High-risk EBV promotes immune evasion in nasopharyngeal carcinoma by upregulating HLA-DP via the encoded BALF2-HR variant

Abstract

Nasopharyngeal carcinoma (NPC) is prevalent in East and Southeast Asia, with genetic factors playing a significant role in its occurrence. The HLA gene region on chromosome 6 is linked to NPC susceptibility, but the mechanisms remain unclear. Epstein-Barr virus (EBV) infection is a well-established cause, with 95% of NPC patients being EBV-positive. Three key variations in the EBV genome (162215_C, 162476_C, 163364_T) in the BALF2 gene are strongly associated with NPC risk. This study finds that the high-risk BALF2 variant (BALF2-HR) upregulates HLA class II molecules, such as HLA-DP, which interacts with LAG-3 on CD8⁺ T cells, inhibiting cytokine secretion and promoting T cell exhaustion, leading to immune evasion and reduced anti-PD-1 efficacy. BALF2-HR also enhances HLA-DP transcription by binding to KPNA2 and facilitating CIITA nuclear translocation. Conjunctive immunotherapy with anti-LAG-3 and anti-PD-1 antibodies significantly improves NPC treatment. This work introduces a new therapeutic strategy for NPC and insights into infection-associated cancers.

Susceptibility of Nasopharyngeal carcinoma (NPC), associated with Epstein-Barr virus infection, is linked with HLA gene region with mechanisms unclear. The authors here show that the high-risk EBV subtype BALF2 upregulates HLA-DP, leading to CD8⁺ T cell exhaustion, and combined anti-LAG-3 and anti-PD-1 therapy improves anti-cancer efficacy.

Introduction

EBV, the first recognized human oncogenic virus¹, infects over 90% of the world’s adult population. However, most infected individuals remain asymptomatic, with only a small fraction developing malignancies. The EBV infects human cells and expresses viral proteins and non-coding RNAs that promote cell proliferation, inhibit cell death, and enable immune evasion. Over time, these changes drive the malignant transformation of infected cells, ultimately leading to the formation of tumors². EBV has been linked to various lymphomas, including Burkitt lymphoma³, Hodgkin lymphoma⁴, and NK/T-cell lymphoma⁵, as well as epithelial malignancies such as NPC, where over 95% of cases are EBV-positive and EBV-associated gastric carcinoma (EBVaGC), which constitutes 10% of gastric cancers⁶, contributing to over 200,000 cases annually and 1.8% of cancer-related deaths worldwide. Among these EBV-associated cancers, only NPC displays a significant geographic concentration^7,8. NPC is most prevalent in North Africa, Southeast Asia, and southern China, accounting for approximately 50% of global NPC cases^6,9,10. The etiology of NPC involves a combination of genetic predisposition, environmental and dietary carcinogens^2,6,11, which also suggests the presence of specific high-risk EBV subtype(s) that contribute to the elevated NPC rates in regions like southern China.

Early genomic analyses identified two major subtypes of EBV: Type 1, which is globally distributed, and Type 2, predominantly found in North Africa and possibly associated with EBV-induced lymphomas¹². Utilizing next-generation sequencing, we uncovered a distinct EBV subtype prevalent among NPC patients in southern China^13,14. Xu et al. subsequently expanded the sample size and performed a case-control association study, confirming a high-risk EBV subtype strongly linked to NPC development¹¹. This subtype is characterized by three specific mutations: 162215_C, 162476_C, and 163364_T, all located within the BamHI-A leftward reading frame 2 (BALF2) gene. This C-C-T high-risk subtype, termed BALF2-HR, is significantly associated with an increased risk of NPC¹¹. BALF2 encodes a single-stranded DNA-binding protein that plays a role in the initiation of early lytic replication and viral particle formation^15–17, and is also expressed during EBV latency¹⁷. However, the exact mechanism by which the high-risk BALF2-HR variant contributes to NPC development, as opposed to the low-risk BALF2-LR variant, is unclear.

NPC tissues are often characterized by a substantial infiltration of lymphocytes, classifying them as “hot” tumors. Recently, significant progress has been made in NPC immunotherapy, particularly with immune checkpoint inhibitors such as monoclonal antibodies targeting PD-1/PD-L1, which have become the first-line treatment for recurrent or metastatic NPC¹⁸. However, around 70% of patients do not respond to anti-PD-1/PD-L1 monotherapy. These non-responders frequently exhibit brief response durations and develop resistance, suggesting that other immune checkpoint molecules may play crucial roles in immune evasion and resistance to immunotherapy in NPC¹⁹. Recent single-cell transcriptome analyses have identified a unique population of tumor cells within NPC tissues that exhibit dual “epithelial-immune” characteristics²⁰. These NPC cells express classical epithelial markers, such as EpCAM, alongside immune molecules like HLA-II and complement proteins, which may enhance their ability to evade immune detection²⁰. As early as 1990, Lu et al. established a strong linkage between the HLA locus and NPC susceptibility through family linkage analysis²¹. Bei et al. further identified several single-nucleotide polymorphisms (SNPs) within the HLA region associated with NPC risk through case-control genome-wide association studies (GWAS)²². However, the regulatory mechanism responsible for the high expression of HLA-II on NPC cells and the role of these molecules in promoting immune evasion and resistance to immunotherapy remains elusive.

In this study, we investigate the role of a high-risk EBV subtype, BALF2-HR, in the progression of nasopharyngeal carcinoma. We discover that BALF2-HR upregulates HLA-II expression, particularly HLA-DP, in NPC cells, which interacts with the immune checkpoint molecule LAG-3 on T cells, promoting T cell exhaustion and immune evasion. Our findings suggest that combining anti-LAG-3 and anti-PD-1 monoclonal antibodies significantly enhances the efficacy of NPC immunotherapy. These results not only improve our understanding of the molecular mechanisms behind NPC immune evasion but also propose a novel therapeutic approach that could optimize immunotherapy outcomes in NPC patients.

Results

BALF2-HR Enhances HLA-II Expression in NPC

Our previous whole-genome sequencing analysis of EBV from 26 NPC biopsy tissues from Hunan Province (a high-incidence region for NPC in southern China)^13,14 revealed that 17 cases carried the C-C-T haplotype at the 162215, 162476, and 163364 loci, 3 cases had the C-C-C haplotype, 5 cases had the C-T-C haplotype, and 1 case had the A-C-C haplotype [Supplementary Fig. 1A]. Through association analysis, Xu et al. identified the BALF2 C-C-T haplotype as the most significantly associated with NPC risk (designated as BALF2-HR), with an odds ratio of 11.71 compared to the low-risk BALF2 A-T-C variant (designated as BALF2-LR) (P = 2.39 × 10⁻²⁴). The C-C-C haplotype showed a weaker association (odds ratio = 3.50, P = 1.22 × 10⁻⁵)¹¹. In this study, we collected another 128 fresh NPC biopsy tissues from Hunan Province [Supplementary Data 1]. Using RT-PCR, we detected EBV in 110 of these samples, while the remaining 18 were EBV-negative. In addition, we collected paraffin-embedded tissue sections from 58 EBV-positive nasopharyngeal carcinoma patients [Supplementary Data 2] and identified the EBV BALF2 genotypes of these 168 NPC samples using Sanger sequencing, revealing that 124 cases had the C-C-T genotype, 17 cases were C-C-C, and 27 cases had other genotypes, further confirming that the C-C-T genotype is most strongly associated with NPC risk [Fig. 1A]. In contrast, in NK/T-cell lymphoma, which also primarily affects the nasopharyngeal region but originates from a different cell type, 66.67% of the EBV BALF2 genotypes were low-risk A-T-C variant²³ [Supplementary Fig. 1B]. We performed immunohistochemical analysis on paraffin-embedded tissue sections from the aforementioned 58 cases of nasopharyngeal carcinoma and found that BALF2 is expressed in EBV-positive nasopharyngeal carcinoma tissues of both low-risk and high-risk subtypes. [Fig. 1B and Supplementary Fig. 1C]. These results suggest that the BALF2 C-C-T genotype represents a relatively unique high-risk EBV subtype and is widely prevalent among nasopharyngeal carcinoma patients in southern China.

Fig. 1. High-Risk EBV-Encoded BALF2 Enhances HLA-II Expression in NPC Cells.

A Sanger sequencing assessing the genotype distribution of high-risk EBV-encoded BALF2 in 168 NPC tissues from Hunan province. B Representative images of BALF2 expression detected by immunohistochemistry in paraffin-embedded tissue sections from 58 nasopharyngeal carcinoma cases, corresponding to Supplementary Fig. 1C. Magnification, × 400, scale bar, 100 μm; Magnification, × 630, scale bar, 50 μm. C HK1 cell lines were transfected with either BALF2-HR (BALF2 C-C-T) or BALF2-LR (BALF2 A-T-C), followed by RNA-seq analysis (3 samples per group) to identify genes regulated by BALF2-HR. Volcano plot displays the RNA-seq results, with significant genes defined by a p-value < 0.05 and |log2FC | > 1. D Volcano plot showing the differential expression of HLA Class II molecules in GSE194333, comparing C666-1 and NP69 cell lines (p-value < 0.05; |log2FC | > 1). E Flow cytometry evaluating the proportion of HLA-II cells among EPCAM⁺ cells in 16 BALF2-HR type and 9 other subtypes of NPC tissues. Original BALF2 genotyping results are shown in Supplementary Fig. 1I. F RT-qPCR assessing HLA-DP, HLA-DQ, and HLA-DR mRNA levels after overexpressing BALF2-HR or BALF2-LR in HK1 cells, and knocking out BALF2 in C666-1 cells. GAPDH serves as a loading control. n = 3 independent experiments. G Western blot assessing HLA-DP, HLA-DQ, and HLA-DR protein levels after overexpressing BALF2-HR or BALF2-LR in HK1 cells, and knocking out BALF2 in C666-1 cells. β-tubulin serves as a loading control. n = 3 independent experiments. H Immunofluorescence assessing HLA-DP, HLA-DQ, and HLA-DR expression after overexpressing BALF2-HR or BALF2-LR in HK1 cells, and knocking out BALF2 in C666-1 cells. FITC is shown in green; DAPI, which stains nuclei, is shown in blue. n = 3 independent experiments. I Flow cytometry assessing HLA-DP, HLA-DQ, and HLA-DR expression after overexpressing BALF2-HR or BALF2-LR in HK1 cells, and knocking out BALF2 in C666-1 cells. MFI represents median fluorescence intensity. The top panel displays the original images; the bottom panel presents the corresponding statistical results. n = 3 independent experiments. Correlations were calculated using an unpaired one-tailed Student’s t test for (E), while one-way ANOVA was applied for (F, I), *p  <  0.05, **p  <  0.01, ***p  <  0.001 and ****p  <  0.0001. Each Western blot and Immunofluorescence were reproduced three times with similar results (G, H). Data are presented as mean ± SD.

