key: cord-0001891-xvc2154y authors: Kreit, Marguerite; Vertommen, Didier; Gillet, Laurent; Michiels, Thomas title: The Interferon-Inducible Mouse Apolipoprotein L9 and Prohibitins Cooperate to Restrict Theiler’s Virus Replication date: 2015-07-21 journal: PLoS One DOI: 10.1371/journal.pone.0133190 sha: 25a160364a2e056a097e4520f64c49c63140da6c doc_id: 1891 cord_uid: xvc2154y Apolipoprotein L9b (Apol9b) is an interferon-stimulated gene (ISG) that has antiviral activity and is weakly expressed in primary mouse neurons as compared to other cell types. Here, we show that both Apol9 isoforms (Apol9b and Apol9a) inhibit replication of Theiler’s murine encephalomyelitis virus (TMEV) but not replication of vesicular stomatitis virus (VSV), Murid herpesvirus-4 (MuHV-4), or infection by a lentiviral vector. Apol9 genes are strongly expressed in mouse liver and, to a lesser extent, in pancreas, adipose tissue and intestine. Their expression is increased by type I interferon and viral infection. In contrast to genuine apolipoproteins that are involved in lipid transport, ApoL9 has an intracytoplasmic localization and does not seem to be secreted. The cytoplasmic localization of ApoL9 is in line with the observation that ApoL9 inhibits the replication step of TMEV infection. In contrast to human ApoL6, ApoL9 did not sensitize cells to apoptosis, in spite of the presence of a conserved putative BH3 domain, required for antiviral activity. ApoL9a and b isoforms interact with cellular prohibitin 1 (Phb1) and prohibitin 2 (Phb2) and this interaction might contribute to ApoL9 antiviral activity. Knocking down Phb2 slightly increased TMEV replication, irrespective of ApoL9 overexpression. The antiviral activity of prohibitins against TMEV contrasts with the pro-viral activity of prohibitins observed for VSV and reported previously for Dengue 2 (DENV-2), Chikungunya (CHIKV) and influenza H5N1 viruses. ApoL9 is thus an example of ISG displaying a narrow antiviral range, which likely acts in complex with prohibitins to restrict TMEV replication. Type I interferons (IFNs) mediate their antiviral effects through the expression of IFN-stimulated genes (ISGs). Recent studies based on large-scale gene knock down and overexpression screenings have evaluated the antiviral activity of hundreds of ISGs acting against RNA and DNA viruses [1] [2] [3] [4] . Some ISG products display direct antiviral activity and sometimes act on a narrow virus range. Others act by regulating signal transduction pathways controlling IFN production and IFN responses and thus act on a broad range of viruses. The emerging picture is that a given virus is controlled by a specific range of ISGs, some of these ISGs being virus-or virus family-specific and others acting in a more general fashion. We recently identified a group of mouse ISG that are not or weakly expressed in primary neurons after IFN-α/β treatment. Among these genes was the gene encoding apolipoprotein 9b (Apol9b). Apolipoproteins are typically associated with the transport of lipids in the organism. Accordingly, human apolipoprotein L1 (ApoL1) was originally described as a member of the high density lipoprotein family, which is involved in cholesterol transport [5] . However, the other members of the ApoL family were classified on the basis of sequence homology to ApoL1 but their functions may have diverged and remain to be characterized. The ApoL family is highly conserved across species [6] . APOL genes have been implicated in diseases such as schizophrenia and osteoarthritis, and are upregulated by both type I and type II IFNs [6, 7] . In human, six ApoL-coding genes (APOL1, APOL2, APOL3, APOL4, APOL5 and APOL6) are clustered on chromosome 22q12 [8] . ApoL1 has been extensively studied and acts as a restriction factor for trypanosome infection [9] . In addition, ApoL1 and other human ApoL members (ApoL1, L2, L3 and L6) have emerged in high-throughput screenings of ISG activity, as proteins with antiviral activity against various classes of RNA viruses [1, 4] . ApoL1 restricted infection of cells by the AR86 strain of Sindbis virus and more modestly by Venezuelan equine encephalitis virus and human parainfluenza virus type 3, but increased infection by Yellow fever virus. ApoL2 slightly inhibited hepatitis C virus replication but had pro-viral activity toward Influenza A virus and respiratory syncytial virus (RSV). ApoL6 was reported to have antiviral activity against two picornaviruses, coxsackie B virus and poliovirus. ApoL6 also displays inhibitory activity against RSV [1, 4] . Overexpression of ApoL6 triggers apoptosis, which suggests that the antiviral effect of ApoL proteins may correlate with cell sensitization to apoptosis [10] . Mouse ApoL9b is a 310 amino acid-long protein, identical by 97% to ApoL9a. ApoL9a and ApoL9b (referred to collectively as ApoL9) are encoded by distinct genes. The murine Apolipoprotein L family is encoded by 12 genes and 1 pseudogene (Apol6, Apol7a, Apol7b, Apol7c, Apol7e, Apol8, Apol9a, Apol9b, Apol10a, Apol10b, Apol11a, Apol11b and pseudogene Apol10c) clustered on chromosome 15 (Fig 1) . Murine ApoL6 is orthologous to human ApoL6, murine ApoL7a to human ApoL5, murine ApoL7b to human ApoL4 and murine ApoL8 to human ApoL2. It is unclear which human protein is orthologous to mouse ApoL9. We previously reported that ApoL9b is an antiviral ISG active against Theiler's murine encephalomyelitis virus (TMEV or Theiler's virus) [11] . The