Genetic and molecular assessment of the relationship between smoking and EBV reactivation: A two-sample Mendelian randomization and in-vitro experimental study

Abstract

Objective

To assess whether modifiable lifestyle factors (particularly tobacco smoking) have a potential causal effect on Epstein–Barr virus (EBV) reactivation proxied by anti-EBV IgG seropositivity using Mendelian randomization (MR), and to explore the underlying mechanism in vitro.

Methods

We performed a two-sample Mendelian randomization (MR) study (retrospective secondary analysis) using de-identified genome-wide association study (GWAS) summary statistics for 83 dietary habits, 5 tobacco smoking behaviors, and 4 sleep traits with anti-EBV IgG seropositivity as the outcome. We further conducted experiments in EBV-positive B-cell lines to examine mechanisms.

Results

Genetically proxied smoking initiation and lifetime smoking were associated with higher odds of anti-EBV IgG seropositivity, whereas an older age at smoking initiation was associated with lower odds. No consistent associations were observed for sleep traits; dietary findings were heterogeneous across discovery and replication cohorts. In vitro, nicotine increased EBV-DNA levels and up-regulated lytic genes (BZLF1, BRLF1) and gp350, with concordant increases in BZLF1 and EA-D proteins; these effects coincided with ROS accumulation and were attenuated by the ROS scavenger NAC.

Conclusions

Our findings support that smoking is associated with EBV seropositivity and nicotine-induced oxidative stress as a plausible mechanism. Modifying smoking behaviors (e.g. delaying initiation and reducing lifetime exposure) may help lower anti-EBV IgG seropositivity.

Introduction

Epstein-Barr virus (EBV) can elicit infectious mononucleosis and other illnesses, including malignancies (such as Hodgkin lymphoma, gastric carcinoma, nasopharyngeal carcinoma, and Burkitt lymphoma), autoimmune diseases (such as rheumatoid arthritis and multiple sclerosis), and lymphoproliferative diseases.^1,2 EBV exists as a long-term chronic infection in >90% of the global population. ³ The virus occurs in a latent state in healthy individuals; however, it can be reactivated intermittently during an individual's lifespan.^4,5 Although most EBV-related cancers entail the latent state of the virus, its lytic cycle participates in malignancy initiation and sustenance by stimulating growth factor and oncogenic cytokine production.^6,7 During EBV lytic replication process, the generated virus particles can elicit chromosomal instability, culminating in aneuploidy and centrosome amplification, thereby escalating the potential for cellular malignant transformation. ⁸ Proteins encoded by EBV lytic genes, including BNLF-2a, BILF, and BGLF5, impede antigen processing and expression of MHC class I molecules through diverse mechanisms, which enables the virus to evade immune surveillance.^9,10 In the EBV lytic reactivation process, the virus synthesizes various antiapoptotic proteins, such as BHRF1 and BARF1, which are crucial in suppressing cell apoptosis and maintaining the survival of the infected cell.^11,12

The triggering factors for EBV reactivation remain elusive, but its relationship with the host immune system is well known. It has been suggested that EBV residing in the latent phase can undergo reactivation under certain pharmacological triggers, such as immunosuppression. ¹³ Additionally, psychological stress, particularly persistent psychological stress, can alter the replication lifecycle of the herpesvirus by impairment of the cellular immune response.^14,15 Analogously, considering the function of the human immune system, which also could be affected by life habits, we supposed that the immune dysregulation linked to a persistent unhealthy lifestyle could stimulate EBV reactivation, substantially increasing susceptibility to the onset of associated diseases.

The occult state of EBV reactivation and atypical clinical manifestations present significant challenges in investigating the activation of the viral and its epidemiological correlation with various lifestyle factors in real-world settings. Mendelian randomization (MR) offers a solution to these difficulties by using genetic variants associated with an exposure as instrumental variables to infer the potential causality of the risk factor with respect to disease. ¹⁶ Additionally, due to the independent assortment of the instrumental variable risk alleles with confounding factors, MR techniques can mitigate confounding by unmeasured or unknown factors. Although anti-EBV IgG seropositivity is an imperfect proxy—it cannot pinpoint the timing of reactivation or quantify its intensity—direct measurement of reactivation in large cohorts is difficult; accordingly, serostatus is often used as a proxied surrogate despite its limitations. ¹⁷ In present study, we conducted a two sample MR analysis for examining the relationship between EBV antibody seropositivity and modifiable lifestyles, including 83 dietary habits, 5 tobacco smoking habits, and 4 sleep behaviors. We also performed in vitro experiments using human EBV-positive cells to elucidate the molecular mechanisms underlying reactivation of EBV by significant risk factors, with an aim to promote a healthy lifestyle for reducing EBV activation and preventing EBV-related diseases.

