Network-targeted TMS modulates task-related striatal activity during motor skill learning

This study demonstrates that applying continuous theta-burst stimulation (cTBS) to the dorsolateral prefrontal cortex can indirectly suppress task-related striatal activity during motor skill learning without altering overall performance, suggesting a potential therapeutic approach for disorders characterized by hyperactive striatal responses.

The Gist

The Big Picture: Trying to "Tune" the Brain's Learning Engine

Imagine your brain is a massive, high-tech city. To learn a new skill (like playing a piano song or, in this case, moving a cursor on a screen with your fingers), different parts of the city need to talk to each other.

Two main neighborhoods are crucial here:

1. The "Boss" (DLPFC): Located in the front of your forehead. This is the planning department. It decides what to do and keeps you focused.
2. The "Engine Room" (The Striatum): Located deep underground in the center of the brain. This is where the actual muscle memory and habit formation happen. It's the engine that turns the plan into action.

The Problem: The "Engine Room" is buried deep underground. You can't stick a wire into it to turn the volume up or down without surgery. It's like trying to fix the engine of a car while the hood is welded shut.

The Solution: The researchers used a clever trick called Network-Targeted TMS. Instead of trying to hit the deep engine directly, they used a magnetic "remote control" (Transcranial Magnetic Stimulation) on the "Boss" (the forehead). The idea is: if you tweak the Boss, the signal travels down the wires to the Engine Room, and the Engine changes its behavior.

---

The Experiment: The "Cursor Game"

The Setup:

• The Players: 50 healthy young adults.
• The Task: They wore a special glove and had to move a cursor on a screen to hit targets using only their finger movements. It sounds simple, but it requires complex coordination between the brain and the hand.
• The Groups: The participants were split into three groups, each getting a different "tuning" of the Boss (DLPFC):
  1. The "Mute" Button (cTBS): A specific magnetic pattern designed to quiet down or inhibit the brain area.
  2. The "Volume Up" Button (20 Hz): A pattern designed to excite or wake up the brain area.
  3. The "Control" Group: No magnetic tuning at all (just the game).

The Goal: See if turning the "Boss" up or down changes how the "Engine Room" (the striatum) behaves while they learn the game.

---

The Results: The Brain vs. The Behavior

Here is where it gets interesting. The researchers expected that if they changed the brain's activity, the people would get better or worse at the game.

1. The Game Score (Behavior): No Change
Surprisingly, everyone played the game about the same. Whether the "Boss" was quieted down, turned up, or left alone, the participants learned the task at roughly the same speed.

• Analogy: Imagine you have a car with a slightly different engine tune. You might expect it to drive faster or slower, but on this specific test track, all the cars finished the lap in the same amount of time.

2. The Brain Scan (fMRI): A Big Difference
Even though the game scores were the same, the brain scans told a different story.

• The "Mute" Button (cTBS on DLPFC): When they quieted down the Boss, the Engine Room (specifically the left anterior caudate, a part of the striatum) became much less active. It was like the engine was idling quietly instead of revving high.
• The "Volume Up" Button (20 Hz): Turning the Boss up did not make the Engine Room rev any louder. It had no effect.
• The Control: The group with no magnetic tuning showed the usual pattern of brain activity as they learned.

The Timing: The effect of the "Mute" button didn't happen instantly. It took about 20–30 minutes after the magnetic treatment for the brain activity to drop. It's like a slow-acting medication; you take the pill, and the effect kicks in a little later.

---

What Does This Mean? (The Takeaway)

1. You can control deep brain areas from the surface.
This study proves that you don't need surgery to influence the deep "Engine Room" of the brain. By targeting the "Boss" (the forehead) with the right magnetic pattern, you can successfully dial down the activity in the deep striatum.

2. "Quiet" works better than "Loud" here.
The "Mute" button (inhibitory stimulation) worked great. The "Volume Up" button (excitatory stimulation) didn't do much. This suggests that for this specific type of learning, the brain might be more sensitive to being calmed down than being hyped up.

3. Brain activity $\neq$ Game performance.
Just because the brain activity changed doesn't mean the person's performance changed. The brain might be working differently (using less energy or a different strategy), but the result on the screen looks the same. This is important because it shows that brain scans can reveal hidden changes that a simple test score misses.

Why Should We Care?

The researchers suggest this could be a big deal for treating diseases.

• Addiction: Addiction is often linked to a "hyperactive" striatum (the Engine Room is revving too high, craving rewards). If we can use this "Mute" button to calm that area down, it might help treat addiction.
• Parkinson's: Conversely, Parkinson's involves a striatum that is too quiet. While this study showed that the "Volume Up" button didn't work, it opens the door to trying different "tuning" methods to wake up that engine.

In a nutshell: The researchers found a way to use a magnetic remote control on the forehead to successfully "turn down the volume" on the deep brain center responsible for habits and learning, without needing to open the skull. It's a promising new tool for understanding and treating brain disorders.

Technical Summary

1. Problem Statement

The striatum is a critical subcortical hub for motor skill learning, but its deep location makes it difficult to modulate directly using non-invasive brain stimulation. While Transcranial Magnetic Stimulation (TMS) can indirectly influence subcortical regions by targeting functionally connected cortical areas (the "network-targeted" approach), it remains unclear whether this method can modulate task-evoked striatal activity during the acquisition of complex motor skills (which impose higher cognitive demands than simple tasks). Furthermore, the specific causal roles of different cortical nodes (Dorsolateral Prefrontal Cortex [DLPFC] vs. Primary Motor Cortex [M1]) and stimulation protocols (inhibitory vs. excitatory) on distinct striatal subregions (anterior vs. posterior) during learning are not fully understood.

