Wei Dai

Nasopharyngeal carcinoma (NPC) is a geographically distinct malignancy closely associated with Epstein–Barr virus (EBV) infection and characterized by extensive epigenetic reprogramming within a highly immunosuppressive tumor microenvironment (TME). High-mobility group box 1 (HMGB1), a multifunctional chromatin-binding protein that can also act as an extracellular damage-associated molecular pattern (DAMP), has emerged as an important regulator of genome organization, transcriptional control, cellular stress responses, and immune signaling. Increased HMGB1 expression has been reported in NPC and is associated with adverse clinicopathological features and poor patient outcomes. Emerging evidence suggests that the diverse biological functions of HMGB1 are influenced by post-translational modifications (PTMs), which affect its subcellular localization, molecular interactions, and extracellular signaling functions. Through these regulatory mechanisms, HMGB1 may transition from a nuclear chromatin-associated protein to an extracellular mediator of immune and inflammatory responses. PTMs including acetylation, phosphorylation, glycosylation, oxidation, methylation, and lactylation have been implicated in regulating HMGB1 trafficking and function, although the specific roles of many of these modifications in NPC remain incompletely characterized. In this review, we summarize current evidence regarding HMGB1 PTMs and discuss their potential implications for EBV-associated NPC, with emphasis on nuclear regulation, immune crosstalk, and therapeutic response. We explicitly distinguish findings directly demonstrated in NPC from mechanistic insights derived from other malignancies and related disease models, and identify areas where proposed mechanisms remain hypothesis-generating rather than experimentally validated in NPC. By integrating current evidence with emerging mechanistic concepts, we highlight key knowledge gaps, unresolved questions, and priorities for future research. A better understanding of PTM-dependent HMGB1 regulation may facilitate the development of novel biomarker and therapeutic strategies for EBV-associated NPC.

Responses to macrophage-directed therapy can be transient because tumors preserve myeloid support through complementary persistence and replenishment. In esophageal squamous cell carcinoma (ESCC), CSF1R inhibition reduced established tumor-associated macrophages but was followed by expansion of Ly6C/CCR2-positive monocytic and Ly6G-positive granulocytic populations. Low-dose decitabine preferentially restricted recruited populations while sparing a LYVE1-associated macrophage state, exposing reciprocal pharmacologic blind spots. Combined treatment suppressed both arms and produced sustained control across patient-derived organoid xenograft, orthotopic, and immunocompetent models. Neutrophil depletion reproduced initial regression but not sustained control, indicating that the recruited escape arm extended beyond Ly6G-positive granulocytes. Single-cell profiling mapped these vulnerabilities onto a treatment-resolved myeloid architecture comprising a C1qa-positive TAM continuum, a C1qa-negative Ccr2/Ly6c2-high inflammatory monocytic-like compartment, and a LYVE1/MRC1-positive tissue-supportive macrophage state. Human ESCC contained corresponding macrophage programs and an adverse-outcome-associated LYVE1-rich niche. These findings identify state-aware coverage of complementary myeloid vulnerabilities as a strategy to overcome escape from macrophage-directed therapy. Statement of Significance: Macrophage-directed therapy can fail when established tumor-associated macrophage states and recruited myeloid populations remain incompletely covered by a single intervention. In ESCC, jointly suppressing complementary persistence and replenishment compartments converted heterogeneous, incomplete responses into sustained control across models, defining a state-aware myeloid coverage principle for biomarker-guided combination therapy.

RNAGAN (version 2.0, https://github.com/ZhaozhengHou-HKU/RNAGAN-2.0.git) is a published foundation model that analyzes single-cell and bulk-level RNA sequencing samples and enables multiple applications that enhance medical insights. Here we applied this model to Nasopharyngeal Carcinoma (NPC) as in-context few-short format (i.e., the model was never trained with any NPC data). We conducted all four supported functions, which include sample stratification, vectorization, pseudo data generation, and marker identification. The results were then used for identifying metastatic NPC and to investigate mechanisms associated with NPC metastasis. Examination with stratification showed that the accuracy of RNAGAN results for evaluating the metastasis risk in NPC patients are comparable to or outcompeted recently published risk estimation linear prediction model. Vectorization results presented consistency across multiple cohorts and RNAGAN model versions. In the task of identifying markers and mechanisms related to NPC metastasis, incorporating pseudo data substantially enhanced the representativeness of single-cohort-based differential expression (DE) analysis. Moreover, RNAGAN identified metastasis-related marker genes base on single cohort, were concordant with the ground truth obtained across multiple cohorts (p=1.05e-9). Regarding biomedical mechanisms, RNAGAN enabled second-order feature extraction, revealing remarkable domination of the protective function of adaptive immune responses (indicated by IL21R levels) over the hazardous function of chronic, non-resolving innate inflammation (indicated by S100A8 levels) against NPC metastasis after first-line treatment. This association demonstrates a high degree of consistency with the external cohort. This study demonstrates the significant value of the foundation model RNAGAN in establishing robust medical applications and insights for cancer types it encounters for the first time without extra training. This work also revealed the important mechanisms for preventing distant metastasis through humoral anti-tumor immunity in NPC. High S100A8 levels trigger an upstream inflammatory cascade via TLR4 and the NLRP3 inflammasome, accelerating IL-1 signaling to drive distant metastasis, potentially through Epithelial-Mesenchymal Transition (EMT). When IL-21R expression is simultaneously high, local B cells capture crucial IL-21 survival signals, overriding the suppressive tissue stress to form functional Tertiary Lymphoid Structures (TLS). These newly assembled lymphoid structures orchestrate the local maturation of IgG-secreting plasma cells, which immediately opsonize and eliminate emerging EMT tumor cells before they can successfully escape into circulation.

