Multiple Viral microRNAs Regulate Interferon Release and Signaling Early during Infection with Epstein-Barr Virus

Acute antiviral functions of all nucleated cells rely on type I interferon (IFN-I) pathways triggered upon viral infection. Host responses encompass the sensing of incoming viruses, the activation of specific transcription factors that induce the transcription of IFN-I genes, the secretion of different IFN-I types and their recognition by the heterodimeric IFN-α/β receptor, the subsequent activation of JAK/STAT signaling pathways, and, finally, the transcription of many IFN-stimulated genes (ISGs).

been reported to express viral gene products upon contact with EBV virions but cannot be transformed in vitro (Gujer et al., 2019 and references therein). Apart from viral proteins, the EBV genome encodes different classes of non-coding RNAs including two long non-coding RNAs, EBV-encoded small RNAs 1 and 2 (EBER1 and 2), circular RNAs (Ungerleider et al., 2018), and 44 mature microRNAs (miRNAs) distributed in clusters along the genome (Fig. 1).
In order to conduct a successful infection, viruses including EBV must evade or counteract the activation of type I interferons (IFN-I). IFN-I production and signaling is a two part system at the center of innate anti-viral immunity. Initially, specialized pattern recognition receptors (PRRs) detect pathogen-associated molecular patterns (PAMPs), which are distinct pathogen structures otherwise absent from non-infected, healthy cells. PRR engagement triggers activation of specific transcription factors which, in turn, induce transcription of IFN-I genes (13 IFN-α subtypes and IFN-β in humans), leading to production and secretion of type I IFNs. Secreted IFN-I binds to a single, heterodimeric IFN-α/β receptor on mammalian cells, thereby activating JAK/STAT signaling pathways and culminating with the transcription of many IFN-stimulated genes (ISGs). These gene products altogether establish a so-called antiviral state, thereby restricting the viral life cycle and/or orchestrating viral clearance in infected and neighbouring cells.
Due to their peculiar faculty to secrete massive amounts of IFN-I, plasmacytoid dendritic cells (pDCs) are a major factor in the IFN-I system. pDCs constitutively express major components of the IFN-I activation pathway like the TLR7 and TLR9 PRRs and the interferon regulatory factor 7 (IRF7) transcription factor. TLR7 and TLR9 are embedded in the endosomal membranes and respectively recognize single-stranded (ss)RNA or unmethylated CpG DNA in the endosomal lumen. Upon binding of PRRs, TLR7 and TLR9 signal through the MyD88 adaptor protein and activate the IRF7 transcription factor. There are hints that EBV is detected by pDCs, which respond to infection by secreting type I IFNs (Lim et al., 2007;Fiola et al., 2010;Quan et al., 2010;Severa et al., 2013). Whether the virus infects human pDCs and the mechanisms by which pDCs detect EBV are not fully understood.
It has been shown that the EBERs, which are abundant viral non-coding RNAs, can be recognized as PAMPs by the RIG-I, TLR3 and TLR7 PRRs and that pDCs further sense the presence of EBV through recognition of unmethylated viral genomic DNA by TLR9 (Fiola et al., 2010). Multiple intracellular sensors recognize exogenous cytosolic DNA, including DNAdependent activator of IFN regulator factors (DAI) (Takaoka et al., 2007), DDX41 (Zhang et al., 2011), absent in melanoma 2 (AIM2) (Hornung et al., 2009), LSm14A (Li et al., 2012), IFN-gamma inducible factor 16 (IFI16) (Unterholzner et al., 2010), and cyclic GMP-AMP (cGAMP) synthase (cGAS) (Sun et al., 2013) which activates the adaptor protein STING (stimulator of IFN genes) to trigger IFN signaling. Human B cells express IFI16, cGAS, and downstream signaling components necessary to induce type I IFN; however, it has been demonstrated that B lymphocytes fail to recognize the presence of genomic EBV DNA through cGAS-STING (Gram et al., 2017).
EBV encodes several proteins that have been described as type I IFN antagonists and have attributed roles in counteracting anti-viral responses. In screening 150 EBV ORFs, EBV LF2, encoding a tegument protein, was identified as an inhibitor of IRF7 dimerization (Wu et al., 2009). BZLF1 also interferes with dimerization of IRF7 (Hahn et al., 2005) and blocks STAT1 tyrosine phosphorylation (Morrison et al., 2001) while BRLF1 controls both IRF3 and IRF7 (Bentz et al., 2010). The EBV kinase BGLF4 impedes IRF3 activity through direct interactions, consequently attenuating type I IFN signaling (Wang et al., 2009). Other viral proteins, such as BGLF5 (van Gent et al., 2011) inhibit production of TLR2 and TLR9.
In this study, we examined the relationship between EBV non-coding RNAs and activation of IFN response pathways. Here, we document that the EBV-encoded miRNAs contribute to the regulation of type I IFN response upon EBV infection, whereas EBERs and LF2 have a neglectable effect in infected human primary B cells. We identified several BART miRNAs which regulate genes involved in the IFN secretion and IFN response pathway. The presence of viral DNA was essential to induce the IFN-α secretion during EBV infection in a TLR9 dependent manner in PBMCs. In a newly established gp350:BlaM fusion assay, we verified that EBV virions enter a subset of pDCs and determined that these infected pDCs are the primary producers of type I IFNs in PBMCs.