To investigate the biological function of EBV BALF2-HR in NPC pathogenesis, we constructed overexpression vectors for BALF2-HR and BALF2-LR (used as a control) and transfected them into the EBV-negative NPC cell line HK1 [Supplementary Fig. 1D]. Differential gene expression analysis, conducted using RNA sequencing (RNA-seq) [Supplementary Data 3], revealed that BALF2-HR significantly upregulated the expression of HLA-II compared to BALF2-LR [Fig. 1C]. We further sequenced the BALF2 gene in the commonly used EBV-positive NPC cell line C666-1 and confirmed that it harbors the BALF2-HR (C-C-T) genotype [Supplementary Fig. 1E]. Analysis of the whole genome expression profile (GSE194333) of C666-1 and immortalized EBV-negative nasopharyngeal epithelial cells, NP69 also showed significantly higher expression of HLA-II in C666-1 cells [Fig. 1D]. Additionally, we analyzed single-cell sequencing data of 4 EBV-negative and 6 EBV-positive NPC biopsy tissues (GSE150825)²⁴, focusing on HLA-II expression. The results demonstrated that HLA-DPA1, DPB1, DQA1, DQB1, DRA, and DRB1 were significantly upregulated in EBV-positive NPC cells [Supplementary Fig. 1F-H].

We also collected fresh human tissue samples from 25 cases of EBV-infected NPC [Supplementary Data 4]. A portion of these samples underwent Sanger sequencing to determine the genotype of the BALF2 gene, revealing that 16 cases carried the high-risk C-C-T subtype [Supplementary Fig. 1I]. Another portion of the samples was dissociated into a single-cell suspension, and flow cytometry was used to assess the expression of HLA-II on the surface of NPC cells (EpCAM⁺). The results showed that HLA-II levels in epithelial cells from the 16 BALF2-HR NPC tissues were significantly higher than those from the 9 NPC tissues with other BALF2 subtypes [Fig. 1E].

In addition to the high-risk EBV-positive NPC cell line C666-1, we also obtained two other cell lines, HONE1-EBV and HK1-EBV, which were infected with the recombinant Akata EBV, kindly provided by Prof. Sai Wah Tsao of the University of Hong Kong²⁵. Sanger sequencing revealed that the Akata EBV carries the low-risk A-T-C BALF2 subtype [Supplementary Fig. 1J]. RT-PCR, Western blot, and immunofluorescence assays confirmed the expression of BALF2 mRNA [Supplementary Fig. 1K] and protein [Supplementary Fig. 1L-M] in all three cell lines. Subsequent Western blot [Supplementary Fig. 2A] and immunofluorescence experiments [Supplementary Fig. 2B] showed that the expression of HLA-II molecules (HLA-DP, HLA-DQ, and HLA-DR) was significantly higher in C666-1 cells carrying BALF2-HR compared to EBV-negative (HONE1, HK1) or BALF2-LR expressing (HONE1-EBV, HK1-EBV) NPC cell lines. Of note, knockout of BALF2 using sgRNA in C666-1 cells carrying high-risk EBV and in HK1-EBV and HONE1-EBV cells infected with low-risk EBV did not alter the latent state of EBV [Supplementary Fig. 2C]. However, overexpression of BALF2-HR or BALF2-LR in the EBV-negative NPC cell lines HK1 and HONE1 [Supplementary Fig. 2D], or knocking out BALF2-HR in C666-1, showed that only overexpressing or knocking out BALF2-HR can significantly upregulate or downregulate the expression of classical HLA-II in NPC cells, as revealed by the results of RT-qPCR [Fig. 1F, Supplementary Fig. 2E], Western blot [Fig. 1G, Supplementary Fig. 2F], immunofluorescence [Fig. 1H, Supplementary Fig. 2G], and flow cytometry [Fig. 1I, Supplementary Fig. 2H]. In contrast, knocking out BALF2-LR in HONE1-EBV and HK1-EBV did not result in a significant change in HLA-II [Supplementary Fig. 3A-C]. These results suggest that the high-risk EBV upregulates HLA-II expression in NPC cells via BALF2-HR, whereas BALF2-LR has no significant effect on HLA-II expression.

BALF2-HR Drives Immune Evasion in NPC and Contributes to Anti-PD-1 Immunotherapy Resistance

HLA class II molecules are primarily expressed on the surface of antigen-presenting cells, where their peptide-binding groove (comprising the α₁ and β₁ domains) binds exogenous antigenic peptides. CD4⁺ T cells recognize the HLA class II-peptide complex through their T-cell receptor (TCR), thereby initiating an immune response²⁶. However, recent studies have shown that they can also bind to the inhibitory immune checkpoint molecule LAG-3 on T cells, thereby suppressing T cell function^27–30. Using flow cytometry, we analyzed the aforementioned 16 NPC biopsy samples infected with high-risk EBV and 9 samples with non-high-risk EBV [Supplementary Data 3]. In NPC tissues infected with high-risk EBV, we observed a significantly reduced proportion of CD8⁺ T cells, along with decreased expression of cytotoxic cytokines in these cells, indicative of a pronounced exhausted phenotype. Notably, the proportions of CD4⁺ T cells, regulatory T cells (Tregs), Th1 cells, Th2 cells, Th17 cells, M2 macrophages, B cells, and NK cells were not significantly altered. These findings suggest that high-risk EBV subtypes may selectively impair CD8⁺ T cell function and promote their exhaustion [Fig. 2A, Supplementary Fig. 4A-D].

Fig. 2. BALF2-HR Promotes Immune Evasion in NPC and is Linked to Poor Response to Anti-PD-1 Immunotherapy.

A Flow cytometry analysis was performed on NPC tissues from 16 high-risk subtype (BALF2-HR) patients and 9 patients with other subtypes. The analysis assessed the proportions of CD8⁺ T cells and the expression of cytotoxic cytokines, including IFN-γ, GZMB, and TNF, within these cells. In addition, the proportions of CD4⁺ T cells, Treg cells, NK cells, B cells, and M2 macrophages were evaluated. The flow cytometry gating strategy is illustrated in Supplementary Fig. 4A-B, with representative images shown in Supplementary Fig. 4C. B Top, Sanger sequencing analysis revealing differences in tumor volume changes post-PD-1 treatment between 39 high-risk subtypes (BALF2-HR) and 19 other subtypes of EBV-positive NPC patients after PD-1 antibody immunotherapy. The difference in treatment response between BALF2-HR and other subtypes was statistically significant (P = 0.0043). Bottom, Immunohistochemical detection of BALF2 expression in these nasopharyngeal carcinoma patients, revealing that BALF2-HR expression levels are closely associated with the efficacy of immunotherapy (P < 0.0001). Complete Response (CR): Disappearance of all target lesions and normalization of tumor markers for at least 4 weeks without new lesions. Partial Response (PR): Reduction in the sum of the diameters of target lesions by ≥ 30% and sustained for at least 4 weeks. Stable Disease (SD): Reduction in the sum of the diameters of target lesions but not reaching PR, or a slight increase but not reaching Progressive Disease. Progressive Disease (PD): Increase in the sum of the diameters of target lesions by ≥ 20% or the appearance of new lesions. C Overexpression of BALF2-HR or BALF2-LR in HK1 cells, and BALF2 knockout in C666-1 cells. Cells were then co-cultured with CD8⁺ T cells for 24 h and treated with 5 µg of IgG, anti-PD-1 antibody, or a combination of anti-PD-1 antibody and anti-LAG-3 antibody. Crystal violet staining was used to assess the viability of tumor cells. n = 3 independent volunteers. D Overexpression of BALF2-HR or BALF2-LR in HK1 cells, and BALF2 knockout in C666-1 cells. Cells were then co-cultured with CD8⁺ T cells for 24 h and treated with 5 µg of IgG, anti-PD-1 antibody, or a combination of anti-PD-1 antibody and anti-LAG-3 antibody. Annexin V-PI/Flow cytometry was employed to detect apoptosis in tumor cells. n = 3 independent volunteers. E Overexpression of BALF2-HR or BALF2-LR in HK1 cells, and BALF2 knockout in C666-1 cells. Cells were then co-cultured with CD8⁺ T cells for 24 h and treated with 5 µg of IgG, anti-PD-1 antibody, or a combination of anti-PD-1 antibody and anti-LAG-3 antibody. Flow cytometry was used to determine the proportions of IFN-γ⁺ TNF⁺ and IFN-γ⁺ GZMB⁺ cells in CD8⁺ T cells. n = 3 independent volunteers. F Knockout of BALF2 in C666-1 cells. Cells were then co-cultured with CD8⁺ T cells for 24 h and treated with 5 µg of IgG, anti-PD-1 antibody, or a combination of anti-PD-1 antibody and anti-LAG-3 antibody. ELISA was used to detect IFN-γ, GZMB, and TNF levels in the co-culture supernatant. n = 3 independent volunteers. G Overexpression of BALF2-HR or BALF2-LR in HK1 cells, and BALF2 knockout in C666-1 cells. Cells were then co-cultured with CD8⁺ T cells for 24 hand treated with 5 µg of IgG, anti-PD-1 antibody, or a combination of anti-PD-1 antibody and anti-LAG-3 antibody. Flow cytometry was used to determine the proportions of PD-1⁺TIM-3⁺ cells in CD8⁺ T cells. n = 3 independent volunteers. Correlations were calculated using an unpaired two-tailed Student’s t test for (A, B), while two-way ANOVA was applied for (D, E, F, G), *p  <  0.05, **p  <  0.01, ***p  <  0.001 and ****p  <  0.0001. Data are presented as mean ± SD.

To further assess the impact of HLA class II molecules on tumor cells in modulating CD4⁺ T cell responses, human PBMCs were isolated, and primary CD4⁺ T cells were purified using CD4 microbeads. These CD4⁺ T cells were co-cultured for a week with dendritic cells and irradiated NPC cells to generate NPC-reactive CD4⁺ T cells³¹, and were later HLA-DR matched with nasopharyngeal carcinoma cells to prevent immune rejection. Subsequently, BALF2-HR or BALF2-LR was overexpressed in HK1 and HONE1 cells, while BALF2-HR was knocked out in C666-1 cells. These modified tumor cells were then co-cultured with NPC-reactive CD4⁺ T cells. Flow cytometry analysis of IL-2, Ki-67, CD69, and FOXP3 expression in CD4⁺ T cells revealed that the high expression of tumor cell HLA-II molecules, mediated by BALF2-HR, partially promoted CD4⁺ T cell activation [Supplementary Fig. 4E], as previously reported³².

