The role of mechanics in axonal stability and development

Abstract

Much of the focus of neuronal cell biology has been devoted to growth cone guidance, synaptogenesis, synaptic activity, plasticity, etc. The axonal shaft too has received much attention, mainly for its astounding ability to transmit action potentials and the transport of material over long distances. For these functions, the axonal cytoskeleton and membrane have been often assumed to play static structural roles. Recent experiments have changed this view by revealing an ultrastructure much richer in features than previously perceived and one that seems to be maintained at a dynamic steady state. The role of mechanics in this is only beginning to be broadly appreciated and appears to involve passive and active modes of coupling different biopolymer filaments, filament turnover dynamics and membrane biophysics. Axons, being unique cellular processes in terms of high aspect ratios and often extreme lengths, also exhibit unique passive mechanical properties that might have evolved to stabilize them under mechanical stress. In this review, we summarize the experiments that have exposed some of these features. It is our view that axonal mechanics deserves much more attention not only due to its significance in the development and maintenance of the nervous system but also due to the susceptibility of axons to injury and neurodegeneration.

1. Introduction

Neuronal cells have evolved to span long distances within an organism by extending thin tubular protrusions called axons and dendrites. Of these axons are of particular interest as they often reach extreme lengths, of the order of a meter in a human and tens of meters in a blue whale. Such spatial scales pose several challenges to the neuronal cells, which include the need to transport materials over long distances, regulation of growth and retraction dynamics, and maintenance of structural integrity.

It is becoming increasingly clear that mechanical forces are critical in neuronal development, stability and function (recently reviewed in [1–3]). For example, it has been shown that mechanical tension regulates axonal initiation [4,5], growth [4,6], and synaptogenesis [7]. It has also been proposed that mechanical tension generated by axons may play an important role in the formation of cortical folds in primates [8,9].

Perturbations to mechanical balance between the cytoskeletal elements cause axonal retraction and this involves actin and microtubule-associated molecular motors, polymerization dynamics, and passive cytoskeletal mechanics [10–12]. Such perturbations to axons in vitro cause morphological changes like axonal beading or spheroids like those seen in pathological conditions, and loss of mechanical balance is implicated in this as well [13,14]. Thus understanding the roles of active and passive forces in axons is expected to be relevant in understanding axonal degeneration as well as rewiring.

Even under normal functioning, axons need to withstand large stretch deformations in animals during movement [15,16]. In extreme cases, it can be as large as 160% as in the case of some baleen whales and hence may have evolved various mechanical strategies for neuro-protection [17]. Extreme stretch leads to axonal damage, which includes stretch injury to nerves, concussion, and traumatic brain injury. Nerve compression resulting from injury or ageing-related conditions also highlights the role of axonal mechanics in debilitating conditions.

In this review, we discuss our current understanding of axonal mechanics in a few select contexts. We restrict ourselves to the cytoskeletal and membrane of the axonal shaft and how their mechanics influences axonal stability. We begin this review by describing some of the salient or unique features of the axonal cytoskeleton. We then give a summary of some of the technical developments in the field that has enabled us to probe axonal mechanical responses of this composite cytoskeleton. The cytoskeletal responses to stretch are divided into short timescale passive and active behaviors followed by long-timescale growth effects mainly for ease of discussion. As the axon is highly anisotropic, radial compression response is discussed separately along with mechanical aspects of calibre maintenance and shape instabilities arising out of cytoskeletal perturbations. Finally, we discuss the often ignored mechanical aspects of the axonal membrane, and its relevance to the setting of optimal calibre and abnormal shape transformations. Throughout this review, we have tried to bring out the complex responses of the axonal ultrastructure focussing our attention mainly on experiments.

2. Axon ultrastructure

The axonal cytoskeleton and its dynamics is central to the mechanical regulation of axonal development and function. While there is significant diversity in the organisation of the cytoskeleton across neuronal types and species, as reviewed recently in [18,19]), certain general organizational features are highlighted here and summarized in Fig. 1.

Fig. 1.

A simplified schematic of the distal axonal cytoskeleton. Typically, crosslinked microtubules form the core of the distal axon and are embedded in a space-filling gel of intermediate filaments (loose arrangement of microtubules interspersed with neurofilaments is also commonly seen). Circumferentially arranged rings of actin filaments form a sub-membrane periodic scaffold and these are interconnected by spectrin tetramers. Other forms of cortical actin and ‘deep’ actin also exist (not shown; see main text for description). A large number of proteins act as cross-linkers connecting the different polymers to each other and to the axonal membrane. Motor proteins generate forces among filaments, and polarized microtubule tracks enable cargo transport (not shown). (adapted from [45] under Creative Commons Attribution Licence)

Axonal microtubules are oriented with their fast growing plus ends pointed distally away from the cell body and undergo growth, shrinkage and severing mediated by enzymes like spastin and katanin [20,21]. Individual microtubules do not span the length of the axon. Estimates of the length of microtubules in C. elegans yields an average of ~6 µm [22], and in vertebrate neurons in vivo, lengths of up to several 100 µm have been reported [23,24]. The density of microtubules per square micrometre in cross-section can range from 10 to 100 depending on the calibre of the axon [25,26]. Electron microscopy reveals that microtubules are often fasciculated in the axon initial segment, however, in the distal axon they appear loosely crosslinked with cross-bridges. These inter-microtubule interactions are governed by microtubule-associated proteins (MAPs) (recently reviewed in [27]), like the tau protein, which can have a profound effect on the mechanical properties of the microtubule network [28].

Vertebrate axons contain parallelly arranged heteropolymeric intermediate filaments that lack polarity and are collectively called neurofilaments. Neurofilaments are dynamic and undergo fission, fusion and bi-directional transport within the axonal shaft [29]. Neurofilaments are a dominant feature of large calibre vertebrate axons and contribute to the maintenance of axonal calibre [30,31]. Inter neurofilament interactions may include side-chain mediated steric or entropic repulsion and ionic interactions, which may depend on the phosphorylation state, and have been proposed to promote space-filling hydrogel networks [32–35].

Axonal actin filaments are organized as sub-plasmalemmal and deep actin structures. A membrane-associated periodic skeleton (MPS) comprising of periodic, circumferential actin rings spaced at ~190 nm and connected by axially aligned spectrin tetramers have been found in all neuronal types [36,37]. It remains likely that other cortical actin networks also exist. Deep actin structures include the highly dynamic actin ‘hot spots’ and the extended ‘actin trails’ emerging from these structures [38]. In addition, there are ‘actin patches’ in the distal axon that give rise to axonal protrusions [39] and developing axons exhibit growth cone-like ‘actin waves’ that propagate from the base of the axon to the tips [40].