Cloning of a reconstituted wild-type EBV strain and mutant derivatives
To study the functions of EBV-encoded miRNAs, we constructed a recombinant EBV that expresses all viral miRNAs at their physiological levels and that can be genetically manipulated in bacteria to generate mutant derivatives. To do so, we used the wt/B95.8 (2089) recombinant virus (Fig. 1) available in our laboratory (Delecluse et al., 1998). This recombinant is based on the B95-8 strain of EBV (Miller and Lipman, 1973) into which a DNA fragment encoding a GFP gene (used for titration), a gene encoding resistance against hygromycin (used for selection in the EBV producer cells), a chloramphenicol acetyltransferase gene (used for selection in bacteria), and the mini-F factor replicon (enabling maintenance in bacteria) have been introduced (Delecluse et al., 1998). We further introduced a DNA fragment from the M-ABA field strain to restore viral genes that are deleted in EBV B95-8 (Bornkamm et al., 1980;Raab-Traub et al., 1980). This repaired B95-8 strain derivative, termed r_wt/B95.8 (6008) (Fig. 1) (Pich et al., 2019), encodes all known EBV miRNAs which are expressed from their genuine promoters at physiological levels. In addition, r_wt/B95.8 also carries the second copy of the lytic origin of DNA replication, oriLyt, (Hammerschmidt and Sugden, 1988) and expresses the LF1, LF2 and LF3 viral proteins from their coding sequences which are also absent in the genome of the B95-8 EBV strain (Supplementary Tab. 1).
Subsequently, we replaced the viral miRNA sequences in r_wt/B95.8 by scrambled sequences similar to our previous approach (Seto et al., 2010) preventing the expression of all viral miRNAs. This strain is called r_ Δ miR ( Fig. 1) (Pich et al., 2019).
The two viral non-coding EBER (Epstein-Barr virus-encoded small RNA) RNAs of 167 and 172 nucleotides in length have been implicated in the activation of type I IFN responses upon EBV infection in B cells (Samanta et al., 2006;Iwakiri et al., 2009). To evaluate their effects on type I IFN activation in our infection model, we deleted their coding sequences in r_wt/B95.8 and r_ΔmiR (Fig. 1). These two strains are called Δ EBER and Δ EBER/ΔmiR, respectively (Pich et al., 2019).
Finally, we mutated the gene encoding the LF2 viral protein that has been described as a type I IFN antagonist (Wu et al., 2009). We introduced a stop codon in the coding sequence of the LF2 gene to prevent its expression in r_wt/B95.8. This EBV strain is termed Δ LF2 (Supplementary Tab. 1).
To conclude, we present here a fully reconstituted EBV strain based on the B95-8 EBV genome. This recombinant EBV genome can be used to produce infectious virus, it can be 6 conveniently genetically manipulated and presumably expresses all viral genes including all miRNAs at their physiological levels from their authentic promoters supported by their regulatory elements. We assume that, despite its chimeric composition, this new strain represents a more physiological model than the widely used B95-8 laboratory strain.

Cytokines secretion by B lymphocytes infected by EBV mutants
We subsequently sought to determine the effects of the different mutations in the reconstituted r_wt/B95.8 EBV reference strain on the secretion of cytokines released from infected human B lymphocytes. Primary human B lymphocytes were infected in vitro with the five virus strains described above (r_wt/B95.8; r_ΔmiR; infection with every virus strain tested, B cells grew in size and started to proliferate. Five days post-infection, infected cells were counted by flow cytometry and seeded at equal densities in fresh medium. Four days later, supernatants were collected and concentrations of IL-12, IL-6, IL-10, and IFN-α were measured by ELISA (Fig. 2).
We showed previously that EBV-encoded miRNAs regulate the expression of human IL-12 . As shown in Figure 2A, B lymphocytes infected with r_ΔmiR secreted more IL-12 than cells infected with r_wt/B95.8, consistent with prior experiments using the B95-8 strain wt/B95.8 (2089) (Fig. 1)  . The deletion of the EBERs or LF2 did not influence IL-12 secretion. Moreover, when the cells were infected with the double mutant Δ EBER/ΔmiR, IL-12 was secreted at concentrations similar to r_ΔmiR infected B cells. These data confirm that the EBV-encoded miRNAs are responsible for altered IL-12 levels in newly infected B lymphocytes while the EBERs and LF2 do not affect IL-12 production.
We have further documented that EBV-encoded miRNAs inhibit the secretion of IL-6 but not IL-10 in the B95-8 context . As shown in Figure 2B, the absence of the viral miRNAs in cells infected with r_ΔmiR led to a substantial increase in IL-6 secretion when compared to r_wt/B95.8-infected cells. This finding demonstrates the importance of the viral miRNAs to regulate IL-6 secretion upon EBV infection. Interestingly, it has been suggested that EBER2 also plays a role in IL-6 secretion (Wu et al., 2007). We did not observe any difference in IL-6 secretion by cells infected with r_wt/B95.8 or Δ EBER, but the Δ EBER/ΔmiR-infected cells did secrete less IL-6 than the r_ΔmiR-infected cells (Fig. 2B).
Results from our infection model suggest that EBERs may stimulate the expression of IL-6 to some extent, but the viral miRNAs largely repress IL-6 secretion. As for IL-12 secretion, Δ LF2 behaved very similar to r_wt/B95.8 (Fig. 2B).
Concerning IL-10 secretion, none of the mutations clearly altered the cytokine secretion by infected B lymphocytes (Fig. 2C). It has to be noted that our ELISA quantification cannot 7 differentiate between human and viral IL-10 encoded by EBV (Moore et al., 1990). Thus, we cannot rule out that human IL-10 expression is regulated by viral miRNAs, EBERs, or LF2 but is compensated by the expression of viral IL-10 encoded by the viral BCRF1 gene.
We next asked whether mutations in the EBV mutant derivatives would impact the secretion of IFN-α by EBV-infected B lymphocytes. As shown in Figure 2D, the absence of viral miRNAs led to increased secretion of IFN-α by r_ΔmiR-infected cells compared to r_wt/B95.8-infected cells. Surprisingly, we did not observe any decrease in secreted IFN-α when comparing Δ EBER-to r_wt/B95.8-infected cells and Δ EBER/ΔmiR-to Δ miR-infected cells. These data are puzzling since EBERs have been reported to interact with the PRR RIG-I and act as type I IFN inducers (Samanta et al., 2006;Samanta et al., 2008;Iwakiri et al., 2009;Iwakiri, 2014;Duan et al., 2015). Additionally, Δ LF2-infected cells did not show any significant difference compared to r_wt/B95.8-infected cells despite the fact that LF2 has been shown to interact with IRF7 and block the expression of type I IFN (Wu et al., 2009).
These discrepancies could be explained by differences in the experimental conditions. What we can ascertain is that, in our experiments with primary cells, the EBERs are not necessary for the expression and secretion of IFN-α by EBV-infected B lymphocytes. Moreover, independent of the presence or absence of the EBERs, the viral miRNAs significantly impact the secretion of IFN-α.
Taken together our data indicate that EBV-encoded miRNAs prevent the secretion of IL-12, IL-6 and IFN-α in newly infected primary human B cells. Whereas the EBERs seem to be partially responsible for the activation of IL-6, their deletion did not cause any significant difference in the secretion of IL-12, IL-10 or IFN-α. We also failed to observe a contribution of LF2 to the secretion of the cytokines analyzed.

