Identification of an N6-methyladenosine-mediated positive feedback loop that promotes Epstein-Barr virus infection.

N6-methyladenosine (m6A) is among the most abundant mRNA modifications, particularly in eukaryotes, and is found in mammals, plants, and even some viruses. Although essential for the regulation of many biological processes, the exact role of m6A modification in virus-host interaction remains largely unknown. Here, using m6A -immunoprecipitation and sequencing, we find that Epstein-Barr virus (EBV) infection decreases the m6A modification of transcriptional factor KLF4 mRNA and subsequently increases its protein level. Mechanistically, EBV immediate-early protein BZLF1 interacts with the promoter of m6A methyltransferase METTL3, inhibiting its expression. Subsequently, the decrease of METTL3 reduces the level of KLF4 mRNA m6A modification, preventing its decay by the m6A reader protein YTHDF2. As a result, KLF4 protein level is upregulated and, in turn, promotes EBV infection of nasopharyngeal epithelial cells. Thus, our results suggest the existence of a positive feedback loop formed between EBV and host molecules via cellular mRNA m6A levels, and this feedback loop acts to facilitate viral infection. This mechanism contains multiple potential targets for controlling viral infectious diseases.

Epstein–Barr virus (EBV), a large double-stranded DNA virus belonging to the γ-herpesvirus subfamily, causes latent infection of more than 95% of the adult population worldwide (1, 2). As a widespread virus causing infectious mononucleosis during primary infection, EBV is also closely implicated in several malignancies in latent infection, including B-cell lymphomas, nasopharyngeal carcinoma (NPC) (3, 4, 5), EBV-associated gastric cancer, Burkitt's lymphoma, and Hodgkin's lymphoma (6, 7, 8, 9). The viral oncogenic factors, such as RNA transcripts and proteins expressing during the latent or lytic phase have been reported to drive pathogenic processes (10, 11).

The EBV immediate-early protein BZLF1 is highly expressed following lytic stimuli induction, and it mediates the transition from the latent to the lytic replication cycle, which is critical for viral spread (12, 13, 14, 15). As a transcription factor, BZLF1 binds to the promoters of many viral lytic genes as well as cellular genes to regulate their expression through EBV–host crosstalk (16). It has been reported that BZLF1 indirectly inhibits the expression of IFN-γ and promotes the expression of VEGF, IL10, and IL13 via viral proteins (10, 17). In addition, BZLF1 was also identified to directly regulate the expression of various host genes, such as FOSB and RASA3 (16).

Krüppel-like factor 4 (KLF4) is a zinc-finger transcription factor belonging to its eponymous family, which is implicated in cell growth, proliferation, and differentiation (18). Recently, there is increasing evidence that KLF4 is involved in other biological processes such as viral infection. For example, KLF4 negatively regulated cellular antiviral response by inhibiting the activation of NF-κB by TNF-α and IL-1β (19). In addition, melatonin was found to stabilize KLF4 by preventing its m6A-dependent mRNA decay (20). However, whether and how KLF4 is regulated by the N6-methyladenosine (m6A) modification via virus–host interaction need to be further elucidated.

m6A is the most prevalent internal modification present in the messenger RNAs of eukaryotes viruses that replicate in the nucleus (21, 22, 23, 24). The modification is introduced by a multicomponent “writer” complex consisting of the enzyme methyltransferase-like 3 (METTL3) and other cofactors including METTL14 and WTAP (25). The resulting m6A residues are recognized by “reader” proteins such as YTHDF2 and YTHDF3, which contain a YTH domain that directly binds to m6A (22, 26, 27, 28). There are multiple lines of evidence that m6A plays an important role in regulating viral infection. Previous studies characterized the m6A RNA methylomes of many RNA viruses, including human immunodeficiency virus, Zika virus, hepatitis C virus, enterovirus 71, murine leukemia virus, and the plant virus alfalfa mosaic virus (29, 30, 31, 32, 33, 34, 35, 36, 37, 38), as well as numerous DNA viruses including hepatitis B virus, herpes simplex virus type 1, simian virus 40, Kaposi's sarcoma-associated herpesvirus, and EBV (39, 40, 41, 42, 43, 44, 45, 46). Viral infection alters the cellular m6A modification level to regulate different biological processes (32, 40, 42, 47). Recently, the role of m6A in EBV-associated tumorigenesis has been investigated by Lang et al. (44), who examined the viral epitranscriptome of EBV-transformed cell lines.

Here, we focused on the modification of cellular genes upon EBV infection in nasopharyngeal cells. We found that EBV infection decreased the METTL3-mediated m6A modification of the host gene KLF4, thereby increasing its expression to promote EBV infection. Mechanistically, the expression of METTL3 was inhibited by BZLF1 during viral infection, which increased the mRNA stability of KLF4. Overall, our results demonstrated a positive feedback loop between BZLF1 and KLF4 that facilitated EBV infection, providing new insights into the crosstalk between virus and host cell.

