Withaferin A inhibits EBV-driven lymphomagenesis through multiple mechanisms, including EBNA1 degradation

Key Points

• •WA induces oxidative stress, NF-κB suppression, and EBNA1 loss in EBV-positive B cells.
• •WA blocks EBV-driven lymphomagenesis in vitro and in vivo by disruption of multiple pathways.

Visual Abstract

Abstract

Epstein-Barr virus (EBV) infects over 90% of the global population and drives multiple aggressive B-cell malignancies, including Burkitt lymphoma, diffuse large B-cell lymphoma, and Hodgkin lymphoma. Standard chemoimmunotherapy regimens can be highly effective, yet Epstein-Barr virus positive (EBV⁺) lymphomas sometimes exhibit poorer responses, higher resistance, and worse survival compared with Epstein-Barr virus-negative (EBV⁻) counterparts. This reflects the virus’s ability to drive immune evasion, alter cell death pathways, and exploit host immune dysfunction, underscoring the potential value of EBV-directed strategies. Withaferin A (WA), a steroidal lactone with known anticancer and anti-inflammatory properties, was evaluated for its efficacy against EBV-associated B-cell non-Hodgkin lymphomas (B-NHL). Across a panel of lymphoma cell lines, WA demonstrated selective cytotoxicity toward EBV⁺ B-NHL, in part through proteasome-dependent degradation of EBNA1 (EBV nuclear antigen 1) and subsequent loss of viral episomes, alongside additional effects on cellular stress and survival pathways. Mechanistic studies revealed that WA collapses antioxidant defenses, drives oxidative stress, and suppresses NF-κB signaling, creating a multipronged disruption of viral and host survival pathways. In primary B-cell models and a cord blood–humanized mouse model of EBV-driven lymphomagenesis, WA inhibited B-cell transformation, reduced splenomegaly and tumor burden, and significantly prolonged survival without evidence of increased viral replication. These findings establish WA as a potent preclinical candidate that selectively targets vulnerabilities unique to EBV-transformed B cells, supporting further optimization and evaluation for EBV⁺ B-cell malignancies.

Epstein-Barr virus (EBV) relies on its EBV nuclear antigen 1 (EBNA1) protein for viral persistence and oncogenesis. Stewart and Damania demonstrate that withaferin A enhances degradation of EBNA1, as well as increasing oxidative stress, and inhibits EBV⁺ tumor cells both in vitro and in a humanized mouse model of EBV-driven lymphomagenesis. This work should revitalize attempts at targeting critical virus-specific proteins with small molecules to treat EBV-driven diseases.

Introduction

Epstein-Barr virus (EBV) is a gammaherpesvirus that infects over 90% of the global population, establishing lifelong latency in B lymphocytes. This persistence contributes to several aggressive lymphoid malignancies, including diffuse large B-cell lymphoma (DLBCL), Burkitt lymphoma (BL), posttransplant lymphoproliferative disease, primary central nervous system lymphoma, and plasmablastic lymphoma. Epstein-Barr virus-positive (EBV⁺) tumors exhibit defined latency programs (latency I-III), characterized by restricted viral gene expression patterns that confer unique molecular profiles compared to their Epstein-Barr virus-negative (EBV⁻) counterparts.

First-line therapy for DLBCL typically involves R-CHOP (rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone), which cures many patients.¹ EBV⁺ DLBCL has often, although not uniformly, been linked to lower response rates and poorer survival, with variability across cohorts reflecting differences in age, comorbidities, and methods of EBV detection.^{2, 3, 4, 5} Similar challenges are observed in other EBV-associated lymphomas,^{6, 7, 8} in which viral latency proteins can promote immune evasion, resistance to apoptosis, and impaired responsiveness to standard regimens.^{9, 10, 11, 12, 13} These clinical hurdles are further compounded in low- and middle-income countries, where EBV-associated cancers are most prevalent. For instance, EBV⁺ BL accounts for 50% to 75% of pediatric cancers in sub-Saharan Africa, where survival rates remain significantly lower than in high-income settings.^{14, 15, 16, 17, 18}

Natural products have historically provided important leads for drug development, particularly in oncology and immunology, because of their structural diversity and ability to engage novel cellular targets.^{19, 20, 21} Withaferin A (WA), a steroidal lactone derived from Withania somnifera (ashwagandha), has demonstrated potent anticancer, anti-inflammatory, and immune-modulating properties.^{22, 23, 24, 25} In tumor models, WA induces cell cycle arrest, inhibits angiogenesis and metastasis, and promotes apoptotic cell death while protecting normal cells from oxidative and chemical stress.^22,23

Here, we evaluate the activity of WA against EBV⁺ B-cell lymphomas. We demonstrate that WA induces proteasone-dependent degradation of EBV nuclear antigen 1 (EBNA1) and that it drives oxidative stress and inhibits NF-κB signaling. Using both in vitro systems and a cord blood–humanized mouse model of EBV-driven lymphomagenesis, we demonstrate that WA selectively targets EBV-transformed B cells, reducing splenomegaly and tumor burden and significantly prolonging survival. These findings establish WA as a multitargeted, preclinical candidate for EBV⁺ lymphomas.

