Neuronal electrical activity is a cascade of biochemical events driven by the passive influx and efflux of ions through voltage-gated channels (Hodgkin and Huxley, 1952). Action potentials are accompanied also by mechanical events that occur in the immediate time scale. In the immediate temporal scale, nano-scale deformations are accompanying the action potential (Iwasa et al., 1980) and heat is released (Tasaki et al., 1989). These mechanical events are minute, but they cannot be explained solely by an electro-chemical view of action potentials. The exact molecular mechanism that leads to this mechanical wave is yet unclear. A physical model of propagation of mechanical solitons was proposed by Heimburg and Jackson (2005). They describe how an action potential may lead to a phase transition of the lipid membrane, that reflects non-linear changes in multiple mechanical and thermodynamic properties (Heimburg and Jackson, 2005), including those described by Iwasa et al. (1980) and Tasaki et al. (1989). This phase transition propagates similar to an action potential, without being dispersed (soliton). Mussel and Schneider (2019) demonstrated theoretically that a propagating acoustic wave in the interface of a lipid membrane and the electrolytic extracellular medium can yield a pulse of electric potential that behaves similar to an action potential. The molecular mechanisms leading to these events are yet to be resolved, and so is their physiological role.

Electrical activity was shown to trigger cytoskeletal changes occurring in the long time scales. Alvarez (1979) demonstrated that a 2 h stimulation of the preganglionic cervical nerve in cats resulted in an increase in the density of microtubules. Hu et al. (2008) imaged cultured hippocampal neurons to show that increasing neuronal activity by repeated perfusion with high KCl medium, enhanced the invasion of microtubules into spines and lengthened the duration of these invasions. Mitsuyama et al. (2008) complemented this result with a similar model of acute hippocampal slices and showed that strong stimulation leads to elongation of microtubules into spines, thus enlarging their protrusions. Merriam et al. (2013) further established the role of Calcium, F-Actin, and Drebrin in regulating the growth of microtubules in spines in similar experiments.

In a set of separate experiments, it was shown that intense electrical activity may lead to beading of axons. In non-living physical systems, as fluids enclosed by a cylindrical membrane, beading is typically linked to a change in the mechanical forces acting on the cylinder. A stable cylindrical system reflects an equilibrium between the stiffness of the cylinder, its surface tension, and the pressure and volume of the enclosed fluid. Once the pressure increases, for example by fluid inflow, minimization of free energy favors an increase in the volume. Given a fixed surface area, volume can be maximized by deforming the cylinder into spheres (thereby the term “beading”). This can be facilitated by reduction of the stiffness of the cylinder (Clerk Maxwell, 1874). Similarly, in neurons, different conditions and manipulations lead to a dramatic segmentation of the neuronal projections into swollen sphere-like segments (beads) separated by long shrunken “necks” (Ochs et al., 1994). Beading was observed in a variety of neuronal pathologies and conditions related to cellular stress like ischemia (Murphy et al., 2008; Risher et al., 2010), spreading depression (Takano et al., 2007), and seizures (Zeng et al., 2007).

To complete this short review of activity related morphological changes, we should mention that, in even longer timescales, neuronal plasticity facilitates morphological changes related to neuronal growth and the formation of networks (Shefi et al., 2005).

All the above mentioned events are necessarily linked to cytoskeletal changes. Currently there is no model that links the mechanical events that occur in different spatio-temporal scales. A model of the relation between cytoskeletal dynamics and the cellular events should include the transition from short and less intense physiological activity to pathological stimulation, the role of different cytoskeletal components linked to the above mentioned dynamics, and the way these affect different cellular structures, as axons, dendrites, and their growth cones.

We hypothesize that electrical activity affects the cytoskeleton immediately. We assume that activity triggers cytoskeletal de-polymerization such that this effect serves a physiological role in reducing the neuronal stiffness thus facilitating the morphological changes that occur later. We present an in vivo experiment that is composed of electron-microscopy imaging and analysis of cytoskeletal changes in axons after electrical stimulation.