Materials and methods

Study design and ethics

A two-sample MR analysis was conducted for examining the potential causal effects of dietary habits, tobacco smoking, and sleep behaviors on anti-EBV IgG seropositivity by utilizing genome-wide association study (GWAS) summary statistics. Three assumptions were considered for causal interpretation of MR estimates (Figure 1). As instrumental variables (IVs), the genetic variations referred to as single nucleotide polymorphisms (SNPs) must meet three assumptions as follows: (1) robustly forecast the exposure; (2) link to outcome only through exposure; and (3) exhibit no relationship with confounding factors affecting the exposure–outcome relationship. ¹⁸ This study was conducted in accordance with the Helsinki Declaration of 1975 as revised in 2024. The two-sample MR used publicly available, de-identified GWAS summary statistics from studies with prior ethics approval and informed consent. Our MR analysis used only publicly available, de-identified summary statistics and involved no interaction with human participants or access to individually identifiable data. Therefore, no additional ethical approval or consent of data source institution was required for the present analysis. This study design was approved by the First Affiliated Hospital of Guangxi Medical University Ethical Review Committee (Approval No. 2025-E0799). The in-vitro experiments used the established human B-cell line Raji only and did not involve human participants or animals. The reporting of this observational MR study conforms to the STROBE and STROBE-MR guidelines (see completed checklist in Table S1 and uploaded as Research Data). ¹⁹

Figure 1.

Study overview and analysis workflow. Schematic of the two-sample Mendelian randomization (MR) analysis and the in-vitro experiments. Genetic instruments (SNPs) for 83 dietary habits, 5 smoking phenotypes, and 4 sleep traits were selected at P < 5 × 10⁻⁶, F > 10, with LD clumping (r² < 0.001, 10-Mb window) and harmonization (excluding palindromic SNPs with 0.2 < EAF < 0.8). The primary MR method was inverse-variance weighted (IVW), with sensitivity analyses including weighted median, MR-Egger, MR-PRESSO, Cochran's Q, and the Steiger directionality test. In-vitro experiments used EBV-positive Raji cells exposed to nicotine (0.1, 1, 10 μM; 24 h) to measure EBV-DNA (BamHI-W qPCR), lytic genes (BZLF1, BRLF1, gp350), lytic proteins (BZLF1, EA-D; immunofluorescence), and ROS (DCFH-DA); N-acetyl-L-cysteine (NAC) was used as a ROS scavenger. Abbreviations: IV, instrumental variable; OR, odds ratio; ROS, reactive oxygen species; NAC, N-acetyl-L-cysteine.

Data sources

We used GWASs primarily involving European ancestry individuals as data sources for the genetic instruments of the modifiable lifestyle factors. Because the anti-EBV IgG outcome GWAS used in this study are both of European ancestry, we restricted the exposure GWAS to primarily European-ancestry datasets to align ancestries across two-sample MR and thereby reduce bias from population stratification and ancestry-specific LD patterns. We selected dietary habits, tobacco smoking habits, and sleep behaviors as typical examples of unhealthy lifestyle habits. A total of 83 individual food items and 83 derived dietary patterns were chosen based on the data of Food Frequency Questionnaire (FFQ) obtained from 449,210 Europeans as indicators of dietary habits from the UK Biobank (UKB) consortium. ²⁰ The UKB fields related to FFQ questions corresponding to dietary habits are available in the UKB Data Showcase (http://biobank.ndph.ox.ac.uk/showcase/). The SNPs linked to tobacco smoking were retrieved from a meta-analysis of the GWAS summary association data, which involved 1,232,091 individuals of mostly European ancestry, including smoking initiation, smoking cessation, age of smoking initiation, and cigarettes per day. ²¹ Lifetime smoking was utilized as a robust phenotype according to the UKB data, which had 462,690 samples dominated by European ancestry. ²² This variable is considered a continuous composite concept of lifetime smoking exposure burden and involves smoking duration, smoking initiation/cessation, and smoking heaviness. Regarding sleep duration, the Sleep Disorder Knowledge Portal (https://sleep.hugeamp.org/downloads.html) was used to acquire the GWAS summary statistics. The participants reported their sleep duration (including naps) at 24-h intervals, with responses indicating hourly increments. GWASs involving European ancestry participants were separately accomplished for short-duration sleep (<7 h/night, n = 106,192 cases) and long-duration sleep (>9 h/night, n = 34,184 cases) relative to sleep duration of >7 h and <9 h (n = 305,742 controls). ²³ To determine insomnia, the participants responded to the question: “Do you have trouble falling asleep at night or do you wake up in the middle of the night?” by choosing from options: “never/rarely,” “sometimes,” “usually,” or “prefer not to answer.” Insomnia was diagnosed if the participants answered “usually”; furthermore, the participants who chose “never/rarely” were considered the control group. European ancestry individuals (345,022 cases and 108,357 controls) were included in Lane et al.'s GWAS. ²⁴ GWAS summary data on anti-EBV IgG seropositivity were extracted from two separate the MRC Integrative Epidemiology Unit (IEU) open GWAS projects, including 5010 cases (ieu-b-4901, European ancestry) and 8735 cases (ebi-a-GCST90006897, European ancestry) (https://gwas.mrcieu.ac.uk/). We selected this serology-based outcome because it is the only EBV-related trait currently available with sufficiently powered GWAS data in two cohorts, enabling a two-sample MR design with replication. While serology is a pragmatic population-level proxy for EBV activity, it does not directly quantify lytic replication at the individual level; we therefore complemented the MR with mechanistic in-vitro experiments to support biological plausibility.

