Species Specificity of Protein Kinase R Antagonism by Cytomegalovirus TRS1 Genes

Abstract

The host antiviral protein kinase R (PKR) has rapidly evolved during primate evolution, likely in response to challenges posed by many different viral antagonists, such as the TRS1 gene of cytomegaloviruses (CMVs). In turn, viral antagonists have adapted to changes in PKR. As a result of this “arms race,” modern TRS1 alleles in CMVs may function differently in cells derived from alternative species. We have previously shown that human CMV TRS1 (HuTRS1) blocks the PKR pathway and rescues replication of a vaccinia virus mutant lacking its major PKR antagonist in human cells. We now demonstrate that HuTRS1 does not have these activities in Old World monkey cells. Conversely, the rhesus cytomegalovirus homologue of HuTRS1 (RhTRS1) fulfills these functions in African green monkey cells, but not rhesus or human cells. Both TRS1 proteins bind to double-stranded RNA and, in the cell types in which they can rescue VVΔE3L replication, they also bind to PKR and prevent phosphorylation of the α-subunit of eukaryotic initiation factor 2. However, while HuTRS1 binds to inactive human PKR and prevents its autophosphorylation, RhTRS1 binds to phosphorylated African green monkey PKR. These studies reveal that evolutionary adaptations in this critical host defense protein have altered its binding interface in a way that has resulted in a qualitatively altered mechanism of PKR antagonism by viral TRS1 alleles from different CMVs. These results suggest that PKR antagonism is likely one of the factors that contributes to species specificity of cytomegalovirus replication.

INTRODUCTION

Cytomegaloviruses (CMVs) are generally considered species specific in their replication patterns (33). Human CMV (HCMV) replicates well in human cells but not in mouse cells, while murine CMV (MCMV) has the opposite host range. However, between more closely related species, the barriers to replication are incomplete. For example, rhesus CMV (RhCMV) can replicate in human cells as well as rhesus cells (2, 29). Although in some cases modification of a single gene can allow a virus to cross a species barrier (24, 38, 40), the limited host range of CMV replication likely involves multiple viral genes that have adapted to support replication in the specific host over millions of years of coevolution. Understanding the changes that have occurred in both host and viral factors has importance for identifying conserved features of the viral life cycle, for assessing the power and limitations of animal models, and for evaluating the risks and barriers to cross-species transmission of viruses.

Like other viruses, CMVs have needed to adapt to multiple host antiviral defenses, including the inhibition of translation by the protein kinase R (PKR) pathway. PKR is activated by binding to double-stranded RNA, dimerization, and autophosphorylation (12, 37). Activated PKR then phosphorylates the α-subunit of eukaryotic initiation factor 2 (eIF2α), resulting in a block to translational initiation and thus to viral replication. Viruses have evolved multiple different mechanisms for interfering with this host defense pathway, underscoring the importance of PKR as a barrier to viral replication (34). HCMV encodes two double-stranded RNA binding proteins, TRS1 (HuTRS1) and IRS1, either of which is sufficient to prevent activation of the PKR pathway, and at least one of these genes is necessary for HCMV replication in human fibroblasts (9, 19, 20, 31).

Analyses of the rates of nonsynonymous-to-synonymous substitutions (the dN:dS ratio) in the PKR alleles among primates have revealed that PKR has been evolving under strong positive selection, likely as a result of an evolutionary “arms race” with viral antagonists (14, 36). At one branch point in the primate lineage leading toward rhesus macaques and African green monkeys (AGMs), PKR acquired a remarkable 22 nonsynonymous changes but 0 synonymous ones (14). These observations stimulated us to investigate the impact that changes in PKR may have had on the function of antagonists encoded by primate CMVs.

Consistent with the hypothesis that the ability of CMV to antagonize PKR may contribute to the host range of viral replication, we found that HuTRS1 blocks PKR activation in human cells but not in Old World monkey cells. The RhCMV homologue of HuTRS1 (RhTRS1) is able to block the PKR pathway in some Old World monkey cells but not in human cells. RhTRS1 and HuTRS1 both bind to double-stranded RNA (dsRNA) and, in the cell type in which each is functional, they bind to PKR. However, HuTRS1 binds to inactive human PKR and prevents its phosphorylation, while RhTRS1 binds to and inhibits the eIF2α kinase activity of AGM PKR after it has been phosphorylated. These results suggest that evolutionary changes in both PKR and the CMV TRS1 genes resulted in qualitatively different binding interactions and mechanisms of antagonism.

MATERIALS AND METHODS

Cells, virus, and infections.

Human fibroblasts (HF), telomerase-immortalized HF (HF-tert; obtained from Denise Galloway, Fred Hutchinson Cancer Research Center [FHCRC]), primary rhesus fibroblasts (RF; obtained from Klaus Früh and Michael Axthelm, Oregon Health Sciences University), telomerase-immortalized RF (Telo RF; obtained from Peter Barry, University of California, Davis [25]), BSC₄₀, and BHK cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% NuSerum (BD Biosciences) as previously described (9). HF with PKR expression knocked down (PKR-kd) and control knockdown HF (ctrl-kd) were produced by cloning of HF-tert lines after transduction with lentiviral vectors encoding an shRNA targeting human PKR or a nonsilencing control shRNA (catalog numbers RHS4430-98844125 and RHS4346, respectively; Open Biosystems).

Vaccinia virus (VV) Copenhagen strain (VC2) (42) and VVΔE3L (4), both obtained from Bertram Jacobs (Arizona State University), and VC2-lacZ, VVΔE3L+E3L, VV-HuTRS1, VV-RhTRS1, and VC2-GFP (described below) were propagated and titers were determined in BHK cells (9). β-Galactosidase (β-Gal) activity in infected cells was measured by a fluorometric substrate cleavage assay (10). HCMV (strain AD-169; ATCC VR-538) was propagated and titers were determined on HF, and RhCMV (strain 68-1; ATCC VR-677; obtained from Peter Barry, University of California, Davis) was propagated and titers were determined on TeloRF cells.

Cell-free translation and yeast two-hybrid plasmids.