Representative micrographs of specific axons from the 100 Hz excitation (Figure 5A) and the 10 Hz (Figure 5B) excitation exhibit loss of densities of axial neurofilaments and microtubules from the lumen compared to the control non-excited group (Figure 5C).

A significant drop in the density of neurofilaments (Figure 6A) and microtubules (Figure 6B) was identified in the axon stimulated groups compared to the controls. For example, 10 Hz stimulations with durations of 1 and 2 min, resulted in neurofilaments density that dropped to 55.8 and 51.1% of the baseline median values, respectively. In the same experiments, microtubules density dropped to 23.7 and 38.5% of the baseline median values, respectively. Similar changes appear also in the 5 Hz stimulation group. However, results of the 100 Hz stimulation group had a much stronger variance, and the cytoskeletal densities dropped to more moderate values. The changes resulting due to stimulations with 10 and 5 Hz were statistically significant for both neurofilaments (P = 0.012 for the 10 Hz–2 min group, compared to the control) and for the microtubules (P = 0.003, P = 0.001, P = 0.002 for the 10 Hz–2 min, 10 Hz–1 min, and 5 Hz–1 min groups, respectively). Changes in the distances to the nearest neighbors, compared to controls were also statistically significant (P = 0.00014, P = 0.00027, P = 7⋅10^–7, for the 10 Hz–2 min, 10 Hz–1 min, and 5 Hz–1 min groups, respectively).

Furthermore, the average distance of both microtubules and neurofilaments from their nearest neighbors significantly increased in the stimulated axons compared to the control (Figure 6C), indicating that these changes are local, throughout the entire axonal sections, rather than the outcomes of specific regions that bias the calculated averages. Interestingly, the control group and the group excited with a 100 Hz stimulation display the highest variability in the densities of the microtubules and neurofilaments and the distance to the nearest neighbor. No statistical significance was found for the comparison of the 100 Hz stimulation group with the control group.

Quantifying the average magnitude of the edges in the micrographs (in between cytoskeletal objects and the lumen) links the electric activity to the presence of electron-dense molecules, such as cytoskeletal debris in the lumen. Indeed, the sharpness of the edges is attenuated in the micrographs of the excited nerves (Figure 6D). The highest variability within axons of the same group was documented in the control group and the group of axons stimulated with 100 Hz. The substantial variability in these two groups is also reflected in the correlation between the densities of microtubules and neurofilaments within each axon, with more “outlayers” in the control and the 100 Hz groups (Figure 7) relative to the other stimulated groups. It should be noted that additional changes appear in the micrographs, especially in these of axons stimulated with a 100 Hz stimulation, on top of the quantified cytoskeletal changes. These include loss of mitochondria, loss of the integrity of the myelin sheath and intrusions appearing in the myelin sheath. Occurrence of changes in the benign stimulation frequencies (5 and 10 Hz) should be further studies.

We show with this in vivo model, that axon stimulation results in immediate rearrangements of the cytoskeleton, reducing the density of the cytoskeletal mesh. We suggest that the changes we identify in the density of neurofilaments and microtubules along the axons (Figure 6) result from activating a defined, highly regulated cytoskeleton cleavage process. We further suggest that this cleavage process is linked in a way that has never been related before to electrical activity. Action potentials are accompanied by an elevation in free calcium levels, both in the nerve cells and in their projections (Callewaert et al., 1996; Jia et al., 2011; Dana et al., 2019). It was further demonstrated in a model of neural axotomy that an elevation in free calcium levels induces activation of the Calpain protease that depolymerizes cytoskeletal proteins (Kamber et al., 2009). Moreover, depolymerization is pivotal in the creation of functional growth cones, following axotomy (Ziv and Spira, 1997). We suggest that a similar mechanism operates after electrical activity. Notice that Ca⁺² concentration gradient across the cellular membrane exceeds these of other ions, since extracellular calcium concentration is ×10⁴ higher than that of the intracellular fluid (Jones and Keep, 1988). Once voltage gated channels open, the inflow of Ca⁺² is very significant, and calcium transients can be detected even after a single action potential (Callewaert et al., 1996; Dana et al., 2019). Indeed, Hanemaaijer et al. (2020) demonstrated, with a model of pyramidal neurons, that single action potentials generate an immediate elevation in the calcium concentrations to near μM levels. The threshold for activation of μ-Calpain is Ca²⁺ concentration of 3–50 μM (Ma, 2013). Thus, these concentrations can be reached momentarily within a single train of action potentials and may trigger a proposed cascade of proteolytic activity – probably initiated in the cellular cortex – in a sub-second timescale. We suggest that in cases of repeated activity this depolymerization reduces the axonal stiffness. This initially results in axonal swelling and later on in beading. In even longer time-scales this activity-induced depolymerization facilitates a reduction of the stiffness that is a pre-requisite for growth. Our observation may thus reflect the first stage that can be followed later by axonal growth plasticity.

