第一部分：骨髓增殖性肿瘤的表型异质性分析及治疗靶点的探索
摘要
背景：经典的BCR-ABL阴性骨髓增殖性肿瘤（myeloproliferative neoplasms, MPN）是一组克隆性造血系统疾病，包括真性红细胞增多症（polycythemia vera, PV）, 原发性血小板增多症（essential thrombocythemia,  ET）和原发性骨髓纤维化（primary myelofibrosis, PMF）。JAK2V617F作为MPN中最常见的突变，是影响MPN表型重要的而非唯一的决定因素。目前的研究已经揭示了某些导致MPN表型异质性的因素，包括驱动突变、额外突变、突变发生顺序、肿瘤负荷和造血干细胞异质性等。然而，单个JAK2V617F突变，甚至是相同的JAK2V617F突变负荷是如何导致不同的疾病表型仍未完全阐明。
目的：揭示MPN表型异质性的影响因素并探索MPN的新型治疗靶点。
方法：对JAK2V617F突变ET和PV患者的二代测序数据进行分析，通过比较基因突变频率及突变负荷，以寻找影响表型异质性的分子学异常。分别对ET、PV患者以及正常人的造血干细胞（hematopoietic  stem  cells,  HSC）和巨核红系祖细胞（Megakaryocyte–erythroid progenitor, MEP）进行转录组分析以探索不同疾病表型在造血分化中差异调节的信号通路。靶向Wnt/β-catenin信号通路在抑制巨核细胞分化和血小板生成中的作用通过体外巨核分化实验进行验证。巨核细胞表型通过以下方法进行验证：光学显微镜下观察细胞形态，免疫荧光法分析CD41和CD61荧光阳性细胞比例和平均荧光强度，巨核集落形成单位分析巨核集落的大小和数量，流式细胞术分析CD41和CD42b双阳性细胞比例以及巨核细胞倍性。对JAK2V617F突变MPN小鼠腹腔注射XAV939后进行血常规分析以证实Wnt/β-catenin抑制剂在体内的降血小板作用。对小鼠骨髓各造血细胞比例进行分析以研究药物降低血小板水平的作用机制。
结果：二代测序数据分析识别了两个影响MPN表型的共存突变。其中SH2B3基因突变主要发生于PV患者，SF3B1基因突变更常见于ET患者。PV患者的JAK2V617F突变负荷常高于ET患者。受试者工作特征曲线界定的区分PV和ET表型的JAK2V617F突变负荷的最佳临界值为34.8%。ET、PV患者以及正常人的HSC和MEP转录组分析结果表明，相比于PV的HSC和MEP，炎症信号通路在ET的HSC和MEP中均上调。此外，Wnt/β-catenin信号通路在ET的HSC至MEP分化过程中特异性上调。体外阻断Wnt/β-catenin信号通路后，巨核细胞分化、成熟以及血小板的生成受抑制。相比于正常人，ET患者来源的造血干祖细胞对抑制剂更加敏感，相同的药物浓度下，其巨核细胞分化的受抑制程度更高。JAK2V617F突变MPN小鼠给予Wnt/β-catenin抑制剂体内给药后，其血小板水平出现明显降低。小鼠骨髓各造血细胞流式免疫表型分析结果表明，Wnt/β-catenin抑制剂降低血小板的作用机制是阻断MEP和巨核祖细胞（Megakaryocyte progenitor, MKP）以及抑制巨核细胞成熟。相比之下，Wnt/β-catenin抑制剂在野生型小鼠中无明显降血小板作用。
结论：本研究在造血分化层面对JAK2V617F突变MPN的表型异质性机制提供了新的见解。Wnt/β-catenin信号通路在ET造血分化中特异性上调。抑制Wnt/β-catenin信号通路能抑制巨核细胞体外分化，同时降低JAK2V617F 突变小鼠的血小板水平，因此可能作为MPN的潜在治疗靶点。

第二部分：真性红细胞增多症血栓风险因素分析及多因素预测模型的建立
摘要
背景：真性红细胞增多症（polycythemia vera, PV）属于经典的BCR-ABL阴性骨髓增殖性肿瘤（myeloproliferative neoplasm, MPN），以成熟红细胞异常增多、高血栓发生率为特征。血栓及血栓相关并发症是PV患者死亡的重要原因。传统的PV血栓分层模型根据年龄以及有无既往血栓将患者分为低危组以及高危组。然而这种二分层的血栓模型可能会忽略某些潜在的风险因素。
目的：建立一个适用于2016年世界卫生组织（World Health organization, WHO）标准诊断的PV患者的多因素血栓预测模型，并对模型进行验证。
方法：纳入本中心PV患者作为模型来源队列，以及外院PV患者作为模型验证队列。统计并分析两个不同队列来源的PV患者的临床信息和二代测序数据。采用多因素Cox回归分析识别血栓风险因素，并在此基础上建立预测模型。模型的区分度采用受试者操作特征曲线下面积（Area Under Curve, AUC）和C指数进行评估，模型的校准度采用校准曲线和拟合优度检验进行评估。不同风险组患者的无血栓生存采用KaplanMeier生存曲线以及log-rank检验进行比较。
结果：本研究共纳入372名PV患者作为模型建立队列，以及外院195名PV患者作为模型验证队列。内部队列中共有136名患者发生了血栓事件，其中122名患者仅发生了动脉血栓，7名患者仅发生了静脉血栓，另有7名患者既有动脉也有静脉血栓。其中，68名患者发生了诊断后血栓，随访期间血栓发生率为3.6/100人年。多因素分析发现年龄≥60岁[风险比（hazard ratio, HR）=2.56, 95%置信区间（confidence interval, CI）1.51–4.35, p<0.001]、心血管风险因素（cardiovascular risk factor, CVF）（HR=4.22, 95%CI  2.00–8.92,  p<0.001）、至少一种血栓高风险突变（包括DNMT3A、ASXL1或BCOR/BCORL1）（HR=4.35, 95%CI 2.62–7.21, p<0.001）和既往血栓（HR=5.93, 95%CI 3.29–10.68, p<0.001）是诊断后血栓的独立风险因素。将非血栓原因所致死亡作为诊断后血栓的竞争风险事件进行竞争风险分析后，发现以上四个因素仍为诊断后血栓的风险因素。按照回归系数β值的权重将年龄≥60岁赋值1分，CVF赋值1.5 分，血栓高风险突变赋值1分，既往血栓赋值2分，建立了一个多因素血栓预后评分系统（MFPSPV）。MFPS-PV模型将患者分为低危（≤1分; 96例, 25.8%）、中危（1.5–2.5分; 146例, 39.2%）及高危（≥3分; 130例, 35.0%）组，三个风险组患者的无血栓生存率具有明显统计学差异（p<0.001）。MFPS-PV在血栓预测的区分度上优于传统模型[C指数：0.87（95%CI 0.83–0.91）vs. 0.80（95%CI 0.74–0.86）；AUC：0.90（95%CI 0.86–0.95）vs. 0.82（95%CI 0.76–0.89）]，且具有良好的校准度。MFPS-PV应用于外部队列仍然表现出较好的血栓预测能力。