The plasmid pEQ1100, which expresses enhanced green fluorescent protein (EGFP) with a C-terminal biotinylation signal and adjacent 6×His tag (designated -BH) was previously described (10). pEQ1180 (formerly called pEQ981 [19]) expresses full-length HuTRS1-H (-H denotes a 6×His tag). A plasmid containing RhTRS1-BH (pEQ1215) was prepared by PCR amplification of RhCMV DNA (provided by Peter Barry, University of California, Davis) using oligos 723 (5′-CCAAAGATCTACCATGCGTCCTCACCGCTCGCCA-3′) and 724 (5′-GCACGGGACGATGAGAACACCAT-3′). The product was digested with BglII and EcoRV and ligated into the BamHI and EcoRV sites in pEQ1068 (8). A series of RhTRS1 deletion mutants was constructed by using pEQ1215 DNA as the template for PCR amplification with the following oligos and cloning the products into pcDNA3.1/V5-His-TOPO (Invitrogen): for RhTRS1(182–695)-H (pEQ1311), we used oligos 868 (5′-ACCATGAGTCCCTCTCCACAAGAC-3′) and 724; for RhTRS1(347–695)-H (pEQ1312), we used oligos 869 (5′-ACCATGGAATATCTGAGCGAGTGGGCT-3′) and 724; for RhTRS1(1–347)-H (pEQ1313), we used oligos 723 and 870 (5′-TTCCAGCCACGGATGTGTAGT-3′); for RhTRS1(1–548)-H (pEQ1314), we used oligos 723 and 871 (5′-TCTCCCCGGCCGTGTTAAAAA-3′).

Plasmids for yeast two-hybrid assays consisted of kinase-dead PKR alleles fused to the GAL4 transcriptional activation domain (AD) in pGAD424 (Clontech) and TRS1 alleles fused to the GAL4 DNA binding domain (BD) in pGBT9 (Clontech). Kinase-dead human PKR-H was first cloned by PCR amplification of pK296RGFP (obtained from Michael Mathews, University of Medicine and Dentistry of New Jersey [43]) using oligos 705 (5′-ACCATGGCTGGTGATCTTTCAGCA-3′) and 708 (5′-ACATGTGTGTCGTTCATTTTTCTC-3′) followed by TOPO cloning to generate pEQ1198. AD-huPKR-H (pEQ1325) was then prepared by PCR amplification of pEQ1198 by using oligos 806 (5′-CCTTGTCGACTCAATGGTGATGGTGATGATG-3′) and 854 (5′-CCTTGGATCCTTATGGCTGGTGATCTTTCAGCA-3′), digesting with BamHI and SalI, and cloning the into the same sites in pGAD424.

Kinase-dead rhesus PKR was inserted into pGAD424 by the following steps. First, pEQ1260, encoding rhesus PKR(wt)-H, was constructed by PCR amplifying and TOPO cloning of rhesus PKR from pSB819:rhPKR (obtained from Nels Elde, FHCRC) with oligos 772 (5′-ACCATGGCTGGTCATCTTGTACCA-3′) and 773 (5′-ATATGTATGTCGTTTTTTCTCTGGGCT-3′). The HindIII and NotI fragment containing rhesus PKR(wt)-B from pEQ1260 was moved into the same sites in pEQ1068 (8) to generate rhesus PKR(wt)-BH(pEQ1271). A kinase-dead rhesus PKR [RhPKR(K295R)-BH(pEQ1282)] was then constructed using pEQ1271 as a template with oligos 800 (5′-GGAAAGACTTACGTTATTAGACGTGTTAAATATAATAGCAAGAAGG-3′) and 801 (5′-CCTTCTTGCTATTATATTTAACACGTCTAATAACGTAAGTCTTTCC-3′) with the Strat-agene QuikChange mutagenesis kit. Finally, AD-rhPKR-H(pEQ1294) was constructed by PCR amplifying rhesus PKR(K295R)-H from pEQ1282 with oligos 805 (5′-CCTTGGATCCTTATGGCTGGTCATCTTGTACCA-3′) and 806, digesting with BamHI and SalI, and ligating the product into the same sites in pGAD424.

To construct kinase-dead AGM PKR(K295R)-H(pEQ1307), the 5′ and 3′ portions of AGM PKR in pSB819:agmPKR (obtained from Nels Elde, FHCRC) were amplified with oligos 801 and 851 (5′-ACCATGGCTGGTGATCTTGCACCA-3′) and oligos 800 and 852 (5′-ACATGTATGTCGTTCCTTTTTCTC-3′), respectively, to introduce the K295R mutation. The products of these reactions were mixed and amplified with oligos 851 and 852, and the product was TOPO cloned. AGM PKR(K295R)-H was then PCR amplified from pEQ1307 using oligos 806 and 853 (5′-CCTTGGATCCTTATGGCTGGTGATCTTGCACCA-3′), digested with BamHI and SalI, and cloned into the same sites in pGAD424, yielding AD-agmPKR-H(pEQ1328).

Because we found that HuTRS1 was not expressed in Saccharomyces cerevisiae, we synthesized yeast codon-optimized forms of both HCMV and RhCMV TRS1 (in pUC57:huTRS1_y-H and pUC57:RhTRS1_y-H, respectively) (GenScript Inc.). To construct BD-huTRS1_y-H(pEQ1284), the huTRS1 open reading frame (ORF) was first moved from pUC57:huTRS1_y-H into pGBT9 as an EcoRI-PstI fragment. This plasmid was then digested with EcoRI, blunted with Klenow, and religated to place the huTRS1 ORF in the correct reading frame. For BD-RhTRS1_y-H (pEQ1297), yeast codon-corrected RhTRS1 was PCR amplified from pUC57:RhTRS1_y-H using oligos 824 (5′-GATCGAATTCATGAGACCACATAGATCCCCT-3′) and 825 (5′-GATCGTCGACTTAATGATGATGATGATGATGATG-3′), cut with EcoRI and SalI, and cloned into the same sites in pGTB9.

VV recombination plasmids and recombinant viruses.

The backbone from which our recombination vectors were derived was pSC11, which contains a lacZ reporter cassette and flanking sequences for recombination into the VV thymidine kinase (TK) locus (5). To enable measurement of viral replication based on β-Gal activity, we constructed a VC2 virus that expresses β-Gal (VC2+lacZ) by homologous recombination with pSC11. VVΔE3L encodes a late promoter:lacZ cassette in place of the E3L gene; therefore, this virus and the derivatives described below all express β-Gal during productive replication (4).

