KSHV-encoded vIRF3 cooperates with cellular IRF4 to drive super-enhancer activity through complex DNA elements

Abstract

The Kaposi’s sarcoma-associated herpesvirus oncoprotein vIRF3 is essential for primary effusion lymphoma (PEL) cell survival and cooperates with cellular interferon regulatory factor 4 (IRF4) to activate super-enhancers (SEs) controlling MYC and IRF4. However, the DNA sequence requirements for this cooperation are unknown. Using reporter assays, we mapped vIRF3/IRF4 responsiveness within the IRF4-SE to an ∼83-bp element activated by vIRF3 alone and cooperatively by vIRF3 and IRF4. vIRF3-dependent activation required an AP-1 motif, whereas cooperation with IRF4 required IRF4 DNA binding and IRF-related motifs outside canonical AP-1-IRF composite elements (AICEs). These motifs were necessary but insufficient outside their native sequence context, indicating a requirement for extended flanking sequences. DNA pulldowns revealed distinct motif requirements for vIRF3 and IRF4 association with the IRF4-SE. A distal MYC-SE required a similarly extended region containing a functional AICE2 motif, with vIRF3 activity depending primarily on its AP-1 half-site. Motif requirements were preserved in integrated reporter assays. BATF promoted vIRF3/IRF4-dependent activation preferentially in integrated reporters and was required to maintain vIRF3 and IRF4 occupancy at the endogenous MYC-SE. These findings define context-dependent DNA and cofactor requirements through which vIRF3 co-opts IRF4-regulated oncogenic enhancers in PEL.

Graphical Abstract

Graphical Abstract.

Introduction

Kaposi’s sarcoma-associated herpesvirus (KSHV, also known as human herpesvirus 8) is the causative agent of primary effusion lymphoma (PEL) [1–3]. The vast majority of PEL tumor cells exhibit the latent KSHV gene expression program. Notably, over 80% of PEL tumors are co-infected by Epstein-Barr virus (EBV). PEL tumor cells display the genetic rearrangements and gene expression features of post-germinal center B cells, including high expression of the hematopoietic lineage transcription factor (TF) interferon regulatory factor 4 (IRF4) [4–6]. Accumulating evidence shows that the overexpression of IRF4 is critical for PEL pathogenesis. Specifically, IRF4 is essential for the survival and proliferation of all PEL-derived cell lines tested to date, irrespective of EBV co-infection [7–10]. PEL shares this IRF4 dependency with several other hematologic malignancies, including multiple myeloma [11], the activated B-cell-like subtype of diffuse large B-cell lymphoma [12, 13], anaplastic large cell lymphoma [14, 15], and adult T-cell leukemia/lymphoma [16]. IRF4 dependency is also observed in EBV-transformed lymphoblastoid cell lines (LCLs) [17, 18], which result from the in vitro immortalization of primary B cells.

In PEL, other IRF4-dependent cancers, and LCLs, IRF4 occupies super-enhancers (SEs) [8, 16, 19–21]. SEs are large clusters of enhancer elements, typically spanning >8 kb, in contrast to <1 kb for regular enhancers [22, 23]. SEs drive exceptionally high transcription from promoters linked in 3D through looping factors. In cancer, SEs have been found to mediate the overexpression of essential proteins that promote cellular survival, proliferation, and other cancer-cell-specific features [24]. One of the most critical SE-linked transcriptional targets of IRF4 in the IRF4-dependent cancers and LCLs is MYC. MYC itself is a TF with roles in proliferation, cell growth, and metabolism [25–27]. The aberrant overexpression of MYC contributes to many cancers, due to genetic amplification or translocation, or transcriptional or post-transcriptional upregulation. IRF4-dependent cancers, including PEL, typically require the SE-mediated transcriptional activation of MYC. IRF4-dependent SEs furthermore drive the overexpression of the IRF4 gene itself and essential IRF4 cofactors [16, 21–24], likely reinforcing IRF4 expression and function.

IRF4 is one of nine human interferon regulatory factor (IRF) TFs [28, 29]. IRFs share a highly conserved N-terminal DNA-binding domain (DBD), followed by a variable linker domain, and a moderately conserved IRF-association domain involved in interactions between IRFs and other protein partners. Several IRFs, including IRF4, additionally contain a C-terminal autoinhibitory region (AR). IRF DBDs interact with the core IRF DNA motif (5′-GAAA, referred to as IRF/GAAA here), which is found in tandem in the interferon-sensitive response element (ISRE; 5′-GAAAnnGAAA, where n is any nucleotide), e.g. within the promoters of interferon-stimulated genes. IRF4 expression is induced downstream of T- and B-cell receptor signaling, such that the magnitude of IRF4 induction correlates with the strength and duration of receptor engagement [30–33]. IRF4’s participation in TF complexes is governed by this variable expression level during immune cell differentiation and the availability of cofactors. IRF4 binds to the consensus IRF site more weakly than other IRFs, due to divergent features in its DBD and autoinhibition through its AR domain [34]. At low IRF4 expression levels, association with other TFs can relieve IRF4 autoinhibition and facilitate binding to composite sites. This paradigm of IRF4 activation is most established for the cooperation of IRF4 with ETS family members, PU.1 (encoded by SPI1) or SPIB, on ETS-IRF (EICE, 5′-GGAAnnGAAA) composite sites [34]. Similarly, IRF4 and AP-1 complexes, comprised of BATF and JUN subfamily members, bind to AP-1–IRF composite elements (AICE, which have the consensus sequence 5′-TGAnTCA-n4-GAAA or GAAATGAnTCA, called AICE1 or AICE2, respectively) [35–37]. IKZF1 (Ikaros) and IRF4 may finally recognize zinc finger-IRF composite sites [38]. At the high concentrations of IRF4 expression that drive plasma cell differentiation, IRF4 homodimerizes on the ISRE, which is critical for the induction of the plasma cell master regulator PRDM1 (also known as BLIMP1) and consequent plasma cell differentiation [31].

The exact identities of the SE elements and TFs that cooperate with IRF4 appear to differ between IRF4-dependent settings. Our published study used ChIP-seq to show that the latently expressed KSHV vIRF3 protein co-occupies several PEL SEs with IRF4 and BATF. Expression of vIRF3 is required for the survival of PEL cell lines in culture [8, 39, 40]. Published ChIP-seq, RNA-seq, Hi-ChIP, western, and reporter data together support a model where vIRF3 and IRF4 positively regulate the PEL SEs linked to the IRF4 and MYC genes, whose high expression is essential for the survival and proliferation of PEL cells [8, 19, 20]. Targeting the vIRF3/IRF4-occupied SE elements linked to IRF4 and MYC using CRISPR interference resulted in reduced expression of the linked genes and loss of the cultures [8, 20]. Similarly, knockdown of either IRF4 or vIRF3 resulted in reduced expression of genes linked to vIRF3/IRF4-occupied SEs [8, 20], including IRF4 and MYC. While BATF is also essential in PEL cells, the reduction in viable PEL cell numbers following knockdown of BATF was delayed by several days compared to the more rapid loss of the culture after the inactivation of vIRF3 or IRF4, leaving the contribution of BATF to vIRF3/IRF4-dependent SE regulation less clear [8]. The chromatin occupancy of JUN family members has not been studied in PEL cells, and individual family members did not score as essential in CRISPR screens [8, 9], suggesting potential redundancies for JUN family members as essential BATF binding partners.

Many questions remain regarding how vIRF3 cooperates with IRF4. vIRF3 shares homology with cellular IRFs in its N-terminal and C-terminal domains. While the N-terminal domain of vIRF3 is predicted to resemble the DBD of cellular IRFs structurally, it lacks the conserved amino acids that mediate the recognition of the IRF motif. Since vIRF3 is, therefore, not expected to bind to the IRF motif, it is unknown how vIRF3 is recruited to SEs. Our published ChIP-seq dataset identifies few high-confidence vIRF3 peaks [8], including ∼60 that map to SEs and overlap IRF4 peaks. HOMER motif analyses indicated significant enrichment of these peaks for composite sites, i.e. ISRE, AICE, and EICE sites, as well as IRF and AP-1 sites. However, the actual TF occupancy and regulatory importance of these sites may depend on the cellular context, and it is unknown whether these motifs mediate vIRF3 and IRF4 function at the essential PEL SEs.

Here, we begin to address these questions by mapping vIRF3/IRF4-responsive elements in the IRF4- and MYC-SEs. Our results show that vIRF3/IRF4-dependent SE activation involves extended and complex genetic elements. In transfected reporter assays, vIRF3 activity depends on critical AP-1 motifs in the IRF4-SE and one of two MYC-SEs, whereas cooperation with IRF4 depends on IRF-related motifs and the DNA-binding ability of IRF4. These sequence requirements are preserved in integrated reporter contexts, where BATF promotes vIRF3/IRF4-dependent activation. BATF furthermore promotes vIRF3 and IRF4 occupancy at the endogenous MYC-SE. Together, these data reveal context-dependent DNA and cofactor requirements through which vIRF3 engages oncogenic SEs and may enable IRF4 to function in previously unknown binding configurations.

Materials and methods

Biological resources

293T/17 (“293T”) (ATCC, CRL-11268) were grown in Dulbecco’s Modified Eagle’s Medium (DMEM, Corning, 10-017-CV) containing 10% Serum Plus™ II Medium Supplement (Sigma–Aldrich 14009C-500ML, Batch Number: 21C421) and 10 μg/ml gentamicin (Gibco, 15710072). 293 (ATCC, CRL-1573) were grown in DMEM containing 10% fetal bovine serum (FBS, Corning, 35-010-CV) and 10 μg/ml gentamicin. BC-3 (ATCC, CRL-2277) were grown in RPMI 1640 (Corning, 10-040-CV), containing 20% FBS and 10 μg/ml gentamicin. BJAB (obtained from Dr Bryan Cullen at Duke University Medical Center) were grown in RPMI 1640 containing 10% FBS and 10 μg/ml gentamicin. All cell lines were grown at 37°C with 5% CO₂. pLS-mP [41] was a gift from Nadav Ahituv (Addgene plasmid # 81225; http://n2t.net/addgene:81225;  RRID:Addgene_81225). The pLS-mP vector used here was reobtained from Addgene and differs slightly from the vector we used previously [8]. LVDP-CArG-RE-GPR was a gift from Stelios Andreadis (Addgene plasmid # 89762; http://n2t.net/addgene:89762;  RRID:Addgene_89762). pGL4.72[hRlucCP] (GenBank Accession number AY738228) and pGL4.23[luc2/minP] (referred to as “pGL4.23” here, catalog number: E8411, GenBank Accession number DQ904455) were purchased from Promega, Madison, WI, USA. pLC-IRF4-IRES-Puro and related lentiviral vIRF3 and BATF expression vectors were described previously [8].

