Kaposi’s sarcoma-associated herpesvirus induces mitochondrial fission to evade host immune responses and promote viral production

Abstract

Mitochondrial dynamics are pivotal for host immune responses upon infection, yet how viruses manipulate these processes to impair host defence and enhance viral fitness remains unclear. Here we show that Kaposi’s sarcoma-associated herpesvirus (KSHV), an oncogenic virus also known as human herpesvirus 8, encodes Bcl-2 (vBcl-2), which reprogrammes mitochondrial architecture. It binds with NM23-H2, a host nucleoside diphosphate (NDP) kinase, to stimulate GTP loading of the dynamin-related protein (DRP1) GTPase, which triggers mitochondrial fission, inhibits mitochondrial antiviral signalling protein (MAVS) aggregation and impairs interferon responses in cell lines. An NM23-H2-binding-defective vBcl-2 mutant fails to evoke fission, leading to defective virion assembly due to activated MAVS–IFN signalling. Notably, we identify two key interferon-stimulated genes restricting vBcl-2-dependent virion morphogenesis. Using a high-throughput drug screening, we discover an inhibitor targeting vBcl-2–NM23-H2 interaction that blocks virion production in vitro. Our study identifies a mechanism in which KSHV manipulates mitochondrial dynamics to allow for virus assembly and shows that targeting the virus–mitochondria interface represents a potential therapeutic strategy.

Mitochondria are critical signalling hubs mediating interactions between viral replication and host immunity¹. They continuously undergo cycles of fusion and fission, driven by dynamin-related GTPases, to adapt to cell demands and innate immune responses². Although viruses have evolved strategies to evade mitochondrial immune functions, the detailed virus–mitochondria interactions, their impact on viral fitness and their potential as therapeutic targets remain poorly understood. To address these questions, we studied Kaposi’s sarcoma-associated herpesvirus (KSHV), a gammaherpesvirus causing Kaposi’s sarcoma (KS), one of the most common malignancies in HIV-infected individuals³. KSHV infection also causes primary effusion lymphoma (PEL), multicentric Castleman’s disease and KSHV inflammatory cytokine syndrome³, conditions associated with high morbidity and limited effective treatments^4,5. Viral lytic replication, occurring through either de novo infection or reactivation, is essential for viral persistence and pathogenesis^6,7. Previous studies primarily relied on gene overexpression systems, limiting insights into authentic viral functions. The development of the KSHV bacterial artificial chromosome 16 (BAC16) system allows studies within a genuine viral replication context⁸. Using BAC16, previous studies identified the viral Bcl-2 homologue (vBcl-2; ORF16) as essential for KSHV lytic replication, independent of its canonical anti-apoptotic and anti-autophagic functions observed in overexpression models^9-11. Notably, vBcl-2 is expressed in KS lesions, suggesting its involvement in disease pathogenesis^12,13.

KSHV vBcl-2 exhibits limited sequence identity with other Bcl-2 family proteins but maintains a similar globular structure¹⁴. Its Bcl-2 homology (BH) domains, BH1, BH2 and BH3, form a hydrophobic cleft (BH3-binding groove) that binds pro-apoptotic proteins (for example, BAX and BAK) and the pro-autophagic protein Beclin 1 (refs. 14,15), whereas the function of its N-terminal BH4 domain remains poorly defined. Although vBcl-2 is proposed to facilitate viral replication by inhibiting apoptosis and autophagy^12,15, substitution of vBcl-2 with other Bcl-2 homologues preserves these classical functions but fails to sustain efficient viral replication¹¹. Furthermore, mutations disrupting its BH3-binding groove, thus abolishing anti-apoptotic and anti-autophagic functions, do not impact viral replication⁹, suggesting an alternative mechanism. Here we demonstrate that vBcl-2 exploits a host nucleoside diphosphate (NDP) kinase to reshape mitochondrial structure, subverting mitochondria-based immune responses to facilitate efficient virion morphogenesis, a therapeutically targetable process crucial for viral propagation.

NM23-H2 is a target of KSHV vBcl-2

To identify vBcl-2-interacting protein(s) important for viral replication, we performed a yeast two-hybrid (Y2H) screen using vBcl-2 as bait and identified the NDP kinase NM23-H2 as a key candidate (Supplementary Table 1). NM23-H2 is one of ten human NDP kinase isoforms (NM23-H1 to NM23-H10) catalysing conversion of NDPs to nucleoside triphosphates (NTPs)¹⁶. Direct in vitro binding assays validated this interaction (Fig. 1a), and co-immunoprecipitation confirmed selective binding of vBcl-2 with endogenous and exogenous NM23-H2, but not with the highly homologous NM23-H1, in BJAB and HEK293T cells (Fig. 1b and Extended Data Fig. 1a,b). To further confirm this interaction during viral replication, we employed iSLK cells latently infected with KSHV BAC16 expressing a fully functional HA-tagged wild-type (WT) vBcl-2 on the viral genome (iSLK-BAC16.vBcl-2 WT)⁹. The cells also express a doxycycline (Dox)-inducible replication and transcription activator (RTA) to trigger lytic reactivation¹⁷. vBcl-2 co-immunoprecipitated with NM23-H2 (not with NM23-H1) upon Dox-induced viral reactivation (Fig. 1c).

Fig. 1 |. vBcl-2 interaction with NM23-H2 is required for KSHV virion production.

a, In vitro GST pull-down (GST PD) assay showing direct interaction between recombinant GST-vBcl-2 and NM23-H2, detected by immunoblotting (IB). Coomassie staining shows the input proteins. b,c, Co-immunoprecipitation (Co-IP) of endogenous NM23-H2 with vBcl-2 in whole-cell lysate (WCL) of BJAB cells stably expressing a vector (Vec) or HA-vBcl-2 (b), and in iSLK-BAC16.vBcl-2 WT cells treated (+) or untreated (−) with Dox and sodium butyrate (NaB) for 60 h (c). d, Co-IP of Flag-NM23-H2, Beclin 1 and BAK with HA-vBcl-2 (WT/mutants). e,f, iSLK-BAC16.vBcl-2 (WT/KO) cells stably expressing Vec or HA-vBcl-2 (WT/mutants) were treated with Dox/NaB for 72 h. Virus-containing supernatants were collected to infect naïve SLK cells. Representative GFP and bright-field (BF) images of SLK cells are shown (e, top) and vBcl-2 expression was detected by IB (e, bottom). Infectious titre was determined by flow cytometry for GFP-positive SLK cells (f, n = 3). g, Co-IP of Flag-NM23-H2 (WT/mutants) with HA-vBcl-2.LC, light chain. h–j, iSLK-BAC16.vBcl-2 WT cells transduced with control or NM23-H2 shRNA along with 3×HA-NM23-H2 rescue plasmids were Dox/NaB treated. Supernatants were collected to infect SLK cells and representative GFP/BF images are shown in h. Infectious titres were determined and are shown in i (n = 3). Protein expression is shown in j. Data in a–e, g, h and j are from one experiment that is representative of three independent experiments (see Source Data for uncropped western blot (WB) images). Data in f and i represent mean ± s.d. analysed using one-way ANOVA with Tukey’s post hoc test; ***P < 0.001, ****P < 0.0001; NS, not significant. Exact P values are provided in Source Data Fig. 1.

Domain mapping revealed that the N-terminal BH4 domain of vBcl-2 is essential for NM23-H2 interaction. Deletion of this region (amino acids 1–18) abolished binding, whereas mutations within the BH3-binding hydrophobic groove (amino acids 34–142; for example, vBcl-2 AAA^9,14) retained NM23-H2 interaction (Fig. 1d and Extended Data Fig. 1c,d). Notably, the E14A mutation in the BH4 domain, previously associated with impaired KSHV replication⁹, specifically disrupted NM23-H2 interaction, without affecting binding to BAK or Beclin 1 (Fig. 1d). Consistently, vBcl-2 E14A inhibited Torin1-induced autophagy as effectively as WT vBcl-2, as indicated by levels of autophagosome-associated LC3 (LC3-II) and the turnover of the autophagic cargo protein p62 in cells treated with mTOR inhibitor Torin 1 (Extended Data Fig. 1e,f). In contrast, the vBcl-2 AAA mutant, unable to bind Beclin 1, failed to inhibit autophagy⁹. Similarly, vBcl-2 E14A (but not vBcl-2 AAA) blocked apoptosis induced by TNF and cycloheximide to an extent comparable to WT (Extended Data Fig. 1g). Thus, vBcl-2–NM23-H2 interaction represents a distinct functional pathway independent of autophagy and apoptosis inhibition.

vBcl-2–NM23-H2 interaction is required for virion production

To address the functional importance of vBcl-2–NM23-H2 interaction in KSHV lytic infection, we compared virion production between the previously reported⁹ vBcl-2-null KSHV BAC16 (BAC16.vBcl-2 KO) containing a stop codon in the N terminus of vBcl-2 and the WT control expressing HA-vBcl-2 (BAC16.vBcl-2 WT) in iSLK cells (Extended Data Fig. 2a). These cells express GFP from the BAC16 genome driven by the EF1α promoter, serving as a marker for viral infection⁸. Compared with iSLK-BAC16.vBcl-2 WT cells, iSLK-BAC16.vBcl-2 KO cells exhibited significantly reduced virion production following reactivation (Extended Data Fig. 2a-d), which was fully rescued by re-expression of WT vBcl-2 or the AAA mutant (Beclin 1/BAK-binding defective but NM23-H2 binding competent), but not by the E14A mutant (Beclin 1/BAK-binding competent but NM23-H2 binding defective) (Fig. 1e,f). To assess whether vBcl-2 is equally important in de novo lytic infection with cell-free virus, we infected HCT116 colorectal cancer cells, which support KSHV lytic replication, with equal titres of iSLK-BAC16-derived vBcl-2 WT or vBcl-2 KO virus. Unlike WT virus, vBcl-2 KO virus failed to yield infectious progeny (Extended Data Fig. 2e-h). These results demonstrate that the vBcl-2–NM23-H2 interaction is critical for efficient KSHV virion production.

NM23-H2 shares 88% sequence identity with NM23-H1 (Supplementary Fig. 1a). Structural superimposition of NM23-H1 and NM23-H2 revealed similar overall topologies, despite differences at specific residues (Supplementary Fig. 1b). Molecular docking analysis showed that NM23-H2 interacts with the N-terminal helix α1 of vBcl-2, consistent with our biochemical data (Supplementary Fig. 1c). Among the residues differing between NM23-H1 and NM23-H2, glutamine 50 (Q50) in NM23-H2 was crucial for interaction with vBcl-2 (Supplementary Fig. 1c). Substitution of Q50 with glutamic acid as in NM23-H1 (Q50E) severely impaired binding, whereas other residue substitutions (A37G, M38L, L41M, R42Q, E46D, H47L, P62A, N69H and K135H) minimally affected binding (Fig. 1g and Extended Data Fig. 2i,j). Notably, the enzymatically inactive NM23-H2 H118C mutant retained vBcl-2 binding, suggesting that the enzymatic activity of NM23-H2 is not required for this interaction (Fig. 1g).

The reduced virus yield observed with the NM23-H2-binding-defective E14A mutant suggested that NM23-H2 and its interaction with vBcl-2 are important for viral replication. Indeed, short hairpin RNA (shRNA)-mediated knockdown of NM23-H2 in iSLK-BAC16.vBcl-2 WT cells significantly decreased virion production, mirroring the phenotype observed in iSLK-BAC16.vBcl-2 KO cells (Fig. 1h-j). Re-expression of shRNA-resistant WT NM23-H2 rescued this defect, whereas neither the NM23-H2-binding-defective Q50E mutant nor the enzyme-dead H118C mutant restored virus yield (Fig. 1h-j). In contrast, NM23-H1 depletion did not reduce virus yield; instead, an increase in virion production was observed (Extended Data Fig. 2k-m). These results establish NM23-H2 as a critical partner for vBcl-2-mediated KSHV lytic replication, distinct from its anti-autophagic and anti-apoptotic roles.

vBcl-2 activates and relocates NM23-H2 to mitochondria

The role of vBcl-2 in regulating host NDPK was previously unknown. We found that expression of WT vBcl-2, but not the NM23-H2 binding-deficient E14A mutant, stabilized NM23-H2 protein levels without altering its mRNA expression (Extended Data Fig. 3a and Supplementary Fig. 2a). Since NM23 proteins function as a hexamer¹⁸, we examined whether vBcl-2 regulates NM23-H2 oligomerization. Cross-linking assays showed that high-molecular-mass oligomeric forms of NM23-H2 increased upon expression of vBcl-2 WT but not the E14A mutant, suggesting that vBcl-2 promotes NM23-H2 oligomerization through direct interaction (Fig. 2a,b). Moreover, vBcl-2 WT, but not the E14A mutant, enhanced NM23-H2 NDPK activity (~3-fold) (Extended Data Fig. 3b), which correlated with increased phosphorylation at the catalytic residue H118 (Extended Data Fig. 3c,d).

Fig. 2 |. vBcl-2 activates and relocates NM23-H2 to mitochondria.

a,b, IB of Flag-NM23-H2 oligomers in HEK293T cells co-transfected with/without HA-vBcl-2 (WT/E14A) and treated with dimethylsulfoxide (DMSO) or disuccinimidyl suberate (DSS) (a). NM23-H2 multimers versus monomers were quantified (b; n = 3 independent experiments). c, Representative STED images of vBcl-2 (red), NM23-H2 (magenta) and TOM20 (green) in iSLK-BAC16.vBcl-2 KO cells expressing mCherry-vBcl-2 (WT/mutants) at 60 h post reactivation. Insets: vBcl-2/TOM20, vBcl-2/NM23-H2 and NM23-H2/TOM20 co-localization. White arrowheads indicate mitochondria-associated NM23-H2. Scale bars, 10 μm. d–f, Quantification of the fraction of NM23-H2 co-localized with TOM20 (d) or vBcl-2 (e), and vBcl-2 co-localization with mitochondria (f) in c (n = 30 cells per group, 3 independent experiments). g–j, IB of mitochondrial/cytosolic fractions (left) and WCL (right) from iSLK-BAC16.vBcl-2 KO cells stably expressing HA-vBcl-2 (WT/mutants) at 60 h post reactivation (g). S.E., short exposure; L.E., long exposure. Relative levels of mitochondrial NM23-H2 (h), cytoplasmic NM23-H2 (i) and mitochondrial vBcl-2 (j) in g are normalized to WCL protein levels (n = 3). k,l, PLAs of NM23-H2-TOM20 in iSLK.BAC16.vBcl-2 KO cells expressing HA-vBcl-2 (WT/mutants), 60 h post reactivation (k). DAPI, nuclei (blue). Insets: PLA puncta (white). PLA puncta per cell (normalized to total NM23-H2 levels) are quantified (l; n = 100 cells, 3 independent experiments). Data in b and h–j are presented as mean ± s.d. and analysed using one-way ANOVA with Tukey’s post hoc test. For boxplot data in d–f and l, the centre line represents the median, the box limits denote the interquartile range (IQR, 25th to 75th percentiles), and whiskers extend from minimum to maximum values; data analysed using Kruskal–Wallis with post hoc Dunn’s test; *P < 0.05, **P < 0.01, ****P < 0.0001. Exact P values are provided in Source Data Fig. 2. See Source Data for uncropped WB images of a and g.