weak expression of Apol9 in IFN-treated mouse primary neurons contributes to the surprising susceptibility of these cells to virus infection. This study aimed at defining the properties of murine ApoL9 proteins and at characterizing their antiviral functions. Ethics statement: Handling of mice (agreement LA1230472) and experimental procedures were conducted in accordance with the EEC directive 86/609/CEE and the related Belgian law of April 6th 2010. The study and protocol used in this study were approved by the ethics committee of the University of Louvain under the agreement # 2010/UCL/MD/031. To minimize the use of laboratory animals, Apol9a and Apol9b were amplified by RT-qPCR from samples of mouse RNA prepared for other experiments. RNA from organs of naïve C57BL/6 mice were obtained from Hermant et al [12] . Samples from FVB/N mice electro-injected with IFN-expressing and control plasmids were from Sommereyns et al. [13] . Brain and spinal cord RNA samples from TMEV-infected mice were obtained from 3 week-old FVB/N mice that were infected intracranially with 10 6 PFU of the DA1 strain of TMEV for the indicated time (unpublished experiment). These mice were euthanized by deep anaesthesia (intraperitoneal administration of a 200 μl mix of Medetomidin hydrochlorid 300 mg/ml (Domitor) and Ketamine 1.5 g/ml (Anesketin)), and perfused with phosphate buffered saline, before tissue collection. Tissues were snap frozen in liquid nitrogen and kept at -80°C until RNA extraction. The murine ApoL family. A. Phylogenetic tree of the murine Apol gene family, generated by the Gene Orthology/Paralogy prediction method available on the Ensemble server (http://www.ensembl.org). Duplication nodes are indicated by black squares. Numbers represent the duplication confidence score. B. Organization of the mouse Apol genes cluster. Underlined regions (Apol7b/9a and Apol7e/9b) share 98% of nucleotide sequence identity and likely arose from a duplication event. C. Alignment of the human ApoL6 BH3 domain with the hypothetical BH3 domain of murine ApoL proteins. Two amino acids, L and D, shown in bold letters, are functionally conserved in all the BH3 domains identified. D. Structural organization of ApoL9 proteins. The putative BH3 domain (residues 81 to 89) is shown in dark gray. The white area (residues 128 to 144) corresponds to a potential transmembrane (TM) domain. Murine ApoL cDNA and protein sequences were retrieved from the NCBI database [14] . Phylogenetic tree computation was generated by Ensembl [15] . TMbase was used to predict transmembrane domains [16] , SignalP 4.1 to predict signal peptide [17] and SecretomeP to predict non-canonical secretory pathway [18] . CELLO [19] and SOSUI [20] were used to predict subcellular localization. Gene expression data and IFN-responsiveness of other members of the ApoL family were obtained from the interferome server [21] . Cell culture, transfections and Brefeldin A treatment L929 cells (ATCC), Neuro-2A (ECACC), HeLa-M [22, 23] and 293T cells [24] were maintained in Dulbecco's modified Eagle medium (Lonza) supplemented with 10% fetal calf serum (FCS) (Sigma) and 100U/ml of penicillin/streptomycin (Lonza). BHK-21 cells (ATCC) were cultured in Glasgow's modified Eagle's medium (GMEM) (Sigma) supplemented with 10% newborn bovine serum (Gibco), 100 U/ml of penicillin/streptomycin (Lonza), and 2.6 g of tryptose phosphate broth per liter (Difco). Plasmid transfections were performed using TransIT-LT1 transfection reagent (Mirus), according to the manufacturer's instructions. For lentivirus production and co-immunoprecipitation experiments, 293T cells were transfected by the calcium phosphate method. Transfection of the TM973 replicon RNA was performed as follows. The pTM973 plasmid carrying the replicon cDNA was linearized with restriction enzyme ClaI and subjected to in vitro transcription with T7 RNA polymerase, using the ribomax T7 kit (Promega, P1300). 0.5 μg of replicon RNA was then transfected in L929 cells (30,000 cells per well seeded in 24-well plates) with 1 μl of mRNA transfection reagent and 1 μl of booster reagent (Mirus, MIR 2250). For Brefeldin A treatment, GolgiPlug (ref 555029, BD Biosciences) was diluted 1000-fold in the culture medium 20 hours after transfection of the cells and 4 hours prior to cell collection for western blot analysis. Interferon IFN-β was produced and quantified, as described previously, from 293T cells transfected with pcDNA3-IFN-β [25] , and diluted in culture medium at 5 units of IFN per ml for cell treatment. Viruses are presented in Table 1 . Theiler's murine encephalomyelitis viruses (TMEV-or Theiler's virus) used in this study were KJ7 and KJ26, GFP/eGFP-expressing derivatives of the persistent DA1 strain and of the neurovirulent GDVII strain, respectively [11] . These viruses contain capsids adapted to infect L929 cells efficiently [26] . KJ7 and KJ26 were produced, as described [27, 28] , from the corresponding full-length cDNA clones pKJ7 and pKJ26. The TMEV replicon expressing the firefly luciferase (called TM973) was constructed by replacing the L-capsid-part of 2A region (nucleotides 1064 to 3925 of pTMDA1-accession JX443418 [29] ), by a DNA segment assembled by PCR and carrying in frame: i) the firefly luciferase ORF PCR-amplified from pGL4.16 (Promega); ii) the 2A coding sequence of FMDV which enables co-translational dissociation of the protein segment translated downstream of this sequence [30] , and