Antagonism of the Protein Kinase R Pathway in Human Cells by Rhesus Cytomegalovirus

ABSTRACT

While cytomegalovirus (CMV) infections are often limited in host range by lengthy coevolution with a single host species, a few CMVs are known to deviate from this rule. For example, rhesus macaque CMV (RhCMV), a model for human CMV (HCMV) pathogenesis and vaccine development, can replicate in human cells, as well as in rhesus cells. Both HCMV and RhCMV encode species-specific antagonists of the broadly acting host cell restriction factor protein kinase R (PKR). Although the RhCMV antagonist of PKR, rTRS1, has very limited activity against human PKR, here, we show it is essential for RhCMV replication in human cells because it prevents human PKR from phosphorylating the translation initiation factor eIF2α, thereby allowing continued translation and viral replication. Although rTRS1 is necessary for RhCMV replication, it is not sufficient to rescue replication of HCMV lacking its own PKR antagonists in human fibroblasts. However, overexpression of rTRS1 in human fibroblasts enabled HCMV expressing rTRS1 to replicate, indicating that elevated levels or early expression of a weak antagonist can counteract a resistant restriction factor like human PKR. Exploring potential mechanisms that might allow RhCMV to replicate in human cells revealed that RhCMV makes no less double-stranded RNA than HCMV. Rather, in human cells, RhCMV expresses rTRS1 at levels 2 to 3 times higher than those of the HCMV-encoded PKR antagonists during HCMV infection. These data suggest that even a modest increase in expression of this weak PKR antagonist is sufficient to enable RhCMV replication in human cells.

IMPORTANCE Rhesus macaque cytomegalovirus (RhCMV) offers a valuable model for studying congenital human cytomegalovirus (HCMV) pathogenesis and vaccine development. Therefore, it is critical to understand variations in how each virus infects and affects its host species to be able to apply insights gained from the RhCMV model to HCMV. While HCMV is capable only of infecting cells from humans and very closely related species, RhCMV displays a wider host range, including human as well as rhesus cells. RhCMV expresses an antagonist of a broadly acting antiviral factor present in all mammalian cells, and its ability to counter both the rhesus and human versions of this host factor is a key component of RhCMV's ability to cross species barriers. Here, we examine the molecular mechanisms that allow this RhCMV antagonist to function against a human restriction factor.

INTRODUCTION

The replication of cytomegaloviruses (CMVs) is generally restricted to a narrow range of host species. For example, human CMV (HCMV) is only known to infect humans in nature and to replicate in human and chimpanzee cells in culture (1). This specificity is due in part to coevolutionary adaptations occurring over the past ∼60 to 80 million years (2) that have honed the ability of these viruses to utilize dependency factors and to evade or counteract restriction factors encoded by their host species. Adaptations in viral genes that have tailored viral replication to the protein landscape in one host may disable their ability to functionally interact with factors that have evolved in other species and thus create barriers to cross-species transmission. However, several nonhuman primate CMVs, including rhesus CMV (RhCMV), African green monkey (Agm) CMV, and squirrel monkey CMV, are somehow able to overcome these species-specific barriers and replicate in human cells (3–5).

One of the most rapidly evolving host cell restriction factors is protein kinase R (PKR) (6, 7). In the presence of double-stranded RNA (dsRNA), which accumulates during many viral infections (8–10), PKR is activated by dimerization and autophosphorylation (11). Activated PKR phosphorylates the α subunit of eukaryotic initiation factor 2 (eIF2α), which in turn tightly binds to and impedes the activity of the guanine nucleotide exchange factor eIF2B, thus inhibiting translation initiation and viral replication. Many viruses, including CMVs, have evolved one or more factors that overcome the PKR pathway and are critical for viral replication and pathogenesis (12, 13). In the case of HCMV, either of two related genes, TRS1 or IRS1, is absolutely essential for replication in normal human fibroblasts (HF), but both are dispensable in cells lacking PKR (8, 14, 15).

Computational analyses have suggested that PKR has been under particularly strong selection during the evolution of Old World monkeys (7). Consistent with this observation, the protein products of HCMV TRS1 and IRS1 (hTRS1 and hIRS1) cannot bind to or inhibit PKR in Old World monkey cells, which likely contributes to the failure of HCMV to replicate in these cells (16). However, while RhCMV TRS1 (rTRS1) does not inhibit human PKR efficiently, RhCMV is still able to replicate in HF (17). Thus, the barrier to CMV cross-species transmission imposed by changes in PKR is not completely predictable based on phylogenetic considerations.

We have focused on RhCMV because of its importance as an emerging model of congenital HCMV pathogenesis and vaccine development (18, 19). These studies were undertaken to clarify how RhCMV is able to replicate in human cells despite observations that have indicated that rTRS1 does not bind to or inhibit human PKR. Since PKR antagonism is essential for HCMV replication in HF, these observations, which were made using heterologous expression systems, predict that RhCMV would not be able to replicate in human cells. However, RhCMV does in fact replicate efficiently in HF. Moreover, its replication is dependent on rTRS1 counteracting human PKR. However, rTRS1 inserted into an HCMV recombinant lacking TRS1 and IRS1 (HCMV[ΔI/ΔT]) is insufficient to enable replication of the virus (HCMV[rT]) in HF. Constitutive expression of a RhTRS1 transgene is sufficient to support the replication of HCMV[rT], but not that of HCMV[ΔI/ΔT]. Along with analyses of rTRS1 expression kinetics during RhCMV infection, our results suggest that a high level of rTRS1 is necessary and sufficient to overcome human PKR. These observations highlight a strategy by which viruses may be able to overcome otherwise resistant host restriction factors and thereby cross species barriers to replicate in a new host.