Identification of cellular transcripts regulated by EBV-encoded miRNAs
Based upon the phenotypes we observed above, we hypothesized that EBV-encoded miRNAs regulate the expression of host cellular genes involved in multiple aspects of the IFN pathways, such as PAMP sensing, production of type I IFN, and/or induction of secondary cytokine signals that encompass innate anti-viral responses. To initially investigate targets of EBV miRNAs that may be involved in these pathways, we first used in silico prediction tools (Garcia et al., 2011) to scan human 3'UTRs for the presence of canonical 7mer and 8mer miRNA seed-match sites. Candidates were cross-referenced with the literature to determine putative targets associated with PRR activation and type I IFN signaling. Among these, we identified potential binding sites for EBV miRNAs in the 3'UTRs of DDX58/RIG-I, RSAD2/Viperin,OAS2,and FYN. 8 To biochemically define EBV miRNA binding sites in human protein-coding transcripts, we performed PAR-CLIP experiments on latently infected DLBCL cell lines (IBL1, IBL4, and BCKN1) which express the full spectrum of EBV BHRF1 and BART miRNAs. Cells were cultured in the presence of 4-thiouridine to label RNAs as previously described (Skalsky et al., 2012). RISC-associated RNAs were immunopurified with antibodies to Argonaute (Ago) proteins and subjected to high-throughput sequencing. Following alignment to a human reference genome, reads were analyzed using two CLIP-seq pipelines (PARalyzer and PIPE-CLIP) to define high resolution Ago interaction sites Chen et al., 2014). To identify host mRNAs reproducibly targeted by EBV miRNAs, we expanded our analysis to include previously published Ago PAR-CLIP datasets from EBV B95-8 and wildtype LCLs and EBV/KSHV+ PEL (Supplementary Tab. 2, Supplementary Fig. 1) (Al Tabaa et al., 2009;Gottwein et al., 2011;Skalsky et al., 2012;Majoros et al., 2013;Skalsky et al., 2014). Derived Ago interaction sites mapping to 3'UTRs of protein coding transcripts were scanned for canonical EBV miRNA seed matches (>=7mer1A). In total, we identified 4,010 individual genes regulated at the 3'UTR level by EBV miRNAs, with 340 genes common to all EBV-infected B cell types investigated (DLBCLs, LCLs, PEL) ( Supplementary Fig. 1E).
To focus on EBV miRNA targets functionally relevant to type I IFN pathways, we then interrogated Ago PAR-CLIP data using Ingenuity Pathway Analysis (IPA, Qiagen). 3,976 of the 4,010 3'UTR target genes were mapped by IPA. Enriched signaling pathways included 'JAK/Stat Signaling' and 'Role of PKR in Interferon Induction and Antiviral Response' (Supplementary Fig. 1F). Among the target genes related to IFN pathways, we identified JAK1, IRF9, STAT1, IRAK2, IKBKB, and interferon receptors (IFNAR1 and IFNAR2).
Together, these data reinforce the observation that EBV miRNAs play a major role in regulating innate immune response pathways.

EBV-encoded miRNAs regulate genes involved in the induction of type I IFN
To confirm miRNA interactions, we first selected targets with relevance to induction of type I IFN (DDX58/RIG-I, RSAD2/Viperin, OAS2, FYN, and IRAK2) and cloned the 3'UTRs of these genes into reporter plasmids. Reporter plasmids were co-transfected into 293T cells with or without an EBV miRNA-expressing plasmid and luciferase expression was assessed (Fig. 3).
When we observed miRNA mediated inhibition of luciferase expression, we mutated the predicted seed-sequences in the reporter plasmid and verified that the miRNA expression did not have an effect on the expression of luciferase ( Supplementary Fig. 2).
It has recently been reported that another EBV miRNA, miR-BART6, acts as a regulator of RIG-I expression and prevents the expression of type I IFN (Lu et al., 2017). Our findings 9 confirm that EBV-encoded miRNAs control the expression of the RIG-I PRR. As the virus has evolved mechanisms to attenuate RIG-I expression, this suggests that RIG-I is involved in efficiently detecting EBV infection. Interestingly, even though the EBERs have so far been described as the primary virus moiety activating RIG-I, our data suggest that there are viral PAMPs recognized by the RIG-I helicase other than EBERs since the Δ EBER-infected cells still secrete IFN-α at levels similar to its wildtype predecessor r_wt/B95.8 (Fig. 2D).
As shown in panels B to E of Fig. 3, we found EBV-encoded miRNAs to also regulate the 3'UTRs of IRAK2, RSAD2/Viperin, FYN, and OAS2. These genes have been described as important factors for the production of type I IFN as well as activation of human pDCs. IFNinducible RSAD2/Viperin promotes IFN-α secretion upon TLR7 and TLR9 activation (Saitoh et al., 2011), while IRAK2 is essential for the expression of type I IFN initiated by TLR7 in pDCs (Kawagoe et al., 2008;Flannery et al., 2011;Wang et al., 2015). A study also showed the importance of a functional IRAK2 for the late stage of cytokines production including IFNα in murine pDCs (Pauls et al., 2013). Modulating the expression of IRAK2 by EBV miRNAs could thus contribute to regulating the release of IFN-α by pDCs via the inhibition of the TLR7 pathway.
OAS2 (2'-5'-oligoadenylate synthetase 2) is an ISG that specifically recognizes dsRNA structures and synthesizes 2'-5'-oligoadenylate structures that activate RNAseL. RNaseL subsequently degrades cellular and viral RNAs and thus inhibits translation in the infected cell. We show that miR-BART1 can regulate the 3'UTR of OAS2 in Figure 3E; however, compared to RSAD2/Viperin, TLR9, TLR7, RIG-I and FYN protein levels which inversely correlated with EBV miRNA presence, OAS2 protein levels were barely upregulated in r_ΔmiR-infected cells (Fig. 3F). Together, these data show that EBV-encoded miRNAs inhibit expression of multiple cellular factors involved in activation, expression, and secretion of type I IFN.