Materials and methods

We used a diverse panel of EBV⁺ and EBV⁻ lymphoma cell lines, primary naïve B cells from healthy donors, and recombinant EBV strains (Akata Bx1 and rB95-8) to evaluate the effects of WA on EBV-driven transformation, proliferation, and survival. Drug sensitivity and viability were assessed using luminescent ATP assays, trypan blue exclusion, dose-response drug screens, and caspase-3/annexin-V–based apoptosis assays. Viral infection of primary B cells was performed at defined multiplicities of infection, with viral activity quantified by transforming units and viral genome copy number. Transcriptional responses and viral load were measured using quantitative RT-PCR and qPCR, while protein expression and signaling pathways (including NF-κB activation and oxidative stress responses) were analyzed by western blotting, NanoLuc reporter assays, ROS-Glo assays, and antioxidant rescue experiments. Flow cytometry was used to assess proliferation, cell-cycle distribution, immunophenotype, and activation markers across infected and cytokine-stimulated B cells. Immunofluorescence microscopy quantified oxidative DNA damage and γH2AX staining in EBV⁺ and EBV⁻ models. For in vivo studies, EBV-infected human cord blood mononuclear cells were engrafted into NSG mice and treated with WA or vehicle to assess lymphomagenesis, tumor burden, viral load, and survival. Spleens and tumors were collected for histopathology, immunohistochemistry, viral DNA quantification, and biochemical analyses. Comprehensive experimental protocols, viral preparation and titration methods, reagent lists, flow cytometry panels, plasmid constructs, and imaging parameters are provided in the supplemental Methods, available on the Blood website.

Cell culture, quantitative reverse transcription PCR, and flow cytometry

All cell lines were grown at 37°C with 5% carbon dioxide. Live cell counts and viability were determined using CellTiter-Glo (Promega, G9241) and trypan blue staining (MilliporeSigma, T8154). The 50% inhibitory concentration was calculated using the GraphPad Prism 9 software. Quantitative polymerase chain reaction (PCR) and genome quantification was conducted using an Applied Biosystems QuantStudio 6 Flex reverse transcription PCR system. Caspase-3 activity was determined using the ApoAlert Caspase-3 Fluorescence Assay kit (Takara Bio, 630215). Flow cytometry experiments were conducted on a MACSQuant VYB (Miltenyi Biotec), gating to live/single cells (supplemental Figure 17), and then analyzed using the FlowJo version 10.8.0 software.

EBV transformation assay

Freshly isolated naïve B cells were infected with the r_wt/B95-8 EBV strain (6008)²⁶ at multiplicities of infection (MOI) ranging from 0.01 to 10. Cells were plated into 96-well plates at 6 × 10⁵ cells per mL, with 24 replicate wells per MOI. After 48 hours, cells were treated with either WA (100 nM; 12 wells) or vehicle control (0.002% dimethyl sulfoxide; 12 wells). The percentage of wells showing B-cell outgrowth at 6 weeks after infection was plotted relative to the MOI. Transformation efficiency was calculated as previously described,²⁷ in which the amount of virus yielding outgrowth in 62.5% of wells corresponds to 1 transforming unit per well. Experiments were performed 3 times with independent donor B cells, and data are presented as mean ± standard deviation.

Cord blood lymphomagenesis model

Human cord blood mononuclear cells were purchased from Stemcell Technologies (Vancouver, Canada), with a unique donor used for each independent experiment. Infected cells were prepared as previously described,²⁸ using 2500 green Raji units per mouse for infection, and then inoculated into 3- to 5-week-old NOD-scid IL2Rgamma^null (NSG) mice. Mice were randomized into treatment groups 5 days after infection. For the dose-ranging study, mice received intraperitoneal injections (150 μL) of WA (1, 5, or 10 mg/kg) or vehicle control (10% dimethyl sulfoxide, 30% Cremophor EL, and 60% Dulbecco phosphate-buffered saline) 3 times per week (Monday, Wednesday, and Friday) for 30 days. At the study end point, mice were euthanized, and spleens and tumors (if present) were harvested, flash frozen, or formalin fixed for further analyses. For the survival experiment, randomized mice received 5 mg/kg WA or vehicle control on the same dosing schedule until humane end points were reached or until 60 days after infection. Mice were monitored regularly for clinical signs of disease, and at necropsy, spleens and tumors were collected for molecular and histopathological analyses. Data on spleen weights, viral load, tumor incidence, and survival were pooled from 3 independent biological replicates.

Results

WA targets EBV-infected B-cell lymphomas

Several compounds targeting pathways critical to EBV infection and persistence were identified, including TH588, BAY 11-7082, baicalein, and 17-DMAG.^{29, 30, 31, 32, 33, 34} WA targets several of these same pathways, such as Hsp90, Sp1, and IκB kinase beta (IKK-β),^{35, 36, 37} suggesting that it could exert broader or more potent effects in B-cell lymphomas. To test this, we used the Akata BL model, which includes coisogenic EBV⁻ (Akata 4E3) and EBV⁺ (Akata BX1) clones. Across a range of concentrations, WA demonstrated greater potency at 24 hours after treatment than previously reported compounds (Figure 1A; supplemental Table 2). EBV⁺Akata BX1 cells were ∼10-fold more sensitive than their EBV⁻ counterparts, with marked reductions in total cell counts (Figure 1B) and increased apoptotic death at submicromolar doses (Figure 1C; supplemental Figure 1A-B). Additional testing across EBV⁺ and EBV⁻ B-cell non-Hodgkin lymphoma (B-NHL) pairs confirmed this preferential cytotoxicity (Table 1). WA also demonstrated activity against EBV⁺ natural killer/T-cell lymphoma lines, which showed greater sensitivity than the EBV⁻ Jurkat line, although direct comparisons are limited by the lack of coisogenic controls. In contrast, epithelial cell lines did not display this selective susceptibility (supplemental Figure 2A-B).

Figure 1.

WA induces apoptosis in EBV⁺ B-NHL cell lines. (A) Dose-response curves for EBV⁻ (Akata 4E3) and EBV⁺ (Akata BX1) cells (mean ± standard deviation of percentage of live cells relative to untreated). (B-D) Quantification of live cell numbers, dead cells (percentage of total), and caspase-3 activity shown as individual data points (n = 6). Statistical analysis for panels B-D: 2-way analysis of variance (ANOVA; factors: EBV status and dose) with Dunnett post hoc test comparing each dose to vehicle within EBV⁺ and EBV⁻ groups. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001. N/A, not available; ns, not significant; Veh, vehicle.