NM23-H2 is mainly found in the cytoplasm¹⁹ but has also been detected in the nucleus²⁰, whereas vBcl-2 was previously reported to localize to mitochondria and nuclei¹¹. Using confocal microscopy, we observed extensive co-localization of HA-vBcl-2 with the outer mitochondrial membrane (OMM) protein TOM20 in HeLa cells but minimal overlap with markers for other organelles including the endoplasmic reticulum (PDI), Golgi (TGN46) and plasma membrane (WGA) (Extended Data Fig. 3e,f). A near-complete overlap was observed between co-expressed vBcl-2 and NM23-H2 that outlined TOM20-positive mitochondria (Extended Data Fig. 3e,g). To assess whether vBcl-2 alters NM23-H2 localization during viral replication, we generated iSLK-BAC16.vBcl-2 KO cells expressing mCherry-tagged vBcl-2 WT, AAA and E14A mutants (Fig. 2c). Super-resolution microscopy of KSHV-reactivated cells expressing WT or AAA (NM23-H2-binding competent) vBcl-2 revealed substantial redistribution of NM23-H2 from the cytosol to a more punctate pattern co-localizing with TOM20, whereas the E14A mutant had no effect (Fig. 2c-f). Subcellular fractionation (Fig. 2g-j) and proximity ligation assays (PLAs) (Fig. 2k,l) confirmed increased mitochondrial localization of NM23-H2 (but not NM23-H1) specifically with WT and AAA vBcl-2, but not E14A.

Although NM23-H2 was previously implicated in transcriptional regulation²⁰, we did not detect nuclear staining of NM23-H2 in HeLa cells (Extended Data Fig. 3e) or in KSHV-reactivated iSLK-BAC16 cells (Fig. 2c). Furthermore, vBcl-2 overexpression in BJAB cells neither broadly altered the transcriptome nor changed the expression of known NM23-regulated genes (Supplementary Fig. 2b,c), underscoring a nucleus-independent mechanism by which vBcl-2 recruits NM23-H2 to mitochondria, promoting its oligomerization and enhancing its enzymatic activity during KSHV lytic replication.

vBcl-2–NM23-H2 interaction induces mitochondrial fission

To explore the functional consequence of mitochondrial recruitment of NM23-H2 by vBcl-2, we analysed mitochondrial morphology in KSHV-reactivated iSLK cells. Upon viral reactivation, over 50% of iSLK-BAC16.vBcl-2 WT cells displayed fragmented mitochondria by 60–72 h post reactivation, compared with less than 5% under non-induced conditions (Extended Data Fig. 4a,b). Fragmentation was evidenced by increased numbers of individual mitochondria per cell and reduced mitochondrial area and perimeter (Extended Data Fig. 4c-e). This phenotype was absent in iSLK-BAC16.vBcl-2 KO cells (Extended Data Fig. 4a-e) but was restored by re-expression of WT vBcl-2 or the AAA mutant, whereas the NM23-H2-binding-deficient E14A mutant had no effect (Fig. 3a,b). Transmission electron microscopy (TEM) confirmed that expression of vBcl-2 WT and vBcl-2 AAA shortened mitochondria and reduced cristae density compared with the control or vBcl-2 E14A (Fig. 3a-d). Similar mitochondrial fragmentation occurred in HCT116 cells following de novo infection with vBcl-2 WT but not KO virus (Supplementary Fig. 3a-c). In fact, ectopic expression of vBcl-2 WT or vBcl-2 AAA (but not E14A) in SLK cells could induce mitochondrial fragmentation, independent of viral replication, which was abolished by depletion of NM23-H2, reducing it to baseline levels (Supplementary Fig. 3d). Consistently, silencing NM23-H2 in KSHV-reactivated iSLK-BAC16.vBcl-2 WT cells resulted in elongated mitochondria, whereas NM23-H1 depletion showed no effect (Extended Data Fig. 4f,g). Importantly, only WT NM23-H2, but not NM23-H2 mutants defective in vBcl-2 binding (Q50E) or kinase activity (H118C), restored mitochondrial fragmentation in rescue experiments (Extended Data Fig. 4f,g). These results indicate that vBcl-2 promotes mitochondrial fragmentation specifically via its interaction with NM23-H2.

Fig. 3 |. vBcl-2 interaction with NM23-H2 induces DRP1-dependent mitochondrial fission.

a, Confocal (top) and TEM (bottom) micrographs of mitochondria in cells expressing Vec or HA-vBcl-2 (WT/mutants) at 60 h post reactivation. Arrows denote fragmented mitochondria. b–d, Quantification of mitochondrial fragmentation from confocal images (b, n = 3, 300 cells per sample), and TEM analysis of mitochondrial length (c) and cristae frequency (d; normalized to length) (n = 36 mitochondria per sample). e–h, Confocal micrographs of mitochondria (e; top) in iSLK-BAC16.vBcl-2 cells expressing control (Ctrl) versus DRP1 shRNAs, Vec versus Flag-DRP1 K38A (green) or treated with Mdivi-1 post reactivation. Mitochondrial fragmentation was quantified in f (n = 3, 100 cells per sample). GFP images (e, bottom) depict virus yield, with titre quantified in g (n = 3) and protein expression indicated in h. i, Co-IP of proteins in cells expressing HA-vBcl-2 (WT/mutants). See Source Data for uncropped images. j–l, Immuno-EM (j) of mitochondrial (M) co-localization of NM23-H2 and DRP1 in iSLK-BAC16.vBcl-2 KO expressing vBcl-2 (WT/mutants). NM23-H2 clustered with DRP1 (white arrowheads) were quantified in k (n = 30 mitochondria per sample). DRP1 particles (red arrows) were quantified (l; n = 40 mitochondria per sample). Scale bars, 10 μm (unless stated otherwise). Data in b, f and g are mean ± s.d. and analysed using one-way ANOVA with Tukey’s post hoc test. For boxplot data in c, d, k and l, the centre line represents the median, the box limits denote the IQR (25th to 75th percentiles), and whiskers extend from minimum to maximum values; data analysed using Kruskal–Wallis with post hoc Dunn’s test (c,d) or two-tailed Mann–Whitney test (k,l); **P < 0.01, ***P < 0.001, ****P < 0.0001. Exact P values are provided in Source Data Fig. 3.

Mitochondrial morphology is regulated by fusion proteins, such as mitofusins (MFN1, MFN2) for outer membrane fusion and optic atrophy1 (OPA1) for inner membrane fusion, and the fission protein dynamin-related protein (DRP1), which is recruited to the OMM by mitochondrial fission factor (MFF) and induces membrane constriction through GTP hydrolysis^21-23. Protein levels of these regulators remained unchanged following vBcl-2 expression (Extended Data Fig. 4h). Moreover, overexpression of fusion protein MFN2 in Dox-induced iSLK-BAC16. vBcl-2 WT cells failed to reverse mitochondrial fragmentation, whereas silencing DRP1 or MFF effectively prevented vBcl-2-induced fragmentation and reduced viral yields (Fig. 3e-h and Extended Data Fig. 4i-k). Supporting this notion, DRP1 inhibition using Mdivi-1, a commonly used DRP1 inhibitor²⁴, or expression of dominant-negative DRP1 (K38A mutant²⁵) markedly reduced mitochondrial fragmentation and viral yields in iSLK-BAC16.vBcl-2 WT cells (Fig. 3e-h), reminiscent of the effects observed following the loss of vBcl-2 or NM23-H2. These results indicate that DRP1 is required for virus-induced mitochondrial fission that favours KSHV lytic infection.

Mitochondrial fission can be influenced by perturbations in mitochondrial integrity²³. However, factors such as mitochondrial membrane potential (ΔΨm), mitochondrial reactive oxygen species (mtROS) production, or mitochondrial DNA (mtDNA) release were unaffected by vBcl-2 expression post reactivation (Supplementary Fig. 4a-c). Although mitochondrial fission has been associated with mitophagy induction²⁶, analysis of the mitochondrial localization of the mitophagy marker PINK1 (ref. 27) showed no significant differences between WT and mutant vBcl-2-expressing cells (Extended Data Fig. 5a). Thus, vBcl-2-induced mitochondrial fission is not a passive consequence of mitochondrial damage but represents an active, virus-driven process facilitating KSHV lytic replication.

vBcl-2 promotes NM23-H2/DRP1 complex and DRP1 activation

Previous studies indicate that NM23 proteins can locally supply GTP to dynamin-related proteins, facilitating membrane remodelling²⁸. Given that vBcl-2 recruits and activates NM23-H2 at mitochondria, we investigated whether vBcl-2 promotes mitochondrial fission by stimulating DRP1 activity via NM23-H2. In HEK293T cells, DRP1 co-immunoprecipitated with NM23-H2 and vBcl-2; depletion of NM23-H2 disrupted this interaction (Extended Data Fig. 5b). Similarly, in KSHV-reactivated cells, WT vBcl-2 and vBcl-2 AAA increased NM23-H2 binding to DRP1, whereas the NM23-H2-binding-deficient E14A mutant did not (Fig. 3i). Notably, the kinase-dead NM23-H2 H118C mutant failed to bind DRP1 even in the presence of vBcl-2 (Extended Data Fig. 5c), suggesting that NM23-H2 enzymatic activity is critical for DRP1 complex formation. Conversely, the GTP-binding-deficient DRP1 K38A mutant retained undiminished interaction with NM23-H2, indicating that DRP1 GTP binding is dispensable for this interaction (Extended Data Fig. 5c).

Using purified proteins, we assessed whether NM23-H2 directly stimulated DRP1 GTPase activity, a key event in mitochondrial fission, through an in vitro GTPase assay²⁸. DRP1 GTPase activity was activated specifically by GTP, but not by ATP and/or GDP. Notably, recombinant NM23-H2 increased DRP1 activity even without added GTP, utilizing GDP and ATP as substrates (Extended Data Fig. 5d). Addition of WT vBcl-2 further enhanced DRP1 activity by ~60%, a boost absent with vBcl-2 E14A (Extended Data Fig. 5d). Immunogold electron microscopy (EM) of KSHV-reactivated iSLK-BAC16.vBcl-2 KO cells confirmed mitochondrial co-clustering of NM23-H2 and DRP1 upon WT vBcl-2 expression (Fig. 3j-l). In contrast, cells expressing vBcl-2 E14A lacked these clusters of NM23-H2/DRP1, although the overall number of mitochondrial DRP1 particles remained similar (Fig. 3j-l). A previous report implicated NM23-H3 at the OMM that may function as a GTP supplier²⁸; however, NM23-H3 knockdown did not affect mitochondria dynamics or virion production in KSHV-reactivated iSLK-BAC16.vBcl-2 WT cells (Supplementary Fig. 5a-d), highlighting the specific role of NM23-H2 in KSHV-mediated mitochondrial regulation.

Together, our findings support a model wherein vBcl-2 recruits and activates NM23-H2 at mitochondria, facilitating NM23-H2/DRP1 complex formation. This promotes local GTP supply to DRP1, driving mitochondrial fission essential for KSHV virion production (Supplementary Fig. 6).

Mitochondrial structural changes reshape immune responses

The mitochondrial network serves as a key platform for activating mitochondrial antiviral signalling protein (MAVS), which forms fibrous aggregates on the OMM to trigger type-I IFN responses^29,30. Given that vBcl-2 induces mitochondrial fission, we examined whether vBcl-2 enhances virion production by suppressing IFN signalling. Upon viral reactivation, iSLK-BAC16 cells infected with vBcl-2 KO virus showed significantly increased IFN-β induction compared with WT virus, particularly at 96 h post reactivation (Extended Data Fig. 6a-c). Re-expression of vBcl-2 WT or vBcl-2 AAA, but not vBcl-2 E14A, in iSLK-BAC16.vBcl-2 KO cells suppressed IFN-β induction, IFN-β secretion and IFN-stimulated gene (ISG) expression, including ISG15 (Fig. 4a-c). Analogous results were observed in Dox-treated TREx BCBL-1-RTA cells³¹, a KSHV-infected PEL cell line carrying a Dox-inducible RTA gene; depletion of vBcl-2 enhanced IFN-β responses (Extended Data Fig. 6d-f). Higher IFN-β expression was also noted during de novo infection with vBcl-2 KO virus compared with WT virus or mock infection (Extended Data Fig. 6g). In fact, expression of vBcl-2 (WT and AAA) per se, but not vBcl-2 E14A, inhibited polyinosine-polycytidylic acid [poly(I-C)]-induced IFN-β responses (Supplementary Fig. 7a-c), an effect reversed by DRP1 knockdown (Supplementary Fig. 7d,e).

Fig. 4 |. Mitochondrial fission dampens virus-induced innate immune response.

a–c, RT–qPCR of IFN-β mRNA (a) and ISG15 mRNA expression (c) in iSLK-BAC16.vBcl-2 KO cells stably expressing Vec, HA-vBcl-2 (WT/mutants) post reactivation for the indicated time. Fold changes were normalized to 18S rRNA. IFN-β in supernatant was measured by enzyme-linked immunosorbent assay in b. n = 3. d, Representative STED super-resolution micrographs of MAVS (magenta) and mitochondria (TOM20, green) in iSLK-BAC16.vBcl-2 WT/KO cells stably expressing mCherry-vBcl-2 (WT/mutants) at 60 h post reactivation. Insets: MAVS distribution. Arrows indicate MAVS puncta and dotted lines denote clustering. Scale bars, 10 μm. e, Diameter of MAVS clusters measured from images in d (n = 99). f, Quantification of MAVS co-localization with TOM20 in d (n = 100 cells per sample). g, Co-IP of endogenous TBK1 with MAVS in iSLK-BAC16.vBcl-2 KO cells stably expressing HA-vBcl-2 (WT/mutants) at 60 h post reactivation. WCLs were immunoblotted for P-TBK1 and IRF3 dimerization, and mitochondrial fraction was used for SDD–AGE analysis of MAVS aggregation. See Source Data for uncropped images. h–j, Densitometric quantification of MAVS aggregation (h), the ratios of p-TBK1/t-TBK1 (i) and IRF3 dimer/monomer (j) in g. p-TBK1, S172-phosphorylated TBK1; t-TBK1, total TBK1. n = 3. Data in d and g are from one experiment that is representative of three independent experiments. Data in a–c and h–j are mean ± s.d. analysed using one-way ANOVA with Tukey’s post hoc test. For boxplot data in e and f, the centre line represents the median, the box limits denote the IQR (25th to 75th percentiles), and whiskers extend from minimum to maximum values; data analysed using Kruskal–Wallis with post hoc Dunn’s test; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Exact P values are provided in Source Data Fig. 4.

