BZLF1 interacts with the chromatin remodeler INO80 promoting escape from latent infections with Epstein-Barr virus

A hallmark of Epstein-Barr virus (EBV) infections is its latent phase, when all viral lytic genes are repressed. Repression results from a high nucleosome occupancy and epigenetic silencing by cellular factors such as the Polycomb repressive complex 2 (PRC2) and DNA methyltransferases that respectively introduce repressive histone marks and DNA methylation. The viral transcription factor BZLF1 acts as molecular switch to induce the transition from the latent to the lytic or productive phase of EBV’s life cycle. It is unknown how BZLF1 can bind to the epigenetically silenced viral DNA and whether it directly reactivates the viral genome through chromatin remodeling. We addressed these fundamental questions and found that BZLF1 binds to nucleosomal DNA motifs both in vivo and in vitro, a property characteristic of bona fide pioneer factors. BZLF1 co-precipitates with cellular chromatin remodeler ATPases, and the knock-down of one of them, INO80, impaired lytic reactivation and virus synthesis. We conclude that BZLF1 reactivates the EBV genome by directly binding to silenced chromatin and recruiting cellular chromatin remodeling enzymes, which implement a permissive state for viral transcription. BZLF1 shares this mode of action with a limited number of cellular pioneer factors, which are instrumental in transcriptional activation, differentiation, and reprogramming in all eukaryotic cells.


INTRODUCTION
Eukaryotic DNA binding sites are often not accessible to their cognate factors because the sites lie within epigenetically silent chromatin and are occupied by nucleosomes. Nucleosomes at binding sites constitute a physical barrier to transcription factors because their binding is often structurally incompatible with DNA wrapped around the histone octamer. Access to nucleosomal sites may be achieved through cooperative and simultaneous binding of several transcription factors that outcompete the histone octamer (Adams and Workman, 1995;Mirny, 2010).
Alternatively, one class of transcription factors, termed pioneer factors (Magnani et al., 2011b;Zaret and Carroll, 2011;Cirillo et al., 1998;Cirillo et al., 2002), can bind their target sequences even on nucleosomal DNA and in silent chromatin and establish competence for gene expression through chromatin remodeling (Zaret and 3 Mango, 2016 for a recent review). Pioneer factors either open chromatin directly through their binding or recruit chromatin modifiers and ATP-dependent chromatin remodeling enzymes that open chromatin to allow access for the transcription machinery (Bartholomew, 2014;Clapier and Cairns, 2009;Längst and Manelyte, 2015). Such pioneer factors play key roles in hormone-dependent cancers (Jozwik and Carroll, 2012), embryonic stem cells and cell fate specification (Smale, 2010;Drouin, 2014), and cellular reprogramming (Iwafuchi-Doi and Zaret, 2014;Soufi et al., 2015). Currently, two-to three-thousand sequence-specific DNA-binding transcription factors in human cells are known (Lander et al., 2001;Venter et al., 2001), but only about a dozen are characterized as pioneer factors.
Certain pioneer factors have peculiar structural characteristics that explain binding to nucleosomal DNA. For example, the winged-helix DNA binding domain of the paradigm pioneer factor FoxA structurally resembles the linker Histone H1, disrupts inter-nucleosomal interactions, opens chromatin, and enhances albumin expression in liver cells (Cirillo et al., 2002;Sekiya et al., 2009). How many other pioneer factors bind to nucleosomal DNA is less well understood, but some directly target partial DNA motifs displayed on the nucleosomal surface (Soufi et al., 2015).
Subsequently, most pioneer factors recruit chromatin remodelers to their binding sites, which open silent chromatin and regulate cell-type specific gene expression (Magnani et al., 2011a;Mayran et al., 2015).
In eukaryotic nuclei, chromatin remodelers mediate the dynamics of nucleosome arrangements and participate in most DNA-dependent processes (Längst and Manelyte, 2015 for a recent overview). They bind to nucleosomes and convert the energy of ATP hydrolysis into the movement, restructuring, or ejection of histone octamers depending on the remodeler. Remodelers are categorized according to their ATPase subunit into four major (SWI/SNF, ISWI, INO80, and CHD) and several minor families and further differentiated by their associated subunits. This range of features reflects specialized functions found in their domains/subunits that mediate direct interactions with modified histones, histone variants, DNA structures/sequences, RNA molecules, and transcription factors. The human genome encodes 53 different remodeler ATPases (Längst and Manelyte, 2015), which are highly abundant chromatin factors with roughly one remodeling complex per ten nucleosomes (Längst and Manelyte, 2015). 4 The Epstein-Barr Virus (EBV) infects more than 95 % of the adult population worldwide with a lifelong persistence in human B cells. The key to EBV's success lies in its ingenious multipartite life, which relies on different epigenetic states of viral DNA (Woellmer and Hammerschmidt, 2013). Initially, EBV establishes a latent infection in all cells it infects (Hammerschmidt, 2015;Kalla et al., 2012). Viral latency is characterized by an epigenetically silenced EBV genome that prevents expression of all lytic viral genes, but usually spares a small set of so-called latent viral genes that remain active. Cellular factors, e.g. the Polycomb repressive complex 2 (PRC2) and DNA methyltransferases respectively introduce repressive histone marks and 5methyl cytosine residues into viral DNA, which ensure the repressed state of all viral lytic genes (Ramasubramanyan et al., 2012;Woellmer et al., 2012).
BZLF1 binds to two classes of BZLF1 responsive elements (ZREs): one class contains a DNA sequence motif reminiscent of the canonical AP-1 binding site, the other class contains a sequence motif with a CpG dinucleotide, which must carry 5methyl cytosine residues for efficient BZLF1 binding Bhende et al., 2004;Flower et al., 2011;Karlsson et al., 2008). Binding of BZLF1 to viral chromatin induces the loss of nucleosomes at certain but not all ZREs with higher than average nucleosome densities (see Figs. 2 and 3 in ref. Woellmer et al., 2012).
This study did not determine whether the initial binding of BZLF1 and loss of nucleosomes are simultaneous events or occur sequentially nor did it identify the molecular mechanisms that underlie these events.
Here we report that the viral factor BZLF1 acts like a pioneer transcription factor. BZLF1 binds mononucleosomal DNA in repressed lytic promoters in vivo and binds to nucleosome core particle DNA in vitro. BZLF1