Table 1.

IC₅₀ values for WA-treated cell lines

Cell line	EBV	Neoplasm origin	Latency	IC50 (μM)
BJAB	−	BL	−	2.70
Akata 4E3∗	−		−	2.20
Akata BX1∗	+		I	0.20
Daudi	+		I	0.24
Mutu I	+		I	0.32
Kem I	+		I	0.27
Rael	+		I	0.37
FL18 EBV−∗∗	−	tFL	−	3.71
FL18 EBV+∗∗	+		II	0.67
OCI-LY1	−	DLBCL	−	1.10
OCI-LY3	−		−	3.90
Toledo	−		−	2.10
DOHH-2 EBV−∗∗∗	−		−	1.10
DOHH-2 EBV+∗∗∗	−		II	0.23
Val	−		II	0.27
Farage	−		III	0.57
LCL-X	+	LCL	III	1.40
LCL-Y	+		III	1.28
LCL-Z	+		III	1.13
LCL-GM12878	+		III	2.00
TRL1	+	PTLD	III	1.10
TRL595	+		III	1.30
BC1	+	PEL	I	2.20
JSC1	+		I	2.00
Jurkat	−	T/NKL	−	32.00
SNT13	+		II	5.00
SNT15	+		II	4.90
SNK6	+		II	3.70

Values were determined by dose-response curves using the trypan blue exclusion assay with 3 to 6 biological replicates. Genetically matched clones are indicated (∗, ∗∗, ∗∗∗).

IC₅₀, 50% inhibitory concentration; PEL, primary effusion lymphoma; PTLD, posttransplant lymphoproliferative disorder; T/NKL, T/natural killer–cell lymphoma; tFL, transformed follicular lymphoma.

WA inhibits EBV’s transformation of B cells

WA’s selective cytotoxicity toward EBV⁺ B-NHL cell lines suggested a mechanism specific to EBV’s manipulation of B lymphocytes. To test this, we treated EBV-infected primary B cells with a low dose of WA (100 nM), a concentration that was not toxic to naïve B cells (Figure 2A). Primary B cells freshly isolated from independent donors were infected with EBV strain Akata BX1 at an MOI of 1. Using the CellTiter-Glo assay, we measured the number of live cells (Figure 2A) and found that WA significantly reduced viability in EBV-infected cells but not in uninfected controls. Infected cells showed a substantial increase in cell death (Figure 2B; 22%-73%) and a fourfold decrease in total cell number (Figure 2C). Time-course analysis revealed that proliferation of EBV-infected B cells was abrogated early after WA treatment, with cell division seemingly halted by 48 hours (Figure 2D), indicating that WA disrupts early events in EBV-driven reprogramming. WA-induced phenotypic changes of EBV-infected B cells, such as reduced clumping and smaller cell size (supplemental Figure 3A), consistent with impaired activation.^38,39 Flow cytometry confirmed that WA significantly reduced the proportion of large, “activated” B cells based on forward scatter profiles (Figures 2E-F; mean values of 65.5% to 23.5%). These phenotypic changes were accompanied by a failure of EBV-infected B cells to progress to lymphoblastoid cell line (LCL) outgrowth even after EBV infection at high MOI (Figures 2G-H).

Figure 2.

WA inhibits EBV’s transformation of B cells. (A) EBV-infected and uninfected B cells from independent donors (n = 6) were treated with WA (100 nM) or vehicle (0.002% dimethyl sulfoxide [DMSO]) and analyzed using the CellTiter-Glo assay; live cell numbers are reported relative to control. (B-C) Cell death and total cell counts for EBV-infected B cells after 8 days of treatment (n = 12) analyzed using paired t tests. (D) Time-resolved total cell counts over 8 days (n = 3) analyzed using a 2-way repeated-measure ANOVA with Šídák multiple comparison testing vehicle vs WA at each time point. (E) Representative fluorescence-activated cell sorter plots showing increased SSC and FSC in EBV-infected B cells after 6 days of treatment. (F) Quantification of FSC changes over 5 days (n = 3) analyzed using a 2-way ANOVA with Šídák multiple comparisons at each time point. (G) Quantification of EBV-infected cell outgrowth in the presence of WA (100 nM) or vehicle (0.002% DMSO) control. (H) Calculation of transformation efficiency from (G). (I) Schematic model of EBV stages of B-cell transformation to LCLs. (J) Similar experiments as in panel A, but WA or vehicle was added at 2, 5, or 14 days after infection (n = 6). ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001. FSC, forward scatter; SSC, side scatter; TU, transforming unit; Veh, vehicle. Panel I adapted from Burton et al.⁴¹

EBV’s in vitro transformation of B lymphocytes can be defined in discreet stages (Figure 2I), which uniquely sensitize infected cells to targeted intervention. For example, ATR and Chk1 inhibitors block transformation when introduced early in infection (0-4 days postinfection [DPI]), whereas interventions at later time points (12 DPI) are less effective.⁴⁰ Likewise, Burkitt-like B cells (5-7 DPI) are sensitive to ferroptosis-inducing agents due to a more restrictive latency programming (latency I).⁴¹ WA intervention was initially given at 0 days after infection (0 DPI). Because WA has been reported to inhibit ATR/Chk1 signaling⁴² and may induce ferroptosis in some contexts,⁴³ we expanded our analysis to additional phases of lymphomagenesis: remodeling (2 DPI), Burkitt-like (5 DPI), and LCL-like (14 DPI). Across all phases, WA demonstrated comparable inhibitory activity (Figure 2J), suggesting that WA either targets multiple stage-specific processes or engages a shared vulnerability that is essential throughout the transformation continuum.

