependent developmental plasticity in the visual system (Drager, 1978; Hubel & Wiesel, 1962; Wiesel & Hubel, 1963) known as the critical period (Dangata et al., 1995, 1996; Hubel & Wiesel, 1970). Because the level of myelination can drastically alter action potential propagation, refinement of myelination may shape early visual signaling patterns and maintain activity after postnatal development. The impact of the specific elements of myelination, such as myelin thickness or internode length, on functional properties of the nerve is not fully understood and consequently, key regulatory factors in mouse optic nerve conduction during postnatal development are not completely characterized.

To investigate the functional role of myelination in the optic nerve, the correlations between axonal conduction and aspects of myelination were studied between postnatal weeks 4 and 12. Functional, molecular, and morphological characteristics of C57/Bl6 optic nerves were assessed, including compound action potentials, expression of myelin protein, myelin thickness, axonal diameter, and node of Ranvier density and maturation. These data were then used to develop a model of optic nerve conduction in which the individual contributions of axon diameter, g-ratio, myelin thickness, and sodium ion channel subtype expression were evaluated. We show that optic nerve axons can be classified into five functional groups and at 5 and 8 weeks, additional responsive axons are recruited and functionally defined axon populations shift toward faster conduction. These functional changes were coincident with increased myelin protein expression and myelin thickness, but not with altered axon or node density. At postnatal week 5, shifts toward large axon diameters correlate with changes in functional properties, while at postnatal week 8, maturation of nodes of Ranvier may drive physiological changes. The simulation model of nerve conduction suggests that sodium ion channel subtype, axon diameter, and myelin thickness can have a greater impact on axon population conduction than g-ratio. This study demonstrates that the optic nerve undergoes an extended period of maturation after initial nerve growth and myelination.

2 |. METHODS

2.1 |. Animals

All experiments were performed in compliance with the approved animal policies of the Institutional Animal Care and Use Committee (IACUC) at the George Washington University School of Medicine and Health Sciences. Male and female C57Bl6 mice aged 4, 5, 6, 8, and 12 weeks old were used in these studies, and no gender-specific differences were observed.

2.2 |. Electrophysiology

Extracellular recordings of the optic nerve were collected from mice aged 4, 5, 6, 8, and 12 weeks using an adaptation of the suction electrode recording method described in Evans et al. (2010). Mice were sacrificed by cervical dislocation, and optic nerves were immediately dissected and placed into an oxygenated (95% O₂/ 5% CO₂) HEPES solution (C₈H₁₈N₂O₄S) (in mM: 140 NaCl, 5 KCL, 2 CaCl₂, 5 Glucose, 10 HEPES) to recover for at least 30 min. Before recording, nerves were moved to a perfusion chamber filled with artificial cerebral spinal fluid (in mM: 126 NaCl, 3 KCl, 2 CaCl₂, 2 MgCl₂ 1.2 NaH₂PO₄, 26 NaHCO₃, 10 Glucose) and maintained at 37°C (Single Channel Temperature Controller, Warner Instruments, TC-324). Nerves were inserted into two suction electrodes (Suction Electrode, A-M Systems, #573000) with the rostral and caudal ends in stimulating and recording electrodes, respectively (Figure 1a). Electrical stimulation of the optic nerve induced simultaneous action potentials resulting in a compound action potential (CAP) response. At the maximum stimulus intensity, the stimulus pulse (Flexible Stimulus Isolator, Iso-Flex, MicroProbes for Life Science) recruits all functional axons providing a measure of total optic nerve function. Variations in axon conduction speeds result in discrete peaks in the waveform that identify functionally-distinct axon populations (Figure 1a’). Tetrodotoxin (a sodium ion channel blocker, 1:1000) was used to isolate stimulus-dependent artifacts which were subtracted from the action potential responses. For each trace, optic nerves were stimulated once every 5 s with a stimulus pulse lasting 50 μs. Responses were amplified by 1000× (BMA-931 Bioamplifier, CWE), filtered at 10 kHz, and acquired at 20 kHz (AxoPatch 200B, Axon Instruments) using Clampex software. A minimum of 10 traces were recorded for each nerve per stimulus intensity, and nerves were recorded within 2 h of isolation.

FIGURE 1.

Recruitment of additional functional axons and fast-conducting axon populations occur at 5 and 8 weeks. Extracellular recordings of wildtype optic nerves (n > 8), aged 4, 5, 6, 8, and 12 weeks, used electrical stimulation (a) to induce compound action potentials (CAPs). CAP waveforms (a’, top left) include contributions from different axon populations, represented as peaks, which were identified by a multi-peak Gaussian fit model (a’, bottom left). Individual axon populations (a’, top right, two of the three peaks are shown in red and blue solid lines) were then defined by latency and peak area (a’, bottom left inset) values in scatter plots (a’, bottom right). Averaged CAP area (b), representing total number of functional axons, increased significantly at 5 and 8 weeks (p = 0.0137). Average response onset (c), defined as the time elapsed after the stimulus, was not significantly different across ages (p = 0.1269) and average response duration (d) was longer at 5 weeks (p = 0.0114). To compare axon conduction across age, peak data from all nerves were pooled (e) and represented in a scatter plot where each peak is defined by peak latency (axon population conduction speed), and peak area (relative number of axons contributing to each peak). K-means cluster-defined groups are outlined (dotted lines) and named by relative speed (“Slow,” “Slow-Intermediate,” “Fast-Intermediate,” “Fast”), where the two fast populations are differentiated (“Fast [Small]” or “Fast [Large]”) by their peak area. The trellis plot (right) shows the distribution of axon populations by cluster identity (columns) and age (rows). At 5 weeks, the “Slow-Intermediate” axon population type is detectable, indicating a shift toward faster conduction. The appearance of the “Fast (Large)” cluster at 8 weeks demonstrates a recruitment of additional fast-conducting axon populations, which is coincident with the loss of the “Slow” cluster, representing the slowest-conducting axons. Abbreviation: aCSF, artificial cerebrospinal fluid

2.3 |. Peak analysis

Traces for each nerve at maximum stimulus intensity were averaged prior to subtraction. The area under the resultant CAP response, response duration, response onset, and multi-peak Gaussian fits of the CAP waveform were calculated using Clampfit software. CAP area, response duration, and onset were compared across age groups using a Kruskal–Wallis analysis of variance (ANOVA) using OriginPro 6.1 software. Peaks for all nerves were described by latency and area (Figure 1a’). To observe differences in peak distribution across age, peak data were pooled for all nerves, and a K-means cluster analysis (OriginPro) was used. Statistically significant (p < .05) clusters were identified, and cluster identity was later cross-referenced with age group.