Recombination plasmid pEQ854 (9) was derived from pSC11 by replacement of the lacZ gene with an EGFP-Puro reporter. EGFP-BH was isolated from pEQ1100 as a BamHI-BclI fragment and inserted into pEQ1131, which is a VV recombination vector derived from pEQ854, by removing the β-Gal reporter gene by digestion with XhoI and ClaI, blunting with Klenow, and religating. The resulting plasmid, pEQ1145, was used to generate VC2-GFP by homologous recombination into VC2. Plasmid pEQ1119, which contains VV E3L-H, was constructed by PCR amplifying E3L from pMT-E3L (obtained from Bertram Jacobs, Arizona State University [6]) using oligos 608 (5′-ACCACCATGGCTAAGATCTATATTGACGAG-3′) and 609 (5′-GAATCTAATGATGACGTAACC-3′), followed by TOPO cloning. To make VVΔE3L+E3L (VVeq1127), a BamHI-BclI fragment from pEQ1119 containing E3L-H was ligated into the BamHI site of pEQ854. The resulting recombination vector (pEQ1127) was then used to construct VVΔE3L+E3L (VVeq1127) by homologous recombination into VVΔE3L.

The VV-HuTRS1 (VVeq1148) recombinant was constructed by replacing the BamHI-BsrG1 fragment containing EGFP in pEQ1145 with TRS1 codons 1 to 760 from pEQ876 (19) followed by recombination into the TK locus of VVΔE3L. To generate VV-RhTRS1, we first made pEQ1135, in which the EFGP-puro cassette was replaced with a neomycin resistance gene, by amplifying pcDNA3.1/V5/His using oligos 624 (5′-GATCCTCGAGACCATGGTTGAACAAGATGGATTGCAC-3′) and 625 (5′-CTAGATCGATACCACAACTAGAATGCAGTG-3′), digesting the product with XhoI and ClaI, and inserting it into the same sites in pEQ854. We then digested pEQ1135 with BamHI and EcoRI, blunted the ends with Klenow, and inserted an RhTRS1-BH cassette derived from pEQ1215 by HindIII digestion, blunting, and PmeI digestion, yielding pEQ1233. VV-RhTRS1 was produced by homologous recombination of pEQ1233 into VVΔE3L to generate VV-RhTRS1 (VVeq1233).

Immunoblot analyses.

Cells were mock infected or infected with CMV or VV (multiplicity of infection [MOI], 3). At various times postinfection, the cells were washed with phosphate-buffered saline (PBS) and then lysed in 2% sodium dodecyl sulfate (SDS). Equivalent amounts of the samples were separated on 10% SDS-polyacrylamide gels, transferred to polyvinylidene difluoride (PVDF) membranes, and probed with one of the following antibodies: Penta-His (Qiagen), PKR (sc-6282 [Santa Cruz Biotechnology, Inc.] or 07-151 [Upstate Biotechnology]), phospho-PKR (T446; 1120-1; Epitomics), actin (A2066; Sigma), TRS1 α999 (31), eIF2α or phospho-eIF2α (Ser51) antibody (both from Cell Signaling Technology, catalog numbers 9722 and 9721, respectively). All purchased antibodies were used according to the manufacturer's recommendations. For all immunoblot analyses, proteins were detected using the Western Star chemiluminescent detection system (Applied Biosystems) according to the manufacturer's recommendations.

Metabolic labeling.

HF, RF, and BSC₄₀ cells were mock infected or infected with the indicated viruses. At 24 h postinfection, the cells were labeled for 4 h with 100 μCi/ml [³⁵S]methionine (Easytag express protein labeling mix; PerkinElmer) in medium lacking methionine. The cells were then washed in PBS and lysed in 2% SDS. Equivalent amounts of protein (50 μg) from each sample were separated on 10% SDS-polyacrylamide gels, dried, and visualized by autoradiography.

Double-stranded RNA binding assays.

dsRNA [poly(rI · rC)]-agarose beads were prepared as previously described (28). The TnT Quick Coupled transcription/translation system (Promega) containing 1 μCi/ml [³⁵S]methionine was used to produce in vitro-translated proteins from the indicated plasmids. Five microliters of lysate in 250 μl of buffer A (100 mM KCl, 20 mM HEPES [pH 7.5], 10% glycerol, 5 mM MgOAc, 1 mM dithiothreitol, 1 mM benzamidine [Sigma] plus 1% NP-40) was incubated with dsRNA-agarose beads (plus carrier Sepharose CL-6B [Sigma-Aldrich]) or with Sepharose CL-6B alone as a binding control for 1 h at 4°C on a rotating mixer. After binding, the beads were pelleted (16,000 × g; 30 s) and washed in buffer A (0.7 ml) 3 to 4 times. The samples were denatured at 95°C for 5 min, separated on 10% SDS-polyacrylamide gels, and analyzed by autoradiography. For analyses of dsRNA binding proteins in cells, lysates were made by washing the cells once in PBS, then lysing them directly on the culture plate in buffer A. The lysates were collected by scraping and incubated on a rotating mixer for 15 min at 4°C, the nuclei were pelleted, and the remaining supernatant was incubated with dsRNA-agarose beads as described above. For competition assays, extracts were preincubated in buffer A with free poly(rI · rC) competitor (200 μg) for 30 min on a rotating mixer prior to addition of dsRNA-agarose beads or control Sepharose CL-6B beads (Sigma-Aldrich). The proteins were separated by SDS-PAGE, transferred to PVDF, and probed with TRS1 α999 antiserum. Lysate corresponding to 3% of the amount used for each binding reaction was analyzed alongside each set of binding reactions.

Yeast two-hybrid assays.

