Brain lesions disrupting addiction map to a common human brain circuit

Substance use disorders affect 8–10% of adults and are a leading cause of death in young people. Existing treatments have modest long-term success, motivating the search for new therapies such as deep brain stimulation, transcranial magnetic stimulation, and surgical lesioning. However, optimal neuroanatomical targets remain unclear. Rare cases of addiction remission following focal brain lesions (for example, after stroke) offer causal insight into human neuroanatomy of addiction remission; prior work implicated the insula but remission has been observed after lesions in diverse regions, leaving localization unresolved. Leveraging lesion network mapping with the human connectome can unify heterogeneous lesion locations by identifying a common brain network underlying a symptom or benefit. This study applies lesion network mapping to identify a brain circuit associated with remission of nicotine addiction and tests generalizability to alcohol addiction risk and specificity relative to other neuropsychological measures.

Prior lesion studies reported that insula damage increases the likelihood of nicotine addiction remission, while other studies highlighted additional regions such as basal ganglia or failed to replicate insula effects, indicating heterogeneity. Circuit-based models of addiction and neuroimaging studies implicate fronto-insular, cingulate, prefrontal, and striatal abnormalities in addiction, including imbalances between dorsal and ventral frontostriatal circuits. Neuromodulation trials (DBS, TMS, surgical lesioning) have targeted insula, anterior cingulate, and frontal cortex with mixed results; an FDA-cleared TMS coil for smoking cessation was designed to target multiple regions. Lesion network mapping has successfully linked disparate lesion locations to common networks and therapeutic targets in other disorders (for example, essential tremor), suggesting a path to refine targets for addiction.

Design: Retrospective analysis of two independent cohorts of active daily smokers at the time of focal brain damage (Iowa n = 67; Rochester n = 62). Outcome groups: addiction remission (quit immediately without difficulty, no relapse, no craving), quitting without remission, and not quitting. Lesion mapping: Lesions manually traced on CT/MRI, transformed to MNI space. Primary analyses:

• Voxel-based lesion-symptom mapping (VLSM) using NiiStat tested associations between lesion location and outcome; study site and lesion size as covariates; reproducibility assessed across cohorts.
• Lesion network mapping (LNM): For each lesion mask (n = 129), seed-based resting-state functional connectivity was computed using two connectomes: a smoker connectome (n = 126) and a normative connectome (n = 1,000). For each seed, individual maps were averaged to create a lesion network. A voxel-wise general linear model related lesion connectivity to outcome, with site as covariate. Multiple comparisons were controlled via threshold-free cluster enhancement in FSL (P < 0.05). Robustness checks included varying smoking cutoffs, covarying lesion size, gray/white matter proportion, age, restricting to strokes or MRI-defined lesions, using a connectome without global signal regression, and mediation by insula damage.
• Dorsal/ventral circuit analysis: Defined dorsal versus ventral striatal ROIs; computed lesion connectivity to these ROIs and tested ROI × group interactions. Structural connectivity analyses:
• Tract-based: Mapped lesions onto 68 white-matter pathways (BCBtoolkit Tractotron); GLM (PALM) tested associations between tract damage and outcome (site and lesion size as covariates), correcting for multiple comparisons.
• Disconnectome: For each lesion, constructed structural disconnection maps using diffusion MRI from 178 controls; computed a disconnectome score reflecting overlap with positive nodes of the addiction remission network; tested association with outcome using PALM (one-tailed, remission > not quitting). Generalizability and specificity:
• Independent cohort (Vietnam head injury study, n = 186) with MAC (MMPI alcoholism proneness) scores: Computed lesion connectivity maps and correlated connectivity with MAC, generating an alcoholism risk network. Spatial correlation and permutation testing assessed similarity to the smoking addiction remission network. Specificity tested against 10 other MMPI domains and 27 additional neuropsychological variables (total 37).
• Case reports: Identified lesions reported to disrupt non-nicotine addictions; computed connectivity and compared to the smoking remission network. Target identification:
• For each brain voxel, computed a connectivity profile and correlated it with the addiction remission network to map voxels best matching (positive) or opposing (negative) the profile (threshold r > 0.75 or r < −0.75). Compared with prior targets (insula, anterior cingulate) and electric field models of BrainsWay H4 (smoking) and H7 (alcoholism) TMS coils.

