Human Cytomegalovirus Major Immediate Early 1 Protein Targets Host Chromosomes by Docking to the Acidic Pocket on the Nucleosome Surface

Abstract

The 72-kDa immediate early 1 (IE1) protein encoded by human cytomegalovirus (hCMV) is a nuclearly localized promiscuous regulator of viral and cellular transcription. IE1 has long been known to associate with host mitotic chromatin, yet the mechanisms underlying this interaction have not been specified. In this study, we identify the cellular chromosome receptor for IE1. We demonstrate that the viral protein targets human nucleosomes by directly binding to core histones in a nucleic acid-independent manner. IE1 exhibits two separable histone-interacting regions with differential binding specificities for H2A-H2B and H3-H4. The H2A-H2B binding region was mapped to an evolutionarily conserved 10-amino-acid motif within the chromatin-tethering domain (CTD) of IE1. Results from experimental approaches combined with molecular modeling indicate that the IE1 CTD adopts a β-hairpin structure, docking with the acidic pocket formed by H2A-H2B on the nucleosome surface. IE1 binds to the acidic pocket in a way similar to that of the latency-associated nuclear antigen (LANA) of the Kaposi's sarcoma-associated herpesvirus. Consequently, the IE1 and LANA CTDs compete for binding to nucleosome cores and chromatin. Our work elucidates in detail how a key viral regulator is anchored to human chromosomes and identifies the nucleosomal acidic pocket as a joint target of proteins from distantly related viruses. Based on the striking similarities between the IE1 and LANA CTDs and the fact that nucleosome targeting by IE1 is dispensable for productive replication even in “clinical” strains of hCMV, we speculate that the two viral proteins may serve analogous functions during latency of their respective viruses.

INTRODUCTION

Nuclear DNA is largely organized and controlled through nucleosomes. Each nucleosome typically assembles 146 to 147 bp of DNA in 1.65 superhelical turns around a core histone octamer composed of a central H3-H4 tetramer flanked by two H2A-H2B dimers (1, 2). Additionally, linker histone H1 binds to the nucleosome at the DNA entry/exit points outside the octamer. Thousands of nucleosomes along the DNA compose a “beads-on-a-string” array, which can further condense into higher-order chromatin, attaining its most compact state during mitosis (reviewed in references 3 and 4).

Herpesviruses transcribe, synthesize, and package their double-stranded DNA genomes in the host cell nucleus, where they contend with and exploit chromatin to aid viral replication and persistence. There are several examples of herpesvirus proteins interacting with cellular mitotic and/or interphase chromatin. The viral proteins attach to chromatin through at least three types of nonmutually exclusive molecular targets: DNA, histones, and chromatin-associated nonhistone proteins. For instance, Epstein-Barr virus (EBV), a gammaherpesvirus, encodes the EBV nuclear antigen 1 (EBNA1), which targets host chromatin by interacting with adenosine/thymidine-rich DNA sequences (5) and EBNA1 binding protein 2 (6–9). In contrast, the latency-associated nuclear antigen 1 (LANA1 or LANA) encoded by another gammaherpesvirus, Kaposi's sarcoma-associated herpesvirus (KSHV), associates with chromatin through complex formation with nonhistone factors, including methyl CpG-binding protein (10), the core histones H2A and H2B (11, 12), and, potentially, linker histone H1 (13). In fact, the crystal structure of a nucleosome complexed with the first 23 LANA amino acids revealed that the viral peptide forms a β-hairpin that specifically interacts with an “acidic pocket” formed by the folded regions of the H2A-H2B dimer (12). Through chromatin interaction, LANA and other viral proteins assume fundamental roles in the infectious cycles of their respective viruses (see Discussion).

The 72-kDa immediate early 1 (IE1) protein (called IE1-72kDa, IE72, or IE1) is a nuclear regulatory phosphoprotein of human cytomegalovirus (hCMV), the prototypic betaherpesvirus. IE1 is expressed from the hCMV genome at the onset of infection. Together with the 86-kDa IE2 protein (called IE2-86kDa, IE86, or IE2), IE1 is the most prominent member of the major IE (MIE) family of hCMV gene products, which have been assigned critical functions in virus-host interaction, including innate immune modulation and transcriptional regulation (reviewed in references 14 to 16). IE1 is required for viral early gene expression and replication in human fibroblasts, at least under conditions of low-input multiplicity (17–19). Association of IE1 with condensed chromatin during mitosis was initially described more than 20 years ago (20) and has henceforward been conspicuous to many researchers (see, e.g., references 21 to 28). The interaction with mitotic chromatin can be observed both during hCMV infection and upon ectopic expression of IE1. Chromosome association was first roughly mapped to MIE exon 4 sequences (20) and, subsequently, to residues 421 to 486 of the 491-amino-acid viral protein (24). Eventually, the 16 carboxy-terminal residues (amino acids 476 to 491) of IE1 were determined to be required and sufficient for mitotic chromatin interaction in transfected cells and were consequently termed the chromatin tethering domain (CTD) (22). The ability for chromosome attachment appears to be evolutionarily conserved between IE1 orthologs of primate CMVs (21, 22, 29). However, despite being a conserved and distinctive feature of IE1, the mechanisms underlying chromosome association by the viral protein have not been determined.

The present study was designed to identify and precisely delineate the molecular interactions anchoring the hCMV IE1 protein to human chromosomes and to determine their impact on viral replication in clinical strains of hCMV.

MATERIALS AND METHODS

Plasmids and mutagenesis.

For expression in human cells, plasmid constructs derived from pcDNA3 (Life Technologies), pCGN (30), pCMV.TetO (a gift from Roger Everett, University of Glasgow, Scotland), pEGFP-C1 (Clontech), and pME18S (31) were used. For expression in Escherichia coli, plasmid constructs derived from pGEX-KG (32) were used.

Plasmid pcDNA-HA-IE1 encodes the hCMV (Towne) 72-kDa IE1 (pUL123) protein tagged with an influenza virus hemagglutinin (HA) epitope (21). Plasmid pcDNA-HA-mIE1 encodes an HA-tagged form of the mCMV IE1 (mIE1, pp89) protein and was constructed by inserting a BamHI-EcoRI fragment from pGEX-mIE1 (21) into the same sites of pcDNA-HA-N (33) (a gift from Ronald Hay, University of Dundee, Scotland). For construction of pcDNA-HA-IE2, encoding an HA-tagged form of the hCMV 86-kDa IE2 (pUL122) protein, a BglII-EcoRI fragment from pEGFP-IE2 (21) was inserted into the BamHI and EcoRI sites of pcDNA-HA-N. Plasmid pCGN-pp71 encodes an HA-tagged form of the hCMV pp71 (pUL82) protein (34). Plasmid pCMV.TetO.IE1 has been described (35). For pCMV.TetO.IE1M483A, pCMV.TetO.IE1NBM, and pCMV.TetO.IE1_1–475, mutant hCMV IE1 (Towne) coding sequences were PCR amplified from the template pEGFP-TNIE1 (21) with primers 483 and 1085, 483 and 1086, and 483 and 695, respectively, and inserted into the HindIII and EcoRI sites of pCMV-TetO. To generate pEGFP-IE1_476–491 and related constructs encoding the enhanced green fluorescent protein (EGFP) fused to the IE1 CTD or to CTD variants with single amino acid substitutions (G476A, G477A, K478A, S479A, T480A, M481A, P482A, M483A, V484A, T485A, R486A, S487A, K488A, A489G, D490A, and Q491A), suitable oligonucleotides (635 and 636; 703 and 734) were annealed and inserted into the BglII and EcoRI sites of pEGFP-C1. A construct replacing the first two IE1 coding triplets in pEGFP-IE1_476–491 with stop codons (pEGFP-Stop) was generated in the same way using oligonucleotides 737 and 738. Likewise, pEGFP-LANA_5–22 was generated by ligating annealed oligonucleotides 639 and 640 via BglII and EcoRI sites with pEGFP-C1. All constructs expressing human histone H2A (H2A.2, type 1B/E) were based on pME-Flag-H2A (36) (a gift from Robert Eisenman, Fred Hutchinson Cancer Research Center, Seattle, WA, USA). To generate constructs encoding Flag-tagged H2A with single (E41A, E56A, E61A, E64A, D72A, D90A, E91A, E92A, and E121A) and triple (E61A E64A D90A and D90A E91A E92A) amino acid substitutions, QuikChange site-directed mutagenesis (Stratagene) was performed according to the manufacturer's instructions using suitable oligonucleotides (787 to 808).

In pGEX-KG, the BamHI and EcoRI sites served for all cloning reactions. Plasmid pGEX-IE1 encodes the hCMV (Towne) 72-kDa IE1 protein fused to glutathione S-transferase (GST) (21). To generate pGEX-IE1_1-475, a PCR product amplified from the template pGEX-IE1 with primers 637 and 638 was digested with BamHI and EcoRI and ligated with pGEX-KG. For construction of pGEX-IE1_476-491 and related constructs encoding GST fused to the IE1 CTD or to CTD variants with single amino acid substitutions, suitable oligonucleotides (635 to 636 and 703 to 734) were annealed and ligated with pGEX-KG. A construct replacing the first two IE1 coding triplets in pGEX-IE1_476-491 with two stop codons (pGEX-Stop) was generated in the same way using oligonucleotides 737 and 738. Likewise, pGEX-LANA_5-22 was generated by ligating annealed oligonucleotides 639 and 640 with pGEX-KG. To construct pGEX-PreS-IE1, a DNA fragment encoding the recognition motif for PreScission protease (LEVLFQGP) between the GST and IE1 sequences was generated by overlap extension PCR (37) from the pGEX-KG and pGEX-IE1templates with primers 484 and 603 to 605. The PCR product was digested with BamHI and EcoRI and ligated with pGEX-IE1. Plasmid pGEX-PreS-IE1 was used as a template for PCR with primers 695 and 838, resulting in a DNA fragment that was digested with BamHI and EcoRI and ligated with pGEX-PreS-IE1 to produce pGEX-PreS-IE1_1-475. Finally, pGEX-PreS-IE1_476-491, pGEX-PreS-LANA_5-22, and pGEX-PreS were generated by ligating annealed oligonucleotides 834 and 835, 836 and 837, and 658 and 659, respectively, with pGEX-PreS-IE1.