Results

EBV infection alters RNA m6A modification and expression of host genes

Previous studies found that viral infection alters the m6A modification of host genes (32, 40, 42, 48). To identify the m6A modification changes in cellular mRNAs following EBV infection, we performed immunoprecipitation and sequencing of methylated RNA (MeRIP-seq) in infected and uninfected nasopharyngeal epithelial cells (Fig. 1A). After extracting the mRNA from NP460 and HK1 cells at 24 h postinfection with EBV, we used an anti-m6A antibody to enrich m6A-modified RNA fragments. This was followed by RNA sequencing of both the input and immunoprecipitated fractions. A significant number of cellular genes with altered m6A modification in response to EBV infection were screened out. As indicated, 25,902 m6A peaks spanning on 11,601 cellular genes in HK1 cells, as well as 27,007 m6A peaks spanning on 12,110 cellular genes in NP460 cells, were identified (Tables S1 and S2). In total, 6758 peaks were downregulated and 1157 peaks were upregulated in HK1 cells (Fig. 1B). In total, 1935 peaks were downregulated and 947 peaks were upregulated in NP460 cells (Fig. 1C). Among these, 553 differential peaks were present in both cell lines (fold change ≧ 1.2), 68 of which showed increased m6A levels, while 485 showed decreased m6A levels, termed as hyper- and hypomethylated peaks, respectively (Fig. 1D). Almost 87% of the altered m6A peaks were hypomethylated, indicating that the m6A levels of total mRNAs of the host were impaired upon viral infection. To verify this hypothesis, we used a cellular RNA m6A level quantification kit to assess the m6A level of EBV-infected HK1, NP460, and NPEC1-Bmil cells. As expected, the m6A levels of cellular genes were lower in the EBV-infected cells (Fig. S1). Next, the hypomethylated genes were screened out for further analysis combined with the differential gene expression.

Figure 1

EBV infection alters mA modification and RNA expression of host genes. A, schematic representation of the MeRIP-seq protocol used to identify differential m6A modification following infection of HK1 and NP460 cells with EBV-GFP. RNA was harvested at 24 h postinfection (hpi). B and C, the differential peaks (fold change ≧ 1.2) identified by MeRIP-seq in HK1 (B) and NP460 cells (C). RPKM, the mean RPKM value of two biological replicates of MeRIP-seq; Up, the peaks with upregulated m6A level in EBV-infected cells; Down, the peaks with downregulated m6A level; ns, not significant. D, the overlap results of the differential peaks in HK1 and NP460 cells from (B) and (C). E, the overlap results of the differential expression level of genes in HK1 and NP460 cells. F, the KEGG pathway enrichment results of the genes with differential expression level. G and H, distribution of the peaks with EBV-induced changes of m6A levels with differential gene expression in HK1 (G) and NP460 cells (H). Differential peaks, fold change ≧ 1.2; differential gene expression, fold change ≧ 1.2. Hyper-down, hypermethylated m6A genes with downregulated expression; Hyper-up, hypermethylated m6A genes with upregulated expression; Hypo-down, hypomethylated m6A genes with downregulated expression; Hypo-up, hypomethylated m6A genes with upregulated expression; Non, no difference; UI, uninfected sample.

As for the differential gene expression, a total of 607 genes were downregulated and 504 were upregulated following EBV infection in HK1 and NP460 cells (Fig. 1E). KEGG pathway analysis indicated that these genes were mainly enriched in the NF-κB signaling pathway (Fig. 1F), which was consistent with a previous report that viral infection primarily triggered the activation of the transcription factors NF-κB and IRF3 to elicit antiviral response (19). Combined analysis of the m6A levels and differential gene expression revealed 3231 and 985 differential peaks in the EBV-infected HK1 cells and NP460 cells, respectively (Fig. 1, G and H and Tables S1 and S2).

Taken together, the results indicated that EBV infection not only decreased cellular m6A modification, but also regulated the mRNA levels of cellular genes which was mainly related to the NF-κB pathway. We therefore set out to further elucidate the crucial genes participating in the regulation.

EBV infection reduces the mRNA m6A modification level of KLF4 and enhances its expression

After overlap analysis of the differential expression genes with m6A hypomethylated in both EBV infected cells, the MeRIP-seq data revealed that 33 hypomethylated genes were differentially expressed following EBV infection (Fig. 2A). Among these genes, KLF4, a transcriptional regulator of the NF-κB signaling pathway, has been reported to induce lytic EBV reactivation in Burkitt lymphoma cells (49, 50). As indicated in our MeRIP-seq data, KLF4 exhibited lower m6A modification and higher expression levels in the context of EBV infection. Furthermore, EBV infection decreased the m6A levels in the KLF4 mRNA (Fig. 2B), indicating that it may be modified and regulated by EBV.

Figure 2

KLF4 promotes EBV infection as a hypo-up gene. A, the m6A and expression levels of the hypo-up and hypo-down genes from the MeRIP-seq data. B, m6A enrichment results of KLF4 by MeRIP-seq. The box region, m6A peak of KLF4; black arrows, primers of KLF4; red arrow, transcription orientation of KLF4. C, the IgG or anti-m6A antibody enrichment of KLF4 mRNA according to RIP-qPCR assay. The percentage of enriched KLF4 mRNA was determined by dividing IP by input. D, the m6A level of KLF4 following the infection of HK1 and NPEC1-Bmi1 cells with EBV-GFP for 24 h. Fold enrichment was determined by dividing the fold change between the IP and the input, and the mean value of the m6A level of KLF4 in the uninfected cells was defined as 1. E, the mRNA expression level of KLF4 following the infection of HK1 and NPEC1-Bmi1 cells with EBV-GFP for 24 h. The KLF4 mRNA expressions levels were normalized to the housekeeping gene ACTB. F, the effects of KLF4 overexpression in HK1 cells after EBV-GFP infection for 24 h were detected by fluorescence microscopy and FACS. Left, representative data from three independent experiments. Scale bars represent 100 μm. Right, quantification of the GFP-positive cells from three independent experiments. The GFP intensity per cell was determined by measuring the mean FITC value by FACS. The mean value of the GFP intensity per cell in the vector control cells was defined as 1. G, the effects of KLF4 overexpression in NPEC1-Bmi1 cells after EBV-GFP infection for 24 h were detected by fluorescence microscopy and FACS. Left, representative data from three independent experiments. Scale bars represent 100 μm. Right, quantification of the GFP-positive cells from three independent experiments. The GFP intensity per cell was determined by measuring the mean FITC value by FACS. The mean value of the GFP intensity per cell in the vector control cells was defined as 1. The shown datas represent as the means ± SD from three biological replicates. ∗∗p < 0.01, ∗∗∗p < 0.001 according to Student's t-test. EBV, EBV-GFP infected cells; UI, uninfected cells.