RESULTS

RhCMV can cross species barriers to replicate in HF.

Although CMVs generally exhibit a high degree of species specificity in their ability to replicate in cell culture, several nonhuman primate CMVs, including RhCMV, have been reported to replicate in human cells (3, 17). To compare the efficiencies of RhCMV and HCMV replication in human and rhesus cells, we infected cells with both viruses and determined the titers of the progeny virus released into the medium over the course of 6 days. HCMV replicated in HF but not at all in rhesus fibroblasts (RF) (Fig. 1A). On the other hand, RhCMV replicated to even higher titers than HCMV in HF, but not as well as it did in RF. In another experiment, we compared the quantities of intracellular versus extracellular virus accumulated at 6 days postinfection (Fig. 1B). Once again, HCMV replicated in HF but not in RF. RhCMV infection produced abundant intracellular and extracellular virus in HF, albeit less than in RF. These data reveal that although RhCMV replicates most efficiently in cells from its host species, it does replicate remarkably well in HF.

FIG 1.

RhCMV can replicate in HF. (A) Measurement of HCMV and RhCMV replication in HF and RF. At the indicated times after infection (MOI = 0.1), the amount of extracellular virus present in the medium was determined by titration on HF^[PKR-ko] (mean titer ± standard deviation [SD]; n = 3). The data are representative of at least three separate experiments. Statistical significance for replication of each virus in HF versus RF was determined using an unpaired t test (***, P < 0.0005). (B) Analysis of the amounts of intracellular and extracellular HCMV and RhCMV produced in HF. Six days postinfection (MOI = 0.1), medium was collected prior to harvesting the cell monolayers. The cell pellets were then collected and freeze-thawed three times to release virus, and the titers of the virus present in the cell supernatants and pellets were determined on HF^[PKR-ko] (mean titer and SD; n = 3). The data are representative of at least three separate experiments. Statistical significance for analogous samples in HF versus RF was determined using an unpaired t test (**, P < 0.005; ***, P < 0.0005).

RhCMV TRS1 is essential for viral replication in HF and for efficient replication in RF.

Previous studies had demonstrated that either of two HCMV proteins, hIRS1 or hTRS1, is absolutely required for HCMV replication in HF because of their ability to inhibit PKR (8, 14, 15). However, rTRS1 expressed in Saccharomyces cerevisiae or in vaccinia virus (VACV) recombinants was unable to bind to or inhibit human PKR efficiently (16), leading us to wonder how RhCMV is able to replicate in HF. One possible explanation is that RhCMV might encode another factor besides rTRS1 that can antagonize human PKR. Many viruses, including herpes simplex virus 1 (HSV-1) and VACV, encode more than one functionally distinct PKR antagonist (12, 13). To address this possibility, we deleted rTRS1 from a bacterial artificial chromosome (BAC) derived from RhCMV strain 68-1 and reconstituted the resulting RhCMV[ΔrT] as described in Materials and Methods. Unlike wild-type (wt) HCMV and RhCMV, RhCMV[ΔrT] was unable to replicate at all in HF (Fig. 2). It did replicate to a very low level in HF expressing rTRS1 (HF+rTRS1) and to nearly wt levels in RF that expressed rTRS1 (RF+rTRS1). Somewhat surprisingly, RhCMV[ΔrT] also replicated to a low but reproducibly detectable level in RF. Most importantly, these data demonstrate that rTRS1 is absolutely required for RhCMV replication in HF.

FIG 2.

RhCMV TRS1 is required for replication in HF and for efficient replication in RF. Shown is replication of HCMV, RhCMV, and RhCMV[ΔrT] in HF, RF, and HF and RF expressing rTRS1 (HF+rT B and RF+rT) (Fig. 5 shows an immunoblot demonstrating rTRS1 expression). Six days after infection (MOI = 0.1), the medium was collected and the amount of extracellular virus present was determined by titration on HF^[PKR-ko] (mean titer and SD; n = 3). The data are representative of at least three separate experiments. Statistical significance for replication of each virus in HF versus each other cell type was determined using an unpaired t test (*, P < 0.05; **, P < 0.005; ***, P < 0.0005; NS, not significant).

RhCMV TRS1 is essential in HF because it antagonizes PKR.

HCMV[ΔI/ΔT] does not replicate at all in HF but replicates well in HF that are PKR deficient, indicating that the only essential function of hTRS1 or hIRS1 is to inhibit PKR (14). To determine whether the role of rTRS1 during RhCMV replication is also to antagonize PKR, we used CRISPR (clustered regularly interspaced short palindromic repeat)/Cas9 and a guide RNA (gRNA) directed against the first double-stranded RNA binding domain in PKR to produce PKR knockout HF (HF^[PKR-ko]). We verified by Sanger sequencing and sequence trace decomposition analysis (20), as well as by immunoblotting (Fig. 3A), that these cells contained mutations that abrogate PKR expression. In addition, we confirmed that VACV with its major PKR antagonist deleted (VACVΔE3L) replicated to wt levels in these cells (Fig. 3B). Next, we infected HF and HF^[PKR-ko] with RhCMV[ΔrT] and monitored the production of virus over the course of 6 days. RhCMV[ΔrT] did not replicate at all in HF but replicated as well as RhCMV did in the HF^[PKR-ko] (Fig. 3C). Thus, the only essential role of rTRS1 in HF is to block the PKR pathway.

