Identification of KDM5B as a crucial epigenetic factor in EBV-induced epithelial tumor progression

To explore the histone modification mechanisms underlying EBV-induced epithelial tumor progression, we initially analyzed single-cell transcriptome data of EBVhigh and EBVlow malignant epithelial cells from 10 NPC tissues (Fig. 1a).18 Differential gene expression analysis identified 40 histone modification enzymes that were upregulated in the EBVhigh NPC cells as compared to the EBVlow cells (Fig. 1b, c and Supplementary Table 1), referencing with an online database of histone modifiers.19 Next, we established an in vitro EBV infection model using GFP-integrated EBV construct to infect two EBV-negative NPC cell lines (S26 and HK-1) and one EBV-negative GC cell line AGS (see Methods). EBV infected or positive (EBV-P) cells were sorted according to their GFP expression levels using flow cytometry, with un-infected cells serving as negative control (NC; Fig. 1a; see Methods). Subsequent transcriptome sequencing analysis of these cell groups pinpointed 11, 35, and 11 upregulated histone modification factors following EBV infection in S26, HK-1, and AGS cells, respectively (Fig. 1d and Supplementary Table 1). Integrative analysis of these histone modification factors identified KDM5B and KDM4B to be consistently upregulated in both biopsy samples and EBV infection models, with KDM5B exhibiting the most significant increase (Fig. 1e and Supplementary Table 1). Western blot analysis corroborated these findings, showing increased protein levels of KDM5B in EBV-positive NPC (HONE1, CNE2, and HK-1) and GC (AGS) cell lines compared to their parental EBV-negative cells (Fig. 1f). Notably, we observed a consistent rise in KDM5B protein expression in both the originally EBV-negative NPC (S26 and HK-1) and GC cell lines (AGS) following EBV infection (Fig. 1g). These findings strongly suggest the pivotal role of KDM5B, whose upregulation appears to be driven by EBV infection, in the progression of EBV-induced epithelial tumors.

Fig. 1

KDM5B is highly expressed and associated with poor survival in EBV-associated epithelial tumors. a Schematic diagram for the screening of histone modifier KDM5B contributing to EBV-associated epithelial tumors. FACS, Fluorescence Activated Cell Sorter. b Histogram showing top 10 upregulated histone modifiers in EBVhigh NPC cell cluster, ranked according to their fold change in relative to the EBVlow cluster in scRNA-seq cohort (n = 10). c Violin plots depicting the normalized expression of KDM5B in EBVhigh and EBVlow clusters from the NPC scRNA-seq data described in (b). d Heatmap illustrating top 8 upregulated histone modifiers in S26, HK-1, and AGS cells post-EBV infection. NC1/2, two replicate negative control groups; EBV-P1/2, two replicate EBV-GFP positive groups. Filled colors from blue to red represent scaled expression levels (normalized −log10P values) from low to high. P-values were calculated by one-sided hypergeometric test and adjusted for multiple comparisons. e Venn diagram showing the overlap between the datasets described in (b) and (d). Numbers are the upregulated genes. f Western blot assay quantifying the protein levels of KDM5B and EBNA1 in EBV-positive NPC (HONE1, CNE2, and HK-1) and GC (AGS) cell lines compared to their parental EBV-negative cells. GAPDH serves as a loading control. g Western blot assay demonstrating the protein levels of KDM5B, EBNA1, and Zta in NPC (S26 and HK-1) and GC (AGS) cell lines pre- and post-EBV infection. Tubulin is used as control. h Representative images of IHC staining for high (bottom) and low (top) protein levels of KDM5B in NPC samples (n = 120). i Survival analysis correlating KDM5B protein expression with overall (OS; left) and disease-free survival (DFS; right) in NPC patients detailed in (h). j Distribution of KDM5B expression levels (left) and metastasis rates (right) in NPC samples with EBV copy numbers greater than 1000 or not. KDM5B expression was determined using IHC staining scores. Comparison was done using Student’s t-test or one-way ANOVA with Bonferroni’s post-test. Data were presented as mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. SD, standard deviation. Scale bar, 50 μm