WA promotes proteasomal degradation of EBNA1 and cellular proteins that stabilize EBNA1

Given the exquisite susceptibility of EBV-infected B cells to WA, we examined viral and host protein dynamics during early infection and in established cell lines. In EBV-infected primary B cells, EBNA1 levels remained low for the duration of infection, whereas LMP1 expression was stable in both vehicle- and WA-treated cells (Figure 3A). In contrast, Zta (BZLF1, a key transcription factor that initiates lytic reactivation) was transiently elevated for the first 4 days, whereas BMRF1, also known as early antigen diffuse (a marker of productive lytic replication), decreased with WA treatment (Figure 3B), indicating that WA does not initiate productive replication. EBV genome copy number declined to ∼41% of control levels after 6 days of WA treatment (Figure 3C), and transcript analysis revealed modest reductions in EBNA1 and LMP1 mRNA with a slight increase in Zta transcripts (supplemental Figure 4A).

Figure 3.

WA induces proteasomal degradation of EBNA1. (A-B) Protein lysates from EBV-infected primary donor B cells treated with vehicle (0.002% DMSO) or WA (100 nM) were collected at the indicated time points and analyzed using sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblotting for latent proteins (A) and lytic proteins (B). The same membrane was sequentially reprobed after stripping; therefore, the same loading control (H3) applies to both panels. (C) EBV genome copy number quantified from infected donor B cells after 6 days of treatment with vehicle or WA (n = 5 donors) analyzed using a paired t test. (D) Representative flow cytometry histograms showing GFP signal shift in EBV⁺ Akata BX1 cells after WA treatment. (E) Quantitative GFP signal analysis in EBV⁺ Akata BX1 cells treated with 250 nM WA (n = 6 replicates); statistical analysis performed using a paired t test. (F) Immunoblot of cell lysates from the experiment in panel E, with EBNA1 fold change normalized to β-actin loading control. (G) Immunoblot from the same experiment as in panel F, including MG132-treated (10 μM) samples, demonstrating proteasome rescue of EBNA1. H3 serves as the loading control. (H) Time-course immunoblot of EBV⁺ Kem I BL cells treated with WA (250 nM) for 12 to 24 hours showing sequential loss of EBNA1, USP7, PLOD1, and Hsp90. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001. Ea-D, early antigen diffuse; eGFP, enhanced green fluorescent protein; GFP, green fluorescent protein; H3, histone 3; MFI, mean fluorescent intensity; Ub, ubiquitin; Veh, vehicle.

In Akata BX1 cells, which harbor a green fluorescent protein (GFP)–tagged EBV genome, WA caused a dose-dependent reduction in GFP signal (Figure 3D-E) and a parallel loss of EBNA1 protein, with near-complete depletion at 250 nM (Figure 3F; supplemental Figure 4B). Genome copy quantification confirmed a sharp drop to ∼11% of baseline levels (supplemental Figure 4C). Across additional BL and DLBCL cell lines, WA consistently depleted EBNA1, and in LMP1-expressing DLBCL lines, LMP1 also declined (supplemental Figure 4D).

Cotreatment with the proteasome inhibitor, MG132, restored EBNA1 levels (Figure 3G), indicating proteasome-mediated degradation. Consistently, MG132 rescued WA-induced cytotoxicity in EBV⁺ Akata BX1 cells (supplemental Figure 6B), whereas little or no change in viability was observed in EBV⁻ Ramos cells (supplemental Figure 6A), confirming that WA cytotoxicity in EBV⁺ B cells is tightly linked to proteasome-dependent destabilization of EBNA1. To further test whether enforced EBNA1 expression could modulate WA sensitivity, we transiently overexpressed EBNA1 in EBV⁻ Ramos cells and in HEK293 cells (supplemental Figure 5). In Ramos cells, EBNA1 overexpression modestly increased the 50% inhibitory concentration with partial rescue of viability, whereas in HEK293 cells, EBNA1 was neither degraded nor did sensitivity change, underscoring the B-cell–specific context of WA activity.

Time-course experiments in the BL Kem I cell line revealed that EBNA1 loss became evident by ∼16 hours and coincided with degradation of host stabilizing proteins including ubiquitin-specific protease 7 (USP7), procollagen-lysine,2-oxoglutarate 5-dioxygenase 1 (PLOD1), and heat shock protein 90 (Hsp90) (Figure 3H). Because USP7 is known to stabilize EBNA1, we tested the USP7 inhibitor, GNE6776, as a control. As reported, USP7 inhibition reduced EBNA1 protein without affecting USP7 itself (supplemental Figure 7A), but this alone did not impair viability (supplemental Figure 7B), indicating that EBNA1 loss is necessary but not sufficient for WA cytotoxicity. Cotreatment with MG132 rescued the loss of USP7, PLOD1, and Hsp90, along with Hsp90 client proteins Cdc2 and c-Myc, and suppressed gamma-phosphorylated histone H2A variant X (γH2AX) accumulation (supplemental Figure 7C), confirming that WA triggers proteasome-dependent degradation of this viral-host protein network.

WA-induced oxidative collapse drives proteasome-mediated EBNA1 degradation in EBV⁺ B-NHLs

WA treatment induced markedly higher oxidative stress in EBV⁺ B-NHLs compared with their EBV⁻ counterparts as measured through hydrogen peroxide levels (Figure 4A). Baseline 8-oxoguanine staining in untreated Akata and FL18 cells revealed elevated oxidative nucleotide pools in EBV⁺ cells (supplemental Figure 8A-B), consistent with previous studies demonstrating that EBV infection drives chronic redox imbalance and accumulation of oxidative lesions in B cells, with EBNA1 upregulating DNA repair enzymes such as MTH1 and OGG1 to buffer this stress.²⁹ Upon WA exposure, these defenses collapsed, with rapid, proteasome-dependent degradation of MTH1, OGG1, and EBNA1 across multiple BL and DLBCL cell lines (Figure 4B; supplemental Figure 8C-D), leading to increased DNA damage marked by γH2AX selectively in EBV⁺ B-NHLs (Figure 4C).

