Estrogen induces the expression of EBV lytic protein ZEBRA, a marker of poor prognosis in nasopharyngeal carcinoma

Abstract Several epidemiological studies have suggested that Epstein–Barr virus (EBV) lytic infection is essential for the development of nasopharyngeal carcinoma (NPC), as the elevation of antibody titers against EBV lytic proteins is a common feature of NPC. Although ZEBRA protein is a key trigger for the initiation of lytic infection, whether its expression affects the prognosis and pathogenesis of NPC remains unclear. In this study, 64 NPC biopsy specimens were analyzed using immunohistochemistry. We found that ZEBRA was significantly associated with a worsening of progression‐free survival in NPC (adjusted hazard ratio, 3.58; 95% confidence interval, 1.08–11.87; p = 0.037). Moreover, ZEBRA expression positively correlated with key endocrinological proteins, estrogen receptor α, and aromatase. The transcriptional level of ZEBRA is activated by estrogen in an estrogen receptor α‐dependent manner, resulting in an increase in structural gene expression levels and extracellular virus DNA copy number in NPC cell lines, reminiscent of lytic infection. Interestingly, it did not suppress cellular proliferation or increase apoptosis, in contrast with cells treated with 12‐O‐tetradecanoylphorbol‐13‐acetate and sodium butyrate, indicating that viral production induced by estrogen is not a cell lytic phenomenon. Our results suggest that intratumoral estrogen overproduced by aromatase could induce ZEBRA expression and EBV reactivation, contributing to the progression of NPC.


| INTRODUC TI ON
Epstein-Barr virus is a human gamma-herpes virus associated with several types of malignancies, including NPC. Recent advances in radiation therapy and chemotherapy have reduced the mortality rate of patients with NPC. However, some NPC populations continue to exhibit poor prognoses, even in the early stages of cancer.
Therefore, more sensitive biological markers must be developed that reflect the poor prognosis of patients with NPC.
In most EBV-associated tumors, EBV infection is predominantly latent. The immediate-early BZLF1 gene product of EBV, ZEBRA, activates the promoters of EBV lytic genes, induces a cascade of viral gene expression, and switches from latent to lytic infection. Recently, several studies have revealed the contribution of lytic infection to the oncogenesis of NPC. 1,2 Some studies have reported that serological anti-ZEBRA titers correlate with diagnosis and prognosis after radiation therapy for NPC. [3][4][5][6] We previously demonstrated that ZEBRA-positive NPC is a risk factor for advanced lymph node metastasis using immunohistochemistry. 7 ZEBRA expression initiates lytic infection triggered by biological or chemical substances in vitro, including TPA, sodium butyrate (SB), calcium ionophores, and anti-Ig. 8 Nonetheless, the physiological stimuli that trigger ZEBRA expression and lytic infection in vivo have not been identified. 8 In addition, whether its expression affects the prognosis remains to be determined.
Estrogen is a risk factor for various tumors, including breast, colon, and lung cancers. [9][10][11][12] Circulating estrogen levels are strongly associated with an increased risk of breast cancer. 9 In addition, some meta-analyses have reported that 20%-30% of breast cancer tissues carry the EBV genome and express viral genes. 13,14 BZLF1 expression was also detected in EBV-positive breast cancer tissues, and anti-ZEBRA titers were correlated with a worse prognosis in these cases. 15 Moreover, aromatase, a cytochrome P450 enzyme complex that mediates estrogen synthesis, is ectopically expressed. 10,11,[16][17][18] Estrogen is activated by binding to the ER, a member of the nuclear receptor family, located in the cytoplasm. 19 Activated ER acts as a transcription factor and influences various biological activities including carcinogenesis.
In this study, we hypothesized that EBV-infected nasopharyngeal cells are exposed to high estrogen levels due to aromatase expression during tumorigenesis. Indeed, we observed that the expression of ZEBRA in NPC was associated with an increase in the expression levels of aromatase and ERα, as well as with the worsening of PFS.

