【作者】 张浩；

【导师】 Zachary J.Smith；

【作者基本信息】 中国科学技术大学 ， 仪器科学与技术， 2025， 博士

【摘要】 细胞器是细胞的基本功能单位,负责执行各种关键的生物学功能。研究细胞器的功能状态对于揭示生命活动机制以及指导疾病的诊断治疗具有重要意义。作为成分复杂的微小亚细胞结构,细胞器的深入研究需要同时对其化学成分和形态结构进行综合分析。然而,当前的细胞器检测技术仍存在局限,荧光显微技术等形态观测手段缺乏化学分析的能力,而质谱和色谱等化学分析手段又无法获取细胞器的形貌和动态,目前仍缺乏能够同时探测细胞器成分和形态的多维度检测方法。拉曼显微技术是一种探测分子振动模式的无标记原位检测技术,能够同时测量样品的化学成分和形态特征,在细胞器研究中具备独特优势。然而,受限于较弱的信号强度,拉曼显微技术在细胞器检测中面临着采集速度慢、检测通量低的技术瓶颈。此外,拉曼显微技术需要长时间曝光才能得到高信噪比光谱,因此很难捕捉那些活细胞中运动细胞器的动态变化。由于拉曼信号强度随样品尺寸减小而快速下降,以上问题在脂滴(约1μm)这类小型细胞器上尤为突出,严重限制了拉曼显微技术在相关研究中的应用。为了解决上述问题,本文进行了如下的研究工作:1.针对拉曼显微技术在脂滴这类小型细胞器上检测通量低的问题,本文提出了一种高通量脂滴光谱检测方法。本研究首先设计并搭建了一套拉曼光谱-相位成像多模态显微系统,然后利用高时空分辨率(200 Hz,245 nm)的相位图像和高准确率(>90%)的脂滴识别算法精确定位脂滴,在此基础上引导拉曼采样激光依次扫描脂滴质心并采集其光谱。随后,通过光谱分析和图像处理精确计算脂滴的脂肪酸组成和形态特征。相比于传统的点扫描拉曼检测方法,该方案将脂滴光谱的采集速度提高了2至3个数量级,能够在10分钟内完成单个细胞内所有脂滴化学成分和形态特征的精确检测。基于该高通量检测方法,本研究以固定细胞为样品采集了超过4万个脂滴的高质量光谱和图像,系统地探究了不同脂滴个体、不同细胞种类以及不同培养环境下的脂滴成分-形态异质性特征,展示了这种基于相位导航的光谱检测方法在细胞器高通量检测中的应用前景。2.针对拉曼显微技术难以检测活细胞中运动脂滴的问题,本文提出了一种运动脂滴光谱检测方法。该方法利用相位图像提供实时的脂滴位置,结合动态分区光谱扫描算法和多线程控制程序,将拉曼采样激光以极高的精度(<300 nm)定位到运动的脂滴上,从而实现了对活细胞脂滴成分变化的长时间、高帧率(2.5 h,1 Hz)追踪分析。此外,结合同步采集的相位图像,还能同时分析脂滴的位置移动、尺寸变化和运动特征。基于该动态脂滴检测方法,我们能够在单细胞器的尺度上对脂滴这类小型细胞器的动态变化进行多维度检测。随后,本研究使用这种方法观测了肝癌细胞摄取花生四烯酸的动态过程,获得了超过11万组脂滴光谱和图像数据,并从成分、形态和动力学三个维度上全面深入地研究这一过程中的脂滴动态变化,展示了这种动态检测方法在研究细胞脂质代谢上的潜力。3.针对拉曼显微技术在线粒体、细胞核等细胞器上检测速度慢的问题,本文在脂滴研究的基础上拓展研究了适用于多种细胞器的快速光谱检测方法。通过结合相位显微技术的全景成像能力、多种细胞器的精确识别算法、针对性的细胞器扫描方式和“光学平均”光谱采集策略,此方法能够在5分钟内得到单个细胞中四种主要细胞器(细胞核、线粒体、脂滴、核仁)的平均光谱,极大提升了细胞器光谱的采集效率。基于这种多细胞器光谱检测方案,本文研究了COS7细胞内不同细胞器的成分差异和细胞对重水的摄取与代谢,验证了这种方案在研究细胞器代谢和药物检测上的应用前景。综上所述,本研究首次提出一种基于相位导航的拉曼显微技术,并展示了其在细胞器研究中的应用。该技术将拉曼显微技术和相位显微技术有机整合,利用相位显微技术的卓越成像能力弥补了拉曼显微镜在形态分析方面的不足,同时基于相位导航优化拉曼显微镜的光谱采样方式,从而显著提高了拉曼光谱采集的通量和速度。此外,本研究也将细胞器的形态信息和化学信息耦合到数据分析中,同步解析了多种细胞器在成分和尺寸上的异质性特征。研究结果表明,该技术为细胞器研究提供了一种可兼顾形态和化学检测的新型高速分析方案。更多还原