Super-resolution stimulated emission depletion (STED) microscopy revealed punctate MAVS distribution in Dox-treated iSLK-BAC16. vBcl-2 WT cells, while vBcl-2 KO cells exhibited fibrous and clustered MAVS staining (Fig. 4d,e and Supplementary Fig. 8), indicative of enhanced MAVS aggregation^30,32, despite their unchanged mitochondrial localization (Fig. 4d-f). Re-expression of vBcl-2 WT or vBcl-2 AAA, but not vBcl-2 E14A, reversed this phenotype in Dox-treated iSLK-BAC16.vBcl-2 KO cells (Fig. 4d-f). Semi-denaturing detergent agarose gel electrophoresis (SDD–AE) analysis confirmed reduced MAVS aggregation in cells expressing vBcl-2 WT and vBcl-2 AAA compared with vBcl-2 E14A, which correlated with impaired MAVS interaction with TANK binding kinase 1 (TBK1), reduced TBK1 phosphorylation and IRF3 dimerization (Fig. 4g,h-j). Production of mini-MAVS, a truncated variant that antagonizes full-length MAVS³³, remained unaffected (Fig. 4g). Notably, depletion of NM23-H2 or DRP1 in KSHV-reactivated iSLK-BAC16.vBcl-2 WT cells reversed vBcl-2-mediated inhibition, resulting in restored MAVS aggregation and IFN-β responses (Extended Data Fig. 6h-j). Thus, vBcl-2-induced mitochondrial fission, mediated by the NM23-H2–DRP1 axis, subverts MAVS-dependent mitochondrial antiviral immunity.

Mitochondrial fission is required for virion morphogenesis

Despite its importance in KSHV virion production, the exact stage at which vBcl-2 functions in the viral lifecycle remains unclear. To clarify this, we examined viral replication kinetics in iSLK-BAC16 cells within 48 h post reactivation, before potential re-infection by released virions. No significant differences were observed between vBcl-2 WT and KO viruses in the mRNA expression of latent (LANA), immediate early (RTA), early (ORF57, ORF11) and late (ORF25) viral genes (Extended Data Fig. 7a). Likewise, viral protein levels and DNA copy numbers were unaffected by loss or re-expression of WT, AAA or E14A vBcl-2 mutants (Extended Data Fig. 7b,c), suggesting that vBcl-2 is largely dispensable for early replication. We next explored vBcl-2’s role in late-stage viral assembly using TEM. iSLK-BAC16.vBcl-2 KO cells expressing WT or AAA vBcl-2 showed abundant nucleocapsids in the nucleus, and enveloped virions in the cytoplasm and extracellular space at 72 h post reactivation (Fig. 5a-c). Conversely, cells expressing vBcl-2 E14A displayed impaired virion formation, characterized by nuclear aggregates with irregular densities (Fig. 5a-c), indicative of disrupted viral assembly compartments³⁴. Detailed analyses revealed that cells expressing vBcl-2 WT or AAA contained angular capsids at multiple maturation stages, including capsids that retained the scaffold but lacked DNA (B-capsid), capsids that expelled the scaffold but failed to encapsidate viral DNA (A-capsid) and mature nucleocapsids (C-capsid) (Supplementary Fig. 9a and Supplementary Table 2). In contrast, vBcl-2 E14A-expressing cells largely lacked mature angular capsids (Supplementary Table 2). As proper capsid assembly enables the transmission of viral nucleocapsids to the cytoplasm, we assessed the effect of vBcl-2 on nuclear egress, using a small KSHV capsid protein ORF65 as an indicator³⁵. KSHV-reactivated cells expressing vBcl-2 E14A showed pronounced nuclear accumulation of ORF65, whereas WT- and AAA-expressing cells exhibited ORF65 dispersal into perinuclear and cytoplasmic areas, suggesting progressive nuclear egress (Fig. 5d-f). Consistently, quantitative (q)PCR analysis showed significantly reduced cytoplasmic-to-nuclear viral DNA ratios in cells expressing E14A post reactivation (Supplementary Fig. 9b). Knockdown of NM23-H2 or DRP1 similarly inhibited ORF65 cytoplasmic release in iSLK-BAC16.vBcl-2 WT cells following reactivation (Extended Data Fig. 7d-f). Moreover, during de novo infection, the vBcl-2 KO virus failed to transport ORF65 into the cytoplasm at 60 h post infection (Extended Data Fig. 7g). These results indicate that vBcl-2-mediated mitochondrial fission specifically promotes late-stage virion assembly and egress.

Fig. 5 |. Mitochondrial fission supports virion assembly through TRIM22 and MxB suppression.

a, TEM of viral particles in iSLK-BAC16.vBcl-2 KO cells expressing HA-vBcl-2 (WT/mutants) 72 h post reactivation. Red arrows, mature nucleocapsids; red arrowheads, enveloped viral particles. Cyto, cytoplasm; Nu, nucleus. b,c Quantification of mature nucleocapsids in the nucleus (b; n = 17 cells per group) and extranuclear space (c; n = 3 independent experiments) in a. d–f, Confocal micrographs of ORF65 (red) in iSLK-BAC16.vBcl-2 KO cells (GFP) expressing HA-vBcl-2 (WT/mutants) post reactivation (d). DAPI, nucleus (blue). The percentage of cells with nuclear/cytoplasmic ORF65 (e; n = 3, 300 cells per sample) and percentage of cytoplasmic ORF65 per cell (f; n = 51 cells per sample) were quantified. g–j, Confocal micrographs of ORF65 (red) in cells transduced with Ctrl or MAVS shRNA post reactivation (g). Indicated protein expression is shown (h). Percentage of cells with cytoplasmic ORF65 (i) and percentage of cytoplasmic ORF65 per cell (j) were quantified (n = 3, 300 cells per sample in i; n = 51 cells per sample in j). k–m, Virus-containing supernatants from Dox/NaB-treated (72 h) iSLK-BAC16.vBcl-2 WT/KO cells transduced with indicated shRNAs were collected to infect SLK cells. Representative GFP/BF images of SLK cells are shown (k), virus titre determined by flow cytometry (l, n = 3) and mRNA expression of indicated genes in k measured by RT–qPCR (m; n = 3). n–p, Virion production in iSLK-BAC16.vBcl-2 WT/KO cells expressing TRIM22 and MxB post reactivation, presented as GFP images in n and quantified in o (n = 3). Indicated gene expression is in p. Scale bars, 10 μm (unless stated otherwise). Data in a, d, g, h, k, n and p are from one experiment that is representative of three independent experiments. See Source Data for uncropped data of h and p. Data in c, e, i, l, m and o are mean ± s.d. analysed using one-way ANOVA with Tukey’s post hoc test. Scatterplots in b, f and j were analysed using Kruskal–Wallis with post hoc Dunn’s test; **P < 0.01, ***P < 0.001, **** P < 0.0001. Exact P values are provided in Source Data Fig. 5.

Suppression of TRIM22 and MxB promotes virion morphogenesis

Our findings indicate that mitochondrial fission induced by vBcl-2 dampens MAVS-mediated IFN signalling and ISG expression. We hypothesized that failure to suppress IFN responses underlies impaired virion assembly and egress in vBcl-2 KO viruses. Indeed, knockdown of MAVS, IFNAR1 (type-I IFN receptor) or IRF3 in Dox-treated iSLK-BAC16. vBcl-2 KO cells restored nuclear-to-cytoplasmic trafficking of ORF65 (Fig. 5g-j and Supplementary Fig. 10a-d) and rescued virion production (Fig. 5k-m), highlighting the MAVS-dependent IFN response as the key factor restricting late-stage morphogenesis upon vBcl-2 loss.

To identify specific ISGs responsible for restricting virion production in the absence of vBcl-2, we examined eight ISGs known to inhibit virus assembly or release (Tetherin, Viperin, VPS4A, CHMP4B, CH25H, ISG15, TRIM22 and MxB)³⁶. All eight ISGs were upregulated in vBcl-2 KO cells (Supplementary Fig. 11a), consistent with increased IFN signalling. Individual depletion of Tetherin, Viperin, VPS4A and CHMP4B showed minimal effects, while CH25H and ISG15 knockdown modestly increased ORF65 nuclear-to-cytoplasmic translocation (Extended Data Fig. 8a-c). Notably, silencing TRIM22 or MxB markedly enhanced ORF65 cytoplasmic release and virion yield; and combined TRIM22 and MxB depletion showed additive effects, increasing viral yield by nearly 80% (Extended Data Fig. 8a-e). Both proteins predominantly localized near the nuclear membrane and/or within the nucleus in infected cells (Supplementary Fig. 11b), and their expression also increased during de novo infection with vBcl-2 KO virus (Extended Data Fig. 8f). Given that MxB restricts HSV-1 capsid assembly³⁶ and TRIM22 broadly inhibits various viruses including HSV-1 (ref. 37), we further validated their antiviral role in KSHV infection. Ectopic expression of TRIM22 or MxB reduced virion yield from KSHV-reactivated iSLK-BAC16.vBcl-2 WT cells (~100-fold) to levels comparable to those from reactivated vBcl-2 KO cells, without further effect in the absence of vBcl-2 (Fig. 5n-p). This reduction correlated with increased nuclear retention of ORF65 without affecting overall protein expression (Extended Data Fig. 8g-i). Similar inhibitory defects were observed in TREx BCBL-1-RTA cells, where overexpression of TRIM22 or MxB reduced cytoplasmic ORF65 and extracellular viral genomes by ~50% (Supplementary Fig. 12a-c). Thus, these results identify TRIM22 and MxB as critical antiviral effectors that must be antagonized by KSHV to enable efficient virion assembly and egress.

Therapeutic targeting of the vBcl-2–NM23-H2 interaction

The critical role of vBcl-2 in virion morphogenesis suggests that it may serve as a promising therapeutic target for limiting viral dissemination. While previous Bcl-2-based inhibitors, such as ABT-263, primarily targeted the hydrophobic BH3-peptide binding groove³⁸, our data indicate that disrupting this region (as in vBcl-2 AAA) did not impair KSHV lytic replication (Fig. 1). In contrast, blocking the N-terminal interaction of vBcl-2 with NM23-H2 abolished virion production by activating mitochondrial antiviral responses. To identify inhibitors specifically disrupting the vBcl-2–NM23-H2 interaction, we conducted high-throughput homogeneous time-resolved fluorescence (HTRF) screening using purified recombinant vBcl-2 and NM23-H2 proteins, testing ~64,000 compounds (Extended Data Fig. 9a). The assay robustly detected vBcl-2–NM23-H2 complex formation in a dose-dependent manner but showed negligible interaction between vBcl-2 and NM23-H1, confirming the specificity of vBcl-2 for NM23-H2 (Extended Data Fig. 9b). After successive optimization and stringent counterscreens to eliminate non-specific hits, we identified Z56833834 (designated as vBcl-2–NM23-H2 inhibitor, VBNI-1) as the lead candidate, which potently and selectively inhibited the vBcl-2–NM23-H2 interaction (half-maximal inhibitory concentration (IC₅₀) ~64 nM) (Fig. 6a,b). Importantly, VBNI-1 did not disrupt vBcl-2 binding to the BH3-peptide even at concentrations exceeding 10 μM, exhibiting over 150-fold selectivity toward the vBcl-2–NM23-H2 interaction compared with the hydrophobic groove (Fig. 6b). Conversely, ABT-263 selectively inhibited BH3-peptide binding without significantly impacting the vBcl-2–NM23-H2 interaction (Extended Data Fig. 9c).

Fig. 6 |. Development of VBNI-1 as a small-molecule inhibitor of the vBcl-2–NM23-H2 interaction.

a, Chemical structure of VBNI-1 (Z56833834). b, HTRF assays showing the inhibitory activity of VBNI-1 for the vBcl-2–NM23-H2 interaction (left) and the vBcl-2–BH3 peptide (EDIIRNIARHLAQVGDSMDR) interaction (right), with IC₅₀ values determined. c–f, Co-IP (c,e) of indicated proteins with vBcl-2 in SLK.HA-vBcl-2 cells treated with different doses of VBNI-1 for 36 h (c,d) or VBNI-1 (2 μM) for different times (e,f). The respective changes in binding were quantified in d and f (n = 3). g,h, Virus yields in VBNI-1-treated iSLK-BAC16.vBcl-2 WT/KO cells post reactivation, determined by infection of SLK cells, are illustrated by GFP images (g) and quantified (h; n = 3). Scale bars, 430 μm. i, Cytotoxicity of VBNI-1 on cells in g (n = 3). j,k, Confocal micrographs of ORF65 (red) in iSLK-BAC16.vBcl-2 cells (j) treated with VBNI-1 (2 μM, 60 h) and in TREx BCBL-1-RTA cells (k) treated with VBNI-1 (4 μM, 48 h) post reactivation. The cytoplasmic staining of ORF65 was quantified (right; n = 3, 300 cells per sample). l–n, STED micrographs of MAVS (magenta) and TOM20 (red) in VBNI-1-treated cells post reactivation. The cells in l with fission and MAVS aggregation were quantified in m (n = 3, with 100 cells per sample) and n (n = 99), respectively. o, IB of P-TBK1 and P-IRF3 (S379) in VBNI-1-treated cells post reactivation. p, RT–qPCR of indicated transcripts in VBNI-1-treated iSLK-BAC16.vBcl-2 cells post reactivation (n = 3). Scale bars, 10 μm (unless stated otherwise). Data in c, e, g, j–l and o are from one experiment that is representative of three independent experiments. See Source Data for uncropped data of c, e and o. Data in d, f, h–k, m and p are presented as mean ± s.d. and analysed using two-tailed Student’s t-test and one-way ANOVA with Tukey’s post hoc test. For boxplot data in n, the centre line represents the median, the box limits denote the IQR (25th to 75th percentiles), and whiskers extend from minimum to maximum values; data analysed using two-tailed Mann–Whitney test; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Exact P values are provided in Source Data Fig. 6.

