Spinal control of motor outputs by intrinsic and externally induced electric field potentials

Abstract

Despite numerous studies on spinal neuronal systems, several issues regarding their role in motor behavior remain unresolved. One of these issues is how electric fields associated with the activity of spinal neurons influence the operation of spinal neuronal networks and how effects of these field potentials are combined with other means of modulating neuronal activity. Another closely related issue is how external electric field potentials affect spinal neurons and how they can be used for therapeutic purposes such as pain relief or recovery of motor functions by transspinal direct current stimulation. Nevertheless, progress in our understanding of the spinal effects of electric fields and their mechanisms has been made over the last years, and the aim of the present review is to summarize the recent findings in this field.

different aspects of spinal control of motor output have been analyzed to different extents, and the present review deals with only one of these. The integrative functions of spinal neurons have been examined most extensively. The fact that spinal neurons do not operate as simple relays but integrate information from a number of sources was first demonstrated with respect to the motoneurons, recognized as the final common path of spinal output (Sherrington 1906). Studies on the integrative properties of spinal interneurons started with premotor interneurons, i.e., interneurons providing direct input to motoneurons (see e.g., Jankowska 1992, 2012). However, these studies soon extended to an increasing number of polysynaptic pathways, especially those involved in voluntary movements and locomotion (Baldissera et al. 1981; Burke 1999, 2007; El Manira 2014; Grillner 1996, 2003; Hultborn 2006; Kiehn 2016; Kiehn and Dougherty 2012; Lundberg 1964, 1975, 1982; McCrea 1992, 2001; McCrea and Rybak 2008; Roberts et al. 2000, 2010; Rossignol et al. 2008; Stuart and Hultborn 2008). All of these interneurons were found to operate as the final common path of complex premotoneuronal input from either peripheral nerves or supraspinal neurons, or both.

Interrelations between supraspinal and spinal neuronal systems became another extensively investigated issue, both introduced and most systematically explored by Lundberg and colleagues (Lundberg 1967, 1975, 1982). Lundberg’s group demonstrated that supraspinal neuronal systems strongly influence spinal reflex responses, whether by their direct actions on motoneurons or on various interneuronal networks. The direct coupling between the corticospinal or other descending tract neurons and motoneurons has turned out to be decisive for only a small fraction of motor commands, even in primates (Lemon 2008; Lemon et al. 2004; Phillips and Porter 1977; Porter and Lemon 1993), showing that spinal interneuronal relays are of a major and not only accessory importance. In addition, interneurons were found both to filter and to adjust any information reaching them, rather than acting only as passive relays of supraspinal motor commands. Among the most extensively investigated relays of these commands are the cervical propriospinal neurons mediating target-directed reaching movements (Alstermark and Isa 2002, 2012; Lundberg 1979, 1999), last-order inhibitory interneurons mediating reciprocal inhibition between antagonist muscles (Jankowska 1992, 2012), and neuronal networks mediating centrally initiated locomotion (El Manira 2014; Grillner et al. 2008; Kiehn 2016; Roberts et al. 2010) or scratching (Berkowitz 2008; Guzulaitis et al. 2014). The organization of less directly coupled supraspinal neuronal systems is currently under investigation (Grillner et al. 2005; Grillner and Robertson 2015).

To ensure that interneuronally relayed reactions are task and context specific, only some interneurons within each of their populations are activated at one time, with several modulatory systems involved in the selection of the most appropriate interneurons in both vertebrates and invertebrates (see e.g., Berkowitz 2008; Berkowitz et al. 2006, 2010; Brownstone et al. 2015; Cropper et al. 2014; Hammar and Jankowska 2003; Hammar et al. 2007; Harris-Warrick 2011; Jing and Weiss 2002; Katz and Harris-Warrick 1999; Rossignol et al. 2008; Sillar et al. 2014). In extreme cases, the task-specific selection of interneuronal subpopulations may, in fact, lead not only to the adjustments of their operation but also to the reversal of reflex actions that they mediate, e.g., flexion instead of extension of a limb (Forssberg et al. 1977) or body movements in opposite directions (Hsu et al. 2017). Among the mechanisms of interneuronal selection, particular attention has been paid to the role of neuromodulators, especially monoamines, which are involved in adjusting the excitability of spinal neurons and thus modifying the operation of their networks (Aggelopoulos et al. 1996; García-Ramírez et al. 2014; Hammar et al. 2004; Hammar and Jankowska 2003; Heckman et al. 2009; Johnson and Heckman 2014; Lundberg 1975, 1982; Noga et al. 2009, 2011). Another focus has been on changes in the way neurons respond, e.g., changes in membrane properties that allow them to respond tonically rather than phasically (Hounsgaard et al. 1988; Hounsgaard and Kjaerulff 1992; Hultborn 2002) or to respond only during certain phases of rhythmic movements.(El Manira 2014; El Manira and Wallén 2000; Grillner 2006; Hultborn and Nielsen 2007; Kiehn 2016; Kiehn and Dougherty 2012; McCrea 2001; McCrea and Rybak 2007, 2008).

