|Home | About | Journals | Submit | Contact Us | Français|
Despite its increasing use in experimental and clinical settings, the cellular and molecular mechanisms underlying transcranial direct current stimulation (tDCS) remain unknown. Anodal tDCS applied to human motor cortex (M1) improves motor skill learning. Here, we demonstrate in mouse M1 slices that DCS induces a long-lasting synaptic potentiation (DCS-LTP), which is polarity-specific, NMDA-receptor dependent and requires coupling of DCS with repetitive low-frequency synaptic activation (LFS). Combined DCS and LFS enhance BDNF-secretion and TrkB-activation, and DCS-LTP is absent in BDNF and TrkB mutant mice, suggesting that BDNF is a key mediator of this phenomenon. Moreover, the BDNF val66met polymorphism known to partially affect activity-dependent BDNF secretion impairs motor skill acquisition in humans and mice. Motor learning is enhanced by anodal tDCS, as long as activity-dependent BDNF secretion is in place. We propose that tDCS may improve motor skill learning through augmentation of synaptic plasticity that requires BDNF-secretion and TrkB-activation within M1.
Transcranial direct current stimulation (tDCS) is a non-invasive brain stimulation technique that has garnered increasing interest due to its modulatory effect on cognitive functions and motor behaviour in healthy subjects and patients with neuropsychiatric diseases (Ferrucci et al., 2008; Hummel et al., 2005; Nitsche et al., 2003a; Reis et al., 2009; Wassermann and Grafman, 2005).
One interesting cortical target for the application of tDCS has been the primary motor cortex (M1), a region crucially involved in motor execution, memory formation and consolidation of motor skills in humans and animals (Muellbacher et al., 2002; Rioult-Pedotti et al., 2000; Ungerleider, 1995). We recently demonstrated that repeated sessions of motor training with concurrent anodal tDCS applied over M1 facilitates learning over multiple days through an enhancement of consolidation (Reis et al., 2009). Despite such encouraging effects of tDCS to modulate human behaviour little is known about its underlying mechanisms.
Electrophysiological data in animals and humans suggest that DCS elicits long-lasting, polarity-dependent changes in neocortex excitability (Bindman et al., 1962; Creutzfeldt et al., 1962; Gartside, 1968; Nitsche and Paulus, 2000; Purpura and McMurtry, 1965). It has also been presumed that tDCS strengthens synaptic connections (Nitsche et al., 2003b; Nitsche et al., 2004; Cheeran et al., 2008), through a mechanism similar to long term potentiation (LTP), a cellular correlate of learning and memory (Bliss and Collingridge, 1993; Martin et al., 2000; Rioult-Pedotti et al., 2000). However, the synaptic effect of DCS in a brain slice, a model system ideally suited for mechanistic studies, is not known. The present study is designed to study the cellular and molecular mechanisms underlying the effect of tDCS on motor skill learning.
Here, we developed a method to investigate the effects of DCS in mouse M1 slices. Field excitatory post-synaptic potentials (fEPSPs) were elicited in layer II/III by stimulating the vertical pathway (layer V → II/III) at 0.1 Hz (test pulse) in slices derived from adult (6–8 weeks old) 129 SvEV mice. The peak amplitude of fEPSPs was chosen to monitor synaptic efficacy because the slope is frequently contaminated by antidromic activity in M1 (Castro-Alamancos et al., 1995). DCS was applied in parallel to the vertical M1 fibers (Fig. 1A) at a field strength (~ 0.75 mV/mm) in the lower range of the estimated field strengths of tDCS used in human M1 (0.22 (Miranda et al., 2006) − 7.7mV/mm (Wagner et al., 2007)). Application of anodal DCS for 15 minutes resulted in potentiation of the fEPSP in most slices (20 of 26 slices, 77%), which began several minutes after DCS onset and outlasted the stimulation duration, i.e. continued to increase even after DCS cessation for almost an hour before a plateau was reached (Fig. 1B, 1C). This result is consistent with the finding that increases in neuronal firing rates also outlast the stimulation duration (Bindman et al., 1962). The mean synaptic efficacy 30 min after DCS application was 118.0% ± 4.5, but 100.0% ± 0.7 in unstimulated control slices (p= .0045 versus DCS). The DCS-induced potentiation (DCS-LTP) lasted as long as a healthy recording could be maintained (120 min: 132.5% ± 8.3; p < .001, Fig. 1C).
