Activation of NIX-mediated mitophagy and replication by an interferon regulatory factor homologue of human herpesvirus

Viral control of mitochondrial quality and content has emerged as an important mechanism for counteracting the host response to virus infection. Despite the knowledge of this crucial function of some viruses, little is known about how herpesviruses regulate mitochondrial homeostasis during infection. Human herpesvirus 8 (HHV-8) is an oncogenic virus causally related to AIDS-associated malignancies. Here, we found that HHV-8-encoded viral interferon regulatory factor 1 (vIRF-1) promotes mitochondrial clearance by activating mitophagy to support virus replication. Genetic interference with vIRF-1 expression or targeting to the mitochondria inhibits HHV-8 replication-induced mitophagy and leads to an accumulation of mitochondria. Moreover, vIRF-1 binds directly to a mitophagy receptor, NIX, on the mitochondria and activates NIX-mediated mitophagy to promote mitochondrial clearance. Genetic and pharmacological interruption of vIRF-1/NIX-activated mitophagy inhibits HHV-8 productive replication. In short, our findings uncover an essential role of vIRF-1 in mitophagy activation for successful lytic replication of HHV-8.


Introduction
Mitochondria are vital energy-generating organelles in eukaryotic cells. They also integrate signals controlling multiple cellular functions such as apoptosis, cell cycle, and development. In addition, recent studies have shown that mitochondria serve as a platform for mediating antiviral signaling pathways that lead to apoptosis and innate immune responses to virus infection 1,2 . For example, proapoptotic BH3-only proteins (BOPs) are elevated 3 and/or activated during virus replication and induce mitochondrial outer membrane permeabilization, a crucial step in the intrinsic apoptotic process that triggers the release of soluble apoptogenic factors from the intermembrane space. In the innate immune response to virus infection, the RIG-I-like receptors (RLRs) RIG-I and MDA-5 recognize cytosolic viral RNA and promote the oligomerization of the mitochondrial antiviral signaling adaptor protein (MAVS; also known as IPS-1, VISA, and Cardif) 4 , which recruits TBK1 and IKKi kinases to activate IRF3 and IRF7 transcription factors 2 .
These activated IRFs induce the expression of type I interferon (IFN) genes, the products of which restrict virus replication. Therefore, successful virus infection and replication are in large part achieved by the ability of viruses to attenuate the innate antiviral responses mediated by mitochondria.
Particular viral proteins have evolved to enable virus infection and replication by modulating the mitochondria-mediated antiviral responses. For example, human herpesviruses encode anti-apoptotic proteins that inhibit the intrinsic apoptosis pathway 1 , and hepatitis C virus encodes a serine protease NS3/NS4A that disrupts RLR signaling and IFN-b production by cleaving MAVS 5 . On the other hand, mitochondria generate reactive oxygen species (ROS) as a byproduct of electron transport complexes. Accumulation of damaged or dysfunctional mitochondria leads to aberrant ROS generation, which can exacerbate apoptosis and augment RLR-MAVS signaling 6,7 . Several viruses are known to effect elimination of infection-altered mitochondria via mitophagy, potentially attenuating apoptosis and innate immune responses 8 .
Mitophagy is a key mechanism of mitochondrial quality control that eliminates aged, dysfunctional, damaged or excessive mitochondria via selective autophagy 9, 10 . A recent study indicates that impaired removal of damaged mitochondria by mitophagy can lead to STINGdependent inflammation and neurodegeneration 11 . PINK1/PARK2-mediated mitophagy is one of the most well-characterized; after the loss of mitochondrial membrane potential, PINK1 is stabilized on the outer mitochondrial membrane (OMM), where it phosphorylates both ubiquitin (Ub) and the Ub-like domain of PARK2, leading to its activation [12][13][14] . In turn, activated PARK2 ubiquitinates mitochondrial proteins such as MFN1/2, DRP1, BCL-2, and VDAC1. Mitophagy receptors such as NDP52 and OPTN bind to and link the ubiquitinated mitochondria to autophagosomes via interaction with autophagy-related protein 8 (ATG8) family members such as LC3 and GABARAP 15 . PARK2-independent mitophagy pathways mediated by other mitophagy receptors such as NIX/BNIP3L, BNIP3, FUNDC1, and PHB-2 have been identified under certain physiological conditions [16][17][18][19] . Nonetheless, viral regulation of mitophagy is an understudied area of investigation.  22,23 . Furthermore, we found that vIRF-1 localizes to mitochondria upon virus replication and suppresses mitochondria-mediated apoptosis and innate immune responses via its inhibitory interactions with proapoptotic BOPs and MAVS 3,24,25 . However, the precise role of mitochondria-localized vIRF-1 in HHV-8 biology remains elusive. In this study, we identify a new function of vIRF-1 in its activation of mitophagy and support of virus replication via this pathway.