In in vitro models, ZEBRA expression is induced by estrogen in an
ERα-dependent manner and contributes to invasion and migration.
Interestingly, we observed that the estrogen-ZEBRA axis resulted in an increase in viral production but not in cell death, in a manner distinct from that of conventional TPA + SB-treated lytic cells.

| Patient characteristics and ZEBRA detection
The present study included Japanese patients with NPC who were diagnosed in the Department of Otolaryngology, Head and Neck Surgery at Kanazawa University Hospital between January 2000 and December 2019. Unavailable tissue specimens were excluded.
Finally, 64 patients were examined ( Table 1). Based on the immunohistochemical results of ZEBRA, we divided the patients into ZEBRA-positive and ZEBRA-negative groups using a 15% positive cutoff value (median value of expression score) and compared the clinical outcomes. The primary outcomes were OS and PFS rates.
This study was approved by the Ethics Committee of Kanazawa University (IRB#2016-033), and written informed consent was obtained from all the patients before enrollment.

| Immunohistochemical analysis
In total, 64 NPC biopsy specimens were examined to determine the expression levels of ZEBRA, aromatase, ERα, and PD-L1. Tumor p = 0.037). Moreover, ZEBRA expression positively correlated with key endocrinological proteins, estrogen receptor α, and aromatase. The transcriptional level of ZEBRA is activated by estrogen in an estrogen receptor α-dependent manner, resulting in an increase in structural gene expression levels and extracellular virus DNA copy number in NPC cell lines, reminiscent of lytic infection. Interestingly, it did not suppress cellular proliferation or increase apoptosis, in contrast with cells treated with 12-O-tet radecanoylphorbol-13-acetate and sodium butyrate, indicating that viral production induced by estrogen is not a cell lytic phenomenon. Our results suggest that intratumoral estrogen overproduced by aromatase could induce ZEBRA expression and EBV reactivation, contributing to the progression of NPC.

K E Y W O R D S
BZLF1, estrogen, lytic infection, nasopharyngeal carcinoma, ZEBRA fixation and immunohistochemical analyses were performed as described previously. 20 The antibodies used in this study are listed in Table S1. The stained sections were independently evaluated by two investigators (HD and SK) who were blinded to clinical data. Aromatase expression was considered positive when cytoplasmic staining was diffused in the absence of nuclear staining, as described previously. 21 ERα and ZEBRA expression was considered positive if there was nuclear-specific dark brown staining, as described previously. 7,22,23 Specimens with at least 10% ERα-, 15% ZEBRA-, and 20% aromatase-positive tumor cells were considered positive. Each cutoff value was determined using the median value of each expression score for immunohistochemical positivity. PD-L1 expression was as-

| Plasmids and retroviral infection
The BZLF1 expression vector pcDNA3-BZLF1 was constructed as described previously. 25 From the plasmid, the cDNA of the pcDNA3-BZLF1 fragment digested with BamHl and XhoI was introduced into the pFB-Neo vector (Addgene, catalog no. 69767), and the plasmid was named pFB-BZLF1. Retroviruses were prepared as described previously. 26  Abbreviations: CRT, chemoradiotherapy; ERα, estrogen receptor α; RT, radiotherapy; SD, standard deviation. * Data are presented as mean and SD and were evaluated by t-test.

| Western blotting
SDS-PAGE and western blotting were performed as described previously. 28 The antibodies used are listed in Table S1.

| Transfection and luciferase assay
To examine the activity of the Zp, we prepared pZp-luc and its derivatives, as described previously and shown in Figure 4A. 30

| Cell proliferation assay
In total, 2.0 × 10 3 cells were seeded into a 96-well plate, and cell proliferation was performed using a CellTiter 96 Aqueous One Solution Cell Proliferation Kit (Promega, catalog no. G3580) as described previously. 32

| Statistical analysis
Statistical significance was set at p < 0.05. All reported p-values were two-sided. All in vitro experiments were performed in triplicate unless otherwise stated. All statistical analyses were performed using EZR (Saitama Medical Center, Jichi Medical University), a graphical user interface for R (R Foundation for Statistical Computing). 33 Detailed information on the materials and methods is available in the supporting information (Appendix S1).