As a standard for absolute quantification of hCMV genome copies by real-time quantitative PCR (qPCR), plasmid pCR-RPPH1-UL54P was constructed by a two-step procedure. First, a PCR product comprising 71 bp of the hCMV UL54 promoter (UL54P) sequence was amplified from an infected cell genomic DNA template with primers 294 and 295 and was ligated with pCR4-TOPO (Life Technologies). Second, the resulting construct (pCR-UL54P) was cleaved with PmeI and ligated with a PCR product comprising 83 bp of the human RNase P RNA component H1 (RPPH1) sequence amplified from an infected-cell genomic DNA template with primers 759 and 765. All oligonucleotide sequences are listed in Table S1 in the supplemental material.

Cells, viruses, and infections.

Human fetal diploid lung fibroblasts (MRC-5) (38) were obtained from the European Collection of Cell Cultures, and early-passage cells (15 to 25 population doublings before senescence) were used in all experiments. MRC-5-derived TetR cells and TetR-IE1 cells expressing inducible wild-type IE1 have been described (35). TetR-IE1_1–475 cells expressing inducible CTD-deleted IE1 were generated via lentiviral transduction analogous to that described for TetR-IE1 cells (35). The H1299 human lung carcinoma cell line (39) was obtained from the American Type Culture Collection. MRC-5 and H1299 cells were grown in Dulbecco's modified Eagle's medium (DMEM) (Life Technologies) supplemented with 10% fetal calf serum (Life Technologies), 100 U/ml penicillin, and 100 μg/ml streptomycin. For TetR, TetR-IE1, and TetR-IE1_1–475 cells, the same medium was further supplemented with 1 μg/ml puromycin and 300 μg/ml G418, and induction of IE1 or IE1_1–475 expression was accomplished by adding doxycycline (1 μg/ml) 72 h prior to collection. All cells were maintained under humidified conditions at 37°C and 5% CO₂. Cultures were regularly screened for mycoplasma contamination.

The wild-type and IE1-deficient viruses of the high-passage-number hCMV Towne strain (TNwt and TNdlIE1, respectively) (40) were described previously (21, 35, 41). The wild-type (TBwt) virus of the low-passage-number hCMV TB40E strain (42) was derived from TB40-BAC4 (43) (a gift from Christian Sinzger, Ulm University, Germany). For the construction of TB40E-based IE1 CTD-deficient bacterial artificial chromosomes (BACs) (pTBIE1_1–475) by en passant mutagenesis, E. coli strain GS1783 (44) carrying TB40-BAC4 was transformed with a DpnI-digested and column-purified PCR product generated using plasmid pLAY2 (45) (a gift from Karsten Tischer, Freie University Berlin, Germany) as a template and oligonucleotides 870 and 871 as primers. “Scarless” removal of CTD-specific sequences by homologous recombination was accomplished by following published protocols (44). To control for inadvertent genetic changes, the en passant strategy was also employed to construct revertant BACs (pTBrvIE1_1–475). To this end, a PCR product comprising MIE exon 4 was generated using TB40-BAC4 as a template and oligonucleotides 876 and 877 as primers. This PCR product was inserted into pUC18 via HindIII and EcoRI sites, resulting in plasmid pUC-MIE. In addition, a PCR product comprising a kanamycin resistance (kan) cassette and an SceI cleavage site was generated, using pLAY2 as a template and oligonucleotides 878 and 879 as primers. Following cleavage with NcoI, this PCR product was inserted into the NcoI site of pUC-MIE, resulting in plasmid pUC-MIE-kan_SceI. Then, GS1783 bacteria carrying BAC pTBIE1_1–475 were transformed with a DpnI-digested and column-purified PCR product generated using plasmid pUC-MIE-kan_SceI as a template and oligonucleotides 880 and 881 as primers, and this was followed by homologous recombination (44). The identity and integrity of pTBIE1_1–475 and pTBrvIE1_1–475 were verified in comparison to TB40-BAC4 by restriction fragment length and DNA sequencing analyses (data not shown).

Allelic exchange to generate IE1 CTD-deficient (pFXIE1_1–475) and revertant (pFXrvIE1) BACs of the low-passage-number hCMV FIX strain (46, 47) utilized the following derivatives of transfer plasmid pGS284 (48) (amplified in E. coli strain S17λpir): pGS284-FXIE1kanlacZ, pGS284-FXMIE, and pGS284-FXIE1_1–475. Plasmid pGS284-FXIE1kanlacZ contains the kan and lacZ genes cloned between sequences flanking the IE1-specific exon 4 of the hCMV FIX MIE transcription unit. The ∼1,000-bp flanking sequences were obtained by PCR amplification using primers 136 and 138 (downstream flanking sequence) or 139 and 140 (upstream flanking sequence) and an EGFP-expressing hCMV FIX BAC (pFXwt) (a gift from Dong Yu, Washington University School of Medicine) as the template. The amplified downstream flanking sequence was cloned into pGS284 via BglII and NotI sites. Following addition of adenosine nucleotide overhangs to the PCR product, the upstream flanking sequence was first subcloned into the pCR4-TOPO vector (Life Technologies) and subsequently inserted via NotI sites into pGS284 carrying the downstream flanking sequence. The kan-lacZ expression cassette was released from plasmid pGEM-kanlacZ (YD-C54) (49) and cloned into the PacI site located between the hCMV flanking sequences in the pGS284 derivative described above. For the construction of pGS248-FXMIE, an ∼3,000-bp sequence of the MIE region was amplified by PCR using the template pFXwt and primers 155 and 156. After phosphorylation, the PCR product was first inserted into the SmaI site of pUC18 to generate pUC18-FXMIE. Plasmid pUC18-FXMIE served as a template for QuikChange site-directed mutagenesis with oligonucleotides 280 and 281, replacing IE1 codon 475 with a stop codon to generate pUC18-FXIE1_1-475. Then, FseI-NotI fragments were excised from pUC18-FXMIE and pUC18-FXIE1_1-475 and cloned into the same sites of pGS284-FXIE1kanlacZ, thereby generating pGS284-FXMIE and pGS284-FXIE1_1-475, respectively. For verification, DNA sequence analysis was completed on all cloned PCR amplification products. Allelic exchange was performed through homologous recombination in E. coli strain GS500, as previously described (21, 48, 49). First, the BAC pFXIE1kanlacZ was generated by recombination of pFXwt with pGS284-FXIE1kanlacZ, followed by selection for kanamycin resistance and LacZ expression. After that, the BACs pFXIE1_1-475 and pFXrvIE1 were made through recombination of pFXIE1kanlacZ with pGS284-FXIE1_1-475 and pGS284-FXMIE, respectively, followed by selection for the loss of kanamycin resistance and LacZ expression. The identity and integrity of pFXIE1_1-475 and pFXrvIE1 were verified in comparison to pFXwt by restriction fragment length and DNA sequence analyses (data not shown). All oligonucleotide sequences are listed in Table S1 in the supplemental material.

Cell- and serum-free virus stocks were produced upon electroporation of MRC-5 cells with BAC clones carrying wild-type (TNwt, FXwt, TBwt), revertant (FXrvIE1, TBrvIE1_1-475), or IE1 CTD-deficient (FXIE1_1-475, TBIE1_1-475) genomes. Stocks of TNdlIE1 viruses were produced in a similar fashion, following electroporation of TetR-IE1 cells (35). All virus stocks were screened for mycoplasma contamination. Titers were calculated by qPCR-based absolute quantification of intracellular viral genome copies following infection of MRC-5 cells as described previously (21, 35) and by comparing the results to a standard curve based on amplifications from plasmid pCR-RPPH1-UL54P (see above) mixed with salmon sperm DNA. For wild-type viruses, titers were also determined by standard plaque assay on MRC-5 cells. For mutant and revertant viruses, numbers of PFU relative to wild-type titers were calculated from intracellular viral copy numbers. Infections were carried out at the input multiplicities indicated in the figure legends and for the durations indicated in the legends or figures on (nearly) confluent MRC-5 cells.

Protein production, purification, and analysis.

For protein production in E. coli, a single colony of the Rosetta strain (Novagen) transformed with pGEX-KG or derivatives was grown by shaking (220 rpm) it overnight at 28°C in Luria-Bertani medium containing ampicillin (50 μg/ml), chloramphenicol (20 μg/ml), and 2% glucose. On the next day, the culture was diluted to an optical density at 600 nm (OD₆₀₀) of 0.1 with fresh prewarmed medium lacking glucose and further grown for 2 to 3 h at 220 rpm and 28°C to an OD₆₀₀ of 0.6. At this point, gene expression was induced by adding isopropyl-β-d-thiogalactopyranoside to a final concentration of 200 μM. Following a 6-h incubation at 220 rpm and 28°C, cells were quick-chilled on ice and collected by centrifugation (∼2,500 × g, 15 min, 4°C). Bacteria were resuspended in a 1/10 culture volume of ice-cold wash buffer (50 mM Tris-HCl [pH 7.5], 150 mM NaCl, 2 mM MgCl₂, 1 mM dithiothreitol [DTT], cOmplete EDTA-free protease inhibitor cocktail [Roche]). After another round of centrifugation (∼2,500 × g, 15 min, 4°C), the pellet was resuspended in 1/25 culture volume of ice-cold wash buffer containing 1 mM phenylmethylsulfonyl fluoride (added freshly). The suspension was snap-frozen in liquid nitrogen and stored at −80°C.

