Causal Evidence for Prefrontal-Motor Coupling in Reward-Responsive Goal-Directed Behavior
- Justin Riddle1,2,3,4,
- Lewis P. Edwards3,4,
- Moria Smoski5,
- Crystal E. Schiller3,
- David R. Rubinow3, and
- Flavio Frohlich3,4,6,7,8,9
- 1Department of Psychology, Florida State University, Tallahassee, Florida 32304
- 2Program in Neuroscience, Florida State University, Tallahassee, Florida 32304
- 3Department of Psychiatry, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina 27599
- 4Carolina Center for Neurostimulation, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina 27599
- 5Department of Psychiatry and Behavioral Sciences, Duke University, Durham, North Carolina 27708
- 6Department of Neurology, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina 27599
- 7Department of Cell Biology and Physiology, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina 27599
- 8Department of Biomedical Engineering, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina 27599
- 9Neuroscience Center, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina 27599
-
Author contributions: J.R., M.S., D.R.R., and F.F. designed research; J.R., L.P.E., C.E.S., and D.R.R. performed research; J.R. analyzed data; J.R. wrote the initial paper. J.R. and F.F. edited the paper.
Abstract
Interventions for major depressive disorder (MDD) using noninvasive brain stimulation often target the left lateral prefrontal cortex to reduce symptoms of anhedonia. However, the electrophysiological mechanism by which symptoms of anhedonia arise from dysfunction of the lateral prefrontal cortex is poorly understood. Furthermore, multiple constructs related to reward processing within the positive valence systems have been found to be blunted with anhedonia. To disentangle components of the positive valence systems, human participants with an active episode of MDD performed an adaptive version of the expenditure of effort for reward task (Adaptive-EEfRT) and three constructs of reward processing were investigated: effort motivation, reward valuation, and reward-responsive goal-directed behavior. Individual differences analysis revealed that symptoms of anhedonia were most strongly related to blunted reward-responsive goal-directed behavior. By recording electroencephalography (EEG), spectral analysis revealed that the amplitude of left prefrontal delta oscillations (2–3 Hz) increased with each reward construct and beta oscillations decreased over the motor cortex contralateral to the hand used for effort exertion. Phase–amplitude delta-beta coupling between the left prefrontal cortex and contralateral motor cortex was significantly increased for each reward construct. When cross-frequency transcranial alternating current stimulation (tACS) was delivered to 26 females and 9 males to mimic prefrontal-motor delta-beta coupling, only reward-responsive goal-directed behavior was improved relative to placebo and active-control tACS. These findings provide causal evidence that the left prefrontal cortex supports reward-responsive goal-directed behavior via cross-frequency coupling. Interventions for anhedonia that target the left prefrontal cortex might be improved by frequency-specific targeting of delta-beta coupling.
- anhedonia
- cross-frequency coupling
- decision-making
- electroencephalography
- major depressive disorder
- transcranial alternating current stimulation
Significance Statement
Anhedonia remains a difficult-to-treat symptom of major depressive disorder (MDD). A better understanding of the neural correlates of anhedonia and the ability to modulate relevant neural circuits will provide an avenue for next-generation interventions. Here, patients with MDD performed a reward-based decision-making task while electroencephalography was recorded. Anhedonia severity was related to reduced reward-responsive goal-directed behavior and reward-responsive goal-directed behavior recruited prefrontal-motor cross-frequency coupling. Noninvasive electrical stimulation was delivered targeting this activity, which successfully engaged the target and increased reward-responsive goal-directed behavior. These findings provide causal evidence that prefrontal-motor coupling supports reward-responsive goal-directed behavior in patients with MDD. The circuit described here is a novel mechanism for translating goals into action that could be targeted by future interventions for anhedonia.
Introduction
Anhedonia, the inability to seek out and experience pleasure, is a core symptom of major depressive disorder (MDD) that remains difficult to treat with many pharmacological interventions occasionally increasing, or introducing, anhedonia (Treadway and Zald, 2011a; Husain and Roiser, 2018; Lambert et al., 2018). While there has been recent success using noninvasive brain stimulation to target the left lateral prefrontal cortex, the mechanism of action for treatment efficacy is currently unknown. A better understanding of the neuropsychological basis for anhedonia symptoms in MDD could lead to the refinement of treatment targets for noninvasive brain stimulation. Furthermore, the impact of noninvasive brain stimulation on neural activity is known to depend on endogenous oscillatory activity within the targeted region (Romei et al., 2016; Riddle and Frohlich, 2021; Wischnewski et al., 2023). When stimulation is aligned with intrinsic neural dynamics, cognition associated with that activity can be enhanced, whereas misalignment may lead to null or detrimental effects (Riddle et al., 2020a; Riddle et al., 2024). Thus, identifying electrophysiological mechanisms by which the left lateral prefrontal cortex supports hedonic and motivational processes is critical for guiding next-generation stimulation approaches that combine temporal and spatial precision.
Symptoms of anhedonia reflect disruptions across multiple reward processing constructs (Treadway and Zald, 2011b; Nusslock and Alloy, 2017), formalized within the Research Domain Criteria (RDoC) Positive Valence Systems (Insel et al., 2010). First, anhedonia is associated with reduced reward responsiveness, as indexed by blunted reward-related positivity (RewP) responses to positive outcomes (Proudfit, 2015). Here, we investigated reward-responsive goal-directed behavior, defined as the impact of reward outcome on subsequent decision-making. Second, people with anhedonia show reduced willingness to exert effort for uncertain rewards, reflecting impairments in action-oriented motivational systems (Treadway et al., 2009; Treadway and Zald, 2011a; Treadway et al., 2012). Third, reduced pleasure from reward can lead to reward devaluation that subsequently reduces reward learning (Vrieze et al., 2013; Kangas et al., 2022). Here, participants with active-episode MDD performed an adaptive version of the expenditure of effort for reward task (Adaptive-EEfRT; Treadway et al., 2009; Treadway et al., 2012; Riddle et al., 2022). The Adaptive-EEfRT dynamically adjusted effort demands to push participants toward their decision indifference point, enabling sensitive measurement of multiple reward constructs. Effort motivation was defined as the proportion of high-effort choices, reward valuation as sensitivity to incentive magnitude, and reward-responsive goal-directed behavior as changes in effort following reward gain versus loss outcomes. EEG was acquired during performance of the Adaptive-EEfRT to investigate the presence of delta-beta coupling and its relationship to these behavioral measures.
A candidate electrophysiological mechanism to target for anhedonia treatment is cross-frequency coupling between low-frequency delta oscillations in the left lateral prefrontal cortex and high-frequency beta oscillations in the motor cortex (Riddle et al., 2022). Delta oscillations (2–3 Hz) have been associated with evidence accumulation during decision-making, while beta oscillations (15–25 Hz) in motor cortex index motor preparation (Tzagarakis et al., 2010; Wyart et al., 2012; Muralidharan et al., 2019; Muralidharan et al., 2022). Coupling between delta and beta oscillations is therefore thought to represent the coordination of a motor response based on contextual information (Wyart et al., 2012; Riddle et al., 2020b) and was recently observed during reward-based decision-making and shown to be reduced in people with anhedonia (Riddle et al., 2022). To establish causal evidence (Bergmann and Hartwigsen, 2021), cross-frequency transcranial alternating current stimulation (tACS) was delivered using a delta coupled to beta electrical waveform as participants performed the Adaptive-EEfRT. tACS was delivered in an active control pattern, theta-gamma, to control for frequency-specificity along with a placebo form of tACS. The impact of delta-beta tACS on reward processing constructs was investigated, and resting-state EEG was acquired following each tACS block to capture the aftereffects of tACS on delta-beta coupling.
