Experimental Evolution Identifies Vaccinia Virus Mutations in A24R and A35R That Antagonize the Protein Kinase R Pathway and Accompany Collapse of an Extragenic Gene Amplification

ABSTRACT

Most new human infectious diseases emerge from cross-species pathogen transmissions; however, it is not clear how viruses adapt to productively infect new hosts. Host restriction factors represent one species-specific barrier that viruses may initially have little ability to inhibit in new hosts. For example, viral antagonists of protein kinase R (PKR) vary in their ability to block PKR-mediated inhibition of viral replication, in part due to PKR allelic variation between species. We previously reported that amplification of a weak PKR antagonist encoded by rhesus cytomegalovirus, rhtrs1, improved replication of a recombinant poxvirus (VVΔEΔK+RhTRS1) in several resistant primate cells. To test whether amplification increases the opportunity for mutations that improve virus replication with only a single copy of rhtrs1 to evolve, we passaged rhtrs1-amplified viruses in semipermissive primate cells. After passage, we isolated two viruses that contained only a single copy of rhtrs1 yet replicated as well as the amplified virus. Surprisingly, rhtrs1 was not mutated in these viruses; instead, we identified mutations in two vaccinia virus (VACV) genes, A24R and A35R, either of which was sufficient to improve VVΔEΔK+RhTRS1 replication. Neither of these genes has previously been implicated in PKR antagonism. Furthermore, the mutation in A24R, but not A35R, increased resistance to the antipoxviral drug isatin-β-thiosemicarbazone, suggesting that these mutations employ different mechanisms to evade PKR. This study supports our hypothesis that gene amplification may provide a “molecular foothold,” broadly improving replication to facilitate rapid adaptation, while subsequent mutations maintain this efficient replication in the new host without requiring gene amplification.

IMPORTANCE Understanding how viruses adapt to a new host may help identify viruses poised to cross species barriers before an outbreak occurs. Amplification of rhtrs1, a weak viral antagonist of the host antiviral protein PKR, enabled a recombinant vaccinia virus to replicate in resistant cells from humans and other primates. After serial passage of rhtrs1-amplified viruses, there arose in two vaccinia virus genes mutations that improved viral replication without requiring rhtrs1 amplification. Neither of these genes has previously been associated with inhibition of the PKR pathway. These data suggest that gene amplification can improve viral replication in a resistant host species and facilitate the emergence of novel adaptations that maintain the foothold needed for continued replication and spread in the new host.

INTRODUCTION

Of the approximately 1,400 known human pathogens, over 60% are zoonotic (1). Furthermore, during the last 80 years, more than 70% of all new and emerging human infectious diseases were acquired from animal sources (2). These emerging zoonoses are often responsible for severe disease in humans, such as HIV-1 infection/AIDS (3) and the recent Ebola pandemic in West Africa (4). Similarly, pathogen transmission between both wild and domestic animals can lead to viral adaptation in species that may serve as intermediate hosts and increase the risk of transmission to humans, as exemplified by the influenza transmission cycle between water fowl, poultry, and swine (5). Because human health and animal health are inextricably linked, as articulated by the One Health concept (6), the pool of potential new pathogens is large and the challenge of detecting emerging pathogens before they lead to epidemics is daunting. Therefore, identifying molecular signatures that may pinpoint pathogens poised to cross species barriers and productively infect new hosts is a pressing need for both human and animal health.

One of the primary barriers to the cross-species transmission of viruses is a group of antiviral proteins collectively known as host restriction factors (7). One of these factors, protein kinase R (PKR) inhibits viral replication by detecting double-stranded RNA (dsRNA), a common by-product of both RNA and DNA virus replication, and subsequently inhibiting protein synthesis (8). In response to this broad antiviral activity, many viruses have evolved at least one mechanism to inhibit the PKR pathway (9).

Vaccinia virus (VACV), the prototypical member of the Orthopoxvirus genus in the family Poxviridae, is a double-stranded DNA virus that replicates in the cytoplasm and has a broad host range. Like most orthopoxviruses, VACV cellular tropism is believed to be regulated by intracellular events rather than at the level of binding and entry (10). VACV encodes several host range genes, including two PKR antagonists: E3L and K3L. E3L antagonizes PKR both by sequestering dsRNA and by directly binding PKR to inhibit autophosphorylation. K3L has considerable homology to eIF2α, from which it is presumed to have evolved following an ancient gene acquisition event. K3L acts as a pseudosubstrate of PKR that competitively inhibits eIF2α phosphorylation. Individually, each of these host range genes is required for replication in some cell types and dispensable in others, suggesting that these two genes have evolved to permit viral replication in a broad host range (11).

PKR itself has been evolving rapidly during primate evolution, likely in order to evade viral antagonists (12, 13). The resulting variation in PKR can confer species-specific resistance to viral antagonists (12, 14). Therefore, cross-species transmission may require a viral antagonist to rapidly adapt to inhibit new variants of PKR under conditions that may permit very little replication. While the mechanism of PKR-mediated inhibition of viral replication has been well characterized, much less is known about the evolutionary processes enabling viruses to overcome PKR in new host species.

We previously reported that genetic amplification of rhtrs1, a rhesus cytomegalovirus gene encoding a weak antagonist of some African green monkey PKR alleles, substantially improves replication of a VACV recombinant with its native PKR antagonists deleted (VVΔEΔK+RhTRS1) in cell lines derived from a broad range of primate species (15). The precise mechanism of PKR antagonism by RhTRS1 is still unclear; however, it is known to inhibit the PKR pathway at a step after PKR autophosphorylation (14), and this phenotype was maintained in virus populations containing rhtrs1 amplifications (15). Gene amplification is a universal mechanism of rapid adaptation in eukaryotes (16, 17), prokaryotes (18, 19), and viruses (15, 20–22), enabling diverse adaptations such as neofunctionalization, antibiotic resistance, and evasion of host restriction factors (reviewed in reference 23). In general, amplification of a gene with weak activity can increase the fitness of an organism through overexpression related to gene dosage effects. In addition, the increase in gene frequency in the population, through both genetic amplification and increased population size, increases the potential for an adaptive mutation to evolve without altering the underlying mutation rate of the organism. Once an adaptive mutation evolves, the selective pressure to maintain the amplification may be relaxed, permitting the collapse of the amplified locus (24). Therefore, we hypothesized that gene amplification may act as a “molecular foothold” to facilitate cross-species transmission.