Reagents and specialized equipment

All restriction enzymes, purchased in HF versions where available, and the Q5 Site-Directed Mutagenesis Kit were from New England Biolabs (Ipswich, MA, USA). We used Taq polymerase (prepared as in [42]) to amplify probes for DNA pulldown assays. Oligonucleotides, Ultramers, and gBlocks used for cloning were purchased from Integrated DNA Technologies (IDT, Coralville, IA, USA) and are listed in Supplementary Table S1. Modified oligos used for DNA pulldown assays and electrophoretic mobility shift assays (EMSAs) were also purchased from IDT and are listed in Supplementary Tables S2 and S3. We performed Gibson Assembly [43] using an in-house protocol that is available on request. We used the Dual-Luciferase® Reporter Assay System (Promega, Madison, WI, USA, catalog number E1980) in combination with a VICTOR® Nivo™ Multimode Plate Reader with Dual Injectors from Perkin Elmer (Waltham, MA, USA). We imaged all gels and Westerns using a LI-COR (Lincoln, NE, USA) Odyssey-M Imaging System. Flow cytometry was performed with a BD FACSymphony™ A1 or FACSCanto™ II. Information for primary and secondary antibodies and control IgG is detailed in Supplementary Table S4. Other reagents are listed in the relevant sections.

Software

FACS data were analyzed in FlowJo v10 (BD).

Statistical analyses

We performed statistical analyses using GraphPad Prism 10 or 11 as detailed for each figure or panel. We performed two independent repeats for EMSAs and expression controls that included well-defined positive controls and had qualitatively reproducible results in these repeats. We performed at least three independent repeats for all figures that involved statistical analyses over replicates.

Constructs and cloning procedures

All newly cloned vectors were verified by Sanger Sequencing (ACGT, Wheeling, IL, USA) or, where the NEB Q5 Site-Directed Mutagenesis Kit was used, by full plasmid sequencing (Plasmidsaurus, South San Francisco, CA, USA).

To clone pGL4.72-RSV-LTR, the Rous sarcoma virus long terminal repeat (RSV-LTR) was polymerase chain reaction (PCR)-amplified from a lentiviral vector using primers 393 and 394 and cloned between the BglII and HindIII sites of pGL4.72[hRlucCP] using T4 DNA ligase.

The following constructs were cloned by Gibson Assembly of the relevant PCR products with XhoI/HindIII-digested pGL4.23[luc2/minP]. For pGL4.23-IRF4-SE^500bp, we inserted a PCR fragment we amplified from the published vector pLS-IRF4-SE-eGFP [8] using primers 4850/4851; for pGL4.23-IRF4-SE^69-151, the relevant fragment was amplified from pGL4.23-IRF4-SE^500bp using primers 5056/5057; for pGL4.23-MYC-SE^500kb a 327bp fragment was inserted, which was originally amplified from genomic DNA and re-amplified using primers 4959/4960; for pGL4.23-MYC-SE^375kb a 299-bp fragment was amplified from genomic DNA using primers 4854/4855. PCR products for pGL4.23-IRF4-SE^500bp-50 bp and 18- to 22-bp deletion mutants were generated using primers 4957 and 5048 (Δ1–50), 4850 and 4958 (Δ451–500), or by overlap PCR using outer primers 5048/5049 and two PCR fragments as templates, which were in turn amplified using forward primer 5049 and mutant-specific reverse primers or mutant-specific forward primers and reverse primer 5048.

The pGL4.23-IRF4-SE^69-151-GAAA1/2/3/4 mutants, GAAA #2 & AP-1 mutant, -ISRE-like mutant, and the pGL4.23-IRF4-SE^500bp-Δ51–151 mutant were cloned using the NEB Q5 Site-Directed Mutagenesis Kit using pGL4.23-IRF4-SE^69-151 or pGL4.23-IRF4-SE^500bp as templates. pGL4.23-IRF4-SE^500bp-Δ51–151 was not used for reporter assays in this paper but served as a template for pLS-mP cloning. To clone the pGL4.23-IRF4-SE^69-151-4GAAA, AP-1/TGAnTCA, and TGAA mutants, we inserted synthesized double-stranded DNA fragments (gBlocks) 5451, 5454, or 5455, respectively, into the XhoI/HindIII-digested pGL4.23 vectors using Gibson Assembly. To clone pGL4.23-IRF4-SE^69-151-R1^old and -R2, we performed Gibson Assembly of the XhoI-HindIII-digested vector with gBlocks (IDT) 5456 and 5457, respectively. pGL4.23-IRF4-SE^69-151-R1^old was not used for reporter assays in this paper but served as a template for cloning below. pGL4.23-IRF4-SE^69-151-R1, which retains the A between the AP-1 and second GAAA motifs, was cloned using the Q5 Site-Directed Mutagenesis Kit, using primers 5618/5619, and pGL4.23-IRF4-SE^69-151-R1^old as template. pGL4.23-IRF4-SE-87–107, 108–129, 87–129 and pGL4.23-R3/R4/R5 were cloned by joining annealed oligos or Ultramers with XhoI/HindIII-digested pGL4.23 vector using T4 DNA ligase.

pGL4.23-MYC-SE^500kb 40 bp deletion mutants were generated by overlap PCR using primers 5048/5049 and two PCR fragments as templates, which were in turn amplified using forward primer 5049 and mutant-specific reverse primers or mutant-specific forward primers and reverse primer 5048. Mutant overlap PCR products were inserted into the XhoI/HindIII-digested pGL4.23 vector using Gibson Assembly. pGL4.23-MYC-SE^500kb-AICE2, AP-1/TGAnTCA, and GAAA mutants and the pGL4.23-MYC-SE^375kb AP-1/TGAnTCA mutant were cloned using the NEB Q5 Site-Directed Mutagenesis Kit and the relevant wild-type (wt) vector as template.

To clone the pLS-IRF4-SE^500bp and -IRF4-SE^69-151 reporters, we used primers 5620/5621 and 5622/5623 to amplify the relevant sequences from pGL4.23-IRF4-SE^500bp and pGL4.23-IRF4-SE^69-151, respectively. The products were cloned into XbaI-digested pLS-mP using Gibson Assembly. pLS-mP-IRF4-SE^500bp-Δ69–151 was cloned by inserting two overlapping PCR fragments amplified from pLS-mP-IRF4-SE^500bp with primers 5620/5523 and 5524/5621 into XhoI/XbaI-digested pLS-mP-IRF4-SE^500bp using Gibson Assembly. To clone the remaining pLS-mP-IRF4-SE^69-151 mutants, we used primers specified in Supplementary Table S1 to amplify the relevant fragments from the corresponding pGL4.23 constructs. The products were inserted into XhoI/XbaI-digested pLS-mP-IRF4-SE^500bp by Gibson Assembly.

LVDP-CArG-RE-GPR was cut using AgeI and EcoRI. Fragments containing the minimal promoter, minP, and minP-IRF4-SE^500bp were amplified with primers 5982/5983 and 5982/5984 from template pLS-mP-IRF4-SE^500bp and inserted into the LVDP-CArG-RE-GPR backbone by Gibson Assembly, resulting in plasmids LVDP-minP and LVDP-minP-IRF4-SE^500bp, which were only used for further cloning here. To replace the original LVDP-CArG-RE-GPR dsRed-Express 2 cassette with mCherry, we amplified an mCherry-containing fragment with primers 6000/6001 and an IRF4-SE^500bp-ZsGreen fragment from LVDP-minP-IRF4-SE^500bp with primers 6002/6003. We obtained the hPGK-cHS4 fragment by restriction digest of LVDP-minP-IRF4-SE^500bp with NheI and XbaI. The three resulting fragments were inserted into XhoI/EcoRI-digested LVDP-minP-IRF4-SE^500bp using Gibson Assembly to generate mLVDP-minP-IRF4-SE^500bp. We generated a minP control vector, mLVDP-minP, by excising the minP fragment from LVDP-minP using AgeI and EcoRI and replacing the corresponding fragment of mLVDP-minP-IRF4-SE^500bp using T4 DNA ligase. mLVDP-minP-IRF4-SE^500bp and mLVDP-minP were only used for cloning in this study. To insert a puromycin resistance cassette downstream of mCherry, we amplified a P2A-Puro fragment from pL2M-MPH-2A-Puro [44] with primers 6058/6059, and mCherry and hPGK fragments from mLVDP-minP-IRF4-SE^500bp (with primers 6060/6061 and 6062/6063, respectively). The resulting three fragments were inserted into BamHI-digested mLVDP-minP-IRF4-SE^500bp to generate pLVDP-mCherry-P2A-Puro-IRF4-SE^500bp. pLVDP-mCherry-P2A-Puro-IRF4-SE^500bp-Δ69–151 and -IRF4-SE^69-151 were cloned by excising the relevant fragments from the corresponding pLS-mP vectors using EcoRI and XbaI and inserting them into EcoRI- and XbaI-digested pLVDP-mCherry-P2A-Puro using T4 DNA ligase. To clone pLVDP-mCherry-P2A-Puro-minP, we excised the relevant fragment from mLVDP-minP using PacI and EcoRI and replaced the corresponding fragment of pLVDP-mCherry-P2A-Puro-IRF4-SE^500bp using T4 DNA ligation. To generate pLVDP-mCherry-P2A-Puro-IRF4-SE^200bp, we amplified the first 200 bp of IRF4-SE^500bp from pGL4.23-IRF4-SE^500bp using primers 6003/6091 and used Gibson Assembly to insert the resulting fragment into EcoRI/XbaI-digested pLVDP-mCherry-P2A-Puro-IRF4-SE^500bp. To generate pLVDP-mCherry-P2A-Puro-IRF4-SE^500bp-AP-1mut, we amplified AP-1 mutant IRF4-SE^500bp fragments using primers 6092/6093 and 6094/6095 and used Gibson Assembly to insert them into EcoRI/XbaI-digested pLVDP-mCherry-P2A-Puro-IRF4-SE^500bp.

To generate pLVDP-mCherry-P2A-Puro-IRF4-SE^500bp-4GAAA mut and ISRE-like mut, we inserted gBlocks 6412 and 6413 into the wt vector that was digested using XhoI and XbaI. pLVDP-mCherry-P2A-Puro-MYC-SE^500kb and an AP-1 mutant version were cloned by inserting PCR fragments generated using primers 5986/6096 and wt or AP-1 mutant templates pGL4.23-MYC-SE^500kb into EcoRI/XbaI-digested pLVDP-mCherry-P2A-Puro-IRF4-SE^500bp using Gibson Assembly. To generate pLVDP-mCherry-P2A-Puro-MYC-SE^500kb/^81–247bp, we amplified the relevant fragment from pGL4.23-MYC-SE^500kb using primers 6097 and 6098 and performed Gibson Assembly with EcoRI/XbaI-digested pLVDP-mCherry-P2A-Puro-IRF4-SE^500bp. To generate AP-1 mutant pLVDP-mCherry-P2A-Puro-MYC-SE^500kb/^81–247bp, we amplified the relevant mutant fragments using primers 6097/6099 and 6100/6098 and cloned them into EcoRI/XbaI-digested pLVDP-mCherry-P2A-Puro-IRF4-SE^500bp using Gibson Assembly.

We cloned pLC-EBFP2-IRES-Puro, pLC-vIRF3-IRES-Puro, and pLC-BATF-IRES-Puro by moving the relevant expression cassettes between published vectors [8], using AgeI and BamHI and T4 DNA ligase. To clone pLC-BATF2-IRES-Puro and pLC-BATF3-IRES-Puro, which express untagged BATF2 and BATF3, respectively, we ordered gBlocks 6182 and 6183 containing HA-tagged expression cassettes, amplified the related untagged expression cassettes using primers 6184/6185 or 6184/6186, and joined them with AgeI- and BamHI-digested pLC-IRES-Puro using Gibson Assembly. pLC-JUNB-IRES-Puro was cloned by inserting a PCR fragment amplified from genomic DNA using primers 5481 and 5482 into the AgeI/BamHI-digested pLC-IRES-Puro backbone, using Gibson assembly. JUNB coding sequences were confirmed to match the reference sequence by Sanger sequencing. To clone pLC-IRF4^R98A/C99A-IRES-Puro, two mutant fragments were amplified using primers 4087/5479 and 5480/4764 and pLC-IRF4-IRES-Puro as template and inserted into the AgeI/BamHI-digested pLC-IRES-Puro backbone using Gibson Assembly.