A yeast two-hybrid assay (16) was used to assay PKR-TRS1 interactions. Saccharomyces cerevisiae pJ69-4α (obtained from Nels Elde, FHCRC [23]), a yeast strain in which the interaction of AD and BD fusion proteins results in activation of His3 and β-Gal reporter genes, was transformed with the indicated plasmids by using high-efficiency lithium-acetate transformation (17). Transformants were plated onto YC medium (yeast complete minimal medium with amino acids containing 2% glucose) -Leu -Trp and incubated for 3 days at 30°C. Colonies were then streaked onto YC -Leu -Trp and YC -Leu -Trp -His plates and incubated at 30°C for 3 to 4 days, after which growth of the transformants was analyzed visually. Protein-protein interactions were also quantified by measuring β-Gal expression, essentially as described in the Clontech user manual for β-Gal detection in yeast extracts (PT3037-1). Briefly, 2-ml aliquots from overnight cultures grown in liquid YC -Leu -Trp medium at 30°C were brought up to 5 ml with yeast extract-peptone-dextrose (YEPD) medium and incubated for an additional 6 to 8 h. Approximately 3 ml of each sample was pelleted (16,000 × g; 30 s), washed in 1.5 ml of Z-buffer (60 mM Na₂HPO₄, 40 mM NaH₂PO₄, 10 mM KCl, 1 mM MgSO₄, 50 mM β-mercaptoethanol), and then resuspended in 300 μl of Z-buffer. After freeze-thawing three times, the supernatant was used for β-Gal assays and for determining protein concentrations.

DNA sequences and alignment.

Predicted amino acid sequences for HuTRS1 and RhTRS1 (GenBank accession numbers FJ616285 and AY186194, respectively) were aligned in Geneious v4.8.5 using the Blosum62 cost matrix with a default gap open penalty of 12 and gap extension penalty of 3. The alignment was refined by hand to maximize pairwise identity.

Nucleotide sequence accession numbers.

The sequences of the yeast codon-optimized HuTRS1 and RhTRS1 genes HuTRS1_y-H and RhTRS1_y-H, respectively, have been deposited in GenBank and assigned accession numbers JQ 62772 and JQ 27773.

RESULTS

Species-specific replication of VV-HuTRS1 and VV-RhTRS1.

The observation that RhCMV can replicate in human cells (2, 29) suggests that it encodes an antagonist capable of inhibiting human PKR. HCMV encodes two such proteins, HuTRS1 and IRS1, but a whole-genome analysis of RhCMV identified a homologue of only HuTRS1 (RhTRS1) (21). Overall amino acid identity between HuTRS1 and RhTRS1 was 35%; however, this homology was not evenly distributed, and some regions shared much higher homology, such as the amino-terminal dsRNA binding domain identified in HuTRS1 (∼53%) (Fig. 1). Because of the sequence homology between HuTRS1 and RhTRS1, we hypothesized that RhTRS1 is likely a PKR antagonist. PKR antagonists from heterologous systems can rescue replication of a VV mutant lacking its own major PKR antagonist, E3L (VVΔE3L) (4, 9). Therefore, we constructed VVΔE3L recombinants containing HuTRS1 (VV-HuTRS1) or RhTRS1 (VV-RhTRS1) to study the host range and mechanism of action of these TRS1 alleles.

Fig 1.

Predicted amino acid sequence alignment of HCMV Towne TRS1 and RhCMV 68-1 TRS1. Black vertical bars indicate amino acid identity, gray vertical bars indicate amino acid differences, and horizontal regions indicate gaps. The previously determined dsRNA binding domain and region required for interacting with PKR in HuTRS1 are noted by bars over the alignment. Arrows below the alignment indicate the sites of RhTRS1 truncation mutants discussed later.

We first infected HF, RF, or AGM cells (BSC₄₀) with VV containing wild-type E3L (VC2+lacZ), VVΔE3L, VV-HuTRS1, or VV-RhTRS1. Each of these viruses contains a lacZ cassette, allowing measurement of β-Gal activity for monitoring viral replication (7). As expected, VC2+lacZ replicated relatively well compared to VVΔE3L in all three cell types (Fig. 2). VV-HuTRS1 replicated well in HF but not in RF or BSC₄₀ cells. In contrast, VV-RhTRS1 replicated well only in BSC₄₀ cells and not in HF or, unexpectedly, not even in RF. Thus, the two TRS1 alleles demonstrated species-specific differences in their abilities to support VVΔE3L replication.

Fig 2.

Replication of VV-HuTRS1 and VV-RhTRS1 is cell type specific. (A) Measurement of viral replication in HF, RF, and BSC₄₀ cells. Triplicate wells of each cell type were mock infected or infected (MOI, 3) with VC2+lacZ (VC2), VV-HuTRS1, VV-RhTRS1, or VVΔE3L. At 24 hpi, viral replication was quantified by measurement of β-Gal as described in Materials and Methods. The mean enzyme activities (+ standard deviations) are shown. (B) Analysis of total protein synthesis by metabolic labeling of infected HF, RF, and BSC₄₀ cells. The same mock-infected or infected cells analyzed in panel A were pulse-labeled with [³⁵S]methionine as described in Materials and Methods, after which cell lysates were prepared and the proteins analyzed by SDS-PAGE and autoradiography.

To clarify the basis for these replication patterns, we monitored overall protein synthesis at 24 h postinfection (hpi) by metabolic labeling, SDS-PAGE, and autoradiography. All cells infected with VVΔE3L synthesized very low levels of new protein, while those infected with VC2+lacZ produced a pattern indicative of ongoing viral protein synthesis at 24 hpi (Fig. 2B). After infection with VV-HuTRS1, HF continued to synthesize proteins while RF and BSC₄₀ did not. After VV-RhTRS1 infection, BSC₄₀ cells continued to synthesize viral proteins while HF and RF did not. Thus, the ability of each TRS1 allele to rescue VVΔE3L replication corresponded to its ability to support continued protein synthesis, suggesting that the restricted replication of VV-HuTRS1 in RF and BSC₄₀ cells and of VV-RhTRS1 in HF and RF may be due to their failure to block the PKR pathway.

Species-specific effects of TRS1 proteins are mediated by the PKR pathway.

The replication patterns noted in Fig. 2 could have been due to differences in expression of the alternative TRS1 proteins or to differences in their activities in the alternative cell types. To distinguish between these possibilities, we first evaluated expression of the TRS1 proteins in an immunoblot assay, making use of the C-terminal His tag present on each protein. In this experiment we used a recombinant, VVΔE3L+E3L, which contains a His-tagged E3L gene inserted into the thymidine kinase locus, in place of VC2+lacZ, to enable comparison of expression among the various PKR antagonists.