FIG 3.

RhCMV TRS1 is essential because it antagonizes human PKR. (A) Expression of PKR in HF and HF^[PKR-ko] and of rTRS1 in HF^[PKR-ko] infected with RhCMV and RhCMV[ΔrT]. Cell lysates were collected and analyzed by immunoblotting with the indicated antisera as described in Materials and Methods. Expression of a viral protein (VP) visualized by stain-free fluorescence to confirm infection is shown. (B) VACV infection confirms the PKR phenotype of HF^[PKR-ko]. The cells were infected with VACV or VACVΔE3L (MOI = 0.1), and viral replication was measured with a β-Gal activity assay (mean enzyme activity and SD; n = 3). (C) Analysis of RhCMV and RhCMV[ΔrT] replication in HF and HF^[PKR-ko]. Triplicate wells were infected with each virus (MOI = 0.1), and on the indicated days postinfection, the amount of extracellular virus present in the medium was determined by titration on HF^[PKR-ko] (mean titer and SD; n = 3). The data are representative of at least three separate experiments. Statistical significance for replication of RhCMV[ΔrT] in HF versus RhCMV and RhCMV[ΔrT] in each other cell type was determined using an unpaired t test (***, P < 0.0005).

RhCMV TRS1 does not prevent phosphorylation of PKR in HF, but it does limit eIF2α phosphorylation.

In our previous studies using a VACVΔE3L recombinant that expresses rTRS1 (VACV[rT]), we had been surprised to discover that rTRS1 did not prevent phosphorylation of PKR, even though it was still able to prevent eIF2α phosphorylation in permissive Agm cells (16). This result suggested that, at least in the VACV system, rTRS1 acts by blocking PKR activity at a step downstream of PKR autophosphorylation and activation. We found the same pattern of increased phosphorylated PKR (PKR-P) but minimal phosphorylated eIF2α (eIF2α-P) accumulation compared to mock-infected cells after infection of HF with RhCMV (Fig. 4). Infection of HF with RhCMV[ΔrT], on the other hand, resulted in increased phosphorylation of both PKR and eIF2α, as was true for cells infected with HCMV[ΔI/ΔT] and VACVΔE3L, as well. The decrease in the amount of PKR in VACVΔE3L-infected HF is likely due to the profound shutoff of host protein synthesis in wt cells infected with the virus, as we have observed in other experiments (14).

FIG 4.

Effects of RhCMV TRS1 on PKR and eIF2α phosphorylation. HF were mock infected or infected with RhCMV, RhCMV[ΔrT], HCMV, HCMV[ΔI/ΔT], or VACVΔE3L. At 48 h postinfection (or 24 h postinfection for VACVΔE3L), 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. The levels of PKR and eIF2α phosphorylation relative to the amount of actin in each sample are indicated below the corresponding immunoblots.

rTRS1 is not sufficient to rescue HCMV[ΔIΔT] replication or to inhibit the PKR pathway in HF.

Although rTRS1 is necessary for RhCMV replication, we previously found that its insertion into VACVΔE3L did not support replication in HF (16). To determine whether it is sufficient in the context of HCMV, we cloned rTRS1 into the HCMV[ΔI/ΔT] BAC and reconstituted the virus (HCMV[rT]) in HF^[PKR-ko]. We then assessed the ability of HCMV[ΔIΔT] expressing hTRS1 (HCMV[hT]), HCMV[rT], and HCMV[ΔI/ΔT] that express green fluorescent protein (GFP) to replicate in HF, HF^[PKR-ko], and a transduced HF cell line that expresses an rTRS1 transgene (Fig. 5A). While HCMV[hT] replicated well in all the cells, HCMV[rT] did not replicate in HF. It did replicate in HF^[PKR-ko] and in the transgenic TRS1-expressing line, albeit more slowly than HCMV[hT] (Fig. 5A). HCMV[ΔI/ΔT] replicated only in the PKR-ko cells and not in the transgenic HF-rTRS1. In addition to monitoring plaque formation by fluorescence microscopy, we analyzed replication of HCMV and HCMV[rT] in HF, HF^[PKR-ko], and two different rTRS1-expressing HF lines by determining the titer of extracellular virus collected on day 6 postinfection (Fig. 5B). HCMV replicated similarly in all of the cells. Although HCMV[rT] did not replicate at all in HF, it did replicate well in HF^[PKR-ko] and in the two rTRS1-expressing cell lines. Thus, the combined rTRS1 expression from the HCMV genome and a cellular transgene is sufficient to support HCMV replication in the absence of TRS1 and IRS1. These results reveal that altering the abundance or timing of expression can enable a weak antagonist like rTRS1 to overcome an otherwise resistant PKR allele.

FIG 5.

