人類運動的精密性與複雜性

運動是一種極為精妙的行為，具有數個關鍵特性。首先，運動必須極其精確才能滿足日常需求。舉例而言，步態準確率若達到 99.9%，仍意味著每行走一千步左右便會絆倒一次，這已屬於需醫療介入的疾病狀態。其次，運動極為複雜。人類的動作幾乎可以組合出無限的模式，形成如喝水、進食、奔跑、跳躍、舞蹈等多種功能性動作。數百年來，醫生與科學家已證實，這些精細的運動控制（即運動協調）由小腦負責。然而，小腦如何計算與調控運動協調的機制仍未明瞭。

小腦頻率編碼與運動障礙之關聯

這種知識上的瓶頸導致小腦相關疾病，例如小腦性共濟失調 (Cerebellar ataxia)，成為無藥可醫的嚴重疾病。此外，原發性顫抖症 (Essential tremor) 是最常見的運動障礙之一，影響約 20% 的老年患者，目前仍缺乏令人滿意的治療方式。同時，曾風靡一時的元宇宙技術，因現有腦機介面無法準確解讀與運動控制相關的神經資訊，導致用戶體驗大打折扣。 

近年來，我們的研究團隊致力於揭開小腦運動控制的奧秘，並取得了重大進展。我們首先觀察不同的小腦疾病，並歸納出一種共通的病理生理學機制。在原發性顫抖症患者中，病患產生了一個固定頻率的不正常動作；而小腦共濟失調患者，則表現出運動節律的喪失。這些發現促使我們提出一項假設：小腦可能透過頻率編碼來控制運動細節。

透過前沿研究解析原發性顫抖症

在探究顫抖的病理生理機制時，我們發現原發性顫抖症患者缺乏一種名為 GluRδ2 的小腦蛋白，導致小腦爬行纖維 (Climbing fiber) 過度生長。這種異常的生長引發小腦的過度共振，進而導致顫抖。此外，我們開發了一種新技術——小腦腦波（cEEG），能夠偵測人類小腦電訊號。研究發現，患者的小腦振盪幅度與疾病的嚴重程度呈正相關。這項發現首次證明，小腦能夠調控運動頻率，並導致固定頻率的非自主運動（即顫抖）。

引領神經工程進入精確控制的新紀元

然而，這種定性機制，缺乏數值精確度，無法準確的反應運動細節。如同物理學與電機工程的要求，我們必須建立數學公式層級的機制，以進行運動控制的神經工程。具體而言，若某位患者的顫抖頻率為 7 Hz，小腦是如何計算出這個「7」的？而另一位患者的顫抖頻率為 6 Hz，相同的小腦神經演算法是否能精準產生「6」而不偏離至其他數值？

為了釐清這種現象的工程層級神經機制，我們先在小鼠實驗中結合了先進的電光技術，包括多陣列神經元記錄與光遺傳學技術。我們發現，小腦是透過神經元群體編碼與神經迴路共振來控制頻率。單一小腦神經元對顫抖頻率的貢獻極為有限，也不精確，但當多個神經元的活動加總後，整體的群體活動會收斂至某一穩定振盪狀態，而該振盪頻率正是與顫抖頻率相符的固定值。此發現為神經工程帶來全新契機——我們不需要精確控制約 500 億顆小腦神經元的個別活動，只需調節整體群體神經元的放電頻率，便可有效操控運動頻率。 

為了在小鼠模型中建立顫抖工程技術，我們使用光遺傳學技術來操控群體神經元的放電模式。例如，透過 13 Hz 的光遺傳刺激，我們能夠誘導小腦群體神經元產生 13 Hz 的振盪，即使個別神經元本身並未以 13 Hz 放電。這種 13 Hz 刺激成功的使健康小鼠產生 13 Hz 的顫抖。此外，我們還能夠抑制小鼠天生的 20 Hz 顫抖，並將其顫抖頻率調整至我們設計的 13 Hz。 

除了在動物模型中的研究，我們亦透過 cEEG 記錄原發性顫抖症患者的顫抖狀態，並發現其小腦振盪與個別患者的顫抖頻率相對應。接下來，我們應用一種非侵入性的神經刺激技術——經顱交流電刺激（tACS），做為治療的嘗試。此技術能夠產生微弱、頻率特定的正弦電流，針對小腦進行調控。我們特意在每位患者的顫抖頻率下，施加一個反相位的刺激。例如，當神經元群體振盪需要更多電流時，我們抽走電流；當神經元群體振盪需要放電時，我們提供額外電流來阻止它。此種反相位 tACS 設計成功干擾了頻率生成，並顯著抑制了原發性顫抖症患者的顫抖現象。這是首次以非侵入性腦刺激技術為原發性顫抖症提供有效的治療。