It is important to note that the above mentioned cytoskeletal entities behave as a composite material due to the close interactions between them. These are not only mediated by inter-polymeric cross-linking but also through the co-regulation of filament dynamics and signalling [41–44]. These co-regulatory activities often preclude the unambiguous identification of the individual contributions of one cytoskeleton system from the collective response. Nevertheless, there has been significant improvement to our understanding of the mechanical properties of the axonal cytoskeleton through the development of a wide variety of tools and techniques, some of which are summarized in the next section.

3. Tools to investigate axon mechanics

A broad variety of tools have been developed over the last several decades to investigate the mechanical properties of axons at various length and time scales. Some of these are schematically summarized in Fig. 2, and here we give a brief description of each with one or two example studies^* (also see [46]). At the whole axon level, the micro-needle method uses a fine, calibrated glass needle as a cantilever to stretch axons by displacing the base of the needle using a linear actuator. The micro-needle either applies a stretch along the axon by attaching the growth cone to the needle using an appropriate adhesive coating ((Fig. 2a) [4,47], or applies a displacement in the transverse direction (Fig. 2b) [48,49]. The axonal strain and needle deflection are measured by video microscopy. Typical force range is ~0.1 nN to ~10 nN. This is perhaps the simplest method, requiring minimal fabrication and equipment. More sophisticated force sensors like micro-machined devices (Fig. 2c) and etched optical fiber cantilevers have also been developed (Fig. 2d). Micro-machined force sensors have been used to stretch ex vivo axons from Drosophila embryos [50]. Like for micro-needles, detection of sensor deflection is via video microscopy. The etched optical fiber method uses a cylindrical cantilever fabricated by etching the tip of an optical fiber. The cantilever deflection is measured by imaging the laser light exiting the fiber on to a silicon position sensitive detector [45,51]. Apart from enabling fast (up to 200 Hz) detection of force, this method allows for the implementation of a variety of software-based feedback algorithms for strain-controlled operation. Fluid drag force generated using micro-fluidic devices is another method used to stretch axons (Fig. 2e) [52,53]. This method allows for easy introduction of cytoskeleton perturbing pharmacological agents to the whole axon or to a small axonal segment, and their washout while monitoring axonal tension or strain via video microscopy.

Fig. 2.

Simplified schematic diagrams showing the basic principle behind some of the techniques that have been used to investigate axonal mechanical responses. The hollow arrows indicate the direction of applied force. (a,b) Micro-needle method, (c) micromachined force sensor, (d) etched optical fiber technique, (e) stretching using microfluidic flow, (f) magnetic bead method, (g) Atomic Force Microscopy, (h) non-invasive probing of Brillouin scattering or axon fluctuations, (i) stretching using embedded magnetic nanoparticles, (j) axon towing using wiper, (k) fast or repeated stretching using elastomer sheet, (j) retraction response after laser ablation, and (j) membrane mechanics using tethers pulled by optical tweezers. Of these, quantitative force measurements are possible in (a–h) and in (m). Techniques (d, g) use a feedback loop to enable various strain or force controlled protocols, (e, f) can stretch axons with controlled force, (f–h) can probe local viscoelastic properties, and (i–l) do not measure viscoelastic properties but can report on axon mechanics or stability or force-induced growth. These techniques are described in more detail in the main text.

While the above-mentioned methods measure whole axon stretch responses, local viscoelastic properties have also been investigated using different techniques. One of the earliest studies used ferromagnetic beads embedded in squid giant axons and pulled using electromagnets [54]. Paramagnetic beads attached to the surface of axons using appropriate protein coatings have also been used to study local lateral extension-response of axons (Fig. 2f) [55]. The external magnetic field generates a constant force on the bead allowing for creep measurements (strain evolution at constant force). Such measurements can also be performed using a commercial Atomic Force Microscope (AFM) by using a cantilever with a rounded tip to prevent puncturing of the axon (Fig. 2g) [56,57]. AFM also allows for compression measurements, frequency response evaluation, and force or strain-controlled operation. Interpretation of magnetic bead and AFM results should be done with care as axons are extremely thin and substrate effects (during compression) or axon bending (during pulling) can influence the measured force. Magnetic bead and AFM methods typically use a force range of ~10 pN to ~1 nN.

Non-invasive methods which take advantage of spontaneous thermal fluctuations instead of an externally imposed force or strain have also been developed (Fig. 2h). Brillouin scattering method is capable of investigating mechanical properties at a subcellular scale [58]. It relies on the inelastic scattering of photons by thermal phonons and hence reports on the longitudinal storage (elastic) and loss (viscous) moduli. Its application to developing spinal tissue in vivo has been demonstrated recently [59], though it is yet to be applied to individual neurons. Lateral thermal fluctuations of the axon have also been used to deduce mechanical properties in a non-invasive manner [60,61].

Very slow stretch is known to induce axonal growth by de novo synthesis of new material and several techniques have been used to study this effect. Apart from micro-needles which can pull one axon at a time at very slow rates [4], use of embedded nanoparticles (Fig. 2i) and towing of large number of axons (Fig. 2j) have also been performed [62,63].

Sudden, large stretching of axons or repeated stretching can lead to non-elastic (plastic) deformation of specific components within the axonal cytoskeleton. Such effects are of interest to learn about axonal threshold to stretch injury. Stretchable substrates on which neurons can be cultured is one tool that has been used to study such effects (Fig. 2k) [64,65]. Axon transection experiments where a microneedle or a laser is used to cut an axon and its retraction and regeneration responses are evaluated are also methods used to study response after injury (Fig. 2l) [66,67].

At the molecular scale, fluorescence-based tension sensors can be extremely valuable in reporting forces within specific cytoskeletal elements. This is especially important considering the composite nature of the axonal cytoskeleton and the difficulty in isolating components by depolymerization without compromising the entire structure. Fluorescence Resonance Energy Transfer (FRET) has been used recently to show that spectrin molecules in axons are held under tension [67], and to show that alternating tension and compression within the spectrin network is necessary for proprioception [68].