Selection of IVs

MR uses genetic variants associated with an exposure (smoking) as instruments to estimate the exposure's effect on the outcome. All exposure-related SNPs with statistically significant GWAS threshold (P < 5 × 10⁻⁶) were extracted as IVs. SNPs showing an F statistic of <10 were excluded for quantifying the strength of the IVs. Furthermore, to confirm the non-dependency of the IVs, linkage disequilibrium analysis was performed under the following threshold conditions: (1) r² < 0.001 and (2) window size = 10 Mb. Palindromic and ambiguous SNPs were excluded by harmonizing processes (0.2 < EAF < 0.8). SNPs displaying potential pleiotropy were omitted based on the MR-pleiotropy residual sum and outlier (MR-PRESSO) test. We re-performed MR analysis for evaluating the robustness.

Statistical analysis of MR

We used a two-sample MR framework as outlined in Study design and ethics (Figure 1). ²⁵ We adopted the inverse-variance weighted (IVW) model as the primary method. The coherence of the causal estimates was determined by weighted median (WM) as well as MR-Egger regression models. In this approach, each SNP was treated as a valid natural experiment to examine its causal effects on the outcome; furthermore, the combined causal effect was evaluated using the outcome as weights for meta-analysis. When horizontal pleiotropy is absent or balanced, the fixed-effect IVW method provides an unbiased estimate. ²⁶

Sensitivity analyses were conducted to estimate potential bias and enhance our results’ reliability. First, the MR-Egger intercept test was performed to determine pleiotropy. The MR-PRESSO method was also applied for identifying outlying SNPs and for confirming any changes in the causal effect after these outliers were removed. ²⁷ Second, because heterogeneity of SNP effect size could cause a bias in IVW estimates, Cochran's Q statistics were calculated for evaluating heterogeneity in instrument effects. ²⁷ p > 0.05 indicated a small effect of heterogeneity in Cochran's Q test. The random-effect IVW model was adopted to estimate the causal effect if heterogeneity was present. Third, an MR Steiger directional test was conducted for validating whether the outcome is influenced by the exposure. ²⁸ If the outcome R² value was remarkably less than the exposure R² value (MR Steiger test; p < 0.05), the exposure and outcome showed no reverse causality relationship. Leave-one-out analysis was performed for assessing the robustness of the causal effect. This involved removing one IV at a single time and conducting the IVW analysis with the remaining IVs for examining how individual variants affect the causal effect.

Measurement of EBV-DNA levels and gene expression detection

Cell DNA was extracted with the QIAamp DNeasy kit (Qiagen, Hilden, Germany) and eluted in 30 μL nuclease-free water. The RNeasy Plus Mini Kit (Qiagen) was utilized for extracting total cellular RNA. A NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, USA) was used for determining purity and quantity of the extracted nucleic acids.