Further validation demonstrated that VBNI-1 treatment dose- and time-dependently blocked vBcl-2 co-immunoprecipitation with NM23-H2 in SLK (Fig. 6c-f) and HEK293T cells (Extended Data Fig. 9d-g) without affecting interactions of vBcl-2 with Beclin 1 or BAK. Moreover, VBNI-1 exhibited no inhibitory activity against NM23-H1 or NM23-H2 NDP kinase activity, suggesting that it is not an ATP-competitive kinase inhibitor (Extended Data Fig. 9h). Docking analysis³⁹ suggested that VBNI-1 binds to the N-terminal α1 helix of vBcl-2 with one of the top hits (energy of −5.6 kcal mol⁻¹) potentially disrupting the vBcl-2–NM23-H2 complex via a steric clash between vBcl-2-bound VBNI-1 and NM23-H2 residues (for example, K128) at the interface (Extended Data Fig. 9i). Surface plasmon resonance (SPR) assays confirmed selective binding of VBNI-1 to vBcl-2 with a dissociation constant (K_D) of 875 nM (Extended Data Fig. 9j). Importantly, minimal cytotoxicity was observed with VBNI-1 in naive SLK and BJAB cells (Extended Data Fig. 9k). Thus, VBNI-1 represents a selective, potent inhibitor targeting the vBcl-2–NM23-H2 interaction with low cytotoxicity.

VBNI-1 inhibits KSHV morphogenesis and virion production

To evaluate the antiviral efficacy of VBNI-1, we assessed its effect on KSHV replication using the established reactivation model¹⁷. Treatment with VBNI-1 dose-dependently reduced virion production in KSHV-reactivated iSLK-BAC16.vBcl-2 WT cells, resulting in over 80% inhibition at 8 μM (Fig. 6g,h and Extended Data Fig. 9l). No measurable VBNI-1 activity was detected in similarly treated iSLK cells but infected with vBcl-2 KO KSHV (Fig. 6g,h and Extended Data Fig. 9l). In addition, the observed antiviral activity was independent of cytotoxicity, as no significant reduction in cell viability occurred with VBNI-1 treatment (Fig. 6i). Similar to vBcl-2 deficiency, VBNI-1 did not impair viral DNA replication, transcription, protein expression or ORF65 levels (Supplementary Fig. 13a-c), but it specifically blocked late-stage virion assembly and egress in iSLK-BAC16.vBcl-2 WT and TREx BCBL-1-RTA cells post reactivation (Fig. 6j,k).

Mechanistically, VBNI-1 treatment retained tubulated mitochondrial structures and prevented mitochondrial fragmentation, leading to enhanced MAVS aggregation on mitochondria during viral reactivation (Fig. 6l-n). Increased MAVS aggregation subsequently activated TBK1 and IRF3 signalling (Fig. 6o), elevating IFN production and ISGs induction, notably of TRIM22 and MxB (Fig. 6p). Silencing TRIM22 and MxB reversed the antiviral effects of VBNI-1 (Extended Data Fig. 9n,o), indicating that these ISGs mediate the inhibitor’s antiviral activity. These findings establish VBNI-1 as a specific inhibitor targeting the vBcl-2–NM23-H2 interaction, which restores mitochondrial immune signalling and effectively inhibits KSHV virion production and dissemination.

Discussion

Mitochondrial membrane dynamics play a critical but context-dependent role in modulating innate immune responses during viral infections⁴⁰; however, the precise molecular mechanisms underlying these dynamics remain less understood. Our study demonstrates that KSHV-encoded vBcl-2, independent of its canonical anti-apoptotic and anti-autophagic functions, recruits and activates the host NDPK NM23-H2 to drive DRP1-dependent mitochondrial fission. While mitochondrial fission was previously reported to enhance antiviral defences during certain viral infections (for example, Dengue virus)⁴¹, KSHV exploits this process to suppress MAVS aggregation and IFN signalling to enable virion morphogenesis, which is otherwise restricted by ISGs such as TRIM22 and MxB (Extended Data Fig. 10). Considering that other gammaherpesviruses also encode Bcl-2 homologues, the mitochondrial fission that we discovered with KSHV may be a conserved evasion strategy among related viruses, as suggested by the fact that KSHV vBcl-2 can be functionally replaced by that of rhesus rhadinovirus (RRV)¹¹ and BHRF1, an Epstein–Barr virus (EBV)-encoded Bcl-2 homologue, also related to mitochondrial fission⁴².

The mechanistic insights provided by our study implicate a role for NM23-H2 as a facilitator of virus-induced mitochondrial fission via local regulation of GTP levels required for DRP1 activity. Although we did not observe an obvious defect in mitochondrial fission in uninfected cells upon NM23-H2 depletion, our observations do not exclude physiological roles of NM23-H2 in assisting DRP1 under normal conditions. Further studies using more sensitive assays or different cellular contexts may unravel previously unappreciated levels of complexity in the regulation of mitochondrial dynamics.

The antiviral effectors TRIM22 and MxB identified herein broaden our understanding of IFN-mediated defences against herpesviruses. Although TRIM22 and MxB individually restrict viral replication in different contexts^36,37, their combined action against KSHV suggests a complex and layered defence mechanism that vBcl-2 must counteract. Future research should focus on delineating whether these proteins directly disrupt viral structural components or host cellular processes critical for virion morphogenesis.

The identification of VBNI-1, a small-molecule inhibitor targeting the virus–host NM23-H2 interface, illustrates the potential therapeutic viability of disrupting mitochondrial dynamics exploited by viruses. Consistent with our mechanistic findings, this compound exhibits inhibitory effects on vBcl-2-induced fission, leading to increased MAVS-dependent IFN response that thwarts virion production in vitro (Extended Data Fig. 10). However, its efficacy in vivo and potential synergy with existing therapies require further evaluation. Given the broader implications of mitochondrial dynamics in the virulence of diverse viruses, development of similar agents could significantly enhance the repertoire of antiviral strategies, potentially extending beyond KSHV.

In conclusion, our study delineates a paradigm in which KSHV manipulates mitochondrial structure to create a permissive environment for viral morphogenesis. By dissecting this mechanism, we not only uncover new vulnerabilities in KSHV’s lifecycle but also highlight the potential of organelle-directed antiviral therapies.

Methods

Cells

Human embryonic kidney cells (HEK293T; ATCC, CRL-3216), HeLa cells (ATCC, CCL-2), SLK cells (provided by J.U.J.⁹) and HCT116 cells (ATCC, CCL-247) were cultured in Dulbecco’s modified Eagle’s medium (DMEM; Sigma-Aldrich); human Burkitt lymphoma B cells (BJAB; DSMZ, ACC 757) were maintained in RPMI 1640 (Sigma-Aldrich). Both media were supplemented with 10% (v/v) fetal bovine serum (FBS; Gibco), 100 U ml⁻¹ penicillin-streptomycin (pen-strep; GenClone) and 2 mM l-glutamine (Gibco). iSLK-BAC16.vBcl-2 WT and iSLK-BAC16.vBcl-2 KO cells (provided by J.U.J.⁹) were cultured in DMEM supplemented with 10% tetracycline (Tet)-free FBS (Takara Bio), 100 U ml⁻¹ pen-strep, 2 mM l-glutamine, 1 μg ml⁻¹ puromycin (Sigma-Aldrich), 250 μg ml⁻¹ G418 (Thermo Fisher) and 250 μg ml⁻¹ hygromycin B (Thermo Fisher), as previously described⁹. TREx BCBL-1-RTA cells (provided by J.U.J.³¹) were cultured in RPMI 1640 supplemented with 10% Tet-free FBS (Takara Bio), 100 U ml⁻¹ pen-strep, 2 mM l-glutamine, 1% sodium bicarbonate, 200 μg ml⁻¹ hygromycin B and 250 μg ml⁻¹ blasticidin (Cayman Chemical), as previously described³¹. All cell cultures were maintained at 37 °C in a humidified atmosphere containing 5% CO₂. Commercially obtained cells were authenticated by the vendors (ATCC and Thermo Fisher) and not revalidated by our laboratory. Cells received from other laboratories were used as provided without further authentication. All cell lines were routinely tested for mycoplasma contamination by PCR.

Plasmids

pCDH-HA-vBcl-2 WT and mutants in Extended Data Fig. 1c were provided by J.U.J.⁹. WT, AAA and E14A vBcl-2 variants were constructed by subcloning the cDNA of the full-length or mutant vBcl-2 into the KpnI/NotI sites of pcDNA5/FRT/TO (Invitrogen) with an N-terminal HA tag, or into the XbaI/BamHI sites of pCDH-CMV-MCS-EF1α-Puro (System Bioscience) with an N-terminal mCherry tag, or into the BamHI/EcoRI sites of pGEX-4T-1 (Amersham) with a GST tag. NM23-H2 cDNAs (including Q50E, H118C and shRNA-resistant NM23-H2) were subcloned into pcDNA5/FRT/TO between BamHI and EcoRI carrying an HA tag, or between KpnI and NotI with a Flag tag at the N terminus, or into pRSETB (Invitrogen) between BamHI and EcoRI carrying 6×His tag. DRP1 cDNA was subcloned into pcDNA5/FRT/TO between BamHI and NotI with an N-terminal Flag tag. Tetherin, Viperin, VPS4A, CHMP4B, ISG15, TRIM22 and MxB cDNAs were from Addgene. pCDH-TRIM22 and -MxB were constructed by subcloning the cDNA of the full-length TRIM22 and MxB into the XbaI/NotI sites of pCDH-CMV-EF1α-Puro (System Bioscience). All mutants were generated by site-directed mutagenesis using Q5 Site-Directed Mutagenesis kit (New England Biolabs), according to manufacturer instruction. All constructs were verified by Sanger sequencing (ABI PRISM 377, Applied Biosystems).

Transfection and gene silencing

Transient DNA transfections of HEK293T cells were performed using the Calcium Phosphate Transfection kit (Takara, 631312) or FuGENE HD transfection reagent (Promega, E2311) according to manufacturer protocols. For transient gene knockdown in TREx BCBL-1-RTA cells, 100 pmol of gene-specific small interfering RNA was electroporated into 3 × 10⁶ cells in 100 μl suspension using the Neon Electroporation transfection system (Invitrogen; 1,350 V, 40 ms pulse width), followed by 48 h of culture. Knockdown efficiency was determined by RT–qPCR using gene-specific probes listed in Supplementary Table 3. For lentiviral gene delivery, pCDH-mCherry-vBcl-2 (WT, AAA and E14A) or pCDH-HA-vBcl-2 (WT, AAA and E14A) constructs were used. For shRNA-mediated gene knockdown, shRNA targeting NM23-H1, NM23-H2, NM23-H3, DRP1, MFF, MAVS, IRF3, IFNAR1, Tetherin, Viperin, VPS4A, CHMP4B, CH25H, ISG15, TRIM22 and MxB, and a scrambled control shRNA were obtained from the TRC lentiviral shRNA libraries. Lentiviruses were produced by co-transfecting HEK293T cells with the lentiviral expression or shRNA plasmids along with packaging constructs pCDH-VSV-G and pCDH-dR8.91, using the calcium phosphate method. Viral supernatants were collected at 60 h post transfection, filtered through a 0.45-μm PES filter (VWR) and concentrated with Lenti-X Concentrator (Takara, 631232). Target cells (SLK, iSLK-BAC16, HEK293T, BJAB) were infected in the presence of 8 μg ml⁻¹ polybrene (Sigma-Aldrich, H9268) for 12 h, then cultured in fresh complete medium. Infected cells were selected at 48 h post infection with 1 μg ml⁻¹ puromycin or 250 μg ml⁻¹ hygromycin B, depending on the selection marker of the lentiviral plasmids. For CRISPR/Cas9-mediated knockout of MAVS, the lentiCRISPRv2 Blast vector (Addgene, 98293) containing the single guide RNA (sgRNA) targeting MAVS was provided by P.F. (University of Southern California (USC)). Lentivirus was produced by co-transfecting HEK293T cells with lentiCRISPRv2 Blast-sgMAVS, pMD2.G (Addgene, 12259) and psPAX2 (Addgene, 12260) for 60 h. iSLK.BAC16 cells were infected with the lentivirus for 48 h, followed by selection with 5 μg ml⁻¹ blasticidin (Cayman Chemical, 14499) for 1 week. Single clones were isolated by limiting dilution, expanded and validated by immunoblot analysis. The sequences of shRNAs and sgRNAs are provided in Supplementary Table 4.

Virus infection, induction and titration

KSHV vBcl-2 WT or vBcl-2 KO virus was produced using iSLK-BAC16. vBcl-2 WT or iSLK-BAC16.vBcl-2 KO cells (provided by J.U.J.⁹). Cells were induced with 1 μg ml⁻¹ doxycycline (Sigma-Aldrich, D9891) and 1 mM sodium butyrate (NaB, Sigma-Aldrich, B5887) for 60–72 h, as previously described⁸. Supernatants were collected by centrifugation (1,500 × g, 10 min, 4 °C), filtered through a 0.45-μm PES filter and titrated as previously described⁸. Briefly, cell-free virus supernatants were 2-fold serially diluted (100 μl per well) in fresh medium and used to infect naïve SLK cells (10⁴ cells per well in 96-well plates seeded 24 h earlier). Plates were immediately centrifuged (1,500 × g, 1 h, 30 °C) and returned to the incubator. At 1 h post infection, the inoculum was removed and replaced with fresh medium. After 24 h, cells were washed with cold PBS, and the percentage of GFP-positive cells was quantified using a BD LSR II 18 flow cytometer (BD Bioscience). Infectious units (IU) were defined as the number of GFP-positive cells per well, and virus titre (infectious units per ml; IU ml⁻¹) was determined using the last well with detectable GFP-positive cells⁴³.

For de novo infection, vBcl-2 WT virus was generated as described above, and the vBcl-2 KO virus was generated in iSLK-BAC16.vBcl-2 KO cells stably expressing vBcl-2 WT. HCT116 cells (1 × 10⁵) pre-seeded in 12-well plates were infected with vBcl-2 WT or vBcl-2 KO viruses at a multiplicity of infection (MOI) of 10. Following a spin inoculation step (1,500 × g, 1 h, 30 °C) and a 1-h incubation at 37 °C, the medium was replaced. Supernatants were collected at 60–72 h post-lytic infection, and infectious virus yield was determined as described.

