Rabbit antibodies against MFF (17090-1-AP, ProteinTech, Manchester, UK) 30, Myc-tag (71D10, 2,278, Cell Signalling Technology, Beverly, MA, United States), FLAG epitope (F7425, Sigma-Aldrich, St. Louis, MO, United States), β-Actin (4,967, Cell Signalling, Danvers, MA, United States), PEX14 (GTX129230, GeneTex, CA,United States) and MAVS (A300-782A, Bethyl Laboratories, TX, United States), and mouse antibodies against MAVS (E-3, SC-166583, Santa Cruz Biotechnology, Dallas, TX, United States), p-STAT1 (Y701, BD Biosciences, San Jose, CA, United States), DLP1 (611,113, BD Bioscience, San Jose, CA, United States), PMP70 (SAB4200181, Sigma-Aldrich, St. Louis, MO, United States), COXIV (4,850, Cell Signalling Technology, Beverly, MA, United States) and α-Tubulin (T9026, Sigma-Aldrich, St. Louis, MO, United States) were used for immunoblotting. Mouse antibodies against PMP70 (SAB4200181, Sigma-Aldrich, St. Louis, MO, United States), Myc epitope (9E10, SC-40, Santa Cruz Biotechnology, Dallas, TX, United States) were used for immunofluorescence analyses. Additionally, mouse antibodies against Myc epitope (9E10, SC-40, Santa Cruz Biotechnology, Dallas, TX, United States) and MAVS (E-3, SC-166583, Santa Cruz Biotechnology, Dallas, TX, United States) were also used for the immunoprecipitation experiments. Species-specific anti-IgG antibodies conjugated to HRP (Bio-Rad, Richmond, CA, United States) or IRDye 800 CW and IRDye 680RD secondary antibodies (LI-COR Biotechonology, Cambridge, UK) were used for immunoblotting and the fluorophores TRITC (Jackson ImmunoResearch, Cambridge, UK) and Alexa 488 (Invitrogen, Waltham, MA, United States) were used for immunofluorescence.

The plasmids STING-FLAG and GFP-RIG-I-CARD (kindly provided by Dr F. Weber, Justus-Liebig Universität Giessen, Germany) and vMIA-Myc (kindly provided by Dr V. Goldmacher, ImmunoGen Inc., Cambridge, MA, United States) were used for mammalian expression.

Mouse embryonic fibroblasts (MEFs) MAVS-PEX cells and MEFs MAVS-KO cells (described in (Dixit et al., 2010), MEFs MAVS-MITO (described below) and human embryonic kidney (HEK) 293T cells (kindly provided by Dr M. J. Amorim, Instituto Gulbenkian para a Ciência, Portugal) were cultured in Dulbecco’s modified Eagle’s medium supplemented with 100 U/ml penicillin, 100 mg/ml streptomycin and 10% fetal bovine serum (all from GIBCO, Thermo Scientific, Waltham, MA, United States) at 37°C in a humidified atmosphere of 5% CO2. MEFs MAVS-PEX cells, MEFs MAVS-MITO and MEFs MAVS-KO cells were transfected with Lipofectamine 3,000 (Invitrogen, Waltham, MA, United States) or microporated with Neon^® Transfection System (Invitrogen, Waltham, MA, United States) (1700 V, width: 20, 1 pulse), following manufacturer’s instructions. HEK293T cells were transfected using 1 mg/ml of Polyethylenimine (PEI, Linear, MW 25000, Polysciences, PA, United States) at a ratio of 1:6 (DNA:PEI). Cells were fixed for organelle morphology or harvested for western blot or co-immunoprecipitation assays, 24–72 h after transfection.

