The human VGLUT3-pT8I mutation elicits uneven striatal DA signaling, food or drug maladaptive consumption in male mice

Persistent habitual behaviors and compulsion are common features of substance use disorders (SUDs) and eating disorders (EDs). The dorsal striatum is central for the shift from goal-directed actions to habitual and compulsive behaviors. Cholinergic striatal interneurons (ChIs), which co-express the vesicular acetylcholine transporter (VAChT) and VGLUT3, modulate striatal circuits via both ACh and glutamate. VGLUT3-dependent glutamate enhances vesicular ACh loading (vesicular synergy). Prior human genetics identified rare VGLUT3 variants, including a heterozygous SLC17A8 p.T8I mutation, in patients with SUDs and in one case with comorbid cocaine addiction and bulimia nervosa. The present study replicates screening for SLC17A8 variants in new patient samples and develops a VGLUT3-p.T8I knock-in mouse to determine molecular, circuit-level, and behavioral consequences. The overarching hypothesis is that p.T8I impairs vesicular synergy leading to reduced striatal cholinergic tone, uneven dorsal striatal dopamine (DA) transmission, accelerated habit formation, and vulnerability to maladaptive consumption of drugs or palatable food, with potential rescue by augmenting cholinergic tone.

ChIs constitute 1–2% of striatal neurons but form dense varicosity plexuses and co-release ACh and glutamate. ACh released from ChIs stimulates DA efflux via nicotinic and muscarinic receptor interactions on DA fibers, whereas ChI-derived glutamate inhibits DA release via metabotropic glutamate receptors. This bidirectional ACh/glutamate control over DA release is prominent in nucleus accumbens and dorsomedial striatum (DMS) but not dorsolateral striatum (DLS). Selective ablation of VAChT in ChIs produces facilitation of habit formation and vulnerability to maladaptive eating with relatively minor effects on psychostimulant-induced locomotion, while VGLUT3 deletion increases anxiety and sensitivity to cocaine reinforcement. Previous work revealed increased frequency of rare VGLUT3 variants in SUD patients, including p.T8I, suggesting VGLUT3 involvement in addiction-related phenotypes. Theoretical frameworks propose that imbalances between goal-directed and habitual systems contribute to EDs and SUDs. Additionally, nicotinic receptor genetics (e.g., CHRNA5 polymorphisms) have been implicated in addiction and possibly eating behaviors, supporting a cholinergic-dopaminergic interface in compulsivity.

