- 學位論文
胰島β細胞在病理與生理狀態時粒線體形態特 徵辨識之研究
Identification of Mitochondrial Morphological Features Specific to Insulin-secreting β-cells at Different Physiological and Pathological Conditions
摘要
摘要 糖尿病的產生是由於胰島素的製造不足、或是分泌的胰島素無法正常作用，會造成高血糖的症狀。有研究指出高血糖會抑制粒線體的複合物而激發氧化壓力，而粒線體的形態變化、動態狀況與氧化壓力及異常的代謝有關，本研究假設不同的糖尿病症狀會影響粒線體形態，並以分泌胰島素的β細胞為目標，建立一套β細胞之粒線體形態分析系統，來驗證此假說。 本研究針對兩種影像進行分析：第一種是來自於細胞中過度表現的調節蛋白之粒線體動態影像，分為控制組、Fis組、Mfn-1組及DNMfn-1組，此種影像的細胞表現中，不同調節蛋白的影像會有不同的粒腺體形態；第二種影像則是在不同糖尿病症狀下的細胞影像，包括控制組、FCCP組、高葡萄糖組、高葡萄糖及棕櫚酸酯組與KCL組，這種影像則可以用來了解糖尿病症狀與粒線體形態間的關係。 當過度表現的蛋白質影像在納入未知組別後可以大幅增加分類準確率，高達100%。被歸在未知組別的影像大多為DN-Mfn1影像及Fis1影像，這兩組影像的粒腺體多為球狀的碎點，太相似所以容易造成分類錯誤；WT-Mfn1的影像則處於粒線體聚合的階段，多為大塊狀，而控制組的影像也因為含有長條狀和網狀，很容易辨別出來，所以WT-Mfn1和控制組比起前兩組影像有更高的準確率。對於不同糖尿病症狀下的影像，本系統的分類準確率為80%。
並列摘要
ABSTRACT Diabetes is due to insufficient secretion or resistant to insulin and results in hyperglycemia. Previous investigation suggested that high glucose might inhibit mitochondrial complex to induce oxidative stress. Because mitochondria morphological changes, mitochondria dynamics, is correlated with oxidative stress and abnormal metabolism, we hypothesize that various diabetic conditions may affect mitochondrial morphology. This study hopes to establish a numerical analysis system for mitochondrial morphology of β-cells, which secret insulin, to test this hypothesis. There are two sets of images for this study. The first set is images of cells over expressing regulatory proteins of mitochondria dynamics, including control, Fis, Mfn-1, and DNMfn-1. This set of images has distinct mitochondrial morphology among cell expressing various regulatory proteins, and used to build the prototype system. The second set is images of cells at various diabetic conditions, including Control, FCCP, High Glucose, High Glucose plus Palmitate and KCl. This set of images is used to understand whether mitochondrial morphology is correlated with various diabetic states. For images of different protein over-expression conditions, the system can accurately classify the image with the percentage up to 100%. Introduction of unknown class has helping in significantly increases the classification accuracy of the system. The unknown class in this image set mostly comes from DN-Mfn1 and Fis1 images. Since in both over expression of DN-Mfn1 and Fis1, mitochondria undergoing fragmentation, no wonder these two types share thef similar trait (majority globular). Over expression of WT-Mfn1 leads to mitochondria aggregation, as can be seen from its images, which consist majority of lump. Control image also has distinct shape that is long tubular and network, so these two latter types stands out from other previous two. For different conditions of stimulant (glucose dataset), the accuracy of the system is 80%.