To knock-down the expression of MFF and DLP1 by RNA interference, 21-nucleotide small interfering RNA (siRNA) duplexes were transfected into MEFs MAVS-PEX and MEFs MAVS-MITO cells using Lipofectamine RNAiMAX (Invitrogen, Waltham, MA, United States) according to the manufacturer’s instructions. Control cells were treated with transfection mix without siRNAs complexes. Cells were assayed for silencing and organelle morphology 72 h after seeding. siRNA oligonucleotides were obtained as pre-designed siRNAs as follows: MFF–sense strand: 5′-CGCUGACCUGGAACAAGGAdTdT-3′ for exon 2 30 (Ambion, Austin, TX, United States); DLP1–sense strand: 5′- UCCGUGAUGAGUAUGCUUUdTdT-3′ 31 (Ambion, Austin, TX, United States).

To generate a MEFs MAVS-MITO stable cell line, MEFs MAVS-KO were transduced with retroviruses, which were first produced by transfecting HEK293T cells with pCL-Ampho and pVSV-G (provided by Dr B. Jesus, University of Aveiro, Portugal), and MSCV2.2 IRES-GFP MAVS-MITO. Twenty-four hours upon transfection, cell media was renewed and 24 h later cell media was collected, filtered and added to MEFs MAVS KO cells, plated 24 h before. Transduced cells were left to grow until full confluence before being split and sorted for low GFP expression level using BD FACSARIA II.

HEK293T cells were infected with SeV (Cantell strain, Charles River Laboratories, Wilmington, MA, United States) with a final concentration of 100 HA units/ml, diluted in serum- and antibiotic- free media. Cells were incubated for 1 h at 37°C and, afterwards growth media containing 20% of FBS was added to cells. Infection continued for 14 h before cells collection.

Cells grown on glass coverslips were fixed with 4% paraformaldehyde in PBS, pH 7.4, for 20 min, permeabilized with 0.2% Triton X-100, for 10 min, blocked with 1% BSA solution, for 10 min, and incubated with the indicated primary and secondary antibodies, for 1 h at room temperature in a humid environment. Cells were then stained with Hoechst (1:2000) for 3 min, before mounting the slide. Confocal images were acquired using a Zeiss LSM 880 confocal microscope (Carl Zeiss, Oberkochen, Germany) using a Plan- Apochromat 63× and 100×/1.4 NA oil objectives, a 561 nm DPSS laser and the argon laser line 488 nm (BP 505–550 and 595–750 nm filters). Images were processed using ZEN Black and ZEN Blue software (Carl Zeiss, Oberkochen, Germany). Digital images were optimized for contrast and brightness using Adobe Photoshop (Adobe Systems, San Jose, CA, United States).

For the evaluation of organelles morphology, around six hundred cells from three independent experiments were counted for each condition, considering the size/shape and number of their peroxisomes or mitochondria. For these analyses, cells were considered as containing “fragmented organelles” when organelles were significantly smaller and in higher number than the ones from the control cells. We considered cells containing “elongated organelles” as those whose organelles had a tubular shape and were significantly longer when compared to the control cells.

To study the interaction between STING and vMIA, MEFs MAVS-PEX cells were co-transfected with vMIA-Myc and STING-FLAG by Lipofectamine 3,000 (Invitrogen, Waltham, MA, United States). Lysates were incubated with anti-MAVS antibody for 2 h at 4°C on a rotary mixer. Then, 50 µL of beads were added to the mixture and rotated for 10 min at room temperature. The complex was washed 3 times with PBS containing 0.1% Tween20 and then resuspended in 3x SDS-sample buffer and boiled for 10 min to elute bound proteins. Untransfected MEFs MAVS-PEX cells were used as negative control for each immunoprecipitation. In all immunoprecipitations, 50 µg of total cell lysate was used as input, and for the output the same volume of input was saved from the cell lysate extracted after incubation with the antibody and beads.

