Promyelocytic leukemia (PML) nuclear bodies (NBs) induce latent/quiescent HSV-1 genomes chromatinization through a PML NB/Histone H3.3/H3.3 Chaperone Axis

Herpes simplex virus 1 (HSV-1) latency establishment is tightly controlled by promyelocytic leukemia (PML) nuclear bodies (NBs) (or ND10), although their exact contribution is still elusive. A hallmark of HSV-1 latency is the interaction between latent viral genomes and PML NBs, leading to the formation of viral DNA-containing PML NBs (vDCP NBs), and the complete silencing of HSV-1. Using a replication-defective HSV-1-infected human primary fibroblast model reproducing the formation of vDCP NBs, combined with an immuno-FISH approach developed to detect latent/quiescent HSV-1, we show that vDCP NBs contain both histone H3.3 and its chaperone complexes, i.e., DAXX/ATRX and HIRA complex (HIRA, UBN1, CABIN1, and ASF1a). HIRA also co-localizes with vDCP NBs present in trigeminal ganglia (TG) neurons from HSV-1-infected wild type mice. ChIP and Re-ChIP show that vDCP NBs-associated latent/quiescent viral genomes are chromatinized almost exclusively with H3.3 modified on its lysine (K) 9 by trimethylation, consistent with an interaction of the H3.3 chaperones with multiple viral loci and with the transcriptional silencing of HSV-1. Only simultaneous inactivation of both H3.3 chaperone complexes has a significant impact on the deposition of H3.3 on viral genomes, suggesting a compensation mechanism. In contrast, the sole depletion of PML significantly impacts the chromatinization of the latent/quiescent viral genomes with H3.3 without any overall replacement with H3.1. vDCP NBs-associated HSV-1 genomes are not definitively silenced since the destabilization of vDCP NBs by ICP0, which is essential for HSV-1 reactivation in vivo, allows the recovery of a transcriptional lytic program and the replication of viral genomes. Consequently, the present study demonstrates a specific chromatin regulation of vDCP NBs-associated latent/quiescent HSV-1 through an H3.3-dependent HSV-1 chromatinization involving the two H3.3 chaperones DAXX/ATRX and HIRA complexes. Additionally, the study reveals that PML NBs are major actors in latent/quiescent HSV-1 H3.3 chromatinization through a PML NB/histone H3.3/H3.3 chaperone axis.


Introduction
Herpes simplex virus 1 (HSV-1) is a human pathogen with neurotropic tropism and the causal agent of cold sores and more severe CNS pathologies such as encephalitis [1]. After the initial infection, HSV-1 remains latent in neuronal ganglia with the main site of latency being the trigeminal (or Gasserian) ganglion (TG). Two transcriptional programs are associated with HSV-1 infection, the lytic cycle and latency, which differ by the number and degree of viral gene transcription. The lytic cycle results from the sequential transcription of all viral genes (approximately 80) and leads to the production of viral progeny. The latency phase, occurring exclusively in neurons, is limited to the abundant expression of the so-called Latency Associated Transcripts (LATs), although physiologically a transitory expression of a limited number of lytic genes is not excluded, making latency a dynamic process [2][3][4].
Following lytic infection of epithelial cells at the periphery, the viral particle enters the axon termini of the innervating neurons by fusion of its envelope with the plasma membrane. The nucleocapsid is then carried into the neuron body by retrograde transport, most likely through the interaction of viral capsid components [5] with microtubule-associated proteins such as dynein and dynactin [6][7][8][9][10]. Once the nucleocapsid reaches the cell body, the virus phenotype changes from the one at the axon termini because most of the outer tegument proteins, including VP16, a viral transactivator that is essential for the onset of lytic infection, remain at the axonal tip [11][12][13]. Hence, when the viral DNA is injected into the neuron nucleus, it does not automatically benefit from the presence of VP16 to initiate transcription of lytic genes. Rather, the balance between lytic and latent transcriptional programs most likely depends on stochastic events and on undescribed neuron-associated factor(s) able to initiate the transcription of VP16 through the activation of neuro-specific sequences present in the VP16 promoter [14]. Without VP16 synthesis, transcription of the viral genes encoding ICP4 (the major transactivator protein) and ICP0 (a positive regulator of viral and cellular gene transcription) is hampered. Hence, ICP4 and ICP0 gene transcription is unlikely to reach the required level to produce these two proteins above a threshold that would favor onset of the lytic cycle. Therefore, in neurons, commitment of the infectious process towards the lytic cycle or latency depends on a race between opposing infection-prone viral components and cellular features with antiviral activities.
Promyelocytic leukemia (PML) nuclear bodies (NBs) (also called ND10) are proteinaceous entities involved in the control of viral infection as part of the cell and nucleus-associated intrinsic antiviral response but also through innate immunity associated with the interferon (IFN) response [15]. Our recent studies have shown that PML NBs tightly associate with incoming HSV-1 genomes in the nucleus of infected TG neurons in mouse models and in primary TG neuron cultures [16,17]. Hence, PML NBs reorganize in structures called viral DNAcontaining PML NBs (vDCP NBs), which are formed at early times during the process of HSV-1 latency establishment and persist during latency per se in a large subset of latently infected neurons in a mouse model of infection [16]. The entrapment of incoming wild type HSV-1 genomes by PML NBs is not a unique feature of latency, because it has recently been shown to occur prior to the onset of lytic infection, as part of the intrinsic antiviral response. HSV-1 genomes trapped in the vDCP NBs are transcriptionally repressed for LATs production [16]. It is known that HSV-1 latency, at least in the mouse model and possibly in humans, is heterogeneous at the single neuron level for the expression of LATs [16,[18][19][20][21][22][23][24][25]. Therefore, although at the entire TG level HSV-1 latency could be a dynamic process from a transcriptional perspective, at the single neuron level, a strict, transcriptionally silent, quiescence can be observed, and vDCP NB-containing neurons are major contributors of this latent/quiescent HSV-1 state. In humans, vDCP NB-like structures have also been observed in latently infected TG neurons [17], suggesting that vDCP NBs are probably molecular hallmarks of the HSV-1 latency process, including in the natural host.
Another essential feature of HSV-1 latency is the chromatinization of its 150-kb genome, which enters the nucleus of the infected cells as a naked/non-nucleosomal dsDNA [26][27][28].
Once the viral genome is injected into the nucleus of the infected neuron, it circularizes, associates with nucleosomes to become chromatinized, and remains as an episome that is unintegrated into the host cell genome [29]. Although latent viral genomes sustain chromatin regulation, essentially through post-translational modifications of associated histones [30][31][32][33][34] not much is known about the mechanisms that induce their chromatinization and which specific histone variants are associated with these latent genomes. In mammals, specific H3 histone variants that differ by a few amino acid residues can influence chromatin compaction and transcriptional activity of the genome. The histone variant H3.3, a specific variant of the histone H3 that is expressed throughout the cell cycle, is deposited in a replication-independent manner, in contrast to H3.1 ( [35] and for review [36]). Interestingly, death domain associated protein 6 (DAXX) and α-thalassemia mental retardation X-linked protein (ATRX), initially identified as a transcriptional repressor and a chromatin remodeler, respectively, are constitutively present in PML NBs, and have now been identified as H3.3-specific histone chaperones [37][38][39]. The other histone H3.3 specific chaperone complex is called the HIRA complex, which is composed of Histone cell cycle regulator (HIRA), Ubinuclein 1 (UBN1), Calcineurinbinding protein 1 (CABIN1), and Anti-silencing function protein 1 homolog A (ASF1a) [35]. The HIRA complex does not normally accumulate in PML NBs except upon entry of the cell into senescence [40,41]. The histone variant H3.3 itself localizes in PML NBs in proliferating and senescent cells, linking PML NBs with the chromatin assembly pathway independently of replication [42][43][44]. Because vDCP NBs contain DAXX and ATRX [16,17,45], their involvement in the chromatinization of incoming HSV-1 genomes and/or long-term maintenance of chromatinized HSV-1 genomes is thus plausible.
Human primary fibroblasts or adult mouse primary TG neuron cultures infected through their cell body with a replication-defective HSV-1 virus, in1374, which is unable to synthesize functional ICP4 and ICP0 under specific temperature conditions, enable the establishment of a latent/quiescent state for HSV-1 [17,[45][46][47]. The latent/quiescent state of HSV-1 in human primary fibroblasts has also been reproduced using engineered HSV-1 unable to express major immediate early genes [48,49]. We have shown that this latent/quiescent state is linked to the formation of vDCP NBs, mimicking, at least concerning this particular structural aspect, the latency observed in a subset of neurons in mouse models and in humans [16,17]. Here, using the in1374-based in cellula model of infection, we showed that vDCP NBs contained not only the DAXX and ATRX proteins but also all the components of the HIRA complex and H3.3 itself. HIRA was also found to co-localize with vDCP NBs in neurons of TG harvested from HSV-1 wild type infected mice. Both DAXX/ATRX and HIRA complex components were found to interact with multiple viral loci by chromatin immunoprecipitation (ChIP). Using the same approaches, we showed that latent/quiescent viral genomes were almost exclusively chromatinized with H3.3, itself modified on its lysine (K) 9 by trimethylation (H3.3K9me3). Most interestingly, we found that H3.3 chromatinization of the viral genomes was dependent on intact PML NBs, demonstrating that PML NBs contribute to an essential part of the chromatinization of the latent/quiescent HSV-1 genomes. Overall, this study shows that the chromatinization of latent HSV-1 involves a PML NB/histone H3.3/histone H3.3 chaperone axis that confers and probably maintains chromatin marks on viral genomes.