vIRF-1 downregulates mitochondrial content during lytic reactivation
We previously reported that the level of MAVS is significantly diminished in the HHV-8infected PEL cell line, BCBL-1, that expresses vIRF-1 following lytic reactivation 25 . Another mitochondrial protein TOM20 was also strongly reduced in the mitochondria of vIRF-1-expressing lytic cells compared to that in vIRF-1-negative neighbor cells ( Supplementary Fig. 1a). Moreover, MitoTracker Red staining of mitochondria showed an apparent decrease in the levels of mitochondria of vIRF-1-expressing lytic cells ( Supplementary Fig. 1b), implying that the reduced levels of MAVS and TOM20 might result from a loss of mitochondrial content rather than a specific inhibition of the expression or stability of the proteins. To further address this notion, we examined the levels of mitochondrial DNA (mtDNA) using a combined approach of immunofluorescence assay (IFA) and fluorescent in situ hybridization (FISH). Indeed, the result showed that mtDNA was less readily detected in vIRF-1-expressing lytic cells (Fig. 1a).
Furthermore, immunoblotting and image cytometry analyses showed that the expression of MTCO2 (mtDNA-encoded cytochrome c oxidase II) was significantly reduced in lytic TRExBCBL-1-RTA (hereafter simply termed iBCBL-1) cells, doxycycline (Dox)-inducible for lytic switch protein RTA 26 , that were treated with Dox for 2 days (Figs. 1b and 1c). These results suggest that mitochondria content of cells could be downregulated, conceivably by vIRF-1, following lytic reactivation of HHV-8.
To next examine whether vIRF-1 is indeed involved in the regulation of mitochondria content, we generated a cell line that stably expresses a short hairpin RNA (shRNA) directed against vIRF-1 in iBCBL-1 cells along with a control cell line expressing shRNA directed against luciferase (Luc). Constitutive knockdown of vIRF-1 was previously shown to induce death of latently HHV-8-infected PEL cell lines 27 . To circumvent this limitation, we used an inducible cytometry analysis showed that, when vIRF-1 was depleted, lytic cells with higher levels of MTCO2 were increased in number (Fig. 1e). Taken together, the results suggest that vIRF-1 plays an important role in the downregulation of mitochondria content during virus replication.

vIRF-1 controls mitochondria content via mitophagy
The cellular content of mitochondria is fundamentally controlled by the balance of anabolic biogenesis and catabolic degradation such as selective autophagy of mitochondria (termed mitophagy). Mitochondrial biogenesis is a complex process requiring the coordinated expression of nuclear and mitochondrial DNAs. Nucleus-encoded mitochondrial transcription factor A (TFAM) is essential for transcription and replication of mtDNA, which are important steps in the mitochondrial biogenesis 28 . Thus, we examined the transcriptional levels of TFAM but observed no significant difference between control and vIRF-1-depleted iBCBL-1 cells that were left untreated or treated with Dox ( Fig. 1f), suggesting that virus replication and vIRF-1 might not influence mitochondrial biogenesis. Therefore, we extrapolated that mitophagy may be involved in vIRF-1 regulation of mitochondrial content during lytic replication. To test this notion, we first examined whether HHV-8 activates mitophagy following lytic reactivation. The results showed that autophagy inhibitors including bafilomycin A1 (Baf A1) and leupeptin blocked a decrease in the MTCO2 levels in Dox-treated (lytically-reactivated) iBCBL-1 cultures (Supplementary Figs. 1c and 1d). Consistent with this, immunoblotting analysis showed that a decrease of MTCO2 protein was inhibited by Baf A1 and chloroquine (CQ), another autophagy inhibitor, but not by a proteasome inhibitor, MG132, in Dox-induced iBCBL-1 cultures ( Supplementary Fig. 1e). We further examined the formation of mitochondria-containing autolysosomes (hereafter simply named mitolysosomes), an end-point readout of mitophagy 29 , using CellLight TM BacMam-labeling of mitochondria and lysosomes (see Material and Methods for details). The results showed that the presence of mitolysosomes was more evident in lytic control iBCBL-1 cells than latent control and both latent and lytic vIRF-1-depleted iBCBL-1 cells (Fig. 1g). Furthermore, electron microscopy imaging demonstrated the apparent presence of mitolysosomes in lytic control cells but not in lytic vIRF-1-depleted iBCBL-1 cells (Fig. 1h). It is noteworthy that mitochondria with disrupted cristae were often observed outside the autophagic vacuoles of lytic vIRF-1-depleted iBCBL-1 cells (Fig. 1h, red arrows). Taken together, our results suggest that vIRF-1 is likely to be involved in activation of mitophagy to control mitochondrial content of cells during virus replication.