Materials and Methods
Experimental design
This study was approved by the institutional review board at the University of North Carolina at Chapel Hill and was preregistered on ClinicalTrials.gov (NCT05084924). Participants were recruited from the community in the Chapel Hill, Durham, and Raleigh area for a placebo-controlled parallel-arm randomized clinical trial. Flyers indicated that the study was recruiting those currently experiencing symptoms of depression. In the first session of the three-session experiment, clinical interviews were conducted to determine eligibility. The second experimental session comprised two phases. In the first phase, baseline EEG was acquired during performance of four blocks of the adaptive version of the expenditure of effort for reward task (Adaptive-EEfRT). After each block, the level of effort exertion required per trial was escalated or decremented based on performance. In the second phase, participants were randomized to receive active delta-beta cross-frequency tACS, active-control cross-frequency tACS with a theta-gamma waveform, or placebo tACS. Eight blocks of tACS were delivered concurrent with performance of the Adaptive-EEfRT. Resting-state EEG without tACS was acquired following each task block to assess target engagement. The third experimental session was a functional MRI session the result of which are not reported here.
The preregistered goal of the study was to deliver tACS to 48 participants. The first participant was enrolled in November
2021, and the study was truncated in July 2023 with 35 participants when the grant period funding the study ended (K99MH126161
to J.R.). With unblinding, it was revealed that 11 participants received delta-beta tACS, 12 received theta-gamma tACS, and
12 received sham tACS. There were no changes to the experimental design after initial enrollment. The sample size of 48 participants
was determined based on a power analysis conducted using the effect size from a previous study that utilized a similar tACS
approach to investigate the impact of delta-beta tACS (versus theta-gamma tACS and placebo tACS) on performance on a hierarchical
cognitive control task in a crossover design (Riddle et al., 2021). In this previous study with 23 participants, an effect size of 0.29 <mml:math display="inline"><mml:mo stretchy="false">(</mml:mo><mml:msubsup><mml:mi>η</mml:mi><mml:mrow><mml:mi mathvariant="normal">p</mml:mi></mml:mrow><mml:mn>2</mml:mn></mml:msubsup><mml:mo
stretchy="false">)</mml:mo></mml:math>(ηp2)
for the interaction effect and a Cohen’s d of 0.48 for the change in performance relative to placebo was reported. Conservatively estimating a correlation between variables
of 0.5 and modeling an interaction in an analysis of variance (ANOVA) with between-participant (intervention arms) and within-participant
(baseline versus tACS) factors, the final sample size of 35 participants was powered at >80% statistical power for the primary
outcome of changing effort exertion. The secondary outcome was the strength of delta-beta cross-frequency coupling. A previous
study used a similar approach to investigate the impact of delta-beta tACS on delta-beta coupling and found an effect size
of 0.38 (Riddle et al., 2021). With 35 participants, the secondary outcome was powered at >95%. While this previous study utilized a crossover design
and the present study employed a parallel arm design, the baseline period of the present study was leveraged as a within-participant
factor to control for individual differences of noninterest.
After initial contact with research staff, phone screening was conducted on 124 participants. During phone screening, preliminary eligibility was assessed: age of 18 to 65, normal or corrected-to-normal vision, no contraindications for tACS or MRI, and self-report of active depression symptoms. Contraindications that warranted exclusion included neurological disorders and current prescriptions of stimulants or benzodiazepines. Following phone screening, 81 participants consented to and completed the 8-item version of the Patient Health Questionnaire (PHQ-8; Kroenke et al., 2001) and the Snaith Hamilton Pleasure Scale (SHAPS; Snaith et al., 1995). If a participant scored 8 or higher on the PHQ-8, then they were invited to complete the first session of the experiment. This criterion was included so that participants were more likely to receive a diagnosis of a current episode of MDD in the first session. In addition, enrollment was stratified such that 50% of participants would be greater than 33 on the SHAPS, the mean SHAPS from a similar population in a previous study (Riddle et al., 2022). Stratification ensured a balanced distribution of anhedonia symptoms by tACS condition. Sixty-eight participants completed the clinical assessment session in which the Columbia Suicide Severity Rating Scale (C-SSRS; Posner et al., 2011), the Hamilton Depression Rating Scale (HAM-D; Hamilton, 1960), the Mini International Neuropsychiatric Interview for the Diagnostic and Statistical Manual version 5 (MINI; Sheehan et al., 1998), and the Maudsley Staging Method (MSM; Fekadu et al., 2009) were administered. Inclusion criteria were a diagnosis of a current episode of MDD. Exclusion criteria were symptoms of psychosis in the past 12 months, lifetime severe alcohol/substance use disorder (A/SUD), current moderate A/SUD, and moderate suicidal risk (intent on the C-SSRS and a score of 3 or greater on the suicide item of the HAM-D). If participants reported suicidal ideation, then a suicide risk assessment was conducted. In addition, participants completed self-report assessments: Behavioral Activation System and Behavioral Inhibition System (BIS/BAS; Carver and White, 1994), Temporal Experience of Pleasure Scale (TEPS; Gard et al., 2006), the State-Trait Anxiety Inventory (STAI; Spielberger, 2010), Beck's Depression Inventory version 2 (BDI-2; Beck and Beamesderfer, 1974), and the Ruminative Responses Scale (RRS; Roelofs et al., 2006). Thirty-seven participants qualified for the study and 35 completed the concurrent tACS and EEG session (26 females, 9 males, ages 18 to 59, mean ± SD 28.9 ± 12.9; Table 1).
During the second experimental session, participants were randomized into one of the three arms of the study and assigned the next available blinding number. Participants completed the SHAPS and STAI and were fitted with tACS and EEG. Four baseline blocks of the Adaptive-EEfRT without tACS were acquired followed by eight blocks concurrent with tACS. After every task block, 2 min of eyes-open resting state was acquired. At the end of the session, participants completed a stimulation side effects questionnaire. If a participant reported a 4 for a side effect on a Likert scale of 0 to 4, then this was defined as an adverse event, and an adverse event structured interview was performed to assess the time course and resolution of the symptom or symptoms.
Adaptive-EEfRT
Participants completed a version of the expenditure of effort for reward task (EEfRT) that was designed for EEG analysis and was described in a previous publication (Riddle et al., 2022). The task was presented using Psychtoolbox-3 for MATLAB (Kleiner et al., 2007). Participants were seated in a comfortable chair with their head on a chin rest that maintained a fixed distance from the monitor. In the EEfRT, participants were faced with a decision to perform a less physically demanding task for a chance to win a low and fixed amount of money or to complete a more physically demanding task for a chance to win more money (Fig. 1A). The easy task provided a fixed incentive at $1, and the hard task was variable from $2.50 to $6.00 in $0.50 increments. The probability of obtaining the reward was 50% after successful completion of the effort expenditure. Effort motivation was defined as the percentage of trials in which the participant chose to perform the hard task. After their decision, participants pressed a button multiple times with their index finger (easy task) or their pinky finger (hard task). Fifteen-pound ankle weights were placed over the forearm and wrist to remind participants not to use their arm muscles to press the buttons. To titrate effort expenditure, participants first pressed a button as fast as possible with their dominant index finger for 7 s and the number of presses for the last 5 s was recorded. The initial titration values of maximum press rate were 27.1 (4.0 SD, 16 to 38). Forty percent of the titrated value was set as easy and 70% for hard. In the Adaptive version of the EEfRT (Adaptive-EEfRT), effort exertion was further ensured to remain close to the indifference point of the individual by increasing or decreasing the difficulty of effort expenditure after each of the four baseline blocks (20 trials each). If a single hard task was successfully completed in the block, then the difficulty was increased by three button presses for hard and one for easy; otherwise, a decrement was performed. Following the baseline of four 5 min blocks, task difficulty was fixed for the eight 5 min stimulation blocks. Participants were not informed about the staircase procedure for task difficulty, nor was their knowledge about this manipulation assessed. The final number of required button presses for the hard task was 30.3 (4.9 SD). In total, the participant completed 12 consecutive blocks (60 min of task) in ∼90 min for the full EEG-tACS session. With this prolonged task demand, participants adapted their decision-making based on their depleted stamina and motivation. In each block, the participant performed the task with either their left or right hand for the entirety of the block. Use of the left or right hand was indicated at the start of each block and the number of blocks for each hand was randomized and balanced. In summary, the Adaptive-EEfRT differs from previous versions in that the difficulty of the task rapidly escalated and physical exertion was requested for over an hour.
