Myelofibrosis (MF), primary (PMF) or secondary (SMF) to polycythemia vera (PPV-MF) or essential thrombocythemia (PET-MF), is characterized by a complex molecular architecture, including mutations in driver (JAK2, CALR, MPL) and other myeloid neoplasm associated genes.¹

Mutations of SF3B1 (splicing factor 3B subunit 1), a component of the U2-small nuclear ribonucleoprotein complex, are uncommon in MF and are reported in <10% of cases.² Conversely, they are present in >80% of cases of myelodysplastic syndromes (MDS) with ring sideroblasts (MDS-RS) and in MDS/MPN with RS and thrombocytosis (MDS/MPN-RS-T) representing a disease-defining genetic event. Indeed, a proposal to consider SF3B1 mutated MDS as an individual entity characterized by an overall better prognosis, has been made.³ In general, the majority of SF3B1 mutations occur within codon 700 resulting in a large-scale downregulation of messenger RNA (mRNA).^{4, 5} The characteristics of SF3B1 mutations outside this hotspot, and their downstream effects, are largely unclear; in this regard, SF3B1 K666N in MDS patients was recently associated with distinctive RNA splicing profile, clinical features and worse outcomes.⁶

While SF3B1 mutations are reportedly lacking prognostic significance in PMF,^{7, 8} no information are available in the setting of SMF. The main objective of the current study was to compare the prevalence and clinical, molecular and prognostic correlates of SF3B1 mutations in SMF versus PMF patients.

After Institutional review board (IRB) approval (#14560), all consecutive patients with PMF and SMF diagnosis, respectively in according to World Health Organization (WHO) and International Working Group for MPN Research and Treatment (IWG-MRT) criteria, and known SF3B1 mutation status, available in our databases, were included in the study. Molecular analysis for driver and non-driver genes was done at diagnosis or first referral, as per previously published methods. Distribution of continuous variables was compared using nonparametric (Mann–Whitney or Kruskal-Wallis) tests, while nominal variables were compared with the Chi-Square test. Time to event analysis (overall [OS] and leukemia-free survival [LFS]) was performed using the method of Kaplan–Meier, with death (for OS), AML progression (for LFS), and allogeneic hematopoietic stem cell transplantation (HSCT; for OS and LFS) used as censors. Statistical analyses were performed with SPSS software, version 28 (IBM-Corp), JMP Pro 15.1.0 software from SAS Institute (Cary, NC) and Statistical Package R version 4.1.1.

In the present study we retrospectively analyzed 520 patients with MF, 325 (62.5%) PMF including 156 (48%) pre-fibrotic and 169 (52%) overt PMF, and 195 (37.5%) SMF, including 85 (44%) PPV-MF and 110 (56%) PET-MF.

Considering PMF patients, median age was 59 years and 62% were males; median follow-up was 4.5 years (range, 0.06–36) (Table S1). Driver mutations distribution was 58% JAK2, 26% CALR, 6% MPL, 12% triple negative (TN) and 2% double driver-mutated. The numbers of documented deaths and leukemic transformations were 121 (37%) and 37 (12%), respectively. The most frequent non-driver mutated genes were ASXL1 (30%), TET2 (20%) and SRSF2 (10%). SF3B1 mutations were detected in 22 patients (7%; 14 overt-PMF and 8 pre-fibrotic MF) as follows: 11 (50%) K666T/N/M, 7 (32%) K700E, 2 (9%) H662Q/Y, 1 (4%) R625L and 1 (4%) N626S. Of the 19/22 SF3B1 mutated patients, 12 (54%) were JAK2V617F, 6 (27%) CALR and 1 (4%) MPL mutated. Considering SF3B1 mutated patients, the most frequent non-driver co-mutated genes included ASXL1 (n = 4; 18%) and TET2 (n = 3; 14%) followed by DNMT3A, EZH2 and TP53 (n = 2; 9% for each). SF3B1 mutated PMF patients were mostly male (86% vs. 60%; p = .01), older (65 vs. 62 years; p = .004) and displayed a quite significant higher platelet count (561 × 10⁹/L vs. 402 × 10⁹/L; p = .06), and lower Hb levels (11.3 g/dL vs. 12.3 g/dL; p = .09) than the wild-type counterpart.