To clarify the interaction between HLA class II molecules and CD8⁺ T cells, primary CD8⁺ T cells were isolated from human PBMCs using CD8 microbeads. These cells were then co-cultured with dendritic cells and irradiated NPC cells for one week to produce NPC-reactive CD8⁺ T cells. Subsequently, NPC-reactive CD4⁺ T cells and NPC-reactive CD8⁺ T cells were co-cultured with tumor cells expressing BALF2-HR. Tumor cell viability, assessed by crystal violet staining, indicated that BALF2-HR exerts its effect predominantly through CD8⁺ T cells rather than CD4⁺ T cells [Supplementary Fig. 4F]. This conclusion was further supported by flow cytometry analysis of tumor cell apoptosis [Supplementary Fig. 4G]. Single-cell RNA sequencing data from the GSE150430 dataset showed that LAG-3 is mainly expressed on CD8⁺ T cells, with relatively lower expression on CD4⁺ T cells [Supplementary Fig. 4H]. These results indicate that in the absence of CD8⁺ T cells, BALF2-HR-mediated CD4⁺ T cell activation is unable to suppress the growth of nasopharyngeal carcinoma cells. Furthermore, flow cytometry and co-culture experiments on nasopharyngeal carcinoma tissues both indicate that BALF2-HR suppresses the cytotoxic function of CD8⁺ T cells. Future studies will investigate the interaction between BALF2-HR and CD8⁺ T cells.

The 58 NPC tissue paraffin sections used in Fig. 1B were from patients who received a combination of conventional treatment and anti-PD-1 (Nivolumab) therapy [Supplementary Data 2]. Of these, 39 cases harbored the high-risk BALF2 C-C-T, while the remaining 19 cases had non-high-risk BALF2 [Supplementary Fig. 5A]. Further analysis of the relationship between BALF2 genotypes and immunotherapy efficacy revealed that among the 39 patients with the BALF2-HR subtype, only 9 achieved a complete response (CR), resulting in a CR rate of 23.08%. In addition, 12 patients showed a partial response (PR), yielding a PR rate of 30.77%. In contrast, among the 19 patients with non-high-risk EBV subtypes, 8 achieved CR (42.10%), and 9 had PR (47.37%). The overall objective response rate (CR + PR) was 53.85% in the BALF2-HR group compared to 89.47% in the non-high-risk group (P = 0.0043) [Fig. 2B]. Immunohistochemical analysis of BALF2 expression in these tissues revealed that only the high-risk BALF2 expression level showed a significant positive correlation with the efficacy of anti-PD-1 therapy. The mean expression value for PD + SD was 8.3, while the mean expression value for PR + CR was 4.8 [Fig. 2B]. These results suggest that the BALF2-HR promotes immune evasion in NPC by upregulating HLA-II and inducing CD8⁺ T cell exhaustion, which may contribute to resistance to anti-PD-1 immunotherapy.

To further validate our hypothesis, we overexpressed BALF2-LR or BALF2-HR in EBV-negative NPC cell lines HONE1 and HK1 or knocked out BALF2 in BALF2-HR-infected NPC cell line C666-1. We found that neither overexpression nor knockout of BALF2-HR affected the proliferation of NPC cells [Supplementary Fig. 5B]. Next, we isolated CD8⁺ T cells and co-cultured them with the aforementioned NPC cells [Supplementary Fig. 5C]. The results showed that overexpression of BALF2-HR significantly enhanced NPC cell survival in co-culture with CD8⁺ T cells, while knocking out BALF2-HR significantly reduced NPC cell survival. When we added a PD-1 monoclonal antibody (nivolumab) to the culture medium, along with an IgG control, there was no significant change in BALF2-HR-induced NPC cell survival. However, combining PD-1 and LAG-3 monoclonal antibodies (Relatlimab) significantly reduced BALF2-HR-induced NPC cell survival [Fig. 2C, Supplementary Fig. 5D]. We also used flow cytometry to assess the apoptosis of NPC cells from the same experiments. Overexpression of BALF2-HR significantly inhibited CD8⁺ T cell-induced apoptosis of NPC cells, while knockout of BALF2-HR significantly promoted NPC cell apoptosis. Adding the PD-1 monoclonal antibody alone did not affect the reduction in apoptosis mediated by BALF2-HR, but the combined use of PD-1 and LAG-3 monoclonal antibodies significantly reversed the reduction in NPC cell apoptosis mediated by BALF2-HR [Fig. 2D, Supplementary Fig. 5E].

In the co-culture model of NPC cells and CD8⁺ T cells, we also assessed T cell functionality and activity using flow cytometry. NPC cells expressing BALF2-HR significantly reduced the proportion of CD8⁺ T cells expressing effector cytokines IFN-γ⁺ GZMB⁺ and IFN-γ⁺ TNF⁺, while knocking out BALF2-HR restored the proportion of these cytokine-expressing CD8⁺ T cells. The addition of PD-1 monoclonal antibody alone did not reverse BALF2-HR-mediated reduction in cytokine expression in CD8⁺ T cells, but the combined use of PD-1 and LAG-3 monoclonal antibodies effectively restored cytokine expression [Fig. 2E, Supplementary Fig. 5F]. ELISA [Fig. 2F, Supplementary Fig. 5G] and RT-qPCR [Supplementary Fig. 5H] assays measuring the expression of effector cytokines IFN-γ, GZMB, and TNF in CD8⁺ T cells yielded similar results. During the exhaustion of CD8⁺ T cells, not only is there a downregulation of effector cytokine expression, but also an increase in the expression of inhibitory receptors, such as PD-1 and TIM-3^31,32. Flow cytometry confirmed that the proportion of PD-1⁺TIM-3⁺ cells among CD8⁺ T cells increased after co-culture with BALF2-HR-expressing NPC cells, while this proportion decreased when BALF2-HR was knocked out. Adding PD-1 monoclonal antibody alone did not reverse BALF2-HR-mediated CD8⁺ T cell exhaustion, but the combined use of PD-1 and LAG-3 monoclonal antibodies significantly inhibited the exhaustion of CD8⁺ T cells [Fig. 2G, Supplementary Fig. 5I]. RT-qPCR [Supplementary Fig. 5J] examination of the expression of CD8⁺ T cell exhaustion-related genes PDCD1, TIM-3, and TOX also supported these findings. These results suggest that BALF2-HR promotes immune evasion and resistance to immunotherapy in NPC by upregulating HLA class II molecules on NPC cells, leading to CD8⁺ T cell exhaustion and impaired cytotoxic function. LAG-3 monoclonal antibody restores T cell function, counteracts CD8⁺ T cell exhaustion, and enhances the efficacy of PD-1 monoclonal antibody therapy.

BALF2-HR Induces Immune Evasion and Immunotherapy Resistance in NPC via Upregulation of HLA-DP

Class II HLA molecules, including HLA-DP, HLA-DQ, and HLA-DR, are crucial components in immune regulation. To further investigate their roles in immune evasion and immunotherapy resistance in NPC, we overexpressed or knocked out HLA-DP, HLA-DQ, and HLA-DR in NPC cell lines [Supplementary Fig. 6A-B]. Crystal violet staining, used to assess tumor cell survival, showed that these HLA-II molecules do not directly influence tumor cell proliferation [Supplementary Fig. 6C]. However, in co-culture with CD8⁺ T cells, overexpression of HLA-DP in NPC cells significantly increased resistance to CD8⁺ T cell-mediated cytotoxicity. In contrast, overexpression of HLA-DQ or HLA-DR had no impact on the survival of co-cultured NPC cells. Using sgRNA to knock out these HLA-II molecules, we found that knocking out HLA-DP significantly reduced the survival of co-cultured NPC cells, while the knockout of HLA-DQ or HLA-DR did not notably affect NPC cell survival. Furthermore, adding PD-1 monoclonal antibody alone to the co-culture system did not affect HLA-DP-mediated survival of NPC cells. However, the combined use of PD-1 and LAG-3 monoclonal antibodies significantly inhibited the survival of HLA-DP-expressing NPC cells [Fig. 3A, Supplementary Fig. 6D]. Similarly, Annexin V-PI staining revealed that overexpression of HLA-DP decreased NPC cell apoptosis, while its knockout increased apoptosis in the co-culture system [Fig. 3B, Supplementary Fig. 6E]. These results indicate that, among classical HLA-II molecules, HLA-DP plays a key role in promoting immune evasion in NPC and contributes to resistance against PD-1 therapy. Since LAG-3 serves as a receptor for HLA-II molecules, its monoclonal antibody can inhibit HLA-DP-mediated immune evasion in NPC cells, thereby enhancing the efficacy of PD-1 immunotherapy.

Fig. 3. HLA-DP Promotes Immune Evasion in NPC and Contributes to Poor Efficacy of Anti-PD-1 Immunotherapy.

A Overexpression or knockout of HLA-DP, HLA-DQ, or HLA-DR in HK1 and C666-1 cells. Cells were then co-cultured with CD8⁺ T cells for 24 h and treated with 5 µg of IgG, anti-PD-1 antibody, or a combination of anti-PD-1 antibody and anti-LAG-3 antibody. Crystal violet staining was used to assess tumor cell viability. The upper panel shows representative images, and the lower panel shows statistical results. n = 3 independent volunteers. B Cells were then co-cultured with CD8⁺ T cells for 24 h and treated with 5 µg of IgG, anti-PD-1 antibody, or a combination of anti-PD-1 antibody and anti-LAG-3 antibody. Annexin V-PI/Flow cytometry was employed to detect apoptosis in tumor cells. n = 3 independent volunteers. C Overexpression or knockout of HLA-DP, HLA-DQ, or HLA-DR in HK1 and C666-1 cells. Cells were then co-cultured with CD8⁺ T cells for 24 hnd treated with 5 µg of IgG, anti-PD-1 antibody, or a combination of anti-PD-1 antibody and anti-LAG-3 antibody. Flow cytometry was used to determine the proportions of IFN-γ⁺ TNF⁺ and IFN-γ⁺ GZMB⁺ cells in CD8⁺ T cells. n = 3 independent volunteers. D Overexpression or knockout of HLA-DP, HLA-DQ, or HLA-DR in HK1 and C666-1 cells. Cells were then co-cultured with CD8⁺ T cells for 24 h and treated with 5 µg of IgG, anti-PD-1 antibody, or a combination of anti-PD-1 antibody and anti-LAG-3 antibody. Flow cytometry was used to assess the proportion of PD-1⁺TIM-3⁺ cells within CD8⁺ T cells. n = 3 independent volunteers. Correlations were calculated using two-way ANOVA for (A–D), *p  <  0.05, **p  <  0.01, ***p  <  0.001 and ****p  <  0.0001. Data are presented as mean ± SD.

Co-culturing NPC cells with CD8⁺ T cells and analyzing via flow cytometry [Fig. 3C, Supplementary Fig. 6F] demonstrated that overexpression of HLA-DP in NPC cells significantly reduced the proportion of CD8⁺ T cells expressing the effector cytokines IFN-γ⁺ GZMB⁺ and IFN-γ⁺ TNF⁺. Knocking out HLA-DP expression in NPC cells using sgRNA restored the proportions of these cytokine-expressing CD8⁺ T cells. The addition of PD-1 monoclonal antibody alone did not reverse the HLA-DP-mediated reduction in CD8⁺ T cell effector cytokine expression. However, the combined use of PD-1 and LAG-3 monoclonal antibodies effectively restored effector cytokine expression in CD8⁺ T cells. ELISA [Supplementary Fig. 6G-H] and RT-qPCR [Supplementary Fig. 6I-J] assays for IFN-γ, GZMB, and TNF in CD8⁺ T cells support these results. Further flow cytometry analysis [Fig. 3D, Supplementary Fig. 6K] showed that NPC cells overexpressing HLA-DP increased the proportion of PD-1⁺TIM-3⁺ cells among CD8⁺ T cells, indicative of T cell exhaustion, while knocking out HLA-DP reduced the proportion of these exhausted CD8⁺ T cells. The addition of PD-1 monoclonal antibody alone did not reverse HLA-DP-mediated CD8⁺ T cell exhaustion, but the combined use of PD-1 and LAG-3 monoclonal antibodies successfully reversed this effect. RT-qPCR [Supplementary Fig. 6L-M] for T cell exhaustion markers such as PDCD1, TIM-3, and TOX, also yielded consistent findings. These results further confirm that HLA-DP plays a pivotal role in immune evasion in NPC by inhibiting T cell cytotoxic function and promoting T cell exhaustion. The combined use of PD-1 and LAG-3 monoclonal antibodies can counteract the inhibitory effects of HLA-DP on T cells, offering a promising strategy to improve immunotherapy efficacy.