EBV-encoded miRNAs regulate the response to type I IFN signaling
Upon secretion, type I IFN interacts with the heterodimeric IFN-α/β receptor (IFNAR) and initiates a signaling cascade through the JAK-STAT pathway. To determine if EBV-encoded miRNAs impact type I IFN responses in addition to inhibiting IFN production, we used a reporter luciferase system whereby type I IFN was added exogenously in the presence of individual EBV miRNAs. We constructed a luciferase reporter plasmid p6898 with the improved firefly luciferase (luc2) gene under control of a chimeric promoter comprising part of the ISG Mx2 promoter and a repetition of five canonical ISREs (Interferon-Stimulated Response Element). A renilla luciferase gene under the control of the weak TK promoter was used as an internal transfection control for subsequent data normalization. We concomitantly 1 0 transfected this vector into 293T cells with individual expression vectors for EBV-encoding miRNAs or an empty vector as a control. 24 h post-transfection, cells were treated with type I IFN and luciferase signals were quantified one day later. As a positive control, we used a miRNA expression vectors for human miR-373 (Fig. 4A), which regulates the expression of JAK1 and IRF9 (Mukherjee et al., 2015), and is thus expected to reduce the response to exogenous IFN. We also cloned and tested the IE1 gene from human cytomegalovirus (HCMV). IE1 is a very potent inhibitor of type I IFN signaling (Piganis et al., 2011) and indeed efficiently suppressed luciferase expression in our assay (Fig. 4A).
Through this functional screen, we identified four EBV-encoded miRNAs that significantly attenuated the response to exogenous type I IFN (miR-BART1, miR-BART16, miR-BART22 and miR-BHRF1-2) as well as multiple other EBV miRNAs that moderately attenuated responses (Fig. 4A). miR-BART16 has recently been shown to regulate the expression of CREB-binding protein (CBP) and thereby inhibits the induction of ISGs (Hooykaas et al., 2017). Targeting of the CBP 3'UTR was confirmed in dual-luciferase 3'UTR reporter assays ( Fig. 4B).
To investigate cellular targets that are likely responsible for the phenotypes observed for the other EBV miRNAs, we examined miRNA targetome datasets ( Supplementary Fig. 1).
Notable targets included JAK1 and IRF9 which are directly involved in JAK/STAT signaling in response to type I IFN. IRF9 is part of the ISGF3 (Interferon-stimulated gene factor 3) complex together with phosphorylated STAT1 and STAT2 and is required to drive the expression of ISGs upon type I IFN signaling (Qureshi et al., 1995). Limiting the expression of IRF9 could be a strategy employed by EBV to prevent ISGF3 complex association and the transcription of ISGs. JAK1 is a kinase and its association with the type I IFN receptor is required to transduce the signal upon ligation of IFN on the receptor. We cloned 3'UTRs of these and other genes into dual-luciferase reporter plasmids and tested them against individual EBV miRNAs. Notably, miR-BART1 inhibited luciferase expression from the IRF9 reporter ( Fig. 4C) while miR-BART3 reduced activity of the JAK1 reporter (Fig. 4D). miR-BART3 had a moderate effect on luciferase expression in our IFN-response assay (Fig. 4A) that could be explained by this finding. We furthermore found that miR-BART2 regulates the To confirm that EBV miRNAs indeed attenuate IFN-mediated activation of the JAK-STAT pathway, we tested a STAT-responsive luciferase reporter harboring SIS-inducible elements 1 1 (SIE) in the presence of EBV miRNAs in 293T cells (Fig. 4F). Consistent with inhibition of JAK1 and IRF9, both miR-BART1 and miR-BART3 suppressed SIE reporter activity following treatment with IFN. qRT-PCR analysis of IFN-treated 293T cells further revealed that ISG54 and ISG56 transcript levels were reduced in the presence of miR-BART1 and miR-BART3 ( Supplementary Fig. 4). These data demonstrate that EBV-encoded miRNAs inhibit the expression of cellular genes involved at every step of the type I IFN pathway, including the response to type I IFN and induction of ISGs.

IFN-α secretion by PBMCs and isolated pDCs infected with mutant EBV strains
We next asked whether PBMCs infected with the designed EBV mutants secrete IFN-α. We It has been reported that plasmacytoid dendritic cells (pDCs) detect EBV infection and respond by releasing type I IFN (Lim et al., 2007;Fiola et al., 2010;Quan et al., 2010; 1 2 Severa et al., 2013). In an attempt to assess the role of viral miRNAs on the sensing of infectious EBV by pDCs, we first depleted pDCs from PBMCs. We infected pDC-depleted PBMCs with r_wt/B95.8 and r_ Δ miR using the same conditions as above and collected the culture supernatants 20 h post-infection. IFN-α release was quantified by ELISA. As shown in Figure 5B, the depletion of pDCs resulted in an almost complete loss of IFN-α secretion by the remaining cells, including B lymphocytes. This finding confirms data by Severa and colleagues (Severa et al., 2013) and supports the idea that pDCs are primarily responsible for the immediate IFN-α release when PBMCs are infected with EBV (Fig. 5A). B lymphocytes barely released IFN-α immediately after PBMC infection (Fig. 5B). Moreover, in the pre-latent phase, EBV-infected B lymphocytes secreted IFN-α, but at a much lower concentration ( Fig. 2D) compared to pDCs (Fig. 5C).
We then isolated pDCs and infected the cells with the five EBV strains. As shown in Figure   5C, r_ Δ miR-infected pDCs secreted more IFN-α than r_wt/B95.8-infected pDCs, consistent with our findings in PBMCs.
Δ EBER-and Δ LF2-infected pDCs released IFN-α similar to r_wt/B95.8-infected cells whereas Δ EBER/ΔmiR-infected pDCs again released the highest amount of IFN-α suggesting that EBERs and viral miRNAs might even work collaboratively and reduce activation of the type I IFN pathway. The absence of LF2 did not seem to cause a marked effect on the secretion of IFN-α by pDCs.
Thus far, our experiments suggest that EBV-encoded miRNAs are directly involved in the regulation of type I IFN responses, especially in regulating gene expression of different pathway components. The overall impact of EBV miRNAs on direct IFN-α release by pDCs appears to be moderate, suggesting that EBV miRNAs act to reduce amplification of the IFN response in the context of PBMCs.

IFN-α release by EBV infected cells is triggered by viral DNA and depends on TLR9 signaling
We identified pDCs as the main producers of IFN-α upon EBV infection, which is in line with published literature (Quan et al., 2010;Severa et al., 2013;Gujer et al., 2019). As described previously, pDCs express TLR7 and TLR9 which sense viral RNA and DNA, respectively, for secreting type I IFNs (Colonna et al., 2004;Reizis, 2019). These TLRs are found in the endosomal compartment and seem to be shifting between endosomes and lysosomes (Ahmad-Nejad et al., 2002). We wanted to analyze if viral DNA is necessary and sufficient for the secretion of IFN-α by PBMCs upon EBV infection. EBV DNA is very rich in CpG dinucleotides, which are non-methylated in virions ( Fig. S3 in Kalla et al., 2010). Similar to bacterial DNA, unmethylated virion DNA is an ideal pattern for recognition by TLR9.
Towards this end, we infected PBMCs with the r_ Δ miR EBV strain or incubated the cells with adjusted concentrations of virus-like particles (VLPs), which are free of viral DNA 1 3 (Hettich et al., 2006). 20 h post infection supernatants were collected and levels of IFN-α were measured. As expected, infection with r_ Δ miR led to a dose-dependent secretion of type I IFN, but PBMCs treated with VLPs did not secrete detectable levels of IFN-α (Fig. 6A, Supplementary Fig. 5). To confirm that the presentation and sensing of viral DNA by pDCs is important for type I IFN release after EBV infection, we treated EBV-infected PBMCs with chloroquine. Chloroquine is an inhibitor of lysosomal functions and blocks the activation of endosomal TLRs including TLR7 and TLR9. Chloroquine treatment of EBV infected PBMCs led to a complete block of IFN-α release compared with PBMCs infected with EBV in the absence of the lysosome inhibitor (Fig. 6B).
The synthetic oligo deoxyribonucleotide ODN2088 is an antagonist, which inhibits the TLR7/8/9 mediated responses in cells. The corresponding oligonucleotide ODN2087 is a TLR7/8 antagonist, but also acts as a TLR9 antagonist control, as it has no effect on TLR9 signaling. To find out whether the IFN secretion in EBV-infected PBMCs is dependent on TLR9 as a sensor for unmethylated CpG-DNA, we infected human PBMCs with EBV r_ Δ miR in the presence of the TLR9 inhibitor ODN2088 or its ODN2087 control oligonucleotide. The inhibition of TLR9 nearly completely blocked the IFN release of EBVinfected PBMCs (Fig. 6C). When the EBV-infected PBMCs were treated with ODN2087 control, a slight drop in IFN-α levels compared to the untreated cells was noticed (Fig. 6C).
Together, these results demonstrate that IFN-α release during EBV infection of PBMCs depends on viral DNA and its detection by TLR9.