tDCS did not improve human motor skill learning in the absence of training (Fig. S1A), suggesting that the ongoing synapse-specific activity during learning (Rioult-Pedotti et al., 2000) may be required to derive the beneficial effect of anodal tDCS. We therefore reasoned that DCS applied to the M1 slice may need to be coupled with simultaneous synaptic activation to induce DCS-LTP. Indeed, in the absence of synaptic activation (no LFS at 0.1 Hz), DCS elicited only a short-lasting increase in fEPSP amplitude, which returned to pre-DCS levels within 10 minutes (Fig. S 1B). Varying the stimulation frequencies while keeping DCS constant showed that frequencies higher (0.2 Hz) or lower (0.0166 Hz) than 0.1 Hz were less effective (Fig. S1C, S1D) indicating that the frequency of synaptic coactivation is relevant for the DCS effect.
Consistent with findings in humans that tDCS polarity is critical in modulating motor cortical excitability (Nitsche and Paulus, 2000) and behaviour (Nitsche et al., 2003a; Reis et al., 2009), DCS-LTP could not be induced when the polarity was reversed (cathodal stimulation; Fig. S1E). In addition, the effect of DCS was not restricted to layer II/III synapses. The fEPSP peak amplitude at layer II/III→V synapses also showed an increase after DCS (108.9 ± 4.0% at 30 min; p= 0.05 versus baseline, Fig. S1F), suggesting that DCS-LTP is a phenomenon that applies to different neuronal populations with comparable geometry.
To further characterize DCS-LTP, we examined factors known to be involved in other forms of long-term plasticity (Malenka and Nicoll, 1999). Pre-treatment with the NMDA receptor antagonist D-APV (50 μM) completely prevented potentiation induction (Fig. 1D), suggesting that this form of plasticity is dependent on NMDA receptor activation. Indirect evidence for this mechanism of action arises from human studies showing that tDCS-induced alteration of M1 excitability after stimulation (as measured by transcranial magnetic stimulation) could be blocked by oral intake of the NMDA antagonist dextromethorphan (Nitsche et al., 2003b). It is also important to consider that DCS may induce disinhibition (attenuation of GABAergic transmission) during stimulation (Stagg et al., 2009; Nitsche et al., 2004), leading to DCS-LTP. However, a local transient “touch” application (Rioult-Pedotti et al., 2000) of the GABAA antagonist bicuculline (BIC, 3.5mM, equivalent to ~0.6 μM BIC in a bath application (Hess et al., 1996)) in the absence of DCS elicited only a transient increase in fEPSP amplitude, which returned to baseline within 20 min. Furthermore, the typical change in the shape of fEPSPs (broadening) indicating disinhibition occurred after application of BIC alone, but not after DCS stimulation (Fig. 1E). Thus, the DCS mediated increase of fEPSP amplitude is most likely not mediated by direct cortical disinhibition.