In this study, we report the emergence of two adaptive mutations in VACV genes A24R and A35R, which arose upon continued experimental evolution of a pool of viruses with amplification of rhtrs1 (15). Notably, these mutations improved replication even in the absence of rhtrs1 amplification, and they arose outside the rhtrs1 locus. Although they appear to function through different mechanisms, either one of these VACV mutations is sufficient to fully rescue VVΔEΔK+RhTRS1 replication in semipermissive African green monkey (AGM) cells and to improve VVΔEΔK+RhTRS1 replication in cells from more-distant primate species. Taken together, these studies suggest a model of cross-species transmission in which transient gene amplification of a weak viral antagonist may act to broadly increase viral species tropism. Subsequent adaptive mutations may evolve not only within the amplified gene (21) but also in extragenic loci that may reveal new host range determinants and virulence factors.

MATERIALS AND METHODS

Viruses and cell culture.

AGM fibroblasts (PRO1190; Coriell Institute for Medical Research) were maintained in minimal essential medium (MEM; Gibco) supplemented with 20% fetal bovine serum (Sigma-Aldrich) and antibiotics (100 U/ml penicillin and 100 μg/ml streptomycin; Gibco). Human foreskin fibroblasts (HFF), rhesus fibroblasts (RF), and BSC40 cells were maintained in Dulbecco's modified Eagle's medium (DMEM; Gibco) supplemented with 10% NuSerum (BD Biosciences) as previously described (25).

VVΔE3L (26) (provided by Bertram Jacobs, Arizona State University), VVΔEΔK+RhTRS1 (13), and VV-βg (VC2-LacZ in reference 14) were propagated, and their titers were determined, in BSC40 cells. VV-A, VV-B, and VV-C (15) were propagated in PRO1190 cells, and their titers were determined in BSC40 cells. All titrations were performed in triplicate. Graphs and 95% confidence intervals were generated in GraphPad Prism 6 (GraphPad software).

Experimental evolution of VV-A, VV-B, and VV-C.

In the initial round of serial passage, confluent 6-well plates of PRO1190 cells were infected with serial 10-fold dilutions of each virus to ensure that individual, well-isolated plaques were obtained. Two days postinfection, 3 plaques were picked from each of the three virus populations (VV-A, VV-B, and VV-C [15]) by cell scraping, resuspended in 100 μl of MEM–20% fetal bovine serum (FBS), and lysed by three freeze-thaw cycles. In subsequent rounds of serial passage, confluent 6-well plates of PRO1190 cells were infected with 1 μl and 10 μl of the preceding plaque lysate to ensure individual, well-isolated plaques. Plaques were picked and lysed by following the above-described methodology. One plaque lysate did not generate any plaques in the third round of plaque purification and was therefore discarded. After three rounds of plaque purification, confluent 6-well plates of PRO1190 were infected with 10 μl of the final plaque lysate/well and incubated at 37°C until the entire well showed cytopathic effects. Cells were collected, pelleted, and resuspended in 1 ml MEM–20% FBS. After three freeze-thaw cycles, virus titers were determined on BSC40 by plaque assays. For the final round, two confluent 10-cm dishes of PRO1190 cell were infected with each of the viruses at a multiplicity of infection (MOI) of 0.1 for DNA isolation and virus production. VACV DNA was purified from infected cell cytoplasmic extracts for genetic and genomic analyses as described previously (27).

Genomic analyses. (i) Library preparation and sequencing.

From each viral pool, 75 ng of viral DNA was sheared and ligated to adaptors by transposition using the Nextera kit (Epicentre) in accordance with the manufacturer's directions, with several modifications: transposition was performed in a 20-μl volume using 0.1 μl Nextera transposase enzyme and “HMW” reaction buffer at a final 1× concentration. Transposition was carried out for 10 min at 55°C, after which sheared, adaptor-ligated templates were transferred directly to PCR mixtures for amplification. In addition, PCR incorporated a sample-specific barcode tag on the reverse primer. Amplified libraries were pooled and cleaned using Ampure beads (Beckman Coulter) at a 1.8:1 ratio of beads to input. The pooled libraries were sequenced on an Illumina HiSeq 2000 instrument (Illumina) using 101-bp forward and reverse reads with a 9-bp index to read the per-sample barcode. Sequences were deposited in the Short Read Archive under accession no. SRP033208.

(ii) Sequence analysis.

Reads from each evolved viral pool were aligned to the VVΔEΔK+RhTRS1 reference genome (15), and single-base and short-indel variants were called using the Genome Analysis Toolkit (28). Custom scripts were used to determine the allele frequency (the fraction of reads supporting variant allele of all aligned reads at a given site) for all variants found in any sample. Copy number was estimated by counting the depth of read coverage within sliding windows of 100 bp, correcting for effects of differences in G+C composition, and dividing by the parental strain read depth. Everted reads at duplication breakpoints were identified using bwasw (29) and custom scripts, as previously described (21). Breakpoint detection was performed by searching for two sequences specific to the amplifications that we had previously identified (15), requiring a match over the entire central 20mer centered around the breakpoint: J2R-L5R, 5′-TAGTAGCCGCACTCGATGGGACATTTCAACGTAAACCGTTTAATAATATTTTGAATCTTATTCCATTATCTGAAATGGTGGTAAAACTAACTGCTGTGTGATGCCTAAACGAAAAATACCCGATCCTATTGATAGATTACGACGTGCTAATCTAGCGTGTGAAGACGATAAATTAATGATCTATGGATTACCATGGATGA-3′; J2R-NeoR, 5′-GCGTATGGCAAACGAAGGAAAAATAGTTATAGTAGCCGCACTCGATGGGACATTTCAACGTAAACCGTTTAATAATATTTTGAATCTTATTCCATTATCTAGCCGGCCACAGTCGATGAATCCAGAAAAGCGGCCATTTTCCACCATGATATTCGGCAAGCAGGCATCGCCATGGGTCACGACGAGATCCTCGCCGTC-3′.

Southern blot analysis.

Approximately 0.5 μg of DNA isolated from each virus was digested with BglII and separated on a 0.5% agarose gel, and the DNA was transferred to a supported nitrocellulose membrane using standard transfer techniques. The membrane was probed with a fragment of rhtrs1 generated by digesting plasmid pEQ1215 (14) with BamHI and EcoRI. The resulting 1.6-kb central fragment of rhtrs1 was gel isolated and ³²P-radiolabeled by random priming.

