等位基因突變頻率於骨髓化生不良症候群患者之預後意義

Abstract

一、背景 骨髓化生不良症候群 (myelodysplastic neoplasms, MDS) 是一種造血幹細胞疾病，臨床表徵與預後差異性大，以造血功能不良、血球低下、細胞化生不良與特定染色體變化為主要表徵，部分患者會快速轉變為急性骨髓性白血病，而基因突變模式與預後息息相關，它是導致疾病發生的重要機轉之一，也是使患者疾病惡化的關鍵因素。如何正確區分高風險患者並給予積極治療是目前臨床上急待解決之問題，目前已有數個可以將患者做風險分級的模型包括：International Prognostic Scoring System (IPSS)、revised IPSS (IPSS-R)、World Health Organization Classification-based Prognostic Scoring System以及MD Anderson Prognostic Scoring System，然而這些模型並不包括基因突變模式，包括等位基因突變頻率 (variant allele frequency, VAF) 之表現，因此本研究分析等位基因突變頻率在骨髓化生不良症候群患者之預後意義，並探討去甲基化藥物治療與異體骨髓幹細胞移植的角色。 二、方法與程序 根據2016年世界衛生組織針對骨髓化生不良症候群的疾病分類與診斷標準，排除先前接受過化學治療或是惡性血液腫瘤病史之患者，總共收集698位原發性骨髓化生不良症候群患者的骨髓檢體，使用TruSight myeloid sequencing panel (Illumina)及HiSeq平台，以次世代基因定序 (next generation sequencing) 的方式完成54個基因突變及等位基因突變頻率之分析，針對CEBPA基因另外使用Sanger定序方式做確認。而FLT3-ITD則使用PCR (polymerase chain reaction)，及毛細管電泳 (fluorescence capillary electrophoresis)方式分析。此研究經臺大醫院研究倫理委員會核可，所有患者皆簽屬臨床試驗同意書 (核可號 201709072RINC)。 三、統計分析 連續變相使用曼惠特尼檢定，類別變相使用費雪或卡方檢定，皮爾森相關係數檢定用以判定等位基因突變頻率與臨床指標的相關性，相關係數大於/小於0.4/-0.4被認定為有意義的正/負相關；無白血病存活率 (leukemia-free survival) 定義為診斷至轉變為急性骨髓性白血病，或診斷至死亡的時間；整體存活率 (overall survival) 定義為診斷至死亡的時間。最大選擇統計量 (maximally selected rank statistics) 用以尋找並檢定等位基因突變頻率與預後有統計相關的閾值，並進一步使用自助法 (bootstrapping) 做內部驗證，所有檢定以p值小於0.05視為有統計意義。 四、結果 甲、病人特性 診斷年齡中位數為66.5歲，男性居多 (63.3%)，追蹤中位數為54.7個月，根據2016年世界衛生組織診斷標準，52%的患者屬於高風險族群 (MDS-excess blasts)，若根據IPSS-R風險模型分類，22.5%及22.2%患者分別屬於高或極高風險類別；24.4%患者有接受去甲基化藥物治療，16.1%患者接受異體骨髓幹細胞移植；23%患者於追蹤期間轉變為急性骨髓性白血病，49.3%患者於追蹤期間內死亡。 乙、基因突變模式 71.5%患者至少帶有一個基因突變，最常見的基因突變為ASXL1 (20.3%)、TET2 (14.3%)、 SF3B1 (13.8%)、RUNX1 (12.6%)、STAG2 (12.5%) 及TP53 (12.3%)。 丙、預後分析 單變相預後分析發現年紀較大，IPSS-R高風險類別，曾接受去甲基化藥物治療，TET2、 IDH2、ASXL1、EZH2、CBL、RUNX1、U2AF1、SRSF2、ZRSR2、STAG2及TP53突變與較短的無白血病存活率和整體存活率有關，而DNMT3A、 BCORL1及NRAS突變則與較短的無白血病存活率相關；接受異體幹細胞移植、女性及SF3B1突變與較長的無白血病存活率和整體存活率相關。而在多變項分析中，IPSS-R高風險類別、DNMT3A、TET2、IDH2、CBL及TP53為較差的獨立預後因子，而女性及接受異體幹細胞移植為較好的獨立預後因子。 丁、等位基因突變頻率與臨床指標及預後的關聯 高IDH2等位基因突變頻率與較低的血色素 (r = -0.496, p = .009) 與骨髓芽細胞比例 (r = -0.432, p = .024) 相關，帶有高等位基因突變頻率之DNMT3A (閾值40%, hazard ratio [HR] 2.87, p < 0.001)、TET2 (45%, HR 2.55, p < 0.001)、ASXL1 (20%, HR 2.24, p < 0.001)、EZH2 (40%, HR 2.12, p = 0.036)、SETBP1 (15%, HR 1.94, p = 0.024)、BCOR (80%, HR 2.49, p = 0.043)、SRSF2 (50%, HR 3.65, p = 0.002)、ZRSR2 (60%, HR 2.91, p < 0.001) 及TP53 (25%, HR 7.84, p < 0.001)，相較於無突變患者其無白血病存活率較短，除了EZH2、SETBP1及BCOR之外，其他基因之高等位基因突變頻率亦與較短的整體存活率相關。在多變項分析中，女性及有接受異體幹細胞移植患者預後較好，年紀大、高風險IPSS-R類別、帶有IDH2及CBL突變的患者無白血病存活率和整體存活率較短，而帶有U2AF1突變、DNMT3A或ZRSR2高等位基因突變頻率患者整體存活率較短。若患者帶有等位基因突變頻率> 25%的TP53突變，其預後相較等位基因突變頻率≤ 25%或不帶有TP53突變的患者差。 若將上述帶有獨立預後預測意義的基因 (與無白血病存活率相關之基因：IDH2、CBL及TP53突變；與整體存活率相關之基因：高等位基因突變頻率之DNMT3A、ZRSR2突變，與IDH2、CBL、U2AF1及TP53突變) 納入現有的風險分類模型 (IPSS-R) 中，可以將患者做更好的分類，舉例來說，原本同樣被分類為低或極低風險的IPSS-R患者，若帶有不好的基因突變，相較於其他同樣類別的患者，預後顯著較差。針對整體存活率，若將這些不好的預後因子納入IPSS-R風險分類模型，分別將有8.9%、17.9%及34.4%自原屬於低或極低風險、中等風險或高風險IPSS-R分類中被重新歸類。 戊、去甲基化藥物及異體幹細胞移植之影響 進一步分析治療對於預後的影響，使用去甲基化藥物及異體幹細胞移植無法改善帶有不好基因的患者預後，然而若針對個別特定基因作分析，若帶有U2AF1突變的患者，接受去甲基化藥物治療可顯著改善其預後。 五、討論及結論 針對等位基因突變頻率與預後之關聯，此研究為目前已知最完整的分析之一，在本研究中發現若帶有高等位基因突變頻率之DNMT3A、TET2、ASXL1、SRSF2、ZRSR2及TP53突變患者預後較差，患者若帶有上述基因之低等位基因突變頻率，其預後與未帶有上述突變之患者相當，而TP53的等位基因突變頻率閾值 (25%) 可以將患者分為高、低及未帶有突變三群，三群患者預後不同，以高者為最差，未帶有突變者最好。