For behavioral analysis, metrics were designed to quantify subdimensions of the Positive Valence Systems of the Research Domain Criteria (RDoC). “Effort motivation” was defined as the percentage of trials in which the participant chose to exert more physical effort, which corresponded to the Effort subconstruct within the Reward Valuation construct of the RDoC. Second, trials were sorted by the reward outcome of the previous trial, and “reward-responsive goal-directed behavior” was defined as the difference in percentage of trials in which the hard task was selected following a win versus following a loss. This metric corresponded to the reward-responsive goal-directed behavior construct of the RDoC. Increased reward-responsive goal-directed behavior reflected an increased rate of choosing hard after winning a reward relative to trials following a loss. Third, “reward valuation” was defined as the influence of the offered incentive on willingness to exert effort. For the higher incentive ($4.50–$6.00) relative to the lower incentives ($2.50–$4.00), we calculated the difference in choosing the high effort exertion per incentive dollar. This metric corresponded to the Reward Valuation construct within the Positive Valence Systems of the RDoC.
High-density electroencephalography
Electroencephalography (EEG) was acquired at 1,000 Hz using a 128-channel elastic net (HydroCel Geodesic Net) with an amplifier (Net Amps 400; Electrical Geodesics). The reference for recording was over Cz and the ground was near POz. The EEG net was placed over the stimulation pads for tACS, and the reference electrode was ensured not to be electrically bridged with the stimulation pads. A photodiode was placed on the screen to synchronize the 60 Hz monitor to the EEG. The baseline Adaptive-EEfRT data and the post-tACS eyes-open resting-state data were preprocessed using the same pipeline, except when noted. The data were filtered between 1 and 58 Hz and downsampled to 200 Hz. Task data were epoched from −2 to 3 s relative to the onset of the decision, and rest data were epoched into 2 s intervals for manual rejection. Epoch rejection was only applied for rare instances when the participant physically adjusted the EEG net or when tACS artifacts persisted into the 2 min recording (task: mean of 0.5, SD of 1.0, range of 0 to 4; rest: mean of 10.9, SD of 11.7, range of 0 to 52). One rest-block was rejected entirely because of a technical issue with the hardware. EEG channels that were near the border of tACS pads were not filled with gel to prevent electrical bridging. Channels were manually flagged for removal with assistance from a custom-built graphical user interface. Artifact subspace reconstruction flagged channels for removal. Spherical interpolation was applied to replace rejected channels (rejected task data: mean of 15.4, SD of 5.0, range of 5 to 24; and rest: mean of 18.8, SD of 5.5, range of 9 to 31). Data were referenced to the global average and independent component analysis was run. Components that corresponded to muscles, eyes, channel noise, or tACS artifacts were removed (task: mean of 25.5, SD of 9.7, range of 9 to 45; rest: mean of 39.7, SD of 7.4, range of 27 to 58). The final analysis was restricted to the 90 channels on or above the 10–20 circumference line.
Time–frequency analysis
Morlet wavelet convolution was performed to localize the electrodes with peak amplitude of low-frequency delta oscillations in the prefrontal cortex and beta-frequency oscillations over the motor cortex. Five-cycle wavelets were used to estimate instantaneous amplitude for frequencies in a logarithmic base 0.05 distribution, selected to approximate the power distribution of the brain (Voytek et al., 2015). First, spectrograms were generated from a priori electrode clusters for the left prefrontal cortex, F3 and surrounding electrodes, and the left motor cortex, C3 and surrounding electrodes. Baseline correction was applied using data before trial onset. Delta (2–3 Hz) and beta (15–25 Hz) amplitude were estimated using band-filtering and the Hilbert transform for each electrode. Mean signal from 100 to 1,000 ms was extracted based on the significant delta activation in the prefrontal cortex and beta decrease in the motor cortex from analysis of the spectrograms. Three contrasts of interest were used that corresponded to the RDoC constructs of interest: effort motivation was trials when high effort exertion was chosen versus low effort exertion; reward-responsive goal-directed behavior was post-win versus post-lose; reward valuation was high incentive versus low incentive. Spatial normalization was performed prior to contrasts using z-normalization. To control for the potential explanation of oscillatory effects as spectral components of the event-related potential (ERP), the ERP was calculated for each condition. Wavelet convolution was run on the ERP for each condition and tested across participants using permutation-based cluster correction.
Phase–amplitude coupling analysis
Phase–amplitude coupling (PAC) was calculated between the phase of delta oscillations over the left prefrontal cortex and
the amplitude of beta oscillations over motor cortex using mean vector length (Eq. 1). A bootstrapped null distribution was calculated by randomly shifting the amplitude time series 1,000 times to estimate
PACZ. PAC was estimated for the Adaptive-EEfRT (100–1,000 ms of the decision epoch) and the post-tACS resting-state recording.
Topographic analysis was performed by calculating PACZ from the left prefrontal site to each electrode across the scalp. Beta amplitude showed greater modulation over the motor
cortex corresponding to the hand used for decision-making (left motor cortex for right hand blocks and vice versa). Thus,
the beta amplitude topography was left–right flipped for left-handed blocks. The data were spatially normalized, and the contrasts
for effort motivation, reward-responsive goal-directed behavior, and reward valuation were calculated:<mml:math display="block"><mml:mrow><mml:mi mathvariant="normal">PAC</mml:mi></mml:mrow><mml:mo>=</mml:mo><mml:mrow><mml:mo>|</mml:mo><mml:mrow><mml:munderover><mml:mo
movablelimits="false">∑</mml:mo><mml:mrow><mml:mi>t</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mi>N</mml:mi></mml:munderover><mml:mrow><mml:mrow><mml:mi
mathvariant="normal">Am</mml:mi></mml:mrow><mml:msub><mml:mrow><mml:mi mathvariant="normal">p</mml:mi></mml:mrow><mml:mi>β</mml:mi></mml:msub><mml:mspace
width=".1em"></mml:mspace><mml:mo>*</mml:mo><mml:mspace width=".1em"></mml:mspace><mml:msup><mml:mi>e</mml:mi><mml:mrow><mml:mi>i</mml:mi><mml:mspace
width=".1em"></mml:mspace><mml:mo>*</mml:mo><mml:mspace width=".1em"></mml:mspace><mml:mrow><mml:mi mathvariant="normal">Phas</mml:mi></mml:mrow><mml:msub><mml:mrow><mml:mi
mathvariant="normal">e</mml:mi></mml:mrow><mml:mi mathvariant="normal">Δ</mml:mi></mml:msub></mml:mrow></mml:msup><mml:mo>/</mml:mo><mml:mi>N</mml:mi></mml:mrow></mml:mrow><mml:mo>|</mml:mo></mml:mrow><mml:mo>.</mml:mo><mml:mspace
width="2em"></mml:mspace><mml:mo stretchy="false">(</mml:mo><mml:mn>1</mml:mn><mml:mo stretchy="false">)</mml:mo></mml:math>PAC=|∑t=1NAmpβ*ei*PhaseΔ/N|.(1)
Cross-frequency transcranial alternating current stimulation
Cross-frequency tACS was delivered using two single-channel neuroConn DC Stimulator Plus devices (NeuroCare Group) with a custom-built cable to splice the cathode of both devices into a single point of current delivery. The stimulators were controlled via MATLAB (MathWorks) using a USB-controlled DAQ (National Instruments) that provided identical waveforms to both devices (Fig. 5A). Randomization by stratification was achieved by random codes that were generated by a researcher not associated with the study. These codes were maintained in a binary file that was accessed by MATLAB code, which ran the computerized task and generated the waveforms for tACS. There were two spreadsheets with six-digit arbitrary codes, one for high anhedonia and one for low anhedonia. Each code mapped to one of the three types of tACS. Participants and experimenters were blinded to the type of tACS delivered.