The HIRA complex components accumulate in the vDCP NBs
The formation of vDCP NBs is a molecular hallmark of HSV-1 latency, and vDCP NBs are present in infected neurons from the initial stages of latency establishment to latency per se in mouse models [16,17]. Using a previously established in vitro latency system [46] consisting of human primary fibroblast cultures infected with a replication-deficient virus (hereafter called in1374) unable to express functional VP16, ICP4 and ICP0, we and others were able to reproduce the formation of vDCP NBs [17,45]. We first verified that vDCP NBs induced in BJ and other human primary cells infected with in1374 at a non-permissive temperature of 38.5˚C, contained, in addition to PML, the proteins constitutively found in the PML NBs, i.e., Sp100, DAXX, ATRX, SUMO-1 and SUMO-2/3 (S1Ai to S1vi Fig, and S1 Table). The DAXX/ATRX complex is one of the two chaperones of the histone variant H3.3 involved in the replicationindependent chromatinization of specific, mostly heterochromatic, genome loci [39]. Interestingly, HSV-1 enters the nucleus of the infected cell as a naked/non-nucleosomal dsDNA and remains during latency as a circular chromatinized episome unintegrated in the host genome [29,50]. It is thus tempting to speculate that the presence of DAXX/ATRX in the vDCP NBs could be linked to a process of initiation and/or maintenance of chromatinization of the latent/quiescent viral genome. The other H3.3 chaperone is known as the HIRA complex and was initially described as specific for the replication-independent chromatinization of euchromatin regions [35,51]. Remarkably, proteins of the HIRA complex are able to bind in a sequence-independent manner to a naked/non-nucleosomal DNA [52], suggesting that the HIRA complex could also participate in the recognition and chromatinization of the incoming naked HSV-1 genome. We thus investigated the localization of all members of the HIRA complex and found that they co-localized with the latent/quiescent HSV-1 genomes at 2 days post-infection (dpi) in BJ and other human primary cells (Fig 1Ai to 1iv, S1 Table). To confirm that the co-localization of members of the HIRA complex with the latent/quiescent HSV-1 could be reproduced in neuronal cells, adult mouse TG neuron cultures were infected with in1374 for 2 days before performing immuno-FISH. Mouse Hira, which was the only protein of the HIRA complex detectable in mouse cells, showed a clear co-localization with a subset of viral genomes (Fig 1B). To analyze whether this co-localization was also reproducible in vivo, immuno-FISH was performed on TG samples from HSV-1-infected mice. Hira was found to co-localize with HSV-1 genomes with the "multiple acute"/vDCP NB pattern (see [17,53,54]) in TG neurons from infected mice at 6 dpi ( Fig 1C) but not with the "single"/vDCP NB pattern (see [16,53,54]) at 28 dpi (Fig 1D), suggesting a dynamic association of this protein with the vDCP NBs.
To analyze this dynamic association, co-localization between incoming HSV-1 genomes and proteins of the PML NBs or of the HIRA complex was quantified at early times from 30 min pi to 6 hpi using a synchronized infection procedure (Fig 1E and S2 Table). Except for the proteins of the HIRA complex, the percentages of co-localization increased with time. Interestingly, at 30 min pi, the percentage of co-localization of HSV-1 genomes with HIRA was significantly higher than with PML (41±7% vs 23±5%, p value = 0.03, Student's t-test, S2 Table). Although DAXX and ATRX also showed, on average, a greater percentage of co-localization with HSV-1 genomes (36±7% and 34±5% at 30 min, respectively) compared with PML, the data were not significant (S2 Table). Moreover, a recent study showed the interaction of at least PML, SUMO-2, and Sp100 with incoming HSV-1 genomes as soon as 1 hpi, which supports our data [55]. The absence of co-localization of mouse Hira with viral genomes with the "single"/vDCP NB pattern in mouse TG neurons at 28 dpi suggested that longer infection times could lead to loss of proteins of the HIRA complex from the vDCP NBs. Infection of BJ cells were reiterated as above, but this time quantifications were performed from 24 hpi to 7 dpi. Strikingly, whereas all the proteins permanently present in the PML NBs remained colocalized with a maximum of 100% of the latent/quiescent HSV-1 genome from 2 dpi until 7 dpi, proteins of the HIRA complex peaked at 2 dpi, and then their co-localization decreased at longer times pi, confirming the temporary association of the HIRA complex with the vDCP NBs ( Fig 1F, and S3 Table).
To definitively show that proteins of the HIRA complex were present in vDCP NBs, immuno-FISH were performed on BJ cells infected for 2 days with in1374 to detect a member of the HIRA complex, HSV-1 genomes, and PML. Strikingly, while proteins of the HIRA complex showed predominant nucleoplasmic staining in non-infected cells (Fig 2i, 2iii, 2v and 2vii), in infected cells all the proteins clearly and systematically accumulated in PML NBs (Fig  2ii, 2iv, 2vi and 2viii). The accumulation of HIRA in PML NBs following infection by HSV-1 has recently been suggested to be part of an interferon-induced antiviral mechanism [56]. Consequently, HIRA, UBN1, CABIN1 and ASF1a co-localized with the latent/quiescent HSV-1 genomes in vDCP NBs (arrows in Fig 2ii, 2iv, 2vi and 2viii). Altogether, these data show that both DAXX/ATRX and HIRA complexes are present within vDCP NBs in neuronal and nonneuronal cells, suggesting a role for these two complexes in latent/quiescent HSV-1 chromatinization.