As in the previous experiment (Fig. 2), VV-HuTRS1 replicated well in HF but not BSC₄₀ cells, while VV-RhTRS1 had the opposite pattern (Fig. 3A). VVΔE3L+E3L replicated moderately well in HF. Although in this experiment it replicated to a low level in BSC₄₀ cells, this virus consistently replicated to a greater extent than the background level of VVΔE3L replication (data not shown). Immunoblot assays revealed that HuTRS1 was easily detectable in HF but was only faintly detectable in VV-HuTRS1-infected BSC₄₀ cells at 24 hpi (Fig. 3B, bottom panel). Conversely, RhTRS1 was detectable in BSC₄₀ cells but not in HF after infection with VV-RhTRS1. E3L was detectable in both cell types infected with VVΔE3L+E3L. Thus, expression of the PKR antagonists correlated with rescue of VV replication.

Fig 3.

Cell-type-specific differences in VV-HuTRS1 and VV-RhTRS1 replication are mediated by PKR. (A) Viral replication in various cell types. The indicated cells were mock infected or infected with VVΔE3L+E3L, VV-HuTRS1, VV-RhTRS1, or VVΔE3L (MOI, 3), and viral replication was measured with a β-Gal activity assay as described in Materials and Methods. Mean enzyme activity (+ standard deviation) for triplicate wells is shown. (B) Expression of virally encoded His-tagged proteins in different cell types. Protein lysates were collected from the infected cells shown in panel A at 24 hpi (bottom blot) or from another set of wells at 6 hpi (upper blot) and were subjected to immunoblot analysis using anti-His antibody as described in Materials and Methods. Arrowheads (right) indicate HuTRS1 and RhTRS1. (C) Expression of HuTRS1 and RhTRS1 in RFs. Lysates from RF cells infected as indicated were prepared at 6 and 24 hpi, and protein expression was analyzed by immunoblotting with anti-His antibody. (D) PKR expression from parental HF-tert cells (HF) or from cells stably transduced with either a PKR-targeting shRNA (PKR-kd) or a control shRNA (Ctrl-kd). Cell lysates were prepared and analyzed by immunoblotting with PKR (sc-6282) and actin antibodies.

We next explored the possibility that the cause of the limited expression of HuTRS1 in BSC₄₀ cells and of RhTRS1 in HF was that the TRS1 variants were unable to block PKR in the nonpermissive cell types. When we examined the expression patterns at 6 hpi, we detected similar amounts of HuTRS1 and RhTRS1 in BSC₄₀ cells. RhTRS1 was also detectable at 6 hpi in HF at a level only slightly lower than HuTRS1. These results revealed that both VV-HuTRS1 and VV-RhTRS1 enter both cell types and express the encoded TRS1 proteins early during infection. However, as a result of an apparent inability of HuTRS1 to prevent the shutoff of protein synthesis in BSC₄₀ cells and of RhTRS1 to perform this function in HF (Fig. 2), the TRS1 proteins fail to accumulate and are absent or nearly absent by 24 hpi in these nonpermissive cells.

We have focused our experiments on HF and BSC₄₀ cells, in which HuTRS1 and RhTRS1 have divergent properties, but we also evaluated expression of these proteins in RF, in which neither one rescues VVΔE3L replication or prevents shutoff of protein synthesis (Fig. 2). At both 6 and 24 hpi, VV-HuTRS1 and VV-RhTRS1 expressed barely detectable levels of HuTRS1 and RhTRS1, respectively, compared to the expression of E3L from VVΔE3L+E3L (Fig. 3C). Although there may be a mechanism that stabilizes specific TRS1 proteins in specific cell types, these results are most consistent with the conclusion that TRS1 proteins that cannot block the shutoff of protein synthesis fail to accumulate due to their reduced production during infection.

If the failure to block PKR activation accounts for the lack of RhTRS1 accumulation in HF, then knocking down PKR in HF should augment RhTRS1 expression. To test this prediction, we transduced HF-tert cells with retroviral vectors containing an shRNA targeting PKR or a control shRNA, prepared clonal derivatives of these cells, and documented the successful knockdown of PKR in an immunoblot assay (Fig. 3D). In the PKR knockdown cells, all viruses, including VVΔE3L and VV-RhTRS1, replicated to a similar extent (Fig. 3A). Notably, RhTRS1 was expressed at a high level even at 24 hpi in these cells (Fig. 3B). The control knockdown cells behaved like the parental HF. All the viruses replicated well and expressed the His-tagged proteins in BHK cells, which are known to be fully permissive for VVΔE3L (27).

These results suggest that RhTRS1 cannot block human PKR, that HuTRS1 cannot block AGM PKR, and that neither TRS1 blocks rhesus PKR. As a result, protein synthesis ceases and little or no TRS1 accumulates by 24 hpi in the nonpermissive cell types. The failure of RhTRS1 to inhibit human or rhesus PKR was unexpected in light of the ability of RhCMV to replicate in HF and RF. We therefore investigated whether the functions necessary for inhibition of human PKR by HuTRS1 were conserved in RhTRS1.

RhTRS1 binds to dsRNA.

Previous studies have suggested that HuTRS1 needs to bind to both dsRNA and to PKR in order to rescue VVΔE3L replication (19, 20). Since RhTRS1 is unable to rescue VVΔE3L replication in human cells (Fig. 2 and 3), we hypothesized that it might lack one of these activities.

To evaluate whether RhTRS1, as expressed in human cells by RhCMV infection, binds to dsRNA, we prepared lysates of HF after mock infection or infection with RhCMV or HCMV. We assessed the dsRNA binding ability of the TRS1 proteins by immunoblot assay of proteins that bound to poly(I · C)-agarose beads in the presence or absence of free competitor poly(I · C) (Fig. 4A). Like HuTRS1, RhTRS1 was detectable in the cell lysates when we used a polyclonal serum raised to HuTRS1 (31). Pull-down assays showed that RhTRS1 and HuTRS1 each bound to the dsRNA beads and that the binding was competed by free poly(I · C).

Fig 4.

RhTRS1 binds to dsRNA. (A) dsRNA binding by HuTRS1 and RhTRS1 following CMV infection. HFs were infected with HCMV or RhCMV (MOI, 3). Cell lysates prepared at 48 hpi were preincubated with no competitor (-) or with free poly(I·C) competitor (+) and then incubated with dsRNA-agarose beads. Bound proteins and total cell lysates (Lys) were analyzed by immunoblot assay using TRS1 antiserum as described in Materials and Methods. (B) Binding of cell-free translated RhTRS1 to dsRNA-agarose. Full-length RhTRS1, the indicated deletion mutants, and the nonbinding control, GFP, were in vitro translated in the presence of [³⁵S]methionine, then bound to dsRNA-agarose (I:C) or naked control beads (NB), and washed, and the bound proteins were analyzed alongside total cell-free lysate (Lys) by SDS-PAGE and autoradiography as described in Materials and Methods.