HCMV[rT] can replicate in HF that express RhCMV TRS1. (A) Fluorescence and phase microscopy of plaque formation in HF, HF^[PKR-ko], and HF+rTRS1. Cells were infected with HCMV[hT], HCMV[rT], and HCMV[ΔI/ΔT] that express GFP, and after 7 days of infection, images were captured as described in Materials and Methods. (B) Replication of HCMV and HCMV[rT] in complementing cells. HF, HF^[PKR-ko], and two HF lines expressing rTRS1 (HF+rT A and B) were infected with each virus (MOI = 0.1), and on day 6 postinfection, the medium from each well was collected and the amount of extracellular virus was determined by titration on HF^[PKR-ko] (mean titer and SD; n = 3). Statistical significance for replication of HCMV or HCMV[rT] in HF versus each other cell type was determined using an unpaired t test (*, P < 0.05; **, P < 0.005; ***, P < 0.0005; NS, not significant). (C) Immunoblot analysis of TRS1 expression from virus-infected cells and rTRS1-expressing cell lines. HF^[PKR-ko] were infected with the indicated viruses (MOI = 3), and at 48 h postinfection, the infected cells were lysed. Equivalent amounts of protein from these and the uninfected rTRS1-expressing cell lines were analyzed by immunoblotting with antisera directed against TRS1 and UL44 as an indicator of viral infection. A portion of the gel showing proteins labeled by stain-free fluorescence is shown as a loading control. (D) Replication of HCMV[rT] in HF treated with ISRIB. HF wt and HF^[PKR-ko] were infected with HCMV[rT], and at 1 h postinfection, the cells were left untreated or treated with 200 nM ISRIB. After 7 days of infection, the images were captured as described in Materials and Methods.

To analyze the expression of hTRS1 and rTRS1 from the viruses used in these studies, HF^[PKR-ko] were infected with each virus, and lysates were collected at 48 h postinfection. Immunoblot analysis was then performed on these lysates, as well as on lysates from the rTRS1-expressing cell lines (Fig. 5C). Antiserum directed against the HCMV DNA polymerase accessory subunit pUL44 was used to indicate the presence of virus in the infected cell lysates. Note that HCMV[rT] and HCMV[ΔI/ΔT] tend to replicate more slowly than HCMV and HCMV[hT], even in HF^[PKR-ko], which likely accounts for the smaller amount of UL44 on day 8 after infection. This observation may reflect an additional nonessential function of TRS1 that augments CMV replication.

Our data show that PKR is able to completely inhibit HCMV[rT] replication in HF, but whether this phenotype is due to the effect of PKR on eIF2α or another PKR substrate is not known. To address this question, we treated cells with ISRIB, a small molecule that potentiates the activity of the guanine exchange factor eIF2B and thereby renders translation initiation relatively insensitive to eIF2α-P (21–23). We infected HF or HF^[PKR-ko] with HCMV[rT], and then the cells were left untreated or treated with 200 nM ISRIB (in 0.2% dimethyl sulfoxide [DMSO]) at 1 h postinfection. HCMV[rT] replicated almost as well in HF treated with ISRIB as it did in HF^[PKR-ko] with or without ISRIB (Fig. 5D). Therefore, eIF2α is the downstream target of PKR that is responsible for preventing HCMV[rT] replication in HF.

RhCMV produces at least as much dsRNA as HCMV does in HF.

Because dsRNA produced during infection by a wide variety of viruses can activate PKR, one potential mechanism by which a virus with a weak PKR antagonist can evade the PKR pathway is by producing less dsRNA, as was recently reported for monkeypox virus (24). To investigate whether such a mechanism might contribute to the efficiency of RhCMV replication in HF, we purified whole-cell RNA from mock-infected cells and cells infected with HCMV or RhCMV at various times postinfection. After filtering equal amounts of mock- or CMV-infected whole-cell RNA onto positively charged nylon membranes, we determined the amount of dsRNA present in each sample by immunoblot assay using an anti-dsRNA antibody (Fig. 6A). Known amounts of reovirus dsRNA were used as a control. Digestion of the samples with single-stranded RNases (ssRNases) revealed that the intensity of the dsRNA signal was not affected by the presence of ssRNA in the samples. Based on the approximately similar signal intensities from 200 ng of whole-cell RNA and from 100 pg of reovirus dsRNA, we estimate that at 48 to 72 h after infection dsRNA makes up roughly 0.05% of the RNA in HCMV- and RhCMV-infected cells. These data also show that the abundance of dsRNA accumulating in RhCMV-infected HF was more than that in HCMV-infected cells at each time point. Averaging the relative quantities of dsRNA from three separate time course experiments indicated that RhCMV produced at least 3 times as much dsRNA as did HCMV at the 24-, 48-, and 72-h time points (Fig. 6B). Thus, the efficient replication of RhCMV in HF cannot be explained by slower accumulation of this PKR-activating ligand.

FIG 6.

RhCMV generates more dsRNA than HCMV does in HF. (A) Production of dsRNA in HF infected with RhCMV and HCMV. HF were mock infected or infected with RhCMV and HCMV (MOI = 3), and at the indicated times postinfection, whole-cell RNA was harvested; 200 ng of mock-infected or infected cell RNA or the indicated amounts of reovirus RNA were blotted onto nitrocellulose and analyzed using antiserum that recognizes dsRNA (9D5 antibody) as described in Materials and Methods. Samples from 48 h postinfection and reovirus RNA were also treated with RNase T1 and RNase A prior to blotting. (B) The amounts of dsRNA generated by RhCMV and HCMV during three separate infections were quantified as described above, and the average quantities from these replicates with standard deviations are shown. The amounts of dsRNA present at each time point were not significantly different from each other, with the exception of dsRNA from mock-infected cells at 48 h compared to RhCMV dsRNA from the same time point (P = 0.0475; unpaired t test). Mock-infected cell dsRNA was not collected at 72 h.