PURPOSE: Nasopharyngeal carcinoma (NPC) is a multifactorial malignancy often diagnosed at an advanced stage due to nonspecific early symptoms. Accurate prognostic stratification is essential for individualized therapy but remains challenging because of biological and clinical heterogeneity. This study aimed to develop and validate a gene expression–based prognostic signature for locally advanced NPC. METHODS: This retrospective biomarker investigation combined transcriptomic profiling and survival analyses. A prognostic model was constructed using LASSO-Cox regression in the discovery cohort (N = 99) from the multicenter, randomized phase III trial NPC-0501, and validated in independent cohorts (N = 133) from Queen Elizabeth Hospital (QEH), Hong Kong, and Sun Yat-sen University Cancer Center (SYS), Guangzhou, using different profiling methods. Mechanisms underlying the signature were explored using single-cell RNA-seq (scRNA-seq) data. The primary outcome was 3-year progression-free survival (PFS), with 5-year overall survival (OS) as a secondary end point. Performance was evaluated using hazard ratios (HRs) and survival probabilities (SP). RESULTS: A 5-gene signature ( SPP1 , IL18BP , PALMD , WDR35 , and SOCS6 ) predicted 3-year PFS in the NPC-0501 cohort (HR, 0.046 [95% CI, 0.011 to 0.190]; P < .001), separating high-risk from low-risk patients (SP, 42.9% v 96.0%; P < .0001). It predicted survival in both validation cohorts, with associations with 5-year OS in QEH (SP, 72.7% v 100%; P < .001) and 3-year PFS in SYS (SP, 70.5% v 93.2%; P = .0059). It outperformed a published metastasis-related model. Single-cell analyses showed SPP1 enrichment in M2 macrophages, linking high risk with an immunosuppressive tumor microenvironment. CONCLUSION: A robust five-gene signature was established and validated for prognostic stratification of locally advanced NPC. Reproducibility across transcriptomic platforms and biological relevance support clinical application to guide personalized treatment.

The advent of artificial intelligence (AI) has brought revolutionary tools for biomedical transcriptomic (RNA-level) research. However, there are persistent constraints including limited interpretations with biomedical concepts such as functional pathways, small sample sizes and substantial time and computing power requirements for AI training. To overcome these limitations, we developed RNAGAN ( https://github.com/ZhaozhengHou-HKU/RNAGAN-1.0.git ), an AI tool with a generative adversarial network (GAN) structure with the objective of enhancing transcriptomic analysis. The network was established based on public human datasets comprising 4.6 million single cells from multiple organs and 5,900 sequenced samples of various cancer types with normal references. A specialized pathway neural layer was embedded to extract activities of predefined pathways from the Human Molecular Signatures Database (MSigDB), or newly learned pathways from single-cell data. The structure of RNAGAN (generator and discriminator) enables four applications after one shared training procedure: 1. single-cell and bulk-level patient stratification or differential diagnosis; 2. analysis of the gene and pathway markers in a selected disease; 3. pseudo data generation when sample size is limited for downstream analysis; 4. vectorization with gene and pathway-level features learned from multiple data sets. RNGAN contributes to the efficient utilization of limited data for transcriptomic studies.