2.4 |. Immunohistochemistry

Mice were anesthetized using Avertin prior to perfusion with 4% paraformaldehyde (PFA) and optic nerves were dissected, postfixed in 4% PFA overnight, and placed in 30% sucrose solution for 24 h. Nerves were embedded in OCT (Optimal Cutting Temperature) compound (Fischer, T), sectioned at 12 μm, and collected on glass slides. Slides were washed with phosphate buffered saline (PBS), incubated in a standard block solution (0.03% Triton X, 10% NGS in 1X PBS), and incubated overnight at 4°C with primary antibodies. After additional washes, slides were incubated with secondary antibodies at room temperature for 1 h, rinsed with PBS, and labeled with DAPI. Primary and secondary antibodies were diluted in blocking solution at the following concentrations: mouse monoclonal antibody against contactin-associated protein 1 (Caspr) (1:700, Millipore, cat# MABN69, RRID:AB_10806491), rabbit polyclonal antibody against sodium channel voltage-gated type VIII alpha subunit (Na_v 1.6) (1:500, Alomone Labs, cat# ASC-009, RRID:AB_2040202), rabbit polyclonal antibody against brain type II Na⁺ channel (Na_v 1.2) (1:500, Alomone Labs, cat# ASC-002, RRID: AB_2040005), goat anti-Mouse IgG Alexa Fluor 594 (1:500, ThermoFisher Scientific, RRID:AB_2534091), and goat anti-Rabbit IgG Alexa Fluor 488 (1:500, ThermoFisher Scientific, RRID:AB_2536525). Images were acquired on a confocal microscope (Zeiss Cell Observer Spinning Disk Confocal) and quantified using Photoshop for a region of interest (50,000 μm²).

2.5 |. Western blot

Mice were sacrificed by cervical dislocation and decapitation after isoflurane anesthesia. Optic nerves were flashfrozen, digested by proteases prior to sonication, and total protein concentration determined using a BCA analysis. An equal concentration of each lysate (2.5 μg/ml) was loaded into a 4%–20% gel (Mini-Protean Tetra Vertical Electrophoresis Cell, BioRad) and separated at 140 V. Proteins were transferred to a nitrocellulose membrane (Mini Trans-Blot Cell, BioRad), incubated for 1 h in a blocking solution (5% BSA in TBS-Triton) followed by primary antibody overnight and secondary HRP-conjugated antibodies for 1 h at room temperature before visualization (Thermo Fisher ECL Western Blotting Substrate). Relative protein concentration was normalized to β-actin (mouse monoclonal antibody against beta-actin, Santa Cruz Biotechnology, cat# 47778, RRID:AB_2223360) as a loading control, and the density of bands was quantified using ImageJ for each of the following antibodies: myelin basic protein (MBP) (rat monoclonal antibody against myelin basic protein, Abcam, cat# ab7349, RRID: AB_305869) and proteolipid protein (PLP) (rabbit polyclonal antibody against proteolipid protein, Abcam, cat# ab28486, RRID: AB_776593).

2.6 |. Scanning electron microscopy

Mice were anesthetized with Avertin prior to perfusion with 2% glutaraldehyde and 4% PFA in 0.1 M sodium cacodylate. Optic nerves were dissected and postfixed for 48 h. Following osmication, dehydration, and embedding in Epon 814, nerves were sectioned either transversely (at the midpoint) or longitudinally, at 120 μm using an ultramicrotome (UC7 Leica Microsystems). These ultrathin sections were placed in silicon wafers and carbon taped in aluminum stubs for scanning electron microscopy (SEM) imaging in a Helios NanoLab 660 dual beam microscope. To maximize the collection of the backscattered electrons and produce TEM-like images, a high contrast solid-state backscatter electron detector, in magnetic immersion mode and 4 μm working distance, was used. The acquisition was performed using 2 kV and 0.40 nA as the current landing. Low-magnification overviews of each analyzed optic nerve were scanned to identify the same areas in each optic nerve to use in myelin/axon quantification; then, high-resolution images of desired areas were performed using 20,000× magnification (acquisition settings: dwell time 5 ms, 3072 × 2048 resolution). Axon diameter and g-ratio were analyzed using ImageJ software and plugin (‘MRI-gratio tools’) to automate axon detection and measurement of axon diameter and cross-sectional area in transverse sections. G-ratios were calculated as the ratio of inner axon diameter to outer myelinated axon diameter, where the diameter was extrapolated from a circle of equivalent measured area. For g-ratio analysis, a minimum of 250 axons were sampled per nerve, and data were combined from central and peripheral regions. Axon density was calculated as the number of axons per image.