EBV-infected pDCs are the main producers of IFN-α
As shown above, type I IFN released immediately after EBV infection of PBMCs almost exclusively originates from pDCs. Our data shown in Figure 5 suggest that viral miRNAs and probably also the two non-coding EBERs modulate the release of IFN-α. Both classes of RNAs are contained in the tegument (i.e. the space between envelope and capsid) in EBV virions and can be delivered to EBV target cells (Pegtel et al., 2010;Jochum et al., 2012;Baglio et al., 2016). IFN-α release by pDCs implies that these cells endocytose EBV particles as B cells do. For pDCs to sense virion DNA and present it to the TLR9 PPR, it is mandatory that EBV particles release their virion DNA probably when endosomes reach the degradative lysosome. In this scenario, it is unclear how viral miRNAs contained in the tegument might modulate the type I IFN response unless they can be delivered into the cytoplasm of pDCs (or any other cell type where they act in conjunction with the RNA-induced silencing complex) prior to lysosomal degradation. We therefore asked if EBV can infect pDCs as has been claimed previously (Severa et al., 2013) and developed an assay to determine the cell types with which EBV particles fuse to deliver their cargo into the cytoplasm.

4
We turned to wt/B95.8 (2089) producer cells (Delecluse et al., 1998) (Fig. 1) as they release very high amounts of infectious EBV upon lytic induction. We transfected the EBV 2089 producer cells with BZLF1 and BALF4 expression plasmids (to induce optimal virus synthesis) (Hammerschmidt and Sugden, 1988;Neuhierl et al., 2002) together with a plasmid (p7180) that expresses a carboxy-terminal fusion of the gp350 glycoprotein with a codonoptimized bacterial ß-lactamase gene. The chimeric protein, termed gp350:BlaM, was readily incorporated into infectious virions (data not shown) similar to HIV particles that contain a chimeric β -lactamase-Vpr protein (BlaM-Vpr). Upon infection the chimeric protein is delivered into the cytoplasm of HIV target cells as a result of virion fusion (Cavrois et al., 2002;Cavrois et al., 2014;Jones and Padilla-Parra, 2016).
We generated wt/B95.8 (2089) Fig. 6, upper panel). To investigate whether EBV can also infect pDCs, we isolated pDCs from PBMCs using a negative selection protocol and repeated the BlaM fusion assay. pDCs infected with gp350:BlaM assembled EBV showed 1 5 approx. 5 % cells with blue fluorescence indicating that a small fraction of pDC fuses with EBV virions (Fig. 7C; Supplementary Fig. 6, lower panel). pDCs incubated with EVs assembled with gp350:BlaM or incubated with EBV lacking gp350:BlaM did not reveal a shift from green to blue cell fluorescence indicating that the novel assay reflects true fusion mediated by infection with EBV ( Supplementary Fig. 6). Our newly established EBV BlaM fusion assay documents that EBV can infect pDCs although at a rather low level compared with the virus's cognate B cells.
Finally, we asked whether the pDCs belonging to the small fraction that EBV infects are the main responders and thereby, main releasors of type I IFN upon viral uptake and fusion. We infected pDCs enriched from PBMCs with gp350:BlaM assembled wt/B95.8 (2089) EBV for four hours, loaded the cells with CCF4-AM and sorted them for green (non-infected) or blue (infected with gp350:BlaM assembled EBV) populations. Identical numbers of sorted green or blue cells were plated in 96-well cluster plates which were incubated at 37°C overnight.
The next day, the supernatants were analyzed for IFN-α concentrations by ELISA. Sorted blue cells, which had been infected with gp350:BlaM assembled EBV showed 15 fold higher IFN-α levels than sorted non-infected cells with green fluorescence (Fig. 7D). When we infected the same number of unsorted pDCs with an identical dose of wt/B95.8 (2089) EBV their supernatant contained low IFN-α levels comparable to the sorted, green pDCs incubated with gp350:BlaM assembled wt/B95.8 (2089) EBV (Fig. 7D). This result was expected given the low prevalence of infected cells in unsorted pDC populations. The data indicate that those pDCs that EBV can infect respond with massive IFN-α release, whereas cells that presumably also take up EBV but degrade the incoming virions in the lysosomal pathway contribute little to IFN-α synthesis.
In summary, our experiments reveal that EBV inefficiently fuses with pDCs and that these infected pDCs are the main producers of INF-α. The IFN-α secretion process depends on the presence of unmethylated EBV DNA, which is sensed by the PPR TLR9. Furthermore, the EBV-encoded miRNAs regulate the expression of genes involved in type I IFN secretion and genes involved in the response to type I IFNs in B cells in the pre-latent phase. In pDCs cells, viral miRNAs seem to modulate cellular genes of the type I IFN pathway, but depending on the composition of the different virus preparations only to a minor extent.
Unexpectedly, our experiments indicate that the abundantly expressed non-coding small EBERs do not seem to have a role in inducing type I IFN in either B cells or pDCs.