Construction of VV-A24R* and VV-A35R*.

Total viral DNA from VV-A1 was prepared as previously described (27). A24R* was PCR amplified from VV-A1 viral DNA using oligonucleotides 989, 5′-GACAAGCGAATGAGGGACGA, and 990, 5′-CCACCAAAACCTGAAACTCC. A35R* was PCR amplified using oligonucleotides 993, 5′-GTCGTTAACGTACGCCGCCA, and 994, 5′-CCAGCATCATTTCTGATTGC. The amplification products were gel purified and cloned into pSC-A using the StrataClone PCR cloning kit (Agilent). Hygromycin B phosphotransferase (HygR) was PCR amplified from pLHCX (Clontech) using oligonucleotides 1009, 5′-GTCACCCGGGTCTAGACTATTCCTTTGCCCTCGGAC, and 1011, 5′-GTCACCCGGGATGGATAGATCCGGAAAGCC. This amplification product was digested with XmaI, gel purified, and cloned into the XmaI site in pSC11 (30). The resulting plasmid was digested with XbaI, and the fragment containing the 7.5K promoter and HygR was gel purified and cloned into the XbaI site of the A24R*- and A35R*-containing plasmids (pEQ1422 and pEQ1423, respectively).

BSC40 cells were infected with VVΔEΔK+RhTRS1 (MOI = 1.0). One hour postinfection, these cells were transfected with pEQ1422 or pEQ1423 using Lipofectamine 2000 (Life Technologies) according to the manufacturer's protocol. Two days postinfection, infected cells were harvested and lysed by three freeze-thaw cycles. These stocks were passaged twice through BSC40 cells and then were subjected to three rounds of plaque purification in the presence of 600 μg/ml hygromycin B, followed by three rounds of plaque purification without hygromycin B. The loss of HygR was confirmed by PCR using oligonucleotides 1009 and 1011, and the presence of A24R* and A35R* was confirmed by Sanger sequencing-purified viral DNA. The viruses were propagated and their titers were determined on BSC40 cells. Externally directed PCR was performed as previously described (15) to confirm that the recombinant virus populations in these stocks contained a single copy of rhtrs1.

Immunoblot analyses.

Cells were mock infected or infected with vaccinia viruses (MOI = 3) and lysed 1 day postinfection in 2% sodium dodecyl sulfate (SDS). Equivalent amounts of the lysates were separated on 10% SDS-polyacrylamide gels, transferred to polyvinylidene difluoride (PVDF) membranes, and probed with one of the following antibodies: anti-PKR (phospho T446) antibody, E120 (ab32036; Abcam), eIF2α (9722; Cell Signaling Technology), phospho-eIF2α (Ser51) antibody (9721; Cell Signaling Technology), or actin (A2066; Sigma). All purchased antibodies were used according to the manufacturer's recommendations. Proteins were detected using the Western Star chemiluminescent detection system (Applied Biosystems) according to the manufacturer's recommendations.

Metabolic labeling.

PRO1190 cells were mock infected or infected with the indicated viruses (MOI = 3). One day postinfection, the cells were labeled for 1 h with 100-μCi/ml l-[³⁵S]methionine/l-[³⁵S]cysteine (EasyTag express protein labeling mix; PerkinElmer) in medium lacking methionine and cysteine and then lysed in 2% SDS. Equivalent amounts of protein from each sample were separated on 10% SDS-polyacrylamide gels, dried, and visualized by autoradiography.

IBT assays.

BSC40 or PRO1190 cells were infected with the indicated viruses (MOI = 0.1). One hour postinfection, virus-containing medium was aspirated from the cells and replaced with fresh medium or medium containing 45 μM isatin-β-thiosemicarbazone (IBT; Pfaltz and Bauer, Inc., kindly provided by Richard Condit), prepared as previously described (31). Two days postinfection, the medium was replaced with 1 ml of fresh medium and cells were frozen at −80°C. After three freeze-thaw cycles, virus titers were determined on BSC40 by plaque assays.

RESULTS

Collapse of the rhtrs1 amplification following passage under continuous selection.

We previously demonstrated that gene amplification of the weak PKR antagonist rhtrs1 is sufficient to rescue replication of the chimeric virus VVΔEΔK+RhTRS1 in both semipermissive African green monkey (AGM) fibroblasts (PRO1190) and nonpermissive human foreskin fibroblasts (HFF) and rhesus fibroblasts (RF) (15). In addition to the replication benefit provided through gene dosage effects, gene amplification is thought to increase the rate of genetic change in an organism without changing the underlying mutation rate by increasing the number of gene copies that are independently evolving in the population (32). Therefore, we hypothesized that adaptive mutations may evolve in the rhtrs1-amplified viruses that could permit a virus with a single copy of rhtrs1 to replicate as efficiently as viruses with multiple copies of the gene. To test this hypothesis, we continued to passage previously described virus pools VV-A, VV-B, and VV-C (15), which have amplifications of the rhtrs1 locus, in semipermissive PRO1190 cells. We plaque purified individual virus isolates during each of three rounds of serial passage and then expanded the virus stocks by two additional bulk passages in these cells.

To assess the replication properties of these plaque-purified viruses, we infected PRO1190 cells with VV-βg, VVΔEΔK+RhTRS1, or the eight plaque-purified virus isolates (MOI = 0.1). Six of these passaged viruses replicated much more efficiently than the initial virus (Fig. 1A). We next analyzed the genomic structure of the rhtrs1 locus in these plaque-purified viruses by Southern blotting to determine whether any of these isolates had undergone contraction of the rhtrs1 locus (Fig. 1B). Four isolates, VV-A2, VV-B2, VV-B3, and VV-C1, retained amplification of the rhtrs1 locus and, like the starting pools, replicated substantially better than VVΔEΔK+RhTRS1. Notably, none of these viruses contained discrete copy numbers of rhtrs1, despite their having been plaque purified three times. Instead, the rhtrs1 locus in each of these isolates was heterogeneous in size, in some cases encompassing a substantial percentage of the viral genome, although the majority of viruses in each population contained only one or two copies of rhtrs1. This result is consistent with gene amplification and contraction being a highly dynamic process in which a viral genome possessing a rhtrs1 amplification likely can produce progeny with a range of gene copy number variants.