We next evaluated the clinical relevance of KDM5B, utilizing immunohistochemistry (IHC) staining assays on specimens from 120 NPC patients (Fig. 1h). Our survival analysis demonstrated that elevated KDM5B expression was significantly correlated with both reduced overall survival (OS) and disease-free survival (DFS) in NPC patients (Fig. 1i). Moreover, upregulated KDM5B expression exhibited a strong association with metastasis in NPC patients (Supplementary Tables 2–3). Importantly, patients with high EBV DNA levels (>1000 copies/ml in plasma) demonstrated significantly higher KDM5B expression and an increased rate of metastasis compared to individuals with lower EBV levels (Fig. 1j). Taken together, these findings corroborate the importance of KDM5B in the progression of EBV-related epithelial tumors.

EBNA1 cooperatively interacts with CEBPB to promote KDM5B transcription

Considering the typical latency of EBV within EBV-related epithelial tumors,5 we investigated how EBV infection influences KDM5B expression in these malignancies. We first examined the correlation of expression between KDM5B and all EBV genes, utilizing transcriptome data from 113 NPC tumor biopsy samples. Notably, we observed a unique positive correlation between KDM5B and EBNA1 in NPC tumor samples (Supplementary Fig. 1a), collaborating with the concurrent upregulation of KDM5B and EBNA1 in EBV-positive or EBV-infected NPC and GC cells (Fig. 1f, g). Next, we separately introduced the EBNA1, together with other two classical EBV latent genes LMP1 and LMP2A, into EBV-positive NPC (HONE1-EBV, C666-1) and GC (AGS-EBV) cells. Real-time quantitative PCR (RT-qPCR) analysis demonstrated that EBNA1 overexpression significantly increased KDM5B expression at both mRNA and protein levels (Fig. 2a, b and Supplementary Fig. 1b), a pattern not observed in cells with LMP1 and LMP2A overexpression (Supplementary Fig. 1c, d). By contrast, EBNA1 knockdown led to a notable decrease in KDM5B expression (Fig. 2c, d and Supplementary Fig. 1e). These observations strongly suggest a specific regulatory relationship between EBNA1 and KDM5B. Considering EBNA1’s role as a transcriptional activator in EBV-associated tumors,20 we conducted luciferase reporter assay, revealing that EBNA1 overexpression significantly enhanced KDM5B transcription activity in NPC and GC cells, while its knockdown resulted in reduced transcription activity (Fig. 2e, f and Supplementary Fig. 1f, g). These findings underscore EBNA1’s regulatory effect on KDM5B transcription, highlighting a specific epigenetic mechanism within EBV-induced epithelial tumor progression.

Fig. 2

EBNA1 interacts with CEBPB to promote KDM5B expression. a RT-qPCR assay quantifying the expression of EBNA1 and KDM5B in EBV positive NPC (HONE1-EBV) and GC (AGS-EBV) cells infected with EBNA1-overexpressing or control vectors. b Western blot assay evaluating the protein levels of KDM5B and EBNA1 in cells described in (a), using GAPDH as a control. c, RT-qPCR analysis showing the expressions of EBNA1 and KDM5B expression in HONE1-EBV and AGS-EBV cells transfected with siRNA against EBNA1 (si-EBNA1#1 and si-EBNA1#2) or negative control (si-NC). d Western blot assay detecting the protein levels of KDM5B and EBNA1 in cells described in (c). GAPDH is used as control. e, f Luciferase assays measuring the transcriptional activity of KDM5B in HONE1-EBV and AGS-EBV cells transfected with EBNA1 or empty vector (e) and EBNA1 siRNAs or control siRNA (f), respectively. g Illustration of putative transcription factor binding sites (BS#1, BS#2, and BS#3) in the KDM5B promoter. h ChIP-qPCR analysis for BS#1, BS#2 and BS#3 at the KDM5B promotor following EBNA1 overexpression in HONE1-EBV and AGS-EBV cells. i Luciferase assays evaluating the KDM5B transcription activity in HONE1-EBV and AGS-EBV cells transfected with either EBNA1 overexpression constructs or empty vector, alongside KDM5B-Luc or mutated transcription factor binding site constructs (CEBPBmt#1, CEBPBmt#2, JUNDmt#3). j Western blot showing EBNA1 co-immunoprecipitation with CEBPB in 293 T cells. k Luciferase assays assessing the KDM5B transcription activity in HONE1-EBV and AGS-EBV cells transfected with the combinations of vector and siRNAs targeting CEBPB or control (si-NC), in the presence or absence of EBNA1. Statistical analysis was conducted using Student’s t-test for two groups and one-way ANOVA followed by Dunnett’s post hoc test for more than two groups, with data presented as mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. ns, no significance. SD, standard deviation