Transient transfection of cell lines
Transfection of DNA into HEK293 cells using polyethylenimine (Polysciences, #24765) was as described (Reed et al., 2006). For protein extracts, 2x10 7 cells per 13 cm cell culture dish were seeded the day before transfection. Each plate was transfected with 30 µg plasmid DNA.

In vitro DNA methylation
CpG methylation of plasmid DNA in vitro was done with the de novo methyltransferase M.SssI (New England Biolabs) and S-adenosyl methionine according to the manufacturer's recommendations.

In vitro reconstitution of chromatin
Drosophila embryo histone octamers were prepared and used for in vitro nucleosome reconstitution via salt gradient dialysis according to Krietenstein et al. (Krietenstein et al., 2012).
Proteins were purified from HEK293 cells transiently transfected with a FLAG-and tandem Strep-tagged BZLF1 expression plasmid . Two days post transfection cells from six 13 cm cell culture dishes were pooled and lysed in 10 ml RIPA-buffer (50 mM Tris-HCl pH 8.0, 150 mM NaCl, 1 % Ipegal, 0.5 % DOC, 0.1 % SDS). Lysates were sonicated as above and BZLF1 protein was purified using analyzed with the aid of a radioisotope scanner (FLA 5100, Fuji) and quantitation of radioactivity signals was performed with the AIDA program (Raytest). For determination of the equilibrium dissociation constant (K D ), these data were fitted to the Hill equation with one-site specific binding using the PRISM 6 program (Graphpad).

Co-immunoprecipitation (Co-IP)
Co-IPs of GFP-fusion or Strep-tag fusion proteins were done with GFP-Trap_A   (Delecluse et al., 1998) were transfected with the siRNAs for three days prior to transient transfection of expression plasmids encoding BZLF1 and gp110/BALF4 to induce virus production in the siRNA-treated cells.

Cell viability assay
siRNA transfected 2089 EBV HEK293 cells were seeded at three different densities into opaque-walled 96-well plate in 100 µl/well. 100 µl of CellTiter-Glo reagent (Promega) was added and mixed for 2 minutes on an orbital shaker to induce cell lysis. The plate was incubated at room temperature for 10 minutes to stabilize the luminescent signal, which was subsequently recorded using a ClarioStar plate reader. Viability of the cells was calculated and expressed as percent of viable cells compared to cells treated with non-targeting siRNA.