vIRF-1 activates NIX-mediated mitophagy
Mitophagy is triggered by activation of specific autophagy receptors localized mainly on the outer mitochondrial membrane (OMM); these proteins interact with ATG8 family members including LC3 and GABARAP via a short-linear motif termed the ATG8-interacting motif (AIM) or LC3-interacting region (LIR), which forms a bridge linking the mitochondria to the autophagosomes 10 . Thus, we hypothesized that vIRF-1 may promote mitophagy by recruiting the mitophagy machinery and/or activating them on the mitochondria. Firstly, we investigated changes in the levels of the mitophagy machinery including mitophagy receptors and LC3 on the mitochondria isolated from latent and lytic iBCBL-1 cells. Consistent with our previous report 25 , vIRF-1 was readily detected in the mitochondrial fraction isolated from lytic iBCBL-1 cells (Fig.   2a). When autophagy is induced, LC3 is processed from a cytosolic form, LC3-I (18 kDa), to the LC3-II (16 kDa) form that is lipidated with phosphatidylethanolamine and associated with the autophagic vesicle membranes. Intriguingly, the LC3B-II form, but not the LC3B-I form, was readily detected in the mitochondria and here exhibited a more than 2-fold increase upon virus replication while total LC3B levels remained unchanged after lytic reactivation (Fig. 2a), indicating that mitophagy may be selectively induced during virus replication. We next examined the levels of cellular mitophagy receptors including NIX (also termed BNIP3L), OPTN, NDP52, p62, NBR1, and FUNDC1. Interestingly, the level of mitochondria-associated NIX was upregulated by more than 2-fold while the other receptors remained essentially unchanged (Fig.   2a). The mitochondrial fission protein DRP1 and the mitochondrial chaperone HSP60 were used as a loading control (Fig. 2a). NIX mRNA expression was not induced by lytic reactivation (Fig.   2b), indicating that NIX protein may be stabilized and/or translocated to the mitochondria during HHV-8 replication. We failed to detect expression of another mitophagy receptor BNIP3, a paralog of NIX, in both latent and lytic iBCBL-1 cells. At present we cannot rule out roles of other mitophagy receptors other than NIX in virus replication-induced mitophagy activation.
Nonetheless, based on our finding of NIX regulation by lytic induction, we investigated further the function of NIX for virus replication-and vIRF-1-induced mitophagy activation in subsequent studies.
We next examined whether vIRF-1 is required for the recruitment of LC3B-II and NIX to the mitochondria. Immunoblotting analyses showed that virus replication-induced translocation of LC3B-II to the mitochondria was abated when vIRF-1 was depleted (Fig. 2c). Similarly, virus replication-promoted mitochondrial localization of NIX and formation of its shifted bands to about 100-kDa, which may represent a homodimer of NIX, were significantly reduced in lytic vIRF-1deficient iBCBL-1 cells (Fig. 2d). Whether NIX dimerization is required for mitophagy activation remains unclear; however, hypoxia-induced dimerization of a paralog of NIX, BNIP3, was proposed to be related to mitophagy activation in vivo 18 . Therefore, we hypothesized that vIRF-1 may activate NIX-mediated mitophagy via promotion of NIX recruitment to and dimerization on mitochondria. To test this possibility, we transiently transfected HeLa.Kyoto cells with Flagtagged vIRF-1 (vIRF-1-Flag) and/or V5-tagged NIX (V5-NIX). Contrary to our expectation, there was no effect of vIRF-1 on promotion of V5-NIX recruitment to and dimerization on mitochondria ( Supplementary Fig. 2). To examine whether NIX is involved functionally in vIRF-1 regulation of mitochondria content, we stably transduced control and vIRF-1-depleted iBCBL-1 cells with Doxinducible NIX shRNA (sh-NIX). Dox-inducible depletion of endogenous NIX and vIRF-1 proteins were verified using immunoblotting in each cell line (Fig. 2e). Image cytometric analysis showed that depletion of NIX alone had a minor effect on the mitochondrial content, but double depletion of NIX and vIRF-1 significantly increased mitochondrial content in lytic cells to an even greater extent than depletion of vIRF-1 (Fig. 2f), indicative of a mechanistic link of vIRF-1 and NIX. This functional interaction was further assessed by transient transfection experiments; surprisingly, overexpression of both vIRF-1-Flag and V5-NIX, but not each individually, could induce a significant decrease in mitochondrial content, as determined by TFAM immunostaining (Fig. 2g).
To rule out a role of nucleus-localized vIRF-1 in the mitochondrial clearance, we used the nuclear localization signal (NLS)-mutated version of vIRF-1 (NLS X ) as previously described 24,25 . The vIRF-1 NLS X variant strongly induced mitochondrial clearance together with V5-NIX (Fig. 2g).
However, the vIRF-1 mutant (DPD) lacking the proline-rich domain (PD) that contains an atypical mitochondrial targeting signal sequence (MTS) 25 could not induce mitochondrial clearance (Fig.   2g). These results suggest that mitochondrial localization of vIRF-1 is required for mitochondrial clearance. Since overexpression of NIX alone did not induce mitochondrial clearance (Fig. 2g), NIX may require an activation step to function as a mitophagy receptor. For example, it is known that in response to high oxidative phosphorylation activity, the small GTPase RHEB (Ras homolog enriched in brain) is recruited to the mitochondria and interacts with NIX to induce mitophagy 30 .