To uncover the regulatory mechanisms of EBNA1 on KDM5B transcription, we examined the presence of EBNA1 binding motif within KDM5B promoter region, which spans −2 kb to +2 kb distance to the transcription start site (TSS). However, no identifiable EBNA1 binding motif was observed, suggesting an indirect regulation role of EBNA1 on KDM5B transcription. Given EBNA1’s known capacity to interact with various cofactors for DNA binding and gene transcription regulation,21 we hypothesized that EBNA1 could facilitate KDM5B transcription through interactions with other transcription factors (TFs). By leveraging the hTFtarget database, we identified 188 TFs with potential binding affinity to the KDM5B promoter. Subsequent correlation analysis using the scRNA-seq data from 10 NPC samples and bulk-seq data from 113 NPC samples identified 14 TFs positively correlated with KDM5B expression, with CEBPB, JUND, and JUN showing particularly strong correlations (Fig. 2g and Supplementary Fig. 1h). Chromatin immunoprecipitation combined with quantitative PCR (ChIP-qPCR) analysis demonstrated that EBNA1 binds to a specific region (−700 bp/−400 bp upstream of the TSS) in the KDM5B promoter in both HONE1-EBV and AGS-EBV cells (Fig. 2h), which encompasses binding sites (BS) for CEBPB#1, CEBPB#2 and JUND#3 (Fig. 2g, h). Luciferase reporter assays indicated that mutations at the CEBPB#2 binding site specifically abrogated EBNA1’s ability to activate KDM5B transcription (Fig. 2i), suggesting a pivotal role of CEBPB in this regulatory mechanism. This was further supported by co-immunoprecipitation assays, confirming the interaction between EBNA1 and CEBPB (Fig. 2j and Supplementary Fig. 1i). Moreover, CEBPB knockdown not only decreased KDM5B transcription but also diminished the EBNA1-induced KDM5B upregulation (Fig. 2k). Additionally, overexpression of EBNA1 mutant without the functional DNA-binding domain (DBD) still increased KDM5B expression, with effect comparable to the intact EBNA1 (Supplementary Fig. 1j), which further supports the indirect regulation of EBNA1 on KDM5B transcription. Collectively, these findings elucidate a mechanism by which EBNA1 activates KDM5B transcription through a cooperative interaction with CEBPB.

To further investigate whether EBNA1 regulation of KDM5B is dependent on the presence of EBV, we first performed ChIP-qPCR experiments with EBNA1 antibody in EBNA1-negative wild-type HONE1 and AGS cells, revealing no significant enrichment of EBNA1 on the CEBPB#1, CEBPB#2 and JUND#3 binding sites in KDM5B promoter (Supplementary Fig. 1k). Moreover, EBNA1 overexpression did not affect KDM5B expression in EBV-negative cells (HK-1, AGS and 293 T; Supplementary Fig. 1l, m). Taken together, these observations suggest that EBNA1-mediated regulation of KDM5B is EBV-dependent and likely involves other EBV-encoded factors.

BZLF1 directly activates KDM5B transcription via binding to its ZREs

Considering the role of EBV lytic cycle in tumor progression,22 we next investigated the potential activation of KDM5B during EBV lytic replication, through treating EBV-positive cell lines with phorbol 12-myristate 13-acetate (TPA) and sodium butyrate (NaB), reagents known to induce EBV lytic reactivation.23 RT-qPCR confirmed an increase in lytic gene expression (BZLF1, BRLF1, BMRF1, and BcLF1) in EBV-positive NPC and GC cells treated with TPA and NaB compared to control group (Fig. 3a and Supplementary Fig. 2a). Notably, a significant upregulation of KDM5B was detected in these cells (Fig. 3a and Supplementary Fig. 2a, b), indicating that EBV lytic activation promotes KDM5B transcription.