Mechanical properties of the axonal membrane and the influence of the underlying cytoskeleton on these have also been investigated, mainly using optical tweezers (Fig. 2m) [69,70]. This is done by extracting a membrane tether or nano-tube by pulling on a micron-sized latex bead and measuring the force response during and after this procedure. Typical tether force is ~10 pN and force resolution of ~1 pN can be easily obtained. Tethers can also be dragged along the axon to gain information on membrane viscous effects, including those arising from membrane-cytoskeleton coupling [70]. Moreover, dynamical regulation of axon tension can be studied using dual-tethers, where one is used to apply a perturbation and the other measures the response [71,72].

Recent investigations conducted using these and other techniques have revealed the vast richness of axonal mechanical properties, which range from passive linear and non-linear viscoelastic responses, active contractile behavior, slow growth response and morphological transformations. In the following sections, we will discuss our current understanding of some of these axonal responses which are important not only in understanding the mechanics of axonal development but also for axonal stability under a variety of degenerative or injury causing conditions.

4. Mechanical responses of axons

As mentioned in Section 2, the axonal interior can be considered as a composite material composed of various interconnected and differently organized filament types. Apart from its anisotropic viscoelastic properties at the micro-scale, which were recognized early [54], the thin, tubular geometry of the axon too makes its mechanical response dependent on the mode of deformation. As a result of these structural and geometric anisotropies, an axon can exhibit distinct responses to radial compression, longitudinal stretching and bending deformations. Moreover, the dynamic nature of the cytoskeleton–imparted by filament turnover, binding-unbinding dynamics of crosslinkers, the action of molecular motors, and local protein synthesis–can render unique time-scale dependent properties to the axonal cytoskeleton. Such responses can be broadly classified as passive viscoelastic, active contractile, and growth responses. In this section, we discuss our current understanding of the passive and active (motor-driven or polymerization-driven) responses based mainly on in vitro experiments conducted on non-myelinated axons or neurites. Although these effects can be coupled and may act simultaneously, this separation makes it convenient to discuss various contributions to the axonal response.

4.1. Viscoelastic stretch response of axons

Among the various deformation modes mentioned above, axonal response to stretch is the most investigated. Early pioneering experiments were performed by the groups of Steven Heidemann and Dennis Bray using calibrated glass micro-needles to pull on either PC12 neurites or chick Dorsal Root Ganglia (DRG) axons in culture. These experiments revealed that axons respond as viscoelastic entities at low forces, can generate active tension, and can also exhibit tension induced growth at higher stretching forces [4,48,73].

When axons are suddenly stretched using a micro-needle by applying a step-like displacement to the base of the needle (Fig. 2a,b), the resulting excess tension on the axon and the axonal strain relaxes with time (see Fig. 3A). It was observed by Dennerll et al. that at forces less than 100 microdynes (1 nN), stretched PC12 neurites exhibited an initial step-like jump in strain (within the time resolution of their video recording) followed by a slower evolution of the strain towards a plateau value with a response time of about 10 min [48]. Corresponding to this, the tension relaxes from a maximum value immediately after the stretch to a non-zero steady-state value, known as rest tension, of about 30 microdynes (0.3 nN). When released from the needle the neurites recovered their initial length within about 1 min. This axon response in the low applied force regime was modelled as a linear spring in series with a Voigt element (a spring and a dashpot in parallel) (Fig. 3B). The former accounts for the fast elastic response and the latter for a longer-term viscoelastic solid-like behavior. Viscoelastic responses of axons have also been measured ex vivo in Drosophila embryos [50]. These axons exhibited a linear force-extension behavior when loaded continuously over a duration of 1-− 2 min, with stiffness in the range of 0.2-− 1.2 nN/μm. In such dynamic loading experiments, viscous stresses too can contribute to the measured force on the sensor. Once the loading was stopped, these axons relaxed to a steady-state tension within 15-− 30 min. Although the experiments discussed here highlighted the viscoelastic nature of the axonal cytoskeleton, many details of the axonal response are more complex than that can be captured by a simple spring and dashpot model as shown in Fig. 3B. This is because the various cytoskeletal elements are differently organized and have specific active as well as passive properties. In the next few subsections we discuss our current understanding of some of these specific contributions, starting with passive response arising out of the actin-spectrin periodic scaffolds seen ubiquitously in axons.

Fig. 3.

(A) Schematic showing axonal response to a step-stretch stretch applied by displacing the base of the micro-needle and the corresponding evolution of axonal strain for the different viscoelasticity models (B-D). The strain response for the combination shown in (B) is shown as a continuous black line, that for (C) is shown in green as a long dashed line, and that for (D) is shown in blue as a fine dotted line. The axonal tension response for model (B) is also shown to indicate that at long times axons relax back to a non-zero rest tension. (B) Spring dashpot model proposed by early studies to explain the viscoelastic response seen in the micro-needle experiments. The spring k₁ can respond instantaneously but the Voigt element (spring k₂ in parallel with the dashpot η₁) gives rise to the viscoelastic relaxation of strain and tension [48]. (C) A model used for explaining active contractile response at the whole axon level. The non-linear element M in parallel to the Voigt element accounts for the effect of molecular motors by making the axon contractile at low deformation speed [49]. (D) At long times stretched axons exhibit growth-induced elongation which, in some cases, appears like a “flow” response. A second dashpot η₂ was added to the passive case to describe the short time elastic, intermediate time viscoelastic, and longer time growth responses of axons [48].

4.1.1. Contribution of the actin-spectrin periodic scaffold to axonal stretch response

Neuronal cells express high levels of spectrins and spectrin mutations lead to abnormalities in neuronal development, the reasons for which are poorly understood [74]. A mechanical role for spectrin emerged when it was shown that when β spectrin is deleted in C. elegans, the long axons break during body movements [75]. This breakage can be prevented if the worm is anaesthetized, suggesting that mechanical strain generated by movement causes snapping. Based on this the authors speculated that pathological conditions arising due to mutations may involve compromised mechanical stability of axons. The more recent discovery of a membrane-associated, periodic arrangement of actin rings interconnected by spectrin tetramers added a new dimension to the question of the nature of its mechanical contribution [37,76]. Incorporation of Fluorescence Resonance Energy Transfer (FRET) based molecular force sensors into β-II spectrin of C. elegans has revealed that the spectrin molecules which interconnect the actin rings are held under tension in the worm axons [67]. This is further supported by measurements of axonal recoil after laser axotomy conducted in the same study. This raises the possibility that at least part of the rest tension discussed under Section 4.1 could be stored in passive form in the actin-spectrin periodic skeleton. How these forces are internally balanced in the neuronal cell is not fully understood.