EBV copy numbers were detected by Premix Ex Taq™ (Probe qPCR) (TaKaRa, Dalian, China), with Namalwa cell line DNA as the standard (reaction conditions were in accordance with the manufacturer's recommendations). The Bio-Rad CFX 96 detection system was used, and all samples were run in triplicate to determine the viral load based on an appropriate standard curve with the following conditions: amplification efficiency of PCR (E) = 95–105% and coefficient of determination (R²) ≥ 0.99). By employing the PrimeScript™ RT Reagent Kit supplemented with gDNA Eraser (TaKaRa), cDNA was reverse transcribed from RNA. RT-qPCR was carried out with the CFX 96 detection system (Bio-Rad) and TB Green^® Premix Ex Taq™ (TaKaRa). Each experiment was conducted in triplicate. The relative gene expression levels were estimated with the 2-^ΔΔct method. Table S4 lists the quantitative primers for EBV genes (reaction conditions were based on the manufacturer's protocol).

Immunofluorescence assay

After fixing the cells with 4% paraformaldehyde, cell smears were prepared. The slides were first blocked for 1 h at room temperature (R/T) with 10% goat serum; subsequently, the slides were incubated overnight with primary antibodies (BZLF1 and Ea-D; dilution: 1:100 and 1:200, respectively; Santa Cruz Biotechnology, sc-53904 and sc-58121, respectively), followed by a 30-min incubation with secondary antibodies (Goat Anti-Mouse IgG H&L Alexa Fluor^® 594 and Goat Anti-Mouse IgG H&L Alexa Fluor^® 488; dilution, 1:500 and 1:500, respectively; Abcam, ab150116 and ab150113, respectively) at R/T. Cell nuclei staining was performed with DAPI. An anti-fade mounting medium was used for mounting the slides. Images were captured under the same exposure conditions, and ImageJ software was used to estimate fluorescence intensity.

RNA sequencing

The TRizol reagent (Life Technologies) was utilized for extracting total RNA from the cells. A NanoDrop 2000 spectrophotometer was used for estimating the purity and concentration of the extracted RNA. The Agilent 2100 Bioanalyzer (Agilent Technologies) was utilized for determining the RNA Integrity Number (RIN). We used samples exhibiting an RIN of >8. Library construction and sequencing were accomplished by Beijing Novogene Bioinformatics Technology Co., Ltd Subsequently, the raw sequencing data were filtered, checked for sequencing error rate, and examined for the distribution of GC content to generate clean reads for further investigations. All quality-controlled clean reads were aligned to the reference genome with HISAT2 software, and sequences with identity of <0.9 and coverage of <0.85 were omitted for excluding redundant sequences from the alignment results. FeatureCounts (version 1.5.0-p3) was employed to calculate the read counts mapped to each gene and the fragments per kilobase of transcript per million mapped reads (FPKM) per gene. DESeq2 software (version 1.20.0) was utilized for analyzing the differential expression of genes between two groups. Genes were considered to be differentially expressed based on |log2(Fold Change [FC])| ≥ 0.00 and adjusted p-value ≤ 0.05. All differentially expressed genes (DEGs) underwent KEGG pathway enrichment analysis with clusterProfiler software (version 3.8.1).

Determination of intracellular ROS levels

To estimate ROS levels, cells were initially digested and then suspended for 25 min in a serum-free medium containing DCFH-DA (10 µM; HY-D0940; MedChemExpress, USA) at 37°C. A flow cytometer (Beckman Coulter, USA) or a microplate reader (Synergy HTX, BioTek Instruments, USA) was utilized for detecting fluorescence intensity with Ex and Em at 485 and 528 nm, respectively.^29,30 FlowJo software (FlowJo LLC, USA; version 10.5.3) was used for analyzing the data.

Statistical analysis for vitro assays

Data are presented as mean ± SD from n = 3 independent biological replicates unless specified; qPCRs were run in technical triplicates. For two-group comparisons we used two-tailed, unpaired Student's t-tests after testing normality (Shapiro–Wilk) and homoscedasticity (Levene/Brown–Forsythe); if assumptions were violated, Mann–Whitney U was applied. For ≥3 group comparisons, we used one-way ANOVA with Dunnett's post-hoc test (or Kruskal–Wallis with Dunn's correction if non-normal). For MR, we report IVW as primary with weighted median/MR-Egger sensitivity, MR-PRESSO for outliers, Cochran's Q for heterogeneity, Egger intercept for directional pleiotropy, Steiger orientation, and leave-one-out. No artificial intelligence (AI) tools (including machine-learning or large language models) were used in the conduct of the research methods, data acquisition, data processing, or statistical analyses.