Two 4.5-by-4.5 cm carbon rubber pads received synchronized current with a 5-by-7 cm return electrode as described previously (Riddle et al., 2021). The tACS pads were applied to the scalp using Ten20 electroconductive paste and impedance was ensured to be under 10 kOhm for each stimulation channel. First, the scalp was marked using a washable pen according to the 10–20 international coordinate system. F3 was located using the BeamF3 algorithm (Mir-Moghtadaei et al., 2015). One target tACS pad was placed just anterior and ventral to F3, centered on AFF5h of the 10–10 system, and the return was centered over FCz (Fig. 5A). The second tACS target pad was approximately posterior to C3, centered on CPP3 of the 10–10 system which was chosen to be equidistant to the return pad [see Riddle et al. (2021) for a similar montage]. The normalized electric field was calculated using the ROAST toolbox (Huang et al., 2018). At a threshold of 0.25 V/m, the predicted neural effect was expected in the left anterior middle frontal gyrus and left motor cortex (Fig. 5B). The left hemisphere was targeted because the left lateral prefrontal cortex is frequently targeted using transcranial magnetic stimulation for the treatment of depression and a meta-analysis revealed that the lateral site was particularly effective at reducing symptoms of anhedonia (Siddiqi et al., 2020).
While tACS was delivered, it was difficult to recover the EEG signal in the targeted frequency bands to assess target engagement [but see (Helfrich et al., 2014; Fehér and Morishima, 2016)]. Stimulation was ramped down at the end of each task block, and 2 min of resting-state EEG was acquired to estimate the residual impact of tACS on neural activity (Riddle et al., 2021). The cross-frequency electrical waveform was 2.5 Hz low-frequency and 20 Hz high-frequency for delta-beta and 6 Hz low with 50 Hz high for theta-gamma (Riddle et al., 2022). The low-frequency component was delivered at 0.6 mA and the high-frequency component at 0.4 mA for a peak current delivery of 1.0 mA per tACS channel. The high-frequency component was 3.5 cycles centered at the peak of the low-frequency component. The sizes of the tACS pads were selected to approximately match the delivered current density: 0.049 mA/cm2 to the target pads and 0.057 mA/cm2 to the return pad. While tACS was delivered the EEG recording computer and the tACS devices were covered by an opaque sheet to blind the experimenters (Riddle et al., 2021). For placebo tACS, electrical stimulation was ramped up, maintained for 30 s, and then ramped back down. This sham-tACS procedure mimics the sensation of adaptation from the skin exposed to an electric current. Furthermore, the use of an active control with a cross-frequency waveform provides the means for assessing the frequency-specificity of tACS irrespective of placebo. Side effects of receiving delta-beta versus theta-gamma tACS were matched in a previous experiment (Riddle et al., 2021).
Statistical analysis
Reward-responsive goal-directed behavior (post-win versus post-lose) and reward valuation (high versus low incentive) were confirmed with paired t tests. A linear mixed effects model was run for each behavioral metric, effort motivation, reward-responsive goal-directed behavior, and reward valuation, using symptom severity as independent variables: depression severity (HAM-D), consummatory anhedonia (inverse TEPS-consummatory), anticipatory anhedonia (inverse TEPS-anticipatory), anxiety (STAI-trait), and general anhedonia (SHAPS). Post hoc correlation analysis was run using Pearson’s correlation. Overall depression severity was used in a partial correlation to assess symptom specificity. As a post hoc analysis, we conducted a linear model as described above but that included the potential confounding influence of exhaustion. Exhaustion was modeled as a within-participant effect of time on task using a median split on trial count. Results are reported as a follow-up to significant findings from the initial analyses and for main effects on constructs of reward.
Time–frequency analysis was performed on left prefrontal cortex and left motor cortex electrodes. Permutation-based cluster correction was performed to identify significant time–frequency activation across all trials. For visualization, follow-up topographic analysis was performed with a stringent alpha threshold (p < 10−6) for delta and beta amplitude from 100 to 1,000 ms. For the three contrasts of interest, effort, reward-responsive goal-directed behavior, and reward valuation the amplitude of delta was compared using paired t tests. Delta topography was tested using cluster correction for contiguous electrodes. Beta amplitude over left and right motor cortex was investigated as a function of the hand used for the block. An interaction was investigated using a two-way ANOVA with factors for hand (left- or right-handed block) and motor cortex (left or right). Delta-beta coupling was investigated using planned t tests for left prefrontal delta phase to left motor beta amplitude across conditions and for the reward-construct contrasts. Topographic analysis was run using cluster correction for contiguous electrodes.
The preregistered primary outcome was a change in effort motivation with delta-beta tACS. A two-way ANOVA was run on effort motivation with factors for tACS (delta-beta tACS, theta-gamma tACS, or sham tACS) and time (baseline and during tACS). Two-way ANOVAs were run for reward-responsive goal-directed behavior and reward valuation. For significant main effects or interactions, post hoc t tests were run. In a post hoc analysis, the analyses above were conducted with the inclusion of a within-participant factor for exhaustion which was estimated using a median split of trials (first half and second half). The reporting of these results was focused on interactions with tACS.
The preregistered secondary outcome was the change in delta-beta coupling from delta-beta tACS. A two-way ANOVA with a between-participants factor for tACS arm and a within-participants factor for the hand used for effort exertion was run. Topographic analysis was conducted using cluster-correction for contiguous electrodes. For a significant effect of tACS on behavior and neural activity, a mediation analysis was run to investigate whether the impact of tACS on behavior was influenced by a neural mediator.
To assess participant blinding, a chi-squared test was run with each arm as a different group to assess for differences in the number of participants that believed that verum tACS was delivered relative to placebo tACS. To assess experimenter blinding, a chi-squared test was run for the experimenters. The difference in side effects between tACS arms was investigated using independent t tests.
