What is the optimal distribution of myelin along a single axon?

Highlights

• Ion channel distribution determines the optimal pattern of partial myelination.

• Short node-internode pairs yield fastest conduction if channel number is conserved.

• Conduction velocity is insensitive to node-internode pair length if channel density is constant.

Abstract

The myelin sheath that insulates some axons in the central nervous system allows for faster signal conduction. Previously, axons were thought to be either unmyelinated or fully myelinated. Recent experimental work has discovered a new pattern of myelination (intermittent myelination) along axons in the mouse brain, in which long unmyelinated axon segments are followed by myelinated segments of comparable length. We use a computational model to explore how myelin distribution (in particular intermittent myelination) affects conduction velocity. We find that although fully myelinated axons minimize conduction velocity, varying the spatial distribution of a fixed amount of myelin along a partially myelinated axon leads to considerable variation in the conduction velocity for action potentials. Whether sodium ion channel number or sodium ion channel density is held constant as the area of the unmyelinated segments increases has a strong influence on the optimal pattern of myelin and the conduction velocity.

Introduction

Some axons in the vertebrate nervous system are wrapped with layers of myelin, which insulate these axons allowing for faster conduction of action potentials. The myelin sheath is produced by Schwann cells in the peripheral nervous system and oligodendrocytes in the central nervous system (CNS) [18], [19]. Unlike Schwann cells, which act on single peripheral axons, oligodendrocytes in the CNS ensheath up to 50 axons [7], allowing them to exert influence on neural processing on a larger scale. An individual oligodendrocyte, or a cluster of neighbouring oligodendrocytes, can have a large number of nearby axons available to myelinate, but experimental data [7] supports the proposition that the choice of which axons to myelinate is not made at random [27]. Moreover, myelination has been shown to be a dynamic process that responds to environmental cues [6], [14].

In a recent study, Tomassy et al. [25] analysed high-resolution maps of myelination by tracing high-throughput electron microscopy reconstructions of single axons of pyramidal neurons in the mouse brain. Analysing neurons in layers II/III from a publicly available dataset of a region of the mouse visual cortex [2], [3], they observed a new pattern of myelination. Historically, axons were thought to be either fully myelinated or unmyelinated. However, when tracing neurons in layers II/III, Tomassy et al. found that 17 out of 22 neurons displayed a pattern of myelination in which myelinated axon segments are interspersed with long unmyelinated segments, and they called this newly identified pattern “intermittent myelination” (IM). In this myelination pattern, unmyelinated sections of these axons were observed to be up to 55 μm long, much longer than typical nodes of Ranvier (approximately 1 μm long).

Since the discovery of intermittent myelination of (excitatory) pyramidal neurons by Tomassy et al. [25], Micheva et al. [15] also observed that the distribution of myelin in layer II/III inhibitory neurons was “patchy” with myelinated segments preferentially located near the cell body. Tomassy et al. [25] noted that neurons in layer II/III are involved in more complex cortical functioning than those found in layers V and VI, where IM was not observed. This raises the possibility that IM may be facilitating more complex neural functioning.

Whilst the evidence that signalling between electrically active axons and oligodendrocyte progenitor cells provides an important cue for inducing OPCs to differentiate into myelinating oligodendrocytes is firmly established, see [16], and the evidence that myelin provides more than just insulation to maximise conduction velocity (CV) continues to mount [9], the mechanisms controlling active myelination remain largely unknown (reviewed in Snaidero and Simons [23]).

The discovery of IM by Tomassy et al. raises several questions. What is the purpose of IM? Does an IM distribution provide any advantages to signal propagation over a fully myelinated pattern? Obviously, the shortest conduction time will be for a fully myelinated axon. However, it is not obvious which distribution of myelin will maximise CV when only a fixed fraction of the length of an axon is to be myelinated. In this instance, the distribution of ion channels also becomes important.

We examine the consequences for CV of the partial myelination of axons. The geometry of our model depends crucially on one parameter, L, which is the length of both the myelinated segments and the unmyelinated segments. This pair is repeated periodically over a fixed length of axon. We only study the effects of myelin distribution on signal CV but we note that there are several other facets of neural information processing that depend on myelin distribution, such as the reliability of action potential (AP) propagation [8] and the energy consumption [11], [20], [21].