Histone H3.3 chaperones interact with incoming viral genome
The co-localization of proteins of the DAXX/ATRX and HIRA complexes with the incoming HSV-1 genomes and their presence in the vDCP NBs suggested an interaction of these proteins with the viral genome, as shown recently for HIRA on a small subset of viral loci [56]. Since DAXX, HIRA, and UBN1 antibodies were not efficient in the ChIP experiments, we constructed cell lines stably expressing myc-DAXX, HIRA-HA, or HA-UBN1 by transduction of BJ cells with lentiviral-vectors (S2 Fig). Cells were infected with in1374 at 38.5˚C and harvested 24 hpi to perform ChIP-qPCR on multiple loci spread over the entire HSV-1 genome, representing promoter or core regions (CDS) of genes of all kinetics (IE/α, E/β, L/γ) (Fig 3A). Cellular glyceraldehyde 3-phosphate dehydrogenase (GAPDH) locus was used as a positive control for enrichment. Significant enrichments compared to controls were detected for all proteins on several viral loci independently of their promoter or CDS status, and with no obvious discrepancy regarding the gene kinetic, confirming the potential interaction of these proteins all along the latent/quiescent HSV-1 genomes. Our immuno-FISH data anticipated a gradual interaction of the four proteins with the incoming viral genomes at early times post infection (see Fig 1E). To verify if this could be measured, ChIP-qPCR were performed at 30 min pi, 2 hpi and 6 hpi, using the same experimental conditions as for the immuno-FISH at early times pi (with synchronization of the infection, see Materials and Methods). The data showed a tendency for a weak interaction with the viral genomes at 30 min pi then an increase at 2 hpi and 6 hpi, although with a lot of variability, probably highlighting the dynamic of the biological events occurring during the initial stages of the infection process (S3B Fig). ATRX showed the more regular increase in its interaction with viral genomes from 30 min to 24 hpi. Overall, the ChIP data correlate with the immuno-FISH, and suggest a dynamic process for the interaction between HSV-1 genomes and proteins of the DAXX/ATRX and HIRA complexes, initiating early after the viral DNA enters the nucleus, and remaining at later times when vDCP NBs are structured.

H3.3 is present in the vDCP NBs and interacts with latent/quiescent HSV-1 genomes
The co-localization of the two histone H3.3 chaperone complexes with viral genomes suggested the chromatinization of HSV-1 latent/quiescent genomes with the histone variant H3.3. Histones H3.1 and H3.3 differ by only 5 amino acids, and, in our hands, no suitable antibody is available that can distinguish both histones by IF or IF-FISH. We thus constructed lentivirus-transduced BJ cell lines expressing a tagged version of either histone (e-H3.1 and e-H3.3) (see Materials and Methods, and [43], S4A and S4B Fig). We confirmed that ectopic expression of e-H3.3 led to its accumulation in PML NBs unlike e-H3.1 (S4C Fig) [42,43]. In1374 infection of BJ e-H3.1/3-expressing cells led to the co-localization of viral genomes almost exclusively with e-H3.3 (Fig 4Ai, 4ii and 4B). Importantly, e-H3.3 co-localized with HSV-1 genomes together with PML in vDCP NBs (Fig 4C). The lack of co-localization of viral genomes with e-H3.1 was in agreement with the absence of any of the H3.1 CAF-1 chaperone subunits (p150, p60, p48) in the vDCP NBs ( Fig 4D, S1 Table). To confirm that e-H3.3, unlike e-H3.1, interacted with HSV-1 genomes, ChIP-qPCR were conducted on the same loci as those analyzed above. As expected, e-H3.3, but not e-H3.1, was highly enriched on the viral genome Immuno-FISH performed in human primary fibroblasts (BJ cells) infected for 2 days with the replication-defective HSV-1 virus in1374. HIRA (i), UBN1 (ii), CABIN1 (iii), ASF1a (iv) (green), and HSV-1 genomes (red) were detected. Scale bars = 5 μm. (B) Immuno-FISH performed in adult mouse primary TG neuron cultures infected for 2 days with in1374. Mouse Hira (mHira, green), HSV-1 genomes (red), and nucleus (inset, gray) are detected. Scale bar = 5 μm. (C) Same as (B) but in TG neurons from 6-day HSV-1wt-infected mice. Scale bars = 10 μm. (D) Same as (B) in TG neurons of 28-day HSV-1wt-infected mice. Scale bars = 10 μm. (E) Quantifications of immuno-FISH performed in BJ cells infected with in1374 at early times pi. Data represent the percentage of co-localization between incoming HSV-1 genomes and representative proteins of the PML NBs (PML, Sp100, SUMO-1, SUMO 2/3) or H3.3 chaperone complex proteins (DAXX, ATRX, HIRA, UBN1, CABIN1). Means from three independent experiments ± SD. The Student's t-test was applied to assess the significance of the results. Ã = p< 0.05 (see S2 Table for  independently of the examined locus ( Fig 4E). Several cellular loci were analyzed as controls for specific enrichments with H3.3 (Enhancer 1 (Enh.1) on chromosome 9, [57]), or H3.1 (leucine-zipper-like transcriptional regulator 1 (LZTR1) on chromosome 22, GEO accession number GSM1135044). Similar data were obtained for all other canonical histones (S5 Fig), confirming that H3.3 association with latent/quiescent HSV-1 genomes is in a nucleosomal context. To confirm that the discrepancy between the binding of e-H3.3 and e-H3.1 to viral genomes was not due to the ectopic expression of histones, we performed similar experiments using antibodies against native proteins. One specific antibody for H3.3, and suitable for ChIP experiments has previously been described [58]. We performed ChIP using antibodies against native H3.1/2 or H3.3 in normal BJ cells infected for 24 h by in1374. The results were similar to those obtained in infected BJ e-H3.3 using the anti-HA antibody (S6 Fig). These data confirmed that no bias was introduced in the ChIP experiments due to the use of tagged histones, and that latent/quiescent HSV-1 genomes are chromatinized with H3.3. The gradual interaction of the four proteins of the H3.3 chaperone complexes with the incoming viral genomes anticipated similar changes in the interaction of H3.3. ChIP-qPCR were performed at 30 min pi, 2 hpi and 6 hpi, using the same experimental conditions as above. The data showed an overall weak or lack of, H3.3 association with the viral genomes at 30 min pi, followed by an increased interaction at 2 hpi and 6 hpi. These data show that the H3.3 chromatinization of the incoming HSV-1 genomes is progressive and follows a kinetic that matches that observed with the proteins of the H3.3 chaperone complexes. The data also fit with recently published data showing the interaction of incoming viral genomes with canonical histones by 2 hpi [55].

