Orbitofrontal cortex control of striatum leads economic decision-making

Economic decision-making involves evaluating available options by integrating sensory information with predicted outcomes to derive subjective value, comparing these values, and selecting actions that maximize expected benefit. Prior studies have identified neural representations of subjective value in the orbitofrontal cortex (OFC), and microstimulation and other manipulations can bias choices, supporting a hypothesized role for OFC in value-based choices. Yet lesion and inactivation results have been mixed, and subjective value signals are also found in medial prefrontal cortex, dorsomedial striatum (DMS), and mediodorsal thalamus, with manipulations of these regions affecting choice behavior. How these regions interact to implement economic choices remains unclear, in part due to technical limitations in primate models. To address this, the authors adapted an economic decision-making task for rats, enabling simultaneous recording and precise manipulation of defined neural populations across brain regions during economic choices.

Multiple lines of work implicate OFC in encoding economic value and influencing choice (e.g., Padoa-Schioppa & Assad; Ballesta et al.), but reports on necessity from lesions/inactivations are contradictory. Other regions, including medial prefrontal cortex, DMS, and mediodorsal thalamus, also encode value-related signals and modulate decision-making when perturbed. Prior primate studies often emphasize goods-space representations in OFC, while rodent studies have reported spatial/action-space encoding in OFC and striatum. Projections from OFC to diverse targets (ventral tegmental area, basolateral amygdala, dorsal/ventral striatum) have roles in credit assignment, value encoding/retrieval, and updating action values. However, the causal contribution of specific OFC outputs, particularly OFC→DMS, to pre-outcome economic choice evaluation had been less defined.

• Subjects: Adult male and female Long–Evans rats. Water-scheduled. Random assignment to experimental groups. Ethical approval obtained.
• Behavioral task: A visually cued economic choice task. On each trial, two side-by-side drifting grating cues indicated reward identity (vertical vs horizontal: blackcurrant- vs lemon-flavored water) and reward size (cue size scaled with drops). After a 2 s cue period, rats nosepoked to the side of the chosen cue to receive reward at the center spout. Inter-trial interval 5–10 s. Training established cue–outcome mappings; full choice sessions used 15 cue combinations, up to 600 trials or 2.5 h. Accuracy defined as choosing the larger-volume reward. Preference scores quantified the drop-number difference at indifference (blackcurrant drops − lemon drops) per animal/session, with short- and long-term stability assessed.
• Control task: No-choice variant presenting a single cue; response required but no spatial choice, to test for nonspecific effects on perception, motor execution, or value recall.
• Optogenetic inactivation screen: After criterion performance, bilateral AAV8-hSyn:SwiChR++-EYFP injections into OFC, DMS, mediodorsal thalamus, or prelimbic cortex with optical fibers placed above targets. During the cue evaluation period, blue light (473 nm, 1 s) initiated SwiChR++ inhibition and red light (635 nm, 1 s) relieved it upon choice initiation. Control animals expressed EYFP only. Effects assessed on psychometric functions, choice latencies, and preference scores.
• Projection-specific inhibition: AAV8-hSyn:eNpHR3.0-NRN-EYFP injected into OFC; fibers implanted bilaterally over DMS or mediodorsal thalamus to inhibit OFC axon terminals specifically during cue evaluation using 594 nm light. Verified OFC projections with oScarlet tracing and whole-brain clearing.
• Electrophysiology: Wireless extracellular recordings from OFC and DMS in freely moving rats using 64-channel silicon probes on microdrives. Spike sorting with Kilosort2/Phy2. Task-modulated units identified via Wilcoxon rank-sum against baseline across ten 500-ms epochs with FDR correction (P<0.001). Linear regressions assessed modulation by task variables, including objective (size) and subjective value (derived from session-specific preferences). Decoding choice side with fourfold cross-validated linear SVMs; computed predicted choice parameter (distance from hyperplane) and cross-correlated OFC vs DMS signals around choice to assess temporal lead–lag relationships on correct vs incorrect trials. Unit counts matched across areas for decoding analyses.
• Statistics: ANOVAs (repeated-measures), t-tests, Pearson correlations; data reported as mean ± s.e.m. Full details in Supplementary Table 1.