Dual-Luciferase reporter assay

293 cells were plated at 1 × 10⁵ cells per well of a 24-well plate 1 day before transfection. Cells were co-transfected with 50 ng of pGL4.23 Firefly luciferase reporter plasmid, 50 ng of an RSV-LTR-driven Renilla luciferase (RLuc) internal control plasmid pGL4.72-RSV-LTR, and 200 ng of pLC-IRES-Puro-based expression plasmids, i.e. 100 ng of pLC-vIRF3-IRES-Puro and/or pLC-IRF4-IRES-Puro, and/or 100 or 200 ng pLC-EBFP2-IRES-Puro. For reporter assays including BATF/JUNB, we also used 100 ng per vector and the total amount of pLC-IRES-Puro plasmids was adjusted to 400 ng with pLC-EBFP2-IRES-Puro as filler. A Master Mix of 50 μl Opti-MEM and Lipofectamine LTX (Invitrogen # 15338100) [DNA(μg): LTX (μl) = 1:3] per reaction was prepared and incubated at room temperature for 5 min. Then the master mix was mixed with plasmid combinations described earlier and incubated at room temperature for 15 min before addition into the wells. Approximately 24 h post-transfection, media were aspirated, and cells were lysed using 100 μl of 1X Passive Cell Lysis Buffer (Promega) per well for 15 min. Five microliters supernatant of each lysate was transferred to a 96-well half-well white plate and Dual-Luciferase reporter assays were performed using the Dual-Luciferase® Reporter Assay System (Promega) according to manufacturer specifications on a Victor Nivo luminescence and fluorescence plate reader with dual injectors.

Lentivirus production and concentration

Lentivirus production was performed as described [44]. Briefly, 293T/17 were plated at 12.5 × 10⁶ cells per 15-cm dish, 4.8 × 10⁶ cells per 10 cm dish, or 8 × 10⁵ per well of a six-well plate 1 day prior to transfection. Lentiviral transfer plasmids, psPAX2, and pMD2.G were co-transfected at 45:35:25 molar ratios using the polyethyleneimine (PEI, Polysciences) transfection method. The medium was changed 4–6 h after transfection. Three days after transfection, the supernatants containing lentiviral particles were harvested and filtered through 0.45 μm-pore-size filters to remove cell debris. pLS-mP-based LVs were concentrated using Amicon Ultra centrifugal filter units (Millipore UFC910024). The concentrated LV preparations were stored in aliquots at −80°C for future use. Filtered pLVDP-containing media were directly stored in aliquots at −80°C without concentration. pLC-IRES-Puro lentiviruses were concentrated by ultracentrifugation as described and stored in aliquots at −80°C [44].

RNA-based and functional titration of lentiviruses

For pLS-mP and pLC-IRES-Puro lentiviruses, we performed both RNA-based and functional titration. For RNA-based titration, lentiviral RNA was purified from supernatants containing viruses using the NucleoSpin® RNA Virus kit (Takara #740956.50). The viral RNA was eluted in 50 μl RNase-free water provided with the kit and subjected to DNase treatment using the RQ1 DNase kit (Promega, #M6101), following the manufacturer’s protocol. Lentiviral genomic RNA copy numbers were quantified by reverse transcription quantitative PCR (RT-qPCR) using the Lenti-X qRT-PCR titration kit (Takara, #631235) following the manufacturer’s instructions, including no template controls (NTC, i.e. dilution buffer or RNase-free water), and no reverse transcriptase controls. A standard curve of threshold cycle (Ct) values versus copy numbers (log scale) was generated based on serial dilutions of the Lenti-X RNA Control template and used to calculate lentiviral copy numbers determined by RT-qPCR of serial dilutions of each viral RNA sample.

In addition to RNA-based titration, we titered each pLS-mP lentivirus and EBFP2- and mCherry-expressing pLC-IRES-Puro controls functionally using flow cytometry. For functional titration of pLS-mP vectors, we plated BC-3 in 12-well plates at a density of 4 × 10⁵ cells/ml and transduced cells with serial dilutions of viral supernatants in the presence of 5 μg/ml polybrene. One day after transduction, the medium was replaced, and cells were split to 3 × 10⁵/ml. Three days post-transduction, the percentage of GFP-expressing cells at each dilution was assessed by flow cytometry and functional titers (transducing units per milliliter) were calculated from dilutions generating single integration events, i.e. yielding <30% of GFP-positive cells. The relative RNA-based titers and functional titers were largely consistent across samples and, therefore, the functional titers were used to calculate the appropriate volume of viral supernatants for subsequent experiments. Functional titration of pLC-IRES-Puro controls in 293 was performed similarly, using transduction conditions outlined in the next section. We used RNA-based relative titers to calculate functional titers for other pLC-IRES-Puro vectors. For pLVDP viruses, we only performed functional titrations in BC-3 and 293 cells as described earlier, calculating the functional titer from the percentage of mCherry-expressing cells 2 days after transduction.

Lentiviral-based reporter assays

For pLS-mP viruses, BC-3 cells were plated at 4 × 10⁵ cells/ml in 12-well plates with 5 μg/ml polybrene. The lentiviruses containing SE-element-driven EGFP reporter were transduced at MOI of ∼1. The medium was changed 1 day after transduction. Three days after transduction, cells were subjected to flow cytometry and EGFP mean fluorescence intensity (MFI) was determined in EGFP-positive cells.

The pLVDP dual-reporter lentiviruses were transduced at an MOI of ∼0.3 into BC-3, BCBL-1 or BJAB cells as described earlier, except that cells were plated at 3 × 10⁵ cells/ml. One day after transduction, medium was replaced with medium containing 1–2 μg/ml Puromycin to select transduced cells. Three days after transduction, cells were subjected to flow cytometry, and ZsGreen mean fluorescence intensity was determined in live, single, mCherry-positive cells.

For Fig. 10, 8 × 10⁴ 293 cells per well were plated in 24-well plates the day before transduction. pLVDP-mCherry-P2A-Puro-IRF4-SE^500bp-wt, or wt and AP-1-mutant pLVDP-mCherry-P2A-Puro-MYC-SE^500kb reporter vectors were transduced at MOI ∼3. Cells were additionally transduced with pLC-IRES-Puro-EBFP2 at MOIs ∼3, 6, 9, and 12 and indicated combinations of pLC-IRES-Puro vectors expressing IRF4, vIRF3, BATF, and JUNB at an MOI of ∼3 each. 293 cells were transduced in the presence of 5 μg/ml polybrene for 24 h and media were changed the following day. Three days after transduction, cells were trypsinized and subjected to flow cytometry as above.

Figure 10.

BATF promotes the vIRF3/IRF4-mediated activation of integrated reporters. (A) ZsGreen reporter expression in 293 cells transduced with pLVDP-mCherry-P2A-Puro-IRF4-SE^500bp, and no additional vectors (no TF), an EBFP2-expressing control vector (at MOIs 3, 6, 9, or 12) to control for all final MOIs used here, or indicated combinations of lentiviral vectors expressing vIRF3, IRF4, BATF, and JUNB (each at MOI 3). Significance was calculated using one-way ANOVA followed by Tukey’s post hoc tests and is relative to the EBFP2 control matched for the total regulator-vector MOI. (B) ZsGreen reporter experiments as in panel A, but with the indicated pLVDP-mCherry-P2A-Puro-IRF4-SE^500bp (at equal MOI ∼3), and transduction with either an EBFP2 control vector (MOI 9) or vectors expressing vIRF3, IRF4, and BATF (each at MOI 3). Significance was calculated using two-way ANOVA followed by Tukey’s post hoc tests. Significance above individual bars is relative to EBFP2 for each reporter; indicated comparisons are between wt and mutant reporters. (C) ZsGreen reporter expression as in panel A but with wt and AP-1 mutant pLVDP-mCherry-P2A-Puro-MYC-SE^500kb (at equal MOI ∼3). Significance was calculated using two-way ANOVA followed by Tukey’s post hoc tests on log₂-transformed ZsGreen MFI values. (D) A subset of the results from panel C showing the individual-factor conditions to visualize lower levels of ZsGreen expression. In panel D, significance above individual bars is relative to the EBFP2 MOI 3 control for the corresponding reporter; brackets compare WT and AP-1-mutant reporters within each regulator condition. Throughout, error bars represent SD from three biological replicates. ns, not significant; *, adj. P < .05; **, adj. P < .01; ^****, adj. P < .0001.

Western blot

293 cells were transfected as described earlier with protein-expressing vector in six-well plates, scaling from other formats used in reporter assays by surface area. Cells were harvested 1 day after transfection. 293 cells were washed twice with phosphate buffered saline (PBS). The adherent cells were lysed in 100–200 μl RIPA buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 1% IGEPAL CA-630, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate) supplemented with protease inhibitors and then scraped and transferred into precooled 1.5-ml tubes. The lysates were incubated on ice for 20∼30 min and sonicated for six cycles (30 s on, 30 s off). Insoluble debris was removed by centrifugation at 16 000 × g for 20 min. The supernatants were transferred to fresh tubes and quantified using bicinchoninic acid (BCA) assays. BC-3 and BCBL-1 cells were subjected to the same procedure, except that the cells were centrifuged to remove the supernatant and then washed with PBS. For whole cell lysates, 20 μg (Fig. 11) or 10 μg protein (other figures) was resolved in Bis-Tris gels and then transferred to 0.2 μm (Fig. 11) or 0.45 μm (other figures) nitrocellulose membrane. Except for BATF, membranes were blocked using Tris-buffered saline (TBS, 20 mM Tris, 137 mM NaCl, pH 7.6) containing 5% w/v non-fat milk powder. Next, membranes were incubated with primary antibody in TBST (TBS containing 0.1% Tween-20) with 5% w/v non-fat milk powder. Membranes were washed three times with TBST for 15 min each time before incubating with IRDye-800 or IRDye-680-conjugated secondary antibody in TBST with 5% milk. Membranes were washed three times with TBST for 15 min each time, rinsed with TBS, and then imaged on a LI-COR Odyssey-M Imaging System. For BATF, we used the same protocol, but blocked membranes using LI-COR Intercept (TBS) Blocking Buffer and used Intercept® T20 (TBS) Antibody Diluent.

Figure 11.

BATF promotes vIRF3 and IRF4 occupancy of the endogenous MYC-SE^500kb. (A) Relative viability of sgBATF- and sgAAVS1-transduced BCBL-1 cultures on day 3 or day 6 after Cas9 induction. Error bars represent SD from four biological replicates, and significance was calculated using an unpaired t-test. ***, P = .0002. (B) Western blots analyzing the expression of the indicated proteins on day 3 after Cas9 induction in the same experiments as in panel A. (C) Quantification of western blot results as in panel B from four biological repeats. Significance was calculated using multiple paired t-tests (Holm-Šídák method). (D–O) ChIP-qPCR quantifying BATF, JUNB, IRF4, and vIRF3 occupancy at the endogenous MYC-SE^500kb, IRF4-SE^500bp, or a control locus (RPL30) in BCBL-1 on day 3 after Cas9 induction. Significance was calculated using one-way ANOVA followed by Tukey’s post hoc tests. Error bars represent SD from three biological replicates in panels D–O and four biological replicates in panel C. ns, not significant; *, adj. P < .05; **, adj. P < .01; ***, adj. P < .001; ^****, adj. P < .0001.