To purify GST and GST fusion proteins, bacteria suspended in wash buffer (see above) were thawed in a water bath at room temperature. For the subsequent workflow, ice-cold buffers and solutions were used, and all steps were carried out on ice or at 4°C. Following addition of lysozyme (150 μg/ml) and benzonase (25 U/ml), the suspension was sonicated five times for 1 min using a Branson model 450 sonifier (duty cycle, 80%; output control, 2) to facilitate cell lysis. The lysate was combined with 1/9 volume of wash buffer containing 10% Triton X-100, rotated for 30 min to solubilize proteins, and centrifuged (20,000 × g, 30 min). The affinity matrix was prepared by washing glutathione-Sepharose 4B (GE Healthcare) consecutively in 10 bed volumes of equilibration buffer (50 mM Tris-HCl [pH 7.5], 150 mM NaCl, 2 mM MgCl₂), 10 bed volumes of blocking buffer (equilibration buffer with 2% bovine serum albumin [BSA]) (1 h under rotation), and another 10 volumes of equilibration buffer. After that, 1.25 ml equilibrated matrix per 1,000-ml culture volume and the supernatant from the bacterial lysate were combined and rotated for 2 h. The sample was then applied to a 10-ml Pierce centrifuge column (Thermo Scientific), and the matrix was washed consecutively with 50 bed volumes of low-salt wash buffer (50 mM Tris-HCl [pH 7.5], 150 mM NaCl, 1 mM DTT, 1% Triton X-100, 1 mM EDTA, cOmplete EDTA-free protease inhibitor cocktail), 50 bed volumes of high-salt wash buffer (50 mM Tris-HCl [pH 7.5], 500 mM NaCl, 1 mM DTT, 1% Triton X-100, 1 mM EDTA, cOmplete EDTA-free protease inhibitor cocktail), and another 50 bed volumes of low-salt wash buffer. After the final washing step, 1 bed volume of low-salt wash buffer was added, the column was sealed, and the 50% slurry containing purified proteins was stored in ice until use. To remove the GST tag from GST-IE1 and GST-IE1_1–475, the protein-loaded matrix was washed with 50 bed volumes of cleavage buffer (phosphate-buffered saline [PBS] with 1 mM DTT), the column outlet was sealed, and 1 bed volume of cleavage buffer containing PreScission protease (50 U/ml; GE Healthcare) was added. Protein cleavage was allowed to proceed in the sealed column for 16 h under rotation. After that, the flowthrough containing IE1 or IE1_1–475 proteins was collected. Protein concentrations were calculated by assuming absorption at 280-nm values of 0.396 or 0.409 for 1 g/1,000 ml (0.1%) IE1 or IE1_1–475, respectively.

For preparation of nucleosomes, H1299 cells grown to confluence were scraped on ice and collected in 15-ml tubes by centrifugation (500 × g, 5 min, 4°C). For the subsequent workflow, ice-cold buffers and solutions were used, and all steps were carried out on ice or at 4°C. Cell pellets, each corresponding to one 10-cm plate, were resuspended in 1 ml micrococcal nuclease (MNase) lysis buffer (10 mM Tris-HCl [pH 7.4], 10 mM NaCl, 3 mM MgCl₂, 150 μM spermine, 500 μM spermidine, 0.5% IGEPAL CA-630) and vortexed while another 4 ml MNase lysis buffer was simultaneously added. Following a 5-min incubation, samples were centrifuged (300 × g, 5 min), the supernatant was removed completely, and nuclei were washed in 500 μl MNase reaction buffer (10 mM Tris-HCl [pH 7.4], 15 mM NaCl, 60 mM KCl, 150 μM spermine, 500 μM spermidine). After that, nuclei were carefully resuspended in 100 μl MNase reaction buffer with 1 mM CaCl₂ and prewarmed for exactly 5 min at 30°C. Extranucleosomal DNA was digested by adding 100 U nuclease S7 MNase from Staphylococcus aureus (Roche) diluted in MNase dilution buffer (5 mM Tris-HCl [pH 7.4], 10 μM CaCl₂). Following incubation for 10 min at 30°C, reactions were stopped with 20 mM EDTA and 2 mM ethylene glycol tetraacetic acid. For coimmunoprecipitations, nucleosome preparations (120 μl) were combined with whole-cell extracts prepared as described above.

Histones were purified by acid extraction from H1299 cell nuclei. Cells grown to confluence were scraped on ice and collected in 15-ml tubes by centrifugation (500 × g, 5 min, 4°C). For the subsequent workflow, ice-cold buffers and solutions were used, and all steps were carried out on ice or at 4°C. Cell pellets, each corresponding to one 15-cm plate, were resuspended in 1 ml hypotonic lysis buffer (10 mM Tris-HCl [pH 7.5], 10 mM NaCl, 3 mM MgCl₂, 10 mM DTT, 0.5% IGEPAL CA-630, EDTA-free protease inhibitor cocktail set III). Then, 4 ml hypotonic lysis buffer was added while samples were concomitantly vortexed at medium speed. Following a 5-min incubation, nuclei were collected by centrifugation (500 × g, 5 min) and washed by repeating the consecutive 1-ml and 4-ml resuspension steps in hypotonic lysis buffer twice (without the 5-min incubation in between). Following centrifugation (500 × g, 1 min), the supernatant was removed completely, the pellet was resuspended in 1 ml 200 mM H₂SO₄, and the nuclei were transferred to 1.5-ml tubes. After that, samples were incubated for 16 h on a rotator. Following extraction, samples were centrifuged (20,000 × g, 30 min) to remove insoluble debris, and 900 μl supernatant was transferred to a new 1.5-ml tube. Histones were precipitated by addition of trichloroacetic acid (TCA) to a final concentration of 35% and overnight incubation at −20°C. After centrifugation (20,000 × g, 15 min), the protein pellet was washed by sonication (Branson sonifier model 450; 10 pulses; duty cycle, 80%; output control, 8) in 1 ml acetone with 50 mM HCl and, subsequently, in 1 ml pure acetone at room temperature. After a final centrifugation step (20,000 × g, 15 min), pellets were dried at room temperature. For binding assays, histones were resuspended in an appropriate volume of reaction buffer by sonication (Branson sonifier model 450; 10 pulses; duty cycle, 80%; output control, 8) and repeated pipetting.

For separation of histones into fractions containing either H2A-H2B or H3-H4, the purification protocol was extended by an ion-exchange chromatography step (50) following acid extraction. Briefly, a 2-ml Pierce centrifuge column was filled with 1 ml sulfopropyl-Sepharose fast flow (Sigma-Aldrich) and washed twice with 8.1 ml wash buffer I (50 mM Tris-HCl [pH 8.0], 200 mM NaCl, 2 mM EDTA). Concurrently, histone extracts (prepared as described above) from six 15-cm dishes of H1299 cells were pooled in a 50-ml tube and combined with 2 volumes of 1 M Tris-HCl (pH 8.0). After verification of the pH (pH 7.0 to 8.0), 200 mM NaCl, 1 mM DTT, and 2 mM EDTA were added, and the mixture was applied to the filled column. The column was washed with 8.1 ml wash buffer II (50 mM Tris-HCl [pH 8.0], 500 mM NaCl, 2 mM EDTA [pH 8.0]) and subsequently with the same volume of wash buffer III (50 mM Tris-HCl [pH 8.0], 600 mM NaCl, 2 mM EDTA). For elution of H2A-H2B, 6.6 ml elution buffer I (50 mM Tris-HCl [pH 8.0], 800 mM NaCl, 2 mM EDTA) was added to the column, of which the first 600 μl was discarded and six 1-ml fractions were collected. Following two 8.1-ml wash steps with elution buffer I, H3-H4 was eluted with 2.5 ml elution buffer II (50 mM Tris-HCl [pH 8.0], 2 M NaCl, 2 mM EDTA) and collected in five 500-μl fractions. A subset of histone fractions (determined by polyacrylamide-sodium dodecyl sulfate [SDS] gel electrophoresis and Coomassie brilliant blue staining [see below]) were pooled and subjected to TCA precipitation as described above.

For protein analysis, samples were mixed with 2× loading buffer (100 mM Tris-HCl [pH 6.8], 4% SDS, 20% glycerol, 0.2% bromphenol blue, 200 mM β-mercaptoethanol), denatured for 5 min at 95°C, and separated in SDS-polyacrylamide gels that were either stained with Coomassie brilliant blue (PlusOne Coomassie blue PhastGel R-350; GE Healthcare) according to the manufacturer's instructions or subjected to immunoblotting, including chemiluminescence detection (SuperSignal West Pico or SuperSignal West Femto; Thermo Scientific) as described previously (51). The following antibodies were used for protein detection following blotting: anti-Flag (α-Flag) (M2; Sigma-Aldrich), α-glyceraldehyde-3-phosphate dehydrogenase (α-GAPDH) (ab9485; Abcam), α-HA (3F10; Roche), α-H2A (ab13923; Abcam), α-H2B (ab1790; Abcam), α-H3 (ab1791; Abcam), α-H4 (62-141-13; Upstate), α-IE1 (1B12 [52]; ab30924 [IE1.G10]; Abcam; 6E1; Santa Cruz), α-IE1/IE2 (MAB810R; Merck Millipore), and horseradish peroxidase-coupled anti-mouse (115-036-003; Dianova), anti-rabbit (AP156P; Chemicon), or anti-rat (112-035-003; Dianova) secondary conjugates.

Protein binding and competition analysis.

