藉由系統性分析配對的原發性和復發性 IDH 野生型膠質母細胞瘤間的基因體和轉錄體差異尋找癌症復發相關的生物標記

Abstract

神經膠質母細胞瘤 (Glioblastoma multiforme，GBM) 是成人中最常見且侵襲性最強的惡性腦部腫瘤，對現有的治療方式仍具高度抗性，且在初次治療後常迅速復發。儘管針對GBM復發所涉及的基因組與轉錄組變化已進行了廣泛的研究，成對的原發及復發 GBM 之間的演化動態仍未完全理解。其中一個主要挑戰是找出與復發時間 (Relapse time/Time to relapse，RT/TTR) 相關的分子標誌，並建立一個能準確預測原發GBM患者RT的穩定預後模型。 在本研究中，我們整合了多個患者匹配的GBM資料集的RNA測序和基因組數據 (Genomic and transcriptomic datasets)，特別關注於異檸檬酸脫氫酶野生型 (IDH wild-type) 的腫瘤。我們探討了RT與原發及復發 GBM之間分子異質性之間的關聯，著重於基因表現模式、腫瘤突變負荷 (Tumor mutational burden，TMB) 和腫瘤微環境。 我們的研究結果顯示，RT與原發及復發 GBM之間轉錄組和基因組的差異程度呈正相關。此外，相較於RT較長的患者，RT較短的患者顯示出非間質型向間質型轉變(Non-mesenchymal-to-mesenchymal transitions)的比例較高，以及更高的間質型亞型表現。我們還觀察到原發及復發 GBM在基因表現模式和TMB上的高度相關性，且這些匹配的原發及復發 GBM基因表現一致性與RT呈負相關。 基於這些觀察結果，我們鑑定出55個與RT相關的基因，並透過單變量和多變量Cox回歸分析建立了一個包含7個關鍵基因（ZSCAN10、SIGLEC14、GHRHR、TBX15、TAS2R1、CDKL1、CD101）的預後模型。該模型產生的風險分數在訓練集及兩個獨立驗證集中與RT呈顯著負相關。此外，該模型有效地將IDH wild-type GBM患者分成兩組，並顯示出顯著不同的無進展生存率 (Progression-free survival，PFS) 結果，且在所有數據集中對1年、2年及3年的PFS率預測表現出色。我們的研究結果為GBM復發時的分子機制提供了新的見解，並提出了潛在的治療靶點以供未來研發策略的參考。

Glioblastoma multiforme (GBM) represents the most pervasive and formidable malignant brain tumor in adults, characterized by its resistance to current therapeutic modalities and a propensity for rapid recurrence post-initial treatment. Despite substantial efforts to elucidate the genomic and transcriptomic landscapes underpinning GBM recurrence, the evolutionary dynamics linking initial/primary and recurrent (I-R/P-R) GBMs remain inadequately defined. A critical challenge lies in discerning molecular determinants predictive of relapse time/time to relapse (RT/TTR) and formulating a reliable prognostic framework to estimate RT in patients with initial/primary GBM (iGBM/pGBM). In this investigation, we leveraged RNA sequencing and genomic datasets from longitudinally collected, patient-matched GBM cohorts, with a specific focus on isocitrate dehydrogenase wild-type (IDH-wildtype) tumors. Our analysis interrogated the interplay between RT and molecular heterogeneity within matched I-R GBMs, emphasizing gene activity dynamics, tumor mutational burden (TMB), and the composition of the tumor microenvironment. Notably, our findings demonstrated a direct correlation between RT and the extent of transcriptomic and genomic divergence observed between initial and recurrent GBMs. Patients with shorter RTs exhibited elevated frequencies of non-mesenchymal-to-mesenchymal transitions and a predominance of mesenchymal subtypes, as opposed to those with prolonged RTs. Moreover, we identified a strong concordance in gene activity patterns and TMB between matched I-R GBMs, which inversely correlated with RT. From these analyses, we pinpointed 55 RT-associated genes and built a prognostic model incorporating seven pivotal genes (ZSCAN10, SIGLEC14, GHRHR, TBX15, TAS2R1, CDKL1, and CD101) through univariate and multivariate Cox regression methodologies. The model's risk scores demonstrated a significant inverse correlation with RT in the training cohort, validated across two independent datasets. Furthermore, the model stratified IDH-wildtype GBM patients into two distinct prognostic subgroups with significantly different progression-free survival (PFS) trajectories, while achieving robust predictive accuracy for one-, two-, and three-year PFS across all cohorts. These findings illuminate the molecular underpinnings of GBM progression and recurrence, offering a deeper understanding of its evolutionary biology. The study also highlights potential molecular targets for therapeutic intervention, paving the way for more personalized and effective strategies in GBM management.