Physical Requirements and Functional Consequences of Complex Formation between the Cytomegalovirus IE1 Protein and Human STAT2^▿,†

Cell culture and virus infections.

All cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 100 U/ml penicillin, and 100 μg/ml streptomycin. Cells were cultured at 37°C in a humidified 5% CO₂ atmosphere. Human fetal diploid lung fibroblasts (MRC-5) 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. The fibrosarcoma cell line 2fTGH, which has been used extensively to study Jak-STAT signaling pathways, was obtained from George Stark (Case Western Reserve University, Cleveland, OH) (54) and maintained in medium containing 250 μg/ml hygromycin. H1299 cells have been described (46). Where applicable, cells were treated with 250 to 1,000 U/ml recombinant IFN-α (PBL Biomedical Laboratories) for various durations.

All infections were carried out on confluent MRC-5 cells with a bacterial artificial chromosome (BAC)-derived wild-type isolate (TNwt; from Hua Zhu, University of Medicine and Dentistry of New Jersey) (40) or with IE1-mutated variants (TNdlIE1 and TNdlIE1AD1-S/P) or a revertant (TNdlIE1rev) of the hCMV Towne strain at input multiplicities of 0.01 to 3 PFU per cell or equivalent infectious genome numbers. Virus was grown on MRC-5 fibroblasts (TNwt and TNdlIE1rev) or IE1-expressing derivatives (TNdlIE1 and TNdlIE1AD1-S/P), and TNwt and TNdlIE1rev titers were determined by standard plaque assay on MRC-5 cells. TNdlIE1 and TNdlIE1AD1-S/P infectious units equivalent to TNwt titers were calculated from relative intracellular genome numbers. To this end, cells were infected with various dilutions of virus stock, and particle adsorption was terminated after 2 h by a 1-min incubation in citrate wash buffer (40 mM citrate [pH 3.0], 10 mM KCl, 135 mM NaCl) followed by two wash steps in medium. DNA was extracted from cells at 8 h postinfection using a DNeasy blood and tissue kit from Qiagen. For viral replication analysis, the same kit was employed to purify DNA from infected culture supernatants. The DNA was used to template real-time PCRs performed in a Roche LightCycler 1.5 employing the hCMV-specific primers 294 and 295 (directed against the UL54 promoter) (Table 1) and a LightCycler FastStart DNA MasterPlus SYBR green I kit (Roche), following the manufacturer's instructions exactly. DNA levels were calculated using the efficiency-corrected relative quantification strategy described in Roche Applied Science Technical Note no. LC 13/2001. The hCMV clinical strains Coz and Par were isolated from blood and urine, respectively, by Stephen Spector (University of California at San Diego). A 5.8-kb fragment covering the major IE transcription units of Coz or Par was amplified by high-fidelity PCR from virion DNA using primers 588 and 589 (Table 1), cloned in vector pCR4-TOPO (Invitrogen), and sequenced (Princeton University Syn/Seq Facility).

Protein production and glutathione S-transferase (GST) pull-down assays.

Plasmid pGEX-IE1 was constructed by excising the 72-kDa hCMV IE1 coding sequence from pcDNA-IE1 (50) via consecutive treatment with HindIII, Klenow polymerase, and EcoRI. The resulting DNA fragment was inserted into the SmaI and EcoRI sites of vector pGEX-KG (23). For the construction of pGEX-IE2, the cDNA encoding the 86-kDa hCMV IE2 protein was derived from pEGFP-IE2 (see below) via BglII/EcoRI digestion and inserted into the BamHI and EcoRI sites of pGEX-KG. The IE1ΔN cDNA lacking nucleotides 4 to 411 from the 5′ end of the full-length hCMV IE1 coding sequence was PCR amplified from plasmid template pEGFP-IE1 (50) using primers 205 and 206 (Table 1). The PCR product was inserted into vector pGEX-KG via BamHI and EcoRI sites to generate pGEX-IE1ΔN. The IE1ΔC sequence, comprising nucleotides 1 to 1038 from the 5′ end of the hCMV IE1 gene, was derived from pEGFP-IE1 by BglII digestion. The large BglII fragment was subsequently inserted into the BamHI site of pGEX-KG to generate pGEX-IE1ΔC. The full-length cDNA of wild-type mIE1 was PCR amplified from pEMBL19-pp89 (a gift from Martin Messerle, Hannover Medical School) (77) using primers 207 and 208 (Table 1). The resulting PCR product was inserted into pGEX-KG via BamHI and EcoRI sites. The identity of each recombinant plasmid and error-free PCR amplification was confirmed by restriction digestion and DNA sequencing (Geneart).