Among a total of 195 SMF patients included in the analysis, median age was 63 years and 55% were male; median follow-up was 3.9 years (range, 0.07–28). There were 80 (41%) documented deaths and 27 (14%) leukemic transformations (Table S2). Driver mutations were distributed as follows: 70% JAK2V617F, 23% CALR, 8% MPL, and 2% TN; 13 (7%) patients were double driver-mutated. The most frequent non-driver mutated genes were ASXL1 (30%), TET2 (18%) and ZRSR2 (8%). SF3B1 mutation was detected in 10 patients (5%) and included 4 (40%) K666T/N/M, 4 (40%) K700E, 1 (10%) G751V, 1 (10%) G742D. All SF3B1 mutated patients carried also a driver mutation; in particular, 5 (50%) were JAK2V617F, 4 (40%) CALR and 1 (10%) MPL mutated. Considering SF3B1 mutated patients, the pattern of most frequent non-driver co-mutated genes included TET2 (n = 4; 40%) followed by CBL, RUNX1 and TP53 (n = 2; 20% for each). SF3B1 mutated SMF patients were older (66 vs. 62 years; p = .09) and displayed splenomegaly less frequently (p = .05). Moreover, correlative analysis revealed that SF3B1 mutation was mostly associated with CBL (p = .004), RUNX1 (p = .02) and KRAS (p = .03) mutations, in comparison with SF3B1 wild-type counterpart.

We then investigated the impact of SF3B1 mutation on OS and LFS. Considering PMF patients, SF3B1 mutation did not impact on OS (HR 1.1; 95% CI 0.6–2; p = .8, Figure 1A) and LFS (HR 0.7; 95% CI 0.2–2.8; p = .6, not shown). Conversely, considering SMF cases, OS was negatively affected by SF3B1 mutation compared to wild-type counterpart (HR 3.2; 95% CI 1.5–7; p = .002, Figure 1B) whereas LFS resulted similar (HR 1.1; 95% CI 0.1–8.4; p = .9; not shown). Of note, in two cases of post-ET SMF an SF3B1 mutation was already present in the ET phase of disease.

With the aim to better decipher the individual contribution of SF3B1 mutation on OS in SMF patients, we performed a Cox regression analysis that included both clinical and molecular variables (Table S3). In addition to SF3B1, univariate analysis identified age ≥ 65 years (HR 5; 95% CI 3.1–8.3; p = .003), leukocytes >25 × 10⁹/L (HR 2, 95% CI 1.1–3.7; 0.03), Hb < 11 g/dL (HR 3.1, 95% CI 2–5.1; p < .0001), platelet count <150 × 10⁹/L (HR 2.7 95% CI 1.6–4.6; p = .0003), circulating blasts ≥3% (HR 4, 95% CI 2.1–7.5; p < .0001) and mutations of CBL (HR 4, 95% CI 1.7–9.3; p = .001), SRSF2 (HR 8.1, 95% CI 2.5–26.7; p = .0006), U2AF1 (HR 2.7, 95% CI 1.2–6; p = .02) and TP53 (HR 2.9, 95% CI 1.4–5.8; p = .003) as risk factors for a reduced overall survival. In the context of all the aforementioned risk factors, multivariable analysis confirmed the independent prognostic contribution of mutated SF3B1 (HR 2.8, 95% CI 1.2–6.3; 0.02), along with older age (p < .0001), hemoglobin <11 g/dL (p = .02), circulating blasts ≥3% (p = .0003) and CBL mutations (p = .01). The low number of cases prevented us by analyzing the individual prognostic contribution of different SF3B1 mutations.

SF3B1, along with SRSF2 and U2AF1 are the three most frequently mutated spliceosome genes in the context of myeloid neoplasms; in PMF, SRSF2 and U2AF1 (mainly Q157) mutations, but not SF3B1, were associated with inferior survival and are incorporated in the mutation-enhanced international prognostic systems (MIPSS)-70/v2.0 scores.^{9, 10} In the present study we analyzed, separately, the clinical, molecular and prognostic correlates of SF3B1 mutations in PMF and SMF. We confirmed the infrequent occurrence of SF3B1 mutations in both PMF (7%) and SMF (5%); 80% of the latter were in the post-ET subset. All the reported spliceosome gene mutations were mutually exclusive. Importantly, in multivariable analysis, also including clinical and molecular variables comprising the MYSEC-PM¹¹ prognostic score, SF3B1 mutations were independently associated with reduced OS in SMF (Table S3). Recently, spliceosome gene mutations occurring in patients with ET (SF3B1, SRSF2 and U2AF1) and PV (SRSF2) were included in MIPSS PV-ET¹² prognostic score, since they adversely affected OS.

The principal limitation of the current study is its retrospective nature; moreover, despite the high number of MF patients included, the number of SF3B1 mutated PMF and SMF cases remained limited, making the findings reported herein largely explorative, requiring validation in larger independent cohorts. However, along with recent data that highlighted a differential prognostic role of ASXL1 mutations in PMF versus SMF,¹³ these findings strengthen the concept that PMF and SMF represent two largely different biological entities. In this regard, the development of integrated prognostic models specific for patients with SMF remains an unmet clinical need.

