High-frequency neuromodulation improves obsessive-compulsive behavior

Abstract

Nearly one billion people worldwide suffer from obsessive-compulsive behaviors^1,2, yet our mechanistic understanding of these behaviors is incomplete, and effective therapeutics are unavailable. An emerging perspective characterizes obsessive-compulsive behaviors as maladaptive habit learning^3,4, which may be associated with abnormal beta-gamma neurophysiology of the orbitofrontal-striatal circuitry during reward processing^5,6. We target the orbitofrontal cortex with alternating current, personalized to the intrinsic beta-gamma frequency of the reward network, and show rapid, reversible, frequency-specific modulation of reward- but not punishment-guided choice behavior and learning, driven by increased exploration in the setting of an actor-critic architecture. Next, we demonstrate that chronic application of the procedure over 5 days robustly attenuates obsessive-compulsive behavior in a non-clinical population for 3 months, with the largest benefits for individuals with more severe symptoms. Finally, we show that convergent mechanisms underlie modulation of reward learning and reduction of obsessive-compulsive symptoms. The results contribute to neurophysiological theories of reward, learning and obsessive-compulsive behavior, suggest a unifying functional role of rhythms in the beta-gamma range, and set the groundwork for the development of personalized circuit-based therapeutics for related disorders.

Extended Data Fig. 1 |. Neuromodulation effects on temperature.

Mean temperature shown across pre-modulation, modulation, and post-modulation periods for each group (passive sham, green; active control/alpha, blue; personalized beta-gamma, red). A significant time × group interaction was observed (F_4,114 = 3.066, P = 0.027, ${\eta }_p^2$ = 0.097, n = 60). There were significant time × group interactions for the beta-gamma and alpha groups (F_2,76 = 4.942, P = 0.012, ${\eta }_p^2$ = 0.115, n = 40) and the beta-gamma and sham groups (F_2,76 = 4.119, P = 0.027, ${\eta }_p^2$ = 0.098, n = 40), but not for the alpha and sham groups (F_2,76 = 0.282, P = 0.686, ${\eta }_p^2$ = 0.007, n = 40). A significant enhancement in temperature was evident only during beta-gamma HD-tACS relative to sham (F_1,38 = 6.409, P = 0.016, ${\eta }_p^2$ = 0.144, n = 40) or to alpha (F_1,38 = 6.311, P = 0.016, ${\eta }_p^2$ = 0.142, n = 40). The effect was rapidly extinguished upon switching off HD-tACS, given that no differences were observed between the alpha and beta-gamma (F_1,38 = 0.056, P = 0.814, ${\eta }_p^2$ = 0.001, n = 40) or the sham and beta-gamma groups (F_1,38 = 0.762, P = 0.388, ${\eta }_p^2$ = 0.02, n = 40) in the post-modulation period. Of note, baseline temperature values were relatively stable and did not significantly differ between groups during the pre-modulation period (alpha versus beta-gamma, F_1,38 = 1.528, P = 0.224, ${\eta }_p^2$ = 0.039, n = 40; alpha versus sham, F_1,38 = 0.072, P = 0.790, ${\eta }_p^2$ = 0.002, n = 40; sham versus beta-gamma, F_1,38 = 1.421, P = 0.241, ${\eta }_p^2$ = 0.036, n = 40). No other parameters showed significant effects. Mixed ANOVAs used the within-participants factor of time (pre-modulation, modulation, post-modulation) and the between-participants factor of group (sham, alpha, beta-gamma). Follow-up univariate ANOVAs within individual modulation periods used the between-participants factor of group (alpha, beta-gamma; alpha, sham; or beta-gamma, sham). Error bars show ±1 s.e.m. *P < 0.05. NS, not significant.

Fig. 1 |. Model integrating beta-gamma activity with reward learning circuitry.