One aspect of spinal control of motor output, however, has remained largely unexplored: control by electric field potentials that accompany spinal activity and are thus intricately linked to all of the above-outlined modulatory systems. This issue has been explored primarily with respect to effects of electric fields on cortical or brain stem neurons (Anastassiou and Koch 2015; Anastassiou et al. 2011; Faber and Korn 1989; Weiss and Faber 2010) and only in a preliminary way on spinal neurons. However, interest in the control of spinal activity by electric fields, both intrinsic and externally induced, has been recently revived by the introduction of transcranial (tDCS) and transspinal (tsDCS) techniques of applying direct current (DC) in clinical practice. Both of these techniques involve passing DC transcutaneously, via an electrode in contact with the skull, or the back, against a remote reference electrode, i.e., in a way noninvasive, safe, and easy to apply. Both tDCS and tsDCS have been found to be beneficial in a variety of situations, improving motor performance as well as general brain functions (see e.g., Ahmed 2013a, 2014b; Bikson et al. 2013; Jackson et al. 2016; Lefaucheur et al. 2017; Nitsche et al. 2008; Nitsche and Paulus 2000; Paulus 2003; 2011; Priori et al. 1998, 2014). Considerable progress already has been made in the analysis of factors that contribute to the effects of DC, such as the mode of application of DC, its intensity, or its timing. With respect to the mechanisms underlying effects of electric fields on the other hand, there has been a consensus that knowledge remains very limited. For example, according to Rahman et al. (2013), “. . . which cellular compartments (somas, dendrites, axons and their terminals) mediate changes in cortical excitability remains unaddressed,” or according to Anastassiou and Koch (2015), “. . . how the effects of endogenous and externally imposed electric fields . . . are mediated and manifested in the brain remains a mystery.” In addition, significantly more attention has been paid to the mechanisms of the short-term or acute effects of the current than of the long-lasting postpolarization effects, especially within the spinal cord.

The aim of this review, therefore, is to summarize progress in the studies of spinal effects of electric fields. The first part of the review focuses on the more extensively investigated effects of externally applied electric fields, and the second on effects of intrinsic field potentials that accompany neuronal activity, in both cases addressing the issues of their mechanisms as well as consequences for spinal control of motor functions.

Effects of Externally Applied Electric Fields

The effects of externally applied DC have been analyzed in a variety of preparations in humans and animals, in vivo and in vitro, on cells and fibers in different parts of the nervous system, and using different experimental approaches. In previous reviews, the theoretical aspects of DC effects have been most extensively dealt with by Faber and colleagues (Faber and Korn 1989; Weiss and Faber 2010) and the therapeutic effects by Nitsche and Paulus and colleagues (see e.g., Lefaucheur et al. 2017; Nitsche et al. 2008). Only issues most relevant to the spinal effects of externally applied DC will therefore be addressed.

Polarity and preferential targets of external electric fields.

effects of transspinal polarization.

When DC was applied via two electrodes in contact with the surface of the feline spinal cord, as illustrated in Fig. 1 with a diagram from Eccles et al. (1962), anodal polarization applied via a dorsally positioned electrode hyperpolarized group Ia muscle sensory fibers traversing the spinal gray matter in the dorsal horn and depolarized them in the ventral horn, whereas cathodal polarization had opposite effects, as indicated by plus and minus signs, respectively. The hyperpolarization was demonstrated by decreased excitability of these fibers to intraspinally applied stimuli and, therefore, a smaller number of these fibers and smaller amplitudes of nerve volleys induced by intraspinal stimuli in the peripheral muscle nerves. Conversely, the depolarization increased the excitability of the stimulated fibers and the nerve volleys in peripheral nerves. As expected, action potentials evoked in the analyzed fibers were larger when the fibers were hyperpolarized (as seen in intra-axonal records from fibers penetrated within the dorsal horn) and smaller when they were depolarized. The depolarization of the fibers within motor nuclei was associated with smaller excitatory postsynaptic potentials (EPSPs) in motoneurons targeted by the fibers, whereas the hyperpolarization was associated with the opposite effects (Eccles et al. 1962).

Fig. 1.

Diagram of changes in membrane potential along group Ia afferents traversing the spinal gray matter during application of direct current (DC) via 2 electrodes contacting the surface of the spinal cord. DC was applied via a large anode close to the dorsal root entry zone and a smaller cathode close to the ventral roots in the study of Eccles et al. (1962). The diagram indicates the sites of hyperpolarizing (+) and depolarizing (−) effects in the regions closest to the anode and cathode, respectively, but also hyperpolarization of motoneurons secondary to the depolarization of their axons, reflecting the intra-axonal and intracellular gradient of potentials. [Adapted from Fig. 1 of Eccles et al. (1962); © 1962 The Physiological Society.]

The effects of DC applied transspinally, using the noninvasive procedures of tsDCS, have been more difficult to localize. When DC was applied between an electrode in contact with the skin at the back, or just over the vertebral column, and a large reference electrode on a distant part of the body, usually the belly, chest, or arm, the intensity of the polarization had to be increased to 1–2.5 mA to be effective. The electric fields evoked in this way were estimated to extend over a few spinal segments, both in human subjects and in animals, and to cover most of the cross section of the spinal cord, including motor nuclei and spinal nerves in these segments (Hernández-Labrado et al. 2011; Parazzini et al. 2014b, 2014c). In addition, in most studies in which the effects of tsDCS were analyzed, the effects were estimated on the basis of changes in responses evoked in motoneurons but depended on input relayed by unspecified premotor neurons, which were most likely spread out throughout large areas of the gray matter, even in the cases of responses evoked by the most selective stimuli. The same DC might thus depolarize some of them while hyperpolarizing the other ones. The wide distribution of neurons that relay effects evoked by cortical or peripheral stimulation to motoneurons affected by tsDCS was indeed directly demonstrated by using multielectrode arrays for recording from the rat spinal cord (Song and Martin 2017; see also Ahmed 2011, 2013a, 2013b, 2014b, 2011; Bocci et al. 2015; Cogiamanian et al. 2011; Priori et al. 2014; Truini et al. 2011). With respect to responses evoked in motoneurons, transspinal polarization applied in anesthetized mice or rats just outside the vertebral column, or via skin, appeared to have effects resembling those of DC applied across the feline spinal cord (see above). This was, for example, the case when the current flow was between a focal electrode placed over the mouse spinal cord and the reference electrode in contact with skin on the belly (Ahmed 2011) or hip [favoring the current flow through axons of hindlimb motoneurons (Ahmed 2013b, 2014a)]. In humans, the results varied depending on the test and configuration of the focal and reference electrodes (for review, see Priori et al. 2014), but in several cases they were as in anesthetized mice and rats (see e.g., Bocci et al. 2015; Cogiamanian et al. 2008; Winkler et al. 2010).

effects of cortical polarization in vivo.