A single colony of the expression strain Escherichia coli M15(pRep4) (Qiagen) transformed with the recombinant pGEX-KG derivatives or empty vector was grown in a shaker overnight at 37°C in Luria-Bertani (LB) medium containing ampicillin (50 μg/ml) and kanamycin (25 μg/ml). The culture was diluted 1:100 in fresh prewarmed medium and further grown for 1 h at 30°C to an optical density at 600 nm of 0.6. After that, gene expression was induced by addition of isopropyl-β-d-thiogalactopyranoside to a final concentration of 0.5 mM. Following a 16-h incubation at 25°C, cells were harvested by centrifugation. After one wash step in ice-cold purification buffer (50 mM Tris-HCl [pH 7.6], 100 mM NaCl, 0.5% Triton X-100) and one freeze-thaw cycle, bacteria were resuspended in 1/50 culture volume of purification buffer with complete mini-protease inhibitor cocktail tablets (Roche). This was followed by sonification (three times for 2 min each; duty cycle, 70%; output control, 5) in a Branson Sonifier 450 and centrifugation for 30 min at 27,000 × g. From the supernatant, GST fusion proteins were affinity purified in a batch procedure using glutathione Sepharose 4B (GE Healthcare) according to the manufacturer's instructions. The purification steps were monitored by sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis followed by Coomassie brilliant blue staining of the protein bands and/or by Western blot analysis.

For binding assays, total extracts from confluent MRC-5 cells were prepared in ice-cold lysis buffer (50 mM Tris-HCl [pH 7.6], 10% glycerol, 100 mM NaCl, 0.5% Triton X-100, complete mini-protease inhibitor cocktail tablets) with short sonification. For each sample, the extract from one 15-cm dish of confluent MRC-5 cells (approximately 1.5 × 10⁷ cells) was mixed with 7.5 μg of full-length GST or GST fusion protein bound to glutathione Sepharose in purification buffer and incubated for 90 min at 4°C with constant rotation. After that, the Sepharose beads were washed four times in GST wash buffer (50 mM Tris-HCl [pH 7.6], 100 mM NaCl, 0.1% Triton X-100), suspended in 2× sample buffer (100 mM Tris-HCl [pH 6.8], 200 mM β-mercaptoethanol, 0.2% bromphenol blue, 20% glycerol, 4% SDS), and heated for 5 min at 95°C. Finally, STAT2 binding was analyzed by SDS-polyacrylamide gel electrophoresis followed by Western blotting.

Immunofluorescence microscopy.

Total RNA purified from MRC-5 fibroblasts infected with hCMV (Towne) was reverse transcribed, and the IE1-specific cDNA was PCR amplified using primers 326 and 327 (Table 1). The resulting PCR product was subjected to EcoRI and BamHI restriction digestion. The IE1-coding DNA fragment was subsequently inserted, in frame with the enhanced green fluorescent protein (EGFP) coding sequence, into the EcoRI and BglII sites of vector pEGFP-C1 (BD Biosciences-Clontech), resulting in plasmid pEGFP-TNIE1. The correct wild-type sequence of the IE1 insert was verified (Geneart). Plasmids pEGFP-IE1ΔN and pEGFP-mIE1 were constructed by subcloning BamHI/EcoRI fragments from pGEX-IE1ΔN and pGEX-mIE1, respectively, into the BglII and EcoRI sites of pEGFP-C1. For pEGFP-ΙΕ1ΔC, the large BglII fragment from pEGFP-IE1 (50) was inserted into the BglII and BamHI sites of pEGFP-C1. The IE2 cDNA from pCGN-IE2 (85) was inserted into vector pcDNA3 via KpnI to generate pcDNA-IE2. From pcDNA-IE2, a HindIII/EcoRI fragment was inserted into the same sites of pEGFP-C1 to generate pEGFP-IE2. A fusion PCR strategy (27) was employed to introduce the internal deletions ΔAD1, ΔS/P, ΔAD1-S/P, Δ387-394, ΔAD2, and ΔAD3 into the IE1 coding sequence. Phusion polymerase (New England Biolabs) and the following primers were used for PCR amplification from template pEGFP-TNIE1: 145, 146 (outer primers used in all PCRs), 386 to 395, 401, and 402 (mutation-specific inner primers) (Table 1). Final PCR products were inserted via HindIII and BamHI sites into vector pEGFP-C1, and error-free amplification of the entire IE1 mutant sequences was confirmed (Geneart). Full-length IE1 and internal deletions ΔAD1, ΔS/P, ΔAD1-S/P, ΔAD2, and ΔAD3 were subsequently transferred to pcDNA-HA-N (a gift from Ron Hay, University of Dundee) (70) via BamHI and EcoRI sites following PCR amplification primed by oligonucleotides 145 and 146 (Table 1) and using the respective pEGFP derivatives as templates. A pSG5-derived plasmid, which encodes a hemagglutinin (HA)-tagged IE1 protein lacking the region spanning AD2 and AD3 (amino acids 421 to 475) (28), was provided by Jin-Hyun Ahn (Sungkyunkwan University, Suwon, South Korea).