Results
Increased anhedonia symptoms were related to reduced reward-responsive goal-directed behavior
The Adaptive-EEfRT (Fig. 1A) was performed by participants with an active episode of MDD (Table 1). Behavior was quantified along three primary metrics: effort motivation, reward valuation, and reward-responsive goal-directed behavior (Table 2). Since the metric for reward-responsive goal-directed behavior was novel, the impact of reward outcome on willingness to exert effort in the subsequent trial was investigated. Overall, participants were near the indifference point of 50% (55.2% ± 23.9%). Participants were more likely to choose the task with higher physical exertion following a reward win (59.6% ± 25.2%) versus a loss (55.0% ± 25.8%; t(34) = 2.22, p = 0.033, d = 0.38; Fig. 1B). Additionally, they were more likely to exert more effort for higher incentives (64.4% ± 26.2%) relative to lower incentives (34.9% ± 29.5%; t(34) = 6.19, p < 0.001, d = 1.05; Fig. 1B). Linear mixed effects models were run using clinical metrics of overall depression severity, anhedonia symptoms, and anxiety symptoms. These models revealed that symptoms of consummatory anhedonia were uniquely related to reward-responsive goal-directed behavior, and there was no relationship with effort motivation or reward valuation (Table 3). Post hoc partial correlation analysis for anhedonia symptoms (sign-flipped TEPS-consummatory) using depression severity as a covariate revealed a negative relationship with reward-responsive goal-directed behavior (r(33) = −0.39, p = 0.02) such that greater symptoms of anhedonia predicted reduced sensitivity to the reward outcome of the previous trial (Fig. 1C). Despite its strong correlation at baseline with anticipatory anhedonia (r(33) = 0.54, p = 0.001), partial correlation revealed a trend-level relationship (r(33) = 0.32, p = 0.066). There was no relationship with effort motivation (r(33) = 0.33, p = 0.06; Fig. 1D) or reward valuation (r(33) = 0.01, p = 0.94; Fig. 1E).
Increased anhedonia symptoms were related to reduced reward-responsive goal-directed behavior (GDB) in the Adaptive Expenditure of Effort for Reward Task (Adaptive-EEfRT). A, The Adaptive-EEfRT comprises three epochs. First, the participant decided between HARD (variable between $2.50 and $6.00) and EASY (fixed at $1). Then, performed physical effort exertion with the corresponding finger. Third, the reward was provided with 50% probability of WIN versus LOSE. The Adaptive-EEfRT was titrated to physical capacity, task difficulty was rapidly escalated after each baseline block, and the task was performed for an extended period. B, Participants were overall responsive to the reward outcome of the previous trial and to the offered incentive. High incentive is $4.50–$6 and low is $2.50–$4. Error bars are SEM. Symptoms of consummatory anhedonia were negatively correlated to (C) reward-responsive GDB, but not (D) effort motivation or (E) reward valuation. Overall depression severity is depicted for comparison. Shaded region is 95% confidence interval. Consummatory anhedonia is the inverse score of the consummatory factor of the TEPS. Symptoms were normalized for display purposes to compare between anhedonia symptoms and overall depression severity. *p < 0.05, **p < 0.001.
Summary statistics of clinical metrics for participants
Summary statistics for behavioral metrics derived from the Adaptive-EEfRT
Linear models for the Adaptive-EEfRT
In a post hoc analysis to investigate the potential confounding influence of exhaustion, we performed linear models with the inclusion of a within-participant factor for time on task (median split of trials over time). This analysis revealed a conservation of the relationship between reward-responsive GDB and consummatory anhedonia when accounting for exhaustion (t = −3.067, p = 0.005) as reported in Table 3 and no effect of exhaustion on reward-responsive GDB (t = −1.344, p = 0.18). There was an effect of exhaustion on effort motivation (t = −3.31, p = 0.002), but accounting for this variance did not change the relationship with any symptoms. Exhaustion did not influence reward valuation (t = −0.248, p = 0.81) or change the relationship with any symptoms. Thus, exhaustion did not influence the relationship between symptoms and constructs of reward.
Prefrontal delta activity is engaged by multiple reward processing construct
Time–frequency analysis was run in the left prefrontal cortex (Fig. 2A). Low-frequency amplitude was increased throughout the decision period from 100 to 1,000 ms. There was a clear theta amplitude modulation from 100 to 500 ms with a separate spectral peak between 2 and 3 Hz with a longer extension in time from 100 to 1,000 ms. The time–frequency window from 2 to 3 Hz and 100 to 1,000 ms was used in subsequent analyses. Delta amplitude was greatest in the central midline and extended into the prefrontal electrodes and bilateral posterior electrodes (Fig. 2B). Investigation of left prefrontal delta power as a function of reward constructs revealed a significant increase in delta amplitude for reward-responsive goal-directed behavior (t(34) = 2.24, p = 0.031, d = 0.38; Fig. 2C) and reward valuation (t(34) = 2.43, p = 0.021, d = 0.41; Fig. 2E), but only a trend-level increase for effort motivation (t(34) = 1.71, p = 0.096, d = 0.29; Fig. 2D). Analysis of the event-related potential (ERP) or its spectral properties could not account for these delta-frequency effects (Fig. S1). Together, these findings demonstrate that left prefrontal delta activity is engaged in reward-based decision-making.
Decision-making recruited prefrontal delta amplitude. A, Spectrogram of left prefrontal cortex (L-PFC) showed increased delta and theta activity with two distinct spectral peaks. Inset depicts the region of interest for left prefrontal electrodes. Black outline are clusters that survived permutation-based correction. The delta oscillatory time–frequency region of interest is highlighted with a rectangle. B, Delta topography across all conditions revealed peak amplitude over the central midline and into the left prefrontal region. For the contrasts used to localize reward constructs, there was (C) increased left prefrontal delta amplitude for reward-responsive goal-directed behavior (GDB), (D) numerical increase for effort motivation, and (E) an increase for reward valuation. Dots are p < 0.05 with cluster correction (k ≥ 3). Dashed circles depict a significant electrode at p < 0.05 that did not survive cluster correction.
Beta amplitude over motor cortex is reduced during response
Spectral analysis of activity in electrodes over the left motor cortex revealed a decrease in beta amplitude (15–25 Hz) from
100 to 1,000 ms (Fig. 3A). The beta desynchronization was distinct from alpha (8–12 Hz) desynchronization, as indicated by two closely spaced Gaussian
components with different center frequencies in the time–frequency spectrogram. Analysis of right-handed blocks separately
from left-handed blocks revealed a maximal decrease in beta amplitude over left motor electrodes with right-handed blocks
and maximal decrease over right motor electrodes for left-handed blocks (Fig. 3B). An interaction was investigated using a repeated-measures two-way ANOVA with factors for hand (right-handed and left-handed
blocks) and motor cortex (left or right electrodes). There was a significant interaction between hand and motor cortex (F(1,34) = 6.50, p = 0.016, <mml:math display="inline"><mml:msubsup><mml:mi>η</mml:mi><mml:mrow><mml:mi mathvariant="normal">p</mml:mi></mml:mrow><mml:mn>2</mml:mn></mml:msubsup><mml:mo>=</mml:mo><mml:mn>0.16</mml:mn></mml:math>ηp2=0.16
), which corresponded to the canonical beta lateralization over motor cortex (t(34) = 2.55, p = 0.016, d = 0.43) such that motor beta decreased greatest when using the contralateral hand for effort exertion (Fig. 3C). There was no main effect of hand (F(1,34) = 0.50, p = 0.48, <mml:math display="inline"><mml:msubsup><mml:mi>η</mml:mi><mml:mrow><mml:mi mathvariant="normal">p</mml:mi></mml:mrow><mml:mn>2</mml:mn></mml:msubsup><mml:mo>=</mml:mo><mml:mn>0.01</mml:mn></mml:math>ηp2=0.01
) or motor cortex (F(1,34) = 0.60, p = 0.45, <mml:math display="inline"><mml:msubsup><mml:mi>η</mml:mi><mml:mrow><mml:mi mathvariant="normal">p</mml:mi></mml:mrow><mml:mn>2</mml:mn></mml:msubsup><mml:mo>=</mml:mo><mml:mn>0.02</mml:mn></mml:math>ηp2=0.02
) on beta amplitude. These analyses assert the independence of the motor beta oscillatory signal from the prefrontal delta
oscillatory signal.
