分析骨髓系腫瘤激酶體基因表現以探索新穎治療

Abstract

背景 人類的激酶體(kinome)涵蓋超過500種蛋白激酶(kinase)，於調控細胞增生、分化、凋亡及代謝等基本細胞過程中扮演關鍵角色。雖然激酶之不正常調控已被證實與多種固態腫瘤相關，其在骨髓系之惡性腫瘤，特別是骨髓化生不良症候群(MDS)與急性骨髓性白血病(AML)中的表現量研究及其臨床意義尚未被充分釐清。本研究透過系統性的藥物-轉錄體(pharmaco-transcriptomic)分析，評估MDS與AML中激酶體表現之圖譜，希望可藉此優化病患之預後分級，並鑑別高風險分子病患族群中的新穎治療把點。

實驗方法 本研究分析來自台大醫院(National Taiwan University Hospital, NTUH)之MDS與AML患者骨髓樣本的轉錄體資料，以研究與整體存活相關之激酶表現。並根據那些與不良預後顯著相關之激酶表現，建構MDS與AML的激酶預後模型。我們運用轉錄體富集(gene set enrichment)分析用於探索相關分子機制；藥物敏感性資料則取得Genomics of Drug Sensitivity in Cancer (GDSC)與Beat AML資料庫。針對激酶標靶治療的細胞實驗驗證，則以AML細胞株進行JNK-IN-8小分子抑制劑的相關藥理實驗。

結果 針對MDS，我們建立了一個新穎的激酶預後模型(KS-MDS)，其整合了七個與不良預後相關的激酶(PTK7、KIT、MAST4、NTRK1、PAK6、CAMK1D 與 PRKCZ)。在AML方面，亦使用相同策略建立一個AML之預後模型(KS-AML)，其包含了十三個預後相關激酶(CDK18、GSK3A、MAP3K6、MAP4K1、MAPK8、NEK3、PDK2、ROCK1、SCYL3、SIK2、SIK3、TRPM7 與 ULK3)。在MDS與AML中，較高的激酶風險分數顯著與高危險度之臨床與分子特徵有關，多變數分析亦確認其為整體存活之獨立不良預後因子。在MDS中，轉錄體分析顯示，KS-MDS高危之病患具有顯著之造血與血癌幹細胞之基因表現特徵。藥物敏感性分析發現VEGFR抑制劑axitinib與PI3K抑制劑taselisib對於KS-MDS表現較高之造血細胞株具有治療效果。在AML中，轉錄體分析顯示KS-AML高危之病患富含KMT2A重組及NPM1突變相關的轉錄特徵。功能性實驗證實，MAPK8 (JNK)的抑制劑JNK-IN-8可抑制AML細胞增生、誘導細胞凋亡，並降低c-JUN、c-MYC與cGAS-STING路徑的活性。此外，JNK-IN-8與BCL2抑制劑venetoclax合併使用，可達到糖解與氧化磷酸化反應的雙重抑制，此為一新穎之對抗AML代謝可塑性的治療策略。

結論 本研究建立了針對MDS與AML之全新激酶體預後模型，能夠補足傳統臨床風險分類，進一步提升預後分級的精確性，並協助找尋具潛力的新穎治療標靶藥物。我們的研究發現激酶體異常表現為高風險骨髓系腫瘤的顯著特徵，並突顯以激酶為標靶的治療策略在抑制癌細胞增生與代謝感受性方面的臨床應用潛力，未來可望用於發展以個人化生物標記為導向的精準治療策略。

Introduction The human kinome, comprising over 500 protein kinases, plays a pivotal role in regulating essential cellular processes, including proliferation, differentiation, apoptosis, and metabolism. While kinase dysregulation has been implicated in numerous solid cancers, its comprehensive characterization and clinical relevance in myeloid malignancies, notably myelodysplastic syndromes (MDS) and acute myeloid leukemia (AML), remain incompletely defined. In this study, we performed a systematic pharmaco-transcriptomic analysis to evaluate kinase expression signatures in MDS and AML, aiming to refine prognostic stratification and identify actionable therapeutic vulnerabilities in high-risk molecular subgroups.

Materials and Methods We analyzed the transcriptomic data from bone marrow (BM) samples from the MDS and AML patients at the National Taiwan University Hospital (NTUH) to identify kinases whose expression correlated with overall survival (OS). kinase-based prognostic models for MDS and AML were constructed by integrating the expression of the most significant adverse-risk kinases. Gene set enrichment analyses were performed to explore associated molecular programs. Drug sensitivity correlations were evaluated using the Genomics of Drug Sensitivity in Cancer (GDSC) and Beat AML databases. Functional validation of kinase-targeted therapy was conducted using the JNK inhibitor JNK-IN-8 in AML cell lines.

Results A kinase-based prognostic model for MDS, termed KS-MDS, was constructed by integrating the expression of seven poor prognostic kinases (PTK7, KIT, MAST4, NTRK1, PAK6, CAMK1D and PRKCZ). A similar approach was applied to the NTUH AML dataset to derive the kinase-based prognostic model for AML, termed KS-AML, incorporating thirteen prognostically relevant kinases (CDK18, GSK3A, MAP3K6, MAP4K1, MAPK8, NEK3, PDK2, ROCK1, SCYL3, SIK2, SIK3, TRPM7 and ULK3). In both MDS and AML, higher kinase-based risk scores were significantly associated with adverse clinical and molecular feature, and multivariate analysis confirmed kinase-based risk scores as independent predictors of poor survival. Transcriptomic profiling revealed that KS-MDS-high patients exhibited enriched hematopoietic and leukemic stem cell (HSC/LSC) gene expression signatures. Drug sensitivity screening identified axitinib, a VEGFR inhibitor, and taselisib, a PI3K inhibitor, as compounds with selective activity against KS-MDS-high hematopoietic cell lines. In AML, transcriptomic analysis revealed significant enrichment of transcriptional signatures associated with KMT2A rearranged and NPM1-mutated AML in the KS-AML-high subgroup. Further functional studies revealed that the MAPK8 (JNK) inhibitor JNK-IN-8 showed dose-dependent inhibition of AML proliferation, induction of apoptosis, by downregulating c-JUN, c-MYC, and the cGAS-STING innate immune signaling pathway. Combined targeting of JNK and mitochondrial metabolism with venetoclax achieved dual inhibition of glycolysis and OXPHOS, indicating a strategy to overcome metabolic adaptability in AML.

Conclusion This integrative kinome-focused analysis identifies novel kinase-based transcriptomic risk scores for MDS and AML, that could improve prognostic stratification beyond current clinical evaluation. Both scores reflect underlying stemness and adverse signaling programs, and guide the identification of candidate therapeutic compounds, including axitinib, taselisib, and JNK-IN-8. These findings establish kinase dysregulation as a central feature of high-risk myeloid neoplasms and highlight the promise of kinase-directed therapies to target both proliferative and metabolic vulnerabilities, and facilitate the development of biomarker-guided therapeutic strategies for these myeloid malignancies.