Figure 4.

WA induces oxidative stress and proteasome-mediated EBNA1 degradation in EBV⁺ B-NHLs. (A) Relative hydrogen peroxide (H₂O₂) levels measured using the ROS-Glo assay in EBV⁻ and EBV⁺ cell lines (n = 5) reported as fold change over untreated control. Statistical comparisons are within each cell line between vehicle (0.002% DMSO) and WA-treated groups, analyzed using 1-way ANOVA with the Dunnett post hoc test. (B) Immunoblot of Akata BX1 cells treated with WA for 24 hours showing dose-dependent loss of EBNA1, MTH1, and OGG1, with increased γH2AX. H3 serves as the loading control. This blot was sequentially reprobed from the membrane shown in Figure 3G. (C) Quantification of γH2AX nuclear staining (total nuclear signal intensity, not discrete foci) from immunofluorescence in EBV⁻ and EBV⁺ Akata cells (5 fields, ∼30 cells each) after WA treatment. Statistical significance assessed using 2-way ANOVA with Dunnett correction, comparing each treatment to vehicle control within EBV status groups. (D) Dose-response curve for Akata BX1 cells pretreated with N-acteylcysteine (NAC; 0-5 mM) for 1 hour followed by WA for 24 hours (n = 3). (E) Immunoblot of Akata BX1 cells treated with vehicle, WA (250 nM), NAC (1 mM), or NAC and WA for 24 hours, showing NAC rescue of EBNA1 and MTH1 and prevention of γH2AX accumulation. (F) Time-course immunoblot of Kem I cells treated with WA (250 nM) showing early loss of MTH1 (∼12 hours), followed by loss of EBNA1 and accumulation of γH2AX (∼16 hours). This blot was sequentially reprobed from the membrane shown in Figure 3H. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001. ns, not significant.

Pretreatment with N-acetylcysteine (NAC) confirmed that oxidative stress drives WA cytotoxicity, rescuing cell viability in a dose-dependent manner and reducing reactive oxygen species (ROS) accumulation (Figure 4D; supplemental Figure 9A-C). Consistent with this mechanism, cotreatment with the proteasome inhibitor MG132, which rescues MTH1 and EBNA1 levels, markedly suppressed WA-induced ROS in both LCLs and Akata BX1 cells (supplemental Figure 9D). Time-course analyses revealed early MTH1 loss (∼12 hours), with DNA damage and EBNA1 degradation detected by ∼16 hours, along with other host chaperones, including USP7, PLOD1, and Hsp90 (Figure 4F). Together, these findings define a model in which EBV⁺ B-NHLs, primed by their baseline redox imbalance, are exquisitely vulnerable to WA-induced oxidative collapse, driving proteasome-mediated degradation of viral and host survival factors, DNA damage, and cell death.

WA disrupts B-cell activation and NF-κB signaling

Because EBNA1 loss alone could not fully explain the early block in EBV-driven transformation, we examined whether WA disrupts signaling pathways that support B-cell activation and survival. In primary B cells stimulated with CpG/immunoglobulin M (IgM), CD40L/interleukin-4 (IL-4), or infected with EBV, treatment with 100 nM WA resulted in distinct viability patterns over a 5-day period (Figure 5A). CpG/IgM-stimulated cells showed an initial ∼50% reduction in viability that plateaued, whereas EBV-infected and CD40L/IL-4–stimulated cells declined steadily to ∼5% to 10% viability by day 5. CellTrace Red assays confirmed reduced proliferation (∼50%) in CpG/IgM-stimulated populations, whereas EBV-infected and CD40L/IL-4–stimulated cells were largely non-proliferative at this stage (supplemental Figure 10A-B).

Figure 5.

WA inhibits CD40/LMP1 activation of B cells. (A) Quantification of live cell numbers using CellTiter-Glo from WA-treated (100 nM) or veh-treated (0.002% DMSO) primary B cells (n = 9 replicates). No statistical analysis is shown. (B) Representative bright-field images of cells at the end point of the experiment in panel A. (C) Quantification of the CD19^highIgM^high B-cell population by flow cytometry after 6 days of WA (100 nM) or vehicle treatment. Statistical significance was determined using 2-way ANOVA with the Dunnett post hoc test, comparing vehicle vs WA within each condition (unstimulated, CD40L/IL-4 stimulated, or EBV infected; n = 5 biological replicates). (D) Time-resolved analysis of cell viability (CellTiter-Glo; top) and ROS accumulation (ROS-Glo; bottom) over 5 days in resting, CD40L/IL-4–stimulated, or EBV-infected primary B cells treated with WA (100 nM; left) or pretreated with NAC (1 mM) followed by WA (100 nM; right). Statistical significance was determined using 2-way ANOVA with the Dunnett post hoc test, comparing day 0 to subsequent time points within each group (n = 3 biological replicates). (E) Quantification of NF-κB–dependent activation markers (CD20, CD38, ICAM-1, PD-L1, and CD23) on EBV-infected B cells following WA (100 nM) or vehicle treatment for 6 days (n = 5 replicates). Statistical significance was assessed using paired t tests. (F) Immunoblots of lysates from LCLs treated with increasing doses of WA, showing loss of p65 phosphorylation at lower doses and total p65 protein loss at higher doses. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001. ns, not significant; P-p65, phosphorylated p65; Unstim., unstimulated; Veh, vehicle.

