Testing both directions of a two-way circuit
The cortico-subthalamic hyperdirect pathway is a fast route by which motor cortex can influence basal ganglia processing, and it is a key target of deep brain stimulation (DBS) in Parkinson's disease. Whether dopamine changes signal transmission along this route in humans was the open question. The study's answer depends on timing: the earliest response survived medication changes, the later one did not.
The design is unusual because it stimulated and recorded in both directions within the same patients. Fifteen patients with idiopathic Parkinson's disease undergoing bilateral DBS electrode implantation had a subdural electrocorticography (ECoG) strip placed over sensorimotor cortex. In the days after surgery, while leads were still externalized, the team delivered direct electrical stimulation through the cortical strip and recorded evoked responses in the subthalamic nucleus (STN) local field potentials, then reversed the direction by stimulating the STN lead and recording cortical responses.
An evoked response here means neural activity time-locked to each stimulation pulse, averaged across pulses. Stimulating cortex probes orthodromic drive down the pathway; stimulating the STN probes the antidromic cortical response. Because both were measured in the same individuals, the comparison is within-patient rather than across groups.
Medication ON and OFF inside the same patients
Recordings were repeated in two dopaminergic states: ON usual clinical levodopa dosing, and OFF after withdrawing all dopaminergic medication for at least 12 hours. The withdrawal was clinically meaningful, with MDS-UPDRS III motor scores of 25.9 ± 9.7 OFF versus 16.6 ± 8.9 ON (p < 0.001).
Paired analyses for cortical stimulation used the 10 patients with both ON and OFF recordings; 2 contributed ON-only and 3 OFF-only data. For STN stimulation, 9 paired datasets were available. The order of ON and OFF sessions was not counterbalanced, because recordings happened early postoperatively and the ON state was prioritized to reduce patient burden during recovery. That design choice matters for how the paired comparisons should be read, though the comparisons themselves are within-patient.
The earliest subthalamic response is dopamine-insensitive
Cortical stimulation produced an earliest STN deflection at 2.6 ± 0.7 ms, followed by medium-latency components at 10.1 ± 1.9, 23.9 ± 3.4 and 38.9 ± 3.5 ms. The first deflection differed significantly from baseline across all patients (p = 0.002) and appeared in nearly all recorded STN channels with consistent morphology. Its latency and shape are what the authors take as evidence of monosynaptic hyperdirect activation.
Dopamine did not change this early peak in any meaningful way. Its amplitude was −3.3 ± 1.7 µV OFF versus −2.6 ± 1.4 µV ON in the 10 paired patients (p = 0.06), a weak difference with a less negative deflection ON, and latency was essentially identical (2.7 ± 0.6 ms OFF versus 2.6 ± 0.8 ms ON, p = 0.36).
The reverse direction behaved the same way. STN stimulation evoked a cortical short-latency component at 3.2 ± 0.3 ms that differed from baseline (p = 0.003), and its amplitude and latency did not differ between medication states (−18.2 ± 9.7 µV OFF versus −22.1 ± 9.8 µV ON, p = 0.367; latency p = 0.288, n = 9). The cortical short-latency latency did not differ significantly from the earliest STN response to cortical stimulation (p = 0.109), which the authors read as reciprocal fast transmission in both directions.
Dopamine amplifies the late subthalamic response
The long-latency picture is different. Cortical stimulation produced a robust STN response between 40 and 100 ms, peaking at 79.9 ± 8.9 ms and deviating from baseline (p = 0.002). In the medication ON state this late peak was more negative and longer-lasting, which the authors call amplification: 1.5 ± 1.2 µV OFF versus −2.2 ± 2.2 µV ON (unpaired n = 13 OFF vs n = 12 ON, p < 0.001; paired n = 10, p = 0.002).
A cluster-based permutation test corrected for multiple comparisons confirmed a longer-lasting negative response ON, restricted to the long-latency window and apparent between 65 and 85 ms. The team also checked whether stimulation frequency (1 Hz in 4 patients versus 5 Hz in 11) explained the effect.
The cortical response to STN stimulation did not show this modulation, so the asymmetry is directional as well as temporal: the slow subthalamic response to cortical input is the dopamine-sensitive element, while the cortical response to subthalamic input stays comparatively stable.
A fast volley and a slower network response
The authors interpret the two components as functionally distinct. The short-latency response is consistent with a fast, likely monosynaptic hyperdirect volley that is hard-wired enough to be unaffected by dopamine. The later response is consistent with polysynaptic or network-level processing in the basal ganglia, which prior non-human primate work has linked to striatal recruitment and indirect pathway signalling.
That mechanistic reading is inference, not direct measurement. The paper states plainly that the specific pathways and synaptic mechanisms behind the long-latency response cannot be resolved from these data, and that the indirect nature of the measurements warrants caution. The authors also note that a contribution of stimulation artefact or volume conduction to the short-latency responses cannot be entirely excluded, though the responses occurred after the artefact ended, persisted under polarity reversal, and showed spatial specificity.
One proposed consequence is speculative and flagged as such: the shorter-lasting OFF-state response, with faster return to baseline, could permit the exaggerated beta-band resonance seen in Parkinson's disease, since pathological synchrony often relies on recurrent network dynamics. The paper calls this framework plausible but inferential and in need of further investigation.
Where the response is largest tracks the clinical target
The responses were not spatially uniform. STN contacts closer to a dorsolateral STN sweet spot, defined from previously published probabilistic mapping of effective stimulation sites, showed more negative evoked amplitudes, with amplitude becoming progressively less negative as distance increased (Pearson r = 0.527, p = 3.981 × 10⁻⁶). Contacts with the largest ON–OFF differences in the short-latency response were mainly in the dorsolateral STN.
This spatial pattern carried a clinical signal. The contact with the largest short-latency STN amplitude matched the electrode level chosen for chronic DBS therapy in 12 of 15 patients (80%), against a chance level of 0.25 under a simplified equal-probability model (two-sided binomial test, p < 0.001). Agreement was assessed at the contact level, not individual segments, because clinical stimulation often co-activated multiple segments within a level. Long-latency amplitudes showed no such clinical correlation.
The authors present this as a possible biomarker for guiding contact selection, but explicitly caution that it rests on a simplified choice model and a limited sample and needs validation in larger prospective studies.
What the dissociation does and does not establish
The result is a within-patient demonstration in 15 patients with predominantly akinetic-rigid Parkinson's disease, recorded in the early postoperative window through externalized leads. It establishes that dopaminergic state changes the late subthalamic response to cortical stimulation while sparing the early one, and that the cortical response to subthalamic stimulation is comparatively stable. It does not identify which polysynaptic routes generate the late response, and the small cohort meant sex and gender were not modeled as covariates.
The practical priority follows from the split rather than from either component alone. If the goal is to track dopamine-sensitive network state, the long-latency subthalamic response between roughly 65 and 85 ms is the informative measure, since it moved with medication while the fast component did not. If the goal is to locate a stimulation target, the short-latency amplitude is the more useful signal, given its 80% level match with the clinically chosen contact. Treating either measure as a general index of pathway engagement would blur the distinction the paper is built on.
Sources used in this answer
Dopamine amplifies subthalamic long-latency responses to cortical stimulation in patients with Parkinson’s disease
In 15 Parkinson's patients, cortical stimulation evoked a short-latency STN response (2.6 ± 0.7 ms) that was preserved across medication states, while a long-latency response (~80 ms) was amplified by dopamine; cortical responses to STN stimulation were unchanged.