Beta oscillations over the motor cortex are lateralized to the contralateral hand used for effort exertion. A, Spectrogram of left motor electrodes revealed distinct beta and alpha amplitude peaks. Black outline are clusters that survived permutation-based correction. The beta oscillatory time–frequency region of interest is highlighted with a rectangle. B, Beta decrease is strongest over the left motor cortex for right-handed blocks and (C) vice versa for left-handed blocks. D, Lateralized beta amplitude was observed over the motor cortices. Dashed circles depict an electrode that was significant at p < 0.05 but did not survive cluster correction.
Prefrontal-motor delta-beta coupling during decision-making
Delta-beta coupling was calculated using mean vector length between the phase of delta oscillations in left prefrontal electrodes and the amplitude of beta oscillations in left motor electrodes (Fig. 4A). For left-handed blocks, the scalp topography of beta amplitude was left–right flipped. Across all conditions, there was significant phase–amplitude coupling between delta and beta oscillations (t(34) = 4.63, p < 0.001, d = 0.78). Topographic analysis showed delta-beta coupling for bilateral motor cortices (contralateral and ipsilateral; Fig. 4B). For reward-responsive goal-directed behavior, delta-beta coupling was significantly recruited (t(34) = 3.27, p = 0.002, d = 0.55). Topographic analysis revealed that coupling was specific to the contralateral motor cortex and the frontal midline (Fig. 4C). Delta-beta coupling was increased for effort motivation (t(33) = 3.88, p < 0.001, d = 0.67) and for reward valuation (t(34) = 3.21, p = 0.003, d = 0.54); topographic analysis revealed coupling with bilateral motor cortices (Fig. 4D,E). Together, these results demonstrated that delta-beta coupling was recruited during reward-based decision-making and was increased by multiple constructs within the Positive Valence Systems.
Delta-beta coupling was recruited by multiple reward constructs. A, Phase–amplitude coupling (PAC) was calculated using mean vector length between delta phase in the left prefrontal cortex (L-PFC) and topographic beta amplitude. For left-handed blocks the topography was left–right flipped to account for beta amplitude lateralization. B, During decision-making (100–1,000 ms after decision onset), delta-beta coupling was increased with bilateral motor cortices. Dots are p < 0.05 after cluster correction of k ≥ 3. C-M1 denotes electrodes over primary motor cortex contralateral to the hand used for effort exertion, and I-M1 is ipsilateral primary motor cortex. C, For reward-responsive goal-directed behavior (GDB), delta-beta coupling was increased to contralateral motor electrodes and frontal midline electrodes. For (D) effort motivation and (E) reward valuation, delta-beta coupling was increased to bilateral motor cortices.
Delta-beta tACS increased prefrontal-motor coupling
Delta-beta tACS was delivered with synchronized current to left PFC and left motor cortex (Fig. 5A) with peak normalized electric field strength in the anterior middle frontal gyrus and primary motor cortex (Fig. 5B). The impact of delta-beta tACS on behavior and delta-beta coupling was investigated by comparison to tACS in an active control cross-frequency coupling pair of noninterest, theta-gamma coupling (Fig. 5C), and to sham tACS (placebo) that mimicked the experience of receiving tACS. Resting-state periods were inserted after each block of the Adaptive-EEfRT to quantify target engagement. A two-way ANOVA was run using factors tACS (delta-beta, active control, or placebo) and hand used for the block (right or left). There was a trend-level main effect of tACS (F(2,32) = 3.26, p = 0.051, η2 = 0.17) and an interaction between tACS and hand used (F(2,32) = 3.16, p = 0.056, η2 = 0.16). This effect was driven by an increase in delta-beta coupling following delta-beta tACS (t(10) = 2.40, p = 0.038, d = 0.72). The effect was stronger following right-handed blocks (t(10) = 4.62, p = 0.001, d = 1.39) than following left-handed blocks (t(10) = 0.11, p = 0.91, d = 0.03). Topographic analysis of right-handed blocks revealed that these effects were specific to the motor cortex (Fig. 5C). There was no change in delta-beta coupling strength for sham tACS (N = 12; t(11) = −0.31, p = 0.76, d = 0.09; Fig. 5D) or theta-gamma tACS (N = 12; t(11) = −1.62, p = 0.13, d = 0.47; Fig. 5E). There was a decrease in delta-beta coupling for right-hand blocks following theta-gamma tACS (t(11) = −3.23, p = 0.008, d = 0.93). These findings provide evidence of target engagement from delta-beta tACS.
Delta-beta tACS increased prefrontal-motor delta-beta coupling. A, Cross-frequency transcranial alternating current stimulation (tACS) was delivered to left prefrontal cortex (L-PFC) and left motor cortex (L-M1) with a return electrode over the frontal midline. B, The normalized electric field shows peak field strength in left anterior middle frontal gyrus and primary motor cortex. C, The targeted cross-frequency coupling rhythm was delta nested beta (green). Following delta-beta tACS, delta-beta coupling between L-PFC and L-M1 significantly increased. D, Sham tACS utilized a placebo ramping protocol that mimicked the sensation of receiving tACS (gray). After sham tACS, there was no change in delta-beta coupling in the targeted regions. E, As an active control, theta nested gamma tACS (purple) was delivered to test the frequency-specificity of delta-beta tACS. After theta-gamma tACS, there was a significant decrease in delta-beta coupling. *p < 0.05 between the a priori regions of interest (outlined regions). L-PFC is F3 and surrounding electrodes where delta phase was extracted. L-M1 is C3 and surrounding. Right-handed blocks are displayed. Dots are p < 0.05, k ≥ 3.
There was a weak, nonsignificant positive association between neural target engagement and the increase in reward-responsive goal-directed behavior from delta-beta tACS (r(9) = 0.48, p = 0.14). Mediation analysis indicated that the effect of delta-beta tACS on behavior was not significantly transmitted through delta-beta coupling (sham as control; ACME p = 0.73). Thus, the observed increase in reward-responsive goal-directed behavior reflects a direct effect of delta-beta tACS rather than an indirect effect through changes in phase–amplitude coupling.
Delta-beta tACS increased reward-responsive goal-directed behavior
Two-way mixed-effects ANOVAs were run for the components of the Positive Valence Systems: reward-responsive goal-directed
behavior, effort motivation, and reward valuation. For reward-responsive goal-directed behavior, there was a trend-level interaction
between time (baseline and during tACS) and tACS (delta-beta, active control, and placebo; F(2,32) = 3.25, p = 0.052, <mml:math display="inline"><mml:msubsup><mml:mi>η</mml:mi><mml:mrow><mml:mi mathvariant="normal">p</mml:mi></mml:mrow><mml:mn>2</mml:mn></mml:msubsup><mml:mo>=</mml:mo><mml:mn>0.17</mml:mn></mml:math>ηp2=0.17
; Table 4). The interaction was driven by a significant increase in reward-responsive goal-directed behavior following delta-beta tACS
relative to placebo (t(21) = 2.57, p = 0.018, d = 1.08) and relative to active control (t(21) = 2.10, p = 0.048, d = 0.90; Fig. 6A). There was no difference between active control and placebo (t(22) = −0.26, p = 0.80, d = 0.11). There was neither an interaction of time and tACS for effort motivation (F(2,32) = 2.02, p = 0.15, <mml:math display="inline"><mml:msubsup><mml:mi>η</mml:mi><mml:mrow><mml:mi mathvariant="normal">p</mml:mi></mml:mrow><mml:mn>2</mml:mn></mml:msubsup><mml:mo>=</mml:mo><mml:mn>0.11</mml:mn></mml:math>ηp2=0.11
; Fig. 6B) nor for reward valuation (F(2,32) = 0.44, p = 0.65, <mml:math display="inline"><mml:msubsup><mml:mi>η</mml:mi><mml:mrow><mml:mi mathvariant="normal">p</mml:mi></mml:mrow><mml:mn>2</mml:mn></mml:msubsup><mml:mo>=</mml:mo><mml:mn>0.03</mml:mn></mml:math>ηp2=0.03
; Fig. 6C).
