The role of neuronal variability in cognitive modulation via prefrontal direct current stimulation Free

This scientific commentary refers to ‘Direct current stimulation modulates prefrontal cell activity and behaviour without inducing seizure-like firing’ by Fehring et al. (https://doi.org/10.1093/brain/awae273).

Soon after humans invented the battery, we tried using it to stimulate the brain. In its modern form, mild electric current delivered in the form of transcranial direct current stimulation (tDCS) is an exciting tool for neuromodulation. Affordable, easy to use and (relatively) painless,¹ it offers a promising path to precision medicine. However, recent years have seen some waning enthusiasm around tDCS on account of reports suggesting that its effects on brain and behaviour may be smaller and less reliable than originally thought. In this issue of Brain, Fehring and colleagues² interrogate the mechanisms of tDCS in primate prefrontal cortex, providing a new impetus for future tDCS work and a novel framework for re-evaluating past findings.

Before delving into Fehring and colleagues’ study,² we must first consider some foundational ideas about the mechanism of action of tDCS. The procedure involves passing a mild electric current between one or more anodal electrodes and one or more cathodes, placed on the scalp (or extra-cephalically). Initial insights into the mechanism of tDCS in humans came from seminal studies on primary motor cortex (M1) and the corticospinal tract conducted by Nitsche, Paulus and colleagues.³ They found that placing an anode roughly over the ‘hand knob’ in area M1 increased corticospinal excitability (CSE) measured at the hand muscles, whereas placing a cathode over M1 decreased CSE.³ These effects lingered for 10–30 min after stimulation and were thought to be mediated by long-term potentiation and long-term depression-like changes in the resting potential of neuronal populations. This mechanistic model led to the assumption that the functional effects of tDCS are primarily driven by polarity-specific changes in cortical excitability.

Yet, an account based on polarity-specific changes in excitability becomes less viable when tDCS is used to target complex neurocomputational mechanisms outside of M1 and the corticospinal tract. Consider the large—and particularly inconsistent—corpus of studies examining cognitive neuromodulation through prefrontal tDCS.⁴ To date, many studies have found significant changes in behavioural assays of attention, memory and executive control after targeting the prefrontal cortex with tDCS.⁵ However, the field is also replete with failures to replicate and, in general, the effects of prefrontal cortex tDCS would appear to be less reliable and weaker than originally believed.⁶

This is where Fehring and colleagues’ work² comes in. Rather than focusing solely on humans, they examined the effects of tDCS in the prefrontal cortex of two Old World monkeys (Japanese macaques; Fig. 1A). By using this non-human primate model, they were able to perform invasive recordings at the primary target region—macaque dorsolateral prefrontal cortex (DLPFC)—while the animals performed a prefrontal short-term memory task (delayed match-to-sample, DMS) (Fig. 1B), with direct translational relevance to humans. The results indicate a more complex mechanism of action for cognitive neuromodulation via prefrontal tDCS than the classic polarity-dependent effects documented in M1.

Figure 1

Cognitive and neuronal effects of prefrontal transcranial electrical stimulation in primates. (A) Schematic of the non-human primate (NHP) transcranial direct current stimulation (tDCS) montage and recording apparatus. (B) Schematic of the delayed match-to-sample (DMS) task used in the study. The task was performed similarly by humans and NHPs, except that humans demonstrated generally superior performance in colour-matching compared to shape-matching, whereas NHPs showed the opposite pattern. (C) In the NHPs, behaviour was modulated by active relative to sham tDCS, with sham sessions showing pre- to post-stimulation response time (RT) slowing that was abolished by active tDCS. Active tDCS also caused a reduction in the coefficient of variability (CoV) in the difficult (for the NHP) colour DMS [CoV(c)] and an increase in CoV in the easier shape DMS [CoV(s)]. (D) Neuronal recordings revealed an increase in firing rate from pre- to post-stimulation in the sham sessions, but this effect was abolished by active tDCS. Notably, both neuronal variability (indexed by the Fano Factor) and task-related activity in units that encoded correct responses at the time of the decision were improved by active tDCS. The analogous behavioural (CoV) and neuronal (Fano Factor) indices of reduced variability may suggest a potential mechanism through which anodal tDCS facilitates more reliable and enhanced performance of prefrontally-mediated tasks. ACC = anterior cingulate cortex.

Open in new tabDownload slide

Fehring and colleagues² found two clear behavioural indicators of modulatory effects on cognition following active tDCS. First, the monkeys demonstrated significant pre- to post-stimulation response slowing, but this effect was attenuated after active but not sham tDCS to DLPFC (Fig. 1C). Additionally, the coefficient of variation in response times (CoV)—thought to reflect executive control fluctuations during the task—was modulated in a task-specific manner with active but not sham tDCS. Specifically, the CoV decreased post-stimulation for the colour variant of the DMS but increased for the shape variant (Fig. 1C). Across all sessions, the CoV was higher and mean response latencies were longer in the colour relative to the shape dimension, suggesting that the colour task may be more difficult for the animals. Previous studies in humans have suggested that tasks with a higher difficulty level are particularly likely to benefit from tDCS,⁷ whereas tasks where performance is close to ceiling at baseline tend to show smaller cognitive modulation effects. This may provide a potential explanation for the task-specific modulation of CoV by prefrontal tDCS in macaques.

Single-cell recordings revealed three intriguing effects (Fig. 1D) that move beyond a polarity-specific account of the mechanisms of action of tDCS. First, while the animals showed within-session increases in DLPFC neuronal activity during sham sessions, this increase was attenuated by active tDCS. Second, the frequency of DLPFC units encoding correct task responses in the decision epoch was increased during active but not sham stimulation. Last, active stimulation reduced the variability of DLPFC neuronal responses, quantified using the Fano Factor index (i.e. mean divided by variance in unit firing rate). Reduced neuronal variability is thought to reflect improved temporal coding in distributed neural circuits⁸ and reduced fluctuations in executive control.⁹ This maps neatly onto the observed task-dependent changes in CoV—suggesting the cognitive benefits of prefrontal tDCS may be driven by reductions in neuronal variability, and increased neuronal gain in prefrontal circuits that regulate the executive control of behaviour. Importantly, the behavioural and neuronal effects of tDCS occurred without increased epileptiform burst-like firing, which is critical for ensuring the safety of prefrontal tDCS in clinical settings.

While this study provides unprecedented insights into cognitive modulation via prefrontal tDCS, several questions remain. For instance, tDCS with a cathodal electrode over the primary target region has produced inconsistent and unreliable cognitive effects. It would therefore be of interest to examine the effects of cathodal tDCS on invasive neuronal recordings in primates. In addition, while local effects of anodal tDCS on prefrontal unit responses and variability are exciting, simulated computational models of electric current flow from tDCS have long suggested the presence of electrical fields far away from the stimulating electrodes.¹⁰ Future studies should therefore include invasive recordings from sites beyond the tDCS electrode, to determine whether the observed effects on firing rate and neuronal variability are local, or occur at the level of distributed task-related brain networks.

Funding

While conducting this work, J.H. was supported by the National Institute on Alcohol Abuse and Alcoholism (R01AA030283), and the National Science Foundation (BCS2237795).

Competing interests

The author reports no competing interests.

References