These results demonstrate that RhTRS1, as expressed in HF after RhCMV infection, can bind to dsRNA. However, as is the case for MCMV (10), it is possible that dsRNA binding by RhTRS1 may require a second RhCMV protein that is produced by RhCMV infection but not by VV-RhTRS1 infection. Therefore, we tested the dsRNA binding properties of RhTRS1 in the absence of other viral proteins by cell-free translation of RhTRS1 followed by dsRNA binding assays. In these experiments, full-length RhTRS1 bound to poly(I · C)-agarose beads but not to control beads, indicating that, like HuTRS1, RhTRS1 alone can bind to dsRNA (Fig. 4B).

RhTRS1 and HuTRS1 are most similar in sequence in their amino-terminal regions, which is where the HuTRS1 dsRNA binding domain is localized (Fig. 1). To assess whether the functional organization of RhTRS1 is similar to HuTRS1, we tested several N- and C-terminal deletions for dsRNA binding activity (Fig. 4B). Deletion of the N-terminal region to codon 182 or beyond eliminated dsRNA binding. Deletion of the C terminus to codon 548 slightly reduced binding, while deletion to codon 347 greatly reduced but did not entirely eliminate binding. Thus, as in HuTRS1, the N-terminal region of RhTRS1 seems to contain the residues required for dsRNA binding.

These data indicate that, like HuTRS1, RhTRS1 binds to dsRNA. Thus, the different abilities of HuTRS1 and RhTRS1 to block the PKR pathway and to support VVΔE3L replication in human cells are not easily explained by any apparent difference in their dsRNA binding activities.

RhTRS1 does not bind to kinase-dead PKR.

An alternative explanation for the failure of RhTRS1 to rescue VVΔE3L replication in HF is that, unlike HuTRS1, it does not bind to human PKR. To test this possibility we cloned yeast codon-optimized RhTRS1 and HuTRS1 genes into the yeast two-hybrid binding domain vector pGBT9. Because wild-type PKR genes from many primate species inhibit yeast growth (14), we used kinase-dead point mutants of the human, rhesus, and AGM PKR genes cloned into the yeast two-hybrid activation domain vector pGAD424. All binding and activation domain plasmid double transformants grew well on control (YC -Leu -Trp) plates (data not shown). Consistent with previous coimmunoprecipitation assays in mammalian cells (20), HuTRS1 directly interacted with human PKR as assessed both by yeast growth on His-minus plates (Fig. 5A) and by expression of β-Gal (Fig. 5B) from the lacZ reporter gene in the yeast strain. HuTRS1 also appeared to bind inefficiently to rhesus PKR but not at all to AGM PKR or to the empty activation domain vector. Despite a small amount of growth of transformants containing RhTRS1 and either human or rhesus PKR on His-minus plates, these yeast cells did not express β-Gal above background levels. Surprisingly, RhTRS1 did not bind to AGM PKR. Thus, RhTRS1 does not bind strongly to kinase-dead PKR from any of the tested primates.

Fig 5.

RhTRS1 does not bind to PKR in the yeast two-hybrid system. (A) Yeast two-hybrid analysis of S. cerevisiae strain pJ69-4α transformed with the indicated binding domain plasmids (outer labels, HuTRS1, RhTRS1, and the parent vector, pGTB9) and activation domain expression vectors (PKR alleles and the vector control, pGAD424) by growth on -His plates. (B) Quantification of protein-protein interactions by β-Gal expression. Yeast double transformants were grown in -Leu, -Trp medium overnight, then in YEPD medium for an additional 6 to 8 h. Lysates were prepared and assayed for β-Gal as described in Materials and Methods.

Differing mechanisms of PKR pathway inhibition by HuTRS1 and RhTRS1.

The puzzling finding that RhTRS1 rescues protein synthesis and VVΔE3L replication in AGM cells but is unable to bind to kinase-dead AGM PKR in a yeast two-hybrid assay suggested that the mechanism by which RhTRS1 functions in AGM cells differs from that of HuTRS1 in human cells. Therefore, we analyzed the PKR activation pathway in more detail in HF and BSC₄₀ cells after infection with VV-HuTRS1, VV-RhTRS1, and control viruses.

We first examined the activation of PKR in these lysates by immunoblot assay, using an antibody directed against human phospho-PKR (T446), which also reacts with the phosphorylated AGM PKR. HuTRS1 prevented PKR phosphorylation in human cells but not in BSC₄₀ cells (Fig. 6). Consistent with its failure to bind either human or AGM PKR (Fig. 5), RhTRS1 did not prevent PKR phosphorylation in HF or BSC₄₀ cells.

Fig 6.

Effects of HuTRS1 and RhTRS1 on PKR and eIF2α phosphorylation. HF and BSC₄₀ cells were mock infected or infected with VC2, VV-HuTRS1, VV-RhTRS1, or VVΔE3L. At 48 hpi the cells were lysed, and equivalent amounts of protein were analyzed by immunoblotting with antibodies directed against total and phospho-PKR, total and phospho-eIF2α, and actin.

We next examined the impact of PKR activation on the next step in the pathway, phosphorylation of eIF2α. As predicted based on the presence of activated PKR, phosphorylated eIF2α accumulated in BSC₄₀ cells infected with VV-HuTRS1 and in HF infected with VV-RhTRS1, but not in HF infected with VV-HuTRS1 (Fig. 6). The abundance of phosphorylated eIF2α in HF after infection with VV-RhTRS1 appeared to be somewhat less than after VVΔE3L infection in this experiment, but that difference was not observed in other similar experiments, and in all cases the eIF2α phosphate levels were much higher after infection with VV-RhTRS1 compared to mock infection or infection with VC2- or VV-HuTRS1. Notably, despite its failure to block PKR phosphorylation, RhTRS1 inhibited eIF2α phosphorylation in BSC₄₀ cells. This effect on eIF2α is consistent with RhTRS1's ability to rescue VVΔE3L replication and protein synthesis in BSC₄₀ cells (Fig. 2 and 3). These results reveal that RhTRS1 inhibits AGM PKR in a qualitatively different manner than the way HuTRS1 inhibits human PKR.