Generation of Raji p4816 cell lines with lentiviral shRNA vectors directed against INO80
Potentially suitable shRNAs were identified using the publicly available web tool siDirect2.0 (http://sidirect2.rnai.jp/). Based on the identified sequences, primers were designed and cloned into the pCDH lentiviral vector (System Biosciences), which was modified and termed p6573 as shown in Figure

Absolute quantification of transcripts using RT-qPCR
Raji p4816

Statistical analysis
We used Prism 6 (GraphPad) for statistical analysis and the two-tailed ratio T-test was applied unless otherwise mentioned.

Loss of histone H3 at repressed lytic promoters follows initial lytic viral reactivation.
We used our established model (Woellmer et al., 2012)  synthesis (Miller et al., 1972). We found that the BZLF1 levels we reach in the Raji inducible system are in a range also found in the small fraction of B95-8 cells that undergo the lytic cycle of EBV (unpublished data).
Next, we performed chromatin immunoprecipitation (ChIP) with cross-linked viral chromatin, which had been fragmented to an average size of 150 base pairs (bp) and an antibody directed against histone H3 indicative of the histone octamer. We detected a partial loss of H3 at promoter sites of certain early lytic genes as reported previously (Woellmer et al., 2012), but only after 15 hours post induction (Fig. 1B). In contrast, H3 levels were unaffected at latent and late lytic promoters (

BZLF1 binds mononucleosomal DNA in viral lytic promoters in vivo.
It is unclear if BZLF1 can bind nucleosomal DNA directly or relies on a mechanism that exposes ZRE motifs, presumably by nucleosome eviction prior to

