IntruductionEsophageal carcinoma is the sixth leading cause of cancer-related mortality worldwide and consists primarily of two histological subtypes: esophageal squamous cell carcinoma (ESCC) and esophageal adenocarcinoma (EAC). ESCC accounts for approximately 90% of global cases and is associated with poor clinical outcomes, with 5-year survival rates ranging from 15% to 25%. This dismal prognosis is largely driven by high rates of disease recurrence following treatment [1]. Radiotherapy remains a central component of ESCC management, employed either definitively or as part of neoadjuvant therapy. However, its efficacy is substantially limited by intrinsic radioresistance. In the landmark RTOG 85-01 trial, definitive concurrent chemoradiotherapy (CCRT) achieved a 5-year overall survival rate of 26% and a median overall survival of approximately 14–17 months [2, 3]. Subsequent studies, including INT 0123, reported 2-year locoregional failure rates of 52% and 56% in the standard-dose and high-dose groups, respectively [4]. Overall, these data underscore the limited long-term efficacy of definitive CCRT, with local recurrence rates remaining high at approximately 40–60% [5, 6]. Ionizing radiation exerts its cytotoxic effects primarily through the induction of DNA double-strand breaks (DSBs), and tumor radioresistance is often attributed to enhanced DSB repair capacity, which contributes to treatment failure and poor survival outcomes [7]. Therefore, delineating the molecular regulators that govern DSB repair and mediate intrinsic radioresistance is critical to overcoming this therapeutic barrier in ESCC.While the two main DNA damage response (DDR) pathways homologous recombination (HR) and non-homologous end joining (NHEJ), are well-established mediators of radioresistance [8,9,10], the upstream regulatory networks remain incompletely understood. Small nucleolar RNAs (snoRNAs), classically known for guiding ribosomal RNA modifications, are increasingly recognized for their non-canonical roles in tumorigenesis and therapy resistance [11, 12]. Beyond their classical roles, certain snoRNAs exhibit oncogenic properties by directly interacting with proteins or modulating non-ribosomal pathways. Notable examples include SNORD6, which facilitates cervical cancer progression by acting as a molecular scaffold to accelerate E6-mediated p53 degradation [13]; SNORD50A/B, which suppresses KRAS activation through direct molecular binding [14], and our previous study demonstrating that SNORD12B promotes AKT-mTOR signaling hyperactivation in ESCC by competitively sequestering PP1α [15]. Emerging evidence further implicates specific snoRNAs in therapeutic resistance. Emerging evidence further implicates specific snoRNAs in therapeutic resistance, including scaRNA2, SNORA74A, SNORA23, and SNORA28 [16,17,18,19], which affect DNA repair or survival signaling. However, the potential roles and molecular mechanisms of snoRNAs in regulating the core DSB repair pathways of HR and NHEJ, and their contribution to intrinsic radioresistance in ESCC, remain largely unexplored, highlighting the critical need to identify clinically relevant snoRNA-DDR axes as therapeutic targets to overcome radioresistance in ESCC.To address this gap, we identify SNORA23 as a pivotal driver of radioresistance and poor prognosis in ESCC. SNORA23 is significantly upregulated in radioresistant cell lines and tumor specimens, where its expression correlates strongly with advanced tumor stage, failure to achieve complete response (CR) following CCRT and reduced overall survival. Mechanistically, we uncover a non-canonical regulatory axis in which SNORA23 directly binds the ARID domain of the histone demethylase KDM5C [20,21,22] through a conserved G11–Ser169 hydrogen bond. This interaction impairs KDM5C’s chromatin association and relieves transcriptional repression of SFPQ, a DNA repair scaffolding gene [23], thereby enhancing SFPQ expression. Functionally, elevated SFPQ serves as a central effector that physically bridges the HR and NHEJ repair machinery. This bridging facilitates enhanced chromatin recruitment of RAD51 and Ku80 to DNA double-strand break sites following irradiation. The clinical relevance of this SNORA23–KDM5C–SFPQ axis is supported by the strong correlation between SNORA23 and SFPQ expression and the independent association of high SFPQ levels with therapeutic resistance and poor clinical outcomes. Importantly, we translate this