In the above study, we used the NCBI reference sequences for HLA-DP, specifically HLA-DPA101:03 and HLA-DPB104:01. Given the population-specific distribution of HLA-DP alleles, deep sequencing of the MHC-II region in the Chinese population³³ revealed that the predominant HLA-DPA1 alleles are *02:02 (48.54%), *01:03 (39.39%), and *02:01 (11.35%), while the major HLA-DPB1 alleles are *05:01 (36.86%), *02:01 (19.74%), and *04:01 (10.60%). To investigate the functional diversity among HLA-DP alleles, we synthesized plasmids encoding the relevant variants and overexpressed them in the HK1 and C666-1 nasopharyngeal carcinoma (NPC) cell lines. These alleles did not significantly affect tumor cell proliferation [Supplementary Fig. 7A-B]. However, co-culture of HLA-DP–overexpressing tumor cells with CD8⁺ T cells revealed that different HLA-DP alleles enhanced tumor cell survival, as shown by crystal violet staining [Supplementary Fig. 7C] and supported by flow cytometry analysis of tumor cell apoptosis [Supplementary Fig. 7D]. Further flow cytometry analysis of CD8⁺ T cells revealed that various HLA-DP alleles differentially suppressed cytotoxic cytokine production and promoted T cell exhaustion [Supplementary Fig. 7E-F]. Together, these results suggest that HLA-DP alleles contribute to immune evasion in NPC to varying degrees, potentially through allele-specific amino acid variations that alter the structure of the antigen-binding groove and affect its interaction with LAG-3.

We overexpressed BALF2-HR in the EBV-negative NPC cell lines HONE1 and HK1 while knocking out HLA-DP, HLA-DQ, or HLA-DR individually. Additionally, in the high-risk EBV-infected NPC cell line C666-1, we knocked out BALF2-HR and overexpressed HLA-DP, HLA-DQ, or HLA-DR. Proliferation of these NPC cells remained unaffected [Supplementary Fig. 8A]. When co-cultured with CD8⁺ T cells and assessed by crystal violet staining, we found that knocking out HLA-DP expression significantly inhibited BALF2-HR-mediated NPC cell survival. Conversely, overexpression of HLA-DP reversed the decreased tumor cell survival caused by BALF2-HR knockout. In NPC cells overexpressing BALF2-HR and simultaneously knocking out HLA-DP, the addition of PD-1 monoclonal antibody enhanced CD8⁺ T cell-mediated cytotoxicity against NPC cells. However, overexpression of HLA-DP reversed the increased sensitivity of BALF2-HR knockout tumor cells to PD-1 monoclonal antibody treatment, leading to increased NPC cell survival and resistance to anti-PD-1 immunotherapy. The combined use of PD-1 and LAG-3 monoclonal antibodies further enhanced CD8⁺ T cell-mediated killing of NPC cells [Fig. 4A, B and Supplementary Fig. 8B]. Flow cytometry analysis of NPC cell apoptosis under these various treatment conditions yielded similar results [Fig. 4C and Supplementary Fig. 8C].

Fig. 4. BALF2-HR Upregulates HLA-DP Expression to Promote Immune Evasion in NPC.

A, B Crystal violet staining assessing the viability of HK1 cells with HLA-DP, HLA-DQ, and HLA-DR knockout while overexpressing BALF2-HR or BALF2-LR, or C666-1 cells with HLA-DP, HLA-DQ, and HLA-DR overexpression while knocking out BALF2. Cells were co-cultured with CD8⁺ T cells for 24 h and treated with 5 µg of IgG, anti-PD-1 antibody, or a combination of anti-PD-1 antibody and anti-LAG-3 antibody. Panel A shows representative images, and panel B shows statistical results. n = 3 independent volunteers. C Annexin V-PI/Flow cytometry detecting apoptosis levels among HK1 cells with HLA-DP, HLA-DQ, and HLA-DR knockout while overexpressing BALF2-HR or BALF2-LR, or C666-1 cells with HLA-DP, HLA-DQ, and HLA-DR overexpression while knocking out BALF2. Cells were co-cultured with CD8⁺ T cells for 24 h and treated with 5 µg of IgG, anti-PD-1 antibody, or a combination of anti-PD-1 antibody and anti-LAG-3 antibody. n = 3 independent volunteers. D Flow cytometry detecting proportions of IFN-γ⁺ TNF⁺ and IFN-γ⁺ GZMB⁺ cells in CD8⁺ T cells co-cultured with HK1 cells with HLA-DP, HLA-DQ, and HLA-DR knockout, while overexpressing BALF2-HR or LR, or C666-1 cells with HLA-DP, HLA-DQ, and HLA-DR overexpression while knocking out BALF2. Cells were co-cultured with CD8⁺ T cells for 24 h and treated with 5 µg of IgG, anti-PD-1 antibody, or a combination of anti-PD-1 antibody and anti-LAG-3 antibody. n = 3 independent volunteers. E ELISA measuring levels of IFN-γ, GZMB, and TNF in the co-culture supernatant of HK1 cells with HLA-DP, HLA-DQ, and HLA-DR knockout, while overexpressing BALF2-HR or BALF2-LR, or C666-1 cells with HLA-DP, HLA-DQ, and HLA-DR overexpression while knocking out BALF2. Cells were co-cultured with CD8⁺ T cells for 24 hours and treated with 5 µg of IgG, anti-PD-1 antibody, or a combination of anti-PD-1 antibody and anti-LAG-3 antibody. n = 3 independent volunteers. F Flow cytometry assessing the proportions of PD-1⁺TIM-3⁺ cells in CD8⁺ T cells co-cultured with HK1 cells with HLA-DP, HLA-DQ, and HLA-DR knockout while overexpressing BALF2-HR or BALF2-LR, or C666-1 cells with HLA-DP, HLA-DQ, and HLA-DR overexpression while knocking out BALF2. Cells were co-cultured with CD8⁺ T cells for 24 h and treated with 5 µg of IgG, anti-PD-1 antibody, or a combination of anti-PD-1 antibody and anti-LAG-3 antibody. n = 3 independent volunteers. Correlations were calculated using two-way ANOVA for (B–F), *p  <  0.05, **p  <  0.01, ***p <  0.001 and ****p  <  0.0001. Data are presented as mean ± SD.

Flow cytometry analysis of CD8⁺ T cell function and activity [Fig. 4D, Supplementary Fig. 8D] demonstrated that overexpression of BALF2-HR in NPC cells decreased the cytotoxic function of co-cultured CD8⁺ T cells. However, knocking out HLA-DP in BALF2-HR-overexpressing NPC cells reversed this suppression of T cell cytotoxic function. Overexpression of HLA-DP in C666-1 cells with BALF2-HR knockout reversed the enhanced CD8⁺ T cell cytotoxic function observed with BALF2-HR knockout. In addition, NPC cells lacking HLA-DP exhibited increased sensitivity to PD-1 monoclonal antibody treatment, while cells overexpressing HLA-DP became resistant to PD-1 blockade. The combination of PD-1 and LAG-3 monoclonal antibodies resulted in significantly improved CD8⁺ T cell-mediated killing of NPC cells compared to PD-1 treatment alone. ELISA [Fig. 4E, Supplementary Fig. 8E] and RT-qPCR [Supplementary Fig. 8F] analyses of CD8⁺ T cell cytotoxic cytokines, including IFN-γ, GZMB, and TNF produced similar findings. Further flow cytometry analysis [Fig. 4F, Supplementary Fig. 8G] confirmed that either knocking out HLA-DP or adding LAG-3 monoclonal antibody could reverse BALF2-HR-mediated CD8⁺ T cell exhaustion and resistance to PD-1 therapy. RT-qPCR [Supplementary Fig. 8H] showed comparable results for exhaustion markers PDCD1, TOX, and TIM-3 in CD8⁺ T cells. These findings indicate that HLA-DP plays a crucial role in mediating BALF2-HR-induced immune evasion in NPC.

HLA-II molecules not only serve as classical ligands for LAG-3 but also act as co-ligands for CD4³⁰. Due to the structural similarity between LAG-3 and CD4, it is possible that CD4 could competitively bind to both LAG-3 and HLA-II. To further assess the binding affinity of HLA-II to LAG-3 and CD4, we overexpressed HLA-II in NPC cells and added 1 µg/mL of recombinant LAG-3 and CD4 proteins to the culture medium [Supplementary Fig. 9A]. Immunofluorescence [Supplementary Fig. 9B] and flow cytometry [Supplementary Fig. 9C-D] confirmed that HLA-II on the surface of NPC cells predominantly binds LAG-3. These findings suggest that HLA-II in NPC cells preferentially interacts with LAG-3 on CD8⁺T cells rather than CD4.

To assess the binding affinity of HLA-DP, HLA-DQ, and HLA-DR to LAG-3 in NPC cells, recombinant proteins of each HLA class II molecule were individually co-incubated with LAG-3. Bio-layer interferometry (BLI) binding assays revealed that HLA-DP binds to LAG-3 with significantly higher affinity than HLA-DQ and HLA-DR [Fig. 5A, Supplementary Fig. 10A]. We individually overexpressed these HLA-II molecules in NPC cells and added 1 µg/mL of recombinant LAG-3 protein to the medium, simulating in vitro interactions between HLA-DP, HLA-DQ, HLA-DR, and LAG-3 [Supplementary Fig. 10B]. Results of immunofluorescence [Fig. 5B, Supplementary Fig. 10C-D] and flow cytometry [Supplementary Fig. 10E-F] showed that NPC cells expressing HLA-DP bound significantly more recombinant LAG-3 protein compared to cells overexpressing HLA-DQ or HLA-DR. Furthermore, when we simultaneously overexpressed HLA-DP, HLA-DQ, and HLA-DR in NPC cell lines, added 1 µg/mL recombinant LAG-3 protein, and incubated for 6 hours, co-immunoprecipitation (Co-IP) assays using HLA-DP, HLA-DQ, HLA-DR, or LAG-3 antibodies confirmed the immunofluorescence and flow cytometry results [Supplementary Fig. 10G-H]. Among the HLA-II molecules tested, HLA-DP exhibited the strongest affinity for LAG-3, suggesting that NPC-produced HLA-DP may facilitate T cell exhaustion by binding to LAG-3 on CD8⁺ T cells.