vIRF-1 binds directly to mitochondria-associated NIX
To examine whether vIRF-1 interacts with NIX on the mitochondria, we performed a coimmunoprecipitation assay with the mitochondrial fraction isolated from lytic iBCBL-1 cells. The result showed that vIRF-1 was indeed co-precipitated with NIX (Fig. 3a). Next, we employed a proximity ligation assay (PLA Duolink, Sigma) to analyze in situ interaction between vIRF-1 and NIX; their interaction was detected in the cytoplasm, potentially in mitochondria, of HeLa.Tokyo cells co-transfected with vIRF-1 and NIX, but not with each alone (Fig. 3b). To examine whether vIRF-1 binds directly to NIX, we performed an in vitro GST pull-down assay using purified recombinant proteins. The result showed that recombinant vIRF-1 protein, which was purified from bacteria as described previously 25 , was co-precipitated with purified GST-NIX but not GST protein (Fig. 3c).
To further characterize this interaction, we next sought to identify the vIRF-1 region(s) responsible for NIX binding. We used an in vitro binding assay using purified GST-NIX and cell lysates of 293T cells transfected with vIRF-1 full-length (FL) or a series of deletion mutants in which segments of 23 to 25 residues were deleted (Fig. 3d). Deletion of the first segment encompassing residues 1 to 23, comprising part of the PD region, completely abrogated NIX binding of vIRF-1 (Fig. 3d). Subsequent mapping and mutagenesis studies were carried out to identify vIRF-1 residues sufficient and required for NIX binding (Fig. 3d). These experiments identified two residues of vIRF-1, asparagine 8 (N8) and phenylalanine 10 (F10), as essential for NIX binding (Fig. 3d). Furthermore, the F10A mutation led to a loss of the ability of vIRF-1 to induce mitochondrial clearance in the presence of NIX (Fig. 3e), showing that vIRF-1 interaction with NIX is required to activate NIX-mediated mitophagy. Collectively, our results indicate that the N-terminal PD region of vIRF-1 contains motifs for vIRF-1 binding to and activating of NIXmediated mitophagy and for mitochondrial targeting 25 .

NIX determinants of physical and functional interactions with vIRF-1
To investigate the intracellular interaction of vIRF-1 and NIX, we employed a luciferasebased protein fragment complementation assay, NanoLuc® Binary Technology (NanoBiT, Promega). The NanoBiT system consists of two small subunits, Large BiT (LgB; 17.6 kDa) and Small BiT (SmB; 1.3 kDa) (Fig. 4a). To find an optimal orientation of the NanoBiT tags for detection of vIRF-1:NIX interaction, Flag-vIRF-1 and V5-NIX were tagged at either the N-or Cterminal end of the NanoBiT subunits ( Supplementary Fig. 4a). Intriguingly, when the NanoBiT tags were fused to the C-terminal end of NIX, the formation of a NIX dimer was greatly reduced ( Supplementary Fig. 4b); this is likely due to its steric hindrance in the C-terminal tail anchor (TA) region, thus interfering with targeting to the OMM and/or TA region-mediated dimerization 31 . At 24 h after co-transfection with the different vIRF-1 and NIX plasmid combinations or single plasmids, a Nano luciferase (NLuc) substrate, furimazine, diluted in phosphate-buffered saline was added to the cell cultures. We identified two binary combinations showing the highest NLuc activity: 1) vIRF-1-SmB plus LgB-NIX and 2) SmB-vIRF-1 plus LgB-NIX ( Supplementary Fig.   4c). To further demonstrate their specific interactions, we generated various deletion and substitution variants together with a negative control, HaloTag (HT)-SmB (Fig. 4b). NIX indeed bound specifically to vIRF-1; the luminescence signal was 10-fold higher when LgB-NIX was coexpressed with vIRF-1-SmB than HT-SmB (Fig. 4c). However, NIX binding to vIRF-1 was significantly diminished when the PD of vIRF-1 was deleted (Fig. 4d). These results are consistent with the data from the precipitation assays above and validated the NanoBiT system for further quantitative analysis of vIRF-1 and NIX interaction. We next examined whether the mitochondrial targeting of NIX is required for binding to vIRF-1. For this assay, we generated the NIX mutant (NIXDTA) that is deficient in the tail anchor (TA) region (residues 188 to 208) and showed that this mutant lost the ability to localize to mitochondria and to form a homodimer (Supplementary Unlike NIX, however, immunoblotting analysis showed that the chimeric NIX proteins did not form homodimers (Fig. 4f). Furthermore, NIX-TA VAMP1B did not induce mitochondrial clearance together with vIRF-1 (Fig. 4g). These results indicate that NIX dimerization via its own TA region is essential for mitophagy activation, and the interaction of vIRF-1 and NIX is not sufficient for triggering mitophagy without NIX dimerization. Thus, we wondered how vIRF-1 activated NIX-mediated mitophagy. ATG8 binding to the LIR motif of NIX is critical in the NIXmediated mitophagy pathway 17 . We thus generated the NIX mutant (NIXDLIR) deficient in the LIR motif (residues 35 to 39) and found that vIRF-1 did not induce mitochondrial clearance together with NIXDLIR (Fig. 4g). Accordingly, we postulated that vIRF-1 may promote the interaction of NIX with ATG8 family members via the LIR. To test this, we employed a co-IP assay using HA-tagged LC3B (HA-LC3B). However, we could not discern any effect of vIRF-1 on NIX and LC3B interaction while vIRF-1, but not vIRF-1 F10A, bound to NIX (Fig. 4h). These results suggest that vIRF-1 may activate NIX-mediated mitophagy by promoting NIX interaction with other member(s) of the ATG8 family or via unknown mechanism(s).