Next, RIP assays were performed using an anti-m6A antibody or IgG control. The results indicated that KLF4 was highly m6A-modified in HK1, NPEC1-Bmi1, and NP460 cells (Fig. 2C). To determine the role of KLF4, we performed MeRIP and qRT-PCR assays to measure its m6A modification and expression levels in EBV infected HK1 and NPEC1-Bmi1 cells. As indicated, the m6A modification level of KLF4 was downregulated in response to viral infection, accompanied by increased expression of KLF4 (Fig. 2, D and E). To further validate the role of KLF4 in EBV infection, we transfected HK1 and NPEC1-Bmi1 cells with a KLF4 plasmid, followed by EBV-GFP infection. The fluorescence microscopy and flow cytometry results revealed that nasopharyngeal epithelial cells with higher expression of KLF4 were more susceptible to acute EBV infection, as revealed by their higher fluorescence ratio (Fig. 2, F and G).

Overall, EBV infection reduced the m6A modification of KLF4 and increased its expression, which promoted the EBV infection of nasopharyngeal epithelial cells. We next investigated the mechanism underlying the EBV-induced decrease of KLF4 m6A modification.

EBV infection reduces the expression of METTL3

Since METTL3 is the key enzyme in the m6A writer complex, we next sought to determine whether EBV infection affected the expression level of METTL3. As predicted, the mRNA level of METTL3 was significantly downregulated in HK1 and NP460 cells after EBV infection (Fig. 3A). It has been reported that EBV can be reactivated by 12-O-tetradecanoylphorbol-13-acetate (TPA) and sodium butyrate (NaB) (51, 52, 53, 54). Consistently, the expression of METTL3 was reduced after induction of lytic virus replication in CNE2EBV cells by TPA/NaB treatment, while expression of BZLF1 was increased (Fig. 3B). Mounting evidence supports the idea that BZLF1 is the major transcription factor regulating numerous viral and cellular genes in response to EBV infection. BZLF1 is not only a transcriptional activator, but also as a repressor (13, 55, 56, 57). We therefore hypothesized that BZLF1 might regulate the promoter activity of METTL3. By analyzing the BZLF1 ChIP-seq data in the NPC cell line HONE1 (58) from the GEO datasets, we found that BZLF1 interacted with the promoter of METTL3 (Fig. 3C). To further verify this result, we constructed a luciferase reporter driven by METTL3 promoter and the luciferase assay indicated that overexpression of BZLF1 dramatically inhibited the promoter activity of METTL3 (Fig. 3D).

Figure 3

BZLF1 suppresses METTL3 promoter activity.A, the mRNA expression level of METTL3 following the infection of HK1 and NP460 cells with EBV-GFP for 24 h. The METTL3 mRNA expression level was normalized to the housekeeping gene ACTB. B, the mRNA expression levels of METTL3 and BZLF1 in CNE2EBV cells following the induction of EBV reactivation with TPA (30 ng/ml)/NaB (2 mM) treatment for 24 h. The METTL3 and BZLF1 mRNA expression levels were normalized to the housekeeping gene ACTB. C, the enrichment of METTL3 promoter using BZLF1 ChIP-seq from the GEO datasets (GSE83354, Input: GSM2200165; BZLF1-ChIP: GSM2200163 and GSM2200164) in HONE1 cells. Red arrow, transcription orientation of METTL3; box region, significant peak region of BZLF1 ChIP-seq. D, luciferase activity of the METTL3 promoter following the transfection of 293T cells with the indicated plasmids for 24 h. The promoter activity was determined by dividing the fold change between the firefly luciferase value and the Renilla luciferase value. The mean value of the METTL3 promoter activity in the vector control cells was defined as 1. The protein level of BZLF1 was analyzed by western blotting using an anti-Flag antibody with β-actin as loading control. E, schematic representation of the full-length BZLF1 protein, two truncation mutants, and two amino acid exchange mutation. A185K, alanine 185 mutated to lysine. F, the luciferase activity of METTL3 promoter following transfection of 293T cells for 24 h with indicated truncated mutant plasmids. The experiments were performed similarly as described (D) except for plasmid transfection. G, luciferase activity of METTL3 promoter following transfection of 293T cells for 24 h with the indicated plasmids encoding amino acid variants. The experiments were performed similarly as described for (D) except for plasmid transfection. The datas represent the means ± SD from three biological replicates. ns, not significant; ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001 according to Student's t-test. DIM, dimerization bZIP domain; DNA, basic DNA binding region; EBV, EBV-GFP infected cells; K12R, lysine 12 mutated to arginine; pMETTL3, the METTL3 promoter in the pGL3-basic plasmid; TA, transactivation domain; UI, uninfected cells.

To understand which domains of BZLF1 mediate the repression of the METTL3 promoter, we generated the truncated variants BZLF1(1-199) and BZLF1(134-245) for the luciferase assay (Fig. 3E). The BZLF1(1-199) truncation could not form dimers due to a deletion of the C-terminal dimerization domain (59, 60). The BZLF1(134-245) truncation lacks the N-terminal transactivation domain and could not trans-activate a reporter construct (13). A recent study found that BZLF1 suppressed the expression of the host gene CIITA's via a mechanism that was independent of dimerization and DNA binding of BZLF1 (57). However, our result indicated that it was BZLF1(134-245) truncation, relying on the dimerization domain, that suppressed the promoter activity of METTL3 and not the BZLF1(1-199) variant (Fig. 3F). The dimerization domain of BZLF1 therefore enables it to suppress METTL3 promoter activity.