Discussion
In this study, we examined early anti-viral responses in primary B-lymphocytes and pDCs infected with EBV and EBV mutants. We found that pDCs are the main source of immediate IFN-α release whereas newly infected B cells release comparably low levels of type I IFNs later in the early, pre-latent phase of infection. In infection experiments with wild-type and mutant EBVs, we determined that viral miRNAs interfere with the secretion of proinflammatory cytokines and IFN-α from newly infected B cells as well as pDCs. Bioinformatic approaches and data mining of our experiments on established EBV-infected B cell lines revealed many cellular candidate genes that are regulated by EBV-encoded miRNAs and are involved in type I IFN secretion and response pathways. Their validation in subsequent analyses confirmed certain already known mRNA targets but added many new cellular genes in the type I upstream and downstream IFN pathways that appear to be governed by EBV's numerous viral miRNAs (Fig. 8).
Whereas EBV readily infects primary B-lymphocytes, its prime target cells, it was uncertain whether the virus can also infect pDCs, which respond by releasing type I IFN. To address this question and to clarify this so far controversial topic, we developed a novel assay based on viral delivery of an enzymatic activity, ß-lactamase, which we fused to gp350, an abundant viral component in the membrane of EBV virions. We found that EBV can readily fuse with a small fraction of pDCs, which, as a consequence, release large amounts of IFN-α compared to the fraction of pDCs that do not fuse with EBV. We also document that pDCs type I IFN release was unequivocally dependent on viral DNA, which is sensed by the endosomal TLR9 receptor.
Our analysis made use of several mutant EBVs to address the contribution of the abundantly expressed small non-coding EBERs and the role of the viral LF2 gene in EBV target cells (Wu et al., 2009). Very surprisingly and contrary to previous publications, deletion of EBV EBERs had no measurable impact on IFN-α release from PBMCs, pDCs, or B-lymphocytes infected with EBV. The cytoplasmic and endosomal pattern recognition receptors RIG-I and TLR3, respectively, have been reported to become activated by EBERs, inducing type I IFN release (Samanta et al., 2006); Iwakiri et al., 2009, J Exp Med, 206, 2091. EBERs were also reported to confer resistance against IFN-α induced apoptosis by binding to PKR and inhibiting its phosphorylation (Nanbo et al., 2002). In light of these publications, one would expect a reduced type I IFN release in infection experiments with EBER deleted mutant EBVs, which was not the case (Fig. 5). Our experiments differed from these published studies that used established B cell lines, mostly Burkitt lymphoma cells, and experimental settings that required the ectopic expression of EBER molecules. In contrast, we infected and analyzed primary human cells and concluded our observations within a couple of days 1 7 after infection. Similarly to EBV mutants devoid of EBERs, deletion of the viral gene coding for LF2 resulted in no measurable change of IFN levels compared with infections with r_wt/B95.8 EBV (Fig. 5). This result is conflicting with the literature describing LF2 as a type I IFN antagonist (Wu et al., 2009). Again, experimental conditions differ substantially as the authors relied on ectopic expression of LF2 in luciferase assays and protein-protein interactions with IRF7 in 293T cells (Wu et al., 2009).
Our findings with B cell infection experiments (Fig. 2) provided a conundrum as these results excluded the non-coding EBERs as viral factors that cause the release of pro-inflammatory cytokines ( Fig. 2A,B) and IFN-α (Fig. 2C) and which the many viral miRNAs regulate. As a consequence, unknown viral PAMPs likely exist and trigger RIG-I or alternative pattern recognition receptors that presumably recognize them. Recently, several KSHV RNA fragments were identified to be sensed by RIG-I in a RNA Pol III independent manner (Zhang et al., 2018). It seems plausible that also in the context of EBV infection certain viral RNAs other than EBERs might trigger RIG-I or other pattern recognition receptors in infected B cells. The situation in this cell type differs from experiments with PBMCs and pDCs where we could identify virion DNA as the culprit and inducer of IFN-α release (Fig. 6) confirming previous reports (Lim et al., 2007;Fiola et al., 2010;Quan et al., 2010;Severa et al., 2013).
Virion DNA is virtually free of methylated cytosine residues very similar to prokaryotic DNA due to a unique strategy of viral DNA replication during the productive, lytic phase of EBV infection (Buschle and Hammerschmidt, 2020).
We established a novel BlaM assay to study the potential of EBV to infect cell populations other than primary B-lymphocytes. The BlaM fusion assay (Fig. 7) identified a minor fraction of pDCs that EBV infects and which releases much higher IFN-α levels than pDCs that were not detectably infected (Fig. 7D). The assay monitors the first steps in viral infectionadhesion, endosomal uptake and fusion of cellular and viral membranes. It makes use of a transmembrane protein which, upon fusion, translocates to the cytoplasmic compartment where the prokaryotic ß-lactamase moiety cleaves its ingenious substrate (Cavrois et al., 2002;Cavrois et al., 2014). Coupling the ß-lactamase reporter domain to gp350 (or another type I transmembrane protein) prevents the secretion of free enzyme and its spontaneous uptake by other cells (Albanese et al., 2020). Consequently, the BlaM assays is free of erroneous background signals supporting the identification of only positive cells in a large population of non-infected cells (Fig. 7C,D) or to detect low level or inefficient fusion of extracellular vesicles with recipient cells (Albanese et al., 2020). The BlaM assay records fusion, only, and is not informative as to whether viral infections are abortive, productive or lead to latent infection. As the assay does not rely on de novo gene expression, it can identify cells that are infected but are resistant to subsequent viral gene expression and 1 8 replication. These cells might still present epitopes of the incoming viral particle to immune cells such as CD4+ T cells contributing to beneficial or adverse immune responses of the infected host organism.
Previously published data show conflicting results regarding pDCs and EBV infection. Fiola et al. postulated that EBV does not establish infection in pDCs but responded to EBV particles or isolated EBV DNA similar to our findings (Fiola et al., 2010). On the other hand, certain isolated EBV strains efficiently infected monocytes which had been stimulated with GM-CSF and IL-4 to induce their differentiation to DCs (Guerreiro-Cacais et al., 2004). In that study, the efficiency of EBV entry into monocytes was monitored by measuring green fluorescence intensity after 48 hours. In our fusion assay, we evaluated the immediate fusion of EBV with pDCs as our incubation period was just four hours before loading the cells with the CCF4-AM substrate, but we found only a small fraction (< 10 %) to undergo fusion (Fig.   7C). In contrast, efficient infection of pDCs through viral binding to the MHC class II molecule HLA-DR and expression of EBV-encoded genes was reported (Severa et al., 2013) together with MHC class II mediated uptake, degradation and subsequent activation of TLR9 followed by IFN-α secretion (Quan et al., 2010). In contrast, we did not observe expression of viral genes (or GFP encoded in the backbone of the viral genomic DNA) in pDCs even when the cells were incubated for up to two days after infection with EBV (data not shown). EBV infection of pDCs and monocytes was reported to reduce biological activities of these cell types (Li et al., 2002;Gujer et al., 2019). Li et al. further postulated that pDCs are not infected, but apoptotic and infected B cells cause the presentation of EBV antigens on MHC class I molecules of DCs. Recently it was reported that pDCs are able to undergo trogocytosis, i.e. to conjugate with antigen presenting cells and to extract surface molecules, including MHC class I molecules, presenting them on their own cell surface. With this mechanism non-infected pDCs could be able to present EBV specific epitopes on MHC class I molecules (Bonaccorsi et al., 2014).
Prior to our study, an accurate and direct method to monitor EBV infection of different cell populations was not available. Our newly introduced gp350-BlaM fusion assay is based on the uptake of EB virions and the fusion of their viral membrane with cellular, possibly endosomal membranes such that the content of the virion is released into the cytoplasm of EBV's host cell. As such, the method can monitor the very first steps of virus entry and fusion and does not depend on subsequent events such as viral de novo transcription for example.
Thus, our novel approach can precisely determine the fraction of cells, with which EBV virions fuse within hours. Using this assay, we demonstrate that EBV virions containing gp350-BlaM fuse with only a small fraction of pDCs, approximately 5 % but with the vast 1 9 majority of primary B-lymphocytes (Fig. 7C). Interestingly, among the abundance of DCs in general, only the small number of pDCs acts upon EBV infection by releasing IFN (Gujer et al., 2019). It also seems as if a so far unknown subpopulation of pDCs might exist that EBV targets (Fig. 7C,D), but their cellular identity is uncertain.
A closer look at the literature shows that there is a network of identified EBV-encoded miRNAs and their regulated targets. The identified target genes are mainly located in adaptive and innate immunity and immune escape mechanisms (Albanese et al., 2017), what also corroborates our hypothesis that miRNAs interfere with type I IFN signaling as part of the innate immune escape during EBV infection (Gallo et al., 2020;Iizasa et al., 2020;Li et al., 2020). With our experiments shown in Figures 3 and 4 and documented in Supplementary Figures 3 and 4, we confirmed a number of host cell transcripts targeted by viral miRNAs, but we also identified several previously unknown effectors and their cellular transcripts (Fig. 8). Already in 2017, through gene expression profiling experiments, miR-BART6-3p was reported to regulate RIG-I and IFN-β responses (Lu et al., 2017). Also, in our study we could confirm that RIG-I is a target of EBV-encoded miRNAs in dual luciferase assays and in recently infected B cells. We observed that miR-BART3 and miR-BART19 are both able to regulate the RIG-I 3´UTR (Fig. 3A). These findings indicate that possibly a combination of different miRNAs during EBV infection leads to a regulation of the same cellular gene. CBP, which we evaluated to be regulated by miR-BART16 (Fig. 4B), was already reported to be target of this viral miRNA (Hooykaas et al., 2017).  Hooykaas et al., 2017;Cox et al., 2015). On the contrary, there are also certain genes that were reported to be regulated by EBV miRNAs, which we could not confirm in 3´UTR luciferase reporter assays. In fact, the majority of candidates from our initial bioinformatics analysis did not pass our criteria as targets when analyzed in dual luciferase reporter assays.
For each candidate, we cloned the 3´UTR as well as introduced point mutations into the suspected seed match sequence to ascertain whether a functional biochemical interaction between the miRNA and its predicted mRNA target could occur. For most candidates of the screen, we were unable to confirm that the predicted miRNA interaction sites were functional.
However, we found that certain cellular proteins are clearly upregulated in r_ Δ miR-infected B cells, such as TLR7 and TLR9 (Fig. 3F) but, in luciferase reporter assays, we failed to identify any viral miRNAs that could regulate the 3'UTRs of these two transcripts ( Supplementary Fig. 7). Thus, TLR7 and TLR9 do not appear to be direct targets of EBV