FIG 1.

Replication of plaque-purified viruses and rhtrs1 locus analysis. Two or three viruses were serially plaque purified from rhtrs1-amplified virus pools VV-A, VV-B, and VV-C. (A) Titers of indicated viruses 48 h postinfection of PRO1190 cells (MOI = 0.1) were measured in BSC40 cells and are represented as means + 95% confidence intervals. (B) Southern blot of purified viral DNA digested with BglII and probed with a ³²P-labeled rhtrs1-specific probe. Numbers on the left are kilodaltons.

We also identified four viruses, at least one from each of the initial populations, in which the rhtrs1 locus had collapsed to a single copy (Fig. 1B). Two of these, VV-B1 and VV-C2, carried a single copy of rhtrs1 and replicated no better than the initial replication-restricted virus, VVΔEΔK+RhTRS1. Most notably, two of the viruses with a single copy of rhtrs1, VV-A1 and VV-A3, replicated over 100-fold better than VVΔEΔK+RhTRS1, suggesting that they had evolved an adaptive mutation that allowed for collapse of the rhtrs1 locus.

Genomic analyses of plaque-purified viruses.

To identify any mutations that might account for the efficient replication of VV-A1 and VV-A3, we performed paired-end Illumina-based deep sequencing on viral DNA isolated from all eight plaque-purified viruses. Consistent with our Southern blotting results (Fig. 1B), VV-B2, VV-B3, and VV-C1 carried an average of two copies of rhtrs1 by read depth analysis, a result similar to our previous analyses of VV-A, VV-B, and VV-C (15). In contrast, plaque-purified isolates VV-A1, VV-A3, VV-B1, and VV-C2 all had only a single copy of rhtrs1 (Fig. 2B). VV-A2 had only a slight elevation in average rhtrs1 copy number (Fig. 2B), even though our Southern blot analysis identified some genomes in this population with multiple copies.

FIG 2.

rhtrs1 copy number variation in plaque-purified viruses. (A) Diagram of the genomic region around rhtrs1 in VΔEΔK+RhTRS1 (top) and the predominant recombination sites J2R-L5R (middle) and J2R-NeoR (bottom). Gene names are indicated below the upper panel. VACV genes are displayed in white, gray, or black; J2R-L and J2R-R denote the 5′ and 3′ fragments of the gene generated during insertion of the rhtrs1-containing cassette into this locus. Exogenous genes are colored blue (RhTRS1) or purple (NeoR). Red arrowheads denote the locations of the sites involved in the recombination events, and the nucleotide positions relative to the VVΔEΔK+RhTRS1 reference genome (15) are noted above each genome. Gray boxes outline the tandem duplication that generates each respective breakpoint. (B) Relative read depth of Illumina sequencing reads from the plaque-purified viruses normalized to VVΔEΔK+RhTRS1. The vertical gray bar highlights the rhtrs1 transgene locus. (C) Copy number of rhtrs1 (rescaled so that the value for VVΔEΔK+RhTRS1 strain is set as 1) plotted against the J2R-L5R breakpoint read frequency (per million mapped reads). The lower left cluster of dots includes VV-A1, VV-A2, VV-A3, VV-B1, and VV-C2 as well as VV-A after one passage in BSC40 cells.

As a complementary assay for amplification, we examined the viral DNA for breakpoint spanning reads specific to the two primary recombination sites that we previously identified (Fig. 2A; see also reference 15). The predominant breakpoint in that study (J2R-L5R) resulted from a recombination event that occurred after nucleotide (nt) 84267 in J2R and before nt 83200 in L5R, and the minor breakpoint (J2R-NeoR) was due to a recombination event after nt 84237 in J2R and before nt 622 of the neoR gene derived from pcDNA3.1, which was inserted into the initial virus as a selectable marker. We detected primarily the J2R-L5R breakpoint in VV-B2, VV-B3, and VV-C1 (Table 1; Fig. 2C). Although we identified a few J2R-L5R breakpoint reads in VV-A2, the majority of the breakpoints in this virus were the J2R-NeoR breakpoint. The latter breakpoint is predicted to generate a smaller fragment after BglII digestion, consistent with our Southern blot results (Fig. 1B). In addition, these breakpoints were present at a lower frequency in VV-A2 than breakpoints in other rhtrs1-amplified viruses, in concordance with read depth analysis, indicating that amplified genomes are a minority in the VV-A2 virus. We also detected a very low level of breakpoint reads even in viruses without detectable amplification by Southern blotting, including VV-βg, suggesting that there may be a low level of gene amplification independent of any selection for amplification.

TABLE 1.

Frequency of J2R-L5R and J2R-NeoR spanning breakpoint reads

Virus	Frequency/106 reads		Absolute no. of reads		Total no. of reads
	J2R-L5R breakpointa	J2R-NeoR breakpointa	J2R-L5R breakpoint	J2R-NeoR breakpoint
VV-βg	6.9	0.0	3	0	437,285
VVΔEΔK+RhTRS1	0.0	0.0	0	0	76,904
VV-A1	13.8	0.0	5	0	362,819
VV-A2	24.2	113.9	7	33	289,717
VV-A3	11.3	0.0	10	0	885,005
VV-B1	16.5	0.0	3	0	181,272
VV-B2	880.1	1.7	526	1	597,672
VV-B3	801.0	0.0	511	0	637,957
VV-C1	788.6	0.0	189	0	239,652
VV-C2	12.3	0.0	9	0	734,517

^aDefined in reference 15 and depicted in Fig. 2A.

Surprisingly, we did not identify a single rhtrs1 mutation in any of the eight virus isolates, including VV-A1 and VV-A3, which both replicate efficiently with only a single copy of rhtrs1. Instead, each plaque-purified virus contained between one and three fixed mutations in VACV genes relative to VVΔEΔK+RhTRS1 (Table 2). Two of these genes, A24R and A35R genes, had fixed mutations in both VV-A1 and VV-A3; therefore, we focused on these two genes as the most likely candidates to explain the improved replication of these two viruses.

TABLE 2.

