毫秒等級三維影像顯微鏡於果蠅腦之研究

Abstract

大腦是生物最重要的器官之一，因為不管是情感、思考、記憶和甚至是生命徵象等大多數行為主要都是由大腦所控制。近百年來，儘管科學家已經對單顆神經元進行了詳盡的研究，我們對腦功能的理解仍然有限。主要原因之一在於大腦的功能並不是由單顆神經元所主導的，而是由全腦神經元之間交錯連接而成且結構複雜的神經網路所控制的。在這個研究裡我們以果蠅作為研究對象，因為果蠅不但具有相當完整的神經結構性圖譜，果蠅腦的大小也足夠小來達成全腦觀察，那為了瞭解功能性神經網路是如何在果蠅的大腦中運作，所需的工具應具有亞細胞的空間解析度以區分單顆神經元、毫秒等級的成像速度以捕捉神經元之間高速的訊號傳遞（像是動作電位）以及接近毫米立方體積的成像尺寸以用於全腦研究。 近年來，雙光子顯微鏡逐漸成為研究活體大腦功能性神經網路的重要工具，因為它不僅能夠提供非侵入性的觀察還能夠達到微米尺度的空間解析度以及深組織穿透深度的光學切片能力來涵蓋三維結構的大腦。為了捕獲神經元之間的快速動態行為，現有的雙光子顯微鏡已經能夠達到毫秒尺度的二維成像速度。然而，神經網絡結構本質上是三維分佈的，由於雙光子顯微鏡需要透過掃描來完成三維成像的特性使得目前雙光子顯微鏡的三維成像速度遠低於毫秒尺度。 因此近十年來許多技術致力於提高三維成像速度，例如壓電致動器、液體透鏡、聲光偏轉器、全像術、空間時域聚焦和層光顯微鏡等等。然而前三種方法是基於單點掃描成像式的技術，由於橫向或軸向掃描速度的限制使得三維成像速度無法達到毫秒尺度。此外，這些技術大多使用單通道偵測器來收光，取樣速度會被螢光衰減週期所限制，導致取樣率不會超過500 MHz，限制了影像速度。後三項技術則是基於廣域照明以及收光來提高成像速度，然而該方式很容易受到散射的影響導致穿透深度不深，因此比較適合用在透明樣本上而不適用於密集的神經網路結構(果蠅大腦)。 為了解決上述的問題，在這項研究中我們開發了一套多焦點多光子體積成像顯微技術，利用將單道光分成32道光，搭配上可調式聲波梯度折射率透鏡以及32通道的光電倍增管，我們達成了每秒500個體積以上的三維成像速度，體積大小約略是邊長為200微米的立方體。透過繞射分光元件產生的32道光做平行掃描，相比於單道光來說可大幅度提高橫向掃描速度，而可調式聲波梯度折射率透鏡可提供約100 kHz的超快軸向掃描速度，從而達到毫秒等級的三維成像。為了能夠補捉可調式聲波梯度折射率透鏡的掃描速度，我們採用了32通道光電倍增管，不但有足夠快的資料取樣速度，整體的資料取樣率更可以提高到10 GHz以上（在這個研究中使用了2.56 GHz），比起單通道偵測器來說整整多了一個數量級。為了證明系統的可行性，我們利用流動的螢光球來測試高速成像，的確能夠捕捉螢光球在流動時的動態行為。希望我們所發展的這套高速多焦點多光子體積成像技術能夠對果蠅全腦功能性神經網路研究有所幫助。

Brain is one of the most important organs because most behaviors, such as emotion, thinking, memory and vital sign, are mainly governed by our brain. Although the behaviors of basic component of brain, so-called neurons, have already been investigated thoroughly for more than one century, our understanding toward brain function is still limited. One of the main reasons is that brain function is not performed by a single neuron, but dominated by sophisticated connections among neurons inside a whole brain, i.e. connectome. In this study, Drosophila was selected as our target not only due to its nearly-complete structural neural connectome but also small enough brain size, which is available for whole brain observation. In order to study how functional connectome works in living brain of Drosophila, the required tool should exhibit (i) sub-cellular spatial resolution to distinguish individual neurons, (ii) temporal imaging speed as high as kHz to capture rapid dynamics among neurons, such as action potential, and (iii) imaging size approaching hundreds of micrometers cubic in volume for whole brain study. Recently, two-photon microscopy has emerged as a powerful tool for in vivo functional connectome study in living brains not only because of its noninvasive property but also due to its ~µm spatial resolution and optical sectioning capability to monitor 3D brain functions with remarkable penetration depth. In order to capture rapid dynamics among neurons, imaging speed of two-photon microscopy in lateral dimension can already reach up to kHz frame rate. Nevertheless, neural network is intrinsically three-dimensionally distributed, and current volumetric two-photon imaging speed is far below kHz due to the requirement of time-consuming sequential axial-scanning. In the past decade, several schemes have been demonstrated to boost volumetric imaging speed, such as piezoelectric actuator, liquid lens, acousto-optic deflector, holography, spatial temporal focusing and light-sheet microscopy. However, the first three methods are single-beam scanning-based techniques, which cannot reach millisecond-scale volumetric imaging due to the limitation of lateral or axial scanning speed. Besides, abovementioned techniques typically use a single-channel detector to collect signal, so that the data acquisition rate is restricted below ~500 MHz under the limitation of fluorescence lifetime. The latter three are based on wide-field illumination and wide-field detection to increase the imaging speed; however, they easily suffer image blurred due to scattering and are limited to transparent samples. Therefore, they are not suitable for dense neuron structures, such as Drosophila brains. In this study, we developed a multi-focal multi-photon volumetric imaging microscopy, based on the combination of 32-channel multi-focal excitation, a tunable acoustic gradient-index (TAG) lens, and a 32-channel PMT, thus reaching unprecedented volumetric imaging rate above 500 volumes per second, in a cubic volume of ~200 µm on each side. The 32-focus parallel scanning is provided by a diffractive optical element (DOE) to considerably enhance lateral scanning speed. The TAG lens offers ~100 kHz ultrafast axial scanning speed, leading to millisecond-scale volumetric imaging. To catch up with the speed of the TAG lens, the 32-channel PMT was adopted which can further boost up data acquisition rate to more than 10 GHz (2.56 GHz is used in this work), one order larger than that of a single-channel detector system. As a proof of concept, we have demonstrated high speed imaging with flowing fluorescence beads, capturing volumetric dynamics of flow motion. Our high-speed multi-focal multi-photon volumetric imaging technique paves the way toward functional connectome study in Drosophila brain.