RhCMV produces more PKR antagonist than HCMV does during infection of HF.

The data shown in Fig. 5 revealed that the added expression of rTRS1 from a cellular transgene was sufficient to enable efficient replication of HCMV[rT]. This effect might have been due to a higher total level of expression of rTRS1 or to the fact that the transgene was expressed prior to the initial time of infection. The fact that HCMV[ΔI/ΔT] did not replicate in transgenic HF+rTRS1 suggested that early expression of rTRS1 alone was insufficient to support HCMV replication. Regardless, these data suggest that differences in expression characteristics between RhCMV and HCMV might explain how RhCMV is able to replicate in HF. We therefore compared the kinetics of expression of rTRS1 during RhCMV infection to those of hTRS1 and hIRS1 during HCMV infection in HF.

Because these analyses relied on antiserum that was raised to the hTRS1 dsRNA-binding domain, we first needed to ascertain whether the antiserum could detect hTRS1 and rTRS1 with the same efficiency. We therefore purified hTRS1 and rTRS1 from cells infected with VACV recombinants expressing 6×His-tagged versions of the proteins. Two dilutions of each protein, as well as hTRS1 purified from a baculovirus vector (25), were visualized by stain-free fluorescence as descried in Materials and Methods to demonstrate that the proteins were present in similar abundances and to validate the linearity of the quantification methods (Fig. 7A, top). The same samples were then immunoblotted, and the proteins were detected with the TRS1 antiserum. The relative intensities of the total protein and of the immunoblot signals were quantified using Image Lab software (Bio-Rad). These analyses revealed that the antiserum does indeed detect both proteins with very similar efficiencies and so can be used to compare the levels of hTRS1 and hIRS1 to that of rTRS1 following infection of HF with HCMV and RhCMV.

FIG 7.

RhCMV produces more TRS1 than HCMV does in HF. (A) Evaluation of the ability of TRS1 antiserum to detect hTRS1 and rTRS1. TRS1 proteins purified from a baculovirus vector (TRS1 ctrl) or from cells infected with VACV recombinants expressing hTRS1 and rTRS1 were detected by stain-free fluorescence and immunoblotting with TRS1 antiserum, and the ratio of the relative immunoblot signal to the relative stain-free protein band intensity was used to compare the affinities of the antiserum for the hTRS1 and rTRS1 proteins. The TRS1 ctrl lanes were loaded with 100 ng (1.0) and 20 ng (0.2) of purified protein, and dilutions containing similar amounts of hTRS1 and rTRS1 were used to determine the relative quantities of each protein on the stain-free gel and the corresponding immunoblot. (B) Accumulation of hTRS1 and hIRS1 compared to rTRS1 in HCMV- and RhCMV-infected cells. HF were infected with HCMV and RhCMV (MOI = 3), and at the indicated times postinfection, the cells were lysed and equivalent amounts of protein (20 μg) were analyzed by immunoblotting with TRS1 antiserum. The baculovirus-expressed TRS1 protein (0.5 μg) used in panel A was included for reference. Dilutions of the RhCMV 96-h lysate (equivalent to 10 μg and 5 μg total protein) were also analyzed, and actin was used as a loading control. Relative amounts of TRS1 (+IRS1) compared to the TRS1 ctrl are indicated below the blots. A TRS1 immunoblot representing median results following multiple repeat time courses is shown. (C) The same lysates analyzed in panel B were analyzed by immunoblotting with HCMV- and RhCMV-specific antisera to pp65 and UL44. Actin was used as a loading control, and a portion of the stain-free gel is also shown.

Finally, we analyzed the expression of these proteins in infected cell lysates (Fig. 7B). At each time point from 24 to 96 h postinfection, rTRS1 accumulated to levels ∼2 to 3 times higher than those of hTRS1 and hIRS1 combined. RhTRS1 migrated as at least four discrete bands (Fig. 7B, RhCMV 96 h, lanes 50% and 25%), suggesting it may undergo a posttranslational modification. We also performed immunoblot analyses on the samples that demonstrated that other RhCMV proteins also accumulated early after infection (Fig. 7C), but their kinetics cannot be quantitatively compared to HCMV protein accumulation due to differences in the antibodies in the two systems. Regardless, these results show that RhCMV may be able to replicate in HF because rTRS1 is expressed to a high level from early times postinfection, which compensates for its relatively weak antagonistic activity against human PKR.

DISCUSSION

The production of PKR antagonists by many different viruses highlights the importance of the PKR pathway as a broadly acting antiviral defense system (12, 13). These factors, which include both RNAs and proteins, utilize a variety of mechanisms to evade or inhibit the PKR pathway. Among CMVs, all known PKR antagonists are proteins encoded by one or more members of the US22 gene family, but the genomic compositions vary among the viruses. For example, mouse CMV (MCMV) has two adjacent genes, m142 and m143, the protein products of which act together in a complex (26). HCMV also possesses two PKR antagonists, hIRS1 and hTRS1, the N-terminal 2/3 of which are identical because the genes are partially included in the repeats flanking the short unique region of the genome, and either is sufficient to inhibit human PKR. Thus far, only one PKR antagonist has been identified in guinea pig CMV (GPCMV), AgmCMV, and RhCMV.