When weak DC was applied to the surface of the cortex, its effects varied depending on the location of the affected neurons. As judged by the frequency of neuronal firing of neurons in the rat, positive current usually facilitated activation of more distant, i.e., deeper located cells, whereas opposite-polarity DC excited more superficial, or nearer, cells (0.1–2.5 µA applied focally; 30–80 µA/mm² over 4-mm² area; Bindman et al. 1964). When DC was applied over the motor cortex, its effects were consistent with direct effects of cortical stimulation on deeply situated feline or primate corticospinal pyramidal tract (PT) neurons, the initial segments of which are most effectively depolarized by surface-positive current pulses (Edgley et al. 1990, 1997; Hern et al. 1962; Phillips 1956; see also Molaee-Ardekani et al. 2013), with the current flowing down the long, vertically oriented apical dendrites, cell bodies, and axons of these neurons. These effects were also consistent with the results of studies on the mode of activation of nerve cells by extracellular stimuli, leading to the conclusion that the effects of the extracellular stimuli are exerted primarily via spread of current to the initial segment of the axon and its depolarization (Gustafsson and Jankowska 1976; McIntyre and Grill 2002; Radivojevic et al. 2016).

The effects of tDCS also differed depending on the location of the neurons within the DC target area as well as the species. For most of the cortically relayed responses in humans, positive tDCS was found to be excitatory (1–2 mA, 0.3–0.6 µA/mm²; Jamil et al. 2017; Lefaucheur et al. 2017). This was particularly the case for responses induced by corticospinal tract neurons, although some cortical neurons were excited by the opposite polarity DC (e.g., neurons situated deep in cortical sulci; see Creutzfeldt et al. 1962; Nitsche et al. 2008). When the subcortically relayed effects of tDCS were estimated, positive tDCS was found to facilitate activation of rubrospinal, vestibulospinal, and reticulospinal neurons in cats (Bączyk et al. 2014; Bolzoni et al. 2013) and humans (Nonnekes et al. 2014), whereas negative tDCS was excitatory in rats (Bączyk and Jankowska 2014). Whether the differences in subcortical effects of tDCS in cats and rats, both anesthetized and under generally similar experimental conditions, depended on the size of the skull and differences in the distances between the electrodes used for tDCS and the explored subcortical regions, or on other factors, could not be defined. If it depended on the size of the skull, the effects found in the cats could be more consistent with the effects of DC on the human brain than effects in rodents. However, the conclusions regarding the subcortical effects of tDCS in both rats and cats were consistent with respect to similar effects of DC on nerve fibers stimulated in the red nucleus, because in both these species the effective tDCS depolarized the interpositorubral fibers. Facilitation by tDCS of subcortically mediated responses evoked by startling acoustic stimuli in humans was likewise compatible with direct tDCS effects on fibers or neurons in the reticular formation (Nonnekes et al. 2014). Thus, if effects of the current within the explored brain region of the various species are related to the depolarizing or hyperpolarizing effects of tDCS in these regions, the conclusions regarding the effects of tsDCS and tDCS are generally in agreement.

effects of uniform electric fields on different cell compartments in vitro.

In view of large areas within which any electric field may affect neurons, the effects of uniform electric fields in vitro have provided particularly valuable information on the distribution of these effects and the basis for interpreting DC effects within the cortex as well as within the subcortical nuclei and spinal cord. In in vitro slice preparations of the hippocampus (Bikson et al. 2004; Kabakov et al. 2012) and motor cortex (Radman et al. 2009, 2013), uniform electric fields were applied by passing DC of a few millivolts, corresponding to ~10 µA, between two large wires positioned in the bath across a slice. The effects of these fields were estimated from population spikes or EPSPs evoked in antidromically or monosynaptically excited cells, allowing the examination of DC effects on selected cell types.

Relating the effects of uniform electric fields in different directions to the morphology of individual cells, as illustrated in Fig. 2, provided an even more unique opportunity to estimate the effects of DC on various compartments of individual cells. Because the distances between the sources of the DC and the analyzed structures in the slices were on the order of millimeters or less (Jefferys 1981), much weaker electric fields could be used, and both the intensity and the direction of the applied current could be modified. The ranges of minimal effective polarization, from a few to 20 mV/mm, were then found to be close to the amplitudes of the effective intrinsic field potentials around Mauthner cells (≤0.4 mV to a few mV; Faber and Korn 1989; Weiss and Faber 2010) and not far from the naturally occurring conditions (see below). Both the experimental data and the resulting models have demonstrated that the excitability of all neuronal compartments is modulated by anodal as well as cathodal polarization. They also demonstrated that “axon terminals may be . . . more susceptible to polarization than somas,” because the increases in the excitability of the presynaptic fibers were generally evoked at a lower cathodal DC intensity than in the excitability of the target cells of these axons (Rahman et al. 2013). In addition, the effects on the cells greatly depended on their geometry and on the direction of the current with respect to the trajectory of the dendrites and initial segment. These results accordingly show that the varying effects of tDCS on cortical neurons at different locations (see above) may be explained by differences in their morphology and location. The same explanation would also apply to differing effects of DC on spinal neurons and fibers in different laminae of the spinal gray matter.

Fig. 2.