Delta-beta tACS increased reward-responsive goal-directed behavior. A, Reward-responsive goal-directed behavior was increased by delta-beta tACS relative to placebo (sham) and active control (theta-gamma). There was no difference in (B) effort motivation or (C) reward valuation for delta-beta tACS relative to placebo or active control. y axis is the difference for tACS relative to baseline. *p < 0.05.
Analysis of variance for tACS effect on behavior
In a post hoc outlier analysis, an outlier for reward-responsive goal-directed behavior during active control tACS was identified. Following removal, delta-beta tACS increased reward-responsive goal-directed behavior relative to active control (t(20) = 3.49, p = 0.002, d = 1.49), and there was still no difference between active control and placebo (t(21) = −1.27, p = 0.22, d = 0.53). The results were strengthened by outlier removal. For effort motivation, an outlier was identified for placebo tACS. After removal, there was a significant difference from sham tACS for delta-beta tACS (t(20) = −3.24, p = 0.004, d = 1.39) and for active-control tACS (t(21) = −3.10, p = 0.005, d = 1.30). However, there was no difference between delta-beta tACS and active-control tACS (t(21) = −0.96, p = 0.35, d = 0.41). While there was an overall decrease in effort motivation, the effect was not different from active-control and, thus, the effect on effort motivation was not frequency-specific.
In a post hoc analysis for the potential confounding influence of exhaustion, we included a within-participants factor of
exhaustion (median split). ANOVA revealed a significant interaction between time (pre- versus during-tACS) and tACS (F(2,32) = 3.40, p = 0.046; <mml:math display="inline"><mml:msubsup><mml:mi>η</mml:mi><mml:mrow><mml:mi mathvariant="normal">p</mml:mi></mml:mrow><mml:mn>2</mml:mn></mml:msubsup><mml:mo>=</mml:mo><mml:mn>0.18</mml:mn></mml:math>ηp2=0.18
) for reward-responsive GDB similar to that shown in Table 4. There were no interactions between exhaustion and tACS. Thus, the impact of delta-beta tACS on increasing reward-responsive
GDB was not influenced by exhaustion. While there was a strong main effect of exhaustion on effort motivation (F(1,32) = 17.29, p = 0.0002; <mml:math display="inline"><mml:msubsup><mml:mi>η</mml:mi><mml:mrow><mml:mi mathvariant="normal">p</mml:mi></mml:mrow><mml:mn>2</mml:mn></mml:msubsup><mml:mo>=</mml:mo><mml:mn>0.35</mml:mn></mml:math>ηp2=0.35
; as seen at baseline), there was no interaction with tACS.
Participants were successfully blinded to tACS
Participants were successfully blinded to whether they received placebo tACS, with 63.6% reporting that they received verum tACS from delta-beta tACS, 33% from theta-gamma tACS, and 91.7% from placebo tACS (Table 5). There was a significant difference between conditions [χ2(2, N = 35) = 8.70, p = 0.01] driven by participants thinking that they did not receive tACS when delta-beta tACS was delivered. Experimenters, on average, were aware of the type of tACS that was delivered with 81.8% reporting real stimulation from delta-beta tACS, 66.7% from theta-gamma tACS, and 33.3% with placebo tACS [χ2(2, N = 35) = 5.96, p = 0.051; Table 5]. Since tACS was repeatedly turned off at the end of each block and the connection was re-established at the start of each block, experimenters likely became aware of the quickly draining batteries for verum tACS and not sham tACS.
Blinding efficacy and side effects
Side effects from tACS were matched across study arms. Participants reported at least one side effect (> 2 on a 0 to 4 Likert scale) 45.5% of the time for delta-beta tACS, 41.7% for theta-gamma tACS, and 66.7% for placebo tACS. The average reported symptom severity was not significantly different between any two conditions (ts < 1, ps > 0.3). For delta-beta tACS, 3 of 11 participants experienced an adverse event, 3 of 12 for theta-gamma tACS, and 1 of 12 for sham tACS (Table S1). In each of these instances, the side effects were deemed to be mild and resolved after the termination of tACS. For delta-beta tACS, one of the adverse events was deemed to be unrelated to the intervention (trouble concentrating) and the other two were related (visual phosphenes). For theta-gamma tACS, two adverse events were deemed unrelated to stimulation (sleepiness) and the other was related (scalp pain). For sham tACS, the single adverse event was unrelated to the intervention (sleepiness). Taken together, the study was participant-blinded such that the behavioral and neural effects of tACS were unlikely to be driven by differences in expectations or the experience of side effects.
Discussion
Participants with MDD performed the adaptive version of the expenditure of effort for reward task (Adaptive-EEfRT), which dissociated multiple reward constructs within the domain of the Positive Valence Systems. Symptoms of consummatory anhedonia were correlated with reduced reward-responsive goal-directed behavior, but not with effort motivation or reward valuation. By recording EEG, neural oscillations involved in reward processing were investigated during decision-making. The amplitude of left prefrontal delta oscillations (2–3 Hz) was increased as a function of reward-responsive goal-directed behavior, effort motivation, and reward valuation. Beta oscillations (15–25 Hz) over bilateral motor cortices were decreased during response preparation but were maximally decreased contralateral to the hand used to make a response. Phase–amplitude coupling was estimated between the delta phase of the left lateral prefrontal cortex and beta amplitude of the motor cortex and was found to be strengthened for each of the reward constructs. When delta-beta tACS was delivered, prefrontal-motor delta-beta coupling was strengthened and reward-responsive goal-directed behavior was increased. These findings provide causal evidence that prefrontal-motor delta-beta coupling plays a role in incorporating reward outcomes into the decision process.
This study is the first to utilize the Adaptive-EEfRT, which improves upon previous iterations by bringing the participant close to their decision indifference point. An indifference point is when a participant reaches a balance between alternative options because the perceived cost/reward value is matched. When investigating individual differences such as the relationship between behavior and symptoms, the titration of the task to the individual is critical to ensure that the distributions of behavioral metrics are normally distributed and not near the ceiling or floor of what can be measured by the task (Riddle and Frohlich, 2023). The Adaptive-EEfRT rapidly changed the level of effort exertion required. While previous studies using the EEfRT found that effort motivation was decreased with greater symptoms of anhedonia, the Adaptive-EEfRT revealed that with a more dynamic titration process reward-responsive goal-directed behavior, and not effort motivation or reward valuation, was reduced in those with greater anhedonia symptoms. This analysis highlights the importance of titrating task difficulty, as the Adaptive-EEfRT employed here might more effectively engage dynamic decision-making. In support of this interpretation, the average decision to exert higher effort was 55.2%, which was just above the indifference point of 50%. While reward responsiveness is typically assessed using the difference between reward and punishment, e.g., the reward-related positive potential (RewP) during the Doors task, here reward-responsive goal-directed behavior incorporated how the reward outcome influences effort exertion. Thus, while this behavioral metric falls under the reward responsiveness construct, it emphasizes the impact of reward on effort exertion. This framing is consistent with the discovery that the RewP is blunted, and perhaps nonexistent, in the absence of a task goal (Liu et al., 2025). Furthermore, the RewP was first quantified as the neural response to receiving performance feedback: correct versus incorrect (Miltner et al., 1997). Altogether, the findings of blunted reward-responsive goal-directed behavior in the Adaptive-EEfRT are consistent with the theory that reward outcome processing should be investigated with respect to goal-directed behavior (Cavanagh and Holroyd, 2026).