• Task behavior: Rats integrated reward quantity and identity. They preferentially chose larger-volume rewards (n=42 rats; one-way repeated-measures ANOVA). Fraction of trials choosing the larger reward was 0.82 ± 0.01. Choice latencies were shorter on easy trials (large volume differences) than difficult trials. Individual juice preferences (blackcurrant vs lemon) were modest but stable across sessions and over ~4 months (significant Pearson correlations).
• Necessity of regions: Optogenetic inhibition of OFC impaired economic decision-making—flatter psychometric curves and slowed latencies on easy trials—and disrupted stable juice preferences (preferences during inhibition did not correlate with uninhibited preferences). Inhibition of DMS similarly flattened psychometric curves and slowed easy-trial latencies but did not alter juice preference scores. Inhibiting prelimbic cortex or mediodorsal thalamus had no detectable effect. EYFP controls showed no effect of illumination. In a no-choice control task, OFC or DMS inhibition did not affect response latency differences by reward size, indicating preserved perception, action execution, and value representations.
• Neural encoding: A large fraction of units were task-modulated (OFC: 1,157/1,329; DMS: 524/656; n=6 rats each). In both regions, encoding was dominated by spatial features (left/right offer amounts and chosen side). OFC neurons were more strongly modulated by subjective value than objective value; this effect was not present in DMS.
• Temporal dynamics and accuracy: Choice side could be decoded with high accuracy from both OFC and DMS activity; decoding accuracy rose earlier in OFC than DMS. Cross-correlations of predicted choice parameters showed OFC led DMS on correct trials (lag −23.93 ± 22.86 ms; OFC leads) but not on incorrect trials, where DMS led OFC (lag 44.82 ± 26.69 ms) (n=30 sessions, 5 rats). SVMs trained on correct trials predicted chosen side equally well on correct and incorrect trials, but the OFC-leading temporal pattern was absent on errors.
• Projection causality: Inhibiting OFC→DMS axon terminals during cue evaluation selectively impaired choices based on reward volume (flatter psychometric curves; increased latencies on easy trials) without altering juice preferences. Inhibiting OFC→mediodorsal thalamus had no effect. Projection inhibition did not affect performance in the no-choice control task. Overall, choice information emerges in OFC before DMS, and direct OFC→DMS transmission is necessary for accurate economic decision-making, particularly for volume-based (objective) value comparisons.

The study addressed how OFC and DMS interact to enable economic decision-making. Behavioral, causal, and electrophysiological evidence converged to show that both OFC and DMS are necessary, with neural encoding in both regions dominated by spatial/action features of the task. Critically, choice-related signals in OFC reliably preceded those in DMS, and the degree of OFC lead predicted choice accuracy, indicating a functional feedforward influence from OFC to DMS. Projection-specific inhibition established that OFC→DMS transmission is required to guide appropriate choices based on reward magnitude, whereas OFC→mediodorsal thalamus is not necessary in this context. The findings suggest that, in rodents, OFC operates in action space to resolve motivational conflict by relaying choice information to DMS, a hub for goal-directed action selection. Differences from primate literature emphasizing goods-space representations may reflect species or task demands. The stronger subjective-value modulation in OFC versus DMS, together with selective impairment of objective value processing by DMS or OFC→DMS inhibition, implies parallel OFC output pathways may differentially support subjective versus objective components of valuation.

This work demonstrates that accurate economic choices in rats depend on a temporally leading representation of choice in OFC that is transmitted to DMS. Both regions are necessary for decision-making, but only the OFC→DMS projection is causally required for volume-based choice computations, aligning with DMS’s role in goal-directed action. The study advances understanding of cortical–striatal mechanisms for value-based decisions and highlights OFC’s action-space processing in rodents. Future research should: (1) dissect contributions of distinct OFC subregions and cell types; (2) map how multiple OFC output pathways (e.g., to amygdala, ventral tegmental area, dorsal/ventral striatum) jointly support subjective versus objective value processing; (3) examine species and task differences (freely moving vs head-fixed) in goods- vs action-space representations; and (4) probe how OFC–DMS dynamics adapt with learning, uncertainty, and habitual control.

• Species and task generalizability: Findings are from freely moving rats in a specific visual choice paradigm; differences from primate goods-space encoding may reflect species or task specifics.
• Regional targeting: Only ventrolateral OFC was targeted; other OFC subregions may contribute differently.
• Projection specificity: Axon tracing quantification cannot distinguish terminals from fibers of passage; causal inhibition targeted terminals but off-target effects along fibers cannot be entirely excluded.
• Blinding and sample size: Data collection/analysis could not be fully blinded for practical reasons; sample sizes were not predetermined statistically (based on prior work).
• Lateralization/movement: Recordings in freely moving animals may influence spatial encoding and lateralization.
• Preference components: While OFC inhibition disrupted both objective and subjective components, DMS and OFC→DMS inhibition selectively affected objective value; the pathway(s) conveying subjective value were not identified here.
• Data/code availability: Primary data and code are available upon request rather than in a public repository.