【Abstract】 Organelles are the fundamental functional units of cells,responsible for performing various critical biological functions.Investigating the functions and states of organelles is of great significance for elucidating the mechanisms of life processes and guiding disease diagnosis and treatment.As minute subcellular structures with complex compositions,the in-depth study of organelles requires comprehensive analyses of both the chemical composition and morphology.However,current organelle detection techniques still have some limitations,morphological observation methods such as fluorescence microscopy lack the ability to perform chemical analysis,while chemical analysis methods such as mass spectrometry and chromatography are unable to capture the morphology and dynamics of organelles.There is still a lack of multidimensional detection methods capable of concurrently probing both the composition and morphology of organelles.Raman microscopy,a label-free in situ detection technique for probing molecular vibrations,is able to simultaneously measure the chemical composition and morphological features of samples,which provides a unique advantage in organelle research.However,Raman microscopy faces the technical bottlenecks of slow acquisition speed and low detection throughput in organelle detection due to the weak signal intensity.In addition,Raman microscopy requires long exposure times to obtain high signal-to-noise spectra,which makes it difficult to capture the dynamic changes of those moving organelles in living cells.Since the Raman signal intensity decreases rapidly with the decrease of sample size,the above problems are particularly prominent in small organelles such as lipid droplets(～1μm),which severely limits the application of Raman microscopy in related research.In order to solve the above problems,this thesis carried out the following research:1.To address the problem of low detection throughput of Raman microscopy on small organelles such as lipid droplets,this thesis proposes a high-throughput lipid droplet spectra detection method.First,this study designed and built a multimodal microscopy system combining Raman spectroscopy and phase imaging.Then this method used phase images with high spatiotemporal resolution(200 Hz,245 nm)and lipid droplet recognition algorithms with high accuracy(>90%)to accurately locate lipid droplets.On this basis,the Raman sampling laser was guided to scan the centroids of lipid droplets in sequence and collect their spectra.Subsequently,the fatty acid composition and morphological characteristics of lipid droplets were accurately calculated through spectral analysis and image processing.Compared with the conventional point scanning Raman detection method,this scheme increases the acquisition speed of lipid droplet spectra by 2 to 3 orders of magnitude,and can complete the accurate detection of the chemical composition and morphological characteristics of all lipid droplets in a single cell within 10 minutes.Based on this high-throughput detection method,this study collected high-quality spectra and images of more than 40,000 lipid droplets in the fixed cells,systematically explored the compositional-morphological heterogeneity of lipid droplets across different lipid droplet groups,cell types,and environmental factors,and demonstrated the application prospects of this phase-guidance-based spectral detection method in high-throughput detection of organelles.2.To address the problem that Raman microscopy is difficult to detect moving lipid droplets in living cells,this thesis proposes a spectral detection method for moving lipid droplets.The method utilizes phase images to provide real-time lipid droplet positions,and combines a dynamic zonal spectral scanning algorithm and a multi-threaded control program to localize the Raman sampling laser to the moving lipid droplets with extremely high accuracy(<300 nm),realizing a long-time,high-frame-rate tracking(2.5 h,1 Hz)and analysis of chemical dynamics of lipid droplets in living cells.In addition,through the phase images acquired synchronously,this method can also simultaneously analyze the positional movement,size change,and motion characteristics of lipid droplets.Using this dynamic lipid droplet detection method,we were able to perform multidimensional detection of dynamic changes in small organelles such as lipid droplets at the single organelle level.Subsequently,we used this method to observe the dynamic absorption process of arachidonic acid by hepatocellular carcinoma cells,obtained more than 110,000 sets of lipid droplet spectra and images,and comprehensively investigated the dynamic changes of lipid droplets in this process in terms of composition,morphology,and kinetics,demonstrating the potential of this dynamic detection method in the study of cellular lipid metabolism.3.To address the slow detection speed of Raman microscopy on organelles such as mitochondria and nucleus,this thesis expands the research of high-speed spectral detection method applicable to a variety of organelles on the basis of lipid droplet study.By combining the panoramic imaging capability of phase microscopy,the precise identification algorithm of multiple organelles,the targeted organelle scanning method and the“optical averaging”spectral acquisition strategy,this method is able to obtain the averaged spectra of the four main organelles(nucleus,mitochondria,lipid droplets and nucleolus)in a single cell within 5 minutes,which greatly improves the acquisition efficiency of the organelle spectra.Based on this multi-organelle spectral detection scheme,this thesis analyzed the compositional differences of different organelles and the cellular uptake and metabolism of heavy water in COS7 cells,and verified the application prospects of this scheme in studying organelle metabolism and drug detection.In summary,this study presents a Raman microscopy technique based on phase guidance and demonstrates its application in organelle research.This technique organically integrates Raman microscopy and phase microscopy,utilizing the excellent imaging capability of phase microscopy to make up for the deficiencies of Raman microscopy in morphology analysis,and optimizing the spectral sampling method of Raman microscopy based on phase guidance,significantly improving the throughput and speed of Raman spectral acquisition.In addition,this study also couples morphological and chemical information of organelles into data analysis,synchronously analyzing the heterogeneous features of multiple organelles in terms of composition and size.The results show that this technique provides a novel high-speed analytical platform for organelle research that allows for simultaneous morphological and chemical detection.更多还原