| Aromatase expression in palatine and pharyngeal tonsil samples
To examine whether aromatase is expressed in normal tonsil tissues under physiological conditions, we first performed immunohistochemical staining of aromatase in 10 palatine and pharyngeal tonsil samples ( Figure 1A

| Expression of ZEBRA is positively associated with the expression levels of ERα and aromatase in NPC
The present study included 64 patients with NPC for immunohistochemical analysis. We assessed whether the expression levels of ERα and aromatase were related to ZEBRA expression ( Figure 1C,D).

Nucleus-specific dark brown staining of tumor cells was considered
positive for ERα and ZEBRA ( Figure 1C). 23 Pearson's correlation analysis of ERα and ZEBRA expression in NPC samples showed a marked association (r = 0.78, p = 1 × 10 −14 ; Figure 1E), suggesting that estrogen levels might be related to the mechanism underlying ZEBRA expression. A weak association was observed between the expression levels of ERα and aromatase (r = 0.49, p < 0.0001; Figure 1F) and aromatase and ZEBRA (r = 0.42, p < 0.001; Figure 1G). Next, we performed dual fluorescence immunostaining for the expression of ERα and ZEBRA in the same tissue sample. Several cells co-expressed both ERα and ZEBRA in NPC tissue samples ( Figure 1H).
No association was found between the expression levels of PD-L1 (TPS and CPS) and ZEBRA when observed by immunohistochemical analysis ( Figure S1A,B). In dual fluorescence immunostaining of PD-L1 and ZEBRA, the expression of ZEBRA was not correlated with the expression of PD-L1 ( Figure S1C).  Figure S2A). Moreover, nine of the 10 female patients were aged over 55 years, indicating that most female patients were postmenopausal. We did not find any difference in aromatase expression levels between male and female patients with NPC (p = 0.85; Figure S2B). Therefore, intratumoral local aromatase expression in women, particularly those who were postmenopausal, was as important as that in men in this study.

| ZEBRA can predict progression-free survival in NPC
Next, we examined whether ZEBRA affects NPC prognosis. The baseline characteristics of the analyzed population of the ZEBRApositive (n = 32) and ZEBRA-negative (n = 32) groups, according to immunohistochemical analyses, are summarized in Table 1.
The baseline characteristics of the patients were similar between the groups. Kaplan-Meier curves for OS and PFS are shown in Figure 1I,J, respectively. The log-rank test revealed that the PFS of ZEBRA-positive patients with NPC was poorer than that of ZEBRAnegative patients with NPC (p = 0.01, Figure 1J); however, the OS was not (p = 0.24; Figure 1I). Therefore, we performed a Cox proportional hazard regression analysis of PFS to determine whether ZEBRA positivity remained a significant variable. Multivariate analysis showed that ZEBRA positivity in NPC was a significant risk factor for poor PFS (adjusted hazard ratio 3.58, 95% confidence interval 1.08-11.87, p = 0.037; Table 2). These analyses indicate that ZEBRA expression is associated with poor prognosis in NPC.
In total, 13 cases of recurrence were observed following first-line treatment, and nine of these cases were treated with chemotherapy or immune checkpoint inhibitors. As described in Figure S3, NPC is susceptible to second-line treatments including chemotherapy and immune checkpoint inhibitors.