• Lesion locations associated with smoking addiction remission were highly heterogeneous; VLSM did not identify significant voxels (P > 0.2) nor replicate across cohorts. • LNM revealed a specific connectivity pattern for addiction remission: stronger positive connectivity to the dorsal cingulate, dorsolateral/lateral prefrontal cortex, and insula, and stronger negative connectivity to medial prefrontal and temporal cortex. Findings were significant after family-wise error correction (PPWE < 0.05) and consistent using smoker or normative connectomes. • Reproducibility: The remission connectivity pattern was similar across cohorts (spatial r = 0.58, permutation P = 0.03) and correlated with remission across cohorts (P = 0.04). Robust to multiple covariates and analytic variants, including lesion size, gray/white matter composition, age, lesion type (stroke-only), imaging modality (MRI-only), and preprocessing (without global signal regression). Mediation by insula damage did not account for the network effect. • Circuit dissociation: Lesion connectivity showed a dorsal–ventral striatal dissociation (ROI × group interaction F(1,101) = 8.4, P = 0.005) and complementary dorsolateral (positive) versus ventromedial (negative) prefrontal patterns. • Neuropsychology: No significant group differences in IQ, executive function, working memory, verbal comprehension, mood, or social introversion; the remission network from this subset matched the full dataset. • Structural connectivity: Damage to 14 tracts associated with remission; two left fronto-insular tracts survived multiple-comparison correction and were within 2 mm of the peak of the functional remission network. Lesions causing remission had higher disconnectome scores between remission-network nodes than non-quitters (P = 0.003). • Generalizability: In the independent cohort, the alcoholism risk network closely matched the smoking remission network (spatial r = 0.65, permutation P = 0.04; controlling for smoking status r = 0.69, P = 0.04). Similarity was driven by connectivity, not lesion location (VLSM spatial r = −0.15, P = 0.83), and was specific to addiction metrics: among 37 other variables, only the alcoholism risk map exceeded chance similarity (one-sided P = 0.04). • External cases: Three case reports of lesions disrupting non-nicotine addictions showed connectivity similar to the smoking remission network (P < 0.05). • Therapeutic targets: Voxels best matching the remission profile identified hubs in the paracingulate gyrus and left frontal operculum/adjacent insula; the strongest negative hub was in the medial fronto-polar cortex, overlapping peak electric fields of FDA-cleared H4 TMS for smoking and H7 TMS used in alcohol dependence.

The study demonstrates that addiction remission following focal brain lesions is not tied to a single anatomical site but to a distributed brain network with specific connectivity features aligned with circuit-based models of addiction. The remission network generalizes from nicotine to alcohol addiction risk, suggesting a shared addiction circuit across substances and specificity to addiction-related measures relative to broader neuropsychological profiles. The observed dorsal–ventral dissociations in striatum and prefrontal cortex are consistent with models positing network imbalances underlying addiction vulnerability and relapse, implying that modulating this balance may promote remission. Convergent functional and structural connectivity analyses support the network’s biological plausibility, including fronto-insular pathways. Importantly, the connectivity profile provides a data-driven approach to refine therapeutic targets for neuromodulation: sites matching the positive profile may be lesion/suppression targets, while sites with the opposite profile may be optimal for excitatory stimulation. The overlap between identified hubs and existing TMS coil fields (fronto-polar cortex) and surgical lesion targets (anterior cingulate vicinity) underscores translational relevance.

Lesions that disrupt addiction converge on a specific human brain circuit characterized by positive connectivity to insula, dorsal cingulate, and lateral prefrontal regions and negative connectivity to medial prefrontal and temporal regions. This circuit is reproducible across cohorts, aligns with structural disconnection patterns, and generalizes to reduced alcoholism risk while remaining specific to addiction-related metrics. Network-based target mapping identifies paracingulate gyrus and left frontal operculum/insula as potential lesion/suppression targets and medial fronto-polar cortex as a candidate for excitatory stimulation, overlapping with effective TMS configurations. Future work should prospectively test these targets in neuromodulation trials, assess side-effect profiles, integrate patient-specific functional and diffusion imaging, and refine multimodal approaches combining structural and functional connectomics.

• Retrospective design with heterogeneous lesion etiologies and locations; reliance on group-level connectomes as proxies for individual connectivity may limit precision. • Addiction remission was assessed only for nicotine in the primary cohorts; the independent cohort lacked pre-injury substance-use data, permitting analysis of alcoholism risk but not remission. • Although robust across multiple controls and covariates, unmeasured behavioral or clinical variables could differ between groups and influence outcomes. • Structural and functional connectivity provide complementary but distinct information; optimal integration remains to be determined. • Proposed neuromodulation targets are inferred from connectivity matching; therapeutic efficacy and side-effect profiles require prospective validation.