|Home | About | Journals | Submit | Contact Us | Français|
A brainstem circuit for encoding the spatial location of sounds involves neurons in the cochlear nucleus that project to medial superior olivary (MSO) neurons on both sides of the brain via a single bifurcating axon. Neurons in MSO act as coincidence detectors, responding optimally when signals from the two ears arrive within a few microseconds. To achieve this, transmission of signals along the contralateral collateral must be faster than transmission of the same signals along the ipsilateral collateral. We demonstrate that this is achieved by differential regulation of myelination and axon caliber along the ipsilateral and contralateral branches of single axons; ipsilateral axon branches have shorter internode lengths and smaller caliber than contralateral branches. The myelination difference is established prior to the onset of hearing. We conclude that this differential myelination and axon caliber requires local interactions between axon collaterals and surrounding oligodendrocytes on the two sides of the brainstem.
The evolution of myelin in the vertebrate brain increased the speed at which neuronal information can be conveyed. Myelin also provides a mechanism to systematically and dynamically adjust conduction times of information flow between neuronal centers (Nave, 2010). While regulation of conduction time along axons is a common occurrence in the peripheral and central nervous system, its functional significance is not always clear (Kimura and Itami, 2009; Seidl, 2014). Recent studies in the avian auditory brainstem show how systematic regulation of conduction time is tied to a well-defined function in a neuronal circuit: Conduction times of binaurally elicited phase locked action potentials must be calibrated to ensure arrival within a small time window at the coincident detector neurons (Carr and Macleod, 2010; Ashida and Carr, 2011). We showed that variations in internode length and axon diameter are used to decrease the speed of signal propagation in the shorter axon branch relative to the longer axon branch (Seidl et al., 2010; Seidl et al., 2014). This conduction velocity regulation results in an equalization of conduction times with the required microsecond precision and enables the resolution of ITDs as small as 10 μs (Fischer and Seidl, 2014).
The mammalian sound localization circuit is comprised of coincidence detector neurons in the medial superior olive (MSO) that receive binaural excitatory inputs from neurons in the anteroventral cochlear nucleus (AVCN; Fig. 1A). Individual AVCN neuron axons bifurcate to send one collateral to the ipsilateral MSO and a second collateral to the contralateral MSO. Conduction velocity must be precisely timed to shift the best response of MSO cells into the physiological range of ITDs (Brand et al., 2002, Fig. 3c,d, when inhibition is blocked best ITDs based on excitation alone shift to 0 ITD; van der Heijden et al., 2013, Fig. 3B; Franken et al., 2015, Fig. 2e, most best ITDs are around 0 ITD). Because the hearing range of the Mongolian Gerbil is similar to humans, it has become an important laboratory animal for studies on sound localization. The physiological range of ITDS in gerbils (the range of naturally occurring ITDs) is ±110–135 μs (Maki and Furukawa, 2005; see also Sterbing et al., 2003). The locations of the MSO nuclei interject a length disparity between the ipsi- and contralateral inputs, creating a major challenge for coincident input of action potential timing from the two ears (Stotler, 1953). We therefore hypothesized that axon diameters and internode lengths differ between ipsilateral and contralateral axons to adjust conduction times along the axonal inputs to provide coincident inputs to MSO neurons at naturally occurring ITDs.
We analyzed internode length and axon diameter in ipsilateral and contralateral branches of the AVCN axons projecting to MSO of the gerbil. Our data suggest that differential myelination and variations in axon diameter in two different branches of individual axon collaterals are a shared principle in the auditory systems of birds and mammals, and are utilized to achieve binaural coincident inputs in the microsecond range.
We used 44 gerbils (Meriones unguiculatus) of either sex, purchased from Charles River Laboratories (Wilmington, MA). Animals were used on the day of delivery to the University of Washington. All procedures were approved by the University of Washington Institutional Animal Care and Use Committee and conformed to National Institutes of Health guidelines.
Animals were anesthetized with isoflurane, decapitated and the brain was extracted from the skull. After extraction, a biocytin crystal (Sigma-Aldrich, St. Louis, MO) was placed onto AVCN on one side and the brain was returned to oxygenated ACSF (Mathews et al., 2010). Following 2–4 h incubation at room temperature (RT), the brain was immersion fixed in 4% paraformaldehyde for 12 h at 4°C. Longer incubation times resulted in tissue deterioration and were avoided. Incubation times were sufficient to label contralateral axons up to 4 mm. After rinsing in phosphate-buffered saline (PBS), the brainstem was coronally sectioned at 75 μm on a vibratome. The resulting sections were counterstained with fluorescent streptavidin (1:1000 Alexa Fluor® 568 or 594, Molecular Probes, Grand Island, NY) and an antibody against the paranodal protein Caspr (1:1000, NeuroMab clone K65/35, NeuroMab, University of California, Davis, Davis CA) in a standard block solution (0.3% Triton X-100 (Sigma-Aldrich), 5% normal goat serum) for at least 6 h at RT or overnight at 4°C. The sections were then rinsed in PBS and exposed to a secondary antibody (1:200, Alexa Fluor 488 goat anti-mouse, Molecular Probes) in block solution overnight at 4°C. After a final rinse in PBS, the sections were coverslipped with Glycergel (Dako, Carpinteria, CA).