For coimmunoprecipitations, subconfluent H1299 cells on 10-cm plates were transfected with 10 μg plasmid DNA using a calcium phosphate precipitation technique (53). Approximately 48 h posttransfection, cells were stored at room temperature for 10 min and cross-linked by adding formaldehyde to a final concentration of 1% directly to the medium. Following a 5-min incubation at room temperature, a glycine solution (125 mM final concentration) was added to each dish to terminate cross-linking. After another 5-min incubation at room temperature, the medium was removed, and cells were washed twice with 10 ml ice-cold serum-free DMEM. For the subsequent workflow, ice-cold buffers and solutions were used, and all steps were carried out on ice or at 4°C. Cells were scraped into 2 ml serum-free DMEM with EDTA-free protease inhibitor cocktail set III (Merck Millipore), and each dish was washed three times with 2.5 ml serum-free DMEM for optimal cell recovery. After that, cells were centrifuged (2,500 × g, 10 min), and the medium was removed completely. The cell pellet was resuspended in 1 ml immunoprecipitation (IP) lysis buffer (50 mM Tris-HCl [pH 7.5], 125 mM NaCl, 500 μM DTT, 0.5% IGEPAL CA-630, 1% Triton X-100, 5 mM EDTA, EDTA-free protease inhibitor cocktail set III), and the suspension was incubated for 5 min. After centrifugation (12,000 × g, 1 min), the pellet was washed in IP lysis buffer, recentrifuged (12,000 × g, 1 min), resuspended in IP lysis buffer, and incubated for 10 min. Then, the cell lysate was sonicated three times for 5 min in a Bioruptor UCD-200 (Diagenode; position H, 30-s on-off cycle) to shear the chromatin. After that, insoluble debris was removed by centrifugation (20,000 × g, 30 min), and 900 μl supernatant was transferred to a new 1.5-ml tube. The following matrices were used for the subsequent immunoprecipitation reactions: anti-Flag M2 affinity gel (Sigma-Aldrich), monoclonal anti-HA–agarose (HA-7; Sigma-Aldrich), mouse IgG-agarose (Sigma-Aldrich), or protein A-agarose/salmon sperm DNA (Merck Millipore). Before use, agarose beads were washed three times in 1 ml IP lysis buffer. To preclear the lysate, 100 μl (20-μl bed volume) mouse IgG-agarose or protein A-agarose/salmon sperm DNA was added, and the mixture was rotated for 1 h. Following centrifugation (20,000 × g, 30 min), 50 μl supernatant was removed to serve as the input sample and 850 μl was transferred to a new 1.5-ml tube. For reactions using protein A-agarose/salmon sperm DNA, the lysate was incubated overnight with α-H3 antibodies (ab1791; Abcam) or IgG from rabbit serum (Sigma-Aldrich). Then, 100 μl (20-μl bed volume) of the respective antibody-coupled or protein A-agarose matrix was added. After rotation for 1 h, samples were centrifuged (100 × g, 1 min), and the supernatant was removed completely. The protein-loaded matrix was resuspended in 100 μl DNase buffer (Ambion) and reacted with 2 μl (4 U) DNase I (Ambion) for 15 min at 25°C. After that, 1 ml IP lysis buffer was added, and the matrix was washed five times in the same buffer. Following the final centrifugation step (100 × g, 1 min), the pellet was resuspended in 45 μl 1× loading buffer, and samples were heated (10 min for output and 5 min for input samples). Before electrophoresis, output samples were centrifuged (16,000 × g, 5 min), and only the supernatant was used for immunoblotting.

For GST pulldown assays, ice-cold buffers and solutions were used, and all steps were carried out on ice or at 4°C. For each reaction, 20 μl (bed volume) glutathione-Sepharose 4B (GE Healthcare) loaded with GST or GST fusion proteins was washed twice in 700 μl binding buffer (50 mM Tris-HCl [pH 8.0], 150 mM NaCl, 1 mM DTT, 10% glycerol, 0.5% Triton X-100, EDTA-free protease inhibitor cocktail set III). Acid-extracted histones from one-half of a 15-cm dish of H1299 cells in 300 μl binding buffer were subjected to one 10-min and two 5-min centrifugations (20,000 × g) to remove insoluble debris. The histone solution was subsequently combined with the washed protein-loaded Sepharose matrix, and 10% of the total volume was removed to serve as input sample. The suspension was rotated for 90 min to facilitate binding. After that, the matrix was washed five times in 700 μl binding buffer, resuspended in 100 μl 1× loading buffer, heated, and analyzed.

To analyze competition between IE1 and LANA for nucleosome/histone binding, synthetic peptides encompassing KSHV LANA residues 5 to 22 (LANA-CTD, GMRLRSGRSTGAPLTRGS) or a mutant amino acid sequence deficient for histone binding (LANA-CTD*, GMRAAAGRSTGAPLTRGS) were purchased from Thermo Scientific and dissolved in water. IE1 and IE1_1–475 proteins were derived from the respective GST fusion proteins by cleavage with PreScission protease (GE Healthcare). Acid-extracted histones in binding buffer were preincubated for 1 h with various concentrations of LANA-CTD or LANA-CTD* (1 mM stock solution), IE1 (95 μM stock solution), or IE1_1–475 (83 μM stock solution). After that, the samples were centrifuged (20,000 × g, 15 min), and the supernatant was reacted with the respective GST or GST fusion proteins coupled to glutathione-Sepharose 4B as described above.

For protein quantification, bands were scanned at 72 dots per inch and subjected to densitometry using Scion Image 4.0.2 software (Scion Corporation), which includes the GelPlot2 extension.

Immunofluorescence and microscopy.

Subconfluent H1299 or MRC-5 cells grown on sterile coverslips in 6-well dishes were transfected with 5 μg plasmid DNA using a calcium phosphate precipitation technique (53). Alternatively, MRC-5 cells were mock or hCMV infected. Approximately 48 h posttransfection or -infection, cells were washed three times with PBS containing 0.05% Tween 20 (PBS-T) and fixed with ice-cold methanol for 20 min at −20°C. After three 5-min washes with PBS-T, samples were blocked for 1 h in PBS-T containing 2% bovine serum albumin (BSA) and reacted for 1 h with the respective primary antibodies in a humidity chamber. The primary antibodies used for immunofluorescence were α-EGFP (ab290; Abcam) or α-IE1 (ab30924 [IE1.G10]; Abcam; sc-69834 [6E1]; Santa Cruz). Following three 5-min washes with PBS-T and a 1-h incubation with the appropriate Alexa Fluor 488- and Alexa Fluor 594-conjugated secondary antibodies (Life Technologies) and 0.2 μg/ml 4′,6-diamidino-2-phenylindole (DAPI) (Roche), coverslips were mounted on glass slides using ProLong Gold (Life Technologies). Slides were analyzed using a Keyence BZ 9000 (generation II) or a Leica DMRX epifluorescence microscope equipped with a digital camera system (Retiga; Q-Imaging), and images were acquired and processed using a BZ II analyzer (Keyence) or Image-Pro Plus (version 6.2) (Q-Imaging) and Adobe Photoshop CS4 software. The extent of overlap between pixels in the green and blue channels was quantified by calculating Pearson's correlations from autocontrasted 8-bit images using ImageJ software (National Institutes of Health) and the Colocalization_Finder (version 1.2) plugin (http://rsb.info.nih.gov/ij/plugins/colocalization-finder.html).

Molecular modeling.

Homology modeling was performed using the PERMOL module (54) implemented in the AUREMOL package (55). The 16 carboxy-terminal residues (amino acids 476 to 491) of hCMV (Towne) IE1 were aligned with 14 amino-terminal residues (amino acids 4 to 17) of KSHV LANA using a program based on the Needleman-Wunsch algorithm contained in PERMOL. The alignment was confirmed with other programs, including ClustalW2 (http://www.ebi.ac.uk/tools/msa/clustalw2) and EMBOSS Stretcher (http://www.ebi.ac.uk/tools/psa/emboss_stretcher), respectively. An alternative alignment was produced by just one of all tested programs (LALIGN; http://www.ebi.ac.uk/tools/psa/lalign). Other possible alignments were excluded, as they did not involve the nucleosome binding motif (NBM) (GMRLRSG) of the LANA CTD. As a template for homology modeling, we used the X-ray structure of the LANA_4–17 nucleosome complex (12) (Protein Data Bank [PDB] accession number 1zla). The modeling was restricted to the interaction site of the LANA peptide with histones H2A and H2B; i.e., LANA amino acids 4 to 17, H2A amino acids 14 to 107, and H2B amino acids 30 to 122 could evolve by use of restrained molecular dynamics, while the remaining part of the histone complex was held rigid during the calculations. For the modeling, the LANA peptide was replaced by the IE1 CTD by using either of the two reasonable amino acid sequence alignments. In a second step, we used PERMOL to generate interatomic distance restraints, dihedral angle restraints, and hydrogen bonds from the H2A-H2B/IE1 CTD model template based on the degree of sequence conservation between the LANA and IE1 CTDs. With this set of restraints, 1,000 structures were calculated per each model using the molecular dynamics program Crystallography & NMR System (56, 57). The 20 best structures in terms of total energy were used for explicit water refinement (58). After the water refinement, a bundle of the 10 lowest-energy structures was selected for each of the two models.

RESULTS

IE1 interacts with nucleosomes in a nucleic acid-independent fashion.

We started the work by employing immunoprecipitation-immunoblotting analysis to investigate whether IE1 and other nuclearly localized CMV proteins are physically associated with nucleosomes. For the first set of binding assays, extracts from cells transfected either with an empty vector or with plasmids encoding epitope-tagged hCMV IE1, murine cytomegalovirus (mCMV) IE1 (mIE1), hCMV IE2, and hCMV pp71 were combined with exogenous nucleosomes prepared by micrococcal nuclease (MNase) digestion of human cell nuclei. Readily detectable amounts of core histones from all four classes (H2A, H2B, H3, and H4) were found to coprecipitate with IE1, while much smaller amounts were detected with IE2. Very little, if any, histone binding was observed in the mIE1, pp71, and empty-vector transfections. Likewise, a nonspecific antibody did not precipitate any appreciable amounts of core histones (Fig. 1A). In a second round of experiments, we checked for coprecipitation of endogenous nucleosomes solubilized by sonication with epitope-tagged IE1, mIE1, IE2, and pp71. Under these conditions, specific core histone binding could be demonstrated only for IE1, not for any of the other viral proteins under investigation (see Fig. S1 in the supplemental material). Importantly, IE1 was found to specifically interact with endogenous core histones not only in plasmid-transfected but also in hCMV-infected cells (Fig. 1B).