Background: Nasopharyngeal carcinoma (NPC) patients who develop distant metastasis have significantly reduced survival rates. Therefore, understanding of metastasis and identifying high‐risk patients are important, and a robust predictive model for accurately assessing the distant‐metastasis risk before treatment is needed for personalised treatment. Methods: NPC patients diagnosed at four Hong Kong public hospitals and at Sun Yat‐Sen University Cancer Center in Guangzhou were selected. Patients were divided into two training cohorts (n = 77 and 30, respectively) and one testing cohort (n = 70). Two independent NPC cohorts collected from Sun Yat‐Sen University Cancer Center (n = 88), and a randomised phase III trial (NPC‐0501) in Hong Kong (n = 81) were used for external validation of the model‐based risk prediction. Results: Our RNA‐based risk score could stratify the patient groups into high and low risk of metastasis and disease progression in two independent external validation cohorts. In predicting NPC 3‐year distant metastasis, the score significantly improved the area under the curve from 84.8% to 90.4% when combined with the known prognostic clinical parameters. This RNA‐based risk score was highly associated with dysregulated functions of B cells and T helper 17 cells and reduced plasma B cells and tertiary lymphoid structure (TLS) formation. The analysis of biopsy samples revealed a significant enrichment of the TLS in non‐metastatic NPC patients. Conclusions: This study improved the accuracy of NPC metastasis prediction and highlight the potential association of TLS against metastatic NPC, encouraging future studies to understand how TLS interacts with NPC to prevent distant metastasis. Furthermore, the multi‐cohort Pareto‐optimisation‐based feature selection approach offers a practical method to explicitly avoid model overfitting and achieve a more robust model. Novelty and Impact: In this multicentre study, we established a new and robust predictive model for NPC distant metastasis using markers selected by a Pareto optimisation approach designed for multi‐cohort data. When combined with clinical parameters, our RNA‐based risk score significantly improved the area under the curve to 90.4%. This study revealed that reduced B‐cell immunity, and TLS formation, may be associated with NPC metastasis, providing insights for the future studies in NPC metastasis.

Epstein‐Barr virus (EBV) is a ubiquitous human virus that is also linked to various human cancers. In EBV‐associated nasopharyngeal carcinoma and gastric carcinoma cells, EBV expresses only essential viral antigens but high levels of long non‐coding RNAs known as BamHI‐A rightward transcripts ( lncBART s). The exact roles lncBART s in the EBV life cycle and the development of EBV‐associated cancers are largely not understood. This study demonstrates that lncBART s play a role in maintaining viral genome replication by affecting the tethering of EBV ori P region to chromosome. Mechanistically, lncBART s interact functionally with a complex consisting of Bromodomain‐containing protein 4 (BRD4), CCTC‐binding factor (CTCF), and EBV nuclear antigen 1 (EBNA1), anchoring their binding at ori P to facilitate EBV episome replication. Additionally, the lncBART s‐BRD4/CTCF complex contributes to the regulation of MYC and BCL2 expression. These findings suggest that lncBART s modulate EBV latency through interactions with ori P region, while lncBART s‐BRD4/CTCF complexes promote host epigenome reprogramming and drive tumorigenesis in EBV‐associated epithelial tumors.

Tumor metastasis leads to most cancer deaths. However, how cellular diversity and dynamic cooperation within the tumor microenvironment contribute to metastasis remains poorly understood. Here we leverage single-cell multi-omics (16 samples, 117,169 cells) and spatial transcriptomics (five samples, 195,366 cells) to uncover the cellular and spatial architecture of esophageal squamous cell carcinoma (ESCC), and characterize an immunosuppressive GPR116⁺ pericyte subset promoting tumor metastasis and immunotherapy resistance. GPR116⁺ pericyte enrichment is transcriptionally regulated by PRRX1, evidenced by pericyte-specific Prrx1 knockout mice. Mechanistically, GPR116⁺ pericytes secrete EGFL6 to bind integrin β1 on cancer cells, activating the NF-κB pathway to facilitate metastasis. Serum EGFL6 serves as a noninvasive biomarker for the diagnosis and prognosis of several tumors. Blocking integrin β1 suppresses metastasis and improves immunotherapy response in animal models of ESCC. Collectively, we provide a spatially resolved landscape of the prometastatic tumor microenvironment in ESCC and highlight the biological and clinical importance of GPR116⁺ pericytes, proposing potential innovative therapeutic strategies for metastatic cancers.

The poor prognosis of resectable esophageal squamous cell carcinoma (ESCC) poses an unmet need to identify early predictive and prognostic genomic biomarkers to improve treatment outcome and risk stratification. Mutational profiling was performed for 171 ESCC patients receiving curative neoadjuvant chemoradiation treatment (nCRT). The discovery cohort included 100 ESCC formalin-fixed paraffin-embedded (FFPE) tumor specimens; the validation FFPE cohort consisted of serial ctDNA samples from 71 patients. The discovery cohort identified hot-spot oncogenic NFE2L2 mutations exclusively localized at DLG and ETGE KEAP1-binding motifs in poor responders associated with incomplete pathological response (P = 0.004). Patients with NFE2L2 mutations in two independent FFPE cohorts had about 2-fold higher risk of death and recurrence. Serial ctDNA analysis further demonstrated oncogenic NFE2L2 mutations detected at post-nCRT were independent prognosticators for recurrence (HR = 5.90; P = 0.005) and survival (HR = 4.75; P = 0.013). Risk stratification based on pathological T and N stages, positive FFPE (HR = 4.50) and ctDNA NFE2L2 mutations (HR = 8.50) identified high-risk groups for recurrence (P = 0.001). Combined FFPE and ctDNA NFE2L2 mutation status predicted nCRT responses (P = 0.05) by ROC analysis. Tracking oncogenic NFE2L2 mutations at pre-treatment and post-surgery or serial ctDNA monitoring during treatment are useful nCRT predictors and independent prognosticators of survival for locally advanced ESCC.