iii) the CRE replication signal of pTMDA1 [31] . The VSV (Strain Indiana) derivative expressing eGFP from an additional transcription unit located between the G and L genes was a gift from Martin Schwemmle (Univ. of Freiburg, Germany) [32] . The MuHV-4 derivative used in this study contains the eGFP gene in a bacterial artificial chromosome cassette [33] and was generated and purified as described previously [34] . Viruses used in this study were produced and titrated on BHK-21 cells. Lentiviral vectors used in this study are listed in Table 2 . Lentiviral vectors used for RNA interference are described in the RNA interference section. Lentiviral vectors used for protein expression were derived from pCCLsin.PPT.hPGK.GFP.pre. [35] . pTM944 is a derivative of this vector, which carries the eGFP coding sequence translated from the TMEV IRES. Transcription of the IRES-eGFP region is driven by the cytomegalovirus (CMV) promoter. Apol9 constructs were cloned in lentiviral vectors pTM942, pTM943 or pTM945. pTM942 is bicistronic vector carrying a phosphoglycerate kinase (PGK) promoter, a multi-cloning site, the IRES of TMEV and the mCherry coding sequence [36] . pTM943 contains the CMV promoter and a multi-cloning site. pTM945 is similar to pTM942 but contains the CMV promoter instead of the PGK promoter. To construct these vectors, the coding sequences of Apol9b and Apol9a were PCR-amplified from C57BL/6 mouse spleen cDNA, using Phusion polymerase (Finnzymes) and cloned in pTM942 using AgeI/BsiWI restrictions sites introduced in the PCR primers. Vectors encoding FLAG-tagged constructs and deletion mutants of Apol9 were obtained by PCR amplification of this gene with primers carrying the FLAG-coding sequence and AgeI and BsiWI restrictions sites that were used to clone the PCR fragment in the lentiviral vectors ( Table 2) . The region coding the putative BH3 domain of ApoL9 was mutated by a two-step PCR with overlapping divergent primers used to introduce the mutations. The amino acid sequence of BH3 was mutated from LHALADHAE to GHAGAAHGE for ApoL9a and GHAGAAHAE for ApoL9b. In all the constructs, the fragments obtained by PCR and their boundaries with the vectors were sequenced to rule out unexpected mutations. Lentivirus particles were produced, as described previously [11] , by transient transfection of 293T cells. Transduction of cells were performed by adding filtered, lentivirus stock in the culture medium of cells seeded at low confluence. The plasmid used to express C-terminally FLAG-tagged mouse IFN-αA is pcDNA3-IFN-αA-FLAG (or pAGE2) [12] . pMK117 is a pLKO.1 [37] derivative constructed by substituting the E2-Crimson coding region for that of the puromycin resistance gene. E2-Crimson was amplified by PCR from pEF1alpha-E2-Crimson vector (Clontech, 63181) with primers introducing BamHI and KpnI restrictions site and cloned into pLKO.1. shRNA sequences targeting the mouse Phb1 and Phb2 transcripts were selected from the RNA mission collection (Sigma) and cloned in pMK117 using AgeI and EcoRI restriction sites introduced in the PCR primers. Constructs pMK118 (shRNA TRCN0000312198) and pMK120 (shRNA TRCN0000302352), targeting coding sequences of Phb1 and Phb2 respectively, gave a more complete inhibition than constructs pMK119 and pMK121 that express shRNA TRCN0000088454 and TRCN0000054428, targeting the coding sequence of Phb1 and the 3'UTR of Phb2, respectively (not shown). pMK118 and pMK120 were thus used for subsequent knock down experiments. Stocks of lentiviral vectors derived from pLKO.1 were generated, as described previously [11] . Efficiency of RNA silencing was assayed 3 days after transduction, by evaluating Phb1 and Phb2 protein expression by Western blot. Caspase 3/7 activity was measured using the Caspase-Glo 3/7 Assay (Promega, G811A). Briefly, 10,000 L929 cells grown in 96 well plates were left untreated, infected with 1 PFU per cell of KJ26 for 10 hours or treated with 1uM staurosporine (Sigma, S-4400) for 7 hours to induce cell apoptosis [38] . After plate equilibration at room temperature, 50 μl of Caspase-Glo 3/7 reagent was added to each well. The plate was placed on a plate shaker, in the dark, for one hour at room temperature and luminescence was measured with a plate-reading GloMax luminometer (Promega). Raw data are presented as relative light units (RLU). In samples from L929 cells transfected with the TM973 replicon, firefly luciferase activity was measured using the luciferase assay system (Promega, E1501), according to the manufacturer's instructions. Luminescence was measured with a GloMax 20/20 luminometer (Promega). Protein extracts were prepared from 293T cells, 24 hours post-transfection, as previously described [39] . FLAG-tagged ApoL9 and IFN-α were detected by western blot using sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) gels containing 12% acrylamide and run in a Tris-Tricine buffer. Blots were probed with anti-FLAG polyclonal (Covance, PRB-132P) and anti-β-actin monoclonal (Sigma, A5441) antibodies. Co-immunoprecipitation assays were conducted on transfected 293T or Neuro-2A cells, or on transduced L929 cells, as previously described [29] . Proteins were immunoprecipitated with anti-FLAG (Covance, PRB-132P), anti-Phb1 (SantaCruz, sc-28259) or anti-Phb2 (SantaCruz, sc-67045) polyclonal antibodies and proteins were detected using SDS-PAGE and immunoblot analysis. After co-immunoprecipitation, samples were resolved using a 10% Tris-Glycine