We observed changes in the density of microtubules that are opposite to some of the earlier results (Alvarez, 1979; Hu et al., 2008; Merriam et al., 2013). It is possible that the differences stem from differences in the experimental models used. However, it is also possible that we detected the initial outcome of the stimulated activity, whereas some of the earlier observations are related to the late dynamics of the cytoskeleton. Given this interpretation, we assume that the process observed here is a reversible physiological process. In low level stimulation it facilitates growth. In more intense stimulation that involves an increased intracellular pressure, it facilitates volumetric extension of the axon, thereby preventing rupture of its membrane. In either case, in vital cells it should be followed by re-polymerization.

Note that the mechanical response described here is linked to “trains” of multiple spikes and not to single action potentials. In addition, the reversible effect of repolymerization most likely occurs in much longer timescales of minutes to hours. Thus, the process described and suggested here may be related to the observations of a propagating mechanical wave (Iwasa et al., 1980; Heimburg and Jackson, 2005; Mussel and Schneider, 2019), but is different from these observations.

As noted, our results demonstrate that the most substantial effect occurs in the 5/10 Hz−1 min stimulation groups. We speculate that the relative elevation in density in the 100 Hz stimulations may be linked to the beading of axons that creates shrunken segments with densely packed cytoskeletal proteins (Ochs et al., 1994). The elevation in the variability of the 100 Hz stimulation group also supports this, as the imaged axonal sections may include swollen or shrunken segments.

Additional experiments that will involve long term tracking of the cytoskeletal dynamics should further test the proposed mechanism of depolymerization by Ca⁺² and Calpain and the proposed morphological kinetics that occur in different time scales.

Finally, we should highlight also the work of Lin and Cantiello (1993) and of Priel et al. (2006) that observed that cytoskeletal elements (actin and microtubules, respectively) act as “transmission lines” or non-linear transistors that amplify the conducted electrical signal in the neuron. If our view of the cytoskeletal dynamics is correct there is a possible link between the de-polymerization and the adaptive nature of neuronal activity. This proposed link between Ca⁺² triggered depolymerization and the results of Lin and Cantiello (1993) and of Priel et al. (2006) conforms with the observations that increase in Ca⁺² level reduces firing frequency, and reduces the propagation velocity of action potentials (Lüscher et al., 1996).

To summarize, in this work we have demonstrated the immediate cytoskeletal changes that accompany neuronal excitation for the first time. Based on EM analysis, we suggest that electrical activity induced depolymerization of microtubules and neurofilaments. Awareness of this phenomenon may help expand our understanding of the biomechanical and chemical events during neuronal excitation. Our work may also have interesting neurophysiological implications on the understanding of the relationship between neuronal function and morphology as means of regulating adaptive neuronal activity, learning, and memory.

The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author.

The animal study was reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) of Tel Aviv University.

UN initiated the study. UN, BY, EP, EO, and AS designed the experiments. BY, TG, EO, TB-K, and AS performed the experiments. BY, YA, ND, TB-K, UN, and EO performed the analysis. UN, BY, TB-K, and EP wrote the manuscript. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.