2.7 |. Optic nerve conduction model

An adaptation of the model described in Arancibia-Cárcamo et al. (2017) was generated in MATLAB to produce simulated compound action potentials (sCAP) using experimentally derived axonal and myelin properties for mouse optic nerves at each age group (4, 5, 6, 8, and 12 weeks of age). The Arancibia-Cárcamo et al. (2017) model (available via Github for public use) simulates the transmembrane potential across nodes of Ranvier along a theoretical rat optic nerve axon using user-defined parameters (axon diameter at internode segments, node length, g-ratio, etc.). Transmembrane potential can be stored at any node and plotted to visualize intracellular action potentials. In the current implementation, optic nerves of 4 mm were simulated using spatial and temporal discretization of 140 μm and 1 μs, respectively. Upon completion of each simulation, intracellular action potentials at the end of each axon were converted to extracellular potentials through spatial and temporal integration of changes in transmembrane potential (Plonsey & Barr, 2007). Some properties (g-ratio, axon diameter, number of nodes, ion channels present at nodes) were experimentally derived, while other values were taken from literature describing the mouse optic nerve anatomy (see Table 1). Simulations of 50, 100, 200, 500, and 1000 axons per optic nerve were collected where the only variable inputs were g-ratio, axon diameter, and sodium ion channel subtype (Na_v 1.6 or Na_v 1.2), which were randomly assigned to each axon within the nerve. For each age group, axon diameter and g-ratio values were assigned by randomly sampling their values from the distribution of experimental measurements. To ensure simulated parameter distributions represented experimental datasets, t-tests were used to determine similarity between simulated and experimental datasets of g-ratio and axon diameter (Figure S1A,B). Sodium ion channel subtype was assigned for each axon using a probability distribution based on the ratio of Na_v 1.6+ to Caspr+ nodes. Na_v 1.6 and Na_v 1.2 ion channel properties were adaptations of the predefined “Fast Na+” and “Persistent Na+” channels in the Arancibia-Carcamo model (Arancibia-Cárcamo et al., 2017; Richardson et al., 2000).

TABLE 1.

Experimental and literature-derived parameters for simulated model of nerve conduction

Parameter	Value	Units
Resting potentiala	−82	mV
Leakage potentiala	−83.38	mV
Na+ reversal potentiala	50	mV
K+ reversal potentiala	−84	mV
Node diameter	0.77	μm
Node lengthb	1.67	μm
Internode lengthc	138	μm
Axon membrane capacitance at node and internoded	0.9	μF/cm2
Myelin membrane capacitanced	0.9	μF/cm2
4-week optic nerve
Mean g-ratio	0.6342	N/A
g-Ratio variance	0.0118	N/A
Mean axon diameter	0.5770	μm
Axon diameter variance	0.0777	μm
Percentage of Fast Nav nodes	17.10	%
Mean myelin thickness (derived)	0.9254	μm
Myelin thickness variance	0.1678	μm
5-week optic nerve
Mean g-ratio	0.6212	N/A
g-Ratio variance	0.0193	N/A
Mean axon diameter	0.7754	μm
Axon diameter variance	0.1003	μm
Percentage of Fast Nav nodes	18.80	%
Mean myelin thickness (derived)	1.246	μm
Myelin thickness variance	0.1924	μm
6-week optic nerve
Mean g-ratio	0.6228	N/A
g-Ratio variance	0.0195	N/A
Mean axon diameter	0.7827	μm
Axon diameter variance	0.1100	μm
Percentage of Fast Nav nodes	26.90	%
Mean myelin thickness (derived)	1.251	μm
Myelin thickness variance	0.2038	μm
8-week optic nerve
Mean g-ratio	0.6044	N/A
g-Ratio variance	0.0145	N/A
Mean axon diameter	0.7904	μm
Axon diameter variance	0.0888	μm
Percentage of Fast Nav nodes	34.70	%
Mean myelin thickness (derived)	1.325	μm
Myelin thickness variance	0.1969	μm
12-week optic nerve
Mean g-ratio	0.6163	N/A
g-ratio variance	0.0185	N/A
Mean axon diameter	0.7981	μm
Axon diameter variance	0.1070	μm
Percentage of Fast Nav nodes	32.50	%
Mean myelin thickness (derived)	1.313	μm
Myelin thickness variance	0.2314	μm

^aValues are from Arancibia-Cárcamo et al. (2017).

^bValues are from Gentet et al. (2000).

^cValues are from Rios et al. (2003).

^dValues are from Butt et al. (1994).

A sCAP was computed for an optic nerve for each age group by summating the extracellular potentials (at the end of each axon) of all axons within age group. Conduction velocity for each axon was calculated as the length of the axon divided by the time elapsed until the measured peak of the action potential. An unsupervised cluster analysis (MATLAB, “clusterdata”) of axon conduction velocities segregated the axons of each nerve into functionally-distinct axon populations. The conduction velocity and corresponding g-ratios, axon diameters, derived myelin thickness, and sodium ion channel subtypes were recorded for each axon simulation. For each axon population within an optic nerve, the linear or nonlinear relationships between conduction velocity and axon diameter, g-ratio, or ion channel subtype were analyzed using MATLAB functions (“polyfit,” “nlfit”) (Figure S2). Because the simulation approach was based on a random sampling of axon diameter and g-ratios for computational efficiency rather than a combinatorial approach, the minimum number of simulated axons required to detect relationships between conduction velocity and axonal properties was determined. Cluster centroids and sCAPs were similar for optic nerve simulations containing 500 and 1000 axons (Figure S1C), so nerve responses generated by 500 axons were used in analysis.

2.8 |. Statistics

For all analyses, individual datasets were first tested for normality to determine if a parametric (ANOVA) or nonparametric (Kruskal–Wallis ANOVA) test would be used to identify significant differences determined by a p-value less than .05. For regression analyses, the meaningful correlation threshold for regression analysis was r² > 0.85, and only fits that meet this threshold are listed in Figure S2C–E.

3 |. RESULTS

3.1 |. Optic nerve function favors faster conduction during postnatal maturation

To explore developmental changes in optic nerve conduction over a 4–12 week postnatal period, we used extracellular nerve recordings in which electrical stimulation of the isolated nerve generates a CAP waveform (Evans et al., 2010; Stys et al., 1991). Measures of the area under the curve (CAP area), representing total nerve function (Figure 1a), varies as a function of postnatal age (Figure 1b) and indicates an unexpectedly long period of nerve maturation. However, the rate of CAP increase was not uniform: significant increases in the average CAP area were observed at 5 and 8 weeks, suggesting a stepwise maturation of functional axons in the optic nerve. Comparison of response onset across age groups found no significant differences (Figure 1c), indicating that at least a subset of optic nerve axons acquired the fastest conduction rate by 4 weeks and remained stable over the following 2 months. By contrast, longer response durations were detected in 5-week-old nerves (Figure 1d), suggesting that there is significant variability in the rate of axon conduction during postnatal development.