The H3.3K9me3 chromatin mark is predominantly found on vDCP NBsassociated latent/quiescent HSV-1 genomes
Both constitutive (H3K9me2, H3K9me3) and facultative (H3K27me3) heterochromatin marks have been found on various loci on latent HSV-1 genomes in vivo [31,33,34]. To analyze the association of these marks with vDCP NBs-associated latent/quiescent HSV-1 genomes, ChIP were performed targeting H3K9me3, H3K27me3 and one euchromatic mark H3K4me2 as a control ( Fig 5). HSV-1 genomes were exclusively associated with H3K9me3 ( Fig 5A), matching previous results obtained using quiescent viruses [59,60]. In contrast H3K27me3 (Fig 5B) or H3K4me2 (Fig 5C) marks were not detected. Cellular genes previously described for their association with either marks were analyzed for the specificity of the antibodies used (Zinc-finger protein 554 (ZNF554)/H3K9me3 [61]); myelin transcription factor 1 (MYT1)/H3K27me3 ( [62]; Actin/H3K4me2). To confirm that the K9me3 modification is present on H3.3 associated with the HSV-1 genomes, Re-ChIP was performed targeting first H3K9me3 then e-H3.3 in infected BJ and BJ e-H3.3 ( Fig 5D). An overall enrichment for H3.3 from samples initially ChIPed with the H3K9me3 antibody was detected only in BJ e-H3.3 and not BJ cells, with 17 viral loci over 31 (55%) showing significant enrichment. The cellular locus, family with sequence similarity 19 member A2 (FAM19A2) specifically enriched with H3.3K9me3 (GEO accession numbers: GSM1358809 (H3.3), and GSM1289412 (H3K9me3)) was used as positive control. These data show that (i) the Re-ChIP experiment is specific of e-H3.3 and (ii) H3.3K9me3 is indeed associated with the vDCP NB-associated HSV-1 latent/quiescent genomes.  Table). These data show that the inactivation of any of the H3.3 chaperone complex affects to a certain extent the fate of vDCP NBs suggesting a connection between the activity of each H3.3 chaperone complex and the formation and/or maintenance of the vDCP NBs.
We then analyzed the potential impact of the loss of vDCP NB stability on the H3.3-dependent HSV-1 chromatinization. We performed H3.3 ChIP in in1374-infected BJ e-H3.3 cells that had been previously depleted for HIRA, UBN1, DAXX or ATRX using one of the previously validated shRNAs (S9A and S9B Fig). The data showed that overall the inactivation of UBN1, DAXX or ATRX, had a weak impact on the association of H3.3 with the viral loci (1 to 3 loci significantly affected over 31, 3.2 to 9.6%) ( Fig 6C). The depletion of HIRA had a relatively greater effect (6/31, 19.4%). To analyze if simultaneous inactivation of both complexes would significantly impact on HSV-1 chromatinization with H3.3, one protein of each complex was inactivated at the same time before performing HSV-1 infection ( Fig 7A). Individual inactivation of HIRA and ATRX is known to lead to the functional inactivation of the HIRA and DAXX/ATRX complexes, respectively [35,52,63,64]. We noticed that the inactivation of HIRA by a siRNA was not as efficient as the shRNA on preventing the association of H3.3 with viral genomes (Fig 7B). This is likely due to differences in the efficiency of the siRNA compared to the shRNA (compare WBs of Figs S7B and 7A), and to the transitory effect of the siRNA compared to the stable effect of the shRNA at the time of the infection (see Materials and Methods). Nonetheless, a significant decrease of the association of H3.3 with a large number of viral loci (20/31, 64.5%) was measured by the simultaneous inactivation of HIRA and ATRX compared to their individual inactivation ( Fig 7B). These results indicate that the DAXX/ATRX complex may compensate for the loss of the HIRA complex on the chromatinization of latent/quiescent HSV-1 genomes with H3.3, and conversely.