There is increasing evidence that sumoylation and DNA-binding activity of BZLF1 are important for regulating cellular genes (61). To further investigate whether the transcriptional repression of the METTL3 promoter by BZLF1 was related to its sumoylation or DNA-binding activity, we generated the BZLF1 (K12R) variant carrying a sumoylation site mutation (61) and the BZLF1 (A185K) variant bearing a binding site mutation (17, 55) for further luciferase assays (Fig. 3E). The K12R mutation did not affect the transcriptional repression activity of BZLF1 at the METTL3 promoter (Fig. 3G), which was consistent with our previous finding that the repression of METTL3 by BZLF1 is independent of its N terminal (Fig. 3F). By contrast, the BZLF1 (A185K) mutant did not exhibit any transcriptional suppression activity at the METTL3 promoter (Fig. 3G), indicating that BZLF1 suppressed METTL3 promoter activity based on DNA-binding activity.

Collectively, our results suggested that BZLF1, a transcription factor that was highly expressed during EBV infection, bound the promoter of METTL3 to suppress its expression and that this transcriptional suppression activity is dependent on the dimerization domain and DNA-binding activity of BZLF1.

METTL3 and YTHDF2 accelerate the mRNA decay of KLF4

Since we found that EBV infection suppressed the expression of METTL3 and promoted the expression of KLF4, we wondered whether METTL3 plays a role in regulating the expression of KLF4. We transfected HK1 and NPEC1-Bmi1 cells with two individual METTL3-specific siRNAs and verified the knockdown efficiency (Fig. 4A). Knockdown of METTL3 markedly increased the expression of KLF4 at both the mRNA and protein levels (Fig. 4, B and C). Consistently, overexpression of METTL3 decreased the expression of KLF4 (Fig. S2A). A RIP assay also revealed that METTL3 interacted with the transcripts of KLF4 in HK1 and NPEC-Bmi1 cells (Fig. 4D). Knockdown of METTL3 decreased the m6A modification level of the KLF4 transcript (Fig. 4E), which was consistent with the finding that EBV infection impaired the m6A modification of KLF4 (Fig. 2, D and E). Since it has been reported that the m6A modification level regulates the decay of mRNA, we hypothesized that METTL3 might regulate the decay of KLF4 to affect its expression through m6A modification. HK1 cells with METTL3 knockdown were treated with actinomycin D, which inhibits the transcription of cellular gene transcription in a time-dependent manner. Subsequent qRT-PCR analysis indicated that the mRNA decay of KLF4 was slowed down in cells with METTL3 knockdown compared with the control cells (Fig. 4, F and G). Thus, METTL3 modified the KLF4 mRNA and promoted its decay.

Figure 4

METTL3 and YTHDF2 accelerate the mRNA decay of the cellular gene KLF4. A and B, METTL3 (A) and KLF4 (B) mRNA levels in HK1 and NPEC1-Bmi1 cells transfected with METTL3-specific siRNAs (siM3-1 or siM3-2) for 24 h. At 24 h post-siRNA transfection, the knockdown efficiency of METTL3 (A) and the KLF4 mRNA level (B) were determined by qRT-PCR. The mRNA expression levels were normalized to the housekeeping gene ACTB. C, METTL3 and KLF4 protein level in HK1 and NPEC1-Bmi1 cells transfected with METTL3-specific siRNAs (siM3-1 or siM3-2) for 24 h were analyzed by western blotting with β-actin as the loading control. The numbers represent the ratio of band intensity of KLF4/β-actin and METTL3/β-actin. D, the relative METTL3-RIP enrichment of the KLF4 mRNA was analyzed by RIP-qPCR in HK1 and NPEC1-Bmi1 cells. The percentage of enriched KLF4 mRNA was determined by dividing IP by input. IgG was used as control. E, the relative m6A enrichment of KLF4 mRNA in HK1 cells transfected with METTL3-specific siRNAs for 24 h. Fold enrichment was determined by dividing the fold change of the IP and the input, and the value of the m6A level of KLF4 in the control siRNA transfected cells was defined as 1. F and G, increasing the half-life of KLF4 mRNA by knockdown of METTL3 in HK1 (F) and NPEC1-Bmi1 (G) cells. The remaining mRNA levels of KLF4 after actinomycin D (2 μg/ml) treatment in METTL3 knockdown or siNC control HK1 cells. The results were normalized to the relative mRNA level at 0 h after the Actinomycin D treatment. The datas represent the means ± SD from three biological replicates. Half-lives (t1/2) were obtained from linear regression analysis. Significance was assesses using two-way ANOVA; ∗∗p < 0.01, ∗∗∗p < 0.001. H, YTHDF2 and KLF4 mRNA levels in HK1 and NPEC1-Bmi1 cells transfected with YTHDF2-specific siRNAs (siY2-1 or siY2-2) for 24 h. The knockdown efficiency of YTHDF2 was determined by qRT-PCR. The mRNA expression levels were normalized to the housekeeping gene ACTB. I and J, increasing the half-life of KLF4 mRNA by knockdown of YTHDF2 in HK1 (I) and NPEC1-Bmi1 (J) cells. The experiments were performed similarly as described for (F) except for the siRNA transfection. The datas represent the means ± SD from three biological replicates. Significance was assessed using two-way ANOVA, ∗∗p < 0.01, ∗∗∗p < 0.001. K, the relative YTHDF2-RIP enrichment of the KLF4 mRNA was analyzed by RIP-qPCR in HK1 and NPEC1-Bmi1 cells. The percentage of enriched KLF4 mRNA was determined by dividing the fold change of IP and the input. IgG was used as control. The datas from A, B, D, E, H, and K represent the means ± SD from three biological replicates. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; according to Student's t-test.