BZLF1 binds mononucleosomal DNA in vitro.
We verified our in vivo finding in a defined and unambiguous in vitro system using electromobility shift assays (EMSAs) with a purified BZLF1 protein and ZREcontaining DNA fragments in their free states or bound as mononucleosomes (Fig.   14 3). The latter bound fragments serve as surrogates for viral chromatin in its repressed state.
The promoter of the early lytic gene BBLF4, which encodes the viral DNA helicase, harbors five ZREs of nine base pairs in length. All five ZREs contain CpG dinucleotides and show a methylation-dependent binding of BZLF1 . We prepared three 156 bp long DNA fragments derived from this promoter region that differed in the positioning of two ZREs, ZRE 3 and ZRE 4 (Fig. 3A). A 156 bp fragment of the BRLF1 coding sequence, which lacks ZREs and is not bound by BZLF1, served as negative control. All four 156 bp fragments, which had been fully CpG methylated using a commercially available de novo CpG methyl transferase which is consistent with the widely observed, low-level but non-specific DNA binding of transcription factors (Fried and Crothers, 1981). BZLF1's binding to free DNA resulted in several shifted bands, which is a common observation .
In contrast, BZLF1's binding to mononucleosomal DNA differed for the three ZREcontaining fragments (Fig. 3B, lower row of panels). BZLF1 did not bind to mononucleosomes without (control fragment) or with only one ZRE (ZRE 3 or ZRE 4). The well-studied yeast transactivator Pho4, which served as a negative control because it does not bind to nucleosomal sites (Venter et al., 1994), did not yield shifted bands with its binding site buried in mononucleosomes but did with free DNA (Fig. S4). However, BZLF1 did bind to the ZRE 3+4 fragment, again with a K D of 15 about 13 nM (Fig. 3C). A truncated BZLF1 protein (aa149-245) that lacks the activation but retains the DNA-binding domain, bound both free and mononucleosomal ZRE 3+4 fragments (Fig. S5) indicating that the DNA-binding domain is sufficient to mediate the pioneer factor-like binding of BZLF1.
The binding of BZLF1 to the mononucleosomal ZRE 3+4 fragment (but not to fragments with single ZRE 3 or ZRE 4 motifs) suggested that BZLF1 requires at least two binding sites for stable binding to a nucleosome. As an alternative interpretation, two ZREs might be required to outcompete the histone octamer for binding such that the shifted complex migrated like a complex of BZLF1 with free DNA. We ruled out this latter possibility by comparing BZLF1 complexes with free and mononucleosomal ZRE 3+4 fragments run in parallel in the same gels (Fig. 3B, bottom row, rightmost panel, and Fig. 3D). The migration position of BZLF1 in complex with free DNA differed from that in complex with mononucleosomal DNA.
An anti-FLAG antibody appropriate for binding FLAG-tagged BZLF1 supershifted the signals and unambiguously identified BZLF1 in both the free and the mononucleosomal DNA shift complexes (Fig. 3D).
Yet another interpretation could be that BZLF1 did not necessarily require two ZREs for binding on a nucleosome, but that a ZRE had to be close to the entry and or exit points of the nucleosomal DNA rather than close to the dyad. It is known that nucleosomal DNA can undergo thermal "breathing" motions that transiently expose DNA sites close to the exit/entry points but much less frequently sites close to the dyad (Anderson and Widom, 2000). To investigate this possibility, we modified the ZRE 3 or ZRE 4 sequences in the ZRE 3+4 fragment by PCR mutagenesis such that we obtained two fragments with only one ZRE located at different positions relative to the entry/exit points termed ZRE 0+4 and ZRE 3+0 (Fig. 4A). EMSAs with BZLF1 and these two constructs in mononucleosomal forms demonstrated that BZLF1's binding to ZRE 0+4 was barely detectable (Fig. 4B, middle panel), relatively strong to ZRE 3+0 (Fig. 4B, right panel), and strongest to ZRE 3+4 (Fig. 4B, left panel). From Hill plots (Fig. 4C)  the different free DNA fragments were in the range of 10 to 20 nM in three independent experiments (Fig. S6).
We made use of the clearly detectable binding of BZLF1 to the single ZRE in the ZRE 3+0 fragment to ask if thermal "breathing" or nucleosomal phasing played a major role to this binding. To do so we altered the position of the single ZRE in the ZRE 3+0 fragment relative to the original position of this ZRE by -5 nucleotides (nt), +10 nt, +15 nt, and +30 nt as shown in Fig. 4A. With these four constructs, we repeated the EMSA analysis and observed robust BZLF1 binding to three of four mononucleosomes tested (-5 nt, +15 nt, and +30 nt), in all cases stronger than to the ZRE 0+4 mononucleosome (Fig. 4D vs. B). BZLF1 binding to ZRE 3+0 +10 nt (Fig.   4A) was not detectable (Fig. 4D, middle panel). This finding discredits a prominent role of DNA "breathing" for BZLF1's binding, especially given that the ZRE is more internal in the ZRE 3+0 +30 nt than in the ZRE 0+4 fragment. The fact that BZLF1 did not bind to ZRE 3+0 +10 nt (Fig. 4D, middle panel) is reminiscent of ZRE 0+4 ( Fig. 4B, middle panel), because both binding sites are positioned similarly, i.e. 18 and 17 nt from the distal ends (Fig. 4A).
The data do not seem to support nucleosomal phasing as a critical determinant either. Shifting of the single ZRE in the fragment ZRE 3+0 by multiples of 15 bp (+15 nt versus +30 nt), which corresponds roughly to one and a half helical turns on the histone octamer surface (44), did not affect BZLF1 binding (Fig. 4D) suggesting that it did not matter which part of the ZRE faces inward or outward from the nucleosome.
Taken together, we conclude that the properties and the position of a given ZRE (ZRE 3 versus ZRE 4) and, to a much larger degree, the cooperation between two ZREs (ZRE 3+4 construct) support BZLF1's binding to a nucleosomal site, but the data do not provide an obvious rule of BZLF1 binding to nucleosomal DNA.

BZLF1 and cellular chromatin remodeling enzymes interact.
In our in vitro experiments with reconstituted nucleosomes we did not observe a histone loss or a disassembly of the nucleosome upon BZLF1's binding because shifted bands characteristic of a BZLF1-free DNA complex (Fig. 3D) or an increase of free DNA (Fig. 3B, (Fig. 5). These results suggested that BZLF1 interacts with subunits of at least two cellular chromatin remodeler families possibly to recruit them to lytic gene promoters and support EBV's lytic reactivation.