Animals in which biocytin placements successfully labeled axon collaterals were included in the data analysis (P10: 5; P20: 5). Sections containing both labeled AVCN to MSO axons and paranodal staining were imaged using a confocal microscope (Fluoview 1000, Olympus, 20x air lens, NA 0.75, Center Valley, PA, or Leica TCS SP8, 20x air lens, NA 0.75 and 60x water, NA 1.4, Leica, Buffalo Grove, IL) and internode length was measured in 3-D as described in Seidl et al. (2010). In most cases we made 2 to 4 individual measurements in a single axon; these were averaged to provide a single measurement for that segment. All well-filled axons that could be isolated from adjacent axons were included in our analyses.
We determined the diameter of axon segments in which internode length was measured. Optical sections were imported to Fiji software (National Institutes of Health, Bethesda, MD) and the full width at half maximum of individual axons segments was determined using the FWHM plugin (John Yew Soon Lim, 2011, Institute of Molecular Biology, National University of Singapore). Diameter was measured in between nodes of Ranvier of axons segments used for analysis of data in Figure 3 and average diameter values of 2–4 individual measurements in a single axon were used for analysis. A recent study measuring axon diameter of spherical bushy axons extending contralaterally to MSO in the gerbil shows similar values for axon diameter as we report, validating our measurements of the contralateral axon (Ford et al., 2015; Figure 1e). In a previous study we validated axon diameter measurements with electron microscopy and found no significant difference between the groups (Seidl et al., 2010).
Images in figures 2B,D,F and H were surface-rendered with Huygens Essential (Scientific Volume Imaging, Hilversum, The Netherlands).
Statistical analyses and graphs were created using Prism (GraphPad Software, La Jolla, CA). Differences between groups were tested with an unpaired t test. All in-text and graphic representations of data illustrate the mean ± standard error of the mean. We performed a linear regression analysis and computed Pearson Product Moment correlations for the data presented in Figure 4 B and D.
We analyzed internode length in AVCN axons of 22 ipsilateral axon branches and 20 contralateral axon branches from 5 P20 gerbils. Individual measurements in a single axon were averaged to provide a single measurement for that collateral. The exact location of the branch point initiating the ipsilateral and contralateral axons is not as evident in the gerbil as in the chicken brainstem. For this reason labeled axon collaterals for analysis were identified by their location proximal to the respective MSO and the fact that they either terminated in the MSO dendrite region or displayed repeated branching, identifying them as terminal arbors (Fig. 1B–D). Internode length measurements were taken proximal to where the axons entered the dendritic area of MSO. Figure 2 shows representative examples of labeled ipsilateral (A,C) and contralateral (E,G) axon collaterals (magenta) that were counterstained with an antibody against the paranodal anchoring protein Caspr (green). Corresponding 3-D reconstructions of the axon segments in Figure 2 confirm the integrity of the labeled fibers and the location of the paranodes (Fig. 2B,D (ipsilateral) and F,H (contralateral)). At P20, average internode length in the ipsilateral axon was 51 ± 6 μm, (SEM). The average in the contralateral axon collaterals was 95 ± 9 μm (Fig. 3A). Internode length was thus 1.85 times longer in the contralateral axon compared to the ipsilateral axon (p = 0.0002).
At P10, 2 days before hearing onset (Woolf and Ryan, 1984), the difference in internode length was already established. We determined internode length in 20 ipsilateral axon collaterals and 17 contralateral axon collaterals from 5 P10 gerbils (Fig. 3B). Average internode length for the ipsilateral axon collaterals was 54 ± 6 μm and 126 ± 13 μm for the contralateral axon collaterals. The distance between nodes of Ranvier in the contralateral axon branch was thus 2.3 times the value of the ipsilateral branch (p < 0.0001).
To further examine the development of axon parameters that are known to influence conduction velocity, we analyzed axon diameter of the axon segments in which internode length was determined. Fiber diameter was measured between nodes of Ranvier identified on axon segments used for analysis for data in Figure 3 and average values were used for analysis. At P20, axon diameter was larger in the contralateral axon segments (1.27 ± 0.03 μm, n = 20) compared to the ipsilateral axon segments (1.00 ± 0.04 μm, n = 22, p < 0.0001, Fig. 4A). Axon diameter measurements are correlated to corresponding internode length values at P20 (Fig. 4B, Pearson r = 0.4805). At P10, we calculated average axon diameter to be 1.04 ± 0.04 μm (n = 21) in the ipsilateral and 1.08 ± 0.04 μm (n = 17) in the contralateral axon segments (Fig. 4C). There was no correlation of individual axon diameter values with corresponding internode length at P10 (Fig 4D, Pearson r = 0.1765).