Fig. 3

BZLF1 activates KDM5B expression via direct binding with the ZREs on its promoter. a RT-qPCR analysis evaluating the mRNA expressions of BZLF1, BRLF1, BMRF1, BcLF1, and KDM5B in HONE1-EBV and AGS-EBV cells treated with TPA + NaB, TPA + NAB + PAA, or vehicle control. b RT-qPCR analysis showing KDM5B expression in EBV negative NPC (HK-1) and GC (AGS) cells transfected with BZLF1-, BRLF1-overexpressing or empty vectors. c Luciferase assays quantifying the KDM5B transcription activity in HK-1 and AGS cells transfected with BZLF1-, BRLF1-overexpressing or empty vectors. d Western blot assay measuring the protein level of KDM5B in HK-1, AGS, and 293 T cells transfected with BZLF1-overexpressing or empty vectors. e Schematic illustration of BZLF1 response elements (ZRE1 and ZRE2) at KDM5B promotor (upper) and their corresponding mutants (ZRE1mt, ZRE2mt, ZRE1/2 mt, bottom) employed in ChIP-qPCR and luciferase assay in (f) and (g). f ChIP-qPCR analyzing the binding of BZLF1 on ZRE1 and ZRE2 at KDM5B promotor, before and after BZLF1 overexpression in HONE1-EBV and AGS-EBV cells. g Luciferase assays evaluating the KDM5B activity in HONE1-EBV and AGS-EBV cells transfected with either KDM5B-Luc or mutant reporter plasmids corresponding to ZRE binding sites (ZRE1mt, ZRE2mt, ZRE1/2mt), following vector or BZLF1 overexpression. Statistical analysis is performed using Student’s t-test for two groups and one-way ANOVA followed by Dunnett’s post hoc test or Sidak’s post hoc test for more than two groups. Results are presented as the mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. ns, no significance. SD, standard deviation

EBV lytic cycle includes the expression of early lytic genes (immediate early and early), DNA replication, expression of late lytic genes, and finally, virus assembly and release.4 To pinpoint the specific phase of lytic replication responsible for the surge in KDM5B expression, we introduced phosphonoacetic acid (PAA), an inhibitor of EBV DNA replication, into the treatment regimen. As expected, PAA treatment led to a notable decrease in the expression of the late lytic gene BcLF1 (Fig. 3a and Supplementary Fig. 2a), highlighting the essential role of the DNA replication phase in the transcription of late lytic genes.4 Intriguingly, despite PAA treatment, the elevated expression levels of KDM5B and early/intermediate lytic genes initially induced by TPA and NaB remained unaffected (Fig. 3a and Supplementary Fig. 2a). This suggests that KDM5B upregulation is specifically associated with the activation of early lytic genes during EBV reactivation, independent of the DNA replication phase.

BZLF1 (encoding Zta) and BRLF1 (encoding Rta) are immediate early lytic genes pivotal for initiating the lytic gene expression cascade by stimulating lytic EBV promoters.4 Given the undetectable expression of these two lytic genes in most NPC samples, no correlation was observed between BZLF1 or BRLF1 and KDM5B (Supplementary Fig. 1a). We next overexpressed BZLF1 and BRLF1 in NPC and GC cells to assess the impact of these genes on KDM5B activation. RT-qPCR and luciferase reporter assays demonstrated that BZLF1, but not BRLF1, significantly enhanced KDM5B expression in EBV-negative cells (Fig. 3b, c). This effect of BZLF1 was consistently observed in EBV-positive cells, where BZLF1 overexpression led to a more than 10–15 folds increase in KDM5B expression, while BRLF1 overexpression resulted in only a slight increase compared to control cells (Supplementary Fig. 2c, d). Additionally, western blot assay confirmed that BZLF1 overexpression increased the protein level of KDM5B in EBV-negative HK-1 and AGS cells, as well as 293 T cells (Fig. 3d). These observations collectively underscore BZLF1 as a crucial modulator of KDM5B expression during EBV lytic replication.