Detailed analysis of the composition of CAP responses revealed reproducible functional differences between populations of axons (represented as individual peaks) at each age examined. These functionally distinct axon populations were grouped into five distinct types based on conduction speed using a K-means cluster analysis on pooled datasets and were termed “Fast (Small/Large),” “Fast-Intermediate,” “Slow-Intermediate,” and “Slow.” Axon populations with the fastest conduction and large peak areas “Fast (Large)” were separated from the “Fast (Small)” type, which had the same range of conduction but fewer contributing axons (Figure 1e,f). Nerves of all ages contained “Fast” and “Fast-Intermediate” axon populations (Figure 1e,f), consistent with the observation that response onset data were similar across all ages evaluated (Figure 1c). The age-related increase in CAP area (Figure 1b) was correlated with the identification of additional functional populations of axons and was most prominent at 5 and 8 weeks. For example, the “Slow-Intermediate” axonal population type was first observed at 5 weeks and was maintained at all older ages. The “Fast (Large)” axonal population was first detectable at 8 weeks, a time at which the “Slow” axon populations were no longer detectable. This loss of “Slow” axon populations persisted until 12 weeks of age (Figure 1f). The observed changes in functionally distinct axon populations between 4 and 12 weeks shift to favor faster conduction with age and indicate an extended period of axonal refinement.

3.2 |. Expression of myelin proteins are dynamic during postnatal optic nerve maturation

To determine whether changes in the levels of expression of myelin proteins were temporally linked to the dynamic functional properties of optic nerve axons during postnatal maturation, the levels of myelin and axonal proteins isolated from optic nerves at 4, 5, 6, 8, and 12 weeks were assessed by western blot analyses. Expression of β-actin was used for comparison of relative protein concentrations of MBP and PLP across different ages. Four isoforms of MBP were detected at all ages and while a general trend toward increased expression was age-dependent, expression levels were dynamic (Figure 2a,b). This was particularly evident with the 14 kDa MBP isoform which increased at 5 weeks, then decreased at 6 weeks, but remained constant at older ages. PLP expression was stable across ages (Figure 2c), suggesting that changes in the relative levels of the majority of myelin proteins were not correlated with observed functional changes.

FIGURE 2.

Expression of 14 kDa MBP isoform is dynamic between 4 and 6 postnatal weeks. Isolated wildtype optic nerves (n = 4) from ages 4, 5, 6, 8, and 12 weeks were analyzed by western blot, using β-actin as a control (a). Relative concentrations of most myelin basic protein (MBP) isoforms (b) were relatively stable across the ages assayed (p > 0.55), showing a general trend toward increased MBP expression by 12 weeks, while more pronounced dynamic changes in the 14 kDa MBP isoform expression (p = 0.032) were observed between 4 and 6 weeks. No significant changes were observed in proteolipid protein (PLP) expression (c, p = 0.771), although there was a similar overall trend in temporal expression. Error bars = SEM

3.3 |. Increased myelin thickness and lower g-ratios correlate with functional axonal changes during postnatal maturation

The observed changes in levels of myelin protein expression may reflect changes in the number of myelinated axons, overall growth of the nerve, or alterations in myelin thickness around individual axons. Thicker myelin confers faster signal conduction, and the thickness of myelin is directly related to axonal diameter through the “g-ratio.” A comparison of g-ratios from a temporal sequence of ultrastructural SEM images (Figure 3a) indicated that lower g-ratios were associated with increasing age (Figure 3b–e). Significant decreases occurred at 5 and 8 weeks, but minimal differences were observed between 8 and 12 weeks (Figure 3e). Assessment of myelin thickness also demonstrated significant increases at 5 and 8 postnatal weeks (Figure 3f). When plotted by axon diameter (Figure 3b–d), these shifts toward lower g-ratios were more pronounced in smaller diameter axons at 5 weeks and affect all axon diameters at 8 weeks. These data are consistent with the demonstrated shift toward faster conduction at 5 and 8 weeks of age and implies functional roles for g-ratio and myelin thickness in the emergence of distinct axonal populations in the optic nerve.

FIGURE 3.

Myelin thickness increases and g-ratios decrease at physiologically relevant time points, 5 and 8 postnatal weeks. Representative scanning electron microscopy (SEM) images (a) for each age (n = 3) demonstrate a variety of myelinated axon diameters. Histograms (b–d, left) of g-ratios (ratio of inner axonal diameter to outer myelinated diameter) compare physiologically relevant time points (between 4 and 5 weeks [b], 6 and 8 weeks [c] and 8 and 12 weeks [d]) and show overall decreases in g-ratio at 5 and 8 weeks and minimal changes between 8 and 12 week g-ratio distributions. Linear fits describing the relationship between g-ratios and axon diameters for these ages (b–d, right) demonstrate drastic decrease in g-ratios for smaller axons at 5 weeks, modest decreases in g-ratio at 8 weeks across all axon diameters, and modest decreases in g-ratio at 12 weeks for small diameters. When compared across age group, average g-ratio (e) was significantly lower between 4 and 5 weeks (p = 0.0025) and 6 and 8 weeks (p = 0.00021), but was not significantly different between 8 and 12 weeks (p = 0.1167), indicating significant shifts toward thicker myelin at 5 and 8 weeks. Similarly, myelin thickness (f) increased significantly at 5 (p = 0.0005) and 8 postnatal weeks (p = 0.0027). Myelinated axon density (g) was not significantly different across age group (p = 0.7523); however, significant increases in average axon diameter (h) between 4 and 5 weeks (p = 0.00064), observed in cumulative axon diameter distributions (i), indicate that axon diameter may be relevant in functional changes at younger ages. Error bars = SEM, scale bar in (a) = 4 μm

Conduction-relevant axonal properties, including axon density and axon diameter, were evaluated across age groups. Axon density remained constant, with a nonsignificant decrease at 12 weeks (Figure 3g) making additional axon recruitment an unlikely contributor to the age-dependent functional changes. The average axon diameter (Figure 3h) and the cumulative distribution of axon diameter (Figure 3i) show significant increases at 5 weeks, which is likely a consequence of overall growth of the nerve. Because large diameter axons would have faster signal propagation, increases in axonal diameter may contribute to faster axon conduction. However, the stability of axon diameters at older ages (6–12 weeks of age) suggests it is less likely to be a significant factor in the functional changes observed at 8 weeks, where additional functional axons are observed.