There is increasing evidence that the m6A writer YTHDF2 participates in the regulation of mRNA degradation (27). To understand the role of YTHDF2 in regulating the expression of KLF4, we used two YTHDF2-targeting siRNAs with confirmed knockdown efficiency (Fig. 4H). As expected, the mRNA level of KLF4 was significantly increased when YTHDF2 was knocked down (Fig. 4H). Moreover, knockdown of YTHDF2 inhibited the decay of the KLF4 mRNA (Fig. 4, I and J). A RIP assay was conducted to detect the association between YTHDF2 and KLF4, and it confirmed that they interacted in both HK1 and NPEC-Bmi1 cells (Fig. 4K).

These result indicated that METTL3 and YTHDF2 accelerated the mRNA decay of KLF4 by increasing the m6A modification level.

KLF4 promotes the expression of BZLF1 and enhances the lytic replication of EBV

Since EBV infection inhibited the expression of KLF4, we reasoned that BZLF1 may act as a viral transcription regulator that controls KLF4. To test this hypothesis, we generated a stable cell line expressing the BZLF1 gene. The mRNA and protein level of KLF4 were significantly increased in cells with increased expression of BZLF1 following DOX treatment, whereas the level of m6A modified KLF4 was reduced significantly, accompanied by a decrease of the METTL3 expression level (Fig. 5, A and B).

Figure 5

KLF4 promotes the expression of BZLF1 and enhances the lytic replication of EBV. A, the mRNA and the m6A levels of KLF4 following BZLF1 overexpression in HK1 cells. The total RNA was extracted at 24 h post doxycycline (1 μg/ml) treatment. The RNA levels were analyzed by qPCR and the m6A level by MeRIP-qPCR. B, the protein expression levels of KLF4, METTL3, and BZLF1 in BZLF1-overexpressed HK1 cells were analyzed by western blotting with β-actin as loading control. The proteins were harvested at 24 h post doxycycline treatment. The numbers represent the ratios of the band intensities of KLF4/β-actin and METTL3/β-actin. C, BZLF1 protein levels in CNE2EBV cells transected with KLF4 or vector control for 24 h. Beta-actin was used as loading control. The numbers represent the ratios of the band intensities of BZLF1/β-actin. D, the effects of KLF4 overexpression in CNE2EBV cells were detected by fluorescence microscopy. The images show representative data from three independent experiments. Scale bars represent 200 μm. E, the average number of EBV copies in CNE2EBV cells with transfected with KLF4 or vector control for 24 h. The EBV copy numbers were detected using qPCR with a specific primer targeting the EBV BALF5 gene region, and GAPDH was used as a reference genome copy. F, BZLF1 protein expression in KLF4 overexpressed HK1 and NPEC1-Bmi1 cells was analyzed by western blotting after acute infection with EBV-GFP for 24 h. At 24 h post transfection of KLF4 or vector control, the cells were infected with EBV-GFP for 24 h. Beta-actin was used as loading control. The numbers represent the ratio of the band intensities in BZLF1/β-actin. G, the promoter activity of BZLF1 in 293T cells transfected with the indicated plasmids for 24 h. The BZLF1 promoter activity was determined by dividing the fold change of the firefly luciferase value with the Renilla luciferase value, and the mean value of the BZLF1 promoter activity in the vector control cells was defined as 1. The protein levels of KLF4, METTL3, and YTHDF2 were analyzed by western blotting using an anti-Flag antibody. Beta-actin was used as loading control. H, a working model for the crosstalk between EBV and host cells. After EBV acute infection, the EBV IE gene BZLF1 was immediately expressed and suppressed the expression of the m6A modification writer METTL3 to regulate cellular m6A levels. Lower expression of METTL3 could increase the RNA stability of KLF4, one of the altered m6A modified cellular genes, by protecting it from YTHDF2-mediated RNA decay. KLF4 could in turn increase BZLF1 transcription to promote lytic EBV infection of nasopharyngeal epithelial cells. The datas represent the means ± SD from three biological replicates. ∗∗p < 0.01; ∗∗∗p < 0.001, according to Student's t-test.

Next, we sought to determine whether KLF4 also regulated the expression of BZLF1. As reported, KLF4 could regulate EBV infection in telomerase-immortalized normal oral keratinocyte epithelial cells by promoting the transcription of the EBV IE gene, BZLF1 (49). Here, we also found that the overexpression of KLF4 promoted the expression of BZLF1 in HK1EBV and CNE2EBV cells persistently carrying EBV-GFP (Figs. 5C and S3A), resulting in stronger fluorescence intensity of EBV-GFP undergoing lytic replication (Figs. 5D and S3B). The qPCR results indicated that the EBV lytic copy number increased twofold in KLF4 overexpressing CNE2EBV cells (Fig. 5E), as well as 1.5-fold in HK1EBV cells, compared to vector control cells (Fig. S3C). Furthermore, overexpression of KLF4 promoted the expression of BZLF1 upon EBV acute infection (Fig. 5F), leading to higher EBV infection efficiency (Fig. 2, F and G). In addition, we found that overexpression of KLF4 significantly increased the promoter activity of BZLF1, while overexpression of METTL3 and YTHDF2 had the opposite effect (Fig. 5G). Thus, KLF4 could promote EBV lytic replication by upregulating the expression of BZLF1, which was inhibited by METTL3 and YTHDF2 via m6A modification.