Immunoprecipitation and immunoblotting

Immunoprecipitation of endogenous or overexpressed epitope-tagged proteins was performed as previously described⁴⁴. Briefly, transfected or infected cells were lysed in a buffer containing 25 mM Tris pH 7.5, 150 mM NaCl, 1 mM EDTA, and 1–2% IGEPAL CA-630 (NP-40) (Sigma-Aldrich, I8896) or Triton X-100 (Sigma-Aldrich, T9284), supplemented with a complete protease inhibitor cocktail (Thermo Fisher, A32965). Lysates underwent freeze–thaw cycles and sonication (15% amplitude, 10 s process time, 5 s push-on, 1 s push-off). For phosphorylation detection of TBK1 and IRF3, cells were pre-treated with 50 nM calyculin A (Thermo Fisher, PHZ1044) for 45 min and lysed in the presence of both protease and phosphatase inhibitors. Cell debris was removed by centrifugation (14,500 × g, 15 min, 4 °C). An aliquot of the cleared lysate was reserved as whole-cell lysate (WCL) control. The remaining lysate was pre-cleared with Sepharose 4B beads (Sigma-Aldrich, 9012-36-6) for 2 h at 4 °C under gentle rotation, then incubated with primary antibodies (1–2 μg ml⁻¹) overnight (12–16 h) at 4 °C. The antibody–antigen complexes were captured with protein A/G agarose beads (Thermo Fisher, 20422) for 2 h at 4 °C. Beads were washed extensively with lysis buffer, and immunoprecipitates were eluted by heating in Laemmli sample buffer (Sigma-Aldrich) at 95 °C for 5 min. Proteins were resolved by SDS–PAGE, transferred to PVDF membranes (Bio-Rad), blocked with 5% non-fat milk (Lab Scientific, bioKEMIX) or BSA (VWR) in PBST and probed with the indicated antibodies. HRP-conjugated secondary antibodies (1:3,000, Thermo Fisher) were used for detection with ProSignal Pico Chemiluminescent HRP Substrate (Genesee Scientific), and signals were visualized using a ChemiDoc Imaging System (Bio-Rad).

Primary antibodies used include: mouse anti-NM23-H1 (sc-514515; 1:1,000), mouse anti-BAK (sc-518110; 1:1,000), mouse anti-MFF (sc-398617; 1:1,000), mouse anti-MFN1 (sc-166644; 1:1,000), mouse anti-MFN2 (sc-515647; 1:500), mouse anti-TBK1(sc-398366; 1:1,000), mouse anti-MxB (sc-271527; 1:1,000), mouse anti-K-bZIP (sc-69797; 1:1,000), mouse anti-ORF57 (sc-135746; 1:1,000), mouse anti-ORF45 (sc-53883; 1:1,000), mouse anti-β-actin (sc-47778; 1:1,000), mouse anti-GAPDH (sc-47724; 1:1,000) from Santa Cruz; mouse anti-Flag (M2) (F1804; 1:2,000) from Sigma-Aldrich; mouse anti-HA (901503; 1:2,000), mouse anti-HA.11(16B12) (901519; 1:1,000), mouse anti-mCherry (8C5.5) (677702; 1:1,000) from BioLegend; rabbit monoclonal anti-TOM20 (D8T4N) (42406S; 1:2,000), rabbit monoclonal anti-LC3B (D11) (3868S; 1:1,000), rabbit anti-p62 (5114S; 1:1,000), rabbit monoclonal anti-DRP1(D6C7) (8570S; 1:1,000), rabbit monoclonal anti-OPA1 (D6U6N) (80471; 1:1,000), rabbit monoclonal anti-phospho-TBK1(S172) (5483S; 1:500), rabbit polyclonal anti-MAVS (3993; 1:1,000) from Cell Signaling Technology; rabbit anti-Histone H3 (A2348; 1:1,000) and rabbit monoclonal anti-phospho-IRF3 (S396) (AP1412; 1:1,000) from ABclonal; rabbit monoclonal anti-IRF3 (703682; 1:1,000) from Invitrogen; rabbit anti-NM23-H2 (GTX33360; 1:200) from GeneTex; mouse anti-NM23-H2 (1D3) (H00004831-M06; 1:200) from Abnova; rabbit anti-ORF73 (LANA) (NBP-07279; 1:1,000) from Novus Biologicals; sheep anti-NM23-H2 (AF6665; 1:1,000) from R&D Systems; rabbit anti-TRIM22 (13744-1-AP; 1:1,000) from Proteintech; rabbit anti-Beclin 1 (11306-1-AP; 1:1,000) from Proteintech; rabbit anti-N1-phosphohistidine (1-pHis) (MABS1330; 1:1,000) from EMD Millipore; and mouse anti-ORF65 (1:500) provided by P.F. and S.-J.G. Secondary antibodies, including goat anti-mouse horseradish peroxidase (HRP) (62-6520; 1:3,000), goat anti-rabbit HRP (65-6120; 1:3,000) and donkey anti-sleep HRP (A16041; 1:3,000) were from Thermo Fisher.

Protein purification and GST pull-down assay

Recombinant proteins were expressed in BL21(DE3) cells transformed with pGEX-4T-1-vBcl-2 (for GST-vBcl-2) or pRSETB-NM23-H2 (for His-NM23-H2). Protein expressions were induced with 1 mM isopropyl β-d-1-thiogalactopyranoside (IPTG; GoldBio, I2481C) at 20 °C overnight. Cells were collected, subjected to 3 freeze–thaw cycles and resuspended in prechilled lysis buffer (125 mM Tris pH 9.0, 300 mM NaCl, 10% glycerol, 0.05% Triton X-100, 1× protease inhibitors, 1 mg ml⁻¹ lysozyme and 1 μg ml⁻¹ DNase I). Lysates were sonicated (3 ×30 s on ice at 60% amplitude with 15 min intervals) and centrifuged to remove debris. The clarified supernatant was incubated with equilibrated glutathione agarose beads (for GST-vBcl-2) or HisPur Ni-NTA resin (Thermo Fisher, 16101, for His-NM23-H2) for 1 h at 4 °C with rotation. After 3 washes with ice-cold PBST (PBS with 1% Triton X-100), GST-vBcl-2 was eluted using 10 mM reduced l-glutathione in 50 mM Tris-HCl, 150 mM NaCl pH 8.0, and His-NM23-H2 was eluted with 100 mM imidazole (VWR) in 20 mM NaH₂PO₄, 300 mM NaCl pH 8.0, followed by desalting to remove imidazole. Protein concentrations were determined using the Pierce BCA Protein Assay kit (Thermo Fisher, 23209).

For the GST pull-down assay, GST-vBcl-2 bound to glutathione agarose beads was incubated with 5 μg purified His-NM23-H2 in Triton X-100 buffer (150 mM NaCl, 5 mM EDTA, 1% (v/v) Triton X-100, 50 mM HEPES pH 7.4 plus protease inhibitor cocktail) at 4 °C for 4 h. Beads were washed extensively and bound proteins were eluted by heating in Laemmli sample buffer at 95 °C for 5 min, followed by immunoblotting.

Disuccinimidyl suberate crosslinking

NM23-H2 oligomerization was assessed by disuccinimidyl suberate (DSS) crosslinking as described previously⁴⁵. Briefly, at 2 days post transduction with Flag-NM23-H2 and/or HA-vBcl-2, HEK293T cells were washed with PBS and resuspended in 500 μl of 2.5 mM DSS (Thermo Fisher, 21655) in PBS for 30 min at r.t. with constant mixing. The reaction was quenched with 50 mM Tris-HCl pH 7.5 for 10 min. Cells were centrifuged at 300 × g for 5 min and lysed in 50 μl of 2% Triton X-100 lysis buffer (25 mM Tris pH 7.5, 150 mM NaCl, 1 mM EDTA and 2% Triton X-100) supplemented with complete protease inhibitor cocktail (Thermo Scientific). Whole-cell lysates were then subjected to immunoblotting.

Docking

Molecular docking simulations were performed using HADDOCK software (v.2.4)⁴⁶ with the crystal structures of human hexameric NM23-H2 (Protein Data Bank (PDB) 1NSK) and KSHV vBcl-2 (PDB 1K3K). Docked structures with the lowest energy scores in HADDOCK were used as input. The coordinate for the NM23-H2 Q50E mutant was generated using the PyMOL mutagenesis wizard (Schrodinger). Docking of VBNI-1, drawn in ChemDraw, with vBcl-2 (PDB 1K3K) was carried out using AutoDock Vina (v.1.2)^39,47; the displayed structural model is one of the top two lowest-energy solutions. PyMOL (Schrodinger) was used for visualization, and the binding affinity of docked complexes was predicted using the PRODIGY web server^48,49.

Mitochondrial fractionation and SDD–AGE

Intact mitochondria were isolated from cultured cells using the Mitochondria Isolation kit (Thermo Fisher, 89874) according to manufacturer protocol. Isolated mitochondria were lysed in RIPA buffer (50 mM Tris-HCl pH 8.0, 150 mM NaCl, 1 mM EDTA, 2% Triton X-100, 0.5% sodium deoxycholate, 1 mM dithiothreitol, 0.1% SDS with protease inhibitors) on ice for 30 min. Protein concentrations were determined using the Pierce BCA Protein Assay kit, after which samples were mixed with Laemmli sample buffer and processed by SDS–PAGE and immunoblotting.

For semi-denaturing detergent agarose gel electrophoresis (SDD–AGE) to analyse MAVS aggregation³⁰, isolated mitochondrial fractions were resuspended in 1× SDD loading buffer (0.5× TBE, 10% glycerol, 2% SDS and 0.0025% bromophenol blue) and loaded onto a vertical 1.5% agarose gel. Electrophoresis was performed at 100 V for 35 min in 0.5× TBE with 0.1% SDS at 4 °C, followed by immunoblotting.

Immunofluorescence staining of cells

Cells seeded on coverslips were washed with PBS and fixed with 4% (v/v) paraformaldehyde (Thermo Fisher, J19943.K2) for 20 min at r.t. Cells were then permeabilized in 0.2% Triton X-100 in PBS for 10 min and blocked with PBS containing 10% goat serum for 1 h at r.t. Primary antibodies diluted in 1% goat serum were applied to the cells for 1.5 h at r.t. or overnight at 4 °C. After 3 washes with PBS containing 0.25% Tween 20, cells were incubated with Alexa 488-, Alexa 568-, Alexa 405- and/or Alexa 647-conjugated secondary antibodies (Invitrogen; 1:800) in 1% goat serum for 1 h at r.t. for confocal microscopy analysis, or incubated with secondary antibodies conjugated with CF680R– (Biotium 20196 or 20192; 1:250) and/or ATTO 647N (Rockland, 611-156-122; 1:250) for 1 h at r.t. for STED microscopy. Following another round of PBS washes, nuclei were counterstained with DAPI (Thermo Scientific, D1306) and coverslips were mounted using VECTASHIELD PLUS Antifade Mounting Medium (Vector Laboratories, H-1900) or ProLong Diamond Antifade mountant (Thermo Scientific, P36965).

For plasma membrane staining, HeLa cells were incubated with 5 μg ml⁻¹ Wheat Germ Agglutinin (WGA)-CF 640R Conjugate (Cayman Chemical, 29026) for 30 min at r.t. before fixation. Cells were then processed as described above. For mitochondria staining, cells were prestained with 200 nM MitoTracker Red CMXRos (Thermo Scientific, M7512) at 37 °C for 20 min before fixation. Cells were subsequently processed as described above.

Primary antibodies used include: mouse anti-TOM20 (F-10) (sc-17764; 1:50), mouse anti-V5 (sc-271944; 1:100), mouse anti-ISG15 (sc 166755; 1:50), mouse monoclonal anti-MxB (H-7) (sc-271527; 1:50), mouse anti-ISG15(F-9) (sc-66755; 1:50) from Santa Cruz; mouse anti-HA (901503; 1:100) from Biolegend; mouse anti-Flag (M2) (F1804; 1:100), rabbit anti-TGN46 (T7576; 1:50) and rabbit anti-TGN46 (T7576; 1:50) from Sigma-Aldrich; sheep anti-NM23-H2 (AF6665; 1:25) from R&D Systems; rabbit monoclonal anti-TOM20 (D8T4N) (42406S; 1:75) from Cell Signaling Technology; rabbit anti-PINK1 (ab23707; 1:50) from Abcam; mouse anti-PDI (RL90) (NB300-517; 1:50) from Novus Biologicals; rabbit polyclonal anti-MAVS (3993; 1:100) from Cell Signaling Technology; rabbit anti-TRIM22 (13744-1-AP; 1:50) from Proteintech; and mouse anti-ORF65 (1:50) provided by P.F. and S.-J.G.

Image acquisition using confocal microscopy and super-resolution microscopy

Confocal microscopy was performed on a Leica TCS SP5 II laser-scanning confocal microscope (Leica Microsystems) equipped with HyD detectors and Leica PMT detectors; four laser lines (405 nm, 488 nm, 561 nm, 633 nm) were used for fluorescence imaging in the blue, green, red and far-red channels, combined with their respective standard emission filter sets. Images were acquired using the HCX PL APO CS (×63, 1.4 NA, oil immersion) objective and the Leica LAS-X software (v.4.8.0.28989). Confocal z-stacks were acquired with optimal z-intervals according to the Nyquist criterion, covering the whole volume of cells from the basal to the apical region. Images were then compiled by ‘max projection’ before analysis in ImageJ.

Super-resolution images were captured on a Leica Stellaris 8 3×TauSTED microscope (×100, 1.4 NA, oil immersion objective, HyD detectors in photon counting mode) using 646 nm and 679 nm excitations with a 775 nm depletion laser. Acquisition parameters were set on the Leica LAS-X software (v.4.8.0.28989) to 1,024 × 1,024 pixels with 5× zoom, yielding a resolution of 22 nm pixel⁻¹. Two-dimensional (2D) deconvolution was performed with Huygens Professional software (v.24.10) (Scientific Volume Imaging) before analysis in ImageJ.

Image processing and analysis

All image analyses were conducted using ImageJ Fiji (v.2.9.0, National Institutes of Health). Maximum intensity projections were generated from deconvoluted z-stack confocal or STED images. Fluorescence intensity was quantified by drawing line scans across regions of interest (ROIs) and applying the ‘Plot Profile’ function. Intensity values for each channel were normalized to their respective maximum value to enable cross-sample comparisons. For individual cell measurements, cells were manually cropped, and background signals were set to zero. The cropped images were converted into binary format, and nuclei were segmented using the DAPI signal via the intensity threshold method. Settings for segmentation and filtering were applied consistently across all datasets. To ensure unbiased quantification, investigators conducted blind counting for all analyses.

For the measurement of NM23-H2 co-localization with vBcl-2 and TOM20 (Fig. 2d-f), mitochondrial areas were defined using TOM20. Intensity threshold segmentation was applied to identify areas corresponding to mitochondria, NM23-H2 and vBcl-2 within each cell. Mitochondria containing NM23-H2 and/or vBcl-2 were determined by overlapping NM23-H2-positive and/or vBcl-2-positive areas with TOM20-positive areas using the ROI manager in ImageJ⁵⁰. The percentage of the NM23-H2 area co-localizing with TOM20 (Fig. 2d) was calculated by dividing the NM23-H2- and TOM20-positive areas by the total NM23-H2 area. Similarly, the percentage of NM23-H2 area co-localizing with vBcl-2 (Fig. 2e) was calculated by dividing the NM23-H2- and vBcl-2-positive areas by the total NM23-H2 area. Co-localization between vBcl-2 and mitochondria (Fig. 2f) was assessed using the ‘JACoP’ plugin in ImageJ, and the Manders’ co-localization coefficient was calculated to quantify the degree of overlap. At least 30 cells randomly selected and pooled from 3 independent experiments were analysed.