Cells lysates were obtained by using a specific lysis buffer (25 mM Tris-HCl pH 7.5, pH 8.0, 50 mM sodium chloride, 0.5% sodium deoxycholate, 0.5% Triton X-100 and supplemented with a protease-inhibitor mix) and by passing them 20 times through a 26-gauge syringe needle. Then, samples were incubated on a rotary mixer for 30 min at 4°C, before being cleared by centrifugation (17,000 x g, 15 min, 4°C). Protein concentration was determined using the Bradford assay (Bio-Rad Protein Assay, Bio-Rad, Hercules, CA, United States). Protein samples were separated by SDS-PAGE on 10% or 12.5% polyacrylamide gels, transferred to nitrocellulose (PROTAN^®, Whatman^®, Dassel, Germany) using a semidry apparatus or wet transfer system (Bio-Rad, Hercules, CA, United States), and analysed by immunoblotting.

Immunoblots were processed after blocking membranes with 5% milk (Molico Skimmed dry milk powder, Nestlé, Vevey, Switzerland) in TBS-T (10 mM Tris pH 8, 150 mM, 0.005% Tween20) and using specific primary antibodies and secondary antibodies diluted in TBS-T. Between incubations, membranes were washed 3 times for 5 min in TBS-T. Immunoblots were scanned with a Bio-Rad GS-800 calibrated imaging densitometer or ChemiDoc™ Touch Gel Imaging System (Bio-Rad, Hercules, CA, United States) for chemiluminescence detection, while Li-COR Odyssey imaging system for fluorescence detection (LI-COR Biotechonology, Cambridge, UK). Images were processed using Quantity One software (Bio-Rad, Hercules, CA, United States) or Image Studio Lite 5.2 (LI-COR Biotechonology, Cambridge, UK). Bands’ quantification was done using the volume tools from Quantity One software (Bio-Rad, Hercules, CA, United States), where the background intensity was calculated using the local background subtraction method.

Twenty-four hours after cells transfection, total RNA was isolated using Nzyol (NZYTech, Lisbon, PT), following manufacturer’s protocol. After quantifying RNA with DS-11 spectrophotometer (DeNovix Inc., Wilmington, DE, United States), 1 µg of total RNA was treated with 1 µL DNase I (Thermo Scientific, Waltham, MA, United States). cDNA was produced from treated RNA using Revert Aid Reverse Transcriptase (Thermo Scientific, Waltham, MA, United States) and Oligo-dT15 primer (Eurofins Genomics, Ebersberg, Germany) following manufacturer’s protocol. For quantitative real-time polymerase chain reaction, 2 μL of 1:10 diluted cDNA was added to 10 μL of 2x SYBR Green qPCR Master Mix (Low Rox) (Bimake, Houston, TX, United States). The final concentration of each primer was 250 nM in 20 μL of total master mix volume. Duplicates of each sample were done, and reactions were run on 7,500 Real-Time PCR System (Applied Biosystems, Waltham, MA, United States). Primer sequences were designed using Beacon Designer 7 (Premier Biosoft, Palo Alto, CA, United States) for IRF1, RSAD2 and GAPDH mouse genes. The oligonucleotides used for mouse IRF1 were 5′-GGT​CAG​GAC​TTG​GAT​ATG​GAA-3′ and 5′-AGT​GGT​GCT​ATC​TGG​TAT​AAT​GT-3’; for mouse RSAD2 were 5′-TGT​GAG​CAT​AGT​GAG​CAA​TGG-3′ and 5′- TGT​CGC​AGG​AGA​TAG​CAA​GA-3’; for mouse GAPDH were 5′-AGT​ATG​TCG​TGG​AGT​CTA-3′ and 5′-CAA​TCT​TGA​GTG​AGT​TGT​C-3’ (Eurofins Genomics, Ebersberg, Germany) GAPDH was used as a reference gene. The thermocycling reaction was done by heating at 95°C for 3 min, followed by 40 cycles of a 12 s denaturation step at 95°C and a 30 s annealing/elongation step at 60°C. The fluorescence was measured after the extension step using the Applied Biosystems software (Applied Biosystems, Waltham, MA, United States). After the thermocycling reaction, the melting step was performed with slow heating, starting at 60°C and with a rate of 1%, up to 95°C, with continuous measurement of fluorescence. Data analysis was performed using the 2^−ΔΔCT method.