| BZLF1 promotes cell invasion and migration
To further clarify whether ZEBRA expression conferred malignant cellular properties, we established a stable NPC cell line expressing ZEBRA. ZEBRA expression was detected by western blotting in HK1-pFB-BZLF1 cells; however, not in HK1-pFB-Neo cells (Figure 2A).
To analyze the effects of BZLF1 on cells, cell growth kinetics were examined. The growth ratio of HK1-pFB-BZLF1 cells was not significantly increased compared with that of HK1-pFB-Neo cells ( Figure 2B). However, according to Matrigel-coated Transwell and wound scratch migration assays, HK1-pFB-BZLF1 cells had a significantly higher number of invasive cells (p < 0.0001; Figure 2C,D) and showed an increase in the rate of migration (p < 0.0001; Figure 2E,F) compared with HK1-pFB-Neo cells. These data suggest that ZEBRA expression upregulates the invasion and migration of tumor cells.

| Transcriptomic analysis of publicly available NPC data
To further validate the immunohistochemical results, RNA-seq analysis was performed using four publicly available data sets ( Table S3).
The flow of sample selection for the NPC GEO data is shown in Figure S4. We excluded 10 EBER-negative NPC samples from the 131 NPC samples from the GEO database.

| Association between BZLF1 and ESR1 expression
In total, 121 NPC data points were divided into BZLF1-positive (n = 59) and BZLF1-negative (n = 62) groups. The expression level of ESR1 mRNA in the BZLF1-positive group was significantly higher than that in the BZLF1-negative group (p = 0.016; Figure 2G). These results support our hypothesis that ERα induces ZEBRA expression in NPC. and BZLF1-negative (n = 42) groups. A trend toward worse PFS was observed in the BZLF1-positive group, although the difference was not statistically significant (p = 0.18; Figure S5A). Additionally, we examined the PFS for morphologically undifferentiated NPCs only (n = 36), and a similar trend was observed (p = 0.24; Figure S5B).
Although we could not detect a significant difference, we hypothesized that the short follow-up duration of patients (average 2.0 ± 0.9 years) in this study could be attributed to the lack of statistical power in this analysis. Nevertheless, it was notable that there was a trend of a relationship between worse prognosis and BZLF1 expression observed in the transcriptomic analysis, comparable with the results of our immunohistochemical analysis.

| Intratumoral estradiol in NPC
ZEBRA expression was found to be associated with ERα expression through immunohistochemical and RNA-seq analyses using publicly available data sets ( Figures 1C-H and 2G). Therefore, we hypoth-

| Estrogen induces BZLF1 transcription in NPC cells
To clarify the causal relationship between ZEBRA and ERα upregulation, we treated EBV-positive NPC cell lines with E2. E2 significantly induced ZEBRA expression in HK1-EBV-eGFP cells exposed to E2

| Estrogen activates the BZLF1 promoter, and the ZII cis-element in BZLF1 promoter is essential for the estrogen effect of BZLF1 expression
We demonstrated that ZEBRA protein synthesis and BZLF1 mRNA production were induced by E2. These results prompted us to examine whether Zp was activated by E2 stimulation using the reporter plasmid Zp (pZp)-luc and its derivatives, as shown in Figure 4A.
E2 significantly induced WT-Zp luciferase activity in HK1 cells (p < 0.001; Figure 4B). Next, to determine which part of the Zp requires the activation of Zp by E2, we further tested reporter vectors with mutated bZIP-binding motifs ( Figure 4C). In the absence of E2, the luciferase activities of the derivatives containing the ZII mutation (pZp-luc mZII and pZp-luc mZII+III) were significantly lower than those of the others, suggesting that ZII of Zp is important for Zp activation in NPC cells ( Figure 4C). Moreover, E2 had little effect on the luciferase activity of the derivatives containing the ZII mutation, including pZp-luc mZII and pZp-luc mZII+III ( Figure 4D). These results demonstrated that ZII is essential for the transcriptional activation of Zp by E2.

