African swine fever virus B66L drives extracellular mitochondrial release to promote systemic inflammation in mice

Abstract

Systemic inflammation is a hallmark of viral infection, but the upstream signals that initiate it remain poorly defined. Here we show that extracellular mitochondria act as inflammatory mediators during infection by vesicular stomatitis virus, influenza A virus, rabies virus, herpes simplex virus 1 and African swine fever virus (ASFV). In ASFV-infected primary porcine alveolar macrophages, the viral protein B66L promotes the capture of damaged mitochondria by autophagosomes while blocking their fusion with lysosomes, causing mitochondria to accumulate outside cells. Extracellular mitochondria are detected in the serum and bronchoalveolar lavage fluid of mice expressing B66L and in the serum of ASFV-infected pigs. Purified extracellular mitochondria trigger inflammatory cytokine production through cGAS-STING signalling and contribute to lung injury in mice. These findings identify virus-associated extracellular mitochondrial release as a pro-inflammatory mechanism during infection.

Systemic inflammation is frequently associated with viral infection and is largely linked to dysregulation of cytokine networks. Here the authors suggest viral infection can result in extracellular release of mitochondria which can drive inflammatory responses.

Introduction

Systemic inflammation is a hallmark of many viral infections and is often a key determinant of disease severity and mortality^1,2. This dysregulated immune response, typified by excessive cytokine release or “cytokine storms,” is commonly attributed to heightened levels of pro-inflammatory mediators such as interleukin-6 (IL-6), tumor necrosis factor (TNF), and interferons^3,4. However, the upstream molecular events that precipitate such widespread inflammatory cascades remain poorly defined. Elucidating the initiating signals of virus-induced systemic inflammation is essential for devising targeted immunomodulatory interventions.

Beyond their canonical role in cellular energy production, mitochondria have emerged as central regulators of innate immunity. Mitochondrial damage can release immunogenic contents, particularly mitochondrial DNA (mtDNA), into the cytosol, where it serves as a damage-associated molecular pattern (DAMP) sensed by cyclic GMP-AMP synthase (cGAS), thereby activating stimulator of interferon genes (STING)-dependent inflammatory signaling^5–7. While cytosolic leakage of mitochondrial components has been extensively studied, whether mitochondria can be secreted into the extracellular space during viral infection, and whether such extracellular mitochondria (EMs) contribute to systemic inflammation, remains an open question⁸.

Autophagy is a crucial process by which cells degrade and recycle proteins and organelles to maintain intracellular homeostasis. This process can be divided into two main stages: (A) autophagosome formation (sequestration of cargo) and (B) delivery of the autophagosome to either (i) the lysosome for degradation (degradative autophagy) or (ii) the extracellular space via the secretory pathway (secretory autophagy)^9,10. Mitochondrial quality control is primarily governed by autophagy, which targets damaged mitochondria for either lysosomal degradation or secretion^10,11. However, in the context of viral infection, autophagy is frequently hijacked to benefit viral replication or facilitate immune evasion^12,13. Certain viruses actively inhibit autophagosome-lysosome fusion, resulting in the intracellular accumulation of autophagic vesicles¹⁴. This disruption raises the possibility that undigested mitochondria could be rerouted for secretion, potentially amplifying inflammation through paracrine signaling¹⁵.

African swine fever virus (ASFV), a large double-stranded DNA virus and the etiological agent of hemorrhagic fever in swine, is associated with high case fatality and profound immunopathology¹⁶. A hallmark of acute ASFV infection is the rapid onset of a cytokine storm, which substantially contributes to disease progression and tissue damage^17,18. Although several ASFV-encoded proteins have been implicated in manipulating host immune responses-such as H240R, which antagonizes NLRP3 inflammasome activation-the upstream triggers responsible for systemic inflammation remain elusive^19,20.

Here, we uncover a mechanism by which the ASFV-encoded protein B66L alters mitochondrial trafficking by diverting mitochondria-containing autophagosomes away from lysosomal degradation toward secretory autophagy. This mitochondrial release required the core autophagy machinery component ATG5 and the selective autophagy receptor TOLLIP. In both ASFV-infected pigs and mice expressing recombinant adenovirus-encoded B66L (rAdV-B66L), EMs accumulated in the serum and bronchoalveolar lavage fluid (BALF), where they acted as potent paracrine damage-associated molecular patterns (DAMPs). These mitochondria were taken up by recipient macrophages, where they induced robust inflammatory responses through the cGAS-STING signaling axis. Our findings delineated a link between viral subversion of mitochondrial quality control and the propagation of systemic inflammation, revealing EMs as potential mediators of innate immune activation and potential targets for therapeutic intervention in viral disease.

Results

ASFV promotes extracellular mitochondria secretion through B66L

To investigate whether ASFV induces extracellular release of mitochondria, we first subjected culture supernatants from ASFV-infected primary porcine alveolar macrophages (PAMs) to high-speed centrifugation, followed by mass spectrometry and Gene Ontology (GO) enrichment analysis. GO analysis revealed a significant enrichment of mitochondrial components, suggesting that ASFV infection is associated with increased extracellular mitochondrial accumulation (Fig. 1a). Infection at increasing MOIs (0.1, 1, and 5) resulted in a dose-dependent increase in both B66L expression and extracellular mitochondrial markers at 24 hpi, supporting a correlation between viral load and extracellular mitochondrial levels (Supplementary Fig. 1a). Time-course analysis at MOI = 1 revealed that B66L expression became detectable at 12 hpi and progressively increased thereafter, closely followed by a concomitant rise in extracellular mitochondrial markers (TOMM20, TOMM70, and TUFM), consistent with increased extracellular mitochondrial markers following B66L expression (Supplementary Fig. 1b).

Fig. 1. ASFV B66L Protein Promotes Mitochondrial Secretion.

a Differentially abundant proteins in supernatants from ASFV-infected PAMs (36 hpi) were identified using a fold change ≥ 1.5 and P < 0.05. Gene Ontology cellular component enrichment was analyzed by Fisher’s exact test with Benjamini–Hochberg correction; the top 10 terms are shown as −log₁₀(P value). b Schematic of the differential centrifugation workflow used to isolate secreted mitochondria from culture supernatants. Created in BioRender. Hu, B. (2026) https://BioRender.com/qffw7zu. c HEK293T, N2a and PAM cells were infected (MOI = 1) with the indicated viruses. At 24 hpi, secreted mitochondria in conditioned media were analyzed by immunoblotting (n = 3). d Serum samples from mock- (n = 3) or ASFV-infected (n = 11) pigs were analyzed by immunoblotting for mitochondrial proteins. e 3D4/21 cells were transfected with empty vector or Flag-B66L for 48 h. Secreted mitochondria and whole-cell lysates (WCL) were analyzed by immunoblotting (n = 3). f, g Extracellular mitochondrial events were quantified from the live-cell imaging experiments shown in Supplementary Fig. 3. In f WSL cells expressing DsRed-Mito were mock treated with IPTG (n = 5), infected with vB66Li without IPTG (n = 5), or infected with vB66Li with IPTG (n = 6). In g, cells were transfected with GFP vector (n = 5) or GFP-B66L (n = 5). Three fields were analyzed per biological replicate. h, Proteinase K protection assay of secreted mitochondria from Vec- or Flag-B66L-transfected 3D4/21 cells (n = 3). i Supernatants from Vec- or Flag-B66L-transfected 3D4/21 cells were passed through a 0.22 μm filter before mitochondrial isolation and immunoblotting (n = 3). j Secreted mitochondria were analyzed by 5-40% iodixanol density-gradient centrifugation and immunoblotting (n = 3). Representative blots in c–e and h–j are from three independent biological experiments with similar results. For panels a, statistical significance was assessed using multiple testing with the Benjamini-Hochberg false discovery rate (FDR) method. For f statistical significance was determined using one-way ANOVA with Tukey’s multiple-comparison test. For g, statistical significance was assessed using two-sided Student’s t-test. Data represent mean ± SD (*, P  <  0.05; **, P  <  0.01; ns, P > 0.05). Source Data 1 are provided as a Source Data file.

To assess whether this phenomenon is specific to ASFV, we infected multiple cell lines with either active or inactivated viruses, including IAV, VSV, RABV, HSV-1, and ASFV. Large extracellular vesicles (LEV) were isolated by differential centrifugation (Fig. 1b), and immunoblotting revealed elevated levels of mitochondrial markers translocase of outer mitochondrial membrane 20/70 (TOMM20/70) and Tu translation elongation factor, mitochondrial (TUFM) in supernatants from cells infected with active-, but not inactivated-, viruses (Fig. 1c and Supplementary Fig. 1a–d). Similarly, serum samples from ASFV-infected pigs contained higher levels of mitochondrial proteins compared with uninfected controls, confirming mitochondrial secretion during natural infection (Fig. 1d).