A quantitative analysis of the mechanical contribution of the axonal actin-spectrin scaffold to stretch deformation was carried out recently using the optical fibre-based force apparatus (Fig. 2d) [45]. This study revealed that actin and spectrin made major contributions to axonal elastic modulus under stretch. The tension vs. extension curves showed a tension buffering response where tension remained nearly constant after an initial increase (Fig. 4A). Correspondingly, the Young’s modulus decreased with increased strain (strain softening response). This decrease could not be accounted for by considering stress relaxation by filament turnover as pharmacological stabilization of either microtubules or actin filaments did not alter the softening response. This came as a surprise as biopolymer networks, within cells and in vitro, typically stiffen (elastic modulus increases) with increasing deformation mainly due to the strong entropic contribution to their polymer elasticity and other polymer network properties [77–79]. Existing models based on springs and dashpots could not account for these responses and the non-monotonic variation in the stress relaxation time as a function of imposed strain observed in this study.

Fig. 4.

(A) Schematic showing the tension response of an axon to the application of multiple step-strains [45]. Axonal tension shows viscoelastic relaxation, reaching a ’steady state’ value in about a few minutes. This steady-state value shows an initial increase with strain and then saturates suggesting a tension buffering process (the dotted line is a guide for the eye). (B) Schematic showing how reversible unfolding of spectrin repeats under imposed stretch may act as a tension buffer in axons as proposed by ref. [45].

The experimental results mentioned above could be explained by a new theoretical model for the actin-spectrin scaffold which considered force-dependent reversible unfolding of the tandem repeat domains of spectrin molecules [45] (Fig. 4B). Indeed such unfolding events have been detected in red blood cells which also contain spectrin scaffold, albeit with a pseudo-hexagonal symmetry [80]. Taken together, the results presented in this section suggest that the axonal actin-spectrin scaffold may act as a shock absorber by allowing spectrin domains to unfold and thereby release tension when stretched to protect axons against damage. The possibility of detachments of spectrin connections to actin rings has also been considered using computational methods [57]. A more composite computational model suggests that the spectrin scaffold may protect microtubules from excess stress to some degree [81]. Indeed, recent evidence from C. elegans identifies a periodically (180 – 200 nm) distributed protein complex coupling axonal and dendritic microtubules to membrane-associated ankyrin [82]. A clear picture of how microtubules themselves respond to fast stretch deformations to the axon is yet to emerge and below we discuss some of the recent findings in this regard.

4.1.2. Contribution of microtubules to passive stretch response

Apart from showing that axons can generate active tension [48], early experiments had also shown that microtubules and actin filament disruptions have opposing effects on axonal tension response [73]. Interestingly, while disruption of actin filaments reduced axonal tension, disruption of microtubules caused an increase in tension. This could either be due to bundled microtubules acting as elements that take compressive stress or due to enhancement of actomyosin contractility triggered by microtubule disruption [43]. The passive response of the highly aligned bundle of rigid microtubules is expected to be qualitatively different from a random network of flexible actin filaments.

In the experiments of Dubey et al. only a slight drop in steady-state tension was observed when microtubules were disrupted using Nocodazole whereas actin disruption caused a drastic drop in the tension [45]. One important difference is that these measurements were done at 25 °C to suppress active processes. In such a scenario, stresses within the bundle may be released either by the stretching of microtubule-associated protein crosslinkers, like tau, or microtubule sliding may occur via crosslink unbinding-rebinding dynamics. The milder than expected effect could also be because the drug only removes dynamic microtubules and leaves behind a relatively stable fraction [13,83]. A reduction in axon stiffness was reported based on experiments where microtubules were perturbed only in short mid segments of axons using a channelled flow [53]. Microtubules have also been reported to break under fast stretch imposed using stretchable substrates (Fig. 2k) to mimic stretch injury [65]. Thus, the viscoelastic response of the cross-linked microtubules under stretch is still an open question.

4.2. Active tension generation in axons

The early experiments of Heidemann and co-workers had clearly shown that chick DRG axons can generate active tension. It was shown that axons, in general, exist in a pre-stressed state with a “rest tension” T_rest ≈ 0−2 nN [48]. By attaching the soma to a calibrated glass needle and allowing the growth cone to advance, it was also shown that growth cones directly pull on the axon [47]. In some cases, the excess axonal tension thus generated was roughly proportional to the distance moved by the growth cone, suggesting an elastic stretching of the axon by the growth cone (growth effects can also contribute to this extension). The question remained as to whether the axonal shaft also acts as a tension generating structure or not. Subsequent experiments by the same group showed that axons whose growth cones were anchored on to a glass needle (Fig. 2a) could recover their rest tension within about an hour after they were slackened by a small displacement of the needle towards the soma [48,84]. In some cases, the recovered tension was about twice the initial rest tension. In such experiments, it is unlikely that the tension is generated solely at the growth cone. More recent experiments using micro-machined force sensors (Fig. 2c) have shown that ex vivo axons of Drosophila embryo can generate contractile stresses with T_rest ≈ 1−13 nN in vivo [50]. Therefore, it seems reasonable that the axon itself should be considered as a contractile cable.

4.2.1. Simplified model for axonal contractility

The linear viscoelastic response model proposed by Heidemann and co-workers was later extended by Bernal et al. to account for active tension generation [49]. Their experiments showed that when PC12 neurites were subjected to a stretch deformation using a micro-needle, they could transition from an initial viscoelastic extension phase to an active contractile phase characterized by a negative strain rate. A new active element was added in parallel to the Voigt element to capture molecular motor activity (Fig. 3C). This active element representing the collective effect of molecular motors has a maximum tension T_a and a characteristic velocity ν₀ such that it generates a total active force $f_a=T_{a}f\left(\dot{\gamma }/{\nu }_0\right)$, where the function $f\left(\dot{\gamma }/{\nu }_0\right)$ decreases with increasing strain rate $\dot{\gamma }\left(t\right)$ and vanishes for $\dot{\gamma }/{\nu }_0>1$. This description is based on the observed force-velocity relation in single molecular motor experiments [85,86]. Solving the resulting constitutive equations reproduces the experimentally observed responses, in particular a transition from a strain relaxation phase to an active contraction phase. The active element parameters T_a and ν₀ as well as the various spring constants and energy dissipation (“viscous effect”) arising from bond-detachment could be estimated by comparing their experiments and model [49].