Based on the results, we proposed a working model for the crosstalk between EBV and host cells (Fig. 5H). After EBV acute infection, the EBV IE protein BZLF1 was highly expressed and suppressed the expression of the m6A writer METTL3 by interacting with its promoter. The reduced expression of METTL3 suppressed the m6A modification of KLF4, which increased the RNA stability of KLF4 by reducing m6A modification reader YTHDF2-mediated mRNA decay. In return, the higher expression of KLF4 promoted EBV lytic infection in epithelial cells. Thus, the BZLF1 and KLF4 serve as a feedback loop to promote EBV infection of the host through crosstalk with METTL3 and YTHDF2.

Discussion

Several recent studies found that viral infection could alter the cellular m6A levels through various pathways. For example, the HIV-1 envelope protein gp120 upregulated the m6A levels of cellular RNAs independently of viral replication (47). Similarly, EV71 and HCMV infection was found to upregulate the expression of the m6A writers METTL3 and METTL14 (37, 62, 63). Previously, Lang et al. (48) had found that lytic reactivation of EBV infection could suppress the expression of the m6A writer METTL14 in LcL, Akata, 293T-EBV, Mutu I, and Mutu III cells. In this study, we found that EBV infection decreased METTL3 expression in nasopharyngeal epithelial cells via the viral transcription factor BZLF1. Interestingly, the cellular m6A modification could in turn influence the virus infection. A recent study found that changes in the m6A modification of the cellular mRNAs RIOK3 and CIRBP influenced their translation and splicing, which subsequently affected flaviviridae infection. Here, we found that EBV infection increased the expression of KLF4 by downregulating the m6A modification of its mRNA, which ultimately promoted EBV infection of the host.

It has been reported that KLF4 is a transcription factor with roles in the cell proliferation, apoptosis, and differentiation (18). KLF4 was also known for its ability to convert differentiated cells into IPS cells together with other transcriptional factors (49). However, little is known about the role of m6A modification of KLF4 upon viral infection. Here, we found that the m6A modification of KLF4 was inhibited by decreased METTL3 expression in response to EBV infection. A previous study found that KLF4 directly interacted with the promoter of BZLF1 to induce its expression and promoted lytic EBV infection (49), but the influence of BZLF1 on the regulation of KLF4 remained unexplored. Here, we found that BZLF1 interacted with the promoter of METTL3 to reduce its expression, which slowed down the mRNA decay of KLF4. Another study suggested that KLF4 can inhibit the recruitment of IRF3 to the IFNβ promoter by directly binding to it, thus inhibiting the activation of NF-κB by TNF-α and IL-1β (19). Based on these findings, we further hypothesized that KLF4 is upregulated during EBV infection and then not only binds to the promoter of the cellular IFNβ gene to repress immune responses, but also binds to the promoter of viral BZLF1 gene to boost its production, which synergistically further facilitate EBV infection.

BZLF1 has been identified as a transcriptional activator of many genes, including almost all lytic genes of EBV (13). However, studies found that BZLF1 also suppressed the expression of several genes, such as TNFR1, TAP2, and CIITA (55, 56, 57). The suppression of CIITA expression and TAP2 promoter activity was found to be dependent on the BZLF1 transcription domain, while the inhibition of TNFR1 promoter activity was dependent on BZLF1 transcription domain and DNA-binding domain. By contrast, the transcriptional repression of HADC3 mediated by sumoylation of BZLF1 protein required the transcriptional activation domain (61). Previous results therefore indicated that the transcriptional activation domain is crucial for the transcriptional suppression activity of BZLF1, without focusing on the importance of the dimerization domain and DNA-binding activity of BZLF1. However, our luciferase assay results indicated that BZLF1 bound to the promoter of METTL3 to suppress its expression, depending on its dimerization domain and DNA-binding activity rather than its transcription activation domain. The detailed mechanism of the transcriptional suppressor function of BZLF1 merits further investigation.

In summary, we provide solid evidence that m6A modification is involved in the crosstalk between EBV and host cells via a positive feedback loop comprising the viral protein BZLF1, the m6A writer METTL3, and the cellular gene KLF4. The detailed mechanisms through which the feedback loop is regulated must be explored further to gain a comprehensive understanding of EBV infection and the role of m6A modification. These findings contribute to our understanding of the role of EBV-induced m6A modification in virus–host interactions and may in turn further advance the development of prophylactic or curative therapies for EBV-associated diseases.

Experimental procedures

Cells and culture conditions

The 293T cell (ATCC, CRL-3216) was purchased from the American Type Culture Collection (ATCC). Bmi1-immortalized primary NPECs (NPEC1-Bmi1) were established in our laboratory as described previously (Life Technology) (64, 65). NPEC1-Bmi1 and NP460 cells were grown in keratinocyte serum-free medium (Life Technology). HK1 cells were grown in RPMI 1640 medium supplemented with 10% fetal bovine serum (FBS, GIBCO) and were a kind gift from Professor Quentin Liu (Sun Yat-sen University, Guangzhou, China) (66). EBV-positive AKATA cells, grown in RPMI 1640 medium supplemented with 5% FBS, were a kind gift from Professor Maria G. Masucci (Karolinska Institute, Sweden). CNE2EBV and HK1EBV cells were cultured in RPMI 1640 medium supplemented with 5% FBS and 500 μg/ml G418. The 293T cells were grown in Dulbecco's Modified Eagle's Medium supplemented with 10% fetal bovine serum.

EBV-GFP preparation and infection

EBV (Akata strain) with an integrated EGFP gene (EBV-GFP) was prepared in AKATA cells as described previously (64, 67). Briefly, EBV-GFP was produced in AKATA cells by reactivation with 0.8% (v/v) of a goat-anti-human IgG (Huayang Zhenglong Biochem Lab). After 72 h post EBV reactivation, the EBV-GFP was harvested and stored at −80 °C. Epithelial cells were exposed to EBV-GFP for 3 h at 37 °C, and the unbound virus was removed by washing twice with Hanks' solution. The infected cells were cultured in fresh medium for 24 h, followed by the quantification of GFP-positive cells via flow cytometry (Beckman Coulter FC500) and fluorescence microscopy.