The observation that RhTRS1 is unable to block PKR phosphorylation but does inhibit phosphorylation of eIF2α led us to investigate the possibility that RhTRS1, unlike HuTRS1, might bind only to the activated form of PKR. In line with this hypothesis, phosphorylated AGM PKR clearly associated with RhTRS1 in BSC₄₀ cells infected with VV-RhTRS1 (Fig. 7, lane 10, PKR-P pull down). To evaluate whether RhTRS1 binds to inactive AGM PKR, we coinfected BSC₄₀ cells with both VV-RhTRS1 and VC2-GFP (lane 11). Under these conditions, PKR remained unphosphorylated, presumably due to expression of E3L by VC2-GFP. Although expression of RhTRS1 was also reduced in the coinfected cells (likely due to the lower MOI of VV-RhTRS1 used and competition between the two viruses), the complete absence of PKR binding to RhTRS1 (lane 11, PKR-tot) suggested that RhTRS1 does not bind to inactive AGM PKR in mammalian cells. We cannot exclude the possibility that under these conditions E3L binds to AGM PKR and sterically prevents RhTRS1 from also binding. However, together with the yeast two-hybrid results, these data support a model in which RhTRS1 binds to AGM PKR only after it has undergone autophosphorylation and is then able to block the eIF2α kinase reaction, consequently allowing continued protein synthesis and viral replication.

Fig 7.

RhTRS1 binds to activated AGM PKR. HF and BSC₄₀ cells were mock infected or infected with VV-HuTRS1, VV-RhTRS1, VC2-GFP, or VVΔE3L (MOI, 3) or coinfected with VV-HuTRS1 or VV-RhTRS1 (MOI, 2) and VC2-GFP (MOI, 4) as indicated. At 24 hpi cell lysates were prepared, and equivalent amounts of each lysate were incubated with nickel-agarose beads. Cell lysates and bound proteins were then analyzed by immunoblotting with antibodies directed against phospho-PKR, total PKR, His, and actin. Arrowheads indicate bound total PKR.

We used a similar strategy to examine the PKR binding properties of HuTRS1 in mammalian cells. Consistent with the prior experiment (Fig. 6), HuTRS1 blocked PKR phosphorylation to a large extent in HF (Fig. 7, lane 1 versus 6, PKR-P). However, it did also bind to the small amount of phosphorylated human PKR that accumulated in VV-huTRS1-infected cells (lane 1). Coinfection of HF with VV-huTRS1 and VC2-GFP reduced phosphorylation of human PKR to background levels and reduced expression of HuTRS1, but the HuTRS1 that was expressed bound to human PKR (lane 2, PKR-tot). These results are consistent with cotransfection experiments in which we consistently found that kinase-dead PKR bound to human TRS1 (data not shown). In BSC₄₀ cells, HuTRS1 bound very weakly to phosphorylated AGM PKR (lane 8, PKR-P). When phosphorylation of AGM PKR was blocked by coinfection with VC2-GFP, we detected no binding to AGM PKR (lane 9, PKR-tot pull down).

Together these results show that HuTRS1 and RhTRS1 have differing binding properties and activities. HuTRS1 binds to inactive human PKR and prevents its autophosphorylation. HuTRS1 also binds to phosphorylated human PKR, and to a slight extent to AGM PKR, but at least in the latter case its binding is unable to block the eIF2α kinase activity of AGM PKR. In contrast, RhTRS1 does not bind to or block autophosphorylation of AGM or human PKR, but it is able to bind to phosphorylated AGM PKR and blocks its eIF2α kinase activity.

DISCUSSION

Pressure from pathogenic viruses can drive the rapid evolution of host genes that have antiviral activities (15). In response, viral antagonists of these genes also undergo rapid evolution. For example, changes in the PKR gene during primate evolution have resulted in modern alleles that have greatly varying sensitivities to inhibition by the poxvirus eIF2α-mimic K3L, which is itself under positive selection (14). Multiple codons in PKR have been evolving rapidly, likely in response to pressure from PKR antagonists encoded by extinct viruses and ancestors of extant viruses.

This evolutionary perspective led us to investigate the impact that PKR adaptations may have had on the species-specific infectivities of CMVs. Since CMVs are believed to have coevolved with their host species, each may have contributed to, as well as adapted to, changes in PKR. Consistent with this view, HCMV replicates in human cells but not in Old World monkey cells (data not shown), and HuTRS1 inhibits human but not Old World monkey PKR (Fig. 2 and 6). The finding that HuTRS1 binds to human PKR and blocks its autophosphorylation (20) (Fig. 5 to 7) can account for its ability to block eIF2α phosphorylation and to maintain protein synthesis in cells infected with HCMV and VV mutants that lack other PKR antagonists (11, 31) (Fig. 2 and 6). We also found that VV-HuTRS1 replicates well in cells from other hominoids, including chimpanzees, orangutans, gorillas, and gibbons (data not shown). In contrast, HuTRS1 does not bind to rhesus or AGM PKR in yeast two-hybrid assays or in AGM cells (Fig. 5 and 7) and is unable to block the PKR pathway in rhesus or AGM cells (Fig. 2, 6, and 7). Taken together, these data suggest that the ability of HuTRS1 to inhibit PKR activation is hominoid lineage specific. One or more of the many changes that arose during the rapid evolution of PKR in Old World monkeys likely accounts for their resistance to HuTRS1 binding and antagonism and may contribute to the inability of HCMV to replicate in Old World monkey cells.