DNA pull-down assays

The 5′-biotinylated primers listed in Supplementary Table S2 were ordered from IDT. To generate probes, we performed multiple PCR reactions using the above pGL4.23 vectors as templates and in-house Taq polymerase. PCR products were precipitated using 1/10 volume of 3 M sodium acetate, 50 μg/ml GlycoBlue, and 3 volumes of 100% ethanol. The mixtures were incubated at −20°C for at least 1 h and then subjected to centrifugation for 30 min at 4°C at 16 000 × g. The pellets were washed twice with ice-cold 70% ethanol, air dried, and resuspended in nuclease-free water. The products were then agarose-gel-purified (IBI Scientific Gel Extraction & PCR Cleanup Kit) and eluted in nuclease-free water. The concentration of the probes was determined using a Qubit Fluorometer.

To prepare nuclear extracts (NEs) used in DNA pull-down assays, 1 × 10⁷ BC-3 cells were centrifuged and washed twice with cold PBS. The cell pellets were resuspended with 400 μl ice-cold Hypotonic buffer [10 mM HEPES-KOH, pH 7.9, 10 mM KCl, 0.1 mM ethyleneglycol- bis(β-aminoethyl)-N,N,Nʹ,Nʹ-tetraacetic acid (EGTA), 1.5 mM MgCl₂, 1 mM dithiothreitol (DTT), 0.5 mM phenylmethylsulfonyl fluoride] and incubated on ice for 10 min. After adding 25 μl 10% Igepal CA-630, samples were subjected to a quick spin to pellet the nuclear fraction. The nuclear pellets were resuspended with 40 μl nuclear extract buffer [20 mM HEPES-KOH, pH 7.9, 420 mM NaCl, 0.2 mM ethylenediaminetetraacetic acid (EDTA), 1.5 mM MgCl₂, 25% glycerol, 0.5 mM DTT, 0.5 mM PMSF, and protease inhibitor cocktail] and incubated on ice for 30 min with brief vortexing every 10 min. The samples were centrifuged at maximal speed for 12 min, the supernatants were collected and subjected to BCA assays for protein quantification.

Fifty micrograms of BC-3 NEs were incubated with 500 ng of biotinylated DNA probes overnight at 4°C in the pulldown buffer (10 mM HEPES-KOH, pH 7.9, 50 mM KCl, and protease inhibitors). Poly[d(I-C)] was added to a final concentration of 10 ng/μl to reduce nonspecific binding. For each sample, 20 μl Dynabeads M280 Streptavidin (Invitrogen #11205D) were pre-washed three times with washing buffer (10 mM HEPES-KOH, pH 7.9, 50 mM KCl, and 0.05% Tween-20) and resuspended with 40 μl washing buffer. The washed beads were added to the protein-probe mix and incubated on a rotator for 2 h at 4°C. The beads were then washed six times with the washing buffer and resuspended in 40 μl resuspension buffer (10 mM HEPES-KOH, pH 7.9, and 50 mM KCl). The resulting samples were subjected to western blot following the description above but using running buffer containing NuPAGE™ Antioxidant (1:400, Invitrogen NP0005) in the upper chamber. DNA pull-down assays using 293T NE were performed similarly, using either 25 or 50 μg NE, with similar relative results.

Nuclear co-immunoprecipitation assay

3 × 10⁶ 293T cells were plated per 10-cm dish 1 day before transfection. Six micrograms of each protein expression plasmid was transfected using the PEI method and the EBFP2 plasmid was used to adjust the total amount of DNA to 12 μg/dish. Two days after transfection, NEs were prepared using the Active Motif (Carlsbad, CA, USA) Nuclear Extract Kit (#40010) according to instructions. Protein concentrations were quantified using BCA assay (Pierce).

Co-immunoprecipitation (Co-IP) was performed with reagents provided with the Active Motif Nuclear Complex Co-IP kit (#54001). Specifically, 100 μg of NEs were incubated with 1.5 μl anti-IRF4 (Cell Signaling, #D9P5H, rabbit) in IP incubation buffer overnight at 4°C. Twenty microliters of Dynabeads protein A magnetic beads (Thermo Fisher, #10002D) were added the next day, and the mixture was incubated for another 2 h at 4°C. The beads were then washed six times and resuspended in IP wash buffer supplied with the kit. The Co-IP samples, together with NEs were analyzed by western blot, using rat anti-IRF4 (Thermo Fisher, clone 3E4, catalog 14-9858-82) as primary antibody to detect IRF4.

Electrophoretic mobility shift assay

5′ IRDye® 700-labeled oligonucleotides were purchased from IDT. The oligos were annealed in annealing buffer (10 mM Tris, pH 7.5–8, 50 mM NaCl, and 1 mM EDTA) in a thermocycler by heating at 95°C for 3 min and annealing at 37°C for 1 h. The resulting probes were diluted to 50 nM working concentration. 3 × 10⁶ 293T cells were plated per 10 cm dish 1 day before transfection. Three micrograms of each protein-expressing plasmid was transfected using the PEI method, using EBFP2 vector to adjust the total amount of DNA to 9 μg/dish. Two days after transfection, NEs were harvested using a nuclear extract kit (Active Motif #40010) following the manual. Protein concentrations were quantified using BCA assay (Pierce). EMSA was performed with reagents included in the Odyssey EMSA Kit (LI-COR, #82907910). Pre-mixed reagents [10× binding buffer, 25 mM DTT/2.5% Tween-20, 1 μg/μl Poly (dI-dC)], 5 μg of NEs, and probes were incubated on ice for 20 min. When performing supershifts, antibodies were added subsequently, and reactions were incubated for another 20 min on ice. Orange loading dye supplied in the kit was added, and samples were separated on DNA retardation gels (Invitrogen™, EC6365BOX) using 1× Tris-borate-EDTA (TBE) as the running buffer (Fisher, BP1333-1). Gels were imaged on an LI-COR Odyssey-M Imaging System.

Inducible BATF KO and ChIP-qPCR

BCBL-1 cells engineered for inducible Cas9 expression and transduced with lentiviral vectors expressing AAVS1 or BATF single guide RNAs (sgRNAs) were reported previously [8]. We used the more efficient BATF sgRNA 2 here, which has the sequence 5′-GGAGAUCAAGCAGCUCACAG. Cas9 was induced using 1 μg/ml doxycycline, and we assessed viability using the CellTiter-Glo® 2.0 Cell Viability Assay (Promega, #G9241), as instructed. We processed cells for ChIP-qPCR (chromatin immunoprecipitation coupled with qPCR) ~72 h after Cas9 induction, using the SimpleChIP Enzymatic Chromatin IP Kit (Magnetic Beads) (Cell Signaling Technology, #9003) largely as instructed and using antibodies listed in Supplementary Table S4. Briefly, we used 4 × 10⁶ cells per sample as instructed. For fixation, we centrifuged cells for 5 min at 360 × g and resuspended the loosened pellets in 1% freshly prepared formaldehyde (Thermo Fisher, #28908), followed by incubation for 8 min on a rocker, prior to quenching. We used 0.375 μl micrococcal nuclease per sample (included in the kit) to digest chromatin and disrupted the nuclei by four cycles of sonication using a Diagenode Bioruptor on the low setting (30 s on, 60 s off). Chromatin input was the equivalent of 8–10 μg DNA per IP, as determined using a Qubit 4 Fluorometer, and we used 10 μg of each antibody per IP. We performed qPCR using primers listed in Supplementary Table S1 with the SimpleChIP® Universal qPCR Master Mix (Cell Signaling Technology, #88989) and a QuantStudio 3 PCR System (Thermo Fisher), according to the manufacturers’ instructions.

Results

vIRF3/IRF4-responsive IRF4-SE sequences map to a complex 83 bp DNA element

We have previously used lentiviral EGFP-based reporter assays in 293 cells to demonstrate that vIRF3 and IRF4 positively regulate a 500-bp sequence centered on the overlapping vIRF3 and IRF4 ChIP-seq peaks in the PEL IRF4-SE, located ∼63 kb upstream of the IRF4 transcription start site [8]. However, vIRF3/IRF4-mediated reporter activation was relatively weak (∼2-fold) and increased moderately to ∼3-fold upon BATF co-expression. The requirement for vector titration furthermore makes the transduction-based assay cumbersome to scale for mutagenesis studies. Transfection-based luciferase reporter assays are more straightforward and can offer a higher dynamic range [45]. To facilitate mapping the sequence requirements for vIRF3/IRF4-mediated IRF4-SE activation, we first tested whether transfection-based dual-luciferase reporter assays recapitulate our published result. We inserted the published 500-bp IRF4-SE element (IRF4-SE^500bp, Supplementary Fig. S1) fragment upstream of the minimal promoter (min. P) in the firefly luciferase (FLuc) reporter plasmid pGL4.23. We then co-transfected the resulting reporter vector with an internal RSV-LTR-driven RLuc control vector (pGL4.72/RSV-LTR) and combinations of vIRF3-, IRF4-, and EBFP2-expressing vectors into 293 cells (Fig. 1A and B). None of the tested combinations affected RLuc expression from the RSV-LTR internal control reporter (Supplementary Fig. S2A). While IRF4 or vIRF3 alone did not significantly activate relative IRF4-SE^500bp FLuc reporter expression (Fig. 1C), IRF4-SE reporter expression was ∼15-fold induced after co-transfection of vIRF3 and IRF4 (Fig. 1C and Supplementary Fig. S2B). These results reproduce the cooperative activation of the IRF4-SE element by vIRF3 and IRF4, with greatly improved sensitivity.

Figure 1.

vIRF3/IRF4-responsiveness of the IRF4-SE in plasmid-based reporter assays. (A) Schematic of the approach used for dual-luciferase reporter assays. See text for abbreviations. (B) Western blot analyses of IRF4 and vIRF3 expression 1 day after transfection of the indicated expression vectors into 293, under the conditions used for reporter assays. GAPDH served as a loading control. BC-3 served as a control for expression levels in PEL cells. Representative of two independent repeats. (C) Results from dual-luciferase reporter assays as described in panel A. We sequentially normalized results from FLuc reporters to those from the RSV-RLuc internal control and then results from TF combinations to those from EBFP2, which we set to 1. Significance is relative to EBFP2 unless indicated by lines. (D) Results from dual-luciferase reporter assays as described in panel C, except that different combinations of BATF and JUNB were included. (E) Western blot analyses of 293 cells transfected with the indicated expression vectors under conditions used in panel D. GAPDH served as a loading control. BC-3 is a control for expression levels in PEL cells. BJAB is a control for expression levels in a KSHV-negative B-cell lymphoma cell line. Representative of two independent repeats. Throughout, error bars represent SD from three biological replicates. ns, not significant; **, adj. P < .01; ^****, adj. P < .0001, calculated using one-way analysis of variance (ANOVA) followed by Tukey’s post hoc tests.