在多變項分析中，高等位基因突變頻率之DNMT3A、ZRSR2突變，與IDH2、CBL、U2AF1及TP53突變為獨立預後因子，若將這些高風險分子預後因子納入現有風險分類系統 (IPSS-R)，可以將病患做更好的風險分級。除此之外，我們首次證實帶有U2AF1突變的患者，接受去甲基化藥物治療可以顯著改善其預後。 骨髓化生不良症候群的病生理機轉複雜，不同分類其預後差異大，隨著基因定序技術的進展，可以更加完整分析基因突變模式對於疾病進展與預後的影響，除了基因突變的有無外，目前已有眾多證據指出等位基因突變頻率跟預後亦有相關，然而目前大多研究都著重分析TP53突變，先前有研究發現分別以20%及50%為閾值，可將帶有TP53突變的患者分成預後不同的三個族群，亦有統合分析 (meta-analysis) 認為以20%為分界，區分高或低突變負荷量 (mutation burden)。除此之外，因骨髓化生不良症候群患者僅有少數人帶有ZRSR2突變，因此先前的研究無法分析ZRSR2之等位基因突變頻率與預後的關係，而本研究首次得以證實ZRSR2高等位基因突變頻率與較差的預後有關。 本篇研究的限制包括其為一回顧性研究，且針對等位基因突變頻率之閾值與風險分類模型的預測力缺少外部效度 (external validation) 的驗證，然這項研究成果仍可提供未來發展新的風險分類模型的雛型與方向，也促進骨髓化生不良症候群患者個人化且精準醫療發展之可能性。

A. Introduction Myelodysplastic syndrome (MDS) is a diverse group of clonal myeloid neoplasms, characterized by clinical and genetic heterogeneity, and increased risk of acute myeloid leukemia (AML) transformation. The accumulation of mutations is involved in MDS pathogenesis, which gives rise to clonal architecture and leads to disease progression. Several prognostic models, including the International Prognostic Scoring System (IPSS), revised IPSS (IPSS-R), World Health Organization Classification-based Prognostic Scoring System and MD Anderson Prognostic Scoring System have been developed to risk-stratify MDS patients. Mounting evidences demonstrate that the addition of mutation data improves the prognostic stratification. In addition to mutational profiles, the variant allele frequency (VAF) of individual mutations also influence the prognosis in MDS patients. In the present study, we performed comprehensive VAF analyses, focusing on the correlation between VAF and survival. We further analyzed the impacts of allogenic hematopoietic stem cell transplantation (HSCT) and hypomethylating agents (HMA) on outcomes considering various VAF in different genes. B. Material and Method A total of 698 primary MDS patients with adequate cryopreserved bone marrow samples for deep-targeted sequencing and IPSS-R data were recruited. The diagnoses were based on the 2016 World Health Organization (WHO) classification. Patients with antecedent chemotherapy/radiotherapy or hematologic malignancies were excluded. This study was approved by the Research Ethics Committee of the National Taiwan University Hospital; and written informed consent was obtained from all participants (approval number: 201709072RINC). TruSight myeloid sequencing panel (Illumina) and the HiSeq platform were adopted to analyze the gene alterations and mutant allele burdens of 54 myeloid-neoplasm relevant genes. Because of the sequencing sensitivity issue, we verified CEBPA mutations via Sanger sequencing. Analysis of FLT3-ITD was performed via polymerase chain reaction (PCR), followed by fluorescence capillary electrophoresis C. Statistical analysis Pairwise comparison between continuous variables was performed using the Mann–Whitney U test, and the Fisher’s exact test or the χ2 test was performed for discrete variables. Pearson’s correlation coefficient was used to assess the strength and direction of the linear relationships between VAF and clinical parameters. The correlation coefficient (r) greater or lower than 0.4/-0.4 