Arrows show anatomical relationships between components of the reward and learning circuitry. Dashed black arrows show the loop of activity between the hippocampus (Hipp), nucleus accumbens (NAcc) and ventral tegmental area (VTA) put forward by Axmacher et al., with solid black arrows showing other major connections involved in learning and motivation. The positive reward feedback beta-gamma effect shown to the right is theorized to derive from generators in the orbitofrontal cortex (OFC) or ventromedial prefrontal cortex (vmPFC)^,, to be modulated by the substantia nigra (SN) and VTA (short red arrow), and to synchronize with beta-gamma rhythms in the ventral striatum (VS), providing a key mechanism by which the communication between frontostriatal regions can result in the transmission of a motivational value signal to the reward circuit in learning or decision-making contexts (long red arrow). Amy, amygdala; dACC, dorsal anterior cingulate cortex; Th, thalamus; VP, ventral pallidum.

Fig. 2 |. Monetary reinforcement learning task and orbitofrontal neuromodulation protocol.

a, Participants chose one of two abstract visual stimuli and later observed the outcome. Following a jittered interstimulus interval (ISI), the outcome depending on the trial type was shown. In reward trials, one stimulus was associated with an 80% probability of winning US$10 and a 20% probability of winning nothing, and the other stimulus had the opposite probability structure. In punishment trials, one stimulus was associated with an 80% probability of losing US$10 and a 20% probability of losing nothing, and the other stimulus had the opposite probability structure. b, Personalized orbitofrontal neuromodulation protocol and current-flow models on three-dimensional reconstructions of the cortical surface. The location and current intensity value of each electrode are shown.

Fig. 3 |. Results of experiment 1, the monetary reinforcement learning task.

a–c, Observed optimal behavioral choices for reward (blue) and punishment (gold) during the pre-modulation (Pre, dotted), modulation (Mod, solid), and post-modulation (Post, dashed) blocks for the passive sham (a), active control/alpha (b) and personalized beta-gamma (c) groups. A significant modulation group × valence × time interaction was observed (F_4,114 = 4.337, P = 0.003, ${\eta }_p^2$ = 0.132, n = 60). The interaction effect was present when comparing beta-gamma with sham (F_2,76 = 3.643, P = 0.036, ${\eta }_p^2$ = 0.087, n = 40) and beta-gamma with alpha (F_2,76 = 7.162, P = 0.001, ${\eta }_p^2$ = 0.159, n = 40), but not when comparing alpha with sham (F_2,76 = 1.286, P = 0.282, n =I 40). During the modulation period, a significant group effect was observed for reward trials (F_2,57 = 17.735, P = 1 × 10⁻⁶, ${\eta }_p^2$ = 0.384, n = 60), with beta-gamma modulation driving the effect (sham versus beta-gamma, F_1,38 = 16.170, P = 2.65 × 10⁻⁴, ${\eta }_p^2$ = 0.299, n = 40; alpha versus beta-gamma, F_1,38 = 35.997, P = 5.69 × 10⁻⁷, ${\eta }_p^2$ = 0.486, n = 40; sham versus alpha, F_1,38 = 0.805, P = 0.375, n = 40), but no differences were observed for punishment trials across groups (all F < 0.291, all P > 0.593, all n = 40). In the beta-gamma group (c, right panel), there was a significant effect of time on reward trials (F_2,38 = 20.159, P = 2 × 10⁻⁶, ${\eta }_p^2$ = 0.515, n = 20), with the modulation period showing significant differences compared with both pre-modulation (F_1,19 = 29.533, P = 3 × 10⁻⁵, ${\eta }_p^2$ = 0.609, n = 20) and with post-modulation (F_1,19 = 23.012, P = 1.25 × 10⁻⁴, ${\eta }_p^2$ = 0.548, n = 20) periods. No differences were observed between pre-modulation and post-modulation periods in reward trials (F_1,19 = 0.822, P = 0.376, n = 20). Beta-gamma modulation did not affect choice behavior in punishment trials over time (all F < 0.370, all P > 0.550, ns = 20). Mixed ANOVAs used the within-participants factors of valence (reward, punishment) and time (pre-modulation, modulation, post-modulation), and the between-participants factor of group (sham, alpha, beta-gamma). Error bars and shaded error bands, ±1 s.e.m. ***P < 0.001. NS, not significant.