Fig. 5. HLA-DP Shows the Highest Binding Affinity for LAG-3 Among HLA Class II Proteins.

A Analysis of LAG-3 binding to HLA-DP. HLA-DP concentrations (top to bottom): 500, 250, 125, 61.5, and 31.25 nM; measured KD = 0.164 nM. n = 3 independent experiments with similar results. B Immunofluorescence assay showing the colocalization of LAG-3 recombinant protein with HLA-DP in NPC cell lines overexpressing HLA-DP, HLA-DQ, or HLA-DR. Scale bar, 20 μm. n = 3 independent experiments with similar results. C.NPC cell lines HONE1 and HK1 overexpressing BALF2-LR or BALF2-HR, and C666-1 cells with BALF2-HR knockout, were co-cultured with CD8⁺ T cells. Cultures were treated with either IgG or LAG-3 monoclonal antibody. Western blot analysis was performed to assess LAG-3 ubiquitination levels. n = 3 independent experiments with similar results. D The same co-culture setup as in (C) was used to perform co-immunoprecipitation (Co-IP) assays, verifying the binding between CD8 and Lck in T cells. n = 3 independent experiments with similar results. E Using the same experimental conditions, Western blot analysis was conducted to measure ZAP70 phosphorylation and TOX expression levels in CD8⁺ T cells. n = 3 independent experiments with similar results. F Immunohistochemistry detecting the correlation of BALF2 and HLA-DP in 39 BALF2-HR and 19 non-BALF2-HR NPC tissue samples before PD-1 immunotherapy. Scale bar, 50 μm. G Immunohistochemistry detecting the expression of HLA-DP in 58 NPC clinical tissue samples before PD-1 immunotherapy. The expression level of HLA-DP was found to be significantly higher in patients with ineffective treatment outcomes (PD + SD, n = 21) compared to those with effective treatment outcomes (PR + CR, n = 37) (mean expression 6.9 vs 4.0, P = 0.0001). Correlations were calculated using the parametric two-tailed Pearson correlation test for (F) and an unpaired two-tailed Student’s t test for (G). Data are presented as mean ± SD.

To further elucidate the molecular mechanism by which LAG-3 inhibits T cell function, we focused on its activation via ubiquitination, a modification known to be essential for LAG-3’s immunosuppressive activity³⁴. BALF2-HR was overexpressed in HONE1 and HK1 tumor cells or knocked out in C666-1 cells, which were then co-cultured with CD8⁺ T cells. In parallel, either IgG or a LAG-3 monoclonal antibody (mAb) was added to the cultures. Co-immunoprecipitation (Co-IP) assays revealed that overexpression of BALF2-HR significantly enhanced LAG-3 ubiquitination in CD8⁺ T cells, whereas knockout of BALF2-HR markedly reduced it. Notably, the LAG-3 mAb blocked the BALF2-HR–induced increase in LAG-3 ubiquitination [Fig. 5C]. Lck (lymphocyte-specific protein tyrosine kinase) is a critical component of the T cell receptor (TCR) signaling pathway. Previous studies have shown that LAG-3 interferes with the interaction between CD8 and Lck^35,36. Co-IP assays demonstrated that BALF2-HR overexpression in tumor cells inhibited CD8–Lck binding, thereby attenuating downstream TCR signaling. This effect was reversed by treatment with LAG-3 mAb [Fig. 5D]. These findings suggest that BALF2-HR interferes with the interaction between CD8 and Lck by promoting the ubiquitination of LAG-3, potentially leading to additional inhibition of downstream TCR signaling.

Phosphorylation of ZAP70 functions as a critical “molecular switch” in TCR signal transduction, while LAG-3 has been shown to promote the expression of the exhaustion-associated transcription factor TOX³⁷. Western blot analysis of TCR signaling components revealed that BALF2-HR overexpression reduced ZAP70 phosphorylation and upregulated TOX expression. Notably, treatment with a LAG-3 mAb restored ZAP70 phosphorylation and suppressed TOX levels [Fig. 5E]. Collectively, these findings demonstrate that BALF2-HR promotes LAG-3 activation via ubiquitination, thereby disrupting the CD8–Lck interaction and attenuating TCR signaling by decreasing ZAP70 phosphorylation. This suppression of TCR signaling lifts the inhibition of TOX, ultimately driving CD8⁺ T cell exhaustion. This mechanistic insight supports the rationale for combination immunotherapy targeting both LAG-3 and PD-1.

To examine how tumor cell HLA-II molecules regulate CD4⁺ T cell TCR signaling, we co-cultured CD4⁺ T cells with tumor cells. We observed that tumor cell HLA-II molecules could enhance the activation of CD4⁺ T cell TCR signaling [Supplementary Fig. 10I]. These results suggest that MHC-II-LAG3 interactions are not a primary mechanism of CD4⁺ T cell suppression, even though anti-LAG3 mAb increased ZAP70 signaling. Therefore, other suppressive pathways involving LAG3 may also affect CD4⁺ T cells. This suggests that, in CD4⁺ T cells, LAG-3 expression is also a crucial factor restricting TCR pathway activation, which warrants further investigation.

We conducted immunohistochemical analysis on paraffin-embedded tissue samples from 58 NPC patients who received conventional treatment combined with PD-1 immunotherapy, as shown in Fig. 1B [Supplementary Data 2], to assess the relationship between BALF2 and HLA-DP expression levels. Our findings indicate a significant positive correlation between BALF2-HR expression and HLA-DP expression in NPC patients [Fig. 5F]. Among these 58 patients, 37 exhibited effective treatment outcomes (PR + CR), while 21 did not respond to treatment (PD + SD). Further analysis revealed that HLA-DP expression was significantly higher in patients who did not respond to treatment compared to those who had effective treatment outcomes [Fig. 5G].

BALF2-HR Interaction with KPNA2 Facilitates CIITA Nuclear Translocation and Upregulation of HLA-DP

To investigate how BALF2-HR upregulates HLA-DP expression, we overexpressed either BALF2-HR or BALF2-LR in the EBV-negative NPC cell line HK1. Co-IP [Fig. 6A] assays combined with high-throughput mass spectrometry identified 318 proteins binding to BALF2-HR and 314 proteins binding to BALF2-LR, with 68 proteins uniquely interacting with BALF2-HR [Supplementary Data 5]. Analysis of gene expression profiles for these 68 proteins in the NPC clinical sample dataset GSE12452 and their correlation with HLA-DP expression revealed that KPNA2, TUBA4B, and MRPL30 were strongly correlated with HLA-DP expression [Supplementary Fig. 11A]. Subsequent Co-IP experiments confirmed that BALF2-HR has a strong binding affinity for KPNA2 [Supplementary Fig. 11B-C]. Protein-protein docking analysis using Alphafold2³⁸ and MOE³⁹ [Supplementary Fig. 11D] predicted that KPNA2 has a higher affinity for BALF2-HR than for BALF2-LR. Using truncation mutants based on Alphafold2-predicted binding sites, Co-IP experiments [Fig. 6B, Supplementary Fig. 11E] demonstrated that BALF2-HR binds the full-length KPNA2 but not its truncation mutants. Further transfection experiments with BALF2-HR and its truncation mutants into HONE1 and HK1 cells confirmed [Fig. 6C] that the 517-850aa (MT-2) region of BALF2-HR is critical for binding to KPNA2, aligning with the binding site prediction from Alphafold2.

Fig. 6. BALF2-HR Facilitates Nuclear Translocation of CIITA and Enhances HLA-DP Expression via Binding to KPNA2.

A Coomassie Brilliant Blue staining was used to detect proteins bound to BALF2-LR and BALF2-HR. HK1 cells overexpressing N-terminal Flag-tagged BALF2-LR or BALF2-HR were subjected to IP with an anti-Flag antibody. n = 3 independent experiments with similar results. B Co-IP analysis to detect interactions between KPNA2 and its truncation variants with BALF2-HR. KPNA2 and its truncation variants (DEL1: 1-111 aa, DEL2: 112-244 aa, DEL3: 245-529 aa) were transfected into HONE1 and HK1 cells overexpressing BALF2-HR. IP was carried out using an anti-Flag antibody to identify bound proteins. n = 3 independent experiments with similar results. C Co-IP analysis to detect interactions between BALF2-HR and its truncation variants with KPNA2. BALF2-HR and its truncation variants (MT1: 1-516 aa, MT2: 517-850 aa, MT3: 851-1134 aa) were transfected into HONE1 and HK1 cells. IP was performed using an anti-Flag antibody to detect bound KPNA2. n = 3 independent experiments with similar results. D Co-IP analysis to detect interactions between BALF2 and KPNA2 or CIITA in HONE1 and HK1 cells. Cells overexpressing BALF2-LR or BALF2-HR were used for IP to assess the binding of BALF2 with KPNA2 and CIITA. n = 3 independent experiments with similar results. E Western blot analysis to detect the nuclear and cytoplasmic distribution of KPNA2 and CIITA. HK1 cells overexpressing BALF2-LR or BALF2-HR and C666-1 cells with BALF2 knockout were analyzed. GAPDH was used as a cytoplasmic marker and histone H3 (H3) as a nuclear marker. n = 3 independent experiments with similar results. F HK1 cells overexpressing BALF2-LR or BALF2-HR and simultaneously knocking out KPNA2, as well as C666-1 cells with concurrent knockout of BALF2 and KPNA2, were analyzed by Western blot to examine the nuclear and cytoplasmic distribution of KPNA2 and CIITA. GAPDH was used as a cytoplasmic marker and histone H3 (H3) as a nuclear marker. n = 3 independent experiments with similar results. G HK1 cells overexpressing BALF2-LR or BALF2-HR with simultaneous KPNA2 knockout, as well as C666-1 cells with concurrent knockout of BALF2 and KPNA2, were analyzed by Western blot to assess the protein expression levels of HLA-DP, HLA-DQ, and HLA-DR. β-tubulin was used as a loading control. n = 3 independent experiments with similar results. H HK1 cells overexpressing BALF2-LR or BALF2-HR and simultaneously knocking out KPNA2, as well as C666-1 cells with concurrent knockout of BALF2 and KPNA2, were analyzed using flow cytometry to assess the expression levels of HLA-DP, HLA-DQ, and HLA-DR. Statistical results are shown in Supplementary Fig. 10M. MFI represents median fluorescence intensity. n = 3 independent experiments.