骨髓纤维化 (MF)、原发性 (PMF) 或继发性 (SMF) 至真性红细胞增多症 (PPV-MF) 或原发性血小板增多症 (PET-MF)，其特征在于复杂的分子结构，包括驱动突变（JAK2、CALR、MPL）和其他骨髓肿瘤相关基因。^1个

SF3B1 （剪接因子 3B 亚基 1）的突变是 U2 小核核糖核蛋白复合物的一个组成部分，在 MF 中并不常见，并且在 <10% 的病例中有报道。²相反，它们存在于 >80% 的骨髓增生异常综合征 (MDS) 伴环状铁粒幼细胞 (MDS-RS) 和 MDS/MPN 伴 RS 和血小板增多症 (MDS/MPN-RS-T) 的病例中，代表疾病定义基因事件。事实上，有人提议将SF3B1突变的 MDS 视为一个整体预后较好的个体实体。³一般来说，大多数SF3B1突变发生在密码子 700 内，导致信使 RNA (mRNA) 的大规模下调。^4、5的特点该热点之外的SF3B1突变及其下游影响在很大程度上尚不清楚；在这方面， MDS 患者中的SF3B1 K666N 最近与独特的 RNA 剪接谱、临床特征和更差的结果相关。^6个

虽然据报道SF3B1突变在 PMF 中缺乏预后意义，^{7, 8}但在 SMF 的情况下没有可用的信息。本研究的主要目的是比较SMF 与 PMF 患者中SF3B1突变的患病率以及临床、分子和预后相关性。

在机构审查委员会 (IRB) 批准 (#14560) 后，所有连续的 PMF 和 SMF 诊断患者，分别根据世界卫生组织 (WHO) 和国际 MPN 研究和治疗工作组 (IWG-MRT) 标准，并且已知SF3B1在我们的数据库中可用的突变状态被包括在研究中。根据先前公布的方法，在诊断或首次转诊时对驱动基因和非驱动基因进行了分子分析。使用非参数（Mann-Whitney 或 Kruskal-Wallis）检验比较连续变量的分布，而将名义变量与卡方检验进行比较。使用 Kaplan–Meier 方法进行事件发生时间分析（总体 [OS] 和无白血病生存 [LFS]），包括死亡（OS）、AML 进展（LFS）和同种异体造血干细胞移植（HSCT） ; 对于 OS 和 LFS）用作审查员。使用 SPSS 软件 28 版 (IBM-Corp)、SAS Institute (Cary, NC) 的 JMP Pro 15.1.0 软件和 Statistical Package R 4.1.1 版进行统计分析。

在本研究中，我们回顾性分析了 520 名 MF 患者，325 名 (62.5%) PMF 包括 156 名 (48%) 前纤维化和 169 名 (52%) 显性 PMF，以及 195 名 (37.5%) SMF，包括 85 名 (44%) PPV-MF 和 110 (56%) PET-MF。

考虑 PMF 患者，中位年龄为 59 岁，62% 为男性；中位随访时间为 4.5 年（范围，0.06–36）（表 S1）。驱动突变分布为 58% JAK2、26 % CALR、6% MPL、12% 三阴性 (TN) 和 2% 双驱动突变。记录在案的死亡人数和白血病转化人数分别为 121 人 (37%) 和 37 人 (12%)。最常见的非驱动突变基因是ASXL1 (30%)、TET2 (20%) 和SRSF2 (10%)。在 22 名患者（7%；14 名显性 PMF 和 8 名纤维化前 MF）中检测到 SF3B1 突变，如下所示：11 名 (50%) K666T/N/M，7 名 (32%) K700E，2 名 (9%) H662Q / Y，1 (4%) R625L 和 1 (4%) N626S。19/22 SF3B1突变患者中，12 例（54%）为JAK2 V617F，6 例（27%）CALR和 1 例（4%）MPL突变。考虑到SF3B1突变患者，最常见的非驱动共突变基因包括ASXL1（n  = 4；18%）和TET2（n  = 3；14%），其次是DNMT3A、EZH2和TP53（n  = 2；9%每个）。SF3B1突变的 PMF 患者大多是男性（86% 对 60%；p  = .01），年龄较大（65 对 62 岁；p  = .004）并且血小板计数显着更高（561 × 10 ⁹/L 对比 402 × 10 ⁹ /L；p  = .06)，并且 Hb 水平低于 野生型对应物（11.3 g/dL 与 12.3 g/dL； p = .09）。