Human genetics and clinical phenotyping: Two samples were investigated. Sample #1 (EDs sample) included 270 female patients (anorexia nervosa, bulimia nervosa, EDNOS) and 71 healthy controls recruited at the Douglas Hospital Eating Disorders Program with structured interviews (Eating Disorders Examination) and DSM-IV-TR diagnoses. Sample #2 (SUDs sample) included 524 outpatients (78% men) genotyped by DNA array for rare and frequent variants in SLC17A6 (VGLUT2), SLC17A7 (VGLUT1), and SLC17A8 (VGLUT3), including p.T8I. Phenotyping used SCID-IV and the Scale for Assessment of Positive Symptoms–Cocaine-Induced Psychosis (SAPS-CIP). Case-control analyses compared p.T8I allele frequencies to gnomAD; within-cases analyses examined associations with SUD phenotypes. QC steps included removal of markers/samples with >2% missingness, ancestry assignment, HWE checks, and relatedness filtering. Statistical tests included Fisher’s exact/Chi-squared, Kruskal–Wallis, Wilcoxon tests, Bonferroni corrections, and Bayesian factors where appropriate. Animals: Male VGLUT3T8I/T8I knock-in mice (C57BL/6N background) were generated by introducing a point mutation in exon 1 (ACC→ATC; T→I at position 8). Genotyping used PCR with specified primers. Mice (2–8 months) were housed under standard conditions. Procedures adhered to European, Canadian, and local animal care guidelines. Molecular/anatomical assays: Immunofluorescence in primary hippocampal neurons and brain sections assessed VGLUT3 and VGLUT3-p.T8I expression/localization; colocalization with bassoon and MAP2 was evaluated. Immunoautoradiography quantified regional VGLUT3 distribution. STED microscopy on striatal synaptic vesicle preparations evaluated nearest-neighbor distances and co-localization rates between VGLUT3 and VAChT. Homology modeling used DgoT structure to predict effects of T8I on VGLUT3 3D conformation. Vesicular transport assays: Synaptic vesicles from mouse striatum were isolated to measure [3H]glutamate and [3H]ACh uptake, with/without exogenous glutamate (1 mM) to assess vesicular synergy. Electrophysiology: In vitro autapse recordings from VGLUT1−/− hippocampal neurons rescued with VGLUT3 or VGLUT3-p.T8I lentiviral vectors measured mEPSC amplitude, frequency, and charge to assess glutamate release properties and release probability. In vivo ACh dynamics: Fiber photometry with the GACh4.3 biosensor expressed in DMS recorded spontaneous ACh events in freely moving mice; event detection used rolling median and MAD-based thresholds, generating z-scored ΔF/F. In vivo DA release: High-speed chronoamperometry measured KCl-evoked DA release in DMS and DLS via Nafion-coated carbon fiber electrodes; analysis focused on maximal ΔDA after local KCl ejection. Behavioral assays: Basal locomotor activity over 24 h; anxiety-like behavior via elevated plus maze and elevated O-maze; marble-burying. Cocaine operant self-administration (FR1→FR3), progressive ratio, extinction, and cue-induced reinstatement. Goal-directed vs habit assessment via sucrose FR1 training and outcome devaluation (sensory-specific satiety). Food addiction model: long-term operant self-administration of chocolate-flavored pellets under FR1 then FR5 with tests for persistence (responding during non-reinforced period), motivation (progressive ratio breaking point), and compulsivity (footshock-paired reinforcement). Binge-like sucrose overconsumption (daily restricted access with two-bottle choice; first-hour vs 4-hour intake; chow preload control). Activity-based anorexia (ABA): running wheel access with progressive food restriction; survival defined as maintaining ≥75% baseline body weight. Pharmacology: donepezil 0.3 mg/kg i.p. daily during ABA in WT and mutants. Statistics: normality/variance checks; two-sided tests; t-tests/Mann–Whitney, Kruskal–Wallis with Dunn’s post hoc, one-/two-way (RM) ANOVAs with Bonferroni/LSD contrasts, Kaplan–Meier with Mantel–Cox and Gehan–Breslow–Wilcoxon tests.