SDS-PAGE and proteins were visualized using PageBlue (Thermo Scientific, 24620). Bands of interest were cut out from the gel and digested with trypsin (50ng/μl in 50 mM NH4HCO3 buffer, PH 8.0). The peptides were analyzed, as previously described [40] , by capillary LC-tandem mass spectrometry in a LTQ XL ion trap mass spectrometer (ThermoScientific, San Jose, CA) fitted with a microelectrospray probe. The data were analyzed with the ProteomeDiscoverer software (ThermoScientific, version 1.4.1), and the proteins were identified with SequestHT against a target-decoy nonredundant human or mouse proteins database obtained from Uniprot. The following parameters were used: trypsin was selected with proteolytic cleavage only after arginine and lysine, number of internal cleavage sites was set to 1, mass tolerance for precursors and fragment ions was 1.0 Da, considered dynamic modifications were + 15.99 Da for oxidized methionine. Peptide matches were filtered using the q-value and Posterior Error Probability calculated by the Percolator algorithm ensuring an estimated false positive rate below 5%. The filtered Sequest HT output files for each peptide were grouped according to the protein from which they were derived. Total RNA preparations, reverse transcription (RT) and quantitative PCR (qPCR) reactions were performed as previously described [41] . qPCR standards consisted of 10-fold dilutions of quantified plasmids carrying the sequence to be amplified: pTM793 for β-actin, pMK72 for Apol9b, pMK68 for Apol9a and pCS40 for Oasl2. These plasmids are derivatives of pCR4-ToPo (Invitrogen) in which the corresponding PCR fragments were cloned. Real-time PCR primer sequences and reference plasmid used are given in Table 3 . HeLa-M cells were fixed for 15 min with PBS containing 4% of paraformaldehyde (PFA). Cells were then permeabilized for 5 min with PBS containing 0.1% of Triton X-100. Blocking occurred for 1 h at room temperature in TNB blocking reagent (Perkin Elmer). Cells were next incubated with the anti-FLAG mouse monoclonal antibody (Sigma-Aldrich, F3165; 1/500) diluted in TNB for one hour. AlexaFluor-488 conjugated secondary antibody (Invitrogen) was incubated for 1 h at a 1/800 dilution in TNB. Finally, cells were mounted with Mowiol for fluorescence microscopy. Antibodies used for double labeling were: anti-transferrin receptor (Abcam mouse monoclonal, ref Ab5339); anti-dsRNA (K1 and J2 mouse monoclonal, English and scientific consulting Bt); anti-TMEV VP1 protein (mouse monoclonal F12B3, [11] ); anti-PBodies (mouse monoclonal anti-p70 S6 kinase, Santa Cruz ref sc-8418 [42] ). For nuclear staining, cells were incubated for 5 min in Hoechst 33258 before the final washing steps. Lipid dropplets were stained at the same step, by incubating the cells for 15 in BODIPY 493503 diluted 1/1000 (Life Technologies). The Golgi compartment was identified after staining glycosylated proteins with Alexafluor564-conjugated wheat germ agglutinin (Molecular probes-W11262). The endoplasmic reticulum was identified by co-transfection of pDsRed-ER as performed in [12] . For staining the endosomal pathway, cells were incubated for 60 min before fixation with 1 mg/ml of 10kDa Dextran-Alexafluor594 (Life technologies D-22913). For mitochondrial staining, cells were incubated for 45 min before fixation with 200 nM mitotracker red FM (Life technologies M-22425). Fluorescence microscopy was performed using a spinning disk confocal microscope equipped with an Axiocam MRm camera (Zeiss). Intensity, contrast, and color balance of images were equilibrated using Zen (Zeiss) and Adobe Photoshop. After dissociation with trypsin-EDTA, cells were suspended in PBS buffer containing 5% of FCS and 0.5% of paraformaldehyde. Data acquisition was done with a LSR Fortessa flow cytometer (BD Bioscience) using the FACSDiva software. Data were analyzed using Results are expressed as mean +/-standard deviations (SD). Statistical significance was analyzed using the two-tailed Mann-Withney U test. NS is indicative of p-values higher than 0.05; Ã p-value < 0.05; ÃÃ p-value < 0.01; ÃÃÃ p-value < 0.001 Mouse Apol genes appear to derive from a common ancestor ( Fig 1A) and are clustered on chromosome 15 (Fig 1B) . Apol9a and Apol9b genes are likely issued from a gene duplication event, since the region encompassing the Apol7b and Apol9a genes shares 98% nucleotides identity with that encompassing the Apol7e and Apol9b genes ( Fig 1B) . Murine ApoL proteins possess a conserved domain named ApoL (referred as pfam05461) whose function is still unknown. Previously, human ApoL6 was reported to exert a pro-apoptotic activity through a Bcl-2 homology (BH) domain, designated BH3 [10] . Sequence alignments predict the presence of BH3 domains in several mouse ApoL proteins, including ApoL9 (Fig 1C) [43] . ApoL9 is mostly hydrophilic but contains a short hydrophobic region (residues 128 to 144), compatible with a transmembrane domain (Fig 1D) . Since anti-ApoL9 antibodies were not available, we constructed vectors encoding N-and C-terminally FLAG-tagged ApoL9b, to define the subcellular localization of the protein by FLAG immunodetection. First, we confirmed that N-terminally and C-terminally FLAG-tagged ApoL9b conserved the ability to exert an antiviral activity against TMEV in L929 cells (Fig 2A) . Next, FLAG-tagged ApoL9b was detected by immunofluorescence in HeLa-M cells transduced with lentiviral constructs. ApoL9 localization was