To probe how BZLF1, or Zta, regulates KDM5B expression, we examined potential Zta-response elements (ZREs), where BZLF1 binds to regulate gene expression,22 within the KDM5B promoter. Intriguingly, we identified two ZREs located at −614 ~ −621 bp (ZRE1) and −940 ~ −947 bp (ZRE2) regions upstream of the TSS in KDM5B (Fig. 3e). To verify the direct interaction between BZLF1 and the KDM5B promoter, we conducted ChIP-qPCR assay with BZLF1-specific antibodies in HONE1-EBV and AGS-EBV cell lines. This assay revealed significant binding of BZLF1 to both the ZRE1 and ZRE2 regions (Fig. 3f), indicating a direct regulatory relationship. Moreover, luciferase assays revealed that individual mutations in ZRE1 (ZRE1mt) or ZRE2 (ZRE2mt) partially reduced the BZLF1-induced activation of KDM5B transcription, whereas simultaneous mutations in both ZREs regions (ZRE1/2mt) completely abrogated this effect (Fig. 3g and Supplementary Fig. 2e). Together with the KDM5B upregulation by BZLF1 observed in EBV-negative cells (Fig. 3b, c), these findings establish that BZLF1, an immediate early EBV lytic gene, directly stimulates KDM5B expression by binding to the ZREs within the KDM5B promoter, independent of EBV infection status.

Given the proximity of the EBNA1 and BZLF1 binding sites on the KDM5B promoter (only 55 bp apart), we wondered whether the regulatory effects of EBNA1 and BZLF1 on KDM5B are independent. To test this, we conducted luciferase assays to evaluate the impact of BZLF1 and EBNA1 on the full-length KDM5B promoter, as well as various mutant constructs, including CEBPBmt#1, CEBPBmt#2, JUND#3, ZRE1mt, ZRE2mt, and ZRE1/2mt, in HONE1-EBV and AGS-EBV cell lines. Our results showed that EBNA1 overexpression enhanced the transcriptional activity of the KDM5B promoter even in the presence of ZRE mutations (Supplementary Fig. 2f). Similarly, BZLF1 overexpression increased the transcriptional activity of the KDM5B promoter containing mutations in the CEBPB and JUND binding motifs (Supplementary Fig. 2g). These findings strongly suggest that the transcriptional activation of KDM5B by EBNA1 and BZLF1 occurs independently.

KDM5B plays an oncogenic role in EBV-associated epithelial tumors

To investigate the role of KDM5B in EBV-associated epithelial tumors, we performed shRNA-mediated knockdown of KDM5B in EBV-positive NPC and GC cells (Fig. 4a). Cell growth and colony formation assays illustrated a significant decline in cell proliferation upon KDM5B knockdown (Fig. 4b and Supplementary Fig. 3a). Furthermore, sphere formation assays demonstrated a significant reduction in tumorigenic capacity of KDM5B-knockdown cells (Fig. 4c and Supplementary Fig. 3b), accompanied by their decreased expression of key stemness markers24 (SOX2, NANOG, OCT4, ABCG2, and BMI-1; Supplementary Fig. 3c). In contrast, KDM5B overexpression in tumor cells led to enhanced cell proliferation, colony formation, and sphere formation capacities (Fig. 4d, e and Supplementary Fig. 3d–f), alongside with increased levels of stemness markers (Supplementary Fig. 3g). These findings collectively underscore the essential role of KDM5B in promoting the oncogenic features of EBV-related epithelial tumors.