PML NBs are essential for H3.3 chromatinization of latent/quiescent HSV-1 genomes
The above experiments were conducted in a context where the cells, although deficient for the activity of one H3.3 chaperone complex at a time, still contained intact PML NBs accumulating Immediate early (IE/α) genes (red), early (E/β) genes (green), late (L/γ) genes (blue). (B) Chromatin immunoprecipitation (ChIP) associated with qPCR performed in in1374-infected normal BJ cells or in1374-infected BJ cells expressing tagged versions of DAXX, HIRA, or UBN1. Infections were performed for 24 h. Anti-myc (DAXX) or anti-HA (HIRA and UBN1) antibodies were used. For ATRX, a native antibody was used, and the results were compared to ChIP with IgG as control. Means from three independent experiments ± SD. Student's t-test was applied to assess the significance of the results. Ã = p< 0.05.
https://doi.org/10.1371/journal.ppat.1007313.g003 e-H3.3 (S7 and S10 Fig). Therefore, we hypothesized that the accumulation of H3.3 within the PML NBs could be one of the key events acting upstream of the H3.3 chaperone complex activity for the induction of chromatinization of the latent/quiescent HSV-1 by H3.3. We analyzed the HSV-1 chromatinization in cells lacking PML NBs. In a previous study conducted in HSV-1 latently infected PML KO mice, we showed that the absence of PML significantly impacted the number of latently infected TG neurons showing the "single"/vDCP NB HSV-1 pattern and favored the detection of neurons containing the "multiple-latency" pattern prone to LAT expression [16,53]. We analyzed the very few neurons showing a "single"/vDCP NB-like pattern in the latently infected PML KO mice for the co-localization of DAXX and ATRX with the viral genomes. We could not detect any of the two proteins co-localizing with the latent HSV-1 genomes (Fig 8Ai to 8vi). Although informative, these in vivo studies did not allow the analysis of the real impact of the absence of PML on the co-localization of the other PML NBassociated proteins with latent HSV-1 genomes, because the neurons showing the "single"/ vDCP NB-like pattern were too few to quantify the effect. We thus depleted PML in normal BJ cells using a PML shRNA-expressing lentiviral transduction approach. We verified the efficiency of the shRNAs against PML in normal BJ cells by IF, RT-qPCR and WB (S11A-S11C Fig). PML-depleted BJ cells were superinfected with in1374, and immuno-FISH was performed at 2 dpi to analyze the co-localization of HSV-1 genomes with DAXX, ATRX, HIRA, and UBN1 ( Fig 8B). Notably, both PML shRNAs gave similar results. The quantification of the data showed that, similarly to the in vivo situation, the depletion of PML significantly decreased the co-localization of DAXX and ATRX with latent/quiescent HSV-1 genomes, leaving HIRA and UBN1 unaffected for their co-localization ( Fig 8C, and S5 Table). Thus, we analyzed whether the failure of DAXX/ATRX to co-localize with the latent/quiescent HSV-1 genomes in the absence of PML NBs, could impact the chromatinization of HSV-1 with H3.3.
We first generated BJ e-H3.3 cells depleted for PML by shRNA-expressing lentiviral transduction similarly to the BJ cells (S11D and S11E Fig). BJ e-H3.3 control or PML-depleted cells were superinfected with in1374 to perform immuno-FISH and analyze the co-localization of HSV-1 genomes with H3.3 ( Fig 9A). Quantification of the data showed a significant decrease in the co-localization of latent/quiescent HSV-1 genomes with H3.3 compared with controls ( Fig 9B), suggesting an impact of the absence of PML NBs on the latent/quiescent HSV-1 association with H3.3. To complement these results at a more quantitative level, we performed ChIP on e-H3. 3. The data showed a major impact of the absence of PML NBs on the H3.3 association with viral genomes, with a significant depletion of H3.3 on multiple loci (21/31, 68%) (Fig 9C). This could not be due to an indirect effect of PML depletion on H3.3 stability because e-H3.3 protein levels were similar in control cells and cells depleted for PML (Fig 9D). Both PML shRNAs gave similar results. To confirm that the absence of PML had an impact on . Means from three independent experiments ± SD. The Student's ttest was applied to assess the significance of the results. ÃÃ = p< 0.01. (C) Immuno-FISH performed in e-H3.3-expressing BJ cells infected for 2 days with in1374. E-H3.3 (gray, green), HSV-1 (gray, red), and PML (gray, blue) were detected. Scale bar = 5 μm. (D) Immuno-FISH performed in normal BJ cells infected for 2 days with in1374. H3.1 CAF chaperone complex proteins p150 (i), p60 (ii), and p48 (iii) (green), and HSV-1 genomes (red) were detected. Nuclei were detected with DAPI (blue). Scale bars = 5 μm. (E) ChIP performed in in1374-infected normal BJ cells (blue), in1374-infected e-H3.1 (red) or e-H3.3 (green) expressing BJ cells. Infections were performed for 24 h. Anti-HA antibody was used for ChIP experiments. Analyzed viral loci were described previously. Cellular loci Enhancer 1 (Enh.1), and leucine-zipperlike transcriptional regulator 1 (LZTR1) are positive controls for deposition of H3.3 and H3.1, respectively. Means from three independent experiments ± SD. The Student's t-test was applied to assess the significance of the results. Ã = p< 0.05, ÃÃ = p< 0.01. (F) ChIP performed in in1374-infected e-H3.3-expressing BJ cells at early times pi, 30 min (blue), 2 h (red), 6 h (green). Anti-HA antibody was used for ChIP experiments. Analyzed viral loci were described previously. Means from three independent experiments ± SD.
https://doi.org/10.1371/journal.ppat.1007313.g004 3 association with latent/quiescent viral genomes, we performed ChIP on in1374-infected control MEF pml +/+ or MEF pml -/cells previously engineered by lentiviral transduction to express e-H3.3 ( Fig 9E). The data confirmed the impaired association of e-H3.3 with latent/ quiescent HSV-1 genomes in the absence of PML, with 26/31 (84%) viral loci significantly impacted (Fig 9F). Cellular loci acid Sensing Ion Channel Subunit 2 (Asic2), and Heme Oxygenase 1 (Hmox1) were used respectively as positive and negative controls for deposition of H3.3 in the absence of Pml as described in [44]. To definitively attribute the lack of deposition of H3.3 on viral loci to the absence of PML, MEF pml -/-;e-H3.3 cells were engineered to allow reexpression, under doxycycline induction, of the isoform I of human PML (PML.I) (Fig9G), which was shown to participate to the HSV-1 antiviral restriction mechanism [65]. The formation of PML NBs after induction of PML.I was visualized by IF ( Fig 9H). ChIPs were then performed on in1374-infected MEF pml -/-;e-H3.3;myc-PML.I cells previously treated or not with doxycycline ( Fig 9I). The data showed that the re-expression of PML.I allowed the re-loading of H3.3 on all the analyzed loci of the latent/quiescent viral genomes with significant results obtained for 21 loci over 31 (68%), demonstrating the essential role of PML/PML NBs in the association of H3.3 with incoming viral genomes. Finally, we wanted to analyze whether the deficit of the H3.3 association with the viral genome in the absence of PML could be compensated by an increase of H3.1 on viral loci. The data from BJ e-H3.1 cells depleted for PML or MEF pml -/-;e-H3.1 cells, and infected with in1374 showed that H3.1 did not replace H3.3 on the viral loci (S12A and S12B Fig). Altogether, these data demonstrate the essential role of PML NBs, probably through the DAXX/ATRX complex activity, in the exclusive H3.3 chromatinization of incoming viral genomes forced to adopt a vDCP NB-associated latent/quiescent pattern due to a deficit in the onset of lytic cycle.