For CMVs and many other viruses, PKR antagonists are critical for viral replication and virulence. This phenotype was nicely demonstrated in a classic study in which the reduced neurovirulence of HSV-1 lacking ICP34.5 was restored to the same level seen with wt HSV-1 in PKR-null mice (27). Similarly, MCMVs lacking m142 or m143 do not replicate in wt cells or in normal mice but replicate to wt levels in PKR-null cells and mice (28). Deletion of GP145 from GPCMV reduces virulence in animals but has only a modest phenotype in cell culture, suggesting that GPCMV might encode a second as-yet-unidentified PKR antagonist (29). However, it is also possible that GPCMV does not need to block PKR activation, as has been reported in a few other viral systems (30, 31). Our studies reveal that rTRS1 is absolutely essential for RhCMV replication in HF. However, RhCMV[ΔrT] did replicate to a low level in RF (Fig. 2), raising the possibility that RhCMV may have a second, weak PKR antagonist that functions only in rhesus cells.

All known CMV PKR antagonists bind to dsRNA. Thus, the relative amounts and kinetics of production of dsRNA could be a determinant of the ability of the virus to replicate. Studies of VACV revealed that deficiencies in RNA degradation can increase the abundance of dsRNA and thereby activate the PKR pathway (32, 33). Conversely, monkeypox virus produces only small amounts of dsRNA and thereby obviates the need for a PKR antagonist (24). Also, a recent report suggested that dsRNA degradation by the Ambystoma tigrinum virus RNase III gene might contribute to evading the PKR pathway by limiting the amount of dsRNA present in infected cells (34). However, our data show that RhCMV produces more dsRNA than does HCMV throughout infection (Fig. 6A). At late times, we estimate that dsRNA may constitute about 0.05% of the whole cell RNA, although a caveat to this estimate is that some fraction of this dsRNA might have been formed by the annealing of complementary ssRNAs during the RNA purification steps. Regardless, our results argue against the possibility that RhCMV uses the strategy of making less dsRNA as a way to supplement the weak activity of rTRS1 and enable replication in HF.

In addition to binding to dsRNA, hTRS1 needs to bind to PKR to block its activity. Direct binding is a feature shared with PKR antagonists encoded by other CMVs, as well as several unrelated viruses (35–37). The species specificity of PKR binding by CMV proteins correlates with their ability to inhibit PKR function. For example, hTRS1 binds well to human PKR but only very weakly to rhesus or Agm PKR in yeast two-hybrid assays and in cells infected with VACV expressing hTRS1, and it inhibits human PKR but not either Old World monkey PKR (16, 38). Likewise, AgmCMV TRS1 binds to and inhibits African green monkey PKR but not human PKR. Moreover, a single amino acid substitution of a codon found in the Agm PKR αG helix into human PKR prevents hTRS1 from binding to and inhibiting PKR (38). Intriguingly, TRS1 from squirrel monkey CMV, the only New World monkey CMV that has been studied, binds to and inhibits both human PKR and Agm PKR.

Despite their more distant evolutionary relationship to HCMV, rodent CMVs encode PKR antagonists that can inhibit human PKR, and at least in the case of MCMV, the proteins have been shown to bind to human PKR. Experimental adaptation of MCMV to HF selected for mutations in several genes, but not in m142 or m143, suggesting that these MCMV factors are able to inhibit human PKR during MCMV infection (39). Also, MCMV in which hTRS1 has been substituted for m142 and m143 is capable of replicating in mouse cells (40), further suggesting there has been a surprising level of conservation of the critical interaction between PKR and its CMV antagonists between rodents and humans that is not the case between Old World monkeys and humans.

The situation with rTRS1 and rhesus PKR is more enigmatic. Despite its coevolutionary association with rhesus macaques, rTRS1 does not bind well to rhesus PKR in yeast two-hybrid assays, and it does not rescue VACVΔE3L replication in RF, although it does bind to and inhibit PKR in African green monkey cells (16, 38). The observation that rTRS1 binds PKR-P, i.e., after the autophosphorylation step, may account for its failure to bind to African green monkey and other PKRs in yeast two-hybrid assays, which used kinase-dead PKR mutants. Regardless, these results raise the conundrum that rTRS1 seems unable to bind to or antagonize human PKR, or even rhesus PKR, in yeast- or VACV-based assays and yet in the context of RhCMV rTRS1 it is critical for blocking PKR function in both human and rhesus cells.

Several lines of evidence suggest that the solution to this puzzle is that abundant early rTRS1 expression enables it to compensate for its weak binding to PKR. While rTRS1 expression either in the context of HCMV[rT] infection or from a cellular transgene was insufficient to support replication, combined expression from both sources is sufficient for nearly wt levels of replication (Fig. 5B). These results are consistent with a prior study in which we found that either amplification of the rTRS1 gene or expression of rTRS1 from a cellular transgene increased replication of a VACV recombinant containing rTRS1 but lacking E3L and K3L (41). We have also reported on a similar phenomenon in which in the absence of E3L, gene amplification and overexpression of VACV K3L, a very weak antagonist of human PKR, facilitated its ability to counteract PKR (42).