Results

MR estimates

Figure 2 presents an overview of significant results from the MR analysis examining the associations of 83 dietary habits, 5 tobacco smoking patterns, and 4 sleep behaviors with anti-EBV antibody seropositivity. Table S2 summarizes the data from the MR analysis. A significantly lower odds ratio (OR) of anti-EBV IgG seropositivity was observed for the following 3 protective risk factors: (1) among current drinkers, drinks usually with meals: yes, it varies, no (Discovery cohort: 0.91 [0.84–0.97], p = 0.008, Replication cohort: 0.56 [0.33–0.95], p = 0.032), (2) among current drinkers, drinks usually with meals: yes vs. no (Discovery cohort: 0.94 [0.90–0.99], p = 0.018, Replication cohort: 0.76 [0.50–1.15], p = 0.197), and (3) milk type: soy milk vs. any other (Discovery cohort: 0.60 [0.42–0.84], p = 0.04, Replication cohort: 0.21 [0.01–6.09], p = 0.36), while 4 dietary habits were markedly associated with a high risk of anti-EBV IgG seropositivity: (1) bread type: white vs. any other (Discovery cohort: 1.08 [1.01–1.16], p = 0.023, Replication cohort: 0.84 [0.49–0.1.45], p = 0.537), (2) tablespoons of cooked vegetables per day (Discovery cohort: 1.03 [0.94–1.14], p = 0.5, Replication cohort: 1.88 [1.02–3.46], p = 0.043), (3) temperature of hot drinks (Discovery cohort: 1.08 [1.02–1.15], p = 0.013, Replication cohort: 1.17 [0.70–1.96], p = 0.54), and (4) beer/cider glasses per month (Discovery cohort: 1.12 [1.03–1.21], p = 0.006, Replication cohort: 0.97 [0.45–2.11], p = 0.946). Additionally, no causal association was found between other dietary habits and anti-EBV IgG seropositivity in both discovery and replication cohorts.

Figure 2.

Forest plot summarizing associations from the two-sample MR analysis. Odds ratios (ORs) with 95% CIs for anti-EBV IgG seropositivity per genetically predicted increase in each exposure are shown for discovery (ebi-a-GCST90006897, 8735 cases) and replication (ieu-b-4901, 5010 cases); the number of SNP instruments used for each exposure is indicated. Estimates are from IVW models with two-sided p values; full results and sensitivity analyses (weighted median, MR-Egger) are provided in Table S2. OR > 1 indicates higher odds of anti-EBV IgG seropositivity; OR < 1 indicates lower odds.

For tobacco smoking phenotypes tested by the IVW analysis, genetically predicted smoking initiation (Discovery cohort: 1.05 [1.01–1.10], p = 0.008, Replication cohort: 0.89 [0.68–1.18], p = 0.426) and lifetime smoking (Discovery cohort: 1.08 [1.01–1.15], p = 0.024, Replication cohort: 2.01 [1.20–3.42], p = 0.008) showed a remarkable association with a high risk of anti-EBV IgG seropositivity. Additionally, the risk decreased with later age at age of initiation (Discovery cohort: 0.89 [0.81–0.97], p = 0.007, Replication cohort: 0.79 [0.39–1.61], p = 0.52).

Genetically predicted short-duration sleep, long-duration sleep, sleep duration, and insomnia and anti-EBV IgG seropositivity exhibited no remarkable associations (Table S2).

Sensitivity analyses

To confirm whether the obtained results were robust, sensitivity analyses were performed, which included MR-PRESSO global test, Cochran's Q test, and MR-Egger intercept test (Table S3). In the MR-Egger intercept test, p >0.05 indicated no horizontal pleiotropy. In the Q test, however, heterogeneity was detected between never eat sugar vs. no sugar restrictions and ieu-b-4901 cohort (Q = 39.34, p = 0.034), spread type: flora + benecol vs. any other and ieu-b-4901 cohort (Q = 8.53, p = 0.036), cereal type: muesli vs. any other and ieu-b-4901 cohort (Q = 39.34, p = 0.034), never eat dairy vs. no dairy restrictions and ieu-b-4901 cohort (Q = 12.54, p = 0.028), bowls of cereal per week and ebi-a-GCST90006897 cohort (Q = 100.74, p = 0.009), overall non-oily fish intake and ebi-a-GCST90006897 cohort (Q = 25, p = 0.016), never eat dairy vs. no eggs, dairy, wheat, or sugar restrictions and ieu-b-4901 cohort (Q = 15.78, p = 0.015), cups of coffee per day and ebi-a-GCST90006897 cohort (Q = 72.49, p = 0.026), tablespoons of cooked vegetables per day and ebi-a-GCST90006897 cohort (Q = 73.71, p = 0.034), frequency of adding salt to food and ebi-a-GCST90006897 cohort (Q = 127.76, p = 0.037), frequency of adding salt to food and ieu-b-4901 cohort (Q = 89.60, p = 0.023), beer/cider glasses per month and ieu-b-4901 cohort (Q = 40.93, p = 0.041), never eat wheat vs. no eggs, dairy, wheat, or sugar restrictions and ebi-a-GCST90006897 cohort (Q = 6.15, p = 0.046), and sleep duration and ebi-a-GCST90006897 cohort (Q = 135.00, p = 0.008). Although certain results showed heterogeneity, it did not invalidate the MR estimates as random-effect IVW; this might stabilize the pooled heterogeneity. Additionally, the absence of pleiotropy in the Egger intercept test suggests no introduction of pleiotropic bias in MR estimates in heterogeneity context (Figures S1). Other analyses revealed no heterogeneity (Table S3). Moreover, no SNP influenced the results in the leave-one-out analysis; the funnel plots were also symmetrical, indicating no violation of the estimates (Figures S1).