HEK 293T cellular fractionation was performed by homogenizing cells in Buffer A (10 mM Tris-HCl pH 7.5, 10 mM KCl, 1.5 mM MgCl2, 0.25 M D-mannitol, supplemented with cOmplete™, EDTA-free Protease Inhibitor Cocktail (Roche, Basel, Switzerland) and passing samples gently through a 26-gauge syringe needle. The homogenate was cleared of nuclei and membranes by centrifugation at 1,000 g for 5 min at 4°C. Mitochondria-enriched fraction pellet was obtained by centrifugation at 2000 g for 10 min at 4°C. The supernatant was then centrifuged again at 25,000 g for 25 min at 4°C to obtain the peroxisome-enriched fraction pellet. Both pellets were gently resuspended in homogenization buffer supplemented with 2% n-Dodecyl β-D-maltoside, to disrupt the organelle membrane without affecting protein’s quaternary structure.

Organelle’s fractions were then processed for the separation of MAVS oligomers by sucrose gradient (as described in (Seth et al., 2005) with minor adaptations). The organelle’s-enriched fractions were loaded in 30–60% sucrose gradients and centrifuged at 170 000 g for 2 h at 4°C. Starting from the top, 7 equal fractions were collected and processed for SDS-PAGE and analysed by immunoblotting.

Statistical analysis was performed in Graph Pad Prism 9 (GraphPad Software, Inc., La Jolla, CA, United States). Data are presented as mean ± standard error mean (SEM). Differences among groups were analysed by one-way ANOVA, followed by Bonferroni’s multiple comparison test; comparisons between two groups were made using unpaired T test, and p-values of <0.05 were considered as significant (****–p < 0.0001, ***–p < 0.001, **–p < 0.01, *–p < 0.05 and ns–non-significant).

Although the contribution of peroxisomes, in concert with mitochondria, to the cellular antiviral response has been established (Dixit et al., 2010), the key differences between the signalling pathways, originating from these two organelles, with distinct kinetics and end products, remains unknown. HCMV vMIA has been shown to inhibit both pathways and, in parallel, induce the fragmentation of both organelles (Castanier et al., 2010; Magalhães et al., 2016). This morphology change has been pinpointed as the main trigger for the vMIA-dependent antiviral signalling inhibition at mitochondria, as it would consequently reduce its association with the ER, hindering the interaction between MAVS and STING (Castanier et al., 2010). At peroxisomes, however, the antiviral signalling inhibition induced by vMIA was shown to be independent of the organelle fragmentation (Magalhães et al., 2016).

These observed differences led us to further investigate the mechanism by which vMIA induces the morphology changes at peroxisomes, in comparison with mitochondria. To that end, we evaluated its dependence on the organelles’ fission machinery, more specifically on the cytoplasmic protein DLP1, the main responsible for the final membrane fission. vMIA’s ability to induce peroxisome fragmentation was analysed upon overexpression of Myc-tagged vMIA (vMIA-Myc) and silencing of DLP1 (via small interference RNA (siRNAs) against DLP1, (siDLP1)) in mouse embryonic fibroblasts (MEFs) that contain MAVS solely at peroxisomes (MEFs MAVS-PEX (Dixit et al., 2010)) (Figure 1) Upon immunolocalization with antibodies against the Myc-tag and the peroxisomal marker PMP70, the cells were examined by confocal microscopy (Figure 1A) and the organelles’ morphological alterations were quantified (Figure 1B,C). As expected, silencing of DLP1 induced peroxisomal elongation when compared with control cells (Figure 1Aa,b and Supplementary Figure S1A). Interestingly, vMIA was no longer able to induce peroxisome fragmentation in the absence of DLP1, as no significant differences in morphology were observed in the presence of the viral protein (Figures 1Ag–i,B). We have also analysed the ability of vMIA to induce peroxisome fragmentation in the absence of MFF, another major player on peroxisomal fission and one of the main anchors of DLP1 at the organelle’s membrane. Upon analysis of MEFs MAVS PEX cells containing siMFF and vMIA-Myc (Figures 1Aj–l,C and Supplementary Figure S1A), no significant changes in peroxisome morphology were detected, when compared to the elongated peroxisomes observed upon silencing of MFF (Figures 1Ac,C). These results clearly show that vMIA depends on a fully functional peroxisome fission machinery to be able to induce the organelle’s fragmentation.