To identify viral determinants responsible for ASFV-induced mitochondrial secretion, we screened ASFV-encoded proteins and identified B66L as a potent inducer (Fig. 1e and Supplementary Fig. 1d). To directly assess the role of B66L during infection, we generated an inducible recombinant ASFV (vB66Li) in which B66L expression is controlled by an IPTG-responsive lac operator–repressor system (Supplementary Fig. 2a). vB66Li replicated efficiently regardless of IPTG treatment (Supplementary Fig. 2b, 2c), while B66L protein expression was strictly dependent on inducer presence, without affecting expression of the late viral protein CD2V (Supplementary Fig. 2d). Using an IPTG-inducible recombinant ASFV (vB66Li), mitochondrial secretion was detectable under both induced and non-induced conditions, but was significantly increased upon B66L induction, demonstrating that B66L enhances this process during infection (Supplementary Fig. 2d). Live-cell imaging demonstrated that extracellular mitochondrial signal was observed under both conditions, but occurred at significantly higher frequency upon B66L induction (Fig. 1f, Supplementary Fig. 3a–c and Supplementary Movie 1–22). Moreover, ectopic expression of B66L alone was sufficient to trigger mitochondrial release in the absence of infection (Fig. 1g, Supplementary Fig. 3d, 3e and Supplementary Movie 23–37), and siRNA-mediated knockdown of B66L during ASFV infection significantly reduced mitochondrial secretion (Supplementary Fig. 1e). Notably, screening of additional viral proteins revealed that M2 from influenza A virus, M proteins from RABV and VSV, and HSV-1 UL12.5 similarly promoted mitochondrial secretion, suggesting that this phenomenon is conserved among diverse viruses (Supplementary Fig. 1f–h).

To determine whether secreted mitochondria were vesicle-associated or free, we performed a proteinase K protection assay. TOMM70, TOMM20, and CD9 were digested by proteinase K, whereas TUFM remained protected unless the membrane was disrupted by Triton X-100, indicating that the secreted mitochondria were largely not enclosed within vesicles (Fig. 1h). Consistently, filtration of supernatants through a 0.22 µm membrane removed mitochondrial proteins but not the exosomal marker CD9, ruling out contamination by small extracellular vesicles (Fig. 1i). Moreover, density gradient centrifugation further showed that mitochondrial markers were enriched in fractions distinct from exosomal CD9 (Fig. 1j).

Together, these results indicate that ASFV infection is associated with increased extracellular accumulation of mitochondrial factors.

Autophagosome formation is required for B66L-mediated damaged-mitochondria secretion

To determine whether autophagosome formation is required for B66L-mediated mitochondrial secretion, we first pharmacologically modulated autophagy. Treatment with the autophagy inhibitor wortmannin markedly suppressed extracellular release of TOMM70, TOMM20, and TUFM, whereas rapamycin enhanced their release in the presence of B66L expression (Fig. 2a). These effects were absent in cells lacking B66L (Supplementary Fig. 4a), indicating that autophagy facilitates B66L-driven mitochondrial secretion. Confocal imaging further revealed that B66L selectively increased colocalization of mitochondria with GFP-GABARAP, but not with GFP-LC3B (Fig. 2b), suggesting a preference for the GABARAP subtype. Consistently, deletion of essential autophagy genes (ATG5 or VPS34) abolished mitochondria-GABARAP colocalization (Supplementary Fig. 4b, 4c). Moreover, knockout of VPS34, ATG5, or ATG7 eliminated B66L-induced mitochondrial secretion without affecting basal release, and B66L-mediated lipidation of GABARAP, but not LC3B, while re-expression of the respective genes restored secretion and lipidation (Fig. 2c–e). Transmission electron microscopy (TEM) analysis confirmed that ASFV infection or B66L expression significantly increased extracellular mitochondria (EMs), whereas knockout of VPS34, ATG5, or ATG7 markedly reduced their numbers in B66L expression cells (Supplementary Fig. 4d, e). Collectively, these data demonstrate that autophagosome formation is essential for B66L-mediated mitochondrial secretion.

Fig. 2. ASFV B66L promotes autophagy-dependent mitochondrial secretion.

a PAM was transfected with a vector expressing B66L for 30 h, and treated with DMSO (Con.), wortmannin (Wor., 10 μM), or rapamycin (Rapa., 5 μM) for 6 h. PAM were then subjected to immunoblot analysis using the indicated antibodies (n = 3). b 3D4/21 cells co-transfected with an empty vector or a vector expressing B66L and GFP-GABARAP/LC3B and DsRed-Mito for 48 h were analyzed by confocal microscopy (n = 5). c–e WT, VPS34-KO, ATG5-KO, or ATG7-KO HEK293T cells were transfected with an empty vector or Flag-B66L for 48 h. Secreted mitochondria were isolated as described in Fig. 1b and subjected to immunoblotting analysis using indicated antibodies (n = 3). f WT- or VPS34-KO 3D4/21 cells transfected with an empty vector or a vector expressing DsRed-Mito for 36 h were subjected to confocal microscopy following immunofluorescence staining with the indicated antibodies. Scale bar: 5 μm (n = 5). g Quantification of Pearson’s correlation coefficient between phosphorylated ubiquitin (p-Ub) and mitochondria from (f) (n = 5). The correlation coefficient was calculated to assess the co-localization of p-Ub and mitochondria in 3D4/21 cells transfected with either an empty vector or a vector expressing DsRed-Mito for 36 h. Representative images and immunoblots in a–f are shown from at least three independent biological experiments with similar results. For a, c–e, g, statistical significance was determined using one-way ANOVA with Tukey’s multiple-comparison test. Data represent mean ± SD (*, P  <  0.05; **, P  <  0.01; ns, P > 0.05). Source Data 1 is provided as a Source Data file.

We next asked whether the secreted mitochondria were damaged. Immunoblotting of conditioned media showed enrichment of Ser65-phosphorylated ubiquitin (p-Ub), a marker of mitochondrial damage^10,21, upon B66L expression; this signal was abolished in autophagy-deficient backgrounds and rescued by gene reintroduction (Fig. 2c-e). Consistently, B66L-expressing cells exhibited increased mitochondrial p-Ub staining, reduced mitochondrial membrane potential (Δψm), elevated reactive oxygen species (ROS), and decreased ATP levels, confirming extensive mitochondrial dysfunction (Fig. 2f, g and Supplementary Fig. 4f–h). Furthermore, we found that B66L disrupted the interaction between TOMM20 and TOMM22, a complex essential for maintaining mitochondrial protein import and function²²(Supplementary Fig. 4i). Together, these findings indicate that B66L induces mitochondrial damage, likely through interfering with the TOMM20-TOMM22 complex.

Importantly, treatment with the mitochondrial complex I inhibitor rotenone or with cycloheximide (CHX), although both strongly induce mitochondrial injury or cellular stress, did not trigger extracellular mitochondrial release (Supplementary Fig. 4j–l). Thus, generic mitochondrial damage or cell stress alone did not recapitulate the increased mitochondrial secretion observed under B66L expression conditions. Instead, B66L actively promotes damaged mitochondria into GABARAP-dependent secretory autophagosomes for extracellular release.

Collectively, these findings demonstrate that B66L specifically harnesses the autophagy machinery to promote damaged mitochondria into GABARAP-dependent secretory autophagosomes, thereby promoting their extracellular release.

B66L promotes secretion of mitochondria through inhibition of autophagosome maturation

To determine whether B66L affects mitophagosome maturation, we used a tandem RFP-GFP-Mito reporter. In this system, mitochondria delivered into acidic lysosomes appear red due to selective quenching of GFP, whereas mitochondria retained in non-acidified vesicles remain yellow. In control cells, CCCP treatment induced the expected GFP-to-RFP shift, reflecting mitophagy completion. By contrast, ASFV infection or B66L expression maintained persistent yellow puncta even under CCCP treatment (Fig. 3a and Supplementary Fig. 5a). Consistently, B66L, as well as IAV M2, VSV M, RABV M, and HSV-1 UL12.5, increased GABARAP lipidation under basal conditions but did not further enhance lipidation in the presence of chloroquine (Supplementary Fig. 5b–d), supporting a block in autophagosome-lysosome fusion.

Fig. 3. B66L promotes mitochondrial secretion by inhibiting autophagosome maturation.

a 3D4/21 cells were transfected with an empty vector or a vector expressing RFP-GFP-Mito for 20 h. Transfected cells were treated with CCCP for 4 h and analyzed by confocal microscopy. Scale bar: 5 μm. Quantification of the number of dots per cell was performed (n = 20). b Schematic diagram of full-length or truncated B66L protein structures. c, d HEK293T cells were cotransfected with vectors expressing GFP-tagged full length B66L, or its CCD1-deletion mutant (GFP-ΔCCD1) or CCD2-deletion mutant (GFP-ΔCCD2) and mcherry-PLEKHM1 (Mch.-PLEK.) and Myc-GABARAP (Myc-GA.) for 48 h. CO-IP assay and immunoblotting analysis were then performed by using indicated antibodies (n = 3). e 3D4/21 cells were transfected with an empty vector or vectors expressing the indicated genes for 20 h. Transfected cells were treated with CCCP for 4 h, fixed, and subjected to confocal microscopy. Scale bar: 5 μm (n = 10). f Quantification of Pearson’s correlation coefficient was performed to assess the co-localization in (e) (n = 10). g 3D4/21 cells were transfected with an empty vector or vectors expressing the indicated genes for 20 h, treated with CCCP for 4 h, fixed, and analyzed by confocal microscopy. Scale bar: 5 μm (n = 10). h Quantification of Pearson’s correlation coefficient was performed to evaluate co-localization in (g) (n = 10). Representative images and immunoblots in a, c, d, e, g, are shown from at least three independent biological experiments with similar results. For a, statistical significance was assessed using two-way ANOVA with Tukey’s multiple-comparison test. For f, h, statistical significance was determined using one-way ANOVA with Dunnett’s multiple-comparison test. Data represent mean ± SD (*, P  <  0.05; **, P  <  0.01; ns, P > 0.05). Source Data 1 are provided as a Source Data file.