3.4 |. Age-related changes are observed in the composition and density of nodes of Ranvier

While the level of myelination can modulate optic nerve axonal conduction, the structure, density, and ion channel composition of nodes of Ranvier may act as a potential regulator. To assess the morphological ultrastructure of nodes, longitudinal sections of optic nerves at 4, 5, and 8 weeks were screened by SEM. At all ages, nodes appeared structurally mature with intact paranodal loops and compact myelin around internode segments (Figure 4a), indicating that fully formed nodes of Ranvier were present in nerves as young as 4 weeks of age. No changes in the structure or morphological organization of nodes of Ranvier were detected during subsequent postnatal maturation. To determine if nodal density was correlated with nerve function, nodes were identified by doublet expression of Caspr, a paranodal marker, in longitudinal nerves (Figure 4b,c). Average nodal density was not significantly different across all ages, although a nonsignificant increase was observed at 5 weeks (Figure 4d). These data indicate that changes in the spatial organization of nodes do not clearly correlate with the observed age-specific functional changes, and given the presence of structurally intact nodes in nerves at 4 postnatal weeks, it is unlikely that node distribution or formation is a major developmental regulator of nerve function.

FIGURE 4.

The proportion of Na_v 1.6+ nodes of Ranvier increases with age, while overall node density remains constant. Representative longitudinal scanning electron microscopy (SEM) images (a) taken from optic nerves at 4, 5, and 8 weeks show fully formed nodes. Nodes of Ranvier were identified by characteristic contactin-associated protein 1 (Caspr) staining and their level of maturity determined by the sodium ion channel subtype, either Na_v 1.2 (immature) or Na_v 1.6 (mature), present at the node (n = 4). Representative images (b) of optic nerves show Na_v 1.6-positive nodes present at all ages and Na_v 1.2-positive nodes (c) were found at 4 weeks (Caspr [red], Na_v 1.6 or Na_v 1.2 [green]). No significant changes in nodal density (d) were observed between 4 and 12 weeks (p = 0.0888); however, the ratio of Na_v 1.6-positive nodes (e) increased with after 6 postnatal weeks (p = 0.0062). Scale bar in (a) = 3 μm; in (b–c) = 1.5 μm, error bars = SEM

The number and ultrastructure of nodes did not show significant changes across age groups; however, the expression of specific ion channels at nodes can impact signal propagation. Previous studies have described a transition from Na_v 1.2 to Na_v 1.6 expression at nodes of Ranvier in the optic nerve (Boiko et al., 2001; Van Wart & Matthews, 2006). Optic nerves across ages were evaluated for expression of Na_v 1.6 (Figure 4b) and Na_v 1.2 (Figure 4c). When the proportion of Caspr-identified nodes with co-expression of the Na_v 1.6 ion channel subtype was normalized to the total number of nodes and compared across age group (Figure 4e), the percentage of Na_v 1.6-positive nodes was increased in older nerves (6, 8, 12 weeks). The Na_v 1.6 ion channel is suited for repetitive firing and may promote faster conduction along axons and therefore could contribute to shifts in axon population types at older ages; however, the shift in ion channel expression was detectable at 6 weeks, suggesting that a critical number of nodes expressing Na_v 1.6 may be necessary for the functional changes observed at 8 weeks.

3.5 |. Simulated compound action potentials are predominately regulated by sodium ion channel subtype, axon diameter, and myelin thickness

Based on the experimental data, an optic nerve conduction model was generated to evaluate simulated compound action potentials (sCAP) and to determine which variables are critical for optic nerve conduction. For each analyzed age, axon simulations were generated using experimentally derived values for axon diameter, sodium ion channel subtype, and g-ratio, which resulted in a range of conduction velocities. The summation of 500 extracellular potentials generated the sCAP, and a cluster analysis defined distinct axon populations within the optic nerve, allowing for the assessment of relationships between conduction velocity and sodium ion channel subtype, axon diameter, myelin thickness, and g-ratio within an axon population by nonlinear regression analysis (Figure 5a,b).

FIGURE 5.

Sodium ion channel type, axon diameter, and myelin thickness can regulate axon population conduction in simulated mouse optic nerves. A schematic of the conduction model is shown in (a). Briefly, simulations of action potentials were generated by inputting experimental values for axon diameter, g-ratio, and sodium ion channel type into a modified version of the axon conduction model described in Arancibia-Carcamo et al. (2017). Resultant intracellular action potentials were converted to extracellular potentials before summation to generate a simulated compound action potential (sCAP). A cluster analysis identified axon populations as clusters and properties of individual axons within those populations were evaluated to determine their significance in nerve conduction. Comparison of conduction velocity and axon diameter (left), myelin thickness (center), or g-ratio (right) of 500 axon simulations for 4, 5, 6, 8, and 12 postnatal weeks (b–f) show distinct axon populations depicted in different colors. Nav 1.6+ axons (x) are specific to faster axon populations, while Nav 1.2+ axons (o) are present in slower axon populations

The number of axon populations identified from sCAPs decreased from 4 to 3 between 6 and 8 week nerve simulations (Figure 5, Figure S2), reflecting a similar refinement of axonal conduction observed in experimental data (Figure 1e), where older ages maintain fewer axon populations. Cluster-defined axon populations composed two major sets with high or low conduction velocities that had specific ion channel subtype expression: axons with Na_v 1.6 ion channels had higher conduction velocities compared to those with Na_v 1.2 ion channels for all age groups (Figure 5b–f). Comparison of axon diameter and conduction velocity within axon populations (Figure 5b–f, left) revealed significant correlations (Figure S2C): for all ages, axons with Na_v 1.6 ion channels had positive linear or exponential relationships with axon diameter, suggesting that Na_v 1.6-expressing axons with large diameters have high conduction velocities; axons with Na_v 1.2 ion channels had a negative parabolic relationship with conduction velocity at all ages, suggesting that an optimal axon diameter exists to achieve the highest conduction velocity in Na_v 1.2-expressing axons. Myelin thickness had a significant, positive exponential relationship with conduction velocity across all age groups (Figure S2D, Figure 5b–f, center). G-ratio did not have significant correlations with conduction velocity for any axon population (Figure S2E, Figure 5b–f, right). Data generated by the nerve conduction model imply that, while ion channel subtype, axon diameter, and myelin thickness can impact axon population conduction, g-ratio is less significant.