KPNA2, a member of the nuclear transport protein family, plays a critical role in the nuclear translocation of various transcription factors and is crucial for tumor formation and progression⁴⁰. The class II major histocompatibility complex transactivator (CIITA) is a key regulator of HLA-II expression⁴¹. To investigate whether BALF2-HR upregulates HLA-II by binding to KPNA2 and facilitating the nuclear translocation of CIITA, we conducted protein-protein docking analyses using Alphafold2 and MOE. These analyses indicated a low affinity between KPNA2 and CIITA alone, but the affinity significantly increased when KPNA2 was complexed with BALF2-HR [Supplementary Fig. 11F]. Co-IP experiments confirmed that BALF2-LR did not bind to KPNA2 or CIITA in HONE1-EBV and HK1-EBV cells, while BALF2-HR bound to both KPNA2 and CIITA in C666-1 cells [Supplementary Fig. 11G]. Overexpression of BALF2-HR or BALF2-LR in EBV-negative NPC cells HONE-1 and HK1 further validated that BALF2-HR binds to both KPNA2 and CIITA, but BALF2-LR does not. In addition, without BALF2-HR, KPNA2 and CIITA did not interact [Fig. 6D].

Nuclear-cytoplasmic fractionation experiments revealed that BALF2-LR, KPNA2, and CIITA were primarily located in the cytoplasm in NPC cells overexpressing BALF2-LR. However, in cells overexpressing BALF2-HR, these proteins were predominantly found in the nucleus. Knockdown of BALF2-HR in C666-1 cells, which are infected with a high-risk EBV subtype, significantly reduced the nuclear presence of both KPNA2 and CIITA [Fig. 6E, Supplementary Fig. 11H]. Immunofluorescence assays confirmed these findings [Supplementary Fig. 11I-J]. Furthermore, knockdown of KPNA2 in BALF2-HR-overexpressing NPC cells resulted in a decreased nuclear distribution of BALF2-HR and CIITA. Similar results were observed in C666-1 cells following KPNA2 knockdown [Fig. 6F, Supplementary Fig. 11K]. These results suggest that BALF2-HR facilitates CIITA nuclear translocation by binding to KPNA2.

In addition, Western blot [Fig. 6G and Supplementary Fig. 11L], flow cytometry [Fig. 6H and Supplementary Fig. 11M-N], and RT-qPCR [Supplementary Fig. 11O] analyses showed that KPNA2 knocking out significantly reversed the upregulation of HLA-DP, HLA-DQ, and HLA-DR induced by BALF2-HR. These findings identify KPNA2 as a key downstream mediator in BALF2-HR-driven upregulation of HLA-II, indicating that BALF2-HR promotes the transcription of HLA class II molecules by binding to KPNA2 and facilitating the nuclear translocation of CIITA.

Combination of LAG-3 and PD-1 Antibodies Enhances Immunotherapy Efficacy in NPC

We demonstrated that BALF2-HR promotes immune evasion and resistance to immunotherapy in NPC cells by upregulating HLA-DP, which binds to LAG-3 on T cells. To assess whether anti-LAG-3 can enhance the efficacy of anti-PD-1 treatment in NPC, we overexpressed BALF2-LR, BALF2-HR, or an empty vector in the EBV-negative NPC cell line HK1 and knocked out BALF2-HR in the EBV high-risk NPC cell line C666-1 using sgRNA, with scrambled sgRNA as a negative control. These five groups of differently treated NPC cells were subcutaneously injected into the right hind limbs of nude mice, which lack functional T cells, to establish xenograft models (40 mice per group). Seven days after injection, each group of tumor-bearing mice was divided into four subgroups (10 mice per subgroup). One subgroup received tail vein injections of PBS, while the other three subgroups received tail vein injections of T cells (5 × 10⁷ cells/mouse) stained with the live-cell fluorescent dye DIR for tracking. The T cells were prepared through co-culture with dendritic cells stimulated by tumor cell lysates to simulate antigen presentation⁴². The three T-cell-receiving subgroups were treated with IgG (50 µg/mouse, negative control), anti-PD-1 (50 µg/mouse), or a combination of anti-PD-1 (50 µg/mouse) and anti-LAG-3 (50 µg/mouse), simulating immunotherapy [Supplementary Fig. 12A].

Live animal imaging seven days after T cell infusion showed minimal T cell distribution within the xenograft tumors overexpressing BALF2-HR. Treatment with anti-PD-1 alone did not restore T cell infiltration in these tumors. However, the combination of anti-PD-1 and anti-LAG-3 significantly increased T cell accumulation at the xenograft site. Similarly, BALF2-HR knockout in C666-1 xenografts led to a notable increase in T cell infiltration. Both anti-PD-1 monotherapy and the combination of anti-PD-1 and anti-LAG-3 further enhanced T cell distribution in these xenografts. These findings suggest that BALF2-HR inhibits T cell survival in tumor tissues, whereas the combined use of LAG-3 and PD-1 antibodies enhances T cell survival in NPC tissues that overexpress BALF2-HR [Fig. 7A, B].

Fig. 7. Combination of LAG-3 and PD-1 monoclonal antibodies enhances immunotherapy efficacy in NPC.

A, B In vivo imaging to monitor the survival and distribution of T cells in xenograft tumor-bearing mice (n = 6 per group). Mice were intravenously injected with DiR-labeled T cells and treated with either anti-PD-1 or a combination of anti-PD-1 and anti-LAG-3. A Representative images. B Statistical results of the DiR fluorescence signal ratio in mice (ratio of DiR fluorescence signal in the right thigh tumor to T cell fluorescence intensity) collected at 1 h, 1 day, 3 days, and 7 days post-injection. The color bar represents the DiR fluorescence intensity of T cells in the mice. C Representative images of subcutaneous tumors in nude mice. T cells were injected seven days after tumor implantation, followed by treatment with either anti-PD-1 alone or in combination with anti-LAG-3. Left: HK1 cells, 35 days post-injection (n = 6 per group). Right: C666-1 cells, 42 days post-injection (n = 6 per group). D Flow cytometry analysis to assess the proportion of CD8⁺ T cells, cytotoxic cytokine secretion, and the proportion of exhausted CD8⁺ T cells. Single-cell suspensions were prepared from tumor tissues of mice harvested 35 days after injection of HK1 cells and 42 days after injection of C666-1 cells (n = 6 per group). Correlations were calculated using two-way ANOVA for (B, D), *p  <  0.05, **p  <  0.01, ***p  <  0.001 and ****p  <  0.0001. These experiments were derived from biological replicates. Data are presented as mean ± SD.

Furthermore, tumor growth was monitored in real-time, and mice were euthanized 5-6 weeks later to collect the tumor tissues [Fig. 7C and Supplementary Fig. 12B, C]. No significant differences in tumor volume or weight were observed between the PBS-treated groups overexpressing BALF2-LR, BALF2-HR, or BALF2-HR knockout. However, in the T cell-treated groups, BALF2-HR overexpression led to larger xenograft tumors compared to controls, while BALF2-HR knockout resulted in significantly smaller tumors. BALF2-HR overexpression conferred resistance to anti-PD-1 therapy, resulting in larger tumor volumes, whereas BALF2-HR knockout increased sensitivity to anti-PD-1, reducing tumor size. The combination of anti-PD-1 and anti-LAG-3 provided the most effective tumor suppression. Flow cytometry of single-cell suspensions from selected xenograft tissues (6 mice per group) in the T cell-treated groups revealed that BALF2-HR overexpression decreased CD8⁺ T cell numbers and increased their exhaustion. Conversely, BALF2-HR knockout increased CD8⁺ T cell numbers and reduced exhaustion. Anti-PD-1 treatment alone did not reverse BALF2-HR-mediated CD8⁺ T cell depletion and exhaustion, but the combination of anti-PD-1 and anti-LAG-3 did [Fig. 7D]. ELISA [Supplementary Fig. 12D] and RT-qPCR [Supplementary Fig. 12E, F] results further supported these findings.

Paraffin-embedded xenograft tissues from the groups described above were analyzed by IHC to confirm BALF2-HR expression efficiency. The BALF2-HR overexpression group exhibited high levels of HLA-DP expression and reduced infiltration of CD8⁺ T cells, whereas the BALF2-HR knockout group displayed low HLA-DP expression with increased CD8⁺ T cell infiltration [Supplementary Fig. 12G–I]. These findings suggest that BALF2-HR facilitates immune evasion in NPC cells in vivo and contributes to resistance against anti-PD-1 monotherapy. However, the combination of anti-LAG-3 and anti-PD-1 therapy was able to reverse BALF2-HR-mediated immune evasion, thereby enhancing the anti-tumor efficacy of anti-PD-1 treatment [Supplementary Fig. 13].

Discussion

We, along with Xu et al., identified a distinct EBV subtype prevalent among NPC patients in southern China^11,13. Xu et al. further demonstrated that high-risk mutations in the EBV gene BALF2 may contribute to the regional predominance of EBV-associated NPC. In southern China and Hong Kong, over 70% of NPC patients harbor high-risk BALF2 variants (162215 C > A, 162476 T > C, 163364 C > T)^11,43, compared to approximately 40% in Singapore⁴⁴. In contrast, these high-risk subtypes are rarely observed in non-endemic regions such as Japan⁴⁵. The BALF2 protein is an early lytic protein in EBV, with prior research primarily focusing on its role in the EBV latency-lytic cycle in lymphocytes^46,47. However, in NPC cells, EBV typically exists in a latent form, predominantly as a type II latency infection⁴⁸. BALF2 expression has been detected in the serum of NPC patients and EBV-positive NPC tissues and cell lines⁴⁹, indicating that BALF2’s main function in NPC is likely not related to initiating EBV lytic replication. Indeed, overexpression of BALF2 in NPC cells showed no impact on EBV latency or lytic replication initiation, nor did it significantly affect NPC cell proliferation. Our study is the first to identify that the BALF2-HR protein, encoded by the high-risk EBV subtype, can significantly upregulate the expression of HLA class II molecules in NPC cells.

NPC not only shows notable geographical and ethnic clustering but also presents unique clinical characteristics, such as pronounced lymphocyte infiltration in tumor tissues. This phenomenon led to NPC being termed “lymphoepithelioma”⁵⁰. Despite the extensive presence of lymphocytes, NPC is able to evade immune responses, raising questions about the underlying mechanisms and the role of Epstein-Barr virus (EBV) in this process. Our previous work demonstrated that EBV contributes to immune evasion by upregulating PD-L1 through encoding circBART2.2⁵¹, miBART11, and miBART17-3P⁴², which partially explains the high PD-L1 expression (over 85%) observed in NPC^52,53. The substantial immune cell infiltration, coupled with high PD-L1 expression, suggests NPC is a promising candidate for immunotherapy. Recent clinical advancements have underscored this potential, with phase III trials showing success in NPC treatment. As a result of these trials, the latest NPC guidelines now recommend immunotherapy as a first-line treatment for high-risk, recurrent, or refractory NPC⁷. While immunotherapy has improved outcomes and provided new hope for NPC patients, a significant number of patients remain unresponsive, suggesting issues of primary resistance and hyperprogression, indicating the involvement of immune inhibitory mechanisms beyond the PD-1/PD-L1 pathway.