FIG 1.

hCMV IE1 interacts with human nucleosomes and all four core histones in a nucleic acid-independent fashion. (A) Results of coimmunoprecipitations from plasmid-transfected cells. H1299 cells were transfected with plasmids encoding the indicated HA-tagged viral proteins or with an empty vector (w/o). Cell extracts were combined with nucleosomes derived from MNase-digested human cell nuclei. Samples were subjected to immunoprecipitation using anti-HA or anti-Flag agarose. Input and output protein samples were separated in 10 or 15% polyacrylamide-SDS gels, and HA-tagged proteins and individual core histones were detected by immunoblotting. See also Fig. S1 in the supplemental material. (B) Results of coimmunoprecipitations from hCMV-infected cells. Following infection of MRC-5 cells with TNwt or TNdlIE1 viruses (3 PFU/cell for 72 h), cells were fixed with formaldehyde and cell extracts were sonicated to solubilize nucleosomes. Samples were subjected to immunoprecipitation using rabbit antibodies to histone H3 or nonspecific rabbit IgG (rbIgG). Input and output protein samples were separated in 15% polyacrylamide-SDS gels, and the IE1 protein and histone H3 were detected by immunoblotting. (C) Results of in vitro GST pulldown assays. Equal volumes of empty glutathione-Sepharose beads or beads carrying GST or GST-IE1 were reacted with acid-extracted human histones in the absence or presence of DNase I and RNase A. Input (8% of output) and output protein samples were separated in 15% polyacrylamide-SDS gels and stained with Coomassie brilliant blue.

To confirm the results obtained by immunoprecipitation-immunoblotting in another type of binding assay, we expressed IE1 fused to glutathione S-transferase (GST) in Escherichia coli and affinity purified the protein on glutathione-Sepharose beads. GST-IE1 beads and beads loaded with only GST or no protein (empty beads) were subsequently reacted with acid-extracted histones, and samples were analyzed by electrophoresis in polyacrylamide gels stained with Coomassie brilliant blue. The four core histones were found to interact with GST-IE1 at approximately equimolar ratios, while only minor binding to GST or to empty beads was observed in these assays (Fig. 1C). IE1 is not believed to bind DNA or RNA directly, and acid-extracted histone preparations are not supposed to contain intact nucleic acids. However, to fully rule out the possibility that the observed IE1-histone interactions are mediated through DNA or RNA, the pulldown assays were repeated in the presence of excess amounts of nucleases (DNase I and RNase A). As expected, the results obtained from DNase- and RNase-treated samples were virtually indistinguishable from those obtained in the absence of nucleases (Fig. 1C).

Taken together, the results from our in vivo and in vitro interaction assays demonstrate that IE1 specifically binds to human nucleosome cores, most likely through direct interaction with histones.

IE1 interacts with core histones via two separable regions with distinct binding specificities.

To investigate whether core histones interact with IE1 residues previously shown to be required for chromosome association (22), we constructed plasmids encoding GST fused either to a carboxy-terminally truncated IE1 lacking the CTD (GST-IE1_1–475) or to an amino-terminally truncated IE1 consisting of only the CTD (GST-IE1_476–491). We also generated a plasmid encoding GST fused to the KSHV LANA amino-terminal CTD (GST-LANA_5–22) (59, 60), which was shown to bind to the H2A-H2B dimer of the nucleosome (12, 61) (Fig. 2A). Following expression in E. coli, the GST fusion proteins were used in pulldown assays with acid-extracted histone preparations as described above for wild-type GST-IE1 (Fig. 1C). With GST-IE1_1–475, a reduction in H2A and H2B binding was evident compared to that with the full-length protein, whereas H3 and H4 binding was not negatively affected. Conversely, GST-IE1_476–491 and GST-LANA_5–22 did not specifically interact with H3 and H4. Instead, both GST-IE1_476–491 and GST-LANA_5–22 displayed selective affinity for H2A and H2B. The interaction between H2A-H2B and GST-IE1_476–491 was less efficient than that between H2A-H2B and full-length GST-IE1 but about equally as efficient as GST-IE1_1–475 and GST-LANA_5–22. No interaction between any of the GST proteins and linker histone H1 was detected (Fig. 2B). Notably, CTD-mediated complex formation between IE1 and H2A or H2B was also observed in coimmunoprecipitations from cells arrested in interphase (G₀ phase), indicating that nucleosome targeting by the viral protein is not restricted to mitosis (Fig. S2).

FIG 2.

hCMV IE1 exhibits two separable histone binding domains with differential specificities for H2A-H2B and H3-H4. (A) Schematic of wild-type and mutant hCMV IE1 and KSHV LANA proteins with relative locations of their CTDs. (B) Results of in vitro GST pulldown assays with acid-extracted unfractionated human histones. Empty glutathione-Sepharose beads or beads carrying GST or the indicated GST fusion proteins were reacted with acid-extracted human histones. Input and output protein samples were separated along with purified recombinant human core histones (H2A, H2B, H3, and H4 from New England BioLabs) in a 15% polyacrylamide-SDS gel and stained with Coomassie brilliant blue. The asterisk marks H1 histones. (C) Results of in vitro GST pulldown assays with acid-extracted human histones separated into H2A-H2B and H3-H4 fractions. Equal volumes of empty glutathione-Sepharose beads or beads carrying GST or the indicated GST fusion proteins were reacted with purified human H2A-H2B or H3-H4. Input (8% of output) and output protein samples were separated in 15% polyacrylamide-SDS gels and stained with Coomassie brilliant blue.

To discriminate between direct and indirect core histone interactions, acid-extracted histones were further purified and separated into fractions highly enriched for either H2A-H2B dimers or H3-H4 dimers/tetramers. As predicted from the preceding experiments (Fig. 2B), GST-IE1_1–475 and GST-IE1_476–491 displayed H2A-H2B binding comparable to that of GST-LANA_5–22 but less efficient than that of full-length GST-IE1. In contrast, GST-IE1 and GST-IE1_1–475 were equally efficient in binding to H3-H4. Finally, for GST-IE1_476–491 and GST-LANA_5–22, no H3-H4 interaction above background was observed (Fig. 2C).

The results from these experiments allow for several conclusions: (i) IE1 binds to core histones through at least two physically separable (and therefore independent) interaction surfaces, i.e., the CTD and unspecified sequences located upstream from the carboxy terminus; (ii) the CTD selectively binds to H2A-H2B dimers through direct interaction, while the upstream histone binding domain directly binds to both H2A-H2B and H3-H4 dimers/tetramers with a preference for H3-H4; and (iii) there are striking similarities in nucleosome binding by the IE1 and LANA CTDs.

Alanine scanning mutagenesis identifies a discrete nucleosome binding motif within the IE1 CTD.

To gain further insight into the physical requirements of IE1-nucleosome complex formation, we decided to determine which individual CTD residues contribute to the interaction. For this purpose, we performed alanine scanning mutagenesis by replacing each of the 16 amino acids comprising the IE1 CTD with alanine (except for A489, which was replaced by glycine) (Fig. 3A). All mutant CTDs were expressed as GST fusion proteins in E. coli and used in pulldown assays with acid-extracted histones. The IE1 CTD pulled down all four core histones in these assays, most likely because the experimental conditions (i.e., higher histone concentrations, leading to a larger proportion of octamers than shown in Fig. 2B) allowed not only for direct (H2A-H2B) but also for indirect (H3-H4) interactions. Again, no binding to linker histone H1 was observed. Interestingly, only four CTD residues (H481, M483, T485, and R486) proved to be essential for histone binding. In addition, six amino acids (S479, T480, P482, V484, S487, and K488) turned out to augment the interaction significantly. In contrast, mutation of all residues upstream of S479 (G476, G477, and K478) and downstream of K488 (A489, D490, and Q491) had no obvious adverse effect on CTD-histone complex formation. In fact, the D490A exchange appeared to even enhance histone binding compared to that of the wild-type CTD (Fig. 3B and C).

FIG 3.

Alanine scanning mutagenesis identifies IE1 CTD residues critical for histone binding. (A) Presentation of wild-type (wt) and mutant IE1_476–491 protein sequences. Amino acids substituted by alanine scanning mutagenesis are highlighted. (B) Results of in vitro GST pulldown assays. Glutathione-Sepharose beads carrying GST or the indicated wild-type (wt) and mutant GST-IE1_476–491 fusion proteins (see also panel A) were reacted with acid-extracted histones. Input (14% of output) and output protein samples were separated in 15% polyacrylamide-SDS gels and stained with Coomassie brilliant blue. (C) Quantitative assessment of GST pulldown assay results. The output bands were quantified by densitometry, and bars represent the ratios of histones H3, H2B, H2A, and H4 to GST-IE1_476–491. Results for the CTD mutants are presented relative to those for the CTD wild-type sample present on the same gel (set to 100%).

In order to link the in vitro histone binding results to cellular chromosome association, we also expressed the wild-type and mutant CTD peptides as fusion proteins with amino-terminal EGFP in human cells. Costaining of the EGFP-CTD proteins with mitotic DNA revealed that each of the same four residues shown to be essential for histone binding (H481, M483, T485, and R486) is also indispensable for chromosome attachment (Fig. 4A and B). Alanine substitution of all four essential CTD residues or individual substitution of M483 also abolished mitotic chromatin association of IE1 in the context of the full-length protein (Fig. 4C). Moreover, in agreement with the binding results, individual exchange of S479, T480, P482, V484, S487, and K488 was linked to an intermediate phenotype, while mutation of residues at the edges of the IE1 CTD had no significant negative (G476, G477, K478, A489, and Q491) or even positive (D490) effects on chromosome association (Fig. 4A and B).

FIG 4.