Complementing these heterologous and overexpression contexts, analyses of cells infected with RhCMV and HCMV revealed that rTRS1 is expressed at an elevated level compared to hTRS1 and hIRS1. Thus, during RhCMV infection, even the relatively small increase in expression of this intrinsically weak antagonist of human PKR may account for efficient replication. Since the interaction between rTRS1 and PKR is a bimolecular reaction, relatively small changes in the abundance of rTRS1 may have nonlinear effects. In support of this idea, our yeast two-hybrid assays showed that rTRS1 does not bind well to human or rhesus PKR (16), but in vectors that express higher levels of rTRS1 and PKR, we did detect binding (data not shown).

We do not yet know the mechanism accounting for the increased expression of rTRS1 by RhCMV. One possibility is that the native rTRS1 promoter is stronger than the hTRS1 promoter, which drives expression of IRS1 and TRS1 in HCMV and of rTRS1 in HCMV[rT]. Another possibility is that RhCMV replicates more rapidly than HCMV, and thus, at any given time point, RhCMV is further along in its replication cycle and therefore accumulates rTRS1 and other proteins expressed faster than HCMV. In fact, at 5 days postinfection, RhCMV plaques are similar in size to HCMV plaques after 7 days of infection (data not shown). The stoichiometry and kinetics of accumulation of dsRNA, PKR antagonists, and PKR, as well as the strength of protein binding interactions, all contribute to the outcome of RhCMV infection in HF.

Along with our previous studies demonstrating that VACV can gain a replication advantage by employing gene amplification as a means of increasing the expression of a weak PKR antagonist (in the form of either VACV K3L or rTRS1), these results demonstrate that increased expression of a viral factor that is on the “offense” in an evolutionary arms race can compensate for weak interactions with its targeted host restriction factor (41–43). In this case, increased rTRS1 expression is likely a result of high-level expression from a single-copy gene. However, because the US22 and other gene families in CMV arose from gene duplication events, CMVs might also utilize gene amplification as a means of gaining a foothold against a resistant host restriction factor. While subsequent adaptations may enable collapse of the amplified locus, consistent with the “accordion” model (42), an alternative outcome is neofunctionalization of the amplified copies, leading to gene families in which only one or two members retain the original function. Whether PKR antagonism was the driving force in the amplification events leading to the US22 gene family is unknown, but the fact that the only two essential genes in the MCMV US22 family are the PKR antagonists indicates that this might be the case. Regardless of the means, these results suggest that increased expression of a weak viral antagonist may be a generalizable mechanism allowing viruses to gain a foothold that facilitates their ability to cross species barriers.

MATERIALS AND METHODS

Cells.

All cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% NuSerum (BD Biosciences). HF were obtained from Denise Galloway (Fred Hutchinson Cancer Research Center), and telomerase-immortalized RF were obtained from Peter Barry (University of California, Davis) (44). HF and RF expressing rTRS1 were produced by transduction of a lentiviral vector (pLHCX) containing an rTRS1 transgene, followed by selection in 100 μg/ml hygromycin B. HF^[PKR-ko] were constructed by transducing HF with the lentiviral vector pCW-Cas9 (a gift from Eric Lander and David Sabatini; Addgene plasmid 50661) containing a gRNA with the genomic PKR target sequence 5′-CCTACCTCCTATCATGTGG-3′. Following selection with 1 μg/ml puromycin and isolation of a clonal population, Sanger sequencing and sequence trace decomposition analysis (20) identified a 1-bp deletion (underlined in the target sequence), suggesting that the deletion is present in both PKR alleles or that one allele has a large deletion undetectable by our analysis methods. These cells were immortalized by transduction with a retroviral vector expressing human telomerase (pLXSN-htert) (45).

Viruses and BAC recombinant viruses.

HCMV (strain AD-169; ATCC VR-538) was propagated on HF, and RhCMV (strain 68-1; ATCC VR-677; obtained from Peter Barry, University of California, Davis) was reconstituted from a BAC as described below and propagated on RF. HCMV lacking both IRS1 and TRS1 (HCMV[ΔI/ΔT]) has been described previously (8). HCMV[rT] was produced by introducing rTRS1 into an HCMV[ΔI/ΔT] BAC as described previously (14). GFP-containing HCMV[hT], HCMV[rT], and HCMV[ΔI/ΔT] BACs were produced by introduction of an Amp-loxP-GFP cassette that allows Amp to be removed along with the BAC plasmid sequences by cotransfection with a Cre expression plasmid during reconstitution of the virus. HCMV[ΔI/ΔT]-GFP was reconstituted in the absence of Cre, so the virus still contains the BAC plasmid sequences, in addition to the GFP cassette.

The RhCMV[ΔrT] mutant was generated by deleting the Rh230 open reading frame (ORF) from the well-characterized RhCMV 68-1 BAC (46) using homologous recombination in Escherichia coli strain EL250 as described previously (47). Briefly, primers containing 50 bp of homology to regions flanking the Rh230 ORF, as well as sequences for amplification of a kanamycin resistance (Kan^r) cassette flanked by FLP recombination target (FRT) sites (forward primer, 5′-AAAGATTGGAATGAAGTGTGGACAAGTGTGTGGGCTTGTAGTTATAGTCTGTAAAACGACGGCCAGT-3′, and reverse primer, 5′-CACTCGTCCCCGTGTTCCTCACCACACCTCCACTCCACTGCTACTTAAACGAAACAGCTATGACCATG-3′) were used for PCR with plasmid pCP015 (48) as a template, and the resulting product was electroporated into E. coli strain EL250 containing the RhCMV 68-1 BAC for in vitro homologous recombination. Deletion mutants were selected on chloramphenicol/kanamycin selection plates. Single bacterial colonies were screened by restriction digestion with XmaI (NEB) and compared to in silico predictions. The Kan^r selection cassette was then deleted by inducing expression of FLP recombinase as described previously (47). The final constructs were analyzed by next-generation sequencing of the entire BAC using an Illumina MiSeq next-generation sequencer and Geneious 8.1.4 software for analysis. The virus was reconstituted by transfecting purified BAC DNA into RF expressing rTRS1 (RF+rTRS1) and was later plaque purified three times on HF^[PKR-ko]. Subsequent stocks of RhCMV[ΔrT] were produced in HF^[PKR-ko]. The reconstituted viral genome was also sequenced using the Illumina MiSeq platform as described above.