Reconstruction of voltage gradients in uniform electric fields of cortical gray matter in a sagittal slice preparation of the mouse motor cortex and degree of polarization of individual cells depending on the current direction. A: cell axes of cortical neurons with respect to the surface of the cortex in different parts of a gyrus and to the direction of radially or tangentially applied electric currents. B: voltage gradient in a region of a uniform radial electric field. C and D: compartment model simulations of morphologically reconstructed pyramidal neurons in layers 5 and 2/3, in a tangentially (C) and radially (D) directed field, indicating their differential polarization. Dep., depolarization; Hyp., hyperpolarization; E, electrical field. [Modified from Fig. 1 in Rahman et al. (2013) with permission; © 2013 The Authors. The Journal of Physiology.]

local effects of dc in the spinal cord.

To circumvent the widespread effects of DC, advantages were also taken of the considerable differences in the density of monophasically applied radial electric fields in a conductive medium at different distances from their source. The highest current density around a focal electrode allowed accordingly the reduction of the area within which the effects of DC were evoked at threshold intensity as well as the restriction of these effects to the cells or fibers closest to the electrode. When local DC was applied to the cortical surface via recording micropipettes, it was noted that the polarized zone was sharply demarcated and that the effects of 0.1- to 0.5-µA DC were detectable within only about a 100-µm distance (Bindman et al. 1964). The effective polarity was nevertheless the same as in the case of larger polarization areas (cf., Bindman et al. 1964; Purpura and McMurtry 1965). In the spinal cord, local application of DC appeared to be of particular importance in view of the considerable variability in the effects of transspinal polarization as well as the wide range of the effects. The intensity of the effective local DC in the spinal cord (Bolzoni and Jankowska 2015) and brain stem (Bączyk and Jankowska 2014) were found to be on the same order as in the cortex, i.e., 0.1–0.5 µA, although intensities of 0.2–0.3 µA were considered preferable in view of the risk of either cathodal or anodal block of transmission by higher intensities. The local cathodal DC was consistently excitatory, whereas anodal DC was inhibitory.

When the susceptibility of spinal cells and fibers to locally applied DC was compared, a common finding was that DC affects fibers to a much greater extent and at a much lower threshold than the somata, fully in line with the conclusions based on in vitro studies of cells and fibers in the hippocampus and cortex (Bikson et al. 2004; Márquez-Ruiz et al. 2012; Rahman et al. 2013), in tissue culture (Radivojevic et al. 2016), or in vivo in the red nucleus (Bączyk and Jankowska 2014; Baldissera et al. 1972). The differentiation between DC effects on spinal cells and fibers was helped by the very favorable morphology of the spinal cord and its input and output neurons. Thus any facilitation of activation of motoneurons could be estimated on the basis of direct recording from their axons in the ventral roots and changes in excitability of the central terminals of sensory fibers by activating them by intraspinal electric stimuli and by recording action potentials evoked in the peripheral nerves (as is diagrammatically indicated in Fig. 3E). A DC-evoked increase in the excitability of the fibers reflected the degree of depolarization of these fibers and resulted in an increased number of the excited fibers, and therefore much larger nerve volleys recorded in the peripheral skin or muscle nerves (with an illustration in Fig. 3A). The nerve volleys in both muscle and skin afferent fibers were found to be considerably increased by DC, whereas the direct effects of DC on neurons at the same location were much weaker. At least, DC applied in the motor nucleus only marginally increased the number of directly activated motoneurons (see Fig. 5E) and allowed the DC-evoked facilitation of synaptically evoked EPSPs (Fig. 3, B and C) or the activation of motoneurons (Fig. 3D) to be attributed to effects of DC at a presynaptic level (Bolzoni and Jankowska 2015). The negligible effects of DC on motoneurons under these conditions were consistent with similarly negligible direct effects of tDCS on rubrospinal neurons (Bączyk and Jankowska 2014) and of tsDCS on human motoneurons (see Bocci et al. 2015; Priori et al. 2014).

Fig. 3.

Effects of local application of cathodal DC in a motor nucleus. A–E, left: experimental arrangements in different series aimed at defining targets of DC action, with the sites of stimulation and recording indicated. A–E, middle: examples of the effects, with blue traces showing control responses and superimposed orange and green traces showing representative responses evoked by the same stimuli during and after DC (0.3 µA) application, respectively. A–E, right: time courses of the mean facilitatory effects of DC using normalized areas of the earliest response components (with respect to the control response areas within the same time window) to monitor the effects of DC (ordinate) within 80 min (abscissa). Blue data points indicate control responses. Black circles indicate responses recorded during successive 5-min periods of DC application. Open and filled diamonds represent responses evoked either between the polarization periods or during the postpolarization period. Orange or green symbols represent mean values of responses illustrated in A–E, middle. Blue dotted horizontal lines show the control level. Vertical dotted lines indicate the end of the last polarization. A: increase in the excitability (depolarization) of muscle afferents, reflected in larger nerve volleys recorded in peripheral nerves (PB). B: increases in EPSPs evoked in motoneurons. C: increases in extracellular field potentials evoked in the motor nucleus by two stimuli. D: monosynaptic field potentials evoked by 2 stimuli, showing a considerable increase of efferent volleys recorded from a ventral root. E: increased excitability of motoneurons during but not after DC application, as monitored by recording efferent volleys from motor axons in the ventral root (VR). [Modified from Figs. 2–8 in Bolzoni and Jankowska (2015); © 2014 The Authors. The Journal of Physiology.]

Fig. 5.

