與思覺失調症之頭顱顏面特徵相關之遺傳變異：由多基因風險分數構成的修飾因子

Abstract

背景與目的 思覺失調症患者及其一等親出現細微身體特徵異常或頭顱顏面特徵的機率較一般人口高；而思覺失調症的部分臨床表徵也被發現與較多的頭顱顏面特徵有關，顯示這些特徵可能扮演修飾因子的角色。根據家族研究，這些特徵具有一定的遺傳率，但目前少有已知的基因標誌與其相關。本研究透過全基因體關聯分析，尋找與思覺失調症患者頭顱顏面特徵有關之遺傳變異，並評估以此建立的多基因風險分數是否可以預測這些特徵。此外，本研究探討這些頭顱顏面特徵與哪些臨床表徵有關，以進一步瞭解這些頭顱顏面特徵相關之多基因在思覺失調症臨床表現上可能扮演的角色。 方法 本研究設計採用全為病例研究，樣本來自台灣精神分裂症之全基因體研究計畫 (Schizophrenia Trio Genomic Study of Taiwan)，總共納入1,057位思覺失調症患者。對每位病人進行7項頭顱顏面特徵的測量、神經認知功能測驗、遺傳研究診斷問卷訪談 (Diagnostic Interview for Genetic Studies)，以及使用精神病基因體聯盟 (Psychiatric Genomics Consortium) 所設計的PsychChip進行60萬個位點的基因型定型。利用基因標誌為基礎所估計之遺傳率，我們挑選該值可估出的三項頭顱顏面特徵 (額突、鼻上翹、頭圍) 進行後續分析。比較具這些頭顱顏面特徵者與不具有任一特徵者，以探討其與遺傳變異位點的相關性。我們先將病人以分層隨機分派方式，分成探索組 (n=527) 與驗證組 (n=530)。針對探索組病人，先以PLINK (v1.07) 軟體估出每一遺傳變異的相關勝算比及其顯著水準。然後再以PRSice (v1.25) 軟體有系統地變化顯著水準值，計算出對應的多基因風險分數。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直接套用在驗證組，看是否仍能成功區分有頭顱顏面特徵者與無這些特徵者。對於納入最佳的多基因風險分數中的遺傳變異，則以Ingenuity Pathway Analysis 軟體進行路徑分析。 結果 比起不具任一頭顱顏面特徵的病人，具有較多頭顱顏面特徵的病人，其混合慣用手的比率較低，且持續專注力的表現較差 (p<0.05)；但並未發現這些特徵與臨床症狀有關聯。在針對基因定型資料進行品質管控並剔除高度連鎖不平衡 (linkage disequilibrium) 的基因位點後，共有196,770個位點被納入後續分析。在探索組樣本中，沒有任何個別位點與頭顱顏面特徵的關聯達到全基因體顯著水準。多基因風險分數分析中，我們發現在顯著水準閾值為0.00105的情況下，由150個位點組合而成的分數，可以在驗證組成功地預測病人頭顱顏面特徵的出現 (p=0.008)。這些遺傳變異可對應至75個已知基因位置；路徑分析的結果暗示它們可能牽涉一些DNA複製與脂質代謝等功能。 結論 本研究成功地建立一組由150個遺傳變異組成的多基因風險分數，通過探索組與驗證組的測試，可以預測思覺失調症病人之頭顱顏面特徵。由路徑分析所暗示的功能，可以作為未來進一步了解這些頭顱顏面特徵相關基因如何修飾思覺失調症的基礎。

Background Patients with schizophrenia and their first-degree relatives show more minor physical anomalies (MPAs) and certain craniofacial quantifications than the general population. Higher frequencies of several clinical characteristics were also observed in schizophrenia patients with those MPAs or craniofacial quantifications, indicating the craniofacial features might serve as a potential modifier. Family studies found moderate heritability of the craniofacial features, but to date no large-scale genetic study was conducted to identify their genetic markers. This study aims to 1) use genome-wide association studies (GWAS) approach to identify potential genetic variants contributing to the presence of the craniofacial features in schizophrenia, and evaluates their genetic contribution by means of polygenic risk score analysis; and 2) examine the relationships between the craniofacial features and clinical characteristics in schizophrenia. Methods This study was a case-only design (n = 1,057), and the study sample was from Schizophrenia Trio Genomic Study of Taiwan project. Several tools were applied for data collection, including a 7-item questionnaire for MPAs and craniofacial quantifications, the Diagnostic Interview for Genetic Studies, neurocognitive tests, and PsychChip microarray for genotyping, which was developed by Psychiatric Genomics Consortium and contained 600,000 markers. We kept MPAs items with available SNPs-based heritability (frontal bossing, nostril anteverted and head circumference beyond one standard deviation) for further analysis. The patients were divided into two groups by stratified randomization, the learning set (n = 527) and the testing set (n=530). The association analysis for searching the related genetic markers was conducted using PLINK (v1.07) in the learning set, by comparing patients with any one of the three features to those without. The significance level and odds ratio of each genetic variant were derived and applied in the testing set for polygenic risk score analysis using PRSice (v1.25), a software featuring finding the best-fit subset of genetic variants. For those genetic variants included in the final polygenic risk score, they were subjected to pathway analysis using Ingenuity Pathway Analysis. Results After data quality control and excluding genetic variants with high linkage disequilibrium, a total of 1,057 schizophrenia patients and 196,770 genetic variants remained for final analysis. Patients with more craniofacial features showed higher frequencies of being fully-left and fully-right handers and poorer CPT performance (p<0.05), but no associations between craniofacial features and clinical symptoms were identified. In the learning set, no single marker reached the genome-wide significance level. A polygenic risk score consisting of 150 genetic variants was significantly associated with patients with more MPAs in the testing set, with a p-value threshold at 0.00105 (p = 0.008). Totally 75 genes were mapped from the 150 genetic variants, and the results of pathway analysis revealed that the functional annotation of these genes were mainly involved in DNA replication or lipid metabolism pathways. Conclusions We successfully identify a polygenic risk score consisting of 150 genetic variants via learning and testing sets that could predict the presence of craniofacial features in schizophrenia patients. Potential functions revealed in pathway analysis could shed light for future studies on the modifier roles of these genetic variants on clinical manifestations of schizophrenia patients.