The interaction between GABARAP/LC3B and PLEKHM1 (Pleckstrin homology domain-containing family M member 1) regulates autophagosome-lysosome fusion²³. We found that B66L selectively disrupted the GABARAP-PLEKHM1 interaction (Supplementary Fig. 5e), without affecting LC3B-PLEKHM1 binding (Supplementary Fig. 5f). Domain analysis revealed two predicted coiled-coil domains (CCDs) in B66L, and mutational studies showed that CCD1, but not CCD2, mediated competitive binding to PLEKHM1, thereby interfered the interaction between PLEKHM1 and GABARAP (Fig. 3b–d). In line with this, ASFV infection or B66L expression, but not CCD1-lacking mutant, inhibited GABARAP-PLEKHM1 colocalization and prevented LAMP1 from associating with mitochondria (Fig. 3e–h, Supplementary Fig. 5g–k), further confirming impaired mitophagosome maturation.

Functionally, PLEKHM1 knockout alone was sufficient to elevate extracellular mitochondrial protein release, and B66L expression did not further enhance this phenotype (Supplementary Fig. 5l). Together, these results demonstrate that B66L promotes secretion of damaged mitochondria by inhibiting autophagosome maturation through disruption of the GABARAP-PLEKHM1 axis.

Autophagy Cargo receptor TOLLIP is required for B66L-mediated secretory of mitochondria

To determine the role of autophagy cargo receptors in B66L-driven mitochondrial secretion, we systematically evaluated autophagy cargo receptors TOLLIP, HDAC6, NDP52, NIX, OPTN, and SQSTM1. Among these candidates, only TOLLIP markedly enhanced the extracellular release of mitochondrial proteins (TOMM20, TOMM70, TUFM) in the presence of B66L (Supplementary Fig. 6a). To test sufficiency, we reintroduced individual cargo receptors into tetra knockout (KO) cells lacking the main mitophagy cargo receptor: NDP52, OPTN, SQSTM1, and TOLLIP²⁴. Strikingly, only TOLLIP reconstitution restored B66L-induced mitochondrial release (Fig. 4a and Supplementary Fig. 6b). Conversely, TOLLIP deficiency suppressed B66L-mediated mitochondrial secretion without affecting basal levels, and re-expression of TOLLIP rescued this defect (Fig. 4b).

Fig. 4. B66L-mediated mitochondrial secretion depends on the autophagy cargo receptor TOLLIP.

a HEK293T tetra-KO cells lacking NDP52, OPTN, TOLLIP and SQSTM1 were co-transfected with empty vector or vectors expressing Flag-B66L, Flag-NDP52, Flag-OPTN, Flag-TOLLIP or Flag-SQSTM1 for 48 h, followed by immunoblotting (n = 3). b WT or TOLLIP-KO HEK293T cells were transfected with empty vector, Flag-B66L or Flag-TOLLIP for 48 h. Secreted mitochondria were isolated as in Fig. 1b and analyzed by immunoblotting (n = 3). c HEK293T cells were co-transfected with GFP-GABARAP and WT TOLLIP, TOLLIP lacking LIR1 (ΔLIR1), LIR2 (ΔLIR2), or both LIR1 and LIR2 (ΔLIR1 + 2) for 48 h, followed by co-immunoprecipitation and immunoblotting (n = 3). d PAM cells were transfected with the indicated constructs for 48 h, and secreted mitochondria in conditioned media were analyzed by immunoblotting (n = 3). e HEK293T cells were transfected with GFP-tagged full-length B66L, or mutants lacking CCD1 (ΔCCD1), CCD2 (ΔCCD2), or both domains (ΔCCD1 + 2) for 48 h. CO-IP assay and immunoblotting analysis of endogenous TOMM20 or TOLLIP were then performed by using antibodies against the indicated proteins (n = 3). f HEK293T cells were transfected with the indicated constructs for 48 h, followed by co-immunoprecipitation and immunoblotting. (n = 3). g 3D4/21 cells were cotransfected with BFP-tagged full length B66L or the indicated CCD deletion mutants together with DsRed-Mito, GFP-GABARAP and Flag-TOLLIP for 36 h, then analyzed by confocal microscopy. Scale bar, 5 μm (n = 3). h PAMs were cotransfected with GFP-tagged full-length B66L or the indicated CCD deletion mutants for 48 h, followed by immunoblotting of endogenous TOMM70, TOMM20, TUFM and phospho-Ser65 ubiquitin (n = 3). Representative images and immunoblots in a–h are shown from at least three independent biological experiments with similar results. For a, b, statistical significance was determined using one-way ANOVA with Tukey’s multiple-comparison test. Data represent mean ± SD (*, P  <  0.05; **, P  <  0.01; ns, P > 0.05). Source Data 1 is provided as a Source Data file.

Previous studies suggested that TOLLIP harbors two putative LC3-interacting region (LIR) motifs. Consistent with this, our biochemical mapping showed that mutation of both LIR motifs simultaneously abolished TOLLIP-GABARAP binding (Fig. 4c), indicating that TOLLIP directly engages GABARAP via its intrinsic LIRs rather than through an adaptor. Importantly, disruption of both LIRs abrogated the ability of TOLLIP to promote B66L-driven mitochondrial secretion (Fig. 4d). Of note, B66L did not alter TOLLIP-GABARAP binding itself (Supplementary Fig. 6c).

We next investigated whether TOLLIP functions as a receptor that links mitochondria to autophagic membranes. Indeed, TOLLIP physically interacted with the mitochondrial TOMM20, an outer membrane protein of mitochondria to evaluate mitophagy²⁵, and this interaction was further enhanced in the presence of B66L (Supplementary Fig. 6d). Mechanistically, B66L simultaneously bound TOLLIP through its CCD1 domain and TOMM20 through its CCD2 domain, thereby bridging TOLLIP to mitochondria (Fig. 4e). Deletion of either CCD1 or CCD2 disrupted B66L-induced TOLLIP-TOMM20 association, impaired TOLLIP-mitochondria colocalization (Fig. 4f, g). Consistently, lack of either CCD1 or CCD2 abolished the effect of B66L on mitochondrial export (Fig. 4h).

Together, these results identify TOLLIP as the essential autophagy cargo receptor for B66L-mediated mitochondrial secretion, acting through direct engagement with GABARAP and B66L-dependent recruitment to mitochondria.

Autophagy is required for B66L-induced secretion of mitochondria in vivo

To assess whether B66L promotes mitochondrial secretion in vivo, mice were infected with recombinant adenoviruses expressing either an empty vector (rAdV) or B66L (rAdV-B66L) (Fig. 5a). Transmission electron microscopy (TEM) of serum samples revealed a pronounced accumulation of damaged mitochondria in rAdV-B66L-infected mice (Fig. 5b, c). Consistently, the serum levels of mitochondrial proteins (TOMM70, TOMM20, TUFM) and Ser65-phosphorylated ubiquitin, a marker of PINK1-dependent mitochondrial damage^10,21, were markedly increased upon B66L expression, but not following B175L expression (Fig. 5d, e and Supplementary Fig. 7a). Notably, these effects were abolished in Tollip- or Atg5-deficient mice, indicating that both cargo loading and autophagosome formation are essential for B66L-driven mitochondrial release.

Fig. 5. B66L promotes release of mitochondria into mice serum and BALF.

a Schematic representation of the mice infected with recombinant adenovirus expressing control or B66L. The 6-week-old C57BL/6 J WT, Atg5-KO, Tollip-KO mice were intranasally infected with rAdV and rAdV-B66L at a dose of 10^9.0 PFU for 72 h, respectively. Created in BioRender. Hu, B. (2026) https://BioRender.com/igkntfa. b The mitochondria in serum from (A) were isolated and subjected to transmission electron microscopy (TEM) observation (n = 5). The schematic illustration was created in part with BioRender. Hu, B. (2026) https://BioRender.com/z5ijl8y. c The number of damaged secretory mitochondria was quantified and subjected to statistic analysis (n = 5). d, e Immunoblotting analysis of mitochondria proteins in serums from (A) (n = 3). f The mice from (a) or infected with rAdV-B175L were euthanized for tissue transmission electron microscopy (TEM) observation in the lungs (n = 5). The schematic illustration was created in part with BioRender. Hu, B. (2026) https://BioRender.com/wiv2sh2. g Schematic for the differential centrifugation assay used to isolate secreted mitochondria from mice bronchoalveolar lavage fluid (BALF). Created in BioRender. Hu, B. (2026) https://BioRender.com/w5fllr0. h, i Immunoblotting analysis of mitochondria proteins in BALF from (a). Animal experiments were performed using independent cohorts of mice (n = 3). Representative images and immunoblots in b, d–f, h, i are shown from at least three independent biological experiments with similar results. For c, statistical significance was assessed using two-sided Student’s t-test. For d, e, h, i, statistical significance was determined using one-way ANOVA with Tukey’s multiple-comparison test. Data represent mean ± SD (*, P  <  0.05; **, P  <  0.01; ns, P > 0.05). Source Data 1 is provided as a Source Data file.