Fig. 4

Oncogenic role of KDM5B in EBV-associated epithelia tumor cells. a Western blot assay confirming the knockdown efficiency of KDM5B in HONE1-EBV, CNE2-EBV, and AGS-EBV cells infected with lentivirus expressing KDM5B shRNAs (sh-KDM5B#1 and sh-KDM5B#2) or control shRNA (sh-Luci). ACTIN serves as a loading control. b Colony formation ability of cells described in (a), with statistics shown at the right. c Sphere formation ability of cells described in (a). d Colony formation results for HONE1-EBV, CNE2-EBV, and AGS-EBV cells infected with lentivirus expressing KDM5B or empty vector, along with corresponding statistics shown at the right. e Sphere formation ability of cells described in (d). f Growth curve of xenograft tumors derived from CNE2-EBV and AGS-EBV cells stably expressing KDM5B shRNAs (sh-KDM5B#1 and sh-KDM5B#2) or control shRNA (sh-Luci). Tumor size (g) and weight (h) for the xenograft tumors excised from (f). i Growth curve of xenograft tumors from CNE2-EBV and AGS-EBV cells stably expressing KDM5B or empty vector. Tumor size (j) and weight (k) for the xenograft tumors excised from (i). l Representative images for IHC staining of Ki-67 in xenograft tumors from CNE2-EBV and AGS-EBV cells with either KDM5B overexpression or control, with statistics summarized at the right. Statistical analysis is conducted using Student’s t-test for two groups and one-way ANOVA followed by Dunnett’s post hoc test for more than two groups, with data are presented as the mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. SD, standard deviation. Scale bar, 50 μm

We next explored the tumorigenic function of KDM5B with in vivo mouse model, utilizing xenografts implanted with EBV-positive epithelial tumor cells that had been genetically modified to either stably knock down or overexpress KDM5B. Notably, xenografts derived from KDM5B-knockdown cells exhibited significantly reduced growth, as evidenced by decreased tumor volume and weight (Fig. 4f–h), and corroborated by diminished Ki-67 staining, indicative of low cell proliferation (Supplementary Fig. 3h). Conversely, xenografts from cells overexpressing KDM5B showed enhanced tumorigenic capacity and a higher proportion of Ki-67 positive cells (Fig. 4i–l). This in vivo evidence further emphasizes the pivotal oncogenic function of KDM5B in the progression of EBV-positive epithelial tumors.

KDM5B suppresses PLK2 expression through H3K4me3 demethylation

To investigate the molecular mechanisms by which KDM5B contributes to tumor progression, we conducted a ChIP-seq analysis to map KDM5B occupancy in HONE1-EBV cells with targeted KDM5B knockdown (using siRNA targeting KDM5B, referred to as si-KDM5B#1/2; see Methods). This analysis revealed a significant presence of KDM5B across promoter regions at genome-wide level (Supplementary Fig. 4a), with a notable reduction in KDM5B occupancy around the TSS (−2 kb to +2 kb) following KDM5B knockdown (Fig. 5a). Transcriptome analysis identified 1943 and 1708 differentially expressed genes in two groups of NPC cells with KDM5B knockdown (Fig. 5b). Due to KDM5B’s function as a transcriptional repressor,25 our investigation focused on genes that were upregulated and contained decreased KDM5B occupancy in their promoters following KDM5B knockdown. This approach led to the identification of 21 genes potentially regulated by KDM5B (Fig. 5c). Further transcriptome analysis revealed eight genes (FAT4, FHL2, GCLM, JPH2, IL17RE, SH3RF2, PLK2, and TPM4) to be consistently downregulated in EBV-infected S26 and HK-1 NPC cells compared to uninfected cells (Data not shown). RT-qPCR analyses in both EBV-negative and -positive epithelial tumor cells, as well as those cells with KDM5B knockdown or overexpression, pinpointed PLK2 as the primary target with the most significant alterations affected by KDM5B (Supplementary Fig. 4b–d). Luciferase reporter assays further corroborated these results, demonstrating that KDM5B knockdown markedly enhanced, whereas its overexpression reduced, PLK2 transcription activity in HONE1-EBV and AGS-EBV cells (Fig. 5d, e).