Mutations identified in plaque-purified viruses relative to vaccinia virus (strain Copenhagen) at greater than 5% allele frequency^a

Position	Reference base(s)	Variant base(s)	Gene	Effectb	VV-A1		VV-A2		VV-A3		VV-B1		VV-B2		VV-B3		VV-C1		VV-C2
					Read depth	Allele frequency (%)	Read depth	Allele frequency (%)	Read depth	Allele frequency (%)	Read depth	Allele frequency (%)	Read depth	Allele frequency (%)	Read depth	Allele frequency (%)	Read depth	Allele frequ ency (%)	Read depth	Allele frequency (%)
39202	T	C	F7L	Mis:K34E:aag→Gag	107	0.9	110	6.4	350	1.7	94	3.2	192	3.7	221	2.3	71	0.0	281	3.6
39202	T	TC	F7L	Indel	109	4.6	111	4.5	354	5.9	99	8.1	196	7.7	222	5.0	72	2.8	287	3.8
39203	A	C	F7L	Mis:N33K:aat→aaG	108	8.3	110	11.8	353	9.1	96	16.7	196	13.5	224	9.8	70	5.7	282	9.2
92103	G	T	J6R	Mis:G268C:ggc→Tgc	322	0.0	233	0.0	697	0.1	148	100.0	484	0.0	528	0.0	215	0.0	662	0.0
142469	C	T	A24R	Mis:T1121 M:acg→aTg	285	100.0	206	99.0	624	100.0	139	0.0	475	0.0	475	0.2	213	0.0	561	0.0
144590	AT	A	A29L	Indel	335	0.3	219	0.9	632	0.2	137	0.0	508	0.4	518	0.4	231	0.0	591	88.5
146037	T	TAA	A31R	Indel	303	85.2	196	0.0	621	0.0	164	0.0	487	0.0	475	0.0	202	0.0	555	0.0
148449	TA	T	A35R	Indel	389	87.7	269	0.0	757	0.0	155	0.0	625	0.0	684	0.0	280	0.0	726	0.0
148499	CA	C	A35R	Indel	352	0.0	231	91.8	696	90.4	137	0.0	574	0.0	595	0.2	253	0.0	687	0.2
149912	G	A	A37R	Mis:S116N:agc→aAc	340	0.0	287	0.0	758	0.1	180	100.0	669	99.6	610	99.5	207	0.0	615	0.2
149922	TAC	T	A37R	Indel	345	0.0	294	0.0	808	0.0	184	0.0	649	0.0	577	0.0	208	85.4	662	0.0
150258	C	CA	A37R	Indel	347	0.0	271	0.0	893	0.0	199	0.0	601	0.0	607	0.0	265	0.0	674	79.7
155472	T	TA	N/A	N/A	329	2.1	224	3.6	728	2.8	178	2.8	466	3.2	571	3.5	195	5.6	622	68.4
178360	TA	T	N/A	N/A	239	1.3	165	0.0	453	1.1	101	87.1	363	1.1	393	1.3	150	0.0	446	1.1

^aAllele frequencies >85% are in bold. N/A, not annotated.

^bFor missense (Mis) mutations, the altered amino acids at the indicated codon are shown, followed by the wild-type (lowercase) and mutant (uppercase) nucleotides at that codon.

A24R is the catalytic subunit of the viral RNA polymerase. Mutations in the VACV RNA polymerase have been implicated in reducing dsRNA accumulation (33, 34); however, mutations in this enzyme have not been associated with activity against PKR. Both VV-A1 and VV-A3 have the same C3362T mutation, predicted to encode a nonsynonymous T1120M amino acid change at the C terminus of A24R (Fig. 3A). It should be noted that, while it was not fixed, this mutation was already present at a high allele frequency (73%) in VV-A (15).

FIG 3.

Amino acid alignments of A24R or A35R. (A) Amino acid alignment of the C terminus of the VVΔEΔK+RhTRS1 A24R gene product with the predicted T1120M mutation (gray box) identified in VV-A1 and VV-A3. (B) Amino acid alignment of the VVΔEΔK+RhTRS1 A35R gene product with the predicted A35R truncation mutants resulting from the single-nucleotide deletions TA211T and CA261C identified in VV-A1 and VV-A3 and with the predicted VARV A35R reading frame (GenBank accession number DQ437580). Differences relative to VVΔEΔK+RhTRS1 before the end of the truncated products are highlighted in gray. Differences between the CA261C and TA211T alleles are underlined in the VV-A3 (TA211T) sequence.

Less is known about A35R; however, it is a highly conserved gene in orthopoxviruses and has been implicated in modulation of the adaptive immune response (35–37). The mutations that evolved in A35R during this study were all single-nucleotide deletions. Interestingly, all three initial virus populations from our previous study, VV-A, VV-B and VV-C, evolved at least one A35R single-nucleotide deletion in a fraction of the genomes (15), although only viruses derived from VV-A retained these deletions after plaque purification. VV-A1 and VV-A3 each had a fixed single-nucleotide deletion in A35R but at different positions; VV-A1 has a TA211T mutation, while VV-A3 has a CA261C mutation. Although these deletions are at different locations, they are both predicted to produce a 92-amino-acid (aa) protein, 17 aa of which would vary between the two viruses due to the different positions of the frame-shifting deletions (Fig. 3B).

VV-A1 and VV-A3 inhibit the PKR pathway with a single copy of rhtrs1.

Our previous study demonstrated that amplification of the rhtrs1 locus improved the ability of VVΔEΔK+RhTRS1 to inhibit PKR. Therefore, we investigated whether the improved replication of VV-A1 and VV-A3 was also associated with inhibition of the PKR pathway, even without rhtrs1 amplification. To address this question, we infected PRO1190 cells with VV-βg, VVΔE3L, VVΔEΔK+RhTRS1, VV-A1, VV-A2, or VV-A3 (MOI = 3) and harvested [³⁵S]methionine-radiolabeled cell lysates 24 h postinfection. As expected, cells infected with VV-βg and VV-A2, which retained some rhtrs1-amplified genomes, expressed more VACV-specific protein than cells infected with either VVΔE3L or VVΔEΔK+RhTRS1 (Fig. 4A). Cells infected with either VV-A1 or VV-A3, which contain only a single copy of rhtrs1, also efficiently expressed VACV proteins at levels substantially higher than those of cells infected with VVΔEΔK+RhTRS1 and comparable to those of VV-βg-infected cells.

FIG 4.