探索小腦共濟失調的頻率編碼異常

雖然原發性顫抖症是常見且重要的疾病，它僅僅是小腦疾病的一個面向。我們進一步研究小腦共濟失調相關的頻率編碼異常，這些疾病的特徵在於運動節律的喪失，且整體發病率可達每 10,000 人中 5 例。

與原發性顫抖症的爬行纖維過度生長相反，我們發現小腦共濟失調患者普遍存在爬行纖維退化的現象，且在遺傳性或非遺傳性的病因皆然。這一病理特徵在脊髓小腦性共濟失調 (SCA1、SCA2、SCA6)、齒紅核蒼白路易體萎縮症 (DRPLA)，以及小腦型多系統萎縮 (MSA-C) 中均可見。爬行纖維的退化導致小腦振盪節律的喪失，這一變化可透過 cEEG 監測到。此外，患者的小腦頻率依賴性振盪減少的程度，與共濟失調的嚴重程度呈現正相關。最終，我們在 SCA1 小鼠模型中透過化學遺傳學技術 (chemogenetics) 提升爬行纖維的活性，成功改善了運動節律與運動表現。

小腦頻率編碼於治療與產業創新之應用

我們的研究證實，小腦透過頻率編碼，來進行精確的動作控制。當振盪過度時，將導致顫抖；而當這種振盪喪失時，則會導致共濟失調。小腦利用神經元群體編碼，使小腦產生整體振盪，而不需每個單一神經元都精確放電。因此，操控小腦的整體振盪頻率，就能精確的操控動作頻率達神經工程的程度，並成為運動障礙的新治療手段。此一系列的發現，為醫療技術的開發提供了可行性。在醫學技術上，可開發一種微侵入性的可程式化裝置，植入硬腦膜上或硬腦膜下（即頭皮下但不直接觸及大腦），進行小腦刺激及頻率操控。如此，不僅對於最常見的運動障礙——原發性顫抖症——有重大潛力，亦能應用於目前無治療方案的小腦共濟失調症—此類疾病在美國食品藥品監督管理局（FDA）是屬於加速審查的路徑。更重要的是，頻率編碼機制能夠為運動控制提供毫秒級精確度，對於未來元宇宙技術、機器人控制，以及腦機介面的發展皆具有深遠影響。

（113年TBF吳火獅醫學獎、台大醫學院藥理所　潘明楷副教授）

Cerebellar frequency coding for motor control: a new opportunity for treating movement disorders and technology development.

The Complexity and Precision of Human Movement

Movement is an extraordinary behavior with several key features. First, it must be highly precise for daily use. A gait accuracy of 99.9% means stumbling once every thousand steps, which is already a disease state requiring medical attention. Second, movement is extremely complex. Human motion involves nearly infinite pattern combinations that enable diverse functional activities such as drinking, eating, running, jumping, and dancing. For hundreds of years, doctors and scientists have confirmed that these intricate motor controls, known as motor coordination, are governed by the cerebellum. However, the precise mechanisms by which the cerebellum computes motor coordination have remained unclear.

Understanding Cerebellar Frequency Coding and Its Role in Movement Disorders

Reflecting this knowledge gap, cerebellar disorders such as cerebellar ataxias are devastating diseases with no effective therapy. Essential tremor, the most common movement disorder affecting nearly 20% of elderly patients, also lacks satisfactory treatment. Additionally, the metaverse—a once highly popular technological trend—has largely failed due to the inability of current brain-computer interfaces to accurately interpret neural information related to motor control, significantly affecting user experience.

Over the past decades, our team has investigated the mystery of cerebellar motor control and achieved significant progress. We first observed different cerebellar diseases and identified a converging pathophysiology. In essential tremor, patients developed motor coordination problems with a fixed frequency. In ataxic patients, motor incoordination manifested as a loss of motor rhythm. These observations led to the hypothesis that the cerebellum controls movement details, termed motor kinematics, via frequency coding.

Investigating Essential Tremor Through Cutting-Edge Research

By studying tremor pathophysiology, we identified that patients with essential tremor lack a cerebellar protein called GluRδ2, which leads to the overgrowth of cerebellar climbing fibers. This overgrowth results in excessive cerebellar oscillations and tremors. We also developed a technology called cerebellar electroencephalography (cEEG), which detects human cerebellar oscillations. The magnitude of these cerebellar oscillations in patients with essential tremor correlates with disease severity. This discovery provides the first evidence that the cerebellum can manipulate motor frequencies and cause tremors—an involuntary movement with a fixed frequency.

Fig. 1. Overgrowth of cerebellar climbing fiber (CF) causes excessive cerebellar oscillations and tremors. From figure S18, Science Translational Medicine. 2020 Jan 15;12(526):eeey1769.