Construction of mutant EBVs
All recombinant EBVs used in this study are based on the maxi-EBV plasmid p2089, which contains the B95-8 EBV genome (Delecluse et al., 1998). The EBV strain B95-8 has a deletion of several pre-miRNA loci, lacks the second lytic origin of DNA replication (oriLyt) and several other genes, which are present in EBV field strains. As described in Pich et al. a new wildtype-like strain based on B95-8 expressing all EBV-encoded miRNAs at physiological level was established. This newly generated reconstituted wildtype strain is called r_wt/B95.8 (6008) and was constructed by introducing a DNA fragment of the EBV strain M-ABA repairing the deletion in B98-5 and the maxi-EBV plasmid p2089 (Pich et al., 2019). Furthermore, parts of the F-plasmid backbone were replaced by introducing an artificial open reading frame encoding eGFP and pac (puromycin N-acetyltransferase conferring puromycin resistance). This new r_wt/B95.8 (6008) strain expresses all viral miRNAs at physiological levels, carries the second oriLyt and expresses LF1, LF2 and LF3 2 1 viral proteins, which are absent in the B95-8 EBV genome. A second EBV strain used in this study is the r_ΔmiR (6338) strain lacking all of the EBV-encoded pre-miRNA loci. It was generated as described in Pich at al. (Pich et al., 2019). Furthermore, the genes of the two viral non-coding EBER RNAs were deleted in r_wt/B95.8 and r_ΔmiR to obtain mutant EBVs termed ΔEBER and ΔEBER/ΔmiR, respectively (Fig. 1) (Pich et al., 2019). Based on genome of r_wt/B95.8 (6008) the viral LF2 gene was disabled by introducing a stop codon in the coding sequence of the LF2 gene. This EBV mutant was called ΔLF2 (Fig. 1, Supplementary Tab. 1).

Cells and culture
Cells used in this study were mostly cultivated in RPMI 1640 medium supplemented with 8 % fetal calf serum (FCS), 100 µg/ml streptomycin, 100 U/ml penicillin, 1 mM sodium pyruvate,

Preparation and quantification of infectious viral stocks
HEK293 virus producer cell lines were established after transfection of the different maxi-EBV plasmids and subsequent selection with puromycin. To generate virus stocks clonal producer cells were transfected using PEI Max with the two plasmids, p509 and p2670, to induce the lytic cycle of EBV. p509 encodes the BZLF1 gene, which triggers lytic phase reactivation (Hammerschmidt and Sugden, 1988) and p2670 the BALF4 gene, which increases the infectivity of the recombinant EBVs (Neuhierl et al., 2002). Three days post transfection supernatants were collected and centrifuged for 10 min at 1,200 rpm and 30 min at 5,000 rpm to remove cell debris. To determine the titers of the different virus stocks we used Raji cells as described (Steinbrück et al., 2015). In detail 1×10 5 Raji cells were incubated with different amounts of virus stocks in a volume of 1 ml for three days at 37 °C.
By flow cytometry the percentage of GFP positive cells was determined and the linear regression equation was calculated as described (Steinbrück et al., 2015). To concentrate the virus, a further ultracentrifugation step with an iodixanol (Optiprep) cushion was used. Therefore, the virus supernatants were loaded on 2-3 ml iodixanol (5 volumes of solution A 2 2 (iodixanol) and 1 volume of solution B (0.85% (w/v) NaCl, 60 mM Hepes, pH 7.4) and centrifuged for 2 h at 28,000 rpm at 4 °C. Concentrated virus in the interphase was transferred to microfiltration tubes (Amicon Ultra-15) and centrifuged at 2000 rpm at 4 °C.
Concentrated virus stocks were quantified using Raji cells as described above.