Reagents

The antibodies used in this study were as follows: anti-BZLF1 (sc53904, Santa cruz, WB); anti-β-actin (66009-1-Ig, Proteintech, WB); anti-Flag-tag (F4049, Sigma, WB); anti-METTL3 (ab195352, Abcam, WB); anti-KLF4 (11880-1-AP, Proteintech, WB); horseradish peroxidase (HRP)-conjugated goat-anti-mouse/rabbit secondary antibodies (#31460, #61-6520, Invitrogen, WB); anti-METTL3 (15073-1-AP, Proteintech, RIP); anti-YTHDF2 (24744-1-AP, Proteintech, RIP). Reagent included actinomycin D (A9415, Sigma); TRIzol (T924, Sigma); and doxycycline (631311, TAKARA).

Plasmids

The METTL3 promoter (−861 to +13 from the transcription start site) and the BZLF1 promoter (−668 to +15) were PCR-amplified using specific primers with an Xho1 restriction enzyme site at the 5′-end and a Hind III site at the 3′-end. The promoters were subcloned into the pGL3-Basic plasmid. Flag-tagged BZLF1 was expressed using the pcDNA3.1+ vector. The expression vectors for Flag-BZLF1 (K12R) and Flag-BZLF1 (S185K) were generated by site-directed mutagenesis of Flag-BZLF1-pcDNA3.1+. The expression plasmids for Flag-BZLF1 (134-245) and Flag-BZLF1 (1-199) were generated from the Flag-BZLF1 plasmid by PCR. The pENTER plasmid and pENTER plasmids with flag tagged METTL3, YTHDF2, or KLF4 were purchased from Vigene Biosciences. To generate BZLF1 overexpressing cell lines, the Flag-BZLF1 coding sequence was PCR-amplified and inserted into pRetroX-TRE3G vector. To generate METTL3-overexpressing cell lines, the Flag-METTL3 coding sequence was PCR-amplified from Flag-METTL3-pENTER and then inserted into the pLVX-dsred-monomer-N1 vector.

Establishment of stable cell lines

The 293FT cells were transfected with the Flag-METTL3-pLVX-dsred-monomer-N1 plasmid or vector control using the packaging plasmids psPAX2 and pMD2G. The transfected cells were incubated at 37 °C for 6 h, after which the medium was replaced with fresh medium. Cell supernatants containing the viral particles were harvested at 48 h after the transfection. HK1 cells were stably transfected with the lentivirus and selected with puromycin (1 μg/ml).

The GP2-293 cells were transfected with the pRetroX-Tet3G or Flag-BZLF1-pRetroX-TRE3G plasmids using the packaging plasmid pVSV-G. Cell supernatants containing the viral particles were harvested at 48 h after transfection. HK1 cells were stably transfected with the retrovirus mixture of Tet3G and Flag-BZLF1-TRE3G and were selected with neomycin (200 μg/ml) and puromycin (1 μg/ml). The stable cell lines were treated with doxycycline at a final concentration of 1 μg/ml to induce Flag-BZLF1 expression. The stably overexpressing cell lines were then identified using qRT-PCR and western blot analysis.

MeRIP-seq and MeRIP-qPCR

The total RNA was isolated using the TRIzol reagent, and the mRNA was purified using the GenEluteTM mRNA Miniprep Kit (MRN10, Sigma). The mRNA was fragmented using the Ambion RNA Fragmentation Reagent (AM8740, Life Technologies) and then purified by ethanol precipitation. Then, 1/10 of fragmented RNA was retained as input for RNA sequencing. For MeRIP, the fragmented RNA was incubated with an anti-m6A antibody (#202003, Synaptic Systems) conjugated to Protein G Dynabeads (Thermo Fisher Scientific) in MeRIP Buffer. Unbound RNA was removed using a magnetic separator and the beads were washed five times with MeRIP buffer. The MeRIP RNA was then extracted with TRIzol. The sequencing of the input RNA samples and MeRIP samples was performed by RIBOBIO. STAR was used for reads mapping and MACS2 was used for m6A peak calling (68, 69). The RNA-seq reads were normalized using the RSEM method (70).

The m6A modification levels of KLF4 were determined using a MeRIP-qPCR assay. After the MeRIP assay, all input samples and MeRIP samples were subjected to qRT-PCR. The fold enrichment was determined by calculating the Ct values of the MeRIP sample relative to the input sample.

siRNA transfection

A total of 2 × 105 cells were seeded into each well of a 6-well plates and grown for 12 h, followed by transfection with the indicated siRNA duplexes, which were delivered using RNAiMAX (Invitrogen) according to the manufacturer's instructions. All the siRNA duplexes were synthesized by RIBOBIO. The single siRNA duplexes were as follows: siNC, siN0000001-1-5 (RIBOBIO); siMETTL3-1, 5′-CAAGTATGTTCACTATGAA-3′; siMETTL3-2, 5′-GACTGCTCTTTCCTTAATA-3′; siYTHDF2-1, 5′-GACCAAGAATGGCATTGCA-3′; siYTHDF2-2, 5′-GCACAGAAGTTGCAAGCAA-3′.

Dual-luciferase reporter assay

The 293T cells were grown in 24-well plates for 24 h, after which the cells were cotransfected with 500 ng of the indicated plasmids, 100 ng of the promoter-reporter plasmid or pGL3-Basic control, and 10 ng Renilla luciferase vector. Firefly and Renilla luciferase activities were analyzed at 24 h posttransfection using the Dual-Luciferase Reporter Assay System (Promega). The relative luciferase activity was calculated as the ratio of firefly luciferase activity to Renilla luciferase activity.