Inhibition of the PKR pathway in PRO1190 cells by efficiently replicating single-copy rhtrs1 viruses. Numbers on the left are kilodaltons. (A) PRO1190 cells were infected with the indicated viruses (MOI = 3.0). Twenty-four hours postinfection, the protein synthesis was monitored by [³⁵S]methionine labeling for 1 h, followed by SDS-PAGE separation and autoradiography. (B) The radiolabeled lysates described in the legend for panel A were analyzed by immunoblotting using the indicated antibodies.

We also investigated the activation state of the PKR pathway in these same cell lysates by immunoblot assays. As expected, VV-βg inhibited phosphorylation of both PKR and eIF2α, while these proteins were both phosphorylated to a greater degree in cells infected either with VVΔE3L or with VVΔEΔK+RhTRS1 (Fig. 4B). Similar to the initial rhtrs1-amplified virus VV-A (15), VV-A2 allowed PKR phosphorylation but fully inhibited eIF2α phosphorylation. Interestingly, both VV-A1 and VV-A3 also permitted partial phosphorylation of PKR relative to VVΔE3L but fully inhibited eIF2α phosphorylation. These data support the hypothesis that VV-A1 and VV-A3 evolved mutations sufficient to inhibit the PKR pathway in PRO1190 cells and permit collapse of the amplified rhtrs1 locus.

Mutations in VACV genes A24R and A35R individually improve the ability of VVΔEΔK+RhTRS1 to inhibit PKR.

Neither A24R nor A35R has ever been implicated in PKR antagonism; however, our results suggest that the A24R* and A35R* mutations present in VV-A1 and VV-A3 may explain the improved replication and PKR pathway phenotypes of these viruses. Therefore, we reconstituted the mutations individually in a VVΔEΔK+RhTRS1 background to determine what effect, if any, each mutation had on viral fitness. Using a transient dominant selection strategy (38), we introduced either the TA211T A35R* mutation encoded by VV-A1 or the A24R* mutation that is common to both VV-A1 and VV-A3 into VVΔEΔK+RhTRS1 to generate VV-A35R* and VV-A24R*, respectively. We infected PRO1190 cells with these new viruses or with VV-βg, VVΔEΔK+RhTRS1, the rhtrs1-amplified virus VV-A, or VV-A1 as controls (MOI = 0.1). Consistent with our previous data, VVΔEΔK+RhTRS1 replicated less well than VV-βG while VV-A and VV-A1 replicated almost as well as VV-βg (Fig. 5). VV-A35R* and VV-A24R* each replicated nearly 100-fold more efficiently than VVΔEΔK+RhTRS1, at a level close to that of VV-A and VV-A1. Thus, A35R* or A24R* each provides a replication benefit to VVΔEΔK+RhTRS1, comparable to the replication benefit provided by rhtrs1 amplification in PRO1190 cells (15).

FIG 5.

Mutations in either VACV gene A24R or A35R increase the replication fitness of VVΔEΔK+RhTRS1 in PRO1190 cells. Titers were determined in BSC40 cells 48 h postinfection (MOI = 0.1) and are reported as means + 95% confidence intervals.

To determine whether VV-A24R* and VV-A35R* can individually inhibit PKR, we infected PRO1190 with VV-βG, VVΔE3L, VVΔEΔK+RhTRS1, VV-A1, VV-A35R*, and VV-A24R* (MOI = 3) and collected [³⁵S]methionine-radiolabeled lysates 24 h postinfection. Autoradiography demonstrated that VV-βg-infected cells strongly expressed VACV proteins, while VVΔEΔK+RhTRS1 only weakly expressed VACV-specific proteins and had reduced total protein synthesis relative to uninfected cells (Fig. 6A). Similar to infection with the adapted virus VV-A1, PRO1190 cells infected with either VV-A24R* or VV-A35R* expressed VACV-specific proteins at levels comparable to infection with VV-βg.

FIG 6.

Mutations in either A24R or A35R result in PKR pathway inhibition in PRO1190 cells. PRO1190 cells were infected with the indicated viruses (MOI = 3.0) and labeled 24 h postinfection with [³⁵S]methionine for 1 h. Lysates were harvested and analyzed by autoradiography (A) or by immunoblotting (B) using the indicated antibodies. (C) PRO1190 cells were infected with the indicated viruses (MOI = 3.0). Cellular lysates were harvested at 6 h (top) or 12 h (bottom) postinfection and analyzed by immunoblotting using the indicated antibodies. Numbers on the left of each panel are kilodaltons.

To determine the stage of the PKR pathway that is inhibited by each individual mutation, we performed immunoblot analyses on the same lysates. Phosphorylation of PKR was increased in cells infected by every virus except VV-βg, but eIF2α phosphorylation levels were very low in lysates of cells infected with VV-A1, VV-A35R*, and VV-A24R* (Fig. 6B), consistent with our previous reports showing that RhTRS1 inhibits PKR activity at a stage after PKR phosphorylation (14, 15). Because we observed a slight decrease in the level of phosphorylated PKR (P-PKR) in cells infected with VV-A1 compared to VV-A35R* or VV-A24R*, we also examined the PKR pathway at earlier time points (Fig. 6C). P-PKR levels were very low at 6 h postinfection and moderate at 12 h after VV-A1 infection. In contrast, P-PKR levels in lysates from VV-A24R*- and VV-A35R*-infected cells were high and comparable to those in VVΔE3L- and VVΔEΔK+RhTRS1-infected cell lysates even by 6 h postinfection. However, we detected very little P-eIF2α in lysates from cells infected with VV-A1, VV-A24R*, or VV-A35R* at any of these time points. Taken together, these data suggest that either A24R* or A35R* is sufficient to inhibit eIF2α phosphorylation but not phosphorylation of PKR in viruses containing RhTRS1. However, the combination of mutations in VV-A1 may act synergistically to delay PKR phosphorylation.

A24R* confers resistance to isatin-β-thiosemicarbazone.