similar in cells transduced with pMK65 and pMK66, expressing N-and C-terminally tagged ApoL9b proteins respectively (not shown). We observed a diffuse cytoplasmic labeling and accumulation of ApoL9b as granulelike structures in part of the cells (Fig 2B) . ApoL9b granules did not co-localize with markers for mitochondria (Fig 2C) , endoplasmic reticulum, endosomes, lysosomes, P-bodies, Golgi vesicles, and lipid droplets or with viral components such as double-stranded RNA or VP1 capsid antigen (data not shown). Interestingly, in contrast to prototypic apolipoproteins, which are secreted, ApoL9 does not contain any predicted signal sequence (Signal P scores were 0.119 for ApoL9a and 0.133 for ApoL9b -cutoff value was 0.5) nor is it predicted to be secreted by a non-canonical secretory pathway by the Secretome P server. Absence of ApoL9 secretion by the classical secretion pathway was further confirmed by showing that Brefeldin A treatment did not trigger accumulation of N-or C-terminally FLAG-tagged ApoL9 proteins in transfected 293T cells, in contrast to C-terminally FLAG-tagged IFN-α that was taken as a secreted protein control (Fig 2D) . We conclude that ApoL9 are cytoplasmic proteins that are not secreted, at least not by the classical secretion pathway. Tissue distribution of Apol9a and Apol9b expression and induction by type I IFN Expression patterns of Apol9a and Apol9b in mouse organs were very similar, in agreement with the fact that their putative gene promoter sequences were highly conserved. In untreated mice, both genes were predominantly expressed in the liver, and to a lesser extent, in pancreas, adipose tissue and intestine ( Fig 3A) . Circulating IFN-α triggered Apol9 upregulation in many tissues ( Fig 3A) . In the liver, upregulation of Apol9 expression by IFN-α was less clear, likely because basal expression of Apol9 was already very high in this organ. In the central nervous system (CNS), the induction of Apol9 was also moderate, presumably because ApoL9 gene transcription is low in neurons and because circulating IFN-α did not cross the blood brain barrier. However, Apol9 expression was upregulated in the brain and spinal cord after infection with Theiler's virus, likely because of local type I IFN production ( Fig 3B) . Accordingly, screening of the GEO microarray database [44] confirmed that Apol9 expression increased with viral infection, in vivo, in many mouse tissues (Table 4 ). In L929 fibroblasts, time course of Apol9a and Apol9b gene induction after IFN-β stimulation was comparable to that of Oasl2 (Fig 3C) . Upregulation of Apol9 gene transcription was almost maximal from 6 hours post-treatment, suggesting that Apol9 expression is directly induced by type I IFN. Taken together, our data indicate that Apol9a and Apol9b are bona fide ISGs that share a similar expression profile. Intriguingly, in spite of the unusual cytoplasmic nature of these apolipoproteins, they are mainly produced by liver cells, like typical apolipoproteins. any, on the other tested viruses. To confirm these results, infectious virus yield was examined by plaque assay. Therefore, cells were transduced with the same Apol9-expressing lentiviral vectors. Three days after transduction, mCherry-positive cells were sorted and cultured for 2 days before infection with the GFP-expressing TMEV, MuHV-4 and VSV derivatives. Infectious virus yield was measured by plaque assay (Fig 4B) . Again, ApoL9 expression significantly reduced TMEV but not MuHV-4 or VSV production. These data confirm the restricted virus range of ApoL9 antiviral activity. Human ApoL6 was reported to induce apoptosis when overexpressed in cells, most likely through its BH3 domain [10, 45] . Since this domain is well conserved in other ApoL proteins including ApoL9 (Fig 1C) , we tested whether ApoL9a and ApoL9b proteins could exert their antiviral activity by sensitizing cells to apoptosis. Apoptosis was measured with a caspase 3/7 assay, in L929 cells that were transduced with an empty vector or with lentiviral vectors expressing Apol9a or Apol9b (Fig 5A) . In the absence of TMEV infection, ApoL9 expression did not increase but rather triggered a minor decrease of caspase activity. Staurosporine strongly increased caspase activity as expected, irrespective of ApoL9 expression. Infection of cells with TMEV induced apoptosis. However, cells expressing Apol9 were slightly but significantly less sensitive to virus-induced apoptosis than control cells (Fig 5A) . This result excludes that ApoL9 may act by sensitizing infected cells to apoptosis. In contrast, ApoL9 appears to restrict viral infection and thereby to limit virus-induced apoptosis. Similar results were obtained by comparing caspase activity in three Apol9b-overexpressing and three control L929 cell clones obtained previously [11] (data not shown). In conclusion, contrary to the reported pro-apoptotic activity of human ApoL6, the antiviral effect of ApoL9 does appear to be mediated by a pro-apoptotic influence of ApoL9. To test whether ApoL9 acts on entry or on replication of TMEV, we used a TMEV replicon (TM973) expressing the firefly luciferase. Replicon RNA was transfected into cells to bypass the virus entry step and luciferase activity was followed as a surrogate marker of TMEV replication. Replicon-derived luciferase activity was measured in L929 cells transduced with the empty pTM942 vector or with the pTM942 derivatives expressing ApoL9 (Fig 5B) . From 10 