Application of sudden stretch to neurites in the transverse direction using a micro-needle could result in large local deformations, possibly leading to local damage or the opening of stretch-activated Ca⁺⁺ channels [87]. Calcium entry could generate calcium spikes that can then modulate the mechanical response of actomyosin cortex by phosphorylating myosin light chains or affecting the cytoskeletal structure. In order to apply a smoother deformation, experiments were later performed using a flow-chamber technique where the axon is deformed using a laminar flow perpendicular to its axis (Fig. 2e) [52]. In such an experiment, the drag force is distributed along the axon and the resulting axonal tension can be calculated from the axonal curvature. These experiments not only confirmed the passive to active transition observed in micro-needle studies but also revealed an oscillatory response in some cases.

The flow bending technique was further exploited to apply cyto-skeleton perturbing drugs to short mid segments (~30 µm) of axons instead of exposing the entire neuron [53]. Local disruption of actin filaments or inhibition of myosin contractility using a Rho-associated protein kinase inhibitor resulted in slackening of the segment exposed to the drug, suggesting a role for actin and myosin in the axonal segment. We discuss this aspect in more detail in the next subsection.

4.2.2. Contribution of actomyosin contractility to longitudinal tension

Apart from the local inhibition of myosin mentioned in the previous section, several other experiments had suggested that the axon shaft has contractile properties driven by myosin-II, with evidence coming from studies on axonal retraction or from straightening of meandering axons [66,88–90].^* Tracking of fluorescently labelled docked mitochondria during straightening of initially meandering axons suggests that the local strain rate can fluctuate between positive and negative values while the net strain rate remains negative [89]. This suggests significant spatial heterogeneity in contractility, at least in early stage chick DRG axons.

Although it is well known that actomyosin networks are contractile and spontaneously generate tension in a variety of cell types, the recent discovery of a periodic array of actin rings in axons has raised new questions. Super-resolution imaging or multiply labelled myosin-II has revealed that myosin mini-filaments bind to actin rings with their heads in registry with the rings and their mid-section (myosin light chains) positioned between rings [91,92]. Whether longitudinal contractility, as described in the earlier sections, can arise from such an arrangement or if it has its origin in the more randomly oriented actin filaments remains an open question [93].

In addition to myosin-II generated forces, the axonal shaft can also generate active stresses and flows driven by microtubule-associated motors, and these are discussed in the next section.

4.2.3. Role of microtubules in axonal force generation

Using microinjection of recombinant dynamitin to interfere with dynein function,^# Baas and co-workers have argued that axonal micro-tubules may get pushed towards the growth cone by dyneins, thereby generating an extensile stress in axons [11,88]. For this, the minus-end directed dynein motor complex may bind to the cortical actin mesh with their cargo domain and push microtubules in the anterograde direction with their motor heads (Fig. 5A). By investigating axonal stability under pharmacological perturbations to microtubules and actin cortex, they proposed that an interplay between active forces–dynein mediated extensile stress on microtubules and actomyosin based contraction–play a critical role in dictating axonal dynamics vis a vis growth and retraction [11].

Fig. 5.

(A) Schematic showing dynein molecular motors (purple) coupling the cortical actin (red) and the microtubule bundles (green) as proposed earlier [11]. The cargo domains bind to the actin and the head groups bind to the polar, aligned microtubules, generating a force which pushes the microtubule bundle to the right relative to the cortex. Arrows indicate the walking direction of the motor heads. The microtubules are assumed to be bound together by crosslinkers (yellow). (B—F) Different scenarios can be speculated if motors are able to crosslink adjacent microtubules within a bundle or between bundles. (B) No relative displacement of filaments can occur if multiple unipolar motors like dynein bind to a pair of microtubules with same polarity such that they are attached to one filament with their cargo domain, which we assume do not detach, and the other with their walking heads. (C) Relative sliding can occur if the filaments are of opposite polarity. (D) Sliding can also occur if, for example in short filaments, all motors are bound with their cargo domains on one tubule and heads on the other. (E) In the case of bipolar motors like kinesin 5, the motors can walk freely towards the plus ends if both filaments have the same polarity, and hence no relative sliding occurs. (F) If filaments are of opposite polarity, they can slide apart by the action of bipolar motors.

Tracking axonal microtubules using photoactivatable tubulin, Reinsch et al. demonstrated en masse anterograde transport throughout the axon [94]. In a few cases translocation speed (max. 117 µm/hr) approached the growth cone velocity (150 µm/hr), while in most axons it was significantly lower. Remarkably, in a few cases, translocation was seen even in the absence of axonal growth motility and microtubules seem to telescope out into the growth cone. Although translocation occurred all along the axon, in some cases the speed increased from the soma end (proximal) to the growth cone end (distal), while in other cases the speed was the same at different locations.

Subsequent experiments using docked mitochondria to track bulk cytoskeletal movement have also shown anterogradely directed bulk cytoskeletal movement in growing axons [10,95]. This translocation, as well as growth, can be prevented by blocking dynein motors but are unaffected if both myosin-II and dynein are blocked simultaneously.

Besides an overall pushing out of microtubules, bidirectional transport of short microtubules has also been observed in axons [96–99]. Within a bundle of long filaments, multiple motors of various types may engage adjacent microtubules as detailed in Fig. 5B-F. This can include dyneins and kinesins like kinesin 1 that may bind to one filament with their cargo domain and to the other with their motor heads, and possibly bipolar motors which can crosslink filaments via their motor domains. Bipolar motors like kinesin Eg5, however, may not participate in the sliding of filaments within a polar bundle [100]. Although little is known on the role of these motor types within axonal microtubule bundles, different scenarios have been explored in silico [101]. Besides this, the resistance to microtubule sliding caused by microtubule crosslinking is expected to increase with microtubule length [102].

In principle, polymerization forces can also generate stresses within bundles of microtubules, as shown by in vitro studies [103], but to our knowledge such possibilities have not been explored experimentally in the axonal context.

4.3. Stretch-induced growth of axons

Animal bodies grow dramatically after the major neural connections have been laid down and axons have extended and found their synaptic partners. Neurons have to adapt to the growth of the surrounding tissue and match the expansion [104,105]. While motor neurons innervating the lower limbs may be ~1 m long in adult humans, their synaptic connectivity is established during the 8th week of gestation when the neuron was only ~1 cm in length. This extensile growth effected by the mechanical tension from the surrounding tissue is referred to as stretch growth or towed growth.

Direct evidence for mechanical tension as a growth stimulus came from micro-needle based axon pulling experiments where the terminal segments of the neurites were lifted and subjected to pulling forces (Fig. 2a). A diversity of neuronal types, including chick sensory [4,5] and forebrain neurons [106], rat hippocampal neurons [107] and retinal ganglion neurons [108] were found to respond to applied tension with extensile growth. In experiments involving lateral displacement of neurons by micro-needles (Fig. 2b), above a threshold force (> 100 microdynes or 1 nN), chick sensory neurons [5] and PC12 neurites [48] elongated at a constant rate indicating active neurite growth (Fig. 3A,D).