Examples of activity-dependent and activity-independent effects of DC. A and B: graphs show changes in the amplitude of field potentials induced in a slice preparation of the mice motor cortex by intracortical stimuli before, during, and after polarization of the slice in a uniform electric field (with the anode on the side of the cortical surface, using 10-µA current corresponding to 0.75 mV/mm for 15 min). DCS, direct current stimulation; LFS, low-frequency stimulation. [A and B are adapted from graphs of Fig. 1C and Supplemental Fig. S1B in Fritsch et al. (2010) in the same scale, with permission of Elsevier.] C: changes in the excitability of group Ia afferents stimulated by single stimuli within the peroneus motor nucleus, monitored by recording nerve volleys in these afferents before and after DC application with the same electrode, as in A. The fibers were stimulated between successive 5-min periods of DC application (0.3 µA) and for 1 h following the last DC polarization period. D: same as C but monitoring the effects of DC on monosynaptic field potential evoked by group Ia afferents in the peroneal motor nucleus. E: changes in the excitability of afferents stimulated by single epidural stimuli, monitored as in C, before and after DC application via the same epidural electrode. The fibers were stimulated before DC application and for an hour following the last of 5 DC polarization periods. [C and D are modified from Figs. 5A and 6J in Jankowska et al. (2016), © 2016 Federation of European Neuroscience Societies and John Wiley & Sons Ltd; E is modified from Fig. 5C in Jankowska et al. (2017).]

The effects of DC on synaptic transmission to neurons at different locations in the spinal gray matter were less consistent (Bolzoni and Jankowska 2015). They showed nevertheless that even weak facilitation of synaptic actions by presynaptic fibers could considerably facilitate activation of the target neurons of these fibers, doubling or even tripling their numbers, as illustrated in Fig. 5D with effects of DC on monosynaptic activation of motoneurons (monosynaptic reflexes).

Sustained DC Effects

A particularly important feature of effects of DC is that these effects outlast the period of DC application. This has been found to be true not only at the cortical level but also at subcortical and spinal levels. In fact, the original description by Bindman et al. (1964) of sustained facilitation of responses of cortical neurons following weak cortical polarization in the rat applies to changes evoked by DC at both the subcortical and spinal levels in all species that have been investigated to date (cat, rat, and mouse). According to the description, “. . . the changes . . . usually did not reach a maximum immediately after a constant current was passed. A period of several minutes usually elapsed before the peak of the effect was reached and also the effect often persisted for a time after the current was switched off.” When the cortical surface was polarized for 5–10 min or longer, the postpolarization changes lasted for hours.

In humans, the aftereffects of tDCS were found to vary depending on the tDCS parameters. In the latest review of tDCS effects by Lefaucheur et al. (2017), the situation is described as follows: “Stimulation of short duration (several seconds) is sufficient to induce these excitability changes, which however do not relevantly outlast the stimulation period (Nitsche and Paulus 2000; Priori et al. 1998). A longer stimulation duration (several minutes) induces excitability changes that can last for one hour or more (Nitsche et al. 2003; Nitsche and Paulus 2001; Priori 2003). . . . M1 excitability changes become steadily significant after the end of tDCS application rather than during stimulation (Santarnecchi et al. 2014).” The long-lasting effects of tDCS are favored by repeated application. For instance, Reis et al. (2009) found that tDCS sessions repeated over 5 consecutive days allowed the retention of postpolarization effects of tDCS on motor skills for at least 3 mo. A comparison of the effects of tDCS repetition at different intervals revealed, in addition, that intervals of 3–20 min promoted the long-lasting tDCS effects more efficiently than either shorter intervals or intervals of 24 h (Jamil et al. 2017; Monte-Silva et al. 2013).

The increased effects of repeated application of DC have also been noted in animal studies, irrespective of whether DC was applied transcranially, transspinally, or locally. In anaesthetized animals, long-lasting DC effects appeared to develop slower than in humans. Nevertheless, subsequent to two to three periods of polarization (with between-polarization periods of 5 min), the facilitation was often observed for up to 2–3 h, and both shortening of the latencies and increases in the amplitude of the tested responses became even more pronounced after the tDCS had been terminated than between the successive periods of DC application (Bączyk and Jankowska 2014; Bączyk et al. 2014; Bolzoni and Jankowska 2015; see also Ahmed 2014a; Ahmed and Wieraszko 2012).

Sustained effects of DC have often been considered to involve the same mechanisms as long-term potentiation (LTP) or depression (LTD) in structures such as the cerebral or cerebellar cortex or hippocampus (for review, see e.g., Fritsch et al. 2010; Lefaucheur et al. 2017; Monte-Silva et al. 2013; Nitsche et al. 2012). This is particularly the case for the long-lasting effects of tDCS on synaptic transmission and on cortical neuronal networks. However, other mechanisms also have been indicated by some nonsynaptic DC aftereffects, particularly by postpolarization increases in the excitability of fibers in peripheral nerves lasting at least 2 min (Ardolino et al. 2005; Esposti R, Bruttini C, Zenoni G, Jankowska E, Cavallari P, and Bolzoni F, unpublished observations). The nonsynaptic effects were considered to include “. . . changes of conformation and function of various axonal molecules involved in transmembrane ion conductance, membrane structure, cytoskeleton, or axonal transport when exposed to a DC field. . . . ” (Lefaucheur et al. 2017). The contribution of synaptic and/or nonsynaptic mechanisms is therefore of prime interest for the interpretation of the sustained subcortical and spinal effects of DC, because the degree of plasticity is much lower at the subcortical and spinal levels than at the cortical level and thus may not involve the same mechanisms.

Sustained increases in the excitability of presynaptic fibers by DC and weak DC effects on postsynaptic neurons in the red nucleus.