hours post-transfection, luciferase activity was significantly lower in all ApoL9-expressing cell populations. We conclude that ApoL9 restricts the replication of TMEV. We then generated L929 cells that overexpress ApoL9 mutants in order to map the ApoL9 domains required for antiviral activity (Fig 5C) . For both Apol9a and Apol9b, deletion of either the C-or the N-terminal 50 residues of ApoL9 abrogated antiviral activity as did the mutation of the putative BH3 domain (amino acids 81-89) (Fig 5D) . These results suggest that mutations may have affected the ApoL9 conformation required for antiviral activity. ApoL9 interacts with cellular prohibitin-1 and -2 Next, we investigated whether ApoL9 would act by interacting with cellular proteins. Therefore, mouse (Neuro-2A) and human (293T) cells were transfected with vectors allowing the expression of N-or C-terminally FLAG-tagged ApoL9b. Cellular partners of ApoL9b were identified 24 hours post-transfection, by co-immunoprecipitation and mass spectrometry. Two protein bands were detectable on Comassie blue-stained gels loaded with products that co-precipitated with ApoL9b in both Neuro-2A and 293T cells (Fig 6A and data not shown) . For both mouse and human cells, bands 1 and 2 were identified as prohibitin 1 and -2 (Phb1 and Phb2) respectively (Fig 6B and S1 Dataset) . Prohibitins (PHBs) 1 and 2 are two ubiquitous multifunctional proteins that may act as chaperones (reviewed by Thuaud, Ribeiro [46] ). Interaction between ApoL9b and PHBs was further confirmed by detection of FL AG-ApoL9b and ApoL9b-FLAG after immunoprecipitation of endogenous Phb1 or Phb2 proteins from cells expressing the FLAG-tagged ApoL9 proteins (Fig 6C) . In these experiments, a N-terminally FLAG-tagged TMEV L Ã protein, used as negative control, was not (Phb1) or very weakly (Phb2) co-immunoprecipitated with prohibitins, supporting the specificity of the ApoL9-prohibitin interaction. The experiment could, however, not be confirmed with endogenous ApoL9 proteins, given the lack of anti-ApoL9 antibodies. ApoL9 and PHBs contribute to antiviral activity To determine whether association with PHBs is necessary for ApoL9 antiviral activity, we used a Phb knock down strategy. L929 cells transduced to co-express mCherry and Apol9b or Apol9a, or mCherry alone (vector), were subsequently transduced with lentiviral vectors coexpressing shRNAs targeting the coding sequence of either Phb1 or Phb2 and the far-red fluorescent E2-Crimson protein [47] , prior to infection with the GFP-expressing KJ7 virus. Knock down of PHBs expression was confirmed by western blot (Fig 7A and data not shown) . As reported previously [48] , knocking down either Phb1 or Phb2 triggered a strong reduction in the amount of both Phb proteins, presumably because stability of Phb1 and Phb2 depends on their mutual interaction. It is worth noting that PHBs knock down triggered a cell growth inhibition from about 5 days post-transduction, which prevented the use of sorted pure populations of PHB knock down cells. Thus, three days after the transduction of ApoL9-expressing cells with vectors expressing shRNA directed against PHBs or control vectors, cells were infected with the GFP-expressing KJ7 virus and analyzed by flow cytometry. As in previous experiments, overexpression of both Apol9a and Apol9b decreased KJ7 replication. Knocking down Phb2 and, to a lesser extent, Phb1 expression in cells that overexpressed Apol9 significantly enhanced but did not fully restore infection (Fig 7B) . Interestingly, knocking down Phb2 also increased infection by TMEV in cells that did not overexpress Apol9, suggesting that this prohibitin has antiviral activity per se, against TMEV. This increase of TMEV infection after knock down of Phb gene expression was unexpected. Indeed, PHBs were previously reported to act as pro-viral factors that facilitate Dengue 2 (DENV-2), Chikungunya (CHIKV) and H5N1 virus entry in cells [48] [49] [50] . We therefore compared the influence of Phb knock down on infection by TMEV (KJ7 or KJ26), MuHV-4, and VSV, using flow cytometry (Fig 7C) . While Phb2 knock down reproducibly favored the infection by both strains of TMEV, it failed to significantly affect MuHV-4 infection and even inhibited VSV infection. We also monitored virus yield, by plaque assay, after PHBs knock down. For this purpose, cells were transduced with the shRNA-expressing lentiviral vectors, at concentrations calculated to yield 90-95% of transduction. Three days after transduction, cell populations were infected with VSV, MuHV-4 and TMEV and virus yield was measured by plaque assay (Fig 7D) . This experiment confirmed the strong pro-viral activity of PHBs toward VSV but failed to show the antiviral effect of PHBs toward TMEV. The discrepancy between the highly reproducible negative effect exerted by PHBs against TMEV infection, measured by flow cytometry and the positive effect on infectious virus particles production may be the consequence of contrasting roles of PHBs at different steps of the virus infection cycle (pro-viral for entry and anti-viral for replication). Alternatively, it may be an indirect effect of the progressive cell growth inhibition that occurs over time, after PHBs knock down. Our data show that Apol9a and Apol9b are two antiviral ISGs that are constitutively expressed in mouse tissues such as liver, pancreas, adipose tissue and intestine. Expression of these genes can be further upregulated by type