Stretch growth of neurites loaded with magnetic nanoparticles subjected to a magnetic field (Fig. 2i) also support the micro-needle based studies [62,109]. These studies, using mouse hippocampal neurons and PC12 neurites loaded with magnetic nanoparticles distributed evenly along the neurites, suggest a close concurrence of stretch growth rates (~0.42–0.66 µm h^-1pN^-1) between these and older studies on chick sensory neurons using micro-needles [5,62,109].

In vivo, axonal extension due to the displacement of the neuronal soma while retaining axonal contacts has been reported during the development of the olfactory placode in zebrafish [6]. Towing of cell bodies away from the growth cones of cultured chick sensory neurons in vitro has reported similar growth responses [110].

‘Extreme’ stretch growth, up to 8 mm/day (0.33 mm/h), ~10 fold higher than typical growth cone translocation rates, with axons reaching lengths up to 10 cm in 2 weeks have been achieved experimentally [111]. Surprisingly, even at such extreme rates, these neurons did not display axonal thinning. On the contrary the calibre increased by ~35% and the axons retained normal electrical properties, organelle distribution and cytoskeleton organization [63,111,112]. Similarly, experimental leg-lengthening studies in rats [113] and rabbits [114] have reported significantly increased intermodal lengths with no accompanying changes in axon calibre, myelination and nerve conduction.

Collectively, these studies indicate an increase in mass addition in response to tension. Consistent with this, blocking protein translation by cycloheximide attenuates towed growth and implicates the involvement of de novo synthesis [62]. Axonal calibre recovery following initial thinning due to micro-needle towed growth suggests intercalated mass addition along the axonal length [110]. These authors have modelled axonal growth as ‘stretch and intercalation’ and suggest that material addition occurs at the stretched region whose extent is determined by the limited propagation of the tensile force due to axonal adhesions [115]. Plasma membrane addition has been investigated largely in the context of growth cone based extension and have reported distal membrane expansion. The clustering of plasmalemmal vesicles and insertion of glycoconjugates at the growth cone support this conclusion [116–119]. However, in light of the evidence for intercalated mass addition during stretch growth, these experiments need to be revisited. The addition of membrane and other proteins, in response to stretch, is likely to involve the secretory pathway and consistent with this expectation accumulation of endoplasmic reticulum cisternae have been reported in axons subjected to stretch growth [62]. Axonal tension is also known to modify axonal transport [120], though the mechanism behind this and the contribution of this process in the targeting and addition of new material are unknown. In addition, evidence for mTOR-dependent stretch induced local protein synthesis has been reported in a sciatic nerve model [121]. While experiments have identified several potential strategies, the relative contribution of these processes awaits systematic evaluation.

The microtubule array is necessary for axon assembly and both microtubule assembly at the growth cone and bulk anterograde trans-location of the microtubule array have been proposed for growth cone based axonal growth [1,10]. In fact, in PC12 neurites microtubule stabilization with paclitaxel eliminated the growth response to applied stretch [48]. Under conditions of ‘extreme’ stretch, rat DRG axons extending at 0.167 mm/h for a period of 14 days, the axonal ultra-structure revealed no change in microtubule density [111]. However, the stretch growth of mouse hippocampal neurons at 6.00 ± 0.15 µm/h for 2 days of magnetic towing resulted in an increase in MT density along the entire axonal shaft in comparison to axons growing without magnetic field based stretch [62]. In these experiments, stretch-growth was found to be sensitive to microtubule depolymerization by nocodazole at doses at which normal tip growth was unaffected. Low doses of paclitaxel affected both tip and stretch growth negatively in this paradigm. The differential effect of nocodazole on stretch-growth highlights the requirement of microtubule assembly for stretch-growth. It is interesting to note that both the ~1 nN towing forces generated by the growth cone and the ~10 pN forces in the magnetic stretch experiments influence microtubule assembly, though it is unclear if the underlying mechanisms are similar.

What senses axonal tension remains elusive. Mechanosensitive ion channels are possible candidates and the role of Piezo channels in axonal growth is well established [122]. While mechanosensitive channels may gate multiple ions, calcium is an attractive possibility as it is known to trigger multiple signalling pathways and remodel the cytoskeleton. However, stretch-growth by magnetic towing revealed attenuated calcium transients [62] consistent with the decreased frequency of calcium spikes observed in fast growing axons [123]. Stretch activated calcium could still play a significant role in initiating the growth response by signalling to the nucleus and triggering transcriptional changes. The stretch-growth of neurons has indeed been associated with transcriptional changes [121,124]. The calcium dynamics is likely to be complex and can be modified by the activity of other stretch activated channels, like stretch activated like potassium channels [68], and therefore await precisely controlled experimentation.

Microtubules may also act as tension sensors in neurons and respond by enhanced growth. Indeed, the tension on the growing end of micro-tubules has been shown to increase the growth rate of microtubules in vitro [125,126]. In the latter studies, low forces of ~1 pN per micro-tubule can significantly increase microtubule assembly and are consistent with low towing forces of ~10 pN per axon, inducing microtubule assembly during stretch-growth of axons. Similarly, the polymerization of actin filaments in vitro is also sensitive to tension [127,128] and may serve to sense axonal stretch. The spectrin MPS has been suggested to respond to sudden stretch by the unfolding of spectrin repeats [45]. It is unclear if continuous loading with weaker forces resulting in stretch-growth can also induce spectrin repeat unfolding but such processes could serve as tension sensitive signal initiators by the force dependent exposure of cryptic binding sites.

Having discussed the viscoelastic stretch responses, active tension generation in axons driven by myosin-II, microtubule based force generation and stretch-induced growth response, we now turn our attention to the role mechanical forces in maintaining axonal calibre and how these affect axonal stability.

4.4. Radial mechanics, axonal calibre and shape stability

As mentioned earlier, the highly anisotropic structural organization of the axonal ultrastructure, together with the thin tubular geometry makes the elastic and contractile responses in the radial direction very different from those along the axonal axis. Probing radial elasticity is much harder owing to the thinness of the axon and the difficulties in decoupling the two directions when local properties are probed. Nevertheless, a few studies have been conducted mainly using magnetic beads attached to the surface or using atomic force microscopy. Such investigations are highly relevant to understanding the regulation of axonal calibre and axonal radial shape stability as will be discussed in the next few subsections.