Pioneering Neural Engineering with Numerical Precision

However, this qualitative pathophysiology lacks numerical precision for frequency computation. As required in physics and electrical engineering, we must establish a mathematical formula-level mechanism to enable neural engineering for motor control. Specifically, if a patient has a "7-Hz" tremor, how does the cerebellum compute this "7"? For another patient with a "6-Hz" tremor, can the same cerebellar neural algorithm generate exactly "6" and not any other number?

To address this engineering-level neuronal mechanism, we incorporated state-of-the-art electro-optical technology in mouse studies, including multi-array neuronal recordings and optogenetics. We discovered that the cerebellum utilizes population neuronal codes and circuit oscillations to control frequency. Specifically, individual cerebellar neurons perform poorly in representing the oscillatory frequency of tremors. However, when the activities of multiple cerebellar neurons are combined, their population activity converges to an oscillatory state with an exact frequency matching the tremors. The discovery of population neuronal codes provides an exciting opportunity for neural engineering. Instead of controlling the activity of individual cerebellar neurons—of which there are approximately 50 billion in humans—we can manipulate motor frequency by biasing population neuronal firings toward a specific frequency.

To establish tremor engineering in mice, we used optogenetics to manipulate population neuron firings. For example, 13-Hz optogenetic stimulation induced 13-Hz population neuronal oscillations, even though none of the individual neurons fired at 13 Hz. Notably, this 13-Hz optogenetic stimulation in the cerebellum caused healthy mice to tremor at 13 Hz. More importantly, we could suppress the innate 20-Hz tremor of a mouse and shift the tremor frequency to our designed 13-Hz stimulation.

Furthermore, population neuronal activities can be recorded using non-invasive technology such as cEEG. To translate therapeutic opportunities to essential tremor patients, we recorded cEEG during tremor episodes and found frequency-dependent cerebellar oscillations at each patient's individual tremor frequency. Next, we applied a non-invasive intervention called transcranial alternating current stimulation (tACS). tACS generates subtle, frequency-dependent sinusoidal currents directed at the cerebellum. We specifically applied these currents at each patient's tremor frequency but in an anti-phase setting. When population neural oscillations required more current at a certain phase, we drained it away. Conversely, when the population neural oscillations were set to discharge current at another phase, we provided additional currents to counteract them. This anti-phase tACS design successfully disrupted frequency generation and suppressed tremors in essential tremor patients. For the first time, we demonstrated a proof-of-concept therapy using non-invasive brain stimulation technology for essential tremor patients.

Fig. 2. The cerebellum encodes tremor frequency via population neuron codes. From figure S17, Science Translational Medicine. 2024 May 15;16(747):eadl1408.

Investigating Cerebellar Ataxias

While essential tremor is prevalent, it represents only one aspect of cerebellar diseases. We next investigated the frequency coding abnormalities—the loss of motor rhythm—associated with cerebellar ataxias, a collection of untreatable diseases with an overall incidence of up to 5 per 10,000 people.

In contrast to the climbing fiber overgrowth observed in essential tremor, we found that climbing fiber regression is a common pathological feature of cerebellar ataxias, regardless of genetic or non-genetic origins. This was evident in patients with spinocerebellar ataxia type 1 (SCA1), SCA2, SCA6, Dentatorubral-Pallidoluysian Atrophy (DRPLA), or multiple system atrophy of the cerebellar type (MSA-C). Climbing fiber regression leads to the loss of cerebellar oscillatory rhythm, which is detectable using cEEG. Consistently, reductions in frequency-dependent cerebellar oscillations correlate with ataxia severity across various etiologies. Finally, using a proof-of-concept therapy in an SCA mouse model, we confirmed that boosting climbing fiber activity via chemogenetics improved both motor rhythm and motor performance.

Fig. 3 The cerebellum encodes tremor frequency via population neuron codes. From figure S19 Science Translational Medicine. 2025 Feb 26;17(787):eadk3922.

Advancing Therapy and Innovation Through Cerebellar Frequency Coding

Our discoveries confirm that the cerebellum contributes to precise motor kinematic control via frequency coding. Frequency control is highly precise due to population neuronal codes that generate collective cerebellar oscillations without requiring precision at the individual neuronal level. Excessive frequency-dependent oscillations lead to tremor, while the loss of such oscillations results in ataxia. Frequency-specific engineering presents a novel therapeutic target for movement disorders. Developing a minimally invasive, closed-loop, programmable device implanted in the epidural or subdural space (beneath the scalp but above the brain) for frequency-dependent stimulation could provide significant biomedical advancements. This technology has the potential to impact both essential tremor—the most prevalent movement disorder—and ataxias, which currently lack effective therapies and are on the accelerated track for FDA approval. More importantly, the frequency coding mechanism introduces sub-second precision to neuronal algorithms for motor control, profoundly impacting not only disease treatment but also advancements in the metaverse and human-level precision in robotics and brain-machine interfaces.

(2024 TBF Wo Ho-Su Medical Award, Associate Professor Ming-Kai Pan)