The neurotrophin BDNF (brain derived neurotrophic factor) is involved in various forms of cortical synaptic plasticity (Akaneya et al., 1997; Lu, 2003) and its secretion depends on calcium and NMDA receptor activation (Balkowiec and Katz, 2002). To test whether BDNF is a critical mediator of the DCS effect we used M1 slices from adult mice carrying a forebrain-specific deletion of the BDNF gene (postnatal excision of the floxed BDNF allele by Cre recombinase, Zakharenko et al., 2003). Slices derived from BDNFflox/flox, cre mice (6–8 weeks old) exhibited no synaptic potentiation after 15 min of DCS exposure, whereas those from the Cre negative BDNFflox/flox littermates displayed intact DCS-LTP (Fig. 2A). Incubating the slice in aCSF containing the BDNF scavenger TrkB-IgG (1.5 μg/ml) for 1.5 hrs prior to DCS abolished DCS-LTP (Fig. 2B), suggesting that activity-dependent BDNF secretion during DCS mediates the fEPSP potentiation.
To investigate the role of the BDNF cognate receptor TrkB in the induction or expression of DCS-LTP we utilized a chemical genetic approach. In TrkBF616A knock-in mice the endogenous TrkB gene is replaced by TrkBF616A, so that TrkB kinase activity can be selectively inhibited by the membrane-permeable small molecule 1NMPP1 (Chen et al., 2005). In the presence of 5 μM 1NMPP1, M1 slices from TrkBF616A mice failed to exhibit DCS-LTP (Fig. 2C, black filled diamonds). In contrast, DCS-LTP could still be induced in TrkBF616A slices in the absence of 1NMPP1 (p=0.003 baseline vs. end of DCS) and continued to increase after DCS cessation (p=0.008 end of DCS vs. end of recording; Fig. 2C, grey filled diamonds). Application of DCS to control slices from C57BL/6J mice in the presence of 1NMPP1 induced normal DCS-LTP (Fig. S2). Interestingly, when 1NMPP1 was applied to TrkBF616A slices at the end of DCS, DCS-LTP could still occur during DCS (p=0.001 baseline vs. end of DCS), but was not further enhanced after DCS cessation (p=0.56 end of DCS vs. end of recording, Fig. 2C, open diamonds). These results suggest that TrkB activation is required for the induction rather than the maintenance of DCS-LTP.
We next determined whether DCS combined with LFS could enhance secretion of endogenous BDNF, leading to TrkB activation in the M1 slice. Western blot analysis (WB) with antibodies against phospho-TrkB and total TrkB was used to determine the ratio of phospho-TrkB/total TrkB. M1 slices were exposed to LFS with or without DCS (control) for 15 minutes and collected for WB 30 minutes after the end of DCS. Remarkably, DCS induced a 1.93 ± 0.2 fold increase in phospho-TrkB over LFS alone (Fig 2D, p= 0.0002). These results suggest that DCS augments the BDNF secretion induced by LFS alone.
To relate activity-dependent BDNF secretion to human motor learning, we first needed to clarify whether learning a particular motor task is BDNF dependent. A condition with complete loss of activity-dependent BDNF secretion is not found in humans. Thus, we studied healthy volunteers with and without the BDNF Val66Met polymorphism, which is known to partially affect activity-dependent secretion of BDNF (18–30% decrease in Met carriers compared to Val/Val subjects, Chen et al., 2006; Egan et al., 2003). The Met allele is associated with a reduction in practice-dependent increase in the amplitude of motor evoked potentials, motor maps (Kleim et al., 2006; McHughen et al., 2009) and a reduced response to noninvasive brain stimulation protocols that alter cortical excitability (Cheeran et al., 2008). We examined motor skill acquisition over five days using a challenging sequential visual isometric pinch force task (Reis et al., 2009, Fig. S3A) in 36 healthy subjects (Val/Val: n=18, Met carriers: n=18; Val/Met (16), Met/Met (2)). While both groups began with comparable baseline performance, Met carriers displayed significantly reduced motor skill acquisition by the end of day 5 relative to Val/Val individuals (Δskill: 2.56 ± 0.2 (Val/Val) versus 1.48 ± 0.29 (Met carriers); Fig. 3A, 3B, 3C, “Sham” columns, p=0.018).