| Interfering with estrogen and ER pathway decreases ZEBRA expression under estrogen treatment
Through immunohistochemical and RNA-seq data analyses using publicly available data sets and E2 exposure experiments in NPC cells, we found that the pathway of E2 and ERα pathways,

| Estrogen affects ZEBRA expression via the estrogen receptor-dependent signaling pathway
Based on the results of luciferase reporter assays of Zp and anti-E2 treatment, we hypothesized that the activation of ERα, which is a transcription factor, is the key to Zp activation. Therefore, we verified whether ERα was necessary for E2-mediated induction of ZEBRA using HK1-EBV-eGFP ( Figure 5D) and NPC-TW01-EBV-NeoR cells ( Figure 5E) transfected with ERα-short interfering RNA (ERα-siESR1). Combined treatment with DMSO and ERα-siRNA inhibited ERα protein expression; however, it did not affect ZEBRA protein expression levels. Combined treatment with E2 and ERα-siRNA inhibited the protein expression levels of both ERα and ZEBRA to a greater extent than EGFP-siRNA-transfected controls ( Figure 5D,E).
These results suggest that the ERα pathway is essential for ZEBRA expression in response to E2 treatment.
Based on these results (Figures 1-4, 5A-E), we hypothesized a predictive model for the mechanism by which aromatase and estrogen induce ZEBRA expression ( Figure 5F). In postmenopausal women and men with NPC, intratumoral aromatase converts androgens in the blood into estrogen, following which intratumorally produced estrogen is exposed to the tumor. Conversely, in premenopausal women with NPC, we hypothesized that there are two pathways of estrogen exposure to the tumor. One pathway involves the production of estrogen in the ovary, which is exposed to the tumor via the blood, and the other pathway, in which the estrogen is locally produced via aromatase, is described above. Estrogen, which is supplied intratumorally through these production pathways, activates

| NPC cells produce and release EBV DNA extracellularly in an estrogen dependent manner
One definition for activating the EBV lytic cycle is that progeny virus particles are produced. As shown in Figure S9, E2 also induces the expression of p18 and gp350/220, which are essential for the production of progeny viruses. To investigate whether E2 activates the EBV lytic cycle completely, we further evaluated the copy number of EBV in cells and medium to determine whether E2 produced EBV DNA and released it extracellularly from NPC cells. There was a significant increase in the quantity of EBV DNA in cells (p < 0.01; Figure 6A) and medium (p < 0.01; Figure 6B) in HK1-EBV-eGFP cells treated with E2 compared with that in untreated cells.

| Estrogen did not decrease the NPC cell viability compared with lytic inducers
Lytic inducers, such as TPA + SB, reportedly induce apoptosis. [34][35][36] Consistently, when EBV-negative NPC-TW01 and NPC-TW01-EBV-NeoR cells were treated with TPA + SB, EBV-infected TW01 cells showed a significant increase in apoptosis compared with that in uninfected cells ( Figure S10). This suggests that EBV infection sensitizes cells to apoptosis induced by TPA + SB.
In contrast, western blotting for HK1-EBV-eGFP ( Figure 6C) revealed that E2 did not induce the expression of PARP, an apoptotic marker. When we assessed the effect of E2 on cellular proliferation, the growth kinetics of neither HK1-EBV-eGFP nor NPC-TW01-EBV-eGFP cells were affected by E2 treatment, whereas treatment with TPA + SB affected the growth kinetics (p < 0.05; Figure 6D,E).
Consistent with these results, flow cytometry also revealed an increase in apoptotic cells in NPC-TW01-EBV-NeoR cells upon TPA + SB treatment, but not with E2 ( Figure 6F,G). Moreover, HK1-EBV-eGFP cells were treated with E2 (10 nM) or TPA + SB, and untreated cells were observed over time using time-lapse microscopy with propidium iodide (PI) staining ( Figure 6H). Although exposure to TPA + SB significantly increased apoptosis, as detected by PI staining, exposure to E2 did not increase the number of apoptotic cells ( Figure 6I). These results suggest that E2 increased viral production, but it did not affect cellular proliferation or apoptosis, unlike conventional lytic inducers such as TPA + SB.