Fig. 4 |. Results of experiment 2, change in obsessive-compulsive symptoms after HD-tACS.

a, Mean OCI-R total score (left panel) and mean OCI-R subscale scores (right panels) with 95% confidence intervals (with mean at center of error bars) shown for each modulation group and timepoint. The shaded region shows the 5 day intervention period. For total OCI-R (left panel), there was a significant group × time interaction (F_4,248 = 6.748, P = 2.43 × 10⁻⁴, ${\eta }_p^2$ = 0.098, n = 64). There was a main effect of time on OCI-R in the beta-gamma group (F_4,124 = 10.278, P = 5.5 × 10⁻⁵, ${\eta }_p^2$ = 0.249, n = 32), and no effect for the control/alpha group (F_4,124 = 0.307, P = 0.80, n = 32). See main text for statistics of pairwise comparisons. For subscales, significant I group × time effects were evident for hoarding (F_4,248 = 2.886, P = 0.030, ${\eta }_p^2$ = 0.044, n = 64), ordering (F_4,248 = 4.234, P = 0.004, ${\eta }_p^2$ = 0.064, n = 64) and obsessing (F_4,248 = 2.817, P = 0.032, ${\eta }_p^2$ = 0.043, n = 64). The beta-gamma modulation group showed significant main effects of time for hoarding (F_4,124 = 8.169, P = 1.16 × 10⁻⁴, ${\eta }_p^2$ = 0.209, n = 32), checking (F_4,124 = 4.827, P = 0.008, ${\eta }_p^2$ = 0.135, n = 32), ordering (F_4,124 = 9.8, P = 3.9 × 10⁻⁵, ${\eta }_p^2$ = 0.24, n = 32) and washing (F_4,124 = 3.029, P =I 0.043, ${\eta }_p^2$ = 0.089, n = 32). Mixed ANOVAs used the within-participants factor of time and the between-participants factor of group (control/alpha, beta-gamma). Follow-up Bonferroni-corrected pairwise comparisons (right panels) in the beta-gamma group (In = 32) revealed a rapid reduction in hoarding (P = 0.02) and ordering (P = 0.012) by the last day of intervention, relative to baseline. These reduced scores remained reduced or continued to drop at 1 month (hoarding, P = 0.009; ordering, P = 0.01), 2 months (hoarding, P = 0.039; ordering, P = 0.002) and 3 months (hoarding, P = 1.9 × 10⁻⁵; ordering, P = 0.003). Of note, relative to baseline, there were significant reductions in checking on day 5 (P = 0.035), at 1 month (P = 0.008) and at 2 months (P = 0.015) after intervention. The alpha group (n = 32) showed no main effects of time (all F < 1.359, all P > 0.257) and no pairwise differences at any timepoint relative to baseline (all P > 0.228). b, Scatter plots with 95% confidence intervals of individual participant total OCI-R scores obtained before the intervention (pre-modulation) shown against the modulation-induced change in OCI-R (that is, baseline minus post-modulation measurement) for each post-intervention timepoint and for each modulation group. The beta-gamma group (n = 32) showed significant Pearson correlations (two-tailed, corrected for multiple comparisons) on day 5 (r₃₂ = 0.624, P = 1.35 × 10⁻⁴), 1 month (r₃₂ = 0.602, P = 2.71 × 10⁻⁴), 2 months (r₃₂ = 0.712, P = 5 × 10⁻⁶) and 3 months (r₃₂ = 0.651, P = 5.6 × 10⁻⁵). *P < 0.05; **P < 0.01; ***P < 0.001.