Mitochondria-localized vIRF-1 promotes survival of lytic cells by activating mitophagy
Given that the PD of vIRF-1 contains the sequences for both mitochondrial targeting and NIX binding, we predicted that deletion of the PD would impede virus replication-induced mitophagy. To test this, we generated an HHV-8 bacterial artificial chromosome 16 mutant (BAC16.DPD) encoding vIRF-1 lacking the PD region, using l-Red recombination techniques described previously 32 . We verified the deletion of the targeted region and the integrity of the mutated versus wild-type BAC16 DNA by sequencing and gel electrophoresis after AvrII or SpeI digestion ( Supplementary Fig. 6). BAC16 (wild type) and BAC16.DPD DNAs were stably transfected into iSLK cells, which express Dox-inducible RTA and are broadly accepted as a model cell line for HHV-8 infection 33 . We first investigated mitochondrial content using flow cytometry analysis of TOM20. Lytic reactivation of the iSLK cells was induced by treatment with 1 mM sodium butyrate (SB) and 1µg/ml Dox. The results showed that the population with higher TOM20 (the M2 region of Fig. 5a) was increased by five-fold in lytic BAC16.DPD cells relative to lytic BAC16 cells while the population with lower TOM20 of both BAC16 and BAC16.DPD cells was increased upon lytic reactivation (the M1 region of Fig. 5a). Therefore, these data support our hypothesis that vIRF-1 plays a critical role in the downregulation of mitochondria content via its PD region. In line with this, lentiviral transduction of vIRF-1 into BAC16.DPD cells promoted an increase in the TOM20 high population (the M1 region) but a decrease in the TOM20 low population (M2 region) (Fig. 5b). Furthermore, we assessed the formation of mitolysosomes in latently and lytically-infected BAC16 and BAC16.DPD cells by immunostaining of TOM20 and LAMP1, a lysosome marker. We used the lysosomal protease inhibitor leupeptin (Leu) to facilitate the detection of mitolysosomes in the iSLK cells (Fig. 5c). The results showed that the formation of mitolysosomes was evident in lytic BAC16 cells, but not in control (BAC16-negative) and lytic BAC16.DPD cells (Figs. 5c and 5d). These genetic studies strongly support our hypothesis that mitochondria-targeted vIRF-1 activates mitophagy during HHV-8 lytic replication.
We previously demonstrated that mitochondria-localized vIRF-1 plays an essential prosurvival role in lytic iBCBL-1 cells 25 . We thus wanted to confirm the results using BAC16-infected iSLK cells. When reactivated with SB/Dox treatment for 2 days, BAC16 iSLK cells exhibited modest morphological changes (cytopathic effects (CPE)) including rounding of cells, characteristic of apoptosis, as assessed by bright-field microscopy ( Fig. 5e). However, BAC16.DPD cells showed stronger CPE (Fig. 5e). Furthermore, we observed that PARP cleavage, which is directly related to apoptotic cell death, was more apparent in BAC16.DPD cells than BAC16 cells upon reactivation (Fig. 5f). Taken together, these results suggest that mitochondrialocalized vIRF-1 mediates, through promotion of mitophagic clearance of mitochondria, protection of lytic cells from mitochondrial-induced cytotoxicity.