Microscopy showed that WA disrupted cellular aggregates in EBV-infected and CD40L/IL-4 cultures but not in CpG/IgM-stimulated cells (Figure 5B). Flow cytometry revealed selective depletion of “active” B-cell subsets, characterized by larger forward scatter and high CD19 and IgM expression (Figure 5C), in EBV-infected and CD40L/IL-4–stimulated cultures. Western blotting demonstrated reduced p100 and p52 protein in CD40L/IL-4–stimulated cells but not in CpG/IgM-stimulated cells or resting B cells (supplemental Figure 11C). IKK-β inhibition with the selective inhibitor, IMD-0354, produced only modest viability loss (∼30%) in CD40L/IL-4–stimulated cells and no significant effect in EBV-infected cells (supplemental Figure 11D), indicating that IKK-β inhibition alone cannot explain WA’s potent cytotoxicity.

WA triggered robust ROS accumulation in EBV-infected and CD40L/IL-4–stimulated cells, correlating with loss of viability (Figure 5D; no pretreatment panel) and MTH1 depletion (supplemental Figure 11E). NAC pretreatment blunted ROS (Figure 5D) and restored viability in WA-treated CD40L/IL-4–stimulated cells. Notably, NAC alone disrupted cellular aggregates in these cultures (data not shown), indicating heightened sensitivity of CD40-driven activation states to redox modulation.

In EBV-infected cells, WA reduced NF-κB–dependent activation markers, including CD20, CD38, ICAM-1, programmed cell death ligand 1 (PD-L1), and CD23 (Figure 5E).^{44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54} In LCLs, low WA doses (250 nM) suppressed p65 phosphorylation without altering LMP1 levels (Figure 5F), whereas higher doses reduced total p65, consistent with proteasome-mediated degradation observed for EBNA1 and USP7. NF-κB luciferase assays confirmed dose-dependent transcriptional suppression (supplemental Figure 12A), and cell cycle profiling revealed impaired proliferation with G1/S arrest (supplemental Figure 12B-C). Cotreatment with MG132 rescued total p65 but not phosphorylation (supplemental Figure 12D), indicating that NF-κB suppression occurs upstream of proteasome-mediated protein loss, likely through direct IKK-β inhibition.

Collectively, these findings show that WA potently suppresses NF-κB signaling, with preferential activity in EBV-infected and CD40-driven B cells. In contrast, CpG/IgM-stimulated cells primarily exhibited reduced proliferation without significant NF-κB inhibition or oxidative stress (data not shown), explaining their relative resistance.

WA promotes survival and reduces splenomegaly in a cord blood–humanized mouse model of EBV-driven lymphomagenesis

Building on the robust in vitro activity of WA, we evaluated its efficacy in vivo using a cord blood–humanized NOD-scid IL2Rgamma^null (NSG) mouse model of EBV-driven lymphomagenesis, which faithfully models the aggressive, high-penetrance EBV⁺ DLBCL observed in patients. Mice were treated with vehicle or WA (1, 5, or 10 mg/kg; Monday, Wednesday, and Friday dosing for 30 days starting 5 days after infection). WA significantly reduced splenomegaly in a dose-dependent manner, with mean spleen weights decreasing from 204 mg in vehicle-treated mice to 111 mg (1 mg/kg), 71 mg (5 mg/kg), and 97 mg (10 mg/kg; Figure 6A). Parallel quantification of splenic viral burden showed a sharp decrease from ∼1.1 × 10⁶ copies per μg DNA in vehicle controls to ∼8.6 × 10⁴, 2.8 × 10⁴, and 1.0 × 10⁵ copies per μg DNA at 1, 5, and 10 mg/kg, respectively (Figure 6B). Based on these findings and favorable tolerability, 5 mg/kg was selected for survival studies, which demonstrated a significant extension of survival compared to vehicle-treated controls (Figure 6C).

Figure 6.

WA increases survival of murine EBV⁺ B-NHLs. (A) Spleen weights from mice treated with vehicle or WA (1, 5, or 10 mg/kg) on Mondays, Wednesdays, and Fridays for 30 days, beginning 5 days after EBV-infected cell inoculation (n = 16 mice per group; pooled from 4 independent experiments with unique cord blood donors). Statistical significance determined using 1-way ANOVA with the Dunnett post hoc test, comparing the vehicle, 5 mg/kg, and 10 mg/kg groups to the 1 mg/kg group. (B) EBV genome copy number from spleens in panel A, normalized to total DNA (n = 16 mice per group). Statistical significance determined using 1-way ANOVA with the Dunnett post hoc test, comparing each treatment group to vehicle. (C) Kaplan-Meier survival curves for vehicle- and WA-treated (5 mg/kg) cohorts (25 mice per group, pooled from 3 independent experiments with unique cord blood donors). Significance determined using the log-rank (Mantel-Cox) test. (D) Immunoblot analysis of spleen lysates (collected on the same dates for direct comparison) from the survival cohort, probing EBV proteins. (E) Tumor lysates from vehicle- and WA-treated mice (collected on the same dates for direct comparison) analyzed using SDS-PAGE and immunoblotting for viral proteins and apoptosis markers. (F) Quantification of Zta, phospho-p65, and cleaved caspase-3 protein levels from the immunoblots in panel E, normalized to actin, and expressed as fold change relative to vehicle tumors. Statistical significance determined using unpaired t tests. (G) Representative histopathology images of tumors from vehicle- and WA-treated cohorts. H&E staining (top) reveals extensive necrosis/apoptosis (≥50% of the tumor) in WA-treated tumors. Immunohistochemistry for CD20 (bottom) shows reduced and patchy CD20 staining in WA-treated tumors relative to vehicle controls. ∗P < .05; ∗∗P < .01; ∗∗∗P < .001; ∗∗∗∗P < .0001. Cl. Casp 3, caspase-3; H&E, hematoxylin and eosin; ns, not significant; Veh, vehicle.