Examination of lung tissue further revealed abundant extracellular mitochondria in intercellular spaces of rAdV-B66L-infected, but not rAdV-B175L-infected, mice (Fig. 5f). Similarly, bronchoalveolar lavage fluid (BALF) from rAdV-B66L-infected mice exhibited increased mitochondrial proteins and Ser65-phosphorylated ubiquitin (Fig. 5h, i), whereas these effects were absent in Tollip- or Atg5-deficient mice and in rAdV-B175L controls (Supplementary Fig. 7b).

Together, these data demonstrate that B66L triggers the secretion of damaged mitochondria in vivo through an autophagy-dependent pathway requiring both TOLLIP-mediated cargo recognition and ATG5-dependent autophagosome formation.

EMs promote inflammation via the cGAS-STING pathway

To assess the inflammatory potential of extracellular mitochondria (EMs) released upon B66L expression, we purified mitochondria from the conditioned media of His-TOMM70-transfected 3D4/21 cells with or without B66L (Fig. 6a). TEM revealed that >60% of EMs from B66L-expressing cells exhibited ruptured outer membranes, compared with <15% in controls (Fig. 6b). Consistently, immunoblotting analysis of equal isolated mitochondria showed elevated Ser65-phosphorylated ubiquitin, confirming enhanced mitochondrial damage (Supplementary Fig. 8a).

Fig. 6. B66L-induced secreted mitochondria promote inflammatory responses.

a Schematic of the affinity purification of mitochondria from the culture medium of cells expressing B66L or control vector. Created in BioRender. Hu, B. (2026) https://BioRender.com/wt4tz4i. b TEM analysis of mitochondria from (a). Five independent preparations with three random fields each were analyzed. Mitochondria were scored as “healthy” (intact membranes and cristae) or “damaged” (fragmented/swollen morphology, disrupted cristae). Data show the percentage of healthy or damaged mitochondria per preparation (n = 5). c Extracellular mitochondria (EMs) from from Vec.- or Flag-B66L-transfected PAM cells were applied to PAM cells for RT-qPCR analysis of IL1B, IL6 and TNFA (n = 15). d EMs obtained from culture medium of HEK293T cells infected with the indicated viruses (MOI:1) were applied to RAW264.7 cells for 12 h, followed by RT-qPCR analysis of IL1B, IL6, and TNFA (n = 15). e EMs from Vec.- or Flag-B66L-transfected 3D4/21 cells were applied to PAM cells in the presence or absence of cGAS or STING inhibitor (n = 15). f Schematic representation of the mice intranasally administered with EMs derived from B66L-expressing 3D4/21 cells or viruses-infected PAM or 293 T cells. Created in BioRender. Hu, B. (2026) https://BioRender.com/4l7578o. g Representative H&E staining of lungs from mice treated for 72 h with EMs derived from B66L-expressing 3D4/21 cells (n = 5). h Serum IL-1β, IL-6 and TNF-α in mice treated for 72 h with mitochondria from empty vector- or B66L-transfected 3D4/21 cells were measured by ELISA (n = 15 mice per group). i Serum IL-1β, IL-6 and TNF-α in mice treated for 72 h with mitochondria from HEK293T or N2a cells infected with the indicated viruses were measured by ELISA (n = 15 mice per group). Representative images and immunoblots in b, g are shown from five independent biological experiments with similar results. For b, c, h, statistical significance was assessed using two-sided Student’s t-test. For d, e, i, statistical significance was determined using one-way ANOVA with Tukey’s or Dunnett’s multiple-comparison test. Data represent mean ± SD (*, P  <  0.05; **, P  <  0.01; ns, P > 0.05). Source Data 1 are provided as a Source Data file.

When applied to recipient PAM (primary porcine alveolar macrophage), these EMs robustly induced IL1B, IL6, and TNFA expression (Fig. 6c). EMs isolated from cells infected with ASFV, IAV, VSV, RABV, or HSV elicited comparable inflammatory responses in PAMs or RAW264.7 cells (Supplementary Fig. 8b and Fig. 6d).

To identify the dominant damage-associated molecular patterns (DAMPs), we tested inhibitors of ATP, succinate, N-formyl peptides, and cGAS. Only cGAS inhibition markedly suppressed cytokine production (Supplementary Fig. 8c, d). Pharmacological inhibition of cGAS or STING confirmed that the response required cGAS-STING signaling (Fig. 6e).

RNAi knockdown further showed that TLR9, AIM2, and ZBP1 depletion modestly reduced cytokine expression, while NLRP3 knockdown selectively impaired IL-1β release, consistent with its role in inflammasome activation. By contrast, cGAS knockdown broadly suppressed levels of IL-6, IL-1β, and TNF-α (Supplementary Fig. 8e–g), highlighting cGAS as the central mediator of EM-induced inflammation.

In vivo, intranasal delivery of EMs from B66L-expressing, but not B175L-expressing, 3D4/21 or viruses-infected PAM, 293 T or N2a cells provoked severe lung injury and elevated IL-1β, IL-6, and TNF-α in serum and multiple organs, whereas EMs from control cells had minimal effects (Fig. 6f-h and Supplementary 8h–j). Similarly, EMs derived from virus-infected cells triggered systemic inflammation (Fig. 6i).

Together, these results demonstrate that B66L-induced EMs are sufficient to trigger robust inflammatory responses through the cGAS-STING pathway.

B66L promotes inflammation through autophagy- and cGAS-STING-mediated EMs release

To explore whether B66L promotes inflammation through autophagy-dependent mitochondrial release, we used a transwell-based co-culture system. We selected Vero cells as donor cells due to their weak innate immune responses and lack of type I interferon production, which minimizes confounding cytokine release caused by B66L expression, thereby isolating the effect of EMs. WT, ATG5-KO, or STING-KO Vero cells were transfected with B66L or control vectors along with GFP-TOMM20 to label mitochondria, and co-cultured with RAW264.7 macrophages (WT, ATG5-KO, or STING-KO) in transwell chambers (Fig. 7a). Mitochondrial transfer occurred only with 1 μm filters but not 0.1 μm, confirming the involvement of large particles such as mitochondria rather than small extracellular vesicles (Fig. 7b). Notably, ATG5 knockout in donor cells completely abolished mitochondrial transfer, whereas knockout of STING in donor cells or knockout of either ATG5 or STING in recipient cells had no effect, indicating that autophagy in donor cells is essential for mitochondrial secretion.

Fig. 7. B66L Induces Inflammation in vivo.

a Schematic of the co-culture system using WT, ATG5-knockout (KO) or STING-KO Vero cells, with or without B66L expression, in the upper chamber and WT, ATG5-KO or STING-KO RAW264.7 cells in the lower chamber. Created in BioRender. Hu, B. (2026) https://BioRender.com/11czpbu. b Representative images showing mitochondrial transfer (green) from donor Vero cells to RAW264.7 cells after 48 h of co-culture. Scale bar, 20 μm (n = 5). c WT Vero cells expressing GFP-labelled mitochondria and either control or B66L were co-cultured with WT RAW264.7 cells for 48 h in transwells with 0.1 μm or 1 μm pore filters. IL-1β, IL-6, TNF-α levels in the medium were measured by ELISA (n = 15). d WT, ATG5-KO, or STING-KO Vero cells expressing GFP-mitochondria and either control or B66L were co-cultured with WT RAW264.7 cells for 48 h using 1.0 μm pore filters. Cytokine levels in the medium were determined by RT-qPCR (n = 15). e WT Vero cells expressing GFP-mitochondria and either control or B66L were co-cultured with WT, ATG5-KO, or STING-KO RAW264.7 cells in 1 μm transwells for 48 h. IL1B, IL6, TNFA levels were analyzed by RT-qPCR (n = 15). f PAM were infected with MOCK, rAdV, or rAdV-B66L and treated with either control or cGAS or STING inhibitor for 4 h. Cytokine levels in culture supernatants were measured by ELISA (n = 15). g Six-week-old C57BL/6 J mice were intranasally infected with rAdV or rAdV-B66L (10⁹ PFU) for 3 days. Serum IL-1β, IL-6, and TNF-α levels were measured by ELISA (n = 15 mice per group). h–i Lung pathology from g was assessed by H&E staining, and the infiltration index was quantified (n = 5 mice per group). Representative images and immunoblots in b, h are shown from five independent biological experiments with similar results. For d, e statistical significance was assessed using two-sided Student’s t-test. For c, f–g, i, statistical significance was determined using one-way ANOVA with Tukey’s multiple-comparison test. Data represent mean ± SD (*, P  <  0.05; **, P  <  0.01; ns, P > 0.05). Source Data 1 are provided as a Source Data file.

Conditioned medium from recipient cells showed that uptake of mitochondrial going through 1 μm filters, but not 0.1 μm, significantly elevated the levels of proinflammatory cytokines IL-1β, IL-6, and TNF-α (Fig. 7c). Size-exclusion filtration (0.22 μm) of media from B66L-expressing PAM cells eliminated cytokine induction, confirming that mitochondria, rather than small EVs, as the inflammatory stimulus (Supplementary Fig. 9a). Similarly, autophagy inhibition by ATG5 knockout or wortmannin treatment suppressed B66L-induced inflammation (Fig. 7d and Supplementary Fig. 9b), while STING knockout in donor cells had no effect, reinforcing the role of donor-cell autophagy in mitochondrial release.

In contrast, knockout of STING, but not ATG5, in recipient cells markedly reduced cytokine levels, suggesting that cGAS-STING sensing in recipient macrophages is required for the inflammatory response (Fig. 7e). Consistently, infection of porcine alveolar macrophages (PAMs) with rAdV-B66L increased pro-inflammatory cytokine expression, which was blocked by cGAS and STING inhibitors (Fig. 7f), supporting the role of the cGAS-STING pathway in sensing EMs.

