Inter- and intra-tumor heterogeneity is a major hurdle in primary liver cancer (PLC) precision therapy. Here, we establish a PLC biobank, consisting of 399 tumor organoids derived from 144 patients, which recapitulates histopathology and genomic landscape of parental tumors, and is reliable for drug sensitivity screening, as evidenced by both in vivo models and patient response. Integrative analysis dissects PLC heterogeneity, regarding genomic/transcriptomic characteristics and sensitivity to seven clinically relevant drugs, as well as clinical associations. Pharmacogenomic analysis identifies and validates multi-gene expression signatures predicting drug response for better patient stratification. Furthermore, we reveal c-Jun as a major mediator of lenvatinib resistance through JNK and β-catenin signaling. A compound (PKUF-01) comprising moieties of lenvatinib and veratramine (c-Jun inhibitor) is synthesized and screened, exhibiting a marked synergistic effect. Together, our study characterizes the landscape of PLC heterogeneity, develops predictive biomarker panels, and identifies a lenvatinib-resistant mechanism for combination therapy.

肿瘤间和肿瘤内的异质性是原发性肝癌（PLC）精准治疗的主要障碍。在这里，我们建立了一个 PLC 生物库，由来自 144 名患者的 399 个肿瘤类器官组成，它概括了亲本肿瘤的组织病理学和基因组图谱，并且对于药物敏感性筛选是可靠的，正如体内模型和患者反应所证明的那样。综合分析剖析了 PLC 异质性，涉及基因组/转录组特征和对七种临床相关药物的敏感性以及临床关联。药物基因组学分析可识别并验证多基因表达特征，预测药物反应，从而更好地对患者进行分层。此外，我们发现 c-Jun 通过 JNK 和 β-catenin 信号传导是乐伐替尼耐药的主要介质。合成并筛选了包含lenvatinib和veratramine（c-Jun抑制剂）部分的化合物（PKUF-01），表现出显着的协同作用。总之，我们的研究描述了 PLC 异质性的特征，开发了预测性生物标志物组，并确定了联合治疗的乐伐替尼耐药机制。