- Human genetics: In the combined EDs+SUDs clinical samples, heterozygous SLC17A8 p.T8I was found in 9/793 cases (1.1%). Compared with gnomAD, the p.T8I minor allele frequency (0.8%) was significantly enriched in patients (Fisher’s exact p = 6.12 × 10^−5). No sex imbalance among carriers (Fisher exact p = 1). In cocaine use disorder (CUD) patients, SLC17A8 mutations were associated with SAPS total and delusion subscale differences (Kruskal–Wallis chi-squared = 8.16, df = 2, p = 0.0169 total). p.T8I carriers showed higher SAPS total scores versus both non-carriers and carriers of other SLC17A8 missense variants (Wilcoxon p = 0.0179 and p = 0.0477). SLC17A7/SLC17A6 missense variants showed no SAPS association (Wilcoxon p = 0.92). Alcohol use disorder prevalence trended higher in non-mutated vs p.T8I vs other VGLUT3 variants (trend test p = 0.013); presence of p.T8I was associated with decreased prevalence of alcohol use disorders (Fisher’s exact p = 0.0303). - VGLUT3 expression/targeting: VGLUT3-p.T8I displayed normal subcellular distribution in cultured neurons, colocalizing with bassoon; brain immunoautoradiography and immunofluorescence showed comparable regional and cellular expression patterns between WT and VGLUT3T8I/T8I mice. STED analysis indicated similar nearest-neighbor distance distributions and ~40% co-localization of VAChT and VGLUT3 on synaptic vesicles in WT and mutants (Chi-squared p = 0.439; K–S p = 0.218). - Structure/function: Homology modeling predicted minimal impact of T8I on VGLUT3 overall structure and N-terminus positioning relative to the transport pore. - Glutamate handling unchanged: Striatal vesicular [3H]glutamate uptake was similar in WT and VGLUT3T8I/T8I (350.7 ± 38.7 vs WT 330.3 ± 50 pmol·mg−1·10 min−1; ANOVA WT vs T8I Tukey p = 0.851), while reduced in VGLUT3-KO (187.7 ± 34.7; WT vs KO p = 0.005). In VGLUT1−/− autapses, mEPSC amplitude, frequency, and charge did not differ between VGLUT3 and VGLUT3-p.T8I rescues (Dunn’s post hoc all p > 0.999), indicating preserved glutamate release probability/RRP cycling. - Vesicular synergy and ACh signaling reduced: In striatal vesicles, glutamate (1 mM) increased [3H]ACh uptake by +113% in WT (p < 0.001) and +74% in VGLUT3T8I/T8I (p = 0.044), with WT > mutant under Glu+ (p = 0.003); basal uptake (Glu−) did not differ (p > 0.999). In vivo GACh4.3 photometry in DMS revealed similar event amplitudes (Wilcoxon p = 0.382) but ~2× lower spontaneous event frequency in mutants (unpaired t-test p = 0.009; Wilcoxon W = 43.5, p = 0.017) and longer inter-event intervals (K–S p = 0.001). - DA signaling uneven across dorsal striatum: KCl-evoked DA release was reduced in DMS by ~31% in mutants (unpaired t-test p = 0.03) with no difference in DLS (p = 0.849). - Basal activity/anxiety: 24 h locomotor profiles were similar (genotype × time F30,420 = 0.794, p = 0.776). Elevated plus maze, O-maze, and marble-burying showed no genotype differences, indicating no overt anxiety alteration. - Cocaine self-administration: Acquisition under FR1/FR3 showed similar active nosepokes overall (genotype F1,23 = 1.641, p = 0.213), though mutants obtained fewer infusions on FR1 days 3–4 (LSD day 3 p < 0.05; day 4 p < 0.01) and fewer mutants reached learning criteria (WT 71% vs T8I 36%; Chi-squared = 6.627, p < 0.05). Progressive ratio breaking points were similar. Extinction proceeded similarly. Cue-induced reinstatement: overall reinstatement in both genotypes; mutants showed higher responding in the second half of the session (genotype × time F11,253 = 2.628, p < 0.01; post hoc 70 and 80 min p < 0.001; 90 min p < 0.05; 100 min p < 0.05), indicating greater relapse vulnerability to drug-associated cues. - Habit bias (outcome devaluation): After 16 FR1 sessions for sucrose, WT reduced responding in devalued vs valued conditions (42.8 ± 6.5 vs 21.0 ± 5.2; paired t-test p = 0.001), whereas mutants did not (36.3 ± 5.95 vs 36.8 ± 9.5; p = 0.945), indicating accelerated habit formation (two-way RM ANOVA Genotype × Value p = 0.0102). - Food addiction criteria: During long-term chocolate pellet self-administration (FR1→FR5), nosepoking did not differ (genotype F1,25 = 0.059, p = 0.811). Persistence, motivation (PR), and compulsivity (shock) showed no significant genotype differences; proportions categorized as addicted vs non-addicted were similar (p = 0.918). - Binge-like sucrose overconsumption: Across 16 days, total 4 h intake was similar (genotype F1,18 = 0.96, p = 0.34). From day 7 onward, mutants consumed more during the first hour (H0–H1; genotype F1,5,270 = 3.066, p < 0.001), with no difference during H1–H4 (genotype F1,18 = 0.027, p = 0.872). Two-bottle sucrose preference was similar (p = 0.341) and chow preload did not abolish the elevated first-hour intake (unpaired t-test p < 0.001), indicating binge-like behavior independent of hunger. - Activity-based anorexia (ABA): During progressive food restriction with running wheels, mutants exhibited greater reductions in food intake and body weight on days 6–8 (food intake genotype F1,18 = 5.437, p = 0.032; body weight genotype F1,18 = 7.364, p = 0.014) and a higher proportion crossing <75% baseline body weight (Kaplan–Meier Mantel–Cox/Gehan–Breslow–Wilcoxon p < 0.001), indicating increased self-starvation vulnerability. - Donepezil rescue in ABA: Daily donepezil (0.3 mg/kg) had no effect in WT (Kaplan–Meier p = 0.819) but significantly reduced the proportion of mutants falling below 75% body weight by improving food intake (Kaplan–Meier p = 0.006; Gehan–Breslow–Wilcoxon p = 0.004), rescuing the self-starvation phenotype.