While neither wild-type A24R nor A35R has previously been implicated in PKR evasion, other VACV RNA polymerase mutants have been shown to decrease the abundance of dsRNA and thereby inhibit activation of the 2′-5′-oligoadenylate synthetase (OAS)/RNase L pathway (31, 33, 34, 39). The A24R mutant in one report (VV-IBT^r90 [34]) was isolated by selecting for resistance to the antiviral drug isatin-β-thiosemicarbazone (IBT), a drug that is thought to increase viral RNA polymerase processivity and thereby generate more dsRNA through the annealing of convergent transcripts. To investigate whether our A24R* mutant might act by a mechanism similar to that of the mutant described by Cresawn et al. (34), we infected AGM-derived cells with VV-βg, VVΔEΔK+RhTRS1, VV-A1, VV-A35R*, VV-A24R*, the IBT-resistant virus VV-IBT^r90, or the IBT-dependent virus VV-G2A (MOI = 0.1) in the presence or absence of 45 μM IBT. Consistent with prior reports, IBT inhibited VV-βg replication more than 20,000-fold, while replication of VV-IBT^r90 was not altered by IBT and VV-G2A replicated well only in the presence of IBT (Fig. 7). In permissive BSC40 cells, VVΔEΔK+RhTRS1 and VV-A35* replicated no better than VV-βg in the presence of IBT. In contrast, VV-A1 replicated 60-fold better than VV-βg, and VV-A24R* replicated 24-fold better than VV-βg in the presence of IBT. We observed a similar pattern in PRO1190 cells (Fig. 7B), although the partial IBT resistance was less pronounced. VV-A1 replicated 13-fold better and VV-A24R* replicated 10-fold better than VV-βg in the presence of IBT, while VV-A35R* and VVΔEΔK+RhTRS1 were both susceptible to IBT. These data suggest that the partial IBT resistance phenotype of VV-A1 is due primarily to A24R*.

FIG 7.

A24R*, but not A35R*, increases VVΔEΔK+RhTRS1 resistance to isatin-β-thiosemicarbazone. BSC40 cells (A) or PRO1190 cells (B) were infected with the indicated viruses (MOI = 0.1) in the presence (black bars) or absence (gray bars) of 45 μM isatin-β-thiosemicarbazone. Titers were determined on BSC40 cells 48 h postinfection. Data are presented as means + 95% confidence intervals.

A24R* and A35R* improve virus replication in primary fibroblasts derived from multiple species.

We previously reported that amplification of the rhtrs1 locus provided a substantial replication benefit in cells derived from multiple primate species beyond the AGM cells used for selection. To determine whether A24R* and A35R* also provide a replication benefit in other cell lines from additional primate species, we infected primary human foreskin fibroblasts (HFF) or rhesus fibroblasts (RF) with our panel of viruses (MOI = 0.1). Just as in PRO1190 cells (Fig. 1A), VV-A1 replicated better than VVΔEΔK+RhTRS1 in either HFF or RF, producing titers comparable to those produced by a VV-A infection in both cell types (Fig. 8). In both HFF and RF, VV-A35R* or VV-A24R* each replicated much more efficiently than VVΔEΔK+RhTRS1. VV-A35R* replication increased approximately 30-fold in both cell types, while VV-A24R* replication increased 100-fold and 30-fold in HFF and RF, respectively. However, these individual mutant viruses did not replicate as well as VV-A or VV-A1, especially in RF. Regardless, these data demonstrate that A24R* and A35R* individually provide a replication benefit that functions in diverse primate-derived cell lines.

FIG 8.

A24R or A35R mutations improve VVΔEΔK+RhTRS1 fitness in HFF and RF cells. Titers after infection of HFF cells (black bars) or RF cells (gray bars) with the indicated viruses (MOI = 0.1). Titers were determined on BSC40 cells 48 h postinfection. Data are displayed as means + 95% confidence intervals.

DISCUSSION

Cross-species pathogen transmission is the primary source of new and reemerging infections in both humans and animals. The pool of potential new pathogens is large, so defining biomolecular signatures that may be predictive of agents poised to cross into new species could improve our ability to anticipate and avert epidemics. Gene amplification may be one such marker of imminent adaptation: it is a broad evolutionary mechanism that has been described across all domains of life, including viruses (15, 20–22, 40). In general, gene amplification provides two potential benefits. First, overexpression due to increased gene dosage can provide a fitness benefit through increased expression of the amplified gene. Second, gene amplification increases the number of independently evolving gene copies in the population, providing additional opportunities for the emergence of adaptive point mutations that confer a selective advantage on their own, obviating the need to maintain the amplification.

We have previously shown that gene amplification of rhtrs1, a weak PKR antagonist, expands the viral species tropism of a recombinant VACV expressing rhtrs1 in place of its native PKR antagonists (15). In the current study, we demonstrate that further passage of this rhtrs1-amplified virus identified two adaptive point mutations that accompanied collapse of the amplified rhtrs1 locus. Unexpectedly, these adaptations arose outside the rhtrs1 locus, in two VACV genes, A24R and A35R, that have not previously been associated with the PKR pathway. These data demonstrate that in addition to adaptations evolving in the amplified locus at functionally important amino acid residues (21), adaptations may also evolve in extragenic loci and potentially identify novel host range determinants and virulence factors.

We also identified two virus isolates, VV-B1 and VV-C2, in which the rhtrs1 locus had surprisingly collapsed to a single copy in the absence of any other adaptive mutations. We hypothesize that this collapse was due to the plaque purification of viruses that had undergone stochastic collapse of the locus during genome replication. A similar observation was reported in a bacterial Lac system of experimental evolution, in which collapse of a genetic amplification, without adaptive mutations, occurred in some members of a colony (41, 42). Gene amplification and collapse are a dynamic process, occurring at rates as high as 10⁻² copy gains and losses per generation in bacteria (43, 44). While our experimental design does not permit a detailed calculation of these rates, several aspects of this study suggest that viral gene amplifications are equally dynamic. For example, although we plaque purified each virus three times, all populations contained a range of rhtrs1 copy numbers except in cases in which the locus collapsed to a single copy (Fig. 1A). Likewise, a single round of passage in permissive AGM-derived BSC40 cells, without plaque purification, was sufficient for collapse of the amplified locus in the VV-A population that we previously described (Fig. 2). These data suggest that genomes already containing a gene duplication generate progeny virions with a range of copy numbers at the locus during each round of replication. However, viruses containing a single copy of rhtrs1 are the predominant species in every gene-amplified population, demonstrating that there may also be selective pressure acting to constrain amplification. Taken together, these data suggest that there is a balance between the selective advantage of the rhtrs1 amplification and the fitness cost of maintaining and packaging this amplification, consistent with observations in bacterial systems (44). Presumably, once an adaptive mutation evolves in an organism, this balance will shift to favor collapse of the locus.