was thought to be positive/negative correlated. Leukemia-free survival (LFS) was defined as the duration from the date of diagnosis to the last follow-up, documented leukemia transformation, or death from any cause, whichever occurred first. Overall survival (OS) was the duration from the date of diagnosis to the last follow-up or death from any cause, whichever occurred first. Maximally selected rank statistics were applied for VAF exploration. All P values were two-sided and considered statistically significant if <0.05. D. Results a. Demographic features The median age was 66.5 years, with male predominance (63.3%). The median follow-up time was 54.7 months. When categorized by the 2016 WHO classification, over half (52.0%) of the patients had myelodysplastic syndrome with excess blasts, including EB1 (23.8%) and EB2 (28.2%). A total of 71.1% patients had IPSS-R intermediate (26.4%), high (22.5%), or very high-risk disease (22.2%). Regarding treatments, 24.4% of patients received HMA and 16.1% underwent allogeneic HSCT. Twenty three percent of patients experienced leukemic transformation and 49.3% died at the end of follow-up. Overall, 71.5% had at least one gene mutation. The most common mutation in the cohort was ASXL1 mutation (20.3%), followed by TET2 (14.3%), SF3B1 (13.8%), RUNX1 (12.6%), STAG2 (12.5%), and TP53 mutations (12.3%). b. Survival analyses In univariable analysis, older age, higher-risk IPSS-R, HMA treatment, and presence of mutations in TET2, IDH2, ASXL1, EZH2, CBL, RUNX1, U2AF1, SRSF2, ZRSR2, STAG2, and TP53 were significantly associated with both shorter LFS and OS, while DNMT3A, BCORL1, and NRAS mutations conferred shorter LFS. Receiving HSCT, female sex, and mutated SF3B1 were favorable factors for LFS and OS. Multivariable analysis showed that older age, higher-risk IPSS-R, and DNMT3A, TET2, IDH2, CBL, and TP53 mutations were independent poor risk factors, while female sex and receiving HSCT were good risk factors for both LFS and OS. c. VAF of mutations and the correlation with clinical parameters and outcomes VAF of IDH2 mutation had impact on clinical features; higher VAF of IDH2 was associated with lower hemoglobin level (r=-0.496, p=0.009) and lower bone marrow blasts percentage (r=-0.432, p=0.024). Compared with wild-type genes, high VAF of mutations in 9 genes, including DNMT3A (cutoff value 40%, HR 2.87, p<0.001), TET2 (45%, HR 2.55, p<0.001), ASXL1 (20%, HR 2.24, p<0.001), EZH2 (40%, HR 2.12, p=0.036), SETBP1 (15%, HR 1.94, p=0.024), BCOR (80%, HR 2.49, p=0.043), SRSF2 (50%, HR 3.65, p=0.002), ZRSR2 (60%, HR 2.91, p<0.001) and TP53 (25%, HR 7.84, p<0.001) were significantly associated with shorter LFS. With the exception of EZH2, SETBP1, and BCOR mutations, high VAF of all other six mutations were also associated with shorter OS. In multivariable analysis, female sex and receiving HSCT were independent favorable factors for both LFS and OS, while older age and higher-risk IPSS-R predicted shorter LFS and OS. Mutant IDH2 and CBL predicted both shorter LFS and OS, while U2AF1 mutation and DNMT3A and ZRSR2 mutations with high VAF were associated poorer OS. Regarding TP53 