Consistent with the human experiment, BDNFMet/Met knock-in mice (Chen et al., 2006) exhibited significantly less skill acquisition on an accelerating rotarod task over 5 days of training compared to wild type littermates (Fig. 3D, 3E, 3F, p=0.047). When faced with the same training protocol, there was also a motor learning deficit in BDNFflox/flox, cre mice compared to their cre negative littermate controls (Fig. 3G, 3H, 3I, p=0.038). Learning curves for the BDNFflox/flox, cre mice (Fig. 3G) were similar to those observed in the BDNFMet/Met mice (Fig. 3D). Taken together, these results suggest that activity-dependent BDNF secretion - particularly in the forebrain - is important for motor learning. Our preliminary attempt failed to reveal deficits in DCS-LTP in M1 slices from BDNFMet/Met mice (data not shown). This is not surprising, given that activity-dependent secretion of BDNF is only reduced by ~30% in BDNFMet/Met neurons (Chen et al., 2006; Egan et al., 2003), and that it took ~ 2 training days to reveal the difference in motor skill gains between genotypes in mice and humans.
As the DCS effect in M1 slices depends on activity-dependent BDNF secretion, BDNF secretion affects motor learning in humans (Fig. 3A, 3C), and anodal tDCS augments motor learning in humans (Reis et al., 2009), we wished to determine the relationship between these findings. We trained 34 subjects (Val/Val: n=17, Met carriers: n=17, Val/Met (16), Met/Met (1)) on the same pinch-force task while applying anodal tDCS to M1. As we predicted, Val/Val subjects experienced greater skill improvement compared to Met carriers (Δskill:3.63 ± 0.27 (Val/Val) versus 2.23 ± 0.26 (Met carriers); Fig. 3C “Anodal” columns, p=0.003; Fig. S3C, S3D). Further analysis by ANOVA showed a significant effect of GENOTYPE (Val/Val vs. Met carriers: p<0.000001) and STIMULATION (sham vs. anodal: p=0.001) on learning. The GENOTYPE × STIMULATION interaction was not significant, suggesting that anodal tDCS may induce a facilitatory effect on BDNF-dependent motor skill learning in both genotypes (Fig. 3C, Fig. S3E): As can be seen in Figure 3C the slopes (dotted lines), i.e. the increase in learning with anodal tDCS, are similar. Nevertheless, we cannot reject the null hypothesis that there is no difference in response to tDCS between Met carriers and Val/Val subjects. Direct comparisons of absolute skill improvement in the 4 groups revealed a significant difference between “Sham” and “Anodal” in Val/Val subjects (p=0.012, Bonferroni-corrected), but not in Met carriers (p=0.25, Bonferroni-corrected). However, these comparisons do not allow testing for relative differences in response to tDCS due to the parallel design of the study.
Here we show that anodal DCS to M1, in combination with synapse specific activation – repeated synaptic activation (LFS) in vitro - induces a form of long-term synaptic plasticity that requires activity-dependent BDNF secretion. This form of synaptic plasticity differs from the conventional NMDA-dependent LTP induced by HFS or pairing. First, most forms of LTP are induced by high frequency stimulation or repeated low-frequency pairing of presynaptic stimulation with postsynaptic depolarization (Malenka and Nicoll 1999). Second, although DCS-LTP is dependent on NMDA receptor activation, the estimated shift of the membrane potential (~ 0.20 mV in layer V pyramidal neurons, and less in layer II/III pyramidal neurons, Radman et al., 2009) by electrical field strengths as used in our study (0.75 mV/mm) is more than 50 times lower than the magnitude needed for the voltage dependent release of the Mg2+ block at the NMDA receptor (Nowak et al., 1984) or for direct induction of action potentials (Spruston et al., 2008). Third, unlike conventional LTP in adult primary motor cortex (Castro-Alamancos et al., 1995), transient suppression of inhibitory transmission in M1 is not required to elicit DCS-LTP. Synaptic co-activation by combined LFS and DCS, which is not present in high frequency stimulation protocols, may explain this phenomenon.