In the transfected reporter assay, the additional co-transfection of BATF and/or JUNB did not increase IRF4-SE^500bp reporter activity, indicating that co-expressed BATF and JUNB were not required under these transfected-reporter conditions. BATF and/or JUNB instead inhibited vIRF3/IRF4-mediated IRF4-SE element activation in this assay (Fig. 1D and E; Supplementary Fig. S3). We chose JUNB here, since it is the preferred interaction partner of BATF in other settings and more highly expressed in PEL cells than other JUN family members ([8, 46], and Supplementary Fig. S3).

Given that the transfection-based assay nevertheless efficiently captures the cooperation between vIRF3 and IRF4, we introduced ∼50 bp deletions to begin mapping the sequences that confer vIRF3/IRF4-dependent IRF4-SE^500bp reporter activation. We co-transfected resulting mutants with expression vectors for the EBFP2 negative control or vIRF3 and IRF4. IRF4-SE^500bp mutants lacking either nucleotides (nts) 51–100 or 101–151 showed strongly reduced reporter activity (Fig. 2A). In contrast, other deletions had a minimal or no impact, suggesting that nt 51–151 harbor the vIRF3/IRF4-responsive element(s). Smaller 18–22 bp deletions within this region, except for Δ51–68 bp, impaired vIRF3/IRF4-mediated reporter activation (Fig. 2B), suggesting that a relatively large ∼83 bp region spanning nt 69–151 mediates vIRF3/IRF4-mediated reporter activation. This fragment (IRF4-SE^69-151) was sufficient for full reporter responsiveness to vIRF3/IRF4 in 293 cells (Fig. 2C) and, therefore, represents a minimal vIRF3/IRF4-responsive element within the IRF4-SE.

Figure 2.

The vIRF3/IRF4-responsive element within the IRF4-SE maps to an 83-bp region. (A) Results from dual-luciferase reporter assays as in Fig. 1C, but co-transfecting wt or 50-bp IRF4-SE^500bp deletion mutant reporters with EBFP2, or vIRF3 and IRF4. For each reporter, results from vIRF3/IRF4-cotransfected samples were normalized to those from EBFP2, as in Fig. 1C, and additionally to the result from the wt IRF4-SE^500bp reporter. min. P: minimal promoter-only Fluc control vector. Significance is indicated relative to the wt vector. (B) Results from dual-luciferase reporter assays as described in panel A, measuring the effect of 18–22-bp deletions between nt 51–151 of the IRF4-SE^500bp. Significance is indicated relative to the wt vector. The reduction of Δ69–86 activity is significant when compared to IRF4-SE^500bp using a one-way ANOVA followed by Dunnett’s post hoc tests (adj. P = .037), and we therefore included this region in panels (C–F). (C) Results from dual-luciferase reporter assays as described in panel A but comparing the IRF4-SE^500bp reporter and a reporter containing nt 69–151 of IRF4-SE^500bp (IRF4-SE^69-151). Significance is relative to min. P unless indicated by lines. (D) Schematic of the pLS-mP [41] lentiviral reporter vector used in this figure, showing wt 5′LTRs, packaging signal (Ψ), Rev response element, antirepressor element #40, a scaffold-attached region, IRF4-SE element, minimal promoter (min. P), EGFP coding sequence, and self-inactivating 3′LTR. (E) Representative histogram comparing lentiviral EGFP reporter expression 3 days after transduction of BC-3 cells with reporters containing the wt IRF4-SE^500bp sequence or the indicated mutants at equal MOI (∼1). (F) Quantification of EGFP mean fluorescence intensity of EGFP-positive BC-3, defined based on an untransduced negative control, over three independent experiments as in panel E. Significance is relative to the wt IRF4-SE^500bp reporter. Throughout, error bars represent SD from three biological replicates for bar graphs. ns, not significant; *, adj. P < .05; **, adj. P < .01; ***, adj. P < .001; ^****, adj. P < .0001, calculated using one-way ANOVA followed by Tukey’s post hoc tests.

To test whether the identified minimal region is similarly critical for IRF4-SE activity in PEL cells, we transduced the PEL cell line BC-3 with lentiviral EGFP-based enhancer element reporters containing the wt IRF4-SE^500bp fragment or Δ51–100, Δ101–151, Δ51–151, or Δ69–151 mutants from above (Fig. 2D–F). Each deletion strongly reduced IRF4-SE^500bp reporter expression, confirming the regulatory importance of the minimal element in PEL cells. Thus, results from both 293 and the PEL cells point to the critical importance of a ∼83-bp minimal vIRF3/IRF4-responsive element within the IRF4-SE, and we therefore focused our subsequent experiments on this region. We note that the minimal IRF4-SE^69-151 element itself was also active in PEL cells, which is shown below in Fig. 6A and B.

Figure 6.

The AP-1 motif is critical for IRF4-SE reporter activity in the BC-3 PEL cell line. (A) Representative histogram comparing EGFP expression 3 days after transduction of BC-3 cells with pLS-mP lentiviral reporters containing the indicated wt or mutant IRF4-SE^69-151 elements at equal MOI (∼1). Quantified over three repeats in panel B. (B) Quantification of EGFP mean fluorescence intensity of EGFP-positive BC-3, defined based on an untransduced negative control, over three independent experiments as in panel A. (C) Schematic of the dual-reporter lentiviral construct used in panels D-F. Abbreviations in addition to those in Fig. 2D are scaffold/matrix attachment (S/MAR) region 8 (sMAR8), herpes simplex virus thymidine kinase polyadenylation signal (HSV TK pA), puromycin resistance cassette (Puro), P2A ribosomal skipping peptide, mCherry coding sequence (mCherry), and chicken hypersensitive site 4 (cHS4) from the β-globin locus. T_actb indicates a sequence downstream of the synthetic pA that facilitates efficient transcription termination of the upstream ZsGreen reporter cassette. Wavy lines indicate expected transcripts after transduction. (D) Quadrant plots to illustrate mCherry and ZsGreen reporter expression in puromycin-selected BC-3 cells 3 days after transduction with the indicated pLVDP-mCherry-P2A-Puro vectors at MOI 0.3. Gated for single, live cells. Numbers in quadrants indicate percentages of the parent population. Representative of three independent repeats, which are further analyzed and quantified in panels E and F. (E) Representative histogram comparing ZsGreen reporter expression in transduced cells as in panel D. Gated as in panel D and additionally for mCherry-positive cells. (F) Quantification of ZsGreen mean fluorescence intensity in transduced BC-3 cells over three independent experiments as in panel E. Significance was compared to min. P. ZsGreen expression from all mutant constructs was furthermore significantly reduced compared to IRF4-SE^500bp, adj. P < .0001, determined using one-way ANOVA followed by Tukey’s post hoc tests. Throughout, error bars represent SD from three biological replicates for bar graphs. ns, not significant; ^****, adj. P < .0001 from one-way ANOVA followed by Tukey’s post hoc tests.

The minimal vIRF3/IRF4-responsive IRF4-SE element lacks a functional AICE motif

Inspection of the identified minimal vIRF3/IRF4-responsive element within the IRF4-SE revealed four IRF/GAAA motifs (Fig. 3A), including one within a candidate ISRE-like motif (TGAAnnGAAA, underlined nucleotides diverge from the consensus). IRF4 tolerates one and even two-nucleotide variations from the IRF/GAAA consensus in the context of composite sites [31, 35], although the sequence above would likely have considerably lower affinity to IRF4 than a consensus ISRE. The second IRF motif is close to a canonical AP-1 motif (TGAnTCA), but not within the 4-nt spacing that is typical for an AICE1 site, which has the consensus sequence TGAnTCA-n4-GAAA. AICE and ISRE motifs are binding sites for IRF4/BATF/JUN complexes and IRF4 homodimers, respectively. In contrast, isolated IRF motifs are not thought to recruit IRF4. EMSAs showed that the IRF4-SE candidate AICE1-like sequence can bind to the BATF-JUNB AP-1 complex but fails to subsequently recruit IRF4 (Fig. 3B). In contrast, we obtained the expected IRF4 recruitment upon BATF-JUNB AP-1 complex binding by a positive control AICE1 motif from IL10 (Fig. 3C), as reported [35]. These results show that the candidate AICE1-like sequence in the minimal vIRF3/IRF4-responsive element is not a functional AICE motif.

Figure 3.

The minimal vIRF3/IRF4-responsive sequence in the IRF4-SE contains a functional AP-1 site, but not an AICE motif. (A) Sequence of nt 69–151 of the IRF4-SE^500bp element. Arrows mark the orientation of candidates for known motifs. (B) EMSAs using a probe containing a candidate AICE1-like motif from the IRF4-SE and NEs from transfected 293T cells expressing combinations of EBFP2, IRF4, BATF, and JUNB as indicated. Arrows mark the orientation of the AP-1 and IRF motifs. Supershift assays were performed using antibodies listed above the panel. Representative of two independent repeats. (C) EMSA as in panel B but using a positive control probe containing the canonical AICE1 site from IL10. Representative of two independent repeats.

The DNA-binding capacity of IRF4 is critical for cooperative IRF4-SE activation

The identified minimal vIRF3/IRF4-responsive element contains several IRF-related motifs, including a candidate ISRE-like site (Fig. 3A), suggesting that IRF4 directly binds these elements to activate reporter expression. We therefore used a published IRF-motif-binding-deficient mutant of IRF4 (IRF4^R98A/C99A) [33, 34, 47] to ask whether the IRF4 DNA-binding capacity is important for cooperative reporter activation by this element. In DNA pulldown assays, a biotinylated fragment spanning the minimal vIRF3/IRF4-responsive region (IRF4-SE^69-151) and an additional 15 bp on either side associated with wt but not R98A/C99A-mutant IRF4, confirming the DNA-binding deficiency of IRF4^R98A/C99A (Fig. 4A). The biotinylated IRF4-SE probe recovered IRF4 independently of vIRF3 co-transfection. In turn, the biotinylated probe recovered vIRF3 independently of whether it was transfected with wt or R98A/C99A-mutant IRF4. Interestingly, wt IRF4 and vIRF3 showed a lower degree of association with a randomized biotinylated DNA fragment (N6, Supplementary Table S2) designed to lack any IRF/GAAA motifs or other known TF-binding sites. This negative control sequence did not activate Fluc reporter expression after co-transfection with vIRF3 and IRF4 (Fig. 4B). We observed similar background association with other biotinylated negative control probes but only marginal recovery in the presence of unbiotinylated probes (see below for quantification). Co-IP of IRF4 and vIRF3 from NEs, as published [8], shows that the IRF4^R98A/C99A mutation does not affect IRF4 association with vIRF3 (Fig. 4C). In reporter assays, the DNA-binding-deficient IRF4 mutant was unable to cooperate with vIRF3 to activate either IRF4-SE^500bp or the minimal IRF4-SE^69-151 region (Fig. 4D and E). Interestingly, we observed significant regulation of the minimal IRF4-SE^69-151 element by vIRF3 alone, and the co-transfection of R98A/C99A-mutant IRF4 did not interfere with this activation by vIRF3 (Fig. 4E). Together, these results suggest that the ability of IRF4 to associate with DNA is critical for its cooperation with vIRF3 on the IRF4-SE element.

Figure 4.