Approximately 2 × 10⁵ 2fTGH cells were plated on coverslips in six-well dishes and transfected with 10 μg of plasmid DNA by calcium phosphate precipitation 24 h thereafter. At 48 h posttransfection, cells were fixed with methanol for 15 min at −20°C, and immunofluorescence staining was performed following a previously published protocol (52). An anti-STAT2 rabbit polyclonal antibody (H-190) or an anti-PML mouse monoclonal antibody (5E10; from Roel van Driel, University of Amsterdam) (68) and an anti-EGFP mouse monoclonal antibody (MAB3580) or an anti-IE1 rabbit polyclonal antibody (rbIE1-1) were used for primary protein detection (Table 2). To generate antibody rbIE1-1, GST-IE1ΔN protein was affinity purified on glutathione Sepharose 4B beads (see above) and eluted with reduced glutathione. A New Zealand White rabbit was immunized three times with 200 μg each of GST-IE1ΔN protein (Davids Biotechnology), and the reactivity of the resulting serum, compared to that of a preimmune serum, was tested by Western blotting and immunofluorescence (data not shown). The target epitopes of antibody rbIE1-1 were roughly mapped to the region between amino acids 138 and 404 in the hCMV IE1 protein (data not shown). Primary antibody incubation was followed by a combination of secondary conjugates, each used at a 1:1,000 dilution (2 μg/ml): Alexa Fluor 594 goat anti-mouse immunoglobulin G (IgG) (heavy plus light chain [H+L]) and Alexa Fluor 488 goat anti-rabbit IgG (H+L) or Alexa Fluor 488 goat anti-mouse IgG (H+L) and Alexa Fluor 594 goat anti-rabbit IgG (H+L) (all highly cross-adsorbed and from Invitrogen). 4′,6-Diamidino-2-phenylindole (DAPI) from Roche was added to the secondary antibody solution at a final concentration of 0.2 μg/ml. Coverslips were mounted on glass slides using ProLong Gold (Invitrogen). Slides were analyzed and images were acquired using a Leica DMRX epifluorescence microscope equipped with a digital camera system (Retiga; Q-Imaging). Images were cropped and processed using Image-Pro Plus 6.2 (Q-Imaging), Meta Imaging series 4.5 (Universal Imaging Corporation), and Adobe Photoshop CS (version 8.0) software.

Immunoprecipitation and Western blotting.

Approximately 48 h following transfection of H1299 cells by calcium phosphate precipitation, whole-cell extracts were prepared in lysis buffer (50 mM Tris [pH 8.0], 10% glycerol, 150 mM NaCl, 0.5% Triton X-100, protease inhibitor cocktail set III [Calbiochem]). Preclearing was performed for 1 h with protein A-agarose (Sigma-Aldrich). Following centrifugation (1 min, 1,000 × g, 4°C), extracts were incubated with monoclonal anti-HA-agarose (clone HA-7; Sigma-Aldrich) for 90 min at 4°C. The agarose beads were subsequently washed four times in immunoprecipitation wash buffer (50 mM Tris [pH 8.0], 0.1% Igepal CA-630, 150 mM NaCl). For immunoprecipitations from hCMV-infected cells, Dynabeads Protein A M-280 (Invitrogen) were used, following the manufacturer's instructions exactly. After the last wash step, beads were resuspended in 2× sample buffer and heated for 5 to 8 min at 95°C. Polyacrylamide-SDS gel electrophoresis and Western blotting were performed as described previously (50). For a list of primary antibodies, see Table 2.

Mutagenesis of hCMV BACs.