Mechanistically, WA treatment reduced EBNA1 protein in splenic lysates (Figure 6D), consistent with in vitro findings of proteasome-dependent EBNA1 destabilization (supplemental Figure 13A-C). In contrast, tumors that survived retained EBNA1 and stable LMP1 expression (supplemental Figure 13D), suggesting that these are escape populations dependent on EBNA1 for persistence. WA-treated mice also had fewer detectable tumors (supplemental Figure 13E), consistent with the observed survival benefit. Tumor lysates demonstrated suppressed phosphorylated-p65, increased Zta expression, and induction of cleaved caspase-3 (Figure 6E), with quantification confirming these changes (Figure 6F). Histopathology revealed widespread necrosis (>50%) and apoptotic morphology (Figure 6G; supplemental Figure 14A), and CD20 immunohistochemistry showed patchy, reduced staining, indicating diminished viable tumor burden (Figure 6G; supplemental Figure 14B). High-magnification hematoxylin and eosin revealed infiltration of lymphocytes, monocytes, and granulocytes in some tumors (supplemental Figure 15), Consistent with these in vivo data, WA-treated LCLs showed partial, abortive lytic activation (supplemental Figure 16). At lower WA doses (250 nM-500 nM), LCLs exhibited modest Zta induction and upregulation of LMP1 but no detectable production of infectious virions, a profile mirrored in EBV-infected primary B cells.

Together, these in vivo findings align closely with our mechanistic data: WA destabilizes EBNA1, induces ROS, suppresses NF-κB signaling, and induces apoptosis while driving partial, nonproductive lytic reactivation. The clear dose responsiveness, significant survival extension at 5 mg/kg, and favorable tolerability underscore WA’s translational potential as a selective therapeutic strategy that exploits the heightened oxidative and proteostatic stress unique to EBV-transformed B cells while sparing uninfected populations.

Discussion

This study provides, to our knowledge, the first comprehensive analysis of WA in models of EBV-associated B-cell malignancies. Across in vitro systems, primary B cells, and a cord blood–humanized mouse model of EBV-driven lymphomagenesis, WA disrupted EBV pathogenesis through a multifaceted mechanism: (1) proteasome-dependent degradation of EBNA1 and its stabilizing partners, (2) collapse of antioxidant defenses leading to oxidative stress and DNA damage, and (3) suppression of canonical and noncanonical NF-κB signaling downstream of CD40 and LMP1. These converging effects destabilized latency, impaired transformation, and triggered apoptotic death in EBV⁺ cells. Although these findings establish WA as a promising lead compound for EBV⁺ lymphomas, they remain a preclinical proof of concept rather than evidence of clinical readiness.

Proteostasis collapse and EBNA1 destabilization

EBNA1 is universally expressed across EBV-associated malignancies and is essential for episome tethering, genome replication, and immune evasion.^{55, 56, 57, 58} Consistent with reports that EBNA1 loss selectively impairs EBV⁺ B cells and prevents LCL outgrowth,^33,34,59,60 we found that WA rapidly and potently reduced EBNA1 through proteasome-dependent degradation, accompanied by loss of viral episomes. Given that EBNA1 transcription is regulated by NF-κB and Sp1 through the Qp promoter,^32,61 WA-mediated suppression of these pathways^{62, 63, 64} likely contributes to reduced mRNA levels. However, EBNA1’s long half-life⁶⁵ and rescue by MG132 indicate that proteasome-mediated degradation is the dominant mechanism.

Multiple host proteins, including USP7, Hsp90, and PLOD1, support EBNA1 stability.^{35,66-67, 68} Pharmacologic inhibition of these factors typically requires prolonged exposure, but WA triggers a rapid, coordinated collapse within 12 to 16 hours, simultaneously degrading USP7, Hsp90, and PLOD1 alongside EBNA1. This multitarget disruption distinguishes WA from single-target approaches and underlies its ability to rapidly eliminate episomes and drive cell death.

Redox imbalance as a driver of sensitivity

Redox imbalance is a well-recognized vulnerability of EBV⁺ cells, created through viral reprogramming that elevates ROS and oxidative stress while activating antioxidant defenses to maintain survival. Previous studies have shown that EBV upregulates enzymes such as MTH1 and OGG1, with EBNA1 playing a key role in sustaining this protective balance.²⁹ Our findings extend this concept by demonstrating how WA exploits this fragile redox state to drive rapid cytotoxicity.

WA has been shown to perturb redox homeostasis in some cancer models by increasing ROS and promoting oxidative stress–induced cell death,^69,70 providing a mechanistic basis for its selective activity in EBV⁺ B cells. In our system, EBV⁺ Akata and FL18 cells exhibited higher baseline ROS compared to their EBV⁻ counterparts, consistent with previous studies. WA further perturbed this balance, triggering sequential MTH1 depletion (by ∼12 hours), EBNA1 loss, and γH2AX accumulation (by ∼16 hours), indicating DNA damage and stress pathway activation. This sequence suggests that oxidative collapse precedes proteostatic destabilization, tipping EBV⁺ cells beyond their tolerance threshold.

Rescue experiments confirmed this model: NAC pretreatment suppressed ROS and restored viability, whereas MG132 cotreatment stabilized MTH1 and EBNA1, blunted ROS accumulation, and prevented cell death. Together, these data support a multitarget mechanism in which WA collapses redox buffering while destabilizing viral (EBNA1) and host (MTH1, OGG1, USP7, Hsp90, and PLOD1) proteins, driving rapid DNA damage and apoptosis. This multipronged disruption is unique to WA and explains its superior potency compared with monoselective agents that require prolonged exposure.