In gerbils that have mature hearing capability we show that internode length is shorter in the ipsilateral axon branch compared to the contralateral branch of AVCN axons that project to MSO. In addition, axon diameter is increased in the contralateral axon segment compared to its ipsilateral counterpart. This finding implies that conduction velocity is regulated differentially in the two collaterals to compensate for the longer travel distance of the contralateral input to MSO. Anatomical restrictions are thus counterbalanced by variations of internode length and axon diameter, and the resulting change in conduction velocity. This difference in internode length is already established at P10, 2 days before the onset of hearing, but the maturation of differential axon diameter appears to lag. Our findings mirror an earlier study on internode length and axon diameter differences in the avian ITD processing circuit (Seidl et al., 2010), leading to an equalization of conduction times (Seidl et al., 2014). Combined, our studies suggest that differential myelination and differences in axon diameter along individual segments of a single axon form a general mechanism to regulate conduction time in the brainstem auditory pathways.
We measured fiber trajectories of labeled axons in 7 sections of 5 different animals. Pathway measurements are a good estimate of axon length (Seidl et al., 2010). Average pathway length from the ipsilateral AVCN to MSO was 1403 ± 771 μm and from the contralateral AVCN to MSO was 3541 ± 380 μm, resulting in an average difference of input length of 2138 ± 102 μm (n = 7). If action potentials were conducted with uniform speed along the AVCN axon, binaural inputs to MSO would differ by 267 to 1069 μs (based on conduction velocities ranging from 8 to 2 m/s). Internode length is a major determinant of conduction velocity (Brill et al., 1977) and axon branches of the analogous circuitry in the bird exhibit a 1.95-fold difference in internode length (Seidl et al., 2010), which translates to a 2.39-fold difference in conduction velocity (Seidl et al., 2014). In the gerbil, internode lengths of AVCN axon branches differ by a factor of 1.85. This difference in internode length can thus compensate for the longer contralateral input length, especially if one considers that internode length is one of several properties regulating action potential conduction velocity.
The avian ITD detection circuit embodies a modified Jeffress model (Carr and Konishi, 1990; Young and Rubel, 1983; Overholt et al., 1992; Köppl and Carr, 2008; Seidl et al., 2010). In contrast, the exact mechanisms of ITD coding in the mammalian auditory brainstem are still being resolved. Originally MSO and its incoming projections were also considered to compose a circuit working according to the Jeffress place theory of sound localization (Yin and Chan, 1990; Smith et al., 1993; Beckius et al., 1999). In vivo recordings in the gerbil and guinea pig, however, show that best ITDs are not distributed across the range of naturally occurring ITDs, but support a population based rate code (McAlpine et al., 2001; Brand et al., 2002; Seidl and Grothe, 2005). In addition to their excitatory inputs, MSO neurons receive prominent glycinergic inhibition (Grothe and Sanes, 1993; Grothe and Sanes, 1994). These inhibitory inputs to MSO are anatomically and physiologically specialized for temporally precise synaptic transmission (Kapfer et al., 2002; Magnusson et al., 2005; Werthat et al., 2008; Couchman et al., 2010) and appear to play a role in the tuning of ITDs (Brand et al., 2002; Pecka et al., 2008; Myoga et al., 2014). The exact role of this inhibition is disputed. Other studies suggest ITD tuning is determined by the relative timing of binaural excitation to MSO (Day and Semple, 2011; van der Heijden et al., 2013; Roberts et al., 2013) and argue against a role of inhibition in ITD tuning (Franken et al., 2015). Regardless of the role of inhibition, however, precisely timed binaural excitation is mandatory for MSO function according to all proposed mechanisms. Our study supports the notion that conduction time of binaural excitatory projections to MSO is precisely regulated. As such, anatomical restrictions are overcome by variations in internode length and fiber diameter. Recent studies by Roberts and colleagues used an elegant in vitro preparation to further show that excitatory input delays to MSO do not differ (Roberts et al., 2013), and thus support our conclusions of temporal compensation of inputs. Our data do not allow any conclusions about the role of inhibition in ITD coding. On the other hand, internode length along axons projecting to the medial nucleus of the trapezoid body (MNTB), which in turn provides an inhibitory input to MSO, appears to be specialized for fast signal transmission (Ford et al., 2015), supporting the concept of temporally precise glycinergic inhibition of MSO neurons (Kapfer et al., 2002).