An IE1 cDNA lacking the exon 4 nucleotides corresponding to amino acids 373 to 420 (AD1 and S/P) was generated using a fusion PCR strategy (27) on template pGS284-MIE (E.-M. Hauer and C. Paulus, unpublished data) with primer pairs 395/140, 394/422, and 140/422. The final PCR product was inserted into vector pCR4-TOPO to generate plasmid pCR4-MIEdlIE1AD1-S/P. Subsequently, a 3.2-kb NheI/SphI fragment from pCR4-MIEdlIE1AD1-S/P covering the mutant IE1 sequence was transferred to pGS284, a derivative of the positive-selection suicide vector pCVD442 carrying ampicillin resistance and sacB genes (66). The resulting transfer plasmid, pGS284-MIEdlIE1AD1-S/P, was used for homologous recombination with the hCMV BAC pTNIE1kanlacZ (E.-M. Hauer and C. Paulus, unpublished data), in which the major IE exon 4 is replaced by a kanamycin resistance (kan) and lacZ cassette (see Fig. S3A in the supplemental material). The BAC also carries a chloramphenicol resistance gene. For allelic exchange via conjugation, the recA⁺ E. coli strain GS500 (66) containing pTNIE1kanlacZ and E. coli S17λpir transformed with the pGS284-MIEdlIE1AD1-S/P donor plasmid were cross-streaked on LB plates (without antibiotics) and incubated overnight at 37°C. To select for cointegrates, 10 ml LB broth containing ampicillin and chloramphenicol was inoculated with bacteria from intersection regions, and cells were grown at 37°C overnight. On the next day, 5 μl of the stationary culture was again inoculated in 10 ml LB medium with ampicillin plus chloramphenicol and incubated for another 8 h at 37°C. This was followed by dilution in 10 ml chloramphenicol containing LB broth and incubation for 15 h at 37°C. Finally, a small culture volume was incubated for 48 h at 30°C on LB plates containing 15 μg/ml chloramphenicol, 20 mg/ml 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside, 200 mg/ml isopropyl-β-d-thiogalactopyranoside, and 5% sucrose. This step allowed selection against the sacB-containing donor vector and nonresolved cointegrates and facilitated identification of positive bacterial clones via blue-white screening. White colonies were replica plated on LB plates containing either ampicillin, kanamycin, or chloramphenicol and incubated at 37°C. Clones resistant to chloramphenicol but not ampicillin or kanamycin were selected, and BAC DNA purified from these cells was further examined by analytical PCR (see Fig. S3B in the supplemental material), restriction enzyme digestion analysis (see Fig. S3B in the supplemental material), and DNA sequencing (Geneart). From two verified BAC clones, pTNdlIE1AD1-S/P-1 and pTNdlIE1AD1-S/P-2, infectious virus was reconstituted by transfection of MRC-5 fibroblasts as described previously (83). The pTNdlIE1-1, pTNdlIE1-2, pTNdlIE1rev-1, and pTNdlIE1rev-2 hCMV BACs were generated in an analogous fashion (see Fig. S3A in the supplemental material), and their construction and reconstitution to infectious virus will be described in detail elsewhere (C. Paulus, T. Knoblach, C. Winterling, and M. Nevels, unpublished data).

RNA interference.

Short interfering RNA (siRNA) duplexes directed against human STAT2 transcripts (sSTAT2-1, r[AAGAUUCUGCAGCAUUUCCCACUCC]; sSTAT2-2, r[AUAUCCGAAGCAGAAGACAUCCUGC]) were purchased from Invitrogen (12937-64). An EGFP-targeted negative control siRNA without related sequences in the human genome (sEGFP, r[GCAAGCUGACCCUGAAGUUC]dTdT) was provided by IBA. MRC-5 cells on 12-well dishes were transfected with 30 nM of individual siRNAs using the reagent Lipofectamine RNAiMAX (Invitrogen) according to the manufacturer's instructions. At 48 h following transfection, cells were subjected to hCMV infection. Quantitations of viral DNA and protein analyses were performed 4 days postinfection.

Nucleotide sequence accession numbers.

The IE1 nucleotide and amino acid sequences from hCMV strains Coz (GQ244522) and Par (GQ244523) have been deposited in the GenBank database (National Center for Biotechnology Information).

Residues in the carboxy-terminal third of the IE1 protein are required for STAT2 interaction.

Our previous work has shown that complex formation between hCMV IE1 and STAT2 can be reproduced by combining a purified GST-IE1 fusion protein expressed in E. coli with total human cell extracts in an in vitro capture assay (53). To narrow down the STAT2 interaction region(s) in the viral polypeptide, we purified bacterially expressed wild-type GST-IE1 and truncated derivatives lacking large parts (137 or 145 amino acids) from the amino (IE1ΔN) and carboxy (IE1ΔC) termini of the full-length protein, respectively. We also prepared unfused GST as well as GST fusions of the hCMV 86-kDa IE2 protein, sharing 85 amino-terminal residues with IE1, and of the mCMV IE1 ortholog (mIE1) (Fig. 1A). The mIE1 gene product is functionally related to hCMV IE1, but the two proteins exhibit only very limited primary sequence similarity (7, 31).

Sufficient amounts of GST and GST-IE fusion proteins were expressed and purified (Fig. 1B), and the identity of each recombinant protein was confirmed by Western blotting (data not shown). Subsequently, glutathione Sepharose beads carrying comparable amounts of full-length proteins were reacted with equal volumes of cell extract prepared from MRC-5 cells (Fig. 1C). Little or no STAT2 was captured when only GST or binding buffer was used for the assays. As expected, wild-type GST-IE1 pulled down readily detectable amounts of STAT2. Similar results were obtained for the GST-IE1ΔN protein indicating that this mutant displays wild-type characteristics with respect to STAT2 binding. For GST-IE1ΔC and GST-IE2, no physical association with STAT2 was detectable, implying that these proteins lack the pertinent interacting domain(s). Surprisingly, however, robust complex formation between mIE1 and human STAT2 was reproducibly observed (Fig. 1C).

