Integration between MCL1 gene co-expression module and the Revised International Staging System enables precise prognostication and prediction of response to proteasome inhibitor-based therapy in individual multiple myeloma

We recently identified a gene module of 87 genes co-expressed with MCL1 (MCL1-M), a critical regulator of plasma cell survival. MCL1-M captures both MM cell-intrinsically acting signals and the signals regulating the interaction between MM cells with bone marrow microenvironment. MM can be clustered into MCL1-M high and MCL1-M low subtypes. While the MCL1-M high MMs are enriched in a preplasmablast signature, the MCL1-M low MMs are enriched in B cell-specific genes. In multiple independent datasets, MCL1-M high MMs exhibited poorer prognosis compared to MCL1-M low MMs. Re-analysis of the phase III HOVON-65/GMMG-HD4 showed that only MCL1-M MMs, but not MCL1-M low MMs, benefited from bortezomib-based treatment. To translate the MCL1-M clustering scheme into a platform for individual diagnosis, we refined the classifier genes and developed a support vector machine-based algorithm. Individual MMs with transcriptome assessed at the RNA-seq or U133 plus 2.0 array platform can be robustly assigned as the MCL1-M high or low subtype with high confidence. Analyses of the MM samples in the HOVON-65/GMMG-HD4 trial and APEX trial reinforce that only MCL1-M high MMs benefit from bortezomib-based treatment with a hazard ratio of 0.58 (P = 0.010) and 0.47 (P = 0.009), respectively. Thus, MCL1-M based subtyping assigns MMs into prognostic and predictive molecular subtypes driven by subtype-specific pathogenic pathways. We also generated our own data set based on 72 newly diagnosed MM samples  from Chaoyang hospital in Beijing. All participants signed the informed consent form, and the study was approved by the institutional ethical review board of the Chao-Yang Hospital, Capital Medical University (Beijing, China). Patients were treated between August 2015 and September 2019, the longest follow-up period was 67 months. Bone marrow CD138+ cells were purified for the preparation of total mRNA for the transcriptome profiling using Affymetrix PrimeView array according to standard protocols.

本研究近期鉴定出一个由87个与MCL1（MCL1-M）共表达的基因构成的基因模块，MCL1是浆细胞存活的关键调控因子。MCL1-M模块既涵盖多发性骨髓瘤（Multiple Myeloma，MM）细胞内在的作用信号，也包含调控多发性骨髓瘤细胞与骨髓微环境相互作用的信号。
据此可将多发性骨髓瘤划分为MCL1-M高表达亚型与MCL1-M低表达亚型。其中MCL1-M高表达亚型富集于前浆母细胞特征基因谱，而MCL1-M低表达亚型则富集于B细胞特异性基因。
在多组独立数据集分析中，MCL1-M高表达的多发性骨髓瘤患者预后较MCL1-M低表达者更差。
对III期临床试验HOVON-65/GMMG-HD4的再分析结果显示，仅MCL1-M高表达的多发性骨髓瘤患者可从基于硼替佐米的治疗中获益，而MCL1-M低表达患者则无此获益。
为将MCL1-M分型方案转化为个体化诊断平台，本研究优化了分类器基因集，并开发了基于支持向量机的分类算法。
通过RNA测序（RNA-seq）或U133 Plus 2.0基因芯片平台完成转录组检测的多发性骨髓瘤样本，可被高置信度地准确归类为MCL1-M高表达或低表达亚型。
对HOVON-65/GMMG-HD4试验与APEX试验中的多发性骨髓瘤样本进行的分析进一步证实，仅MCL1-M高表达的多发性骨髓瘤患者可从基于硼替佐米的治疗中获益，二者的风险比分别为0.58（P=0.010）与0.47（P=0.009）。
综上，基于MCL1-M的分型方案可将多发性骨髓瘤划分为由亚型特异性致病通路驱动的预后型与预测型分子亚型。
本研究同时构建了自建数据集，纳入72例来自北京朝阳医院的初诊多发性骨髓瘤样本。
所有受试者均签署了知情同意书，本研究经首都医科大学附属北京朝阳医院伦理审查委员会批准。
患者的治疗时段为2015年8月至2019年9月，最长随访时长达67个月。
按照标准实验流程，研究人员纯化骨髓CD138+细胞以提取总mRNA，并采用Affymetrix PrimeView基因芯片开展转录组谱分析。