Effects of nicotine on EBV replication in Raji cells

Motivated by prior evidence that cigarette-smoke exposure promotes EBV lytic gene expression ³¹ and given that nicotine is a principal constituent of cigarette smoke, we investigated whether nicotine promotes EBV lytic activation in EBV-positive B cells. Initially, we examined whether nicotine enhances EBV replication in Raji cells. Treatment of Raji cells with 0.1, 1, or 10 μM nicotine for 24 h dose-dependently increased EBV-DNA content (Figure 3(a), p < 0.05). Moreover, the assessment of EBV lytic gene expression showed elevated expression levels of the early lytic genes BRLF1 and BZLF and the late lytic gene gp350 (Figure 3(b) to (d)). Immunofluorescence analysis indicated an upregulation of BZLF1 and EA-D protein expression in Raji cells following nicotine treatment (Figure 3(e) to (g)). Induction of the immediate-early transactivators BZLF1 and BRLF1, together with increases in EBV DNA, EA-D, and the late gene gp350, suggests entry into the lytic cycle rather than isolated transcriptional changes. Taken together with the increased EBV DNA copy number, these data are consistent with nicotine promoting EBV lytic reactivation in Raji cells.

Figure 3.

Effect of nicotine on EBV replication in EBV-positive B-cell lines. (a) Relative fold increase in EBV DNA in the BamHI-W region of EBV-positive Raji cells untreated or treated with 0.1, 1, and 10 μM nicotine by real-time quantitative PCR. (b-d) Expression of the viral lytic phase genes BZLF1, BRLF1, and gp350 was measured in Raji cells after treatment with 0.1, 1, and 10 μM nicotine for 24 h. (e-g) Immunofluorescence experiments were performed to assess the expression of EBV lytic proteins. Data represent mean ± SD (n = 3). One-way ANOVA with Dunnett's post-hoc test vs. control was used for multi-dose comparisons; two-tailed, unpaired Student's t-tests were used where only two groups were compared. p < 0.05 was considered significant. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Transcriptomic response to nicotine in Raji cells

To evaluate how nicotine exerts biological effects on EBV-positive cell lines, Raji cells were subjected to 10 μM nicotine treatment for 24 h and subsequently underwent transcriptome sequencing. The raw sequencing data were aligned and quantified to obtain the gene expression profiles for each sample. Subsequently, we used principal component analysis (PCA) for visualizing the expression patterns; the findings showed distinct separation between samples pre- and post-nicotine treatment. As depicted in Figure 4(a), 56.88% and 12.00% of the variance were attributed to the first and second principal components, respectively. These findings suggest that nicotine treatment induces notable transcriptional disparities in Raji cells.

Figure 4.

Transcriptome changes in Raji cells following nicotine treatment. (a) PCA showing global expression separation between control and 10 μM nicotine (24 h) groups (n = 4 per group). (b) Volcano plot of DEGs; thresholds: |log₂(FC)| ≥ 0, adjusted p ≤ 0.05. KEGG enrichment of up-regulated (c) and down-regulated (d) DEGs. Bubble size represents gene counts; color encodes adjusted p value. Abbreviations: DEG, differentially expressed gene; adj.P, adjusted p value (Benjamini–Hochberg).