When DC was applied in the red nucleus in anesthetized cats and rats, it caused long-lasting increases in the excitability of terminal branches of interpositorubral fibers providing input to rubrospinal neurons and facilitated transsynaptic activation of these neurons by electrical stimuli to a greater extent than by their direct activation. The increased excitability of these presynaptic fibers was also demonstrated by recording from single interpositorubral neurons in the cerebellum and comparing the probability of their antidromic activation before, during, and after DC application in the red nucleus (Bączyk and Jankowska 2014). Cathodal tDCS doubled the probability of antidromic activation of these neurons by stimuli applied in the red nucleus (e.g., from 40 to 50 to 100 responses per 100 stimuli). Very similar effects were obtained when the polarizing current was applied using tDCS (over the motor cortex) and when it was applied locally within the red nucleus itself at a much lower intensity (<1 µA) (Bączyk and Jankowska 2014; Bolzoni et al. 2013). In both cases, similarly long-lasting aftereffects (more than 2 h) were evoked.

Sustained increases in the excitability of presynaptic fibers by intraspinally applied local DC and weak DC effects on α-motoneurons.

Increased excitability of presynaptic fibers stimulated within the spinal cord was found not only during local DC application but also during the postpolarization period. The sustained effects on presynaptic fibers were then as polarity dependent as the effects of tDCS, but were consistently facilitatory when cathodal DC was applied and depressive when anodal DC was applied in anesthetized rats (Bolzoni and Jankowska 2015; Jankowska et al. 2016). As illustrated in Fig. 3A, the activation of group I afferent fibers in a motor nucleus and cutaneous, or group II, muscle afferents in the dorsal horn was increased during subsequent applications of cathodal DC as well as during the directly following between-polarization periods. Thereafter, the excitability either continued to increase during at least 30 min of the postpolarization period or tended to decline but remained above the prepolarization level throughout this period.

These changes were matched by facilitation of synaptic actions of group Ia afferents in the feline motor nucleus, as reflected by increases in both intracellularly recorded EPSPs (Fig. 3B) and the synaptic field potentials evoked by these afferents (Fig. 3C). Both increased during and between the successive periods of DC polarization and outlasted the last DC application. However, when changes in the excitability of motoneurons to stimuli applied in the motor nucleus were monitored by recording discharges in a ventral root (Fig. 3E), the increased excitability of these neurons was seen during, but not after, local DC application; i.e., it did not match the sustained changes in the excitability of presynaptic fibers.

Sustained increases in the excitability of afferent fibers stimulated within the dorsal columns by epidurally applied DC.

Changes in the excitability of nerve fibers evoked by intraspinally applied DC provide indications of the effects of intrinsic electric fields that might be evoked under natural conditions. However, intraspinal polarization would not be easy to use for therapeutic purposes in humans, unless used together with intraspinal stimulation, explored by the Mushahwar and Prochazka groups (Holinski et al. 2016; Mushahwar et al. 2000; Prochazka 2016; Prochazka and Mushahwar 2001). DC polarization, in contrast, might be reasonably easy to combine with the clinically routinely used epidural stimulation.

In addition, the effects of epidurally applied DC appear to be even more potent than the effects of intraspinally applied DC (Jankowska et al. 2017). In acute experiments in rats, this was examined by comparing the effects of DC on nerve volleys evoked by epidural stimulation of sensory fibers in the dorsal columns and recorded in peripheral nerves. Both the epidural stimulation and epidural DC polarization required stronger stimuli (although not exceeding DC of 50 and 1 µA, respectively) than when applied intraspinally, but DC-evoked increases in nerve volleys in cutaneous and/or muscle afferents were often severalfold. However, in contrast to the intraspinally evoked effects, the increases in the excitability of sensory fibers in the dorsal columns were evoked within only a few seconds, and 15–30 s of cathodal polarization sufficed for postpolarization effects lasting 1–2 h. These findings are illustrated in Fig. 4B, which shows the antidromically conducted nerve volleys recorded every second, in Fig. 4, A and C, which shows the averages of these volleys, and in Figs. 4D and 5E, which show the time course of their changes. These facilitatory effects provide another example of the sustained, nonsynaptic effects of DC, because epidurally applied DC would only secondarily facilitate activation of the target cells of fibers excited by epidural stimuli following increases in orthodromically conducted nerve impulses in the stimulated fibers and their subsequent synaptic actions.

Fig. 4.

Example of the effects of DC on epidurally stimulated afferent fibers. A–D: nerve volleys recorded in the peroneal nerve in response to epidurally applied stimuli (18 µA, at 1 Hz). A: averaged records of control volleys (n = 10). B: control records of single volleys immediately preceding DC application. C: single volleys during the first 12 s of DC application (1 µA, via the same electrode used for epidural stimulation). D: averaged records, as in A, but 5 min after the termination of 1 min of DC application. Dotted horizontal blue line in A–C indicates the original amplitude of the volleys. Continuous gray bar indicates the period of DC application. E: time course of changes in the area of the first volley component, in %control (ordinate), during the postpolarization period (abscissa). contr, Control. [Modified from Figs. 3 and 4 in Jankowska et al. (2017).]

Activity dependence/independence of long-lasting effects of DC in the spinal cord.

Several research groups have demonstrated that sustained facilitation of synaptic actions by tDCS is activity dependent, i.e., that it requires DC to be coupled with repetitive synaptic activation (Fritsch et al. 2010; Nitsche et al. 2012; see also Kim et al. 2017).