mutations, patients with high VAF had the worst outcomes compared to those with wild type TP53 or low VAF. The presence of poor-risk mutations (DNMT3A and ZRSR2 mutations with high VAF, mutant TP53, IDH2, CBL and U2AF1 for OS; mutant TP53, IDH2 and CBL for LFS) could re-stratify the IPSS-R risk groups. For instance, in the IPSS-R low and very low-risk group, the patients with poor-risk mutations had an OS significantly shorter than those without poor-risk mutations. Considering OS, the incorporation of the molecular data in the IPSS-R could reclassify 8.9% (18/202) of IPSS-R very low/low-risk patients to intermediate-risk subgroup, 17.9% (33/184) of IPSS-R intermediate to high-risk subgroup, and 34.4% (54/157) of IPSS-R high to very high-risk subgroup. d. Impact of hypomethylating agents and HSCT on survival in patients with poor-risk mutations The use of HMA or HSCT could not significantly improve LFS or OS in patients with at least one of the poor-risk mutations. However, focusing on specific mutations, patients harboring U2AF1 mutation had similar LFS and OS compared with those with wild-type U2AF1 if they received HMA treatment. E. Discussion and Conclusion To the best of our knowledge, the present study was one of the most comprehensive researches that investigated the clinical significance of VAF in a large number of myeloid-malignancies related gene mutations in MDS patients. We demonstrated that high VAF of DNMT3A, TET2, ASXL1, SRSF2, ZRSR2 and TP53 mutations were significantly associated with shorter LFS and OS. Patients with low VAF of DNMT3A, TET2, ASXL1, SRSF2, and ZRSR2 mutation, respectively had outcome comparable to those with the wild-type gene. For TP53 mutation, the VAF level could separate patients into three risk groups with distinct outcomes. Further, in multivariable analysis, high VAF of DNMT3A and ZRSR2 mutations independently predicted poorer OS. Moreover, the presence of mutations in TP53, IDH2, CBL, and U2AF1 also conferred a worse prognosis. With the advances of next-generation sequencing technologies, which are more powerful for comprehensive mutation analysis and more sensitive to identify rare variants and mutations with low-frequency. In addition to mutations, mounting evidences have shown VAF of mutations also have clinical significances. One of the previous studies demonstrated that TP53 mutated patients could be segregated into three groups with distinct outcomes using 20% and 50% as cutoff values for VAF. A meta-analysis suggested a threshold of 20% as a rough line between high and low clone burden of TP53 mutation. Due to the relatively lower rate of ZRSR2 mutation in MDS, data for its clinical impacts are limited. Here, we demonstrated the association between ZRSR2 mutation and poor outcomes and found that the prognostic impact of ZRSR2 mutation depended on its VAF. The limitations of our study include its retrospective nature and the lack of external validation to confirm the prognostic significance of the VAF cutoff levels we set. However, our data fostered our understanding of the mutation burden of the diseases and provided future patient-tailored therapeutic avenues.