I IFN. Both ApoL9 isoforms display antiviral activity against Theiler's virus but not against VSV, MuHV-4 or a lentiviral vector derived from HIV-1. ApoL9 thus belongs to the group of ISG displaying a narrow antiviral range. Other members of the Apol gene family (Apol6, l7a, l7c, l9a and l9b) are upregulated by type I and type II IFN but have different tissue distributions. Apol7a is highly expressed in the liver, like Apol9, while Apol6 is mainly expressed in adipose tissue and mammary glands and Apol7c in CD8+ cells and lymph nodes [21] . The function of these proteins is still elusive. From their IFN-inducible property, it can be anticipated that they also act as antiviral proteins and, given their tissue distribution, that they contribute to shape the tissue tropism of viruses. In contrast to ApoL1, ApoL9 lacks a typical signal sequence. We failed to detect FLAGtagged ApoL9 in the supernatant of cells transfected with constructs expressing either N-terminally or C-terminally tagged proteins (data not shown) and brefeldin A treatment did not lead to ApoL9 intracellular accumulation ( Fig 2D) . Furthermore, immunofluorescent detection did not show any association of ApoL9 with components of the secretory pathway, such as the endoplasmic reticulum or the Golgi apparatus. Our interpretation of these data is that ApoL9 is not secreted, at least not by the classical secretory pathway. This conclusion contrasts with the recent report of Sun et al. who reported that ApoL9 may be secreted by macrophages to enhance epithelial cell proliferation [51] . To reconcile these observations, one could hypothesize that ApoL9 proteins may be secreted in a cell type-specific fashion. In line with a cytoplasmic localization of ApoL9, this protein was found to inhibit the replication of a TMEV replicon that was transfected into cells. ApoL9 did not, however, act by sensitizing cells to apoptosis. ApoL9 immunolocalization did not provide a clear hint toward the mode of action of this protein. In part of the cells, ApoL9 formed granule-like structures. In the absence of anti-ApoL9 antibodies to detect the endogenous protein, we cannot prove that these granules are formed in physiological conditions. However, a recent work reports the appearance of similar granules in the case of human ApoL2 [52] . Neither ApoL2 nor ApoL9 granules colocalized with tested subcellular compartments. Nevertheless, most ApoL9 molecules were detected as diffuse cytoplasmic proteins and it is therefore difficult to rule out their association with specific cell components or with components of virus replication complexes. By co-immunoprecipitation, we found that ApoL9 interacts with cellular prohibitins Phb1 and Phb2, both in mouse and human cells. PHBs are highly conserved and widely expressed proteins, which are present in many different cellular compartments including the cytoplasm, mitochondria and cell surface [53, 54] . They interact with many proteins, including viral proteins and are sometimes considered as chaperones (reviewed by Thuaud, Ribeiro [46] ). Knocking down Phb2 expression did not completely abrogate the antiviral protection conferred by ApoL9 overexpression (Fig 7B) . This suggests that PHBs association with apoL9 is not strictly required for ApoL9 antiviral activity. Alternatively, the incomplete abrogation of ApoL9 activity may result from incomplete Phb knock down. It is tempting to speculate that PHBs may act as chaperones to stimulate the antiviral activity of ApoL9. More surprisingly, Phb2 knock down slightly but reproducibly increased cell susceptibility to TMEV replication, independently of ApoL9 expression. This suggests that PHBs may, per se, have an antiviral activity. This is contrasting with the strong proviral effect of PHBs observed for VSV (Fig 7B and 7C ) Dengue virus 2 (DENV-2), Chikungunya virus (CHIKV) and H5N1 influenza virus [48] [49] [50] . In conclusion, our data suggest that ApoL9a and ApoL9b are two IFN-induced cytoplasmic proteins that antagonize TMEV by cooperating with prohibitins. These apolipoproteins largely differ from genuine lipid transport-associated lipoproteins, both by their intracytoplasmic localization and by their antiviral function. Why ApoL9 isoforms are particularly weakly expressed in neurons is still unclear. Supporting Information S1 Dataset. ApoL9-interacting proteins. Excel file showing the complete set of proteins identified by mass spectrometry, in bands 1 and 2 of Commassie blue-stained gels (see Fig 6A) loaded with FLAG-ApoL9 immunoprecipitates from Neuro-2A and 293T cells. (XLSX) A diverse range of gene products are effectors of the type I interferon antiviral response Systematic identification of type I and type II interferoninduced antiviral factors A short hairpin RNA screen of interferon-stimulated genes identifies a novel negative regulator of the cellular antiviral response Pan-viral specificity of IFN-induced genes reveals new roles for cGAS in innate immunity Human High Density Lipoprotein Apolipoprotein Expressed by the Pancreas The apolipoprotein L family of programmed cell death and immunity genes rapidly evolved in primates at discrete sites of host-pathogen interactions Gene expression analysis in schizophrenia: reproducible up-regulation of several members of the apolipoprotein L family located in a high-susceptibility locus for schizophrenia on chromosome 22 The human apolipoprotein L gene cluster: identification, classification, and