Alanine scanning mutagenesis identifies IE1 CTD residues critical for mitotic chromatin association. (A) Results of colocalization analyses in mitotic cells. H1299 cells were transfected with plasmids encoding EGFP or the indicated wild-type (wt) and mutant EGFP-IE1_476–491 fusion proteins (Fig. 3A). Cells were fixed with methanol, immunostained for EGFP, and counterstained for DNA with DAPI. Representative individual and merged images of EGFP and DAPI signals are presented. (B) Quantitative assessment of colocalization analysis results. The extent of overlap between pixels in the green and blue channels was quantified by calculating Pearson's correlation coefficients (1 = perfect positive correlation, 0 = no correlation, −1 = perfect negative correlation). Bars represent means from at least three randomly selected mitotic cell images with standard deviations. Student's t test was performed to check for statistical significance of differences between the wild-type CTD and the respective mutant (*, P < 0.1; **, P < 0.01). (C) Assessment of essential CTD residues in the background of full-length IE1. H1299 cells were transfected with pCMV.TetO-derived plasmids encoding wild-type IE1, IE1_1–475, or full-length IE1 with alanine substitutions of all four (IE1 NBM) or a single (IE1 M483A) CTD residue determined to be essential in the experiments shown in panels A and B. Cells were fixed with methanol, immunostained for IE1, and counterstained for DNA with DAPI. Representative individual and merged images of IE1 and DAPI signals are presented. (D) Delineation of the nucleosome binding motif (NBM) within the IE1 CTD as deduced from data shown in panels A, B, and C and Fig. 3. Amino acids determined to be essential for or accessory to histone binding and chromatin association are highlighted.

These results identify a 10-amino-acid nucleosome binding motif (NBM) (STHPMVTRSK) within the IE1 CTD, of which amino acids H481, M483, T485, and R486 are individually essential for nucleosome core targeting and chromosome attachment (Fig. 4D).

Histone binding by the IE1 CTD is directed by acidic residues in H2A.

In addition to identifying the viral determinants of IE1-nucleosome complex formation, we set out to probe the histone-specific contributions. Much of the LANA-nucleosome interaction is mediated by negatively charged residues in H2A, composing the acidic pocket (12). Given the similarities in histone binding between the LANA and IE1 CTDs (Fig. 2), each of the nine negatively charged amino acids found in H2A (including acidic pocket residues E56, E61, E64, D90, E91, and E92) (Fig. 5A) was individually replaced with alanine. Additionally, we constructed two triple mutants where H2A residues E61, E64, and D90 or D90, E91, and E92 were simultaneously changed to alanine. Subsequently, human cells were transfected with plasmids encoding epitope-tagged wild-type or mutant H2A, and pulldown assays with acid-extracted histones and GST-IE1_476–491 were conducted (Fig. 5B). Interestingly, only mutations in acidic pocket residues E56, E61, E64, and D90 and the two triple substitutions proved to abolish IE1-H2A binding, while all other mutations (E41A, D72A, E91A, E92A, and E121A) had little, if any, effect on this interaction.

FIG 5.

Site-directed mutagenesis identifies human H2A residues critical for interaction with IE1. (A) H2A residues targeted by mutagenesis and evaluated for contributions to IE1 binding. The complete sequence of human histone H2A.2 (H2A type 1B/E) is shown, with acidic residues in bold letters and amino acids that form the acidic pocket underlined. (B) Results of in vitro GST pulldown assays performed with the IE1 CTD. Empty glutathione-Sepharose beads or beads carrying GST-IE1_476–491 were reacted with acid-extracted human histones from H1299 cells transfected with an empty vector (w/o) or plasmids encoding the indicated Flag-tagged H2A proteins. Input and output protein samples were separated in 15% polyacrylamide-SDS gels and stained with Coomassie brilliant blue. Flag-H2A proteins were detected by immunoblotting. (C) Results of coimmunoprecipitations performed with full-length IE1 protein. H1299 cells were simultaneously transfected with an empty vector (w/o) or plasmids encoding the indicated wild-type (wt) and mutant Flag-H2A proteins and plasmids encoding either HA-tagged full-length IE1 or HA-IE1_1–475. Cells were fixed with formaldehyde, and cell extracts were sonicated to solubilize nucleosomes. Samples were subjected to immunoprecipitation using anti-HA or mouse IgG (mIgG) agarose. Input and output protein samples were separated in 10 or 15% polyacrylamide-SDS gels, and Flag- and HA-tagged proteins were detected by immunoblotting. (D) Quantitative assessment of results. The α-Flag bands shown in panel C were quantified by densitometry, and bars represent the ratios of output to input signal intensities relative to that of the H2A wild-type sample present on the same blot (whose intensity was set to 100%).

To confirm these results in the context of the full-length IE1 protein, we performed additional immunoprecipitation-immunoblotting assays (Fig. 5C and D). Again, the E41A, D72A, E91A, E92A, and E121A substitutions did not diminish (but rather enhanced) IE1 binding relative to that of wild-type H2A. However, we reproducibly found reduced binding between the viral protein and E56A, E61A, E64A, and D90A H2A mutants and the D90A E91A E92A H2A mutant. In addition, the E61A E64A D90A mutant exhibited highly impaired IE1 binding. As expected, neither of the histone mutants was completely defective for binding to the viral protein, most likely due to the second histone binding domain located upstream of the CTD (Fig. 2B and C). Correspondingly, IE1_1–475 retained some affinity for H2A (Fig. 5C and D).

These results demonstrate that several negatively charged H2A residues (E56, E61, E64, and D90) composing the nucleosomal acidic pocket, but not acidic residues outside the pocket, selectively direct the interaction with the IE1 CTD.

The IE1 and LANA CTDs compete for binding to nucleosome cores and chromosomes.

Our results indicate that IE1 targets human chromatin via interaction between its CTD and the acidic pocket formed by H2A-H2B on nucleosome cores (Fig. 2 to 5), which closely resembles the situation described for LANA (12). To further test the idea that the IE1 and LANA CTDs target the same nucleosomal surface, we asked whether the presence of the IE1 CTD is compatible with or competitive to nucleosome binding by the LANA CTD. First, we utilized a competition pulldown assay to address this question. GST-LANA_5–22 was reacted with acid-extracted histone octamers and increasing molar ratios of purified IE1. As observed beforehand for IE1 (Fig. 3B), histone binding by LANA extended to all four core histone species under these conditions. However, IE1 diminished complex formation between the LANA CTD and core histones in a dose-dependent manner (Fig. 6A). Compared to full-length IE1, the CTD-deficient protein (IE1_1–475) was much less efficient in competing with GST-LANA_5–22 for histone binding (Fig. 6B). Very similar results were obtained when a synthetic peptide encompassing the LANA CTD (LANA-CTD) was used to compete with binding between GST-IE1_476–491 and core histones, while a mutant peptide (LANA-CTD*) had no effect (Fig. 6C).

FIG 6.

The IE1 and LANA CTDs compete for binding to human core histones. (A) Full-length IE1 competes with GST-LANA_5–22 for histone binding. (B) IE1 lacking the CTD (IE1_1–475) is less active in competing with GST-LANA_5–22 for histone binding than the full-length protein. (C) A peptide encompassing LANA_5–22 (LANA-CTD), but not a mutant peptide (LANA-CTD*), interferes with histone binding to GST-IE1_476–491. Acid-extracted human histones were combined with solvent or with the indicated soluble IE1 proteins (A, B) or LANA peptides (C) and then subjected to in vitro GST pulldown assays with glutathione-Sepharose beads carrying GST-LANA_5–22 (A, B) or GST-IE1_476–491 (C). Input and output protein samples were separated in 15% polyacrylamide-SDS gels and stained with Coomassie brilliant blue. See also Fig. S3 in the supplemental material.

To gain in vivo support for our assumption that the IE1 and LANA CTDs compete for binding to nucleosomes, we expressed HA-tagged IE1 and LANA_5–22 fused to EGFP in human cells and examined their localization by immunofluorescence microscopy. In mitotic cells, both EGFP-LANA_5–22 and HA-IE1 localized predominantly to condensed chromatin when expressed individually (with HA or EGFP, respectively), conforming to previous observations (see, e.g., references 12, 20, 22, 24, and 59). However, simultaneous expression of the two proteins resulted in partial release of EGFP-LANA_5–22 and almost complete displacement of HA-IE1 from mitotic chromatin. As expected, HA-IE1 lacking the CTD (HA-IE1_1–475) did not attach to chromosomes (Fig. S3). These results strongly suggest that IE1 and LANA share the same binding site on the nucleosome.

The IE1 CTD is predicted to form a β-hairpin recognizing the acidic pocket on the nucleosomal surface.

The notion of a shared binding site for IE1 and LANA on the nucleosome also suggests that key residues responsible for CTD-histone interaction may be conserved between the two viral proteins. Therefore, we generated alignments between the IE1 and LANA CTD sequences. The algorithm implemented in PERMOL (54, 62) and most other tested programs (e.g., ClustalW2, European Bioinformatics Institute) produced an alignment with three identical residues, one conserved exchange, and two semiconserved substitutions between LANA_4–17 and IE1_476–491 (Fig. 7A). However, the same number of identical residues was also found with an alternative alignment (see Fig. S4A in the supplemental material) generated by another program (LALIGN; European Bioinformatics Institute). The two alignments differ mainly in the way the RS sequence (R486 to S487) in the IE1 CTD is positioned relative to either of two RS sequences (R9 to S10 and R12 to S13) in the LANA CTD. Other possible alignments were excluded, as they did not involve the LANA residues known to be required for nucleosome binding (12).

FIG 7.

The IE1 CTD is predicted to adopt a β-hairpin structure that docks with the acidic pocket formed by H2A-H2B on the nucleosomal surface. (A) Amino acid sequence alignment of IE1_476–491 and LANA_4–17 (asterisk, identical residue; colon, conserved substitution; period, semiconserved substitution) used for homology modeling. (B) Bundles of the 10 lowest total energy structures for IE1_476–491 after explicit water refinement (RMSD = 0.038 nm) based on the sequence alignment in panel A and the LANA_4–17-nucleosome crystal structure (PDB accession number 1zla). (C) Predicted intra- and intermolecular hydrogen bonds in the IE1_476–491-nucleosome complex based on the sequence alignment in panel A and the LANA_4–17-nucleosome crystal structure (PDB accession number 1zla). Black arrows symbolize predicted interactions between the IE1 CTD and H2A or H2B residues also observed in the LANA-nucleosome complex, and red arrows represent predicted interactions unique to the IE1-nucleosome complex. Arrow tails define the donor, and arrowheads point at the acceptor of a possible hydrogen bond. (D) Front view of detailed total best energy homology model of the molecular interaction between IE1_476–491 and the H2A-H2B dimer based on the sequence alignment in panel A and the LANA_4–17-nucleosome crystal structure after explicit water refinement. The model is presented as displayed in PyMOL (http://www.pymol.org) (97), with the H2A-H2B dimer in surface representation. Blue color represents positively charged and red color negatively charged side chains. The H2A residues predicted or experimentally determined to be critical for binding to IE1_476–491 (Fig. 5) are indicated. The IE1_476–491 peptide is shown in a ball-and-stick representation. IE1 residues P482 and R486 are marked by arrows for better orientation. (E) Model of the molecules in panel D after a 90° turn in the indicated direction. See also Fig. S4 and Movie S1 in the supplemental material.