Fig. 5

KDM5B represses PLK2 expression through H3K4me3 demethylation at its promoter. a ChIP-seq enrichment analysis showing KDM5B occupancy around the transcription start site (TSS) of target genes, ranging from -2 kilobases (kb) to +2 kb, following KDM5B knockdown (si-KDM5B#1/2), alongside with control (si-NC). b Heatmap showing the differential expressed genes (DEG) affected by KDM5B knockdown (si-KDM5B#1/2). Filled colors from blue to red represent expression fold change from low to high, in the KDM5B knockdown group compared to the control group. c Scheme diagram illustrating the integrative analysis of ChIP-seq and RNA-seq data for two KDM5B knockdown sets, followed by intersecting their shared genes. Luciferase assays evaluating the PLK2 transcription activity in HONE1-EBV and AGS-EBV cells with KDM5B knockdown (d) by shRNAs (#1/#2) or overexpression (e). f ChIP-Seq profiling KDM5B occupying signals in PLK2 loci following KDM5B knockdown by siRNAs (#1/2). ChIP-qPCR analysis evaluating the binding of KDM5B and H3K4me3 at PLK2 promotor in HONE1-EBV and AGS-EBV cells with KDM5B knockdown by shRNAs (g) or overexpression (h). Statistical analysis is conducted using Student’s t-test for two groups and one-way ANOVA followed by Dunnett’s post hoc test for more than two groups. Data are presented as the mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. SD, standard deviation

Given that KDM5B exerts its transcriptional repression through removal of the trimethyl modification of H3K4 (H3K4me3, a marker for transcriptional activation),25 we assessed its role in activating PLK2 promoter. ChIP-seq analysis demonstrated a substantial reduction in KDM5B occupancy in the −500 bp to −200 bp region upstream of the PLK2 TSS in NPC cells upon KDM5B knockdown (Fig. 5f). Furthermore, ChIP-qPCR with KDM5B and H3K4me3 antibodies revealed that KDM5B knockdown reduced its occupancy and enhanced H3K4me3 modification at PLK2 promoter in both NPC and GC cells (Fig. 5g). In contrast, KDM5B overexpression led to increased KDM5B occupancy and a decreased H3K4me3 modification at the promoter (Fig. 5h). These findings collectively indicate that KDM5B downregulates PLK2 expression through reducing H3K4 trimethylation at the PLK2 promoter. Additionally, we also observed a significant reduction in KDM5B enrichment and a corresponding increase in H3K4me3 enrichment at the promoter regions of several other candidate genes, including FAT4, FHL2, GCLM, JPH2, IL17RE, SH3RF2, and TPM4, following KDM5B knockdown (Supplementary Fig. 4e). These results are consistent with the ChIP-seq analysis (Supplementary Fig. 5), highlighting the critical role of KDM5B in regulating global histone modification processes.

KDM5B facilitates tumor progression through suppressing PLK2 in EBV-associated epithelial tumors

To explore the role of PLK2 in EBV-related cancers, we first examined its effects on NPC and GC cell lines. We observed that PLK2 overexpression significantly inhibited, while its knockdown notably enhanced, both growth and colony formation in EBV-positive epithelial tumor cells (Fig. 6a, b and Supplementary Fig. 6a–d). This aligns with the well-known tumor suppressor function of PLK2 in various human cancers.26,27 To further investigate whether PLK2 counteracts KDM5B-driven tumor progression, we introduced simultaneous overexpression or knockdown of KDM5B and PLK2 in HONE1-EBV, CNE2-EBV and AGS-EBV cells. Intriguingly, PLK2 overexpression remarkably attenuated the increased cell growth, colony formation, and sphere formation capacities of KDM5B-overexpressing EBV-positive NPC and GC cells as compared to the control groups (Fig. 6c–e and Supplementary Fig. 7a). Furthermore, KDM5B knockdown significantly inhibited NPC cell proliferation, while the suppressive effects were largely rescued by PLK2 knockdown (Supplementary Fig. 7b–d), suggesting that PLK2 is a major downstream mediator of KDM5B-induced tumor progression.

Fig. 6

PLK2 mediates KDM5B-promoted tumor progression in EBV-associated epithelial tumors. a Western blot analysis showing the protein level of PLK2 in CNE2-EBV, HONE1-EBV, and AGS-EBV cells infected with lentivirus expressing PLK2 or empty vector. GAPDH is used as control. b Colony formation ability of cells described in (a), with statistics summarized at the right. c Cell growth curves of HONE1-EBV, CNE2-EBV, and AGS-EBV cells stably overexpressing KDM5B (KDM5B) or not (control vector), followed by infection of lentivirus expressing PLK2 or empty vector. d Colony formation ability of cells described in (c), with statistics presented at the right. e Sphere formation ability of HONE1-EBV and AGS-EBV cells described in (c). f Tumor growth curve of xenograft from CNE2-EBV and AGS-EBV cells described in (c). Tumor size (g) and weight (h) for the xenograft tumors excised from (f). i Representative IHC staining images showing the protein levels of KDM5B and PLK2 in NPC samples (NPC #1 and NPC #2). j Comparison of PLK2 expression level in NPC patients (n = 120) with high or low KDM5B expression based on IHC staining scores. Statistical analysis is conducted by Student’s t-test for two groups and one-way ANOVA followed by Sidak’s post hoc test for more than two groups. Data are presented as the mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. SD, standard deviation. Scale bar, 100 μm