The destabilization of vDCP NBs induces the recovery of HSV-1 transcriptional activity and the formation of replication compartments
vDCP NB-associated latent genomes have been shown to be transcriptionally silent for the LAT expression in vivo [16], and for the expression of a reporter gene in vitro in mouse TG neuron cultures [17], and in human primary fibroblasts [45]. Moreover, it is known that the viral protein ICP0 induces the destabilization of PML NBs [66] and is essential for HSV-1 reactivation in vivo [67], and for the transcriptional de-repression of a silenced viral genome in vitro [45,59,60]. However, it is not known if the transcriptional recovery is correlated to the destabilization of the vDCP NBs. We analyzed if latent HSV-1 genomes trapped in vDCP NBs were definitively silenced or could resume a transcriptional program leading to replication of viral genomes provided that vDCP NBs were destabilized. ICP0 or its non-functional RING finger mutant (ICP0ΔRF) were expressed from BJ-eTetR/cICP0 or BJ-eTetR/cICP0ΔRF cells harboring vDCP NBs for 4 days (Fig 10). HSV-1 in1374 infected BJ-eTetR cells were used as controls. Expression of ICP0 or ICP0ΔRF was induced for 24 h, 48 h or 72h at the permissive temperature for in1374 replication (32˚C) (S13 Fig). Transcription of the reporter (LacZ) gene was measured by RT-qPCR to analyze the transcriptional recovery of the vDCP NB-associated latent/quiescent viral genomes (Fig 10A). The addition of doxycycline in infected BJ-eTetR or BJ-eTetR/c ICP0ΔRF cells did not lead to any significant transcription of the LacZ gene. Only  cICP0 cells (Fig 10B) but not BJ-eTetR/cICP0ΔRF (S15 Fig) showed a clear disappearance of the vDCP NBs. Concomitantly, only BJ-eTetR/cICP0 cells showed the formation of replication compartments (RCs) indicating that the virus is in the process of lytic phase following vDCP NBs destruction by ICP0. To confirm that the lytic transcriptional program was indeed occurring, viral transcripts of all kinetics were analyzed (Fig 10C). Twenty four and 48 h post ICP0  induction, lytic genes were expressed with a clear switch towards the γ genes (UL44/gC and US6/gD) at 48 h confirming the onset of the lytic transcriptional program. Expression of ICP0ΔRF did not enable the re-expression of viral genes. These data show that vDCP NBsassociated latent/quiescent HSV-1 genomes can resume transcription and a lytic program provided that the vDCP NBs are destabilized, suggesting that these genomes are not definitively silenced, and could participate to the reactivation process of HSV-1.

Discussion
The HSV-1 genome enters the nucleus of infected neurons, which support HSV-1 latency as a naked/non-nucleosomal DNA. Many studies have described the acquisition of chromatin marks on the viral genome concomitantly to the establishment, and during the whole process, of latency. Paradoxically, although it is undisputable that these chromatin marks will predominantly be associated with latency and reactivation, few data are available for the initiation of the chromatinization of the incoming viral genome. Here, we demonstrate the essential contribution of PML NBs in the process of chromatinization of incoming HSV-1 genomes meant to remain in a latent/quiescent state. We showed that PML NBs are essential for the association of the histone variant H3.3 with the latent/quiescent HSV-1.
Two members of the HIRA complex, HIRA and ASF1a, were previously shown to be involved in H3.3-dependent chromatinization of HSV-1 genomes at early times after infection in non-neuronal and non-primary cells favoring the onset of the lytic cycle [68,69]. Moreover a recent study highlighted the interaction of HIRA with quiescent HSV-1 and plasmid DNA in primary human fibroblasts [56]. Our in vivo data in TG neurons and in vitro data in infected human primary fibroblasts or adult mice TG neuron cultures, show that all the proteins of the HIRA complex accumulate within specific nucleoprotein structures called the viral DNA-containing PML NBs or vDCP NBs. vDCP NBs contain transcriptionally silent HSV-1 genome that we previously demonstrated in vivo to be associated with the establishment of latency from the early steps of neuron infection [17]. Additionally, our data show that: (i) the mouse Hira protein, in vivo, and all the components of the HIRA complex, in cultured cells, temporarily accumulate in vDCP NBs, and (ii) significantly greater amount of incoming HSV-1 genomes co-localize with HIRA compared with PML at very early times pi (30 min). These data suggest that the HIRA complex could also be involved to some extent in the establishment of HSV-1 latency by the initial recognition of the incoming naked/non-nucleosomal viral DNA and the chromatinization of non-replicative HSV-1 genomes intended to become latent. In this respect, a recent study suggested an anti-viral activity associated with HIRA against HSV-1 and murine cytomegalovirus lytic cycles [56]. To that extent, although they are both functionally essential for the activity of the HIRA complex [35,51,64,70], our data show that the depletion of HIRA has a greater effect compared to the UBN1 depletion, on the H3.3 association with the viral genomes. This could be simply explained by a better efficiency of the HIRA, compared to the UBN1, shRNAs. Alternatively, HIRA was shown to be recruited to UV-induced DNA damage independently of UBN1 (see figure S2D in [71]), and to participate to the loading of newly synthesized H3.3 on chromatin [72]. Therefore, the depletion of HIRA could indirectly and/or directly impact on two initial events occurring concomitantly to the entry of the viral genomes in the nucleus; first a signaling pathway associated to the detection of DNA breaks present in incoming viral DNA as suggested in [55,73]; and second the chromatinization process per se. If these two events are linked it could explain the differences observed between the HIRA and UBN1 depletion on the loading of H3.3 on the viral genomes. Experiments are in progress to investigate this.
Interestingly, proteins of the HIRA complex have been previously shown to be able to directly bind to naked DNA in a sequence-independent manner, in contrast to DAXX and ATRX [52]. Nevertheless, our ChIP data highlight the interaction of viral genomes with DAXX and ATRX, but we cannot assert that the two proteins directly interact with naked DNA. The gamma-interferon-inducible protein 16 (IFI16), a member of the PYHIN protein family, has been described as a nuclear sensor of incoming herpesviruses genomes, and suggested to promote the addition of specific chromatin marks that contribute to viral genome silencing [74][75][76][77][78][79][80][81]. A proteomic study determining the functional interactome of human PYHIN proteins revealed the possible interaction between ATRX and IFI16 [82]. Thus, it will be interesting to determine in future studies if IFI16 and H3.3 chaperone complexes physically and functionally cooperate in the process of chromatinization of the latent/quiescent HSV-1 genome.
One of the main finding of our study is the demonstration of the essential contribution of PML NBs in the H3.3-dependent chromatinization of the latent/quiescent HSV-1 genomes. A close link between PML NBs and H3.3 in chromatin dynamics has been demonstrated during oncogene-induced senescence (OIS). In OIS, expression of the oncogene H-RasV12 induces DAXX-dependent relocalization of neo-synthesized H3.3 in the PML NBs before a drastic reorganization of the chromatin to form senescence-associated heterochromatin foci [42,43]. Hence, the contribution of the PML NBs in the deposition of H3.3 on specific cellular chromatin loci has also been reported [43,44]. The present study shows that the absence of Pml in HSV-1wt latently infected Pml KO mice, or the depletion of PML by shRNA in BJ cells infected with in1374, significantly affects the co-localization of DAXX and ATRX, but not HIRA and UBN1, with latent/quiescent HSV-1 genomes, confirming previous studies for DAXX and ATRX [45]. Taken together with the impaired association of H3.3 with the viral genomes in the absence of PML NBs, these data suggest that a significant part of the latent/quiescent HSV-1 genome chromatinization by H3.3 could occur through the activity of the DAXX/ATRX complex in association with the PML NBs.
Given the particular structure formed by the latent/quiescent HSV-1 genome with the PML NBs, our study raises the question of the possible acquisition of a chromatin structure within the vDCP NBs. The individual inactivation of DAXX, ATRX, HIRA, or UBN1 significantly impacts the co-localization of the latent/quiescent HSV-1 genomes with PML, and hence the formation of vDCP NBs. However, it only mildly affects the association of H3.3 with viral genomes, suggesting an absence of correlation between the formation of vDCP NBs and H3.3 chromatinization. However, our data show that the depletion of DAXX, ATRX, HIRA, or UBN1 does not modify the accumulation of e-H3.3 at PML NBs, leaving intact the upstream requirement of H3.3 accumulation in PML NBs for H3.3-dependent viral chromatin assembly. We have recently shown that vDCP NBs are dynamic structures that can fuse during the course of a latent infection [17]. It is thus possible that incoming viral genomes can be dynamically associated with vDCP NBs to be chromatinized, and in the absence of any of the H3.3 chaperone complex subunit, this dynamic can be perturbed, resulting in some viral genomes that do not show a co-localization with PML. Given that depletion of none of the four proteins affects the structure of the PML NBs, and considering the essential role of PML NBs in the H3.3 chromatinization of the viral genomes, this possibility cannot be ruled out. The depletion of H3.3, which almost exclusively participates in latent/quiescent HSV-1 genome chromatinization compared to H3.1/2, does not prevent the formation of vDCP NBs (S16 Fig), and is rather in favor of a chromatinization of the viral genome in the vDCP NBs. It is unlikely that canonical H3.1/2 could replace H3.3 for the chromatinization of the incoming HSV-1 genomes prior to the formation of the vDCP NBs. Indeed, our multiple immuno-FISH and ChIP assays failed to detect H3.1/2 and/or H3.1/2 chaperones that associate or co-localize with viral genomes. Nonetheless, we cannot rule out a possible replacement of H3.3 with another H3 variant for the chromatinization of viral genomes before their entrapment by the PML NBs to form vDCP NBs.
Our data show that the vDCP NBs-associated HSV-1 genomes are chromatinized with H3K9me3, and the Re-ChIP assays confirm an association with H3.3K9me3, but not H3K27me3. In vivo, it has been shown that both H3 modifications could be found on latent HSV-1 genomes [31,33,34]. One simple explanation could reside in the heterogeneity of latent genomes distribution within the nuclei of the infected neurons in the in vivo mouse and/or rabbit models of latency [16,17,54], however this would need to be formally demonstrated. Though, vDCP NBs-associated HSV-1 genomes remain compatible with the transcription of lytic genes provided that the vDCP NBs are destabilized by ICP0, a viral protein known to be required for full in vivo reactivation [67], and to erase chromatin marks associated with latent/ quiescent viral genomes in vitro [59]. Therefore, vDCP NBs are not a dead end for the virus life cycle, and HSV-1 latently infected neurons containing vDCP NBs are likely to contribute to the process of reactivation.
Altogether, our study demonstrates the essential role of a PML NB/H3.3/H3.3 chaperone axis in the process of chromatinization of viral genomes adopting a vDCP NB pattern, which represents an essential structural and functional aspect of HSV-1 latency establishment. Given the involvement of H3.3 in the chromatinization of other latent herpesviruses belonging to different sub-families than HSV-1, such as EBV [83] and HCMV [58], as well as adenovirus type 5 [84], this pathway of chromatinization is likely to play a major role in the biology of the whole Herpesviridae family, and possibly of other DNA viruses such as adenoviruses, papillomaviruses, hepatitis B virus, and retroviruses.