To verify the results from the in vitro binding assays in intact cells, we performed transfection experiments with plasmids expressing EGFP-tagged wild-type IE1, IE1ΔN, IE1ΔC, IE2, and mIE1 proteins. Our prior work has demonstrated that binding of IE1 to STAT2 correlates with specific intranuclear colocalization patterns (53). Thus, the nuclear distribution of the individual viral IE proteins in spatial relation to endogenous STAT2 was examined by double-labeling indirect immunofluorescence microscopy (Fig. 1D). In agreement with previous observations (53), the majority (≥90%) of interphase cells displayed a rather diffuse nuclear staining (excluding nucleoli) for both IE1 and STAT2. However, intensive colocalization of STAT2 with EGFP-IE1 but not EGFP alone at punctate nuclear structures was observed in a subset (≤10%) of cells. The fact that most of the IE1/STAT2-containing dots stained positive for the PML protein (data not shown) identified these structures as ND10, which is a well-established target of IE1 (reviewed in references 42 and 73). In our experimental setting, STAT2 was recruited to ND10 only in the presence of the hCMV protein, not in IE1-negative cells. Likewise, STAT2 was relocalized to nuclear dots by mIE1, although costaining with the cellular protein was less pronounced than with hCMV IE1. Similar to IE1 and mIE1, the IE1ΔN, IE1ΔC, and IE2 proteins also localized to the nucleus, although nuclear targeting of IE1ΔN was less efficient than that of the wild-type, most likely due to one missing nuclear localization signal (amino acids 1 to 24) (Fig. 1A) (35, 78). Moreover, no clear dot pattern was observed for IE1ΔN, IE1ΔC, and IE2, reflecting the lack of critical residues for efficient ND10 targeting (2, 29, 35, 36, 49, 78). Notably, occasional small IE1ΔC- and IE2-positive nuclear dots did not costain for STAT2. In mitotic (metaphase) cells, wild-type IE1 and, to a lesser extent, IE1ΔN were found to be enriched at condensed chromatin (Fig. 1D). In contrast, mIE1, IE1ΔC, and IE2 failed to localize to mitotic chromatin, consistent with the fact that these proteins do not contain the previously defined IE1 chromatin-tethering domain (amino acids 476 to 491) (Fig. 1A) (58, 78). Importantly, STAT2 displayed exclusion from metaphase chromatin in nontransfected cells but was recruited to mitotic chromosomes by IE1 and IE1ΔN, in accordance with the results from our binding experiments (Fig. 1C).

These results indicate that residues in the carboxy-terminal 145 amino acids of the hCMV IE1 protein are required for complex formation with STAT2. In contrast, the amino-terminal 137 amino acids of IE1, including the region shared with IE2, do not contribute significantly to this interaction. Interestingly, STAT2 binding seems to be evolutionarily conserved between the hCMV and mCMV IE1 orthologs.

The STAT2 binding region of IE1 is characterized by evolutionarily conserved LC motifs and a predicted disordered structure.

As yet, no experimental data on the three-dimensional structures of the CMV major IE proteins are available. To explore the structural architecture of the STAT2-interacting carboxy-terminal 150 residues of the IE1 protein in silico, we used the Simple Modular Architecture Research Tool (SMART) (37, 63) and tools available on the PredictProtein server (http://www.predictprotein.org/) (59). Based on the SEG algorithm developed by Wootton and Federhen (80), both SMART and PredictProtein identified four low-complexity (LC) regions of strong compositional bias in the hCMV IE1 (Towne) sequence (Fig. 2A): (i) a 14-residue (positions 373 to 386) acidic region referred to here as AD1, (ii) a 15-amino-acid serine- and proline-rich region between positions 395 and 409 (S/P), (iii) a 25-residue (positions 421 to 445) acidic region (AD2), and (iv) another 25-residue acidic element between amino acids 451 and 475 (AD3). AD2 and AD3 together have been designated the acidic domain of IE1 (e.g., see references 28 and 58). Interestingly, the SEG program predicted the same LC domain architecture in the carboxy-terminal parts of IE1 proteins from a variety of different laboratory-adapted and clinical hCMV strains, including 13 sequences available through GenBank (National Center for Biotechnology Information) and two sequences from virus isolates (Coz and Par) first described in this work (Fig. 2A). In some cases, AD1, S/P, and the amino acids between these two regions were recognized as one continuous LC domain (hCMV type II versus type I) (see Fig. S1A in the supplemental material). Furthermore, the number (two to four), approximate lengths, and relative positions of carboxy-terminal LC motifs were remarkably highly conserved between hCMV IE1 and the respective orthologs of primate and nonprimate CMVs (see Fig. S1A in the supplemental material). On the other hand, such motifs were rarely present in protein regions outside the carboxy-terminal domains, and the few LC sequences identified there were not positionally conserved between the orthologs. Intriguingly, even the carboxy-terminal regions of rat CMV IE1 and mIE1 were specifically enriched in LC motifs (see Fig. S1A in the supplemental material).