4 |. DISCUSSION

Our detailed analysis of mouse optic nerve maturation identified a number of factors that may contribute to the postnatal development of signaling patterns critical for visual function. First, we show that optic nerve development is characterized by dynamic changes in nerve function and is considerably more protracted than previously thought, extending into at least the third month of life. Second, we identify functionally discrete populations of optic nerve axons with different conduction characteristics within the nerve. Third, at 5 and 8 weeks of age, there is a transition toward fast-conducting axons and a loss of slow-conducting axons, which correlates with changes in 14 kDa MBP expression, increases in myelin thickness, and decreases in g-ratios. Early shifts to larger axon diameters and later changes in sodium ion channel composition at nodes of Ranvier may differentially regulate nerve conduction. Fourth, a multifactorial model based on morphological analyses of the mouse optic nerve supports the regulatory role of axon diameter, myelin thickness, and sodium ion channel subtype in axonal conduction.

4.1 |. Novel functional properties of the mouse optic nerve during maturation

Varied axon diameter and the distribution of myelination in the optic nerve imply diversity in axonal conduction speed. This has been described as functionally-distinct axon populations in electrically evoked CAPs (Evans et al., 2010; Stys et al., 1991) which have been used to study the effects of anoxia (Tekkök et al., 2003), energy metabolism (Allen et al., 2006; Evans et al., 2013; Meakin et al., 2007), demyelinating diseases or injury (Bagchi et al., 2014; Gutiérrez et al., 1995; Ma et al., 2013), and monocular deprivation (Etxeberria et al., 2016). Our study, characterizing nerve function during postnatal development, identifies dynamic changes in CAP waveforms between 4 and 12 weeks of age, indicating that functional plasticity extends far beyond early nerve development. New functionally-distinct axon population types arise at 5 and 8 weeks, coincident with increases in total nerve function. The fastest axon populations are conserved across age group, consistent with unchanged response onsets, while functionally distinct axon populations shift at 5 and 8 weeks to favor fast conduction through increased recruitment of fast-conducting and loss of slow-conducting axons. These changes reflect a period of axonal refinement in the optic nerve, which likely optimizes the coding of visual signals and may be a consequence of experience-dependent functional plasticity as part of an extended critical period in visual development. Shifts in axon population conduction may be a mechanism of organizing visual information by either promoting synchronicity or distinguishing signals carrying specific visual information. The visual functionality associated with specific axon populations is unknown; however, studies suggest that axon populations may reflect distinct RGC subtypes (Evans et al., 2010) or axons with glucose-dependence (Allen et al., 2006; Meakin et al., 2007).

Previous studies evaluating mouse CAPs have grouped multiple ages together (i.e., 6–8 weeks, 8–12 weeks) and classified animals older than 6 weeks as adults or not specified age (Cases et al., 2017). Studies establishing the three-peak CAP waveform (Evans et al., 2010) used albino CD-1 mice that have abnormal distributions of contralateral- and ipsilateral-projecting RGCs (Jeffery et al., 1994; Prieur & Rebsam, 2017) compared to nonalbino C57Bl6 mice and may have a different composition of axon populations. Studies using C57BL6 mice at 4 and 5 weeks (Etxeberria et al., 2016; Evans et al., 2010) describe comparable measures of total CAP area, response onset, and duration to the studies described here, and report notable variation in number of peaks in the waveforms, ranging from 1 to 3. Analysis of axon populations in other studies (Evans et al., 2010) grouped them by their relative speed within each nerve prior to comparison between treatment groups (i.e., all 1st, 2nd, or 3rd peaks are compared across treatment groups). Our study accounts for the inherent variability of nerve responses by pooling peak data and distinguishing differences in nerve conduction by cluster analysis. However, a power analysis indicated that a relatively large sample size (n = 8) was required to determine whether statistically significant differences were detectable. A potential confound in these studies is that poor nerve viability, as a result of glucose or oxygen deprivation (Meakin et al., 2007; Tekkök et al., 2003), may alter nerve responses. The nerves in the current study were continuously perfused with an oxygenated and glucose-rich media during the length of the recording and selection criteria for nerve responses required CAP amplitudes to be greater than 10% of the observed noise, thereby eliminating extremely low-functioning nerves from analysis. In addition, the recordings were all completed within 30–90 min of optic nerve dissection, falling within the 2 h period before CAP area is reported to decline (Evans et al., 2010). Thus, the CAP waveforms in this study likely reflect true variation in axonal responses to an electrical stimulus.

4.2 |. Role of myelin protein expression and myelin thickness in regulating optic nerve function during maturation

Loss of myelin protein observed in MBP-deficient mouse models has been shown to correlate with deficits in neuronal function (Poggi et al., 2016; Sinha et al., 2006; Wang et al., 2012), yet the detailed functional significance of myelin distribution remains unclear. We observed limited changes in the relative expression of PLP or MBP isoforms compared to β-actin, with the exception of the 14 kDa MBP isoform which was dynamic between 4 and 6 weeks. While physiological changes in nerve function at 5 weeks coincided with increased expression of the 14 kDa MBP isoform, no functional changes were correlated with the decrease in this isoform at 6 weeks. Little is known about the role of this isoform in myelin compaction or synthesis, but our observations suggest that while increases in this isoform may contribute to the generation of fast conducting axons (the “Slow-intermediate” type), it may not be necessary to maintain axonal function. The western blot analyses do not distinguish between myelin and cytoplasmic-associated protein; however, our ultrastructural analyses of the optic nerve found increased myelin thickness and decreased g-ratios at 5 and 8 weeks with no change at 6 weeks. These trends suggest that 14 kDa MBP may facilitate the development of thicker myelin at functionally relevant time points and that a process of myelin rearrangement may occur at 6 weeks which maintains nerve function in the absence of additional myelin protein. The overall trend toward the increased expression of myelin proteins from 4 to 12 weeks may reflect a coincident increase in myelination and overall nerve growth.