In vivo, rAdV-B66L infection significantly elevated serum levels of IL-1β, IL-6, and TNF-α in wild-type (WT) mice, but not in Atg5-, Tollip-, Cgas-, or Sting-deficient mice (Fig. 7g), indicating that either autophagy or the cGAS-STING axis is essential for the inflammatory response. Hematoxylin and eosin (H&E) staining revealed pronounced immune cell infiltration and lung tissue damage in rAdV-B66L-infected WT mice, but not in the Atg5-, Tollip-, Cgas-, or Sting-deficient mice (Fig. 7h, i).

Similarly, LEVs from ASFV-infected PAM cells significantly induced inflammation and lung damage in mice. However, when ATG5 or TOLLIP expression was knocked down in ASFV-infected PAM cells, LEVs failed to elevate IL1B, IL6, and TNFA levels, nor did it cause lung damage, confirming that autophagy is essential for the secretion of EMs (Supplementary Fig. 9c–e).

In contrast, purified EMs from ASFV-infected WT-PAM cells did not induce inflammation or lung damage in Cgas-, or Sting-deficient mice (Supplementary Fig. 9f–h), indicating that the cGAS-STING pathway is required for EM-induced inflammation.

Together, these findings demonstrate that B66L promotes inflammation by inducing autophagy-dependent EMs release from donor cells, which is sensed by the cGAS-STING pathway in recipient immune cells (Fig. 8).

Fig. 8. A mechanistic model for inflammation triggering by virus-induced extracellular mitochondria (EMs).

The African swine fever virus (ASFV)-encoded protein B66L facilitates the sequestration of damaged mitochondria into autophagosomes while concurrently inhibiting their fusion with lysosomes, leading to the extracellular release of intact mitochondria. These extracellular mitochondria are subsequently internalized by recipient macrophages, where they activate the cyclic GMP-AMP synthase (cGAS)-stimulator of interferon genes (STING) pathway. This activation induces the production of proinflammatory cytokines, including IL-1β, IL-6, and TNF-α, across multiple organs, culminating in systemic inflammation and lung tissue damage. Created in BioRender. Hu, B. (2026) https://BioRender.com/c0qvkl9.

Discussion

Our study uncovers a mechanism by which viruses, particularly African swine fever virus (ASFV), are associated with systemic inflammation through the release of EMs. This process is enhanced by the viral protein B66L, which recruits TOLLIP to facilitate the packaging of damaged mitochondria while concurrently inhibiting autophagosome-lysosome fusion, leading to their extracellular secretion. These EMs, characterized by ruptured outer membranes and elevated levels of Ser65-phosphorylated ubiquitin, accumulate in both serum and bronchoalveolar lavage fluid of infected hosts. Functionally, their presence activates the cGAS-STING signaling pathway in recipient immune cells, culminating in the production of proinflammatory cytokines and subsequent lung tissue damage.

This discovery refines the traditional understanding that systemic inflammation during viral infections is primarily driven by cytokine overproduction. By identifying virus-induced EMs as upstream activators of inflammation through the cGAS-STING pathway, our study reveals how mitochondrial release amplifies the inflammatory response. Unlike previous reports where mitochondrial components like mtDNA are passively released during cell injury, our findings suggest that viral proteins can modulate host autophagy pathways in a manner that enhances extracellular mitochondrial accumulation. This mechanism enhances inflammatory responses and provides a pathway through which viruses exacerbate disease pathology.

Our findings show that TOLLIP functions as a bona fide autophagy receptor, directly engaging GABARAP via its LIR motifs and recognizing the mitochondrial outer membrane protein TOMM20 as a cargo. Notably, B66L enhances the TOLLIP-TOMM20 interaction, amplifying TOLLIP’s receptor function to promote mitochondrial secretion. While these findings establish TOLLIP as an autonomous receptor in our system, we cannot exclude that under physiological conditions it may act in an adaptor-like manner in cooperation with other receptors.

Our findings build upon previous studies, such as those by Woo Hyun Shin et al., which showed that TOLLIP negatively regulates mitophagy²⁶. However, it is believed that when mitophagy is deficient, alternative mechanisms, such as mitochondrial secretion, is activated to handle damaged mitochondria^10,27. Interestingly, our study demonstrates that TOLLIP mediates secretory autophagy, facilitating the release of damaged mitochondria into the extracellular space. This positions TOLLIP as a non-classical autophagic cargo receptor involved in mitochondrial release rather than degradation. In line with these findings, recent work by Hamid Reza Nouri et al. showed that Tollip knockout mice exhibit reduced extracellular secretion of mtDNA and its presence in bronchoalveolar lavage fluid (BALF) following IAV infection²⁸, further confirming TOLLIP’s critical role in regulating mitochondrial and mtDNA secretion during viral infections. Collectively, these results emphasize TOLLIP’s dual role in autophagy: as a negative regulator of mitophagy and a key mediator of mitochondrial secretion. Our study shows how viruses exploit TOLLIP to manipulate mitochondrial fate and modulate immune responses via secretory autophagy, expanding the current understanding of TOLLIP’s involvement in mitochondrial dynamics during infection.

Our study demonstrates that ASFV, via B66L, selectively modulates GABARAP-dependent mitophagy while sparing LC3B-dependent autophagy. B66L promotes mitochondrial packaging for extracellular secretion and simultaneously blocks autophagosome-lysosome fusion, preventing mitochondrial degradation. This dual function preserves immunogenic extracellular mitochondria (EMs), which act as potent activators of innate immunity. Interestingly, this immune activation contrasts with the activity of other ASFV proteins, such as L83L, which suppresses the cGAS-STING pathway by targeting STING for degradation via TOLLIP recruitment, although the autophagy subtypes involved remain unclear²⁹. Together, these observations suggest that ASFV exploits distinct autophagy pathways to selectively control the fate of specific substrates, creating a cellular environment that both modulates host immunity and favors viral replication. This paradoxical interplay between immune activation and suppression highlights a sophisticated viral strategy in which different viral proteins fine-tune host defenses to maximize viral fitness.

A previous study demonstrated that the ASFV protein p17 induces mitophagy by facilitating the interaction between SQSTM1 and TOMM70, leading to the engulfment of mitochondria by LC3B-dependent autophagosomes³⁰, which suggests a pathway toward mitochondrial degradation or secretion. In contrast, our study shows that the ASFV protein B66L recruits damaged mitochondria to TOLLIP and directs them into GABARAP-dependent autophagosomes. However, B66L inhibits the fusion of these GABARAP-dependent autophagosomes with lysosomes, promoting the extracellular secretion of mitochondria via secretory autophagy. Notably, B66L does not affect the fusion of LC3-dependent autophagosomes with lysosomes. These findings indicate that ASFV employs distinct autophagic pathways depending on the viral protein involved, regulating different ATG8 family proteins. Specifically, ASFV uses LC3B-dependent mitophagy for mitochondrial degradation, while GABARAP-dependent secretory autophagy facilitates mitochondrial release. Both mechanisms contribute to the virus’s ability to modulate host immune responses and support viral replication by clearing damaged mitochondria and influencing cellular stress responses.

B66L may employ a mechanism similar to that of α-synuclein, disrupting the interaction between TOMM20 and TOMM22 and thereby impairing mitochondrial protein import. Extending this concept to other viruses, proteins such as IAV M2, VSV M, RABV M, and HSV-1 UL12.5 have also been reported to induce mitochondrial damage^31–34. Notably, IAV M2 and RABV M have been shown to inhibit autophagosome maturation^35,36. Our findings indicate that these viral proteins, including B66L, share two conserved functional features: triggering mitochondrial damage and blocking autophagic flux-both of which synergistically promote mitochondrial secretion. This evidence supports the notion that mitochondrial release may reflect a broader viral-associated phenomenon to divert damaged mitochondria from degradation toward secretory autophagy. By amplifying pro-inflammatory signaling through extracellular mitochondria while suppressing mitophagy-dependent antiviral defenses, viruses may create a cellular environment conducive to their replication and persistence.

In summary, our study reveals a previously unrecognized mechanism of virus-induced systemic inflammation mediated by EMs. It demonstrates how ASFV B66L manipulates host autophagy and immune sensing machinery to promote the release of DAMP-like EMs, triggering cGAS-STING-dependent inflammation. This work expands the current understanding of DAMPs in viral infections and underscores the therapeutic potential of targeting the EM-cGAS-STING axis to alleviate tissue damage and excessive inflammation caused by viral pathogens. While basal mitochondrial release may reflect a host stress-adaptive response, our data demonstrate that viral B66L expression quantitatively further promotes this pathway, thereby converting a stress-associated process into a pro-inflammatory amplification mechanism.

Methods

Ethics statement

All the animal experiments were conducted in accordance with procedures approved by the Animal Ethical and Welfare Committee for Institutional Animal Care and Use Committee (IACUC) of Zhejiang University (approval nos. ZJU20220495, ZJU20250834 and ZJU20260078).