| Estrogen does not interfere with the cell cycle progression of EBV-positive tumor cells
To identify the mechanism by which the E2-induced lytic cycle does not induce apoptosis, we performed RNA-seq on HK1-EBV-eGFP cells treated with E2 or TPA + SB and subjected them to GO enrichment analysis. In TPA + SB-treated cells, GO terms for the "mitotic cell cycle process" (GTSE1, MKI67, and CDCA5) were downregulated ( Figure S11A). Moreover, several GO terms related to cell cycle progression, such as "regulation of cell cycle process," "DNA replication," and "positive regulation of cell cycle process" were downregulated in TPA + SB ( Figure S11A). In TPA + SB-treated cells, GO terms for the "extracellular matrix" (ANGPTL6, LRRC24, and AZGP1) were upregulated ( Figure S11B). In contrast, there was no significant enrichment of GO terms in E2-treated cells that were related to the cell cycle ( Figure S11C,D). KEGG pathway analysis demonstrated that downregulated genes of HK1-EBV-eGFP cells treated with TPA + SB were significantly involved in the "cell cycle," "Fanconi anemia pathway," and "progesterone-mediated oocyte maturation" ( Figure S12A-C). The KEGG map of the cell cycle signaling pathways is shown in Figure S12D. In the KEGG map of the cell cycle, 15 genes were significantly downregulated in HK1-EBV-eGFP cells treated with TPA + SB. These results suggest that TPA + SB induces G1 arrest and apoptosis, which are disadvantageous for cell survival and viral replication; however, E2 can successfully circumvent these two events.

| DISCUSS ION
ZEBRA is a transcriptional activator, and its expression alone is adequate for the initiation of the entire lytic cycle. 3  induced lytic replication appears to be more advantageous for both cellular proliferation and viral production than conventional methods.
Plasma EBV-DNA levels reportedly correlate with tumor volume and stage of EBV-positive NPC, which is associated with poor treatment outcome. 37,[40][41][42][43]  ZEBRA is also involved in the invasion and migration of tumor cells, which explains its association with the poor prognosis of NPC.
Others have reported that ZEBRA induces the expression of MMP1, MMP3, and MMP9, which are associated with angiogenesis and extracellular matrix degradation. 2,7,45,46 Moreover, the ZEBRA protein reportedly contributes to tumorigenesis via other mechanisms such as genome instability, tumor-promoting inflammation, and immune evasion. 47 The causal relationship and contribution of each mechanism to invasion, migration, metastasis, and survival remains to be determined.
Tissue-specific expression of estrogen is induced by the conversion of androgens in the blood to estrogen and is mediated by the tissue-specific expression of aromatase. Recently, the carcinogenic properties associated with estrogen overexpression, mediated by tissue-specific expression of aromatase, have gained worldwide attention. Aromatase activity has been detected not only in gonadal tissues but also in some extragonadal tissues, including adipose tissues, skin fibroblasts, liver, and brain. 48 Local expression of aromatase is a risk factor for colon and lung cancers. 11,18 Moreover, we demonstrated that aromatase was expressed in the normal pharyngeal tonsil, and the expression levels of ZEBRA, ERα, and aromatase were correlated in NPC. Collectively, our study proposes a noncanonical, tumorigenic role for paracrine/intracrine estrogen, not limited to breast and uterine tumors, developed by endocrine estrogen. 49 In conclusion, we demonstrated the significance of ZEBRA expression, induced by local estrogen, in NPC development.

Atsushi Kaneda is currently an editorial board member in Cancer
Science. The authors declare no conflicts of interest associated with this manuscript.

S U PP O RTI N G I N FO R M ATI O N
Additional supporting information may be found in the online version of the article at the publisher's website.