Recent discoveries, enabled by single-cell sequencing, have revealed that epithelial-derived NPC cells can exhibit immune-like characteristics, including the expression of oncogenic immunoglobulins (Ig)⁵⁴ and genes associated with immune cell functions, particularly HLA class II molecules. This phenomenon has been termed the epithelial-immune dual characteristics of tumor cells²⁰. However, the link between EBV infection and these dual characteristics in NPC cells has not yet been reported. Classic HLA class II proteins, encoded by highly polymorphic loci such as HLA-DP, HLA-DQ, and HLA-DR, consist of non-covalently linked polypeptide chains (α, β)^55,56. While HLA class II molecules were traditionally thought to have similar functions, recent studies have uncovered distinct roles for different HLA class II molecules. For example, HLA-DP has been identified as a ligand for NKp44⁵⁷ and is associated with graft-versus-host disease⁵⁸. HLA-DQ influences the risk of relapse following hematopoietic stem cell transplantation⁵⁹, and HLA-DR polymorphisms are closely linked to vaccine responsiveness⁶⁰. However, NPC cells exhibit some immune cell features, including high expression of HLA class II molecules and immunoglobulins such as Igα and Igκ^20,61. Our findings establish a connection between a unique EBV subtype prevalent in southern China and the epithelial-immune dual characteristics of NPC, mediated through the upregulation of HLA class II molecules by the EBV-encoded BALF2-HR protein. Some epithelial-derived tumors also express HLA class II molecules⁶², and our study offers new insights into their functional role in tumor cells. Consequently, further investigations into other HLA-II–expressing tumors are warranted and may hold significant clinical relevance.

The classical receptor for HLA-II molecules is the inhibitory immune checkpoint LAG-3, which is highly expressed on exhausted T cells⁶³. LAG-3, also known as CD223, was identified in 1990 as a CD4 homolog, with recent single-cell sequencing data showing high expression on CD8⁺ T cells, especially upregulated in the EBV-positive NPC microenvironment⁶⁴. The immunosuppressive function of LAG-3 is not solely dependent on HLA-II molecules. Ligands such as Galectin-3 (Gal-3), fibrinogen-like protein 1 (FGL1), and liver sinusoidal endothelial cell lectin (LSECtin) provide MHC-II-independent binding pathways for LAG-3, contributing to its inhibitory effects⁶⁵. As a result, our research has demonstrated that even after the HLA-DP knockout, combination therapy with PD-1 and LAG-3 mAbs remains more effective than PD-1 monotherapy. In EBV-associated pediatric Hodgkin lymphoma, elevated expression of PD-1 and LAG-3 on tumor-infiltrating lymphocytes correlates with poorer patient survival.⁶⁶. LAG-3 maintains TOX expression on CD8⁺ T cells⁶⁴ and works in synergy with PD-1 to drive T cell exhaustion and inhibit IFN-γ-dependent anti-tumor immunity⁶⁷. LAG-3 is the third immune checkpoint pathway successfully targeted for cancer therapy. In 2022, the FDA approved the combined use of anti-PD-1 and anti-LAG-3 (Nivolumab + Relatlimab) for melanoma, and multiple clinical trials are ongoing for other tumors^65,68. Notably, a phase Ib/II clinical trial in NPC (NCT05102006) demonstrated that, compared to other solid tumors, patients with NPC exhibited improved responses to combined PD-1 and LAG-3 monoclonal antibody therapy⁶⁹. Our study found that the EBV-encoded BALF2-HR protein promotes immune evasion in NPC by upregulating HLA-DP, which binds to LAG-3 on CD8⁺ T cells, leading to T cell exhaustion. Although the BALF2-HR variant shows regional specificity, the immunological mechanisms identified in this study—such as PD-1 and LAG-3–mediated T cell exhaustion—are likely applicable to other high-risk EBV-associated NPC patients worldwide.

In addition, the HLA locus on human chromosome 6 is a critical genetic susceptibility factor for NPC, alongside the association of high-risk EBV subtypes, such as the BALF2-HR haplotype^22,70. As early as 1990, Zeng et al. established a strong connection between the HLA locus and NPC susceptibility through family linkage analysis²¹. Zeng’s team further identified several single-nucleotide polymorphisms (SNPs) within the HLA region associated with NPC risk through case-control genome-wide association studies²². More recently, Xu et al. demonstrated that the interaction between HLA susceptibility genes and high-risk EBV subtypes accounts for over 60% of NPC risk⁸. This highlights the importance of exploring how unique EBV genotypes and the genetic background of NPC-susceptible populations, particularly HLA polymorphisms, contribute to the incidence of NPC. Our findings that BALF2-HR upregulates HLA class II molecules, leading to NPC immune evasion and resistance to immunotherapy, may help explain the regional variations in NPC incidence. Investigating how genetic susceptibility (e.g., HLA loci) and environmental factors (e.g., EBV infection) interact to drive the high incidence of NPC in southern China will provide deeper insights into the etiology and pathogenesis of NPC, offering a foundation for developing targeted treatment strategies.

The specific role of BALF2-HR in NPC development is shaped by complex interactions among host genetic factors, environmental exposures, and viral evolution. This unique synergy, which includes HLA mutations and environmental factors such as nitrite consumption, is exclusive to NPC and absent in other EBV-associated cancers, highlighting BALF2’s region-specific involvement in NPC pathogenesis. Given the critical role of the BALF2-HR subtype in NPC pathogenesis and its regulatory interplay with HLA class II molecules, future studies should focus on elucidating the interaction mechanisms between high-risk EBV subtypes and host HLA polymorphisms across diverse genetic backgrounds. This will help clarify how these factors jointly contribute to immune evasion and the geographic variation in NPC incidence. Meanwhile, research should expand to include other epithelial-derived tumors that express HLA-II molecules to better understand the prevalence of BALF2-HR–associated signaling within the tumor immune microenvironment and its potential implications for cancer immunotherapy. In addition, our study found that the high expression of HLA-II molecules, mediated by BALF2-HR, can partially activate CD4⁺ T cells. Although MHC-II-LAG3 interactions themselves might not directly suppress CD4⁺ T cells, other potential suppressive mechanisms involving LAG3 could still impact CD4⁺ T cell function. The further impacts of this activation on the tumor microenvironment remain to be explored in future studies. Moreover, novel combination therapies targeting both PD-1 and LAG-3 checkpoints warrant prospective clinical evaluation in NPC patients harboring high-risk EBV subtypes, as well as in related cancers. A systematic investigation of the interactions among high-risk EBV variants, host genetic susceptibility, and immune regulatory networks will provide critical insights into NPC pathogenesis and inform precision therapeutic strategies for NPC and similar malignancies.

In conclusion, our study demonstrates that the high-risk EBV-encoded BALF2-HR protein directly interacts with KPNA2, facilitating the nuclear translocation of CIITA and thereby enhancing the expression of HLA class II molecules, such as HLA-DP. HLA-DP, in turn, interacts with LAG-3 on CD8⁺ T cells, leading to the suppression of their cytotoxic activity and inducing T cell exhaustion, which contributes to immune evasion in NPC. Furthermore, our findings indicate that combining anti-PD-1 and anti-LAG-3 therapies holds great promise for the immunotherapy of NPC patients with high-risk EBV subtypes. These insights uncover a novel molecular mechanism underlying T cell exhaustion and dysfunction in NPC, highlighting how high-risk EBV subtypes contribute to the development of epithelial-immune dual characteristics in NPC cells. This research provides a strong rationale for using combined anti-PD-1 and anti-LAG-3 immunotherapy in patients with NPC infected by high-risk EBV subtypes.

Methods

Cell lines

The nasopharyngeal carcinoma (NPC) cell lines HONE1, HK1, and the EBV-positive NPC cell line C666-1 used in this study were maintained in our laboratory at the Cancer Research Institute of Central South University. The HONE1-EBV and HK1-EBV cell lines were generously provided by Professor Tsao from the University of Hong Kong⁷¹ and Professor Li from Southern Medical University⁷². Cells were cultured in RPMI 1640 medium (Gibco, 61870036, USA) supplemented with 10% fetal bovine serum (Vazyme Biotech Co., Ltd, F103-01, China) and 1% penicillin-streptomycin (NCM Biotech, C100C5, China) at 37 °C with 5% CO₂.

Clinical NPC Samples and recruited volunteers

The study involved three groups of clinical NPC tissue samples. The first group comprised 128 NPC tissues used for RNA extraction, quantitative real-time PCR (RT-qPCR), and subsequent Sanger sequencing to identify BALF2 subtypes (see Supplementary Data 1). The second group included 25 fresh NPC tissues prepared for single-cell suspensions and flow cytometry (see Supplementary Data 3). The third group consisted of paraffin-embedded sections from 58 NPC patients treated with PD-1 antibody, which were utilized for immunohistochemistry (IHC) and Sanger sequencing to identify BALF2 subtypes (see Supplementary Data 4). Human primary PBMCs were obtained from healthy volunteers. All uses of human material were confirmed by histopathological examination, and obtained with approval from The IRB of School of Basic Medical Sciences, Central South University or The IRB of Cancer Research Institute, Central South University and informed consents are obtained from patients and recruited volunteers. In addition, the study utilized the GSE150825 database to analyze the differential expression of HLA-II molecules in four EBV-negative and six EBV-positive NPC tissues, and the GSE194333 database to compare HLA-II molecule expression between NP69 and C666-1 cell lines.

Immunohistochemistry (IHC)

IHC was performed using the Elivision plus (mouse/rabbit) IHC Kit (Fuzhou Maixin Biotech. Co., Ltd, KIT-9902, China). Tissue sections were deparaffinized, rehydrated, and subjected to antigen retrieval by boiling in EDTA (Beyotime, P0084, China) buffer for 15 min. After washing with PBS (Servicebio, G0002, China), endogenous peroxidase activity was blocked, and the sections were outlined with a hydrophobic pen. The sections were then blocked with goat serum, incubated overnight with primary antibodies, and subsequently incubated with a secondary antibody at room temperature for 10 minutes. Following this, the sections were treated with a biotin-peroxidase solution, developed with a DAB solution, counterstained with hematoxylin, dehydrated, and then mounted. Images were captured using an Olympus BX51 fluorescence microscope (Olympus, Japan). Staining intensity and positive rates were assessed independently by two experienced pathologists. Staining intensity was scored as follows: 0 = negative, 1 = weak, 2 = moderate, and 3 = strong. Positive rates were scored as 0 = negative, 1 = 1–25%, 2 = 26–50%, 3 = 51–75%, and 4 = 76–100%.

Vectors, sgRNA, and cell transfection

The BALF2-LR and BALF2-HR overexpression vectors, along with sgBALF2, sgHLA-DP, sgHLA-DQ, and sgHLA-DR, were obtained from Beijing Tsingke Biotech Co., Ltd. The overexpression vectors for HLA-DP, HLA-DQ, and HLA-DR were constructed using the Pc3.1 vector. Transfection of the overexpression vectors, control plasmids, sgRNA, and negative controls into NPC cell lines was performed using the Neofect DNA transfection reagent (Neofect Biotech, Beijing, China). The sequence information for BALF2-LR and BALF2-HR is provided in Supplementary Data 6.

