利用多鳥嘌呤重複序列探討卵巢癌之演化過程

Abstract

上皮性卵巢癌是全球十大惡性腫瘤之一，其中亮細胞卵巢癌的組織學分型在亞洲，尤其在台灣具有較高的發病率和較差的存活率，這加強了對該疾病進行深入剖析的緊迫性。癌症發展過程中會受到不同的因素影響，免疫環境則是當中相當重要的一個因素。在研究中發現“熱”腫瘤通常預後會較好，但亮細胞卵巢癌(OCCC)亞型卻顯示出相反的結果，但目前沒有直接的原因解釋這個現象。所以我們希望可以利用演化的方式去疾病進展和免疫相互作用的動態環境，以揭示選擇壓力如何影響疾病進展的生物意義。本研究會利用多鳥嘌呤指紋圖譜計算樣本間的遺傳距離，並結合突變率來重建與細胞分裂相關的系譜，以了解其相互作用。目前我們已完成十個病人的系譜重建，包括六個免疫熱病例和四個免疫冷病例。“熱”腫瘤的受精卵到最近共同祖先(MRCA）細胞分裂平均為3888次，略高於“冷”腫瘤的984次，這意味著熱腫瘤的突變負荷較高。“熱”腫瘤的變異係數平均為34%，而“冷”腫瘤為28.7%。“熱”的腫瘤多鳥嘌呤分佈的標準差的變異數的為0.98，而“冷”腫瘤的變異數為1.00。根據共祖先率(coalescence ratio)的結果顯示，免疫熱腫瘤的平均比率為0.67，而免疫冷腫瘤的比率為0.56。而這表明“熱”腫瘤可能由初始創始細胞的高突變負荷定義，並影響了後續的克隆結構。相反，“冷”腫瘤可能因選擇壓力低，導致更多的克隆掃蕩形成一個相對同質性高的克隆結構。另一方面，當同時發現患者有子宮內膜癌(EC)和卵巢癌(OC)的情況下對臨床診以及治療上會是一種挑戰，並且兩者會常具有相同的組織學樣形態-子宮內膜樣亞型，導致潛在的誤診和治療不足。為了解決這一問題，我們提出了一種基於多聚鳥嘌呤指紋圖譜的工具，用於根據遺傳相似性區分疾病狀態。在分析34個同時性子宮內膜癌和卵巢癌(SEOC)的病例時，結果顯示有1個樣本有高遺傳距離與DNA核酸錯配修復(MMR)狀態不匹配的情況，這表明該工具有可能定義MMR狀態。核酸錯配修復缺陷(MMRd)病例被排除在分析之外。然後，我們計算了皮爾遜相關係數的結果合併無進展生存期數據的共祖先率(coalescence ratio)，發現高相似性的樣本具有不同的分群，一群顯示高相似性和低疾病無惡化存活期，而一群則顯示高相似性和高疾病無惡化存活期。鑑於SEOC病例中EC和OC不同存活具有一定差異，我假設高相似性的群中包括兩種不同的疾病。我們會使用機器學習的方法針對定序資料進行分析，以識別這些高相似性但不同存活率的樣本組，用以確定是否存在兩種不同疾病。在總結中，我們的系譜分析表明，免疫相互作用對OCCC的疾病進展有不同影響，但需要更多病例來驗證這一觀察結果。SEOC的區分工具可以識別隱藏的MMRd腫瘤並區雙重原位癌症和轉移癌症，為準確的原發地預測和改善預後提供關鍵資訊。

Epithelial ovarian cancer (EOC) is a top 10 malignancy worldwide. Ovarian clear cell carcinoma (OCCC), a subtype of EOC, has a higher prevalence and worse survival in Asian populations, especially in Taiwan, highlighting the urgent need for a deeper understanding of the disease. The survival outcome of the endometrioid-like subtype is optimal, but it poses a unique clinical challenge when it coexists with endometrial cancer due to their similar histological features. In this thesis, we aimed to dissect them by understanding their evolution (disease progression). Although “hot” tumors usually have better prognoses, ovarian clear cell carcinoma (OCCC) shows the opposite. We aimed to understand how the interaction between mutation burden and the immune microenvironment shaped selection processes by building phylogenetic trees utilizing indels in polyguanine sequences. Among the ten phylogenies we constructed (hot: 6; cold: 4), the zygote-to-MRCA distance in “hot” tumors averages 3,880 cell divisions, slightly higher than the 984 divisions observed in “cold” tumors. “Hot” tumors also exhibited higher intra-tumor heterogeneity (34%), defined by the average coefficient of variance, compared to “cold” tumors (28.7%). The stutter distribution for immune-hot tumors had a mean variance of 0.98, compared to 1 for immune-cold tumors. The coalescence ratio shows an average of 0.67 in immune-hot tumors, compared to 0.56 in immune-cold tumors. This suggests that “hot” tumors might be defined by the high mutation burden of the initial founder cell, which subsequently influences the clonal structure. In contrast, “cold” tumors might experience low selection pressure, leading to more clonal sweeps and a homogeneous clonal structure. Utilizing this method, on the other hand, we studied synchronous endometrial and ovarian cancer (SEOC) utilizing a cohort of 39 cases in which 8 cases with high tumor mutation burden were identified by our approach compared with 7 cases by IHC staining. The coalescence ratio was calculated by using Pearson correlation coefficient to distinguish cases with double primary tumor (6/25) from metastatic diseases (19/25). However, we found that genetic similarity does not correlate with survival outcomes, indicating the importance of recognizing the tumor's origin. Therefore, we applied well-established machine learning methods to sequencing data to identify the origins of the tumors. In summary, our phylogenetic analysis suggested that immune interactions impacted OCCC progression differently. Immune-hot tumors tended to have a high mutation burden and complex clonal structure, while immune-cold tumors showed lower mutation burden and homogenous clonal structure. Additional cases are needed to confirm these findings. Polyguanine fingerprinting could serve as a better tool for identifying cases with high TMB and for distinguishing between DPC and MC, offering a promising alternative for clinical practice.