Blockade of the μ-opioid receptor (MOR) by naltrexone reduces relapse risk in a subpopulation of alcohol-dependent patients. Previous positron-emission-tomography (PET) studies using the MOR ligand [11C]carfentanil have found increased MOR availability in abstinent alcoholics, which may reflect either increased MOR expression or lower endogenous ligand concentration. To differentiate between both effects, we investigated two cohorts of alcoholic subjects using either post-mortem or clinical PET analysis. Post-mortem brain tissue of alcohol-dependent subjects and controls (N=43/group) was quantitatively analyzed for MOR ([3H]DAMGO)-binding sites and OPRM1 mRNA in striatal regions. [11C]carfentanil PET was performed in detoxified, medication free alcohol-dependent patients (N=38), followed by a randomized controlled study of naltrexone versus placebo and follow-up for 1 year (clinical trial number: NCT00317031). Because the functional OPRM1 variant rs1799971:A>G affects the ligand binding, allele carrier status was considered in the analyses. MOR-binding sites were reduced by 23-51% in post-mortem striatal tissue of alcoholics. In the PET study, a significant interaction of OPRM1 genotype, binding potential (BPND) for [11C]carfentanil in the ventral striatum, and relapse risk was found. Particularly in G-allele carriers, lower striatal BPND was associated with a higher relapse risk. Interestingly, this effect was more pronounced in the naltrexone treatment group. Reduced MOR is interpreted as a neuroadaptation to an alcohol-induced release of endogenous ligands in patients with severe alcoholism. Low MOR availability may explain the ineffectiveness of naltrexone treatment in this subpopulation. Finally, low MOR-binding sites are proposed as a molecular marker for a negative disease course.