Quantitative reverse-transcription PCR

For gene expression analysis, total cellular RNA was isolated from cultured cells using the TRIzol Reagent and 1 μg of RNA was reverse-transcribed using the GoScript Reverse Transcription System (A5001, Promega) in a 20 μl reaction mixture. The mRNA level was evaluated by real-time PCR using the LightCycler 480 SYBR Green I Master Mix (4887352001, Roche) on a Roche LightCycler 480 instrument. All the gene expression values were normalized to the housekeeping gene ACTB. For the quantification of EBV genome copies, the cellular genomic DNA was isolated from cultured cells using the EZNA DNA kit (D3396-02, Omega BioTek) according to the manufacturer's instructions. EBV copies were detected using qPCR with specific primers targeting the EBV BALF5 gene region, and GAPDH was used as a reference genome copy. The qPCR primers were as follows: KLF4: 5′-ATGCTCGGTCGCATTTTTGG-3′ and 5′-AATTCGCCCGCTCAGATGAA-3′; ACTB: 5′-GTGAAGGTGACAGCAGTCGGT-3′ and 5′-AAGTGGGGTGGCTTTTAGGA-3′; METTL3: 5′-TTGTCTCCAACCTTCCGTAGT-3′ and 5′-CCAGATCAGAGAGGTGGTGTAG-3′; YTHDF2: 5′-GTTGGTAGCGGGTCCATTACT-3′ and 5′-GGTCTTCAGTTTAGGTTGCTGT-3′; BZLF1: 5′-CCCAGTCTCAGACATAACCC-3′ and 5′-CAGGCTGTGGAGCACCAATG-3′; GAPDH-DNA: 5′-GTCAAGGCTGAGAACGGGAA-3′ and 5′-AGCCACACCATCCTAGTTGC-3′; and BALF5-DNA: 5′-CGGAAGCCCTCTGGACTTC-3′ and 5′-CCCTGTTTATCCGATGGAATG-3′.

RNA immunoprecipitation qPCR (RIP-qPCR)

The RIP assay was performed using the Magna RIP Quad RNA-Binding Protein Immunoprecipitation Kit (Millipore, 17-704). For the endogenous METTL3-RIP or YTHDF2-RIP in HK1 and NPEC1-Bmi1 cells, the cells were plated onto a 10 cm dish and grown for 24 h. The cells were then lysed with lysis buffer and 1% of the lysate was left as the input sample. The rest of the lysate was incubated with the anti-METTL3 antibody, anti-YTHDF2 antibody, or a rabbit IgG control conjugated to Protein A/G Dynabeads in 1 ml RIP buffer. After incubation overnight at 4 °C, the beads were washed five times with wash buffer, and the RIP RNA samples and input samples were collected by adding the TRIzol reagent to the beads. The input and all the RIP samples were subjected to qRT-PCR. The relative RIP enrichment was determined by calculating the Ct values of the RIP sample relative to the input sample.

Measurement of total cellular m6A levels

Total RNAs were extracted from the indicated cells using TRIzol reagent. The total m6A content was measured using an m6A RNA methylation quantification kit (cat. no. P-9005; Epigentek) according to the manufacturer instructions. Briefly, 200 ng of each sample and the standard control were bound to the assay wells. The assay wells were incubated with the capture antibody, detection antibody, enhancer solution, detection solution, and stop solution in that order. Finally, the signal was measured using a microplate reader at 450 nm. The relative m6A quantification was calculated according to the standard curve.

Measurement of mRNA stability

The HK1 or NPEC1-Bmi1 cells were transfected with the indicated siRNAs for 24 h. The cells were then treated with actinomycin D (2 μg/ml). RNA was collected at the indicated time after treatment and examined by qRT-PCR. The turnover rates and half-lives of mRNAs were estimated according to previously published methods (71, 72). After actinomycin D treatment, the changes in KLF4 mRNA remaining at the given time relative to 0 h were calculated using the Ct values as described previously. The half-lives were derived from linear regression fitting of ln (relative RNA remaining) as a function of time using the following equation: t1/2 = ln (2)/slope.

Western blot analysis

Western blotting analyses were performed as described previously (67). Briefly, the cells were lysed in RIPA buffer containing a protease inhibitor mixture (Roche) and incubated on a rocker at 4 °C for 10 min. The protein concentrations of the lysates were measured using a BCA protein assay kit (23227, Pierce) and normalized to equal amounts of protein, separated by SDS-PAGE, transferred to a polyvinylidene difluoride (PVDF) membrane, and incubated with the indicated primary antibodies. Then, PVDF membranes were washed and incubated with species-specific HRP-conjugated secondary antibodies and visualized by enhanced chemiluminescence (ECL, Millipore).

Statistical analysis

The datas were expressed as the means ± SD from at least three independent experiments. Two-group comparisons were conducted using the two-tailed unpaired Student's t-test. Statistical analyses were performed using GraphPad Prism 8 (GraphPad Software). In all results, ns denoted “not significant”, ∗p < 0.05, ∗∗p < 0.01 and ∗∗∗p < 0.001.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request. The authenticity of this article has been validated by uploading the key raw data onto the Research Data Deposit public platform (www.researchdata.org.cn), under the RDD number RDDB2021001089. The raw sequence data reported in this paper have been deposited in the Genome Sequence Archive of the BIG Data Center at the Beijing Institute of Genomics, Chinese Academy of Science, under accession number HRA000681 (accessible at http://bigd.big.ac.cn/gsa-human).

Supporting information

This article contains supporting information.

Conflict of interest

The authors report no conflicts of interest in this work.