It should be pointed out, however, that we cannot exclude that DCS-LTP shares characteristics of spike timing dependent plasticity (STDP). DCS has been shown to amplify spike timing in response to current steps (Radman et al., 2007) and to increase spontaneous neuronal firing rates (Bindman et al., 1962; Purpura and McMurtry, 1965). These mechanisms may increase the probability of coincident pre-/postsynaptic firings (pairing protocol; Malenka and Nicoll 1999), and therefore may also contribute to DCS-LTP induction. Hypothetically, this could explain the dependency of DCS-LTP on NMDA receptor activation. Moreover, protocols used to induce STDP were reported to increase BDNF secretion (Tanaka et al., 2008). Supporting BDNF’s involvement in timing-dependent plasticity in the human motor system, subjects carrying the Met allele show an altered response to combined transcranial magnetic and afferent nerve stimulation (Cheeran et al., 2008).
A substantial proportion of neuronal BDNF is secreted in the pro-form (proBDNF), which is subsequently converted to mature BDNF (mBDNF) by extracellular proteases, e.g. plasmin (Pang et al. 2004). In the hippocampus, the conversion to mBDNF is required for late-phase LTP (Pang 2004). This process is critically dependent on the co-release of proBDNF and tissue plasminogen activator (Nagappan et al., 2009). Although the role of such processes within M1 is not certain, it is tempting to speculate that DCS in conjunction with LFS triggers the co-release of proBDNF and proteases, the latter of which may promote the cleavage of proBDNF to mature BDNF. Mature BDNF-induced TrkB activation would then promote long-lasting potentiation of synaptic efficacy in M1.
Previous work suggested a role of BDNF in motor system plasticity and short term motor performance (Cheeran et al., 2008; Kleim et al., 2006; McHughen et al., 2009). We now report behavioural alterations in prolonged motor skill learning when activity-dependent BDNF secretion is reduced. Our data are consistent with the view that tDCS is beneficial to motor learning when BDNF release occurs through training (in humans) or through LFS (in M1 slices). In the absence of activity-dependent BDNF secretion (e.g. DCS alone in the slice or tDCS in the absence of training in humans), the beneficial effects of DCS may not materialize. However, it should be noted that we have not established that a decrease in activity-dependent BDNF secretion in Met carriers attenuated the effect of anodal tDCS on motor skill learning. Given the small effect of the polymorphism on activity-dependent BDNF secretion (30% for Met/Met and 18% for Val/Met, Egan et al., 2003, Chen et al., 2006) it is conceivable that greater behavioural gains by the Val/Val genotype in response to tDCS may not be effectively detectable by our approach. Further studies using different techniques or animal models may resolve this issue.
While our mouse slice and human behavioural experiments were intended to complement each other to enable deeper insights into the mechanisms of tDCS, one should not neglect the significant species differences between mice and humans and the difficulties to fully translate slice experiments to human conditions. Moreover, factors other than BDNF may contribute to the augmenting effect of DCS on synaptic strength on one hand and motor learning on the other hand. Future work will be needed to investigate additional mechanisms, e.g. release of other neuromodulators and the cascades and messenger systems that they activate.
We propose that DCS could serve as an experimental tool to induce synaptic plasticity in vitro in brain regions that do not respond to conventional protocols. From a clinical perspective, our results may have implications for the treatment of conditions marked by impaired long-term synaptic plasticity and/or reduced cortical BDNF levels, e.g. neurodegenerative diseases (see Mattson 2008 for review). Understanding the cellular and molecular basis of therapeutic tools such as tDCS and their interaction with individual genetic predispositions could contribute to an advanced development of more individualized treatments.
Detailed supplementary methods accompany this paper online.
Experiments were approved by the NINDS Animal Care and Use Committee, in full compliance with the Guide for Care and Use of Laboratory Animals of the National Research Council.