Virus strains, mice and virus inoculation, primary mouse TG neuron cultures, cells
The HSV-1 SC16 strain was used for mouse infections and has been characterized previously [85]. The HSV-1 mutant in1374 is derived from the 17 syn + strain and expresses a temperature-sensitive variant of the major viral transcriptional activator ICP4 [86] and is derived from in1312, a virus derived from the VP16 insertion mutant in1814 [87], which also carries a deletion/frameshift mutation in the ICP0 open reading frame [88] and contains an HCMV-lacZ reporter cassette inserted into the UL43 gene of in1312 [89]. This virus has been used and described previously [17,45]. All HSV-1 strains were grown in baby hamster kidney cell (BHK-21, ATCC, CCL-10) and titrated in human bone osteosarcoma epithelial cells (U2OS, ATCC, HTB-96). In1374 was grown and titrated at 32˚C in the presence of 3 mM hexamethylene bisacetamide [90].
Mice were inoculated and TG processed as described previously [16]. Briefly, for the lip model: 6-week-old inbred female BALB/c mice (Janvier Labs, France) were inoculated with 10 6 PFU of SC16 virus into the upper-left lip. Mice were sacrificed at 6 or 28 dpi. Frozen sections of mouse TG were prepared as described previously [16,92]. For the eye model: inoculation was performed as described previously [93]. Briefly, prior to inoculation, mice were anesthetized by intra-peritoneal injection of sodium pentobarbital (50 mg/kg of body weight). A 10-μL drop of inoculum containing 10 5 PFU of 17syn+ was placed onto each scarified corneal surface. This procedure results in~80% mouse survival and 100% infected TG.
Primary mouse TG neuron cultures were established from OF1 male mice (Janvier lab), following a previously described procedure [17]. Briefly, 6-8-week-old mice were sacrificed before TG removal. TG were incubated at 37˚C for 20 min in papain (25 mg) (Worthington) reconstituted with 5 mL Neurobasal A medium (Invitrogen) and for 20 min in Hank's balanced salt solution (HBSS) containing dispase (4.67 mg/mL) and collagenase (4 mg/mL) (Sigma) on a rotator, and mechanically dissociated. The cell suspension was layered twice on a five-step OptiPrep (Sigma) gradient, followed by centrifugation for 20 min at 800 g. The lower ends of the centrifuged gradient were transferred to a new tube and washed twice with Neurobasal A medium supplemented with 2% B27 supplement (Invitrogen) and 1% penicillin-streptomycin (PS). Cells were counted and plated on poly-D-lysine (Sigma)-and laminin (Sigma)coated, eight-well chamber slides (Millipore) at a density of 8,000 cells per well. Neuronal cultures were maintained in complete neuronal medium consisting of Neurobasal A medium supplemented with 2% B27 supplement, 1% PS, L-glutamine (500 μM), nerve growth factor (NGF; 50 ng/mL, Invitrogen), glial cell-derived neurotrophic factor (GDNF; 50 ng/mL, Pepro-Tech), and the mitotic inhibitors fluorodeoxyuridine (40 μM, Sigma) and aphidicolin (16.6 μg/ mL, Sigma) for the first 3 days. The medium was then replaced with fresh medium without fluorodeoxyuridine and aphidicolin.
Primary  (Sigma, F7524), L-glutamine (1% v/v), 10 U/mL penicillin, and 100 mg/mL streptomycin. BJ cell division is stopped by contact inhibition. Therefore, to limit their division, cells were seeded at confluence before being infected at a multiplicity of infection (m.o.i.) of 3, and then maintained in 2% serum throughout the experiment. Infections of BJ cells for short times (from 30 min to 6 h) were performed by synchronizing the infection process with a pre-step of virus attachment to the cells at 4˚C for one hour. The infection medium was then removed, and the temperature was shifted to 37˚C to allow a maximum of viruses to simultaneously penetrate into the cells.