DESeq2 was utilized for differential gene expression analysis in pairwise comparisons, with the criteria of |log2(FC) | ≥ 0 and p-adjust ≤ 0.05 to filter out DEGs. We identified 1452 DEGs between the nicotine treatment and control groups, which included 743 and 709 upregulated and downregulated genes, respectively (Figure 4(b)). To investigate the mechanism of nicotine-induced activation of EBV-positive cell line viruses, significantly upregulated or downregulated genes were subjected to KEGG enrichment analysis to determine distinct functional pathways between the groups. Our analysis showed remarkable enrichment of virus infection-related pathways among the upregulated DEGs (Figure 4(c)), including HIV1 infection and human cytomegalovirus infection, EBV infection, oxidative phosphorylation, and chemical carcinogenesis reactive oxygen species, while downregulated DEGs showed enrichment in pathways associated with cell cycle, apoptosis, and cancer, for example, transcriptional dysregulation in cancer and microRNAs in cancer (Figure 4(d)).

Reactive oxygen species under nicotine exposure

Based on the transcriptomic data, the significantly enriched pathways were related to oxidative stress (OS) and reactive oxygen species (ROS), which are associated with the activation of EBV lytic replication according to the previously reported literature. Therefore, we speculated that the mechanism through which nicotine activates EBV is related to its ROS-related impact. To assess how nicotine treatment affects ROS levels in Raji cells, which predominantly carry a latent EBV infection, we used the DCFH-DA fluorescent probe for detecting intracellular ROS levels. Nicotine treatment elevated ROS levels in Raji cells (Figure 5(a) to (c)). To understand the function of ROS in nicotine-induced EBV lytic activation, we evaluated the effects of ROS inhibitors on such activation. Treatment with ROS inhibitors reduced ROS levels and EBV-DNA content as compared to the untreated group (Figure 5(d)). Concurrently, the mRNA expression levels of BRLF1, BZLF1, and gp350 lytic genes displayed a similar pattern (Figure 5(e) to (g)). Immunofluorescence experiments showed a decrease in BZLF1 and EAD expression levels following treatment with ROS inhibitors as compared to untreated controls (Figure 5(h) to (j)). Thus, nicotine-induced activation of EBV lytic replication is associated with ROS accumulation.

Figure 5.

Nicotine treatment induces the accumulation of reactive oxygen species in Raji cells. (a) Extracellular ROS level was measured by a microplate reader. (b and c) ROS levels in Raji cells determined by flow cytometry. (d-g) Effect of nicotine on the EBV-DNA load and the relative expression levels of lytic genes in Raji cells after NAC treatment was assessed by real-time quantitative PCR. (h-j) Immunofluorescence experiments were performed to assess the expression of EBV lytic proteins. NAC is a general ROS scavenger. Data represent mean ± SD (n = 3). Two-way ANOVA (NAC × nicotine) including the interaction term was used for multi-factor comparisons, with multiplicity-controlled post-hoc tests; two-tailed, unpaired Student's t-tests were used where only two groups were compared. p < 0.05 was considered significant. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Discussion

MR is an observational approach that strengthens inference about potential causality by using genetic variants as instruments, but it does not constitute definitive proof of cause and effect. The causality in our study as a probabilistic inference under the MR assumptions, rather than a proven mechanistic relationship. In this study, although we found evidence of a link between 7 of 83 dietary patterns derived from individual food items and anti-EBV IgG seropositivity, the results for bread type: white vs. any other and beer/cider glasses per month should be cautiously addressed, as the effect direction of the two dietary patterns was discordant in discovery and replication cohorts. One drinking-with-meals phenotype showed a protective association that replicated. We offer a host-mediated hypothesis: eating with alcohol typically lowers peak blood-alcohol concentration and enhances first-pass metabolism, which may attenuate alcohol-induced immune dysregulation and oxidative stress.^32–34