sites of distribution Apolipoprotein L-I is the trypanosome lytic factor of human serum Apolipoprotein l6, a novel proapoptotic Bcl-2 homology 3-only protein, induces mitochondria-mediated apoptosis in cancer cells Inefficient type I IFN-mediated antiviral protection of primary mouse neurons is associated with the lack of apolipoprotein L9 expression IFN-epsilon is constitutively expressed by cells of the reproductive tract and is inefficiently secreted by fibroblasts and cell lines IFN-lambda (IFN-lambda) is expressed in a tissuedependent fashion and primarily acts on epithelial cells in vivo The NCBI BioSystems database TMBASE-A database of membrane spanning protein segments Locating proteins in the cell using TargetP, Sig-nalP and related tools Feature-based prediction of non-classical and leaderless protein secretion Prediction of protein subcellular localization Physicochemical factors for discriminating between soluble and membrane proteins: hydrophobicity of helical segments and protein length Interferome v2.0: an updated database of annotated interferon-regulated genes Enhancement of rhinovirus plaque formation in human heteroploid cell cultures by magnesium and calcium Analysis of mutation in human cells by using an Epstein-Barr virus shuttle system Characterization of the murine alpha interferon gene family Adaptation of Theiler's virus to L929 cells: mutations in the putative receptor binding site on the capsid map to neutralization sites and modulate viral persistence Protein 2A is not required for Theiler's virus replication Molecular cloning of the complete genome of strain GDVII of Theiler's virus and production of infectious transcripts Evasion of antiviral innate immunity by Theiler's virus L* protein through direct inhibition of RNase L Analysis of the aphthovirus 2A/2B polyprotein 'cleavage' mechanism indicates not a proteolytic reaction, but a novel translational effect: a putative ribosomal 'skip'. The Journal of general virology A coding RNA sequence acts as a replication signal in cardioviruses Fusion-active glycoprotein G mediates the cytotoxicity of vesicular stomatitis virus M mutants lacking host shut-off activity. The Journal of general virology Cloning and mutagenesis of the murine gammaherpesvirus 68 genome as an infectious bacterial artificial chromosome Proteomic characterization of murid herpesvirus 4 extracellular virions Gene transfer by lentiviral vectors is limited by nuclear translocation and rescued by HIV-1 pol sequences Inefficient type I interferon-mediated antiviral protection of primary mouse neurons is associated with the lack of apolipoprotein l9 expression A lentiviral RNAi library for human and mouse genes applied to an arrayed viral high-content screen Caspase and proteasome activity during staurosporin-induced apoptosis in lens epithelial cells IFN-ε is constitutively expressed by cells of the reproductive tract and is inefficiently secreted by fibroblasts and cell lines The protein-disulfide isomerase DsbC cooperates with SurA and DsbA in the assembly of the essential beta-barrel protein LptD Cardiovirus leader proteins are functionally interchangeable and have evolved to adapt to virus replication fitness Mammalian stress granules and processing bodies Structure of the BH3 domains from the p53-inducible BH3-only proteins Noxa and Puma in complex with Mcl-1 NCBI GEO: archive for functional genomics data sets-update Apolipoprotein L6, induced in atherosclerotic lesions, promotes apoptosis and blocks Beclin 1-dependent autophagy in atherosclerotic cells Prohibitin ligands in cell death and survival: mode of action and therapeutic potential A rapidly maturing far-red derivative of DsRed-Express2 for whole-cell labeling Identification and characterization of prohibitin as a receptor protein mediating DENV-2 entry into insect cells Identification of prohibitin as a Chikungunya virus receptor protein Identification of human host proteins contributing to H5N1 influenza virus propagation by membrane proteomics Type I interferons link viral infection to enhanced epithelial turnover and repair Apolipoprotein L2 contains a BH3-like domain but it does not behave as a BH3-only protein Reversal of obesity by targeted ablation of adipose tissue Vi polysaccharide of Salmonella typhi targets the prohibitin family of molecules in intestinal epithelial cells and suppresses early inflammatory responses Testosterone and interleukin-1beta increase cardiac remodeling during coxsackievirus B3 myocarditis via serpin A 3n. American journal of physiology Heart and circulatory physiology Gene expression analysis of host innate immune responses in the central nervous system following lethal CVS-11 infection in mice. Japanese journal of infectious diseases Roles of vaccinia virus genes E3L and K3L and host genes PKR and RNase L during intratracheal infection of C57BL/6 mice Guanylate binding protein 4 negatively regulates virus-induced type I IFN and antiviral response by targeting IFN regulatory factor 7 Interleukin-6 is a potential biomarker for severe pandemic H1N1 influenza A infection pandemic H1N1 influenza virus elicits similar clinical course but differential host transcriptional response in mouse, macaque, and swine infection models We are grateful to Stéphane Messe and Muriel Minet for expert technical assistance, and Nicolas Dauguet (Ludwig Institute for Cancer research, Brussels) for his kind and professional help in flow cytometry and cell sorting.