Antibodies and reagent

Anti-TOMM70 (A21210), anti-TOMM20 (A19403), anti-TUFM (A6423), anti-ATG5 (A19677), anti-TOLLIP (A21551), anti-CD9 (A24130), anti-TLR9 (A14642), anri-ZBP1 (A28338) rabbit polyclonal antibodies (pAb), and anti-ACTIN (AC026), anti-ATG7 (A21895), anti-LC3B (A19665), anti-GABARAP (A4335), anti-cGAS (A27527), anti-NDP52 (A24021), anti-SQSTM1 (A19700), anti-GFP(AE078) rabbit mAb were purchased from ABclonal Technology (Wuhan, China). Rabbit mAb against ATG7 (81760-1-RR) was purchased from Proteintech (Wuhan, China). Mouse mAb against Flag (F1804) was purchased from Sigma-Aldrich (St. Louis, MO, USA). Rabbit polyclonal antibodies against Myc (R1208-1), OPTN (ER2001-02), and AIM2 (HA500270) were purchased from Huaan Biology Technology (Hangzhou, China). Rabbit mAb against phospho-UB (Ser65; 62802), Rabbit mAb against PLEKHM1 (77092), Rabbit mAb against VPS34 (4263), Rabbit mAb against NLRP3 (15101) were purchased from Cell Signaling Technology (Danvers, MA, USA). Alexa Fluor 647 dye (A-21236) was purchased from ThermoFisher (Waltham, MA, USA). Horseradish peroxidase (HRP)-labeled goat anti-mouse (5220-0341) or anti-rabbit (5220-0336) IgG was purchased from KPL (Milford, MA, USA). Dimethyl sulfoxide (DMSO; 472301) was purchased from Sigma-Aldrich. Rapamycin (HY-10219), wortmannin (HY-10197), NH₄CL (HY-Y1269), Bafilomycin A1 (HY-100558), chloroquine phosphate (1118000), C-176 (HY-112906), G150 (HY-128583) were purchased from MedChemExpress (NJ, USA). NP-40 cell lysis buffer (P0013F) and MG132 (S1748) were purchased from Beyotime (Shanghai, China). ChamQ Universal SYBR qPCR Master Mix (Q711-02) and ExFect^TM transfection reagent (517051) were purchased from Vazyme (Nanjing, China). PMSF (P0100) was purchased from Solarbio Biological Technology (Beijing, China). BsmBI (EG22507S) was purchased from Yugong Biotech (Lianyungang, China). The inactive sera of pigs naturally infected with African swine fever virus were kindly provided by Dr. Pei Zhou from Huazhong Agricultural University. Rhodamine 123 (C2008S, Beyotime), MitoSO™ Red (S0061S, Beyotime), ATP test kits (S0026, Beyotime).

Cell, virus, and animals

HEK293T (ATCC, CRL-3216), 3D4/21 (ATCC, CRL-2834), Vero E6 (ATCC, CRL-1586), N2a (ATCC, CCL-131), and RAW264.7 (ATCC, TIB-71) cells were cultured in Dulbecco’s modified Eagle medium (DMEM; Gibco, 12430054) supplemented with 10% fetal bovine serum (Gibco, 10099158). These cell lines were authenticated by PCR assays using species-specific cytochrome c oxidase I primers, followed by sequencing of the amplified products, to confirm species identity and exclude interspecies cross-contamination. WSL cells were kindly provided by Prof. Jun Han and Prof. Hanchun Yang, China Agricultural University, Beijing, China, and were not authenticated after receipt. Primary porcine alveolar macrophages (PAMs) were originally isolated in our laboratory and maintained for subsequent experiments. As primary cells, PAMs were not subject to cell line authentication. The CADC_HN09 were respectively reserved in the State Key Laboratory of Veterinary Biotechnology. The 6-week-old C57BL/6 J background mice were purchased from Shanghai Slac Co., Ltd. (Shanghai, China).

Virus isolates, propagation, purification, and titration

African swine fever virus (ASFV) strain CADC_HN09 reserved in the State Key Laboratory of Veterinary Biotechnology were used throughout this study. The parental virus and all recombinant viruses were propagated in primary porcine alveolar macrophages (PAMs) as indicated. Cells were infected at the specified multiplicity of infection (MOI) and incubated under standard culture conditions. At the indicated time points post infection, culture supernatants were collected and clarified by low-speed centrifugation to remove cell debris.

Viral stocks were either used directly as clarified supernatants or further concentrated by ultracentrifugation through a sucrose cushion when indicated. Virus titers were determined by plaque assay on WSL cells, as specified in the figure legends. Titers were calculated using standard plaque counting or the Reed–Muench method and expressed as PFU / mL or TCID₅₀ / mL, respectively. All infections were performed using viral stocks with comparable titers to ensure experimental consistency.

Plasmid construction

The B66L gene was amplified from the genome of the ASFV strain CADC_HN09. The amplified coding sequence was cloned into the pEGFP-C3 (Clontech, 6082), pEBFP-N1 (NovoPro, V006569), pCMV-Flag-N (Clontech, 635688), or pCMV-Myc-N (Clontech, 635689) vector, as indicated. Truncated B66L variants were generated by PCR amplification using pEGFP-B66L or pEBFP-B66L as the template with the following primers:

ASFV B66LΔCCD1-S: GATATAAAAAGAGCACTTTTAGTCGTATTGAGC

ASFV B66LΔCCD1-R: CTCAATACGACTAAAAGTGCTCTTTTTATATC

ASFV B66LΔCCD2-S: GTATTTTTCTTGTTATTCTTATTAATGAAAACA

ASFV B66LΔCCD2-R: GTTTTCATTAATAAGAATAACAAGAAAAATAC

For each gene, the double-deletion mutant (ΔCCD1 + 2) was generated by further deleting CCD2 based on the ΔCCD1 construct, using CCD2-specific primers and the ΔCCD1 plasmid as the template.

TOLLIP (NM_019009.4) was amplified from HEK293T cell cDNA and cloned into the pCMV-Flag-N vector. Truncated TOLLIP variants were generated by PCR amplification using Flag-TOLLIP as the template with the following primers:

TOLLIPΔLIR1-S: TGGACGACCGCATTGCCCCGGAGTCCCTGAGGCA

TOLLIPΔLIR1-R: TGCCTCAGGGACTCCGGGGCAATGCGGTCGTCCA

TOLLIPΔLIR2-S: CAAGGTGGAGGACAAGAGCGGGAGGCAGGGGGA

TOLLIPΔLIR2-R: TCCCCCTGCCTCCCGCTCTTGTCCTCCACCTTG

The TOLLIP double-deletion mutant (ΔLIR1 + 2) was generated by PCR amplification using LIR2-specific primers with the ΔLIR1 construct as the template.

RNA interference (RNAi)

The siRNAs targeting the indicated genes were designed based on the corresponding mouse transcript sequences. Target sequences were selected using the DSIR website, and sequence specificity was evaluated by BLAST against the mouse transcriptome. The siRNA sequences were as follows:

cGAS siRNA, sense 5′-GGAUUGAGCUACAAGAAUAUU-3′ and antisense 5′-UAUUCUUGUAGCUCAAUCCUG-3′;

TLR9 siRNA, sense 5′-AGCUCAACCUGUCCUUCAAUU-3′ and antisense 5′-UUGAAGGACAGGUUGAGCUUG-3′;

AIM2 siRNA, sense 5′-GGAAGGAAGACAAGAGAUAUU-3′ and antisense 5′-UAUCUCUUGUCUUCCUUCCUG-3′;

ZBP1 siRNA, sense 5′-GGACAUAGAAAGCUCUCAAGA-3′ and antisense 5′-UUGAGAGCUUUCUAUGUCCUG-3′;

NLRP3 siRNA, sense 5′-CACUCAUGAUUGACUUCAAUG-3′ and antisense 5′-UUGAAGUCAAUCAUGAGUGUG-3′.

Transfection was performed using Lipofectamine RNAiMAX (13778075, Thermo Fisher, MA, USA) at a final siRNA concentration of 20 nM according to the manufacturer’s protocol. Cells were collected at the indicated time points for downstream analyses.

Generation of knockout cell lines using CRISPR/Cas9 gene editing

Knockout (KO) cell lines, including single and multiple (up to tetra) knockouts, were generated using the CRISPR/Cas9 system. Guide RNAs (gRNAs) were designed to target a common exon shared by all known splicing variants of each gene. Synthetic oligonucleotides (Sigma) encoding the gRNA sequences were annealed and cloned into the lentiCRISPR v2 vector (Addgene plasmid #52961) according to the standard protocol. All gRNA inserts were sequence-verified prior to use.

For cell line generation, HEK293T, 3D4/21, Vero, and RAW264.7 cells were transfected with the verified CRISPR constructs using Lipofectamine 3000 (Thermo Fisher Scientific) and incubated for 24 h. Transfected cells were then subjected to puromycin selection (2 µg/mL for 48-72 h) to enrich for successfully transduced populations. Surviving cells were diluted and plated at low density to obtain single-cell clones. Individual clones were screened for loss of the target protein by immunoblotting. Genomic DNA was extracted from candidate KO clones, and the target loci were amplified by PCR and subjected to Sanger sequencing to confirm frameshift-inducing indels. The corresponding genomic regions from parental wild-type cells were sequenced in parallel as controls. Details of gRNA sequences, target sites, and genotyping primers are provided in Supplementary Tables 1, 2.

For generation of multiple knockout lines, sequential transfections or simultaneous co-transfections of multiple gRNA constructs were performed. Specifically, the TOLLIP single knockout (KO) was first generated in wild-type HEK293T cells. Subsequently, gRNAs targeting NDP52, SQSTM1/p62, and OPTN were simultaneously introduced into the TOLLIP-KO background to obtain TOLLIP/NDP52/SQSTM1/OPTN tetra-KO cells. All resulting clones were validated by both immunoblotting and sequencing to confirm the absence of the respective gene products.