mechanistic insight into therapeutic potential by demonstrating that targeted suppression of SNORA23 using antisense oligonucleotides (ASOs) effectively disrupts the SNORA23-KDM5C-SFPQ axis, impairs DSB repair, reactivates radiation-induced apoptosis, and synergizes with radiotherapy to inhibit ESCC xenograft growth, without detectable systemic toxicity.Collectively, our findings reveal a mechanism by which SNORA23 modulates KDM5C activity to influence DNA damage response. SNORA23 governs intrinsic radioresistance by acting as a chromatin occupancy modulator of the epigenetic regulator KDM5C. This interaction derepresses SFPQ, a DNA repair orchestrator that integrates HR and NHEJ execution. This study establishes the SNORA23-KDM5C-SFPQ axis as a critical determinant of cell fate under radiation-induced cytotoxic stress, highlighting SFPQ’s unique role in coordinating dual DNA repair programs and providing a compelling rationale for targeting SNORA23 to overcome radioresistance and improve therapeutic outcomes in ESCC.ResultsSNORA23 drives radiotherapy resistance and predicts poor prognosis in ESCCWe developed radioresistant ESCC subclones (KYSE30R, KYSE450R) through fractionated irradiation to investigate whether snoRNAs contribute to ESCC intrinsic radioresistance. These subclones exhibited pronounced radioresistance, as evidenced by increased clonogenic survival with reduced sensitization enhancement ratios (SERs) (Fig. 1A, B), attenuated irradiation-induced apoptosis and decreased G2/M arrest following irradiation (Fig. S1A, B), and diminished PARP/caspase-3 cleavage (Fig. S1C). Multiplex qPCR profiling revealed broad alterations in snoRNA expression between radioresistant and parental cells, identifying 66 differentially expressed snoRNAs, including 32 upregulated and 34 downregulated transcripts (Fig. 1C). Among these, SNORA23 emerged as the most significantly upregulated snoRNA, as confirmed by independent qPCR validation (Fig. S1D, E). To evaluate the clinical relevance of SNORA23 in radiotherapy response, we quantified its expression in tumor specimens from 104 patients with locally advanced ESCC undergoing concurrent chemoradiotherapy (CCRT). SNORA23 levels were significantly elevated in advanced T-stage and node-positive tumors (Fig. 1D, E; Table 1). ESCC patients undergoing CCRT failing to achieve complete response (CR) exhibited higher SNORA23 expression compared to CR patients (Fig. 1F), with an inverse correlation between SNORA23 and CR rate (Fig. 1G), indicating its role in intrinsic radioresistance. High SNORA23 expression was also associated with inferior overall survival (Fig. 1H). Time-dependent ROC analysis supported SNORA23’s prognostic value for predicting 3-year survival (Fig. 1I). Critically, both univariate and multivariate regression established SNORA23 as an independent adverse prognostic factor (Fig. 1J, K; Table S1). Collectively, these results nominate SNORA23 as a mechanism-linked biomarker of intrinsic radiotherapy resistance and poor prognosis in advanced ESCC, where its upregulation predicts CCRT failure and mortality risk.Fig. 1: High SNORA23 expression correlates with radioresistance and poor prognosis in ESCC patients.Full size imageA Representative colony formation assay of parental and radioresistant post-irradiation at different doses. B Survival fraction curves of ESCC parental and radioresistant cell lines after exposure to increasing doses of irradiation and multitarget single-hit model analysis determining parameters (N, D₀, Dq) and SER values in ESCC parental and radioresistant cell lines. C Heatmap displaying differentially expressed snoRNAs ( | FC | >2) identified by multiplex qPCR profiling analysis in radioresistant versus parental cell lines. SNORA23 expression levels in ESCC patients with advanced T-stage (D), positive N-stage (E), and non-CR (F) after radiotherapy. G CR rate post-radiotherapy stratified by SNORA23 expression. H Kaplan-Meier OS curves in 104 CCRT ESCC patients stratified by primary tumor SNORA23 expression levels. I Time-dependent ROC curve assessing SNORA23’s prognostic accuracy for 3-year OS (AUC = 0.842). Forest plots from univariate (J) and multivariate (K) Cox regression analyses. Cell-based experiments were performed in biological triplicates. Data are shown as mean ± SD. Analyses: Mann-Whitney U test (D–F), Chi-square test (G), Log-rank test (H), Cox proportional hazards regression model (J, K). Statistical significance, *P