For the tobacco smoking phenotypes, lifetime smoking exhibited an apparent risk to anti-EBV seropositivity, which showed a statistically concordant change between the discovery and replication cohorts. The results also indicated that the initiation of smoking with older age has a lower risk of anti-EBV IgG seropositivity. Our findings agreed with results from some previous case–control studies.^35–37 Most clinical epidemiological investigations, for example, cohort and case-control studies, rely on the observational data, which makes it difficult to avoid confounding factors. ³⁸ However, MR research can address this issue by using genetic variants as tools to determine potential causal links between clinical outcomes and risk factors. ³⁹ MR research resembles randomized controlled trials because genetic variant segregation is random and unaffected by environmental factors, ensuring equal distribution of confounders across groups. ⁴⁰ This allows MR research to provide potential causal evidence. ³¹ Previous studies have mostly examined smoking type (inhaled or not inhaled) and cumulative amount of smoking, while our study included diverse smoking phenotypes, such as smoking initiation and lifetime smoking. In in vitro study, cigarette smoke extract promoted the replication of EBV, stimulated expression of the immediate-early transcriptional activators Rta and Zta, and elevated BFRF3 and gp350 transcriptional expression levels in the lytic state. ⁴¹ Viral infections, notably by EBV, frequently lead to elevated ROS levels in cells, with Zebra protein expression and virus lytic replication potentially further elevating cellular ROS levels.^42–44 Hence, factors inducing EBV reactivation and ROS production could trigger a detrimental feedback loop, thereby amplifying the risk of EBV-related diseases. Our investigation on the impact of nicotine treatment on cellular OS revealed a notable increase in ROS levels following 10 μM nicotine treatment for 24 h. Pretreatment with the ROS inhibitor NAC reduced ROS levels and suppressed nicotine-induced enhancement of viral DNA levels. This supports the hypothesis that the primary role of nicotine in EBV activation may stem from ROS response it induces in cells. While the MR analysis shows an association between smoking and EBV antibody status, and our cell experiments demonstrate a mechanistic plausibility (nicotine triggering lytic gene expression), these results do not establish the causal pathway between increased antibody seropositivity and the lytic gene upregulation. The prospective studies measuring EBV DNA antibody kinetics in smokers vs non-smokers are warranted.

Several limitations of our study should be acknowledged. First, anti-EBV IgG seropositivity is a serologic proxy rather than a direct measure of EBV lytic reactivation. Seropositivity reflects prior exposure and the host immune response and cannot resolve antigen-specific titers, timing, or distinguish latent versus lytic states, which may introduce outcome misclassification and attenuate causal estimates. To mitigate these issues, we examined two independent GWAS datasets (discovery and replication), performed MR-Egger, MR-PRESSO, Cochran's Q, and Steiger tests, and triangulated the MR findings with cell-based experiments showing that nicotine increases EBV-DNA and lytic gene/protein expression and induces ROS, effects that were reversed by NAC. Together, these steps increase confidence in the interpretation while acknowledging the limits of a serologic proxy. Second, our analyses were restricted to European-ancestry GWAS to align exposure and outcome ancestries and reduce stratification; this limits generalizability to non-European populations and does not rule out residual population structure within Europeans. Large resource cohorts (e.g. UK Biobank) may also exhibit selection/participation bias (healthy-volunteer effects), which could propagate into GWAS and hence MR instruments. Third, the in-vitro component provides mechanistic plausibility but has limited external validity: we used a single EBV-positive B-cell line (Raji), and focused on nicotine rather than the full cigarette-smoke milieu; dose/time regimens may not recapitulate chronic exposures. Thus, these experiments support but do not prove clinical reactivation and should be followed by studies measuring EBV antibody kinetics in cohorts (smokers vs non-smokers). In parallel, extend MR (multi-ancestry/within-family, multivariable, colocalization/pleiotropy checks) and replicate mechanisms across EBV-positive B-cell and epithelial models (nicotine vs. smoke extract) to strengthen causal inference and translational relevance.

Conclusion

This two-sample Mendelian randomization and in-vitro study suggests that smoking may promote Epstein–Barr virus (EBV) reactivation. Genetic liability to smoking initiation and greater lifetime exposure were associated with higher odds of anti-EBV IgG seropositivity, whereas later initiation was associated with lower odds. In EBV-positive B cells, nicotine increased EBV DNA and lytic gene expression, and these changes were attenuated by ROS scavenging, consistent with an oxidative-stress–mediated mechanism. If confirmed in prospective and interventional studies, efforts that prevent uptake, and support cessation may help reduce EBV reaction.

Supplemental Material

sj-zip-1-sci-10.1177_00368504251392601 - Supplemental material for Genetic and molecular assessment of the relationship between smoking and EBV reactivation: A two-sample Mendelian randomization and in-vitro experimental study

Supplemental material, sj-zip-1-sci-10.1177_00368504251392601 for Genetic and molecular assessment of the relationship between smoking and EBV reactivation: A two-sample Mendelian randomization and in-vitro experimental study by Weiren Xiang, Nan Shi, Zhenqiu Luo, Fangfang Chen, Shi Luo, Yu Ren, Jingyu Li, Xiang Bin, Guangyao He, Xiang Yi, Wei Xia and Anzhou Tang in Science Progress