Symptoms of consummatory anhedonia, but not anticipatory anhedonia, were related to blunted reward-responsive goal-directed behavior. Since the Adaptive-EEfRT required effort motivation and goal-directed processing, one might expect that anticipatory anhedonia might be more predictive of individual differences in performance (Berridge et al., 2009). However, reward-responsive goal-directed behavior quantified the impact of the experience of the reward on subsequent decision-making. The relationship to consummatory anhedonia suggests that reduced hedonic experience during reward receipt likely explains the reduction in goal updating following the reward. While we did not assess apathy, future research should refine the symptom specificity of reduced reward-responsive goal-directed behavior by accounting for the general effect of apathy.
The left prefrontal cortex was used as a target based on its known role in approach motivation (Harmon-Jones, 2003) and because of the successful use of transcranial magnetic stimulation to the left prefrontal cortex to reduce symptoms of anhedonia (Siddiqi et al., 2020). Consistent with this decision, delta oscillations were increased specifically in left prefrontal electrodes as a function of all three reward constructs: reward-responsive goal-directed behavior, effort motivation, and reward valuation. Thus, the choice to target left prefrontal delta oscillations was an optimal selection based on the data. Nonetheless, there are a multitude of waveform and montage parameters to consider (Riddle and Frohlich, 2021; Wischnewski et al., 2023). Future research could investigate whether left prefrontal delta tACS without the use of a cross-frequency waveform is sufficient to modify reward-responsive goal-directed behavior. Alternatively, effort motivation and reward valuation might be engaged by left prefrontal delta tACS, whereas the incorporation of a cross-frequency waveform to synchronize the left prefrontal and motor cortices is wholly different.
The investigation of neural activity during decision-making in reward-based decision-making tasks is novel because most previous studies were focused on the processing of the reward outcome. Nonetheless, reward-based decision-making likely resembles other forms of decision-making. For example, a previous series of studies that identified prefrontal-to-motor delta-beta coupling found greater delta-beta coupling when contextual control was required (Riddle et al., 2020b; Riddle et al., 2021). Here, delta-beta coupling was strengthened for higher incentives, high effort motivation, and following receipt of reward. While speculative, top-down signaling from the prefrontal cortex might reflect the incorporation of contextual information into the decision-making process when a high-effort decision is more likely to be chosen. Future studies could use an enriched decision landscape where uncertainties about the probability of reward are introduced, and behavior is modeled using reinforcement learning models. The importance of considering contextual information in the decision process could be manipulated with the prediction that delta-beta coupling will track with the degree of contextual control engaged.
In the previous work that investigated delta-beta coupling, the hand used for decision-making and effort exertion was not manipulated as it was here. Delta-beta coupling increased to both the motor cortex contralateral to the hand being used and the motor cortex that was ipsilateral. Delta-beta coupling was strengthened to bilateral motor cortices with greater effort exertion and for higher incentives. However, as a function of reward-responsive goal-directed behavior, delta-beta coupling was only increased to the motor cortex contralateral to the actively used hand and uniquely increased coupling to the frontal midline. Frontal midline structures, e.g., the orbitofrontal cortex and the anterior cingulate cortex, have been associated with the processing of the value of a reward (Wallis and Kennerley, 2011; Shenhav et al., 2016). Thus, the unique pattern of delta-beta coupling as a function of reward-responsive goal-directed behavior might reflect the incorporation of the contextual relevance of reward value processed in midline structures with the active control of the motor cortex used to make an action selection. Given that reward-responsive goal-directed behavior alone showed unilateral prefrontal-motor coupling, the application of delta-beta tACS to only one hemisphere might have disrupted the balanced bilateral delta-beta coupling of effort motivation and reward valuation, which could partially explain why there was no significant impact on these behavioral metrics.
The electrical waveform used for the active control mimicked theta-gamma coupling, which was used previously to target neural activity associated with working memory (Polania et al., 2012; Jaušovec and Jaušovec, 2014; Alekseichuk et al., 2016; Violante et al., 2017; Wolinski et al., 2018; Reinhart and Nguyen, 2019) and rule set maintenance (Riddle et al., 2021). Here, theta-gamma tACS did not result in any change in behavior and produced a decrease in delta-beta coupling. There is a rich literature that investigates the complementary roles of beta and gamma oscillations for supporting top-down and bottom-up information processing, respectively (Bastos et al., 2015; Michalareas et al., 2016). In fact, these two modes of information transfer are preserved across the cerebral cortex and instantiated in the laminar profile where deep layers display gamma oscillations linked to feedforward signaling and superficial layers display beta oscillations related to feedback signaling (Mendoza-Halliday et al., 2024). Thus, the delivery of theta-gamma coupled tACS might have disrupted delta-beta coupling by virtue of the inverse relationship between beta and gamma activity.
As with any scientific study, the present investigation has limitations. First, the experimenters became unblinded to the type of tACS that was delivered. This influence was mitigated by the fact that unique codes were used for each participant such that the experimenter could not unconsciously influence the ongoing experiment. Furthermore, experimenters did not distinguish active-control tACS from delta-beta tACS. Second, the preregistered primary outcome was an increase in effort motivation with delta-beta tACS. Thus, the change in reward-responsive goal-directed behavior, while logically following from analysis of the baseline data, was not expected. Relatedly, there was not a significant negative relationship between effort motivation and symptoms of anhedonia as has been previously found (Riddle et al., 2022; Treadway et al., 2012). One explanation is that the adaptive component used here was novel. By pushing the decision criteria closer to the indifference point of the individual, the relationship of net effort motivation to anhedonia symptom severity might have been lost. Finally, there was a theoretical risk that noninvasive brain stimulation increased risk-taking, a symptom of mania. However, in the present study, participants became more sensitive to the reward of the previous trial and showed a numerical decrease in their rate of choosing the higher incentive option, which suggests that risk-taking was not increased.
In conclusion, these data provide additional evidence for a robust coupling signal between the prefrontal cortex and the motor cortex during reward-based decision-making. The coupling of prefrontal delta phase to motor beta amplitude is not epiphenomenal from coincident occurrence but is causally related to the processing of reward and the decision to exert effort because targeted neurostimulation engaged this mechanism and systematically altered behavior. These data provide a critical foundation for the development of more targeted neuromodulation approaches to specific electrophysiological mechanisms.
Footnotes
- Received July 4, 2025.
- Revision received December 15, 2025.
- Accepted January 13, 2026.
-
This work was supported by the National Institute of Mental Health (K99/R00 MH126161 to J.R.; R01 MH124387 to F.F.) and by the Foundation of Hope (Seed Research Grant to J.R.). The funding organizations did not influence the reporting of results. This manuscript does not reflect the opinions of the funders.
-
J.R., L.P.E., M.S., and C.E.S. have no financial disclosures. D.R.R. has received research funding from the Buszucki Foundation, NIH, and Sage Therapeutics. He has served on scientific advisory boards for the Foundation of Hope, Sage Therapeutics, and Sensorium Therapeutics and clinical advisory boards for Embark neuro and Felicity Pharma. He has served as consultant for Aldeyra Therapeutics, Brii Biosciences, and GH Research–Ireland; and he has stock options in Sage Therapeutics and Sensorium Therapeutics. F.F. is a shareholder of EPI and receives payments from EPI for chairing the scientific advisory board. He is listed as an inventor on patents issued to UNC-Chapel Hill on noninvasive brain stimulation.
-
This paper contains supplemental material available at: https://doi.org/10.1523/JNEUROSCI.1289-25.2026
- Correspondence should be addressed to Justin Riddle at justin.riddle{at}fsu.edu or Flavio Frohlich at Flavio_frohlich{at}med.unc.edu.
- Copyright © 2026 the authors