In order to analyse whether an analogous dependence occurs at mitochondria, we performed similar analyses in MEFs MAVS-MITO cells (where MAVS is present solely at mitochondria). Mitochondria morphology was analysed by confocal microscopy upon immunolocalization with antibodies against Myc and the mitochondrial marker TIM23, in cells expressing vMIA-Myc in the presence or absence of DLP1 or MFF (siDLP1 or siMFF) (Figure 2). Surprisingly, vMIA was still able to induce mitochondria fragmentation in the absence of either DLP1 or MFF, totally reverting the organelle elongation observed upon silencing of any of these proteins (Figures 2Ag–l,B,C, Supplementary Figure S1B). These results were also observed upon analysis of mitochondria morphology in MEFs MAVS PEX cells, in the absence of DLP1 or MFF (Supplementary Figure S1C Figures 3Ag–l,D,E). This data indicates that, contrarily to peroxisomes, vMIA is able to interfere with mitochondria morphology in a fission machinery-independent manner.

The independence on the organelle morphology for the vMIA-induced antiviral signalling inhibition at peroxisomes (Magalhães et al., 2016), contrarily to what was observed at mitochondria (Castanier et al., 2010), suggests that these two processes may be less correlated than initially anticipated. In order to conclude on whether the peroxisomal and mitochondrial fragmentation prompted by vMIA is in any way dependent on MAVS signalling, we analysed the organelles’ morphology in cells lacking MAVS and in the presence of the viral protein. To that end, MEFs MAVS-KO cells were transfected with vMIA-Myc, and confocal microscopy analyses were performed upon immunolocalization with antibodies against PMP70 or TIM23. As shown in Figure 3A, even in the absence of MAVS, vMIA was able to induce a strong fragmentation of both peroxisomes (Figure 3B) and mitochondria (Figure 3C). These results demonstrate that the organelles morphology and the mechanism of vMIA-mediated inhibition of the MAVS signalling are less related than anticipated.

Nevertheless, as the mitochondria fragmentation induced by vMIA has been associated to a decrease in MAVS-STING interaction and the consequent inhibition of the immune response (Castanier et al., 2010), we analysed whether the same would be observed upon peroxisomal fragmentation. We started by analysing the occurrence of an interaction between the peroxisomal MAVS and STING, as this had not yet been demonstrated. To that end, we transfected MEFs MAVS-PEX with a FLAG-tagged STING (STING-FLAG) and, 24 h after, performed co-immunoprecipitation analyses with a pull-down against MAVS. As shown in Figure 3D, STING interacts with the peroxisomal MAVS. In order to investigate whether this interaction is eventually compromised by vMIA, these experiments were also performed upon co-transfection with vMIA-Myc. As shown in Figures 3D,E, the presence of the viral protein does not inhibit the interaction between STING and peroxisomal MAVS, contrarily to what had been observed at mitochondria. These results, together with the distinct dependencies on organelle morphology, point out major differences between the mechanisms involved in the modulation of the peroxisomal and mitochondrial antiviral signalling by vMIA.