Different BZLF1 domains mediate the interaction with SNF2h versus INO80.
The differential binding of the bZIP constructs to SNF2h versus INO80 indicated that different domains of BZLF1 might mediate these interactions. We extended our co-immunoprecipitations to include more BZLF1 derivatives (Fig. 6A). We transiently transfected HEK293 cells with the BZLF1 constructs, which were co-expressed with the GFP-tagged chromatin remodeler ATPase subunits, SNF2h or INO80. The BZLF1 expression plasmids were adjusted to obtain similar protein levels (Fig. 6B).
Cell lysates were again treated with benzonase and DNase I, immunoprecipitations were done with GFP-binder coupled sepharose beads and the co-precipitated BZLF1 was detected by Western blotting using the antibody directed against a motif within BZLF1's bZIP domain. The immunoblots ( Fig. 6C;  We followed the experimental work by the Tsukiyama lab (Gelbart et al., 2005), who used a catalytically inactive remodeler mutant to map its targets genome-wide. We to these viral promoters. This effect can also be seen to some degree at the control locus (HPRT1) and a latent viral gene (EBER) (Fig. S8), which might indicate a global alteration of chromatin structure upon lytic phase induction, because BZLF1 binds to more than 10 5 binding sites in cellular chromatin (unpublished data).

shRNA mediated knock-down of INO80 reduces transcriptional reactivation of certain early lytic genes of EBV.
Next, we asked whether INO80 levels might be critical to activate viral lytic genes upon expression of BZLF1. We engineered lentiviruses to stably express shRNAs directed against INO80 transcripts in Raji p4816 cells (Fig. S9A, B) and tested the timely expression of selected viral lytic genes upon addition of doxycycline by RT-qPCR.
As can be seen in Fig. 7A, two shRNAs efficiently reduced the steady-state levels of INO80 protein in Raji cells. We next analyzed the transcriptional activation of four early viral genes (BMRF1, BNLF2a, BRLF1, BBLF4) in three different Raji p4816 cell lines stably transduced with shRNA_INO80_1, shRNA_INO80 _2, or a non-targeting shRNA_nt control (Fig. S9A, B). The knock-down of INO80 resulted in a reduced activation of BMRF1 and BNLF2a eight hours post induction (Fig. 7B). In contrast to these two 'early responding' genes, BRLF1 and BBLF4, which have considerably slower kinetics of induction, showed a very modest reduction of their transcript levels 15 hours post induction, only ( Fig. 7B and Fig. S10). We concluded from this experiment that INO80 may play a critical functional role in the early phase of viral reactivation at certain viral promoters of early lytic genes.

siRNA-mediated knock-down of INO80 inhibits de novo synthesis of virus.
As BZLF1 is the crucial trigger for viral reactivation and interacts with at least two cellular chromatin remodelers, we hypothesized that at least one of them should be necessary for lytic induction. To test this hypothesis, we used an siRNA knock-down strategy to assess the roles of SNF2h or INO80 in 2089 EBV HEK293 cells (Delecluse et al., 1998). Upon transient transfection of a BZLF1 expression plasmid (Hammerschmidt and Sugden, 1988), this 2089 EBV HEK293 producer cell line releases infectious virus, which can be quantified by assaying infected, GFP-positive Raji cells by flow cytometry (Steinbrück et al., 2015).
The 2089 EBV HEK293 cells were treated for three days with siRNA pools targeting the SMARCA5 gene encoding SNF2h or the INO80 gene or with a non-20 targeting control siRNA. The respective knock-down efficiencies were assessed by Western blotting (Fig. 8A). In these siRNA-treated cells virus synthesis was initiated by transient co-transfection of expression plasmids encoding BZLF1 together with gp110/BALF4 as described (Neuhierl et al., 2002). Expression of gp110/BALF4 increases virus infectivity by about a factor of ten (Neuhierl et al., 2002). Three days after plasmid co-transfection, cell supernatants were collected and defined volumes were used to infect Raji cells. The fractions of GFP-expressing Raji cells were determined by flow cytometry after three additional days such that the virus concentrations could be calculated (Fig. 8B).
Steady-state protein levels of SNF2h or INO80 were modestly reduced after three days of siRNA treatment (Fig. 8A). Cells treated with non-targeting siRNAs or with an SNF2h-specific siRNA pool (Fig. S9C) did not differ significantly in the levels of released, infectious EBV (Fig. 8B). However, cells treated with an INO80-specific siRNA pool released significantly fewer viral particles (Fig. 8B). This observation is notable given the only modest diminution of INO80 by the siRNA treatment (Fig. 8A).
None of the siRNA pools directed against SNF2h or INO80 had an adverse effect on cell viability (Fig. S9D) suggesting that the reduced virus synthesis after siRNA knockdown is a specific and INO80-related effect.
Together, these results support a critical role for INO80 and for BZLF1's acting as a pioneer factor in EBV lytic activation.