结论：MFPS-PV首次整合了临床特征和二代测序数据，将分子遗传学指标纳入血栓预测中，在2016年WHO标准诊断的PV患者中表现出良好的血栓预测能力。

Part  1:  Molecular  analysis  of  phenotypic  heterogeneity  in  JAK2V617Fpositive myeloproliferative neoplasms reveals a potential target for therapy
Abstract
Background: Classical BCR-ABL-negative  myeloproliferative  neoplasms  (MPNs)  are  a group  of  clonal  hematopoietic  disorders, including  polycythemia  vera  (PV),  essential thrombocythemia  (ET),  and  primary  myelofibrosis.  JAK2V617F  is  the  most  frequent mutation in MPNs. It is an important but not the only determinant of MPN phenotype. Recent studies  have  revealed  several  factors  contributing  to  phenotypic  heterogeneity  of  MPN, including  driver  mutation,  additional  mutation, mutational  order,  tumor  loads  and hematopoietic stem cell (HSC) heterogeneity. However, it is still not fully elucidated how a single  JAK2V617F  mutation,  and  even  the  same  tumor  load,  lead  to  different  disease phenotypes.
Objective: The aim of the study is to unveil factors involved in phenotypic heterogeneity and to identify novel therapeutic targets for MPN.
Methods: Next-generation sequencing (NGS) was performed on JAK2V617F+ET and PV patient  samples  to  find  molecular  markers  revealing  phenotypic  heterogeneity.  RNA sequencing  was  performed  on  HSCs  and  megakaryocyte-erythroid  progenitors  (MEPs)  of healthy donors and MPN patients to find signaling pathways differentially regulated during hematopoietic differentiation of different disease phenotypes. The effects of targeting Wnt/β-catenin signaling on inhibiting megakaryocyte (MK) and platelet production were confirmed by  in  vitro  MK  differentiation. The  phenotype  of  MK  was  identified  using the  following methods: observation of cell morphology under microscope, analysis of the proportion and mean  fluorescence  intensity  of  CD41- and  CD61-positive  cells  by  immunofluorescence,analysis of the size and number of MK colonies by colony forming units-MK assays, analysis of the  proportion  of  CD41- and  CD42b-positive  cells  and MK ploidy by flow  cytometry analysis. JAK2V617F+MPN mice were injected with XAV939, and then underwent blood routine test to validate the effect of thrombocytosis remission of Wnt/β-catenin inhibitor in vivo. Mice bone marrow hematopoietic stem progenitor cells (HSPCs) were analyzed to study the mechanism of action of the drug in reducing platelet levels.