在分析中总共包括 195 名 SMF 患者中，中位年龄为 63 岁，55% 为男性；中位随访时间为 3.9 年（范围，0.07-28）。有 80 例 (41%) 记录在案的死亡和 27 例 (14%) 白血病转化（表 S2）。驱动突变分布如下：70% JAK2 V617F、23% CALR、8% MPL和 2% TN；13 名 (7%) 患者发生双重驱动突变。最常见的非驱动突变基因是ASXL1 (30%)、TET2 (18%) 和ZRSR2 (8%)。在 10 名患者 (5%) 中检测到 SF3B1 突变，包括 4 名 (40%) K666T/N/M、4 名 (40%) K700E、1 名 (10%) G751V、1 名 (10%) G742D 。所有SF3B1突变患者也携带驱动突变；特别是，5 个（50%）是JAK2 V617F、4 个（40%）CALR和 1 个（10%）MPL突变。考虑到SF3B1突变患者，最常见的非驱动共突变基因的模式包括TET2（n  = 4；40%），其次是CBL、RUNX1和TP53（n  = 2；各 20%）。SF3B1突变的 SMF 患者年龄较大（66 岁与 62 岁；p  = .09）并且出现脾肿大的频率较低 ( p  = .05)。此外，相关分析表明，SF3B1与SF3B1野生型对应物 相比，突变主要与CBL ( p  = .004)、RUNX1 ( p  = .02) 和KRAS ( p = .03) 突变相关。

然后我们研究了SF3B1突变对 OS 和 LFS的影响。考虑到 PMF 患者，SF3B1突变不会影响 OS（HR 1.1；95% CI 0.6–2；p  = .8，图 1A）和 LFS（HR 0.7；95% CI 0.2–2.8；p  = .6，未显示). 相反，考虑到 SMF 病例，与野生型对应物相比，OS 受到SF3B1突变的负面影响（HR 3.2；95% CI 1.5–7； p  = .002，图 1B），而 LFS 结果相似（HR 1.1；95% CI 0.1 –8.4；p  = .9；未显示）。值得注意的是，在两个 ET 后 SMF 病例中，SF3B1突变已经存在于疾病的 ET 阶段。

为了更好地解读SF3B1突变对 SMF 患者 OS 的个体贡献，我们进行了包括临床和分子变量的 Cox 回归分析（表 S3）。除SF3B1外，单变量分析确定年龄≥ 65 岁（HR 5；95% CI 3.1–8.3；p  = .003），白细胞 >25 × 10 ⁹ /L（HR 2，95% CI 1.1–3.7；0.03） , Hb < 11 g/dL (HR 3.1, 95% CI 2–5.1; p  < .0001), 血小板计数 <150 × 10 ⁹ /L (HR 2.7 95% CI 1.6–4.6; p  = .0003), 循环母细胞 ≥3% (HR 4, 95% CI 2.1–7.5; p < .0001) 和CBL 突变(HR 4, 95% CI 1.7–9.3; p  = .001)，SRSF2（HR 8.1，95% CI 2.5–26.7；p  = .0006）、U2AF1（HR 2.7，95% CI 1.2–6；p  = .02）和TP53（HR 2.9，95% CI 1.4–5.8；p  = .003) 作为降低总生存期的危险因素。在上述所有风险因素的背景下，多变量分析证实了SF3B1突变（HR 2.8，95% CI 1.2–6.3；0.02）以及年龄较大 ( p  < .0001)、血红蛋白 <11 g/ dL ( p  = .02)，循环母细胞 ≥3% ( p  = .0003) 和CBL突变 ( p =.01）。病例数少阻止了我们分析不同SF3B1突变的个体预后贡献。

SF3B1以及SRSF2和U2AF1是髓系肿瘤中突变最频繁的三个剪接体基因；在 PMF 中，SRSF2和U2AF1（主要是 Q157）突变而非SF3B1与较差的存活率相关，并被纳入突变增强型国际预后系统 (MIPSS)-70/v2.0 评分中。^{9, 10}在本研究中，我们分别分析了PMF 和 SMF 中SF3B1突变的临床、分子和预后相关性。我们确认SF3B1很少发生PMF (7%) 和 SMF (5%) 的突变；后者的 80% 属于后 ET 子集。所有报告的剪接体基因突变都是相互排斥的。重要的是，在多变量分析中，还包括构成 MYSEC-PM ¹¹预后评分的临床和分子变量，SF3B1突变与 SMF 中 OS 降低独立相关（表 S3）。最近，在 ET（SF3B1、SRSF2和U2AF1）和 PV（SRSF2）患者中发生的剪接体基因突变被包括在 MIPSS PV-ET ¹²预后评分中，因为它们对 OS 产生不利影响。

当前研究的主要局限性在于其回顾性；此外，尽管包括了大量的 MF 患者，但SF3B1突变的 PMF 和 SMF 病例的数量仍然有限，使得本文报告的发现主要是探索性的，需要在更大的独立队列中进行验证。然而，随着最近的数据强调ASXL1突变在 PMF 与 SMF 中的不同预后作用，¹³这些发现强化了 PMF 和 SMF 代表两个截然不同的生物实体的概念。在这方面，针对 SMF 患者的综合预后模型的开发仍然是一个未满足的临床需求。