Frozen sections
Frozen sections of mouse TG were generated as previously described [92]. Mice were anesthetized at 6 or 28 d.p.i., and before tissue dissection, mice were perfused intracardially with a solution of 4% formaldehyde, 20% sucrose in 1X PBS. Individual TG were prepared as previously described [92], and 10-μm frontal sections were collected in three parallel series and stored at -80˚C.

DNA-FISH and immuno-DNA FISH
HSV-1 DNA FISH probes consisting of cosmids 14, 28 and 56 [94] comprising a total of~90 kb of the HSV-1 genome were labeled by nick-translation (Invitrogen) with dCTP-Cy3 (GE Healthcare) and stored in 100% formamide (Sigma). The DNA-FISH and immuno-DNA FISH procedures have been described previously [16,92]. Briefly, infected cells or frozen sections were thawed, rehydrated in 1x PBS and permeabilized in 0.5% Triton X-100. Heat-based unmasking was performed in 100 mM citrate buffer, and sections were post-fixed using a standard methanol/acetic acid procedure and dried for 10 min at RT. DNA denaturation of the section and probe was performed for 5 min at 80˚C, and hybridization was carried out overnight at 37˚C. Sections were washed 3 x 10 min in 2 x SSC and for 3 x 10 min in 0.2 x SSC at 37˚C, and nuclei were stained with Hoechst 33258 (Invitrogen). All sections were mounted under coverslips using Vectashield mounting medium (Vector Laboratories) and stored at 4˚C until observation.
For immuno-DNA FISH, cells or frozen sections were treated as described for DNA-FISH up to the antigen-unmasking step. Tissues were then incubated for 24 h with the primary antibody. After three washes, secondary antibody was applied for 1 h. Following immunostaining, the cells were post-fixed in 1% PFA, and DNA FISH was carried out from the methanol/ acetic acid step onward.
The same procedures were used for infected neuronal cultures except that the cells were fixed in 2% PFA before permeabilization.

Western blotting
Cells were collected in lysis buffer (10 mM Tris-EDTA, pH 8.0) containing a protease inhibitor cocktail (Complete EDTA-free; Roche) and briefly sonicated. Protein extracts were homogenized using QiaShredders (Qiagen). The protein concentration was estimated by the Bradford method. Extracted proteins were analyzed by Western blotting using appropriate antibodies (see below).

Microscopy, imaging, and quantification
Observations and most image collections were performed using an inverted Cell Observer microscope (Zeiss) with a Plan-Apochromat ×100 N.A. 1.4 objective and a CoolSnap HQ2 camera from Molecular Dynamics (Ropper Scientific) or a Zeiss LSM 800 confocal microscope. Raw images were processed using ImageJ software (NIH).
Quantitative PCR was performed using Quantifast SYBR Green mix (Qiagen) and the MX3005P apparatus (Agilent/Stratagene). Primers were used at a final concentration of 1 μM. Their sequences and target genes are provided in Tables 2 and 3.

Re-ChIP
Cells were processed similarly to ChIP until addition of the antibody. Two μg of the first antibody were pre-incubating with agarose beads coupled to protein A (Millipore 16-157) in PBS-0.5% BSA overnight at 4˚C under shaking. The beads were washed twice with IP Buffer 1X (150mM NaCl, 5mM Tris pH8; 0.5mM EDTA; 0.1% SDS; 1% Triton), then the chromatin was incubated with the antibody/beads overnight at 4˚C. Beads were then successively washed for 5 min at 4˚C under constant shaking once in "low salt" buffer, once in "high salt" buffer, once in "LiCl" buffer and twice in TE buffer. Chromatin-antibody complexes were eluted at 37˚C for 30 min with 100 μL of Re-ChIP elution buffer (1% SDS, 0.1M NaHCO3, 10mM DTT). Fifty μL were kept for analysis of the first capture efficiency, then the other 50 μL were diluted 20 times with IP buffer 1X and incubated overnight at 4˚C with the second antibody pre-incubated with agarose beads coupled to protein A. The beads were washed twice with IP Buffer 1X, then successively washed for 5 min at 4˚C under constant shaking, once in "low salt" buffer, once in "high salt" buffer, once in "LiCl" buffer and twice in TE buffer. Chromatin-antibody complexes were then eluted at 65˚C for 30 min under constant shaking with 200 μL of elution buffer (1% SDS, 0.1M NaHCO3). Input and IP products were de-crosslinked overnight at 65˚C a with 20 mg/mL of proteinase K (Sigma) and 10 mg/mL of RNAse A (Sigma). DNA was then purified by phenol-chloroform/ethanol precipitation, resuspended in water, and kept at -20˚C until use for qPCR.

Antibodies
The antibodies used for immunofluorescence, ChIP and WB are provided in Tables 4-6.  infected for 2 days with the replication-defective HSV-1 virus in1374. PML (i), Sp100 (ii), SUMO-1 (iii), SUMO 2/3 (iv), ATRX (v), DAXX (vi) (green), and HSV-1 genomes (red) were detected. Scale bars = 5 μm. RT-qPCR to detect DAXX, ATRX, HIRA, and UBN1 mRNA was performed, and the results were compared to a control shRNA (shCTRL). Means from three independent experiments ± SD. The Student's t-test was applied to assess the significance of the results. Ã = p< 0.05, ÃÃ = p< 0.01. (B) WB for detection of decreases in DAXX, ATRX, HIRA, and UBN1 proteins in normal BJ cells or BJ cells transduced with shRNA-expressing lentiviruses (48 h post-transduction). Actin was detected as a loading control. Two shRNAs were tested for each protein. Cells were infected with in1374 for 24 h. Anti-HA antibody was used for the ChIP experiments. The analyzed viral loci were described previously. Data represent means from two independent experiments ± SD. (TIF)