Based on the two most likely alignments and the available X-ray structure of the LANA amino-terminal peptide bound to the nucleosome core (PDB accession number 1zla), molecular-dynamics-based homology modeling of the IE1 CTD-histone complex was performed. From the patterns of hydrogen bonds, interatomic distances, and dihedral angles (54, 62), two sets of restraints (one for each alignment) were created and used to calculate 1,000 structures by restrained simulated annealing. The 10 best structures each in terms of lowest total energy were refined in explicit water, resulting in structural bundles (Fig. 7B and see Fig. S4B). The results suggest that the IE1 CTD can adopt a β-hairpin (two antiparallel β-strands connected by a reverse turn) resembling the LANA CTD structure. The structural bundle deduced from the first of the two alignments is well defined, with a backbone root mean square deviation (RMSD) of 0.038 nm (Fig. 7B). In this model, the β-hairpin formed by the IE1 CTD is stabilized by four main-chain intramolecular hydrogen bonds, and there are 10 intermolecular hydrogen bonds between CTD residues and histones H2A-H2B, including amino acids forming the acidic pocket (Fig. 7C). A three-dimensional representation of this model indicates excellent shape and charge complementarity between the IE1 CTD and the acidic pocket formed by H2A-H2B, including H2A residues E56, E61, E64, and E92 (Fig. 7D and E and Movie S1). The model derived from the second alignment is structurally much less well defined, exhibiting a backbone RMSD of 0.200 nm (Fig. S4B). Here, the CTD engages in only three intra- and five intermolecular hydrogen bonds (Fig. S4C).

Although both interaction models largely comply with our in vitro and in vivo data, the in silico results and conclusions from mutagenesis experiments (see Discussion) favor the model shown in Fig. 7.

The IE1 CTD is dispensable for productive replication of clinical hCMV strains.

To address how IE1-nucleosome interaction may impact the course and outcome of infection, BAC clones of the hCMV low-passage-number strains TB40E (TBwt) and FIX (FXwt) were used to construct mutant viruses specifically lacking the IE1 CTD (TBIE1_1–475 and FXIE1_1–475, respectively). We also generated revertant viruses (TBrvIE1_1–475 and FXrvIE1) to control for inadvertent genetic changes. The IE1_1–475 proteins expressed from the mutant genomes were detected at kinetics and steady-state levels comparable to those of full-length IE1 (Fig. 8A to B), and the mutant proteins were confirmed to be inactive for chromosome association (Fig. 8C and D) in hCMV-infected cells.

FIG 8.

Steady-state levels and chromatin association of wild-type and mutant IE1 proteins in hCMV-infected cells. (A) IE1 and IE2 steady-state protein levels in hCMV TB40E infections. MRC-5 cells were mock infected or infected with TBwt, TBrvIE1_1–475, or TBIE1_1–475 at 3 PFU/cell, and viral IE1/IE2 protein levels were monitored over time by separation in 10% polyacrylamide-SDS gels and immunoblotting. Detection of the cellular GAPDH protein served as a loading control. (B) IE1 and IE2 steady-state protein levels in hCMV FIX infections. MRC-5 cells were mock infected or infected with FXwt, FXrvIE1, or FXIE1_1–475 at 3 PFU/cell, and viral IE1/IE2 protein levels were monitored over time by separation in 10% polyacrylamide-SDS gels and immunoblotting. Detection of the cellular GAPDH protein served as a loading control. (C) Mitotic chromatin association of IE1 in hCMV TB40E infections. MRC-5 cells were mock infected or infected with TBwt, TBrvIE1_1–475, or TBIE1_1–475 at 3 PFU/cell for 48 h, and the localization of the viral IE1 protein relative to cellular mitotic chromatin (stained with DAPI) was detected by indirect immunofluorescence using a 1:1 mix of antibodies IE1.G10 and 6E1. (D) Mitotic chromatin association of IE1 in hCMV FIX infections. MRC-5 cells were mock infected or infected with FXwt, FXrvIE1, or FXIE1_1–475 at 3 PFU/cell for 48 h, and the localization of the viral IE1 protein relative to cellular mitotic chromatin (stained with DAPI) was detected by indirect immunofluorescence using antibody IE1.G10.

Following infection of permissive fibroblast cells (MRC-5), two independent clones each of TBIE1_1–475 and FXIE1_1–475 did not exhibit significantly altered replication from that of the corresponding wild-type and revertant strains, either at high or at low input multiplicities, as determined by quantification of extracellular viral DNA and infectious particles (Fig. 9). These findings are consistent with results from a recent study testing a CTD-deleted mutant of the hCMV high-passage-number strain Towne (23). Thus, nucleosome targeting by IE1 appears to be entirely dispensable for normal productive hCMV infection in fibroblasts, suggesting a function during nonproductive infection and/or infection of other cell types (see Discussion).

FIG 9.

The IE1 CTD is not required for efficient productive hCMV infection. (A) Single-step replication analysis of TB40E viruses. MRC-5 cells were infected with TBwt, TBrvIE1_1–475, or TBIE1_1–475 (two independent clones) at a high-input multiplicity (3 PFU/cell), and extracellular viral genomes were monitored over time by qPCR-based relative quantification. Data represent means and standard deviations from duplicate infections each measured twice (the value for TBwt at 0 days postinfection was set to 1). (B) Peak virus titers in high-multiplicity TB40E infections. Extracellular infectious particles from 7 days postinfection of the experiment depicted in panel A were quantified by standard plaque assay on MRC-5 cells. Data represent means and standard deviations from duplicate infections each measured twice. (C) Multistep replication analysis of TB40E viruses. MRC-5 cells were infected with TBwt, TBrvIE1_1–475, or TBIE1_1–475 (two independent clones) at a low input multiplicity (0.03 PFU/cell), and extracellular viral genomes were monitored over time by qPCR-based relative quantification. Data represent means and standard deviations from duplicate infections each measured twice (the value for TBwt at 0 days postinfection was set to 1). (D) Peak virus titers in low-multiplicity TB40E infections. Extracellular infectious particles from 16 days postinfection of the experiment depicted in panel C were quantified by standard plaque assay on MRC-5 cells. Data represent means and standard deviations from duplicate experiments each measured twice. (E) Single-step replication analysis of FIX viruses. MRC-5 cells were infected with FXwt, FXrvIE1, or FXIE1_1–475 (two independent clones) at a high input multiplicity (3 PFU/cell), and extracellular viral genomes were monitored over time by qPCR-based relative quantification. Data represent means and standard deviations from duplicate infections, each measured twice (the value for FXwt at 0 days postinfection was set to 1). (F) Peak virus titers in high-multiplicity FIX infections. Extracellular infectious particles from 6 days postinfection of the experiment depicted in panel E were quantified by standard plaque assay on MRC-5 cells. Data represent means and standard deviations from duplicate infections, each measured twice. (G) Multistep replication analysis of FIX viruses. MRC-5 cells were infected with FXwt, FXrvIE1, or FXIE1_1–475 (two independent clones) at a low-input multiplicity (0.05 PFU/cell), and extracellular viral genomes were monitored over time by qPCR-based relative quantification. Data represent means and standard deviations from duplicate infections, each measured twice (the value for FXwt at 0 days postinfection was set to 1). (H) Peak virus titers in low-multiplicity FIX infections. Extracellular infectious particles from 15 days postinfection of the experiment shown in panel G were quantified by standard plaque assay on MRC-5 cells. Data represent means and standard deviations from duplicate infections, each measured twice.

DISCUSSION

Structural aspects of the IE1-nucleosome interaction.

Nucleosomes are the repeating centerpieces of chromatin (reviewed in references 3 and 4). Within the nucleosome core, the disordered histone tail domains are known to engage in interactions with numerous different proteins, while the folded regions are believed to function primarily in compacting and constraining the DNA. However, the nucleosome surface is highly contoured and differentially charged (1, 2). The most distinctive feature of this surface is the acidic pocket, a negatively charged and concave patch generated by several highly conserved glutamate and aspartate residues contributed by histone H2A and, to a lesser extent, H2B (reviewed in references 63 and 64). In fact, the acidic pocket is the only negatively charged area on the otherwise positive or hydrophobic nucleosomal surface. The pocket serves no apparent role in maintaining the structure of the nucleosome but is specifically recognized by histone H4 tails of adjacent nucleosomes (1, 2) and an increasing number of nonhistone cellular and viral proteins. The first protein, besides histone H4, reported to bind to the acidic pocket was LANA encoded by the gammaherpesvirus KSHV (12). Subsequently, human interleukin 33 (IL-33) (61), Drosophila melanogaster regulator of chromosome condensation 1 (RCC1) (65), human high-mobility-group nucleosomal 2 (HMGN2) protein (66), and Saccharomyces cerevisiae silent information regulator 3 (Sir3) (67) were all shown or predicted to interact with the acidic pocket of the nucleosomal core. Furthermore, Foamy virus group-specific antigen (Gag) and Borna disease virus ribonucleoprotein (RNP) were proposed to bind to host chromosomes via H2A-H2B (68, 69).