Activity- and task-specific modulation of neuronal networks in humans by tDCS has been related to modulation of already activated neuronal networks, and according to Bikson et al. (2013), “subthreshold neuromodulation may reflect changes in ongoing processes.” The activity dependence of DC-evoked facilitation in cortical slices in the experiments of Fritsch et al. (2010) was more dramatic because the DC failed to facilitate synaptic transmission in these slices unless it was applied together with low-frequency stimulation. This effect is illustrated in Fig. 5, which shows the continuous increase in compound EPSPs evoked by intracortical stimuli during the 1–2 h following 15 min of polarization associated with these stimuli (Fig. 5A), whereas any facilitation during DC application without stimulation declined to control levels within a few minutes (Fig. 5B). Sustained effects of tDCS, and their slow increase during 30 min of the postpolarization periods, were, however, also found after a 15-min period of DC application during which corticospinal neurons were not activated, at least not intentionally (Jamil et al. 2017). Some effects of tDCS (e.g., synaptic) might thus be activity dependent, whereas other effects, in particular, the nonsynaptic ones, might be activity independent. It was therefore critical to investigate whether the same mechanism underlay the synaptic and nonsynaptic effects of DC by verifying whether the sustained DC-evoked increases in fiber excitability depend or do not depend on the activation of the fibers. When this question was addressed, several facilitatory effects of DC in the spinal cord were found not to require stimulation during DC application. Such activity independence has been demonstrated in the case of DC-induced increases in the excitability of fibers evoked in the absence of concomitant activation of these fibers during polarization, as well as increases of monosynaptic field potentials evoked by these fibers (Jankowska et al. 2016). Activity-independent facilitation evoked by intraspinally applied DC (0.3 µA) is illustrated in Fig. 5, C and D, whereas Fig. 5E depicts an even stronger and more rapid facilitation evoked by epidural polarization (1 µA).

Whether activity-dependent changes in synaptic activity evoked by DC and the activity-independent effects of DC on nerve fibers are more or less independent, and, if so, how they are integrated, remain open questions. It would be also important to know whether nerve fibers in other regions of the nervous system are similarly affected, and to what extent long-lasting depolarization of these fibers might contribute to any sustained central postsynaptic DC-evoked changes.

Effects of DC on complex spinal neuronal systems.

As illustrated in Fig. 3, locally applied DC or locally generated intrinsic field potentials may have multiple direct and indirect effects. DC applied at a distance must, therefore, have even more widespread effects, as has been demonstrated directly (Song and Martin 2017) as well as deduced from the gradients of electric fields (Parazzini et al. 2011, 2014a; Rahman et al. 2013; see Fig. 2). Electric fields may thus have an impact on a variety of neurons and spinal neuronal systems, depending on how and the region within which they are applied.

Direct effects of tsDCS on spinal motoneurons and on monosynaptic activation of these neurons by muscle afferents appeared to be very weak (Bocci et al. 2015; Priori et al. 2014), unless under conditions where the motoneurons were specifically addressed (Ahmed 2013b; Ahmed and Wieraszko 2012). The effects of tDCS on one category of spinal interneurons, those mediating reciprocal inhibition of motoneurons, which are located just outside the motor nuclei (see e.g., Jankowska 2012), were likewise difficult to demonstrate (Lackmy-Vallée et al. 2014). Nevertheless, the effects of tDCS on widespread neuronal networks mediating flexion-extension, locomotion, or centrally initiated movements were substantial (Ahmed 2011, 2013a, 2014a; Hubli et al. 2013; Song and Martin 2017; see Bocci et al. 2015; Priori et al. 2014). Figure 6 illustrates the effects of spinal polarization on background firing level of samples of neurons in both the dorsal and ventral horn as well as on their responses to single stimuli applied to the motor cortex in rats.

Fig. 6.

Examples of increased firing rate in spinal neurons during cathodal DC application and decreased firing rate during anodal polarization. A–C: raster plots and peristimulus time histograms of 2 neurons before and during transspinal DC application (c-sDCS, cathodal spinal cord DCS; a-sDCS, anodal spinal cord DCS). D: mean changes in the firing rate of a sample of neurons in regions of interest in the dorsal (dROI) or ventral horns (vROI) in response to stimulation of the motor cortex. [Modified from Fig. 4 in Song and Martin (2017).]

The dependence of effects of DC on the mode of its application may be illustrated in two spinal modulatory systems: presynaptic inhibition and postactivation depression. In anesthetized rats, both presynaptic inhibition and postactivation depression were facilitated by local cathodal DC and weakened by anodal DC, whereas the effects of tsDCS were found not to be polarity specific, because they were facilitated by both cathodal and anodal tsDCS (Kaczmarek et al. 2017). In humans, the effects of tsDCS were not consistent, because postactivation depression of the monosynaptic actions of Ia afferents on motoneurons (H-reflex) was enhanced by cathodal tsDCS but was either weakened or largely unaffected by anodal tsDCS (Knikou et al. 2015; Lamy et al. 2012; Winkler et al. 2010). Before tsDCS is used for therapeutic purposes in humans to weaken exaggerated spinal reflexes or to enhance excessively weakened ones, it would thus be important to verify the effects of tsDCS in both animals and humans, to avoid any adverse effects, especially in combination with other treatments used.

Effects of Intrinsic Electric Fields in the Spinal Cord

Experimental evidence for electric interactions in the nervous system has been found between adjacent nerve fibers or nerve cells in vertebrates as well as invertebrates (for reviews see Anastassiou and Koch 2015; Anastassiou et al. 2011; Faber and Korn 1989; Qiu et al. 2015; Weiss and Faber 2010). Such interactions were found under circumstances when sufficiently strong electric field potentials were generated around neurons, or when gap junctions were present between them, and when they appeared to act as extracellularly applied electric stimuli depolarizing or hyperpolarizing the neighboring neurons or fibers. However, within the vertebrate spinal cord, effects of electric fields have apparently only been demonstrated in one study, by Nelson (1966). In this study, negative field potentials were evoked around selected spinal motoneurons by stimulating motoneuron axons in the ventral root filaments (at a strength below the threshold for these neurons). The antidromic field potentials induced in this way were shown to increase motoneuron excitability by facilitating induction of action potentials when paired with subthreshold excitatory postsynaptic potentials (EPSPs) evoked by dorsal root stimulation. They likewise reduced the threshold for activation of the motoneurons by intracellularly applied current pulses. In both cases, the effects were reminiscent of effects of locally applied subthreshold stimulus pulses or DC, but the facilitation only occurred during a period of ~1 ms, corresponding to the duration of the antidromic field potential (Fig. 7, A and B₃).