Discussion
EBV takes advantage of the host cell's epigenetic machinery to establish a stable latent infection. Upon infection, the viral DNA is epigenetically naïve, i.e. free of histones and devoid of methylated CpG dinucleotides Kintner and Sugden, 1981;Fernandez et al., 2009). In the course of establishing the latent phase, the host cell's epigenetic machinery compacts the viral DNA into nucleosomal arrays, introduces repressive histone modifications, and initiates the methylation of the majority of viral CpG dinucleotides . As a consequence, viral promoters, with the exception of those of a few latent genes, are silenced during EBV's latent phase. Densely positioned nucleosomes, repressive histone marks 21 introduced by Polycomb proteins, and extensive DNA methylation keep the virus in a strictly latent, dormant mode (Ramasubramanyan et al., 2012;Woellmer et al., 2012).
EBV can escape from latency and enter the lytic, productive phase when its host B cells terminally differentiate to plasma cells (Laichalk and Thorley-Lawson, 2005).
In the lytic phase, the loss of nucleosomes increases the accessibility of viral DNA to binding transcription factors. The removal of repressive and the gain of active histone marks reactivate the promoter regions of early viral lytic genes, enabling the virus to replicate its DNA, express late viral genes, and produce viral progeny to infect new B cells (Hammerschmidt, 2015 for a recent review).
The viral transcription factor BZLF1, which is induced upon terminal plasma cell differentiation, is the switch triggering a transition from the latent to the lytic phase.
First, BZLF1 induces the expression of several early lytic EBV genes by binding sequence-specifically to ZREs in their promoter regions. Many ZREs need to contain 5-methyl cytosine residues to permit BZLF1's binding and thus CpG methylation of viral DNA is a prerequisite to express certain, essential, early lytic genes Kalla et al., 2012). Second, BZLF1 enables viral DNA replication. It binds to the lytic origin of DNA replication and promotes the recruitment of components of the viral DNA replication machinery to this viral origin to support its function (Schepers et al., 1993). Third, BZLF1 directly or indirectly causes the loss of nucleosomes in the promoter regions of viral lytic genes, which correlates with their expression (Hammerschmidt, 2015).
We and others have hypothesized that the efficient reactivation of silenced, inactive viral chromatin is forced by a presumed pioneer function of the BZLF1 transcription factor (Woellmer et al., 2012), chromatin alterations (Adamson and Kenney, 1999;Zerby et al., 1999), and/or the additional recruitment of chromatin remodelers (Woellmer et al., 2012). Our results, both in vivo (Fig. 2) and in vitro ( Figs. 3 and 4), show that BZLF1 can bind mononucleosomal DNA in promoter regions of early lytic genes known to be regulated by BZLF1. We have also shown that BZLF1 can bind ZREs in nucleosomes close to the nucleosome dyad (Fig. 4D).
The nucleosome-binding activity is encoded within the C-terminal part of BZLF1 encompassing the bZIP domain and C-terminus and does not depend on BZLF1's 22 transactivation domain (Fig. S5). BZLF1's binding does not eject nucleosomes in vitro (Fig. 3B, D) in the absence of other molecular machines. This observation is in line with our initial data (Fig. 1B) indicating that the binding of BZLF1 precedes a decrease in nucleosomal occupancy at early lytic promoters by hours.
Transcriptional activation of lytic viral genes (BMRF1 and BNLF2a in Fig. 7B for example) appears to precede nucleosome ejection as shown in Fig. 1B (15  chromatin remodeling enzymes to sites in viral promoters, dramatically reducing their histone occupancy (Neely et al., 1999;Herrera and Triezenberg, 2004). The ATPase subunit SNF2h has been reported previously to promote HSV-1 immediate-early gene expression as well as replication and might also interact with VP16 (Bryant et al., 2011).
Similarly, the INO80 remodelers are able to slide nucleosomes, exchange histones, regulate transcription, and are involved in DNA repair and cell cycle checkpoint control (Shen et al., 2000;Tsukuda et al., 2005;van Attikum et al., 2007).