The observation of gene amplification after only a few passages in several independent experiments (15, 21, 22) suggests that gene amplification may be a frequent occurrence during VACV replication. Our deep-sequencing data revealed rare instances of a specific J2R-L5R breakpoint in PRO1190-adapted viruses that have only a single copy of rhtrs1 (Fig. 1 and Table 1) in the vast majority of genomes, similar to some reported duplication frequencies in bacterial systems (44–46). We also observed a similar rate of duplication at this specific breakpoint in VV-βg, which does not have rhtrs1 at all and in which amplification at this locus would be unlikely to provide a replication benefit. Duplications are expected to be generally neutral or detrimental (44). Therefore, these observations suggest that a steady-state frequency of gene duplication in poxviruses may reflect a balance between de novo formation of duplications and their subsequent loss through collapse or competition. Rare viruses in the population with these preexisting duplications would thus be primed for rapid adaptation in situations where additional gene copies would provide a selective advantage, such as replication in the presence of more-resistant host restriction factors.

We identified mutations in two VACV genes, A24R and A35R, accompanying contraction of the rhtrs1 locus in adapted viruses. Because these adaptive mutations evolved at so far a distance from the rhtrs1 locus and because the rhtrs1 amplification is so unstable, we are unable to determine whether these second-site mutations arose in genomes that contained rhtrs1 amplifications or if they arose independently in the populations. Regardless, neither A24R nor A35R has been previously associated with the PKR pathway. A24R encodes the catalytic subunit of the viral RNA polymerase, and in this study we identified a C-terminal mutation that is sufficient to inhibit PKR in cell lines derived from a variety of primate species and to provide partial resistance to IBT in a VVΔEΔK+RhTRS1 background. Previously, Cresawn et al. identified a different mutation in the VACV RNA polymerase that confers IBT resistance by decreasing processivity of the holoenzyme (34). This mutation also conferred RNase L resistance, presumably because the decreased transcription processivity reduced the overlap of convergent transcripts and consequently reduced the production of dsRNA that would activate oligoadenylate synthetases and thus RNase L at late stages of VACV replication. While the C-terminal mutant that we identified in this study does not confer the same level of IBT resistance as those described by Richard Condit's laboratory, it does reduce sensitivity to IBT.

Of note, two very recent reports demonstrated that inhibiting the degradation of dsRNA by engineered mutations in VACV decapping genes and reducing expression of the exoribonuclease Xrn1 resulted in accumulation of dsRNA, activation of the PKR and OAS/RNase L pathways, and reduction of viral replication and virulence (47, 48). Together with our results, these studies suggest that the rate of production and destruction of RNAs might be finely tuned to enable sufficient viral gene expression to support replication but not so much as to activate host antiviral dsRNA sensors. However, VV-A1- and VV-A24R*-infected cells still phosphorylate PKR, which might not be expected if A24R* did in fact inhibit dsRNA production, suggesting that A24R*-mediated evasion of PKR may be more complex than simple downregulation of dsRNA production. Further work will be necessary to determine whether A24R* polymerase processivity may be intermediate between the wild-type polymerase and those described by the Condit laboratory, or if A24R* evades the PKR pathway in some way not directly mediated through dsRNA production. Regardless of the exact mechanism, our data demonstrating that A24R* improves VVΔEΔK+RhTRS1 replication and enables PKR inhibition in a variety of primate species provide compelling evidence that A24R mutations mediate a previously unrecognized mechanism for poxviral evasion of PKR. Furthermore, the evolution of A24R mutations under a variety of selective pressures, such as IBT or PKR, suggests that mutations in this gene may be a general mechanism for adapting to other dsRNA sensors in a variety of host species (33, 34).

A35R is an in vivo virulence gene that modulates the adaptive immune response through the regulation of major histocompatibility complex class II (MHC-II) and cytokine expression (35–37), yet it has not previously been implicated in modulation of the innate immune system. In this study, we demonstrate that a single-nucleotide deletion in A35R confers the ability to inhibit PKR from multiple primates in a VVΔEΔK+RhTRS1 background. However, A35R* does not confer IBT resistance, suggesting that this mutation works through a different mechanism from that of A24R*. In this study, we describe the effect of the single-nucleotide deletion at nucleotide position 211, although in total we identified five single-nucleotide deletions (Table 2) (15) that evolved during passage of VVΔEΔK+RhTRS1 in PRO1190 cells, including the deletion at position 261 that was fixed in VV-A3, a virus that also replicates well with only a single copy of rhtrs1. It remains to be determined whether these other A35R deletion mutants also have improved activity against PKR. The high frequency of mutations in A35R is striking given that this gene is highly conserved among all known orthopoxviruses, with the notable exception of variola virus (VARV). VARV, which replicates only in humans, encodes a truncated form of A35R predicted to express a 60-amino-acid protein (GenBank accession no. DQ437580), a gene product somewhat shorter than the 92-amino-acid gene product predicted to be expressed by VV-A1 and VV-A3. Since A35R* improves the replication of VVΔEΔK+RhTRS1 in RF and PRO1190, as well as HFF, it seems unlikely that the truncation of A35R in VARV contributes to the restriction of this virus to humans. Regardless, our observations suggest that A35R may be a multifunctional poxviral host range determinant, modulating both the innate and adaptive host immune responses.

Cross-species pathogen transmission is an important source of new diseases in both humans and animals. In this study, we demonstrate the evolution of two adaptive point mutations correlated with the collapse of an amplified rhtrs1 locus during passage under continuous selective pressure. This study presents the first experimental evidence of extragenic adaptive mutations correlated with collapse of a gene amplification in viruses. These adaptive mutations were also present at high frequencies in viruses with rhtrs1 amplification (Table 2) (15). Due to the dynamic nature of the amplification as well as the distance between the amplified locus and the point mutations, we have been unable to determine whether these different adaptations arose consecutively or independently. Although neither A24R nor A35R has previously been associated with the PKR pathway, either one of these mutations fully rescued VVΔEΔK+RhTRS1 replication in AGM-derived cells and also provided a moderate replication benefit in human- and rhesus-derived cells. Thus, this study identifies two new genes that mediate evasion of PKR pathway activation in cells from a variety of primate species. Taken together, these studies suggest a potential model for viral adaptation to new hosts in which gene amplification occurs early and broadly increases viral species tropism. As the virus replicates in new hosts, adaptive mutations can arise not just in the amplified gene but also elsewhere in the genome and can be sufficient to permit collapse of the amplified locus while maintaining the foothold needed for continued replication and spread in the new host.