Moreover, in vivo xenograft models clearly showed that PLK2 overexpression significantly curtailed the tumor growth boosted in CNE2-EBV and AGS-EBV cells due to KDM5B overexpression (Fig. 6f–h and Supplementary Fig. 7e). These observations highlight the capacity of PLK2 to counterbalance the tumorigenic function of KDM5B. Further IHC staining analyses demonstrated a notable negative correlation between PLK2 and KDM5B expression levels in an NPC cohort (n = 120; Fig. 6i, j). Kaplan-Meier survival analysis revealed a significant association between elevated PLK2 expression and improved overall and disease-free survival (Supplementary Fig. 7f), contrasting with the adverse association of KDM5B overexpression. Notably, patients with concurrent high KDM5B and low PLK2 expression exhibited the worst outcomes in terms of both OS (left) and DFS (right) survivals (Supplementary Fig. 7g). These findings strongly suggest that KDM5B may facilitate tumor progression of EBV-associated epithelial tumors by suppressing PLK2.

To investigate if the effects of KDM5B and PLK2 are EBV-dependent, we conducted parallel knockdown and overexpression experiments in EBV-negative HONE1, CNE2, and AGS cell lines. Our results demonstrated that KDM5B knockdown had a significantly weaker suppressive effect on cell proliferation in EBV-negative cells compared to EBV-positive cells (Supplementary Fig. 8a, b). Similarly, PLK2 knockdown induced a smaller proliferative effect in EBV-negative cells (Supplementary Fig. 8a, b). Notably, western blot analysis revealed that KDM5B knockdown in EBV-negative cells did not result in PLK2 upregulation (Supplementary Fig. 8c), a phenomenon observed in EBV-positive cells (Supplementary Fig. 7b). These findings strongly suggest that the regulation of PLK2 by KDM5B is dependent on the presence of EBV. Moreover, weaker effects of both KDM5B and PLK2 overexpression on cell proliferation were observed in EBV-negative cells compared to EBV-positive cells (Supplementary Fig. 8d–f). Collectively, these results underscore the critical role of the KDM5B/PLK2 axis in EBV-driven tumor progression and highlight its potential as therapeutic target in EBV-associated epithelial tumors.

KDM5B/PLK2 axis mediate EBV-induced activation of the PI3K/AKT/mTOR pathway

Given the pivotal role of the PI3K/AKT/mTOR signaling pathway in the progression of EBV-associated epithelial cancers,28,29 we investigated the effects of KDM5B and PLK2 on this pathway in NPC and GC cells. Western blot analyses revealed that EBV infection led to KDM5B upregulation and concurrently PLK2 downregulation in S26 and AGS cells, alongside with the PI3K/AKT/mTOR pathway activation reflected by the increased levels of p110α (a catalytic subunit of PI3K), p-AKT (Ser 473), and p-mTOR (Ser 2448) (Fig. 7a and Supplementary Fig. 9a). In contrast, KDM5B knockdown resulted in PLK2 upregulation and decreased the expression of these key signaling molecules in these cells upon EBV infection (Fig. 7a and Supplementary Fig. 9a). Moreover, while Zta initiated the activation of PI3K/AKT/mTOR pathway in HONE1-EBV and AGS-EBV cells, KDM5B knockdown abolished the Zta-induced effects on this signaling cascade (Fig. 7b and Supplementary Fig. 9b). Conversely, KDM5B overexpression led to the activation of the signaling cascade in these cells (Fig. 7c and Supplementary Fig. 9c), consistent with the finding in prostate tumor progression.