RNA Extraction and RT-qPCR

RNA was extracted from NPC cell lines using TRIzol reagent (Invitrogen™, 15596026CN, USA) and converted to cDNA with the HiScript cDNA Synthesis Kit (Vazyme, R223, China). Quantitative real-time PCR (RT-qPCR) was conducted in triplicate using SYBR Green (Vazyme, Q312, China), with GAPDH as the internal reference gene. Primer sequences are detailed in Supplementary Data 6.

Western blot

Total proteins were extracted from NPC cells using RIPA buffer (Shanghai Epizyme, PC101, China) supplemented with protease inhibitors (Shanghai  Epizyme, GRF101, China) and phosphatase inhibitors (Beyotime Biotechnology, Shanghai, China). Proteins were separated by 10% ExpressCast PAGE (NCM Biotech, P2012, China) and transferred to a PVDF membrane (Vazyme, E801, China). Membranes were blocked and incubated overnight with primary antibodies, followed by a 1 h incubation with secondary antibodies at room temperature. Protein bands were visualized using ECL detection reagent (Coolaber, SL1350, China), and images were captured with the Mini Chemiluminescent Imaging and Analysis System. Details of the antibodies used are provided in Supplementary Data 7.

Mass spectrometry analysis

BALF2-LR and BALF2-HR were overexpressed in the EBV-negative HK1 cell line, followed by Co-IP using an anti-BALF2 antibody. The precipitated proteins were analyzed via SDS-PAGE and high-throughput mass spectrometry (Jingjie PTM BioLab Co., Ltd.) to identify interacting proteins. The mass spectrometry results are provided in Supplementary Data 5.

Co-Immunoprecipitation (Co-IP)

Transfected cells were collected 48 h post-transfection and lysed using IP lysis buffer (50 mM Tris [pH 7.4], 150 mM NaCl, 1% NP-40, 0.1% SDS) supplemented with protease inhibitors (Keygen, KGB5106-1, China). A portion of the cell lysate was saved as an input control, while the remaining lysate was incubated with primary antibodies at room temperature for 2 h. The lysates were then incubated with A/G beads (Selleck, B23202, China) at 4 °C overnight. After washing the beads, proteins were eluted by boiling in 95 °C SDS sample buffer and analyzed by Western blotting.

Bio-layer interferometry (BLI)

Bio-layer interferometry (BLI) was performed at 25 °C using the GatorPrime biosensor system (GatorBio) equipped with ProA probes. Fc-tagged LAG-3 (10 µg/ml) was immobilized onto the biosensors for 1 min, followed by a 2 min wash in PBS to establish the baseline. The biosensors were then immersed in wells containing varying concentrations of HLA-DP, HLA-DQ, or HLA-DR for 5 min to assess binding. Compound recombinant proteins were purchased from MedChemExpress (Monmouth Junction, NJ, USA). This was followed by a 10-minute dissociation phase in PBS buffer. Data were analyzed using GraphPad Prism 9, applying a standard 1:1 binding model. Each sample was tested in two independent experiments.

Immunofluorescence

Transfected cells were plated in 24-well plates and fixed with 4% paraformaldehyde (Servicebio, G1101, China) for 15 minutes. After blocking with 5% BSA (Servicebio, GC305006, China) for 30 minutes, cells were incubated overnight with primary antibodies, followed by a 1 h incubation with fluorescent secondary antibodies at room temperature. Nuclei were stained with DAPI (Invitrogen, D1306, USA) for 10 min. Images were captured using a confocal microscope (Ultra-View Vox, Perkin-Elmer, Waltham, MA, USA).

Flow cytometry

Transfected tumor cells and CD8⁺ T cells were co-cultured in 6-well plates at a 1:10 ratio for 24 h. Tumor cell apoptosis was evaluated by collecting and digesting the co-cultured cells, followed by staining with an apoptosis detection kit (BD Pharmingen) and analysis using flow cytometry (BD LSRFortessa). To measure T cell cytokines, co-cultured CD8⁺ T cells were treated with 10 µg/mL BFA (Selleck, S7046, China) for 6 h, then stained with CD3 and CD8 antibodies, permeabilized, and further stained with fluorescently labeled cytokine antibodies. T cell exhaustion was assessed by staining co-cultured CD8⁺ T cells with antibodies targeting CD3, CD8, PD-1, and TIM-3. Data analysis was conducted using FlowJo v10 software (Treestar). The flow gating strategy is shown in Supplementary Figs. 4 and 13.

Enzyme-linked immunosorbent assay (ELISA)

Supernatants from co-cultured tumor and CD8⁺ T cells, as well as single-cell suspensions from tumors of nude mice, were collected after 24 h and analyzed for cytokines using ELISA kits (Abcam, USA). Cytokine levels of IFN-γ, GZMB, and TNF were measured with a microplate reader equipped with SoftMax Pro 7 (Molecular Devices).

Crystal violet staining

Tumor cells cultured in 6-well plates were fixed with 4% paraformaldehyde for 15 minutes and stained with 0.1% Crystal Violet solution (Coolaber, SL7081, China). The stained cells were scanned using a scanner, and then the stain was dissolved in 1 mL of DMSO (Coolaber, CD4731C, China). Absorbance was measured at 570 nm in 96-well plates.

Isolation of human primary T cells and DC

Peripheral blood from healthy volunteers was diluted 1:1 with saline and subjected to Ficoll-Hypaque (Cytiva, 17144003, USA) density gradient centrifugation to isolate peripheral blood mononuclear cells (PBMCs). Monocytes were further enriched using CD14⁺ magnetic microbeads (Miltenyi Biotec, 130-050-201, Germany). The isolated monocytes were seeded at a density of 1.5 × 10⁶ cells/ml in 24-well plates and cultured for five days in complete RPMI 1640 medium supplemented with GM-CSF (100 ng/ml) and IL-4 (50 ng/ml). The purity of the resulting dendritic cells (DCs) was evaluated by flow cytometry.

Primary CD4⁺ T cells (Miltenyi Biotec, 130-045-101, Germany) and CD8⁺T cells (Miltenyi Biotec, 130-045-201, Germany) were isolated from PBMCs using magnetic separation with CD4/CD8 microbeads. To generate NPC-reactive T cells, primary CD8⁺T cells from HLA-A–matched donors and primary CD4⁺ T cells from HLA-DR–matched donors were co-cultured with autologous dendritic cells and irradiated NPC cells for one week^42,51,73,74. T cells were then activated and expanded using the T Cell Activation/Expansion Kit (Miltenyi Biotec, 130-091-441, Germany) and cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum (Umedium, 3022 A, China), IL-2 (Novoprotein, Shanghai, C013, China), IL-7 (Novoprotein, Shanghai, C013, China), and IL-15 (Novoprotein, Shanghai, CX47, China) for 7–10 days. The purity of CD4⁺ and CD8⁺ T cell populations was verified by flow cytometry. All primary T cells used in the co-culture experiments were provided by 3 independent volunteers.

Animal experiments

Two hundred female Balb/c Nude mice (4–6 weeks old, female, N-0007) were purchased from Silaikejingda Experimental Animal Co. (Hunan, China) and housed at the Animal Experiment Center of Hunan Cancer Hospital, in a Non-Specific Pathogen-Free (NSPF) environment. The experimental and control groups were housed separately to avoid any cross-contamination or bias. After preparation, HK1 cells transfected with BALF2-LR or BALF2-HR, and C666-1 cells transfected with sgBALF2, were subcutaneously injected into the right thigh inguinal region of the mice. Tumor formation was visually observed seven days after the initial treatment. Activated T cells were administered via tail vein injection. Some groups also received PD-1 antibody treatment or a combination of PD-1 and LAG-3 antibodies. Tumor volume was measured weekly. After 35 days for the HK1 cell lines and 42 days for the C666-1 cell lines, the mice were euthanized by cervical dislocation. Tumor tissues were excised, weighed, fixed, paraffin-embedded, and sectioned for immunohistochemistry. Single-cell suspensions from tumor tissues were prepared for flow cytometry to assess T cell function and exhaustion. Cytokine expression levels in tumor tissues were analyzed using ELISA and RT-qPCR.

To evaluate the role of BALF2-HR in promoting tumor immune evasion in vivo and the therapeutic efficacy of PD-1 and LAG-3 antibodies, 200 mice were randomized into 5 groups (40 mice per group). These groups included those injected with HK1 cell lines (Vector control, BALF2-LR overexpression, and BALF2-HR overexpression) and those injected with C666-1 cell lines (Negative control and BALF2 knockout). Additionally, the groups were divided based on treatment (10 mice per group): one groups received activated T cell therapy seven days post-injection, one groups received activated T cell therapy with PD-1 antibody, and one groups received activated T cell therapy with the combination of PD-1 and LAG-3 antibodies. All animal experiments were approved by the Experimental Animal Ethics Committee of Central South University. The maximum tumor size permitted by the institutional ethical board was no more than 1500 mm³ and/or not exceeding 10% of the animal’s body weight. All experiments strictly adhered to these limits. Tumor size was measured using digital calipers, and tumor volume was calculated using the following formula: Tumor volume = 0.52 × Length × Width². Tumor measurements were performed two to three times per week. Animals were monitored daily for general health status, body weight, behavior, and signs of discomfort or distress. Criteria for early termination included: tumor volume exceeding the approved ethical limit, tumor ulceration or necrosis, rapid tumor growth affecting mobility or normal behavior, body weight loss greater than 15–20%, or any signs of severe distress such as lethargy, impaired movement, or abnormal posture.

Live-cell imaging of T cells

In vivo distribution of activated T cells in mice was evaluated using an in vivo imaging system (Bruker, USA). T cells were labeled with the deep red live-cell fluorescent dye DiR (ThermoFisher, D12731, China) prior to injection via the tail vein. Distribution was monitored at 1 h, 1 day, 3 days, and 7 days post-injection under 3% isoflurane anesthesia (RWD, R510-22-16, China).

Statistical analysis

All statistical analyses were conducted using GraphPad Prism 8 software (GraphPad, USA). The Student’s t test was employed to assess significant differences between two groups, while ANOVA was used for multiple comparisons. All data are presented as mean ± standard deviation (SD). A p-value of < 0.05 was considered statistically significant.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Conceptualization and methodology: Z.Z., J.G., and W.X. Investigation: Y.M., Q.W., L.S., Q.Y., Q.H., P.R., Z.C., D.W., H.Q., P.C., Z.G., F.W., B.X., M.Z., M.T., G.L., and C.G. Formal analyses: Y.M., Q.W., L.S., Q.Y., and Z.G. Writing-original draft: Y.M. and Q.W. Writing-review & editing: Z.Z., J.G., and W.X. All authors read and approved the final manuscript.

Peer review

Peer review information

Nature Communications thanks Said Dermime, who co-reviewed with Queenie Fernandes, Takumi Kumai and Hiroyoshi Nishikawa for their contribution to the peer review of this work. A peer review file is available.

Data availability

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD074503. The RNA-seq raw sequencing data have been deposited in the Gene Expression Omnibus (GEO) under accession number GSE319431. All other data supporting the findings of this study are available from the corresponding author upon reasonable request. Source data are provided in this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-70725-4.