While INO80 has not been implicated in the reactivation of herpes viruses, the EBV nuclear proteins EBNA-LP and EBNA2, which are necessary for lymphoblastoid cell 23 line growth and survival have been found associated with an INO80 remodeler complex (Portal et al., 2013). Our results indicate that BZLF1's transactivation domain interacts with the INO80 ATPase subunit (Fig. 6C)  and H3K4me2 but not by histone acetylation (Cirillo and Zaret, 1999;Lupien et al., 2008). FoxA can even favor H3K4me2 deposition (Smale, 2010). The pioneer factor PBX1 is also capable of reading specific epigenetic signatures such as H3K4me2 (Magnani et al., 2011a;Berkes et al., 2004), and PU.1 reprograms the chromatin landscape through the induced deposition of H3K4me1 (Heinz et al., 2010). The viral transcription factor BZLF1 might read and decipher epigenetic modifications as well.
In fact, we confirmed in ReChIP experiments that BZLF1 co-occupies H3K4me1 in the viral promoters of early lytic target genes (Fig. 2B). 24 The viral transcription factor BZLF1 belongs to the basic leucine-zipper (bZIP) family of transcription factors and contains a variant of the leucine zipper motif responsible for the coiled-coil structure (Farrell et al., 1989;Chang et al., 1990;Lieberman and Berk, 1990). BZLF1 forms homodimers and binds DNA motifs via its two long bZIP helices (Petosa et al., 2006). The basic region of each helix contacts the major groove and the zipper region forms a coiled-coil stabilized by BZLF1's Cterminal tail. The closest relative of BZLF1 is the cellular c-Fos/c-Jun heterodimer AP1 transcription factor (Farrell et al., 1989). Its binding to nucleosomal DNA is reduced compared to free DNA sequence motifs, but nucleosomal DNA binding severely affects the structure of the underlying nucleosome, which can facilitate the subsequent binding of additional transcription factors (Ng et al., 1997). It thus appears that BZLF1 has optimized this fundamental function of AP-1 transcription factors to support EBV's escape from repressed chromatin.
EBNA1 was also proposed to have similarities with the paradigmatic pioneer factor FoxA1 in a recent review (Niller and Minarovits, 2012), but to our knowledge, our biochemical and functional data identify BZLF1 as a bona fide pioneer factor of Epstein-Barr virus. Two domains in EBNA1 mimic the AT-hooks of certain cellular high mobility group proteins (Altmann et al., 2006;Hung et al., 2001) and promote the mobility of the linker histone H1 indicative of an EBNA1 intrinsic remodeling function, which is independent of cellular chromatin remodelers (Coppotelli et al., 2013).
The pioneer factor BZLF1 could share certain functions with its cousin, VP16 of Herpes simplex virus. Herpes simplex virus DNA is not (Leinbach and Summers, 1980;Muggeridge and Fraser, 1986) or only selectively associated with nucleosomes during lytic infection (Oh et al., 2015 and references therein), and partially conflicting data suggest that either histone chaperones (Oh et al., 2012), chromatin modifying co-activators (Herrera and Triezenberg, 2004), or chromatin remodelers (Neely et al., 1999) are responsible for the lack of histones on Herpes simplex DNA. Interestingly, the transactivation domain of VP16 was found to associate directly with members of the SWI/SNF remodeling complex (Neely et al., 1999) activating in vitro transcription from a nucleosomal template (ibid). This latter finding is reminiscent of BZLF1's transactivation domain interacting with the 25 chromatin remodeler INO80 (Fig. 6C), which seems to be critical for efficient lytic infection of EBV (Fig. 7).
It has also emerged from recent work (Soufi et al., 2015) that key factors in cellular reprogramming to yield induced pluripotent stem cells share critical functions with pioneer factors. Our current findings suggest that EBV has acquired this principle and puts it to use with its BZLF1 factor to reprogram viral latent chromatin within hours and to promote escape from latency.
Taken together, our experiments suggest that BZLF1 is a bona fide pioneer transcription factor (Zaret and Mango, 2016), which recruits cellular machines for opening up repressed viral chromatin. It remains to be shown how BZLF1 can interact with nucleosomal DNA at high structural resolution, because BZLF1's in vitro binding does not seem to follow known rules of nucleosome phasing (Fig. 4D).
Future experiments should solve this conundrum.