Growth curves were performed by collecting the supernatants from cells infected with the indicated viruses at the specified times, and then titers were determined on HF^[PKR-ko]. For each growth curve, cells were infected, and at 1 h postinfection, the cells were washed 3 times with PBS, after which the medium was replaced. For day 0 samples, the medium was collected after replacing the medium, and the cells were again refed. Determination of the amount of intracellular virus was performed in the same way, except that the infected cells were collected, resuspended in 1 ml of medium, and freeze-thawed three times to release virus prior to determining titers. GFP-expressing viruses were visualized, and images were captured at ×4 magnification on an Evos FL imaging system (ThermoFisher Scientific). ISRIB (Sigma) was added to the cells at 1 h postinfection.

VACVΔE3L (49) was obtained from Bertram Jacobs (Arizona State University), and VACV (VC2 expressing lacZ) was constructed as previously described (16). β-Galactosidase (β-Gal) activity in infected cells was measured by a fluorometric substrate cleavage assay (26).

Protein immunoblot analyses.

Samples for immunoblotting were prepared by lysing cells with 2% SDS. The lysates were sonicated in a bath sonicator to disrupt nuclear DNA, and then proteins were separated by SDS-polyacrylamide gel electrophoresis on gels containing 0.5% 2,2,2-trichloroethanol to allow stain-free fluorescent visualization of proteins (50), transferred to a polyvinylidene difluoride (PVDF) membrane (Millipore), and probed with the indicated antibodies using the Western Star chemiluminescent detection system (Applied Biosystems) according to the manufacturer's recommendations. The antibodies used in these experiments included eIF2α L57A5 (number 2103), phospho-eIF2α Ser51 (number 3597), and PKR D7F7 (number 12297), all from Cell Signaling Technology, as well as phospho-PKR T446 (ab32026; AbCam), actin (A2066; Sigma), HCMV pp65 and ICP36 (UL44) (both from Virusys Corp.), RhCMV pp65b (51), and RhCMV UL44 (52). Polyclonal rabbit antiserum that recognizes the HCMV TRS1/IRS1 dsRNA-binding domains, as well as the RhCMV TRS1 dsRNA-binding domain (α999), has been described previously (8). All the purchased antibodies were used according to the manufacturers' recommendations. Immunoblot images were captured and quantified with a ChemiDoc Touch imaging system and Image Lab software (Bio-Rad Laboratories, Hercules, CA).

dsRNA quantification.

HF were mock infected or infected with RhCMV or HCMV (multiplicity of infection [MOI] = 3). At various times postinfection, RNA was purified from the infected cells using TRIzol reagent (Life Technologies) according to the manufacturer's instructions. RNA concentrations were determined with a NanoDrop 2000 (ThermoFisher). A total of 200 ng of each sample was spotted onto a Hybond N⁺ nylon membrane using a Bio-Dot apparatus (Bio-Rad) and UV cross-linked using a Stratalinker 1800 (Stratagene). To determine whether ssRNA was causing any background signal, the same amount of RNA was digested with RNase A (Sigma) and RNase T1 (Thermo Scientific) in RNase A buffer (10 mM Tris-Cl, pH 7.5, 5 mM EDTA, 300 mM NaCl) for 20 min at 37°C prior to spotting onto a Hybond N⁺ membrane. The membrane was probed for dsRNA using monoclonal antibody 9D5 (10) (obtained from Shigeo Yagi, California Department of Public Health) using a Western Star chemiluminescent detection system (Applied Biosystems), and dsRNA quantities were calculated using Image Lab software as described above. Reovirus RNA was used as a standard, with some samples treated with RNase as described above.

Quantification of hTRS1 and rTRS1 by immunoblot assay.

Recombinant TRS1 protein was purified from a baculovirus vector as previously described (25). 6×His-tagged hTRS1 and rTRS1 expressed from VACV recombinants were purified from infected cell lysates using Ni-nitrilotriacetic acid (NTA) resin (5 Prime Inc.). Dilutions of each protein were run on an SDS-polyacrylamide gel, and total protein was visualized by stain-free fluorescence as described above. The proteins were then transferred to a PVDF membrane and probed with the TRS1 dsRNA-binding domain antiserum. Images were captured and band intensities from the immunoblot and gel were calculated as described above in order to compare the efficiencies of binding of the antiserum to hTRS1 and rTRS1.

Statistical analysis.

All statistical analyses were performed using an unpaired, two-tailed t test. If unequal variances were observed for unpaired sample sets (F test for unequal variance), an unpaired t test with Welch's correction was performed. Average differences between the means that were less than 10-fold were not considered biologically relevant. Statistical analyses were performed using Prism 7 software (GraphPad).