The DNA-binding capacity of IRF4 is required for its cooperation with vIRF3. (A) DNA pull-down assays using NEs from 293T transfected with the indicated protein expression vectors confirm that DNA-binding-deficient IRF4^R98A/C99A cannot associate with a biotinylated probe comprised of nt 54–166 of the IRF4-SE^500bp (IRF4-SE^54-166, marked by a star). W and M denote wt and R98A/C99A-mutant IRF4, respectively. N6 denotes a biotinylated randomized control probe (see Supplementary Table S2), B- denotes an unbiotinylated IRF4-SE^54-166 probe. Recovered proteins were analyzed by western blotting for IRF4 and vIRF3. Representative of two independent repeats. (B) Results from dual-luciferase reporter assays as described in Fig. 2A, showing that the N6 control sequence does not respond to co-transfection of vIRF3 and IRF4. IRF4-SE^69-151 and minimal promoter (min. P) served as positive and negative controls, respectively. Significance is relative to IRF4-SE^69-151 unless indicated by lines. (C) Nuclear Co-IP with anti-IRF4 antibody, as published [8], but using nuclear lysates of 293T that were transfected with the indicated plasmids, shows that IRF4^wt (W) and IRF4^R98A/C99A (M) associate with vIRF3 similarly. Co-precipitates were analyzed by western blotting for IRF4 and vIRF3. Representative of two independent repeats. (D) Results from dual-luciferase reporter assays in 293 using the IRF4-SE^500bp reporter, as described in Fig. 1C, but including the IRF4^R98A/C99A mutant to compare the effect of IRF4^wt and IRF4^R98A/C99A. (E) As in panel D but using the IRF4-SE^69-151 reporter. Significance is relative to EBFP2 unless indicated by lines. Throughout, error bars represent SD from three biological replicates. ns, not significant; ***, adj. P < .001; ****, adj. P < .0001, calculated using one-way ANOVA followed by Tukey’s post hoc tests.

Separate sequences mediate IRF4-SE activation by vIRF3 and cooperation with IRF4

Given the importance of IRF4’s ability to bind to DNA for its cooperation with vIRF3, we directly tested the roles of the IRF/GAAA-related motifs and the AP-1 motif in the context of the minimal IRF4-SE^69-151 reporter. For this, we mutated each IRF/GAAA motif alone, the four IRF/GAAA motifs together, the adjacent IRF/GAAA #2 and AP-1 motifs together, the ISRE-like motif, the AP-1 motif, and the TGAA candidate ISRE half-site (Fig. 5A). Upon co-transfection of 293 with expression vectors encoding vIRF3 and IRF4, mutation of each individual IRF/GAAA motif led to a modest, but significant, reduction in reporter activity (Fig. 5B). Mutating the four IRF/GAAA motifs together strongly decreased IRF4-SE^69-151 activity. Mutation of both the IRF/GAAA #2 and AP-1 sites reduced reporter activity to background level. Mutation of the AP-1 site (TGAnTCA) had a significantly stronger effect than eliminating the closely spaced IRF motif (IRF/GAAA #2) (adj. P = .0047, one-way ANOVA with Tukey’s multiple comparisons test). In contrast, the strong reduction in reporter activity after mutating the GAAA #2 and AP-1 sites together or the AP-1 site alone was not significantly different. Mutating the ISRE-like motif or either candidate half site (i.e. IRF/GAAA #3 or TGAA) strongly impaired reporter activity, and any differences between these three mutants did not reach significance in this assay. Interestingly, only mutations that disrupted the AP-1 motif caused significantly reduced reporter activation by vIRF3 alone (Fig. 5C). These results suggest that the AP-1 motif is critical for the response to vIRF3, while the IRF/GAAA-related motifs, including three apparently isolated IRF/GAAA motifs and an ISRE-like motif, are critical for the cooperation of vIRF3 with IRF4. Thus, the elements required for the basic response to vIRF3 and its cooperation with IRF4 are genetically separable. Importantly, the AP-1 site and the IRF/GAAA and ISRE-like motifs do not form a known composite site but instead cooperate within a complex and extended element.

Figure 5.

The vIRF3 responsiveness of the IRF4-SE depends on the AP-1 site, while cooperation with IRF4 depends on IRF/GAAA motifs, including a candidate ISRE. (A) Sequences of wt IRF4-SE^69-151 and mutants used in this figure. IRF/GAAA motifs, the AP-1 site, and the TGAA IRF-like half site of the candidate ISRE are marked by arrows. Mutated nucleotides are in lowercase. Unchanged nucleotides are shown as lines. (B) Dual-luciferase reporter assays showing relative wt and motif-mutant IRF4-SE^69-151 reporter activities following co-transfection with vIRF3 and IRF4. Normalization to results obtained with EBFP2 was performed as in Fig. 2A, and significance compared to the wt reporter was determined using one-way ANOVA followed by Dunnett’s post hoc tests. Error bars represent SD from three biological replicates for bar graphs. ns, not significant; ^****, adj. P < .0001. (C) Dual-luciferase reporter assays as described in panel B but testing transfection of vIRF3 only. Normalized as in panel B. Results from the GAAA #2 & AP-1 and AP-1 mutants differed significantly from the wt IRF4-SE^69-151, determined using one-way ANOVA followed by Dunnett’s post hoc tests. Error bars represent SD from three biological replicates for bar graphs. ns, not significant; *, adj. P < .05.

IRF/GAAA, ISRE-like, and AP-1 motifs mediate IRF4-SE activity in PEL cells

We next confirmed the importance of the IRF/GAAA motifs, the ISRE-like motif, and the AP-1 motif using lentiviral EGFP-based IRF4-SE^69-151 reporters in the PEL cell line BC-3 (Fig. 6A and B), using the vector shown in Fig. 2D. While the minimal IRF4-SE^69-151 element was sufficient to increase EGFP expression in BC-3, reporter activation was weaker than with the IRF4-SE^500bp element (compare Figs 2D–F and 6A, B), not allowing for efficient distinction of EGFP-expressing cells from untransduced cells. Notably, IRF4-SE^69-151 showed improved activity over IRF4-SE^500bp in 293 cells. The reduced activity of IRF4-SE^69-151 in BC-3 may reflect differences between the transfection- and transduction-based assays, the slightly different ratios of vIRF3 and IRF4 expression we achieve in 293 compared to those in BC-3 (Fig. 1B), the different cellular context, and positive regulatory roles for sequences outside the core element in BC-3 that are not captured in 293. It is furthermore possible that sequences outside nt 69–151 are subject to unphysiological negative regulation in 293. Since SEs are defined by their large size and contain densely packed regulatory elements, the presence of additional regulatory elements in the IRF4-SE is expected. Despite this, mutations in this construct confirm the importance of the adjacent IRF/GAAA #2 and AP-1 sites and the ISRE-like motif in IRF4 reporter activation in PEL cells (Fig. 6A and B).

To more efficiently distinguish transduced from untransduced cells and thereby increase the resolution of this assay, we developed a lentiviral vector that combines a promoter/enhancer element-driven ZsGreen reporter cassette with an insulated mCherry-P2A-Puro^R expression cassette (Fig. 6C), based on a reported design [48]. Inclusion of the mCherry-P2A-Puro^R cassette enables convenient lentiviral titration and reporter analysis upon selection with puromycin and/or gating for mCherry-expressing cells (Fig. 6D and Supplementary Fig. S4). Using this approach, we showed that deletion of the minimal 83-bp responsive element (nt 69–151), mutation of the AP-1 site, the 4 GAAA motifs, or the ISRE-like motif in the context of the IRF4-SE^500bp reporter strongly impaired its activity in BC-3 and BCBL-1 PEL cells (Fig. 6D–F; Supplementary Fig. S4B and C), confirming the requirement for these sequences in PEL cells. The IRF4-SE^500bp element did not activate reporter expression in the KSHV-negative B cell line BJAB (Supplementary Fig. S4D and E), confirming the published PEL-specific activity of this element [8]. Interestingly, in the context of this lentiviral vector, a shorter IRF4-SE^200bp element, which includes the first 200 bp of IRF4-SE^500bp, conferred robust activation in BCBL-1 but not BC-3 cells, suggesting that pLVDP may present a more stringent context than pLS-mP above (Fig. 6E and F; Supplementary Fig. S4B and C).

The association of the IRF4-SE with vIRF3 and IRF4 is genetically separable

To explore the role of the identified motifs in recruiting IRF4 and vIRF3 to the minimal vIRF3/IRF4-responsive element, we generated biotinylated DNA fragments carrying a subset of the IRF- and AP-1-related mutations. Pulldowns with two biotinylated negative controls showed similar background association with IRF4 and vIRF3 in BC-3 NEs (probes N2, N6, Fig. 7A–C). The WT IRF4-SE fragment recovered 2–3-fold more IRF4 and vIRF3 than the negative control pulldowns, suggesting sequence-specific association of these proteins, despite background binding. Interestingly, mutation of the four GAAA motifs or the ISRE-like motif reduced the association of IRF4 to background levels. These results suggest that the sequence-specific recruitment of IRF4 depends on GAAA motifs, and, most importantly, an ISRE-like element with a non-canonical TGAA 5′-half site. The 4GAAA or ISRE-like mutations did not reduce sequence-specific association with vIRF3 (Fig. 7C). Instead, mutation of the IRF/GAAA #2 and AP-1 sites significantly reduced association with vIRF3, consistent with the importance of the AP-1 motif in the assays above. Mutation of the IRF/GAAA #2 and AP-1 sites did not decrease IRF4 association in this assay. Together, these data suggest that the association of vIRF3 and IRF4 with the minimal IRF4-SE element is genetically separable, at least in the context of transfected reporters and this in vitro assay using biotinylated DNA fragments.

Figure 7.

vIRF3 and IRF4 association with the vIRF3/IRF4-responsive elements depends on distinct motifs. (A) DNA pull-down assays, as in Fig. 4A, but using BC-3 NEs and a subset of the motif mutants from Fig. 5. Probes are biotinylated unless marked with (B-). N2/6 are randomized negative control probes (see Supplementary Table S2). Representative of three independent repeats, which are quantified in panels B and C. Quantification of IRF4 (B) and vIRF3 (C) pulldown efficiency from three independent experiments as shown in (A). Significance is relative to wt. Error bars represent SD from three biological replicates. ns, not significant; *, adj. P < .05; **, adj. P < .01; ***, adj. P < .001; ^****, adj. P < .0001 from one-way ANOVA followed by Tukey’s post hoc tests.

IRF- and AP-1-related motifs are insufficient for the vIRF3/IRF4 responsiveness of the IRF4-SE

Due to the critical role of the IRF, ISRE-like, and AP-1 motifs, we next asked if these sites alone are sufficient to confer vIRF3/IRF4 responsiveness in an otherwise randomized context. We generated reporters with randomized sequences of all nucleotides outside the motifs mentioned above (R1 and R2). We additionally generated reporters that preserved known motifs but randomized shorter sequences separating them (R3–R5, Fig. 8A). Of these reporters, only R4, which preserves sequences separating the AP-1 and ISRE-like motifs, had detectable reporter activity in 293 cells, achieving ∼30% of wt reporter activation upon co-transfection of vIRF3 and IRF4 (Fig. 8B). Taken together with results from above, this result suggests that the IRF, ISRE-like, and AP-1 motifs are critical for the activity of the minimal vIRF3/IRF4-responsive element but insufficient. In BC-3 cells, none of the partially randomized mutants substantially activated the lentiviral IRF4-SE^69-151 EGFP reporter (Fig. 8C and D), confirming that the IRF- and AP-1-related motifs are also insufficient in this context. The differential activity of R4 in 293 and BC-3 may be due to reporter integration status and/or differences in cellular context, including the availability or stoichiometry of vIRF3, IRF4, and cofactors.