Unlike HCMV, RhCMV replicates in both Old World monkey cells and human cells (2, 29). This fact, along with studies showing that the PKR antagonists encoded by both HCMV and MCMV are essential in order for the viruses to replicate at all in cell culture (10, 31, 32, 44), suggests that RhCMV likely has the ability to block human PKR. Yet RhTRS1, the predicted RhCMV antagonist based on homology to HCMV, is unable to block the PKR pathway in human cells. Surprisingly, RhTRS1 also appears unable to block rhesus PKR, as judged by its inability to rescue protein synthesis or VVΔE3L replication in RF. However, RhTRS1 is able to block eIF2α phosphorylation in BSC₄₀ cells (Fig. 6) and can rescue VVΔE3L replication in several other AGM cell lines (data not shown). We have not ruled out the possibility that RhTRS1 acts in BSC₄₀ cells by blocking another of the cellular eIF2α kinases. However, VVΔE3L replicates in cells in which PKR has been knocked out or down (45, 46) (Fig. 3), and RhTRS1 does bind to activated AGM PKR (Fig. 7). Thus, our results support the conclusion that RhTRS1 functions by blocking PKR in AGM cells, but not in human cells or even in cells from its host species, the rhesus macaque. We found that VV-RhTRS1 does not replicate well in five other RF lines (provided by Peter Barry) (data not shown), so it is unlikely that the RF cell line we used in these experiments is in some way not representative of RF. However, we have not yet tested VV-RhTRS1 in a panel of primary AGM cells in order to clarify whether its ability to replicate in BSC₄₀ cells is a property of the PKR gene or somehow related to the fact that these are an established cell line. HuTRS1 functions well in both primary and transformed human cells (9, 19, 31), leading us to hypothesize that the ability of TRS1 alleles to antagonize PKR tracks more closely with the PKR gene than with the transformation state of the cells, but additional studies are needed to clarify this issue.

Another possibility is that RhCMV really is an AGM CMV isolate. In this regard, RhCMV strain 68-1 was isolated from rhesus macaques that were housed with AGM (2). However, this strain clearly replicates in rhesus macaques (30, 41), and its sequence differs substantially from that reported for AGM CMV (1, 3). Thus, it seems most likely that RhCMV really is a rhesus macaque virus, albeit one that may be able to cross species barriers.

The fact that RhTRS1 appears unable to block PKR in human or rhesus cells raises the question as to how RhCMV is able to replicate in these cells. It is possible that VVΔE3L-based assays do not accurately model events occurring during CMV infections. For example, perhaps RhCMV infection produces a relatively small amount of dsRNA or for some other reason only activates PKR to a low level compared to VV-RhTRS1 infection, and thus even weak antagonism of human or rhesus PKR by RhTRS1 may be sufficient to support replication of RhCMV. This explanation would not account for the differences in host range of VV-RhTRS1 and VV-HuTRS1, which likely produce similar amounts of dsRNA. Even if VV requires a more potent PKR antagonist than do CMVs, our experiments illuminate a clear difference in the potencies of HuTRS1 and RhTRS1 in cells from different species. Another explanation for the ability of RhCMV to replicate in human and rhesus cells while VV-RhTRS1 does not is that RhCMV might encode a second factor that acts with RhTRS1 to block human and rhesus PKR, analogous to the MCMV system in which two viral genes, m142 and m143, are both required to block the PKR pathway and for viral replication in mouse cells (10, 31, 32, 44). It is also possible that RhCMV encodes another gene (or genes) that functions independently of RhTRS1 to block PKR in human and rhesus cells, while RhTRS1 serves this function in AGM cells. Such an arrangement would be analogous to the VV system, in which E3L is critical for replication in many cell types but not in BHK cells, while K3L is required for efficient PKR antagonism in BHK cells but few others (27). Construction and analysis of a RhCMV with deletion of its TRS1 gene will be important for distinguishing among these possibilities.

Regardless of how RhCMV is able to block PKR in order to replicate in human and rhesus cells, our studies of the effects of RhTRS1 in AGM cells reveal a new mechanism by which a CMV dsRNA binding protein can block the PKR pathway. Most significantly, unlike HuTRS1, RhTRS1 does not block PKR autophosphorylation but does still block the next step in the PKR pathway, the phosphorylation of eIF2α. It can thereby prevent the shutoff of protein synthesis and allow viral replication. HuTRS1 and RhTRS1 appear to have different PKR binding properties. HuTRS1 binds to inactive human PKR and inhibits its activation. HuTRS1 can also bind to phosphorylated human PKR. On the other hand, RhTRS1 binds to AGM PKR only after it has been phosphorylated (Fig. 4 and 7). Thus, TRS1 proteins may differ in their recognition of higher-order structural determinants, such as surfaces created or exposed by the conformational changes that PKR undergoes as a result of binding to dsRNA, dimerization, and autophosphorylation (12). It may be that RhTRS1 binding to activated AGM PKR interferes with the conformational change necessary to reposition serine 51 of eIF2α during the kinase reaction (13).

RhTRS1 is not the only viral protein that blocks the PKR kinase after the autophosphorylation step. The VV K3L gene encodes a PKR pseudosubstrate that mimics eIF2α and binds preferentially to activated PKR, preventing phosphorylation of eIF2α (12). Respiratory syncytial virus infection appears to allow PKR activation but blunts its eIF2α kinase activity, possibly as a result of the viral N protein binding to PKR and shifting its association from eIF2α to protein phosphatase 2A (18). Although the effects of blocking PKR's eIF2α kinase activity before or after autophosphorylation are similar with respect to translational initiation, it is possible that there are different consequences of the two mechanisms. For example, antagonist-bound activated PKR might still be able to phosphorylate other potential substrates (22), although additional studies will be needed to determine whether the viral antagonists block these kinase reactions too.

Our studies reveal substantial differences in TRS1-PKR interactions and mechanisms in primates. It therefore seems paradoxical that the PKR antagonists from a much more distantly related virus, MCMV, function in human cells and that HuTRS1 can function in mouse cells (10, 44). It may be that under some assay conditions, such as overexpression resulting from transient transfection (10), the dsRNA binding properties of these antagonists are sufficient to block the PKR pathway (and possibly other dsRNA-activated pathways) and allow viral replication. Such a mechanism may explain an N-terminal deletion mutant of E3L that binds to dsRNA but not to PKR and enables viral replication in HeLa cells (39). However, this mutant fails to prevent eIF2α phosphorylation or protein synthesis inhibition at late times after infection in cell culture, fails to rescue pathogenesis in mice, and is unable to antagonize PKR in a yeast assay (26, 35). Another possibility is that during adaptation to some of the many changes arising in PKR during Old World monkey evolution, RhTRS1 lost its ability to interact with surfaces of ancestral PKR alleles that were maintained in both the hominoid and rodent lineages. Further genetic and structural dissection of the interactions of PKR alleles and their CMV antagonists will be needed to resolve this puzzle and to identify the specific settings in which PKR antagonism contributes to the species specificity of CMV replication.