For the measurement of PINK1 co-localization with TOM20 (Extended Data Fig. 5a), intensity threshold segmentation was applied to individual cells to identify areas corresponding to TOM20 and PINK1. Co-localization was assessed using the ‘JACoP’ plugin in ImageJ to calculate Manders’ coefficient to quantify the degree of overlap. At least 100 cells randomly selected from 10–15 HPFs across 3 independent experiments were analysed.

For mitochondrial morphology analysis (for example, Extended Data Fig. 4a), images were acquired as tile-scan z-stacks with a 0.2-μm z-step and compiled into maximum intensity projections. Mitochondrial morphology was classified as fragmented (predominantly short and spherical), tubulated (predominantly highly elongated with interconnectivity) or intermediate (predominantly short tubular, neither connected nor spherical), as described previously^51,52. At least 100 cells per condition from 3 independent experiments were analysed.

For the analysis of mitochondrial number, area and perimeter, high-quality single-cell images were randomly selected for pre-processing (background subtraction, Sigma Filter Plus; Enhance Local Contrast, and Adjust Gamma). Thresholding was performed using the ‘2D threshold’ function of the Mitochondria Analyzer 2.1.0 plugin in ImageJ⁵³. Mitochondrial number, mean area and mean perimeter in a whole cell were calculated with a minimum area of 0.1 μm² on the basis of 50–100 cells per condition using the ‘2D Analysis’ function of the Mitochondria Analyzer (ImageJ).

For the measurement of MAVS aggregation (for example, Fig. 4d), from deconvoluted super-resolution images, 225 μm² ROIs of MAVS were randomly selected. MAVS clusters were segmented using a threshold of 2% of the maximum signal intensity, and their sizes were measured using the maximum feret diameter, as described previously⁵⁴. Clusters smaller than 200 nm were excluded as non-oligomerized. Approximately 100 clusters in ROIs pooled from at least 15 cells from 3 independent experiments were analysed. Boxplots with individual points were generated using GraphPad Prism v.10.

For the analysis of virus egress (for example, Fig. 5), lytically induced cells were immunostained for KSHV capsid protein ORF65 and counterstained with DAPI, and images were acquired as tile-scan z-stacks with a 0.2-μm step size using a confocal microscope and compiled into maximum intensity projections. The nucleus was delineated using the DAPI signal, and intensity threshold segmentation was applied to both the ORF65 and nuclear channels to define ROIs corresponding to viral capsids and nuclei, respectively. To refine the nuclear segmentation and improve the discrimination between perinuclear and intranuclear ORF65 signals, the DAPI images were denoised using the ‘Median’ filter with a 0.389-μm diameter, as previously described⁵⁵. For quantification of the percentage of cytoplasmic ORF65 per cell, intranuclear ORF65 regions were first identified by overlapping the ORF65-positive ROIs with the nuclei ROIs using the ROI Manager in ImageJ⁵⁰. Cytoplasmic ORF65 signals were then determined by subtracting the intranuclear signal from the total ORF65 signal. The percentage of cytoplasmic ORF65 was calculated by dividing the cytoplasmic ORF65 area by the total ORF65 area for each cell. Approximately 50 cells per condition were analysed. For the measurement of the percentage of cells exhibiting cytoplasmic ORF65, ~300 cells were randomly selected from 15–20 HPFs across 3 independent experiments. Cells were scored as positive for cytoplasmic ORF65 if a distinct signal was observed in the cytoplasmic compartment.

Proximity ligation assay

For NM23-H2-TOM20 proximity ligation assays (PLAs) (Fig. 2k), iSLK-BAC16 cells were seeded on 18 mm coverslips and induced for lytic reactivation. At 60 h post reactivation, cells were washed once with PBS, fixed with 4% paraformaldehyde in PBS for 15 min at r.t. and permeabilized with 0.2% Triton X-100 in PBS for 5 min. PLA was performed using the Duolink In Situ Red Starter Kit Mouse/Rabbit (Sigma-Aldrich DUO92101) according to manufacturer instructions. Briefly, after permeabilization, samples were blocked and then incubated with primary antibodies against NM23-H2 and TOM20 (Cell Signaling Technology, 42406S) for 1.5 h, followed by incubation with anti-rabbit MINUS and anti-mouse PLUS PLA probes. Ligation, rolling-circle amplification and subsequent washes were performed according to manufacturer instruction. Finally, coverslips were mounted using a DAPI-containing mounting medium. All incubations were performed in a humidity chamber using 500 μl per well. Images were acquired on a Leica TCS SP5 II confocal microscope using ×63 1.4 NA oil immersion objective. PLA puncta were quantified using the ‘analyze particles’ algorithm in ImageJ as previously described⁵⁶. The number of PLA puncta per cell was normalized to the average NM23-H2 PLA signal per cell in each condition, as determined using NM23-H2 primary antibodies from two different species⁵⁷.

Immunogold labelling and EM

iSLK-BAC16.vBcl-2 KO cells with complemented expression of vBcl-2 (WT, AAA or E14A) were induced for lytic reactivation for 60–72 h and fixed in 2% formaldehyde/2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.4; Electron Microscopy Sciences) for 30 min at r.t. Cells were pelleted, post fixed in 2% osmium tetroxide (Electron Microscopy Sciences, 19152) in 0.2 M cacodylate buffer for 1 h at r.t., rinsed in distilled H₂O, en bloc stained with 2% uranyl acetate (Electron Microscopy Sciences, 541-09-3), dehydrated through a graded ethanol series and embedded in EMbed-812 (Electron Microscopy Sciences). Thin sections were collected, contrast stained with 2% uranyl acetate and lead citrate, and imaged using a JEOL 1010 transmission electron microscope equipped with a Hamamatsu digital camera and AMT Advantage NanoSprint500 software (v.7.00).

Mitochondrial lengths were measured by an observer blinded to experimental condition. Six to eight cell profiles (with nuclei) per condition were randomly selected, and mitochondrial length was measured using the freehand line tool in ImageJ after scale calibration⁵⁸. Cristae density was determined by manually counting the number of cristae per mitochondrion and normalizing to mitochondrial length, as described previously⁵⁸.

For the measurement of KSHV capsid morphogenesis and nuclear egress, 15–20 randomly selected EM images per condition were analysed by blinded investigators; viral capsids undergoing maturation within the nucleus, or dissipated into the cytoplasm and/or extracellular space were individually counted.

For immunegold labelling, ultrathin sections were incubated with primary antibodies (control IgG, anti-NM23-H2 or anti-DRP1; 1:10 dilution), followed by secondary antibodies (goat anti-rabbit IgG conjugated to 10 nm gold particles (Electron Microscopy Science, 25109) and goat anti-mouse IgG conjugated to 6 nm gold particles (Electron Microscopy Science, 25124); 1:50 dilution). Images were acquired as above. DRP1/NM23-H2 clusters were defined as regions containing 3 or more gold particles (6 nm and 10 nm), with each particle located within 20 nm of its nearest neighbour (measured in ImageJ). Approximately 30 mitochondria randomly selected from 10–15 images across 3 independent experiments were manually counted and recorded for cluster analysis and DRP1 quantification.

RT–qPCR analysis

Total RNA was extracted using the Direct-zol RNA Miniprep kit (Zymo Research, R2050) following manufacturer protocol. RNA quality and quantity were examined using a NanoDrop Lite spectrophotometer. cDNA synthesis was performed using the iScript cDNA Synthesis kit (Bio-Rad 1708841), and 50 ng cDNA was used for qPCRs with PerfeCTa SYBR Green Master SuperMix (Quantabio, 95054-02K) on a CFX Opus 96 Real-time PCR System (Bio-Rad). Gene-specific primers (Supplementary Table 3) were used, and relative mRNA expression was calculated using the comparative threshold (ΔΔC_T) method with 18S rRNA as the internal control. All reactions were performed in triplicates.

To assess KSHV DNA replication in cells, total DNA was extracted using the DNeasy Blood and Tissue kit (QIAGEN, 69504) according to manufacturer instruction, followed by qPCR with ORF73-specific primers. Viral DNA copy numbers were normalized to host cellular GAPDH.

To quantify progeny virus genomes in the medium, culture supernatant containing virus was treated with 100 U ml⁻¹ DNase I (New England Biolabs, M0303S) for 15 min at 37 °C to remove non-encapsidated viral DNAs. The reaction was stopped with 50 mM EDTA and heated at 75 °C for 15 min. Viral DNA was then extracted using the QIAamp DNA mini kit (QIAGEN, 56304) and 5 μl of the eluted DNA was used for qPCR with ORF73-specific primers. A standard curve generated from serial dilutions of BAC16 DNA (1–10⁹ genome copies per reaction) was used to calculate viral genome copy numbers, which were expressed as copies per ml.

All primer sequences are listed in Supplementary Table 3.

Mitochondrial membrane potential, ROS and cytosolic mtDNA

Analysis of mitochondrial membrane potential was performed by flow cytometry in cells incubated for 30 min at 37 °C with 100 nM tetramethylrhodamine methyl ester (TMRM; AnaSpec, 88065) in complete medium, along with LIVE/DEAD fixable dye (Thermo Fisher). Live cells were gated and analysed on a BD LSR II flow cytometer, and the geometric mean fluorescence intensity was determined using FlowJo v.10 (BD Bioscience). The gating strategy for flow cytometry is in Supplementary Fig. 15.

Mitochondrial ROS were measured using MitoSOX (Thermo Scientific, M36008). Briefly, cells induced for 60 h lytic reactivation were loaded with 5 μM mitoSOX in complete medium at 37 °C for 15 min, then washed 3 times with PBS. Fluorescence intensity was measured with a FLUOStar Optima microplate reader (BMG LabTech) and normalized to control values.

Measurement of cytosolic mtDNA was performed as previously described⁵⁹. Briefly, cells induced for 60 h lytic reactivation were split into 2 aliquots. One aliquot underwent total DNA extraction using the DNeasy Blood and Tissue kits (QIAGEN) to quantify total mtDNA. The other aliquot was homogenized in 200 μl of mild digitonin (ACROS Organics, 11024-24-1) lysis buffer (150 mM NaCl, 50 mM HEPES pH 7.4, 20 μg ml⁻¹ digitonin), followed by 3 rounds of centrifugation at 980 × g for 3 min to pellet intact cells. The resulting supernatant was centrifuged at 17,000 × g for 10 min to obtain a cytosolic fraction free of nuclear, mitochondrial and endoplasmic reticulum contamination. DNA was extracted from this fraction using the DNeasy Blood and Tissue kits (QIAGEN). qPCR was performed on both cytosolic DNA and whole-cell DNA extracts using primers for human mtDNA (Cytb and Nd4), with whole-cell C_T values serving as normalization controls. The purity of the cytosolic fraction was verified by immunoblotting for specific organelle markers.

Cell viability analysis

Adherent cells were plated at 1 × 10⁴ cells per well and suspension cells at 5 × 10⁴ cells per well in 96-well plates. After overnight incubation, cells were treated with compounds at the indicated concentrations and durations. Viability was assessed using the WST-1 reagent (Roche, 50-100-3319) according to manufacturer instructions. Briefly, the medium was aspirated, and 10 μl of WST-1 in complete medium was added to each well, followed by incubation for 0.5–4 h at 37 °C with a brief 1-min agitation to ensure proper mixing. Absorbance was then measured using a FLUOStar Optima microplate reader.

Apoptosis and autophagy analyses

Apoptosis analysis was performed as previously described⁶⁰. Briefly, HEK293T cells stably expressing vBcl-2 (WT or mutant) were seeded at 1 × 10⁶ cells per well in 6-well plates and incubated for 24 h. Cells were then treated with medium containing 10 ng ml⁻¹ human TNF (PEPROTECH, 300-01A) and 5 μg ml⁻¹ cycloheximide (Sigma-Aldrich, C4859) for 4 h. Apoptotic cells were detected using the FITC Annexin V Apoptosis Detection kit with propidium iodide (PI; BioLegend, 640914) according to manufacturer instructions and analysed on a BD LSR II flow cytometer. Cells were gated as live (Annexin V/PI-double negative), early apoptotic (Annexin V-positive, PI-negative) or late apoptotic/necrotic (Annexin V/PI-double positive).

For assessment of autophagy, HEK293T cells stably expressing vBcl-2 (WT or mutant) were treated with 50 nM Torin 1 (Selleck Chemicals, S2827) for 3 h. Cells were lysed in ice-cold RIPA lysis buffer (50 mM Tris-HCl pH 8.0, 150 mM NaCl, 1 mM EDTA, 2% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS) supplemented with protease inhibitor cocktail for 30 min, followed by centrifugation at 16,000 × g for 10 min at 4 °C. Cleared lysates were diluted in 2× SDS–PAGE loading buffer and analysed by immunoblotting for LC3 and p62.

Luciferase reporter assay

HEK293T cells were seeded in 12-well plates (~10⁵ cells per well) and transfected with 400 ng IFN-β luciferase reporter plasmid (provided by P.F.), 40 ng TK-RLuc plasmid and 2 μg ml⁻¹ poly(I-C) using FuGene transfection reagent. At the indicated times, luciferase activity was measured with the Dual-Luciferase Reporter Assay System (Promega E1980) on a FLUOStar Optima microplate reader. Firefly luciferase activity was normalized to Renilla luciferase activity, and fold induction was calculated relative to mock-transfected controls.

Enzyme-linked immunosorbent assay

Culture media from transfected or reactivated cells were centrifuged to remove cell debris, and IFN-β levels were quantified using the LumiKine Xpress hIFN-β 2.0 kit (Invivogen, luex-hifnbv2), following manufacturer instructions. IFN-β concentrations (pg ml⁻¹) were calculated using a standard curve.

NDP kinase activity

A standard pyruvate kinase–lactate dehydrogenase coupled assay was used to examine the NDP kinase activity of recombinant human NM23-H2 as previously described⁶¹ with minor modification. Briefly, recombinant NM23-H2 (GeneTex, 200 ng, GTX67595-pro) was pre-incubated with 250 ng purified vBcl-2 (WT and E14A) and added to a reaction mixture containing 50 mM Tris-HCI pH 7.4, 50 mM KCI, 5 mM MgCl₂, 1 mg ml⁻¹ BSA, 0.1 mM phosphoenolpyruvate, 0.1 mg ml⁻¹ NADH (Cayman Chemical), 0.2 mM GDP, 2 U ml⁻¹ pyruvate kinase (Sigma-Aldrich) and 2.5 U ml⁻¹ lactate dehydrogenase (Sigma-Aldrich). The reaction was initiated by adding 0.2 mM ATP (Thermo Fisher, R0441), and absorbance at 340 nm was recorded every 10 s for 14 min using a FLUOStar Optima microplate reader. NDP kinase activity is presented as the percentage decrease in NADH absorbance.