Compositionally biased amino acid sequences are regarded as hallmarks of intrinsically unstructured or disordered protein domains which have no single well-defined tertiary structure in their native, functional state (reviewed in reference 15). Frequently, disordered domains function via binding to a structured partner, thereby undergoing a disorder-to-order transition. Several in silico predictors of protein disorder have been developed (reviewed in reference 14), including IUPred (http://iupred.enzim.hu) (13). IUPred recognizes disordered regions from the primary sequence based on the assumption that globular proteins are composed of amino acids which have the potential to form a large number of favorable interactions; in contrast, intrinsically unstructured proteins adopt no stable structure, because their amino acid composition does not allow sufficient favorable interactions to form. Using IUPred, we identified disordered structures at the amino- and carboxy-terminal ends of the hCMV IE1 protein (amino acids 1 to ∼50 and ∼380 to 491, respectively) with very high confidence, while the central region, between amino acids 50 and 378, was predicted to constitute a structured globular domain (Fig. 2B). Very similar results were obtained with other computational tools for structural disorder prediction, including DisEMBL (http://dis.embl.de) (38) (data not shown). Strikingly, the location of the putative disordered region at the IE1 carboxy terminus and its peak scores almost exactly overlap the AD1, S/P, AD2, and AD3 LC motifs, suggesting a structural relation. Furthermore, the strong tendency for structurally disordered carboxy-terminal domains is highly conserved among IE1 proteins of all human and animal CMVs examined, including chimpanzee, rhesus macaque, African green monkey, rat, and mouse virus strains. In fact, extremely high disorder tendency scores were determined for the extended carboxy-terminal regions of mIE1 and rat CMV IE1 proteins (see Fig. S1B in the supplemental material). By comparison, hCMV IE2 is predicted to be rather globally disordered with a relatively structured carboxy-terminal region (Fig. 2B).

Thus, despite limited primary sequence similarity, IE1 proteins of CMVs from diverse mammalian species share the presence of LC motifs and a putatively unfolded carboxy-terminal domain. The predicted structural disorder of this domain has likely evolved to enable binding to multiple host cell proteins, including STAT2.

The LC elements in the IE1 protein constitute core and accessory interaction sites for STAT2.

To examine the relative contributions of each carboxy-terminal LC element to STAT2 binding, we constructed and cloned IE1 mutant cDNAs with individual deletions of the AD1, AD2, AD3, and S/P sequences (Fig. 3A). An AD1-S/P double deletion was also generated. Wild-type and mutant IE1 cDNAs were transiently expressed in H1299 cells, and coimmunoprecipitation assays were performed (Fig. 3B). The results show that the ΔAD1 and ΔS/P mutants were severely (≥80% and ≥60%, respectively) compromised for binding to STAT2 (Fig. 3B and C). Combined mutation of AD1 and S/P (ΔAD1-S/P) resulted in less than 3% of wild-type binding activity (Fig. 3C). In contrast, loss of AD2 or AD3 affected IE1-STAT2 complex formation only moderately (by <40%). In fact, the ΔAD3 mutant exhibited a measurable binding defect only when the amounts of coprecipitated STAT2 were normalized to the slightly varying input IE1 protein levels (Fig. 3C; note that this normalization is relevant only if the overall amounts of IE1 are limiting for complex formation with STAT2, which may not be the case given the high abundance of the overexpressed viral protein).

In addition to the binding assays, double-labeling fluorescence microscopy was performed to study the spatial distribution of the mutant IE1 proteins relative to endogenous STAT2 (Fig. 4). Consistent with the binding data, the ΔAD2 and ΔAD3 mutants displayed characteristics similar to those of the wild-type regarding colocalization with STAT2 at ND10 and mitotic chromatin, although fewer ND10s stained double positive for ΔAD2 and STAT2 than for ΔAD3/STAT2 and wild-type IE1/STAT2 (Fig. 4B). Surprisingly, deletion of either AD1 or S/P did not significantly affect colocalization with STAT2 at ND10. However, both mutations individually diminished sequestration of STAT2 at mitotic chromatin. As in the coimmunoprecipitations, the most severe phenotype was observed for the ΔAD1-S/P protein, which was entirely inactive for STAT2 recruitment to either ND10 or chromatin despite the fact that this mutant localized efficiently to both nuclear compartments (Fig. 4A). Deletion of the short sequence (amino acids 387 to 394) between the AD1 and S/P elements did not detectably affect IE1-STAT2 subnuclear codistribution (data not shown), verifying that the two LC elements are the critical determinants in this interaction.