Cell viability assay

PAM cells were seeded in 96-well plates and treated with G150 (0.25 or 2.5 μM), A438079 (2 or 20 μM), itaconate (2 or 20 mM), or BVT173187 (0.03 or 0.3 μM). Cell viability was measured using a Cell Counting Kit-8 (CCK-8) assay following the manufacturer’s instructions. Absorbance at 450 nm was recorded with a microplate reader, and viability was calculated relative to the untreated control group. Data were obtained from independent biological experiments with technical replicates averaged within each biological replicate.

Mitochondria isolation from extracellular milieu

The detailed protocol used in this study was adapted from Dr. Frezza’s report³⁷. Briefly, to isolate mitochondria released into the extracellular milieu, porcine alveolar macrophages (PAMs) were infected with ASFV, and culture supernatants were collected at the indicated time points. The supernatants were first cleared of intact cells and large debris by centrifugation at 200 × g for 10 min, followed by a second centrifugation at 3000 × g for 15 min. The clarified supernatants were then subjected to differential centrifugation at 10,000 × g for 30 min at 4 °C to pellet large extracellular vesicles and mitochondria. The resulting pellet, referred to as the crude extracellular mitochondrial fraction, was washed once with ice-cold PBS and immediately used for downstream assays, including immunoblotting, mitochondrial membrane potential measurement, and functional analyses.

Recombinant adenovirus production, concentration and purification

Recombinant adenovirus was generated following the protocol outlined in the Adenovirus User Manual-Genemedi. Briefly, ASFV B66L was cloned into pDC316-mCMV-EGFP (YRgene, VXY0585). The constructs were co-transfected into HEK293A cells with pBHGlox (delta) E1, 3Cre (Biovector NTCC, Inc, BioVector 912752). Virus was collected through three freeze-thaw cycles and then used to infect fresh HEK293A cells. After five rounds of expansion, the titration of purified virus was determined through a multiplicity of infection (MOI) test.

Animals and mouse breeding and infection

The Atg5-KO mice were kindly provided by Qiming Sun (Department of Biochemistry and Department of Cardiology of Second Affiliated Hospital, Zhejiang University School of Medicine). The Sting1-KO (Strain NO. T012747), Cgas-KO (Strain NO. T003215), and Tollip-KO (Strain NO. T012006) mice on a C57BL/6 J genetic background were purchased from Jiangsu GemPharmatech (Nanjing, China), a national resource center for genetic engineering mice (Supplementary Table 3). All mice were on a C57BL/6 J genetic background. Six-week-old mice (female) were maintained in a specific pathogen-free barrier facility under a 12 h light/12 h dark cycle at an ambient temperature of 20-22 °C and relative humidity of 50-60%. All animal experiments were conducted in accordance with protocols approved by the Institutional Animal Care and Use Committee (IACUC) of Zhejiang University (approval nos. ZJU20220495, ZJU20250834, and ZJU20260078).

Six-week-old female C57BL/6 J mice (n = 15 mice per group) were intranasally exposed to 10^9.0 PFU of recombinant adenovirus (rAdv and rAdv-B66L). On day 4 post-infection, serum samples were collected to observe secreted mitochondria and detect inflammatory factors. Subsequently, the mice were euthanized for the assessment of lesions, inflammatory factors, and tissue transmission electron microscopy (TEM) observation in the lungs. For histopathological analysis, the lungs were excised, fixed with 4% PFA, embedded in paraffin, sectioned, and stained with hematoxylin and eosin solution (H.E). Additionally, on day 4 post-infection, the lungs were fixed with 2.5% glutaraldehyde at 4 °C, embedded in resin, sectioned, and stained with uranyl acetate and alkaline lead citrate for 5 to 10 min, respectively, for TEM observation. The mice were housed in specific-pathogen-free conditions, maintaining a temperature range of 20-22 °C, humidity at 50%-60%, and with ad libitum access to food and water.

Isolation of secreted mitochondria

Cultured cells expressing 6xHis-tagged TOMM70 were cultivated for a duration of 48 h. Subsequently, the cultured medium was harvested and subjected to centrifugation at 200 × g for 10 min and 3000 × g for 10 min to pelletize both floating cells and debris. The resulting supernatant, containing secreted mitochondria, was then aseptically transferred to a fresh tube and further centrifuged at 10,000 × g for 30 min.

Transmission electron microscopy (TEM) and quantification of mitochondrial damage

Mitochondria were isolated from the culture supernatants of empty vector (EV)-transfected or B66L-expressing cells by differential centrifugation. For each condition, five independent cell cultures were prepared, and mitochondria were purified from their supernatants. Ultrathin sections were examined using a Hitachi H-7650 transmission electron microscope (Hitachi, Tokyo, Japan), and images were acquired using Digital Micrograph software (v1.71.38). Three randomly selected fields were acquired per preparation. Mitochondria were categorized as “healthy” if they exhibited intact double membranes and well-organized cristae, or as “damaged” if they displayed fragmented/swollen morphology, disrupted cristae, or vacuole-like structures. The percentage of healthy or damaged mitochondria was calculated from five independent biological replicates per condition

Preparation of cell supernatants for mass spectrometry

Supernatants from mock-treated or African swine fever virus (ASFV)-infected primary porcine alveolar macrophages (PAMs) at 36 h post infection were collected and clarified by centrifugation to remove cell debris. Proteins in the supernatants were prepared for liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis, including protein extraction, enzymatic digestion and peptide desalting. Three biologically independent samples were analyzed for each group. The resulting peptide samples were submitted to Applied Protein Technology (Shanghai, China) for LC-MS/MS identification.

Gene ontology (GO) enrichment analysis

Differentially abundant proteins from the culture supernatant of ASFV-infected primary porcine alveolar macrophages (PAMs) were identified based on a fold change ≥ 1.5 and p < 0.05, which were set as the significance thresholds. Gene Ontology (GO) annotation was conducted using the Gene Ontology Consortium database via the DAVID platform. Enriched GO terms in the Cellular Component (CC) category were determined using Fisher’s exact test, and p-values were corrected for multiple testing with the Benjamini-Hochberg false discovery rate (FDR) method. GO terms with an adjusted p-value < 0.05 were considered significantly enriched. The top 10 enriched CC terms were visualized by ranking the -log10(p-value) values, which were plotted using the ggplot2 package.

Proteinase K (PK) protection assay

The crude mitochondrial suspension was divided into equal aliquots of 200 µl and centrifuged at 10,000 × g for 30 min at 4 °C. The mitochondrial pellets underwent the following treatments: (1) a pellet was re-suspended in 500 µl of Solution A (250 mM sucrose, 5 mM NaN3, 2 mM EGTA, 20 mM HEPES-Na, pH 7.4) and incubated on ice for 10 min (untreated mitochondria); (2) a pellet was re-suspended in 500 µl of Solution A containing 20 µg/ml of PK and incubated on ice for 10 min (surface-exposed OMM proteins’ digestion); (3) a pellet was re-suspended and incubated with 400 µl of 1% (v/v) TritonX-100 in Solution A on ice for 20 min and then incubated with 500 µl of Solution A containing 20 µg/ml of PK on ice for 10 min (all mitochondrial proteins’ digestion).

RT-qPCR

Total RNA was isolated with RNAiso Plus (9109; Takara, Shiga, Japan) according to the manufacturer’s instructions. DNase I (M0303, NEB, New England Biolabs, USA) was used to remove DNA. The RevertAid RT reverse transcription kit (K1691, Thermo Fisher, MA, USA) was employed for reverse transcription according to the manufacturer’s instructions. The relative abundance of transcripts was measured using ChamQ Universal SYBR qPCR Master Mix (Q711-02; Vazyme, Nanjing, China), the LightCycler 96 sequence detector system (Roche), and the primer in follow:

B66L-S: ATGGATATAAAAAGAGCAC

B66L-R: AAGAATGTTTTCATTAATAAG

Swine IL1B-S: ATTCAGGGACCCTACCCTCTC

Swine IL1B-R: ATCACTTCCTTGGCGGGTTC

Swine IL6-S: ACAAAGCCACCACCCCTAAC

Swine IL6-R: CGTGGACGGCATCAATCTCA

Swine TNFA-S: CGTCGCCCACGTTGTAGCCAAT

Swine TNFA-R: GCCCATCTGTCGGCACCACC

Swine ACTIN-S: TGGCGCCCAGCACGATGAAG

Swine ACTIN-R: GATGGAGGGGCCGGACTCGT

Co-immunoprecipitation (Co-IP)

The HEK293T cells were co-transfected with the indicated recombinant plasmids for 48 h. Subsequently, the cells were collected and lysed using NP40 buffer. Cellar lysates were incubated with anti-Flag mAb and protein A/G beads for 4 h at 4°C. After centrifugation, the supernatant was removed, and the pellets were resuspension in washing buffer. Centrifugation and resuspension of the pellets in fresh washing buffer were repeated five times. Finally, the pellets were lysed in lysis buffer for Western blot analysis.