GTPase assay

DRP1 GTPase activity was measured as previously described²⁸ with some modifications. Briefly, HEK293T cells transfected with FLAG-DRP1 for 48 h were lysed in Triton X-100 lysis buffer (150 mM NaCl, 5 mM EDTA, 1% (v/v) Triton X-100, 50 mM HEPES pH 7.4) supplemented with protease inhibitor cocktail, and the lysate was clarified by centrifugation at 20,000 × g for 15 min at 4 °C. Flag-DRP1 was immunoprecipitated using anti-Flag antibody (12 h at 4 °C) and protein A/G beads (2 h at 4 °C). After washing with Triton X-100 buffer containing 500 mM NaCl, the beads were equilibrated in GTPase assay buffer (20 mM Tris pH 8.0, 10 mM dithiothreitol, 2.5 mM EDTA, 0.3 mg ml⁻¹ BSA, 0.1 mg ml⁻¹ cardiolipin (Sigma-Aldrich, C0563)) supplemented with protease inhibitor cocktail. Bead-bound DRP1 were incubated at 37 °C for 1 h with 100 ng recombinant NM23-H2, 125 ng purified vBcl-2 (WT and E14A) and 1 mM nucleotides as indicated. GTPase activity was measured using the QuantiChrom ATPase/GTPase Assay kit (BioAssay Systems, DATG-200) following manufacturer instructions. After adding the malachite green detection reagent, absorbance at 620 nm was measured with a FLUOStar Optima microplate reader, and activity was expressed as a percentage relative to the control (1 mM GTP).

ATPase activity assay

ATP hydrolysis by recombinant NM23-H2 and NM23-H1 was measured using the Transcreener ADP² FI Assay kit (BellBrook, 3013-1K) according to manufacturer protocol. Assays were conducted in black opaque 384-well plates at r.t. Briefly, 5 μl of 0.2 nM NM23-H2/H1 in assay buffer (10 mM HEPES pH 7.4, 100 mM NaCl, 5 mM MgCl₂, 0.01% Triton X-100) at 2× final concentration was dispensed into each well. Test compounds were added in 100 nl of 100% dimethylsulfoxide (DMSO) using an Echo 650 acoustic liquid handler. The reaction was initiated by adding 5 μl of 1 μM ATP (final concentration) containing 2 μg ml⁻¹ ADP-IR antibody and 4 nM ADP-Alexa594 tracer. Fluorescence (excitation/emission: 577/641 nm) was recorded every 2 min for 30 min using a ClarioStar plate reader (BMG LabTech). Progress curves were fitted by linear regression, and the slopes were normalized to calculate percent inhibition, with 100% inhibition defined as the slope in the absence of NM23-H2/H1 and 0% inhibition as the slope in their presence.

Yeast two-hybrid assay

To identify vBcl-2-interacting proteins, yeast two-hybrid assay was performed using Matchmaker Gold Yeast Two-Hybrid System (Takara, 630489) according to a previously published method⁶². Yeast strain AH109 harbouring a Gal4-vBcl-2 full-length fusion plasmid was grown overnight in synthetic dropout (SD)/-Trp medium to ~10⁷ cells per ml. The culture was diluted into 1 l of warmed Trp medium to an optical density (OD₆₀₀) of 0.2~0.3 and grown to exponential phase. Cells were collected and washed twice with water and once with TE buffer (Clontech). The cell pellet was resuspended in 8 ml of 10 mM Tris-HCl pH 7.5, 1 mM EDTA and 0.1 M Li-acetate (LiOAc). Transformations were performed by mixing the cell suspension with 1 mg of transforming DNA and 20 mg of single-stranded salmon sperm DNA, followed by the addition of 60 ml of 40% polyethyleneglycol 3350 in Tris-EDTA-LiOAc. After thorough mixing, the cells were incubated at 30 °C with agitation for 30 min, heat shocked at 42 °C for 15 min, pelleted, resuspended in YPD medium and plated on selective medium. Library screening and plasmid recovery were performed according to manufacturer instructions.

RNA-seq

Total RNA was extracted from 3 independent samples of each BJAB cell line using the Direct-zol RNA MiniPrep kit following manufacturer instruction. RNA was treated with DNase I, and 500 ng of total RNA was used to prepare 3’ mRNA-seq libraries with the QuantSeq 3’ mRNA-Seq FWD Library Prep kit (LEXOGEN, 015.24) for IIIumina according to manufacturer protocol. Library quality and size were assessed using Agilent Tapestation with DNA 5000 Screentape (Agilent), and library quantification was performed using real-time PCR (Kapa Biosystems). Pooled libraries were sequenced on an Illumina NextSeq 2000 system using high-output, single-end 75-bp reads. Reads were aligned to the human genome (UCSC, hg38) using STAR aligner⁶³, and differential gene expression was determined using DESeq2 (ref. 64). Genes with an average read count <10 or with an outlier value were excluded from the analysis. Differential expression plots for significant genes (P < 0.05, [log₂FC] > 1) were generated using the limma and ggplot2 packages in R.

HTRF-based high-throughput screening assay

An HTRF assay was developed to evaluate the in vitro interaction between vBcl-2 and NM23-H2 in white opaque 384-well low-volume microplates (Greiner, 784075). To enhance solubility and yield, a mutant vBcl-2 protein [vBcl-2 (1–146) with N67D/V117A] was used, which has been demonstrated to retain function in viral lytic replication^9,65. For assay optimization, 5 μl of various concentrations of purified GST-vBcl-2 at 2× final concentration was pre-incubated with 1 nM anti-GST-d2 HTRF acceptor (PerkinElmer) in assay buffer (50 mM HEPES pH 7.2, 5 mM MgCl₂, 0.02% Triton X-100) for 30 min at r.t. This mixture was dispensed into the wells, followed by the addition of 5 μl of various concentrations of purified His-NM23-H2 (2× final concentration) pre-incubated with 0.5 nM anti-His-Tb HTRF donor (PerKinElmer). The reaction was incubated for 2 h at r.t. HTRF signals were measured on a ClarioStar Plus plate reader using a 337 nm excitation filter with 665 nm (acceptor) and 620 nm (donor) emission filters, and the HTRF ratio was calculated as (intensity_665nm/intensity_620nm) × 10,000.

For high-throughput compounds screening, 5 μl of assay buffer containing 20 nM GST-vBcl-2 and 1 nM anti-GST-d2 HTRF acceptor was dispensed into each well using a Biotek MicroFlo dispenser (1 μl cassette). Next, 100 nl of 1 mM test compounds were added using the Janus MDT Nanohead. After a 15-min pre-incubation, 5 μl of 4 nM His-NM23-H2 and 0.5 nM anti-His-Tb in assay buffer was added to each well (final DMSO concentration, 1%). Following a 2 h incubation at r.t., HTRF signals were measured as described above. Inhibition was normalized such that 100% inhibition corresponded to the HTRF signal in the absence of vBcl-2, and 0% inhibition to the signal with both vBcl-2 and NM23-H2 in the absence of test compounds.

For dose-response experiments, test compounds were serially diluted (1:3.16, semi-log) in 100% DMSO at 100× final concentration in a source plate. Then, 100 nl was transferred to the assay plate using the Janus NanoHead. IC₅₀ values were determined by nonlinear regression fitting of the normalized data to a 4-parameter dose-response model using XLFit5 (IDBS).

Surface plasmon resonance assay

SPR experiments for VBNI-1 binding to vBcl-2 were performed on a Biacore T200 system (Cytiva) using standard amine coupling. Briefly, a CMD700M carboxymethyldextran sensor chip (Xantec Bioanalytics) was conditioned with 0.1 M sodium borate pH 9.0 containing 1 M NaCl for 180 s at 10 μl min⁻¹, then activated with 100 mM EDC (1-ethyl-3-(3-dimethylaminopropyl) carbodiimide) in water and 100 mM NHS (N-hydroxysuccinimide) in 50 mM MES pH 5.0 for 600 s at 10 μl min⁻¹. vBcl-2 (100 nM in 10 mM sodium acetate pH 5.0) was injected until 20,000 RU of proteins were immobilized, and the remaining activated sites were blocked with 1 M ethanolamine for 180 s at 10 μl min⁻¹. The running buffer was then switched to 20 mM HEPES pH 7.4, 100 mM NaCl, 5 mM MgCl₂, 0.005% Tween 20 and 2% DMSO before compounds injection. Test compounds, serially diluted 1:3, were injected at 30 μl min⁻¹ for 60 s, followed by a 180-s dissociation phase. Background-subtracted sensograms were fitted to one-site kinetic model using Biacore T200 Evaluation Software (v.3.2.2).

Procedures for chemical synthesis of Z56833834 (VBNI-1)

2-(Benzo[d]oxazol-2-ylamino)-5,6-dimethylpyrimidin-4(3H)-one (VBNI-1) was synthesized following the previously reported procedures^66,67, as shown below. The detailed synthetic procedure is available in Supplementary Note.

Statistical analysis

Statistical analysis was performed using GraphPad Prism 10. Data are expressed as mean ± s.d. unless otherwise noted. All experiments were independently repeated at least three times. The Student’s t-test was used for comparisons between two groups. For multiple groups, analysis of variance (ANOVA) with Tukey’s post hoc test or Kruskal–Wallis test with post hoc Dunn’s test was applied, as specified in the respective figure legends. A two-sided P ≤ 0.05 was considered statistically significant (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; NS, not significant).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Extended Data

Extended Data Fig. 1 |. KSHV vBcl-2 targets NM23-H2 independently of inhibition of autophagy and apoptosis.

a, Co-immunoprecipitation (co-IP) of vBcl-2 with NM23-H2 in HEK293T cells co-transfected with HA-vBcl-2 and Flag-NM23-H2. Actin serves as a loading control. b, Co-IP of endogenous NM23-H2 with vBcl-2 in HEK293T cells expressing empty vector (Vec) or HA-vBcl-2. c, Schematic of WT vBcl-2 and deletion mutants, with binding to NM23-H2 determined by co-IP of Flag-NM23-H2 with HA-vBcl-2 in HEK293T cell lysates. BH (Bcl-2 homology) domains and the conserved ₈₄W₈₅G₈₆R residues in BH1 domain are highlighted. TM, transmembrane domain. +, strong binding; −, no binding. d, Co-IP of Flag-tagged NM23-H2 with HA-tagged wild-type (WT) or mutant vBcl-2 in HEK293T cells. The binding results are summarized in c. e, IB analysis of LC3-I/LC3-II and p62 in HEK293T cells stably expressing Vec or HA-vBcl-2 (WT or mutants) with or without Torin 1 (50 nM, 3 h). f, Densitometric quantification of LC3-II/LC3-I (red bars) and p62/actin (blue) from e. n = 3. g, Flow cytometry analysis of apoptosis in HEK293T cells stably expressing Vec or HA-vBcl-2 (WT or mutants) treated with TNF-α (10 ng/ml) and CHX (5 μg/ml) for 4 h. Early apoptotic cells (Annexin V-positive and PI-negative) were quantified (n = 3). Data in a,b,d,e are from one experiment that is representative of three independent experiments. See Source Data for uncropped data of a,b,d,e. Data shown in f,g are mean ± s.d. analyzed by one-way ANOVA followed by Tukey’s post hoc test for comparisons among multiple groups. *, p < 0.05; **, p < 0.01; ****, p < 0.0001; ns, not significant. Exact P values are provided in Source Data Extended Data Fig. 1.

Extended Data Fig. 2 |. vBcl-2 interaction with NM23-H2 is required for infectious virion production.

a, Experimental schematic to assess vBcl-2 knockout (KO) effects on KSHV reactivation. iSLK-BAC16.vBcl-2 WT and iSLK-BAC16-vBcl-2 KO cells were induced (Dox/NaB, 60-72 h), and supernatants infected naïve SLK cells for 24 h (results in b-d). b, Bright field (BF) and GFP images of SLK cells infected with virus from induced iSLK-BAC16 cells. Scale bars, 50 μm. c,d, Virus titres determined by two-fold serial dilution of supernatants collected 72 h post-reactivation (a) and quantification of GFP-positive SLK cells per well by flow cytometry at 24 h post-infection (c), expressed as infectious units/ml (d). e, Schematic for assessing vBcl-2 KO impact on KSHV de novo infection in HCT116 cells infected with vBcl-2 WT or vBcl-2 KO KSHV at MOI = 10, and supernatant collected were used to infect SLK cells for 24 h (results in f-h). f, GFP images of SLK cells infected in e. Scale bars, 430 μm. g, Flow cytometry analysis of GFP-positive SLK cell percentages and mean fluorescence intensity from e. h, qPCR analysis of viral genome copies in supernatants from e. (n = 3 independent experiments). i, Close-up view of site variations between NM23-H2 (yellow) and -H1 (cyan) in the structural model of NM23-H2 and vBcl-2. The corresponding residues of NM23-H2 and NM23-H1 are shown in stick. The ΔG for the hydrogen (H)-bonding interactions of NM23-H2 Q50 was noted⁶⁵, which was lost when Q50 was replaced by E50. j, Co-IP of Flag-NM23-H2 (WT/mutant) with HA-vBcl-2 in HEK293T cells. k-m, iSLK-BAC16.vBcl-2 WT cells transduced with ctrl or NM23-H1 shRNA induced as above and supernatants infected SLK cells. GFP/BF images are shown in k, virus titer determined in l (n = 3), and protein expression shown in m. Data in j,m are representative of three independent experiments (see Source Data for uncropped data). Data in d,h,l are mean ± s.d. analyzed by two-tailed Student’s t test and one-way ANOVA with Tukey’s post hoc test. *, p < 0.05; **, p < 0.01; ****, p < 0.0001. ns, not significant. Exact P values are provided in Source Data Extended Data Fig. 2.