Myelin thickness appears to have differential effects after initial myelination: when myelin thickness increased (g-ratios decreased) at 5 weeks, the new “Slow-Intermediate” axon population type is generated; when myelin thickness increased (g-ratios decreased) at 8 weeks, the “Slow” type is lost and more axons contribute to the existing fast axon type. This suggests a major role for myelin thickness, rather than myelin protein concentration, as a regulator of functionally distinct axon populations. We speculate that during initial myelination, developmental factors allow oligodendrocytes to increase myelin thickness over a diverse set of axons. By 6 weeks of age, optimal myelin thickness may be reached by axons contributing to all population types except for the “Slow” type, and increased conduction speeds and loss of the “Slow” type observed at 8 weeks could be a consequence of additional wraps of myelin around the slowest-conducting axons. While it is difficult to experimentally isolate the effect of myelin thickness in axonal conduction, previous studies in the optic nerve support that myelin thickness can impact function: hypomyelination and altered nodal properties result in visual deficits in demyelinated rat (Roncagliolo et al., 2006) and mouse (Lehman & Harrison, 2002; Mironova et al., 2016) models and hypermyelination occurring in mice with constitutively active Akt rapidly increases optic nerve conduction before severe anatomical changes resulted in conduction deficits (Yu et al., 2011). Notably, the observed correlation between dynamic myelination and functional plasticity supports previous literature describing the functional relevance of myelin adaptation in cognitive processes, such as learning and memory. Myelination may contribute to the refinement of neural activity generated by axonal properties, and the current study suggests that it may act as a developmental regulator of visual processing.

4.3 |. Role of specific adaptations in axon diameter, node density, and sodium ion channel expression in optic nerve function

Additional factors likely contribute to the age-associated regulation of functionally distinct axon populations during optic nerve maturation, including axon diameter, localization of nodes of Ranvier, and sodium ion channel expression. Variations in axon diameter have been shown to compensate for asynchronous neural activity, which would otherwise occur due to anatomical constraints in axonal signal conduction, such as axon lengths in rat olivocerebellar neurons (Lang & Rosenbluth, 2003; Sugihara et al., 1993) and mouse thalamocortical projections (Salami et al., 2003). Our data show that functional changes occurring in 5-week-old nerves are consistent with observed increases in axon diameter which would increase conduction speed. While changes in axon diameter were not reported in previous morphological studies (Dangata et al., 1995, 1996), we did observe stable numbers of axons, consistent with previous studies, indicating that RGC axonal growth was not a source of new functional axons. Because changes in axon diameter were limited to early ages, it may be that small axons with insufficient signal propagation trigger an increase in axon diameter, increasing both the likelihood of signal transmission and conduction speed, resulting in nerve responses with more functional, fast-conducting axons.

The placement and composition of nodes of Ranvier have been shown to regulate interaural time differences in avian auditory signaling (Seidl, 2014; Seidl et al., 2010) and in the mouse, these properties likely function to code specific frequencies in the olivocerebellar tracts (Ford et al., 2015). In addition, Exteberria et al. (2016) have shown that monocular deprivation at eye opening delays optic nerve conduction velocity of P32 mice and shortens internode lengths, indicating that experience-dependent activity can alter optic nerve myelination. Morphologically mature nodes of Ranvier were present in the youngest optic nerves, and no significant changes in density of Caspr+ nodes were observed during subsequent maturation. This suggests that dramatic changes in node distribution and by association, internode length, are not correlated with physiological nerve responses during the studied interval. However, comprehensive neuronal tracing or 3D reconstruction of individual myelinated RGC axons may detect subtle, functionally-relevant changes in node density and internode length.

While activity at nodes of Ranvier is not directly assessed in this study, the observed changes in sodium ion channel subtype expression occurring when the slowest-conducting axons were lost and additional fast-conducting axons were recruited suggest that ion channel subtype may be important for axon functionality. In the optic nerve, a developmental shift in expression of Na_v 1.2 (expressed along unmyelinated regions of axons) to the Na_v 1.6 (expressed at mature nodes as myelination occurs) at nodal regions is coincident with shifts from individual spike activity to repetitive firing in RGC axons (Van Wart & Matthews, 2006). In addition, myelindeficient shiverer mice lack Na_v 1.6 channels and rely on Na_v 1.2 for successful signal propagation along their axons (Boiko et al., 2001). We found that expression of Na_v 1.6 channel at Caspr+ nodes increased at 6 postnatal weeks, prior to changes in nerve function. We speculate that a process of node maturation, where Na_v 1.6 is the dominant ion channel, occurring at 6 postnatal weeks affects conduction in a subpopulation of axons, which becomes functionally-relevant by 8 weeks. Our study suggests that the composition of ion channels at nodes of Ranvier is likely an important factor in axonal conduction during postnatal development of the optic nerve. However, further analysis of sodium ion channel subtype expression, ion channel density, and assessments of the functional relevance of the Na_v 1.6 subtype at nodes of Ranvier would help clarify the physiological relevance of ion channel subtype expression.

4.4 |. Optic nerve conduction model predicts substantial impact of sodium ion channel subtype, axon diameter, and myelin thickness on axon population conduction velocity, but limited influence of g-ratio

The experimental data suggest that a number of parameters may affect optic nerve axonal conduction during maturation. To gain insights into the relative importance of each parameter, a model of optic nerve conduction was developed. Modeling optic nerve conduction presents a considerable challenge due to the number of parameters involved and dispute over the weights of these parameters based on discrepancies in experimental data collection and analysis. Some simulations of nerve conduction suggest that internode length is the most efficient and energetically favorable regulator (Arancibia-Cárcamo et al., 2017), while others indicate that myelin thickness (and myelin capacitance) determines axonal conduction (Moore et al., 1978; Schmidt & Knösche, 2019). We developed a novel model of mouse optic nerve conduction which allows for the investigation of the relationship between specific axonal or myelination characteristics (axon diameter, myelin thickness, g-ratio, sodium ion channel subtype) and axonal conduction. Because this model uses experimentally derived values where possible, the simulated nerve responses more closely reflect true biological responses when compared to a model that relies exclusively on values derived from the literature, which may not reflect comparable optic nerve recording conditions. The current model is not a perfect representation of nerve conduction and relies on several assumptions: (1) Axons are uniformly myelinated with no unmyelinated regions along the axon or variations in number of wraps or myelin compaction, (2) the only functionally relevant ion channels present at nodes are sodium, potassium, and leak ion channels, (3) the number and density of ion channels are constant and uniform for all nodes, (4) the CAP is a simple summation of extracellular action potentials and has no influence from the nearby activity of glial cells or other axons, and (5) the level of stimulation will be sufficient to ensure successful propagation along the length of the axon, therefore omitting conditions where action potentials fail to reach the end of the optic nerve.