Immunofluorescence assay and confocal microscopy

Cells were plated on glass-bottom cell culture plates, were transfected with the indicated plasmids, and treated with the indicated drugs, or cultured in medium. After 24 h or 36 h, the cells were fixed with 4% paraformaldehyde in PBS for 15 min at room temperature, and permeabilized with 0.1% TritonX-100 for 10 min. The permeabilized cells were blocked with 5% nonfat milk for 1 h. After washing three times with PBST, the cells were incubated with the indicated antibodies for 2 h at 37 °C. The cells were washed with PBS and then incubated with Alexa Fluor 647 dye (ThermoFisher, A-21236) for 1 h. Fluorescence images were acquired using an LSM780 laser-scanning confocal microscope (Zeiss, Oberkochen, Germany) with ZEN software (v3.9).

RFP-GFP-mito mitophagy assay

3D4/21 cells were co-transfected with an empty vector or a B66L-expressing plasmid together with the tandem RFP-GFP-Mito reporter. Twenty-four hours post-transfection, mitophagy was induced with carbonyl cyanide m-chlorophenyl hydrazone (CCCP, 10 µM, 4 h) or vehicle (DMSO). Cells were analyzed by laser-scanning confocal microscopy using identical acquisition settings across conditions. Under basal conditions, mitochondria appear yellow (GFP + /RFP + ) due to colocalized signals. Upon mitophagy induction, delivery of mitochondria to acidic autolysosomes quenches GFP while RFP remains stable, producing red-only puncta that mark mitolysosomes. For quantification, at least 20 randomly selected cells per group from three independent experiments were analyzed. After background subtraction and consistent thresholding in Fiji (ImageJ), the “Analyze Particles” function was applied to enumerate yellow (GFP + RFP) and red-only (RFP) puncta per cell. Data are expressed as mean ± SD, and statistical significance was determined using two-tailed Student’s t-tests.

Histological Analysis and Infiltration Index

Wild-type and gene knockout mice were intranasally infected with recombinant adenovirus expressing B66L (Ad-B66L) or control adenovirus. Lungs were collected, fixed in 4% paraformaldehyde, paraffin-embedded, sectioned (5 μm), and stained with H&E. Images were acquired by bright-field microscopy. The infiltration index, defined as the percentage of infiltrated area relative to alveolar area, was calculated by converting RGB images to greyscale, generating image pyramids (1, 2, 4, 8×), and applying thresholds to detect low-intensity pixels across all levels. Infiltrated regions were dilated with an 8-pixel disk operator. Alveolar area was segmented by excluding blood vessels and bronchioles ( > 100 pixels). The ratio was used as a quantitative index of lung damage and repair.

Construction of inducible B66L recombinant ASFV

An inducible recombinant ASFV expressing B66L was designed following the lac operator/repressor strategy previously established for ASFV essential genes. Briefly, the B66L coding sequence was placed under the control of a synthetic late ASFV promoter derived from p72 and containing an Escherichia coli lac operator sequence, allowing transcriptional repression by the LacI repressor and inducible activation upon addition of isopropyl-β-D-thiogalactopyranoside (IPTG). The lacI gene was expressed from an ASFV early/late promoter (pU104L) to ensure constitutive production of the repressor throughout the viral replication cycle.

The inducible cassette, together with a reporter gene (EGFP) for selection, was inserted into the ASFV genome by homologous recombination using left and right flanking regions corresponding to the target genomic locus. Recombinant viruses were generated by transfecting permissive cells with the linearized transfer plasmid followed by infection with parental ASFV in the presence of IPTG. Recombinant virus populations were isolated by multiple rounds of plaque purification under inducing conditions. The genomic structure of the recombinant virus was verified by PCR analysis and sequencing to confirm correct insertion and integrity of the inducible cassette.

Plaque assay

Plaque assays were performed using confluent Vero cell monolayers cultured in six-well plates. Cells were infected with the recombinant inducible B66L ASFV or the parental CADC_HN09 strain as a reference control. Following a 1 h adsorption period at 37 °C, the viral inoculum was removed and replaced with DMEM supplemented with 2% fetal bovine serum and 0.6% Noble agar, either containing or lacking 1 mM isopropyl-β-D-thiogalactopyranoside (IPTG). After incubation for 5-7 days, cells were fixed with formaldehyde, the agar overlay was carefully removed, and plaques were visualized by staining the monolayers with 1% crystal violet.

Multiple-step growth curve analysis

To determine replication kinetics, Vero cells grown in 24-well plates were infected with the recombinant inducible B66L virus or parental CADC_HN09 at a multiplicity of infection (MOI) of 5 PFU per cell. After virus adsorption for 1 h, cultures were maintained in DMEM containing 2% fetal bovine serum, in the presence or absence of 1 mM IPTG. At the indicated time points postinfection, both cells and supernatants were collected, subjected to sonication to release intracellular virus, and viral titers were quantified by plaque assay on Vero cells under inducing conditions (1 mM IPTG).

Time-lapse confocal imaging and quantification of extracellular mitochondria

Time-lapse imaging was performed using an LSM 900 fast ultra-high-resolution laser-scanning confocal microscope with Airyscan 2 detection (Zeiss, Oberkochen, Germany) controlled by ZEN software (v3.9). WSL cells stably expressing DsRed-Mito were either mock-treated with IPTG, infected with the inducible recombinant ASFV vB66Li in the absence or presence of IPTG, or transfected with GFP empty vector or GFP-B66L as indicated. Live-cell imaging was conducted under physiological conditions (37 °C, 5% CO₂).

For each condition, five independent cells from five independent biological experiments were recorded and analyzed. For each selected cell, three randomly chosen fields of view containing clearly defined plasma membrane boundaries were analyzed. Extracellular mitochondria were defined as discrete DsRed-Mito-positive structures located outside the visually delineated cell boundary. The number of extracellular mitochondria per field was manually quantified using ImageJ (NIH) and averaged per cell before statistical analysis.

Quantitative data are presented as mean ± SD. Statistical comparisons between groups were performed using appropriate parametric tests as indicated in the figure legends. Scale bar, 2 μm.

Transmissible electron microscopy

The transfected cells were scraped from the plates or the lungs of Recombinant adenovirus infected mice fixed with 2.5% glutaraldehyde in phosphate buffer (0.1 M, pH 7.0) for 4 h, and then postfixed with 1% OsO4 in phosphate buffer for 1-2 h. The fixed cells were washed three times in the phosphate buffer (0.1 M, pH 7.0) for 15 min and subsequently dehydrated in a gradient ethanol. They were then placed in Eppendorf containing Spurr resin and heated at 70˚C for more than 9 h. The embedded specimen was sectioned using a LEICA EM UC7 ultratome. The sections were stained with uranyl acetate and alkaline lead citrate for 5 to 10 min respectively, and observed using a Hitachi Model H-7650 TEM.

ELISA

Use mouse- and pig-derived IL-1β, IL-6, and TNF-α ELISA kits to detect inflammatory factors in mouse serum and bronchoalveolar lavage fluid and pig serum according to the instructions.

Transwell assay

To assess the role of mitochondrial transfer in triggering inflammation in recipient cells, a Transwell co-culture system was employed. RAW246.7 (1 × 10⁵ cells/well) were seeded in the lower chamber, while Vero cells (1 × 10⁵ cells/well) transfected with either an empty vector or a construct encoding B66L were placed in the upper insert. Following co-culture, the medium from RAW246.7 was collected and analyzed by ELISA to quantify the levels of pro-inflammatory cytokines, including IL-1β, IL-6, and TNF-α.

Statistics and reproducibility

Image analysis was performed using ImageJ (v2.3.0). Statistical analyses were performed using GraphPad Prism (v10.4). Differences between two groups were assessed using unpaired two-sided Student’s t-tests. Comparisons among multiple groups were analyzed using one-way or two-way ANOVA followed by Dunnett’s or Tukey’s multiple-comparison test, as specified in the figure legends. Cell-based experiments were performed on separate days using independently prepared cell cultures, and each independent experiment was considered one biological replicate. For animal experiments, one mouse was considered one biological replicate. Exact P values are provided where possible in the figures or Source Data 1 and Source Data 2 files. Data are presented as mean ± SD (*, P  <  0.05; **, P  <  0.01; ns, p > 0.05).

Sample sizes are indicated in the relevant figure legends and were chosen based on our previous experience, preliminary data, and common practice in the field. No statistical method was used to predetermine sample size. No data were excluded from the analyses. For animal studies, mice were randomly assigned to experimental groups. Histopathological scoring, immunofluorescence quantification, and TEM-based mitochondrial classification were performed in a blinded manner. Representative H&E, immunofluorescence, electron microscopy, and immunoblot images are from independent experiments with similar results. Unless otherwise stated, all experiments were independently performed at least three times.

Reporting summary

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

Author contributions

B.H. and L.L. conceived the project, designed the study, and wrote the manuscript. B.H. and J.Z. supervised the study, administered the project, and acquired funding. L.L., Z.H., and G.X. led the cell-based, animal, tissue, and molecular experiments performed in vivo and in vitro, and analyzed the data. L.L., Z.H., G.X., F.W., J.B., R.F., Y.Z., and Y.L. performed the experiments. J.W. and J.L. contributed to discussion of the results and provided technical and resource support. B.H. reviewed the manuscript for intellectual content. L.L. and B.H. revised the manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Data availability

The protein mass spectrometry raw data generated in this study have been deposited in the ProteomeXchange Consortium via the iProX partner repository under accession code PXD069379 (https://www.iprox.cn//page/project.html?id=IPX0013782000). Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-73537-8.