2. Methodology

Experimental Design & Participants

• Participants: 50 healthy young adults (right-handed) completed a two-day protocol.
• Groups: Participants were randomly assigned to three TMS groups:
  1. DLPFC-cTBS: Continuous Theta-Burst Stimulation (inhibitory) on the Dorsolateral Prefrontal Cortex.
  2. M1-cTBS: Continuous Theta-Burst Stimulation on the Primary Motor Cortex.
  3. DLPFC-20Hz: High-frequency (20 Hz) rTMS (excitatory) on the DLPFC.
• Baseline: Data from 30 participants from a previous study (Choi et al., 2020) using the same task without TMS served as the "No-stim" control group.
• Timeline:
  • Day 1: Calibration of a data glove, familiarization, resting-state fMRI, localizer session, and T1 structural imaging to define individual TMS targets.
  • Day 2: TMS application (immediately followed by fMRI), followed by a motor learning task inside the scanner.

Stimulation Protocols

• Targets:
  • M1: "Hand-knob" area (MNI: -38, -22, 52).
  • DLPFC: Left DLPFC (MNI: -42, 32, 32).
• Intensity: 80% of the Resting Motor Threshold (RMT).
• Protocols:
  • cTBS: 40 seconds of triplet pulses at 50 Hz (inhibitory).
  • 20 Hz rTMS: 20 minutes of 40-pulse trains (excitatory).

Task & Data Acquisition

• Task: A complex de novo motor skill learning task involving a 5×5 grid. Participants used a 14-sensor MR-compatible data glove on their right hand to control a cursor, aiming to reach specific targets. The task involved 3 fMRI runs (approx. 8 mins each), totaling 291 trials.
• Imaging: 3-T Siemens Prisma scanner. High-temporal resolution EPI (TR = 460 ms) was used to capture dynamic changes.
• Analysis:
  • Behavioral: Success rate (proportion of time cursor held in target) analyzed via Linear Mixed-Effects (LME) models.
  • fMRI: Voxel-wise General Linear Model (GLM) with performance as a parametric modulator.
  • Regions of Interest (ROIs): Eight striatal subregions (Anterior/Posterior Caudate and Putamen in both hemispheres) defined by the intersection of TTatlas and Harvard-Oxford atlases.
  • Statistical Modeling: LME models tested the interaction between Run (time: 10, 20, 30 mins post-stim) and Group to assess time-dependent modulation.

3. Key Contributions

• Network-Targeted Causality: Provides causal evidence that stimulating the DLPFC (a cortical node) can specifically downregulate activity in the connected anterior striatum (subcortical node) during complex motor learning.
• Temporal Specificity: Demonstrates that the modulatory effects of cTBS are time-dependent, emerging significantly 20–30 minutes post-stimulation, aligning with known neurophysiological windows for cTBS effects.
• Regional Specificity: Shows that the effect is specific to the anterior caudate (linked to the cognitive loop) rather than the putamen (sensorimotor loop) or posterior striatum, and is specific to the DLPFC target rather than M1.
• Protocol Specificity: Highlights that inhibitory cTBS effectively modulates striatal activity, whereas excitatory 20 Hz rTMS on the DLPFC did not produce significant changes in this specific network during early learning.

4. Results

Behavioral Performance

• Learning Rate: All TMS groups showed significantly faster learning rates (steeper improvement slopes) compared to the No-stim baseline group.
• Overall Performance: There was no significant difference in the overall success rate (average performance) between the four groups. This suggests that while neural processing changed, the behavioral output was not drastically altered at the group level, possibly due to individual variability in baseline strategies.

fMRI & Neural Modulation

• Time-Dependent Effects: No significant group differences were observed at 10 minutes post-stimulation. However, at 30 minutes, significant differences emerged.
• DLPFC-cTBS Effect: This group showed a significant reduction in task-evoked activity in the Left Anterior Caudate (ipsilateral to stimulation) compared to the baseline.
  • Statistical significance: $T(45) = 3.22$, corrected $p = 0.019$.
  • This reduction was specific to the anterior caudate; no significant modulation was found in the putamen or posterior caudate.
• Other Groups:
  • M1-cTBS: Did not significantly modulate activity in the posterior striatum (putamen) or caudate.
  • DLPFC-20Hz: Did not induce significant excitatory effects on striatal activity.
• Interaction: The LME analysis confirmed a significant Run $\times$ Group interaction, indicating that the trajectory of striatal activation over time differed specifically for the DLPFC-cTBS group.

5. Significance and Implications

• Mechanistic Insight: The study supports the hypothesis that the DLPFC and anterior striatum form a "cognitive" loop critical for the early stages of complex motor learning. Inhibiting the DLPFC suppresses the associated striatal activity, likely reflecting a downregulation of dopaminergic signaling (given prior PET evidence linking DLPFC stimulation to caudate dopamine release).
• Clinical Potential: The ability to non-invasively suppress hyperactive striatal responses via network targeting suggests a potential therapeutic avenue for disorders characterized by striatal hyperactivity, such as addiction or OCD.
• Methodological Advancement: The study validates the use of combined TMS-fMRI to probe network-level plasticity in deep brain structures, overcoming the limitations of direct subcortical stimulation.
• Future Directions: The authors suggest that while cTBS showed inhibitory effects, other excitatory protocols (e.g., iTBS) or combining TMS with emerging techniques like Temporal Interference Stimulation (TIS) could be explored to treat conditions involving striatal hypoactivity (e.g., Parkinson's disease).

Conclusion:
The paper demonstrates that network-targeted cTBS applied to the DLPFC selectively and temporally suppresses task-related activity in the anterior striatum during motor skill learning, providing a causal link between cortical stimulation and subcortical modulation without necessarily altering gross behavioral performance metrics.