Extended Data Fig. 3 |. KSHV vBcl-2 regulates NM23-H2 activities.

a, IB (top) and densitometric quantification (bottom) of NM23-H2 stability in cells transduced with Vec or HA-vBcl-2 (WT/E14A) and treated with CHX (n = 3). b, Coupled pyruvate kinase-lactate dehydrogenase assay (Methods) showing vBcl-2-driven NDPK activity of recombinant NM23-H2 preincubated with GST-vBcl-2 (WT/E14A) or Vec (n = 3 independent experiments). c, in vitro assay showing that increasing amount of recombinant GST-vBcl-2 (WT or E14A) modulate histidine phosphorylation of purified His-NM23-H2 in the presence of ATP, detected by IB with anti-1-pHis antibody. Coomassie blue staining shows input of each purified proteins as indicated. d, Densitometric quantification of 1-pHis levels in (c) normalized to NM23-H2 input (n = 3 independent experiments). e-g, Representative confocal images of in HeLa cells (e) co-transfected with HA-vBcl-2 and Flag-NM23-H2 showing the distributions of vBcl-2 (green) and NM23-H2 (red) relative to organelle markers (magenta), including PDI (ER), TGN46 (trans-Golgi), TOM20 (mitochondria), and WGA (wheat germ agglutinin; plasma membrane). Scale bars, 10 μm. The percentage of vBcl-2 overlapping with each marker (f) and vBcl-2 colocalization with NM23-H2 (g) were quantified (n = 50 cells pooled from three independent experiments). Data in a,c,e are from one experiment that is representative of three independent experiments. See Source Data for uncropped data of a,c. Data in a,b,d represent mean ± s.d. analyzed by one-way ANOVA followed by Tukey’s post hoc test. Box-plot data in f,g are presented as median (center line), interquartile range (IQR; box, 25th to 75th percentiles), and whiskers extending from minimum to maximum values; and analyzed using Kruskal-Wallis with post hoc Dunn’s test. *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001; ns, not significant. Exact P values are provided in Source Data Extended Fig. 3.

Extended Data Fig. 4 |. Characterization of the impacts of vBcl-2 on mitochondrial morphology.

a, Confocal micrographs of mitochondria in iSLK-BAC16.vBcl-2 WT/KO cells induced by Dox/NaB for indicated time and stained with MitoTracker (red) and DAPI (blue). b, Quantification of mitochondrial morphology (tubulated, intermediate, and fragmented) in a (n = 3, with 100 cells/experiment). c-e, Analysis of mitochondrial numbers per cell (c), mean area (d) and perimeter (e) in cells in a (n = 94 cells/group, 3 independent experiments). f,g, Confocal micrographs of mitochondria (red) in cells transduced with indicated shRNAs plus NM23-H2 rescue plasmids post-reactivation (f) with corresponding quantification of fragmentation (g; n = 3, 300 cells/sample). h, IB of the indicated mitochondrial fission and fusion factors in mitochondrial fractions and WCL from iSLK-BAC16.vBcl-2 KO cells stably expressing HA-vBcl-2 (WT/mutants) at 60 h post-reactivation. TOM20, control for mitochondrial fraction. GAPDH, a loading control for WCLs (see Source Data for uncropped WB). i, Confocal micrographs of mitochondria (TOM20, green) in iSLK-BAC16. vBcl-2 KO cells expressing mCherry-vBcl-2 (red) and transduced with Vec, an MFN2-encoding plasmid, or with Ctrl shRNA, DRP1 shRNA, MFF shRNAs at 60 h post- reactivation. j, Percentage of cells in (i) with mitochondrial fission was quantified (n = 3, with 100 cells/experiment). k, Infectious virus titer in cells in (i) with MFF depletion was quantified by infecting SLK cells (top, n = 3) and GFP images are shown (left). Scale bars, 10 μm. Data in a,f,h,i,k are from one experiment that is representative of three independent experiments. Data in b,g,j,k represent mean ± s.d. analyzed by one-way ANOVA followed by Tukey’s post hoc test. Box-plot data in c-e are presented as median (center line), interquartile range (IQR; box, 25^th to 75^th percentiles), and whiskers extending from minimum to maximum values; and analyzed using Kruskal-Wallis with post hoc Dunn’s test. ***, p < 0.001; ****, p < 0.0001; ns, not significant. Exact P values are provided in Source Data Extended Data Fig. 4.

Extended Data Fig. 5 |. Impacts of vBcl-2 on mitochondrial function and integrity.

a, Representative confocal micrographs (left) and quantification (right) of PINK1 (magenta) colocalization with TOM20 (green) in iSLK-BAC16. vBcl-2 KO cells expressing mCherry-vBcl-2 (red) at 60 h post-reactivation (n = 100 cells/sample, 3 independent experiments). Scale bars, 10 μm. b, Co-IP of vBcl-2 and endogenous NM23-H2 with DRP1 in HEK293T cells expressing NM23-H2 shRNA or Ctrl shRNA and transduced with Vec or HA-vBcl-2. c, Co-IP of indicated proteins in HEK293T expressing mCherry-vBcl-2 (WT/E14A). The relative amount of HA-NM23-H2 co-immunoprecipitated with DRP1 was normalized to total HA-NM23-H2 in WCL (right; n = 3 independent experiments). d, DRP1 GTPase activity in the presence/absence of purified NM23-H2, vBcl-2 (WT/E14A), and indicated nucleotides (n = 3 independent experiments). Data in a-c are representative of three independent experiments. See Source Data for uncropped data of b,c. Data in a,c,d are mean ± s.d. analyzed by one-way ANOVA with Tukey’s post hoc test. Exact P values are provided in Source Data Extended Data Fig. 5.

Extended Data Fig. 6 |. KSHV vBcl-2 attenuates MAVS-mediated IFN response.

a-c, RT-qPCR analysis of IFNB1 (a) and ISG15 (c) mRNA levels (normalized to 18S rRNA), and IFN-β production (b) in supernatant measured by ELISA in iSLK-BAC16.vBcl-2 WT and vBcl-2 KO cells induced with Dox/NaB for the indicated time. n = 3. d-f, RT-qPCR analysis of vBcl-2 transcripts (d, n = 3), IFNB1 transcripts (e, n = 3), and ISG15 transcripts (f, n = 3) in TREx BCBL-1-RTA cells transfected with two independent vBcl-2-specific siRNA for 12 h and induced with Dox (1 μg/ml) and Nab (1 mM) for 48 h. g, RT-qPCR of IFNB1 mRNA in HCT116 cells infected with vBcl-2 WT or vBcl-2 KO KSHV at 10 MOI for 60 h (n = 3). h, Representative STED micrographs of MAVS (magenta) and TOM20 (green) in iSLK-BAC16.vBcl-2 WT cells transduced with Ctrl vs. DRP1 shRNAs or NM23-H2 shRNAs at 60 h post-reactivation. Insets highlight MAVS distribution. Arrows indicate MAVS puncta and dotted lines denote clustering. Data are from one experiment that is representative of three independent experiments. i, Diameter of MAVS clusters measured from images in h (n = 100). j. RT-qPCR of ISG15 mRNA expression in h (n = 3). Data in a-g represent mean ± s.d. obtained from three independent experiments performed in triplicates, and analyzed by two-tailed t test (a-c) or one-way ANOVA followed by Tukey’s post hoc test (d-g). Box-plot data in i are presented as median (center line), interquartile range (IQR; box, 25^th to 75^th percentiles), and whiskers extending from minimum to maximum values; and analyzed using Kruskal-Wallis with post hoc Dunn’s test. *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001; ns, not significant. Exact P values are provided in Source Data Extended Data Fig. 6.

Extended Data Fig. 7 |. The NM23-H2 binding-deficient vBcl-2 E14A mutant KSHV is defective in viral capsid assembly and nuclear egress during the late phase of lytic reactivation.

a, RT-qPCR of viral gene expression in iSLK-BAC16.vBcl-2 cells (WT and KO) stably expressing Vec, HA-vBcl-2 (WT/mutants) at indicated time post-reactivation with Dox/NaB (n = 3). b, IB of viral proteins in WCLs from (a) at 48 h post-reactivation. IE/E, immediate early and early genes; L, late genes. c, Relative viral genome copy number changes in cells from (a; n = 3). d, Representative confocal micrographs of ORF65 (red) in iSLK-BAC16. vBcl-2 WT cells expressing Ctrl shRNA, NM23-H2 shRNAs, or DRP1 shRNAs after 60 h Dox/NaB induction. iSLK-BAC16.vBcl-2 KO cells were included as a control. DAPI stains nuclei (blue). Green signals (GFP) mark virus-infected cells. e, Percentage of cells with cytoplasmic ORF65 in (d) was quantified (n = 3, with 100 cells/group). f, IB for ORF65, NM23-H2, and DRP1 expression in cells in (d). g, Confocal micrographs of ORF65 (red) in HCT116 cells infected with vBcl-2 WT or vBcl-2 KO KSHV at 10 MOI for 60 h (left). The percentage of cells with cytoplasmic ORF65 was quantified (right; n = 3, with 100 cells/group). Scale bars, 10 μm. Data in b,d,f,g are from one experiment that is representative of three independent experiments. See Source Data for uncropped data of b,f. Data shown in a,c,e,g represent mean ± s.d. analyzed by Student’s two-tailed t test or one-way ANOVA followed by Tukey’s post hoc test. ***, p < 0.001; ****, p < 0.0001; ns, not significant. Exact P values are provided in Source Data Extended Data Fig. 7.

Extended Data Fig. 8 |. Identification of IFN-induced antiviral factors that restrict vBcl-2-dependent capsid assembly and nuclear egress.

a, Representative confocal images of ORF65 staining in iSLK-BAC16.vBcl-2 KO cells transduced with ISG-specific shRNA as indicated. Scale bars, 10 μm. b, Quantification of the percentage of cells from (a) with cytoplasmic ORF65 staining (100 cells/experiment, n = 3). c, RT-qPCR validation of ISG knockdown in (a, n = 3). d,e, TRIM22 and MxB suppress infectious KSHV production. iSLK-BAC16.vBcl-2 KO cells were transduced with Ctrl or ISG-specific shRNA and induced with Dox/NaB for 72 h. Supernatants were harvested and infectious virus yield was determined by infecting SLK cells for 24 h. Representative GFP/BF images in SLK cells are shown (d). Scale bars, 430 μm. Virus yield was quantified as the percentage of GFP-positive SLK cells (e). DKD, double knockdown (n = 3). f, RT-qPCR analysis of TRIM22 and MxB mRNA in HCT116 cells infected with vBcl-2 WT or vBcl-2 KO KSHV at MOI = 10 for 60 h (n = 3). g-i, Confocal images of ORF65 staining in iSLK-BAC16.vBcl-2 WT/KO cells expressing TRIM22 and MxB post-reactivation. Percentage of cells with cytoplasmic ORF65 (h; n = 3, 300 cells/sample) and the percentage of cytoplasmic ORF65 per cell (i; n = 49 cells/sample) were quantified. Scale bars, 10 μm. Data in a,d,g are from one experiment that is representative of three independent experiments. Data in b,c,e,f,h represent mean ± s.d. analyzed by Student’s two-tailed t test or by one-way ANOVA followed by Tukey’s post hoc test. Scatter plots in i are analyzed with Kruskal-Wallis with post hoc Dunn’s test. *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001; ns, not significant. Exact P values are provided in Source Data Extended Data Fig. 8.

Extended Data Fig. 9 |. VBNI-1 disrupts KSHV vBcl-2 interaction with NM23-H2 and impairs late stage KSHV lytic replication.

a, HTRF assay. His-NM23-H2 is bound by an anti-His-Tb HTRF donor fluorophore. GST-vBcl-2 is bound by an anti-GST-d2 HTRF acceptor. vBcl-2 binding to NM23-H2 produces FRET. The addition of small molecule inhibitor(s) reduces FRET. b, HTRF showing dose-dependent vBcl-2 binding to NM23-H2 (3 nM; red) or NM23-H1(25 nM; blue fitted by nonlinear regression. c, HTRF comparison of vBcl-2 binding to NM23-H2 vs. BID BH3 peptide in the presence of ABT-263; IC₅₀ values indicated (n = 3). d-g, Co-IP of endogenous NM23-H2, Beclin1, and BAK with vBcl-2 in HEK293T cells expressing HA-vBcl-2 after VBNI-1 treatment at increasing doses for 36 h (d,e) or with VBNI-1 (2 μM) at increasing timepoints (f,g). Interactions with indicated binding partners were quantified in (e,g; n = 3). h, The effects of VBNI-1 on NM23-H1/H2 kinase activity measured by Transcreener ADP²-FI assay; inhibition curves with IC₅₀ values shown. i, Docking model (left) shows the electrostatic surface of vBcl-2 harbouring VBNI-1 (sphere). Overlay (right) of vBcl-2-VBNI-1 with vBcl-2-NM23-H2 highlights potential steric clash between VBNI-1 and NM23-H2 interface residue K128. For clarity, the vBcl-2 molecule bound to VBNI-1 is not shown. j, SPR analysis of VBNI-1 interaction with vBcl-2 across concentrations. k, Cytotoxicity of VBNI-1 in SLK (blue) and BJAB (red) cells; median effective concentration (EC₅₀) values indicated. l, Dose-dependent inhibition of KSHV production by VBNI-1 in reactivated iSLK-BAC16.vBcl-2 WT but not KO cells; EC₅₀ values indicated. m-o, Virus production (m,n) and gene expression (o) following VBNI-1 treatment and indicated shRNA knockdowns post-reactivation. GFP/BF images shown in m; quantified in n (n = 3). Scale bars, 430 μm. Data in d,f,j,m are from one experiment that is representative of three independent experiments. See Source Data for uncropped data (d,f). Data in e,g,n,o represent mean ± s.d. analyzed by one-way ANOVA followed by Tukey’s post hoc test. **, p < 0.01; ****, p < 0.0001; ns, not significant. Exact P values are provided in Source Data Extended Data Fig. 9.

Extended Data Fig. 10 |. Schematic representation of the mechanistic action of vBcl-2 manipulation of mitochondrial dynamics for immune subversion and its therapeutic targeting by VBNI-1.

vBcl-2 interaction with host NM23-H2 allows for activation of DRP1-mediated mitochondrial fission, which reduces MAVS aggregation and dampens the downstream type I IFN response, leading to productive virion assembly and nuclear egress. This virus-host interaction can be specifically antagonized by VBNI-1. VBNI-1 blocks the late phase of viral assembly/egress and thwarts virion production through a mechanism that involves induced expression of the ISGs TRIM22 and MxB. See main text for details.

Supplementary Material

Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41564-025-02018-3.

Data availability

The KSHV vBcl-2 gene sequence (gene ID 4961447) was obtained from NCBI. The crystal structures of the human NM23-H2 (PDB 1NSK) and KSHV vBcl-2 (PDB 1K3K) complex were obtained from the Protein Data Bank. The raw RNA sequencing data for BJAB cells expressing vBcl-2 WT (labelled as ‘W’), vBcl-2 E14A (labelled as ‘E’) and empty vector (labelled as ‘V’) have been deposited at NCBI GEO database under accession number GSE292579. RNA sequence reads were aligned to the human genome (UCSC hg38). This database is available at https://www.genome.ucsc.edu/cgi-bin/hgGateway. Source data are provided with this paper.