Schematic of the animal experiment, histology, electrode design, and high-fat (HF) intake summary. (A) Schematic of the experimental design: electrode implantation, followed by recovery period (7 d), 1-h daily HF access (days 0–18), and intervention period (days 11–18). (B–D) Histology examination revealing the implant locations, and design of the eight-contact multielectrode array, which had six designated recording contacts (in blue/green) and two designated stimulating contacts (in red). (E–G) Binge-like behavior developed and stabilized by day 10 of 1-h daily HF exposure, indicated by significant increase in daily HF intake with >25% of daily caloric intake in 1 h. *P < 0.05.

Raw local field potentials (LFPs), power spectral density, time–frequency analyses of the nucleus accumbens (NAc) LFPs, delta power characterization, and system block diagram of the RNS setup. (A–C) Raw LFP samples are shown during the onset (dotted gray line) of house chow as a control and high-fat (HF) consumption on days 0 and 10 (before and after the development of binge-like behavior, respectively). (D–F) Mean power spectrogram of NAc LFPs immediately before and after the onset of chow and HF consumption on days 0 and 10. (G) Power spectral density analysis of NAc LFPs immediately before (2-s window) chow and HF intake on days 0 and 10, averaged across individual mice, revealing higher power in low-frequency oscillations immediately before the onset of HF intake on day 10. (H) Mean delta power significantly increased immediately before the onset of HF intake on day 10 compared with HF intake on day 0 and chow intake. (I) Delta power percent change over baseline during the onset of HF consumption on day 10 (normalized to the 1-h period of HF exposure). (J) Delta power peak distribution before the onset HF consumption on day 10. (K) Mean power spectrogram of NAc LFP during the onset of juvenile interaction. (L) System block diagram of the RNS setup, which consisted of a 1× follower cable for unit amplification, a headstage for analog/digital conversion, a digital filter, a computer for synchronizing neural electrophysiological and behavioral data, a prototype biomarker detector (Neurostimulator; model RNS-300; NeuroPace), a constant-current stimulator, and a charge-coupled device camera for synchronized behavioral recording. **P < 0.01. See also and .

Schematic of the intervention period during 1-h daily high-fat (HF) exposure in mice, and result summary of different electrical stimulation protocols on HF intake and juvenile interaction. (A) Representative nucleus accumbens (NAc) LFP delta oscillations (in red) at the onset of HF intake detected by the integrated responsive neurostimulation (RNS) system. (B) Schematic of the intervention period. Each intervention session [deep brain stimulation (DBS), manually triggered stimulation (Man), responsive neurostimulation (RNS), randomly applied stimulation (Rnd)] was followed by one washout session (Off). (C–F) The effects of different stimulation protocols on HF consumption. DBS, Man, and RNS significantly reduced HF intake. (G) Reduction in HF intake induced by DBS, manually triggered stimulation, RNS, and randomly applied stimulation, compared with off stimulation. The reduction in HF intake induced by the manually triggered stimulation and RNS was significantly higher than randomly applied stimulation. (H) The number of stimulations bouts (1 bout = 10 s) delivered during Man and RNS were significantly lower than DBS. (I) DBS of the NAc significantly reduced juvenile interaction time, while RNS showed no effect on this behavior. (J) Neither DBS nor RNS of the NAc during HF exposure altered locomotor activity. (K) Real-time place preference test suggested that NAc stimulation was neither rewarding nor aversive. (L and M) Sensitivity and specificity of delta biomarker on RNS day in all six mice. In total, there were 179 HF pellet approaches, of which 124 were detected by the RNS system (sensitivity, 0.693). There were also 1,241 correct rejections (stimulation off when no HF approach occurred) and 685 stimulations triggered when no HF approaches were observed (false stimulation; specificity, 0.644). *P < 0.05, **P < 0.01, and ***P < 0.001. See also and .

Multiunit and coherence analyses. (A) Identification, principal-component (PC) analysis, and average waveforms of nucleus accumbens (NAc) neurons recorded on day 10 during HF intake. (B) Raster plot of NAc firing represented for each trial under different conditions. (C) NAc spike rate on day 10 during HF intake was significantly higher than chow and day 0 HF intake. (D) Representative example of NAc LFPs (Bottom) and unit activities (Top), showing phase synchrony with low-frequency oscillations. (E) Delta spike–field coupling revealed that the coherence on day 10 during HF intake was significantly higher than chow and day 0 HF intake. (F) NAc spike rate significantly correlated with delta power on day 10 during HF intake. ^#P < 0.1, *P < 0.05, **P < 0.01, and ****P < 0.0001.

Schematic of functional neuroimaging and local field potentials (LFPs) recording during the monetary incentive delay (MID) task in human subject. (A) Schematic of the MID task, which consists of cue onset, anticipation phase, target onset, and outcome phase. (B) Functional magnetic resonance imaging (fMRI) showing area activated by gain vs. nongain anticipation in the nucleus accumbens (NAc) (white circle; Z > 2.54; cluster, four 3-mm cubic voxels). (C) Blood oxygen level-dependent (BOLD) signal changes during MID task extracted from activated voxels in the left NAc averaged by condition, indicating NAc activation during high reward anticipation compared with baseline [high reward:baseline, T₍₁₇₎ = 3.23, P < 0.01, uncorrected; low reward:baseline, high punishment:baseline, low punishment:baseline, n.s.]. (D) Electrode contact locations in the NAc for LFP recording using preoperative 7-T MRI merged with postoperative computed tomography scan. Coronal view (trajectory view not shown) demonstrates most posterior aspect of electrode trajectory with the following entry anterior–posterior commissure coordinates: x = 41.31, y = 43.39, z = 43.59; 34.2° from midsagittal plane; 60.3° from axial plane; and coordinates for the ventral-most extent of the recording lead: x = 6.03, y = 15.07, z = −6.60. The measurement of 7.5 mm indicates the span of the lead. (E and F) Raw LFPs during baseline and anticipation of high reward. (G) Power density analysis indicating delta range as the frequency region of interest during anticipation period. (H) Power density analysis revealing significant increase in delta power during anticipation of high reward compared with baseline. (I) Normalized NAc LFP power spectrogram (averaged across individual trials), indicating increased delta band (1- to 4-Hz) power during anticipation of high reward (Insets: frequency range from 0 to 50 Hz). **P < 0.01. BL, baseline; HP, high punishment; HR, high reward; LP, low punishment; LR, low reward. See also –.