Male 6 to 8 weeks old mice were anesthetized with CO2 and decapitated. Brains were quickly removed and placed in ice-cold buffer containing (in mM): NaCl 125, KCl 2.5, NaH2PO4 1.25, NaHCO3 25, CaCl2 1, MgCl2 2, and glucose 11. Coronal slices (thickness 400 μm) were cut with a Vibratome 1000 (The Vibratome Company, St. Louis, MO, USA) and incubated in an interface recording chamber perfused with carbogenated (CO2 95%, O2 5%; pH 7.4) artificial cerebrospinal fluid (aCSF) containing (in mM): NaCl 125, KCl 1.75, NaH2PO4 1.25, NaHCO3 25, CaCl2 2, MgCl2 1, and glucose 11 for at least 1 hour at 33–34° C before recording. Data were collected from slices at 33–34° C using an Axoclamp 200A amplifier and pClamp 9.2 software (Axon Instruments, via Molecular Devices, Toronto, Canada). Field EPSPs were recorded from mouse motor cortex slices.
Commercially available human tDCS electrodes (Amrex-Zetron Inc., Carson, CA, USA), modified in size, were used for DC field application to the slices placed in an interface chamber (Warner Instruments, Hamden, CT, USA). The electrical field induced by steady direct current stimulation (Stimulator 2100, A-M Systems, Carlsborg, WO, USA) was mapped within the slice using an Axoclamp 2A amplifier (Molecular Devices, Toronto, Canada). One electrode was kept in a fixed position as a reference. A second electrode was moved through the field in defined steps using a micromanipulator. Based on the field measurements (Fig. 1A), a stimulation intensity of 10 μA was chosen for all DCS applications (induced field strength ~ 0.75 mV/mm).
Behavioural and genotyping studies were approved by the local Institutional Review Boards and in compliance with the declaration of Helsinki. All subjects gave written informed consent to participate in the study. Healthy subjects practiced a sequential visual isometric pinch force task (Suppl. Fig. S3A) which entailed the modulation of pinch force exerted onto a force transducer to move a horizontal screen cursor quickly and accurately between a start position and different target gates (Suppl. Fig. S3B). The skill measure per training block was calculated in each subject based on the bivariate observation of movement time and error rate using a mathematical model described previously (Reis et al., 2009). Anodal tDCS (current density 0.04 mA/cm2; total charge 0.048 C/cm2) or sham tDCS was applied daily for 20 minutes during training, targeting the left M1 hand knob. The cathode was placed over the contralateral forehead. Please refer to the Suppl. Methods for a detailed description of training and tDCS. After testing data distributions for normality (Kolmogorov-Smirnov test), group differences in skill improvement were assessed by a factorial 2×2 ANOVA (factors GENOTYPE (Met carrier vs. Val/Val) and STIMULATION (Anodal tDCS vs. Sham tDCS)) and two-tailed t-test comparisons with Bonferroni adjustment.
A four-lane accelerating rotarod (rod diameter: 3 cm; AccuRotor, AccuScan Instruments Inc., Columbus, OH) was used to assess motor skill acquisition in BDNFMet/Met mice, BDNFflox/flox, cre mice and the corresponding wildtype littermates. Velocity of the rod was set to increase from 0 to 40 rpm in 5 min. Mice were placed in separate compartments. After familiarization, the latency to fall off the rod was automatically recorded by light beam breaks for six consecutive trials daily. Rotarod data were analyzed by one-way ANOVA.
This research was supported by the NINDS Intramural Research Program, NIH (BF, JR, HMS) and the Alexander von Humboldt Foundation Germany (JR). We thank JW Krakauer and PA Celnik for helpful comments and for their involvement in the previous characterization of the motor task utilized. We thank F Lee, A Morozov and D Ginty (transgenic animals), D Ide and A Mitz (field measurements), and D Brandsch and K Spies (data analysis). All authors disclose any financial conflict of interest.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.