The survival benefit derived from sorafenib treatment for patients with hepatocellular carcinoma (HCC) is modest due to acquired resistance. Targeting cancer stem cells (CSC) is a possible way to reverse drug resistance, however, inhibitors that specifically target liver CSCs are limited. In this study, we established two sorafenib-resistant, patient-derived tumor xenografts (PDX) that mimicked development of acquired resistance to sorafenib in patients with HCC. RNA-sequencing analysis of sorafenib-resistant PDXs and their corresponding mock controls identified EPH receptor B2 (EPHB2) as the most significantly upregulated kinase. EPHB2 expression increased stepwise from normal liver tissue to fibrotic liver tissue to HCC tissue and correlated with poor prognosis. Endogenous EPHB2 knockout showed attenuation of tumor development in mice. EPHB2 regulated the traits of liver CSCs; similarly, sorted EPHB2High HCC cells were endowed with enhanced CSC properties when compared with their EPHB2-Low counterparts. Mechanistically, EPHB2 regulated cancer stemness and drug resistance by driving the SRC/AKT/GSK3β/β-catenin signaling cascade, and EPHB2 expression was regulated by TCF1 via promoter activation, forming a positive Wnt/β-catenin feedback loop. Intravenous administration of rAAV-8-shEPHB2 suppressed HCC tumor growth and significantly sensitized HCC cells to sorafenib in an NRAS/AKT-driven HCC immunocompetent mouse model. Targeting a positive feedback loop involving the EPHB2/β-catenin axis may be a possible therapeutic strategy to combat acquired drug resistance in HCC.Significance: This study identifies a EPHB2/β-catenin/TCF1 positive feedback loop that augments cancer stemness and sorafenib resistance in HCC, revealing a targetable axis to combat acquired drug resistance in HCC.

SUMMARY Transcription, the first step of gene expression, is carried out by the enzyme RNA polymerase (RNAP) and is regulated through interaction with a series of protein transcription factors. RNAP and its associated transcription factors are highly conserved across the bacterial domain and represent excellent targets for broad-spectrum antibacterial agent discovery. Despite the numerous antibiotics on the market, there are only two series currently approved that target transcription. The determination of the three-dimensional structures of RNAP and transcription complexes at high resolution over the last 15 years has led to renewed interest in targeting this essential process for antibiotic development by utilizing rational structure-based approaches. In this review, we describe the inhibition of the bacterial transcription process with respect to structural studies of RNAP, highlight recent progress toward the discovery of novel transcription inhibitors, and suggest additional potential antibacterial targets for rational drug design.