Figure 8.

The identified motifs are insufficient for full vIRF3/IRF4 responsiveness of the IRF4-SE. (A) Schematic of IRF4-SE^69-151 mutants with intact identified motifs and fully or partially randomized context (R1–R5). Mutant sequences are in lowercase with gray shading. Previously identified motifs are in color and unchanged. (B) Dual-luciferase reporter assays showing relative wt and randomized-mutant IRF4-SE^69-151 reporter activities following co-transfection with vIRF3 and IRF4. Normalization to results from EBFP2 and wt IRF4-SE^69-151 was performed as in Fig. 2A. Significance is indicated relative to the wt vector unless indicated for specific comparisons. (C) Histogram comparing EGFP expression 3 days after transduction of BC-3 cells with pLS-mP lentiviral reporters containing the indicated wt or mutant IRF4-SE^69-151 elements at equal MOI (∼1). Representative of three independent repeats. (D) Quantification of EGFP mean fluorescence intensity in EGFP-positive BC-3 cells, defined based on an untransduced negative control, over three independent experiments as in panel C. (E) DNA pull-down assays, as above, using the indicated sequences and BC-3 NEs. Probes were biotinylated, except when marked (B-). N6 and N2 are randomized negative control probes as in Fig. 7A and Supplementary Table S2. Representative of three independent repeats, which are quantified in panels F and G. Quantification of vIRF3 (F) and IRF4 (G) pulldown efficiency from three independent experiments as shown in panel (E). Significance is relative to wt unless indicated. (H) Dual-luciferase reporter assays showing relative wt and randomized-mutant IRF4-SE^69-151 reporter activities following expression of vIRF3 only. Normalization to results from EBFP2 and the wt reporter was performed as in Fig. 2A. Significance is relative to the wt reporter, unless indicated for specific comparisons. (I) Dual-luciferase reporter assays showing relative wt and short IRF4-SE fragment reporter activities following transfection with vIRF3 only. Normalization to results from EBFP2 and the wt reporter was performed as in Fig. 2A. Significance is relative to the wt reporter, unless indicated for specific comparisons. Throughout, error bars represent SD from three biological replicates for bar graphs. ns, not significant; *, adj. P < .05; **, adj. P < .01; ***, adj. P < .001; ^****, adj. P < .0001 from one-way ANOVA followed by Tukey’s post hoc tests.

To further characterize the role of the sequences separating the IRF-related motifs, we subjected mutants R1, R3, R4, and R5, together with the N2 and N6 control fragments, to DNA pulldown assays with BC-3 NEs. Sequences R1, R3, and R5 did not specifically associate with vIRF3, while mutant R4, which preserves the sequences separating the AP-1 and ISRE-like motifs and was vIRF3/IRF4-responsive in 293 reporter assays, showed wt levels of vIRF3 association (Fig. 8E and F). Interestingly, mutant R4 failed to specifically recruit IRF4 from BC-3 NEs, although it retains all IRF/GAAA-related motifs and the ISRE-like element, potentially explaining why this reporter was not active in BC-3 cells (Fig. 8E and G). These pulldown data suggest that the sequences separating the AP-1 and ISRE-like sites promote recruitment of vIRF3 to the IRF4-SE. Accordingly, mutant R4 was fully responsive to vIRF3 alone in the context of the minimal IRF4-SE^69-151 reporter (Fig. 8H). Furthermore, shorter fragments that retained the AP-1 motif but lacked any IRF motifs responded to vIRF3 (Fig. 8I). Interestingly, a fragment encompassing nt 87–129 of the IRF4-SE^500bp conferred significantly greater activation than a fragment including only nt 87–107, suggesting that the role of the sequences separating the AP-1 site and the ISRE-like motif is not simply due to the immediate sequence context downstream of the AP-1 site but involves a relatively extended sequence. Together with the data above, these results confirm that vIRF3 and IRF4 recruitment to the IRF4-SE depends on genetically separable sequence features and requires not only IRF, ISRE-like, and AP-1 motifs but also additional sequences.

The PEL MYC-SEs are vIRF3/IRF4-responsive SEs

We next sought to extend our analyses to other candidates for vIRF3/IRF4-responsive SEs. For this, we chose the two PEL MYC-SEs, located ∼500 and 375 kb downstream of the MYC transcription start site, respectively, due to their prominent occupancy by vIRF3 and IRF4 and the central importance of MYC for PEL cell survival [7, 8, 19, 20, 49]. We inserted the SE sequences spanning the overlapping vIRF3/IRF4 ChIP-seq peaks into the transfected firefly reporter vector (Supplementary Fig. S5). As for the IRF4-SE, vIRF3 and IRF4 cooperatively activated both MYC-SEs (Fig. 9A and B). Interestingly, the downstream MYC-SE^500kb element generated much higher reporter activity in 293 cells than the upstream MYC-SE^375kb element. While the MYC-SE^500kb reporter was furthermore substantially activated by vIRF3 alone, the MYC-SE^375kb reporter was not. Neither element was activated by IRF4 alone. Forty base-pair deletion mutants mapped the region required for the vIRF3/IRF4 responsiveness of the stronger MYC-SE^500kb element to an ∼87-bp region spanning nt 121–207 (Fig. 9C), similar in size to the ∼83-bp vIRF3/IRF4-responsive region within the IRF4-SE. This region contains a canonical AICE2 motif near its 5′ end (Fig. 9D), which we confirmed as functional using EMSAs (Supplementary Fig. S6A and B). Mutating the AICE2 motif greatly reduced reporter activation after co-transfection of vIRF3 and IRF4, while AP-1 and IRF half-site mutants had more subtle, but significant effects (Fig. 9E). In contrast, the response to vIRF3 depended as much on the AP-1 half-site as the entire AICE2 motif, while mutating the IRF half-site had no significant effect (Fig. 9F). These results suggest that vIRF3 requires the AP-1 half-site of the AICE2 motif, while cooperation with IRF4 may in part depend on the IRF half-site.

Figure 9.

vIRF3/IRF4-responsiveness of the two PEL MYC-SEs. (A) Dual-luciferase reporter assays showing the relative activity of an FLuc reporter containing a 327-bp element centered on the vIRF3/IRF4 ChIP-seq peaks in the MYC-SE located ∼500 kb downstream of the MYC transcription start site (MYC-SE^500kb, sequence in Supplementary Fig. S5A), after co-transfection with expression vectors for the indicated proteins into 293. Results were normalized as in Fig. 1C, i.e. first to results from the RSV-LTR-RLuc reporter and then to results from EBFP2, which were set to 1. Results are from seven independent repeats. Significance is relative to the EBFP2 control unless indicated by lines. (B) As in panel A but using a reporter containing a 299-bp fragment centered on the vIRF3/IRF4 ChIP-seq peaks in the MYC-SE located ∼375 kb downstream of the MYC transcription start site (MYC-SE^375kb, sequence in Supplementary Fig. S5B). Data were normalized as in panel A. Results are from five independent repeats. Significance is relative to the EBFP2 control. (C) Results from dual-luciferase reporter assays showing wt or mutant MYC-SE^500kb activities after expression of vIRF3 and IRF4. Results from vIRF3/IRF4 co-transfection were normalized as in Fig. 2A, i.e. to results from the internal RSV-LTR-RLuc control, EBFP2, and the wt construct, which was set to 1. Results are from three independent repeats. Significance is relative to the wt reporter. (D) Sequence of MYC-SE^500kb nt 121–207. The candidate AICE2 site is marked in color. (E) Dual-luciferase reporter assays as in panel C but using wt or AICE2 site mutant MYC-SE^500kb. Wt and mutant AICE2 motif sequences are next to each bar. Results are from four independent repeats. Significance is relative to the wt reporter. (F) Dual-luciferase reporter assays as in panel E but after expression of vIRF3 or EBFP2 only. Results are from four independent repeats. Significance is relative to the wt reporter. (G) Dual-luciferase reporter assays as in panel E but using wt and mutant MYC-SE^375kb reporters. The sequence of the wt or mutant AP-1 motif is next to each bar. Results are from three independent repeats. Significance is relative to the wt reporter. (H) Histogram comparing ZsGreen reporter expression in BC-3 transduced with wt and mutant pLVDP-mCherry-P2A-Puro-MYC-SE^500kb. Gated for single, live, mCherry-positive cells. Representative of three independent repeats, which are quantified in panel I. (I) Quantification of ZsGreen mean fluorescence intensity in mCherry-positive BC-3 cells over three independent experiments as in panel H. Significance is relative to min. P unless indicated by lines. Data were acquired together with Fig. 6D, and these experiments therefore share the min. P control. Throughout, error bars represent SD from at least three biological replicates for bar graphs, as specified in each panel. ns, not significant; *, adj. P < .05; **, adj. P < .01; ***, adj. P < .001; ^****, adj. P < .0001 from one-way ANOVA followed by Tukey’s post hoc tests.

In contrast to the IRF4-SE, co-transfection of vIRF3 and IRF4 with BATF alone moderately improved activation over that achieved by transfection of vIRF3 and IRF4 alone, while co-transfection of both BATF and JUNB did not enhance the response of the MYC-SE^500kb element to either vIRF3 alone or vIRF3 and IRF4 together (Supplementary Fig. S6C). Despite a functional AICE2 motif, co-transfection of IRF4, BATF, and JUNB was furthermore not sufficient to activate the transfected MYC-SE^500kb reporter in 293 cells (Supplementary Fig. S6C). The less active MYC-SE^375kb reporter does not contain a canonical AICE motif. Mutating a candidate AP-1 site within MYC-SE^375kb furthermore did not affect reporter activity after IRF4 and vIRF3 co-transfection (Fig. 9G); however, we note that this reporter contains additional motifs that match the AP-1 consensus less stringently (see Supplementary Fig. S5B for the sequence of this element). Using the dual lentiviral reporter vector from Fig. 6C, we confirmed the requirement for the MYC-SE^500kb AP-1 site in the PEL cell line BC-3 (Fig. 9H and I; Supplementary Fig. S6D). As for the IRF4-SE, shortening the reporter to a fragment spanning only nucleotides 81–247 of the original vector reduced its activity, suggesting that additional positive regulatory sequences exist in the 327 bp MYC-SE^500kb enhancer element.

BATF promotes vIRF3- and IRF4-dependent activation of integrated lentiviral reporters

BATF and/or JUNB co-transfection did not promote the activation of a plasmid-based IRF4-SE reporter (Fig. 1D and Supplementary Fig. S3B), while we had previously observed modestly enhanced activation of an integrated reporter upon BATF co-expression [8]. In contrast, BATF co-transfection activated a transfected MYC-SE^500kb reporter about twofold over the activity we obtained with vIRF3 and IRF4 only. These results suggest that the role of BATF may depend on the presence of a functional AICE motif as well as the integration status of the reporter. To test this idea directly, we performed integrated lentiviral reporter assays using the dual-color lentiviral reporter vectors from Fig. 6C. In these experiments, we observed significant IRF4-SE^500bp reporter activation only upon co-expression of vIRF3, IRF4, and BATF (Fig. 10A). These results further suggest that pLVDP is a more stringent context for activation than pLS-mP, which was moderately activated by the combined expression of vIRF3 and IRF4 in side-by-side experiments as published (not shown), potentially due to the inclusion of additional insulator/regulatory elements in pLVDP. Deletion of nt 69–151 or mutation of the 4 GAAA motifs, the AP-1 motif, or the ISRE-like motif reduced pLVDP-IRF4-SE^500bp reporter activity to background levels (Fig. 10B), validating the importance of the sequences identified in transfection-based assays as critical in this context.