4.4.1. Mechanical response to radial deformations

Atomic force microscopy studies (Fig. 2g) on axons treated with pharmacological agents to disrupt microtubules, neurofilaments and actin filaments showed that microtubules made the most significant contribution to lateral compression modulus followed by neurofilaments and then the cortical actin [129]. The elastic modulus measured in this study, which was of the order of 9 kPa, using Hertz contact model was much higher than that reported by other AFM experiments [56,130]. Subsequent measurements of rate-dependent compression moduli for axons using AFM show that the elastic modulus decreases from 800 Pa at short times (0.1 s) to 80 Pa at long times. In addition, this study suggests the existence of distinct relaxation times [56]. Using data from axons treated with cytoskeleton perturbing drugs and using finite element modelling, the authors claim that these distinct responses arise from the thermal fluctuations of microtubules, membrane relaxation and cytosol viscosity. A recent study has also investigated compression induced degeneration of axons and has shown that mechanical susceptibility differs significantly for DRGs and hippocampal neurons [131].

Rheological measurements performed using magnetic tweezers to pull on paramagnetic beads attached to axons (Fig. 2f) revealed a predominantly viscous-like response for neurites whereas the soma showed more solid-like behavior, possibly arising from the contribution from the nucleus [55]. Moreover, while the soma exhibited a stress-stiffening response, the neurite fluidized with applied stress (softening). The mean Young’s modulus, however, was higher for the neurite (7 kPa) as compared to the soma. As in the AFM studies mentioned above, microtubules along with dynein activity and neurofilaments were found to be the principal load bearing elements in neurites under radial deformation and disruption of actin filaments did not have any significant effect. In addition, it was observed that the neurite became more viscous-like when grown on softer substrates [55].

The studies mentioned above highlight the anisotropic nature of neurite mechanical properties. The contribution of microtubules to lateral deformations is perhaps not surprising considering that they are relatively stiff polymers, heavily crosslinked by microtubule-associated proteins, and connected to the rest of the cytoskeletal elements. Neurofilaments are expected to contribute significantly under radial compression due to their lateral side-arm interactions and are known to be important for setting axon calibre. Next, we’ll discuss how these filaments along with the axonal membrane participate in stabilizing axonal tubular geometry.

4.4.2. Mechanical regulation of axonal calibre

Axonal calibre varies widely across species and neuron types, typically in the range of 0.1–10 µm, and reaches ~1 mm in the case of the giant squid axon. Maintenance of an optimum calibre is essential for faithful propagation of the action potential and may also involve energetic considerations [132,133]. Active regulation of osmotic pressure, regulation of synthesis of cytoskeletal components, regulation of phosphorylation state of neurofilaments, etc are potentially involved in determining axonal thickness. In a given neuronal cell in vitro, the axonal diameter remains roughly constant all along its length and how the radius is set and the uniformity is maintained are also not fully understood. While the regulatory mechanisms may be complex, some insights can be gained regarding the mechanical forces that compete against each other in maintaining axonal calibre.

The axonal membrane is under tension and this tension has been measured using optical tweezers [70,134]. It is known that a synthetic lipid membrane tube under tension σ attains an equilibrium radius R₀ which is dictated by its bending modulus B and the membrane tension as given by $R_0=\sqrt{B/2\sigma }$. Using measured values for membrane modulus and tension, this radius is ~0.1 µm, much smaller than a typical axonal diameter. Indeed, bare membrane tethers pulled out of axons attain this narrow diameter after tension is allowed to reach steady state [70,134]. This suggests that the axonal caliber is set by a balance of membrane forces and cytoskeletal forces which may include osmotic as well as elastic contributions [135]. In particular, the steric and elastic contributions from the crosslinked microtubule core [136–138], the entropic pressure arising out of neurofilaments [32,139], neurofilament phosphorylation state dependent repulsion of the side arms [140], and the mechanics of the periodically spaced actin rings [91] are all expected to balance the membrane tension to attain a stable diameter for the axon. Moreover, just as actomyosin contractility leads to longitudinal tension generation in axons, there is recent evidence that it may also generate circumferential tension and affect axon calibre [138]. The radial contractility could also have its origin in the periodically spaced actin rings although the mechanisms are yet to be deciphered [93]. Further evidence for the interplay between membrane tension and elastic forces in dictating axonal calibre or loss of it comes from the observation of shape instabilities in axons as discussed in the next subsection.

Several studies on neurons from a wide range of species have revealed that axonal calibre increases during the propagation of action potential [141–143]; reviewed in [144,145]. Though some possible mechanisms have been discussed in literature, the nature of electro-mechanical coupling is an interesting area for further research.

4.4.3. Shape instabilities of axons

The normal cylindrical geometry of the axon discussed above can be lost under a variety of abnormal conditions. One of the most common morphological transformation is the appearance of a series of swellings all along the axon, commonly referred to as axonal beading or axonal spheroids [14]. Axonal beading can occur as a result of sudden stretch to nerve fibers [146], mechanical perturbations to axons in vitro [147], or as a result of a variety of neurodegenerative conditions like Alzheimer’s and Parkinson diseases [148–150], ageing [151,152], multiple sclerosis [153], ischemia [154], etc. In certain cases, like in stretched nerve fibers and in beading induced by sudden osmotic swelling, biophysical analysis shows that excess membrane tension is the main driving force for the appearance of periodic swellings and this is resisted by the cyto-skeleton [155,156]. This is because the membrane area and hence interfacial energy of a tense membrane enclosing a fixed volume is reduced by a peristaltic shape as compared to a uniform cylinder. At high enough membrane tension, this leads to a greater gain in membrane energy as compared to the energy cost for deforming the cyto-skeleton (see appendix in ref. [13] for a simplified explanation). A recent investigation using pharmacological disruption of cytoskeletal elements, especially microtubules, suggest that even under normal membrane conditions, disruption of the cytoskeleton can destabilize axons causing beading [13]. It is suggested that competition between membrane tension and bulk elasticity dictates the stable axonal shape with a reduction in the elastic modulus leading to the appearance of beaded shapes. It was also shown in this study that the axonal swellings propagate retrogradely, leading to axonal atrophy and that the propagation direction may be set by gradients in microtubule stability close to the growth cone.

Thus, mechanical balance between membrane tension and bulk elastic forces arising from the cytoskeletal elements play a crucial role in maintaining axonal calibre and shape stability. However, how axonal membrane tension is regulated either during growth or in mature axons is only beginning to be understood. Several quantitative investigations suggest that the nature of axonal membrane mechanics is distinct from other cell types, and we summarize some of these unique features in the next section.

