重建图像显示出长程投射的神经元在鼠脑内是如何延伸的

煞费苦心描绘大脑中蜿蜒曲折的神经环路的科学家们不久将会发现他们的研究领域即将实现工业化。位于中国苏州的一个大型研究设施将于下个月启动高分辨率脑图谱的绘制。

典型实验室通常只有一到两台脑成像系统，而该新建立的研究机构将拥有50套自动化的成像系统，对小鼠全脑进行快速切片，获取每一层的高清晰度图像并将其它们重建为三维图像。该研究机构的合作者之一，来自美国西雅图Allen脑研究所的分子生物学家HongkuiZeng表示，“这种工厂化的研究规模将极大地加速脑科学的发展”。同时她还表示，“以工业化的形式大规模、标准化地产生数据将改变神经科学已有的研究方式”。

来自中国武汉华中科技大学的生物医学成像专家骆清铭表示，该研究机构也将开展人脑成像，目标是成为一个国际中心，以帮助从脑疾病到类脑人工智能的研究者绘制各种神经连接。骆清铭是该新科研机构——华中科技大学苏州脑空间信息研究院的领导者。该机构拥有5年共计4.5亿人民币（约合6700万美元）的科研经费，将在未来招募120位科学家和技术人员。自称脑空间信息学研究者的骆清铭是该机构高速成像系统的发明人。

热点话题

该研究机构的另外一位合作者，来自美国纽约冷泉港实验室的神经科学家JoshHuang表示，“毫无疑问需求是巨大的”。高通量、快速全脑成像可以改变神经生物学家对大脑中的神经元究竟是如何联系这一问题的理解，就像在21世纪初高通量测序帮助遗传学家解析人类基因组一样。他说，“这项研究将在多个物种构建细胞分辨脑图谱方面产生巨大影响”。

哺乳动物大脑包含数百万个神经元，人脑甚至包含上百亿的神经元。这些神经元有成千上万种不同的细胞类型，在细胞形状、大小和基因表达方面有很大的差异。神经科学家希望绘制出结构以及它们如何互相作用将有助于揭示它们的功能（Nature 548,150-152:2017）。位于德国马格德堡的莱布尼茨神经生物学研究所的脑成像专家Jürgen Goldschmidt表示，通过比较多个大脑中特定类型的神经元，神经生物学家有可能从中发现疾病或者某种后天习得的行为对神经元结构的影响。

但是这样的绘制过程通常需要数月甚至数年时间的努力。这个过程包括用金刚石刀将厘米尺度的小鼠脑切成15000张超薄切片，通过化学或荧光标签对每一张切片进行染色以辨别特定特征，用显微镜对每一张切片进行成像，并最终将这些图像重建成一个三维图谱。

高速成像

这正是该研究院的用武之地。合作者表示，该院拥有大量能够进行高速和高分辨率成像的系统。HongkuiZeng表示，这种成像系统能够在两周内完成其他如超分辨共聚焦成像系统需要数月才能完成的小鼠全脑成像。

当今年二月在美国马里兰州贝塞斯达参加美国脑计划会议的研究者们看到了一张一个神经元覆盖了鼠脑大量通路的图像时，他们感受到了这一技术的能力（参见Nature543, 14–15; 2017）。该工作正是由Allen脑研究所神经生物学家ChristofKoch领导的团队与骆清铭的研究团队合作完成，该图像中神经细胞的丰富投射提示该神经元在协调全脑输入输出以形成意识中可能扮演了重要角色。

骆清铭表示，苏州脑空间信息研究院将会产生海量数据，一个小鼠脑的数据量已经达到了8TB。而人脑的体积大约是小鼠脑的1500倍，如果以研究院现有速度需要一台成像系统至少花费20年时间完成数据采集。目前骆清铭正致力于提高速度并采用多套成像系统并行工作。

骆清铭渴望与世界各地的研究者合作，目前已经和Allen脑研究所、冷泉港实验室和斯坦福大学的研究者展开合作。同时骆清铭也表示，有兴趣的人很多以至于难以满足每一位研究者的需求，“受限于原有服务能力，此前我们不得不婉拒了部分研究者”。

原文如下：

China launches brain-imaging factory

Hub aims to make industrial-scale high-resolution brain mapping a standard tool for neuroscience

David Cyranoski 16 August 2017

Neuroscientists who painstakingly map the twists and turns of neural circuitry through the brain are about to see their field expand to an industrial scale. A huge facility set to open in Suzhou, China, next month should transform high-resolution brain mapping, its developers say.