The present study identifies the first betaherpesvirus protein targeting the acidic pocket on the nucleosome. Previous work has demonstrated that the hCMV IE1 protein attaches to human mitotic chromosomes through a CTD located at the carboxy terminus (20, 22, 24). However, the mechanism of interaction has not been addressed. Although hCMV encodes several chromatin-associated proteins (e.g., see references 70 and 71; reviewed in references 72 to 76), nucleosome binding appears to be a rather unique feature of IE1. For instance, the hCMV IE2 protein is known to interact with DNA (see, e.g., references 77 to 80), histones H3-H4 (81), and several histone-associated proteins (see, e.g., references 51 and 81 to 83); however, IE2 neither associates with condensed chromatin (21, 25) nor with nucleosomes (this work), at least not to the same extent as IE1. Surprisingly, even mIE1 has little affinity for nucleosomes and does not colocalize with mitotic chromatin (21, 84; this work) despite reportedly binding to DNA and core histones (85, 86). Accordingly, IE1 orthologs of known rodent CMVs (including mCMV and rat CMV strains) do not seem to exhibit functional CTD sequences (data not shown). However, CTD and NBM sequences are highly conserved across IE1 proteins of primate CMVs (Fig. 10) (22). The African green monkey CMV IE1 ortholog has also been shown to associate with metaphase chromosomes (29). At the same time, sequence similarity across the full-length orthologous IE1 proteins is limited (Fig. 10). Thus, any available evidence points to nucleosome targeting being a primate-specific viral adaptation and a distinguishing feature of primate CMV IE1 proteins.

FIG 10.

The IE1 NBM is selectively conserved through primate CMV evolution. (A) Alignment of orthologous IE1 sequences from hCMV (Towne) and all nonhuman primate CMV isolates present in GenBank (National Center for Biotechnology Information) (98), including chimpanzee CMV (panine herpesvirus 2), African green monkey (simian) CMV (cercopithecine herpesvirus 5), rhesus macaque CMV (Macaca mulatta CMV and macacine herpesvirus 3, 68-1 strain), and baboon CMV. The multiple-sequence alignment was generated using ClustalW2 (European Bioinformatics Institute; http://www.ebi.ac.uk/tools/msa/clustalw2) with default settings and rendered using Jalview 2.7 (http://www.jalview.org) (99). The extent of amino acid conservation is visualized by shades of blue. CTD and NBM sequences are marked. (B) Table presenting GenBank accession numbers of orthologous IE1 sequences from primate CMVs and percent amino acid sequence identities of full-length IE1 proteins, IE1 CTDs, and IE1 NBMs relative to the corresponding hCMV IE1 (Towne) sequence based on pairwise alignments (IE1) or the multiple-sequence alignment shown in panel A (CTD, NBM). (C) Sequence logo illustrating extent of amino acid conservation in orthologous IE1 CTD sequences from primate CMVs. The sequence logo was generated using WebLogo 2.8.2 (http://weblogo.berkeley.edu) (100) with the default color code. Amino acids are numbered according to their positions in the hCMV (Towne) IE1 sequence. Numbers corresponding to amino acids forming the hCMV IE1 NBM are printed in bold type, and residues essential for nucleosome and chromosome interaction are underlined.

Our results indicate that the IE1 NBM directly and selectively recognizes the nucleosome core through the H2A-H2B dimer. In fact, IE1 does not detectably interact with linker histone H1, and nucleic acids are not required for nucleosome binding by the viral protein. The latter finding matches the long-standing assumption that hCMV IE1 does not directly bind to DNA (86), although our data do not exclude this possibility. Despite the fact that the CTD is sufficient for H2A-H2B binding, nucleosome core interaction, and chromosome association, histone binding by IE1 is clearly not restricted to the carboxy-terminal domain. Instead, another histone binding region, which functions independently from the CTD/NBM, must exist upstream of amino acid 476. In contrast to the NBM, this region directly binds to all four core histones, with a preference for H3-H4. The fact that all tested H2A mutants retain some affinity for full-length IE1 suggests that the upstream histone binding domain may interact with H2A-H2B residues outside the acidic patch. It is tempting to speculate that negatively charged residues within three proximate acidic stretches (termed acidic domains 1 to 3 [AD1 to AD3]) (21) between amino acids 373 and 475 of IE1 might engage in electrostatic interactions with basic residues of the H3-H4 tetramer and the H2A-H2B dimers. Moreover, the observation that IE1_1–475 does not associate with chromosomes and is not required for nucleosome interaction suggests that the upstream histone binding region may interact with “free” histones rather than nucleosomes.

In most published cases of interactions with the nucleosomal acidic pocket (H4, LANA, IL-33, RCC1, and HMGN2), arginine (and serine)-containing linear motifs within conformationally flexible protein regions contribute to H2A binding. Our work identifies the 10-amino-acid sequence STHPMVTRSK to be this motif in IE1 (NBM). Results from proton nuclear magnetic resonance spectroscopy suggest that the IE1 CTD is highly mobile and that this domain is natively unstructured (data not shown), as has been previously predicted (21). Nonetheless, our results strongly suggest that the IE1 CTD forms a loop connecting two hydrogen-bonded antiparallel β-sheets (β-hairpin), at least when complexed with H2A-H2B (“induced fit”). The predicted IE1 β-hairpin closely resembles the structure previously reported for the complexed LANA CTD (12). Consequently, the modes of interaction with nucleosomes are strikingly similar for IE1 and LANA; this is evident not only from our binding and competition experiments but also from comparisons between our preferred model and the known crystal structure of the IE1 and LANA CTD-nucleosome complexes, respectively. According to the published structure (12), LANA_4–17 interacts with E61, E64, D90, and E92 of H2A (and several residues of H2B). Likewise, our preferred model displays interactions between IE1_476–491 and H2A E61, E64, and E92 (and H2B). In addition, substitution of H2A D90 has a strongly adverse effect on IE1 binding that cannot be directly inferred from the proposed hydrogen bond patterns. However, the imidazole ring of IE1 H481 is expected to be positively charged in the complex and may thus be involved in an electrostatic interaction with the nearby negatively charged side chain of H2A D90. Most likely, the D90A exchange also alters the geometry of the IE1 H481 binding site within the acidic pocket. Interestingly, our model predicts an additional interaction of IE1_476–491 with H2A E56 which has not been described for LANA. The importance of E56 for the IE1-nucleosome interaction is reinforced by our mutational analyses. Moreover, the fact that E56 is not predicted to contribute to H2A-IE1 binding by our alternative model adds further experimental support in favor of the preferred model. Taken together, our structural data indicate that LANA and IE1 target nucleosomes through molecular interactions that are highly similar, albeit not identical.

Potential functions of the IE1-nucleosome interaction.

Our structural analyses strongly suggest that the IE1 CTD has specifically evolved to fit the acidic pocket of the nucleosome and that the CTD/NBM sequences are highly and selectively conserved through primate CMV evolution. These findings clearly point to an important role of nucleosome targeting by IE1 in hCMV infection. However, no such role has been reported so far.

Our prior work has demonstrated that nucleosomes are not confined to cellular chromatin but also form on hCMV nuclear DNA (87–89). We have also recently shown that global nucleosome occupancy and dynamics across hCMV genomes are largely controlled by IE1-dependent mechanisms (88). Thus, IE1 may target nucleosomes for active remodeling of viral and/or cellular chromatin. Notably, interaction between the acidic pocket contributed by H2A-H2B and a positively charged region of the histone H4 tail from an adjacent nucleosome has been implicated in the formation of higher-order chromatin structures (90, 91). The H4 tail has no significant sequence similarity with the IE1 or LANA CTDs (data not shown) and seems to adopt a conformation different from that of the viral peptides. Moreover, the binding sites of LANA (and hence also of IE1) and H4 in the acidic pocket do not overlap (2, 12). Nonetheless, the available structures and models predict that IE1 and LANA may displace the H4 tail from the pocket, raising the possibility that the viral proteins might adversely affect chromatin condensation. Conceivably, global chromatin decondensation may serve as a mechanism through which viruses enhance the permissiveness of the cellular and/or viral genome to transcriptional activation and/or other DNA-based processes. Chodaparambil and colleagues have examined the effect of LANA CTD binding on folding and self-association of nucleosome arrays. Against expectations, they found that LANA stabilizes self-association of nucleosomes and promotes cellular heterochromatin formation (92). However, these findings somewhat conflict with other reports (e.g., reference 93). Likewise, IL-33 appears to regulate chromatin compaction by promoting nucleosome-nucleosome interactions (61). It remains to be determined whether IE1 has any positive or negative effects on higher-order chromatin structure.

The CTD has been shown to be entirely dispensable for complementing the defective replication of an IE1-deficient mutant hCMV (Towne) in human fibroblasts (22). Moreover, mutant viruses of both laboratory-adapted and clinical hCMV strains expressing IE1_1–475 instead of the full-length protein do not display any obvious phenotypic differences from the parental wild-type viruses in these cells (reference 23 and this work). Thus, nucleosome binding by IE1 appears to be irrelevant for productive hCMV infection in fibroblasts. However, IE1-nucleosome interaction may serve an important function in cell types supporting nonproductive (latent) hCMV infections. More than a decade ago, it was first reported that LANA tethers KSHV episomes to host mitotic chromosomes (13, 94), and LANA turned out to be necessary and sufficient for KSHV episome persistence in the absence of other viral genes. By bridging KSHV DNA and host chromosomes, LANA facilitates nuclear retention and segregation of viral episomes to daughter nuclei during viral latency (reviewed in reference 95). Very similar mechanisms are used by EBNA1 of EBV and the early 2 (E2) proteins of human and bovine papillomaviruses, which also tether their respective genomes to cellular chromosomes for efficient maintenance during cell division (reviewed in reference 96). As opposed to other DNA viruses, including both alpha- and gammaherpesviruses, the mechanism of viral genome persistence during latency in any of the betaherpesviruses has remained a mystery. Thus, it is highly tempting to speculate that the IE1-nucleosome interaction described in this work may contribute to hCMV genome tethering and maintenance.