Several key findings were evident from the simulation model. Clustering of axon populations in sCAPs mimicked experimental trends across age group, where the complexity of axon population conduction was reduced in older nerve responses. This supports the hypothesis that the diversity of axon population conduction is linked to one or multiple observed, morphological properties of the nerve. The model demonstrated that sodium ion channel subtype can determine fast or slow axon populations, which is consistent with known properties of the two subtypes: Na_v 1.2 can sustain slower signal conduction along unmyelinated axon segments (Boiko et al., 2001), and Na_v 1.6 is suited for repetitive firing (Van Wart & Matthews, 2006) and reflects node maturity. Interestingly, the model also revealed an interaction between axon diameter and Na_v 1.6/1.2 expression, where large diameter axons expressing Na_v 1.6 will exponentially increase conduction velocity, while Na_v 1.2-expressing axons have their fastest conduction at an optimal axon diameter. These data support the idea that axon diameter and ion channel expression may be mechanisms of organizing visual signals. We speculate that morphological changes between 4 and 6 weeks (increased axon diameter and expression of Na_v 1.6+ nodes) contribute to increased range of axon population conduction, favoring faster conduction overall. However, in the optic nerve, expression of the number and density of sodium ion channel subtypes at nodes may be variable along the axon, resulting in a greater diversity of axon population conduction.

Although increased myelin thickness at key functional time points (Figures 2 and 3) suggests it may contribute to increased axonal conduction, our model indicates that its effect may be indirect and not specific to axon populations. Increased myelin thickness resulted in exponential increases in axonal conduction across all axon populations at all ages suggesting that myelin thickness contributes to overall increases in conduction speed, but does not have a differential effect on axon population conduction. Consequently, myelin thickness does not directly distinguish between axon populations and is a less critical regulator of nerve conduction compared to axon diameter or sodium ion channel subtype expression. Lastly, our model showed that g-ratio does not have a strong relationship with axon population conduction velocity, suggesting that the ratio of myelin thickness to axon diameter may not be a critical factor in altering the range of nerve conduction in this system. Observed decreases in g-ratio may instead reflect the development of mechanisms that regulate axonal function independent of the supramaximal nerve stimulation evaluated in this study, such as a visually-evoked response where additional wraps of myelin may ensure signal propagation of subthreshold neural activity. Additionally, g-ratio may promote synchrony or temporally separate ganglion cell firing patterns in response to a specific visual stimulus, which would not be represented by the current simulation model. While the properties that define nerve function are likely multifaceted, our model implicates sodium ion channel subtype, axon diameter, and to a lesser extent, myelin thickness, as potentially important regulators of the development of functionally-distinct axon populations.

4.5 |. Additional factors modulating axonal conduction in the optic nerve

Our study has identified several key factors contributing to signal propagation along RGC axons, yet additional properties may govern optic nerve conduction. Voltage-gated potassium ion channels, K_v 1.1 and K_v 1.2, are localized to juxtaparanodal regions in optic nerve axons and could play a role in refining axon population conduction. Ectopic expression of K_v 1.1 in internode segments of demyelinated optic nerves resulted in reduced CAP areas which were partially restored by application of 4-aminopyridine (Bagchi et al., 2014). This implies a role for K_v 1 currents in maintaining signal propagation at nodes of Ranvier. A comprehensive study of the density and distribution of potassium and other ion channels at nodes during optic nerve mouse development would provide insight into the functional role of potassium ion channels. In addition, node of Ranvier length may alter ion conductance through varied expression of sodium and potassium ion channel subtypes. Preliminary assessment of node of Ranvier length in normal optic nerves aged 4, 5, and 8 postnatal weeks using SEM suggested no clear differences in node length (data not shown) and consequently was kept constant in our conduction model.

Expression of ion channels along proximal regions of RGC axons may be relevant in axonal conduction. The axon initial segment is populated with sodium and potassium ion channels and is associated with altering neuronal properties, such as improving sensitivity to stimuli (Kuba et al., 2006). RGC axon initial segment length may be specific for distinct RGC subtypes (Raghuram et al., 2019), which implies a potential role in distinguishing axon populations in the optic nerve. Our evaluation of optic nerve function lacks the axon initial segment; however, analysis of axon initial segments using a novel electrophysiology preparation where the retina is attached to the optic nerve would provide insight into the relationship between functionally-distinct axon populations in the nerve and specific RGC subtypes.

5 |. CONCLUSION

This study describes significant heterogeneity in optic nerve axonal properties and an extended period of plasticity in nerve function and myelination during postnatal development in the mouse. Refinement of axonal function favoring faster conduction correlates with increased myelin thickness and altered molecular composition of myelin; however, regulation by axon diameter and ion channel expression at nodes of Ranvier may be age-specific. Modeling CAP responses revealed that sodium ion channel subtype, axon diameter, and myelin thickness may have greater impact on nerve function than g-ratio. While mechanisms underlying axon population conduction remain complex, the correlation between remodeling of oligodendrocyte myelination and the distribution of axon populations highlights the precise temporal control of glial cells in CNS function. While specific genetic and pharmacological tools targeting selected properties of myelin are still needed to further evaluate the dependence of axonal function on myelin without disrupting neighboring myelination properties, the extended dynamic nature of myelin and its functional relevance makes it a valuable target for regenerative therapies of demyelinating diseases.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.