Results: Using NGS technology, two concurrent mutations that may affect phenotype were identified, including mutations in SH2B3, which is primarily prevalent in PV, and SF3B1, which is more commonly mutated in ET. Higher JAK2V617F allele burden was associated with PV phenotype. A receiver operating characteristic curve identified JAK2V617F allele burden  of  34.8% as the optimal  cut-off  level  to  distinguish  PV  and  ET.  Transcriptomic analysis of HSCs and  MEPs  from  ET,  PV and healthy  donors  showed  that  inflammatory signaling pathways were elevated in both HSCs and MEPs of ET, unlike in PV HSCs and MEPs. Notably, Wnt/β-catenin signaling was uniquely upregulated during ET hematopoietic differentiation  from  HSC  to  MEP,  and  inhibiting  Wnt/β-catenin  signaling  blocked  the formation of MK and platelets in vitro. CD34+cells derived from ET patients were more sensitive  to  Wnt/β-catenin  inhibitor compared with that  from  healthy  donors,  and  MK differentiation from ET cells was more inhibited at the same drug concentration. Consistently, Wnt/β-catenin inhibitor administration decreased platelet counts in JAK2V617F+MPN mice by blocking MEPs and megakaryocyte progenitors and by inhibiting MK maturation. While in wild-type mice, Wnt/β-catenin inhibitor did not significantly reduce platelet counts.
Conclusion: Our findings provide new insights into the mechanisms underlying phenotypic 
differentiation  of  JAK2V617F+ PV and  ET at  the  level  of  hematopoietic  differentiation. Wnt/β-catenin signaling is specifically upregulated during ET hematopoietic differentiation. Inhibiting Wnt/β-catenin signaling can inhibit MK differentiation in vitro and reduce platelet  counts in vivo, indicating that Wnt/β-catenin signaling may serve as a potential therapeutic target of MPN.

Part 2: Development and validation of a multiple factor-based prognostic score system of thrombosis in polycythemia vera
Abstract
Background:  Polycythemia  vera  (PV)  is  a  subtype  of  classic  BCR-ABL-negative myeloproliferative neoplasm (MPN), characterized by abnormally elevated mature red blood cells  and  high  incidence  of  thrombosis. Thrombosis and  thrombotic  complications  are important causes of death for patients with PV. The conventional stratification of thrombosis divides patients into low-risk and high-risk groups according to age and previous thrombosis.However, this two-tiered thrombosis stratification may ignore some potential risk factors.
Objectives: This  study  aimed  to  develop  and  validate  a  multiple  factor-based  prediction  model of thrombosis for the 2016 World Health Organization (WHO)-defined PV.
Methods: Patients from our institution were included as training cohort, and patients from another  center  were  included  as external  validation  cohort. Clinical  and  next-generation sequencing (NGS) data from two cohorts of patients with PV were analyzed. Multivariable Cox regression analysis was conducted for identification of thrombotic risk factors and model development. The  discrimination  of  the  model  was  evaluated  by  the  area  under receiver operating  characteristic curve (AUC) and  the  C  index. The calibration  of  the  model  was evaluated by the calibration curve and the goodness-of-fit test. Thrombosis-free survival of patients  in  different  risk  groups  was displayed using  Kaplan-Meier  survival  curves and compared by log-rank test.
Results: The study involved 372 patients in the training cohort and another 195 patients in the external validation cohort. In the training cohort, thrombotic events were recorded in 136 patients (36.6%), including 122 patients (32.8%) with only arterial thrombosis, 7 patients (1.9%) with only venous thrombosis, and 7 patients (1.9%) with both arterial and venous thrombosis. There were 68 patients (18.3%) with thrombosis after diagnosis, corresponding to  an  incidence  rate  of  3.6/100  patient-years  during  follow-up. Multivariable  analysis indicated that age≥60 years (hazard ratio [HR] 2.56, 95% confidence interval [CI] 1.51–4.35, p<0.001),  cardiovascular  risk  factors  (HR  4.22, 95%CI  2.00–8.92,  p<0.001),  at  least  one high-risk  mutation  for  thrombosis (mutations  in  DNMT3A, ASXL1,  or  BCOR/BCORL1) (HR 4.35, 95%CI 2.62–7.21, p<0.001) and previous thrombosis (HR 5.93, 95%CI 3.29–10.68, p<0.001) were independent risk factors of thrombosis. Those risk factors remained significant contributors to thrombosis after accounting for the competing risk of death. According to the weight of β coefficient, those factors above were assigned 1 point, 1.5 points, 1.5 points and 2 points respectively. A multiple factor-based prognostic score system of thrombosis (MFPSPV) was then  developed,  classifying  patients  into  low-risk  (≤1  point; n=96,  25.8%), intermediate-risk (1.5–2.5 points; n=146, 39.2%) and high-risk (≥3 points; n=130, 35.0%)groups. Patients in the three groups had significantly different thrombosis-free survival rates (p<0.001). The MFPS-PV outperformed the conventional model in discrimination power (Cstatistic: 0.87 [95%CI 0.83–0.91] vs 0.80 [95%CI 0.74–0.86]; AUC: 0.90 [95%CI 0.86–0.95] vs. 0.82 [95%CI 0.76–0.89]). The MFPS-PV was well-calibrated and remained consistent during external validation.
Conclusions: The MFPS-PV, integrating clinical characteristics and NGS data for the first time, successfully incorporated molecular genetic indicators into thrombosis prediction, and showed excellent accuracy and utility for the 2016 WHO-defined PV.