A cell-penetrating vIRF-1 PD peptide impairs mitophagy and mitochondrial dynamics
To test the possibility that a short peptide containing the mitochondrial-and NIX-targeting sequences of vIRF-1 may competitively inhibit vIRF-1-activated mitophagy, we synthesized a peptide (TAT-PD1) comprising vIRF-1 residues 1-25 fused to HIV-1 TAT residues 47-57 (Fig.   6a). The TAT-only peptide (hereafter termed TAT) was used as a control. Firstly, we examined whether TAT-PD1 can target to the mitochondria within cells. For immunoblotting detection of TAT-PD1, we used rabbit anti-vIRF-1 antibody, which recognizes an epitope in the N-terminus (residues 1 to 22). Consistent with our previous report 25 , a majority of TAT-PD1 could be detected in the mitochondrial fraction, in particular, the mitochondrial detergent resistant membrane fraction, but not the cytosolic fraction, of iBCBL-1 cells (Fig. 6b). Our IFA analysis showed that TAT-PD1, but not TAT, could inhibit the mitochondrial clearance induced by vIRF-1 and NIX in HeLa.Kyoto cells and led to an accumulation of the mitochondria at the perinuclear area in the cells (Fig. 6c). Next, we examined whether TAT-PD1 influences mitochondrial content in HHV-8-infected cells. Flow cytometric analysis of TOM20 showed that TAT-PD1, but not TAT, elicited an increase of the cell population with a higher mitochondrial content in lytic iBCBL-1 cells (compare M3 between Fig. 6d (ii) and (iv)). Also, TAT-PD1 effectively diminished the cell population with lower mitochondrial content (compare M1 between Fig. 6d (ii) and (iv)), indicating that TAT-PD1 could promote an accumulation of the mitochondria by blocking mitochondrial clearance in lytic cells. Intriguingly, TAT-PD1 also led to an accumulation of the M3 population in latent cells, but to a lesser extent compared to lytic cells (Fig. 6d (iii)), implying that TAT-PD1 might interfere with basal mitophagy, presumably mediated by vIRF-1 that is expressed at low levels in latent PEL cells or by a vIRF-1-independent mitophagy pathway.
Most strikingly, electron microscopy analysis revealed that TAT-PD, but not TAT alone, induced a substantial enlargement of the mitochondria in lytic iBCBL-1 cells (Figs. 6e and 6f). This may be attributed to an accumulation of fused mitochondria and/or inhibition of mitochondria fission by the peptide, together with concomitant inhibition of mitophagy. Interestingly, overexpression of vIRF-1 induced mitochondrial fragmentation, and TAT-PD1, but not TAT, inhibited vIRF-1-induced mitochondrial fragmentation (Fig. 6g). The fragmentation could be mediated by a block of fusion and/or promotion of fission. To test this, we used a dominant negative variant (K38A) of DRP1, a mitochondrial fission protein and found that vIRF-1-induced mitochondrial fragmentation was inhibited by the variant (Fig. 6h), suggesting that vIRF-1 may induce mitochondrial fragmentation potentially via DRP1. As mitochondria fission is often associated with initiation of mitophagy 34 , it is likely that vIRF-1-induced mitochondrial fission may contribute to the promotion of mitophagy although it remains to be determined the exact mechanisms by which vIRF-1 orchestrates mitochondrial dynamics to induce mitophagy.
Interestingly, the electron microscopy image showed that TAT-PD1 induced nuclear condensation in lytic cells (Fig. 6e), indicative of apoptotic cell death. Supporting this observation, annexin V staining analysis showed that TAT-PD1 significantly promoted apoptotic cell death of lytic iBCBL-1 cells (Fig. 6i).

Mitophagy is essential for HHV-8 productive replication
We previously demonstrated that mitochondria-localized vIRF-1 contributes to HHV-8 productive replication 3,25 . We sought to verify these findings using the Dox-inducible system described above. Consistent with our published data, production of encapsidated virions as well as expression of lytic antigens including ORF45 and K8.1 decreased by 40 to 50% in lytic vIRF-1-  (Fig. 2f), suggesting that vIRF-1/NIX-mediated mitophagy is essential for successful lytic replication. Furthermore, we examined the effect of the mitophagy inhibitory peptide on virus productive replication. As expected, TAT-PD1, but not TAT, significantly inhibited the production of encapsidated HHV-8 virions in iBCBL-1 cells that were treated with Dox for 48 and 72 h (Fig. 7c). Interestingly, spontaneous virus production in the absence of Dox was greatly suppressed by TAT-PD1 (Fig. 7c), conceivably due to a cytotoxic effect caused by an accumulation of mitochondria as shown in Fig. 6d. Furthermore, a mitophagy inhibitor, liensinine, apparently inhibited basal and lytic production of HHV-8 virions at a final concentration of 20 µM (Fig. 7d), and a mitochondria fission inhibitor, Mdivi-1, also inhibited production of HHV-8 virions from lytic iBCBL-1 cells with a half maximal inhibitory concentration (IC50) of about 25 µM (Fig. 7e). Our findings demonstrate the importance of mitophagy, promoted by vIRF-1, for productive replication of HHV-8.