5. Axonal membrane mechanics and tension homeostasis

Membrane tension influences multiple cellular processes, including vesicle fusion, cellular migration, long-range coordinating of cytoskeleton remodelling and mechanotransduction. Membrane flow is expected to result in millisecond scale transduction of mechanical tension across the cell to coordinate long distance signalling. However, recent work in non-motile cells shows that local perturbations of membrane tension remain static over many minutes and fail to show long distance propagation [72]. In these cells, membrane cortex attachments are thought to resist flow and prevent the propagation of membrane tension.

However, the axonal plasma membrane displays distinctive properties that merit attention. Several studies have established the criticality of membrane addition for growth cone based axonal elongation [157]. Axons exhibit retrograde flow of the axonal membrane driven by a membrane tension gradient between the neuronal cell body and the growth cone, as revealed by tracking of membrane attached beads and optical tweezer based measurements [69]. Further, in the axons of chick DRG neurons and rat hippocampal neurons, membrane tether attachment points slide with relative ease along the axonal length suggesting a weaker and differently organised coupling between the membrane and the cortex compared to other cell types which were studied [70,158]. This is consistent with significantly lower peak forces necessary to pull membrane tethers from Xenopus retinal ganglion cell neurons compared to fibroblasts [159].

A recent study has directly tested membrane tension propagation in rat hippocampal neurites using a dual tether assay, where one tether is used to apply a perturbation and the tension change in the other situated at a distance is monitored. Contrary to immobile non-neuronal cells, rapid tension propagation over significant distance is observed [158]. Interestingly, tension propagation was significantly higher in axons compared to dendrites. The latter observation corresponds with a higher density of membrane cortex attachments in dendrites as inferred from reduced diffusivity of GPI-anchored proteins. The same study used tether-induced pearling instability in axons to estimate the speed of membrane tension propagation to ~20 µm/s 1000 fold more than propagation velocities estimated in non-motile HeLa cells [72].

Axons have a membrane-associated periodic skeleton, which may provide a reduced surface area for membrane coupling and result in weaker membrane cortex attachments than what is observed in nonneuronal cells. This may explain some of the above mentioned measured differences in membrane-cytoskeleton coupling and tension propagation speeds.

These findings highlight the possibility of axonal membrane tension propagation as a mediator of axonal growth and function. Ultrafast coupling of vesicle fusion and compensatory endocytosis at synaptic terminals have been associated with membrane tension [160] and the rapid propagation of membrane tension in axon terminals have been associated with rapid synaptic turnover [71]. Local relaxation of membrane tension has also been associated with increased axonal branching [158].

The site of axonal membrane addition has been traditionally thought to be restricted to distal regions with several studies reporting clustering and fusion of plasmalemmal precursor vesicles at the growth cone [157, 161]. However, experiments subjecting neurons to hyperosmotic shock have reported a 20% increase in the area within a minute followed by a uniform recovery to the original area and volume [156]. As cellular membranes are expected to rupture beyond 3-−5% stretch, these experiments suggest buffering of excess tension by intercalated addition of membrane lipids throughout the axonal length. Whether axons have a membrane reservoir capable of rapidly increasing the membrane area in response to tension remains unclear. In non-neuronal cells, membrane ruffles, invaginations and mechanosensitive caveolae rapidly flatten to buffer a sudden increase in membrane tension [162]. Such mechanisms are yet to be explored in the case of neurons.

6. Conclusions and future directions

In this review, we have summarized some of the mechanical and dynamical processes within the axon that impinge on axonal development, stability and function. These results along with several other lines of investigation have resulted in a paradigm shift in how we view the axonal cytoskeleton and membrane. The discovery of new ultrastructural features like the membrane-associated periodic skeleton and the demonstration of perturbations to this structure leading to axonal fragility in vivo raises the question as to how exactly this structure influences axonal stability. There is now ample evidence that the actin based cytoskeleton involves myosin-II driven contractile responses, but the microscopic details of how myosins act on the various actin structures within the axon need to be established. Although the polar, aligned organization of microtubules within the axons is relatively simple, there is growing evidence that at least a significant fraction exhibit polymerization and decay dynamics and are acted upon by molecular motors to generate active stresses. These active processes may lead to a dynamic mechanical balance that are critical for axonal stability. This may explain why overstabilization, as well as destabilization of microtubules adversely, affect axonal functioning. Although we have not paid much attention to neurofilaments in this review, they too are known to be highly dynamic and how this filament system affect axonal stability is not fully understood. Another less looked-into component is the axonal membrane and recent experiments show that membrane tension too influences axonal morphology and it is likely that this tension is maintained via feedback mechanisms that may involve tension dependent exocytosis and exocytosis. Finally, coming to the long timescale stretch growth of axons, although many details of this process is now well established, not much is known about how mechanical tension is sensed and how it is translated to regulated mass addition.

While the starkly different organization of the different axonal cytoskeletal components and the differences in the microscopic mechanisms by which they contribute to axonal mechanics makes these investigations fascinating, the interdependencies between the structures make pharmacological or genetic deletion of specific components difficult to interpret. However, the diversity in the organisation of the axonal cytoskeleton can be exploited to test the specific contributions of certain elements. The lack of intermediate filaments in arthropods [163] may facilitate understanding the contribution of neurofilaments to neurons. Similarly, the distribution of intermediate filaments and microtubules varies within neuron types [19] and can inform on the effect of micro-tubule bundling. Along the same lines, neurons seem to lack the periodic actin-spectrin scaffold early in development [89,164] and can be contrasted with mature neurons.

Co-evolution of theoretical models along with experiments is another strategy that has made much progress and need to be pursued further. In this review we have focused mainly on experiments and we refer the reader to a recent review on theoretical developments [165].

Neurodegeneration is one of the vexing problems of modern health care as longevity is increasing at a rapid pace due to our success in treating various other critical illnesses. This situation demands a closer look at how the dynamically regulated mechanical balance within the axon get affected under such conditions. This may provide alternative strategies to prevent axonal atrophy or loss of function and/or may inform on relevant upstream events that lead to neurodegeneration. Another strong motivation comes from the fact that axons, being extremely thin and long, are highly impacted by sudden stretch during normal functioning or under injury-causing conditions. How the ultrastructural organization of the axons consisting of elements with different mechanical properties protects axons against damage is also a wide-open area for further research.