Oxidative stress and NF-κB suppression in EBV-infected B cells

CD40 activation is critical for B-cell activation, proliferation, and germinal center maturation, with NF-κB as a key downstream effector.^{71, 72, 73} LMP1 mimics constitutive CD40 signaling but cannot fully substitute for CD40 in vivo.⁷¹ Consistent with this biology, WA produced context-dependent cytotoxicity: EBV-infected and CD40L/IL-4–stimulated B cells exhibited rapid loss of viability, ROS accumulation, and MTH1 depletion, whereas CpG/IgM-stimulated cells showed partial growth inhibition and relative resistance. Disruption of cellular aggregates and depletion of activated CD19^highIgM^high subsets further support a model in which WA selectively targets highly activated, NF-κB–dependent cells.

Mechanistically, WA potently suppressed canonical NF-κB signaling, with loss of p65 phosphorylation and transcriptional activity at lower doses and proteasome-dependent degradation of total p65 at higher doses. MG132 rescued p65 protein levels but not phosphorylation, consistent with suppression occurring upstream via direct inhibition of IKK-β.^37,74 Both CD40 and LMP1 require IKK-β for canonical and noncanonical NF-κB activation^{75, 76, 77}. However, selective IKK-β inhibition alone did not block early EBV-driven proliferation (supplemental Figure 11D),⁷⁸ highlighting that WA’s additional effects, including EBNA1 destabilization and oxidative stress, are key to its potent cytotoxicity. NAC rescue experiments further confirmed that ROS synergizes with NF-κB suppression to drive cell death.

In vivo efficacy and abortive lytic activation

In the humanized mouse model of EBV-driven lymphomagenesis, WA demonstrated robust, dose-dependent antitumor activity, reducing splenomegaly, lowering splenic viral burden, and significantly extending survival at 5 mg/kg doses. Although plasma levels were not directly measured, previous studies suggest that this dose yields concentrations (∼2 μM-3 μM)⁷⁹ comparable to those of effective in vitro levels, supporting pharmacological relevance. Tumor incidence was reduced by ∼30%, and histopathology revealed extensive tumor necrosis and apoptosis, with diminished CD20 staining consistent with reduced viable tumor burden. These results highlight WA’s systemic antitumor effects, although intermittent drug exposure and tumor heterogeneity may allow for the persistence of drug-tolerant cells.

Mechanistic analysis of tumors revealed suppression of phospho-p65, induction of cleaved caspase-3, and upregulation of Zta without downstream lytic protein expression, mirroring in vitro observations of abortive lytic activation in EBV-infected B cells and LCLs. This supports a model in which WA-mediated NF-κB suppression, coupled with ROS-driven stress, destabilizes latency, drives nonproductive lytic gene expression, and culminates in apoptosis without viral spread. LMP1-driven NF-κB signaling is known to maintain latency^{30,80, 81, 82}; disruption of this pathway by WA parallels the activity of proteasome inhibitors such as bortezomib, which is under clinical evaluation with rituximab in posttransplant lymphoproliferative disease.⁸³ However, WA uniquely destabilizes EBNA1, offering an additional, tumor-specific mechanism of action.

Differential sensitivity and altered response to WA

The reduced sensitivity of LCLs to WA relative to EBV⁺ B-NHL cell lines suggests context-dependent mechanisms of resistance. In LCLs and tumors, WA upregulated LMP1 and Zta without inducing other lytic proteins or infectious virions, indicating abortive lytic activation, whereas EBV⁺ B-NHL lines exhibited loss of LMP1 and no Zta induction. Persistence of EBNA1 in residual tumors despite significant reduction in tumor burden and extended survival further suggests that a subset of EBV-infected cells withstands WA-induced stress.

It has been reported that EBNA1-null viruses can still form LCLs via genome integration through an undefined mechanism.^84,85 This reduced sensitivity may reflect viral genome integration, which is common in LCLs and has also been reported in some EBV⁺ BLs (supplemental Table 3).^{71,84, 85, 86, 87, 88} Such integration may alter viral gene regulation and could contribute to the higher tolerance to WA observed in these models. Other factors, including pharmacokinetic (PK) or microenvironmental differences between in vitro and in vivo systems or intrinsic differences in oxidative stress tolerance or NF-κB dependency may also play a role. Future studies will include genome integration mapping, longitudinal dosing and PK analyses, and mechanistic dissection of abortive lytic activation to better define resistance pathways.

Collectively, our findings position WA as a potent, multitargeted preclinical candidate for EBV-associated B-cell malignancies. By concurrently suppressing NF-κB signaling, inducing oxidative stress, and destabilizing EBNA1, WA selectively disrupts viral and host pathways central to lymphomagenesis. The dose-dependent efficacy, significant survival benefit, and favorable tolerability observed in the humanized model underscore its translational promise while highlighting the need for careful pharmacokinetic, toxicology, and formulation studies. WA is a naturally occurring constituent of W somnifera (ashwagandha), long used in traditional medicine, particularly in South Asia.^22,89 A phase 1 trial in India tested a standardized ashwagandha extract in patients with osteosarcoma, in which it was well tolerated but WA was undetectable in plasma, reflecting limitations of oral extract formulations and underscoring the need for improved delivery strategies.⁹⁰ More recently, a phase 1/2 trial (ClinicalTrials.gov identifier: NCT05610735) was initiated in recurrent ovarian cancer to evaluate ashwagandha tablets with Doxil, aiming to define the maximum tolerated dose and preliminary efficacy. Although encouraging, purified WA itself has not yet entered oncology trials. In the future, WA should be considered as a promising preclinical lead requiring further optimization and formal safety studies for EBV-associated malignancies. Notably, WA’s ability to destabilize EBNA1, a protein expressed across all EBV-driven malignancies, distinguishes it from other targeted strategies such as proteasome inhibitors.⁸³ Rational combination studies will be critical to enhance efficacy, address partial resistance, and advance WA or optimized analogs into early-phase clinical testing.

Conflict-of-interest disclosure: The authors declare no competing financial interests.