The study integrates human genetics with a humanized mouse model to demonstrate that the SLC17A8 p.T8I variant is enriched in clinical populations with SUDs and associates with increased psychotic symptom severity in CUD. Mechanistically, p.T8I leaves VGLUT3 expression, targeting, glutamate vesicular uptake, and glutamate release intact, but partially blunts vesicular synergy, reducing ACh vesicular loading and spontaneous ACh release in DMS. This cholinergic hypofunction yields an asymmetric DA signaling profile—reduced KCl-evoked DA release in DMS with preserved DLS—consistent with prior models in VAChTcKO mice and with the known ACh/glutamate control over DA release differing across dorsal striatal compartments. The resulting DMS–DLS imbalance is sufficient to bias action control toward habits, facilitating cue-induced relapse to cocaine seeking and promoting maladaptive eating phenotypes, including binge-like first-hour sucrose overconsumption and enhanced vulnerability to ABA-induced self-starvation. Pharmacologically, enhancing cholinergic tone with the acetylcholinesterase inhibitor donepezil corrected the self-starvation vulnerability in mutants, supporting causality of cholinergic deficits and suggesting a translational intervention. The findings align with broader cholinergic-dopaminergic frameworks of compulsivity and may generalize beyond p.T8I to other polymorphisms (e.g., CHRNA5) that modulate cholinergic control of striatal DA. Collectively, the data propose that uneven dorsal striatal DA transmission (DMS < DLS) driven by impaired ACh vesicular synergy represents a shared circuit mechanism contributing to compulsive features across SUDs and EDs.

The VGLUT3-p.T8I variant, identified in patients with SUDs and one with comorbid bulimia, induces a selective deficit in striatal cholinergic vesicular synergy without impairing glutamatergic function. This reduces ACh release and DMS DA output while sparing DLS, biasing behavior toward habit formation and increasing vulnerability to cue-induced cocaine relapse, binge-like sucrose intake, and self-starvation in an ABA paradigm. Augmenting ACh tone with donepezil rescues the self-starvation phenotype, nominating pro-cholinergic strategies as potential treatments for EDs (notably anorexia nervosa) and possibly other compulsive disorders. Future work should elucidate the molecular partners and mechanisms by which VGLUT3’s N-terminus mediates vesicular synergy, assess sex differences and environmental interactions, validate these mechanisms across diverse human populations (especially underrepresented ancestries), and evaluate cholinesterase inhibitors or other pro-cholinergic agents in controlled clinical trials.

Human genetics analyses involved a relatively small number of p.T8I carriers, limiting power and generalizability; ancestry-related reference panel gaps, particularly among African and admixed populations, constrain precise population frequency estimates. Clinical associations (e.g., SAPS differences) may be influenced by unmeasured confounds despite controls for route/dose of cocaine. The mouse experiments used only males; potential sex-specific effects require study. Translational relevance from mouse models (e.g., ABA, binge-like sucrose) to human EDs has inherent constraints. While donepezil reversed mutant self-starvation, safety/tolerability concerns (e.g., cardiovascular and gastrointestinal side effects) necessitate cautious clinical translation. The precise molecular mechanism linking VGLUT3’s N-terminus to vesicular synergy remains unresolved.