與思覺失調症之神經認知缺損相關之遺傳變異：由多基因風險分數構成的修飾因子

Abstract

背景與目的：思覺失調症為高遺傳率且受多基因影響的疾病。這類病人的認知功能缺損是核心病徵之一，而且家庭研究顯示病人的這些認知缺損有相當程度來自遺傳因子的貢獻。這些發現顯示有些遺傳變異可能扮演修飾因子的角色，影響思覺失調症病人的認知功能。然而，目前對於影響病人認知功能缺損的相關遺傳變異仍不清楚。本篇研究主要目的為 (1) 利用全基因體關聯分析，建立預測思覺失調症病人之認知功能缺損的多基因風險分數；(2) 針對這些與思覺失調症認知功能缺損相關之遺傳變異，利用生物資訊系統，探索他們可能參與之生物途徑或相關的基因網絡。 方法：本研究樣本來自於台灣精神分裂症之全基因體研究計畫 (Schizophrenia Trio Genomic study of Taiwan)。每位病人接受遺傳研究診斷問卷訪談 (Diagnostic Interview for Genetic Studies) 、2項神經認知功能測驗及其他檢查，並抽血。認知功能的測量包括威斯康辛卡片分類測驗 (Wisconsin Card Sorting Test)與持續注意力測驗 (Continuous Performance Test)。病人的DNA以精神病基因體協會 (Psychiatric Genomics Consortium) 所設計的PsychChip進行約60萬點的基因型定型。最後樣本為有完整認知功能測驗及基因型資料的905位病人。先以主成分分析將認知功能指標簡化，然後以分層隨機分派的方式，將病人分成探索組 (n = 453) 與驗證組 (n = 452)。每組病人都以主成分分數的中位數為切點，分成表現較好者與較不好者，再探討這二分式後果與遺傳變異的相關性。針對探索組，每一個遺傳變異點與二分式後果的相關勝算比及其顯著水準，先用PLINK軟體估算。再由PRSice軟體以每0.00005為一間隔，有系統地計算出相對應的多基因風險分數，並從中找出解釋力最佳的一組遺傳變異。再將相關的多基因風險係數套用至驗證組，看是否仍能夠區分病人的認知表現好壞。得到最佳的多基因風險分數後，利用Ingenuity Pathway Analysis (IPA) 軟體分析潛在的生物途徑與基因網路。 結果：主成分分析得到2個認知功能的主成分，分別為執行功能主成分 (解釋總變異的46.31%) 與注意力主成分 (解釋總變異的21.42%)。針對基因定型資料的品質管控並去除高度連鎖不平衡的基因位點後，共196,770個遺傳變異納入分析。針對執行功能主成分之二分式後果，在顯著水準閾值為0.0014的情況下，由190個遺傳變異組成的多基因風險分數，可在驗證組中成功地預測執行能力之好壞 (p = 0.002)，並可解釋2.79%的變異量。然而針對注意力主成分之二分式後果，沒有一組遺傳變異可以在驗證組中成功地區別注意力之好壞。就可預測執行功能主成分的190個遺傳變異，有90個可以被IPA辨認至84個已知基因。基因網絡的分析結果發現，最大的網絡是以NF-κB為核心，暗示它們可能牽涉一些與發炎及免疫反應相關之功能。 結論：本篇研究成功地建立一組由190個遺傳變異組成的多基因風險分數，成功通過探索組與驗證組的測試，可以預測思覺失調症病人之認知功能好壞。由基因路徑分析所暗示的功能，可以作為未來進一步了解這些基因如何修飾思覺失調症認知缺損的基礎。

Background: Schizophrenia is a disease with high heritability, and has been considered to be affected by multiple genes. Cognitive deficits are one of the core symptoms of schizophrenia patients; in addition, family studies show that these cognitive impairments have substantial genetic contributions. Furthermore, previous studies have indicated that some genetic variants may be modifier factors and influence the clinical manifestations of schizophrenia patients. However, it remains little investigated about which genetic variants that affect schizophrenia patients’ cognitive impairment. This study aims to 1) search for genome-wide genetic variants for cognitive impairment in schizophrenia patients by means of polygenic risk scores (PRSs) approach; and 2) find possible pathways or networks involved by these genetic variants using bioinformatics tools. Methods: Subjects of this study were part of the participants from the Schizophrenia Trio Genomic study of Taiwan (S-TOGET). Every participant received an interview using Diagnostic Interview for Genetic Studies and two types of cognitive assessments, among other examinations, and was drawn for blood sample. Both the Wisconsin Card Sorting Test (WCST) and the Continuous Performance Test (CPT) were used to assess patients’ cognitive performance. Patients’ DNA was genotyped using a PsychChip containing 600,000 genetic variants, which was designed by Psychiatric Genomics Consortium for genome-wide association studies. The final sample size was 905 patients who had complete information of the two neurocognitive tests and PsychChip genotyping data. We used principal component analysis (PCA) to extract information from the cognitive indices; then, patients were divided into the learning set (n = 453) and the testing set (n = 452) using stratified randomization. For individual principle components, patients from each group were divided into better or poor group by the median of the principle component score. Then, the association of each genetic variant with the dichotomous outcome in the learning set was estimated using the software PLINK to derive the odds ratio and significant level. Then, we used the software PRSice at a grid size of 0.00005 systematically to obtain the polygenic risk score and chose the best-fit subset of genetic variants. The best-fit polygenic coefficients from the learning set were applied directly in the testing set to see whether the score could differentiate the binary cognitive outcome among schizophrenia patients. The genetic variants included in the final polygenic risk score were then subjected to pathway analysis using the software Ingenuity Pathway Analysis. Results: The PCA analysis identified two principal components of cognitive performance: Executive Function Component (explaining 46.3% of the total variance) and the Attention Component (explaining 21.4% of the total variance). After quality control and removing genetic variants with high linkage disequilibrium, there were 196,770 genetic variants left for further analysis. For the dichotomous Executive Function Component, at a significant level of p = 0.0014, a polygenic risk score consisting of 190 genetic variants could successfully predict the dichotomous Executive Function Component in the testing set (p = 0.002), explaining 2.8% of the variance. However, for the dichotomous Attention Component, we did not find a set of genetic variants from the learning set that could successfully predict the counterpart outcome in the testing set. For the 190 genetic variants included in the polygenic risk score for the dichotomous Executive Function Component, 90 of them could be mapped to 84 known genes. The results of network analysis indicated that the network with NF-κB as the center was the biggest network, suggesting that the genetic variants of the polygenic risk score might be involved in some immune-related function. Conclusions: We successfully identified a polygenic risk score consisting of 190 genetic variants via learning and testing sets that could predict schizophrenia patients’ cognitive performance. Potential functions revealed in the network analysis might shed some light for future studies in the modifier roles of these genetic variants on cognitive impairment in schizophrenia patients.