Discussion
The role of autophagy in HHV-8 lytic infection has been controversial. The HHV-8encoded lytic proteins K7 and vBcl-2 suppress autophagy by interacting with autophagy proteins Rubicon and Beclin 1, respectively 35,36 , reportedly promoting virus replication. In addition, other lytic proteins including K1, vIL-6, vGPCR, and ORF45 may inhibit the initiation of autophagy by activating PI3K/AKT/mTOR signaling 37 . However, these studies are sharply in contrast to another report that found that RTA protein, a switch for turning on lytic reactivation, can activate autophagy to facilitate HHV-8 lytic replication 38 . A recent report also showed that autophagy is activated concurrently with HHV-8 lytic induction and is essential for viral lytic gene expression using multiple PEL cell lines 39 . Intriguingly, the authors proposed that the fusion of autophagosomes with lysosomes may be blocked at the late steps of virus replication to allow virus particles to be delivered to the extracellular milieu via autophagic vesicles. These results suggest that autophagy might be subject to different temporal regulation for HHV-8 productive replication.
Nonetheless, the role of selective autophagy in HHV-8 infection has not previously been reported.
Herein, we have provided experimental evidence for selective autophagy of mitochondria (mitophagy) in cells lytically infected with HHV-8, and further demonstrated that mitochondrialocalized vIRF-1 activates mitophagy by directly interacting with the mitophagy receptor, NIX, to facilitate HHV-8 replication.
Our experiments with Dox-inducible double silencing of vIRF-1 and NIX in lytically HHV-8-infected PEL cells have demonstrated that vIRF-1 and NIX mechanistically linked for mitochondrial clearance. Conversely, overexpression of both vIRF-1 and NIX, but not each alone, promoted mitochondrial clearance in HeLa.Kyoto cells. This cooperation is likely to be mediated by vIRF-1 activation of NIX-mediated mitophagy rather than vice versa in that vIRF-1 utilized the LC3 interacting motif of NIX for mitochondrial clearance (Fig. 4g).
How does vIRF-1 activate NIX-mediated mitophagy? NIX forms a unique dimeric structure resistant to ionic detergents (e.g. sodium dodecyl sulfate) and reducing reagents (e.g. dithiothreitol) and its dimerization is functionally implicated in apoptosis and possibly mitophagy [40][41][42] . However, the precise details of the mechanisms by which dimerization contributes to the functions of NIX remain elusive. In fact, when overexpressed, NIX was able to readily dimerize but did not induce mitochondrial clearance by itself ( Supplementary Fig. 2 and Fig. 2g).
Moreover, co-transfected vIRF-1 induced mitochondrial clearance without promoting further dimerization of NIX. However, the NIX variants, NIX-TA VAMP1B and NIX-TA FIS1 , lost the ability to form a homodimer and to induce mitochondrial clearance when co-expressed with vIRF-1 although they could localize to mitochondria and bind to vIRF-1 (Fig. 4). Based on these data, NIX dimerization may be a prerequisite but not sufficient to mediate vIRF-1-activated mitophagy.
A recent report showed that the phosphorylation of serine residues adjacent to the LIR of NIX stabilized the interaction with LC3B and promoted NIX-mediated mitophagy 43 . Because vIRF-1 did not enhance the interaction of NIX and LC3B, however, it is unlikely that vIRF-1 induces the NIX phosphorylation. Nevertheless, it is worth investigating in future studies if vIRF-1 promotes NIX interactions with the ATG8 family members other than LC3B. It is also conceivable that vIRF-1 activates other mitophagic pathways mediated by NIX. For example, NIX serves as a substrate of the E3 ubiquitin ligase PARK2, and then ubiquitinated NIX recruits the autophagy receptor NBR1 to the mitochondria 44 . Intriguingly, NIX was also shown to induce lysosomal degradation of mitochondria via MIEAP-induced accumulation of lysosome-like organelles within mitochondria, designated as MALM 45 . For future investigations, it would be interesting to examine if vIRF-1 also activates the non-canonical pathways of NIX for mitochondrial clearance.
In this study, we have also discovered a novel role of vIRF-1 in inducing mitochondrial fragmentation, potentially via a mitochondrial fission protein, DRP1, although the exact mechanism of action remains to be elucidated. Asymmetric mitochondrial fission, in which damaged or dysfunctional mitochondrial components are partitioned from intact components, has been postulated to play a key role in linking the altered mitochondrial components to the mitophagy machinery 46-48 . Interestingly, BNIP3 can induce both mitochondrial fragmentation and mitophagy 49,50 . Notably, DRP1 binding to BNIP3 and recruitment to mitochondria and mitochondrial fission appear to be a prerequisite for BNIP-mediated mitophagy in cardiomyocytes.
Whereas NIX shares about 52% sequence identity (64% sequence similarity) with BNIP3, it is unclear whether NIX also plays a role in mitochondrial fission. Our data showed that overexpression of NIX often induced a clustering of mitochondria at the perinuclear area rather than fragmentation ( Fig. 2g and Supplementary Fig. 5). Thus, it is likely that vIRF-1-induced mitochondrial fission may concomitantly contribute to the initiation of NIX-mediated mitophagy during virus replication.

Competing Interests
The authors declare no competing interests.

Materials & Correspondence
Young Bong Choi to whom correspondence and material requests should be addressed.  Table 1). For replacement of the NIX tail-anchor (TA) region with the TA region of other tail-anchored proteins, overlap extension PCR was carried out as previously described 52 and the fused genes were cloned into vectors pBiT1.1-N (Promega) and pICE_V5 32 .