Generation of gp350:BlaM assembled extracellular vesicles (EVs) and their quantitation
To assemble extracellular vesicles (EVs) with gp350:BlaM we transfected HEK293 cells with the p7180 plasmid, which codes for a fusion of gp350 and a codon-optimized BlaM (gp350:BlaM). Three days after transfection supernatants were harvested as described above and EVs were sedimented by ultracentrifugation at 24,000 rpm for 2 h. The EV stocks were quantified using the Elijah assay. 2×10 5 Elijah cells (Rowe et al., 1985) were incubated with different volumes of EVs or a calibrated wt/B95.8 (2089)

Isolation of B cells from adenoids and infection
Adenoid biopsies were washed with PBS and mechanically disintegrated with two sterile scalpels in a sterile petri dish. The cell suspension was filtered through a 100 µm mesh cell strainer. To obtain a maximum of single cells this procedure was repeated. The cell suspension was brought to 50 ml with PBS, mixed and centrifuged at 300 g for 10 min. The for 15 min at 4 °C. A depletion of CD3 positive cells was done by magnetic bead sorting. The cells were washed in 10 ml staining buffer and resuspended in pre-warmed medium. After counting the cells were infected with titered virus stocks at a density of 1×10 6 cells/ml. The next day the cells were centrifuged, resuspended in fresh medium at the same density and cultivated. Four days later (on day five post-infection) the cells were counted using calibrated APC beads as volume standard using a flow cytometer. The cells were incubated for 4 subsequent days at 37 °C at a density of 7×10 5 cells/ml in a total volume of 2 ml in 24-well cluster plates. Afterwards, supernatants were collected for ELISA analysis.

Protein lysates from infected B cells and Western blot immunostaining
To determine the protein expression in infected B cells, B cells were isolated as described above. Infection of the isolated cells with the indicated virus stocks was performed at an initial density of 1×10 6 cells/ml. After 24 hours the cells were centrifuged and resuspended in fresh medium at the same density. Cells were incubated at 37 °C for 4 or 8 additional days.
Cells were washed in cold PBS and resuspended in RIPA lysis buffer complemented with protease and phosphatase inhibitors. Cell lysates were frozen at -80 °C. After thawing on ice, the lysates were mixed and centrifuged at 13,000 rpm for 10 min at 4 °C. Supernatants were collected and the protein amount was determined using the Bradford assay. Protein concentrations of the lysates were adjusted using RIPA lysis buffer, Lämmli buffer was added and identical protein amounts of the different samples were loaded on TGX Stain-free Precast gels purchased from Biorad. After the run the gels were activated by a 45 sec UV exposure and electroblotted onto nitrocellulose membranes. The membrane was blocked and incubated with the indicated primary antibodies ( Fig. 3F and 4E (#3460-1A-6) and IL-10 (#3430-1A-6). The assays were performed as described in the manufacturer´s protocol using 60-80 µl of cell supernatants.

Isolation and infection of PBMCs and pDCs
PBMCs were isolated from fresh whole blood samples or buffy coats. The blood was diluted 1:3 or more in PBS and 30 ml of the diluted blood sample was layered onto a 15 ml Ficoll Hypaque cushion in a 50 ml tube. The tubes were centrifuged at 1,850 rpm at 10 °C for 30 min. PBMCs were carefully collected from the turbid interphase and were transferred to a fresh tube. Cells were washed with PBS three times at decreasing centrifugation parameters

PAR-CLIP datasets and Ingenuity pathway analysis.
EBV B95-8 or wild-type LCLs and EBV+ PEL PAR-CLIP datasets are previously described; Ago PAR-CLIP datasets for AIDS-related DLBCL cell lines (IBL1, IBL4, and BCKN1) were generated as previously described Skalsky et al., 2012;Majoros et al., 2013;Skalsky et al., 2014). AIDS-related DLBCL cell lines were kindly provided by Dr. Ethel Reads were processed using the FASTX-toolkit (http://hannonlab.cshl.edu/fastx_toolkit/), aligned to the human genome (hg19) using Bowtie (-v 3, -l 12, -m 10, --best --strata), and Ago interaction sites were defined using the PARalyzer v1.5 ) and 2 6 PIPE-CLIP (Chen et al., 2014) pipelines. For PARalyzer analysis, reads are mapped to the genome and reads that overlap by at least one nucleotide are placed into groups. Groups are then analyzed for T>C conversions and regions with a higher likelihood of T>C conversions are posited as interaction sites so long as the minimum read depth of three reads per site is met. 3'UTR Ago interaction sites were annotated and scanned ad hoc for canonical seed matches (>=7mer1A) to mature EBV miRNAs.
To define major cellular pathways regulated by EBV miRNAs, we used Ingenuity Pathway Analysis (IPA) (Qiagen). 3,976 human genes harboring 3'UTR miRNA interaction sites (>=7mer1A) as identified in PAR-CLIP (EBV+ BC1 cells, EBV B95-8 or wild-type LCLs and EBV+ DLBCL cells) were queried. Enriched canonical pathways related to IFN-I signaling were defined by IPA, and individual pathways were selected for visual representation. Genes targeted by EBV miRNAs that modulate these enriched pathways were curated and subject to further validation.

(B-D)
293T cells were co-transfected with the indicated wild-type luciferase reporter plasmids (WT) for CBP, IRF9, and JAK1 or reporter plasmids in which the predicted seed sequences were mutated (mut) together with or without a miRNA-encoding plasmid. The luciferase expression in these cells was assessed and normalized to lysates from cells co-transfected with the wild-type 3´-UTR reporter and empty plasmid (∅). P-values were calculated using the one-way ANOVA test. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Supplementary Figure 7. EBV's miRNAs do not target TLR7 and TLR9 3'UTRs
293T cells were co-transfected with the indicated wild-type luciferase reporter plasmids (WT) for TLR7 or TLR9 with or without a miRNA-encoding plasmid. Luciferase expression in these cells was assessed and normalized to lysates from cells co-transfected with the wild-type 3´-UTR reporter and the empty miRNA expression plasmid (Ctrl).