Curiously, Huh et al. recently identified a region exactly comprising the AD2 and AD3 motifs (amino acids 421 to 475) as being critical for IE1-STAT2 physical interaction (28). To compare the relative contributions of AD2-AD3 and AD1-S/P to STAT2 interaction in our binding and colocalization assays, we transfected H1299 cells with plasmids expressing full-length IE1 or one of the two deletion mutants (see Fig. S2 in the supplemental material). The coimmunoprecipitations confirmed that the ΔAD1-S/P IE1 protein is severely defective for STAT2 binding, whereas the Δ421-475 mutant exhibited binding characteristics resembling wild-type in our hands (see Fig. S2B in the supplemental material). Moreover, in contrast to IE1 lacking AD1-S/P, the Δ421-475 protein colocalized with STAT2 at condensed chromatin or ND10 in most mitotic or interphase cells, respectively (see Fig. S2C in the supplemental material). After all, quantitative image analysis (analogous to Fig. 4B) revealed a ∼20% reduction in IE1-STAT2 colocalization efficiency at ND10 associated with the Δ421-475 mutation compared to the wild-type (data not shown).

These results demonstrate that mutation of individual carboxy-terminal LC motifs, including combined deletion of AD1 and S/P, does not significantly affect protein stability, nuclear import, or subnuclear targeting of IE1. However, the presence of AD1 and S/P elements is critical for IE1-STAT2 interaction. The two proximal LC motifs cooperate to form a core interface for STAT2 binding, while the adjacent AD2 element appears to have an ancillary function in this interaction (Fig. 5). The fact that colocalization with STAT2 at chromosomes and at ND10 does not cosegregate in each IE1 LC mutant might indicate that the viral protein forms distinct STAT2-containing complexes in the chromatin and interchromatin compartments of the cell nucleus.

STAT2 binding and ND10 disruption are genetically separable activities of IE1.

The experiments whose results are shown in Fig. 4 had already indicated that none of the four conserved LC motifs in the IE1 carboxy-terminal domain is required for condensed chromatin association or ND10 targeting by the viral protein. In addition, we examined all our IE1 internal deletion mutants by immunofluorescence microscopy following transient transfection for their capacity to disrupt ND10 integrity (Fig. 6A). As expected, the wild-type viral protein was able to efficiently trigger relocalization of the ND10 principal marker protein PML from a punctate to a rather homogenous nuclear distribution in most transfected cells, while the dot structures remained intact in all IE1-negative interphase cells. The effect of each mutant tested (ΔAD1, ΔS/P, ΔAD2, ΔAD3, and ΔAD1-S/P) on ND10 integrity was indiscernible from that of the full-length IE1 protein. In fact, even the ΔAD1-S/P mutant, which proved to be completely defective for STAT2 interaction (Fig. 3 and 4), disrupted ND10 with wild-type characteristics (Fig. 6A). Therefore, we conclude that the failure of some IE1 variant proteins to bind to and colocalize with STAT2 efficiently does not result from global structural perturbations induced by the mutations but rather specifically reflects the absence of critical STAT2 binding motifs. Moreover, these results demonstrate that STAT2 interaction and ND10 disruption are genetically separable, distinct activities of the hCMV IE1 protein.

To confirm these results in the context of an authentic hCMV infection, we constructed a mutant virus that specifically lacks the STAT2 core binding site (AD1-S/P) of IE1 using a BAC-based allelic exchange strategy (see Fig. S3A in the supplemental material). IE1 expressed from the resulting mutant virus, TNdlIE1AD1-S/P, was compared with results obtained with matching wild-type and IE1-null viruses for STAT2 binding following infection of human fibroblasts (Fig. 6B). In concordance with our transfection assays (Fig. 3), the ΔAD1-S/P protein, although stably expressed from the viral genome, completely failed to bind to endogenous human STAT2 under conditions that allowed binding to full-length IE1 (Fig. 6B). Furthermore, neither the efficiency nor the temporal kinetics of ND10 disruption was detectably different between the wild-type virus and TNdlIE1AD1-S/P, whereas ND10 remained intact after infection with TNdlIE1 (Fig. 6C and data not shown). These results strongly support our view that binding to STAT2 and interaction with ND10 are two separable and, most likely, unrelated activities of the hCMV IE1 protein.