Mutagenesis of HHV-8 bacmid BAC16
The BAC16 genome was edited using a two-step seamless Red recombination in the context of E. coli strain GS1783 (a kind gift from Greg Smith) as previously described 53 . Briefly, PCR amplification was performed to generate a linear DNA fragment containing a kanamycin resistance expression cassette, an I-SceI restriction enzyme site, and flanking sequences derived from HHV-8 genomic DNA, each of which includes a 40-bp copy of a duplication. The direct connection of the translational initiation codon with codon 76, which results in deletion of the proline-rich domain (PD, 2 to 75 residues) of vIRF-1, was placed in the middle of the duplication ( Supplementary Fig. 6a). This fragment was purified and then electroporated into GS1783 cells harboring BAC16 and transiently expressing gam, bet, and exo, which are expressed in a temperature-inducible manner from the lambda Red operon in GS1783 chromosome. The integrated Kan R /I-SceI cassette was cleaved by I-SceI enzyme that was inducibly expressed by treatment with 1% L-arabinose, resulting in a transiently linearized BAC16. A second Redmediated recombination between the duplicated sequences results in recirculation of the BAC DNA and seamless loss of the Kan R /I-SceI cassette. Kanamycin-sensitive colonies were selected via replica plating. The BAC DNAs were purified using the NucleoBond BAC 100 kit (Clontech).
The recombination area was amplified by PCR and the mutation was verified by DNA sequencing of the PCR amplicon ( Supplementary Fig. 6a). Gross genome integrity and the deletion of the PD region were verified using digestion with AvrII and SpeI restriction enzymes and agarose gel analysis of digestion profiles (Supplementary Fig. 6b).

Immunological assays
Antibodies used in immunological assays including immunoblotting, immunoprecipitation, ImageJ was used for all image preparation and analysis.

Quantification of mitochondrial DNA
The amount of mitochondrial DNA (mtDNA) in lytic iBCBL-1 cells was determined using a combined approach of IFA and fluorescence in situ hybridization (FISH). Firstly, iBCBL-1 cells were treated with Dox for 2 days and IFA was performed with rabbit anti-vIRF-1 antibody as described above to identify lytically infected iBCBL-1 cells. After incubation with an anti-rabbit IgG Alexa 594-conjugated secondary antibody, the cells were post-fixed in Image-iT™ fixative solution for 15 min in the dark. After washing in PBS three times, the cells were incubated with 0.1 µg/µl RNase A for 1 h at 37℃, permeabilized in ice-cold 0.7% Triton X-100 and 0.1M HCl for 10 min on ice, and denatured in 50% formamide and 2x SSC at 80℃ for at least 30 min.
Coverslips were dehydrated by successive rinsing in 70%, 80%, and 95% ethanol and air-dried, and then hybridized with mtDNA-specific probes overnight at 37℃ in a dark and humid chamber.
The probes were prepared by nick translation of six overlapping PCR products (see (Supplementary Table 3 for primer sequences) representing the entire human mtDNA using FISH Tag™ DNA green kit (Thermo Fisher Scientific). Following washing, coverslips were mounted, and cells were imaged as above. The amount of mtDNA were measured using ImageJ. Twenty-four hours after transfection of 293T cells in a 6-well plate with the indicated genes in the NanoBiT vectors, cells were washed in PBS and resuspended in 1 ml of Opti-MEM™ I reduced serum medium (Thermo Fisher Scientific), and 100 µl of the cell suspension was transferred to a 96-well white plate in triplicate. Furimazine (N1110, Promega), a NanoLuc®-luciferase substrate, was diluted in PBS at a ratio of 1 to 50 and 25 µl of the diluted substrate was added to each well.
After 5 min of incubation, the luciferase activity in each well was measured by GloMax® 96 microplate luminometer (Promega).

Apoptosis assays
Dead cells existing prior to the experiments were removed using Histopaque-1077 (Sigma) or Dead Cell Removal kit (Miltenyl Biotec). Apoptotic cells were identified by staining with FITCannexin V as described by the manufacturer's protocol (BioLegend). Also, apoptotic cell death was monitored by detecting the cleaved product of PARP1 via immunoblotting.

Real time-quantitative PCR (RT-qPCR) analysis of gene expression
Total RNAs were isolated using the RNeasy mini kit (Qiagen). First-strand cDNA was synthesized from 1 µg of total RNA using SuperScript IV reverse transcriptase (Thermo Fisher Scientific) with random hexamers. RT-qPCR was performed using an ABI Prism 7500 system (Applied Biosystems) with the FastStart SYBR green/ROX master mix (Sigma-Aldrich).
Reactions were performed in a total volume of 25 µl and contained 50 ng of reverse-transcribed RNA (based on the initial RNA concentration) and gene-specific primers. PCR conditions included an initial incubation step of 2 min at 50 o C and an enzyme heat activation step of 10 min at 95 o C, followed by 40 cycles of 15 seconds at 95 o C for denaturing and 1 min at 60 o C for annealing and extension.

HHV-8 replication assays
For determination of encapsidated HHV-8 genome copy number, viral DNA was purified using Quick-DNA™ Viral Kit (Zymo Research) following pretreatment of the culture supernatants containing HHV-8 virions with DNase I (New England BioLabs) at 37℃ overnight. RT-qPCR was performed as described above with LANA primers (Supplementary Table 3). BAC16 DNA was used as a standard for the calibration curve. Also, the expression levels of lytic antigens including ORF45 and K8.1 were determined by immunoblotting.

Reagents
Chemical reagents were purchased from the following companies: MG132, Cell Signaling           vIRF-1 D P G Q R P N P F G A 12 6 A P N P F G A 6 12       h Mitochondria area,