Where typical laboratories might use one or two brain-imaging systems, the new facility boasts 50 automated machines that can rapidly slice up a mouse brain, snap high-definition pictures of each slice and reconstruct those into a 3D picture. This factory-like scale will “dramatically accelerate progress”, says Hongkui Zeng, a molecular biologist at the Allen Institute for Brain Science in Seattle, Washington, which is partnering with the centre. “Large-scale, standardized data generation in an industrial manner will change the way neuroscience is done,” she says.

The institute, which will also image human brains, aims to be an international hub that will help researchers to map neural connectivity for everything from studies of Alzheimer’s disease to brain-inspired artificial-intelligence projects, says Qingming Luo, a researcher in biomedical imaging at the Huazhong University of Science and Technology (HUST) in Wuhan, China. Luo leads the new facility, called the HUST-Suzhou Institute for Brainsmatics, which has a 5-year budget of 450 million yuan (US$67 million) and will employ some 120 scientists and technicians. Luo, who calls himself a “brainsmatician”, also built the institute’s high-speed brain-imaging systems.

Hot topic

“There will be large demand, for sure,” says Josh Huang, a neuroscientist at Cold Spring Harbor Laboratory in New York, which is also partnering with the Chinese institute. Access to high-throughput, rapid brain mapping could transform neuro-scientists’ understanding of how neurons are connected in the brain, he says — just as high-throughput sequencing helped geneticists to untangle the human genome in the 2000s. “This will have a major impact on building cell-resolution brain atlases in multiple species,” he says.

Mammalian brains have millions of cells, and human brains even have billions. And the cells come in some 10,000 different types, marked by differences in shape, size and the genes they express. Neuroscientists hope that mapping out the structures and how they interact will help to reveal their functions (see Nature 548, 150–152; 2017). By comparing particular neuron types across multiple brains, scientists might be able to pick out the effects of a disease or a learned behaviour on cell structure, says Jürgen Goldschmidt, a brain-imaging researcher at the Leibniz Institute for Neuro-biology in Magdeburg, Germany.

But such maps often require months or years of effort. The process involves shaving centimetre-long mouse brains into 15,000 ultrathin slices with a diamond blade, staining each layer with chemicals or fluorescent tags to pick out particular features, imaging each layer with a microscope and then reconstructing the images into a 3D map.

High-speed mapping

That’s where Luo’s institute can help. Its vast number of machines have impressive speed and resolution, collaborators say. According to Zeng, the devices can gather the same amount of detail on a mouse brain in two weeks as would require months using other technologies, such as super-resolution confocal imaging.

Participants at a February meeting of the US BRAIN initiative (Brain Research through Advancing Innovative Neurotechnologies) in Bethesda, Maryland, were treated to a display of the technology’s capabilities when they were shown an image of a neuron that wrapped all of the way around a mouse brain (see Nature 543, 14–15; 2017). Allen Institute neuroscientist Christof Koch, whose team did the work in collaboration with Luo’s group, suggests the extensive reach of the neuron shows that the cell has a role in coordinating inputs and outputs across the brain to create consciousness.

The Suzhou institute will generate a huge amount of data: each mouse brain map alone will be 8 terabytes, Luo says. But the volume of a human brain is nearly 1,500 times that of a mouse brain; it would take a single machine around 20 years to digitally reconstruct one at the institute’s current rate. Luo aims to increase the speed of his machines and to use multiple devices in parallel.

Luo is keen for worldwide collaboration; along with the Allen Institute and Cold Spring Harbor Laboratory, Stanford University in California is forming a partnership with the centre. But Luo says that interest is so high that he won’t be able to accommodate everyone. “We are already turning people down.”

Nature 548,268-269 (17 August 2017) doi:10.1038/548268a

原文链接：

https://www.nature.com/news/china-launches-brain-imaging-factory-1.22456