Similar experiments with the MYC-SE^500kb enhancer element showed that BATF also enhanced the vIRF3/IRF4-mediated activation of this element (Fig. 10C), although we additionally obtained measurable activation by IRF4 alone, vIRF3 alone, and co-expression of IRF4 and vIRF3 (Fig. 10C and D, see Supplementary Tables S5 and S6 for a full statistical analysis). These experiments furthermore confirmed the importance of the AICE2 AP-1 half-site in the integrated lentiviral context.

The preferential requirement for BATF in integrated reporter assays may reflect a role of BATF on chromatinized templates and is consistent with the pioneer-like activity of BATF [50, 51], which has been shown to bind non-accessible sites and promote chromatin accessibility in T cells. While transfected plasmids associate with nucleosomes, their chromatinization is more variable and less dense than that of genomic sites [52]. At the MYC-SE, where we detect enhancement by BATF in both transfected and integrated reporters, BATF likely additionally facilitates IRF4 recruitment at the AICE2 motif through direct interaction [35–37].

BATF promotes IRF4 and vIRF3 occupancy of the endogenous MYC-SE

To assess the role of BATF at the endogenous IRF4- and MYC-SE sequences in PEL cells, we performed inducible CRISPR/Cas9-mediated BATF knockout experiments in BCBL-1 cells. Three days after Cas9 induction, cultures expressing a validated BATF sgRNA were as viable as cultures expressing the safe-harbor-directed control guide sgAAVS1 [8] (Fig. 11A). At this time point, BATF was efficiently depleted, while vIRF3 and JUNB expression levels were unaffected, and expression of IRF4 and MYC was modestly reduced (Fig. 11B and C). Interestingly, most JUNB protein migrated reproducibly at a lower molecular weight upon BATF KO, consistent with an altered post-translational modification pattern. BATF KO cultures had measurably reduced viability at later time points (Fig. 11A, day 6), confirming the essentiality of BATF in BCBL-1.

ChIP-qPCR analysis of the endogenous MYC-SE^500kb element on day 3 after Cas9 induction confirmed robust BATF occupancy in sgAAVS1 cells and efficient depletion of BATF from this site in BATF KO cells (Fig. 11D). JUNB, IRF4, and vIRF3 robustly co-occupied this site with BATF and their occupancy was strongly reduced upon BATF KO (Fig. 11E–G). In the case of IRF4, its occupancy of the MYC-SE^500kb element was reduced to a greater degree than its expression (adj. P = .0001, one-way ANOVA with Tukey’s multiple comparisons test). These results suggest that vIRF3 and IRF4 cannot maintain full occupancy at the endogenous MYC-SE^500kb element in the absence of BATF. Given the presence of a functional AICE element, the reduction in IRF4 occupancy may partially reflect a direct role of BATF in its recruitment. BATF may furthermore promote the occupancy of IRF4, vIRF3, and JUNB by maintaining chromatin accessibility, in line with its documented role in promoting genome accessibility in T cells.

The recovery of BATF at the IRF4-SE was less efficient than at the MYC-SE^500kb and BATF KO reduced BATF occupancy to background levels observed at the negative control locus RPL30 (∼0.5% of input, Fig. 11H and L). JUNB occupancy mirrored that of BATF (Fig. 11I and M). IRF4 occupancy of the IRF4-SE was reduced but mirrored the reduction in IRF4 expression and remained ~10-fold higher than at the RPL30 locus (Fig. 11J and N). vIRF3 occupancy did not change significantly upon BATF KO (Fig. 11K and O). These data are consistent with a less dramatic role of BATF in maintaining IRF4 and vIRF3 occupancy at the IRF4-SE than at the MYC-SE. Together with the reporter data above, these results directly validate a critical role for BATF in promoting IRF4 and vIRF3 occupancy at the MYC-SE, while suggesting a less prominent role in maintaining their occupancy at the IRF4-SE, potentially through effects on chromatin accessibility.

Discussion

The latent KSHV protein vIRF3 is essential for the survival of PEL cells [40]. One critical role of vIRF3 is to promote the SE-mediated overexpression of the essential proteins IRF4 and MYC [8]. We show that vIRF3 and IRF4 cooperatively activate reporter elements derived from the IRF4-SE and distal MYC-SE^500kb. Responsiveness of these elements depends on unusually long, ~80–90-bp sequences, consistent with complex regulatory sites.

Notably, IRF4 alone failed to activate the transfected reporters, whereas vIRF3 alone activated the minimal IRF4-SE element and MYC-SE^500kb but not MYC-SE^375kb or the full IRF4-SE^500bp reporter. In these assays, IRF4 required cooperation with vIRF3 for full activation of both reporter elements, with an additional contribution of BATF at the MYC-SE^500kb. At the IRF4-SE, the DNA sequences required for vIRF3-mediated activity and association were genetically separable from those mediating its cooperation with IRF4, arguing that vIRF3 association does not require DNA-bound IRF4 in these assays. vIRF3 activity depended on AP-1 sequences in both elements. In the case of the MYC-SE^500kb, the AP-1 motif forms a canonical AICE2 element with an adjacent IRF/GAAA motif; however, the impact of IRF/GAAA half-site mutation on vIRF3/IRF4-mediated activation was modest, suggesting that additional sequences may contribute to the regulation of this element by IRF4. In the case of the IRF4-SE, the AP-1 site did not participate in a canonical AICE motif but participated in non-canonical cooperation with an ISRE-like element and apparently isolated IRF/GAAA motifs. These results highlight the essential role of vIRF3 in PEL-SE engagement and suggest that vIRF3 enables IRF4 to function in configurations that diverge from known composite binding motifs or are suboptimal. A second MYC-SE, MYC-SE^375kb, was not activated by vIRF3 alone and was significantly, but weakly activated by co-transfection of vIRF3 and IRF4. This element may use an AP-1-independent mode of vIRF3 recruitment, or vIRF3 might be recruited through a less optimal AP-1 site that remains to be identified.

For the IRF4-SE, the AP-1, IRF/GAAA, and ISRE-like motifs were insufficient in fully randomized flanking contexts. However, retention of the extended sequence between the AP-1 and ISRE-like motifs in R4 preserved vIRF3 association, vIRF3-only responsiveness, and partial vIRF3/IRF4 responsiveness in 293 cells, indicating that these intervening sequences contribute to vIRF3 association and function. These results point to complex sequence requirements for vIRF3 recruitment and may help explain why our prior ChIP-seq analysis revealed only few high-confidence vIRF3 peaks [8]. All ∼60 SE-associated vIRF3 peaks in this dataset overlap IRF4 peaks, suggesting that vIRF3 specifically associates with regions that are enriched in IRF4-binding sites and likely recruit IRF4 in other settings, through different cofactors. Accordingly, these sites overlapped IRF4-occupied SEs in other IRF4-dependent settings [8]. While we cannot exclude that the low number of vIRF3 peaks is related to the performance of available vIRF3 antibodies, our data above favor a model in which vIRF3 targets a restricted set of SEs through multifactorial and context-dependent interactions, in cooperation with IRF4.

BATF is an established IRF4 partner that is best studied in T cells. Our published ChIP-seq data identified BATF occupancy at several vIRF3/IRF4-occupied SEs [8], and our ChIP-qPCR experiments confirm robust BATF and JUNB occupancy at the MYC-SE^500kb, whereas BATF enrichment at the IRF4-SE was modest, approximately twofold above the RPL30 control and much lower than at MYC-SE^500kb. BATF knockout furthermore strongly reduced JUNB, IRF4, and vIRF3 occupancy at the MYC-SE^500kb before detectable effects on cell viability, establishing an important role for BATF in maintaining TF occupancy at this locus. In contrast, BATF promoted vIRF3/IRF4-dependent activation of both the IRF4- and MYC-SEs in integrated reporter assays. Together, these results suggest that BATF preferentially promotes vIRF3/IRF4 function in integrated or endogenous chromatin compared with transfected plasmid contexts, consistent with its established role in promoting chromatin accessibility [50, 51]. At the MYC-SE^500kb, the functional AICE2 motif may additionally enable BATF to directly promote IRF4 recruitment. Although JUNB co-occupies the MYC-SE with BATF, its functional contribution and that of other JUN family members remain unclear. Future studies examining JUN family occupancy and individual or combinatorial loss of JUN proteins in PEL cells will be important to define their roles in BATF-, IRF4-, and vIRF3-dependent SE regulation.

Our mutagenesis of the IRF4-SE revealed that isolated IRF/GAAA motifs, outside of canonical composite elements, can contribute to cooperative activation by vIRF3 and IRF4. Although isolated IRF motifs are generally considered low-affinity sites for IRF4, our findings suggest that IRF4 may use such sites in the presence of vIRF3 or other regulators. While DNA pulldown assays did not show enhanced IRF4 recruitment in the presence of vIRF3, these assays may not fully capture the chromatin-level dynamics in PEL cells. Follow-up studies should therefore assess the influence of vIRF3 on IRF4 occupancy in native chromatin.

Because IRF4 or vIRF3 knockdown rapidly compromises PEL cell viability [7–9, 40], we relied on ectopic expression experiments to isolate their roles. These approaches have inherent limitations, such as imperfect stoichiometry, absence of B-cell or PEL-specific chromatin features, and cell-type-specific expression of other TFs. However, we validated key findings in PEL cells using reporter assays and DNA pulldowns and tested BATF function at endogenous SEs by ChIP-qPCR, confirming the importance of the mapped AP-1 and IRF-related motifs for activity and revealing distinct sequence requirements for vIRF3 and IRF4 association. There are several important questions that should be addressed in future work. Specifically, the molecular basis for vIRF3 recruitment to DNA remains unresolved. The N-terminal region of vIRF3 folds similarly to the IRF DBD but lacks the conserved residues involved in recognition of the IRF/GAAA motif by IRFs. While vIRF3 is therefore not thought to bind to IRF/GAAA motifs directly, non-canonical DNA binding to complex motifs may have been missed. Similarly, it is unknown how vIRF3 associates with IRF4 and whether this association is direct, although our results suggest that association of vIRF3 and IRF4 is independent of IRF4’s ability to bind to DNA.

In summary, our study defines sequence features underlying vIRF3-mediated SE activation in PEL cells and contributes to our understanding of how this KSHV protein co-opts host transcriptional machinery to sustain the oncogenic transcriptional program in PEL. Our findings reveal complex and context-dependent requirements for vIRF3/IRF4-mediated SE activation and provide a starting point for dissecting SE regulation in PEL and other IRF4-dependent contexts.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

None declared.

Funding

This work was supported by the National Cancer Institute (NCI) of the National Institutes of Health (NIH) under grant numbers R01-CA247619 (to E.G.) and R50-CA221848 (to E.T.B.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Data availability

No new data were generated in addition to those presented in the figures and tables. All vectors are available on request from E.G.