Fig. 7.

Example of the facilitation of synaptically evoked activation of motoneurons by antidromically evoked field potentials around neighboring motoneurons, as demonstrated by Nelson (1966). The records illustrate the possibility of inducing an action potential (B₃) by combining stimulation of group Ia afferents in the medial and lateral gastrocnemius nerve with stimulation of ventral root fibers under conditions where the 2 stimuli were subthreshold when applied separately (B₁, B₂). B₂ shows that nerve stimulation by itself only evoked an EPSP, whereas B₁ shows that ventral root stimulation by itself only evoked an intracellularly recorded extracellular field potential (<10 mV), in contrast to effects of the stronger VR stimulation that was followed by antidromic activation of the motoneuron. Records in B₁ and B₂ are superimposed in B₃ as well as at a higher amplification in A. [Adapted from Fig. 1 in Nelson (1966).]

Facilitation of activation of spinal motoneurons was found to be evoked by antidromic field potentials of a few millivolts, and even field potentials around a single spinal motoneuron appeared to be sufficient. Furthermore, such field potentials were found to be recordable for distances exceeding 0.5 mm (Nelson 1966). Effective electric field potentials of similar amplitudes (≤0.4 mV to a few mV) were found around Mauthner cells in the fish brain stem (Faber and Korn 1989). Field potentials around other spinal neurons have not been systematically analyzed, but those evoked by low-threshold skin and group II muscle afferents in Rexed’s laminae III and IV are likely to be among the largest ones; they amount to several millivolts when a number of these afferents are activated synchronously. Large field potentials would also be expected around the spontaneously firing dorsal horn neurons examined recently by Rudomin’s group, the activity of which was concluded to be reflected in spontaneously appearing electric field potentials in laminae III and IV as well as in cord dorsum potentials (Chávez et al. 2012). The cord dorsum potentials were of 100–150 µV and were related to spontaneously appearing field potentials of, on average, 200–300 µV at a depth of 1–1.5 mm (Chávez et al. 2012; Manjarrez et al. 2000), but individual field potentials giving rise to cord dorsum potentials at such distances must have been much larger.

Faber and Korn (1989) stressed that whenever extracellular field potentials are evoked, “some current must flow across the membranes of neighboring cells and influence them.” They considered, therefore, the field effects to be another class of synaptic input to a neuron that supplements chemically mediated synaptic transmission. The advantage of field effects would be that they occur with no synaptic delay and may shorten the latency of synaptic responses in a way similar to what happens in mixed electrotonic and chemical excitatory synapses. More recently, Weiss and Faber (2010) pointed out that electrical fields might also modify chemical synaptic transmission, for example, by clearing negatively charged transmitter molecules or by influencing voltage-gated channels involved in exocytosis (Sylantyev et al. 2008; Voronin 2000).

Another effect of electric fields might be to initiate or to arrest tonic neuronal discharges associated with plateau potentials, long-lasting potentials of all-or-none character, initiated by a sufficiently strong cell depolarization and terminated only when the cell has been repolarized due to a voltage-dependent noninactivating Ca²⁺ conductance, and giving rise to sustained discharges (Hounsgaard et al. 1988; Hultborn 2002). As illustrated in Fig. 8 with records of Hounsgaard et al. (1988), a short-lasting depolarization of a neuron may induce discharges that outlast the excitatory input, stopping only when the neuron is becoming hyperpolarized. This has been demonstrated with intracellularly applied current pulses or properly timed synaptic input, but the effects of extracellular electric fields may not be negligible and in this way contribute to the modulation of spinal neuronal activity.

Fig. 8.

Example of tonic discharges of a motoneuron (top trace) induced by intracellularly applied depolarizing current pulse (bottom trace) and terminated by hyperpolarizing current pulses. Responses were evoked under conditions in which intracellular pulses were able to evoke plateau potentials, as demonstrated by Hounsgaard et al. (1988). If extracellularly evoked current pulses have a similar effect, intrinsic field potentials might likewise trigger sustained neuronal activity or terminate it, and in this way contribute to the shaping of motor reactions. [Adapted from Fig. 3A in Hounsgaard et al. (1988); © 1988 The Physiological Society.]

Recent studies have also revealed that electric coupling (gap junctions) is much more common than previously thought. As judged by the presence of connexin36, a protein that mediates electrical communication via gap junctions, as well as the effects of gap junction blockers, gap junctions exist in the spinal cord between both motoneurons and other spinal neurons (Bautista et al. 2012, 2014). Gap junctions might thus contribute to spreading out of an even very weak degree of polarization of neurons by either intrinsic or external electric fields to other neurons and to the strengthening of the interactions between them.

Taken together, such observations indicate a variety of ways in which spinal activity may be adjusted by electric fields generated around individual neurons and neuron populations. However, those pointed out above refer to only a few possibilities at a cellular level, and the possibilities of modulation at a molecular level would be endless and might involve all aspects of short- and long-term synaptic plasticity (Zucker and Regehr 2002).

In conclusion, a great deal is already known about the effects of both intrinsic and external field potentials and that they must be combined with the effects of other spinal control systems outlined at the beginning of this review. However, how all these effects integrate to ensure optimal motor output, and which mechanisms underlie effects of the electric fields, remain open questions.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author.

AUTHOR CONTRIBUTIONS

E.J. researched, wrote, and approved the manuscript and prepared figures.