STAT2 interaction-dependent and -independent activities of IE1 contribute to type I IFN resistance and efficient replication of hCMV.

Our prior work has shown that an IE1-null mutant hCMV (CR208) (22) lacking all of major IE exon 4 is hypersensitive to exogenous human IFN-α compared to the corresponding wild-type virus (53). Moreover, neutralization of endogenously produced IFN-β partially complemented the low-multiplicity-dependent growth phenotype of the IE1-deficient virus. These findings suggest that IE1 promotes hCMV replication at low input multiplicities, at least in part, by antagonizing the antiviral type I IFN response (53).

In order to investigate to what extent hCMV IFN resistance depends on the IE1-STAT2 interaction, we compared the replication efficiencies of TNwt, TNdlIE1rev, TNdlIE1AD1-S/P, and TNdlIE1 viruses in the presence and absence of exogenously added IFN-α. Multistep growth analyses after low-multiplicity infection revealed that, as expected, TNdlIE1 strains were severely attenuated in terms of replication kinetics and peak titers compared to the parental virus (Fig. 7A). TNdlIE1AD1-S/P viruses displayed an intermediate phenotype between revertant and TNdlIE1 viruses in these analyses. The differentially attenuated phenotypes of ΔAD1-S/P and IE1-null mutants likely result, at least in part, from varying sensitivities to IFN-β produced from the infected fibroblasts. In support of this view, the TNdlIE1AD1-S/P mutant also exhibited an intermediate phenotype between revertant and IE1-deficient viruses when exogenous IFN-α was added to high-multiplicity infections (Fig. 7B; also, see Fig. S4 in the supplemental material). To obtain further support for these findings, we performed a series of Western blotting experiments (Fig. 7C). In the absence of exogenous IFN-α, IE1 and IE2 steady-state protein levels were comparable between wild-type and TNdlIE1AD1-S/P viruses, and slightly less IE2 was found in the TNdlIE1 infections. However, TNdlIE1 produced markedly reduced levels of early (ppUL44) and late (pp28) viral proteins compared to TNwt, confirming the fact that IE1 activates early gene expression (20, 22). In comparison, accumulation of pUL44 and pp28 was much less severely affected in the TNdlIE1AD1-S/P mutant. When TNwt-infected cells were treated with IFN-α, ppUL44 and pp28 but not IE2 accumulated to significantly lower levels than in nontreated infections, reflecting partial sensitivity of hCMV early gene expression to exogenous type I IFN. However, protein (including IE2) production from TNdlIE1 was hardly detectable, indicating that without IE1, hCMV is not able to initiate replication in the presence of large amounts of IFN-α. Again, TNdlIE1AD1-S/P showed intermediate characteristics between the wild-type and IE1-null phenotypes in this experiment (Fig. 7C).

To further substantiate the contribution of STAT2 interaction to the attenuated phenotype of the TNdlIE1AD1-S/P virus, we monitored viral replication following siRNA-mediated knockdown of STAT2 gene expression (Fig. 8). Two siRNAs designed to target the human STAT2 mRNA (sSTAT2-1 and sSTAT2-2) and a negative control siRNA directed against EGFP (sEGFP) were tested for their effects on STAT2 protein levels (Fig. 8A). GAPDH was used as a control protein in these assays. As expected, the EGFP-specific siRNA had no detectable effect on STAT2 or GAPDH levels. Of the two STAT2-directed siRNAs, only one duplex (sSTAT2-2) caused considerable (>80%) STAT2 knockdown, whereas the other one (sSTAT2-1) had little if any effect (Fig. 8A). Consistently, transfection of sSTAT2-2 but not sSTAT2-1 or sEGFP restored the ∼10-fold defect in viral titers associated with the ΔAD1-S/P mutation to wild-type levels (Fig. 8B and C). At the same time, sSTAT2-2 partially rescued defective replication of the TNdlIE1 mutant, and neither one of the two STAT2 siRNAs had any obvious effect on TNwt and TNdlIE1rev virus titers (Fig. 8B and C). Interestingly, STAT2 knockdown by sSTAT2-2 transfection also more than compensated the negative effect of exogenous IFN-α on TNdlIE1AD1-S/P titers (Fig. 8C). This likely indicates that, in the absence of IE1-STAT2 interaction, hCMV replication is hypersensitive not only to exogenous but also to endogenous type I IFN.

Taken together, our results strongly support the view that the relative resistance of hCMV gene expression and replication to type I IFNs crucially depends on the IE1 protein and that interaction between the viral protein and STAT2 via the AD1 and S/P LC motifs contributes significantly to this phenotype. Nonetheless, additional mechanisms independent of STAT2 binding must exist that account for the role of IE1 in counteracting the antiviral IFN response and in promoting viral replication.