To further unravel the mechanism by which vMIA inhibits the MAVS signalling at peroxisomes, we once more evaluated the importance of the fission machinery proteins DLP1 and MFF. Upon silencing of any of these proteins in MEFs MAVS-PEX cells, vMIA-Myc was transfected and, 24 h after, the MAVS-dependent antiviral signalling was stimulated by overexpressing a constitutively active form of RIG-I (GFP-RIG-I-CARD, as in (Yoneyama et al., 2004; Magalhães et al., 2016). In correlation with our previous results, the absence of DLP1 did not hinder the peroxisomal antiviral response, leading to similar expression levels of ISGs (IRF1 and RSAD2) when comparing with the ones obtained in the stimulated control cells (Figure 4A). However, our results clearly show that, at peroxisomes, MFF’s absence strongly impairs the vMIA-mediated inhibition of IRF1 and RSAD2 mRNA expression (Figure 4A). MFF, hence, rises as the first identified peroxisomal protein to play an important role on the mechanism of vMIA-mediated inhibition of the peroxisome-dependent antiviral response.

In order to determine whether the same dependency is observed at mitochondria, similar experiments were performed in MEFs MAVS-MITO cells. Surprisingly, in these cells, neither the lack of DLP1 nor MFF had significant effects on ISGs expression (Figure 4B).

The importance of MFF for the antiviral signalling inhibition mediated by vMIA seems to be, hence, exclusive for the peroxisomal response.

One of the crucial steps of the cellular RIG-I/MAVS antiviral signalling, is the oligomerization of MAVS at the peroxisomal and mitochondrial membranes upon interaction with RIG-I, forming functional high molecular weight prion-like aggregates, essential for the activation of the downstream proteins (Seth et al., 2005; Hou et al., 2011). To further unravel the mechanism of action of vMIA towards the peroxisomal MAVS signalling, and as we had previously observed that this viral protein interacts with MAVS, we analysed its effect on MAVS oligomerization upon infection with Sendai virus (SeV) in HEK293T cells. As these cells contain MAVS at both peroxisomes and mitochondria, we used differential centrifugation to prepare peroxisome-enriched fractions and analyse solely the oligomerization of peroxisomal MAVS. Gradient assay experiments were then implemented to separate peroxisomal MAVS in different density fractions, depending on its oligomerization degree. As shown in Figure 5A, upon infection with SeV, confirmed by the increase in p-STAT1 (Figure 5E), peroxisomal MAVS oligomers appear at the higher density fraction. Interestingly, this is no longer observed in the presence of vMIA (Figure 5A). These results suggest that this viral protein, likely through direct interaction (previously reported in (Magalhães et al., 2016)), prevents the formation of MAVS oligomers, inhibiting the downstream antiviral signalling (Figures 5A,E).

In order to unravel whether this oligomerization inhibition also occurs at the mitochondrial membranes, we have performed similar analysis on mitochondria-enriched fractions isolated by differential centrifugation. As shown in Figure 5B, the results were in all similar to the ones obtained in peroxisomes, demonstrating that, vMIA is also able to inhibit mitochondrial MAVS oligomerization and, consequently, the downstream antiviral signalling (Figures 5B,E).

The dependency on MFF for the vMIA influence towards the peroxisomal antiviral response, led us to investigate whether this protein would be involved in the observed inhibition of MAVS oligomerization. Similar gradient separation analysis of peroxisome-enriched fractions was performed in MFF silenced (via siMFF transfection) HEK293T cells infected with SeV. Interestingly, our results show that, in the absence of MFF, vMIA is no longer capable of inhibiting the formation of peroxisomal MAVS oligomers, and hence the downstream antiviral signalling (Figures 5C,E).

On the other hand, similar experiments performed in mitochondrial-enriched fractions of HEK293T cells infected with SeV, indicate that vMIA is still able to inhibit MAVS oligomerization in the absence of MFF and, hence, hinder the downstream antiviral signalling (Figures 5D,E). These results correlate to our previous observation that MFF is not essential for the vMIA-mediated inhibition of the antiviral signalling at mitochondria.

Altogether our results demonstrate that vMIA inhibits MAVS oligomerization at peroxisomes in an MFF-dependent manner and highlight important differences between its mechanisms of action towards both organelles.