Variations in myelination of axons are a common occurrence in the central nervous system (E.g., Bennett, 1970; Waxman, 1971; Sugihara et al., 1993; Lang and Rosenbluth, 2003; Salami et al., 2003; Tomassy et al., 2014; for review see Kimura and Itami, 2009; Seidl, 2014; de Hoz & Simons, 2015 ), yet the functional implications of this variability are not always clear. The ITD detection networks of birds and mammals provide a unique preparation in which the normal function of this network can only be realized when the timing of action potentials emanating from the two ears are regulated in the microsecond range. Therefore, the selection pressure to modulate conduction times is clear and serves a well-defined computational function. We find that contralateral fibers display interspersed short internodes whose length is only a fraction of a regular internode. This is not without precedence, as short internodes have been described before (Waxman, 1971; Waxman and Melker, 1971). Variations in the geometry of the myelin sheath may provide a mechanism for temporal adjustment or more complex transformations of neural information in axons. It seems possible that short internodes are introduced for fine tuning conduction times to achieve specifically timed inputs to MSO.
Axon diameter in the contralateral axon segments is on average larger than in the ipsilateral segments at P20. This suggests that fiber caliber contributes to an increase in signal transmission speed in the contralateral axon branch projecting to MSO. However, axon diameter does not differ between the two axon branches at P10. Therefore, it appears that the intraaxonal difference in internode length, already established at P10, precedes differentiation in axon diameter.
Node of Ranvier assembly, node spacing, and axon diameter are controlled by myelinating glia (de Waegh et al., 1992; Garcia et al., 2003; Court et al., 2004; Susuki and Rasband, 2008). Studies of mature systems indicate that axon diameter and internode length in the peripheral nervous system are highly correlated (E.g., Vizoso and Young, 1948; Hess and Young, 1949; for review see Waxman, 1975). Our data support a significant correlation between these parameters in individual axons at P20. Both axon diameter and internode length are increased in the contralateral axon segments relative to the ipsilateral segments. Surprisingly, there appears to be no overall difference in axon caliber between the collaterals and the correlations between parameters do not approach significance at P10. This suggests that internodal length and axon caliber can be independently regulated.
We have now described data in homologous pathways in two divergent species indicating that separate collaterals of individual parent axons precisely regulate the interactions with surrounding neural elements to achieve markedly different conduction velocities. This results in precise integration of the timing of information from different sources to the postsynaptic neurons. While the resulting functional attributes are obvious for the processing of binaural acoustic information, the fundamental principal described in these papers has importance for understanding the temporal integration of signals required for all sensory information coding and motor coordination, particularly fine motor skills. Our data and interpretations thereby warrant some speculations about how the precise regulation of action potential velocity in these separate collaterals of the same parent axons is maintained and how it emerges during ontogeny, a period of substantial brain growth (Wilkinson, 1986). The ipsi- and contralateral axon collaterals of individual AVCN neurons are spatially separated within the brainstem. Differentially expressed genes along the pathways of the ipsilateral and contralateral trajectories could lead to differences in molecules in axons or in oligodendrocytes specifying the differences in internode length. To date no such molecule has been identified. There are multiple lines of evidence to indicate that excitatory neuronal activity instructs myelination (Barres and Raff, 1993; Demerens et al., 1996; Stevens et al., 1998; Ishibashi et al., 2006; Li et al., 2010; Goldsberry et al., 2011; Bergles et al., 2000; Lin et al., 2005; Káradóttir et al., 2005; Ziskin et al., 2007; Mangin et al., 2008; Gibson et al., 2014; Fannon et al., 2015). Specifically, electrically active axons can induce myelin formation by vesicular release of glutamate that signals NMDA receptors (NMDAR) on nearby OLs (Wake et al., 2011). Neuregulin (NRG1) appears to increase NMDAR currents in oligodendrocytes 6-fold (Lundgaard et al., 2013). Thus, specific NRG1-expression could enhance the sensitivity of oligodendrocytes to electrical activity from axons, caused by spontaneous activity before hearing onset (Jones et al., 2007; Walsh and McGee, 1988), and thus influence myelination. It has been demonstrated that different axons myelinated by a single oligodendrocyte can have myelin sheaths of different thicknesses, with myelin thickness dictated by the axon (Waxman and Sims, 1984). Other factors, such as Wnt (Guo et al., 2015), may influence myelination and internode length regulation as well. Future studies are needed to unravel the mechanisms of conduction velocity regulation and we think the auditory brainstem can serve as a useful tool to study how internode length is regulated, particularly in conjunction with transgenic mouse models (Pfrieger and Slezak, 2012).
We thank David J. Perkel for comments on the manuscript. A.H.S. was supported by the Virginia Merrill Bloedel Hearing Research Center, the Hearing Health Foundation and NIH grant DC011343. This work was also supported by NIH grants DC003829 and DC004661.
Competing financial interests
The authors declare no competing financial interests.